Characterization of the Immunogenomic Landscape of Follicular Lymphoma

Background: Follicular Lymphoma (FL) is an indolent, but incurable form of non-Hodgkin lymphoma (NHL), with variable clinical behavior. The immune environment plays a major role in dictating the biology of FL. For example, gene expression profiling experiments have demonstrated that recurring "immune response" gene signatures in diagnostic FL specimens were more accurate in predicting clinical outcomes than gene expression patterns in FL cells themselves. Furthermore, immune cells, including T cells, are often found infiltrating or surrounding malignant follicles in FL, and our group has identified the presence of highly clonal T cell populations in most FL samples, suggesting that FL cells express antigens that drive the expansion of lymphoma-specific T cells. In addition, inactivating mutations in B₂ microglobulin are relatively common in FL, indicating that down-regulation of antigen expression may occur as a result of host immune surveillance. Lastly, FL is responsive to immunotherapies, such as PD-1 antibody blockade. Importantly, the FL antigens that promote spontaneous immune recognition remain unknown. Emerging data has revealed a critical relationship between the genetic landscape and the immunogenicity of cancer. Somatic mutations in malignant cells encode for altered proteins (i.e. neoantigens) that are "sensed" as foreign by T cells, and the mutational "load" of a given tumor has direct implications for the host immune response.

Methods: To address the hypothesis that somatic mutations generate neo-antigens which are the targets of T cell recognition in patients with FL, whole exome sequencing (WES) was performed on FACS-purified FL B cells (CD10⁺CD19⁺ cells) and matched T cells from 7 FL biopsy samples. HLA A, B and C haplotypes were predicted using PolySolver, OptiType, Phlat and Seq2hla software with >90% concordance. HLA binding affinity was predicted for all possible 8-11-mer peptides generated from each somatic mutation and corresponding wild-type peptides using NetMHCpan v2.8. These tiled peptides were analyzed for their predicted binding affinities (IC₅₀ nM) to HLA class I molecules. An IC₅₀ <150 nM was considered a predicted strong binder, ≥150 but ≤500 nM an intermediate to weak binder, and >500 nM a non-binder.

Results: A median number of 71 somatic mutations were identified in each sample (range, 8-135). Non-synonymous single nucleotide variants (SNV) were cataloged, and in silico analysis was utilized to predict potential high-affinity epitopes capable of binding patient-specific HLA molecules. We achieved an average of 90% proportion of nucleotides > Phred quality score 30 of total bases, according to the base calling accuracy (range, 89.02%-92.06%). G>A/C>T transitions were the most common nucleotide substitutions observed. Consistent with published reports, recurrent somatic mutations were identified in EZH2, CREBBP, IGLL5, MUC21, PCLO, and PIM1. We also identified recurrent mutations in BCL2 and MLL2 (frameshift insertions) known to dominate the mutational landscape of FL. Mutations in HLA class II alleles were common, occurring in 3/7 samples. In silico analysis was next employed to identify putative neo-antigens. A total of 139 candidate neo-antigenic peptides were predicted across all patients. On average, 19 neo-antigens were predicted for each patient (range, 3-41). No public neo-antigens were identified, suggesting that the neo-antigens arising in FL are patient-specific. This observation is consistent with what has been described for other cancer subtypes. Functional validation of lead candidate neo-antigen peptides is ongoing.

Conclusion: While overall numbers of somatic mutations are modest in FL, putative neo-antigens were identified in all patient samples analyzed. Our findings suggest that immune therapies incorporating neo-antigen-specific approaches, such as personalized vaccines, should be explored in FL patients.