PerspectiveCell Biology

Cutting and Pasting Antigenic Peptides

See allHide authors and affiliations

Science  23 Apr 2004:
Vol. 304, Issue 5670, pp. 525-527
DOI: 10.1126/science.1097508

T lymphocytes recognize virally infected cells or tumor cells expressing virus-derived or tumor-specific peptides, respectively, bound to major histocompatibility complex (MHC) molecules. CD8 T cells recognize MHC class I-peptide complexes, whereas CD4 T cells recognize MHC class II-peptide complexes. The peptides recognized by CD8 T cells are normally 8 to 10 residues in length and are generated by antigen-presenting cells (APCs) through proteasomal degradation of cytosolic proteins, followed in many cases by trimming of the protein's amino (N) terminus. The peptides are shunted into the endoplasmic reticulum by a specific transporter, and here they bind to assembling MHC class I molecules [reviewed in (1)]. The complexes are then transported to the APC surface, where the T cell detects them and then kills any offending infected cell or tumor cell that it encounters. Many MHC class I-associated peptides and the proteins from which they originate have been identified. Predictably, the peptides are contiguous sequences within the proteins and are generated by simple proteolytic events. Recently, however, two antigenic peptides have been described that are derived by fusion of two distinct shorter sequences from the same protein, the consequence of an excision-and-splicing event (2, 3). On page 587 of this issue, Vigneron et al. (3) argue that peptide splicing is catalyzed by the proteasome during protein degradation.

Both reports describe peptides that are recognized by tumor-specific cytotoxic CD8 T cells (2, 3). In the first study (2), T cells were isolated from a patient with renal cell carcinoma. The tumor peptide recognized by the T cells was bound by the MHC class I allele HLA-A3 and derived from fibroblast growth factor-5 (FGF-5). Antigenic peptides can be identified by expressing truncated versions of the protein in a cell line expressing the appropriate MHC class I allele and asking whether the mutant proteins retain the ability to confer recognition by the T cell. In this case, deletion of amino acids 161 to 172 prevented recognition, which would normally indicate that the antigenic peptide lies between or overlaps one of these residues (2). However, deleting amino acids 212 to 220 also eliminated recognition. A peptide incorporating both residues 172 and 212, the minimum requirement, would clearly be longer than 8 to 10 amino acids. After considerable effort, the authors determined that the antigenic peptide was actually a nonamer consisting of residues 172 to 176 fused to residues 217 to 220. In the second report, Vigneron and colleagues obtained from a melanoma patient CD8 T cells that recognized a peptide from the melanocyte glycoprotein gp100 bound to the class I allele HLA-A32 (3). This antigenic peptide also proved to be a nonamer, beginning at residue 40 and ending at residue 52 with amino acids 43 to 46 excised from the sequence.

How does the excision process take place? In the first case cited, the precise mechanism was not established, although it clearly happened at the level of protein and not RNA. Stop codons inserted between the DNA segments encoding the two parts of the nonamer prevented its generation, indicating that translation of the complete sequence was essential. Also, addition of high concentrations of a synthetic peptide, corresponding to amino acid residues 172 to 220 and incorporating the excised 40 residues, to an HLA-A3-positive B cell line conferred recognition by the CD8 T cells. Generation of the spliced peptide was confirmed by purifying it from extracts of the cultured B cell line.

The authors of the second study went further. First, Vigneron and co-workers introduced a 13-amino acid oligomer (13-mer) precursor peptide (residues 40 to 52 of gp100) into the cytosol of HLA-A32-positive cells by electroporation. They then used a specific inhibitor to show that active proteasomes were required to excise the four residues from the 13-mer and confer T cell recognition. Second, they showed that purified proteasomes in solution could generate the nonamer directly from the 13-mer.

Could the proteasome directly splice together two shorter peptides to make the antigenic peptide? The answer is no: Addition of the N-terminal trimer and the carboxyl (C)-terminal hexamer to proteasomes failed to generate the nonamer. Such a finding implies that the splicing mechanism must be tied to proteolysis, and this insight led the investigators to their key experiment. When a peptide containing the N-terminal seven residues of the 13-mer was added to proteasomes along with the C-terminal hexamer, the spliced nonameric peptide was produced. To understand the meaning of this, one must consider the mechanism by which the proteasome cleaves a peptide bond (see the figure). The proteasome is a barrel-shaped structure of four rings, the outer two containing seven α subunits and the inner two containing seven β subunits. The active sites lie within the barrel and involve the N-terminal threonine residues of a subset of the β subunits. Cleavage involves nucleophilic attack by the threonine hydroxyl group on the peptide bond carbonyl group, generating a covalent acyl-β subunit conjugate that is rapidly hydrolyzed to release the peptide fragment (4). The authors propose, with good experimental support, that the terminal amino group of a second peptide confined within the proteasome can attack the acyl-enzyme intermediate instead of a water molecule. This regenerates a peptide bond between two previously separated amino acids (see the figure).

Peptide splicing by the proteasome.

Two separate peptide sequences within a protein (red and blue) are covalently connected after sequential cleavage steps to generate an MHC class I-binding peptide. In step 1, the amide bond at the N terminus of the C-terminal peptide (blue) is attacked by the active-site threonine residue of the proteasome β subunit. The bond is cleaved and an acyl-enzyme intermediate with the rest of the protein is formed as the C-terminal peptide is released. Hydrolysis in step 2 breaks the bond of the acyl-enzyme intermediate. In step 3, the amide bond at the C terminus of the N-terminal peptide (red) is similarly attacked, releasing the sequence between the two peptides. In step 4, the C-terminal peptide generated in step 2—still confined to the interior of the proteasome—attacks the acyl-enzyme intermediate formed in step 3, splicing the two peptide sequences by reforming an amide bond (step 5). To produce an antigenic peptide that can bind to an MHC class I molecule, the two peptides should sum to 8 to 10 residues. The additional remaining N-terminal amino acids are removed by aminopeptidases acting in the cytosol or the endoplasmic reticulum.


The mechanism that deletes 40 amino acids from FGF-5 is likely to be the same as that which excises four amino acids to produce the gp100 antigenic peptide. How common and how efficient is the process? Many MHC class I-binding peptides have been isolated and sequenced, and no spliced peptides have previously been described. One could therefore argue that splicing is uncommon. Alternatively, because the peptides identified chemically are the dominant peptides from the hundreds bound to the MHC class I molecules of a cell, the process may be common but inefficient. Cytotoxic CD8 T cells are famously sensitive, detecting very small numbers of MHC class I-peptide complexes on the surface of APCs. In the most extreme example, Sykulev et al. (5) even suggest that a single MHC class I-peptide complex could be detected. Examples of peptides inefficiently generated but recognized by CD8 T cells abound in the literature. Some come from proteins derived from alternate reading frames, 5′- or 3′-untranslated sequences, or even introns [reviewed in (1)]. A recent example was produced by aberrant initiation of translation at a leucine codon instead of the normal methionine codon (6).

It is telling that the two spliced peptides cited here are derived not from foreign proteins but from normal human proteins. This is commonly the case for peptides recognized by tumor-specific CD8 T cells. MHC class I molecules bound to rare self peptides that do not induce immunological tolerance probably constitute the only antigenic complexes normally available to CD8 T cells for tumor recognition. The CD8 T cells activated by such complexes potentially could react with normal tissues, which raises the interesting possibility that spliced peptides might occasionally be the targets of CD8 T cells in autoimmune diseases.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
View Abstract

Stay Connected to Science

Navigate This Article