Antigenic prediction of C. albicans Sap2
In this study, in silico experiments (dry lab) were performed to stimulate Class II HLA antigen presentation of C. albicans virulent antigens. Unlike viruses, C. albicans is an extracellular pathogen. Thus, after the fungal antigen has been processed, the digested peptides would be presented in the context of Class II HLA molecules to CD4+ T cells (28). Hence, the Hotspot Hunter program, which is a suitable prediction tool for Class II HLA-DRB1, was chosen as a computational method and attached to CandiVF (25). The program uses an ANN method and HMM as predictive engines for identifying antigenic clusters of peptides that are able to fit into the groove of Class II HLA molecules. ANN has been used for the prediction of peptides that bind to both multiple Class I and Class II HLA molecules, with a sensitivity and specificity close to 80% (29).
HMM is a novel predictive engine for T cell epitope prediction (30). The use of HMM in the prediction of peptide antigenicity has been demonstrated in Class I HLA-A2, but not in Class II HLA. The model gave a high accuracy of prediction (30). Notably, difficulty in Class II HLA epitope prediction is due to the length of peptides presented in the groove of HLA, which are approximately 11–30 residues. Therefore, the predictions were combined and selected as longer peptides containing a cluster of predicted peptides. Noguchi et al. were the first group to combine HMM with the SSS algorithm for optimization of HMM structure, and use this for the prediction of peptides that bind to Class II HLA-DRB1*0101 (30). They have demonstrated that S-HMM prediction accuracy is comparable to fully connected HMM and ANN methods.
Immunogenicity of predicted peptide epitopes
Validation of epitope prediction demonstrated that stimulation of PBMC by P11, P17 and P31 consistently led to the strongest proliferative indices and the highest number of IL-2 producing clones. Sequence comparison between these three peptides using in silico prediction results revealed that 50% of the amino acid sequence at C-terminal of P17, and 85% of the amino acid sequence at C-terminal of P31, were within the predicted areas. P11, on the contrary, lay outside the predicted region. Since there are mixtures of macrophages, B and T cells in PBMC, and these cells can interact with one another resulting in production of IL-2 by T cells and macrophages, the ELISpot assay for IFN-γ was performed to identify only T cell proliferative clones. The IFN-γ assay suggested that P17 and P31 were T-cell epitopes, because insignificant numbers of IFN-γ producing colonies were detected in P11 stimulated samples. When the P11 sequence was compared to previously published results (12), the peptide contained a B-cell epitope in the middle of the sequence (Fig. 5). Only a truncated form of B-cell epitope is present in the flanking P10 and P12 overlapping peptides, which show a lower degree of proliferative responses upon stimulation. It has previously been found that T and B cells recognize the same or very closely overlapping sites on a protein (31–33). Such a relationship between T helper specificity and B cell specificity for the same protein antigen can be explained by the T-B reciprocity hypothesis (34). These relationships play a possible role in regulating both arms of immune responses.
Figure 5. Sequence of synthetic peptides, P17 and P31 (upper panel) in comparison to the predicted Sap2 T cell epitope (highlighted). Lower panel displays Sap2 B-cell epitope (highlighted) lays within the sequence of P10–12.
Download figure to PowerPoint
It is important to note that even though IL-4 is the important cytokine in the regulation of IgE synthesis, our analysis of IL-4 levels in in vitro conditions showed no significant correlation with Th1 or Th2 response. Unlike IL-13, there was an abnormal balance of IL-4/IFN-γ production in the tested samples. This may be due to the fact that IL-13 expresses for longer periods than IL-4, as has previously been stated by Katagari et al. (35).
A three dimensional structure analysis of Sap2 showed that P31 lies on the surface of the Sap2 molecule while P17 is less accessible to external molecules (Fig. 6). Indeed, a general perception of B cell binding sites is that they tend to be exposed or protrude out of the molecule, or be assembled as topographic antigenic sites. In contrast, T cells specific for processed antigens are limited to recognizing short segments of continuous structures thereby limiting to primary and secondary structures. The structural analysis showed that P17 and P31 possess β-sheets at the central part of the peptides. The β-sheets carry hydrophobic amino acids whose side chains are pointed towards the inside of the Sap2 molecule (Fig. 6). P17 has one β-sheet which contains -AYSL-(β154-β157) (Fig. 6) while P31 shows two β-sheets with -DVVFNFS-(β261-β267) and -AKIS-(β270-β273) sequences. These β-sheets are enriched with the hydrophobic amino acids Y, L, V, F and I that are common anchor residues at the peptide position 1 for HLA-DRB1 binding (Fig. 6) (36).
Figure 6. The three dimensional map of P17 and P31 peptides (signified in orange) on Sap2 molecule. Amino acid residues (signified in blue) represent N-terminal region of peptides which are outside the predicted area.
Download figure to PowerPoint
The experimental validation showed that Hotspot Hunter can identify some promiscuous peptides that enable stimulation of T cell activation from volunteers of eight HLA-DRB1 supertypes. The reported allele frequency of the eight HLA-DRB1 supertypes among 140 Thais and Singaporean Chinese was approximately 34.2% and 38.9%, respectively (37). In the case of C. albicans Sap2, P31 was an example of a peptide within the predicted areas that could stimulate T cells from all blood donors. Nevertheless, its predictive binding score did not correlate with the immunogenicity of the peptide. This peptide consists of an amino acid sequence that spans 90% of the predicted peptide and is present in all Sap2 varieties. Additionally, both P17 and P31 could stimulate T cells of a donor that carries the HLA-DRB1 allele (*12/12) other than the eight predicted supertypes. Thus it is possible that P17 and P31 could promiscuously stimulate the T cell response of a broader population than previously expected.
PBMC proliferation and ELISpot assays do not verify the actual binding of a peptide to HLA-DRB1, but rather measure the response of T helper cells from HLA-DRB1 allele-matched individuals. In addition, there is a possibility that the stimulated peptides, such as P17 and P31, could be presented in the context of HLA-DQ and HLA-DP. However, it has previously been reported that most fungal infections on skin and mucosa induce HLA-DR, but not HLA-DQ, expression on the surface of the cells (38). A marked expression of HLA-DR antigens has been shown throughout the epithelium of patients with oral candidiasis by indirect immunohistochemistry (38). The only epithelial type that has been shown to display HLA-DQ is that of the oral cavity (38). Another study, on T cell reactivity to allergic bronchopulmonary aspergillosis, has demonstrated that 19 of 21 T cell clones specific to the Asp F1 antigen of Aspergillus fumigatus were restricted by HLA-DR molecules, and only the two remaining clones by HLA-DP molecules (39). Interestingly, HLA class II involvement in Asp f1 presentation in all patients studied has been restricted to one or a few of the alleles of a given DRB1 genotype (39). All in all, it appears that HLA-DR plays a more important role than other HLA molecules in fungal antigen presentation.
In summary, we have demonstrated the utility of immunomics, which is the database construction and prediction of T-cell epitopes using a computational approach, followed by experimental validation. With the genome and proteome data of many pathogens growing rapidly, such an approach significantly improves the efficacy of extraction and analysis of biological research. This study has also demonstrated the use of bioinformatics tools to accelerate immunological research through a concept of “reverse vaccinology” which represents a paradigm shift as compared to conventional approaches to vaccine development (40). The latter approach is time-consuming and can identify only antigens that can be purified. The combination of large-scale screening by informatics and targeted validation experiments defines a knowledge-based approach to epitope-driven vaccine discovery and design.