Assessment of Patient-Derived Xenograft Growth and Antitumor Activity: The NCI PDXNet Consensus Recommendations

Although patient-derived xenografts (PDXs) are commonly used for preclinical modeling in cancer research, a standard approach to in vivo tumor growth analysis and assessment of antitumor activity is lacking, complicating comparison of different studies and determination of whether a PDX experiment has produced evidence needed to consider a new therapy promising. We present consensus recommendations for assessment of PDX growth and antitumor activity, providing public access to a suite of tools for in vivo growth analyses. We expect that harmonizing PDX study design and analysis and access to a suite of analytical tools will enhance information exchange and facilitate identification of promising novel therapies and biomarkers for guiding cancer therapy.

Integration of Computational Docking into Anti-Cancer Drug Response Prediction Models

Cancer is a heterogeneous disease in that tumors of the same histology type can respond differently to a treatment. Anti-cancer drug response prediction is of paramount importance for both drug development and patient treatment design. Although various computational methods and data have been used to develop drug response prediction models, it remains a challenging problem due to the complexities of cancer mechanisms and cancer-drug interactions. To better characterize the interaction between cancer and drugs, we investigate the feasibility of integrating computationally derived features of molecular mechanisms of action into prediction models. Specifically, we add docking scores of drug molecules and target proteins in combination with cancer gene expressions and molecular drug descriptors for building response models. The results demonstrate a marginal improvement in drug response prediction performance when adding docking scores as additional features, through tests on large drug screening data. We discuss the limitations of the current approach and provide the research community with a baseline dataset of the large-scale computational docking for anti-cancer drugs.

Combining the Tyrosine Kinase Inhibitor Cabozantinib and the mTORC1/2 Inhibitor Sapanisertib Blocks ERK Pathway Activity and Suppresses Tumor Growth in Renal Cell Carcinoma

Current treatment approaches for renal cell carcinoma (RCC) face challenges in achieving durable tumor responses due to tumor heterogeneity and drug resistance. Combination therapies that leverage tumor molecular profiles could offer an avenue for enhancing treatment efficacy and addressing the limitations of current therapies. To identify effective strategies for treating RCC, we selected ten drugs guided by tumor biology to test in six RCC patient-derived xenograft (PDX) models. The multitargeted tyrosine kinase inhibitor (TKI) cabozantinib and mTORC1/2 inhibitor sapanisertib emerged as the most effective drugs, particularly when combined. The combination demonstrated favorable tolerability and inhibited tumor growth or induced tumor regression in all models, including two from patients who experienced treatment failure with FDA-approved TKI and immunotherapy combinations. In cabozantinib-treated samples, imaging analysis revealed a significant reduction in vascular density, and single-nucleus RNA sequencing (snRNA-seq) analysis indicated a decreased proportion of endothelial cells in the tumors. SnRNA-seq data further identified a tumor subpopulation enriched with cell-cycle activity that exhibited heightened sensitivity to the cabozantinib and sapanisertib combination. Conversely, activation of the epithelial-mesenchymal transition pathway, detected at the protein level, was associated with drug resistance in residual tumors following combination treatment. The combination effectively restrained ERK phosphorylation and reduced expression of ERK downstream transcription factors and their target genes implicated in cell-cycle control and apoptosis. This study highlights the potential of the cabozantinib plus sapanisertib combination as a promising treatment approach for patients with RCC, particularly those whose tumors progressed on immune checkpoint inhibitors and other TKIs.

SIGNIFICANCE

The molecular-guided therapeutic strategy of combining cabozantinib and sapanisertib restrains ERK activity to effectively suppress growth of renal cell carcinomas, including those unresponsive to immune checkpoint inhibitors.

Abstract 5407: A pan-cancer PDX histology image repository with genomic and pathological annotations for deep learning analysis

Patient-derived xenografts (PDXs) model human intra-tumoral heterogeneity in the context of the intact tissue of immunocompromised mice. Histological imaging via hematoxylin and eosin (H&E) staining is performed on PDX samples for routine assessment and, in principle, captures the complex interplay between tumor and stromal cells. Deep learning (DL)-based analysis of large human H&E image repositories has extracted inter-cellular and morphological signals correlated with disease phenotype and therapeutic response. Here, we present an extensive, pan-cancer repository of nearly 1,000 PDX and paired human progenitor H&E images. These images, curated from the PDXNet consortium, are associated with genomic and transcriptomic data, clinical metadata, pathological assessment of cell composition, and, in several cases, detailed pathological annotation of tumor, stroma, and necrotic regions. We demonstrate that DL can be applied to these images to classify tumor regions with an accuracy of 0.87. Further, we show that DL can predict xenograft-transplant lymphoproliferative disorder, the unintended outgrowth of human lymphocytes at the transplantation site, with an accuracy of 0.97. This repository enables PDX-specific investigations of cancer biology through histopathological analysis and contributes important model system data that expand on existing human histology repositories. We expect the PDXNet Image Repository to be valuable for controlled digital pathology analysis, both for the evaluation of technical issues such as stain normalization and for development of novel computational methods based on spatial behaviors within cancer tissues.

Citation Format: Brian S. White, Xing Yi Woo, Soner Koc, Todd Sheridan, Steven B. Neuhauser, Shidan Wang, Yvonne A. Evrard, John David Landua, R Jay Mashl, Sherri R. Davies, Bingliang Fang, Maria Gabriela Raso, Kurt W. Evans, Matthew H. Bailey, Yeqing Chen, Min Xiao, Jill Rubinstein, Ali Foroughi pour, Lacey Elizabeth Dobrolecki, Maihi Fujita, Junya Fujimoto, Guanghua Xiao, Ryan C. Fields, Jacqueline L. Mudd, Xiaowei Xu, Melinda G. Hollingshead, Shahanawaz Jiwani, PDXNet consortium, Tiffany A. Wallace, Jeffrey A. Moscow, James H. Doroshow, Nicholas Mitsiades, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis G. Carvajal-Carmona, Alana L. Welm, Bryan E. Welm, Ramaswamy Govindan, Shunqiang Li, Michael A. Davies, Jack A. Roth, Funda Meric-Bernstam, Yang Xie, Meenhard Herlyn, Li Ding, Michael T. Lewis, Carol J. Bolt, Dennis A. Dean, Jeffrey H. Chuang. A pan-cancer PDX histology image repository with genomic and pathological annotations for deep learning analysis. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5407.

Abstract 5720: Combination therapies in matched 3D in vitro and in vivo preclinical models of rare and recalcitrant cancers from the National Cancer Institute’s Patient-Derived Models Repository

There is a major need in oncology drug development to establish predictive preclinical assays with high translational relevance to patient responses. The National Cancer Institute’s Patient-Derived Models Repository (https://pdmr.cancer.gov) offers a collection of highly characterized models from a variety of cancer types including rare and recalcitrant malignancies and tumors from patients of diverse ancestry. This collection includes matched sets of patient-derived cell lines, organoids, and xenografts (PDXs), which allows comparisons of drug responses from in vitro and in vivo assays performed with the same patient-derived tumor model. A high-throughput screen was conducted using matched sets of patient-derived organoids and cell lines. Patient-derived cell lines were grown as 3D multicellular spheroids mixed with endothelial cells and mesenchymal stem cells. The patient-derived organoids were 100% tumor cells and were plated in 5% basement membrane extract supplemented with growth factors and cytokines. All drugs were tested at concentrations up to their reported clinical Cmax values and cell viability for individual drug treatments and drug combinations were assayed using CellTiter-Glo 3D after seven days drug exposure. Prior to the endpoint viability measurements, growth curves for spheroid median volume and organoid median surface area were calculated from a series of non-invasive brightfield images collected every 12 hours. For some drug combinations, differential responses were observed between the matched organoids and multicellular spheroids, potentially reflecting the contribution of the stromal component in the spheroids. Overall, the drug-dependent growth responses observed from the two 3D in vitro models (i.e., multicellular spheroids and organoids) were frequently comparable to those observed in vivo from PDXs. For example, the in vitro activities of several drug combinations including: BAY1895344 + temozolomide, erlotinib + cediranib, entinostat + talazoparib, and selumetinib + abemaciclib, demonstrated good agreement with the responses observed in vivo. However, among the drug combinations tested ixazomib + panobinostat showed the greatest cytotoxicity in vitro but had no activity in the matched PDX models. The availability of matched patent-derived cell lines, organoids and PDXs provides an opportunity to learn about the features of assay methodologies and data analyses that influence the successful translation of preclinical results between in vitro and in vivo systems. The results of this study are encouraging, but also highlight discrepancies that will be important to investigate, understand and address in order to improve translational capacity of future assays. This project was funded in part with federal funds from the NCI, NIH, under contract no. HHSN261201500003I.

Citation Format: Thomas S. Dexheimer, Thomas Silvers, Rene Delosh, Russell Reinhart, Chad Ogle, Zahra Davoudi, Eric Jones, Debbie Trail, John Carter, Justine Mills, Kyle Georgius, Howard Stotler, Michelle Norris, Shannon Uzelac, Suzanne Borgel, Tiffanie Minor, Luke Stockwin, Michael Mullendore, Kevin Plater, Keegan Kalmbach, Jessica Steed, Matthew Murphy, Gareth Bliss, Carrie Bonomi, Kelly Dougherty, Marion Gibson, Kevin Cooper, Dianne Newton, Cindy R. Timme, Yvonne A. Evrard, Melinda G. Hollingshead, Nathan P. Coussens, Ralph E. Parchment, James H. Doroshow, Beverly A. Teicher. Combination therapies in matched 3D in vitro and in vivo preclinical models of rare and recalcitrant cancers from the National Cancer Institute’s Patient-Derived Models Repository. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5720.

Abstract 6072: Chromosomal aneuploidy, whole-genome doubling and mutational signatures in NCI PDMR models

Introduction: Structural variants (SVs) are a unique class of mutations which have certain therapeutic implications for the tumor. Certain SVs, such as chromosomal aneuploidy, whole-genome doubling (WGD), have specific therapeutic implications. The underlying cellular processes present in the tumor are reflected in mutational signatures. Here, we describe the landscape of chromosomal aneuploidy, WGD and mutational signatures in the National Cancer Institute’s Patient-Derived Models Repository (NCI PDMR) to facilitate the investigation of their roles in therapeutic responses of the preclinical models.

Method: Chromosome arm-level aneuploidy was called by scoring at the individual arm level if >90% of the arm copy number (CN) was gained/lost based on whole-exome sequencing (WES) data. Aneuploidy score was defined as number of arms with aneuploidy. WGD was determined by derived allelic specific CN, purity and ploidy from tumor/normal matched samples and permutation test. Mutational signatures (COSMIC v3) including single base substitutions (SBS), doublet base substitutions (DBS), small insertions and deletions (ID) and CN signatures were derived using SigProfiler for specimens with somatic mutations and CNs.

Results: A large fraction (85%) of patient-derived xenograft (PDX) models (N=755) have at least one arm -level aneuploidy. Certain chromosomes and arms (7, 8, 17p and 18) are more frequently aneuploid, which might be biased due to the overrepresentation of gastrointestinal cancer in the cohort. Histology specific differences were observed in the frequency of arm level aneuploidies. For example, synovial sarcoma (SYNS) and endometrioid carcinoma (UEC) have much lower level of aneuploidy than non-small cell lung cancer (NSCLC) or clear cell renal carcinoma (ccRCC) models. 61% of PDX models (N=277) have WGD, in which certain histologies have more WGD [NSCLC: 81%, head and neck squamous cell carcinomas (HNSCC): 71%] than others. Samples having WGD have a higher degree of aneuploidy and chromosomal instability. WGD and aneuploidy remain stable along the passages in 78% PDX models. Intra-model heterogeneity of WGD was observed due to lineage difference. Mutational signatures (SBS6,15,20) indicating concurrent DNA polymerase epsilon (POLE) mutation and defective DNA mismatch repair were highly enriched in microsatellite instability-high models (p<0.01, Fisher’s exact test). Among 30 PDX models where the patients had known platinum-based chemotherapy history, 40% of them had an identifiable platinum chemotherapy treatment signature (SBS31 or DBS5). Chromothripsis associated amplification signature (CN8) was enriched in models with WGD (p<0.05).

Conclusion: We have characterized chromosomal aneuploidy, WGD and mutational signatures in NCI PDMR models. The models with SVs can be utilized in preclinical drug studies to understand their role in therapeutic response in patients.

Citation Format: Li Chen, Biswajit Das, Ting-Chia Chang, Yvonne A. Evrard, Chris A. Karlovich, Alyssa Chapman, Brandie Fullmer, Ashley Hayes, Ruth Thornton, Nikitha Nair, Shahanawaz Jiwani, Lindsay Dutko, Kelly Benauer, Gloryvee Rivera, Corinne Camalier, John Carter, Suzanne Borgel, Tiffanie Miner, Chelsea McGlynn, Justine Mills, Shannon Uzelac, Tia Shearer, Lauren Hicks, Michelle Norris, Carley Border, Sergio Alcoser, Thomas Walsh, Michael Mullendore, Michelle Eugeni, Dianne Newton, Melinda G. Hollingshead, P. Mickey Williams, James H. Doroshow. Chromosomal aneuploidy, whole-genome doubling and mutational signatures in NCI PDMR models. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6072.

Abstract 2050: Molecular subclassification of NCI PDMR breast cancer models using PAM50 gene expression signature

Background: Breast cancer is the second most common cancer in women. In 2022, it accounted for 15% of total new cancer cases and is the number four cause of death among all cancer types. To benefit from precision medicine, distinguishing molecular subtypes for prognosis and treatment in a clinical setting is essential. While intrinsic subtype classification from NGS results of patients is well established, the approach has not been comprehensively described for patient-derived xenograft (PDX) models, which have been shown to be powerful in translational research. The National Cancer Institute's Patient-Derived Models Repository (NCI PDMR; https://pdmr.cancer.gov) provides rich information in developing the method.

Materials and Methods: Normalized gene expression data of breast cancer PDX and patient specimens (originators) were extracted using tximport and DESeq2 based on RNA-seq analysis. The immunohistochemistry (IHC) was used to determine the status of ER, PR and HER2 receptor expression in these tumor specimens. The PAM50 classification was performed by the R package Genefu. For further analysis, the PAM50 centroids for all 5 subtypes were also obtained from Genefu.

Results: Using the RNA-seq data from 43 PDX models (180 PDX samples, 4~6 samples/model), we were able to predict subtypes at the model level based on the PAM50 method: There are 1 Luminal A subtypes, 5 Luminal B; 6 Her2; 30 Basal and 1 Normal, which encompasses the whole spectrum of PAM50. Thirty originators were also included and there are 8 Luminal A, 9 Luminal B, 2 Her2 and 11 Basal. With the matched 11 originators and the PDX models, 91% of their predicted subtypes are identical; 0.80 Cohen’s kappa was obtained, indicating high inter-rater agreement. We also described subsequent analysis with IHC data-based subtypes. For the 10 originators having IHC-based subtypes, 90% agreement was observed; for 24 PDX models with IHC data, 88% was observed. Of all the 180 PDX samples, 33 of the 43 PDX models (77%) have consistent predicted PAM50 molecular subtypes across different passages and lineages. Within the discordant samples, we observed cases such as a mixture of luminal B and Basal, which can be reasonably interpreted by AR positive signal from IHC. The discrepancy encourages further PDX subclassification from the Basal subtype.

Conclusions: Using our high-throughput gene expression profiles from many patients and samples from patient derived models, we have demonstrated the feasibility of applying classic PAM50 classification algorithm, which was originally developed with microarray data, to be able to recognize the expression signals from our RNA-seq data. Overall, this study should set a primer for the identification of PDX-based subtypes, starting from breast cancer.

Citation Format: Peter I. Wu, Lindsay Dutko, Shahanawaz Jiwani, Li Chen, Biswajit Das, Ting-Chia Chang, Yvonne A. Evrard, Chris A. Karlovich, Alyssa Chapman, Brandie Fullmer, Ashley Hayes, Ruth Thornton, Nikitha Nair, Kelly Benauer, Gloryvee Rivera, Thomas Forbes, John Carter, Suzanne Borgel, Tiffanie Miner, Chelsea McGlynn, Justine Mills, Shannon Uzelac, Tia Shearer, Lauren Hicks, Michelle Norris, Carley Border, Sergio Alcoser, Thomas Walsh, Michael Mullendore, Michelle Eugeni, Dianne Newton, Melinda G. Hollingshead, P. M. Williams, James H. Doroshow. Molecular subclassification of NCI PDMR breast cancer models using PAM50 gene expression signature [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2050.

Data augmentation and multimodal learning for predicting drug response in patient-derived xenografts from gene expressions and histology images

Patient-derived xenografts (PDXs) are an appealing platform for preclinical drug studies. A primary challenge in modeling drug response prediction (DRP) with PDXs and neural networks (NNs) is the limited number of drug response samples. We investigate multimodal neural network (MM-Net) and data augmentation for DRP in PDXs. The MM-Net learns to predict response using drug descriptors, gene expressions (GE), and histology whole-slide images (WSIs). We explore whether combining WSIs with GE improves predictions as compared with models that use GE alone. We propose two data augmentation methods which allow us training multimodal and unimodal NNs without changing architectures with a single larger dataset: 1) combine single-drug and drug-pair treatments by homogenizing drug representations, and 2) augment drug-pairs which doubles the sample size of all drug-pair samples. Unimodal NNs which use GE are compared to assess the contribution of data augmentation. The NN that uses the original and the augmented drug-pair treatments as well as single-drug treatments outperforms NNs that ignore either the augmented drug-pairs or the single-drug treatments. In assessing the multimodal learning based on the MCC metric, MM-Net outperforms all the baselines. Our results show that data augmentation and integration of histology images with GE can improve prediction performance of drug response in PDXs.

Abstract 3091: Patient-derived organoid drug responses corroborate known target-drug interactions for selected anticancer agents

Patient-derived organoids (PDOrgs) are heterogeneous three-dimensional cellular clusters that have been shown to recapitulate the tumor histology and genetic alterations of their originating tissue. Numerous studies suggest the in vitro drug responses of tumor organoids align with in vivo responses. In this study, we evaluated fourteen anticancer agents against a cohort of PDOrgs from three disease histologies: colon, pancreatic, and non-small cell lung adenocarcinoma. The PDOrgs were obtained from the National Cancer Institute’s Patient-Derived Models Repository (https://pdmr.cancer.gov): a resource that offers clinically annotated and molecularly characterized models. The PDOrg models were selected for specific genetic variants of KRAS and BRAF, or different RNA levels of ABCB1, an ATP-dependent efflux pump. The approved and investigational agents were selected to target specific genetic variants and pathways: KRAS G12C covalent inhibitors (sotorasib and MRTX-1257), RAS pathway inhibitors (BAY-293, BI-3406 and TNO-155), BRAF V600E/K inhibitors (dabrafenib and encorafenib), ABCB1 substrates (paclitaxel, doxorubicin, 5-FU, AZD-1775, and SN-38), and ABCB1 non-substrates (gemcitabine and trametinib). The goal of the study was to assess whether the sensitivities of PDOrgs to therapeutic agents matched these genetic profiles under standard in vitro conditions. PDOrgs were seeded into 384-well microplates, in a semi-automated fashion, and exposed to nine concentrations of each anticancer agent for six days followed by cell viability assessment by CellTiter-Glo 3D. Data analysis was performed using GRmetrics, an R package for calculation and visualization of concentration-response metrics based on growth rate inhibition (https://git.bioconductor.org/packages/GRmetrics). These data demonstrated that PDOrgs harboring a KRAS G12C variant were uniquely sensitive to sotorasib and MRTX-1257 and were, overall, more sensitive to the other RAS pathway targeting agents. Conversely, PDOrgs harboring wild type RAS and other KRAS variants were largely unresponsive to these targeted agents. Likewise, only PDOrgs harboring the BRAF V600E variant were sensitive to dabrafenib and encorafenib. For the majority of PDOrgs, the pharmacological responses to agents that are ABCB1 substrates inversely correlated with ABCB1 RNA expression. This study demonstrates the ability of organoids to serve as useful models for evaluating therapeutic responses to anticancer agents, including identifying known target-drug associations. Moreover, the technical conditions, as well as the selected PDOrgs and therapeutic agents, may be used as a reference set for the validation of a fully automated PDOrg screening system. This project was funded in part with federal funds from the NCI, NIH, under contract no. HHSN261201500003I.

Citation Format: Curtis Hose, Erik Harris, John Connelly, Petreena S. Campbell, Mariaestela Ortiz, Eric Jones, Dianne Newton, Yvonne A. Evrard, Melinda Hollingshead, Ralph Parchment, Beverly A. Teicher, Nathan P. Coussens, James H. Doroshow, Annamaria Rapisarda. Patient-derived organoid drug responses corroborate known target-drug interactions for selected anticancer agents [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3091.

Abstract 1913: Quality control workflows developed for the NCI Patient-Derived Models Repository using low pass whole genome sequencing and whole exome sequencing

Background: The National Cancer Institute's Patient-Derived Models Repository (NCI PDMR; pdmr.cancer.gov) is developing a variety of patient-derived xenograft (PDX) models for pre-clinical drug studies. All NCI PDMR models undergo quality control (QC) processes. Two unique QC challenges are: a) to assess genomic stability across PDX model passages; and b) to confirm the suitability of PDX-derived cancer associated fibroblasts (CAFs) as germline surrogates when blood is not available. Multiple bioinformatics QC assessments have been developed to measure the genomic fidelity in these PDX models using low-pass whole genome sequencing (LP-WGS) and in CAFs using whole exome sequencing (WES).

Methods: LP-WGS was performed on 502 PDX samples from 38 models of rare cancer across passages 2 through 9 and WES was performed on 92 CAFs from 32 different histologies. In the QC workflow for estimating the genomic stability of passages within models, BBSplit was used for the assessment of human/mouse DNA content. CNVkit was utilized for copy number (CN) detection. The fraction of genome changed was calculated by comparing the copy numbers of each passage sample to the original patient sample. To evaluate purity of CAFs, three QC steps were constructed: a) plot of SNP variant allele frequency (ideogram); b) variant annotation using OncoKB (www.oncokb.org); c) percentage of genomic loss of heterozygosity (LOH), based on a set of ~800,000 heterozygous SNPs from a population-level genomic database (gnomAD) based on WES data.

Results: PDX models showed genomic stability in CN profile when measured by LP-WGS. Human tumor DNA content remains stable ranging from 75-85% across different tiers of PDX passages from Donor +1 to Donor +6 and more. No models showed statistically significant evolution in CN profile, given the average 5 samples per model in each tier of passages. The QC workflow for CAFs generated five categories based on SNP ideograms, the presence/absence of oncogenic variants and LOH. Following observations were made: a) 72.5% CAFs were confirmed as matched diploid CAFs (category 1); b) 6.6% of CAFs were diploid and had >= 1 germline oncogenic variant - classified as category 2. CAFs in category 1&2 were suitable as germline surrogates; c) 12% of CAFs (category 3) showed putative polyploidy on SNP ideograms with no oncogenic variants and suitable for somatic variant calling; d) 8.8% of CAFs (category 4) had polyploidy and oncogenic variants present; e) LOH high CAF (category 5) - we identified a CAF with 42% LOH, later confirmed to be a tumor cell line by immunohistochemistry (IHC). Other CAFs (n=91) showed little variance, ranging from 0.6%-1.7% LOH.

Conclusions: We developed standard QC workflows to evaluate genomic stability of PDX models during passaging and qualify CAFs as germline surrogates for pre-clinical study.

Citation Format: Ting-Chia Chang, Li Chen, Biswajit Das, Yvonne A. Evrard, Chris A. Karlovich, Tomas Vilimas, Alyssa Chapman, Nikitha Nair, Luis Romero, Anna J. Lee Fong, Amanda Peach, Brandie Fullmer, Lindsay Dutko, Kelly Benauer, Gloryvee Rivera, Erin Cantu, Shahanawaz Jiwani, Nastaran Neishaboori, Tomas Forbes, Corinne Camalier, Luke Stockwin, Michael Mullendore, Michelle A. Eugeni, Dianne Newton, Melinda G. Hollingshead, Mickey P. Williams, James H. Doroshow. Quality control workflows developed for the NCI Patient-Derived Models Repository using low pass whole genome sequencing and whole exome sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1913.

Abstract 3120: Method development for generation of PDX models from rapid autopsy samples for the NCI patient-derived models repository

NCI’s Patient-Derived Models Repository (NCI PDMR; pdmr.cancer.gov) has developed a variety of patient-derived models across most solid tumor histologies. These models are early passage, genetically characterized and associated with limited patient treatment history. As part of this effort, the NCI PDMR worked with the University of Nebraska Medical Center Rapid Autopsy Program and Johns Hopkins University Legacy Gift Rapid Autopsy Program to develop and optimize methods for collection, processing, and shipping of autopsy tumor material to maintain viability during overnight transit for use in patient-derived model development. These methods have been successfully transferred to two other participating rapid autopsy programs. To date, 412 autopsy tumor samples from 76 consented patients have been received for model development; 348 shipped overnight in media for next day implantation into NSG host mice and 64 cryopreserved prior to shipping for a comparative assessment of take-rate versus fresh tumor samples. On average 3-8 tumor samples, primary and metastatic, were collected post-mortem from the truncal region of each patient. Histologies include Pancreatic adenocarcinoma (n=43), Cholangiocarcinoma (n=6), Prostate adenocarcinoma (n=6), and 21 others with 1-2 patients/histology. The overall age range of enrolled patients was 5-88yo. The post-mortem cold ischemic time for collections ranged from 1.5 to 20 hours with a median of 3h (avg. 3.75h; outlier >11h removed). Collection methods were optimized to reduce contamination and increase viability of tumor tissues for successful PDX model generation. Of 348 fresh tumor samples collected to date, 69 PDX models from 33 patients have been generated (range 1-6 models/patient) and an additional 55 samples are being monitored for growth in passage 0. The largest public single-patient PDX model sets are for melanoma (899932-113-R, n=6) and two pancreatic adenocarcinomas (521955-158-R, n=6, 217524-143-R, n=4). Important methods for reducing contaminants in autopsy tumor material include sterilization of the surface of the body prior to opening, use of sterile fields, using separate sterile instruments for each collection site, rinsing the surface of the resected tumor tissue, and use of antibiotics in the collection media. The now established SOPs are publicly available on the NCI PDMR website (pdmr.cancer.gov/sops). We recommend incorporating as many of these methods as possible within the limitations of your individual site. Of the 69 models developed to date, 48 are publicly available from the NCI PDMR while the rest are undergoing quality control process prior to public release. Models developed from autopsy material provide a research tool to investigate tumor evolution, differences between primary and metastatic lesions, and assessment of differences in therapeutic response based on differences in the tumor biology.

Citation Format: Yvonne A. Evrard, Michelle Eugeni, Michelle Ahalt-Gottholm, Carrie Bonomi, Suzanne Borgel, Thomas C. Caffrey, John Carter, Ting-Chia Chang, Li Chen, Kevin Cooper, Biswajit Das, Emily Delaney, Kelly Dougherty, Eleonora Duregon, Stephanie Ecker, Joe Geraghty, Marion Gibson, Lauren Hicks, Jenna Hull, Sharon Int Veldt, Shahanawaz Jiwani, Chris A. Karlovich, Jade Loewenstein, Candace Mallow, Chelsea McGlynn, Justine Mills, Tiffanie Miner, Jowaly Schneider, Tia Shearer, Savanna Styers, Shannon Uzelac, Paul Grandgenett, Michael Hollingsworth, Jody E. Hooper, P. Mickey Williams, Melinda Hollingshead, James H. Doroshow. Method development for generation of PDX models from rapid autopsy samples for the NCI patient-derived models repository [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3120.

Abstract 1202: A repository of PDX histology images for exploring spatial heterogeneity and cancer dynamics

Patient-derived xenografts (PDXs) recapitulate intratumoral spatial heterogeneity and simulate a tumor microenvironment in which human immune and stromal cells in the PDX are replaced over passages by murine cells partially lacking immune function. Histological imaging enables exploring the spatial heterogeneity and dynamics of cancer, stromal, and immune cell interactions as correlates of tumor stage and therapeutic response over passages. We created a repository of curated, haematoxylin and eosin (H&E) images as a community resource for addressing these questions.

Images were generated at five sites within the NCI’s PDX Development and Trial Centers Research Network (PDXNet) and the NCI Patient-Derived Models Repository. Over 900 images, including 739 from PDXs and 190 from paired patients, are hosted on the Seven Bridges Genomics Cancer Genomics Cloud. They represent 42 cancer subtypes, including breast cancer (n=134), colon adenocarcinoma (COAD; n=94), pancreatic cancer (n=87), lung adenocarcinoma (LUAD; n=80), melanoma (n=71), and squamous cell lung cancer (LUSC; n=65). Paired human/PDX images are available for each of these cancers. Human and/or PDX images generated following patient treatment are available for 37 of the subtypes. Most images are from early passages (P0: 158; P1: 292; P2: 152; P3: 69; >P3: 55). Annotations include sex, age, race, ethnicity, and, for most images, pathological assessment of tissue-level percent cancer, stromal, and necrotic cell content (n=639) and tumor stage (n=650). RNA and exome sequencing data are available for 99 and 228 images, respectively, matched at the patient or sample level.

Quality control was performed using HistoQC. Cells were segmented and labeled as neoplastic, necrotic, immune, stromal, or other using Hover-Net and predictions of total neoplastic cell area correlated with whole-slide pathological assessment of cancer cell percentage (COAD: r=0.51; LUSC: r=0.59). HD-Staining, another classification approach, was applied to a subset of images and our clinical annotations will facilitate validation of this and related methods. Features of 512 x 512 pixel tiles were computed using the Inception V3 convolutional neural network pre-trained on ImageNet. Unsupervised clustering of these features demonstrate inter-patient heterogeneity within pathologist-annotated tumor regions. A classifier developed using pathologist-annotated cancer, stromal, and necrotic regions and trained on the features in LUSC images (n=10 images) achieved a cross-validation accuracy of 96% for cancer tiles across (n=5) LUAD images. Accuracy was lower for stromal classification (90%), likely reflecting current limitations of our small, but growing, labeled training set.

Our repository of clinically-annotated PDX H&E images should aid the community in studying spatial heterogeneity and in training deep learning-based image analysis methods.

Citation Format: Brian S. White, Xingyi Woo, Soner Koc, Todd Sheridan, Steven B. Neuhauser, Akshat M. Savaliya, Lacey E. Dobrolecki, John D. Landua, Matthew H. Bailey, Maihi Fujita, Kurt W. Evans, Bingliang Fang, Junya Fujimoto, Maria Gabriela Raso, Shidan Wang, Guanghua Xiao, Yang Xie, Sherri R. Davies, Ryan C. Fields, R Jay Mashl, Jacqueline L. Mudd, Yeqing Chen, Min Xiao, Xiaowei Xu, Melinda G. Hollingshead, Shahanawaz Jiwani, PDXNet Consortium, Yvonne A. Evrard, Tiffany A. Wallace, Jeffrey A. Moscow, James H. Doroshow, Nicholas Mitsiades, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis G. Carvajal-Carmona, Alana L. Welm, Bryan E. Welm, Michael T. Lewis, Ramaswamy Govindan, Li Ding, Shunqiang Li, Meenhard Herlyn, Michael A. Davies, Jack A. Roth, Funda Meric-Bernstam, Carol J. Bult, Brandi Davis-Dusenbery, Dennis A. Dean, Jeffrey H. Chuang. A repository of PDX histology images for exploring spatial heterogeneity and cancer dynamics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1202.

Abstract 80: Genomic characterization of PDX models from rare cancer patients in the NCI Patient-Derived Models Repository

Background: The National Cancer Institute’s Patient-Derived Models Repository (NCI PDMR; https://pdmr.cancer.gov) has developed a large number of patient-derived xenograft (PDX) models from a diverse set of rare cancers. These models have been genomically characterized using whole-exome sequencing (WES) and RNAseq. The resource provides a unique opportunity to explore the genomic features of rare tumor models in NCI PDMR and to understand the oncogenic processes in pre-clinical models to identify biomarkers associated with therapeutic responses.

Methods: Genomic characterization was done in 4-6 PDX samples across multiple passages and lineages from each model. As the samples exhibited a high level of genomic stability within each model, consensus mutation and copy number variation (CNV), microsatellite instability (MSI), genomic loss of heterozygosity (LOH), homologous recombination deficiency score (scarHRD), and mutational signature data were generated from WES. Fusions were identified from RNASeq data using Star-Fusion and FusionInspector. Gene set enrichment analysis was conducted from the gene expression data obtained from RNAseq.

Results: 1) 233 PDX models have been developed and characterized from more than 45 different rare malignancies. Most frequent cancer types are different sarcomas (n=63), head & neck squamous cell carcinoma (n=61), and malignant fibrous histiocytoma (MFH) (n=11); 2) TP53 was the most frequently altered gene, mutated in 51% of models, followed by NOTCH1 (16%) and PIK3CA (11%). In terms of CNVs, ovarian epithelial cancer (OVT) showed relatively high chromosomal instability, while uterine endometrioid carcinoma (UEC) and synovial sarcoma (SYNS) had low instability; 3) MSI-H was observed in only 7 models. Esophageal adenocarcinoma (ESCA), OVT, and cervical squamous cell carcinoma (CESC) had high scarHRD and genomic LOH scores, while both scores were low in UEC and anal squamous cell carcinoma (ANSC). COSMIC v2 mutational signature 3 is significantly associated with a high scarHRD score (p-value < 0.01, Wilcoxon rank-sum test); 4) Characteristic fusions were observed in certain sarcoma models: SS18-SSX1 and ASPSCR1-TFE3 fusions were observed in SYNS and alveolar soft part sarcoma (ASPS) models respectively. EWSR1-FLI1 fusion was present in 2 out of 3 Ewing sarcoma (ES) models. 5) Gene set enrichment analysis from RNASeq data showed that epithelial-mesenchymal transition score could accurately distinguish carcinoma from sarcoma models, confirming the divergent gene expression programs.

Conclusion: Comprehensive genomic characterization of NCI PDMR models generated from rare cancers solves an unmet need in the community. It will serve as a valuable resource for translational researchers interested in pre-clinical drug development and discovery.

Citation Format: Li Chen, Rini Pauly, Ting-Chia Chang, Biswajit Das, Yvonne A. Evrard, Chris A. Karlovich, Tomas Vilimas, Alyssa Chapman, Nikitha Nair, Luis Romero, Anna Lee Fong, Amanda Peach, Shahanawaz Jiwani, Nastaran Neishaboori, Lindsay Dutko, Kelly Benauer, Gloryvee Rivera, Erin Cantu, Corinne Camalier, Thomas Forbes, Michelle Gottholm-Ahalt, John Carter, Suzanne Borgel, Chelsea McGlynn, Candace Mallow, Emily Delaney, Tiffanie Miner, Michelle A. Eugeni, Dianne Newton, Melinda G. Hollingshead, P. Mickey Williams, James H. Doroshow. Genomic characterization of PDX models from rare cancer patients in the NCI Patient-Derived Models Repository [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 80.

Drug response prediction in patient-derived xenografts with data augmentation and multimodal deep learning

e13572

Background: Prediction of drug response is a critical research area in precision oncology and has been previously explored with large drug screening studies of cancer cell lines (CCLs). Patient-derived xenografts (PDXs) are an appealing platform for preclinical drug studies because the in vivo environment of PDXs helps preserve tumor heterogeneity and usually better mimics drug response of patients with cancer compared to CCLs. Methods: We investigate multimodal neural network (NN) and data augmentation for drug response prediction in PDXs. The multimodal NN learns to predict response using drug descriptors, gene expressions (GE), and histology whole-slide images (WSIs) where the multi-modality refers to tumor features only. The NN uses late integration where separate subnetworks are used to encode the input feature types before concatenation and prediction layers. Median tumor volume per treatment group is assessed relative to the control group to create a binary variable representing response. The data include twelve single-drug and 36 drug-pair treatments resulting in 2,556 single-drug and 2,203 drug-pair response values. Pathology and omics data from 487 PDXs from NCI's Patient Derived Models Repository are used as tumor feature model inputs. We explore whether the integration of WSIs with GE improves predictions as compared with models that use GE alone. We use two methods to address the limited number of response values in the dataset: 1) homogenize drug representations which allows to combine single-drug and drug-pairs into a single dataset, 2) augment drug-pair samples by switching the order of drug features which doubles the sample size of all drug-pair samples. These methods enable us to combine single-drug and drug-pair treatments which results in 6,962 responses, allowing us to train multimodal and unimodal NNs without changing architectures or the dataset. Results: Prediction performance of three unimodal NNs which use GE (um1, um2, and um3) are compared to assess the contribution of data augmentation methods. NN um1 that uses the full dataset which includes the original and the augmented drug-pair treatments as well as single-drug treatments significantly outperforms NNs (p-values < 0.01) that ignore either the augmented drug-pairs (um2) or the single-drug treatments (um3). In assessing the contribution of multimodal learning, results show that the multimodal NN (mm) outperforms both unimodal NNs that ignore either the GE (um4) or the WSIs (um1). However, the improvement of mm over um1 is not statistically significant (p-value < 0.26). Conclusions: Our results show that data augmentation and integration of histology images and GE can help improve prediction performance of drug response in PDXs.[Table: see text]

Abstract P097: Comparative single cell transcriptome profiling of primary tumors, CTCs and metastatic sites from a bladder cancer PDX model

Background: A PDX bladder cancer model, BL0293-F563, grows large subcutaneous tumors, spontaneously metastasizes to the liver and bone, and sheds high numbers of circulating tumor cells (CTCs). This PDX model provides a unique opportunity to explore the relationships between primary tumors, CTCs and metastatic cell subpopulations. Methods: BL0293-F563 tumors (available from the NCI Patient-Derived Models Repository [https://pdmr.cancer.gov/] and originally developed by Jackson Laboratories) were implanted into NSG mice and and primary tumors, metastatic nodules in the liver, and blood were collected at maximal allowable tumor burden. Tumor tissue was dissociated using Miltenyi Tumor Dissociation Kit with OctoDissociator, and Human CTCs were enriched from whole mouse blood through negative selection with anti-mouse CD45 and anti-mouse MHC-1 magnetic beads. Single cell sequencing was done using 10X Genomics 3’ gene expression assay v3.1. Sequencing libraries were prepared using 10X Genomics Chromium and 3’ gene expression kit v3.1. Data processing and analysis was done using 10X Genomics’ Cell Ranger pipeline, Seurat, and consensus non-negative matrix factorization. Results: Using Seurat FindNeighbors, cells in the aggregated dataset were classified into 17 distinct clusters. All clusters were comprised of cells from multiple sites (primary tumor, CTCs, metastases), but three clusters were enriched in CTCs and one cluster was composed of mostly primary tumor cells. All clusters exhibited an epithelial-like gene expression signature score, suggesting that CTC shedding was occurring without prominent epithelial-mesenchymal transition. Consistent with expected differences in oxygenation states, CTC-enriched clusters exhibited a lower hypoxia gene expression score than primary tumor and metastasis-enriched clusters. CTC-enriched clusters also showed higher expression of oxidative phosphorylation genes, suggesting metabolic differences between CTCs and cells from primary tumors and metastases. Based on Human Primary Cell Atlas phenotype prediction, several clusters were associated with stem cell like phenotypes. Additionally, two of three CTC-enriched clusters had elevated expression of mitosis-associated genes, suggesting that at least some populations of CTCs are not quiescent but actively cycling. Conclusions: Utilizing single cell gene expression profiling, we have linked the gene expression profile of CTCs to specific cell subpopulations in primary tumors and metastases. We show that CTC-enriched cell clusters appear to maintain an epithelial phenotype. Subpopulations of CTC cells exhibited enrichment of stemness-associated transcripts and features of active cell cycling.

Citation Format: Tomas Vilimas, Brandie Fullmer, Alyssa Chapman, Li Chen, Ting-Chia Chang, Rini Pauly, Biswajit Das, Chris Karlovich, Yvonne A. Evrard, Howard Stotler, Michelle M. Gottholm-Ahalt, Tara Grinnage-Pulley, Melinda G. Hollingshead, James H. Doroshow, P. Mickey Williams. Comparative single cell transcriptome profiling of primary tumors, CTCs and metastatic sites from a bladder cancer PDX model [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2021 Oct 7-10. Philadelphia (PA): AACR; Mol Cancer Ther 2021;20(12 Suppl):Abstract nr P097.

Comprehensive characterization of 536 patient-derived xenograft models prioritizes candidatesfor targeted treatment

Development of candidate cancer treatments is a resource-intensive process, with the research community continuing to investigate options beyond static genomic characterization. Toward this goal, we have established the genomic landscapes of 536 patient-derived xenograft (PDX) models across 25 cancer types, together with mutation, copy number, fusion, transcriptomic profiles, and NCI-MATCH arms. Compared with human tumors, PDXs typically have higher purity and fit to investigate dynamic driver events and molecular properties via multiple time points from same case PDXs. Here, we report on dynamic genomic landscapes and pharmacogenomic associations, including associations between activating oncogenic events and drugs, correlations between whole-genome duplications and subclone events, and the potential PDX models for NCI-MATCH trials. Lastly, we provide a web portal having comprehensive pan-cancer PDX genomic profiles and source code to facilitate identification of more druggable events and further insights into PDXs' recapitulation of human tumors.

Abstract 3007: Imaging characterization of NCI patient derived model repository (PDMR) xenografts

The purpose of this work is to provide imaging data on patient derived xenografts models that are available from the National Cancer Institute Patient-Derived Models Repository (https://pdmr.cancer.gov) that may be useful to investigators in deciding which models to use in testing therapeutics or designing co-clinical trials. Using imaging to monitor progress of a treatment or evaluate development or prevention of metastases has advantages in minimizing the number of mice required as well as providing more information about metabolic status and tumor heterogeneity.

NSG mice were implanted with each model and monitored with non-invasive imaging. The core imaging methodologies were T2W magnetic resonance imaging (MRI), to visualize heterogeneity and determine the usefulness of MRI to detect metastatic disease; 18F fluorodeoxyglucose PET/CT, to evaluate aerobic glycolysis; and 3'-Deoxy-3'-(18)F-fluorothymidine PET/CT, to evaluate proliferation. This work is ongoing and to date we have performed MRI on over 80 models with diverse tumor histology with a subset also undergoing FDG PET (77 models) and FLT PET (36 models).

These models and their imaging characteristics will be presented. Nineteen models have demonstrated metastatic disease that could be detected by non-contrast MRI imaging with pathological confirmation, including ones derived from melanoma, lung, colon, pancreatic, uterine, anal, and head & neck. Four of these (two colon adenocarcinoma, one melanoma and one pancreatic adenocarcinoma) have undergone more extensive characterization evaluating cohorts to determine location, timing of appearance, and penetrance of metastasis. These experiments evaluated cohorts of animals who were monitored as the primary tumors grew and others for whom the primary tumor was excised to encourage metastatic spread. The results of this metastatic characterization will be summarized. Additionally, the image data and the appropriate Standard Operating Procedures (SOP's) will be posted in The Cancer Imagining Archive (TCIA) https://www.cancerimagingarchive.net/ for investigators to access. The goal of this poster is to make investigators aware of the availability of this complementary imaging data as they consider research on models accessible from the NCI PDMR.

Citation Format: Paula M. Jacobs, James L. Tatum, Joseph D. Kalen, Lilia V. Ileva, Lisa A. Riffle, Keita Saito, Nimit L. Patel, Jessica Phillips, Melinda Hollingshead, Yvonne A. Evrard, Michelle Gottholm-Ahalt, Chelsea Sanders, Amy James, Simone Difilippantonio, Elijah F. Edmondson, James H. Doroshow. Imaging characterization of NCI patient derived model repository (PDMR) xenografts [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3007.

Abstract 3012: Patient-derived models of rare cancers in the National Cancer Institute's patient-derived models repository

There is an unmet need for preclinical models of rare cancers and rare disease sub-types. The National Cancer Institute's Patient-Derived Models Repository (NCI PDMR; https://pdmr.cancer.gov) is developing quality-controlled, early-passage, clinically-annotated patient-derived tumor xenografts (PDXs), in vitro tumor cell cultures (PDCs), cancer associated fibroblasts (CAFs), and patient-derived organoids (PDOrg) and has focused on addressing unmet needs in the preclinical model space including developing models from adult and pediatric patients with rare cancers. To date, NCI has created and molecularly characterized over 150 preclinical models of rare cancer including indications such as Hurthle cell carcinoma, osteosarcomas, Merkel cell carcinomas, salivary gland cancers, synovial sarcomas, and carcinosarcomas. Rare cancer models developed to date will be reviewed and their histopathologic and molecular characteristics compared to that reported in the clinical setting. A pipeline to identify fusion proteins in these rare cancers such as the Ewing sarcoma EWSR1-FLI1 fusion and NAB2-STAT6 fusions in solitary fibrous tumors (SFT) has been implemented. Four malignant peripheral nerve sheath tumors (MPNST) PDX models are available for researches; these models were developed from patients diagnosed between the ages of 37-68. At the time of model development, two patients were treatment naïve and two had prior radiotherapy. Two of the MPNST PDX models have NF1 oncogenic mutations, three have deep deletions in CDKN2A/B, and three have a mutation in either EED or SUZ12 consistent with the reported molecular characteristics of patients with MPNST. Also of clinical relevance, of two mesothelioma models available, one carries an NF2 driver mutation and the other BAP1 and LATS2 and a PDX model for Hurthle cell carcinoma has wide-spread loss of heterozygosity (LOH 80%). Models for other rare cancers are in development, including four cholangiocarcinoma PDXs with histopathologic confirmation that are currently being expanded for molecular characterization and distribution. Funded by NCI Contract No. HHSN261200800001E

Citation Format: Cindy R. Timme, Sergio Y. Alcoser, Devynn Breen, John Carter, Ting-Chia Chang, Alice Chen, Li Chen, Kristen Cooley, Biswajit Das, Emily Delaney, Michelle A. Eugeni, Michelle M. Gottholm-Ahalt, Tara Grinnage-Polley, Jenna Hull, Chris Karlovich, Kimberly Klarmann, Shahanawaz Jiwani, Candace Mallow, Chelsea McGlynn, Justine Mills, Malorie Morris, Michael Mullendore, Dianne Newton, Tia Shearer, Jesse Stottlemyer, Shannon Uzelac, Thomas Walsh, P. Mickey Williams, Yvonne A. Evrard, Melinda G. Hollingshead, James H. Doroshow. Patient-derived models of rare cancers in the National Cancer Institute's patient-derived models repository [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3012.

Abstract 3010: Single agent response comparisons in a large-scale, preclinical trial of rare cancer PDXs by the National Cancer Institute's patient-derived models repository

The National Cancer Institute's Patient-Derived Models Repository (NCI PDMR; https://pdmr.cancer.gov) is performing a large-scale preclinical study with 39 patient-derived xenograft (PDX) models of rare cancers (including mesothelioma, MPNST, osteosarcoma, Merkel cell carcinoma) treated with 56 novel therapeutic combinations (targeted and cytotoxic agents) in an exploratory, n-of-4 arm, study design. Drug combinations with additive activity may undergo clinical evaluation in patients with rare cancers. PDX tumors are treated with a set of 8 combinations plus relevant vehicle controls while in parallel enough PDXs are serially passaged for the next passage and drug set. Every serial passage undergoes several quality control assessments that serve as go/no-go criteria. Combinations that show promising responses (e.g., regression or durable tumor growth inhibition) are repeated along with the single agent arms to determine if the response is driven by the combination or only one of the agents. We are currently at the half-way point in the overall study and here report interim results for the early combination agents that have single agent data for comparison. In a combination of a VEGFi and EGFRi, 6/37 models achieved a partial regression (30% shrinkage for more than one consecutive time point) and 17/37 had tumor growth inhibition while drug was on board. Single agent studies have been completed for 17/37 models with this combination and 7/9 responses were due to at least an additive effect of the combination. In contrast, while an HDACi + nucleoside analog combination had 16/36 responsive models, response in most of the single agent studies was due to only one of the agents. As part of this study, 3 models have been identified that have responded to at least 50% of the combinations tested possibly indicating a hypersensitive phenotype: two Merkel cell carcinomas (n=28 and 32) and one Neuroendocrine carcinoma (n=27). There is no immediate link between mechanism of action of the agents in the combinations, and the two Merkel cell carcinoma responses only had a moderate overlap. Finally, two Rhabdomyosarcoma models in the study have been the least responsive models to date. Funded by NCI Contract No. HHSN261200800001E

Citation Format: Yvonne A. Evrard, Sergio Y. Alcoser, Suzanne Borgel, Devynn Breen, John Carter, Tiffanie Chase, Alice Chen, Li Chen, Kristen Cooley, Biswajit Das, Emily Delaney, Lyndsay Dutko, Shannon Ecker, Thomas Forbes, Kyle Georgius, Michelle M. Gottholm-Ahalt, Tara Grinnage-Pulley, Sierra Hoffman, Chris Karlovich, Kimberly Klarmann, Shahanawaz Jiwani, Justine Mills, Malorie Morris, Michael Mullendore, Dianne Newton, Gloryvee Rivera, Howard Stotler, Jesse Stottlemyer, Savanna Styers, Cindy R. Timme, Debbie Trail, Shannon Uzelac, Tomas Vilimas, Thomas Walsh, Nikki Walters, P. Mickey Williams, Melinda G. Hollingshead, James H. Doroshow. Single agent response comparisons in a large-scale, preclinical trial of rare cancer PDXs by the National Cancer Institute's patient-derived models repository [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3010.

Abstract 3009: A systematic review of the tumor growth metrics of patient-derived xenograft (PDX) models in the literature and in NCI PDXNet centers

Background: Despite increasing utilization of patient-derived xenografts (PDXs) in early drug development, there are no agreed upon metrics for assessment of PDX growth inhibition for agents given alone or in combination. In the present study, we aim to investigate what metrics are being used in the literature, as well as among the National Cancer Institute PDX Development and Trial Centers Research Network (PDXNet) investigators.

Methods: Relevant PDX literature was identified and retrieved using an information retrieval tool, RetriLite, to search for articles that met following criteria: 1) Published between 01/2018 through 12/2019; 2) Published in a journal with impact factor of 10 or above; 3) Search terms included: Cancer, PDX(s), patient derived xenograft(s), and patient-derived xenograft(s). Exclusion criteria included: 1) Brain tumors; 2) Immune-oncology/non-solid tumors; 3) Studies with no detailed information; 4) studies from PDXNet investigators. In addition, a questionnaire regarding PDX analysis practices was distributed to NCI PDXNet investigators and responses were analyzed.

Results: Sixty-five studies with relevant information were included in this systematic literature review and 15 NCI PDXNet PIs from all six centers responded to the survey representing the general practice in the network. The most commonly used tumor growth assessment metric was comparisons in tumor volumes in different treatment arms, used by 33 (51%) of 65 PDX papers and 13 (87%) of 15 PDXNet investigators. Thirteen different growth metrics were reported in the PDX literature and ten different metrics were used by PDXNet investigators. PDXNet investigators were more likely to use growth metrics analogous to clinical endpoints compared to the PDX literature, including percent change of tumor volume (80% vs 17%), event-free survival (EFS: 40% vs 11%), and overall survival (33% vs 8%). PDXNet investigators were also more likely to assess objective response rate (ORR) compared to the PDX literature (60% vs 12%); several different cutoffs were used for defining response and progression. For combination therapy, most investigators and literature compared tumor volumes across treatment arms, with few looking at measures of synergy or dynamic effects and with variable utilization of other metrics such as OR and EFS. In PDX literature, of the 40 papers with combination therapies presented, at least one monotherapy control arm was missing in 7 (18%), and four (10%) only compared growth with the no treatment control arm.

Conclusions: In summary, there is great variability in growth metrics used in the PDX community. To better use PDXs as preclinical models and increase the reproducibility of treatment effect on PDXs, a joint effort is needed to harmonize approaches in PDX growth assessment.

Citation Format: Dali Li, Min Jin Ha, Yvonne A. Evrard, Huiqin Chen, Lisa M. McShane, Jeffrey Grover, Jing Wang, Bingliang Fang, Timothy DiPeri, Michael T. Lewis, Lawrence Rubinstein, Jack A. Roth, Jeffrey H. Chuang, James H. Doroshow, Jeffrey A. Moscow, Funda Meric-Bernstam. A systematic review of the tumor growth metrics of patient-derived xenograft (PDX) models in the literature and in NCI PDXNet centers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3009.

TPM, FPKM, or Normalized Counts? A Comparative Study of Quantification Measures for the Analysis of RNA-seq Data from the NCI Patient-Derived Models Repository

Background

In order to correctly decode phenotypic information from RNA-sequencing (RNA-seq) data, careful selection of the RNA-seq quantification measure is critical for inter-sample comparisons and for downstream analyses, such as differential gene expression between two or more conditions. Several methods have been proposed and continue to be used. However, a consensus has not been reached regarding the best gene expression quantification method for RNA-seq data analysis.

Methods

In the present study, we used replicate samples from each of 20 patient-derived xenograft (PDX) models spanning 15 tumor types, for a total of 61 human tumor xenograft samples available through the NCI patient-derived model repository (PDMR). We compared the reproducibility across replicate samples based on TPM (transcripts per million), FPKM (fragments per kilobase of transcript per million fragments mapped), and normalized counts using coefficient of variation, intraclass correlation coefficient, and cluster analysis.

Results

Our results revealed that hierarchical clustering on normalized count data tended to group replicate samples from the same PDX model together more accurately than TPM and FPKM data. Furthermore, normalized count data were observed to have the lowest median coefficient of variation (CV), and highest intraclass correlation (ICC) values across all replicate samples from the same model and for the same gene across all PDX models compared to TPM and FPKM data.

Conclusion

We provided compelling evidence for a preferred quantification measure to conduct downstream analyses of PDX RNA-seq data. To our knowledge, this is the first comparative study of RNA-seq data quantification measures conducted on PDX models, which are known to be inherently more variable than cell line models. Our findings are consistent with what others have shown for human tumors and cell lines and add further support to the thesis that normalized counts are the best choice for the analysis of RNA-seq data across samples.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-021-02936-w.

Converting tabular data into images for deep learning with convolutional neural networks

Convolutional neural networks (CNNs) have been successfully used in many applications where important information about data is embedded in the order of features, such as speech and imaging. However, most tabular data do not assume a spatial relationship between features, and thus are unsuitable for modeling using CNNs. To meet this challenge, we develop a novel algorithm, image generator for tabular data (IGTD), to transform tabular data into images by assigning features to pixel positions so that similar features are close to each other in the image. The algorithm searches for an optimized assignment by minimizing the difference between the ranking of distances between features and the ranking of distances between their assigned pixels in the image. We apply IGTD to transform gene expression profiles of cancer cell lines (CCLs) and molecular descriptors of drugs into their respective image representations. Compared with existing transformation methods, IGTD generates compact image representations with better preservation of feature neighborhood structure. Evaluated on benchmark drug screening datasets, CNNs trained on IGTD image representations of CCLs and drugs exhibit a better performance of predicting anti-cancer drug response than both CNNs trained on alternative image representations and prediction models trained on the original tabular data.

Learning curves for drug response prediction in cancer cell lines

BACKGROUND

Motivated by the size and availability of cell line drug sensitivity data, researchers have been developing machine learning (ML) models for predicting drug response to advance cancer treatment. As drug sensitivity studies continue generating drug response data, a common question is whether the generalization performance of existing prediction models can be further improved with more training data.

METHODS

We utilize empirical learning curves for evaluating and comparing the data scaling properties of two neural networks (NNs) and two gradient boosting decision tree (GBDT) models trained on four cell line drug screening datasets. The learning curves are accurately fitted to a power law model, providing a framework for assessing the data scaling behavior of these models.

RESULTS

The curves demonstrate that no single model dominates in terms of prediction performance across all datasets and training sizes, thus suggesting that the actual shape of these curves depends on the unique pair of an ML model and a dataset. The multi-input NN (mNN), in which gene expressions of cancer cells and molecular drug descriptors are input into separate subnetworks, outperforms a single-input NN (sNN), where the cell and drug features are concatenated for the input layer. In contrast, a GBDT with hyperparameter tuning exhibits superior performance as compared with both NNs at the lower range of training set sizes for two of the tested datasets, whereas the mNN consistently performs better at the higher range of training sizes. Moreover, the trajectory of the curves suggests that increasing the sample size is expected to further improve prediction scores of both NNs. These observations demonstrate the benefit of using learning curves to evaluate prediction models, providing a broader perspective on the overall data scaling characteristics.

CONCLUSIONS

A fitted power law learning curve provides a forward-looking metric for analyzing prediction performance and can serve as a co-design tool to guide experimental biologists and computational scientists in the design of future experiments in prospective research studies.

Conservation of copy number profiles during engraftment and passaging of patient-derived cancer xenografts

Patient-derived xenografts (PDXs) are resected human tumors engrafted into mice for preclinical studies and therapeutic testing. It has been proposed that the mouse host affects tumor evolution during PDX engraftment and propagation, affecting the accuracy of PDX modeling of human cancer. Here, we exhaustively analyze copy number alterations (CNAs) in 1,451 PDX and matched patient tumor (PT) samples from 509 PDX models. CNA inferences based on DNA sequencing and microarray data displayed substantially higher resolution and dynamic range than gene expression-based inferences, and they also showed strong CNA conservation from PTs through late-passage PDXs. CNA recurrence analysis of 130 colorectal and breast PT/PDX-early/PDX-late trios confirmed high-resolution CNA retention. We observed no significant enrichment of cancer-related genes in PDX-specific CNAs across models. Moreover, CNA differences between patient and PDX tumors were comparable to variations in multiregion samples within patients. Our study demonstrates the lack of systematic copy number evolution driven by the PDX mouse host.

Ensemble transfer learning for the prediction of anti-cancer drug response

Transfer learning, which transfers patterns learned on a source dataset to a related target dataset for constructing prediction models, has been shown effective in many applications. In this paper, we investigate whether transfer learning can be used to improve the performance of anti-cancer drug response prediction models. Previous transfer learning studies for drug response prediction focused on building models to predict the response of tumor cells to a specific drug treatment. We target the more challenging task of building general prediction models that can make predictions for both new tumor cells and new drugs. Uniquely, we investigate the power of transfer learning for three drug response prediction applications including drug repurposing, precision oncology, and new drug development, through different data partition schemes in cross-validation. We extend the classic transfer learning framework through ensemble and demonstrate its general utility with three representative prediction algorithms including a gradient boosting model and two deep neural networks. The ensemble transfer learning framework is tested on benchmark in vitro drug screening datasets. The results demonstrate that our framework broadly improves the prediction performance in all three drug response prediction applications with all three prediction algorithms.

Abstract 1118: Absence of mouse-specific tumor evolution in patient-derived cancer xenografts

Patient-Derived Xenografts (PDXs) are preclinical models largely used to study tumor biology and drug response. Recent literature highlighted the possibility that growth of human tumors in a mouse microenvironment imposes a selection driving mouse-specific genetic evolution of PDXs, which may compromise their reliability as human cancer models. Conversely, independent studies observed a conservation of the genomic landscape during PDX engraftment and passaging.

We noticed that PDX genetic evolution was particularly evident in studies based on copy number aberration (CNA) inferred from gene expression data, while it was negligible when DNA-based CNA profiles were employed. Therefore, in a joint international effort of the EurOPDX and PDXNet consortia, we assembled a dataset of 37 hepatocellular and 54 gastric carcinoma tumor or PDX samples with matched RNA-based and DNA-based CNA profiles. We found that DNA-based CNA profiles invariably yield higher concordance between patient's tumor and derived PDXs than those inferred from RNA. RNA-based profiles displayed poor concordance with matched DNA-based profiles, and much lower resolution, so that they missed many focal copy number events detected by DNA-based methods. These results revealed that CNA measurements cannot be accurately estimated by expression data and that a systematic reassessment of CNA dynamics in PDXs based on DNA data is required.

To this aim, we generated CNA profiles by low-pass whole genome sequencing (WGS) of 87 colorectal and 43 breast cancer triplets, each composed of matched patient's tumor (PT) and PDX at early (PDX-early) and later (PDX-late) passage. In this way, for each tumor type, we generated three perfectly matched PT, PDX-early and PDX-late cohorts and performed CNA recurrence analysis by GISTIC in each cohort. The hypothesis was that if the mouse host induces a selective pressure capable of shaping the CNA landscape during PDX engraftment and propagation, GISTIC analysis would highlight systematic and progressive changes, from the PT to the PDX-early cohort, and then to the PDX-late cohort. Notably instead, the CNA profiles of the PT and PDX-early/late cohorts were virtually indistinguishable, with no progressive accumulation or loss of CNA during PDX passage and only minor changes not functionally related or associated to cancer-driver or actionable genes. These results were not consequence of insufficient capture of the CNA repertoire, since the GISTIC profiles recapitulated those generated by TCGA for colorectal and breast cancer. In summary, our analyses highlighted that while RNA-based CNA inferences have inadequate resolution and accuracy to study genomic evolution in PDXs, DNA-based CNA profiles confirm retention of CNAs in PTs and PDXs, excluding a systematic mouse driven selection via copy number changes. Ultimately, these results support the robustness of PDXs as preclinical models for predicting drug response.

Citation Format: Jessica Giordano, Xing Yi Woo, Anuj Srivastava, Zi-Ming Zhao, Michael W. Lloyd, Roebi de Bruijn, Yun-Suhk Suh, Francesco Galimi, Andrea Bertotti, Adam Lafferty, Alice C. O'Farrell, Elodie Modave, Diether Lambrechts, Petra ter Brugge, Violeta Serra, Elisabetta Marangoni, Rania El Botty, Jong-Il Kim, Han-Kwang Yang, Charles Lee, Dennis A. Dean, Brandi Davis-Dusenbery, Yvonne A. Evrard, James H. Doroshow, Alana L. Welm, Bryan E. Welm, Michael T. Lewis, Bingliang Fang, Jack Roth, Funda Meric-Bernstam, Meenhard Herlyn, Michael Davies, Li Ding, Shunqiang Li, Ramaswamy Govindan, Jeffrey A. Moscow, Carol J. Bult, Claudio Isella, Livio Trusolino, Annette T. Byrne, Jos Jonkers, Jeffrey H. Chuang, Enzo Medico, EurOPDX consortium & PDXNET consortium. Absence of mouse-specific tumor evolution in patient-derived cancer xenografts [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 1118.

Abstract 3916: Patient-derived organoid and cell culture models from the NCI Patient-Derived Models Repository (NCI PDMR) preserve genomic stability and heterogeneity of patient tumor specimens

Background: The National Cancer Institute (NCI) has developed a Patient-Derived Models Repository (PDMR; https://pdmr.cancer.gov) of preclinical models including patient-derived xenografts (PDX), organoids (PDOrg) and patient-derived cell cultures (PDC). Extensive clinical annotation and genomic datasets are available for these preclinical models. However, it is unclear if the molecular profiles of the corresponding patient tumors are stably propagated in these models. We have previously demonstrated that PDX models from the NCI PDMR faithfully represent the patient tumors both in terms of genomic stability and tumor heterogeneity. Here, we conduct an in-depth investigation of genomic representation of patient tumors in the PDOrgs and PDCs.

Methods: PDOrgs (n=64) and PDCs (n=94) were established from tumor fragments (i.e., initiator specimens) obtained either from patient specimens or from PDX specimens of early passage. For some models (n=19), both PDOrgs and PDCs were generated from the same tumor tissue; in fewer cases (n=4), PDCs were established from organoids derived from patient specimens. Whole Exome Sequencing and RNA-Seq were performed on all PDCs and PDOrgs, and data were compared with patient specimens or early passage PDXs.

Results: A majority of the PDOrgs and PDCs have stably inherited the genome of the corresponding patient specimens based on the following observations: (1) &gt;87% of PDOrgs and PDCs maintained similar copy number alteration profiles compared with the initiator specimens of the preclinical model; (2) the variant allele frequency (VAF) of clinically relevant mutations remained consistent between the PDOrgs, PDCs, and the initiator specimens, with none of the PDCs or PDOrgs deviating by &gt;15% VAF; and (3) clinically relevant biomarkers (e.g., MSI, LOH, mutational signatures etc.) are concordant amongst the PDOrgs, PDCs, and the initiator specimens. We observed that the majority of SNVs and indels present in the initiator specimens were also found in the PDOrgs and PDCs, suggesting almost all the tumor heterogeneity was preserved in these preclinical models.

Conclusions: This large and histologically diverse set of PDOrgs and PDCs from the NCI PDMR exhibited genomic stability and faithfully represented the tumor heterogeneity observed in corresponding patient specimens. These preclinical models thus represent a valuable resource for researchers interested in pre-clinical drug or other studies.

Citation Format: Biswajit Das, Yvonne A. Evrard, Li Chen, Rajesh Patidar, Tomas Vilimas, Justine N. McCutcheon, Amanda L. Peach, Nikitha V. Nair, Thomas D. Forbes, Brandie A. Fullmer, Anna J. Lee Fong, Luis E. Romero, Alyssa K. Chapman, Kelsey A. Conley, Robin D. Harrington, Shahanawaz S. Jiwani, Peng Wang, Michelle M. Gottholm-Ahalt, Erin N. Cantu, Gloryvee Rivera, Lindsay M. Dutko, Kelly M. Benauer, Vishnuprabha R. Kannan, Carrie A. Bonomi, Kelly M. Dougherty, Joseph P. Geraghty, Marion V. Gibson, Savanna S. Styers, Abigail J. Walke, Jenna E. Moyer, Anna Wade, Mariah L. Baldwin, Kaitlyn A. Arthur, Kevin J. Plater, Luke Stockwin, Matthew R. Murphy, Michael E. Mullendore, Dianne L. Newton, Melinda G. Hollingshead, Chris A. Karlovich, Paul M. Williams, James H. Doroshow. Patient-derived organoid and cell culture models from the NCI Patient-Derived Models Repository (NCI PDMR) preserve genomic stability and heterogeneity of patient tumor specimens [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3916.

Abstract 3554: Genomic landscape of acquired uniparental disomy in NCI PDMR patient derived xenograft models

Background: Acquired Uniparental Disomy (aUPD) is relatively common in cancer. Occurrence of aUPD is more frequent in some tumor histologies (e.g., serous ovarian, colorectal) and may be relevant for choice of therapy. The Patient-Derived Models Repository (PDMR; https://pdmr.cancer.gov) developed by The National Cancer Institute (NCI) includes patient-derived xenograft (PDX) models from multiple tumor histologies with different passages and lineages. The associated clinical annotation and genomic data make it possible to assess the prevalence of aUPD in the PDMR cohort and the stability of aUPD in different passages and lineages within a PDX model.

Methods: High tumor purity in the PDX specimens (after removal of mouse reads representing the stroma) enabled highly accurate assessment of loss of heterozygosity (LOH). Variants called by GATK Haplotype caller from whole exome sequencing (WES) data were used to identify segments of homozygosity using BCFtools/RoH (runs of homozygosity). The RoH segments were then intersected with the bed file for chromosome arms to get %LOH at the arm level. If %LOH on a chromosome arm was &gt;90%, we considered the sample to have aUPD at the arm level. WES was also used to look for associations between DNA damage repair (DDR) pathway alterations and aUPD.

Results: We made the following observations: a) aUPD was observed most frequently in chr18q (75/427, 17.6%) and chr3p (69/427, 16%) of PDX models; b) aUPD was observed more frequently in certain tumor histologies, e.g., clear cell renal cell carcinoma (6/8), small cell lung cancer (3/4) and non-small cell lung cancer (25/38); c) extensive aUPD was observed in 4 PDMR models (&gt;50% of evaluated chromosome arms in these models have aUPD); d) aUPD was not observed in some tumor histologies, i.e., synovial sarcoma, uterine endometrioid carcinoma; e) in the vast majority of PDMR models (&gt;90%), aUPD is maintained faithfully across lineages and through multiple passaging; f) subclonal aUPD events were observed in some models across different lineages; g) significant enrichment of double strand DNA break repair (DSBR) pathway alterations was observed in PDMR models without aUPD (p=0.0007, Fisher's exact test) suggesting defects in DSBR are not associated with aUPD; and h) aUPD was rarely observed in MSI-high models (1/30) suggesting mutual exclusivity of mismatch repair (MMR) pathway defects and aUPD.

Conclusion: We observed a relatively high frequency of UPD in the PDMR models (at least 1 arm of a chromosome). UPD was more frequently observed in specific chromosomal arms. The frequency of aUPD was higher in some tumor histologies and absent in others. aUPD was stably maintained across passages and lineages, although some heterogeneity was observed. Our data suggest aUPD is not associated with defects in DSBR and MMR pathways. Preclinical drug studies using NCI PDMR models may suggest appropriate therapeutic options for cancers with aUPD.

Citation Format: Rajesh Patidar, Li Chen, Chris A. Karlovich, Biswajit Das, Yvonne A. Evrard, Tomas Vilimas, Justine N. McCutcheon, Amanda L. Peach, Nikitha V. Nair, Thomas D. Forbes, Brandie A. Fullmer, Anna J. Lee Fong, Luis E. Romero, Alyssa K. Chapman, Kelsey A. Conley, Robin D. Harrington, Shahanawaz S. Jiwani, Peng Wang, Michelle M. Gottholm Ahalt, Erin N. Cantu, Gloryvee Rivera, Lindsay M. Dutko, Kelly M. Benauer, Vishnuprabha R. Kannan, Suzanne D. Borgel, John P. Carter, Jesse M. Stottlemyer, Tiffanie L. Miner, Devynn R. Breen, Emily T. Delaney, Chelsea A. McGlynn, Candace N. Mallow, Marianne Radzyminski, Shannon N. Uzelac, Sergio Y. Alcoser, Tara L. Grinnage-Pulley, Michelle A. Eugeni, Dianne L. Newton, Melinda G. Hollingshead, Paul M. Williams, James H. Doroshow. Genomic landscape of acquired uniparental disomy in NCI PDMR patient derived xenograft models [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3554.

Abstract 5056: Quality control efforts in a large-scale, preclinical trial of rare cancer PDXs by the National Cancer Institute's patient-derived models repository (NCI PDMR)

The National Cancer Institute's Patient-Derived Models Repository (NCI PDMR; https://pdmr.cancer.gov) is performing a large-scale multi-year preclinical study with 39 PDX models of rare cancers (mesothelioma, MPNST, osteosarcoma, Merkel cell carcinoma, etc) treated with 56 novel therapeutic combinations in an exploratory, n-of-4 arm, study design. Combinations that show promising responses (e.g., regression or durable inhibition of tumor growth) will be repeated along with the single agent arms to determine if the response is driven by the combination or only one of the agents. In order to do this in a timely fashion, relatively speaking, the PDX tumors are serially passaged and each passage is treated with a set of 8 combinations plus relevant vehicle control(s) while in parallel enough PDXs are retained to be expanded for the next passage and drug set. Every serial passage undergoes several quality control assessments that serve as go/no-go criteria including pathology assessment, human:mouse DNA content assessment, and low pass whole genome sequencing to determine the average fraction of genome changed compared to the original donor material. If there is a QC failure, the PDX model is restarted from early passage cryo-material (passage 1-2). An additional quality control effort is to bookend the combination studies with the first set of agents to see if tumor response is similar across passages. To date, most of the models have demonstrated a high degree of stability, though a couple of models have moved toward murine content and have been restarted from early passage material so all drug combinations can be tested. DNA and RNA are retained from all passages so a full NGS evaluation can be performed at a later date. This effort has been ongoing for over a year and the first bookend studies are beginning to be tested to determine if response at first and last passage of the study are consistent with each other, given the constraints of the inherent heterogeneity of the models themselves. Single agent studies of drug combinations that demonstrated a response in 30%-50% of the models tested are also underway to determine which combinations have a more than additive effect compared to the single agents. Promising combinations will be moved forward to early phase clinical trials for these rare cancers.

Funded by NCI Contract No. HHSN261200800001E

Citation Format: Yvonne A. Evrard, Biswajit Das, Sergio Y. Alcoser, Suzanne Borgel, Devynn Breen, John Carter, Tiffanie Chase, Alice Chen, Lily Chen, Kristen Cooley, Emily Delaney, Raymond Divelbiss, Lyndsay Dutko, Thomas Forbes, Kyle Georgius, Michelle Gottholm-Ahalt, Tara Grinnage-Pulley, Sierra Hoffman, Chris Karlovich, Shahanawaz Jiwani, Justine Mills, Malorie Morris, Michael Mullendore, Dianne Newton, Rajesh Patidar, Gloryvee Rivera, Howard Stotler, Jesse Stottlemyer, Savanna Styers, Debbie Trail, Shannon Uzelac, Thomas Vilimas, Abigail Walke, Thomas Walsh, Nicole Walters, Peng Wang, P. Mickey Williams, Melinda Hollingshead, James H. Doroshow. Quality control efforts in a large-scale, preclinical trial of rare cancer PDXs by the National Cancer Institute's patient-derived models repository (NCI PDMR) [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5056.

Systematic Establishment of Robustness and Standards in Patient-Derived Xenograft Experiments and Analysis

Patient-derived xenografts (PDX) are tumor-in-mouse models for cancer. PDX collections, such as the NCI PDXNet, are powerful resources for preclinical therapeutic testing. However, variations in experimental and analysis procedures have limited interpretability. To determine the robustness of PDX studies, the PDXNet tested temozolomide drug response for three prevalidated PDX models (sensitive, resistant, and intermediate) across four blinded PDX Development and Trial Centers using independently selected standard operating procedures. Each PDTC was able to correctly identify the sensitive, resistant, and intermediate models, and statistical evaluations were concordant across all groups. We also developed and benchmarked optimized PDX informatics pipelines, and these yielded robust assessments across xenograft biological replicates. These studies show that PDX drug responses and sequence results are reproducible across diverse experimental protocols. In addition, we share the range of experimental procedures that maintained robustness, as well as standardized cloud-based workflows for PDX exome-sequencing and RNA-sequencing analyses and for evaluating growth. SIGNIFICANCE: The PDXNet Consortium shows that PDX drug responses and sequencing results are reproducible across diverse experimental protocols, establishing the potential for multisite preclinical studies to translate into clinical trials.

Genomic characterization of preclinical models derived from primary and metastatic sites from rapid autopsy patients in PDMR

e13506

Background: The National Cancer Institute has developed a repository of preclinical models [Patient-Derived Models Repository (NCI PDMR, https://pdmr.cancer.gov )] including patient derived xenografts (PDXs), organoids (PDOrgs) and in vitro tumor cultures (PDCs) from patients with solid tumor cancer histologies. A subset of these preclinical models is derived from post-mortem collections from rapid autopsies representing the end point in disease progression. Clinical annotations and genomic datasets associated with these models provide a unique opportunity to study tumor evolution, mechanistic insights into the metastatic process, and treatment resistance. Methods: To date, 43 PDXs, 21 PDCs, and 23 PDOrgs using rapid autopsy specimens from 8 primary and 35 metastatic sites of 18 patients have been developed by the Biological Testing Branch (DTP, DCTD, NCI Frederick, MD) for the PDMR. Whole exome (WES) and total transcriptome (RNASeq) data were processed to generate mutation, copy number alteration (CNA) and gene expression data. Multi-model lineage trees were reconstructed based on putative somatic variants for all the models derived from all patients. The fraction of the genome affected by CNA was compared both within and across PDX models. Results: Most of the rapid autopsy PDX models (32/43) are derived from pancreatic adenocarcinoma (PAAD) patients (13/18), with metastatic specimens originating from sites including liver, colon, omentum, and lung. Driver mutations are present in all preclinical model specimens derived from the same patient. For instance, KRAS p.G12D is present in all patient-derived model specimens derived from PAAD patient 521955. The fraction of the genome affected by CNA remains stable within a PDX model across passages (n = 24, mean = 6.39%, sd = 5.90%). However, we found that this increased when comparing PDX models derived from metastatic sites versus the primary site (n = 19, mean = 16.92%, sd = 10.46%). This indicates presence of tumor heterogeneity between metastatic and primary sites. The lineage tree for models from patient 521955 indicates that one liver metastasis has a unique seeding event compared to the other 4 metastatic sites. Unsupervised clustering analysis on gene expression data also confirms the observed tumor site relationships. Conclusions: Our data demonstrate the potential use of these preclinical models available from the NCI PDMR. These models provide a unique resource for preclinical studies in tumor evolution, metastatic spread mediators, and drug resistance.

AI Meets Exascale Computing: Advancing Cancer Research With Large-Scale High Performance Computing

The application of data science in cancer research has been boosted by major advances in three primary areas: (1) Data: diversity, amount, and availability of biomedical data; (2) Advances in Artificial Intelligence (AI) and Machine Learning (ML) algorithms that enable learning from complex, large-scale data; and (3) Advances in computer architectures allowing unprecedented acceleration of simulation and machine learning algorithms. These advances help build in silico ML models that can provide transformative insights from data including: molecular dynamics simulations, next-generation sequencing, omics, imaging, and unstructured clinical text documents. Unique challenges persist, however, in building ML models related to cancer, including: (1) access, sharing, labeling, and integration of multimodal and multi-institutional data across different cancer types; (2) developing AI models for cancer research capable of scaling on next generation high performance computers; and (3) assessing robustness and reliability in the AI models. In this paper, we review the National Cancer Institute (NCI) -Department of Energy (DOE) collaboration, Joint Design of Advanced Computing Solutions for Cancer (JDACS4C), a multi-institution collaborative effort focused on advancing computing and data technologies to accelerate cancer research on three levels: molecular, cellular, and population. This collaboration integrates various types of generated data, pre-exascale compute resources, and advances in ML models to increase understanding of basic cancer biology, identify promising new treatment options, predict outcomes, and eventually prescribe specialized treatments for patients with cancer.

Abstract 1074: The PDX Data Commons and Coordinating Center (PDCCC) for PDXNet in support of preclinical research

Patient-Derived Xenografts (PDX) are proven models to study novel drugs or drug combinations and test hypothesis in preclinical studies. The overarching goal of the PDXNet is to coordinate the development of appropriate PDX models and methods for preclinical drug testing to advance CTEP clinical development of new cancer agents.

The PDXNet is an NCI-funded consortium of six PDX Development and Trial Centers (PDTCs) and one PDCCC. Four PDTCs are responsible for developing PDXs and executing specific preclinical trials focused on cancer types including breast cancer, melanoma, and lung cancer. The other two recently awarded centers are specifically focused on minority PDX models and preclinical trials. Besides the PDTCs, the NCI Patient-Derived Models Repository (PDMR) at the Frederick National Laboratory for Cancer Research (FNLCR) is also providing models and data to the PDXNet. The PDCCC is responsible for coordination and developing standards for PDX generation as well as data analysis and metadata harmonization. The PDX Data Commons is built on top of existing NCI resources, leveraging the Cancer Genomics Cloud maintained by Seven Bridges Genomics, where PDXNet data is co-located with TCGA and other large-scale datasets. The PDCCC is co-led by experts from the Jackson Laboratory, providing scientific leadership in xenograft methods and cancer biology to ensure the promulgation of standards that are well-suited for the PDX community.

A new portal has been set up at https://www.pdxnetwork.org/ to serve as the point of access to PDXNet resources. In addition, we established ongoing network-wide meetings to facilitate knowledge exchange, held PDXNet portal trainings, and set up working groups to tackle specific challenges. For instance, the Data Ontology working group has been working towards building a common data ontology model specifically for PDX datasets. We are in the process of annotating the very first dataset using this new ontology on the PDXNet portal. Also, the Workflows working group has been working on building and benchmarking various RNA-seq and whole exome sequencing analysis workflows to standardize data processing between PDXNet grantees and create a harmonized PDXNet dataset. These PDX models and the accompanying data will be opened to the community for data mining and/or preclinical research.

The PDXNet is a strong step toward building a consensus around PDX models, so that the power for discovery can be expanded by making multi-institutional PDX cohorts a reality. As the coordination center, we are also working closely with the EuroPDX project to exchange standards and knowledge to support the PDX community with a set of standards going forward. The PDCCC is a central part of this process to systematically capture and analyze the variables most influential to PDX models and share protocols and tools to make PDXs an interchangeable research currency for preclinical discovery.

Citation Format: Jacqueline Rosains, Anuj Srivastava, Wingyi Woo, Vishal Sarsani, ZiMing Zhao, Javad Noorbakhsh, Ogan D. Abaan, Christian Frech, Jack DiGiovanna, Ryan Jeon, Steve Neuhauser, Peter Robinson, Yvonne A. Evrard, Carol Bult, Jeffrey A. Moscow, Brandi Davis-Dusenbery, Jeffrey H. Chuang. The PDX Data Commons and Coordinating Center (PDCCC) for PDXNet in support of preclinical research [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1074.

Abstract 4524: Comparison of PDX, PDC, and PDOrg models from the National Cancer Institute’s Patient-Derived Models Repository (PDMR)

The National Cancer Institute (NCI) has developed a Patient-Derived Models Repository (PDMR) comprised of quality-controlled, early-passage, clinically-annotated patient-derived tumor xenografts (PDXs), in vitro tumor cell cultures (PDCs), cancer associated fibroblasts (CAFs), and patient-derived organoids (PDOrg). NCI has focused on generating models to complement existing PDX collections and address unmet needs in the preclinical model space. These models are offered to the extramural community for research use (https://pdmr.cancer.gov), along with clinical annotation and molecular information (whole exome sequence, gene expression using RNASeq), via a publicly accessible database. Currently, over 200 PDX models, 50 PDC models, and 100 CAF models are available for distribution to the US research community. Approximately 50 PDOrg models will be released in early 2019. As part of its rare cancer initiative, the NCI is also targeting the collection of infrequently-observed tumor histologies to advance both biological investigations and drug development efforts for under-studied malignancies. Comparison of matched models, models where more than one model type are available (e.g., PDX and PDC), demonstrate a high degree of concordance across the model types. Genetic stability across the models is assessed using multiple criteria including genetic assessment of CNVs and presence of driver mutations. Optimal CNV assessment uses whole exome sequence data corrected for cellularity in the patient specimen using germline reads and corrected for cellularity in the PDX specimens by subtraction of the mouse reads. Histomorphologic comparison of PDXs and cell line xenografts (CLX) generated from in vitro PDCs and PDOrgs also overall show a high degree of concordance, though loss of features and dedifferentiation can be observed in some models. Overall these models demonstrate a high degree of conservation at the genetic and pathologic level when compared to the patient tumor. These models can provide researchers the ability to perform high- or mid-throughput screening in 2D or 3D culture followed by targeted selection of PDX models for in vivo studies. Funded by NCI Contract No. HHSN261200800001E

Citation Format: Yvonne A. Evrard, Dianne Newton, Biswajit Das, Sergio Y. Alcoser, Kaitlyn Arthur, Mariah Baldwin, Carrie Bonomi, Suzanne Borgel, John Carter, Tiffany Chase, Alice Chen, Lily Chen, Nikki E. Craig, Vivekananda Datta, Emily Delaney, Raymond Divelbiss, Kelly Dougherty, Thomas Forbes, Kyle Georgius, Joe Geraghty, Marion Gibson, Michelle M. Gottholm-Ahalt, Tara Grinnage-Pulley, Kelly Hedger, Sierra Hoffman, Chris Karlovich, Wiem Lassoued, Shahanawaz Jiwani, Candace Mallow, Chelsea McGlynn, Mallorie Morris, Jenna Moyer, Mike Mullendore, Matt Murphy, Rajesh Patidar, Kevin Plater, Marianne Radzyminski, Nicki Scott, Luke H. Stockwin, Howard Stotler, Jesse Stottlemyer, Savanna Styers, Debbie Trail, Tomas Vilimas, Anna Wade, Abigail Walke, Thomas Walsh, P. Mickey Williams, Melinda G. Hollingshead, James H. Doroshow. Comparison of PDX, PDC, and PDOrg models from the National Cancer Institute’s Patient-Derived Models Repository (PDMR) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 4524.

Abstract 1056: Xenograft-associated B cell lymphoproliferative disease (XABLD) as a surrogate model to study Epstein-Barr virus (EBV) driven B cell Diseases

Background: Patient-derived tumor xenografts (PDX) are powerful tools to study cancer biology, cancer genomics and developmental therapeutics. A common problem in the development of PDX models is proliferation of atypical lymphocytes at the implantation site, which often overtake or limit the growth of the original tumor. This atypical lymphocyte proliferation has been described as XABLD in our PDX models. In this study, we characterized XABLD cases by morphology, immunophenotyping and genomic profiling. We hypothesize that XABLD tumors are morphologically and phenotypically similar to EBV-driven post-transplant lymphoproliferative disease (PTLD) and diffuse large B cell lymphoma (DLBCL). XABLD is a surrogate model to study EBV-driven PTLD and DLBCL.

Materials and Methods: Models were generated from patient tissue collected under NCI Tissue Procurement Protocol (clinicaltrials.gov: NCT00900198) and CIRB Tissue Procurement Protocol 9846 for development of models for NCI’s Patient-Derived Models Repository (https://pdmr.cancer.gov). Specimens were implanted subcutaneously in NOD/SCID/IL2Rg null (NSG) mice and animal health was monitored throughout the study. Tumors in mice with suspected XABLD were harvested and reviewed by histology and immunohistochemical analysis for CD45, B and T cell markers, EBV status, B-cell clonality assay. All samples were also classified by the Lymph2Cx NanoString cell of origin assay and transcriptome profiling.

Results: XABLD cases were found to originate from both solid tumor and circulating tumor cell implants. XABLD is a rapidly growing tumor positive for CD45, CD20, and LMP1 stains, 36 of 42 cases are strongly positive for PD-L1 stain. 39 of 42 cases exhibited an activated B cell (ABC) phenotype with evidence of elevated NF-kB signaling. Most cases were monoclonal for IGK/IGH and contained high numbers of tumor infiltrating CD8-positive T-cells with associated high mRNA expression of activated T cell markers.

Conclusion: The clinical presentation, morphology and molecular characteristics of XABLD cases were similar to EBV-driven DLBCL. As the XABLD models exhibited frequent PD-L1 expression and marked infiltration of CD8-positive T cells, they may be useful for in vitro evaluation of checkpoint inhibitor response and T cell antitumor activity.

Grant Support: This project has been funded in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Citation Format: Tomas Vilimas, Gloryvee Rivera, Brandie Fullmer, Wiem Lassoued, Lindsay Dutko, Amanda Peach, Corinne Camalier, Li Chen, Rajesh Patidar, Suzanne Borgel, John Carter, Howard Stotler, Raymond Divelbiss, Jesse Stottlemyer, Michelle M. Gottholm-Ahalt, Michelle Crespo-Eugeni, Sean McDermott, William Jacob, Liqiang Xi, Pallavi Galera, Yvonne A. Evrard, Melinda G. Hollingshead, Elaine S. Jaffe, Mark Raffeld, Biswajit Das, Chris Karlovich, Vivekananda Datta, James H. Doroshow, P. Mickey Williams. Xenograft-associated B cell lymphoproliferative disease (XABLD) as a surrogate model to study Epstein-Barr virus (EBV) driven B cell Diseases [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1056.

Integrative analyses of signaling and DNA damage repair pathways in patient-derived xenograft (PDX) models from NCI’s patient-derived models repository (PDMR)

3111

Background: Patient-derived xenografts (PDXs) are increasingly being used in translational cancer research for preclinical drug efficacy studies. The National Cancer Institute (NCI) has developed a Patient-Derived Models Repository (NCI PDMR; pdmr.cancer.gov ) of PDXs with clinical annotation, proteomics, and comprehensive genomic datasets to facilitate these studies. Here, we present an integrative genomic, transcriptomic, and proteomic analysis of critical signaling and DNA damage repair pathways in these PDX models, which represent 9 common and multiple rare tumor histologies. Methods: 304 PDX models from 294 patients were established from various solid tumor histologies from patients with primary or metastatic cancer. Whole Exome Sequencing, RNA-Seq and Reverse Phase Protein Array (RPPA) were performed on 2-9 PDXs per model across multiple passages. An integrative workflow was applied on multiple data sets to detect pathway activation. Results: We profiled 10 signaling and 5 DNA repair pathways in the PDMR dataset. We observed that: (i) a large fraction (40%) of PDX models have at least 1 targetable mutation in the RTK/RAS and/or PIK3CA pathways; (ii) 131 models (45%) have putative driver and oncogenic mutations and copy number variants (CNVs) in the WNT, TGFRb , NRF2 and NOTCH pathways. In addition, 17% of PDX models have targetable mutations in DNA damage repair pathways and 20 PDMR models have a DNA mismatch repair defect (MSI-H). We confirmed activation of the signaling pathways in a subset of PDX models by pathway enrichment analysis on gene expression data from RNASeq and phosphoprotein-specific antibody binding data from RPPA. Activation of DNA repair processes was confirmed by enrichment of relevant mutational signatures and loss of heterozygosity in these PDX models. Conclusions: Genomic analysis of NCI PDMR models revealed that a large fraction have clinically relevant somatic alterations in key signaling and DNA damage repair pathways. Further integrative analyses with matched transcriptomic and proteomic profiles confirmed pathway activation in a subset of these models, which may prioritize them for preclinical drug studies.

Measuring phospho-MET by multiplex immunofluorescence to aid in selection of patients with MET activation in tumors

3131

Background: Currently, patient selection criteria for clinical testing of MET inhibitors are limited. Robust studies selecting patients based on MET protein expression, MET gene amplification, or mutations have not met their efficacy goals. Development of microscopy-based assays to quantify levels of phospho-MET (pMET) in tumors has been hampered by poor antibody specificity. Here, we present the development and validation of a robust, highly specific multiplex immunofluorescence assay (IFA) that measures pY1235-MET and total MET in tumor tissue. Methods: This assay utilizes antibodies to pY1235-MET (NCI-23111), total MET (D1C2), and plasma membrane (PM) marker Na+/K+-ATPase, each conjugated to a different Alexa Fluor dye. We used tumor tissue from crizotinib-treated SNU5 xenograft models to demonstrate pY1235-MET assay fitness-for-purpose and cross-platform assay concordance with our validated pMET ELISA. In addition, this IFA was validated by phospho-peptide competition using custom tissue microarrays (TMA) derived from patients with colorectal carcinoma (CRC). Finally, we developed quantitative algorithms to assess pY1235 MET levels in the plasma membrane and nucleus using PM and DAPI masks, respectively. Patient-derived xenograft models (PDX) were obtained from NCI’s Patient-Derived Models Repository (www.pdmr.cancer.gov). Results: The prevalence of high pY1235-MET expression in CRC patient specimens was greater than expected; of the 64 TMA cores evaluated, 29 (45%) and 19 (29%) had high pY1235-MET and total MET levels, respectively, as defined by mean marker area of ≥ 30 μm2/cell. To address the potential utility of pY1235-MET as a diagnostic biomarker, we examined 15 CRC PDX models by pMET ELISA and IFA. Two CRC tumor models were positive for pY1235-MET expression in both assays. The pY1235-MET IFA results and gene expression data were used to select PDX models for ongoing preclinical trials of potent MET inhibitors. Conclusions: This novel pY1235-MET IFA will enable clinicians to address the utility of activated MET as a biomarker for patient selection and/or prediction of response in clinical trials of MET inhibitors. Funded by NCI Contract No. HHSN261200800001E.

Predicting tumor cell line response to drug pairs with deep learning

BACKGROUND

The National Cancer Institute drug pair screening effort against 60 well-characterized human tumor cell lines (NCI-60) presents an unprecedented resource for modeling combinational drug activity.

RESULTS

We present a computational model for predicting cell line response to a subset of drug pairs in the NCI-ALMANAC database. Based on residual neural networks for encoding features as well as predicting tumor growth, our model explains 94% of the response variance. While our best result is achieved with a combination of molecular feature types (gene expression, microRNA and proteome), we show that most of the predictive power comes from drug descriptors. To further demonstrate value in detecting anticancer therapy, we rank the drug pairs for each cell line based on model predicted combination effect and recover 80% of the top pairs with enhanced activity.

CONCLUSIONS

We present promising results in applying deep learning to predicting combinational drug response. Our feature analysis indicates screening data involving more cell lines are needed for the models to make better use of molecular features.

Abstract 1038: Xenograft-associated B cell lymphoproliferative disease as a surrogate model to study Epstein-Barr virus (EBV) driven lymphoma of the elderly

Background: Patient-derived tumor xenografts (PDX) are powerful tools to study cancer biology, cancer genomics and developmental therapeutics. A common problem in the development of PDX models is proliferation of atypical lymphocytes at the implant site, which often overtake or limit the growth of the original tumor. This atypical proliferation has been described as Xenograft-Associated B cell Lymphoproliferative Disease (XABLD) in our PDX models. In this study, we characterized XABLD cases by morphology, immunophenotyping and genomic profiling. We hypothesize that XABLD tumors are morphologically and phenotypically similar to EBV-driven lymphoma of the elderly and may function as a surrogate model for that lymphoma. Materials and Methods: Models were generated from patient tissue collected under NCI Tissue Procurement Protocol (clincialtrials.gov: NCT00900198) and CIRB Tissue Procurement Protocol 9846 for development of models for NCI's Patient-Derived Models Repository (https://pdmr.cancer.gov). Specimens were implanted subcutaneously in NOD/SCID/IL2Rg null (NSG) mice and animal health was monitored throughout the study. Tumors in mice with suspected XABLD were harvested and reviewed by histology and immunohistochemical analysis for CD45, B and T cell markers and EBV status. All samples in this study were classified by the Lymph2Cx NanoString cell of origin assay and transcriptome profiling. Results: XABLD-associated mice had rapidly growing CD45-positive tumors at the implantation site. Histopathological features were consistent with EBV-driven diffuse large B-cell lymphoma (DLBCL) primarily of polymorphous subtype. All XABLD specimens were diffusely positive for CD20 and EBNA, and most cases contained tumor infiltrating CD8-positive T-cells. Out of 42 cases, 36 were PD-L1-positive and 26 were PD-1-positive by IHC. 39 cases exhibited an activated B cell (ABC) phenotype, which is predominant in EBV-positive DLBCL. Conclusion: XABLD development has been seen across multiple patient histologies from both solid tumor and circulating tumor cells tissues of origin. The clinical presentation, morphology and molecular characteristics of XABLD cases were similar to EBV-driven DLBCL. As DLBCL is an aggressive disease with limited treatment options, our early-passage XABLD models may be useful in the preclinical evaluation of new therapies for EBV-positive DLBCL. Grant Support: This project has been funded in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Citation Format: Tomas Vilimas, Gloryvee Rivera, Brandie Fullmer, Wiem Lassoued, Lindsay Dutko, William Walsh, Amanda Peach, Corinne Camalier, Li Chen, Rajesh Patidar, Suzanne Borgel, John Carter, Howard Stotler, Raymond Divelbiss, Jesse Stottlemyer, Margaret Defreytas, Michelle M. Gottholm-Ahalt, Michelle A. Crespo-Eugeni, Sean McDermott, Yvonne A. Evrard, Melinda G. Hollingshead, Biswajit Das, Chris Karlovich, Vivekananda Datta, James H. Doroshow, P. Mickey Williams. Xenograft-associated B cell lymphoproliferative disease as a surrogate model to study Epstein-Barr virus (EBV) driven lymphoma of the elderly [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1038.

Abstract IA21: NCI's Patient-Derived Models Repository: Generating models from racial and ethnic minorities

The National Cancer Institute (NCI) has developed a Patient-Derived Models Repository (PDMR) comprising quality-controlled, early-passage, clinically annotated patient-derived xenografts (PDXs) to serve as a resource for public-private partnerships and for academic drug discovery efforts. The first 100 models were released in May 2017, are clinically annotated with pathologic and molecular information available in a publicly accessible database, and are available to the extramural community for research use (pdmr.cancer.gov). The PDMR was established by NCI at the Frederick National Laboratory for Cancer Research (FNLCR) in direct response to discussions with academia and industry; the oncology community's highest-priority need was identified as better preclinical models that more faithfully reflect the patient's tumor and are associated with the patient's treatment history. NCI has focused on collecting specimens from patients with cancers that are under-represented in many other PDX collections such as head and neck, pancreatic, bladder and gynecologic cancers, melanomas, and sarcomas. Recently, NCI has also increased its focus on generation of PDXs from racial and ethnic minorities through new funding opportunities. The initial goal is to have 20-25% of the models in the PDMR from racial and ethnic minorities in diagnoses that are most pertinent to cancer health disparities. The overall goal of NCI is to create a long-term home for at least 1,000 models such that sufficient biologic, clinical, and population diversity is represented to allow researchers to ask questions such as: what is the impact of tumor heterogeneity on target qualification or clinical response; do PDXs more faithfully represent the human tumor for pharmacodynamic assay and predictive marker development; or can an adequately powered preclinical PDX clinical trial lead to better evaluation of therapies for future clinical use?

Grant Support: This project has been funded in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. government.

Citation Format: Yvonne A. Evrard. NCI's Patient-Derived Models Repository: Generating models from racial and ethnic minorities [abstract]. In: Proceedings of the Tenth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2017 Sep 25-28; Atlanta, GA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2018;27(7 Suppl):Abstract nr IA21.

Abstract 1029: The PDX Data Commons and Coordinating Center (PDCCC) for PDXNet

Patient-Derived Xenografts (PDX) are powerful models to study tumors' drug-response in the context of personalized medicine. In the PDX model settings, by virtue of expanding the patient's tumor sample, testing multiple drug or drug-combinations can be executed rapidly and has no ethical limitations. However, there are major issues around standards that need to be addressed to make these models widely accessible and usable.

The overarching goal of the PDXNet is to coordinate the development of appropriate PDX models and methods for preclinical drug testing to advance CTEP clinical development of new cancer agents. In an effort to standardize protocols for PDX generation as well as data analysis and metadata harmonization, we are building a data storage, sharing, and analysis platform that harmonizes PDXNet data with other large datasets and analysis workflows. The PDX Data Commons is built on top of existing NCI resources, leveraging the Cancer Genomics Cloud maintained by Seven Bridges Genomics, where PDXNet data is co-located with TCGA and other large-scale datasets. The PDCCC is co-led by experts from The Jackson Laboratory, providing scientific leadership in xenograft methods and cancer biology to ensure the promulgation of standards that are well-suited for the PDX community. In addition, the PDCCC is responsible for establishing studies to identify best-practices for PDX data analysis and metadata schemas. The data collected as part of the PDXNet is currently stored on the PDXNet portal that has a query interface for identifying models for pre-clinical trials. Simultaneously, we administer training activities and research pilots to build synergies within the PDXNet, enhancing the ability of the PDXNet to develop clinical trials from PDX studies.

In PDXNet, besides the PDCCC, there are 4 PDX Development and Trial Centers (PDTCs) responsible for executing specific pre-clinical trials focused around cancer types including breast cancer, melanoma, and lung cancer. Data generated by the PDTCs will be hosted by the PDCCC, and metadata will be collected based on schemas developed by the network for systematic ontological analysis. These PDX models, in coordination with the NCI Patient-Derived Models Repository (PDMR) at the Frederick National Laboratory for Cancer Research (FNLCR) will be shared with the broader community. In addition, PDTC's will collaborate with non-PDXNet investigators for PDX studies through an administrative supplement program supported by the NCI.

The PDXNet is a strong step toward building a consensus around PDX models, so that the power for discovery can be expanded by making multi-institutional PDX cohorts a reality. The PDCCC is a central part of this process to systematically capture and analyze the variables most influential to PDX models and share protocols and tools to make PDXs an interchangeable research currency for pre-clinical discovery.

Citation Format: Anurag Sethi, Anuj Srivastava, Xingyi Woo, Vishal Sarsani, Ziming Zhao, Javad Noorbakhsh, Christian French, Jack DiGiovanna, Ogan D. Abaan, Steve Neuhauser, Peter Robinson, Yvonne A. Evrard, Carol J. Bult, Jeffrey A. Moscow, Brandi Davis-Dusenbery, Jeffrey H. Chuang. The PDX Data Commons and Coordinating Center (PDCCC) for PDXNet [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1029.

PDX-MI: Minimal Information for Patient-Derived Tumor Xenograft Models

Patient-derived tumor xenograft (PDX) mouse models have emerged as an important oncology research platform to study tumor evolution, mechanisms of drug response and resistance, and tailoring chemotherapeutic approaches for individual patients. The lack of robust standards for reporting on PDX models has hampered the ability of researchers to find relevant PDX models and associated data. Here we present the PDX models minimal information standard (PDX-MI) for reporting on the generation, quality assurance, and use of PDX models. PDX-MI defines the minimal information for describing the clinical attributes of a patient's tumor, the processes of implantation and passaging of tumors in a host mouse strain, quality assurance methods, and the use of PDX models in cancer research. Adherence to PDX-MI standards will facilitate accurate search results for oncology models and their associated data across distributed repository databases and promote reproducibility in research studies using these models. Cancer Res; 77(21); e62-66. ©2017 AACR.

Abstract 3840: The National Cancer Institute’s patient-derived models repository (PDMR)

The National Cancer Institute (NCI) is developing a Patient-Derived Models Repository (PDMR) comprised of quality-controlled early-passage clinically-annotated patient-derived xenografts (PDXs) and in vitro patient-derived cell cultures (PDCs), including tumor cell and cancer-associated fibroblast cell cultures, to serve as a resource for public-private partnerships and for academic drug discovery efforts. These PDMs will be clinically-annotated with molecular information (whole exome sequence, RNASeq) available in a publicly accessible database and will be available to the extramural community for research use. The PDMR was established by NCI at the Frederick National Laboratory for Cancer Research (FNLCR) in direct response to discussions with academia and industry; the oncology community’s highest priority need is better preclinical models that more faithfully reflect the patient’s tumor and are associated with the patient’s treatment history. NCI has focused on collecting specimens from patients with cancer that are under-represented in many other PDX collections such as head and neck, pancreatic, bladder, ovarian and small cell lung cancers, melanomas and sarcomas. In addition, NCI is increasing its focus on creating PDXs from minority/underserved populations and will soon be expanding to include pediatric cancers. The PDMR generates the majority of its PDXs by subcutaneous implantation; however certain histologies have better take-rates in either orthotopic or alternate implant sites. All SOPs and quality-control standards developed by the PDMR as well as those shared by collaborators will be posted to the public web site that houses the PDMR database. The overall goal of NCI is to create a long-term home for at least 1000 models such that sufficient biological and clinical diversity is represented to allow researchers to ask questions such as: what is the impact of tumor heterogeneity on target qualification or clinical response; do PDXs more faithfully represent the human tumor for pharmacodynamic assay and predictive marker development; or can an adequately powered preclinical PDX clinical trial lead to better evaluation of therapies for future clinical use? Grant Support: This project has been funded in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Citation Format: Yvonne A. Evrard, Michelle Ahalt-Gottholm, Sergio Alcoser, Carrie Bonomi, Suzanne Borgel, John Carter, Biswajit Das, Vivekananda Datta, Cheryl Davis, Kelly Dougherty, Michelle Eugeni, Marion Gibson, Catherine Karangwa, Jason Lih, Dianne Newton, Han Si, Shivaani Kummar, Larry Rubinstein, Alice Chen, P. Mickey Williams, Melinda G. Hollingshead, James H. Doroshow. The National Cancer Institute’s patient-derived models repository (PDMR) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3840. doi:10.1158/1538-7445.AM2017-3840

Promise and limits of the CellSearch platform for evaluating pharmacodynamics in circulating tumor cells

Circulating tumor cells (CTCs), which are captured from blood with anti-epithelial cell adhesion molecule (EpCAM) antibodies, have established prognostic value in specific epithelial cancers, but less is known about their utility for assessing patient response to molecularly targeted agents via measurement of pharmacodynamic (PD) endpoints. We discuss the use of CellSearch (Janssen Diagnostics, LLC, Raritan, NJ) CTC isolation technology for monitoring PD response in early phase trials. We present representative data from three clinical trials with the poly(ADP-ribose) polymerase (PARP) inhibitor veliparib (ABT-888) suggesting that CTCs can be used to measure PD effects. However, while often leading to hypothesis-generating information, our experience points to the difficulty in obtaining sufficient EpCAM-expressing CTCs from patients with advanced disease to reach statistically significant conclusions about PD effects from each trial. Overall, the level of phenotypic heterogeneity observed in specimens from patients with advanced carcinomas suggests caution in the use of cell-surface differentiation marker-based methods for isolating CTCs.

Pharmacodynamic analyses in a multi-laboratory network: lessons from the poly(ADP-ribose) assay

Clinical pharmacodynamic assays need to meet higher criteria for sensitivity, precision, robustness, and reproducibility than those expected for research-grade assays because of the long duration of clinical trials and the potentially unpredictable number of laboratories running the assays. This report describes the process of making an immunoassay based on commercially available reagents "clinically ready". The assay was developed to quantify poly(ADP-ribose) (PAR) levels as a marker of PAR polymerase inhibitor activity for a proof-of-concept phase 0 clinical trial at the National Cancer Institute (NCI) and subsequent clinical trials. In this publication, we retrospectively examine the measures taken to validate the published PAR immunoassay and outline key lessons learned during the development and implementation of these procedures at both internal and external clinical trial sites; these measures included optimizing PAR measurements in tumor biopsies and peripheral blood mononuclear cells (PBMCs), reagent qualification, analytical validation and assay quality control, instrument qualification and method quality control, and support for external laboratories.

Pharmacodynamic Response of the MET/HGF Receptor to Small-Molecule Tyrosine Kinase Inhibitors Examined with Validated, Fit-for-Clinic Immunoassays

PURPOSE

Rational development of targeted MET inhibitors for cancer treatment requires a quantitative understanding of target pharmacodynamics, including molecular target engagement, mechanism of action, and duration of effect.

EXPERIMENTAL DESIGN

Sandwich immunoassays and specimen handling procedures were developed and validated for quantifying full-length MET and its key phosphospecies (pMET) in core tumor biopsies. MET was captured using an antibody to the extracellular domain and then probed using antibodies to its C-terminus (full-length) and epitopes containing pY1234/1235, pY1235, and pY1356. Using pMET:MET ratios as assay endpoints, MET inhibitor pharmacodynamics were characterized in MET-amplified and -compensated (VEGFR blockade) models.

RESULTS

By limiting cold ischemia time to less than two minutes, the pharmacodynamic effects of the MET inhibitors PHA665752 and PF02341066 (crizotinib) were quantifiable using core needle biopsies of human gastric carcinoma xenografts (GTL-16 and SNU5). One dose decreased pY1234/1235 MET:MET, pY1235-MET:MET, and pY1356-MET:MET ratios by 60% to 80% within 4 hours, but this effect was not fully sustained despite continued daily dosing. VEGFR blockade by pazopanib increased pY1235-MET:MET and pY1356-MET:MET ratios, which was reversed by tivantinib. Full-length MET was quantifiable in 5 of 5 core needle samples obtained from a resected hereditary papillary renal carcinoma, but the levels of pMET species were near the assay lower limit of quantitation.

CONCLUSIONS

These validated immunoassays for pharmacodynamic biomarkers of MET signaling are suitable for studying MET responses in amplified cancers as well as compensatory responses to VEGFR blockade. Incorporating pharmacodynamic biomarker studies into clinical trials of MET inhibitors could provide critical proof of mechanism and proof of concept for the field. Clin Cancer Res; 22(14); 3683-94. ©2016 AACR.

Abstract 1186: Creating clinical target validation groups via quality assured transfer of robust clinical pharmacodynamic (PD) assays from the NCI: clinical implementation of a HIF1α immunoassay in tumor biopsies

Robust pharmacodynamic (PD) assay results are valuable for informing go/no-go decisions about preclinical and clinical development of new agents and for identifying optimal combinations of targeted agents. The NCI's Division of Cancer Treatment and Diagnosis (DCTD) develops and validates PD assays to obtain accurate information about target engagement in first-in-human clinical trials. The Pharmacodynamic Assay Development and Implementation Section (PADIS) and National Clinical Target Validation Laboratory (NCTVL) were established at SAIC-Frederick to develop and validate PD assays by validating analytical performance, demonstrating fitness-for-purpose for the clinical protocol, and finalizing companion standard operating procedures (SOPs) for specimen handling and processing. Proven clinical assays are transferred from the NCI to requesting sites in academia and industry through laboratory-based certification and training, centralized access to current SOPs, assistance with assay transfer, and participation in the assay's Quality Assurance Plan. One such validated assay is the HIF1α Immunoassay, which is currently in clinical evaluation at NCTVL. Procedures to clinically implement HIF1α analysis in tumor biopsies from patients treated with angiogenesis inhibitors required consideration of many factors including the low abundance and oxygen sensitivity of the analyte. A customized extraction process for HIF1α was developed and optimized using mouse xenograft biopsies. Use of a customized extraction buffer containing 2-hydroxyglutarate, which was degassed to minimize dissolved oxygen, was shown to improve analyte recovery. Optimized extraction procedures also include use of an automated ceramic bead homogenizer to improve both the efficiency and reproducibility of extraction of HIF1α over sonication. The immunoassay procedure uses lyophilized reagents from a commercially available kit and a modified procedure together with quality control specimens that serve to qualify the performance of each run. Finally, pre-defined data analyses and quality control evaluation criteria are used to perform a semi-automated process to analyze data and evaluate acceptability of performance for both the assay run and individual clinical specimens. In total, five SOPs have been developed to define procedures from collection and freezing of the needle biopsies in the radiology suite to data analyses and reporting. Together, these defined procedures are being used to perform PD analyses as key parts of two clinical trial evaluations at the NCI and will be launched to the community once clinical utility is demonstrated. Funded by NCI Contract No HHSN261200800001E.

Citation Format: Ralph E. Parchment, Robert J. Kinders, Sook Ryun Park, Katherine V. Ferry-Galow, Yvonne A. Evrard, Jiuping Ji, Shivaani Kummar, Joseph E. Tomaszewski, James H. Doroshow. Creating clinical target validation groups via quality assured transfer of robust clinical pharmacodynamic (PD) assays from the NCI: clinical implementation of a HIF1α immunoassay in tumor biopsies. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 1186. doi:10.1158/1538-7445.AM2013-1186