Optimizing drug sensitivity assays in patient-derived tumor organoids: a comparison of IC50 estimation methods and experimental parameters

Abstract

Patient-derived tumor organoids (PDOs) hold immense potential for personalized drug sensitivity testing, but accurate efficacy determination is crucial for clinical translation. This study investigated factors influencing the accuracy and reproducibility of drug sensitivity measurements in PDOs, focusing on half-maximal-inhibitory-concentration (IC50) calculation methods, drug concentration numbers, and plate types. PDOs were established from six primary cancer tissues, including two cervical resections, one lung biopsy, one lung pleural effusion, one breast biopsy, and one gastric resection. They were subjected to drug sensitivity assays with 21 single/combined treatments, encompassing chemotherapy and targeted therapy drugs, with concentrations standardized in fold. Utilizing 6- and 12-concentration setups, IC50 derived from GraphPad-Dose-response–Inhibition (DRI), LC-logit, and LC-probit methods were compared. Relative changes (RCs) in IC50 and area-under-the dose-response-curve (AUC) between setups and the impact of plate type on cell viability measurements were assessed. In the 12-concentration setup, no significant IC50 differences were observed among the calculation methods. Notably, GraphPad-DRI and LC-logit exhibited minimal RCs between the 6- and 12-concentration setups (0.035 and −0.033, respectively), indicating accurate IC50 quantification even with fewer drug concentrations. AUC correlated strongly with GraphPad-DRI-derived IC50 (R = 0.858) and demonstrated lower variance between technical replicates. Furthermore, opaque-bottom plates yielded higher precision in cell viability measurements compared to transparent-bottom plates. This study provides valuable insights into optimizing drug sensitivity testing in PDOs. By demonstrating the robustness of specific IC50 calculation methods and the feasibility of using fewer drug concentrations, this study contributed to the standardization and reliability of PDO-based drug sensitivity assays.

Introduction

Patient-derived tumor organoid (PDO) has emerged as a powerful tool in cancer research, offering a more physiologically relevant platform for drug sensitivity testing compared to traditional two-dimensional cell cultures [1]. PDOs recapitulate the genetic and phenotypic heterogeneity of the original tumor, enabling personalized medicine approaches to identify effective treatment strategies [2, 3]. The selection of “Organoids” as the “Method of the Year” by Nature Methods in 2017 [4], along with a report by Vlachogiannis et al. in 2018 demonstrating a strong correlation between in vitro PDO data and clinical responses [3], has led to the widespread use of organoid technology in translational studies [5–7] and drug discoveries [8–10]. Notably, the US Food and Drug Administration now allows the promotion of a drug or biologic to human trials without requiring animal testing, suggesting that toxicity testing based on animal alternatives such as human-derived organ-chips or organoids may be sufficient [11]. This shift also highlighted the critical importance of accurate and reliable drug sensitivity assays using human-derived organoids for the successful clinical translation of PDO technology [3, 12].

Plotting biological responses against their causal stimuli (typically on a logarithmic scale) often results in a sigmoid curve [13]. While this sigmoid relationship adequately describes many biological systems, applying transformations such as logit [14], probit [15], and angular [16] can linearize these curves. Linearization offers several advantages. First, it improves accuracy in estimating IC50—Linear regression on linearized data often provides more robust and reliable estimates of IC50 values compared to non-linear fitting of the original sigmoid curve. Second, it simplifies statistical analysis, as linear models are generally easier to analyze statistically and enable more straightforward comparisons between different drugs or experimental conditions. Third, it enhances visualization by making it easier to visualize and interpret linearized dose-response relationships, thereby aiding in identifying trends and patterns in the data. In most biological systems, logit (for curve fitted with logistic sigmoid function) and probit (for curve fitted with cumulative normal distribution function) transformations yield an accurate linearization [13, 17–19]. However, to our knowledge, no study has directly compared the half-maximal inhibitory concentration (IC50) values obtained using logit and probit methods in PDO drug sensitivity tests, nor has the impact of these methods on reproducibility when transitioning from a higher to a lower number of drug concentrations been investigated.

The IC50 is a widely used metric for assessing drug efficacy in various models, including PDOs [19]. However, calculating IC50 can be challenging in certain scenarios, including (i) the complexity to determine a suitable dilution factors and number of concentrations that sufficiently cover the required range to accurately estimating the IC50 [13] and (ii) the requirement to deal with drug-resistant tumors or heterogeneous responses [20, 21]. In such cases, alternative parameters like the area under the dose-response curve (AUC) may provide a more robust assessment of drug sensitivity. While numerous publications have reported on drug sensitivity assays with PDOs [22, 23], there is limited study on how to enhance consistency across different operators. Continuous effort on improving data consistency and minimizing operational errors are crucial for establishing standard protocols for a more robust organoid assays in drug assessment.

First, this study aimed to compare the IC50 obtained from different IC50 estimation methods provided by GraphPad Prism software and R package, with former allowed logit sigmoid transformation and latter allowed both logit and probit transformations. Second, this study also investigated the impact of experimental parameters, such as the number of drug concentrations tested (6 vs. 12 setup), on the accuracy and reproducibility of drug sensitivity measurements. Next, the utility of AUC as an alternative to IC50 for drug sensitivity testing in PDOs was evaluated. Lastly, a proficiency test was conducted with five experienced operators to explore ways to improve reproducibility/consistency of organoid drug sensitivity results, based on influence of a technical factor, that is, the use of transparent-bottom plate with opaque supporting wall versus opaque cell culture plates during luminescent signal collection, on the precision of cell viability measurements.

Materials and methods

Patient specimens

Six primary cancer tissues were collected between March and November 2023. These included two cervical resections, one lung biopsy, one lung pleural effusion, one breast biopsy, and one gastric resection. All tissue samples were collected from individual patients, including five females and one male, with ages ranging from 36 to 80 years (Table 1).

Table 1.

Demographics and clinical information of six patients included in this study and drugs used for sensitivity test.

Sample	Gender	age	Date of collection	Surgical procedure	Cancer tissue	Histological type of primary tumor	Pathological staging (TNM)	Drugs or combination screened
Cervical cancer 1	Female	57	2023-05-11	Resection	Cervical	NA	NA	Vitamin C
Cervical cancer 2	Female	36	2023-08-25	Resection	Cervical	Cervical adenocarcinoma	NA	Cisplatin; Carboplatin
Lung cancer 1	Male	66	2023-11-15	Pleural effusion	Lung	lung squamous cell carcinoma	Ct3n2m0-IIIB	Gemcitabine; Anlotinib; Cisplatin; Paclitaxel; Docetaxel
Lung cancer 2	Female	76	2023-10-18	Biopsy	Lung	lung adenocarcinoma	NA	Anlotinib; Gefitinib; Crizotinib; Ceritinib; Erlotinib; Afatinib; Alflutinib; Osimertinib; Almonertinib
Breast cancer	Female	80	2023-03-21	Biopsy	Breast	Breast Cancer	IV	Cisplatin; Carboplatin; Pemetrexed; Alflutinib; Almonertinib; Pemetrexed + Carboplatin; Pemetrexed + Cisplatin; Almonertinib + Cisplatin; Pemetrexed + Cisplatin + Alflutinib
Gastric cancer	Male	70	2023-08-09	Resection	Gastric	Gastric adenocarcinoma	NA	Pirarubicin; Vitamin C

Abbreviations: NA, not available.

All patient samples were collected at Hangzhou Cancer Hospital and processed in the laboratory within 48 h, with sample shipping temperature maintained at 2–8°C. The use of human materials in this study has received ethical approval from the Hangzhou Cancer Hospital (HZCH-2021 No.010). Prior to the collection of samples, written consent for research purposes was obtained from each donor.

Cold instruments were used during samples’ collection, and cancer-rich areas were targeted while avoiding areas with high stiffness, necrosis, and large fat masses. For surgical and endoscopic specimens, at least 0.5 g was collected. For biopsy specimens, the total length was greater than 3 cm. For malignant effusion specimens, at least 300 ml was collected.

Preparation of malignant effusion for organoid culture

The establishment of organoid culture from malignant effusion was modified from a published protocol [24], with the optimized protocol as follows. Before processing the samples, the percentage of cells with a diameter greater than 12 μm was measured using a cell counter. This diameter threshold allowed for the exclusion of most immune cells. Samples meeting the acceptance criteria, with over 30% of cells having a diameter greater than 12 μm, were selected for PDO construction. Upon arrival at the laboratory, the sample was first transferred to a 250 ml centrifuge tube, centrifuged at 300 g for 5 min to collect the cell sediment. Next, the sediment was washed more than five times with cold Dulbecco’s Phosphate Buffered Saline (DPBS) buffer (BDBIO, China). If the cell sediment appears red, indicating the presence of red blood cells, a cell lysis step was performed using the Red Blood Cell Lysis Solution (Sangon Biotech, China).

Preparation of tissue samples for organoid culture

The preparation of tissue samples was modified from the same published protocol [24], with the optimized protocol detailed below. Before processing the samples, the tissue was washed thoroughly using cold DPBS. The tissue was rinsed repeatedly until the washing solution appeared clear (approximately 5–10 times). Then, the tissue was cut finely into pieces of about 0.5–1 mm³, as uniformly as possible, using sterile scissors. Depending on the size of the tissue, add 3 to 6 ml of tissue digestion solution (AimingMed Hangzhou, China). The tissue suspension was placed on a shaker at 37°C for 30–60 min for digestion. Every 20 min, the tissue digestion solution was pipetted to facilitate the process until the cells in the tissue are dissociated into single cells or small cell clusters. After digestion, a 100 μm filter was used to remove remaining tissue fragments. The filtrate was then centrifuged at 300 g for 5 min to obtain the cell sediment.

Establishment, passaging and characterization of tumor organoids

Tumor organoid establishment was performed according to the manufacturer’s protocol of MasterAim Organoid Culture Medium (AimingMed Hangzhou). The cell sediment was resuspended in 1 ml of Advanced DMEM/F12 medium (Thermo Fisher Scientific, MA, USA), and the number of cell clusters was counted using either an automatic cell counter or a manual counting plate. The cells were then resuspended at a density of 4 × 10³ cells/μl in Matrigel (Corning, NY, USA). The resuspended cells were inoculated into a cell culture plate pre-warmed to 37°C, ensuring that the droplets were placed in the center of the culture well. For a 24-well plate, 50 μl per well was inoculated. The culture plate was allowed to rest at 37°C for 5 min, then inverted and placed for 25 minutes. Afterwards, 500 μl of Complete MasterAim Organoid Culture Medium (AimingMed Hangzhou) was added.

When passaging the organoids, culture medium was removed with a pipette, 0.5 ml per well of TrypLE Express (Thermo Fisher Scientific) was added to the organoid culture plate. Matrigel domes were dispersed, and then digested at 37°C for 5–10 min. When most organoids were digested into cell clusters (about 20–50 μm in size) as observed under a microscope, twice the volume of DPBS was added to stop the digestion.

For the hematoxylin and eosin (H&E) staining and Immunohistochemistry (IHC) analysis of tissues and organoids, cancer tissues and organoids were fixed in 4% paraformaldehyde at 4°C for 24 h, dehydrated, and embedded in paraffin for sectioning. Organoids were embedded in agarose before paraffin processing. Sections were stained with H&E for morphology or processed for IHC to detect KI67, CEA, CK7, P16, and P40 (Abcam, USA).

Drug sensitivity assay with tumor organoid

Cultured organoids were digested into cell clusters as per the passaging steps, then centrifuged at 300 g for 5 min. Cell sediments were resuspended in 60% cold Matrigel (100 cells/μl) and inoculated into a 384-well plate by 10 μl per well (Corning). The entire step was performed on ice. After inoculation, 384-well plate was placed in a 37°C incubator for 30 min, then 45 μl of the complete MasterAim Organoid Culture Medium (AimingMed Hangzhou) was added into each well. Two days after the culturing, drug sensitivity experiments were conducted for six 4-fold and/or twelve 2-fold serially diluted concentrations, respectively, with triplicates, starting from standardized 4-fold unit. The opaque white cell culture plates (781080, Greiner Bio-One, Austria) were used in all the drug sensitivity tests. Drug concentration testing range for each drug in this study was determined by considering the Cmax value from DrugBank 6.0 [25], with reference on IC50 data derived from cancer cell lines from the Genomics of Drug Sensitivity in Cancer [26]. The Cmax was positioned at the midpoint of the dilution series (Drug concentration at 4-fold unit for each drug is provided in Supplementary Data Table). All the drug compounds were purchased from the MedChemExpress (China). Seventy-two hours after drug administration, CellTiterGlo 2D (Promega, WI, USA) luminescent reagent was used to read the luminescence data with a microplate reader (Tecan, Switzerland).

Factor on plates’ type during luminescent signal collection

A separate study was conducted to investigate influence of culture plate types on reproducibility of test results. Five individual operators were tasked to perform same drug sensitivity testing on Cisplatin and Gemcitabine drugs combination in the lung cancer pleural effusion sample, using transparent-bottom plates with white supporting wall (781098, Greiner Bio-One) and opaque white cell culture plates (781080, Greiner Bio-One) during luminescent signal collection for cell viability measurement, respectively.

Data analyses

Relative cell viability percentage of each tested sample with replicates was derived according to published protocols [24, 27]. IC50 and AUC were determined based on the relative cell viability percentage, using the GraphPad Prism 8.2 (GraphPad, CA, USA) and the Ecotox R package version 1.4.4. The IC50 of an experiment set was determined using non-linear regression curve fit model from GraphPad Prism, according to the software’s instructions. Briefly, under the Dose-response—Inhibition (DRI) module, [Inhibitor] vs. normalized response—variable slope was selected to obtain the IC50. On the other hand, the AUC (i.e. Total Area in the software result) was obtained using the Analyze—XY analyses—area under curve module. Separate IC50 values were determined using Lethal Concentration Logit (LC-logit) and Probit (LC-probit) from the Ecotox (R package), respectively, according to the package’s instructions, for comparison purpose. The RC in this study was calculated with difference of IC50/AUC derived from 6-concentrations versus 12-concentrations (i.e. values from 12-concentration minus that from 6-concentration), that was subsequently normalized (i.e. divided) using IC50/AUC derived from matching 12-concentration value. A P-value below .05 was considered statistically significant in this study.

Results

This study evaluated 21 unique single or combined drug treatments, including 8 chemotherapy agents (Vitamin C, Cisplatin, Carboplatin, Gemcitabine, Paclitaxel, Docetaxel, Pirarubicin, and Pemetrexed) and 9 targeted therapy (Anlotinib, Gefitinib, Erlotinib, Afatinib, Alflutinib, Osimertinib, Crizotinib, Ceritinib, and Almonertinib) agents. The selection of the 21 drug treatments was based on clinical relevance, efficacy data, and established guidelines. The drugs were chosen in consultation with clinicians, primarily following first-line chemotherapy regimens for specific cancer types. For example, Cisplatin and Carboplatin are first-line treatments for cervical cancer, while Gemcitabine, Paclitaxel, Docetaxel, and Cisplatin are commonly used in lung cancer. PDOs derived from patient tissues were subjected to drug sensitivity testing using up to nine single or combined cancer drugs, resulting in a total of 28 12-concentration sensitivity tests. All the 28 12-concentration tests were subjected to IC50 estimation using the GraphPad-DRI, LC-logit, and LC-probit methods. Due to the varying working concentrations of different drugs, standardized drug concentration values were used to facilitate better comparison of the dose response curves. As shown in Supplementary Data Table, the highest concentration for each drug was standardized to a 4-fold unit, followed by six 4-fold and twelve 2-fold serial dilutions. Histopathology preservation in patient-derived organoids (PDOs) from various tissue types is demonstrated in the Supplementary Figure. Figure 1 shows Day-3 brightfield images of PDOs obtained from different tissue types that were either treated with Cisplatin, Pirarubicin, or negative control (without drug treatment) in a 384-well cell culture plate.

Figure 1.

Brightfield images (Post 72 h) of organoids derived from different tumor tissue types that were either treated with Cisplatin, Pirarubicin (with a range of drug concentrations) or negative control (without drug treatment) in a 384-well cell culture plate. The top bar shows standardized drug concentration values in fold unit

Out of the 28 test results, 11 (39%) yielded an IC50 fold unit exceeding 4.0 in at least one IC50 estimation method and were thus considered beyond the limit of quantification. The remaining 17 (61%) results were subjected to RC comparison study, comparing IC50/AUC derived from 6-concentration versus 12-concentration results. Pairwise comparisons were conducted on RC values obtained using the respective GraphPad-DRI, GraphPad-AUC, LC-logit, and LC-probit methods.

Pairwise comparisons of the GraphPad-DRI, LC-logit, and LC-probit methods revealed no statistically significant differences in the derived IC50 values across the 17 drug sensitivity experiments, with medians (ranges) of 0.672 (0.067–3.875), 0.498 (9.47 × 10⁻⁸–2.759), and 0.464 (1.68 × 10⁻⁸–3.666), respectively (Fig. 2A; Table 2). Separately, in comparisons of IC50/AUC values derived from 6-concentration versus 12-concentration experiments, GraphPad-DRI and LC-logit methods exhibited the lowest RCs, with median RCs of 0.035 and -0.033, respectively (Table 2). Noteworthy, no statistical difference was found between the 2 groups of RC values (P = .0890, Fig. 2B). A positive RC indicated that the 12-concentration results produced a higher IC50/AUC than its corresponding 6-concentration experiment, while a negative RC indicated that the 12-concentration results produced a lower IC50/AUC that its corresponding six-concentration experiment. On the other hand, the LC-probit method, with a median RC of −0.077, exhibited significantly lower RC compared to the LC-logit method (P =.0079).

Figure 2.

Parallel coordinates plots comparing (A) half-maximal inhibitory concentration (IC50) derived using 12-concentration experiments and calculated using GraphPad Dose-response—Inhibition (DRI) module, Lethal Concentration (LC)-logit, and LC-probit methods and (B) relative change (RC) values derived using GraphPad-DRI, GraphPad-AUC, LC-logit, and LC-probit methods. Paired Wilcoxon signed-rank test was used to compare the value differences between the methods in pairwise

Table 2.

Methods and statistics of half-maximal inhibitory concentration (IC50) and area under the drug response curve (AUC) derived from 12-concentration experiments and relative change (RC) when comparing values derived from 6-concentration to that from 12-concentration experiments.

Method	Analyzed value	Median IC50/AUC (range) by 12 concentrations	Median relative change (RC, range)
Graphpad-IC50	Cell viability	0.672 (0.067 to 3.875)	0.035 (−0.515 to 2.169)
LC-Logit-IC50	Cell viability	0.498 (9.47 × 10−8 to 2.759)	−0.033 (−0.893 to 1.223)
LC-Probit-IC50	Cell viability	0.464 (1.68 × 10−8 to 3.666)	−0.077 (−0.936 to 1.190)
Graphpad-AUC	AUC	0.382 (0.102 to 0.803)	0.160 (−0.246 to 0.467)

Notably, the AUC derived from the GraphPad-AUC method achieved high positive correlation with the IC50 estimated from the GraphPad-DRI method (R = 0.858, P < .001; Fig. 3). Lastly, dose-response curves for the Cisplatin and Gemcitabine drug combination tested using both transparent-bottom (Fig. 4A) and opaque-bottom (Fig. 4B) plates, by the five operators, were generated. Analysis of these curves revealed greater precision, both between and within operators, when using opaque-bottom plates. Notably, higher variability in cell viabilities within triplicate testing of an operator was observed at lower drug concentrations. The maximum standard deviation observed with transparent-bottom plates was 16.25 at a 0.06250-fold concentration, while the maximum standard deviation with opaque-bottom plates was 14.49 at a 0.00391-fold concentration. Comparing between operators, the median standard deviation of cell viabilities was 7.87 (range: 2.49–17.32) for transparent-bottom plates and 3.93 (range: 0.86–8.51) for opaque-bottom plates. Consequently, the opaque-bottom plates yielded lower CV for both IC50 and AUC estimates (0.23 and 0.09, respectively) compared to the transparent-bottom plates (0.49 for IC50 and 0.21 for AUC, Table 3).

Figure 3.

Correlation between area under the drug response curve (AUC) derived from the GraphPad-AUC method and half-maximal inhibitory concentration (IC50) derived from the GraphPad Dose-response–Inhibition method

Figure 4.

Drug response curves show variabilities of cell viabilities (derived from luminescent signal) obtained from 5 operators using (A) transparent plate and (B) opaque plate. Sigmoidal (logistic) function was used for curve fitting for each operator. The table shows the mean and standard deviation (in bracket) of cell viability triplicates by each operator obtained at different concentrations

Table 3.

Comparison of drug sensitivity measures for Cisplatin and Gemcitabine drugs combination in a lung cancer pleural effusion sample, repeated by five individual operators on transparent and opaque plates, respectively.

Method/calculated values	Operators					Mean (SD)	CV
	OP1	OP2	OP3	OP4	OP5
Transparent plate
IC50	0.17	0.47	0.14	0.40	0.47	0.33 (0.16)	0.49
AUC	0.24	0.33	0.44	0.36	0.32	0.34 (0.07)	0.21
Opaque plate
IC50	0.22	0.19	0.19	0.22	0.12	0.19 (0.04)	0.23
AUC	0.27	0.26	0.25	0.28	0.22	0.26 (0.02)	0.09

Abbreviations: IC50, half-maximal inhibitory concentration; AUC, area under the dose-response curve; SD, Standard Deviation; CV, coefficient of variation.

Discussion

Noteworthy, the initial standardized drug concentration (4-fold unit) used for the 2- or 4-fold serial dilutions was determined based on the Cmax value, representing the maximum serum concentration that a drug achieves after administration but before a second dose, which is also considered as safe dosage in human body. When an IC50 estimation exceeded the upper quantifiable range (i.e. the 4-fold unit), the drug was considered insensitive to the tested PDO. Conversely, if an IC50 estimation fall below the lower quantifiable range (i.e. the 0.00391-fold unit), the drug could still be considered potentially sensitive to the PDO.

In this study, the GraphPad-DRI method was selected as the reference for comparison due to its widespread use in conventional pharmacological research and PDO drug sensitivity assays [1, 27]. No significant difference in IC50 was observed among the GraphPad-DRI, LC-logit, and LC-probit methods when using the 12-concentration setup (Fig. 2A). However, when comparing IC50 results between 6-concentration result and corresponding 12-concentration result, GraphPad-DRI and LC-logit demonstrated minimal and comparable RC values (medians of 0.035 and -0.033, respectively; Table 2 and Fig. 2B). This suggested that these two methods may offer greater accuracy and/or reduced transitional errors for laboratories transitioning to a 6-concentration experimental setup for IC50 drug sensitivity quantification. In comparison, LC-probit produced RC values with median of −0.077. Given that both GraphPad-DRI and LC-logit utilized a logistic sigmoid function to fit the dose-response curve[14], whereas LC-probit employed a cumulative normal distribution function[15], this suggested the logistic sigmoid may be more accommodating of requirement for an IC50 estimate with at least two concentrations that predicted responses both below and above 50% [13].

In general, organoid viability is assessed by measuring ATP levels with luminescent signals, whereas organoid morphology is often observed before lysing for ATP quantification. Consequently, transparent-bottom plates with white/black supporting wall are commonly used in tumor organoid drug sensitivity assays [1, 7]. However, the first two lower concentrations were often found to exhibit higher luminescent flux compared to the higher concentrations in this study. We hypothesized that this phenomenon may be attributed to luminescent signal leakage through the transparent bottoms of the plates. The lower variance found with opaque cell culture (opaque bottom) plates in this study supported this hypothesis. To reduce experimental costs and improve imaging efficiency of organoid morphology, one may perform drug treatments in transparent plates for easy sample handling and transfer equal amounts of cell lysates to an opaque plate for cell viability measurement via ATP levels.

Assessing clinical consistency between organoid drug sensitivity and patient clinical response is a fundamental goal of tumor organoid research [28]. While both AUC and IC50 are employed to evaluate drug sensitivity across various models [1], IC50 is more prevalent in organoid studies [1]. However, PDOs are more commonly collected from patients with advanced cancer, chemotherapy often encounters resistance [20], potentially hindering IC50 calculation and introducing variability across experiments. The AUC is often used as a robust parameter aiming to compare potency and efficacy of one drug agent across different tissues [28]. Admittedly, only a maximum of 3 individual tissues were tested for any of the 21 single/combined drugs in this study (Supplementary Data Table). Despite this limitation, a strong correlation between AUC values across multiple tissues and IC50 values obtained using the GraphPad-DRI method was observed (Fig. 3). Moreover, AUC values demonstrated lower variability across technical replicates (Table 3). These results suggest that AUC may serve as a reliable alternative for evaluating the efficacy of different drugs on a given tissue, particularly in scenarios where IC50 calculation is challenging. AUCs were not calculated for the LC-logit and LC-probit methods, as these values or the corresponding fitted dose-response curve functions are not currently provided by the Ecotox R package.

As more clinical and research laboratories are adopting PDO drug sensitivity tests, it is essential to establish commonly accepted standards for organoid culture and detection methods to enhance the reliability of test results across different labs. While some studies have suggested introducing automated sample/liquid handling equipment into organoid culture [29, 30], the current techniques are not yet mature, and most research institutions still rely on manual labor. Despite existing published standards [1, 31] and consensus guidelines [32], there is limited discussion on how to reduce technical errors or improve operators' proficiency in generating more reliable test result. This study highlighted the importance of standardized operational procedures and measurement criteria in proficiency testing, demonstrating that using opaque-bottom plates improves result reproducibility among experienced operators and may enhance consistency across different laboratories.

While this study provided valuable insights into drug sensitivity testing using PDOs, it is important to acknowledge the limitation posed by the small sample size. The number of patient samples recruited in this study (N = 6) may not fully capture the heterogeneity and complexity of various cancer types [21]. Consequently, the findings might not be generalizable to a broader patient population. Future studies with larger and more diverse cohorts are necessary to validate and expand upon these results. Besides, conducting all experiments in triplicates may not be satisfactory for such studies. Drug screening using PDOs faces challenges, primarily the labor-intensive nature of current culture methods and limited tissue availability, which can hinder high-throughput screening and PDO generation. To address these limitations, laboratories can consider automating the culture process for increased efficiency. Furthermore, to maximize the utility of limited PDOs, a more deterministic approach to drug panel selection is recommended. This involves close collaboration with the patient’s physician to select drug panels based on clinical relevance and established guidelines. Ideally, this selection process would be further enhanced by incorporating a prediction/optimization model to identify drugs with a higher likelihood of efficacy for the specific patient’s tissue.

In conclusion, this study has important implications for improving the accuracy and reproducibility of drug sensitivity measurements in PDOs. This study was the first to demonstrate that using a 6-concentration experimental setup (Given the same min-max of concentration range as one with higher concentrations setup) with either the GraphPad-DRI or LC-logit method can provide accurate IC50 estimations, potentially streamlining workflows without sacrificing data quality. This study also highlighted the potential for signal leakage when using transparent-bottom plates and recommends the use of opaque plates or transferring lysates to opaque plates for luminescence measurements to minimize variability. Furthermore, the use of AUC was proposed as a robust alternative to IC50, particularly in cases where drug resistance makes IC50 calculation challenging. Finally, the proficiency test underscored the need for standardized procedures and emphasized that seemingly minor technical factors, such as plate selection, can significantly impact result reproducibility across different operators and laboratories. These findings may contribute to the ongoing effort to standardize PDO drug sensitivity testing and enhance its reliability in both clinical and research settings.

Author contributions

Yidan Chen (Investigation [equal], Methodology [equal], Project administration [lead], Visualization [equal]), Jian Zhang (Project administration [equal]), Bin Zhang (Project administration [equal]), Hongkai Lee (Conceptualization [equal], Methodology [equal], Supervision [equal]), Suyao Xie (Investigation [equal], Visualization [equal]), Xueqin Chen (Funding acquisition [equal]), Mingliang You (Funding acquisition [supporting], Supervision [supporting]), Chongyang Shen (Supervision [equal]), and Bing Xia (Conceptualization [equal], Funding acquisition [lead], Supervision [equal]), Huayang Xing (Conceptualization [lead], Supervision [lead])

Supplementary data

Supplementary data is available at Biology Methods and Protocols online.

Conflict of interest statement. None declared.

Funding

This work was supported by Science and Technology Development Project of Hangzhou (Grant No. 202004A19 and 202204A08). The publication-related fees was sponsored by Grant 202204A08.

Data availability

The data underlying this article are available in the article and in its online supplementary material.