BackgroundDiffuse large B-cell lymphoma (DLBCL) is the most common lymphoma subtype accounting for ∼40% of all non-Hodgkin lymphomas. Although R-CHOP has significantly improved outcome in DLBCL, 35% of patients still experience relapsed/refractory disease. Further investigation into the genomic architecture of DLBCL is needed to determine the biological correlates that underlie treatment failure. Immune cell signatures have been shown to impact survival in DLBCL (Lenz et al, NEJM 2008). Recurrent mutations and/or copy number alterations (CNAs) in genes that impact the tumor's immune-cell crosstalk have recently been described, but their clinical impact is unknown. Moreover, mutations and copy-number loss involving genes that foster immune escape has been implicated as a dominant oncogenic mechanism in DLBCL (Challa-Malladi et al, Can Cell 2011). We sought to determine the relationship between genetic alterations involving immune-cell crosstalk pathways and survival in a cohort of DLBCL patients treated with R-CHOP. MethodsFollowing our initial analysis of 91 whole transcriptome sequencing (RNA-seq) including 40 with whole genome sequencing, we validated SNVs in the genes of interest by Sanger sequencing in the tumor and matched germline DNAs if available. RNA-seq-derived gene expression data were also used to determine the cell of origin (COO) distinction. Affymetrix SNP 6.0 microarrays were used to ascertain the CNAs in all 91 DLBCL samples. Clinical outcome correlations were analyzed by Cox regression and the log-rank test in the 82 patients treated with R-CHOP (median follow-up; 82 months). ResultsUsing next-generation sequencing, we identified 97 SNVs in 16 genes (B2M, CD274, CD276, CD58, CD83, CIITA, NLRC4, NLRC5, RFXAP, RFXANK, TNFRSF14, TNFRSF25, TNFSF9, TAP1, TAP2 and TAPBP) that were selected based on an extensive literature review implicating these genes in immune-cell crosstalk pathways in DLBCL. Among them, eight genes were mutated in more than one case including B2M (7%), CD58 (7%), TNFRSF14 (7%), CIITA (4%), TNFSF9 (2%), NLRC5 (3%), CD83 (11%), and RFXAP (2%). These eight loci were further analyzed for CNAs, gene expression and outcome correlations. SNP 6.0 microarrays revealed either focal heterozygous or homozygous deletions in B2M (22%), CD58 (14%), TNFSF9 (10%), and heterozygous deletions in TNFRSF14 (12%), CIITA (3%), NLRC5 (7%), CD83 (2%), and RFXAP (7%). Amongst the 91 cases, 35 cases (38%) had no genetic aberration. 56 cases (62%) had at least one aberration, while 29 (32%) had multiple genetic aberrations indicating that genetic aberrations associated with immune-cell crosstalk are not mutually exclusive in DLBCL. Mutations and CNAs were not COO specific, with the exception of TNFRSF14, which was significantly restricted to the GCB-subtype (p=0.001).We evaluated outcome correlations for each genetic aberrations in the 82 cases with uniform therapy, including two recognized major histocompatibility class (MHC) groups; MHC class I (B2M and NLRC5) and MHC class II (CIITA, RFXAP and CD83). In univariate analysis, the presence of genetic aberrations in the MHCI group and NLRC5 individually were prognostic (5yr PFS, 59% vs 78%, p=0.025, 44% vs 77%, p=0.005, respectively), but not the MHCII group or other individual loci. In subgroup analyses according to COO, both MHCI and MHCII genetic aberrations were associated with inferior 5yr PFS in the ABC subtype (MHCI, 27% vs 79%, p<0.0001) (MHCII, 40% vs 68%, p=0.018), but not in GCB. We next assessed the prognostic impact of the total number of genetic hits and revealed that patients with at least one genetic hit had a 5yr OS and PFS of 65% and 57%, respectively, while those with no genetic hit had a 5yr OS and PFS of 100% and 97%, respectively (both comparisons, p<0.0001). In a Cox model of PFS including IPI and COO, any genetic hit in immune-cell crosstalk pathways remained strongly predictive (HR=15.1 [2.0-113.2], p= 0.008) and was independent of both IPI (HR= 2.1 [1.1-4.0], p= 0.030) and COO (HR= 0.6 [0.3-1.4], p=0.26). ConclusionsRecurrent genetic alterations involving genes related to immune-cell crosstalk in DLBCL are frequent (56%), co-occur and independently contribute to outcome in patients treated with R-CHOP. MHCI and MHCII perturbations are only prognostic in the ABC subtype. [Display omitted] Disclosures:No relevant conflicts of interest to declare.