Cancer Vaccines and Personalized Neoantigen Therapy

Cancer vaccines bridge vaccinology and precision oncology through two distinct clinical paradigms: prophylactic cancer vaccines, which immunize healthy individuals against oncogenic viruses to prevent malignant transformation altogether; and therapeutic cancer vaccines, which prime and expand tumor-specific CD8⁺ cytotoxic T lymphocytes (CTLs) and CD4⁺ helper T cells in patients with established malignancies to eradicate micrometastatic residual disease and prevent tumor recurrence. While prophylactic vaccines against Hepatitis B Virus (HBV) and Human Papillomavirus (HPV) have reduced virus-associated liver and cervical cancer incidence by over 80% to 90% in vaccinated cohorts, early therapeutic vaccines targeting unmutated self-proteins struggled to overcome central thymic tolerance and tumor microenvironment immunosuppression.

Over the past decade, the convergence of clinical whole-exome sequencing, deep-learning Major Histocompatibility Complex (MHC) binding prediction, rapid mRNA-LNP synthesis, and immune checkpoint inhibitors enabled individualized neoantigen specific therapy (iNeST). By encoding 20 to 34 patient-specific somatic neoantigens—mutated peptides present exclusively in tumor cells and foreign to the host thymus—personalized cancer vaccines generate high-affinity, polyfunctional T-cell responses demonstrated in randomized clinical trials to halve recurrence risk in resected melanoma and induce long-term disease-free survival in pancreatic cancer.

HBV and HPV Oncogenesis and Population Cancer Prevention

Approximately 12% to 15% of all human cancers worldwide are attributable to chronic infections with oncogenic viruses (alongside Helicobacter pylori and Schistosoma haematobium). Prophylactic recombinant virus-like particle (VLP) vaccines prevent two major viral malignancies by eliciting high-titer neutralizing antibodies that block primary viral infection:

• Hepatitis B Virus (HBV) and Hepatocellular Carcinoma (HCC): Chronic HBV infection drives persistent necroinflammatory liver injury, oxidative DNA damage, regenerative cirrhosis, and insertional mutagenesis of viral DNA into host chromosomal fragile sites (such as the TERT promoter) alongside expression of the viral transactivator protein HBx. Introduced as a universal infant immunization program in Taiwan in 1984, recombinant HBsAg VLP vaccination reduced chronic HBV carriage in children from ~10% to <0.5% and decreased the incidence of pediatric and young-adult hepatocellular carcinoma by more than 80%, establishing the first clinical proof that a vaccine could prevent a human solid tumor.
• Human Papillomavirus (HPV) and Anogenital/Oropharyngeal Carcinomas: High-risk mucosal alpha-papillomaviruses—principally HPV-16 and HPV-18 (causing ~70% of cervical cancers and >85% of HPV-positive oropharyngeal cancers), plus HPV-31, 33, 45, 52, and 58 (bringing total coverage in the 9-valent Gardasil 9 vaccine to ~90% of cervical cancers)—drive malignancy through two viral oncoproteins expressed upon integration into basal epithelial cells. The viral E6 oncoprotein recruits the host ubiquitin ligase E6AP (UBE3A) to polyubiquitinate and degrade the tumor suppressor p53 (TP53) via the 26S proteasome, while activating telomerase (hTERT). Simultaneously, the viral E7 oncoprotein binds and degrades the retinoblastoma protein (pRb, RB1), releasing E2F transcription factors to force unscheduled entry into S phase. Large-scale national registry linkage studies in Sweden, Scotland, and England demonstrated that routine adolescent HPV L1 VLP vaccination at ages 12 to 13 achieves an 87% to >90% reduction in invasive cervical carcinoma.

Why Shared Tumor-Associated Antigen (TAA) Vaccines Struggled

For three decades, first-generation therapeutic cancer vaccines targeted Shared Tumor-Associated Antigens (TAAs)—non-mutated proteins encoded in the normal human germline genome that are either overexpressed in tumors (such as HER2/neu, MUC1, WT1, and hTERT), restricted to a specific tissue lineage (differentiation antigens such as gp100, Melan-A/MART-1, and tyrosinase in melanoma, or prostatic acid phosphatase [PAP] in prostate cancer), or normally silenced in adult somatic tissues except immune-privileged germ cells and placenta (Cancer-Testis Antigens [CTAs] such as MAGE-A3 and NY-ESO-1).

Although the autologous dendritic-cell/leukapheresis vaccine sipuleucel-T (Provenge)—loaded ex vivo with a fusion protein of PAP and GM-CSF—earned FDA approval in 2010 after extending median overall survival by 4.1 months (25.8 vs. 21.7 months) in metastatic castration-resistant prostate cancer (and intravesical live-attenuated Bacillus Calmette–Guérin [BCG], introduced by Alvaro Morales in 1976, remains standard innate/adaptive immunotherapy for non-muscle-invasive bladder cancer), large Phase III trials of shared TAA peptide and protein vaccines (such as the MAGRIT trial of MAGE-A3 in non-small-cell lung cancer) failed to improve disease-free survival. Two fundamental immunological barriers limited TAA vaccines:

1. Central Thymic Tolerance and Clonal Deletion: Because TAAs are unmutated self-proteins, medullary thymic epithelial cells (mTECs) ectopically express them in the thymus under the control of the transcriptional regulator Autoimmune Regulator (AIRE) and FEZF2. During T-cell development, immature thymocytes bearing T-cell receptors (TCRs) with high affinity for self-peptide–MHC complexes undergo negative selection (clonal deletion) via BIM-mediated apoptosis or differentiate into FOXP3⁺ regulatory T cells (T_reg). Consequently, the surviving peripheral T-cell repertoire capable of recognizing shared TAAs consists of low-affinity clones with modest cytolytic potency.
2. Peripheral Tolerance and Tumor Microenvironment (TME) Immunosuppression: Even when low-affinity TAA-specific T cells are expanded in peripheral blood, their infiltration and effector function within solid tumors are blunted by tumor-expressed PD-L1 (CD274) ligating T-cell PD-1, immunosuppressive cytokines (TGF-β, IL-10, VEGF), metabolic depletion of tryptophan via indoleamine 2,3-dioxygenase (IDO1), myeloid-derived suppressor cells (MDSCs), and downregulation of tumor MHC Class I molecules (β₂-microglobulin loss).

Somatic Neoantigens and Escape from Central Tolerance

Pioneered by Uğur Şahin and Özlem Türeci, Catherine J. Wu, Robert D. Schreiber, and Steven A. Rosenberg, Tumor-Specific Neoantigens (TSAs) are mutated peptides that arise de novo in malignant cells from somatic non-synonymous single-nucleotide variants (SNVs), insertion–deletion frameshift mutations (which generate novel downstream amino acid sequences with high immunogenicity), or chromosomal gene fusions. Because somatic mutations occur selectively in the tumor genome and are absent from normal host tissues and the embryonic thymus, neoantigen-specific T cells bypass central thymic deletion and retain high-affinity TCRs comparable to T cells directed against foreign viruses. Furthermore, because normal cells do not express the mutated peptide, neoantigen vaccination avoids on-target autoimmune destruction of healthy organs.

Sequencing, Deep-Learning Neoepitope Prioritization, and Manufacturing

Because over 95% of somatic mutations in a given patient's tumor are unique ("private" passenger mutations) rather than shared across patients, neoantigen vaccines must be custom-designed and manufactured on a single-patient basis (n = 1 batch manufacturing) within a rapid 4- to 6-week vein-to-vein turnaround time:

1. Tumor–Normal Genomic and Transcriptomic Profiling: DNA is extracted from a patient's surgically resected tumor (or core biopsy) alongside matched normal peripheral blood mononuclear cells (PBMCs) and subjected to whole-exome sequencing (WES) to call somatic variants and calculate clonal variant allele fractions (VAF) (prioritizing truncal clonal mutations shared by all tumor cells over subclonal mutations). Simultaneously, tumor RNA-Seq verifies that the mutated allele is actively transcribed, and high-resolution 4-digit HLA Class I (HLA-A, -B, -C) and Class II (HLA-DR, -DP, -DQ) typing defines the patient's antigen-presenting repertoire.
2. Deep-Learning Neoepitope Prediction: Bioinformatic pipelines flanked by mass-spectrometry immunopeptidomics train neural networks—such as NetMHCpan-4.1, MHCflurry, and MixMHCpred—to score every candidate 8–11 amino acid peptide (for MHC Class I) and 15–25 amino acid peptide (for MHC Class II) spanning each somatic mutation. Algorithms integrate proteasomal C-terminal cleavage probability, TAP1/TAP2 endoplasmic reticulum transport efficiency, predicted peptide–MHC binding affinity (IC₅₀ < 50–500 nM or eluted ligand %Rank < 0.5%), peptide–MHC complex half-life stability, and dissimilarity to the wild-type self-proteome (agretopicity, particularly when the mutation creates a new primary anchor residue at position 2 or the C-terminus).
3. Concatenamer mRNA or Long-Peptide Synthesis: The top 20 to 34 ranked patient-specific neoepitopes (each encoded as a ~25–27-mer polypeptide centered on the somatic mutation, connected by non-immunogenic glycine–serine linkers into one or two poly-epitope mRNA concatenamers and flanked by an N-terminal MHC Class I signal peptide and a C-terminal MHC Class I trafficking domain [MITD] that routes translated polypeptides into both MHC Class I and Class II processing compartments) are synthesized in automated, small-footprint cGMP microfluidic reactors. Including CD4⁺ MHC Class II neoepitopes alongside CD8⁺ epitopes is required because CD4⁺ helper T cells license conventional type 1 dendritic cells (cDC1, via CD40L–CD40) and secrete IL-2 and IL-21 to sustain cytotoxic CD8⁺ T-cell clonal expansion and prevent exhaustion.

Clinical Breakthroughs: KEYNOTE-942 in Melanoma and BNT122 in Pancreatic Cancer

When administered in the adjuvant setting (after surgical resection of the primary tumor, when residual tumor burden and systemic immunosuppression are minimal) in combination with PD-1/PD-L1 checkpoint blockade, personalized neoantigen vaccines have achieved landmark clinical outcomes:

• mRNA-4157 (V940) + Pembrolizumab in High-Risk Melanoma (KEYNOTE-942): Developed by Moderna and Merck, mRNA-4157 is an LNP-formulated mRNA vaccine encoding up to 34 patient-specific neoantigens. In the randomized Phase IIb KEYNOTE-942 trial in patients with resected high-risk Stage IIIB–IV cutaneous melanoma, combining adjuvant mRNA-4157 (up to 9 intramuscular doses) with the anti-PD-1 monoclonal antibody pembrolizumab reduced the risk of recurrence or death by 44% to 49% (Hazard Ratio HR = 0.51–0.56 at 2- to 3-year follow-up) and reduced the risk of distant metastasis or death by 62% to 65% (HR = 0.35–0.38) compared to standard-of-care pembrolizumab monotherapy, regardless of tumor mutational burden or PD-L1 status.
• Autogene Cevumeran (BNT122) in Pancreatic Ductal Adenocarcinoma (PDAC): Developed by BioNTech and Genentech and evaluated by Vinod Balachandran and Luis Rojas at Memorial Sloan Kettering Cancer Center (Nature 2023), autogene cevumeran encodes up to 20 patient-specific neoantigens packaged as an intravenous uridine mRNA–lipoplex (RNA-LPX) nanoparticle with a net negative charge that selectively targets splenic dendritic cells and uses unmodified uridine to activate endosomal TLR7/8 and systemic IFN-α. Despite PDAC being a notoriously low-mutation, immunologically "cold" malignancy with an ~88% 5-year mortality rate, sequential administration of surgery, a single dose of anti-PD-L1 atezolizumab, 8 priming doses of intravenous autogene cevumeran, and mFOLFIRINOX chemotherapy induced high-magnitude de novo polyfunctional neoantigen-specific CD8⁺ T cells (comprising up to 10% of all circulating CD8⁺ T cells) in 50% of patients (8 of 16). At a median follow-up of 1.5 to 3+ years, vaccine responders had a median recurrence-free survival that was not reached (with functional vaccine-induced T-cell clones persisting for multiple years), whereas non-responders relapsed at a median of 13.4 months.

Why is it essential for personalized cancer vaccines to activate CD4+ helper T cells in addition to CD8+ killer T cells?

Although CD8⁺ cytotoxic T lymphocytes directly kill MHC Class I-expressing tumor cells via perforin and granzyme B, CD8⁺ T cells primed without CD4⁺ helper T-cell support become dysfunctional and undergo rapid exhaustion. Neoantigen-specific CD4⁺ T cells bind MHC Class II on cDC1 dendritic cells and deliver CD40L costimulation ("dendritic cell licensing") alongside IL-2 and IL-21, which upregulates CD70/CD80/CD86 on the dendritic cell, remodels the tumor vasculature via IFN-γ, and enables durable CD8⁺ memory differentiation.

Why are therapeutic cancer vaccines most effective in the surgical adjuvant setting rather than in bulky metastatic disease?

Large, bulky metastatic tumors establish a deeply immunosuppressive microenvironment characterized by physical stromal barriers, hypoxia, lactic acidosis, high densities of MDSCs and T_reg cells, and genetic loss of HLA Class I antigen presentation. Administering a neoantigen vaccine in the adjuvant setting—immediately after surgical removal of the visible tumor—allows vaccine-primed T cells to expand in an intact immune system and eliminate microscopic residual cancer cells at a high effector-to-target ratio.

Why does autogene cevumeran (BNT122) use unmodified uridine RNA-lipoplexes intravenously, whereas prophylactic COVID-19 vaccines use m1Ψ-modified mRNA intramuscularly?

Prophylactic viral vaccines aim to maximize antigen protein expression for B-cell antibody production while minimizing systemic fever and interferons, so they use N¹-methylpseudouridine (m¹Ψ) intramuscularly. In contrast, eradicating cancer requires massive systemic expansion of cytotoxic CD8⁺ T cells, which strictly depends on high Type I Interferon (IFN-α) signaling during dendritic-cell priming. Intravenous RNA-LPX nanoparticles carrying unmodified uridine selectively home to the spleen and engage TLR7 in plasmacytoid and conventional dendritic cells, triggering an IFN-α pulse that drives potent T-cell expansion.

1. Sahin, U., Derhovanessian, E., Miller, M., Kloke, B. P., Simon, P., Löwer, M., et al. (2017). "Personalized RNA mutanome vaccines mobilize poly-specific therapeutic immunity against cancer." Nature, 547(7662), 222–226. https://doi.org/10.1038/nature23003
2. Ott, P. A., Hu, Z., Keskin, D. B., Shukla, S. A., Sun, J., Bozym, D. J., et al. (2017). "An immunogenic personal neoantigen vaccine for patients with melanoma." Nature, 547(7662), 217–221. https://doi.org/10.1038/nature22991
3. Rojas, L. A., Sethna, Z., Soares, K. C., Olcese, C., Pang, N., Patterson, E., et al. (2023). "Personalized RNA neoantigen vaccines stimulate T cells in pancreatic cancer." Nature, 618(7963), 144–150. https://doi.org/10.1038/s41586-023-06063-y
4. Weber, J. S., Carlino, M. S., Khattak, A., Meniawy, T., Ansstas, G., Taylor, M. H., et al. (2024). "Individualised neoantigen therapy mRNA-4157 (V940) plus pembrolizumab versus pembrolizumab monotherapy in resected melanoma (KEYNOTE-942): a randomised, phase 2b study." The Lancet, 403(10427), 632–644. https://doi.org/10.1016/S0140-6736(23)02268-702268-7)