Binding to a specific MHC class I or class II allele is a prerequisite for a peptide to be available for T cell recognition. Understanding this interaction at the molecular level is therefore of great interest not only from a structural point of view (Garboczi et al., 1996; Garcia et al., 1996), but even more for designing vaccines and identifying potential T cell epitopes in autoimmune diseases and tumor-specific immune responses. Therefore, based on the known association of autoimmune diseases with particular HLA class I or class II alleles— e.g., HLA-DRB1*1501 with multiple sclerosis (Vogt et al., 1994) and HLA-DRB1*0401 with rheumatoid arthritis (Nepom and Erlich, 1991; Hammer et al., 1995)—the knowledge of al-lele-specific MHC anchor motifs has been employed to predict T cell epitopes within candidate antigens whenever the respective proteins were not easily available or large sets of overlapping peptides to span such a protein could not be afforded (Margalit et al., 1987; Sette et al., 1989; Rothbard and Gefter, 1991; Roberts et al., 1996; Honeyman et al., 1997, 1998; Mamitsuka, 1998; Milik et al., 1998; Sette and Sidney, 1998). Since immunodominance of peptides within a given protein often correlates with their binding affinity, and this in turn with the presence of the MHC binding motif, this approach is reasonable, although a number of factors need to be considered. For practical purposes, the following discussion outlines how the current knowledge can be employed for such epitope predictions.
Known MHC binding motifs are not listed here, because these data have rapidly expanded in recent years and are now easily accessible via the Internet (see Internet Resources). In particular, the Web site SYFPEITHI, a database of MHC ligands and peptide motifs (version 1.0) by Rammensee and colleagues, allows easy access to peptide sequences that have been shown to be recognized in the context of a specific MHC molecule, as well as to the MHC anchor motif, relevant references, and, for some MHC class I alleles, the prediction of epitopes in peptide sequences of interest. Another site, BioInformatics and Molecular Analysis Sec-
tion (BIMAS; K. Parker, R. Taylor, and colleagues), ranks potential 8- to 10-mer peptides according to their predicted half-time of dissociation to HLA class I molecules (Parker et al., 1994). Finally, Brusic et al. (1998b) have compiled large numbers of MHC-binding peptides in their database of MHC binding peptides (MHCPEP, version 1.3; see Internet Resources). These databases are constantly updated and expanded. Furthermore, a number of prediction algorithms and programs have been published (Margalit et al., 1987; Roberts et al., 1996; Honeyman et al., 1997, 1998; Mamit-suka, 1998), and constant refinement is sought by using artificial neural networks and other techniques (Fleckenstein et al., 1996; Altuvia et al., 1997; Honeyman et al., 1998; Milik et al., 1998; Falcioni et al., 1999; Sturniolo et al., 1999; Reche et al., 2002).
Some of the factors that need to be taken into account in trying to predict peptide agretopes have been mentioned above. Class I-associated peptides are usually eight or nine amino acids long. Within this sequence, the positioning of the MHC anchor amino acids is crucial and of relatively greater importance than the anchors in MHC class II molecules. However, the entire peptide backbone and amino acid side-chain interactions, their size, hydrophobicity, and charge all contribute to the MHC-peptide interactions and thus ultimately to binding affinity. As mentioned above, the peptide-binding groove of MHC class II molecules is open at either end and, consequently, MHC class II molecules have been shown to accommodate significantly longer peptides (i.e., twelve to twenty four amino acids), with much longer ones being eluted from MHC class II alleles (Chicz et al., 1993; Vogt et al., 1994). However, peptide lengths down to five and three amino acids, respectively, have been described to be recognized by HLA class I- or class II-re-stricted T cells (Reddehase et al., 1989; Hemmer et al., 2000). Unlike class I-binding pep-tides, the peptide-binding motif can be shifted to the N- or C-terminal end, repetitive or multiple motifs may result in binding in different registers, and one peptide may contain binding motifs for a number of different MHC class II alleles, and thus appear to be a promiscuous binder (Sette and Sidney, 1998; see below). Typical MHC class II anchor motifs contain three or more anchor positions with characteristic relative spacing (see below), and often a peptide does not show all three anchor amino acids, but only two or one. Furthermore, it's important to note that peptides without any obvious binding motif have been found experimentally to bind well. This observation is due to the fact that each amino acid contributes more or less independently to binding in a positive or negative way (Hammer et al., 1993; Southwood et al., 1998), and therefore negative or positive influences in certain positions (e.g., the absence of a favoring amino acid in an anchor position) can be compensated by others. The amino acids and their end modifications may further influence binding in an allele-non-specific fashion (i.e., N-terminal acetylation often improves binding, but proline may decrease it by modifying the peptide structure).
In the section below, examples are given for class I and class II molecules, respectively. These examples show how the knowledge of allele-specific anchor motifs can be employed for the prediction of T cell epitopes.
Was this article helpful?