Thirty-eight patients with gynecological cancer were enrolled during 2020–2022. Tumors from 31 patients were successfully generated into organoids. Among them, 28 PDOs can be expanded, including 15 with ovarian cancer, 11 with endometrial cancer, and 2 with cervical cancer. Among the ovarian cancer PDOs, four were high-grade serous carcinoma (HGSC), four were mucinous, three were clear cells, one was carcinosarcoma, and one was endometrioid (Table 1).

Different pathologic types presented with different appearances (Fig. 1A). HGSC organoids displayed papillary branching, with a glandular-like protruding contour. Endometrioid organoids showed a glandular pattern with smaller cells and more confluent surface. Mucinous cancer organoids showed vanished glandular architecture and a simple, non-stratified cell lining outside with mucin-like content. Clear cell cancer organoids had polyhedral, flattened cells with vacuolated cytoplasm.

OV-PDOs show interpatient morphology differences. A Phase-contrast of PDOs under culture with matrix gel and appropriate growth factors. B H&E staining of PDOs shows epithelial invaginations and folding as well as a round, cystic phenotype with lumen formation

As shown in Fig. 1A, two HGSC organoids (HGSC-4 and HGSC-5) from different patients displayed different cell densities and arrangements, suggesting its interpatient heterogeneity. Interestingly, within a single patient, organoid sizes, cellular densities, and protruding contours were distinct; these morphologic differences thus also indicate intrapatient heterogeneity (HGSC-4, HGSC-5, and HGSC-7, Supplementary Figure S1).

With hematoxylin and eosin (H&E) staining, mucinous organoid cells were spheroid with the cell apical at inner side and basal at outer side. The organoid cells had a high nuclear-cytoplasmic ratio, and nuclear polymorphism (Fig. 1B). High-power field showed that organoid cells had vacuolated and secretions. We further assessed organoid biomolecular preservation with H&E and immunohistochemical staining of organoids, parent tumors, and Human Protein Atlas (https://www.proteinatlas.org/) references.

The characteristics of organoids and parental tumors were compared. We presented case HGSC-5, which is a p53 null-expression patient. As shown in Fig. 2A, the organoids displayed a high nuclear-cytoplasmic ratio, irregular nuclei, and prominent nucleoli, recapitulating all malignancy features. p53 immunostain showed null expression in HGSC PDOs and was matched to the parent tumors. In addition, the strong positive PAX8 and weak positive WT1 were compatible with parent tumors and a p53 null-expression case in the Human Protein Atlas. In primary mucinous ovarian cancer, the immunohistochemistry expression of PAX8 is usually low [15]. Mucinous carcinoma PDOs presented with diffuse CK7 positive, focal PAX8 positive, and CK20 negative, matching the parent tumors and reference data (Fig. 2B). This cumulative evidence indicates the successful establishment of ovarian cancer-derived organoids, of which the morphology and molecular characteristics were consistent with parent tissues.

Patient-derived organoids morphologically and molecularly matched the parent tumors. H&E stain and immunohistochemistry of PDOs in HGSC (HGSC-5, A) and mucinous ovarian cancer (MC-4, B) in paired tumor (upper), OV-PDOs (middle), and database (lower)

We tested drug responses in 10 OV-PDOs using guideline-recommended chemotherapeutic agents, including paclitaxel, cisplatin, carboplatin, epirubicin, doxorubicin, gemcitabine, topotecan, and olaparib. The drug sensitivity analysis is presented in Fig. 3 and estimated IC50 values are shown in Supplementary Figure S2. The areas under the dose–response curves for each drug were also calculated (Fig. 4). Each PDO had a unique dose–response curve for each drug, indicating interpatient heterogeneity even within the same histology type. A lower area under the curve (AUC) indicates a more sensitive response, as demonstrated in Fig. 4 (smaller circles indicate better drug choices).

Drug testing and personalized therapy of ovarian cancer in OV-PDOs. Dose–response curves of 10 OV-PDOs treated with cisplatin, carboplatin, paclitaxel, gemcitabine, epirubicin, doxorubicin, topotecan, and olaparib. Dots represent five-repetition means. Error bars represent five-repetition standard error of the mean. The statistical analysis of drug response at 0.1 µM was calculated using the chi-square test

Area under the drug response curve values mapped to the balloon plot. AUC for a fixed concentration range. Circle color and size indicates AUC results. AUC can be seen as average efficacy and compared across patients

The HGSC-4 OV-PDO was relatively sensitive to epirubicin (AUC = 0.496) and topotecan (AUC = 0.550). The HGSC-6 OV-PDO was relatively sensitive to paclitaxel (AUC = 0.507), gemcitabine (AUC = 0.382), and topotecan (AUC = 0.461). The HGSC-5 OV-PDO responded poorly to all drugs (AUC = 0.691–0.831). For the clear cell ovarian cancer (CCC) OV-PDOs, three were more sensitive to topotecan (AUC = 0.425–0.570) and CCC-3 was more sensitive to gemcitabine (AUC = 0.335). Between the two mucinous OV-PDOs, MC-5 was more sensitive to gemcitabine (AUC = 0.239) and topotecan (AUC = 0.253). These results demonstrate an OV-PDOs-guided precision therapy approach.

The clinical courses of four patients, who included platinum-sensitive, resistant, and refractory profiles, were correlated with their OV-PDOs drug response results (Fig. 5).

Drug sensitivity compatibility between PDOs and clinical data. Summary timeline of the platinum-sensitive (HGSC-4, A), resistant (EM-2, B), and refractory (HGSC-5, C; CCC-2 D) treatment plans. The reference range of CA 125 is 0–35 units/mL. Circle with straight lines indicates the time of sample collection

Patient HGSC-4 had stage IIIC platinum-sensitive HGSC and had undergone primary suboptimal debulking surgery (Fig. 5A). After surgery, she received adjuvant chemotherapy with paclitaxel and carboplatin, based on guideline recommendations. The tumor markers decreased and the clinical image revealed partial tumor response after six cycles. Because of intolerable neurological toxicity, especially hand numbness, the regiment was changed to liposomal doxorubicin (Lipodox). After six cycles of Lipodox, she was tumor-free. There was no tumor recurrence during the following 18 months. In vitro drug testing revealed that this patient’s OV-PDOs (Fig. 4, HGSC-4) were sensitive to taxanes (paclitaxel, AUC = 0.546), platinum (carboplatin, AUC = 0.565), and anthracycline (epirubicin, AUC = 0.496), which is compatible with her clinical course.

Patient EM-2 had platinum-resistant, stage IIB, dedifferentiated endometrioid ovarian cancer. She had a suboptimal debulking surgery following adjuvant chemotherapy, with six cycles of paclitaxel and carboplatin to achieve complete remission (Fig. 5B). After four months, tumor recurrence at the presacral area was found. She had chemotherapy with three cycles of Lipodox, but the tumor progressed. She then underwent chemotherapy with gemcitabine and cisplatin, and a second, optimal debulking operation. There was no evidence of disease for 16 months (at the time of manuscript preparation). Her OV-PDOs (Fig. 4, EM-2), derived at the second debulking surgery, showed resistance to paclitaxel (AUC = 0.802), cisplatin (AUC = 0.846) and Lipodox (doxorubicin, AUC = 0.820) and relative sensitivity to gemcitabine (AUC = 0.692) and topotecan (AUC = 0.578). Thus, topotecan may be a better drug choice in the event of future recurrence.

Patient HGSC-5 had platinum-refractory, stage IIIC HGSC. She received four cycles of neoadjuvant chemotherapy with paclitaxel and carboplatin, followed by optimal interval debulking (Fig. 5C). Tumor progression developed after two further cycles of paclitaxel and carboplatin. Chemotherapy was then shifted to Lipodox. Tumor progression occurred again after three cycles of Lipodox, at which time gemcitabine was administered. The tumor still progressed after three cycles. The patient began palliative care and expired a few months later. Her OV-PDO drug tests (Fig. 4, HGSC-5) revealed multiple drug resistances to these chemotherapeutic agents (paclitaxel, AUC = 0.765; carboplatin, AUC = 0.766; doxorubicin, AUC = 0.831; gemcitabine, AUC = 0.728).

Patient CCC-2 had platinum-refractory, stage IIIC CCC. Metastatic para-aortic lymph node was found by the general surgeon, and she received three cycles of neoadjuvant treatment with bevacizumab, paclitaxel, and cisplatin at the gynecologic department (Fig. 5D). Because of proteinuria and nephrotoxicity, she had paclitaxel and carboplatin for two more cycles. Imaging showed a growing para-aortic tumor; therefore, she underwent interval suboptimal debulking surgery. She received adjuvant chemotherapy with gemcitabine and carboplatin. However, she was found to have jaundice before the next chemotherapy cycles, and imaging revealed rapid tumor growth. Family counseling led to the decision to take palliative care for the rest of her life, and she expired two weeks later. Her OV-PDOs (Fig. 4, CCC-2), derived at the interval debulking surgery, presented relative resistance to paclitaxel (AUC = 0.754), carboplatin (AUC = 0.610), and gemcitabine (AUC = 0.718).