C57BL/6 and BALB/c mice were purchased from the Experimental Animal Center of Zhongshan University (Guangzhou, China). Transgenic Foxp3^DTR mice, Act-mOVA mice and OT-II mice on a C57BL/6 background were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Foxp3^GFP mice on a C57BL/6 background were kindly provided by Professor Zhexiong Lian (South China University of Technology, Guangzhou, China). Mice were bred in the Laboratory Animal Center of Shenzhen People’s Hospital (Shenzhen, China) in pathogen-free conditions. Syngeneic C57BL/6 × C57BL/6 and allogeneic C57BL/6 × BALB/c mating combinations were established. Each female mouse at 10-12 weeks of age was cocaged with one male. The point at which a vaginal plug was detected was designated as embryonic day 0.5 (E0.5). Mice were sacrificed at different gestational age to investigate proportions of Treg cells in the thymi, peripheral blood, spleens, LNs or deciduae. Non-pregnant mice were set up as negative controls.

Foxp3^GFP females were mated to OT-II males to generate OT-IIFoxp3^GFP mice. Spleen CD4⁺CD62L⁺GFP^- naïve T cells were isolated from female Foxp3^GFP or OT-IIFoxp3^GFP mice. Female C57BL/6 mice were mated to male Act-mOVA mice to generate fetal antigen mOVA that can be recognized by CD4⁺ cells from OT-II mice (26). Spleen CD4⁺CD62L⁺GFP^- naïve T cells (2 × 10⁶ cells) from female Foxp3^GFP mice or OT-IIFoxp3^GFP mice were injected into the tail veins of pregnant C57BL/6 mice mated to Act-mOVA mice on E9.5. Pregnant mice were sacrificed on E13.5, and CD4⁺GFP⁺ Treg cells in the peripheral blood, spleens and deciduae were analyzed using flow cytometry.

For Treg cell depletion, syngeneic pregnant Foxp3^DTR mice were injected i.p. with DT (Sigma, Saint Louis, MO, USA) at a dosage of 60 µg/kg of body weight on E9.5, E10.5 and E11.5. Control mice received phosphatebuffered saline (PBS). For Treg cell reconstitution, 2 × 10⁵ purified Treg cells from the thymi and spleens of C57BL/6 mice, or in vitro cultured iTreg cells were injected into the tail veins of Foxp3^DTR mice 2 h before DT injection. Pregnant mice were sacrificed on E13.5. At this stage of gestation, feto-placental units undergoing resorption can be clearly distinguished from their viable counterparts on the basis of size and the presence of extensive tissue wasting and hemorrhage. The number of total embryos, resorbed and viable embryos per mouse was counted. The embryo resorption rate (ERR) was calculated as: ERR (%) = number of resorbed embryos/number of total embryos × 100. Treg cells in the thymi, peripheral blood, spleens and deciduae were analyzed using flow cytometry. Deciduae were collected for calculating Th1 and Th17 cell percentages using flow cytometry.

Thymi, spleens or LNs including cervical, axillary, mesenteric and inguinal LNs were collected and mashed through a 40 μm cell strainer (Thermo Fisher Scientific, Waltham, MA, USA) to obtain single-cell suspensions. Mononuclear cells were then enriched by lysing red cells using a red cell lysis buffer (Biolegend, San Diego, CA, USA).

Peripheral blood samples were collected from the vena orbitalis, heparinized, and purified by centrifugation on a Ficoll-Hypaque Premium (GE Healthcare, Pittsburgh, PA, USA).

Uterine tissues were collected, washed and cut into small pieces and digested with 1 mg/ml Dispase II (Roche, Basel, Switzerland) and 0.1 mg/ml DNase I (Roche) at 37°C for about 20 min in a shaking water bath. When single or clumps of cells were observed under the microscope, the released cells were separated from undigested tissue pieces by filtering through a 40 µm cell strainer. Mononuclear cells were purified over the Ficoll-Paque Premium by centrifugation at 400 g for 30 min at 20°C.

Naïve T cells were isolated from the spleens of Foxp3^GFP mice using a mouse CD4⁺CD62L⁺ naïve T cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instructions. In brief, spleen mononuclear cells were labeled with a CD4⁺ T Cell Biotin-Antibody Cocktail, washed and incubated with anti-biotin microbeads. Non-CD4⁺ T cells were then magnetically depleted using the midiMACS separator. CD4⁺ T cells were collected and incubated with anti-CD62L microbeads. CD4⁺CD62L⁺ naïve T cells were sorted using the miniMACS separator. The purity of isolated naïve T cells were detected using flow cytometry.

Thymus and spleen mononuclear cells of Foxp3^GFP mice on E12.5 were isolated. CD4⁺ cells were enriched using a mouse CD4⁺ T cell isolation kit (Miltenyi Biotec) according to the manufacturer’s instructions, and CD4⁺GFP⁺ Treg cells were sorted by a FACSAria cell sorter (Becton Dickinson, Franklin Lakes, NJ, USA). In brief, the single-cell suspensions were labeled with a CD4⁺ T Cell Biotin-Antibody Cocktail, washed and incubated with anti-biotin microbeads. Non-CD4⁺ T cells were then magnetically depleted using the midiMACS separator. Subsequently, enriched CD4⁺ T cells were collected, and CD4⁺GFP⁺ Treg cells were sorted using the FACS-Aria cell sorter. The purity of isolated CD4⁺GFP⁺ Treg cells were analyzed using flow cytometry.

Spleen CD4⁺CD62L⁺GFP^- naïve T cells of Foxp3^GFP mice were isolated. The cells (1 × 10⁶/ml) were then cultured in the presence of 2.5 μg/ml anti-CD3 Ab (Biolegend), 1 μg/ml anti-CD28 Ab (Biolegend), 100 U/ml IL-2 (R&D Systems, Minneapolis, MN, USA), and 10 ng/ml TGF-β (R&D Systems) for 96 h in vitro. The ratio of CD4⁺GFP⁺ iTreg cells was detected using flow cytometry, and the iTreg cells were sorted for later experiments.

For Treg cell analysis, cells were incubated with fluorescence-conjugated Abs against CD45, CD3, CD4 and CD8 for 30 min at 4°C, and then washed and fixed in a fixation buffer. Cells were re-suspended in a permeabilization wash buffer and incubated with a fluorescence-conjugated Ab against Foxp3. Isotype controls were established using matched fluorescence-labeled isotype control Abs to account for nonspecific staining. Immunostained cells were analyzed on a FACS Fortessa flow cytometer (BD Biosciences, San Jose, CA, USA) using the FlowJo software (Ashland, OR, USA). The percentages of Foxp3⁺ cells in the CD45⁺CD3⁺CD4⁺CD8^- cell population were analyzed.

For Foxp3^GFP mice, cells were incubated with aforementioned fluorescently labeled Abs against CD45, CD3, CD4 and CD8 for 30 min at 4°C. Cells were washed and fixed in a fixation buffer. The percentage of GFP⁺ cells in the CD4⁺GFP⁺ cell population was analyzed. In selected experiments, cells were incubated with fluorescently labeled Abs against CD45, CD3, CD4 and CD8, CCR2 and CCR6. The expression of CCR2 and CCR6 on CD45⁺CD3⁺CD4⁺CD8^-GFP⁺ cells was detected. For naïve T cell analysis, cells were mixed with fluorescently labeled Abs against CD45, CD3, CD4 and CD8, and CD62L for 30 min at 4°C. The percentages of CD45⁺CD3⁺CD4⁺CD62L⁺ cells in CD45⁺CD3⁺CD4⁺CD8^- cells were analyzed.

For analysing intracellular factors, cells were incubated with fluoresence-conjugated Abs against CD45, CD3, CD4 and CD8 for 30 min at 4°C. Cells were washed, fixed, re-suspended in a permeabilization wash buffer, and incubated with fluorescence-conjugated Abs against IFN-γ and IL-17. Percentages of IFN-γ⁺ and IL-17⁺ cells in the CD45⁺CD3⁺CD4⁺CD8^- cell population were analyzed. Fluorescence-conjugated Abs, matched fluorescence-labeled isotype control Abs, fixation buffer, and permeabilization wash buffer were all purchased from BioLegend. Flow cytometry Abs were summarized in Table 1 .

The uteri of the non-pregnant mice, syngeneic and allogeneic pregnant mice on E13.5 were quickly removed and placed in liquid nitrogen and transferred to the -80°C refrigerator for later use. Total RNA was extracted from the uteri using Trizol extraction, and the content and purity of total RNA were determined by spectrophotometry. cDNA was obtained according to the instructions of the PrimeScript™ RT Reagent Kit with gDNA Eraser (Perfect Real Time) (TaKaRa, Beijing, China). The mRNA levels of ccl2, ccl7, ccl12,ccl17, ccl19, ccl20, ccl21, cxcl9, cxcl10 and cxcl11 in the deciduae were detected using PowerUp™ SYBR™ Green Master Mix (Applied Biosystems, Foster City, CA, US) and QuantStudio fluorescent quantitative RT-PCR system (Thermo Fisher Scientific) with β-actin as the internal reference. The method of 2^⁃△△Ct relative quantitation was used to compare the difference of target mRNA expression between groups. The primers of the chemokines were bought from Shanghai Shenggong Biological Company (Shanghai, China) and shown in the Table 2 .

The uteri of the non-pregnant mice, syngeneic and allogeneic pregnant mice on E13.5 were collected and stored at -80°C. Tissue protein was extracted, and the bicinchoninic acid (BCA) Protein Quantification Kit (Thermo Fisher Scientific) was used to detect protein expression levels. The prestain protein marker and protein samples to be tested were loaded in the glue and electrophoresed, and then transferred to the PVDF membrane, and sealed by 5% skimmed milk. The membrane was then incubated with anti-mouse CCL2, CCL7, CCL12 and CCL20 Abs, washed and incubated with horseradish peroxidase (HRP)-coupled goat-anti-mouse secondary Ab IgG (H+L). The membrane was then washed, and hypersensitive chemiluminescence substrate (Thermo Fisher Scientific) was added for color exposure. Image J software (NationalnInstitutes of Health, Bethesda, MD, USA) was used to detect the gray value of the strip. The expression of target proteins CCL2, CCL7, CCL12 and CCL20 were analyzed semi-quantitatively using GAPDH as the internal reference protein. Western blot Abs were summarized in Table 1 .

Isolated thymus CD4⁺GFP⁺ Treg cells (1 × 10⁵) were seeded in the upper chamber of a 24-well plate (5-μm pore; Sigma). A volume of 600 μl recombinated chemokines including 20 ng/ml CCL2, 500 ng/ml CCL7, 50 ng/ml CCL12 and 10 ng/ml CCL20 were loaded in the lower chamber, respectively. Basal culture medium in the lower chamber served as a negative control. After 24 h, cells in the lower chamber were collected. The numbers of CD4⁺GFP⁺ cells were calculated using flow cytometry. Migration index was expressed as fold change of the numbers of CD4⁺GFP⁺ cells relative to those isolated from basal culture medium controls. The assay was carried out in triplicate and repeated three times independently.

All statistical analyses were performed using SPSS 19.0 software (Chicago, IL, USA). Data were analyzed by one-way ANOVA with Bonferroni test among three or more groups, or independent Student’s t-test between two groups. Results were given as mean ± SD. A p value of < 0.05 was considered statistically significant between analyzed groups.

The percentages of thymus, spleen and LN CD4⁺Foxp3⁺ Treg cells in CD4⁺ T cells neither differ significantly among pregnant and non-pregnant groups, nor differ significantly between syngeneic and allogeneic mice from E4.5 to E19.5 ( Figures 1A, C, D ). However, as shown in Figure 1B , the proportions of blood CD4⁺Foxp3⁺ Treg cells in CD4⁺ T cells significantly increased in both syngeneic and allogeneic mice on E13.5 as compared with non-pregnant mice (p < 0.01 for both comparisons), while proportions of blood Treg cells at different gestational ages did not differ between syngeneic and heterogeneic pregnant mice.

As shown in Figure 2 , the percentages of decidua CD4⁺Foxp3⁺ Treg cells in both syngeneic and allogeneic mice significantly increased as the pregnancy progressed, peaked on E13.5, and then decreased till E19.5. The percentages of decidua Treg cells in allogeneic mice on E10.5, E13.5 and E16.5 were significantly higher than those in syngeneic mice (p < 0.01, respectively).

As shown in Figure 3A , compared with non-pregnant group, levels of CCL2, CCL7, CCL12 and CCL20 mRNA in the deciduae significantly increased in both syngeneic mice (p < 0.01, p < 0.05, p < 0.05 and p < 0.01) and in allogeneic mice (p < 0.01, respectively). In addition, levels of CCL2, CCL7, CCL12 and CCL20 mRNA in the deciduae of allogeneic mice were significantly up-regulated compared with those of syngeneic mice (p < 0.05, p < 0.05, p < 0.05 and p < 0.01). However, levels of CCL17, CCL19, CCL21, CXCL9, CXCL10 and CXCL11 mRNA in the deciduae of pregnant mice did not differ significantly from non-pregnant mice (data not shown).

Since mRNA levels of CCL2, CCL7, CCL12 and CCL20 significantly increased in the deciduae of pregnant mice compared with non-pregnant mice, we analyzed protein levels of those chemokines in the decidua. As shown in Figures 3B, C , protein levels of CCL2, CCL12 and CCL20 in the deciduae of syngeneic mice (p < 0.05, p < 0.01 and p < 0.05) and allogeneic mice (p < 0.01, p < 0.01 and p < 0.05) were significantly higher than those of non-pregnant mice. In addition, the level of CCL12 in the deciduae of allogeneic mice significantly increased compared with that of syngeneic mice (p < 0.01). However, we did not find significant differences in CCL7 levels among the three groups ( Figures 3B, C ).

In the case the mRNA levels of CCL2, CCL7, CCL12 and CCL20 in the deciduae significantly increased, we wondered whether those chemokines can attract thymus Treg cells into the deciduae by chemotactic experiments in vitro. As shown in Figure 3D , the migration indexes of CCL2, CCL7, CCL12 and CCL20 groups were all significantly increased compared with the control group without chemokines (p < 0.05, p < 0.01 and p < 0.01).

Since mRNA and protein levels of CCL2, CCL12 and CCL20 in the decidua significantly increased in pregnant mice, we wanted to know levels of the receptors CCR2 and CCR6 of those chemokines on thymus and blood Treg cells. As shown in Figures 4A-C , compared with non-pregnant mice, while the expression of CCR2 on thymus Treg cells significantly up-regulated in syngeneic and allogeneic mice (p < 0.01 for both comparisons), the expression of CCR6 on thymus Treg cells significantly up-regulated in allogeneic mice (p < 0.01). In addition, the expression of CCR2 and CCR6 on thymus Treg cells significantly increased in allogeneic mice as compared to syngeneic mice (p < 0.01 for both comparisons). Similarly, as shown in Figures 4D-F , compared with non-pregnant mice, while the expression of CCR2 on peripheral blood Treg cells significantly up-regulated in syngeneic and allogeneic mice (p < 0.01 for both comparisons), the expression of CCR6 on peripheral blood Treg cells significantly up-regulated in allogeneic mice (p < 0.01). Besides, the expression of CCR2 (p < 0.01) and CCR6 (p < 0.05) on peripheral blood Treg cells significantly increased in allogeneic mice as compared to syngeneic mice.

In order to investigate the role of fetal antigen-specific pTreg cells in maternal-fetal tolerance, we first investigated whether pTreg could be induced from fetal antigen-specific naïve T cells using transgenic OT-II mice and fetal surrogate antigen mOVA. We demonstrated that pTreg cells were induced in vivo from mOVA-specific CD4⁺CD62L⁺GFP^- naïve T cells, since transfusion of CD4⁺CD62L⁺GFP^- naïve T cells that from OT-IIFoxp3^GFP mice induced more CD4⁺GFP⁺ pTreg cells in the peripheral blood, spleens and deciduae of recipient mice than transfusion of naïve T cells from Foxp3^GFP mice (p < 0.01, respectively) ( Figure 5 ).

In order to investigate the role of Treg cells in pregnancy, pregnant Foxp3^DTR mice were i.p. injected with DT to deplete Treg cells to set a murine inflammatory abortion model. Our results showed that 3 consecutive injections of DT at the dosage of 60 μg/kg weight almost completely deplete body Treg cells as demonstrated by nearly no Treg cells in the thymi, blood, LNs, spleens and deciduae without obvious maternal morbidity or mortality. Treatment with DT did not change total fetuses, but markedly increased the ERR in Foxp3^DTR mice (p < 0.01) ( Figure 6 ), suggesting an important role of Treg cells in maternal-fetal tolerance.

Meanwhile, we investigated whether thymus- or extrathymus-derived Treg cells play a role in pregnancy tolerance by adoptive transfer of Treg cells from different sources into DT-injected Foxp3^DTR mice. As shown in Figure 6A , total fetuses did not differ among groups. Adoptive transfer of thymus (p < 0.01) or spleen CD4⁺GFP⁺ Treg cells (p < 0.01), or in vitro generated iTreg cells (p < 0.05) into DT-treated Foxp3^DTR mice significantly decreased ERRs compared with DT-treated Foxp3^DTR mice, which did not differ from those of vehicle-treated control mice ( Figures 6B, C ).

Since Th1 and Th17 cells play an important role in pregnancy loss (27), we wondered whether the immune tolerant effects by adoptive transfer of Treg cells were through inhibiting Th1 or Th17 cells. We found that treatment with DT significantly increased the percentages of peripheral blood IFN-γ⁺ cells and IL-17⁺ cells in CD45⁺CD3⁺CD4⁺ cells compared with controls (p < 0.01 for both comparisons, Figures 7A, B ), indicating that deficiency of Treg cells increases systemic Th1 and Th17 cells. Compared with DT-treated Foxp3^DTR mice, adoptive transfer of thymus or spleen CD4⁺GFP⁺ Treg cells significantly decreased the percentages of peripheral blood IFN-γ⁺ cells (p < 0.01 for both comparisons), which did not differ from those of vehicle-treated control mice ( Figure 7A ). In addition, compared with DT-treated Foxp3^DTR mice, adoptive transfer of in vitro generated iTreg cells significantly decreased the percentages of peripheral blood IFN-γ⁺ cells (p < 0.01), which were still higher than those of vehicle-treated control mice (p < 0.01), and DT-treated Foxp3^DTR mice with adoptive transfer of thymus Treg cells (p < 0.01) or spleen Treg cells (p < 0.05, Figure 7A ).

Similarly, treatment with DT significantly increased the percentages of decidua IFN-γ⁺ cells (p < 0.01) and IL-17⁺ cells (p < 0.05) in CD45⁺CD3⁺CD4⁺ cells compared with controls ( Figures 7C, D ), indicating that deficiency of Treg cells also increases maternal-fetal interface Th1 and Th17 cells. Compared with DT-treated Foxp3^DTR mice, adoptive transfer of thymus and spleen Treg cells or in vitro generated iTreg cells significantly decreased the percentages of decidua IFN-γ⁺ cells (p < 0.01, respectively), which did not differ from those of vehicle-treated control mice ( Figure 7C ).Compared with DT-treated Foxp3^DTR mice, adoptive transfer of thymus and spleen Treg cells or in vitro generated iTreg cells decreased the percentages of peripheral blood and decidua IL-17⁺ cells, but the differences did not reach statistical significance ( Figures 7B, D ).