DAPT and gliotoxin, were obtained from Merck Millipore (Darmstadt, Germany). The compounds were reconstituted in dimethyl sulfoxide (DMSO). Human cell lines derived from melanoma (518A2), HCC (HEP3B, SNU398, Huh7), pancreas-CA (PANC1), and breast-CA (HCC38, MDA-MB-468) were cultured in RPMI 1640 supplemented with 10% heat inactivated fetal calf serum (FCS), 2 mM Glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin (all reagents were obtained from Gibco, Life Technologies Inc., Paisley, United Kingdom). HEP3B (HB-8064), SNU398 (CRL-2233), PANC1 (CRL-1469), HCC38 (CRL-2314), and MDA-MB-468 (HTB-132) cell lines were obtained from the American Type Culture Collection (ATCC, Rockville, MA, United States). The cell line Huh7 (JCRBO 403) was obtained from the National Institute of Biomedical Innovation (Osaka, Japan). The melanoma cell line 518A2, characterized by the BRAF V600E mutation and a CDKN2A exon 2 deletion, was obtained from Leiden University. The generation of the HCC cell line HCC-3 was described previously (Winter et al., 2008). Cells were incubated with the indicated concentrations of inhibitors or with equal amounts of solvent.

Flow cytometry was performed on a FACSCalibur using CellQuest Pro software version 5.2.1 (Becton Dickinson, San Jose, CA, United States). Annexin V, and propidium iodide staining was performed to estimate the percentages of cells undergoing apoptosis using a kit from Bender Med. Systems Inc. (Vienna, Austria). The percentage of apoptotic cells was calculated as sum of propidium iodide (PI)-/Annexin V (Ax)+ (early apoptosis) and PI+/Ax+ (late apoptosis/necrosis) cells.

3 × 10⁷ cells were lysed in 1 ml hypotonic buffer (10 mM HEPES, pH 7.9; 1.5 mM MgCl₂; 10 mM KCL) containing 0.15% NP-40 at 4°C for 10 min. The nuclear proteins were extracted from the nuclear fraction by suspending the nuclei in 100 μl extraction buffer (300 mM KCl; 1.5 mM MgCl₂; 20 mM HEPES, pH 7.9; 0.2 mM EDTA; 25% Glycerin) at 4°C for 20 min with constant agitation. A CSL site spanning oligonucleotide (5′-CAGCCCTGTGGGAACTTGCTG-3′) was annealed with the reverse complementary strand and served as probe. EMSA for the detection of NOTCH/CSL complexes were performed essentially as described (Hubmann et al., 2002).

The N1^IC (bTAN 20) and N2^IC (C651.6DbHN) antibodies used for supershift/interference and western blot assays were obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Department of Biological Science, Iowa City, IA, United States). The NFκB p65 (RelA) and ACTB Antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, United States). Western blotting was performed according to standard protocols.

For NOTCH2 cellular localization, human HCCs tissue array slides were obtained from SuperBioChips Laboratories, Seoul, South Korea. Heat induced antigen retrieval was performed in 10 mM citrate buffer at pH 6. After incubation with the NOTCH2 primary antibody (C651.6DbHN), slides were incubated with biotinylated secondary antibody, followed by alkaline phosphatase-streptavidin and chromogen. The stained slides were evaluated by a pathologist to evaluate NOTCH2 localization.

Total RNA was extracted using the TRI Reagent^®isolation system (Sigma–Aldrich, St Louis, MO, United States). Moloney murine leukemia virus (M-MLV) reverse transcriptase and GoTaqPCR kits (Promega) were used for semiquantitative RT-PCR using primer sets as follows: HEY1, forward 5′-ATACGCCTGCATTTACCAGC-3′ and reverse 5′-TCAATTGACCACTCGCACAC-3′. Primer sets for NOTCH1, NOTCH2, and ACTB were published elsewhere (Hubmann et al., 2013). Real-time quantitative RT-PCR (qPCR) for NOTCH2 was performed with TaqMan^®-probes (Hs01050717_m1) purchased from Applied Biosystems (Thermo Fisher Scientific, Waltham, MA, United States).

siRNA duplexes (siRNAs) for NOTCH2 (ON-TARGETplus^TM) and controls (RISC-free Co-siRNA, and siGLO red transfection indicator) were obtained from Dharmacon (Lafayette, CO, United States). Transfection of siRNAs into HCC cell lines was performed by using the lipid reagent siLentFect^TM from Bio-Rad Laboratories. The transfection efficiency was determined by FACS and varied from 80 to 90%.

The redox status was analyzed by a DCFDA (2′,7′-dichlorofluorescin diacetate) containing ROS detection assay using a kit from Abcam (Cambridge, United Kingdom) according to the manufacturer’s instructions. Within the cells, DCFDA is deacetylated and oxidized by ROS into highly fluorescent DCF (2′,7′-dichlorofluorescin) which is measured by flow cytometry.

Pathogen-free, 4–6 week old, female athymic nude mice (Harlan Winkelmann, Borchen, Germany) were housed under sterile conditions and treated according to the regulations of the local animal ethics committee (BMBWK-66.009/0055-BrGT/2006). Hundred microliter of a tumor cell suspension in PBS containing 9 × 10⁶ 518A2 cells were inoculated subcutaneously into the lower right and left flank of each mouse as described previously (Krepler et al., 2004). When mean tumor volume reached approximately 75 mm³ (based on caliper measurements), mice were randomly assigned to treatment groups. One hundred and fifty microliter gliotoxin solution or an equal amount of vehicle control (Vitamin-E derivative) was intraperitoneal injected according to the dosing schedule. The trial was stopped when control mice reached a mean tumor volume of 1 cm³.

Statistical significance of differences among treatment groups was calculated by one-way ANOVA and Bonferroni’s test using SPSS software (SPSS Inc., Chicago, IL, United States). Differences in mean tumor volume between two time points within treatment groups were calculated by using the Wilcoxon matched pairs test. P-values less than 0.05 were considered to be of statistical significance.

The oncogenic properties of NOTCH receptors are tightly linked with their nuclear localization and their transcriptional activity (Jeffries and Capobianco, 2000; Tando et al., 2013; Ressel et al., 2014; Saito et al., 2016). In the nucleus, the intracellular domain of NOTCH (N^IC) forms a ternary transcription factor complex on CSL responsive genes (Ntziachristos et al., 2014). Therefore, we analyzed nuclear extracts from human cell lines derived from melanoma (518A2), HCC (Hep3B, HCC-3, SNU398, Huh7), and pancreas-CA (PANC-1) for DNA-bound NOTCH/CSL complexes on a CSL consensus site (GTGGGAA) spanning probe by EMSA. We found that NOTCH is highly active in these cell lines with exception of Huh7 (Figure 1A, lane 1). Addition of N2-Ab to the EMSA reaction (supershift/interference assay) completely disrupted the formation of the transcription factor complexes (Figure 1A, lane 3) whereas N1-Ab had no effect (Figure 1A, lane 2). This shows that NOTCH2 is the dominant nuclear active NOTCH receptor in these cell lines. Interestingly, the well differentiated HCC cell line Huh7 did not display any DNA-bound NOTCH complexes and, thus, served as negative control for canonical NOTCH/CSL signaling in our experiments (Hayashi et al., 2015).

In order to analyze the efficiency of GSI in these cell lines, we next tested the sensitivity of DNA-bound NOTCH2 complexes to the widely used and selective GSI DAPT. As shown in Figure 1A lane 4, treatment of 518A2, HEP3B, HCC-3, SNU398, and PANC-1 cells with 5 μM DAPT, a concentration which we previously showed that it inhibits NOTCH2 signaling in GSI sensitive CLL cases (Hubmann et al., 2010, 2013), had no effect on the NOTCH2 transcription factor complex after 1 day of incubation.

Based on our recent data which showed that the Aspergillum derived secondary metabolite gliotoxin is a potent NOTCH2/CSL transactivation inhibitor in CLL cells (Hubmann et al., 2013), we tested the effect of gliotoxin in these cell lines. As shown in Figure 1A lane 5, exposure to gliotoxin at 0.2 μM (=64 ng/ml) for 24 h completely blocked the formation of DNA-bound N2^IC complexes in nuclear NOTCH2 positive cell lines.

In terms of cell viability, DAPT had almost no effect while gliotoxin efficiently induced apoptosis. Interestingly, the pro-apoptotic effect of gliotoxin was restricted to nuclear NOTCH2 positive cells while the well differentiated nuclear NOTCH2 negative HCC cell line Huh7 was found to be resistant to gliotoxin treatment (Figure 1B).

Since NOTCH2 is considered as a tumor suppressor gene in certain breast cancer cell lines (O’Neill et al., 2007; Kim et al., 2016), we additionally tested HCC38 and MDA-MB-468 breast cancer cells for their NOTCH activity and sensitivity to gliotoxin. HCC38 cells were found to be positive for NOTCH2/CSL complexes whereas no remarkable NOTCH activity could be detected in MDA-MB-468 cells (Supplementary Figure 1A). In HCC38 cells, gliotoxin inhibited NOTCH2/CSL complexes (supplementary Figure 1A) and induced apoptosis in a dose dependent manner (Supplementary Figure 1B) whereas MDA-MB-468 cells were resistant to gliotoxin treatment (Supplementary Figure 1B).

In order to substantiate the relation between the differentiation grade and the nuclear localization of NOTCH2, we performed immunohistochemical analysis of human HCC tissue sections obtained from patients with less differentiated and more differentiated tumors. In accordance with published data (Hayashi et al., 2015), staining primary human HCC tissues with anti-NOTCH2 antibodies showed a prominent nuclear localization of NOTCH2 in less differentiated HCC tissues with a more immature cellular morphology as indicated by a high nuclear to cytosol (N/C) ratio. In contrast, a predominant cytoplasmic localization of NOTCH2 was found in more differentiated HCC tissues with low H/C ratio, underlining the clinical relevance of our cell line data (Figure 1C). A global overview about the expression and localization of NOTCH2 in cancer tissue arrays is presented in the human protein atlas¹ (Uhlen et al., 2015).

We next tested the effect of gliotoxin on total NOTCH2 protein expression in nuclear NOTCH2 positive (SNU398) and in nuclear NOTCH2 negative (Huh7) HCC cells by western blotting. The wild type NOTCH2 receptor is anchored as cleaved heterodimer on the cell surface. This consists of a 180-kD NOTCH2 extracellular domain (N2^EC) and a 120-kD NOTCH2 transmembrane form (N2^TM). After ligand binding, the 100-kD NOTCH2 intracellular domain (N2^IC) is released from the N2^TM form by γ-secretase cleavage. As shown in Figure 1D, both cell lines express mainly N2^IC (100-kD) whereas N2^TM (120-kD) was only weakly detectable in Huh7 cells (Figure 1D). Gliotoxin efficiently inhibited the expression of NOTCH2 in SNU398 cells but not in Huh7 cells suggesting that gliotoxin specifically targets NOTCH2 expression in nuclear N2^IC active SNU398 cells. However, further studies on the subcellular localization of NOTCH2 and dynamics of NOTCH2 migration upon exposure to gliotoxin would be of a major interest for getting deeper insight on the mechanism of action of gliotoxin in terms of NOTCH regulation.

The nuclear NOTCH2 activity in SNU398 cells was reflected by the transcription of its target gene HEY1 (Figure 1D). Gliotoxin but not DAPT inhibited HEY1 transcription in SNU398 cells. This result supports the hypothesis, that deregulation of NOTCH2 signaling in certain cell lines is probably caused by the expression of truncated, ligand independent N2^IC forms which do not require γ-secretase for processing and function.

Since gliotoxin may have a wide range of downstream targets (Scharf et al., 2016) and may exert its apoptotic effect via reactive oxygen species (ROS) production and via inhibition of the redox sensitive transcription factor NFκB (Gardiner et al., 2005), we evaluated the effect of gliotoxin on the nuclear expression of the NFκB subunit p65 (RelA) (Gloire and Piette, 2009). As shown in Figure 1D, gliotoxin had no influence on the expression of p65 in nuclear extracts of both cell lines. Furthermore, gliotoxin did not influence the redox status in gliotoxin sensitive SNU398 cells as determined by a cellular ROS detection assay (Figure 1E). In summary, these results clearly show that the induction of apoptosis by gliotoxin, at least in this cell line, is associated with the transcriptional NOTCH2/CSL activity, is independent of the GSI sensitivity of N2^IC, and is not associated with effects on the redox status of the treated cells.

We next compared the effect of NOTCH2 inhibition by gliotoxin and siRNA in the HCC cell lines SNU398 and HCC-3. Gliotoxin remarkably inhibited NOTCH2 mRNA expression (Figure 1F, left panel), the formation of NOTCH2/CSL complexes (Figure 1G, left panel), and induced apoptosis (Figure 1H, left panel) within 12 h of incubation. In contrast, NOTCH2 siRNA downregulated NOTCH2 mRNA expression (Figure 1F, right panel) but had no influence neither on NOTCH2/CSL complexes (Figure 1G, right panel) nor on cell viability (Figure 1H, right panel) within 2 days of incubation. This indicates that gliotoxin may target NOTCH2 signaling at the transcription factor level which in turn disrupts a positive feedback loop of NOTCH2 mRNA expression (Artavanis-Tsakonas et al., 1999). In contrast, downregulation of NOTCH2 mRNA alone seems to be insufficient to target NOTCH2 signaling. This might be explained by the stability of the NOTCH2/CSL complex on DNA. Together, these results confirmed that the induction of apoptosis by gliotoxin is tightly linked with its inhibitory effect on the formation of the NOTCH2/CSL complex on DNA (Hubmann et al., 2013).

Since many in vivo investigations and clinical trials with GSI have been already reported (Espinoza and Miele, 2013; Andersson and Lendahl, 2014) or are still ongoing², we focused the in vivo investigation on gliotoxin treatment using our established 518A2 melanoma xenograft mouse model (Krepler et al., 2004).

We found that intraperitoneal (ip) application of gliotoxin on a single day (5 mg/kg in the morning and 2.5 mg/kg in the evening) was well tolerated and the mice did not show any study limiting side effects. Therefore, we applied this gliotoxin dosing schedule to the treatment groups on day 11 (early stage tumor model; group A, n = 6) and on day 25 (late stage tumor model; group B, n = 6), respectively. We then determined the effect of gliotoxin on the tumor mass in treated animals compared to the control group (n = 8) by serial caliper measurements (Figure 2).

In the early stage tumor model, application of gliotoxin on day 11 revealed a significant decrease of the mean tumor volume on day 14 (control group versus treatment group A: 115 mm³ vs. 83 mm³; P = 0.008) with the greatest tumor mass reduction after 2 weeks (68%) (Figure 2A). This effect was even more pronounced in the late stage tumor model, where the same dosing schedule was applied on day 25 (Figure 2B). This caused a significant decrease of the mean tumor volume on day 28 (control group versus treatment group B: 576 mm³ vs. 218 mm³, P = 0.005) with the greatest tumor mass reduction after 1 week (77%) (Figure 2B). Therefore, we conclude that gliotoxin is highly active in this melanoma xenograft mouse model and, thus, is therapeutically relevant for NOTCH2 associated malignancies.