Regulatory T cells induce a privileged tolerant microenvironment at the fetal-maternal interface

Authors


Abstract

The mechanisms underlying immune tolerance during pregnancy are poorly understood. In this regard, Treg seem to play an important role in mediating maternal tolerance to the fetus. We proposed a crucial role of T regulatory cells (Treg) in avoiding immunological rejection of the fetus after observing diminished number and function of Treg in abortion-prone mice. We further confirmed the protective role of Treg during pregnancy by transferring pregnancy-induced Treg into abortion-prone mice, which prevented rejection. Here, we analyzed the mechanisms involved in Treg-mediated protection. As expected, Treg therapy prevented abortion, while expanding the peripheral and thymic Treg population. Surprisingly, the decidual levels of the Th1 cytokines IFN-γ and TNF-α were not diminished after therapy. Interestingly, the mRNA levels of leukemia inhibitory factor, TGF-β and heme oxygenase-1 at the fetal-maternal interface were dramatically up-regulated after Treg transfer, while the levels of indolamine 2,3-dioxygenase remained unchanged. Our data suggest that Treg treatment can not prevent T cell infiltration or high Th1 levels but is able to create a privileged tolerant microenvironment at the fetal-maternal interface, further shedding light onto the molecular mechanisms involved in pregnancy tolerance.

Abbreviations:
foxp3:

Forkhead box P3

HO:

Heme oxygenase

IDO:

Indolamine 2,3-dioxygenase

LIF:

Leukemia inhibitory factor

TBS:

Tris-buffered saline solution

Treg:

Regulatory T cells

Introduction

During pregnancy, the maternal immune system has to tolerate the persistence of paternal alloantigens without affecting anti-infectious immune responsiveness. Despite years of research in this area, the cellular and molecular mechanisms enabling the maternal immune system to support its semiallogeneic fetus are still poorly understood. Medawar 1 first proposed the similarities of a growing fetus with an allograft, which is still the more accurate theory to explain the paradigm of mammalian reproduction. Since normally the maternal immune system does not reject its semiallogeneic concept, pregnancy has been thought to be a state of immunological tolerance 1, 2.

Local mechanisms seem to play an important role in evading immune attack, since maternal alloreactive lymphocytes are not systemically depleted. The specialized fetal trophoblasts in contact with maternal uterine tissue might contribute to tolerance by several mechanisms, such as depleting tryptophan 3, inactivating NK cells through HLA-G expression 4 or provoking apoptosis of activated maternal lymphocytes 5. Incomplete tolerance might result in disturbed pregnancy like spontaneous abortion and pre-eclampsia. Th1/Th2 cytokine balance with Th2 predominance has been seen as a very important mechanism determining the survival of the fetus in the maternal uterus 617. However, mice genetically deficient for IL-4 and IL-10 do not show disturbed pregnancy 18, suggesting that Th2 cells are not essential for normal pregnancy and alloreactive Th1 cells must be differently regulated, as already proposed 1921.

CD4+CD25+ regulatory T cells (Treg) have recently been described as a unique subpopulation of T cells 2224. They were confirmed to play a major role in preventing autoimmunity and tolerating allogeneic organ grafts 2526. The acceptance of paternally derived tumor cells during pregnancy 2 supports the involvement of systemic regulatory processes in pregnancy. Aluvihare et al.27 suggested an important role of Treg in a model of normal murine pregnancy. We have recently reported that in the well-established murine abortion model CBA/J × DBA/2J, the accumulation of paternal alloantigen-specific Th1 cells at the decidua of abortion mice seems to be due to insufficient generation of pregnancy-induced Treg 28. BALB/c-mated CBA/J females showed on the contrary augmented number of Treg during pregnancy 28. Abortion-prone mice showed a diminished Treg activity, as analyzed by IL-10 secretion and forkhead box P3 (foxp3) protein expression 28. Moreover, fetal rejection could be completely prevented by adoptive transfer of Treg exclusively from normal pregnant mice 28. Other authors confirmed the important role of Treg in human pregnancy 29, 30.

In our model, as both mating combinations share the paternal MHC type, Treg generated to BALB/c may specifically protect maternal anti-DBA/2J responses as well, by both direct or indirect allorecognition, which might be the dominant pathway for immunoregulation by CD4+CD25+ cells 31, 32. Here, we analyzed the molecular and cellular mechanisms involved in Treg-associated pregnancy protection after transfer of Treg.

Results

The adoptive transfer of Treg from normal pregnant mice significantly diminished the abortion rate

Abortion-prone mice presented significantly augmented abortion rates as compared to normal pregnant animals, as expected (median: 20 vs. 0%, p<0.01). We further isolated CD4+CD25+ cells from thymus and spleen of 14-day normal pregnant mice (BALB/c-mated CBA/J females) and transferred 2×105 of them adoptively into pregnant abortion-prone mice (days 0–2 of pregnancy) using our established protocol 28. The transfer of CD4+CD25+ cells from normal pregnant mice completely prevented spontaneous abortion (median: 0%, p<0.05 compared to abortion-prone mice). The implantation rates were comparable between all groups (p>0.05). Abortion and implantation rates from all groups are shown in Table 1.

Table 1. Abortion and implantation rates
GroupPercentage of abortiona)Implantation rate (total number)
  1. a) The data are expressed as medians. Minimum and maximum values are shown in parentheses.

  2. b) Abortion-prone mice showed a statistically significant augmentation in the abortion rate compared to normal pregnant mice (**p<0.01) as analyzed by the non-parametric Mann-Whitney U-test.

  3. c) Mice treated with Treg showed a diminished abortion rate compared to the abortion-prone group (*p<0.05) as analyzed by the non-parametric Mann-Whitney U-test.

BALB/c-mated CBA/J 0 (0–16.6)6
DBA/2J-mated CBA/J 20 (0–60)**b)6
DBA/2J-mated CBA/J treated with Treg from BALB/c-mated 0 (0-33.3)*c)7

The number of systemic CD4+CD25+ cells as well as the local levels of foxp3 and neuropilin-1 mRNA were augmented after Treg transfer

To assess whether the transfer of pregnancy-induced Treg was able to increase the number of Treg in the recipients, the percentage of CD4+CD25+ cells was analyzed by flow cytometry in spleen, thymus and decidua. As we previously reported 28, abortion-prone mice presented diminished number of CD4+CD25+ splenocytes and thymocytes. The transfer of Treg provoked a significant augmentation of the proportion of CD4+CD25+ cells, presumably Treg, in both, spleen and thymus (Fig. 1A, B). When analyzing the percentage of CD4+CD25+ cells at the fetal-maternal interface, namely in the decidua, we were not able to confirm a diminution in the CD4+CD25+ cell population (Fig. 1C). On the contrary, a slight augmentation in the number of CD4+CD25+ cells could be observed, which may indicate activated cells. In fact, CD25 is used in the decidua as an activation marker rather than a marker for Treg 28, 33.

Figure 1.

Expansion of the CD4+CD25+ subpopulation, probably Treg, in the spleen and thymus. Abortion-prone mice presented significantly fewer (A) CD4+CD25+ Treg splenocytes and (B) thymocytes when compared to normal pregnant mice (p<0.05) as analyzed by the non-parametric Mann-Whitney-U test. The treatment of abortion-prone mice with Treg from CBA/J x BALB/c significantly expanded the Treg population in (A) the spleen (p<0.05) and (B) the thymus (p<0.05). (C) In decidua, comparable percentages of CD4+CD25+ T cells could be observed in normal pregnant, abortion-prone and Treg-treated mice, maybe due to the fact that CD25 is also an activation marker and this population may not only represent Treg but also activated cells. All data are presented as median ± 75% quartiles. *p<0.05 and **p<0.01 as analyzed by the non-parametric Kruskal-Wallis-test followed by the Mann-Whitney-U test. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

For having a more confidential marker of Treg activity, we analyzed the foxp3 and neuropilin-1 34 mRNA levels in whole tissue by real-time RT-PCR (decidua and placenta, respectively). We observed a diminution in the foxp3 and neuropilin-1 levels at the fetal-maternal interface in abortion prone-mice. Interestingly, the Treg application slightly augmented the foxp3 levels and significantly augmented neuropilin-1 levels (Fig. 2A, B). Neuropilin-1 is claimed to be a novel and very sensitive marker for Treg 34. Our data strongly suggest an expansion of the Treg population in animals treated with CD4+CD25+ cells.

Figure 2.

Increased Treg activity at the fetal-maternal interface from mice treated with Treg. Accurate markers of Treg function, (A) foxp3 and (B) neuropilin-1 mRNA levels were analyzed by real time RT-PCR in decidua and placenta respectively. Abortion-prone mice presented slightly diminished levels of foxp3 and neuropilin-1 as compared to normal pregnant mice. The levels of foxp3 were slightly up-regulated after Treg-therapy, while the placental neuropilin-1 levels were significantly augmented after Treg transfer, indicating an augmentation in Treg activity. All data are presented as median ± 75% quartiles. **p<0.01 as analyzed by the non-parametric Kruskal-Wallis-test followed by the Mann-Whitney-U test. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

Because foxp3 is known to be expressed in Treg (as confirmed in our positive control shown in Fig. 3A), but no reports indicate its expression in other cells at the murine fetal-maternal interface, we analyzed the protein expression by immunofluorescence. We observed fewer foxp3+ cells infiltrating the decidua from abortion-prone mice when compared to normal pregnant animals. Besides, Treg treatment augmented the number of foxp3+ cells infiltrating the deciduas (Table 2, Fig. 3B). Surprisingly, giant cells also expressed foxp3 (Fig. 3C). Foxp3 was recently discovered and encodes a novel member of the forkhead family of transcription factors. This structural domain is required for nuclear localization and DNA binding. Its product scurfin, transiently expressed in heterologous cells, represses transcription of a reporter containing a multimeric forkhead binding site. Upon over-expression in CD4 T cells, scurfin attenuates activation-induced cytokine production and proliferation 35. Until now, it was unknown whether foxp3 is also expressed in placental cells. Having found a positive foxp3 immunofluorescent staining in giant cells, we decided to confirm these results by analyzing the foxp3 mRNA expression in a cell line derived from giant cells, namely the Rcho-1 cells 36 derived from rat placenta. In fact, we confirmed rat foxp3 mRNA expression in this cell line (4.3×10–5 as normalized to β-actin).

Figure 3.

Foxp3 protein expression. (A) Positive staining control. Green cells represent CD4+ cells, while red cells (intranuclear staining) depict cells positive for foxp3. Please note that all foxp3+ cells are also positive for CD4. (B) This picture depicts foxp3+ infiltrating lymphocytes (arrows) in decidual samples from normal pregnant mice as analyzed by immunofluorescence. The picture was taken using an ocular magnification of 20 ×. (C) This picture, ocular magnification 40 ×, shows giant cells which are positive for the transcription factor foxp3.

Table 2. Qualitative analysis of foxp3 expression at the fetal-maternal interfacea)
Groupfoxp3+ infiltrating immune cellsfoxp3+ giant cells
  1. a) Data are presented as medians; we refrain from making statistical analysis of our qualitative staining since only five samples from each group were analyzed.

BALB/c-mated CBA/J 1.55
DBA/2J-mated CBA/J 12.5
DBA/2J-mated CBA/J treated with Treg from BALB/c-mated CBA/J 24

No changes in the Th1 cytokine production or Th1 mRNA levels could be observed after the transfer of Treg, which was effective in avoiding abortion

Because Th1 production at the fetal-maternal interface has been associated with abortion, we analyzed the impact of the Treg transfer on the ability of decidual immune cells to secrete Th1 cytokines. For doing this, we isolated decidual immune cells and stimulated them unspecifically with PMA/ionomycin, which is known to stimulate almost all memory/effector T cells to produce cytokines and would be a mirror of the in vivo situation. As expected 28, decidual immune cells from abortion-prone mice produced significantly more TNF-α than decidual lymphocytes from mice having normal pregnancy (Fig. 4A). Surprisingly, the transfer of Treg into abortion-prone mice, which was effective in preventing abortion, only slightly diminished the production of TNF-α by decidual immune cells and this diminution was not statistically significant (Fig. 4A). The production of IFN-γ by decidual lymphocytes was slightly augmented in the abortion-prone group compared to the normal pregnant group (Fig. 4B). Again, Treg transfer could not prevent high IFN-γ secretion by decidual lymphocytes.

Figure 4.

Th1 cytokine production was not modified after Treg transfer. (A) TNF-α secretion by decidual cells. Decidual cells from abortion-prone mice secrete, as already known, more TNF-α than decidual cells from normal pregnant mice. Transferring Treg from BALB/c-pregnant CBA/J mice into abortion-prone mice only slightly diminished the levels of secreted TNF-α. (B) IFN-γ levels. Decidual cells from abortion-prone mice secrete slightly more IFN-γ than decidual cells from normal pregnant mice. Transferring Treg from BALB/c-pregnant CBA/J mice into abortion-prone mice did not diminish the levels of secreted IFN-γ. (C) mRNA levels of TNF-α were augmented in abortion-prone mice compared to normal pregnant mice and could not be diminished after transfer of Treg. (D) No significant differences in the mRNA levels of IFN-γ could be observed between normal pregnant mice and mice undergoing abortion. Transfer of Treg cells from normal pregnant mice into abortion-prone mice. Augmented the IFN-γ mRNA levels. All data are presented as median ± 75% quartiles. *p<0.05 and **p<0.01 as analyzed by the non-parametric Kruskal-Wallis-test followed by the Mann-Whitney-U test. Represents outlier values. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

Because not only immune cells, but also trophoblasts and decidual cells are able to secrete cytokines, we decided to analyze the mRNA levels for both, TNF-α and IFN-γ by real-time RT-PCR in whole decidual and placental tissue separately. We confirmed a significant augmentation of TNF-α mRNA in decidual tissue from mice undergoing abortion compared to normal pregnant mice (Fig. 4C). The transfer of Treg had no effect on the TNF-α mRNA levels in decidual tissue. In placental tissue, no differences could be observed in the TNF-α mRNA levels between all groups (Table 3). We observed again no significant differences in the IFN-γ mRNA levels between normal pregnant and abortion-prone mice in decidua. The transfer of Treg surprisingly up-regulated the decidual IFN-γ mRNA levels, as can be observed in Fig. 4D. No differences could be observed in the placental levels of IFN-γ mRNA between all groups (Table 3).

Table 3. Th1 and Th2 mRNA levels (2–ΔcT) in whole placental tissuea)
CBA/J femalesTNF-αIFN-γIL-10
  1. a) Data are presented as medians. Minimum and maximum values are shown in parentheses. No statistically significant differences could be observed between any of the groups as analyzed by the non-parametric Kruskall-Wallis test followed by the Mann-Whitney U-test between two particular groups.

BALB/c-mated 1.1×10–4(5.9×10–6–1.6×10–4)1.24×10–5(5.9×10–6–1.6×10–4)5.8×10–6(1.3×10–6–2.5×10–5)
DBA/2J-mated9.7×10–5(2.2×10–5–1.8×10–4)2.5×10–5(2.1×10–6–2.5×10–4)6.5×10–6(6.0×10–7–2.9×10–5)
DBA/2J-mated and treated with Treg from BALB/c-mated CBA/J 2.1×10–4(1.1×10–5–3.8×10–4)3.6×10–5(1.1×10–5–3.8×10–4)1.43×10–5(1.3×10–6–3.4×10–5)

The transfer of Treg, which was effective on avoiding abortion, did not induce an augmentation of the classical Th2 cytokines, IL-10 and IL-4

Having found no suppressed Th1 cytokine production, we wondered whether the Treg transfer was effective in augmenting the Th2 cytokine production as a possible explanation of the therapeutic effect of the Treg transfer. We evaluated the IL-10 and IL-4 production by flow cytometry after unspecific stimulation with PMA/ionomycin and the IL-10 mRNA levels by real-time RT-PCR. We observed no differences in the levels of secreted IL-10 (Fig. 5A) or IL-4 (data not shown) by flow cytometry. Total IL-10 mRNA levels were only slightly augmented in decidua after Treg treatment (Fig. 5B). Placental mRNA levels for IL-10 were only slightly and not significantly diminished after treatment with Treg (Table 3). Interestingly, IL-10 production was significantly augmented in splenocytes after treatment with Treg (data not shown). All these data indicate that although Th2 cytokines are slightly augmented after Treg therapy, they are not involved in the success of the therapy.

Figure 5.

Th2 cytokine production was not modified after Treg cell transfer. No statistically significant differences could be observed in secreted (A) IL-10 protein or (B) IL-10 mRNA levels between all experimental groups as analyzed by the non-parametric Kruskal-Wallis-test. All data are presented as median ± 75% quartiles. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

Treg treatment provoked local changes leading to a privileged tissue microenvironment characterized by augmented TGF-β, leukemia inhibitory factor and heme oxygenase-1 levels

A tempting hypothesis by Waldmann and collaborators 23 claims that Treg may work in conjunction with tissues to establish a privileged microenvironment. These authors proposed some associated pathways such as IL-10 and TGF-β expression, indolamine 2,3-dioxygenase (IDO) expression or heme oxygenase (HO) induction 23. In this regard, we analyzed the mRNA and protein levels of some so-called tolerant molecules, namely IDO, TGF-β, HO-1 and leukemia inhibitory factor (LIF) in samples from animals receiving Treg and compared these results with those obtained from abortion-prone or normal pregnant mice.

Pregnant mice exposed to IDO inhibitor exhibited dramatically increased tendencies to lose conceptuses in allogeneic but not syngeneic mating combinations 3739. The pharmacological effects of IDO inhibitor on pregnancy outcome suggested that cells expressing IDO provide an immunosuppressive barrier that protects allogeneic conceptuses from maternal T cell immunity. All these data suggested that T cell-dependent, antibody-independent activation of maternal complement is a risk factor in allogeneic pregnancy and that IDO minimizes this risk. Our data first show, contrary to the Mellor's hypothesis, that abortion-prone animals do not express less IDO at the fetal-maternal interface compared to normal pregnant animals (Fig. 6). Recently, Mellor and Munn 40 proposed in a review article as mechanism of action of Treg that they might condition APC to acquire regulatory function by inducing the APC to express IDO. They based this hypothesis on the observations made on DC, which are able to suppress T cell responses upon IDO expression 41. Our data do not confirm this speculation, since the transfer of Treg, which was able to rescue from abortion, did not change the IDO levels, at least at the mRNA level (Fig. 6).

Figure 6.

IDO mRNA levels were not modified after Treg transfer. No changes could be observed in IDO mRNA levels between normal pregnant and abortion-prone mice. Moreover, the injection of Treg from normal pregnant mice into abortion prone-mice did not provoke an augmentation in the levels of IDO mRNA. All data are presented as median ± 75% quartiles. Statistical significance was analyzed by the non-parametric Kruskal-Wallis-test. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

When analyzing the protein levels of TGF-β at the fetal-maternal interface by immunohistochemistry, we observed a significant augmentation in the expression of this tolerant cytokine after the transfer of Treg as compared to abortion-prone mice (Table 4). The TGF-β expression in the treated group was even higher than in the normal pregnant controls (Fig. 7A–E). Especially spongiotrophoblasts, giant cells and labyrinthic cells from the placenta as well as decidual cells show a high expression of TGF-β after Treg therapy (Table 4, Fig. 7D). Analysis of the mRNA levels of TGF-β in decidual tissue depicted a trend towards an augmentation after Treg therapy (Fig. 7F). This augmentation, however, did not reach significant levels. Analysis of mRNA in whole placental tissue did not show differences between the groups (data not shown), probably due to the heterogeneity of the cell types present in this tissue.

Table 4. TGF-β protein expression at the fetal-maternal interface as analyzed by immunohistochemistrya)
GroupDecidual cellsGiant cellsSpongiotrophoblastsLabyrinthic cells
  1. a) Immunohistochemistry data are presented as medians. Minimum and maximum values are shown in parentheses.

  2. b) *p<0.05 vs. BALB/c-mated CBA/J as analyzed by the non-parametric Mann-Whitney U-test.

  3. c) *p<0.05 vs. DBA/2J-mated CBA/J as analyzed by the non-parametric Mann-Whitney U-test.

  4. d) *p<0.05 and **p<0.01 vs. BALB/c-mated CBA/J as analyzed by the non-parametric Mann-Whitney U-test.

BALB/c-mated CBA/J0.50(0.25–1)0.50(0.25–1.25)0.87(0.50–2)1.00(0.50–2.50)
DBA/2J-mated CBA/J0.75*b)(0.50–1.50)0.50(0.50–1.50)1.25(0.75–2.50)1.12(0.50–2.50)
DBA/2J-mated CBA/J treated with Treg from BALB/c-mated CBA/J1.00**d)(0.50–1.25)1.00*d)(0.50–1.25)1.50 (0.75–2.50)2.00*c)**d)(1.25–3)
Figure 7.

TGF-β levels were augmented after Treg transfer from normal pregnant mice into abortion-prone mice. (A)-(E) show representative pictures of TGF-β staining; (A) and (B) are representative samples from normal pregnant mice (BALB/c-mated CBA/J), (C) is a representative sample from abortion-prone mice and (D) a representative sample from mice rescued from abortion after transfer with Treg. (E) depicts the negative staining control. Semi-quantitative analysis of TGF-β staining is shown in Table IV. (F) The addition of Treg cells leads to a trend towards augmentation of the TGF-β mRNA levels. (A, C–E) have an ocular magnification of 20 ×, while (B) has 40 ×. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

Further, we recently reported HO-1 to be involved in pregnancy tolerance, since abortion-prone mice presented diminished HO-1 placental levels and the pharmacological or gene therapy-based augmentation of HO-1 could prevent fetal rejection 4244. Interestingly, Choi et al.45 recently reported that foxp3 expression can induce HO-1 expression. Having found augmented foxp3 levels in mice receiving Treg transfer, we wondered whether HO-1 expression was augmented. Very interestingly, HO-1 mRNA levels were slightly up-regulated in placenta and decidua from animals receiving Treg, suggesting an important role for HO-1 in Treg-induced tolerance (Fig. 8). Interestingly, the in vivo up-regulation of HO-1 by means of cobalt-protoporphirin significantly augmented the levels of neuropilin-1, a novel marker for Treg 34, confirming the link between both systems 43.

Figure 8.

HO-1 mRNA levels were slightly up-regulated after transfer of Treg. The levels of the cyto-protective HO-1 molecule were up-regulated in (A) placenta and (B) decidua after Treg transfer. However, this up-regulation was not significant as analyzed by the non-parametric Kruskal-Wallis test followed by the Mann-Withney-U test. All data are presented as median ± 75% quartiles. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

We last analyzed the levels of LIF, a protein which is essential for implantation 46 and has recently been reported to be involved in graft acceptance 47, 48. Interestingly, the Treg transfer provoked a dramatic increase in the LIF mRNA in decidua, as shown in Fig. 9. LIF mRNA levels were comparable between both groups when analyzing whole placental tissue (data not shown). These data suggest a crucial role for LIF in Treg-induced pregnancy tolerance, as it plays in graft tolerance 47, 48.

Figure 9.

The mRNA levels of LIF were significantly up-regulated in decidua after Treg transfer. All data are presented as median ± 75% quartiles. ***p<0.01 as analyzed by the non-parametric Kruskal-Wallis-test followed by the Mann-Whitney-U test. □ normal pregnant mice (CBA/J × BALB/c), ▪ abortion-prone mice (CBA/ J × DBA/2J), ▪ Treg-treated abortion-prone mice.

All together, these data indicate that Treg transfer creates a privileged microenvironment at the fetal-maternal interface characterized by high TGF-β, LIF and HO-1 levels. Our data further suggested that IDO is not being involved in Treg-induced tolerance at least at the fetal-maternal interface, confirming data on normal pregnancy in IDO-deficient mice 49.

Discussion

During pregnancy, unique tolerance mechanisms are activated from them we know very little despite years of intense research. In humans, the maternal immune system has to tolerate during 9 months the persistence of paternal alloantigens without affecting anti-infectious immune responsiveness. There is enough evidence now that the fetal-placental interface is not a rigid, close structure but does interact with the maternal immune system. Therefore, tolerant mechanisms have to exist, since happily, most pregnancies end in a living baby, without allorejection. Failure of these tolerance mechanisms might result in abnormal pregnancy, such as spontaneous abortion and pre-eclampsia. Using a murine model, we recently demonstrated for the first time the important consequences of a deficient activity of Treg during pregnancy, namely immunological spontaneous abortion 28. Abortion-prone mice showed a diminished Treg activity and fetal rejection could be completely prevented by adoptive transfer of Treg exclusively from normal pregnant mice 28. Recent data suggest that similar mechanisms should exist in humans 30.

In the present study, we concentrated on the mechanisms by which pregnancy-induced Treg generate tolerance and enable fetal survival in a well-established model of murine abortion, namely the CBA/J × DBA/2J abortion-prone combination 17. From our recent study, we know that Treg may be antigen-specific during pregnancy (at least in our murine model), since the transfer of Treg from BALB/c (H2d)-pregnant CBA/J (H2k) females could prevent fetal rejection while the transfer from Treg obtained from virgin naive CBA/J females could not 28. This suggests that alloantigen (H2d in our model) stimulation of Treg is required for mediating their protective effects in vivo. As both mating combinations share the paternal MHC type, Treg generated to BALB/c can specifically protect maternal anti-DBA/2J responses as well, both by direct or indirect allorecognition, which might be the dominant pathway for immunoregulation by CD4+CD25+ cells 31, 32.

Here, we aimed to analyze the pathways activated after transfer of CBA/J × BALB/c Treg into DBA/2J-mated CBA/J females. We confirm that the transfer of pregnancy-induced Treg expanded the CD4+CD25+ Treg population in DBA/2J-mated CBA/J females systemically (spleen) and in thymus. When analyzing the decidual Treg population, we were not able to find an expansion of this population in decidua, probably due to the fact that the measurement of CD4+CD25+ cells is not a good choice for determining the Treg subpopulation and may indicate the presence of activated cells 24, 28, 33. We therefore analyzed the protein and mRNA levels for foxp3 and neuropilin-1 as more accurate Treg activity markers. These data indicate an expansion of the Treg population locally after Treg transfer, since foxp3 and especially neuropilin-1, a novel Treg marker 34, were up-regulated. It is tempting to speculate that these Treg are acting locally by either blocking deleterious pathways or stimulating tolerant pathways.

The “Th1/Th2 paradigm” proposes that Th2-type cytokines such as IL-4 and IL-10 may favor the maintenance of mammalian pregnancy 6, 1012, whereas the excessive production of Th1 cytokines (IL-2, IFN-γ, TNF-α) would mediate the rejection of the fetus at the feto-maternal interface 10, 1317. Thus, we wondered whether the protective effects observed after transfer of pregnancy-induced Treg were related to a down-regulation of “abortive” cytokines (Th1) and/or to an up-regulation of pregnancy-protective (Th2) cytokines. Contrary to our expectations, we were not able to observe a diminished production of Th1 cytokines, namely TNF-α and IFN-γ, in animals receiving pregnancy-induced Treg when using flow cytometry. The mRNA levels for both confirmed that no diminished synthesis of Th1 cytokines is taking place after therapy, indicating that some other local protective mechanisms have to be activated to avoid the already proven deleterious effect of Th1 cytokines at the fetal-maternal interface.

Second, we investigated the protein production and mRNA levels of Th2 cytokines. Astoundingly, the production of IL-10 or IL-4 was not up-regulated in mice being treated with Treg as analyzed by flow cytometry. Consequently, the mRNA levels for IL-10 were only slightly and not significantly up-regulated after the therapy, confirming that neither a Th1 diminution nor a Th2 augmentation is related to the prevention of fetal rejection following Treg therapy. This further suggests that Treg may promote tolerance by activating alternative pathways. These results are not surprising taking into account that the combination of IL-4 and IL-10 knockout in mice resulted in normal pregnancies, suggesting that neither maternal nor fetal production of IL-4 or IL-10 is crucial for the completion of pregnancies 18. As we already proposed 28, it is tempting to speculate that in the IL-10-/IL-4-knockout mice, the lack of IL-10 or IL-4 – which are definitively related to successful pregnancy outcome in human and mice 6, 9, 50, 51 – could be “covered” or “by-passed” by high levels of other immunosuppressive molecules, i.e. TGF-β 52, 53.

Accordingly, we next analyzed the levels of TGF-β after Treg therapy. In fact, they appeared to be locally up-regulated after treatment with Treg. Augmented local TGF-β levels seem therefore to be related with Treg-induced tolerance during pregnancy. Moreover, TGF-β has been reported to be one of the molecules involved in at least in vitro Treg action 24, 54. Interestingly, the transfer of Treg as therapy against abortion is effective only very early in pregnancy, before implantation is taking place 28. Accordingly, as demonstrated by Robertson (reviewed in 55), TGF-β present in male seminal fluids appears to be activated in the female reproductive tract immediately after insemination. This would lead to immune activation necessary for implantation. This immune activation would not activate the rejection of male antigens because of the presence in seminal plasma of immunomodulatory molecules such as TGF-β 55. It remains to be elucidated whether TGF-β secreted locally by Treg has the same function than the male TGF-β and whether in the abortion-prone combination a defective TFG-β is involved in defective or incomplete tolerance at the fetal-maternal interface.

Treg were shown to inhibit in vitro Th1 but not Th2 responses 56 as well as proliferation of Th1 effector cells by secreting IL-10 and TGF-β 57. Treg can exert their actions not only by secreting IL-10 and other immunomodulatory cytokines but also by contact-dependent inhibition, particularly of APC 58. Since we did not find any evidences that the in vivo actions of Treg include inhibition of Th1 responses or stimulation of Th2 responses, we searched for additional experimental evidences for other pathways.

Waldmann et al.23 recently proposed in a very provocative article that regulation operates within a local microenvironment involving presentation of antigen. Linked suppression would further involve some kind of tissue privilege probably coordinated by Treg 23. A normal placenta, as a tolerant tissue, is expected to continuously release alloantigens which would be processed by APC from the maternal immune system under “non-inflammatory”, “non-dangerous” circumstances as proposed for other in vivo models 23. This danger-free antigen presentation (probably peripherally during pregnancy) would provoke the expansion of Treg, and further enable them to become an important population as long as the pregnancy status lasts. Indeed, conversion from CD4+CD25 into CD4+CD25+ with Treg characteristics in the periphery has been already been confirmed in an experimental model 57. Once the antigen disappears (at birth), Treg would diminish in number and, consequently, their effects would disappear, which would explain the non-acceptance of paternal grafts after pregnancy. Thus, the removal of “danger-free, tolerance-enforcing” antigen would result in the loss of tolerance 23, 58.

As the Waldmann's hypothesis claims, Treg may work in conjunction with tissues to establish a state of privilege in the tissue microenvironment. Treg may affect other pathways to finally induce this privilege. These authors proposed some associated pathways as IL-10 and TGF-β expression, IDO expression or HO induction with consequently enhanced tissue levels of CO 23. Recent data regarding IDO suggested the existence of alternative pathways leading to tolerance during pregnancy 3740. However, very recent data revealed that the mating combinations involving H2-matched or -miss-matched IDO-deficient parents presented completely normal pregnancies 49. These results lead to the suggestion that the importance of the IDO system in pregnancy outcome has been over-estimated. Our data suggest no participation of IDO in Treg-induced tolerance, contrary to a recent hypothesis by Mellor and Munn 40, proposing that during pregnancy, Treg might condition APC to acquire regulatory function by inducing the APC to express IDO. However, due to the normality of pregnancy despite the lack of IDO 49, it seems tempting to speculate that IDO might not be involved in Treg-induced tolerance, at least at the fetal-maternal tolerance.

Further, we recently reported HO-1 to be involved in pregnancy tolerance. Abortion-prone mice express low HO-1 levels in placental and decidua 42. Moreover, the pharmacological or gene therapy-based augmentation of HO-1 can rescue the pups from maternal rejection 43, 44. Furthermore, Choi et al.45 have recently reported that foxp3 expression can induce HO-1 expression. Thus, we wondered whether HO-1 expression was augmented at the fetal-maternal interface from animals receiving Treg, which had augmented foxp3 mRNA levels. Very interestingly, HO-1 mRNA levels were increased in placenta and decidua from animals receiving Treg, suggesting a clue role for HO-1 in Treg-induced tolerance. These novel data nicely support the hypothesis by Waldmann et al.23 on a privileged microenvironment involving HO-1 expression and previous own data on the importance of HO-1 during pregnancy 4244, 59, 60.

Finally, we analyzed the mRNA levels of LIF, a very interesting candidate for maternal tolerance, since it has been found to be essential for implantation success 46, 61. Animals lacking LIF gene produce normal blastocysts that fail to implant into LIF-deficient uterus but are capable of implantation in a wild-type uterus 46. It is also known that intraperitoneal injection of LIF on day 4 of pregnancy into LIF–/– mice restored implantation 62. One of the important actions of LIF is to enhance the expression of a subset of progesterone-regulated genes 61. Very interestingly, LIF was further found to be up-regulated in CBA/J mice naturally accepting skin allografts from BALB/c mice 48, indicating that it may be also involved in allotolerance. To the best of our knowledge, no previous data indicate a role for LIF after the implantation period, i.e. related to allotolerance at the fetal-maternal interface. In our study, LIF mRNA levels were found to be dramatically augmented after Treg application, suggesting an important role of LIF in Treg-induced tolerance at the fetal-maternal interface.

In summary, our data suggest that Treg at the fetal-maternal interface may be effective in avoiding allo-rejection by interacting with tissues and creating a “tolerant” privileged microenvironment characterized by high expression of TGF-β, HO-1 and LIF but can not prevent Th1 cell infiltration at the fetal-maternal interface, shedding light onto the molecular mechanisms involved in pregnancy tolerance.

Material and Methods

Animals and pregnancy outcome

Male DBA/2J mice were purchased from Charles River (USA) and CBA/J females were purchased from Charles River (France) through Charles River, Sulzfeld, Germany. BALB/c males were acquired from Harlan Winkelmann (Borchen, Germany). All animals were housed in a barrier facility. Animal care and experimental procedures were conducted in accordance with the Guide for the care and use of agricultural animals in agricultural research and teaching, published by the Consortium for Developing a Guide for the Care and Use of Agricultural Animals in Agricultural Research and Teaching, 1988, the National Research Council (NRC) and conformed to requirement of the state authority for animal research conduct (LaGetSi, Berlin, 0070/03).

As already confirmed in several studies, the mating combination CBA/J × DBA/2J represents the abortion-prone group 17, 63. Two-month-old CBA/J females were paired with 2–4-month-old DBA/2J or BALB/c males, checked twice a day for vaginal plugs and separated from the males if mated. The day of vaginal plug detection was considered as day 0 of pregnancy. The DBA-plugged CBA/J females were randomized and received PBS (n=12) or 2×105 Treg from 14-day BALB/c-pregnant CBA/J females (n=10). PBS-treated BALB/c-mated females served as normal pregnant controls (n=10). All injections were made i.v. at days 0–2 of pregnancy. Please note that adoptive transfer of cells had a comparable effect when done on day 0, 1 or 2 of pregnancy, while it was not effective when applied later on (beginning at day 4), as we described previously 28. The data presented in the present study are the means of two experiments.

On day 14 of pregnancy, the females were sacrificed, the uteri removed and the implantation sites documented. The abortion (resorption) sites could be identified by their small size accompanied by a necrotic, hemorrhagic appearance, compared with normal embryos and placentas. The percentage of abortion was calculated as the ratio of resorption sites and total implantation sites (resorption plus normal implantation sites) as described previously 17.

Treg isolation

CD4+CD25+ Treg were isolated from a mixture containing spleen and thymus cells from non-pregnant or 14-day normal pregnant mice using magnetic beads as described elsewhere 28, following the instructions of the manufacturer (MACS; Miltenyi Biotech, Germany). The purity of the preparations was between 96 and 98% in all experiments. After isolation, Treg were washed twice with cold PBS, counted, diluted at 2×105 in 200 μL PBS and injected i.v. into pregnant mice.

Sample collection

The uterine horns were opened longitudinally, and the feto-placental unit separated from the uterine implantation sites. The whole placental and decidual unit was separated individually from the respective embryo and its implantation site. Further, placenta and deciduas were carefully separated from each other for flow cytometry and molecular biology but kept together for immunohistochemistry. For flow cytometry, decidual samples were cut in small pieces and collected in Hanks’ balanced salt solution containing no Ca2+ and no Mg2+ (Sigma, Taufkirchen, Germany). For RNA isolation, placental and decidual tissues were carefully washed with cold sterile PBS pH 7.40, snap-frozen, and kept at –80°C till use. For RNA isolation, we exclusively harvested tissues from healthy implantation sites since previous studies from our laboratory failed to obtain RNA from resorption sites. For immunohistochemistry, tissues were snap-frozen and kept at –80°C till use.

Flow cytometry

Decidual immune cells were obtained as previously described 17. Spleen, thymus or decidual cells were incubated for 4 h with 50 ng/mL PMA (Sigma), 1 μg/μL ionomycin (Sigma) and 20 μM monensin (Sigma) in RPMI (Gibco, Invitrogen, Karlsruhe, Germany) with 10% FCS (Biochrom AG, Berlin, Germany) in a humidified incubator at 37°C and 5% CO2. Thereafter, the cells were washed, incubated with the surface antibodies for 10 min at 4°C in darkness and fixed with a 1% paraformaldehyde solution overnight at 4ºC. On the following day, after permeabilizing the cells with 0.2% saponin, intracellular antibodies were added for 20 min at 4°C in darkness. The cells were washed and read on a FACScan flow cytometer from BD (Germany). Washing steps were performed using FACS washing buffer [1% BSA (Sigma) and 0.1% sodium azide (Merck, Darmstadt, Germany)]. The following mAb (purchased from BD) were used: FITC-conjugated rat anti-mouse CD4 (rat IgG1), Cy5-conjugated anti-mouse CD8, PE-labeled anti-mouse CD25, IL-4, IL-10, IFN-γ and TNF-α (rat IgG2b). FITC-, Cy5- or PE-conjugated rat IgG1 or Ig2b were used as negative control in separate tubes.

Real-time RT-PCR

Tissue (placenta or decidua, 100 mg) were treated with 1 mL Trizol (Gibco, Life Technologies, Germany) and disaggregated using a homogenizer (Ultra Turrax T8; Ika, Germany). The RNA was then extracted with chloroform, precipitated with absolute ethanol, washed and finally diluted in RNAse-free water. The RNA was quantified by reading ultraviolet absorbance at 260 nm.

Total RNA (2 μg) were placed for 2 min on ice and added with dNTP (2.5 mM; Amersham Pharmacia), Dnase I (2 U/mL; Stratagene) and RNase inhibitor (40 U/mL; Promega) mixed in reaction buffer. The mix was incubated for 30 min at 37°C and further heated to 75°C for 5 min. The addition of the reverse transcriptase (200 U/mL; Amersham) and RNAse inhibitor in diethylpyrocarbonate (Sigma)-water started the reverse transcription. This reaction mixture was incubated at 42°C for 60 min followed by incubation at 94°C for 5 min.

Amplification reactions (13 μL) for murine foxp3, IFN-γ, TNF-α, IL-10, TGF-β and HO-1 as well as for rat foxp3 consisted of 2 μL cDNA, 6.25 μL mastermix containing PCR buffer, dNTP, MgCl2 and Ampli-Taq DNA Polymerase (Eurogentec, Berlin, Germany), 3 μL of the primer mix, 1.25 μL water and 0.5 μL of the fluorescent probes. PCR reactions were performed as follows: 2 min at 50°C followed by an initial denaturation step of 10 min at 95°C, followed by 15 s at 95°C and 1 min at the appropriate annealing temperature for 40 cycles. All samples were normalized to their β-actin mRNA content. All reactions were performed on the ABI Prism 7700 Sequence Detection System (Perkin Elmer Applied Biosystems). Primer and probe sequences are available upon request.

For LIF, IDO and neuropilin-1, real-time PCR was carried out with 2 μl cDNA using the I-Cycler (Bio-Rad, Germany) and the DNA binding dye SYBR Green for the detection of PCR products. Primer sequences are also available upon request. Real-time PCR assays have all been validated for their specificity, linearity and precision over the relevant range for each primer of sets in both approaches, either by using a fluorescent probe or the SYBR Green system.

Immunohistochemistry

For TGF-β3 staining, paraffin sections containing placenta and decidua were de-waxed, washed twice with Tris-buffered saline solution (TBS) pH 7.4 and treated with 3% hydrogen peroxide in methanol for 20 min at room temperature to block the endogenous peroxidase activity. The tissues were blocked with 5% BSA in TBS for 20 min at room temperature. Then, the samples were incubated overnight at 4°C with a 1:50 dilution of the primary antibody (rabbit against TGF-β3; Santa Cruz Biotechnology; please note that the antibody does not cross react with TGF-β1 or TGF-β2). After washing, the samples were stained with the secondary goat anti-rabbit biotinylated antibody (DAKO, Glostrup, Denmark) diluted 1:100 in 5% BSA for 1 h at room temperature. Then, horseradish peroxidase-conjugated solution (DAKO) was added for 30 min at room temperature after washing in TBS. Finally, the sections were developed with 3,3-diaminobenzidine (Vector, Peterborough, UK), counterstained with Hemalaun (Roth) and mounted. Negative controls were performed by replacing the primary antibody with 5% BSA in TBS, by 1% normal rabbit serum or by incubating the antibody previously with a commercial blocking peptide (Santa Cruz Biotechnology) for 2 h, as described by the manufacturer.

The pattern and intensity of staining of the different cell types of placenta and decidual samples were evaluated by two independent observers using a light microscope (Zeiss Axiophot) in a ×200 magnification (×20 objective and ×10 ocular). The degree of staining in each placental cell type, as well as in decidua, was graduated semi-quantitatively as: negative (–), weak (+), moderate (++), high (+++) or intense (++++), and these marks were later converted into numerical scores (0, 1, 2, 3 and 4, respectively) for a semi-quantitative analysis.

Immunofluorescence

Snap-frozen, Tissue-Tek-embedded placentas (five for each group) were cut at 5–7 μm, fixed for 10 min in cold acetone and use for immunofluorescence. We used an already established double-immunofluorescence staining protocol for foxp3 and CD4 64. The first antibodies were diluted in 10% BSA (self-made polyclonal rabbit antibody against murine foxp3 at a concentration of 5 μg/mL and FITC-labeled anti-CD4 antibody from BD Pharmingen diluted 1:200). The secondary antibody was applied in a 1:100 dilution (594 goat anti rabbit; Alexa, Leiden, The Netherlands). The sections were washed and mounted using a ready-to-use mounting medium with DAPI as counterstaining (Vectashield, Germany). Negative controls were performed by replacing the first antibody with 10% BSA, or 5% non-immune serum.

All sections were analyzed under a fluorescence microscope by two independent and blind observers. As positive controls, slides containing pieces of spleen from normal non-pregnant females were included. For decidual tissue, only foxp3 staining was carried out. The number of infiltrating foxp3+ cells in the deciduas was semiquantified as follows: no cells: 0; few cells in 10× field (1–3): 1; moderate number of cells in 10× field (4–6): 2; elevated number of cells in 10× field (7–10): 3. The intensity of staining in giant cells was analyzed as already described for TGF-β3. For foxp3 staining at the fetal-maternal interface, we refrained from making any statistical analysis due to the few number of samples (n=5) employed for each group.

Data analysis and statistics

All data are presented as medians or medians ± 75% quartiles. Please note that circles and asterisks in the box plots depict outliers, which were included in the statistical analysis. As suggested by our statistical advisor due to the characteristics of the groups, analysis of the differences between all groups was performed using the Kruskall-Wallis test, while differences between two groups were calculated using the non-parametric Mann-Whitney U-test. p<0.05 was considered as statistically significant.

Acknowledgements

The authors are very grateful to Dr. Brigitte Wegner (Institute of Medical Biometry, Charite, Berlin, Germany) for her advice in statistics, to Dr. Katja Kotsch (Institute of Medical Immunology) for designing the LIF primers and to Prof. Michael Soares for providing the Rcho-1 rat cell line. The present work was supported by grants from the DFG (Ze526/4–1) and the Charité (UFF2003-109 and 2003-230) to A.C.Z. M.L.Z. is a recipient of a PhD fellow from the Charité.

Footnotes

  1. 1

    WILEY-VCH

  2. 2

    WILEY-VCH

  3. 3

    WILEY-VCH

  4. 4

    WILEY-VCH

  5. 5

    WILEY-VCH

  6. 6

    WILEY-VCH

  7. 7

    WILEY-VCH

  8. 8

    WILEY-VCH

  9. 9

    WILEY-VCH

Ancillary