Epitope dominance was analyzed by converting peptide responses to a residue-by-residue epitope score, which corresponds to the average response to all peptides that contain the residue. In the study by Bohle et al., the specificity of T-cell lines or T-cell clones from 13 allergic individuals were mapped using a series of 12-mer peptides spanning the complete sequence of Ves v 5 with 9-residue overlaps [2]. This density of coverage results in exceptionally good resolution of epitope positions. Proliferative responses were considered positive when the stimulation index (SI) was greater than 4 for lines cultured from the peripheral blood mononuclear cells (PBMC) of allergic individuals. Epitope-mapping data for the other allergens were handled similarly, and relevant sources and features of the analysis are presented in Table 1.

Calculation of epitope scores, alignment of datasets, and analysis of correlation vs. offset were performed using Microsoft Excel. Significance tests were performed using GraphPad Prism. Residue-stability profiles were calculated using the COREX/BEST implementation at with default values for all parameters [23]. Entropy factors were estimated with the COREX/BEST implementation, but they typically resulted in residue-stability profiles with less dispersion than could be obtained with a slightly lower entropy factor. Thus, the entropy factor was adjusted downward from the estimated value by 0.02. Residue solvent accessibilities were calculated using MOLMOL [24]. Crystal structures used for calculation of residue-stability and solvent accessibility are listed in Table 1. Since an high-resolution structure was not available for Cry j 1, an homology model was obtained on the basis of the structure of Jun a 1 (80% identical) using SwissModel [25]. For analysis of sequence entropy, homologous proteins were identified by the method noted in Table 1 and aligned using ClustalW [26]. Sequence entropy calculations were performed using BioEdit [27]. Significance of correlations was evaluated using t tests implemented in GraphPad Prism. In order to properly represent the sampling frequency of Ves v 5 epitope-mapping data, the significance tests were applied to paired datasets from which two-thirds of the points had been removed by retaining every third point.

For the identification of flexibility maxima in Ves v 5, profiles of B-factor, solvent-exposed surface area, and sequence entropy were smoothed with a 15-residue moving-window average. A "first-derivative" profile was generated by taking the difference between smoothed values for the current residue and the preceding residue. The first-derivative profile was smoothed with a 9-residue moving window average, and local maxima were assigned to residues where the smoothed first derivative became negative. For the identification of minima in COREX stability, a first-derivative profile was generated as described above using the raw COREX stability profile. The first-derivative profile was smoothed with a 7-residue moving window average, and local minima were assigned to residues where the smoothed first derivative became positive. Flexibility data for the other allergens were processed similarly, and the relevant sources and processing parameters are presented in Table 1.

The pattern of CD4+ epitope dominance that was observed for Ves v 5 in allergic individuals supports the hypothesized relationship to allergen structure. The number of allergenic epitopes is similar to the number of peaks of structural flexibility or stability, and the epitopes tend to be situated between the flexible and inflexible regions. The most frequent spacing of epitopes between flexible and inflexible regions (10 residues C-terminal from a flexibility maximum and 4–6 residues N-terminal from a stability maximum) is remarkably similar to the spacing of 12 residues C-terminal from the flexibility maximum that was described for T-helper epitopes in the outer domain of HIV gp120 [31].

The spacing observed in Ves v 5 is different from that observed in influenza hemagglutinin, wherein epitopes most frequently occurred 10 residues C-terminal from the stability maximum, rather than the flexibility maximum [30]. However, the pattern in hemagglutinin was a mirror-image of the pattern in Ves v 5 and gp120. In hemagglutinin, epitopes were on the N-terminal side of flexible segments, rather than the C-terminal side. For all three antigens/allergens, epitopes tended to be excluded from the most flexible sites in the proteins.

In the following discussion, patterns of epitope dominance will be discussed in the context of a mechanism by which allergen structure directs CD4+ T-cell epitope dominance. The mechanism is hypothesized to have the following elements:

1. Epitope dominance is due to preferential presentation of certain peptide-MHCII complexes. Although at least one report argues that the abundance of epitope presentation has little influence on immunodominance [32], other studies suggest that antigen processing and presentation have a potent influence on immunodominance [19,33,34]. We take the position that, especially in regard to promiscuously dominant epitopes (which are presented by more than one allele of MHCII), the cause of the dominance is the preferential presentation of the epitope.

2. Proteases and MHCII co-mingle in the antigen-processing compartment and therefore can compete for the sequences that satisfy requirements for binding to both proteases and MHCII. The following observations support this postulate. Several proteases are implicated in the processing of both antigens and the MHCII-bound invariant chain [35-37]. The proteolytic separation of two MHCII-bound epitopes was found to be a rate-limiting step in presentation of the epitopes [38]. The level of activity of asparagine endopeptidase (AEP) can control the presentation of an epitope that contains a cleavage site for AEP [39].

3. The antigen/allergen remains in a native-like conformation through the initial proteolytic nicking of the protein and/or loading of a fragment into an MHCII. Studies of protein folding and stability have provided examples in which proteins retain elements of native-like structure at low pH [9], following proteolytic nicking [40,41], and when parts of the protein are demonstrably unfolded [42,43].

4. Proteases and MHCII preferentially bind to antigen/allergen sequences that have low conformational stability and adequate affinity for the binding site. To some extent, these two properties could be mutually exclusive. For example, hydrophobic sidechains can stabilize binding of an antigen segment to both proteases [44] and MHCII [45,46], but hydrophobic sidechains also tend to be buried in structurally stable elements of protein structure [47], where they are unavailable for binding. Thus, there exists a three-way competition for interactions with antigen/allergen sequences that involves intramolecular folding, binding to the protease, and binding to the MHCII.

5. Proteases bind shorter flexible segments of antigen/allergen than MHCII [11,48]. It follows that the on-rates of proteases are faster than the on-rates of MHCII because the smaller binding sites require less reordering of the polypeptide and because short flexible segments occur with higher frequency than the long flexible segments in the natively folded antigen/allergen.

6. MHCII bind stably to epitopes that have adequate binding affinity. Thus, the MHCII protect the bound segments from proteolysis [49]. However, the kinetics of MHCII binding and dissociation are modulated by DM, which responds to APC activation [7,50].

7. The extensive and stable interactions of antigen/allergen segments with the MHCII provide a driving force for unfolding the antigen. Although we are unaware of any direct evidence supporting this postulate, peptide binding to MHCII has been described as cooperative process akin to protein folding [51]. In the absence of coupling to any energy source, the assembly of the peptide-MHCII complex and disassembly of the antigen/allergen structure may be considered two sides of a thermodynamic equilibrium.

These elements have been incorporated into a model for processing and presentation of allergenic epitopes (Fig. 7). The model illustrates the presentation of two epitopes but it is not intended to suggest that both epitopes are presented from every molecule, nor is it to suggest that other epitopes would not also be presented. These two epitopes are examples of two types, "deep" and "shallow", which are distinguished by the order of proteolysis and peptide loading. In the upper pathway, proteolysis precedes loading of the "deep" epitope. In the lower pathway, loading of the "shallow" epitope precedes proteolysis. Although it is possible that an antigen/allergen would be cleaved at multiple sites and adjacent epitopes be presented from each fragment, the fact that most allergic individuals mount a response to only one or two epitopes suggests that the more likely scenario is the presentation of one or two epitopes after a single proteolytic event. Subsequent proteolytic cleavages may produce fragments that are too small and unstable to resist complete proteolytic degradation.

Six of seven peaks of epitope dominance in Ves v 5 lie adjacent to peaks of flexibility. In the model, this pattern of epitope dominance is shown to arise from an initial endoproteolytic nick in a flexible segment, followed by loading of an adjacent epitope in the MHCII (Fig. 7, upper pathway). The regularity of this relationship is highlighted in the plot illustrating the flexibility maxima as single data points (Fig. 3). These six epitopes are effectively excluded from the most flexible regions, and they overlap the least flexible regions. Since these epitopes include residues that are buried in the protein interior, they are called "deep" epitopes. These epitopes are partially buried or otherwise sequestered from MHCII binding by three-dimensional structure until proteolytic nicking at a nearby site renders the epitope more accessible.

The dominance peak at residue 25 of Ves v 5 is exceptional in that it lies squarely over a flexibility maximum as defined by several criteria. This "shallow" epitope requires no proteolysis and little unfolding for loading into the MHCII. The dominance of this epitope probably results from a combination of good flexibility/accessibility and protection from proteolysis by continued association with the MHCII through multiple cycles of DM-catalyzed dissociation and rebinding (Fig. 7, lower pathway).

The positions of epitopes in venom allergen Api m 1 and pollen allergens Phl p 1 and Cry j 1 were not so regularly spaced from a maximum or minimum of flexibility that significant correlations could be identified at a single offset (data not shown). Nevertheless, most of the epitopes were excluded from the center of flexible segments. This pattern is consistent with cleavage of these allergens in the flexible regions, followed by loading into the MHCII of the newly generated fragments. Apparent exceptions to this general trend are discussed in the following.

All three dominance peaks in Api m 1 overlap the adjacent peaks of flexibility, which would seem to be incompatible with the "proteolysis-first" mechanism. However, as for most antigens/allergens, the naturally-processed MHCII ligands for Api m 1 have not been characterized, and thus we do not know the exact position of the N-termini. In the available study of Api m 1, epitopes were mapped using an irregular series of peptides that spanned the sequence in an average of 8-residue steps. Thus, it is possible that that the epitopes could be refined to smaller sequences whose N-terminal ends coincide with the most flexible, protease-sensitive sites, which would be completely consistent with the "proteolysis-first" mechanism.

For Phl p 1, the epitope(s) near residue 35 seems to be a strong candidate for the "binding-then-proteolysis" pathway because the epitope is centered on a 12-residue segment that was disordered in the protein crystal. Presumably, this segment has good affinity for the MHCII; and therefore it resists DM-catalyzed dissociation from the MHCII and is protected from proteolysis. The weakly immunogenic epitopes near residues 145 and 184 coincide with regions of flexibility, as defined by all four criteria. The modest immunogenicity of these flexible regions could be related to their location near the N-terminus of the second major domain of Phl p 1. An initial cleavage on the N-terminal side of the epitope at residue 145 may yield an independent molecule. N-terminal disordered segments may be particularly good ligands for MHCII. This hypothesis is consistent with the exceptional immunogenicity of the N-terminal epitope in Ves v 5 and Phl p 1, and it is also consistent with the regular pattern of epitopes on the C-terminal flank of flexible sites in Ves v 5 and HIV gp120, which suggested that the MHCII binds near the N-terminus of proteolytic fragments.

The regular pattern of epitope dominance in Cry j 1 is strikingly similar to the regular pattern of flexible sites in that protein. In the ribbon diagram, the basis for the patterns is apparent in the regular turns of the pectin-lyase-like beta helix. Turns of the beta helix recur at an average interval of 29 residues, and the flexible sites are distributed in a stripe down one face of the beta helix. Presumably, any one of these flexible sites can serve as the initial site for proteolytic cleavage, followed by loading of an adjacent epitope.

The bias toward dominance of epitopes on the C-terminal side of flexible sites in Ves v 5 suggests that a feature of either the loading mechanism or the structure of the protein favors loading of an N-terminal sequence after an initial proteolytic cleavage. However, the number of epitopes illustrating this bias constitutes too small of a sample to establish such a relationship. Although the epitopes tend to be closer to flexible sites on the C-terminal side, their positions are nearly centered on the stable/inflexible segments, as indicated in Ves v 5 by the anti-correlations with flexibility at offsets of 4–6 residues and the correlation with COREX residue stability at zero offset. In a compiled analysis of epitope spacing in nine antigens/allergens, epitopes were found to occur with equal frequency on the N-terminal and C-terminal sides of flexible sites [52]. On the basis of the data available, the existence of a bias toward loading the N-terminal or C-terminal fragment in any given antigen/allergen must be regarded as anecdotal.

Although many details of the allergen processing and presentation mechanism remain to be elucidated, the exclusion of epitopes from flexible sites in allergens suggests that the three-dimensional structure of the allergen exerts a strong influence on the pattern of CD4+ epitope dominance by targeting the initial sites of proteolytic processing.

The authors declare that they have no competing interests.

SJM participated in the analysis of Ves v 5 structure, the correlation of structural features with epitope dominance, and drafted, read, and approved the manuscript. SJL conceived of the study, and participated in its design and coordination and drafted, read, and approved the manuscript.