S845 PREDICTED NEOANTIGENS PROVIDE PROGNOSTIC INSIGHT IN DLBCL

Abstract

Background:Somatic mutations in tumor cells can generate neoantigens (NAs) recognizable by T‐cells as part of a patients’ immune surveillance. Evasion of immune surveillance is a hallmark of cancer that impacts clinical outcomes. Diffuse Large B‐cell Lymphomas (DLBCL) harbor high tumor mutational burden (TMB), but the prevalence and contribution of predicted NAs on clinical outcomes in patients (pts) have not been fully characterized.Aims:Apply a clinically available, targeted comprehensive genomic profiling (CGP) platform examining the predicted neoantigen burden (NAB) and specific NAs to provide prognostic insight and inform vaccine strategies in DLBCL.Methods:Biopsy samples, clinical outcomes, and CGP (n = 465 genes) data were available from 499 pts enrolled in the phase 3 de novo DLBCL GOYA trial (NCT01287741; ITT = 1418; obinutuzumab (G)‐CHOP vs rituximab (R)‐CHOP), with CGP alone from an additional 690 DLBCL routine clinical care samples from the FMI Database (DLBCL‐FM), of which 82% were from pts with relapse/refractory (R/R) disease (47% of the examined subset were evaluable). CGP was used to estimate TMB, HLA type (OptiType) and NA prediction (NetMHCpan); multivariate cox regression was used to correlate TMB, NA with PFS, EFS, and OS. All models were adjusted for treatment, age, sex, IPI, and cell‐of‐origin (Nanostring). Results are reported descriptively with 95% CIs.Results:We calculated a median TMB of 10.5 and 9.7 mut/Mb for the GOYA and DLBCL‐FM cohorts, respectively and both had a median of 3 predicted NAs per pt. The majority of pts (86%) were predicted to have ≥1 NA. TMB correlated with NAB in both GOYA (0.56 Spearman coefficient) and DLBCL‐FM (0.55 Spearman coefficient). While TMB did not correlate with outcomes, the presence of a NA correlated with worse EFS (HR 1.72 (1.01, 2.94)) but not PFS (HR 1.55 (0.87, 2.76)) or OS (HR 1.72 (0.79, 3.77)). Loss of B2 M, a critical component of MHC I presentation, was enriched among pts with high NAB (≥5)(OR = 1.8, p = 0.01), with a similar trend for a lower NAB cut‐off (≥1)(OR = 1.96, p = 0.08) in the DLBCL‐FM cohort (largely R/R), but not among GOYA (1L) pts (low NAB (OR = 1.25, p = 0.72); high NAB (OR = 1.15, p = 0.6)). Among BCL2 alterations, 25.9% were predicted NAs, which correlated with poorer outcomes (presence/absence of a BCL2 NA: OS HR = 3.24 (1.59, 6.62), PFS HR = 2.22 (1.21, 4.06)). As a continuous measure, the number of predicted BCL2 NAs was also associated with worsening outcomes (OS HR = 1.44 (1.18, 1.76), PFS HR = 1.33 (1.10, 1.61)); moreover, thresholding highlighted strong OS/PFS associations, (≥2 BCL2 NAs OS HR 5.61 (2.18, 14.42); ≥3 BCL2 NAs OS HR 7 (2.05, 23.96); ≥4 BCL2 NAs OS HR 12.98 (3.64, 46.27)). BCL2 short variant NAs were more likely in patients with BCL2‐high expression by IHC (25/35 BCL2 NA are IHC+), and patients with both BCL2‐high expression (IHC) and BCL2 NA experienced poor overall survival compared to patients with BCL2‐high expression without NA (IHC+/NA+: OS HR = 4.40 (1.88, 10.32), PFS HR = 3.10 (1.49, 6.44)).Summary/Conclusion:Pts with DLBCL are frequently predicted to have at least one NA, with an enrichment in B2 M loss among patients with a high NAB in the largely R/R DLBCL cohort. BCL2 NAs correlate with poorer outcomes among de novo DLBCL patients treated with R/G‐CHOP, even when accounting for known prognostic factors such as COO, age, BCL2‐high IHC and IPI. Moreover, the presence and number of BCL2 NAs was prognostic for PFS and OS. These insights may inform future personalized cancer vaccine strategies in DLBCL.