The YCH is an automated system designed for selecting and isolating various types of cells ranging from single cells to PDOs, individually from a source to a destination plate (Fig. 1). The efficacy and precision of transferring PDOs to 384-well screening plates using the YCH system were optimized.
Fig. 1: Overview of YCH mode of action.
Step 1 “Detection”: morphological and phenotypic features of cellular structures in the source plate are obtained by high-throughput image analysis. Step 2 “Selection”: PDO features (20 distinct parameters) are instantly visualized in a histogram. Combining multiple features and threshold limits as selection criteria allows a target group of PDOs to be automatically selected. Step 3 “Transfer”: selected single PDOs are picked and dispensed damage-freely using gentle pipetting action. Step 4 “Verification”: successful PDO isolation is confirmed by imaging the destination plate. Verifying process completion is possible by imaging the source plate before and after PDO transfer.
Assay seeding optimization
The first step of the YCH flow is PDO detection and selection, which is based on brightfield imaging of 6-well microgrid plates (source plates). For precise transfer, it is critical that PDOs do not share their microgrid square with other PDOs. We optimized the seeding density in the source plates to ensure sufficient PDOs for transfer that meet the selection criteria. Optimization experiments were performed using metastatic CRC PDO 1 and CRC PDO 2. We assessed three seeding densities: 1000, 2000, and 4000 PDOs per well in a 6-well source plate. The YCH consistently detected a number of PDOs per well that closely matched the original seeded quantity, with all seeding densities matching 80–90% of the detected PDOs (Fig. 2A, B) and meeting the criteria for identifying single PDOs (circularity 0.4–1 and area 300–8000 μm2, Fig. 2C). While an average of 10 grids out of more than 15,000 grid squares containing multiple PDOs were detected for all seeding densities (Fig. 2D), visual microscopic inspection showed that grids containing multiple PDOs were prevalent after seeding 4000 PDOs (Fig. 2E). Consequently, a seeding density of 2000 PDOs per grid square was selected for subsequent experiments.
Fig. 2: Optimization of PDOs seeding densities into 6-well microgrid (source plate) and detection using YCH.
A PDOs detected by YCH after seeding different numbers of PDOs into the source plate. B Similar to A, but plotted as a percentage of intended to be seeding. C Percentage of detected PDOs within selection criteria (300–8000 μm2 PDO size and circularity 0.4–1.0). D Similar to C, with the additional criteria of having 1 PDO/grid. Representative brightfield image of source plate’s wells captured after seeding 1000 (E), 2000 (F) and 4000 (G) PDOs per well.
The second step is transferring the PDOs from the source to the destination plate. Therefore, we determined the transfer efficiency of the YCH, depending on aspiration and dispense efficiency. After finalizing the transfer of target PDOs selected based on distinct optimized parameters including, PDO area, shape and roundness, the YCH software (Analyzer) verified the accuracy of aspiration. This was done by capturing brightfield images of both the source and destination plate and comparing images obtained from the source plate before and after transferring and images of the destination plate after dispensing. On average, in 80% of the cases, the PDOs were correctly aspirated from the source plate and dispensed into the destination plate. Different PDO numbers per well of the 384-well destination plate were tested to explore the downscaling of drug screening experiments, including 1, 2, 5, and 10 PDOs per well. PDOs were monitored 5 days after transfer to assess the effects, if any, of handling and transfer using the YCH. sing 5 or 10 PDOs per well resulted in transfer efficiency of 80-100% with reduced variability when compared to 1 or 2 PDOs per well, regardless of ECM percentages (Fig. 3).
Fig. 3: Transfer efficiency from source plate to destination plate using YCH.
A PDO number transferred presented as raw counts (B) % of planned, 5 days after transfer, of 1, 2, 5 or 10 PDOs per well.
The YCH was initially designed to transfer a single PDO per tip from a source to a destination plate. Since this is time-consuming and limits the number of PDOs that can be handled on a single day, further development and testing allowed the transfer of 3 PDOs within the same tip without affecting transfer accuracy and/or assay quality. The transfer time was reduced to 1 second per PDO when transferring 3 PDOs instead of 1 PDO within the same tip (Supplementary Fig. 1).
Initially, we observed PDO clumping upon transfer to the destination 384-well plates when using higher extracellular matrix (ECM) percentages. To address this issue, we optimized the percentage of ECM incorporated into the plates while ensuring optimal PDO growth. ECM plays an essential role in supporting epithelial cell polarization and proliferation. This fosters the development of accurate PDO structures with enhanced functionality15. To determine the ECM percentage that would promote the best PDO growth and viability while minimizing clumping in the screening plates, we investigated varying concentrations (0%, 1%, 2%, and 5%, Fig. 3 and Supplementary Fig. 2). The least PDO clumping was visually observed in wells containing 2% or 5% ECM. Therefore, we selected 5% ECM as the optimal concentration for further testing. A higher ECM concentration supports the accurate development of organoid structures and enhances their functionality16.
Assay readout optimization
In current in vitro cell-based drug screening assays, the ATP-based CellTiter-Glo readout is the gold standard for evaluating cellular growth and viability following drug treatment17,18,19. While this luminescence-based readout is fast and scalable, it is susceptible to non-viability-related ATP fluctuations. Another emerging readout is CyQUANT™ Cell Proliferation Assay, which only stains the DNA of living cells by using a fluorescent DNA-binding dye and a quencher. The signal can be quantified by fluorescence measurements using a plate reader, but staining is also suitable for image-based analysis. Therefore, a CyQUANT image-based readout was explored to determine PDO viability (Supplementary Fig. 3A). The image-based readout was first optimized using 250 PDOs per well in 384-well plates seeded with the Multidrop Combi. It was observed that when using the Multidrop Combi for CyQUANT addition, PDO detached from the bottom of the plate (Supplementary Fig. 4), leading to difficulties in segmenting the PDOs in the image analysis. Thus, it was investigated if the detachment could be reduced by gently adding lower volumes with increased dye concentrations (e.g. 40 μL 2x, 20 μL 3x or 10 μL 5x concentrated), using a Tecan Fluent liquid handling platform. We also assessed if this would require prolonged incubation time to obtain the same signal. Similar signals were obtained for all conditions tested, independent of the type of quantification (image-based fluorescence intensity measures, image-based area measures, or reader-based fluorescence measures, Supplementary Fig. 3B). Therefore, the addition of 20 μL 3x concentrated dye and 1 h incubation was selected as the optimal condition, as it minimizes the added liquid volume, reducing sample detachment while still achieving an effective staining signal with a shortened incubation period. PDO detachment was effectively prevented, and improved assay quality was observed for all CyQUANT measures compared to CellTiter-Glo (Supplementary Fig. 3C). While all tested readouts successfully captured dose-response curves for staurosporine titration (Supplementary Fig. 3D), the lower assay quality observed with the CellTiter-Glo readout emphasized the need for further optimization of the CyQUANT viability readout when using a reduced number of PDOs to ensure robust assay performance.
Subsequently, results were evaluated using low PDO numbers per well transferred by the YCH, specifically 1, 2, 5 and 10 PDOs per well. A small assay window was observed with the CellTiter-Glo readout (Fig. 4). Image-based CyQUANT area and fluorescence intensity measurements both showed better assay windows, with clear differentiation between DMSO (negative control) and Staurosporine (positive control), which was not observed with the plate reader-based CyQUANT assays (Fig. 4). Additionally, as no assay window was observed when seeding 1–2 PDOs per well, seeding densities of 5–10 PDOs per well were selected for further optimization.
Fig. 4: Assay window for low PDO numbers.
Assay window (positive and negative control wells) for different readout measurements using low PDO numbers of one PDO model. Data points represent technical replicates (n = 10 DMSO (0.1%) and n = 4 Staurosporine). The black line represents the median, numbers indicate Z’-factor and %CV respectively.
To determine the effect of PDO size on assay reproducibility, PDOs from one model were size selected to transfer 5 or 10 PDOs with different diameter ranges (20–100 μm, 40–100 μm and 40–70 μm) to 384-well plates upon which assay quality was compared using image-based CyQUANT intensity and area measurements. Interestingly, the exclusion of PDOs between 20–40 μm (selection of PDOs between 40–100 μm) increased assay quality when seeding 5 PDOs per well. Similar assay quality was obtained after seeding 5 PDOs per well of 40–100 μm when compared to 10 PDOs of 20–100 μm (Fig. 5A, B). No clear differences were observed between 40–70 μm and 40–100 μm size ranges, which is likely caused by the limited fraction of 70–100 μm sized PDOs in the population. Overall, the highest assay quality (Z’ = 0.41–0.49, %CV = 20–24%) was obtained for CyQUANT intensity and area measurements, respectively after transferring 10 (20–100 μm) PDOs per well.
Fig. 5: Assay quality for different PDO sizes.
A Assay window for different CyQUANT measurements comparing 20–100, 40–100 and 40–70 μm organoid diameter data points represent technical replicates (n = 8 DMSO and n = 4 Staurosporine), numbers indicate Z’-factor and %CV respectively. B Distribution of different organoid sizes within a PDO culture. Results shown are derived from one PDO model.
To validate these findings, we repeatedly compared all parameters with sufficient assay quality. We compared (1) different readouts (CellTiter-Glo vs. CyQUANT), (2) PDO seeding densities (5 vs. 10 PDOs per well), and (3) the seeding method (Multidrop Combi vs. YCH). Similar to previous experiments, the highest assay quality was obtained after seeding 10 PDOs per well with CyQUANT fluorescence intensity (Supplementary Fig. 5A). Increased assay quality was observed for low PDO numbers when seeded with the YCH compared to the Multidrop Combi (Supplementary Fig. 5B). Most importantly, the amount of input material required for YCH reduced substantially to 0.5–5.3% of the necessary input for conventional screening (Table 1).
Table 1 Overview of the number of PDOs needed as input to seed a full 384-well plate
Assay verification
The next step was to confirm the feasibility of using 10 PDOs per well for drug screening purposes. Further tests compared drug screening results from 250 PDOs per well against 10 PDOs per well of a 384-well screening plate.
Determine diversity in response within a PDO model
Given that tumor complexity and diversity is accurately represented in PDOs, our aim was to confirm that reducing the number of PDOs screened does not alter drug sensitivity. Therefore, we examined if PDO response to the commonly used chemotherapeutic agent 5-FU was consistent, by seeding 10 PDOs per well using the YCH, and analyzing single PDOs. One PDO model exhibited a single distribution, suggesting a homogeneous population, regardless of the concentration of 5-FU used (Fig. 6A). In contrast, treatment with the IC50 of 5-FU in another PDO model resulted in two populations, indicating a heterogeneous population (Fig. 6B). Notably, this effect was induced by 5-FU treatment, as evidenced by the single distribution observed at day 0 (Fig. 6B, left graph). Taken together, these findings confirm that YCH is a valuable tool for studying PDO diversity in drug responses by transferring single PDOs and analyzing each individually across wells of a screening plate.
Fig. 6: PDO diversity in response to 5-FU treatment.
CyQUANT area and intensity measurements of CRC PDO 1 (A) and CRC PDO 2 (B) PDOs in response to treatment with IC10, IC30 and IC50 of 5-FU when seeding 10 PDOs per well. Data points represent individual PDOs seeded in 3–6 technical replicates, depending on the condition. Concentrations shown in graph: IC10 is 2.60 µM, IC30 is 33.2 µM and IC50 is 164.2 µM.
PDO sensitivity to EGFR inhibition with minimal seeding density
The relation between CRC sensitivity to EGFR inhibition with panitumumab and the RAS/BRAF mutation status is well known20,21. Exploratory, we investigated if these sensitivity profiles can also be captured using only 10 PDOs per well with the CyQUANT readout in viability assays. Two PDO models were selected for this experiment: one BRAF and KRAS wildtype (WT) PDO and another with a BRAF V600E mutation. The RAS/BRAF wildtype PDO was sensitive, whereas the BRAF V600E mutated PDO was resistant to panitumumab, independently of the PDO seeding density (10 vs. 250 PDOs per well) or readout (CyQUANT vs. CellTiter-Glo, Fig. 7).
Fig. 7: PDO response curves to panitumumab by BRAF status.
Response of BRAF mutant metastatic CRC PDO (PDO 1) and KRAS/BRAF WT CRC PDO (PDO 3) to EGFR inhibitor panitumumab seeding 10 PDOs per well with CyQUANT readout or seeding 250 PDOs per well and CellTiter-Glo readout. Data represents the Mean ± SD of technical replicates (n = 4).
Assay application
Proof-of-concept (POC) clinical validation study on mCRC cancer PDOs
As a proof of concept (POC), eight PDOs from mCRC patients who had received 5-FU and oxaliplatin combination therapy in the clinic were selected for drug screens with 5-FU and oxaliplatin monotherapy after seeding 10 PDOs per well and using CyQUANT as readout. Similar drug screens using a 10-point concentration range were performed after seeding 250 PDOs per well. The drug response was quantified using the area under the drug response curves (AUC). Interestingly, PDO models, on average, demonstrated similar or increased growth when seeded with low numbers (10 PDOs per well) when compared to high numbers (250 PDOs per well, Supplementary Fig. 6 and Supplementary Fig. 7). Average area fold change for 10 PDOs per well = 5.19, average area fold change for 250 PDOs per well = 3.06 (Supplementary Fig. 6B). Despite this disparity, the response to 5-FU and oxaliplatin was comparable using 10 or 250 PDOs per well (R = 0.85, p = 0.0079 and R = 0.75, p = 0.034, Fig. 8A). The mean difference in AUC between screens using 10 and 250 PDOs per well was 0.03, with limits of agreement ranging from −0.34 to 0.41 (Supplementary Fig. 8A) Additionally, to explore the potential for further miniaturizing the assay from 10 PDOs per well to 5 PDOs per well with fewer concentration points to further reduce TAT, 5-FU was tested using a 5-point concentration range. This was done after seeding 5 PDOs per well for eighteen mCRC PDOs derived from metastases, five of which were also screened with 10 PDOs per well. The AUCs of 5 and 250 and 5 and 10 PDOs per well were significantly correlated (R = 0.67, p = 0.00079 and R = 0.99, p = 0.0012, Fig. 8B, C), demonstrating robust PDO response measurements with reduced input material. The mean difference in AUC between screens using 5 and 250 PDOs per well was −0.22, with limits of agreement ranging from −0.65 to 0.2 (Supplementary Fig. 8B).
Fig. 8: Correlation of PDO responses to 5-FU (and oxaliplatin).
A Correlation of PDO responses to 5-FU and oxaliplatin using 10 PDOs per well and 250 PDOs per well. PDOs are derived from metastases. B Correlation of PDO responses to 5-FU using 5 PDOs per well and 250 PDOs per well. All 18 PDOs are derived from metastases. C Correlation of PDO responses to 5-FU using 5 PDOs per well and 10 PDOs per well. All 5 PDOs are derived from metastases. Measurements shown are the mean of 2 and 4 biological replicates for 5 PDO per well and 10 PDOs per well, respectively. AUCs for 250 PDOs per well are based on CyQUANT intensity growth rate inhibition (GR) curves. AUCs for 5 and 10 PDOs per well are based on mean CyQUANT intensity drug response curves.
Investigating the PDOs’ responsiveness to chemotherapeutic agents holds promise for personalized medicine in cancer treatment. Therefore, we compared drug response measurements using 10 PDOs per well from eight patients with clinical outcomes. We observed a significant correlation between mCRC PDO response to 5-FU and oxaliplatin and the clinical outcomes of mCRC patients. Results show that the PDO responses to 5-FU and oxaliplatin mirror the progression-free survival (PFS, R = −0.85, p = 0.0075 and R = −0.89, p = 0.0032) and overall survival (OS, R = −0.81, p = 0.015 and R = −0.57, p = 0.14) from the moment of metastatic diagnosis in corresponding patients (Fig. 9). Reducing the number of PDOs from 250 to 10 per well did not weaken the correlations with patient outcomes (Supplementary Fig. 9). This alignment underscores the potential utility of miniaturized PDO assays as predictive tools for treatment outcomes in metastatic CRC.
Fig. 9: Correlation of PDO responses using 10 PDOs per well and size change metastatic lesions, PFS and OS of patients treated with 5-FU and oxaliplatin combination treatment.
A Correlation between tumor size change and PDO response to 5-FU and oxaliplatin. B Correlation between patient progression-free survival and PDO response to 5-FU and oxaliplatin. C Correlation between patient overall survival and PDO response to 5-FU and oxaliplatin. Data points represent different PDOs. Measurements shown are the mean of different biological replicates. AUCs are based on mean CyQUANT intensity drug response curves.
