During pregnancy, the immune system has two vital functions: to protect the mother against infections and to tolerate the semiallogeneic fetus. The fact that the maternal immune system does not mount an aggressive immunological reaction against the placenta and the fetus, which bear paternal, and thus foreign, antigens has attracted considerable interest. While it was earlier believed that the maternal immune system could not detect the growing fetus, it is now known that the maternal immune system not only becomes aware of the fetus but also actively tolerates it. Several mechanisms have been proposed to explain tolerance, which might co-exist to ensure compensatory mechanisms when one of them is absent. Increased understanding of these mechanisms should provide insight into pregnancy complications, such as spontaneous abortion and preeclampsia, which may arise partly from incomplete tolerance.

Explaining the paradox

The paradoxical nature of the immunological relationship between a pregnant mother and her antigenically foreign child was first recognised by Peter Medawar in 1953 (Ref. Reference Medawar1). He proposed three reasons as to why the fetus does not usually provoke an immunological reaction from its mother: (1) anatomical separation of the embryo from its mother, by a barrier considered at that time to be impermeable to maternal cells; (2) antigenic immaturity of the fetus; and (3) immunological inertness of the mother. However, as a result of increasingly precise research methods, none of these explanations is valid anymore.

First, it is now known that in species with hemochorial placentation (as in humans and rodents), the mother and the fetus are not separated by an intact layer of epithelial cells; trophoblasts are thus exposed to maternal immune cells, which have the potential to develop a strong immune response to reject the foreign tissues. Bidirectional transfer of cells between mother and fetus is common during human and murine pregnancy, and can lead to microchimaerism (persistence of fetal cells in the mother for decades) (Refs Reference Billington2, Reference Andrassy3, Reference Yan4, Reference Khoshrotehran5, Reference Nelson6 and Zenclussen, A.C., unpublished). Chaouat and colleagues (Ref. Reference Chaouat, Kolb and Wegmann7) as well as Wegmann (Ref. Reference Wegmann8) revised the concept of the anatomical separation and proposed the ‘trophoblast barrier–immunotrophism mechanism’, where the trophoblast cells play a protective role for the rejectable fetus within the maternal immune system, not by creating a barrier but by actively contributing to tolerance through the secretion of soluble factors.

Second, the fetus may not be antigenically immature: dendritic cells (DCs) derived from fetal skin, which are in direct contact with maternal blood and therefore capable of entering the maternal circulation, express major histocompatibility complex (MHC) class II antigens and are potent accessory cells in polyclonal T-cell responses (Ref. Reference Elbe-Burger9). Third, with regard to immunological inertness of the mother, there is clear evidence that maternal T-cell recognition of fetal antigens can lead to antigen-specific tolerance. In mice, T cells expressing transgenic T-cell receptors (TCRs) specific for paternal antigens recognise paternal alloantigens expressed by the fetus, and this results in transient tolerance of the T cells (Ref. Reference Tafuri10). Thus, not only does the maternal immune system respond to the fetus, it is also actively tolerised by it (Ref. Reference Jiang and Vacchio11). Besides, systemic immune activation occurs in pregnancy, and women do not usually suffer from immunosuppression during pregnancy. In this regard, the fetal–maternal interface can be considered as a site of immune privilege, where fetal tissue directs the maternal immune response towards a protective, tolerant one.

Since Medawar's hypothesis, a variety of mechanisms have been proposed to explain the survival of the semiallogeneic fetus in pregnancy (Table 1 and references therein), particularly with regard to the active generation of tolerance. These mechanisms may act through distinct or overlapping pathways, and some of them are associated with the generation or action of regulatory T cells (Treg). Redundancy in the systems allowing pregnancy success is suggested by the fact that normal pregnancies are observed in mice lacking the T-cell suppressor IDO (Ref. Reference Baban35) or T helper 2 (Th2) cells (Refs Reference Svensson36, Reference Fallon37). The Th1–Th2 cytokine balance and Treg have been seen as particularly important in pregnancy success and are discussed further below. This review also summarises recent work on the enzyme HO-1, which may act partly by modulating immune responses to the fetus.

Table 1

Some of the current theories explaining the survival of the semiallogeneic fetus within the maternal uterus

Abbreviations: DC, dendritic cell; HO-1, heme oxygenase 1; IDO, indoleamine 2,3-dioxygenase; LIF, leukaemia inhibitory factor; MHC, major histocompatibility complex; NK, natural killer; PDL-1, programmed death ligand 1; Treg, regulatory T cells.

Th1–Th2 cytokine balance in pregnancy

Over the past decade, the Th1–Th2 cytokine balance has been seen as a very important mechanism determining the survival of the fetus in the maternal uterus (Refs Reference Lin27, Reference Raghupathy38, Reference Saito39, Reference Piccinni40). Production of Th2-type cytokines such as the interleukins IL-4 and IL-10 locally at the fetal–maternal interface was thought to favour the maintenance of mammalian pregnancy (Refs Reference Lin27, Reference Piccinni40), whereas a Th1-dominant pattern [up-regulation of IL-2, IL-6, interferon γ (IFN-γ) and tumour necrosis factor α (TNF-α)], as seen during miscarriage, was thought to mediate fetal rejection (Refs Reference Lin27, Reference Raghupathy38, Reference Zenclussen41) by provoking an inflammatory milieu incompatible with pregnancy success. Other immunosuppressive molecules, especially transforming growth factor β (TGF-β), a Th3 cytokine, have also been proposed to play an important role during pregnancy (Refs Reference Gorivodsky42, Reference Dünker and Krieglstein43).

This importance of the Th1–Th2 balance for pregnancy outcome was questioned recently because normal pregnancies were observed in mice that lacked Th2 cytokines, as in IL-10/IL-4 knock-out mice (Ref. Reference Svensson36) and even in mice with quadruple gene deletions leading to lack of IL-4, IL-5, IL-9 and IL-13 (Ref. Reference Fallon37). This suggests that Th2 cells are not essential for normal pregnancy, and other mechanisms may prevent alloreactive Th1 attack on the fetus. Furthermore, up-regulated levels of Th1 cytokines observed during miscarriage might not be the main cause of abortion but rather a result of immunological rejection and inflammation. Nevertheless, IL-4 and IL-10 are related to successful pregnancy outcome in humans (reviewed in Ref. Reference Dealtry, O'Farrell and Fernandez44), and the administration of IL-10 significantly decreased the abortion rate in the CBA/J x DBA/2J murine model (which has a high rate of abortion) (Ref. Reference Chaouat45). In addition, although IL-10 is not essential for maternal immune tolerance, maternal IL-10 is a determinant of the growth trajectory of progeny in the uterus and after birth (Ref. Reference White46). Recent evidence outlined below has shed more light on a role for Th1 cells in pregnancy failure.

Regulatory T cells in pregnancy

CD4 ⁺CD25 ⁺  Treg are a unique subpopulation of T cells (reviewed in Ref. Reference Sakaguchi65) known to play a major role in preventing autoimmunity and tolerating allogeneic organ grafts (Refs Reference Waldmann66, Reference Sakaguchi67), acting either by cell–cell contact or by secreting immunomodulatory factors such as IL-10 and TGF-β. The acceptance of paternally derived tumour cells during pregnancy (Ref. Reference Tafuri10) supports the involvement of systemic regulatory processes in pregnancy. The existence of a regulatory T-cell population was first proposed in the early 1980s by Chaouat and Voisin (Ref. Reference Chaouat and Voisin68), although the identification of this population was hampered by the techniques available at that time.

Recent studies have shed more light on the role of Treg in pregnancy. During normal pregnancy in mice, there is an expansion of Treg (Ref. Reference Aluvihare, Kallikourdis and Betz29) and, interestingly, mice undergoing abortion have a diminished number of Treg as compared with normal pregnant mice (Ref. Reference Zenclussen30). Additionally, human studies support the important role of Treg in pregnancy. Normal human pregnancy is associated with an increase in the circulating Treg (Refs Reference Somerset69, Reference Sasaki70). Furthermore, the proportion of decidual CD4 ⁺CD25 ⁺  T cells was significantly lower in specimens from spontaneous abortion compared with specimens from induced abortions (Ref. Reference Sasaki70).

The action of Treg: generation of a tolerant microenvironment?

There is no clear evidence that Treg actions in pregnancy include inhibition of Th1 responses (Ref. Reference Zenclussen72). An interesting theory presented by Waldmann and collaborators postulates that Treg might affect several cellular pathways to establish a state of privilege in the tissue microenvironment (Ref. Reference Waldmann66). Some of the putative protective pathways induced after Treg action are IL-10 and TGF-β production, enhanced IDO expression, or HO-1 induction (Ref. Reference Waldmann66).

Interestingly, increased TGF-β protein is found in mice rescued from abortion by Treg transfer (Ref. Reference Zenclussen72); thus augmented local TGF-β levels might be related to Treg-induced tolerance during pregnancy, at least in this particular model of spontaneous abortion. HO-1 mRNA levels were also increased in placenta and decidua in animals that received Treg, suggesting a critical role of HO-1 in Treg-induced tolerance. Interestingly, HO-1 up-regulation was related to an augmentation in the Treg marker neuropilin-1 (Ref. Reference Sollwedel73), which might indicate cross-talk between both systems during pregnancy (see later). Dramatically augmented local levels of LIF after Treg application suggest an important role of LIF in Treg-induced tolerance at the fetal–maternal interface (Ref. Reference Zenclussen72). LIF is essential for implantation success: animals lacking the LIF gene produce normal blastocysts that fail to implant into a LIF-deficient uterus but are able to implant in a wild-type uterus (Ref. Reference Stewart31). Interestingly, LIF was also found to be up-regulated in CBA/J mice that naturally accepted skin allografts from BALB/c mice (Ref. Reference Metcalfe, De and Muthukumarana74), indicating that it may also be involved in tolerance of allografts. Thus, Treg at the fetal–maternal interface may be effective in avoiding allorejection by interacting with tissues and creating a ‘tolerant’ privileged microenvironment characterised by high expression of TGF-β, HO-1 and LIF (Fig. 1).

Figure 1

Hypothetical tolerance mechanisms operating at the fetal–maternal interface. Briefly, we propose that during pregnancy, regulatory T cells (Treg; characterised by Foxp3 and neuropilin-1 expression) are generated after contact with male antigens. This population expands in the periphery as a result of continuous alloantigen release from the placenta. At the fetal–maternal interface, these cells are thought inhibit maternal effector cells and further up-regulate tolerant molecules, such as heme oxygenase 1 (HO-1), leukaemia inhibitory factor (LIF), transforming growth factor β (TGF-β) and interleukin 10 (IL-10). HO-1 and LIF are expressed mainly by trophoblasts, whereas TGF-β can also be released by immune cells, including Treg. Together, these mechanisms may contribute to the tolerance towards the fetal antigens. APC, antigen-presenting cell.

Heme oxygenase 1 during pregnancy

The enzyme HO degrades heme, a protein involved in several vital cellular processes including cell differentiation and proliferation (Refs Reference Abraham, Friedland, Levere and Brown75, Reference Abraham76, Reference Beri and Chandra77, Reference Sassa and Nagai78, Reference Ponka79). It is synthesised in all human nucleated cells, although its importance varies between tissues. Senescent red blood cells, a major source of excess heme, have to be removed from the circulation and are then degraded by the reticulo-endothelial system in the spleen, liver and bone marrow (Ref. Reference Ponka79). Excess free heme can cause cell damage and tissue injury as heme catalyses the formation of reactive oxygen species, resulting in oxidative stress (Refs Reference Vercellotti80, Reference Jeney81); in addition, it can easily intercalate in cellular and intracellular membranes, further destabilising the cytoskeleton (Refs Reference Balla82, Reference Ryter and Tyrrell83). It is thus of vital importance that the amount of free heme is tightly controlled to maintain cellular homeostasis and avoid pathological conditions. HO breaks down the porphyrin ring to yield equimolar amounts of biliverdin, free iron (Fe^2 +), and carbon monoxide (CO) (Ref. Reference Beri and Chandra77). The beneficial effects of HO are related to the avoidance of the toxic accumulation of free heme as well as with the antioxidant and protective functions of bilirubin and biliverdin (Refs Reference Otterbein and Choi84, Reference Yesilkaya, Altinayak and Korgun85). Furthermore, the intermediate metabolite CO has been postulated to have protective functions since it promotes reactions that tend to protect the tissues by stimulating blood vessel relaxation (Refs Reference Vedernikov, Graser and Vanin86, Reference Ryter and Choi87).

To date, three isoforms of heme oxygenase have been identified: HO-1, HO-2 and HO-3 (Refs Reference Ryter and Tyrrell83, Reference Maines88). HO-1 and HO-2 share little similarity in amino acid sequence (40%), whereas the HO-2 and HO-3 isoform are far more homologous (90%). All HO isoforms are highly conserved among species. HO is expressed in prokaryotic bacteria as well as in fungi, plants and humans. HO proteins are microsomal and are anchored to the endoplasmic reticulum.

HO-1, -2 and -3 are products of different genes, and show distinct expression patterns and regulation (Ref. Reference Montellano89). Under normal physiological conditions, most cells express low or undetectable levels of HO-1 protein, while HO-2 is constitutively expressed. Very little is known about HO-3. HO-1 gene expression is highly inducible by several factors including heavy metals, bacterial lipopolysaccharides, hypoxia, hyperoxia, heat shock, ischaemia, UV radiation, H₂O₂, cytokines, nitric oxide, stimuli that deplete cellular glutathione stores, and its substrate heme. HO-1 gene expression is often cell-type- and species-specific (Ref. Reference Morse and Choi90). HO-2 transcription is up-regulated by only a few agents such as opiates and adrenal glucocorticoids (Refs Reference Li and Clark91, Reference Liu92). HO-1 is active even at low substrate concentrations. By contrast, HO-2 has low activity at low substrate concentrations, and even when it is at its maximal activity, it is limited to less than 10% that of HO-1 (Ref. Reference Maines and Panahian93). Specific activity of HO in different organs greatly varies. The highest HO activity is found in spleen, testis, and brain. HO-2 is abundant in brain, testis, placenta and liver. HO-3 has been identified in brain, kidney, liver, heart, testis, and spleen and has been reported to have only a very poor heme-degrading capacity (Ref. Reference McCoubrey, Huang and Maines94).

Up-regulation of HO-1 allows the acceptance of mouse and rat allografts (Refs Reference Woo95, Reference Chauveau96), while its down-regulation or absence is directly related to acute graft rejection (Ref. Reference Soares97). With regard to pregnancy, immunohistochemical staining shows that HO-1 and HO-2 are expressed in different trophoblast types of the placenta of rodents and humans (Refs Reference Zenclussen98, Reference Ihara99, Reference Zenclussen100). The expression of HO-1 protein and mRNA peaks during implantation [Ritschel, S. (2006) Diploma Thesis. Technische Fachhochschule, Berlin, Germany].

Heme protein catabolism by trophoblast-derived HO would be expected to prevent accumulation or recirculation of free heme, which could be extremely toxic to the mother and fetus. Indeed, down-regulated levels of HO-1 and HO-2 have been observed at the fetal–maternal interface in mice undergoing abortion (Ref. Reference Zenclussen98). Furthermore, human miscarriage and preeclampsia have been associated with diminished placental HO levels (Refs Reference Barber101, Reference Zenclussen102). A microsatellite polymorphism of the HO-1 gene (leading to low levels of HO-1 expression) is also associated with idiopathic recurrent miscarriage in humans (Ref. Reference Denschlag103). Interestingly, HO-1 knock-out mice are not able to reproduce (Ref. Reference Poss and Tonegawa104), suggesting a crucial role for HO-1.

Conclusions and clinical implications

Although many mechanisms have been proposed to explain fetal tolerance, further research is necessary to elucidate fully the complex net of interactions guaranteeing implantation of a ‘foreign’ fetus, its growth and normal placental development. Nevertheless, there is enough experimental and clinical evidence to conclude that both Treg and HO-1 are very important for pregnancy outcome (Fig. 1). Future work is likely to focus on investigating their mechanisms of action and their interaction with other relevant pathways. Novel approaches such as microarray technologies may help in revealing new molecules associated with tolerance. The great advantage of the ongoing animal models of fetal tolerance is that tolerance can be achieved by manipulating early events.

Understanding of the interactions of components implicated in prevention of rejection of the fetus is essential for the design of therapies to prevent pregnancy complications such as spontaneous abortion and preeclampsia. The concept of redirecting the maternal immune system to recognise and tolerate paternal cells was introduced in the late 1970s by vaccinating mothers with cells from their partners before pregnancy (Refs Reference Taylor and Faulk115, Reference Takeshita116, Reference Hill117); this was considered a promising treatment because blood transfusion given before transplants was reported to decrease the rejection of organ allografts. Although very controversial (Refs Reference Takeshita116, Reference Hill117, Reference Christiansen, Nielsen and Kolte118), this treatment has been shown to be effective in cases of immunological spontaneous abortions (Ref. Reference Mowbray119). It was recently reported at a scientific meeting (Ref. Reference Marget120) that the treatment of patients with lymphocytes from their partners led to augmented levels of Foxp3, suggesting generation of Treg. Although speculative, the idea of using maternal Treg specific for paternal antigens as a therapy is tempting. Murine data show that it is possible to generate ex vivo antigen-specific cells with suppressor activity. Whether this therapy is an applicable one remains to be discussed in the light of ethical and safety issues.