Growth differentiation factor 15 induces cisplatin resistance through upregulation of xCT expression and glutathione synthesis in gastric cancer

2.1 Differential expression and survival analyses in clinical databases

Gene Expression Profiling Interactive Analysis was used to analyze the gene expression levels of GDF15 and GFRAL between normal and tumor tissues of gastric cancer patients (http://gepia.cancer-pku.cn/).²¹ The OS and PFS of gastric cancer patients expressing different levels of GDF15 were determined using the Kaplan–Meier plotter with the GDF15 gene (Affy ID 221577_x_at) and GSE15459 dataset (https://kmplot.com/analysis/).²²

2.2 Patient samples

A total of 100 patients with gastric cancer who received curative surgery were included. Patients who underwent palliative surgery or who had unresectable tumors were excluded. The operated specimens were fixed and wax embedded immediately after the procedure. The definition of patient characteristics and subsequent clinical follow-up were modified from a previous study.²³

2.3 Immunohistochemistry analyses

Initially, clinical samples from 100 patients with gastric cancer were analyzed by IHC staining using a previous method.²⁴ Because some tumor tissues were absent from the collected slices, we evaluated GDF15 and GFRAL expressions in 87 and 92 tumor slices, respectively. The GDF15 (sc-377195; Santa Cruz Biotechnology) and GFRAL (PA5-24545; Invitrogen, Thermo Fisher Scientific) Abs were used, and the expression of GDF15 or GFRAL in tissue samples was assessed with a semiquantitative system. The IHC analysis was independently performed by a pathologist (Li AFY). SPSS software (version 20.0; SPSS Inc.) and R software (The R Project for Statistical Computing) were used to analyze the statistical calculations and undertake the Kaplan–Meier survival analysis.

2.4 Serum and culture medium GDF15 analyses

The GDF15 levels of available serum samples derived from gastric cancer patients (n = 97) and from cell culture medium (12 h and normalized by cell number) were analyzed using a GDF15 ELISA (ELH-GDF15; Raybiotech) according to the user manual.

2.5 Cell culture and reagents

AGS, NUGC-3, and TSGH9201 human gastric cancer cells and human embryonic kidney 293 T (HEK293T) cells were cultured in a 5% CO₂ and 37°C incubator. The gastric cancer cells were cultured with a medium RPMI-1640 (Gibco) containing 10% FBS (Gibco) and 1% penicillin–streptomycin (Biological Industries). HEK293T cells were maintained in DMEM (Gibco) containing 10% FBS, 1% antibiotics, 1% nonessential amino acids, and 1% L-glutamine (Biological Industries). Cisplatin (P-4394), SSA, BSO, and NAC were obtained from Sigma-Aldrich (Merck). The GDF15 neutralizing Ab (MAB957) and recombinant human GDF15 (Escherichia coli-expressed) protein (#9279-GD-050) were obtained from R&D Systems. The purified mouse IgG2b κ isotype control Ab (400302) was obtained from BioLegend. SPP86 was purchased from Tocris Bioscience (Bio-Techne). SB-431542 was purchased from Cayman Chemical.

2.6 RNA extraction and quantitative real-time PCR

TRIzol reagent (Invitrogen) was used to extract cellular RNA. RNA (5 μg) was used for reverse transcription to cDNA with RevertAid reverse transcriptase (Invitrogen). The StepOne System (Thermo Fisher Scientific) and KAPA SYBR FAST (Roche, Merck) were used to undertake real-time PCR as previously reported.²³ The oligonucleotide primer sequences were: GAPDH-f: 5′-CCGTCTAGAAAAACCTGCC-3′; GAPDH-r: 5′-GCCAAATTCGTTGTCATACC-3′; GDF15-f: 5′-GTGTTGCTGGTGCTCTCGTG-3′; GDF15-r: 5′-CGGTGTTCGAATCTTCCCAG-3′; xCT-f: 5′-TCATTGGAGCAGGAATCTTCA-3′; and xCT-r: 5′-TTCAGCATAAGACAAAGCTCCA-3′.

2.7 Western blot analysis

Protein samples were prepared using a radioimmunoprecipitation assay lysis buffer.²³ Samples were further heated and analyzed with SDS-PAGE. After transferring and blocking, targeted proteins were detected with specific primary/secondary Abs and ECL reagents using GE Healthcare luminescence fluorescence image capture system (AI680). The MultiGauge image software (Fujifilm) was used for quantification. The PERK (sc-32577) Ab was purchased from Santa Cruz Biotechnology. The p-GCN2 (ab75836) and p-PKR (phospho T446, ab32036) Abs were purchased from Abcam. Eukaryotic initiation factor 2α (AHO0802), p-eIF2α (44-728G), GFRAL (PA5-24545), and α-tubulin (A11126) Abs were obtained from Invitrogen. Antibodies against GCN2 (#65981), PKR (#12297), PERK (#5683), xCT (#12691), Akt (#4685), p-Akt T308 (#2965), p-Akt S473 (#9271), STAT3 (#9139), p-STAT3 (#9145), ERK1/2 (#9101), and p-ERK1/2 (#9102) were purchased from Cell Signaling Technology. Heme-regulated eIF2α kinase Ab was purchased from MyBiosource (MBS2538144). Activating transcription factor 4 (10835-1-AP) Ab was obtained from Proteintech.

2.8 Small interfering RNA

The mixed siRNA solution was prepared according to a previous study.²³ Mixed siRNA solutions (ON-TARGET plus SMART Pool human siRNAs: scramble siRNAs, D-001810-01-05; GDF15, L-019875-00-0005; GFRAL, L-032083-02-0005; PKR, L-003527-00-0005; HRI, L-005007-00-0005; and GCN2, L-005314-00-0005) were added to the cell culture dish for 48–72 h (Dharmacon, GE Healthcare).

2.9 Sulforhodamine B assay for cell growth curve and drug sensitivity

At the end of experiments, cells were fixed by trichloroacetic acid (Sigma-Aldrich), stained by sulforhodamine B (Sigma-Aldrich), dissolved with Tris Base (10 mM; J.T. Baker), and determined at optical density 510 nm with Sunrise absorbance microplate reader (Tecan) as described previously.²³

2.10 Intracellular ROS and mitochondrial ROS levels

After treatment, cells were stained with DCF (Molecular Probes, Thermo Fisher Scientific) diacetate or MitoSOX Red (Molecular Probes) according to a previous study.²³ The fluorescence intensity of DCF and MitoSOX Red was determined by FACSCalibur flow cytometry (Becton Dickinson). Cells (≥10,000) were evaluated by CellQuest software (Becton Dickinson).

2.11 Propidium iodide exclusion assay

After treatment, cells were harvested and stained with PI solution (Sigma-Aldrich). The PI fluorescence intensity was determined using a flow cytometer according to a previous study.²⁵ Cells (≥10,000) were collected and the proportion of cells with PI exclusion was evaluated using CellQuest software.

2.12 Transwell migration assay

Cells were seeded onto Transwell inserts in a serum-free medium. Transwell inserts were incubated using a 24-well plate with serum-containing medium. The inserts were then fixed with methanol and visualized with Liu's A and B (Tonyar Biotech) according to a previous study.²³ A total of four view fields in each group were analyzed under a microscope with 100–200× magnification and evaluated with ImageJ software (NIH).

2.13 Overexpression of GDF15

The plasmid mixture of pcDNA plasmid carrying the ORF of GDF15 (pcDNA-GDF15) or empty vector (pcDNA) and Lipofectamine 3000 transfection reagent (Invitrogen) was added to the cells with an antibiotic-free medium for 12 h according to the manufacturer's instructions. Subsequently, the cells were cultured with fresh medium for 48 h.

2.14 Reporter luciferase activity assay

The cells were transfected with a mixture of plasmid and Turbofect transfection reagent (Thermo Fisher Scientific). According to a previous study,¹⁹ enhanced green fluorescent protein was cotransfected to normalize the luciferase activity. The pGL3 vectors (Promega) carrying the different (WT, ARE-mutant, and AARE-mutant) xCT luciferase reporter promoters were constructed in a previous study.¹⁹ The reporter activity was analyzed using the Bright-Glo Luciferase Assay System (Promega) according to the manufacturer's instructions.

2.15 Glutathione level assay

Determination of reduced and total GSH levels was carried out by a GSH Colorimetric Assay Kit (ab239709; Abcam) according to a previous study.²⁰

2.16 Statistical analyses

Data are presented as mean ± SEM. A p value of less than 0.05 was regarded as a significant difference. Statistical analysis was carried out with SigmaPlot (Systat) and graphs were plotted with Prism software (GraphPad).

3.1 GDF15 gene expression is elevated in gastric tumors and high serum GDF15 level is correlated with poor prognosis in gastric cancer patients

Using the GEPIA online database, we found that the gene expression level of GDF15 in tumor tissues was significantly higher than that in normal tissues of patients with gastric cancer (Figure 1A). In addition, the Kaplan–Meier plotter showed that high GDF15 gene expression had significantly lower OS and PFS rates in patients with gastric cancer than in those with low GDF15 gene expression (Figure 1B).

We further used IHC stain to determine the protein expression of GDF15 in gastric tumor tissues and found that GDF15 was highly expressed in male patients (Table 1). However, high GDF15 protein expression did not interfere with the OS or DFS rates (Table 1). To evaluate whether circulated GDF15 can predict the prognosis in gastric cancer, we determined the serum level of GDF15 using ELISA. The upper quartile of serum GDF15 level in this study was used to define the high and low levels. We found that high serum GDF15 levels were predominantly observed in male patients or those aged ≥65 years (Table 2). Low serum GDF15 levels were more frequently observed in gastric cancer patients with positive nodal involvement and increased depth of cancer invasion (Table 2). The gastric cancer patients with high serum GDF15 levels had lower OS and DFS rates compared to patients with low serum GDF15 levels (Figure 1C and Table 2). These results suggest that serum GDF15 levels might be a reliable prognostic marker for gastric cancer.

3.2 Expression of GDF15 is essential for cell proliferation and migration

To clarify the role of GDF15 in gastric cancer, siRNA against GDF15 and pcDNA-GDF15 were used. The siGDF15 could ameliorate cell proliferation in human gastric cancer cells (Figure 2A). In contrast, GDF15 overexpression enhanced the cell growth (Figure 2B). Moreover, knockdown and overexpression of GDF15 repressed and increased cell migration, respectively (Figure 2C,D). These results suggest that GDF15 expression is essential for cell proliferation and migration in human gastric cancer cells.

3.3 Growth differentiation factor 15 is abundant in cisplatin-resistant gastric cancer cells and responsible for cisplatin resistance

To determine the GDF15-mediated chemoresistance, we compared the GDF15 expression between previously established CisR human gastric cancer cells^{19, 20} and their parental cells. The gene expression of GDF15 in CisR cells was significantly higher than that in parental cells (Figure 3A). Similarly, GDF15 protein expression and secreted GDF15 levels were significantly increased in CisR cells compared to parental cells (Figure 3B,C). Moreover, knockdown of GDF15 significantly decreased cisplatin resistance in CisR cells (Figure 3D). Using the PI exclusion assay, we further validated that the knockdown of GDF15 can increase cisplatin sensitivity in CisR cells (Figure 3E).

Based on the fact that GDF15 is a secreted growth factor that acts through autocrine or paracrine signaling, we treated CisR cells with GDF15 neutralizing Ab and found that it can increase cisplatin sensitivity (Figure 3F). In contrast, treatment with rhGDF15 reduced cisplatin sensitivity in parental cells (Figure 3G). Moreover, overexpression of GDF15 reduced cisplatin sensitivity in parental cells (Figure 3H). High GDF15 could decrease cisplatin sensitivity in human gastric cancer cells.

3.4 Growth differentiation factor 15 upregulated xCT expression through eIF2α-ATF4 signaling and increased GSH levels contribute to cisplatin resistance

Intracellular GSH has been reported to inactivate cisplatin and contribute to resistance of cisplatin.¹⁸ We previously found that CisR human gastric cancer cells contained higher intracellular GSH levels than parental cells.^{19, 20} To evaluate whether GDF15 can modulate intracellular GSH levels, we determined the intracellular total and reduced form of GSH levels in CisR cells following GDF15 knockdown. Knockdown of GDF15 decreased the total and reduced GSH levels in CisR cells (Figure 4A,B). Moreover, consistent results were observed in CisR cells following GFRAL knockdown (Figure 4A,B). Knockdown of GDF15 significantly increased the intracellular and mitochondrial ROS levels in CisR cells under cisplatin treatment (Figure 4C,D). These results suggest that increased GDF15 in CisR cells can modulate intracellular GSH and ROS levels. We previously revealed that the cystine-uptake transporter xCT-increased intracellular GSH levels was crucial for cisplatin resistance in human gastric cancer cells.^{19, 20} Knockdown of GDF15 decreased the gene and protein expressions of xCT in CisR cells (Figure 4E,F). The elevated GDF15 might upregulate xCT and increase intracellular GSH levels, which downregulates cisplatin sensitivity in human gastric cancer cells.

Transcription factors ATF4 and NF-E2-related factor 2 (Nrf2) can upregulate xCT gene expression.²⁶ Figure 4F also shows that knockdown of GDF15 reduced ATF4 expression and the phosphorylation level of eIF2α in CisR cells. To determine the regulation of xCT, we transfected AGSCisR cells with pGL3 luciferase reporter vectors carrying the different types (WT, ARE-mutant, or AARE-mutant) of xCT promoters (Figure S1A). Knockdown of GDF15 significantly reduced the promoter activities of ARE-mutant and WT xCT but did not affect the activity of AARE-mutant xCT promoter (Figure 4G). These results suggest that GDF15 might upregulate xCT expression through the eIF2α-ATF4 pathway and increase intracellular GSH levels in CisR cells.

To validate the GDF15 on upregulation of eIF2α-ATF4-xCT and intracellular GSH levels, we transfected with pcDNA-GDF15 to overexpress GDF15 in parental cells. Overexpression of GDF15 increased the gene and protein expression of xCT (Figure 5A,B) and upregulated the eIF2α-ATF4 pathway (Figure 5B). Overexpression of GDF15 also significantly increased the activities of WT and ARE-mutant xCT promoters, but mildly increased the activity of AARE-mutant xCT promoter in HEK293T cells (Figure 5C). Additionally, SSA (an xCT inhibitor²⁷) and BSO (a GSH synthesis inhibitor²⁸) can reverse GDF15 overexpression-decreased cisplatin sensitivity (Figure 5D). These results indicate that GDF15 overexpression upregulates the eIF2α-ATF4-xCT pathway and increases intracellular GSH levels, which could downregulate cisplatin sensitivity in gastric cancer cells.

Growth differentiation factor 15 belongs to a novel TGF-β superfamily of cytokines^{29, 30} and triggers cellular response through the Smads pathway.^{31, 32} Some lines of evidence showed that GDF15 might contribute to chemo- or radioresistance through the Smads pathway.^33-35 A previous study mentioned that recombinant GDF15 might be contaminated by TGF-β and the effect and signaling transduction of GDF15 must be interpreted cautiously.³⁶ In our study, we found that both GDF15 overexpression and rhGDF15 could induce cisplatin resistance (Figure 3G,H). In addition, GDF15 neutralizing Ab might sensitize the cisplatin-resistant gastric cancer cells to cisplatin (Figure 3F). We further used SB-431542 (an inhibitor of the TGF-β family type 1 receptors/ALK5, ALK4, and ALK7 that preferentially inhibit Smad2/3³⁷) to dissect the role of the Smads pathway in the GDF15-mediated cisplatin resistance. However, the cisplatin sensitivity of the AGSCisR cells was not changed by SB-431542 (Figure S1B,C). These results suggest that the Smads pathway in GDF15-mediated cisplatin resistance might be minimal.

3.5 Growth differentiation factor 15-GFRAL-enhanced eIF2α-ATF4-xCT pathway and cisplatin resistance are mainly mediated by ROS-activated GCN2

Four eIF2 kinases (GCN2, HRI, PERK, and PKR) are responsible for the phosphorylation of eIF2α.^38-41 To elucidate the upstream regulator of eIF2α in CisR cells, we determined whether siGDF15 and siGFRAL affected the phosphorylation levels of GCN2, PKR, and PERK. Knockdown of GDF15 or GFRAL decreased the phosphorylation levels of GCN2 and PKR, as well as the eIF2α phosphorylation and the expression of ATF4 and xCT in CisR cells (Figure 6A,B). Additionally, we used SPP86 (an inhibitor of RET, the GFRAL downstream target^{42, 43}) to evaluate the GDF15-GFRAL-RET downstream regulation. Similar to thyroid cancer cell lines,⁴² the treated dose of SPP86 could effectively inhibit the RET-ERK and Akt signaling (Figure S1D). SPP86 could suppress the GCN2, PKR, and eIF2α phosphorylation and the ATF4 and xCT protein expressions in CisR cells (Figure 6C,D). Our results indicated that GCN2 and PKR could be crucial for the GDF15/GFRAL-mediated eIF2α-ATF4-xCT signaling in CisR cells.

Although the levels of PKR phosphorylation can be downregulated in CisR cells after knockdown of GDF15 or GFRAL and treatment with SPP86 (Figure 6), knockdown of PKR did not reverse the GDF15 overexpression-activated eIF2α phosphorylation or protein expressions of ATF4 and xCT in parental cells (Figure 7A). Moreover, siRNA against HRI could not reverse the GDF15 overexpression-activated eIF2α phosphorylation or the ATF4 and xCT protein expressions (Figure 7B). Furthermore, knockdown of HRI and PKR did not sensitize CisR cells to cisplatin (Figure 7C). Importantly, the ATF4 and xCT protein expression and phosphorylation of eIF2α were suppressed by knockdown of GCN2 in GDF15-overexpressed cells (Figure 7D). Our result revealed that GDF15 activates GFRAL-GCN2-xCT signaling and decreases cisplatin sensitivity in human gastric cancer cells.

General control nonderepressible 2 could not only be activated under amino acid deprivation, but also be triggered by UV irradiation, ATP-competitive kinase inhibitors, and oxidative stress.^{19, 41, 44-47} Moreover, GDF15 overexpression might increase the oxidative stress.^{48, 49} In order to dissect the role of ROS in the GDF15-regulated GCN2, we assessed the intracellular ROS levels under GDF15 overexpression. We found that GDF15 overexpression might elevate the intracellular and mitochondrial ROS levels (Figure S1E,F). Furthermore, NAC (a ROS scavenger⁵⁰) could mitigate the GDF15 overexpression-elevated GCN2 phosphorylation (Figure S1G). These results suggest that ROS might be a critical mediator for GCN2 activation under GDF15 overexpression.

3.6 Glial cell-derived neurotrophic factor family receptor a-like required for GDF15-mediated cancer progression of gastric cancer cells

As a GDF15 receptor, GFRAL is important for nutritional homeostasis.^7-9 Although it has been identified in the nervous system and in some cancer cells, the exact role of GFRAL in gastric cancer remains unknown.^{51, 52} The GEPIA online database showed that the gene expression of GFRAL was not significantly different between the tumor and normal counterparts of gastric cancer patients (Figure 8A). To further validate the role of GFRAL in patients with gastric cancer, we determined the GFRAL protein expression with IHC stain and characterized their clinical importance. Table 3 shows that the GFRAL protein was highly expressed in malignant gross appearance, such as Borrmann type 3 and 4; however, GFRAL expression was downregulated in advanced malignant cell grade, worse Lauren's histology, and increased nodal involvement. In addition, gastric cancer patients with high GFRAL protein expression did not affect the clinical outcomes (Table 3).

To characterize the importance of GFRAL, GFRAL siRNA was used in human gastric cancer cells. Similar to the effects of GDF15 knockdown, knockdown of GFRAL mitigated cell proliferation and decreased migration (Figures 8B–D). Importantly, knockdown of GFRAL sensitized CisR cells to cisplatin (Figure 8E). These results suggest that GDF15 and its receptor, GFRAL, are essential for gastric cancer malignant progression.