-Author name in bold denotes the presenting author
-Asterisk * with author name denotes a Non-ASH member
Clinically Relevant Abstract denotes an abstract that is clinically relevant.

PhD Trainee denotes that this is a recommended PHD Trainee Session.

Ticketed Session denotes that this is a ticketed session.

4357 Anatomic Genetic Heterogeneity Is Uncovered By Concurrent Cell-Free DNA and Tissue Profiling and Predicts Treatment Resistance in Diffuse Large B-Cell Lymphoma

Program: Oral and Poster Abstracts
Session: 621. Lymphomas: Translational – Molecular and Genetic: Poster III
Hematology Disease Topics & Pathways:
Research, Lymphomas, Non-Hodgkin lymphoma, Translational Research, Genomics, Bioinformatics, Diseases, Lymphoid Malignancies, Biological Processes, Emerging technologies, Technology and Procedures, Profiling
Monday, December 9, 2024, 6:00 PM-8:00 PM

Jordan S Goldstein, MD, MSc1, Florian Scherer, MD2, Brian J. Sworder, MD, PhD3, Maximilian Diehn, MD, PhD4*, David Kurtz, MD, PhD5 and Ash A. Alizadeh, MD, PhD6

1Department of Medicine, Divisions of Oncology and Hematology, Stanford University, Palo Alto, CA
2Department of Hematology, Oncology and Stem Cell Transplantation, Faculty of Medicine, University of Freiburg, Freiburg, Germany
3Division of Hematology and Oncology, University of California, Irvine, Irvine, CA
4Department of Radiation Oncology, Stanford University Medical Center, Stanford, CA
5Department of Medicine (Oncology), Stanford University, Palo Alto, CA
6Department of Medicine, Divisions of Oncology and Hematology, Stanford University, Stanford, CA

Background: Diffuse large B-cell lymphoma (DLBCL) displays clinical and molecular heterogeneity, seen both inter- and intra-patient. Within DLBCL patients, the evaluation of anatomic genetic heterogeneity has been limited by single-site biopsies. Cell-free DNA (cfDNA) can provide more comprehensive spatial genomic profiling. We sought to characterize and quantify DLBCL anatomic heterogeneity using cfDNA in relation to tissue biopsies and explore its impact on outcomes.

Methods: We used a previously reported cohort of patients with DLBCL, high-grade B-cell lymphoma or transformed low grade lymphoma, who received front-line anthracycline-based chemoimmunotherapy (Kurtz et al., JCO 2018). We included patients with pre-treatment tumor, plasma, and germline samples and outcome data available. We applied targeted next-generation sequencing by CAPP-Seq. Somatic mutations were identified by paired analysis of tumor or plasma with germline DNA and filtered for clonal hematopoiesis.

We assessed the differences between private cfDNA mutations and those shared with tissue, as single nucleotide variants (SNVs), coding mutations, and driver mutations, defined as mutations contributing to DLBCL initiation or progression. To quantify anatomic heterogeneity, for all SNVs identified in cfDNA, we evaluated and developed linear models for log-transformed variant allele fractions (VAF) in plasma relative to tissue. We assessed metrics including slope, root mean square error (RMSE), and Spearman correlation and their association with clinical features, clonality, and PFS.

Results: Of 75 patients, 99% had unique private plasma SNVs, 68% had private plasma coding mutations, and 46% had private plasma driver mutations. Of all driver mutations, 40% were shared between plasma and tissue and 27% were private to plasma. The most common plasma driver gene mutations occurred in BCL2, PIM1, SOCS1, TP53 and CARD11 genes, while CARD11, TNFAIP3 and BCL2 were the most commonly mutated driver genes private to plasma. 24 driver mutations (47%) were more likely to be detected as private plasma SNVs than shared, including CARD11, TNFAIP3, PR2Y8 BCL6, and MLL2.

Among mutations found in cfDNA, SNVs shared with tissue had higher allelic levels than SNVs private to plasma with median VAF 6% vs 1.2% respectively (p < 0.0001). Coding SNVs in cfDNA were more likely to be private to plasma than non-coding SNVs (17% vs 6%, p < 0.0001) and were associated with lower allelic levels than non-coding SNVs (median VAF 2% vs 5% respectively, p < 0.0001), potentially due to AID-related hypermutation in non-coding regions. Among coding mutations in cfDNA, driver mutations were associated with higher VAFs and more likely to be shared with tissue (59%) than private to plasma (38%).

Linear models of SNV plasma VAF relative to tissue VAF were generated for each patient. Higher slope (change in plasma SNV VAF relative to tissue) was associated with worse PFS as a continuous variable (HR = 10, p = 0.0005) and when stratified at median 0.26 (HR = 2.6, p = 0.04). RMSE of the linear models > median 0.22 was associated with worse PFS (HR = 3.3, p = 0.01). This was particularly the case among transformed low-grade lymphoma patients, who had 100% PFS at 3 years for RMSE < 0.22 compared to 19% for RMSE > 0.22 (p = 0.01). Correlation between plasma and tissue VAF was not significantly associated with outcomes (HR = 3.5, p=0.09). Number of subclonal variant clusters identified in cfDNA was associated with slope (r = 0.58, p < 0.0001), correlation (r = 0.50, p < 0.0001), and RMSE (r = 0.32, p = 0.005) as well as worse PFS (HR = 1.3, p = 0.04). Neither slope, RMSE nor correlation were associated with IPI, age, stage, COO, or histology; higher slope and RMSE remained associated with worse PFS when controlling for these clinical features. Plasma mean VAF was associated with slope (r = 0.55) and correlation (r = 0.53). Controlling for mean VAF, slope (p = 0.01) and RMSE (p = 0.02) remained significantly associated with worse PFS.

Conclusion: DLBCL exhibits substantial intra-patient anatomic heterogeneity uncovered by cfDNA profiling, through identification of mutations absent in single site biopsies and comparison of plasma and tissue genotype profiles. Increased anatomic heterogeneity is independently associated with DLBCL treatment resistance. Coupling pre-treatment liquid with tissue biopsy provides a comprehensive assessment of DLBCL genotype with prognostic implications.

Disclosures: Scherer: Roche Sequencing Solutions: Research Funding; Gilead Sciences: Research Funding; Takeda: Research Funding; AstraZeneca: Honoraria; Servier: Honoraria. Kurtz: Foresight Diagnostics: Current Employment, Current equity holder in private company, Current holder of stock options in a privately-held company, Membership on an entity's Board of Directors or advisory committees, Patents & Royalties. Alizadeh: Gilead: Consultancy; Foresight: Consultancy, Other: Scientific Co-founder; Roche: Consultancy; Forty Seven: Other: stock; Pharmacyclics: Consultancy; CiberMed: Consultancy, Other: Scientific Co-founder; CARGO Therapeutics: Divested equity in a private or publicly-traded company in the past 24 months; ADC Therapeutics: Consultancy; Adaptive Biosciences: Consultancy; BMS: Research Funding.

*signifies non-member of ASH