A dual-arm peptide vaccine against KRAS, EGFR, PD-L1 and VEGFR2 for non-small cell lung cancer in Vietnamese patients

Technical Summary: CRUISER, a Dual-Arm Peptide Vaccine for NSCLC in Vietnamese Patients

Problem Statement
Therapeutic vaccines for non-small cell lung cancer (NSCLC) are typically designed against global HLA supertypes or European-weighted allele sets. However, the immunogenetics of Southeast Asian populations, particularly the Kinh ethnic group in Vietnam, diverge significantly from these reference sets. In Vietnamese NSCLC patients, activating EGFR mutations occur in ~35% of cases and KRAS mutations in ~22%, often driven by specific codon 12/13 mutations. Current supertype-based constructs fail to exploit the convergence of these recurrent neoantigens with high-frequency, population-specific HLA alleles (e.g., HLA-A*11:01). Furthermore, effective cancer immunotherapy requires addressing both cellular immunity (T-cell activation) and the immunosuppressive tumor microenvironment (PD-L1 checkpoint inhibition and VEGFR2-driven angiogenesis). A single antigen format cannot simultaneously satisfy the structural requirements for T-cell epitopes (linear peptides) and antibody targets (conformational receptor surfaces), necessitating a multi-component design that has not been previously optimized for the Vietnamese HLA landscape.

Methodology
The study presents CRUISER, a computationally designed, dual-arm vaccine restricted to a Vietnamese-specific HLA reference panel. The workflow involved four primary stages:

1. Population-Specific HLA Reference Panel: A reference set was compiled from high-resolution genomic data of the Vietnamese Kinh population, comprising 12 HLA Class I alleles and 13 Class II specificities. This panel excluded global supertypes to ensure relevance to the target demographic.
2. Epitope Screening Cascade:
   • Pool A (T-cell arm): 92 candidate neoepitopes (derived from KRAS and EGFR mutations) were screened against the Vietnamese HLA panel. Candidates underwent a four-layer cascade: antigenicity (VaxiJen), toxicity (ToxinPred/ToxDL), allergenicity (AllerTOP/AlgPred), and cytokine-inducing capacity (IFN-γ/IL-4 prediction) to favor Th1 polarization.
   • Pool B (Antibody arm): Peptides were selected from the ligand-binding faces of PD-L1 and VEGFR2 based on structural criteria (surface exposure and overlap with contact residues) rather than antigenicity scores. These were cyclized via disulfide bonds to stabilize conformation.
3. Construct Assembly and Modeling:
   • CRUISER-A: A 273-residue recombinant protein fusing the XCL1 chemokine (adjuvant) to a cassette of 8 epitopes (5 Class I, 3 Class II) and a PADRE helper sequence, capped with a T4 fibritin trimerization domain.
   • Pool B: Four disulfide-cyclized synthetic peptides.
   • Validation: The construct was modeled using AlphaFold2, refined, and validated for physicochemical properties, solubility, and allergenicity.
4. Molecular Dynamics (MD) and Interaction Energy: To verify the hypothesis that fusing XCL1 to the antigen cassette does not disrupt receptor binding, three independent 100 ns MD simulations were performed on the XCL1:XCR1 complex within the fusion protein. These were benchmarked against a positive control of free XCL1 bound to XCR1 using MM-PBSA interaction energy calculations.
5. Immune Simulation: C-ImmSim was used to predict the immune response profile (antibody isotypes, T-cell populations, cytokine production) following a three-dose schedule.

Key Results

• Population Coverage: The final epitope set achieved a combined population coverage of 73.35% in Vietnam. This represents a significant gain over global (46.78%) and European (41.23%) coverage, driven primarily by Class I alleles (HLA-A11:01, C08:01, C*03:04). The mean number of epitopes recognized per individual was 2.08 in Vietnam versus 0.95 worldwide.
• Epitope Selection: Only 17 of 92 candidates survived the screening cascade. Notably, the most frequent Vietnamese Class II allele, DRB1*12:02, was absent from the final construct because no epitope restricted by it met all safety and potency thresholds.
• XCL1 Fusion Stability: MD simulations confirmed that the fusion of the epitope cassette to XCL1 did not disrupt receptor engagement. The fusion complex maintained a stable interface with XCR1, preserving the seven-transmembrane fold of the receptor and the XCL1 core. The MM-PBSA interaction energy was consistently negative and comparable to the free chemokine control, with the fusion complex maintaining a robust hydrogen bond network.
• Immune Response Prediction: Simulations predicted a Th1-polarized response characterized by sustained cytotoxic T-lymphocyte populations, high IFN-γ production, and a shift from IgM to IgG isotypes.
• Pool B Design: Four cyclic peptides were identified. Two targeted PD-L1 and two VEGFR2. One candidate (EP076) required an explicit risk annotation due to discordant allergenicity predictions but was retained for its unique position on the VEGF-binding face.

Key Contributions

1. Population-Restricted Design: The study quantifies the advantage of tailoring vaccine epitopes to a specific population's HLA profile, demonstrating that a "one-size-fits-all" supertype approach would forfeit approximately one-third of potential coverage in the Vietnamese population.
2. Validated Adjuvant Fusion: Unlike previous studies that rely solely on docking, this work provides computational evidence (via MD and MM-PBSA benchmarking) that the XCL1 chemokine retains its functional engagement with XCR1 even when constrained within a large fusion protein.
3. Dual-Arm Architecture: The design successfully separates cellular and humoral immunity into two distinct arms (Pool A and Pool B) to resolve the structural incompatibility between linear T-cell epitopes and conformational antibody targets.
4. Structure-First Selection: For the antibody arm, the study prioritizes structural location (ligand-binding faces) over antigenicity scores, successfully eliminating high-scoring candidates that mapped to non-functional regions of the receptor.

Significance and Claims
The authors claim that CRUISER represents a significant methodological advancement in immunoinformatics by moving beyond global supertypes to a population-specific strategy that yields quantifiable gains in predicted coverage. The work establishes that:

• Population restriction is a determinant of potency, not merely a procedural refinement.
• Chemokine-based dendritic cell targeting can be structurally validated through rigorous MD simulations rather than assumed.
• Separating T-cell and antibody arms allows for the simultaneous targeting of neoantigens and wild-type checkpoint/angiogenic proteins without generating junctional neo-epitopes or conflicting structural demands.

The paper explicitly maintains a modest tone, stating that all findings are computational predictions requiring experimental validation. The authors note that the Class II coverage (26.76%) is limited by the stringent screening criteria and the absence of DRB1*12:02 binders, a deficit partially mitigated by the inclusion of the universal PADRE helper epitope. They emphasize that the MM-PBSA values are relative and not absolute binding constants, and that the final construct must be tested for immunogenicity in HLA-transgenic mice and Vietnamese peripheral blood mononuclear cells before clinical application.