Transcriptional Activation of Gstp1 by MEK/ERK Signaling Confers Chemo-Resistance to Cisplatin in Lung Cancer Stem Cells

Lung cancer management remains a challenge due to its asymptomatic and late presentation when it is metastatic. The clinical response to the first-line platinum-based chemotherapy in patients with advanced lung cancer is disappointing due to the development of chemoresistance. Chemoresistance is a complex phenomenon. Mechanistic research using experimental models has yielded limited clinical results to help increase understanding for overcoming resistance. While the role of lung CSCs in conferring multidrug resistance has been postulated, experimental evidence remains associative and lacks in depth mechanistic inquisition. In the present study, using mouse and human lung adenocarcinoma cell lines and their respective paired CSC derivative cell lines that we generated, we identified cancer stem cell component of lung adenocarcinoma as the source that confers multidrug resistance phenotype. Mechanistically, Gstp1 confers cisplatin resistance in mouse and human lung CSC models, both in vitro and in vivo. Further, transcriptional activation of Gstp1 expression by MEK/ERK signaling underlies cisplatin resistance in lung CSC cells. Moreover, we show that GSTP1 expression is a poor diagnostic and prognostic marker for human lung adenocarcinoma, thus is of high clinical relevance. Taken together, we have provided mechanistic understanding of the lung CSC in mediating chemoresistance.


HIGHLIGHTS
• The cancer stem cell component of lung adenocarcinoma is the source that confers multidrug resistance phenotype. • Gstp1 expression is elevated in lung CSC cells which can be further increased upon treatment with a panel of chemotherapy drugs. • Gstp1 confers cisplatin resistance in mouse and human lung CSC models, both in vitro and in vivo. • Providing the first experimental evidence that transcriptional activation of Gstp1 expression by MEK/ERK signaling underlies cisplatin resistance in lung CSC cells. • GSTP1 expression is a poor diagnostic and prognostic marker for human lung adenocarcinoma thus is of high clinical relevance.

INTRODUCTION
Lung cancer is the most common cause of cancer-related deaths in the world (1). The high mortality rate (51-99%) of lung adenocarcinoma is due to it being asymptomatic, it having late presentation when it is metastatic and becoming resistant to anti-cancer therapies (2). In spite of the development of new therapeutic strategies, the outcome of patients with metastatic lung cancer has barely improved over the past few decades, and the overall 5-year survival rate remains very low (10-15%) (3,4). Lung adenocarcinoma is the most common histological type of lung cancer, comprising ∼60% of non-small cell lung cancers (NSCLC) (5). Although platinum-based chemotherapy represents the standard first-line treatment for patients with advanced NSCLC, therapeutic outcome is disappointing due to the development of chemo-resistance, relapse, and distant metastases (6,7). Mechanistic understanding of the involvement of commonly studied multidrug resistant genes using human lung adenocarcinoma cell lines has yielded limited clinical success in overcoming chemo-resistance thus far. According to the CSCs theory, tumorigenesis, and cancer progression are due to a subset of phenotypically distinct cells characterized by unlimited self-renewal and enhanced clonogenic potential (8)(9)(10). Lung CSCs are shown to be associated with higher recurrence rates (11,12). In agreement with this hypothesis, lung cancers that manifest stem cell signatures are associated with multidrug resistance (including cisplatin) and with disease relapse (12)(13)(14). However, in depth characterization and mechanistic investigation of multidrug resistance in lung CSCs were lacking, partially due to the lack of stable cellular models of lung CSC.
Glutathione S-transferases (GSTs) are phase II detoxifying enzymes involved in the maintenance of cell integrity, oxidative stress and protection against DNA damage by catalyzing the conjugation of glutathione to a wide variety of electrophilic substrates (15)(16)(17). Gstp1, a member of the GST family, is directly involved in the detoxification of cisplatin via the formation of cisplatin-glutathione adducts, which indicates that Gstp1 may play a role in the acquisition of resistance to this platinum compound (18,19). Even though a growing number of studies have demonstrated that Gstp1 plays a key role in the development and maintenance of malignancy in several tumor types (20)(21)(22), mechanistic understanding of Gstp1 in mediating chemoresistance in lung cancer is sketchy. Its role in mediating chemoresistance in CSCs is unknown.
The MAPK pathway, including MEK/ERK, JNK, and p38 kinase, plays a pivotal role in cell survival, proliferation and migration of tumor cells (23)(24)(25). While several studies reported activation of the MEK/ERK cascade in response to cisplatin treatment in several forms of cancer, the consequence of such activation on cell survival remains controversial (26)(27)(28)(29)(30)(31)(32). Few studies reported the activation of GST gene expression by MEK/ERK signaling in breast cancer (33)(34)(35). Up until the Abbreviations: CSCs, cancer stem cells; LCSCs, lung cancer stem cells, NSCLC, non-small cell lung cancers; LUAD, lung adenocarcinoma. present study, regulation of Gstp1 expression in lung CSCs has not been examined.
In the present study, we employed the lung CSCs derived from mouse parental Lewis lung carcinoma cell line (LLC-Parental) and human cancer cell lines H1299, which were named LLC-SD and H1299-SD, respectively. The stem cell properties of LLC-SD in vitro and in vivo had been characterized (36)(37)(38). Using the stable mouse and human lung CSC models that we generated and characterized, we clearly revealed lung CSCs as the cellular component that confer multidrug chemoresistance both in vitro and in vivo. We also demonstrate that the underlying mechanism of cisplatin chemo-resistance is the transcriptional activation of Gstp1 by MEK/ERK signaling. We also show the clinical relevance of these findings and establish GSTP1 as a poor diagnostic and prognostic marker, as well as a therapeutic target for human lung adenocarcinoma. Since the expression changes of Gstp1 in response to chemo-therapeutic agents were specific to the lung CSCs, not the non-stem-like parental cells, Gstp1 may also serve a lung CSC marker for evaluating the CSC content in clinical lung samples.

Cell Culture and Cell Lines
LLC-Parental and H1299-Parental cells were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China) and cultured in DMEM high glucose supplemented (Hyclone) containing 10% FBS (Gibco). LLC-SD and H1299-SD cells were stem counterpart cells, which were maintained in DMEM/F12based normal stem cell media (Hyclone), supplemented with 20 ng/ml EGF (BD), 20ng/ml FGF (BD) and 2% B27 (Gibco). All cell lines were cultured at 37 • C, in a 5% CO 2 atmosphere with 95% humidity.

Reverse Transcription Quantitative Real-Time Polymerase Chain Reaction (RT-qPCR)
Total RNA was isolated using TRIZOL (Takara) according to the manufacturer's instructions. cDNA was reverse transcribed from 2 µg of total RNA. 39 cycles of PCR amplification were performed using 95 • C for 30s, 95 • C for 5s, and 60 • C for 30s for each cycle. TBP was used as a loading control. The sequences of PCR primers are listed in Table 1.

Flow Cytometric Apoptosis Assay
At 24 h following transfection or inhibitor treatment, the indicated cells were treated with cisplatin at a final concentration of 10 µM. After 24 h of cisplatin treatment, the floating and attached cells were collected separately and then combined. 0.5-1 million cells of each single-cell suspension was stained with the Annexin V-fluorescein isothiocyanate (FITC) and propidium iodide (PI), and analyzed with a BD LSR. Unstained cells were included as negative controls for each FACS analysis. The data were analyzed using BD FACS software.

Western Blot
Western blot was performed according to a standard method. The total protein was extracted with RIPA buffer (Beyotime), according to the manufacturer's instructions. The concentration was determined by the BCA method. Twenty-five micrograms of protein was used to run on a 10% polyacrylamide (Beyotime) gel and transferred to a PVDF membrane underneath. The membrane was blocked in 5% non-fat milk (Bio-Rad). Primary antibodies and secondary antibodies (Proteintech) were used according to manufacturer's protocol. ECL method was used and the blot was developed under the gel electrophoresis imager (Bio-Rad). The following primary antibodies were used: anti-Gstp1 (Abcam), anti-Mek1 (upstate), anti-p-Mek1/2 (Cell Signaling Technology), and anti-GAPDH (Proteintech).

Lentivirus Packaging and Infection
Stable knockdown or overexpression of Gstp1 in LLC-SD cells was achieved with infection of LLC-SD cells with either pLL3.7-shN.C.(mouse) or pLL3.7-shGstp1(mouse) and LV4-Vector(mouse) or LV4-OEGstp1(mouse), respectively, in the presence of 8 µg/ml polybrene (Genepharma, China) for 12 h, and the medium was refreshed. Seventy-two hours after the infection, the efficiency of infection was measured under a fluorescent microscope and by RT-qPCR.

Clinical Samples
We collected 43 tumor samples from The Affiliated Hospital of Southwest Medical University after informed consent approved by Southwest Medical University of Institutional Review Boards. Specimens were prepared from paraffin embedded (FFPE) tissue. Genomic Total RNA from FFPE tissues was extracted using the Tiangen RNAprep Pure FFPE Kit (Tiangen, China). RNA yield was evaluated using a Nanodrop 2000 spectrometer (NanoDrop Technologies Inc., DE) and Qubit 3.0 Fluorometer (Thermo Fisher Scientific, MA). All of the formalin fixed paraffin embedded was obtained from the department of pathology in the affiliated hospital of southwest medical university, which was used to identify the cancer progress as the stage of lung cancer and received the approval of the institutional ethics committee of southwest medical university medical center. Moreover, this study obtained written informed consent from the participants.

Cisplatin Resistance Assay
Single cell suspension (1 × 10 3 ) mixed with 50 µL Matrigel Matrix (Corning) at a 1:1 ratio was injected subcutaneously into both insides of the hind legs of each 6-8 weeks old female BALB/c nude mice (Beijing Huafukang bioscience company, Beijing, China). Tumor growth was measured every 2 days, and tumor volume was calculated as V = (length × width 2 )/2. Treatment with cisplatin started when the tumor volume of the subcutaneous xenograft tumors reached about 100 mm 3 . 16 mg/kg cisplatin was administrated every other day intraperitoneally (qilu-pharmaceutical). After 9 days of treatment, when mouse body weights were decreased by 20%, mice were killed in a humane manner. Experimental mice were raised in an accredited specific pathogen-free Animal Facility at Chongqing medical University. All protocols were approved by the Institutional Animal Care and Use Committee of Chongqing medical University and conducted with humane animal care.

Bioinformatics Analyses
The lung cancer dataset, which included 535 lung adenocarcinoma tissues and 59 normal lung tissues, was obtained from The Cancer Genome Atlas project (TCGA, https://portal.gdc.cancer.gov/). Student's t-test was then used to compare GSTP1 expression levels between lung cancer and normal lung tissues. Gene set enrichment analysis (GSEA) was performed to identify pathways associated with GSTP1 mRNA expression levels in the TCGA lung cancer dataset. GSEA software was obtained from the Broad Institute (http://www. broad.mit.edu/gsea).The expression levels of GSTP1 genes in the selected cancers were analyzed using Oncomine. For this, we compared clinical specimens of cancer vs. normal patient datasets. In order to reduce our false discovery rate, we selected p < 0.01 and gene rank, with the top 10% as a threshold. We analyzed the results for their p-values, fold change, and cancer subtype. In many instances we found several significant correlations in different tumor types, but we showed only lung adenocarcinoma except for sub-forms. TCGA survival analysis used the website http://www.oncolnc.org/. Additionally, Kaplan-Meier Plotter database was used to draw the overall survival (OS), progression-free survival (FS) and post-progression survival (PPS) curves.

Statistical Analysis
Data were analyzed by one-way analysis of variance and Student's independent t-test using GraphPad Prism software and was presented as mean ± SD or SEM indicated in figure legends. Differences were considered statistically significant when P < 0.05.

RESULTS
The The lung CSCs LLC-SD cells were derived from the LLC-Parental cell line and demonstrated high capability of self-renewal and enhanced tumorigenicity as we characterized and reported (36)(37)(38). LLC-SD cells exhibited spheroid growth in vitro and can be stably maintained in culture ( Figure 1A). The multi-drug resistant phenotype of LLC-SD was manifested by their resistance to a wide range of structurally and functionally unrelated chemotherapy drugs. LLC-SD cells exhibited more than 3folds more resistance to cisplatin, paclitaxel and 5-fluorouracil compared with that in LLC-Parental cells ( Figure 1B). LLC-SD cells were more resistant to undergo apoptosis upon all drug treatments than the paired LLC-Parental cells ( Figure 1B). LLC-SD exhibited 12.8-, 30.0-, and 50.6-fold resistance to cisplatin, paclitaxel and 5-fluorouracil, respectively, as determined by IC 50 ( Table 2). To discern the potential genes that mediate multidrug resistance in the LLC-SD cells, we examined the expression of known multidrug resistance genes, including Gstp1, Abca2, Mrp1, Abcg2 and Mdr1, by RT-qPCR. Gstp1, Abca2, Mrp1 expression was robustly upregulated in LLC-SD, compared with that in LLC-Parental ( Figure 1C).
Cisplatin is one of a number of platinum (Pt)-containing compounds used for the treatment of a variety of malignancies, including ovarian and testicular cancer, and remains the first-line of chemotherapy for patients with advanced NSCLC (39). Hence, we further determined the expression of Gstp1, Abca2, Mrp1 before and after cisplatin treatment in LLC-SD and LLC-Parental cells. While the expression of all three genes in LLC-Parental cells was inhibited upon cisplatin treatment, Gstp1 and Mrp1 expression was elevated in LLC-SD cells after cisplatin treatment ( Figure 1D) and the most pronounced increase was seen in Gstp1 (Supplementary Figure 3). Thereafter, we chose to focus on the Gstp1 gene and cisplatin for mechanistic investigation as GSTs are known to form adducts with platinum compound (18). Treatment of LLC-SD cells with an Gstp1 inhibitor, Ezatiostat, reversed the resistance to cisplatin at all doses evaluated, while Ezatiostat had less of an effect on LLC-Parental cells determined by Cck-8 assay (Figure 1E). Flow cytometric analysis confirmed these findings (Figure 1F). Taken together, these observations demonstrate that it is the CSC component of the LLC-Parental cells, i.e., the LLC-SD cells, that confer resistance to cisplatin, which appears to be mediated by Gstp1.

Blocking Gstp1 Reversed Cisplatin Resistance Both in vitro and in vivo
To confirm the involvement of Gstp1 in mediating chemoresistance of lung CSCs, we deregulated Gstp1 expression by generating LLC-SD cell lines: (i) in which Gstp1 was constitutively inhibited by lentiviral-mediated Gstp1 shRNA (shGstp1) silencing, and (ii) in which Gstp1 was up-regulated by lentiviral-mediated Gstp1 (OEGstp1) overexpression (Figures 2A,B). Chemosensitivity of LLC-SD to cisplatin varied with changes in Gstp1 expression. Inhibition of Gstp1 by shGstp1 resulted in a left-shifted cell survival curve indicating sensitization (Figure 2C, left). In contrast, Gstp1 overexpression led to enhanced resistance to cisplatin, as evident by the right-shifted cell survival curve (Figure 2C, right). The Cck-8 assay results were confirmed by the apoptosis assay ( Figure 2D).
To determine whether Gstp1 mediates cisplatin resistance in vivo, four cell lines that differ in Gstp1 expression were injected subcutaneously into the nude mice (Methods, 1000 cells/side). After 11 days of injection, when subcutaneous xenograft tumors growth reached ∼100 mm 3 , cisplatin treatment was administrated (Methods). The experiment was terminated on Day 9 since the start of the treatment after 4 treatments wereadministered. Consistent with our in vitro finding, Gstp1 regulates LLC-SD tumor sensitivity to cisplatin, measured by visual examination of tumor size (Figure 2Ei), tumor growth rate (Figure 2Eii) and tumor weight (Figure 2Eiii).
Taken together, Gstp1 regulates LLC-SD sensitivity to cisplatin both in vitro and in vivo (Figures 1, 2).

MEK/ERK Signaling Pathway Mediates Cisplatin Resistance in LLC-SD Cells
It is reported that activation of the mitogen activated protein kinase (MEK/ERK) pathway may mediate resistance to cisplatin in ovarian cancer cells by regulating GST-π expression (40). Gene set enrichment analysis (GSEA) was performed to evaluate signaling pathways that were associated with Gstp1 expression in the TCGA lung cancer samples. The results revealed that Gstp1 expression was positively correlated with the MEK/ERK signaling pathway (Supplementary Figure 1A), which implied that Gstp1 may be affected by MEK/ERK signaling pathway in lung cancer. To investigate the relationship between MEK/ERK signaling pathway and Gstp1, we first evaluated whether Gstp1 expression is subjected to MEK/ERK signaling pathway regulation. Treatment of LLC-SD cells with MEK/ERK inhibitor PD98059, at the same time of receiving cisplatin treatment, prevented the up-regulation of Gstp1 expression (Figure 3A). In LLC-SD cells that exhibit resistance to cisplatin both in vitro and in vivo (Figures 1, 2), LLC-SD proliferative activity was sensitive to the inhibition of the MEK/ERK signaling pathway by PD98059 (Figure 3Bii). Further, PD98059 treatment significantly reduced LLC-SD resistance to cisplatin (Figure 3Bii). In contrast, PD98059 treatment did not alter cisplatin sensitivity in LLC-P cells (Figure 3Bi) that were sensitive to cisplatin's growth inhibition (Figure 1B). In agreement with the Cck8 analyses, PD98059 augmented cisplatin-induced apoptosis in LLC-SD cells (Figure 3Cii), and it restored the ability of cisplatin to induce apoptosis in LLC-SD to a level that is comparable to cisplatin-induced apoptosis in LLC-P cells (Figure 3Ci). Based on this set of observations, we hypothesized that cisplatin-induced increase in Gstp1 expression (Figure 1D), that was subjected to PD98059 inhibition (Figure 3A), might be a result of cisplatin-induced activation of the MEK/ERK signaling pathway leading to transcriptional activation of Gstp1 expression. Consistent with this hypothesis, enhancement of Mek1 phosphorylation was only seen in cisplatin-resistant LLC-SD cells and not in the LLC-Parental cells that were sensitive to cisplatin (Figure 3D).
To further investigate the role of the MEK/ERK signaling pathway in mediating cisplatin resistance in LLC-SD cells, we generated LLC-SD derivative cell lines in which Mek1/2 expression was stably inhibited by RNA interference (shMek1/2) ( Figure 4A). Mek1/2 silencing prevented cisplatin-induced increase in Gspt1 expression in LLC-SD cells ( Figure 4B). Consequently, cisplatin resistance was reduced, as determined by the Cck8 assay ( Figure 4C) and apoptosis assay (Figure 4D) in vitro and by tumor growth and tumor weight in vivo (Figures 4Ei-iii), upon inactivation of the MEK/ERK pathway by Mek1/2-knockdown. This set of observations are consistent with PD98059 treatment, confirming the role of MEK/ERK signaling in mediating resistance to cisplatin in LLC-SD lung CSC cells.

Gstp1 Confers MEK/ERK Mediation of Cisplatin Chemoresistance in LLC-SD Cells
To order the molecular relationship between activation of MEK/ERK signaling and transcriptional activation of Gstp1 in mediating cisplatin chemoresistance, Gstp1 expression was restored in Mek1/2 inhibited LLC-SD-shMek1/2 cells by lentiviral-mediated overexpression of Gstp1 (Figures 5A,B). Restoration of Gstp1 expression in LLC-SD-shMek1/2 cells reversed the cisplatin sensitive phenotype seen in LLC-SD-shMek1/2 cells to cisplatin resistance seen in shMek1/2+OEGstp1 cells, measured by Cck-8 proliferation assay ( Figure 5C) and the apoptosis assay ( Figure 5D) in vitro, and the tumor transplantation assay in vivo ( Figure 5E). These results indicate that Gstp1 acts downstream of MEK/ERK signaling pathway. Further, MEK/ERK signaling mediates cisplatin resistance in lung CSC LLC-SD cells by transcriptional activation of Gstp1.

High Gstp1 mRNA Levels Are Associated With Clinical Accelerated Disease Progression and Poorer Survival in Lung Adenocarcinoma
To verify whether the findings we made in LLC-SD mouse model of lung adenocarcinoma are applicable to lung CSC models derived from human adenocarcinoma, the human lung cancer cell line H1299 was used to generate H1299-SD cells (Figure 6Ai) using the same approach that we developed (36)(37)(38). Cisplatin resistance phenotypes observed in LLC-SD cells were similarly manifested in H1299-SD cells (Figures 6Aii,iii). GSTP1 expression was also increased in H1299-SD compared with that in H1299 cells (Figure 6Aiv). More importantly, inhibition of GSTP1 restored cisplatin sensitivity in H1299-SD cells in vitro (Figures 6Bi-iii). Further, the promoting effect of Frontiers in Oncology | www.frontiersin.org GSTP1 on cisplatin resistance is regulated by MEK/ERK signaling ( Figure 6C). This set of observations confirmed our main findings in the LLC-SD model and both collectively demonstrate that "lung CSC component is the cause of chemoresistance to cisplatin, which is conferred by MEK/ERK transcriptional activation of GSTP1." To assess the clinical relevance of GSTP1 in lung adenocarcinoma, lung adenocarcinoma mRNA sequencing datasets from The Cancer Genome Atlas (TCGA) and ONCOMINE database were analyzed. The differential GSTP1 expression between the cancer tissue and the normal lung was drastic and the elevation in lung adenocarcinoma tissues was profound (Figures 6Di,ii, Supplementary Figure 1B). Moreover, we collected 43 clinical paraffin samples of lung adenocarcinoma, including 22 cases of early adenocarcinoma (T1 and T2) and 21 cases of advanced adenocarcinoma (TIII and TIV). We analyzed GSTP1 mRNA expression in this clinical cohort and found higher GSTP1 expression in advanced lung adenocarcinoma compared to that in early stage lung adenocarcinoma ( Figure 6E). Thus, GSTP1 expression appears to correlate with disease staging. The low GSTP1 expression in the non-cancerous lung tissue and staging-wise elevation could be explored for lung cancer diagnosis and merits further investigation.
We next analyzed the prognostic value of GSTP1 expression by examining the relationship between GSTP1 expression and lung adenocarcinoma progression using the publicly accessible TCGA (analyzed in Oncolnc website) and Kaplan-Meier Plotter (analyzed in Kaplan-Meier Plotter website) databases. High GSTP1 mRNA levels are correlated with faster disease progression and a lower rate of survival in lung adenocarcinoma (Figure 6F,  Supplementary Figure 1C). Thus, high GSTP1 expression predicts poor prognosis.
Taken together, the bioinformatics analyses indicate that the findings we made using the mouse and human lung carcinoma CSC cellular models, regarding GSTP1 mediation of cisplatin resistance, are of clinical relevance and importance.

DISCUSSION
Lung cancer management remains a challenge due to it being asymptomatic, having late presentation when it is metastatic and becoming resistant to anti-cancer therapies. Although platinumbased chemotherapy represents the standard first-line treatment for patients with advanced NSCLC, therapeutic outcome is compromised by the development of chemoresistance, relapse and disease recurrence (6,7). Chemoresistance is a complex phenomenon, which may involve but is not limited to the reduction of drug accumulation, enhancement of DNA repair, impediment to apoptosis and alterations in cell cycle. Mechanistic research of the involvement of commonly studied multidrug resistant genes using human lung adenocarcinoma cell lines has yielded limited clinical results to help increase the understanding for overcoming resistance. While the role of lung CSCs in conferring multidrug resistance has been postulated, experimental evidence remains associative and lacks in depth mechanistic inquisition (12)(13)(14), partially due to the lack of stable cellular models of lung CSC.
The present study has made the following novel findings that contribute to an improved mechanistic understanding of clinically observed chemoresistance to platinum-based chemotherapies in advanced lung cancer: First, lung CSCs are the source of chemoresistance. We developed and characterized the paired mouse Lewis lung adenocarcinoma cell line, LLC-Parental, and its CSC derivative cell line, LLC-SD (36)(37)(38), as well as paired human lung adenocarcinoma cell line, H1299-Parental, and its CSC derivative cell line, H1299-SD, that we generated using the same approach for this study (Figure 6) to overcome the obstacle of lacking stable lung CSC cellular models for mechanistic investigation. The sensitivity of both the LLC-Parental and H1299-Parental to chemotherapies, in deer contrast of the multidrug resistant phenotypes observed in the LLC-SD and H1299-SD CSCs, helped us to clearly identify the CSC component of both the mouse and human lung adenocarcinoma cell lines as the cellular component to confer multidrug resistance phenotype, in particular to cisplatin (Figures 1, 6). Thus, we have provided concrete experimental evidence demonstrating the involvement of lung CSCs in chemoresistance and have overcome the obstacle of the lack of stable cellular models of CSCs, including lung CSCs. Second, expression of Gstp1 is elevated in lung CSC cells, which can be further increased upon treatment with a panel of chemotherapy drugs. Using the two sets of cellular models for the screening of drug resistant genes that mediate multidrug resistant phenotype in the mouse and human lung CSC cells, we identified Gstp1 among other candidates that was consistently expressed at higher levels in LLC-SD and H1299-SD cells and could be further increased upon treatment with a panel of chemotherapy drugs, in particular with cisplatin (Figures 1, 6).
Third, Gstp1 confers cisplatin resistance in mouse and human lung CSC models, both in vitro and in vivo. Using lentiviralmediated gene transfer, we generated LLC-SD (Figure 2) and H1299-SD (Figure 6) derivative cell lines, in which Gstp1 expression was either inhibited or up-regulated. Through the bidirectional modulation of Gstp1 expression and in vitro (both LLC-SD and H1299-SD) (Figures 2, 6) and in vivo (LLC-SD) (Figure 2) characterization of the sensitivity to cisplatin, we can make a rather convincing conclusion that high Gstp1 expression confers cisplatin resistance in experimental lung CSCs.
In the present study, we found elevated levels of Gstp1 in response to multiple-drug treatment (Figure 1,  Supplementary Figure 2), although Gstp1 function is mostly studied under the context of platinum category of chemotherapeutic agents, due to its ability to remove the agents from the cells via the formation of Pt-glutathione. However, our findings are consistent with the literature reports that Gstp1 expression could be regulated by transcriptional factors, including activating protein-1 (AP-1) and nuclear factor erythroid-2-related factor 2 (Nrf2), in response to multiple anticancer drugs, including chlorambucil, cyclophosphamide, cisplatin, DOX and mitoxantrone (41)(42)(43). Thus, Gstp1mediated chemoresistance may not be limited to cisplatin and related anticancer agents, thus merits further studies.
Fourth, we have provided the first experimental evidence that transcriptional activation of Gstp1 expression by MEK/ERK signaling underlies cisplatin resistance in lung CSC cells. Prior to our study, it was reported that in addition to MEK/ERK, both JNK/SAPK and p38 pathways were also affected by cisplatin treatment (44). In addition, Lin et al. reported that Gstp1 expression can be upregulated by phosphorylation of Erk2 and by activating Nrf2 in response to methionine restriction (45). In the present study, we focused on examining MEK/ERK transcriptional regulation of Gstp1 expression in response to cisplatin treatment in lung CSCs. Through lentiviral mediated siRNA interference of Mek1/2 expression, we confirmed the requirement of MEK/ERK activation to confer cisplatin resistance both in vitro and in vivo (Figure 3). We didn't stop here and took one step further which was to order the molecular relationship between activation of MEK/ERK signaling and transcriptional activation of Gstp1 in mediating cisplatin chemoresistance. It was achieved by restoration of Gstp1 expression in Mek1/2 silenced LLC-SD-shMek1/2 cells, which manifested concomitant Gstp1 down-regulation, by overexpressing of Gstp1 (Figures 5A,B). Restoration of Gstp1 expression in LLC-SD-shMek1/2 cells reversed the cisplatin sensitive phenotype seen in LLC-SD-shMek1/2 cells to cisplatin resistance seen in shMek1/2+OEGstp1 cells, both in vitro and in vivo (Figure 5). These results confirm that transcriptional activation of Gstp1 by MEK/ERK signaling pathway mediates cisplatin resistance in experimental lung CSCs.
Fifth, GSTP1 expression is a poor diagnostic and prognostic marker for human lung adenocarcinoma, thus is of high clinical relevance. Prior to this study, the clinical relevance of GSTP1 has not been well-explored. We conducted multiscale analyses including the analysis of publicly available gene expression databases of human adenocarcinoma of lung, gene expression analysis of clinical paraffin blocks of lung adenocarcinoma as well as mouse and human lung adenocarcinoma CSC models. We found that GSTP1 expression appears to correlate with disease staging (Figure 6). The low GSTP1 expression in the non-cancerous lung tissue and staging-wise elevation could be explored for lung cancer diagnosis and merits further investigation. We also found that high GSTP1 mRNA levels are correlated with faster disease progression and a lower rate of survival in lung adenocarcinoma (Figure 6F, Supplementary Figure 1C). Thus, high GSTP1 expression predicts poor prognosis. Further, since the expression changes of GSTP1 in response to chemotherapeutic agents were specific to the lung CSCs, not the non-stem-like parental cells, GSTP1 may also serve a lung CSC marker for evaluating the CSC content in clinical lung samples.
Taken together, the bioinformatics analyses indicate that the findings we made using the mouse and human lung carcinoma CSC cellular models, regarding Gstp1 mediation of cisplatin resistance, are of clinical relevance and importance.

ETHICS STATEMENT
This study was carried out in accordance with the recommendations of Institutional Animal Care and Use Committee of Chongqing medical University. The protocol was approved by Institutional Animal Care and Use Committee of Chongqing medical University. Experimental mice were raised in an accredited specific pathogen-free Animal Facility at Chongqing medical University. All protocols were approved by the Institutional Animal Care and Use Committee of Chongqing medical University and conducted with humane animal care. All of the formalin fixed paraffin embedded were obtained from department of pathology in the affiliated hospital of southwest medical university and received the approval of the institutional ethics committee of southwest medical university medical center. Moreover, this study obtained written informed consent from the participants.

AUTHOR CONTRIBUTIONS
JL and TY substantially contributed to conception and design, acquisition of data, analysis and interpretation of data, and drafting the article. YL, LK, ZS, and DL acquired part of the data. JW and HX contributed to conception and design, revising it critically for important intellectual content, and final approval of the version to be published.