RANKL-mediated harmonious dialogue between fetus and mother guarantees smooth gestation by inducing decidual M2 macrophage polarization

Decidual macrophages (dMφ) contribute to maternal–fetal tolerance. However, the mechanism of dMφ differentiation during pregnancy is still largely unknown. Here, we report that receptor activator for nuclear factor-κ B ligand (RANKL), secreted by human embryonic trophoblasts and maternal decidual stromal cells (DSCs), polarizes dMφ toward a M2 phenotype. This polarization is mediated through activation of Akt/signal transducer and activator of transcription 6 (STAT6) signaling, which is associated with the upregulation of histone H3 lysine-27 demethylase Jmjd3 and IRF4 in dMφ. Such differentiated dMφ can induce a Th2 bias that promotes maternal–fetal tolerance. Impaired expression of RANKL leads to dysfunction of dMφ in vivo and increased rates of fetal loss in mice. Transfer of RANK+Mφ reverses mouse fetal loss induced by Mφ depletion. Compared with normal pregnancy, there are abnormally low levels of RANKL/RANK in villi and decidua from miscarriage patients. These results suggest that RANKL is a pivotal regulator of maternal–fetal tolerance by licensing dMφ to ensure a successful pregnancy outcome. This observation provides a scientific basis on which a potential therapeutic strategy can be targeted to prevent pregnancy loss.

Pregnancy constitutes a major challenge to the maternal immune system, which must tolerate fetal alloantigen encoded by paternal genes. 1,2 The disturbance of maternal-fetal immune regulation is associated with several complications of human pregnancy, including spontaneous abortion (SA), intrauterine growth restriction (IUGR) and preeclampsia. [3][4][5] Accumulating evidence indicates that decidual macrophages (dMφ), the second largest decidual leukocyte population during the first trimester (~20%) following decidual NK cells (dNKs, 50-70%), are involved in several processes required for a successful pregnancy, including trophoblast invasion, as well as tissue and vascular remodeling. 6,7 However, the mechanisms responsible for dMφ differentiation and polarization at the maternal-fetal interface remain largely unexplored.
Of note, two distinct states of polarized activation of macrophages have been recognized: the classically activated (M1) macrophage phenotype and the alternatively activated (M2) macrophage phenotype. [8][9][10] Bacterial moieties such as LPS and TH1 cytokine interferon-γ (IFN-γ) polarize macrophages toward the M1 phenotype. These M1 macrophages are characterized by high interleukin (IL)-12 and IL-23 and low IL-10 production, and accordingly can kill intracellular microorganisms and induce Th1 immunity. In contrast, M2 polarization was originally discovered as a response to the Th2 cytokines IL-4 and IL-13, the anti-inflammatory cytokine IL-10, M-CSF, glucocorticoids and immune complexes. They generally share characteristics such as high IL-10 and low IL-12 and IL-23 production, anti-inflammatory and tissue remodeling properties, and scavenging of apoptotic cells and debris, and therefore have been considered to be important regulators of the immune response.
The dMφ were classified as resembling an M2 phenotype. 11 However, there is still ambiguity with regard to the distinct functions of the dMφ subset. 4,6,12 Recent research has revealed that first-trimester dMφ can be divided into two distinct subsets, CD209 + and CD209 − dMφ. 6,12 In comparison with CD209 − dMφ, CD209 + dMφ express high levels of the scavenger receptor CD163, the phagocytic receptors CD206 and CD304, and the CD209 ligand ICAM-3, and low levels of CD11c, which are associated with spiral arteriole remodeling. 6 Receptor activator of NF-κB ligand (TNFSF11, also known as RANKL) and its tumor necrosis factor (TNF)-family receptor RANK are essential regulators of osteoclast differentiation and thereby fundamental aspects of bone physiology, bone remodeling, 13,14 lymph node formation, 15 establishment of thymic microenvironment, 16 mammary gland development during pregnancy 17,18 and bone metastasis of cancer. 19 Osteoprotegerin (OPG) is a decoy receptor for RANKL. By binding RANKL, OPG blocks the RANKL-RANK interaction. Osteoclasts are derived from monocyte/macrophage precursors. 13 However, the role of RANKL in inducing macrophage differentiation and functional regulation at the maternal-fetal interface is largely unknown.
In this study, we investigated the effect of RANKL from human embryonic trophoblasts and maternal DSCs on dMφ differentiation and maternal-fetal immune tolerance, and we analyzed the relationship of RANKL production at the interface with miscarriage.

Results
The crosstalk between fetus and mothers leads to high levels of RANKL/RANK expression at the maternal-fetal interface. To investigate the role of RANKL/RANK signaling at the maternal-fetal interface, we first analyzed the expression of RANKL and found that both embryonic trophoblasts from villi and maternal DSCs from decidua are positive for RANKL in human first-trimester pregnancy (Figures 1a and  b). As observed by immunohistochemistry, RANKL expression was located both in cell membrane and cytoplasm ( Figure 1a). Similar results for RANKL expression levels were obtained by ELISA and flow cytometry (FCM) of the isolated trophoblasts and DSCs. In comparison with DSCs, trophoblasts secreted more soluble RANKL (sRANKL) and expressed higher level of membrane RANKL (mRANKL), and the co-culture of trophoblasts and DSCs produced higher levels of sRANKL ( Figure 1b).
To identify target cells of RANKL at the maternal-fetal interface, we analyzed the expression of RANK in decidual leukocyte cells (DLCs) primarily isolated from human decidual tissue. Among DLCs, the percentage of RANK + cells in CD45 + CD14 + dMφ, as well as decidual NKT and CD3 + T cells was 80% ( Supplementary Figure 1a), which was 4.45-fold higher than that of CD14 + monocyte cells from peripheral blood (pMo) (Figure 1c).
Interestingly, further phenotypic analysis showed that both RANK + pMo and RANK + dMφ expressed higher levels of M2polarized macrophages markers (10) (the scavenger receptor CD163, the phagocytic receptors CD206 and CD209) and M1 co-stimulatory molecules CD80, CD86, HLA-DR, CD11c and cytokine IL-12p40, compared with RANK − cells (Figure 1d and Supplementary Figure 1b). These significant differences in M2 phenotype between RANK + pMo and RANK − pMo was maintained and even augmented in dMφ. Conversely, the M1 advantages possessed by RANK + pMo gradually weakened in RANK + dMφ. In comparison with RANK − dMφ, the expression pattern of RANK + in dMφ suggests that RANK signaling may regulate the differentiation and function of dMφ.
Embryonic extravillous trophoblasts (EVT) are in direct contact with maternal DSCs and DLCs, and the interaction between these cell subsets has an important role in maintaining maternal-fetal tolerance, which is required for a successful pregnancy. 3 To investigate this intercellular crosstalk, we cocultured primary human trophoblasts, DSCs and dMφ (T+D +M). FCM analysis revealed markedly increased expression of RANKL in trophoblasts and DSCs and RANK on dMφ in the T+D+M co-culture (Figures 1e and f). The percentage of RANK + dMφ was reduced to 60% after culture alone for 48h. The presence of embryonic trophoblasts in the absence of DSCs led to a frequency of 80% RANK + dMφ (Figures 1e and f). Taken together, these data suggest that embryonic trophoblasts may have a crucial role in the regulation of maternal dMφ differentiation and function through the RANKL-RANK interaction.
RANKL from trophoblasts and DSCs triggers M2 differentiation of dMφ and Th2 bias. To investigate the potential effect of RANKL on dMφ, we directly co-cultured the purified CD14 + dMφ with RANKL-overexpressed DSCs and JEG-3 cells (human placental choriocarcinoma cell line) (RANKL + D +J). Compared with control DSCs and JEG-3 cells (Ctrl-D+J), RANKL + D+J resulted in the elevation of CD206, CD209, CD163 and IL-10 and the decline of IFN-γ, IL-12/23p40, CD80 and CD86 in CD14 + dMφ (Figures 2a and b). In contrast, blocking the RANKL-RANK interaction with a neutralizing antibody against RANKL (α-RANKL) or OPG protein could reverse the expression of CD80 and CD86 and the release of IL-10, IL-12p40 and IL-23 induced by RANKL of trophoblasts and DSCs (Supplementary Figure 2). RANKL participates in the regulation of monocyte migration by inducing the secretion of chemokine, such as CCL22 and CCL2. 20,21 However, current results show that RANKL was not involved in the regulation of the M2-recruited chemokines CCL17, CCL22 and CCL24 and the M1-recruited chemokines CXCL9 and CXCL10 of dMφ (data not shown). We have previously reported that RANKL stimulates DSCs to secrete the chemokine CCL2. 22 Our study suggests that RANKL may recruit peripheral monocyte to decidua, and further polarize them toward M2 dMφ.
Evidence suggests that the RANKL-RANK interaction increases macrophage/dendritic cell (DC) survival and enhances the induction of T-cell response. [23][24][25] However, published evidence shows that RANKL-mediated modulation of DCs in mucosal tissues increases the number of CD4 + CD25 + regulatory T (Treg) cells and promotes immunosuppressive activity toward foreign antigens, such as food or commensal bacteria in the intestines. [26][27][28] However, the molecular mechanism underlying RANKL on Mφs and DCs remains unclear. We further investigated the regulation of these RANKL-instructed dMφ on the differentiation of decidual Figure 1 The crosstalk between fetus and mothers leads to high levels of RANKL/RANK expression at the maternal-fetal interface. (a) RANKL expression in villi and decidua of normal pregnancy (n = 12) by immunohistochemistry. Original magnification: ×200. (b) RANKL secretion by primary trophoblasts (1 × 10 5 cells per well) and DSCs (1 × 10 5 cells per well) (n = 6) from normal pregnant women by ELISA after culture for 24-96 h (left). RANKL production by trophoblasts alone, DSCs alone and the co-culture of trophoblasts and DSCs (n = 6) for 48 h (right). (One-way ANOVA). (c) We isolated primary PBMCs from peripheral blood (n = 6) and DLC from deciduas of normal pregnant women (n = 6), and then analyzed RANK expression on pMo and dMφ from normal pregnant women by labeling with anti-CD14, RANK and CD45 antibodies. (Student's t-test). (d) Further analysis of the phenotype of RANK + and RANK − pMo and dMφ from normal pregnant women (n = 24) by FCM. (One-way ANOVA). (e and f) Trophoblasts, DSCs and/ or dMφ (n = 6) were co-cultured at a 1 : 1 : 1 ratio for 48 h, and then RANKL expression on CK7 + trophoblasts and CK7 − DSCs, and RANK on CD14 + dMφ were evaluated by FCM, respectively. (One-way ANOVA). Tro: human trophoblasts. pMo: human peripheral blood monocytes; dMφ: human dMφ. Data are expressed as the mean ± S.E.M. *Po0.05, **Po0.01 and ***Po0.001. NS: no statistical difference naive T cells. After pre-culture with trophoblasts and DSCs, we collected dMφ, and then co-cultured them with decidual naive T cells for 5 days (Figure 2c). We observed that either α-RANKL or OPG abolished the stimulatory effect on the IL-10 and Th2 transcription factor GATA-3 and the inhibitory impact on tumor necrosis factor-α (TNF-α) and Th1 transcription factor T-bet in CD4 + T cells mediated by dMφ pre-treated with trophoblasts and DSCs (D+T-dMφ) (Figures 2d-g). In addition, blocking RANKL in the T+D+M co-culture further inhibited IL-4 secretion but stimulated IFN-γ production of CD4 + T cells (Figures 2d-g). However, RANKL-instructed dMφ had no effect on decidual Treg cell differentiation (data not shown).
As a potent inducer of decidual M2 Mφ, 6 the increased IL-10 in dMφ and decidual CD4 + T cells induced by RANKL may further amplify the impact of RANKL on M2 polarization of dMφ. These data provide strong evidence that RANKL is (Student's t-test). (c-g) After 48 h of culture with or without trophoblasts and DSCs and treatment with or without recombinant human OPG protein (rhOPG, 100 ng/ml) or antihuman RANKL neutralizing antibody (α-RANKL, 5 μg/ml), CD14 + dMφ (n = 6) were collected and co-cultured with autologous decidual naive T cells at ratios of 1 : 1 (c). The decidual naive T cells had been previously activated with anti-CD3 (5 μg/ml), anti-CD28 (1 μg/ml) and rhIL-2 (20 U/ml) for 3 days, and then collected. After 5 days of co-culture, the expression of GATA-3 and T-bet in CD4 + T cells (e-g) were analyzed by FCM; alternately, these CD4 + T cells were collected and reactivated with anti-CD3 and anti-CD28 alone for another 24 h, and then the secretion levels of IL-4, IL-10, TNF-α and IFN-γ (d) were analyzed by ELISA. expressed at the maternal-fetal interface, and support the presence of a positive regulatory loop between trophoblasts and dMφ to induce maternal-fetal immune tolerance during pregnancy.
The effect of RANKL on dMφ is dependent on the activation of the Akt/STAT6-Jmjd3/IRF4 signaling pathway. Of note, mRANKL or sRANKL cleaved by matrix metalloproteinases (MMPs) or a disintegrin and metalloproteases (ADAMs) binds RANK and then mainly activates the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) pathway to control osteoclastogenesis through adaptor molecules such as TRAFs and Gab2, 29-31 thus regulating osteoimmunology by pro-and anti-inflammatory effects on the immune system. 32 To identify the downstream pathway through which RANKL drives M2 polarization during pregnancy, we assessed the impact of RANKL on different signaling pathways in dMφ. T+D+M led to changes in the phosphorylation of several signaling pathways such as Akt, NF-κBp65 and c-Jun N-terminal kinases (JNK) (data not shown), suggesting that several pathways should be involved in the functional regulation on dMφ. Notably, α-RANKL or OPG specifically inhibited the activation of Akt and STAT6 (a master regulator of M2 genes in mice downstream of IL-4R) [33][34][35] in dMφ (data not shown). Culture with RANKL + -D+J gave rise to the increased level of Akt and STAT6 phosphorylation in dMφ compared with the culture with Ctrl-D +J (Figures 3a and b). The Akt signal inhibitor (Akti) LY294002 partly reversed the level of STAT6 phosphorylation ( Figure 3c).
Compared with the control, the release of IL-10 and IL-12p40 and IL-23 by dMφ increased and decreased, respectively, under co-culture with RANKL + -D+J, and these effects could be clearly impaired by Akti or STAT6 inhibitor (STAT6i) AS1517499 35 treatment ( Figure 3d). Interestingly, treatment with STAT6i resulted in M1 differentiation and Th1 bias in the mice uterus in vivo ( Figure 3e). Therefore, these data suggest that RANKL induces M2 differentiation by activating the Akt/STAT6 signaling pathway.
It has been reported that Jmjd3 regulates M2 polarization by inhibiting the transcription of typical M1-associated genes and inducing IRF4. [36][37][38] Unlike IRF5, 39 IRF4 has been recognized as an essential transcription factor for M2 polarization and the expression of M2 signature genes such as Arg1, Ym1 and Fizz1 in mice. 40 We investigated the effects of RANKL on these transcription factors. Blocking RANKL resulted in a marked decrease in Jmjd3 and IRF4 but not IRF5 transcription in the co-culture of D+T-dMφ (Supplementary Figure 3). Conversely, there was an increase in the transcription of Jmjd3 and IRF4 in RANKL + -D+J-dMφ, and both Akti and STAT6i treatment could partially abrogate these effects on Downregulation of RANKL expression leads to murine decidual Mφ dysfunction and fetal loss. In comparison with non-pregnant mice, uterine Mφ (uMφ) from pregnant mice had a high level of RANK (Supplementary Figure 4). To further explore the role of absent RANKL at the maternalfetal interface during the differentiation of Mφ in vivo, we first evaluated RANKL/RANK expression in the uterus in normal pregnancy and abortion-prone matings. In CBA/J × DBA/2 matings as an abortion-prone model, decreased RANK expression was detected in F4/80 + uMφ on gestational days 5 and 9, compared with CBA/J × BALB/c matings as a mouse model of normal pregnancy (Figure 4a).
To investigate the influence of RANKL/RANK signaling on outcome of pregnancy in vivo, we evaluated embryoabsorbing level between RANKL − / − and WT pregnant mice. The embryo-absorbing site could be macroscopically distinguished as hemorrhagic spots and necrosis at late gestation (gestational day 14, Figure 4g). The RANKL − / − mice were more susceptible to fetal loss than WT mice (Figures 4g and  h). These findings provide evidences of a key role of RANKL in the regulation of dMφ differentiation and function, promoting maternal-fetal tolerance to support normal pregnancy. Abnormal suppression of RANKL expression contributes to uMφ dysfunction and fetal loss in vivo.
Adoptive transfer of RANK + Mφ relieves murine embryo absorption induced by Mφ depletion. The depletion of Mφ in pregnant mice using Clodronate Liposomes ( Supplementary Figures 5a and b) led to a significant decrease in RANKL in uterine DSCs (uDSCs) and placental trophoblasts (pTros) (Supplementary Figure 5c), suggesting that dMφs are involved in the maintenance of high levels of RANKL at the maternal-fetal interface. To provide insight into the role of RANKL-instructed dMφ in maternal-fetal immune regulation and pregnancy outcome in vivo, we evaluated the effect of dMφ depletion and adoptive transfer of RANK + Mφ on these processes. Next, we isolated RANK + and RANK − Mφs from mouse spleen and observed that these RANK + Mφs, like human RANK + pMo, had high levels of M1 and M2 phenotype molecules (Supplementary Figure 6a). To investigate the process of RANK + Mφ differentiation in the uterus in vivo, we labeled these RANK +   transferred group, PKH-67-RANK + Mφ recruited to uterus presented high levels of CD206 and CD209, and low levels of CD86 and similar levels of CD80 (Figure 5b). In addition, the transfer of PKH-67-RANK + Mφ led to a Th2 bias in mouse uterus (Figure 5c), an increase in Akt and STAT6 activation (Figure 5d), and a high level of IRF4 (Figure 5e) in uMφ.
Subsequently, we observed that Mφ depletion caused a significant increase in fetal loss (Figures 5f and g). To further identify the role of RANKL-instructed Mφ in ameliorating fetal loss, we transferred RANK + or RANK − Mφ to Mφ-deleted pregnant mice and found that adoptive transfer of RANK + Mφ could significantly relieve the murine embryo absorption induced by Mφ depletion (Figures 5h and i).
The suppression of RANKL/RANK expression and dMφ dysfunction in human miscarriage. The imbalance of maternal-fetal immunoregulation has been previously reported in pregnancy complications 3-5 such as miscarriage, preeclampsia and IUGR. Therefore, we evaluated RANKL/ RANK expression in the tissue from patients with miscarriage during the first trimester. In comparison with normal pregnancy, we observed a decrease in RANKL expression in trophoblasts and DSCs (Figure 6a), as well as reduced RANK expression on CD14 + dMφ (Figures 6b and c) in miscarriage, accompanied by a decreased frequency of dMφ with an M2 phenotype and an increase in the M1 phenotype (Figures 6d and e). Taken together, the suppression of RANKL/RANK signaling may result in dMφ dysfunction and further trigger miscarriage during the first trimester.

Discussion
Collectively, we have demonstrated that RANKL derived from embryonic trophoblasts and maternal DSCs drives dMφ polarization toward an M2 phenotype by activating Akt/STAT6 signaling and enhancing the transcription of IRF4 and Jmjd3; finally, it contributes to the formation and maintenance of maternal-fetal tolerance by further inducing Th2 bias (Figure 7). This maternal-fetal dialog mediated by RANKL guarantees a smooth gestation by creating a tolerant microenvironment. The process of maternal-fetal tolerance formation is not only derived from a maternal behavior for immune adaptation to pregnancy but also, most importantly, the fetus can instruct the mother's immune system to adapt to it.
In comparison with the peripheral, the M2 phenotype advantage of RANK + dMφ becomes more prominent in the decidua. The levels of M1 phenotype markers in RANK + dMφ are also higher than those in RANK − dMφ, but still very low, similar to the level of RANK − pMo. Similarly, it has been reported that both CD209 + and CD209 − dMφ stimulate the release of proinflammatory cytokines such as IL-6 and TNF-α after LPS stimulation in vitro. 12 These data emphasize the complexity of dMφ biology. During normal pregnancy, the M2 advantage of dMφ at the maternal-fetal interface is relative and mainly depends on the local microenvironment. This advantage may be disrupted by intrauterine infection and lead to an M1 advantage to limit infection. The expression of proinflammatory molecules in dMφ may align more with the theory that immune activation is required to facilitate trophoblast invasion and implantation, as well as the establishment of fetal-maternal tolerance during the first trimester.
The human maternal-fetal interface is characterized by intimate contact between the maternal decidua and extravillous cytotrophoblast cells that invade the decidua. Trophoblasts can influence the maternal immune system during pregnancy by expressing soluble and cell surface molecules, such as HLA-G, 42,43 IDO 44 and anti-inflammatory cytokines. [45][46][47] These molecules limit the proliferation and activation of T cells, antigenpresenting cells and NK cells in decidua. In our present study, we found that the crosstalk between embryonic trophoblasts and maternal DSCs and dMφ contributes to the accumulation of RANKL expression at the maternal-fetal interface. RANKL expressed by trophoblasts and DSCs induces M2 differentiation of dMφ and further drives the Th2 bias, suggesting that RANKL/ RANK signaling has a critical role in dMφ differentiation and maternal-fetal tolerance. It is noteworthy that trophoblasts upregulate RANK expression specifically on dMφ. These findings further highlight the core role of trophoblasts in dMφ differentiation regulation. Rather than traditional NF-κB signaling under the RANKL/RANK axis, we found that Akt/STAT6-Jmjd3/ IRF4 signaling is required for M2 differentiation of dMφ induced by RANKL at the maternal-fetal interface in vitro and in vivo. Further studies should clarify the molecular mechanisms by which RANKL specifically activates Akt/STAT6 signaling in dMφ.
In human pregnancy, embryo implantation in the receptive endometrium triggers a series of responses collectively called decidualization. During decidualization, endometrial stromal cells (ESCs) undergo steroid hormone-dependent proliferation and differentiation into decidual cells. 48 Interestingly, pregnancy-associated hormones (PAHs, such as estrogen and prolactin) upregulate RANKL and RANK levels, downregulate OPG expression, and further affect osteoclastogenesis at distinct stages of development. 49,50 RANKL/RANK system also controls the incidence and onset of progestindriven breast cancer and physiological thermoregulation in females under the control of sex hormones. 51,52 Therefore, high levels of PAH during pregnancy may also be one of the important factors leading to such high levels of RANKL/RANK at the maternal-fetal interface.
The decidua has been considered a specialized mucosal wall of the uterus. Research examining other mucosal tissues (skin and intestine) shows that epidermal and Peyer's patchderived DCs stimulated with RANKL induce immunosuppressive activity by modulating surface barrier DCs and increasing the expansion and function of Treg cells. [26][27][28] Our results partially echo the immunosuppressive effect of RANKL in the mucosa. This function in the decidua is independent of the regulation of Treg differentiation.
In comparison with normal pregnancy, we observed that RANKL in trophoblasts and DSCs and RANK on dMφ in patients with miscarriage were greatly decreased. Moreover, the dMφ phenotype during human and mouse pregnancy wastage shows an M1 predominance. RANKL − / − mice presented uMφ dysfunction and increased fetal loss. This deregulation of uMφ supports an inflammatory environment that further triggers abortive processes. 53 Therefore, our study reveals a novel pathogenic role of abnormal RANKL/RANK signaling at the maternal-fetal interface during SA in humans and mice. Trials conducted in vivo also showed that RANKL − / − mice had no significant influence on the total number of embryo implantations (data not shown). However, our unpublished data show that either endogenous or exogenous RANKL directly stimulates the proliferation and enhances the invasiveness of human trophoblasts, partially echoing its role in tumor cells. 19 We propose that the lack of RANKL in vivo may result in a decrease in trophoblast proliferation and invasion, but to a certain extent, it will also  create a proinflammatory microenvironment. This inflammatory pattern during the initial stage of pregnancy may be conducive to the invasion and implantation of trophoblasts. Therefore, the overall effect of absent RANKL signaling in vivo may not affect embryo implantation. However, with advancing pregnancy, abnormally low levels of RANKL will result in miscarriage via the M1 dMφ-triggered disorder of maternal-fetal immune tolerance. Therefore, further research is needed to elucidate the cause of low RANKL/RANK expression in miscarriage patients.
In conclusion, as shown in Figure 7, accompanied by the implantation of blastocyst during normal pregnancy, PAH trigger ESCs to differentiate into DSCs and further induce high levels of RANKL expression and CCL2 release. The latter allows the recruitment of peripheral monocytes into decidua. Embryonic trophoblasts that are deeply implanted in decidua are in close contact with DSCs and DLCs. The dialog of these cells not only further increases RANKL expression in trophoblasts and DSCs but also enhances the sensitivity of RANK on dMφ to RANKL by upregulating RANK expression. Subsequently, the RANKL-RANK interaction drives dMφ to M2 differentiation. After being instructed, these dMφ will create and maintain a maternal-fetal tolerant microenvironment.
Once the maternal-fetal interface presents abnormal low level of RANKL, dysfunction of dMφ and then miscarriage will occur. Therefore, our study provides a potential target molecule, RANKL, for the identification of new strategies to prevent and treat pregnancy complications.

Materials and Methods
Patient and sample collection. All procedures involving participants in this study were approved by the Human Research Ethics Committee of Obstetrics and Gynecology Hospital, Fudan University (Shanghai, China), and all subjects completed an informed consent to collect tissue samples. First-trimester human peripheral blood was obtained from 41 women with clinically normal pregnancies (age: 27.45 ± 7.21 years; gestational age at sampling: 48.35 ± 7.6 days (mean ± S. D.)]), which were terminated for nonmedical reasons. Human villi tissues were obtained from 172 women with clinically normal pregnancies (age: 29.88 ± 6.91 years; gestational age at sampling: 49.17 ± 9.34 days [mean ± S.D.]) (Termination for nonmedical reasons), or from 12 women with SA (age: 31.09 ± 4.28 years; gestational age at sampling: 47.95 ± 10.1 days (mean ± S.D.)). Decidual tissues were obtained from 135 women with clinically normal pregnancies (age: 27.19 ± 5.61 years; gestational age at sampling: 51.09 ± 8.72 days (mean ± S.D.)) (termination for nonmedical reasons) and 23 women with spontaneous miscarriage (age: 31.54 ± 5.71 years; gestational age at sampling: 53.06 ± 5.8 days (mean ± S.D.)). All pregnant women were confirmed by ultrasound and blood tests, and the women with spontaneous miscarriage because of endocrine, anatomical, and genetic abnormalities, as well as infection were excluded. Mice. RANKL heterozygote mice were obtained from the Jackson Laboratories (Sacramento, CA, USA) and subsequently maintained in the Laboratory Animal Facility of Fudan University (Shanghai, China). RANKL knockout (RANKL − / − ) mice and wildtype littermates (WT) were obtained by mating of male and female RANKL heterozygote mice. A group of adult female C57BL/6 mice were purchased from the Laboratory Animal Facility of Fudan University and used for this study. They were usually maintained for 2 weeks in the animal facility before use. The Animal Care and Use Committee of Obstetrics and Gynecology Hospital, Fudan University approved all animal protocols.
Isolation and culture of human trophoblasts, DSCs and DLCs. The villi and decidua tissues from the first-trimester pregnancy were put immediately into ice-cold Dulbecco's modified Eagle's medium (DMEM high D-glucose; Gibco, Grand Island, NY, USA), transported to the laboratory within 30 min after surgery Figure 7 RANKL-mediated harmonious dialog between the fetus and mother guarantees a smooth gestation by inducing decidual M2 macrophage polarization. Together with blastocyst implantation during normal pregnancy, PAHs trigger ESCs to differentiate into DSCs, further inducing high levels of RANKL expression and CCL2 release. The latter recruits more peripheral monocytes to the decidua. Embryonic trophoblasts that are deeply implanted in decidua can closely contact DSCs and DLCs. This dialog not only further increases RANKL expression in trophoblasts and DSCs, but it also enhances the sensitivity of RANK to RANKL by upregulating RANK expression on Mφ. Subsequently, the RANKL-RANK interaction drives Mφ to M2 differentiation (low expression of CD80 and CD86, high secretion of IL-10, and low level of IL-12 and IL-23) by activating Akt/ STAT6-Jmjd3/IRF4 signaling. After education, these Mφs will create and maintain a maternal immune tolerance microenvironment (increase in Th2 and decrease in Th1). After the development of abnormally low RANKL expression at the maternal-fetal interface, the dysfunction of Mφ, the imbalance of maternal-fetal immune regulation and then abortion will occur. Tro: trophoblasts and washed in Hank's balanced salt solution for isolation of human trophoblasts from villi, and DSCs and DLCs from deciduas according to a previously described method. 46 This method supplies a 95% purity of vimentin − cytokeratin (CK)7 + trophoblast cells and greater than 98% vimentin + CK7 − DSCs and CD45 + DLCs, which was confirmed by FCM analysis.
Detection of RANK expression on peripheral blood mononuclear cells (PBMCs) and DLCs. The PBMCs were isolated from the peripheral blood of pregnant women who were terminated for nonmedical reasons. Next, FCM was performed to analyze the expression of RANK on pMo and dMφ by labeling anti-human CD14, RANK and CD45 antibodies. In addition, we further evaluated the phenotype of RANK + and RANK − pMos and dMφ purified from PBMC and DLC (n = 24) by FCM.
Purification of dMφ and decidual naive T cells. We isolated dMφ and decidual naive T cells from DLCs by MACS (Miltenyi Biotec, Bergisch Gladbach, Germany) and performed FCM analysis with standard protocols.
The intracellular phosphorylation level of Akt and STAT6. The intracellular phosphorylation level of Akt and STAT6 in dMφ after 24-h culture with JEG-3 and DSCs was analyzed using BD Phosflow antibodies, according to standard protocols.
The transcription of Jmjd3, IRF4 and IRF5. After 24-h co-culture, the transcription level of Jmjd3, IRF4 and IRF5 in dMφ was analyzed by real-time PCR, according to standard protocols. The primer sequences were designed and synthesized by TaKaRa Biotechnology Co., Ltd (Tokyo, Japan) as described in Supplementary Table 1.
Co-culture of dMφ and decidual naive T cells. After 48 h of culture with trophoblasts/JEG-3 cells and DSCs, CD14 + dMφ were collected and washed three times with phosphate-buffered saline to remove excess cytokines. The remaining cells were co-cultured with autologous decidual naive T cells at a 1 : 1 ratio. The decidual naive T cells were pre-activated with 5 μg/ml anti-CD3 (OKT3; 16-0037; ebioscience), 1 μg/ml anti-CD28 (CD28.2; 16-0289; ebioscience), and 20 U/ml rhIL-2 (202-IL-010, R&D Systems) for 2 days, and collected, washed, and then incubated with culture medium only. After 5 days of co-culture, the naive T cells were reactivated with anti-CD3 and anti-CD28 for 24 h before the supernatants were collected. The expression of GATA-3 and T-bet in decidual naive T cells, and the secretion level of IL-4, IL-10, and TNF-α and IFN-γ were analyzed by FCM and ELISA (R&D Systems), respectively.
Animals and experimental design. We divided female C57BL/6 mice (age: 8 weeks old, weight: 20-23 g) into two groups by using a random number table by body weight, age and family: the adoptive transfer of RANK + Mφ group and the adoptive transfer of RANK − Mφ group. This was an unblinded trial.
The differentiation of uMφ and uCD4 + T cells, the activation of Akt and STAT6, and the level of IRF4 in uMφ were analyzed by FCM, and the level of fetal loss was counted at day 14 of gestation in WT and RANKL − / − pregnant mice. The day of appearance of a copulatory plug was arbitrarily designated as day 0 of gestation. The embryo absorption rate and implantation number were counted on day 14 of gestation. The percentage of fetal loss (the embryo absorption rate) was calculated as follows: percentage fetal loss = R/(R+V) × 100, where R represents the number of hemorrhagic implantatio (sites of fetal loss) and V stands for the number of viable, surviving fetuses.
In addition, mouse uterine tissues were removed, minced on ice and digested with an enzyme mix of Liberase and Dispase (Invitrogen, Carlsbad, CA, USA). uMφ were isolated from mouse uterus by MACS (Miltenyi Biotec) to analyze the transcription of Jmid3 and IRF4 (Takara Bio Inc.,Tokyo, Japan) by RT-PCR at day 10 of gestation. The primer sequences were described in Supplementary Table 1.
For Mφ depletion and Mφ adoptive transfer in pregnant C57BL/6 mice, and Clodronate Liposomes were injected intraperitoneally at day 1 (200 μl) and day 4 (100 μl) of gestation, and then the expression of RANKL in Vimentin + uSC and CK7 + pTros were analyzed at day 7 by FCM. RANK + and RANK − Mφs were isolated from mouse spleen and labeled with PKH-67, and then they were transferred to Mφdepleted pregnant mice at day 5 of gestation. Subsequently, the differentiation of uMφ and uCD4 + T cells, the activation of Akt and STAT6, and the IRF4 level in uMφ were analyzed by FCM at day 10 of gestation, and the level of fetal loss was counted at day 14 of gestation in the RANK + transfer group and RANK − transfer group.
To investigate the role of STAT6 and Jmjd3 signals in uMφ and uCD4 + T cells, the pregnant C57BL/6 mice were intraperitoneally injected with STAT6i AS1517499 (200 μl at the concentration of 2 mg/kg) or JMJD3i GSK J4 HCl (200 μl at a concentration of 4.48 mg/kg) in (n = 5 mice per group) at day 4, and then the differentiation of uMφ and u CD4 + T cells and the IRF4 level in uMφ cells at day 10 were determined by FCM.
Statistics. The continuous variable is shown as the mean ± S.E.M. Continuous variables were analyzed by Student's t-test in case of two groups and by one-way ANOVA using Tukey' s post-hoc test in multiple groups. The embryo resorbing rate was analyzed using an adjusted t-test. All analyses were conducted with SPSS 16.0 Statistical Package for the Social Sciences software (IBM SPSS, Armonk, NY, USA). Po0.05 was considered statistically significant.