S236: high-breadth sequencing of circulating tumor dna identifies novel classification of hodgkin lymphoma

Background: Hodgkin lymphoma (HL) arises from transformed B-cells and is one of the most frequent malignancies in young adults. Current clinical challenges include long-term side effects of front-line therapies and difficult-to-treat relapsed/refractory disease. A deconvolution of different HL subtypes based on genetic and tumor microenvironment (TME) features might be the key for an improved understanding of HL pathogenesis and thus improve therapeutic efforts. Aims: To decipher biologically distinct subtypes of HL noninvasively, thus ultimately enabling further individualization of HL therapies. Methods: We performed circulating tumor (ct)DNA sequencing of n=243 patients with HL from the German Hodgkin Study Group (GHSG) HD21 (NCT02661503) and NIVAHL (NCT03004833) trials. We performed non-negative matrix factorization to detect genetically distinct clusters of HL and gene expression profiling of the tumor tissue to assess potential differences in the TME between these clusters. Results: ctDNA sequencing was successful in the majority of patients and yielded mutational frequencies comparable to previously published cohorts by others and us (Spina et. al., Blood, 2018; Camus et. al., Haematologica, 2020; Sobesky et. al., Med, 2021). Using recurrent mutations and copy number variations (CNV) as features, we were able to detect three genetically distinct clusters of HL by non-negative matrix factorization. Interestingly, these clusters showed significant differences regarding several clinical factors. Cluster 1 was defined by high-level 9p24.1 amplifications (harboring PD-1 and PD-L1), a modest mutational burden, high frequency of B-symptoms as a sign of general inflammation, and impaired early response to therapy. In contrast, Cluster 2 depicted a particularly low mutational burden and frequent evidence of Epstein-Barr-Virus (EBV) infection. Cluster 3 was mainly characterized by a high mutational burden, frequent alteration in genes involved in previously described core HL oncogenic pathways and an absence of B-symptoms. Next, we assessed human leukocyte antigen (HLA) haplotypes. Several haplotypes were protective or predisposing for HL when comparing haplotype frequencies to healthy donors. Using HLA haplotypes to predict neoantigens, we found differences in potential neoantigen frequencies between clusters. Strikingly, when assessing the TME through gene expression profiling of tumor tissue, we were again able to detect three distinct subtypes. These subtypes showed high concordance with the previously defined genetic clusters (Fig 1). While the TME of Cluster 1 and Cluster 2 showed significantly different gene expression profiles, Cluster 3 was characterized by a downregulation of most genes enriched in the other clusters. Cluster 2 was mainly characterized by the presence of cytotoxic CD8 cells, NK-cells, neutrophils, and eosinophils, thus indicating an inflammatory TME driven by myeloid cells as well as an adaptive immune response. In contrast, Cluster 1 showed enriched fractions of B-cells, Tfh-cells and Treg-cells, supporting an evasion of the immune response by the underlying HL. Summarizing their individual features, we named the three detected clusters Inflammatory immune escape HL (Cluster 1), EBV/EBV-like HL (Cluster 2), and Oncogenic driven HL (Cluster 3). Summary/Conclusion: We here present a novel biological classification of HL that consists of three distinct subtypes exhibiting different genetic, clinical and TME features. Consequently, this classification might improve the biological understanding of HL and pave the way towards further individualization of HL therapy.Keywords: Hodgkin’s lymphoma, Genetic, liquid biopsy, ctDNA