Pregnancy Impairs the Innate Immune Resistance to Salmonella typhimurium Leading to Rapid Fatal Infection¹

Salmonella enterica serovars are Gram-negative bacteria that can cause systemic infection and enteritidis. Typhoid fever in humans caused by S. enterica serovar typhi contributes to >600,000 deaths per year. S. enterica serovar typhimurium (ST)⁵ induces gastroenteritis in humans, as well as livestock, and is a major foodborne pathogen in the developed world (1). The mouse model of ST infection mimics human typhoid, causing disseminated disease in many peripheral organs including spleen, liver, and lymph nodes (2). The genetic background of the mice strongly influences outcome of infection; C57BL/6 mice succumb within 7 days even to a low dose, whereas 129×1/SvJ mice develop a chronic infection that lasts 60–90 days. Susceptibility of C57BL6/J mice has been attributed to a mutation in the NRAMP gene, whereas 129×1/SvJ mice have a normal NRAMP gene which confers resistance (3).

ST is an intracellular bacterium that resides within the modified phagosomes of APCs (4). Innate immunity that includes NK cells, NKT cells, dendritic cells, and neutrophils are all critical in controlling the early primary infection (5). Inflammatory cytokines produced by innate immune cells including TNF-α, IFN-γ, IL-12, and IL-18 are also important for curtailment of infection during the first week before onset of adaptive immunity (5, 6). CD4⁺ T cell response to ST infection is generally detectable only beyond 7 days of infection (7), whereas CD8⁺ T cell response is substantially delayed until the second week postinfection (8). Overall, ST appears to have evolved many mechanisms to evade the host immune system and establish chronic infection.

In humans, the high-risk populations for Salmonella infections include the young, old, pregnant, transplant patients, and HIV-infected individuals (9, 10). Salmonella serovars also lead to pregnancy loss in livestock (11). Despite evidence for increased susceptibility of the immunocompromised host to many Salmonella species, the underlying mechanisms remain largely unknown. Pregnancy confers a transient altered immune status in the maternal host as responses are biased toward type 2 (humoral) and away from type 1 (inflammatory and cell-mediated) phenotype (12, 13). This shift in the immune state during pregnancy, while pronounced at the maternal-fetal interface, may to some extent modulate systemic immunity. Pregnancy can have deleterious effect on the outcome of infections such as leishmaniasis, malaria, toxoplasmosis, and listeriosis (14, 15, 16). Furthermore, autoimmunity such as systemic lupus erythematosus in which the principal pathology is autoantibody production tends to flare up during pregnancy, whereas rheumatoid arthritis, an inflammatory disorder, is ameliorated in the maternal host (17). However, most infections during pregnancy, while detrimental to the fetus, do not lead to fatal maternal outcome.

The mechanisms of host response to infections during pregnancy need to be understood to enable better management of epidemic outbreaks and to reduce the risk of horizontal and vertical spread of disease. In particular, very little is known about the specific host-pathogen interactions of intracellular bacteria with an immunocompromised host. In this study, using the mouse model of infection, we have defined some of the factors that influenced ST pathogenesis in pregnant hosts. We show that ST rapidly proliferates in the placenta leading to invasive fetal disease within 14 h. Additionally, pregnant hosts become fatally susceptible to infection due to defective systemic innate immunity that is plausibly triggered by altered immune responsiveness to ST in the placental environment.

Virulent ST strain SL1344 (18) and the mutant strain ssaR were generated and maintained as previously described (19). Briefly, bacteria were grown in liquid culture in brain-heart infusion (BHI) medium (Difco Laboratories). At mid-log phase (OD₆₀₀ = 0.8), bacteria were harvested and frozen at −80°C (in 20% glycerol). CFU were determined by performing serial dilutions in 0.9% NaCl, which were spread on BHI agar plates.

129×1/SvJ mice, 6–8 wk of age, were purchased from The Jackson Laboratory. Mice were maintained in the animal facility at the Institute for Biological Sciences (National Research Council Canada, Ottawa, Ontario, Canada) in accordance with the recommendations of the Canadian Council on Animal Care. One male and two females were housed per cage overnight and mating was determined by the presence of a vaginal plug on the following morning. The day of vaginal plug observation was defined as pregnancy day 0. For infection, frozen stocks of bacteria were thawed and diluted in 0.9% NaCl. Mice were inoculated with 1 × 10³ organisms suspended in 200 μl of 0.9% NaCl, via the lateral tail vein (i.v.). Infection was initiated at various stages of pregnancy: middle (days 10–12) and late (days 14–15), and in nonpregnant age-matched mice. For depletion of IL-6 in vivo, mice were treated with 100 μg/100 μl (i.p.) of functional grade anti-mouse IL-6 Ab (clone MP5-20F3; eBioscience) twice (day of initiation of infection and 2 days later).

Mice were euthanized by CO₂ asphyxiation and spleen, liver, placentas/uteroplacental tissue, and fetal liver were aseptically removed. For the spleen, single-cell suspensions were obtained by squishing the organ between frosted ends of a glass slide. Liver, placentas (pooled per mouse), and fetal liver (pooled per mouse) were homogenized using a motorized homogenizer. In some cases, wherein it was difficult to discern the placentas of resorbed fetuses, the entire uteroplacental unit was homogenized. An aliquot of the cell suspension or homogenate was lysed with water for 30 s, and then evaluated for the numbers of viable bacteria. Ten-fold serial dilutions of the tissue homogenates, in 100-μl volume of 0.9% NaCl were plated on BHI agar. For assessing bacterial burdens at early time points after infection, the entire volume of tissue homogenate was spread onto several plates. Colonies were counted after 24 h of incubation at 37°C.

Fetal resorptions were identified by the notably smaller size and necrotic or hemorrhagic appearance of the fetus and/or placenta when compared with normal viable fetuses and/or as resorbing necrotic scars in the uterus. The percentage resorption rate was calculated using the formula R/(R + V) × 100, where R is the number of resorbing fetuses and V is the number of viable fetuses per animal.

Single-cell suspensions of spleens were analyzed for the various immune cell types based on their surface expression of various markers. For staining with all Abs, cells were first incubated on ice (10⁶ cells in 100 μl of PBS plus 1% BSA) with anti-mouse CD32/CD16 (FcγII/III receptor). After 10 min, 3–5 μl of different FITC- or PE-labeled anti-mouse Abs were added and incubated for an additional 30 min on ice. Abs against the following cell surface markers were used to identify the various immune cell types: B220, CD4, CD8, TCRγδ, CD25, F480, MAC-1, CD11c, Gr-1, NK1.1, DX5, CD94, and Ly49D. All Abs were purchased from BD Biosciences. After 30 min, cells were washed and fixed in 1% formaldehyde in PBS and acquired on an EPICS XL flow cytometer (Beckman Coulter). Analysis was done using EXPO software (Beckman Coulter).

Uterine tissue from nonpregnant or placenta from pregnant mice were cut into small pieces, and treated with 0.5 mg of collagenase type IV (Worthington Biochemical) for 10 min (37°C), and then a single-cell suspension was obtained by squishing the tissue with a plunger over a cell strainer. Cells were subject to Percoll Plus (GE Healthcare Biosciences) gradient centrifugation and the lymphocyte interface was collected. Cells were then stained with anti-DX5 Ab, similar to the staining procedure outlined for spleen cells.

Cytotoxicity of NK cells in vitro was assessed in a ⁵¹Cr-release assay on YAC-1 target cells. Murine NK-sensitive target cells, YAC-1 (Moloney murine leukemia virus-induced lymphoma cells) as well as NK-insensitive cells, P815 (mastocytoma cells, H-2^d) were propagated in RPMI 1640 medium additionally supplemented with 8% FBS and 10 μg/ml gentamicin (R8 medium) 37°C, 8% C0₂. Both cell lines were obtained from the American Type Culture Collection (ATCC). For the assay, 5 × 10⁶ target cells were labeled with 50 μCi ⁵¹Cr (Amersham Pharmacia Biotech) for 1 h at 37°C and washed three times. Effector cells were prepared by tweezing the spleens between the frosted ends of two sterile glass slides in R8 medium. Cells were subsequently passed through Falcon 2360 cell strainers (BD Labware), centrifuged, and resuspended in R8 medium. Various ratios of effectors and targets were cocultured for 4 h at 37°C in 96-well round-bottom tissue-culture plates (Falcon). The supernatants were collected, and radioactivity was detected by gamma counting. The percentage of specific lysis was calculated using the formula: 100 × ((experimental cpm − spontaneous cpm)/(total cpm − spontaneous cpm)). The percentage-specific killing obtained at various E:T ratios was also converted to lytic units, wherein 1 lytic unit represents the number of effectors yielding 15% specific lysis of 2.5 × 10⁴ YAC-1 targets.

Blood was obtained by cardiac puncture under anesthesia, before euthanasia of mice, and collected in microtainer serum separator tubes (BD Biosciences). After the blood was allowed to clot at 4°C, the serum was separated by quick high-speed centrifugation, and stored at −70°C. Levels of serum IL-12, IL-6, IFN- γ, and IL-10 were assayed by sandwich ELISA. The following Ab pairs were used: R4-A62 (ATCC HB170) and XMG1.2-biotin for IFN-γ; SXC4 and SXC1-biotin for IL-10. IL-6 and IL-12 Ab pairs were purchased from BD Biosciences. Cytokine standards were purchased from ID Labs. Duplicate standard curves encompassing several doubling dilutions of the standard were included on each plate. All serum samples were assayed at the same time to minimize day-to-day error in cytokine detection. The detection limits for cytokines in the ELISAs were 200 pg/ml IFN-γ, 200 pg/ml IL-10, 200 pg/ml IL-6, 100 pg/ml IL-12p40.

Placentas from individual mice were dissected out, pooled, and snap-frozen in a dry ice/100% ethanol bath. In the case of ST-infected pregnant mice, where it was not always possible to discern individual placenta, the whole uteroplacental unit was snap-frozen. Total RNA was extracted using the Qiagen RNeasy Mini kit according to the instructions of the manufacturer along with rapid mechanical lysis. Briefly, the placentas or uteroplacental units were cut into pieces and lysed in 1 ml of lysis buffer in a MiniBeadbeater 3110BX (BioSpec Products) with glass beads (φ = 0.5 mm and φ = 0.1 mm; BioSpec Products). Total RNA from homogenates was extracted and treated with RNase-free DNase I (Roche Applied Science) for 30 min at 37°C. DNase was then removed according to the instructions of the manufacturer. A total of 2–5 μg of total RNA was taken for cDNA synthesis. cDNA was synthesized using AncT primers (Sigma-Aldrich). RNA was made linear at 65°C for 5 min and cDNA was synthesized in a 40-μl reaction volume containing: 1.5 μl of AncT primers (100 pM/μl), 8 μl 5 × first-strand buffer, 4 μl of DTT (100 mM), 5 μl of dNTP (5 mM), 1 μl of RNase OUT (40 U/μl), 2 μl of Superscript II (200 U/μl) (Invitrogen Life Technologies), and 15 μl of RNA template. Reverse transcription was performed in a Thermo Cycler 9700 (Applied Biosystems) at 42°C for 15 min and 45°C for 2 h. Identical samples not treated with Superscript II were also prepared as controls to measure DNA contamination. The remaining RNA template was hydrolyzed with 1 M NaOH at 65°C for 5 min and neutralized with 1 M HCl. cDNA was purified using Microcon YM-30 centrifugal filter units (Millipore). The number of amplicons was measured by quantitative real-time PCR using gene-specific primers and quantitative PCR SYBR green supermix (ABgene). Primers were designed using Primer Express 2.0. The primer pairs used are listed in Table I. β-actin was used as an internal reference control. Ten-fold dilutions of cDNA were used as template to generate the standard curve for each primer-template set (1×, 1/10×, 1/100×, 1/1000×). This standard curve was run together with triplicate reactions of the uncharacterized samples. PCR was performed in sealed tubes in a 96-well microtiter plate in an ABI Prism 7000 thermocycler (Applied Biosystems). The 25-μl reaction consisted of 12.5 μl of quantitative PCR SYBR green supermix, 2.5 μl of primer mix (1.5 pM/μl each), and 10 μl of template. Thermal conditions were as follows: activation at 95°C for 15 min, followed by 40 cycles of denaturation at 95°C for 15 s, annealing at 60°C for 1 min, and extension at 72°C for 1 min. Fluorescence was measured during the annealing step and plotted against the amplification cycle. Relative quantitative analysis of the data was extrapolated from the standard curve. Primer efficiencies were 98–100%.

Nonparametric Mann-Whitney or Student’s t test, as appropriate, and stated in the figure legends were used to determine the statistical significance of the experimental data.

Normal C57BL/6J mice succumbed to ST infection (given i.v.) within 7 days even to a low dose (10²), whereas 129×1/SvJ mice resisted infection even with 10³ dose of ST (i.v.), and developed a chronic infection that lasts for ∼60 days, followed by clearing (Fig. 1,a). In contrast >60% of 129×1/SvJ mice infected with ST in midpregnancy (days 10–12), succumbed to infection within 7 days (Fig. 1,b). The median survival in the pregnant group was 6 days. This correlated to a rapid increase in bacterial burden in the organs of pregnant mice. When ST infection was initiated in middle or late pregnancy, the bacterial burden in the spleen was profoundly increased (∼1000-fold) in comparison to the burden achieved in the spleens of nonpregnant age-matched mice (Fig. 1,c). The exacerbation of ST infection observed in pregnant mice also correlated to substantial bacterial colonization of the placenta. Infection with 10³ CFU of ST resulted in >10⁷ bacteria in the uteroplacental units of individual mice within 3 days (Fig. 1,d). This resulted in a >75% fetal loss (Fig. 1 e). Furthermore, infection in late pregnancy (days 14 and 15) induced premature labor in many mice (data not shown), again demonstrating the highly detrimental effect of ST infection on pregnancy outcome.

A highly attenuated mutant strain of ST, ssaR, is defective for the Salmonella pathogenicity island II (SPI-II) type III secretion, and is unable to secrete type III effectors proteins, and hence is unable to survive or rapidly proliferate within phagocytic cells (20). 129×1/SvJ mice were infected with 10³ CFU of ssaR in middle pregnancy (days 10–12) and the bacterial burden was determined in the spleen (Fig. 2,a) and liver (Fig. 2,b) on day 3 of infection in age-matched nonpregnant and pregnant mice. The bacterial counts were similar in the peripheral organs of both nonpregnant and pregnant mice, suggesting that host resistance to infection remained intact. However, ssaR infection was deleterious to pregnancy outcome, as colonization of the placenta was observed (Fig. 2,c). Furthermore, this correlated to ∼50% fetal resorption rate (Fig. 2 d). Thus, ssaR infection results in less severe systemic bacterial burden in the pregnant host, despite placental colonization and adverse fetal outcome.

To address the mechanism(s) responsible for the rapid loss of host resistance to ST infection during pregnancy, we evaluated the phenotype and numbers of the various immune cell types in the spleen (Figs. 3 and 4). Four groups of mice were included in the study: naive (nonpregnant), naive pregnant, infected (nonpregnant, age matched), and infected pregnant. First, in comparison to naive mice, pregnant mice in the absence of any infection exhibited significantly reduced percentages of CD4⁺ and CD8⁺ T cells, whereas differences in other cell types were marginal and insignificant (Fig. 3). ST infection of nonpregnant 129×1/SvJ mice induced increased percentages of many innate immune cell types defined by the expression of F480, MAC-1, CD11c, and Gr-1 (macrophages, dendritic cells, and neutrophils). Most striking was the significant increase in NK1.1⁺ and DX5⁺ cells, indicative of a rapid onset of NK cell response to infection which was lacking in infected-pregnant mice. Both subsets of DX5⁺CD94⁺ and DX5⁺Ly49D⁺ were lower in pregnant-infected mice (Fig. 3). The overall numbers (Fig. 4) of NK cells and their subsets in the spleen was also markedly decreased in infected-pregnant mice, reiterating the lack of effective NK response against ST infection during pregnancy. Infected-pregnant mice also showed significant reductions in the percentages of dendritic cells, neutrophils, and TCRγδ T cells as evident from the decreased splenic expression of CD11c, Gr-1, and TCRγδ, respectively (Fig. 3). Calculating the overall numbers of these innate immune cell subsets also supported their decrease in the spleen (data not shown). Infected-pregnant mice exhibited lower percentages of MAC-1⁺ cells but differences in percentages of F480⁺ cells were statistically insignificant compared with infected nonpregnant mice. Thus, whether macrophage numbers are modulated in infected-pregnant mice is unclear. The percentages of T and B cells defined by the expression of CD4, CD8, CD25, and B220 were also lower in infected-pregnant mice in comparison to nonpregnant controls. However, as T cells showed a constitutive reduction in pregnant animals even in the absence of infection, it was deduced that ST did not cause any further reduction of adaptive immune cell types. Furthermore, effects of ST infection were studied on day 3 of infection, an early time point when adaptive immunity may be expected to play little role. In summary, it appeared that the ability of the host to elicit a rapid recruitment and activation of systemic innate immune response to infection was severely compromised during pregnancy.

NK cells are the first line of defense against many infections, and are capable of mounting a cytotoxic response against infected target cells. Age-matched nonpregnant and pregnant mice were challenged with ST, and the ability of splenic effectors to kill NK-sensitive targets was evaluated on day 3 of infection. Splenic effectors from infected-nonpregnant mice killed YAC-1 targets effectively, and exhibited negligible killing on NK-insensitive P815 target cells (Fig. 5,a). In contrast, splenic effectors from infected-pregnant mice exhibited weak killing of YAC-1 target cells (Fig. 5 b). Stringently comparing the data based on lytic U/10⁶ spleen cells, the level of NK cytolytic activity on day 3 of ST infection was ∼2 U/10⁶ spleen cells in infected-nonpregnant mice and was significantly lower, <0.5/10⁶, for infected-pregnant mice (p = 0.003). The weak cytolytic activity was not attributable to kinetics of the assay, as NK activity was minimal before day 3 (data not shown).

Uterine NK (uNK) cells constitute a distinct population during pregnancy and are important for early implantation. It has been suggested that uNK cells are recruited from splenic lineage cells (21, 22). Furthermore, uNK cells have been implicated in inflammation-induced pregnancy loss (23). Thus, it is possible that the reduced splenic NK activity noted in ST-infected pregnant hosts is a result of redistribution of NK subsets to the uteroplacental environment. We therefore evaluated the numbers of DX5⁺ NK cells in the uterus (Fig. 6). In nonpregnant mice, the uNK cells accounted to 2–5% of the uterine lymphocyte population. This number did not change significantly in the naive pregnant uterus. However, there was a clear decrease in the uNK cell numbers of infected nonpregnant mice suggesting redistribution of NK lineage cells to other lymphoid organs. In contrast, in pregnant mice, on day 3 postinfection uNK cell numbers did not show any change in response to infection (Fig. 6). Furthermore, their numbers remained at levels comparable to naive pregnant mice, even at 24 h after infection (data not shown).

The data presented thus far support a defective innate cell recruitment/activation in pregnant mice. As NK cells showed the most significant activation in response to infection in nonpregnant hosts, we examined in additional detail the differences in NK cell function and distribution. Although NK cell response was diminished in the spleen, it appeared that uNK numbers remained stable in pregnant-infected host. Thus, it was not clear whether NK cells could be solely responsible for the pathology of pregnancy loss and/or defective systemic resistance to infection. Indeed, a highly virulent pathogen such as ST may evoke multiple interactions with host immune cells at various sites of infection. Thus, in the next series of experiments, we addressed the role of inflammatory cytokines.

We measured cytokine levels in the serum of individual mice. IL-12 production is important for influencing the function of many cell types including dendritic cells, NK cells, and T cells. In comparison to noninfected naive mice, ST infection evoked increased production of serum IL-12 on day 3. However, this increase in serum IL-12 did not occur in pregnant ST-infected mice (Fig. 7,a). In contrast, serum IL-6 levels (Fig. 7 b) were elevated in infected-pregnant mice in comparison to infected-nonpregnant or naive controls. In contrast, serum levels of IFN-γ and IL-10 were similar in both infected-nonpregnant and -pregnant mice on day 3 of ST infection (data not shown). TNF was not detected in the serum of any of the groups (data not shown). Thus, ST-infected pregnant mice exhibited a skewed peripheral inflammatory response characterized by overt production of IL-6 and decreased production of IL-12.

Maintenance of healthy pregnancy is associated with placental secretion of anti-inflammatory cytokines IL-10 and TGF-β (24, 25). As ST infection was causing a dramatic increase in fetal loss, we sought to ascertain the relative levels of inflammatory (IL-6, IL-18, TNF, IL-12p40, and IFN-γ) and anti-inflammatory cytokines (IL-10 and TGF-β) in the uteroplacental tissue. This was achieved by performing quantitative RT-PCR on samples from individual mice and normalizing values to expression levels of β-actin. Between days 15 and 18 of gestation, uteroplacental tissue from ST-infected mice (day 3 postinfection) showed a significant increase in the expression of cytokines IL-6 (∼40-fold increase), TNF-α (∼8-fold increase), IL-18 (∼3.5-fold increase), and IL-10 (∼10-fold increase) relative to expression levels in the placenta of healthy noninfected mice (Fig. 8). IFN-γ expression levels also showed an increased trend in infected-placental tissue, but lacked statistical significance. In contrast, the expression levels of IL-12p40 and TGF-β were similar in both healthy and infected uteroplacental tissue. Thus, uteroplacental cytokine expression also suggested overt production of IL-6 in pregnant hosts.

Of the various cytokines, IL-6 appeared to be dramatically increased both in the serum and placenta. Thus, we sought to determine whether this bore any direct consequences to bacterial burden or fetal loss. We therefore depleted IL-6 in vivo by Ab treatment during the 3-day window of ST infection in pregnant mice. Fig. 9,a shows that anti-IL6 treatment of pregnant ST-infected mice significantly decreased the splenic bacterial burden, relative to untreated pregnant-infected mice. Blocking the function of IL-6 in vivo did not reduce placental bacterial load or resorption rates (Fig. 9, b and c). Anti-IL6 treatment of nonpregnant mice did not alter their overall resistance to ST infection. Overall, it appears that the increased production of IL-6 may be one of the mechanisms responsible for impaired control of ST in the periphery.

Although blocking the IL-6-mediated inflammation in the pregnant host reduced systemic ST infection, it failed to reduce placental infection and fetal resorptions. Therefore, a kinetic study was undertaken to elucidate the rate of infection in various tissues. Fig. 10 demonstrates the bacterial burden in the spleen and placenta at 2–72 h after infection. After infection with 10³ dose, bacteria could be detected in the spleen and placenta as early as 2 h after infection, with ∼200 bacteria detected in the spleen and <50 detected in the placenta. After 14–30 h, splenic bacterial burden showed very little increase from the initial 2-h counts and was similar in both nonpregnant and pregnant hosts. In contrast, the bacterial burden in the placenta had increased to 10⁵ by 14 h, and 10⁸ by 27–30 h. Differences in the splenic bacterial burden between pregnant and nonpregnant hosts became evident only at 72 h. These results suggest that although ST may not show a predilection for the placenta at 2 h, it proliferates profoundly and unabatedly in the placenta. Indeed, it appears that the bacterial numbers double approximately every hour in the placenta. Furthermore, the placental infection appeared to be highly invasive, as colonization of fetal liver occurred by 14 h, and ∼10³–10⁴ bacteria were recovered by 24 h (n = 5 mice, fetal livers of litter from each mother were pooled, homogenized, and plated). By 48 h, we could not clearly discern the fetus in many animals. Thus, the propensity of ST to proliferate in the placenta appears to acutely breach the placental barrier.

Typhoid fever is a major public health problem in developing countries, whereas the incidence of nontyphoidal salmonellosis caused by enteritidis species is on the rise even in the developed world. Immunocompromised individuals are at high risk for bacteremia and systemic spread (26). Complications in pregnancy due to Salmonella infections include endomyometritis, salpingitis, chorioamnionitis, transplacental infection and septic abortion, neonatal septicemia, and meningitis, and there are a few reported cases of life-threatening septicemia in the mother (9, 27). Abortions due to Salmonella serovars is also common in live stock, leading to huge economic losses (11). Thus, there is clinical evidence supporting an ineffective response to Salmonella species during pregnancy. However, to our knowledge, there have been no reports on the immune mechanisms that modulate pathogenesis of Salmonella infection in pregnant hosts.

Our data implicate a two-way Salmonella-induced immunopathology that adversely affects both the fetus and mother. The first event of pregnancy loss appears to be triggered by the rapid proliferation of Salmonella in the secluded placental environment. Two hours after infection, only ∼5% of the infective dose (∼50 bacteria) reached the placenta, yet this number grew astoundingly to >10⁵–10⁷ by 14–30 h. In contrast, relatively higher numbers (∼200 bacteria) homed to the spleen, but they expanded only marginally in the first 24 h. This marked contrast in the ability of Salmonella to proliferate in the spleen vs placenta is suggestive of a unique escape and/or invasive mechanism in the pregnant host. Maternal malaria is associated with sequestration of plasmodium-infected erythrocytes in the placenta (28). Similarly, Listeria monocytogenes uses actA filament-mediated entry (29) to favorably proliferate in the placenta, despite initial tropism to lymphoid tissue (30, 31). Whether the rampant placental ST infection is due to accumulation of infected cells from other organs or preferential trophoblast entry needs further study.

Our study adds Salmonella to the list of pathogens such as Toxoplasma, Chlamydia, Listeria, and Plasmodium that preferentially invade the placenta and may cause abortion through local inflammation (15, 16, 31, 32). The placental interface does express TLRs and can trigger the innate cascade (32, 33) to protect the host. Indeed, ST infection is also associated with increased placental expression of IL-6, IL-18, TNF, and IFN-γ. It is possible that the interaction of Salmonella LPS with its TLR-4 ligand on the trophoblast triggered the profound local inflammation and trophoblast apoptosis (33).

The second rather surprising event in Salmonella-induced immunopathology is the profound and rapid loss of systemic host resistance, resulting in the conversion of a normally resistant strain of mouse to a susceptible one. Overt inflammation can cause substantial fetal damage (34, 35) and thus it is understandable that infections that augment inflammation cause pregnancy loss. This may be an evolutionary advantage to avoid wastage of resources on a damaged offspring and preserve the mother. However, it appears that if the pathogen is highly evasive and virulent, such as Salmonella, the deviation of inflammatory responses during pregnancy can also result in catastrophic host outcome. Increase in splenic bacteria was preceded by heavy placental infection. Therefore, it is possible that reverse trafficking of ST from placenta to the spleen occurred. There is precedence for trafficking of bacteria from placenta to spleen in the case of L. monocytogenes infection (30), although effective immunity to Listeria in the systemic compartment ensues. Contrastingly, despite high ST burden by day 3 in the spleen of pregnant hosts, recruitment and activation of many innate immune cell types (NK cells, neutrophils, dendritic cells) was reduced in comparison to nonpregnant hosts.

One possibility was that pregnancy prevented activation of cell types such as NK cells as a perturbation in their peripheral numbers can lead to spontaneous abortion and pre-eclampsia (35, 36, 37). Pregnancy hormones, estrogen, progesterone, and prolactin may have dampening effects on peripheral NK cells (38). However, in the absence of infection, we observed no marked difference in the overall numbers of splenic NK cells in the pregnant and nonpregnant mice. Thus, it is likely that the peripheral expansion/recruitment/activation of NK cells evoked by infection may be inefficient during pregnancy. The Ly49 receptor family are constituted as homodimers that bind directly to their MHC class I ligands and Ly49D is considered to be stimulatory (39). In contrast, CD94 on NK cells is coexpressed as a heterodimer with NKG2, and the resulting receptor may be stimulatory or inhibitory dependent on the associating NKG2 isoform (39). The decrease of both NK1.1⁺Ly49D⁺ and NK1.1⁺CD94⁺ subsets in infected-pregnant mice suggests lack of stimulatory receptor activation during pregnancy, and consequent weak NK cell function, resulting in impaired clearance of infected cells. Indeed, we noted that depleting NK cells (with anti-NK1.1 Ab treatment) increased (1.6-fold) bacterial burden in nonpregnant 129×1Sv/J mice suggesting a role for NK cells in facilitating ST clearance (data not shown). However, the apparent decrease in peripheral NK cells during pregnancy may be a consequence of their redistribution to the uterus (21). uNK cells are important for decidualization (22), however, they express mainly NK inhibitory receptors and have weaker cytolytic activity (40). Nevertheless, uNK cells have been implicated in triggering inflammation-induced fetal demise later in gestation (23). In contrast, uNK cells were not increased in ST-infected pregnant hosts. The rapid advanced pathology evoked by ST may have masked any early increase in uNK cells. Furthermore, the virulent nature of ST probably evoked a complex cellular and cytokine cascade involving multiple events. However, the decrease in uNK cells pursuant to infection in nonpregnant mice may have aided NK cell recruitment to other infected compartments. In contrast, the lack of change in uNK cell numbers in the pregnant-infected host may indicate their failure to redistribute to systemic tissue. It is then plausible that resident uNK cells in the pregnant-infected host responded to local infection and contributed to placental pathology and, by failing to migrate to the spleen, also led to breakdown of systemic resistance.

IL-12 is produced by monocyte lineage cells after recognition of pathogen-specific molecular patterns. Consistent with the reduced recruitment of dendritic cells and macrophages, infected-pregnant mice displayed reduced serum IL-12 amounts. This may be a consequence of the cross-talk between NK cells and dendritic cells (41), as IFN-γ produced by NK-cells can stimulate IL-12 production. Alternatively, the lack of dendritic cell activation may have contributed to weaker NK cell response as IL-12 can drive NK cell expansion and function. IL-12 plays an essential role in Salmonella pathogenesis: exogenous administration of IL-12 improves survival times in susceptible mice and anti-IL-12 Ab treatment exacerbates infection (42); immunity against oral Salmonella infection is associated with a rapid increase in IL-12p40 mRNA production in a Peyer’s patch (43). In humans, IL-12 deficiency increases susceptibility to ST infection (44). In contrast IL-12 is highly deleterious to pregnancy (45, 46). Interestingly, placental expression of IL-12p40 was also not increased, despite increased levels of other inflammatory cytokines. Furthermore, we noted that pregnant mice were hypersensitive to in vivo administration of IL-12 becoming acutely ill within a day (data not shown). Thus, it makes sense that ST-infected pregnant hosts may be down-modulating IL-12 production. Regulatory mechanisms may exist during pregnancy to prevent the deleterious expression of IL-12 in response to infection.

Despite the reduced activation of many innate cell types in the spleen of pregnant-infected hosts, at the uteroplacental interface, there was substantial inflammatory response characterized in particular with the overt up-regulation of IL-6. Indeed, IL-6 was also elevated in the serum of pregnant ST-infected mice. How such dichotomous inflammation (reduced systemic innate cell type activation and IL-6 increase) is regulated in the ST-infected pregnant host is unclear. IL-6 is a pleiotropic cytokine with contrasting effects on varied cell types and is also associated with suppressive immune effects (47). The function of IL-6 is controlled through soluble IL-6R trans-signaling (48). Interestingly, IL-6 promotes neutrophil apoptosis leading to suppression of neutrophil infiltration and resolution of acute inflammation while steering transition to acquired immunity (49). Neutrophils are the main source of early IFN-γ production critical for activation of NK cells. Furthermore, IL-6 also has complex differential effects on differentiation and activation of dendritic cells and macrophages (49, 50). Thus, the overt production of IL-6 may have lead to the deleterious consequence of lack of sufficient recruitment of inflammatory cell types in the systemic compartments of pregnant ST-infected hosts.

Interestingly, blocking IL-6 activity in vivo restored systemic host resistance, reiterating the inappropriate presence of this cytokine in pregnant ST-infected mice. A recent study demonstrated opposing effects of IL-6 and IL-12 on downstream signaling events, and ST survived in larger numbers in IL-6-treated macrophages (51). Thus, the elevated IL-6 levels may have facilitated unabated growth of ST in pregnant hosts. Nevertheless, while anti-IL-6 was effective in reducing bacterial burdens in the spleens of pregnant mice, it failed to do so in the placenta, and consequently did not prevent fetal demise. Considering the extremely high levels of placental IL-6 expression, it is possible that anti-IL-6 Ab failed to provide sufficient neutralization at the fetomaternal interface in a timely manner. Indeed, the phenomenal expansion in numbers of ST from 2 to 14 h (50 to >10⁵ bacteria) implies a doubling time of ∼1 h. Thus, the early innate immune response at the placental interface was not sufficient to curb bacterial replication. It is also possible that the absence of IL-12 expression by the placenta favored ST proliferation. Moreover, fetal loss may have been irreversible and rapid as placental carnage occurred before the host using IL-6 neutralization to its benefit.

Pregnancy poses a high risk against other intracellular infections such as malaria, listeriosis, and tuberculosis (31, 52, 53). However, despite adverse pregnancy outcome, normally, the host does not fatally succumb to infection. Indeed, even in areas of endemic infections, successful pregnancies occur. Moreover, there is no undue mortality in women of reproductive age. Thus, the maternal host is probably capable of effectively combating many infections, despite “altered” immune state. This raises the question as to why ST infection resulted in such dramatic consequences for the host. Pathogens differ in their virulence, replication rate, intracellular habitats, and nature of inflammation evoked, and these factors influence their interaction with the host. For example, Listeria proliferates rapidly but is localized in the cytoplasm allowing the host innate and adaptive immune response to rapidly control pathogen replication (54). Alternatively, mycobacteria reside within the phagosome and can avert host adaptive immunity, but its slow replication leads to chronic infection (55). In contrast, Salmonella not only proliferates rapidly, but it remains in a modified phagosome (4). It is also a highly virulent organism which devotes >4% of its genome to virulence mechanisms (56). Thus, even in the fully competent host, adaptive immunity to Salmonella is substantially delayed (8), making it important for innate immunity to control infection. Consequently, even a slightly deviated and/or delayed cytokine/cellular innate response during pregnancy may lead to catastrophic host outcome. The primary interaction of ST with the placental interface probably triggered a defective innate immune cascade leading ultimately to breakdown of systemic resistance. Our study reiterates the multifaceted cytokine and cellular regulatory network operative in pregnancy that highly virulent pathogens may use to deregulate host innate responses. The knowledge that some infections may evoke rapid maternal immunopathology and fatality bears implications for control of epidemic outbreaks.

We gratefully acknowledge the technical assistance provided by Renu Dudani and Ahmed Zafer. This is National Research Council Publication Number 42520.

The authors have no financial conflict of interest.