Chowdhury, Rakin

Abstract Background Primary CNS lymphoma (PCNSL) is a rare, aggressive diffuse large B-cell lymphoma confined to the CNS, with poor outcomes. Virally associated PCNSL, including AIDS-related PCNSL (AR-PCNSL) and CNS post-transplant lymphoproliferative disorder (CNS-PTLD) accounts for ~10% of cases and is strongly linked to EBV infection and poor prognosis. In our prior Blood (Gandhi et al. 2021) study, targeted sequencing showed that EBV⁺ PCNSL lacks canonical MYD88, CD79B, and PIM1 mutations and retains intact HLA class I/II, suggesting distinct immunobiology. However, questions remain regarding the immune microenvironment, broad mutational landscape, and intratumoral heterogeneity. Here, we present the first large-scale comprehensive integrated analysis of virally associated PCNSL to define subtype-specific vulnerabilities and therapeutic implications. Methods FFPE brain biopsies from 72 PCNSL patients (40 EBV- de novo PCNSL, 32 EBV+ PCNSL including 18 AR-PCNSL and 14 CNS-PTLD) were analyzed. Immune gene expression was profiled using the NanoString PanCancer Immune Panel. Whole exome sequencing (WES) defined recurrent driver mutations, tumor mutational burden (TMB), and LymphGen classification (excluding A53 subtype). Single nuclei RNA sequencing (snRNA-seq) using the 10x Genomics Chromium Flex platform characterized cellular heterogeneity. Results WES demonstrated a marked reduction in canonical mutations (MYD88L265P, CD79B, PIM1, CARD11) in CNS-PTLD (0%) and AR-PCNSL (29%) vs. de novo PCNSL (87.5%; p<0.0001). AR-PCNSL had a distinct mutational signature when compared with CNS-PTLD and de novo PCNSL. NOTCH1/2 mutations were significantly enriched in AR-PCNSL (71%) vs. CNS-PTLD (9%) and de novo PCNSL (33%) (p<0.01). NLRC5 mutations were present in 64% of AR-PCNSL cases, higher than the 25% observed in de novo PCNSL and were absent in CNS-PTLD. TMB was significantly elevated in AR-PCNSL (mean 26.27 mutations/Mb) vs. de novo PCNSL (17.22) and CNS-PTLD (11.01; p=0.0018) underscoring pronounced genomic instability and suggesting HIV may exert a mutagenic influence beyond immunosuppression. EBV driven subtypes lacked MCD features and were more often classified as ST2 or unassigned. Despite NOTCH2 mutations, AR-PCNSL cases were not classified under the BN2 subtype, further highlighting limitations in current molecular classification systems to accurately stratify virally associated PCNSL. Immune transcriptomic profiling revealed further distinctions within EBV+ PCNSL. CNS-PTLD showed increased expression of genes related to antigen presentation (p<0.0001), cytotoxicity (p=0.0019), and T cell function (p<0.0001), suggesting retained immune responsiveness. AR-PCNSL exhibited significant suppression of HLA class I/II genes (p=0.0027 and p<0.0001 respectively), despite intact genomic loci, implicating post-transcriptional or epigenetic silencing. Immune cell composition revealed reduced macrophage infiltration and suppressed CD4 expression in AR-PCNSL. CNS-PTLD showed higher expression of checkpoint molecules LAG3, TIM3, and CTLA4, consistent with T cell exhaustion. snRNA-seq profiling of AR-PCNSL identified distinct malignant and non-malignant cell populations, including a reduced CD4⁺ compartment and altered microglial/macrophage populations and cell states compared to de novo PCNSL, corroborating bulk transcriptomic data, amongst other ongoing analysis. Conclusion This study represents the largest and most comprehensive immunogenomic investigation of virally associated PCNSL. AR-PCNSL and CNS-PTLD lack canonical PCNSL driver mutations and instead harbor distinct mutational signatures, including NOTCH pathway alterations in AR-PCNSL. These findings underscore divergent pathogenic mechanisms and support tailored therapies. AR-PCNSL is marked by high TMB, suppressed antigen presentation, and reduced CD4⁺ infiltration, indicating profound immune dysfunction. CNS-PTLD retains immune functionality but shows checkpoint upregulation and T cell exhaustion. Despite both being EBV-driven and CNS-restricted, they display distinct molecular and immunological programs, suggesting context specific EBV mechanisms. These findings challenge existing classification systems and support CNS-specific molecular frameworks to guide therapeutic interventions including BTK inhibition for NOTCH-mutated lymphoma. They lay the foundation for precision approaches in EBV⁺ PCNSL, including immune restoration, antigen-specific T cell therapy, and epigenetic modulation.

Background: A subset of follicular lymphoma (FL) patients will experience a ‘high-risk defining event’ (HRDE) that portends poor survival due to early progression of disease or transformation to diffuse large B cell lymphoma (TFL). Progression of disease within 24 months (POD24) from front-line immunochemotherapy (ICT) identifies patients with a subsequent five-year OS of 50% compared to a reference group of 90% (Casulo et al. JCO 2015). Although a robust surrogate endpoint for OS in clinical trials (Casulo et al. Blood 2022), it remains to be established if the OS of patents is influenced by the nature of the HRDE. In this study, we examined whether characteristics and clinical trajectory of those with relapsed/progressive FL is the same as that of patients experiencing TFL. Methods: We conducted an international retrospective analysis of newly diagnosed FL patients requiring treatment across 14 academic centres between 2002 and 2021 in the era of rituximab and PET-CT staging. Patients were included with grade 1-3A FL requiring first-line (1L) treatment (ICT, rituximab monotherapy or radiotherapy). We excluded patients with untreated FL, de novo TFL and those with <24 months follow up provided they did not experience a HRDE. Baseline clinical, imaging and laboratory values, treatment details and outcomes were collected. We defined HRDE groups as persistent (relapse or progression) FL within 24 months of 1L treatment (FL24), early TFL (<24 months of 1L treatment) or late TFL (>24 months of 1L treatment). Patients not experiencing a HRDE formed the FL reference group. Landmark overall survival (OS) analysis was performed by Cox regression using risk status as a time-varying covariate for time of HRDE. For the FL reference group, landmark was defined at 24 months after 1L treatment. Results: 501 patients met the inclusion criteria and were categorised as: reference FL (n=378), FL24 (n=50), early TFL (n=28), late TFL (n=15). 24 patients who died within 24 months of 1L treatment without documentation of preceding HRDE and were omitted. Median follow up for patients alive beyond 24 months was 5.1 years. Median age at diagnosis was 60 years, 91% had ECOG performance status 0-1 and 38% were high-risk (>3) FLIPI. 1L chemotherapy regimens included: bendamustine (59%), CHOP (24%), CVP (8%), other (9%); a majority received rituximab (80%) vs obinutuzumab (20%). Maintenance therapy was used in 54% of pts. Expectedly, those experiencing HRDE had an inferior OS compared to the reference group (HR 5.23, p<0.001)(Figure 1). However, among HRDE group, median OS significantly differed: early TFL patients (3.0 yrs) and FL24 (12.6yrs)(HR 2.27, p=0.02) (Table 1). There was no difference in OS after HRDE between early TFL and late TFL ( p=0.81). Biopsy was performed in 84% of TFL cases with no difference in OS between biopsy-proven vs clinically suspected cases ( p=0.23). Multivariate OS analysis assessing HRDE and treatment demonstrated no significant interaction between HRDE group and 1L chemotherapy ( p=0.52), monoclonal antibody ( p=0.29) or maintenance therapy ( p=0.27). Several established prognostic features at baseline were enriched in patients who went on to experience HRDE including: ECOG >1 ( p<0.001), raised LDH ( p<0.001), advanced-stage disease (( p<0.011) and high risk FLIPI ( p=0.005). Baseline FDG-PET was available in 259 patients. Adverse PET metrics were enriched in only early TFL, and not FL24 or late TFL patients, including high SUV max( p=0.009), total metabolic tumor volume ( p=0.015) and total lesional glycolysis ( p=0.01). Conclusions: HRDEs after 1L treatment are associated with inferior OS in FL compared with standard-risk patients. Unlike POD24, this study demonstrates disparate clinical outcomes between patients experiencing early progression with persistent FL and those experiencing TFL. Adverse metabolic characteristics on baseline FDG-PET also suggest biological distinction between HRDE subsets. These findings have important implications for epidemiological studies, biomarker development and clinical trials aiming to improve outcomes in this underserved high-risk FL cohort.