To examine the effect of TTFields on the viability of PDAC cell lines, we cultured BxPc-3, BxGEM and AsPC-1 cells in specially manufactured 96-well culture plates with electrodes to induce TTFields at frequencies of 150 kHz by a generator ( Figure 1A ). The cells in these culture plates were incubated at 37°C or 38.5°C for up to six days. The plates were then evaluated daily by MTT assay, which measures the cellular metabolic activity as an indicator of cell viability, proliferation and cytotoxicity. The values of each group were set to 0 at the start of the experiment at day 1. Whereas the MTT signal of cells cultured at 37°C in the presence or absence of TTFields continuously increased, it was even higher when the cells were cultured at 38.5°C ( Figure 1B ). A pronounced inhibition of viability was only seen upon the combination of hyperthermia and TTFields. These results became even clearer upon setting the controls to 1 and presenting the data in bar charts. While TTFields tended to increase the viability at 37°C, they significantly decreased it at 38.5°C after 3 days ( Figure 1C ). After 6 days, however, TTFields largely decreased the viability even at 37°C and almost completely at 38.5°C. Representative images of the cell morphology under microscopic magnification confirm the above results ( Figure 1D ). Interestingly, mild hyperthermia paired with TTFields failed to substantially inhibit the viability of the nonmalignant pancreatic ductal cell line CRL-4023, even 6 days after treatment ( Figure 1E ).

To obtain information on the influence of mild hyperthermia paired with TTFields on stem cell properties, we studied the ability to form colonies. BxPc-3, BxGEM and AsPC-1 cells were treated with TTFields at 37°C or 38.5°C for 72 h. Then, the cells were reseeded at clonal density in 6-well plates, and colony formation was evaluated 14 days later. Whereas TTFields at 37°C only marginally reduced colony formation, they strongly reduced it at 38.5°C ( Figure 2A ). In contrast, hyperthermia alone significantly increased colony formation. For examination of the long-term effect, the surviving cells of each group were collected and equal cell numbers were reseeded at clonal density without further treatment. The resulting, so-called “second-generation” colonies were analyzed 14 days later. TTFields at 37°C reduced colony formation to approximately 50% to 60%, and TTFields in the presence of 38.5°C nearly completely abolished colony formation. Additionally, hyperthermia alone did not enhance colony formation, as observed in the short-term treatment. Similar results were obtained when the cells were treated and grown as spheroids ( Figure 2B ). Under these conditions, the effects were more pronounced upon reseeding the cells for second-generation spheroid formation. Next, we examined the effect of TTFields combined with mild hyperthermia or of mild hypothermia alone on migration and performed scratch assays. The closure of the wounded region was examined 12 h and 24 h later. The gap was significantly larger upon combination of TTFields and 38.5°C compared to 38.5°C alone ( Figure 3 ). In contrast, the same experiments at 37°C were largely ineffective (data not shown). This result suggests again that the combination therapy also inhibits migration more than each single treatment option alone.

To further highlight gene expression changes, we treated BxGEM cells at 37°C and 38.5°C and used untreated cells cultured at 37°C as control. Seventy-two hours later, the RNA was harvested and a gene array analysis was performed. A heat map was generated and demonstrated well-separated gene clusters between the groups ( Figure 4A ). This finding was further highlighted by a volcano plot analysis of the array results, which exhibited a greater spreading of differentially regulated genes when TTFields plus were combined with 38.5°C compared to TTFields alone at 37°C ( Figure 4B ). Then, a KEGG enrichment analysis was performed, which led to the identification of 20 significantly up- or downregulated pathways closely related to tumorigenesis ( Figure 4C and Table 1 ). Among the differentially regulated pathways were the ubiquitin proteasome system and autophagy, DNA repair, DNA replication, MAPK signaling and cell cycle regulation. Next, a GSEA analysis was performed and resulted in enrichment plots of autophagy, cell cycle and DNA replication ( Figure 5A ). The most significantly differentially regulated genes belonging to these pathways are shown in heat maps ( Figure 5B ). To identify the related signaling networks, the open source software platforms Cytoscape and String were utilized, which showed Tp53, PNCA, and MAD2L2 in the center of two cell cycle/DNA replication signaling networks and AKT1 in the center of an autophagy-related signaling network ( Figure 5C ). Thus, there is an interaction of the identified differentially regulated genes in biological signaling networks.

For validation of our previous results, BxPc-3, BxGEM and AsPC-1 cells were treated with TTFields plus hyperthermia at 38.5°C, whereas cells cultured at 37°C without further treatment served as control. Seventy-two hours later, RT-qPCR was performed to examine the expression of the before identified target genes CyclinB, CDK1, GADD45B, MAD2L2, MCM6, TP53, ATG5, TSC1 and AKT1. The expression of each gene was normalized to the expression of GAPDH, and the gene expression of the TTField plus 38.5°C group was evaluated relative to the 37°C control, which was set to 1. GADD45B, TP53, ATG5 and TSC1 were significantly upregulated in all cell lines, whereas MAD2L2, MCM6 and AKT1 were significantly downregulated ( Figure 6A ). In contrast, the expression levels of CyclinB and CDK1 did not appreciably change. Most importantly, the expression patterns largely confirmed the expression data from the microarray assay, which is underlined by comparison of the results of both assays ( Table 2 ).

To investigate the effect of TTFields on the cell cycle, the cells were cultured at 37°C or 38.5°C in the presence or absence of TTFields. Seventy-two hours later, the cells were stained with propidium iodide and examined by flow cytometry. TTFields combined with mild hyperthermia led to an increase of the G2/M phase, and a decrease of the G1 phase, whereas a statistically significant S phase arrest was not observed ( Figure 6B ).

To examine the influence induction of autophagy, the cells were cultured at 37°C or 38.5°C in absence or presence of TTFields. Seventy-two hours later, the expression of the autophagic flux indicator p62 and the presence of the autophagy marker LC3-II were examined by Western blot analysis. Whereas p62 expression decreased upon TTField treatment at 37°C and 38.5°C, indicating autophagy signaling, LC3-II was only clearly visible after combining TTFields with mild hyperthermia, as shown by a representative Western blot, quantification by Image J and a diagram with the means and standard deviations; also, the crude Western blot images are provided ( Figure 6C and Supplementary Figure S1 , data not shown). These data indicate that autophagy signaling occurred and was strongest when TTFields were combined with mild hyperthermia.

To demonstrate the clinical relevance of candidate genes, which we identified by gene array analysis, we utilized TCGA public database and identified the availability of expression data paired with clinical data for DDIT4/REDD1, which is involved in cell growth, proliferation and survival; TSC1, which acts as a tumor suppressor; MCM6, which is involved in the cell cycle and DNA replication; and ORC1, which is involved in the initiation of DNA replication. Kaplan-Meier plots were created to visualize the relevance of expression of these genes in PDAC tissue and the related clinical outcome of patients ( Figures 7A, B ). Whereas high expression levels of DDIT4, MCM6 and ORC1 were related to significantly shorter survival, high expression of the tumor suppressor TSC1 was related to significantly longer survival, as expected. Together, these data confirm the relevance of the identified TTField plus hyperthermia-induced candidate genes identified by gene array analysis.

We combined for the first time TTFields and mild hyperthermia of 38.5°C and demonstrate that this combination is more effective than each single treatment. Thus, the application of these two well-tolerated modalities may provide an efficient new treatment approach. A logical next step would now be the evaluation by in vivo experiments with mice, for which, however, the complex technical equipment would first have to be further developed. Then it would also have to be tested to apply the TTFields for quite short time in a repeated way with an interval time between repetitions of days, because such a protocol and its effects are interesting for the clinical application. As both individual treatments, TTFields and hyperthermia, are already in clinical use, their direct, combined application in patients, as co-treatment along with chemotherapy, would also be conceivable.

The established human pancreatic cancer cell lines BxPc-3 and AsPC-1 and the human hTERT-HPNE immortalized nonmalignant pancreatic duct cell line CRL-4023 were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and cultured as described (55). Gemcitabine-resistant BxPc-3 cells (BxGEM) were selected from parental cells by continuous gemcitabine treatment with increasing concentrations up to 200 nM for more than one year (56). Negative mycoplasma cultures were confirmed by monthly mycoplasma tests. The cell lines were recently authenticated by a commercial service (DSMZ, Braunschweig, Germany).

Recently we developed in-house a device for applying TTFields to cells growing in 96-well flat-bottom plates and evaluated its functionality in control experiments (57). This TTField device enabled us to treat PDAC cells with an alternating voltage between two insulated wires. The antenna wires were arranged diametrically to each well within a 96-well plate and formed, together with a variable capacitor, the capacitive part of an electrical resonant circuit. The inductive part of the resonant circuit was the secondary coil of a transformer whose primary coil was connected to a function generator (HP 3310A, Hewlett Packard, USA) and a customized preamplifier. An alternating low-voltage signal with a fixed frequency between 100 and 300 kHz was generated in the primary coil, and the resonant circuit was tuned to resonance using the variable capacitor. At resonance, voltage amplitudes U_eff of up to 500 V were generated in the resonant circuit. We calculated the maximum value of the electric field strength in the area of the cells under highly simplified conditions. The antenna wires ended at the bottoms of the 96-well plates on which the cells grew. Considering only the 2-dimensional case on the bottom plane of the 96-well plates, the antenna wires can be approximated as point-like charges, and Coulomb’s law in equation (1) was used for the calculation.

In this equation, Q is the field-generating charge, e → is the unit vector for the field direction, ϵ ₀ = 8.85 * 10^–12 As/Vm is the vacuum permittivity, ϵ_r is the relative permittivity and r is the distance to the charge. The total field is formed as a superposition of 2-point charges according to equation (1). The calculations were carried out only for the straight line connecting the two-point charges. The field effects of charges outside the calculation plane were ignored. With the applied voltage U_eff = 500 V, we calculated the charge Q with equation (2):

with the capacitance C of the antenna wires.

The total capacitance C was calculated according to equation (3), taking into account, that there were different dielectric materials between the antenna wires: air, plastic, and the saline cell medium.

C_air , C_P , and C_Med were calculated with equation (4):

with k = air, P, Med and A = πd_wL_w /2, d_w = 0.5mm, L_w = 10mm.

C_air is the capacitance of the air gap between the antenna wire and the adjacent well wall using ϵ_air = 1 and d_air = 3 mm, C_P is the capacitance of the plastic well wall using ϵ_p = 2 and d_P = 1 mm, and C_Med is the capacitance of the cell medium using ϵ_Med = 75 and d_Med = 5 mm. Depolarization factors due to the geometry of the various dielectrics were not taken into account.

For treatment of cell lines with TTFields, the cells were seeded in 96-well plates at a density of 2000-4000 cells/ml in 200 µl of medium per well. After installation of the antenna wires for the treatment with TTFields, the plates were placed in the incubator. The cell culture medium was not changed during treatment.

PDAC cell lines were resuspended at a concentration of 2 to 4x 10³ cells/ml, and 200 µl per well of a 96-well microplate were seeded. We minimized the counting errors by continuous shaking of the cell suspension and the use of a multi-head pipette for cell seeding. Also, the different cell concentrations from 2 to 4x 10³ cells/ml per batch enabled us to took those for evaluation in which the control group grew to nearly 100% density after 6 days. After treatment, 20 µL of the 5 mg/mL yellow MTT tetrazolium salt (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) were added to each well. The plates were incubated for 3.5 h at 37°C. The principle of this assay is that viable, metabolically active cells, but not dead or dying cells, contain NAD(P)H-dependent oxidoreductase enzymes, which reduce the yellow MTT to purple formazan crystals. By microscopic inspection, we detected the formation of formazan crystals and subsequently the medium was carefully removed, 200 µL DMSO was added and the cells were agitated on a shaker for 10 min at 37°C until the formazan crystals were completely dissolved. Finally, the optical density of the purple colour, which reflects mitochondrial activity, was measured at 570 nm by the use of an ELISA reader.

Proteins were harvested by the use of RIPA Lysis Buffer and the protein concentration was determined by the use of the BCA Protein Assay Kit (both from Abcam, Cambridge, UK). After denaturation by boiling for 5 min, the samples were incubated on ice and then separated along with a commercial protein ladder by SDS-PAGE. A semi-dry system was used for the transfer of separated proteins in the gel to a PVDF membrane. The membrane was blocked by shaking in 3% BSA solution, followed by incubation with primary antibodies and thereafter with IRDye® infrared.

Dye-conjugated secondary antibodies (LI-COR Biosciences GmbH, Bad Homburg, Germany). The infrared intensity was measured by the use of an Odyssey CLx Infrared Imaging System (LI-COR Biosciences GmbH). The primary antibodies were rabbit polyclonal anti-LC3B and anti-P62 (Abcam, Cambridge, UK) and rabbit monoclonal anti-GAPDH antibody (Abcam).

Seventy-two hours after treatment with TTFields at 37°C or 38.5°C as described above, the cells were harvested, washed and 5 ml ice-cold 70% EtOH was added drop-by-drop while vortexing. Subsequently, the samples were incubated at -20°C overnight. After PBS-washing, the cells were resuspended in 0.5 mL of the PI/RNase Staining Buffer (Becton Dickinson, Heidelberg, Germany) and stored for 15 min at room temperature. The cell cycle was analyzed by flow cytometry and the use of a FACSCalibur™ device (Becton Dickinson). The measured data were assessed with FLOWJO software (FLOWJO, LLC, Ashland, USA) to determine the cell fractions in the G1, S, and G2+M phases of the viable cell population.

Seventy-two hours after treatment with TTFields at 37°C or 38.5°C as described above, the cells were detached by trypsinization and reseeded at low density in 6-well plates (BxPc-3 and BxGEM, 1000 cells/well; AsPC-1, 800 cells/well) in six-fold approaches. The cells were incubated in a 37°C incubator for 10-14 days without changing the cell culture medium until considerably appropriate colonies were observed under the microscope in plates containing untreated control cells. Subsequently, the cells were washed with 10 mL PBS and 2 mL of 3.7% paraformaldehyde (PFA), was added for 10 min at room temperature. The fixation solution was removed, and 2 mL of 70% EtOH were added for 10 min. Finally, the cells were stained with 0.05% Coomassie blue, followed by washing with water. The plates were dried overnight. Colonies consisting of a minimum of 50 cells were counted under a stereomicroscope. The percentage of plating efficiency was evaluated by normalising the values obtained for treated cells to the values of non-treated cells. After normalisation, the value of non-treated cultures was set to 1. For the analysis of the second generation of colony formation, living cells were harvested and re-plated at low density in 6-well plates.

Seventy-two hours after treatment with TTFields at 37°C or 38.5°C as described above, the cells were seeded at a density of 1x 10³ cells/mL in 500 µL NeuroCult™ NS-A basal serum-free medium supplemented with 20 ng/mL hEGF, 10 ng/mL bFGF and 2 μg/mL heparin (STEMCELL Technologies Cambridge, US) per well in 24-well Sphera Low-Attachment Surface plates (ThermoFisher Scientific, Waltham, MA, USA) for spheroid formation. The spheroids were photographed after 5 days, and cell spheroids were identified according to their typical shape and size and the percentage of spheroids was calculated. To evaluate secondary spheroid formation, equal numbers of surviving cells of the first round of spheroid-formation were reseeded. For quantification of the percentage of spheroid forming cells, the cells were seeded at one cell per well in 96-well plates. Wells with more than one cell were excluded from evaluation.

The cells (1x 10³/well) were seeded in 96-well plates and grown to >90% confluence overnight. The cell layer was scratched with the fine end of a 10 μL pipette tip (time 0). Then the cells were exposed to TTFields at 37°C or 38.5°C. The closure of the wounded region was evaluated 12 h and 24 h later. Pictures were taken by the use of a Nikon Eclipse TS 100-F inverted microscope equipped with a camera. The images and the closure of the gap were analyzed with the computer program Image J (58).

Seventy-two hours after treatment with TTFields at 37°C or 38.5°C as described above, the cells were harvested and total RNA was isolated by the use of the RNeasy^® Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The concentration was determined by the use of a Nanodrop 2000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA) and the RNA was stored at -80°C until use. A total of 100 ng mRNA were reverse transcribed using the High Capacity RNA to cDNA™ Kit according to the instructions of the manufacturer (ThermoFisher Scientific, Waltham, MA, USA). The cDNA was diluted in 200 µl RNAse-free water and 1 μL cDNA was used immediately as template for real-time PCR, which was performed with 1 μL of forward and reverse primers for the genes of interest and SSoAdvanced SYBR Green Supermix (Qiagen, Hilden, Germany). The primer sequences were created by the use of PrimerBank, available for free online (https://pga.mgh.harvard.edu/primerbank/), and the primer sequences for MAD2L2, GADD45B, MCM6, Cyclin B1 (CCNB1), CDK1, ATG5, TSC1, TP53 and AKT1 are shown in Table 3 . The PCR conditions were 10 min 95°C as initial denaturation step, followed by 40 cycles of 15 s denaturation at 95°C, 1 min annealing at 60°C, followed by cooling down to 4°C. All reactions were run in duplicate. Melt curve analysis for each pair of primers and agarose gel electrophoresis of the PCR products ensured the specificity of the primers. The gene expression level of each target gene was normalised to that of GAPDH. The results are presented as relative expression value (REV) by using the 2^−ΔΔCt method of relative quantification given in equation (6).

By referring to each REV value of the target gene, the fold change (FC) can be calculated using equation (7).

mRNA was isolated from untreated or TTField-treated BxGEM cells grown at 37°C or 38.5°C using the RNeasy Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. Microarray analyses was performed at the Microarray-Analytic Center of the Medical Faculty Mannheim using Clariom™ D Assays (Thermo Fisher Scientific, Dreieich, Germany). Biotinylated antisense complementary DNA (cDNA) was prepared based on a standard Affymetrix labeling protocol with the GeneChip^® WT Plus Reagent Kit and the GeneChip^® Hybridization, Wash and Stain Kit. Thereafter, hybridization on the chip was performed in a GeneChip hybridization oven 640, staining was performed in the GeneChip Fluidics Station 450, and the chip was then scanned with a GeneChip Scanner 3000. The custom CDF version 22 with ENTREZ-based gene definitions was used to annotate the arrays (59). The raw fluorescence intensity values were normalized by applying quantile normalization and RMA background correction. One-way analysis of variance (ANOVA) was applied; a fold change of 1.5 was used for the selection of differentially expressed genes using commercial SAS JMP10 Genomics version 6 (SAS Institute, Cary, USA). A false positive rate/false discovery rate (FDR) <0.15 was considered to be the level of significance. The accession number of the gene array at ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) is #E-MTAB-10270 (BxGEM, control, TTF, TTF+hyperthermia 38.5°C).

The gene array-derived list of differentially expressed genes was further correlated for their biological function and involvement in signaling pathways. For the selection of differentially expressed genes, an absolute value of the logarithmic fold change (log FC) ≥1.3 and a cutoff of P<0.05 were chosen to identify statistically significant pathways.

Heat maps and volcano plots were created with the free software environment for statistical computing and graphics R (https://www.R-project.org/).

Gene set enrichment analysis (GSEA) was performed using the fgsea package available in the open-source software Bioconductor (https://bioconductor.org/packages/release/bioc/html/fgsea.html).

The freely available online database resource KEGG (Kyoto Encyclopedia of Genes and Genomes, https://www.genome.jp/kegg/) was used for the selection of relevant biological functions and pathways with enrichment scores of P<0.05.

The open source platform String 11.0 (https://string-db.org) was used to collect, score and integrate publicly available sources of protein–protein interaction data and to supplement it with calculations and predictions. By the use of the Cytoscape StringApp ( http://apps.cytoscape.org/apps/stringapp), we visualized the identified protein-protein-interaction network based on the obtained differentially expressed candidate genes.

The Cancer Genome Atlas (TCGA) is a publicly available online database (https://www.cancer.gov/about-nci/organization/ccg/research/structural-genomics/tcga) with over 20,000 primary cancer and matched normal samples spanning 33 cancer types with matched molecular and clinicopathological data. TCGA was used for evaluation of the identified target genes DDIT4, TSC1, MCM6 and ORC1 by Kaplan-Meier analysis using the online available Kaplan-Meier Plotter (http://kmplot.com/analysis/index.php?p=service). mRNAs expression data of the above-mentioned target genes from human PDAC tumor tissue were available and used to analyze the overall survival (OS) and recurrence-free survival (RFS) of patients. The database divides patient samples into high expression groups and low expression groups according to the median values of mRNA expression and validates them by a Kaplan–Meier survival curve. Information on number of patients, median values of mRNA expression, 95% confidence interval (CI), hazard ratio (HR), and P-value can be found on the Kaplan–Meier Plotter web page (http://kmplot.com/analysis/index.php?p=service). A P-value <0.05 was considered as statistically significant. The log-rank test was used to calculate the statistical significance of the differences observed among the Kaplan-Meier curves.

Statistical analyses were performed using JMP14.0.0 (SAS Institute, Cary, USA) or Prism 7.0 (GraphPad Software, San Diego, California, USA). For most of the experiments, Dunnett’s tests were used to calculate the P values; nontreated cells at 37°C served as the control group. For cell migration, Student’s t-tests of independent samples were used to calculate the P values. All experiments were repeated a minimum of three times, and the data are presented as the mean ± standard deviation (SD). The null hypothesis was rejected when P <0.05 (*P <0.05 and **P <0.01).