Immediately after implantation, maternal lymphocytes accumulate around the embryo and lymphocytes are activated. Administration of anti-CD25 antibody to mice decreases systemic Treg levels for 3 days. In mice, implantation occurs 4.5 days after mating; therefore, when anti-CD25 antibody was administered 2.5 days after mating, 2 days before implantation, the number of Tregs decreased by about 70% at the time of implantation [22]. When Tregs were depleted by anti-CD25 antibody, implantation failure occurred in the mating of BALB/C♀ and C57BL/6♂ (allo-pregnancy), but was achieved in the mating of BALB/C♀ and BALB/C♂ (syngeneic pregnancy) [21]. Similar findings have been reported previously [29, 30]. It can thus be concluded that Tregs are essential for implantation in allogeneic-pregnant mice. Similar to that in mice, Treg numbers have been found to be reduced in cases of implantation failure in humans [31].

Plaks et al. [32] reported that DCs increase in the uterus during implantation and implantation failure occurs in both allogeneic and syngeneic pregnancies when CD11c⁺ DCs are depleted during implantation. Furthermore, they demonstrated that DCs are valuable for angiogenesis, proliferation, and differentiation of decidual stromal cells [32]. This was the first report to show that DCs, innate immune cells, play an important role not only in immunity but also in implantation. Although DCs are cells that migrate between organs in the body, there has been little information on whether the increase in DC numbers during implantation is due to migration from outside the uterus or proliferation in the uterus or whether the uterine DCs change DC subsets. Using KikGR mice, the authors irradiated the uterus with UV light, causing DCs in the uterus to change to a red coloration, and subsequently examined the migration of DCs (Figure 1) [24]. Furthermore, we performed a 6-color analysis of mononuclear cells in the uterus using flow cytometry and analyzed their properties [24]. Immediately after intercourse, DCs in the uterus underwent maturation, changed into antigen-presenting cells, and migrated to the uterine regional lymph nodes. Subsequently, a large number of external immature DCs, which induce immune tolerance, flowed into the uterus. Some of the uterine DCs remained in the uterus after coitus, and we found that these DCs expressed PD-L2, which induces immunosuppression (Figure 1, Table 4). In other words, after intercourse, DCs in the uterus mature and migrate to the uterine regional lymph nodes, where they may present paternal antigens to Tregs. Furthermore, we found that immature DCs, which induce immune tolerance, migrated into the uterus from outside the uterus at the time of implantation of the allogeneic embryo (Table 4). In addition, it was observed for the first time that the DCs that stay in the uterus express PD-L2 that induces immunosuppression and change to tolerogenic DCs, which induce immune tolerance; thus, DCs in the uterus can induce tolerance against allogeneic embryos (Figure 1, Table 4). Interestingly, PD-L2 on uterine resident DCs did not change during syngeneic pregnancy (Table 4).

CD11b is expressed on M1 macrophages and M2 macrophages. When M1 and M2 macrophages (pan-macrophages) are depleted during implantation using CD11b-DTR mouse, the vascular structure of the corpus luteum is destroyed, progesterone production is reduced; and implantation failure occurs (Table 5) [33]. However, it is unclear whether M1 or M2 macrophages are important for implantation. CD206 is known to be a specific marker for M2 macrophages. Using CD206-diphtheria toxin receptor transgenic (CD206DTX) mice, we depleted CD206⁺M2 macrophages during implantation by administering a small amount of diphtheria toxin [34]. Implantation failure occurred, but there were no histological differences in the corpus luteum, and plasma progesterone levels were also normal. Thus, it is speculated that M1 macrophages are important in luteogenesis. In vitro fertilization (IVF) of eggs derived from mice with depleted CD206⁺M2 macrophages and transfer of embryos into the uterus of wild-type mice resulted in a normal implantation rate [34]. In other words, the embryos developed normally, even without CD206⁺M2 macrophages. On the other hand, in the endometrium of CD206⁺M2 macrophage-depleted mice, proliferation of endometrial epithelial cells continued [34]. Macrophages are classified into M1 macrophages, M2 macrophages and other macrophages. Since CD11b is expressed on all macrophages, the following can be inferred from the combination of Care’s papers and ours. Therefore, CD11b⁺M1 or CD11b⁺ other macrophages are necessary for luteogenesis, and CD206⁺M2 macrophages are necessary for the endometrial epithelium to stop proliferating and transform into a differentiated endometrial epithelium suitable for implantation (Table 5). Further investigation revealed that when M2 macrophage numbers were reduced, inflammatory cytokines such as TNF-α were produced by M1 macrophages, which increased Wntβ expression in the endometrial epithelium, and in turn increased β-catenin expression; this led to increased fibroblast growth factor (FGF) 18 expression, which proliferated the endometrial epithelium, resulting in implantation failure [34]. Retinoblastoma (RB) protein, encoded by Rb1, is a prominent inducer of cell cycle arrest. Akaeda et al. [35] reported that uterine-specific Rb1 knockout mice showed implantation failure despite normal progesterone levels in the blood and normal embryo quality, while endometrial epithelial cells showed persistent cell proliferation. They reported implantation failure in this model. Our data and Akaeda’s study showed that excessive proliferation of endometrial epithelial cells at the implantation window period induces implantation failure. Endometrial biopsies were performed at the implantation stage in human implantation failure cases. The ratio of M2/M1 macrophage mRNA was significantly lower in the unsuccessful pregnancy group than in the successful pregnancy group in the subsequent IVF embryo transfer (IVF-ET) [34]. It is possible that M2 macrophage levels are lower in patients with clinically normal blood progesterone levels but implantation failure occurs due to insufficient differentiation in the secretory phase of the endometrium, determined through endometrial dating or endometrial receptivity analysis.

Inflammation is necessary for implantation; however, excessive inflammation can lead to implantation failure [36]. Excessive inflammation can occur when M1 macrophages are dominant over M2 macrophages, when the Th1/Th2 ratio is Th1 dominant, when mature DCs are dominant over immature DCs, and when there are few Tregs (Figure 2). As a treatment for such cases, M1 macrophages can be induced to differentiate into M2 macrophages by administering M-CSF or G-CSF to optimize the M1/M2 macrophage ratio [37, 38]. In some cases, G-CSF has been reported to improve implantation rates in patients with recurrent implantation failure [39]. There is also a report that G-CSF tended to increase clinical pregnancy rates in IVF-ET patients (RR 1.62, 95% CI: 0.86–3.08). However, no statistically significant differences were observed [40]. For example, administration of G-CSF in patients with decreased M2 macrophages may improve pregnancy rates. The use of immunosuppressive agents such as tacrolimus [41] and anti-TNF-α antibodies [42] may also improve implantation failure in M1 macrophage predominant cases by controlling excessive inflammation. Tregs regulate inflammation and depletion of Tregs at implantation induces implantation failure [21, 29, 30]. If we can develop a method to increase the number of Tregs in the uterus during the implantation period, it may be applied to the treatment of implantation failure. Vitamin D supplementation [43] and high-dose intravenous immunoglobulin (IVIG) therapy [44] may also induce Treg formation and improve implantation failure [45]. Broad spectrum antibiotics may help treat implantation failure in the presence of endometritis [45].

On the other hand, mechanical stimulation of endometrial cells by the blush technique and induction of differentiation into M1 macrophages by GM-CSF may improve implantation failure in situations where inflammation is low (Figure 2). However, a recent multicenter randomized control study revealed that therapeutic endometrial scratching did not increase the implantation rate [46]. Patient selection is necessary to demonstrate clinical improvement by endometrial scratching.

Human miscarriages occur at a frequency of 10–15% in the 20s and 30s, but the frequency of fetal chromosome abnormalities increases with age, especially in women over 40 years of age, and the frequency of miscarriages increases to 40–60% [47]. Approximately 60–80% of miscarriages are due to fetal chromosome abnormalities (mostly aneuploidies), resulting in failure of the fetus to survive. It is well known that chromosomal structural abnormalities of patients and/or partners, uterine malformations, hypothyroidism, and antiphospholipid antibodies are risk factors for RPL, and these risk factors are identified in only 30–40% of cases [47–49]. In Japan, 65% of cases have no risk factors identified despite various tests [48]. We hypothesized that miscarriages with unknown risk factors and normal fetal karyotypes are caused by abnormal maternal immunity; therefore, we mainly examined Tregs in the decidua and peripheral blood in such cases.

Immunohistological examination of the localization of Tregs using Foxp3 immunostaining revealed that in normal pregnancies, Tregs were clustered in the decidua basalis near the implantation site. In miscarriages with abnormal fetal karyotypes, Tregs in the decidua basalis were present in the same number as in normal pregnancies. In contrast, the localization of Tregs in the decidua basalis was significantly decreased in miscarriages with normal fetal karyotypes [50]. Interestingly, there was no difference between the three groups in the non-implantation area, that is, changes in Treg localization were observed only around the implantation area in miscarriage cases of normal fetal karyotypes [50].

Foxp3 is a marker of Tregs but is also expressed on activated T cells that have no immunoregulatory ability [51]. Foxp3 and CD45RA antigen expression can be used to identify effector Tregs with strong immunosuppressive activity (Foxp3^high CD45RA⁺), and activated T cells (Foxp3^dim CD45RA⁻), naive Tregs with weak immunosuppressive activity (Foxp3^dim CD45RA⁻), which have no immunosuppressive activity [51]. These Foxp3⁺ cells were examined in normal pregnancy, miscarriages with abnormal fetal karyotype, and miscarriages with normal fetal karyotype. It was found that Foxp3^high effector Tregs numbers were significantly decreased and Foxp3^dim activated T cell numbers were significantly increased in the decidua in miscarriages with normal fetal karyotypes [52]. These results suggest that rejection occurs in normal fetal karyotype miscarriages because the number of effector Tregs, which protect the fetus from maternal immune cells, decreased in the decidua and the number of activated T cells, which attack the fetus, increased in the decidua in humans. On the other hand, peripheral blood effector Treg, naive Treg, and activated T cell populations were the same among the three groups [52], suggesting that abnormal immune status was only present at the feto-maternal interface.

In humans, it is not possible to identify fetal antigen-specific Tregs as in the mouse experimental system. The diversity of TCRs that recognize antigens is known to be more than 2 × 10⁷. Therefore, if Tregs recognize fetal antigens locally, it was presumed that Tregs with the same TCR repertoire sequence would increase clonally because Tregs are stimulated to proliferate by recognition of fetal or paternal antigens. We isolated effector Tregs from peripheral blood and decidua basalis using flow cytometry and performed reverse transcription polymerase chain reaction (RT-PCR) on each cell to sequence TCR sequences from Foxp3 mRNA⁺ cells, which are markers of Tregs. Foxp3⁺ Tregs with the same TCR sequence were defined as clonal Tregs and used as surrogate markers for fetal antigen-specific Tregs. The results showed that clonal Tregs were rarely present in peripheral blood but were present in early pregnancy decidua at a median of 4.5% (1.4–10.8%) [53]. In addition, all three cases that we were able to analyze in the previous and next pregnancies, respectively, had Tregs with a common TCR sequence between the previous and current pregnancies. Since there is more than 2 × 10⁷ TCR diversity, the likelihood of having the same TCR sequence is only 1/2 × 10⁷. So it is possible that the previous and current fetuses shared the same major histocompatibility antigens or other minor antigens. It is highly likely that mother Tregs recognize the same major histocompatibility antigens or other minor antigens expressed on trophoblasts and clonally expand. Next, we examined clonal Tregs in decidua and compared them among the three groups of normal pregnancy cases, miscarriages with abnormal fetal karyotypes, and normal fetal karyotypes, and found no differences [53, 54]. But the frequency of effector Tregs significantly decreased in miscarriage with normal fetal karyotype. In the case of miscarriage with normal fetal karyotype may be caused by anon-specific immunosuppression rather than paternal antigen-specific tolerance.

Cytotoxic T cells that recognize the paternal antigens attack the fetus resulting in miscarriage in animal model. There is an increased number of effector memory (EM) CD8⁺ T cells in the decidua that may attack the fetus in human [55, 56]. Interestingly, clonal CD8⁺ T cells are present in EM CD8⁺ T cells but are rarely present in naive CD8⁺ T cells in humans [55]. Expression of programmed cell death 1 (PD-1) in cytotoxic T cells results in decreased cytotoxic activity. Kinder et al. [57] reported that PD-1 and lymphocyte activating gene-3 (LAG-3) selectively suppress CD8⁺ T cell cytotoxicity with fetal antigen stimulation in mouse. In cases of normal fetal karyotype miscarriage, the number of clonally expanded EM CD8⁺ T cells among CD8⁺ T cells was significantly increased by approximately 30% compared to that in normal pregnancies (15%). In the early pregnancy period, approximately 80% of clonal EM CD8⁺ T cells are PD-1⁻; therefore, the increased frequency of clonal EM CD8⁺ T cells in miscarriages with normal fetal karyotype shows that the frequency of cytotoxic CD8⁺ T cells against fetal antigens increased. In addition to the decrease in Treg levels, the increase in PD-1⁻ cytotoxic CD8⁺ T cell numbers in the miscarriage cases with normal fetal karyotype was thought to be responsible for the attack on the fetus.

If Tregs are to be used as a treatment for miscarriage in the future, drug therapy to increase uterus specific Treg pools may be effective [54].

Excessive inflammation has been shown to cause miscarriage not only in mice but also in humans, and vitamin D supplementation [56] and high-dose IVIG therapy [58, 59], which increase Treg cell numbers, have been reported to reduce miscarriage in RPL. Anti-TNF-α antibodies have also been reported to reduce miscarriage in cases of RPL [60]. On the other hand, there have been no reports on whether a low inflammatory state can cause RPL; therefore, such RPL is thought to be rare.

As we have discussed, the induction of immune tolerance is necessary to maintain pregnancy. The epidemiological risks of PE and miscarriage are different (Table 6). PE is more common in first pregnancies, and the risk is lower in women who have had two or more pregnancies, including previous miscarriages. Even in multiparous women with new partners, the risk of PE was higher in the first pregnancy after a partner change, suggesting that the deficiency of paternal antigen-specific tolerance is associated with the risk of PE. On the other hand, the amount of Tregs in the decidua was lower in miscarriage cases with normal karyotype fetuses, which induced different epidemiological features between PE and RPL.

Seminal plasma induces paternal antigen-specific Tregs; therefore, the increased risk of PE in condom use, short cohabitation, IVF and sperm donation (artificial insemination by donor, AID) pregnancies may be related to inadequate induction of paternal antigen-specific Tregs. In current IVF-ET, embryos are transferred to the uterus without seminal plasma priming, which may result in insufficient induction of paternal antigen-specific Tregs, a lower implantation rate, and a higher risk of PE.

In conventional pregnancies, the fetus is half-self and half-father, but in oocyte donation pregnancies, embryo-donation pregnancies, and hydatidiform moles, the fetus is completely allografted to the maternal host. In this case, stronger immune tolerance is required for maintenance of allogeneic pregnancy, and if it is insufficient, the frequency of both PE and miscarriage may increase.

From the abovementioned findings, it can be inferred that PE is characterized by a decrease in fetal (paternal) antigen-specific Treg numbers, while miscarriage is characterized by a decrease in overall Treg numbers rather than antigen-specific Treg numbers; however, this has not been fully elucidated scientifically.

The authors were the first to show that Treg reduction in peripheral blood was observed in PE [61]. We also showed, for the first time, that a decrease in Treg levels in peripheral blood was observed in PE [61], and we found a significant decrease in effector Treg levels among Treg levels [62]. Placental bed biopsy of oocyte-donated pregnancies was performed to examine the localization of immunocompetent cells by immunostaining. The numbers of macrophages, NK cells, CD4⁺ T cells, and Tregs were found to be reduced [63]. Since these cells are involved in the remodeling of uterine spiral arteries by trophoblasts [12, 16, 17, 64, 65], cytokeratin 7 (CK7) staining of trophoblasts around spiral arteries showed that remodeling failure of spiral arteries occurred in oocyte-donation pregnancies, which was proven to be a factor contributing to the high risk of PE [63].

To prove the fetal (paternal) antigen-specific tolerance deficiency in PE cases, we examined Treg clonality using the analysis of TCR repertoire in single cell level. The results showed that the number of clonal Tregs in the decidua increased to a median of 20.9% (15.4–28.1%) in late normal pregnancy but significantly decreased to a median of 9.3% (4.4–14.5%) in PE [53]. These results are consistent with epidemiological findings and demonstrate, for the first time, that one of the etiologies of PE is inadequate induction of fetal (paternal) antigen-specific tolerance. In the future, treatment of PE will require induction and administration of paternal antigen-specific Tregs, which will pose one more hurdle than treatment to increase the overall number of Tregs. The percentage of clonally expanded EM CD8⁺ T cells in the decidua basalis of PE cases was approximately 20%, which was not significantly different from the 15% observed in normal pregnancy cases. In contrast, the PD-1 expression rate in clonally expanded EM CD8⁺ T cells was significantly lower in PE cases, with a median of 40%, and 65% in normal pregnancy cases [56]. In other words, PD-1⁻CTL cells were significantly increased in PE cases compared to those in normal pregnancy cases. This suggests that the fetus is attacked by an increase in PD-1⁻ cytotoxic CD8⁺ T cell numbers, in addition to a decrease in the number of Tregs that recognize fetal antigens.

The etiology of PE is thought to be shallow placentation due to failure of EVT invasion. In normal pregnancy, EVT invasion extends up to 1/3 of the uterine wall, but in PE, the invasion is shallower, resulting in inadequate remodeling of the spiral artery, which in turn is thought to cause PE. This deep invasion of EVT into uterine muscle wall is specific to primates, making it difficult to model in rodents for experimental use. However, occlusion of uterine artery induces PE-like symptoms in rodents [66]. Using this model, it has been reported that Tregs are decreased in this model, and injection of Tregs improves the PE-like symptoms [67], suggesting that Tregs may be useful for the treatment of PE.

NK cells recognize HLA-C expressed on trophoblasts by killer cell immunoglobulin-like receptors (KIRs), and KIR AA activates NK cells. Interestingly, the combination of the lack of KIR AA on maternal NK cells and the expression of HLA-C2 in the fetus have been reported to increase the risk of PE [68]. In this combination, NK cells may not be activated and cytokine secretion such as VEGF that promotes angiogenesis may be inadequate. Indeed, VEGF production by T cells and NK cells is decreased in PE [69]. Although there is one study that does not support this report [70], it is interesting to note that the risk of PE increases when the immune system is insufficiently activated.

On the other hand, when PE develops, the Th1/Th2/Th17/Treg balance becomes Th1 cell dominant [71], the number of Th17 cells increases [72], and the number of Tregs decreases [73]. In addition, the number of clonal cytotoxic T cells that attack the fetus increases, and PD-1 expression on these cells also decreases [56]. These results indicate that the immune system is activated in PE. This is contrary to the results of previous studies, which showed that insufficient activation of the immune system is a risk factor for the development of PE [68]. A certain amount of immunoactivation is necessary for pregnancy, but too much activation may lead to PE.

Since seminal plasma exposure is important for paternal antigen-specific Tregs induction, it may be possible to prevent the development of PE if oral contraceptives are recommended instead of condom contraceptive methods, with careful attention to sexually transmitted diseases. However, it is unclear whether this would be effective since there are no such intervention trials.

In preforming IVF-ET, there is no paternal semen exposure in the normal course of the operation. Although it has been reported that intravaginal seminal plasma administration increased the pregnancy rate in IVF-ET [28], whether it suppresses the development of PE has not been reported yet, and we look forward to future results.

If cell therapy using Tregs will be used, it would require the administration of in vitro expanded paternal antigen-specific Tregs is desirable. Because paternal antigen-specific Tregs into PE cases decrease in PE.

PD-1 agonistic antibodies may be effective to manage PE because PD-1 expression is decreased in cytotoxic T cells, but at present, there are no such clinical trials.

IL-10 administration [74] and anti-TNF-α antibodies [75] may be effective in regulating systemic inflammation at the onset of PE. Although their efficacy has been reported in animal studies, there are no clinical trials at this time, and their efficacy is unknown. Thus, as the immune abnormalities in PE are gradually being identified, immunological treatments may be developed in the near future.