Patient-derived Xenograft Samples Research Articles

Abstract PO2-28-08: A spatial transcriptomic study of a triple-negative breast cancer (TNBC) patient-derived xenograft (PDX) model of residual disease refractory to conventional chemotherapy

Abstract TNBC is one of the most aggressive subtypes of breast cancer. Combating chemotherapy resistance is critical to improving quality of care and reducing fatality among TNBC patients. In order to understand the mechanisms of chemoresistance in TNBC tumors, it has been proposed that the spatial interactions between cell types in the tumor microenvironment (TME) could offer insights into the differential response to therapy among tumors, and metastatic potential of certain tumors over others. Thus, this study utilizes spatially resolved transcriptomic (SRT) technology to profile the spatial interactions in the TNBC TME. With the goal of better understanding chemoresistance in TNBC and providing a proof-of-concept for new technology, we have launched a longitudinal SRT study of a TNBC PDX of residual disease before, during, and after adriamycin and cyclophosphamide (AC) treatment (Tx). SRT by 10X Genomics Visium was performed on vehicle, AC-treated residual tumor (21 days post-Tx), and AC-treated regrown tumor (50 days post-Tx when tumors regrew to starting tumor volume). This study is divided into two parts: 1) development of a computational pipeline; 2) spatial colocalization analysis of residual and regrown tumors to understand chemoresistance. PART 1: Due to the lack of specialized tools for processing and sorting xenograft reads from SRT data, we developed the Xenomake pipeline, which combines a xenograft sorting algorithm (Xengsort) and spatial barcode demultiplexing pipeline to assign reads into the host and graft organisms for each spatial spot. RESULTS: Xenomake permits clustering the spatial spots into stroma-rich (enriched for mouse mRNAs), and epithelial-rich (enriched for human mRNAs) regions. We show that Xenomake can find differential cytokine production in the stroma and epithelium. Since PDX data separate the tumor into stroma and epithelium by organism, the pipeline enables fine-tuned downstream analysis such as stromal-stromal and stromal-epithelial interactions. Xenomake is thus generally applicable for SRT involving PDX samples. PART 2: Previously we detailed patterns of tumor regression into a residual tumor state, followed by uncontrolled regrowth in the absence of treatment in multiple PDX models of TNBC. Although using PDX models necessitates immune-compromised mice, several types of stromal populations are present and analyzable in these models. Thus, using organism-assigned reads processed by Xenomake, we computationally inferred the localization pattern of stromal cell types in Visium spots including cancer associated fibroblasts (CAF), macrophages (MP), endothelial cells (ENDO), monocytes, and perivascular-like cells. We compared the spatial localization profiles (SLP) of stromal cell types across samples. With human reads, we computed the spatial pathway activity map (SPAM) for the ~30 HALLMARK pathways across samples. We correlated the patterns of SPAM with stroma cell-type SLP to survey the stroma-epithelial interactions across samples. RESULTS: Stroma cells are generally distributed in the tumor periphery in vehicle and AC50, while in AC21 there is a notable increase in stroma abundance and stroma infiltration within tumor mass. SLP of MP (Cd68+ and Csf1r+) is correlated with ENDO (Pecam1+), and with CAF (Acta2+ and Pdgfrb+). Additionally, all 3 cell types are correlated with Vim and Cd44 expression. Although all samples show enrichment of MP, ENDO, and CAF, they interact differently with pathway activities of adjacent epithelial cells. In vehicle and AC50, the 3 cell types are colocalized with OXPHOS, MYC targets, E2F pathway, while in AC21, a switch to colocalization with EMT is observed. Furthermore, hypoxia response and glycolysis display anti-correlation with stroma cell types, meaning that these pathway activities are further away from stroma and occupy distinct territories. The results suggest stroma-tumor metabolic crosstalk and ways of targeting residual disease. Citation Format: Benjamin Strope, Katherine Pendleton, William Bowie, Gloria Echeverria, Qian Zhu. A spatial transcriptomic study of a triple-negative breast cancer (TNBC) patient-derived xenograft (PDX) model of residual disease refractory to conventional chemotherapy [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-28-08.

Nextflow pipeline for Visium and H&E data from patient-derived xenograft samples

We designed a Nextflow DSL2-based pipeline, Spatial Transcriptomics Quantification (STQ), for simultaneous processing of 10x Genomics Visium spatial transcriptomics data and a matched hematoxylin and eosin (H&E)-stained whole-slide image (WSI), optimized for patient-derived xenograft (PDX) cancer specimens. Our pipeline enables the classification of sequenced transcripts for deconvolving the mouse and human species and mapping the transcripts to reference transcriptomes. We align the H&E WSI with the spatial layout of the Visium slide and generate imaging and quantitative morphology features for each Visium spot. The pipeline design enables multiple analysis workflows, including single or dual reference genome input and stand-alone image analysis. We show the utility of our pipeline on a dataset from Visium profiling of four melanoma PDX samples. The clustering of Visium spots and clustering of H&E imaging features reveal similar patterns arising from the two data modalities.

Abstract 2758: Stemness subtype gastric cancer accelerate hematogenous metastasis

Abstract Oncogenic progression and metastasis correlate with the acquisition of stem cell-like characteristics and the loss of differentiated phenotype. Poorly differentiated primary tumors often lead to unfavorable oncologic outcomes due to a heightened likelihood of metastasis to distant organs across various cancers. However, the relationship between stemness traits and metastasis in gastric cancer (GC) remains insufficiently explored. In this investigation, we analyzed bulk RNA-sequencing from microdissected primary tumor specimens and clinical data to pinpoint molecular subtypes linked to distant organ metastasis. Transcriptomic data from patient-derived xenograft (PDX) samples and available single-cell RNA datasets validated our bulk RNAseq findings. Comprehensive analysis highlighted the molecular traits of GC tied to hematogenous metastasis. Microdissection of primary tumor FFPE samples facilitated bulk RNA sequencing, encompassing samples from 33 metastatic and 31 metastasis-free patients, with 8 paired normal samples. Most samples contained less than 10% of tumor microenvironment (TME) components like fibroblasts, endothelial cells, and immune cells. Using non-negative matrix factorization (NMF) clustering, we achieved unsupervised subtyping of the gastric cancer gene expression matrix. Permutation tests based on subsampling indicated consistent high-level cophenetic coefficients at k=2, suggesting robust bifurcation of transcriptomic data into two clusters. One showed augmented stemness traits while the other emphasized differentiated mucosal genes, labeled as stemness and differentiated subtypes, respectively. GSEA highlighted a significant boost in the Hallmark angiogenesis score in the stemness subtype, not attributed to endothelial cell infiltration as verified by CIBERSORTx. This subtype also exhibited a marked decrease in hematogenous metastasis-free survival (HMFS) (p=0.008, log-rank test). Further, PDX samples were scrutinized. Post classification of the PDX graft expression matrix via logistic regression, we examined the endothelial cell fraction, determined by CIBERSORTx. Stemness subtype PDX grafts revealed a significant elevation in the host (mouse) endothelial cell fraction (p<0.05). In addition, stemness PDX patient showed significant decreased HMFS (p = 0.03). We also categorized the pseudobulk expression matrix of malignant cells into stemness or differentiated subtypes using GSE183904 scRNA cohort. scRNA sample analysis consistently presented an elevated endothelial cell fraction in the stemness GC subtype. The outcomes suggest a propensity for angiogenesis in the TME of the stemness subtype of GC. In conclusion, our findings demonstrate that the stemness subtype of GC accelerates hematogenous metastasis, as evidenced through clinical data, bulk RNAseq, PDX, and scRNAseq data. Citation Format: Seungho Lee, Jaeun Yoo, Seungbok Lee, Hyun Myong Kim, Kyoungyun Jeong, Yie-Ri Yoo, Ji-Yeon Shin, Kyoung Un Park, Hye Seung Lee, Seong-Ho Kong, Do Joong Park, Hyuk-Joon Lee, Han-Kwang Yang. Stemness subtype gastric cancer accelerate hematogenous metastasis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2758.

Abstract 4215: Single-cell proteomic characterization of the tumor microenvironment in humanized NOG-EXL patient-derived xenograft (PDX) models

Abstract Introduction: CD34+ hematopoietic stem cell (HSC) derived humanized mice engineered to express human cytokine transgenes are promising models for the evaluation of novel immunotherapeutic treatments. While flow cytometry-based analyses have confirmed the presence of many human immune cell populations within humanized mice bearing patient-derived tumor xenografts (PDX), comprehensive high-dimensional immunophenotyping of the tumor microenvironment (TME) of these models remains limited. Beyond this, the relative influence of PDX vs. donor CD34+ HSC populations on TME composition is poorly understood. Here, we characterize the quantity and diversity of immune cell subpopulations in the TME, bone marrow, and spleen of humanized NOG-EXL (huNOG-EXL) mice using Cytometry by Time-Of-Flight (CyTOF) data with > 60 features. Methods: We performed CyTOF analysis on 250 dissociated samples from primary tumors (n = 15; n = 6 head and neck cancer, n = 6 ovarian cancer, n = 3 renal cell carcinoma), as well as xenograft (n = 59), spleen (n = 88), and bone marrow (n = 88) samples from huNOG-EXL mice (Taconic). Each sample was profiled with two partially overlapping antibody panels; an innate immunity-focused panel (40 features), and an adaptive immunity-focused panel (40 features). Results: CyTOF data were collected from over 10 million live immune cells for each of the adaptive and innate antibody panels. Unbiased clustering of single-cell marker expression revealed clusters corresponding to all major lymphoid and myeloid cell types. Therapeutically relevant cell states were identified including exhausted CD8+ T cells marked by PD-1 and Tim-3 expression, and PD-L1+ macrophages. Hierarchical clustering of bone marrow and spleen samples suggested that systemically, immune composition was not driven by tumor-intrinsic factors alone with many samples clustering on the basis of CD34+ stem cell donor ID. Conversely, clustering PDX samples revealed strong consistency in xenografts originating from the same primary tumor sample regardless of the chosen stem cell donor. The immune composition of primary tumor samples showed greater similarity with matched PDX samples versus unmatched PDX samples, further supporting that tumor-intrinsic factors drive immune composition in xenografts. Conclusions: We provide a high-resolution characterization of the immune phenotypes present in huNOG-EXL PDX models across three cancer types. These data captured substantial heterogeneity in the immune composition of humanized xenograft samples, and we demonstrate that immune composition in humanized xenografts is strongly driven by tumor-intrinsic features. Overall, this resource data supports the use of these models for immuno-oncology studies with future work focused on functionally validating specific immune cell populations of interest. Citation Format: Daniel Stueckmann, Jalna Meens, Sally Zhang, Shirley Hui, Sujana Sivapatham, Sephane Chevrier, Jennifer Pfeil, Lisa Martin, Maria Komisarenko, Philip Dube, Susan Prendeville, Hartland Jackson, Antonio Finelli, Gary Bader, Bernd Bodenmiller, Keith A. Lawson, Laurie Ailles. Single-cell proteomic characterization of the tumor microenvironment in humanized NOG-EXL patient-derived xenograft (PDX) models [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4215.

Abstract 2504: Determining tumour homologous recombination repair status in PDX and clinical samples using automated RAD51 scoring

Abstract PARP inhibitors (PARPi) as monotherapy are approved in a wide variety of cancers that show genetic alterations that result in homologous recombination repair (HRR) deficiency. Current genomic scar tests for HRR deficiency (HRD) are predictive of response to PARPi on a population scale but they may fail to provide a real-time functional status of HRR in an individual. RAD51 is a critical DNA repair factor that functions in HRR to safeguard the integrity of the genome and could be a useful patient selection biomarker for PARPi response.During HRR, RAD51 forms filamentous structures to facilitate repair, which can be visualised by direct immunofluorescence (IF) as nuclear foci. Quantification of these RAD51 foci has been shown to be a functional readout of HRR status, however current manual counting of foci is challenging and can result in high inter-observer variability. In order to standardise RAD51 quantification, we aimed to develop an image analysis solution.Tumour samples from patient derived xenograft (PDX) studies and human tissues from commercial sources and clinical trials were co-stained with IF for RAD51 and geminin, as a marker of cells in S/G2 phase of the cell cycle. Manual scoring was performed as reported previously. Visiopharm was used to create and image analysis pipeline in allowing delineate tumour epithelial areas, detection of the geminin positive nuclei and RAD51 foci within these and the quantification of the foci.We have successfully developed an image analysis solution to select geminin positive proliferating tumour cells from PDX and clinical material, then quantify the RAD51 foci within these cells. There is a good correlation with manual scores with a similar accuracy to the manual inter-observer variability (PDX: R2 = 0.76, F1 = 0.95 [n=34]; breast cancer samples: R2 = 0.72, F1 = 0.92 [n=33]). Developing and image analysis solution has challenges due to several factors such as pan-nuclear staining of non-focus forming RAD51 being variable across samples, manual scoring using a wider Z-plane, susceptibility of RAD51 to pre-analytical factors and the fact that a duplex IF stain with geminin is required to mark actively replicating cells.The frequent inactivation of HRR is known to sensitise cancer cells to PARP inhibitors, we demonstrate a quantitative functional readout of HRR competency using RAD51 foci staining and an image analysis algorithm. This solution may enable the use of RAD51 foci as a patient selection biomarker for PARP inhibitors by providing an objective, robust and reproducible algorithm for scoring. Citation Format: Jack Robertson, Jack Ashton, Gemma N. Jones, Gordon Mills, Tugba Y. Ozmen, Matthew J. Rames, Soonyoung Park, Paul Waring, Natalia Lukashchuk, Elizabeth Harrington. Determining tumour homologous recombination repair status in PDX and clinical samples using automated RAD51 scoring [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2504.

Abstract 4190: Patient-derived xenograft (PDX) models with differential HER2 expression for preclinical evaluation of HER2-targeted therapies

Abstract Human epidermal growth factor receptor 2 (HER2) is a critical therapeutic target in gastric cancer. However, the varying expression levels of HER2 in patients present a significant challenge in selecting the appropriate HER2-targeted therapies. Patient-derived xenograft (PDX) models offer a valuable tool to address this clinical concern. We established a panel of PDX models derived from gastric cancer patients, each representing a distinct range of HER2 expression levels. These models were subsequently treated with HER2-targeted drugs, including Trastuzumab (Herceptin), Margetuximab, and Enhertu. Our results demonstrated differential drug responses corresponding to the varying HER2 expression levels in the PDX samples. Trastuzumab, Margetuximab, and Enhertu exhibited distinct efficacies across the HER2 expression spectrum. This underscores the significance of selecting the right therapeutic approach based on the HER2 status of individual patient. Our collection of PDX models with differential HER2 expression levels provides a robust platform for preclinical evaluation of HER2-targeted therapies. This approach not only addresses the clinical challenge of HER2 heterogeneity but also facilitates a more personalized and effective treatment strategy for gastric cancer patients. Citation Format: Xiaoliu Yang, Shiying Guo, Huiyi Wang, Xu Su, Jing Zhao, Xiang Gao, Cunxiang Ju. Patient-derived xenograft (PDX) models with differential HER2 expression for preclinical evaluation of HER2-targeted therapies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4190.

Abstract 868: A novel unsupervised machine learning model applied to neuroblastoma single-cell RNA-seq data reveals a drug-induced mesenchymal-like gene expression program

Abstract High-risk neuroblastoma is one of the most difficult to treat pediatric cancers, with a survival rate of only about 50% and significant long-term consequences from current chemotherapy. Preclinical models such as cell lines and mice are the backbone of drug development and experimental-mechanistic oncology, but the development of new treatments is hampered in part by a lack of understanding of the direct clinical relevance of such data collected in preclinical models. Despite this, few formal methods have been developed to determine how these various models represent/resemble primary patient tumors. Here, we present the first comprehensive single-cell RNA-seq analysis of neuroblastoma across an extensive cohort of patient tumors and a variety of preclinical model systems (n = 126 total samples assembled - the largest cohort of its kind). By building an innovative unsupervised machine learning method, which we term “automatic consensus nonnegative matrix factorization” (acNMF), we have integrated and contrasted the transcriptional landscapes of patient tumors with those of cell lines, patient-derived xenografts (PDX), and genetic mouse models (GEMM).Using these tools, we discovered the dominant adrenergic gene expression programs found in neuroblastoma patient tumors were preserved across all preclinical models. However, the presumptive chemo-resistant mesenchymal-like programs, while identifiable in cell lines, were primarily restricted to subpopulations of cancer-associated fibroblasts and Schwann-like cells in vivo. Surprisingly, a mesenchymal-like program could be acutely chemotherapy-induced in GEMM and was evident in pre-treated patient and PDX samples, suggesting a previously uncharacterized mechanism of therapy escape. In addition to these core findings, our computational tools were able to further delineate the classical neuroblastoma adrenergic and mesenchymal gene expression programs, discovering for example, novel subpopulations of cancer associated fibroblasts. These behaviors were conserved across tumors and most preclinical models, which we validated by RNA in situ hybridization, which is a high resolution ultra sensitive spatial transcriptomics technology. Our work cautions against overreliance on traditional preclinical models without recognizing their limitations and we offer a nuanced, high-resolution view of neuroblastoma pre-clinical systems for advancing therapeutic development. We have launched an open-source web resource, featuring this integrated map to aid the scientific community in further exploration of these data and hypothesis generation (available at http://pscb.stjude.org). Citation Format: Rich Chapple, Xueying Liu, Sivaraman Natarajan, Margaret I. Alexander, Yuna Kim, Anand Patel, Christy W. LaFlamme, Min Pan, William C. Wright, Hyeong-Min Lee, Yinwen Zhang, Meifen Lu, Selene C. Koo, Courtney Long, John Harper, Chandra Savage, Melissa D. Johnson, Thomas Confer, Walter J. Akers, Michael A. Dyer, Heather Sheppard, John Easton, Paul Geeleher. A novel unsupervised machine learning model applied to neuroblastoma single-cell RNA-seq data reveals a drug-induced mesenchymal-like gene expression program [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 868.

Abstract 5987: Targeting the spliceosome in high-risk B-cell acute lymphoblastic leukemia

Abstract The outcome for children with high-risk B-cell acute lymphoblastic leukemia (B-ALL) is poor. Disease relapse is speculated to be due to leukemia cells escaping treatment. Our group discovered a unique cell subpopulation in B-ALL that has the capacity to initiate leukemia and is resistant to treatment. Transcriptome studies of this cell population highlight a unique RNA regulation process, specifically RNA splicing. In B-ALL, dysregulated splicing is reported to be associated with drug resistance and disease relapse. SF3B1 is a core component of the spliceosome and an essential protein in the RNA splicing process. SF3B1 inhibition is therapeutic in many cancers. In our current studies, we investigated the therapeutic potential of pladienolide B (Plad-B), an SF3B1 inhibitor, in B-ALL. SF3B1 protein expression was significantly higher in cell lines (Reh and JM1) and 28 primary B-ALL samples (14 each for standard-risk and high-risk), regardless of the risk group, than in normal B-cells (NBs) and hematopoietic stem cells (HSCs). Plad-B showed significant dose-dependent cytotoxicity in the cell lines with IC50 of 1.2nM and 0.6nM, respectively, and three harvested high-risk patient-derived xenograft (PDX) samples with almost the same IC50 as the cell lines. Plad-B did not show cytotoxicity in NBs and HSCs at the same tested concentrations. In vivo efficacy of Plad-B was tested using an Reh xenograft mouse model and a high-risk PDX model. Plad-B, as a single drug treatment, significantly prolonged survival in both models (p<0.01). G2/M cell cycle arrest and apoptosis induction were observed at 24 hours after Plad-B treatment in Reh and JM1. Splicing events were examined by RNA-seq in the treated cells at different time points (15 min, 30 min, and 60 min). Plad B demonstrated rapid splicing inhibition as early as 15 min post-treatment and, at 60 min, splicing was inhibited in 1,669 genes, including apoptosis-associated genes. 2,625 differential splicing events were observed in these genes with ~96% from intron retention and exon skipping. Furthermore, 202 genes showed significant changes in their expression over this time frame. Most of the genes were rapidly downregulated in the treatment group compared to the control, and some were slowly downregulated. There were also some genes which were upregulated, either rapidly or slowly, compared to the control. We demonstrated that Plad-B induced short pro-apoptotic spliced isoforms, instead of anti-apoptotic forms, in BCL2L1 and MCL-1, as early as 60 min after treatment. In conclusion, these data demonstrated the therapeutic potential of SF3B1 inhibition in high-risk B-ALL. Plad-B rapidly inhibited splicing in many genes and downregulated the anti-apoptotic forms of apoptosis-associated genes, leading to cell apoptosis. In future studies, we will identify the downstream targets of Plad-B and further investigate the mechanism of rapid apoptosis induction by SF3B1 inhibition. Citation Format: Yuki Murakami, Hiroaki Konishi, Clifford Tepper, John McPherson, Noriko Satake. Targeting the spliceosome in high-risk B-cell acute lymphoblastic leukemia [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5987.

Enhanced potency of immunotherapy against B-cell precursor acute lymphoblastic leukemia by combination of an Fc-engineered CD19 antibody and CD47 blockade.

CD19-directed immunotherapy has become a cornerstone in the therapy of B-cell precursor acute lymphoblastic leukemia (BCP-ALL). CD19-directed cellular and antibody-based therapeutics have entered therapy of primary and relapsed disease and contributed to improved outcomes in relapsed disease and lower therapy toxicity. However, efficacy remains limited in many cases due to a lack of therapy response, short remission phases, or antigen escape. Here, BCP-ALL cell lines, patient-derived xenograft (PDX) samples, human macrophages, and an in vivo transplantation model in NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) mice were used to examine the therapeutic potency of a CD19 antibody Fc-engineered for improved effector cell recruitment (CD19-DE) and antibody-dependent cellular phagocytosis (ADCP), in combination with a novel modified CD47 antibody (Hu5F9-IgG2σ). For the in vivo model, only samples refractory to CD19-DE monotherapy were chosen. Hu5F9-IgG2σ enhanced ADCP by CD19-DE in various BCP-ALL cell line models with varying CD19 surface expression and cytogenetic backgrounds, two of which contained the KMT2A-AFF1 fusion. Also, the antibody combination was efficient in inducing ADCP by human macrophages in pediatric PDX samples with and adult samples with and without KMT2A-rearrangement in vitro. In a randomized phase 2-like PDX trial using seven KMT2A-rearranged BCP-ALL samples in NSG mice, the CD19/CD47 antibody combination proved highly efficient. Our findings support that the efficacy of Fc-engineered CD19 antibodies may be substantially enhanced by a combination with CD47 blockade. This suggests that the combination may be a promising therapy option for BCP-ALL, especially in relapsed patients and/or patients refractory to CD19-directed therapy.

Abstract C060: Characterization of a collection of patient-derived xenograft models for pancreatic ductal adenocarcinoma translational research

Abstract To conduct successful translational research for pancreatic ductal adenocarcinoma (PDAC), it is imperative to have preclinical models that accurately reflect the intricate molecular, genomic, and histological features of the disease. Patient-derived xenografts (PDXs) have emerged as invaluable platforms for studying tumor biology and personalized therapeutic intervention. Here we present a collection of 32 PDAC PDX models established and fully characterized in our institution. This PDAC PDX collection was generated with tissue from both primary and metastatic lesions, comprehensively representing the disease spectrum. Genomic profiling using next-generation sequencing revealed a remarkable concordance between the PDX samples and their original patient tumor tissue. Notably, identifying common driver mutations associated with PDAC further validates the clinical relevance of this PDX collection. Moreover, histopathologic examination confirmed the similarity between PDX tumor samples and their original patient tissue samples. Interestingly, we also optimized a protocol to generate tumoroids from PDAC PDXs. Leveraging the broad spectrum of disease represented in our collection will facilitate addressing critical research questions, identifying novel therapeutic targets, and ultimately improving patient outcomes of PDAC. Citation Format: Mariana Yáñez, Enrique Javier Arenas, Teresa Macarulla, Tian Tian, Joaquín Arribas, Sofía Llorente, Carles Fabregat, Marta Escorihuela, Florian Castet, Jessica Querol, Maria Teresa Salcedo. Characterization of a collection of patient-derived xenograft models for pancreatic ductal adenocarcinoma translational research [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Pancreatic Cancer; 2023 Sep 27-30; Boston, Massachusetts. Philadelphia (PA): AACR; Cancer Res 2024;84(2 Suppl):Abstract nr C060.

Vascularized tissue on mesh-assisted platform (VT-MAP): a novel approach for diverse organoid size culture and tailored cancer drug response analysis.

This study presents the vascularized tissue on mesh-assisted platform (VT-MAP), a novel microfluidic in vitro model that uses an open microfluidic principle for cultivating vascularized organoids. Addressing the gap in 3D high-throughput platforms for drug response analysis, the VT-MAP can host tumor clusters of various sizes, allowing for precise, size-dependent drug interaction assessments. Key features include capability for forming versatile co-culture conditions (EC, fibroblasts and colon cancer organoids) that enhance tumor organoid viability and a perfusable vessel network that ensures efficient drug delivery and maintenance of organoid health. The VT-MAP enables the culture and analysis of organoids across a diverse size spectrum, from tens of microns to several millimeters. The VT-MAP addresses the inconsistencies in traditional organoid testing related to organoid size, which significantly impacts drug response and viability. Its ability to handle various organoid sizes leads to results that more accurately reflect patient-derived xenograft (PDX) models and differ markedly from traditional in vitro well plate-based methods. We introduce a novel image analysis algorithm that allows for quantitative analysis of organoid size-dependent drug responses, marking a significant step forward in replicating PDX models. The PDX sample from a positive responder exhibited a significant reduction in cell viability across all organoid sizes when exposed to chemotherapeutic agents (5-FU, oxaliplatin, and irinotecan), as expected for cytotoxic drugs. In sharp contrast, PDX samples of a negative responder showed little to no change in viability in smaller clusters and only a slight reduction in larger clusters. This differential response, accurately replicated in the VT-MAP, underscores its ability to generate data that align with PDX models and in vivo findings. Its capacity to handle various organoid sizes leads to results that more accurately reflect PDX models and differ markedly from traditional in vitro methods. The platform's distinct advantage lies in demonstrating how organoid size can critically influence drug response, revealing insights into cancer biology previously unattainable with conventional techniques.

The E3 ubiquitin ligase NEDD4 regulates chemoresistance to 5-fluorouracil in colorectal cancer cells by altering JNK signalling

Colorectal cancer (CRC) is the second leading cause of cancer deaths. Though chemotherapy is the main treatment option for advanced CRC, patients invariably acquire resistance to chemotherapeutic drugs and fail to respond to the therapy. Although understanding the mechanisms regulating chemoresistance has been a focus of intense research to manage this challenge, the pathways governing resistance to drugs are poorly understood. In this study, we provide evidence for the role of ubiquitin ligase NEDD4 in resistance developed against the most commonly used CRC chemotherapeutic drug 5-fluorouracil (5-FU). A marked reduction in NEDD4 protein abundance was observed in a panel of CRC cell lines and patient-derived xenograft samples that were resistant to 5-FU. Knockout of NEDD4 in CRC cells protected them from 5-FU-mediated apoptosis but not oxaliplatin or irinotecan. Furthermore, NEDD4 depletion in CRC cells reduced proliferation, colony-forming abilities and tumour growth in mice. Follow-up biochemical analysis highlighted the inhibition of the JNK signalling pathway in NEDD4-deficient cells. Treatment with the JNK activator hesperidin in NEDD4 knockout cells sensitised the CRC cells against 5-FU. Overall, we show that NEDD4 regulates cell proliferation, colony formation, tumour growth and 5-FU chemoresistance in CRC cells.

Abstract C047: Comprehensive curation of label-free proteomics data for 186 patient-derived xenograft (PDX) models in HuBase™

Abstract Introduction. HuBase™, a Crown Bioscience Inc. database, hosts pharmacological and genomic data from over 2500 patient-derived xenograft (PDX) models. It is widely used in pharmacology for model selection and biomarker analysis. Recognizing the pivotal role of proteomics data in unraveling intricate molecular mechanisms underlying biological processes and diseases, we acknowledge the challenge posed by the significant sequence similarity between mouse and human homologs in PDX samples, which hampers the accurate characterization of proteomics data using traditional run-then-separate protocols. In this study, building upon our previously published separate-then-run protocol1, we established a label-free proteomics quantification pipeline to accurately determine the protein expression level of 186 PDX models in a systematic manner. The resulting proteomics data have been seamlessly integrated into HuBase™, constituting the first comprehensive collection of PDX proteomics data derived from mouse cell-depleted samples. Methods. Mouse stromal cells were removed from the 186 PDX models with Mouse Cell Depletion Kit. The mouse depleted samples were analyzed using the 4D Data-Independent Acquisition (4D-DIA) technology. Quality control (QC) samples were measured at every 15 MS runs to assess technical variability. Raw MS data was processed with Spectronaut® 17 to generate label-free quantification (LFQ), which were log-transformed for subsequent analysis. To address batch effects, the Combat method in the sva R package was used on 2 different batches. Additionally, a subset of 10 models without mouse cell removal, underwent the same proteomics quantification pipeline for comparative analysis with their mouse depleted counterparts. Results. We quantified the expression levels of over 12,000 proteins per sample, with a significant number of proteins showing positive correlations with mRNA expression in all PDX samples. After the batch effect correction, we observed no systematic discrepancies between the two batches, and the median intensity of QC samples remained comparable. Through a comparative analysis of 10 pairs of PDX models with and without mouse stromal cell removal, we found differentially expressed proteins were enriched in the extracellular matrix for non-mouse depleted samples, which could be attributed to the upregulation of these proteins in mouse stromal cells. The significant presence of some mouse proteins underscores the importance of removing mouse stromal cells prior to proteomic characterization to avoid adversely confounding subsequent analyses leading to misinterpretations. Conclusion. Powered by state-of-the-art 4D-DIA technology, we have curated the first comprehensive PDX proteomics dataset derived from mouse-depleted samples, offering additional value for PDX model selection and exploration of potential biomarkers. References. 1. Shi, Z., Mao, B., Chen, X., Hao, P. & Guo, S. Mouse Stromal Cells Confound Proteomic Characterization and Quantification of Xenograft Models. Cancer Research Communications 3, 202–214 (2023). Citation Format: Hengyuan Liu, Xiaobo Chen, Sheng Guo, Binchen Mao, Rekha Pal. Comprehensive curation of label-free proteomics data for 186 patient-derived xenograft (PDX) models in HuBase™ [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr C047.

Identification of vulnerabilities for targeting BCL-2 family members in T-Cell Acute Lymphoblastic Leukemia

T-Cell acute lymphoblastic leukemia (T-ALL) is a disease caused by the malignant transformation of T-cell lineage progenitors. With the use of intensive chemotherapies survival rates have improved, but outcomes particularly of relapsed patients remain poor. In addition, currently used intensive chemotherapies are associated with high rates of treatment-related morbidity and mortality, emphasizing the need to develop new and improved therapies. The intrinsic apoptosis pathway, one of the key pathways controlling cell death, is dysregulated in many cancers and contributes to leukemogenesis and treatment failure. The main steps of this pathway are controlled by proteins of the B-cell lymphoma (BCL-2) family at the outer mitochondrial membrane. Hence, targeting this pathway by BH3-mimetics has emerged as an effective new treatment strategy in different cancers. Venetoclax is a selective BCL-2 inhibitor and is successfully used in CLL and AML, but heterogeneous sensitivity to venetoclax has been described in ALL and inhibitors of other BCL-2 family members including BCL-XL-selective A-1331852 and MCL-1 selective AZD5991 have been developed. Moreover, a dual inhibitor of the anti-apoptotic BCL-2 family members BCL-2 and BCL-XL (AZD4320) with its dendrimer conjugate (AZD0466) has recently shown anti-tumor activity in hematologic cancer models with manageable toxicity. The aims of this study are to evaluate susceptibilities of T-ALL to inhibitors of BCL-2 family proteins, to identify markers of response, to elucidate mechanisms of action and to assess combination effects. First, we analyzed the anti-leukemia activities of inhibitors targeting BCL-2 (venetoclax), BCL-XL (A-1331852) and MCL-1 (AZD5991) and of the dual BCL-2/BCL-XL inhibitor AZD4320 in T-ALL cell lines (N=6) and a series of patient-derived xenograft (PDX) samples (N=8). Inhibition of MCL-1 alone was not effective in the cell lines tested (EC50 >1 µM), but heterogeneous sensitivity was found in PDX samples (EC50 <1µM in 3/8 samples). In cell lines, sensitivity for BCL-2 inhibition was exclusively found in a cell line with an early T-cell precursor ALL phenotype (Loucy). Similar to cell lines, we found insensitivity for BCL-2 inhibition in all PDX samples tested (N=8). For inhibition of BCL-XL, we found heterogeneous sensitivities with 4/6 cell lines and 5/8 PDX samples showing an EC 50 <1 µM. Interestingly, BH3-profiling confirmed BCL-2 dependence in Loucy and BCL-XL dependence in the other cell lines. Importantly, dual inhibition of BCL-2 and BCL-XL showed clear responses in cell lines and PDX samples (median EC 50s 796.1 nM and 721.6 nM). Notably, sensitivity was associated with priming of the BH3-peptide BAD (r s=0.89, p=0.03), providing a marker of response. We next sought to elucidate the molecular mechanisms by which different BH3-mimetics induce cell death in T-ALL. Analyzing protein interactions of BCL-2 family members by co-immunoprecipitation studies, we found that AZD4230 acts on-target by releasing BIM from BCL-2 and BCL-XL, but with compensatory increased protein complexes of BIM and MCL-1. Accordingly, combined inhibition of BCL-2 and BCL-XL with inhibition of MCL-1 leads to release of BIM from all three anti-apoptotic proteins enabling downstream apoptosis signaling. These findings suggest that combined inhibition of BCL-2 and BCL-XL together with MCL-1 may provide synergistic benefits. To further validate these data, we applied dynamic BH3-profiling and determined dependencies of T-ALL samples on BCL-2 family members in response to BCL-2/BCL-XL inhibition. Here, we found that the MCL-1 dependence strongly increases upon BCL-2/BCL-XL inhibition, suggesting MCL-1 as a target for combined inhibition. Based on these results, we performed dose-response matrix analyses to test the effects of combined BCL-2/BCL-XL and MCL-1 inhibition. Most importantly, combined inhibition showed synergism in all cell lines and PDX samples tested (Bliss scores >10), including leukemias being insensitive to single compounds. Taken together, we found vulnerabilities of T-ALL to BCL-2 family inhibition, particularly to dual BCL-2/BCL-XL inhibition by AZD4320, and we demonstrated on-target activities. Using BH3-profiling, we identified BAD-priming as a marker of response for AZD4320 and MCL-1 dependence as a resistance mechanism that can be targeted by combination treatment, suggesting further clinical evaluation.

Drug Response Profiling of Childhood T-Cell Acute Lymphoblastic Leukemia to Inform Functional Precision Medicine

Background/Aims: Despite the advancement in risk-directed, chemotherapy-based protocols, the prognosis of children with T-cell acute lymphoblastic leukemia (T-ALL) especially for those who were refractory to existing treatments or relapsed from remissions remains poor. The complexity of mutational landscape and lack of actionable genomic lesions collectively pose significant challenges for implementing targeted therapies based solely on genomics. To address this issue, this study seeks to utilize drug sensitivity and resistance test (DSRT) to directly assess the ex vivo response of primary T-lymphoblasts to a portfolio of targeted agents, aiming to establish its drug sensitivity profile and identify drug candidates of immediate clinical relevance to tailor functional precision medicine. Methods: A serum-free, mesenchymal stem cell (MSC)-based culture system was adopted to perform ex vivo DSRT on primary or patient-derived xenograft (PDX) samples of pediatric T-ALL. The latter was generated by transplanting NSG mice with paucicellular specimens with splenocytes collected at leukemia engraftment as the sources of patient-derived T-lymphoblasts. The 42-drug panel consisted of 35 targeted agents (1 in Phase I trials, 6 in Phase II trials, 6 in Phase III trials, 22 FDA-approved) and 7 standard chemotherapeutics. The DSRT tested six serial drug concentrations ranging from 0.1 to 10,000 nM. After 96 hours of co-culturing with MSCs, T lymphoblasts were identified by CD7 staining, while apoptotic cells were characterized by Annexin V/7-AAD. The half maximal inhibitory concentration (IC50) of individual drugs were curated from the dose-response curves. Unsupervised clustering was performed to map the overall drug response pattern. A case series of relapsed or refractory T-ALL were managed with targeted therapies guided by drug profiling results and evaluated for clinical responses. Results: DSRT was conducted on 19 primary and 14 PDX pediatric T-ALL samples from 20 patients. For primary samples, two had been tested for 42 drugs, ten were tested for 16 drugs and the rest underwent testing with varying numbers of drugs depending on cellularity and sample quality. All PDX samples were subjected to testing for 42 drugs. PDX samples retained the original pattern of drug response in matched patient materials (R 2 = 0.77-0.94, 8 paired samples). Unsupervised clustering identified 5 distinct groups of drug response: (i) highly active compounds with median IC50 <10 nM (4 drugs including the proteasome inhibitors bortezomib and carfizomib, BCL2 inhibitor navitoclax and HDAC inhibitor panobinostat); (ii) generally active compounds (median IC50 <230 nM, 10 drugs), including BCL2 inhibitors, MCL1 inhibitors, JAK2/3 inhibitors and MDM2 inhibitors; (iii) compounds with bimodal activities (wide IC50 ranges, 4 drugs), including dasatinib (tyrosine kinase inhibitor), sirolimus and everolimus (mTOR inhibitors) and standard chemotherapeutics; (iv) generally inactive drugs with sporadic exceptions (12 drugs); and (v) completely inactive compounds (12 drugs). Four children with T-ALL received treatment according to drug profiling results. Two patients were in first relapse, one in third relapse and one in a non-remission state. Three of these patients received bortezomib treatment, while the fourth patient received venetoclax. All patients responded to the treatments with complete or partial remission. One patient was bridged to allogeneic stem cell transplantation, and two were bridged to CD7-directed CAR therapies. Conclusions: Our study unveiled a pediatric-specific drug response profile for T-ALL and highlighted opportunistic drug candidates with clinical relevance. The successful implementation of functional precision medicine in our clinic elucidated its practicality and potential to enable personalized treatment for high-risk pediatric T-ALL.

Secondary Lesions and Sensitivity to Signaling Inhibitors of Pediatric iAMP21 B-Cell Precursor Acute Lymphoblastic Leukemia

Introduction Intrachromosomal amplification of chromosome 21 (iAMP21) B-cell precursor acute lymphoblastic leukemia (BCP-ALL) in children is a high-risk subtype for which targeted drugs are lacking. In this study we aimed to determine the frequency of secondary lesions and investigated the cellular sensitivity for candidate targeted drugs. Methods We performed total RNA sequencing on 28 iAMP21 and 28 B-other (negative for sentinel fusion genes) pediatric samples to determine the frequency of secondary lesions in newly diagnosed patients. A panel of 18 patient derived xenografts (PDX) of 8 primary iAMP21 ALL samples was generated, and secondary lesions were validated by PCR, RT-PCR, and whole exome sequencing. To test sensitivity, primary or PDX cells were exposed ex vivo to a concentration range of gilteritinib (FLT3 inhibitor), trametinib (MEK1/2 inhibitor), or ruxolitinib (JAK1/2 inhibitor). Results Secondary lesions in the cytokine receptor gene FLT3 were enriched in iAMP21 compared with B-other ALL including internal tandem duplications (ITD) and other activating lesions (50.0% vs. 10.7%, p=0.003). Lesions in genes encoding cytokine receptors CRLF2 and IL7R had a similar frequency between iAMP21 and B-other cases (25% vs. 17.9%, p=0.75 and 7.1% vs. 0%, p=0.49 respectively). Inactivating lesions in SH2B3, the downstream negative regulator of JAK/STAT and FLT3 signalling, were more frequent in iAMP21 cases vs. B-other (46.4% vs. 7.1%, p=0.002), while the frequency of JAK1 and JAK2 mutations did not differ (3.6 vs. 0% and 10.7 vs. 10.7% p = 1 for both). All SH2B3, CRLF2 and JAK lesions were retained in PDX samples, whereas in contrast FLT3-ITD was retained in only 2 of 5 PDX. Gilteritinib sensitivity did not differ between iAMP21 and B-other cases (median LC50 1.24 µM, range 0.33 to 4.85 µM vs. median LC50 1.21 µM, range 1.20 to 1.3 µM, p=0.95). Grouping iAMP21 cases by FLT3 and SH2B3 status, samples with both FLT3-ITD and SH2B3 lesion had the highest sensitivity to gilteritinib (median LC50 0.39 µM, range 0.35 to 0.43 µM). Samples with only FLT3 or SH2B3 lesion did not show increased sensitivity compared to those without a lesion (p>0.5 for both). Median ruxolitinib sensitivity did not statistically differ between iAMP21 and B-other cases (median LC50 8.38 µM, range 1.00 to >10 µM, vs. median LC50 0.48, range 0.21 to >10 µM; p=1) although extreme resistance to ruxolitinib seemed more frequent in iAMP21 cases (6 out of 12 cases) and was not related to FLT3 or SH2B3 status. CRLF2-rearranged ( CRLF2r) iAMP21 cases were slightly more sensitive to ruxolitinib than those lacking CRLF2r, although this difference did not reach significance (median LC50 4.72 µM, range 1.0-4.85 vs. median LC50 >10 µM, range 2.66 to >10 µM; p=0.08). The highest sensitivity was found in the only case with both a CRLF2r and a JAK1 mutation (LC50 of 1.00 µM). A large variation was present in trametinib sensitivity amongst both iAMP21 and B-other cases (median LC50 0.11 µM, range 0.017 to >5 µM vs. median LC50 0.18, range 0.018 to >5 µM; p=0.72). More than half of iAMP21 cases were sensitive irrespective of RAS-pathway lesion status. Conclusion iAMP21 leukemias are enriched in FLT3 and in SH2B3 lesions which, when co-occurring, affect sensitivity to FLT3 inhibition by gilteritinib but do not affect JAK-inhibition by ruxolitinib. This suggests these lesions act synergistically and might bypass downstream JAK/STAT signalling. This might also explain the observed sensitivity to RAS-pathway inhibition irrespective of secondary lesions. These results suggest that further research into FLT3 and RAS signalling inhibitors might lead to better treatment options for pediatric iAMP21 BCP-ALL.

Phenotypic Drug Response Profiling Identifies Asparaginase-Based Synergistic Combinations for Very High Risk Acute Lymphoblastic Leukaemia

In acute lymphoblastic leukaemia (ALL), empirical multi-drug chemotherapy combinations lead to >85% cure rates in children. But 10-15% of patients show suboptimal response and as drugs are dosed to tolerance, further intensification is not possible. An exception is the key antileukaemic drug, asparaginase (ASNase), which is not associated with dose-related toxicity. We evaluated the use of ex-vivo co-culture based drug response profiling (DRP) to identify novel synergistic and effective drug combinations with ASNase for patients with suboptimal responses to standard chemotherapy in ALL. Patient-derived xenograft (PDX) ALL cells from 5 patients with high-risk B-ALL (3 B-other, 2 IKZF plus) were treated with 120 drugs in 5-point serial dilutions in combination with or without an IC 50 dose of ASNase, in co-cultures with hTERT-immortalized primary bone marrow stromal cells. Eight clinically relevant agents with promising sensitizing activity as identified by decreased area-under-the-curve with addition of ASNase were identified (SEL, selinexor; ELTA, eltanexor; VEN, venetoclax; BZM, bortezomib; navitoclax; carfilzomib; mitoxantrone, birinapant). We validated combinations of ASNase with the 8 selected drugs plus 4 drugs used in induction therapy (prednisolone, dexamethasone, vincristine and daunorubicin) using a 4x4 drug matrix in 15 high-risk ALL PDX samples. Synergy scores were calculated using the zero-interaction potency (ZIP) model. Synergy (ZIP score ≥10) or additivity (ZIP score >0 and <10) with ASNase was identified for exportin-1 inhibitors (XPO1i; selinexor, eltanexor), venetoclax and proteosomal inhibitors (bortezomib, carfilzomib) in 100% (15/15), 86.6% (13/15) and 80-86.6% (12/15 and 13/15) samples respectively. Antagonism (ZIP score <0) was noted in more than one-third (6/15) of samples with ASNase in combination with steroids or vincristine. Three-drug combinations were evaluated ex-vivo in these 15 PDX samples. Synergy was identified with addition of low doses of BZM or VEN to the ASNase-SEL combination (ASNase-SEL-BZM and ASNase-SEL-VEN). Next, 4-6 mice per treatment arm were transplanted intravenously with 1 million ALL cells from 4 high-risk B-ALL patients. Randomized cohorts were treated with vehicle, ASNase-BZM-ELTA (ABE) and ASNase-dexamethasone-vincristine (ADV). Leukemic burden was assessed weekly using flowcytometry. The mice tolerated both treatment arms well. Platelet counts recovered within one week after an initial drop in the ABE arm. Both treatment groups had significant improvement in event free survival (EFS) as compared to the control group (ABE p-value 0.0082; ADV p-value 0.0325) though no significant differences were noted between the 2 treatment arms (p-value 0.8971). Three patients with high-risk medullary relapsed/refractory BCP-ALL, 2 not in morphological remission and 1 at time of very early relapse, received a combination of ASNase-BZM-VEN-SEL (ABVS) after consent. One patient for whom DRP suggested sensitivity to both SEL and VEN achieved molecular remission (molecular minimal residual disease < 10 -5) after one course of therapy remaining in complete remission 11 months later. Two patients with sensitivity to SEL but not to VEN as assessed by DRP, had progressive disease. ABVS was well tolerated in these patients with no grade 3-4 adverse events. Our results identify XPO1 inhibitors as a new class of drugs with promising anti-leukaemic activity in BCP-ALL. Co-culture based DRP shows promise in identifying alternative novel sensitive synergistic combinations ex-vivo with decreased toxicity for patients who have an inadequate response to standard therapy.

Calibrated Release IL15 Bivalent CD33 and/or FLT3 and NOT Emcn Logic Gated Gene Circuit CAR-NK Cell Therapy (SENTI-202) in Venetoclax Resistant Patient Derived Xenograft Acute Myeloid Leukemia Models

Venetoclax (VEN) as combination therapy has improved response rates and overall survival of patients with acute myeloid leukemia (AML). However, once AML relapse prognosis is dire with a 5.3 month median survival (Brandwein, 2020). While chimeric antigen receptor (CAR) cell therapies have revolutionized the treatment landscape for lymphoid hematological malignancies, development of such therapies for AML has been challenging, in part due to the heterogeneity of the disease. Currently, there is a paucity of individual AML targets that are consistently expressed across AML subpopulations. Furthermore, the expression of these AML targets is not restricted to tumor cell populations, often resulting in off-tumor toxicity against healthy cell populations. SENTI-202 employs OR and NOT Logic Gating to overcome these challenges. SENTI-202 represents an innovative preclinical CAR-NK cell therapy, engineered to exploit a powerful CD33 OR FLT3 NOT EMCN Logic Gate gene circuit, in conjunction with calibrated release IL-15 (crIL15) expression. The CD33 OR FLT3 (OR GATE) activating CAR (aCAR) concurrently targets two AML antigens, thus permitting a broader therapeutic window against AML LSCs (CD33+/-, FLT3+), blasts (CD33+, FLT3+/-), and more specifically, VEN-resistant (VEN-r) AML with a CD33+ monocytic phenotype (Pei, 2020). The NOT EMCN (NOT GATE) inhibitory CAR (iCAR) is a pivotal safeguard, selectively shielding healthy EMCN+ Hematopoietic Stem Cells and Progenitor Cells (HSCs/PCs) from potential off-tumor toxicity. Previously, SENTI-202 exhibited remarkable efficacy in targeting and eliminating CD33 and/or FLT3 expressing AML cell lines, while concurrently safeguarding healthy HSCs/PCs (Garrison et al., ASH, 2022). To further validate the efficacy and broaden the utility of SENTI-202, we conducted assessments on clinically relevant preclinical AML models, along with VEN-r AML patient-derived xenograft (PDX) models. We first assessed the in vitro cytotoxicity and cytokine profiles of CAR-NK cells possessing the SENTI-202 OR Gate and crIL15 against primary R/R AML patient samples (n=11), where we observed significantly higher cytotoxicity compared to unengineered NK-cells (68% vs 19.7%) (p=0.0005, E:T ratio 1:1 for 24h) and increased TNF-a and IFN-g secretion. To further explore CAR-NK cell anti-tumor activity, we screened 10 different PDX samples and discovered that while the median frequency of CD33 and/or FLT3 expression was 99.9% (97.48-100), individual CD33 and FLT3 expression level varied, underscoring AML proteomic heterogeneity and the clinical need for OR Gate approaches. Our initial PDX studies used clinically relevant VEN/HMA-r AML PDX cells (designated as PDX1) obtained from a patient with FLT3-ITD, GATA2, and NRAS mutations, who initially responded to VEN/decitabine treatment but later relapsed. To facilitate in vivo tracking of AML and disease progression, the VEN-r PDX cells were engineered with Akaluc, an exceptionally brighter luciferase compared to conventional luciferase. Following confirmation of AML engraftment in bone marrow through bioluminescence imaging (BLI), the CAR-NK cells, along with unengineered NK-cells, were infused at a dose of 30 million cells per infusion, once a week for three consecutive weeks. The persistence of the CAR-NK cells was detectable for up to 2 weeks after the last infusion. Notably, the infusion of CAR-NK cells resulted in the clearance of bone marrow (BM) disease as evident from BLI one week after the second infusion of CAR-NK cells while unengineered NK-cells failed to clear bone marrow disease. Assessment of CAR-NK cell efficacy in two different AML PDX models is ongoing, along with additional in vitro and in vivo studies that further demonstrate the ability of the SENTI-202 NOT EMCN Logic Gate to significantly protect healthy HSCs from off-tumor toxicity. In conclusion, CAR-NK cells expressing the SENTI-202 OR Gate and crIL15 demonstrate robust anti-tumor activity against primary AML cells within in vitro cytotoxicity assays and in vivo AML PDX models, including VEN-r AML. These promising preclinical results highlight the potential of SENTI-202 as a targeted and selective therapy for AML, offering a new avenue for overcoming the challenges posed by the heterogeneity of the AML proteomic landscape.

Unraveling Diverse Mechanisms of Complex Structural Variant Interactions through Multiomic Data in Multiple Myeloma

Introduction: Structural variants (SVs) play a role in proliferation of plasma cells that characterizes multiple myeloma (MM). Several hallmark SVs have been identified in MM, including translocations involving the immunoglobulin heavy chain locus and several partner genes, as well as deletions and duplications affecting key genes such as TP53, RB1, and MYC. Complex SVs such as templated insertions, chromothripsis, and chromoplexy are associated with disease progression but have not yet systematically been studied. Methods: Paired short-read (Illumina, average depth 75x), long-read HiFi (PacBio, 16x), optical genome mapping (OGM; Bionano Genomics, 341x) and Micro-C (Cantata Bio, 45x) data were sequenced from 8 relapsed MM patient-derived xenograft samples and 2 cell lines, covering t(11;14), t(4;14), and hyperdiploid subgroups. Reads were aligned to the hg38 reference genome while complex SVs (chromothripsis, chromoplexy and templated insertion) were systematically identified and accurately reconstructed by combining short-read (Manta), HiFi (pbsv), OGM, and Micro-C data (EagleC). Chromosome organization including loops, topologically associated domains (TADs) and A/B compartments were called from Micro-C data (Mustache and Juicer). Haplotype phasing was conducted using small variants detected from HiFi data. De novo assembly of the genome was conducted using HiFi and Micro-C data and compared to the reference genome. Reconstructed derivative chromosome interactions due to complex SVs were further identified by aligning Micro-C reads to the newly assembled genomes. Previously identified super-enhancers from H3K27Ac ChIP-seq from MM cell lines were mapped onto these SVs and expression of target genes assessed by RNA-seq. Results: 1,340 SVs were found in 10 samples using OGM of which 828 were identified in short-read WGS data. 16 complex SVs including chromothripsis, chromoplexy and templated insertions were detected in 10 samples, and we describe two (Fig.1) in detail here. A complex templated insertion consisting of a t(2;6;8) was found in a t(11;14) PDX sample with copy number gains at breakpoints which could not be resolved from short-read data alone. OGM data revealed a single contig consisting of DNA from chromosomes 2, 6, and 8. Micro-C data revealed interactions between these chromosomes, specifically the gained regions on chromosomes 2 and 8 showed more interactions with chromosome 6 than with other chromosomes, indicating insertions into chromosome 6. As a result, both MYC at chromosome 8 and GALM at chromosome 2 were inserted into the TXNDC5 locus, containing a known MM super-enhancer, leading to the over-expression of both MYC and GALM (log 2FC=1.7 and 1.0) in this sample. A de novo genome assembly indicated neo-TAD formation encompassing the super-enhancer and both genes. In coMMpass dataset, high expression of GALM and MYC are significantly associated with inferior overall (p=0.002 and 0.05, logrank test) and progression-free survival (p=0.01 and 0.09). Another complex templated insertion consisting of a t(3;4;16) was identified in a hyperdiploid PDX sample, again with copy number gains at breakpoints. Although the short-read data indicated an assembly similar to the previous templated insertion SV, the OGM and Micro-C data indicated a 3-way chromosome swap where the short arm of chromosome 3 was translocated to the short arm of chromosome 4, the short arm of chromosome 4 was translocated to the long arm of chromosome 16, and the long arm of chromosome 16 was translocated to the short arm of chromosome 3. Separate contigs for the t(4;16), t(3;16) and t(3;4) were confirmed by OGM data and only 2-way chromosomal interactions were seen in the Micro-C data. This complex event led to numerous oncogenes juxtaposed to super-enhancers, which caused over-expression of target genes such as ACD and RANBP10 (log 2FC=1.0 and 0.4), both of which were on chromosome 16 and are significantly associated with inferior overall (p=0.02 and p=0.004) and progression-free survival (p=0.003 and p=0.005). Conclusion: By utilizing multiomic techniques we were able to reconstruct complex SV events, showing different mechanisms of generating templated insertion events and identifying their biological consequences. We demonstrated how these complex events could potentially facilitate tumor proliferation via engaging the super-enhancers and regulating the expression of key oncogenes.

Targeting ADSS2 Enhances BH3 Mimetics Treatment Induced Mitochondrial Apoptosis in AML Cells

Recently, the BH3 mimetic-Bcl2 inhibitor venetoclax (VEN), has been approved by FDA in 2018 for the treatment of patients with AML in combination with a hypomethylating agent (HMA) such as azacytidine (AZA). This synergy is likely due to HMA-induced mitochondrial vulnerability associated with cytoplasmic RNA metabolism. Nonetheless, most AML patients treated with the combination eventually relapse, possibly due to persistence of tumor cells not dependent on antiapoptotic proteins (Bcl2, Mcl1). There is a pressing need to develop targeted therapeutics sensitizing AML cells to mitochondrial apoptosis, particularly for those patients failed from VEN associated treatment. To achieve the goal, we performed a CRISPR/Cas9 Knockout (KO) screen (using Bim peptide-based dynamic ΔBH3 profiling as read-out) targeting genes (n=534) involved in cytosolic nucleotides metabolism pathways to increase mitochondrial vulnerability in a BH3 mimetic-refractory AML cell line (HEL). That search identified ADSS2, which catalyzes the first committed/rate-limiting step of AMP biosynthesis from IMP, emerged as the top hit, exhibiting highest ΔBH3 priming. Subsequent BH3 priming assays validated that ADSS2 KO significantly enhanced sensitivity to BH3 mimetics across various AML cell lines, including MOLM13, SKM1, OCI-AML3, and HEL cells. ADSS2 KO also led to a noteworthy reduction in the IC50 values of both VEN and the Mcl1 inhibitor S63845 in these cell lines. Additionally, the synergistic effect between sgRNA targeting ADSS2 and VEN or S63845 was also confirmed in Cas9 expressing cells from two AML patient-derived-xenograft (PDX) samples. We next transplanted AML cells (form one PDX samples) with or without ADSS2-KO into NGS mice, then treated mice with VEN. Notably, VEN treatment (100 mg/kg, QDx14, PO) significantly reduced tumor burden in ADSS2-KO AML mice compared to WT AML mice. Additionally, ADSS2 KO combined with VEN treatment group displayed a survival advantage compared to single VEN treatment group. Further mechanistic investigations revealed that ADSS2 KO led to a remarkable depletion of AMP rather than ATP, and a robust repression of AMPK activity in HEL cells. To assess whether AMPK inhibition mimics the effect of ADSS2 KO, AMPKα-deficient (Mx1-Cre/Prkaa1 fl/fl/ Prkaa2 fl/fl) MLL-AF9 (MA9) transformed murine AML cells were assessed. Indeed, AMPKα depletion substantially enhanced VEN treatment efficiency ex-vivo. Notably, enforced overexpression of a constitutively active AMPKa2 (T172D) construct in a VEN sensitive cell line SKM1, promoted cell resistance to mitochondrial apoptosis by VEN or S63845. Moreover, T172D expression can also abolish ADSS2-KO mediated mitochondrial vulnerability to VEN or S63845 treatment, indicating the pivotal role of AMPK signaling in mediating ADSS2 function. There are no reported ADSS2 inhibitors. To explore ADSS inhibitor, we conducted a fragment-based virtual screen of 8,000,000 compounds (MolPort) library and requested the top-ranking hits for ADSS2 catalysis evaluation by ELISA. Notably, the top two hits AS104 and AS71 exhibited the most potent inhibitory effects (AS104 IC50=2.43µM, AS71 IC50=3.85µM). NMR-STD analysis was conducted in a cell-free system, which unequivocally confirmed the physical binding of inhibitors and the ADSS2 protein. To further investigate the interactions between AS71 and AS104 with ADSS2 in a cellular context, both compounds were subjected to a cellular thermal shift assay (CETSA). Notably, treatment with AS71 and AS104 led to substantial shifts in ADSS2 protein thermal stability, providing evidence of their interactions with ADSS2 within the cellular environment. In both HEL cells and a primary VEN-resistant AML samples, treatment with AS71 and AS104 significantly increased ΔBH3 priming. Collectively, our observation revealed that targeting ADSS2 is critical for sensitizing AML cells to BH3 mimetics. We are now conducting preclinical assessments of ADSS2 inhibitors.