High-Throughput Drug Screening of Clear Cell Ovarian Cancer Organoids Reveals Vulnerability to Proteasome Inhibitors and Dinaciclib and Identifies AGR2 as a Therapeutic Target

Abstract

There are currently no effective treatments available for clear cell ovarian cancer (CCC). In this study, we aimed to identify effective drugs for CCC through high-throughput drug screening (HTDS) using ovarian cancer organoids and determine novel therapeutic targets based on the biological characteristics of CCC through omics analysis. An ovarian cancer organoid biobank was established, and HTDS was conducted using CCC organoids based on libraries of 361 and 4,560 compounds. The efficacy of the identified drugs was verified in in vitro and in vivo experiments using a patient-derived organoid xenograft mouse model. Transcriptome analysis was performed to identify genes related to the pathways targeted by the identified drugs in CCC and to assess their potential as therapeutic targets. Proteasome inhibitors and dinaciclib were extracted using HTDS and shown to inhibit tumorigenesis in vitro and in vivo. CCC, like multiple myeloma, exhibited activated endoplasmic reticulum (ER) stress and unfolded protein response (UPR), and treatment with proteasome inhibitors further enhanced ER stress and UPR, ultimately leading to cell death. Transcriptome analysis identified anterior gradient-2 (AGR2) as a key gene involved in UPR in CCC. CRISPR knockout of AGR2 suppressed cell proliferation, increased sensitivity to proteasome inhibitors, and reversed platinum resistance in CCC. AGR2 knockout also upregulated Schlafen 11, contributing to platinum sensitivity. ER stress and the UPR are activated in CCC, and proteasome inhibitors disrupt this balance, ultimately leading to cell death. AGR2 may serve as a potential therapeutic target in CCC.

Significance: Proteasome inhibitors and dinaciclib are identified as effective drugs for CCC. CCC has a high basal UPR, and proteasome inhibition may disrupt this balance. AGR2 is involved in the UPR of CCC, and inhibiting AGR2 further enhances the UPR and confers platinum sensitivity, making it a potential therapeutic target.

Figure 1

Establishment of a CCC organoid biobank. A, Clinicopathologic characteristics of the 11 cases of CCC from which organoids were derived. NED, no evidence of disease; OS, overall survival; PFS, progression-free survival. B, ssGSEA showing an activation of the CCC signature in CCC organoids. MM, multiple myeloma. C, Genomic characteristics of the 11 CCC organoids analyzed by exome sequencing (left) and 43 CCC tissues analyzed by WES [Itamochi and colleagues (23); right].

Figure 2

HTDS for CCC organoids using a 384-well plate. A, ATP luminescence-based HTDS system. B, Two independent HTDS showed reproducible results in two CCC organoids. C, Schematic presentation of HTDS with 361 compounds for six CCC organoids. HTDS identified 17 compounds with inhibition efficiencies over 70%. D, Dose–response curve of eight compounds for nine CCC organoids. The IC₅₀ of each compound in each organoid is shown.

Figure 3

An HTDS with 4,560 compounds identified PIs and dinaciclib as candidate drugs for CCC, demonstrating effectiveness in vivo. A, Schematic presentation of HTDS with 4,560 compounds for two CCC organoids (18-015 and 19-042). DHFR, dihydrofolate reductase; HDAC, histone deacetylase. B, Results (% control) of eight PIs in the HTDS in two CCC organoids. C, Dose–response curve of three PIs (bortezomib, carfilzomib, and ixazomib), dinaciclib, paclitaxel, and cisplatin in nine CCC organoids (18-015, 19-001, 19-010, 19-042, 19-044, 19-055, 19-076, 19-079, and 18-148). The IC₅₀ values of each compound in each organoid are shown. D, Histopathology of a PDOX (19-042) and its original tumor. H&E, hematoxylin and eosin. E,In vivo experiments using PDOX treated with bortezomib or dinaciclib. Bortezomib treatment significantly reduced tumor volume (P = 0.004) and tumor weight (P = 0.013) without body weight loss. Dinaciclib treatment showed a tendency to reduce tumor volume (P = 0.093) and significantly reduce tumor weight (P = 0.030) without body weight loss. DUB, deubiquitinase; NAMPT, nicotinamide phosphoribosyltransferase; N.S., not significant; PLK, polo-like kinase; TOPK, T-LAK cell-originated protein kinase.

Figure 4

The activation of the UPR in CCC, with a higher level of activation observed following treatment with bortezomib. A, GSEA of the IRE1α pathway, PERK pathway, and ATF6 pathway in CCC, multiple myeloma from TCGA, and normal ovary from the Genotype-Tissue Expression database. B, PCA of ssGSEA using ATF6, IRE1α, and PERK pathways. TCGA-OV, TCGA ovarian cancer; TCGA-PAAD, TCGA pancreatic adenocarcinoma. C, Bright-field image of CCC organoid (19-042) treated with bortezomib (50 nmol/L, 24 hours) or dinaciclib (500 nmol/L, 24 hours). D, Immunoblot analysis of UPR markers (IRE1α, XBP1s, PERK, and ATF6) and an apoptosis marker (cleaved caspase-3) following treatment with bortezomib (10 or 50 nmol/L, 24 hours) or dinaciclib (100 or 500 nmol/L, 24 hours). E, TEM images after treatment with bortezomib (50 nmol/L, 24 hours) or dinaciclib (500 nmol/L, 24 hours). The arrowhead indicates the ER. Bortezomib treatment at 50 nmol/L significantly induced ER lumen dilatation (P < 0.001), whereas dinaciclib treatment did not. The arrow indicates the abnormal lysosome. Scale bars, 1 μm (top) and 2 μm (bottom). MM, multiple myeloma.

Figure 5

AGR2-high CCC organoids are sensitive to bortezomib, and AGR2 knockout reduces tumor growth. A, Heatmap showing the IC₅₀ of drugs selected through the HTDS of 361 compounds (Fig. 2) of nine organoids (18-015, 19-079, 19-055, 19-076, 18-148, 19-042, 19-001, 19-010, and 19-044). High, green; low, magenta. B, Genes shared in at least two of the UPR-related pathways (GOBP_IRE1_mediated_unfolded_protein_response, GOBP_regulation_of_PERK_mediated_unfolded_protein_response, and GOBP_ATF6_mediated_unfolded_protein_response). AGR2 was significantly downregulated in the bortezomib-resistant 18-015 CCC organoid (P = 0.0049). C, Read count of AGR2 in RNA sequencing of bortezomib-resistant (18-015) and other CCC organoids. D, Immunoblot analysis of UPR-related genes, including AGR2 in bortezomib-resistant (18-015) and other CCC organoids (19-001, 19-042, 19-055, 19-076, and 19-079). E, Immunoblot analysis of AGR2-KO CCC organoid (19-042) confirmed knockout of AGR2. F,AGR2-KO organoids exhibited significantly slower growth than the control (P = 0.014). G, Cell cycle analysis of the AGR2-KO CCC organoid (19-042) showed a significant increase in the G₂–M fraction compared with the control (P = 0.028).

Figure 6

AGR2 knockout sensitizes CCC organoids to carboplatin through upregulation of SLFN11 and induces UPR and ER stress. A, Dose–response curve of bortezomib, dinaciclib, and carboplatin in AGR2-KO organoids. AGR2-KO organoids showed increased sensitivity to bortezomib (P = 0.007 and 0.012) and carboplatin (both P = 0.010) but not to dinaciclib. B, Correlation between AGR2 and SLFN11 mRNA in the Cancer Cell Line Encyclopedia dataset (Broad, 2019; ref. 52). AGR2 expression was significantly related to low SLFN11 expression (P = 5.087 × 10⁻¹⁴). C, Immunoblot analysis of UPR pathway genes and SLFN11 in AGR2-KO CCC organoids. D, Activation of the UPR pathway in AGR2-KO CCC organoids (Hallmark_unfolded_protein_response, P = 0.0022; GOBP_ATF6_mediated_ unfolded_protein_response, P = 0.00012; GOBP_IRE1_mediated_unfolded_protein_response, P = 0.00073; and GOBP_regulation_of_PERK_mediated_unfolded_protein_response, P = 0.76, compared with the control). E, TEM images the of AGR2-KO CCC organoid. The arrowhead indicates the ER. AGR2 knockout showed significant ER lumen dilatation (clone B, P < 0.001 and clone C, P < 0.001). F, Higher AGR2 expression in CCC (n = 51) than in HGSC (n = 43; P < 0.001). AGR2 expression was not correlated with progression-free survival in CCC (P = 0.74). G, Schematic representation of the identified mechanisms in this study. CCC exhibits high UPR, which is further upregulated by PIs to induce apoptosis. There are many AGR2-high CCCs, and AGR2 inhibition upregulates UPR and SLFN11 expression, leading to growth inhibition and platinum sensitivity.