Computational structure‐based approach to study chimeric antigens using a new protein scaffold displaying foreign epitopes

The identification and recombinant production of functional antigens and/or epitopes of pathogens represent a crucial step for the development of an effective protein‐based vaccine. Many vaccine targets are outer membrane proteins anchored into the lipidic bilayer through an extended hydrophobic portion making their recombinant production challenging. Moreover, only the extracellular loops, and not the hydrophobic regions, are naturally exposed to the immune system. In this work, the Domain 3 (D3) from Group B Streptococcus (GBS) pilus 2a backbone protein has been identified and engineered to be used as a scaffold for the display of extracellular loops of two Neisseria gonorrhoeae membrane proteins (PorB.1b and OpaB). A computational structure‐based approach has been applied to the design of both the scaffold and the model antigens. Once identified the best D3 engineerable site, several different chimeric D3 displaying PorB.1b and OpaB extracellular loops were produced as soluble proteins. Each molecule has been characterized in terms of solubility, stability, and ability to correctly display the foreign epitope. This antigen dissection strategy allowed the identification of most immunogenic extracellular loops of both PorB.1b and OpaB gonococcal antigens. The crystal structure of chimeric D3 displaying PorB.1b immunodominant loop has been obtained confirming that the engineerization did not alter the predicted native structure of this epitope. Taken together, the reported data suggest that D3 is a novel protein scaffold for epitope insertion and display, and a valid alternative to the production of whole membrane protein antigens. Finally, this work describes a generalized computational structure‐based approach for the identification, design, and dissection of epitopes in target antigens through chimeric proteins.


| INTRODUCTION
2][3][4][5] These subunit vaccines are particularly attractive since they are safer compared to killed or live-attenuated preparations, and they direct the immune response against the most immunogenic and protective antigens. 6,73][14] Some of them are part of the external membrane of the pathogen through an extended hydrophobic portion. 15The presence of large hydrophobic domains makes these membrane antigens insoluble when recombinantly produced.Moreover, the hydrophobic regions, naturally not exposed to the immune system, in the recombinant refolded membrane antigens result to be accessible deviating the immune response.An alternative strategy could be the identification and production of the most immunodominant regions/epitopes.For their identification, different approaches can be used: (i) protein or peptide array, (ii) genetic manipulation, and (iii) Hydrogen Deuterium Exchange (HDX) experiments. 16,17The genetic manipulation of the strain is time-consuming and requires the use of pathogens. 17,180][21] An ideal scaffold should be able to maintain its properties in terms of stability and solubility while displaying target epitopes.
The domain 3 (D3) of Group B Streptococcus (GBS) pilus backbone protein (BP-2a) is one of the four domains of the GBS BP-2a and the most immunogenic pilus domain able to induce functional antibodies. 223 is a small and highly stable protein (̴ 15 kDa), due to the presence of an internal isopeptide bond and can be easily produced as soluble His-tagged protein in Escherichia coli. 22,23These characteristics make D3 an attractive protein to be tested as a scaffold for displaying foreign epitopes.
The emergence of Neisseria gonorrhoeae strains that are resistant to nearly all classes of antibiotics available for treatment underscores the urgent need for new prevention, for this reason, a vaccine targeting this pathogen is highly needed. 24,25Porin B (PorB) and the opacity-associated proteins (Opa) are two of the most highly abundant outer membrane proteins involved in N. gonorrhoeae pathogenicity. 26PorB, represents 60% of the total outer membrane protein content of N. gonorrhoeae and it exists as two possible allelic variants: 1a and 1b 27 .These forms are mutually exclusive and show high degrees of sequence homology with the highest variability within the loop regions. 28,29The gonococcal strain FA1090 expresses the allelic form 1b of PorB, while OpaB is one out of a possible 11 Opa variants encoded by the FA1090 genome. 30The structure of the Opa family members is predicted to be composed of a membrane-spanning eight-stranded β-barrel, connected by four extracellular loops. 31The specificity for host receptors is conferred by the semi-and hyper variable regions found in these flexible loops. 31Despite that both PorB and OpaB are potential vaccine targets, their hydrophobic nature makes their production as recombinant proteins challenging. 27,32,33In fact, the recombinant expression in E. coli leads to the accumulation of the proteins in inclusion bodies making the purification and refolding processes laborious, expensive, and time consuming. 33In addition, the 3D structure of both antigens is not present in the protein data bank (PDB) and only the structure of PorB homologues is reported. 34n this study, taking advantage of the recent developments in the field of artificial intelligence (AI)-based structure prediction methods (AlphaFold2 35 ), the 3D structures of both PorB.1b and OpaB have been accurately predicted.The structure-based approach allowed the identification in the D3 scaffold of six flexible and unstructured loops potentially engineerable with foreign epitopes.Moreover, the structural analysis of predicted models of PorB.1b and OpaB led to the identification of 12 extracellular loops (8 for PorB.1b and 4 for OpaB) to be extrapolated and displayed onto D3 scaffold.Overcoming the difficulties related to the production of entire insoluble membrane proteins, several different constructs have been recombinantly produced in E. coli.Biochemical characterization allowed the identification of the best engineerable D3 site.The correct display of target epitopes on the scaffold has been demonstrated by X-ray crystallography and immune assays allowing the (i) identification of immune dominant loops in the two model antigens and (ii) the recognition of native proteins by the mouse sera raised with the recombinant chimera.

structure prediction of target antigens
The amino acidic sequences encoding for OpaB and PorB.1b proteins have been extrapolated from the genome of N. gonorrhoeae strains FA1090 and F62 (used only for the sequence alignment reported in the supplementary information) collected in public databases for molecular typing and microbial genome diversity (PubMLST). 36tarting from the prediction of the three-dimensional structures, the sequences of OpaB and PorB.1b target loops as well as the flexible regions of D3 scaffold have been identified.The structure of PorB.1b and OpaB has been computationally predicted with Swiss model 37 and AlphaFold2, 38 while the structure of D3 used as scaffold is available in the PDB (pdb code: 2XTL 22 ).
The structural investigation of both antigens and scaffold has been performed with PyMOL. 39igure representation has been performed with PyMOL and ChimeraX. 39,40

| Design and production of recombinant proteins
Chimeric D3 displaying OpaB and PorB.1b loops have been designed by inserting the target epitope into a flexible site of D3 scaffold.The genes encoding for the designed molecules were produced as DNA strings by GeneArt (Thermo Fisher Scientific) and cloned into pET15b+ (Merck-Sigma) with enzyme-free cloning strategy using In-fusion cloning kit (Takara).The plasmid pET29b+ encoding D3-GFP (inserted in D3Loop2) was purchased by Twist Bioscience.Codon usage has been optimized for the expression in the E. coli host cell and 6xHis-TEV tag has been added at N-term of each chimera.
Protein expression has been achieved in the cytoplasm of E. coli BL21(DE3)t1r provided by NEB.Cells were grown in HTMC expression medium (Glycerol 15 g/L; Yeast Extract 30 g/L, MgSO 4 x7H 2 O 0.5 g/L; KH 2 PO 4 5 g/L; K 2 HPO 4 20 g/L; KOH 1 M to pH final 7.35 ± 0.1) for 16-18 h at 20°C, then induction was performed by adding IPTG 1 mM to the cell culture for 24 h.Protein expression and solubility have been checked with SDS-PAGE analysis by loading 15 μL of sample (total and soluble fraction) mixed with loading sample dye and reducing agent (Thermo-Fisher Scientific, Italy) on 4%-12% polyacrylamide gels.The percentage of soluble protein has been evaluated in relation to the total amount of expressed protein by analyzing the gel image with ImageJ software. 41The soluble proteins were chemically extracted by using the CelLytic reagent (Merck-Sigma-Aldrich, Italy) solubilized in distilled water followed by centrifugation to remove cell debris.The recombinant soluble proteins were purified from the supernatant with IMAC chromatography.The sample has been applied onto a 5-mL His-trap FF crude column (Cytiva) previously equilibrated with 1X PBS.The flowthrough has been separately collected and the resin has been washed with 50 mL of 1X PBS with 20-mM imidazole, while the interest protein has been eluted with 300-mM imidazole in PBS.Buffer was exchanged in PBS by using 10-kDa cutoff spin concentrator (Millipore Amicon Ultra).Protein purity was checked by SDS-PAGE analysis and protein concentration was determined by measuring the absorbance at 280 nm with Nanodrop and adjusted according to the corresponding molar extinction coefficient calculated with Expasy ProtParam tool. 42Tested molecules have a molecular weight between 14 and 21 kDa and the calculated molar extinction coefficient was in the range of 16 960 and 29 910 M −1 cm −1 .
Recombinant OpaB protein has been expressed in the inclusion bodies (IB) of E. coli Bl21DE3t1r with growth in Luria-Bertani expression medium for 16 h at 37°C and then induction was performed by adding IPTG 1 mM to the cell culture for 6 h.The protein has been extracted from IB with 8 M urea and maintained in this buffer for protein quantification with Nanodrop and western blot analysis.

| Nano differential scanning fluorimetry
The analysis has been performed with Tycho NT 6 (Nanotemper).A capillary was filled with the sample [0.5 mg/mL] and a linear temperature ramp from 25 to 91°C was applied to unfold proteins for 3 min.During the scanning, the increase of intrinsic tryptophan or tyrosine fluorescence was recorded and the 'melting temperature' or 'Tm', corresponding to the midpoint of the transition from folded to unfolded was evaluated.Samples were measured in triplicate and data were analyzed in GraphPad Prism 9.

| Western blot
Three micrograms of purified recombinant proteins was transferred to a nitrocellulose membrane after an SDS-PAGE run using i-transfer and i-Blot Mini/regular kit (Thermo Fisher Scientific, Italy).The membrane was then blocked with 3% milk in PBS+ 0.1% TWEEN (Sigma-Aldrich, Italy) for 1 h at room temperature (RT).After the incubation with primary antibodies (Ab), produced in mice, and secondary antibodies (Sigma-Aldrich SAB3701214-1), the membrane was washed three times with PBS + 0.1% TWEEN for 5 min under shaking in order to remove unbound Ab.The result was detected using the colorimetric reaction happening between horseradish peroxidase (HRP) conjugated with the secondary antibody and its substrate 4-chloro-1-naphthol (Bio-Rad, Italy).

| GFP fluorescence assay
The fluorescence of recombinant purified D3-GFP protein was measured via fluorescence spectroscopy on a 96well black flat bottom using a Tecan Infinite M200 reader (Tecan, Mannedorf, Switzerland).The excitation wavelength was 395 nm (bandwidth 10 nm) and the emission spectrum was recorded with gain 80 at 448 nm.The intensity recorded for the D3 alone in PBS buffer at 448 nm was subtracted from that recorded for GFP and D3-GFP to remove the background.A final volume of 100 μL of 0.85 mg/mL of each protein sample was transferred into the 96-well black flat bottom plate (Greiner Bio-One, Frickenhausen, Germany) and the fluorescence emission was recorded.

| Luminex assay
Luminex Magplex beads were equilibrated at RT and prepared for use according to the manufacturer's instructions.A quantity of 100 μL (1.25 × 10 6 ) of resuspended beads was transferred to a LoBind Eppendorf tube and placed into a magnetic separator for 2 min.The supernatant was removed, and the beads were washed with water and activated for 20 min with NHS and EDC (10 μL of each 50 mg/ mL solution diluted in dH 2 0) in 100 mM of monobasic sodium phosphate pH 6.2 buffer and washed twice with 50 mM MES pH 5. The activated beads were incubated for 2 h with 20 μg/mL of molecule of interest resuspended in 50 mM MES pH 5. Coupled beads were finally washed twice with 1X PBS + 0.05% Tween and stored in 500 μL of assay buffer (1X PBS, 0.05% TWEEN and 0.5% BSA) at 4°C.After each step described above, beads were resuspended by vortex for approximately 20 s and placed into a magnetic separator for 2 min.
Standard sera and sera from mice immunized with outer membrane vesicle (OMV) or recombinant proteins were pre-diluted in assay buffer and then 3-fold dilution was performed in 50-μL final volume for each well of Grainer microtiter plate.Fifty microliters of coupled beads (3000 beads/well) was added to sera and the plate was incubated for 1 h at RT in the dark on a plate shaker at 700 rpm.Unbound Ab was removed by washing plates three times with 200 μL of PBS using an automatic plate washer with a magnetic plate holder.Each well was then loaded with 50 μL of secondary antibody [2.5 μg/mL] conjugated with R-phycoerythrin-AffiniPure 1:200 diluted in assay buffer and incubated for 1 h at RT in the dark on a plate shaker at 700 rpm.After washing, beads were suspended in 100 μL of PBS and analyzed with Bioplex 200.Data were acquired in real time by Bioplex Manager Software 6.2 (BioRad, Italy).

| Ethics statement and in vivo studies
Animal treatments were performed in compliance with Italian laws and approved by the institutional review board (Animal Ethical Committee) of GSK Vaccines Siena, Italy.
Ten micrograms of each purified recombinant chimera or OMV produced by N. gonorrhoeae FA1090, with low endotoxin level (<0.1EU/μg), mixed with AS01 (used in combination with recombinant proteins) or Alum (OH) (used for OMV) adjuvants have been used to immunize intraperitoneally ten 7-week-old CD1 female mice.Three different immunizations based on recombinant proteins were performed on days 0, 21, and 37, while the immunizations based on the OMV were performed on days 1, 29, and 57.Sera samples have been collected at each point.

| Protein crystallization and structure determination
Purified recombinant protein D3PorBLoop5 was concentrated until it reached a concentration of 18 mg/mL.Using a Crystal Gryphon robot (Art Robbins Instruments), 384 different crystallization conditions were tested by using 200-nL reservoir and 200-nL protein sample.The best crystals were grown for 6 days in buffer containing 0.1 M HEPES with 20% w/v jeff ED-2001 as precipitant at pH 6.5.Crystals were soaked in the original mother liquor supplemented with 15% ethylene glycol prior to cryo-cooling in liquid nitrogen.Diffraction of the crystals was performed at beamline ID30A-1 of the European Synchrotron Radiation Facility (ESRF).We collected 900 images at 100 K, at wavelength λ = 0.96546 Ås.Data were processed using autoProc 43 and they were reduced using Scala within the CCP4 program suite. 44Crystals of the D3Loop5 chimera belong to space group F222, with the asymmetric unit containing two copies and a solvent content of 56.8% (Matthew coefficient of 2.85 Å3/Da).The structure of the D3PorBLoop5 was determined at 2.6 Å resolution by molecular replacement with Phaser 45 using two separate search models obtained from homology modeling simulation (Swiss model 37 ) and from 2XTL data collected in PDB database.Rigid body and restrained refinement were carried out with Refmac5 (from CCP4i suite 44 ).Structure quality was assessed using Molprobit, 46 while protein-protein interface areas were analyzed and calculated using the Protein Interfaces, Surfaces, and Assemblies service (PISA) 47 by PDBePISA.Figures were generated using PyMOL. 39 3 |RESULTS

| Structural analysis of GBS D3
allowed the identification of potential engineerable sites D3 is one of the four domains that builds up the backbone protein of GBS pilus 2a (BP-2a) which, together with ancillary proteins, forms the entire pilus.The structural analysis of BP-2a (PDB: 2XTL) 22 revealed the presence of only three out of four domains: D2, D3, and D4.Compared to the other domains, D3 is the smallest and folds into a β-barrel structure independently protruding from BP (Figure 1).The six spanning β-strands, forming the β-barrel, are connected by six flexible loops that are potentially engineerable.Despite the small size, in vivo studies revealed that D3 is highly immunogenic in mice. 22,23Considering all these data, D3 has been chosen as a potentially suitable scaffold for the display of foreign epitopes.To the best of our knowledge, the 3D structures of both antigens, PorB.1b and OpaB, are not publicly available hampering future structure-based antigen design.Only in silico prediction based on sequence homology has been reported (Supplementary Figure S1). 29,48However, recent advances in the structural prediction field provided deep learning techniques that allowed models with a level of accuracy comparable to the experimental structures.In this study, the artificial intelligence (AI) approach named AlphaFold2 (AF2), 35 a cutting-edge deep learning algorithm that outperforms the latest protein folding competition CASP14 49 has been applied.AF2 prediction of PorB.1b revealed the canonical 3-fold PorB symmetry with 16-stranded β-barrel, short turns connecting the strands on the periplasm, and long interstrand loops on the extracellular part of the pore for each of the monomer (Figure 2A).Interestingly, it has been observed that two of the eight loops present a secondary structure.Loop5 (L5) is structured as a β-hairpin, which is solvent exposed and oriented toward the central channel pore of the monomer, while Loop3 presents two short-helical turns, and it is predicted to be directed inside the pore (Figure 2A).The per-residue measurement of model local confidence (per-residue local distance difference test pLDDT) revealed high confidence for the β-barrel predicted region (pLDDT ≥50) but low confidence (pLDDT ≤30) in the region of loops.This is probably due to the intrinsic flexibility of these regions.
The OpaB model is in agreement with the secondary structure prediction of eight antiparallel β-strands, forming a barrel structure in the bacterial outer membrane, linked by four extracellular loops 50 (Figure 2B).Two out of four loops, namely, Loop2 and Loop3, showed the same β-hairpin structure previously observed in PorB.1bLoop5.The pLDDT score revealed high confidence for the β-strands region (pLDDT ≥50) but low confidence in the OpaB/L2-L3 region (pLDDT ≤30).
3.3 | D3 site 2 is the optimal site for the insertion of foreign epitopes.
The identification of the best epitope insertion site has been performed considering the expression level, solubility (Figure 3A), thermal stability (Figure 3B,C), and the ability to preserve the epitope conformation (Figure 3D) of six different chimeras.The analysis has been performed by inserting the longest epitope identified in the model antigens (PorB.1bLoop3)into each of the six D3 sites.All chimeras were expressed in soluble form in E. coli at levels comparable to the empty D3 and the percentage of soluble chimera is reported in the Supplementary Figure S2.Only engineering D3 site 4 generated a less soluble chimera incorrectly folded (Figure 3A).In fact, for this chimera, a transition between the folded and unfolded states was not appreciated by nano differential scanning fluorimetry analysis suggesting that the protein is not properly structured (Figure 3B).By contrast, the chimera generated by engineering of all the other D3 sites presented a shift in the measured fluorescence at temperatures in the range of 45-76°C (Figure 3C), suggesting that they are folded.The calculated Tms of chimeras is lower than the Tm detected for the scaffold alone (88°C), reasonably due to the insertion of a long and flexible foreign portion.Among all tested positions, the engineerization of site 1 and site 2 led to the formation of the most stable chimeras.Notably, the higher initial ratio of 350/330 nm detected for site 1 chimera suggests the presence of partially unfolded or aggregated portions, 51 whereas the D3 site 2 shows a profile similar to the empty D3 scaffold with the lowest initial 350/330 nm ratio.Another important factor for the choice of the best insertion position is the ability to preserve the native epitope conformation.For this reason, the 3D structure of each chimeric D3 displaying PorB.1bLoop3 has been computationally predicted (Supplementary Figure S3).By this analysis, PorB.1bLoop3 resulted in maintaining its α-helix conformation when inserted into D3 site 2 and site 6 (Figure 3D).In all the other sites, the epitope is partially structured with respect to its predicted native conformation in the whole protein.Combining all these data, the engineerization of D3 site 2 resulted in the most promising strategy to produce soluble and stable chimeras able to correctly display the epitope of interest.

| D3 correctly displays gonococcal epitopes and allowed the identification of immunodominant loops
The structural investigation of PorB.1b and OpaB models allowed the identification of eight and four extracellular loops, respectively.These 12 different epitope amino acid sequences have been extrapolated from the whole proteins and inserted into D3 site 2. All chimeras have been obtained as His-tagged soluble proteins in the cytoplasm of E. coli.Structural characterization of purified proteins revealed that the insertion of a large epitope (tested up to 34 amino acids) did not destroy scaffold structure.In fact, the thermostability analysis revealed that each molecule is properly folded with a Tm ranging from 70 to 82°C (Supplementary Figure S4).Furthermore, to identify the immunodominant loops in PorB.1b, a western blot with two different mouse sera was performed.This analysis revealed that the antiserum raised against the recombinant purified PorB.1b (α-rPorB) recognized only PorB.1b Loops 1-3-5 and 6, whereas the serum raised against gonococcal outer membrane vesicles (α-OMV-FA1090), containing 70% of PorB.1b and almost 20% of Opa family proteins, recognized only the Loop5 (Figure 4Ai,ii).This result was further confirmed by Luminex-based analysis (Figure 4Bi).
A similar experiment was conducted with the four loops of the OpaB protein displayed on the D3 scaffold revealing that α-OMV-FA1090 serum recognized clearly only D3OpaBLoop2 and Loop3 (Figure 4Aiii).In fact, while D3OpaBLoop1 was poorly recognized in western blot by α-OMV-FA1090, it was not detected by the same serum in Luminex assay nor was D3OpaBLoop4 (Figure 4Bii).D3OpaBLoop1 appears to be weakly detected in the western blot, but not in the Luminex assay (signal similar to empty D3).This could depend on the different conditions of the two assays.Moreover, the in vivo study conducted in mice revealed that the designed chimeras elicit immune responses against the target epitopes.In this sense, α-D3PorBLoop5 and α-D3OpaBLoop2 antisera recognized in western blot the native proteins present in total cell-extracts and in OMV of N. gonorrhoeae FA1090 (Figure 4C).In accordance, due to the high sequence diversity of PorB.1bLoop5 and OpaBLoop2 between strains FA1090 and F62 (Supplementary Figures S5 and S6), neither tested sera recognized the total extract of F62 strain (Figure 4C).As expected, α-D3PorB.1bLoop5and α-D3OpaBLoop2 antisera recognized in western blot the purified recombinant full-length proteins (PorB or OpaB) the constructs used to raise the sera (D3PorB.1bLoop5and D3OpaBLoop2) and the D3 empty scaffold (Supplementary Figure S7) highlighting the immune response raised also against the D3 scaffold and not only against the target foreign epitope.

| D3PorB.1bLoop5 crystal structure confirmed the preservation of native epitope conformation.
The computational structural prediction of PorB.1bLoop5 has shown that Loop5 adopts a β-hairpin structure in native PorB.1b and the same structure is also computationally predicted in the D3Loop5 chimeric protein.To further elucidate the conformation of Loop5 PorB.1b in D3 scaffold and validate the in silico analysis of PorB.1b structure, recombinant pure D3PorBLoop5 has been used to conduct a crystallization trial.Diffracting crystals were obtained and the 3D structure of D3PorBLoop5 has been solved by X-ray crystallography.Electron density maps were of high quality and allowed the model building and structure refinement to a final resolution of 2.6 Å (Table 1).The crystallographic unit cell contains two chains arranged in a mirror image.The second chain is rotated at 180° along y axes compared to the first chain (Figure 5A).Although crystallization was carried out using the entire D3Loop5 chimera (139 a.a.), 8 N-terminal residues were absent in the density maps of both chains (Figure 5A) while 7 N-terminal residues were missing only in the density map of chain B. Based on the computational prediction, this is reasonably due to the high flexibility of these regions.The interface analysis performed with PISA revealed that the interface area represents about 9% of the entire surface and the two monomers are taken together by multiple hydrogen bonds occurring between residues located in a β-strand (from residue 98 to 114).In addition, for each chain, the presence of the internal isopeptide bond between residues K43 and N146 has been detected (Figure 5C).This result, in accordance with the data published by Nuccitelli   suggests that the D3 engineerization with a epitope did not alter its structure.Moreover, the foreign epitope displayed has maintained its native conformation.Density map of PorB.1bLoop5 region confirms the β-hairpin organization of this epitope (Figure 5D).This result validates the computational structural prediction reported above of PorB.1b.In addition, the model of chimeric D3PorB.1bLoop5has been computationally predicted with AF2 and compared with the crystal structure obtained (Figure 5B).The structures are aligned with an overall calculated root mean square deviation (RMSD) of 2.9 A°.In particular, the structure alignment presents the highest similarity between the structures of the scaffold with an RMSD of 1.5 A°.While the structure of PorB.1bLoop5 between crystallographic and predicted model presents a higher structural diversity with an RMSD of 5.8 A° (Figure 5B).Reasonably, the structural diversity detected is due to the flexibility of this region, while its secondary structure as β-hairpin is maintained.These results suggest that the engineerization of D3 site 2 allows the preservation of the epitope native conformation without destroying the scaffold structure.Atomic coordinates of D3PorB.1bLoop5have been deposited in the Protein Data Bank under accession PDB code 8C27.

GFP protein when inserted in D3Loop2
The green fluorescent protein (GFP) from the jellyfish Aequorea victoria is a soluble protein displaying visible fluorescent light at a wavelength of 508 nm after being excited with ultraviolet light. 52To produce a fluorescent signal, GFP must form and maintain its tight β-barrel structure.For this reason, GFP is usually used as a protein reporter to distinguish proteins that correctly fold and are soluble when expressed in E. coli from those which misfold and aggregate. 53In this study, GFP was used as a model protein to investigate if D3Loop2 could accommodate and correctly display an entire protein, instead of a loop only, as demonstrated for PorB and OpaB proteins.The AF-2 3D structural model of D3-GFP was generated suggesting that flexible residues at N-term and C-term of GFP could allow the correct β-barrel structure of GFP (Figure 6A).After cell lysis, the protein samples containing D3-GFP and GFP alone immediately appeared highly yellow suggesting the presence of folded GFP in the cell lysate.The two proteins were successfully purified from E. Coli cytoplasm by IMAC chromatography, and the pure fluorescent protein samples were visualized by SDS-PAGE (Figure 6B), showing the bands at the correct MW (41 kDa for D3-GFP, 14.7 kDa for alone, and 26.7 kDa for GFP alone) and by UV-light exposure (Figure 6C).Fluorescence of D3-GFP and GFP was measured at 0.85 mg/mL protein concentration in triplicate, confirming a similar fluorescence intensity for D3-GFP and GFP alone.No fluorescence was detected for D3 alone or PBS buffer, as expected.

| DISCUSSION
The use of an engineerable protein scaffold can be particularly useful in vaccine development for different purposes: (i) dissection of protein antigen to identify the immunogenic regions (ii) to direct the immune response against the epitope of interest, (iii) to overcome the difficulties to produce insoluble protein antigens by focusing only on the soluble portions.In this work, the entire process for the identification and production of a new protein scaffold displaying foreign epitopes has been presented.Usually, chimeric proteins are generated by fusing the foreign epitopes at the N-or C-term of the scaffold.However, in the case of loops extrapolated from the protein of interest, finding an internal region of the scaffold for the insertion can be beneficial in terms of preservation of foreign epitope conformation.For this reason, potentially engineerable sites should be identified in the flexible portion of the scaffold and not in the regions presenting secondary structures.Considering all these factors, domain 3 of GBS pilus protein, containing six flexible loops potentially engineerable, has been identified and tested as an ideal scaffold for foreign loops display.As target antigen, two important membrane proteins (PorB.1b and OpaB) of Neisseria gonorrhoeae have been selected for loop extrapolation and display on D3 scaffold.To overcome the difficulties related to the lack of structural information about the selected target antigens, AlphaFold2 has been applied to predict their 3D structures with high accuracy.The structural analysis revealed that both proteins are organized as β-barrel with eight (PorB.1b)and four (OpaB) extracellular loops.This result is in accordance with other reported predictions based on the sequence homology. 54,55However, given the high sequence variability of loops among protein variants, the major structural diversity is detected for the loop conformation.In fact, a different secondary structure has been predicted by AF2 for some of the extracellular loops, differently from what has been previously observed by using comparative modeling prediction tools. 27,56  Although Loop1 and Loop4 were highly conserved among strains and variants, they are not able to induce the production of specific antibodies. 30The immunization with chimeric D3 constructs induced an immune response against the D3 protein scaffold as expected and not only against the specific target epitope inserted.GBS D3 has been previously demonstrated to be a protective portion of the Pilus 2a backbone protein, inducing functional antibodies. 22Further studies could investigate if the response raised by the D3 chimeric scaffold could maintain its functionality.Moreover, the crystal structure resolution of chimeric D3 displaying PorBLoop5 has experimentally confirmed that the epitope conformation is maintained upon engineerization, validating the computational structural prediction of PorB.1b.Moreover, to assess whether more complex epitopes or structured antigens could keep their native conformation when displayed on the D3 scaffold, the entire GFP aminoacidic sequence was inserter in D3 scaffold Loop2 and the resulting chimera was produced as recombinant protein.After purification, the typical yellow color of GFP at room light was appreciated in all fractions containing D3-GFP chimera demonstrating the correct folding of GFP protein inserted into the D3 scaffold Loop2.
All these data show that D3 is an ideal protein scaffold easily engineerable.It can correctly display tested target epitopes/antigens, thus maintaining their physiochemical properties.D3 scaffold has been successfully used to dissect two membrane protein antigens (PorB.1band OpaB) allowing the identification of three more immunogenic epitopes (PorB.1bLoop5,OpaBLoop2, and Loop3).PorB and OpaB are two main outer membrane proteins of N. gonorrhoeae, an important human pathogen, for which an effective vaccine is highly needed.The chimeric constructs obtained, and the identification of immunogenic epitopes could be of interest in the contest of developing a N. gonorrhoeae vaccine.
The strategy adopted in this study requires an extensive structural investigation of both protein scaffold and antigens.The computational structure prediction was crucial for the identification of PorB.1b and OpaB extracellular loops and for the evaluation of best D3 engineerable site.The work-plan presented here is based on a multidisciplinary approach including the in silico protein design, structural and biochemical analysis, and in vivo studies leading to the identification of a new optimal protein scaffold, immunodominant epitopes and to validate the computational structure In conclusion, the D3 scaffold and the approach described here to study the most challenging class of proteins (membrane proteins) can be further applied to other proteins.Future studies could be focused on the investigation of additional application of D3 scaffold for vaccine development and antibody characterization.

F
I G U R E 3 (A) SDS-PAGE analysis of expression and solubility of chimeric D3 displaying PorB.1bLoop3 inserted into six different sites.T: total fractions, S: soluble fractions of E. coli extract.(B) Unfolding profiles of each chimeric D3 evaluated with nano differential scanning fluorimetry.The error bars indicate the standard deviation calculated on three measurements.(C) Inflection temperatures corresponding to the medium melting temperature of each molecule calculated as inflection point of the curve.(D) Cartoon representation of predicted 3D structure of D3 displaying PorB.1bLoop3 into sites 2 and 6 and PorB.1b.Black arrows indicate the two α-helix structures of Loop3 PorB.1b detected in the native protein and maintained after the insertion into D3 sites 2 and 6.

F I G U R E 4
Western blot analysis and Luminex assay of chimeric D3 displaying PorB.1b loops testing (A-i) α-rporB1b and (A-ii and B-i) α-OMV-FA1090 sera.Western blot (A-iii) and Luminex assay (B-ii) for chimeric D3 displaying OpaB loops.Western blot analysis of total lysate of FA1090 and F62 strains and purified OMV from FA1090 tested with α-porB Loop5 sera (C-ii) and α-OpaB Loop2 (C-i).X axis of Luminex graph reports the sera dilutions.

F I G U R E 5
Crystal structure resolution of D3PorB.1bLoop5.(A) Cartoon representation of crystal structure obtained of D3porBloop5 crystallographic dimer (chain A in green and chain B in cyan).Bottom alignment between amino acid sequence of chimeric protein (D3PorB.1bLoop5)and amino acid sequence of crystallized protein (crystal).Black rectangle highlights the N-terminal residues absent in both chains of the crystal.Yellow rectangle highlights the seven C-terminal residues absent in the crystal structure of chain B. (B) Structure alignment of computationally AF2-derived 3D model (dark gray) of D3PorB.1bLoop5 and crystal structure (green) chain A and RMSD values calculated with Pymol.(C) Isopeptide bond detected in both chain A (top) and chain B (down) with the respective density map around the residues N146 and K43 involved in the bond.(D) Graphical representation of density map detected about the PorB.1bLoop5.has been selected to perform a scanning engineerization of each D3 site.The production biochemical characterization of six different chimeras indicated that D3 site 2 is the best insertion point.Based on that, all the other PorB.1b and OpaB loops have been extrapolated and inserted in this position.By combining western blot and Luminex analysis, PorB.1bLoop5 and OpaBLoop2 and Loop3 have been identified as the most immunogenic loops.This is in accordance with the information reported by Cole et al., who defined Loop2 and Loop3 as the hypervariable loops of Opa proteins containing the most immunogenic and functional epitopes.

F
I G U R E 6 D3 scaffold correctly displays folded GFP: (A) Structural model of D3-GFP generated with Alphafold2 and visualized with ChimeraX.D3 structure is colored in dark green and GFP in light green.(B) SDS-PAGE of purified D3-GFP, GFP alone, and D3 empty.(C) Fluorescence signals of GFP constructs (D) Thermal stability (nano-DSF) of D3-GFP construct.
22 al.,22 Among all PorB.1b and OpaB loops, PorB.1bLoop3Statistics for the highest resolution shell are shown in parentheses.R sym = S hkl S i jIi(hkl) − ⟨I(hkl)⟩ j∕S hkl S i Ii(hkl) .R work = SjjF(obs)j − jF(calc)jj ∕ S jF(obs)j.R free is the same as for Rwork but calculated for 5% of the total reflections that were chosen at random and omitted from refinement. Note: