On Thursday 29 September 2022,Helle Rus Povlsen will defend her PhD thesis "Development of Immunoinformatics Methods for Improved Rational Identification of T cell Epitopes".
Time: 13:30
Place: Building 303A, auditorium 49
Supervisor: Professor Morten Nielsen
Co-Supervisor: Associate Professor Leon Eyrich Jessen
Assessment committee:
Professor Ole Lund, DTU Health Tech
Director Marie Toussaint, GSK
Head of Bioinformatics Frederik Otzen Bagger, Rigshospitalet
Chairperson:
Associate Professor Elena Papaleo
Abstract:
The research projects presented in this thesis are centered around T cell specificity. T cells play a crucial role in maintaining health by eliminating intruding pathogens and malignant cell changes. This ability is granted via the T cell receptor (TCR), which interacts with peptides presented by MHC molecules on the surface of host cells. To ensure broad protection against any potential pathogen, the immune system has evolved to generate highly diverse TCRs which may recognize a wide range of targets. However, such a complex system is inevitably very challenging to study. Nevertheless, this thesis has been dedicated to investigate T cell specificity via popular experimental methods and develop immunoinformatic tools and analyses to enhance the yield of such methods.
The first project of the thesis provides a detailed overview of the full spectrum of exact antigens (minimal peptide epitope) within the viral genome of SARS-CoV-2. Using NetMHCpan 4.1 (Reynisson & Alvarez, 2020), we selected 2204 potential HLA binding peptides (9–11 amino acids) for experimental evaluation. These peptides were predicted to bind one or more of ten prevalent MHC-I molecules leading to a total 3141 peptide-MHC specificities for experimental evaluation. Using our large-scale T cell detection technology based on DNA-barcoded peptide-MHC multimers, we mapped T cell recognition throughout the complete SARS-CoV-2 genome, identified the exact epitopes recognized by SARS-CoV-2-specific CD8+ T cells, and characterized immunodominance of these epitopes in COVID-19 disease. Broad T cell recognition toward SARS-CoV-2-derived peptides was also identified in SARS-CoV-2 unexposed healthy individuals, with a large overlap in the peptide-MHC complexes recognized in the two groups. However, T cell recognition was substantially enhanced in the patient group, with SARS-CoV-2-reactive T cells accounting for up to 27% of all CD8+ T cells. Furthermore, we have evaluated the phenotypic characteristics of SARS-CoV-2-specific T cells and correlated their activation signatures with disease severity.
However, multimer staining only provides shallow insight into the complexity of TCR recognition of pMHCs. In order to truly understand the rules that govern T cell specificity, we employed single-cell sequencing, enabling the capture of TCRαβ-chains, the cognate pMHC provided by DNA-barcoded multimers, and hashing antibodies in the second project of the thesis. As single-cell data is polluted with multiple confounding factors, we developed a method to efficiently remove noise and retain accurate pairing of TCR-pMHC. The data-driven method called ATRAP (Accurate T cell Receptor and Antigen Pairing) consists of a series of filtering approaches to obtain increasingly accurate TCR-pMHC pairing. The first filtering step is based on identifying expected targets by comparing distributions of all pMHCs detected within a clonotype consisting of 10 or more clones. The key is to study clonotypes in ensemble rather than individually, to averaged out deviations. Based on single-cell matches with the expected targets, we could compute an accuracy score by which thresholds for pMHC counts were optimized. Another key step of ATRAP filtering is ensuring HLA correspondence between pMHC and the HLA haplotype of the T cell donor, assuming that a T cell is absolutely restricted to the allele for which it was selected during the thymocyte maturation process. We evaluated these filters by comparing similarity of TCR sequences. We showed an increased distinction between similarities of TCRs binding the same peptide (intra-specificity) relative to TCRs binding different peptides (inter-specificity), indicating that the ATRAP framework reliably removed noisy observations. We further evaluated the method by correlating the ATRAP filtered data with responses detected by multimer staining.
In the third and final project, we benchmarked ATRAP against a recently released method (ICON) to learn the advantages and disadvantages of each approach. The two methods distinctively differed by their prioritization between specificity and sensitivity of detecting TCR-pMHC pairs. ICON efficiently removed most of the ambiguous specificity annotations and thereby produced a dataset ready to use for modeling purposes. In contrast, ATRAP was designed to include ambiguous specificity annotations to enable study of cross-reactivity, however, at the expense of including more false positive observations. The two methods performed on par, with a slightly better distinction between intra- and inter-specificities provided by ATRAP.