Abstract
CD3 x CD20 bispecific antibodies (BsAb)have been approved for use in relapsed/refractory (R/R) diffuse large B-cell lymphomas (DLBCLs) based on its deep and durable response and manageable toxicity profiles. While emerging data suggest that tumor microenvironment may associate with treatment response, these findings have not yet been validated in the FDA-approved agents Epcoritamab and Glofitamab. This study aimed to integrate genomic profiling, gene set enrichment, and tumor microenvironment (TME) analysis to identify biomarkers predictive of response to Epcoritamab and Glofitamab in patients with R/R DLBCLs.
We performed whole exome sequencing (WES) on 15 patients and whole transcriptome sequencing (WTS) on 15 patients. Among the patients, 13 of them have both WES and WTS data. For WES, raw sequencing reads were aligned to the human reference genome (hg38) using the Burrows-Wheeler Aligner, Variant calling was conducted using VarScan and ANNOVAR. For WTS, STAR was applied to perform the alignment and HTSeq was used for gene expression quantification. The downstream analyses include CIBERSORTx and GSEA. Clusters were determined using Immune Quadrant (Tumuluru et al). Xenium spatial transcriptomics was performed on a subset of cases.
15 patients with available pre-BsAb tissue biopsy were included. The cohort predominantly consisted of diffuse large B-cell lymphoma (DLBCL)(n=11, 3 germinal center B-cell–like [GCB]; 8 non-GCB), with 2 cases of double-hit lymphoma, and 2 cases of follicular lymphoma transformed to DLBC. The overall response rate (ORR) and complete response (CR) rate to BsAb were 40.0% and 26.7%, respectively. The median progression-free survival (PFS) and overall survival (OS) of the entire cohort was 2.4 months (95% confidence interval [CI] 1.8-2.9 months) and 10.1 months (95%CI 3.4-16.8 months), respectively. CRS occurred in 7 patients, including 6 patients with grade 1 CRS and 1 patient with grade 2 CRS.
The most frequently mutated genes were TP53 (53.3%, n=8), KMT2D (40%, n=6), CREBBP (33.3%, n=5), CARD11 (26.7%, n=4), IGLL5 (26.7%, n=4), and MYC (26.7%, n=4). Additional recurrent mutations included BCL2, DTX1, DUSP2, NOTCH2, SPEN, and TNRC18 (each 20%, n=3).
The cohort was segregated into 3 clusters defined by cell-of-origin (COO) and immune-related gene set expression scores. These clusters consisted of 5 activated B cell-like (ABC) cold, 4 GCB cold, and 6 GCB hot LBCLs. No ABC hot was identified. There were more CR/PR in the GCB hot cluster compared to cold GCB/ABC (67% vs 25%; p=0.13). No significant associations were found between mutations, clusters and outcomes (ORR, CR, CRS, PFS, OS), likely due to limited number of cases.
Immune deconvolution via CIBERSORTx revealed that GCB hot cases had significantly higher levels of total T cells compared with GCB cold (p=0.0095) and ABC cold (p=0.0087), indicating a “hot” TME that may favor immunotherapeutics. No significant difference was observed in T cell subtypes, B cells, NK cells, macrophages, or dendritic cells levels between groups. Three cases were available for Xenium analysis (GCB hot; GCB cold; ABC cold) which showed correlation with cluster groups with the GCB hot tumor having a high proportion of T cells as compared to the cold tumors (36% vs 2.5%).
This integrative genomic and immune profiling study identified distinct molecular clusters in R/R DLBCL patients treated with CD3 x CD20 bispecific antibodies. Mutations did not correlate with response or cluster but GCB hot tumors demonstrated significantly higher total T cell infiltration compared to cold clusters and a trend for better response. These findings suggest that immune microenvironment profiling may inform BsAb therapy selection, with GCB hot tumors potentially representing a more immunologically favorable subset. Further validation in larger cohorts is warranted to establish these biomarkers for clinical decision-making in BsAb therapy selection.
