TNF- Is a Critical Effector and a Target for Therapy in Antiphospholipid Antibody-Induced Pregnancy Loss
http://www.100md.com
免疫学杂志 2005年第1期
Abstract
The antiphospholipid syndrome (APS) is characterized by recurrent fetal loss, intrauterine growth restriction, and vascular thrombosis in the presence of antiphospholipid (aPL) Abs. Our studies in a murine model of APS induced by passive transfer of human aPL Abs have shown that activation of complement and recruitment of neutrophils into decidua are required for fetal loss, and emphasize the importance of inflammation in aPL Ab-induced pregnancy loss. In this study, we examine the role of TNF- in pregnancy complications associated with aPL Abs in a murine model of APS. We show that aPL Abs are specifically targeted to decidual tissue and cause a rapid increase in decidual and systemic TNF- levels. We identify the release of TNF- as a critical intermediate that acts downstream of C5 activation, based on the fetal protective effects of TNF- deficiency and TNF blockade and on the absence of increased TNF- levels in C5-deficient mice treated with aPL Abs. Our results suggest that TNF- links pathogenic aPL Abs to fetal damage and identify TNF blockade as a potential therapy for the pregnancy complications of APS.
Introduction
The antiphospholipid syndrome (APS)3 is characterized by arterial and venous thromboses and pregnancy loss that occur in the presence of antiphospholipid (aPL) Abs (1). APS is common in patients with systemic lupus erythematous, and it is also found in individuals without other autoimmune features. Although it is clear that the specific antigenic reactivity of aPL Abs and their targeting to decidual tissue are critical to their effect, the pathogenic mechanisms that result in fetal injury in vivo are incompletely understood.
Therapy for pregnant women with APS is focused on preventing thrombosis, but anticoagulation is only partially successful in averting miscarriage and carries risks for fetus and mother. Although experimental models have emphasized the role of thrombosis in placental tissue, histopathologic findings in placentas from women with APS argue that proinflammatory factors may contribute to injury (2, 3). Our recent studies showing that activation of complement and recruitment of neutrophils into decidua are required for fetal loss in a murine model of APS underscore the importance of inflammation in aPL Ab-induced pregnancy loss (4). However, the specific effectors of fetal injury remain unknown. Because TNF- is released when infiltrating inflammatory cells are activated by complement split products and elevated levels of TNF- have been associated with pregnancy failures, we considered the possibility that TNF- may have a role in pregnancy complications associated with APS.
Evidence from human pregnancy studies points to a strong association between maternal Th2-type immunity and successful pregnancy, whereas Th1-type immune reactivity is associated with pregnancy loss (5). Mitogen-stimulated PBMC, harvested at the end of the first trimester, have consistently been shown to produce higher levels of IL-4, IL-5, IL-6, and IL-10 in patients who went on to have normal pregnancies than in those who had recurrent spontaneous miscarriages. In contrast, levels of TNF-, IL-2, and IFN- were uniformly higher in patients who subsequently had miscarriages (6, 7). Indeed, the preponderance of Th2 cytokines relative to Th1 cytokines in the local milieu of the fetus is considered to be essential for its survival. The proinflammatory Th1-dominant response that underlies clinical pregnancy failure is dependent on immunologic factors that may be amplified by environmental stimuli, such as LPS, autoantibodies, including aPL Abs, and stress (8, 9). Importantly, TNF- is sensitive to modulation by such environmental factors.
TNF- is a multifunctional Th1 cytokine with roles in regulating hormone synthesis, placental architecture, and embryonic development (10), and elevated TNF- levels are associated with miscarriage (11) and pre-eclampsia (12). Patients with recurrent pregnancy loss, both with and without aPL Abs, evidence enrichment of TNF- promoter alleles associated with increased cytokine levels (11, 13). Increased placental levels of TNF- have been associated with pregnancy failure in mice (14), administration of TNF- increases abortion rates (6), and blockade of TNF- has been shown to prevent stress-induced miscarriage in murine models of abortion (15). That TNF- can directly promote tissue damage in pregnancy has been suggested by in vitro studies in which TNF--activated maternal monocytes bind to LFA-1 on placental syncytiotrophoblasts and induce apoptosis (16).
In this study, we examine the role of TNF- in aPL Ab-induced fetal injury in a murine model of APS in which pregnant mice receive human IgG containing aPL Abs. Passive transfer of IgG from women with recurrent miscarriage and aPL Abs results in fetal loss and growth restriction (17). The frequency of fetal resorption in aPL Ab-treated mice is 40% compared with <10% in mice treated with IgG from healthy individuals, and the average weight of surviving aPL Ab-treated fetuses is reduced by 50% (4, 13). We have shown that activation of complement, specifically C5a-C5a receptor interactions, is a necessary intermediary step for these clinically relevant deleterious effects of aPL Abs (4). However, the downstream pathogenic mediators of placental and fetal damage induced by aPL Abs and the sites and agents for optimal therapeutic interventions to prevent APS are not yet clear. In this study, we test the hypotheses that aPL Abs targeted to decidual tissue lead to the release of TNF-, that TNF- is a critical mediator in aPL Ab-induced damage, and that blockade of TNF- might be an effective treatment to prevent pregnancy loss.
Materials and Methods
Mice
Adult mice (6–8 wk) were used in all experiments. BALB/c mice were purchased from Taconic Farms. TNF--deficient, TNFtm1Gkl/J (TNF–/–) mice and TNF+/+ background strain (B6129S6) were obtained from The Jackson Laboratory (18). C5-deficient mice (C5–/–) (B10.D2-H2dH2-T18c Hc°/o2SnJ) and C5-sufficient mice (C5+/+) (C57BL/10SnJ) were also obtained from The Jackson Laboratory (19, 20). Procedures that involved mice were approved by the local Committee on Animal Use in Research and Education and were conducted in strict accordance with Guidelines for the Care and Use of Laboratory Research Animals promulgated by the National Institutes of Health.
Preparation of aPL and other Abs
Human IgG containing aPL Abs (aPL-IgG) was obtained from patients with APS (characterized by high-titer aPL Abs (>140 GPL U), thromboses, and/or pregnancy losses) (1). Normal human IgG (NH-IgG) was obtained from healthy non-autoimmune individuals. IgG was purified by affinity chromatography using protein G-Sepharose chromatography columns (Amersham Biosciences) as previously described (13). The generation, structure, and specificity of the human IgG1 aPL mAb (mAb 519) was previously described (21). All IgG samples were treated to deplete endotoxin with Centriprep ultracentrifugation devices (Millipore) and determined to be free of endotoxin using the Limulus amebocyte lysate assay.
Murine pregnancy model
Female mice were mated with previously isolated males, and the presence of a vaginal plug was defined as day 0 of pregnancy. On days 8 and 12 of pregnancy, mice were treated with i.p. injections of aPL-IgG (10 mg) or NH-IgG (10 mg). In some studies, mice were treated on days 8 and 12 of pregnancy with i.v. injections of human aPL mAb (100 μg) (21). Mice were sacrificed on day 15, uteri were dissected, fetuses and placentas were weighed, and fetal resorption frequency was calculated (number of resorptions divided by the total number of formed fetuses and resorptions). Resorption sites are easily identified and result from the loss of a previously viable fetus at that site. To inhibit TNF-, BALB/c mice were treated with polyethylene glycol (PEG)-conjugated soluble TNF-R type I (PEG sTNFRI; Amgen) on days 8, 10, and 12 of pregnancy (5 mg/kg, i.p). Mice received PEG sTNFRI 1 h before aPL-IgG injections on days 8 and 12. To inhibit C5, mice were treated on days 8 and 10 with anti-C5 mAb (1 mg, i.p.) (22) or murine IgG, as a control.
Detection of plasma levels of Abs and TNF-
Mice were bled on day 8 of pregnancy 30, 60, and 90 min, and 2, 4, 6, 8, 10, 12, 18, 24, 30, 36, and 42 h after administration of aPL-IgG or NH-IgG injection. Anticardiolipin (aCL) activity was measured by an ELISA according to the manufacturer’s instructions (Sigma-Aldrich). Levels of human IgG and mouse TNF- were measured by standard ELISA methods (Zeptometrix and OptEIA Pharmingen, respectively).
Immunohistochemistry
Deciduae were removed from mice on day 8 of pregnancy, 30, 60, 90, and 120 min after administration of aPL-IgG or NH-IgG, frozen in OCT compound, and cut into 10-μm sections. After quenching endogenous peroxidase with 1% H2O2 in methanol and blocking nonspecific binding sites with normal goat serum (Cappel), sections were incubated with goat anti-mouse TNF-, goat anti-human IgG (Sigma-Aldrich), or goat anti-mouse C3 (Cappel), followed by anti-goat IgG conjugated to HRP (Sigma-Aldrich). Bound HRP was detected with diaminobenzidine. Sections were counterstained with hematoxylin.
Statistical analyses
Data are expressed as mean ± SD. Student’s t test was used to compare fetal resorption rates and fetal weights between groups. A probability of <0.05 was used to reject the null hypothesis.
Results
APL Abs localize to decidual tissue in pregnant mice
We have previously shown that passive transfer of human IgG from APS patients into pregnant mice causes inflammation within deciduae and fetal resorption or growth restriction (4, 13). These in vivo studies demonstrated a direct pathogenic role for aPL Abs. In vitro studies in human and murine placentas have shown that trophoblast cell membranes are targets for 2GPI-dependent and 2GPI-independent aPL Abs (23), suggesting that these Abs are specifically targeted to the placenta. In this study, we investigated the tissue localization and kinetics of aPL Ab handling in an in vivo model. BALB/c mice were treated with human aPL-IgG (10 mg, i.p.) or NH-IgG on day 8 of pregnancy, and the presence of human IgG was assessed in tissue by immunohistochemistry. We detected human IgG in mouse deciduae within 30 min after i.p. injection of aPL-IgG into pregnant BALB/c mice (Fig. 1a). Analysis of kidney, lung, and liver sections from these animals showed no evidence of human IgG deposition (data not shown). In deciduae of mice treated with IgG from healthy individuals, there was no detectable human IgG deposition (Fig. 1b). To demonstrate that Abs reactive with aPL, rather than other autoantibodies or xenoreactive Abs that may be present in polyclonal human IgG, are deposited in deciduae, we treated pregnant mice with human aPL mAb (100 μg, i.v.). The results of immunohistochemical studies of decidual tissue from mice treated with human aPL mAb were similar to those obtained with polyclonal aPL-IgG (Fig. 1c).
FIGURE 1. aPL Abs are targeted to the decidua. a–c, BALB/c mice were treated with aPL-IgG (10 mg, i.p.), NH-IgG (10 mg, i.p.), or human aPL mAb (100 μg, i.v.) on day 8 of pregnancy (n = 6–8 mice/group). Mice were sacrificed 60 min after injections, and histologic sections of deciduae were stained with anti-human IgG Abs. In mice treated with polyclonal aPL-IgG (a) or human aPL mAb (c), human IgG was present throughout decidual tissue. In contrast, there was minimal IgG deposition in mice treated with NH-IgG (b). d and e, Pregnant mice (on day 8) and nonpregnant mice were treated with aPL-IgG (n = 5), and serial blood samples were collected and analyzed for plasma levels of human IgG and aCL reactivity. Levels of human IgG (d) and aCL Abs (e) increased to a similar extent in pregnant and nonpregnant mice. Although total human IgG levels were not different in pregnant and nonpregnant mice, aCL titers fell more rapidly in pregnant mice (p < 0.01), suggesting that aCL-reactive human IgG is specifically removed from circulation.
To determine whether aPL-reactive Abs are specifically targeted to decidual tissue, we compared the handling of aPL-IgG in pregnant mice with that in nonpregnant mice. Analysis of serial blood samples obtained after injection of aPL-IgG showed that, whereas levels of total human IgG in the circulation did not differ between pregnant and nonpregnant mice (Fig. 1d), titers of human aCL-reactive IgG were persistently and significantly lower in pregnant mice (e). The plasma volume of pregnant mice does not increase significantly by day 8, as measured by Evans blue dye dilution (pregnant vs nonpregnant, 1.07 ± 0.06 vs 1.10 ± 0.08 ml); therefore, the decreased levels of aCL-reactive IgG cannot be explained by a dilutional effect (24). Rather, our results suggest that aPL Abs are selectively targeted and bound to decidual tissues in pregnant mice, and thereby removed from the circulation.
Treatment with aPL Abs causes a rapid increase in decidual and systemic TNF-
To determine whether deposition of aPL-IgG in deciduae causes local TNF- release, we performed immunohistochemical studies of decidual tissues from day 8 of pregnancy removed 30, 60, 90, and 120 min after the i.p. injection of aPL-IgG. Intense staining for TNF- was evident by 30 min after administration of aPL-IgG (Fig. 2a) and persisted through 120 min. The time course for the appearance of TNF- in decidual tissue was similar to that of human IgG deposition. Immunohistochemistry studies indicated that trophoblasts, as well as infiltrating inflammatory cells, were producing TNF- (Fig. 2b). There was no TNF- in deciduae of mice treated with NH-IgG (Fig. 2, c and d).
FIGURE 2. Treatment with aPL Abs causes increased TNF- in deciduae and in plasma. a–d, BALB/c mice were treated with aPL-IgG (10 mg, i.p.) or NH-IgG (10 mg, i.p.) on day 8 of pregnancy. Mice were sacrificed 60 min after injections, and histologic sections of deciduae were stained with anti-mouse TNF- Ab. a and b, In representative sections of decidual tissue from mice treated with aPL-IgG, there was extensive intense staining for TNF- throughout decidual tissue and surrounding the necrotic embryonic debris (ED). Trophoblasts (*) and leukocytes (arrows) expressed TNF-. c and d, In mice treated with NH-IgG, there was minimal TNF- present, and the developing embryo (E) was intact. e, Pregnant and nonpregnant mice (n = 5–7 mice/group) were treated with aPL-IgG or NH-IgG on day 8, and serial blood samples were collected at 0, 10, 20, 30, 60, 90, 120, 240, 360, and 480 min after injections and assayed for levels of TNF-. In pregnant mice treated with aPL-IgG, there was a significant increase in plasma TNF- at 20 min (p < 0.001) after treatment, which peaked at 60 min and persisted for 8 h. aPL-IgG did not cause an increase in TNF- levels in nonpregnant mice. Levels were similar to those in mice treated with NH-IgG.
The administration of aPL-IgG to pregnant mice also caused a rapid and sustained increase in serum concentrations of TNF-. Within 20 min after i.p. injection of aPL-IgG, circulating levels of TNF- increased dramatically, peak levels occurred at 60 min, and the elevation of TNF- persisted for >8 h (Fig. 2e). Importantly, TNF- levels were not increased in pregnant mice that received NH-IgG or in nonpregnant mice treated with aPL-IgG (Fig. 2e). Taken together, our results show a temporal relationship between binding of human IgG to deciduae, local expression of TNF-, and elevation of systemic TNF-. That the increase in circulating TNF- was evident only in pregnant mice treated with aPL-IgG suggests that aPL Abs localized to deciduae initiate the production of proinflammatory cytokines by trophoblasts and leukocytes.
Blockade of TNF- improves pregnancy outcomes in mice treated with aPL Abs
To directly test whether elevated local or systemic TNF- contributes to aPL Ab-induced pregnancy loss, we studied mice with a targeted deletion of TNF-. If TNF- is a key intermediate in fetal injury, then TNF–/– mice should be protected from pregnancy loss. In the TNF+/+ background strain, aPL-IgG caused a >4-fold increase in the frequency of fetal resorption compared with treatment with NH-IgG (43 ± 9 vs 10 ± 5%; p < 0.001). In mice lacking TNF-, the harmful effect of aPL Abs on pregnancy outcome was markedly reduced. In fact, the frequency of fetal resorption in TNF–/– mice that received aPL-IgG was not significantly different from that of mice treated with IgG from healthy individuals (aPL-IgG vs NH-IgG, 24 ± 4 vs 17 ± 9%; p = NS), but it was significantly lower than that observed in TNF+/+ mice treated with aPL-IgG (p < 0.02) (Fig. 3a).
FIGURE 3. TNF- deficiency and TNF- blockade protect mice from aPL Ab-induced fetal loss. a, TNF+/+ and TNF–/– mice were treated with aPL-IgG (10 mg, i.p.) or NH-IgG (10 mg, i.p.) on days 8 and 12 of pregnancy. Fetal resorption frequency was determined on day 15 of pregnancy (n = 5–8 mice/group). In TNF+/+ mice, aPL-IgG induced a >4-fold increase in the frequency of fetal resorptions compared with NH-IgG (*, aPL-IgG vs NH-IgG, p < 0.001). In contrast, in TNF–/– mice, aPL-IgG did not cause an increase in fetal resorption frequency (aPL-IgG vs NH-IgG, p = NS; aPL-IgG, TNF+/+ vs TNF–/–, p < 0.02). b, Pregnant BALB/c mice were treated with aPL-IgG or NH-IgG on days 8 and 12 of pregnancy, and some from each group also received PEG-sTNFRI (5 mg/kg, i.p) on days 8, 10, and 12 of pregnancy (n = 5–8 mice/group). Administration of PEG-sTNFRI along with aPL-IgG reduced pregnancy losses by 50% (*, aPL-IgG vs aPL-IgG plus sTNFRI, p < 0.005), but PEG sTNFRI caused a modest in increased fetal loss in mice treated with NH-IgG (NH-IgG plus PEG sTNFRI vs NH-IgG, p < 0.02).
As an alternate approach to test the hypothesis that TNF- is a mediator of aPL Ab-induced pregnancy complications, we blocked TNF- activity with high-affinity monomeric PEGylated type I TNFR. PEG sTNFRI has been shown to have potent in vivo anti-inflammatory activity in a number of murine models of rheumatoid arthritis (25, 26). In pregnant BALB/c mice treated with PEG sTNFRI on days 8, 10, and 12, fetal resorption frequency was not increased by administration of aPL-IgG (aPL-IgG vs NH-IgG, 21 ± 4 vs 19 ± 6%) (Fig. 3b). Although there was modest increase in fetal loss associated with PEG sTNFRI in mice treated with NH-IgG (NH-IgG plus PEG sTNFRI vs NH-IgG, 19 ± 6 vs 8 ± 6%; p < 0.02), blockade of TNF- in aPL Ab-treated mice decreased pregnancy loss by 50% (aPL-IgG vs aPL-IgG plus PEG sTNFRI, 42 ± 6 vs 21 ± 4%; p < 0.005) (Fig. 3b). The magnitude of the improvement in fetal survival was comparable to that observed in TNF–/– mice (Fig. 3a).
Activation of complement is required for TNF- release induced by aPL Abs
We have previously identified complement component C5 and, particularly, its cleavage product, C5a, as key mediators of fetal injury caused by aPL Abs, and have shown that C5–/– mice are completely protected from aPL Ab-induced fetal resorption and growth restriction (4). C5a attracts and activates neutrophils, monocytes, and mast cells, and stimulates the release of inflammatory mediators, including TNF-. To determine whether there is a relationship between activation of complement, release of TNF-, and pregnancy loss caused by aPL Abs, we studied C5-deficient mice. If TNF- is a critical effector of fetal injury that acts downstream of C5, we would expect to find no elevation of TNF- in C5-deficient mice treated with aPL-IgG. In the C5+/+ background strain, there was a rapid and sustained increase in circulating TNF- in response to aPL-IgG Abs (Fig. 4a), comparable to that seen in BALB/c mice (Fig. 2c). In contrast, in C5–/– mice, there was no increase in TNF- after aPL-IgG; levels were similar to those in mice treated with NH-IgG (Fig. 4a).
FIGURE 4. APL Ab-induced increases in TNF- are downstream of complement activation. a, C5+/+ and C5–/– mice were treated with aPL-IgG (10 mg, i.p) or NH-IgG (10 mg, i.p.) on day 8 of pregnancy, and serial blood samples were collected and analyzed for TNF- levels (n = 5–8 mice/group). In C5+/+ mice, TNF- levels increased rapidly and remained elevated after administration of aPL-IgG (p < 0.005), whereas in C5–/– mice there was no change in TNF- after treatment with aPL-IgG. TNF- levels did not change in mice that received NH-IgG. b–d, Decidual tissue from mice sacrificed 60 min after injections was stained with anti-mouse TNF- Ab. Although there was extensive TNF- production and deposition deciduae from C5+/+ mice treated with aPL-IgG (b), minimal TNF- was present in the absence of C5 activation due to either C5 deficiency (c) or treatment with anti-C5 mAb (1 mg, i.p.) (d).
Immunohistochemical studies of decidual tissues from mice lacking C5 documented the absence of local TNF- release in response to aPL-IgG. There was extensive staining for TNF- in deciduae of C5+/+ mice that had received aPL-IgG (Fig. 4b), but only minimal TNF- in C5–/– mice (c). Similarly, blockade of C5 cleavage with anti-C5 mAb, which we have previously shown to completely protect pregnancies from aPL Ab-induced injury (4), also prevented local TNF- release (Fig. 4d). Similar results were obtained in mice treated with a C5a receptor antagonist peptide (data not shown). Collectively, our data indicate that TNF-, a mediator of pregnancy loss, is produced in response to complement activation by aPL Abs.
Discussion
In this paper, we report experiments in a murine model of APS induced by passive transfer of human aPL Abs that show that aPL Abs are preferentially bound by decidual tissues where they induce local complement activation, production of inflammatory mediators, and ultimately pregnancy loss. We have identified the release of TNF- as a critical intermediate that acts downstream of C5 activation and links pathogenic aPL Abs to fetal damage. Our conclusions are based on the fetal protective effects of TNF- deficiency and TNF blockade and on the absence of increased TNF- levels in C5-deficient mice treated with aPL Abs. Our results identify TNF blockade as potential therapy for the pregnancy complications of APS.
Although APS is a systemic disease, its most common clinical manifestation is miscarriage. In fact, APS has emerged as a leading cause of pregnancy loss and pregnancy-related morbidity, including intrauterine growth restriction, premature births, and pre-eclampsia (27). Up to 20% of women with recurrent miscarriage have aPL Abs, and in 15% of otherwise apparently normal women, aPL Abs are the sole explanation for recurrent fetal loss (28, 29). That aPL Abs are specifically targeted to trophoblasts is consistent with these clinical observations. During trophoblast differentiation, phosphatidylserine is externalized on the trophoblast outer leaflet where it provides a target for aPL Abs (27, 30). Our in vivo studies show that aPL Abs are selectively removed from the circulation in pregnant mice and specifically bound to decidual tissues where they activate complement via the classical pathway (4). Effectors downstream of complement activation, such as TNF-, provide a means by which these Abs cause pregnancy failure. Indeed, we noted that the appearance of TNF- in the decidua and in the circulation was coincident with complement activation within the inflamed decidua and on the extraembryonic membranes (13). Both observations were made as early as 30 min after aPL-IgG injection, indicating that TNF- may be an early indicator of embryonic or fetal damage.
The deleterious effects of TNF- on pregnancy may be direct and indirect. TNFRs expressed on the trophoblast modulate cell proliferation and differentiation in normal pregnancy (10, 31). Expression of TNFRI is constitutive, whereas expression of TNFRII is developmentally programmed (32). In vitro studies have shown that TNF- induces apoptosis of cytotrophoblasts, suggesting that aberrant expression of TNF- may have harmful effects on placental development and function (33, 34, 35). However, regulated TNF- expression in the developing placenta may be essential at specific stages of pregnancy, because PEG sTNFRI-induced blockade of TNF- in mice treated with NH-IgG was associated with a modest increase in pregnancy failure. Finally, deficient hormone synthesis by the corpus luteum as a consequence of increased TNF- may also interfere with pregnancy survival (36, 37, 38, 39).
TNF- production may induce abortion indirectly by activating endothelial cells and leukocytes. Our previous work showing that mice lacking ICAM-1 are protected from aPL Ab-associated pregnancy complications emphasizes the importance of endothelial activation in fetal injury (40). TNF--stimulated neutrophils and monocytes release inflammatory mediators, including reactive oxidants, proteolytic enzymes, and complement components, that directly damage decidual tissue, accelerate alternative pathway activation, enhancing C5 cleavage, and thereby trigger release of more TNF-, which further amplifies local inflammation. In addition, stimulation of effector cells by cytokines alters their response to complement activation products, and a major effect of C3a and C5a is recruitment of leukocytes that release such cytokines. It is also possible that trophoblasts amplify local damage by secreting TNF- themselves, in response to complement split products or assembly of the terminal membrane attack complex of complement, C5b-9, on the cell membrane. Indeed, in an experimental model of dilated cardiomyopathy, chronic Ig-mediated complement activation in myocardium contributes to progression of disease by inducing TNF- expression in cardiac myocytes through sublytic C5b-9 (41). Placental tissues exposed to hypoxia-reoxygenation in vitro (the stimulus associated with activation of complement (42)) express and secrete increased levels of TNF-, underscoring the possibility that trophoblasts themselves are a source of proinflammatory cytokines (43).
In summary, our findings demonstrate that TNF- is a critical mediator of aPL Ab-induced injury, and that it is released in response to complement activation. Our results fit into a larger schema consistent with evidence from murine models of myocarditis, sepsis, and rheumatoid arthritis, and suggest that tissue damage induced as a result of complement activation is mediated by production of TNF- (42, 44). However, because blockade of TNF- does not completely protect pregnancies, it is likely that other effector pathways contribute to fetal demise. The variability in the effectiveness of TNF- inhibition that we observed is in keeping with the heterogeneous responses to TNF- blockade seen in murine models of Ab-induced arthritis (45) and in patients with rheumatoid arthritis. Thus, inhibitors of TNF- may provide effective treatment for some patients with APS and recurrent pregnancy loss. In conclusion, therapies directed at TNF-, already in use to treat other inflammatory diseases (46, 47, 48), merit more intense scrutiny in pregnant women with APS, but the therapeutic window for TNF- blockade in pregnancy needs to be carefully defined.
Acknowledgments
We are grateful to our colleagues for generously providing reagents, Dr. Ulrich Feige (Amgen) for PEG sTNFRI, Dr. Michael Holers (University of Colorado Health Sciences Center, Denver, CO) for anti-C5 mAb, and Dr. Paolo Casali (University of California, Irvine, CA) for human aPL mAb. We thank Marta Guerra for help in preparing the manuscript.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported in part by a grant from the National Institutes of Health (AI055007) and by the Alliance for Lupus Research and the Mary Kirkland Center for Lupus Research at Hospital for Special Surgery.
2 Address correspondence and reprint requests to Dr. Jane E. Salmon, Hospital for Special Surgery, 535 East 70th Street, New York, NY 10021. E-mail address: salmonj@hss.edu
3 Abbreviations used in this paper: APS, antiphospholipid syndrome; aPL, antiphospholipid; aPL-IgG, human IgG containing aPL Abs; NH-IgG, normal human IgG; PEG, polyethylene glycol; sTNFRI, soluble TNF-R type I; aCL, anticardiolipin.
Received for publication July 8, 2004. Accepted for publication October 4, 2004.
References
Wilson, W. A., A. E. Gharavi, T. Koike, M. D. Lockshin, D. W. Branch, J. C. Piette, R. Brey, R. Derksen, E. N. Harris, G. R. Hughes, et al 1999. International consensus statement on preliminary classification criteria for definite antiphospholipid syndrome: report of an international workshop. Arthritis Rheum. 42:1309.
Out, H. J., C. D. Kooijman, H. W. Bruinse, R. H. Derksen. 1991. Histopathological findings in placentae from patients with intra-uterine fetal death and anti-phospholipid antibodies. Eur. J. Obstet. Gynecol. Reprod. Biol. 41:179.
Magid, M. S., C. Kaplan, L. R. Sammaritano, M. Peterson, M. L. Druzin, M. D. Lockshin. 1998. Placental pathology in systemic lupus erythematosus: a prospective study. Am. J. Obstet. Gynecol. 179:226.
Girardi, G., J. Berman, P. Redecha, L. Spruce, J. M. Thurman, D. Kraus, T. J. Hollmann, P. Casali, M. C. Caroll, R. A. Wetsel, et al 2003. Complement C5a receptors and neutrophils mediate fetal injury in the antiphospholipid syndrome. J. Clin. Invest. 112:1644
Mellor, A. L., D. H. Munn. 2000. Immunology at the maternal-fetal interface: lessons for T cell tolerance and suppression. Annu. Rev. Immunol. 18:367.
Chaouat, G., E. Menu, D. A. Clark, M. Dy, M. Minkowski, T. G. Wegmann. 1990. Control of fetal survival in CBA x DBA/2 mice by lymphokine therapy. J. Reprod. Fertil. 89:447
Raghupathy, R.. 2001. Pregnancy: success and failure within the Th1/Th2/Th3 paradigm. Semin. Immunol. 13:219.(Jessica Berman, Guillermi)
The antiphospholipid syndrome (APS) is characterized by recurrent fetal loss, intrauterine growth restriction, and vascular thrombosis in the presence of antiphospholipid (aPL) Abs. Our studies in a murine model of APS induced by passive transfer of human aPL Abs have shown that activation of complement and recruitment of neutrophils into decidua are required for fetal loss, and emphasize the importance of inflammation in aPL Ab-induced pregnancy loss. In this study, we examine the role of TNF- in pregnancy complications associated with aPL Abs in a murine model of APS. We show that aPL Abs are specifically targeted to decidual tissue and cause a rapid increase in decidual and systemic TNF- levels. We identify the release of TNF- as a critical intermediate that acts downstream of C5 activation, based on the fetal protective effects of TNF- deficiency and TNF blockade and on the absence of increased TNF- levels in C5-deficient mice treated with aPL Abs. Our results suggest that TNF- links pathogenic aPL Abs to fetal damage and identify TNF blockade as a potential therapy for the pregnancy complications of APS.
Introduction
The antiphospholipid syndrome (APS)3 is characterized by arterial and venous thromboses and pregnancy loss that occur in the presence of antiphospholipid (aPL) Abs (1). APS is common in patients with systemic lupus erythematous, and it is also found in individuals without other autoimmune features. Although it is clear that the specific antigenic reactivity of aPL Abs and their targeting to decidual tissue are critical to their effect, the pathogenic mechanisms that result in fetal injury in vivo are incompletely understood.
Therapy for pregnant women with APS is focused on preventing thrombosis, but anticoagulation is only partially successful in averting miscarriage and carries risks for fetus and mother. Although experimental models have emphasized the role of thrombosis in placental tissue, histopathologic findings in placentas from women with APS argue that proinflammatory factors may contribute to injury (2, 3). Our recent studies showing that activation of complement and recruitment of neutrophils into decidua are required for fetal loss in a murine model of APS underscore the importance of inflammation in aPL Ab-induced pregnancy loss (4). However, the specific effectors of fetal injury remain unknown. Because TNF- is released when infiltrating inflammatory cells are activated by complement split products and elevated levels of TNF- have been associated with pregnancy failures, we considered the possibility that TNF- may have a role in pregnancy complications associated with APS.
Evidence from human pregnancy studies points to a strong association between maternal Th2-type immunity and successful pregnancy, whereas Th1-type immune reactivity is associated with pregnancy loss (5). Mitogen-stimulated PBMC, harvested at the end of the first trimester, have consistently been shown to produce higher levels of IL-4, IL-5, IL-6, and IL-10 in patients who went on to have normal pregnancies than in those who had recurrent spontaneous miscarriages. In contrast, levels of TNF-, IL-2, and IFN- were uniformly higher in patients who subsequently had miscarriages (6, 7). Indeed, the preponderance of Th2 cytokines relative to Th1 cytokines in the local milieu of the fetus is considered to be essential for its survival. The proinflammatory Th1-dominant response that underlies clinical pregnancy failure is dependent on immunologic factors that may be amplified by environmental stimuli, such as LPS, autoantibodies, including aPL Abs, and stress (8, 9). Importantly, TNF- is sensitive to modulation by such environmental factors.
TNF- is a multifunctional Th1 cytokine with roles in regulating hormone synthesis, placental architecture, and embryonic development (10), and elevated TNF- levels are associated with miscarriage (11) and pre-eclampsia (12). Patients with recurrent pregnancy loss, both with and without aPL Abs, evidence enrichment of TNF- promoter alleles associated with increased cytokine levels (11, 13). Increased placental levels of TNF- have been associated with pregnancy failure in mice (14), administration of TNF- increases abortion rates (6), and blockade of TNF- has been shown to prevent stress-induced miscarriage in murine models of abortion (15). That TNF- can directly promote tissue damage in pregnancy has been suggested by in vitro studies in which TNF--activated maternal monocytes bind to LFA-1 on placental syncytiotrophoblasts and induce apoptosis (16).
In this study, we examine the role of TNF- in aPL Ab-induced fetal injury in a murine model of APS in which pregnant mice receive human IgG containing aPL Abs. Passive transfer of IgG from women with recurrent miscarriage and aPL Abs results in fetal loss and growth restriction (17). The frequency of fetal resorption in aPL Ab-treated mice is 40% compared with <10% in mice treated with IgG from healthy individuals, and the average weight of surviving aPL Ab-treated fetuses is reduced by 50% (4, 13). We have shown that activation of complement, specifically C5a-C5a receptor interactions, is a necessary intermediary step for these clinically relevant deleterious effects of aPL Abs (4). However, the downstream pathogenic mediators of placental and fetal damage induced by aPL Abs and the sites and agents for optimal therapeutic interventions to prevent APS are not yet clear. In this study, we test the hypotheses that aPL Abs targeted to decidual tissue lead to the release of TNF-, that TNF- is a critical mediator in aPL Ab-induced damage, and that blockade of TNF- might be an effective treatment to prevent pregnancy loss.
Materials and Methods
Mice
Adult mice (6–8 wk) were used in all experiments. BALB/c mice were purchased from Taconic Farms. TNF--deficient, TNFtm1Gkl/J (TNF–/–) mice and TNF+/+ background strain (B6129S6) were obtained from The Jackson Laboratory (18). C5-deficient mice (C5–/–) (B10.D2-H2dH2-T18c Hc°/o2SnJ) and C5-sufficient mice (C5+/+) (C57BL/10SnJ) were also obtained from The Jackson Laboratory (19, 20). Procedures that involved mice were approved by the local Committee on Animal Use in Research and Education and were conducted in strict accordance with Guidelines for the Care and Use of Laboratory Research Animals promulgated by the National Institutes of Health.
Preparation of aPL and other Abs
Human IgG containing aPL Abs (aPL-IgG) was obtained from patients with APS (characterized by high-titer aPL Abs (>140 GPL U), thromboses, and/or pregnancy losses) (1). Normal human IgG (NH-IgG) was obtained from healthy non-autoimmune individuals. IgG was purified by affinity chromatography using protein G-Sepharose chromatography columns (Amersham Biosciences) as previously described (13). The generation, structure, and specificity of the human IgG1 aPL mAb (mAb 519) was previously described (21). All IgG samples were treated to deplete endotoxin with Centriprep ultracentrifugation devices (Millipore) and determined to be free of endotoxin using the Limulus amebocyte lysate assay.
Murine pregnancy model
Female mice were mated with previously isolated males, and the presence of a vaginal plug was defined as day 0 of pregnancy. On days 8 and 12 of pregnancy, mice were treated with i.p. injections of aPL-IgG (10 mg) or NH-IgG (10 mg). In some studies, mice were treated on days 8 and 12 of pregnancy with i.v. injections of human aPL mAb (100 μg) (21). Mice were sacrificed on day 15, uteri were dissected, fetuses and placentas were weighed, and fetal resorption frequency was calculated (number of resorptions divided by the total number of formed fetuses and resorptions). Resorption sites are easily identified and result from the loss of a previously viable fetus at that site. To inhibit TNF-, BALB/c mice were treated with polyethylene glycol (PEG)-conjugated soluble TNF-R type I (PEG sTNFRI; Amgen) on days 8, 10, and 12 of pregnancy (5 mg/kg, i.p). Mice received PEG sTNFRI 1 h before aPL-IgG injections on days 8 and 12. To inhibit C5, mice were treated on days 8 and 10 with anti-C5 mAb (1 mg, i.p.) (22) or murine IgG, as a control.
Detection of plasma levels of Abs and TNF-
Mice were bled on day 8 of pregnancy 30, 60, and 90 min, and 2, 4, 6, 8, 10, 12, 18, 24, 30, 36, and 42 h after administration of aPL-IgG or NH-IgG injection. Anticardiolipin (aCL) activity was measured by an ELISA according to the manufacturer’s instructions (Sigma-Aldrich). Levels of human IgG and mouse TNF- were measured by standard ELISA methods (Zeptometrix and OptEIA Pharmingen, respectively).
Immunohistochemistry
Deciduae were removed from mice on day 8 of pregnancy, 30, 60, 90, and 120 min after administration of aPL-IgG or NH-IgG, frozen in OCT compound, and cut into 10-μm sections. After quenching endogenous peroxidase with 1% H2O2 in methanol and blocking nonspecific binding sites with normal goat serum (Cappel), sections were incubated with goat anti-mouse TNF-, goat anti-human IgG (Sigma-Aldrich), or goat anti-mouse C3 (Cappel), followed by anti-goat IgG conjugated to HRP (Sigma-Aldrich). Bound HRP was detected with diaminobenzidine. Sections were counterstained with hematoxylin.
Statistical analyses
Data are expressed as mean ± SD. Student’s t test was used to compare fetal resorption rates and fetal weights between groups. A probability of <0.05 was used to reject the null hypothesis.
Results
APL Abs localize to decidual tissue in pregnant mice
We have previously shown that passive transfer of human IgG from APS patients into pregnant mice causes inflammation within deciduae and fetal resorption or growth restriction (4, 13). These in vivo studies demonstrated a direct pathogenic role for aPL Abs. In vitro studies in human and murine placentas have shown that trophoblast cell membranes are targets for 2GPI-dependent and 2GPI-independent aPL Abs (23), suggesting that these Abs are specifically targeted to the placenta. In this study, we investigated the tissue localization and kinetics of aPL Ab handling in an in vivo model. BALB/c mice were treated with human aPL-IgG (10 mg, i.p.) or NH-IgG on day 8 of pregnancy, and the presence of human IgG was assessed in tissue by immunohistochemistry. We detected human IgG in mouse deciduae within 30 min after i.p. injection of aPL-IgG into pregnant BALB/c mice (Fig. 1a). Analysis of kidney, lung, and liver sections from these animals showed no evidence of human IgG deposition (data not shown). In deciduae of mice treated with IgG from healthy individuals, there was no detectable human IgG deposition (Fig. 1b). To demonstrate that Abs reactive with aPL, rather than other autoantibodies or xenoreactive Abs that may be present in polyclonal human IgG, are deposited in deciduae, we treated pregnant mice with human aPL mAb (100 μg, i.v.). The results of immunohistochemical studies of decidual tissue from mice treated with human aPL mAb were similar to those obtained with polyclonal aPL-IgG (Fig. 1c).
FIGURE 1. aPL Abs are targeted to the decidua. a–c, BALB/c mice were treated with aPL-IgG (10 mg, i.p.), NH-IgG (10 mg, i.p.), or human aPL mAb (100 μg, i.v.) on day 8 of pregnancy (n = 6–8 mice/group). Mice were sacrificed 60 min after injections, and histologic sections of deciduae were stained with anti-human IgG Abs. In mice treated with polyclonal aPL-IgG (a) or human aPL mAb (c), human IgG was present throughout decidual tissue. In contrast, there was minimal IgG deposition in mice treated with NH-IgG (b). d and e, Pregnant mice (on day 8) and nonpregnant mice were treated with aPL-IgG (n = 5), and serial blood samples were collected and analyzed for plasma levels of human IgG and aCL reactivity. Levels of human IgG (d) and aCL Abs (e) increased to a similar extent in pregnant and nonpregnant mice. Although total human IgG levels were not different in pregnant and nonpregnant mice, aCL titers fell more rapidly in pregnant mice (p < 0.01), suggesting that aCL-reactive human IgG is specifically removed from circulation.
To determine whether aPL-reactive Abs are specifically targeted to decidual tissue, we compared the handling of aPL-IgG in pregnant mice with that in nonpregnant mice. Analysis of serial blood samples obtained after injection of aPL-IgG showed that, whereas levels of total human IgG in the circulation did not differ between pregnant and nonpregnant mice (Fig. 1d), titers of human aCL-reactive IgG were persistently and significantly lower in pregnant mice (e). The plasma volume of pregnant mice does not increase significantly by day 8, as measured by Evans blue dye dilution (pregnant vs nonpregnant, 1.07 ± 0.06 vs 1.10 ± 0.08 ml); therefore, the decreased levels of aCL-reactive IgG cannot be explained by a dilutional effect (24). Rather, our results suggest that aPL Abs are selectively targeted and bound to decidual tissues in pregnant mice, and thereby removed from the circulation.
Treatment with aPL Abs causes a rapid increase in decidual and systemic TNF-
To determine whether deposition of aPL-IgG in deciduae causes local TNF- release, we performed immunohistochemical studies of decidual tissues from day 8 of pregnancy removed 30, 60, 90, and 120 min after the i.p. injection of aPL-IgG. Intense staining for TNF- was evident by 30 min after administration of aPL-IgG (Fig. 2a) and persisted through 120 min. The time course for the appearance of TNF- in decidual tissue was similar to that of human IgG deposition. Immunohistochemistry studies indicated that trophoblasts, as well as infiltrating inflammatory cells, were producing TNF- (Fig. 2b). There was no TNF- in deciduae of mice treated with NH-IgG (Fig. 2, c and d).
FIGURE 2. Treatment with aPL Abs causes increased TNF- in deciduae and in plasma. a–d, BALB/c mice were treated with aPL-IgG (10 mg, i.p.) or NH-IgG (10 mg, i.p.) on day 8 of pregnancy. Mice were sacrificed 60 min after injections, and histologic sections of deciduae were stained with anti-mouse TNF- Ab. a and b, In representative sections of decidual tissue from mice treated with aPL-IgG, there was extensive intense staining for TNF- throughout decidual tissue and surrounding the necrotic embryonic debris (ED). Trophoblasts (*) and leukocytes (arrows) expressed TNF-. c and d, In mice treated with NH-IgG, there was minimal TNF- present, and the developing embryo (E) was intact. e, Pregnant and nonpregnant mice (n = 5–7 mice/group) were treated with aPL-IgG or NH-IgG on day 8, and serial blood samples were collected at 0, 10, 20, 30, 60, 90, 120, 240, 360, and 480 min after injections and assayed for levels of TNF-. In pregnant mice treated with aPL-IgG, there was a significant increase in plasma TNF- at 20 min (p < 0.001) after treatment, which peaked at 60 min and persisted for 8 h. aPL-IgG did not cause an increase in TNF- levels in nonpregnant mice. Levels were similar to those in mice treated with NH-IgG.
The administration of aPL-IgG to pregnant mice also caused a rapid and sustained increase in serum concentrations of TNF-. Within 20 min after i.p. injection of aPL-IgG, circulating levels of TNF- increased dramatically, peak levels occurred at 60 min, and the elevation of TNF- persisted for >8 h (Fig. 2e). Importantly, TNF- levels were not increased in pregnant mice that received NH-IgG or in nonpregnant mice treated with aPL-IgG (Fig. 2e). Taken together, our results show a temporal relationship between binding of human IgG to deciduae, local expression of TNF-, and elevation of systemic TNF-. That the increase in circulating TNF- was evident only in pregnant mice treated with aPL-IgG suggests that aPL Abs localized to deciduae initiate the production of proinflammatory cytokines by trophoblasts and leukocytes.
Blockade of TNF- improves pregnancy outcomes in mice treated with aPL Abs
To directly test whether elevated local or systemic TNF- contributes to aPL Ab-induced pregnancy loss, we studied mice with a targeted deletion of TNF-. If TNF- is a key intermediate in fetal injury, then TNF–/– mice should be protected from pregnancy loss. In the TNF+/+ background strain, aPL-IgG caused a >4-fold increase in the frequency of fetal resorption compared with treatment with NH-IgG (43 ± 9 vs 10 ± 5%; p < 0.001). In mice lacking TNF-, the harmful effect of aPL Abs on pregnancy outcome was markedly reduced. In fact, the frequency of fetal resorption in TNF–/– mice that received aPL-IgG was not significantly different from that of mice treated with IgG from healthy individuals (aPL-IgG vs NH-IgG, 24 ± 4 vs 17 ± 9%; p = NS), but it was significantly lower than that observed in TNF+/+ mice treated with aPL-IgG (p < 0.02) (Fig. 3a).
FIGURE 3. TNF- deficiency and TNF- blockade protect mice from aPL Ab-induced fetal loss. a, TNF+/+ and TNF–/– mice were treated with aPL-IgG (10 mg, i.p.) or NH-IgG (10 mg, i.p.) on days 8 and 12 of pregnancy. Fetal resorption frequency was determined on day 15 of pregnancy (n = 5–8 mice/group). In TNF+/+ mice, aPL-IgG induced a >4-fold increase in the frequency of fetal resorptions compared with NH-IgG (*, aPL-IgG vs NH-IgG, p < 0.001). In contrast, in TNF–/– mice, aPL-IgG did not cause an increase in fetal resorption frequency (aPL-IgG vs NH-IgG, p = NS; aPL-IgG, TNF+/+ vs TNF–/–, p < 0.02). b, Pregnant BALB/c mice were treated with aPL-IgG or NH-IgG on days 8 and 12 of pregnancy, and some from each group also received PEG-sTNFRI (5 mg/kg, i.p) on days 8, 10, and 12 of pregnancy (n = 5–8 mice/group). Administration of PEG-sTNFRI along with aPL-IgG reduced pregnancy losses by 50% (*, aPL-IgG vs aPL-IgG plus sTNFRI, p < 0.005), but PEG sTNFRI caused a modest in increased fetal loss in mice treated with NH-IgG (NH-IgG plus PEG sTNFRI vs NH-IgG, p < 0.02).
As an alternate approach to test the hypothesis that TNF- is a mediator of aPL Ab-induced pregnancy complications, we blocked TNF- activity with high-affinity monomeric PEGylated type I TNFR. PEG sTNFRI has been shown to have potent in vivo anti-inflammatory activity in a number of murine models of rheumatoid arthritis (25, 26). In pregnant BALB/c mice treated with PEG sTNFRI on days 8, 10, and 12, fetal resorption frequency was not increased by administration of aPL-IgG (aPL-IgG vs NH-IgG, 21 ± 4 vs 19 ± 6%) (Fig. 3b). Although there was modest increase in fetal loss associated with PEG sTNFRI in mice treated with NH-IgG (NH-IgG plus PEG sTNFRI vs NH-IgG, 19 ± 6 vs 8 ± 6%; p < 0.02), blockade of TNF- in aPL Ab-treated mice decreased pregnancy loss by 50% (aPL-IgG vs aPL-IgG plus PEG sTNFRI, 42 ± 6 vs 21 ± 4%; p < 0.005) (Fig. 3b). The magnitude of the improvement in fetal survival was comparable to that observed in TNF–/– mice (Fig. 3a).
Activation of complement is required for TNF- release induced by aPL Abs
We have previously identified complement component C5 and, particularly, its cleavage product, C5a, as key mediators of fetal injury caused by aPL Abs, and have shown that C5–/– mice are completely protected from aPL Ab-induced fetal resorption and growth restriction (4). C5a attracts and activates neutrophils, monocytes, and mast cells, and stimulates the release of inflammatory mediators, including TNF-. To determine whether there is a relationship between activation of complement, release of TNF-, and pregnancy loss caused by aPL Abs, we studied C5-deficient mice. If TNF- is a critical effector of fetal injury that acts downstream of C5, we would expect to find no elevation of TNF- in C5-deficient mice treated with aPL-IgG. In the C5+/+ background strain, there was a rapid and sustained increase in circulating TNF- in response to aPL-IgG Abs (Fig. 4a), comparable to that seen in BALB/c mice (Fig. 2c). In contrast, in C5–/– mice, there was no increase in TNF- after aPL-IgG; levels were similar to those in mice treated with NH-IgG (Fig. 4a).
FIGURE 4. APL Ab-induced increases in TNF- are downstream of complement activation. a, C5+/+ and C5–/– mice were treated with aPL-IgG (10 mg, i.p) or NH-IgG (10 mg, i.p.) on day 8 of pregnancy, and serial blood samples were collected and analyzed for TNF- levels (n = 5–8 mice/group). In C5+/+ mice, TNF- levels increased rapidly and remained elevated after administration of aPL-IgG (p < 0.005), whereas in C5–/– mice there was no change in TNF- after treatment with aPL-IgG. TNF- levels did not change in mice that received NH-IgG. b–d, Decidual tissue from mice sacrificed 60 min after injections was stained with anti-mouse TNF- Ab. Although there was extensive TNF- production and deposition deciduae from C5+/+ mice treated with aPL-IgG (b), minimal TNF- was present in the absence of C5 activation due to either C5 deficiency (c) or treatment with anti-C5 mAb (1 mg, i.p.) (d).
Immunohistochemical studies of decidual tissues from mice lacking C5 documented the absence of local TNF- release in response to aPL-IgG. There was extensive staining for TNF- in deciduae of C5+/+ mice that had received aPL-IgG (Fig. 4b), but only minimal TNF- in C5–/– mice (c). Similarly, blockade of C5 cleavage with anti-C5 mAb, which we have previously shown to completely protect pregnancies from aPL Ab-induced injury (4), also prevented local TNF- release (Fig. 4d). Similar results were obtained in mice treated with a C5a receptor antagonist peptide (data not shown). Collectively, our data indicate that TNF-, a mediator of pregnancy loss, is produced in response to complement activation by aPL Abs.
Discussion
In this paper, we report experiments in a murine model of APS induced by passive transfer of human aPL Abs that show that aPL Abs are preferentially bound by decidual tissues where they induce local complement activation, production of inflammatory mediators, and ultimately pregnancy loss. We have identified the release of TNF- as a critical intermediate that acts downstream of C5 activation and links pathogenic aPL Abs to fetal damage. Our conclusions are based on the fetal protective effects of TNF- deficiency and TNF blockade and on the absence of increased TNF- levels in C5-deficient mice treated with aPL Abs. Our results identify TNF blockade as potential therapy for the pregnancy complications of APS.
Although APS is a systemic disease, its most common clinical manifestation is miscarriage. In fact, APS has emerged as a leading cause of pregnancy loss and pregnancy-related morbidity, including intrauterine growth restriction, premature births, and pre-eclampsia (27). Up to 20% of women with recurrent miscarriage have aPL Abs, and in 15% of otherwise apparently normal women, aPL Abs are the sole explanation for recurrent fetal loss (28, 29). That aPL Abs are specifically targeted to trophoblasts is consistent with these clinical observations. During trophoblast differentiation, phosphatidylserine is externalized on the trophoblast outer leaflet where it provides a target for aPL Abs (27, 30). Our in vivo studies show that aPL Abs are selectively removed from the circulation in pregnant mice and specifically bound to decidual tissues where they activate complement via the classical pathway (4). Effectors downstream of complement activation, such as TNF-, provide a means by which these Abs cause pregnancy failure. Indeed, we noted that the appearance of TNF- in the decidua and in the circulation was coincident with complement activation within the inflamed decidua and on the extraembryonic membranes (13). Both observations were made as early as 30 min after aPL-IgG injection, indicating that TNF- may be an early indicator of embryonic or fetal damage.
The deleterious effects of TNF- on pregnancy may be direct and indirect. TNFRs expressed on the trophoblast modulate cell proliferation and differentiation in normal pregnancy (10, 31). Expression of TNFRI is constitutive, whereas expression of TNFRII is developmentally programmed (32). In vitro studies have shown that TNF- induces apoptosis of cytotrophoblasts, suggesting that aberrant expression of TNF- may have harmful effects on placental development and function (33, 34, 35). However, regulated TNF- expression in the developing placenta may be essential at specific stages of pregnancy, because PEG sTNFRI-induced blockade of TNF- in mice treated with NH-IgG was associated with a modest increase in pregnancy failure. Finally, deficient hormone synthesis by the corpus luteum as a consequence of increased TNF- may also interfere with pregnancy survival (36, 37, 38, 39).
TNF- production may induce abortion indirectly by activating endothelial cells and leukocytes. Our previous work showing that mice lacking ICAM-1 are protected from aPL Ab-associated pregnancy complications emphasizes the importance of endothelial activation in fetal injury (40). TNF--stimulated neutrophils and monocytes release inflammatory mediators, including reactive oxidants, proteolytic enzymes, and complement components, that directly damage decidual tissue, accelerate alternative pathway activation, enhancing C5 cleavage, and thereby trigger release of more TNF-, which further amplifies local inflammation. In addition, stimulation of effector cells by cytokines alters their response to complement activation products, and a major effect of C3a and C5a is recruitment of leukocytes that release such cytokines. It is also possible that trophoblasts amplify local damage by secreting TNF- themselves, in response to complement split products or assembly of the terminal membrane attack complex of complement, C5b-9, on the cell membrane. Indeed, in an experimental model of dilated cardiomyopathy, chronic Ig-mediated complement activation in myocardium contributes to progression of disease by inducing TNF- expression in cardiac myocytes through sublytic C5b-9 (41). Placental tissues exposed to hypoxia-reoxygenation in vitro (the stimulus associated with activation of complement (42)) express and secrete increased levels of TNF-, underscoring the possibility that trophoblasts themselves are a source of proinflammatory cytokines (43).
In summary, our findings demonstrate that TNF- is a critical mediator of aPL Ab-induced injury, and that it is released in response to complement activation. Our results fit into a larger schema consistent with evidence from murine models of myocarditis, sepsis, and rheumatoid arthritis, and suggest that tissue damage induced as a result of complement activation is mediated by production of TNF- (42, 44). However, because blockade of TNF- does not completely protect pregnancies, it is likely that other effector pathways contribute to fetal demise. The variability in the effectiveness of TNF- inhibition that we observed is in keeping with the heterogeneous responses to TNF- blockade seen in murine models of Ab-induced arthritis (45) and in patients with rheumatoid arthritis. Thus, inhibitors of TNF- may provide effective treatment for some patients with APS and recurrent pregnancy loss. In conclusion, therapies directed at TNF-, already in use to treat other inflammatory diseases (46, 47, 48), merit more intense scrutiny in pregnant women with APS, but the therapeutic window for TNF- blockade in pregnancy needs to be carefully defined.
Acknowledgments
We are grateful to our colleagues for generously providing reagents, Dr. Ulrich Feige (Amgen) for PEG sTNFRI, Dr. Michael Holers (University of Colorado Health Sciences Center, Denver, CO) for anti-C5 mAb, and Dr. Paolo Casali (University of California, Irvine, CA) for human aPL mAb. We thank Marta Guerra for help in preparing the manuscript.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 This work was supported in part by a grant from the National Institutes of Health (AI055007) and by the Alliance for Lupus Research and the Mary Kirkland Center for Lupus Research at Hospital for Special Surgery.
2 Address correspondence and reprint requests to Dr. Jane E. Salmon, Hospital for Special Surgery, 535 East 70th Street, New York, NY 10021. E-mail address: salmonj@hss.edu
3 Abbreviations used in this paper: APS, antiphospholipid syndrome; aPL, antiphospholipid; aPL-IgG, human IgG containing aPL Abs; NH-IgG, normal human IgG; PEG, polyethylene glycol; sTNFRI, soluble TNF-R type I; aCL, anticardiolipin.
Received for publication July 8, 2004. Accepted for publication October 4, 2004.
References
Wilson, W. A., A. E. Gharavi, T. Koike, M. D. Lockshin, D. W. Branch, J. C. Piette, R. Brey, R. Derksen, E. N. Harris, G. R. Hughes, et al 1999. International consensus statement on preliminary classification criteria for definite antiphospholipid syndrome: report of an international workshop. Arthritis Rheum. 42:1309.
Out, H. J., C. D. Kooijman, H. W. Bruinse, R. H. Derksen. 1991. Histopathological findings in placentae from patients with intra-uterine fetal death and anti-phospholipid antibodies. Eur. J. Obstet. Gynecol. Reprod. Biol. 41:179.
Magid, M. S., C. Kaplan, L. R. Sammaritano, M. Peterson, M. L. Druzin, M. D. Lockshin. 1998. Placental pathology in systemic lupus erythematosus: a prospective study. Am. J. Obstet. Gynecol. 179:226.
Girardi, G., J. Berman, P. Redecha, L. Spruce, J. M. Thurman, D. Kraus, T. J. Hollmann, P. Casali, M. C. Caroll, R. A. Wetsel, et al 2003. Complement C5a receptors and neutrophils mediate fetal injury in the antiphospholipid syndrome. J. Clin. Invest. 112:1644
Mellor, A. L., D. H. Munn. 2000. Immunology at the maternal-fetal interface: lessons for T cell tolerance and suppression. Annu. Rev. Immunol. 18:367.
Chaouat, G., E. Menu, D. A. Clark, M. Dy, M. Minkowski, T. G. Wegmann. 1990. Control of fetal survival in CBA x DBA/2 mice by lymphokine therapy. J. Reprod. Fertil. 89:447
Raghupathy, R.. 2001. Pregnancy: success and failure within the Th1/Th2/Th3 paradigm. Semin. Immunol. 13:219.(Jessica Berman, Guillermi)