Identification of Genetic Subtypes in Follicular Lymphoma

Introduction: Follicular lymphoma (FL) is the most common indolent non-Hodgkin lymphoma subtype and is characterized by heterogeneous clinical trajectories. While certain genetic abnormalities are highly prevalent, heterogeneity exists also at the molecular level. Yet, unlike diffuse large B-cell lymphoma, where genetic subtypes have been described in multiple studies, the taxonomy of FL has not been clearly established to date. While clusters defined by STAT6 mutations or aberrant somatic hypermutation have been described in FL by Crouch et al. (Blood Advances, 2022), its landscape of molecular alterations has been insufficiently dissected into biologically distinct subgroups. The recognition of such subtypes may pave the way for treatment individualization. Here, we sought to define novel genetic subtypes in FL by means of targeted DNA sequencing and clustering.

Methods: Samples of FL grade 1 to 3A (i.e. classic FL) were accrued from biorepositories of participating centres in Canada, Northern Europe and Australia, as well as from two prospective clinical trials (E4402 and E2408). We performed hybridization-based targeted DNA-sequencing of 57 genes that are recurrently mutated in FL. Moreover, we included targeted DNA sequencing data from a previously published dataset (PMID 27959929). All cases were analyzed using a uniform computational pipeline. Samples were removed from the analysis if the mean on-target coverage was <50X. Single nucleotide variants were identified using Mutect2 and we applied a variant allele fraction cut-off of 0.1. Bernoulli mixture-model clustering was applied using variants as features for the identification of genetic subtypes. Cluster stability was assessed by bootstrap consensus. Clusters were tested for their association with outcome measures. As a further means to annotate genetic clusters, we generated genome-wide methylation data using the Illumina EPIC platform.

Results: Following quality control, the total number of pre-treatment samples available for clustering was 715 (543 from formalin-fixed and paraffin-embedded tissue samples and 172 from fresh-frozen). The average number of mutations and mutated genes per sample were 5.9 (range 1-20) and 4.8 (range 0-14), respectively. An analysis of interactions between somatic mutations revealed both expected (e.g. mutual exclusivity of CREBBP and EP300 mutations) and unexpected patterns (e.g. co-occurring mutations of GNA13 and MEF2B), overall justifying the pursuit of unsupervised clustering. After consensus clustering, the optimal clustering solution resolved 5 genetically distinct groups, henceforth referred to as C1 to C5 (Figure). The C1 cluster was characterized by variants in genes involved in the mTOR signaling pathway (ATP6AP1, RRAGC, ATP6V1B2); the C2 cluster by low mutation burden; the C3 cluster by mutations involving TNFAIP3, TP53 and MYD88; the C4 cluster by GNA13 and MEF2B mutations; and the C5 cluster by STAT6, CREBBP, EZH2 and TNFRSF14 mutations. The number of mutations located in RGYW or WRCY motifs, suggestive of somatic hypermutation, was highest in the C4 and the C3 clusters (P < 0.001). Applying a previously reported methylation-based clock (epiCMIT) to the subset of samples with methylation data (n=345), we found that samples in the C2 cluster had the lowest proliferative history. Concordantly, the C2 cluster was characterized by a lower degree of global hypomethylation and lower average methylation of CpG islands, whereas samples in clusters C1, C3, C4 and C5 were characterized by not just longer proliferative histories, but also by higher mean methylation in CpG islands. In terms of clinical associations, the C2 cluster was significantly associated with limited-stage disease presentation. In an exploratory outcome analysis of 165 patients treated with immunochemotherapy, the C3 and C5 clusters were associated with inferior time to progression.

Conclusion: Here, we identified novel genetic subtypes that are distinguishable not just by unique sets of mutations, but also by their clinical associations and methylation profiles. The novel genetic profiles shed light on core pathways distinctly tied to each subtype. Targeted DNA sequencing of a limited gene panel represents a potentially accessible method of subtyping FL in the clinical setting, with implications for the understanding of responses to targeted therapy.

Johnson:Merck, AbbVie, Roche, Gilead: Consultancy. Baetz:Astrazeneca: Honoraria; Gilead: Honoraria; Roche: Honoraria; Servier: Honoraria; BeiGene: Honoraria. Gandhi:Beigene: Research Funding; Janssen: Research Funding. Steidl:Abbvie: Consultancy; Bayer: Consultancy; Bristol Myers Squibb: Consultancy; Curis Inc: Consultancy; Epizyme: Research Funding; Roche: Consultancy; Seattle Genetics: Consultancy; Trillium Therapeutics: Research Funding. Kahl:TG Therapeutics: Consultancy; AstraZeneca: Consultancy, Research Funding; ADT Therapeutics: Consultancy; Roche: Consultancy; Genentech: Consultancy, Research Funding; Abbvie: Consultancy, Research Funding; MEI: Consultancy; AcertaPharma: Consultancy; Pharmacyclics: Consultancy; Celgene/BMS: Consultancy, Research Funding; Beigene: Consultancy, Research Funding; Kite: Consultancy; Janssen: Consultancy; Incyte: Consultancy; Hutchmed: Consultancy, Research Funding; Genmab: Consultancy; Seattle Genetics: Consultancy; Research To Practice: Speakers Bureau. Kridel:Abbvie: Research Funding.