Epigenetics as a mechanism of lineage extinction and relapse of B-cell lymphoma treated with CD19 CAR T-cell therapy Authors: Alireza Torabi, Jason Love, David Wu, Alexandre Hirayama, Kikkeri Naresh Published: 2026 ABSTRACT Visual Abstract TO THE EDITOR: Resistance and relapse following targeted therapy in B-cell lymphomas are often mediated by antigen loss or modulation,1,2 clonal evolution with the acquisition of additional gene alteration,3 treatment failure (T-cell exhaustion in chimeric antigen receptor [CAR] T-cell therapy), or changes in tumor microenvironment.4 Loss of the target antigen can result from a variety of mechanisms, including gene mutations or deletions, alternative splicing, trogocytosis, antigen masking, and epigenetic downregulation.5 Antigen-negative relapses can occur due to hypermethylation of the targeted gene promoter.6 However, in these circumstances, the overall B-cell programming remains almost intact other than loss of the targeted antigen. Lineage switch (LS) and infidelity, driven by immune pressure, represent significant pathways for relapse following targeted therapy.7 Although relapsed tumors retain their clonal relatedness, they undergo a fundamental shift in identity and phenotype. Although the precise drivers of LS remain poorly understood, epigenetic reprogramming has emerged as a central process that allows cancer cells to acquire stem cell–like properties, bypass lineage constraints, and evade therapeutic pressure.8,9 Recent studies have specifically linked epigenomic reprogramming to LS and relapse,10 highlighting global methylation of B-cell lineage genes as a key factor following targeted treatment.11 We previously reported a case of relapsed aggressive B-cell lymphoma after CD19 CAR T-cell therapy with a null phenotype, meaning the loss of all the B-cell markers as well as common leukocyte antigen, CD45.12 This dedifferentiated tumor expressed CD56 and showed BCL2 gene rearrangement, indicating a clonal relation to the B-cell lymphoma before the immunotherapy. However, the underlying mechanism of losing the B-cell antigen and transcription factors remained unclear. DNA was extracted from formalin-fixed paraffin-embedded tissue samples of pre– and post–CAR T-cell therapy biopsy specimens. We did not undertake prior enrichment of specific cell populations. We investigated the methylation status of a relapsed/refractory B-cell lymphoma after CAR T-cell therapy, trying to identify the epigenomic changes and key genes involved in dedifferentiation and tumor evolution after CAR T-cell therapy. Genome-wide methylation profiling comparing the pretreatment to the post–CAR T-cell therapy revealed a significant hypermethylation shift in differentially methylated cytosines (DMCs) and differentially methylated regions (DMRs). DMC and DMRs were identified using the DSS package after filtering for cytosines with read depth ≥5. DSS was used to perform Wald tests with Benjamini-Hochberg adjustment, defining significance as an false discovery rate (FDR) ≤0.05 and methylation difference ≥0.1. Figure 1A shows the heat map of the top 3000 DMC and DMR sites. Using this approach, genome-wide methylation profiling comparing the pretreatment to the post–CAR T-cell therapy samples revealed significant enrichment of biologically relevant gene programs, most notably the developmental process (P = 1.1 × 10-308) followed by the multicellular organismal process (P = 1.1 × 10-308), the regulation of cellular process (P = 9.1 × 10-292), and the response to stimulus (P = 1.4 × 10-254). Differentially methylated gene heat map, differentially methylated B-cell markers, immunoglobulin heavy chain (IgH) locus methylation status, and differentially methylated top genes. (A) Heat map of the top 3000 DMC and DMR sites, pre– and post–CAR T-cell therapy. Blue denotes hypermethylation and yellow denotes hypomethylation. (B) Hypermethylation of B-cell antigens and transcription factors after CAR T-cell therapy. (C) IgH locus on chromosome 14. Each dot represents CpG-containing region within the IgH locus. Hypermethylation is more pronounced in Eμ/Sμ enhancer, constant regions, and 3′RR enhancers. (D) A total of 219 significantly differentially expressed genes are represented in terms of their measured expression change (x-axis) and frequency of genes measured at a given expression change (y-axis). Cγ/Cε/Cα, constant region genes; DMC, differentially methylated CpG; DMR, differentially methylated region; Eμ/Sμ, intronic enhancer and switch μ region; FC, fold change; IGHD, diversity segment; IGHJ, joining segment; IGHV, variable gene segment; 3′RR (hs3a-hs4), 3′ regulatory region enhancers; TF, transcription factor. Nearly all major diagnostic B-cell surface antigens, including CD19, CD20 (MS4A1), CD22, CD37, CD74, CD79A, and CD79B, demonstrated large markedly hypermethylated CpG proportions, indicating epigenetic promoter silencing (Figure 1B). Core B-lineage transcriptional regulators also exhibited significant hypermethylation burden in PAX5, EBF1, BCL6, PRDM1, IRF4, OCT2 (POU2F2), and BOB1 (POU2AF1) genes. Similarly, CD45/PTPRC, which is not targeted by CAR T cells, showed multiple significantly hypermethylated CpG regions, consistent with immune-selective promoter constraint in the posttreatment clone. The hypermethylation also included IgH enhancer regions (Figure 1C). A total of 219 differentially expressed genes were identified out of a total of 62 563 genes in Advaita Knowledge Base (Figure 1D). For large-scale DNA methylation studies, Advaita Knowledge Base provides a precise framework for mapping epigenetic modifications to specific functional pathway perturbations. The top hypermethylated genes were tumor suppressor genes (FHIT and CUX1), genes involved in mTOR and Wnt/β-catenin pathways such as RPTOR and BCL9L genes, respectively, and genes involved in immune response (NFATC1 and MAD1L1). The hypomethylated genes are those involved in development and stemness activation (HOX family, MEIS2, DLX5, IRX5, PAX9, and ZIC2) and those involved in growth, adhesion, and migration (IGF1R, IGF2, BMP2, LAMA4, and PDGFC). FZD8, which is crucial for the Wnt/β-catenin pathway, was also hypomethylated. These data were analyzed in the context of pathways curated from the Kyoto Encyclopedia of Genes and Genomes database.13,14 The results of the pathway analysis including top pathways, biological processes, and molecular functions with associated P values, are illustrated in supplemental Figure 1. This analysis identified ribosome upregulation as one of the top enriched pathways. We performed additional immunostaining of the post–CAR T-cell tumor beyond those described in our earlier report.12 c-MYC immunostain showed strong nuclear staining, indicating a high proliferative phenotype, and β-catenin showed only membranous/cytoplasmic staining without nuclear staining (Figure 2), arguing against the canonical activation of the WNT pathway. Immunostains on the post–CAR T-cell tumor. (A) Strong cMyc expression; (B) weak LEF1 nuclear staining; (C) cytoplasmic/membranous β-catenin staining. CAR T-cell therapy eliminates antigen-visible B-cell lymphoma clones but spares proliferation-competent ones such as MYC-driven cells. Original magnification ×20 for panels A-C. Silbert et al reported a large series of cases of acute leukemias that showed a LS after immunotherapy.15 The basis of the LS in acute leukemia is poorly understood, and several mechanisms, such as the presence of primitive and stem cell–like populations and actual transdifferentiation, have been proposed.16 Transcriptomic and epigenomic signatures of the blast population have shown differentially methylated regions of the B-cell precursors during maturation.17 Transcriptional reprogramming and chromatin accessibility have been associated with LS in KMT2A-rearranged acute leukemia.10 In mature lymphoid neoplasms, LS after CAR T-cell therapy has been reported.11,18 In these reports, LS in relapsed malignancies was associated with silencing of B-cell genes through promoter hypermethylation. This case also elucidates on the role of epigenetics in LS. Following CAR T-cell therapy, B-cell antigen expression was shut down as a result of hypermethylation of B-cell genes, while sparing the proliferation-related genes such as MYC, as further evidenced by high MYC expression. Despite the persistence of t(14;18) after CAR T-cell therapy, BCL2 immunostain was negative.12 Methylation analysis showed epigenetic silencing at the translocated IgH enhancer hub, a CpG-dense region prone to epigenetic modification.19 IgH methylation changes can therefore suppress enhancer-driven BCL2 allelic transcription even when translocated, without implying lineage transformation. Pathway analysis after CAR T-cell therapy revealed upstream activation of GLI3 and UGT8 (FDR = .004 and .01, respectively). Although FZD8, a WNT receptor activating the β-catenin pathway, was among the top hypomethylated genes, the WNT3A pathway itself was not significantly activated (P = .09); this was further corroborated with membranous/cytoplasmic β-catenin localization on immunostaining. The recurrent post–CAR T-cell tumor has likely upregulated FZD8 as a compensatory survival mechanism, yet the canonical Wnt/β-catenin transcriptional program has remained inactive due to membrane sequestration of β-catenin by the E-cadherin complex or a shift toward noncanonical Wnt signaling pathway (Wnt/PCP). Epigenetic dysregulation remains a critical driver of tumor adaptation.20 The hypomethylation of oncogenes and hypermethylation of tumor suppressors contribute to malignant evolution, whereas epigenetic reprogramming can reawaken stem cell–like features that promote relapse under therapeutic pressure.21,22 Genes such as HOX, MEIS, and DLX5 exemplify this process.23 Normally associated with early development, their activation in cancer supports self-renewal, dedifferentiation, and impaired maturation. DLX5 gene, essential for neural crest differentiation,24 becomes hypomethylated and transcriptionally active in several cancers and can enhance proliferation through MYC promoter activation.25 Similarly, MEIS family genes function as HOX cofactors,26 modulating chromatin accessibility and sustaining embryonic transcriptional programs; MEIS2-related neuronal differentiation is paralleled here by CD56 expression.27 Among tumor suppressors, FHIT hypermethylation was particularly pronounced. The silencing of large fragile-site genes such as FHIT and concurrent activation of developmental transcription factors enable malignant cells to remodel lineage identity, evade immune pressure, and restore fitness following CAR T-cell therapy.28 In summary, we propose that selective pressure from CD19-targeted CAR T-cell therapy eliminated CD19+ lymphoma cells, facilitating the expansion of epigenetically reprogrammed clones and subsequent relapse with a CD19-null phenotype. Genome-wide methylation analysis demonstrated the silencing of B-cell identity genes, the activation of developmental and stem-like programs, and remodeling at fragile-site loci. These findings suggest that epigenetic lineage extinction and dedifferentiation represent an important resistance pathway, distinct from established mechanisms such as target gene mutation, deletion, or T-cell exhaustion. Although this case highlights lineage silencing as an emerging phenomenon, we anticipate that complete lineage extinction remains infrequent. Larger studies are warranted to delineate the frequency and clinical impact of such occurrences across B-cell lymphomas because we have not identified a dedifferentiated phenotype (of this nature) in other instances of post–CAR T-cell therapy recurrence observed at our center. Conflict-of-interest disclosure: The authors declare no competing financial interests. Acknowledgments Contribution: A.T. conducted the methylation study; A.T. and K.N. reviewed the data and wrote the manuscript; and all authors critically reviewed the manuscript. Footnotes Data are available from the corresponding author, Alireza Torabi (alirezat@uw.edu), on request. The full-text version of this article contains a data supplement. Supplementary Material Supplemental Methods, Figure, Table, and Reference References Associated Data Supplementary Materials Supplemental Methods, Figure, Table, and Reference