Live Recombinant Salmonella Typhi Vaccines Constructed to Investigate the Role of rpoS in Eliciting Immunity to a Heterologous Antigen

We hypothesized that the immunogenicity of live Salmonella enterica serovar Typhi vaccines expressing heterologous antigens depends, at least in part, on its rpoS status. As part of our project to develop a recombinant attenuated S. Typhi vaccine (RASTyV) to prevent pneumococcal diseases in infants and children, we constructed three RASTyV strains synthesizing the Streptococcus pneumoniae surface protein PspA to test this hypothesis. Each vector strain carried ten engineered mutations designed to optimize safety and immunogenicity. Two S. Typhi vector strains (χ9639 and χ9640) were derived from the rpoS mutant strain Ty2 and one (χ9633) from the RpoS+ strain ISP1820. In χ9640, the nonfunctional rpoS gene was replaced with the functional rpoS gene from ISP1820. Plasmid pYA4088, encoding a secreted form of PspA, was moved into the three vector strains. The resulting RASTyV strains were evaluated for safety in vitro and for immunogenicity in mice. All three RASTyV strains were similar to the live attenuated typhoid vaccine Ty21a in their ability to survive in human blood and human monocytes. They were more sensitive to complement and were less able to survive and persist in sewage and surface water than their wild-type counterparts. Adult mice intranasally immunized with any of the RASTyV strains developed immune responses against PspA and Salmonella antigens. The RpoS+ vaccines induced a balanced Th1/Th2 immune response while the RpoS− strain χ9639(pYA4088) induced a strong Th2 immune response. Immunization with any RASTyV provided protection against S. pneumoniae challenge; the RpoS+ strain χ9640(pYA4088) provided significantly greater protection than the ISP1820 derivative, χ9633(pYA4088). In the pre-clinical setting, these strains exhibited a desirable balance between safety and immunogenicity and are currently being evaluated in a Phase 1 clinical trial to determine which of the three RASTyVs has the optimal safety and immunogenicity profile in human hosts.


Introduction
Vaccines based upon live, attenuated pathogens were originally developed to prevent infection by homologous pathogens. The success of these vaccines ((BCG, oral polio (Sabin), MMR and Ty21a)) raised the possibility that a live, attenuated pathogen could be used to stimulate immunity not only against itself, but also heterologous pathogens by expressing foreign antigens. This idea is particularly attractive for pathogens that require a complex or mucosal immune response for protection, since other vaccination strategies are not as effective at stimulating these kinds of immune responses. The success of the chemically attenuated S. Typhi Ty21a vaccine [1] indicated the feasibility of developing rationally attenuated, highly immunogenic live S. Typhi vaccines.
Early success with live attenuated homologous bacterial vaccines led to the development of strategies to modify the vaccine strains for use as vectors to deliver a variety of protective antigens via the mucosal route. This is particularly advantageous for pneumococcal protein antigens as delivery via the mucosal route elicits secretory as well as serum antibody responses, resulting in protection against carriage as well as systemic disease [2,3,4]. Recombinant attenuated Salmonella enterica serovar Typhimurium vaccines (RASVs) expressing the pneumococcal antigen gene pspA have been constructed and demonstrated to induce protective immunity in mice against challenge with virulent S. pneumoniae [5,6]. Translating what works in S. Typhimurium for mice into a S. Typhi-based vaccine for humans has led to mixed results. Although DgalE [7], DaroC, DaroD [8], Dcya Dcrp [9] and DphoPQ [10,11] mutants of S. Typhimurium were completely attenuated in mice and induced protective immunity to challenge with wild-type S. Typhimurium, S. Typhi DgalE [12], DaroC DaroD and Dcya Dcrp [13] mutants were not adequately attenuated and caused significant reactogenicity in humans. It thus became necessary to devise new strategies to overcome reactogenicity issues. For example, the strains such as (Dcya Dcrp-cdt) [14,15], CVD908-htrA [16] and an DaroC DssaV S. Typhi Ty2 [17] are reported to be safe and immunogenic in humans. However, a phoP mutant of S. Typhimurium, completely attenuated in mice [10], led to development of a safe and immunogenic DphoPQ S. Typhi vaccine strain [18].
Although RASVs have been successfully used to express recombinant antigens and induce protective immunity in animals [19], recombinant attenuated S. Typhi vaccines (RASTyV; vaccine strains synthesizing heterologous antigens) have not been sufficiently effective in human studies to justify commercial development [20]. There are several reasons for the lack of success using live S. Typhi vectors in humans. The major difficulty has been in establishing a balance between eliciting a strong immune response to the vectored antigen and reactogenicity. Nonreactogenic strains are often hyperattenuated and induce poor immune responses, even to Salmonella antigens [14,18]. It is likely that the recombinant S. Typhi strains do not colonize lymphoid tissues such as the gut-associated lymphoreticular tissues (GALT), mesenteric lymph nodes and spleen to a sufficient level to stimulate a robust immune response against the foreign antigen. The host immune system may also be ''distracted'' by the Salmonella carrier, mounting a strong response against Salmonella antigens instead of the vectored antigen.
Recently, we have begun to understand another possible reason for the poor results with live RASTyV strains expressing protective antigens from a diversity of pathogens. It was reported that rpoS mutations attenuate S. Typhimurium [21] and this led shortly thereafter to the demonstrated role of RpoS in regulating genes on the S. Typhimurium virulence plasmid [22,23]. S. Typhimurium strains with rpoS mutations have markedly diminished abilities to colonize the Peyer's patches in mice [24]. Virulence plasmid-cured S. Typhimurium strains colonize Peyer's patches as well as wild-type strains [25], indicating that chromosomal genes controlled by rpoS must be of critical importance for Peyer's patch colonization. We thus theorized that rpoS mutations not only attenuate Salmonella but also likely lessen their immunogenicity. This fits with past experiences when it was reported that S. Typhi Ty2, which had been used by most groups constructing RASTyV strains to evaluate in humans, was an rpoS mutant [26]. When administered at high doses, S. Typhimurium rpoS mutants can protect against homologous, but not heterologous challenge [26,27]. In our own unpublished studies, we examined the immunogenicity of rpoS mutants. In a dose response study, we found that introduction of an DrpoS mutation reduces the protective efficacy of a live S. Typhimurium strain (unpublished results). Introduction of a regulated rpoS expression cassette [28] into a recombinant attenuated S. Typhimurium vaccine strain results in a substantial reduction in the immune response directed against the vectored antigen (unpublished results). Therefore, we postulated that an RpoS + S. Typhi RASV strain would be more effective at inducing protective immune responses to recombinant antigens than an RpoS 2 S. Typhi (e.g. Ty2) vaccine strain.
Based on our current knowledge about the RpoS phenotypes of S. Typhi strains, it can be inferred that attenuated RpoS + S. Typhi strains are more virulent in humans than equivalent RpoS 2 S. Typhi strains with the same attenuating mutations [13]. Use of RpoS + S. Typhi antigen delivery vectors will therefore necessitate degrees of attenuation in excess of what is necessary to attenuate S. Typhi Ty2 strains for humans, but these means of attenuation must not compromise immunogenicity.
S. Typhimurium infection in mice is often used as a model of typhoid fever, due to the host-restriction of S. Typhi and the absence of a small animal model to evaluate typhoid fever caused by S. Typhi. We have recently developed a number of new systems in S. Typhimurium designed to enhance the safety and immunogenicity of RASVs, including regulated delayed in vivo attenuation [28,29], regulated delayed antigen synthesis [30] and regulated cell lysis [31]. We have also modified antigen, repressor and plasmid vector sequences to enhance transcription and translation efficiencies to maximize protective antigen synthesis in vivo. In addition, we have also explored other means to enhance immunogenicity and safety, including the DsopB mutation, which reduces Salmonella-induced fluid secretion in the intestines [32,33,34] and enhances the immunogenicity of a vectored antigen [35]. Many of these features were combined to create S. Typhimurium strain x9558 [34]. In addition, this strain also has a DasdA mutation, allowing use of the antibiotic resistance-free Asd + balanced-lethal plasmid maintenance system [36]. When complemented with an Asd + plasmid that directs expression of the pneumococcal gene pspA, x9558 has been shown to be safe, immunogenic and capable of eliciting an immune response that protects against challenge with virulent Streptococcus pneumoniae, in both adult [37,38] and infant [34,39] mice.
S. pneumoniae is a formidable bacterial pathogen, causing disease with high morbidity and mortality even in regions where antibiotics are readily available. S. pneumoniae is estimated to kill 1-2 million children under the age of 5 years each year in developing countries, accounting for 20-25% of all deaths in this age group [40]. Antibodies to pneumococcal capsular polysaccharides can protect against fatal infection, but this protection is serotype specific. Current vaccines based on capsule polysaccharides therefore provide coverage only for the specific serotypes included in the vaccine. There are 91 distinct capsular serotypes of S. pneumoniae [41,42], and geographic differences in serotype prevalence have resulted in suboptimal protection in many countries. Recent reports have shown that although carriage of vaccine serotypes was reduced in immunized individuals, the vacated niche was promptly occupied by non-vaccine serotypes [43,44,45]. This ''replacement carriage'' has translated into a significant increase in cases of invasive disease caused by non-vaccine serotypes in conjugate vaccine recipients [46]. Because of these concerns, we focused our attention on developing a vaccine based on pneumococcal proteins, such as PspA, that contribute to virulence and are common to all serotypes [47].
In this work, we constructed three new recombinant attenuated S. Typhi vaccines (RASTyV) derived from S. Typhi Ty2, its RpoS + derivative [48] and ISP1820 expressing pspA from an Asd + expression plasmid. We introduced a constellation of mutations nearly identical to those present in S. Typhimurium strain x9558 [34,38]. In particular, two mutations, Dpmi and DtviABCDE, affecting synthesis of two major S. Typhi surface antigens, O-antigen and Vi capsule, respectively, were included in an attempt to reduce the host immune response to Salmonella, and, hopefully, to enhance the immune response to the vectored antigen, PspA. We evaluated each strain for safety, persistence in the environment and in human blood and immunogenicity in mice.

Characterization of S. Typhi strains in vitro
All the S. Typhi constructions were characterized for type I fimbriae production in static LB broth cultures [59] by yeast agglutination [60] and for motility in motility medium (bioMériux, Marcy I'Etoile, France). Lipopolysaccharide (LPS) profile was evaluated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis then visualized by silver staining [61]. Plasmid profiles were verified by alkaline lysis and agarose gel (0.5%) electrophoresis [62]. Biochemical profiles were determined in the presence and absence of 0.2% arabinose using the API 20E system (bioMériux). The presence of Vi, O and flagella antigens were evaluated by agglutination assays. The presence or absence of RpoS was determined by addition of hydrogen peroxide to cultures to detect activity of the RpoS-dependent catalase, KatE [63,64], and by glycogen accumulation when streaked on Kornberg glycogen indicator agar [65]. The presence of each chromosomal mutation was verified by PCR using the appropriate DNA primers ( Table 3). Stability of Asd + plasmids in vaccine strains was determined as described [66].

SDS-PAGE and immunoblot analyses
To evaluate arabinose dependent regulation of gene expression, strains were grown in NB medium with 0.05% arabinose at 37uC with aeration. When the culture reached an OD 600nm of 0.8, it was diluted 1:10 into fresh NB without arabinose and grown to an OD 600 of 0.8. This process was repeated 6 times (,20 generations). One ml of culture was collected from each passage and prepared for western blot analysis [54]. Total protein was normalized, samples were separated on a 10% (wt/vol) sodium dodecyl sulfate-polyacrylamide gel and transferred onto nitrocellulose membranes. Fat-free milk powder dissolved in phosphate buffered saline (PBS) (5%, wt/vol) supplemented with 0.05% Tween 20 (PBS-T) was used for blocking. The membrane was incubated with an appropriate anti-rabbit polyclonal antibody (1:10,000), anti-Fur, anti-Crp, and anti-LacI [28], anti-PspA [30], anti-GroEL (Sigma, St. Louis, MO, USA) or anti-mouse monoclonal antibody (1:1,000) anti-RpoD (Neoclone, Madison, WI, USA) for 1 h at room temperature, washed three times with PBS-T, and then incubated with a 1:10,000 dilution of alkaline phosphatase-conjugated anti-rabbit immunoglobulin G (IgG) Table 3. Oligonucleotides used to verify the genotype of the RASTyV strains. F: forward primer; R: reverse primer.  (Southern Biotech, Birmingham, AL). Color was developed with nitroblue tetrazolium and 5-bromo-4-chloro-3-indolylphosphate (BCIP) (Sigma).

Salmonella subcellular fractionation
Cultures were grown in NB to an OD 600 of 0.6 and centrifuged at 5,8006g for 10 min. Periplasmic fractions were prepared by a modification of the lysozyme-osmotic shock method [67] as previously described [6]. The supernatant fluid was saved for analysis of secreted proteins. Equal volumes of outer membrane proteins (OMPs), periplasmic, cytoplasmic, and supernatant fractions and total lysate samples were separated by SDS-PAGE for western blot analysis.

Deoxycholate sensitivity assay
RASTyV strains and wild-type Salmonella Typhi strains were grown in KT broth to a cell density of approximately 1610 9 CFU/ml (late log phase). The growth media for the RASTyV strains were supplemented with 0.2% arabinose and 0.2% mannose. Cultures were harvested by centrifugation at 25006g for 20 min at room temperature. Cells were resuspended in PBS and allowed to acclimate for one hour. Following acclimation, the cells were pelleted and the PBS was removed. Cells were transiently exposed to sodium deoxycholate in PBS (0, 0.1, 0.25, 0.5, 1.0, 2.5 and 5.0%) for 2 h at 37uC (final concentration 1610 9 CFU). The number of viable cells after deoxycholate treatment was assessed by duplicate plating on LB agar+0.2% arabinose. This experiment was repeated three times.

Survival of Salmonella strains in chlorinated water, surface water and in raw sewage
Chlorinated water was prepared by dissolving a crushed chlorine pool tablet into distilled water to a final concentration of 4,5 ppm (chlorine concentration was determined by the SenSafe chlorine testing kit (Industrial Test Systems, Rock Hill, SC, USA)). Chlorinated water was used within 10 min of preparation. Frozen stocks were thawed at 37uC, and then residual phytone-glycerol was removed by washing cells with PBS. Cells were suspended in 1 ml of PBS to a concentration of 160.2610 9 CFU/ml. The actual titer was measured by plating serial dilutions onto LB agar+0.2% arabinose. 1610 9 CFU of cells were diluted into 19 ml of chlorinated water, and viability was assessed by plating at 10, 30 and 60 min after the addition of chlorine. Samples were prepared in triplicate and plated in duplicate at all time points.
Raw sewage samples were obtained from the Kyrene Water Reclamation Facility in Tempe, AZ and were used within one week of collection. Untreated surface water samples were obtained from Tempe Canal in Tempe, AZ and were used within one week of collection. The sewage was diluted 1:1000 in PBS prior to use to reduce the number of background organisms to ,1610 5 CFU/ml. S. Typhi wild-type and RASTyV strains from phytone-glycerol frozen stocks were thawed, and the cells washed with PBS, then inoculated into triplicate 20 ml surface water or sewage samples to a concentration of 1610 8 CFU/ml. Samples were incubated at ambient temperature (21-25uC) for the duration of the assay. Viability of the Salmonella strains was assessed on days 0, 1, 3, 7 and 10 after inoculation by plating on MacConkey agar+1% lactose and % mannose. For untreated surface water cultures containing wild-type ISP1820 or x9633(pYA4088), brilliant green agar (Difco) containing 0.2% arabinose and 0.2% mannose was used in place of MacConkey agar. Each sample was plated in duplicate for each time point.
Plates were compared to pure cultures streaked on identical media. All colonies suspected to be Salmonella were patched onto Bismuth-sulfite, XLT-4 and MacConkey agar+1.0% maltose+/ 20.2% arabinose for verification. The limit of detection for this assay was 1610 4 CFU/ml.

Human Subjects
Individuals who had received a vaccine directed against S. Typhi were excluded from participating in the study. Up to 50 ml of blood was collected from individual volunteers by venipuncture and treated with sodium heparin to prevent coagulation [68]. Samples were used as whole blood or processed to isolate monocytes within 24 h of collection.
Survival of Salmonella strains in whole human blood and in human peripheral monocytes RASTyV strains containing pYA4088, wild-type Salmonella Typhi strains and Ty21a were grown in KT broth to a cell density of approximately 1610 9 CFU/ml (late log phase). The growth media for the vaccine strains were supplemented with 0.2% arabinose and 0.2% mannose. Ty21a was grown in the presence of 0.05% galactose (this concentration was found to allow complete LPS O-antigen synthesis without reducing the strain growth rate). In some experiments, blood components were inactivated by incubation at 55uC for one hour. For survival of Salmonella strains in whole human blood, cells were diluted in phosphate buffered saline (PBS), pH 7.4 and inoculated into untreated or heat-inactivated fresh human blood to a density of approximately 1610 6 CFU/ml. Inoculated samples and uninoculated controls were incubated at 37uC for the duration of the assay. Bacterial viability was assessed by plating onto LB agar with 0.2% arabinose at 0, 3, 6 and 18 h after inoculation. The assay was performed in duplicate and was repeated 3 times using blood samples from different individuals. The limit of detection for this assay was less than 10 CFU/ml.
For intracellular survival of Salmonella in human peripheral monocytes, peripheral blood monocytes were isolated from human blood by Ficoll-Paque PLUS (GE Healthcare) gradient [69]. Monocytes were washed with Hank's buffered salts solution (Invitrogen, Carlsbad, CA, USA), resuspended in RPMI 1640 (Invitrogen) without serum at a concentration of approximately 5610 5 cells/ml and allowed to adhere to Corning Costar 24-well plates for 1 h at 37uC, 5.0% CO 2 . Non-adherent cells were removed after 1 h by washing with PBS and RPMI 1640 without serum was added to the monolayer. The bacterial invasion assay was carried out essentially as described [70]. Salmonella strains were suspended in RPMI 1640 at a concentration of approximately 5610 6 CFU/ml and allowed to invade the human monocytes for 1 h at 37uC in 5% CO 2 (MOI of 10:1). After one hour, cells were treated with 100 mg/ml gentamicin to kill extracellular bacteria. The number of intracellular bacteria was assessed at 1, 4 and 24 h after inoculation. Monocytes were lysed with 0.1% sodium deoxycholate for 5 min and samples were serially diluted in PBS and plated on LB agar+0.2% arabinose. This assay was performed in duplicate and was repeated three times using cells isolated from different individuals.

Sensitivity of RASTyV strains to guinea pig serum complement
The complement sensitivity assay was carried out essentially as described [71]. Briefly, RASTyV strains with pYA4088 and wild-type Salmonella Typhi strains were grown in KT broth to a cell density of approximately 1610 9 CFU/ml (late log phase).
The growth media for the RASTyV was supplemented with 0.2% arabinose and 0.2% mannose. Cells were diluted to approximately 1610 6 CFU/ml in PBS and were exposed to 22% purified guinea pig complement in PBS (Calbiochem, San Diego, CA, USA) in the presence or absence of Group D 1 LPS O-antigen antibody (BD Bioscience, Franklin Lakes, NJ, USA). Reactions were incubated 3 h at 37uC. Complement-resistant cells were enumerated by plating on LB agar+0.2% arabinose. The assay was conducted in duplicate, and was repeated a minimum of 3 times for each strain.

Mice
Female and male BALB/c mice, 6 to 7 weeks old, were obtained from the Charles River Laboratories. Mice were acclimated for 7 days after arrival before starting the experiments. Newborn mice (less than 24 hours old) were obtained from pair-wise mating of female and male BALB/c mice.

Distribution of Salmonella bacteria in newborn mice
Groups of newborn mice were orally inoculated with 10 ml containing 1610 9 CFU of Salmonella vaccine strains, wild-type strains, or Ty21a. Mice were euthanized and necropsied at various times. Spleen, liver, and whole intestines were collected on days 3 and 7 post-infection. Tissues were weighed and homogenized in a final volume of 1 ml buffered saline with gelatin (BSG) [72] and serial dilutions plated onto MacConkey agar plates containing 1% lactose, and with or without 0.05% arabinose and 0.2% mannose to determine the number of viable bacteria. Plates were incubated at 37uC for at least 18 h. We also used 900 ml of homogenized tissues to inoculate 5 ml tetrathionate broth (Difco) for Salmonella enrichment. Samples that were negative by direct plating and positive by enrichment were recorded as 10 CFU/g. Samples that were negative by both direct plating and enrichment were recorded as 0 CFU/g.

Immunization of mice
RASTyV vector strains harboring plasmids pYA4088 (pspA) or pYA3493 (empty vector) were grown in LB broth with 0.05% arabinose and 0.2% mannose overnight at 37uC as standing cultures that were diluted 1:100 in the same medium, prewarmed, and grown with aeration (180 rpm) at 37uC to an OD 600 of 0.8 to 0.9. Bacteria were collected by centrifugation at room temperature and resuspended in BSG to densities appropriate for the inoculation route and dose. To determine the actual CFU/dose, serial dilutions of the RASTyV strains were plated onto MacConkey agar supplemented with 1% lactose, with or without 0.05% arabinose and 0.2% mannose. Seven-week-old mice were inoculated intranasally with 160.2610 9 CFU of RASTyV vector strains carrying either the pspA expression plasmid pYA4088 or control plasmid pYA3493 in 10 ml. Mice were boosted with the same dose of the same strain six weeks later. Blood was obtained by mandibular vein puncture at biweekly intervals. Blood was incubated at 37uC for 60 min and the clot was pelleted by centrifugation. Serum was removed from the whole-blood samples and stored at 220uC. Vaginal-wash samples were collected at biweekly intervals and stored at 220uC as described [6]. Sera were collected 2, 4, 6 and 8 weeks after the initial vaccination and serum IgG responses to rPspA, S. Typhi LPS and S. Typhi OMPs were measured. This experiment was performed twice, with each group (8 mice) receiving approximately the same dose of vaccine, and the results from both experiments were similar and have been pooled for analysis.

Antigen preparation
PspA protein was purified as described previously [6]. S. Typhi LPS was obtained from Sigma. Serovar Typhi outer membrane proteins (SOMPs) were purified from S. Typhi strain x9633 as described previously [6]. The PspA-Rx1 clone was a kind gift from Susan Hollingshead at the University of Alabama at Birmingham.

Enzyme-linked Immunosorbent Assay (ELISA)
Sera from all mice in a group were pooled for analysis. ELISA was used to measure IgG antibodies against S. Typhi LPS, SOMPs and rPspA in serum, IgG1 and IgG2a in serum and IgA in vaginal washes against rPspA as previously described [6,38]. Absorbance was recorded at 405 nm using an automated ELISA plate reader (model EL311SX; Biotek, Winooski, VT). Absorbance readings that were 0.1 higher than BSG control values were considered positive [38].

IL-4 and IFN-c ELISPOT assays
At week 7, spleen cells were harvested from 3 mice per group. Cells from each spleen were assayed by ELISPOT in triplicate wells as previously described [73]. Briefly, PVDF membrane plates (Millipore) were washed with sterile H 2 O, and coated with 100 ml of anti-IL-4 or anti-IFN-c mAbs (BD PharMingen, San Jose, CA, USA) at 5 mg/ml in PBS overnight at 4uC. The wells were washed with PBS and blocked with RPMI containing 10% FCS. Then, 50 ml of cell medium (RPMI-1640 supplemented with 10% FCS, 2 mM L-glutamine, 100 IU/ml penicillin, and streptomycin and 1% HEPES) and 50 ml of spleen cells (10 6 per well) in cell medium with or without stimulation with rPspA at 5 mg/ml were added per well and incubated in the plates overnight in 5% CO 2 at 37uC. The next day, the cell suspensions were discarded and the plates washed with PBS-T. Biotinylated anti-IL-4 or anti-IFN-c mAb (BD PharMingen) at 5mg/ml in PBS-T with 1% FCS was added and incubated at room temperature for 2 h. After washing with PBS-T, 100 ml/well of avidin peroxidase diluted 1:1,000 (vol/vol) in PBS-T containing 1% FCS were added, followed by incubation for 1 h at room temperature. After washing with PBS-T, 100 ml of AEC (3-amino-9-ethylcarbazole, BD Bioscience) was added per well. Spots were developed for 15 min at room temperature. Plates were dried and analyzed by using an automated CTL ELISPOT Reader System (Cellular Technology, Shaker Heights, OH, USA).

Pneumococcal challenge
At week 10, mice were challenged by intraperitoneal injection with 1610 4 CFU of S. pneumoniae WU2 in 100 ml BSG [5]. The 50% lethal dose (LD 50 ) of S. pneumoniae WU2 in BALB/c mice was 1610 2 CFU. Challenged mice were monitored daily for 15 days.

Statistical analysis
Numerical data are presented as arithmetic means for bacterial number data and geometric mean and standard deviation in all other assays. Mann-Whitney U Test (version 5.0; GraphPad Software, Inc.) was used for comparing the distribution of S. Typhi in tissues of newborn mice, survival of S. Typhi strains in peripheral human mononuclear cells, in human blood, chlorinated tap water, raw sewage and surface water. An ANOVA (SPSS Software) analysis, followed by LSD (Least Significant Difference) method, was used to evaluate differences in antibody titer and cytokine-secreting cells response discerned to 95% confidence intervals. The Kaplan-Meier method (SPSS Software) was applied to obtain the survival fractions following i.p. challenge of intranasally immunized mice. P,0.05 was considered statistically significant.
All of these mutations have been previously described [28,29,38,48], with exception of DasdA33. The DasdA33 defined deletion mutation encompasses a 1,104 base pair deletion including the ATG start codon but not including the TAG stop codon. The upstream flanking regions of S. Typhi asdA gene share only 55% similarity with S. Typhimurium. Therefore, the flanking regions used to construct DasdA33 were amplified from the S. Typhi Ty2 genome. Although most of these mutations have been described, the details of construction for some of the mutations are described here for the first time. The DP crp527 ::TT araC P BAD crp, DP fur81 ::TT araC P BAD fur, Dpmi-2426 and DsopB1925 mutations, attenuate S. Typhimurium for virulence in the oral mouse model [28,29] (and unpublished data). In the two former mutations, the crp and fur genes are under transcriptional control of the araC P BAD promoter. When cells are grown in broth culture in the presence of arabinose, crp and fur are expressed. After immunization, when the cells reach host tissues where free arabinose is not available [31], crp and fur are no longer expressed and the cells become attenuated [28]. As expected for DP crp527 ::TT araC P BAD crp mutants, all three RASTyV strains formed red colonies on MacConkey maltose agar in the presence of arabinose, and white colonies in the absence of arabinose [28]. The DP fur81 ::TT araC P BAD fur phenotype was tested using CAS plates. In absence of arabinose, a yellow ring was observed around the colony, indicating siderophore production. In presence of arabinose, the indicative yellow ring was not observed as expected [28]. The OMP profile showed that the iron-uptake OMPs, repressed by Fur, were down regulated in presence of arabinose and up regulated in absence of arabinose (data not shown; see reference [29]). During strain construction we observed that a DP fur33 ::TT araC P BAD fur S. Typhimurium mutant synthesized too much Fur to be optimally attenuated after growth in 0.2% arabinose [28]. We therefore replaced the DP fur33 ::TT araC P BAD fur deletion/insertion with the DP fur81 ::TT araC P BAD fur deletion/insertion, in which the ATG start codon of the fur gene was changed to GTG [28]. In S. Typhimurium DP fur81 ::TT araC P BAD fur strains produce less Fur than DP fur33 ::TT araC P BAD fur strains [28]. Arabinose-regulated synthesis of Crp and Fur in all three RASTyV was confirmed by western blot analysis (Fig. 2).
The phenotypes of DP phoPQ107 ::TT araC P BAD phoPQ in the absence and presence of arabinose and PhoP + strains were tested by 3XD plates supplemented with 50 mg/ml of 5-bromo-4-chloro-3-indolyl-phosphate to reveal acid phosphatase activity. Complementarily, the expression of PhoP was tested by western blot (data not shown). However, during the course of strain construction, we observed that a DP phoPQ107 ::TT araC P BAD phoPQ mutant of S. Typhimurium is only moderately immunogenic, although highly attenuated [28]. We attempted several modifications of the construct to change the level of phoP expression, similar to the strategy we used to improve the fur construct, but none of the changes significantly improved the immunogenicity profile [28]. Therefore, we decided against using this mutation in our S. Typhi vaccine vectors. Rather than completely remake our three vector strains, we replaced the DP phoPQ107 mutation with the wild-type phoP promoter from S. Typhi ISP1820 in all strains (Table 1; Fig. 1).
The DagfBAC811 mutation serves to reduce the bacterium's ability to form thin aggregative fimbriae [74] thus reducing its ability to survive and persist in the environment [75]. The D(gmdfcl) deletion blocks the conversion of GDP-mannose to GDPfucose, preventing colanic acid production [76,77], which may contribute to biofilm formation in vivo [78]. The DagfBAC811 and D(gmd-fcl) mutations were confirmed by PCR using the relevant primers shown in Table 3 (data not shown).
The S. Typhi Vi capsular polysaccharide is involved in immune evasion and has immunosuppressive effects on the host [79]. Also, the Vi antigen inhibits bacterial adhesion and invasion of intestinal epithelia [80]. TviA, a key regulator of Vi antigen synthesis, also represses expression of genes required for synthesis of flagella and the invasion-associated type 3 secretion system through repression of the flagellar regulators flhDC and fliZ, resulting in reduced invasion and reduced motility [81]. We deleted tviABCDE, genes required for the regulation and synthesis of the Vi capsule [82]. Wild-type strains with the DtviABCDE10 deletion showed an increase in type 3 secretion when grown in LB broth (data not shown) and all three of the final vaccine strains were resistant to Vi-specific phage and did not agglutinate in presence of anti-Vi antiserum. Additional advantages for using a Vi 2 strain in our constructs is that there will be no immune interference in individuals who have been vaccinated with Vi or Vi-conjugate vaccines and, along with the pmi deletion, serves to remove major surface antigens to decrease the host immune response against Salmonella.
The DsopB1925 deletion was introduced because sopB mutations in Salmonella reduce fluid accumulation in ligated ileal loops [32] and enhance the host immune response against vectored antigens [35,83]. The DaraBAD23 deletion was introduced into our S. Typhimurium vaccine strains to preclude acid formation due to arabinose fermentation during in vitro growth, as this leads to a cessation of growth [28]. Although S. Typhi does not utilize arabinose as carbon source, the araBAD sequences of S. Typhi Ty2 and S. Typhi CT18 are identical to the S. Typhimurium sequence. The DaraBAD23 deletion was introduced into the ISP1820 derivative. No differences in arabinose utilization were detected on either MacConkey plates containing only arabinose or on minimal media containing arabinose as the sole carbon source, and therefore we did not introduce the DaraBAD23 deletion into the S. Typhi Ty2 lineage.
Prior to the introduction of the DasdA33 mutation, a functional rpoS gene from S. Typhi ISP1820 was introduced into the S. Typhi Ty2 derivative. The DasdA33 mutation was then introduced into both the S. Typhi Ty2 and S. Typhi Ty2 RpoS + derivatives, yielding three attenuated vaccine strains, two derived from Ty2, x9639 and x9640 (RpoS + ), and one from ISP1820, x9633 (Table 1; Fig. 1). The presence or absence of RpoS was confirmed by evaluating catalase activity and glycogen accumulation as described in the Materials and Methods. All strains had the expected phenotype: RpoS + strains S. Typhi x9640 and S. Typhi x9633 were positive for catalase activity and glycogen accumulation and RpoS 2 strains S. Typhi Ty2 x9639 were negative for both phenotypes.
The API20E profiles of each of the RASTyV strains (containing pYA4088) were compared to the non-recombinant wild-type parents in the presence and absence of 0.2% arabinose. In the presence of arabinose, all three vaccine strains exhibited biochemical profiles consistent with Salmonella; API code 4104540 for x9633(pYA4088) and 4004540 for x9639(pYA4088) and x9640(pYA4088). As expected, in the absence of arabinose, the profile for each strain was different, 4004000 for the Ty2 derivatives x9639(pYA4088) and x9640(pYA4088) and 4104000 for the ISP1820 derivative x9639. The change in identification is the result of the lack of fermentation of all Crp-dependent sugars (crp is not expressed in the absence of arabinose). In the presence of arabinose, the API results for strain x9633(pYA4088) deviated from the wild-type S. Typhi ISP1820 in respect to a positive reaction for ornithine decarboxylase (ODC). During the strain construction process, the positive ODC reaction was observed following the introduction of the DP crp527 ::TT araC P BAD crp and DP fur33 ::TT araC P BAD fur mutations. Neither deletion alone was sufficient to induce this phenotype: the phenotype was only observed when the mutations were combined and was observed in independently-derived strains, suggesting that this was not a spontaneous mutation elsewhere in the chromosome. However, it is unclear how the combination of DP crp527 ::TT araC P BAD crp and DP fur33 ::TT araC P BAD fur could result in synthesis of a functional ornithine decarboxylase (ODC) in S. Typhi ISP1820. While it is possible that a functional ornithine decarboxylase gene was introduced from S. Typhimurium through the transduction process used in strain construction, this is unlikely, as the ODC + phenotype was only observed in DP crp527 ::TT araC P BAD crp and DP fur33 ::TT araC P BAD fur double mutants.
Type I fimbriae were synthesized in all three S. Typhi mutant strains when they were grown statically in LB broth containing 0.05% arabinose. Type I fimbrial synthesis was not assessed in the absence of arabinose, due to poor growth of the strains in arabinose-free media. LPS synthesis in all three vaccine strains was dependent of the presence of D-mannose in the media as expected due to the Dpmi-2426 mutation (data not shown).

Regulated protein synthesis in RASTyV strains
The Asd + plasmid vector pYA4088 (Table 2), which complements the DasdA33 deletion, carries DNA that codes for aa 3,285 of the S. pneumoniae gene pspA [30], fused at the amino-terminal end to the b-lactamase type 2 secretion signal that directs PspA secretion out of the cytoplasm and into the periplasm and supernatant [6]. Plasmid pYA4088 and the parent plasmid, pYA3493 ( Table 2) that does not carry pspA gene sequence [6], were introduced into the three S. Typhi vaccine strains. Plasmids pYA4088 and pYA3493 were maintained and stable in all strains for at least 60 generations. Each strain was grown in the presence of arabinose and then diluted into media without arabinose and serially passaged in arabinose-free media. Samples were taken after each passage and synthesis of proteins encoded by arabinoseregulated genes was assessed by western blots. We obtained similar results for all three strains. The results from strain S. Typhi ISP1820 x9633(pYA4088) and x9633(pYA3493) are shown in Figure 2A. As arabinose was diluted away, synthesis of Crp, Fur and LacI decreased, but PspA synthesis increased, as expected.
To verify PspA secretion, we fractionated cells from each strain and determined the level of PspA in each fraction. Our results indicated that greater than 50% of the PspA-Rx1 was found in the periplasm and supernatant for all three strains. Typical results, for S. Typhi ISP1820 x9633(pYA4088) are shown in Figure 2B. These results are consistent with what has been reported for S. Typhimurium [6,30].

Sensitivity of RASTyV strains to transient deoxycholate exposure
We noticed during the strain construction process that the RASTyV strains grew more slowly on bile-containing media such as MacConkey agar than the wild-type S. Typhi. This raised the possibility that one (or more) of the mutations introduced into the vaccine strains negatively impacted survival in the presence of bile, and thus the RASTyV strains might have difficulty surviving transit through the duodenum where bile from the gall bladder is introduced into the intestine. Therefore, we assessed survival of the RASTyV strains in the presence of deoxycholate compared to wild-type S. Typhi. Deoxycholate concentrations in the duodenum range from 0-1%, and may rise as high as 4% in the gall bladder [84]. The viability of all strains decreased as the amount of deoxycholate increased, but the behavior of the RASTyV strains differed depending on their parental background (Fig. 3). The S. Typhi Ty2-derived vaccine strain, x9639(pYA4088) survived at a significantly lower rate than the wild-type S. Typhi at all concentrations of deoxycholate (P,0.05). The remaining RAS-TyV strains were significantly more sensitive to deoxycholate than the wild-type strains at most concentrations tested (P,0.05). However, for all vaccine strains, the decrease in resistance compared to wild-type resulted in a drop in cell numbers of less than 10-fold, suggesting that the vaccine strains should be able to transit the intestinal tract, but may not persist for long duration in the intestinal contents.

RASTyV strain survival in chlorinated water, in sewage and untreated surface water
In addition to being avirulent in the host, a live attenuated vaccine should possess some containment features to reduce its survival and persistence in nature to preclude vaccination of individuals who did not elect to be vaccinated [85]. We evaluated the survival of the RASTyV strains compared to their wild-type S. Typhi counterparts in chlorinated water, sewage and surface water, conditions designed to mimic the environment and to address concerns regarding the impact of releasing live attenuated, genetically-modified organisms into the environment. Since the chlorine concentration in our local tap water was found to vary significantly from day to day, we elected to mimic the chlorination process that occurs during municipal water treatment. All S. Typhi strains were extremely sensitive to chlorinated water experiencing several logs of killing after a 30-minute exposure (Fig. 4A).
In sewage, the titers of the RASTyV strains dropped steadily after one day, decreasing more than 3 logs in titer over the 10-day period (Fig. 4B). The wild-type strains survived significantly better than the vaccine strains (P,0.01). A similar trend was observed in the untreated surface water. The number of viable RASTyV bacteria decreased steadily after inoculation at a more rapid rate than the wild-type strains (Fig. 4C). The wild-type strains survived significantly better than the RASTyV strains (P,0.01).

Survival of RASTyV strains in human blood, peripheral human mononuclear cells, and in guinea pig complement
The development of rapid ex-vivo assays using human blood and freshly elutriated peripheral blood mononuclear cells (PBMCs) have been demonstrated as reliable tools for determining attenuation of S. Typhi for vaccine research and development [68,86]. 1610 6 CFU of S. Typhi was added to duplicate 1.5 ml blood aliquots from volunteers. The numbers of all the RASTyV strains and Ty21a dropped below the detection threshold of 10 CFU/ml within 3 h and the numbers did not increase at later time points (Fig. 5A). In contrast, the wild-type S. Typhi ISP1820 and Ty2 parent strains survived significantly better than their vaccine derivatives x9633(pYA4088) and x9639(pYA4088), respectively (P,0.01). Surprisingly, the Ty2 RpoS + strain x8438 survived poorly in whole human blood. No killing was observed for any of the strains in heat-inactivated blood samples (Fig. 5B), indicating that the poor survival of the RASTyV strains is due to their inability to resist the microbicidal effects of active blood components.
An important part of Salmonella pathogenesis is the ability of the bacterium to survive and grow in macrophages [86,87]. Therefore, we evaluated the ability of our strains to survive in human monocyte-derived macrophages (Fig. 5C). The three RASTyV strains and Ty21a were significantly reduced in their ability to persist and survive in PBMCs compared to their wildtype parents at 24 h after infection (P,0.01). The ISP1820 derivative, x9633(pYA4088) survived better than any of the other vaccine strains, while the Ty2 derivatives, x9639(pYA4088) and Ty21a were the most sensitive to killing.
Complement is an important blood component with bactericidal effects. We evaluated the relative sensitivities of the RASTyV strains and their wild-type counterparts to complement. Both wildtype and RASTyV strains were sensitive to antibody-dependent complement killing when incubated in the presence of anti-Salmonella O-antigen group D 1 antibody (Fig. 5D). In all cases, the vaccine strains were more sensitive to complement than their wildtype counterparts (P,0.01). In the absence of S. Typhi-specific antibody, the wild-type strains were completely resistant to the effects of complement. The RASTyV strains remained sensitive to complement-mediated killing even in the absence of opsonizing antibody. This is consistent with the work done by R. Looney showing that the Vi capsule blocks C3 deposition on the surface of S. Typhi, and strains lacking Vi, such as the RASTyV strains used in this study, are susceptible to complement killing via the alternative (non-antibody) pathway [88].

Distribution of S. Typhi strains in tissues of newborn mice
Newborn mice (24 hours old) inoculated orally with 160.2610 9 CFU of RASTyV strains, wild-type parents or S. Typhi Ty21a survived without any symptoms of disease. We chose to use the oral route in these experiments to establish that our strains can overcome host defenses (e.g. low pH of stomach, presence of antibacterial substances in the gut) allowing it to invade the host mucosa and reach deeper tissues (e.g. spleen and liver). The bacterial burden in tissues of newborn mice were determined at 3 and 7 days post infection (Fig. 6). All strains were able to persist in the intestine for 7 days, although the Ty2 derivatives did so at significantly lower numbers than their wild-type parents (P,0.05) (Fig. 6A). All vaccine strains were less able to persist in the liver and spleen compared to the wild-type strains ( Fig. 6B and Fig. 6C). The ISP1820 derivative x9633(pYA4088) was present in both tissues on day 3, but was cleared by day 7.

Antibody responses in mice after intranasal immunization with recombinant Salmonella vaccines
Groups of seven-week-old mice were intranasally immunized with 160.2610 9 CFU each test strain. All mice immunized with strains expressing pspA developed anti-PspA antibodies (Fig. 7A). Strain x9640(pYA4088) (Ty2 RpoS + ) induced significantly higher anti-PspA IgG titers than the ISP1820 derivative x9633(pYA4088) at all time points (P,0.05 or P,0.01), and higher than the RpoS 2 Ty2 derivative x9639(pYA4088) at 4 weeks, 6 weeks and 8 weeks (P,0.05). Anti-PspA titers increased in all groups immunized with strains expressing pspA after the second immunization at 6 weeks. After this boost, anti-PspA IgG antibody titers in x9639(pYA4088) immunized mice were significantly higher than in mice immunized with x9633(pYA4088) (P,0.05). No anti-PspA IgG was detected in mice immunized with BSG (data not shown).
All attenuated Salmonella strains induced significant anti-LPS titers (Fig. 7B) as early as two weeks post inoculation. After the second immunization, significant boosting of serum antibody responses to LPS was observed (P,0.05 or P,0.01). Similar results were obtained for serum IgG responses against SOMPs (Fig. 7C).
Mucosal IgA anti-PspA responses were detected by week 2 in mice immunized with x9639(pYA4088) (RpoS 2 ), but not in mice immunized with the two RpoS + strains (P,0.01) (Fig. 7D). By 4 weeks, anti-PspA IgA was detected in the x9640(pYA4088) group. Statistical differences were seen between groups at each time point and titers reached a maximum by 8 weeks, after the boost that occurred at 6 weeks. IgA responses in mice immunized with x9633(pYA4088) were the slowest to develop (Fig. 7D).

Antigen-specific stimulation of IL-4 or IFN-c production
To further evaluate the Th1/Th2 immune response, ELISPOT was used to determine numbers of IFN-c (Th1-associated) and IL-4 (Th2-associated) secreting splenocytes from immunized and control mice in response to PspA (Fig. 9). Splenocytes from mice immunized with the RASTyV strains carrying pYA4088 produced very low levels of PspA-specific IFN-c secreting cells (Fig. 9A). In contrast, splenocytes from the mice immunized with any of the RASTyV strains carrying pYA4088 produced significantly higher levels of PspA-specific IL-4 secreting cells than those of the empty vector controls (P,0.05 or P,0.01) (Fig. 9B). Of note, the results with strain x9639(pYA4088), which induced a strong IL-4 response and an undetectable IFN-c response, was consistent with the strong Th2 response indicated by the IgG1/IgG2a analysis (Fig. 9B).

Evaluation of protective immunity
To examine the ability of RASTyV strains expressing pspA to protect against pneumococcal infection, mice were challenged intraperitoneally with 1610 4 CFU of S. pneumoniae WU2 four weeks after the boost. S. pneumoniae WU2 produces PspA that is Figure 5. Survival of RASTyV strains in human blood, peripheral human mononuclear cells, and guinea pig complement. The bactericidal effects of (A) active whole blood and (B) heat-treated blood were compared by incubating 1610 6 CFU of each strain in 1.5 ml normal or heat-inactivated human whole blood. Bacterial survival was measured by spread plating at the indicated times after inoculation. **, P,0.01, for each RASTyV compared to its wild-type parent at 18 h. (C) S. Typhi survival in peripheral human mononuclear cells at 1 h, 4 h and 24 h after inoculation with 5610 6 CFU. **, P,0.01, for each RASTyV compared to its wild-type parent at 24 h, and significant differences between Ty21a and RASTyV are indicated. The assay was performed in duplicate and was repeated at least 3 times using blood from different individuals. The limit of detection was less than 10 CFU/ml. (D) S. Typhi survival in guinea pig complement three hours after inoculation with 1610 6 CFU. **, P,0.01, for the RASTyV strain compared to its wild-type parent at 3 h. Data shown are the arithmetic means of triplicate samples. doi:10.1371/journal.pone.0011142.g005 cross-reactive with PspA Rx1 encoded in pYA4088 [89]. All mice immunized with RASTyV expressing pspA were significantly protected compared with BSG and empty vector control groups (P,0.01) (Fig. 10). The protection afforded by the Ty2 derivatives, RpoS + x9640(pYA4088) was significantly greater than that afforded by the ISP1820 derivative x9633(pYA4088) (P,0.05).

Discussion
Our ultimate goal is to design an RASTyV that synthesizes multiple pneumococcal antigens. One important often overlooked step in that process is to identify an optimal Salmonella vector strain. Therefore, we constructed, characterized and evaluated in mice three recombinant attenuated S. Typhi strains derived from ISP1820 and Ty2 (with and without a functional rpoS gene) synthesizing one antigen, the S. pneumoniae pneumococcal surface protein PspA, to evaluate which strain will make the best vector. One consideration in our strain design strategy was to test our hypothesis that immune responses induced to protective antigens will be superior in RpoS + S. Typhi strains.
One of the most daunting problems encountered when evaluating an S. Typhi vaccine is making a pre-clinical determination of its safety and/or level of attenuation. Typically, an identical mutation is introduced into S. Typhimurium and the virulence of the strain is determined by oral administration to mice. While this has proven to be an accurate first approximation, it is clear that the interaction between S. Typhimurium and mice is not identical to the interaction between S. Typhi and humans. In some cases, attenuation of an S. Typhi mutant has been evaluated by injecting low doses of the vaccine strain intraperitoneally with hog gastric mucin into outbred mice [90]. This method has proven reliable for demonstrating the reduced virulence of a variety of S. Typhi strains that are attenuated by deletion or mutation of genes that result in a growth defect, such as aro mutants [91] and Ty21a [12]. However, the validity of this model is undermined by the observation that a S. Typhi phoP mutant, which has no growth defect, is just as virulent as wild-type strains [92] in this model despite the fact that both S. Typhi and S. Typhimurium DphoPQ mutants have been shown to be safe in Phase I clinical trials [18,93]. In this study, we examined survival of the RASTyV strains in whole blood, in PMBC's and in the presence of complement as a surrogate for safety. Our strains were exquisitely sensitive to whole blood, with no detectable survivors after three hours (Fig. 5A). The vaccine strains were also more susceptible than their parent strains to killing by PBMCs (Fig. 5C), and by guinea pig complement (Fig. 5D). The ISP1820 derivative, x9633(pYA4088), survived better in PBMCs than either of the Ty2 derivatives, which had similar survival curves. The survival profiles of our vaccine strains in whole blood and PBMCs was similar to the licensed typhoid vaccine strain Ty21a, indicating that their attenuation phenotypes resulted in a desirable safety profile.
Interestingly, strain x8438, the wild-type Ty2 RpoS + strain, had a survival profile in whole blood that was more like the RASTyV than wild-type Ty2. The reason for this is unclear, but it may be related to the fact that Ty2, isolated in 1918, has, presumably, been an rpoS mutant for many years. During that time, it is possible that it has acquired a number of suppressor mutations that enhanced its survival. Restoration of the mutant rpoS allele may have led to a phenotypically less fit strain, at least with respect to survival in blood.
This result does not appear to be related to Vi antigen production, since both S. Typhi x8438 (Ty2 RpoS + ) and S. Typhi ISP1820 make comparable amounts of Vi, although both Typhi LPS (B), and S. Typhi OMP (C), and mucosal IgA responses to rPspA (D) were measured by ELISA using pooled sera and vaginal washes from BALB/c mice intranasally immunized with the indicated strains carrying either plasmid pYA3493 (control) or pYA4088 (PspA). Mice were boosted at week 6. ELISAs were performed twice with identical results. Significant differences were indicated *, P,0.05; **, P,0.01. No immune responses were detected to any antigen tested in mice immunized with only BSG or in preimmune sera from vaccinated mice (reciprocal titer ,1:50). doi:10.1371/journal.pone.0011142.g007 strains make less Vi than Ty2 [48]. One difference between these strains that may account for these observations is that S. Typhi ISP1820 produces more LPS O-antigen than x8438 [48,82] and the production of long-chain O-antigen has been Figure 8. Serum IgG1 and IgG2a responses to rPspA measured by ELISA. Anti-rPspA IgG1 and IgG2a titers in pooled sera from BALB/c mice intranasally immunized with the indicated RASTyV strains at various times. Mice were boosted at week 6. The ratios of IgG1:IgG2a at 8 weeks were 1:1 for x9633(pYA4088) (A) and x9640(pYA4088) (C) immunized mice respectively; and 8:1 for x9639(pYA4088) (B) immunized mice. All ELISAs were performed twice with identical results. doi:10.1371/journal.pone.0011142.g008 Splenectomies were performed on euthanized BALB/c mice 7 weeks after the primary immunization and one week after the boost. Buffer controls were also included. Splenocytes were harvested from 3 mice per group, and cells from each spleen were assayed in triplicate. Each symbol represents the results from a single well. The results from each well are presented as ELISPOTS per million splenocytes minus any background (#4) ELISPOTS, from unpulsed mock controls. There were no significant differences between vaccine strains and the control group for secretion of IFN-c. For secretion levels of IL-4, **, P,0.01 for x9633(pYA4088) and x9639(pYA4088) versus controls BSG group or empty vector group, *, P,0.05 for x9640(pYA4088) versus BSG group or x9640(pYA3493), and for x9639(pYA4088) versus x9640(pYA4088) as indicated. doi:10.1371/journal.pone.0011142.g009 associated with complement resistance [94]. With regard to environmental safety, all strains were sensitive to chlorinated water and the RASTyV strains were less able to survive in raw sewage compared to their wild-type parents (Fig. 4). These results indicate that the RASTyV strains have a diminished capacity to persist in the environment.
In day-old mice, the three RASTyV strains exhibited different capacities for persistence of the intestinal tract. Higher numbers of RpoS + strains S. Typhi x9633(pYA4088) and S. Typhi x9640(pYA4088) were isolated than of RpoS 2 strain S. Typhi x9639(pYA4088) (Fig. 6). Although our data are limited by the fact that we harvested the entire intestines, including contents, these results are consistent with the observation that RpoS regulated genes are necessary for invasion and colonization of M cells overlying the GALT [24,26]. In spite of this shortcoming, the data indicate that an RpoS + phenotype confers some advantage in the intestinal environment. All three of the wildtype strains were able to persist in the liver and spleen, while their attenuated counterparts were unable to maintain detectable numbers in these tissues. The inability of attenuated S. Typhi strains to persist in neonatal spleen and liver is consistent with a previous report [95] although in that study 7-day-old mice were inoculated intranasally. To our knowledge, this is the first report showing persistence in day-old mice with orally administered wild-type S. Typhi. We note that despite the ability of wild-type S. Typhi ISP1820, Ty2, and Ty2 RpoS + to persist in neonatal mouse tissues, this does not lead to a lethal infection. (Santander, unpublished observation).
The three RASTyV strains stimulated different immune responses to PspA. The RpoS + vaccine strains S. Typhi x9633(pYA4088) and S. Typhi x9640(pYA4088) induced mixed Th1-and Th2-type serum immune responses against rPspA and induced lower titers of IgA mucosal secretions than the RpoS 2 vaccine strain S. Typhi x9639(pYA4088). Despite the poor induction of IFN-c and the comparatively strong induction of IL-4, all three strains induced high levels of PspA-specific serum antibodies (Figs. 7 and 9). Mice immunized with S. Typhi x9640(pYA4088) or S. Typhi x9639(pYA4088) produced the highest titers of PspA-specific serum antibody and experienced the highest level of protection, significantly greater than the protection seen in mice immunized with the ISP1820 derivative S. Typhi x9633(pYA4088) (Fig. 10). It is not clear what the basis of these differences is and it is difficult to make predictions concerning which of the three strains will be the most immunogenic in humans, but it is clear that all three RASTyV strains are immunogenic and protective in the intranasal mouse model.
We have designed and fully evaluated three new RASTyV strains that include new features including regulated delayed in vivo attenuation and regulated delayed in vivo synthesis of the pneumococcal antigen PspA. We have demonstrated that these strains are immunogenic in mice, are highly susceptible to killing in human blood, and are poorly adapted for survival in the environment. Thus, the three RASTyV strains demonstrate a balance between a desirable safety profile and immunogenicity and should be completely safe when used in a clinical setting. The results from these preclinical studies pave the way for a phase 1 clinical trial, currently underway, where we are evaluating all three strains for safety and immunogenicity to determine the best S. Typhi vector to enable construction of an anti-pneumococcal vaccine. The clinical trial will allow us to address the importance of rpoS in RASTyV immunogenicity. Based on the results of the current clinical trial and ongoing work in our laboratory, we plan to make adjustments to the best of these three strains to further enhance safety to allow its use in younger and/or immunocompromised or malnourished populations and to incorporate additional protective pneumococcal antigens, with the goal of inducing protective immunity against all pneumococcal serotypes. Figure 10. Evaluation of protective efficacy. Groups of eight 7-weekold mice were intranasally immunized twice at 6-week intervals with the indicated strains and challenged intraperitoneally with 1610 4 CFU of S. pneumoniae WU2 4 weeks later. The experiment was performed twice. Both experiments gave similar results, and the data have been pooled. The protection from each RASTyV strains was significantly different compared with control groups. *, P,0.05 for survival of mice immunized with x9640(pYA4088) (Ty2 RpoS + ) compared with survival of mice immunized with x9633(pYA4088) (ISP1820). doi:10.1371/journal.pone.0011142.g010