Specific CTL epitopes restricted to the A*0201 HLA-A supertype were predicted for the HA proteins of H7N9, H5N1, and H9N2 influenza viruses. Three motifs 438VLLENQKTL, 455NLYNKVKRA, and 523KILTIYSTV named A, B, and C, respectively, were conserved between the H5N1 and H9N2 viruses. The HA proteins of H7N9 and H9N2 viruses possessed two similar CTL epitopes, 36TLTENNVPN and 324KLAVGLRNV (D and E, respectively), while a conserved motif was not detected between H7N9 and H5N1. The D and E motifs were attached on both sides of the A/-B/-C peptides using EAAAK repeats, and the resulting peptides were fused to HK-1 (Figure 1). The kozak sequence was introduced to increase the efficiency of translational initiation. Bioinformatics analyses were performed on two HA CTL/HK-1 constructs named Ⅰ and Ⅱ. The physicochemical properties of the constructs are listed in Table 1.
Figure 1. Schematic models of HA CTK/HK-1 constructs (A: construct Ⅰ; B: construct Ⅱ) including the conserved CTL epitopes restricted to the A*0201 HLA-A supertype among the H7N9, H5N1, and H9N2 inﬂuenza viruses attached to HK-1.
Table 1. physicochemical properties of HA CTL/HK-1 constructs
The mRNA structures of the HA CTL/HK-1 peptides were analyzed using RNAfold, and the minimum free energy levels were estimated to be -32.40 kcal/mol for construct Ⅰ and -31.36 kcal/mol for construct Ⅱ. These values indicate good stability of the mRNA structures. Secondary structure analysis (Figure 2) revealed that construct Ⅱ was composed of random coils (19.72%) and α-helixes (80.28%). Construct Ⅰ had higher α-helixes content (83.10%) and lower random coils content (16.90%) than construct Ⅱ. No β-sheets were detected in these constructs. 3D structures were modeled (Figure 3) and evaluated, and the best model was chosen for further analysis.
Figure 2. PSIPRED graphical result of secondary structure prediction of the two HA CTL/HK-1 chimera proteins, construct Ⅰ (A) and construct Ⅱ (B). The three conserved CTL epitopes restricted to the A*0201 HLA-A supertype between H5N1, and H9N2 inﬂuenza viruses are shown in red and others in black boxes.
On the basis of the results obtained from the alignment program in SWISSMODEL, the crystal structure of H9 influenza virus (ljsh.1.B) which showed 74.04% sequence identity was selected as the template. To select the best chimeric protein model, the generated 3D HA CTL/HK-1 models were compared. Construct Ⅰ was modeled with 93.3% confdence and 74% identity using the single highest scoring template. Construct Ⅱ was modeled with 94.1% confidence and the same identity as construct Ⅰ. The models were evaluated according to their template modeling scores (TM-scores), root mean square deviations (RMSDs), energy minimization, and Ramachandran plots. The estimated values for both constructs are listed in Table 2. The Ramachandran plots for the two HA CTL/HK-1 peptides were generated to assess the models (Figure 4). The plots revealed that the φ and ψ torsion of angles of all residues in the models were clustered around the secondary structure regions that defne backbone conformation. However, the torsion angles of construct Ⅱ were less widely distributed than those of construct Ⅰ, and they showed better clustering within the allowed regions of the plot.
Table 2. The QMEAN4 Z-score, TM-score and homology modeling RMSD of HA CTL/HK-1 chimeric proteins
Figure 4. Evaluation of HA CTL/HK-1 model quality by Ramachandran plot analysis. (A) construct Ⅰ: 85.1% of the residues were present in favored region, 10.8% in the allowed region, and 4.1% in the outlier region; (B) construct Ⅱ: 91.9% of the residues were in favored region, 2.7% in the allowed region, and 5.4% in outlier region. The φ and ψ torsion angles of amino acid residues in the proteins were reasonably accurate.
The overall quality of the models was validated; Z-scores of -3.63 and -4.27 were obtained for constructs Ⅰ and Ⅱ, respectively (Figure 5). Local geometry was evaluated on the basis of torsion angle potential, atomic distance-dependent interaction potential, and buried status of the residues (solvation free energy), using the QMEAN Z-score (Table 2). The higher Z-score (-1.65) obtained for construct Ⅰ may be attributed to differences in the torsion and solvation energies of the two constructs.
According to the Kolaskar and Tangaonkar antigenicity scale, the most likely antigenic determinants for construct Ⅰ were at positions 12-18, 23-29, and 31-58, with an average antigenic propensity of 1.028. For construct Ⅱ, the antigenicity plot predicted antigenic determinants at regions 4-18, 23-29, and 31-46 with an average antigenic propensity of 1.021 (Figure 6). According to the Levitt scale, amino acid residues at these positions fall inside the beta turn region (1.000 as the threshold level). The predicted segments from both sequences may be antigenic, and may elicit antibody response.
Figure 6. Antigenicity prediction plot of HA CTL/HK-1 chimera peptides, predicted using the Kolaskar-Tongaonkar method. Regions with antigenic propensity scales > 1 are predicted as antigenic regions. The average antigenicity of construct Ⅰ (A) was 1.028 at the 12AAKVLLE18, 23LEAAAKN29, and 31AKKILTIYSTVKLAVGLRNVEAA58 motifs. The average antigenicity of construct Ⅱ (B) was 1.021 at the 4VGLRNVEAAAKVLLE18, 23LEAAAKN29, and 31AKKILTIYSTVT46 motifs.
All possible mutant chimeric peptides were generated by a mutating single residue at a time, and the ability of each mutant to induce IFN-γ was predicted. Sorting of the mutants revealed that replacement of T at position 1 with K increased the probability of inducing IFN-γ from -4.352 to 2.572 in construct Ⅰ. Substitution of E at position 50 with F, H, K, L, M, Q, R, W, or Y in construct Ⅱ raised the score from -4.352 to 3.0. The HLA-A*0201-binding properties of the altered epitopes were predicted, and the results revealed that the substitutions had no effect on the HLA-binding affnity.