Figure 1.
Comparison of specificity of in vivo immune responses with in silico predictions using various algorithms.
Individual rabbit (Panel A, n = 8 animals) or mouse (Panel B, n = 10 animals) immune sera were analyzed by ELISA and reported as OD values (bar graph). X-axis indicates the peptide ID and span; Y-axis indicates the OD value of the ELISA after subtracting the background measured for pre-immune sera. Horizontal line indicates threshold OD as described in Materials and Methods. Brackets indicate the segments of the protein predicted to be immunogenic by the KTApred (black), the Bepipred (blue) or the Discotope (green) algorithms. Open circles (grey) above individual bars identify peptides predicted by the ABCpred prediction tool. Asterisks above individual bars indicate positive responses detected by MALDI-TOF MS analysis. The MALDI-TOF data are the mean of three independent experiments using pooled immune sera. Responses from pooled pre-immunes were subtracted.
Figure 2.
Top-ranked structure predictions of the CelTOS protein using Rosetta (A), i-TASSER (B), and QUARK (C) as backbone traces.
Predicted conformational epitopes by Discotope are shown as colors for epitopes I (yellow), II (orange), and III (red). Regions predicted to be epitopes by Discotope but not found to be antigenic in peptide scans are shown in magenta.
Figure 3.
Consolidation of structural properties from in silico predictions and in vivo immune responses.
Computational epitope predictions (gray) are shown for ABCpred, Discotope, and Bepipred. Experimental epitope mapping using antibody peptide scanning (black) from both rabbit and mouse anti-PfCelTOS serum antibodies are significant above an OD-cutoff of the mean background responses plus three standard deviations from the mean. Computational secondary structure propensities for α-helix (blue), coiled (red) regions and disordered propensity (green) reported in a relative scale −1 to 1. All computational epitope definitions are based on classifications using default score cutoff values for Discotope and Bepipred. Score cutoff for ABCpred was optimized to maximize accuracy. Only results from computational methods with statistically significant predictions (p<0.001) are shown.
Table 1.
B-cell epitope mapping accuracy to mouse anti-PfCelTOS serum antibodies.
Table 2.
B-cell epitope prediction using second generation methods.
Figure 4.
Proteasomal and Cathepsin cleavage maps of CelTOS.
The proteasomal and Cathepsin D, L and S peptide fragments were separated by UFLC, analyzed on an LCMS-IT-TOF mass spectrometer and then identified using the MASCOT data base. The peptides derived from the proteasomal degradation of CelTOS (blue lines) are denoted above the sequence, and Cathepsin D (red lines), Cathepsin L (green lines) and Cathepsin S (purple lines) are denoted below the sequence, represent the identified peptides. The data shown is representative of two separate experiments.
Figure 5.
Fine specificity of PfCelTOS-specific T cells.
Reactivity was determined by ELISpot analysis measuring PfCelTOS-specific IFN-γ responses. Mouse splenocytes from three strains (inbred BALB/c and C57BL/6 and outbred ICR) were tested against a panel of 43 overlapping peptides (AA = amino acid position within the protein). Putative binding to indicated MHC class I and class II (in bold) alleles was determined by Rankpep analysis. Underlined amino acids designate predicted binding motif for indicated MHC allele. Shading and intensity of shading indicates the magnitude of the T cell response after ex vivo stimulation with the peptides.