Dissecting the contribution of O-Antigen and proteins to the immunogenicity of Shigella sonnei generalized modules for membrane antigens (GMMA)

GMMA are exosomes released from engineered Gram-negative bacteria resembling the composition of outer membranes. We applied the GMMA technology for the development of an O-Antigen (OAg) based vaccine against Shigella sonnei, the most epidemiologically relevant cause of shigellosis. S. sonnei OAg has been identified as a key antigen for protective immunity, and GMMA are able to induce anti-OAg-specific IgG response in animal models and healthy adults. The contribution of protein-specific antibodies induced upon vaccination with GMMA has never been fully elucidated. Anti-protein antibodies are induced in mice upon immunization with either OAg-negative and OAg-positive GMMA. Here we demonstrated that OAg chains shield the bacteria from anti-protein antibody binding and therefore anti-OAg antibodies were the main drivers of bactericidal activity against OAg-positive bacteria. Interestingly, antibodies that are not targeting the OAg are functional against OAg-negative bacteria. The immunodominant protein antigens were identified by proteomic analysis. Our study confirms a critical role of the OAg on the immune response induced by S. sonnei GMMA. However, little is known about OAg length and density regulation during infection and, therefore, protein exposure. Hence, the presence of protein antigens on S. sonnei GMMA represents an added value for GMMA vaccines compared to other OAg-based formulations.


Results
GMMA production and characterization. GMMA displaying the OAg on their surface and carrying wild type lipid A were produced from a S. sonnei 53G strain (Ss OAg+) ( Table 1). Further mutations were introduced in this GMMA producer strain to decrease the number of the lipid A acyl chains, thus obtaining two alternative strains carrying a lipid A with reduced endotoxicity: Ss ΔhtrB OAg+ and Ss ΔmsbB OAg+ (Table 1). GMMA not displaying the OAg on their surface were produced from a S. sonnei 53G strain cured from the big virulence plasmid (Ss OAg−). The same lipid A modifications were also applied to the OAg-negative strain to obtained detoxified GMMA (Ss ΔhtrB OAg− and Ss ΔmsbB OAg−) ( Table 1).
All GMMA were fully characterized and showed an average radius in the range of 30-50 nm as measured by High Performance Liquid Chromatography-Size Exclusion Chromatography (HPLC-SEC) coupled with Multiangle Light Scattering (MALS) ( Table 2). Absence of free soluble proteins or DNA fragments was confirmed by HPLC-SEC analysis performed on the entire particles (Fig. S1a). The expected lipid A structure was confirmed by Matrix-Assisted Laser Desorption/Ionization-Time Of Flight Mass Spectrometry (MALDI-TOF MS) analysis and the ability of the different GMMA to activate Toll-like receptor 4 (TLR-4) was verified in a Monocyte Activation Test (MAT) (Fig. S1b,c). OAg-positive GMMA displayed a low molecular weight (LMW) OAg population, a medium molecular weight (MMW) OAg population and a high molecular weight group 4 capsule (G4C) 20 with average size of approximately 2.0 kDa, 20 kDa and 200 kDa respectively (Fig. 1). All polysaccharide species shared the expected structure, consisting of a disaccharide repeat containing 2-acetamido-2-deoxy-l-altruronic acid (l-AltNAcA) and 2-acetamido-2-deoxy-l-fucose (FucNAc4N), as verified by Nuclear Magnetic Resonance  Table 2). For the OAg-negative GMMA, the absence of OAg repeats was confirmed by High-Performance Anion-Exchange Chromatography coupled with Pulsed Electrochemical Detection (HPAEC-PAD) analysis, with lack of the characteristic sugars of the OAg repeating unit. Also, HPLC-SEC analysis on extracted sugars showed absence of G4C and MMW OAg and only presence of core molecules at 1.8 kDa (Fig. 1).
Proteomic analysis of S. sonnei GMMA. Proteomic analysis was performed by Liquid Chromatography coupled with Mass Spectrometry (LC-MS/MS). A global set of 250 proteins were quantified analyzing the six different GMMA samples (Table S1). We decided to focus on the 30 most abundant proteins in each GMMA preparation, which accounted for > 95% of the total protein composition of each sample. We obtained a list of 49 proteins in total for the six samples (Table S2). The localization of the quantified proteins was predicted according to PSORTb software. The analysis predicted ~ 90% of the total protein amount for each GMMA sample to be either outer membrane proteins or periplasmic proteins (Table S2), thus confirming that GMMA are generated by outer membrane blebbing and not by bacterial lysis. Around 2-5% of the total protein amount had an unknown prediction of localization in all samples but only a minority of proteins was predicted to be inner membrane or cytoplasmic. Next, we compared the abundance of each protein in the different GMMA samples (Table S2 and Fig. 2). In particular, we compared the OAg+ GMMA with penta-acylated lipid A to each other and to the OAg+ GMMA with hexa-acylated lipid A (Fig. 2). The results of the comparisons are reported as XY plots where proteins showing similar abundances in the two compared samples lay between the two dotted lines, while proteins showing at least a twofold difference in one of the two samples lay outside the dotted lines. Only proteins accounting for at least 1% of the total composition and varying of at least twofold in the compared samples were considered significantly different and were depicted as red circles. The different comparisons showed that the majority of proteins are similarly represented in the different GMMA samples, while only 10 proteins are slightly variable. In particular the most evident changes, due to the two modifications of the lipid A structure, can be highlighted in a decrease on protein abundance (i.e. OmpA, OmpD, OmpX, OmpF, FkpA) or an increase (i.e. DegP, UshA, TolC).   Serological response. Independently from the different OAg dose, Ss ΔhtrB OAg+ GMMA and Ss ΔmsbB OAg+ GMMA elicited comparable levels of anti-LPS specific total IgG both 27 days after the first immunization and 14 days after the second one, with titers at day 42 being statistically higher than those at day 27. Also OAg-negative GMMA elicited anti-LPS IgG titers, even if significantly lower than those elicited by vaccination with the corresponding OAg-positive GMMA. Anti-LPS titers observed after vaccination with OAg-negative GMMA were reasonably due to the presence of anti-LPS core antibodies (considering that S. sonnei LPS was used as coating antigen in ELISA) and were similar in the two groups of mice immunized with GMMA lacking the OAg (Fig. 3a). Anti-LPS specific IgM resulted to be lower for mice immunized with Ss ΔhtrB OAg+ GMMA compared to Ss ΔmsbB OAg+ GMMA, whereas titers were comparable between the two groups of mice immunized with OAg-negative GMMA. Again, anti-LPS specific IgM were higher after vaccination with OAg-positive GMMA than with OAg-negative GMMA (Fig. S2a). All anti-LPS specific IgG subclasses (IgG1, IgG2a, IgG2b, and IgG3) were elicited upon immunization with both OAg-positive and OAg-negative GMMA (Fig. S3). LPS-specific IgG subclasses were comparable among the two groups immunized with OAg-positive GMMA and the two groups of mice immunized with OAg-negative GMMA respectively, both at study day 27 and 42. Interestingly, only the groups immunized with OAg-negative GMMA were characterised by IgG1/IgG2 ratios < 1 (Fig. S2c).
Protein-specific total IgG elicited by vaccination with OAg-negative GMMA were significantly higher than those elicited by vaccination with the corresponding OAg-positive GMMA both after the first and the second immunization. Moreover, IgG elicited by Ss ΔhtrB OAg+ GMMA were higher than those induced by immunization with Ss ΔmsbB OAg+ GMMA (Fig. 3b).
Also anti-protein specific IgM were higher for the group immunized with Ss ΔhtrB OAg+ GMMA compared to Ss ΔmsbB OAg+ GMMA, whereas antibody levels were comparable between the two groups of mice immunized with OAg-negative GMMA. Anti-GMMA proteins specific IgM were lower after vaccination with OAg-positive GMMA than with corresponding OAg-negative GMMA (Fig. S2b). As for anti-LPS specific IgG subclasses, all anti-GMMA proteins specific IgG subclasses (IgG1, IgG2a, IgG2b, and IgG3) were elicited upon immunization with both OAg-positive and OAg-negative GMMA (Fig. S3). IgG subclasses were comparable among the two groups immunized with OAg-negative GMMA, both at study day 27 and 42, whereas anti-GMMA proteins specific IgG subclasses were higher after immunization with Ss ΔhtrB OAg+ GMMA compared to Ss ΔmsbB OAg+ GMMA. IgG1/IgG2 ratios were < 1 for all the groups. IgG1/IgG2 ratio obtained with Ss ΔhtrB OAg+ GMMA was statistically different from that of Ss ΔmsbB OAg+ GMMA, the latter being lower (Fig. S2d).
Vaccine-elicited antibody functionality. Functionality of the sera was evaluated against an OAg-positive S. sonnei strain through a serum bactericidal assay (SBA) based on luminescence (Fig. 3c) 21 . Similar killing of S. sonnei was observed after immunization with OAg-positive GMMA and no killing after immunization with OAgnegative GMMA. We wondered if this could be due to the fact that anti-proteins antibodies could not bind to the bacterial surface proteins shielded by the OAg chains. Therefore, we also evaluated bactericidal activity of sera against an OAg-negative S. sonnei strain (Fig. 3d). SBA titers for sera raised against OAg-negative GMMA were higher compared to those raised against OAg-positive GMMA. Sera raised against Ss ΔmsbB OAg+ GMMA Vaccine-elicited antibody binding to bacteria. To further confirm that anti-proteins antibodies were not able to bind to the surface of bacteria due to the presence of the OAg chains, the same bacteria used for SBA were stained with pooled sera representing the four immunization groups and analyzed by Fluorescence Activated Cell Sorting (FACS). When OAg-positive bacteria were stained with sera from mice immunized with OAgpositive GMMA, the observed Mean Fluorescence Intensities (MFI) were similar and much higher than the MFI given by staining with sera elicited after immunization with OAg-negative GMMA (Fig. 3e). However, the signal given by sera raised against OAg-negative GMMA was low but detectable, even if the binding of these antibodies did not result in killing of OAg-positive bacteria. This signal may be due to the binding of antibodies directed to the lipid A core (as observed in ELISA, see Fig. 3a). When OAg-negative bacteria were stained with the same sera, the differences among the groups reflected those observed in ELISA and SBA (Fig. 3f).
Identification of the immunodominant protein antigens in GMMA. GMMA surface protein immunoprecipitation (IP) coupled with Mass Spectrometry (MS) was performed with pooled sera from mice immunized with OAg-positive GMMA (Ss ΔhtrB OAg+ or Ss ΔmsbB OAg+ GMMA) to detect and identify the immunodominant protein antigens, therefore the antigens on GMMA surface that are mainly recognized by the two sera. GMMA from Ss OAg− were used for the immunoprecipitations and the precipitated proteins were identified and quantified by LC-MS/MS. A total of 15 outer membrane proteins, which are most likely important for the observed serum bactericidal activity, were identified and quantified in the pulled down samples (Table 3). Of these, 7 proteins were pulled down exclusively with sera raised against Ss ΔhtrB OAg+ GMMA. While among the 8 remaining proteins, 4 antigens were pulled down in greater amounts (at least twofold more) with sera raised against Ss ΔhtrB OAg+ GMMA compared to sera raised against Ss ΔmsbB OAg+ GMMA. Taken together, these results confirmed not only that Ss ΔhtrB OAg+ GMMA elicited higher anti-GMMA proteins antibody titers compared to Ss ΔmsbB OAg+ GMMA, as also seen by ELISA, SBA and FACS analysis (Fig. 3), but also that the antigens recognized by these antibodies are different. Interestingly, the differences observed in the amount of each antigen pulled down by the two sera do not correlate with the amount of the same antigens in the GMMA used to immunize mice, and despite some of the immunodominant proteins were comparably

Discussion
GMMA have been proposed as an attractive platform for the development of an OAg-based vaccine against S. sonnei. S. sonnei GMMA have demonstrated to be able to induce high levels of anti-LPS specific IgG both in animal models and in humans 6,14,17,18 . However, the quality and the relative contribution of protein-specific antibodies to the overall immune response induced by GMMA have never been fully elucidated and dissected from the immune response directed to the OAg. Other groups are working on vaccines against Shigella which are based on protein antigens 8,22 . Here we tested GMMA differing for the presence and the relative density of OAg chains on their surface, while the length of OAg chains was similar among the OAg-positive GMMA. The two alternative detoxification strategies resulted in different OAg densities on GMMA 23 . Indeed, the htrB mutation had a strong impact on the overall OAg amount which was largely decreased compared to the parental strain and the alternative msbB double mutant. On the same line, we verified through proteomic analysis that the majority of proteins are similarly represented in the different GMMA samples. Moreover, the proteins detected and their relative abundance were in line with the results obtained for similar OAg-negative GMMA preparations in previous studies 15,24 , further confirming the reproducibility of GMMA production. When tested in mice, OAg-positive GMMA were able to elicit higher anti-LPS specific antibody titers compared to OAg-negative GMMA, as expected. Anti-LPS total IgG elicited after immunization with OAg-negative GMMA were probably due to the presence of lipid A core on the surface of OAg-negative GMMA as well as in the ELISA plate coating antigen. The group of mice immunized with Ss ΔmsbB OAg+ GMMA received a dose of OAg which was around eight times higher compared to the group of mice immunized with Ss ΔhtrB OAg+ GMMA, but the same protein dose. This difference in the OAg dose did not impact on the anti-LPS response because the tested OAg doses were already at the plateau of the LPS-specific humoral response elicited by S. sonnei GMMA in mice, as verified in previous dose response studies (data not shown). In order to fully elucidate if OAg density can play a role on the anti-LPS response induced, ad hoc studies testing OAg-positive GMMA at lower and not immune-saturating doses, conducted in a dose-response manner, will be necessary and are envisaged together with allometric studies.
On the other hand, despite the same protein dose was used, higher OAg doses corresponded to lower antiprotein responses, suggesting an immuno-interference between OAg and protein antigens. It is known that mice orally inoculated with the same dose of bacteria having reduced LPS synthesis can enhance production Table 3. Proteins identified through immunoprecipitation of GMMA proteins with sera from groups immunized with OAg-positive GMMA. GMMA surface immunoprecipitation was performed with pooled sera from mice immunized with OAg-positive GMMA to detect the protein antigens that are mainly recognized by the two sera and therefore to identify the immunodominant protein antigens in both GMMA preparations. OAg-negative GMMA carrying wild-type lipid A were used for the immunoprecipitation experiments and the precipitated proteins were identified and quantified by LC-MS/MS. A total of 15 outer membrane proteins were identified and quantified in the pulled down samples. NA not applicable. www.nature.com/scientificreports/ of anti-outer membrane proteins antibodies compared to immunization with wild-type bacteria 25 . Moreover, protein immunoprecipitation coupled with mass spectrometry, performed with sera from mice immunized with OAg-positive GMMA, allowed to detect the immunodominant protein antigens in both Ss ΔhtrB and SS ΔmsbB OAg+ GMMA preparations and suggested that OAg density can have an impact on quantity and quality of the immune response against membrane proteins. These results also confirm that GMMA are not only a delivery system for the S. sonnei OAg since they also display additional protein antigens which are strongly immunogenic.
Although an immune correlate of protection against shigellosis has not yet been established, two well characterized assays are available to measure the functionality of the vaccine-induced antibodies against Shigella spp: SBA and opsonophagocytic killing (OPK) 10 .
Even though both OAg-positive and OAg-negative GMMA were able to elicit anti-protein antibodies, sera elicited against OAg-positive GMMA were able to kill S. sonnei in SBA, whereas sera raised against OAg-negative GMMA did not show any complement-mediated bactericidal activity on OAg-positive S. sonnei. Although we cannot rule out the contribution of anti-protein antibodies in OPK, both functional assays follow the same response pattern and have been attributed primarily to antibodies specific for the OAg 26 . In line with this, we observed lack of binding of anti-protein antibodies to OAg-positive bacteria, which is a prerequisite for both complement-mediated or phagocyte-mediated killing. Therefore, both functional assays may not be suitable to assess efficacy for protein-based vaccine candidates. However, by setting up a serum bactericidal assay with an OAg-negative S. sonnei strain, we were able to demonstrate for the first time the bactericidal activity of sera that are not targeting the OAg. Indeed, when the bacterial surface proteins are not shielded from antibody binding by the OAg chains, anti-protein antibodies are able to interact with their targets as proven by FACS analysis and to mediate complement deposition and bacterial killing as proven by SBA. Additional studies looking at cellular immune responses would be beneficial to further explore the role of proteins on GMMA immunogenicity.
From this study we concluded that the OAg is indeed a key antigen for functional immunity against S. sonnei strains, but antibodies against protein antigens are anyway generated upon immunization with S. sonnei GMMA and are functional against S. sonnei when bacterial surface is not shielded by OAg chains. This is an important aspect and an added value of GMMA vaccines, since Shigella pathogenesis is complex 27 , OAg length is modulated in vivo 28,29 , and antibodies against targets other than OAg might be important in protecting against the spread of the infection in specific organs or time of infection.

Materials and methods
Bacterial strains and generation of mutants. Shigella sonnei 53G was chosen as parent strains. The list of Shigella mutant strains used in this study and their abbreviated identifications are listed in Table 1. Loss of the major virulence plasmid pSS was obtained by selecting a colony with white appearance on congo red agar and genotype confirmed by the absence of the origin of replication and the plasmid encoded gene wzy 15 . The null mutation of msbB1, msbB2, htrB, tolR, wbg, virG was obtained by replacing the gene of interest ("gene") with an antibiotic resistance cassette or auxotrofic marker, by homologous recombination using lambda red recombeneering system 30 . The tolR gene was replaced by kanamycin, the msbB2, htrB and wbG genes were replaced by chloramphenicol, the msbB1 gene by erythromycin antibiotic cassettes, respectively. The virG gene was replaced by the nadA and nadB genes from E. coli. Ss ΔmsbB OAg+ strain was generated in this study using the same primer sets specific for tolR, msbB1, and virG genes previously reported 12,14,15,20 in addition to a specific primer set for the replacement of the msbB2 gene with the chloraphenicol resistance cassette from plasmid pKD3 30 (Forward primer: taaaatattaatgatgattatggtaggggcattcgcactaaataatgaaaGTG TAG GCT GGA GCT GCT TC, Reverse primer: acaactagtggaaatacctgtactttataatttcaagggtacgggtccgcCAT ATG AAT ATC CTC CTT AG).
GMMA production and characterization. S. sonnei GMMA were produced and purified as previously described 14 . Total protein content was estimated by microBCA using bovine serum albumin (BSA) as a reference following the manufacturer's instructions (Thermo Scientific), total OAg amount by HPAEC-PAD 31 , purity and particle size was detected by HPLC-SEC/MALLS 32 , lipid A structure was determined by MALDI-TOF MS 12 , extracted OAg was characterized by HPLC-SEC and NMR 33 . Behaviour of different GMMA has been also confirmed by MAT using human PBMC 12 . Despite the presence of antibiotic resistance cassettes in the final GMMA-producing strains, antibiotics are not needed during large-scale growth 14 and possible nucleic acids contaminations in GMMA preparations are monitored and kept below the regulatory acceptance criteria for use in humans.

Proteomic analysis.
For quantitative proteomic analysis, one hundred micrograms of GMMA were TCA precipitated as previously described 34,35 and the protein pellet was resuspended in 50 mM ammonium bicarbonate containing 0.1% (w/v) RapiGest SF (Waters) and 5 mM DTT and heated at 100 °C for 10 min. Digestions were performed overnight at 37 °C with 2.5 μg trypsin (Promega). Digestions were stopped with 0.1% (v/v) formic acid, desalted using OASIS HLB cartridges (Waters) as described by the manufacturer, dried in a Centrivap Concentrator (Labconco) and resuspended in 50 μL of 3% (v/v) ACN and 0.1% (v/v) formic acid. Peptide mixtures were stored at − 20 °C until further analysis.
For each GMMA sample, three (technical replicates) LC-MS/MS acquisitions were performed as previously described 34,35 . The percentage of each protein in the total sample was calculated according to the corresponding peak area (averaged between the three techincal replicates) and the theoretical molecular weight (MW) using the following formula: www.nature.com/scientificreports/ The mass spectrometric raw data were processed with the PEAKS software ver. 8 (Bioinformatics Solutions Inc.) for de novo sequencing, database matching identification and label free quantification, as previously described 35 . Protein identification from MS/MS spectra was performed against S. sonnei 53G database (UniProt code UP000194501-4779 ORF entries) combined with common contaminants (human keratins and autoproteolytic fragments of trypsin) with a FDR set at 0.1%. The localization of the quantified proteins was predicted according to PSORTb software ver. 3 (Brinkman Laboratory, https ://www.psort .org/psort b/).

Assessment of anti-S. sonnei LPS and anti-S. sonnei GMMA proteins immune responses in mice.
Pre-immune sera and sera collected four weeks after the first and two weeks after the second immunization were analysed by ELISA 14 for anti-S. sonnei LPS total IgG, IgG1, IgG2a, IgG2b, IgG3, IgM content using S. sonnei LPS as plate coating antigen (at the concentration of 0.5 µg/mL in Phosphate Buffer Saline, PBS) and for anti-S. sonnei GMMA proteins total IgG, IgG1, IgG2a, IgG2b, IgG3, IgM content using Ss OAg-GMMA as plate coating antigen (at the concentration of 1 µg/ml in PBS).

Assessment of serum bactericidal activity against S. sonnei.
Single sera collected at day 42 were assayed in SBA based on luminescent readout as previously described 36 against OAg-positive S. sonnei strain (Ss WT OAg+) and OAg-negative S. sonnei strain (Ss WT OAg−). Heat inactivated (HI) sera were serially diluted in PBS in the SBA plate (10 µL/well). The starting dilution of each serum in the assay was 1:100 (final dilution) for SBA conducted with S. sonnei OAg-positive strain and 1:500 (final dilution) for SBA conducted with S. sonnei OAg-negative strain, followed by threefold dilutions steps up to 7 dilution points, plus one control well with no sera added. A 4-parameter non-linear regression was applied to raw luminescence (no normalisation of data was applied) obtained for all the sera dilutions tested for each serum; an arbitrary serum dilution of 10 15 was assigned to the well containing no sera. Fitting was performed by weighting the data for the inverse of luminescence 2 and using GraphPad Prism 7 software (GraphPad Software).
Results of the assay are expressed as the IC50, represented by the reciprocal serum dilution that is able to reduce the luminescence signal of 50% compared to the negative control (and thus causes 50% growth inhibition). Titers lower than the minimum measurable assayed were assigned a value of half of the first dilution of sera tested (50 and 250 respectively for SBA assessment of OAg-positive and OAg-negative strains).
Statistical analysis. Significant differences between ELISA responses and SBA titers were evaluated using the non-parametric two-tailed Mann-Whitney test. The comparison was performed between the two groups immunized with OAg-positive GMMA, the two groups immunized with OAg-negative GMMA and between the groups immunized with GMMA sharing the same lipid A structure. FACS analysis. Shigella sonnei strains (Ss WT OAg+ and OAg−) were grown overnight at 37 °C in LB supplemented with 20 µg/mL chloramphenicol. Bacteria were then treated as prevoiusly described 35 . They were pelleted and washed with PBS at 8000×g for 5 min. Bacteria were then blocked with PBS containing 3% (w/v) Bovine Serum Albumin (BSA) for 15 min and incubated with mouse sera diluted in PBS + 1% (w/v) BSA (1:500) for 1 h. After washes, samples were incubated with Alexa Fluor 647 goat anti-mouse IgG (1:500) (Molecular Probes) for 30 min. Finally, bacteria were fixed with 4% (w/v) formaldehyde for 20 min and flow cytometry analysis was performed using FACS Canto II flow cytometer (BD Biosciences).
Protein A/G UltraLink beads (50 µL) were added to the lysates and incubated overnight at 4 °C, following the manufacturer's protocol. Following overnight incubation, beads were washed twice with solubilization buffer %Protein X = Avg Area Protein X * MWProtein X Avg Area Protein * MWProtein