Escherichia coli M15 strain transformed with the plasmid pQE-30 containing the pspa94-pdt fusion gene from Goulart et al. (2013) was cultured in 10 L bioreactor at 30°C using 6 L chemically defined HDF medium (Seeger et al., 1995), kanamycin (25 μg/mL) and ampicillin (100 μg/mL). Induction was carried out using 1 mM isopropyl ß-D-1-thiogalactopyranoside (IPTG). Cells were harvested by centrifugation and frozen at −80°C prior to use. Before purification, cells were suspended in lysis buffer (10 mM Na-phosphate, 0.1% Triton X-100, 5 mM EDTA, 5 mM PMSF), lysed with a high pressure continuous homogenizer PandaPLUS 2000 (GEA Group, Düsseldorf, Germany), incubated with 0.1% cetyltrimethylammonium bromide (CTAB) agitating for 1 h to precipitate endotoxins (Figueiredo et al., 2017). The lysate was centrifuged and the soluble fraction was filtered. Several purification trials were carried out using anion exchange (Q-Sepharose FF), immobilized metal affinity (IMAC-Sepharose FF), hydrophobic interaction (Phenyl-Sepharose FF High Sub), and size exclusion chromatographic (Sephacryl S-200 HR) methods.

Western blot using anti-pneumococcal whole cell vaccine (PWCV) polyclonal serum (Gonçalves et al., 2014), anti-PspA94 and anti-PdT (Goulart et al., 2013) was performed with rPspA-PdT samples before (clarified lysate) and after purification. These samples were applied three times in the SDS-PAGE (10%) and transferred to the membrane, which was stained with Ponceau S (Supplementary Figure S1) and split into three parts, each one incubated with one of the antibodies. Detection was achieved with 3,3′-diaminobenzidine (DAB).

Proteolytic activity on casein was determined using Gonçalves et al. (2003) protocol. Casein (1% w/v) was used as a substrate in 0.02 M Tris-HCl buffer, pH 8.0, 0.02% sodium azide. Casein solution (0.4 mL) was incubated with 0.1 mL of sample at 37°C for 48 h. Non-hydrolyzed casein was precipitated by adding 0.4 mL of 50% (w/v) trichloroacetic acid (final concentration: 10%), tubes were kept on ice for 30 min and centrifuged at 17,000 g for 10 min; absorbance at 280 nm was measured before and after incubation. Trypsin (Sigma-Aldrich, San Louis, MO, United States) at 0.5 mg/mL and distilled water were used as positive and negative controls, respectively.

The presence of hydrolytic enzymes in the samples was also evaluated by in-gel zymography (Fernández-Resa et al., 1995; Leber and Balkwill 1997; Vandooren et al., 2013). Sodium dodecyl sulfate (SDS) polyacrylamide gels 12% were copolymerized with either 0.1% gelatin or 0.1% casein. Samples were prepared under non-reducing conditions, trypsin 0.5 mg/mL solution was used as positive control and electrophoresis was carried out for 90 min at 120 V. After the run, gels were washed twice for 30 min and twice for 10 min with 2.5% Triton X-100 solution for SDS removal, and then incubated in collagenase buffer (50 mM Tris-HCl, 200 mM NaCl, 5 mM CaCl₂, pH 7.6) for 18 h at 37°C. Gels were stained (25% ethanol, 5% acetic acid, 0.1% Coomassie Blue G-250) overnight and destained (10% ethanol, 5% acetic acid) for 2 h. Control gels were cast without substrates to show protein bands of the samples (Supplementary Figure S2).

For the detection of serine proteases, the chromogenic reagent Nα-benzoyl-L-arginine-p-nitroanilide, L-BAPNA (Hieng et al., 2004), was used as substrate, as described by Knittel et al. (2016). This substrate was prepared at a concentration of 115 mM in 50 mM Tris-HCl buffer, pH 7.5. Incubation of substrate and sample was carried out in 96-well plates containing: 160 μl of 50 mM Tris-HCl buffer pH 7.5, 20 μl of substrate solution and 20 μl of sample at different concentrations (diluted in the same buffer) for 40 min at 37°C. Bothrops jararaca crude venom (5 μg per well) and buffer instead of sample were used for the positive and negative control, respectively. After incubation, absorbance was measured at 405 nm.

For the detection of metalloproteases, Abz-AGLA-EDDnp reagent (Freitas-de-Sousa et al., 2017; Mancuso et al., 2018) was used as a substrate at a final concentration of 200 μM, according to Knittel et al. (2016). This substrate was diluted before use to 400 μM in 50 mM Tris-HCl buffer with 10 mM CaCl₂, 150 mM NaCl and 0.05% (v/v) Brij^® 35, pH 7.5. The incubation of the substrate with the sample was carried out in 96-well plates containing 50 μl of diluted substrate and 50 μl of sample at different concentrations. Absorbance was monitored for 10 min, with measurements every 1 min at 320 and 420 nm wavelengths. Jararhagin isolated from B. jararaca venom (5 μg/mL) and buffer instead of sample were used for positive and negative control, respectively. Specific activity of metalloproteinases was calculated in relative fluorescence units (RFU/min/µg).

For determination of the degradation site of the hybrid molecule, a sample of degraded rPspA94-PdT was loaded into immobilized metal affinity chromatography (IMAC) to separate both parts of the degraded fusion. The N-terminal His-tagged fragment remained bound to the resin, while the C-terminal fragments (without His-tag) were recovered in the flowthrough. Then, the flowthrough fraction was loaded into high performance liquid chromatography (HPLC) on a butyl (C4) stationary phase column. For the equilibration, 0.1% trifluoroacetic acid (TFA) solution was used and elution was carried out with the same solution and gradient from 0% to 100% acetonitrile. Elution absorbance at 280 nm was monitored in real time on a chromatogram and the major peak was manually collected. This main peak was concentrated and its N-terminus portion was sequenced using Edman degradation method (Edman et al., 1950).

Two new constructs were proposed, which would join the antigens with different molecular linkers: A flexible linker (FL), formed by glycines and serines, whose sequence is N- GGGGSGGGGS- C; and a rigid linker (RL), which forms an alpha-helix structure, whose sequence is N- AEAAAKEAAAKA -C. In addition, within the proline-rich region, the non-proline sequence block (NonPro), and the last block of prolines of PspA were removed from this new construct (Hollingshead et al., 2000), since recombinant PspA molecules that do not have this non-Pro region, such as PspA4Pro (Moreno et al., 2010; Figueiredo et al., 2017) and PspA3 (Carvalho et al., 2012), were more stable than recombinant PspA245 in whose sequence the region was maintained (Barazzone et al., 2011). Also, the new constructs were cloned into pET28a, eliminating nine pQE-30 derived amino acids from beginning of the tandem fusion protein, three (RGS) before six histidine tag (His-tag), and another six (GSACEL) after His-tag and before the first PspA amino acid.

Overlap extension PCR was carried out with high-fidelity polymerase (Q5 high fidelity DNA polymerase; New English Biolabs, Ipswich, MA, United States), where two PCR are performed in parallel to amplify the genes of each protein with a linker sequence and a third PCR in series to join the products of the previous reactions (Ho et al., 1989; Reikofski and Tao 1992; Bryksin and Matsumura 2010). The plasmid pQE-30 carrying pspa94-pdt gene (Goulart et al., 2013) was used as template to construct the two new hybrid proteins. Primers were designed to insert the linker region between pspa and pdt (Table 1). Optimal conditions for the reactions were defined after evaluation of different annealing temperatures (70°C–55°C) and different concentrations of magnesium chloride (1–4 mM). The genes were amplified in a thermocycler at the following optimal conditions: 2 min at 94°C, 30 cycles of 94°C for 30 s, 60°C for 60 s and 72°C for 90 s each and 72°C for 5 min, using 4 mM MgCl₂. PCR products were analyzed using agarose gel electrophoresis. Gel bands were purified using a GFX PCR DNA Purification and Gel Band Purification Kit (Cytiva, Marlborough, MA, United States) and nucleic acid concentration was determined using NanoDrop (Epoch Microplate Spectrophotometer, Bio-Tek).

High-fidelity polymerase used for the PCR produces fragments with blunt ends, so the PCR products were purified and an adenylation reaction with Taq DNA polymerase (GoTaq DNA polymerase, Promega) was carried out before inserting the PCR product into the cloning vector. Reaction was carried out with a final concentration of 2.5 mM MgCl₂ and 0.2 mM dATP in a thermocycler for 18 min at 70°C followed by 2 min at 72°C. Reaction product was used directly in the ligation reaction with the cloning vector pGEM-T Easy Vector (Promega, Madison, WI, United States). Competent E. coli DH5α cells were transformed with the plasmids carrying the genes of pspa-fl-pdt (flexible) or pspa-rl-pdt (rigid). With the constructs obtained, colonies were screened; restriction maps and sequencing were performed. The gene sequences were submitted to GenBank database under the accession numbers OP871266 (pspa-fl-pdt) and OP871267 (pspa-rl-pdt). Minipreparation was performed with Illustra plasmidPrep Mini Spin kit (Cytiva, Marlborough, MA, United States). Commercial plasmid pET-28a (Novagen) was digested with the enzymes NcoI and XhoI and treated with calf intestinal alkaline phosphatase (Promega, Madison, WI, United States) to remove the phosphate groups from the 5′ends and prevent their recircularization. Subsequently, cloning vectors with the inserts were digested with the same restriction enzymes and inserts were cloned into the expression vector pET-28a, providing a C-terminal His-tag to both fusion proteins. Resulting constructs pET-28a-pspa-rl-pdt (rigid) or pET-28a-pspa-fl-pdt (flexible) were respectively transformed into the expression strains E. coli BL21 (DE3) (Novagen) and E. coli BL21 (DE3) Rosetta (Novagen).

Glycerol stocks (10%) in chemically defined HDF medium (Korz et al., 1995) were prepared for each producer clone, ensuring the cells were still in exponential growth phase. Kanamycin (30 μg/mL) was added for both clones and chloramphenicol (34 μg/mL) only for E. coli BL21 (DE3) Rosetta. Main cultures were performed at 30°C in Tunair™ shaken flasks using HDF medium with glycerol as carbon source, and protein production was induced in mid-log-phase cultures with 0.4 mM IPTG. Five hours after induction, cells were harvested by centrifugation and frozen at −20°C.

Cell pellets were suspended with lysis buffer (20 mM Na-phosphate, 500 mM NaCl, 0.1%, Triton X-100, 1 mM PMSF), lysed at high pressure continuous homogenizer PandaPLUS 2000 (GEA Group, Düsseldorf, Germany) and incubated with 0.1% CTAB agitating for 1 h to precipitate endotoxins (Figueiredo et al., 2017). Both rPspA-FL-PdT and rPspA-RL-PdT were produced in soluble form, thus, soluble fraction was collected after centrifugation, pH was adjusted to 7.5 and imidazole was added to a final concentration of 1 mM. The material was then filtered (0.45 µm) prior to chromatography. Hybrid proteins were purified through IMAC with Ni²⁺ charged MiniChrom Fractogel^® Metal Chelate (Merck Millipore, Darmstadt, Germany) column in an ÄKTA avant system. Elution was carried out with 20 mM Na-phosphate pH 7.5, 500 mM NaCl, 200 mM imidazole, and elution fraction was loaded into Sephacryl S-200 (Cytiva, Marlborough, MA, United States) size exclusion column. Elution was then carried out with Na-phosphate buffer saline pH 7.4. Purified fractions were sterile filtrated, stored at −20°C, and used within 3 months for the immunizations. Total protein was quantified by DC™ Protein Assay Reagent Kit (Bio-Rad, Hercules, CA, United States) and bovine serum albumin (BSA) was used as standard. Purity was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and gel densitometry (Loccus Pix DS 5000 and LabImage 1D; Loccus, Cotia, Brazil). The relative quantity of target proteins was calculated by Eq. 1 r e l a t i v e   q u a n t i t y   % = i n t e n s i t y   o f   t a r g e t   p r o t e i n   b a n d ∑ i n t e n s i t y   o f   a l l   b a n d s   i n   t h e   l a n e   × 100 (1)

Purified samples obtained from size exclusion chromatography were used to evaluate the stability of rPspA-FL-PdT and rPspA-RL-PdT periodically at 4°C and −20°C. Aliquots were prepared for each period of time (1, 2, 3, 6, and 12 months) and evaluated by SDS-PAGE and Western blot. Chemiluminescence detection was performed with anti-polyHistidine-Peroxidase conjugated antibody (Sigma-Aldrich, San Louis, MO, United States). Additionally, high performance size exclusion chromatography (TSKgel G3000PWXL column, Tosoh Bioscience, Tokyo, Japan) was carried out to quantitatively assess the stability of the fusion proteins, using 18-month sterile filtrated purified samples stored in Na-phosphate buffer saline (pH 7.4) at 4°C or −20°C. The fusion protein peak and all other peaks of the chromatogram detected by UV₂₈₀ were analyzed and compared regarding percentage of total chromatogram area (%area).

Secondary structures and stability of purified rPspA-FL-PdT and rPspA-RL-PdT at different temperatures were verified by circular dichroism (CD) spectroscopy, using a JASCO J-810 spectropolarimeter (Japan Spectroscopic, Tokyo, Japan). rPspA-FL-PdT and rPspA-RL-PdT samples were prepared in 20 mM phosphate buffer pH 7.0. Measurements were taken from 185 to 260 nm and final CD spectra were obtained from the mean of five measurements. The fusion proteins were heated (1°C/min) from 2°C to 98°C and cooled back to 2°C (1°C/min), and their melting temperature (Tm) was calculated using Origin software (OriginLab Corporation) by fitting the sigmoid curves using Boltzmann function and finding the minimum value of its derivative at X-axis, which represents the sigmoid inflection point. Deconvolution was calculated based on the Dichroweb (Whitmore and Wallace, 2004) online database and the CDSSTR (Sreerama and Woody, 2000) algorithm.

The user-friendly Colab version of AlphaFold2 software (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb) was used to predict the molecular structure of the hybrid proteins, with and without linkers, using default MMseqs2 parameters (Mirdita et al., 2022). The predicted unrelaxed models presented as “rank 1” by AlphaFold2 were used as input in GalaxyWEB (Heo et al., 2013) for refinement (https://galaxy.seoklab.org/cgi-bin/submit.cgi?type=REFINE) and BIOVIA Discovery Studio Visualizer (Dassault Systèmes, Vélizy-Villacoublay, France) was used to visualize the refined models. GalaxyWEB output generates five new models and provides information regarding molecular dynamics; the chosen model should present the highest “GDT-HA” (global distance test - high accuracy), “RMSD” (root-mean-square deviation), “Rama favored”, and the least “MolProbity”. The refined model for each protein was used as input in PROCHECK (Laskowski et al., 1993) (https://saves.mbi.ucla.edu/) for stereochemical evaluation by Ramachandran plot and in ProSA-web (Sippl, 1993; Wiederstein and Sippl, 2007) (https://prosa.services.came.sbg.ac.at/prosa.php) for comparing Z-scores with solved protein structures. Additionally, Qualitative Model Energy Analysis (Benkert et al., 2009) (QMEAN, https://swissmodel.expasy.org/qmean/) was used for model quality estimation of each protein and for identification of potentially problematic regions in the sequences.

Pneumococcal strain A66.1 (serotype 3 carrying PspA from family 1 and clade 2) was maintained as frozen stock (−80°C) in Todd-Hewitt broth supplemented with 0.5% yeast extract (THY) with 10% glycerol. Before the experiments, the isolate was plated on blood agar for overnight growth, then cultured in THY up to OD_600nm 0.4–0.5 and harvested by centrifugation. Newly prepared stocks were frozen with 10% glycerol and serial dilutions were plated on blood agar for determining viable cell concentration by colony-forming units (CFU/mL).

This study was carried out in accordance with the rules issued by the National Council for Control of Animal Experimentation (CONCEA). The protocol was approved by the Ethic Committee on Animal Use of the Butantan Institute (CEUAIB), Permit Number: 1905090218. Six-week-old female specific-pathogen-free (SPF) BALB/c mice (5-6 animals per group), from the Medical School of the University of São Paulo (São Paulo, Brazil), were immunized subcutaneously (total volume: 200 µl) with three doses of rPspA-FL-PdT or rPspA-RL-PdT, varying amounts of protein per dose (5, 10 or 20 µg) at 15-day intervals, using sterile saline solution 0.9% and aluminum hydroxide (Alum) as adjuvant (50 µg per dose), totalizing seven groups. On the day of the immunization, frozen aliquots stored at −20°C from recently purified proteins were thawed and adsorbed with Alum before inoculating them to the animals. The same purification batch of each protein was used for the whole experiment. The adjuvant alone in saline was used as negative control. Fourteen days after each immunization, the animals were bled by retro-orbital puncture and antibody production was evaluated by enzyme-linked immunosorbent assay (ELISA). 21 days after the third immunization, mice were anesthetized through the intraperitoneal route with a xylazine/ketamine solution (25 mg/Kg of xylazine and 50 mg/Kg of ketamine) and challenged through the intranasal route with 1 × 10⁵ CFU of strain A66.1. The suspension of pneumococci in 50 μl of sterile saline solution 0.9% was inoculated into one nostril. This procedure ensures the solution reaches the lungs and does not remain only in the upper airways (Rodrigues et al., 2018). After challenge, mice were monitored twice per day, and lethargic animals with reduced mobility were euthanized with a lethal dose of a xylazine/ketamine solution (60 mg/Kg of xylazine and 300 mg/Kg of ketamine). Survival rate curves were analyzed by Log-rank test (Mantel-Cox).

ELISA was carried out in 96-well plates (Nunc MaxiSorp^®, Thermo Fisher Scientific) coated overnight at 4°C with 5 μg rPspA-FL-PdT or rPspA-RL-PdT in 0.1 M carbonate-bicarbonate buffer pH 9.6 (depending on the group analyzed). Non-coated wells were used as blocking control. Plates were incubated for 30 min at 37°C, washed three times with PBS-Tween 0.05% (PBST) and then blocked with 200 μl of 5% skim milk in PBS for 1 h at 37°C. The plates were washed three times with PBST. In the first assay (Supplementary Figure S3), pooled samples of each dose of each group were prepared, diluted in PBS with 1% BSA, with an initial dilution factor of 1:1,000 for sera after the first dose and 1:4,000 after the second and third doses. Then, 200 μl were added to the first wells and two-fold serial dilutions were carried out to the end of the plate so that every well contained 100 μl. In the second assay, a previous dilution was chosen based on titration of the pools, and the sera from all animals were individually diluted in PBS with 1% BSA and added to the plate. Animal sera were not added to the wells corresponding to the blank (PBS with 1% BSA). Plates were incubated for 2 h at 37°C. Plates were washed three times with PBST and then 100 μl of horseradish peroxidase-conjugated anti-mouse IgG (Sigma-Aldrich, San Louis, MO, United States) diluted in PBS with 1% BSA (1:5,000) were added, and then plates were incubated for 1 h at 37°C. Plates were washed three times with PBST and the development was performed with a solution of OPD (1.0 mg/mL) and H₂O₂ (5 μL/mL of 30% H₂O₂) in citrate-phosphate buffer pH 5.0 (0.1 M sodium citrate; 0.2 M monobasic sodium phosphate) for 15 min at room temperature in the dark. Reaction was stopped by adding 50 μl of 4 M H₂SO₄ and absorbance was measured at 492 nm wavelength. The same protocol was followed to analyze antibodies against individual proteins, but the coating was carried out with rPspA94 or rPdT and all groups were analyzed for each protein. In these experiments, a titration of pooled groups (Supplementary Figure S4) was also carried out first and then, each animal was analyzed individually. Statistical analysis of the specific IgG produced in each group was performed by one-way ANOVA with a Tukey’s Multiple Comparison Test.

In the first purification trial, rPspA-PdT purity achieved 85.5% after purification (Figure 1A). Fifteen days after the purification process, the sample was analyzed by SDS-PAGE and degradation of the hybrid protein was observed (Figure 1A). Thus, Western blotting was carried out using anti-PspA94, anti-PWCV polyclonal serum or anti-PdT to detect both antigens in the clarified lysate and in the purified sample where degradation was observed (Figure 1B). When the membrane was incubated with anti-PWCV, two bands at the expected molecular weights of the individual rPspA94 and rPdT proteins were observed (Figure 1B), which may indicate that the hybrid protein was breaking between the two proteins genetically linked in this construction. The lanes incubated with anti-PspA94 or anti-PdT indicated the lower band corresponds to PspA moiety and the upper band to PdT (Figure 1B). Clarified lysate lanes also showed degradation of the fusion protein, but the individual antigens were not clearly separated as on purified sample. From these results, the hypothesis that the protein was degrading during the purification process started to be investigated.

It is also important to point out that rPspA presents “apparent molecular mass” in SDS-PAGE different than expected (Yother et al., 1992), which explains why electrophoretic profile of rPspA-PdT hybrid molecules differs from their theoretically molecular mass of ∼88 kDa.

The presence of proteolytic activity was investigated by two methods in recently clarified lysate and/or a partially purified sample from a Phenyl-Sepharose chromatography (containing 20% ethanol). Casein digestion assay was carried out using partially purified sample from a Phenyl-Sepharose elution fraction containing 0.4 M NaCl, where degradation was observed 15 days after storage at 4°C. No proteolytic activity was observed in this sample (Figure 2A). The presence of hydrolytic enzymes in the samples was also evaluated by in-gel zymography with gelatin or casein. No proteolytic activity was detected either for gelatin (Figure 2B) or casein (Figure 2C). As the gel is incubated with buffer containing essential cofactors, proteolytic enzymes are able to degrade the substrate within the gel. If there is activity, proteolytic zones become visible after staining. Such zones were only visible for the positive control.

Additional highly sensitive assays were carried out to specifically detect serine proteases and metalloproteases. Samples utilized for these assays were clarified lysate and a partially purified fraction from a Q-Sepharose containing 0.2 M NaCl, where degradation was observed. For the serine protease detection, the samples were prepared to contain 5.0, 0.5, and 0.05 μg of rPspA-PdT per well. Absorbance values at 405 nm measured after incubation with L-BAPNA substrate are shown in Figure 2D. Sample values did not exceed 10% of the positive control absorbance, indicating no serine protease activity was detected by this method.

For the metalloprotease detection, the same samples and dilutions were used, corresponding to 5.0, 0.5, and 0.05 μg of rPspA-PdT per well. Figure 2E shows relative fluorescence units (RFU) measured over time (min). It is possible to observe the metalloprotease jararhagin degrading the substrate linearly over time (positive control). However, the samples maintained their RFUs between −13 and 52 units, oscillating around the noise level. Thus, no metalloproteinase activity was detected in the samples by this method.

In conclusion, no protease activity was detected by any of the four methods we carried out, leading to different strategies to investigate how and why the fusion protein was degrading.

Degraded rPspA-PdT protein had its fragments separated by IMAC-Sepharose column charged with Ni²⁺. As this hybrid protein has a His-tag at the N-terminus, the C-terminal portion resulting from degradation was collected in the flowthrough fraction of metal affinity chromatography, while the remaining protein and the N-terminal portion resulting from degradation were collected in the elution fraction with 250 mM imidazole. Flowthrough was loaded onto HPLC and the major peak fraction was collected, concentrated, and subjected to N-terminal sequencing by Edman degradation. Sequencing revealed the main cleavage site was between the two proteins rPspA and rPdT, because the N-terminal portion of the main HPLC peak was identified as the first amino acids of PdT primary sequence (LEMANKAVND, where “LE” are the amino acid residues derived from XhoI cloning site and “MANKAVND” are the residues coming from PdT sequence). These results led to the strategy of inserting molecular linkers between the antigens to increase stability and overcome this issue. As stability is required for licensing a new vaccine, this construct was not evaluated in further experiments.

Using overlap extension PCR (Figure 3A), two new genes were constructed: pspa-FL-pdt, containing a flexible molecular linker between the protein genes; and pspa-RL-pdt, with a rigid linker separating the antigens. Initially, two PCRs were carried out in parallel to build pspa-FL and FL-pdt, both genes with a portion of the linker sequence (Figures 3B, D); then another PCR was carried out to join the previous products and obtain the full pspa-FL-pdt gene (Figure 3F). The same procedure was carried out for pspa-RL-pdt (Figures 3C, E, G). The genes were inserted in pET28a and E. coli strains transformed with the respective expression vectors.

Hybrid proteins with linkers were produced from the transformed E. coli strains in chemically defined medium, pellets were lysed, incubated with 0.1% CTAB, centrifuged and proteins were purified by IMAC and size exclusion chromatography (SEC). Analysis of each step was carried out by SDS-PAGE for rPspA-FL-PdT (Figure 4A) and rPspA-RL-PdT (Figure 4B); relative quantity of fusion proteins on each fraction is described in band percentage. Elution of SEC was filtered through 0.22 µm membrane and stored at −20°C until mice immunization.

Purified proteins were sterile filtrated, maintained in Na-phosphate buffer saline (pH 7.4) at 4°C or −20°C, and evaluated periodically by SDS-PAGE and Western blot. Purified rPspA-FL-PdT was stable for at least 12 months at both 4°C (Figure 5A) and −20°C (Figure 5B). rPspA-RL-PdT degraded after 6 weeks storage at 4°C (Figure 5C). Although SDS-PAGE showed that rPspA-RL-PdT remained stable at −20°C for at least 12 months, Western blot indicated the protein underwent partial degradation (Figure 5D). High performance size exclusion chromatography also assessed the stability of the fusion proteins by comparing peak %area, using 18-month sterile filtrated purified samples stored in Na-phosphate buffer saline (pH 7.4) at 4°C or −20°C. In the case of rPspA-FL-PdT stored at −20°C, the fusion protein peak corresponded to 97.1% of total area, while another peak detected showed 2.9% (Supplementary Figure S5). When rPspA-FL-PdT was stored at 4°C, the main peak decreased to 87.6%, and few other peaks were detected, the major one representing 9.15% of total chromatogram area (Supplementary Figure S6). For rPspA-RL-PdT stored at −20°C for the same time, the main peak representing the fusion protein showed 91.6% of total chromatogram area, and other two peaks were detected, representing 4.6% and 3.85% (Supplementary Figure S7). When rPspA-RL-PdT was stored at 4°C for 18 months, the fusion protein peak corresponded to only 38.6% of total chromatogram area (Supplementary Figure S8).

Purified rPspA-FL-PdT and rPspA-RL-PdT were analyzed by circular dichroism spectra to determine the secondary structures before and after heating the samples (Figures 6A, B, respectively). In both spectra before heating, two valleys were observed (at approximately 208 nm and 220 nm), suggesting an alpha-helix-rich structure, which is expected for the PspA fragment in the hybrid molecule (Jedrzejas et al., 2001; Carvalho et al., 2012; Figueiredo et al., 2017). The complete spectra obtained after heating showed similar shapes but suggested a discrete loss of secondary structure. This was confirmed by deconvolution, by the percentage of secondary structures before and after heating (Figures 6A, B). β-sheets, absent in the PspA structure, were identified because they are present in the pneumolysin structure (Lawrence et al., 2015; Marshall et al., 2015). The heating and cooling curves obtained at 222 nm (Figures 6C, D), showed that the proteins lose part of their organized structures during heating, but regain part of them when cooled back. The melting temperature (Tm), at which the protein loses half of its original structure, was calculated and both hybrid molecules showed very similar values (Figures 6C, D), suggesting the difference observed regarding stability is not due to their Tm. rPspA-PdT without linker was not analyzed since the protein was unstable.

Since we could not find a clear difference in CD spectra that could justify the higher stability of rPspA-FL-PdT, we investigated the effects of linkers on the fusion protein stability using in silico modeling. The refined and validated models for each hybrid protein showed some features in the structure that could explain differences in stability (Figure 7A). In the rPspA-PdT model (without linker), the disordinate region indicated in pink represents the sequence that was removed in the new constructs. The removal of this sequence and addition of the rigid linker slightly altered the folding; meanwhile, the flexible linker insertion clearly changed the overall folding and seemed to alter the secondary structure surrounding its site. In the model quality estimation by QMEAN scores (Figure 7B), the highest score was obtained by rPspA-FL-PdT, while rPspA-PdT scored a more distant value and the scores for rPspA-RL-PdT remained almost the same for the tandem fusion protein. Analyzing the Ramachandran plots (Figure 8A), rPspA-PdT was the only model that presented residues in disallowed regions, and rPspA-FL-PdT showed more residues in the most favored regions. rPspA-PdT and rPspA-RL-PdT fell slightly off the Z-score range for solved native protein structures (Figure 8B).

BALB/c mice received three subcutaneous immunizations with 15-day interval containing different doses of rPspA-FL-PdT or rPspA-RL-PdT in sterile PBS with 50 µg Alum; control groups received only Alum with 0.9% sterile saline. Specific IgG antibody levels were measured 15 days after each immunization. Statistical differences between the two different fusion proteins (rPspA-FL-PdT and rPspA-RL-PdT) and the different doses (5 μg, 10 μg and 20 μg) were compared (Figures 9A–C). Initially, the groups were compared only within each immunization. Then, the first, second and third immunizations within each of the groups were also compared. While the difference between the first and second immunization was notable (p < 0.0001) for all groups, there was no statistically significant difference between the second and third doses. Although vaccination with rPspA-RL-PdT led to higher antibody production when compared with rPspA-FL-PdT groups of the same doses, the differences were not considered statistically significant (except for FL20 versus RL20 after the first immunization, p < 0.0001). While there was no statistical differences between the second and third doses regarding the antibodies against the entire fusion proteins, we could observe interesting differences for the antigens separately (Figure 10). In general, the results regarding IgG anti-PspA are similar to the antibody levels against the entire fusion proteins, i.e., differences are much more evident comparing the first and second doses, and a plateau is reached after the second dose (Figures 10A, C, E). The third dose elicited significantly higher antibodies anti-PspA only for groups immunized with 5 μg and 10 µg of rPspA-FL-PdT (Figures 10A, C), while for 20 µg or for rPspA-RL-PdT in any amount, the increase of antibody levels was not significant (Figure 10E). However, we could observe remarkable differences when comparing IgG levels anti-PdT (Figures 10B, D, F). The only difference that was not statistically significant was observed in the group immunized with 5 µg of rPspA-FL-PdT when comparing the first and second doses (Figure 10B).

Subcutaneous immunization with rPspA-FL-PdT or rPspA-RL-PdT was also evaluated regarding its ability to protect against intranasal lethal challenge with pneumococcal strain A66.1 delivered to the lungs 21 days after the third dose. All immunized mice, except one in the group that received 10 μg of rPspA-FL-PdT, survived the challenge. Comparing the survival curve (Figure 9D) of the control group to all other groups (except FL10), the difference was significant (p = 0.002); comparison between control and FL10 was also significant (p = 0.0071); and comparison between FL10 and the other groups was not considered statistically different (p = 0.3173). The only immunized animal that did not survive showed marked piloerection and reduced mobility, being euthanized on the fourth day. All other immunized animals showed no apparent clinical signs of infection.