Distinct genomic landscape & response dynamics of HIV-associated lymphomas identified by integrated noninvasive host and virome profiling

Abstract

Abstract Background: HIV infection increases lymphoma risk, with many cases linked to co-infection with latent Epstein-Barr Virus (EBV). Immunosuppression and oncogenic viruses contribute to aggressive cancers that often present at advanced stages and progress rapidly. Prior studies of HIV-associated lymphomas have generally been restricted to relatively small cohorts and pretreatment tumor specimens, and there is a lack of integrated methods for tracking viral dynamics alongside host tumor and immune responses. We characterized the genomic landscapes from a large HIV-positive (HIV+) lymphoma cohort as contrasted with HIV– counterparts, revealing how viral reactivation and immune function influence cancer development. Methods: We comprehensively profiled 305 blood samples from 122 patients, including 51 HIV+diffuse large B cell lymphoma (HIV+/DLBCL), 10 HIV+ Burkitt lymphoma (HIV+/BL), and 61 HIV–/DLBCL patients. HIV+ patients were treated with SC-EPOCH-RR (NCT000019253). Tumor EBV status was determined by Epstein-Barr-encoded small RNA (EBER) ISH. Samples were profiled by VirCAPP-Seq targeting 180 viral species (Garofalo, Blood 2019), CAPP-Seq targeting 186 B-cell lymphoma genes (Newman, Nat Med 2014), and a custom panel targeting HLA genes. Results: To capture virome dynamics, we applied unsupervised clustering to serial samples of HIV+ lymphomas using viral family abundances. We identified 3 distinct Viral abundance clusters (V1-3), each reflecting a unique spectrum of active species, with gradually decreasing viral abundances from V1 to V3. V1 reflected high abundance across anelloviruses (AV), polyomaviruses, EBV, and other herpesviruses. V2 had high abundance of AV, but low EBV abundance. V3 had low abundance across all viral families. Consistent with the degree of immunosuppression, pretreatment samples were enriched in V1. Patients achieving remission had a gradual clearance of diverse viruses during and after therapy, transitioning towards V2/V3 clusters over time. In contrast, patients experiencing progression (PD) or associated death remained enriched in V1 at time of PD. We genotyped EBV subtypes with adequate genomic coverage of HIV+ lymphoma alongside EBV+ cell lines, classical Hodgkin lymphomas, and post-transplant lymphoproliferative disorders. HIV+ lymphomas had higher prevalences of EBV type 2 and type 1/2 co-infections (p<0.05) compared to immunocompetent lymphomas. Hosts with evidence of HLA-A/B supertype homozygosity had a higher risk of death from PD (p<0.01). We also found that while antiretroviral treatment (ART) suppressed circulating HIV RNA as expected, HIV cfDNA was not affected by ART, likely measuring latent HIV infected reservoirs. Pretreatment mutant ctDNA levels correlated with cell-free EBV (cfEBV) levels in EBER+HIV+ lymphomas (p<0.01). EBER– cases exhibited more frequent and diverse somatic alterations, including evidence of genomic hypermutation from microsatellite instability. HIV+cfEBV-/DLBCL showed higher mutant frequencies of TP53 and epigenetic modifier genes, while HIV+cfEBV+/DLBCL had fewer distinguishing recurrent mutations. Compared to HIV–/DLBCL, HIV+/DLBCL showed more frequent and diverse TP53 alterations but fewer MYD88, CD79B, and PIM1 alterations. Classifying by LymphGen (Wright, Cancer Cell 2020), we identified a higher prevalence of unclassified cases (“Other”, p<0.05) and a lower prevalence of MCD than HIV-/DLBCL, with EBER+ cases more often unclassified than EBER– cases. We applied PhasED-Seq (Kurtz, Nat Biotechnol 2021) to monitor ctDNA minimal residual disease (MRD) during chemoimmunotherapy in serial HIV+ cases. The presence of MRD after 3 cycles of therapy strongly predicted disease-free survival (DFS) (p<0.05). Considering diverse clinical, genomic and viromic features, cfEBV level, IPI, and non-GCB histology each remained independently prognostic of DFS outcomes (p<0.05) in HIV+ lymphomas. Conclusions: Noninvasive genomic profiling is readily feasible and identifies distinct viral, immune, and somatic alteration signatures that characterize HIV-associated lymphomas and distinguish them from immunocompetent counterparts. Broad viral activity at diagnosis predicts poorer outcomes in HIV+ lymphomas, underscoring the need for controlling viral reactivation. The diversity of host and EBV subtypes and host MRD response dynamics holds promise not only for monitoring treatment efficacy but could also inform early lymphoma detection in people living with HIV.