Bacterial strains and plasmids used in this study are listed in Table 1 . Escherichia coli and S. Typhimurium UK-1 derivative strains were routinely cultured in LB broth [19] or on LB agar at 37°C. S. Typhimurium UK-1 mutant strains were supplemented with 50 μg/ml of diaminopimelic acid (DAP), 0.05% arabinose, 0.1% mannose and 0.1% rhamnose when necessary for bacterial growth as described in previous work (29, 52–54). For animal experiments, S. Typhimurium χ12341 was cultured in LB broth with necessary supplements. Overnight cultures were diluted 1:100 into fresh medium and grown with aeration (200 rpm) to an optical density at 600 nm of ~0.85. Bacteria were then centrifuged at 5,000 × g for 15 min at room temperature and resuspended in buffered saline with 0.01% gelatin (BSG) (55). The H. pylori strain SS1 (a kind gift from Prof. James G. Fox, Massachusetts Institute of Technology) was grown on Brucella agar supplemented with 5% sheep blood, 25 μg/ml trimethoprim, 100 μg/ml vancomycin, 3.3 μg/ml polymyxin B, 50 μg/ml amphotericin B, 200 μg/ml bacitracinin and 10 μg/ml nalidixic acid in an anaerobic jar with a microaerophilic gas generating kit (BD, USA) for 5 days at 37°C.

The regulated delayed lysis vector pG8R111 ( Figure 1A ) for synthesis and delivery of antigens has a P_trc-regulated synthesis of encoded protein antigens for delivery by cell lysis and araC P_araBAD-regulated murA and asdA genes with GTG start codons to lessen translation efficiency. The pG8R111 has a weaker SD AAGGCAA to further reduce the production of AsdA. The P22 P_R located with opposite orientation to the transcription of the araC P_araBAD GTG-murA GTG-asdA genes is repressed by the C2 repressor made during the growth of χ12341 with arabinose. C2 concentration decreases due to cell division in vivo to cause P_R-directed anti-sense mRNA synthesis to block translation of residual asdA and murA mRNA. Transcription terminators (TT) flank all plasmid domains for regulated lysis, replication, and gene expression so that expression in one domain does not interfere activities of adjacent domain. pG8R114 ( Figure 1B ) is derived from pG8R111 with a much-improved optimized β-lactmase signal sequence (56), the fusion of molecules to the bla SS in pG8R114 leads to delivery of molecules to the periplasm that results in increased production of outer membrane vesicles (OMVs) and releasing into the supernatant fluid surrounding PIESV cells that enhance induced immune responses (57, 58). We made constructs to synthesize nine putative protective H. pylori antigens according to the antigen gene sequences in H. pylori SS1 (2, 59–67) ( Table 1 ). These constructs were evaluated for synthesis and delivery and protective immunity induction in mice. Antigens included VacA, CagA, UreB, UreA, HpaA, BabA, HopM, Hp-NAP and a chimeric antigen. To design the chimeric antigen with the most antigenic fragments of FliD, UreB, VacA and CagA, bioinformatic tools (65) were used to identify T-cell and B-cell epitopes. Sequences encoding FliD (1–600), UreB (327–385), VacA (744–805) and CagA (51–100) polypeptides were accessed from GenBank. To assist epitope exposure, flexible glycine-serine (GS) linkers were included between the gene segments. Sequences were codon-optimized to have a high-level expression in S. Typhimurium and GC contents adjusted to be closer to that for Salmonella. All genes were synthesized by Biomatik (Cambridge, Ontario, Canada). To detect the synthesis of recombinant proteins, a 6 His-tag sequence was added at the 3′ end of each gene before the stop codon. Based on the presence or absence of a signal peptide encoded in each antigen gene, antigens were divided into two groups. For those antigens without a signal peptide, the synthesized genes were cloned into pG8R111 whereas for those H. pylori antigens with signal sequences, these were removed, and the codon-optimized sequences were inserted into pG8R114 with fusion to the bla SS ( Table 1 ). Considering the results obtained from the initial protection experiments (data not shown), four antigens were selected for further study. The codon-optimized H. pylori genes were fused with P_lpp in front of the second gene and then were cloned under control of the P_trc promoter of pG8R111 or pGR114 ( Table 1 ) ( Figure 1 ). Plasmids carrying genes were finally electroporated into χ12341. To obtain purified recombinant Hp-NAP, UreA, UreB and HpaA, sequences encoding each antigen were cloned into pET28a (+) ( Table 1 ) using NcoI at N-terminal and XbaI at C-terminal sites. pET28a derivatives with inserts were electroporated into E. coli BL21(DE3) ( Table 1 ) for synthesis and purification of gene products.

Measurement of plasmid stability is described previously (55). Briefly, vaccine strains grown overnight (G0) were diluted 1:1000 into pre-warmed fully supplemented LB grown (permissive growth conditions) with aeration for 12 h at 37°C. This process was repeated for approximately 50 generations (G5) and the proportions of cells possessing the Asd⁺ plasmids were determined for each culture. The percentage of clones holding the plasmids from each culture was obtained by counting the colonies grown on LB agar without and with DAP. The ability of these clones to synthesize H. pylori antigens was also checked after 50 generations using western blotting.

S. Typhimurium strain χ12341 carrying recombinant plasmids (PIESV-Hp) or empty vectors (PIESV-empty vector) were cultured in LB containing 0.05% arabinose, 0.1% mannose and 0.1% rhamnose at 37°C. LB containing kanamycin 50 μg/ml was used to culture E. coli BL21 carrying recombinant plasmids as well. When the bacteria reached an OD600 of 0.6, 1 mM IPTG was added to the cultures to induce heterologous H. pylori protein syntheses. Protein samples from bacterial cultures were separated and analyzed by SDS-polyacrylamide gel electrophoresis with transfer to nitrocellulose membranes as previously described (68). Recombinant proteins were detected by anti-6xHis peroxidase (Roche, Basel, Switzerland) (1/40,000 dilution) for 2 h. E. coli BL21carrying pG8R289 (ureA-6xHis), pG8R290 (ureB-6xHis), pG8R291 (hpaA-6xHis) or pG8R292 (napA-6xHis) were grown overnight at 37°C in LB broth supplemented with 50 μg/ml kanamycin. The procedures for protein synthesis and purification have been described in our previous study (4). Prestained Protein Ladder, 10 to 180 kDa (Thermofisher, USA) was used as the protein marker.

All animal experiments were approved by the University of Florida Institutional Animal Care and Use Committee. Six-to eight-week-old SPF female C57BL/6 mice (n=10/group), were purchased from Charles River Laboratories (Wilmington, MA). Mice were acclimated for one week after arrival before starting the experiments.

Immunization procedures followed the previous description (4, 68, 69). Briefly, food and water were not given to mice for 6 h prior to the immunization and re-supplied 30 min later. Mice were orally immunized with 20 μl BSG containing 10⁹ CFU of each vaccine strain, combination of strains, or with 20 μl BSG alone as the negative control on day 0 and boosted on days 14 and 28. Blood samples were collected individually on days 0, 14, 28, 42 and 72 and serum collected individually for analysis of antibody responses. Mice (10 mice) were immunized with 10⁹ CFU of each PIESV synthesizing and delivering different antigens. A mixture of PIESV strains delivering pG8R111 and pG8R114 (PIESV-empty vectors) were used as controls.

To assess whether the immunization of mice with Salmonella delivering H. pylori antigens was able to reduce bacterial burden of H. pylori in the stomachs of infected mice, mice were infected with 5×10⁸ H. pylori SS1 thrice at one day intervals two weeks after the last immunization. Four weeks post infection, mice were euthanized and their stomachs removed and H. pylori CFUs quantified. To gain CFUs, the quantitative bacterial culture of mouse stomach was used. Briefly, a half section of the stomach from each euthanized mouse was completely homogenized, serially diluted, and then plated on selective medium as described above.

An enzyme-linked immunosorbent assay (ELISA) was used to specifically evaluate serum total IgG, IgG1 and IgG2c titers in immunized mice. 96-well polystyrene plates (Nalge Nunc. Rochester, NY, USA) were coated with purified recombinant HpaA, UreA, UreB and Hp-NAP (5 mg/ml). After over-night incubation, the plates were washed three times with TBST buffer (Tris-buffered saline, pH 7.4, containing 0.05% Tween 20) and followed by blocking with 300 µl PBS containing 10% FBS for 2 h at 37°C. After adding serial dilutions of mouse sera to the plates, they were incubated for 2 h at room temperature and then washed with TBST. HRP-conjugated goat-anti-mouse IgG1, IgG2c or total IgG (Southernbiotech, Birmingham, AL, USA) were added to the wells and incubated for 90 min at 37°C. Following the last washing step, specific reactivity was calculated by the addition of 50 µl/well of the TMB substrate (Thermofisher, USA). The reaction was stopped by adding 15 µl of 2 M H₂SO₄. Next, the absorbance at 450 nm was measured. To define a cut-off value for the test, the mean OD plus three-fold SD of sera from mice immunized with BSG was calculated at a 1:100 dilution. ELISA titers were calculated as the reciprocal of the last serum dilution providing an OD higher than the cut-off (70). To assess stomach mucosal IgA production, the secretory IgA was extracted by incubation of mouse stomachs in PBS containing 5% non-fat dry milk, 4-(2-aminoethyl)-benzenesulphonyl fluorid (Calbiochem), 1 mg/ml 3.25mM aprotinin, 10 mM leupeptin (Sigma), and bestatin (Boehringer-Mannheim Biochemicals). After extensive vortexing and centrifugation (16,000 g for 10 min), the supernatants were used for the determination of antibody titers. An ELISA was used as described above.

Mouse stomachs were obtained 75 days post the first immunization. Stomach tissues were fixed in 10% neutral buffered formalin for at least 24 h. The trimmed tissues were processed, paraffin-embedded, sectioned at 5 μm, and stained with hematoxylin and eosin. Selected sections of stomach tissues were also evaluated with Warthin-Starry and Gram stains to identify bacteria. The pathologist was blinded as to the experimental groups and treatments of the study. Histologic samples of both glandular and squamous portions of the stomach were examined and evaluated and scored based on intensity 0 to 3 for criteria including mucosal inflammation and type, submucosal inflammation and type, non-H. pylori bacteria, mucosal ulceration, and hyperkeratosis of squamous stomach.

SPSS computer software, version 17.0. was used to do statistical analysis. The statistical differences between two groups were studied by t-test and results among several groups were obtained by one factor analysis of variance (ANOVA) and Tukey’s post hoc test. The statistical border for accepting significance was P< 0.05. Data were envisaged using the GraphPad Prism 7 program.

The construction and properties of χ12341 have been recently described (29). Three sugars, arabinose, mannose and rhamnose which are not present in animal tissues, were used to regulate the virulence trait. The ΔPmurA::TT araC P_araBAD murA and ΔasdA::TT araC P_araBAD c2 mutations regulate synthesis of diaminopimelic acid and muramic acid, two essential constituents of peptidoglycan, which enable PIESV lysis in the absence of DAP and inability to synthesize the MurA enzyme without arabinose in vivo (53). The Δpmi ΔwaaL ΔpagL::TT rhaRS P_rhaBAD waaL mutations collectively provide regulated delayed attenuation and cause cessation in the synthesis of LPS O-antigen in the absence of the sugars mannose and rhamnose in vivo (52, 74, 75). The deletion Δ(wza-wcaM) enhances complete lysis, blocks the synthesis of cell surface polymers to prevent biofilm formation and enhances induced immunities (76). The ΔrelA::araC P_araBAD lacI TT mutation confers the regulated delayed antigen synthesis phenotype that enhances PIESV colonization and induction of immune responses by the arabinose-dependent regulated LacI synthesis so that P_trc-regulated genes on plasmid vectors are gradually expressed as LacI is diluted at each cell division during in vivo growth (77–79). The ΔrecF mutation decreases inter- and intra-plasmidic recombination to stabilize PIESVs (80). The ΔsifA mutation enables PIESV escape from the Salmonella-containing vesicles (SCVs) to cause PIESV cells to lyse in the cytosol (81, 82). This enables delivery of protective antigens to the proteasome for class I presentation to generate CD8-, CD17- and NKT-dependent immunities.

Codon-optimized sequences encoding all nine H. pylori antigens were successfully inserted into the plasmid vectors pG8R111 or pG8R114 ( Figures 1A, B ) to yield the plasmids listed in Table 1 . The synthesis of each antigen was confirmed by western blot using a monoclonal antibody against the 6His-tag ( Figure 1C ) to yield the plasmids listed in Table 1 . Results of the plasmid stability experiment showed that all recombinant plasmids encoding antigens except HopM were stable in the Salmonella vaccine strain. However, only 30% of Salmonella cells lost the plasmid specifying HopM synthesis after 50 generations of growth under permissive conditions. Based on results obtained from initial protection experiments, UreA+UreB (pG8R230) and HpaA+Hp-NAP (pG8R262) ( Figures 1D, E ), the set of antigens providing considerable protection, were selected to be expressed by pG8R111 and pG8R114, respectively. χ 12341(pG8R230) and χ 12341(pG8R262) specified synthesis of the four encoded antigens after IPTG induction ( Figure 1F ). We also designed and expressed a chimeric antigen including protective parts of four four putative antigenis (FliD, VacA, CagA and UreB) of H. pylori using pG8R111 in PIESV χ 12341 ( Figure 1G ).

To determine whether immunization with vaccine candidates lowers the bacterial burden in the stomachs of infected mice, we ascertained CFUs of H. pylori using quantitative bacterial culture procedures. Immunized mice were infected with H. pylori SS1 two weeks after the last immunization. Then, four weeks after challenge, the stomachs of euthanized mice were removed, minced, homogenized, serially diluted and then plated on the selective agar medium. As shown in Table 2 ; Figure 2A , higher levels of protection were observed in mice immunized with the Salmonella vaccine PIESV-Hp strains carrying pG8R230(UreA+UreB) and pG8R262(HpaA+HpNAP). Notably, seven out of 10 immunized mice in this group showed sterile protection and the other three mice showed a significant reduction in a load of bacteria compared to mice receiving BSG or empty vector ( Figure 2A ).

To study the humoral response against antigens in mice immunized with PIESV vector strains delivering four antigens, sera were obtained at several time points over three months following initial immunization. Immunization of mice induced strong and specific immunoglobulin G (IgG) responses against each component of the cocktail vaccine including pG8R230(UreA+UreB) and pG8R262(HpaA+HpNAP), in which IgG2c (Th1-related isotypes) titers were usually slightly higher than those of the IgG1 subtype ( Figure 2B ). The total IgG and IgG2c titers against each antigen started to increase during the second week after the first immunization and peaked after 75 days. However, IgG1 titers peaked at 30 days post first immunization and titers fell after 75 days. These results indicated that immunization of mice with Salmonella vaccine vectors carrying pG8R230(UreA+UreB) and pG8R262(HpaA+HpNAP) elicit a Th1-biased immune response. To evaluate whether immunization with PIESV strains delivering these four antigens also provoked mucosal immune responses, gastric IgA production was evaluated for each antigen in stomachs of immunized mice. As shown in Figure 2B , the immunized mice with the cocktail of four antigens significantly augmented gastric mucosal IgA titers against each component of the cocktail vaccine. These findings indicate that immunization of mice with a combination of pG8R230(UreA+UreB) and pG8R262(HpaA+HpNAP) synthesized and orally delivered by two PIESV vector strains provokes both specific systemic and mucosal humoral immune responses.

Antimicrobial peptide plays a important role in immune responses to H. pylori (83, 84). To obtain a deeper understanding of the roles of antimicrobial peptides and immunological markers in the protection provided in immunized mice, ten days after infecting the tested mice with H. pylori SS1, mouse stomachs and spleens were obtained. Reg3a, Reg3b, CXCL1, CXCL2, and CXCL5 in stomach tissues were investigated by qPCR. These chemokines and antimicrobial peptides were selected since they have been shown to work in correlation with IL-17 (85, 86). Spleen samples were further analyzed for T-cell markers and cytokines. In the stomachs of immunized mice after infection with H. pylori SS1, the expression of CXCL2 increased two-fold while the expression of the remaining genes did not change ( Figure 4A ).

In response to H. pylori SS1 infection, IFN-γ+ CD4 T cells increased in the spleens of immunized mice compared to the controls. Additionally, no significant changes in TNF+ or IL-17A+ CD4 T cells in the spleen for any test groups were seen. There were also no significant changes in IFN-γ+, TNF+ or GranzymeB+ CD8 T cells in the spleen for any test group ( Figure 4B ).

The stomach tissues were evaluated and scored based on severity 0 to 3 for criteria including mucosal inflammation and type, submucosal inflammation and type, presence of non-H. pylori bacteria, mucosal ulceration, and hyperkeratosis of squamous portions of the stomach. The scale 0-3 refers to 0=normal, 1=mild, 2=moderate, 3=severe. The inflammation ranged from nominal to severe both in the mucosa and submucosa of both the glandular and squamous portions of the stomach and consisted of lymphocytes, plasma cells with or without smaller numbers of neutrophils. Mice in control group had large amounts of hyperkeratosis in the squamous portions of the stomach which are often associated with moderate to large numbers of large, Gram-positive, rod-shaped bacteria on the surface or within the laminated keratin. The hyperkeratosis and associated bacteria are suggestive of hyporexia or anorexia for a relatively prolonged period of at least 2 days. In anorectic rodents, the build-up of excess keratin is presumably caused by reduced mechanical removal by the passage of food (87). The hyperkeratosis was not seen in a single area as suggested at the limiting ridge, but throughout all of the squamous lined areas of the forestomach.

However, significant weight loss was not seen in any group of mice. Overall, the mice receiving the PIESV-empty vector had relatively higher inflammation and numbers of non-H. pylori bacteria (gram-positive rod-shaped bacilli) compared to immunized mice. These findings might suggest the role of other gram-positive bacteria as well as H. pylori infection in the development of stomach complications. Additionally, these findings demonstrated that the vaccine candidate not only protects against H. pylori SS1 but also against other non-H. pylori flora, which may have a role in stomach complications ( Figure 5 ).