CD4+ T-Cell Epitope Prediction by Combined Analysis of Antigen Conformational Flexibility and Peptide-MHCII Binding Affinity

Tysheena P. Charles, Daniel L. Moss, Pawan Bhat, Peyton W. Moore, Nicholas Kummer, Avik Bhattacharya, Samuel J. Landry, Ramgopal R. MettuBiochemistry 2022 · DOI

TL;DR

This paper introduces a novel computational approach, Antigen Processing Likelihood (APL), that combines conformational flexibility and peptide-MHCII binding affinity to predict CD4+ T-cell epitopes. This combined method significantly improves prediction accuracy compared to using either factor alone, offering a more robust tool for vaccine design and understanding immune responses.

What Problem Does This Paper Address?

Predicting CD4+ T-cell epitopes is crucial for developing effective vaccines and understanding immune responses. CD4+ T cells recognize peptide fragments presented by MHC class II (MHCII) molecules. However, predicting which peptides will be presented and elicit a response is challenging due to several factors:

  • Antigen Processing Complexity: Antigens are broken down by proteases within cellular compartments (endolysosomes) before binding to MHCII. This process is influenced by antigen folding, protease activity, and the cellular environment.
  • MHCII Binding Variability: MHCII molecules can bind peptides in multiple "registers" (different starting points), and their binding specificity is influenced by many factors, making simple binding affinity predictions insufficient.
  • Epitope Dominance: Not all potential epitopes are equally immunogenic. "Dominant" epitopes are those that elicit the strongest T-cell response, and their emergence is a complex interplay of antigen processing and presentation.

Existing prediction methods primarily focus on peptide-MHCII binding affinity, but these often fail to accurately predict actual T-cell responses. This paper aims to bridge this gap by incorporating the influence of antigen structure and its impact on processing.

Methods

The core of this work is the development and application of an Antigen Processing Likelihood (APL) algorithm. Here's a breakdown of the key methodological aspects:

  • APL Algorithm: This algorithm integrates multiple data sources to predict the likelihood of a region within an antigen being processed into a T-cell epitope. The primary inputs are:
  • Conformational Stability Data: Derived from X-ray crystal structures, this includes measures like sequence entropy, crystallographic b-factors, COREX scores, and solvent-accessible surface area. These parameters help identify regions that are rigid or flexible.
  • Protease Sensitivity/Resistance: This is incorporated by analyzing fragmentation patterns from limited proteolysis experiments. Regions that are cleaved are considered more accessible to proteases.
  • Linear Antibody Epitope (F-LE) Reactivity: This measures antibody binding to linear peptide fragments, providing an indirect indicator of solvent accessibility and potential epitope regions.
  • Combined Prediction: The APL score is then combined with peptide-MHCII binding affinity predictions (obtained from the NetMHCII 2.3 server) to generate a final prediction score.
  • Parameter Optimization: The algorithm's parameters (weights for input data sources, magnification factors, loop/flank sizes) were optimized using a leave-out cross-validation approach to maximize the positive predictive value (PPV).
  • Experimental Validation: The predictive models were validated using experimental data from mice immunized with a plague vaccine candidate (F1-V). This involved mapping CD4+ T-cell epitopes through splenocyte proliferation assays and mapping linear B-cell epitopes.

Datasets

The study utilized several datasets for training and evaluation:

  • Plague Vaccine Candidate F1-V: This recombinant protein, composed of Caf1 and LcrV subunits, served as a primary model antigen for detailed experimental mapping of CD4+ T-cell and linear B-cell epitopes.
  • Benchmark Antigen Sets: The authors evaluated their APL algorithm on two benchmark sets:
  • C57BL/6 Mouse Epitope Data: This dataset comprised 13 diverse antigens for which CD4+ T-cell epitopes had been systematically mapped in C57BL/6 mice.
  • Human Epitope Data: A dataset of 12 antigens with mapped CD4+ T-cell epitopes in humans was also used.

Key Insights and Ideas

  • Structure Dictates Processing: The study strongly suggests that the three-dimensional structure of an antigen significantly influences its processing and, consequently, CD4+ T-cell epitope dominance. Conformationally stable regions tend to resist proteolysis, while flexible or disordered regions are more accessible to proteases.
  • APL Captures Processing Likelihood: The APL algorithm, by integrating conformational flexibility and protease sensitivity, effectively models the likelihood of a region being processed into an epitope. This is a crucial step beyond simply predicting peptide binding.
  • Synergy of APL and MHCII Binding: Combining APL with peptide-MHCII binding affinity predictions yields significantly better epitope prediction accuracy than either method alone. This highlights that both the availability of a peptide (processing) and its ability to bind MHCII are critical.
  • Biochemical Data as a Proxy for Structure: Linear antibody epitope reactivity and protease sensitivity data can serve as valuable proxies for conformational flexibility, especially when high-resolution crystal structures are unavailable. This broadens the applicability of the APL approach.
  • Dominant Epitopes Near Stable Regions: Dominant CD4+ T-cell epitopes were consistently found at the edges of conformationally stable antigen segments, suggesting a mechanism where proteolysis at flexible boundaries liberates stable peptide fragments for MHCII presentation.

Key Results

  • Improved Prediction Accuracy: The combined APL and MHCII binding prediction method achieved an accuracy of 36% for a sample of 13 diverse antigens in C57BL/6 mice.
  • APL Alone is Powerful: For human immunity, where MHCII allele diversity is high, APL alone achieved 42% prediction accuracy when using a stringent scoring threshold.
  • Experimental Validation with F1-V: The study successfully mapped CD4+ T-cell epitopes in F1-V immunized mice and found that dominant epitopes were associated with stable antigen segments, aligning with the APL model.
  • Biochemical Data Sufficiency: Using only F-LE and proteolysis data (without crystallographic information), APL achieved significant accuracy in predicting CD4+ T-cell epitopes in CBA mice.

Limitations and Future Work

  • Requirement for Structural Data: A significant limitation of the APL algorithm is its reliance on high-resolution three-dimensional structures (typically X-ray crystallography). While homology models can be used, this can be a barrier for some antigens.
  • Optimization of Parameters: The authors note that while their optimization approach is robust, further refinement of parameter sets for specific antigens or contexts could yield even better results.
  • Further Investigation of Biochemical Data: The study suggests that further efforts to improve the prediction of proteolytic sensitivity from biochemical data could enhance epitope prediction, especially when structural data is scarce.
  • Broader Applicability: While tested on a limited set of antigens, the authors propose that APL could be a valuable tool for predicting CD4+ T-cell epitopes in human immune responses, particularly in complex scenarios with diverse MHCII alleles. Future work could focus on expanding the training datasets and refining the algorithm for broader applications.

Summary updated

Abstract

Antigen processing in the class II MHC pathway depends on conventional proteolytic enzymes, potentially acting on antigens in native-like conformational states. CD4+ epitope dominance arises from a competition among antigen folding, proteolysis, and MHCII binding. Protease-sensitive sites, linear antibody epitopes, and CD4+ T-cell epitopes were mapped in plague vaccine candidate F1-V to evaluate the various contributions to CD4+ epitope dominance. Using X-ray crystal structures, antigen processing likelihood (APL) predicts CD4+ epitopes with significant accuracy for F1-V without considering peptide-MHCII binding affinity. We also show that APL achieves excellent performance over two benchmark antigen sets. The profiles of conformational flexibility derived from the X-ray crystal structures of the F1-V proteins, Caf1 and LcrV, were similar to the biochemical profiles of linear antibody epitope reactivity and protease sensitivity, suggesting that the role of structure in proteolysis was captured by the analysis of the crystal structures. The patterns of CD4+ T-cell epitope dominance in C57BL/6, CBA, and BALB/c mice were compared to epitope predictions based on APL, MHCII binding, or both. For a sample of 13 diverse antigens, the accuracy of epitope prediction by the combination of APL and I-A<sup>b</sup>-MHCII-peptide affinity reached 36%. When MHCII allele specificity was also diverse, such as in human immunity, prediction of dominant epitopes by APL alone reached 42% when using a stringent scoring threshold. Because dominant CD4+ epitopes tend to occur in conformationally stable antigen domains, crystal structures typically are available for analysis by APL, and thus, the requirement for a crystal structure is not a severe limitation.
BibTeX
@inproceedings{charles2022cd,
  author = {Tysheena P. Charles and Daniel L. Moss and Pawan Bhat and Peyton W. Moore and Nicholas Kummer and Avik Bhattacharya and Samuel J. Landry and Ramgopal R. Mettu},
  title = {{CD4+ T-Cell Epitope Prediction by Combined Analysis of Antigen Conformational Flexibility and Peptide-MHCII Binding Affinity}},
  booktitle = {Biochemistry},
  year = {2022},
  doi = {10.1021/acs.biochem.2c00237},
}
Figure