Comparison of two supporting matrices for patient-derived cancer cells in 3D drug sensitivity and resistance testing assay (3D-DSRT)

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

] and pediatric cancer[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

], and utilized such cultures for the characterization of cancer genomics and systematic drug sensitivity and resistance testing (DSRT), e.g. functional precision medicine studies. Moreover, HT drug screening has shown its potential for guiding early clinical trials with novel- and investigational agents[

[19]

Murumägi A.
et al.

STRN-ALK rearranged pediatric malignant peritoneal mesothelioma – Functional testing of 527 cancer drugs in patient-derived cancer cells.

]. Patient stratification and clinical/molecular phenotypic sub-grouping of patients has additionally shown to be influenced by HT drug screening[

[20]

Snijder B.
et al.

Image-based ex-vivo drug screening for patients with aggressive haematological malignancies: interim results from a single-arm, open-label, pilot study.

]. Such assays for drug response testing are heavily exploited for primary drug target evaluation. However, they may be of great interest also at later stages of treatment design and disease profiling when second-line treatments are needed e.g. because of developed resistance to first-line treatments. While HT-screening methods have been well-established for 2D cell cultures, this approach may in some cases have limited or misleading significance for predicting clinical responses[

21

Garnett M.J.
et al.

Systematic identification of genomic markers of drug sensitivity in cancer cells.

,

22

Singh T.
Neal A.S.
Moatamed N.A.
Memarzadeh S.

Exploring the Potential of Drug Response Assays for Precision Medicine in Ovarian Cancer.

Google Scholar

,

23

Liu L.
Yu L.
Li Z.
Li W.
Huang W.R.

Patient-derived organoid (PDO) platforms to facilitate clinical decision making.

]. As drug response rates vary between different cell models, there is an urgent call for improvement by using e.g., patient-derived primary cells that better predict tumor cell sensitivity and resistance. Alongside the increased variety of culture models, HT screening methods need to match with cell model-specific requirements. As an example, early progress in HT screens with 3D cultures have been made when using non-adherent plates[

[24]

Lenin S.
et al.

A drug screening pipeline using 2d and 3d patient-derived in vitro models for pre-clinical analysis of therapy response in glioblastoma.

,

[25]

Friedrich J.
Seidel C.
Ebner R.
Kunz-Schughart L.A.

Spheroid-based drug screen: considerations and practical approach.

]. Additionally, development of analytical tools as well as novel readouts, e.g., image-based single cell analytics, are urgently needed for understanding cell-type specific drug responses, particularly when studying more complex cell models of heterogenous diseases, such as cancer[

26

Madoux F.
et al.

A 1536-Well 3D Viability Assay to Assess the Cytotoxic Effect of Drugs on Spheroids.

,

27

Grexa I.
et al.

SpheroidPicker for automated 3D cell culture manipulation using deep learning.

,

28

Carragher N.
et al.

Concerns, challenges and promises of high-content analysis of 3D cellular models.

].

In this study, we first established, as a proof-of-concept assay, a 3D-DSRT pipeline for the hepatocellular HepG2 cell line in 384-well plates in the presence of GrowDex as a supporting matrix. Furthermore, we demonstrated how Matrigel and GrowDex can be applied in 3D culture conditions for primary OvCa PDCs and for 3D-DSRT of OvCa PDCs of two patients. PDCs were treated with a selected library of 52 approved and investigational oncological compounds in clinically relevant concentration series in various screening conditions, and the sensitivity to the drugs was measured by cell viability measurement, visualized as drug sensitivity and resistance score (DSS). Finally, we show the comparison of drug screening of OvCa PDCs in 2D vs. 3D in different matrices and discuss potential benefits and challenges of this approach for high-content (HC) screening applications.

2. Materials and Methods

2.1 Cell culture

HepG2 cells (ATCC, HB-8065) were cultured in DMEM supplemented with 10% FBS, 1x penicillin/streptomycin and 1% L-glutamine. To optimize the 3D conditions for drug testing, the HepG2 were cultured with different concentrations (0.1%-1%) of GrowDex (UPM Biomedicals, Finland) to allow the formation of spheroids at the 96-well plates and 384-well plates (7 days).

Ovarian cancer (OvCa) PDCs were obtained from low-grade serous ovarian cancer (LGSOC; ascites sample, OvCa1; FMOC02_3) and high-grade serous ovarian cancer patient (HGSOC; cancer tissue sample, OvCa2; FMOC11)[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

]. Shortly, tissue was minced into small pieces with a sterile scalpel and underwent enzymatic digestion using a Tumor Dissociation kit (Miltenyi Biotec) and gentle MACS Dissociator (Miltenyi Biotec) to obtain single-cell suspension. After centrifugation at 300 x g, red blood cells were removed from both ascites (OvCa1) and cancer tissue-derived cells (OvCa2) using Red Blood Cell Lysis Solution according to manufacturer's protocol (Miltenyi Biotec).

OvCa1 PDCs were cultured in serum-free stem cell media DMEM-F12, supplemented with 20 ng/ml EGF (Corning), 10 ng/ml FGF (Invitrogen), B27 (Thermo Fisher Scientific) and primocin (Invivogen). OvCa2 PDCs were cultured according to the protocol published by Liu et al.[

[1]

Liu X.
et al.

ROCK inhibitor and feeder cells induce the conditional reprogramming of epithelial cells.

,

[29]

ElHarouni D.
et al.

iTReX: Interactive exploration of mono- and combination therapy dose response profiling data.

]. The PDCs have earlier been characterized to identify their somatic mutations, copy number changes, and transcriptomic profile, including fusion genes[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

]. Thereby, the PDCs were confirmed to be representative models of the tumor tissues from which they were derived from[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

].

To test the spheroid formation and growth the OvCa1 and OvCa2 PDCs were cultured in four GrowDex concentrations 0.8%, 0.4%, 0.2%, and 0.1% in clear bottom 96-well ULA plate (Corning, #3474). Shortly, 10 000 cells/well were embedded in a total volume of 100 µl of diluted GrowDex in cell culture medium, and 25 µl of fresh medium was added either twice to wells (top of the matrix -cell layer), and 3 times to highest matrix concentration. Cultures were let to propagate for 14 days. The experiment was repeated with 0.2% and 0.1% GrowDex, and 10% and 5% Matrigel in 96 & 384-well ULA plates.

2.2 Immunostaining of cells and imaging

The spheroids were grown in different GrowDex concentrations, followed by 4% paraformaldehyde fixation, 0.3% Triton-X-100 permeabilization, blocking of nonspecific signal with 0.2% BSA and staining with primary and secondary antibodies at 37°C and RT, respectively, for 1 h. The markers included Hoechst33342 (nuclear stain, Molecular Probes), mAb against Ki-67 (cell proliferation, Dako), and Alexa Fluor 647 Phalloidin (F-actin, Invitrogen). For live cell fluorescence marker staining and imaging, the cells were plated in 0.1% GrowDex or 10% Matrigel on 384-well ULA plates in two different cell concentrations. For PDCs, 3000 or 6000 cells were plated/well and for HepG2 cells, 2500 cells were plated/well. The cell-matrix suspension was plated (25µL/well), centrifuged (300 x g, 3 min) and incubated (120 h at 37°C and 5 % CO₂). Remaining dyes (Hoechst 33342; Thermo Fisher Scientific, TMRE; Abcam and DRAQ7; Thermo Fisher Scientific) were added at the incubation end-point in an equal, 25 µL volume. Cells were centrifuged (at 300 x g, 20 s) and incubated for 1 h to let the dyes enter the cell-matrix suspension. Imaging was performed with Opera Phenix (PerkinElmer) confocal spinning disk microscope (10x air objective, NA 0.3 or 40x water objective, NA 1.1) or for brightfield microscopy, with Leica DM IL LED Inverted Microscope (5x/10x objective). Images were further processed for 3D with Volocity software or Harmony 4.9 (PerkinElmer) for visualization. Segmentation and quantification of spheroid regions was performed with Harmony 4.9 as described in Supp. Table 7. In short, spheroid regions were segmented from 10x magnification maximum projection (MIP) brightfield images with a linear classifier based on SER texture features. Selected regions were then quantified for their morphological features and the number of objects.

2.3 Drugs and plate layouts

HepG2 cells were screened for drug efficacy using a library consisting of 35 drugs plated in five concentrations covering a 10,000-fold range (HepG2 cells, Supp. Table 1A). The drugs were selected based on earlier results of FO4b library drug screen with 527 drugs (FO4b library, Supp Table 1B) of HepG2 cells either tested in 2D or in 5% Matrigel on pre-drugged plates. 35 drugs were selected for our targeted library: 31 had a cell viability based drug sensitivity score (DSS) value > 10 in both 2D and Matrigel -cultured HepG2 (e.g. cells were sensitive to these drugs), and four drugs (Simvastatin, sorafenib, atorvastatin, and doxorubicin) with DSS < 6 (e.g. cell viability was not affected by these drugs).

For PDCs, DSRT was performed with a selected panel of 52 drugs (OvCa PDCs, Supp. Table 1C) in 5 different concentrations at 384-well plates; for OvCa1 PDCs all conditions were performed in replicates/triplicates. Shortly, the panel of 52 drugs was designed based on study by Murumägi et al. 2022, in which a wide set of PDCs from OvCa patients (16 representative PDCs derived from 13 HGSOC, LGSOC and mucinous –MUCOC- patients) have been screened in a high throughput drug testing in 2D. The drug panel of 52 drugs includes several actionable drugs that could be repurposed in the clinic in the treatment of ovarian cancer. Shortly, the panel includes, for example, 13 chemo drugs, commonly used in the clinic, 29 kinase inhibitors, most of which are FDA approved/clinical trials, as well as rapalogs (approved), Parp inhibitors (two of which approved), and Bcl-2 inhibitors (approved).

The compounds were diluted to 100% dimethyl sulfoxide (DMSO) or water. The compounds were either transferred directly to the plates without cells (i.e. pre-drugged plates with a maximum storage of 2 months in MultiPod (Roylan Developments) system with controlled humidity) or added on top of the pre-cultured cells using Echo Acoustic Dispenser (Labcyte). The DSRT pipeline for 2D cell cultures at the High Throughput Biomedicine Unit, Institute of Molecular Medicine Finland, FIMM has been described in detail[

30

Liu X.
et al.

Conditional reprogramming and long-term expansion of normal and tumor cells from human biospecimens.

,

31

Pemovska T.
et al.

Individualized systems medicine strategy to tailor treatments for patients with chemorefractory acute myeloid leukemia.

,

32

Pietiainen V.
et al.

The High Throughput Biomedicine Unit at the Institute for Molecular Medicine Finland: High Throughput Screening Meets Precision Medicine.

]. DMSO (0.1%) was included as a negative control and benzethonium chloride (BzCl, 100 µM) as a positive control for cell viability measurement, quality control analysis and for the calculation of DSS.

2.4 Optimizing the 3D drug testing automation for HepG2 in GrowDex

To optimize the drug testing in GrowDex, pre-culturing or pre-drugging methods were tested using HepG2 cells. HepG2 cells were mixed with 0.4% GrowDex (See Fig. 1 for schematic drawing, and Supp. File 1 for the GrowDex 3D-DSRT protocol), and seeded to 384-well plates (2500 cells/well; Corning Ultra Low Attachment (ULA) 384-well plates #3827 for spheroids). For adherent cells, 750 cells were seeded/well to Corning TC-treated 384-well plates (#3764). In both cases, the cells were pre-cultured for 2 days (48 h) as spheroids in GrowDex (or monolayer in 2D) prior to adding the drugs with Echo Acoustic Dispenser (Labcyte Inc). After 72 hour incubation with drugs at 37°C and 5 % CO₂, 25 µl of CellTiter-Glo (CTG, Promega Madison, WI, USA) was added to the wells, incubated for 20 min on an orbital shaker (600 rpm), and the luminescence was detected with PHERAstar FS plate reader (BMG Labtech).

Figure 1Schematic overview of the screening pipeline used in the 3D-DSRT for HepG2 cells in GrowDex and Matrigel. Two drugging approaches were used (pre-culturing method and pre-drugging method), and the data was analyzed using our publicly available web-based Breeze software. Based on HepG2 optimization screens, the pre-culturing method was selected for OvCa PDCs for 3D Drug Sensitivity and Resistance Testing (3D-DSRT).
View Large Image
Figure Viewer
Download (PPT)

Alternatively, HepG2 cells were seeded to pre-drugged plates. Shortly, cells in growth medium were mixed gently in pre-diluted GrowDex and cell-GrowDex -suspension was dispensed to Corning ULA 384-well plates (3500 cells/well in 25 μL) using BioMek with 384-well plate compatible multichannel tips. For 2D drug testing with pre-drugged plates, 1500 cells/well were seeded to pre-drugged Corning TC-treated plates. The experiments were performed in duplicates of which the screens with results of highest Z’ were used for follow-up analysis and visualization.

2.5 Cell growth assay

The growth of HepG2 cells and OvCa2 PDCs in 3D was assessed in 0.4% GrowDex or 10% Matrigel and 0.1% GrowDex or 10% Matrigel, respectively, over the time course of 0, 24, 48, 72 and 120 h. Cells were plated on 384-well ULA plates and incubated for the designated time period. The assay was terminated at respective timepoint by addition equal plating volume (25 μL) of CTG. Plates were shaken on an orbital plate shaker for 15 min and centrifuged for 3 min at 300 x g. Luminescence counts (RLUs) were detected with PHERAstar FS plate reader (BMG Labtech). 

2.6 Drug testing in 2D and 3D for PDCs

For 2D DSRT, the PDCs were cultured as above, but seeded (750 cells/well) to pre-drugged 384-well TC plates (Corning, #3674) and incubated for 3 days at 37°C, 5 % CO₂ prior to the cell viability measurement as above.

For 3D drug testing of OvCa PDCs, 3 000 cells/well were embedded in 0.1% GrowDex or in 10% Matrigel with growth medium. 25 μL of the suspension was then transferred with Biomek FXp (Beckman Coulter) or CyBio Felix (Analytic Jena) 384-multichannel head (for cytotoxic drugs in a separate experiment) to 384-well ULA plates (Corning, #3827). The protocol for GrowDex-based 3D-DSRT is included in Supp. File 1. For Matrigel -based 3D-DSRT, see Feodoroff et al. for a detailed protocol[

Feodoroff M.
et al.

Protocol for 3D drug sensitivity and resistance testing of patient-derived cancer cells in 384-well plates.

]. Air bubbles were removed from wells with a slow centrifugation (300 x g, 20 s). The plates were incubated 48 h at 37°C, 5 % CO₂ prior to addition of drugs with Echo Acoustic dispenser as described above. The plates were then gently shaken with an orbital plate shaker for 1 min and the cells were incubated with drugs for 72 h, prior to the addition of CTG for cell viability measurement as described above.

For a separate cytotoxic drug screening, PDCs with matrices were seeded to 384-well plates (6 replicates for each drug concentration in dublicate plates; n=12 for each drug concentration), and drugs (bortezomib, romidepsin, CUDC-907, and sepantronium bromide) were added for 72 h prior to the cell viability measurement, as described above (see Fig 1., pre-cultured 3D-DSRT protocol scheme).

2.7 Quality Control, Data Analysis and Statistics

Raw plate reader output data were run through the in-house R scripts (available at GitHub) and later run through Breeze analysis pipeline (https://breeze.fimm.fi) for quality control (QC) assessment, including Z prime, and SE/SDs for positive and negative controls, dose response quantification and visualizations. The dose response curves were fitted for each compound and the DSS were defined as described earlier[

[33]

Kulesskiy E.
Saarela J.
Turunen L.
Wennerberg K.

Precision Cancer Medicine in the Acoustic Dispensing Era: Ex Vivo Primary Cell Drug Sensitivity Testing.

]. The heatmaps were created using Morpheus (https://software.broadinstitute.org/morpheus/), and the clustering was based on one minus Pearson correlation with average linkage method.

Standard deviation calculations and correlation analysis were performed using GraphPad Prism 9 (GraphPad Software, La Jolla, California, USA). Data was correlated between selected datasets with Pearson correlation coefficients. Pearson r is indicated in each graph. Confidence interval was set at 95%. Analysis of dose responses was performed by initial log-transformation of the concentration range and followed by nonlinear regression analysis with a confidence interval set at 95%. The graphs were created using BioRender, GraphPad Prism 9 or scripts in the Breeze.

3. Results

3.1 Optimized 3D drug testing for HepG2 cell spheroids in GrowDex hydrogel

The drug screening assay in the GrowDex hydrogel was first set-up using a commercially available and widely studied hepatocellular carcinoma cell line HepG2, shown to form spheroids in different matrices[

Feodoroff M.
et al.

Protocol for 3D drug sensitivity and resistance testing of patient-derived cancer cells in 384-well plates.

,

Yadav B.
et al.

Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies.

], including GrowDex[

Yadav B.
et al.

Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies.

]. First, we confirmed the ability of HepG2 cells to form spheroids in different GrowDex concentrations (0.125-0.25-0.5-1%) on 96-well plates (Fig. 2A). As the spheroids with a regular shape were formed in 0.25% GrowDex, we further titrated the optimal GrowDex concentration in 0.2% and 0.4% GrowDex in 384-well plates (Fig. 2B). 0.4% GrowDex concentration was selected for drug testing as it was still accurately delivered using robotic handling (our unpublished data) and the cells formed clear spheroid structures in this concentration, as in Fig. 2A-B.

Figure 2Optimization of 3D-DSRT assay for HepG2 cells in GrowDex hydrogel. HepG2 cells were cultured as spheroids A. in 0.125-0.25-0.5-1% GrowDex in 96-well plates and B. in 0.2% and 0.4% Growdex in 384-well ULA plates. Cells were imaged with a light microscope, 10x objective (A; scale bars 100μm) and 5x objective (B; scale bars 100 μm). C. Summary of the DSRT quality control measurement results for six different conditions compared for HepG2 (HepG2 Adherent precultured = HepG2 adherent pre-cultured 2D DSRT, HepG2 Adherent 72 h/5 days = HepG2 adherent pre-drugged 2D-DSRT for 72 h and 5 days; and HepG2 GrowDex precultured = HepG2 spheroids in pre-cultured 3D-DSRT in GrowDex, HepG2 GrowDex 72 h/5 days = HepG2 in Growdex, pre-drugged 3D-DSRT for 72 h and 5 days; see also Fig. 1). The Z’ revealing the separation between the distributions of the positive and negative controls, was > 0.7 for all drug plates, demonstrating high-quality of the screens. CVs for DMSO (negative control) were <10. BzCl was used as a positive control (high CTG values). (CV=The coefficient of variation; SSMD = strictly standardized mean difference; data from the Breeze. The table was created with BioRender.com.) D. Heatmap showing the column- and row-wise hierarchical clustering of drug responses based on the DSS values of HepG2 cells. The pre-cultured spheroids cluster separately but closely with adherent cell co-cultures based on the DSS values. The DSS values of pre-drugged screens (in which the cells have been added directly on top of the drugs) cluster under the same main branch of adherent and GrowDex -screens. The length of incubation (72 h or 5 days) of drugs with cells does not have a major impact on DSS values. The bar on top: red signal = high DSS/sensitivity to the drug; white signal: low DSS/ no sensitivity. E. HepG2 cells grown in GrowDex, imaged with 40x objective using Opera Phenix confocal microscope. Live cell staining was performed at 120 h timepoint using Hoechst for nucleai, TMRE for active mitochondria and DRAQ7 for nucleai of non-viable cells. Images from separate fluorescent channels for each marker as well as a merged fluorescencet image are represented here (scale bars 50 μm).
View Large Image
Figure Viewer
Download (PPT)

For optimization screens with HepG2 cells in GrowDex, 35 drugs (each in 5 concentrations) were selected based on their efficacy (DSS >10) in earlier 2D and 3D (5% Matrigel) DSRT (Supp. Table 1B, Supp. Table 2 for 2D DSRT & Supp. Table 3 for 5% Matrigel-3D-DSRT). Here, our in-house 2D-DSRT was used as quality reference[

[38]

He L.
et al.

Methods for high-throughput drug combination screening and synergy scoring.

]. We tested different conditions and timepoints in parallel in both 2D (1-3) and 3D-DSRT (4-6) on 384-well plates (see Fig. 1): 1) Adherent Pre-cultured): cells seeded 48 h prior to adding the compounds, followed by 72 h incubation with the drugs (total assay time, 5 days), 2) Adherent 72 h: cells added on top of pre-drugged plates for 72 h, 3) Adherent 5 days: cells added on top of pre-drugged plates for 5 days, 4) GrowDex Pre-cultured: cells in matrix seeded 48 h prior to adding the compounds, followed by 72 h incubation with the drugs, 5) GrowDex 72 h: cell suspension in matrix added on top of pre-drugged plates for 72 h and 6) GrowDex 5 days: cell suspension in matrix added on top of pre-drugged plates for 5 days. The cell viability measurement (luminescence counts; based on cellular ATP amount), was performed at the end of the DSRT assays. The data was analyzed using in-house Breeze pipeline that included plate-specific heatmaps, graphs of column-wise signal distribution as well as other values for quality control of each assay plate. These include e.g. the Z-Prime (or Z’) factor (Z’ factor), a widely-used measure of signal to noise (S/N) ratio. It shows the separation between the distributions of the positive (here BzCl), and negative (here DMSO) controls[

[36]

Ramaiahgari S.C.
et al.

A 3D in vitro model of differentiated HepG2 cell spheroids with improved liver-like properties for repeated dose high-throughput toxicity studies.

]. Typically, the HT screening assay should have Z’ > 0.5. As shown in Fig. 2C, the quality control analysis revealed the Z’ > 0.70 for the HepG2 2D/3D-DSRT assays, suggesting that they were robust and of good quality.

The dose-response curves, IC50 values and DSS values for HepG2 screens are shown in Supp. Table 4 and exemplified in graphs in Supp. Fig 1D. It was observed that if a drug was effective in one condition, the response was typically similar in other conditions, despite of small variations in the DSS values. The most effective drugs, as shown based on DSS values > 10, were mainly similar for HepG2 cells screened either in pre-cultured or pre-drugged conditions (72 h). Drugs with DSS values > 10, depending also on the shape of the dose-response curve and IC50 values are typically considered effective[

[15]

Malani D.
et al.

Implementing a Functional Precision Medicine Tumor Board for Acute Myeloid Leukemia.

,

[17]

Saeed K.
et al.

Clonal heterogeneity influences drug responsiveness in renal cancer assessed by ex vivo drug testing of multiple patient-derived cancer cells.

,

Yadav B.
et al.

Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies.

].

However, in the pre-culture conditions, the drug efficacy was generally lower (Supp. Fig. 1A-C). Few drugs such as omacetaxine (conventional chemotherapeutic, protein synthesis inhibitor) and LY3999120 (a pan-RAF inhibitor) were less effective in screens performed with pre-cultured cells. In contrast carfilzomib (proteasome inhibitor) and alespimycin (HSP90 inhibitor) were less effective only in HepG2 cells pre-cultured as spheroids in GrowDex.

The clustering (based on one minus Pearson correlation with average linkage method) of different DSRT conditions based on DSS values was performed (Fig. 2D). Both pre-cultured conditions clustered separately, and pre-drugged conditions clustered under the same main branch. This branch was further divided to the pre-drugged adherent and GrowDex screens. Cells pre-cultured in GrowDex showed the smallest correlation with any DSRT screens with adherent cells (correlation values 0.46-0.56).

For drug testing of PDCs in 3D-DSRT conditions we selected pre-cultured conditions, where the spheroids are already formed by the time drugs are added. In this pre-culturing 3D-DSRT set-up, we further evaluated the effects GrowDex on the growth of HepG2 cells by measuring the cell viability in 384-well plates at five different time points (0-24-48-72-120 h, Supp. Fig 1D). We also tested the 10% Matrigel as a comparison. The growth rate of HepG2 cells in both 0.4 GrowDex and 10% Matrigel increased until 72 h, quantified by the ATP present, which indicates the presence of metabolically active cells (CTG measurement). After 48 h the growth of spheroids was slowing down in Matrigel while it still increased in GrowDex (Supp. Fig 1D).

To determine the amount of the spheroids/well and their diameter, HepG2 cells were grown in 384-well plates for 5 days in 0.4% GrowDex prior to the imaging them with HC microscope in a bright-field mode. The image analysis showed that one well consisted of 107±6.4 spheroids, and their average diameter was 82±7.7µm (pre-culturing protocol, 5 days, n=images covering 5 × 384-well plate wells). The spheroids grown in GrowDex were stained for nuclear marker (live and dead cells), DRAQ7, which indicates non-viable cells and TMRE (sequestered by active mitochondria). While few DRAQ7+ cells were observed, the TMRE+ staining of cells in spheroids indicated that they had active mitochondria, thus confirming the viability of the stained spheroids (Fig 2E).

Importantly, the results confirmed that GrowDex can be accurately delivered using robotics to 384-well plates when compared to our conventional DSRT with adherent cells, and could therefore be utilized for 3D-DSRT with PDCs on 384-well plates.

3.2 Optimizing the formation of OvCa PDC spheroids in different matrices

To set up the DSRT platform for functional precision medicine, in which the PDCs are utilized in drug screening to discover the patient -specific drug sensitivities, we utilized OvCa PDCs established from another study[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

], in which they have been genetically characterized to match with the mutational background of the original tumor they were derived from. First, we tested different culture conditions for OvCa PDCs at 96-well plates (Supp. Fig. 2A), and then in 384-well plates (Fig. 3A; Supp. Fig. 2B, with different time points). The PDCs formed spheroids in 5% and 10% Matrigel and both GrowDex conditions (0.1% and 0.2%) in their specific PDC culture media in both 96- and 384-well plates (see Materials and Methods). Based on these experiments, 10% Matrigel and 0.1% GrowDex were selected for our 3D-DSRT of OvCa PDCs.

Figure 3Optimization and performance of 3D-DSRT for OvCa PDCs. A. Microscopic images of OvCa2 PDCs grown in 0.1% GrowDex and 10% Matrigel on 384-well plates at 48 h timepoint when drugs are added in the pre-culturing 3D DSRT method. PDCs were imaged with 10x objective of a light microscope (scale bars 100 μm; zoom-in images are shown). B. The DSRT quality control measurement results for replicates^*) of OvCa1 drug screens and OvCa2 screens. Z’ > 0.64 was obtained for all screens. (CV= the coefficient of variation; SSMD = strictly standardized mean difference; data from the Breeze. The table was created with BioRender.com.) C. Heatmap showing the column- and row-wise hierarchical clustering of drug responses based on the DSS values of OvCA PDC (52 drugs). The DSS values of adherent cultures of both patient samples (e.g. 2D-DSRT) cluster separately from 3D-DSRT results. The bar on top: red signal = high DSS/sensitivity to the drug; white signal: low DSS/no sensitivity. D. Selected drugs showing difference in 2D vs. 3D-DSRT. While differences in DSS values between PDCs from two different patients were observed, some kinase inhibitors were in general more effective in 3D-DSRT than in 2D-DSRT, while PDCs in 2D-DSRT were more sensitive to conventional chemotherapeutics. Results of replicate follow-up screens (average DSS), for comparison of drug responses in different DSRTs (2D, 3D; Matrigel/GrowDex) of OvCa1 and OvCa2. Here, the DSS of 3D screens (Matrigel, GrowDex) were compared to the earlier measured DSS in 2D.  E. Dose response curves of OvCa1 to a selected panel of drugs including conventional chemotherapeutics and MEK1/2 inhibitors. Different OvCa1 PDC culture conditions are combined in each graph. Concentration ranges of each drug (**Supp. File 2**.) have been log-transformed prior to performing the non-linear regression. Confidence level was set at 95% for each regression analysis.
View Large Image
Figure Viewer
Download (PPT)

3.3 Drug testing of OvCa PDCs in 3D matrices

We tested the sensitivity of OvCa1 and OvCa2 PDCs to a selected panel of 52 drugs (Supp. Table 1C) in 2D (adherent) and 3D assay as optimized (10% Matrigel; 0.1% GrowDex). First, PDCs were seeded to 384-well plates and pre-cultured in 2D (adherent) or 3D conditions (0.1% GrowDex or 10% Matrigel) for 48 h to allow spheroid formation prior to addition of drugs for 72 h. The drugs were then added with an acoustic dispenser in five different concentrations, covering the physiologically relevant concentration range, to 384-well plates with pre-cultured PDCs (see Fig. 1, pre-culturing method). Similarly to HepG2 drug screening, DMSO with cells was used as a negative control, and a generally toxic drug, Benzethonium Chloride (BzCl; with cells) was used as a positive control for visualization and statistics of the quality of the screens (Fig. 3B, Supp. Fig. 3A). The DSRT data was analyzed using the Breeze software for the QC and the calculation of DSS values[

[37]

Malinen M.M.
Palokangas H.
Yliperttula M.
Urtti A.

Google Scholar

]. Based on the QC analysis, Z prime (Z', or Z factor) values were > 0.64 for all screens (Fig. 3B), indicating good quality for HT screening. The technical replicates of OvCa1 screens performed relatively robustly in the assay, with Z' values of 0.75 and 0.75 (GrowDex) and 0.65 and 0.76 (Matrigel) (Fig. 3B). To confirm the quality of the 3D-DSRT in GrowDex, the additional OvCa1 screen was performed, (OvCa1 PDC 0.1% GD Screen 3; Z’ 0.73, SSMD 11 and CV for cells in DMSO 9.1). All DSS values of each drug (average, mean, and SDs) are separately listed for all OvCa1 3D-DSRT screens in Supp. Table 5.

Hierarchical clustering of DSS values of OvCa PDCs screened in 2D (adherent cells) or in 3D (Matrigel/GrowDex) clustered to 2D and 3D -condition specific clusters, and within 3D conditions, to the patient-specific clusters. The average DSS values of the replicate screens performed for OvCa1 were included (see also Supp. File 5 for triplicates in GrowDex 3D-DSRT). Internal correlation between OvCa1 replicate screens was calculated with Pearson correlation at 95% confidence interval. Pearson r was 0.9949, 0.9821 and 0.9806 for adherent, 10% Matrigel and 0.1% GrowDex screens, respectively. Clustering of all the data (Supp. Fig. 3B) indicates similar clustering of OvCa1 PDC replicates in 3D-DSRT. The data of the screens, including the EC50, IC50 and DSS values as well as graphs for each drug are available in Supp. File 2.

We also systematically compared the DSS-based drug sensitivities of PDCs in GrowDex or Matrigel 3D-DSRT to each other or to those obtained in 2D-DSRT (adherent conditions). OvCa1 PDCs were highly resistant to most of the tested compounds, but showed patient-specific sensitivities (DSS >10) in both 2D and 3D-DSRT to navitoclax (2D and 3D-DSRT), tanespimycin (HSP90 inhibitor) and apitolisib (PI3K/mTOR inhibitor) as well as to topotecan (Topoisomerase I inhibitor/Camptothecin analog) and gemcitabine (antimetabolite, nucleoside analogue). OvCa1 PDCs showed high sensitivity to EGFR inhibitors, especially in 3D-DSRT with Matrigel, afatinib being a highest scored EGFR/ERBB2 inhibitor (DSS 20.3). Of MEK1/2 inhibitors, refametinib was the most effective in both 3D matrices (DSS 11.9 for GrowDex and 19.1 for Matrigel). Interestingly, mitotic inhibitors, vinorelbine and paclitaxel were more effective in 2D- than 3D-DSRT. The same profile was observed with DSRT assays of OvCa2 PDCs, where many conventional chemotherapeutics (docetaxel, paclitaxel and vinorelbine) were more effective in 2D than 3D-DSRT, whereas MEK1/2 inhibitors, such as pimasertib and refametinib, showed higher DSS values in 3D than 2D-DSRT (Fig. 3C-E, Supp. Fig. 2 C and Supp. Table 5 for OvCa PDC 3D-DSRTs). With OvCa2, the main hits in 3D-DSRT included MEK1/2 inhibitors pimasertib, refametinib and selumetinib, as well as PI3K/mTOR inhibitor apitolisib, with minor differences in DSS values of Matrigel vs. GrowDex 3D-DSRT. Instead, multi-kinase inhibitor dasatinib and HSP90 inhibitor tanespimycin were more effective in Matrigel 3D-DSRT.

To evaluate the effect of different culture conditions on drug sensitivities and resistance, we compared the DSRT results from 2D and 3D Matrigel/GrowDex screens based on DSS values (Fig. 4A-F). The Pearson correlation coefficients between GrowDex 3D-DSRT and 2D-DSRT were 0.6664 for OvCa1, and 0.4128 for OvCa2 (Fig. 4A-B). The corresponding correlation values for Matrigel 3D-DSRT and 2D-DSRT were 0.4544 for OvCa1 and 0.4964 for OvCa2 (Fig. 4C-D). For OvCa1 PDC DSRT assays, the correlation of Matrigel-3D-DSRT vs. GrowDex-3D-DSRT was 0.9392, whereas for OvCa2 PDCs it was 0.6849 (Fig. 4E, F). The results of OvCa1 screens show higher correlation values between adherent and GrowDex culture conditions than with Matrigel, whereas OvCa2 showed similar correlation values between adherent cultures and both 3D culture methods.

Figure 4Comparison and correlation of OvCa PDCs in 2D (adherent) -DSRT and 3D-DSRT in 0.1% GrowDex and 10% Matrigel. A-D: Graphs show the DSS value -based comparison of OvCa1 and OvCa2 PDCs exposed to the drugs in 3D-DSRT, either 0.1% GrowDex or 10% Matrigel vs. adherent 2D-DSRT culture conditions. The r value indicated the Pearson correlation coefficients of DSS values of two compared screens. The tip hit drugs as well as drugs showing differences between two compared screens are annotated in the graphs. Afatinib, used efficiently in the clinics to treat the patient based on functional precision medicine approach
[
[18]
Murumägi A.
et al.
Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.
*British Journal of Cancer.* 2022; 2022: 1-13https://doi.org/10.1038/s41416-022-02067-z
Crossref

Scopus (1)

Google Scholar
]
is highlighted. Graphs in E and F show similar comparison of DSS values obtained by using two different matrices, GrowDex or Matrigel in 3D-DSRT assay. G. The number of OvCa2 spheroids/well and or the spheroid diameter (μm) in 0.1% GrowDex or 10% Matrigel at two different cell amounts in 384-well plate assay (5 days). The measures were obtained from HC brightfield microscope images by quantitative image analysis (see Materials and Methods), n= 5-6 wells/analysis. H. Representative 40x fluorescence microscopy live cell images of OvCa2 PDCs in 0.1% GrowDex and 10% Matrigel at 120 h timepoint. PDCs were stained using fluorescent Hoechst, TMRE and DRAQ7 labels. Brightfield image, single fluorescence channel and merged images are shown. (Scale bar 50 μm).
View Large Image
Figure Viewer
Download (PPT)

Finally, to evaluate the effect of generally cytotoxic drugs, which were not originally included in our library of 52 drugs, OvCa PDCs were treated with 5 concentrations of bortezomib (proteasomal inhibitor), CUDC-907 (HDAC1/2/3/10, PI3Kalpha inhibitor), romidepsin (a selective inhibitor of histone deacetylase) and sepantronium bromide (survivin inhibitor) in Matrigel and GrowDex. The dose responses of these drugs, shown in Supp. Fig. 4A, were mostly similar between matrices, and PDCs in 3D-DSRT were highly sensitive to these drugs (2 replicates with 12 data points/drug concentrations are shown). In Supp. Table 6, we further provide statistics of obtained values for all five the concentration ranges for every cytotoxic compound and control. This indicated that spheroids in 3D-DSRT are sensitive to the generally cytotoxic drugs, although the matrices could interfere with the drug effects.

The amount and the size of the spheroids (OvCa2 PDCs on 384-well plates) in both 3D-DSRT conditions were evaluated from brightfield images, as these features could also affect the compound efficacy. The number of spheroids/well and the spheroid diameter for two different cell concentrations (3000 and 6000 cells/well) of OvCa2 PDCs in GrowDex and Matrigel are shown in Fig. 4G. Cells grown in GrowDex (3000 cells/well as used in the 3D-DSRT) were observed to form smaller spheroids, which were less uniform in size (66.7 mm ± 21.6 diameter) than those cultured in Matrigel (86.4 mm ± 3.8 diameter). Also, there were app. 2x less spheroids in GrowDex than Matrigel when 3000 cells/well were used (see Fig. 4G). The variation of spheroid size in GrowDex could possibly be due to some single PDCs not assembled into the spheroids but remaining as single cells with a low capacity to divide. Addition of more cells (6000 vs 3000 cells/well) did not greatly affect spheroid size in GrowDex (1.1 x diameter) while in Matrigel the size increased upon the increased cell amount (1.4 x diameter) (Fig. 4G). Here, we also evaluated the viability of OvCa2 PDCs in both matrices (0-5 days viability measurement). In Matrigel, the PDC spheroids were growing for 5 days, while in GrowDex their viability reduced, suggesting that some PDCs may not survive well in GrowDex or become senescent (Supp. Fig. 5A). To ensure that the spheroids remain viable in the optimized assay for longer periods of time, we stained them with live stains; TMRE (active mitochondria), DRAQ7 (dead cell marker) and Hoechst (nuclear marker) (Fig. 4H). OvCa2 PDCs grown in both matrices showed the presence of TMRE+ cells, while also some DRAQ7 positive dead cells were observed, mostly loosely around the spheroids, further confirming that the spheroids are still metabolically active at 5 days.

Spheroids grown in 0.1% GrowDex were also fixed and stained with Phalloidin (for actin) and Ki-67 for detection of proliferative cells in a low-throughput proof-of-concept experiment (Supp. Fig. 5B). Ki-67 positivity was observed, suggesting that PDCs can retain their proliferative capacity in GrowDex, when assembled to small spheroids.

Altogether, these results demonstrate differences in drug responses performed in 2D vs. 3D conditions, which should be taken into account in any type of profiling of drug sensitivities of cell lines or PDCs, both in drug discovery, and functional precision medicine. They also support the use of GrowDex as a matrix in the 3D-DSRT assays, similarly to a widely used Matrigel.

4. Discussion

In this study, we first established the robust 3D-DSRT assay with automated cell seeding and drug addition for HepG2 cells in GrowDex hydrogel. The 3D-DSRT assay, which included spheroid formation prior to drug addition, was then used to systematically screen sensitivities of PDCs for two OvCa PDC patients against a panel of 52 drugs. We compared the results of 3D-DSRT performed in the presence of either of two matrices, animal-free GrowDex hydrogel or animal-derived Matrigel, to those of adherent 2D-DSRT cells. As a conclusion, both matrices can support 1) the formation of PDC spheroids and 2) the robust automated 3D-DSRT assay. The results of drug sensitivities of OvCa PDCs, on top of the patient-specific efficacies, suggested that certain drugs are selectively effective either in 2D or 3D conditions.

As HepG2 cell line has earlier been shown to form metabolically active spheroids in hydrogels in several studies[

Yadav B.
et al.

Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies.

,

[38]

He L.
et al.

Methods for high-throughput drug combination screening and synergy scoring.

,

[39]

Zhang J.-H.
Chung T.D.
Oldenburg K.R.

A Simple Statistical Parameter for Use in Evaluation and Validation of High Throughput Screening Assays.

], we used this cell line for setting up the assay both as a conventional DSRT (pre-drugged plates), as well as for testing the 3D set-up for GrowDex and Matrigel[

Feodoroff M.
et al.

Protocol for 3D drug sensitivity and resistance testing of patient-derived cancer cells in 384-well plates.

]. Drug responses of adherent HepG2 cells (2D) and HepG2 cells grown in GrowDex (3D) to 35 investigated drugs when measured as cell viability values (CTG), reflecting the cellular ATP amount, were mostly similar based on DSS values. This could propose that GrowDex as such does not greatly interfere or slow down drug effects (e.g., drug uptake to the spheroids, or solubility of the drugs) compared to 2D culture conditions. We also observed that drugs generally considered to be toxic, such as CUDC-907, were active in low concentrations in both HepG2 cells (35 drugs screen) and OvCa PDCs in 3D-DSRT (cytotoxic drug screen). However, specific studies to measure the drug uptake need to be performed, especially if large spheroids/organoids are screened, to confirm the effiecient drug uptake in the presence of matrix. Also, effects of higher GrowDex concentrations as well as cell types used (e.g. properties of spheroids formed) should be further studied. Similar drug responses observed in 2D and 3D-DSRT of HepG2 cells may partly be due to the limited set of drugs tested, but as well as to the adaptation of HepG2 cell line to culture conditions, and therefore a limited effect of matrix addition.

Although HepG2 cells growth in GrowDex was faster than in Matrigel, imaging revealed that the spheroids contained metabolically active cells in both matrices (TMRE staining of mitochondria), and while some dead cells (DRAQ7 staining) were observed, they mostly surrounded the spheroids. However, the spheroids in GrowDex were much smaller than in Matrigel, likely due to the absence of the cell-matrix interactions in GrowDex, which works more like a physical support for spheroids.

Different cell lines or PDCs require preliminary optimization of GrowDex concentration to identify optimal conditions suitable for a particular sample[

[8]

Lou Y.R.
et al.

The Use of Nanofibrillar Cellulose Hydrogel As a Flexible Three-Dimensional Model to Culture Human Pluripotent Stem Cells.

,

40

Potdar S.
et al.

Breeze: an integrated quality control and data analysis application for high-throughput drug screening.

,

41

Li Y.
Kumacheva E.

Hydrogel microenvironments for cancer spheroid growth and drug screening.

,

42

Thimm, G., Gawlitta-Gorka, E., Sorg, G. & Flotow, H. High Throughput Cytotoxicity Testing with HepG2 Cells Grown in 3D Culture | UPM Biomedicals. https://www.upmbiomedicals.com/resource-center/application-notes/high-throughput-cytotoxicity-testing-with-hepg2-cells-grown-in-3d-culture/.

Google Scholar

,

43

Rantala, J. & Paasonen, L. Solid tumor derived cell line BT474 and ascites metastasis derived cell line COLO205 cultures in GrowDex | UPM Biomedicals. https://www.upmbiomedicals.com/resource-center/application-notes/solid-tumor-derived-cell-line-bt474-and-ascites-metastasis-derived-cell-line-colo205-cultures-in-growdex/.

Google Scholar

,

44

Meng, Y., Sheard, J. & Bashford, A. High-content quantitation of cancer stem cells from a glioblastoma cell line cultured in 3D using GrowDex-T hydrogel | Molecular Devices. https://www.moleculardevices.com/en/assets/app-note/dd/img/high-content-quantitation-of-cancer-stem-cells-from-glioblastoma-cell-line-cultured-in-3d-using-growdex-t-hydrogel.

Google Scholar

]. In our study, two OvCa PDCs from two patients were observed to form spheroids in a very low concentration (0.1%) of GrowDex. For example, patient-specific matrix preferences (GrowDex/Matrigel) have been observed e.g. for clear cell renal carcinoma PDCs (Pietiäinen et al., unpublished). However, this may not be dependent only on the properties of GrowDex but from what type of tissue the cells are derived from, thereby requiring specific matrix protein-interactions or growth factors to form spheroids or organoids[

[43]

Rantala, J. & Paasonen, L. Solid tumor derived cell line BT474 and ascites metastasis derived cell line COLO205 cultures in GrowDex | UPM Biomedicals. https://www.upmbiomedicals.com/resource-center/application-notes/solid-tumor-derived-cell-line-bt474-and-ascites-metastasis-derived-cell-line-colo205-cultures-in-growdex/.

Google Scholar

]. In addition, with primary cells, such as PDCs used here, it is important to aim to screen as early passages as possible. We have observed that PDCs can become senescent already in relatively low passages or after thawing them for re-culture - their ability to form spheroids could therefore also reduce. In addition, the PDCs have a tendency to mutate in a long-term cell culture, pontentially causing differences in the drug testing results compared to those of PDCs of the early cell culturing passages.

A 3D-DSRT pipeline was further established for OvCa PDC spheroids in Matrigel[

Feodoroff M.
et al.

Protocol for 3D drug sensitivity and resistance testing of patient-derived cancer cells in 384-well plates.

], and GrowDex. Our assay quality assessment showed that similarly to HepG2 cell line, the PDCs can be accurately delivered to 384-well plates with both GrowDex and Matrigel using robotics in a 3D-DSRT assay. We also observed that room temperature handling of GrowDex made it easier to perform the assay in HT manner than with Matrigel, where ice and cold blocks need to be used with automation (see Supp File 1). The comparison of the DSS values between the replicate screens (OvCa1 PDCs) revealed a strong correlation, implicating high reproducibility and robustness of our assay. Furthermore, both matrices used in our 3D-DSRT assay for OvCa PDCs with 52 drugs resulted in relatively well-correlating drug response patterns (DSS value comparison), especially in case of OvCa1, suggesting the similar suitability of both widely utilized Matrigel, and GrowDex hydrogel for drug testing in 3D when the automation and assay quality are considered. Importantly, live imaging of OvCa spheroids – similarly to HepG2 cells, showed that they are metabolically active, although some dying/dead cells were observed in 5 days as well as the reduction of growth rates, here in GrowDex. Both matrices have their limitations and benefits: Here, Matrigel was better supporting the cell growth and formation of larger spheroids with more homogenous size, whereas GrowDex in low concentrations was more robust to use in the automation needed for HC/HT screening. Larger comparative screens and PDCs derived from other cancer types are required to further study the potential differences in targeted drug responses in different 3D growth supporting matrices as has been shown for example for pancreatic cancer[

[45]

Rinner B.
et al.

MUG-Mel2, a novel highly pigmented and well characterized NRAS mutated human melanoma cell line.

].

When we compared the drug sensitivities of OvCa PDCs grown in GrowDex or Matrigel to those obtained in our 2D-DSRT set-up with pre-plated cells (adherent conditions on TC polystyrene plates[

[32]

Pietiainen V.
et al.

The High Throughput Biomedicine Unit at the Institute for Molecular Medicine Finland: High Throughput Screening Meets Precision Medicine.

]), we observed that some of the mitotic and microtubule inhibitors (such as docetaxel targeting cell division) were more effective in 2D. It is likely that drugs targeting the mitotic pathways give a weaker response in 3D due to slower cell proliferation and growth compared to 2D cultures. Similar comparison of 2D vs 3D drug testing of urachal adenocarcinoma showed relatively many drugs being less effective in 3D[

[7]

Mäkelä R.
et al.

Ex vivo modelling of drug efficacy in a rare metastatic urachal carcinoma.

]. Similarly, the non-small cell lung cancer (NSCLC) cell line Colo699 has shown reduced response to vinorelbine in 3D (microtissues) vs 2D[

[46]

Munne P.M.
et al.

Compressive stress-mediated p38 activation required for ERα + phenotype in breast cancer.

]. In turn, certain kinase inhibitors, in particular MEK1/2 inhibitors, were more effective in 3D conditions (DSS difference 2-fold). Higher sensitivity to MEK inhibitors have also been seen in 3D colon cancer models compared to planar cells[

[47]

Malinen M.M.
et al.

Differentiation of liver progenitor cell line to functional organotypic cultures in 3D nanofibrillar cellulose and hyaluronan-gelatin hydrogels.

]. Drugs with certain functional mechanisms may affect differently depending on whether they are tested on adherent 2D cultures or 3D spheroids[

[45]

Rinner B.
et al.

MUG-Mel2, a novel highly pigmented and well characterized NRAS mutated human melanoma cell line.

]. Interestingly, hydrogel and animal-derived matrices used in our 3D-DSRT provide completely different type of microenvironment and support for cells, as spheroid formation is mostly mechanically supported is obtained by hydrogel, while both mechanical and matrix-cell interaction are supported by Matrigel. This was also observed in our assay-set up, regarding the amount and size of the spheroids formed.

To fully evaluate whether 3D-DSRT in different matrices in comparison to 2D-DSRT can give more physiologically relevant results, the effects of the same drugs potentially used to treat a particular patient would need to be systematically followed. In our study, we showed that OvCa1 PDCs were in our 3D set-up - particularly with Matrigel, sensitive to EGFR inhibitors, such as afatinib (a dual EGFR and ERBB2 inhibitor). Based on our functional precision cancer medicine approach, including the in-house DSRT of PDCs of OvCa1 patient, the patient was treated in clinics with afatinib and has showen a durable response to the treatment[

Murumägi A.
et al.

Drug response profiles in patient-derived cancer cells across histological subtypes of ovarian cancer: real-time therapy tailoring for a patient with low-grade serous carcinoma.

].

Whether the 3D-DSRT is an absolute requirement for some drugs, such as MEK inhibitors, to reveal their physiologically relevant activity is an important issue to point out in future studies with larger drug libraries than in this study. While in our study the same growth media was used in all DSRT assay conditions for PDCs from the same patient, the impact of the media used in the DSRT has also to be taken into account. Different nutrients and supplements in the cell media affect for example on cell proliferation, and may affect drug responses - therefore, the effects of media currently used for establishing and drug testing of any patient -derived organoid or other type of PDC cultures in functional precision medicine should be fully evaluated[

[48]

Hou S.
et al.

Advanced Development of Primary Pancreatic Organoid Tumor Models for High-Throughput Phenotypic Drug Screening.

].

The requirements of matrix proteins differ between cell types, and therefore experiments with other cell lines or primary cells/PDCs need to be optimized case by case. Drug responses of adherent (2D) OvCa PDCs and OvCa PDCs grown in 3D revealed remarkable differences in the DSRT assay with 52 drugs. The systematic follow-up studies evaluating the effects of 3D-matrices, growth media, and assay set-up (such as U-bottom, ULA multi-well plates or microfluidics platforms) on PDCs derived from different tumor types, are required to find the optimal physiological conditions for HT 3D-drug screening. Furthermore, to fully validate the usage of the 3D-DSRT platforms for the functional precision medicine, the ex vivo/in vitro drug responses of PDCs need to be compared to the clinical drug responses of the patient, as shown with OvCa1 patient case[