Comparison of Tissue-Selective Proinflammatory Gene Induction in Mice Infected with Wild-Type, DNA Adenine Methylase-Deficient, and Flagellin-Deficient Salmonella enterica

Abstract

Mutants of Salmonella enterica serovar Typhimurium deficient in DNA adenine methylase (Dam) are attenuated for virulence in mice and confer heightened immunity in vaccinated animals. In contrast, infection of mice with wild-type (WT) strains or flagellin-deficient mutants of Salmonella causes typhoid fever. Here we examined the bacterial load and spatiotemporal kinetics of expression of several classes of host genes in Peyer's patches, the liver, and the spleen following oral infection of mice with WT, dam mutant, or flagellin-deficient (flhC) Salmonella. The genes evaluated included inflammatory (interleukin-1β [IL-1β], tumor necrosis factor alpha), chemokine (macrophage inflammatory protein 2), Th1/Th2 indicator (IL-12p40, IL-4), and interferon system (beta interferon [IFN-β], IFN-γ, protein Mx1 GTPase, RNA-dependent protein kinase, inducible nitric oxide synthase, suppressor of cytokine signaling 1) beacons. We showed that maximal interferon system and proinflammatory gene induction occurred by 5 days after infection and that the levels were comparable for the WT and flhC strains but were significantly lower for the dam mutant. Additionally, host gene expression in systemic tissues of individual animals was dependent on the bacterial load in the Peyer's patches for mice infected with WT, dam mutant, or flhC mutant Salmonella as early as 8 h after infection. Moreover, a bacterial load threshold in the Peyer's patches was necessary to stimulate the host gene induction in the liver and spleen. Taken together, these results suggest that bacterial load and the accompanying strain-specific cytokine signature are important determinants of the host innate immune response and associated disease manifestations observed in dam mutant Salmonella-infected animals compared to the immune response and disease manifestations observed in WT and flhC mutant Salmonella-infected animals.

Salmonella infection is a significant problem globally, and it is estimated that there are 1.4 million cases annually in the United States alone (www.cdc.gov). Salmonella strains are intracellular pathogens that initiate infection by migration through the intestinal epithelia, followed by invasion and subsequent multiplication within host epithelial and phagocytic cells (6). Infection with Salmonella enterica serovar Typhimurium generally results in nonlethal gastroenteritis in humans but causes lethal disseminated bacteremia and typhoid fever in mice that effectively model human S. enterica serovar Typhi infection. The bacterial gene encoding DNA adenine methylase (Dam) modulates bacterial gene expression through N-6 methylation of adenosine residues in GATC sites (21). Although not required for Salmonella viability, Dam affects a broad range of processes, including DNA replication and repair, chromosome segregation, insertion element transposition, and surface protein expression (15, 20, 32).

Salmonella mutants lacking functional Dam are highly attenuated for virulence in the mouse model and confer immunity against subsequent wild-type (WT) Salmonella infection (19, 20). Oral infection of mice with dam mutant Salmonella is characterized by impaired colonization of systemic tissues, such as the liver and spleen, dampened induction of interferon (IFN)-stimulated gene (ISG) expression, and eventual clearance of the infection (19, 20, 37). Little is known about the mechanistic basis of the differences in virulence between dam and WT Salmonella strains observed in mice. Among the possible contributing factors that lead to the altered virulence of dam mutant Salmonella that has been observed are reduced multitissue cytokine responses (37), decreased invasion of intestinal epithelial cells, and reduced M-cell cytotoxicity (15).

The host responses to pathogen infection include recognition of conserved pathogen-associated molecular patterns (PAMPs) by members of the Toll-like receptor (TLR) family. Recognition of bacterial PAMPs activates signaling pathways leading to the induction of proinflammatory cytokines and chemokines, as well as their downstream target genes (23). The host genes induced following Salmonella infection include genes encoding prototypical inflammatory cytokines, such as tumor necrosis factor alpha (TNF-α) and IFN-γ (38, 50); chemokines, such as macrophage inflammatory protein 2 (MIP2) (8); type I IFN, such as IFN-β; and the targets of IFN and cytokine signaling, such as the inducible nitric oxide synthase (iNOS) that is necessary for resistance to Salmonella infection (1, 37, 48). Deficiencies in TNF-α, interleukin-12 (IL-12), or IFN-γ also have been shown to increase the severity of Salmonella infection in mice lacking these factors (26, 50).

Our previous work demonstrated that there was greater induction of targets of both type I IFN-α/β- and type II IFN-γ-stimulated genes in mice infected with the WT than in mice infected with Salmonella dam mutant strains at a late time after infection, 6 days (37). By contrast, no significant difference in IFN system-inducing capacity was observed between the WT and dam Salmonella strains following infection of cultured macrophages or epithelial cells (39). Furthermore, infection of macrophages with a Salmonella flhC mutant that does not express flagellin proteins (7, 36) activated an ISG response similar to that activated by the WT strain, but the flhC mutant elicited a significantly lower innate immune response than the WT in epithelial cells (39).

In order to further characterize the differences in innate immunity-inducing capacity between the dam mutant and WT strains in mouse infections compared with cell culture infections and to further define the contribution of bacterial flagellin to the induction of mediators of the host inflammatory response in vivo, we analyzed the relative levels of expression of an expanded set of inflammatory cytokine genes and their targets in mice at various times following oral infection. Expression was measured in different mouse tissues at early (8 h) and intermediate (2 days) times following infection with WT, dam, or flhC Salmonella and was compared to late gene expression (5 days). For our analyses, we assessed the relative transcript levels of representative type I IFN system genes (encoding IFN-β, protein Mx1 GTPase [Mx], RNA-dependent protein kinase [PKR], and suppressor of cytokine signaling 1 [SOCS1]), type II IFN system genes (encoding IFN-γ and iNOS) (35), inflammatory genes (encoding IL-1β and TNF-α), chemokine genes (encoding MIP2), and Th1/Th2 indicator genes (encoding IL-12p40 and IL-4). We found that infection of mice with Salmonella induced tissue-selective induction of the innate immune response that was similar at early and middle time points for the three strains but was significantly elevated for the WT and flhC strains compared to the dam mutant at the late time point after infection (5 days). Furthermore, cytokine gene induction was observed in individual animals early after infection (8 h), despite the absence of detectable bacteria in the liver and spleen, and induction correlated with the initial bacterial load in the Peyer's patches.

MATERIALS AND METHODS

Bacterial strains.

Salmonella pathogenic strains used in this study were derived from S. enterica Typhimurium strain ATCC 14028 (CDC 6516-60). The dam mutant contained a dam102::Mud-Cm insertion (MT2116). The lacZ transcriptional fusion to the flhC flagellar gene (flhC5496::MudJ) was obtained from K. Hughes (University of Utah) and was transduced into strain 14028 (MT2425 flhC5496::MudJ [TH3928]) (3, 7, 39). WT dam⁺ flhC⁺ strain MT2057 contained a zjf7504::MudJ transcriptional fusion which was used to discern it from other Salmonella strains which were inherently Lac⁻ (9).

Infection of mice and determination of bacterial load.

Oral infection was performed as described previously (37). Briefly, inocula were prepared by growing mutant and WT bacteria overnight in Luria-Bertani broth at 37°C with shaking. Female BALB/c mice (Charles River Laboratory) that were 6 to 8 weeks old were perorally infected by gastrointubation with 10⁹ CFU in 0.2 ml of 0.2 M sodium phosphate buffer (pH 8.0) or were mock infected with buffer alone. The oral 50% lethal dose of the dam strain is >10⁹ CFU (20), whereas the 50% lethal doses of the flhC mutant and WT strains are similar, ∼10⁵ CFU (36). Mice were examined daily following challenge for morbidity and mortality. Mice were sacrificed at 8 h or 2 or 5 days after infection as indicated below. Tissues were excised and processed rapidly for RNA analyses and for bacterial load determination. To determine the numbers of bacteria in host tissues, moribund mice were sacrificed and bacteria were recovered from host tissues and plated for colony counting. The host tissues assayed included the Peyer's patches (the Peyer's patches proximal to the ileum-cecum junction), liver, and spleen. Homogenate derived from ∼0.1 g of liver or ∼0.05 g of spleen or Peyer's patches was typically plated. The threshold of detection was <20 CFU/g for the liver and <40 CFU/g for the spleen and Peyer's patches. Reconstitution control experiments revealed that the rates of recovery of the WT, dam, and flhC Salmonella strains from both liver and spleen homogenates were >92% as measured by plating for determination of CFU after 30 min of incubation at 4°C. All animal experimental protocols were approved by the Institutional Animal Care and Use Committee.

RNA isolation.

Mice were sacrificed, and the liver, spleen, and Peyer's patches were rapidly excised, immediately frozen in liquid nitrogen, and then stored at −80°C. Tissue was homogenized with Trizol reagent (Invitrogen) using 2-ml glass-to-glass homogenizers (Fisher), and total RNA was prepared as specified by the Trizol reagent protocol. RNA concentrations were determined spectrophotometrically.

qPCR analysis.

Reverse transcription was carried out as described previously (16, 39, 43) using 2.5 μg of total RNA, random hexamer oligonucleotide (Promega) primer, and Superscript II RNase H reverse transcriptase (Invitrogen). Quantitative PCR (qPCR) then was carried out using SYBR green Supermix (Bio-Rad) and a Bio-Rad MyIQ real-time PCR thermocycler. For amplification of mouse genes, previously described primer pairs were used, as follows: for the IFN-β gene, primers described by Ogasawara et al. (29); for the IFN-γ, IL-1β, IL-12p40, IL-4, and TNF-α genes, primers described by Overbergh et al. (30); for the PKR gene, primers described by Al-Khatib et al. (2); for the SOCS1 gene, primers described by Wormald et al. (51); and for the MIP2 gene, primers described by Singer and Sansonetti (40). The following additional primer pairs were used: for the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene, forward primer 5′-GCCTTCCGTGTTCCTACCC-3′ and reverse primer 5′-TGCCTGCTTCACCACCTTC-3′; for the Mx gene, forward primer 5′-TGTGGAAGCAAGCAAGCC-3′ and reverse primer 5′-AAGACAGTTAGGAGCAGGATC-3′; and for the iNOS gene, forward primer 5′-TCCCGAAACGCTTCACTTCC-3′ and reverse primer 5′-GATGCTGCTGAGGGCTCTG-3′. Quantification of the qPCR product was performed using the MyIQ software (Bio-Rad). Relative target gene transcript amounts were normalized to the level of the GAPDH gene, relative to the average of the normalized values obtained for uninfected liver.

Statistical analysis of data.

The statistical significance of differences between gene expression values for individual groups of infected animals and the cognate uninfected mouse values were determined using the Mann-Whitney nonparametric U test (95% confidence interval). Two-tailed P values of 0.05 or less were considered statistically significant in each test.

RESULTS

dam mutant differs from WT and flhC mutant Salmonella strains in spatiotemporal colonization following oral infection of mice.

Bacterial load is a parameter that may affect the magnitude of the innate immune response. As an initial step in determining the role of bacterial Dam and flagellin in the induction of innate immunity in vivo, mice were either mock infected or infected orally with a WT, dam, or nonflagellated flhC Salmonella strain, and the bacterial loads were determined. Infected mice were sacrificed at 8 h or 2 or 5 days after infection. Peyer's patch, liver, and spleen tissue samples were harvested, and the bacterial loads (CFU/g) were measured. The bacterial loads in the Peyer's patches were comparable for the WT, dam, and flhC strains at both 8 h and 2 days following oral infection (Fig. 1). Additionally, the median level of colonization of the Peyer's patches with the Salmonella dam mutant was 10- to 40-fold lower than the level of colonization with the WT or flhC mutant at 5 days postinfection (Fig. 1).

FIG. 1.

Comparison of bacterial counts in tissues from mice infected with WT, dam mutant, or flhC mutant Salmonella. Mice were orally infected with a strain, and bacterial loads were determined for tissue samples taken from the liver, Peyer's patches, and spleen at 8 h or 2 or 5 days postinfection. Solid bars indicate the median value calculated for each experimental group of eight mice, except for the day 5 group of nine mice infected with the dam mutant. Individual mice within an experimental group were designated with numbers, as follows: 1 to 8, mock-infected mice; 11 to 18, mice infected with the WT at 8 h; 21 to 28, mice infected with the dam mutant at 8 h; 31 to 38, mice infected with the flhC mutant at 8 h; 41 to 48, mice infected with the WT at 2 days; 51 to 58, mice infected with the dam mutant at 2 days; 61 to 69, mice infected with the flhC mutant at 2 days; 71 to 78, mice infected with the WT at 5 days; 81 to 89, mice infected with the dam mutant at 5 days; 91 to 98, mice infected with the flhC mutant at 5 days. The thresholds of detection (designated by the horizontal line) were <20 CFU/g for the liver and <40 CFU/g for the spleen and Peyer's patches. UI, mock infected.

The WT and flhC Salmonella strains displayed comparable abilities to colonize both the liver and spleen (Fig. 1) and induced significant typhoid illness at 5 days postinfection. Maximal colonization of the liver and spleen by the WT and flhC strains was seen at 5 days postinfection, when the number of CFU/g was ∼100-fold higher than that on day 2. The colonization of the spleen and liver by the flhC mutant strain was slightly delayed compared to the colonization by the WT as measured by median bacterial counts at 2 days after infection (Fig. 1). By contrast, the dam mutant strain failed to detectably colonize the liver and spleen at any time examined, and the dam mutant-infected mice displayed no overt symptoms of illness, in agreement with our previous observations (20, 37).

WT and flhC and dam mutant Salmonella elicit similar immune response gene induction profiles in mice at 8 h, but there is differential induction at 5 days postinfection.

Relatively little is known regarding the ability of dam mutant Salmonella to induce inflammatory cytokine and chemokine gene expression beyond the reduced ISG expression of a limited set of ISGs seen previously when a dam strain was compared to a WT strain (17, 37). Furthermore, the contribution of bacterial flagellin to the initiation of the innate immune response relative to the contribution of Dam-regulated genes is unknown. Therefore, RNA from the Peyer's patches, livers, and spleens of groups of mice infected with the WT or flhC mutant strain was compared to RNA from mice infected with the dam mutant strain. Differential activation of several proinflammatory genes was found to be strain and spatiotemporally dependent. The statistical significance of observed differences in gene expression patterns was determined by the Mann-Whitney U test. The median values and associated P values for the gene expression analyses are shown in Table S1 in the supplemental material for the groups of infected mice; the results for individual animals are shown in Fig. 2 to 5.

FIG. 2.

Comparison of inflammatory cytokine and chemokine transcript levels in tissues recovered from mice infected with WT, dam mutant, or flhC mutant Salmonella. Total RNA was isolated from individual tissue samples taken from the liver, Peyer's patches, or spleen at 8 h or 2 or 5 days postinfection and from mock-infected controls (UI). RNA samples were analyzed by reverse transcription and real-time qPCR for (A) IFN-γ, (B) TNF-α, (C) IL-1β, and (D) MIP2 expression as described in Materials and Methods. Transcript levels were normalized between samples using GAPDH. Normalized transcript levels in the liver, Peyer's patches, and spleen are all expressed relative to the average transcript level present in the livers of uninfected mice, indicated by the horizontal line at an ordinate value of 1. The second horizontal lines shown for the Peyer's patches samples and for the spleen samples indicate the median transcript levels in the Peyer's patches and spleens of uninfected animals, respectively, relative to the levels in the livers of uninfected animals. Individual mice within an experimental group are identified by number. The solid bars indicate the median values calculated for each experimental group of mice; individual mice within an experimental group are designated as described in the legend to Fig. 1. UI, mock infected.

FIG. 5.

Comparison of IL-12p40 and IL-4 transcript levels in tissues recovered from uninfected mice and mice infected with WT, dam mutant, or flhC mutant Salmonella. Analyses were performed with RNA from mock-infected mice (UI) or mice infected with a WT, dam mutant, or flhC mutant strain as described in the legend to Fig. 2. RNA samples were analyzed for (A) IL-12p40 and (B) IL-4 expression.

(i) IFN-γ, TNF-α, IL-1β, and MIP2.

Significant induction (P < 0.05) of IFN-γ, TNF-α, and IL-1β was observed in the liver 5 days following infection with all three Salmonella strains (Fig. 2). For WT and flhC Salmonella, the induction was >100-fold for IFN-γ and TNF-α and >50-fold for IL-1β. The induction by the WT and flhC strains typically was five- to sevenfold greater than the induction by the dam mutant (Fig. 2) (see Table S1 in the supplemental material). The transcript levels of IFN-γ, TNF-α, and IL-1β were only modestly increased above the high basal level in the spleen and Peyer's patches following infection (Fig. 2A, B, and C). Note that the basal levels of IFN-γ (Fig. 2A), TNF-α (Fig. 2B), and IL-1β (Fig. 2C) were generally lowest in the liver and highest in the spleen, as has been reported previously for the levels of iNOS (34), CIITA (37), IFN-γ, IL-12, and TNF-α (47, 49, 52). The basal levels of the MIP2 chemokine transcript were similar in the liver, Peyer's patches, and spleen, and there was robust induction (>100-fold induction in the liver and spleen; ∼15-fold induction in the Peyer's patches) in all three tissues at 5 days after infection with the WT and flhC strains (Fig. 2D). The dam mutant, by contrast, was a poor and, in most cases, insignificant inducer of MIP2 (Fig. 2) (see Table S1 in the supplemental material). While the median values for IFN-γ, TNF-α, IL-1β, and MIP2 gene expression at 8 h for groups of infected mice were not statistically significantly increased compared with the values for the uninfected control group, induction of these inflammatory genes was detected in the livers of several individual animals at 8 h postinfection with all strains tested (Fig. 2), even though no bacteria were recovered from the liver or spleen at this early time point (Fig. 1). By 2 days after infection, the statistical significance increased and the animal-to-animal variability decreased.

(ii) iNOS, Mx, PKR, and SOCS1.

In order to assess the relative abilities of the Salmonella dam and flhC mutant strains to affect the expression of beacon IFN-stimulated genes, we determined the relative transcript levels of the ISGs encoding the following proteins (35, 51): iNOS (Fig. 3A), SOCS1 (Fig. 3B), Mx (Fig. 3C), and PKR (Fig. 3D). The patterns of expression of these four genes differed in the liver, Peyer's patches, and spleen following infection (Fig. 3). For iNOS (Fig. 3A), significant induction was seen in the liver, which increased with time after infection from ∼30- to 50-fold at 2 days to >3,000-fold at 5 days. The Peyer's patches and spleen, by contrast, displayed high basal iNOS transcript levels relative to the levels in the liver. However, significant iNOS induction was detected in both tissues at 5 days only after infection with the WT and flhC strains but not after infection with the dam mutant strain (Fig. 3) (see Table S1 in the supplemental material). A pattern comparable to that for iNOS transcript levels (Fig. 3A) was seen for SOCS1 transcript levels (Fig. 3B) following infection with the three Salmonella strains (WT, flhC mutant, and dam mutant). Again, the inducing capacities of the flhC mutant and WT Salmonella strains were similar and greater than the inducing capacity of the dam mutant.

FIG. 3.

Comparison of cytokine-responsive gene expression in tissues recovered from uninfected mice and mice infected with WT, dam mutant, or flhC mutant Salmonella. Analyses were performed with RNA from mock-infected mice (UI) or mice infected with the WT, dam mutant, or flhC strain as described in the legend to Fig. 2. RNA samples were analyzed for (A) iNOS, (B) SOCS1, (C) Mx, and (D) PKR expression.

Induction of Mx was most prominent in the liver, but the transcript level was significantly increased only at 5 days after infection (Fig. 3C) (see Table S1 in the supplemental material). A pattern similar to that for Mx transcript levels (Fig. 3C) was seen for PKR transcript levels (Fig. 3D) in the liver, but no statistically significant increase was seen in the Peyer's patches or spleen.

(iii) IFN-β.

When the relative IFN-β transcript levels were determined by reverse transcription-qPCR (Fig. 4), the small differences in median values between groups of infected mice were not statistically significant.

FIG. 4.

Comparison of type I IFN-β transcript levels in tissues recovered from uninfected mice and mice infected with WT, dam mutant, or flhC mutant Salmonella. Analyses were performed with RNA from mock-infected mice (UI) or mice infected with a WT, dam mutant, or flhC mutant strain as described in the legend to Fig. 2.

Wild-type, dam mutant, and flhC mutant Salmonella strains elicit differential production of Th1/Th2 adaptive immune cytokines.

The adaptive immune response is broadly divided into Th1/cellular and Th2/humoral branches that are regulated through the action of signaling molecules. Among these responses, Th1 and Th2 polarization of T cells can be distinguished in part by the production of the IL-12 and IL-4 cytokines, respectively (28, 44).

(i) IL-12p40.

Infection with either the WT or the flhC mutant resulted in greater induction of the IL-12p40 transcript in the liver than infection with the dam mutant (Fig. 5A). The median induced IL-12p40 transcript levels seen at day 5 after infection were more than those seen at day 2, which were more than those seen at 8 h (Fig. 5) (see Table S1 in the supplemental material). By contrast, in the Peyer's patches, a significant decrease in the IL-12p40 RNA level from the high basal level was seen in mice at 5 days after infection with the WT and flhC mutant strains (Fig. 5A) (see Table S1 in the supplemental material).

(ii) IL-4.

The pattern of IL-4 transcript expression (Fig. 5B) differed markedly from that of IL-12p40 transcript expression (Fig. 5A). Infection with the Salmonella dam mutant caused a modest but significant increase in the IL-4 transcript level in the liver after 5 days, but no significant change was seen with either the WT or flhC mutant strain (Fig. 5B) (see Table S1 in the supplemental material). Furthermore, in the Peyer's patches and spleen, a decrease in the level of IL-4 expression was observed at 2 and 5 days after infection. Taken together, these results suggest that the Salmonella dam mutant exhibited a reciprocal IL-12/IL-4 cytokine ratio relative to that exhibited by the WT or flhC mutant.

Induction of proinflammatory cytokine gene expression in the liver 8 h postinfection shows that there is a critical bacterial load dependence in the Peyer's patches that is independent of the flhC and dam mutations.

Colonization of the liver and spleen was not detectable at 8 h after oral infection of mice (Fig. 1), whereas induction of multiple inflammatory cytokines was observed in the livers of individual animals (Fig. 2). When the relative liver transcript levels of IFN-γ, TNF-α, IL-1β, and MIP2 at 8 h after infection for individual mice were plotted against the bacterial load found in the Peyer's patches of the same animals at 8 h, induction of inflammatory gene expression appeared to require a critical threshold bacterial load of ∼10⁴ CFU/g (Fig. 6). The critical threshold bacterial loads were similar for all three bacterial strains and were comparable for induction of IFN-γ (Fig. 6A), TNF-α (Fig. 6B), IL-1β (Fig. 6C), and MIP2 (Fig. 6D).

FIG. 6.

Cytokine gene induction in the liver at early times after infection correlates with the bacterial load in the Peyer's patches. Transcript levels of (A) IFN-γ, (B) TNF-α, (C) IL-1β, and (D) MIP2 in the liver at 8 h after infection with a WT, dam mutant, or flhC mutant strain, as shown in Fig. 2, were plotted against the corresponding bacterial load present in the Peyer's patches at 8 h after infection. A best-fit curve for each Salmonella strain is shown. The numbers indicate data for individual mice designated as described in the legend to Fig. 1.

DISCUSSION

An important component of the host's innate immune response to microbial infection, including Salmonella infection, is the production of proinflammatory cytokines and chemokines and activation of the IFN response (38, 45, 50). While much has been learned about the specific bacterial defects observed following loss of Dam, including bacterial membrane instability (32) and altered transcription of virulence factors (3, 4), comparatively little is known regarding the ability of a Salmonella dam mutant to elicit the innate immune response. We used WT and flagellin-deficient Salmonella flhC mutant strains together with a dam mutant to further delineate the characteristics of the innate immune response elicited following oral infection of mice. Several important points emerged from our findings.

First, infection of mice with Salmonella by the natural oral route resulted in robust induction of several proinflammatory and IFN system genes, including the genes encoding IFN-γ, iNOS, SOCS1, TNF-α, IL-1β, MIP2, and IL-12p40. By contrast, comparatively modest induction of the type I IFN Mx and PKR gene beacons was observed, especially following infection with dam bacteria. These results, obtained at 5 days after infection, are consistent with the results at 6 days postinfection obtained in our initial study, which examined a very limited set of IFN system genes (37). The results described here further suggest that the reduced induction of innate immune and IFN system genes elicited following infection with the dam mutant compared to the induction obtained with WT Salmonella was largely the result of restricted colonization and reduced bacterial load with the dam mutant. Importantly, the dam mutant did not evoke overly robust proinflammatory cytokine induction and subsequent septic shock, as observed with WT Salmonella.

While differential activation of the innate immune system by bacteria deficient for defined PAMPs clearly can occur in defined cell types in culture (23, 39, 42, 53), we found no significant difference in the gene expression profiles between mice infected with the WT and mice infected with the Salmonella flhC mutant, at least in the Peyer's patch, spleen, and liver tissues, which consisted of multiple cell types. These findings are consistent with the notion that the actions of bacterial constituents other than flagellin (for example, peptidoglycan via TLR2, lipopolysaccharide via TLR4, or CpG-rich DNA via TLR9) provide functionally redundant activation of the host innate immune system in vivo (23).

We found reduced colonization of the liver and spleen by the flhC mutant compared to the WT at day 2 but no difference in colonization or gene induction at day 5. Redundancy in the immune response evoked by WT Salmonella has also been established with knockout mice, as infection of TLR5^+/+ animals with Salmonella and infection of TLR5^−/− animals with Salmonella result in similar mortality, but mice deficient for both TLR4 and TLR5 succumb more rapidly following WT infection (14). As we show here, because the Salmonella nonflagellated mutant induced an immune response similar to that induced by the WT, the ability to produce flagellin is not an obligatory requirement for maximal activation of the innate immune response. Conceivably, the role of flagellin following Salmonella infection may be most pertinent in the context of the kinetics of infection rather than modulation of a global inflammatory immune response. Deficiency in the flagellin-TLR5 system results in delayed infection following oral infection in the mouse model due to reduced activation of TLR5-expressing intestinal lamina propria cells (36, 46).

Alternatively, another possibility is that in selective microenvironments within the mouse, expression of flagellin may be suppressed following WT infection, and thus the WT bacteria may appear to the host to be functionally equivalent to the nonflagellated flhC mutant in the context of recognition by the innate immune system (10, 11).

Although no salmonellae were detected in the liver or spleen with any of the strains at 8 h postinfection (Fig. 1), inflammatory gene induction was observed in individual animals (Fig. 2). This induction in the liver at 8 h after infection correlated well with the bacterial load at the primary site of infection, the Peyer's patches (Fig. 6). The early proinflammatory gene induction seen in the liver and the spleen may have been due to the action of cytokines produced in the Peyer's patches or, alternatively, due to the release of bacterial PAMPs, such as lipopolysaccharide, peptidoglycan, or flagellin, from the Peyer's patches into the circulatory system, followed by filtering through the hepatic portal system or clearance by the reticuloendothelial system (41). The induction in the liver observed at 8 h postinfection, expressed as a function of the bacterial load in the Peyer's patches, was remarkably strain independent.

The greater induction of IL-4 seen following dam mutant infection may have been due to enhanced release of bacterial surface components, such as flagellin, in the Peyer's patches, which then were able to influence a shift towards a Th2 phenotype in infected mice (12, 13). The pronounced decreases in the levels of the IL-4 and IL-12 transcripts seen in the spleens of WT- and flhC mutant-infected mice may conceivably have reflected a redistribution of the splenic cell populations that affected cytokine expression during Salmonella infection. Indeed, during the course of WT Salmonella infection the population of cells in the spleen has been demonstrated to undergo a significant shift from lymphocytes to macrophages and neutrophils both in number and as a percentage (25). Such a fundamental shift in cell type distribution may in part account for the diminished expression of Th1/Th2 cytokines in the spleen.

The relatively high basal gene expression that we found in the spleen compared to the liver of uninfected mice is consistent with the observations of other workers, including observations for iNOS (34), CIITA (37), IFN-γ, IL-12, and TNF-α (47, 49, 52). The reason for the different basal transcript levels observed in different tissues is not completely clear, but the difference may be due to different requirements for the cognate gene products in the general maintenance of tissue physiology or for the function of defined cell types, such as lymphocytes, that differentially populate the tissues tested (18, 27). Some of the beacon genes most strongly induced (for example, the TNF-α and IFN-γ genes) have been reported to be integral to the host innate defense against Salmonella (26, 50).

In contrast to the robust IFN-β induction obtained in macrophage cell culture (39), no significant increase in the IFN-β transcript level was detected in mice following Salmonella infection. However, because induction of the Mx transcript was evident and because Mx is inducible by type I IFN but not by other signaling pathways as far as is known (35), this implies that a type I IFN was indeed induced even though, surprisingly, no IFN-β transcript was detected at any of the times examined. One possible reason for the absence of detectable changes in IFN-β RNA in vivo may be the timing of the analysis, as the temporal induction of and subsequent decrease in IFN-β transcript levels in Salmonella-infected J774 cells in culture are characterized by a very rapid onset, sharp maximum, and rapid decay (39). Since type III IFNs are known to signal in a manner similar to that of type I IFNs (31), another possibility is that IFN-β may not be the inducer of type I IFN beacon genes in an infected animal but rather is one of the IL-28 type III IFN-like proteins that is responsible for the induction of the transcripts indicative of type I IFN bioactivity.

Infection with the dam strain evoked a system-wide early immune response similar to that evoked by the WT that is sufficient to induce protective immunity but does not cause the morbidity and mortality associated with typhoid illness (20, 37). The robust cytokine induction found in the WT- and flhC mutant-infected mice can be problematic, especially when it results in septic shock (5, 14, 22, 24, 33). By contrast, low-level activation of the innate immune response, as we observed in dam mutant-infected mice, may be important for clearance of the invading pathogen.