This protocol describes a semi-automated method for medium- to high-throughput organoid drug screenings and microscope-agnostic, automated image analysis software to quantify and visualize multiparametric, single-organoid drug responses to capture intratumor heterogeneity.
Patient-derived tumor organoids (PDTOs) hold great promise for preclinical and translational research and predicting the patient therapy response from ex vivo drug screenings. However, current adenosine triphosphate (ATP)-based drug screening assays do not capture the complexity of a drug response (cytostatic or cytotoxic) and intratumor heterogeneity that has been shown to be retained in PDTOs due to a bulk readout. Live-cell imaging is a powerful tool to overcome this issue and visualize drug responses more in-depth. However, image analysis software is often not adapted to the three-dimensionality of PDTOs, requires fluorescent viability dyes, or is not compatible with a 384-well microplate format. This paper describes a semi-automated methodology to seed, treat, and image PDTOs in a high-throughput, 384-well format using conventional, widefield, live-cell imaging systems. In addition, we developed viability marker-free image analysis software to quantify growth rate-based drug response metrics that improve reproducibility and correct growth rate variations between different PDTO lines. Using the normalized drug response metric, which scores drug response based on the growth rate normalized to a positive and negative control condition, and a fluorescent cell death dye, cytotoxic and cytostatic drug responses can be easily distinguished, profoundly improving the classification of responders and non-responders. In addition, drug-response heterogeneity can by quantified from single-organoid drug response analysis to identify potential, resistant clones. Ultimately, this method aims to improve the prediction of clinical therapy response by capturing a multiparametric drug response signature, which includes kinetic growth arrest and cell death quantification.
In recent years, in vitro cancer drug discovery, drug screening, and fundamental research have been transitioning from the use of traditional two-dimensional (2D) cancer models with immortalized cell lines to more physiologically relevant three-dimensional (3D) cancer models. This has spurred the adoption of tumor spheroids with established cancer cell lines, which recreate more complex cell-to-cell interactions and structures present in solid tumors. Currently, patient-derived tumor organoids (PDTOs) are the most advanced and physiologically relevant 3D cancer model available for in vitro cancer research, as they provide additional advantages over tumor spheroids, namely the heterogeneity found in cancer patients1. PDTOs are established from tumor tissue originating from cancer patients, and therefore retain both the tumor phenotype and genotype. As such, PDTOs are becoming invaluable for fundamental and translational cancer research and have the potential to greatly improve precision oncology2.
Despite their promising potential, these sophisticated 3D in vitro cancer models are often underutilized due to a lack of advanced analysis methods. The most commonly used assay determines the number of viable cells in the PDTO via the quantification of intracellular ATP3. These assays are normally single-timepoint, bulk analyses, thus overlooking critical time-dependent responses and neglecting clonal responses. Specifically, the ability to monitor the growth of PDTOs (growth rate) and their response to specific therapies is of high interest4,5. The normalized drug response (NDR), which scores drug response based on the growth rate normalized to a positive (ctrl+) and negative control (ctrl-) condition, has also recently been reported to be a crucial metric for evaluating cancer drug sensitivity with cell-based screening, though this was predominantly done for 2D cell lines6. Therefore, more sophisticated analysis methods are needed to fully take advantage of these more clinically representative and complex 3D cancer models. Microscopy is considered a powerful approach to study the complexity of these organoid models7.
This paper describes a method for monitoring kinetic drug responses in 3D cancer models, using conventional widefield microscopes and live-cell imaging systems. Adaptations were made to the protocol described by Driehuis et al.4 to be compatible with automation using a pipetting robot, digital drug dispenser, and live-cell imaging system to increase reproducibility and reduce the number of ‘hands-on’ hours of labor. This method allows for medium- to high-throughput drug screening of both tumor spheroids with established cancer cell lines (see Supplementary Table S1 for tested cell lines), as well as the PDTOs, in a 384-well microplate and multi-organoid format. By using a convolutional network machine learning process, automated identification and tracking of individual tumor spheroids or PDTOs could be performed solely from brightfield imaging and without the use of fluorescent live-cell labeling dyes8. This is highly advantageous, as most identification with brightfield imaging requires manual annotation (which is laborious and time-consuming) or requires the addition of fluorescent dyes, which can confound drug responses related to photoxicity-induced oxidative stress9.
The resulting image analysis software developed in-house extends the functionality of conventional live-cell imaging systems, as 3D image analysis modules are either not available, platform-restricted, or not compatible with 384-well microplates and whole-well imaging. In addition, these modules are often highly priced and offer limited bulk organoid readouts. Therefore, this method is highly relevant for scientists who have access to widely available live-cell imaging systems and aim to extract more information about a drug response compared to the gold-standard but rudimentary ATP-based assay. With the addition of specific cell death indicators, cytostatic drug responses can be distinguished from cytotoxic responses, thus providing further insight into mechanistic drug actions currently unattainable from single-timepoint analysis. Finally, live-cell imaging allows for individual organoid tracking to obtain single organoid drug response metrics to capture response heterogeneity and identify potential resistant subclones.
The goal of this method and the associated image analysis software is to implement low-cost automation in organoid drug screening to limit user intervention and reduce variability in handling, image analysis, and data analysis. To make this software available to researchers, it is microscope- and platform-agnostic, and a cloud-based application is made available. Thus, by supporting conventional live-cell imaging systems, we also aim to improve their functionality for 3D culturing applications and analysis.
Human pancreatic ductal adenocarcinoma (PDAC) patient-derived organoids were used. Tissue resection fragments were obtained from patients undergoing curative surgery at the Antwerp University Hospital. Written informed consent was obtained from all patients, and the study was approved by the UZA Ethical Committee (ref. 14/47/480). Details related to all materials, reagents, equipment, and software used in this protocol are provided in the Table of Materials. An overview of the workflow is presented in Figure 1. Example data is provided in the supplementary material to reproduce the protocol.
1. Day 0: Preparation of 2- or 3-day old organoids
2. Days 2 – 3: Harvest and seed 2- or 3-day old organoids
3. Day 4: Drug treatment and reagent dispensing with digital drug dispenser
4. Acquire images with the live-cell imager
NOTE: For the growth rate and NDR, a scan at timepoint 0 (T0 = start treatment) must be acquired 1-2 h after adding Cytotox Green.
5. Image and data analysis
The automated pipetting protocol ensures an even distribution of PDAC_060 PDTOs in all the columns of the 384-well microplate (Figure 2A). As expected, a variation in the number and mean area of PDTOs was observed between the wells (Figure 2A,B). The total survival area (total brightfield area – total green area) combines the label-free organoid segmentation with the fluorescence-based cell death signal and is, in our experience, the most robust parameter to study drug responses over time (Figure 2C)8. To account for variations in cell seeding and organoid size, growth rate-based metrics should be used to reduce variations between replicates, as shown by the reduced error bars in Figure 2D versus Figure 2C, and a higher Z-factor indicating a strongly improved drug screen quality (Figure 2E).
The NDR dose-response curve (Figure 2G), normalized to ctrl- and ctrl+, is clearly superior to the GR dose-response curve (Figure 2F), normalized to ctrl-, as it increases the separation of the drug response curves and more accurately represents cytotoxic drug responses. An example of the associated images for ctrl-, ctrl+, and 400 nM gemcitabine/80 nM paclitaxel-treated PDTO is shown in Figure 3. An interesting observation is that the cytotoxic effect of gemcitabine was dominant in the combination therapy as no added value of paclitaxel was observed.
Next, two additional PDTO lines, PDAC_052 and PDAC_087, were used. A clear difference in growth rate between these lines was observed (Figure 4A), which supports the use of GR metrics. Again, NDR dose-response curves (Figure 4C) resulted in an increased dynamic range and separation between the three different patients compared to the GR curves (Figure 4B). Furthermore, the protocol allows for the determination of NDR over time and shows that PDAC_052 and PDAC_060 had a very similar cytostatic drug response to a low dose of gem-pac (Figure 4D), while a clear differential cytostatic versus cytotoxic response could be observed for the middle (Figure 4E) and high doses (Figure 4F) of gem-pac. These drug responses were consistent with the clinical responses observed in the patients (Figure 4G).
Finally, a major benefit of the approach and software is that single-organoid drug responses can be quantified to study response heterogeneity and identify potentially resistant subclones. Figure 5 provides a clear overview of the clonal dynamics of the different patients and shows that PDAC-087 had the most resistant subclones following treatment, which is consistent with the aggressive and highly resistant disease observed in the patient. Interestingly, this patient was also the least sensitive to the ctrl+ staurosporin.
Figure 1: Workflow overview. Please click here to view a larger version of this figure.
Figure 2: Seeding accuracy and drug response metrics. (A) Organoid counts/well of PDAC_060 PDTOs seeded in a 384-well microplate using the pipetting robot. Each dot represents the count in a single well and plots are separated by the 384-well microplate columns. (B) Mean PDTO area/well. (C) Total survival area (total brightfield area – total green area) and (D) growth rate (total survival area normalized to T0 = 1) of PDAC_060 PDTOs treated with a 5:1 ratio of gemcitabine/paclitaxel. (E) Z-factor as a metric for assay quality. (F) Growth rate-dose response curve normalized to ctrl- and (G) normalized drug response curve normalized to ctrl- and ctrl+. Error bars indicate mean ± SD of two wells. Abbreviations: PDAC = pancreatic ductal adenocarcinoma; PDTO = patient-derived tumor organoid; GR = growth rate; NDR = normalized drug response. Please click here to view a larger version of this figure.
Figure 3: Example images. Representative images of PDAC_060 PDTO treated with vehicle (ctrl-), 400 nM gemcitabine/80 nM paclitaxel, and 2 µM staurosporin (ctrl+). The left column shows brightfield images, the middle column shows the Cytotox Green fluorescent signal, and the right column shows the label-free annotated brightfield images using the organoid analysis module. Scale bars = 100 µm. Abbreviations: PDAC = pancreatic ductal adenocarcinoma; PDTO = patient-derived tumor organoid; GemPac = gemcitabine/paclitaxel. Please click here to view a larger version of this figure.
Figure 4: Comparing interpatient drug response. (A) Comparison of growth rate (based on total survival area) of PDAC_052, PDAC_060, and PDAC_087 PDTO lines. (B) Growth rate-dose response curve normalized to ctrl- and (C) normalized drug response curve normalized to ctrl- and ctrl+. Kinetic NDR of a (D) low, (E) middle, and (F) high dose of gemcitabine/paclitaxel (5:1 ratio). (G) PDAC patients' clinical characteristics. Error bars indicate mean ± SD of two wells. Abbreviations: PDAC = pancreatic ductal adenocarcinoma; PDTO = patient-derived tumor organoid; GR = growth rate; NDR = normalized drug response; FFX = folfirinox. Please click here to view a larger version of this figure.
Figure 5: Single organoid metrics. Single organoid dose response based on cell death (green area/brightfield area) and area (brightfield) of PDAC_052, PDAC_060, and PDAC_087 PDTOs treated with vehicle (ctrl-), 400 nM gemcitabine/80 nM paclitaxel, and 2 µM staurosporine (ctrl+). Green regions indicate viable organoids; blue regions indicate x-as range of GemPac and ctrl+ plots. Abbreviations: PDAC = pancreatic ductal adenocarcinoma; PDTO = patient-derived tumor organoid; GemPac = gemcitabine/paclitaxel. Please click here to view a larger version of this figure.
Cell suspension stock | Cells/Drop | # Drops (20 µL) | Stock (1/3) | ECM (2/3) |
1.13 × 107 cells/mL | 75,000 | 10 | 75 uL | 150 µL |
1.13 × 107 cells/mL | 75,000 | 5 | 40 uL | 80 µL |
Table 1: Dilution for plating in ECM domes. Abbreviation: ECM = extracellular matrix.
Compound | Stock concentration | Dilution | Working concentration | Solvent | Well concentration | Comments |
Cytotox Green | 1 mM (DMSO) | 1/10 | 10 µM | DMSO | 60 nM | Cell death marker |
Cytotox Red | 1 mM (DMSO) | 1/10 | 10 µM | DMSO | 250 nM | Cell death marker |
Caspase 3/7 Green | 5 mM (DMSO) | 1/2 | 2.5 mM | DMSO | 2.5 µM | Apoptotic marker |
Hoechst | 20 mM (H2O) | 1/200 | 100 µM | 0.33% Tween/PBS | 50 nM | Nuclear marker |
Staurosporin | 10 mM (DMSO) | / | 1 – 10 mM | / | 2 – 5 µM | Positive control |
Gemcitabine | 10 mM (DMSO) | / | 1 – 10 mM | / | Titration | Chemotherapy |
Paclitaxel | 10 mM (DMSO) | / | 1 – 10 mM | / | Titration | Chemotherapy |
Cisplatin | 5 mM (0.9% NaCl) | 1/2 | 2.5 mM | 0.6% Tween/PBS | Titration | Chemotherapy |
Table 2: Example dilutions of frequently used drugs and fluorescent reagents. Each compound needs to be dissolved in either 100% DMSO or 0.3% Tween/PBS.
Supplementary Table S1: Overview of compatible cancer cell lines. Static: spheroids are not migratory. Merge: spheroids migrate toward each other and merge together. Please click here to download this File.
Supplementary File 1: Organoid seeding solution calculation tool. Please click here to download this File.
Supplementary File 2: STL file for 3D printing custom labware 'Microplate Holder'. Please click here to download this File.
Supplementary File 3: STL file for 3D printing custom labware '2 x 25 mL Reservoir Holder'. Please click here to download this File.
Supplementary File 4: JSON file for custom labware pipetting robot 'Microplate Holder'. Please click here to download this File.
Supplementary File 5: JSON file for custom labware pipetting robot '2 x 25 mL Reservoir Holder_WithCooler'. Please click here to download this File.
Supplementary File 6: JSON file for pipetting robot protocol 'Plating_ PDO_384well_Cooled_Row2-23'. Please click here to download this File.
Supplementary File 7: Overview of the pipetting robot desk setup. (A) Cooling elements and (B) reservoir and microplate. Please click here to download this File.
Supplementary File 8: TDD file for protocol of the digital drug dispenser. Please click here to download this File.
Supplementary File 9: XML file for protocol of the live-cell imager for brightfield and fluorescence imaging. Please click here to download this File.
Supplementary File 10: Example plate map. Please click here to download this File.
Supplementary File 11: Example input file for NDR R script. Abbreviation: NDR = normalized drug response. Please click here to download this File.
Supplementary File 12: Normalized drug response NDR R script. Abbreviation: NDR = normalized drug response. Please click here to download this File.
Supplementary File 13: Example output file of NDR R script GR values. Abbreviations: GR = growth rate; NDR = normalized drug response. Please click here to download this File.
Supplementary File 14: Example output file of NDR R script with GR values transposed. Abbreviations: GR = growth rate; NDR = normalized drug response. Please click here to download this File.
Supplementary File 15: Example output file of NDR R script NDR values. Abbreviation: NDR = normalized drug response. Please click here to download this File.
Supplementary File 16: Example output file of NDR R script with NDR values transposed. Abbreviation: NDR = normalized drug response. Please click here to download this File.
Medium- to high-throughput PDTO drug screening often relies on readouts that only extract a fraction of information that organoids could potentially provide. It has become increasingly clear that, in order for the rapidly evolving organoid technology to realize greater scientific and clinical potential, more advanced 3D assays, readouts, and analysis methods are critically required. Here, an advanced screening pipeline is described, which not only increases the reproducibility, but also considerably enhances the clinical translatability by incorporating an AI-driven, live-cell imaging readout. In addition to analysis software developed in-house, the use of the normalized drug response metric (NDR) is implemented, which clearly demonstrates its ability to define patient-specific differences in treatment response6.
The inclusion of this normalization metric will undoubtedly be of tremendous value, recalling that numerous studies aim to delineate treatment responses based on minor differences in area under the curve (AUC) or half-maximal inhibitory concentration (IC50) (as most of the dose-response curves overlap/are located close to each other)11,12. Growth rate metrics have already been implemented in organoid drug screening protocols using the ATP-based assay but rely on the normalization of reference wells lysed at timepoint 04. In contrast, this method allows for intrawell growth-rate normalization, which not only accounts for interpatient differences in PDTO growth rate but also interwell differences resulting from variations in seeding density and plate location-dependent effects to increase reproducibility. Furthermore, we adapted the NDR to further increase the separation of interpatient PDTO response by including a positive control for normalization6,8.
Furthermore, the analysis, which is compatible with high-throughput and automation formats, can accurately detect individual organoid responses, enabling the quantification of subclonal resistance-the major driving force of tumor relapse and progression13. For example, although PDAC052 and PDAC060 showed a good response to the treatment in vitro (based on the NDR), the additional single-organoid analysis was able to detect a small (bigger population with PDAC060) population of subclones that do not respond to the treatment. Interestingly, this corresponded highly with the clinical observation, given that PDAC052 and PDAC060 had a durable response (no tumor activity detected) but eventually were both diagnosed with local tumor progression (due to the presence of resistant clones). Compared to the conventional 3D readouts (ATP-based assay and size/numbers), this advanced screening pipeline is expected to increase the predictive performance by extracting more clinically relevant information out of these ‘patients-in-the-lab’. This hypothesis is now being tested by screening clinical PDTO samples in the authors’ laboratory with this method to correlate ex vivo with in vivo response and clinical outcome.
To obtain more insights into the mechanisms of a drug response, conventional fluorescent live-cell imaging reagents, in addition to cytotoxicity dyes, are compatible with this method to study mechanisms of cell death. We have previously shown the compatibility of this method with the Sartorius Caspase 3/7 Green Reagent to study caspase-dependent induction of apoptosis following cisplatin treatment8. The compatibility with other dyes to study oxidative stress (CellROX reagents) or hypoxia (Image-iT Hypoxia reagents) remains to be tested. However, these reagents have already successfully been used in 3D in vitro models14,15.
The image analysis software is also compatible with other plate formats or culturing methods (e.g., microcavity plates, ECM domes) if clear, in-focus images of the organoids can be captured. This is often challenging for organoids cultured in domes since they grow in different z-planes, which requires z-stacking functionality of the microscope that is not always available. Therefore, we advise the use of flat-bottom ULA 384-well microplates to ensure images of sufficient quality.
In addition, the analysis is compatible with other live-cell imaging systems, as previously shown for phase-contrast images captured with an IncuCyte ZOOM system8. A limitation of the Spark Cyto live-cell imaging system that was used in this manuscript is the one-plate capacity for kinetic measurements. However, the Spark Motion expansion increases its capacity to up to 40 microplates that can be screened in bulk. The compatibility of the software developed in-house will be expanded to these and other systems to offer a platform-agnostic solution, with the goal to standardize and automate image and data analysis pipelines. The web-based application will also include interactive graphing tools and automated drug metric calculations, as shown in this paper, to reduce manual analysis time.
The label-free PDTO segmentation algorithm was trained and tested on various in-house grown spheroid and PDTO models with distinct morphological differences (solid, semi-solid, cystic), and can consequently detect these with high accuracy8. A limitation of the model is that the inclusion of cystic PDTOs increased the unwanted detection of bubbles present in the well following seeding. However, overnight incubation was sufficient to remove most of these bubbles, allowing for a qualitative timepoint 0 scan. The accuracy of the organoid image segmentation and the method needs to be validated by other users, and based on their feedback, the software can be trained further to obtain a robust and automated image analysis algorithm. In addition, we aim to obtain more clinical data to correlate the ex vivo drug response quantified by this method to the clinical response in the patient to identify the best parameters to predict therapy response and further develop this method for functional precision cancer medicine16.
The authors have nothing to disclose.
Part of this research was funded by donations from different donors, including Dedert Schilde vzw and Mr Willy Floren. This work is partially funded by the Flemish Research Foundation, 12S9221N (A.L.), G044420N (S.V., A.L., E.G), 1S27021N (M.L), and by the Industrial Research Fund of the University of Antwerp, PS ID 45151 (S.V., A.L., C.D.). Figure 1 was created with BioRender.com.
6-well plate | Greiner | 657160 | |
8-Channel p300 (GEN 2) pipette | Opentrons | ||
300 µL Tips | Opentrons | ||
384-well flat-bottom ULA microplate | Corning | 4588 | minimum volume 50 µL |
384-well flat-bottom ULA Phenoplate | Perkin Elmer | 6057802 | minimum volume 75 µL |
A8301 | Tocris Bioscience | 2939 | |
ADF+++ | Advanced DMEM/F12, 1% GlutaMAX, 1% HEPES, 1% penicillin/streptomycin | ||
Advanced DMEM/F-12 | ThermoFisher Scientific | 12634 | |
B27 | ThermoFisher Scientific | 17504044 | |
Breathe easy sealing membrane | Sigma-Aldrich | Z380059 | |
Caspase 3/7 Green | Sartorius | 4440 | |
Cell Counting Slides for TC10/TC20 | Bio-Rad Laboratories | 1450017 | |
CellTiter-Glo 3D | Promega | G9681 | ATP-assay |
Cooler for 25 mL reservoir | VWR (Diversified Biotech) | 490006-908 | |
Cooling element 12 x 8 x 3 cm | Bol.com | 9200000107744702 | For custom microplate holder OT-2 |
Cultrex Organoid Harvesting Solution | R&D systems | 3700-100-01 | |
Cultrex PathClear Reduced Growth Factor BME, Type 2 | R&D systems | 3533-010-02 | extracellular matrix (ECM) |
Cytotox Green | Sartorius | 4633 | |
Cytotox Red | Sartorius | 4632 | |
D300e | Tecan | Digital drug dispenser | |
D300e Control v3.3.5 | Tecan | Control software D300e | |
FGF10 | Peprotech | 100-26 | |
Full Medium | ADF+++ supplemented with 0.5 nM WNT surrogate-Fc-Fusion protein, 4% Noggin-Fc Fusion Protein conditioned medium, 4% Rpso3-Fc Fusion Protein conditioned medium, 1x B27, 1 mM N-acetyl cysteine (NAC), 5 mM nicotinamide, 500 nM A83-01, 100 ng/mL FGF10, and 10 nM Gastrin | ||
Gastrin | Sigma-Aldrich | G9145 | |
Gemcitabine | Selleck Chemicals | S1714 | |
GlutaMAX | ThermoFisher Scientific | 35050 | |
HEPES | ThermoFisher Scientific | 15630056 | |
Hoechst 33342 Solution (20 mM) | ThermoFisher Scientific | 62259 | |
Human pancreatic ductal adenocarcinoma (PDAC) patient-derived organoids | Biobank@uza (Antwerp, Belgium; ID: BE71030031000; Belgian Virtual Tumorbank funded by the National Cancer Plan) | ||
N-acetyl-cysteine | Sigma-Aldrich | A9165-25G | |
Nicotinamide | Sigma-Aldrich | N0636-100G | |
Noggin-Fc Fusion Protein conditioned medium | Immunoprecise | N002 | |
Opentrons App v6.0.1 | Opentrons | OT-2 control software | |
Opentrons Protocol Designer Tool | Opentrons | https://designer.opentrons.com/ | |
Orbits data compression tool | www.orbits-oncology.com or contact corresponding author | ||
Orbits image analysis webapp | University of Antwerp | www.orbits-oncology.com or contact corresponding author | |
OT-2 | Opentrons | Pipetting robot | |
Paclitaxel | Selleck Chemicals | S1150 | |
Pasteur Pipette 230 mm | Novolab | A33696 | |
Peniciline-Streptomycin | ThermoFisher Scientific | 15140 | |
Prism 9 | GraphPad | ||
Rspo3-Fc Fusion Protein conditioned medium | Immunoprecise | N003 | |
Spark Cyto 600 | Tecan | Live-cell imaging and multi-mode platereader | |
SparkControl v3.1 | Tecan | Spark Cyto control software | |
Staurosporine | Tocris Bioscience | 1285 | |
Sterile 25 mL reservoir | VWR (Diversified Biotech) | 10141-922 | |
T8 plus cassette | Tecan | ||
TC20 | Bio-Rad Laboratories | automated cell counter | |
TrypLE | ThermoFisher Scientific | 12604-021 | dissociation enzyme |
Tween-20 | Acros Organics | 233360010 | |
WNT Surrogate-Fc-Fusion protein | Immunoprecise | N001 | |
Y-27632 | Selleck Chemicals | S1049 |