REVIEW ARTICLE: Tolerance to the Foetal Allograft?


Gérard Chaouat U 782 INSERM, Université Paris Sud, and Hôpital Antoine Béclère, 32 rue des carnets 92141 Clamart, France


Citation Chaouat G, Petitbarat M, Dubanchet S, Rahmati M, Ledée N. Tolerance to the Foetal Allograft? Am J Reprod Immunol 2010

In this review, we will detail the concept of tolerance and its history in reproductive immunology. We will then consider whether it applies to the foetal–maternal relationship and discuss the mechanisms involved in non-rejection of the foeto-placental unit.


In June 1980, I attended the Gusberg Festschrift, organised by Norbert Gleicher, which resulted in the founding of AJRI and ASRI. The opening lecture by R.E. Billingham was entitled ‘Mechanisms or factors’ and proposed to explain exemption from rejection of the allogeneic foeto-placental unit. For this AJRI celebration issue, ASRI has requested a review on tolerance, a topic of great interest to me since 19741 and the Medawar paradigm.2 Even if we avoid the cardinal role of inflammation in implantation, as exemplified by leukaemia inhibitory factor knockout (KO) mice, does the tolerance concept apply to pregnancy? Are mothers really ‘tolerant’ to paternal alloantigens of ‘the foetal allograft’? Inversely, could the term ‘tolerance’ be misused?

Part 1: Tolerance or not tolerance? that is the question

Immunological Tolerance: The Classics

It is obvious that one should first define tolerance3 and only then question if allopregnancy fulfils the criteria systemically or locally. Conversely, does allopregnancy induce tolerance to paternal alloantigens? Let us examine the definition of tolerance and its historical background, excluding the ‘TLX’ theory [trophoblast lymphocyte cross-reactive antigen-X].4 R.H. Schwartz5 defines it as ‘a physiologic state in which the immune system does not react destructively against the components of an organism that harbours it, or against antigens that are introduced to it’. Jan Klein (Natural History of the Major Histocompatibility Complex) speaks of ‘inability of the immune system to respond specifically to a stimuli, to which it does have the potential to respond’. These reflect different perceptions: the first being a total lack of response, as was found by early studies of high- or low-zone tolerance carried out by Mitchison, Chiller, Weigle, Kolsch. For review, see reference.6 These studies were carried out using soluble antigens, such as bovine serum albumin or human gamma globulin. Others see tolerance as a more complex phenomenon involving active mechanisms. Indeed, in Medawar’s classical transplantation tolerance,7 animals do not mount any response whatsoever towards the graft, even when rechallenged at a spatial/temporal distance. Current thinking indicates a total absence of antigen-triggered cytokine production linked to clonal deletion. Tolerance is not long-lived in the case of induction in adults, as opposed to being lifelong for self-tolerance or neonatally alloinduced.

With regard to mechanisms, tolerance can rely either passively on immediate clonal deletion or either after an immune response by exhaustive immunisation – mostly after exposure to infectious agents – or be actively acquired or maintained, by suppressor/regulatory T cells, this involving also ‘suppressor memory’.8 This memory explains the differences in primiparity versus multiparity for ‘tolerance’ or preeclampsia. For transplantation, Hasek observed ‘split tolerance to living cells’, characterised by a total lack of cytolytic T lymphocytes (CTL) but the presence of an alloantibody response.9 This concept applies rather well to pregnancy.10 Moreover, in enhancement/facilitation phenomenon, continuous coexistence of antibodies and CTLs can be demonstrated.11 But concepts of antibody-mediated self-tolerance collapsed when Zinkernagel and Doherty demonstrated self-tolerance MHC restriction, as alloantibodies are unrestricted. For these ‘active’ processes, Schwartz’s definition is the closest and applies to pregnancy, still too often viewed as total anti-paternal unresponsiveness, despite evidence of immunotrophism. But do we need to speak of ‘tolerance’ for pregnancy? Some do not think so.

Immunological Tolerance: The Heretics

The self- /non-self-theory has pitfalls, and pregnancy is the main one for opponents. Antonio Countinho sees the immune system as: (a) networks, including anti-self natural autoantibodies12 and idiotype/anti-idiotype antibodies/T-cells. This has been relatively poorly studied in allopregnancy, despite reports13,14 that might be relevant to effects of intravenous immunoglobulins (IVIG) for recurrent spontaneous abortions (RSA). Matzinger’s ‘danger theory’15 stems from discussions on pregnancy with Robert Schwab (June 16, 1998 New York Times). For her, it implies that the immune system does not function by self /non-self, but instead reacts to ‘danger’ signals such as inflammation, apoptosis, and bleeding. Thus, healthy foetuses are not rejected, simply because they do not send alarm signals. However, should they become infected, the mother, in clearing infection, also rejects the foetus. ‘The danger model’ predicted an important role for antigen presenting cells (APCs) in turning tolerance on or off, and specific ‘danger receptors’, subsequently identified as Toll-like receptors. It offers an apparently elegant, though tautological, explanation of allopregnancy as ‘it does not elicit danger’. Polly Matzinger states further: ‘reproduction cannot be a danger’.... ‘it does not make evolutionary sense’. This also explains why a conceptus still thrives in a pre-immunised host, as grafting produces micro wounds and local bleeding, while the foetus does not seem to do so.

My Personal Approach of the Danger Model and Pregnancy

Danger was enunciated before the 1989–1991 papers describing implantation as requiring local inflammation and ignores that invasion is accompanied by apoptosis,16 local bleeding and in equids there are zones of quasi rejections in the placenta with a massive maternal lymphocytic infiltrate.17 It is also difficult to explain by the danger model why, in murine abortion,18 some foetuses are rejected, whereas in the same mother, others are not, both being not infected. However, CBA × DBA/2 embryos can be rescued by pre-culture in CSF-conditioned medium before transfer to a CBA foster mother, suggesting that these embryos are not fully ‘normal’.19 Danger might explain why the CBA × DBA/2 system is environmentally dependent,20 although surprisingly, the LPS content of faeces does not correlate with abortion.21 Danger, however, does not explain why CBA × DBA/2 and DBA/2 × CBA matings are seen differentially (gene imprinting experiments of A. Paldi) and why immunisation against paternal MHC antigens corrects ‘danger’,22 even how immunisation permits pregnancy in case of donkey embryos implanted in mare (the donkey in horse pregnancy).15 Finally, Matzinger did not envisage alloantigen-specific mechanisms regulating only the anti-foetal reactions. Moreover, what she describes is exactly opposite to some cases of infections, such as local, e.g. uterine Listeria. In this case, local immune responses are seriously impaired during pregnancy, shielding the infection, until it spreads to the point that the systemic immune system cannot cope with it, even though the potential for a systemic immune response is there.23 Similarly, pregnancy impairs resistance to Salmonella leading to rapid, fatal infection.24

Confusion on the Immunological Nature of Pregnancy

As stated by Loke and Moffet in their review in Nature Immunology,25‘ruminations about the immune system during pregnancy are mostly centred on the acquisition of maternal tolerance to the allogeneic foetus’. This view is probably too simplistic. Failure to distinguish between the local and systemic immune response has led to a great deal of confusion. Another problem is dealing with pregnancy as if it had first to cope with the adaptive immune system. But placentation appeared in the Devonian,26 in an era where T cells did not exist, and evolution had barely produced IgM. But NK cells did already exist. Such an immune system still exists nowadays in some sharks. The adaptive immune system later adapted to pregnancy, which then used it for immunotrophism27 and local vessel remodelling, selecting a specialised unique population, uterine NK cells.

Some too Often Forgotten but Classic Experiments

After Medawar’s reflections, his collaborator Billingham started experiments28 with Alan Beer and Judith Head. What they found was that the mother is not systemically, or even locally, tolerant to paternal alloantigens. Let us recall that an animal A, made tolerant to B, accepts any B tissue. Tolerance is incomplete or absent if some tissues are accepted but others more immunogenic are rejected. Thus, the classical challenge is skin allografts, not weakly immunogenic tumours as painstakingly defined by Brent, Billingham, and Medawar28 at ‘the birth of transplantation biology’.29,30 The results of Beer and Billingham’s skin graft studies in a first pregnancy are so easily reproducible, even with strongly immunogenic tumours,31–33 that it is surprising they are so often ignored: (1) in first pregnancy, synpregnant or allopregnant mice do reject HY (male) syngeneic skin grafts exactly as virgins. MHC identical, minor loci different grafts are also rejected as in virgin hosts; (2) MHC alloskingrafts have a similar fate, albeit with rejection kinetics of, at best, 2 days, which is attributed to gestational corticoids; (3) not least, intrauterine grafts at a minimal distance from the implantation site will enjoy only slightly prolonged survival, unless placed in the decidua basalis itself, or in pseudo pregnant decidua, where these sites behave then as immunologically privileged; (4) finally, Woodruff34 showed that foetal tissues grafted in the leg will be rejected during pregnancy.

Immunisation and Tolerance Challenge Studies Question the ‘Tolerance’ Concept

In a tolerant animal, a new, e.g. post-induction of tolerance, challenge by immunisation will not induce rejection of any matched tissue. Indeed, Mitchison, and later on Lanman, showed a ‘lack of harm to foetus from sensitisation of the mother’.35 And if ‘pregnant mice are not primed’, but can be primed by immunisation during pregnancy,36 this does not induce abortion, while one sees in such cases, rejection of paternal grafts if not adjacent to the implantation site. We have seen such a phenomenon in the CBA/J strain which is an ‘alloantibody producer’ and is one of the strains where suppressor T cells (Ts) were first demonstrated in pregnancy. Anti-paternal MHC immunisation prior to pregnancy results in the induction of circulating active anti-paternal CTLs with rejection of a paternal tumour strain allograft.37 And, as for Beer and Billingham’s study,38 the placentae in such immunised mice were bigger than the controls. So there is no classical systemic tolerance in the first pregnancy. It must be mentioned here that the H-2 Kb-transfected P815 mastocytoma used by Tafuri39 is by far not as immunogenic as skin or a methylcholanthrene sarcoma, and ‘after delivery (21–28 days), the ability to reject P815-Kb grafts was restored’, which is in marked contrast with a real tolerance which lasts far longer and survives the removal of the challenging tissue. Similarly, the more immunogenic JR-5 fibrosarcoma cells, or Lewis lung tumour (LLT), of Robertson’s group40–42 are also rejected post-delivery. The sole case when such allotumour is not rejected is enhancement1 but only in the so-called alloantibody ‘producer’ strains.1,43

Foetal Cells in Maternal Blood

As pointed out by Loke, ‘micro-chimerism’ is seen in mice and humans.44,45 Some foetal cells, mostly trophoblasts, engraft eventually, especially in the bone marrow. Such cells can persist until 27 years post-delivery.46 So there is a real ‘tolerance’-like phenomenon to some foetal cells, the mechanisms by which they escape destruction, seeming to be the same as for local trophoblasts. But as exemplified by their detection after abortion, one can observe ‘rejection of foetal allograft’ and ‘tolerance’ to foetal cells.

Pregnancy in Tolerant Animals

Finally, pregnancy should not be affected by tolerance to paternal alloantigens, but tolerance negatively affects pregnancy. Female rats made specifically tolerant before pregnancy to paternal alloantigens produce smaller F1 foeto-placental units,38 as do anti-CD4-treated or nude mice.47 In the Beer and Billingham experiments, even in tolerant animals with reduced placental weights, allogpregnancies still yielded the biggest placenta and foetuses.38 This remained incomprehensible until it was made clear that NK cells participate in the ‘immunotrophic’ phenomenon.48 The final conclusions by Beer and Billingham were clear cut. Pregnant animals were not systemically tolerant, and ‘some active immune mechanism linked to allorecognition of the foetus by the mother was required for a fully successful pregnancy’; a conclusion reiterated strongly in the title of several of their papers49 and at the origin of Alan Beer’s ‘treatments’ of RSA by alloimmunisation which we do not discuss here. So it was known until the 1970s that the foeto-placental unit behaves exactly the opposite of a tolerated allograft: tolerance makes it smaller, and immunisation makes it thrive.

Local Immune Reaction/Suppression of Anti-Foetal Cellular Responses

Could this active immune reaction, nevertheless, be creating a local tolerance? This reaction was examined in detail again by Beer, Scott and Billingham and was found not only at the foetal–placental interface itself but also in the para-aortic draining lymph nodes.38,49,50 Their removal partially alleviated, what was not yet named, ‘immunotrophism’.38 In 8 non- immunised animals, foeto-placental weights were significantly lower in those animals whose lymph nodes were excised. The magnitude of this effect is strain dependent. This positive reaction was shown, later on, to be maximal in abortion-prone models, as immunisation prevents foetal loss,51 the root of the immunotrophism theory.27,51

Tolerance induced by multiparity

Multiparity is markedly different from a classical graft. In this case (allograft on a virgin recipient), a second similarly incompatible graft suffers second set rejection. But in every mammalian species, placental and foetal weight, and often litter size, are increased by multiparity. The only known exception is in the CBA × DBA/2 matings, where a second DBA/2 pregnancy increases foetal losses in some CBA/J mice, termed then ‘bad mothers’. Nevertheless, even in this strain, many adverse effects are seen only in the first pregnancy, offering a murine model of preeclampsia.52 Moreover, multiparity induces real, long-lasting systemic tolerance to male skin grafts53 and tolerance or hypo-responsiveness towards paternal MHC allografts.53,54 In both cases, the effects are transferable by injection of thymus-derived suppressor cells, e.g Ts. So in conclusion to this first part, instead of classical ‘tolerance’, it seems preferable to speak as Billingham does of non-rejection of the foetus or eventually to speak of a ‘transient, local tolerance-like phenomenon’, accompanied in certain strains/ species by a ‘transient systemic anti-paternal hypo-responsiveness’, which can eventually lead to a ‘complete state of systemic tolerance induced by multiparity’ to paraphrase Kaliss.55

Part 2: Immunology of the feto–maternal relationship

Humoral Compartment

In many species or strains of mice, B cells produce anti-paternal alloantibodies, even in the first pregnancy. These strains are called the alloantibody ‘producer’ strains, but the overwhelming majority are ‘non-producers’.43 In ‘producers’, the ‘natural’ antibody is non-complement-fixing IgG1.1,43 Isotype switching to IgG1 is seen in pregnancy of pre-immunised, non-producers, but a significant proportion of the antibody are still IgG2.43 IgG1 predominance leads to the concept that tolerance in pregnancy was a proof of the facilitation concept.1,11 But what then of the non-producers? Moreover, there are species, such as primates, in which an anti-paternal cytotoxic alloantibody response is observed as early as first pregnancy, and this is the case for human alloantibodies.56 For most authors, such antibodies are mainly associated with graft rejection, so there must be local protection. Let us mention also here the ‘asymmetric’ antibodies.57 It ensues that (1) an immunoregulatory mechanism operating at a systemic level must explain the isotype switch, such as secretion of IL-4 by the placenta; and (2) locally, the foetus must have a strong protection against complement-mediated cytotoxicity, as it is possible to obtain hyperacute paternal graft rejection by injection of anti-paternal MHC antiserum in mice without compromising foetuses. The protection is threefold: (1) There is quick turnover and engulfment of anti-MHC placenta bound antibodies. The placenta acts as an active ‘sponge’,58 which might explain the different fate of MHC and OVA transgenic foetuses after immunisation; (2) The placenta expresses complement regulatory proteins; and (3) The placenta secretes Th2 cytokines.59

Cellular Immune Responses

As in a classical T-cell response, the size of draining lymph nodes (DLN) increases during pregnancy,38,50 as first evidence that the maternal T cells are ‘aware’ of the conceptus as said later by Tafuri.39 In a second, similar MHC mismatch pregnancy, a recall flare phenomenon is observed in DLNs, showing that the mother ‘remembers’ the first allopregnancy. In vitro, anti-paternal lymphocytes, or anti-trophoblast mixed lymphocyte reactions (MLRs), in a normal first pregnancy never generate CTL (and neither pregnancy nor abortions ever induce CTLs in vivo), but authors vary on MLR kinetics; a primary one for some authors and a secondary response for most. Transgenic models are available for T and B cells. Colette Kanellopoulos has shown that placental giant cells migrate into bone marrow and delete some immature B cells.60 For T cells, in vivo studies by Tafuri et al.39 yielded clear evidence for T cells being transiently specifically unresponsive/anergic. But we repeat that responsiveness and T-cell phenotype are restored after delivery,39 while with HY transgenic, Jiang and Vacchio demonstrated that T cells specific for foetal antigens decrease in an antigen-specific manner during pregnancy and remain low post-partum, consistent with clonal deletion61 and contrasting with the ‘accumulation’ reported by Mellor.62 The remaining clonotypic T cells are unresponsive to antigenic stimulation (anergy), but at the T-cell receptor level, the number of co-receptors is not down-regulated.61 Thus, anergy and clonal deletion coexist.

Mechanisms Involved in the Acceptance of Foeto-Placental Unit

Anergy or equivalent

The zeta chain of CD3 co-receptor is abnormally phosphorylated.63,64 This can be obtained in MLR by incubating responder cells with supernatant of placental explant cultures or purified heat-resistant material.64 Cells allostimulated in the presence of this material will not respond in a second MLR with the same stimulator MHC, whereas they will do so against a third party. The T-cell anergy observed in such a case is transient, requiring continuous presence of the active moiety which has been identified as being a prostaglandin (PGE2).65 This explains the above-reported anergy 39,63,64 seen by Tafuri and others. A similar activity has been traced in the blood in the form of placental exosomes.66 A suppressive role can be exerted by embryo-associated suppressor factors, such as pre-implantation factor (PIF), as described by Barnea,67 as well as by the well-described regeneration and tolerance factor (RTF) cytokine of Beaman and Hoversland,68 whose neutralisation by a monoclonal antibody clearly induces abortion.68 indoleamine 2, 3-dioxygenase (IDO), which is expressed by trophoblasts, also induces profound T-cell anergy. Indeed, neutralisation of IDO induces abortion solely in allopregnancies with rates varying with the mating combination.69 IDO KO mice breed, which is often presented as a negative argument, but these are synpregnancies not allopregnancies. The physiological situation for this requires IDO KO in two different strains.

Mechanisms of clonal deletion

Two mechanisms can explain clonal deletion. First, Fas/Fas ligand interaction: outer trophoblasts express Fas ligand with a weaker expression at term. Activated T cells express Fas, and the interaction of Fas with FasL induces death by apoptosis. Thus, any anti-paternal alloantigen T cells are immediately destroyed when binding trophoblasts.70 Such T cell encounters in the periphery (bone marrow) with deported trophoblasts would explain micro-chimerism. However, allopregnancies are normal in double Fas/FasL matings.71 Another mechanism with similar consequences is the secretion of sHLA-G, which kills activated T cells.72 Clonal deletion becomes, as a consequence, deeper, as pregnancy progresses, and reverts in absence of a placenta.

A Privileged Environment: Cytokine Balance at the Foetal–Placental Interface

The Th1/Th2 paradigm73 supposes a shift to Th2 predominance during pregnancy, which at the foetal–placental interface would create a transient hypo-responsive (privileged) site. Indeed, the main Th2 cytokine, IL-10, is present at both sides of the foetal–placental interface,59,74 and IL-10 prevents resorptions in CBA × DBA/2 matings.75 However, IL-10 KO mice or deletion of 4 Th2 by KO simultaneously in one mouse76 does not affect foetal health. But Sharma and Robertson have shown data that while IL-10 KO mice develop normally, they are more susceptible to LPS-induced abortion,77,78 somehow linking IL-10 with ‘danger’. Finally, three more mechanisms should be mentioned, mostly on the ‘uterine side’: TGF-beta produced locally by null cells;79 progesterone-induced blocking factor (PIBF);80 and suppressor/regulatory T cells (Ts/Tregs). TGFs, which are also strong immunosuppressants, are the sole growth factors being also immunosuppressive. A deficiency of a DLN suppressor factor was first noted in the CBA × DBA/2 mating. The factor proved to be a TGFβ2 analogue.79 TGF-beta has important immunodeviating capacities during implantation. Trophoblast MHC class I recognition elicits progesterone receptor (PgR) expression on hitherto PgR-lymphocytes, which in the presence of high doses of progesterone, seen only at the placental–foetal interface, induces PIBF secretion itself.80 All of these mechanisms are redundant, and the soluble factors act at high doses, thus only locally, creating a quasi-immunologically privileged site without affecting systemic immunity.

Real tolerance and multiparity

Historically, pregnancy-enhancing antibodies were first reported in multiparous mice and even in ‘non-producer’ strains. Multiparity induces transferable-specific hypo-responsiveness or even true tolerance to either HY or paternal alloantigens.53,54 Placental products, be them placental extracts or water-soluble material obtained from these, co-injected with alloantigenic cells, induce systemic antigen-specific LyT2+ Ts.81 These were traced in the first pregnancy in mice and in rats by Baines and Liburd. Similarly, antigen-specific MHC-restricted Ts were found in humans.82 Controversies about in vitro assays can still be traced in proceedings of the Gusberg meeting.83 In the 1980s, we studied, in detail, the in vitro properties and mode of action of these suppressor cells (specificity, mediation by a soluble factor). A part of these studies was carried out with anti I–J antisera, as many other labs working on suppression did at the time.

Death and Resurrection of Regulatory T Cells

Lee Hood’s demonstration that the I–J region does not exist while properties of the suppressor factor(s) of Gershon and Cantor were more and more improbable doomed Ts. For an excellent revision of the history of Ts, see references.84,85 We nevertheless still tested/published the role of Ts in CBA × DBA/2 matings.51 As reviewed, in,86 the CD25 and Foxp3 markers again boosted Ts on the forefront. Yet the I–J trauma lead to a more benign denomination of ‘regulatory T cells’ (Tregs), rather than ‘CD4+ Ts’, which we first saw in 1981, but termed ‘inducers’ .87 CD8+ cells are still important partners, as shown in studies by Arck, Clark and coworkers.88

First Demonstrations of T Regs in Murine Pregnancy ‘Tolerance’

Aluvihare and Darasse convincingly demonstrated that CD4+ CD25+ elimination causes foetal deaths in allopregnancy by transfer or direct in vivo experiments.89,90 Saito traced/ quantified Foxp3 cells (T regs) in human decidua as well as regulatory NK/T cells.91 Robertson and coworkers92 showed that the Foxp3 marker decreases in unexplained infertility endometrial biopsies. These, and Fainbolm, detected periodic T reg modulation during the menstrual cycle, peaking in the late follicular phase.93,94 For Fainbolm, T regs from patients with RSA are ‘functionally deficient’,93 and T reg decidual recruitment correlates with expression levels of CCL3, CCL4, CCL5, CCL22, and CX3CL1.90 Finally, placenta-dependent CD8+ T regs have been demonstrated by Shao et al.,95 and this is reminiscent of earlier data in mice about LyT2 Ts.81

HLA-G. Intrinsic resistance: is there a need for tolerance?

Could the placenta escape immune attack by resisting effector cell lysis? We have discussed the Fas/Fas ligand interaction. Membrane and soluble HLA-G (sHLA-G) also play a role, including sHLA-G secretion by the MHC-syncytiotrophoblast. Moreover, trophoblasts (and choriocarcinomas) are resistant intrinsically to cell-mediated lysis.96–98 This resistance is independent of HLA-G.99,100 The once debated soluble factors96,101,102 had properties which fits with what is now known of soluble HLA-G, be it sHLAG1/ G2 characteristics. The activity is very similar to what we have described crudely in the 1980s as CTL and NK inhibitory factors (CTL and NK IF).101,102

Abortogenic conditions by activation? Break of ‘tolerance’?

However, lymphokine-activated killer cells (LAKCs) lyse trophoblast, and activated NK cells cause abortion.18,103 This suggests that ‘tolerance’ can be broken by systemic activation. In this context, immunisation with OVA induces abortion in OVA transgenic mice,41 an occurrence not seen with classical transgenics. Thus, a ‘modified’ placenta can be seen/rejected as a transplant, but OVA at the trophoblast surface does not necessarily have the high turnover of MHC class I.58 If this explanation is correct, OVA immunisation should 13 not affect an OVA–MHC recombinant protein transgenic foetus. This bears interest also, as in a Greek study, many patients with RSA were virus positive.104 The forced induction of class II alloantigen on the placenta to induce abortion, as reported by Athanassakis et al., is very controversial, as Mattson’s group did not reproduce it. IDO blockade of abortions is mediated by CD4+, not by CD8+ T cells,69 pointing towards a crucial role of local macrophages and complement.

A Role for Non-T and Non-B Cells

Antigen-presenting cells, notably dendritic and CD11+ cells, are involved in the creation of a privileged local microenvironment,105 while also being crucial for decidualisation/implantation.106 In CBA × DBA/2 matings, syngeneic dendritic cell therapy increases local CD8+, γδ T cells, TGF-beta1, and PIBF, correlating with decreased abortion rates.105,107 A pivotal role was shown for galectin-1 (Gal-1), an immunoregulatory glycan-binding protein, synergising with progesterone. For the influence of stress in pregnancy, we direct readers to recent reviews.108,109

The innate immune system


Maternal non-rejection of the foetus also necessitates local regulation /cohabitation with the local innate immune system. Cytotoxic alloantibodies in many species call for complement regulation, and indeed activation of complement is abortifacient.110,111 Also, differential levels of MBL (mannan-binding lectin) are observed in CBA × DBA/2 versus CBA × BALB/c mice and in human patients.112 But complement is regulated at the fetal–placental interface by placental regulatory proteins. Mice made KO for crry destroy their embryos even in syngeneic pregnancy.113 MCP and DAF play this role in humans. Hence, prevastatin is to be tested for abortion and preeclampsia therapy.

NK Cells

We will not detail uNK cells and angiogenesis, but according to the missing self-theory, MHC-negative trophoblast, while protected against T-cell effectors by lack of target molecules, should be destroyed by NK cells. The low lytic activity of uNK cells, per se, might seem to be a protection. In fact, while syncytia cannot be destroyed as easily by a single ‘hole’ and offers considerable capacity of self-repair, one should recall that activated NK cells are abortifacient as also seen in ‘natural’ CBA × DBA/2 matings.18 This activation is controlled by the NK-repressing activity of the already detailed HLA-G, placental factors, PIBF, and IL-10. But besides NK cell ‘attack’ of the placenta, there are reports suggesting for RSA, as for preeclampsia, an improper combination of maternal killer Ig like receptor (KIR) and foetal HLA-C.114 When mice are injected with poly(I:C), abortion occurs because uterine NK cells are activated. Similarly, the human uterine NK cells can be activated towards cytotoxicity. The final activity of NK cells is governed by a balance of inhibition and activation by the trophoblast ligands/NK cell receptor interactions. El Costa et al. have shown that engagement of NKp46 receptor, but not NKp30 receptor on decidual NK cells, triggers cytotoxicity. Such cytotoxic potential is negatively controlled by NKG2A inhibitory receptor co-engagement.115 This and other studies on NK cell KIR repertoire in spontaneous abortions suggest that uNK cells, and in some circumstances systemically activated blood NK cells, can ‘reject the foetal allograft’ as seen in break of transplantation tolerance.

New partners?

More partners, such as NKT cells and inhibitory NKT (iNKT) cells, are emerging in tolerance. As a recent example, alpha beta(+) CD161(+) NKT cells have been shown to reside in the decidua and may play an important role in foetal tolerance, and this is reinforced by demonstration of expression of CD1d on trophoblasts.116,117 Linking ‘tolerance’ and immunotrophism, decidual iNKT cells are strongly polarised towards GMCSF expression, and CD1d expression is linked to trophoblast differentiation.117 Another subset certainly playing a role is Th17 cells, which can be involved in rejection. Galectin regulates this subset. Interestingly, FoxP3/IL-17 dysregulation is seen in preeclampsia, and we have obtained data linking IL-17 with implantation failure. Other cytokines important in this respect are Ebi3 (IL-27) and its derivative IL-35, an immunosuppressor expressed at interface in mice118 and by activated T regs. Another emerging modulator is IL-22, regulator of Th17, IL-17, IL-23 also regulating in many systems G-CSF, a matter of importance in view of CSF role in embryo implantation potential and foetal tolerance.119

Back to ‘Danger’ Predictions: Toll-Like Receptors

As stated earlier, the danger theory predicted Toll-like receptors and the initial steps of pregnancy as an inflammatory, Th-1-dominated stage. This suggests that Toll-like receptors play a cardinal role in early adhesion/invasion and participate in the promotion of foeto-maternal tolerance. We will not substitute here the excellent reviews of Mor and Abraham,120 but recall in the context that the system includes regulation of Toll-like receptors by ligands as regulators of T reg function. Data suggest that a ‘break of tolerance’ can be linked to response to local danger, as strongly suggested by CBA × DBA/2 matings, with a role for MD1. Similarly, TLR9-triggered activation in IL-10 KO mice amplifies uterine neutrophil and macrophages and their migration to the placental zone, with high pregnancy losses.78

‘Tolerance’ Induction Before Implantation

Finally, ‘priming’ for ‘tolerance’ might start before implantation. Using a very elegant model of paternal strain tumour growth, Robertson’s group has shown that mating induces an inflammatory response,40–42 as seen already in the 1990s,121 which causes systemic antigen-specific hypo-responsiveness involving Tregs, TGF-beta and CSFs.122 But paternal strain tumours are rejected post-pregnancy. Thus, ‘tolerance’ is rather hypo-responsiveness. Seminal fluid is required as are the cells in the ejaculate. Therefore, ‘tolerance’ is prepared before implantation,122 also possibly via embryo signals such as PIF67 and follicular fluid G-CSF .


In conclusion, transient hypo-responsiveness, but not classical tolerance, exists in the uterus and to a lesser extent, systemically. This is not because of a single mechanism – each one acting as back up, should others fail. Considerable progress has been made since I began my research in 1974. For this anniversary issue, I recall that at the New York Mount Sinai hospital 1980 meeting, these questions were raised. Nowadays, although experiments were then ‘basically correct’,83 one is impressed by the complexity unravelled which testifies for the strength and development of our field.

Note: An extended version of this review (350 references, 15100 words, Word format) will be sent by email upon request to: