Cutting Edge: Cytokine-Dependent Abortion in CBA × DBA/2 Mice Is Mediated by the Procoagulant fgl2 Prothombinase¹

Spontaneous abortions (resorptions) in DBA/2-mated CBA/J female (CBA/J × DBA/2) mice is thought to represent a rejection of the semiallogeneic fetoplacental unit by activated NK cells and activated macrophages (1, 2). These cells infiltrate maternal mesometrial decidua at the site of implantation. Quantitation has been accomplished by carefully counting the number of asialoGM1⁺ NK cells and F4/80⁺ MHC II⁺ cells per microscope field (2). Infiltration begins on day 6.5 of gestation, 2 days after implantation has occurred, and abortions begin after day 8.5 of pregnancy (2, 3). The frequency of implantation sites with such an infiltrate is proportional to the percentage of embryos that resorb (2). Murine resorptions are characterized by focal necrosis at the junction of the fetal trophoblast with decidua, an infiltrate of polymorphonuclear leukocytes (with some lymphocytic cells) at sites of necrosis and along the walls of large vessels in decidua, and by thrombosis and hemorrhage (2, 4, 5, 6). The cytokines TNF-α and IFN-γ play an important role in abortions, as their administration increases the abortion rate and specific antagonists decrease the abortion rate (7, 8, 9, 10). It has been proposed that macrophage-derived TNF-α stimulates NK cells to produce γ-IFN, which further activates the macrophages, as occurs in the early defense response to infectious agents (1, 11). There are other potential sources of TNF-α and IFN-γ, and systemic Th1-type responses may cause abortions via augmenting levels of such cytokines (12); IL-2 may also cause abortions by contributing to NK-macrophage activation at the fetomaternal interface (7). Pregnancy tends to bias T cell responses along a Th2 pathway, and alloimmunization against the paternal H-2^d Ag also shifts the intracellular cytokine phenotype of γδ T cells in the uterus away from Th1 and toward Th2 (13, 14).

The mechanism by which the implanted embryo is killed is unknown. The current paradigm holds that the fetal trophoblast cells, which form the interface between the embryo and maternal tissues, are damaged or killed (1). Trophoblast is not susceptible to lysis by TNF-α, NK cells, or macrophages, but may be killed by cytokine-activated NK cells (i.e., LAKs)⁵ and may undergo apoptosis in response to TNF-α and IFN-γ in vitro (1, 15). Haddad et al. (16) have recently suggested macrophage-derived nitric oxide may mediate lethality and that the role of NK cells is to produce γ-IFN to activate the macrophages. This makes sense because LAK generation requires IL-2, a cytokine not detected in decidua of DBA/2-mated CBA/J mice, and administration of aminoguanidine, an inhibitor of inducible nitric oxide synthase, prevented abortions in this system (16). On the other hand, IL-12 may substitute for IL-2 in generating LAKs from NK cell precursors, and nitric oxide, a short range mediator, may act primarily by increasing the local cytotoxic activity of TNF-α (17, 18). Aminoguanidine also impairs polyamine metabolism and interferes with macrophage activation (19). Further, an i.p. injection of anti-asialoGM1 Ab after day 8.5, when abortions have already begun to occur, stopped further embryo deaths (3); one would not expect production of NO by activated macrophages to be arrested so quickly, and the data suggest direct NK cell-mediated killing.

We have directly tested the role of cytokines TNF-α and IFN-γ and of NK cells and macrophages in causing abortions, using in vivo cell depletion techniques and mice deficient in the response to IFN. We show that neither NK cells nor macrophages are required for abortions, that the cytokines act on the mother and not on embryonic trophoblast, and that the embryos die from ischemia due to activation of vascular endothelial cell procoagulant, which causes thrombosis and inflammation. This appears similar to the mechanism whereby TNF-α causes ischemic necrosis of nonantigenic tumors (20).

Inbred mice of strains CBA/J and DBA/2 were obtained from Iffa Credo, l’Arbresle, France. C57BL/6J and DBA/2 mice were obtained from The Jackson Laboratory, Bar Harbor, ME. C57BL/6 mice with knockout of the IFN response element, IRF-1, were generated as previously described (21) and bred in at the Ontario Cancer Institute, Toronto, Canada. CBA/J mice were maintained in the Paris colony under conventional open-top wire cage conditions with food and water ad libitum and a 12-h light-dark cycle. Mice in the Toronto colony were maintained in a barrier facility. Female CBA/J, C57BL/6, or C57BL/6 IRF−/− mice were mated by overnight cohabitation with a DBA/2 (IRF1+/+) male, and the morning of sighting a vaginal plug was defined as day 0.5 of pregnancy.

Pregnant CBA/J mice were depleted of NK cells by i.p. injection of 0.2 ml rabbit IgG anti-asialoGM1 Ab (Nordic, Tebu, St. Quentin les yvelines, France) on day 6.5 of gestation; PBS was used as a control as it has been previously shown to be equivalent to nonimmune rabbit IgG (1). Macrophage depletion was performed by i.p. injection of 100 mg/kg silicon dioxide (Sigma, St. Louis, MO) twice a week for 4 wk before mating, as described in (22). Affinity-purified rabbit IgG neutralizing Ab to mouse procoagulant (fgl2 prothombinase) was prepared as previously described (23, 24); the mice were given an i.p. injection of 0.2 ml of a 1/50 dilution of a 5.5 mg/ml preparation of anti-fgl2 or control rabbit Ab each day beginning on day 3.5 of gestation. Hormonal support of pregnancy sufficient to replace ovarian function was provided in some experiments by injecting 6.7 ng of 17β-estradiol + 1 mg progesterone (Sigma) in 0.1 ml oil i.m. daily beginning on day 4.5 of gestation (25). One hundred micrograms of rat monoclonal IgG2b anti-mouse granulocyte Ab RB6-8C5 (PharMingen, San Diego, CA) (26) or isotype control was injected i.p. on day 6.5 of pregnancy. TNF-α (Ref. 7 and R&D Systems, Minneapolis, MN), 1000 or 2000 U, and/or murine rIFN-γ (Ref. 7 and R&D Systems), 1000 U, was injected i.p. on day 7.5 of pregnancy. In some experiments, rIL-10 was used as described (9). On day 13.5 of pregnancy, the mice were killed and the number of resorbing and healthy embryos was counted. In some experiments, the uteri were snap frozen, 5-μm sections were cut, and the tissues were stained with rat monoclonal F4/80 Ab (Caltag, Tebu Nordic) to macrophages. Briefly, tissue sections were incubated with a 1/30 dilution of F4/80 in PBS for 30 min, and binding was detected using peroxidase-streptavidin with biotin-labeled anti-rat IgG2b second Ab (Serotec, Kidlington, Oxford, U.K.) (27).

Four to ten mated mice per treatment group were used. The significance of differences in the pooled resorption rate was tested by χ² or Fisher’s exact test where appropriate.

Experiment 1, Table I, shows that i.p. injection of TNF-α boosted the abortion rate of DBA/2-mated CBA/J mice in a dose-dependent manner. If the mice had received anti-asialoGM1 Ab treatment, the background rate of abortion decreased, as expected (1, 3, 7), and TNF-α no longer had a significant effect. These data supported the model TNF-α → NK → activated NK → kill embryo. To ensure adequate levels of endogenous macrophage-derived TNF-α, we repeated the experiment and added IFN-γ. Experiment 2 shows that IFN-γ alone boosted the abortion rate in PBS-pretreated mice to the level achieved with TNF-α, and addition of TNF-α had no significant supplementary effect. In NK cell-depleted mice, IFN-γ failed to boost abortions. This suggested that the model IFN-γ → macrophages → activated to produce NO → embryo death was not correct. However, when IFN-γ and TNF-α were administered together, >80% of the implanted embryos aborted. This suggested an obligatory synergy/codependence; in NK cell-depleted mice, TNF-α does not work because the NK cell source of IFN-γ has been eliminated, and IFN-γ fails because macrophages dependent on NK cell-derived IFN-γ have stopped producing TNF-α, and the i.p. injected cytokine does not stimulate TNF-α production quickly enough for both cytokines to be present simultaneously. A direct NK or macrophage killing mechanism seemed an unlikely explanation for abortions. To further test this idea, we repeated the experiment using macrophage-depleted mice. Experiment 3 shows that macrophage depletion reduced the abortion rate. It can be seen that macrophage depletion had no significant effect on the 80% abortion rate produced by injecting TNF-α + IFN-γ. Tissue staining for F4/80⁺ macrophages confirmed that the silica treatment had been effective and the cytokine treatment did not cause a macrophage infiltration (data not shown). TNF-α + IFN-γ may act synergistically to suppress production of essential gestational hormones by the ovary (28), and such an inhibition could cause abortions (4, 25). However, ovarian failure should have caused 100% abortions (4). Further, when we gave hormone replacement therapy as described in Materials and Methods (25), there was no effect on either the background rate of abortion or the high rate of abortion produced by 2000 U TNF-α + 1000 U IFN-γ (35/41, 86%, n = 5, control group vs 37/45, 82%, n = 5, cytokine-treated group; not statistically different).

TNF-α and IFN-γ act synergistically to induce apoptosis in human trophoblast cell cultures (15). The results shown in Table I could be explained by a direct apoptotic action on trophoblast. However, the cytokine CSF-1 is present in vivo, which may abrogate the apoptotic effects of TNF-α and IFN-γ (15, 29). To test for a direct effect on trophoblast in mice, we mated IFN-γ-unresponsive IRF1−/− females (21, 30) to DBA/2 (+/+) males so that fetal trophoblasts express IRF1 but maternal tissues do not. The IRF1 knockout gene was on the C57BL/6 rather than the CBA/J background; C57BL/6 mice have low spontaneous abortion rates and lack the NK/γδ T cell infiltrate seen in CBA/J mice (14). Nevertheless, as shown in Table II, C57BL6 × DBA/2 pregnancies aborted dramatically when the cytokine treatment was given. This validated the conclusion drawn from Table I that an NK cell/macrophage infiltrate in decidua was not necessary for cytokines to trigger abortions, supporting the hypothesis that the role of the NK/γδ T cell infiltrate, which only occurs in response to CBA/J × DBA/2F1 trophoblast, is to enhance the levels of IFN-γ and TNF-α required to trigger abortion. By contrast, pregnant IRF−/− females had low background abortion rates and were completely resistant to TNF-α + IFN-γ. These data suggested that the cytokines might be acting on the mother in the CBA/J × DBA/2 model and not on trophoblast to induce abortions.

Since neither macrophages nor NK cells seemed necessary for TNF-α + IFN-γ to act, the most logical target appeared to be the maternal uterine vascular endothelium. These cytokines stimulate surface expression of procoagulant (fgl2 prothombinase, which is distinct from tissue factor); clotting initiated by fgl2 or tissue factor is known to lead to ischemic damage in a variety of inflammatory disease models such as hepatitis and endotoxic shock (23, 24, 31). Table III shows the effect of treatment of pregnant CBA/J × DBA/2 mice with Ab to fgl2. The background rate of abortion was reduced to 4%, similar to the frequency of chromosome abnormalities in mouse embryos (32). Further, the high abortion rate induced by TNF-α + IFN-γ was almost completely prevented.

Granulocytes are commonly seen in resorption sites and contribute to lysis of TNF-α + IFN-γ-activated endothelial cells (33). Nitric oxide, a mediator implicated in abortions in CBA × DBA/2, and IL-1, an additional cytokine participating in these abortions, also contributes to granulocyte → endothelial cell damage (8, 16, 33). Thrombin is known to activate endothelial cells to release IL-8, and IL-8 promotes infiltration of granulocytes, which digest tissues (5, 33, 34). To determine whether granulocytes contributed to abortions as predicted, we injected pregnant mice with a monoclonal anti-granulocyte Ab that is known to block granulocyte-mediated tumor rejection in vivo (26). As shown in Table IV, anti-granulocyte Ab partially reduced the spontaneous abortion rate and significantly abrogated the effect of TNF-α + IFN-γ. As a control, a significant reduction in the spontaneous abortion rate similar to that achieved with anti-fgl2 was confirmed using rIL-10 as described (9). A similar result was obtained with anti-granulocyte Ab in a second experiment (which did not include rIL-10) and combined with the data in Table IV, the control spontaneous abortion rate of 30% (resorptions/total implants = 29/96) was significantly reduced to 13% (11/87) (p < 0.01, by χ²), TNF-α + IFN-γ boosted abortions to 87% (77/89) (p < 0.001) and anti-granulocyte Ab reduced the latter to 13% (13/98)(p < 0.001). The partial effect of anti-granulocyte Ab (reduction of abortion rate to 12–16%) could reflect a contributory but non-obligatory role of these cells for abortion or an incomplete cell depletion in vivo by the dose of Ab used.

The neovascular supply to tumors and implanted embryos appears uniquely sensitive to the effects of endotoxin (which triggers TNF-α production) and to injection of TNF-α. There were no signs of systemic illness noted in our cytokine-treated mice. It is unknown why the vascular supply to the embryo is so sensitive, but teleologically, autoamputation via ischemia provides a primitive mechanism for preserving the function of those organs required for survival during times of stress. An obligatory synergy of TNF-α and IFN-γ, which we noted in our mice after in vivo depletion of asialoGM1⁺ cells or of macrophages, has been shown for lethality in endotoxic shock (35). Further support for the idea that more than one cytokine must be present to abort the embryo has been provided by a study of TNF-α in CBA/H mice in which concomitant infection was necessary (36); one may suggest that the role of infection is to increase the production of other cytokines needed to deliver the termination signal to the uterine vasculature. An obligatory synergy in induction of premature parturition has also been reported (37). Concerning protection against abortion, TGF-β₂-producing γδ cells in the uterine lining appear to be important (14, 38). TGF-β₁ protects the endothelium from granulocyte-mediated reperfusion injury (39). However, it is unclear whether TGF-β₁ and/or TGF-β₂ explain why 15 to 20% of implants survive TNF-α + IFN-γ or whether another mechanism is involved. TGF-β₂ and IL-10 may prevent abortions in the CBA/J × DBA/2 model by acting on NK cells, macrophages, and other potential sources of Th1-type cytokines rather than by altering the endothelium (9, 14).

An increase in activated blood CD56⁺16⁺ NK cells in uterine endometrium/decidua, Th1 cytokine responses to trophoblast, and a deficiency of TGF-β₂-producing nonclassical CD56⁺16⁻ decidual NK/γδ cells has been reported in approximately 50% of women with recurrent spontaneous abortions (40, 41, 42, 43, 44, 45). In humans, in contrast to the mouse, approximately one-half of recurrent abortions may be attributed to chromosome anomalies in fetal trophoblast, and NK/γδ T cells would not be expected to have a role in the mechanism of such losses in contrast to the abortion of chromosomally normal embryos (45). Interestingly, Kohut et al. (46) have recently reported that women who are more likely to abort chromosomally normal embryos show more marked histologic evidence of thrombosis and inflammation (villitis and vasculitis). Vascular injury, rather than direct NK/macrophage cytotoxicity on fetal trophoblast, may explain abortion of normal embryos in both mice and humans. Prothombin activation at endothelial and trophoblast cell surfaces may also occur in abortions due to antiphospholipid Abs (47); therefore, coagulation triggered by different initiators may represent a final common pathway for pregnancy termination.