In silico analysis and in vivo assessment of a novel epitope-based vaccine candidate against uropathogenic Escherichia coli

Uropathogenic Escherichia coli (UPEC) are common pathogens in urinary tract infections (UTIs), which show resistance to antibiotics. Therefore, there is a need for a vaccine to reduce susceptibility to the infection. In the present study, bioinformatics approaches were employed to predict the best B and T-cell epitopes of UPEC virulence proteins to develop a multiepitope vaccine candidate against UPEC. Then, the efficacy of the candidate was studied with and without Freund adjuvant. Using bioinformatics methods, 3 epitope-rich domains of IutA and FimH antigens were selected to construct the fusion. Molecular docking and Molecular dynamics (MD) simulation were employed to investigate in silico interaction between designed vaccine and Toll-like receptor 4 (TLR4). Our results showed that the levels of IgG and IgA antibodies were improved in the serum and mucosal samples of the vaccinated mice, and the IgG responses were maintained for at least 6 months. The fusion protein was also able to enhance the level of cytokines IFN.γ (Th1), IL.4 (Th2), and IL.17. In challenge experiments, all vaccine combinations showed high potency in the protection of the urinary tract even after 6 months post first injection. The present study indicates that the designed candidate is able to evoke strong protective responses which warrant further studies.


Result immuno-informatics analyses.
Defining linear B-cell epitopes. Since B-cell epitopes have an important role in humoral responses, full-length sequences of FimH and IutA were subjected to linear B-cell epitope prediction. Twenty mer epitopes with a cutoff more than 0.8 were selected using BCPred and IEDB servers. IEDB showed several continuous predicted epitopes in both FimH and IutA proteins. Therefore, the regions containing the highest number of epitopes predicted by two servers were selected to increase the accuracy of prediction (Table 1).
Defining T-cell epitopes. According to multiple alleles of MHC-I and II, high score predicted epitopes were determined for both proteins. Sixteen 9-mer epitopes were predicted as MHC-I binders, and 21 epitopes with high binding affinity score of < 50 IC 50 nM were predicted as MHC-II binders. The selected human and mouse MHC-I and MHC-II epitopes are summarized in Table 2 and the output data of each tool are provided in Supplementary Table S1-S4.
Designing of epitope-rich domains for final vaccine construct. Based on immuno-informatics analysis, B-cell, MHC-I, and MHC-II epitopes with the highest scores were selected. The final vaccine construct was made with     3D structure prediction and validation. Five 3D models of the designed vaccine were modeled by I-TASSER, and those with maximum C-score were selected. Template modeling score (TM score) of the best model was 0.74 ± 0.15 that indicated the structural similarity based on the structural alignment. According to ProSA, Z-score of the selected model was − 2.02, which was in the range of scores for native proteins with similar sizes (Fig. 2). Ramachandran plot showed that 71.6% of residues were located in the favored regions. Refinement of the predicted structure improved the allowed amino acids, and Z-score of the refined model to 92.30% and − 2.8, respectively. In Molprobity analysis, all-atom clash and Molprobity scores were calculated to be 8.93 and 2.01, respectively. In addition, the expected RMSD was calculated to be 3.51 by TM-align (Fig. 2B). The 3D model of final construct was visualized by PyMOL Viewer (Fig. 2C).
Prediction of discontinuous B-cell epitopes. According to the important role of conformational epitopes in humoral response, conformational B cell epitopes of the designed sequence were predicted on the basis of protein-antibody interaction. Discontinuous peptides with a score value of 0.7 or more were selected, and the scores showed surface protein atoms responsible for binding to antibodies. The compositions of amino acids, the number of amino acids, sequence location, as well as the score values are summarized in Table 3. Furthermore, 3D representation of the predicted discontinuous epitopes of the final protein is shown in Supplementary Fig. S1.
Molecular docking of TLR4-protein vaccine. The best docked model was selected based on the total free energy (− 1829.9) and best interaction tendency to TLR4 (Fig. 3A). Hydrophobic interactions, as an important factor for initiating innate immune responses, were dominant signs for the interaction between TLR4 and the designed vaccine (Fig. 3B).
Molecular dynamic simulations. The molecular dynamics (MD) simulation showed that the potential energy of the simulated system remained stable during the simulation time. The RMSD plots for TLR4 and the designed construct are shown in Fig. 3C and confirmed the stability of the designed construct in interaction with TLR4.
Sequence analyses. The results of homology alignment of the selected epitopes in FimH and IutA with those deposited in National Centre for Biotechnology Information (NCBI) showed that these sequences were conserved among UPEC strains (> 97% for FimH and > 90% for IutA) (Supplementary Figs. S2 and S3).
Codon optimization and cloning. The codon optimization of the DNA sequence to the codon usage of E. coli was carried out by Biomatik Company and OPTIMIZER server. The gene was cloned into pET28a expression vector using NcoI and HindIII restriction sites with a poly histidine-tag (6x-His tag) at C-terminus of the protein.
The fidelity of cloning was finally verified by gel electrophoresis, PCR, restriction map analysis (Fig. 4A), and sequencing.
Purification and confirmation of the recombinant protein. The protein expression was confirmed by SDS-PAGE and Western blot. The SDS-PAGE (Fig. 4B) and Western blot showed a 45 kDa band as expected by Mw calculations. The recombinant protein was purified from E. coli DE3 lysates by applying onto Ni-NTA affinity column. The purified protein was verified by SDS-PAGE (Fig. 4C) and Western blot analysis (Fig. 4D). The LAL test indicated a low amount of LPS (< 0.5 EU/ml) in the protein solution.
Determination of IgG responses in human cases. The levels of IgG response against the recombinantly produced protein construct were measured in humans who had been previously infected with a pathogenic UPEC. Those infected showed a statistically significant higher level of IgG as compared to the humans that had no history of infection with a UPEC strain (p < 0.05) (Fig. 5).
Humoral immune responses in the immunized mice. The mice groups that received fusion protein alone or with Freund adjuvant showed higher IgG responses compared to the control groups that received just PBS or Freund. Moreover, data analysis indicated that adding the Freund to fusion protein resulted in a statistically significant higher IgG response than inoculating the protein alone (Fig. 6A).
The evaluation of longevity of humoral responses revealed that mice vaccinated with fusion protein, and fusion plus Freund could significantly enhance IgG responses after the second immunization until day 45, and the immune responses they had induced remained steady up until 6 months post first vaccination. Overall, IgG responses induced by fusion and fusion plus Freund were significantly higher than those in control groups in all weeks. Moreover, the induced IgG response by fusion and fusion plus Freund was statistically different throughout the six months after the first vaccination (Fig. 6A).  Mucosal antibody responses. Mucosal responses were determined in urine samples collected 2 weeks after the last vaccination. According to the results shown in Fig. 7, mice vaccinated with both fusion and fusion plus Freund could enhance the IgG levels in comparison with the control groups. We found that Freund adjuvant could not significantly raise the IgG level of the fusion protein as compared to the fusion alone group. It was also observed that the vaccine formulations, including fusion and fusion plus Freund could not significantly enhance the mucosal IgA levels when compared to the control groups ( Fig. 7).

Measurement of cytokine responses.
According to the ELISA data, higher levels of IFN.γ, IL.4, and IL.17 cytokines were measured in supernatant of the splenocytes isolated from immunized groups as compared to control groups (p < 0.001). These findings showed that immunization with fusion plus Freund significantly increased the IFN.γ levels compared to the fusion protein alone (p < 0.001). While there was no statistical significant difference between the production of IL.4 and IL.17 in mice immunized with the fusion plus Freund, and the fusion alone (Fig. 8).
Vaccine efficacy against kidney and bladder challenge. Challenge experiments were performed using UPEC strain CFT073 48 and 180 days after the first vaccination on the mice that were immunized with three doses. As a result of the first challenge, the mice groups immunized with the fusion or the fusion plus Freund showed a significant reduction (10 2 -10 3 folds) in bacterial loads in both bladder and kidney organs as compared to the control groups (p < 0.05) (Fig. 9A1,B1). We also observed that there was no significant decrease in the levels of UPEC in the bladder and kidney of the mice that received the fusion alone and the fusion plus Freund. In addition, the results of the challenge after 6 months indicated that the bacterial levels in the vaccinated groups were significantly reduced (10 2 -10 3 folds) in the bladder (p < 0.01) and kidneys (p < 0.01) as compared to the control groups ( Fig. 9A2,B2).

Discussion
Immunotherapy is the most potent way to prevent infectious diseases such as UTIs caused by UPEC strains. Safety, immunogenicity, and induction of protective immunity against a broad spectrum of UPEC strains are the ideal criteria of a vaccine against UPEC 24 . Bioinformatics programs could define immune dominant B and T cell epitopes of antigens that play important roles in pathogenicity, and induction of immune responses 25,26 . The present study aimed to design and construct a novel multi-epitope protein from UPEC strain based on in silico methods to evoke protective immune responses.
Along with all the positive features of poly epitope vaccines, the main problem is their low immunogenicity. The usage of epitope rich domains is one of the strategies to overcome this problem which was applied in this construct 15 . These domains could induce robust T and B-cell immune responses without having the limitations of single epitopes 15 . Due to the importance of T-cell mediated immunity in the levels of B-cell responses in the protection against UPEC infections 17 , we decided to include MHC-I epitopes in the construct together with the MHC-II epitopes.
Given the importance of linkers in designing of vaccine candidates, we decided to apply a peptide epitope for T-helper cells instead of routine linkers, derived from HLA PADRE. In fact, PADRE has the capacity to evoke effective CD4 + T cell responses in inducing both the humoral and cellular responses 27,28 . Previous studies showed that FimH could induce innate host responses by interaction with TLR4 expressed on immune cells 6,23 . Tolllike receptors are among the pattern recognition receptors (PRRs) that can induce different aspects of immune responses such as secretion of pro-inflammatory cytokines 29 . Although the best linear B-and T-cell epitopes of FimH were located in amino acids 25-160 (Tables 1, 2), the results of defining the conformational epitopes showed that amino acids 166-168 of FimH were also among the selected discontinuous epitopes. Furthermore, the construct composed of amino acids 25-168 in FimH showed better confidence score and stability than the construct made of residues 25-160 in FimH. Finally, based on the linear and conformational epitopes and also evaluation of structures, we decided to design a construct which composed of two domains of IutA (336-377 and Table 3. Conformational epitopes of the designed protein as predicted by ElliPro.
It was found that the humans who were infected with the UTIs caused by UPEC strains developed humoral response against the vaccine candidate. The fact that humans infected by UPECs managed to respond to the vaccine protein showed that the predicted B cell epitopes, at least some of them, were indeed identified by clinical samples validating the predictions.
The use of a proper adjuvant is another strategy to promote the immune responses and direct them to a desired direction 30 . We decided to use a well-established adjuvant such as Freund to examine how much it improved the immunogenicity of the designed candidate. Freund has been used as one the most effective adjuvants for production of high titers of antibodies 31 , and induction of cell-mediated immunity (Th1) 32 . Our www.nature.com/scientificreports/ findings showed that the fusion protein without adjuvant could evoke significant humoral (IgA, IgG1, and IgG2a) and cellular (IFN.γ and IL.4) responses, demonstrating the ability of the candidate in raising a mixed response of type 1 (Th1) and type 2 (Th2). Freund could significantly increase the levels of total IgG and IFN.γ (Th1) as compared to the fusion alone. It should be considered that the high levels of Th1 could be effective for the eradication of intracellular reservoirs of UPEC in the bladder. It is also possible that there was a correlation between the production of IL.17 and IFN.γ, since secretion of IL.17 has helped to induce the high titer of IFN.γ by the vaccine combinations 33,34 . Unlike some studies indicating mucosal antibody responses were not increased in subcutaneous immunization 35 , our study demonstrated that mucosal IgG response was induced in the subcutaneous route of administration. Wiser et al. also reported that the use of different epitope rich domains in multi-epitope vaccines from subcutaneous route had the potential to induce IgA responses 15 .
In the bladder challenge, we found that all vaccinated mice significantly reduced bacterial levels in the bladder two days after bladder infection, with a trend towards the reduction of bacterial load in kidneys. Among the candidates, the fusion admixed with Freund showed the highest protection in the bladder and kidneys against E. coli. Interestingly, this vaccine combination revealed the full protection against the kidney infection in 40% of the immunized mice after the first and second challenge assay. It was found that the humoral and cellular (IFN.γ and IL.17) responses induced in the full protected mice (2 of 5 mice) were in the highest levels compared to the other immunized mice. Therefore, the high levels of protection among the mice could be attributed to the humoral (mucosal or systemic) or cellular responses or a mixture of both 33,36 . There may also be a direct relation between the decrease of UPEC levels recovered from the bladders and decrease of infection in the kidneys, thus the reduction of UPEC infection in the bladder may have resulted in the kidney protection.
Memory B cells are essential in the prevention against UTIs, especially recurrent UTIs 13 . In our study, the immunized groups with and without Freund maintained high levels of the IgG response until 6 months after the first vaccine dose. In addition, the repetition of the mouse challenge model after 6 months post first immunization confirmed the efficacy of sustained responses in the protection of urinary tract against UPEC. This finding was one of the prominent features of our designed vaccine candidate which could be effective in the eradication of bacterial reservoirs for a long time.
Our study has a limitation. This study was the first step of developing a novel vaccine against UTIs caused by UPECs. In this step we assessed the efficacy of the vaccine only with UPEC strain CFT073 and it was not possible to evaluate its efficacy in mice with clinical UPEC strains. We are going to evaluate the efficacy of this vaccine with different UPEC strains in the future.
In the present study, we designed and constructed a novel multi-epitope vaccine candidate from UPEC strain based on in silico methods. This study showed the potential of the candidate in the induction of immune responses, and the protection of urinary tract against UPEC. Therefore, our designed candidate could be a promising candidate against UTIs caused by UPEC which requires further investigations. Our results could also be useful in gaining insight towards the potential of epitope-based construct as an important protective and therapeutic approach for bacterial immunization.  Primary sequence analysis and domain identification. IutA (Accession No. AAN82071.1) and FimH (Accession No. AAN83822.1) sequences from UPEC strain CFT073 were retrieved from National Centre for Biotechnology Information (NCBI) (www.ncbi.nlm.nih.gov). TM BETA software was used to identify the extracellular domains of proteins (https ://psfs.cbrc.jp/tmbet a-net/).
Prediction of MHC-I binding epitopes. As A*0201 is the most common MHC-I bound allele in human 40 , this allele was chosen for epitope prediction. MHC-I binding epitopes in BALB/c mice were also considered in the www.nature.com/scientificreports/ prediction to have strong T cell responses in mice. 9mers epitopes were selected, because most HLA molecules have strong preferences for binding to 9mers 41 . Three different algorithms were employed for prediction of HLA-I epitopes, including SYFPEITHI (cut off score > 20) 42 , Net MHC 4.0 (an artificial neural network), and RANKPEP which utilize a position specific scoring matrix (PSSM) to predict MHC-I restricted epitopes 43 .
Prediction of MHC-II binding epitopes. Nine human HLA-II super type alleles, as well as Ad and Ed mouse alleles were chosen for MHC-II epitope prediction. IEDB server (https ://tools .immun eepit ope.org/mhcii /) was used to predict MHC-II epitopes using NetMHCII and Consensus methods 44 . RANKPEP and SYFPEITHI servers were used for identification of mouse MHC-II epitopes.
Vaccine features. Protein's tertiary structure prediction, refinement, and validation. I-TASSER (https :// zhang lab.ccmb.med.umich .edu/I-TASSE R) that identifies similar structure templates from the Protein Data Bank (PDB) was used to predict the tertiary structure of the designed sequence 49 . Then, the best-modeled structure was refined by GalaxyRefine (https ://galax y.seokl ab.org/cgi-bin/submi t.cgi?type=REFIN E) 50 . The refined model was finally analyzed using protein structure analysis (ProSa) (https ://prosa .servi ces.came.sbg.ac.at/prosa .php) 51 , Molprobity, and Ramachandran Plot Analysis resource (RAMPAGE) 52 . Analysis of amino acid sequences. The amino acid sequences of FimH and IutA used for construction of the poly-epitope were aligned with the sequences of UPEC strains available in NCBI and analyzed with BLAST (www.ncbi.nih.gov) tool.

Molecular interaction studies of
Codon optimization and cloning of the designed gene. The designed construct was translated reversely to DNA sequence. After codon optimization and adding a 6 His-tag to C-terminus of the sequence, the construct was cloned into pET28a expression vector (Novagen, USA) by Biomatik Company (Canada). Expression of the recombinant protein was induced by adding different concentrations of isopropyl-beta-thio galactopyranoside (IPTG) and assayed using SDS-PAGE and Western blot analysis using His tag monoclonal antibody (Sigma, USA) at dilution of 1:1000. The expressed protein was then purified using His tag affinity chromatography on Ni-NTA column (Qiagen), according to the protocol. Limulus amebocyte lysate (LAL) test was performed to determine the level of LPS contamination. Finally, the purified protein was dialyzed overnight and quantified using Bradford method.

Assessment of IgG responses in human.
After purification of the recombinant protein, possible IgG response against the protein was assessed in serum of previously UPEC-infected people (n = 10). An ELISA procedure was used for this, testing serial serum dilution from 1:10 to 1:1000 as described in the next section. Human sera from non-infected people were used as negative control. Assessment of humoral immune responses in mice. The antibody responses including IgG, IgG1, IgG2a, and IgA were quantified by enzyme-linked immunosorbent assay (ELISA) as described previously 55,56 . Briefly, purified fusion protein at a concentration of 10 μg/ml (1 μg/well) was coated onto microtiter plates (Greiner, Germany), and incubated overnight at 4 °C. The boundless positions in plates were blocked with 3% bovine serum albumin (BSA) (Sigma, USA) for 2 h at 37 °C, followed by 4 times washing with PBS-T buffer (PBS1x + Tween 20) and Statistical significance of the differences between the mice groups were determined by Kruskal-Wallis analysis (Dunn's multiple comparison test) and are shown in between brackets with asterisks or P values. The single asterisk shows statistical significance of control mice over vaccine combinations (p < 0.05). Double asterisks indicate statistical significance of control mice over vaccine formulations (p < 0.01).
Cytokine assay. For cytokine assessment, mice (n = 5) were sacrificed 45 days after the first vaccination and the spleens were isolated and homogenized. Then, the homogenized splenocytes were cultured at a density of 3 × 10 5 and stimulated with fusion protein (10 μg/ml) to collect the supernatants after 3 days. Finally, the levels of IFN.γ, IL.17, and IL.4 were measured by the DuoSet ELISA kit (R&D Systems, USA).
Bladder challenge assay in the immunized mice. Bladder challenge assay was performed on days 48 and 180 after the first vaccination to evaluate the efficacy and sustainability of developed immune responses. Five mice out of 15 immunized mice were used for the first challenge and the remaining 5 mice were kept until day 180 were used for the second challenge assay. In brief, E. coli strain CFT073 was introduced transurethrally by micro catheter (B&D, USA) into the bladder of anesthetized mice with a mixture of ketamine and xylazine (70 mg/ kg + 5 mg/kg) (Alfasan, the Netherland). After one week, kidneys and bladders of the challenged mice were isolated, homogenized, and cultured in LB agar medium to count the appeared colonies.
Statistical analysis. The statistical SPSS software (SPSS package program version 19, IBM Corporation) was used for analysis of immune responses. All experiments were conducted in three independent experiments and results were expressed as average ± S.D. Analysis of variance (ANOVA) was used for statistical analysis of the immune responses including total IgG, IgG isotypes, IgA, and cytokine levels (IFN.γ, IL.4, and IL.17), followed by Tukey or Games-Howell tests. Comparison of IgG responses between the infected humans with UPEC with humans that had no history of infection with UPEC was performed by Student t-test. Analysis of the challenge experiments was done using Prism program (GraphPad), version 6, with Kruskal-Wallis and Mann-Whitney. P value less than 0.05 was considered as a statistically significant difference between groups.