Number of predicted tumour-neoantigens as biomarker for cancer immunotherapies

Abstract

Background: Cancer immunotherapies, like immune checkpoint inhibition, aim to boost the immune response towards the tumour. Tumours possessing a high mutational load are expected to be sensitive to immune therapy, since they likely expressing a large number of neo-antigens. Presentation of somatic mutation-derived neoantigens via HLA molecules on the surface of tumour cells is important for their recognition and destruction by cytotoxic T cells. Concordantly, mutational load of the tumour, the quality and quantity of those neoantigens were shown to be of high prognostic and therapeutic value. Aim of this work was to establish a workflow using somatic tumour sequencing data to calculate mutational load and identify tumour-specific neo-antigens which may serve as biomarkers for the efficacy of personalised cancer immunotherapies. Methods: Somatic mutations were identified by exome sequencing of normal and tumour tissue. The mutational load is defined as the number of somatic SNV-, InDeland essential splice site mutations per megabase of coding DNA. Truncating mutations in tumour suppressor genes as well as somatic mutations with an in-house frequency of-1% are not accounted. After HLA typing using exome data, computer algorithms (SYFPEITHI, netMHC-3.0 and netMHCpan-2.4) were applied to predict those neoantigen-peptides which bind with high affinity to the patient's HLA class I molecules. While the total number of neoantigen-peptides may already be predictive for the efficacy of immune checkpoint inhibition, we also established sophisticated selection criteria to prioritise the predicted neoantigens for targeting by cancer vaccines or adoptive T cell therapies Results: To-date, 107 tumour samples have been analysed from 102 patients suffering from cancers of diverse origin. The mutational load ranged from 0.5 to 51 mutations/ megabase in this group. The automatically determined mutational load correlates with the number of the individually verified non-synonymous somatic mutations. As expected, the number of non-synonymous somatic mutations correlated well with the total amount of predicted neoantigens. A median of 100 somatic missense SNVs led to a median of 303 neoantigen-peptides predicted to bind at least to one of the patients' HLA alleles. For 67% (median) of mutations one or more neoantigen-peptides were predicted (median= 60 peptides/patient). These parameters can be useful in the future to estimate the quantity of neoantigens, if the mutational load or the number of somatic non-synonymous mutations is known. Conclusions: Tumour panel or exome sequencing and subsequent neoantigen prediction have revealed that the number of somatic non-synonymous SNVs and total number of predicted neoantigens correlate well with each other (ratio 1:3). How well both parameters predict clinical outcome of personalised cancer immunotherapies needs to be further verified in large prospective trials. For targeted immunotherapies (e.g. cancer vaccines or adoptive T cell therapies) sophisticated selection criteria were established to prioritise the most promising neoantigen targets. Drugs targeting immune checkpoint inhibitors were successful in few tumour entities with high mutational load. The mutational load could serve as biomarker for a broad and differentiate applicability of immune-checkpoint-inhibitors.