Proteomics-driven identification of potential vaccine candidates can be a sound approach for selecting promising antigens, which are elicited against environmental stimuli analogous to host response upon pathogen invasion and are physiologically relevant for pathogens within the host [102]. Availability of pathogen proteome information upon infection in host can provide opportunities for in silico mining of novel vaccine candidates, and this approach has been utilized for in silico design of an epitope-based vaccine against Theileria parasites of ruminants [103]. For a dimorphic human parasite like Leishmania, it is more important to target human stage associated antigenic proteins that are physiologically important for parasites to infect and establish in new host. In recent years, several studies utilized immunoinformatic approaches of epitope screening in designing epitope-based vaccines. Khatoon et al. [104], Singh et al. [105] and Vakili et al. [106] have previously reported the theoretical potential of in silico designed vaccines for visceral leishmaniasis. Notably, in a later study by Vakili et al. [107], the group further evaluated successfully the immunogenic potential of the multi-epitope vaccine, derived in part from known antigens, by administering the chimeric construct in experimental mice. This suggests that the in silico designed vaccines with epitopes derived from appropriate protein targets have the potential to progress toward advanced phases of vaccine development for visceral leishmaniasis. While the in silico studies by Khatoon et al. [104] and Singh et al. [105] largely utilized available genomic databases of L. donovani to select vaccine targets, Dikhit et al. [11, 108] performed thorough investigations involving in silico, in vitro and in vivo analysis to screen and validate immunogenic epitopes obtained from proteins that are increasingly expressed at infective stage of parasite. Such highly expressed proteins are likely important for physiological and/or infective process of the parasite and thus can be more effective vaccine targets. In this report, we took an approach to select such amastigote proteins in terms of contrasting abundance or specificity (abundant up to the level of detection) from comparative proteome profiles of L. donovani promastigotes and amastigotes. Based on propensity of those proteins for secretion in vitro and/or having secretory signal sequence, we further combined immunoinformatic tools to identify candidate antigens that have secretory potential. A comparison of methodological and outcome features among several studies that have employed in silico design and evaluation of epitope-based candidate vaccines against visceral leishmaniasis to date has been summarized in Table 4. Overall, our reported vaccine construct was found to be comparable to the earlier exclusively-in silico reports in terms of antigenicity, population coverage and receptor interaction. However, experimental studies remain crucial to validate the immunogenic potential of the designed vaccine.
Analyzing amastigote secretome through intra-macrophagic studies is considered difficult, while significant difference in secretome between amastigotes and promastigotes is unlikely based on relatively low stage-specific differences in gene expression [27]. However, due to dynamicity in the relationship between mRNA and protein abundance as L. donovani adapts to amastigote condition, comparative level of abundance can be more reliable indicator. Hence, our screening approach is relevant within the context. Perhaps, the most studied amastigote-specific vaccine candidate in L. donovani happens to be a cellular stress countering abundant surface antigen, A2, which has shown to confer whole or epitope specific efficacy in multiple immunization models [8, 109]. The vaccine construct reported in current study comprised of immunogenic T-cell specific epitopes, as predicted immunoinformatically from thirteen amastigote associated proteins. Five of them are known to associate with virulence in mammalian host (fructose-1,6-bisphosphatase, putative protein disulfide isomerase, putative lipophosphoglycan biosynthetic protein, leishmanolysin and cysteine protease), while others have putative roles in countering host induced stress response (thioredoxin-like protein, glutathione peroxidase, stress-inducible protein STI1 homolog), host-microbicidal activity regulation (proteasome endopeptidase) and protein synthesis (elongation factor 2). Three proteins were uncharacterized according to the proteomic studies. Protein domain and homology (to proteins of other Leishmania species) suggest that two of these proteins may potentially play role in drug resistance phenotype (E9BUW4) and protection from intracellular stress (E9BDB8), while the specific function of alpha/beta hydrolase domain containing protein (E9BQ40) in amastigotes has not yet been deciphered. On the other hand, lack of reports on experimental evaluation of immunogenic potential of several Leishmania proteins, which are included in our predicted set of antigenic proteins, is apparent. Among the thirteen proteins of current interest, six proteins (elongation factor 2, proteasome endopeptidase complex, putative protein disulfide isomerase, leishmanolysin, cysteine protease, putative lipophosphoglycan biosynthetic protein) or their species homologs are known to have proven immunoreactive properties (Table 1). Nevertheless, increased abundance of these proteins suggest their likely role of pathological significance in host invasion and/or survival. The antigenicity scores suggest that several truly produced and potentially immunogenic proteins, which have not yet been explored experimentally, could be abundantly produced in amastigote form of parasite. Immunological evaluation of these amastigote stage associated proteins may potentiate unraveling of novel Leishmania antigens in future.
In the context of functional roles of selected proteins, our designed vaccine has the potential to benefit the host by generating appropriate immune response both in early and progressive phase of systemic infection. Furthermore, almost all of the epitopes were found in corresponding proteins of L. infantum indicating potential cross-protection against this visceralizing species. Most of the VL cases are reported from the endemic zones of Indian subcontinent, East Africa and South America. Thus, in designing an epitope based subunit vaccine, it is important to estimate the fractions of population in the target endemic zones based on HLA genotypic frequencies. The immunogenic non-self CTL epitopes in the vaccine modeled here is estimated to cover 96.8%, 91.7% and 93.9% percent allelic population of Brazil, India and Sudan, respectively, with experimentally evaluated truly binding affinity [110], while for HTL epitopes, it is almost 100% for each of these population. The vaccine construct has antigenic properties while it was not found to be an allergen. The structure was found thermodynamically stable and surface-soluble, while the core is hydrophobic- a favorable feature for antigen processing. Vaccine specific, but not parasite protein specific humoral response was predicted, and this can be used as biomarker of vaccine efficacy [46, 111] without eliciting parasite specific B-cell response. Moreover, the construct structure showed good binding affinity in previously reported binding cavity of TLR4 [112-115]. The structural interface between TLR4 and the peptide adjuvant (APPHALS) used here has been extensively studied, in which the position occupied by the adjuvant peptide in the TLR4-MD2 complex has been suggested to be varying depending on its position in vaccine model and canonical activation of receptor was proposed to be mechanized by insertion of peptide adjuvant in MD2 [116]. Since we used already activating but hypo-responsive TLR4-MD2 crystal structure removed of LPS for docking [117], it was not possible to predict into agonistic behavior of bound vaccine. Nevertheless, our docking model was consistent with the observation of non-MD2 (non-canonical) binding of adjuvant linked peptide, while suggesting that vaccine intrinsic segment may have more affinity than the peptide adjuvant for binding to TLR4.
Overall, binding interface along with MD simulation of docked complex in solvent system hints at sufficiently stable cross-link of TLR4 and MD2 with no major bond rearrangement between TLR4-MD2, and TLR4*-MD2 heterodimer formations. Although, the simulation time was short, this is reasonable as none of the vaccine residues interacted at crucial MD2 binding sites [112]. On the other hand, H-bond was found to increase between TLR4 ECDs (where vaccine is bound to one TLR4 ECD) in vaccine bound form compared to the unbound TLR4, which suggests potential event of positive interactions and movement between the ECDs. Besides, reduction in electrostatic surface potential at the vaccine bound TLR4 interface was observed after docking, which was consistent in post-simulation structural interface. Simultaneously, it was observed that a homo-dimer destabilizing His458-His458* repulsion [114] at pre-dock TLR4 was nullified and superseded post-dock by a solvent stable pi-hydrophobic interaction. It is thus possible that change in interpolated charge difference between pre-dock and post-dock TLR4 interface could have contributed to the bonding rearrangement between TLR4 ECDs. Notably, this rearrangement also involved participation of other critical histidine (His431, His555) residues at TLR4-TLR4* interface [118] unlike the unbound structure. Overall, these events are congruent with non-canonical TLR4 activation model mediated by microbial peptides, metals and cationic lipid nano-carriers, which are suggested to not confer canonical interaction with other monomers but to induce bond rearrangement among receptor monomers upon interaction [112-115]. Although the exact mechanism remains to be elucidated, our observation suggests that the vaccine construct may possess characteristic peptide feature of a non-canonical TLR4 ligand [119, 120], which may facilitate TLR4-TLR4* dimerization for downstream activation of immune cells.
The trends of backbone RMSD, Rg and H-bond of the vaccine bound complex over the simulation period complied with structural flexibility rather than rigidity of the complex. The RMSF values of complex side-chain indicate that the higher fluctuations in TLR4 were of those residues, which are vaccine unbound and located in solvent exposed loop mostly at or around glycines [121]. Increased residual fluctuation at LRR10-12 and around Gly397 may also be attributed to the mutations introduced at the position 299 and 399 in TLR4 structure (4G8A) as reported in [117], which was used to dock the vaccine protein. Nevertheless, it is unlikely that vaccine interaction would induce dissociation in structural interface of natural TLR4-MD2 since none of the highly fluctuating TLR4 residues had any direct interaction with vaccine or MD2.
Simulation outcome of hypothetical immunization in VL susceptible HLA alleles (hypothetical heterozygous combination) was consistent with the predicted immunogenicity of vaccine. Furthermore, we showed that the simulation outcome can be dynamic for different constructs when we used the same criteria in simulation program and the same HLA profile to test two known vaccine candidates for VL. Importantly, for these peptides, IL-10 production was reported previously as either prominent (peptide-2) or lessened (peptide-1) in comparison to SLA in vitro. It is not expected that simulation results will reflect experimental outcome, however, we observed a general trend of difference in immunosuppressive cytokine (e.g., IL-10) induction potential between the two peptides from simulation outcome- with peptide-2 having more potent IL-10 induction capacity. Although statistical significance could not be inferred from simulation plots, the difference seems consistent with experimental result. Understandably, the predicted epitopes (not shown) in simulation program did not comply mostly with our target set of epitopes due to difference in epitope prediction algorithm [45]. However, when compared to the simulation outcome of the known peptides, the general trend was comparable to both peptides for IFN-γ induction, while TGF-β and IL-10 were predicted to be considerably less pronounced than that by peptide-2. Besides IL-10, TGF-β has potent immunosuppressive properties, enhances disease progression, and may prevent cure and protective immunity development against leishmaniasis [122, 123]. Thus, the simulation prediction of higher propensity of the construct to induce a more Th1 polarized response rather than Th2 is consistent with our desired immunogenicity.
Despite difference in epitope set, simulation dynamics over time can be extrapolated for the estimated set of epitopes of our construct since it is also comprised of diverse T-cell epitopes and vaccine-specific B-cell immunogenic regions as predicted by several immunoinformatic tools. It has been proposed previously that the simulation dynamics can be consistent with a realistic immunization process in terms of primary and secondary immune responses [45]. Likewise, clearance of antigen, production of antibody, and development and persistence of memory B cells as well as CD4+ T cells over several months were assumed in the simulation outcome. For primary activation and maintenance of CD8+ T cells, CD4+ T cells (both Th1 and Th2 type) [124] are believed to be required [125, 126], where, cytokines such as IFN- γ, IL-2 and IL-4 could be involved [127, 128, 129, 130]. The simulation outcome suggests high levels of IFN-γ and IL-2, which may potentiate CD8+ T cell expansion. On the other hand, it is unlikely that the vaccine would trigger clonal expansion of epitope specific T-cells since we combined potent epitopes from several amastigote-associated proteins of comparable affinity, and it was consistent with the simulation dynamics for repeated exposure of 12 doses, as indicated by Simpson index (D). Rather, high level of IL-2 production can be expected for diverse epitope mediated immune response functional over long time in vaccine mediated immunity.
Experimental validation is utmost to prove this computational work. Next phases of the reverse vaccinology approach would ideally involve assessing the recombinant immunogenic protein expressed in E. coli (strain K12) system as proposed here, in vitro stimulation of peripheral blood mononuclear cells from active VL patients as well as healthy endemic people for cytokine production, and evaluation in challenge model. While a multi-epitope vaccine molecule generated by using reverse vaccinology approach can induce specific responses in in vivo and in vitro assays, a single recombinant molecule can also reduce the cost of production [131, 132]. The in silico designed vaccine reported here confers substantial immunogenic potential to be considered for in vitro experimental evaluation in the next phase of the study.