Showing papers by "Sung-min Kim published in 2022"

1286 Structure-based prediction of neoantigens paired with T cell receptors on phenotype-selected CD8+ tumor-infiltrating lymphocytes

Abstract:

Background

Vaccination by tumor neoantigens are promising immunotherapy by providing more tumor-reactive T cell pool, which could be boosted by anti-PD-1. Identification of neoantigens cognate to tumor-reactive tumor-infiltrating lymphocytes (TILs) is critical for clinical efficacy of neoantigen vaccines. Here, we developed in silico neoantigen prediction platform by structure-based pairing with T cell receptors (TCR) on pre-existing CD8+ TILs selected by single-cell transcriptome profiles. Neoantigens derived by our strategy reflect in vivo immunogenicity and tumor reactivity.

Methods

Tumor resections from solid cancer patients are subject to whole exome and transcriptome sequencing. In addition, TILs from the patients are subject to scRNAseq/scTCRseq to stratify TCRs of TILs into target TILs and non-target TILs by single-cell transcriptome profiles. Neoantigen epitopes are filtered with HLA binding/immunogenicity and prioritized by tricomplex (TCR-peptide-HLA) structure-based TCR binding score from our platform Vacinus. To evaluate the immunogenicity of selected neoantigens, the neoantigen peptides are tested for in vitro IFNg ELISPOT assay using peripheral blood mononuclear cells (PBMCs) from the same patients, which sensitively detects antigen-experienced T cells.

Results

We have screened the immunogenicity of 286 neoantigens derived from Vacinus platform for 34 solid cancer patients. In particular, we could detect the immunogenic neoantigens in the majority of Hepatocellular carcinoma (HCC) patients (13/14) which had low mutational burden with median 95 (46-373) non-synonymous mutations. Single-cell transcriptome of CD8 TILs from HCC revealed that TILs had primarily exhausted/cytotoxic phenotype (target TIL) and non-exhausted memory phenotypes (non-target TIL). We applied TCRs derived from target or non-target TILs to select cognate neoantigens predicted by Vacinus platform, which are tier1 and tier2 neoantigens, respectively. Interestingly, when those neoantigens were tested for immunogenicity with IFNg ELISPOT assay, higher T cell responses were detected in tier1 (31%) than tier2 neoantigens (17%), reflecting tier1 neoantigens have more capability to induce in vivo immunogenicity in cancer patients. Using mouse tumor models, we are investigating therapeutic efficacy of tier1 neoantigens.

Conclusions

It is feasible to develop cancer neoantigen vaccine in HCC which has low mutational burden. Neoantigens paired with TILs in an activated state were identified to have greater immunogenicity compared to TILs in a memory state, underscoring the selection of target TILs. Neoantigen prediction by structure-based pairing of neoantigens and phenotype-selected TILs showed promising potential to better select therapeutically-relevant cancer vaccines.

Ethics Approval

The study with samples from HCC patients was approved by the Asan Medical Center Institutional Review Board (IRB) (2022-0263)

Markerless registration approach using dynamic touchable region model

Abstract: Markerless registration is required for image‐guided surgery; it has limited accuracy due to the ambiguity of the correspondence point set. In this study, we proposed a registration framework to improve registration accuracy for markerless registration using a dynamic touchable region model (DTRM).

74 DeepTCRMatch: An effective way of computing T cells antigen specificity

Abstract: achieves accuracy 0.3 for an independent dataset. We also test super-vised task: whether our model can induce probable cognate antigens for a given CDR3. Our model achieves precision 0.7. Conclusions Our deep learning model can extract k -mer infor-mation that only represents antigen specificity. This informa-tion is an invaluable numerical vector for computing similarity of antigen specificity. By doing this, our model can solve the one-of-many problem and predict the antigen specificity. In the future, our model will improve its performance as a size of training dataset grows. preprint.

Structure-based prediction of neoantigens paired with T cell receptors on pre-existing CD8+ tumor-infiltrating lymphocytes.

Abstract: e14583 Background: Tumor neoantigens have ability to expand tumor-specific T cell immunity, which could be boosted by anti-PD-1 immunotherapy. Therapeutically relevant neoantigens should be presented on patient HLAs and recognized by tumor-reactive T cells. Despite tumor-infiltrating lymphocytes (TILs) have enriched tumor-reactive T cell populations, most tumor-specific TILs exhibit exhausted phenotype and have limitations to in vitro screening. Here, we developed in silico structure-based prediction of neoantigen discovery pipeline by pairing with T cell receptors (TCRs) on pre-existing CD8+ exhausted tumor-infiltrating lymphocytes. Methods: Tumor resections from patients with liver, stomach, colorectal, and lung cancers are subject to whole exome and transcriptome. In addition, TILs from the patients are subject to scRNA/TCRseq to select TCRs of exhausted T cells, which are mostly thought to have tumor-reactivity. Neoantigen epitopes are prioritized from our in silico pipeline including HLA binding and TCR-peptide-HLA structure-based pairing score. To evaluate the immunogenicity of selected neoantigens, the neoantigens are tested for in vitro IFNg ELISPOT assay using peripheral blood mononuclear cells (PBMCs) from the same patients, which sensitively detects antigen-experienced T cells. Results: For 5 patients with T cell-reactive neoantigens, among 23 neoantigens selected by our in silico pipeline and tested in vitro, 10 neoantigens showed positive pre-existing T cell responses by matched patients. Then, we analyzed gene expression profiles of TILs paired in silico with in vitro T cell response-positive or negative neoantigens (pTILs and nTILs, respectively) combined with TILs from 23 solid cancer patients. Interestingly, pTILs were mostly mapped to the clusters which highly expressed markers of exhaustion and cytotoxicity, while nTILs showed random distribution across all the clusters. Conclusions: Neoantigen prediction using structure-based pairing of neoantigens and phenotype-selected TILs showed promising potential to better select therapeutically-relevant cancer vaccines. This data supports further investigation of our in silico pipeline into more patient samples and preclinical studies in mouse tumor models.