Genetic signatures of CNS relapse in DLBCL patients profiled with next-generation sequencing (NGS).

7016 Background: CNS relapse of DLBCL is a devastating event with median survival <6 months. While the overall incidence is low (~5%), various factors impact the risk of CNS relapse including CNS-IPI, cell-of-origin (COO), and MYC rearrangement. Recent studies using NGS have defined DLBCL genetic subgroups with distinct biology. The MCD and C5 subgroups, defined by LymphGen and DLB class , respectively, have a shared genetic signature characterized by MYD88 L265P and CD79B mutations, immune evasion, and extranodal tropism with enrichment in CNS lymphomas. We investigated the impact of MCD/C5 signature on the risk of CNS relapse in a large cohort of DLBCL patients (pts) profiled with NGS. Methods: We performed a retrospective analysis of 155 DLBCL pts treated with chemoimmunotherapy and profiled with the UCSF500 NGS panel, which includes 529 genes and copy number data. FISH was performed to evaluate for MYC , BCL2 and BCL6 rearrangements. Pts with CNS involvement at diagnosis were excluded. Data were extracted from medical records including demographics, WHO-HAEM5 diagnosis, COO per Hans algorithm, CNS-IPI, and treatment history. LymphGen and DLB class subtypes were determined using online classifiers incorporating NGS and FISH data. We identified LymphGen and DLB class subtypes enriched in pts with CNS relapse. We performed univariable and multivariable analysis (MVA) (Cox proportional hazards model) to assess the impact of COO, CNS-IPI, and genetic signature on the cumulative incidence (CI) of CNS relapse. Results: The median age was 61. 60% were male, 57% non-Hispanic white, 21% Asian, 16% Hispanic, and 3% Black. DLBCL subgroups included DLBCL-NOS (60%), transformed DLBCL (15%), Richter transformation (8%), high-grade B-cell lymphoma (HGBCL)-double hit (5%), primary mediastinal LBCL (4%), DLBCL, post-transplant (3%), HGBCL-NOS (3%), and T-cell histiocyte-rich LBCL (2%). 67% were stage III-IV, 51% non-GCB type, and 21% high-risk CNS-IPI. LymphGen subtypes included EZB-MYC- (26%), ST2 (12%), MCD (9%), EZB-MYC+ (7%), BN2 (7%), N1 (3%), and composite cases (3%); 34% were unclassifiable. DLB class subtypes included C1 (13%), C2 (3%), C3 (52%), C4 (8%), and C5 (11%); 14% were not classified given <50% confidence in assignment. Median follow-up was 30 months. 18 pts experienced CNS relapse; these cases were enriched for MCD and C5 subtypes (28% each). The 2-year CI of CNS relapse in pts with MCD and/or C5 subtypes (N=21) was 20.4% vs. 6.4% in non-MCD/C5 cases (HR 3.46, p=0.007). Notably, 3 MCD/C5 pts had late CNS relapse (>3 years from diagnosis). In MVA, MCD/C5 signature was an independent risk factor for CNS relapse (HR 4.51, 95% CI 1.42-14.32, p=0.011) along with MYC rearrangement (p=0.036) and non-GCB COO (p=0.028). Conclusions: In this heterogeneous DLBCL cohort, MCD/C5 genetic signature was an independent risk factor for CNS relapse. Incorporation of CNS-active agents should be prioritized for this subgroup.