T-cell Receptor Binding Research Articles

Differential roles of kinetic on- and off-rates in T-cell receptor signal integration revealed with a modified Fab’-DNA ligand

Antibody-derived T-cell receptor (TCR) agonists are commonly used to activate T cells. While antibodies can trigger TCRs regardless of clonotype, they bypass native T cell signal integration mechanisms that rely on monovalent, membrane-associated, and relatively weakly binding ligand in the context of cellular adhesion. Commonly used antibodies and their derivatives bind much more strongly than native peptide major histocompatibility complex (pMHC) ligands bind their cognate TCRs. Because ligand dwell time is a critical parameter that tightly correlates with physiological function of the TCR signaling system, there is a general need, both in research and therapeutics, for universal TCR ligands with controlled kinetic binding parameters. To this end, we have introduced point mutations into recombinantly expressed α-TCRβ H57 Fab to modulate the dwell time of monovalent Fab binding to TCR. When tethered to a supported lipid bilayer via DNA complementation, these monovalent Fab'-DNA ligands activate T cells with potencies well-correlated with their TCR binding dwell time. Single-molecule tracking studies in live T cells reveal that individual binding events between Fab'-DNA ligands and TCRs elicit local signaling responses closely resembling native pMHC. The unique combination of high on- and off-rates of the H57 R97L mutant enables direct observations of cooperative interplay between ligand binding and TCR-proximal condensation of the linker for activation of T cells, which is not readily visualized with pMHC. This work provides insights into how T cells integrate kinetic information from TCR ligands and introduces a method to develop affinity panels for polyclonal T cells, such as cells from a human patient.

HLA A*24:02-restricted T cell receptors cross-recognize bacterial and preproinsulin peptides in type 1 diabetes.

CD8+ T cells destroy insulin-producing pancreatic β cells in type 1 diabetes through HLA class I-restricted presentation of self-antigens. Combinatorial peptide library screening was used to produce a preferred peptide recognition landscape for a patient-derived T cell receptor (TCR) that recognized the preproinsulin-derived (PPI-derived) peptide sequence LWMRLLPLL in the context of disease risk allele HLA A*24:02. Data were used to generate a strong superagonist peptide, enabling production of an autoimmune HLA A*24:02-peptide-TCR structure by crystal seeding. TCR binding to the PPI epitope was strongly focused on peptide residues Arg4 and Leu5, with more flexibility at other positions, allowing the TCR to strongly engage many peptides derived from pathogenic bacteria. We confirmed an epitope from Klebsiella that was recognized by PPI-reactive T cells from 3 of 3 HLA A*24:02+ patients. Remarkably, the same epitope selected T cells from 7 of 8 HLA A*24+ healthy donors that cross-reacted with PPI, leading to recognition and killing of HLA A*24:02+ cells expressing PPI. These data provide a mechanism by which molecular mimicry between pathogen and self-antigens could have resulted in the breaking of self-tolerance to initiate disease.

Lipidomic scanning of self-lipids identifies headless antigens for natural killer T cells

To broadly measure the spectrum of cellular self-antigens for natural killer T cells (NKT), we developed a sensitive lipidomics system to analyze lipids trapped between CD1d and NKT T cell receptors (TCRs). We captured diverse antigen complexes formed in cells from natural endogenous lipids, with or without inducing endoplasmic reticulum (ER) stress. After separating protein complexes with no, low, or high CD1d-TCR interaction, we eluted lipids to establish the spectrum of self-lipids that facilitate this interaction. Although this unbiased approach identified fifteen molecules, they clustered into only two related groups: previously known phospholipid antigens and unexpected neutral lipid antigens. Mass spectrometry studies identified the neutral lipids as ceramides, deoxyceramides, and diacylglycerols, which can be considered headless lipids because they lack polar headgroups that usually form the TCR epitope. The crystal structure of the TCR-ceramide-CD1d complex showed how the missing headgroup allowed the TCR to predominantly contact CD1d, supporting a model of CD1d autoreactivity. Ceramide and related headless antigens mediated physiological TCR binding affinity, weak NKT cell responses, and tetramer binding to polyclonal human and mouse NKT cells. Ceramide and sphingomyelin are oppositely regulated components of the "sphingomyelin cycle" that are altered during apoptosis, transformation, and ER stress. Thus, the unique molecular link of ceramide to NKT cell response, along with the recent identification of sphingomyelin blockers of NKT cell activation, provide two mutually reinforcing links for NKT cell response to sterile cellular stress conditions.

TSpred: a robust prediction framework for TCR-epitope interactions using paired chain TCR sequence data.

Prediction of T-cell receptor (TCR)-epitope interactions is important for many applications in biomedical research, such as cancer immunotherapy and vaccine design. The prediction of TCR-epitope interactions remains challenging especially for novel epitopes, due to the scarcity of available data. We propose TSpred, a new deep learning approach for the pan-specific prediction of TCR binding specificity based on paired chain TCR data. We develop a robust model that generalizes well to unseen epitopes by combining the predictive power of CNN and the attention mechanism. In particular, we design a reciprocal attention mechanism which focuses on extracting the patterns underlying TCR-epitope interactions. Upon a comprehensive evaluation of our model, we find that TSpred achieves state-of-the-art performances in both seen and unseen epitope specificity prediction tasks. Also, compared to other predictors, TSpred is more robust to bias related to peptide imbalance in the dataset. In addition, the reciprocal attention component of our model allows for model interpretability by capturing structurally important binding regions. Results indicate that TSpred is a robust and reliable method for the task of TCR-epitope binding prediction. Source code is available at https://github.com/ha01994/TSpred.

HeteroTCR: A heterogeneous graph neural network-based method for predicting peptide-TCR interaction

Identifying interactions between T-cell receptors (TCRs) and immunogenic peptides holds profound implications across diverse research domains and clinical scenarios. Unsupervised clustering models (UCMs) cannot predict peptide-TCR binding directly, while supervised predictive models (SPMs) often face challenges in identifying antigens previously unencountered by the immune system or possessing limited TCR binding repertoires. Therefore, we propose HeteroTCR, an SPM based on Heterogeneous Graph Neural Network (GNN), to accurately predict peptide-TCR binding probabilities. HeteroTCR captures within-type (TCR-TCR or peptide-peptide) similarity information and between-type (peptide-TCR) interaction insights for predictions on unseen peptides and TCRs, surpassing limitations of existing SPMs. Our evaluation shows HeteroTCR outperforms state-of-the-art models on independent datasets. Ablation studies and visual interpretation underscore the Heterogeneous GNN module’s critical role in enhancing HeteroTCR’s performance by capturing pivotal binding process features. We further demonstrate the robustness and reliability of HeteroTCR through validation using single-cell datasets, aligning with the expectation that pMHC-TCR complexes with higher predicted binding probabilities correspond to increased binding fractions.

Dynamic allostery in the peptide/MHC complex enables TCR neoantigen selectivity.

The inherent cross-reactivity of the T cell receptor (TCR) is balanced by high specificity, which often manifests in confounding ways not easily interpretable from static structures. We show here that TCR discrimination between an HLA-A*03:01 (HLA-A3)-restricted public neoantigen derived from mutant PIK3CA and its wild-type (WT) counterpart emerges from motions within the HLA binding groove that vary with the identity of the peptide's first primary anchor. The motions form a dynamic gate that in the complex with the WT peptide impedes a large conformational change required for TCR binding. The more rigid neoantigen is insusceptible to this limiting dynamic, and with the gate open, is able to transit its central tryptophan residue underneath the peptide backbone to the contralateral side of the HLA-A3 peptide binding groove, facilitating TCR binding. Our findings reveal a novel mechanism driving TCR specificity for a cancer neoantigen that is rooted in the dynamic and allosteric nature of peptide/MHC-I complexes, with implications for resolving long-standing and often confounding questions about the determinants of T cell specificity.

Development of a Stable Peptide-Major Histocompatibility Complex (MHC) via Sortase and Click Chemistry.

T cells play a crucial role in antitumor immune responses and the clearance of infected cells. They identify their targets through the binding of T-cell receptors (TCRs) to peptide-major histocompatibility complex (pMHC) molecules present in cancer cells, infected cells, and antigen-presenting cells. This interaction is often weak, requiring multimeric pMHC molecules to enhance the avidity for identifying antigen-specific T cells. Current exchangeable pMHC-I tetramerization methods may overlook TCRs recognizing less stable yet immunogenic peptides. In vivo applications targeting antigen-specific T cells demand the genetic synthesis of a pMHC fusion for each unique peptide antigen, which poses a significant challenge. To address these challenges, we developed a sortase and click chemistry-mediated approach for generating stable pMHC molecules. Leveraging sortase technology, we introduced an azide click-handle near the N-terminus of β2m, proximal to the MHC-peptide-binding groove. Simultaneously, the peptide was engineered with a multi glycine linker and a C-terminal alkyne click-handle. Azide-alkyne click reactions efficiently immobilized the peptide onto the MHC molecule, providing a versatile and efficient method for pMHC generation. The resulting peptide-clicked-MHC specifically binds to its cognate TCR and remains stable for over 3 months at 4 °C in the absence of any additional free peptide. The stability of the pMHC and its affinity to cognate TCRs are influenced by the linker's nature and length. Multi glycine linkers outperform poly(ethylene glycol) (PEG) linkers in this regard. This technology expands the toolkit for identifying and targeting antigen-specific T cells, enhancing our understanding of cancer-specific immune responses, and has the potential to streamline the development of personalized immunotherapies.

Enhanced T cell receptor specificity through framework engineering.

Development of T cell receptors (TCRs) as immunotherapeutics is hindered by inherent TCR cross-reactivity. Engineering more specific TCRs has proven challenging, as unlike antibodies, improving TCR affinity does not usually improve specificity. Although various protein design approaches have been explored to surmount this, mutations in TCR binding interfaces risk broadening specificity or introducing new reactivities. Here we explored if TCR specificity could alternatively be tuned through framework mutations distant from the interface. Studying the 868 TCR specific for the HIV SL9 epitope presented by HLA-A2, we used deep mutational scanning to identify a framework mutation above the mobile CDR3β loop. This glycine to proline mutation had no discernable impact on binding affinity or functional avidity towards the SL9 epitope but weakened recognition of SL9 escape variants and led to fewer responses in a SL9-derived positional scanning library. In contrast, an interfacial mutation near the tip of CDR3α that also did not impact affinity or functional avidity towards SL9 weakened specificity. Simulations indicated that the specificity-enhancing mutation functions by reducing the range of loop motions, limiting the ability of the TCR to adjust to different ligands. Although our results are likely to be TCR dependent, using framework engineering to control TCR loop motions may be a viable strategy for improving the specificity of TCR-based immunotherapies.

EPIC-TRACE: predicting TCR binding to unseen epitopes using attention and contextualized embeddings.

T cells play an essential role in adaptive immune system to fight pathogens and cancer but may also give rise to autoimmune diseases. The recognition of a peptide-MHC (pMHC) complex by a T cell receptor (TCR) is required to elicit an immune response. Many machine learning models have been developed to predict the binding, but generalizing predictions to pMHCs outside the training data remains challenging. We have developed a new machine learning model that utilizes information about the TCR from both α and β chains, epitope sequence, and MHC. Our method uses ProtBERT embeddings for the amino acid sequences of both chains and the epitope, as well as convolution and multi-head attention architectures. We show the importance of each input feature as well as the benefit of including epitopes with only a few TCRs to the training data. We evaluate our model on existing databases and show that it compares favorably against other state-of-the-art models. https://github.com/DaniTheOrange/EPIC-TRACE.

627. Bioinformatic Approach to Design a Plasmodium falciparum PfRipr Multi-Epitope Vaccine Construct

Abstract Background Malaria infections remain an enormous contributor to global fatalities with an estimated death toll of 619,000 in 2021 and 77% of deaths being children aged under 5 years. Transmitted via mosquito, Plasmodium falciparum is the most lethal parasite of its genus but has evaded many treatment and vaccine efforts due to its complex life cycle and redundant invasion mechanisms. Epitope-based vaccines hold significant promise for malaria vaccine development due to their ease of development and ability to target dominant regions in antigenically variable pathogens. The recently characterized protein P. falciparum merozoite Rh5 interacting protein (PfRipr) is nonredundant, highly conserved, and essential for erythrocyte invasion, making it an ideal target for a bloodstage malaria vaccine. Merozoite Invasion Model of binding and insertion of PRCR complex to erythrocyte basigin in merozoite invasion. Adapted from Scally et al. 2022. Methods Using P. falciparum sequences collected from Burkina Faso and Uganda, we assessed the immunogenic potential of PfRipr epitopes with regard to T-cell receptor binding and B-cell recognition. T-cell receptor binding was predicted using NetMHCpan searching against MHC I and II alleles with high regional frequencies. Using an in-silico 3D model of PfRipr predicted via AlphaFold, tertiary structures of all PfRipr sample sequences were predicted via SWISS-MODEL then analyzed by ElliPro to identify linear and discontinuous B-cell epitopes. Putative epitopes were filtered using allele coverage, conservation, antigenicity, and allergenicity. B-Cell Epitope Prediction Protocol for B-cell epitope prediction using AlphaFold for template modeling, Swiss-Model for homology-based modeling of sample sequences, and ElliPro for continuous and discontinuous epitope prediction.​ Epitope Sorting and Filtering Protocol for sorting T-cell epitopes using Average Allele Coverage (percentage of alleles in which the protein sequence was identified as an epitope) and B-cell epitopes using ElliPro Score (Protrusion Index representing surface accessibility of the epitope).​ Results Between the two datasets, there were 19 matching epitopes with 7 MHC I, 9 MHC II, 1 linear, and 2 discontinuous. These epitopes were used to design a multi-epitope-based bloodstage vaccine construct against P. falciparum. MHC Epitopes Partial list of MHC epitopes found in Burkina Faso and Uganda datasets. B-Cell Epitopes B-cell epitopes found in Burkina Faso and Uganda datasets. PfRipr 3D Structure with B-cell Epitopes 3D model of PfRipr predicted via AlphaFold with labeled B-cell epitopes Conclusion To validate their predicted immunogenicity, epitopes can be further investigated using in silico protein stabilization and docking simulations, in vitro methods such as HLA stabilization or T-cell activation assays, and in vivo methods using transgenic mouse models. The pipeline of immunoinformatic analyses formulated in this project can be further applied to P. falciparum sequence datasets collected from other malaria endemic regions to develop vaccines effective against circulating strains. Multi-Epitope Vaccine Construct Schematic representation of hypothetical multi-epitope vaccine construct. FliC is added to the N-terminus to stimulate and amplify immune response. The EAAAK linker is used to separate the bifunctional fusion protein domains. AAY linkers are used to amplify MHC I and II epitopes and prevent junctional epitope binding. GPGPG linkers between B-cell epitopes function to minimize junctional epitopes and retain conformational-dependent immunogenicity. A six amino acid Histidine (6H) chain was added to the C-terminus to aid during lab purification processes. Disclosures All Authors: No reported disclosures

GGNpTCR: A Generative Graph Structure Neural Network for Predicting Immunogenic Peptides for T-cell Immune Response.

Identifying the interactions between T-cell receptor (TCRs) and human antigens is a crucial step in developing new vaccines, diagnostics, and immunotherapy. Current methods primarily focus on learning binding patterns from known TCR binding repertoires by using sequence information alone without considering the binding specificity of new antigens or exogenous peptides that have not appeared in the training set. Furthermore, the spatial structure of antigens plays a critical role in immune studies and immunotherapy, which should be addressed properly in the identification of interacting TCR-antigen pairs. In this study, we introduced a novel deep learning framework based on generative graph structures, GGNpTCR, for predicting interactions between TCR and peptides from sequence information. Results of real data analysis indicate that our model achieved excellent prediction for new antigens unseen in the training data set, making significant improvements compared to existing methods. We also applied the model to a large COVID-19 data set with no antigens in the training data set, and the improvement was also significant. Furthermore, through incorporation of additional supervised mechanisms, GGNpTCR demonstrated the ability to precisely forecast the locations of peptide-TCR interactions within 3D configurations. This enhancement substantially improved the model's interpretability. In summary, based on the performance on multiple data sets, GGNpTCR has made significant progress in terms of performance, universality, and interpretability.

In Silico Prediction of Plasmodium falciparum PfRipr Epitopes as Vaccine Candidates

Abstract Background The global malarial death toll in 2020 was ~627,000, 77% of which occurred in children <5 years of age. Plasmodium falciparum is the most lethal parasite of its genus but has evaded vaccine development due to its complex life cycle and redundant invasion mechanisms. Epitope-based vaccines hold significant promise due to their ability to target dominant and/or conserved regions in highly antigenically variable pathogens, including malaria. The recently characterized P. falciparum merozoite Rh5 interacting protein (PfRipr) is nonredundant, highly conserved, and essential for erythrocyte invasion, making it an ideal target for a bloodstage malaria vaccine. (Figure 1) We sought to analyze P. falciparum sequences from highly endemic areas using bioinformatic approaches that identify candidate PfRipr epitope-based bloodstage vaccine antigens for inclusion in a multiantigen vaccine. Figure 1: Model of binding and insertion of PRCR complex to erythrocyte basigin in merozoite invasion. Adapted from Scally et al. 2022. (https://doi.org/10.1038/s41564-022-01261-2 , https://creativecommons.org/licenses/by/4.0/) Methods Using two P. falciparum sequence datasets from Burkina Faso and Uganda, we assessed the immunogenic potential of PfRipr epitopes for T-cell receptor binding and B-cell recognition. T-cell receptor binding was predicted using NetMHCpan and NetMHCIIpan searching against MHC I and II alleles with high regional frequencies. Using an in-silico 3D model of PfRipr predicted via AlphaFold, tertiary structures of all PfRipr sample sequences were predicted via SWISS-MODEL then analyzed by ElliPro to identify linear and discontinuous B-cell epitopes. (Figure 2) Putative epitopes were filtered using binding thresholds, allele coverage, conservation, antigenicity, and allergenicity. (Figure 3) Figure 2: Protocol for B-cell epitope prediction using AlphaFold for template modeling, Swiss-Model for homology-based modeling of sample sequences, and ElliPro for continuous and discontinuous epitope prediction. Figure 3: Protocol for sorting T-cell epitopes using Average Allele Coverage (percentage of alleles in which the protein sequence was identified as an epitope) and B-cell epitopes using ElliPro Score (Protrusion Index representing surface accessibility of the epitope). Results The Burkina Faso PfRipr dataset predicted 1729 unique T-cell and 51 unique B-cell epitopes, and 28 T-cell and 4 B-cell epitopes remained post-filtration. Epitope prediction using the Ugandan dataset predicted 576 unique T-cell and 53 unique B-cell epitopes, and 57 T-cell and 7 B-cell epitopes remained post-filtration. Between the two datasets, there were 19 matching epitopes with 7 MHC I, 9 MHC II, 1 linear, and 2 discontinuous. (Tables 1-2, Figure 4) These epitopes were used to design a multi-epitope-based vaccine construct with a Flagellin adjuvant and peptide linkers to separate functional regions for immune processing. (Figure 5) Figure 4: 3D model of PfRipr with labeled B-cell epitopes. Predicted tertiary structure adapted from AlphaFold database. (https://doi.org/10.1038/s41586-021-03819-2, https://doi.org/10.1093/nar/gkab1061, https://creativecommons.org/licenses/by/4.0/) Figure 5: Schematic representation of hypothetical multi-epitope vaccine construct. FliC is added to the N-terminus to stimulate and amplify immune response. The EAAAK linker is used to separate the bifunctional fusion protein domains. AAY linkers are used to amplify MHC I and II epitopes and prevent junctional epitope binding. GPGPG linkers between B-cell epitopes function to minimize junctional epitopes and retain conformational-dependent immunogenicity. A six amino acid Histidine (6H) chain was added to the C-terminus to aid during lab purification processes. Conclusions Immunoinformatic tools employed in sequence identified candidate PfRipr epitopes used to design a bloodstage vaccine construct against P. falciparum. To validate their predicted immunogenicity, epitopes can be further investigated using in silico protein stabilization and docking simulations, in vitro methods such as HLA stabilization or T-cell activation assays, and in vivo methods using transgenic mouse models. The pipeline of immunoinformatic analyses formulated in this project can be further applied to a larger database of diverse P. falciparum sequences collected from other malaria endemic regions to develop vaccines effective against circulating strains.

Tumor Reactive T Cells Reside in the Bone Marrow and Are Associated with Clinical Response to Hematological Malignancies

T cell receptors (TCRs) play a crucial role in orchestrating cellular immunity in both health and disease, particularly in the context of cancer. Although immunotherapy and stem cell transplantation have shown promising clinical responses in hematological malignancies, these have not been linked to the presence of endogenous tumor reactive T cells. Moreover, the specificities and phenotypes of such tumor recognizing TCRs in these largely low immunogenic entities are unknown. In this study, we addressed this gap by developing an approach to map TCR specificities and transcriptional phenotypes of human bone marrow-resident T cells directly ex vivo, focusing on patients who were treatment-naïve at the time of biopsy. This allowed us to generalize our findings beyond immune checkpoint inhibitor-treated patients, a limitation of previous studies in solid cancers. To discover T cells that specifically recognize cancer cells or viral epitopes, we developed a multiplexed profiling approach using microfluidic reaction chambers, optical cytokine detection and TCR sequencing of patient-derived primary T cells. We were able to confidently map TCR recognition of 25,957 bone marrow-resident T cells from n=20 hematological cancer patients. Based on the acquired functional data and reverse phenotyping of screened TCRs to their original cell state in situ, we built a bone marrow T cell atlas charting >300,000 expression profiles in human bone marrow-resident T cells, their TCR specificities as well as their longitudinal clonal dynamics upon therapeutic perturbations in n=40 patients with newly diagnosed multiple myeloma (MM) and acute myeloid leukemia (AML). Our approach detected tumor-reactive T cell clones in the bone marrow of all screened individuals with varying frequency (1-5% of T cells), which were largely found to display a CD8+ effector-memory or progenitor-exhausted phenotype. In contrast, non-tumor reactive clones were either CD4+ or enriched for viral specificities. We identified a conserved signature of anti-tumor reactivity across patients. This signature included the ITGB1 gene, which encodes CD29 and was uniformly expressed in tumor reactive T cells. Importantly, ITGB1 expression was specifically restricted to tumor reactive, compared to virus-specific or bystander T cells. In two independent mouse tumor models, we found that CD29 protein was significantly upregulated on homed T cells carrying tumor reactive TCRs compared to non-reactive TCRs. A causal link between tumor reactive T cells and clinical response to hematological cancers would be strengthened by finding that these cells specifically recognize autologous cancer epitopes and clonally expand as part of a clinical anti-tumor response. We therefore cloned patient-individualized TCR libraries and performed immunoprecipitation of MHC class I : peptide complexes from cancer cells. We validated TCR binding against ~1-20 epitopes per patient, which were mostly derived from cancer/testis antigens (CTAs) and novel open reading frames (nORFs). Finally, we classified tumor reactive TCRs in three clinical cohorts: Notably, we observed selective expansion of tumor reactive TCRs upon autologous stem cell transplantation, which was associated with deep clinical response in newly diagnosed MM patients. Conversely, the failure of tumor reactive TCRs to clonally expand on-treatment underlined clinical non-response to bispecific BCMAxCD3 antibody therapy (MM) or combination therapy with immune checkpoint inhibitors and Azacytidine (AML). We show here that, unexpectedly, clinically relevant endogenous anti-tumor reactivity lies in a rare subset of bone marrow-resident T cells. The recurrent detection of these immunotherapy-responsive TCRs in patients with MM or AML revealed the selective clonal expansion of tumor-reactive TCRs following autologous stem cell transplantation. We describe a conserved gene signature of tumor reactivity that diverges from lymphocytes infiltrating solid tumors and includes CD29 as an indicator of antigen-specific T cell homing. Our findings facilitate the antigen-agnostic identification of tumor-reactive TCRs with potential for future patient-individualized cell therapies and suggest the relevance of immune responses against alternative epitopes, such as cancer/testis antigens, in hematological malignancies with low mutational loads.

Determining epitope specificity of T-cell receptors with transformers.

T-cell receptors (TCRs) on T cells recognize and bind to epitopes presented by the major histocompatibility complex in case of an infection or cancer. However, the high diversity of TCRs, as well as their unique and complex binding mechanisms underlying epitope recognition, make it difficult to predict the binding between TCRs and epitopes. Here, we present the utility of transformers, a deep learning strategy that incorporates an attention mechanism that learns the informative features, and show that these models pre-trained on a large set of protein sequences outperform current strategies. We compared three pre-trained auto-encoder transformer models (ProtBERT, ProtAlbert, and ProtElectra) and one pre-trained auto-regressive transformer model (ProtXLNet) to predict the binding specificity of TCRs to 25 epitopes from the VDJdb database (human and murine). Two additional modifications were performed to incorporate gene usage of the TCRs in the four transformer models. Of all 12 transformer implementations (four models with three different modifications), a modified version of the ProtXLNet model could predict TCR-epitope pairs with the highest accuracy (weighted F1 score 0.55 simultaneously considering all 25 epitopes). The modification included additional features representing the gene names for the TCRs. We also showed that the basic implementation of transformers outperformed the previously available methods, i.e. TCRGP, TCRdist, and DeepTCR, developed for the same biological problem, especially for the hard-to-classify labels. We show that the proficiency of transformers in attention learning can be made operational in a complex biological setting like TCR binding prediction. Further ingenuity in utilizing the full potential of transformers, either through attention head visualization or introducing additional features, can extend T-cell research avenues. Data and code are available on https://github.com/InduKhatri/tcrformer.

The potential role of infusions of T regulatory cells in inducing and maintaining liver allograft tolerance.

1_nse8422sKaltura.

A transfer-learning approach to predict antigen immunogenicity and T-cell receptor specificity.

Antigen immunogenicity and the specificity of binding of T-cell receptors to antigens are key properties underlying effective immune responses. Here we propose diffRBM, an approach based on transfer learning and Restricted Boltzmann Machines, to build sequence-based predictive models of these properties. DiffRBM is designed to learn the distinctive patterns in amino-acid composition that, on the one hand, underlie the antigen's probability of triggering a response, and on the other hand the T-cell receptor's ability to bind to a given antigen. We show that the patterns learnt by diffRBM allow us to predict putative contact sites of the antigen-receptor complex. We also discriminate immunogenic and non-immunogenic antigens, antigen-specific and generic receptors, reaching performances that compare favorably to existing sequence-based predictors of antigen immunogenicity and T-cell receptor specificity.

BERTrand-peptide:TCR binding prediction using Bidirectional Encoder Representations from Transformers augmented with random TCR pairing.

The advent of T-cell receptor (TCR) sequencing experiments allowed for a significant increase in the amount of peptide:TCR binding data available and a number of machine-learning models appeared in recent years. High-quality prediction models for a fixed epitope sequence are feasible, provided enough known binding TCR sequences are available. However, their performance drops significantly for previously unseen peptides. We prepare the dataset of known peptide:TCR binders and augment it with negative decoys created using healthy donors' T-cell repertoires. We employ deep learning methods commonly applied in Natural Language Processing to train part a peptide:TCR binding model with a degree of cross-peptide generalization (0.69 AUROC). We demonstrate that BERTrand outperforms the published methods when evaluated on peptide sequences not used during model training. The datasets and the code for model training are available at https://github.com/SFGLab/bertrand.

CD1 displays its own negative regulators.

After two decades of thestudy of lipid antigens that activate CD1-restricted T cells, new studies show how autoreactive αβ T-cell receptors (TCRs)can directly recognize the outer surface of CD1 proteins in ways that are lipid-agnostic. Most recently, this lipid agnosticism has turned to negativity, with the discovery of natural CD1 ligands that dominantly negatively block autoreactive αβ TCR binding to CD1a and CD1d. This review highlights the basic differences between positive and negative regulation of cellular systems. We outline strategies to discover lipid inhibitors of CD1-reactive T cells, whose roles in vivo are becoming clear, especially in CD1-mediated skin disease.

Complimentary electrostatics dominate T-cell receptor binding to a psoriasis-associated peptide antigen presented by human leukocyte antigen C∗06:02

Psoriasis is a chronic skin disease characterized by hyperproliferative epidermal lesions infiltrated by autoreactive T cells. Individuals expressing the human leukocyte antigen (HLA) C∗06:02 allele are at highest risk for developing psoriasis. An autoreactive T cell clone (termed Vα3S1/Vβ13S1) isolated from psoriatic plaques is selective for HLA-C∗06:02, presenting a peptide derived from the melanocyte-specific autoantigen ADAMTSL5 (VRSRRCLRL). Here we determine the crystal structure of this psoriatic TCR-HLA-C∗06:02 ADAMTSL5 complex with a stabilized peptide. Docking of the TCR involves an extensive complementary charge network formed between negatively charged TCR residues interleaving with exposed arginine residues from the self-peptide and the HLA-C∗06:02 α1 helix. We probed these interactions through mutagenesis and activation assays. The charged interface spans the polymorphic region of the C1/C2 HLA group. Notably the peptide-binding groove of HLA-C∗06:02 appears exquisitely suited for presenting highly charged Arg-rich epitopes recognized by this acidic psoriatic TCR. Overall, we provide a structural basis for understanding the engagement of melanocyte antigen-presenting cells by a TCR implicated in psoriasis while simultaneously expanding our knowledge of how TCRs engage HLA-C.

Decoupling peptide binding from T cell receptor recognition with engineered chimeric MHC-I molecules

Abstract Human leucocyte antigens (HLAs) display a repertoire of epitopic peptides on the cell surface for T cell recognition. Amino acid polymorphisms in the peptide-binding groove of different HLA allotypes result in distinct antigenic repertoires, providing a basis for adaptive immunity to emerging pathogens at the population level. T cell receptors (TCRs) recognize pHLA antigens through interactions with both peptide and heavy chain residues. In this work, we use structural data from peptide:MHC-I and pMHC:TCR structures to identify residues important for peptide and/or TCR binding, and we outline a fixed-backbone computational design approach for engineering synthetic molecules that combine peptide binding and TCR recognition surfaces from existing HLA allotypes. X-ray crystallography demonstrates that chimeric molecules bridging divergent HLA alleles can bind selected peptide antigens in a specified backbone conformation. Moreover, in vitro tetramer staining and biophysical binding experiments using chimeric pMHC-I molecules presenting established antigens further demonstrate the requirement of TCR recognition on interactions with HLA framework residues, as opposed to interactions with peptide-centric Chimeric Antigen Receptors (CARs). Our results underscore a novel, structure-guided platform for developing synthetic HLA molecules displaying tumor-associated antigens with modified TCR interaction surfaces, which could further serve as platforms to elicit alloreactive T cell responses against self-antigens that are overexpressed in cancer. Finally, we gain a deeper understanding of the evolution and functional divergence of the different HLA allotypes by demonstrating an example of convergent evolution. Supported by grants from NIAID (5R01AI143997), NIGMS (5R35GM125034), and NIDDK (5U01DK112217)