A modified 384‐well‐device for versatile use in 3D cancer cell (co‐)cultivation and screening for investigations of tumor biology in vitro

Abstract Pancreatic cancer exhibits a worst prognosis owed to an aggressive tumor progression i.a. driven by chemoresistance or tumor‐stroma‐interactions. The identification of candidate genes, which promote this progression, can lead to new therapeutic targets and might improve patient's outcome. The identification of these candidates in a plethora of genes requires suitable screening protocols. The aim of the present study was to establish a universally usable device which ensures versatile cultivation, screening and handling protocols of cancer cells with the 3D spheroid model, an approved model to study tumor biology. By surface modification and alternative handling of a commercial 384‐well plate, a modified device enabling (i) 3D cultivation either by liquid overlay or by a modified hanging drop method for (ii) screening of substances as well as for tumor‐stroma‐interactions (iii) either with manual or automated handling was established. The here presented preliminary results of cell line dependent dose‐response‐relations and a stromal‐induced spheroid‐formation of the pancreatic cancer cells demonstrate the proof‐of‐principle of the versatile functionality of this device. By adapting the protocols to automation, a higher reproducibility and the ability for high‐throughput analyses were ensured.

clinical testing [14] which can be attributed to the artificial character of the performed 2D-assays which barely reflect important features of solid tumors in vivo -cell-cell and cell-matrixinteractions in three dimensions [15]. A better in vitro model is the 3D spheroid, developed during cultivation of adherent cells under non-adhesive conditions, e.g. in hanging drops [16] or on agarose coated surfaces (liquid overlay) [17]. The presence of less artificial cell-cell and cell-matrix-interactions within 3D cultures [14] leads to in vivo-like growth kinetics [18], metabolic rates [19] and chemoresistances [20] and make achieved data more credible in respect of subsequent animal testing compared to 2D culture. However, implementation of 3D cultures to largescale studies is cost-intensive and time-consuming.
The aim of the present study was to establish a device with universal versatility for 3D cancer cell cultivation and screening to ensure a high degree in flexibility of choosing the perfect 3D (co-)cultivation method and screening approach of any cell sample.

Manual protocol for substance screenings in the modified 384-well-device
The substance screening should fit the following parameter: 72 h treatment of 3D cultures 24 h after cell seeding. The exponential proliferation phase of untreated cells should be ensured for 96 h. Therefor growth kinetics were preliminary performed with the LO-method. Due to its lower volume, substrate limitations occur more likely in LO compared to MHD-method. 100, 300, 500, 800, and 1000 cells per well were seeded with the respective medium supplemented w/wo doxycycline in a 4-fold dissemination. On day 4 after seeding, cultures were fed with fresh medium, in case of LO by addition of 10 μL and in case of MHD by adding 17.5 μL after 17.5 μL were aspired. CellTiterGlo R 3D assay was performed on day 2, 4, 6, and 8 after seeding as specified by the manufacturer and luminescence was measured on the Synergy H1 Multidetection Reader (BioTek, Bad Friedrichshall, Germany). Relative luminescence counts of each cell density were plotted against time in culture.
For substance screenings, cells were seeded in the determined density and the respective medium w/wo doxycycline with the LO-or MHD-method. After cultivation under standard conditions for 24 h, gemcitabine or paclitaxel (Sigma Aldrich) was dissolved with the respective medium w/wo doxycycline. Dilutions were submitted to the cells in increasing concentration (0-100 μM). In case of LO-method, 10 μL of the substance was added to each well and plates were cultivated under standard conditions. In case of MHD-method, plates were incubated the right way up for 10 min to allow cell sedimentation. 17.5 μL medium was aspirated per well and 17.5 μL of the substance was added. Plates were incubated upside down under standard conditions. CellTiterGlo R 3D assay was performed 72 h after treatment as mentioned above with the following modification: prior to the addition of assay reagent, medium was aspirated, in case of LO-method 10 μL and in case of MHD-method 25 μL. Relative luminescence units from untreated cells were adjusted to 100% and luminescence intensities from treated cells were scaled to it. IC 50 -values were calculated on the basis of linear interpolation of adjacent values.

Manual protocol for a coculture assay in the modified 384-well-device
PDAC cells were mixed with increasing amounts of HPSC (0-100%) in 20% intervals in PDAC medium w/wo doxycycline. These mixtures were seeded at a density of 1000 cells per well either with LO-or MHD-method. Plates were cultivated under standard conditions and monitored microscopically and photo documented 48 h after seeding (inverted microscope IX81 and software CellˆP, Olympus Deutschland GmbH, Hamburg, Germany).

Pipet robot and software
Automated handling was performed with the Genesis 100/8 robot and the software Genesis 4.2 (Tecan, Männedorf, Switzerland). The robot housing (Fornax, Neustadt/Wied, Germany) was completely closed and guaranteed for antiseptic conditions through disinfection by 1.5% v/v lysoformine (Dr. Hans Rosemann GmbH, Berlin, Germany) and UV irradiation. All experiments were performed with conductive, pre-sterilized tips (Fornax, Neustadt/Wied, Germany).

Adaption of substance screening and coculture assay to automated handling
The modification of the 384-well plates was adapted to automation. Coating with agarose or pHEMA was performed according to the parameters in 2.2 in part with homemade heating elements to warm up the agarose and the tips during the coating process. Except for preparation of cell suspension and gemcitabine dilutions, all steps of the protocol mentioned in 2.3 and 2.4., except for the invertation and cultivation of the plates, were adapted to automation with suitable devices under use of the liquid detection level of the robot.

Experimental procedure and valuation of the modified 384-well-device
The modified 384-well-device affords the opportunity of different 3D cell cultivation methods performed either manually or automated. An explanation of the handling followed by a numeration of the main dis-and advantages of the device is given in Fig. 1.

The modified 384-well device fulfills all requirements for 3D cell cultivation
To determine the linear cell proliferation range during the substance screening process, growth kinetics without substance were previously performed within the modified 384-well-device. A density and cell line dependent growth behavior was observed (Supporting Information Figs. 1 and 2). This is congruent to the finding that the proliferation index of 3D cultures depends on the cell type [23]. Furthermore, doxycycline had no effect on the proliferation of wildtype cells and PaTu-8988t clone II but inhibited expansion of PaTu-8988t clone I. The determination of a differential proliferation in various cell lines within the present study confirms the usability of the modified 384-well-device to identify proliferation influencing candidate genes. Another important information is that a density of 300-500 cells per well allows an exponential proliferation of every analyzed cell line at least for up to 96 h after seeding.

The modified 384-well-device ensures manual and automated substance screenings within different 3D cultivation methods
In a next step, operational capability of the modified 384-welldevice for substance screenings was tested. Experiments with PaTu-8988t wildtype cells (3D cell cultivation, microscopic determination, and substance screening) were performed manually before the protocols were implemented to automated handling.
In the present study a slight tightening of the 3D cultures in LO-compared to MHD-method was observed ( Fig. 2A and B). Zanoni already reviewed, that different 3D cultivation techniques differentially influence the shape of the cultures [24], showing the necessity of being flexible in choosing the right cultivation technique for each cell sample. By introducing the modified 384-well-device to the substance screening procedure, the MHD-method was determined as unsuitable for manual handling due to the necessity of a defined immersion depth for pipetting. Instead, the LO-method led to reproducible results when handled manually: Dose-responserelations were recorded with a mean IC 50 -value of 17.8 nM (Fig.  2C). After the automation, reproducible and comparable IC 50values could be determined with LO-(17.0 nM) as well as MHDmethod (16.5 nM), depicting that both methods are suitable for substance screenings. Furthermore, IC 50 -values from automated handling were comparable to those from manually performed experiments with the LO-method, depicting that automation did not influence the process.
Additionally, a comparison of the dose-response-relation reveals slight increased IC 50 -values of 2D (23.0 nM) compared to 3D cultures (16.5-17.8 nM; Fig. 2C). This fact disagrees with the omnipresent opinion that 3D cultures are more resistant compared to 2D cultures [25]. Among others, the limited diffusion of substances and hypoxia within tight spheroids are discussed to be involved in the drug resistance [26]. The cultures in the present study are packed very loosely, even 8 days after seeding (Supporting Information Fig. 2), so the mentioned mechanisms for chemoresistance may be nonexistent. Aside from that, it is well accepted that 3D cultures can also show an increased sensitivity compared to 2D cultures, e.g. toward specific substances like kinase inhibitors [27]. Anyway, to exclude any impracticalness of our modified 384-well-device and substance screening procedure, experiments were performed with HT29-cells which form tight spheroids 24 h after seeding (Fig. 2D). Again, comparable IC 50 -values could be observed for LO-(48.5 μM) and MHD-method (44 μM). Furthermore these values show a huge difference to the IC 50 -value from 2D culture (0.04 nM), confirming the suitability of the modified 384-well-device for substance screenings of 3D cultures. Finally, it leads to the conclusion that the shape of 3D cultures and their response to chemotherapy is cell-line-dependent and not influenced by the modified 384well-device, contrary, it helps to deal even with loosely packed 3D cultures.

The modified 384-well-device ensures substance screenings in various cell lines
In a last step the versatility of the modified 384-well-device was tested with independent doxycycline-inducible PaTu-8988t mutants. A differential influence of gemcitabine on clone I and II was detected ( Fig. 2E and F). Clone I is impaired in proliferation through doxycycline and since substance screening was performed under submission of doxycycline, the detected resistance might be an artefact due to less expansion (Supporting Information Fig. 1). Those clones can be rated as dormant cells and might therefore not be useful for the screening of substances which target dividing cells [28]. Convenient supplementary analyses of these clones can be performed with the assay presented by Wenzel et al. [29]. It is also possible that the respective mutated genes are involved in the resistance mechanisms of both clones. The underlying mechanisms will be a topic for continuative investigations, e.g. by extension of the protocols to transcriptomic and metabolic profiling of selected clones.
To sum up, with the present proof-of-principle study, the usability of the modified 384-well-device for cell proliferation and substance screenings with different cultivation methods and operation modes was confirmed. The goal of the established assay was not to detect underlying resistance mechanisms but genes with an influence on proliferation and/or chemo-response, as presented. These genes are the basis for further investigation of the molecular mechanisms of chemo-resistance and can lead to the identification of new therapeutic options.

The modified 384-well-device is suitable for cocultivation of PDAC and stromal cells in different 3D cultivation methods and lead to the identification of stromal-induced spheroid-formation
To determine the usability of the modified 384-well-device for the cultivation and screening of cocultures, mixtures from PaTu-8988t wildtype and HPSC were either cultivated with LO-or MHD-method (Fig. 3A). A comparable HPSC-induced spheroid-formation of the PaTu-8988t cultures in both cultivation methods was observed. Cultures rose in round shaped structure starting with 60% HPSC. Doxycycline had no influence on the shape of the cocultures. The stroma-induced spheroid-formation in PaTu-8988t wildtype cells can be rated as tumor-stroma-interactions comparable to previous observations in other entities [30]. Furthermore, the influence of pancreatic stellate cells on cancer progression in vivo is indisputable and justifies the usage of HPSC as stromal compartment in the in vitro protocol [31].

The modified 384-well-device is suitable for detection of cell-line dependent tumor-stroma-interactions
In a last step, the usability of our modified 384-well-device to detect a possible differential influence of HPSC on sphereformation of different PDAC was analyzed by the introduction of two doxycyclin-inducible mutants from PaTu-8988t and Suit2-007 wildtype cells as a second PDAC cell line to the protocol (Fig. 3B). In comparison to the respective wildtype cells, clone I showed a tighter and clone II a looser packed shape. The possibility of cell-line-dependent characteristics of 3D cultures was discussed previously [23]. Nevertheless, in both mutants in ac-cordance to the wildtype cells, 60% HPSC led to the compaction of the cells to a tight spheroid. Other conditions were observed with the Suit2-007 wildtype cells: an amount of 80% HPSC was necessary for the stromal-induced spheroid-formation. A differential influence of stromal cells in dependency of the cell line could be observed earlier [30].
To sum up, the modified 384-well-device is useable for the screening of different cell lines regarding their spheroidformation in the presence of stromal cells under different cultivation methods. The spheroid culture is an approved model to study tumor biology in vitro [32], and the correlation between a compact spheroid-formation and tumorigenicity in mouse models is already confirmed [33]. So the identification of a stromal-induced spheroid-formation can give a hint to in vivolike tumor-stroma-interactions. Findings from the screenings with the modified 384-well-device can act as basis for further investigations. Molecular mechanisms of the interaction can be assessed with the device e.g. by implementing feeding of supplements or potential inhibitors of the interaction.

Concluding remarks
A modified 384-well-device for the investigation of chemoresistance and tumor-stroma-interaction as key players for tumor progression was established. Compared to other commercially available devices it provides many advantages (Fig. 1). Due to the possibility of automation, a plethora of different genes can be assessed regarding their influence on tumor progression in a high-throughput manner and with a high statistic assurance. An extension of the protocols e.g. to transcriptomic or metabolomic profiling of interesting candidates is needed to identify molecular mechanisms but nevertheless, the results from the presented protocols serve as basis for this. Requirements for the availability of the protocols in personalized medicine are also given. As a logical implication to dose-response-relations of defined cocultures and due to miniaturization, low-sized samples like biopsies can be analyzed to improve patient's outcome.

Practical application
With the present study, a modified 384-well-device for investigations of tumor biology was established. The device shows versatilities regarding (i) cultivation methods, (ii) screening approaches, in detail chemo-response and tumor-stroma-interaction, and (iii) operation mode, thus experimental flexibility. Besides the practicability for 2D cultures, the device was adapted to 3D cell cultivation which better mimics in vivo situations. By applying cells of different genetic background to the protocols, different genes can be analyzed regarding their influence on tumor biology. The possibility to adapt all steps to automation ensures high-throughput analyses with a high degree of reproducibility. By introducing patient samples to the protocols, their use in therapy-associated diagnostics could improve patient's outcome in the future.