Multiple immunodominant O-epitopes co-expression in live attenuated Salmonella serovars induce cross-protective immune responses against S. Paratyphi A, S. Typhimurium and S. Enteritidis

Salmonella enterica subsp. enterica (S. enterica) is a significant public health concern and is estimated to cause more than 300,000 deaths annually. Nowadays, the vaccines available for human Salmonellosis prevention are all targeting just one serovar, i.e., S. Typhi, leaving a huge potential risk of Salmonella disease epidemiology change. In this study, we explored the strategy of multiple immunodominant O-epitopes co-expression in S. enterica serovars and evaluated their immunogenicity to induce cross-immune responses and cross-protections against S. Paratyphi A, S. Typhimurium and S. Enteritidis. We found that nucleotide sugar precursors CDP-Abe and CDP-Par (or CDP-Tyv) could be utilized by S. enterica serovars simultaneously, exhibiting O2&O4 (or O4&O9) double immunodominant O-serotypes without obvious growth defects. More importantly, a triple immunodominant O2&O4&O9 O-serotypes could be achieved in S. Typhimurium by improving the substrate pool of CDP-Par, glycosyltransferase WbaV and flippase Wzx via a dual-plasmid overexpressing system. Through immunization in a murine model, we found that double or triple O-serotypes live attenuated vaccine candidates could induce significantly higher heterologous serovar-specific antibodies than their wild-type parent strain. Meanwhile, the bacterial agglutination, serum bactericidal assays and protection efficacy experiments had all shown that these elicited serum antibodies are cross-reactive and cross-protective. Our work highlights the potential of developing a new type of live attenuated Salmonella vaccines against S. Paratyphi A, S. Typhimurium and S. Enteritidis simultaneously.

Introduction Salmonella belongs to the family Enterobacteriaceae and is a medically important pathogen for both humans and animals. Based on its surface antigenic composition, Salmonella is currently divided into more than 2600 serotypes [1]. However, 99% of human and animal infections are caused solely by one subspecies, Salmonella enterica subsp. enterica (S. enterica). S. enterica is estimated to cause more than 300,000 deaths annually [2,3], mostly in developing countries. According to their clinical manifestations and presentations, S. enterica has traditionally been divided into 'typhoidal serovars' and 'non-typhoidal serovars' [4,5]. For example, human hostrestricted S. enterica serovar Typhi (S. Typhi) and Paratyphi A (S. Paratyphi A) are the leading causes of systemic infections known as typhoid and paratyphoid fever [2], respectively. In contrast, broad host-ranged S. enterica serovar Typhimurium (S. Typhimurium) and Enteritidis (S. Enteritidis) generally induce self-limiting gastroenteritis in healthy individuals [6]. However, non-typhoidal serovars may become invasive when the host are infants, young children or immunocompromised adults, causing a life-threatening infection involving the bloodstream, meninges, and other normally sterile sites [7]. Invasive Non-typhoidal Salmonella (iNTS) disease is a severe illness with a case fatality ratio of approximately 15%. In sub-Saharan Africa, where the iNTS is observed to be a particular threat [8][9][10], S. Typhimurium and S. Enteritidis were the most frequently isolated iNTS pathogens [8], accounting for more than 80%. Unfortunately, widespread antimicrobial resistance among iNTS isolates is threatening the effectiveness of amenable antibiotic treatments [11]. To date, vaccines are regarded as one of the most economical and effective ways to prevent salmonellosis.
Immunity to Salmonella, induced by natural infection or vaccination, is serotype-specific [12]. In S. enterica serovar, this serotype specificity is largely determined by the O-antigen polysaccharide [13] or Vi capsule [14]. The Vi capsule is produced by S. Typhi, while the Oantigen is widely found in other S. enterica serovars. The O-antigen polysaccharide is the outermost part of the lipopolysaccharides (LPS), which is a structurally diverse polymer and repeats in a diverse range of numbers. The LPS is found exclusively in the outer leaflet of the Salmonella membrane [15]. Approximately 2×10 6 LPS molecules cover *75% of the cell surface, thus resulting in a formidable barrier limiting the antibodies from accessing the bacterial surface [16]. Nowadays, the only licensed single antigen vaccine against Salmonella infections is rationally based on their surface polysaccharides, i.e., the Vi capsule [17] or Vi capsule glycoconjugate vaccines [18]. However, multi-valent strategies are being explored in clinical development. For example, a bivalent outer membrane vesicle approach, also referred to as Generalized Modules for Membrane Antigens (GMMA) [19], targets S. Enteritidis and S. Typhimurium, and a trivalent glycoconjugate approach targets S. Enteritidis, S. Typhimurium, and S. Typhi [20,21]. Both of them involve in O-antigens formulation. Consequently, an OAgbased vaccine covering the other frequently isolated strains (i.e., S. Paratyphi A, S. Typhimurium and S. Enteritidis) is predictably desirable.
The O-antigen gene cluster of S. Paratyphi A, S. Typhimurium and S. Enteritidis are all located between the galF and gnd gene in the chromosome [25,26] (S1 Fig). The main differences among these gene clusters are the regions responsible for synthesizing the 3,6-dideoxyhexosyl side-branch sugars, and the side-branch sugars are transferred from CDP-sugar precursors by glycosyltransferase WbaV [27,28]. The CDP-sugars synthesis pathways are all initiated from glucose-1-phosphate but diverge after DdhD. CDP-abequose synthase (Abe) and CDP-paratose synthase (Prt) reduce the keto group at C-4 to give either the galactose stereochemistry for CDP-Abe or the glucose stereochemistry for CDP-Par. CDP-Par 2-epimerase (Tyv) is responsible for the synthesis of CDP-tyvelose [27] (Fig 1B). Meanwhile, the Wzx is a multi-transmembrane protein with enormous sequence diversity that flips oligosaccharide substrates with varying degrees of preference [29,30].
Previously, we and others have shown that the immunodominant O-epitope of one S. enterica strain could be converted to another via chromosomally genetic manipulation [31][32][33]. For example, the O4 O-serotype in S. Typhimurium could be converted into O9 of S. Enteritidis by replacing the gene abe with prt-tyv D1 without sacrificing its immunogenic properties [31,33]. Similarly, S. Typhimurium with abe-wzx B1 -wbaV B1 gene replaced by prt-tyv A1 -wzx A1 -wbaV A1 from S. Paratyphi A could convert its original O4 O-serotype into O2 and still retained an excellent immunogenic property [32]. So, we later come up with an interesting question. Could it be possible to simultaneously express immunodominant O2, O4 and O9 O-epitopes in one strain? In

Ethics statement
All animal experiments were conducted in compliance with the Animal Welfare Act and regulations stated in the Guide for the Care and Use of Laboratory Animals, which was approved by the Institutional Animal Care and Use Committee (IACUC) of Southwest University (IACUC-20210525-02).

Bacteria, plasmids, and culture conditions
The bacteria and plasmids used in this study are listed in Table 1. E. coli, S. enterica serovars and their derivatives were aerobically grown at 37˚C in Luria-Bertani (LB) broth or on LB agar. Allelic exchange in S. enterica serovars was achieved by sacB gene-based counter selection on LB agar plates that contained 10% sucrose with no sodium chloride added and incubated at 30˚C [34]. When necessary, chloramphenicol was added at 25 μg/ml to select transconjugants. D-Alanine (D-Ala) (50 μg/ml) was added for the growth of Δalr ΔdadB strains [35]. Diaminopimelic acid (DAP) was added at 50 μg/ml for the growth of Δasd mutant χ7213 [36]. In vitro growth rates of Salmonella strains were determined by optical density measurements.

Molecular and genetic procedures
Molecular biology techniques were performed following standard methods [37], and details regarding the primers used in this study are listed in the S1 Table. DNA concentration and purity were measured using a Nanodrop ND-2000 spectrophotometer (Thermo Fisher Scientific), and DNA fragments were cyclized by Circular Polymerase Extension Cloning (CPEC) method [38]. Insertion or deletion mutant strains construction. As antibiotic resistance is not permitted in live vaccine strains, sacB gene-based sucrose counter-selectable suicide vectors were used to construct unmarked insertion or deletion mutations in S. enterica serovars [34]. Take the construction of L001 (SA-ddhc::abe) as an example; the upstream and downstream homologous regions were amplified from S. Paratyphi A using primer pairs P11/P12 and P15/P16. The abe gene was amplified from S. Typhimurium using primer pairs P13/P14 and the suicide vector backbone was amplified from pRE112 using primer pairs P1/P2. After purification, these three fragments were cyclized by CPEC method, resulting in a new suicide plasmid pHY005 (112ddhc::prt A1 ). The conjugational transfer of pHY005 to S. Paratyphi A was performed using the suicide vector donor strain χ7213 [39]. Transconjugants with the first homologous recombination event were selected on chloramphenicol agar without supplemental DAP. The positive clones were inoculated in fresh LB media without chloramphenicol addition. The second homologous recombination event, resulting in excision of the suicide vector from the S. Paratyphi A chromosome, was selected on 10% sucrose LB plates without sodium chloride and grown at 30˚C. Successful gene insertion mutations were confirmed by PCR screening and DNA sequencing. Other insertion mutations are constructed following the same procedures. The insertion mutations constructed for this study are illustrated in S2 Fig, and the primer pairs used for each DNA fragment amplification were labeled accordingly. As for the crp and cya gene deletion mutations, the suicide plasmids 112-Dcrp and 112-Dcya were constructed previously [33]. The subsequent deletion mutation processes were similar to those described above for the insertion mutations. All successful gene deletion or insertion mutations were confirmed by DNA sequencing.
Recombinant plasmids construction. Briefly, prt, tyv, wzx and wbaV genes were cloned from S. Paratyphi A or S. Enteritidis using the same primer pairs P19/P20. To distinguish them from each other, we added an "A1" suffix subscript after each gene cloned from S. Paratyphi A, as S. Paratyphi A belongs to S. enterica serogroup A1. Similarly, a "D1" suffix subscript was added to each gene cloned from S. Enteritidis, as S. Enteritidis belongs to S. enterica serogroup D1. Next, the vector backbone was cloned from pSC101-asd and p15a-dadB using the same primer pairs P17/P18, as these two plasmids all contained a Ptrc promoter-multiple cloning site (MCS)-5ST1T2 terminator cassette. After purification, the DNA fragment prt A1 -tyv A1wzx A1 -wbaV A1 and the linearized pSC101-asd vector were cyclized by CPEC method, resulting in pSC101-asd-O2. Similarly, the DNA fragment prt D1 -tyv D1 -wzx D1 -wbaV D1 and the linearized p15a-dadB vector were cyclized by CPEC method, resulting in p15a-dadB-O9.

Phenotype characterization of mutant strains
Mutant strains phenotype evaluations included the LPS silver stain and western blot, slide agglutination test, P22 transduction studies, growth rates, motility test, and minimum inhibitory concentration (MIC) test of deoxycholate (DOC) and polymyxin B. These methods were all reported previously, and a complete description of all methods employed for these phenotype evaluations is provided in S1 Text.

Virulence determination and colonization in mice
Six-week-old female BALB/c mice were purchased from Dashuo Biotechnology Co., Ltd. (Chengdu, China). To determine the 50% lethality dose (LD 50 ), bacteria were grown to OD 600 of 0.8 to 0.9 and harvested by centrifugation at 3,452 × g at room temperature. The centrifuged S. Typhimurium, S. Enteritidis or their derivatives were resuspended and adjusted to the appropriate OD 600 value by buffered saline with gelatin (BSG) [40]. In contrast, the centrifuged S. Paratyphi A or its derivatives were resuspended and adjusted to the appropriate OD 600 value by 10% hog gastric mucin (Sigma). Six mice per group were infected orally with 20 μl of BSG or intraperitoneally injected with 500 μl 10% hog gastric mucin containing various doses of bacteria, ranging from 1 × 10 4 CFU to 1 × 10 8 CFU. Mice were monitored for mortality or signs of significant morbidity daily. The LD 50 was calculated using the method of Reed and Muench. To evaluate colonization, three mice per group were orally inoculated with 20 μl of BSG containing 1 × 10 9 CFU bacteria. On days 4 and 8 post-inoculation, Peyer's patches, spleen and liver samples were collected. Samples were homogenized, dilutions were plated onto MacConkey and LB agar to determine viable counts.

Vaccination and immune response measurement
Thirty mice per group were vaccinated orally on day 0 with 20 μl BSG containing approximately 1 × 10 9 CFU vaccine strains and boosted on day 14 with the same dose. Blood samples and vaginal secretions were collected from randomly selected twelve mice in each group on day 28 after the booster immunization. On day 35, mice in each group were randomly subdivided into three small subgroups; mice in each subgroup were challenged orally with 5 × 10 7 CFU of S. Typhimurium and S. Enteritidis or injected intraperitoneally with 1 × 10 4 CFU of S. Paratyphi A (~100 times the LD 50 ) [40,41]. S. Paratyphi A, S. Typhimurium and S. Enteritidis LPS were purified as described previously [42] and used as coating antigens to measure the immune responses. IgM, IgG and IgA antibodies specific to S. Paratyphi A, S. Typhimurium and S. Enteritidis LPS in the serum or vaginal secretions were measured using the quantitative enzyme-linked immunosorbent assay (ELISA) as described previously [43]. Antibody concentrations were calculated based on absorbance values and the standard curve.

Serum bactericidal activity assay
The serum bactericidal assay (SBA) was performed as previously described with a few modifications [44]. Briefly, log-phase cultures of S. Paratyphi A, S. Typhimurium and S. Enteritidis were grown to an OD 600 of 0.6 in LB broth. After centrifugation and resuspension, the logphase cultures were diluted in PBS to a concentration of approximately 1× 10 4 CFU/ml. Sera samples from vaccinated mice were serially diluted from 1:100 to 128,000, and nonimmune sera were serially diluted from 1: 10 to 1,280. Optimal SBA results were achieved by combining 25 μl of active baby rabbit complement (BRC) (25% final concentration) with 15 μl of PBS, 50 μl of diluted mice pooled sera, and 10 μl of diluted bacteria (~350 CFU). In total, 10 μl of the mixture from each well was spread on LB agar plates after 60 min to assess the bactericidal activity. The spread LB agar plates were incubated overnight at 37˚C, and the viable CFU were counted the next day. The negative control contained only bacteria and complement. The serum bactericidal antibody titer was defined as the reciprocal of the highest serum dilution that produced >50% killing in relation to the killing observed for the control wells. The titers were determined from the mean bacterial count from triplicate wells.

Statistical analysis
Data were analyzed using GraphPad Prism 5 software (Graph Software, San Diego, CA) by one-way or two-way ANOVA of variance followed by Tukey's multiple-comparison post-test. Kaplan-Meier survival curve comparisons were calculated by comparing two groups at each time point through the log-rank (Mantel-Cox) test. The data were expressed as the means ± SD. P<0.05 was considered statistically significant.

S. enterica mutants exhibiting double or triple immunodominant Oserotypes
The primary goal of this study is to try to exhibit O2, O4 and O9 O-serotypes simultaneously in one S. enterica serovar. To begin with, we explore the possibility of co-expressing O2 & O4 or O4 & O9 O-epitopes first. According to the O-antigen gene cluster analysis (S1 Fig), the major difference among S. Paratyphi A, S. Typhimurium and S. Enteritidis lie in the genes responsible for the synthesis of the side-branch sugar (Fig 1B), i.e., the abe, prt and tyv genes for CD-Abe, CD-Par and CDP-Tyv biosynthesis respectively. Consequently, we inserted the abe gene from S. Typhimurium between the ddhc and prt gene of S. Paratyphi A to test the potential of coexpressing O2 and O4 O-epitopes (S2A Fig). The results of bacterial agglutination ( Table 2) and western blot (Fig 2A) showed that the SA-ddhc::abe mutant could react positively to O2 and O4 antisera, respectively, indicating the co-existence of O2 and O4 O-epitopes in the outer membrane of SA-ddhc::abe mutant. To be more persuasive, we inserted the prt A1 gene from S. Paratyphi A similarly between the ddhc and abe gene of S. Typhimurium (S2B Fig) to see if these coexpressing phenomena still existed. Consistently, the bacterial agglutination (Table 2) and western blot results (Fig 2A) all indicated that the ST-ddhc::prt A1 mutant exhibited O2 and O4 Oepitopes simultaneously. At the same time, the abe gene was inserted between the ddhc and prt D1 gene of S. Enteritidis (S2C Fig) and the prt D1 -tvy D1 gene was inserted between the ddhc and abe gene of S. Typhimurium (S2D Fig). Unsurprisingly, either SE-ddhc::abe or ST-ddhc:: prt D1 -tyv D1 could independently agglutinate with O4 and O9 antisera obviously ( Table 2). However, we did not observe clear anti-O9 positive bands of SE-ddhc::abe or clear anti-O4 positive bands of ST-ddhc::prt D1 -tyv D1 in western blot results ( Fig 2B). Unfortunately, we failed to obtain a S. Paratyphi A, S. Typhimurium or S. Enteritidis mutant that could express O2, O4 and O9 Oepitopes simultaneously through chromosomally genetic modifications.
We have noticed that S. Enteritidis possess both functional prt and tyv genes and it turns out to be a dominant O9 O-serotype, while the tyv gene in S. Paratyphi A is frameshift and yet it turns out be a dominant O2 O-serotype. So, we speculate that when prt and tyv are coexpressed at a similar level, the CDP-Par will be mostly converted into CDP-Tyv, resulting in an O9 dominant O-serotype. Therefore, we hypothesize that it might be possible to express O2, O4 and O9 O-epitopes simultaneously by improving the CDP-Par substrate pool. Based on that assumption, we have built a dual-plasmid expression system. The ST-Δalr ΔdadB L058 ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB-O9) ++ + ++ O2&O4&O9 a serum used in this assay were anti-Par O2, anti-Abe O4 and anti-Tyv O9 antiserum. b agglutination observed immediately was indicated as "+++", within 30 sec as "++", more than 1 min as "+".

Phenotype characterizations of S. enterica mutants
As we have achieved double or triple immunodominant O-epitopes co-expression in S. enterica mutants, we would like to know whether or not these O-serotypes' co-existence would influence their phenotype characterizations. Firstly, the LPS profiles of these mutant strains were visualized by silver staining. Most mutants can synthesize a full length of LPS when compared with their wild-type parent strains, except for ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB), which exhibited a decrease in LPS synthesis (Fig 3). Next, the accessibilities of mutants' O-antigen structure for Phage P22 were examined. The number of transductions obtained from double or triple O-serotypes exhibiting mutants was similar to that of their wild-type parent strains (Table 3), indicating that their O-antigen backbones were unaffected. Secondly, the growth rate of these mutant strains and their wild-type parent strains were all evaluated, and there was no significant difference between them (Fig 4). Meanwhile, their sensitivity to the bile salt DOC and the cationic antimicrobial peptide polymyxin B were tested. The DOC MICs had no difference among these strains, whereas the polymyxin B MICs for double or triple O-epitopes co-expressing mutant strains were twofold lower than their wild-type parent strains. All mutants retained similar swimming motility as their wild type parent strains (Table 3). Taken together, all these results indicated that double or triple O-epitopes co-expressing had a minor influence on the phenotype changes of our constructed S. enterica mutants.

Virulence and colonization of the S. enterica mutants in BALB/c mice
The LD 50 values of wild-type S. Typhimurium S100 and S. Enteritidis S246 were approximately 10 5 CFU, whereas the LD 50 values of double or triple O-epitopes co-expressing mutant strains were approximately 10 6 or 10 7 CFU, showing around 10 times or 100 times attenuation (Table 3). Since S. Paratyphi A is a human host-restricted pathogen, its virulence attribute was evaluated in a lethal murine model that requires suspending the bacteria in 5-10% hog gastric mucin and then injecting the suspension intraperitoneally [45]. In this model, the LD 50 value of wild-type S. Paratyphi A was approximately 100 CFU, showing a highly virulent attribute.

Immune responses induced by live attenuated S. enterica vaccines
To assess the immunogenicity of these vaccine candidates, mice were inoculated orally with approximately 10 9 CFU of each strain on day 0 and boosted on day 14 with the same doses. Anti-S. Paratyphi A, anti-S. Enteritidis and anti-S. Typhimurium LPS serum antibodies were measured on day 28. The IgG immune response are depicted in Fig 6. Mice vaccinated with L008 (O2&O4) and L083 (O2&O4&O9) mounted a significantly higher anti-S. Paratyphi A LPS immune responses than those of ST-Δcrp Δcya or SE-Δcrp Δcya. Similarly, mice vaccinated with SE-Δcrp Δcya, L009 (O4&O9) and L083 (O2&O4&O9) mounted a significantly higher anti-S. Enteritidis LPS immune responses than those of ST-Δcrp Δcya. Except for SE-Δcrp Δcya, mice vaccinated with other vaccine candidates could all mount a significantly higher anti-S. Typhimurium LPS responses. We did observe some cross-immunogenicity between ST-Δcrp Δcya and SE-Δcrp Δcya, but it had long been proved to be the common trisaccharide backbone and the conserved core oligosaccharide. So far, we did not obtain valid evidence showing that there were detectable cross-immune responses among immunodominant O2, O4 and O9 O-epitopes unless they were co-expressed simultaneously. L008 (O2&O4) could mount a significantly higher anti-S. Paratyphi A LPS immune response than those vaccinated with ST-Δcrp Δcya (O4), showing an improved cross-immune response against S. Paratyphi A. Consistently, L009 We also evaluated the serum IgM (S5 Fig) and vaginal washes IgA antibodies immune responses (S6 Fig). A similar trend of IgM and IgA immune responses has been observed compared to IgG antibodies. Unsurprisingly, L008 (O2&O4), L009 (O4&O9) and L083 (O2&O4&O9) could induce a significantly higher amount of cross-reactive IgM antibodies than their parent strains. However, the elicited IgM antibodies were even higher than their IgG counterpart, which is unexpected. A significantly higher level of secreted IgA antibodies against S. Paratyphi A and S. Enteritidis were observed in mice vaccinated by L083 (O2&O4&O9) when compared to its parent strain ST-Δcrp Δcya. However, L009 (O4&O9) could elicit a higher level of anti-S. Enteritidis LPS IgA compared to L083 (O2&O4&O9) and the level of anti-S. Paratyphi A LPS IgA was similar between L008 (O2&O4) and L083 (O2&O4&O9). Negative control groups did not mount a detectable immune response. All these ELISA results had shown that our vaccine candidates could induce promising crossimmune responses.

Antibody-dependent complement-mediated S. enterica killing
To evaluate the functional capacities of antibodies induced by our vaccine candidates, we performed the serum bactericidal assays (SBAs) using pooled serum from immunized mice or BSG control. Baby rabbit complement titration with serum samples from immunized mice had a significantly higher complement-mediated killing activity versus those from nonimmunized mice (Fig 7). The serum antibodies induced by ST-Δcrp Δcya (O4) or SE-Δcrp Δcya (O9) exhibited a high level of SBA activity against its homologous serotype strains. However, they had a limited SBA activity against heterologous serotype strains. On the contrary, the serum antibodies induced by L008 (O2&O4) and L009 (O4&O9) significantly improved SBA activity against heterologous serotype strains S. Paratyphi A and S. Typhimurium, respectively. More importantly, the bactericidal abilities induced by L083 (O2&O4&O9) covered all tested heterologous serotype strains, significantly improving SBA activity against S. Paratyphi A, S. Typhimurium and S. Typhimurium, simultaneously.

Protective efficacy of live attenuated S. enterica vaccines against wild-type S. Typhimurium and S. Enteritidis
The immune protection against wild-type S. Paratyphi A, S. Typhimurium and S. Enteritidis were evaluated in the murine model (Fig 8). Unfortunately, most of the vaccinated mice succumbed to the 100 times LD 50 challenge from S. Paratyphi A. Only three mice, one out of twelve in L008 (O2&O4) and two out of twelve in L083 (O2&O4&O9) vaccinated group, survived in the end. However, mice vaccinated by L008 (O2&O4) and L083 (O2&O4&O9) could generally survive as long as 4 days, while the other vaccinated groups could barely survive a single day, showing a positive correlation of cross-protection. Meanwhile, mice vaccinated by ST-Δcrp Δcya (O4) or SE-Δcrp Δcya (O9) could survive a 100 times LD 50 challenge from S. Typhimurium or S. Enteritidis, respectively, indicating a high level of homologous protection. However, these mice succumbed to the heterogenous challenge. Intriguingly, mice vaccinated

Discussion
Traditionally, Salmonella vaccine developments followed the strategies of targeting one particular serovar or serotype, leaving the cross-protection largely unexplored. However, recently, a bivalent outer membrane vesicle (GMMAs) approach targeting S. Enteritidis and S. Typhimurium [19] and a trivalent glycoconjugate approach targeting S. Enteritidis, S. Typhimurium, and S. Typhi [21] are showing the light of multi-valent strategies. In this study, we explored another way of improving the cross-protections against S. Paratyphi A, S. Typhimurium and S. Enteritidis by O-antigen O-epitopes rational design.
The antisera of patients who recovered from paratyphoid fever or iNTS diseases mainly target the O-antigen polysaccharides of S. enterica serovars, indicating the importance of anti-OAg antibodies in patients' adaptive immune responses. The O-serotype specificity of S. Paratyphi A, S. Typhimurium and S. Enteritidis is determined by the 3,6-dideoxyhexosyl sidebranch residues of the trisaccharide backbone (Figs 1 and S1) [22]. Therefore, our strategy was to exhibit double or triple immunodominant O-serotypes in one S. enterica serovar. In order to achieve that goal, we first explored the possibility of co-expression of two O-epitopes. As the results had shown, either inserting the abe gene in front of the prt A1 gene in S. Paratyphi A (SA-ddhc::abe) or the prt A1 gene in front of the abe gene in S. Typhimurium (ST-ddhc::prt A1 )  (Table 2). Consistently, positive bands were observed in the western blot assays when using O2 and O4 antisera as the primary antibody (Fig 2A). These data all indicated that O2 and O4 could be simultaneously expressed in either S. Paratyphi A or S. Typhimurium background. Similarly, we inserted the abe gene in front of the prt D1 gene in S. Enteritidis (SE-ddhc::abe), and the prt D1 gene in front of the abe gene in S. Typhimurium (ST-ddhc::prt D1 -tyv D1 ) to testify the possibility of O4 and O9 O-serotypes exhibition (S2 Fig). We observed a clear bacterial agglutination with O4 and O9 antisera in both SE-ddhc::abe and ST-ddhc::prt D1 -tyv D1 (Table 2). However, the western blot assay did not show anti-O9 positive bands in SE-ddhc::abe sample or anti-O4 positive bands in ST-ddhc::prt D1 -tyv D1 sample (Fig 2B). The possible explanation for this might be the unbalanced expression of O4 and O9 O-epitopes, as this phenomenon has already existed naturally in wild-type S. Enteritidis. In wild-type S. Enteritidis or S. Typhi, the CDP-Par and CDP-Tyv synthesize at a similar level. However, the immunodominant Oserotype of S. Enteritidis or S. Typhi is exclusively O9, with even undetectable bacterial agglutination to O2 antiserum. On the other hand, S. Paratyphi A exhibiting O2 O-serotype is due to the frameshift mutation in tyv gene [47], interrupting the synthesis of CDP-Tyv from CDP-Par, leaving the S. Paratyphi A with no choice but utilizing the CDP-Par as its substrate for side-branch sugar attachment. Lastly, we failed to express O2, O4 and O9 O-serotypes in either S. Paratyphi A, S. Typhimurium or S. Enteritidis strain by chromosomally genetic modification.
It seems that S. enterica serovars have a different preference for the CDP-Abe, CDP-Par and CDP-Tyv substrates when they are synthesized at a similarly level. At least, it is evident that S. enterica prefers CDP-Tyv rather than CDP-Par [28]. Therefore, we hypothesized that achieving an O2, O4 and O9 O-serotype phenotype might be possible by increasing the synthesis of CDP-Par. So, we built a dual-plasmid expression system in a S. Typhimurium mutant, i.e., ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd or p15a-dadB). As the WbaV glycosyltransferase and Wzx flippase might have influenced synthesizing and transferring of different side-branch O-units, we amplified the whole prt A1 -tyv A1 -wbaV A1 -wzx A1 genes from S. Paratyphi A and inserted them into pSC101-asd vector, resulting in pSC101-asd-O2 in brief (S2 Fig). At the same time, we amplified the whole prt D1 -tyv D1 -wbaV D1 -wzx D1 genes from S. Enteritidis and inserted them into p15a-dadB vector, resulting in p15a-dadB-O9 in brief. Predictably, ST-Δalr ΔdadB ΔrecF Δasd harboring both pSC101-asd-O2 and p15a-dadB-O9 would increase CDP-Par synthesis, as there are at least two copies of prt gene transcribing but with only one normal tyv gene. Meanwhile, the synthesis of CDP-Abe was unaffected. The bacterial agglutination assays demonstrated that ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB-O9) could agglutinate with O2, O4 and O9 antiserum individually, indicating that we had successfully achieved a triple immunodominant O-serotypes phenotype in S. Typhimurium background (Table 2). However, we did not receive a consistent result from western blot assays, which showed a predominantly anti-O9 positive band (Fig 2C). This inconsistency between the bacterial agglutination and western blot assay might indicate the amount or distribution of O2, O4 and O9 O-epitopes in ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB-O9) is most likely unequal or unbalanced. We could not explain clearly how these O-epitopes are arranged in the outer membrane, either homogeneously or heterogeneously attached to each of the O-antigen polysaccharides (S7 Fig), or merely mixed irregularity. It is also possible that new irrelevant epitopes are created and, in this case, a portion of the antibody response may not be functional.
Liu etc. had shown that the glycosyltransferase WbaV of S. enterica is in fact more effective in attaching the Tyv to the O-antigen common trisaccharide backbone than Par, and this inefficiency of low Par utilization could be primarily improved by overexpressing the wbaV gene [28]. Their findings were reconfirmed in this study. Meanwhile, Hong etc. had also shown that the Wzx flippases have a strong preference for their cognate substrate [29,30]. Furthermore, they demonstrated that the absence of the side-branch resides would dramatically decrease the efficiency of Wzx flippases [29]. However, no direct evidence had been put forward to illustrate the influence of different side-branch resides upon Wzx translocation efficiency. Our previous report demonstrated that the S. Typhimurium could tolerate well Tyv side-branch O-units but reject Par side-branch O-units [32], indicating that nonnative substrate does influence Wzx translocation efficiency. However, in this study, we further developed these theories by showing that S. Typhimurium did not exhibit severe rejection against CDP-Par if accompanied by CDP-Abe. Moreover, S. Typhimurium could utilize CDP-Par as its side-branch building block even in the presence of CDP-Tyv.
After successfully constructing a series of double or triple immunodominant O-serotype mutant strains, we would like to know their phenotype characterizations. The phage P22 tailspike protein recognizes and hydrolyzes the repetitive O-antigen polysaccharides of S. enterica serogroup A, B and D at the Rha-Gal (1!3)-glycosidic linkages during infections [48]. So, we first performed a P22 phage transduction assay to check the O-antigen integrality of our mutant strains. The transduction numbers of our constructed mutants were similar to their wild-type parent strains (Table 3), which indicated that our insertion mutations did not change the essential common trisaccharide backbone of O-antigen polysaccharides. Next, the LPS profile of each mutant strain was examined. It is obvious that ST-ddhc::prt A1 and SEddhc::abe mutants could exhibit a full length of LPS (Fig 3A), while ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB) or ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd, p15a-dadB-O9) showed a decreased synthesis of LPS length (Fig 3B). Surprisingly, ST-Δalr ΔdadB ΔrecF Δasd (pSC101-asd-O2, p15a-dadB-O9) exhibited a full length of LPS synthesis, indicating a higher level of CDP-Par synthesis might be the critical factor in achieving a triple O-serotype phenotype in a S. enterica strain. However, this assumption is not strictly demonstrated in this study. Meanwhile, we also conducted other phenotype evaluation assays (Table 3), such as the DOC and polymyxin B MICs, to roughly mimic the intestinal environment of the natural oral infection route and swimming abilities, an indicator of bacteria surface "wettability" [49]. After all these tests, we selected ST-ddhc::prt A1 , SE-ddhc::abe and ST-Δalr ΔdadB ΔrecF Δasd (pSC101asd-O2, p15a-dadB-O9) for further cross-immunogenicity evaluation. To guarantee their safety in the murine model, the global regulator cya and crp were deleted in these mutant strains [50], resulting in a series of potential vaccine candidates. The vaccine candidates showed good colonization in Peyer's patches, livers, and spleens ( Fig 5).
We are primarily interested in whether or not our live attenuated vaccine candidates could elicit effective cross-immune responses. To evaluate that, we applied ELISA assays to detect the antibodies raised against the O-antigen polysaccharides of wild-type S. Paratyphi A, S. Typhimurium and S. Enteritidis (Fig 6). Consistent with the bacterial agglutination assays, L008 (O2&O4) could induce a significantly higher anti-O2 antibody response than ST-Δcrp Δcya (O4). Meanwhile, L009 (O4&O9) could induce a significantly higher anti-O4 antibody response than SE-Δcrp Δcya (O9). Most importantly, L083 (O2&O4&O9) could simultaneously induce significantly higher anti-O2, anti-O4 and anti-O9 antibody responses in mice when compared to the negative control, SE-Δcrp Δcya (O9) and ST-Δcrp Δcya (O4), respectively. These data showed that our vaccine candidates could elicit effective cross-immune responses. Although a high level of anti-S. Typhimurium, anti-S. Enteritidis and anti-S. Paratyphi A LPS serum antibodies were obtained, we would like to know whether or not these raised antibodies are indeed functional. The in vitro bactericidal data indicated that serum complement-mediated S. Paratyphi A, S. Typhimurium and S. Enteritidis killing depended upon anti-O2, anti-O4 and anti-O9 antibodies (Fig 7). Moreover, we challenged all vaccinated mice with wild-type virulent S. Paratyphi A, S. Typhimurium or S. Enteritidis strain (Fig 8). Compared with the negative control, ST-Δcrp Δcya (O4) and SE-Δcrp Δcya (O9) vaccinated mice could receive good homologous protection but failed to heterologous challenge. However, mice vaccinated by double or triple O-epitopes vaccine strains received a significantly higher protection rate even when challenged with heterologous wild-type virulent strains, showing adequate cross-immunity protection.
However, there are some limitations to this study. The S. Typhimurium and S. Enteritidis used in SBA assay are all animal isolates, which could not simply be equal to the human blood isolates. Other reports have shown that anti-LPS antibodies stimulate low complement-dependent killing against S. Enteritidis human blood isolates [51]. Furthermore, S. Typhimurium ST313, which dominates in sub-Saharan Africa, is highly serum resistant and expresses distinct transcriptional patterns that may aid in escaping this killing mechanism [52]. It is a good future research direction to illustrate how they escape this complement-dependent killing mechanism.
In summary, we expressed double or triple immunodominant O-epitopes in a S. Typhimurium or S. Enteritidis background strain and proved they are effective in inducing cross-immunity and cross-protection. Significantly, the strategies we present in this study are not limited to S. enterica O-antigen polysaccharides but have applicability for generating cross-protection for many other important human and animal pathogens. Notable examples include Shigella flexneri and many pathogenic Escherichia coli. and S. Typhimurium (C) were determined by ELISA. A significantly higher level of IgG2a specific to the S. Paratypi A LPS compared to IgG1 was observed in L008 (O2&O4) and L083 (O2&O4&O9) ( ��� , P<0.001). A significantly higher level of IgG2a specific to the S. Enteritis LPS compared to IgG1 was observed in SE-Δcrp Δcya ( ��� , P<0.001). A significantly higher level of IgG2a specific to the S. Typhimurium LPS compared to IgG1 was observed in ST-Δcrp Δcya, L008 (O2&O4), L009 (O4&O9) and L083 (O2&O4&O9) ( ��� , P<0.001). The antibody concentrations were calculated using a standard curve. All of the measured sample concentrations were within the standard curve range. The error bars represent the standard deviation of the means. These data are representative of at least two independent experiments. (TIF)