Importance of therapeutic neoantigen selection
It is highly important to carefully refine the chosen selection of tumour epitopes used in therapeutics, to induce an immune response (only) against the expected tumour antigens. Selection of falsely identified epitopes in the final therapy will not correctly redirect the immune system against the tumour and could potentially even induce immune-related adverse events (irAEs) if the immune system is brought towards imbalance. Overstimulation can push the immune system to supraphysiological levels with a subsequent risk of auto-immune disorders.
On the other hand, it is essential that the refined set of antigens is representative for the whole tumour, taking subclonality within the tumour and its metastases into account. As not all alteration events are shared across tumour cells, their neoantigen repertoire knows strong differences. For this purpose, the sample gathering protocol of myNEO aims to get a representative outlook of the tumour block and the technology tools are able to convolute results from multiple biopsy samples. Next to that, the bioinformatic filtering prefers tumour alterations with high allelic frequency that are functionally important (driver mutations) to increase the probability of selecting highly clonal epitopes.
To limit immune escape, wherein outgrowth occurs of tumour cells that do not possess the mutation targeted by therapy (or have reverted it), multiple valuable well-represented alterations are targeted at the same time.
"myNEO carefully selects 20 well-validated immunogenic neoantigens for each individual patient - out of the predicted list containing hundreds of peptides - to target during immunotherapy"
After the sequencing-based neoantigen prediction, the selected neoantigens can be validated on different levels. Firstly, the alteration that caused the tumour-specific epitope can be confirmed via Sanger-seq / Amplicon-based sequencing on a second biopsy sample. This confirmation step is usually performed just before therapy administration to ensure that the patient is still possessing the alteration (and its occurrence was not altered due to a previous round of chemo-/radiotherapy).
Secondly, the presence of the antigen on the tumour surface is confirmed. This involves performing a tandem mass spectrometry analysis of the HLA (I&II) ligandome on the tumour tissue, the only unbiased methodology to interrogate the repertoire of (naturally presented) HLA-binding peptides. However, performing an MS-analysis is not fit for routine clinical use due to its large biopsy sample requirements, extensive time- and labour costs, and limited sensitivity. To partially cope with these MS limitations, a cross-patient employable deep-learning algorithm (neoMS) predicts which antigens are most likely to be presented on the patient tumour cells. Due to its extensive training on over 2.5 million datapoints, the tool highlights the synergy possible between analytical and computational tools.
Thirdly, immunogenicity screening assays confirm whether the presented antigens on the tumour surface can elicit an immune response. These require peripheral blood mononuclear cells (PBMCs) or tumour-infiltrating lymphocytes (TILs) wherein stimulation of T-cells against certain neoantigens are observed. Several types of assays are available such as peptide-HLA tetramer assays, neoantigen-pulsed COS7 stimulation ELISPOT assays, and screening experiments of tandem minigenes or peptide pools. The most optimal strategy depends on the required set-up, but they are all characterised by certain key disadvantages.
"myNEO has developed the neoMS and neoIM algorithms, predicting antigen presence on the tumour surface and epitope immunogenicity across populations respectively. These tools highlight the synergy possible between analytical and computational tools."
Using ex-vivo assays as a filter for neoantigen selection limits the spectrum of T-cell reactivity to existing (and possibly ineffective) T-cell responses, not considering de novo response. Due to the cost- and labour intensiveness it is unfeasible for a high TMB tumour to screen all possible neoantigens. Besides the differential agretopicity index (DAI) proposed by Duan et al. that has shown suboptimal results, no in-silico tools exist that predict true epitope immunogenicity. For this purpose, myNEO has developed a tool predicting pattern-based epitope immunogenicity across populations. This neoIM algorithm predicts immunogenicity of tumour-specific surface peptides based on a selection of 10 principal components resulting out of a PCA analysis of 500 selected amino acid parameters.
Want to learn more about this?