Immunopathology of Early Pregnancy

One of the important discoveries in reproduction is that the reproductive process uses for its own purpose lymphokines and cytokines which were initially defined as being of immunologic nature, first traced within the immune system. This is not very surprising, since most of these molecules are more signaling factors and communication molecules between cells of various lineages than molecules strictly restricted to the immune system, such as immunoglobulins (Ig) and T-cell receptors. The field has now evolved from the idea of a host-graft relationship to the concept of an integrated dialogue between the mother and the conceptus, more akin, for some, to a host-tumour relationship.

that the reproductive process uses for its own purpose lymphokines and cytokines which were initially defined as being of immunologic nature, first traced within the immune system. This is not very surprising, since most of these molecules are more signaling factors and communication molecules between cells of various lineages than molecules strictly restricted to the immune system, such as immunoglobulins (Ig) and T-cell receptors. The field has now evolved from the idea of a host-graft relationship to the concept of an integrated dialogue between the mother and the conceptus, more akin, for some, to a host-tumour relationship.
Yet, in fact, the paradigm of acceptance and/or rejection of the foetus by its mother is still heavily embedded in the thinking of many reproductive immunologists. This is not surprising, since the trend of research that led to modern reproductive immunology was initiated by Sir Peter Medawar's disquisition on "the riddle of the foetal allograft. ''1 It is therefore not all disconcerting that, in 1997, one of Medawar's colleagues, Leslie Brent, in his recent book A History of Transplantation Immunology, still depicts the materno-fetal relationship as "Nature's (almost) perfect allograft, ''z nor that pioneering work in the field is due to another collaborator of Leslie Brent and Peter Medawar, Rupert E. Billingham. 3 However, the paradigms have evolved. As you will see in this brief presentation, there is still room for discussing, and re-envisaging, with the use of murine models and data in humans, the concept and mechanisms of "rejection" of the foetal allograft. But the question of the tolerance of the concep-tus has evolved. First, we have known for a long time that we can and must dissociate local and systemic events, but this idea has not yet fully penetrated the thinking of the general public. It is possible to observe rejection of paternal strain tissues after preimmunisation of the mother without compromising pregnancy 4,s (in fact, the babies born in such situations are in most circumstances heavier), and we have personally verified that. 6 It ensues that systemic immune responses (which we do not want at all to negate, having taken a rather indisputable part in their discovery7) are, at best, a testimony of maternal recognition of the foeto-placental unit. Indeed, pregnancy is perfectly possible in the absence of humoral (antibody) response, s,9 Indeed, if tolerance to the foetus was alloantibody mediated, it would be an exception, because tolerance is MHC restricted, and the Nobel prize awarded to Zinkernagel and Doherty recognizes the consequences of that phenomenon. With rare exceptions, antibodies are not MHC restricted, and attempts to say that tolerance was antibody mediated should be in immunology history. This is not completely true, since there is not yet a satisfactory explanation of the enhancement phenomenon, possibly because of the controversy over IgG1/ IgG2 in mice, which uselessly shifted the debateenhancing antibodies in rats, are part cytotoxic.
Similarly, we do not negate, of course, the existence of systemic suppresser cells, 7 but pregnancy is perfectly viable in animals experimentally depleted of these. 1 The repeated enunciation in France that pregnancy depends on an alloantibody and suppresser T-cell-mediated immune response is simply outdated since around [1979][1980][1981], and claims that unfortunately are expressed in mainstream immunology meetings and in print that "reproductive immunology has to teach the mainstream immunologists how the immune system works" as a "facilitation reaction" with "important role for enhancing antibodies.., and suppresser cells" have disastrous effects: First, mainstream immunologists repeatedly dismiss, in a more and more irritated fashion, such preposterous advances, and, as such, would consider as a backlash the whole field as not serious; Second, those claims have led some clinicians to exacerbate the role of maternal antibodies, clouding unconsciously the issue of early pregnancy loss.
Third, they distract from the main advances that have indeed been going on, leading to molecular understanding of the field.
For indeed, there has been progress, such as Tom Wegmann's seminal idea that possibly we should re-envisage our view of the immune system in pregnancy: instead of seeing the foetus as confronted by a threat of rejection by its mother--the "immunotrophic" concept postulated from a discovery made with the routine abortion model (the CBA DBA/2 system), 11,1z we recognize that immune reaction could be in some cases beneficial for fetal survival. 13 Then, it ensued the demonstration in vitro, then in vivo, that indeed T-cell derived cytokines were growth factors for the placental trophoblasts, 14 followed by in vivo demonstration of T-cell control of placental growth ls,6 and treatment of a murine abortion model (CBA x DBA/2) by purified or recombinant cytokines. 7 In the very same period, David Clark had made the discovery that this precise model (CBA x DBA/ 2) was deficient in local, decidual associated suppresser factor, 8 and later on showed that this factor was molecularly related to  We are now faced with events occurring mostly locally, and with the concept of a cytokine network at the foeto-placental interface. The events are in fact different whether one looks at parturition (where there is good evidence for involvement of interleukin [ILl-l, IL-6, IL-8, and tumor necrosis factor [TNF] e-z,zs) or the established pregnancy which is indeed immunologically characterised by transient acceptance of relatively weakly immunogenic tumor grafts of parental strain origin in the first pregnancy, linked to a T-cell anergy state and suppresser T-cell mediated multi-pregnancy induced tolerance to paternal alloantigens. [26][27][28][29] In the latter case, the close examination by polymerase chain reaction (PCR) of cytokine profiles shows that it is more complicated than initially thought, with variations throughout pregnancy of which the physiological significance is yet unclear, 3 and which we will not detail here in the human. Similarly, the profiles are not simple in mice. 131 And we now come to the topic of this communication, early pregnancy. Nowhere is the understanding of the cytokine network more interesting, and nowhere has it led to more surprising discoveries, opposed to the paradigm of the acceptance of the foetal allograft, and so far away from the initial framework (which aas indeed once useful) of the facilitation reaction.
Early implantation requires "inflammatory" molecules, the consequence, in part, of a quasiinflammatory reaction locally, followed by an immediate return to non-inflammatory conditions. It is probable, from what we know in animal models, that each of these steps are absolutely required for further successful pregnancy. But before that, a word is required about the antigenicity of the embryo in mice and humans.

AN ALLOGRAFT?
It is a tautology, a "Lapalissade" to say that to be alloantigenic, an embryo must express foreign tissues antigens. This is not totally true. Failure to express major histocompatibilty antigens has another counterpart. It activates cells which are what Charlie Loke calls the most primitive immune system, 31 the Natural Killer cells (NKs). That it was so is a relatively recent discovery, which owes a lot to the theory of the "missing self" originally proposed by Klas Karre. 3z Therefore, it ensues that the trophoblasts cannot theoretically be completely neutral, otherwise there would be an NK-mediated rejection reaction: the NKs express two sorts of receptors, which are not yet fully understood, albeit an impressive body of evidence is accumulating (see Immunological Reviews 158). Killer Activating Receptors (KARs), whose engagement, in a 74 INFECTIOUS DISEASES IN OBSTETRICS AND GYNECOLOGY neoclassical fashion, triggers NK lytic pathways (and possibly others, such as cytokine secretion, especially when the Killer Inhibitory Receptors [KIRs] are engaged). The second, and most important, are KIRs. The KIRs arc activated when engaged by the reconnaissance of the presence of MHC molecules (most often in a given minor histocompatibility antigen context), which explains "easily" some "oddities" which we published for the sake of honesty on the CBA xDBA/2, system and which were then rejected by proponents of the MHC-T-cell duo theory (proponents for whom we, incidentally, have the greatest respect), triggered so as to inhibit NK lytic pathway.
An added level of complexity for immunologists and clinicians is that while MHC is (relatively) conserved through evolution as a member of the super Ig family, KIRs and KARs are not at all related between mice and humans. Why? Somehow, the trophoblast faces a dilemma: express MHC, which might trigger a classical. T-cellderived rejection reaction, or repress completely MHC, which will defuse the KIRs and trigger NKmediated rejection.
This dilemma is partly solved in the human. The trophoblast expresses, as far as its extravillous component is concerned, a specific antigen, HLA-G, with no variations (or a few amino acids) from one individual to the other, but it is now known that it has some variability. 3s HLA-G has little peptide presenting capacity (but it does have it!), and, as a truncated molecule, it is not recognised as "polymorphic." (It has few antigenic disparities between individuals, if any, that can be "seen" by T-cell  The situation would thus seem simple in humans: the extravillous trophoblast is where interaction with decidual lymphocytes takes place and therefore where NK and NK-like cells accumulate and are defused. The lack of classical MHC molecules would render the trophoblast a non-target and a non-inductor of a classical T-cell-mediated response. In addition, there is no bioactive IL-2 in decidua, albeit an IL-2-1ikc material has been traced. 39 The situation is more complex than that, even for those who would like to make an exception in the case of primates: in humans, the expression of HLA-G is restricted to extravillous cytotrophoblasts, villous cyto-, and syncytiotrophoblasts that are uniformly MHC-negative, but a mirror situation is observed in baboons, and in the rhesus monkey, HLA-G homologue is a pseudo gene (although its function could be exerted by a yet uncharacterised protein). 4,41 So, this is already a snag.
Two further snags are to be discussed: First, how could the totally MHC-negative human syncytiotrophoblast, a fact not dismissed since its description by Faulk, 4z,43 not trigger NK-mediated lysis? This is as yet unexplained by proponents of the HLA-G paradigm. Second, we now know that HLA-C is expressed by extravillous trophoblast. 37 It has a restricted polymorphism and interacts with NKs, but the theory must somehow be twisted to make it act as purported for the monomorphic HLA-G, for despite a certain paucity of reports, there are T-cell-mediated responses to HLA-C.
Finally, in all species other than the primates, there is expression of classical, polymorphic class I molecules, generally precisely on the layers of the placenta that confront the maternal immune system (44,45,46). In mice K, D, and L are expressed on the spongiotrophoblast, perfectly accessible to antibodies and cells. The same is true in horses for equine leukocyte antigens, in pigs for swine leukocyte antigens, and, according to Twink Allen, in elephants.
There are three alternative hypotheses for the problem: One is that man is man, or, as Peter Johnson wrote (The Immunologist, 1996:4;p172), "differences in placentation between humans, rodents, and other species can make direct comparisons largely meaningless." The second one is, as Leslie Brent says when discovering the status of HLA-G in other species, that HLA-G suppresses as would any proper MHC-ligand in the proper cells (that is, by the sort of experiments in which the missing self was discovered to be true) and that HLA-G has likely another function, since there is no equivalent in mice, etc, that warrants its evolutionary appearance (a specific peptide or hormone intracellular carrier, for example, yet uncovered).
The third one is that there are in other species functional equivalents of HLA-G, which could reside in the already discovered monomorphic or poorly polymorphic MHC class molecules.
None of these theories, incidentally, is mutually exclusive.
However, for class II (I-a in mice, HLA-DR in humans), the situation is more simple. In no species do the trophoblast layers express, class II alloantigens, and expression of class II on trophoblasts induced by azacytidine results in regular abortion in mice. 47 Why? The question is unsettled, since the mechanisms have not been explored, but one should recall here in the same vein that expression of class I in mice is restricted to spongiotrophoblast, with no expression being found on labyrinth.
Indeed, derepression of class I expression in the labyrinth is found in interferon y induced abortion, but one does not know if it is a consequence or a cause of abortion. 48 So, we are in quicksand about the "allograft status of the embryo." As far as the preimplantation embryo is concerned, it is simpler to say that there is no class I expression (nor class II) whatsoever on the gametes, blastocysts, extracellular cell mass or ectoplacental cell cone (EPC) in any species, including, when studies could be done, humans. 49 This is somehow troublesome for some aspects of the recent HLA-G theories: the negative EPC cells certainly are facing a very important NK accumulation. so And yet, the EPC resists perfectly well NKmediated lysis. Conversely, aborted embryos in many species at that early stage display MHC expression on EPC cells.
The solution may be simple: at that stage, the cells of the EPC are intrinsically resistant to cellmediated lysis, sl-s3 most probably because cellmediated death involves target cell participation and the pathways of apoptosis. Like other cells secreting high amounts of steroids, trophoblasts are resistant to steroid-induced apoptosis and thus to cell-mediated death. However, they are sensitive to LAKCs mediated cell death, s4 but for that we will see that there is in addition a lot of immunosuppressive material in the vicinity, and that abortion might not be mediated by trophoblast death. Such an immunosuppression takes place after im-plantation, which on the contrary calls for inflammation, and is a Thl response.

INFLAMMATION AND IMPLANTATION
In France, the early stages of pregnancy are difficult to study in humans, despite advances in medically assisted reproduction. The Loi Huriet requests full informed consent before taking a biopsy sample, and in the very "sensitive" area of reproductive tract organs, this is quite a deterrent, as opposed to the ease with which blood samples are obtained.
In the implantation period, where some events are very transiently associated with blastocyst adhesion and initial invasion of uterine walls, it is, of course, inconceivable and unlawful to deliberately transfer an in vitro fertilisation (IVF) for the sole purpose of aspiration or, even worse, (total) removal of uterine tissues for immunological investigations. As stated, animal models are not necessarily fully pertinent, and thus Y.W. Loke said at the 15th World Congress on Fertility and Sterility in Montpellier that the only valid model for human pregnancy is "the human species itself." But, this does not solve fully the problem of the periimplantation period. Macaques are not fully relevant, because of antigenic disparities, like baboons, at the trophoblast level with human situation, and chimpanzees are an endangered and protected species of extremely limited availability. For these reasons, in vitro alternative models have been developed, which may or may not be totally relevant, complemented by studies in animals, mostly in mice.

Animal Studies
The studies performed in animals have revealed a salient and totally unexpected aspect contrasting with the established pregnancy: the preimplantation uterus undergoes an "inflammatory-like" reaction, with transient influx, after mating in mice and rats, there of lymphocytes and macrophages, and seminal fluid is required, since this is not seen in females with fallopian tubal ligation or in vasectomised males, ss-s7 As expected, an increased secretion of IL-1 or, IL-1 13, TNF or, IL-6, and mRNA is detected in the uterus, paralleled by the pregnancy-associated continuous rise in production of CSF-1 which starts by then. ss's9 Except for CSF-1, those inflammatory cytokines return to basal levels 76 INFECTIOUS DISEASES IN OBSTETRICS AND GYNECOLOGY by days 3 and 4, but CSF-1 secretion continues to increase throughout pregnancy. 58,s9 In rats, local injection of the PAF-acether antagonist BN 50081 completely prevents implantation, whereas the untreated contralateral horn is fully receptive, ss,s6 Three cytokines are, in our opinion, crucial, as established in mutant or "knock-out" animals: CSFs, IL-1, and LIF. CSFs (CSF-t, S-CSF, -CSF) CSF-1 deficient oSP (osteopetrotic, or op/op) mutant mice have a profound pregnancy defect, and males have also reproductive functions. CSF-1 receptor, c-fins, is expressed after the 2-4 cell stage until the EPC outgrowth and later the spongiotrophoblast and is also found at very high levels in the maternal then the primary decidual zone and finally in the sole decidua basalis. 6z S-CSF and its receptor (c-kit, which maps in the dominant white spotting W locus in the mouse), is also important, since mutations in the W locus t result in sterility in mice, often associated with abnormal embryonic development. 6z GM-CSF was the first cytokine with IL-3 involved in the immunotrophic theory 13-16 and is found in the preimplantation uterus. Tartakowsky has shown that culture of CBA x DBA/2 embryos (a murine abortion model 11,z) in GM-CSFor CSF-l-containing medium partly corrects the deficient implantation rate seen when transferring these embryos in CBA/J mothers. 63 GM-CSF receptor is expressed in the EPC and later spongiotrophoblast cells and in the early implantation uterus. 64'65 Recently, in an elegant series of experiments, Robertson implied the inflammatory response and more peculiarly the GM-CSF. In agreement with studies conducted to prove the immunotrophic theory, s,16,66 she found that GM-CSF-deficient mice have smaller litter sizes and placental weights than controls and implies GM-CSF and components in the semen ejaculate in hyporesponsiveness to paternal alloantigens. 67 IL-I IL-1 might play a mandatory role in pregnancy. According to Simon and Polan, 68 IL-1 receptor antagonist completely prevents successful implantation in mice. However, Colin Stewart et al. (Stewart, personal communication and answers at many meetings) have examined at the Roche Institute for Molecular Biology IL-1 deficient mice with perfectly normal reproductive function, nor did we obtain implantation failure in mice by in vivo injection of neutralising anti-IL-1 antiserum.

LIF
By gene knock-out technology, Stewart has elegantly shown that maternal production of LIF is mandatory for successful implantation: LIFdeficient mice, obtained by "gene knockout" are fertile, but sterile. LIF-/LIFembryos implant in normal CD1 recipient mice, but LIF +/LIF + embryos do not implant in LIF-/LIFmice. The defect is corrected by recombinant HILDA LIF via an osmotic pump. 69 Human Studies IL-I IL-1 a, 13 expression is found in human endometrial epithelium throughout the menstrual cycle, reaching a peak in the late luteal phase. 7,7 Immunohistochemistry shows that IL-1 receptor antagonist in human endometrial cells is at higher levels during the follicular phase than during the early and mid luteal phases. 72 IL-1 is also found in trophoblasts and IL-1 receptor type I on syncytiotrophoblasts, suggesting an autocrine and paracrine role of IL-1 in human implantation. 7 Indeed, it has been reported that IL-1 is a modulator of the decidualisation itself, 74 and it is also controlling hCG production by human trophoblasts in culture. 7 The production and role of IL-1 by early human embryos is controversial; some authors made it a predictor of successful implantation in humans, 76,77 while others did not detect it at all. 34,78 Differences in culture media, conditions, etc. obviously could be invoked. In a co-culture of endometrial epithelial cells and or stromal cells with embryos, IL-1, IL-1 receptor antagonist, and IL-1 receptor type were found in about 56% of cells from embryos cultured with epithelial cells but not stromal cells, apparently correlated with pregnancy success. 79 LIF LIF is present in human endometrium, with consensus for its expression in the second part of the cycle, and during pregnancy, this in variance with mice. 8-8z,84 Optimal uterine secretion coincides with the implantation window and expression of   believe we have an objective assay that can complement immunohistochemistry, analysis on cytofuge smears, and in vitro production in explant culture.
It is important, in this context, to emphasize that recombinant human HILDA LIF is available, and women could benefit from substitutive therapy. While we feel confident enough to go forward with such a procedure, the clinicians would prefer we first correct a defect outside pregnancy before going further on.
This leads to questions about the mechanisms of action of LIF and, indeed, the inflammatory response. It is our working hypothesis that the local inflammation is required for optimal expression on placental trophoblast and uterine decidual cells of adhesion molecules, which are necessary for implantation to occur, such as trophoblast laminin and We believe that an important part of early pregnancy loss is due to abnormal processes of this early reaction. It is noticeable that dysregulations are already known, such as in preeclampsia, where trophoblasts do not down-regulate 13 4, nor is o 1 upregulated. 94 In addition, the effects of cytokines on the matrix degrading protease are also to be encompassed in that respect. 31 TNF There is early and temporary expression of relatively high levels of TNF by invading extravillous trophoblasts 9s which could, as seen in other systems, up-regulate adhesion molecules and selectin. 96 This inflammatory reaction, however, has to be then very quickly down-regulated, since during the early post implantation phase it is abortogenic. There are many other cytokines that we will not discuss here, for sake of space, but we cannot not mention here CSF-1 and its receptor, whose programme is very parallel to what is seen in rodents and which has a similar role, 97-11 mainly as placental trophoblast growth factor acts as the other immunotrophic cytokine, IL-3 lz In this commu-   nication, however, we would like to mention tau and placental interferons.

Placental Interferons
Corpus luteum maintenance in ovine does not depend on non-chorionic gonadotropin, but on trophoblastin whose trophoblast secretion occurs in the periimplantation, also named ovine trophoblast protein (oTP). It has pleiotropic activities. 13,14 It exerts a wide variety of effects, such as local antiviral properties, preparation of the uterus for optimal implantation by promoting 2 5 A synthetase activity, and local cytostatic properties, making it an ideal candidate for early immune suppression. It is likely to be involved in control of the Thl/Th2 balance. 1s-8 It is in fact an interferon of the new Th I/Th2 BALANCE AND LOCAL IMMUNOSUPPRESSION As soon as one completes the implantation period, in humans and in mice, an abnormal Thl reaction to trophoblast causes early pregnancy loss, recurrent spontaneous abortion, and infertility. Several mechanisms are involved. For example, the HELLP syndrome (hemolysis, elevated liver function tests, and low platelet counts) is linked to an elevated IL-12 serum level and possibly an enhanced IL-2 level in the serum 11,11 of unclear origin, since despite recent molecular evidence for the contrary112, no IL-2 is assumed to reside in decidua and placenta. This is especially important, since in a variety of animal systems, activated NKs (by IL-2 or IL-4, e. g, LAKCS), are abortifacient, and indeed LAKCS can lyse trophoblast cells in vitro, s4 but not normal NKs or CTls. Thus, there has been a quest as indeed "built-in" in the "allograft model" for local immunosuppression for the mechanisms of early pregnancy loss and its control.
The Thl cytokines, be they IL-2, interferon y, or TNF, are abortifacient in vivo, the latter at any stage of pregnancy, in abortion-prone and nonabortion-prone systems 17'113 '114'116'117 Interferon / and TNF act in synergy. Their control is exerted by a variety of non-cytokinic mediators (prostaglandins, 1-25 OH dihydroxyl cholecalciferol in the decidua, placental low MW suppresser factors), as well as by suppressive cytokines such as TJ6 (the "pregnancy-associated cytokine") which has been cloned initially in mice. Monoclonals against TJ6 are abortifacient at early stages, and deficient expression of TJ6 is reported to be associated with pregnancy loss in human, especially if the defect is on CD 19-and CD S6-positive peripheral and decidual lymphocytes. Such a defect could be predictive of risk of pregnancy loss. s-e Another regulator which is, like TJ6, progesterone dependent is the progesterone induced blocking factor (PIBF), secreted by activated T lymphocytes, lee PIBF alters the Thl/Th2 balance towards Th2 and corrects activators of NKK (poly IC) and natural (CBA x DBA/2) induced abortion. z3-zs Its expression by maternal lymphocytes has been shown to correlate with (perhaps even require) implantation. z6 The decidua-associated transforming growth factor 13-2 analogue 9 is secreted by uterine non-T, non-B cells in mice and decidual transforming growth factor [3-2 secreting cells in humans, lz6 It is a potent immunosuppressive factor, and it has been reported that its production was defective in women with repeated miscarriages, lz6 Finally IL-10 is clearly important, since in animal models pregnancy is very obviously a Th2 phenomenon. e7 In humans, it is secreted by placental trophoblasts and choriocarcinoma, lzs,z9 and, interestingly, its secretion is enhanced by GM-CSF. IL-10 is also produced by isolated decidual cells. 3 In mice, its secretion is more controversial, l'e But, in murine models, IL-10 with IL-3 suppress local action of IL-2 and gamma interferon, biasing the maternal immune response towards a Th2 profile. Deneys (UCL, Bruxelles, Belgium) has shown that in pregnant women, TNF was down-regulated and IL-4 up-regulated when measured in the serum, a pattern not seen in recurrent aborters (Deneys, personal communication). Further proof of the role of IL-10 is the effect observed in infections. Leish-mania biases the responses towards Thl, and there is less IL-10 but more pregnancy failure. 133,134 EFFECTOR MECHANISMS What are the exact mechanisms of early pregnancy loss? It is clear that interferon and gamma interferon are involved, very often as a consequence of either abnormal maternal recognition of the conceptus/semen, or local, undetected "a trs bas bruit" infection, with peculiar insistence on mycoplasma. 17,18, [133][134][135][136] The data of Hill strongly suggest it is also the case in humans, e.g., Thl immunity causes recurrent abortions and sterility, with emphasis on gamma interferon. 137,138 But what are the exact mechanisms?
First, the cells and the "immunological events" need to be reconsidered, even for immunotrophism; interest is leaning now toward NKs rather than T cells. Anne Croy has shown that placentae of NK-deficient mice are grossly hypotrophic, leading to premature foetal death, 139 shifting t.he "immunotrophic" paradigm from T cells to NK cells.
Thus, the production of immunotrophic cytokines might be in fact mostly (but not exclusively) under direct control of NK cells more than T cells, and "allorecognition" of pregnancy might, in fact, be exerted by these NKs rather than in a classical fashion by T cells. In consequence, a hitherto little suspected role could be envisaged for HLA-G: the control of the production of immunotrophic cytokines and not solely the local defusing of NK mediated cytolytic functions (a role that was forecasted by Y W Loke31). In the CBA x DBA/2 model of natural immune abortion, together with the poly IC 12 U model, (an NK activator), 4z we explored the role of NKs in cooperation with P. Kourilsky. 4,43,44 Anti MHC H-2d (BALB/c) alloimmunisation prevents the effect of Poly IC or Poly I Poly C12 U as well as CBA x DBA/2 fetal loss. But, not all selective alloimmunisation prevents the CBA x DBA/2 foetal loss, and the genetic patterns of the effective immunising splenocytes (MHC restricted, minor loci dependent), albeit apparently representing a single trait, lz,41 are so odd to comprehend in terms of "classical" T-cell recognition (BALB/c works in CBA xDBA/2, B10,D2 does not-both are H-2d) that some T-cell and MHC proponents preferred to dismiss the data as non reproducible (which they were not, being reproduced at present by more than 20 labs, with significant developments) or (we quote) "absolute bullshit." But, we then demonstrated that immunisation with Kd or Ld transfected L cells prevented resorptions and used that system to study the effect ofamino acid mutations on the border of the pocket of the MHC molecule in that system. These were shown to affect recognition by maternal cells promoting optimal foetal survival when immunisation was performed using H-2 d mutant transfected L cells. We then used various mutations that were artificially induced and localized on the border of the molecule's groove or in the peptide binding site. Some H-2d mutations on the helixes are recognized as foreign to H-2d and afford foetal protection against Poly IC. With P. Kourilsky, we wondered whether NK recognition of discrete determinants on the MHC molecule was not involved and, with help of Sylvie Delassus, Jean Pierre Abastado, Claude Roth, and Jos Even, we immunised CBA/J with L cells transfected with H-2d molecules mutant on positions 65, 69, which do not affect peptide presentation and 114 as a control. These protected against CBA x DBA/2 fetal loss, and, more important, enhanced IL-10 production in the supernatant of placental and decidual explants as already described using a commercial ELISA of high sensitivity, while treatment with asialo GM1 abolishes this effect.
At this stage, we propose the following working hypothesis: the NK cell repertoire is heteroclitic and consists of the recognition of selected altered domains of MHC molecules, whether that alteration results from mutations or allosteric conformational changes induced by (in the context of) background genes. Such a recognition either switches lytic/cytostatic pathway (macrophage activation, NK activation, TNF secretion, and balance of the CD4 system towards a Thl profile) to IL-4 and IL-12 secretion by the NKs or pushes the system towards IL-10 secretion, and secretion by NKs or under the influence of NKs of the so-called "immunotrophic" cytokines. Thus, the concept would integrate both immunotrophism and the Thl/Th2 balance.
As far as the effector mechanisms are involved, it is already well known that TNF and interferon are causative and act in synergy, 8,4s,146 but other mediators have been implied, such as nitric oxide released by activated macrophages. 47-1s There is bSignificant increase in abortion rate, P < 0.005 by X 2. cSignificant increase in abortion rate compared to PBS control, P < 0.005 by 2; significant difference compared to lower dose of TNF-c, P < 0.05. dSignificant reduction in abortion rate by anti-asialo GMI antibody compared to PBS control, P < 0.005 by X 2. eNo significant boosting of abortion rate compared to PBS injected anti-asialo GMI-treated group. fResult from two independent experiments giving same result have been pooled. g/-interferon (--IFN) significantly boosted abortion rate, P < 0.005 by X 2. hTNF-c was given at 1000 v and 2000 in separate experiments with /-IFN and gave similar results; the data have been pooled for ease of presentation. The abortion rate was significantly boosted, P < 0.005 by X 2. Untreated CBA/J female mice mated to DBA/2 males. i1000 v /-IFN + 2000 TNFsignificantly boosted abortion rate, P < 0.005 by X 2.
kCBA/J mice injected twice a week for 4 weeks with 100 mg/kg silicon dioxide before mating significantly reduced abortion rate, P < 0.05 by X 2. no doubt that asialo GMI+ cells or cells of the NK lineage are involved, since their transfer, once activated, causes abortion, and, since they do accumulate at the site of embryo resorption, their modulation positively (Poly IC activation) or negatively (anti-asialo GM1 treatment) influences parallel resorption rates, lsl-1s4 We have studied this problem with David Clark, taking into account the fact that TNF and gamma interferon arc by themselves abortifacients, and performing cell deletion to learn the cellular mechanism triggered, lss In Table 5, intraperitoneal injection of TNF-c enhanced CBA xDBA/2 resorptions, but anti-asialo GM1 antibody both decreased the background rate and prevented the action of TNF-o. When we added /-interferon, it had as expected an abortifacient effect, but that was not seen in NK-celldepleted mice, suggesting that the Baines model 149'1s (-interferon macrophages activated to produce NO embryo death was not correct.
However, when /interferon and TNF-o were administered together, more than 80% of the implanted embryos aborted. This confirmed the already observed syncrgy/codependencc and sug-gested that in fact it might be crucial. Indeed, NK cell depletion suggested that TNF-o could not find enough NK-derived /-interferon, while /interferon alone fails because macrophages which depend on NK-cell-derived -interferon have and thus did not produce TNF-o, and the intraperitoneally injected cytokine does not stimulate TNF-o production quickly enough.
The action of NKs or macrophage has been attributed to a direct killing of trophoblast. In silica treated macrophage-depleted mice, the abortion rate was reduced, but this treatment was ineffective in TNF-o plus /-interferon treated animals, whose macrophage depletion was checked by the anti F4/ 80+ MoAb. Ovarian failure could have happened after silica, but this was ruled out since it should have caused abortions in 100% of cases, and the results were unaltered hormone replacement therapy treated animals.
The most logical target appeared to be the maternal uterine vascular endothelial cell, since the cytokines stimulate surface expression of procoagulant (flg/2-prothrombinase, which is distinct from tissue factor) and clotting. Table 6 shows that an- dSignificant reduction in spontaneous abortion rate P < 0.001 compared to no cytokine control group, Fisher's Exact test. eSignificant reduction in abortion rate P < 0.001 compared to cytokine-treated controls, Fisher's Exact test. No significant difference compared to anti-flg/2-treated mice which did not receive an injection of cytokines. tibody to fig/2 prothrombinase reduced the background rate of abortion to 4% and prevented the effects of TNF-ot plus y-interferon. Since granulocytes are observed in resorption sites and lyse of TNF-ot plus -interferon-activated endothelial cells, we tested the injection of pregnan't mice with a monoclonal anti-granulocyte antibody which is known to block granulocyte-mediated tumor rejection in vivo. ls6 Anti-granulocyte antibody .partially reduced the spontaneous abortion rate and significantly abrogated the effect of TNF-ot plus yinterferon (as control, see effects of rlL-10) (Table  7). Thus, vasculitis, rather than a direct cytotoxic action on fetal trophoblast, leads to abortion.
In this context, and in keeping with the aforementioned r61e of NK cells in cytokine production, concerning protection against abortion, TGF-132producing y T cells in the uterine lining appear to be important. 1sT,is8 These cells are producing TGF-[32 and IL-10, albeit, by in situ and immunohistochemistry, we find IL-10, IL-3, and IL-4 in murine spongiotrophoblast mainly, with traces in the decidua (Cayol, DEA Paris 7, Sept 1997, and in preparation; See also photos 1,2, and 3), in agreement with what is reported in humans, lzS,lz9

STRESS-INDUCED ABORTION
In a series of models, we have studied stressinduced abortion. Stress (be it contention, ultrasonic stress, etc.) can induce abortion in mice, and this can be prevented by alloimmunisation. 14s This phenomenon has been studied in detail. It involves pathways which are by some aspects very similar to those discussed above but involving substance P, CD8 T cells, and alterations of mast cells. The discussion of stress-induced pathways could be by itself a review, and readers should refer to other published papers, lsS-16s This model is of importance in discussing early human pregnancy loss. Of interest is the fact that a monoclonal antibody, BA 11, prepared by R. Jalali in James Mowbray laboratory, blocks both classical and stress-mediated abortion. 166,167 To finish, we would like to say we do not negate systemic events, we simply relativise them. As stated at the beginning of this paper, we do not negate the existence of allograft enhancement during pregnancy, nor peripheral T-cell hyporesponsiveness. 26-29'67' [168][169][170] We merely believe they reflect local events, for the most part, including the hitherto undiscussed placental suppresser factors.
These, that we have suspected since 1980,171 have proved to be elusive, 17z-174 but with a continuous effort in mice, 17s-18z we have delineated the active moiety in human and murine placental supernatants, and this induces T-cell anergy lsl,lsz in a very similar manner to staphyloccocal enterotoxin, 183 explaining, in our view, the T-cell unresponsiveness observed locally 184 whose aforementioned phenomenons (especially the report by TafurizT) are in our opinion a mere reflection. In that respect, since the anergy is transient and reversible, it is interesting to note that optimal secretion of the factor in mice is seen by cells of the invasive ECP, 17z and that there is a local deficiency in suppresser materials in the CBA x DBA/2 window of abortion. 73 CONCLUSIONS There are several causes of early pregnancy loss, which we are only beginning to uncover. Several, as in the murine system, have different etiologies, but use a final "rejection-like" common pathway, where TNF and gamma interferon are important. But several others are likely to be due to NK failure dTNF-a + y-IFN given as in Tables and 2. Significant increase in abortion rate compared to group I, P < 0.001 by X and Fisher's Exact test. eSignificant reduction in abortion rate compared to group 3, P < 0.001 by 2 and Fisher's Exact test. Pooled result from groups 2 and 4, 15/102 15% abortion rate, significantly less than 30% rate in group I, P < 0.05 by X 2. fSignificant reduction in abortion rate compared to group I, P < 0.0012 by Fisher's Exact test. No significant reduction compared to group 2 (P 0.096) or pooled groups 2 and 4 (P 0.075). Clark  or misrecognition, hence improper cytokine secretion and lack of growth factors. Several others are due to defects of the local inflammatory reaction (endometriosis, with preexisting high levels of TNF, is among these), and it is likely that, as in the LIF system, we will see discrete defects leading to definition of molecular abnormalities, with defects in adhesion molecules being involved.
Thus, early pregnancy loss and recurrent miscarriage, though due to "immunological-like" circuitry, are likely to be split up into different syndromes. It is our opinion in that respect that it is not at all surprising that only one woman out of 10 or 11 benefits from purely imunologic treatment. We have to be able to define which women will benefit by rigorous immunological criteria. (Unfortunately, the leukocyte immunisation saga has lead to many charlatan, unconscious, or preposterous theorisations, sometimes with disastrous consequences; women are not guinea pigs nor mice! We have always wondered what was the real basis for the MHC linked, disequilibrium antigen, as well as how a system like the CBA xDBA/2-which was described as minor loci dependent, MHC restricted and in which the good father (BALB/c) and the bad father (DBA/2) were both H-2d-could have been taken as an example of the proof for the need for absence of HLA homology! This is just an example, but we could--and, in fact, we are due tom write horrid things about anti-paternal antibodies, to illustrate what was said at this paper's onset.) This has to be said, because otherwise what we learn from the discoveries of complex cytokine networks at present is that by unfounded treatments we are at risk of altering other useful pathways and induce the lack of remission of transient infertility.
The "stress" saga tells us about the importance of neuroimmunoendocrine pathways, which also should not be a surprise.
To the impatience of clinicians, we recall what H. Metzger said when he was cloning the IgE receptor: "Haste is waste." We believe we have made significant progress in the past five years in understanding, with the help of animal models, in unraveling pathways which are operational indeed.
But, ae nowa have to tabe into account that added levd of complexity. We know it sounds difficult for clinicians, because we have a jargon, but we cannot escape the truth. Yes, a cytobine netaor/ is operating in early pregnancy, and, in certain cases, we might be found guilty if we were treating patients as if ignoring its existence. 7. Chaouat G, Voisin GA, Escalier D, Robert P: Facilitation reaction (enhancing antibodies and suppressor cells) and rejection reaction (sensitised cells) from the mother to the paternal antigens of the conceptus. Clin Exp Immunol 35:13-24, 1979. 8. Rodger JC: Lack of requirement for a maternal humoral immune response to establish or maintain a successful allogeneic pregnancy. Preeclampsia is associated with abnormal expression of matory cytokines in murine pregnancy: alloimmunization prevents abortion but does not affect the induction of preterm delivery. Cell Immunol 157:328, 1994. 146. Gendron