Search

Supplementary Table 9: Differentially expressed proteins in LDE vs. control treated Ptch;p53 SI-CSC and SD-CSC tumors

Supplementary Table 5: mutational profiles for Fsmo;GFAP-cre SD-CSC tumors

Supplementary figures and Tables 1 & 2

The emergence of treatment resistance significantly reduces the clinical utility of many effective targeted therapies. Although both genetic and epigenetic mechanisms of drug resistance have been reported, whether these mechanisms are stochastically selected in individual tumors or governed by a predictable underlying principle is unknown. Here, we report that the dependence of cancer stem cells (CSC), not bulk tumor cells, on the targeted pathway determines the molecular mechanism of resistance in individual tumors. Using both spontaneous and transplantable mouse models of sonic hedgehog (SHH) medulloblastoma treated with a SHH/Smoothened inhibitor (SMOi), sonidegib/LDE225, we show that genetic-based resistance occurs only in tumors that contain SHH-dependent CSCs. In contrast, SHH medulloblastomas containing SHH-dependent bulk tumor cells but SHH-independent CSCs (SI-CSC) acquire resistance through epigenetic reprogramming. Mechanistically, elevated proteasome activity in SMOi-resistant SI-CSC medulloblastomas alters the tumor cell maturation trajectory through enhanced degradation of specific epigenetic regulators, including histone acetylation machinery components, resulting in global reductions in H3K9Ac, H3K14Ac, H3K56Ac, H4K5Ac, and H4K8Ac marks and gene expression changes. These results provide new insights into how selective pressure on distinct tumor cell populations contributes to different mechanisms of resistance to targeted therapies. This insight provides a new conceptual framework to understand responses and resistance to SMOis and other targeted therapies.Significance:The mechanism by which individual tumors become resistant to targeted therapies is thought to be unpredictable. This study provides novel insights into how selective pressure on cancer stem versus bulk tumor cells drives distinct and predictable mechanisms of resistance to targeted therapies. This finding paves a way for future treatment strategies that incorporate anticipated resistance mechanisms in devising second-line therapies in a personalized manner.

Supplementary Table 8: DEG in LDE vs control in SD-CSC Ptch;p53 tumors

Supplementary Table 6: DEG in LDE vs control treated fSMO;GFAP-cre tumors

Supplementary Table 3: Mutational profiles for Ptch;p53 SD-CSC tumors

Supplementary Table 4: mutational profiles for Ptch;p53 SI-CSC tumors

Supplementary Table 7: DEG in LDE vs. control treated SI-CSC Ptch;p53 tumors

To cure a patient's cancer is to eradicate invasive cells from the ecosystem of the body. However, the ecologic complexity of this challenge is not well understood. Here we show how results from eradications of invasive mammalian species from islands—one of the few contexts in which invasive species have been regularly cleared—inform new research directions for treating cancer. We first summarize the epidemiologic characteristics of island invader eradications and cancer treatments by analyzing recent datasets from the Database of Invasive Island Species Eradications and The Cancer Genome Atlas, detailing the superior successes of island eradication projects. Next, we compare how genetic and environmental factors impact success in each system. These comparisons illuminate a number of promising cancer research and treatment directions, such as heterogeneity engineering as motivated by gene drives and adaptive therapy; multiscale analyses of how population heterogeneity potentiates treatment resistance; and application of ecological data mining techniques to high-throughput cancer data. We anticipate that interdisciplinary comparisons between tumor progression and invasive species would inspire development of novel paradigms to cure cancer.

Table S1: (A) Summary of success rates for six mammalian invasive species under different eradication strategies. 95% confidence interval (CI) was calculated using binomial proportion CI with Jeffreys prior; (B) Summary of success rates (Kaplan-Meier five-year survival rates) for different tumor types under different treatment options. 95% CI was calculated using exponential Greenwood CI for Kaplan-Meier estimator and beta product procedure.

The supplementary file contains additional methods, analyses, and details of workflows as referenced in the manuscript, including all supplementary figures and tables. A full table of contents and descriptions are provided within.

Purpose:Genomic and transcriptomic alterations during metastasis are considered to affect clinical outcome of colorectal cancers, but detailed clinical implications of metastatic alterations are not fully uncovered. We aimed to investigate the effect of metastatic evolution on in vivo treatment outcome, and identify genomic and transcriptomic alterations associated with drug responsiveness.Experimental Design:We developed and analyzed patient-derived xenograft (PDX) models from 35 patients with colorectal cancer including 5 patients with multiple organ metastases (MOMs). We performed whole-exome, DNA methylation, and RNA sequencing for patient and PDX tumors. With samples from patients with MOMs, we conducted phylogenetic and subclonal analysis and in vivo drug efficacy test on the corresponding PDX models.Results:Phylogenetic analysis using mutation, expression, and DNA methylation data in patients with MOMs showed that mutational alterations were closely connected with transcriptomic and epigenomic changes during the tumor evolution. Subclonal analysis revealed that initial primary tumors with larger number of subclones exhibited more dynamic changes in subclonal architecture according to metastasis, and loco-regional and distant metastases occurred in a parallel or independent fashion. The PDX models from MOMs demonstrated therapeutic heterogeneity for targeted treatment, due to subclonal acquisition of additional mutations or transcriptomic activation of bypass signaling pathway during tumor evolution.Conclusions:This study demonstrated in vivo therapeutic heterogeneity of colorectal cancers using PDX models, and suggests that acquired subclonal alterations in mutations or gene expression profiles during tumor metastatic processes can be associated with the development of drug resistance and therapeutic heterogeneity of colorectal cancers.

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.

Supplementary Materials and Methods, Supplementary Figures S1-S10 and Supplementary Tables S1-S6

Personalized medicine holds great promise for improving cancer outcomes, yet there is a large inequity in the demographics of patients from whom genomic data and models, including patient derived xenografts (PDX), are developed and for whom treatments are optimized. In this study we develop a genetic ancestry pipeline for the Cancer Genomics Cloud, which we use to assess the diversity of models currently available in the National Cancer Institute (NCI) supported PDX Development and Trial Centers Research Network (PDXNet). We show that there is an over-representation of models derived from patients of European ancestry, which is consistent with other cancer model resources. We discuss these findings in the context of disparities in cancer incidence and outcomes among demographic groups in the US. For example, for the top cancer health disparities affecting African Americans and Latinos, there is a significant lack of ethnic/race appropriate models needed to advance pre-clinical research and personalized clinical treatment. For stomach and liver tumors, which represent disparities in these two minority populations, there are only three available models derived from patients from such backgrounds. Fortunately, ongoing NCI-funded efforts in minority focused PDXNet centers are actively addressing these gaps. We further discuss these results in the context of power analyses to highlight the immediate need for the development of models from minority populations to address cancer health equity in personalized medicine. Citation Format: Brian J. Sanderson, Paul Lott, Katherine Chiu, Juanita Elizabeth Quino, April Pangia Vang, Michael W. Lloyd, PDXNet Consortium, Anuj Srivastava, Jeffrey H. Chuang, Luis G. Carvajal-Carmona. Development and application of genetic ancestry reconstruction methods to study diversity of patient-derived models in the NCI PDXNet Consortium [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 1946.

Over half of BRAF mutant melanoma patients with response to targeted therapy recur within 15 months. Resistance is postulated to arise from certain clonal populations that enter a slow cycling persister state, evade treatment with relative dormancy, and repopulate the tumor when reactivated. Via longitudinal profiling we define expression states of clonal populations and track their evolutionary progression. BRAF mutant patient derived melanoma xenografts were treated with BRAF/MEK inhibitors through maximum tumor size reduction and re-growth. Spatial transcriptomics (ST) was performed at 5 timepoints across 94 days of treatment, tracking clonal populations with intact tissue structure. Deep learning on H&E-stained images automatically detected histological features that co-localize with expression levels. A novel computational pipeline performed clustering, differential expression and pathway enrichment analysis, copy number variation detection, and pseudotemporal ordering, allowing spatial recreation of clonal phylogenies and cell fate trajectory inference during the treatment course. Three clonal categories were defined; Sensitive: predominant in treatment-naïve samples, Persister: remaining clusters in maximally shrunken specimens, Resistant: re-emergent, fast-proliferating. Analysis of clonal composition identified GAPDH and CCND1 as highly expressed in sensitive clones, suppressed in persisters, and reactivated in resistant clones; TYRP1, DCT, MITF, and NGFR showed an opposite pattern. Persister clones evolved from oxidative phosphorylation toward glycolysis and MAPK regulation. Spatial analysis showed increased glycolysis:oxidative phosphorylation ratio in centrally located sensitive clusters, with more balanced peripheral expression. Resistant clones showed upregulated DUSP6 expression and enrichment in Orexin signaling and MAPK regulation. Imaging- and expression-based clones were highly concordant. Spatial profiling during melanoma treatment identified transient persister and emergent resistant clones, defining expression profiles and phenotypic features. Altered ratio of oxidative to glycolytic metabolic pathways was a hallmark of clonal evolution in both space and time. Specific MAPK pathway re-entry points were identified as candidate resistance mechanisms, suggesting potential therapeutic targets. Deep learning features derived from histological images showed good correlation with ST profiles, providing a promising method for clonal tracking via images alone. This longitudinal experiment mimics a tumor’s clinical course, inducing persistence and resistance via treatment that parallels that seen by patients. Combining ST and imaging techniques, we provide insight into clonal dynamics with novel spatiotemporal resolution, defining an evolutionary roadmap to acquired treatment resistance. Citation Format: Jill Carol Rubinstein, Sergii Domanskyi, Todd B. Sheridan, Brian J. Sanderson, SungHee Park, Jessica Kaster, Haiyin Li, Olga Anczukow, Meenhard Herlyn, Jeffrey H. Chuang. Spatiotemporal profiling defines persister and resistance signatures in targeted treatment of melanoma [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 1164.

High-resolution multiplexed tissue imaging such as imaging mass cytometry (IMC) has enabled quantitative cell characterization with preserved tissue architecture. However, the subcellular resolution information has not been fully leveraged yet, as most analyses have focused on cellular or compartment level features. Here, we have developed a robust computational approach integrating cellular spatial and molecular features in multiplexed images to quantify immune synapses. Our approach enables computational detection of subcellular enrichment of T cell co-receptors at immune synapses, which are key regulators of T cell functions in the tumor microenvironment (TME). We quantified immune synapse strength between T cells and various antigen-presenting cells in tumor regions in one melanoma whole tissue immunofluorescence dataset and two independent melanoma IMC datasets covering 99 patients. We observed that sub-localizations of immune synapse-related molecules are highly correlated, e.g. CD3 with CD8. These co-localizations are cell type-specific, with differential behaviors in CD4+ or CD8+ T subtypes. Intra-tumoral T-cell-myeloid synapses are associated with improved T cell proliferation (p = 2.8E-4). In addition, we also observed that cytotoxic T cell - melanoma synapses in immune rich regions are stronger in immune checkpoint inhibition therapy responders than in non-responding patients (p = 0.013). Our approach enables computational resolution of spatial inter-cellular interactions within the TME, is applicable across imaging modalities, and facilities assessment of biological and clinical significance. Citation Format: Zichao Liu, Victor G. Wang, Jan Martinek, Ali Foroughi pour, Jie Zhou, Karolina Palucka, Jeffrey H. Chuang. Computational analysis of immune synapses in melanoma tumor microenvironment [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 5883.

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.

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.

To cure a patient's cancer is to eradicate invasive cells from the ecosystem of the body. However, the ecologic complexity of this challenge is not well understood. Here we show how results from eradications of invasive mammalian species from islands—one of the few contexts in which invasive species have been regularly cleared—inform new research directions for treating cancer. We first summarize the epidemiologic characteristics of island invader eradications and cancer treatments by analyzing recent datasets from the Database of Invasive Island Species Eradications and The Cancer Genome Atlas, detailing the superior successes of island eradication projects. Next, we compare how genetic and environmental factors impact success in each system. These comparisons illuminate a number of promising cancer research and treatment directions, such as heterogeneity engineering as motivated by gene drives and adaptive therapy; multiscale analyses of how population heterogeneity potentiates treatment resistance; and application of ecological data mining techniques to high-throughput cancer data. We anticipate that interdisciplinary comparisons between tumor progression and invasive species would inspire development of novel paradigms to cure cancer.

The emergence of primary and acquired treatment resistance significantly reduces the clinical utility of many effective targeted therapies. Both genetic and epigenetic mechanisms of drug resistance have been reported in literature; however, whether these mechanisms are stochastically selected in individual tumors or governed by a predictable underlying principle is unknown.Here, we report that one can predict a priori the resistance mechanism that will arise in individual SMO inhibitor (SMOi)-resistant SHH medulloblastoma (MB), based on different CSC phenotypes in each tumor. We show that the dependence of cancer stem cells (CSCs), not bulk tumor cells, on the targeted pathway (sonic hedgehog (SHH) pathway) determines the molecular mechanism of resistance in individual tumors. Using both spontaneous (Fsmo;GFAP-cre) and transplantable (Ptch+/-;p53) mouse models of SHH MB treated with a Smoothened inhibitor, sonidegib/LDE225, we show that genetic-based resistance occurs only when the CSCs depend on the targeted pathway. In contrast, SHH MBs containing SHH-dependent bulk tumor cells but SHH-independent CSCs (SI-CSCs), acquire resistance through epigenetic reprogramming. Mechanistically, we discovered that the elevated proteasome activity in SMOi-resistant SI-CSC MBs alters the tumor cell maturation trajectory through enhanced degradation of specific epigenetic regulators, including the histone acetylation machinery. Consequently, SMOi-resistant SI- SMOi-resistant SI-CSC exhibit a global reduction of H3K9Ac, H3K14Ac, H3K56Ac, H4K5Ac, and H4K8Ac marks and gene expression changes. These results provide new insights into how selective pressure on distinct tumor cell populations contributes to different mechanisms of resistance to targeted therapies and implicate histone acetylation in the process. This information can be clinically exploited to understand responses and resistance to SMOis and other targeted therapies. Citation Format: Joshy George, Yaohui Chen, Nourhan Abdelfattah, Keiko Yamamoto, Scott Adamson, Jong Min Choi, Brad Rybinski, Anuj Srivastava, Parveen Kumar, Min Gyu Lee, David S. Baskin, Wen Jiang, Betty Y. Kim, William Flavahan, Jeffrey H. Chuang, Sung Yun Jung, Kyuson Yun. Cancer stem cells, not bulk tumor cells, predict mechanisms of resistance to SMO inhibitors in SHH medulloblastoma [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 3961.

Colorectal cancer is one of the most common cancers in both men and women. Patient stratification in colorectal cancers is challenging and affects therapy decisions and patient care. In this study, we develop an integrative Inception V3-based deep learning approach to stratify TCGA-COAD patients from H&E images, showing that combining deep learning extracted morphological features with clinical and mutational status improves patient stratification. Morphological features within tumor regions can distinguish patients by short or long overall survival (OS) (OS5 years, AUC 0.81±0.12), while they are less informative for intermediate OS (35 years, AUC 0.78±0.20), and again found that combining morphological features and clinical data is superior to only using a single datatype. Our deep learning models were also able to perform reliable tumor segmentation (AUC>0.92). However, survival predictions based on deep learning tumor segmentation were inferior to those based on pathologist tumor annotations (OS5 years, AUC 0.79±0.14). Furthermore, we show that the cross-talk between different data modalities is informative of patient risk. More precisely, models that directly combine image and clinical features performed superior to a naïve Bayes classifier combining predictions of models trained on each data type (integrative model AUC=0.81±0.06 and naïve Bayes AUC=0.76±0.11, p Citation Format: Jie Zhou, Ali Foroughi pour, Rafic Beydoun, Hany Deirawan, Fayez Daaboul, Fahad Shabbir Ahmed, Jeffrey H. Chuang. Integrative deep learning analysis identifies cross-talk between morphology, mutation, and clinical variables for colon adenocarcinoma patient outcomes [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 467.

Metastasis is a major risk factor for poor melanoma outcome, but mechanisms supporting distant organ colonization by melanoma are not fully understood. Here, we found that metastatic melanoma tumors from patients are infiltrated by CD33+ myeloid cells. To determine the role of CD33+ cells in melanoma metastasis, we used NSG mice humanized by engraftment of human CD34+ hematopoietic progenitor cells and transgenic expression of human hematopoietic cytokines SCF/GM-CSF/IL-3 (SGM3). Humanized (h)NSG-SGM3 mice enabled development of human CD33+ myeloid cells in the bone marrow and peripheral tissues, and when implanted subcutaneously with human melanoma cell line, supported melanoma colonization of the spleen, liver, lung, and kidneys. Melanoma growth in distant organs was dependent on host SGM3 expression and facilitated by human CD33+ myeloid cells. Deeper characterization attributed this activity to a rare human IL-3- and SCF-dependent CD33+CD11b+CD117+ progenitor cell subset comprising Citation Format: Chun I. Yu, Jan Martinek, Te-Chia Wu, Kyung In Kim, Joshy George, Elaheh Ahmadzadeh, Rick Maser, Florentina Marches, Patrick Metang, Pierre Authie, Vanessa K. Oliveira, Victor G. Wang, Jeffrey H. Chuang, Paul Robson, Jacques Banchereau, Karolina Palucka. Human KIT+myeloid cells facilitate visceral organ colonization by melanoma [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 1750.

Intratumoral heterogeneity is an essential aspect of cancer biology, as the diversity of cancer cells can increase challenges of designing effective treatment. Cancer cells within the same tumor evolve and compete for shared resources and may also respond differently to drug treatments. In previous work, by closely analyzing triple negative breast cancer xenografts, we have identified a system of closely-related subclonal populations within a tumor that respond differentially to chemotherapy. To explore the dynamics of competition and cooperation between these subclones, we derived cell lines from each and tagged them with fluorescent markers. We then co-cultured these cell lines in 3D organoids and tracked them using Opera Phenix high content screening system (3D confocal fluorescence microscopy). To determine the number of cells and the location of each subclones within organoids, we developed a 3D cell segmentation pipeline. Our pipeline relies on a two-step process where organoids are segmented first and then individual cells are detected within each organoid. Preliminary analysis of hundreds of organoids suggests that the two cell types preserve their relative ratio and sensitivity to treatment regardless of organoid size, suggesting that the system is in a state of ecological coexistence. We will present results on cellular mixing during growth and treatment and the implications for adaptive therapy ecologies in tumors. Citation Format: Patience Mukashyaka, Javad Noorbakhsh, Pooja Kumar, Elise Courtois, Olga Anczukow, Paul Robson, Edson Liu, Jeffrey Chuang. Intratumor heterogeneity quantification using 3D segmentation of organoids co-culture systems [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 3144.

Deep learning has become a popular tool for analyzing hematoxylin and eosin (H&E) stained whole slide images (WSIs) and has been utilized to study conserved spatial behaviors across cancers [1]. Deep learning models are black-boxes and difficult to interpret. We propose the concept of an abstract morphological gene, hereafter called mone, defined as the features encoding the morphology at each region of a WSI. Mones averaged over WSIs share distributional similarities with bulk-level expressions. Such similarities allow using tools originally developed for studying gene expressions. We study >22000 H&E slides of 19 cancers of the cancer genome atlas (TCGA) using the inceptionV3 network. TSNE plots suggest mones detect tumors and distinguish tissues of origin. We obtained AUCs above 95% for one-versus-all predictions of mone-based classifiers detecting tumors and tissue of origin. We also obtained cross-classification accuracies comparable to [1] using a mone-based logistic regression model. Differential mone analysis identifies pan-cancer and cancer-specific mones differentiating tumor and normal slides. Mone 893 is a pan-cancer feature and an indicator of dense cellular regions. We identified this mone in breast cancer WSIs and validated its consistent behavior in ovarian and lung cancers. Differential mone analysis comparing formalin-fixed paraffin-embedded (FFPE) and fresh frozen slides identifies deep learning features which may be affected by frozen tissue artifacts. Removing such features is essential for developing models not confounded by tissue artifacts. While recent deep learning models predict expressions from WSIs (see [2] for examples), here, we use them to identify the morphological features mones encode. Integrative mone-gene co-expression analysis suggests mone 893 heavily correlates with genes in integrin signaling and inflammation mediated by cytokines and chemokines pathways in breast cancer. Mone 869 heavily correlates with expression of COL8A1 in ovarian cancer. Expression of collagen genes is associated with poor prognosis and drug resistance in ovarian cancer [3,4]. These findings indicate the significant associations between individual deep learning-defined features and both genetic and prognostic quantifications. References [1] Noorbakhsh, et al. Deep learning-based cross-classifications reveal conserved spatial behaviors within tumor histological images, bioRxiv 715656 (2020) [2] Schmauch, et al. A deep learning model to predict RNA-Seq expression of tumours from whole slide images. Nat Commun 11, 3877 (2020) [3] Zhang, Wei, et al. Integrating genomic, epigenomic, and transcriptomic features reveals modular signatures underlying poor prognosis in ovarian cancer. Cell reports 4.3 (2013) 542-553 [4] Januchowski, et al. Increased expression of several collagen genes is associated with drug resistance in ovarian cancer cell lines. Journal of Cancer 7.10 (2016) 1295 Citation Format: Ali Foroughi Pour, Jonghanne Park, Jeffrey H. Chuang. MONE: A construction for interpreting deep learning features in pathology slides [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 197.

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.

We created the PDX Network (PDXNet) Portal to provide an intuitive way for researchers to explore and understand the models, sequencing data, and bioinformatics workflows generated by NCI's PDXNet consortium for research access (https://portal.pdxnetwork.org/). The portal also provides metrics for PDXNet's activities, data processing protocols, and training materials for processing PDX data. The PDXNet Portal highlights model and data resources that include 216 new models across 29 cancer types. The most prevalent cancers represented in the PDX model dataset include invasive breast carcinoma (30.6%), melanoma (18.1%), and adenocarcinoma (14.4%). PDXNet teams have provided 2263 sequencing files from 356 samples across 204 patients, comprising whole exome (82.9%) and RNA seq files (17.1%). The most prevalent cancers represented in the PDXNet sequencing data set include Breast Pleural Effusion (27.2%), Breast Poorly Differentiated (12.5%), and Lung Adenocarcinoma (9.6%). The portal also provides access to 9492 sequencing files across 78 disease types that include 2594 samples across 463 patients uploaded from the NCI Patient-Derived Model Repository. The dataset includes both whole exomes (52.8%) and RNA seq (47.2%) data. The PDMR samples include PDX (82.7%), primary tumor (5.7%), normal germline (5.5), organoid culture (3.2), and Mixed Tumor Culture (2.9). The PDMR dataset also has multiple passages: P0 (21.8%), P1(39.5%), P2 (25.6%), and P3 (8.5%). These models and data resources support ten PDXNet Pilot activities, multiple publications, and international collaborations. PDXNet has also developed a set of 13 robust, validated, and standardized workflows for processing PDXNet whole-exome and RNA seq data. Collectively, these workflows allow for the standardized processing of PDX and complementary human tissues (normal and tumor). Our plan is to continuously update the model and data lists on the PDX portal as resources are generated. We expect that the PDXNet generated models, scheduled to grow to 1000 new models by 2022, will support multi-agent treatment studies, determination of mechanisms of sensitivity and resistance, and pre-clinical trials for example through the COMBO-MATCH program. The robust standard workflows currently processing all PDX sequencing data may also facilitate harmonizing data across studies. Lastly, we expect that the generated sequencing data will support computational approaches for studying cancer evolution and the mechanisms underlying cancer treatments. Citation Format: Soner Koc, Mike Lloyd, Steven Neuhauser, Javad Noodbakhsh, Anuj Srivastava, Xing Yi Woo, Ryan Jeon, Jeffrey Grover, Sara Seepo, Christian Frech, Jack DiGiovanna, PDXNet Consortium, Yvonne A. Evard, Tiffany Wallace, Jeffrey Moscow, James H. Doroshow, Nicholas Mitsuade, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis Carvarjal-Carmona, Alana Welm, Bryan Welm, Michael T. Lewis, Govindan Ramaswamy, Li Ding, Shunquang Li, Meenherd Herlyn, Mike Davies, Jack Roth, Funda Meric-Bernstam, Peter Robinson, Carol J. Bult, Brandi Davis-Dusenbery, Dennis A. Dean, Jeffrey H. Chuang. Advancing PDX research through model, data, and bioinformatics with the PDXNet Portal [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 3017.

Histopathological images are an integral data type for studying cancer. We show pre-trained convolutional neural networks (CNNs) can be systematically applied across cancer types, enabling comparisons to reveal shared spatial behaviors. We develop CNNs with a common architecture trained on 19 cancer types of The Cancer Genome Atlas (TCGA), analyzing 14459 hematoxylin and eosin scanned frozen tissue images. Our CNNs are based on the Inception-V3 network and classify TCGA pathologist-annotated tumor/normal status of whole slide images in all 19 cancer types with consistently high AUCs (0.995±0.008). Remarkably, CNNs trained on one tissue are effective in others (AUC 0.88±0.11), with classifier relationships recapitulating known adenocarcinoma, carcinoma, and developmental biology. Moreover, classifier comparisons reveal intra-slide spatial similarities, with an average tile-level correlation of 0.45±0.16 between classifier pairs on the TCGA test sets. In particular, the TCGA-trained classifiers had average tile-level correlation of 0.52±0.09 and 0.58±0.08 on hold-out TCGA lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) test sets, respectively. These relations are reflected on two external datasets, i.e., LUAD and LUSC whole slide images of Clinical Proteomic Tumor Analysis Consortium. The CNNs trained on TCGA achieved cross-classification AUCs of 0.75±0.12 and 0.73±0.13 on LUAD and LUSC external validation sets, respectively. These CNNs had average tile-level correlations of 0.38±0.09 and 0.39±0.08 on LUAD and LUSC validation sets, respectively. Breast cancers, bladder cancers, and uterine cancers have spatial patterns that are particularly easy to detect, suggesting these cancers can be canonical types for image analysis. This study illustrates pre-trained CNNs can detect tumor features across a wide range of cancers, suggesting presence of pan-cancer tumor features. These shared features allow combining datasets when analyzing small samples to narrow down the parameter search space of CNN models. Citation Format: Javad Noorbakhsh, Saman Farahmand, Ali Foroughi pour, Sandeep Namburi, Dennis Caruana, David Rimm, Mohammad Soltanieh-ha, Kourosh Zarringhalam, Jeffrey H. Chuang. Deep learning identifies conserved pan-cancer tumor features [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-003.

Deep learning models are gaining popularity for pathology image analysis in cancer and have successfully been used to build disease and mutation status prediction models [1,2]. However, there are several shortcomings to these models. They typically tile whole slide images with little or no pre-processing leaving small artifacts, such as stitching artifacts, intact in the training data. The network may correlate these artifacts with class labels instead of, or in addition to, biologically meaningful morphological features. While one may argue transfer learning can be used to guard against such artifacts, (1) to the best of our knowledge no rigorous study has been performed to validate this idea given whole slide images, and (2) transfer learning-based deep learning models, such as the one studied in [2], show sensitivity to small artifacts in the data. That being said, transfer learning based methods usually borrow from models trained on the image-net dataset, in which each image contains one object we wish to classify. In contrast, pathology images contain morphological features at different scales spread through the slide, which may be best understood as texture-type features. This begs the question of whether transfer learning models, when used as black-box universal feature extractors, miss important morphological features due to the different nature of features encountered in pathology images. To address these issues we propose a wavelet based convolutional neural network, called AzinNet, inspired by the wavelet based network of [3]. AzinNet uses wavelet transform to (1) denoise the data and remove small artifacts, and (2) create a hierarchy of wavelet channels representing morphological features at different scales. We train and test AzinNet on TCGA whole slide images, and compare it with transfer learning based models inputted with raw and denoised images. We also compare robustness of AzinNet and transfer learning-based models to small perturbations of the data. Initial results suggest AzinNet achieves a higher area under curve and is more robust to data perturbations. References: [1] Noorbakhsh, J., Farahmand, S., Soltanieh-ha, M., Namburi, S., Zarringhalam, K., & Chuang, J. (2019). Pan-cancer classifications of tumor histological images using deep learning. bioRxiv, 715656. [2] Fu, Y., Jung, A. W., Torne, R. V., Gonzalez, S., Vohringer, H., Jimenez-Linan, M., Moore, L., & Gerstung, M. (2019). Pan-cancer computational histopathology reveals mutations, tumor composition and prognosis. bioRxiv, 813543. [3] Fujieda, S., Takayama, K., & Hachisuka, T. (2017). Wavelet convolutional neural networks for texture classification. arXiv preprint arXiv:1707.07394. Citation Format: Ali Foroughi Pour, Javad Noorbakhsh, Jeffrey Chuang. AzinNet: A wavelet convolutional neural network for pathology image analysis [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 2098.

Mechanisms contributing to immunotherapy resistance in patients are the object of intense studies. The myeloid cells play a major role in tumors and include cells with different functions that can be grossly summarized as: (1) Antigen capture for presentation (dendritic cells, DCs) or for degradation (macrophages); (2) Tissue repair (macrophages) and (3) effector function (mast cells, monocytes, granulocytes). However, the functional status of myeloid cells in human tumors is not completely understood. To study intact tumor microenvironments, we have established a comprehensive approach for cellular and molecular analysis. Polychromatic immunofluorescence and histocytometry showed that CD14+ cells represent the majority of the total tumor immune infiltrate. Furthermore, while majority of the T cells are located outside of cancer nests, most of the T cells present in the melanoma tissue are in direct contact with CD14+ cells rather than melanoma cells. The distribution of CD14+ cells shows two distinct patterns: CD14+ cells within cancer nest (intratumoral) are in close interactions with melanoma cells and are loaded with melanoma protein; CD14+ cells in the tissue surrounding cancer nest (stromal) do not show melanoma protein cargo. Using customized immunofluorescence guided laser capture micro-dissection, we harvested CD3+ T cells based on their tissue location and CD14+ cells based on their tissue location and melanoma protein load for downstream analysis. Both stromal and intratumoral CD14+ cells display a macrophage like identity with RNA expression of so called “M2 like” macrophages markers, yet we detect a significant degree of variability in the expression levels for certain markers. Transcriptional profiling showed that CD14+ cells clustered according to their tissue localization, i.e., intratumoral vs. stromal, interestingly this localization imprint was less pronounced for T cells, where samples clustered preferentially by patient, leading to much lower number of localization specific differentially expressed genes. Computational analysis, revealed distinct gene signatures associated with different inflammatory and metabolic pathways in intratumoral and stromal CD14+ cells. Thus, transcriptome differentiates functional status of CD14+ cells related to their localization within tumor. Further the intra/stromal CD14+ signature clusters patient with significantly better long term survival across multiple cancer types in TCGA, which in metastatic melanoma was linked with dendritic cells signature in the stroma. Finally, combining our CD3+ and CD14+ LCM data we identified localization specific pairs of receptor/ligand interactions between myeloid and T cells. Citation Format: Jan Martinek, Kyung In Kim, Jianan Lin, Te-Chia Wu, Hannah Borouchov, Ananya Gulati, Lili Sun, Victor Wang, Joshy George, Philipp Henrich, Florentina Marches, Anthony Rongvaux, Michael Chiorazzi, Jeffrey H. Chuang, Paul Robson, Richard Flavell, Jacques Banchereau, Karolina Palucka. Transcriptional profiles of CD14+ cells in situ in melanoma reveal plasticity, localization dependent function and specific T cell interactions [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 3441.

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, which could impact the capacity of PDXs for faithful modeling of patient treatment response. Such results contrast with reports that have observed genomic fidelity of PDX models with respect to the originating patient tumors and from early to late passages by direct DNA measurements (DNA sequencing or SNP arrays). Here we resolve these contradicting observations by systematically evaluating CNA changes and the genes they affect during engraftment and passaging in a large, internationally collected set of PDX models, comparing both RNA and DNA-based approaches. The data collected, as part of the U.S. National Cancer Institute (NCI) PDXNet (PDX Development and Trial Centers Research Network) Consortium and EurOPDX consortium, comprises 1548 patient (PT) and PDX datasets (1451 unique samples) from 509 models derived from American, European and Asian cancer patients, spanning across 16 tumor types. By assessing copy number changes by pairwise (PT-PDX or PDX-PDX) correlation and residual analysis to control for systematic biases, our study demonstrates that prior reports of systematic copy number divergence between PTs and PDXs are incorrect, and confirms the high retention of copy number during PDX engraftment and passaging. Moreover, only a small proportion of models show large CNA discordance between the samples, suggesting that the variations observed in PDX are mainly due to rare clonal selection of individual tumors rather than murine pressures. This large scale data analysis also reveals several other findings that clarify the evolutionary process in PDXs. We do observe larger deviations between PT-PDX than in PDX-PDX comparisons, likely due to dilution of PT signal by human stromal cells. Interestingly, we found that a major contributor to the differences between PDX samples is lineage-specific drift associated with splitting of tumors into fragments during PDX propagation. We observed no significant enrichment of cancer-related genes in PDX-specific CNAs across models, suggesting the lack of systematic copy number evolution driven by the PDX mouse host. Moreover, CNA differences between patient and PDX tumors were comparable to variations in multi-region tumor samples or intra-patient samples. Thus concerns about the genetic stability of the PDX system are likely to be less important than the spatial heterogeneity of solid tumors themselves. This result is consistent with our results on lineage effects during passaging, which indicate that intratumoral spatial evolution is the major reason for genetic drift. Our in-depth tracking of CNAs throughout PDX engraftment and passaging confirms that tumors engrafted and passaged in PDX models maintain a high degree of molecular fidelity to the original patient tumors and their suitability for pre-clinical drug testing. This work also finely enumerates the copy number profiles in hundreds of publicly available models, which will enable researchers to assess the suitability of each for individualized treatment studies. Citation Format: Xing Yi Woo, Jessica Giordano, Anuj Srivastava, Zi-Ming Zhao, Michael W. Lloyd, Roebi de Bruijn, Yun-Suhk Suh, Jong-Il Kim, Han-Kwang Yang, Charles Lee, Dennis A. Dean, Brandi Davis-Dusenbery, Yvonne A. Evrard, James H. Doroshow, Alana L. Welm, Bryan Welm, Michael T. Lewis, Bingliang Fang, Jack A. Roth, Funda Meric-Bernstam, Meenhard Herlyn, Michael Davies, Li Ding, Shunqiang Li, Ramaswamy Govindan, Claudio Isella, Jeffrey A. Moscow, Livio Trusolino, Annette Byrne, Jos Jonkers, Carol J. Bult, Enzo Medico, Jeffrey H. Chuang, PDXNET consortium & EurOPDX consortium. Conservation of copy number profiles during engraftment and passaging of 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 1673.

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.

Tumor-infiltrating lymphocytes (TIL) hold strong prognostic value in determining patient survival as well as response to immune checkpoint blockade across many tumor types, especially metastatic melanoma. Tumors often evolve mechanisms to prevent TIL access and/or function to allow for continued growth. One such mechanism is the recruitment and reprogramming of tumor-associated macrophages (TAM) which suppress TILs through a variety of pathways. The interactions between TAMs and TILs that result in a suppressive tumor microenvironment (TME) have been studied in vitro and with spatially-agnostic sequencing, leaving the location-specific contributions underexplored. Interrogating the TAM-TIL interactions in an intact TME will uncover novel mechanisms responsible for TAM function and may be key to reversing TIL immunosuppression, improving anti-tumor immune responses, survival, and potentially therapy response. We use histocytometry, a multiplex quantitative tissue imaging method capable of capturing whole-slide tumor sections, to explore the human metastatic melanoma TME in situ and resolve the spatial contributions of TAM-TIL interactions. We demonstrate a T-cell preference for physical contact with phagocytic TAMs in the tumor but non-phagocytic TAMs in the stroma. We further model immune synapses from histocytometry to probe the function underlying the TAM-TIL preference based on the directionality of signaling components. The model suggests T-cells in contact with macrophages in the tumor are communicating more frequently with the phagocytic TAMs as compared to their non-phagocytic counterparts, indicating that this interaction extends beyond a simple proximity preference. In addition, proliferating T-cells communicate more frequently with macrophages than non-proliferating T-cells but this effect limited to the intratumor space. Together we demonstrate the functional consequences of physical TAM-TIL interactions with unprecedented spatial resolution. This work unravels important TAM interactions that shape the TME which have not been previously appreciated, providing novel insight into the forces modifying T-cell function in the TME and revealing new TAM biology to explore further. Citation Format: Victor G. Wang, Jan Martinek, Ananya Gulati, Hannah Boruchov, Kelly Ray, Karolina Palucka, Jeffrey H. Chuang. Multiplex fluorescent imaging of the metastatic melanoma microenvironment reveals spatially-dependent macrophage interactions modulating tumor-infiltrating T-cells [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 2832.

Hypoxia-driven adenosine accumulation in tumor microenvironment is believed to play an important role in tumor evasion of immune response. Adenosine via interaction with its A2a receptor, expressed on immune cells, acts as a negative regulator of cytotoxic T-cell immune activity. Increased A2aR expression was observed in tumor infiltrating lymphocytes in freshly resected NSCLC specimens. Higher expression of A2aR was associated with worse prognostic findings in head and neck squamous cell carcinoma. Conversely, it correlated with favorable clinical outcome in non-metastatic lung adenocarcinoma. It has been hypothesized that A2aR antagonism could lead to enhanced efficacy for immune checkpoint blockade. A2aR antagonists have demonstrated favorable in vitro and ex vivo response in NSCLC and ongoing clinical trials are testing these drugs in conjunction with established immunotherapy. However, little is known about the role of A2aR in NSCLC patients. We investigated how A2aR expression affected tumor immune landscape, PD-L1 expression and clinical outcomes in a large pool of NSCLC patients. Methods: We obtained A2aR expression for NSCLC patients from TCGA; adenocarcinoma (ADC), n=586; squamous cell carcinoma (SCC), n=517. The data was arranged into 4 quartiles according to A2aR expression mRNA-seq z-scores, defining the highest quartile as A2aR-high and the lowest quartile as A2aR-low. We evaluated how A2aR expression correlated with a) immune landscape, b) PD-L1 expression, c) tumor mutational burden (TMB) and neoantigen burden, and d) clinical outcome. Results: In ADC, A2aR-high was significantly associated with lower infiltration of activated CD4 and CD8 T cells when compared with A2aR-low patients (12% vs 48% ; 9% vs 47%, all p Citation Format: Pedro Viveiros, Bhoomika Sukhadia, Pratyusha Nunna, Keon Woo Park, Jeffrey Chuang, Lihua Zou, Young Kwang Chae. Implications of the A2a receptor (A2aR) on tumor microenvironment in non-small cell lung cancer (NSCLC) [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 523.

Patient-derived xenograft (PDX) models are in vivo models of human cancer that have been used for translational cancer research and therapy selection for individual patients. The Jackson Laboratory (JAX) PDX resource has over 450 models representing more than 20 different types of cancer. The models undergo rigorous quality control and are genomically characterized to identify somatic mutations, copy number alterations, and transcriptional profiles. Bioinformatics workflows for analyzing genomic data obtained from human tumors engrafted in a mouse host (i.e., Patient-Derived Xenografts; PDXs) must address challenges such as discriminating between mouse and human sequence reads and accurately identifying somatic mutations and copy number alterations when paired non-tumor DNA from the patient is not available for comparison. Here we describe bioinformatics analysis workflows and guidelines (https://github.com/TheJacksonLaboratory/PDX-Analysis-Workflows) that we developed specifically for the analysis of genomic data generated from PDX tumors. Our workflows incorporate commonly used software and public databases but are tailored to address the specific challenges of PDX genomics data analysis through parameter tuning and customized data filters and result in improved accuracy for the detection of somatic alterations in PDX models. We also report a gene expression-based classifier that can identify EBV-transformed tumors. Finally, to demonstrate the effectiveness of our workflows, we show the overall concordance of the genomic and transcriptomic profiles of the PDX models in the JAX PDX resource with relevant tumor types from The Cancer Genome Atlas (TCGA). Using the reliable results obtained from the PDX genomics data analysis, we are able to compare the patient tumor with different PDX passages, perform classification analysis to verify the annotations of PDX tumors, as well as associate genomic signatures of each PDX tumor with results from dosing studies. Acknowledgements The data analysis workflows reported in this publication were partially supported by the National Cancer Institute of the National Institutes of Health under Award Number P30CA034196. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The public portal for JAX PDX data is supported by R01CA089713. Citation Format: Xing Yi Woo, Anuj Srivastava, Joel H. Graber, Vinod Yadav, Vishal Kumar Sarsani, Al Simons, Glen Beane, Stephen Grubb, Guruprasad Ananda, Grace Stafford, Jeffrey H. Chuang, Susan D. Airhart, R. Krishna Karuturi, Joshy George, Carol J. Bult. Genomic data analysis workflows for tumors from patient-derived xenografts (PDXs): Challenges and guidelines [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 1075.

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.

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.

Background: Triple negative breast cancer (TNBC) is the most aggressive subtype of invasive breast cancer that lacks ER, PR, and HER2 expression. It is a heterogeneous disease with several molecular subtypes: basal-like1 (BL-1), basal-like 2 (BL-2), mesenchymal (M), and luminal androgen receptor (LAR). Treatment for TNBC is normally limited to chemotherapy, and relapse is common. Here we report molecular alterations associated with TNBC metastasis by analyzing the genomic profiles of paired primary and metastatic TNBCs. Methods:50 paired TNBCs were identified through an IRB-approved protocol via the City of Hope (COH) Biospecimen Repository. DNA mutation and mRNA expression profiles of 10 paired primary and metastatic TNBCs were analyzed. DNA mutations were identified using exome sequencing by FoundationOne®. Affymetrix Human Genechip 2.0 was used for mRNA expression profiling. Raw data were normalized and processed using Expression Console, and linear regression was performed using Limma to identify the differentially expressed genes between primary and metastatic TNBCs. Results: DNA mutation profiling showed that multiple mutations occurred within genes covering pathways of PI3K/AKT/mTOR, DNA repair, RAS/MAPK, cell cycle, and growth factor receptor signaling, reconfirming genomic heterogeneity of TNBCs. Gene expression profiles were analyzed for Lehmann's TNBC molecular subtypes (BL-1, BL-2, M, and LAR). Six of ten TNBCs showed phenotype shift between the primary and metastatic TNBCs. Several unique gene expression patterns were identified by comparing the paired TNBCs. CCNE1 and TPX2 were co-overexpressed in metastatic TNBCs compared to paired primaries. This mirrored prior studies in ovarian cancer, where co-overexpression of CCNE1 and TPX2 were found related to clonal resistance against chemotherapy. Splicing factors TRA2B and SRSF7 were also over-expressed in metastatic TNBCs compared with primaries. The analysis studying the association of CCNE1 and TPX2 with overall survival is ongoing using TCGA. Conclusion: Overall, these results show the comparative changes between primary and relapsed TNBCs and indicate the heterogeneity of molecular mechanisms of recurrence. CNNE1 and TPX2 may represent important genes involved in TNBC metastasis. Further analyses including a total of 50 paired TNBCs are currently underway. Study Contact: Yuan Yuan MD PhD, City of Hope Comprehensive Cancer Center; Duarte, CA 91030; Email: yuyuan@coh.org Citation Format: Zhao Z-M, Yost S, Yuan Y-C, Warden C, Chuang J, Yuan Y. Mechanisms of recurrence: Paired analysis of primary and metastatic triple negative breast cancer [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD8-02.

Population heterogeneity within tumors is essential to the development of drug resistance. However, precise quantification of cellularity levels of subpopulations, and in particular how they evolve in response to treatment, has been challenging. Here we describe the high precision characterization of subclonal evolution within triple-negative breast cancer patient-derived xenografts (PDXs) generated from three patients. For each patient model, we established multiple PDXs and treated them in cohorts of 5-10 mice each for the therapies cyclophosphamide, doxorubicin, cisplatin, or docetaxel, with treatments lasting one month. In all three patient models, the average behavior across mice was a reduction in size in response to docetaxel, but growth under doxorubicin and cyclophosphamide. For cisplatin treatment, one of the three models showed tumor shrinkage while the other two models showed continued growth but at a rate lower than the doxorubicin or cyclophosphamide cohorts. To determine the evolutionary behaviors underlying these observations, we initially performed exome panel sequencing of 34 residual tumor samples from these and untreated xenografts. Computational mutation and copy number analysis indicated sample-specific differences in tumor populations both in response to treatment and due to genetic drift. However, they also revealed measurement uncertainties related to exome capture efficiency, locus-specific read counts, and computational copy number estimation that limited quantitative inference of evolutionary behaviors. To solve this problem we used droplet digital PCR (ddPCR) to measure variant allele frequencies and local copy number at selected loci from the prior residual tumors as well as additional samples from replicate cohort treatments and cultured conditionally reprogrammed progenitor cells. In total we performed 1665 ddPCR measurements across 150 cancer samples. These ddPCR measurements reduced sample-specific uncertainty in variant allele frequency to ~2% and copy number to ~0.2, allowing for precise identification of subclones and their cellularities in each sample. We observed several common modes of evolution within these tumors including selective sweeps, spatial diffusion, and stable coexistence between distinct subpopulations. In the samples from one patient model, we observed frequent symbiotic growth of two distinct subpopulations having differential cisplatin sensitivity, such that the degree of clonal selection during cisplatin treatment was proportional to the tumor volume change on a mouse-to-mouse basis. This study demonstrates how high precision genomic characterization across comparatively treated samples can reveal treatment-relevant subclonal ecology, as well as the mutations that distinguish populations with different behaviors. Citation Format: Hyunsoo Kim, Pooja Kumar, Francesca Menghi, Eliza Cerveira, Mallory Romanovitch, Guru Ananda, Joshy George, Henry Chen, Susan Mockus, Chengsheng Zhang, Yan Yang, James Keck, R. Krishna Murthy Karuturi, Carol Bult, Charles Lee, Edison Liu, Jeffrey H. Chuang. High-precision quantification of clonal evolution during comparative treatments of triple negative breast cancer [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 2909. doi:10.1158/1538-7445.AM2017-2909

Individual tumors, including the aggressive and difficult to treat triple-negative (ER-/PR-/HER2-) breast cancers (TNBCs) are heterogeneous collections of cells with multiple subclonal populations each contributing to the tumor. While subclonal heterogeneity is likely responsible for the development of drug resistance, identification of how tumor cell populations change over time has been difficult, largely because of the challenges in resampling tumor tissue at close time points. Here we quantify tumor evolution in human patient-derived xenografts implanted into NSG mice, which we use to test subclonal heterogeneity as a function of location within a tumor, propagation time, and drug treatment. We used high-depth (∼400x) sequencing of a targeted panel of 358 genes to quantify somatic mutation allele frequencies from 6 spatially-separated and 8 temporally-propagated xenograft samples derived from the same TNBC patient tumors. Samples ranged in age from 2-4 months post-engraftment. Although we observed a few low frequency mutations distinguishing samples, overall we found that allele frequencies of somatic mutations were well-preserved on this time scale. We then generated replicate xenografts from the same patient tumor and treated them respectively with cisplatin, doxorubicin, cyclophosphamide, docetaxel, or vehicle control for 25 days. Although again somatic mutations showed few differences in allele frequency across samples, substantial variations were seen when data were analyzed for copy number alterations. To confirm these effects we repeated the treatments for xenografts derived from two additional TNBC patients. Again we observed strong changes in tumor heterogeneity at the copy number level. This effect was particularly, but not exclusively, apparent in tumors with the greatest response to therapy. We further verified these measurements through Sanger and digital PCR sequencing on the treated mice and other mice in the same cohorts. Using a multi-sample xenograft propagation, dissection, sequencing, and computational analysis protocol, we have shown that tumor subpopulation changes in response to treatment can be quantified and distinguished from spatial or temporal effects, even for treatment time courses as short as 1 month. In triple negative breast cancer these variations are most apparent at the level of copy number variation. Our study demonstrates how patient-derived xenografts can provide detailed resolution of tumor population evolution during the manifestation of resistance. Citation Format: Hyunsoo Kim, Pooja Kumar, Francesca Menghi, Joshy George, Guru Ananda, Susan Mockus, Chengsheng Zhang, Nicholas Larson, Henry C. Chen, Yan Yang, James Keck, R. Krishnamurthy Karuturi, Charles Lee, Carol Bult, Edison Liu, Jeffrey H. Chuang. Evolution during propagation and treatment of patient-derived triple negative breast cancer xenografts. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 2411.