Abstract A017: A patient derived xenograft repository capturing clinical and molecular heterogeneity of large B-cell lymphoma

Abstract

Large B-cell lymphomas (LBCLs) are clinically and molecularly heterogeneous malignancies with a rapidly evolving therapeutic landscape. The increasing use of CD19 chimeric antigen receptor T cell therapy (CART19) has created urgent need for preclinical models that capture LBCL heterogeneity and tumor intrinsic resistance mechanisms. Existing LBCL patient-derived xenograft (PDX) models pre-date the CART19 era, have a small number of cases to fully represent the molecular spectrum, and are not openly accessible. We therefore established X-LYMPH (Xenografts of Lymphoma), a publicly available, molecularly annotated LBCL PDX repository that captures the heterogeneity of LBCL. PDX models were generated from core needle biopsies implanted into the renal capsule of immunodeficient NSG mice. From 133 implanted LBCL biopsies (98 diffuse large B-cell lymphoma (DLBCL), 19 high-grade B-cell lymphoma, 6 primary mediastinal large B-cell lymphoma (PMBL), 10 other subtypes), we established 48 PDX models from 45 patients with take rates of 30-53% depending on histology. Together with existing LBCL PDX models from the Public Repository of Xenografts (PRoXe) and Weill Cornell Medicine (WCM), a total of 80 models were characterized by whole exome sequencing (WES), low-pass whole genome sequencing (lpWGS), DLBCL90 NanoString transcriptional subtyping, RNA-seq, ATAC-seq, tissue microarray immunohistochemistry (IHC), and fluorescence in situ hybridization (FISH). WES revealed recurrent mutations in established LBCL drivers and models spanned multiple LymphGen genetic subtypes including EZB, BN2, MCD, ST2 and A53. Transcriptional subtyping identified germinal center B-cell-like (GCB, n=30), activated B-cell-like (ABC, n=36), PMBL (n=4), and unclassified (n=10) subtypes, with 15 GCB cases positive for dark zone signature. Among the 48 models, 23 were derived from post-CART19 progression biopsies including 19 where CART19 was immediately prior therapy. Molecular characterization of these models showed 4 PDXs harbored CD19 mutations, including two truncating mutations reducing CD19 protein expression and two point-mutations in the FMC63-binding domain. Additionally, 6 models showed CD19 protein loss by IHC, with RNA-seq demonstrating selective CD19 transcript downregulation and ATAC-seq revealing epigenetic silencing at the CD19 locus. To evaluate tumor cell-intrinsic CART19 resistance, we further treated 7 PDX models in vivo with TCR-knockout CART19 cells generated from healthy donors via CRISPR editing. A CART19-naïve model achieved complete tumor clearance by 4 weeks, whereas all 6 post-CART19 progression models showed persistent disease with different resistance patterns, including variable T cell infiltration ranging from intratumoral penetration with modest tumor reduction to peripheral-only or failure of infiltration with rapid progression. Overall, X-LYMPH provides a publicly available resource for mechanistic investigation of LBCL biology and CART19 resistance, and preclinical evaluation of novel therapies for high-risk lymphoma patients.