Computationally Designed Proteins Mimic Antibody Immune Evasion in Viral Evolution

Infectious Diseases Microbiology Biochemistry Life Sciences Medicine Immunology Bioengineering
computational design immune evasion viral evolution vaccine evaluation SARS-CoV-2

Academic Background

The continuous mutation of SARS-CoV-2 has posed ongoing challenges to the efficacy of vaccines and antibody therapies. Traditional evaluation methods can only test against existing variants and fail to predict future immune escape mutations. To address this, Noor Youssef and colleagues developed the EVE-Vax (Evolutionary Variant Evaluation for Vaccines) computational platform, which designs poly-mutant spike proteins to simulate potential future antigenic evolution pathways of the virus, enabling proactive assessment of the broad-spectrum effectiveness of vaccines and therapeutics.

The core scientific questions of the study include:
1. Limitations of immune escape prediction: Existing experimental methods (e.g., deep mutational scanning, DMS) can only test single-point mutations or limited combinations and rely on patient serum data.
2. Feasibility of antigen design: Poly-mutant proteins often lose functionality—how can computational design preserve their infectivity and immune escape properties?
3. Timeliness of vaccine evaluation: How to preemptively predict variant escape from current vaccines.

Source of the Paper

  • Research Team: A multidisciplinary collaboration involving Harvard Medical School, the Broad Institute of MIT and Harvard, UMass Chan Medical School, and others. Corresponding authors include Debora S. Marks, Jeremy Luban, et al.
  • Journal: Immunity, June 10, 2025, Volume 58, Pages 1–11.
  • DOI: 10.1016/j.immuni.2025.04.015.

Research Process and Results

a) Research Workflow

1. EVE-Vax Algorithm Design

  • Input Data: Integrated evolutionary sequences, structural information, and antibody-binding site data of SARS-CoV-2 spike proteins.
  • Three Key Constraints:
    • Fitness: Predicted the impact of mutations on viral survival using evolutionary models.
    • Accessibility: Evaluated whether mutation sites were located in antibody-accessible regions.
    • Dissimilarity: Predicted the disruption of mutations to existing antibody binding.
  • Poly-Mutant Combinations: Generated double and multi-mutant combinations from top 1% escape-scoring single mutations, resulting in 83 designed spike variants.

2. Experimental Validation

  • Pseudovirus Infectivity Testing:

    • Samples: 83 designed proteins constructed as pseudoviruses on 5 variant backgrounds (B.1, BA.4⁄5, BA.2.75, etc.).
    • Methods: Lentiviral vector packaging with luciferase reporter gene assays to measure infectivity.
    • Results: 90% (75⁄83) of designed proteins retained infectivity, significantly higher than random mutation libraries (% functionality).

  • Neutralization Antibody Escape Analysis:

    • Serum Samples: 9 panels of human polyclonal sera (including convalescent, vaccinated, and breakthrough infection cases).
    • Neutralization Assays: Measured pseudovirus susceptibility to serum neutralization (geometric mean ID50 titers).
    • Key Findings:
    • The designed BA.2.75-4C variant (with G339D/L452R/Q493R/K529L) showed 5.3-fold higher neutralization resistance than the natural BA.2.75 variant, approaching that of the later-emerging XBB variant (7.2-fold).
    • B.1-background designs (e.g., B.1-4A) exceeded the neutralization resistance of early variants like Alpha and Delta.

3. Vaccine Evaluation Applications

  • Bivalent mRNA Vaccine: Designed XBB variants predicted escape by later variants (e.g., CH.1.1), aligning with real-world observations.
  • Nanoparticle Vaccines: Comparison in non-human primate models showed that mosaic RBD-based nanoparticle vaccines (Mosaic-8B) elicited broader neutralizing antibodies than homotypic vaccines.

b) Key Results and Logical Chain

  1. Computational Design Validity: EVE-Vax successfully predicted escape mutations (e.g., L452R in XBB) observed in natural variants within 12 months.
  2. Balance of Functionality and Escape: Poly-mutant designs maintained infectivity while mimicking natural variant neutralization profiles.
  3. Proactive Evaluation Value: Designed antigens revealed vaccine vulnerabilities in advance, such as the weak protection of bivalent vaccines against XBB.1.5.

Conclusions and Impact

  1. Scientific Value:

    • Proposed the first generalizable framework for simulating viral antigen evolution via computational design, applicable to SARS-CoV-2 and other highly mutable viruses (e.g., influenza, HIV).
    • Demonstrated that deep learning models can outperform experimental methods (e.g., DMS) in predictive power.

  2. Practical Value:

    • Vaccine Development: Accelerates broad-spectrum vaccine design, avoiding a reactive “chasing variants” approach.
    • Therapeutic Evaluation: Provides early warnings for updating monoclonal antibody therapies.

Research Highlights

  • Methodological Innovation: First integration of evolutionary models with antigen design, achieving 90% functional protein designs.
  • Predictive Accuracy: Designed variants predicted immune escape features of natural variants 12 months in advance.
  • Interdisciplinary Integration: Combined computational biology, structural biology, and immunological experimental validation.