HLA and antigen presentation in vaccine design
Before you start: human leukocyte antigen (HLA), beta-2 microglobulin (B2M), T-cell receptor (TCR).
You will be able to: Explain how HLA type constrains a candidate peptide without proving immune recognition.
A personalized vaccine design must consider both the tumor's candidate antigens and the patient's antigen-presentation system. Human leukocyte antigen (HLA) is the human major histocompatibility complex (MHC). These molecules display peptides that T cells can inspect.
Before starting: HLA is the canonical definition; TCRs explain recognition.
Class I and class II
| System | Main display context | Conventional reader |
|---|---|---|
| Class I, including HLA-A, HLA-B and HLA-C | Most nucleated cells; commonly intracellular protein-derived peptides | CD8 T cells |
| Class II, including HLA-DR, HLA-DP and HLA-DQ | Mainly professional antigen-presenting cells, with inducible expression elsewhere | CD4 T cells |
Typical class-I peptides are shorter than class-II peptides, but lengths and processing vary. CD4 help can support a response; some CD4 cells also have cytotoxic functions. Class II is not merely a preliminary substitute for class-I recognition.
Type, display and recognition
A predicted fit is one filter; each subsequent claim needs its own evidence.
An HLA type identifies inherited allele versions. People can share alleles; the type is not a unique barcode. A mutation can produce a candidate peptide suited to one allele and poorly suited to another.
Tumor cells may lose an allele or alter presentation machinery. B2M, beta-2 microglobulin, supports classical class-I HLA structure. B2M loss can compromise many class-I targets without removing class-II presentation by the same mechanism. RNA expression alone does not establish assembled surface peptide–HLA.
Prediction quality is allele-specific
Algorithms use experimental binding or presentation data and sequence features. Training coverage differs by allele. HLA-C often has lower surface abundance than HLA-A/B, and allele representation can affect model confidence. That does not make HLA-C broadly more permissive, nor does one atlas's predicted yield establish that a patient's HLA-C targets are absent.
New binding, presentation or recognition experiments can address uncertainties, but these assays test different things. Calibrating a predictor requires held-out evaluation; cytokine responses cannot simply be treated as peptide-binding measurements.
Escape does not select the rescue
Losing a peptide or the relevant HLA display can limit a vaccine-induced TCR (T-cell receptor) response. In TNBC-MERIT, a relapse specimen illustrated B2M-associated presentation escape. It is a mechanistic observation from a small study, not a personal recurrence probability.
Surface-target chimeric antigen receptors (CARs) or engagers can avoid that particular recognition requirement. NK (natural killer)-cell missing-self biology may also become relevant. Effective killing still depends on activating signals, target availability, trafficking and clinical evidence. HLA loss does not make a tumor automatically susceptible to any of these therapies.
Try it
A candidate binds HLA strongly in a prediction. What additional question matters most before calling it an effective vaccine target?
Explain it back: Whether the relevant tumor cells process and display it, whether useful T cells recognize it, and whether that response safely changes clinical outcomes.
Explain it back
HLA matching constrains design; presentation and recognition must be distinguished from predicted binding.
Takeaway
HLA matching constrains design; presentation and recognition must be distinguished from predicted binding.
Next: Immunopeptidomics. Case-specific interpretation belongs in molecular profile and the HLA/B2M question.
Sources
Checked October 8, 2026.
- Sarkizova et al., HLAthena — primary presentation dataset and prediction methods.
- TNBC-MERIT primary paper.
- FDA tebentafusp review — allele-specific peptide–HLA recognition.