An organoid-guided platform for ovarian cancer: enabling prediction of patients’ chemotherapy response

Abstract

Background

Organoids represent a new platform for drug screening and personalized medicine. However, the difficulty of organoid construction limits their wide application.

Methods

We collected 153 tumour samples from patients with epithelial ovarian tumours. The associations between patient characteristics and organoid generation were analysed via chi-square tests and Fisher’s exact tests. Univariate and multivariate logistic regression analyses were performed to identify independent factors influencing organoid development. We conducted a drug screen on 20 organoids to predict the drug response of clinical patients retrospectively and prospectively.

Results

One hundred fifty-three organoids were developed from 153 ovarian tumour patients, with a 57.52% success rate. Preoperative low CA153 levels (OR (95% CI): 3.44 (1.39, 9.00), P = 0.009) and high CA199 levels in patients (OR (95% CI): 0.20 (0.06, 0.57), P = 0.004) correlated with successful organoid generation, whereas other clinical features were not significantly correlated with ovarian tumour organoid generation. The subgroup analyses further showed that CA153 and CA199 were two independent factors influencing organoid construction. Ovarian cancer organoids can retain the pathological and genetic characteristics of the original tumour tissues. The PDOs were able to predict the prior clinical responses of these patients with an efficiency rate of 100%. The prospective prediction efficiency of PDOs was 80%.

Conclusions

Preoperative CA153 and CA199 levels were found to be independent factors influencing ovarian tumour organoid generation. The drug responses of most PDOs to paclitaxel and carboplatin were consistent with the clinical treatment outcomes.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13048-026-02178-8.

Introduction

Ovarian cancer (OC) is the most lethal gynaecological malignancy. A recent annual report indicated that OC accounted for approximately 19,680 newly diagnosed cancer cases and 12,740 cancer-related deaths in the USA [1]. Owing to nonspecific symptoms and limited screening methods, 70% of patients are diagnosed with ovarian cancer at an advanced stage, and the overall survival rate is 30.2% [2]. Advanced-stage cancer recurs and develops drug resistance after initial platinum-based chemotherapy [3].

Due to the high heterogeneity and strong resistance of ovarian cancer, preclinical model-guided drug screening is highly important. Although cancer cell lines are the most commonly used model in medical studies, they cannot recapitulate the molecular characteristics of tumours in vivo [4]. Animal xenograft models utilizing human-derived cancer cells are highly expensive and time consuming [5]. Thus, organoids have emerged as an innovative and powerful preclinical research model. The success rate of organoid construction in vitro far exceeds that of stable cancer cell line establishment from the same tissue [6]. These methods can be applied to reveal the genomic and mutational landscape and screen for drug sensitivity in clinical patients [7]. As organoid technologies have developed, organoid biobanks for various cancers, such as colorectal cancer [8], breast cancer [9], and cervical cancer [10], have been established and have been shown to serve as drug screening platforms. Research on ovarian cancer organoids is increasing, and these models, serving as a guide for drug screening, represent a novel direction for treatment decisions in clinical patients [7, 11–13].These resources enable studies to clarify cancer development and explore innovative therapeutic regimens [14].

However, the success rate of organoid generation varies among different tumour tissues [15–17]. In addition to the tumour type, the discrepancy in the success rate relies on a series of clinicopathological and experimental factors. Dustin Deming et al. suggested that paucicellular tissue, necrotic components and contamination contributed to cases of failure [18]. Guang-Wen Cao et al. reported that a larger tumour size, microvascular invasion (MVI), macrovascular invasion, advanced TNM stage, and advanced Barcelona Clinic Liver Cancer (BCLC) stage influenced the success rate of constructing hepatocellular carcinoma (HCC) organoids [19]. The sample size, purity, and access to tumour tissue (biopsy or surgical procedure) are considered to impact patient-derived organoid (PDO) generation [6, 20]. However, Markus H. Heim et al. reported no significant correlations between a comprehensive set of clinical data and the growth and success rates of HCC organoids [21]. Ovarian cancer organoids were generated by Kopper et al. and Maenhoudt et al., with success rates of 65% and 56%, respectively [7, 11]. Kopper et al. transplanted OC organoid orthotopically or subcutaneously into immunodeficient mice to perform drug screening test. The results were consistent with these clinical observations [11]. Maenhoudt et al. used OC-derived organoids for in vitro drug screening involving several chemotherapeutic agents [7]. De Witte et al. just used five patients-derived PDOs to reflect patients’ clinical response to chemotherapy and targeted drugs [13]. Hill et al. generated short-term PDOs of ovarian cancer to identify targetable DNA damage repair defects and predict the clinical response of DNA repair inhibitors [22].

Organoids, as 3D in vitro models, have been utilized by some researchers with distinct construction methods [23–25], while others just employed them for in vitro drug sensitivity testing. However, there is a scarcity of studies tracking and comparing clinical treatments, and only a few publications have incorporated clinical data, albeit with small sample sizes. The establishment of ovarian cancer organoids remains a challenge. This study aims to analyze experimental data alongside patients’ clinical characteristics to identify factors influencing organoid construction, with the goal of providing insights for optimizing ovarian cancer organoid models in the future. Additionally, by integrating organoid drug sensitivity testing with neoadjuvant therapy, adjuvant therapy for initially treated patients, and clinical outcomes in recurrent cases, we seek to comprehensively evaluate the application value of organoids in these areas, identify patients with drug-resistant ovarian cancer at an early stage, and thoroughly assess the reliability of ovarian cancer organoids as an in vitro drug prediction model.

Methods

Construction of ovarian tumour organoids

Specimen selection and collection

Organoid-related research was approved by the Ethics Committee of Chongqing University Cancer Hospital (CZLS2020274-A). The samples were collected from patients with epithelial ovarian cancer who were admitted to the hospital for surgery. The sample selection criteria are as follows. Primary tumour or metastatic lesions were obtained, including those in the peritoneum, intestine, and lymph nodes, from patients who underwent tumour debulking surgery or biopsy. Recurrent lesions were obtained by a second debulking surgery or laparoscopic abdominal exploration. With the assistance of pathologists, we collected fresh surgical samples to generate organoids. The included tumour tissues included high-grade serous ovarian cancer (HGSOC), low-grade serous ovarian cancer (LGSOC), ovarian endometrioid carcinoma (EOC), ovarian mucinous carcinoma (MOC), and serous/mucinous borderline ovarian tumour (SMBOT) samples. The exclusion criteria include patients with incomplete clinical data, insufficient samples, or poor sample quality. Fresh tumour tissues typically appeared as pink, cauliflower-like regions or nodules. We should avoid the areas ablated by the harmonic scalpel to collect tissue specimens. The pale white, light yellow and dark red tissues presented poor tissue activity. The collected tumour samples, which were maintained in ice-cold culture media, were transferred to the laboratory for further processing within one hour.

Tissue digestion and embedding of cells in suspension

The retrieved tissue was immediately washed with ice-cold PBS. Blood, adipose tissue, necrotic tissue, and epithelial components were removed. Pink, fresh, and tender tissues were subjected to immunohistochemical staining, cryopreservation, and organogenesis (Fig. 1A). The mechanical shearing and enzyme digestion methods were combined. The tissue pieces were gradually digested with 2X TryPLE™ (Gibco, Catalogueg #: A12177-02) at 37 °C for 10–40 min and then filtered through a Falcon^® 100 μm cell strainer (Corning) to obtain a cell suspension. The cell precipitate was obtained by centrifugation at 1200 rpm for 5 min at 4 °C. Erythrolysis was performed if red blood cells remained. Finally, the properly centrifuged cell precipitate was suspended in Matrigel^® (Corning, Catalogueg #: 356231) and plated in a 24-well plate (1 × 10⁶ cell in 50 µl droplets). After coagulation for 20–30 min, the cells embedded in Matrigel were cultured in a 5% CO₂ incubator at 37 °C with modified organoid medium [11] consisting of advanced DMEM/F12 supplemented with 1% penicillin/streptomycin, 1% GlutaMAX, 10 mM HEPES, 1:50 B27 supplement, 1.25 mM N-acetyl-L-cysteine, 250 ng/ml recombinant human R-spondin-1, 100 ng/ml recombinant Noggin, 10 mM nicotinamide, 100 µg/ml primocin, 500 nM A83-01, 10 ng/ml recombinant human EGF, 500 ng/ml hydrocortisone, 37.5 ng/ml recombinant human Heregulinβ-1, 10 ng/ml recombinant human FGF10, 10 nM β-oestradiol, 10 µM forskolin, and 10 µM Y-27,632 to generate organoids (Table S1).

Fig. 1.

Diagram depicting the generation of ovarian tumour organoids and their morphological characteristics. A Workflow of the construction of PDOs derived from primary ovarian tumour and metastasis tissue. Fresh tumour lesion is necessary for the successful generation of PDOs, necrosis and normal tissue should be identified and removed. The collected tissue was departed into three parts for organogenesis, cryopreservation, and IHC testing. B The tumour tissue was digested by Trypsin (1X), and the obtained cells were derived into PDOs. The small cell clusters were advantageous to successfully generate organoids, compared with the obtained scattered single cells. C The varied morphology of ovarian tumour organoids among different PDOs

Morphological and pathological characterization of PDOs

The growth and evolution process of the PDOs was dynamically observed under a Leica inverted microscope. For PDO characterization, cell recovery solution (Corning, Catalogue #: 354253) was used to dissociate PDOs from Matrigel at 4 °C for 40 min, followed by centrifugation (300 × g, 5 min, 4 °C). Paraformaldehyde (4%) was added to fix the PDO precipitate overnight. The pellet was embedded in paraffin according to standard immunohistochemical procedures. The biological characteristics of the generated PDOs were verified via haematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining in reference to the parental tumours. Four-micrometer-thick paraffinized PDO sections were subjected to deparaffinization, rehydration, and subsequent antigen retrieval. The sections were then immersed in PBS containing 0.3% Triton X-100 for 20 min for permeabilization, followed three rinses with PBS for 5 min each. Goat serum was used for blocking for 60 min. The sections were incubated with primary antibodies overnight at 4 °C. The primary antibodies used for IHC included anti-PAX8 (MXB Biotechnologies, RMA-0817, 1:100), anti-p53 (MXB Biotechnologies, MAB-0674, 1:1), and anti-Her2 (MXB Biotechnologies, kit-0043, 1:1) antibodies. The sections were then incubated with the corresponding horseradish peroxidase-labelled secondary antibodies for 60 min at room temperature. The DAB reaction mixture (Invitrogen, Catalogue #: 34065) was subsequently added. The sections were stained with haematoxylin dye for nuclear staining and preserved with neutral resin. A pathological section scanner (KONFOONG Bioinformation, MAGSCANNER, KF-PRO-005-HI) was used to collect images.

Genomic analysis of parental tumours and organoids

DNA was extracted from tumour tissues and matched PDOs using the DNeasy Blood & Tissue Kit (Qiagen, Germany) according to the manufacturer’s protocol. A total of 0.2 µg of DNA per sample was used as the input material for DNA library preparation. The sequencing library was generated using the NEBNext^® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA, Catalogue #: E7370L) according to the manufacturer’s recommendations, and index codes were added to each sample. Quality control was applied to guarantee meaningful downstream analysis, and we used Fastp (version 0.23.1) to perform a basic statistical analysis of the quality of the raw reads [26]. Clean data were mapped to the human reference genome GRCh38 using Burrows Wheeler Aligner (BWA) software [27] and Samblaster [28] to generate BAM files. Sambamba [29] was subsequently used to sort the BAM files and mark duplicate reads according to the chromosome position. Somatic mutations were detected by comparing each cancer sample to the matched reference blood leukocytes. The somatic SNVs were detected by MuTect [30], whereas the somatic InDels were identified by Strelka [31]. Somatic CNAs (copy number alterations) were detected by analysing BAM files for read depth variations using Control-FREEC through a comparison of the tumours or organoids to reference blood leukocytes [32], and ANNOVAR was used to annotate the results to acquire genes in specific regions [33]. An analysis of the mutational signature was performed using the R package MutationalPatterns (v1.10.0) to calculate the optimal contribution of COSMIC signatures and determine the genomic context for all somatic SNVs in tumour tissues and organoids [34].

Clinicopathological characteristics of the patients

The clinicopathological features of the enrolled patients and experimental factors of the PDOs were collected by two gynaecologists, and these features were mutually checked before data analysis. The clinicopathological factors included age; FIGO stage (2009); BRCA status; surgery method; tumour size (the maximum diameter of the tumour); tissue source; preoperative serum cancer antigen 125 (CA125), HE4, CEA, and CA153 levels; pathology; disease status; and the proportion of Ki-67-positive cells. In addition, the experimental factors included the cell count and cell viability, which were detected with an RWD automatic counting instrument (C100-SE/C100).

Drug screening of patient-derived organoids

The successfully established organoids were digested into small spheres of 40 μm. Occasionally, the organoids are large and cannot be digested into small, uniform spheres. Therefore, we split the organoids by a combination of mechanical dissociation and TrypLE enzymatic digestion. Subsequently, the split organoids were passed through a Falcon^® 40 µM cell strainer to remove large organoids and obtain uniform, small organoids. The filtered organoids were centrifuged at 1200 rpm for 5 min at 4 °C and then resuspended in culture medium. Organoids were seeded into ultralow-attachment 384-well plates at a density of approximately 300 organoids/8µl in a 50% Matrigel/culture medium mixture. Two days after plating, the growth of the organoids was restored. A six-point dilution series of each drug was prepared and dispensed in culture medium lacking Y-27,632. The drug concentrations of paclitaxel (Yangtze River Pharmaceutical Group) were 100 µM, 10 µM, 1 µM, 0.1 µM, 0.01 µM, and 0.001 µM, and those of carboplatin (MedChemExpress, HY-17393) were 500 µM, 200 µM, 100 µM, 50 µM, 10 µM, and 1 µM. The tested drugs were replaced on the third day. The culture medium without Y-27,632 was designated the control group. Y-27,632 inhibits ROCK1 and ROCK2 by competing with ATP for binding to the catalytic site [35], which may interfere with assay results and lead to false positives. Cell viability was analysed using an ATPlite (CellTiter-Glo^® (Promega)) assay in accordance with the manufacturer’s instructions following 6 days of drug incubation, and the results were normalized to those of the corresponding control. The data were analysed with GraphPad Prism 8 software. The IC50 and AUC values were computed through the application of nonlinear regression (curve fit) and the equation log(inhibitor) against the normalized response [36]. Each drug dilution was replicated three times.

Statistical analysis methods

Successful organoid generation refers to producing a sufficient quantity of organoids for short-term experiments, and presenting morphological features of a whole tumour and could be passaged. The continuous variables were converted to categorical variables according to the cut-off values, which were determined by constructing receiver operating characteristic (ROC) curves. The categorical variables are described as counts (percentages). Pearson’s chi-square test and Fisher’s exact test were employed for the univariate analysis. The Spearman correlation coefficient was calculated to assess collinearity among these independent variables. Variables exhibiting statistical significance in the univariate analysis were included in multivariate binary logistic regression analysis, which was performed to evaluate factors independently influencing organoid generation. P values (odds ratios [ORs] and 95% confidence intervals [95% CIs]) were calculated to show the results of univariate and multivariate analyses. All the tests were two-sided. A P value < 0.05 was considered to indicate statistical significance. R version 4.0.5 was used for the statistical analysis with the gtsummary and pROC packages.

Data Availability

The data generated in this study are available within the article and its supplementary data files.

Results

Establishment and growth characteristics of ovarian tumour organoids

In total, we established 153 ovarian cancer organoids (overall success rate 57.52%) from 153 patients, including 123 patients with newly diagnosed ovarian cancer and 30 patients with recurrent ovarian cancer. Tumours from patients with newly diagnosed ovarian cancer had a higher success rate of organoid establishment than those from patients with recurrent ovarian cancer (76.67% vs. 52.85%). PDOs were generated from patients with diverse pathologies, including 126 HGSOC, 7 LGSOC, 5 endometrioid carcinoma, 4 mucinous carcinoma, 1 poorly differentiated cancer, and 10 borderline tumours. HGSOC tumours accounted for 82.35% of the samples, and the success rate was 52.38%; among these samples, those derived from newly diagnosed HGSOC had a lower success rate of 46.46%.

Organoids are usually generated successfully within 2–3 weeks and need to be passaged, which was consistent with Kopper’s results [11]. The growth of organoids varies with different conditions of the tumour tissue and cell suspensions. An effective tissue composition is vital for successful organoid generation. First, the fresh pink samples were handled immediately after detachment. Notably, the tumour and normal tissues, as well as the fresh and necrotic components, were identified well (Fig. 1A). The adipose tissue, epithelial components and necrotic parts were removed from the collected samples. The remaining tumours were divided into three parts for cryopreservation, immunohistochemical diagnosis, and organoid generation. The collected samples were subsequently digested via mechanical and enzymatic methods. The viability of the obtained cell suspensions varied across diverse tissue conditions. If the samples were off-white (rotted-like) or fresh pink, the viability of the cell suspension was generally less than 70% or more than 80%, respectively. In addition, the mixed state of cell clusters and single cells was a critical factor in the success rate of organoid culture (Fig. 1B). Owing to the high heterogeneity of ovarian cancer, the digested samples presented different states of cell suspension, with variable digestion times. Occasionally, ovarian tumour tissues were rapidly separated into a single-cell suspension after digestion and presented high viability, but they failed to generate PDOs (Fig. 1B).

PDOs maintain the histological characteristics of original tumour tissues

Initially, suspensions of single cells and cell clusters with Matrigel were seeded in 24-well plates. The formation and growth process of the organoids were observed dynamically. The organoids displayed diverse morphologies, including the following four categories: dense, cystic, solid-cystic and cellular cohesiveness (Fig. 1C). The solid-cystic structures showed the mixture of slight-moderate dense PDOs and thick-wall cystic PDOs with cells inside. A wide morphological spectrum was observed in distinct histological subtypes. HGSOC organoids displayed all the morphologies, with varying degrees of density and cell cohesiveness. Furthermore, mature cystic or solid-cystic ovarian organoids exhibited some folds and invaginations. The dense PDOs presented as black solid balls or irregular forms. Similarly, EOC organoids were dense with poor light transmittance and irregular morphologies. LGSOC organoids displayed a solid-cystic appearance with multiple lumens protruding outwards or exhibiting cellular cohesiveness. SMBOT organoids exhibited uniform solid or cystic morphologies.

For the comparison of PDOs and parental tumours, haematoxylin and eosin (H&E) staining and immunohistochemistry (IHC) were performed to evaluate the cellular characteristics and the expression of ovarian cancer biomarkers (e.g., PAX8, P53, and Her2 for HGSOC), respectively, in the samples. H&E staining revealed that the PDOs preserved the specific heterogeneous morphologies of the parental tumours. The cytological characteristics, including large and deeply stained nuclei, an irregular arrangement of cancer cells, and a decreased cytoplasmic ratio, were retained in PDOs (Fig. 2A). Next, specific tumour biomarkers of ovarian cancer were detected to verify the pathology of PDOs and parental tumours. The results revealed that the expression patterns of specific tumour biomarkers were consistent between PDOs and the corresponding tumours. For example, strong positive expression of PAX8 and P53 was detected in newly diagnosed HGSOC tissues, as well as in the organoids derived from them, and P53 and Her2 presented positive expression patterns in recurrent HGSOC tumours and PDOs (Fig. 2B). The degree of positive expression remained highly consistent in the parental tumours and the corresponding PDOs.

Fig. 2.

H&E and immunohistochemistry staining (IHC) comparing PDOs with corresponding parent tumours. A H&E staining was performed in parent tumour tissue and the derived PDOs, such as the newly diagnosed HGSOC and recurrent MOC tumour. The result showed the characteristics of parent tumours were retained in PDOs, including the large and deep stained nuclei, irregular arrangement of cancer cells, and a decreased cytoplasmic ratio. (the first row: tumours; the second row: organoids), (scale bar of the figures in the left column: 100μm; scale bar of the enlarged blue area: 50μm). B IHC of the specific tumour biomarkers in parent tumours and the corresponding PDOs, such as P53 and PAX8 in newly diagnosed HGSOC, P53 and Her2 in recurrent HGSOC. (scale bar of the figures in the left column: 100μm; scale bar of the enlarged blue area: 50μm)

PDOs recapitulate the genetic heterogeneity of the parental ovarian tumours

We analysed whether PDOs retain the genetic alterations of their corresponding tumours by performing whole-exome sequencing (WES) of three pairs of primary ovarian cancer tissues and matched organoids. Generally, it will take 2–3 weeks from tumour to PDOs. The timing of organoid genomic heterogeneity analysis is synchronized with drug screening, for better evaluating the drug response of organoids compared to the original tumour [11]. With the qualified sequencing data (Table S2), a further downstream analysis was performed. The tumour mutational burden (TMB), a measure of noninherited mutations per megabase of DNA, indirectly reflects the ability of tumour cells to produce neoantigens [37]. Figure 3A shows that the somatic TMB was not significantly different between primary tumours and the corresponding organoids (P = 0.576). The total number of somatic mutations (SNVs and InDels) in the coding sequence region and nonsynonymous mutations were similar in the matched tumour tissues and PDOs. Furthermore, we constructed a heatmap to show specific genes with a high frequency of mutation among samples (Fig. 3C). Two representative comparisons (ovarian cancer patients P4 and P5) between representative PDOs and paired primary tumour maintained similar mutation patterns (number and types) of high-frequency genes, whereas the P2 organoid maintained three-eighths of high-frequency genes detected in P2 the tumour tissue. The components of tumour tissues are more complex than those of PDOs, which have a higher purity of cancer cells [6]. Similarly, using in-house software [38], we found that the mutation patterns of cancer-predisposing genes and driver genes remained highly consistent in paired tumours and organoids (Fig. 3B). A more in-depth analysis revealed that the point mutation characteristics (Fig. 3D–E) were similar between PDOs and corresponding tumours. Most of the detected driver genes and predisposition genes remain consistent between the tumour and organoids, such as ATRX, NCOA4, NRG1, etc. Notably, PDOs could acquire new mutational fingerprints (e.g. MSH2, RUNX1, PBRM1) or lose genetic information(e.g. SVEP1, ROS1, DDX6) from the parental tumours. Briefly, all these findings suggested that the genomic heterogeneity of parental tumours was often captured well in PDOs.

Fig. 3.

Ovarian cancer organoids retain the genetic characteristics of the original tumour tissues. A Comparison of tumour mutational burden (TMB), number of total mutations in CDS region and non-synonymous mutations of SNV and InDel in patient-derived organoids and primary tumours. B The top 30 predisposing genes and driver genes in organoids and parent tumour tissues. The top bar chart represents mutation burden in each sample. C Heat-map analysis of the top 30 mutations and mutation types in organoids and the corresponding tumours. D Different point mutation characteristics in organoids and corresponding primary ovarian tumours. E Different contributions of point mutation types in matched PDOs and parental tumours

Factors influencing PDO growth

Organoids can be generated from diverse tumour samples with different clinicopathological characteristics. We converted the continuous variables to binary variables according to the cut-off values (Fig. S1B), which were determined from the ROC curves (Fig. S1A). These binary categorical variables were then used to analyze their association with the success rate of organoid establishment. A univariate analysis of all the OC organoids (Table 1) revealed similar success rates across patients of various ages (p = 0.090), surgical methods (p = 0.120), BRCA statuses (P = 0.400), and obtained cell counts (p = 0.200) and cell viability rates (p = 0.063). However, our results revealed that preoperative tumour biomarkers are likely important for organoid generation. CA153 (P < 0.001) and HE4 (P = 0.003) had negative impacts on overall organoid generation, whereas higher CA199 expression was associated with a greater likelihood of successful organoid development (P = 0.002). CA153 and CA199 were reported to exhibit high diagnostic performance individually [39, 40]. The results did not reveal any significant differences in the CEA or CA125 levels. Additionally, we found that Ki67 (%), the tumour size, FIGO stage, pathology, disease status, and tissue source were associated with successful organoid generation (P < 0.05) (Table 1). Furthermore, CA153 and HE4 levels were strongly correlated, with a Spearman correlation coefficient of 0.652. Similarly, a strong relationship was observed between the tissue source and disease status (Spearman’s correlation coefficient = -0.753) (Fig. S2). Thus, the variables CA153 levels and the tissue source with smaller P values were included in the multivariate analysis, excluding HE4 levels and the disease status. The results revealed that CA199 (OR 95% CI: 0.21 (0.07, 0.60), P = 0.005) and CA153 (OR 95% CI: 3.07 (1.22, 8.10)), P = 0.019) levels were two independent factors affecting organoid generation (Table 2).

Table 1.

Univariate analysis of the association between the characteristics of patients and organoid generation

Variables	All OC org.#				Newly diagnosed OC org.				HGSOC org.				Newly diagnosed HGSOC org.
	Success N = 88	Failure N = 65	χ²	P #	Success N = 65	Failure N = 58	χ²	P	Success N = 66	Failure N = 60	χ²	P	Success N = 46	Failure N = 53	χ²	P
Ki67 (%)			4.281	0.039			3.592	0.058			0.058	0.800			0.351	0.600
≦ 45	35(41%)	16 (25%)			27 (42%)	15 (26%)			14 (22%)	14 (23%)			9 (20%)	13 (25%)
> 45	51(59%)	49(75%)			37 (58%)	43 (74%)			51 (78%)	46 (77%)			37 (80%)	40 (75%)
Missing	2	0			1	0			1	0
CA125 (U/ml)			2.898	0.089			1.157	0.300			2.906	0.088			0.666	0.400
≦ 205	32 (36%)	15 (23%)			14 (22%)	8 (14%)			25 (38%)	14 (24%)			9 (20%)	7 (13%)
> 205	56 (64%)	49 (77%)			51 (78%)	49 (86%)			41 (62%)	45 (76%)			37 (80%)	45 (87%)
Missing	0	1			0	1			0	1			0	1
CA153 (U/ml)			12.133	< 0.001			10.231	0.001			9.847	0.002			8.005	0.005
≦ 32.5	48 (61%)	17 (30%)			31 (54%)	12 (24%)			33 (54%)	13 (25%)			19 (45%)	8 (17%)
>32.5	31 (39%)	39 (70%)			26 (46%)	38 (76%)			28 (46%)	39 (75%)			23 (55%)	38 (83%)
Missing	9	9			8	8			5	8			4	7
CA199 (U/ml)			9.366	0.002			8.487	0.004			6.290	0.012			5.504	0.019
≦ 6.79	10 (11%)	20 (32%)			6 (9.4%)	17 (30%)			10 (15%)	20 (34%)			6 (13%)	17 (33%)
> 6.79	77 (89%)	43 (68%)			58 (91%)	39 (70%)			56 (85%)	38 (66%)			40 (87%)	34 (67%)
Missing	1	2			1	2			0	2			0	2
HE4 (U/ml)			8.894	0.003			5.673	0.017			6.356	0.012			3.286	0.070
≦ 450	60 (76%)	31 (52%)			39 (70%)	25 (47%)			44 (71%)	27 (48%)			26 (62%)	21 (43%)
> 450	19 (24%)	29 (48%)			17 (30%)	28 (53%)			18 (29%)	29 (52%)			16 (38%)	28 (57%)
Missing	9	5			9	5			4	4			4	4
CEA (U/ml)			1.609	0.200			0.759	0.400			1.479	0.200			0.851	0.400
≦ 0.5	27 (32%)	26 (42%)			20 (32%)	22 (40%)			23 (35%)	26 (46%)			16 (35%)	22 (44%)
> 0.5	58 (68%)	36 (58%)			42 (68%)	33 (60%)			43 (65%)	31 (54%)			30 (65%)	28 (56%)
Missing	3	3			3	3			0	3			0	3
Age (years)			2.878	0.090			3.916	0.048				> 0.9			NA	> 0.9
≦ 45	15 (17%)	5 (7.7%)			14 (22%)	5 (8.6%)			5 (7.6%)	5 (8.3%)			5 (11%)	5 (9.4%)
> 45	73 (83%)	60 (92%)			51 (78%)	53 (91%)			61 (92%)	55 (92%)			41 (89%)	48 (91%)
Tumour size (cm)			5.998	0.014			NA	> 0.9			7.331	0.007			NA	0.700
≦ 3.75	20 (23%)	5 (7.8%)			4 (6.2%)	3 (5.3%)			18 (27%)	5 (8.5%)			4 (8.7%)	3 (5.8%)
> 3.75	68 (77%)	59 (92%)			61 (94%)	54 (95%)			48 (73%)	54 (92%)			42 (91%)	49 (94%)
Missing	0	1			0	1			0	1			0	1
Cell viability			3.449	0.063			1.461	0.200			5.780	0.016			2.565	0.110
≦ 0.90	70 (80%)	43 (67%)			49 (77%)	38 (67%)			55 (85%)	39 (66%)			36 (80%)	34 (65%)
> 0.90	17 (20%)	21 (33%)			15 (23%)	19 (33%)			10 (15%)	20 (34%)			9 (20%)	18 (35%)
Missing	1	1			1	1			1	1			1	1
Cell number (*104)			1.625	0.200			0.492	0.500			1.618	0.200			0.674	0.400
≦ 210	26 (30%)	13 (21%)			16 (25%)	11 (20%)			20 (31%)	12 (21%)			12 (27%)	10 (20%)
> 210	61 (70%)	50 (79%)			48 (75%)	45 (80%)			45 (69%)	46 (79%)			33 (73%)	41 (80%)
Missing	1	2			1	2			1	2			1	2
FIGO# stage			5.508	0.019			9.025	0.003			0.895	0.300			NA	0.300
I-II	21 (24%)	6 (9.2%)			18 (28%)	4 (6.9%)			9 (14%)	5 (8.3%)			6 (13%)	3 (5.7%)
III-IV	67 (76%)	59 (91%)			47 (72%)	54 (93%)			57 (86%)	55 (92%)			40 (87%)	50 (94%)
Disease status			5.601	0.018							6.483	0.011
Newly diagnosed	65 (74%)	58 (89%)							46 (70%)	53 (88%)
Recurrence	23 (26%)	7 (11%)							20 (30%)	7 (12%)
Pathology			7.706	0.006			8.290	0.004
Non-HGSOC#	22 (25%)	5 (7.7%)			19 (29%)	5 (8.6%)
HGSOC	66 (75%)	60 (92%)			46 (71%)	53 (91%)
Surgery			2.375	0.120			0.743	0.400			0.620	0.400			0.000	> 0.9
Cytoreductive surgery	62 (70%)	38 (58%)			43 (66%)	34 (59%)			43 (65%)	35 (58%)			27 (59%)	31 (58%)
Laparoscopic biopsy	26 (30%)	27 (42%)			22 (34%)	24 (41%)			23 (35%)	25 (42%)			19 (41%)	22 (42%)
BRCA status			0.793	0.400			0.929	0.300			0.081	0.800			0.114	0.700
Negative	40 (61%)	30 (53%)			29 (62%)	26 (52%)			29 (53%)	27 (50%)			20 (53%)	23 (49%)
Positive	26 (39%)	27 (47%)			18 (38%)	24 (48%)			26 (47%)	27 (50%)			18 (47%)	24 (51%)
Missing	22	8			18	8			11	6			8	6
Tissue source			6.993	0.008			2.189	0.140			9.510	0.002			3.118	0.077
Metastasis	32 (36%)	11 (17%)			10 (15%)	4 (6.9%)			29 (44%)	11 (18%)			9 (20%)	4 (7.5%)
Adnexa	56 (64%)	54 (83%)			55 (85%)	54 (93%)			37 (56%)	49 (82%)			37 (80%)	49 (92%)

This table presented the univariate analysis of all organoids and three subgroups, including newly diagnosed ovarian tumours, HGSOC, and newly diagnosed HGSOC. The results showed different variables impacted the generation of PDOs in all organoids and subgroups

org^# Organoids, P^# Pearson’s Chi−squared test, FIGO^# International Federation of Gynecology and Obstetrics, HGSOC^# High grade serous ovarian cancer

The bold P values (<0.05) indicated the variables were the influencing factors for organoid generation in each group in univariate analysis

Table 2.

Multivariate analysis of the characteristics of ovarian tumour patients related to organoid generation

Variables	All OC org#.		Newly diagnosed OC org.		HGSOC org.		Newly diagnosed HGSOC org.
	OR# (95% CI#)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Age (years)				0.130
=<45			1.00 (Ref)
> 45			2.89 (0.77, 12.6)
Ki 67 (%)		0.072
≤ 45%	1.00 (Ref)
> 45%	2.58 (0.93, 7.52)
CA153 (U/ml)		0.019		0.009		0.002		0.002
≤ 32.5	1.00 (Ref)		1.00 (Ref)		1.00 (Ref)		1.00 (Ref)
> 32.5	3.07 (1.22, 8.10)		3.75 (1.44, 10.6)		4.10 (1.70, 10.7)		5.72 (2.00, 19.3)
CA199 (U/ml)		0.005		0.017		0.006		0.013
≤ 6.79	1.00 (Ref)		1.00 (Ref)		1.00 (Ref)		1.00 (Ref)
> 6.79	0.21 (0.07, 0.60)		0.23 (0.06, 0.72)		0.19 (0.05, 0.58)		0.21 (0.05, 0.67)
Tumour size (cm)		0.100				0.600
≤ 3.75	1.00 (Ref)				1.00 (Ref)
> 3.75	3.00 (0.85, 11.9)				1.60 (0.67, 4.43)
FIGO# Stage		0.150		0.200
I-II	1.00 (Ref)		1.00 (Ref)
III-IV	2.44 (0.74, 8.99)		2.43 (0.62, 10.9)
Tissue source		0.300				0.400
Metastasis	1.00 (Ref)				1.00 (Ref)
Adnexa	1.83 (0.63, 5.55)				2.07 (0.42, 12.7)
Pathology		0.600		0.800
Non-HGSOC#	1.00 (Ref)		1.00 (Ref)
HGSOC	0.67 (0.13, 3.69)		0.82(0.18, 3.99)
Cell Viability						0.300
≤ 0.90					1.00 (Ref)
> 0.90					1.70 (0.67, 4.43)

This table showed multivariate analysis in all organoids and three subgroups, including newly diagnosed ovarian tumours, HGSOC, and newly diagnosed HGSOC

org^# organoids, OR# Odd ratio, 95% CI^# 95% Confidence Interval, FIGO^# International Federation of Gynecology and Obstetrics, HGSOC^# High grade serous ovarian cancer

The bold P values (<0.05) indicated the variables (CA153, CA199) were independent factors of the generation of PDOs

Additionally, we evaluated these variables in relation to organoid establishment in three subgroups, namely, primary ovarian cancer, HGSOC, and primary HGSOC organoids. The univariate analysis revealed that CA153 levels (P = 0.001), CA199 levels (P = 0.004), HE4 levels (P = 0.017), age (P = 0.048), FIGO stage (P = 0.003) and pathology (P = 0.004) were correlated with the successful generation of primary OC organoids (Table 1). In the HGSOC organoids, CA153 levels (P = 0.002), CA199 levels (P = 0.012), HE4 levels (P = 0.012), tumour size (P = 0.007), disease status (P = 0.011), tissue source (P = 0.002) and cell viability (P = 0.016) played significant roles in the univariate analysis (Table 1). According to the multivariate analysis (Table 2), CA153 levels had a negative impact on primary ovarian tumour and HGSOC organoids (P = 0.009 and P = 0.002, respectively), whereas CA199 levels had a positive impact on the generation of primary ovarian tumours and HGSOC organoids (P = 0.017 and P = 0.006, respectively). Furthermore, univariate and multivariate analyses indicated that CA153 and CA199 levels also affected the generation of primary HGSOC organoids.

It has been revealed that serum tumour markers CA153 and CA199 are linked to ovarian cancer prognosis [40]. In summary, CA153 and CA199 levels play highly significant roles in independently influencing the generation of all ovarian tumour organoids.

Use of organoid drug screening to predict clinical efficacy

Twenty organoids for drug screening were derived from 19 ovarian cancer patients who were treated with a combination of paclitaxel and platinum, paclitaxel and carboplatin (TC regimen) were commonly used in clinical practice. The clinical efficacy of the combined therapeutic regimens in 19 patients was 78.95% (Tables S3–5). We tested the sensitivity of these patient-derived organoids to carboplatin and paclitaxel. Figure 4A shows that paclitaxel and carboplatin had cytotoxic effects on PDOs at various drug concentrations. PDOs with an IC50 lower than the Cmax of paclitaxel (4.1 µM) [41] and carboplatin (87 µM) [42] are considered sensitive to the drug. Since the AUC is highly correlated with the IC50 (Fig. 4F), the AUC can assist in determining drug sensitivity when the IC50 cannot be calculated. The AUCs of the four PDOs (blue in Fig. 4B) were lower than the AUC_cut−off value (0.69), which indicated that they were sensitive to paclitaxel. Additionally, the graph of the viability of the PDOs (Fig. 4B) clearly showed that the organoids exhibited significantly greater sensitivity to paclitaxel than to carboplatin. The concordance rates (red and black) of drug screening and clinical treatments with paclitaxel and platinum were 75% and 90%, respectively (Fig. 4C).

Fig. 4.

Ovarian cancer organoids show the patient-specific response to paclitaxel and carboplatin. A Phase-contrast pictures of PDOs illustrated that the survival viability increased as the concentration of paclitaxel decreased and that of carboplatin decreased as well. B Drug response curves of PDOs treated with paclitaxel and carboplatin, individually. When the IC50 (half - maximal inhibitory concentration) is greater than the Cmax (maximum plasma concentration), the PDOs are considered to be resistant to paclitaxel or carboplatin. (Res, resistant; green and blue, Sensitive). C For all PDOs with treatment of paclitaxel can carboplatin, individual scatterplot shows IC50 values identified by drug screening of the matched patients, with clinical responses indicated on the right for comparing the consistency. (CR, complete response; PR, partial response; PD, progressive disease; SD, stable disease). D Drug response curves of four PDOs, which were treated with carboplatin alone or carboplatin added with 1 uM paclitaxel, show that carboplatin combined with paclitaxel elicits a better therapeutic response. E Drug response curves of a patient-derived organoid treated with carboplatin combined with different doses of paclitaxel, indicated that the combined therapy of paclitaxel and carboplatin exhibits a better drug response. F Scatterplots indicating significant positive correlation between AUC and IC50 values for paclitaxel and carboplatin

However, considering that patients clinically receive TC combination therapy, we jointly analysed the results of drug screening for paclitaxel and carboplatin. It has been reported in the literature that the drugs in a combination regimen may target different tumour subclones or exert synergistic effects. Even if the drugs are antagonistic, or a particular drug has weak efficacy on its own, their combined use can still yield clinical benefit [43, 44]. If PDOs are sensitive to paclitaxel or carboplatin in drug tests, TC combination therapy is considered effective for patients. In total, the drug screening results of four PDOs (Org-124, Org-137, Org-222, and Org-225) did not match the clinical response of patients to the combined therapy, with a predicted efficacy of 80%. Org-124 and Org-137 were sensitive to paclitaxel but not carboplatin, which suggested that these two patients may be sensitive to TC therapy. However, new lesions appeared in patient N124 within six months after TC adjuvant treatment, and the CA125 level decreased to the reference value (16 U/mL). Patient N137 experienced disease progression during TC adjuvant treatment. Org-222 was resistant to paclitaxel and carboplatin. However, patient N222 achieved partial remission after TC therapy, with a decrease in CA125 levels and a reduction in the lesion clinically. Org-225 was sensitive to paclitaxel but not carboplatin, while patient N225 did not respond to TC therapy, with an increase in the number of tumour lesions. Furthermore, we tested the efficacy of carboplatin alone and carboplatin combined with paclitaxel in the same PDO. The results showed that paclitaxel could decrease the IC50 of carboplatin and increase the sensitivity to carboplatin (Fig. 4D). As the dose of paclitaxel increased, the sensitivity of combination therapy increased (Fig. 4E). Therefore, the combined analysis of the response to paclitaxel and carboplatin often can predict patients’ clinical response, including neoadjuvant chemotherapy and postoperative adjuvant therapy.

PDO pharmacophenotyping accurately mirrors the prior treatment outcomes of the respective patients

In addition, we retrospectively collected data on the therapeutic regimens and treatment outcomes of 7 ovarian cancer patients in a blinded fashion by two professional gynecologists. When opinions differ, a third physician makes the judgment. The PDOs were able to predict the prior clinical responses of these patients with an efficiency rate of 100%. Three patients (N248, N263, and N264) received paclitaxel and carboplatin as neoadjuvant chemotherapy (Table S4). Prior to interval cytoreductive surgery, the patients achieved a partial response, as evaluated according to the RECIST 1.1 criteria. Moreover, they achieved complete remission 6 months after adjuvant chemotherapy and were regarded as platinum-sensitive patients (Table S5). Consistently, their PDOs also responded to carboplatin or paclitaxel (IC50 < Cmax). Five patients whose PDOs were sensitive to both paclitaxel and carboplatin were diagnosed with platinum-sensitive recurrent ovarian cancer. The N225 PDO was sensitive to paclitaxel but not carboplatin, and it was concordant with the retrospective clinical response. The N222 PDO was resistant to paclitaxel and carboplatin in the drug screening test, which is inconsistent with the prospective and retrospective clinical responses. Overall, our PDO chemosensitivity profile largely paralleled the retrospective clinical data from the corresponding patients.

Discussion

Ovarian cancer and its microenvironment exhibit a high degree of heterogeneity [45, 46]. Many factors can affect the construction of ovarian cancer organoids, making culture much more difficult than that of colorectal cancer. Despite a drug resistance rate of 70% in ovarian cancer, current research on treatment efficacy predictions for these patients remains insufficient [47]. Therefore, guiding subsequent precise clinical treatment through organoid-based drug screening is particularly important.

In total, we constructed 153 ovarian cancer organoids derived from 153 patients, with an overall success rate of 57.52%. The culture efficiency of newly diagnosed OC was greater than that of recurrent OC (76.67%>52.85%). Hans et al. established 56 organoids from 32 different patients, with a success rate of 65%. Among them, the construction success rate of HGSOC organoids was 55% [19]. In our study, the success rate of constructing HGSOC organoids was 52.38%, which was close to the result reported by Hans et al. The difference in the proportions of HGSOC and non-HGSOC patients (82.35% vs. 33.93%) might be the main reason for the discrepancy in the overall success rates. Hugo et al. constructed PDOs mainly for HGSOC, with an overall construction efficiency of 44%. The culture success rate for HGSOC specifically was 36% [7]. These results are consistent with the results of the univariate analysis of organoid construction in our study, which revealed that the pathology significantly influences the efficiency of PDO construction. However, the multivariate analysis indicated that only CA153 and CA199 levels are independent factors influencing the generation of ovarian cancer organoids. Confounding factors or mediating variables, such as tissue resource, FIGO stage and cell viability, may influence the pathological type. The sample size must be expanded for further verification. In addition, Hans et al. and Wojciec et al. explored the composition of the culture media and discovered that it plays a crucial and indispensable role in the construction of PDOs [11, 48]. In our study, we built upon Kopper’s culture medium and modified it to subsequently generate ovarian cancer organoids(Table S1) [11].

In our study, we found that the corresponding organoids were derived ovarian cancer patients with low CA153 levels and high CA199 levels with increased efficiency. CA153 is the soluble form of MUC1. An oncogenic C-terminal transmembrane subunit (MUC1-C), encoded by MUC1, induces changes in chromatin accessibility and gene expression in association with driving cancer stem cell (CSC) state, self-renewal capacity and tumourigenicity [49, 50]. MUC1 could facilitate the symmetric division of tumour spheroidal cells and maintain the stemness of cancer stem-like cells [51]. However, Shengze Li et al. found that CD133^High/MUC1^High cells could differentiate into CD133^High/MUC1^Low cells after cultivation in cell culture medium, whereas the CD133^High/MUC1^Low cells only produced CD133^Low/MUC1^Low cells [51].

Sialyl Lewis a is the antigenic epitope of tumour marker CA19-9. E-selectin ligands sialyl-Lewis A and X (sLeA/X) play vital roles in the E-selectin binding and endothelial adhesion, facilitating tumour cell proliferation (2D) or 3D colony formation [52]. The data from Natoni et al. strongly suggested that the presence of SLea/x is essential for proper P-selectin binding, leading the cells to displaying a robust adhesion [53]. Factors influencing cell proliferation, stemness, and adhesion capabilities are closely related to the formation of organoids. In summary, we speculate that CA153 and CA199 may be important factors of organoid development through the above mechanisms, but further experimental validation is still required. Noteworthy, this result may suggest that the tumour tissues should be classified by CA153 and CA199 level and that stratified construction with diverse culture media should be performed. Our findings may open an avenue for researchers, enabling them to explore diverse culture systems, increase the construction efficiency, and promote the application of organoid drug screening platforms.

In our study, we illustrated the ability of PDOs as a drug screening platform to retrospectively analyse and prospectively predict the clinical response. Different PDOs exhibit distinct heterogeneity in drug responses. As avatars for drug testing, PDOs can be used to screen personalized therapeutic strategies for ovarian cancer patients. Chen et al. conducted drug sensitivity tests using breast cancer organoids and documented a significant correlation between the drug response and previous clinical treatment response, which is consistent with the results of our drug screen [54]. We examined the drug response of recurrent ovarian cancer organoids to paclitaxel and carboplatin. The majority of these organoids were sensitive to TC treatment, which is consistent with the previous clinical platinum-sensitive recurrence status of the parental tumours. Furthermore, the organoids, which were generated from primary tissues collected from interval cytoreduction surgeries, had a significant response to TC, consistent with the efficacy of the neoadjuvant chemotherapy. The corresponding patients achieved partial remission.

We also generated PDOs derived from patients with newly diagnosed and platinum-sensitive recurrent ovarian cancer. The first-line treatment for these patients is platinum-based regimens (paclitaxel + carboplatin). These parameters were evaluated by performing a drug sensitivity test. The results showed that most cases were consistent with the patients’ prospective treatment efficacy. The CA125 level of patient N124 decreased to within the normal range after completing TC chemotherapy. However, the disease recurred with platinum resistance four months later. During this period, the tumour cells possibly underwent remodelling, gradually reducing their response to TC treatment and eventually leading to the development of drug resistance [55–57]. Patients N137 and N225 experienced disease progression after 2–3 cycles of TC treatment. Notably, Org-124, Org-137, and Org-225 were sensitive to paclitaxel but resistant to carboplatin. Based on the evaluation method proposed by Helen et al. and Chen et al. [54, 58], we considered these patients to be sensitive to TC combination therapy. Org-222 was resistant to paclitaxel and carboplatin, but the corresponding patient achieved partial remission after TC treatment in the clinic. The discrepancy between the pharmacophenotyping results and the clinical response to TC therapy could be attributed to the complex tumour ecosystem and the high heterogeneity of cancer cells [13]. Although there is a similar pattern of genetic mutations between the tumour tissue and the derived organoids, but some of the high-frenqucy genes are inconsistent between the two. Intertumour and intratumour heterogeneity may contribute to this discrepancy. Tumours are ecosystems in which cancer and noncancer cells interact and evolve in complex and dynamic ways. The tumour microenvironment remodels ovarian cancer cells, resulting in differences in the drug responses of PDOs originating from primary or metastatic sites [59, 60]. The co-culture system by co-culturing ovarian cancer cells with immune cells or stromal cells can more accurately reflect the drug response of tumour cells [61]. A 3D co-culture model of human ovarian cancer organoids with cancer-associated fibroblasts (CAFs) suggested that CAFs may mediate organoid growth and promote resistance to paclitaxel and cisplatin [62].

Additionally, we found some differential genes in parental tumour tissues and the derived organoids in the sequence-based heterogeneity analysis. Two organoids exhibit loss of SVEP1 and TP53, while two others show newly acquired RUNX1 alterations. RUNX1 is closely associated with chemoresistance. In EOC cells, RUNX1 is reported as a novel regulator of IGF1R during acquirement of chemoresistance [63]. RUNX1 functioned as a transcription factor to drive expression of PDGFB, leading to resistance to palbociclib by enhancing the Akt pathway and suppressing senescence [64]. It’s reported that PDOs could capture tumour clonal evolution, which varies across individuals [20]. Organoids may acquire chemoresistance when they newly evolve RUNX1. TP53 mutations in tumour tissues [65, 66] and SVEP1 in CAFs [67] may mediate chemoresistance in ovarian cancer, while the loss of them in organoids may increase the chemosensitivity. It is possible that a TP53-wildtype cancer stem-like cell population, which was a minor component in vivo, was preferentially expanded under our specific 3D culture conditions. The small number of sequencing cases in the article is a limitation, and more samples can be added for validation.

Our study illustrated the feasibility of the use of ovarian cancer organoids as a drug screening platform. Moreover, we analysed the factors influencing the construction of PDOs, improving the efficiency of PDO generation. We will focus on ovarian cancer patients with multidrug resistance later for the discovery and preclinical testing of novel therapeutic strategies. Due to the lack of information on the immune microenvironment, vascularization, and endocrine regulatory systems in vitro, agents that are catalysed by liver enzymes, target tumour angiogenesis or regulate hormone levels cannot be evaluated in PDOs [54]. However, organoid chips are being investigated to develop a coculture system of organoids with immune cells and vascular epithelial cells [68]. This advancement will significantly facilitate the application of organoids in precision medicine.

Conclusions

We generated an ovarian cancer organoid biobank from the tumour tissues of patients with a broad spectrum of treatment statuses. CA153 and CA199 levels were independent factors associated with PDO generation. The drug responses of organoids from these patients to paclitaxel and carboplatin are, to a certain extent, consistent with their clinical treatment responses. In the future, organoids can serve as a preclinical drug screening platform to guide the clinical treatment of ovarian cancer patients. Ovarian cancer organoids, as 3D experimental models in vitro, provide a personalized drug screening platform for ovarian cancer patients and promote the development of precision oncology.

Abbreviations

BCLC

Barcelona Clinic Liver Cancer

BWA

Burrows Wheeler Aligner

CAFs

Cancer-associated fibroblasts

CNAs

Copy number alterations

CA125

Cancer antigen 125

95% CIs

95% confidence intervals

CR

Complete response

EOC

Ovarian endometrioid carcinoma

HCC

Hepatocellular carcinoma

HGSOC

High-grade serous ovarian cancer

H&E

Haematoxylin and eosin

IHC

Immunohistochemistry

LGSOC

Low-grade serous ovarian cancer

MOC

Ovarian mucinous carcinoma

MVI

Microvascular invasion

OC

Ovarian cancer

ORs

Odds ratios

PDO

Patient-derived organoid

PR

Partial response

PD

Progressive disease

SD

Stable disease

SMBOT

Serous / mucinous borderline ovarian tumour

TMB

Tumour mutational burden

Authors' contributions

LW and XPZ collected the primary tumour tissues and generated PDOs. LW performed the drug screening of PDOs. QXJ and MSH collected the clinicopathological data of patients. LFM and LZ prepared the figures. QLL and HJW preformed the statistics analysis and made the tables. Finally, LW originally drafted the manuscript. DLZ and HXW initiated the study and finalized the manuscript. All authors read and approved the final manuscript.

Funding

This work was funded by Chongqing Science and Technology Bureau (Grant No. cstc2022jxjl20039), Talent Program of Chongqing (Grant No. cstc2024rcih-bgzxm0162, YXGD202403), Chongqing Health Commission (Grant No. 2023ZDXM029 and 2023MSXM043), Scientific and Technological Research Program of Chongqing Municipal Education Commission (Grant No. KJQN20230013), the Project for Enhancing Scientific Research Capabilities of Chongqing University Cancer Hospital (Grant No. 2023nlts009), and Beijing Health Alliance Charitable Foundation (Grant No. BJHA-CRP-089), Chongqing Shapingba District Health Commission (Grant No. 2023SQKWLH021), and Wu Jieping Medical Foundation (Grant No.320.6750.2022-22-12).

Data availability

The data generated in this study are available within the article and its supplementary data files.

Declarations

Ethics approval and consent to participate

The study was performed in accordance with the Declaration of Helsinki and had patients’ informed consent. On November 28, 2023, the Ethics Committee of Chongqing University Cancer Hospital approved the study titled “Exploratory Study on the Guidance of Organoid Drug Sensitivity Testing for the Precise Treatment of Recurrent Ovarian Cancer” (Approval No.: CZLS2020274 - A). Every human participant provide their consent in the study. We ensured the protection of their privacy.

Consent for publication

Consent for publication has been obtained from all participants, and we have taken measures to protect their anonymity.

Competing interests

The authors declare no competing interests.