Quantitative modeling of the effect of antigen dosage on B-cell affinity distributions in maturating germinal centers

Marco Molari, Klaus Eyer, Jean Baudry, Simona Cocco, Rémi Monasson

TL;DR

This paper introduces a quantitative model for B-cell affinity maturation in germinal centers, revealing that optimal antibody affinity is achieved at intermediate antigen dosages. By combining this model with experimental data from immunized mice, the authors provide new insights into the complex mechanisms of immune response, particularly the role of selection permissiveness.

What Problem Does This Paper Address?

Generating effective antibodies against pathogens is a cornerstone of immunology and vaccination. This process, known as affinity maturation (AM), occurs in specialized structures called germinal centers (GCs). While it's understood that AM involves B-cell mutation and selection, the precise influence of factors like antigen (Ag) dosage on the efficiency and outcome of this process remains complex and not fully understood. Previous studies have suggested a non-monotonic relationship between Ag dosage and antibody affinity, implying an "optimal" dose rather than a simple "more is better" scenario. However, the underlying mechanisms and how to quantitatively predict this optimal response have been challenging to elucidate. This paper aims to address this by developing a computational model that can quantitatively describe and predict the B-cell affinity distribution during AM, specifically investigating how Ag dosage and other factors influence this process.

Methods

The authors developed a stochastic model for affinity maturation that simulates the evolution of B-cell binding energies within germinal centers. This model incorporates key biological processes:

  • Cell duplication and Somatic Hypermutation (SHM): B-cells divide and acquire random mutations, some of which affect affinity.
  • Ag-binding selection: B-cells must bind to Ag presented on follicular dendritic cells (FDCs) to survive.
  • T-cell help selection: B-cells compete for survival signals from T follicular helper cells, which are influenced by Ag presentation.
  • Differentiation: Surviving B-cells differentiate into plasma cells (PCs) or memory cells (MCs).

A key innovation is the development of a deterministic version of this model, which accurately approximates the stochastic behavior in the limit of large cell populations. This deterministic model allows for detailed mathematical analysis, particularly using an eigenvalue equation to describe the asymptotic behavior of the affinity distribution as a "traveling wave."

Furthermore, the authors employed a maximum-likelihood-based inference method using a Parallel Tempering algorithm to fit the model's parameters to experimental data. This allowed them to infer parameters that are difficult to measure directly, such as the "permissiveness" of selection.

Datasets

The study utilized experimental data from mice immunized with Tetanus Toxoid (TT). Specifically, they measured the affinities of single IgG-secreting cells (IgG-SCs) extracted from mouse spleens following various immunization protocols. These protocols involved different Ag dosages and delays between injections. The authors also generated artificial datasets from their model to validate their inference procedure.

Key Insights and Ideas

  • Non-monotonicity of Ag Dosage: The model and experiments confirm that optimal antibody affinity is not achieved at the highest Ag doses. Instead, intermediate dosages strike a balance between sufficient selection pressure and sustained GC activity, leading to the best overall affinity.
  • Permissiveness in GC Selection: The inferred parameters suggest that germinal center selection is "permissive." This means that even B-cells with lower affinity have a non-zero chance of surviving and contributing to the antibody repertoire, which might be crucial for dealing with diverse or evolving pathogens.
  • Quantitative Link Between Model and Experiment: The ability of a single set of model parameters to accurately reproduce experimental affinity distributions across various immunization schemes highlights the power of combining quantitative modeling with statistical inference for understanding complex biological processes.
  • Stochasticity vs. Determinism: The deterministic model provides a powerful analytical tool that accurately captures the average behavior of the more complex stochastic model, simplifying analysis while retaining key biological insights.

Key Results

  • The model accurately reproduces experimental affinity distributions of IgG-SCs from immunized mice under varying Ag dosages and injection delays.
  • The average population affinity shows a non-monotonic dependence on Ag dosage, with an optimal intermediate dose identified.
  • Inferred parameters suggest that GC selection is permissive (parameter 'a' = 0.12) and stochastic (parameter 'b' = 0.66), meaning low-affinity cells have a chance to survive, and high-affinity cells can still be eliminated.
  • The model's deterministic version effectively predicts the asymptotic behavior of the affinity distribution as a traveling wave, with growth rate and maturation speed dependent on Ag concentration.
  • The study provides a framework for inferring biologically relevant parameters (like selection permissiveness) that are not easily measured experimentally.

Limitations and Future Work

  • Model Simplifications: The authors acknowledge that the model simplifies certain aspects, such as assuming cell duplication is independent of affinity and not explicitly modeling antibody feedback. They also note that the model doesn't account for the initiation of GC reactions at very low Ag dosages.
  • Experimental Sensitivity: Measurements are limited by experimental sensitivity, meaning very low or very high affinities cannot be precisely quantified.
  • Future Directions: The authors suggest extending the model to predict outcomes for more complex immunization protocols (e.g., continuous delivery), applying the inference framework to new datasets, and investigating the role of selection permissiveness in more detail. They also propose exploring the impact of affinity-affecting mutations more deeply, potentially by incorporating "shape space" representations.

Summary updated

Abstract

Affinity maturation is a complex dynamical process allowing the immune system to generate antibodies capable of recognizing antigens. We introduce a model for the evolution of the distribution of affinities across the antibody population in germinal centers. The model is amenable to detailed mathematical analysis and gives insight on the mechanisms through which antigen availability controls the rate of maturation and the expansion of the antibody population. It is also capable, upon maximum-likelihood inference of the parameters, to reproduce accurately the distributions of affinities of IgG-secreting cells we measure in mice immunized against Tetanus Toxoid under largely varying conditions (antigen dosage, delay between injections). Both model and experiments show that the average population affinity depends non-monotonically on the antigen dosage. We show that combining quantitative modeling and statistical inference is a concrete way to investigate biological processes underlying affinity maturation (such as selection permissiveness), hardly accessible through measurements.
BibTeX
@inproceedings{molari2020quantitative,
  author = {Marco Molari and Klaus Eyer and Jean Baudry and Simona Cocco and Rémi Monasson},
  title = {{Quantitative modeling of the effect of antigen dosage on B-cell affinity distributions in maturating germinal centers}},
  booktitle = {eLife},
  year = {2020},
  doi = {10.7554/elife.55678},
}

Loading related content...

Figure