A GMMA-CPS-Based Vaccine for Non-Typhoidal Salmonella

Non-typhoidal Salmonella are a major cause of gastroenteritis worldwide, as well as causing bloodstream infections in sub-Saharan Africa with a high fatality rate. No vaccine is currently available for human use. Current vaccine development strategies are focused on capsular polysaccharides (CPS) present on the surface of non-typhoidal Salmonella. This study aimed to boost the amount of CPS purified from S. Typhimurium for immunization trials. Random mutagenesis with Tn10 transposon increased the production of CPS colanic acid, by 10-fold compared to wildtype. Immunization with colanic acid or colanic acid conjugated to truncated glycoprotein D or inactivated diphtheria toxin did not induce a protective immune response in mice. However, immunization with Generalized Modules for Membrane Antigens (GMMAs) isolated from colanic acid overproducing isolates reduced Salmonella colonization in mice. Our results support the development of a GMMA-CPS-based vaccine against non-typhoidal Salmonella.


Introduction
Salmonella enterica species are the major cause of foodborne diseases, typhoid fever and gastroenteritis [1]. Typhoid fever, common in developing countries, is mainly caused by typhoidal Salmonella (TS) serovars: S. Typhi and S. Paratyphi [2]. Gastroenteritis occurs worldwide and is primarily caused by non-typhoidal Salmonella (NTS) serovars: S. Typhimurium and S. Enteritidis [2]. The global incidence of NTS gastroenteritis is estimated at~94 million cases annually, with 155,000 deaths, and~80.3 million cases are contracted via foodborne transmissions [3]. NTS gastroenteritis is usually self-limiting in immunocompetent individuals and severe cases can be treated with antibiotics. However, there is an increase in multi-drug-resistant isolates [4], and in sub-Saharan Africa, NTS serovars associated with gastroenteritis globally are the leading cause of bloodstream infections that are distinct from typhoid fever [5,6]. A protective vaccine against NTS serovars and implementation of effective food safety procedures are needed to reduce the global incidence of gastroenteritis.
Capsular polysaccharides (CPS) are carbohydrate-based polymers synthesized and expressed on bacterial surfaces [7]. The CPS that have been identified in NTS are colanic acid, lipopolysaccharide (LPS) and O-Ag capsule [8][9][10]. Colanic acid composed of repeating units of glucose, galactose and fucose [8] is produced at temperatures below 30 • C

Bacterial Strains, Media and Growth Conditions
The bacterial strains used in this study are listed in Table 1. For standard growth, strains were inoculated from frozen stocks onto LB agar (lysogeny broth, 1% NaCl, 1.5% agar) supplemented with an appropriate antibiotic (50 µg/mL kanamycin (Kan), 34 µg/mL chloramphenicol (Cam) or 5 µg/mL tetracycline (Tet)) and grown overnight at 37 • C. Isolated colonies were used to inoculate 5 mL LB broth and the culture was incubated for 18 hours at 37 • C with agitation at 200 rpm. For EPS purification, overnight cultures were grown for 5 days at 28 • C on agar supplemented with 1% glucose, 0.05% yeast extract, 10 mM Na 2 HPO 4 , 0.1% NH 4 Cl, 0.3% KH 2 PO 4 , with or without 40 mM MOPS (3-

Generation of S. Typhimurium Mutant Strains
The lambda red recombinase knockout procedure [34] was used to generate S. Typhimurium ∆lon, ∆tolR, ∆yihW, ∆yih and ∆yihVW mutants. Primers containing 50-nucleotide sequences on either side of lon, tolR, yih, yihW or yihVW (Table 2) were used to amplify the cat gene from pKD3 or kan gene from pKD13 using Phusion High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MS, USA). The PCR products were purified and electroporated into S. Typhimurium 14028 cells containing pKD46. Mutants were selected by growth at 37 • C on LB agar supplemented with 10 µg/mL Cam or 50 µg/mL Kan. Cam R isolates were re-streaked onto LB agar containing 34 µg/mL Cam. Mutations were transduced into S. Typhimurium strains with a clean background using P22 phage [35]. The cat or kan gene was resolved from the chromosome using pCP20 [34]. DNA sequencing of PCR products from the chromosomes of mutant Salmonella isolates was used to verify the loss of lon, tolR, yih, yihW or yihVW. Primers used in this study are listed in Table 2.

Generation of Transposon Mutants
To generate S. Typhimurium ∆bcsA ∆yihW transposon mutants [30,32], S. Typhimurium LT2 harboring pNK972 was infected with P22 lysate of S. Typhimurium TT10423 containing Tn10dtet on F'. The resulting transductants (~100,000 colonies) were pooled and a P22 S. Typhimurium LT2 Tn10dtet transducing fragment library was generated, following the method outlined by Maloy [35]. The resulting P22 phage lysate was used to infect S. Typhimurium ∆bcsA ∆yihW and plated on 1% tryptone supplemented with 10 µg/mL Tet. The resulting transductants (10,000 to 20,000) were visually screened for the presence of a mucoid morphology, which was indicative of high CPS production. To identify the site of transposon insertion, nested PCR was performed on purified genomic DNA using primers TL or TR and ARB1 or ARB6. The product of the first reaction was further amplified using primers UniversalTn and ARB2. PCR products were purified and sequenced. To identify the site of Tn10 insertion, DNA sequence was mapped to S. Typhimurium 14028 using Geneious 9.0.5 (https://www.geneious.com)

Generation of Plasmid Vectors
To generate pBR322-yihVW, the DNA region containing yihVW was PCR-amplified from S. Typhimurium 14028 genomic DNA using primers yihVWFOR and yihVWREV ( Table 2) and Phusion polymerase (Thermo Fisher Scientific, Waltham, MS, USA), following reaction conditions as recommended by the manufacturer. The resulting PCR product was digested with AatII and PstI and ligated into AatII/PstI-digested pBR322 prior to electroporation into S. Typhimurium 14028. Positive clones were selected by growth on LB agar supplemented with 7 ug mL −1 Tet. The generation of pBR322-yihVW plasmid was confirmed by sequencing using primers yihVWseqF and yihVWseqR ( Table 2). The generation of yihUTSRQPO and yihVW promoter region fused to luxCDABE operon has previously been described [9].

Purification and Neutral Monosaccharide Analysis of Capsular Polysaccharide
The capsular polysaccharide (CPS) purification protocol was adopted from Gibson et al. [9] with some modifications. To separate EPS from other cellular materials, S. Typhimurium cells scraped off the agar surface were resuspended in 1% phenol, mixed vigorously by vortexing and sedimented by centrifugation (16,000× g, 4 • C, 4 h). Crude EPS was precipitated from the supernatant by adding 4 volumes of ice-cold acetone while stirring continuously with a glass rod. The resulting precipitate was stored overnight at −20 • C to allow for further precipitation, sedimented by centrifugation (6000× g, 4 • C, 15 min) and air-dried. Crude EPS was resuspended in water, dialyzed in water for 48 h (MW cut-off 10 kDa; SnakeSkin dialysis tubing, Thermo Fisher Scientific) and lyophilized. The lyophilized polysaccharide was dissolved in buffer A (15 mM NaOAc, 0.05% Triton X-100, pH 5.5) and 0.01% sodium azide. The dissolved sample was heated twice at 37 • C for 15 min prior to loading onto Q Sepharose FF xk50/11.5 and washing with 2 column volumes of buffer A. Materials were eluted with a stepwise gradient of buffer B (1.5 M NaOAc, 0.05% Triton X-100, pH 5.5), which was increased sequentially from 17% (1.25 column volumes) to 50% (1.25 column volumes) to 100% (1.5 column volumes). CPS-specific serum was used to identify fractions containing cross-reactive material by western blot.
For size exclusion chromatography, colanic acid containing fractions were pooled, concentrated (10 MWCO centrifuge filters, Amipore) and filtered through a 0.22 µm syringe tip filter, prior to loading on to the Superdex S300 prep grad xk26/95 column. The column was washed with 50 mM NH 4 HCO 2 pH 7.72. Colanic acid containing fractions were identified using Western blot and concentrated with a Millipore Amicon Ultra-15 Centrifugal Filter Device, dialyzed in water for 48 h (MW cutoff 10 kDa; SnakeSkin dialysis tubing; Thermo Fisher Scientific) and lyophilized.
The dry polysaccharide samples were treated by thermostable α-amylase and αglucosidase, followed by extensive dialysis against de-ionized water and freeze-drying to make sure the samples were glycogen-free [36]. The lyophilized samples were then hydrolyzed to monosaccharides by trifluoacetic acid, reduced with sodium borohydride, and acetylated using acetic anhydride [37]. The resulting alditol acetates were analyzed on an Agilent HP 5890 Series II gas chromatograph coupled to a flame ionization detector (GC-FID) fitted with a Rtx-2330 capillary column (15 m length × 0.32 mm internal diameter × 0.2 µm film thickness; Restek Corporation). Inlet temperature was kept consistently at 250 • C and an injection volume of 1 µL was used for all injections. For each sample run, the oven temperature was programmed to hold at 180 • C for 2 min, increase from 180 to 230 • C at 4 • C/min, from 230 to 260 • C at 8 • C/min, followed by holding at 260 • C for 10 min, and the inlet program was programmed to increase from 5.0 to 6.0 psi at 0.10 psi/min, from 6.0 to 7.0 psi at 0.20 psi/min and from 7.0 to 8.0 psi at 0.40 psi/min, followed by holding at 8.0 psi for 10.75 min (total run time 28.25 min). Relative monosaccharide composition (mol%) was calculated by referring to the FID responses of monosaccharide standards. Experiments were conducted in triplicate.

Endotoxin Removal from Purified EPS
For endotoxin removal by acid hydrolysis, purified crude EPS were dissolved in 1% acetic acid and heated in an oil bath at 110 • C for 100 min [38]. The solution was centrifuged for 10 min at 8000 rpm. Pellet was discarded and the supernatant was lyophilized before colanic acid isolation using anion exchange and size exclusion chromatography.
For endotoxin removal using Triton X-114, purified colanic acid was dissolved in precondensed Triton X-114, prepared by dissolving 20 mL of Triton X-114 and 16 mg of 2, 6-Ditert-butyl-4-methylphenol at 4 • C in 10 mM Tris-HCl (pH 7.4), 150 mM NaCl solution [39]. The mixture was incubated at 30 • C until it separated into a detergent-depleted aqueous phase and a detergent-rich phase. The large aqueous phase was discarded and replaced with an equal volume of 10 mM Tris-HCl (pH 7.4), 150 mM NaCl. The solution was mixed and incubated at 30 • C to enable phase separation. After three rounds of phase separation, the detergent-rich phase with a Triton X-114 concentration of 11%, was collected and stored at room temperature [38].
For endotoxin removal using pre-condensed Triton X-114, purified EPS was dissolved in water to a final concentration of 0.5 mg/mL and mixed with 11% Triton X-114 to a final Triton X-114 concentration of 1% (w/v). The resulting cloudy solution was stirred for 30 min at 4 • C until the solution became clear. The mixture was incubated at 37 • C for 30 min to induce separation into two phases, and subsequently centrifuged at 1200× g for 30 min at 25 • C. The lower phase containing LPS was discarded and the upper phase containing EPS was mixed with 11% Triton X-114 to a final concentration of 2%. LPS extraction was repeated twice as described above. Centrifugation speed and time were increased by 1000× g and 30 min for each round of purification to obtain the best separation.
To remove detergent from the EPS-containing solution after LPS removal, the Triton X-114-treated samples were mixed with three volumes of methanol-chloroform (two volumes of methanol and one volume of chloroform) and incubated at room temperature for 30 min to enable phase separation. The Triton X-114 containing lower phase was eluted and the colanic acid containing upper phase was mixed again with the methanolchloroform solution. This procedure was repeated two additional times. Residual methanolchloroform was removed using water aspiration. Samples were dialyzed for 48 h in ddH 2 O at 4 • C and lyophilized.

Conjugation of EPS to Truncated Glycoprotein D (tgD) and CRM197
Twenty mg of purified colanic acid dissolved in 0.15 M NaCl (2 mL final volume) was allowed to react with 13 mg of sodium cyanoborohydride (CNBr, dissolved in 16 µL chloroform), and the pH of the solution was maintained at pH 10.5 using NaOH (0.5 M). After 15 min at room temperature, adipic dihydrazide (ADH) was added to a final concentration of 12 mg/mL. HCl (0.5 M) was used to reduce the final pH to 8.5. The reaction mixture was incubated at room temperature overnight with stirring. The solution was dialyzed twice against 20 mM MES buffer (pH 6.0) using a 10 kDa cut-off MWCO dialysis tubing for 3 hours at 4 • C [40]. Inactivated diphtheria toxin CRM197 was obtained from Fina Biosolutions LLC (Rockville, MD, USA). The purification of truncated glycoprotein has been previously described [41].
The resulting CNBr-treated CPS (5 mg/mL, 3 mL) was allowed to react with 1.5 mL tGD or CRM197 (2 mg/mL) in the presence of 12 mg EDC (dissolved in 20 mM MES buffer pH 6.0) at room temperature overnight. The resulting solution was dialyzed against 0.15 M NaCl using 10 kDa MWCO dialysis tubing for 3 h at 4 • C. CPS conjugated to tGD (CA-tgD) or CRM197 (CA-CRM197) was loaded onto a Sephacryl S300 xk26/94 column. The column was washed with 50 mM PBS, 15 mM NaCl, pH 7.4 solution and 10 mL fractions were collected. Fractions containing conjugated material were identified based on size after western blot analysis with EPS-specific serum.

Outer Membrane Vesicles (GMMAs) Production and Purification
For GMMA production, bacteria grown overnight at 37 • C were used to inoculate 1% tryptone broth (without antibiotics) to an optical density of 0.03 at 600 nm. Cultures were incubated at 30 • C, for 18 h with agitation (200 rpm). Culture supernatants were collected by centrifugation at 8000× g for 10 min, filtrated with a 0.45 µm filter, and concentrated by dialysis (MW cut-off 10 kDa; SnakeSkin dialysis tubing, Thermo Fisher Scientific) against polyethylene glycol (PEG) 2000. GMMAs were pelleted by ultra-centrifugation at 186,000× g for 2 h at 4 • C. The resulting pellets were resuspended in endotoxin-free water, filtered with a 0.22 µm filter, lyophilized, and resuspended in PBS.

Vaccine Formulation
Vaccine antigens, colanic acid (CA), CA-tgD and CA-CRM197 were formulated with triple combination adjuvant consisting of Polyinosinic:polycytidylic acid (Poly (I:C)) (Thermo Fisher Scientific), host defense peptide (HDP) and polyphosphazene (PCEP) (Idaho National laboratory) immediately prior to administration. Formulations were prepared by first mixing 10 µg of PIC and 20 µg HDP in PBS and incubating at room temperature (RT) for 15 min. CA, CA-tgD or CA-CRM197 were added before the addition of 10 µg of PCEP, making a final ratio of 1:2:1 of PIC:HDP:PCEP (TriAdj). Mixtures were incubated in the dark for 15 min at RT prior to administration.

Murine Immunization Experiments
Six-to eight-week-old female BALB/c mice were purchased from Charles River Laboratories (Kingston, ON), while C57BL/6 mice were purchased from Jackson Laboratory (Bar Harbor, ME). Mice were assigned to cage groups using a randomization table prepared in Microsoft Excel and individual mice were marked with ear notches.

Murine Infection Experiments
For competitive index (CI) experiments, two groups of six C57BL/6 mice were challenged with a mixed inoculum consisting of 1:1 ratio of Kan R and Cam R S. Typhimurium 14028 strains containing the sig70_16 luxCDABE construct [42]. Mice were challenged via oral gavage with a total CFU of~10 7 .
Immunized C57BL/6 or BALB/C mice were challenged 21 days after the final immunization by oral gavage with 10 7 CFU of Kan R S. Typhimurium containing the sig70_16 luxCDABE construct.
Infected mice were weighed daily and monitored for clinical signs of infection. Mice that had a >20% drop in weight were humanely euthanized. All mice were humanely euthanized 4-7 days post-infection. Spleen, liver, mesenteric lymph nodes (MLN) and cecum were collected from each mouse. Blood was collected from C57BL/6 mice.
Collected organs were placed in a 2 mL Eppendorf Safe-Lock tube containing 1 mL of phosphate-buffered saline (PBS) and a 5 mm steel bead (Product# 69989; Qiagen, Hilden, Germany) and homogenized using a mixer mill (MM400; Retsch Gmbh, Haan, Germany) for 5 min at 30 Hz. Serial dilutions of organ homogenates and blood were plated on LB agar supplemented with 50 µg/mL Kan and the number of Salmonella CFU were enumerated. Organs collected from mice infected with a mixed inoculum were plated on both Kan and Cam agar. The CI values were calculated as follows: (CFU mutant/CFU wildtype) output /(CFU mutant/CFU wildtype) input .

ELISA
Serum IgG levels specific for CA, tgD, CRM197 and GMMAs were measured by ELISA. Briefly, 96-well plates were coated with CA (2 µg per well), tgD, CRM197 or GMMA (0.1 µg per well) in coating buffer at 4 • C overnight. Wells were blocked for 30 min at RT with 5% skim milk dissolved in Tris-buffered saline (TBS) containing 0.05% tween 20 (TBST). Sera were serially diluted in TBST, starting at 1:100 in 5-fold dilutions. Alkaline phosphatase-conjugated goat-anti-mouse IgG at a dilution of 1:1500 was used to detect bound IgG. The reaction was visualized with p-nitrophenyl phosphate, and absorbance was read at 405 nm with a reference wavelength of 409 nm.

Statistical Analysis
Statistical analyses were performed using GraphPad Prism version 8.0.0 for Mac OS (GraphPad Software, San Diego, CA, USA).
The normality of all datasets was assessed using the Shapiro-Wilk normality test. For luciferase assays, if the data were not normally distributed, comparisons were made using the Mann-Whitney test. If data were normally distributed, comparisons were made using the unpaired t-test.
For murine infections, if the data were normally distributed, comparisons were performed using unpaired t-test with Welch's corrections. If any data sets were not normally distributed, comparisons were performed using the Mann-Whitney test.
For CI experiments, if the data were normally distributed, a one-sample t-test was used to determine if the mean was significantly different from one. If the data were not normally distributed, the Wilcoxon signed-rank test was used to determine if the median CI was significantly different from one.

The yih Operons Are Not Involved in S. Typhimurium Virulence
The divergently transcribed yihUTSRQPO (yihU) and yihVW operons have been identified as the operons controlling O-Ag capsule biosynthesis in Salmonella [9]. Using bioluminescence, White et al. [43] showed that the yihU operon is expressed during infection. Since virulence factors are often optimal vaccine candidates, we wanted to determine if the yih operons are involved in Salmonella virulence. We performed a competitive index experiment in C57BL/6 mice orally challenged with a 1:1 mixture of S. Typhimurium 14028 (wildtype) and S. Typhimurium 14028∆yih. At 4-7 days post-infection, mice were euthanized and bacterial loads in the spleen, liver, MLN and cecum were enumerated. An equal number of wildtype and S. Typhimurium 14028∆yih were recovered from all organs tested (Figure 1), suggesting that the yih operons are not involved in Salmonella virulence.

YihW Represses Expression of the Yih Operons
One of the problems of commercializing CPS-based vaccines is the difficulty to produce large quantities of CPS at an industrial scale. Therefore, our first objective was to boost O-Ag capsule production in Salmonella serovar Typhimurium by increasing the expression of the yih operons. It is often difficult and inefficient to purify CPS in the presence of cellulose because it non-specifically binds the materials in the extracellular matrix [31]. Therefore, a cellulose-negative strain was used for Salmonella CPS purification.
Genes in the yihUTSRQPO operon are homologous to carbohydrate metabolism genes, while genes in the yihVW operon are homologous to regulatory genes (yihV have homology to kinases, while yihW has homology to glycerol-3-P regulon repressor) [9]. To determine if genes in the yihVW operon have a regulatory effect on the expression of the yih operons, yihUTSRQPO and yihVW promoter::lux fusions were used to measure the promoter activity of both operons in S. Typhimurium. Deletion of cellulose synthase (∆bcsA) had no effect on the expression of yihU (Figure 2A), and increased yihVW expression by~70-fold ( Figure 2B). Plasmid-based overexpression of yihVW had no effect on yihVW expression but reduced yihU expression by~20-fold. Deletion of yihVW or yihW alone increased yihU expression by~100-fold (Figure 2A). Our results are consistent with YihW being a transcriptional repressor of the yihUTSRQPO operon.

YihW Represses Expression of the Yih Operons
One of the problems of commercializing CPS-based vaccines is the difficulty to produce large quantities of CPS at an industrial scale. Therefore, our first objective was to boost O-Ag capsule production in Salmonella serovar Typhimurium by increasing the expression of the yih operons. It is often difficult and inefficient to purify CPS in the presence of cellulose because it non-specifically binds the materials in the extracellular matrix [31]. Therefore, a cellulose-negative strain was used for Salmonella CPS purification.
Genes in the yihUTSRQPO operon are homologous to carbohydrate metabolism genes, while genes in the yihVW operon are homologous to regulatory genes (yihV have homology to kinases, while yihW has homology to glycerol-3-P regulon repressor) [9]. To determine if genes in the yihVW operon have a regulatory effect on the expression of the yih operons, yihUTSRQPO and yihVW promoter::lux fusions were used to measure the promoter activity of both operons in S. Typhimurium. Deletion of cellulose synthase (ΔbcsA) had no effect on the expression of yihU (Figure 2A), and increased yihVW expression by ~70-fold ( Figure 2B). Plasmid-based overexpression of yihVW had no effect on yihVW expression but reduced yihU expression by ~20-fold. Deletion of yihVW or yihW alone increased yihU expression by ~100-fold ( Figure 2A). Our results are consistent with YihW being a transcriptional repressor of the yihUTSRQPO operon.
To determine if increased expression from the yihU operon would lead to increased production, O-Ag capsule was purified from S. Typhimurium ΔbcsA and S. Typhimurium ΔbcsA ΔyihW strains. Two times more crude CPS was purified from S. Typhimurium ΔbcsA ΔyihW compared to S. Typhimurium ΔbcsA. To isolate O-Ag capsule from crude CPS, anion and size exchange chromatography were performed and O-Ag capsule containing fractions were identified using EPS-specific serum. Similar quantities of O-Ag capsule were purified from S. Typhimurium ΔbcsA and S. Typhimurium ΔbcsA ΔyihW after chromatography ( Table 3), suggesting that the increase in yihU expression did not lead to increased production of O-Ag capsule.

Effect of Precursor Sugars on yihUTSRQPO Promoter Activity
The lack of adequate amounts of precursor sugars could be one possible explanation for why 100-fold increased expression of the yihUTSRQPO operon in S. Typhimurium ΔbcsA To determine if increased expression from the yihU operon would lead to increased production, O-Ag capsule was purified from S. Typhimurium ∆bcsA and S. Typhimurium ∆bcsA ∆yihW strains. Two times more crude CPS was purified from S. Typhimurium ∆bcsA ∆yihW compared to S. Typhimurium ∆bcsA. To isolate O-Ag capsule from crude CPS, anion and size exchange chromatography were performed and O-Ag capsule containing fractions were identified using EPS-specific serum. Similar quantities of O-Ag capsule were purified from S. Typhimurium ∆bcsA and S. Typhimurium ∆bcsA ∆yihW after chromatography ( Table 3), suggesting that the increase in yihU expression did not lead to increased production of O-Ag capsule.

Effect of Precursor Sugars on yihUTSRQPO Promoter Activity
The lack of adequate amounts of precursor sugars could be one possible explanation for why 100-fold increased expression of the yihUTSRQPO operon in S. Typhimurium ∆bcsA ∆yihW did not lead to increased production of O-Ag capsule. Hence, we wanted to determine if increased amounts of O-Ag capsule could be isolated from S. Typhimurium ∆bcsA ∆yihW grown in media supplemented with precursor sugars required for O-Ag capsule biosynthesis. S. Typhimurium O-Ag capsule is made up of tetrasaccharide repeating units of galactose, rhamnose, mannose and abequose, with abequose and galactose residues partially substituted with a glucose side chain [17]. Using a luciferase reporter assay, we examined the effect of precursor sugars on yihU expression. The addition of mannose, rhamnose and galactose to the growth media increased yihU expression approximately 5-fold in S. Typhimurium ∆bcsA ( Figure 3A), while only the addition of mannose slightly increased yihU expression in S. Typhimurium ∆bcsA ∆yihW ( Figure 3B). However, increased gene expression in the presence of precursor sugars had no measurable effect on O-Ag capsule production (data not shown). This indicated that the lack of increased O-Ag capsule production by S. Typhimurium ∆bcsA ∆yihW was not due to the absence of an adequate amount of precursor sugars.

Overproduction of CPS in Transposon Mutants
Since increased expression of the yih operons and supplementation of growth media with precursor sugars did not boost the amount of O-Ag capsule purified from S. Typhimurium ∆bcsA ∆yihW, we hypothesized that other regulators besides yihW may be negatively regulating the biosynthesis of O-Ag capsule. To identify unknown factors that negatively regulate CPS biosynthesis in Salmonella, random mutagenesis was performed in S. Typhimurium ∆bcsA ∆yihW using Tn10dtet transposon. It was expected that random mutagenesis in S. Typhimurium ∆bcsA ∆yihW, with~100-fold increased expression of yihU, would lead to a mutant strain that overproduces O-Ag capsule.
precursor sugars on yihU expression. The addition of mannose, rhamnose and galactose t growth media increased yihU expression approximately 5-fold in S. Typhimurium ∆bcsA ure 3A), while only the addition of mannose slightly increased yihU expression in S. T murium ∆bcsA ∆yihW ( Figure 3B). However, increased gene expression in the presence o cursor sugars had no measurable effect on O-Ag capsule production (data not shown). indicated that the lack of increased O-Ag capsule production by S. Typhimurium ∆bcsA ∆ was not due to the absence of an adequate amount of precursor sugars. Bacterial colonies overproducing CPS often have mucoid and watery surfaces that can be visually detected [11,44], hence colony morphology was used to identify CPS overproducing strains. After random mutagenesis, seven potential high-CPS-producing isolates with mucoid morphologies were identified (named Tn10A-G). Nested PCR was used to identify the genomic sites of Tn10 insertions (Table 3), and crude CPS was purified from each mutant. At least 2 times more crude CPS was isolated from Tn10 C, D, E and G compared to S. Typhimurium ∆bcsA ∆yihW (Table 3). Following our standard O-Ag capsule purification procedure,~10 times more material was obtained from mutant strain Tn10C, hence subsequent experiments were performed with this mutant. Using luciferase assays, we showed that yihU and yihVW expressions were similar in Tn10C and S. Typhimurium ∆bcsA ∆yihW (Figure 2), suggesting that increased yihU expression was not responsible for the increased amount of CPS produced in the Tn10C strain.

Increased CPS Production by Salmonella Grown in Buffered Media
O-Ag capsule was purified from S. Typhimurium grown in EPS media with a carbon to nitrogen ratio of 10:1. We reasoned that a drop in pH due to increased glucose metabolism may reduce growth and limit the amount of CPS produced by S. Typhimurium ∆bcsA ∆yihW Tn10C. Therefore, we investigated the effect of buffering the pH on CPS production. Approximately four times more crude CPS was isolated from strain Tn10C grown in EPS media buffered with MOPS compared to unbuffered media (Table 3).

Overproduction of Colanic Acid in S. Typhimurium
Since immune serum raised to whole Salmonella EPS [9] was used to guide the purification process, we performed monosaccharide composition analysis to characterize the final purified CPS. The amounts of fucose, galactose and glucose present in our crude and final purified CPS showed that colanic acid was purified from S. Typhimurium ∆bcsA ∆yihW Tn10C, and not O-Ag capsule (Table 4). Furthermore, detection of rhamnose and mannose suggests the presence of contaminating LPS in purified colanic acid.
WcaJ is a glycosylase that is part of the wca gene cluster important for the biosynthesis and transport of colanic acid in Salmonella [45]. To verify the overproduction of colanic acid, we generated a S. Typhimurium ∆bcsA ∆yihW Tn10C ∆wcaJ strain. Deletion of wcaJ resulted in the loss of the mucoid morphology associated with CPS overproduction (data not shown), further indicating that colanic acid was being over-produced from S. Typhimurium ∆bcsA ∆yihW Tn10C.

Removal of Contaminating LPS from Purified CPS
LPS is a common contaminant of purified CPS from Gram-negative bacteria, and since it is toxic in high amounts, the presence of LPS is not desirable in vaccine formulations [46]. Triton X-114 or acid hydrolysis with acetic acid are routinely used to remove LPS from purified CPS [38,39]. To determine the most efficient method for LPS elimination, purified colanic acid was treated with either Triton X-114 or acetic acid.
Triton X-114 is a non-ionic detergent that separates into detergent-rich and detergentpoor phases at room temperature [46]. When mixed with purified CPS, LPS preferentially goes into the detergent-rich phase, due to non-polar interactions between lipid A and Triton X-114 detergent. Several rounds of separation with Triton X-114 are usually performed to reduce the amount of LPS present in purified CPS [38]. We performed three rounds of LPS removal with Triton X-114, however increasing amounts of material were lost with each separation round. Only 50 mg of colanic acid was isolated from 1000 mg of crude CPS.
Following acid hydrolysis with acetic acid, the endotoxic portion of LPS, lipid A, is cleaved off the core oligosaccharide and forms precipitates which can be separated by centrifugation. After lipid A removal, colanic acid was purified using anion exchange and size exclusion chromatography. Approximately 200 mg of colanic acid was purified from 1000 mg of crude CPS.
After treatment with Triton X-114 or acid hydrolysis, the limulus amebocyte lysate (LAL) assay was used to quantify the amount of LPS (i.e., lipid A) present in purified colanic acid. About 4.9 × 10 4 endotoxin unit (EU)/mL was present after three rounds of LPS removal with Triton X-114, while 1.4 × 10 4 EU/mL of LPS was present after treatment with acetic acid followed by size exclusion chromatography. LPS at a concentration between 1 × 10 3 and 4.55 × 10 6 EU/mL has been shown to be safe when administered to a 20 g mouse [46,47]. Hence, both methods reduced LPS to a safe level, however, less material was lost when acid hydrolysis was used for LPS elimination.

Immune Response Induced by Colanic Acid
The objective of this study was to boost the amount of CPS purified from S. Typhimurium for immunization trials. Random mutagenesis with Tn10dtet transposon aimed at increasing the amount of O-Ag capsule purified from S. Typhimurium led to the overproduction of colanic acid. Colanic acid has previously been associated with survival in non-host environments [11,12], however since it is a conserved CPS, we wanted to determine if immunization with colanic acid could induce a protective immune response against Salmonella.
After primary immunization, two booster immunizations were administered at 21-day intervals. Immunization with colanic acid induced an anti-CPS acid IgG response in C57BL/6 ( Figure 4A) and BALB/C ( Figure 4B) mice. the carrier proteins alone, indicating that successful conjugation had occurred (Sup mentary Figure S1). C57BL/6 mice were immunized with 50 µg of colanic acid conjug to CRM197 (CA-CRM197), while BALB/c mice were immunized with 1 µg of colanic conjugated to tgD (CA-tgD). Immunization with colanic acid or CA-CRM197 ind similar levels of anti-CPS IgG in C57BL/6 mice ( Figure 4A). Immunization with co acid or CA-tgD also induced similar levels of anti-CA IgG in BALB/c mice (Figure however, mice were immunized with 50 times less CA-tgD. Immunization with CA-CRM197 or CA-tgD induced robust anti-CRM197 IgG anti-tgD IgG levels, respectively ( Figure 4C,D). These results indicate that immuniza with colanic acid alone or colanic acid conjugated to a carrier protein induced an imm response in mice.  To generate a specific and strong T-cell-dependent response, CPS are often conjugated to carrier proteins [49,50]. Colanic acid was conjugated to truncated bovine herpesvirus 1 glycoprotein D (tgD) [51,52] or the inactive form of diphtheria toxin (CRM197) [53]. In both cases, we detected high molecular weight material running above the size of the carrier proteins alone, indicating that successful conjugation had occurred (Supplementary Figure S1). C57BL/6 mice were immunized with 50 µg of colanic acid conjugated to CRM197 (CA-CRM197), while BALB/c mice were immunized with 1 µg of colanic acid conjugated to tgD (CA-tgD). Immunization with colanic acid or CA-CRM197 induced similar levels of anti-CPS IgG in C57BL/6 mice ( Figure 4A). Immunization with colanic acid or CA-tgD also induced similar levels of anti-CA IgG in BALB/c mice ( Figure 4B), however, mice were immunized with 50 times less CA-tgD.
Immunization with CA-CRM197 or CA-tgD induced robust anti-CRM197 IgG and anti-tgD IgG levels, respectively ( Figure 4C,D). These results indicate that immunization with colanic acid alone or colanic acid conjugated to a carrier protein induced an immune response in mice.

Immunization with Colanic Acid Does Not Induce a Protective Immune Response
To determine if immunization with colanic acid or colanic acid conjugated to a carrier protein induced a protective immune response, mice were orally challenged with 10 7 CFU of S. Typhimurium three weeks after the final immunizations. Four to seven days postchallenge, mice were euthanized, and liver, spleen, cecum, MLN or blood were collected for bacterial enumeration. Approximately 1-log less Salmonella was recovered from C57BL/6 mice immunized with CA-CRM197 compared to mice immunized with PBS or colanic acid ( Figure 5A). No significant difference in Salmonella CFU levels was detected from BALB/c mice immunized with colanic acid, CA-tgD or PBS (control) ( Figure 5B). These results indicated that an immune response induced by colanic acid alone or conjugated to tgD was not protective against future Salmonella infections.

Immunization with Colanic Acid Does Not Induce a Protective Immune Response
To determine if immunization with colanic acid or colanic acid conjugated to a carrier protein induced a protective immune response, mice were orally challenged with 10 7 CFU of S. Typhimurium three weeks after the final immunizations. Four to seven days postchallenge, mice were euthanized, and liver, spleen, cecum, MLN or blood were collected for bacterial enumeration. Approximately 1-log less Salmonella was recovered from C57BL/6 mice immunized with CA-CRM197 compared to mice immunized with PBS or colanic acid ( Figure 5A). No significant difference in Salmonella CFU levels was detected from BALB/c mice immunized with colanic acid, CA-tgD or PBS (control) ( Figure 5B). These results indicated that an immune response induced by colanic acid alone or conjugated to tgD was not protective against future Salmonella infections. Figure 5. Amounts of S. Typhimurium recovered from mice previously immunized with colanic acid. (A) C57BL/6 mice were immunized with phosphate buffered saline (PBS), colanic acid (CA) or CA conjugated to inactivated diphtheria toxin (CA-CRM197). (B) BALB/c mice were immunized with PBS, CA or CA conjugated to truncated glycoprotein D (CA-tgD). All immunized mice were orally challenged with 10 7 CFU of Salmonella, and 4-7 days post-infection, the liver, spleen, cecum, mesenteric lymph node (MLN) and blood (C57BL/6) were harvested. Organs were homogenized before plating on LB agar supplemented with kanamycin. The log10 CFU values of Salmonella recovered from individual organs from each mouse are shown. The dashed line represents the limit of Figure 5. Amounts of S. Typhimurium recovered from mice previously immunized with colanic acid. (A) C57BL/6 mice were immunized with phosphate buffered saline (PBS), colanic acid (CA) or CA conjugated to inactivated diphtheria toxin (CA-CRM197). (B) BALB/c mice were immunized with PBS, CA or CA conjugated to truncated glycoprotein D (CA-tgD). All immunized mice were orally challenged with 10 7 CFU of Salmonella, and 4-7 days post-infection, the liver, spleen, cecum, mesenteric lymph node (MLN) and blood (C57BL/6) were harvested. Organs were homogenized before plating on LB agar supplemented with kanamycin. The log 10 CFU values of Salmonella recovered from individual organs from each mouse are shown. The dashed line represents the limit of detection of 100 CFU. For each group of mice, the black line represents the median values. Statistical significance: Not significant (ns): p > 0.05.

Immunogenicity of GMMAs
GMMAs are targets for vaccine development because they are rich sources of outer membrane antigens that can often induce a strong immune response [20]. We wanted to determine if immunization with GMMAs purified from CPS overproducing isolates would induce a protective immune response against Salmonella. The deletion of tolR affects the stability of the linkage between the inner and outer membrane and results in increased GMMA production [54,55]. A tolR mutation has previously been linked to increased GMMA production in Salmonella [22], hence we purified GMMAs from colanic acid overproducing Tn10C with a tolR deletion (Tn10C GMMAs). As controls, we also purified GMMAs from S. Typhimurium ∆tolR (wildtype GMMAs) and S. Typhimurium ∆tolR ∆lon (Lon GMMAs). Lon is a cytoplasmic serine protease responsible for the proteolytic cleavage of several proteins, and S. enterica strains with lon mutations have been shown to overproduce colanic acid [56].
Groups of six C57BL/6 mice were immunized intramuscularly with 50 µg of GMMAs purified from each strain. Two booster immunizations of 5 µg were given to each mouse at 3-week intervals. Serum from immunized mice was screened for the presence of antibodies that were specific to GMMAs and CPS. Immunization with GMMAs purified from wildtype, ∆lon or Tn10C strains induced a significant amount of anti-CPS antibodies 42 days after primary immunization ( Figure 6A). There were no statistically significant differences between the amount of anti-CPS IgG induced by wildtype, Lon and Tn10C GMMAs.
creased GMMA production in Salmonella [22], hence we purified GMMAs from colanic acid overproducing Tn10C with a tolR deletion (Tn10C GMMAs). As controls, we also purified GMMAs from S. Typhimurium ΔtolR (wildtype GMMAs) and S. Typhimurium ΔtolR Δlon (Lon GMMAs). Lon is a cytoplasmic serine protease responsible for the proteolytic cleavage of several proteins, and S. enterica strains with lon mutations have been shown to overproduce colanic acid [56].
Groups of six C57BL/6 mice were immunized intramuscularly with 50 µg of GMMAs purified from each strain. Two booster immunizations of 5 µg were given to each mouse at 3-week intervals. Serum from immunized mice was screened for the presence of antibodies that were specific to GMMAs and CPS. Immunization with GMMAs purified from wildtype, Δlon or Tn10C strains induced a significant amount of anti-CPS antibodies 42 days after primary immunization ( Figure 6A). There were no statistically significant differences between the amount of anti-CPS IgG induced by wildtype, Lon and Tn10C GMMAs.
Immunization with wildtype, Lon or Tn10C GMMAs induced increasing amounts of anti-GMMAs IgG on days 21, 42 and 63 ( Figure 6B-D). Our results indicate that immunization with GMMAs induces CPS-and GMMA-specific immune responses.

Immunization with GMMA Reduced Salmonella Colonization of Mice Organs
To assess the level of protection induced by GMMAs, three weeks after the final immunizations, mice were orally challenged with 10 7 CFU of S. Typhimurium. All mice were euthanized 4-7 days after challenge and protection was assessed by the levels of S. Typhimurium colonization in the liver, spleen, cecum, MLN and in blood samples.
S. Typhimurium was not recovered from 3 of 5 mice immunized with wildtype GM-MAs, and 4 of 5 mice immunized with Tn10C GMMAs, whereas 4 of 5 mice immunized with Lon GMMAs and 5 of 6 non-immunized control mice had significant levels of S. Typhimurium colonization (Figure 7). Based on these results, we hypothesize that im-munization with wildtype and Tn10C GMMAs induced a partially protective immune response against S. Typhimurium.
ized with Lon GMMAs and 5 of 6 non-immunized control mice had significant levels of S. Typhimurium colonization (Figure 7). Based on these results, we hypothesize that immunization with wildtype and Tn10C GMMAs induced a partially protective immune response against S. Typhimurium.
Mice immunized with wildtype or Tn10C GMMAs showed ~5-log reduction of bacterial colonization in the liver and spleen, and ~3-log reduction in the cecum, MLN and blood, compared to control mice (Figure 7). Mice immunized with Lon GMMAs showed a 1-log reduction in the liver, spleen, cecum and MLN and a 4-log reduction in the blood compared to control mice (Figure 7). Our results indicate that immunization with GMMAs reduces bacterial colonization of mice organs.

Discussion
Immune responses directed against CPS form the basis for some of the most successful human vaccines. CPS conjugated to carrier proteins are the basis for most CPS-based vaccines currently licensed for use, against human pathogens such as Haemophilius influenzae type b, Neisseria meningitidis, Streptococcus pneumoniae and Salmonella serovar Typhi Figure 7. Ability of GMMAs to protect mice against lethal challenge of S. Typhimurium. C57BL/6 mice were immunized with PBS and GMMAs purified from S. Typhimurium ∆tolR (wildtype), S. Typhimurium ∆tolR ∆lon (Lon) and S. Typhimurium ∆tolR ∆bcsA ∆yihW Tn10dtet (Tn10C). Immunized mice were orally challenged with 10 7 CFU of S. Typhimurium 14028, and 4-7 days post-infection, mice were euthanized, and liver, spleen, cecum, MLN and blood were collected for bacterial enumeration. The log 10 CFU of Salmonella recovered from each mouse is shown. The dotted line represents the limit of detection of 100 CFU. The black line represents the median log CFU values determined from each group of mice.
Mice immunized with wildtype or Tn10C GMMAs showed~5-log reduction of bacterial colonization in the liver and spleen, and~3-log reduction in the cecum, MLN and blood, compared to control mice ( Figure 7). Mice immunized with Lon GMMAs showed a 1-log reduction in the liver, spleen, cecum and MLN and a 4-log reduction in the blood compared to control mice ( Figure 7). Our results indicate that immunization with GMMAs reduces bacterial colonization of mice organs.

Discussion
Immune responses directed against CPS form the basis for some of the most successful human vaccines. CPS conjugated to carrier proteins are the basis for most CPS-based vaccines currently licensed for use, against human pathogens such as Haemophilius influenzae type b, Neisseria meningitidis, Streptococcus pneumoniae and Salmonella serovar Typhi [57][58][59]. There is currently no licensed vaccine against NTS. Due to the success of Vi CPS-based vaccines against typhoid fever [60,61], there is renewed interest in developing a CPS-based vaccine against NTS serovars that cause gastroenteritis and invasive blood stream infections.
Colanic acid is a common CPS mainly associated with survival of Salmonella cells in adverse environmental conditions [12]. However, it has been shown to contribute to Salmonella biofilm formation on mammalian cell lines and chicken intestinal epithelium [61]. The role of SrfA in CPS biosynthesis is unknown; however, insertional inactivation of srfA led to increased colanic acid production. SrfA is a virulence effector involved in modulating the host inflammatory response by promoting the activation of NF-κB signaling [62,63]. Since colanic acid may be produced in vivo during infection, we wanted to determine if a colanic acid-specific immune response could provide protection against lethal S. Typhimurium challenge. Similar CFU of Salmonella were recovered from mice immunized with colanic acid or PBS (control). Salmonella overproducing colanic acid has previously been shown to be susceptible to the bactericidal activity of human serum and to phagocytemediated killing [44,64]. Based on these results, we conclude that immunization with colanic acid does not induce a protective immune response. Although similar CFU values of S. Typhimurium were recovered from the organs of CA-tgD-and PBS-immunized mice, the reduced CFU recovered from mice immunized with CA-CRM197 suggests that colanic acid conjugated to CRM197 can induce a partially protective immune response in mice.
GMMAs are attractive vaccine candidates because they are comprised of multiple antigens that can stimulate an immune response [65]. Several studies have focused on developing GMMAs purified from bacteria with different genetic modifications as vaccine candidates for NTS [29]. GMMAs purified from S. Typhimurium with truncated LPS or deficient in flagellin production have been shown to induce cross-reactive antibody responses and cross-protection against Salmonella serovar Choleraesuis and Enteritidis [66,67]. Our results showed that GMMAs purified from colanic overproducing Salmonella reduced bacteria colonization in mice organs. Although Tn10C and Lon GMMAs were isolated from colanic acid overproducing strains, immunization with Tn10C GMMAs reduced bacterial colonization to a greater extent compared to Lon GMMAs. This result suggested the presence of different antigens on Tn10C GMMAs as compared to Lon-GMMAs. This conclusion is supported by several studies that have shown that the composition of GMMAs can vary with regards to growth phase, growth condition and genetic mutation present in bacteria [68][69][70]. Based on our results, we propose that engineered GMMAs with high CPS content should be developed as vaccine candidates. LPS-OAg and GMMAs alone have been shown to induce protective immune responses against NTS in mouse models of infection [14,29]. However, none of these antigens have been successfully developed as a vaccine against NTS. Since GMMAs indue a more diverse immune response [70], and LPS-OAg is the target of protective immune response against NTS [14][15][16], we speculate that GMMAs with high concentrations of LPS-OAg will induce a diverse and cross-protective immune response against NTS.
The S. enterica O-Ag capsule was thought to be a potential vaccine candidate because it has similar repeating units as LPS O-Ag, which is the target of protective immunity against NTS [14][15][16]. Gibson et al. [9] speculated that precursor sugars synthesized by the LPS O-Ag machinery are modified, assembled and translocated out of the cell as O-Ag capsules by gene products from the divergent yihUTSRQPO and yihVW operons [9]. The increased expression of yihUTSRQPO in a ∆yihW mutant is consistent with YihW being a repressor of the yih operons. The repressive effect of YihW has also been described in E. coli, where three yihW binding sites were identified in the yih promoter region [71,72]. Overexpression of genes often leads to the overproduction of specific proteins or CPS. However, a 100-fold increased expression of the yihUTSRQPO operon in the ∆yihW mutant did not result in increased production of O-Ag capsule. The introduction of precursor sugars into growth media often leads to increased CPS biosynthesis [73][74][75]. However, O-Ag capsule biosynthesis was not increased in the presence of excess precursor sugars. In addition, random mutagenesis aimed at increasing the amount of O-Ag capsule produced by S. Typhimurium ∆bcsA ∆yihW led to increased colanic acid production. Based on these results, we speculate that the yih operons are not involved in biosynthesis of the O-Ag capsule.
Using enzymatic assays with purified proteins and mass spectrometric analysis of products, genes in the yih operons have been shown to be required for the catabolism of sulphoquinovose (SQ) by E. coli [76,77]. SQ is a major constituent of the human diet and about 10 pentagrams are produced by photosynthetic organisms annually [78]. Analysis of Salmonella mutants indicated that genes in the yih operons are involved in the colonization and persistence of Salmonella on plant surfaces [79,80], biofilm formation [81], resistance to desiccation [9] and resistance to killing by human immune serum [13]. Due to the high nucleotide (80%) and protein (80-92%) sequence identity with the E. coli operon, we speculate that the yih operons are also involved in SQ catabolism by S. enterica. However, due to the phenotypes associated with mutant strains, there is a possibility that genes in the yih operons have dual functionality. In Salmonella, the yih operons may be involved in SQ catabolism and in the synthesis of an unknown CPS or the modification of the antigenic content of the outer membrane. The serum used by Gibson et al. [9] for the initial identification of the yih operons was generated against whole Salmonella EPS fraction and was found to be cross-reactive to an uncharacterized CPS [9]. This might explain the lack of CPS in S. Typhimurium ∆yihO observed by Marshall and Gunn [13] using confocal microscopy [9,13]. A role for the yih operons in the modification of the antigenic content of the outer membrane provides an explanation for the exclusive production of phase 1 flagellin FliC by S. Typhimurium ∆yih mutants, and the production of~25% to 45% more short LPS (with 1 to 8 LPS O-Ag repeating units) by S. Typhimurium ∆yih mutants compared to wildtype [13]. More research is required to clarify the role of the yih operons in Salmonella.
Based on our research, we predict that the O-Ag capsule described by Gibson et al.  [82][83][84]. We speculate that the O-Ag capsule and VL-O-Ag could be the same CPS because they both have similar banding patterns on SDS-PAGE [9,13], are composed of similar tetrasaccharide repeat units [17] and have been implicated in similar functions [13,83]. Although the O-Ag capsule was found to be glycosylated at the tyvelose and galactose residues, high molecular weight LPS with varying glycosylation levels has been previously described in Salmonella [84]. Hence, the O-Ag capsule may be VL-O-Ag glycosylated at the tyvelose and galactose sugars. More research is needed to determine the structure of this polysaccharide and others that are present on the cell surface of S. enterica isolates.

Conclusions
Although CPS has been the focus of subunit vaccine development against NTS, none have been tested in clinical trials. The immune response to LPS-O-Ag is strain-specific, while colanic acid does not induce a protective immune response. Research on CPS-based vaccines against NTS will have to focus on identifying novel CPS that can be developed as vaccines. Currently, GMMAs may be the best strategy for vaccine development, because they are highly immunogenic, deliver multiple antigens in their natural context and can induce a cross-protective immune response. With the rise in multi-drug-resistant isolates and the increasing incidence of invasive bloodstream infections, a vaccine is urgently needed to reduce the prevalence of NTS infections.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.