|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: SB YW KR AKS DK AS NB DB. Performed the experiments: SB YW KR BC AKS AM AS DK NB. Analyzed the data: SB YW KR AKS DK AS NB DB. Wrote the paper: SB KR DB.
Invasive Salmonella infection is an important health problem that is worsening because of rising antimicrobial resistance and changing Salmonella serovar spectrum. Novel vaccines with broad serovar coverage are needed, but suitable protective antigens remain largely unknown. Here, we tested 37 broadly conserved Salmonella antigens in a mouse typhoid fever model, and identified antigen candidates that conferred partial protection against lethal disease. Antigen properties such as high in vivo abundance or immunodominance in convalescent individuals were not required for protectivity, but all promising antigen candidates were associated with the Salmonella surface. Surprisingly, this was not due to superior immunogenicity of surface antigens compared to internal antigens as had been suggested by previous studies and novel findings for CD4 T cell responses to model antigens. Confocal microscopy of infected tissues revealed that many live Salmonella resided alone in infected host macrophages with no damaged Salmonella releasing internal antigens in their vicinity. In the absence of accessible internal antigens, detection of these infected cells might require CD4 T cell recognition of Salmonella surface-associated antigens that could be processed and presented even from intact Salmonella. In conclusion, our findings might pave the way for development of an efficacious Salmonella vaccine with broad serovar coverage, and suggest a similar crucial role of surface antigens for immunity to both extracellular and intracellular pathogens.
Salmonella infections cause extensive morbidity and mortality worldwide. A vaccine that prevents systemic Salmonella infections is urgently needed but suitable antigens remain largely unknown. In this study we identified several antigen candidates that mediated protective immunity to Salmonella in a mouse typhoid fever model. Interestingly, all these antigens were associated with the Salmonella surface. This suggested that similar antigen properties might be relevant for CD4 T cell dependent immunity to intracellular pathogens like Salmonella, as for antibody-dependent immunity to extracellular pathogens. Detailed analysis revealed that Salmonella surface antigens were not generally more immunogenic compared to internal antigens. However, internal antigens were inaccessible for CD4 T cell recognition of a substantial number of infected host cells that contained exclusively live intact Salmonella. Together, these results might pave the way for development of an efficacious Salmonella vaccine, and provide a basis to facilitate antigen identification for Salmonella and possibly other intracellular pathogens.
Enteric fever caused by systemic Salmonella infection causes tremendous morbidity and mortality worldwide . Current control strategies become increasingly inefficient as a result of increasing antimicrobial resistance ,  and emergence of Salmonella serovars that are not covered by currently available safe vaccines , . This situation generates an urgent medical need for novel Salmonella vaccines with broad serovar coverage.
Early killed whole-cell vaccines containing mixtures of different serovars provide broad protection, but cause unacceptable adverse reactions . As an alternative to whole-cell vaccines, subunit vaccines containing a few defined Salmonella components could minimize adverse reactions. Indeed, vaccines containing the capsular polysaccharide Vi antigen provide moderate protection and excellent safety . On the other hand, serovars Paratyphi A and non-typhoidal Salmonella (NTS) that cause an increasing number of invasive salmonelloses , lack the Vi antigen and are therefore not covered by Vi vaccines . Apart from Vi, few Salmonella antigens have been identified, and all of these provide at best moderate levels of protection against challenge infection with virulent Salmonella strains in the commonly used mouse typhoid fever model. Moreover, antigens such as flagellin  and OmpD  are poorly conserved among relevant serovars.
For extracellular pathogens with antibody-mediated immunity, protective antigens must be surface-exposed , and this enables an effective strategy for priorization of antigen candidates . Humoral response to surface antigens can also contribute to immunity to intracellular pathogens such as invasive Salmonella . Indeed, Vi which induces protective antibody responses in human vaccinees, forms an extracellular capsule around Salmonella Typhi . Two additional antigens that confer partial immunity in the mouse typhoid fever model, flagellin  and SseB , are also part of Salmonella surface structures (flagella, translocon complex of a type III secretion system). Furthermore, outer membrane preparations (but not the outer membrane component lipopolysaccharide) have been suggested to mediate protective humoral immune responses against extracellular Salmonella bacteremia  and attenuated Salmonella strains in the mouse model , . A number of porins such as OmpC, OmpD, and OmpF are highly abundant in such outer membrane preparations suggesting that they might represent the actual protective antigens , , .
However, immunity to Salmonella critically depends also on CD4 T cells . Unfortunately, protective T cell antigens seem to be rare, and priorization of candidates is difficult since relevant antigen properties for CD4 T cell responses remain unclear , , . One key precondition for protective responses is expression of the respective Salmonella antigen during infection , and some data suggest that highly abundant antigens might be particularly well recognized by CD4 T cells , . Antigen in vivo expression can be deduced from various complementary approaches including screening of promoter trap libraries , , proteomics , serum antibody response –, as well as mutant virulence phenotypes.
In addition to antigen expression, antigen immunogenicity could play a major role. Antigen detection by cognate CD4 T cells requires antigen processing and presentation of the resulting small peptides by major histocompatibility (MHC) class II molecules. Peptide sequence properties that are characteristic for well recognized epitopes, can be used for genome-wide prediction of promising antigens . However, a large number of non-protective antigens contain putative high-score epitopes , ,  which could compromise the discriminatory power of this approach.
Experimental detection of immune responses to an antigen in convalescent individuals that have survived infection, demonstrates that this antigen was expressed in vivo and could be recognized by the immune system –. Indeed, this approach has been recently shown to facilitate identification of protective Chlamydia antigens . On the other hand, many immunodominant antigens in convalescent individuals lack protective efficacy, while a number of protective antigens may induce immune responses below the detection threshold during natural infection .
Another antigen property that can affect CD4 T cell responses is antigen localization. In particular, secreted or surface-associated antigens might induce particularly strong cellular immune responses because of superior processing, kinetic advantages compared to internal antigens, and/or physical association with pathogen-associated molecular patterns (PAMP) such as lipopolysaccharide that provide potent stimuli for innate and adaptive immunity , –. Indeed, secretion/surface localization has been widely used to prioritize candidates for antigen identification. However, antigens with likely internal localization can also induce specific CD4 T cell responses that mediate protection against various intracellular pathogens , .
Taken together, relevant antigen properties for CD4 T cell mediated immunity to intracellular pathogens remain poorly characterized, and this impairs antigen priorization for vaccine development. To address this issue, we compared here 37 diverse Salmonella antigens in a mouse model that closely mimics human typhoid fever . The results suggested that recognition of surface-associated antigens might be necessary to detect and combat live intracellular Salmonella, whereas recognition of internal antigens would mediate futile non-protective attack of already dead Salmonella. In conclusion, we propose a similar crucial role of surface-associated antigens for immunity to both extracellular and intracellular pathogens.
To determine immune responses to Salmonella antigens, we selected 21 broadly conserved Salmonella proteins. We selected several subunits of the SPI-2 type III secretion system since the putative translocon subunit SseB of this system showed promising protectivity in previous studies , . We also included several porins since a previous study had shown that OmpD conferred protection against an attenuated Salmonella mutant . To explore the role of antigen localization we selected additional proteins localized in Salmonella cytosol, inner membrane, periplasm, and outer membrane/surface. To explore the role of antigen abundance, we determined absolute quantities of more than 1100 Salmonella in infected mouse spleen. Specifically, we purified Salmonella from infected mouse spleen using flow cytometry as described . We determined absolute protein quantities in these ex vivo purified Salmonella using shot-gun proteomics with 30 isotope labeled reference peptides and the iBAQ quantification method  (for detailed description see Materials and Methods section). From these data, we selected additional antigens with a large range of abundances (Tab. 1).
To determine potential cross-protection between different serovars, we cloned the corresponding genes from Salmonella enterica serovar Typhi (except for OmpD which was obtained from serovar Typhimurium since it is absent in serovar Typhi). We expressed the proteins as C-terminal His6-fusions in E. coli followed by Ni-affinity chromatography purification. We purified the control antigen GFP-His6 using the same protocol.
We determined immune responses to these antigens in genetically resistant, convalescent mice that had survived infection with virulent Salmonella enterica serovar Typhimurium. We detected antigen-specific CD4 T cells in spleen using a sensitive CD154 assay  and measured serum IgG antibody responses using ELISA. All tested antigens were recognized by CD4 T cells (Fig. 1A; Tab. 1), many of which secreted IFNγ or IL-17 upon stimulation. Both cytokines play crucial roles in immunity to Salmonella . Frequencies of responsive CD4 T cells were in the same range as for flagellin, which has been considered an immunodominant antigen . These data suggested that Salmonella infection elicited a broad cellular immune response against a large number of in vivo expressed antigens from all Salmonella compartments in agreement with data observed for S. Typhi infected human patients . There was no correlation between in vivo antigen abundance as determined by proteome analysis of ex vivo purified Salmonella, and CD4 T cell frequency or cytokine profile (Tab. 1).
Serum antibody responses revealed similar broad recognition of antigens from several Salmonella compartments (Fig. 1B) in agreement with previous data for human typhoid fever patients –, . Interestingly, the three immunodominant humoral antigens T2461, PhoN, and PcgL were all highly expressed in vivo (Tab. 1) suggesting a potential impact of antigen dose on antibody responses to Salmonella, although responses to minor antigens did not correlate with antigen abundance. PhoN has been previously recognized as an immunodominant antigen .
Many of the tested Salmonella antigens were capable to induce cellular and humoral immune responses. To test if these responses could confer protective immunity, we tested the 21 recombinant Salmonella antigens in immunization/challenge infection experiments in genetically susceptible BALB/c mice. Based on the results, we selected 16 additional Salmonella antigens primarily from the outer membrane, and tested them using the same experimental immunization/challenge approach (however, we did not measure their immunogenicity in convalescent mice). For simplicity, we discuss results for both antigen sets together. Out of 37 tested antigens, only few antigens enabled prolonged survival after oral challenge infection with virulent Salmonella compared to control immunization with the unrelated antigen GFP (Fig. 2; Tab. 1; poor survival of PhoN-vaccinated animals confirmed recently published data ). In fact, only two antigens (T0937 and T2672) mediated protective immune responses with P-values below 0.05 in our small experimental groups of only five mice per antigen. Replicate experiments with larger group sizes might yield statistical significant results for additional candidates such as SseB that has already been shown to be protective in two independent previous studies. Such experiments will be required to select individual antigens for vaccine development in future studies. On the other hand, the primary focus of this study was to identify antigen properties that correlate with protectivity. For this purpose, the somewhat noisy survival times detected with small animal groups were still helpful.
As an example, survival times did not correlate with CD4 T cell responses (Fig. 3A) or serum antibody levels (Fig. 3B) during natural infection of resistant mice. This could partially reflect differences in MHC class II haplotypes (H2d in BALB/c vs. H2b in 129/Sv), courses of infection, and potential differences in Salmonella biology in susceptible vs. resistant mice. However, a recent large-scale study reports comprehensive immunogenicity data for BALB/c mice and other mouse strains that had been immunized with attenuated Salmonella, as well as for human patients . Several antigens that prolonged survival of immunized BALB/c mice after Salmonella challenge infection in our experiments, elicit detectable antibody responses in various mouse strains including BALB/c. However, none of these antigens was found to be immunodominant  and antibodies to antigens IroN and CirA with the longest survival times were not detected in this and previous studies. This could reflect differential antigen expression in virulent vs. attenuated Salmonella, different routes of administration, and/or differential expression at various stages of disease progression. Together, these data provide no evidence for immunodominance in convalescent or immune individuals as a prerequisite for protectivity.
Interestingly, in vivo expression levels also did not correlate with survival times (Fig. 4A). In fact, the two antigens that enabled the longest survival, IroN and CirA, had in vivo expression levels that were below our detection threshold. By comparison, antigens T2461 and PhoN were highly expressed in vivo and induced potent CD4 T cell and humoral responses in convalescent individuals, yet failed to prolong survival (in agreement with previous observations ).
In contrast to immunogenicity and in vivo abundance, antigen localization seemed to be crucial (Fig. 4B). In fact, antigens enabling prolonged survival times were exclusively associated with the Salmonella surface, either as experimentally validated outer membrane-associated lipoproteins , as outer membrane proteins, or as the translocon complex of the type III secretion system encoded by Salmonella pathogenicity island two (SPI-2) (Tab. 1). These data suggested distinct immune responses to Salmonella outer membrane/surface antigens that fundamentally differ from those to internal antigens.
On the other hand, surface localization alone was not sufficient for protectivity. As examples, membrane proteins PgtE, PagC, and Tsx were highly expressed in vivo and PgtE and PagC elicited potent CD4 T cell responses in convalescent individuals (Fig. 1A). PagC is also well recognized by antibodies and CD4 T cells of human typhoid fever patients . However, PagC, PgtE, and Tsx failed to prolong survival. Interestingly, structural models revealed that these proteins were largely buried in the outer membrane bilayer (Fig. 5), and their extracellular loops contained at most one predicted CD4 T cell epitope each, and only up to two linear antibody epitopes, respectively. Importantly, key amino acids in exposed T cell epitopes differed among Salmonella serovars which might have impaired cross-protectivity of serovar Typhi antigens against serovar Typhimurium challenge infection. Similar observations were also made for non-protective TolC, OmpC, OmpD, and OmpF. By contrast, antigens IroN, CirA, and FepA that enabled extended survival after challenge infection, had extracellular loops with several highly conserved T and B cell epitopes (Fig. 5). Further studies with larger data sets will be required to validate the relevance of these structural properties for protectivity.
The strong bias for surface-associated Salmonella antigens might have been expected based on previous data for model antigens suggesting superior immunogenicity of surface antigens compared to internal antigens , –. However, these model antigen data were in striking contrast to results from us and others demonstrating comparable immune responses to autologous Salmonella antigens from all Salmonella compartments (Fig. 1A,B). Furthermore, there was no obvious correlation between immunogenicity and survival times (Fig. 3A,B).
To better understand these discrepancies between model antigens and autologous Salmonella antigens, we re-visited the impact of antigen localization using a well-characterized, sensitive model system in which a MHC II-restricted T cell epitope from ovalbumin comprising amino acids 319 to 343 (OVA) is recognized by adoptively transferred cognate T cell receptor transgenic CD4 T cells , . We targeted the OVA epitope to different Salmonella compartments by fusing it to various proteins with known localization: GFP_OVA (cytosol ), OVA_MglB (periplasm ), Lpp_OVA (inner leaflet of the outer membrane ), and OVA_AIDA (outer leaflet of the outer membrane ) (Fig. 6A). To modulate expression levels, we used ribosome binding sites with differential translation initiation efficiency . We expressed these fusion proteins from an in vivo inducible promoter  in an attenuated Salmonella enterica serovar Typhimurium aroA strain . Antigen expression and localization was validated in in vitro cultures using cell fractionation followed by western blotting, trypsin treatment, and antibody binding (Fig. S2). Interestingly, small fractions of both outer membrane antigens LPP_OVA and partially processed OVA_AIDA were released to the extracellular surroundings when expressed at high levels (Fig. S2C) in agreement with previous findings for similar proteins –.
We infected BALB/c mice with Salmonella strains by intragastric gavage of 1010 CFU. All Salmonella strains colonized intestinal Peyer's patches with peak tissue loads of 3×104 to 1.5×105 CFU at day seven post infection as observed before for attenuated Salmonella aroA . All constructs stably maintained their respective ovalbumin-expression plasmids (>80% at 7 days post infection). To determine antigen-specific CD4 T cell induction, we adoptively transferred OVA-specific TCR-transgenic CD4 T cells one day prior to Salmonella infection. OVA-specific T cells upregulated the early activation marker CD69 and formed blasts in mice infected with Salmonella expressing ovalbumin model antigens, but not in mice infected with control Salmonella (Fig. 6B) as observed previously . CD4 T cell induction kinetics were similar for all constructs and consistent with our previous observations  suggesting a response to Salmonella in situ antigen expression, but not to the inoculum , .
To compare T cell responses against the various Salmonella constructs, we measured T cell blast formation at peak Salmonella colonization at day seven post infection. Salmonella tissue loads varied somewhat between individual mice but for each construct, there was a linear relationship between the number of ovalbumin-specific DO11.10 blasts and Salmonella loads (Fig. 6C) in agreement with our earlier observations . To determine the specific immunogenicity of each Salmonella strain, we calculated the average ratio of DO11.10 CD4 T cell blasts per viable Salmonella (i.e., the slopes in Fig. 6C) . The data revealed comparable immunogenicity of model antigens GFP_OVA and OVA_MglB (Fig. 6D). In contrast, high-level expression of surface-associated LPP_OVA and OVA_AIDA induced superior responses that clearly surpassed responses even to saturating amounts  of internal GFP_OVA.
The OVA_AIDA fusion protein contained a fragment of the virulence factor AIDA from enteropathogenic E. coli and a cystein-deficient variant of the cholera toxin B subunit from Vibrio cholerae . Both components might have stimulatory effects ,  that could potentiate ovalbumin immunogenicity. To test this potentially confounding factor, we compared Salmonella expressing a suboptimal level of cytosolic GFP_OVA  (some 54.000 copies per Salmonella cell) to Salmonella expressing the same amount of GFP_OVA together with AIDA and cholera toxin B. Both strains induced DO11.10 T cell blasts with similar efficacy (Fig. 6D) suggesting that AIDA and cholera toxin B expression had no impact on the immunogenicity of Salmonella-encoded OVA.
Taken together, these findings suggested that antigens from all Salmonella compartments could induce specific CD4 T cell responses, but highly expressed outer membrane-associated antigens were clearly superior in agreement with previous observations in other model systems. However, these data were in striking contrast to responses to autologous Salmonella antigens (see discussion).
The fundamentally superior protectivity of surface-associated Salmonella antigens might reflect their unique accessibility to antigen processing and presentation in infected host cells in contrast to internal Salmonella antigens that are shielded by the Salmonella envelope, and thus remain invisible for the host immune system until Salmonella is damaged and the bacterial cell breaks open. To detect intact and damaged Salmonella in infected tissues, we used cytosolic GFP as a marker for internal antigens.
Salmonella expressing GFP from the chromosomal in vivo induced locus sifB were readily detected in infected tissue homogenates using flow cytometry  (Fig. 7A). Flow cytometric counts for GFP+ Salmonella closely correlated with viable counts as determined by plating (Fig. 7A, inset) suggesting that detectable GFP levels were present in all live Salmonella.
Confocal microscopy of infected spleen and liver sections revealed many particles that were stained by a polyclonal antibody to Salmonella lipopolysaccharide, had typical Salmonella size and morphology, and contained GFP (Fig. 7B) as previously observed  suggesting that these particles represented live intact Salmonella. In addition, we also detected numerous lipopolysaccharide-positive particles with distorted shapes that lacked detectable GFP (Fig. 7C), and likely represented killed and partially degraded Salmonella. Such particles were absent in non-infected control sections. Some Salmonella killing during acute infections had previously been proposed –. We observed some infected cells containing both intact and damaged Salmonella, but a large number of live Salmonella resided alone (or together with other live Salmonella) in infected cells with no detectable dead Salmonella. In such infected cells, internal Salmonella antigens were thus shielded and inaccessible for immune recognition.
There is an urgent medical need for an efficacious Salmonella vaccine with broad coverage of invasive serovars. One important bottleneck in the development of such a vaccine is the identification of suitable protective antigens. In this study, we identified broadly conserved S. Typhi antigen candidates that prolonged survival after S. Typhimurium challenge infection in the mouse typhoid fever model. The protectivity of some of these candidates should be confirmed with larger experimental groups to select the best antigen candidates for vaccine development in future studies.
Two siderophore receptors (IroN, CirA) enabled the longest survival (Tab. 1) consistent with previous studies that revealed siderophore receptors including IroN as promising vaccine antigens in other models –. Interestingly, siderophore receptors are induced by iron starvation and/or activation of the PhoPQ two component sensory system . IroN and CirA induction could thus contribute to increased protective efficacy of membrane preparations from iron-starved Salmonella , or live attenuated Salmonella phoQ24 with constitutive hyperactivation of the PhoP response regulator .
On the other hand, all identified antigens still provided at most partial protection against challenge infection with virulent Salmonella suggesting a need for additional antigens. Unfortunately, protective Salmonella antigens might be rather rare as even among the 37 tested in vivo expressed antigens that were all highly immunogenic during infection, only a small minority enabled prolonged survival. OmpC, OmpD, and OmpF were previously proposed as potential protective antigens based on data obtained for enriched Salmonella membrane preparations. However, all three antigens failed to protect in our model. This could reflect higher stringency of our model (challenge infection with virulent Salmonella vs. highly attenuated mutant Salmonella), denatured three-dimensional structures of our recombinant antigen preparations vs. native antigens, and/or presence of undetected minor protective antigens (such as IroN and CirA) besides OmpC, OmpD, and OmpF in the previously used outer membrane antigen preparations.
Additional protective Salmonella antigens could be identified by comprehensive immunization/challenge experiments, but this would require extensive animal experimentation. Antigen priorization using relevant antigen properties could help to narrow down the number of antigen candidates to more practical numbers. Unfortunately, some previously proposed antigen properties seemed to have limited relevance for protectivity in our model. This included Salmonella in vivo expression levels, sequence-based antigenicity predictions, and immunodominance in convalescent individuals. Poor correlation of antigen immunodominance with protective efficacy has also been observed in tuberculosis . On the other hand, immune recognition in convalescent individuals can still provide valuable information about antigen expression during at least some stage of infection that might be difficult to obtain otherwise , . Such data thus could greatly help to prioritize antigen candidates .
In contrast to antigen abundance and immunodominance, surface-association appeared to be an essential prerequisite. Surprisingly, some surface-associated proteins that enabled prolonged survival also included lipoproteins which were likely to reside in the inner leaflet of the outer membrane facing the internal periplasmic space with no exposure to the outside. It is possible that some lipoproteins might flip across the outer membrane as observed for other Gram-negative bacteria . Moreover, some lipoprotein fraction is constantly released to the outside through outer membrane vesicle shedding , .
Several mechanisms could contribute to the striking superiority of surface-associated antigens. Antibody responses are important for full protection against virulent Salmonella , and protective antibody responses must be directed against surface antigens . On the other hand, CD4 T cells are even more important for immunity to Salmonella at least in the mouse typhoid fever model , and it is unclear why CD4 T cells should respond to surface-associated antigens in a fundamentally different way compared to the much larger number of internal antigens.
In fact, early cell culture experiments suggested no impact of Salmonella antigen localization on CD4 T cell recognition of infected cells . However, in this study a large amount of antigen was already present in the inoculum, and rapid killing of the majority of phagocytosed Salmonella  would have released this antigen from all Salmonella compartments. Several subsequent in vivo studies suggested that surface-associated model antigens might have intrinsically higher immunogenicity compared to internal model antigens , –. However, the various model antigen targeting constructs could have differed in antigen in vivo expression levels, antigen stability, and epitope processing. Fusion partners could also have direct immunomodulatory effects. We therefore re-visited this issue and tried to control some of these factors. Our results clearly supported the previous finding of superior immunogenicity of highly expressed surface-associated model antigens in Salmonella.
In surprising contrast to these data from model antigens, however, humoral and cellular immune responses in Salmonella-infected convalescent mice did not show any bias for surface-associated autologous Salmonella antigens in this as well as in a recent large-scale study . Broad recognition of antigens from all pathogens compartments has also been observed in Salmonella Typhi-infected or Chlamydia-infected human patients , , , , . Model antigens and autologous antigens were also discordant with respect to the impact of antigen abundance. Specifically, our data for ovalbumin model antigens in this and a previous study , as well as similar findings for Mycobacterium bovis BCG overexpressing Ag85b , suggested that high in vivo expression levels enhance antigen immunogenicity. However, for autologous Salmonella antigens in vivo expression levels did not correlate with protectivity. Striking discrepancies between results for model antigens vs. autologous antigens have also been observed in other pathogens .
Some of the discrepancies could reflect technical issues. In particular, strong expression of foreign surface model antigens might induce subtle alteration in Salmonella in vivo properties such as increasing outer membrane vesicle shedding or alterations in protein secretion that could affect antigen presentation and immune recognition. Furthermore, model antigens might not be representative of autologous antigens that may have been shaped by host/pathogen co-evolution selecting for weak immunogenicity. Regardless of the actual causes of these discrepancies, our data indicated that in contrast to evidence from model antigens, protective Salmonella surface-associated antigens were not more immunogenic compared to internal antigens.
As an alternative explanation, surface-associated antigens might become more rapidly available for immune recognition compared to internal antigens that are only released after some pathogen damage. This could be relevant since early immune responses might facilitate infection control . In the mouse typhoid fever model, however, a detectable fraction of Salmonella is rapidly killed early during infection as observed in this and previous studies ,  similar to events during Mycobacterium infection . Consistent with these observations, CD4 T cell induction kinetics in the ovalbumin model system were similar for Salmonella strains with internal or surface-associated OVA-expression.
Instead, we propose an alternative explanation based on the observation that many live Salmonella resided alone, or together with other live Salmonella, in infected host cells with no dead Salmonella releasing their internal antigens. As a consequence, Salmonella internal antigens remained inaccessible for antigen processing and presentation in these cells. In contrast, surface-exposed Salmonella antigens, or antigens released by outer membrane vesicle shedding, could be accessible for processing and presentation to cognate CD4 T cells for initiation of protective anti-Salmonella effector mechanisms (Fig. 8). In comparison, CD4 T cells recognizing internal Salmonella antigens would have limited impact on infection control because they miss many cells containing live Salmonella and instead direct their responses to host cells containing already dead Salmonella. According to this model, surface-associated antigens thus differ fundamentally from internal antigens because they are uniquely accessible in host cells containing only live Salmonella.
Surface-associated/secreted antigens have been shown to be crucial for CD8 T cell-dependent immunity to Listeria infection , . Our data suggested that such antigens might also be crucial for CD4 T cell mediated immunity to Salmonella and potentially other intracellular pathogens. Interestingly, some internal antigens have been shown to confer partial protection in infectious diseases caused by intracellular pathogens such as Leishmania  and Mycobacterium . In these infections live and dead pathogens often co-occur in the same host microenvironments ,  suggesting that both internal and surface-associated antigens might be available for T cell recognition and initiation of antimicrobial immune effector mechanisms targeting both live and already dead pathogens . We speculate that full protection might still require immune detection of all live pathogens including those that reside in microenvironments with yet no accessible internal antigens from dead pathogens. Further studies are required to test this hypothesis.
This study suggested novel Salmonella antigens that conferred partial protection against virulent Salmonella in a stringent typhoid fever model. High sequence conservation among relevant Salmonella serovars and cross-protection of serovar Typhi antigens against serovar Typhimurium challenge infection, suggested that some of these antigens might help to pave the way for a broadly protective vaccine against systemic Salmonella infection. In addition, our findings suggested that surface-associated antigens might represent particular promising antigens for both humoral and cellular immunity to Salmonella, since recognition of surface antigens uniquely enables detection and destruction of live Salmonella in relevant host microenvironments. This crucial importance of antigen localization could facilitate discovery of additional protective antigens for Salmonella and potentially other intracellular pathogens.
All animal experiments were approved (license 2239, Kantonales Veterinäramt Basel-Stadt) and performed according to local guidelines (TschV, Basel) and the Swiss animal protection law (TschG).
Antigens were PCR-amplified from Salmonella enterica serovar Typhi Ty2 (or Salmonella enterica serovar Typhimurium SL1344  for ompD), cloned as His6-fusions by conventional ligation into pET22b, or by Enzyme Free Cloning into plasmid pLICHIS , and overexpressed in E. coli BL21. GFP_His6 was cloned as control antigen. Antigens were purified from washed inclusion bodies using immobilized metal ion affinity chromatography (Protino Ni TED 1000, Macherey Nagel) followed by ion exchange chromatography (Ion exchange spin columns, Pierce Thermo Scientific, cationic or anionic resins depending on antigen isoelectric point).
Salmonella expressing the green fluorescent protein (GFP) were sorted infected using flow cytometry as described . Preparation of tryptic peptides and analysis by LC-MS/MS was done essentially as described  with some modifications. Given the limited sample material Protein LoBind tubes and pipette tips (Axygen) were used throughout the procedure. Frozen FACS sorted Salmonella pellets were resuspended in 15 µl lysis buffer (100 mM ammonium bicarbonate, 8 M urea, 0.1% RapiGest) and sonicated for 2×30 seconds. The released proteins were reduced and alkylated, and first digested for 4 hrs with sequencing grade LysC peptidase (10 ng/µl; Promega) before overnight trypsin digestion. The detergent was cleaved by adding 2M HCL and 5% TFA to final concentrations of 50 mM and 0.5% respectively, and incubating for 45 min at 37°C. Prior to centrifugation to remove the cleaved detergent (14,000×g, 10 min, 4°C) a mixture containing 32 heavy labeled reference peptides were added to the samples (5*10−5 fmoles per Salmonella for expected “high” abundance proteins, 5*10−6 fmoles per Salmonella for expected “low” abundance proteins; Tab. S1). The recovered peptides were desalted on C18 reverse-phase spin columns (Macrospin columns, Harvard apparatus), dried under vacuum and subjected to LC-MS/MS using an LTQ-Orbitrap-Velos instrument (Thermo-Fischer Scientific). The amount of material analyzed in a single shot in the MS depended on the infection load, and corresponded to peptides derived from between 5*105 and 2*106 sorted Salmonella, plus contaminating mouse material which escaped detection in the cell sorter . We analyzed samples from seven independently infected mice. In order to increase the number of Salmonella protein identifications, MS-sequencing was focused on previously identified peptides from Salmonella using the recently developed inclusion list driven workflow . Each sample was analyzed twice in succession in the MS to verify technical reproducibility. Peptides and proteins were database searched against a decoy database consisting of the SL1344 genome sequence (ftp://ftp.sanger.ac.uk/pub/pathogens/Salmonella/), GFP_OVA, 204 frequently observed contaminants, all mouse entries from SwissProt (Version 57.12), and all sequences in reversed order (total 42502 entries) using the Mascot search algorithm. The search criteria were set as follows: full tryptic specificity was required (cleavage after lysine or arginine residues); 2 missed cleavages were allowed; carbamidomethylation (C) was set as fixed modification; oxidation (M) as variable modification. The mass tolerance was set to 10 ppm for precursor ions and 0.5 Da for fragment ions. The false discovery rate was set to 1% for protein and peptide identifications. In addition to Salmonella proteins a substantial number of mouse proteins were identified in the samples as previously noted . Absolute quantities were determined for those 18–20 “anchor” proteins that were detected along with a corresponding labeled AQUA peptide using the Trans-Proteomic Pipeline (TPP,V4.4.0). We then used the iBAQ method  to establish absolute quantities of all remaining protein identifications, with a linear model error of between 47 and 60%.
Translational fusions of the ovalbumin peptide containing amino acids 319 to 343 to various proteins with differential targeting in the Salmonella cell were constructed by PCR cloning. All fusion genes were cloned into a pBR322-derived plasmid backbone  downstream of a Salmonella genome fragment containing the in vivo inducible pagC promoter  and ribosomal binding site 1 (AAGAA) or 2 (AGCAG) for low or high translation initiation efficiencies . To generate ova_aida, coding sequence for the ovalbumin peptide (ova) was inserted between the signal peptide derived from cholera toxin B and the HA tag in plasmid pLAT260 . A control plasmid coding for CTB_AIDA and GFP_OVA was also constructed. To generate lpp_ova, lpp without the C-terminal lysine codon that can cross-link to peptidoglycan , was amplified from E. coli DH5α and fused with ova and a C-terminal HA tag. To generate ova_mglB, mglB gene without the signal peptide sequence was amplified from E. coli DH5α and fused with a ctB signal sequence followed by ova and the HA tag. The construction of gfp_ova has been described . The various plasmids were transformed into attenuated Salmonella enterica serovar Typhimurium aroA SL3261 .
Ovalbumin expression was assessed by western blotting with a polyclonal antibody to ovalbumin (Sigma) that recognizes the OVA peptide comprising amino acids 319 to 343 . Salmonella outer membranes were prepared by extraction with L-lauryl sarcosinate as described . Periplasm was prepared by chloroform extraction as described . Culture supernatants were sterile filtered (0,2 µm pore size) and subjected to TCA precipitation . To assess ovalbumin surface accessibility, intact or lysed Salmonella cells were treated with 50 µg ml−1 trypsin at 37°C for 10 min. In addition, Salmonella were stained with an antibody to the HA tag, and examined by fluorescence microscopy.
Female 8 to 12 weeks old 129/Sv mice were obtained from Charles River. Mice were orally infected with 109 CFU Salmonella enterica serovar Typhimurium SL1344  from late log cultures using a round-tip stainless steel needle. Control mice were sham-infected. Mice were sacrificed 6 months after infection. Splenocytes were isolated and tested for antigen-specific CD4 T cell responses as described . Unstimulated T cells from convalescent mice as well as antigen-stimulated T cells from naïve control mice showed only weak background responses (Fig. S1). Some antigens gave also weak responses for T cells from convalescent mice (depending on the individual mouse). Together, these data suggested that antigen-nonspecific background responses to E. coli contaminants that might have been present in trace amounts in our antigen preparations did not result in unspecific T cell responses in our assay. Plasma was tested for antigen-specific IgG responses using ELISA with an IgG calibration curve for absolute quantification.
Female, 8 to 12 weeks old BALB/c mice were obtained from Charles River. Groups of 5 mice were immunized subcutaneously with 10 µg antigen emulsified in complete Freund's adjuvant followed by a second immunization with incomplete Freund's adjuvant four weeks later. After additional four weeks, mice were orally infected with 6×105 CFU Salmonella enterica serovar Typhimurium SL1344  from late log cultures using a round-tip stainless steel needle. Infected BALB/c were monitored twice daily and sacrificed when moribund.
BALB/c and DO11.10 mice  were bred in the Bundesamt für gesundheitlichen Verbraucherschutz und Veterinärmedizin (Berlin, Germany) under specific-pathogen free conditions. Adoptive transfer of 4×106 DO11.10 T cells into syngenic age- and sex-matched BALB/c mice was performed one day before infection as described . For infection, attenuated Salmonella strains carrying expression cassettes for various ovalbumin fusion proteins were grown to late log phase and harvested. Bacteria were washed twice and resuspended in LB containing 3% sodium bicarbonate. Doses containing ca. 1010 cfu in 100 µl were administered intragastrically to chimeric mice with a round-tip stainless steel needle. At various time points post infection, mice were anesthetized and sacrificed. DO11.10 T cell blast formation was determined by flow cytometry as described . Aliquots of the same Peyer's patch preparations were treated with 0.1% Triton x-100 to release intracellular Salmonella for CFU determination by plating, and for quantitation of GFP_OVA in vivo expression levels by two-color flow cytometry as described . Many TCR tg models show substantial clonal expansion upon antigen stimulation. However, in our Salmonella model we observe only weak and variable accumulation of tg CD4 T cells in infected tissues which might reflect the fact that even at peak Salmonella loads only about 1 ng antigen is present . Instead, blastogenesis as measured by CD69 upregulation and increased forward scatter provides a sensitive antigen-specific readout.
BALB/c mice with Salmonella loads of 106 to 107 in spleen and liver were sacrificed. 10 µm cryosections were stained with polyclonal rabbit antibodies to Salmonella lipopolysaccharide (SIFIN) and anti-CD68 (abcam) followed by Alexa 546-conjugated goat anti-rabbit and Alexa 647-conjugated goat anti-rat antibodies (Invitrogen). Sections were examined by confocal microscopy (Leica, SP5).
Structural models for selected Salmonella outer membrane antigens based on solved structures of homologues were obtained from SWISS-MODEL  available at http://swissmodel.expasy.org. Linear B-cell epitopes were predicted using FBCPred  available at http://ailab.cs.iastate.edu/bcpreds/predict.html using an epitope length of 14 and 90% specificity. Peptides that bind to MHC II I-Ad and/or I-Ed were predicted using RANKPEP  available at http://imed.med.ucm.es/Tools/rankpep.html with a binding threshold yielding 85% sensitivity for detection of well-defined epitopes in MHCII haplotype databases (the default setting of RANKPEP).
Representative antigen-specific CD4 T cell responses in convalescent and control mice. Cells were stimulated with various antigens (shown are examples for T2461 and T0937). CD4, CD154-double positive cells representing responding cells were gated (upper panels) and analyzed for IFNγ and IL-17 (lower panels). Background responses in control mice were subtracted form responses in convalescent mice and reported in Tab. 1.
Expression and localization of ovalbumin epitope fusion proteins in Salmonella. A) Anti-ovalbumin immunoblot of total Salmonella cell lysates (3×107 cfu) of strains expressing either low (“lo”) or high (“hi”) levels of ovalbumin fused to different proteins. Expected molecular weights were: GFP_OVA, 30 kDa; OVA_MglB, 38 kDa; Lpp_OVA, 11 kDa; OVA_AIDA, 67 kDa. B) Localization of various fusion proteins. OVA_MglB was detected in isolated periplasm fractions (Ppl.) in similar quantities as in whole cell lysates (Lys.). Lpp_OVA was detected in isolated outer membrane fractions. It was unaccessible for trypsin degradation in intact Salmonella but readily digestible in isolated membrane fractions. Immunostaining of intact Salmonella with a fluorescent antibody showed no detectable signal. OVA_AIDA was detected in isolated outer membranes and accessible to trypsin digestion even in intact Salmonella suggesting surface localization. This was confirmed by immunostaining. C) Immunoblot of culture supernatants of 4.5×1011 CFU (TCA precipitation). Endogenous Salmonella proteins with apparent molecular weights of ca. 23 and 67 kDa, respectively, cross-react with the anti-ovalbumin polyclonal antibody (empty arrowheads). These bands were also detected in non-recombinant Salmonella. In addition, an OVA-containing protein of around 11 kDa was released from Lpp_OVA expressing Salmonella (black arrowhead), whereas a 30 kDa fragment was released from Salmonella expressing high amounts of OVA_AIDA.
List of isotope labeled peptides used for protein quantification.
We thank Claus Lattemann, Thomas Meyer, and Reinhold Förster for helpful discussions and generous support.
DB and co-workers are funded by the Swiss National Science Foundation (31003A-121834), SystemsX (RTD project BattleX) and Deutsche Forschungsgemeinschaft (SFB641-A9, SPP1316 Bu971/6). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.