Long-term patient-derived ovarian cancer organoids closely recapitulate tumor of origin and clinical response

Abstract

Background: Ovarian cancers are the second cause of death from gynecological cancers worldwide, due to a late diagnosis combined with the development of resistance to chemotherapy. However, half of these cancers present alterations in Homologous Recombination (HR), making them sensitive to inhibitors of the PARP protein (PARPi), involved in DNA repair. Nevertheless, identifying patients who respond to chemotherapy and selecting those eligible for PARPi remains a challenge for clinicians. In this context, the use of Patient-Derived Tumor Organoids (PDTO) for predictive functional testing represents an interesting prospect for clinical decision making.

Methods: Here we established a panel of 37 long-term PDTO models of various histological subtypes from 31 ovarian cancer patients. Histological and molecular profiles of PDTO were compared to tumor sample of origin using immunohistochemical analyses and global approaches (copy number variation and transcriptomic profiling). PDTO models were exposed to standard drugs for ovarian cancer patients, including PARPi, and response was assessed using viability assay. To further define the HR status of PDTO, we performed a functional assay evaluating the ability of PDTO to initiate HR (RECAP test) using automated histo-imaging quantitative analysis of RAD51 foci, as well as an NGS analysis based on the sequencing of an HR-related genes panel to obtain a Genome Instability Score (GIS).

Results: We demonstrated that PDTO mimicked histological and expression of tumor markers of paired tumors. Moreover, non-negative matrix factorization approach revealed that PDTO recapitulated the transcriptomic profile of the cancer component from their sample of origin. Screening of chemotherapeutic drugs showed that PDTO exhibit heterogeneous responses, and that response of PDTO from high-grade serous ovarian carcinoma to carboplatin recapitulated patient response to first-line treatment. Additionally, the detection of HRD phenotype of PDTO using functional assay was associated with the results of the HRD test Genomic Instability Scar (GIScar).

Conclusion: Although larger-scale investigations are needed to confirm the predictive potential of PDTO, these results provide further evidence of the potential interest of ovarian PDTO for functional precision medicine.

Keywords: Functional assay; Homologous recombination; Ovarian cancer; Patient-derived tumor organoids.

Fig. 1

Establishment and characterization of a PDTO panel from ovarian tumors. (A) Pie-chart showing the histological subtype, sampling time, sample origin and FIGO stage of the 31 patients from which PDTO lines were established. (B) Pie-chart showing the sample origin of the 37 ovarian PDTO lines. (C) Representative pictures of PDTO of various morphologies. A = Ascites-derived PDTO and T = Tumor–derived PDTO. Scale bar = 100 μm. (D) Hematoxylin and Eosin (HE) staining and immunohistochemistry analysis of ovarian cancer protein markers ki67, PAX8, p53 and HNF1β in tumor tissues and paired PDTO. Scale bar = 200 μm

Fig. 2

Ovarian PDTO lines capture genomic and transcriptomic features of parental tumor. (A) Genome-wide heatmap of DNA copy number gains (blue) and losses (red) of tumor tissues and paired PDTO. A = Ascites-derived PDTO and T = Tumor–derived PDTO. (B) Distribution of tumor samples and paired PDTO on UMAP clusters based on gene expression profile following NMF (k = 12). * PDTO that do not have a matched tumor sample counterpart

Fig. 3

Ovarian PDTO lines recapitulate patient response to carboplatin. (A) Flow chart of the PDTO selection for the comparison with patient clinical outcome. (B) Dose-response curves of the 9 selected PDTO, each curve is representative of at least two independent experiments. A = Ascites-derived PDTO and T = Tumor–derived PDTO. (C) Summary of the IC50 (µM) and AUC obtained in B. for the 9 PDTO models. (D) Representation of the carboplatin normalized AUC z-score (n = 9). (E) Comparison of the PDTO response to carboplatin and clinical outcome, expressed as complete surgery, platinum-free interval (PFI), progression-free survival (PFS), CA-125 normalization and KELIM score. (F) Dot plots comparing PDTO response to carboplatin, expressed as normalized AUC z-score, and clinical outcome expressed as PFI. Data were analyzed using unpaired two-sided Mann–Whitney test. (G) Kaplan-Meier plot comparing PFI of the resistant PDTO group (n = 5) and the sensitive PDTO group (n = 4). Data were analyzed using Gehan-Breslow-Wilcoxon test

Fig. 4

Drug screening of PDTO standard-of-care therapies in ovarian cancer. (A) Heatmap showing PDTO response to carboplatin, paclitaxel, olaparib, niraparib, doxorubicin and gemcitabine, expressed as normalized AUC z-score (n = 29 models). A = Ascites-derived PDTO and T = Tumor–derived PDTO. (B) Heatmap showing response of PDTO models derived from the same patient (n = 2 patients). (C) Heatmap showing response of PDTO models derived from recurrent ovarian cancer (n = 4 patients)

Fig. 5

PDTO can be used to assess homologous recombination deficiency. (A) Correlation between the carboplatin and olaparib PDTO response, expressed as normalized AUC z-score (n = 29). (B) Dot plots comparing PDTO response to olaparib, expressed as normalized AUC z-score, and the genomic instability score HR status (n = 29). Data were analyzed using unpaired two-sided Mann–Whitney test. (C) Schematic representation of the RECAP test protocol. (D) DAPI staining to visualize nuclei (blue) and immunostaining of cyclinA2 (green) and RAD51 (red) in BRCA1/2-wt (OV-174_T) and BRCA2-mut (OV-156_A) PDTO models. (E) Principal Component Analysis (PCA) of RECAP test results (n = 23). A = Ascites-derived PDTO and T = Tumor–derived PDTO. (F) Dot plots comparing PDTO RECAP test PC1 and the genomic instability score HR status (n = 23). Data were analyzed using unpaired two-sided Mann–Whitney test. (G) Kaplan-Meier plot comparing PFI of the RECAP HRD group (n = 8) and the RECAP HRP group (n = 13). (H) Kaplan-Meier plot comparing PFI of the GIScar HRD group (n = 6) and the GIScar HRP group (n = 15). (I) Median PFI of OC patients from PDTO were derived depending on the HR status determined by GIScar analysis and RECAP test