Characterization Of Patient-Derived Xenograft Models Of Endometrial Cancer.

The appropriate models to use for studies of cancer development, progression, drug response and resistance to therapy remain hotly contested. Leading models include cell lines, genetically-modified mice, patient-derived organoids and patient-derived xenografts (PDX). Progress in studies of endometrial cancer is hampered in part by the lack of extensively characterized research models. Our group recently performed next-generation sequencing on the five most commonly-studied endometrial cancer cell lines, Ishikawa H, ECC1, Hec50co, KLE and RL95-2. Our data, which includes histology, mutation screening, MLH1 promoter methylation, copy number variation and microsatellite instability (MSI), conclude that none of the five models cleanly fit any extant post-hoc classification scheme, including the four clusters defined by TCGA: POLE ultra-mutated (cluster 1), MSI hypermutated (cluster 2), copy-number low (endometrioid, cluster 3), and copy-number high (serous-like, cluster 4). We hypothesized that cell lines may not fully represent the true biology of endometrial cancer. Therefore, we have now established PDX models of endometrial cancer in hopes of creating a more representative set of research tools. We generated PDX models by implanting fresh endometrial tumor tissue (hysterectomy specimens) subcutaneously into athymic mice. Of 21 implanted tumors, 8 successfully grafted (take rate=38%). After the initial growth in mice (passage 1 or P1), tumors could be excised, cryopreserved and stored for extended periods in liquid nitrogen. Cryopreserved tumors retain the original phenotype of the P1 engrafted PDX tumor, maintain the ability to proliferate and can be utilized for subsequent larger-scale xenograft studies. Histology is preserved for 5-7 passages, including the architectural pattern of abnormal glands, sheets and occasional papillary configurations. Additionally, some PDX models could be cultured in 2D. We next undertook an extensive morphologic and molecular characterization of three PDX models, denoted EC-PDX1, 2 and 3. EC-PDX1 is serous by histology but contains a POLE mutation (P286R) along with mutations in MSH2, ARID1A and TP63, suggestive of a cluster 1 POLE ultramutated phenotype. Expression of estrogen and progesterone receptors (ER and PR) were low or absent. EC-PDX1 was sensitive to cisplatin in vivo. EC-PDX2 is a high-grade endometrioid model devoid of ER, PR and p53 by immunohistochemistry. Mutational analysis identified a frame shift mutation in exon 8 of p53, which results in a premature stop codon and supports the null staining. These data suggest that EC-PDX2 is most similar to copy-number high cluster 4. EC-PDX3 is low-grade endometrioid with robust expression of ER and PR and wild-type levels of p53. This model harbors activating mutations in PI3KCA (E545K) and KRAS (G13D) and a stop gained mutation in ARID1A. Based on these features, we bin EC-PDX3 as a copy-number low cluster 3. We conclude that PDX models may be superior to cancer cell lines for some studies of the biology and therapeutic response of endometrial cancer. Citation Format: Yuping Zhang, Eric J. Devor, Donghai Dai, Jesus Gonzalez-Bosquet, Kristina W. Thiel, Kimberly K. Leslie. Characterization of patient-derived xenograft models of endometrial cancer [abstract]. In: Proceedings of the AACR Virtual Special Conference: Endometrial Cancer: New Biology Driving Research and Treatment; 2020 Nov 9-10. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(3_Suppl):Abstract nr PR009.