The human PC cell lines ASPC-1 (CRL-1682) and BXPC-3 (CRL-1687) were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). Twist (sc-38604-V) and GDF15 (sc-39798-V) shRNA lentiviral particles, control shRNA lentiviral particles-A (cat. no. sc-108080) and mouse anti-human monoclonal Twist antibody (Twist2C1a; cat. no. sc-81417), mouse anti-human GDF-15 monoclonal antibody (G-5; cat. no. sc-377195) and mouse anti-human matrix metalloproteinase-2 monoclonal antibody (MMP-2; cat. no. sc-53630) were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA). The SensoLyte 520 MMP-2 Assay kit (cat. no. AS-71151) was purchased from AnaSpec (Fremont, CA, USA). The QCM ECMatrix 24-well (8 μM) Fluorimetric Cell Invasion Assay kit (cat. no. ECM554) was purchased from Chemicon (Millipore, Billerica, MA, USA). Superfect transfection reagent was purchased from Qiagen (Valencia, CA, USA). The p38 mitogen-activated protein kinase (MAPK) Assay kit (cat. no. 9820) was purchased from Cell Signaling Technology, Inc. (Beverly, MA, USA). Human Twist cDNA was subcloned into a pcDNA 3.1 expression vector (16). Full-length human GDF15 cDNA (MCG: 4145) vector was purchased from Invitrogen Life Technologies (Carlsbad, CA, USA). The human GDF15 expression vector (pcDNA3-GDF15) was constructed by subcloning the GDF15 cDNA vector following digestion with EcoRI and NotI into the pcDNA3.1 expression vector (Invitrogen Life Technologies). Puromycin, G418, cisplatin, the selective p38 MAPK inhibitor SB203580 and all chemicals of reagent grade were purchased from Sigma-Aldrich (St. Louis, MO, USA).

The Twist or GDF15 expression vector was transfected into cells using Superfect transfection reagent (Qiagen) according to the manufacturer's instructions. Pools of stable transductants were generated via selection with G418 (600 μg/ml) according to the manufacturer's instructions. The Twist or GDF15 shRNA lentiviral particles contain expression constructs encoding target-specific 19–25 nt (plus hairpin) shRNA designed to specifically knock-down Twist or GDF15 gene expression. The control shRNA lentiviral particles contain a scrambled shRNA sequence that does not lead to specific degradation of any cellular mRNA and was used as a negative control. Lentiviral transduction was performed in ASPC-1 and BXPC-3 cells. Pools of stable transductants were generated via selection with puromycin (4 μg/ml) according to the manufacturer's instructions (Santa Cruz Biotechnology, Inc.).

RNA was prepared from cells using TRIzol reagent followed by purification using a TURBO DNA-free kit (Ambion, Austin, TX, USA). The cDNAs were synthesized using SuperScript II reverse transcriptase (Invitrogen Life Technologies). qPCR was performed on the LightCycler thermal cycler system (Roche Diagnostics, Indianapolis, IN, USA) using an SYBR Green I kit (Roche Diagnostics) according to the manufacturer's instructions. The results were normalized against that of the housekeeping gene glyceral-dehyde-3-phosphate dehydrogenase (GAPDH) in the same sample. The primers used are as follows: Human GDF15, forward 5′-CGGTGAATGGCTCTCAGATG-3′ and reverse 5′-CAGGTCCTCGTAGCGTTTCC-3′; human GAPDH, forward 5′-GACTCATGACCACAGTCCATGC-3′ and reverse 5′-AGAGGCAGGGATGATGTTCTG-3′. Each experiment was repeated three times in duplicate.

In vitro cell invasion assays were performed with the QCM ECMatrix 24-well (8 μM) Fluorimetric Cell Invasion Assay kit (Chemicon; Millipore) according to the manufacturer's instructions (17,18). The kit used an insert polycarbonate membrane with an 8 μM pore size. The insert in the invasion kit was coated with a thin layer of ECMatrix. Cell invasion was determined by fluorescence. Each experiment was repeated three times in duplicate.

Cells were dissolved in 250 μl of 1X SDS loading buffer (62.5 mm TrisHCl, pH 6.8, 2% SDS, 25% glycerol, 0.01% bromphenol blue, 5% 2-mercaptoethanol) and incubated at 95°C for 10 min. Equal quantities of proteins for each sample were separated by 10% SDS-polyacrylamide gel and blotted onto a polyvinylidene difluoride microporous membrane (Millipore). Membranes were incubated for 1 h at room temperature with a 1:500 dilution of the following primary antibodies: Anti-Twist, anti-GDF-15, and anti-MMP-2, and then washed and revealed using horseradish peroxidase-conjugated bovine anti-mouse secondary antibodies (cat. no. sc-2371; Santa Cruz Biotechnology, Inc., 1:5,000, 1 h). Peroxidase was revealed with a GE Healthcare ECL kit (Shanghai, China). Three independent experiments were performed for each western blot analysis.

MMP-2 activity was measured with the SensoLyte 520 MMP-2 Assay kit (AnaSpec) according to the manufacturer's instructions (19,20). The supernatants were collected and then incubated with 4-aminophenylmercuric acetate and MMP-2 substrate. The fluorescence intensity at Ex/Em wavelengths of 490 nm/520 nm were used as a measure of MMP-2 activity. Each experiment was repeated three times in duplicate.

Cells were plated in triplicate in 96-well plates at a density of 5,000 cells. After 24 h of incubation, the medium was replaced by fresh medium with or without various concentrations of cisplatin (0.1, 0.25, 0.5, 1.0, 1.5, 3.0, 6.0, 15.0, 30.0, or 55.0 mM) (Sigma-Aldrich). Subsequently, cell viability was assayed 48 h later using a modified MTT assay as previously described (21). The half maximal inhibitory concentration (IC50) values were defined as the concentrations resulting in a 50% reduction in growth compared with control cell growth.

p38 MAPK activity was measured using the p38 MAPK Assay kit (Cell Signaling Technology, Inc.) according to the manufacturer's instructions (22). Briefly, cells were directly lysed in the culture dishes. Cell lysates were sonicated and centrifuged at 20,000 x g for 10 min at 4°C. The supernatant containing equivalent quantities of protein (200 μg) was incubated by gentle rocking with 20 μl of immobilized mouse anti-human phospho-p38-MAPK monoclonal antibody (28B10; cat. no. 9216; Cell Signaling Technology, Inc.; 1:500) for 16 h at 4°C. The immunoprecipitates were washed twice with the lysing buffer and pelleted by centrifu-gation at 20,000 × g for 10 min at 4°C. The p38 MAPK assay was performed using activating transcription factor 2 (ATF2) fusion protein (2 μg) as a substrate in the presence of 200 μM ATP and 1X kinase buffer according to the manufacturer's instructions. Samples were resolved on a 12% SDS-PAGE gel and visualized by autoradiography.

Statistical analyses were performed with SPSS for Windows 19.0 (SPSS, Inc., Chicago, IL, USA). All data values are expressed as the mean ± standard deviation. Comparisons of means among multiple groups were performed with one-way analysis of variance followed by post hoc pairwise comparisons using Tukey's test. P<0.05 was considered to indicate a statistically significant difference.

To investigate the functional interaction between Twist and GDF15 in PC cells, Twist and GDF15 were stably overexpressed in ASPC-1 and BXPC-3 human PC cells by stable transfection. By contrast, the cells were also stably transduced with lentiviral shRNAs to knock down Twist and GDF15, respectively. As shown in Fig. 1, Twist and GDF15 were constitutively expressed in ASPC-1 and BXPC-3 cells. Compared with the controls, Twist was overexpressed >4.8 fold and knocked down >80% in ASPC-1 and BXPC-3 cells, respectively; GDF was overexpressed >4.2 fold and knocked down >80% in ASPC-1 and BXPC-3 cells, respectively. GDF15 expression in the cells increased (by >3.3 fold) and decreased (>60%) in parallel with Twist overexpression and knockdown, respectively. By contrast, overexpression and knockdown of GDF15 had no significant effect on Twist expression (Fig. 1). Our pilot study suggested that Twist regulates GDF15 expression in PC cells by a p38 MAPK-dependent mechanism (data not shown). Therefore, a selective p38 MAPK inhibitor SB203580 (10 μM) was included in all experiments in the present study (23). As shown in Fig. 1, the p38 MAPK inhibitor had no significant effect on the constitutive expression level of Twist, whereas it eradicated Twist-induced GDF15 expression in PC cells. RT-qPCR assays revealed a similar data trend (Fig. 2), suggesting that Twist regulates GDF15 expression at the mRNA level.

To examine the individual effect of and interaction between Twist and GDF15 on PC cell invasion, in vitro cell invasion assays were performed. Compared with the controls, overexpression of Twist increased cell invasion by 3.4 and 2.1 fold in ASPC-1 and BXPC-3 cells, respectively, which was eradicated by knockdown of GDF15 or SB203580 (10 μM; Fig. 3). By contrast, knockdown of Twist decreased cell invasion by ~60% in ASPC-1 and BXPC-3 cells, respectively, which was completely reversed by overexpression of GDF15 (Fig. 3). Compared with the controls, overexpression of GDF15 increased cell invasion by 3.7 fold in ASPC-1 cells and 2.3 fold in BXPC-3 cells, whereas knockdown of GDF15 decreased cell invasion by ~85% in the two cell lines (Fig. 3).

MMPs have a critical role in cancer cell invasion (24). Among the different MMPs assessed, MMP-2 expression was found to be significantly altered by Twist and GDF15 in PC cells (data not shown). Compared with the controls, overexpression of Twist increased MMP-2 expression by 3.9 and 2.5 fold in ASPC-1 and BXPC-3 cells, respectively, which was eradicated by knockdown of GDF15 or SB203580 (10 μM; Fig. 4). By contrast, knockdown of Twist decreased MMP-2 expression by 51% in ASPC-1 and 43% in BXPC-3 cells, which was completely reversed by overexpression of GDF15 (Fig. 4). Compared with the controls, overexpression of GDF15 increased MMP-2 expression by 4.1 fold in ASPC-1 cells and 2.5 fold in BXPC-3 cells, whereas knockdown of GDF15 decreased MMP-2 expression by ~80% in the two cell lines (Fig. 4). A similar data trend was observed with MMP-2 activity (Fig. 5).

To examine the individual effect of and interaction between Twist and GDF15 on PC chemoresistance, cisplatin IC50 values were examined in PC cells. A higher IC50 value was considered to correspond with clinical chemoresistance to cisplatin. As shown in Fig. 6, after 48 h of cisplatin treatment, the cisplatin IC50 values for ASPC-1 and BXPC-3 cells were 5.6 and 6.1 μM, respectively. Overexpression of Twist significantly increased the IC50 values to 20.5 and 12.7 μM, respectively, which was eradicated by knockdown of GDF15 or SB203580 (10 μM; Fig. 6). By contrast, knockdown of Twist decreased the IC50 values to 2.3 and 3.4 μM, which was completely reversed by overexpression of GDF15 (Fig. 6). Overexpression of GDF15 increased the IC50 values of ASPC-1 and BXPC-3 cells to 23.2 and 15.4 μM, respectively, while knockdown of GDF15 decreased the IC50 values to 1.3 and 2.1 μM, respectively (Fig. 6).

The above results suggested that Twist promotes PC cell invasion and chemoresistance to cisplatin largely through regulating GDF15 expression by a p38 MAPK-dependent mechanism. Therefore, the individual effect of and interaction between Twist and GDF15 on p38 MAPK activity was next examined, which was measured by phosphorylation of ATF2, a substrate of activated p38 MAPK (22). As evidenced by increased levels of phosphorylated ATF2, overexpression of Twist induced p38 MAPK activity by 4.2 and 3.9 fold in ASPC-1 and BXPC-3 cells, respectively, which was eradicated by SB203580 (10 μM) but not knockdown of GDF15 (Fig. 7). By contrast, knockdown of Twist decreased p38 MAPK activity by ~70% in ASPC-1 and BXPC-3 cells, which was not significantly affected by overexpression of GDF15 (Fig. 7). Compared with the controls, overexpression and knockdown of GDF15 demonstrated no significant effect on p38 MAPK activity (Fig. 7).