Prediction of Conserved Peptides of Paracoccidioides for Interferon-γ Release Assay: The First Step in the Development of a Lab-Based Approach for Immunological Assessment during Antifungal Therapy

Impaired antigen-specific cell-mediated immunity (CMI) is a primary immunological disturbance observed in individuals that develop paracoccidioidomycosis (PCM) after exposure to Paracoccidioides spp. Restoration of Paracoccidioides-specific CMI is crucial to stop the antifungal treatment and avoid relapses. A convenient and specific laboratory tool to assess antigen specific CMI is required for the appropriate clinical treatment of fungal infections, in order to decrease the time of antifungal therapy. We used an interferon-γ release assay strategy, used in the diagnosis of latent tuberculosis infection, to address our aims in this study. Information on proteins secreted by two well-studied representative strains—Paracoccidioides brasiliensis (Pb18) and P. lutzii (Pb-01)—were explored using PubMed or MEDLINE. From 26 publications, 252 proteins were identified, of which 203 were similar according to the Basic Local Alignment Search Tool. This enabled a selection of conserved peptides using the MEGA software. The SignalP-5.0, TMHMM, IEDB, NetMHC II, and IFNepitope algorithms were used to identify appropriate epitopes. In our study, we predicted antigenic epitopes of Paracoccidioides that could bind to MHC class II and induce IFN-γ secretion. These T cell epitopes can be used in the development of a laboratory tool to monitor the CMI of patients with PCM.


Introduction
Paracoccidioidomycosis (PCM) is a systemic mycosis endemic to Latin America and is caused by fungi belonging to the genus Paracoccidioides [1]. Based on studies of nuclear and mitochondrial genealogy, five species of Paracoccidioides-P. brasiliensis, P. lutzii, P. americana, P. restripiensis, and P. venezuelensis-are reportedly responsible for PCM [2]. However, P. brasiliensis and P. lutzii are the primary representative species used in clinical, molecular, morphological, and immunological studies on fungi-host interplay, with the findings having implications in laboratory diagnosis [1,3].

Bioinformatics Prediction Programs
Screening for peptides of P. brasiliensis and P. lutzii that would cause IFN-γ release in IGRA involved the following steps: 1. target protein screening, 2. characterization of target proteins based on conserved region (CDD)-related proteins associated with energy production, cellular respiration, and immune defense, 3. search for conserved regions of the target proteins, 4. search for homology of amino acid sequences between the two species, 5. search for conserved antigenic regions, 6. prediction of immunogenicity of the binding peptide to class II MHC molecule, and 7. prediction of IFN-γ secreted by T cells. The study design has been depicted in Scheme 2.

Bioinformatics Prediction Programs
Screening for peptides of P. brasiliensis and P. lutzii that would cause IFN-γ release in IGRA involved the following steps: 1. target protein screening, 2. characterization of target proteins based on conserved region (CDD)-related proteins associated with energy production, cellular respiration, and immune defense, 3. search for conserved regions of the target proteins, 4. search for homology of amino acid sequences between the two species, 5. search for conserved antigenic regions, 6. prediction of immunogenicity of the binding peptide to class II MHC molecule, and 7. prediction of IFN-γ secreted by T cells. The study design has been depicted in Scheme 2.

Investigation of Amino Acid Sequences
The first step comprised identification of upregulated proteins secreted in the yeast phase using GenBank (Available online: https://www.ncbi.nlm.nig.gov/pubmed (accessed on 02 April 2019).

Identification of Conserved Peptides from Protein Sequences of P. brasiliensis and P. lutzii
To identify conserved peptides, the MEGA software v.7.0 (Molecular Evolutionary Genetics Analysis) was used [16]. The tool showed variable and conserved regions along the sequences in pairs obtained in the alignment. From the conserved regions, sequences that were 15 amino acids in length (peptides) were selected for antigenicity and immunogenicity analyses.

Prediction of Antigenic and Immunogenic T Cell Epitope
Antigenicity prediction is essential for proper peptide synthesis because it indicates whether a peptide can bind the MHC class II molecule and form an antigen complex that is recognized by T cell receptors [17]. To determine which peptides would bind the MHC class II molecule, two predictors-signal webserver v. 3.0 and TMHMM webserver v.2.0-were used to indicate signal peptide cleavage sites, with a cutting bridge <0.4 on the D axis [18]. Thus, the selected peptides would remain intact without undergoing cleavage, thereby integrating the MHC molecule. TMHMM webserver v.2.0 analysis revealed that the peptides in the host cell were in the extracellular portion of the cell membrane. A threshold of >1 was considered for extracellular peptide regions [19]. To predict immunogenicity, we used two commonly accessed IEDB data sets (Free Epitope Database and Prediction; available online: https://www.iedb.org/ (accessed on 02 May 2019) and NetMHCII webserver v.2.3 (Available online: http://www.cbs.dtu.dk/services/NetMHCII (accessed on 02 May 2019). IEDB is a data bank that integrates different algorithms for prediction of the binding affinity between peptides and HLAs. Results were obtained via NetMHCpanII, NNAlign, SMMalign, Storniolo, and Consensus methods, all based on the <1% rank for strong binders [20]. In this study, epitopes were predicted for the seven subsets of HLA alleles existing in the overall global and Brazilian populations (HLA-DP, HLA-DQ, and HLA-DR). NetMHC II webserver v. 2.3 is an algorithm that can predict the binding affinity between peptides and HLAs (HLA-DR, HLA-DP, and HLA-DQ). However, this tool is based on the percentage rank and affinity, which is calculated using the formula 1-log(IC 50 nM)/log(50.000). Both methods classify potential ligands with values of <1 [21].

Prediction of IFN-γ-Inducing MHC Class II Binders
The IFNepitope algorithm was used to predict if the selected epitopes would induce secretion of IFN-γ [16]. Antigenic regions were explored to confirm the ability of these epitopes to induce IFN-γ secretion via MSV, classifying inducers as a positive and non-inducers as a negative. Positive epitopes presented with a score of approximately 0.0.

Validation Screening
To validate our results, the ESAT-6 protein was used as a model. This protein was used in commercial IGRA tests for LTBI and represents strong immunogenic epitopes, which are recognized during tuberculosis infection [22]. The ESAT-6 sequence was downloaded (GenBank: ABD98021.1) and submitted for analysis using the following algorithms: Signal P, TMHMM, NetMHC II, IEDB, and IFNepitope. These data are shown in ( Figure S1).

Selected Articles and Protein Inclusion Criteria
Of the 150 articles, 25 containing information on the identity of the extracted amino acid sequences in GenBank corresponding to P. brasiliensis and P. lutzii were included; these included articles that described yeast proteins obtained from preparations of fungal culture, plasma, or peripheral blood mononuclear cells without the intervention of any drugs. Upregulated proteins associated with the fungal cell wall structure, respiration, defense, virulence, and induction of the immune response were included. The sequence identities were searched for in the introduction, methodology, results, or discussions of the articles listed in Table 1.

Analysis of Proteins Based on Their Conserved Domain Database
First, 252 proteins were classified according to their family based on data from the conserved domain database (CDD); 39 proteins related to nuclear activity and with multiple functions were excluded. Similarly, G-type proteins, such as GPI glyceraldehyde aldolase, fructose-1,6-bisphosphate aldolase, glyceraldehyde-3-phosphate, and GTPases, were excluded because the peptide ligands could not dissociate from the host intracellular membrane; these would not be consequently expressed on the cell membrane associated with the MHC molecule.
Another 63 proteins were excluded for not fulfilling any of the similarity criteria. A total of 127 proteins were included based on their query score; however, despite having an E-value cutoff equivalent, two proteins--aqualysin (PADG_04168) and 1,3 β-gluconase (PADG_07461)--exhibited <90% identity and were excluded. A total of 125 antigen sequences were selected for further analysis ( Table 2). Table 2. Alignment of amino acid sequences of P. brasiliensis and P. lutzii by the BLAST tool.

Protein
Identification E-Value Identity Pb18 Pl 01  SignalP identified 19 proteins with signal peptides in most of the P. brasiliensis and P. lutzii sequences. However, of the nineteen proteins, the TMHMM server identified only six sequences that exhibited transmembrane propellers. Nevertheless, these proteins were included in the peptide analyses as they revealed extracellular regions in most sequences. In this study, only seven sequences were intracellular and were excluded. Immunogenic analysis by NetMHC and IEDB of the sequences indicated that optimal epitopes were present in the following proteins-interalpha trypsin (PADG_06178), chitin synthase class VII (>ABV31248.1), peroxisomal hydratase-dehydrogenase-epimerase (PADG_08651), and phosphoenolpyruvate carboxykinase (PAAG_08203).
NetMHC II analysis of interalpha trypsin identified the HLA-DRB10101 alleles related to MHC epitope 1 as the binding partner-MSAFSRMTASLGFSK (15 amino acids, amino acid 510-526). The results showed a percentage rank of 0.07 and an affinity calculation of 0.9, thereby indicating this epitope as the strongest ligand in the group. For epitope 1, no signal peptide was found by the SignalP algorithm, whereas TMHMM located it outside the cell membrane ( Figure 1).

J. Fungi 2020, 6, x FOR PEER REVIEW 10 of 19
SignalP identified 19 proteins with signal peptides in most of the P. brasiliensis and P. lutzii sequences. However, of the nineteen proteins, the TMHMM server identified only six sequences that exhibited transmembrane propellers. Nevertheless, these proteins were included in the peptide analyses as they revealed extracellular regions in most sequences. In this study, only seven sequences were intracellular and were excluded. Immunogenic analysis by NetMHC and IEDB of the sequences indicated that optimal epitopes were present in the following proteins-interalpha trypsin (PADG_06178), chitin synthase class VII (>ABV31248.1), peroxisomal hydratase-dehydrogenaseepimerase (PADG_08651), and phosphoenolpyruvate carboxykinase (PAAG_08203).
NetMHC II analysis of interalpha trypsin identified the HLA-DRB10101 alleles related to MHC epitope 1 as the binding partner-MSAFSRMTASLGFSK (15 amino acids, amino acid 510-526). The results showed a percentage rank of 0.07 and an affinity calculation of 0.9, thereby indicating this epitope as the strongest ligand in the group. For epitope 1, no signal peptide was found by the SignalP algorithm, whereas TMHMM located it outside the cell membrane ( Figure 1).
For peroxisomal protein hydratase (PADG_06851), NetMHC II identified alleles corresponding to the MHC (HLA-DRB10101) for epitope 3-RAYALLFSKLGAAVV (15 amino acids, amino acid 326-341)-with a rank percentage of 0.5 and affinity calculation of 0.8. For epitope 3, no signal peptide was found with the SignalP algorithm, and it was located in the outer part of the membrane ( Figure  3).
For phosphoenolpyruvate carboxykinase (PAAG_08203), NetMHC II identified alleles corresponding to MHC (HLA-DRB10101) for epitope 5-ERVSIIANPAVASLY (15 amino acids, amino acid 118-132)-with a rank percentage of 1.6 and affinity calculation of 7.5. For epitope 4, no signal peptide was found with the SignalP algorithm, and it was located on the outside of the cell membrane ( Figure 4). For class VI chitin synthase (ABV31248.1), NetMHC II identified alleles corresponding to MHC (HLA-DRB10101) for peptide 2-FDFYYLLTSASTPA (15 amino acids, amino acid 193-208)-considering the rank percentage of 0.05 and affinity calculation of 0.9. For epitope 2, no signal peptide was found by SignalP and TMHMM in the extracellular portion of the membrane (Figure 2).
For peroxisomal protein hydratase (PADG_06851), NetMHC II identified alleles corresponding to the MHC (HLA-DRB10101) for epitope 3-RAYALLFSKLGAAVV (15 amino acids, amino acid 326-341)-with a rank percentage of 0.5 and affinity calculation of 0.8. For epitope 3, no signal peptide was found with the SignalP algorithm, and it was located in the outer part of the membrane (Figure 3).

Prediction of IFN-γ inducing MHC Class II Binders
To determine which of the four epitopes could lead to the best, possible, abundant secretion of IFN-γ, three different analyses were performed using IFNepitope. To confirm the results, we used ESAT-6 as a control, as shown in Table 4.

Discussion
In the present study, in silico analyses allowed the prediction of immunogenic epitopes from P. brasiliensis and P. lutzii antigens. In this prediction, four peptides with fifteen amino acids restricted to the HLA class II molecule were identified as the optimal epitopes in the recognition of T cells.
These peptides were obtained from interalpha trypsin, chitin synthase, peroxisomal hydratase-dehydrogenase-epimerase, and phosphoenolpyruvate carboxykinase. In general, these proteins are produced when the fungus undergoes a mechanism of adaptation to the intracellular environment, in order to survive in a latent state [28,32,33,41,48]. The low-oxygen environment allows the fungus to survive via metabolic adaptation, thereby enabling the fungus to evade intracellular defense mechanisms while remaining protected by the endosome. To stay into the endosome, the fungus inhibits the phagolysosome maturation, lyses the host cell, detoxifies oxidative or nitrosative reagents, and uses different metabolic pathways to obtain energy from the available nutrients [26,49,50]. This mechanism is characterized by interactions between the fungus and host molecules and with the components of the cellular matrix, thereby causing delays in immune response induction [51]. Despite efforts required for fungal evasion, infected macrophages can act as antigen-presenting cells APCs, and can process and present antigens to TCD4 + cells. Recognition promotes the differentiation of TCD4 + cells into auxiliary T (Th1) cells, which when activated secrete IFN-γ, further activating the microbicidal function of alveolar macrophages and stimulating the presentation of antigens at the infected sites [52].
The alignment of protein sequences using BLAST made it possible to acquire data on conserved peptides of P. brasiliensis and P. lutzii ( Table 2). The aligned sequences when compared with GenBank sequences exhibited high similarity, thereby indicating that a conserved peptide could be identified. As P. brasiliensis and P. lutzii are detected at a higher frequency in Brazil [4] compared to that detected in the remaining regions of the world, only proteins of P. brasiliensis and P. lutzii were prioritized. As recommended, the alignment of protein sequences was assessed by checking the codons of the genome between the two species (ClustalW). This allowed us to select conserved peptides and to confirm the conservation of immunogenic epitopes predicted by the immunogenicity prediction analysis tools. Next, antigenicity and immunogenicity analyses were performed for these conserved peptide regions with reference to 125 genomic protein sequences. Thus, it was possible to identify the best HLA class II epitopes [53].
Four strong candidate peptides for stimulation of the in vitro cellular immune response were found--MSAFSRMTASLGFSK, FDVFYYLLTSASTPA, RAYALLFSKLGAAVV, and ERVSIIANPAVASLY. MSAFSRMTASLGFSK has been identified in interalpha trypsin peptide. This protein stimulates cellular differentiation [32]. High levels of trypsin have been observed in the sputum of patients with pulmonary cystic fibrosis. This is because the protein induced mucus hypersecretion and promoted the leukocyte recruitment to the inflammation site [54].
FDVFYYLLTSASTPA is part of the chitin synthase protein sequence, which forms α-glucan, a fungal cell wall layer in yeast. Studies have shown that high amounts of chitin synthase secreted by the fungus via proteolytic enzymes of the phagolysosome are used to reconstruct and maintain the cell wall [45,55,56]. Recently, this immunomodulatory activity of chitin synthase was demonstrated in the maturation of dendritic cells infected by P. brasiliensis, in vitro [57].
RAYALLFSKLGAAVV is a peroxisomal hydratase-dehydrogenase-epimerase peptide and studies have shown that the secretion of this protein is increased under stress conditions in fungal cultures and phagocytes infected by Paracoccidioides. It inhibits oxidative reactions and detoxifies molecules released by the lysosome [26,29,51].
ERVSIIANPAVASLY is a phosphoenolpyruvate carboxykinase peptide abundant in macrophages infected by Paracoccidioides [30]. Immunogenic epitopes of this antigen have already been identified in Leishmania major and L. donovani and have been used to develop a recombinant vaccine for leishmaniases. In animal models, this vaccine has shown promising results since the phosphoenolpyruvate epitopes activate TCD4 + lymphocytes, thereby promoting clonal expansion and increasing IFN-γ secretion [58].
In antigenicity analysis, proteins with a peptide signal at the beginning of the sequence or those lacking the signal exhibited higher affinity for antigenic epitopes. To determine this, we used two predictors-SignalP and TMHMM. SignalP revealed which peptides would be located in extracellular regions of proteins. The obtained results were confirmed using TMHMM, which showed that the peptides were located on the host cell membrane surface, thereby indicating an association with MHC. Both predictors have been used in studies analyzing proteins secreted by Fusarium oxysporum and Alternaria brassicicola for elucidation of proteins that are intracellular molecules and are involved in the pathogenesis of fungal infections [59].
Our four candidate peptides were extracellular molecules; in contrast, seven proteins were excluded because they presented signal peptides and cleavage sites in almost all peptide regions, thereby indicating an association with intracellular catalytic mechanisms. This may increase the costs associated with the production of these peptides as they could be cleaved during in vitro stimulation in blood monocytes. These analyses were possible because SignalP, in particular, is a predictor that classifies proteins as secretory or non-secretory in addition to indicating cleavage sites in the image. Thus, proteins that will be transported outside the cell and those that are localized inside the cytoplasm can be identified [60].
Although some proteins having a GPI anchor (e.g., glyceraldehyde dehydrogenase and fructose 1.6 bisphosphate) show no secretory signals, the presence of signal peptide in the C terminal chain has been considered [53]. This signal peptide was confirmed by the transmembrane helix analysis (TMHMM); therefore, although some proteins were antigenic, they were excluded as the peptides could not disintegrate from the membrane and cross the host cell membrane.
The results of the predictions performed by the NetMHCII and IEDB algorithms were promising as the epitopes with the highest binding affinity demonstrated values close to 0. The analysis of seven different subsets of HLA existing in the population showed that the four peptides identified exhibited excellent affinity, considering their IC50 and NNAlign rank [61].
To compare our results, we used the ESAT-6 protein of M. tuberculosis that was used as a positive control in the IGRA. In NetMHCII, ESAT identified two strong binding epitopes of rank <50% (epitope 1, QWNFAGIEAAASAIQ; epitope 2, WFAIEAAASAIQG). The IEDB showed greater binding affinity for HLA-DRB1 * 07: 01, HLA-DRB1 * 15: 01, HLA-DRB5 * 01: 01, HLA-DRB4 * 01: 01, HLA-DRB3 * 01: 01, and HLA-DRB1 * 03: 01 in the global population and to HLA-DRB1 * 07: 01 and HLA-DRB1 * 03: 01 in the Brazilian population. However, in the antigenicity analysis, the peptides QWNFAGIEAAASAIQ and WFAIEAAASAIQG did not show a peptide signal in SignalP and were located outside the host cell membrane as per TMHMM. Comparing the epitopes of ESAT-6 with the ones we found, the highest coverage of class II MHC was noted for the highest-class coverage percentage of 0% and <50%.
Epitope 4 induced the highest IFN-γ secretion in silico, via MHC class II. This result was similar to that found for the ESAT-6 epitope. Epitopes 2 and 3 have been shown to induce greater production of IFN-γ. Only epitope 2 did not show any relation with IFN-γ. We used the IFNepitope algorithm to confirm the induction of cytokines such IL-4 by epitope 2. However, for this purpose, it is essential to utilize the support vector machine (SMV). Furthermore, the relationship of IL-4 shows the involvement of these antigenic regions with the humoral immune response [20] and may be used in other contexts. In summary, the results show that epitope 4--ERVSIIANPAVASLY--can be most easily synthesized and incorporated into the IGRA laboratory tool for diagnosis and monitoring of PCM.

Conclusions
The present study predicted, in silico, four conserved epitopes present in P. brasiliensis and P. lutzii with potential for in vitro stimulation of T lymphocytes. The epitopes demonstrated high affinity by human allelic subsets in the Brazilian population. The epitope ERVSIIANPAVASLY showed the best performance in the induction of INF-γ, as compared to the epitope of ESAT-6 used in tests for IGRA. Therefore, the epitope identified herein can be used in the development of an IGRA for PCM, which, in combination with clinical assessments, can assist in the clinical follow-up of patients with PCM.