Inhibition of ERK1/2 in cancer-associated pancreatic stellate cells suppresses cancer–stromal interaction and metastasis

We obtained PDAC specimens from patients who underwent pancreatectomy for at our institution. KPC (LSL-Kras G12D/+; LSL-Trp53R172H/+; Pdx-1-Cre) transgenic mice were described previous [18]. Tissues were embedded, sliced, stained and observed using an optical microscope (BZ-X710; Keyence, Osaka, Japan).

We blocked endogenous peroxidase activity with methanol containing 0.3% hydrogen peroxidase. Antigen retrieval was performed by boiling samples in a microwave oven (citrate buffer, pH 6.0). Human pancreatic tissues and KPC mice tissues were sliced to 4 μm, incubated with rabbit anti-phospho-ERK1/2 (#4370, Cell Signaling, Technology, Danvers, MA, USA) overnight at 4 °C and stained with EnVisionþ System-HRP Labeled Polymer Anti-Rabbit (#K4003; Dako, Glostrup, Denmark). The staining was performed using serial sections.

We used the following PCC lines: AsPC-1, CFPAC-1(American Type Culture Collection, Manassas, VA, USA), Panc-1 (Riken BioResource Center, Ibaraki, Japan), SUIT-2 (Japan Health Science Research Resources Bank, Osaka, Japan), and BxPC-3 (National Kyushu Cancer Center, Fukuoka, Japan). All PCCs were maintained in DMEM (Sigma Chemical Co., St. Louis, MO, USA) supplemented with 10% FBS at 37 °C with humidified 90% air and 10% CO₂. Human pancreatic ductal epithelial (HPDE) cells were obtained from Dr. M.-S. Tsao (University of Toronto, Canada) and maintained in HuMedia-KG2 medium (KK-2150S Kurabo, Osaka, Japan). We established human PSCs from fresh pancreatic cancer surgical specimens using the outgrowth method [19‐21], as described in our previous reports. The isolated cells were confirmed to be PSCs by their spindle-shaped morphology, and immunofluorescence staining for αSMA, vimentin, CD90, glial fibrillary acidic protein, and nestin, but not CK19 [22, 23]. They were used within eight passages for each assay. Immortalization of PSCs was conducted as described previous [23]. All PSCs were maintained in DMEM (Sigma-Aldrich Co., Tokyo, Japan) supplemented with 10% fetal bovine serum, streptomycin (100 mg/ml), and penicillin (100 mg/ml) at 37 °C in a humidified atmosphere containing 10% CO₂. HPaSteC cells (#3830; ScienCell Research Laboratories, Carlsbad, CA, USA) were maintained according to the manufacturer’s instructions using Stellate Cell Medium (#5301; ScienCell). PCCs from primary tumors in KPC mice were established using an outgrowth method [19], and isolated cancer cell lines were maintained as described [24]. ERK1/2 inhibitor used in vitro (S7101, Selleck Chemicals, Houston, TX, USA) and in vivo (HY-50846, MCE, NJ, USA) were reconstituted following the manufacturer’s recommendations and used at the indicated doses. Chloroquine phosphate was purchased from Sigma-Aldrich (#PHR1258), dissolved in phosphate-buffered saline to 10 mM, and stored at ⁻20C until used.

PDAC organoids were established from KPC PCCs, which were established using the outgrowth method as described [21]. PCCs were embedded in growth factor-reduced Matrigel (Cat#356231; BD Bioscience, CA, USA), and cultured in human complete medium at 37 C° for 14 days [21, 25]. Human complete medium was AdDMEM/F12 (Cat#12634–010; Invitrogen, CA, USA), medium supplemented with 1 M HEPES (Invitrogen), GlutaMax (Cat#35050–061; Invitrogen), penicillin/streptomycin (Cat#15140122; Invitrogen), B27 (Cat#17504044; Invitrogen), N-acetyl-l-cysteine (Cat#A9165; Sigma-Aldrich Co.), Wnt-3a (Cat#5036-WN-010; R&D Systems, MN, USA), R-Spondin 1 (Cat#120–38; Peprotech, NJ, USA), Noggin (Cat#120-10C; Invitrogen), epidermal growth factor (EGF, Cat#AF-100-15; Peprotech), fibroblast growth factor (FGF, Cat#100–26; Peprotech), nicotinamide (Cat#N0636; Sigma-Aldrich Co.), Y-27263 (Cat# Y0503; Sigma-Aldrich Co.) and A83–01 (Cat#2939/10; R&D Systems).

Total RNA was extracted from cultured cells using a High Pure RNA Isolation Kit (Roche Diagnostics, Mannheim, Germany) and DNase I (Roche Diagnostics, Sigma-Aldrich), according to the manufacturers’ instructions. We performed qRT-PCR using a QuantiTect SYBR Green Reverse Transcription-PCR kit (Qiagen, Tokyo, Japan) and a CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA). Our specific primer sequences were purchased from Sigma-Aldrich (Tokyo, Japan). Primer sequences are listed in Additional file 5: Table S1. We normalized mRNA expression levels to 18S rRNA levels.

The invasiveness and migration capacities of PCCs were assessed by determining the number of cells invading or migrating across transwell chambers as previously described [24, 26]. For invasion assays, PCCs (1 × 10⁵ cells/250 μl) were seeded in the upper transwell chamber (8-μm pore size; Becton Dickinson, Franklin Lakes, NJ) containing 100 mL of reconstituted Matrigel-coated membrane (20 mg/well, BD Biosciences, Bedford, MA). Thereafter, cells were incubated for 36–48 h and the number of invading PCCs was counted. Cell migration assays were performed with PCCs using the same protocol as the invasion assay without a Matrigel-coated membrane. Cells were allowed to migrate, and were counted 18–24 h after cell seeding into the upper chamber. In co-culture experiments, PSCs were seeded in 24-well plates (#353504; Corning) at a density of 5 × 10⁴ cells/well. At 24 h after seeding, medium was replaced with 750 μL of fresh DMEM containing 10% FBS. Transwell chambers (8-μm pores; Becton Dickinson) were placed in 24-well dishes, and then PCCs, which had been suspended in 250 μL of the same medium (1 × 10⁵ cells/mL) were seeded into the transwell chambers. After incubation for the indicated time, migration and invasion were evaluated by counting the cells that had invaded to the lower chamber. In both assays and at each time point, invaded or migrated cells at the bottom of the chamber were fixed with 70% ethanol, stained with hematoxylin and eosin, and counted in 5 random fields at 100× magnification (BZ-X710; Keyence Corporation, Osaka, Japan). Each experiment was performed in triplicate and repeated at least three times.

Western blotting was performed as described previously [27]. Cells were lysed in Pro-Prep (InTron Biotechnology, Seongnam, Korea) and proteins were separated on 4–15% Mini-Protean TGX Precast Gels (Bio-Rad) and transferred to Trans-Blot Turbo Mini PVDF Transfer Packs (Bio-Rad) using a Trans-Blot Turbo Transfer Starter System (Bio-Rad). Membranes were incubated overnight at 4–8 °C with anti-ERK1/2 (#4695, Cell Signaling Technology, Danvers, MA, USA), anti-phospho-ERK1/2 (#4370, Cell Signaling Technology), anti-αSMA (#M0851; Dako, Japan), anti-E-cadherin (#3195S, Cell Signaling Technology), anti-vimentin (#5741, Cell Signaling Technology), anti-collagen type I (sc-8783 Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-collagen type VI (sc-47,712, Santa Cruz Biotechnology), anti-MMP2 (#13132, Cell Signaling Technology), anti-MMP3 (#sc-6839, Santa Cruz Biotechnology), anti-MMP14 (AB8345, Millipore, Temecula, CA, USA), anti-IL-6 (#ab9324, Abcam, Cambridge, MA, USA), anti-p16 (MABE1328, Millipore), anti-p15 (MABE1664, Millipore), anti-CC8 (#9496S, Cell Signaling Technology), anti-fibronectin (sc6952, Santa Cruz Biotechnology), anti-LC3 (#2775S; Cell Signaling Technology), anti-AKT (#4060S, Cell Signaling Technology), anti-phospho-AKT (#4691S, Cell Signaling Technology), and anti-b-actin (ab8227; Abcam), and then probed with appropriate secondary antibodies (Cell Signaling Technology). Immunoblot signals were detected by enhanced chemiluminescence with ChemiDoc XRS (Bio-Rad).

Cells (1 × 10³ cells/well) were seeded in 96-well plates (Greiner Bio-One, Frickenhausen, Germany) and cell viability examined using the CellTiter-Glo Luminescent Cell Viability Assay Kit (G7570, Promega) following the manufacturer’s instructions. Background was subtracted using values from wells containing only culture medium.

BALB/c AJcl nu/nu female mice were purchased from Clea (Tokyo, Japan) and transported to our institution at 4 weeks old. After 1 week of acclimation, each of the 20 nude mice were splenic implanted with 5 × 10⁴ PDAC organoids and 5 × 10⁴ PSCs and randomized into four groups for treatment with PBS (control), ERK1/2 inhibitor (25 mg/kg dissolved in 100 μL PBS) alone, CQ (50 mg/kg dissolved in 100 μL PBS) alone, or with a combination of ERK1/2 inhibitor and CQ. One week after implantation, mice were treated intraperitoneally with either vehicle, SCH772984, Chloroquine or combination according to the dosing schedule indicated in the figure legends. The mouse liver metastasis tissues were fixed in formaldehyde, embedded in paraffin, and cut into 4-μm-thick sections. All mouse experiments were approved by the Ethics Committee of Kyushu University.

For in vitro experiments, values are expressed as mean ± standard deviation. Comparisons between two groups were made using Student’s t-test. P < 0.05 was considered significant.

Three consecutive rounds of in vivo selection were performed by using splenic xenografts of PDAC cell lines SUIT-2 and AsPC-1 cells, from which metastatic lesions were harvested to establish metastatic SUIT-2 (SLMS) and AsPC-1 (SLMA) cells (Fig. 1a). We then investigated in vitro characteristics of the SLMS and SLMA cells. The SLMS cells had an apparent spindle-shaped morphology compared with their parental SUIT-2 cells (Fig. 1b). Migration, invasion and proliferation capacities of highly metastatic (HM) PCCs were significantly greater than their parental PCCs (Fig. 1c, d; Additional file 1: Figure S1A). We confirmed that SLMS cells occurred in liver metastases more frequently than did their parental SUIT-2 cells in vivo (Fig. 1e; Additional file 1: Figure S1B). As we had verified the upregulated aggressiveness of metastatic cells, we compared results of phospho-kinase array analysis between SLMS cells and parental SUIT-2 cells, and found altered expression of several phosphorylation kinases HMPCCs, including upregulation of p53 (S15), p53 (S46), AKT (S473), ERK1/2 (T202/Y204), AMPKa1 (T183), and downregulation of p70S6 kinase (T241/S424) and AMPKa2 (T172) (Additional file 1: Figure S1C). The greatest increase in phosphorylation level was seen in ERK1/2 (T202/Y204). To validate the accuracy of array data, we used western blotting to evaluate expression of p-ERK1/2 in HMPCCs and parental cells. We also found the mesenchymal marker vimentin was upregulated in HMPCCs (Additional file 1: Figure S1D). These data indicated that established HMPCCs were more aggressive and tumorigenic than their parent cells.

In previous reports, p-ERK1/2 expression in PCCs was demonstrated in pancreatic cancer tissues. Using database of Human Protein Atlas (available from www.protenatlas.org), we found expression of ERK1 and/or ERK2 in both tumor and stromal cells (Additional file 2: Figure S2A, B). However, analysis of The Cancer Genome Atlas [28, 29] shows expression of ERK2, but not ERK1, correlates with poorer overall and disease-free survival in PDAC (Additional file 2: Figure S2C, D). We therefore investigated p-ERK1/2 expression in PDAC samples obtained in our institution (Additional file 6: Table S2) and detected p-ERK1/2 on both tumor cells and stromal cells (Fig. 2a). We also assessed p-ERK1/2 expression in PDAC derived from our KPC mouse model (LSL-KrasG12D/+; LSL-Trp53R172H/+; Pdx-1-Cre) and found p-ERK1/2 expression in primary tumor cells, stromal cells and liver metastases (Fig. 2b). Next, we investigated the p-ERK1/2 expression of various pancreatic cells including HPaSteC cells (normal PSC cells), cancer-associated PSCs, HPDE cells and PCCs. PSCs showed high p-ERK1/2 expression, even compared with PCCs (Fig. 2c). Interaction between PCC and PSC is a key process in pancreatic cancer progression [30]. To validate the involvement of p-ERK1/2, we co-cultured PCCs and PSCs, using the transwell system. Compared with monocultured cells, p-ERK1/2 expression was significantly upregulated in PCCs when co-cultured with PSCs (Fig. 2d). Previous study showed that PSCs can promote migration, invasion and EMT process via the regulation of E-cadherin and vimentin expression in PDAC cells [31]. We also found that cell migration and invasion of AsPC-1 and SUIT-2 were enhanced when indirectly co-cultured with PSCs (Fig. 2e and f).

Next, we investigated the effect of ERK inhibition on PCCs and PSCs. We chose a novel ERK1/2-selective pharmacologic inhibitor, SCH772984, which has shown cellular potency in tumor cells with BRAF, NRAS, or KRAS mutations and induces tumor regressions in xenograft models at toxicity-free doses [32]. First, we examined the effects of SCH772984 on viability of parental PCCs and HMPCCs. The IC₅₀ values of SCH772984 on AsPC-1 and SUIT-2 cells were 1291 nM and 1180 nM, respectively (Fig. 3a, b), compared with 424.2 nM and 847.7 nM, respectively, for HM SLMA and SLMS cells, which indicates HMPCCs are more sensitive to ERK1/2 inhibitor (Fig. 3c, d). As expression of p-ERK1/2 in PDAC is reportedly related to EMT [33], we investigated changes in kinase phosphorylation in HMPCCs after ERK1/2 inhibition. Upregulation of the epithelial cell marker, E-cadherin, and downregulation of the mesenchymal marker, vimentin, were observed through western blotting (Fig. 3e), which indicates that inhibiting p-ERK1/2 leads to suppression of EMT in HMPCCs.

As high expression of p-ERK1/2 was only detected in PSCs (Fig. 2c), we hypothesized inhibiting ERK1/2 in PSCs would be more efficient than in PCCs. We established immortalized PSCs from a pancreatic cancer specimen obtained at our institution [34]. We observed a change from spindle-like shapes to round shapes among these PSCs after 72 h of SCH772984 treatment (Fig. 4a). The two primary cultures of PSCs were more sensitive to SCH772984, with IC₅₀ values of 321 nM and 89 nM, respectively, compared with the HMPCCs (Fig. 4b, c). When we investigated changes in expressions of related cytokines and chemokines after SCH772984 treatment, we found senescence marker p15, p16, fibrosis marker α-SMA, fibronectin, Collagen Type I and Collagen Type IV were upregulated; and MMP2, MMP3, IL-6 (which are related to cell invasiveness and malignancy) were downregulated (Fig. 4d, e). These data are consistent with the results of the previous study, which showed that p16 induces cellular senescence and stable growth arrest without a senescence-associated secretory phenotype [35]. As inhibition of CDK4/6, a downstream target of ERK1/2, reportedly upregulated drug-induced autophagy in breast cancer [36], we investigated the effect of ERK inhibition on autophagy in PSCs. We found that autophagy marker LC-3II protein expression was upregulated. Our results suggest that inhibition of ERK did not induce the reversion of PSC from activated phenotype to quiescent type, but to cellular senescence, which may be another activated phenotype.

Because PCCs and PSCs showed similar reactions to SCH772984, we investigated the effect of SCH772984 on cancer–stromal interaction. Using a transwell indirect co-culture system (Fig. 5a), we found SCH772984 did not reduce PCC migration or invasion capacity when treated with the lower IC50 dose for PSCs. However, this lower dose of SCH772984 inhibited PCC migration and invasion when co-cultured with PSCs (Fig. 5b, f). In addition, a direct co-culture cell viability assay revealed SCH772984 suppressed proliferation of PCCs and PSCs co-cultured at the lower PSC IC50 dose (Fig. 5c-e and g-i). However, we did not obtain a similar finding for monocultured PCCs. These results suggest that treatment with SCH772984 could preferentially target PSCs to suppress cancer–stromal interaction.

To evaluate therapeutic efficiency, we combined SCH772984 with an effective autophagy inhibitor CQ, which is shown to suppress PSCs activation via inhibition of autophagy activity [23]. A further morphology change was observed in PSCs after the combined treatment (Fig. 6a; Additional file 3: Figure S3a). Compared with a 4-fold increase in viability of control PSCs, the combined treatment remarkably restricted proliferation of PSCs (Fig. 6b; Additional file 3: Figure S3b); It also downregulated α-SMA and Collagen Type I expression, and upregulated senescence markers p15 and p16, compared with SCH772984 treatment alone (Fig. 6c; Additional file 3: Figure S3c). Primary cultured PSCs partly exhibited positive beta-galactosidase staining as shown in the previous reports [37]. We performed senescence β-galactosidase staining to investigate the correlation between p15/p16 expression and autophagy during PSC cellular senescence. We found SCH772984 treatment did not increase cellular senescence in PSCs, possibly due to autophagy activation, although p15/p16 expression was upregulated, and combined SCH772984 + CQ remarkably induced PSC cellular senescence (Fig. 6d; Additional file 3: Figure S3d). Therefore, combining SCH772984 with CQ suppressed drug-induced autophagy of SCH772984 and led to cellular senescence in PSCs.

Next, we evaluated the therapeutic effect of combine treatment of SCH772984 with CQ in PCCs. Single agent and/or combine treatment of SCH772984 and CQ didn’t affect the migration capacity of PCCs. SCH772984, but not CQ, significantly suppressed invasion of PCCs. And this effect was not further promoted by combine treatment with CQ. However, this effect was not further promoted by combine treatment with CQ (Fig. 6e; Additional file 3: Figure S3e). In addition, cell viability of PCCs was significantly suppressed by single agent of ERK inhibitor or CQ, and combination treatment further suppressed cell viability compared to single agent (Fig. 6f; Additional file 3: Figure S3f). Furthermore, the EMT transition was downregulated of PCCs following SCH772984 alone treatment and/or combination treatment. PCCs didn’t upregulated the cellular senescence markers p15 and p16 unlike PSCs. Instead, we observed an induction of CC8 expression, indicating combination treatment induced significant apoptosis of PCCs (Fig. 6g; Additional file 3: Figure S3g).

ERK1/2 inhibitor was identified as a potential therapeutic agent for primary liver cancer in organoid xenograft experiments [38]. A pancreatic tumor organoid was shown to recapitulate the histology and gene expression of its parental tumor [25, 39]. We therefore generated cancer organoids for 2 weeks using KPC-derived PDAC cells as described previous (Fig. 7a) [21]. Organoid or 2D-cultured cells were then splenic transplanted into nude mice with KPC-derived PSCs. Consistent with previous reports [17, 32], SCH772984 treatment did not cause any toxic effects such as body weight loss. Compared to 2D-cultured cells, there was no significant increase of liver metastasis in splenic xenograft experiment using organoid. However, the implanted tumor derived from organoids frequently restored high p-ERK1/2 expression, which was consistent with the original tumor (Additional file 4: Figure S4). Because of these results, we chose organoid xenografts for the following experiments. KPC cancer organoids were splenic co-transplanted into nude mice with KPC-derived PSCs. One week after implantation, mice were treated intraperitoneally with either vehicle, SCH772984, CQ or combination according to the dosing schedule indicated in the figure legends (Fig. 7b). After 13 days of treatment, their liver metastases were harvested and evaluated (Fig. 7c). Compared with controls, the combined treatment group remarkably decreased the metastatic nodules (14 vs. 2, average) in the liver (Fig. 7d), liver volume (2.53 vs. 1.29cm³, average) and liver weight (1.65 g vs. 1.29 g, average), although SCH772984 alone and CQ alone also decreased metastatic nodules (14 vs. 5, 14 vs. 6, average) (Fig. 7e, f). Histologic analysis by using serial sections demonstrated that Ki67 expression was downregulated in the combined treatment (6% positive) group compared with control (44% positive) group and/or single-agent treatment (18% positive for SCH772984 and 36% positive for CQ) groups (Fig. 7g). The corresponding rectangles indicated reduction of p-ERK1/2 expression in a-SMA-positive PSCs compared with controls. In addition, masson’s trichrome stain demonstrated significant decrease in the expression of collagen fibers (7% positive) in combined treatment group compared with control (12% positive) group and/or single-agent treatment (22% positive for SCH772984 and 14% positive for CQ) groups (Fig. 7g).