In-silico design and evaluation of an epitope-based serotype-independent promising vaccine candidate for highly cross-reactive regions of pneumococcal surface protein A

Streptococcus pneumoniae (pneumococcus) is an opportunistic pathogen and is a major cause of morbidity and mortality worldwide, with more than 98 serotypes based on their polysaccharide capsules [1, 2]. In 2018, the global pneumococcal burden was appraised to be 26.7 occurrences per 1,000 people, resulting in over 1,000,000 deaths [3, 4]. Effective treatment of pneumococcal diseases concerning antibiotic selection is a growing concern because of the increasing multi-drug resistance pattern of pneumococci [5, 6]. For the prevention of pneumococcal diseases, licensed vaccines are based on polysaccharide capsules of the most prevalent pneumococcal serotypes. The limited coverage of the licensed vaccines, broad geographical variation in circulating serotypes, non-vaccine serotype replacement, and the prevalence of non-encapsulated pneumococci from patients with invasive pneumococcal disease (IPD) are key reasons for an attempt to overcome the pneumococcal vaccine limitations and design the novel serotype-independent vaccines [7‐11]. Pneumococcal protein-based vaccine (PPV) formulation is a cost-effective and promising candidate for serotype-independent vaccine development [12, 13]; and many pneumococcal conserved cell-surface proteins have already been identified as ideal antigens for PPV in recent years [6, 14, 15].

Pneumococcal surface protein A (PspA) is a very important virulence factor that has been widely studied and is present in all pneumococcal strains [3, 16, 17]. Various active or passive immunization studies using rPspAs demonstrated that animal models were protected against the lethal challenge of pneumococci [7]. Furthermore, the administration of PspA in early human adult clinical trials has been reported [13]. Another study demonstrated that PspA immunization provides more comprehensive protection than Prevnar pneumococcal conjugate vaccine [18].

The N-terminal end of PspA, which is more variable due to mutation accumulation [19], has protection-eliciting epitopes, that have been divided into three regions A, B, and C [20]. The B-region of PspA is serologically variable and forms the basis of classifying PspA into three families with six clades. This region is identified as a clade-defining region (CDR) and comprises two prominent families. Family 1 contains Clades 1 and 2, and Family 2 is made up of Clades 3, 4, and 5. These two families are exhibited in almost 100% of clinical isolates from adult IPD and non-IPD children. Finally, Family 3 is composed of Clade 6, which is extremely rare among pneumococci, and it has been reported that the percentage of Clade 6 in pneumococcal strains is less than 1%. So in many studies, this clade is excluded from the study [20]. Analysis of the CDR sequence showed that the sequences belonging to the same clades demonstrated a sequence identity of ≥ 90% and those of different families ≤ 55% sequence identity [21]. Previous studies have shown high levels of cross-reactivity between different PspA fragments within the B-region of PspA [7, 22]. The C region of PspA is the Proline-Rich Domain (PRD), characterized by the presence of repetitive motifs of proline residues, and highly conserved 22-amino acid immunogenic epitopes called the Non-Proline Block (NPB). Although this region has a partly variable sequence, it is serologically highly cross-reactive and elicits antibodies against the PRD region, which can passively protect mice from lethal pneumococcal disease [21, 23].

Increasing evidence strongly proposes that a single protein, especially PspA from one family or clade, will not be sufficient to stimulate protection against all pneumococcal strains [12, 24, 25]. Higher levels of cross-reactivity have been reported within the same family, not between families, and the family-elicited immunity is clade-dependent [7, 22]. Therefore, at least one fragment from each of two prominent families has been considered for PspA-based vaccines to extend protection [22]. Akbari et al. showed that immunization of mice with PspAB1-5 (B region of N-terminal from all PspA clades) led to higher protection than PspA4ABC (A, B, and C regions of PspA Clade 4) in pneumococcal challenges [7]. Other studies showed that the combined vaccine candidate composed of two segments of each PspA family exhibited varying degrees of cross-reactivity and protection. Piao et al. showed that in three constructs of the PspA, including N-terminal and proline-rich regions from PspA families 1 and 2, immunization with PspA2 + 4 and PspA2 + 5 exhibited no protection against pneumococcal challenge with two Clades 1 and 3. Also, the binding capacity of the anti-PspA3 + 2 specific IgG to the surface of pneumococci with PspA Clades 1–4 was high, but not for Clade5. Finally, they concluded that PspA3 + 2 has an advantage over PspA2 + 4 and PspA2 + 5 [7, 25]. Akbari et al. also suggested that all B and C regions of all clades should be used in PspA-based vaccine designs to achieve the full level of cross-reactivity and cross-protection against all pneumococci [7]. Therefore, an essential step for PspA-based vaccine design is to cover and overcome clade-dependent immunity against all pneumococcal strains expressing all PspA families by selecting immunodominant truncated domains of all PspA clades focusing on two cross-reactive B and C-regions. This can be achieved through cost and time-benefit approaches such as immunoinformatics tools in vaccine development. Many studies have reported that immunoinformatics, reverse vaccinomics, or computational immunological approaches are reliable, accurate, quick, and cost-effective methods, with a broad collection of available and powerful tools for epitope-based vaccine design and vaccine development [26‐30]. Therefore, the present study is the first attempt to use immunoinformatics tools for epitope mapping analysis of the N-terminal sequence of all five PspA clades. Then we designed and constructed the PspA_1-5c+p vaccine candidate and evaluated the PspA_1-5c+p protection against pneumococcal infection by immunization of mice with recombinant PspA_1-5c+p. We also evaluated the cross-reactivity ability of the anti-PspA_1-5c+p antibody against pneumococcal strains representing both PspA families and the functional activity of the anti-PspA_1-5c+p antibody. The findings suggest the potential use of this vaccine candidate as a novel serotype-independent PspA-based pneumococcal vaccine with a strong cross-reactivity response. The schematic procedure of this research has been shown in Fig. 1.

A successful serotype-independent PspA-based vaccine against pneumococcus is a vaccine that comprises multiple immunological surface components with high cross-reactivity feature including various N-terminal domains of the PspA families [24, 66]. So, the purpose of designing this study was first to improve the PspA-based vaccine potency and efficacy using immunoinformatics tools as the first line of vaccine design. Then make experimentally a new construct with highly conserved and variable regions with high antigenic binding epitopes of B- and T-cell, with emphasis on cross-reactive regions of PspA N-terminal to evaluate the covering immune response against pneumococcal PspA clades.

To our knowledge, this is the first study on the immunoinformatics-based design of PspA families-based vaccines. We first modeled, refined, and validated the 3D structure of five different PspA clades using computational approaches. Then, we also predicted linear and conformational B-cell, and T-cell epitopes, especially at cross-reactive regions of PspA families 1 and 2 using various databases. According to B-cell epitope prediction servers (BCPred, IEDB, and Ellipro), the cross-reactive regions of each clade were analyzed. These regions had at least five B-cell epitope sequences of ~ 6–25mer in length with VaxiJen scores of 0.5 to 1. Some of the predicted epitopes had antigenicity scores of 2 to 3.3. The antigenicity score, surface accessibility, flexibility, hydrophilicity, beta-turn features, and conformational B-cell epitopes of predicted immunodominant regions were also considered suitable for designing cross-reactive PspA-based vaccines. As not all antibodies against PspA are protective; therefore, understanding which epitopes can elicit a protective response is critical [67]. So epitope mapping of PspA is one of the most widely used methods for identifying these epitopes. McDaniel et al. (1994) showed that the protection-eliciting regions of PspA were localized at 192–260 amino acid regions of PspA from the strain Rx1 using four of the nine monoclonal antibodies [68]. In accordance with McDaniel, we analyzed the B-cell epitopes of strain Rx1 as a clade 2 using immunoinformatics databases, and our results showed that the most predicted epitopes were located in the predicted region by McDaniel, which has thirteen sequences of 7–104 mer in length with VaxiJen scores of 0.5 to 1. Findings from McDaniel’s experimental study have the potential to support our predictions. Therefore, we also used this immunoinformatics prediction method for epitope mapping of other PspA clades. In agreement with Singh et al. [69] we used MHC-II binding epitopes prediction servers. Then the predicted epitopes for strain 435/96 (clade1) were compared with the predicted epitopes by Singh and showed almost similar results that could be considered for inducing IFN-γ and IL-4 production. So, we analyzed other PspA clades for MHC-II binding epitope prediction. Some predicted HTL epitopes were also predicted as B-cell epitopes, so we selected the immunodominant truncated CDR regions of each clade.

Mukerji et al. classified the proline-rich domain (PRD) of the PspA into three relatively distinct groups [21]. On the other hand, these PRD regions, especially the PKPEQP motif and non-proline block (NPB) sequence, can elicit protection against pneumococcal infection. As Daniels et al. indicated that when mice were immunized by group 2 PRD, they have been shown protection against the challenge test by a pneumococcal strain with group 3 PRD. These results showed cross-protection against epitopes shared by different groups of PRD [21, 70]. PRD group’s motifs have also been reported to be linear epitopes, and human antibodies can recognize all three PRD groups [21]. In completing the study by Mukerji et al. [21], we used the repetitive motif sequences from all three PRD groups and NPR sequence as a highly conserved and immunogenic domain in PRD of PspA to cover all diversity and cross-protection of the PRD groups. Finally, the designed construct has been named PspA_1-5c+p representing the CDR and proline regions of five PspA clades. We used the rigid linker EAAAK between each truncated domain to make the least interaction between domains and maintain the best three-dimensional structure and accessible B-cell conformational epitopes. According to the literature, many natural linkers have alpha-helical structures, which are stable and rigid spacers to keep a fixed distance that is used for separating the functional domains. Another advantage of rigid linkers compared to flexible linkers is that the flexible linkers lead to low expression yields with loss of biological activity[49].

The designed PspA_1-5c+p construct was assessed for its physicochemical characteristics. The PspA_1-5c+p construct was expected to be acidic in nature, depending on the theoretical isoelectric point. The aliphatic index (indicating thermostability) and grand average of hydropathicity (GRAVY) were estimated at 82.23 and -0.997, respectively. The negative GRAVY value means that the protein has a hydrophilic nature and may interact with water molecules. The in-vivo half-life, as an estimation of time for destroying half the amount of protein after synthesis in the cell, was estimated at 30, 20, and 10 h in mammalian, yeast, and E. coli, respectively. Although the instability index was computed at 40.12, which categorizes the protein as unstable (II of > 40 indicates instability), the experimental result of the recombinant PspA_1-5c+p expression and purification showed that this protein was stable [51]. The molecular weight of the PspA_1-5c+p construct was 67.93 kDa. The SDS-PAGE and western bot results of recombinant PspA_1-5c+p expression confirmed the estimated molecular weight of PspA_1-5c+p construct. It has been reported that proteins with a molecular weight of less than 100 kDa are suitable for vaccine design due to their easy expression and purification steps [46]. Therefore, this designed protein was an acceptable vaccine candidate. Also, the results of the codon adaptation index (CAI) and GC content of 0.84 and 42.97%, respectively, showed a good efficiency of the final vaccine transcription and translation in the E. coli host. So that, the Gibbs free energy after sequence optimization for PspA_1-5c+p construct mRNA was -445.5 kcal.mol⁻¹, showing the lowest free energy and stable structure, and there was no observed unsuitable pseudo-knot or loop at 5’of mRNA. These computational results were confirmed with the expression of PspA_1-5c+p in E. coli BL21 using 1 mM IPTG. The presence of a 67 kDa recombinant PspA_1-5c+p sharp band by 12% SDS-PAGE showed acceptable expression and codon optimization. The computationally predicted overexpression and soluble feature of PspA_1-5c+p using the SOLpro prediction were validated by purification of recombinant PspA_1-5c+p under the native condition in Ni–NTA affinity chromatography with a high concentration of 0.8 mg/ml. The expression of the PspA_1-5c+p construct was verified by the Western blot on PspA_1-5c+p using an anti-His tag antibody.

Understanding the secondary and tertiary structures of the target protein is critical to vaccine design. The secondary structure of PspA_1-5c+p contained 83.22% alpha-helix, 0.49% extended strand, and 16.28% random coil using the GOR V prediction server. It has been reported that the important shapes of “structural antigens” are natively unfolded protein regions and alpha-helical coiled-coil peptides. Both structural forms can be retreated into their native structure and therefore be identified by antibodies naturally induced in response to infection [55]. The PspA_1-5c+p 3D structure was modeled using the I-TASSER server. This server is one of the best and most widely used servers for designing three-dimensional protein structures. I-TASSER server uses the multiple threading alignments from PDB to identify structural templates and designs the 3D structures using repetitive fragment assembly simulations [71]. According to many recent papers which have cited to I-TASSER server for protein 3D structure predictions [72‐75], the I-TASSER server was ranked as the No 1 server for protein structure prediction in recent community-wide CASP7, CASP8, CASP9, CASP10, CASP11, CASP12, CASP13, and CASP14 experiments. It was also ranked as the best for function prediction in CASP9. The server is in active development with the goal to provide the most accurate protein structure and function predictions using state-of-the-art algorithms [36]. Using structural refinement servers, we could improve the overall quality factor of the initial PspA_1-5c+p 3D model predicted by I-TASSER from 89.66% to 98.14%, and in the Ramachandran plot, disallowed region residues were reduced from 1.6% to 1.2% after the refinement process. Ramachandran plot also revealed that most of the residues are located in the favored and allowed regions (98.8%), demonstrating that the overall model quality is satisfactory. The structural refinement servers optimized the hydrogen-bonding network, minimized the atomic energy of the model, and improved the 3D structure by molecular dynamics simulation. In this study, the MD simulation was applied to verify the stability and flexibility of the structure of the designed PspA_1-5c+p protein. Analysis of the MD simulation trajectory revealed that the designed structure of the PspA_1-5c+p reaches a stable state with low deviations from 50 to 85 ns. This can indicate the stability of the 3D structure. In addition, using the RMSF plot, we found that the C-terminal of PspA_1-5c+p protein is the fluctuating region of the protein. During the simulation, the fluctuation of this region occurred around 0.7 nm. Nonetheless, the rest of the protein had a fluctuating value of less than 0.35 nm. These residues (C-terminus region of the protein) have more freedom of action in the environment due to the coil structure. Furthermore, the ClusPro and DimPlot results of PspA_1-5c+p and HLA-DRB1*01:01 (the most common binding allele in the Iran population [46]) docking complex showed the lowest energy binding of -744.3 kcal.mol⁻¹ and 64 cluster members indicating good binding affinity and coupling of this protein with human MHCII via sixteen hydrogen bonds and six salt bridges. However, in order to improve and examine the precise interaction between the protein and HLA-DRB1*01:01, the docking between the T-cell epitope placed in the groove of HLA-DRB1*01:01 chains with the T lymphocyte receptor (TCR) [76] or docking of the human ternary complex of the T-cell receptor, peptide-MHCII molecule, and CD4 are recommended [77]. Since the PspA has a lactoferrin binding domain in the CDR region[59], to furthermore validation of the 3D structure of the modeled PspA_1-5c+p, we docked the PspA_1-5c+p protein with human lactoferrin N-lobe (HLF). We demonstrated the PspA_1-5c+p protein can be attached to HLF molecules effectively via both regions representing PspA Families 1 and 2 in PspA_1-5c+p protein with the lowest energy binding of -1128.9 and -987.2 kcal.mol⁻¹ and maximum cluster members of 58 and 80, respectively. In coordination with the docked control molecule (PDB id: 2PMS) [59], in two models of the PspA_1-5c+p-HLF docked complex, most residues of HLF that have been in contact with CDR residues include Arg4, Arg5, Arg25, Arg28, Arg31, Arg40, Gln14, Gln24, and lys39. It has been reported that the negatively charged surface of PspA helices can interact with the highly cationic lactoferricin moiety of lactoferrin and inhibit its bactericidal effect against pneumococci. Our results were in line with the study conducted by Senkovich et al. and could show a good 3D structure of the CDR region in the PspA_1-5c+p vaccine formulation that could bind to HLF correctly. Senkovich et al. also suggested that inhibition of this interaction using small molecules or antibodies may permit lactoferrin’s natural bactericidal effects to preserve the host from pneumococcal colonization and infection and can be used for designing therapeutic strategies for the prevention and treatment of pneumococcal diseases [59]. Therefore, further studies can be performed to evaluate the binding of antibodies generated against PspA_1-5c+p to PspA on the surface of the different pneumococcal strains in the presence of the labeled human lactoferrin.

The results of predicting the conformational B-cell epitopes of the PspA_1-5c+p construct showed that after designing the structure, the conformational B-cell epitopes of each clade with a score of > 0.5 could be identified by the ElliPro server. These results can be indicated by the high potential of the PspA_1-5c+p to stimulate humoral immunity with the help of antibodies. One of the first steps in confirming a vaccine candidate is immunoreactivity detection using the serological test. According to the antigenicity score of 0.77 for the final PspA_1-5c+p construct from the Vaxijen server, this protein was considered a good antigen to stimulate the immune system. The experimental results confirmed and validated the computational antigenicity analysis of this protein. This protein was able to raise anti-PspA_1-5c+p IgG titers in immunized mice with PspA_1-5c+p construct compared to the control group (p < 0.0001) at different times of administration (Fig. 12). In addition, using immunoinformatics predictions, PspA_1-5c+p was considered a non-toxic, and non-allergen. So that, in the experimental results, this protein provided a very good and effective immunological response without causing any allergenicity or toxicity in the animal model. So that, after injection of the PspA_1-5c+p construct, we did not observe any increase in body temperature, weight loss, allergic reaction, sensitivity, or restlessness in the animal model. As in past studies conducted on the PspA protein, there were no reports of any deleterious nature of PspA. Sanofi Pasteur has also studied phase 1 of the clinical trial of PspA [9, 74, 75]. In this study, we demonstrated that anti-PspA_1-5c+p IgG reacted strongly with no significant difference (p-value = 0.2) against the surface of all three pneumococcal strains representing both PspA families. These results can indicate the high coverage of the cross-reactivity and binding ability of the anti-PspA_1-5c+p IgG among different used PspA clades, and cover the limitation of different cross-reaction levels in the PspA-based construct designed so far. In this context, Akbari et al. demonstrated that an antibody against the PspAB_1-5 antigen containing the single B region from all clades compared to PspA₄ABC could increase the cross-reactivity against pneumococcus strains representing Clades 1, 2, and 5. However, the strong binding ability of the anti-PspAB_1-5 antibody was against strain ATCC 6305 (Clade 2) with an optical density of ~ 2.1. Although, they suggested that for the construction of a PspA-based vaccine, the B region from all clades should be included [7] but is not sufficient due to the significant difference observed between optical densities of the cross-reactivity ability of the anti-PspAB_1-5 antibody against all three pneumococcus strains [7]. In this study, no different cross-reactivity ability of the anti-PspA_1-5c+p antibody was seen against two PspA families. In contrast to Akbari et al., our whole-cell ELISA results showed the optical density of the cross-reactivity ability of the anti-PspA_1-5c+p antibody was the same between three stains (Clades 1, 2, and 5) and increased to 3. This increase in the tendency of anti-PspA_1-5c+p antibody to bind to the bacterial surface may be due to two factors: the use of all cross-reactive truncated domain of CDRs together with highly conserved NPB region and using repetitive proline-rich motifs that cover the diversity of each clade. This study was also able to solve problems related to cross-reactivity differences in the studies of other research that used the various recombinant PspA proteins consisting of N-terminal and proline-rich regions from two PspA families or each region alone [23, 25, 78‐80].

We also applied the Opsonophagocytosis test to assess the in vitro potential protective effects of PspA-based vaccines against pneumococcus strains representing both PspA families. The gold standard in-vitro test for assessing the polysaccharide-base pneumococcal vaccine effectiveness is the Opsonophagocytosis assay [16, 81]. Opsonophagocytosis is thought to be considered an important function in the host defense for the elimination of pneumococci. This process is started by complement activation in the presence of antibodies that are attached to the surface of pneumococci. Then, using phagocytic cells, pneumococci are swallowed and killed [16, 81]. The results showed that the anti-PspA_1-5c+p antibodies act as a good opsonin for killing pneumococcal strains and can attach to the native protein from each PspA clade on the surface of pneumococcal strains representing both PspA families.

We also analyzed the complement-mediated killing activity of anti-PspA_1-5c+p antibody as a serum bactericidal assay against three strains of pneumococcus, expressing two PspA families. The highest bactericidal activity was detected at a 1:4 dilution in order to kill more than 50% of pneumococci compared to the control group. No significant difference was seen between the antibody’s activities against three strains of pneumococcus. These results suggest that this antibody not only has a high titer with strong and uniform cross-reactivity coverage against three pneumococcal strains but also has high bioactivity for pneumococcal clearance using complement or phagocytic cells. Goulart et al. reported that the level of complement-mediated antibody-dependent phagocytosis depends on the similarity between anti-PspA antibodies and PspA that are expressed on the pneumococcal surface [80].

In addition, according to immune simulation servers, PspA_1-5c+p was predicted to compose antigenic regions that have the potency to induce IL-4 and IL-10 cytokines. Furthermore, the PspA_1-5c+p construct was predicted to have many IFN-γ inducing MHC class II binding peptides throughout its sequence. These bioinformatics results showed that PspA_1-5c+p might induce both humoral and cellular immune pathways. Overall, these results show the success of the bioinformatics tool in designing a PspA-based vaccine candidate to cover the cross-reactivity of the vaccine candidate against all used PspA clades. As the efficacy and reliability of the immunoinformatics approach have been proven in a lot of pioneering work regarding the design and development of epitope-based vaccines [26, 27, 47, 82, 83]. Our results are in accordance with these studies. In this context, Ahmadi et al. designed a novel Hla-MntC-SACOL0723 fusion protein using immunoinformatics tools. They then showed that this fusion protein could elicit high specific IgG titer with high opsonophagosytosis’s killing activity against S. aureus resulting in a decrease in the bacterial burden in the spleen and kidneys [47]. Hasanzadeh et al. also demonstrated that the computational design of their epitope‑based vaccine candidate could induce immune responses and provide high potency in the protection of the urinary tract against uropathogenic Escherichia coli (UTEC) [83].

The limitation of this study was the lack of access to standard pneumococcal strains expressing other clades of PspA for assessing the full cross-reactive feature of the anti-PspA_1-5c+p IgG. In the future, we will resolve the mentioned limitation and also analyze the profile of subclasses of specific IgG1 and IgG2a against PspA_1-5c+p construct immunization, levels of the IL-4 and IFN- γ cytokines, and the protection ability of this construct in immunized groups against pneumococcal infections to confirm our computational immune simulation results.

Our experimental data revealed that immunoinformatics helps us to design protective serotype-independent vaccine candidates. Experimental assessments on three clades of PspA showed promising results with a strong cross-reactivity feature that should be further investigated in vitro and in vivo experiments with other pneumococcal clades to confirm the full cross-reactivity and cross-protection.