Peptide Vaccines Explained: The Science and the Clinical Record

Immunology

Peptide vaccines use short synthetic fragments of a protein — sometimes just eight to ten amino acids — to teach T cells to recognise a target. The idea has been in clinical testing for over three decades, it has produced genuinely striking phase I immunology, and it has not yet produced an approved product.

The appeal is obvious. If the immune system ultimately recognises short peptide fragments displayed on MHC molecules, why deliver a whole pathogen or a whole protein? Synthesise the fragment that matters, inject it, and skip everything else. That reasoning is sound as far as it goes. The gap between it and a working vaccine is where the interesting science lives.

The problem with injecting exactly the right fragment

A minimal MHC class I epitope needs no processing. That sounds like an advantage, and it is the central difficulty. Because it needs no processing, it can load directly onto MHC class I on essentially any nucleated cell — including cells that carry no costimulatory molecules and are broadcasting no danger signal. The T cell sees its antigen in a context that says “normal tissue,” and behaves accordingly.

A 2007 study made this concrete. Mice given the minimal ovalbumin epitope OVA257–264 in incomplete Freund's adjuvant transiently activated CD8 effector T cells that then failed to undergo secondary expansion or to kill target cells, because peptide leaking out of the oily depot was being presented systemically without any danger signal. Two fixes restored function: adding a class II helper peptide, or simply extending the same CTL epitope to a 30-mer. The long peptide was predominantly presented in the inflamed draining lymph node instead.

Synthetic long peptides work because they cannot skip processing. A 30-amino-acid peptide has to be internalised and cut up before it can be displayed, which means only professional antigen-presenting cells can present it, and it routes through both class I and class II pathways — so CD8 killing and CD4 help arrive together.

The depot can be the enemy

Adjuvant choice is not a detail. A 2013 Nature Medicine study vaccinated mice with a melanoma peptide in incomplete Freund's adjuvant and found that the primed, tumour-specific CD8 T cells accumulated not in the tumour but at the persisting, antigen-rich vaccination site, where they became dysfunctional and were deleted. A formulation that did not persist long-term at the injection site produced a better antitumour response. A vaccine can succeed at generating T cells and still fail by parking them in the wrong place.

Modern regimens therefore lean on immune-activating adjuvants rather than depots alone. Poly-ICLC — a synthetic complex of polyinosinic–polycytidylic acid, poly-L-lysine and carboxymethylcellulose — engages endosomal TLR3 and cytosolic MDA5, upregulating CD80 and CD86 on dendritic cells and driving IL-12 and type I interferon. Montanide ISA 51 remains in use as a water-in-oil emulsion providing depot release and local APC recruitment. TLR ligands can also be covalently conjugated onto the long peptide itself, putting the antigen and its danger signal on one molecule.

Neoantigen vaccines, and the number nobody advertises

The most active area is personalised cancer vaccination. The workflow is elegant on paper: sequence the tumour and matched normal exome, use RNA-seq to confirm the mutation is expressed, predict which mutated peptides will bind the patient's HLA alleles, synthesise the winners, and vaccinate.

The 2017 NeoVax melanoma trial did exactly this. Of 10 patients enrolled with resected high-risk melanoma, 8 had vaccines designed (13–20 immunising peptides each, 15–30 amino acids, in four pools with poly-ICLC) and 6 initiated vaccination, all completing five priming and two booster doses. Vaccine-induced polyfunctional CD4 and CD8 T cells targeted 58 of 97 (60%) and 15 of 97 (16%) of the unique neoantigens respectively. Adverse events were mild. It is a six-patient phase I with no control arm and no efficacy endpoint — and it is one of the most influential papers in the field.

The honest constraint came three years later. The TESLA consortium had 25 independent teams predict immunogenic epitopes from shared tumour sequencing data across six subjects. Of 608 top-ranked predicted peptides tested by pMHC-multimer assay, 37 were immunogenic — six percent. A model integrating presentation and recognition features filtered out 98% of non-immunogenic peptides at precision above 0.70, validated on a further 310 epitopes. Prediction is improving. It is still throwing away roughly fifteen of every sixteen top-ranked candidates.

What the clinical record shows

Two glioblastoma trials found real intratumoral biology. In the 2019 NeoVax glioblastoma study, 8 patients received up to 20 long peptides with poly-ICLC after resection and radiotherapy; patients who received dexamethasone generated no circulating response, while those who did not generated polyfunctional CD4 and CD8 responses, with neoantigen-specific T cells detectable in the tumour afterwards. GAPVAC-101 enrolled 16 newly diagnosed patients and used a two-stage design — unmutated “warehouse” peptides first, then de-novo mutated peptides in 11 patients — reporting median PFS of 14.2 months and median OS of 29 months, alongside two anaphylactic reactions and one increase in cerebral oedema. Both are single-arm.

The clearest efficacy signal for a peptide vaccine is not in oncology drug development at all. In a 2009 NEJM trial, 20 women with HPV-16-positive grade 3 vulvar intraepithelial neoplasia received three or four vaccinations with long peptides spanning HPV-16 E6 and E7 in incomplete Freund's adjuvant. Roughly half showed complete regression of their lesions, correlating with induced HPV-16-specific immunity, with no adverse event above CTCAE grade 2.

Against that, the two large randomised trials both missed. Rindopepimut, an EGFRvIII 14-mer conjugated to KLH, was tested in ACT IV: 745 glioblastoma patients randomised across 165 sites, median OS 20.1 months with vaccine versus 20.0 months with control, terminated for futility. Tecemotide, a 25-amino-acid MUC1 lipopeptide in a liposome, was tested in START: 1,513 stage III NSCLC patients after chemoradiotherapy, median OS 25.6 versus 22.3 months, adjusted HR 0.88, p = 0.123 — not significant. A predefined concurrent-chemoradiotherapy subgroup looked better, but a subgroup is not a result.

Nothing you can get is a peptide vaccine

No peptide vaccine holds FDA or EMA marketing approval. The products people assume are peptide vaccines are not:

• HPV vaccines are recombinant virus-like particles built from the complete L1 major capsid protein, produced in yeast and self-assembled into non-infectious particles — whole protein, not peptide.
• Sipuleucel-T (Provenge), the only FDA-approved therapeutic cancer vaccine (2010), is an autologous cell product: a patient's own CD54-positive cells activated in vitro with PA2024, a recombinant PAP–GM-CSF fusion protein.
• CIMAvax-EGF, approved in Cuba (conditional 2008, definitive 2014), is an EGF–P64k protein conjugate with no FDA or EMA authorisation.

Peptide vaccines outside oncology

AN1792, a synthetic aggregated amyloid-beta 42 peptide with QS-21 adjuvant, was the first serious attempt at an Alzheimer's vaccine. In the interrupted phase IIa, 300 patients received AN1792 and 72 received placebo; the trial was halted when meningoencephalitis-consistent findings occurred in 18 of 298 treated patients (6%) versus none of 74 on placebo. Only 59 of 300 (19.7%) developed the predetermined antibody response.

CYT006-AngQb, angiotensin II conjugated to a Q-beta virus-like particle, went the other way. In 72 patients with mild-to-moderate hypertension, the 300 mcg dose reduced mean ambulatory blood pressure by 9.0/4.0 mmHg, with a much larger effect on the early-morning surge, and no serious adverse events. Cat-PAD, seven Fel d 1-derived T-cell epitope peptides of 13–17 amino acids designed not to bind IgE, produced a persistent two-year symptom reduction in early studies and then showed no difference from placebo in a 2016 phase 3.

The pattern across all of it is consistent: peptide vaccines reliably generate measurable immunity, and converting that immunity into a clinical endpoint has been the unsolved part. Related problems show up elsewhere in peptide therapeutics — anti-drug antibodies are the mirror image of the same immunology, and peptide degradation constrains what can be delivered and where.

Frequently asked questions

Is the HPV vaccine a peptide vaccine?

No. Gardasil-9 and similar HPV vaccines are made from recombinant virus-like particles built from the complete L1 major capsid protein of each HPV type, produced in yeast and self-assembling into non-infectious particles. That is a whole-protein vaccine, not a short synthetic peptide.

Has any peptide vaccine been approved?

Not by the FDA or EMA. Sipuleucel-T, approved in 2010, is an autologous cell therapy using a recombinant PAP–GM-CSF fusion protein. Cuba's CIMAvax-EGF is an EGF–P64k protein conjugate. Neither is a synthetic peptide vaccine.

Why do peptide vaccines need adjuvants?

A bare 8–10mer carries no danger signal, degrades quickly, and can load onto MHC class I on cells that lack costimulation. In mouse work, a minimal epitope in incomplete Freund's adjuvant primed CD8 T cells that then failed to expand or kill, while extending the same epitope to 30 amino acids restored function. Hence long peptides plus an immune-activating adjuvant such as poly-ICLC.

How reliable is neoantigen prediction?

On its own, not very. In the TESLA consortium study, 25 teams predicted neoantigens from shared sequencing data; of 608 top-ranked predicted peptides tested experimentally, only 37 (6%) were immunogenic. A refined model filtered out 98% of the non-immunogenic candidates at precision above 0.70, so it is improving, but the raw hit rate for computationally selected epitopes is about one in sixteen.

Informational only — not medical advice · 21+. VialHelp does not sell peptides and does not recommend any product or vendor. Consult a qualified healthcare professional for anything health-related.