TGF-β induced EMT and stemness characteristics are associated with epigenetic regulation in lung cancer

Transforming growth factor-β (TGF-β) promotes tumor invasion and metastasis by inducing epithelial-mesenchymal transition (EMT). EMT is often related with acquisition of stemness characteristics. The objective of this study was to determine whether EMT and stemness characteristics induced by TGF-β might be associated with epigenetic regulation in lung cancer. A human normal lung epithelial cell line and four lung cancer cell lines were treated with TGF-β. Transcriptome analysis of BEAS-2B and A549 cells incubated with TGF-β were analyzed through next-generation sequencing (NGS). Western blotting was carried out to investigate expression levels of epithelial and mesenchymal markers. Wound healing and Matrigel invasion assay, sphere formation assay, and in vivo mice tumor model were performed to evaluate functional characteristics of EMT and stemness acquisition. To investigate whether activation of EMT and stem cell markers might be involved in epigenetic regulation of lung cancer, experiment using a DNA methyltransferase inhibitor (5-azacytidine, AZA), methylation-specific PCR (MSP) and bisulfite sequencing were performed. NGS revealed changes in expression levels of EMT markers (E-cadherin, N-cadherin, fibronectin, vimentin, slug and snail) and stem cell markers (CD44 and CD87) in both BEAS-2B and A549 cells. Functional analysis revealed increased migration, invasion, sphere formation, and tumor development in mice after TGF-β treatment. Expression of slug and CD87 genes was activated following treatment with AZA and TGF-β. MSP and bisulfite sequencing indicated DNA demethylation of slug and CD87 genes. These results suggest that TGF-β induced EMT and cancer stemness acquisition could be associated with activation of slug and CD87 gene by their promoter demethylation.

www.nature.com/scientificreports/ for tumor progression 5,6 . During the transition, epithelial cells can obtain invasive and migratory properties to become cells having stemness characteristics 3 . Cancer stem cell (CSC) has the ability to self-renew and initiate tumor formation. Thus, it is intrinsically resistant to therapy. EMT inducers such as TGF-β, Wnt, Notch, NF-kB, and ERK/MAPK pathways can promote stem cell characteristics of solid tumors [7][8][9][10] . During EMT and stemness acquisition, epigenetic mechanisms such as DNA methylation and histone modifications are involved in the regulation of EMT and stemness-related genes [11][12][13][14][15] . However, whether EMT and stemness characteristics induced by TGF-β might be associated with epigenetic regulation in lung cancer remains unclear. Thus, the aim of the present study was to evaluate the effect of TGF-β induced EMT on stemness acquisition of lung cancer cells and determine the possible epigenetic mechanisms involved in the development of lung cancer.

Methods
Cell culture and transcriptome analysis. A human normal lung epithelial cell line and four lung cancer cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). BEAS-2B cells were cultured in DMEM/F/12 medium. A549, H292, H226, and H460 cells were maintained in RPMI 1640 medium. Each of the medium was supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 100 μg/mL streptomycin, and 250 ng/mL amphotericin B. Cells were maintained at 37 ℃ in a 5% CO2 humidified atmosphere. To analyze the effect of TGF-β, cells were treated with 10 ng/mL TGF-β (R&D Systems, Minneapolis, MN, USA).
We used not only representative NSCLC cell lines (one adenocarcinoma cell line, NCI-A549; one squamous cell carcinoma cell line, NCI-H226), but also other lung cancer cell lines (one large cell carcinoma cell line, NCI-H460; one mucoepidermoid pulmonary carcinoma cell line, NCI-H292) for the supportive information. We want to confirm the effect of TGF-β on EMT and stemness acquisition, which is a general feature regardless of aggressiveness of lung cancer cell lines including normal lung cell line although BEAS-2B might not the best control for adenocarcinomas.
Transcriptomes of BEAS-2B and A549 cells treated with TGF-β for 72 h were analyzed as described previously 16 . Transcriptome analysis by next-generation sequencing (NGS) in Fig. 1 was based on two samples (BEAS-2B and A549) to screen EMT and stemness genes induced by TGF-β treatment.
Western blot analysis. Cells were lysed using radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitors for 20 min. Proteins (20 μg) extracted from cells were separated on SDS-PAGE gel and transferred to a nitrocellulose membrane. The membrane was blocked for 1 h with blocking buffer (5% non-fat milk in PBS containing 0.1% Tween 20) and then incubated with primary antibodies dissolved in blocking buffer at 4 ℃ overnight. Membranes were washed with TBS-T buffer (PBS with 0.1% Tween 20) for 5 min three times and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at RT for 1 h. The membrane was developed using an ECL Kit (Amersham Pharmacia Biotech, Little Chalfont, Buckinghamshire, UK) with X-ray film.
Migration, Matrigel invasion and sphere formation assay. Migration, Matrigel invasion and sphere formation assay were performed as described previously 16 . After overnight incubation of lung cells, wells were manually scratched using a 200-μL pipette tip on the bottom of the well. Migrated cells were cultured in serumfree medium with or without 10 ng/mL TGF-β for 72 h.
After treatment with TGF-β for 72 h, 2.5 × 10 4 cells were seeded onto Matrigel-coated membrane with Falcon Cell Culture Inserts in 24-well plates (pore size, 8 μm; Corning Inc., Corning, NY, USA) containing serum-free medium. Inserts were placed into 24-well culture plates that contained medium with 5% FBS. The ImageJ software was used to count the number of invaded cells.
Animal model. Mice experiments were carried out according to guidelines approved by the Institutional Animal Care and Use Committee of the Catholic University Medical School (Approval no. CUMC-2015-0080-01). Based on methods described in previous studies 17, 18 , we prepared an animal subcutaneous implantation model using BALB/c nude mice for the analysis of stemness characteristics. After incubating BEAS-2B, A549, and H226 cells with TGF-β for 72 h, 1 × 10 5 , 1 × 10 6 , or 5 × 10 6 cells were injected into each side of the mouse flank. Cells untreated with TGF-β were implanted on the left and cells treated with TGF-β were implanted on the right. Tumor development and growth were checked three times per week. At 2 weeks after injection, mice were sacrificed to obtain tumor samples. Methylation-specific PCR. Methylation-specific PCR was performed to investigate slug and CD87 meth- www.nature.com/scientificreports/ 10 μM of each primer, and 250 ng of template DNA using MyCycler™ PCR machine (Bio-Rad, Hercules, CA, USA). PCR products were separated with 2% agarose gel and detected using ChemiDoc™ Imaging Systems (Bio-Rad, Hercules, CA, USA) and ImageJ software.
Bisulfite sequencing PCR. Bisulfite sequencing PCR was evaluated using bisulfite-modified DNA at CD87 promoter region with the following primers: forward, TTT GGG GAT AGA GTT GTG ATT; reverse, CCT CAA TTA AAC CCT ATT CCAA. PCR products were then sequenced.
Statistical analysis. Data are presented as mean values ± standard error. Considering the small sample size, analyses were performed using non-parametric Wilcoxon rank-sum tests to identify significant differences between two groups. All reported P values are one-tailed P-value for a pre-specified direction in the alternative because the hypothesis of interest was only one-directional based on previous studies 19,20 regarding the induction of EMT and stemness acquisition by TGF-β treatment. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results
Transcriptome analysis following TGF-β treatment using NGS. To investigate the effect of TGF-β on mRNA expression in lung cells (BEAS-2B and A549), next-generation sequencing was performed for transcriptome analysis. Obvious differences in mRNA expression patterns were shown between cells treated with and without TGF-β (Fig. 1A). To analyze the effect of TGF-β on EMT and stemness characteristics, various EMT and stem cell markers were examined. Expression levels of E-cadherin, the epithelial marker, were decreased 5.1-to 6.1-fold while those of mesenchymal markers (N-cadherin, fibronectin, vimentin, slug and snail) were increased more than twofold in cells treated with TGF-β (Fig. 1B). Among stem cell marker candidates, CD44 and CD87 expression levels were increased around twofold following TGF-β treatment (Fig. 1C). Fold changes of various EMT and stem cell markers are presented in Table 1.

Expression of TGF-β induced EMT markers.
The protein expression levels of each marker were evaluated with a series of cell lines, BEAS-2B, A549, H292, H226 and H460. Consistent with transcriptome analysis, E-cadherin expression was decreased in BEAS-2B, A549, and H292 cells treated with TGF-β. Expression levels of N-cadherin, fibronectin and snail were increased in BEAS-2B, A549, H292 and H460 cells and those of vimentin and slug were increased in BEAS-2B and A549 cells treated with TGF-β although their expression levels differed among lung cell lines (Fig. 2).

Functional assessment of EMT following TGF-β treatment.
To examine whether TGF-β could augment migration and invasion, wound healing and Matrigel invasion assay were performed. Wound healing assay revealed that TGF-β increased migration of lung cells toward the center of the scratched area (Fig. 3A). Matrigel invasion assay indicated that TGF-β significantly increased transwell invasion of lung cells (Fig. 3B).

Acquisition of stemness in vitro and in vivo.
Sphere formation assay has been used to indicate stem cell growth. Sphere formation rates of all lung cells treated with TGF-β were increased compared to those untreated with TGF-β (Fig. 4A).
Tumor-forming ability of cells treated with TGF-β was compared following injection into BALB/c nude mice. The potential of lung cells to form a tumor was analyzed as the rate of tumor development based on size and weight. Tumor size and weight were higher in TGF-β treated cell groups than those in TGF-β untreated cell groups ( Fig. 4B and C).

Activation of slug and CD87 by its promoter demethylation. To examine whether EMT and stem
cell marker expressions were activated aberrantly, we investigated expression levels of EMT and stem cell markers after treatment with a DNA methyltransferase inhibitor AZA. AZA treated cells showed increased slug and CD87 expression levels compared with TGF-β (−) control (n = 3, P < 0.05; Fig. 5A). These findings suggest that DNA demethylation is associated with the activation of slug and CD87 expression.
DNA from lung cells was converted with sodium bisulfite. Unmethylation and methylation levels at slug and CD87 promoter regions were then compared among lung cells treated with or without TGF-β using methylationspecific PCR. After TGF-β treatment, unmethylation levels of slug and CD87 genes were increased whereas, methylation levels of slug and CD87 genes were decreased in both BEAS-2B and A549 cells (P < 0.05; Fig. 5B and C).
To indicate CpG site methylation pattern of CD87 promoter region, bisulfite sequencing was analyzed in BEAS-2B and A549 cells. Representative bisulfite sequencing results of six CpG sites (underlined letters) within a 100-bp promoter region of CD87 gene are presented in Fig. 6. After TGF-β treatment, demethylated CpG sites are indicated by red letters in Fig. 6. Heterozygote C/T double peaks (Y) were found at several sites.

Discussion
The objective of this study was to examine the effect of TGF-β on possible EMT and stem cell markers in the induction of EMT and stemness acquisition. Moreover, we investigated whether the mechanism of EMT and stemness acquisition was related to the activation of possible EMT and stem cell markers by epigenetic change of related genes. Various lung cell lines were used because we want to confirm whether the effect of TGF-β in the EMT and stemness acquisition is a common feature despite different aggressiveness of lung cancer cell lines  www.nature.com/scientificreports/ including normal immortalized lung cell line, BEAS-2B. We previously reported that CXCR4 activation by its aberrant promoter demethylation was associated with hypoxia-induced acquisition of EMT and cancer stem cell characteristics in lung cancer 16 .
In the present study, next-generation sequencing was performed for transcriptome analysis to investigate the effect of TGF-β on expression levels of EMT and stem cell markers. TGF-β induces the demethylation of H3K27me3 in Snail1 promoter and overexpresses Snail1, which cause EMT. Referring to these results, it's possible  . (B, C) Tumor formation abilities in BEAS-2B, A549, and H226 cells treated with or without TGF-β were compared following injection into BALB/c nude mice. Numbers of mice that developed tumors were higher in TGF-β treated cell groups compared to those in TGF-β untreated cell groups. Weight and size of the tumors were enhanced following injection of BEAS-2B (1 × 10 6 and 5 × 10 6 ), A549 (5 × 10 6 ), and H226 (5 × 10 6 ) cells treated with TGF-β compared with those in TGF-β untreated cell groups. Data are presented as mean ± SE (*P < 0.05, † P < 0.01). www.nature.com/scientificreports/ that the mechanism of TGF-β to induce EMT is correlated with Smad, which is the upstream molecule of Snail in the TGF-β signaling cascade 21,22 . Based on previous evidences, the following candidate EMT and stem cell markers were chosen: EMT markers, E-cadherin, N-cadherin, fibronectin, vimentin, α-SMA, slug, snail, Twist1, Twist2, ZEB1, ZEB2 and ZO-1; stem cell markers, CD44, CXCR4, ABCG2, ALDH1A1, EpCAM, CD90, Nanog, SOX2, SSEA4, CD166, BMI-1, nestin, Musashi-1, CD87, MET, SLC3A2, c-Myc, and KLF4 (Table 1). Among these EMT markers, N-cadherin, fibronectin, vimentin, slug, and snail were enhanced. Results were validated by www.nature.com/scientificreports/ western blotting. Among these stem cell markers, CD44 and CD87 expression levels were increased. This result was verified by real-time RT-PCR. Experimental conditions including TGF-β concentration, duration of TGF-β exposure, and the type of cell lines might have influenced experimental outcomes. TGF-β induced EMT was demonstrated by a reduction in epithelial marker (E-cadherin) and an induction in mesenchymal markers (N-cadherin, fibronectin, vimentin, slug and snail). These changes of EMT markers were functionally validated with wound healing and Matrigel invasion assay. All lung cell lines (i.e., normal cell line BEAS-2B and four cancer cell lines A549, H292, H226, and H460) displayed increased migration and transwell invasion. These results correspond with previous studies showing that TGF-β can induce EMT 23,24 .
Slug and snail are zinc finger transcription factors that regulate EMT in primary human cancers including pancreatic cancer, breast cancer, gastric cancer, lung cancer, and ovarian cancer [25][26][27] . Cell surface receptor CD44 and urokinase receptor CD87 can be candidate stem cell markers for multiple solid tumors including breast cancer, head and neck cancer, and ovarian cancer [28][29][30][31] . Although CD44 was analyzed to have 26 stemness signature, we could not find CD87 as a stemness marker from the StemChecker 32 . We designed our experiments to screen the stemness related genes with NGS, which was used by previous studies 30,33 indicating CD87 might be a potential stemness marker in SCLC and breast cancer. The results of these studies provided the stemness genes and their physiological meaning, which are the main reason why we selected the stemness markers.
EMT marker slug can be induced by TGF-β treatment in prostate cancer and lung cancer 34,35 . Stem cell marker CD87 can be activated by TGF-β in melanoma cells 36 . Previous reports have suggested that expression levels of EMT markers (such as slug and snail) and stem cell markers (such as CD44) can be regulated by DNA methylation of promoter regions of these genes 13,[37][38][39] . However, to the best of our knowledge, there have been few studies on DNA methylation associated with CD87 expression.
Methylation analysis was performed in cells exposed to TGF-β for 72 h to determine whether DNA methylation might play a role in EMT and stemness phenotype. DNA methyltransferase inhibitor AZA was used to evaluate whether TGF-β induced activation of EMT markers (slug and snail) and stem cell markers (CD44 and CD87) were related to epigenetic mechanism (DNA demethylation). We found that both TGF-β and AZA could activate expression of slug and CD87. However, AZA treatment did not activate expression of snail or CD44 (data not shown).
As far as we know, there have been few studies on the molecular mechanism of TGF-β induced EMT and stem cell markers or their functional relation in lung cancer. The present study confirmed that slug and CD87 were activated by TGF-β, which were related with their aberrant promoter demethylation. www.nature.com/scientificreports/ The present study suggest that TGF-β-induced EMT and cancer stemness acquisition are related with slug and CD87 activation by their aberrant promoter demethylation. However, further study is required on the detailed epigenetic mechanism and the development of potential therapeutics for lung cancer.