SCIENCE SIGNALING | RESEARCH ARTICLE IMMUNOLOGY

Tim-3 signaling in peripheral NK cells promotes maternal-fetal immune tolerance and alleviates pregnancy loss Yanhong Li,1,2,3* Jiayuan Zhang,4* Di Zhang,1,2,3 Xiaowu Hong,5 Yu Tao,1 Songcun Wang,1,2,3 Yuanyuan Xu,1 Hailan Piao,1 Weijie Yin,4 Min Yu,1,6 Yin Zhang,1,7 Qiang Fu,1 Dajin Li,1,2,3 Xing Chang,4† Meirong Du1,2,3†

Copyright © 2017 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works

INTRODUCTION

How the semiallogeneic fetus can develop without being attacked or rejected by the maternal immune system has been a fascinating question in reproductive biology and immunology. Complex immunoregulation is required for maternal immune tolerance toward the semiallogeneic fetus and the maintenance of pregnancy (1, 2). Once maternal-fetal immune tolerance is disturbed, various complications of pregnancy occur (3, 4). Spontaneous abortion is the most common complication of gestation, occurring in about 15% of human pregnancies (5). When a spontaneous abortion occurs, the risk of subsequent miscarriage is markedly increased. Recurrent miscarriage (RM) is defined as two or more consecutive miscarriages. More than 50% of these events remain unexplained after standard investigations (6). There is also no reliable predictor or effective prevention strategy for RM. A systematic review and meta-analysis demonstrated that the most commonly used biomarkers [serum human chorionic gonadotropin (hCG) and progesterone] are 1 Laboratory for Reproductive Immunology, Hospital of Obstetrics and Gynecology, Fudan University Shanghai Medical College, Shanghai 200011, China. 2Key Laboratory of Reproduction Regulation of National Population and Family Planning Commission, Shanghai Institute of Planned Parenthood Research, Institute of Reproduction and Development, Fudan University, Shanghai 200032, China. 3Shanghai Key Laboratory of Female Reproductive Endocrine Related Diseases, Shanghai 200011, China. 4Key Laboratory of Stem Cell Biology, Institute of Health Sciences, Shanghai Jiao Tong University School of Medicine and Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Shanghai 200031, China. 5Department of Immunology, School of Basic Medical Sciences, Fudan University, Shanghai 200032, China. 6 In Vitro Fertilization–Embryo Transfer Center, Hospital of Obstetrics and Gynecology, Fudan University Shanghai Medical College, Shanghai 200011, China. 7Department of Obstetrics, Hospital of Obstetrics and Gynecology, Fudan University Shanghai Medical College, Shanghai 200011, China. *These authors contributed equally to this work. †Corresponding author. Email: [email protected] (X.C.); [email protected] (M.D.)

Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

not useful in predicting pregnancy outcome (7). Therefore, an accurate diagnostic biomarker to predict miscarriage is urgently needed. Multiple mechanisms for evading the rejection of embryos by the maternal immune system have been proposed (8–10). As a critical regulator of active immune tolerance, inducible regulatory T cells (iTregs) express the transcription factor Foxp3 and increase in number in the peripheral blood and decidua during pregnancy, playing indispensable roles in the generation of maternal-fetal tolerance in mice and humans (11, 12). Natural killer (NK) cells with a high functional diversity have many regulatory roles in pregnancy. Although much attention has been paid to studying decidual NK (dNK) cells in pregnancy because of their marked abundance (13–15), NK cells in the peripheral blood (pNK) may also play indispensable roles in pregnancy. These cells participate in increasing the number of T helper 2 (TH2) cells while decreasing the number of TH1 cells during gestation (16, 17). In addition, an increased proportion of CD56+CD16+ NK cells is found in women with recurrent pregnancy loss, which is accompanied by an increase in the cytotoxic activity of pNK cells (18–20). Nevertheless, the precise role of pNK cells during normal pregnancy (NP) and their regulatory mechanisms remain poorly understood. T cell immunoglobulin and mucin-containing protein 3 (Tim-3), a type I membrane protein, was initially identified on terminally differentiated TH1 cells. The identification of galectin-9 (Gal-9) as a ligand for Tim-3 has now firmly established the Gal-9–Tim-3 interaction as an important regulator of TH1 immunity by inducing TH1 cell apoptosis (21, 22). Tim-3 is also found on TH17 and T cytotoxic 1 cells. Tim-3 acts as a negative regulatory molecule to suppress cell-mediated immune responses and promote immune tolerance (23–25). The engagement of Tim-3 by Gal-9 suppresses allograft rejection and improves the survival of allogeneic skin grafts (26). Because embryos express paternal 1 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

Pregnancy loss occurs in about 15% of clinically recognized pregnancies, and defective maternal-fetal immune tolerance contributes to more than 50% of these events. We found that signaling by the type I membrane protein T cell immunoglobulin and mucin-containing protein 3 (Tim-3) in natural killer (NK) cells had an essential protective role during early pregnancy. Tim-3 on peripheral NK (pNK) cells was transiently increased in abundance during the first trimester of pregnancy, which depended on interleukin-4 (IL-4)–signal transducer and activator of transcription 6 (STAT6) and progesterone signaling. Tim-3+ pNK cells displayed immunosuppressive activities, including the production of anti-inflammatory cytokines and the induction of regulatory T cells (Tregs) in a transforming growth factor–b1 (TGF-b1)–dependent manner. Tim-3 on pNK cells was stimulated by its ligand galectin-9 (Gal-9), leading to signaling by the kinases c-Jun N-terminal kinase (JNK) and AKT. In recurrent miscarriage (RM) patients, Tim-3 abundance on pNK cells was reduced and the immunosuppressive activity of Tim-3+ pNK cells was impaired. Compared to Tim-3+ pNK cells from donors with normal pregnancies, RM patient Tim-3+ pNK cells exhibited changes in DNA accessibility in certain genetic loci, which were reversed by inhibiting accessible chromatin reader proteins. Furthermore, Tim-3+ pNK cells, but not Tim-3− pNK cells, reduced fetal loss in abortionprone and NK cell–deficient mice. Together, our findings reveal a critical role for Tim-3–Gal-9 signaling–mediated immunoregulation by pNK cells in maternal-fetal immune tolerance and suggest that Tim-3 abundance on pNK cells is a potential biomarker for RM diagnosis.

SCIENCE SIGNALING | RESEARCH ARTICLE

RESULTS

Tim-3+ NK cells are transiently increased in number in peripheral blood during early human pregnancy To characterize Tim-3 function in maternal-fetal immune tolerance, we first analyzed the expression of Tim-3 in peripheral blood mononuclear cells (PBMCs) during pregnancy. Leukocytes, NK cells, CD11c+ DCs, and CD14+ monocytes all showed a substantial cell surface abundance of Tim-3 (fig. S1, A and B). Although DCs and monocytes expressed Tim-3, the percentages of these cell populations that were Tim-3 did not vary during pregnancy (fig. S1C). In contrast, the percentage of Tim-3+ NK cells in the first trimester of pregnancy was considerably greater than that in either nonpregnant donors or those in the second trimester (Fig. 1, A and B). Among pNK cells (including two major subsets, CD56dim and CD56high NK cells), nearly 90% of CD56dim NK cells expressed Tim-3, whereas only about 40% of CD56high NK cells were Tim-3+ (Fig. 1C). Together, these data suggest that the number of Tim-3+ NK cells is transiently increased in the peripheral blood during the first trimester of pregnancy, a critical period for establishing maternal immune tolerance to the fetus. Tim-3+ NK cells display immunosuppressive activity during pregnancy To further investigate the role of Tim-3+ NK cells in human pregnancy, we analyzed the cytokines that they produce. Compared to Tim-3− NK Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

cells, Tim-3+ NK cells produced more anti-inflammatory cytokines, including TGF-b1, IL-10, and IL-4, but less of the pro-inflammatory tumor necrosis factor–a (TNF-a) (Fig. 1, D and E). The abundance of perforin (Fig. 1E) and the extent of cytotoxicity against HTR-8/SVneo cells (a first trimester human trophoblast cell line) (Fig. 1F) were also substantially reduced in Tim-3+ NK cells compared to Tim-3− NK cells. Similar findings were obtained from a comparison of the abundances of the mRNAs encoding these cytokines between these two subpopulations of NK cells (Fig. 1G). These results suggest that Tim-3+ NK cells have a robust immunosuppressive activity, which was further confirmed by microarray analysis (Fig. 1, H and I). A gene ontology (GO) analysis revealed that cytokines and cytokine activity represented a major transcriptional difference between Tim-3+ and Tim-3− pNK cells in early pregnancy (fig. S2). These data suggest that Tim-3+ pNK cells have a high capacity to produce anti-inflammatory cytokines and a diminished cytotoxicity toward trophoblasts. These traits are conducive to an immune tolerant microenvironment. The cell surface abundance of Tim-3 on NK cells is enhanced through IL-4–STAT6 or progesterone signaling We next investigated the factors that contributed to the increased abundance of Tim-3 on NK cells in early pregnancy. Because of the predominance of TH2 over TH1 cytokines during normal gestation (31, 32), we treated pNK cells with the TH2 cytokine IL-4 and found that Tim-3 abundance was substantially increased (Fig. 2A). Treatment with the TH1 cytokine IFN-g reduced the abundance of Tim-3 under both basal and IL-4–stimulated conditions (Fig. 2A). Moreover, IL-4 activity was dependent on its downstream transcription factor STAT6, because a STAT6 inhibitor (A77-1726) abrogated the increased Tim-3 abundance induced by IL-4 (Fig. 2, B and C). We also observed the potential regulatory effects of pregnancy-associated hormones on Tim-3 abundance on NK cells. Progesterone exerted a biphasic regulatory effect on Tim-3 abundance, which depended on its concentration (Fig. 2D). Progesterone at concentrations of 0.01 to 1.0 nM increased the abundance of Tim-3 on pNK cells in a dosedependent manner, whereas greater concentrations of progesterone (>100 nM) had an inhibitory effect, suggesting that physiological concentrations of progesterone promoted the cell surface expression of Tim-3. However, estradiol and b-hCG did not affect Tim-3 abundance on NK cells (fig. S3). Together, these data suggest that the abundance of Tim-3 on pNK cells increased during early pregnancy through IL-4–STAT6 or progesterone signaling. Gal-9–Tim-3 interactions induce the immunosuppressive activity of NK cells in maternal blood Previous studies indicated that the immunoregulatory effect of Tim-3 depends on the interaction with its ligand, Gal-9 (21, 33, 34). Thus, we treated pNK cells with recombinant human Gal-9 (rhGal-9) and detected substantial increases in the amounts of TGF-b1 and IL-10 produced by pNK cells but a decreased amount of TNF-a production (Fig. 2E). The activity of rhGal-9 was dependent on Tim-3, because an anti–Tim-3 neutralizing antibody blocked rhGal-9 activity, whereas a neutralizing antibody against CD44, another target receptor for Gal-9, failed to abrogate the effect of rhGal-9 on NK cells (Fig. 2E). Similarly, rhGal-9 suppressed the production of perforin (Fig. 2E) and inhibited pNK cytotoxicity (Fig. 2F), and these effects were abrogated by the anti–Tim-3 antibody but not the anti-CD44 antibody. Therefore, these data suggest that Gal-9 might modulate pNK cell function toward an immune tolerant state by binding Tim-3. 2 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

antigens, NP is often regarded as a successful allogeneic transplantation. In this regard, it is tempting to explore whether Gal-9–Tim-3 signaling plays a role in protecting the fetus from being attacked by the maternal immune system. In addition to being found on T cells, Tim-3 is also found on innate immune cells, including mast cells, macrophages, dendritic cells (DCs), and NK cells (23, 25, 27). Tim-3 not only serves as a marker of cellular activation or maturation but also suppresses NK cell cytotoxicity (28). An increase in the cell surface abundance of Tim-3 on NK cells leads to NK cell dysfunction during chronic viral infections (29, 30). Therefore, Tim-3 exerts diverse regulatory effects on NK cells under different immune contexts, involving numerous pathological and physiological processes. Whether Tim-3–mediated NK cell functions influence maternal-fetal immune tolerance and pregnancy outcome is currently unclear. Here, we investigated the role of Tim-3 on pNK cells during early pregnancy in mice. We found that Tim-3+ NK cells proliferate in the first trimester through signaling by interleukin-4 (IL-4) and signal transducer and activator of transcription 6 (STAT6) or by progesterone signaling. Tim-3+ NK cells displayed tolerant phenotypes and induced the differentiation of naïve CD4+ T cells into iTregs in a transforming growth factor–b1 (TGF-b1)–dependent manner. Furthermore, in patients with RM, NK cells had reduced cell surface expression of Tim-3, and Tim-3+ NK cells produced less anti-inflammatory cytokines and lost their ability to induce iTreg generation. A loss of the suppressive activity of Tim-3+ NK cells in RM was associated with a global change in chromatin accessibility, which was partially reversed by the inhibition of accessible chromatin reader proteins. In abortion-prone (AP) or NK cell–deficient mice, the adoptive transfer of Tim-3+ NK cells reduced fetal loss, whereas the transfer of Tim-3− NK cells had no effect. Our data suggest that the Gal-9–Tim-3 axis plays a role in the maintenance of NP through the regulation of pNK cell functions, suggesting that Tim-3+ pNK cells might be a promising biomarker for the prediction of RM.

SCIENCE SIGNALING | RESEARCH ARTICLE A

B

C

Non 1st 2nd

CD56

CD56

Tim-3 β-Tubulin R1 R2

38.6%

89.3%

CD3

Tim-3

3.37%

30.1%

0.52%

37.4%

8.63%

8.51%

Perforin

28.5%

IFN-γ

43.1%

TNF-α

2.4%

IL-4

15.9%

IL-10

TGF-β1

D 59.6%

Tim-3+ NK

72.7%

Tim-3– NK

F

–Log10 (P)

H

0 2 4 6 8 10 13 16 19 22 25

E

G

I

Tim-3– NK (n = 100)

Tim-3+ NK (n = 865 )

–3.0 –2.0 –1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Log2

Fold change

(Tim3+/Tim3–)

To identify the signaling pathways downstream of Gal-9–Tim-3, we stimulated pNK cells with rhGal-9 and examined the phosphorylation of key kinases involved in NK cell function. We found that rhGal-9 substantially increased the abundances of total AKT and phosphorylated AKT (pAKT), as well as of phosphorylated c-Jun N-terminal kinase (pJNK) and phosphorylated extracellular signal–regulated kinase 1/2 (pERK1/2) (Fig. 2G). Furthermore, the specific inhibition of the JNK or AKT signaling pathways, but not ERK1/2 signaling, eliminated the effect of rhGal-9 on pNK cells (Fig. 2H). These results suggest that the Gal-9–Tim-3 interaction may induce a tolerant phenotype in pNK cells through the activation of JNK and AKT signaling. In addition to these downstream signaling molecules, we also analyzed the expression of genes encoding transcription factors that are associated with NK cell Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

development. After rhGal-9 stimulation, the abundances of mRNAs of ID2, BLIMP1, and GATA3 were substantially increased, whereas that of EOMES was markedly reduced (Fig. 2I). Similar results were found at the protein level (Fig. 2J). The altered expression of ID2, BLIMP1, GATA3, and EOMES induced by rhGal-9 was abrogated by the specific inhibition of JNK or AKT (Fig. 2K). All ex vivo analyses suggested a Gal-9–dependent mechanism, whereby Tim-3 inhibits pNK cell activity. The immunosuppressive activity of Tim-3+ NK cells is impaired in patients with RM To evaluate the potential relevance of Tim-3+ NK cells in maternal-fetal tolerance, we first quantified these cells in patients with RM by flow cytometry. Although the frequency of Tim-3+ NK cells was comparable 3 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

CD56

Fig. 1. Human Tim-3+ NK cells increase in number and have an immune tolerant phenotype during early pregnancy. (A to C) Tim-3 abundance on NK cells is increased during the first trimester of pregnancy. (A and B) The percentage of peripheral CD56+ cells from nonpregnant donors (Non, 10 donors), first-trimester donors (1st, 30 donors), and second-trimester donors (2nd, 9 donors) that were Tim-3 + was determined by flow cytometric analysis (A), whereas the abundance of Tim-3 in the same cells was determined by Western blotting analysis (B). (C) The percentages of CD3−CD56+, CD3−CD56dim, and CD3−CD56bright pNK cells in early pregnancy that were Tim-3+ were determined by flow cytometry. Data are from 30 donors. (D and E) Human Tim-3+ pNK cells show an immunosuppressive phenotype. Purified Tim-3+ or Tim-3− CD56+ pNK cells were stimulated with phorbol 12-myristate 13-acetate, ionomycin, and brefeldin A for 4 hours. The cells were analyzed by flow cytometry to detect the indicated cytokines and perforin. (D) Numbers in the plots show the percentage of cells in the boxed regions from a representative experiment. IFN-g, interferon-g. (E) Data are means ± SEM of the percentage of cells positive for the indicated marker from six independent experiments. (F) Tim-3+ NK cells have a reduced cytotoxicity toward trophoblast cells. Cytotoxicity was determined with CytoTox 96 Non-Radioactive Cytotoxicity Assay. The trophoblast cell line HTR-8/SVneo was used as the target (T) cell for NK cells. Data are means ± SEM of nine samples per group. (G) Tim-3+ and Tim-3− pNK cells were analyzed by real-time polymerase chain reaction (PCR) to determine the relative abundance of the indicated mRNAs. Data are means ± SEM of three experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test. (H and I) Tim-3+ and Tim-3− NK cells were sorted from the PBMCs of a pregnant donor at first trimester. RNA was extracted, and global gene expression was analyzed by microarray. Data are representative of six donors. The unique gene signatures based on a >2-fold difference in expression for Tim-3− NK cells (100 probes, gray) and Tim-3+ NK cells (865 probes, yellow) are highlighted. (I) The heat map depicts the expression of genes associated with immune tolerance and activation in Tim-3− and Tim-3+ NK cells.

SCIENCE SIGNALING | RESEARCH ARTICLE B

A

Fig. 2. The IL-4–STAT6 pathway and progesterone promote the increased number of Tim-3+ NK cells in early pregnancy, whereas Gal-9–Tim-3 110 kDa 17.1% 28.1% 63.4% 36.9% pSTAT6 interactions suppress NK cell activation. (A and 100 kDa STAT6 B) NK cells isolated from nonpregnant donors 54 kDa β-Tubulin were left untreated (Ctrl) or were treated with IL-4 CD56 (100 ng/ml), IFN-g (35 ng/ml), or both for 48 hours. The cells were then analyzed by flow cytometry to C D determine the percentage of Tim-3+ cells (A) and by IL-4+vehicle IL-4+A77-1726 Western blotting to detect the phosphorylation 32.1% 57.4% of STAT6 (B). pSTAT6, phosphorylated STAT6; tSTAT6, total STAT6. (C) NK cells were stimulated with IL-4 alone or in the presence of the STAT6 inhibitor A77-1726 for 48 hours. The cells were then CD56 analyzed by flow cytometry to determine the perE F centage of Tim-3+ cells. *P < 0.05, **P < 0.01, and ***P < 0.001 by analysis of variance (ANOVA). DMSO, dimethyl sulfoxide. (D) NK cells stimulated with the indicated concentrations of progesterone were analyzed by flow cytometry to determine the percentage of Tim-3+ cells. ***P < 0.001 by Student’s t test. Data in (A) to (D) are means ± SEM of three independent experiments. (E) pNK cells from early pregnant donors were stimulated by rhGal-9 alone or in the presence of blocking antibodies against Tim-3 or CD44, as indicated, for G H 48 hours. The cells were then analyzed by flow cyGal-9 0 10 30 60 (min) tometry to determine the percentage of cells pospNF-κB 110 kDa itive for the indicated cytokines or perforin. Data 100 kDa NF-κB are means ± SEM of 12 donors. IgG, immunopAKT 60 kDa AKT globulin G. (F) NK cells were treated with rhGal-9 60 kDa pJNK 48 kDa in the presence or absence of the indicated JNK 48 kDa blocking antibodies before being incubated with pERK 44 kDa target cells at the indicated ratios to determine ERK 44 kDa cytotoxicity as described in Materials and Methods. GAPDH 37 kDa Data are means ± SEM of four experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA. (G) NK cells were stimulated with rhGal-9 for the indicated times before being analyzed by Western blotting with antibodies against the indicated proteins. Bar graphs show pooled densitometric data from three experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test. pNF-kB, phosphorylated nuclear factor kB; GAPDH, glyceraldehydeI J K Ctrl Gal-9 3-phosphate dehydrogenase; NS, not significant. 17 KDa ID2 90 KDa (H) Purified NK cells were stimulated with rhGal-9 BLIMP1 GATA3 48 KDa in 15 mM LY294002, 10 mM SP600125, 10 mM U0126, EOMES 17 KDa or DMSO, as indicated. The cells were analyzed by 37 KDa GAPDH flow cytometry to determine the percentage of cells positive for the indicated cytokines or perforin. Data are means ± SEM of 10 experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by ANOVA. (I and J) NK cells were left unstimulated or were stimulated with rhGal-9 for 48 hours and then analyzed by real-time PCR to determine the relative abundances of the indicated mRNAs (I) and by Western blotting with antibodies against the indicated proteins (J). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test. (K) NK cells were left untreated or treated with rhGal-9 in the presence or absence of the indicated inhibitors. The relative abundances of the indicated mRNAs were then determined by real-time PCR analysis. Data in (I) to (K) are means ± SEM of three experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA. IL-4

IFN-γ IL-4+IFN-γ

Tim-3

Tim-3

Ctrl

Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

peripheral immune cells (Fig. 3C) were reduced in patients with RM compared to controls. Compared to Tim-3+ NK cells from control donors with NPs, Tim-3+ pNK cells from patients with RM produced less of the anti-inflammatory cytokines TGF-b1, IL-10, and IL-4 but more pro-inflammatory cytokines, 4 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

between patients with RM and those control donors with NPs, the cell surface abundance of Tim-3 [as assessed by measurement of the mean fluorescence intensity (MFI)] on pNK cells was substantially decreased in patients with RM (Fig. 3A). Furthermore, the plasma Gal-9 concentration (Fig. 3B) and the amount of intracellular Gal-9 produced in

SCIENCE SIGNALING | RESEARCH ARTICLE

Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

5 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

scape of chromatin accessibility in their Tim-3+ and Tim-3− NK cells. An assay for transposase-accessible chromatin sequencing (ATAC-seq) was used to analyze global chromatin accessibility (35). To determine whether global chromatin accessibility reflected the differences among samples, we performed a principal components analysis (PCA). We found that the samples formed three well-separated groups, including Tim-3+ NK cells from RM patients, Tim-3− NK cells from RM patients, and NK cells (either Tim-3+ or Tim-3−) from normal donors (Fig. 4A). PC1, which accounted for 50% of the total variance, separated the Tim-3+ cells from the Tim-3− NK cells of RM patients. PC2, which accounted for 15% of the variance, distinguished the NK cells of normal donors from those of RM patients (Fig. 4A). Note that the NK cell subsets from the same normal donor tended to cluster together, suggesting that individual differences overshadowed the differences between Tim-3+ and Tim-3− cells in normal donors. Similarly, hierarchical clustering confirmed the separation between patient NK cells and normal donor NK cells (Fig. 4B). These results suggest that the chromatin accessibility of pNK cells can capture different pregnancy outcomes. Thus, these findings suggest that Fig. 3. Tim-3+ NK cells from RM patients have reduced immunosuppressive activity. (A) The relative abundances of Tim-3+ pNK cells represent an indicator Tim-3 on pNK cells from donors with NP and patients with RM were determined by flow cytometry. Data are means ± SEM of of pregnancy status, which prompted 20 donors per group. (B) The concentrations of soluble Gal-9 in the plasma of NP donors and RM patients were determined our desire to investigate the underlying by enzyme-linked immunosorbent assay. Data are means ± SEM of six NP donors and eight RM patients. (C) PBMCs from NP mechanism by which Tim-3+ NK cells + donors and RM patients were analyzed by flow cytometry to determine the percentages of Gal-9 cells among the indicated + − are involved in gestation. immune cell subsets. Data are means ± SEM of six NP donors and eight RM patients. (D) Tim-3 NK cells (left) and Tim-3 NK We next compared the Tim-3+ NK cells (right) from NP donors and RM patients were analyzed by flow cytometry to determine the percentages of cells that + cells from RM patients to those from were positive for the indicated cytokines or perforin. Data are means ± SEM of 8 NP donors and 10 RM patients. (E) Tim-3 and Tim-3− NK cells from NP donors and RM patients were analyzed to determine their cytotoxicity, as described earlier. Data normal donors and identified 260 reare means ± SEM of nine donors per group. *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test. gions with statistically significantly reduced accessibility but 360 regions including TNF-a and IFN-g, as well as more perforin (Fig. 3D). The with significantly increased accessibility in the cells from the RM pacytotoxicity of Tim-3+ pNK cells from patients with RM was also tients (P < 0.05) (Fig. 4C). On the basis of published chromatin immuenhanced (Fig. 3E), suggesting that there were functional defects in noprecipitation sequencing (ChIP-seq) results for human CD56+ NK the Tim-3+ NK cells of patients with RM. However, cytokine produc- cells, the regions with reduced accessibility in the RM patients were astion by the Tim-3− NK cells and their cytotoxicity were comparable be- sociated with H3K4me1hi H3K4me3low (the two forms of methylation tween RM patients and donors with NPs (Fig. 3, D and E). Accordingly, of Lys4 on histone H3) histone modifications and were distal to the immunosuppressive activity of Tim-3+ NK cells was markedly re- transcription start sites (TSSs), suggesting that these regions were duced in patients with RM, which may account for the uncontrolled mainly enhancers. In contrast, regions with increased accessibility in allogeneic response to the fetus. the RM patient cells were enriched for H3K4me1low H3K4m3hi and were proximal to the TSS, suggesting that they were mainly promoters Chromatin accessibility of genes encoding anti- and (Fig. 4, D and E). These data suggest that promoters and enhancers pro-inflammatory cytokines is altered in Tim-3+ NK cells were regulated by different mechanisms in Tim-3+ NK cells. from patients with RM Furthermore, one of the regions with the most diminished accessiTo systemically characterize the difference in Tim-3+ pNK cells bility in the Tim-3+ NK cells from RM patients was the 3′ end of the between RM patients and donors with NPs, we assessed the land- TGFB1 gene, which was associated with a reduced promoter accessibility

SCIENCE SIGNALING | RESEARCH ARTICLE NP Tim-3+

RM Tim-3+

NP Tim-3–

RM Tim-3–

100

NP1+ NP3– NP2+ NP2–

NP3

0

NP3+ NP1–

RM2

RM1– RM2– RM1+

RM1

Normal

C

D

Accessibility decreased in RM

Accessibility unchanged

RM2+

RM1+

RM2–

PC1 (50%)

NP1–

200

RM1–

100

NP3+

0

NP2–

RM2+

–100

NP2+

–200

NP1+ NP3–

–200

B

NP2

NP1

–100

PC2 (15%)

200

A

Patients Accessibility increased in RM

-Log10 P(adj)

TGF-β1

POTEA MIR5583-1

4

H3K4me1

2

E P < 0.05

0 -4

-2 0 2 Log2 (patient/normal)

4

Distance to TSS (bp)

F

G 55

Promoter

3'Enhancer

25

55

25

55

25

55

25

55

25

55

25

55

25

Enhancer

NP 1 NP 2 RM 1 RM 2 NP 1 NP 2 RM1

Tim-3–

25

Tim-3+

55

RM2

H

TGFB1

IL10

of TGFB1, suggesting that this region functions as an enhancer to coordinate with the TGFB1 promoter (Fig. 4, C and F). This enhancer in the Tim-3− NK cells from RM patients showed a similar accessibility to that of the enhancer in Tim-3− NK cells from normal donors, suggesting that a cell-specific mechanism regulates TGFB1 expression in Tim-3+ NK cells (Fig. 4F). Similarly, the accessibility of the IL10 locus was also reduced in the Tim-3+ NK cells from RM patients (Fig. 4G). These data suggest that Tim-3 activation in NK cells during pregnancy Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

promotes the accessibility of loci encoding anti-inflammatory cytokines, which enhances their expression and induces immune tolerance to the fetus. A decreased expression of Tim-3, together with other defects, may inhibit the accessibility of these genes encoding anti-inflammatory cytokines in RM patients and contribute to immune-based rejection of the fetus. Chromatin-accessible regions are usually associated with histone acetylation, which is recognized by bromodomain-containing proteins 6 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

NFIX

H3K4me3

Tim-3+ 6

Fig. 4. Changes in chromatin accessibility in Tim-3+ NK cells are involved in the abnormal cytokine profile associated with RM. (A and B) Tim-3+ and Tim-3− NK cells were sorted from NP donors (n = 3) or RM patients in the first trimester (n = 2). Global chromatin accessibility was analyzed by ATAC-seq and was compared using PCA (A). On the basis of a hierarchical clustering analysis, RM patient Tim-3+ NK cells displayed distinct global chromatin accessibility compared with NK cells from NP donors (B). (C) ATAC-seq analysis was used to compare chromatin accessibility between Tim-3+ NK cells from NP donors and RM patients. Regulatory regions showing a statistically significant difference (P < 0.05) are labeled in red. (D and E) Tim-3+ NK cells from NP donors and RM patients were analyzed to assess chromatin accessibility at enhancer and promoter regions. Regulatory regions that showed reduced accessibility (left), unaltered accessibility (middle), or increased accessibility (right) were analyzed for the presence of H3K4me3 and H3K4me1 (D) or for distance to the closest TSS (E). bp, base pair. (F and G) Tim-3+ NK cells from NP donors and RM patients were analyzed to assess accessibility at the TGFB1 (F) and IL10 (G) loci. Top to bottom: The genome browser tracks of ATAC-seq for Tim-3+ NK cells from two healthy donors and two RM patients and then those for the Tim-3− NK cells from the same individuals. The numbers on the left indicate the ATAC-seq reads after normalization to 10 × 106 mapped reads. (H) Sorted Tim-3+ and Tim-3− NK cells from NP donors and RM patients were treated with the indicated concentrations of JQ1, and the percentages of the NK cells that were positive for TNF-a (left) and IFN-g (right) were determined by flow cytometry. Data are means ± SEM of three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test.

SCIENCE SIGNALING | RESEARCH ARTICLE (36). To investigate the relationship between altered chromatin accessibility and cytokine-encoding gene expression, we treated the Tim-3+ or Tim-3− NK cells with JQ1, an inhibitor of the BET family of bromodomain proteins (37). We observed that the percentages of Tim-3+ NK cells from RM patients that were positive for the inflammatory cytokines TNF-a and IFN-g were substantially reduced by JQ1 (Fig. 4H). Treatment with JQ1 also reduced the percentages of Tim-3− NK cells from RM patients with normal donors that were positive for TNF-a and IFN-g, although to a lesser degree. Furthermore, JQ1 had no effect on the percentages of Tim-3+ NK cells from normal donors that were positive for inflammatory cytokines (Fig. 4H). The production of TGFb1 and perforin was unaffected by JQ1 (fig. S4). These data suggest that the increased pro-inflammatory cytokine production by Tim-3+ NK cells from patients with RM was at least partially caused by the increased accessibility of the loci of the corresponding genes.

Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

7 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

Tim-3+ NK cells from RM patients are less potent than those from normal donors in promoting iTreg differentiation TGF-b1 is a key factor in inducing the differentiation of naïve CD4+ T cells to iTregs. Our earlier results (Figs. 3D and 4, C and F) showed that the Tim-3+ NK cells from normal donors produced large amounts of TGF-b1 compared to those from RM patients, which also showed the most diminished promoter accessibility of TGFB1. Therefore, we hypothesized that Tim-3+ NK cells could promote the generation of iTregs and that this activity of Tim-3+ NK cells might be impaired in patients with RM, which could contribute to the pathogenic role of this NK cell population in RM. To investigate this hypothesis, we first cultured Tim-3+ or Tim-3− NK cells with CD4+CD45RA+ naïve T cells from the same donor at early pregnancy. Coculture with Tim-3+ NK cells, but not with Tim-3− NK cells, induced the production of Foxp3 in the CD4+ T cells, and this was inhibited by an anti–TGF-b1 neutralizing antibody (Fig. 5A). The production of TGF-b1 in the iTregs was also promoted by Tim-3+ NK cells (Fig. 5B). In addition, we investigated the in vitro capacity of the iTregs that were induced by Tim-3+ NK cells to inhibit proliferation and IFN-g production from CD4+CD25− T cells. We found that the proliferation of CD4+CD25− T cells in response to anti-CD3 and anti-CD28 antibodies was markedly inhibited by peripheral naïve CD4+CD25+ T cells (pTregs) and CD4+CD25+Foxp3+ T cells induced by Tim-3+ NK cells (iTregs) (Fig. 5C). IFN-g production by the CD4+CD25− T cells was also reduced by pTregs and iTregs (Fig. 5D). These results suggest that Tim-3+ NK cells not only induce the differentiation of iTregs but also promote their immunosuppressive function in a TGF-b1–dependent manner. We next compared Tim-3+ NK cells from RM patients and matched controls for their capacity to induce the differentiation of CD4+CD45RA+ naïve T cells into Tregs. In coculture experiments with naïve CD4+ T cells, Tim-3+ NK cells from RM patients were less capable than control cells to induce Foxp3+ T cell generation (Fig. 5, E and F). Using a crossover approach to compare the effect of Tim-3+ NK cells derived from RM patients and from controls on Treg differentiation and function, we found that the percentage of Foxp3+ and TGF-b1+ cells induced by RM patient Tim-3+ NK cells was substantially less than that induced by control Tim-3+ NK cells (Fig. 5G). This is consistent with our observation that Tim-3+ NK cells from RM patients produce less TGF-b1 compared with those from controls. From these results, we conclude that Tim-3+ NK cells in maternal peripheral blood can promote iTreg differentiation and enhance iTreg function through TGF-b1 but that this capacity is attenuated in RM patients.

Tim-3 blockade disrupts immune tolerance and induces pregnancy failure To further explore the function of Tim-3 in pregnancy, we used a wellestablished AP mouse model (38, 39). Consistent with observations in human pregnancies, the percentage of Tim-3+ NK cells in the first trimester of pregnancy was substantially greater than in the second trimester in mice (fig. S5A). In addition, in the AP mice, the abundances of Tim-3 and Gal-9 on NK cells were decreased (fig. S5B), which was accompanied by abnormal cytokine production and cytotoxicity (fig. S5, C and D). The conserved function of Tim-3+ NK cells between humans and mice further highlights the importance of the Tim-3 signaling in the immunoregulation of NK cells during pregnancy. We next treated the AP and NP mice with a Tim-3–blocking antibody, RMT3-23. This antibody substantially decreased litter size and increased the resorption rate in NP and AP mice (Fig. 6, A to C). The fetus size for mice treated with RMT3-23 was also markedly decreased (Fig. 6, D and E), suggesting that Tim-3 blockade had an adverse effect on embryo development. In addition, we found that the NK cell number in peripheral blood declined markedly (fig. S6A) in mice treated with RMT3-23. Flow cytometric analysis showed that the greatest abundance of Tim-3 was on mature CD11b single-positive (SP) cells but that the lowest abundance of Tim-3 was on immature CD27 SP cells (fig. S6, B and C). Moreover, Tim-3 abundance was gradually increased during the process of NK cell differentiation from CD34+ cells to mature NK cells (fig. S6D). Together, these data suggest that Tim-3 is closely associated with NK cell development and maturation. Furthermore, Tim-3 blockade inhibited the production of anti-inflammatory cytokines but promoted pro-inflammatory cytokine production by the splenic NK cells in both NP and AP mice (Fig. 6F). Similar results were obtained in dNK cells (fig. S7A). However, we observed that activating Tim-3 by administering recombinant mouse Gal-9 (rmGal-9) reduced embryo resorption in AP mice but not in normal mice (Fig. 6, A and D). The RMT3-23–induced fetal loss in both NP and AP mice was not reversed by rmGal-9 administration (Fig. 6, A and C). Consistent results were obtained for cytokine production and the cytotoxicity of NK cells in the spleen (Fig. 6F) and decidua (fig. S7A). These observations suggest a possible role for Gal-9 and Tim-3 signaling in the regulation of the immune tolerant propensity of NK cells in pregnancy. To assess the signaling pathways downstream of Gal-9 and Tim-3, the relative abundances of pJNK and pAKT were measured by performing intracellular staining. The inhibitory effect of RMT3-23 on JNK and AKT phosphorylation in NK cells was observed, and the administration of rmGal-9 failed to reverse that inhibitory effect. However, we detected substantial JNK and AKT phosphorylation in NK cells from mice treated with rmGal-9 (Fig. 6G). Furthermore, NK cells from RMT3-23–challenged mice had decreased amounts of the transcriptional regulators Id2, Blimp-1, and Gata-3 but an increased amount of Eomes compared to NK cells from control mice. Conversely, NK cells from mice treated with rmGal-9 exhibited increased amounts of Id2, Blimp-1, and Gata-3 but a decreased amount of Eomes (Fig. 6H). Consistent results were observed in the decidua (fig. S7, B and C). Thus, the regulatory effect of Tim-3 and Gal-9 on NK cells during pregnancy might depend on the JNK and AKT signaling pathways and might be mediated through the modulation of the transcription factors Id2, Blimp-1, Gata-3, and Eomes. The in vitro experiments revealed that Tim-3+ NK cells can induce Treg generation. Here, our in vivo studies showed that interfering with

SCIENCE SIGNALING | RESEARCH ARTICLE

Foxp3

A

Con

Tim-3– NK

2.83

4.32

Tim-3+ NK Tim-3+ NK + α-TGF-β1 9.34

31.4

NP

CD4

TGF-β1

B

Con

Tim-3– NK

Tim-3+ NK Tim-3+ NK + α-TGF-β1

NP

Foxp3

C CD25

: iTreg : pTreg : CD4+CD25– T cells : pTreg+CD4+CD25– T cells : iTreg+CD4+CD25– T cells EdU

Foxp3

E

Con

Tim-3– NK 5.41

3.94

Tim-3+ NK Tim-3+ NK + α-TGF-β1 17.5

10.2

RM

CD4

TGF-β1

F

Con 0.64

Tim-3– NK

Tim-3+ NK Tim-3+ NK + α-TGF-β1

1.27

7.61

3.49

RM 2.35

4.62

16.3

10.1

Foxp3

G

the Tim-3-Gal-9 interaction notably affected the phenotype and function of NK cells from pregnant mice. Next, we wondered whether the Tregs would show some consistent changes. Our results showed that the proportions of CD25+Foxp3+ Tregs in blood and spleen were reduced in mice treated with RMT3-23 compared to those in untreated mice (Fig. 6, I and J). Consistent with this observation, these Tregs displayed an impaired immunosuppressive capacity and reduced TGFLi et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

b1 production (Fig. 6, I and J). Furthermore, the administration of rmGal-9 led to an increase in the proportion of CD25+Foxp3+ Tregs with enhanced TGF-b1 production in the blood and spleen (Fig. 6, I and J). Consistent results were observed in the decidua (fig. S7D). These data suggest that the Tim-3–Gal-9 signaling pathway facilitates the immunosuppressive activity of NK cells and the induction of Tregs during pregnancy. 8 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

D

Fig. 5. Tim-3+ NK cells from RM patients are less potent than those from controls in inducing iTreg differentiation through TGF-b1. (A and B) Naïve CD4 + T cells isolated from the peripheral blood of NP donors were stimulated with anti-CD3 and anti-CD28 antibodies for 5 days before being cocultured in the presence or absence of anti–TGF-b1 antibody with Tim-3+ or Tim-3− NK cells obtained from NP donors. The percentages of Foxp3+CD4+ T cells (top) and TGF-b1+Foxp3+CD4+ T cells (bottom) were determined by flow cytometry. Data in the bar graphs are means ± SEM of nine donors. (C and D) Tim-3+ NK cell–induced CD4+CD25+ T cells (iTregs), pTregs, and CD4+CD25− cells in the indicated combinations were restimulated with anti-CD3 and anti-CD28 antibodies. (C) Their immunosuppressive function was assessed by measurement of 5-ethynyl2′-deoxyuridine (EdU) incorporation. (D) The percentages of the indicated cell populations that were positive for IFN-g were determined by flow cytometry. Data are means ± SEM of nine donors. (E and F) Naïve CD4+ T cells isolated from the peripheral blood of NP donors were stimulated with anti-CD3 and anti-CD28 antibodies for 5 days before being cocultured in the presence or absence of anti–TGF-b1 antibody with Tim-3+ or Tim-3− NK cells obtained from RM patients. The percentages of Foxp3+CD4+ T cells (top) and TGF-b1+Foxp3+CD4+ T cells (bottom) were determined by flow cytometry. Data in the bar graphs are means ± SEM of 12 donors per group. (G) Tim-3+ and Tim-3- NK cells from NP donors and RM patients were analyzed by flow cytometry to determine the percentages of Foxp3+CD4+ T cells (left) and TGF-b1+Foxp3+CD4+ T cells (right). Data are means ± SEM of three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA.

SCIENCE SIGNALING | RESEARCH ARTICLE A

(

)

(

)

(

)

(

)

(

)

(

)

(

)

(

)

D

(

)

(

)

(

)

(

(

)

(

)

(

)

(

C

) ( ( ( ( ) ( ( ( (

): NP+PBS ): NP+RMT3-23 ): NP+Gal-9 ): NP+RMT3-23+Gal-9 ): AP+PBS ): AP+RMT3-23 ): AP+Gal-9 ): AP+RMT3-23+Gal-9

F

pAKT

pJNK Cells

Isotype PBS RMT3-23 Gal-9 RMT3-23+Gal-9 Fluorescence intensity

H

Spleen

Blood Foxp3

I

PBS

RMT3-23

Gal-9 RMT3-23+Gal-9

6.8

1.84

13.9

2.2

11.6

5.21

22.1

6.1

J

E

Fig. 6. Tim-3 blockade breaks maternalfetal tolerance and impairs pregnancy maintenance. (A) Representative pictures showing the number of embryos per uterus from NP and AP mouse models treated with RMT3-23 and Gal-9 alone or in combination, as indicated by conditions (i) to (viii). (B) Analysis of the percentage of fetal resorption in the NP and AP mouse models treated as indicated in (A). (C) Summary of litter sizes at gestational day (GD) 14.5 in the NP and AP mouse models that were treated as indicated in (A). (D) Representative pictures showing embryos from the NP and AP mouse models that were treated as described in (A). PBS, phosphate-buffered saline. (E) Analysis of fetal size (volume) in the NP and AP mouse models at GD 14.5 after treatment as indicated in (A). (F) Spleens from the NP and AP mouse models treated with the indicated combinations of RMT2-23 and Gal-9 were analyzed by flow cytometry to determine the percentages of NK cells that were positive for the indicated cytokines. Data are means ± SEM of five to eight mice per group. (G) Left: Splenic NK cells from the AP mice treated as described in (A) were analyzed by flow cytometry to determine the relative abundances of pJNK and pAKT. Right: Calculation of the MFIs for pJNK and pAKT. Data are means ± SEM of five to eight mice per group. (H) Splenic NK cells from the AP mice treated as described in (A) were analyzed to determine the percentage of cells positive for the indicated transcriptional regulators. Data are means ± SEM of five to eight mice per group. (I and J) Left: Peripheral blood (top) and splenic (bottom) CD4+ T cells from AP mice that were treated as described in (A) were analyzed by flow cytometry to determine the percentage of Foxp3+ cells. Representative dot plots are shown. Right: Calculation of the percentages of Foxp3+CD4+ T cells in the blood and spleen under the indicated conditions. Data are means ± SEM of five to eight mice per group. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA.

CD25

Tim-3+ NK cells facilitate pregnancy maintenance To confirm the function of Tim-3+ NK cells in NP, we purified Tim-3+, Tim-3−, or total NK cells from mice with NP and adoptively transferred these cells into AP mice (Fig. 7A). We observed that the resorption rates of mice that received Tim-3+ NK cells and total NK cells from the mice with NP were substantially reduced. The adoptive transfer of Tim-3− NK cells from mice with NP had no obvious effects on the resorption rate. In contrast, Tim-3+ NK cells and total NK cells isolated from the Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

AP mice failed to reverse the pregnancy outcome (Fig. 7, B and C). In addition, the effect of Tim-3+ NK cells from AP mice was rescued after treatment with IL-4 or progesterone, which was concomitant with the increased Tim-3 abundance on the treated cells (Fig. 7, B to D). Furthermore, we found a marked increase in the proportion of CD25+Foxp3+ Tregs in the blood of AP mice that received Tim-3+ NK cells and total NK cells from mice with NP (Fig. 7E). The transfer of Tim-3+ NK cells and total NK cells from the AP mice had no effect on the proportion of 9 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

G

B

SCIENCE SIGNALING | RESEARCH ARTICLE CD25+Foxp3+ Tregs in the blood of the recipient AP mice. Nevertheless, Tim-3+ NK cells and total NK cells isolated from the AP mice treated with IL-4 or progesterone had a similar function as those cells isolated from mice with NP that increased the proportion of CD25+Foxp3+ Tregs in the blood of the recipient mice (Fig. 7E). Furthermore, the transfer of Tim-3+ NK cells and total NK cells isolated from mice with NP led to the increased production of TGF-b1 by CD25+Foxp3+ Tregs in the

A

Tim-3+ NK Tim-3– NK NK

FACS

Normal pregnancy mouse

Spleen

blood. A similar effect was observed after AP mice received Tim-3+ NK cells and total NK cells isolated from the AP mice treated with IL-4 or progesterone (fig. S8A). Consistent results were obtained in the spleen and decidua (fig. S8, B to E). The potent capacity of Tim-3+ NK cells to induce the production of iTregs and improve the pregnancy outcome was further confirmed by using the Nfil3−/− pregnant mouse model. Because they are deficient

Abortion-prone mouse

Tim-3+

NK Tim-3– NK NK FACS

Tim-3+ NK Tim-3– NK NK

FACS NK + IL-4

Abortion-prone mouse

Spleen

Tim-3+ NK Tim-3– NK NK

FACS

Ctrl Nfil3

+/+

mouse

Spleen

B Tim-3+ NK of AP treated with IL-4

Tim-3+ NK Tim-3– NK

Nfil3–/– mouse

C

Tim-3+ NK of NP

NK of NP

Tim-3+ NK of AP treated with P

Tim-3+ NK of AP

D

E

Tim-3+ NK Tim-3– NK

G Nfil3

Nfil3

–/–

–/–

+ Ctrl

+ Tim-3+ NK

H

Spleen

F

Blood Foxp3

Ctrl 11.9

13.6

22

26.2

9.02

13.8

Spleen

Nfil3–/– + Tim-3– NK

Blood TGF-β1

CD25 4.48

20.2

12.3

24

9.8

9.03

CD25

Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

10 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

FACS NK + Progesterone

Abortion-prone mouse

Fig. 7. Tim-3+ NK cells alleviate fetal loss in AP and NK-deficient mice. (A) Schematic for the adoptive transfer of different NK cell subsets from the indicated pregnant mice to AP pregnant mice or Nfil3−/− pregnant mice. Top: Tim-3+ NK cells, Tim-3− NK cells, and total NK cells were sorted from NP mice and transferred into the tail vein of AP pregnant mice. Middle: Total NK cells sorted from the AP models were transferred directly into the tail vein of pregnant AP mice or were treated with IL-4 or progesterone for 48 hours in vitro before being sorted by fluorescence-activated cell sorting (FACS) into Tim-3+ and Tim-3− NK cells and then transferred to AP pregnant mice. Bottom: Tim-3+ and Tim-3− NK cells were sorted from normal C57BL/6 (Nfil3+/+) pregnant mice and transferred directly into the tail vein of Nfil3−/− pregnant mice. (B) Representative pictures of embryos in the uterus of AP mice that had received the indicated NK cells. (C) NK cells isolated from AP mice were left untreated (Ctrl) or were treated with IL-4 or progesterone (P) before being analyzed by flow cytometry to determine the percentage of Tim-3+ cells. (D) Analysis of fetal resorption in AP mice that had received the indicated NK cells. Data are means ± SEM of 9 to 12 mice per group. (E) Flow cytometric analysis of the percentage of CD25+Foxp3+ cells within the peripheral blood CD4+ T cells of AP pregnant mice that had received the indicated NK cells. (F) Representative pictures of embryos in the uterus of control Nfil3−/− mice or those that received either Tim-3+ or Tim-3NK cells. (G) Analysis of fetal resorption in pregnant control Nfil3−/− mice or those that received the indicated NK cells. Data are means ± SEM of five mice per group. (H) Left: Flow cytometric analysis of the percentages of CD25+Foxp3+ T cells and TGF-b1+CD25+Foxp3+ T cells in control Nfil3−/− mice or those that received the indicated NK cells. Right: Measurement of the percentages of the indicated cells. Data are means ± SEM of five mice per group. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA.

SCIENCE SIGNALING | RESEARCH ARTICLE in NK cells (40), Nfil3−/− pregnant mice display an increased rate of embryo resorption than do the control mice. We sorted Tim-3+ and Tim-3− NK cells from the spleens of Nfil3+/+ mice (which have NP) and injected them into Nfil3−/− pregnant mice. The adoptive transfer of Tim-3+ NK cells, but not Tim-3− NK cells, substantially reduced the embryo resorption rate of the recipient Nfil3−/− mice (Fig. 7, F and G). Furthermore, we found that the proportion of CD25+Foxp3+ Tregs was increased in the blood and spleen of Nfil3−/− mice that received Tim-3+ NK cells. The production of TGF-b1 by the CD25+Foxp3+ Tregs was also substantially greater than that in the mice that received Tim-3− NK cells. The transfer of Tim-3− NK cells had no effect on the proportion of CD25+Foxp3+ Tregs or their production of TGF-b1 (Fig. 7H). Consistent results were observed in the decidua (fig. S8, F and G). These data suggest that Tim-3+ NK cells with the capacity to induce Treg differentiation contribute to immune tolerance and pregnancy maintenance.

NK cells are crucial in the establishment and maintenance of pregnancy because of their functional diversity, including their roles in placental vascular remodeling and immunoregulation. Emerging evidence indicates an association between the abnormality of NK cells and the occurrence of RM. A meta-analysis showed that women with RM have statistically significantly more pNK cells than those with a healthy pregnancy (7, 20). For patients with a history of RM, the increased activity of pNK cells might be predictive of subsequent miscarriage. In addition, cytokine production by pNK cells is dysregulated in women with a history of reproductive failure (41). These studies suggest that women with RM have altered pNK parameters, which can therefore be recommended as biological markers to identify RM and assess the effect of immunotherapy. However, the specific role of pNK cells in pregnancy is poorly defined, and little is known about the underlying molecular mechanism. Here, we described Tim-3+ NK cells with tolerance phenotypes that are characterized by an increased production of anti-inflammatory cytokines (including TGF-b1, IL-10, and IL-4), a reduced production of proinflammatory cytokines (including TNF-a), and reduced cytotoxicity toward trophoblast cells. Furthermore, this subtype of NK cells efficiently induced Treg differentiation. These results suggest that Tim-3+ pNK cells can directly contribute to the anti-inflammatory microenvironment and, more selectively, facilitate iTreg proliferation, which is conducive to maternal-fetal immune tolerance. The systemic generation of Tregs during early gestation is pivotal for successful pregnancy in both human and mouse models. The importance of Tregs for NP has been demonstrated in studies showing that the absence of Tregs can impair murine pregnancy, whereas the adoptive transfer of Tregs can prevent fetal rejection (42, 43). However, the mechanism underlying Treg generation during early pregnancy is unclear. Here, we found that the accessibility of TGFB1 in Tim-3+ pNK cells from patients with RM was reduced, resulting in a marked decrease in the production of TGF-b1 and a subsequent deficiency in Treg differentiation. Thus, our results suggest that previous reports of diminished Treg responses in patients with RM may be attributable to abnormal Tim-3+ NK cells. These findings provide new insights into the mechanism of RM and the versatility of NK cells (44). We found that Tim-3 abundance on pNK cells was decreased in patients with RM, indicating that Tim-3+ NK might be a predictor of aberrant pregnancy. However, the regulation of Tim-3 abundance on NK cells is unclear. Here, we found that the amount of cell surface Li et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

11 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

DISCUSSION

Tim-3 on pNK cells was increased during the first trimester in NP because of the typical TH2 polarization signaling by the IL-4–STAT6 pathway and physiological concentrations of progesterone. However, increased concentrations of progesterone showed an inhibitory effect on Tim-3 abundance, which may explain the decreased abundance of Tim-3 on NK cells in the second trimester of pregnancy, when the progesterone concentration is several times greater than that in early pregnancy. Estrogen and b-hCG showed no effect on Tim-3 abundance on pNK cells, which is consistent with the wide use of progesterone, but not estrogen or b-hCG, in the clinical treatment of early threatened abortion. Moreover, progesterone promotes TH2 responses by inhibiting TH1 cytokine production (45, 46). Thus, we suggest that progesterone induces Tim-3 abundance on pNK cells through a direct effect as well as an indirect effect by regulating the TH2-TH1 balance. This result provides a potential theoretical basis for the clinical application of progesterone in the treatment of RM. Accompanying the deficient Tim-3 expression on pNK cells in RM, we found a substantial decrease in the concentration of the natural ligand Gal-9 in the blood circulation. Furthermore, rmGal-9 administration reduced embryo resorption in AP mice, and the disordered cytokine production of NK cells was efficiently reversed, consistent with the results of our in vitro experiment. We therefore propose that the Tim-3–mediated immunomodulation by pNK cells during pregnancy is enhanced by Gal-9. In addition, Gal-9–Tim-3 activated JNK and AKT, leading to increased ID2, BLIMP1, and GATA3 mRNA abundance and decreased EOMES mRNA abundance. These transcription factors are associated with NK cell development and functions (47–49). Previous reports demonstrated that Tim-3 ligation enhances the activation of the MAPK (mitogen-activated protein kinase) and NF-kB pathways (50). Tim-3 enhances phosphoinositide 3-kinase (PI3K)–AKT activation and A20 activity, inhibiting Toll-like receptor 4 (TLR4)–stimulated NF-kB activation (51). Nevertheless, we showed that JNK and AKT were downstream mediators of Gal-9–Tim-3 signaling. Differences among these studies might be explained by the specific intracellular tail of Tim-3 found in different cell types (50, 52). Thereby, the influence of Gal-9 on pregnancy outcome and the underlying mechanism of its interaction with Tim-3 indicate its potential application in the diagnosis and treatment of RM. We also showed that Tim-3 blockade substantially increased embryo resorption and led to the appearance of dysfunctional pNK cells, further confirming the crucial modulatory effects of Gal-9– Tim-3 on pNK cells in pregnancy. Note that blocking Tim-3 decreased pNK cell numbers (fig. S6A), which is consistent with Tim-3 activation by Gal-9 inducing the expression of genes encoding transcription factors associated with NK development. To find the mechanism underlying the decrease in the number of pNK cells after RMT3-23 treatment, we analyzed the expression of Tim-3 on CD27−CD11b− (double-negative), CD27+CD11b− (CD27 SP), CD27+CD11b+ (doublepositive), and CD27−CD11b+ (CD11b SP) populations among NK cells (CD3−Nkp46+) in mice. The results showed that the greatest abundance of Tim-3 was on mature CD11b SP cells but that the lowest abundance of Tim-3 was on immature CD27 SP cells (fig. S6, B and C). Consistently, flow cytometric analysis of human samples showed that about 80% of mature CD56dim NK cells had cell surface Tim-3, whereas only about 40% immature CD56bright NK cells were Tim-3+ (Fig. 1C). These data imply that Tim-3 might be associated with NK cell development and maturation. Moreover, we analyzed the published microarray data (28) and found that Tim-3 expression was gradually increased during the process of NK cell differentiation from CD34+ cells to mature NK cells

SCIENCE SIGNALING | RESEARCH ARTICLE

MATERIALS AND METHODS

Human sample collection This study was approved by the Research Ethics Committee of Obstetrics and Gynecology Hospital, Fudan University (Shanghai, China). All subjects signed informed written consent for the collection and study of samples. Peripheral blood samples were obtained from normal nonpregnant women, women with clinically normal pregnancy, and patients with RM that occurred during the first trimester. NPs were terminated for nonmedical reasons, and RMs were classified as unexplained after the exclusion of maternal anatomic or hormonal abnormalities or paternal and maternal chromosomal defects. Flow cytometry and cell sorting Cells were washed and incubated with appropriate fluorochromeconjugated antibodies for staining. Flow cytometric analysis was performed on a CyAn ADP analyzer (Beckman Coulter), and data were analyzed with FlowJo version 6.1 software (Tree Star). For flow cytometric sorting, cells were stained with specific antibodies and isolated on a BD FACSAria cell sorter (BD Biosciences). Western blotting analysis Whole-cell protein extracts were prepared by lysing cells in radioimmunoprecipitation assay buffer supplemented with proteinase inhibiLi et al., Sci. Signal. 10, eaah4323 (2017)

26 September 2017

tors and phosphatase inhibitors (Beyotime Institute of Biotechnology). Protein yield was quantified using the bicinchoninic acid protein assay. After denaturation, equal amounts of protein were separated by SDS– polyacrylamide gel electrophoresis before wet-transfer onto polyvinylidene difluoride membranes (Amersham Biosciences). Nonspecific binding sites were blocked by incubating the membranes with 5% bovine serum albumin (BSA) in tris-buffered saline (TBS) with 0.1% Tween 20 (TBS-T) for 1 hour, which were rinsed and incubated with primary antibody solutions (1:500 for Tim-3; 1:2000 for ID2, BLIMP1, GATA3, and EOMES; 1:1000 for STAT6, pSTAT6, pERK1/2-Thr202/Tyr204, pNF-kB p65, pJNK-Thr183/Tyr185, pp38, pAKT, ERK1/2, NF-kB p65, JNK, p38, and AKT; and 1:3500 for GAPDH) diluted in blocking buffer (5% BSA and 1× TBS-T) overnight at 4°C. All of the antibodies used for Western blotting analysis were purchased from Cell Signaling Technology. Primary antibodies were removed by washing the membranes four times in TBS-T, and the membranes were then incubated for 1 hour with horseradish peroxidase–conjugated secondary antibody (1:5000). After three washes with TBS-T, immunopositive bands on the blots were visualized with the enhanced chemiluminescence detection system (Merck Millipore). Quantitative real-time PCR Total RNA was extracted from pNK cells with TRIzol Reagent (Invitrogen) and then reverse-transcribed into first-strand complementary DNA (cDNA) (RR036A, TaKaRa Biotechnology) according to the manufacturer’s instructions. The synthesized cDNA was amplified with specific primers (Sagon) and SYBR Green (TaKaRa Biotechnology) with ABI PRISM 7900 Sequence Detection System (Applied Biosystems). Triplicate samples were examined in each condition. A comparative threshold cycle (CT) value was normalized for each sample by the 2−DDCT method. NK cell cytotoxicity assay The cytotoxicity activity of NK cells was determined with the CytoTox 96 Non-Radioactive Cytotoxicity Assay Kit (G1780; Promega) according to the manufacturer’s instructions. pNK cells (effector cells; 100 ml) at concentrations of 1.0 × 106/ml, 5.0 × 105/ml, 2.5 × 105/ml, and 1.3 × 105/ml were mixed with 100 ml of HTR-8/SVneo cells (target cells) at a concentration of 2.0 × 104/ml, resulting in E:T ratios of 50:1, 25:1, 12.5, and 6.5:1, respectively. The percentage of cytotoxicity was calculated after correcting for lactate dehydrogenase (LDH) release from HTR-8/SVneo cells using the following formula: Percentage cytotoxicity = 100 × (Experimental values − Culture medium background)/(Maximum LDH release − Culture medium background). Experimental values represent LDH release activity from cocultures of effector and target cells. LDH culture medium background represents LDH release from trophoblasts. Maximum LDH release represents LDH release from trophoblasts that were lysed by sonication. Murine models of pregnancy Male DBA/2 and female CBA/J mice (8 to 10 weeks old) were purchased from Beijing HFK Bioscience Co. Male Balb/c mice (8 to 10 weeks old) were purchased from the Department of Laboratory Animal Science, Fudan University. All animals were kept under specific pathogen–free conditions. All of the experimental procedures involving animals were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (China), and permission was approved by the Human Research Ethics Committee of Obstetrics and Gynecology Hospital of Fudan University. CBA/J × Balb/c 12 of 15

Downloaded from http://stke.sciencemag.org/ on September 26, 2017

(fig. S6D). In addition, Ndhlovu et al. (28) demonstrated that Tim-3 marks human NK cell maturation. We speculate that Tim-3 might be a maturation marker of NK cells in both humans and mice. Treating mice with RMT3-23, the antibody against Tim-3, had a similar effect to that of the administration of antibodies to either NK1.1 or asialo-GM1. These two antibodies are usually used in NK cell depletion assays. That is, RMT3-23 administration may inhibit the maturation of NK cells. Because mature CD11b SP NK cells are the major subset of peripheral blood NK cells (accounting for 70 to 80% of the total) in mice and nearly 80% of CD11b SP NK cells are Tim-3+ (fig. S6, B and C), RMT-23 led to a marked reduction in the number of pNK cells. On the basis of these results, we suggest that Gal-9–Tim-3 signaling not only regulates pNK cell function but also influences pNK cell development. In conclusion, we identified a distinct NK cell subpopulation, Tim-3+ NK cells, with immunosuppressive activity in early pregnancy (fig. S9). The reduced abundance of Tim-3 on NK cells was accompanied by disordered anti- and pro-inflammatory cytokine profiles in patients with RM. Moreover, the Treg differentiation–inducing activity of Tim-3+ NK cells was attenuated in patients with RM. Together, these results suggest a protective role for this NK cell subset in pregnancy. Currently, the most commonly used biomarkers (serum hCG and progesterone) are regarded as ineffective in predicting pregnancy outcome (7). Thus, an accurate and reliable diagnostic biomarker is urgently needed to reflect the maternal-fetal tolerance status and to enable preventative intervention. The Tim-3+ NK cells defined in our study may serve as biological markers during early pregnancy for predicting the occurrence of RM. Our findings also underscore the important functional diversity that exists within human NK cell subpopulations and provide insights into the molecular mechanisms by which NK cells mediate immune tolerance. Because the embryo expresses paternal antigens foreign to the mother, it has been regarded as an allograft or a pseudotumor. Paradigms from our study may also be useful for investigating transplantation immunology and tumor immunology.

SCIENCE SIGNALING | RESEARCH ARTICLE mating combinations (NP model) were established as follows: female CBA/J mice were mated in natural cycling with male Balb/c mice. CBA/J × DBA/2 mating combinations (AP model) were established as follows: female CBA/J mice were mated in natural cycling with male DBA/2 mice. Detection of a vaginal plug was chosen to indicate day 0.5 of gestation. Pregnant females received three injections of RMT3-23 (anti–Tim-3 antibody) intraperitoneally at doses of 500, 250, and 250 mg at days 3.5, 5.5, and 7.5, respectively. In other instances, pregnant mice were injected with rmGal-9 intraperitoneally at a dose of 100 mg at days 3.5, 5.5, and 7.5, respectively. The percentage of resorbed embryos was calculated as follows: resorbed embryos/total embryos × 100.

Expression profiling by microarray Tim-3+ pNK cells and Tim-3− pNK cells from six women with NP were sorted by FACS. Total RNA was directly isolated with TRIzol according to the manufacturer’s instructions. RNA was analyzed on a NanoDrop spectrophotometer (Thermo Fisher Scientific) and a Bioanalyzer (Agilent) to determine RNA yield and integrity. RNA was subjected to one round of amplification and biotinylation with the Amino Allyl MessageAmp II aRNA Amplification Kit (Thermo Fisher Scientific). Biotinylated RNA was hybridized to human genome HOA 6.2 chips. Gene expression analysis was performed using the Whole Genome OneArray Microarray (Phalanx Biotech Group). The log2 ratio of each probe was calculated by pairwise combination and error-weighted average. Probes showing a log2 ratio of ≥1 or ≤−1 and a P value of