The search and identification of immunogenic peptide epitopes has emerged as a promising strategy for developing a multi-epitope vaccine against toxoplasmosis [57, 63,64,65]. In this context, the application of bioinformatics tools in immunology, combined with reverse vaccinology, has significantly impacted the discovery of new candidates owing to its rapid and promising outcomes [66,67,68,69]. These approaches facilitate comprehensive scanning of potential pathogen antigens through molecular modeling, enabling the identification of interactions with host receptors and major histocompatibility complexes [70]. Additionally, Artificial Intelligence (AI) methods, such as Deep Learning and Artificial Neural Network architectures, have the potential to automate epitope identification by analyzing large datasets, thus accelerating the discovery of more effective vaccine candidates [71].

Recognizing the potential of in silico analysis, the first part of this study involved developing Artificial Neural Network (ANN) architectures and utilizing various bioinformatics tools to predict and select immunogenic peptides from T. gondii. This was accomplished either by analyzing the complete proteome of different strains of the parasite or by focusing on protein products from overexpressed genes in the protozoan during infection of human PBMC. Peptides with the highest probability of interacting with relevant HLA-I complexes that met the majority of the criteria in the in silico strategy were selected.

Interestingly, the 9-mer peptides with the highest predicted probabilities by ANNs across the seven parasite strains showed a consistent pattern of hydrophobic and aliphatic amino acids, such as leucine (L) and valine (V), at positions 2 and 9, respectively (Fig. 3A). Similarly, for 10-mer peptides, a pattern of hydrophobic, aliphatic, and aromatic residues (F and L) was observed at positions 1 and 2, with V and L at the C-terminal position (Fig. 3B). These findings are significant because this pattern of hydrophobic amino acids at these positions correspond to anchoring sites within HLA-I binding pockets [60]. It has also been reported that the B and F pockets of HLA-I, which interact with residues at these positions, exhibit chemical specificity for aliphatic, aromatic, and predominantly hydrophobic amino acids [72].

Additionally, we observed minor variations in the chemical specificities of the residues based on the HLA supertype (Fig. 4). For HLA-A*02, leucine (L) and valine (V) were predominantly found at these positions (Fig. 4A), in agreement with Sette et al. [72], where pocket B interacts with small aliphatic residues (such as L) and pocket F is specific for aliphatic and small hydrophobic residues (such as V) [72]. For the HLA-A*24 supertype, residues F/Y/W were identified at position 2 and I/F/W at the C-terminal position (Fig. 4B), which aligns with the chemical specificity of pocket B for aromatic residues (such as F and W), and with the broader specificity of pocket F, which interacts with aromatic, aliphatic, and hydrophobic residues [72]. Finally, for HLA-B*35, all the selected peptides contained proline (P) at position 2 and phenylalanine (F) at the C-terminal position (Fig. 4C). Sette et al., similarly reported specificity for proline in pocket B of the B*35 supertype, and affinity for aromatic, aliphatic, and hydrophobic residues in pocket F [72].

Beyond identifying the properties of peptides for potential HLA-I interactions, we observed that positions 4–6 in most of the epitopes with the highest predicted probabilities contained at least one of the following residues: F, I, W, or A. These residues have previously been linked to immunogenicity [28]. It has been described that, aromatic side chains, such as phenylalanine (F) and others (I, W, A), are positively associated with immunogenicity [49]. Conversely, small residues such as serine (S), K, M, and Q, are negatively associated with this characteristic [49]. Notably, it has been reported that amino acids at positions 4–6 are critical for interaction with T-cell receptor (TCR) residues and are therefore essential for activating the immunogenic response [49]. Functional evidence of this interaction may be linked to our experimental results, particularly by the production of IFN-γ, which is likely produced by CD8 + T cells from seropositive individuals that recognize T. gondii peptides.

Additionally, a key aspect of our in silico strategy was the review and characterization of the proteins from which the predicted peptides were derived. These proteins are located in the membrane and secretory organelles, such as rhoptries and micronemes, as well as in the cytosol, endoplasmic reticulum, nucleus, and mitochondria, which are associated with metabolic and catalytic functions, as previously reported by Cardona et al. [28]. and McMurtrey et al. [23]. This suggests that not only surface or secreted proteins are crucial for inducing an immune response in humans. This is particularly significant, as most of the epitopes evaluated for T. gondii to date have been derived from surface antigens and secretory proteins, such as SAG, GRA, ROP, MIC proteins, and more recently, ROM and MAG antigens [73]. Therefore, this study broadens the scope of potential proteins and peptides that may naturally activate cellular immune responses against T. gondii.

Moreover, we found that the genes encoding the proteins from which the peptides were derived were expressed across all three parasite stages that affect humans— oocysts, bradyzoites, and tachyzoites —and that the peptides were conserved among the different T. gondii strains analyzed (according to ToxoDB). These features make the peptides promising candidates for future vaccine development, as they may induce an immune response against infections at any stage or strain (typical or atypical) of the parasite [74]. Other studies have focused on identifying conserved regions as potential vaccine candidates. A recent in silico study aimed at designing a multi-epitope subunit vaccine for visceral leishmaniasis selected conserved B and T cell epitopes from four different Leishmania strains derived from four distinct antigenic proteins [75], and demonstrated that the vaccine triggered a pro-inflammatory response, including the proliferation of activated T and B lymphocytes, suggesting a promising vaccine construct.

After characterizing the proteins from which the peptides were derived, additional conditions were analyzed in the selection strategy. One of these conditions addresses a key feature of immunogenic peptides processed via the MHC-I pathway: a positive prediction for proteasomal cleavage and TAP transport. As described in the literature, in the conventional MHC class I antigen processing pathway, proteins are degraded by proteasomes into fragments of approximately 3–18 amino acids, only a subset of which reaches the transporter associated with antigen processing (TAP) located in the endoplasmic reticulum (ER) membrane. In the ER, these peptides are loaded onto MHC I molecules before being transported to the cell surface for presentation to T cells [25, 76]. A previous study found that approximately 15% of all peptides generated from a protein were successfully transported to the ER, with approximately 2.5% binding to an MHC molecule. Ultimately, approximately 50% of the peptides present on the cell surface are recognized by T cell receptors [78]. Thus, efficient prediction is crucial for identifying peptides that can effectively activate the T lymphocytes. At this step of the analysis, the complete sequences of the T. gondii proteins from which the peptides were derived were loaded into IEDB analysis resources, utilizing the combined antigen-processing predictor [77]. In this web server, proteasomal cleavage was predicted, where the score indicated the likelihood that the peptides contained the necessary residues for cleavage at the C-terminal end. Additionally, TAP transport prediction assessed the ability of the peptide to be generated with its extended N-terminal precursor [78]. Interestingly, all peptides previously selected by neural networks showed a positive score for proteasomal cleavage and TAP transport (total score > 0.5). Therefore, no peptides were excluded at this stage of the analysis.

As the final criterion in the in silico analysis, peptide alignment with human protein sequences was performed to eliminate those with identities greater than 70%, to avoid cross-reactivity and potential autoimmune responses. This threshold was chosen based on recent studies on epitope-based vaccine development targeting other infectious diseases. For example, in Chagas disease, T. cruzi epitopes for T and B cells were identified for inclusion in a multi-epitope vaccine, ultimately retaining 18 peptide sequences with identities below 70% to any human protein or microbiome-associated protein [79]. In our study, some peptides predicted by the neural network were excluded, including one derived from the RON3 protein (FLLDFLLYV) and another from an efflux transporter protein (LYLLHSWTW), as they showed 77% or higher identity with human proteins.

Regarding the experimental phase of the study, characterization of eligible individuals revealed that 38% were positive for the HLA-A*02 allele, 68% were positive for the HLA-A*24 allele, and 22% were positive for the HLA-B*35 allele. These frequencies are generally consistent with previous reports for A*02 (23–50%) and B*35 (16–23%), but the A*24 frequency was higher than those reported in studies from Colombia, South America, and other regions (20–36,5%) [28, 34, 36,37,38,39, 80,81,82,83]. This suggests that the HLA-A*24 supertype is more prevalent in our study population [54].

Following the identification of peptides and selection of individuals with relevant HLA-I alleles, peripheral blood mononuclear cells (PBMC) from individuals previously infected with T. gondii were used as a model to assess the immune response stimulated by the peptides. As reported in the literature, PBMC immune responses to T. gondii-derived peptides are predominantly characterized by elevated IFN-γ levels [84]. Specifically, in order to find immunogenic peptide epitopes, we decided to test our selected peptides with people with chronic asymptomatic infection because they have successful control of the infection and did not show pathological consequences (e.g., retinochoroiditis). Therefore, ELISpot and flow cytometry assays were developed with PBMCs from these individuals, in order to evaluate peptide-stimulated cytokine production and cytotoxic responses. The 9mer and 10mer aminoacids peptides were evaluated in ex vivo assays, but only those with 9 residues stimulated a relevant IFN and cytotoxic response.

Regarding the results of the ELISpot assay revealed that peptide P1 (FLFAWITYV), related to HLA-A*02, derived from the palmitoyl-transferase DHHC3 protein of T. gondii, induced an average of approximately 40 spot-forming cells (SFCs) per 250,000 PBMCs (Fig. 5A). When extrapolated to 1 × 106 of PBMC per well, as reported in previous studies [28, 64], this corresponds to roughly 160 SFCs. Similarly, peptide P8 (VFAFAFFLI), derived from a potassium ion channel protein and restricted to HLA-A*24, generated a stronger response (Fig. 6A) with an extrapolated value of 400–600 SFCs per 1 × 106 de PBMC. Peptide P6 (YPIAPSFAM), derived from a putative microneme protein restricted to HLA-B*35, showed an extrapolated value of approximately 200 SFCs per/1 × 106 PBMC (Fig. 7A). These values are comparable to those obtained in prior IFN-γ ELISpot assays for immunogenic T. gondii peptides restricted to HLA-A*11:01 tested in PBMC from seropositive individuals [64]. This study evaluated five peptides from membrane and secretory proteins, including SAG1224 − 232(KSFKDILPK), SAG2C13 − 21(STFWPCLLR), GRA589 − 98(AVVSLLRLLK), SRS52A250 − 258 (SSAYVFSVK), and GRA6164 − 172 (AMLTAFFLR), which stimulated approximately 100–200 SFCs per 1 × 106 de PBMC [64]. This indicates that our peptides, selected through the rational strategy and derived from cytosolic or metabolism-associated proteins, elicit immune responses comparable to those generated by secretory or surface proteins of the parasite, which are recognized as immunogenic.

It is important to emphasize that the selection of peptides restricted to HLA-I was intended to activate the cellular response primarily mediated by CD8 + T lymphocytes (TL), which have been recognized as crucial for protection against T. gondii [21]. CD8 + TL mediate their effector functions through the production of cytokines such as IFN-γ and tumor necrosis factor-alpha (TNF-α), and/or through cytolytic mechanisms, where cytotoxic T lymphocytes (CTL) are the protagonists. These responses are crucial for maintaining control not only against T. gondii but also against a range of intracellular infections [85]. The CTL response is the key effector arm of the immune system for protective immunity to control intracellular infections and is elicited by small linear epitopes from processed antigens [86].

When CD8 + TL were stimulated, the predominant cells involved in the cytotoxic response, as evaluated by flow cytometry, were central memory (CM) CD8 + T cells, which responded more robustly to stimulation with P1, P6, and P8 peptides in most of the individuals assessed. Although the magnitude of this activation was relatively low (< 5%) in terms of the percentage of CD8 + T cells positive for the degranulation marker CD107a (Figs. 9 and 10, and 11), there were significant different when compared with the negative control. Additionally, the observed percentages of CD107a + cells were comparable to those previously reported for human PBMC stimulated with cytomegalovirus peptides (~ 0.42%) [58].

We found a higher proportion of CD107a-positive CM CD8 + T cells responding to peptide stimulation than effector memory (EM) T cells, which is consistent with a recent study that evaluated Leishmania braziliensis epitopes in central and effector memory CD4 + and CD8 + T cells from individuals with cutaneous leishmaniasis and healthy controls. The study reported significantly higher frequencies of CM CD8 + cells in affected individuals following peptide stimulation, indicating that these cells act as reservoirs of antigen-specific cells [87]. Additionally, it has been noted that CM T cells have greater proliferative capacity compared to EM T cells after antigenic stimulation, especially when previously exposed to the same antigenic source [88]. Therefore, our peptides could induce a peptide-specific cellular response and proliferation of CD8 + CM T cells.

On the other hand, peptide P8 induced response in PBMC from both, chronically asymptomatic and seronegative individuals with no statistically significant differences between the groups (Fig. 8). To explain the promiscuous behavior of this peptide, we hypothesized that P8 may activate not only memory T lymphocytes in seropositive individuals, but also innate immune cells in seronegative individuals. It is plausible that the peptide interacts with other innate immune receptors [89], potentially triggering the expression of transcription factors and the transcription of cytolytic effector genes, leading to the production of proinflammatory cytokines and/or the release of cytotoxic granules [90]. The analysis of these innate immune cells in seronegative individuals stimulated with P8 is a perspective of this study.

The final aspect of the analysis focused on understanding how the proteins from which peptides were derived entered the antigen MHC-I processing pathway. P6, originating from an MIC-type protein, is likely to be secreted into the host cell by secretory organelles, micronemes. The protein probably follows the conventional MHC-I processing pathway, consistent with the literature and most reports on T. gondii [25, 73]. In contrast, P1 and P8, which are derived from the cytoplasmic proteins of the parasite, it is hypothesized to have been processed via alternative mechanisms, such as autophagy. Autophagy is an intrinsic cellular recycling system [76], a catabolic pathway generally used to degrade cytoplasmic material, and is capable of mediating pathogen destruction [91]. It can be stimulated by both the innate and adaptive immune responses [92]. It has been proposed that autophagy can intersect with the MHC-I presentation pathway to alert CD8 + T cells to intracellular infections [76]. Based on this hypothesis, autophagy likely plays a dual role, not only in facilitating intracellular pathogen destruction but also in enhancing immune surveillance by presenting pathogen-derived peptides, as seen with T. gondii in this case.

Considering these results, P1, P6, and P8 peptides from T. gondii could be part of a multi-epitope vaccine candidate that can activate memory and innate immune responses. The literature reports that multi-epitope vaccines more accurately mimic antigen processing and presentation during a natural infection, and thus induce more efficient protective immunity than whole-protein vaccines [19]. Due to the complexity of the parasite life cycle and the variability of parasite antigens, the development of multi-epitope vaccines against T. gondii is an attractive alternative to previous vaccine approaches [17].

In summary, this study offers valuable contributions to the discovery, identification, and evaluation of T. gondii immunogenic peptides as potential components of future multi-epitope vaccines against human toxoplasmosis. This study integrated the application of in silico strategies with ex vivo methodologies, employing human PBMC models and various immunological techniques. However, additional methodological approaches, such as analyzing granzyme B and perforin as part of cytotoxic activity and other Th1 cytokines, studying larger populations, and utilizing animal models, are necessary to validate the proposed peptide candidates further. Furthermore, this combined in silico and ex vivo approach could be applied to other pathogenic microorganisms in order to find peptide sequences that could trigger an immune response.