EXT1 methylation promotes proliferation and migration and predicts the clinical outcome of non‐small cell lung carcinoma via WNT signalling pathway

Abstract DNA methylation is important for lung cancer prognosis. In this work, it is aimed to seek novel biomarkers with DNA methylation‐expression‐pathway pattern and explore its underlying mechanism. Prognostic DNA methylation sites and mRNAs were screened in NSCLC data set from TCGA, and further validated using the samples retrospectively collected, and EXT1 was identified as a potential target. Gene body methylation of three CpG sites (cg03276982, cg11592677, cg16286281) on EXT1 was significantly associated with clinical outcome, and the EXT1 gene expression also predicted prognosis. The expression level of EXT1 was also correlated with its DNA methylation level. This observation was further validated in a new data set consist of 170 samples. Knocking down of EXT1 resulted in decreased proliferation and migration. EXT1 targets were analysed using GSEA. It is found that the WNT signalling is the potential downstream target of EXT1. Further analyses revealed that the EXT1 targets the beta‐catenin and effect migration rate of NSCLC cell lines. The WNT signalling inhibitor, XAV‐939, effectively disrupted the migration promotion effect induced by EXT1. In summary, EXT1 methylation regulates the gene expression, effects the proliferation and migration via WNT pathway and predicted a poor prognosis for NSCLC.


| INTRODUC TI ON
Lung cancer is currently the leading cause of deaths in cancers. In China, 733 300 new cases and 610 200 deaths were estimated in 2015. Among the subtypes, non-small cell lung carcinoma, which includes lung squamous cell carcinoma and adenocarcinoma, is the main subtype 1 of lung cancer. The carcinogenesis and prognostic complexity of non-small cell carcinoma were reflected by the genomic and environmental heterogeneity. 2,3 DNA methylation is the most well-known epigenetic modification on mammal DNA, and the methylation pattern across genome significantly influence the DNA long-term interaction, and chromatin structure and thus play vital roles in many biological processes, including carcinogenesis and cancer development. For example, the methylation status of APC and RARB was significantly altered during NSCLC carcinogenesis, and its methylation is associated the prognosis of NSCLC. 4 Similarly, the promoter methylation of hOGG1 was shown to increase the risk of NSCLC carcinogenesis. 5 Using the promoter methylation status of four genes (p16, CDH13, RASSF1A and APC), Brock et al developed a discrimination model for stage I lung cancer carcinogenesis and validated with external dataset. 6 Additionally, LRP12 methylation was shown to predict the clinical response of NSCLC to carboplatin. 7 However, none of these biomarkers were used clinically, due to the unknown mechanism of these genes.
The aim of this study was to screen the genes that effect the cancer development via DNA methylation alteration in non-small cell lung carcinoma. Data in The Cancer Genome Atlas were selected and analysed, and EXT1 (exostosin glycosyltransferase 1) was identified as candidate gene. Subsequent analyses revealed that the DNA methylation of EXT1 regulates the expression level of EXT1 and thus promotes proliferation and migration. Further analyses revealed that WNT signalling pathway is the downstream of EXT1.

| Gene screening
The gene expression data of non-small cell lung carcinoma in TCGA data set, which consist of lung squamous cell carcinoma and lung adenocarcinoma, were downloaded from UCSC Xena website (https://xena.ucsc.edu/). The primary tumour samples were enrolled whereas the other samples (metastatic and normal samples) were excluded. The gene expression values were transformed to log2(FPKM + 1) values from the normalized data. Cox univariate regression was implemented by correlating the gene expression levels and overall survival information, and P values were calculated (P1). In addition, the samples were divided into high-expression and low-expression groups according to the mRNA abundances. Subsequently, log rank test was carried out evaluate the survival difference between high/low-expression group (P2). Genes with both P1 < 0.001 and P2 < 0.001 were identified as survival-related genes (SRG).
DNA methylation data based on Illumina 450K array of NSCLC sample were also downloaded from UCSC Xena. The sample enrolment criteria were implemented as expression data. The raw data were normalized using quantile algorithm. Correlation between DNA methylation level and overall survival was analysed using both Cox univariate regression and log rank test, as expression data processing procedure, as survival-related CpGs. Methylation sites significantly associated with survival using both mRNA (SRG) and methylation levels (SRC) were retained for further analyses.
The annotation information of the DNA methylation platform (Illumina 450K) was downloaded from the manufacture provided website. Genes annotated with at least 3 CpG sites were retained for further analysis.

| RNA extraction, cDNA preparation and qRT-PCR
Written consents have been achieved from the patients involved in this study. Total RNA from lung cancer cell lines (NCI-H520 and A549, squamous cell carcinoma and adenocarcinoma) and the enrolled NSCLC samples was extracted using TRIzol (Invitrogen, CA) by following the manufacturer provided steps. The raw quality of the isolated RNA was evaluated by both Nanodrop 2000 and agarose gel electrophoresis. The cDNA was synthesized by templating 1-2 μg total RNA with random primers and reverse transcriptase (Invitrogen, Los Angeles, CA, USA). Real-time PCR was implemented using the SYBR Green kit (Applied TaKaRa, Tokyo, Japan) and ABI PRISM 7900 sequence detector. In this step, 18s RNA was used as the endogenous control, and the relative mRNA levels in different batches were calculated according to its relative Ct values and 18s RNA Ct values.

| Proliferation and migration assays
For migration assay, Transwell filter champers (Costar, Corning, NY, USA) were used according to the manufacturer provided manual. After culturing for 12 hours, six random microscopic fields were selected, and cell number in each field of each group was counted, and each experiment was repeated for 3 independent times. For the NSCLC cancer cell line proliferation assays, siEXT1 and control cells (4 × 10 3 cells/well) were plated into 100 μL growth medium inside 96-well plates for different time. Cell density of each period was evaluated by measuring cell density OD values with the Cell Counting Kit 8 assay (Dojindo Laboratories, Kumamoto, Japan) according to manufacturer provided manual.

| Staining and treating with XAV-939
H520 cells were seeded on a 6-well plates and were treated with 2 μM lycopene for 6 hours; afterwards, the cells were fixed with cold methanol, treated with blocking 1% BSA, 0.1% gelatin buffer for 1 hour, incubated with β-catenin antibody for 1 hour, washed with PBS, treated using FITC-conjugated mouse anti-goat IgG antibody (Santa Cruz Biotechnology, Shanghai, China) for 1 hour, added 5 μg/mL of DAPI and place the cells to laser scanning confocal microscope analysis (Zeiss LSM510; Carl Zeiss AG Corporate, Shanghai, China) to generate β-catenin was detected using confocal microscopy. XAV-939 (Selleck, Huston, Texas) was diluted to 5 µM according to the manufacture provided manual and added to the cell lines to further diluted to 5 nM in both cell lines.

| Statistical analyses
All analyses were implemented on R software and packages. Survival analyses were implemented on R package 'survival'. Gene Set Enrichment analyses were carried out using java software 'GSEA' 8 released. Student's t test was used to assay clinical indicator comparison, the proliferation, DNA methylation and migration difference, and P < 0.05 was considered statistically significant.

| Identification of genes on methylationexpression-prognosis axis for NSCLC
To systematically screen the genes in methylation-expressionprognosis way, the non-small cell lung carcinoma samples, including lung adenocarcinoma and lung squamous cell carcinoma data set in The Cancer Genome Atlas (TCGA) data, were downloaded, TA B L E 1 The distribution of selected CpG sites and genes. These sites and genes listed were significantly associated with NSCLC survival combined and analysed. The DNA methylation difference between normal and NSCLC was identified, and CpG sites significantly associated with survival were also selected. In addition, gene expression profile was also analysed. Genes correlated with clinical outcome were also identified. Genes occurred simultaneously in both DNA methylation, and gene expression were retained. Furthermore, genes with multiple CpG sites inside its gene elements (TSS, promoter, gene body, UTR) were used for further analyses. The results were shown in Table 1. Genes with at least three CpG sites associated with clinical outcome of NSCLC were identified. The candidate genes were DKK1, EXT1, KIAA1324 and TRIM15. Among these genes, DKK1 has been widely reported for its prognostic value in NSCLC, 9 and prognostic value of TRIM15 has been emphasized. 10 Although the effect of KIAA1324 was also reported in NSCLC. 11 Thus, the EXT1 DNA methylation and its function were studied in this work.

| Prognostic value of EXT1 in NSCLC
The prognostic value of EXT1 was analysed on both expression and DNA methylation levels. As expected, the NSCLC samples with high EXT1 expression were shown to significantly associated with worse prognosis (median survival months: 42.3, 95% CI: 36.9-56.0), compared with the low-expression group (median survival month: 57.9 95% CI: 57.9-72.5, P = .00065), as shown in Figure 1A. The prognostic values of cg03276982, cg11592677 and cg16286281 methylation were also evaluated. As shown in Figure 1B

| Prognostic validation of EXT1 mRNA and methylation
Despite that the gene selection considered both mRNA expression and DNA methylation, the EXT1 selection was based on the data from TCGA cohort. Thus, the good performance of EXT1 for prognosis may resulted from over-fitness. To exclude such reason, 170 newly primary NSCLC samples were collected from the 900th Hospital of Joint Logistic Support Force. Gene expression and DNA methylation levels of cg03276982, cg11592677 and cg16286281 were quantified using qRT-PCR and Pyrosequencing methods, respectively. As expected, the expression of EXT1 was significantly associated overall survival of NSCLC (Figure 2A), and the similar DNA methylation pattern was observed in these three CpG sites ( Figure 2B-D). In consistent with previous results, the mRNA level of EXT1 was significantly and negatively associated with DNA methylation level of cg11592677 and cg16286281 ( Figure 2E,F), but not cg03276982 (not shown). Collectively, these results indicate that the prognostic value of DNA methylation and mRNA abundance of EXT1 is robust across cohorts.

| The Relationship between EXT1 and clinicopathological indicators
The correlation between EXT1 and clinical indicators was also analysed in our own data set. Overall, the expression of EXT1 and methylation of CpG sites were independent from clinical indicators, with

F I G U R E 3 The correlation between EXT1 biomarkers and clinical indicators.
The categorized clinical indicators are mostly independent from EXT1 expression (A), and methylation status of cg03276982 (B), cg11592677 (C) and cg16286281 (D) a few exceptions. As shown in Table 2, the expression of EXT1 was significantly associated with gender, and the methylation status of cg03276982 was correlated with age and primary tumour stage. In addition, relationship between clinical categories and EXT1 expression/ methylation was also analysed. As shown in Figure 3, most clinical indicators are independent from EXT1 expression/methylation. Cox multivariate regression analysis was carried out to evaluate the importance of EXT1 and other clinical indicators. As shown in Table 3, EXT1 and all these CpG sites were significantly associated overall survival, whereas the other clinical indicators were not, indicating that EXT1 and its CpG sites are important indicator and powerful biomarker for the prognosis of NSCLC.

| The impact of oncogene EXT1 on cell proliferation and migration in NSCLC cell lines
To further investigate the role of EXT1 in NSCLC, two different NSCLC cell lines, NCI-H520 and A549, were selected for functional validation. In this part, the impact of EXT1 on proliferation and migration was assayed. After knocking down of EXT1 in H520 and A549 ( Figure 4A), the proliferation rate of both cell lines was determined and compared with the control group, using CCK9 assay in several days. As expected, the cell lines with EXT1 knocking down has a significantly lower proliferation rate than the control group ( Figure 4B) from the 2nd day since culturing, indicating the role of EXT1 in cell cycle. Afterwards, the migration ability was also evaluated and compared between experimental and control group, and the migration rate of EXT1 knocking down group has a significantly less migration ability than the control ( Figure 4C). All results above indicate that the EXT1 gene expression facilitates cell proliferation and migration in NSCLC cell lines.

| Pathways associated with EXT1
To investigate the potential target genes downstream, Gene Set Enrichment Analysis (GSEA) was implemented by analysing EXT1-high F I G U R E 4 Impact of EXT1 on proliferation and migration. After knocking down of EXT1 with siRNA (A-B), the proliferation rate in NCI-H520 (C) and A549 (D) was significantly reduced, and the migration rate was also decreased by visualized by microscope (E) and counting (F) [Colour figure can be viewed at wileyonlinelibrary.com] and EXT1-low group, which were divided by median expression value, by using the data in TCGA data set. As shown in Figure 5A, cell cycle ( Figure 5B), P53 signalling ( Figure 5C), DNA repair and WNT signalling pathway ( Figure 5D) were significantly enriched.

| EXT1 effects the migration of NSCLC cell lines via WNT signalling pathway
As the impact of WNT signalling pathway has been emphasized in the other non-cancer cell lines, 12 we seek to investigate whether the EXT1 promotes cancer progression by WNT signalling pathway.
After knocking down of EXT1 by siRNA, the active beta-catenin protein abundance was compared with the control group. As expected, the beta-catenin protein level was significantly decreased in both H520 and A549 cell lines ( Figure 6A,B). For further validation, the beta-catenin was visualized using confocal microscope.
The beta-catenin was observed, and its abundance is less in siEXT1 cell line (siEXT1-H520) than the control (scramble-H520), as shown in Figure 6C. To further validate whether the EXT1 promotes the migration via WNT signalling, the EXT1 overexpression H520 and A549 cell line were constructed ( Figure 6D), and the migration rate

| D ISCUSS I ON
EXT1 (exostosin glycosyltransferase 1) is an endoplasmic reticulumresident type II transmembrane glycosyltransferase involved in the chain elongation step of heparan sulphate biosynthesis. It is ubiquitously expressed across tissues, in consistent with its promoter structure. 13 This gene has been shown to be paly critical roles in lung development. 14 This gene has been reported to have a frequent mutation in multiple osteochondromas, 15,16 and it is also a frequently mutated gene in lung cancer. 17  In addition to their anti-inflammatory effect in smokers, omega-3 polyunsaturated fatty acids (ω-3 PUFAs) supplements are suggested to modulate pivotal pathways underlying the progression of lung cancer. 26 As in the case of hepatocellular carcinoma, the latter may occur through inhibition of Wnt/β-catenin signalling pathway. 27 Thus, ω-3 PUFAs act similarly to the WNT signalling inhibitor, XAV-939 which effectively disrupts the migration promotion effect of EXT1. Consequently, administration of ω-3 PUFAs may be potentially effective in counteracting the EXT1 methylation-mediated lung cancer-promoting effects. Collectively, this study revealed that gene body hypermethylation of EXT1 results in EXT1 overexpression, activates WNT signalling pathway and predicts the survival of NSCLC.

CO N FLI C T O F I NTE R E S T
The authors declare no (potential) conflict of interest.

E TH I C S A PPROVA L A N D CO N S E NT TO PA RTI CI PATE
Written consents have been achieved from the patients involved in this study, and this study is approved by Ethnic Committee of The 900th Hospital of Joint Logistic Support Force.

CO N S E NT FO R PU B LI C ATI O N
None.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data were all accessible via the accession provided in the material section.