Glioblastoma-Derived Three-Dimensional Ex Vivo Models to Evaluate Effects and Efficacy of Tumor Treating Fields (TTFields)

Simple Summary In glioblastoma, tumor recurrence is inevitable and the prognosis of patients is poor, despite multidisciplinary treatment approaches involving surgical resection, radiotherapy and chemotherapy. Recently, Tumor Treating Fields (TTFields) have been added to the therapeutic set-up. These alternating electric fields are applied to glioblastoma at 200 kHz frequency via arrays placed on the shaved scalp of patients. Patients show varying response to this therapy. Molecular effects of TTFields have been investigated largely in cell cultures and animal models, but not in patient tissue samples. Acquisition of matched treatment-naïve and recurrent patient tissues is a challenge. Therefore, we suggest three reliable patient-derived three-dimensional ex vivo models (primary cells grown as microtumors on murine organotypic hippocampal slices, organoids and tumor slice cultures) which may facilitate prediction of patients’ treatment responses and provide important insights into clinically relevant cellular and molecular alterations under TTFields. Abstract Glioblastoma (GBM) displays a wide range of inter- and intra-tumoral heterogeneity contributing to therapeutic resistance and relapse. Although Tumor Treating Fields (TTFields) are effective for the treatment of GBM, there is a lack of ex vivo models to evaluate effects on patients’ tumor biology or to screen patients for treatment efficacy. Thus, we adapted patient-derived three-dimensional tissue culture models to be compatible with TTFields application to tissue culture. Patient-derived primary cells (PDPC) were seeded onto murine organotypic hippocampal slice cultures (OHSC), and microtumor development with and without TTFields at 200 kHz was observed. In addition, organoids were generated from acute material cultured on OHSC and treated with TTFields. Lastly, the effect of TTFields on expression of the Ki67 proliferation marker was evaluated on cultured GBM slices. Microtumors exhibited increased sensitivity towards TTFields compared to monolayer cell cultures. TTFields affected tumor growth and viability, as the size of microtumors and the percentage of Ki67-positive cells decreased after treatment. Nevertheless, variability in the extent of the response was preserved between different patient samples. Therefore, these pre-clinical GBM models could provide snapshots of the tumor to simulate patient treatment response and to investigate molecular mechanisms of response and resistance.


Introduction
Glioblastoma (GBM) is one of the most aggressive primary brain tumors in adults with a median survival time of 16-18 months and a five-year survival rate of 6% for male

Patient-Derived Primary Cells, Cell Lines, and Cell Culture
To generate PDPC, necrotic areas and blood vessels were removed from the intraoperatively obtained tumor tissue and the latter then separated using a homogenizer. The homogenized tumor material was cultured in 25 cm 3 cell culture flasks (Corning, New York, NY, USA) in Dulbecco's Modified Eagle's Medium (DMEM) containing 1 g/L glucose, sodium pyruvate, 3.7 g/L NaHCO 3 and L-glutamine and supplemented with 20% v/v heat-inactivated fetal calf serum (FCS), 2 × non-essential amino acids (NEAA, 100× stock, add 10 mL to 500 mL medium) (all from Gibco, Carlsbad, CA, USA) and 1.5% vitamin C (Sigma-Aldrich, St. Louis, MO, USA) at 37 • C, 5% CO 2 , and 95% humidity until an adherent cell layer was formed [27]. A GBM cell line U87MG (CLS, Eppelheim, Germany) was cultured in DMEM supplemented with 10% v/v FCS, 2 × NEAA, 3 mM L-glutamine, 100 U/mL penicillin, and 100 mg/mL streptomycin (Invitrogen, Carlsbad, CA, USA) as a monolayer under the same conditions.

Preparation of Organotypic Hippocampal Brain Slice Cultures (OHSC)
OHSC were prepared as described previously [23]. Briefly, mice 5-8 days postpartum (p5-p8) were decapitated following ethical guidelines. The brain was dissected under the microscope and glued with the dorsal surface facing downward into the sample tube of a vibratome (Precisionary Instruments, Greenville, SC, USA) ( Figure 2A) with histoacryl glue (B. Braun, Tuttlingen, Germany) ( Figure 2B,C). The tube was filled with molten agarose (Sigma-Aldrich, St. Louis, MO, USA) ( Figure 2D). A cooling block pre-cooled at −80 °C and then stored on ice until use ensured rapid hardening of the agarose ( Figure 2E). The sample tube was clamped into the vibratome, which was set to an advance of 3.5 and an oscillation of 6 and 350 µ m thick slices were generated ( Figure 2F). Slices were collected in a preparation tray filled with minimal essential medium (MEM) supplemented with 1% penicillin/streptomycin, 1% L-glutamine (all from Gibco, Carlsbad, CA, USA) and 1% glucose (Sigma-Aldrich, St. Louis, MO, USA). The hippocampus was isolated, transferred using a wide glass pipette and cultured on inserts with a semi-permeable membrane of 0.4 µ m pore size (Greiner Bio-one, Frickenhausen, Germany) in a 24-well plate (Corning Costar, New York, NY, USA) containing brain slice medium (MEM supplemented with 25% normal horse serum, 25% Hank's Balances salt solution (HBSS), 1% penicillin/streptomycin, 1% L-glutamine (all from Gibco, Carlsbad, CA, USA), vitamin C and 1% glucose (both from Sigma-Aldrich, St. Louis, MO, USA) at 35 °C, 5% CO2 and 95% humidity. It was vital that excess medium be aspirated from the OHSC. The slices obtain their nutrient supply from the medium via the pores of the insert and must not be covered by any liquid.
In addition, the slice should be placed as centrally and planar as possible onto the insert. After a cultivation period of two weeks, experiments could be started.

Preparation of Organotypic Hippocampal Brain Slice Cultures (OHSC)
OHSC were prepared as described previously [23]. Briefly, mice 5-8 days postpartum (p5-p8) were decapitated following ethical guidelines. The brain was dissected under the microscope and glued with the dorsal surface facing downward into the sample tube of a vibratome (Precisionary Instruments, Greenville, SC, USA) ( Figure 2A) with histoacryl glue (B. Braun, Tuttlingen, Germany) ( Figure 2B,C). The tube was filled with molten agarose (Sigma-Aldrich, St. Louis, MO, USA) ( Figure 2D). A cooling block pre-cooled at −80 • C and then stored on ice until use ensured rapid hardening of the agarose ( Figure 2E). The sample tube was clamped into the vibratome, which was set to an advance of 3.5 and an oscillation of 6 and 350 µm thick slices were generated ( Figure 2F). Slices were collected in a preparation tray filled with minimal essential medium (MEM) supplemented with 1% penicillin/streptomycin, 1% L-glutamine (all from Gibco, Carlsbad, CA, USA) and 1% glucose (Sigma-Aldrich, St. Louis, MO, USA). The hippocampus was isolated, transferred using a wide glass pipette and cultured on inserts with a semi-permeable membrane of 0.4 µm pore size (Greiner Bio-one, Frickenhausen, Germany) in a 24-well plate (Corning Costar, New York, NY, USA) containing brain slice medium (MEM supplemented with 25% normal horse serum, 25% Hank's Balances salt solution (HBSS), 1% penicillin/streptomycin, 1% L-glutamine (all from Gibco, Carlsbad, CA, USA), vitamin C and 1% glucose (both from Sigma-Aldrich, St. Louis, MO, USA) at 35 • C, 5% CO 2 and 95% humidity. It was vital that excess medium be aspirated from the OHSC. The slices obtain their nutrient supply from the medium via the pores of the insert and must not be covered by any liquid. In addition, the slice should be placed as centrally and planar as possible onto the insert. After a cultivation period of two weeks, experiments could be started.

Seeding of Fluorescence-Labeled GBM Cells and Organoids onto OHSC
The cells were fluorescently labeled for easy visualization of U87MG cells, PDPC and organoids growing on OHSCs. U87MG and PDPC were transfected with green fluorescent protein (GFP) utilizing the pmaxGFP plasmid (Lonza, Cologne, Germany) in combination with nucleofection using the Amaxa Cell Line Nucleofector Kit V (Lonza, Cologne, Germany), as detailed elsewhere [27,28]. Briefly, cells were detached with 0.25% trypsin/EDTA (Carl Roth, Karlsruhe, Germany) and suspended in cell culture medium. For each transfection, 1 × 10 6 cells were centrifuged for 10 min at 300× g at room temperature.

Seeding of Fluorescence-Labeled GBM Cells and Organoids onto OHSC
The cells were fluorescently labeled for easy visualization of U87MG cells, PDPC and organoids growing on OHSCs. U87MG and PDPC were transfected with green fluorescent protein (GFP) utilizing the pmaxGFP plasmid (Lonza, Cologne, Germany) in combination with nucleofection using the Amaxa Cell Line Nucleofector Kit V (Lonza, Cologne, Germany), as detailed elsewhere [27,28]. Briefly, cells were detached with 0.25% trypsin/EDTA (Carl Roth, Karlsruhe, Germany) and suspended in cell culture medium. For each transfection, 1 × 10 6 cells were centrifuged for 10 min at 300× g at room temperature. The supernatant was discarded, and the cells were re-suspended in 100 µL of Nucleofector solution V. In a 1.5 mL reaction tube, cells were mixed with 2 µg pmaxGFP plasmid, transferred to transfection cuvettes and electroporated with a transfection program U29. After transfection, cells were transferred to the wells of a 6-well plate containing 1 mL cell culture medium each. The cells were allowed to recover for 2-3 days before further use. Approximately 1 × 10 5 cells were taken up in 10 µL cell culture medium and spread onto the surface of the OHSC. After 2-3 days, microtumor growth and invasion could be detected using an inverted fluorescence microscope LEICA DMI 3000 B (Leica, Wetzlar, Germany).
Organoids were mechanically minced with scalpels into pieces of 100 µm size and incubated in 10 µm Carboxyfluorescein succinimidyl ester (CFSE) in PBS for 15 min as provided in the CellTrace TM CFSE Cell Proliferation Kit (Invitrogen, Carslbad, CA, USA) and following the manufacturer's instructions. The solution was then replaced with fresh GBO medium and incubated at 37 • C for another 30 min. The next day, organoids fluoresced at an excitation wavelength of 488 nm and were placed onto the OHSC using a pipette.

Generation of Patient-Derived Organotypic Tumor Slice Cultures
Generation of organotypic tumor slice cultures was based on a publication by Merz et al. [25]. After surgical tumor resection, the tissue was directly transferred to Hibernate A medium and stored on ice. The tumor tissue was carefully freed from necrosis and blood vessels and cut into approximately 2 cm × 0.5 cm pieces using a scalpel. Preparation of slices was performed as described above for the OHSC (Figure 2). The slices were embedded in agarose ( Figure 2G) and had to be carefully cut out with a scalpel before they could be transferred to the center of the semi-permeable membrane with a wide glass pipette. The vitality of slices, histopathology and tumor content were assessed by an experienced neuropathologist. Tumor slices could be cultured in brain slice medium at 35 • C, 5% CO 2 and 95% humidity for up to 6 days. The medium was changed every second day.

TTFields Treatment
The inovitro™ laboratory research system (Novocure, Haifa, Israel) was used for TTFields administration. U87MG cells and PDPC cultured as monolayers were treated as described previously [29]. Briefly, glass coverslips with 20 mm diameter (Hartenstein, Würzburg, Germany) were placed into inovitro ceramic dishes (Novocure, Haifa, Israel). Cells were trypsinized and plated onto the coverslips by placing 350 µL cell culture medium containing 30,000 cells as a drop in their center. Cells attached during a 20 h incubation at 37 • C and 5% CO 2 . The medium was replaced by 2 mL fresh cell culture medium, the ceramic dishes were placed onto a base plate connected to a TTFields generator and TTFields at 200 kHz were applied with an intensity of 1.7 V/cm for 72 h. The medium was renewed every 48 h. Control cells were kept under the same conditions without TTFields application. To evaluate TTFields effects, cells were trypsinized and counted using the Scepter 2.1 cell counter (Merck, Darmstadt, Germany).
To treat OHSC and tumor slices with TTFields, ceramic dishes with high walls (Novocure, Haifa, Israel) were utilized ( Figure 3). The holders, containing the inserts with semi-permeable membranes, were placed into the dishes and 2.5 mL brain slice medium was pipetted into the dishes outside the inserts ( Figure 3A-E). OHSC and tumor slices, respectively, were transferred as described above ( Figure 3F,G). To avoid condensation water to drop onto the slice surfaces, a 12 mm coverslip was placed over each insert. The ceramic dish was covered with parafilm and closed with a lid to minimize evaporation of the medium ( Figure 3C,H,I). TTFields were applied for 72 to 96 h at 200 kHz and 1.5 V/cm. Condensation water was carefully aspirated every day, and the medium was changed every second day. To evaluate microtumor growth on OHSC, images were taken using a LEICA DMI 3000 B microscope, LEICA DFC450 camera and LAS V4.5 software (all Leica, Wetzlar, Germany). Tumor size was determined on the fluorescence images using the Measure Tool from the open-source program Fiji (Image J 1.53c) [30,31]. Tumor slices were fixed in 4% formalin (Carl Roth, Karlsruhe, Germany) for 24 h at 4 • C and then transferred to phosphate buffered saline (PBS) (Sigma-Aldrich, St. Louis, MO, USA) for immunohistochemical and immunofluorescence staining.

Immunohistochemical and Immunofluorescence Staining
The fixed brain slices were dehydrated, embedded in paraffin, and sliced into 3 µ m thick sections. Next, standardized hematoxylin and eosin (HE) (Carl Roth, Karlsruhe, Germany) staining was performed for histology. For immunohistochemical staining, Ki67 (ab16667, Abcam, Cambridge, UK) and GFAP (sc33673, santacruz, Dallas, TX, USA) antibodies were used at a dilution of 1:1000 and 1:100 in antibody dilution buffer (DCS Innovative Diagnostik Systeme, Hamburg, Germany), respectively, and incubated overnight at 4 °C. Protein expression was visualized using the secondary antibodies AlexaFluor488 and AlexaFluor555 (both from Invitrogen, Carlsbad, CA, USA) in a 1:1000 dilution, and incubated for 1 h at room temperature. Finally, the slices were mounted using Fluoroshield mounting medium containing DAPI (Abcam, Cambridge, UK). Five representative fields of view per slide were photographed with the LEICA DMI 3000 B microscope with standardized settings at 40× magnification and analyzed for staining intensity via the

Immunohistochemical and Immunofluorescence Staining
The fixed brain slices were dehydrated, embedded in paraffin, and sliced into 3 µm thick sections. Next, standardized hematoxylin and eosin (HE) (Carl Roth, Karlsruhe, Germany) staining was performed for histology. For immunohistochemical staining, Ki67 (ab16667, Abcam, Cambridge, UK) and GFAP (sc33673, santacruz, Dallas, TX, USA) antibodies were used at a dilution of 1:1000 and 1:100 in antibody dilution buffer (DCS Innovative Diagnostik Systeme, Hamburg, Germany), respectively, and incubated overnight at 4 • C. Protein expression was visualized using the secondary antibodies AlexaFluor488 and Alex-aFluor555 (both from Invitrogen, Carlsbad, CA, USA) in a 1:1000 dilution, and incubated for 1 h at room temperature. Finally, the slices were mounted using Fluoroshield mounting medium containing DAPI (Abcam, Cambridge, UK). Five representative fields of view per slide were photographed with the LEICA DMI 3000 B microscope with standardized settings at 40× magnification and analyzed for staining intensity via the batch processing function of the open-source program Fiji (ImageJ 1.53c) [30,31]. The macro settings are described elsewhere [32].

Statistical Analysis
Statistical analysis was performed using GraphPadPrism 9 software (GraphPad Software, San Diego, CA, USA). Statistical significance was defined by unpaired 2-tailed t-tests, and by ANOVA. p < 0.05 was considered to be significant. In the box plots the boxes represent the median with the 25% and 75% quartile and the whiskers the minimum and maximum of the data set. Descriptive statistics were obtained to calculate the tumor size and its growth or decrease. The mean value was expressed as percentage with standard error of the mean (±SEM). The difference was also shown with ± SEM as well as the 95% confidence interval (CI). All experiments were performed at least in triplicates.

Patient Cohort
To establish the described ex vivo GBM models, tumor samples of eight GBM patients were utilized ( Table 1). The tumors were classified according to the most recent WHO classification of tumors of the central nervous system [26]. Only GBM, IDH-wildtype CNS WHO grade 4 were used. A low amount of tumor material prevented execution of all models with the same samples in this proof of principle study.

Patient-Derived GBM Primary Cells Display Variable Responses to TTFields at 200 kHz
Previously published data revealed that application of TTFields at 200 kHz significantly decreased the number of GBM cells [11,12,29,33]. In our experiments, we treated U87MG cells and three PDPC monolayer cell cultures with TTFields at 200 kHz for 72 h. These cells responded to TTFields to a variable extent ( Figure 4A). Whereas the cell numbers of U87MG cells were reduced to 41.75% ± 13.76% compared to the untreated control (p = 0.0033, CI = −90.76 to −25.74) and those of GBM-1 were diminished to 49.7% ± 10.87% (p < 0.0001, CI = −69.19 to −31.41), GBM-2 and GBM-3 were non-responsive to TTFields treatment ( Figure 4A).

Cell Proliferation Decreases and Apoptosis Increases in Human Organotypic GBM Tumor Slice Cultures when Treated with TTFields
Human organotypic GBM tumor slice cultures were treated with TTFields at 200 kHz for 72 h. HE staining revealed a general decrease in cell number and a concomitant increase of apoptotic cells in the treated slices compared to untreated controls ( Figure 6A). While staining for the proliferation marker Ki67 remained basically constant over time in GBM-7 (mean expression 10.38% ± 2.14% at 0 h vs. 8.63% ± 0.71% 72 h later), and only slightly dropped to 5.0% ± 1.32% (p = 0.0298, CI −6.84 to −0.41) when TTFields were applied ( Figure 6B,C), GBM-8 displayed a completely different picture. There was an increase of Ki67 positivity by 9.17 ± 1.994 percentage points from 6% ± 1.99% (0 h) to 15.17% ± 2.01% (72 h) observable not only in untreated control slices (p = 0.0004, CI 4.89 to 13.44), but even more pronounced by 19.0 ± 2.05 percentage points to 25.0% ± 2.1% after 72 h TTFields treatment (p < 0.0001, CI 14.61 to 23.39) (Figure 6B,D). Ki67 staining only identifies cell cycle entry and does not reveal if the cell cycle will be completed [34,35]. Since TTFields

Cell Proliferation Decreases and Apoptosis Increases in Human Organotypic GBM Tumor Slice Cultures when Treated with TTFields
Human organotypic GBM tumor slice cultures were treated with TTFields at 200 kHz for 72 h. HE staining revealed a general decrease in cell number and a concomitant increase of apoptotic cells in the treated slices compared to untreated controls ( Figure 6A). While staining for the proliferation marker Ki67 remained basically constant over time in GBM-7 (mean expression 10.38% ± 2.14% at 0 h vs. 8.63% ± 0.71% 72 h later), and only slightly dropped to 5.0% ± 1.32% (p = 0.0298, CI −6.84 to −0.41) when TTFields were applied ( Figure 6B,C), GBM-8 displayed a completely different picture. There was an increase of Ki67 positivity by 9.17 ± 1.994 percentage points from 6% ± 1.99% (0 h) to 15.17% ± 2.01% (72 h) observable not only in untreated control slices (p = 0.0004, CI 4.89 to 13.44), but even more pronounced by 19.0 ± 2.05 percentage points to 25.0% ± 2.1% after 72 h TTFields treatment (p < 0.0001, CI 14.61 to 23.39) (Figure 6B,D). Ki67 staining only identifies cell cycle entry and does not reveal if the cell cycle will be completed [34,35]. Since TTFields have been shown to induce G1 cell cycle arrest [36], we wondered whether this increase in Ki67 would be sustained, and incubated the GBM-8 slices for another 24 h with TTFields. At 96 h total treatment, the percentage of Ki67 positive cells within the slice dropped to 3.0% ± 1.14% (p < 0.0001, CI −27.74 to −16.26), while Ki67 rose to 26.5% ± 3.12% in control cells (p = 0.0120, CI 3.09 to 19.57) ( Figure 6B,D). have been shown to induce G1 cell cycle arrest [36], we wondered whether this increase in Ki67 would be sustained, and incubated the GBM-8 slices for another 24 h with TTFields. At 96 h total treatment, the percentage of Ki67 positive cells within the slice dropped to 3.0% ± 1.14% (p < 0.0001, CI −27.74 to −16.26), while Ki67 rose to 26.5% ± 3.12% in control cells (p = 0.0120, CI 3.09 to 19.57) ( Figure 6B,D).  Table 1.

Discussion
In drug screening, effects of potential candidates are initially tested using in vitro models before embarking on a more costly, albeit more specific and detailed, in vivo study. Although in vitro systems are efficient and entail simple procurement, they often fail to demonstrate physiological interactions that occur in vivo. Nonetheless, even though the employment of in vivo methods more closely resembles clinical settings, humans and animals still differ in responses [37]. Therefore, the use of 3D tumor models such as spheroids and organoids has provided an opportunity to perform in vitro experiments using materials obtained from patient tissue, bridging the gaps between in vitro, pre-clinical and clinical set-ups. The use of spheroids and organoids eliminate the limitation of two-dimensional 2D models and are more consistent with in vivo studies, since they can contain several cell types such as stromal cells [17,18]. Moreover, they can be developed to resemble the tumor microenvironment more closely, allowing functional investigations of drug responses as well as metastasis, such as combining the 3D model with biomaterials [17], or engineering them to mimic physiological environments by using tumor-on-a-chip technology [38]. Three-dimensional models have been successfully used for drug screening studies for breast, lung and colon cancer [20][21][22]. Studies in GBM using 3D systems are

Discussion
In drug screening, effects of potential candidates are initially tested using in vitro models before embarking on a more costly, albeit more specific and detailed, in vivo study. Although in vitro systems are efficient and entail simple procurement, they often fail to demonstrate physiological interactions that occur in vivo. Nonetheless, even though the employment of in vivo methods more closely resembles clinical settings, humans and animals still differ in responses [37]. Therefore, the use of 3D tumor models such as spheroids and organoids has provided an opportunity to perform in vitro experiments using materials obtained from patient tissue, bridging the gaps between in vitro, pre-clinical and clinical set-ups. The use of spheroids and organoids eliminate the limitation of two-dimensional 2D models and are more consistent with in vivo studies, since they can contain several cell types such as stromal cells [17,18]. Moreover, they can be developed to resemble the tumor microenvironment more closely, allowing functional investigations of drug responses as well as metastasis, such as combining the 3D model with biomaterials [17], or engineering them to mimic physiological environments by using tumor-on-a-chip technology [38].
Three-dimensional models have been successfully used for drug screening studies for breast, lung and colon cancer [20][21][22]. Studies in GBM using 3D systems are also not far behind, as more research favoring them emerge [19,39,40]. However, none of these models has ever been combined with TTFields treatment ex vivo.
TTFields added to the standard maintenance therapy have improved the median progression-free and overall survival of newly diagnosed GBM patients by 2.7 and 4.9 months, respectively. Above that, they more than doubled the 5-year survival from 5% to 13% [16]. TTFields are mostly well tolerated, with systemic toxicity for the TTFields plus standard treatment group being comparable to the standard therapy group, with mild to moderate skin toxicity, e.g., skin rash and eczema underneath the arrays, occurring in 52% of patients in the TTFields group [16,41]. In order to reach highest efficacy, TTFields should be applied for ≥18 h each day on average [42]. It has been discussed that carrying the device, which weighs 1.3 kg, on a daily basis might restrict patients from daily activities and might hamper social life [43]. In addition, to place the arrays, patients need to shave their scalps, which could lead to stigmatization [43]. However, preliminary data from the recent TIGER trial do not support such apprehensions [44]. However, TTFields hardware is costly and maintenance is expensive [10]. Moreover, not all patients respond to TTFields to the same extent; some gain only a slight advantage, while others survive for 5 years or longer [16]. A standard sub-group analysis based on, e.g., MGMT promoter methylation status, extent of resection, Karnofsky performance index (KPI), or age, failed to identify which patients respond better to TTFields, as a survival benefit was demonstrated in all sub-groups [16]. Thus, patients who benefit from TTFields should be chosen wisely to allocate TTFields treatment efficiently. Information on mechanisms leading to TTFields resistance is very limited, and most molecular effects of TTFields have been investigated in cell cultures or animal models only [15]. Hence, the analysis of patient tumor samples before and after TTFields treatment would be desirable. However, we know from our own clinical experience that a bias could exist amongst such patients. Their TTFields device usage might vary [42], as well as the total treatment duration [16]. In addition, the time between end of the TTFields treatment and occurrence of relapse differs from patient to patient. Finally, patients with a reduced KPI might not be re-operated upon, while the time until relapse and following re-surgery might be prolonged in other patients. These factors could limit the availability of tissue for research. On the other hand, tumor samples of especially these patient groups are most interesting in terms of investigating treatment resistance related to molecular and cellular differences of inter-patient tumor heterogeneity and post-therapeutical changes.
To address these limitations, we tested several patient-derived ex vivo tumor tissue culture methods to identify patients who might benefit from TTFields therapy prior treatment, and to investigate short-and long-term effects of TTFields, including molecular prerequisites leading to TTFields responsiveness or resistance. At first, the standard GBM cell line U87MG and different PDPC monolayer cultures were treated with TTFields. As expected, there was a high variability of TTFields response, with some PDPC not reacting at all, possibly reflecting the patients' tumor sensitivity for TTFields treatment. However, when grown as microtumors on OHSC, the cells became more sensitive. Thus, three-dimensional growth might better represent the in vivo situation and lead to more reliable results, as also has been shown for other cancer entities, such as breast cancer [20]. Nevertheless, while this primary cell-only approach might be easily done from a technical point of view, it has restricted validity as a screening system for clinical application. Due to the rupture of cell-cell-contacts during lysis, duration of cultivation, lack of tumor microenvironment, hypoxic gradients, and medium components such as serum, PDPC cultures not only changes their antigen surface expression patterns, but also undergoes molecular and transcriptional changes and thus no longer represent the parental tumor characteristics [19,24,39]. New ex vivo models such as organoids or tumor slice cultures might overcome these hurdles [17][18][19]39,40].
Organoids represent the histological characteristics, cellular diversity, gene expression, and mutational profiles of their corresponding parental tumors, as was successfully demon-strated for breast, lung and colon cancer, as well as GBM [18,[20][21][22]40,45]. In addition, they can be generated quickly and reliably within two weeks from intraoperatively gained tissue [24]. Thus, they would be available for TTFields testing already two weeks after surgery, and results of TTFields screenings could be expected at a time when patients completed radiotherapy and could start with TTFields application. Our data show that organoids grown on OHSC respond to TTFields treatment by interpatient heterogeneous appearance and growth patterns, with TTFields causing shrinkage of the microtumors to varying extents. Thus, patient-derived GBM tumor organoids represent an ideal and flexible model to not only test personalized therapies by correlating mutational profiles with responses to TTFields, but also to investigate changes in organoid cell structures and molecular protein expression patterns caused by TTFields. Since organoids can be propagated over long time periods without changing their properties [24], they also should be suitable to investigate long-term TTFields effects in a patient related setting.
In addition to the above-mentioned paternal specifics, patient derived organotypic tumor slice cultures retain the tumor microenvironment, including neural cells, and therefore are even closer to the in vivo tumor situation of GBM patients [25]. Freshly sliced, the tissue is ready to use, but on the downside, slicing is a delicate method highly dependent on tissue quality, and is susceptible to deficiencies. Handling these cultures requires a high level of experience, and they are viable only for a couple of days. On the other hand, adjacent slices from the same tumor region can be generated and cultured for better comparison of different experimental conditions. Especially, molecular alterations can be visualized in slice cultures. As a proof of concept, we stained slice cultures for Ki67, a well-established proliferation marker in pathological evaluation of tumor tissue [35]. It was intriguing that the percentage of Ki67 positive cells in one of the investigated GBM samples increased over time despite TTFields treatment, which was expected to interfere with cell proliferation [11][12][13]. However, Ki67 staining identifies cells which have entered the cell cycle and is high in G1, S, G2 and M phase, but does not give an indication about the cells' later fate [34,35]. TTFields can cause cell cycle arrest, which might lead to accumulation of Ki67-positive cells for a certain time. Indeed, when staining the slices after 96 h instead of 72 h treatment, there was a massive drop in Ki67 positivity, probably due to Ki67 degradation and cell death. While this observation is based on only one single tumor and should not be overinterpreted, it proves the feasibility of investigating TTFields-induced molecular changes in cultured tumor tissue slices. Therefore, both organoids and tumor slice cultures have different advantages, which complement each other.
Taken together, the screening methods we describe demonstrate high feasibility for use in patients who are being considered for TTFields treatment. As has been suggested by Gilazieva et al., it is advisable to combine different methods and model systems to obtain reliable results when studying tumor response to therapies [18]. The same 3D systems may likely apply to other future treatment modalities in which the same concept would prove beneficial. Nonetheless, our focus is to provide clinicians an aid in discerning the likelihood of patients to be good candidates for TTFields therapy, for proper allocation.
Like every other study, ours has some limitations. First, the ex vivo data were not matched with the clinical data of the respective patients. Second, we performed only a small number of experiments for each system. As the main goal of this study was to prove the technical feasibility of combining patient derived organotypic ex vivo culture systems with TTFields, we now plan to continue our studies by matching experimental data with the clinical course of the patients in a prospective bench to bedside and back approach.

Conclusions
In this small study we were able to identify differences in the inter-patient treatment response not only when using PDPC cultures, but especially when utilizing patient-derived organoids grown on OHSC and tumor slice cultures. This establishes our methodology as a powerful tool to screen for patients that might benefit from TTFields treatment, as well as to elucidate cellular alterations within the cultured tissue. Last but not least, these models will shed light onto molecular mechanisms of treatment response and resistance, especially if they are used in combination.

Patents
U.S. Provisional Patent Application No. 63/409,525 is based on the work reported in this manuscript.