MALARIA-INDUCED MURINE PREGNANCY FAILURE: DISTINCT ROLES FOR IFN-γ AND TNF^,

Abstract

While an important role for excessive pro-inflammatory cytokines in compromise of pregnancy has been established, an immunological basis for malaria-induced fetal loss remains to be demonstrated. In this study, the roles of IFN-γ and TNF in Plasmodium chabaudi AS-induced fetal loss in mice were directly investigated. Pregnant IFN-γ null mutant mice experienced a more severe course of infection compared to intact C57BL/6 mice, characterized by high parasitemia, severe anemia and marked weight loss. However, fetal loss was delayed in these mice relative to intact controls. Because IFN-γ^−/− mice exhibited sustained levels of plasma TNF, the role of this cytokine was examined. Whereas splenic tnf expression in C57BL/6 mice was highest three days before peak parasitemia, increased placental expression relative to uninfected mice was sustained, indicating that locally produced TNF may be important in malaria-induced pregnancy failure. Indeed, antibody neutralization of TNF resulted in preservation of embryoes until day 12 of gestation, a time point at which all fetuses are lost in untreated mice. Histological analysis revealed that TNF ablation preserved placental architecture while placentae from untreated infected mice had widespread hemorrhage and placental disruption, with fibrin thrombi in some maternal blood sinusoids. Consistent with a role for cytokine-driven thrombosis in fetal loss, expression of pro-coagulant tissue factor was significantly increased in the placentae of infected C57BL/6 mice but was reduced in mice treated with anti-TNF antibody. Together, these results suggest that IFN-γ contributes to malaria-induced fetal loss, but TNF is a critical factor which acts by inducing placental coagulopathy.

Introduction

Despite a recent significant expansion of interest in placental malaria, which is characterized by the sequestration of cytoadherent Plasmodium falciparum in the maternal blood space of the human placenta and associated inflammatory cell infiltrate and tissue damage, the mechanisms that are central to malaria-induced poor birth outcomes remain poorly understood. In the context of highly endemic malaria, where a major adverse outcome for the fetus is low birth weight, the accumulation of maternal immune cells, as well as production of proinflammatory cytokines and chemokines in the placenta, are important features. The latter are thought to be derived from both maternal and fetal cells in the placenta (1–3). In contrast, P. falciparum infection in nonimmune pregnant women or during an epidemic has been shown to be more severe and can cause high rates of abortion, stillbirth and preterm labor (4). The immunologic basis for these outcomes is unknown. We have recently developed a mouse model to investigate the immunologic and molecular mechanisms involved in malaria-induced fetal loss (5). In this model, C57BL/6 (B6)³ mice infected at day 0 of pregnancy abort their fetuses at mid-gestation. Pregnancy loss occurs following high systemic production of proinflammatory cytokines, IFN-γ and IL-1β, and splenic production of TNF, together with high levels of soluble TNF receptor II (submitted for publication). High systemic production of IL-10, while protecting the mice against TNF-induced excessive weight loss and anemia (6), is apparently inadequate to block the deleterious, embryotoxic effects of these proinflammatory cytokines.

Production of IFN-γ during early stages of infection is essential for protection against primary P. chabaudi AS infection in B6 mice (7). IFN-γ, primarily produced by NK cells and T cells, is a pluripotent cytokine that has been shown to regulate over 200 genes in a wide variety of cells and tissues (8). During malarial infection, IFN-γ activates macrophages to produce TNF and other soluble mediators such as nitric oxide and reactive oxygen species (7). TNF, a multifunctional cytokine produced by macrophages, T and B cells and mast cells, is involved in immunoprotection against infection, but also in inflammation, autoimmunity and pathophysiology of many diseases (9). During malarial infection, TNF has been implicated in both protection and pathogenesis. During blood stage malaria infection in mice, this cytokine is associated with splenomegaly (10), weight loss, and anemia (11). In humans, excessive TNF is associated with cerebral malaria (12) and malarial fever (13); a lower IL-10 to TNF ratio in plasma is associated with anemia in children (14). In placental malaria, TNF is associated with a local inflammatory response and low birth weight (15, 16).

In pregnant rodents, small quantities of IFN-γ at appropriate locations are thought to be beneficial for normal pregnancy (17), and TNF is involved in normal embryonic growth and development (18). However, IFN-γ and TNF or TNF receptor null mutant mice can reproduce normally, suggesting that these cytokines may not be essential for successful pregnancy. Nonetheless, IFN-γ produced in excess can have an abortifacient effect (19). Aberrant production of TNF during pregnancy increases fetal resorptions in mice (20) and is linked to recurrent spontaneous abortion in humans (21). Despite these associations between elevated levels of proinflammatory cytokines and poor pregnancy outcome, the exact mechanism(s) by which fetal loss is induced remain unclear. Interestingly, inflammation and thrombosis are linked in many diseases (22) and in pregnancy loss (23). Inflammatory mediators such as TNF can induce the expression of tissue factor (TF), a key enzyme involved in initiation and propagation of thombus formation (24). Thus, it is noteworthy that TNF is elevated (15) and excessive fibrin deposition and accumulation of macrophages staining positive for TF are found in the maternal intervillous space of the P. falciparum-infected placenta (25). Furthermore, both human (26) and murine (27) malarial infections are characterized by a systemic pro-coagulant state and consumption of coagulation factors.

In this study, P. chabaudi AS infection in pregnant (IP) mice selectively depleted of TNF and in IFN-γ gene null mutant mice was employed to directly test the roles of TNF and IFN-γ in malaria-induced fetal loss. The results show that IFN-γ deficiency, which does not preclude TNF production, plays an incremental role in pregnancy success but is insufficient to rescue pregnancy in malaria-infected animals. On the other hand, neutralization of TNF results in mid-gestational pregnancy success that is indistinguishable from uninfected mice. Placentae of aborting mice have sustained tnf expression, and exhibit hemorrhage, fibrin thrombi formation and enhanced TF protein expression, but anti-TNF treatment results in reduced placental TF and preservation of placental architecture. Together these results suggest that TNF plays a pivotal role in malaria-induced placental pathology and fetal loss during malaria infection, potentially as a participant in dysregulated coagulation.

Material and Methods

Mice and parasites

C57BL/6 (B6) mice originally purchased from The Jackson Laboratory, Bar Harbor, ME and IFN-γ null mutant (IFN-γ^−/−) mice (B6.129S7-Ifng^tm1Ts) obtained from Dr. Rick Tarleton (University of Georgia) were maintained and bred by brother-sister pairing for a maximum of 10 generations at the University of Georgia Animal Resources facility in accordance with the guidelines of the University of Georgia Institutional Animal Care and Use Committee. Eight to ten week old female mice were used for all the experiments. P. chabaudi AS originally obtained from Dr. Mary M. Stevenson (McGill University and the Montreal General Hospital Research Institute, Quebec, Canada) was maintained as described previously (5) and was used for all experiments.

Experimental design

B6 or IFN-γ^−/− female (infected, pregnant; IP) mice were infected i.v. on gestation day 0 with 1000 P. chabaudi AS infected RBCs (iRBCs) per 20 grams of body weight. Infected, non-pregnant (INP) mice, and sham-injected, uninfected, pregnant (UP) mice were used as infection and pregnancy controls, respectively. Mice were sacrificed on gestation day or experiment day (ED) 6, 8, 9, 10, 11 and 12 to assess pregnancy outcome and development of immune responses. Parasitemia, hematocrit and body weight were monitored as described previously (5).

Anti-TNF antibody treatment

Infected B6 mice and UP controls were injected i.p. with 100 µg of anti-TNF mAb (clone MPG-XT3, Upstate, Lake Placid, NY, USA) or with rat IgG (Sigma) as a control on ED 6, 8, 9, 10 and 11. Mice were sacrificed on ED 12 or immediately with evidence of abortion (bloody mucoid discharge from the vagina) (5).

Cytokine ELISA

Plasma IFN-γ, TNF and IL-10 levels were determined using OptEIA ELISA kits (BD Pharmingen, San Diego, CA) or DuoSets (R&D Systems, Minneapolis, MN) according to manufacturer’s instructions. Limits of detection were 8 pg/ml for Il-10 and TNF and 15 pg/ml for IFN-γ.

Quantitative real-time RT-PCR

Total RNA was isolated from fetoplacental units and spleens using the RNeasy kit (Qiagen, Valencia, CA, USA) following the manufacturer’s protocol and stored at −85°C. Contaminating genomic DNA was removed by digesting with RNAse (RQ1 RNAse free DNAse, Promega Madison, WI, USA) as recommended by the manufacturer. First strand cDNA was synthesized from 1 µg of obtained total RNA using the Improm-II Reverse Transcription System (Promega). Quantitative, real-time RT-PCR was carried out using specific primers for tissue factor (tf) (forward: 5’tcagttcatggagacggagac-3’ and reverse: 5’-ggttgtgtctcggtaaggtaa-3’), tumor necrosis factor (tnf) (forward: 5’-gtaacccgttgaacccatt-3’ and reverse: 5’-cacttggtggtttgctacgac-3’), and 18s RNA (forward: 5’-gtaacccgttgaaccatt-3’ and reverse: 5’-ccatccaatcggtagtagcg-3’) (MWG-Biotech Inc., High Point, NC, USA). All primers were validated for use in comparative real-time PCR using the ABI Prism 7500 thermocycler and analyzed with the Sequence Detection System (SDS) software (Applied Biosystems, Foster City, CA, USA). No template and no reverse-transcription controls were included to verify absence of genomic DNA contamination. The ΔΔCt method of analysis was used with the 18s RNA as normalizing gene and cDNA from UP mice as the calibrator. Results are expressed as fold increase over UP controls.

Histology and immunohistochemistry

Uteri were harvested on ED 9, 10, 11, and 12 and fixed in buffered formalin for 48 h. Tissues were subsequently paraffin embedded and processed for histology and immunohistochemistry (IHC) studies. H&E stained and unstained placental sections (5 µm thick) from IP and UP mice were prepared and indirect immunolocalization of TF was performed using unstained sections. The rabbit ABC staining system (Santa Cruz Biotechnology, Inc, Santa Cruz, CA, USA) was used for IHC according to the manufacturer’s instructions. Briefly, following deparaffinization, rehydration, and unmasking in a pressure cooker with 1X Declere solution (Cell Marque, Rocklin, CA) for 5 min, sections were blocked for nonspecific binding with goat serum and then incubated with rabbit anti-TF antibody (American Diagnostica Inc, Stamford, CT) at a 1:100 dilution overnight at 4°C. Sections were then incubated with the biotinylated secondary antibody and the amplification system, with ultimate target detection using diamino-benzidine (DAB) chromogen. Non-specific rabbit IgG (1:100) was used as a primary antibody negative control. Sections were counterstained with hematoxylin (Vector Laboratories, Inc, Burlingame, CA) and mounted with Flo-Texx (Lerner Laboratories, Pittsburg, PA). A semi-quantitative method was used to score the TF immunostaining. Sections were evaluated independently by two authors (D.S. and T.N.) and assigned values 0 to 4, where 0 designates negative; 1, weak and/or diffuse staining ; 2, moderate staining; 3, strong, focal staining; and 4, strong, diffuse staining.

Statistical Analysis

Data analysis was performed using SAS statistical software package (version 8.02; SAS Institute, Inc., Cary, NC). The significance of differences among group means in the case of normally distributed data was determined using Proc GLM or Mann-Whitney U test. Tukey’s Studentized (HSD) range test was used to perform multiple pairwise group comparisons. In cases of non-normally distributed data, analysis was done with the nonparametric Wilcoxon rank sum test, with proc multtest for adjustment of P values in multiple pairwise comparisons. P values of ≤0.05 were considered to be significant.

Results

P. chabaudi AS-infected IFN-γ ^−/− mice experience a more severe course of infection

P. chabaudi AS infection in B6 mice results in significant increases in the levels of IFN-γ in the plasma, spleen and placental cell culture supernatants from IP mice relative to UP mice during ascending and peak parasitemia, corresponding to mid-gestational fetal loss in this model (5) (submitted for publication). To directly assess the role of IFN-γ in the observed fetal loss, IFN-γ^−/− mice were mated and infected with P. chabaudi AS. Consistent with a previous report (7), the development of parasitemia in null mutant INP (data not shown) and IP mice (Fig. 1) was accelerated relative to intact B6 mice. Parasitemia peaked in IFN-γ^−/− IP mice on ED 10, one day earlier than B6 mice, and was significantly higher on ED 8 through 10 (P ≤0.014). Although parasitemia in IP and INP IFN-γ^−/− mice dropped to less than 6% on ED 14, none of the mice survived beyond ED 15 (data not shown).

Figure 1. Course of P. chabaudi AS infection in IFN-γ^−/− mice.

Percent parasitemia, hematocrit and body weight of null mutant and control mice are shown. Mice were sacrificed at ED 6, 8, 9, 10, 11, and 12 and data pooled from two serial sacrifice studies. Starting sample sizes: UP IFN-γ^−/−, n = 23 and IP IFN-γ^−/−, n = 28. Intact B6 mice were sacrificed at ED 12 or at abortion; sample sizes, UP, n = 8; IP, n = 7. Data represent mean ± SEM. The Y axis on weight panels begins at 15 g to avoid compression and poor visualization of the data. Statistical differences (all P < 0.05): *IFN-γ^−/− versus B6; ^†IFN-γ^−/− IP versus all groups; ^§B6 UP versus B6 IP; ^¶UP versus IP; ^‖IFN-γ^−/− UP versus IFN-γ^−/− IP.

It has been proposed that mortality in IFN-γ^−/− mice results from failure to recover from malaria-induced anemia (7). In parallel with accelerated parasitemia, IP IFN-γ^−/− mice developed anemia rapidly and exhibited significantly lower hematocrit on ED 10 compared to UP IFN-γ^−/− and IP and UP B6 mice (P<0.05; Fig. 1), as well as INP mice (data not shown). Hematocrit continued to decrease to its nadir (nearly 20% of the normal level) on ED 11, but at this time point was not significantly different between IFN-γ^−/− and B6 IP mice. Whereas B6 mice, regardless of pregnancy, recover from anemia during resolution of infection (5), IFN-γ^−/− mice failed to regain hematocrit even after experiencing a decline in parasitemia levels (Fig.1).

Both intact B6 (5) and IFN-γ^−/− IP mice initially gained weight as their pregnancies developed (Fig. 1). However, IFN-γ^−/− IP weight plateaued at ED 7 and 8, and on ED 10, when parasitemia peaked, dropped below starting weight, one day before a similar drop in intact B6 mice. Infected B6 mice regained weight as they started clearing parasitemia and recovering from anemia (Fig. 1), but IP IFN-γ^−/− mice exhibited a downward trend in body weight (Fig. 1) until they succumbed to infection (data not shown).

Pregnancy outcome in P.chabaudi AS-infected IFN-γ^−/− mice

Contrary to expectation, IP IFN-γ^−/− mice, like intact B6 mice, experienced malaria-associated failure of pregnancy (Table I and Table II). B6 IP mice resorbed 79% of embryos by ED 11, but only 11% were inviable in IP null mutant mice, a level that was indistinguishable from UP IFN-γ^−/− mice (Table II). On ED 12, a time point at which B6 mice had no remaining viable embryoes (5), 60% were still viable in IFN-γ^−/− mice (Table I). These results suggest that while IFN-γ may contribute to malaria-induced pregnancy loss, other factors are likely to be involved.

Table I.

Fetal loss is delayed in infected, pregnant IFN-γ^−/− mice^a

ED	IP	Nc	UP	Nc	Pd
	Resorptionsb		Resorptionsb
6	0/44	6	0/29	5	NA
8	0/48	6	0/30	4	NA
9	0/33	4	0/40	5	NA
10	6/52	6	0/22	3	0.17
11	4/34	4	1/24	3	0.39
12	39/65	9	0/19	3	0.0001

^aMice were sacrificed on the EDs indicated to assess pregnancy success. Resorptions were scored by examination of uteri under a dissection microscope.

^bTotal number of resorptions per total number of fetuses

^cNumber of mice per group.

^dFisher’s exact test. NA = not applicable.

Table II.

Pregnancy outcome in IFN-γ^−/− mice and B6 mice treated with neutralizing antibodies to TNF^a

Infection or treatment group	N	Mean # of viable/ total fetus	% Resorption	P value
IP IFN-γ−/−	4	8/9	12	0.003*
IP B6	7	1/7	79	0.0065#
UP IFN-γ−/−	3	8/8	4
UP B6	8	6/6	1
IP IgG (at abortion)	5	0/4	100	0.0051**
IP anti-TNF (ED 12)	5	7/8	15	NS†
UP IgG (ED12)	4	7/8	13
UP anti-TNF (ED 12)	3	7/9	15

^aTo assess pregnancy success, IFN-γ^−/− and intact B6 mice, including those treated with rat IgG, were sacrificed on ED 11. Anti-TNF-treated mice and IgG-treated UP mice were sacrificed on ED 12. Resorptions were scored by examination of uteri under a dissection microscope. Statistical comparison of resorption rates

^*IP IFN-γ^−/− versus IP B6

^#IP B6 versus UP B6

^**IP IgG versus IP anti-TNF

^†IP anti-TNF versus UP IgG and UP anti-TNF.

Systemic TNF expression in IP IFN-γ^−/− mice

Previous studies have shown that virgin, P. chabaudi AS-infected IFN-γ^−/− mice are capable of producing TNF, an abortifacient cytokine, although in smaller amounts compared to intact mice (7). Consistent with that report, IP IFN-γ^−/− mice exhibited high, sustained plasma levels of TNF (ED 8 through 12) relative to UP and INP IFN-γ^−/− mice (Fig. 2a). To assess whether the uncontrolled production of TNF was related to a defect in immunoregulatory activity (28), plasma IL-10 in IP and UP IFN-γ^−/− mice was measured. INP IFN-γ^−/− mice exhibited sustained high plasma IL-10 levels corresponding to ascending and peak parasitemia, but IL-10 production in IP IFN-γ^−/− mice peaked early during infection and by ED 12 was undetectable (Fig. 2b). UP IFN-γ^−/− mice produced little to no IL-10.

Figure 2. Systemic TNF and IL-10 levels in IFN-γ^−/− mice.

IFN-γ^−/− IP, INP, and UP mice were sacrificed at indicated time points and plasma samples were assayed for A: IL-10 and B: TNF by ELISA. Box plots show interquartile range with median. Number of mice sacrificed (IP, UP, and INP, respectively) at each ED was as follows: at ED 6, n = 5, 3, and 4; at ED 8, n = 4, 4, and 3; at ED 9, n = 4, 5, and 4; at ED 10, n = 5, 3, and 3; at ED 11, n = 4, 4, and 3; at ED 12, n = 6, 3, and 6. *P <0.05, **P < 0.025.

Ablation of TNF rescues pregnancy

Because pregnancy loss in IFN-γ^−/− IP mice occurred following sustained systemic TNF production, the role of this cytokine was directly assessed. IP B6 mice were treated with neutralizing TNF mAb injections five times during ascending and peak parasitemia. The treatment had no effect on development or resolution of parasitemia and did not impact weight in IP or control UP mice (Fig. 3). However, in comparison to the control IgG-treated IP group, anti-TNF-treated IP mice regained body weight on ED 11. In addition, pregnancy outcome was substantially improved in TNF-ablated IP mice (Table II). While all IgG-treated IP mice aborted or resorbed their embryos by ED 11, the anti-TNF IP group had only 15% resorptions on ED 12, which was comparable to the rate of spontaneous loss observed in the UP anti-TNF (15%) and IgG (13%) groups. Because TNF was not detected in the plasma from mice treated with neutralizing antibodies (data not shown), the results suggest that TNF is pivotal in malaria-induced fetal loss.

Figure 3. Course of P. chabaudi AS infection in B6 mice treated with anti-TNF mAb.

Percent parasitemia, hematocrit and body weight of anti-TNF mAb and control IgG mice are shown. Starting sample sizes: anti-TNF and IgG IP groups, n = 5; UP IgG group, n = 4; UP anti-TNF group, n = 3. Data represent mean ± SEM. The Y axis on weight panels begins at 15 g to avoid compression and poor visualization of the data. Statistical differences (all P < 0.05): *UP versus IP; ^†UP anti-TNF versus IP; ^§all UP versus IP anti-TNF; ^¶UP IgG versus IP IgG; ^‖all UP versus IP IgG; ^#IP IgG versus all anti-TNF.

To study the effect of anti-TNF treatment on production of other critical cytokines, plasma levels of IFN-γ and IL-10 were measured at the time of sacrifice (ED 11 for IP IgG and ED 12 for other groups) (Fig. 4). Both TNF ablated and control IgG-treated IP mice had robust IFN-γ and IL-10 responses in the plasma, in contrast to UP mice, in whom these cytokines were very low or undetectable.

Figure 4. Production of IL10 and IFN-γ by anti-TNF and control antibody-treated mice.

Infections and anti-TNF or control IgG treatments were performed as described in materials and methods. IP IgG mice were sacrificed at the time of abortion (ED 11). IP anti-TNF, UP anti-TNF, and UP IgG mice were sacrificed on ED 12. Box plots show interquartile range with median. Number of mice sacrificed for each group was as follows: IP IgG, n = 5; IP anti-TNF, n = 5; UP IgG, n = 4 and UP anti-TNF, n = 5. *P ≤ 0.020, **P ≤ 0.004.

TNF gene expression is sustained in the placenta

Given the apparent role of TNF in mediating fetal loss in infected mice, the major sources of this cytokine are of interest. We have recently shown that isolated primary murine trophoblasts exposed to iRBCs in vitro secrete TNF (submitted for publication), suggesting that the placenta itself may be an important contributor of this fetotoxic cytokine. Indeed, tnf expression, although very high in the IP spleen on ED 8, decreased relative to UP mice in the days preceding fetal loss, while expression in the placenta, albeit not as intense as in the spleen, was sustained from ED 8 to ED 10 (Table III). Furthermore, aborting mice tending to have higher levels of placental tnf expression than non-aborting mice on ED 10 and 11 (EDs combined (mean ± SEM): 6.6 ± 1.3, aborting versus 2.6 ± 0.6, non-aborting, P = 0.06).

Table III.

TNF gene expression in B6 spleen and placenta during ascending parasitemia^a

	ED 8	ED 9	ED 10
Spleen (n)	8.2 ± 0.4 (5)	3.1 ± 0.3 (6)	1.4 ± 0.8 (5)
Placenta (n)	2.8 ± 0.6 (5)	2.4 ± 0.7 (5)	2.8 ± 0.2 (7)

^aRelative quantity of TNF gene expression in the placenta and the spleen of Plasmodium chabaudi AS-infected B6 mice at different time points was determined by RT-PCR. Placenta and spleen were harvested at sacrifice on the indicated EDs and processed for RNA. The 18S rRNA gene served as the reference gene. The results are expressed as fold increase (mean ± SEM) over the placenta and the spleen of UP mice. Sample size indicated in parentheses.

Placenta sections from aborting mice exhibit massive hemorrhage and placental disruption in association with upregulated TF

To assess the pathological basis of malaria-induced, TNF-associated pregnancy loss in P. chabaudi AS-infected mice, histological examinations were performed. Relative to UP mouse placenta on ED 11 (Fig. 5a), IP mice undergoing abortion had widespread placental hemorrhage, thinning of the labyrinth (where critical maternofetal nutrient and gas exchange occur), and generalized disruption of placental architecture (Fig. 5b). Remarkably, placenta from ED 12 TNF-ablated mice exhibited normal architecture (Fig. 5c), consistent with the high survival rate of embryoes in these mice (Table II). In addition to substantially altered placental architecture, some maternal blood sinusoids in the IP B6 placenta contained fibrin thrombi (Fig. 5d). Given the previous observation that TF expression is upregulated in the malaria-infected human placenta (25), and the known positive relationship between TNF and TF in pathologic placenta (29), the expression of this potent procoagulant was assessed (Table IV). The tf mRNA was increased on EDs 9 through 11 in IP mice relative to control UP mice, and was slightly higher in aborting compared to non-aborting mice at these time points (ED 10 and 11 combined (mean ± SEM): 3.2 ± 0.7, aborting (n=10) versus 2.4 ± 0.3, non-aborting (n=7), P > 0.05). To verify that changes in tf gene expression resulted in changes in TF protein expression in the placenta, IHC in uteri harvested from ED 9 to ED 12 was performed. Significant immunoreactivity was found in the IP B6 placenta on ED 10 (Fig. 5g) and 11 (Table V), especially within mononuclear trophoblast cells surrounding maternal blood sinusoids (Fig. 5g). Although most ED 9 placentae did not exhibit strong TF staining (Table V), placenta from one IP mouse that aborted earlier than is typical (ED 9) had very strong staining in mononuclear trophoblast cells (staining score: 3.0; Fig. 5f). Staining of syncytiotrophoblast and trophoblast giant cells, which, unlike mononuclear trophoblast cells, are not in contact with maternal blood, was not observed (Fig. 5f and not shown). Occasional TF+ monocytes were also detected in the maternal placental blood (Fig. 5f). No reactivity was seen in sections stained with control antibody (Fig. 5e). TF immunoreactivity of cells in contact with maternal blood suggests that TF activation in these areas of the placenta may be responsible for the thrombosis observed in the placentae of mice infected with P. chabaudi AS. In this context, it is noteworthy that cells expressing TF were commonly associated with tissue disruption (not shown). Semi-quantitative scoring for TF protein expression revealed significantly higher levels in placenta sections from IP mice on ED 10 and 11 compared to UP mice (Table V; P < 0.0054). Additionally, a significantly higher proportion (76%) of IP mice aborting at ED 10 and 11 expressed TF with a score of 1 or above compared to non-aborting IP mice (24%; P = 0.016). To identify the role played by TNF in placental TF expression during malarial infection, placental sections from mice treated with anti-TNF mAb were also examined by IHC. Placentae from TNF-ablated IP mice at ED 12 exhibited significantly reduced TF staining (Fig. 5i) compared to IP mice treated with rat IgG (Table V, P = 0.0021), and was not different from staining seen in TNF-ablated (Fig. 5h) and IgG-treated UP mice (Table V, both P > 0.05). This result suggests a close relationship between TNF and TF expression in the placenta during malarial infection.

Figure 5. Histopathological and immunohistochemical analysis of the P. chabaudi AS-infected mouse placenta.

A–D. H&E staining. E–I. Immunolocalization of tissue factor (TF) by IHC. Intense brown staining indicates specific immunolocalization of TF, with blue counterstaining with hematoxylin. A. Normal placenta at ED 11 shows three layers: the uterine wall and decidua (u), junctional zone (j; containing spongiotrophoblast), and labyrinth (l). The embryo is indicated as “e”. B. IP mouse placenta at ED 11. Reduced thickness of the labyrinth, extensive hemorrhage in placental tissue (arrows) and disruption of the embryo (e) are evident. C. Anti-TNF antibody-treated mouse placenta at ED 12. Note the normal appearance of the placental architecture. D. Fibrin thrombi (arrows) in maternal blood sinusoids of the junctional zone in an ED 11 IP mouse. E. Section of placenta from the junctional zone/labyrinth interface in a mouse aborting at ED 9 stained with control IgG. F. Section from same location as shown in E, exhibiting strong expression of TF (brown color; arrows) on mononuclear cytotrophoblast (mt) surrounding a maternal blood sinusoid (ms). Syncytiotrophoblasts (st) are not stained. A maternal monocyte (arrowhead) in the maternal blood also stains intensely for TF. G. Intense TF staining in mononuclear cytotrophoblast at ED 10 between the labyrinth and the junctional zone, with spongiotrophoblast (sp). H. Placenta section from the labyrinth/junctional zone interface from an ED 12 UP mouse treated with TNF neutralizing antibody stained with anti-TF antibodies. Trophoblast giant cells (tgc) and other cell types do not express TF. I. Minimal placental TF-specific staining at the labyrinth/junctional zone interface in an ED 12 IP mouse treated with TNF neutralizing antibody. Actual size represented by image: A, B, C: 3330 µm; D: 250 µm; E, F: 247 µm; G, 116 µm; H, I: 200 µm.

Table IV.

TF gene expression in B6 placenta preceding and at the time of abortion^a

	ED 9	ED 10	ED 11
tf expression (n)	2.3 ± 1.4 (5)	2.8 ± 0.2 (7)	2.9 ± 0.6 (10)

^aRelative quantity of tf expression in the placentae of Plasmodium chabaudi AS-infected B6 mice at different time points was determined by RT-PCR. Placentae were harvested at sacrifice on the indicated EDs and processed for RNA. The 18S rRNA gene served as the reference gene. The results are expressed as fold increase (mean ± SEM) over the placentae of UP mice.

Table V.

Semi-quantitative scoring of immunhistochemical staining of tissue factor in infected and uninfected placenta^a

Infection / Treatment Group	ED (n)	TF Score
IP	9 (5)	1.0 ± 0.7
IP	10 (7)	3.2 ±0.3*
IP	11 (8)	3.0 ± 0.3†
UP	9 (5)	0.0 ± 0.0
UP	10 (5)	0.5 ± 0.3
UP	11 (5)	0.8 ± 0.3
UP	12 (5)	1.0 ± 0.6
IP IgG	11 (5)	3.5 ± 0.5§
IP anti-TNF	12 (5)	1.3 ± 0.5
UP IgG	12 (4)	0.8 ± 0.5
UP anti-TNF	12 (3)	0.3 ± 0.3

^aMice were sacrificed on the EDs shown to semi-quantitatively score TF staining intensity within the placental tissue of B6 mice infected with P. chabaudi AS. Note that all untreated or IgG-treated IP mice aborted by ED 11 and thus are not represented on ED 12. Mean ± SEM for each group is shown.

^*IP versus UP, P = 0.0044

^†IP versus UP, P = 0.0054

^§IP IgG versus IP anti-TNF and UP IgG, P = 0.021.

Discussion

Poor birth outcomes in areas highly endemic for malaria are associated with P. falciparum sequestration in the placenta, local production of pro-inflammatory cytokines, and significant placental pathology (30). Well-described, characteristic pathological features include excessive perivillous fibrin deposition, trophoblast basement membrane thickening, syncytiotrophoblast necrosis, and impaired uteroplacental blood flow (reviewed in ref. (31)). Pathology associated with placental malaria in low endemic areas, on the other hand, has received relatively little attention (32), and the high rates of fetal loss associated with malaria epidemics in naïve populations (33) remain unexplained. Furthermore, the mechanisms by which malaria induces early gestational pregnancy loss are poorly understood. Because experimental studies of severe malaria infection during human pregnancy are not possible due to ethical and logistical constraints, we established a mouse model to explore mechanisms of malaria-induced fetal compromise (5). In this model, P. chabaudi AS infection during early pregnancy invariably results in abortion and fetal resorption in naïve B6 mice (5). Mid-gestational fetal loss is associated with increased systemic, splenic and placental production of IFN-γ TNF, and IL-1β, together with high levels of soluble TNF receptor II (submitted for publication). In the present study, the relationship between inflammatory cytokines and coagulation during placental malaria and their influence on pregnancy outcome are investigated for the first time. The results suggest that pro-inflammatory cytokine responses induced by malaria lead to activation of coagulation in the placenta, which compromises placental function and ultimately contributes to death of the embryo.

B6 mice, regardless of pregnancy, have robust cytokine responses and are able to clear infection with P. chabaudi AS (5) (submitted for publication). In the absence of IFN-γ, a pivotal cytokine in the response to this parasite (7), IP mice experienced significantly higher parasitemia, severe anemia and marked weight loss compared to intact mice and succumbed to infection by ED 15. This is consistent with studies demonstrating that depletion of IFN-γ, and IFN-γ receptor deficiency, are associated with prolonged acute phase parasitemia and greater mortality due to P. chabaudi AS infection (7, 34, 35). Moreover, IFN-γ production is correlated with protection against P. falciparum infection (36), including at the placental level (37). Despite increased disease severity, however, pregnancy outcome in IFN-γ^−/− mice was improved relative to intact B6 mice. At ED 12, 40% of embryos were viable in null mutant mice, in contrast to IP B6 mice, in which all embryos by this time are resorbed or actively expelled (5). The known cytotoxic effect of IFN-γ on fetal trophoblasts (38) could explain this differential outcome. However, IFN-γ^−/− mice ultimately experienced profound malaria-induced pregnancy loss, with all conceptuses being nonviable by ED 15, which suggested that another factor is sufficient to induce pregnancy loss in these mice. Indeed, from ED 8 to 12, IP IFN-γ^−/− mice had elevated systemic levels of TNF, a known placentotoxic factor (38). While IFN-γ is a major inducer of TNF production, malarial parasites and their byproducts such as insoluble hemozoin (39) can directly activate splenic monocytes or macrophages to produce TNF (40). While additional studies will be required to determine the specific stimuli and cell sources of TNF in IFN-γ^−/− mice, the present study shows that in intact B6 mice the spleen is a major site of TNF gene expression, with placentae contributing relatively less, but for a longer period of time. It is of interest that elevated TNF mRNA expression in the B6 spleen (and liver) early during infection correlates with resistance to Plasmodium chabaudi AS infection, whereas delayed TNF gene expression in the liver and excessive levels of TNF protein in serum later during infection correlate with susceptibility (41). Thus, the impact of TNF on outcome of malarial infection depends on the timing, magnitude and site of its expression. It will be important in future studies using this model to assess the specific cell types producing TNF and to establish the relative contributions of the spleen and placenta, and their roles in outcome of infection.

Consistent with a central role for TNF in malaria-induced fetal loss, antibody neutralization of this cytokine in intact P. chabaudi AS-infected B6 mice conferred significant improvement in pregnancy outcome. Whereas B6 mice typically lose weight and abort or resorb all embryos by ED 12, mAb-treated mice maintained their weights and resorbed only 15% by this day, a rate indistinguishable from UP mice. This result was not unexpected, given the potent abortifacient effect of TNF. Exogenous administration of TNF to P. vinckei–infected mice induced abortion (42), and blockade of this cytokine prevented fetal loss in a murine model of stress-induced abortion (43).

Production of IL-10 is critical for control of TNF production in P. chabaudi AS infection (6). Thus, it is noteworthy that IL-10 production was suppressed in IFN-γ^−/− mice; its appearance early during infection but subsequent decline may contribute to TNF persistence in these mice. In infected intact B6 mice, IL-10 is produced at high levels systemically, but is absent at the placental level (submitted for publication). During normal pregnancy, IL-10 is thought to play an important role in preserving the fetal allograft through suppression of abortifacient cytokines such as TNF and IFN-γ (44). Thus, poor uteroplacental IL-10 responses, in the context of sustained TNF production, may contribute to compromise of pregnancy in P. chabaudi AS-infected mice.

While the present results provide compelling evidence that TNF is critical for malaria-induced fetal loss, the exact mechanism by which this cytokine compromises pregnancy is not clear. One possibility is that TNF acts directly on the trophoblast through TNF receptors, initiating an apoptotic program, ultimately leading to profound destruction of placental tissue. We are currently conducting studies with TNFRI/II^−/− mice to examine this possibility. Alternatively, TNF induces thrombosis in the placenta through upregulation of procoagulants such as TF (29). TF is the primary cellular initiator of blood coagulation (45). It acts by binding to coagulation factor VIIa and ultimately yields thrombin, which converts fibrinogen to fibrin. Control of TF expression and activity is critical for successful pregnancy as evidenced by profound pregnancy loss in thrombomodulin-deficient mice (46). In these mice, uncontrolled TF expression, leading to activation of protease-activated receptors on trophoblast by thrombin, and formation of fibrin, and subsequently fibrin degradation products, directly contribute to trophoblast growth arrest and death (46). Upregulated TF also compromises murine pregnancy by causing coagulation-induced severe ischemic injury in the placenta (23). TF, together with TNF, is implicated in the pathogenesis of human pre-eclampsia (29), a condition characterized by profound placental dysfunction and infant low birth weight. Thus, a pathogenic cycle of pro-inflammatory cytokines and TF activation, recently suggested to underlie the pathogenesis of cerebral malaria (47), may be involved in malaria-induced pregnancy loss. There is ample evidence that coagulation is hyperactivated in human malaria (26), and has also been reported in non-pregnant P. yoelli-infected mice (27). Consistent with this, placentae from aborting IP mice had evidence of fibrin thrombus formation in maternal placental blood sinusoids, and IHC revealed significantly increased TF expression on trophoblast cells surrounding these sinusoids. TF is constitutively expressed by human syncytiotrotrophoblast cells and placental perivascular cells (48) and is upregulated in response to TNF (29). The significantly reduced placental TF expression by trophoblasts in contact with maternal blood in anti-TNF treated mice confirms this relationship and indicates that TNF induced in response to P. chabaudi AS infection is a likely trigger for enhanced trophoblast expression of TF. Although TF is also expressed by activated monocytes, as was demonstrated in human placental malaria, together with excessive perivillous fibrin deposition (25), monocytes are unlikely to be major sources of placental TF in this model, since they do not accumulate in the placentae of P. chabaudi–infected B6 mice (J. Poovassery et al., unpublished data; Fig. 5f). However, in addition to TF expression by trophoblast and/or monocytes in the malaria-infected placenta, it is noteworthy that the membrane of P. falciparum-infected erythrocytes has surface-exposed phosphatidylserine, an important co-factor for clot formation (49). Accumulation of iRBCs in the maternal intervillous space, which is a characteristic feature of P. falciparum infection during pregnancy in humans and in P. chabaudi AS- (5) and P. berghei-infected rodents (50), could therefore also contribute to placental coagulopathy in both mice and humans. Finally, it remains to be determined to what extent coagulation is perturbed in the periphery of IP mice, as was shown in P. yoelli-infected mice (27), and the importance of this coagulation for pregnancy outcome. Making this determination will be informative and potentially clinically relevant, particularly for human placental malaria, definitive diagnosis of which is not possible until after delivery. Moreover, the ability of TNF ablation to block peripheral coagulation has not been studied, and could prove to be an important intervention for prevention of adverse outcomes in malaria (13).

Considered all together, the present findings suggest that in P. chabaudi AS-infected mice, TNF-exposed trophoblasts upregulate TF, and, perhaps together with accumulated iRBCs, galvanize a local coagulation cascade in the maternal blood spaces of the placenta. The resultant deposition of a fibrin meshwork (23) promotes local ischemia and disruption of placental function, including nutrient and gas exchange, together culminating in tissue destruction, hemorrhage, and fetal death. Alternatively, both TNF and coagulation factors and their byproducts may directly induce trophoblast cell death (46), and, ultimately, placental failure and fetal death. Definitive establishment of coagulopathy as a major player in malaria-associated fetal compromise awaits further detailed study; we are currently endeavoring to finely characterize the relationships between coagulation and inflammatory placental malaria in studies of naturally exposed pregnant women, in an in vitro trophoblast model system (3, 51), and in continuing work with the mouse model described here. Nonetheless, these data provide compelling evidence that TF upregulation and excessive fibrin deposition in human placental malaria may be underappreciated as critical players in the tissue damage that culminates in intrauterine growth restriction (31) or pre-term labor (52). The recent demonstration that ablation of coagulation by pharmacologically blocking TF in abortion-prone mice can significantly improve pregnancy outcome (23) raises the possibility that such a therapeutic approach could be a novel means for reducing the risk for low birth weight in association with placental malaria.

To conclude, these findings suggest that malarial infection during pregnancy induces cytokine-driven coagulopathy in the placenta, contributing to damage and dysfunction of the placenta and embryo. Documenting the mechanistic basis of placental damage by pro-inflammatory cytokines and coagulation will provide critical insights into the pathogenesis of malaria-induced compromise of pregnancy and ultimately contribute to the development of new strategies for prevention of poor birth outcomes and fetal loss induced by placental malaria.