HLA-A*02:01 and Splice-Derived Neoantigen Presentation

HLA-A*02:01 (Human Leukocyte Antigen A*02:01, encoded within the Major Histocompatibility Complex [MHC] Class I locus on human chromosome 6p21.3 across 8 exons; 365 amino acids including a 24-residue signal peptide, assembling with non-covalently bound β₂-microglobulin [β2m, encoded on Chr15q21.1]) is the single most frequent polymorphic MHC Class I allele in human populations—carried by 40% to 50% of individuals of European and North American ancestry, ~20–35% of Hispanic and East Asian populations, and ~15–25% of African populations. On the plasma membrane of every nucleated human cell, the HLA-A*02:01 heavy chain (α1, α2, and α3 extracellular domains) cradles an 8-to-11-amino-acid (canonical 9-mer) intracellular peptide** inside its α1/α2 polymorphic groove, displaying a continuous snapshot of the intracellular proteome to the αβ T-cell receptor (TCR) and CD8 co-receptor of cytotoxic T lymphocytes (CTLs).

Historically, cancer neoantigen discovery focused almost exclusively on non-synonymous single-nucleotide missense mutations (SNVs). However, SNV-derived neoantigens differ from self-peptides by only a single amino acid (often subject to central thymic tolerance) and are scarce in low-tumor-mutational-burden (low-TMB) malignancies such as acute myeloid leukemia (AML), uveal melanoma, pediatric gliomas, prostate carcinoma, and estrogen-receptor-positive breast cancer. Over the past decade, immunopeptidomic mass spectrometry (LC–MS/MS) and transcriptome-wide long-read sequencing have revealed that aberrant alternative splicing, intron retention, cryptic exon inclusion, pangenome VNTR frameshifts, and spliceosome factor mutations (SF3B1, U2AF1, SRSF2, ZRSR2, or pharmacological PRMT5 inhibition) generate a vast reservoir of "dark-matter" splice-derived neoantigens. Because frame-shifted splice junctions produce novel amino acid sequences downstream of the junction that bear zero homology to the germline self-proteome, splice-derived peptides bound to HLA-A*02:01 (K_d / IC50 < 50 nM) elicit potent, un-tolerized CD8⁺ T-cell responses and serve as premier targets for TCR-engineered T cells (TCR-T) and bispecific ImmTAC (tebentafusp) therapeutics.

α1/α2 Superdomain Architecture and the Six Specificity Pockets (A–F)

Solved in 1987–1991 by Pamela Bjorkman, Don Wiley, and Jack Strominger (PDB: 1HLA, 1HHK), the X-ray crystal structure of human HLA-A*02:01 complexed with β₂-microglobulin (β2m, 99 residues) established how MHC Class I molecules bind thousands of sequence-diverse peptides with nanomolar affinity (t1/2 ≈ 6–24 hours at 37 °C):

1. The α1/α2 Peptide-Binding Platform (residues 1–180): Exon 2 (α1 domain, residues 1–90) and Exon 3 (α2 domain, residues 91–182) fold into an eight-stranded antiparallel β-pleated sheet floor topped by two long antiparallel α-helices (α1 and α2 helices) separated by a ~10–12 Å cleft (~25 Å long). Unlike MHC Class II grooves (which are open at both ends and bind 13–25-mer peptides), the ends of the HLA-A*02:01 cleft are closed off by bulky conserved residues (Tyr59/Tyr171 at the N-terminus; Tyr84/Thr143/Trp147 at the C-terminus), restricting canonical ligands to 8, 9, or 10 residues (with longer 11–12-mer peptides forced to bulge out of the center of the groove toward the TCR).
2. Conserved Backbone Hydrogen-Bond Clamps (A and F Pockets): At the A-pocket, a conserved cluster of tyrosine residues (Tyr7, Tyr59, Tyr159, and Tyr171) forms a rigid hydrogen-bond network with the free positively charged N-terminal α-amino group (NH3+) of peptide position 1 (P1). At the opposite F-pocket, Tyr84, Thr143, Lys146, and Trp147 hydrogen-bond to the negatively charged C-terminal α-carboxylate (COO−) of position 9 (P9/PΩ).
3. Polymorphic Primary Anchor Pockets (B-Pocket at P2 and F-Pocket at P9): Sequence polymorphism among the >25,000 known human HLA-A, -B, and -C alleles is concentrated in the six specificity pockets (A through F) lining the floor and walls of the α1/α2 groove, maintained across human populations by pathogen-driven balancing selection (heterozygote advantage, dN/dS ω > 3.0 at peptide-contacting codons). In HLA-A*02:01, two deep nonpolar pockets dictate its signature hydrophobic binding motif (x-[L/M/I]-x-x-x-x-x-x-[V/L/I]):

• The B-Pocket (accommodating Primary Anchor P2): Lined by hydrophobic and uncharged residues Phe9, Met45, Val67, and Glu63/Lys66, creating a deep nonpolar cavity that strongly prefers Leucine (Leu) or Methionine (Met) (and secondarily Ile, Val, or Gln) at position 2 (P2).
• The F-Pocket (accommodating C-Terminal Primary Anchor P9 / PΩ): Lined by Asp77, Leu81, Tyr116, Tyr123, and Ile124, which preferentially accommodates the aliphatic side chain of Valine (Val) or Leucine (Leu) (and Ile/Ala) at position 9 (P9).

Immunoproteasome Cleavage, TAP1/TAP2 Transport, and ERAP1 Trimming

For an aberrant splice-derived polypeptide to be presented by HLA-A*02:01 on the tumor cell surface, it must traverse the four-step endogenous MHC Class I antigen processing pathway:

1. Pioneer-Round Translation & Proteasomal Cleavage: Aberrant mRNAs containing retained introns, frameshifted cryptic exons, or unannotated alternative splice junctions undergo translation by ribosomes—including pioneer-round translation prior to nonsense-mediated decay (NMD)—generating defective ribosomal products (DRiPs) and truncated polypeptides. Ubiquitinated polypeptides are degraded by the 26S constitutive proteasome or the IFN-γ-inducible immunoproteasome (PSMB8/LMP7, PSMB9/LMP2, PSMB10/MECL-1), whose chymotrypsin-like active site (LMP7) preferentially cleaves immediately after hydrophobic Val, Leu, and Phe residues, directly generating the exact P9/PΩ C-terminus required by the HLA-A*02:01 F-pocket.
2. TAP1/TAP2 Translocation into the Endoplasmic Reticulum (ER): Cytosolic peptides of 8 to 16 residues are pumped into the ER lumen by the heterodimeric ATP-binding cassette transporter TAP1–TAP2 (Transporter Associated with Antigen Processing).
3. ERAP1/ERAP2 N-Terminal Molecular Ruler Trimming: Inside the ER, Endoplasmic Reticulum Aminopeptidase 1 (ERAP1) trims extended N-terminal residues down to a precise 9-mer (8–10-mer) length without cleaving the C-terminal anchor.
4. Peptide-Loading Complex (PLC) Proofreading: Tapasin (TAPBP), ERp57, and calreticulin hold empty HLA-A*02:01–β2m heterodimers in an open conformation and catalyze peptide exchange until a high-affinity peptide (IC50 < 50–500 nM, scored by NetMHCpan-4.1 with eluted ligand percentile rank %Rank < 0.5% for strong binders) locks the groove shut for Golgi export to the plasma membrane.

Five Mechanistic Sources of Tumor-Specific Splice Neoantigens

Unlike single-nucleotide missense (SNV) neoantigens—where 8 of the 9 amino acids in a 9-mer remain identical to wild-type self-protein—splice-derived neoantigens (neojunctions) arise from five distinct classes of RNA processing alterations that frequently shift the translational open reading frame (+1 or +2 frameshift), creating entirely foreign peptide tails (10 to >50 amino acids long) that contain multiple HLA-A*02:01 L/M (P2) and V/L (P9) anchor motifs:

1. Somatic Spliceosome Hotspot Mutations (SF3B1, U2AF1, SRSF2, ZRSR2):

• SF3B1 (HEAT repeat hotspot mutations K700E, R625H/C, H662Q): Present in ~20–30% of myelodysplastic syndromes (MDS), ~15% of chronic lymphocytic leukemia (CLL), ~20–25% of uveal melanoma, and ~3% of breast and pancreatic cancers. Mutant SF3B1 weakens interactions with SUGP1 and selects aberrant upstream branchpoints, driving recognition of cryptic 3′ splice sites (AG′) located 10–30 nucleotides upstream of canonical 3′ splice sites across hundreds of transcripts (BRD9, UQCC1, PPP2R5A, MAP3K7). When the inserted intronic segment is not a multiple of 3 nucleotides, the resulting frameshift generates shared, public SF3B1-mutant neoantigens presented on HLA-A*02:01 and HLA-A*24:02.
• U2AF1 (S34F / Q157P in AML and lung adenocarcinoma) and SRSF2 (P95H/L/R): Alter 3′ splice-site (YAG) or exonic splicing enhancer (CCNNGG vs GGNNGG) recognition, generating recurrent exon-skipping and cryptic-exon neoepitopes.

1. Intron Retention (IR) and Pharmacological Spliceosome Perturbation: Widespread intron retention is a hallmark of almost all solid and hematologic malignancies compared to matched normal tissues—and can be pharmacologically induced by PRMT5 inhibitors (in MTAP-deleted tumors), RBM39 sulfonamide molecular glue degraders (indisulam, E7820), or SF3B1 modulators (H3B-8800). Translation into the 5′ end of a retained intron before encountering an intronic stop codon produces highly immunogenic HLA-A*02:01-restricted intron-retention neoantigens.
2. Cis-Acting Somatic Splice-Site Mutations and Deep Intronic Cryptic Exons: Somatic mutations at 5′ donor (GT) or 3′ acceptor (AG) dinucleotides (frequently affecting TP53, GATA3, MET exon 14 skipping, and RB1) or deep intronic mutations that create a novel donor site induce exon skipping or pseudo-exon inclusion.
3. Chromosomal Gene Fusions and Post-Translational Proteasomal Peptide Splicing: Fusion transcripts (BCR–ABL1, FUS–DDIT3, TMPRSS2–ERG, DNAJB1–PRKACA in fibrolamellar carcinoma) create junctional neoepitopes; additionally, the 26S proteasome can catalyze transpeptidation (proteasomal cis-spliced peptides) that ligate two non-contiguous fragments of the same protein into an HLA-A*02:01 epitope.
4. Subtelomeric and Intronic VNTR / Transposable Element Exonization: Epigenetic loss of DNA methylation or N6-methyladenosine (m6A) regulation reactivates intronic LINE-1, Alu, and endogenous retroviral (HERV/LTR) cryptic splice acceptors ("TE-chimeric transcripts", such as JETs) and variable-number tandem repeat (VNTR) splice junctions, yielding tumor-shared antigens across ovarian, lung, and renal cell carcinomas.

Why TCR-Based Therapeutics Outperform CAR-T for Intracellular Splice Antigens

Conventional Chimeric Antigen Receptor (CAR)-T cells use an antibody single-chain variable fragment (scFv) and can recognize only extracellular cell-surface proteins (CD19, BCMA, HER2), which represent <10% of the human proteome. Because spliceosome factors, transcription factors, and splice-derived neojunction proteins reside in the nucleus and cytoplasm, they are invisible to CAR-T cells—but are robustly processed and presented on the cell surface as 9-mer peptide–HLA-A*02:01 (pHLA) complexes.

Two clinical modalities specifically target HLA-A*02:01–peptide complexes:

1. Autologous and Off-the-Shelf TCR-Engineered T Cells (TCR-T, e.g., afamitresgene autoleucel): Patient T cells are transduced with an affinity-optimized αβ TCR specific for an HLA-A*02:01-restricted tumor or splice neoantigen (requiring as few as 1 to 50 pHLA copies per tumor cell for cytolytic degranulation of perforin and granzyme B).
2. Soluble Bispecific ImmTACs (Immune Mobilizing Monoclonal TCRs Against Cancer, e.g., Tebentafusp): An engineered, disulfide-stabilized soluble TCR (affinity-matured to picomolar K_d ≈ 10–50 pM against the target pHLA-A*02:01 complex, compared to ~1–50 μM for natural TCRs) fused to an anti-CD3 single-chain antibody effector. Once bound to the tumor cell's HLA-A*02:01–peptide complex, the anti-CD3 arm recruits and activates any polyclonal circulating T cell regardless of its native TCR specificity—demonstrated clinically by FDA approval of tebentafusp (HLA-A*02:01–gp100 YLEPGPVTA) in metastatic uveal melanoma and advancing for shared SF3B1, KRAS, and PRAME splice/cancer-testis epitopes.

Why are splice-derived frameshift neoantigens often more immunogenic than single-nucleotide missense (SNV) neoantigens?

A single-nucleotide missense mutation (SNV) changes only 1 of the 9 amino acids in a canonical 9-mer peptide. Unless that single substitution creates a brand-new P2/P9 anchor or points directly upward at positions P4–P7 toward the TCR complementarity-determining region 3 (CDR3), the mutant peptide looks nearly identical to the wild-type self-peptide, for which high-avidity T cells were deleted in the thymus during central tolerance. By contrast, an out-of-frame (+1 or +2) alternative splicing or intron-retention event alters the entire downstream amino acid sequence (10–60+ novel residues before a premature stop codon). This foreign tail typically contains multiple completely non-self 9-mer peptides in which all TCR-facing residues (P3–P8) differ from the self-proteome, engaging high-affinity naive CD8+ T-cell repertoires.

Why doesn't Nonsense-Mediated mRNA Decay (NMD) prevent presentation of frameshifted splice neoantigens on HLA-A*02:01?

Although the Nonsense-Mediated Decay (NMD) surveillance machinery (UPF1–UPF2–UPF3B and the exon junction complex EJC) degrades mRNAs carrying premature termination codons (PTCs) located >50–55 nucleotides upstream of an exon–exon junction, NMD requires at least one "pioneer round" of ribosome translation to physically displace EJCs and detect the PTC. Crucially, Jonathan Yewdell and colleagues established that the truncated polypeptides synthesized during this pioneer round of translation (DRiPs) are co-translationally ubiquitinated and fed directly into the 26S proteasome and TAP1/TAP2–HLA-A*02:01 pathway—meaning that even an mRNA with low steady-state abundance due to NMD can efficiently populate cell-surface HLA-A*02:01 molecules.

How do tumor cells evade HLA-A*02:01 splice neoantigen recognition, and how can resistance be monitored?

Under selective pressure from CD8+ T cells or TCR-T therapy, tumor clones evolve immune escape through three primary genetic mechanisms: (1) Somatic Loss of Heterozygosity at the HLA locus (HLA LOH on Chr6p21.3), deleting the chromosome copy carrying HLA-A*02:01; (2) Biallelic truncating mutations in B2M (β₂-microglobulin on Chr15q21.1), which prevent folding and surface transport of all HLA-A, -B, and -C Class I molecules (though complete B2M loss removes inhibitory killer-cell immunoglobulin-like receptor [KIR] ligands and sensitizes the tumor to Natural Killer [NK] cell cytotoxicity); and (3) mutations in JAK1/JAK2 or TAP1/TAP2, which block IFN-γ-induced upregulation of the antigen-presentation machinery.

1. Bjorkman, P. J., Saper, M. A., Samraoui, B., Bennett, W. S., Strominger, J. L., & Wiley, D. C. (1987). "Structure of the human class I histocompatibility antigen, HLA-A2." Nature, 329(6139), 506–512. https://doi.org/10.1038/329506a0
2. Kahles, A., Lehmann, K.-V., Toussaint, N. C., et al. (2018). "Comprehensive analysis of alternative splicing across tumors from 8,705 patients." Cancer Cell, 34(2), 211–224. https://doi.org/10.1016/j.ccell.2018.07.001
3. Smart, A. C., Margolis, C. A., Pimentel, H., et al. (2018). "Intron retention is a source of neoepitopes in cancer." Nature Biotechnology, 36(11), 1056–1058. https://doi.org/10.1038/nbt.4239
4. Lu, S. X., De Neef, E., Thomas, J. D., et al. (2021). "Pharmacologic modulation of RNA splicing enhances anti-tumor immunity." Cell, 184(15), 4032–4047. https://doi.org/10.1016/j.cell.2021.05.038