GPX2 promotes EMT and metastasis in non‐small cell lung cancer by activating PI3K/AKT/mTOR/Snail signaling axis

Abstract Lung cancer, with non‐small cell lung cancer (NSCLC) being the main subtype, is the leading cause of cancer death worldwide, which is mainly due to the cancer metastasis. Glutathione peroxidase 2 (GPX2), an antioxidant enzyme, is involved in tumor progression and metastasis. Nevertheless, the role of GPX2 in NSCLC metastasis has not been clarified. In this study, we found that GPX2 expression was elevated in NSCLC tissues and high GPX2 expression was correlated with poor prognosis in patients with NSCLC. In addtion, GPX2 expression was related to the patient's clinicopathological features, including lymph node metastasis, tumor size, and TNM stage. Overexpression of GPX2 promoted epithelial–mesenchymal transition (EMT), migration, and invasion of NSCLC cells in vitro. Knockdown of GPX2 showed the opposite effects in vitro and inhibited the metastasis of NSCLC cells in nude mice. Furthermore, GPX2 reduced reactive oxygen species (ROS) accumulation and activated the PI3K/AKT/mTOR/Snail signaling axis. Therefore, our results indicate that GPX2 promotes EMT and metastasis of NSCLC cells by activating the PI3K/AKT/mTOR/Snail signaling axis via the removal of ROS. GPX2 may be an effective diagnostic and prognostic biomarker for NSCLC.


| INTRODUCTION
Lung cancer is one of the most frequent cancer types and is the leading cause of cancer death worldwide in 2020. 1 Around 18% of global cancer deaths are caused by lung cancer, which is partly due to the fact that most patients are diagnosed with metastases. 1,2 The 5-year survival rate of lung cancer is only 10%-20% in most countries. 3 Histologically, lung cancer is divided into small cell lung cancer and non-small cell lung cancer (NSCLC), of which the latter accounts for about 85%. 4 Thus, it is particularly urgent to identify novel diagnostic and prognostic markers for NSCLC. Meanwhile, the molecular mechanisms of NSCLC progression and metastasis also need to be explored to provide effective therapeutic targets for patients with metastatic disease.
Glutathione peroxidase 2 (GPX2) is a member of the glutathione peroxidase family, which can reduce hydrogen peroxide and fatty acid hydroperoxides to protect cells from the oxidative damage. [5][6][7][8] GPX2 is mainly expressed in the gastrointestinal tract including esophagus and liver, 9 and is also upregulated in some cancers, such as gastric cancer, colorectal cancer, esophageal cancer, liver cancer, and breast cancer. [10][11][12][13][14] Expression of GPX2 is associated with the growth, 13,15 metastasis, 13,16 and drug resistant 17,18 of cancer cells, as well as low patient survival rate. 10,15,19 Nevertheless, other studies have pointed tumor-suppressive function of GPX2 20,21 and showed that low GPX2 expression led to poor patient survival. 20,22 Thus, GPX2 seems to have a Janus-faced function as an oncogene and a tumor suppressor, which might depend on the tumor types and the different stages of malignant transformation. 5 In this study, we analyzed GPX2 expression in various human tumors with The Cancer Genome Atlas (TCGA) data and found that GPX2 was particularly highly expressed in lung cancer, including both lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). We validated GPX2 expression in NSCLC tissues by regular PCR and immunohistochemistry (IHC) and analyzed the correlation of GPX2 expression with clinicopathological features and survival. Moreover, we investigated the effects of GPX2 on epithelial-mesenchymal transition (EMT), migration, invasion, and metastasis of NSCLC cells in vitro and in vivo, and even explored the molecular mechanism of its action. GPX2 may play a carcinogenic role in NSCLC and may be a potential therapeutic target as well as prognostic and diagnostic biomarker for NSCLC.

| Data source and preprocessing
The gene expression data and clinical information of human tumors, including 594 LUAD-related samples (535 LUAD samples and 59 adjacent normal tissues) and 551 LUSC-related samples (502 LUSC samples and 49 adjacent normal tissues), were obtained from TCGA database (https://portal.gdc.cancer.gov). Level 3 high-throughput sequencing-fragments per kilobase of transcript per million mapped reads data were transformed into transcripts per million reads (TPM) by using R software package. The differential expression of GPX2 between tumor and normal tissues was analyzed by R software. Data were visualized using R package ggplot2. The diagnostic value of GPX2 in lung cancer was evaluated by receiver operating characteristic (ROC) curve.

| Sample collection
In total, 293 NSCLC samples with complete clinicopathological data, including 252 paraffin-embedded NSCLC samples, 31 paraffin-embedded NSCLC samples and their related lymph nodes, as well as 10 fresh NSCLC samples, were collected from the Second Affiliated Hospital of Dalian Medical University. The use of these samples was approved by the Ethics Committee of the Second Affiliated Hospital of Dalian Medical University. Patients included in this study were initially diagnosed as LUAD or LUSC. Patients who received preoperative chemoradiotherapy and/or died of surgical complications were excluded from the study. All participants provided written informed consent.

| Cell culture and treatment
Human NSCLC cell lines NCI-H520, NCI-H358, NCI-H1299, NCI-H460, and A549, as well as two normal lung cell lines (HLF and HBE) were obtained from the American Type Culture Collection (ATCC). LUAD cell line Anip973 was kindly provided by the Harbin Medical University. LUAD cell line SPC-A-1 was purchased from the Cell Bank of the Chinese Academy of Sciences. The cells were grown in RPMI-1640 media (Gibco) added with 10% fetal bovine serum (FBS, Gibco) and 1% penicillinstreptomycin at 37°C with 5% CO 2 in a humidified incubator. The NSCLC cells were treated with LY294002 (20 μM) for 24 h. For N-acetyl-L-cysteine (NAC), the cells were treated with a concentration of 2.5 mM for 24 h.

| Construction of stable cell lines with GPX2 overexpression or knockdown
The lentiviruses with GPX2 overexpression or knockdown as well as negative control were obtained from GenePharma Co., Ltd. and then infected the NSCLC cells with polybrene (Sigma). Stable cell lines with GPX2 overexpression or knockdown as well as their corresponding control cells were screened with puromycin. The shRNA sequences targeting GPX2 are as follows: shRNA1, 5′-GGGAG AAG GTA GAT TTC AATA-3′; shRNA2, 5′-GCCGC ACC TTC CCA ACC ATCA-3′; shRNA3, 5′-GCACA ACC ACC CGG GAC TTCA-3′.

Immunohistochemistry (IHC)
For ICC, three 10 cm dishes of NSCLC cells were collected and centrifuged. The cell pellet was fixed in formalin for 30 min and subsequently dehydrated with 95% ethanol at room temperature for 2 h. The cell pellet was wrapped in filter paper and fixed with formalin again. After dehydration, the cell pellet was embedded in paraffin. Then, the paraffin-embedded cell block was cut into 3-μm thick section like tissue block. ICC or IHC staining was performed according to reagent manufacturer's instructions. Primary antibodies including anti-GPX2 (#ab137431, Abcam, 1:600), anti-E-cadherin (#3195, CST, 1:500), and anti-vimentin antibodies (#5741, CST, 1:500) were incubated at 4°C overnight. After that, they were incubated with reaction enhancer and goat anti-mouse/rabbit immunoglobulin G (PV-9000; Beijing Zhongshan Golden Bridge Biological Technology) and colored with DAB and hematoxylin. Yellow granules in the cytoplasm of lung cancer cells were defined as positive, while unstained cells were negative. The proportion and staining intensity of positive cells were evaluated by semiquantitative method. Proportion of positive cell was scored as 0 (0%), 1 (1%-25%), 2 (26%-50%), 3 (51%-75%), and 4 (>75%). The staining intensity score was divided into 0 (no staining), 1 (light yellow), 2 (brownish-yellow), and 3 (brown). The two types of scores were then multiplied to determine the expression level: scores below 6 (<6) are low expression, while scores of 6-12 are high expression.

| Migration and invasion assays
In the wound healing experiment, cells were seeded into the 6-well plate. When the cell density was 90%-100%, vertical cross lines were drawn on the cells using a 200 μL sterile pipette tip. Cells were incubated in RPMI-1640 media with 2% FBS for 24 h in a humidified incubator. Several fields were selected and photographed at both 0 and 24 h. Scratch width was measured by Image J software. Mobility is represented by the ratio of the difference in scratch width between 0 and 24 h to the width at 0 h. The experiment was performed in triplicates.
In Transwell assay, cell suspension was put into an upper chamber of Transwell plate coated with or without Matrigel (BD Biosciences), and 600 μL medium containing 10% FBS was added to the lower chamber. After cultured in humidified incubator, cells in the upper chamber were wiped off, while cells on the outside of the upper chamber were fixed with methanol, stained with crystal violet solution, and photographed under a microscope. The numbers of cells in each field were counted.

| Lung metastasis assay
Male BALB/c nude mice (6 weeks old) were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. Ten mice were divided into two groups randomly. 1 × 10 6 GPX2-stable knockdown or control A549 cells were injected into the mouse tail vein (n = 5 for each group). Nude mice were fed for 8 weeks and then sacrificed. Lungs were removed, photographed, fixed, and embedded in paraffin. Then, they were sectioned followed by hematoxylin and eosin (H&E) or IHC staining. The animal study was approved by the Institutional Animal Care and Use Committee of Dalian Medical University.

| Measurement of reactive oxygen species (ROS) levels
DCFH-DA (Beyotime Biotechnology), a cellular permeable fluorescent probe, was used to detect ROS levels. In brief, adherent cells or cell suspension were incubated with DCFH-DA probe for 20 min at 37°C according to the manufacturer's instructions. The green fluorescence intensity of cells was observed under a fluorescence microscope or detected by flow cytometry after washing.

| RNA sequencing
The integrity of RNA was evaluated using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies). After the library was qualified, samples were sequenced by the Illumina Novaseq platform. Differential expression analysis was carried out using the edgeR R package (3.18.1). The p-values were adjusted with the Benjamini & Hochberg method for controlling the false discovery rate. Genes with adjusted p-value (Padj) < 0.05 and |log2 (fold change)| > 1 were considered as differentially expressed genes (DEGs). Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of DEGs were performed by using the clusterProfiler R package.

| Statistical analysis
Data were performed by using SPSS 19.0, GraphPad Prism 7.0, or Microsoft Excel. Quantitative results are expressed as mean ± SD. Comparison between two selected groups was assessed by Student's t-test (unpaired, two-tailed). Chi-squared test was used to analyze the association between GPX2 expression and patients' clinicopathological features. The overall survival (OS) was analyzed and compared by Kaplan-Meier method and log-rank test between groups. Prognosis factor was performed by Cox regression model. Spearman's rank correlation coefficient and chi-squared test were used to analyze the correlation between GPX2 expression in NSCLC tissues and that in lymph node metastases according to the IHC scores. The diagnostic ROC curve was used to demonstrate the diagnostic value of GPX2 expression in differentiating cancer (NSCLC, LUAD, or LUSC) tissues from noncancerous tissues. p < 0.05 was considered statistically significant.

| GPX2 is highly expressed in NSCLC
A pan-cancer analysis based on TCGA database showed that GPX2 mRNA was differentially expressed in multiple human tumors, including LUAD, LUSC, breast invasive carcinoma (BRCA), colon adenocarcinoma (COAD), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), and prostate adenocarcinoma (PRAD), in both non-paired ( Figure 1A) and paired samples ( Figure 1B). Specifically, we found that the expression of GPX2 mRNA was significantly higher in both LUAD and LUSC tissues than their adjacent normal tissues ( Figure 1C-F). In order to confirm the high expression of GPX2 in lung cancer, we detected GPX2 expression in 10 paired NSCLC tissues and adjacent normal tissues by using PCR. As shown in Figure 1G, GPX2 mRNA was expressed in 5 out of 10 cancer tissues, but was hardly detected in matched adjacent normal tissues. To validate GPX2 protein level is also upregulated in NSCLC, we performed IHC staining in another cohort of 252 paired NSCLC tissues and adjacent normal tissues. The staining intensity of GPX2 was divided into no staining, light yellow, brownish-yellow, and brown ( Figure S1). As expected, GPX2 protein was highly expressed in both LUAD and LUSC relative to matched adjacent normal tissues ( Figure 1H,I). GPX2 protein was expressed in 166 out of 252 NSCLC samples: 86 samples with high GPX2 expression and 80 samples with low GPX2 expression ( Figure 1J). However, in matched adjacent normal tissues, GPX2 protein was only sporadically expressed in bronchiolar epithelial cells of 14 samples ( Figure 1J).

| GPX2 expression is associated with the poor prognosis of patients with NSCLC
To explore the clinical significance of GPX2, we classified GPX2 expression into high expression group and low expression group of 252 cases of NSCLC tissues according to the immunohistochemical scores. GPX2 protein expression was significantly correlated with gender (p < 0.001), histological type (p < 0.001), lymph node metastasis (p < 0.001), tumor size (p = 0.017), and TNM stage (p = 0.022), as determined by chi-squared test (Table 1). We subsequently explored the relationship between GPX2 expression and overall survival using Kaplan-Meier analysis and log-rank test. As shown in Figure 2A-C, NSCLC, LUAD, or LUSC patients with high GPX2 expression had a significantly shorter overall survival than those with low GPX2 expression (p < 0.001 for NSCLC and LUAD patients; p = 0.017 for LUSC patients). To further assess the prognostic value of GPX2 in NSCLC, univariate and multivariate Cox regression analyses were employed. Univariate Cox regression analysis showed that GPX2 expression, lymph node metastasis, tumor size, and TNM stage were all significant prognostic factors for NSCLC patients (all of them, p < 0.001; Figure 2D and Table 2). In multivariate Cox regression analysis, GPX2, as well as tumor size and TNM stage, was identified as an independent prognostic factor for NSCLC patients (p = 0.013, hazard ratio (HR) = 1.600; Figure 2E and Table 2). Furthermore, we stratified patients with clinicopathological factors and analyzed the association of GPX2 expression with overall survival of these patients. Kaplan-Meier analysis showed that high GPX2 expression could also predict poor overall survival in patient subgroup with tumor size >3 cm (p = 0.001; Figure 2F), positive lymph node metastasis (p = 0.038; Figure 2G), TNM stage I (p = 0.010; Figure 2H), or TNM stage III (p < 0.001; Figure 2H), but not in patients with tumor size ≤3 cm (p = 0.553; Figure 2F), negative lymph node metastasis (p = 0.817; Figure 2G), or TNM stage II (p = 0.160; Figure 2H). Taken together, these findings suggest that GPX2 expression is associated with the poor prognosis of NSCLC patients, especially in patients with tumor size >3 cm, positive lymph node metastasis, TNM stage I, or TNM stage III.
To explore the correlation between GPX2 and metastasis, we then detected the expression of GPX2 by IHC in the tissues (primary cancer tissues, adjacent normal tissues, and lymph nodes) of 31 NSCLC patients with positive lymph node metastasis. We found that GPX2 was highly expressed in the metastatic lung cancer cells of 13 NSCLC patients, but was not expressed in lymphocyte of lymph nodes ( Figure 2I). The expression of GPX2 was higher in primary cancer tissues and lymph node metastases than that in adjacent normal lung tissues (p < 0.001; Figure 2J). Spearman's rank correlation coefficient analysis and chisquared test confirmed a strong positive correlation between GPX2 expression in primary NSCLC tissues and that in lymph node metastases (Spearman r = 0.834, p < 0.001; Figure 2K and Table 3). These results suggest that GPX2 is closely related to the lymph node metastasis of NSCLC.
Moreover, we evaluated the diagnostic value of GPX2 expression in differentiating lung cancer tissue from noncancerous tissue by ROC analysis. The area under the curve (AUC) of GPX2 in NSCLC, LUAD, and LUSC was 0.793, 0.767, and 0.835, respectively ( Figure 2L), consistently with the analysis result based on TCGA database ( Figure 2M), suggesting that GPX2 has high sensitivity for NSCLC diagnosis, and may be an ideal diagnosis biomarker for NSCLC.

| GPX2 induces EMT in NSCLC cells
Based on the important clinical significance of GPX2 in NSCLC, we next aimed to investigate whether GPX2 plays a critical role in NSCLC progression and metastasis. To test it, we first evaluated the expression of GPX2 in seven NSCLC cell lines by using qPCR method and found that GPX2 expression in A549 and H520 cells were greatly higher than that in other NSCLC cell lines ( Figure 3A). Western blotting and immunocytochemical staining further confirmed this result ( Figure 3B and Figure S2A). Therefore, we selected A549 and H520 cells with high GPX2 expression to construct GPX2-stable knockdown cell lines and selected Anip973 cells with low GPX2 expression to generate GPX2-stable overexpression cell line, by lentivirus-mediated methods. The GPX2-stable overexpression or knockdown cell lines were successfully confirmed by qPCR and Western blotting ( Figure S2B and Figure 3C). Interestingly, we observed that GPX2-stable overexpression Anip973 cells showed a mesenchymallike morphological feature compared with control cells (Figure 3D), indicating that overexpression of GPX2 may induce EMT in NSCLC cells. Indeed, by detecting EMTrelated markers with Western blotting, we observed that GPX2 overexpression led to downregulated E-cadherin but upregulated vimentin in Anip973 cells, while GPX2 knockdown had the opposite effects in A549 or H520 cells ( Figure 3E,F). Moreover, after GPX2 was knocked down, the mRNA levels of E-cadherin and vimentin were also changed as their protein levels ( Figure 3G). It is wellknown that one of the classical mechanisms of E-cadherin downregulation is that transcription of E-cadherin is inhibited by some transcription factors such as Snail. [23][24][25][26] Thus, we then examined whether GPX2 can also regulate the expression of Snail. As shown in Figure 3E,F, Snail was upregulated in GPX2-stable overexpression Anip973 cells compared with control cells, while it was downregulated in GPX2-stable knockdown A549 or H520 cells. To verify the important role of Snail in GPX2-induced E-cadherin and vimentin changes, we employed siRNAs to silence Snail in GPX2-stable overexpression Anip973 cells. Western blotting assay showed that knockdown of Snail reversed the GPX2 overexpression-induced downregulation of E-cadherin and upregulation of vimentin ( Figure 3H).

| GPX2 promotes migration and invasion of NSCLC cells
As EMT is closely related to cell migration and invasion, 27 we thus employed the wound healing and Transwell assays to investigate the role of GPX2 in migration and invasion of NSCLC cells in vitro. Wound healing assay showed that overexpression of GPX2 promoted the scratch healing capacity of Anip973 cells ( Figure 4A,B), while knockdown of GPX2 suppressed this capacity in A549 or H520 cells ( Figure 4C-E). Consistently, in  Transwell assays, overexpression of GPX2 significantly enhanced the migration and invasion capacities of Anip973 cells ( Figure 4F,G), while knockdown of GPX2 exhibited the opposed effects in both A549 and H520 cells ( Figure 4H-K). Therefore, these results showed that GPX2 promotes the migration and invasion capacities of NSCLC cells in vitro. In addition, knockdown of Snail greatly reversed the GPX2 overexpression-induced migration and invasion of Anip973 cells ( Figure S3), indicating that Snail is required for the GPX2-induced NSCLC cell migration and invasion.

| Knockdown of GPX2 inhibits the metastasis of NSCLC cells in vivo
To investigate the effect of GPX2 on metastasis of NSCLC cells in vivo, we injected the GPX2-stable knockdown or control A549 cells into the tail vein of nude mice. After 8 weeks, lungs were collected from nude mice and processed for H&E or IHC staining ( Figure 5A). The mice injected with GPX2-stable knockdown A549 cells developed less lung metastases than the mice injected with control A549 cells ( Figure 5B-E). We subsequently detected the EMT-related markers in these lung tissues by IHC staining and found that knockdown of GPX2 upregulated the expression of E-cadherin but decreased the expression of vimentin ( Figure 5F). Collectively, these findings suggest that knockdown of GPX2 can inhibit the metastasis ability of NSCLC cells in vivo.

| GPX2 downregulates ROS levels and activates PI3K/AKT/mTOR signaling pathway
As an antioxidant enzyme, GPX2 could participate in tumorigenesis by scavenging ROS, and is also important for the inhibitory function of YAP in LUAD progression via regulating ROS. 15,28 To validate the scavenging effect of GPX2 on ROS in NSCLC cells, we detected the ROS levels in GPX2-stable overexpression or knockdown NSCLC cell lines by using DCFH-DA probe, which can be oxidized to fluorescent DCF by intracellular ROS. As expected, the cellular fluorescence intensity of GPX2-stable overexpression Anip973 cells, observed under a fluorescence microscope, was weaker than that of control cells ( Figure S6A). In contrast, the fluorescence intensity of GPX2-stable knockdown A549 or H520 cells was stronger than that of control cells ( Figure 6A,B). Furthermore, by measurement of the cellular fluorescence intensity with flow cytometry, we found that overexpression of GPX2 deceased the fluorescence intensity of Anip973 cells ( Figure 6C), while knockdown of GPX2 increased the fluorescence intensity of A549 or H520 cells ( Figure 6D,E). These results suggest that GPX2 can reduce the levels of ROS in NSCLC cells.
To explore the potential mechanism by which GPX2 promotes tumor metastasis, RNA sequencing in GPX2stable knockdown or control A549 cells was performed. In total, 296 DEGs with 249 upregulated and 47 downregulated genes were identified, when GPX2 was knocked down by shRNA1 (sh1 vs. shNC dataset; Figure S4, left), while 150 DEGs with 102 upregulated and 48 downregulated genes were screened out, when GPX2 was knocked down by shRNA2 (sh2 vs. shNC dataset; Figure S4, right). GO enrichment analysis of these DEGs showed that extracellular matrix, proteinaceous extracellular matrix, peptidase inhibitor activity, and so on were enriched in both two data sets ( Figure 6F).
We also made KEGG analysis to find GPX2-related signaling pathway. In sh1 vs. shNC dataset, the main enrichment pathways were PI3K-AKT signaling pathway, cell adhesion molecules (CAMs), Ras signaling pathway, and so on ( Figure 6G, left). In sh2 vs.
shNC dataset, the main enrichment pathways were PI3K-AKT signaling pathway, Ras signaling pathway, Relaxin signaling pathway, and so on ( Figure 6G, right). Collectively, KEGG analysis indicates that PI3K/AKT signaling pathway is the important downstream pathway regulated by GPX2.
Therefore, we then examined the expression of PI3K/ AKT pathway-related proteins in GPX2-stable overexpression or knockdown NSCLC cells. Interestingly, GPX2 overexpression increased the PI3K and AKT phosphorylation (p-PI3K and p-AKT) levels but not their total proteins in Anip973 cells, while GPX2 knockdown decreased p-PI3K and p-AKT levels in A549 or H520 cells ( Figure 6H). To further confirm the effect of GPX2 on PI3K/AKT pathway, we evaluated the level of mTOR phosphorylation (p-mTOR), which is the direct downstream target of PI3K/AKT pathway. Consistently, the changes in p-mTOR after GPX2 overexpression or knockdown were similar to those in p-PI3K and p-AKT ( Figure 6H). Moreover, treatment of GPX2-stable overexpression Anip973 cells with LY294002 (an inhibitor of PI3K) could reverse the upregulation of p-AKT and p-mTOR caused by GPX2 overexpression ( Figure S5). To determine whether GPX2-mediated ROS removal is involved in PI3K/AKT/mTOR pathway activation, we then used NAC, a ROS scavenger, to clear ROS in GPX2-stable knockdown H520 cells. As expected, ROS scavenger NAC strongly reversed the GPX2 knockdown-induced ROS accumulation and inhibition of PI3K/AKT/mTOR pathway ( Figure S6B and Figure 6I). Furthermore, it has been reported that inhibition of mTOR can reduce the level of Snail. 29 Thus, these results, together with the effect of GPX2 on Snail as described above, suggest that GPX2 reduces ROS accumulation and activates PI3K/AKT/ mTOR/Snail signaling axis in NSCLC cells.

| DISCUSSION
Lung cancer is the leading cause of cancer-related death worldwide, of which NSCLC is the most common subtype. 1,4 The histological type and stage of lung cancer determine the treatment options and prognosis. F I G U R E 3 GPX2 induces EMT in NSCLC cells. (A, B) The expression of GPX2 in seven NSCLC cell lines was detected by using qPCR (A) and Western blotting (B). β-Actin was used as reference. (C) The overexpression or knockdown efficiency of GPX2 in stable NSCLC cell lines was verified by Western blotting. shNC, shRNA negative control; sh1/2/3, shRNA1/2/3 targeting GPX2. (D) Cell morphology of GPX2-stable overexpression or control Anip973 cells. (E, F) EMT-related proteins were detected by using Western blotting upon GPX2 overexpression or knockdown. Representative results (E) and corresponding quantification (F) are shown. (G) The mRNA levels of EMTrelated markers in GPX2-stable knockdown H520 cells were detected by using qPCR. β-Actin was used as reference. (H) Western blotting showing the effect of Snail knockdown on the expression of EMT-related markers in GPX2-stable overexpression Anip973 cells. siNC, siRNA negative control; siSnail, siRNA targeting Snail. *p < 0.05, **p < 0.01, and ***p < 0.001 were analyzed by Student's t-test between two selected groups.
Clinically, new biomarkers for NSCLC diagnosis and prognosis are urgently needed. In this study, we identified GPX2 as a potential prognostic and diagnostic biomarker for NSCLC. Functional studies showed that GPX2 contributed to the invasion and metastasis abilities of NSCLC cells. Furthermore, we also explored PI3K/AKT/mTOR/Snail signaling axis as an important mechanism for GPX2-mediated invasion and metastasis of NSCLC cells, which might provide a basis for individualized and precise treatment of patients with NSCLC.
GPX2, one member of glutathione peroxidase family, is important for the reduction in hydrogen peroxide and fatty acid hydroperoxides. 5,7,8 GPX2 has been reported to be associated with colonic inflammation, such as inflammatory bowel diseases. 30 Interestingly, GPX2 seems to have anti-inflammatory function, thus preventing inflammation-triggered carcinogenesis. [31][32][33] However, high expression of GPX2 was discovered in some cancers [10][11][12][13][14] and plays important roles in cancer progression and metastasis. 13,15,16 In this study, GPX2 expression was analyzed in various human tumors with TCGA data and the results showed that GPX2 mRNA level was significantly higher in both LUAD and LUSC tissues than in their adjacent normal tissues. It has been reported that GPX2 is upregulated in LUAD tissues as well as in cisplatin-resistant LUAD cells and tissues, thus promoting cisplatin resistance in LUAD. 17 Nevertheless, the clinical significance and the role of GPX2 in lung cancer progression and metastasis are still unclear. We thus used IHC to validate GPX2 expression in our cohort with 252 paired NSCLC tissues and adjacent normal tissues and found that GPX2 protein was also highly expressed in NSCLC, but hardly expressed in adjacent normal tissues.
Although GPX2 is highly expressed in some tumors, the clinical prognostic significance of GPX2 in various cancers is different. Several studies have clarified that high GPX2 expression is associated with poor prognosis in patients with gastric cancer or liver cancer, 10,19 while other studies reported that expression of GPX2 is related to a better survival in esophageal cancer. 12 This may be due to the distinctive roles of GPX2 in different cancer types. Although the clinical value of GPX2 in predicting overall survival in patients with lung cancer has been preliminary studied, 17,34 their conclusions are mainly based on the Kaplan-Meier Plotter database or TCGA database, which only reflects mRNA level, not protein level. In the current study, by using our collected NSCLC samples and IHC method, we found that high GPX2 protein level was associated with worse overall survival in patients with NSCLC, especially in patients with tumor size more than 3 cm, positive lymph node metastasis, TNM stage I, or TNM stage III. However, GPX2 expression was not associated with overall survival in patients with TNM stage II, which may be partly due to the small number of TNM stage II patients (n = 59). Importantly, GPX2 could even serve as an independent prognostic factor for NSCLC patients. In addition, according to the AUC, GPX2 may be a good pathological diagnostic marker for NSCLC, which could effectively differentiate cancer from noncancerous tissues. Collectively, these results suggest that GPX2 may serve as an important prognostic and diagnostic biomarker for NSCLC.
Moreover, in this study, we observed that GPX2 expression was associated with clinicopathological features, including lymph node metastasis, tumor size, and TNM stage. In addition, the expression of GPX2 in primary NSCLC was positively correlated with that in lymph node metastases. These results suggest that GPX2 may be involved in lymph node metastasis. After overexpression of GPX2 in Anip973 NSCLC cells, we found that the morphology of Anip973 cells changed from epithelioid to mesenchymal properties, indicating that GPX2 is able to promote the EMT process. Suppression of E-cadherin is a hallmark of EMT, which is partly due to the transcriptional repression by transcription factors, such as Snail. [23][24][25][26] Indeed, overexpression of GPX2 increased the expression of Snail and vimentin, decreased the expression of E-cadherin, and promoted the migration and invasion of NSCLC cells in vitro. However, GPX2 knockdown had the opposite effects on NSCLC cells in vitro, and inhibited lung metastasis of NSCLC cells in nude mice, collectively suggesting that GPX2 is involved in EMT and lung cancer metastasis.
To explore the mechanism by which GPX2 promotes NSCLC invasion and metastasis, we conducted RNA F I G U R E 4 GPX2 promotes migration and invasion of NSCLC cells. (A, B) GPX2-stable overexpression or control Anip973 cells grown to 90%-100% were subjected to the wound healing assay. Representative results (A) and corresponding quantification (B) are shown. (C-E) GPX2-stable knockdown (sh1 and sh2) or control (shNC) A549/H520 cells grown to 90%-100% were subjected to the wound healing assay.
Representative results (C, D) and corresponding quantification (E) are shown. (F, G) Transwell assays showing the migration and invasion capacities of GPX2-stable overexpression or control Anip973 cells. Representative results (F) and corresponding quantification (G) are shown. (H-K) Transwell assays showing the migration and invasion capacities of GPX2-stable knockdown or control A549/H520 cells. Representative results (H, J) and corresponding quantification (I, K) are shown. *p < 0.05, **p < 0.01, and ***p < 0.001 were analyzed by Student's t-test between two selected groups. sequencing in GPX2-stable knockdown or control A549 cell line. KEGG analysis showed that PI3K/AKT signaling pathway was enriched after GPX2 knockdown, which provided a basis for subsequent exploration of this signaling pathway. Indeed, Western blotting assay revealed that the levels of p-PI3K, p-AKT, as well as p-mTOR, a downstream target of PI3K/AKT signaling, were all downregulated in GPX2-stable knockdown NSCLC cells, but upregulated in GPX2-stable overexpression NSCLC cells. Interestingly, studies have reported that PI3K/AKT signaling can be regulated by ROS. [35][36][37] In our study, we found that GPX2 could reduce the levels of ROS in NSCLC cells, and ROS scavenger NAC reversed the GPX2 knockdowninduced ROS accumulation and inhibition of PI3K/ AKT/mTOR pathway, indicating that GPX2-mediated ROS removal is involved in PI3K/AKT/mTOR pathway activation. Moreover, it has been reported that inhibition of mTOR with mTOR kinase inhibitors could effectively decrease the levels of Snail and inhibit the cancer cell invasion and metastasis. 29 Here, our data showed that knockdown of Snail reversed the GPX2 overexpression-induced EMT as well as migration and invasion. Thus, our results together with literature reports suggest that GPX2 can promote EMT and metastasis of NSCLC cells by activating PI3K/AKT/ mTOR/Snail signaling axis via downregulating ROS. Therefore, inhibition of PI3K/AKT/mTOR activity by targeting GPX2 might be an effective therapeutic strategy to inhibit tumor metastasis. In summary, GPX2 was upregulated in both primary NSCLC and lymph node metastases and was correlated with clinicopathological characteristic, such as lymph node metastasis. High GPX2 expression was associated with the poor prognosis of patients with NSCLC. GPX2 promoted EMT, invasion, and metastasis of NSCLC cells. Mechanistically, GPX2 downregulated ROS levels and activated PI3K/AKT/mTOR/Snail signaling axis ( Figure 6J). Our findings revealed the important role of GPX2 in NSCLC and identified GPX2 as a potential therapeutic target and biomarker for patients with NSCLC.

AUTHOR CONTRIBUTIONS
S. Shao, X. Liu, and F. Peng conceived and designed this study. F. Peng and Q. Xu collected data. F. Peng, Q. Xu, X. Jing, X. Chi, Z. Zhang, X. Meng, X. Liu, and J. Yan carried out the experiments and analyzed the results. F. Peng wrote the initial paper. X. Liu and S. Shao revised the paper. S. Shao provided technical and material support. All authors read and approved the final manuscript.