Ganglioside GM2, highly expressed in the MIA PaCa-2 pancreatic ductal adenocarcinoma cell line, is correlated with growth, invasion, and advanced stage

Gangliosides, a group of glycosphingolipids, are known to be cell surface markers and functional factors in several cancers. However, the association between gangliosides and pancreatic ductal adenocarcinoma (PDAC) has not been well elucidated. In this study, we examined the expression and roles of ganglioside GM2 in PDAC. GM2+ cells showed a higher growth rate than GM2− cells in the adherent condition. When GM2– and GM2+ cells were cultured three-dimensionally, almost all cells in the spheres expressed GM2, including cancer stem cell (CSC)-like cells. A glycolipid synthesis inhibitor reduced GM2 expression and TGF-β1 signaling in these CSC-like cells, presumably by inhibiting the interaction between GM2 and TGFβ RII and suppressing invasion. Furthermore, suppression of GM2 expression by MAPK inhibition also reduced TGF-β1 signaling and suppressed invasion. GM2+ cells formed larger subcutaneous tumors at a high incidence in nude mice than did GM2– cells. In PDAC cases, GM2 expression was significantly associated with younger age, larger tumor size, advanced stage and higher histological grade. These findings suggest that GM2 could be used as a novel diagnostic and therapeutic target for PDAC.

There were no notable morphological differences between GM2-and GM2+ cells in adherent culture conditions. To compare the characteristics of GM2-and GM2+ cells, we sorted MIA PaCa-2 based on GM2 expression level. FACS-reanalysis of sorted cells showed that the fraction of GM2+ cells in cells sorted from GM2 negative or positive populations were approximately 0% (GM2-populations) or 95% (GM2+ populations), respectively (Fig. 2a). These reanalyzed results confirm that the GM2-and GM2+ cells were well GM2+ cells in adherent culture conditions exhibit high growth rates and are highly sensitive to anti-cancer drugs. We also investigated the functional properties of GM2+ cells. The rate of GM2+ cell growth in adherent conditions was significantly higher than that of GM2-cells (Fig. 3a). The effects of three commonly used anti-pancreatic cancer drugs, gemcitabine, 5-FU, and abraxane, on GM2+ cells were evaluated to determine the CSC properties of the cells. CSCs possess an effective efflux pathway that confers resistance to anticancer drugs. The survival rates of GM2+ cells were lower than those of GM2-cells after treatment with the anticancer drugs at either 10 or 100 μM (Fig. 3b). Furthermore, we examined the levels of expression of four genes encoding potential anticancer drug transporters. Real-time PCR analysis revealed that expression of ABCG2, ABCC1, and ABCC2 were not significantly different between GM2-and GM2+ cells (Fig. 3c). We further examined stemness of GM2+ cells using real-time PCR analysis of CSC markers. Of the markers assayed, only CD24 had higher levels of expression in GM2+ cells than in GM2-cells, while Nestin was lower in GM2+ cells (Fig. 3d). Another method commonly used to examine CSC characteristics, especially self-renewal ability under the floating condition 4 , is the sphere formation assay. ATP assays showed that the number of cells in the spheres was not different in GM2 + and GM2-cells (Fig. 3e), indicating no differences in sphere-forming capability between the www.nature.com/scientificreports www.nature.com/scientificreports/ two types of cells. Hence, GM2+ cells in adherent culture conditions exhibited high growth rates and were highly sensitive to anti-cancer drugs but did not have remarkable stem cell characteristics compared with GM2-cells.
GM2 is expressed in CSC-like sphere cells induced by 3D culture conditions. We recently reported that pancreatic CSCs can be enriched in vitro using anchorage-independent 3D culture 22,23 and employed this technique to further investigate the relationship between GM2 expression and stem cells. FACS analysis of GM2 after 3D culture showed that almost all cells in the spheres expressed GM2 in MIA PaCa-2 (96.1%) whereas the fraction of GM2+ cells in other cell lines were not significantly changed, except in PK-59 in which GM2+ cells were still less than 15% of cells in the spheres (Fig. 4a). Real-time PCR analysis of stemness markers in MIA PaCa-2 showed that each stemness markers were increased in sphered cells compared with adherent-cultured cells (Fig. 4b). These results suggest that CSC-like MIA PaCa-2 cells induced in 3D culture expressed GM2. We next examined the potential for GM2 synthesis in sorted GM2-and GM2+ cells in adherent culture conditions (Fig. 2a) after re-culture under adherent or 3D culture. Under adherent culture, the expression pattern of GM2 in both sorted GM2-and GM2+ cells returned to the steady state of unsorted adherent-cultured cells after 7 days of re-culture (Fig. 4c). By contrast, sorted GM2+ and GM2-cells expressed high levels of GM2 after 7 days of re-culture under 3D culture (Fig. 4c). Thus, 3D culture induced GM2+ CSC-like cells and GM2-cells have potential to synthesize GM2 and did so in 3D culture conditions. GM2+ CSC-like cells exhibit responsiveness against TGF-β1 via interaction with GM2 and TGF-receptor. Next, to investigate the role of GM2 in GM2+ CSC-like MIA PaCa-2 cells, we used N-(5'-adamantane-1'-yl-methoxy)-pentyl-1-deoxynojirimycin (AMP-dNM), which is a specific inhibitor of glucosylceramide synthase that can be used to study the functional roles of endogenous gangliosides without affecting ceramide levels 24,25 . Treatment with AMP-dNM lowered GM2 expression and that of other gangliosides (GM3, GM1, and GD1a) on sphere cells ( Fig. 5a and S1). Morphological analysis of control and AMP-dNM-treated MIA PaCa-2 sphere cells using TEM showed that there were few microvilli on the surface of sphere cells and that vacuoles (asterisks) were noticeable in AMP-dNM-treated cells (Fig. 5b). The growth of sphere cells was not affected by AMP-dNM treatment (Fig. S2A), and real-time PCR analysis showed that expression of stemness markers in sphere cells was not affected by AMP-dNM treatment (Fig. S2B). Furthermore, anti-cancer drug resistance determined by ATP assays showed that anti-cancer drug resistance was higher in sphere cells and was accompanied by increased expression of transporters (ABCG2 and ABCC2) compared with adherent-cultured cells and were not negatively affected in sphere cells treated with AMP-dNM (Fig. S3). These results suggest that CSC-like cells enriched in 3D culture conditions have high GM2 expression, but that GM2 does not functionally affect stemness.
Epithelial-mesenchymal transition (EMT) has recently been reported to be associated with CSCs and malignant behaviors in cancer [26][27][28] . We investigated whether GM2 + sphere cells were responsive against TGF-β1, an EMT inductive factor, and could further the contribution of GM2 to TGF-β1 signaling. Western blot analysis showed that TGFβRI and TGFβRII in sphere cells were expressed to the same extent as in adherent-cultured cells, regardless of AMP-dNM treatment (Fig. 5c), and that activation of TGF-β1 signaling, as indicated by phosphorylation of Smad2/3 in TGF-β1-treated sphere cells, was attenuated by AMP-dNM treatment (Fig. 5d). These results suggest that GM2 contributes to the regulation of TGF-β1 signaling. TGF-β1 signaling occurs dynamically after interaction between TGFβRI and TGFβRII, which are localized in the lipid rafts 29 . GM2 is also located on lipid rafts complexed with integrin and modulates the downstream signaling pathway 30 . Immunoprecipitation analysis was performed to confirm the involvement of GM2 in TGF-β1 signaling. Interaction of GM2 and TGFβRII was observed in TGF-β1-treated sphere cells and this interaction was inhibited by AMP-dNM treatment (Fig. 5e). Furthermore, expression of EMT markers, such as N-cadherin and Slug, in conjunction with activation of TGF-β1 signaling was observed (Fig. 5f). Furthermore, increased invasion in TGF-β1 treated sphere cells was also suppressed by AMP-dNM treatment, determined using invasion assays with matrigel-coated chamber (Fig. 5g). Taken together, these data suggest that expression of GM2 was linked to CSCs, and that GM2 correlated with regulation of TGF-β1 induced EMT and is involved in invasiveness of MIA PaCa-2 cells.

GM2 expression in 3D culture conditions is regulated by MAPK pathway.
To clarify the contribution of GM2 expression to signaling in CSC-like cells, we examined GM2 expression after treatment with inhibitors of several signaling pathways that are considered to be involved in sphere formation and CSC self-renewal, such as the FGF/EGF and JAK/STAT3 pathways 31,32 . Among signal inhibitors, we found that PD0325901 (MEK inhibitor) decreased GM2 expression in sphere cells, a decrease that was accompanied by a reduction of sphere size (Fig. 6a,b). Next, we used real-time PCR to examine whether glycosyltransferases and/or NEU3 were associated with GM2 synthesis, finding that NEU3 expression was downregulated and all others were increased in sphere cells compared with adherent-cultured cells (Fig. S4). By contrast, PD0325901 treatment significantly decreased ST3GAL5 expression in sphere cells compared with non-treated sphere cells (Fig. S4). As shown in Fig. 2b, ST3GAL5 contributes to synthesis of all gangliosides at the end of ganglioseries. This suggests that inhibition of GM2 in sphere cells via PD0325901 may be dependent on attenuated expression of ST3GAL5. Furthermore, PD0325901 treatment attenuated activation of TGF-β1 signaling (Fig. 6c) and further inhibited increases in invasion of TGF-β1-treated sphere cells (Fig. 6d). These results suggest that suppression of GM2 expression by treatment with an MEK inhibitor suppressed the increased invasion associated with activation of TGF-β1 signaling in CSC-like cells.
GM2 is expressed in human PDAC cases. In order to clarify the expression and roles of GM2 in human PDAC cases, we performed immunohistochemical analyses on 117 human pancreas tissues. Normal pancreatic ducts in the PDAC tissues were negative or faintly positive for GM2. Islet cells and acinar cells were also negative for GM2. GM2 was positive in 18.8% (n = 22) of PDAC cases (Fig. 7b and Table 1). In pancreatic intraepithelial neoplasia (PanIN) lesions close to PDAC cells, GM2 was positive in 2.6% (n = 1) of low-grade PanIN cases and positive in 3.6% (n = 1) of high-grade PanIN cases. The number of cases with GM2 expression was significantly higher in PDAC tissues than in low-grade and high-grade PanIN (p = 0.01). There were no significant differences in GM2 expression between low-grade PanIN and high-grade PanIN (low grade PanIN vs high-grade PanIN, p = 0.672).

Discussion
Several gangliosides have previously been identified in CSCs 32 . For example, ganglioside GD2 was identified as a CSC marker in breast cancer 33 . Reduction of GD2 by inhibition of GD3 synthase in breast cancer cells reduced the CSC population and CSC-associated properties, including in vivo tumor formation 33 . Another report showed that GD2, GD3, GM2, and GD1a are highly expressed in breast CSCs, but only GD2 and GD3 were suggested to contribute to sphere formation and cell motility 34 . In the current study, we report for the first time that GM2 are highly expressed in pancreatic CSC-like cells. Although the association between GM2 with pancreatic CSCs is unknown, GM2 is associated with metastasis in melanoma and lung carcinoma 20,21 . Here we found that GM2 may www.nature.com/scientificreports www.nature.com/scientificreports/ be involved in invasion of the adjacent extracellular matrix through regulation of TGF-β1 signaling in pancreatic CSCs. High invasiveness to stromal tissues of GM2+ cells may lead to larger tumor size in animal experiments and human PDAC tissues. Future investigations focusing on GM2, e.g. using anti-GM2 antibodies, may lead to new PDAC therapies targeting GM2. Downregulation of NEU3, a plasma membrane-associated sialidase that modulates ganglioside content by removing sialic acid, is known to contribute to increases in GM2 expression 35 . As shown in this study, NEU3 expression is reduced in adherent cultured GM2+ cells compared with GM2-cells and in CSC-like cells, indicating that the level of NEU3 expression may be involved in regulation of GM2 expression. It is necessary to further clarify what factors are involved in the expression of GM2 in CSC-like cells.
In this study, we demonstrated that AMP-dNM treatment could inhibit TGF-β1 signaling and invasion, presumably by inhibiting the interaction between GM2 and TGFβRII. It has been previously reported that AMP-dNM treatment improves glucose tolerance, reduces hepatic steatosis, and enhances insulin response in rodent models of type 2 diabetes 25,36 . Furthermore, AMP-dNM treatment was demonstrated to reduce the development of atherosclerosis by lowering plasma cholesterol levels in APOE*3-Leiden and low-density lipoprotein receptor -/-mice 37 . Thus, the effects of AMP-dNM on age-related diseases, including diabetes and arteriosclerosis, are known at the individual level. Incidence of pancreatic cancer is also increased in the aged population, and changes in glycosylation that occur with aging are thought to correlate with carcinogenesis and tumor progression. Therefore, the suppressive effects of AMP-dNM on tumor invasion to stromal tissues may be expected in pancreatic cancer, which is frequently seen in PDAC.
Several monoclonal antibodies (mAbs), such as the chimeric anti-EGFR mAb cetuximab, have been successfully introduced into clinical practice to treat cancer patients 38 . For example, it has been demonstrated that humanized anti-GM2 antibodies inhibited production of multiple organ metastases and prolonged survival of the SCID mouse in a metastasis model induced by GM2-expressing small-cell lung cancer cells 21 . GM2 is predominantly detected in poorly differentiated types of pancreatic cancer and correlates with growth and invasion, therefore therapeutics using anti-GM2 antibodies may be expected to have potential in treatment of pancreatic cancer. On the other hand, it has been shown that anchorage-independent growth and lung metastasis in triple-negative breast cancer were inhibited by MEK inhibitors 39 . In the present study, we showed for the first time that treatment with an MEK inhibitor attenuated GM2 synthesis in pancreatic CSC-like cells, suppressed anchorage-independent growth and increased invasion associated with activation of TGF-β1 signaling. Further www.nature.com/scientificreports www.nature.com/scientificreports/ study will be required, but the MEK inhibitor, which acts on MAPK signaling, may have clinical potential for preventing invasion in GM2 + pancreatic cancer cases.
It is known that gangliosides are secreted from cancer cells by shedding and have pleiotropic effects, such as regulation of tumor growth 40 , angiogenesis 41,42 , and immune modulation 43 . The exosome contains a cell membrane antigen, including ganglioside 44 , so the exosome-expressed ganglioside would also be secreted into the blood. Furthermore, it has been reported that GM2 can be quantified in plasma using LC/ESI-MS/MS 45 , a method that could be used to detect GM2 derived from cancers in the blood. In the current study, prominent GM2 expression was observed in MIA PaCa-2 cells (established from poorly differentiated cells obtained from a PDAC patient), in poorly differentiated human pancreatic cancer tissues and in pancreatic CSC-like cells. Hence, there is the possibility that shed GM2 and/or exosome bound GM2 could be detected in the blood of patients with early stage, poorly differentiated pancreatic cancer.
In conclusion, we showed that a PDAC cell line expressing GM2 in 2D-culture exhibited a high growth rate and high tumor initiation, pancreatic CSC-like cells expressing GM2 in 3D-culture exhibit responsiveness against TGF-β1 resulting in promotion of invasion, and GM2 expression is associated with growth and advance of human PDAC (Fig. 7c). Therefore, further study will be required for development of an early detection method for GM2 expressing pancreatic cancers and novel therapeutic strategies targeting GM2.

Perineural invasion
No (n = 9) 1 (11. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods Cell culture. Cell culture was performed as previously described 46,47 . The human PDAC cell lines PK-1, PANC-1, PK-59, and MIA PaCa-2 were obtained from the Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer, Tohoku University (Sendai, Japan). PK-45P, PK-8, T3M-4, and KP-4 human PDAC cells were provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. Cells were grown in growth medium (RPMI 1640 medium containing 10% fetal bovine serum) at 37 °C under a humidified 5% CO 2 atmosphere. For EMT induction, cells were cultured for 48 h in growth medium containing 10 ng/mL TGF-β1 (Peprotech, Rocky Hill, NJ, USA). For 3D culture, cells in growth medium were plated at 1.0 × 10 4 cells/well or 3.0 × 10 3 cells/well in 24-well ultra-low attachment plates (Corning Inc. Kennebunk, ME, USA) or 96-well ultra-low attachment plates (Thermo Fisher Scientific, Waltham, MA, USA), respectively. The spheres were aspirated after 7 days using micropipettes and placed in microcentrifuge tubes for use in further experiments. Photographs of the spheres were taken using a sphere analyzing device, Cell3iMager duos (SCREEN Holdings Co., Ltd., Kyoto, Japan).
Cases and tissue samples. A total of 117 cases of PDAC were examined. All patients underwent surgical resection at the Tokai University Hospital (Kanagawa, Japan) between 2007 and 2017. Only cases of conventional PDAC were included in the analysis. Patients who received preoperative chemotherapy or who were diagnosed as special histological types of cancer, such as adenosquamous carcinoma, mucinous carcinoma, undifferentiated carcinoma or invasive intraductal papillary mucinous neoplasm, were excluded. All surgical materials were fixed in formalin and embedded in paraffin. Formalin-fixed and paraffin-embedded (FFPE) tissue samples were cut into 4 μm-thick sections and stained with H&E. Pathological TNM (pT/pN/pM) staging and histological grade were classified according to the Union for International Cancer Control, eighth edition 48 . Overall survival was defined as the time interval between surgery and death or the date of the last patient visit. PanIN contained in the PDAC sections was evaluated. PanIN was classified as either low or high grade 49  Immunoblotting. Immunoblotting and immunoprecipitations were performed as previously described 51 .

Heterotopic implantation of PDAC cells.
In order to clarify in vivo tumorigenicity of GM2-or GM2 + PDAC cells, 1 × 10 5 cells/animal (n = 5, 5 sites per cell line) were subcutaneously injected into the bilateral flank of nine-week old, female, athymic mice (BALB/cAJcl-nu/nu; CLEA Japan Inc, Tokyo, Japan). Tumor volume was calculated weekly using the formula: volume = a × b 2 × 0.5, where a is the longest diameter and b is the shortest. The animals were monitored for 5 weeks. All experiments were approved by the Animal Experiments Committee of Nippon Veterinary and Life Science University and were performed in accordance with Guidelines for Animal Experiments by the Nippon Veterinary and Life Science University.
Immunohistochemistry. Immunohistochemical analysis of GM2 expression was performed by cutting FFPE blocks into 4 μm-thick sections and staining with anti-GM2 antibody (dilution 1:1000; TCI) using BOND-MAX (Leica Biosystems) according to the manufacturer's instructions. Normal kidney tissue was used as a positive control expression. GM2 expression in PDAC or PanIN cells that was equal to or stronger than that in the positive control was considered positive. The percentage of membranous and/or cytoplasmic cells that stained positively for GM2 in the total PDAC cells or total cells in a single PanIN lesion was evaluated. GM2 expression was considered positive when 5% of the cancer cells in the tissues were stained. Ethics statements. All experiments and methods were performed in accordance with relevant guidelines and regulations. Specifically, primary human pancreatic cancer specimens were obtained with consent from patients who were admitted to Tokai University Hospital. All participants provided written informed consent for this investigation. The investigation was performed under approval of the Institutional Review Board at Tokai University (permission number: 17R220) and Tokyo Metropolitan Hospital and Institute of Gerontology (permission number: R17-50). All animal experiments were performed under the approval of the Animal Experiments Committee of Nippon Veterinary and Life Science University (permission number: 30K-27).

Statistical analysis. Statistical analyses for in vitro studies were performed using EZR (Saitama Medical
Centre, Jichi Medical University; http://www.jichi.ac.jp/saitama-sct/SaitamaHP.files/statmedEN.html; Kanda, 2012) or Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA). Results are expressed as means ± standard deviation (SD) from three independent experiments. Unpaired Student's t-test and one-way ANOVA with Tukey's HSD test were used for comparing two or more groups, respectively. Results of tumor size were expressed as means ± standard error (SE) and a Mann-Whitney test was used. Statistical analyses of immunohistochemistry were conducted using IBM SPSS Statistics software, version 25.0 (IBM, Chicago, IL, USA). Fisher's exact test or Pearson's χ2 test was used to analyze the relationship between clinicopathological factors and GM2 expression and the changes in GM2 expression during carcinogenesis in low grade PanIN, high-grade PanIN and PDAC. Overall survival curves and median survival times were plotted using the Kaplan-Meier method, and compared using the log-rank test. All data were representative of at least two independent experiments. p values < 0.05 were considered significant.