The human pancreatic cancer cell lines PANC-1 and MIA PaCa-2 were obtained from the Type Culture Collection of the Chinese Academy of Science (Shanghai, China). Both were cultured in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin-streptomycin at 37°C with 5% CO2.

The clinical characteristics of ELK3 and ZEB1 expression in PDAC patients were analyzed using TMAs containing 70 pairs of pancreatic cancer samples, purchased from Shanghai Outdo Biotech (Shanghai, China). TMAs contain patients’ complete clinicopathological information and follow-up data. The application of TMAs complied with relevant regulations, and present study was approved by the Ethics Committee of Shanghai General Hospital.

After deparaffinization and dehydration, the TMAs were boiled in sodium citrate solution (0.01 M, pH 6.0) for 15 min. Next, 3% hydrogen peroxide was used to block endogenous peroxidase activity. Next, the TMAs were incubated with ELK3 antibody (1:200, Sigma HPA001600) and ZEB1 antibody (1:200, Sigma HPA027524), respectively, at 4°C overnight. The following day, the TMAs were incubated with secondary antibodies for 1 h at room temperature. IHC staining scores were based on staining intensity and staining area. The staining intensity was divided into four levels: 0 (negative staining), 1 (weak staining), 2 (moderate staining), and 3 (strong staining). Staining areas were scored as 0 (0–10%), 1 (10–25%), 2 (25–50%), 3 (50–75%), and 4 (75–100%). The final score was calculated by multiplying the above two scores. Total score of ≤ 4 indicated low expression and > 4 indicated high expression. Assessed by three proficient pathologists independently, the score applied for both ELK3 and ZEB1 expression.

Stable knockdown and overexpression of ELK3 were achieved by construction of lentivirus vector (OBiO Biotechnology, Shanghai, China). PANC-1 and MIA PaCa-2 cells were cultured in 6-well plates. When the cells reached 70% confluence, they were infected with appropriate lentiviruses in the presence of 6μg/ml polybrene (Hanbio Biotechnology, Shanghai, China). Infected cells were selected using 4 μg/ml puromycin (Sigma, United States) for 2 weeks. Transfection efficiency was determined by qRT-PCR and western blotting analysis.

The overexpression vector and small interfering RNAs (siRNAs) specifically targeting genes were synthesized by RiboBio (Guangzhou, China). Transfection of plasmids or siRNAs in pancreatic cancer cells was performed using Lipofectamine^TM2000 (Invitrogen, United States 11668-019) following the manufacturer’s instructions. The transfection efficiency was determined by qRT-PCR and western blotting analysis.

Total RNA was isolated from cells using RNAiso Plus Reagent (TakaRa) and was reverse-transcribed to cDNA using PrimeScript^TM RT reagent kit (TakaRa). Then, SYBR^® Premix Ex Taq^TM (TakaRa) was used to amplify cDNA. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as the internal control to normalize data. The relative gene expression of mRNAs was determined by the 2^–ΔΔCt method.

Pancreatic cancer cells were lysed in RIPA buffer containing protease and phosphatase inhibitors. BCA Protein Assay Kit (Beyotime Biotechnology, China) was used to measure the concentration of protein. Protein samples were separated by SDS-PAGE at 90 V for 2 h and then transfected into PVDF membranes for 2 h. After blocking in 5% fat-free milk for 1.5 h, the membranes were incubated with primary antibodies at 4°C overnight. Next day, the membranes were incubated with secondary antibodies for 2 h. Finally, the membranes washed using TBST and detected by ECL chemiluminescent reagent (Millipore, United States).

Pancreatic cancer cells were cultured in 6-well plates. When the cells grown to full confluence, a scratch wound was made in the center of the well using a 200 μl plastic micropipette tip. Wound healing images were captured at 0 and 24 h after injury. The width of wound healing was quantified and compared with baseline values.

Cancer cells (2–5 × 10⁴) in serum-free medium were seeded into upper chamber of 24-well plates (Corning, United States) with an 8.0 μm pore size polycarbonate membrane without (migration) or with (invasion) matrigel. The medium in the bottom chamber contained 10% FBS as the chemoattractant. After 24 h, the cells in the upper chamber were removed using a cotton swab. The cells that migrated and invaded to lower chamber were stained with 0.1% crystal violet and counted in five random fields under a microscope.

When cells seeded in the confocal plates grown into 50–60% confluence, they were washed with PBS, fixed in 4% paraformaldehyde for 20 min, permeabilized with 0.1% Triton X-100 for 20 min and blocked with 5% goat serum for 1 h at room temperature. The cells were then incubated with primary antibodies at 4°C overnight followed by incubation with fluorophore-conjugated secondary antibody for 1 h. Following washing, the samples were stained with DAPI and imaged using a confocal microscope (Leica, Germany). The primary antibodies were listed as follows: E-cadherin (1:100, Cell Signaling Technology 3195S), N-cadherin (1:100, Cell Signaling Technology 13116S), Vimentin (1:250, Abcam ab92547), and β-catenin (1:250, Abcam ab32572).

ChIP experiments were performed using 1 × 10⁷ pancreatic cancer cells with the ChIP kit (Cell Signaling Technology) according to the manufacturer’s protocol. Briefly, PDAC cells were cross-linked with 1% formaldehyde. Then the cells were lysed and chromatin was harvested and fragmented using enzymatic digestion. The chromatin was immunoprecipitated using ChIP-grade antibody against ZEB1 (Sigma, United States). After immunoprecipitation, the protein-DNA cross-links were reversed and the DNA was purified, followed by qRT-PCR. The enrichment was calculated using the following formula: percentage = 2% × 2^{(C[T]Input Sample – C[T] IP Sample)}. The qRT-PCR products were used for DNA electrophoresis and visualized by ethidium bromide staining.

The luciferase reporter plasmids (pGL3-Luc containing the ELK3 promoter, pGL3-Luc containing intact binding site#2 in the ELK3 promoter and pGL3-Luc containing mutant binding site#2 in the ELK3 promoter) were synthesized by GenePharma (Shanghai, China). PDAC cells were co-transfected with approximately 80 ng of luciferase reporter plasmids and 100 ng of si-ZEB1 or ZEB1 overexpression vectors using Lipofectamine^TM 2000 (Invitrogen, United States). After incubation for 36 h, luciferase activities were detected using the Dual Luciferase Reporter Assay System (Promega, United States) and calculated with the ratio of firefly luciferase/Renilla luciferase activity.

The animal experiments in this study were approved by the Animal Care Committee of Shanghai General Hospital. To study primary tumor growth in vivo, 4-week-old male BALB/c nude mice were chosen and maintained under specific pathogen-free conditions. The mice were randomly divided into four groups (n = 5), and stable cell lines (1.0 × 10⁷) were subcutaneously injected into the armpit of the nude mice. Tumor volumes were estimated in accordance with the formula (length × width²)/2 and measured twice a week. After 21 days, the mice were sacrificed and subcutaneous tumors were removed and weighed. For in vivo lung metastasis model, the nude mice were randomly divided into four groups (n = 5), and 2 × 10⁶ cells were injected through the tail vein. Six weeks later, all of the mice were sacrificed, and the lungs were removed, imaged, paraffin-embedded and stained with hematoxylin and eosin (H&E).

SPSS 22.0 was conducted for statistical analysis. Significant correlations between ELK3/ZEB1 expression and clinicopathological features of PDAC patients were analyzed by Student’s t-test or the Mann-Whitney U-test and Pearson χ²-test. Kaplan-Meier method and log-rank test was used to evaluate survival differences. P < 0.05 was considered statistically significant for all tests.

To clarify the role of ELK3 in PDAC, we first analyzed the expression of ELK3 using the Oncomine database. The mRNA level of ELK3 was significantly upregulated in pancreatic cancer tissues in two datasets (Badea Pancreas Statistics, P = 6.36E-9; Segara Pancreas Statistics, P = 8.46E-5) (Figure 1A). Furthermore, the overexpression of ELK3 in PDAC was confirmed by analyzing two NCBI Gene Expression Omnibus (GEO) datasets (GSE15471, P = 4.78E-10; GSE71987, P = 1.08E-7) (Figure 1B) and The Cancer Genome Atlas (TCGA) dataset (P < 0.05) (Figure 1C). The Kaplan-Meier curve analyses revealed that elevated ELK3 level indicated poorer overall survival (OS) and relapse free survival (RFS) (P < 0.05) (Figure 1D). To further determine the expression level of ELK3, immunohistochemical (IHC) analysis of the tissue microarray (TMA) containing 70 PDAC tissues and corresponding normal tissues were conducted. As shown in Figures 1E,F, the expression level of ELK3 was higher in PDAC tissues than in adjacent normal tissues. Overall, we conclude that ELK3 is frequently elevated in PDAC, and the role of ELK3 remains to be explored.

To investigate the functional roles of ELK3 in PDAC, ELK3 expression was verified by qRT-PCR. The result showed that ELK3 was overexpressed in several pancreatic cancer cells compared with normal pancreatic ductal epithelial cell (HPDE6-c7) (Supplementary Figure 1A). PANC-1 and MIA PaCa-2 cells were selected to knock down and overexpress ELK3, which have been used in our previous experiments (Zhao et al., 2018). We constructed three shRNAs (sh-1, sh-2, sh-3) and a lentiviral overexpression vector targeting ELK3. The qRT-PCR results showed that ELK3 levels were significantly down-regulated or up-regulated in PDAC cells transfected with the indicated shRNAs or overexpression vector, respectively (Supplementary Figure 1B). Among the three shRNAs, sh-1 (sh-ELK3) was selected for further study because it had the highest inhibitory efficiency. In addition, the successful knockdown and overexpression of ELK3 were confirmed at protein levels (Supplementary Figure 1C). Colony formation and EdU assays revealed that knockdown of ELK3 suppressed the proliferation of both PANC-1 and MIA PaCa-2 cells (Figures 2A,B). Conversely, overexpression of ELK3 had the opposite effect on cell proliferation (Figures 2A,B). The wound-healing assay demonstrated that ELK3 depletion inhibited the mobility of PANC-1 cells, while forced expression of ELK3 increased the migration speed of them (Figure 2C). Correspondingly, the effect was confirmed by transwell migration and matrigel invasion assays (Figure 2D). Additionally, similar results were obtained in MIA PaCa-2 cells (Figures 2C,D). In conclusion, our findings indicate that ELK3 may be involved in cell proliferation, migration and invasion, acting as a positive regulator.

To verify the function of ELK3 in pancreatic tumor growth and metastasis in vivo, we injected pancreatic cancer cells with stable knockdown or overexpression of ELK3 into the armpit or tail vein of nude mice. The results showed that tumors in the sh-ELK3/MIA PaCa-2 group grew more slowly than those in sh-NC/MIA PaCa-2 group, and this phenomenon was also reflected by tumor volume and final tumor weight (Figures 3A–C). Additionally, the volume and weight of xenograft tumors in ELK3/PANC-1 group were significantly higher than the control tumors of PANC-1 cells (Figures 3D–F). In the lung metastasis model, we discovered that the number of metastatic nodules in mice injected with sh-ELK3/MIA PaCa-2 cells were less than in mice injected with sh-NC/MIA PaCa-2 cells (Figures 3G,H), while a higher number of metastatic nodules was observed in mice injected with ELK3/PANC-1 cells than in those injected with Ctrl/PANC-1 cells (Figures 3I,J). Taken together, these in vivo results suggest that ELK3 plays an important role in pancreatic tumor growth and metastasis.

As we all know, EMT process plays important roles in cancer cell invasion and tumor metastasis (De Craene and Berx, 2013). TGFβ is a potent inducer of EMT, and TGFβ stimulation can irritate changes in cell morphology and biological behavior (Neuzillet et al., 2015; Su et al., 2020). Our results showed that ELK3 was associated with malignant progression of pancreatic cancer. This prompted us to explore the underlying effects of ELK3 on TGFβ-induced EMT process. Western blot and confocal immunofluorescence analysis indicated that TGFβ treatment markedly decreased E-cadherin and increased N-cadherin and Vimentin expression (Figures 4A,B). However, in sh-ELK3/PANC-1 and sh-ELK3/MIA PaCa-2 cells simultaneously treated with TGFβ, these molecular events induced by TGFβ were abolished by ELK3 depletion (Figures 4A,B). Additionally, ELK3 knockdown also effectively quenched the wound healing, cell migration and invasion abilities induced by TGFβ in PANC-1 and MIA PaCa-2 cells (Figures 4C,D). These results demonstrate that ELK3 is crucial for TGFβ-induced EMT in PDAC cells.

Considering the importance of β-catenin signaling in the development of cancer and EMT process (Ghahhari and Babashah, 2015; Krishnamurthy and Kurzrock, 2018), we were inspired to explore whether ELK3 could regulate the Wnt/β-catenin signaling pathway in pancreatic cancer. To verify the effects of ELK3 on Wnt/β-catenin, we performed western blot assay. The results showed that neither knockdown nor overexpression of ELK3 significantly affected the total β-catenin level (Figure 5A). However, ELK3 depletion decreased the level of nuclear β-catenin and increased the cytosolic β-catenin levels, whereas ELK3 overexpression had the opposite effects on the subcellular location of β-catenin (Figure 5A). TOP-Flash and FOP-Flash luciferase reporters were used to further test the activity of the Wnt/β-catenin signaling pathway. As shown in Supplementary Figure 2, the TOP/FOP luciferase activities in PDAC cells transfected with sh-ELK3 groups were much lower than those in the control group, and were significantly higher in ELK3 overexpressed cells. Furthermore, to determine whether β-catenin is essential for the functions of ELK3, ELK3/PANC-1, and ELK3/MIA PaCa-2 cells were transfected with siβ-catenin. We found that β-catenin suppression dampened ELK3-mediated cell proliferation (Figure 5B and Supplementary Figure 3), wound healing (Figure 5C), migration and invasion (Figure 5D). These data confirmed Wnt/β-catenin signaling pathway plays a vital role in ELK3-mediated pancreatic cancer progression.

To dissect the molecular mechanism of ELK3 overexpression in PDAC, we first explored the genetic or epigenetic dysregulation of ELK3 in pancreatic adenocarcinoma (TCGA, Firehose Legacy) from the cBioPortal database. However, we discovered no evidence regarding the dysregulation of ELK3 at the genetic or methylation levels (Supplementary Figure 4A), suggesting that genetic alterations (mutation, amplification, and deletion) and methylation modification may not be the main causes of the overexpression of ELK3 in PDAC. Then, we would explore its overexpression at the transcriptional level. As one of the most important EMT-inducing transcription factors, ZEB1 not only transcriptionally represses but also activates some EMT-related genes, and its overexpression promotes tumorigenesis and metastasis in human carcinomas (Wellner et al., 2009; Sanchez-Tillo et al., 2011; Krebs et al., 2017). A recent study showed that ZEB1 could collaborate with ELK3 to regulate gene expression (Cho et al., 2019). Thus, we explore whether ZEB1 could transcriptionally activate ELK3 expression in PDAC. Analyzing from the JASPAR database, we found the binding motifs of ZEB1 and five potential ZEB1 binding sites on the ELK3 promoter (Figures 6A,B and Supplementary Figure 4B). QRT-PCR and western blotting analysis demonstrated that forced expression of ZEB1 significantly increased ELK3 mRNA and protein levels, while ZEB1 deletion exhibited an opposite effect (Figures 6C,D and Supplementary Figures 4C,D). ChIP-qPCR results indicated that ZEB1 could interact with ELK3 promoter within the −641 to −631 bp region (Figures 6E,F). In addition, we found a significantly decreased ZEB1 enrichment in the ELK3 promoter following ZEB1 silencing, while ZEB1 overexpression increased the occupancy of ZEB1 in the ELK3 promoter (Figure 6G). To further investigate the regulatory role of ZEB1 on ELK3 transcription, we constructed wild-type (WT) and mutant (Mut) reporter plasmids. For the mutant plasmid, several bases were replaced in the binding site#2, and the wild type reporter contained intact binding site#2 (Figure 6H). Luciferase reporter assays showed that overexpression of ZEB1 could activate the luciferase activity of WT plasmids, but failed to activate Mut reporters (Figure 6I). To sum up, we concluded that ZEB1 binds to the region between −641 and −631 bp of the ELK3 promoter to activate its’ transcriptional activity in pancreatic cancer.

ZEB1 has been shown to promote the proliferation and metastasis of PDAC cells (Krebs et al., 2017). Since ZEB1 could increase ELK3 levels in PDAC, we investigated whether ELK3 was necessary for mediating the effect of ZEB1 on the cellular proliferation and metastasis of PDAC. As shown in Figure 7A and Supplementary Figure 5A, ZEB1-enhanced cell proliferation was inhibited by ELK3 knockdown. Moreover, ZEB1-enhanced cell migration and invasion ability was decreased when ELK3 was knockdown (Figures 7B,C and Supplementary Figures 5B,C). As an EMT-activator, western blot and immunofluorescence results showed that ZEB1 could promote the EMT process of PDAC cells, while this effect was reversed by ELK3 knockdown (Figures 7D,E and Supplementary Figures 5D,E). In summary, these results demonstrated that ELK3 was important for the oncogenic effect of ZEB1 on PDAC progression.

First, we found that ZEB1 expression was also upregulated in GSE15471 (P = 1.75E-07) and GSE71987 (P = 6.79E-06) datasets (Supplementary Figure 5A). Scatter plot analysis showed a positive correlation between the mRNA levels of ZEB1 and ELK3 (GSE15471, R = 0.6792, P < 0.0001; GSE71987, R = 0.8890, P < 0.0001) (Supplementary Figure 5B). In addition, the protein level of ZEB1 was examined in above 70 paired pancreatic cancer tissues and paracancerous tissues, and the results of IHC analysis showed that ZEB1 was highly expressed in pancreatic cancer tissues compared with matched normal tissues (Figures 8A,B). Moreover, nearly 64.3% of pancreatic cancer samples with higher expression of ZEB1 presented stronger ELK3 staining, while approximately 71.2% of those with lower ZEB1 expression exhibited weaker ELK3 staining (Figure 8C). Pearson correlation analysis confirmed the positive correlation between ZEB1 and ELK3 proteins in TMAs (R = 0.848, P < 0.0001) (Supplementary Figure 5C). Based on the median expression of ELK3 or ZEB1 in TMAs, the samples were divided into ELK3 high expression group and ELK3 low expression group or ZEB1 high expression group and ZEB1 low expression group. As shown in Table 1 and Supplementary Figure 5D, ELK3 expression was significantly higher in pancreatic cancer tissues of T3 stage, N1stage, distant metastasis M1 stage and AJCC stage IIB-IV than in these of T1–T2 stage, N0 stage, M1 stage and AJCC-IIA stage, respectively (P < 0.05 for all). ZEB1 expression was positively correlated with pathological grade, N stage and AJCC stage (P < 0.05 for all, Table 2 and Supplementary Figure 5E). Kaplan-Meier survival analysis showed that patients with higher ELK3 and ZEB1 expression levels were both associated with worse overall survival (OS) (Figures 8D,E). Moreover, the combination of these two elements demonstrated that pancreatic cancer individuals with the expression of ZEB1^highELK3^high had an even worse OS rate than any other groups (P = 0.0014) (Figure 8F). Taken together, ZEB1 and ELK3 predicted poor survival in clinical samples and may be indicators of efficient prognostic factors in PDAC patients.