GDF15-Mediated Cholesterol Metabolism Promotes EMT and Metastasis in Esophageal Cancer

Background: GDF15 is a potential biomarker for patients with esophageal cancer (EC). However, the mechanistic role of GDF15 in the invasion and metastasis of EC remains poorly understood. Methods: We determined the expression and function of GDF15 in esophageal cancer cells (ESCCs) and in patient tissue samples using western blotting, migration and invasion assays, immunohistochemistry, Co-IP assays, and quantitative real-time-PCR. In addition, a pulmonary metastatic nude mouse model was used to determine the function of GDF15. We then supplemented our experimental results with database analysis to validate our ndings. Results: GDF15 was upregulated in EC, and was associated with poor differentiation, high metastasis rates and worse prognosis. GDF15 knock-down reduced the migration and invasion of ESCCs. Co-IP assays demonstrated its association with SCAP, while GDF15 knock-down decreased SCAP levels. SCAP overexpression reversed the migration, invasion and EMT in GDF15-siRNA ESCCs. However, after incubation with β-cyclodextrin (β-CD), the ability of migration and invasion was weakened, EMT was reversed again. Migration, invasion, and EMT were enhanced in GDF15-siRNA ESCCs cultured in the presence of cholesterol and were similar to GDF15-siRNA ESCCs overexpressing SCAP. In vivo, knockdown of GDF15 inhibited lung metastasis of ESCCs and was reversed by SCAP overexpression or high cholesterol diet. Increased lung metastasis after SCAP overexpression was partially suppressed by intraperitoneal injection of β-CD. In addition, we determined that GDF15 was a direct target of miR-1324, miR-1324 was down-regulated in EC tissues. MiR-1324 upregulation resulted in decreased GDF15 expression and metastasis in ESCCs. Conclusions: We demonstrated that SCAP mediated GDF15-induced the invasion and metastasis of EC by regulating cholesterol metabolism. In addition, GDF15 was shown to ESCCs, transition; β-CD, β-cyclodextrin; MIC-1, macrophage inhibitory cytokine-1; NAG-1, nonsteroidal anti-inammatory drug-activated gene-1; SCAP, element-binding protein; SREBP2, sterol regulatory element-binding protein 2; ER, endoplasmic reticulum; rhGDF15, recombinant human GDF15; RT-qPCR,


Introduction
EC is a common malignant tumor with a worldwide prevalence [1]. An estimated 246,000 new cases were diagnosed in China in 2015, with approximately 188,000 dying from the disease [2]. The 5-year overall survival rate is below 20% despite improvements in medical facilities and treatment methods. The poor prognosis of EC is largely due to its high metastatic rates [3]. Hence, it is important to identify the molecular mechanisms underlying metastasis of EC to aid in the development of new treatment strategies for EC patients.
Growth differentiation factor 15 (GDF15), is a divergent member of the TGF-β superfamily. GDF15 is also known as macrophage inhibitory cytokine-1(MIC-1) or nonsteroidal anti-in ammatory drug-activated gene-1 (NAG-1). In ammation, obesity, cardiovascular diseases, and tumors have been shown to induce abnormal expression of GDF15 [4,5]. Previous studies have shown that GDF15 expression is elevated in certain human tumors, and dysregulation of GDF15 has been correlated with tumor progression and poor clinical outcomes [6][7][8][9]. GDF15 has served as a potential biomarker for patients with EC [10]. However, the biological function of GDF15 in EC remains poorly understood. Previous reports regarding the role of GDF15 in the invasion and metastasis of EC are rare, with the underlying mechanism yet to be deciphered.
Sterol regulatory element-binding protein cleavage-activating protein (SCAP) is responsible for regulating the activation of the sterol regulatory element-binding protein 2 (SREBP2). When cellular cholesterol levels decline, SCAP dissociates from the endoplasmic reticulum (ER) and aides SREBP2 to the Golgi.
SREBP2 then undergoes protease cleavage to release its N-terminal domain, which then translocates into the nucleus to activate its target genes (HMGCR, LDLR) for cholesterol synthesis and uptake [11,12].
Cholesterol, a cellular membrane component, has a variety of functions, which includes signal transmission and regulating the activation of membrane receptors [13]. In addition, previous studies have suggested that cholesterol could promote colorectal cancer progression and inhibit cell apoptosis [14].
Cholesterol levels are higher in EC tissues compared to corresponding tumor-adjacent tissues [15]. To the best of our knowledge, few reports are available about the role of SCAP or cholesterol in the metastasis of EC. After analyzing protein interaction databases, we found an interaction between GDF15 and SCAP. We investigated the importance of this interaction to determine the underlying mechanism of GDF15 in the metastasis of EC.
MicroRNAs are an important class of non-coding endogenous RNAs that are involved in the posttranscriptional regulation of gene expression. This is accomplished by the binding of miRNAs to the 3′untranslated region (3′-UTR) of target mRNAs [16]. Aberrant expression of speci c miRNAs have been implicated in EC development and progression, i.e., miR-450a-5p has been shown to inhibit autophagy and enhance radio-sensitivity in EC [17], while miR-338-5p has been shown to reverse chemoresistance and inhibit invasion of EC [18]. The GDF15 mRNA transcript has a putative target sequence for miR-1324 in the 3′ UTR and has been predicted to be a direct target of miR-1324 using Targetscan. MiR-1324 has been previously identi ed as a tumor suppressor in hepatocellular carcinoma growth and metastasis [19].
Whether miR-1324 may also be dysregulated and function as a GDF15 regulator in EC is yet to be deciphered.
In the present study, we investigated the clinical signi cance and biological function of GDF15 in EC. We demonstrated that GDF15 downregulation inhibited tumor metastasis. Mechanistically, GDF15 regulated intracellular cholesterol levels by interacting with SCAP. The GDF15/SCAP axis promoted migration and invasion of ESCCs and was targeted by miR-1324. The results of our study indicated the signi cance of the GDF15/SCAP axis for EC metastasis and suggested that this pathway could be therapeutically targeted.

Tumor Samples
This study was approved by the Ethics Committee of Zhongshan Hospital of Fudan University, Shanghai, China (B2016-100). EC tumor and matched para-cancer tissues were obtained from twenty patients undergoing treatment at the Zhongshan Hospital of Fudan University, from 2018 to 2019. Follow-up and clinical pathology data were collected from these twenty patients. All tissues were stored at -80°C until required.
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) Total cellular RNA was extracted using TRIzol (Invitrogen; Thermo Fisher Scienti c, Inc.). The All-in-One™ miRNA qRT-PCR reagent kit (GeneCopoeia Inc. Maryland) was used to reverse-transcribe total cellular RNA into cDNA and then ampli ed using miRNA speci c primers. Target gene expression was quanti ed using the 2 -∆∆Cq method and normalization to U6 levels.
Western blot analysis RIPA buffer containing PMSF (Beyotime Institute of Biotechnology) was used to lyse cells on ice. The BCA protein assay (Pierce; Thermo Fisher Scienti c, Inc.) was then used to measure the protein concentration of the samples. SDS-PAGE was used to separate total protein (20 μg/well). Proteins were then transferred onto a PVDF membrane (EMD Millipore), blocked using 5% low-fat milk for one hour, and then incubated with the relevant primary antibodies at 4˚C overnight. The next day, membranes were incubated with secondary antibodies (Beyotime Institute of Biotechnology) for 1 hour at 37˚C. Protein bands were visualized using ECL (Pierce; Thermo Fisher Scienti c, Inc.). Target protein levels were normalized to β-actin levels .

Migration and invasion assays
Cellular migration and invasion were determined using Transwell Chambers (Corning Inc., USA). Brie y, 5 × 10 4 cells were added to the upper chamber (Matrigel (Corning Inc., USA), (diluted 1:9) pre-coated for invasion) with 200 μl of serum-free media. The bottom chamber contained 600ul of complete media.
After 24 h, cells on the upper surface of the membrane were removed, and cells that had migrated to the lower membrane were xed and stained. Cell numbers were analyzed using a microscope (Leica Microsystems, Germany).

Immunohistochemistry
Lung tissues from nude mice and clinical samples were xed at room temperature with 4% paraformaldehyde for approximately10 hour before para n embedding. Tissues were then sectioned into 4-μm thick slices. Depara nization and rehydration were performed using xylene and a graded series of ethanol respectively. Hematoxylin and Eosin (H&E) (Beyotime Institute of Biotechnology) was used to stain lung tissues. Clinical samples were incubated with speci c primary antibodies against GDF15 (cat.no. ab223539; Abcam), E-cadherin (cat.no. ab40772; Abcam), Vimentin (cat.no. ab92547; Abcam), SCAP (cat.no. ab125186; Abcam), SREBF2 (cat.no. ab30682; Abcam), or HMGCR (cat.no. ab174830; Abcam) at 4˚C overnight followed by incubation with the appropriate secondary antibody (cat. no. A0208, Beyotime Institute of Biotechnology) for 30 minutes at room temperature. DAB was used as the chromogen. Images were captured using a microscope (Leica Microsystems, Germany).

Co-immunoprecipitation (CoIP) assays
Total proteins were extracted using the Thermo Scienti c IP Lysis Buffer (Cat. No. 87787) containing PMSF. Cell lysates were incubated with antibodies against GDF15 (cat.no. ab223539; Abcam) or SCAP (cat.no. ab125186; Abcam) at 4 °C on a shaker overnight, followed by incubation with 50 µl Protein A/G PLUS-Agarose beads (Cat. No.sc-2003; Santa Cruz Biotechnology, Inc.) for an additional 10 h. Beads were then washed 4 times with 1 mL Lysis Buffer before denaturation using 2 × SDS sample buffer at 100 °C for 20 min for subsequent western blot assays.

Dual-Luciferase Reporter Assay
Target genes of miR-1324 were identi ed using Targetscan. Wild-type and mutant 3 UTR sequence of RNA uorescence in situ hybridization (FISH) miR-1324 and GDF15 expression in EC specimens were determined using FISH. Speci c probes were purchased from RiboBio (Guangzhou, China). After tissues were depara nized and dehydrated as described above, sections were hybridized with the speci c probes in hybridization buffer overnight at 37°C in a dark moist chamber. The next day, slides were washed and counterstained with DAPI. Images were captured using a uorescence microscope (Leica Microsystems, Germany).

Animal experiments
All animal experiments were approved by the ethics committee of Zhongshan Hospital of Fudan University, Shanghai, China. This study was performed according to the guidelines of the Shanghai Medical Experimental Animal Care Commission. Tumor implantation was performed using KYSE150 cells, which included cells transfected with NC-siRNA+NC-OE, GDF15-siRNA+NC-OE, and GDF15-siRNA+SCAP-OE. 1 × 10 6 cells were injected into the tail veins of BALB/c nude mice (18-20 g, 4-6 weeks). Two weeks later, mice injected with GDF15-siRNA+ NC-OE and GDF15-siRNA+ SCAP-OE were randomly divided into two groups, resulting in ve groups of n=5 for each group. Mice in the GDF15-siRNA+ NC-OE group were fed a modi ed cholesterol-rich diet containing 2% cholesterol (cat.no.TP 06106; Trophic Animal Feed High-tech Co., Ltd. Nantong, China), while the remaining groups were fed a control diet for 8 weeks. One group of GDF15-siRNA+ SCAP-OE mice were intraperitoneally injected with β-CD, while the other groups were administered a vehicle control twice a week. Two months later, mice were euthanized and the lungs harvested for H&E staining.

Statistical analysis
Data analysis was performed using the SPSS 24.0 software (SPSS, Inc.). Mean ± standard deviation was used to represent all values. One-way analysis of variance or student's unpaired t-test was used to compare three or two groups, respectively. p<0.05 was considered statistically signi cant.

GDF15 expression levels in EC and adjacent tissues
Using the gene expression TCGA database, we observed that GDF15 expression was higher in EC compared to adjacent normal tissue (Fig. 1A). This was consistent with our western blot results (Fig. 1B). High GDF15 expression levels predicted poor differentiation, higher metastasis rates and worse prognosis in patients with EC ( Figures 1C-E). In addition, GDF15 expression levels were negatively correlated with Ecadherin and positively correlated with vimentin (Fig. 1F). Hence, we hypothesized that GDF15 may be involved in invasion and metastasis of EC.
Effect of GDF15 on the migration and invasion of ESCCs.
Transwell chambers were used to determine the migration and invasion ability of ESCCs. As shown in Fig. 2A-2B, GDF15-siRNA ESCCs had reduced migration and invasion ability but was restored after incubation with rhGDF15 (100ng/ml) for three culture generations. Target genes associated with EMT were then analyzed. As shown in Fig. 2C, vimentin, snail, and slug levels were decreased and E-cadherin levels were increased in GDF15-siRNA ESCCs, suggesting reduced EMT. After incubation with rhGDF15, their expression levels were restored. These results suggest that GDF15 promotes migration and invasion of ESCCs.

SCAP mediated GDF15-induced migration and invasion of ESCCs.
It has been previously reported that GDF15 could promote colon cancer invasion and metastasis by activating c-Fos [20]. We investigated whether GDF15 played an additional role in EMT and metastasis of EC. As shown in Fig.3A, an interaction between GDF15 and SCAP was observed using the protein interaction database. This interaction was con rmed using CoIP assays (Fig.3B). Interestingly, the expression of SCAP and its downstream target genes, i.e., SREBF2 and HMGCR were reduced in siRNA-GDF15 ESCCs (Fig.3C). In addition, SCAP, SREBF2 and HMGCR expression levels in EC were higher compared to adjacent normal tissues (Fig.3D). We observed that SCAP overexpression could up-regulate SREBF2 and HMGCR expression levels (Fig.4A), and reverse the migration and invasion ability of GDF15-siRNA ESCCs (Fig.4B-4C), increase the expression of vimentin, snail, and slug, and decrease E-cadherin expression (Fig.4D). One of the main functions of SCAP is to regulate intracellular cholesterol levels [11]. SCAP overexpression has been reported to promote the accumulation of intracellular cholesterol in THP-1 macrophages [21]. To demonstrate the role of SCAP in promoting migration, invasion, and EMT of GDF15-siRNA ESCCs, β-CD, a reagent used to deplete intracellular cholesterol levels, was added to GDF15-siRNA ESCCs overexpressing SCAP. As shown in Fig.4E-4G, after β-CD incubation, migration and invasion ability were reduced, and EMT was reversed. In addition, invasion and EMT were enhanced after GDF15-siRNA ESCCs were incubated with cholesterol ( Fig.4H-4J), and were similar to GDF15-siRNA ESCCs overexpressing SCAP. In vivo experiments were consistent with these ndings. Knockdown of GDF15 expression inhibited ESCC lung metastasis and was reversed by SCAP overexpression or high cholesterol diet. The effect of SCAP overexpression on lung metastasis was partially offset by the intraperitoneal injection of β-CD (Fig.6A). These results suggest that SCAP mediated GDF15-induced migration and invasion of ESCCs was through cholesterol metabolism.

GDF15 is a target of miR-1324
Accumulating evidence has demonstrated that miRNAs play an important role in the progression of human cancers [22]. GDF15 mRNA transcript has a putative target sequence for miR-1324 in the 3′ UTR and was predicted to be a direct target of miR-1324 by Targetscan (Fig.5A). We validated this nding using luciferase reporter assays (Fig.5C-5D). In addition, subcellular co-localization analyses demonstrated a cytoplasmic interaction between GDF15 and miR-1324 in EC tissues (Fig.5F). The function and molecular mechanism of miRNA-1324 in EC were then investigated. As shown in Fig.7E, the expression of miR-1324 in adjacent non-tumor tissues was higher compared to EC tissues. After transfection with miR-1324 mimics, miR-1324 expression levels increased in ESCCs (Figure 5 B), while there was a decrease in GDF15, SCAP, SREBF2, and HMGCR expression levels and reduced migration, invasion and EMT (Fig.5G-5J). In vivo results were consistent with these observations. Mice injected with ESCCs transfected with lentivirus overexpressing miR-1324 had reduced lung metastasis (Fig.6B-6C). These results suggest that miR-1324 could reduce the invasion and metastatic ability of ESCCs by targeting GDF15.

Discussion
EC is a kind of human malignancy that is prevalent worldwide and endangers human life and health [1]. It is characterized by an insidious onset, low survival rate, high morbidity, and mortality. The 5-year overall survival rate is below 20%, with poor prognosis closely correlated with its high metastatic rate [3]. Hence, it is important to identify the underlying mechanisms of metastatic EC. A previous study indicated that GDF15 could serve as a potential biomarker for EC [10]. GDF15 not only promoted bone metastasis of prostate cancer but also induced colon cancer formation in senescence-associated tissue microenvironment [6,23]. However, the role of GDF15 in the invasion and metastasis of EC remains poorly understood, with potential mechanisms yet to be deciphered.
As a divergent member of the TGF-β superfamily, GDF15 is considered to be a promoter for the development and progression of some cancers. Additionally, it plays a role in apoptosis resistance, invasion, and metastasis of tumors [24][25][26]. In the present study, we observed that GDF15 expression levels in EC were higher compared to adjacent normal tissues (Fig. 1A-1B). In addition, high expression levels predicted poor differentiation, higher aggressiveness and worse prognosis in EC patients (Fig. 1C-1E). EMT is a process characterized by loss of cell-cell adhesion and increase migration and invasion ability. The trans-differentiation from quiescent epithelial cells into motile mesenchymal cells is essential for tumor progression [27,28]. We observed that GDF15 expression levels were negatively correlated with adhesion molecule E-cadherin levels and positively correlated with mesenchymal marker vimentin (Fig.  1F). To further understand the biological role of GDF15 in the invasion and metastasis of esophageal carcinoma, ESCCs were transfected with siRNA against GDF15. As shown in Fig. 2, migration and invasion ability were reduced and EMT was inhibited in GDF15-siRNA ESCCs. This was reversed when cells were incubated with rhGDF15. These results were further con rmed using in vivo mouse models (Fig.6A).
Next, we investigated the mechanism of GDF15 to promote invasion and metastasis of EC. We observed an interaction between GDF15 and SCAP (Fig.3A-3B). TGFβ1 has been reported to activate SREBF2 in kidney mesangial cells, and SREBF2 activation was dependent on SCAP [29,30]. In this study, knockdown of GDF15 decreased expression of SCAP, SREBF2, and HMGCR in siRNA-GDF15 ESCCs (Fig.3C), while SCAP overexpression could increase SREBF2 and HMGCR expression levels (Fig.4A) and reverse the migration, invasion, and EMT of GDF15-siRNA ESCCs (Fig.4B-4D). As a sterol sensor protein, SCAP forms a complex with SREBP2 to regulate its activation. This in-turn mediates de novo synthesis and uptake of cholesterol [11,12]. To further understand the role of SCAP in enabling invasion and migration of GDF15-siRNA ESCCs, β-CD was used to deplete intracellular cholesterol levels in GDF15-siRNA ESCCs overexpressing SCAP. β-CD incubation reduced the effect induced by SCAP overexpression in GDF15-siRNA ESCCs (Fig.4E-4G). Cholesterol addition restored migration, invasion and EMT in GDF15-siRNA ESCCs (Fig.4H-4J). In vivo animal experiments validated our ndings (Fig.6A).
Numerous studies have demonstrated that aberrant miRNA expression plays an important role in EC.
MiR-203a has been shown to inactivate the PI3K/AKT/mTOR pathway and inhibit EC development [31], while microRNA-10b induces cisplatin resistance by targeting PPARγ in EC [32]. However, the function and molecular mechanism of miRNA-1324 in EC progression are yet to be deciphered. The GDF15 mRNA transcript has a putative target sequence for miR-1324 in the 3′ UTR and was predicted to be a direct target of miR-1324 using Targetscan (Fig.5A). We validated these ndings using luciferase reporter assays ( Fig.5C-5D). miR-1324 expression levels in adjacent non-tumor tissues were higher compared to EC tissues. After transfection with miR-1324 mimics, miR-1324 expression levels increased in ESCCs (Figure 5 B), expression levels of GDF15, SCAP, SREBF2, and HMGCR decreased, migration and invasion were inhibited, and EMT reversed (Fig.5G-5J). In vivo, mice injected with ESCCs transfected with lentivirus overexpressing miR-1324 had reduced lung metastasis (Fig.6B-6C). miR-1324 was identi ed as a tumor suppressor in hepatocellular carcinoma growth and metastasis [19]. Our results suggest that miR-1324 could reduce the invasion and metastasis of ESCCs by targeting GDF15.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.