Chemokine Gene Silencing in Decidual Stromal Cells Limits T Cell Access to the Maternal-Fetal Interface

See allHide authors and affiliations

Science  08 Jun 2012:
Vol. 336, Issue 6086, pp. 1317-1321
DOI: 10.1126/science.1220030

Keeping Baby Safe

Because half the genes from a developing fetus are inherited from the father, from the mother's perspective, a fetus is “foreign.” How, then, does the maternal immune system tolerate the fetus? Nancy et al. (p. 1317) found that careful regulation of cellular recruitment signals allowed for fetal tolerance in mice. Although high numbers of T cells localized to the myometrium layer of the uterine wall in pregnant mice, very few T cells were found in the decidua, the uterine tissue that encapsulates the fetus and placenta. Thus, in pregnancy, regulation of immune cell localization may allow for organ-specific immune tolerance.


The chemokine-mediated recruitment of effector T cells to sites of inflammation is a central feature of the immune response. The extent to which chemokine expression levels are limited by the intrinsic developmental characteristics of a tissue has remained unexplored. We show in mice that effector T cells cannot accumulate within the decidua, the specialized stromal tissue encapsulating the fetus and placenta. Impaired accumulation was in part attributable to the epigenetic silencing of key T cell–attracting inflammatory chemokine genes in decidual stromal cells, as evidenced by promoter accrual of repressive histone marks. These findings give insight into mechanisms of fetomaternal immune tolerance, as well as reveal the epigenetic modification of tissue stromal cells as a modality for limiting effector T cell trafficking.

Besides being essential for reproductive success, the ability of the allogeneic fetus and placenta to avoid rejection by the maternal immune system during pregnancy (i.e., fetomaternal tolerance) has served as a paradigm for the study of organ-specific immune tolerance (1). Recent work on this problem has made use of a mouse mating system in which wild-type females are crossed with males hemizygous for the Act-mOVA transgene (2) to generate concepti expressing a transmembrane form of the model antigen chicken egg ovalbumin (OVA) from the ubiquitously active β-actin promoter (3, 4). As early as embryonic day 7.5 (E7.5), OVA is expressed at high levels by trophoblasts directly contacting the uterus (i.e., at the maternal/fetal interface) (3), thus exposing maternal tissue to a surrogate fetal/placental antigen that should in principle allow for T cell priming and render the conceptus susceptible to attack by antigen-specific cytotoxic T lymphocytes (CTLs).

This mating system has been used to show that fetal rejection is in part prevented by mechanisms that minimize the activation of naïve T cells with fetal/placental specificity (3, 5). However, antigen-specific fetal loss still does not occur when systemic antifetal/placental CTL activity is experimentally induced in late gestation (3). Thus, a fail-safe mechanism also exists to protect the conceptus from activated CTLs. To visualize this phenomenon more directly, we asked whether pregnant mice, immunized with soluble OVA before mating, would show Act-mOVA–specific fetal loss on E10.5 after being rechallenged with OVA plus the adjuvant combination of agonistic CD40 antibodies plus polyinosinic:polycytidylic acid [poly(I:C)] on E5.5 (6). We studied this period of early gestation because the behavior of OVA-specific T cells would not be influenced by the systemic release of fetal/placental OVA, which starts on ~E10.5 (3). Strikingly, pregnant mice bore the expected Mendelian proportion of Act-mOVA+ concepti (17 of 26 total embryos from n = 3 pregnant mice), even though 14 to 20% of splenic CD8 T cells were OVA-specific at the time the mice were killed (fig. S1). Thus, fetal rejection does not occur even when memory T cells with known fetal/placental specificity are reactivated in early pregnancy.

To find possible explanations for this observation, we evaluated the distribution of reactivated memory T cells in the uteri of C57BL/6-mated mice. Consistent with the ability of effector T cells to infiltrate peripheral tissues even in the absence of a localized antigen source (7, 8), E8.5 mice rechallenged with OVA plus adjuvant on E5.5 showed large numbers of CD3+ T cells distributed throughout the segments of myometrium (and associated submyometrial stroma) overlying each implantation site (Fig. 1, A and B), as well as throughout the myometrium and endometrium of the undecidualized uterine segments between implantation sites (i.e., interimplantation sites) (Fig. 1, C and D). In contrast, CD3+ cells in the decidua appeared sparse (Fig. 1B), with tissue densities remaining at levels similar to those seen throughout the uteri of mice that were not OVA-rechallenged. CD3+ cells in the decidua were most prominent within blood vessels; however, most were extravascular in the implantation site–associated myometrium and interimplantation sites (Fig. 1, E and F). Together, these results suggested that the decidua had a reduced capacity for T cell accumulation, possibly due to an intrinsic inability to recruit T cells from the blood. Accordingly, reactivated memory T cells were also unable to infiltrate decidual tissue encapsulating OVA-expressing concepti (fig. S2) or in hormonally pseudopregnant females with oil-induced artificial deciduomas.

Fig. 1

The decidua resists infiltration by reactivated memory T cells. (A to D) B6CBAF1/J (H-2b/k) females were given 3 × 105 OVA-specific OT-I CD8+ T cells to maximize potential T cell accumulation and were immunized with OVA protein 2 to 3 weeks before mating to C57BL/6 males. On E5.5, pregnant mice either received no additional treatment [(A) and (C)] or were intravenously injected with 0.5 mg OVA and adjuvant [CD40 antibodies plus poly(I:C)] [(B) and (D)]. The mice were killed on E8.5, and tissue immunostaining using CD3-specific antibodies (red) was performed on cross sections of implantation sites [(A) and (B)] and interimplantation sites [(C) and (D)]. The two poles of the decidua (mesometrial and antimesometrial) are indicated; 4′,6-diamidino-2-phenylindole (DAPI) counterstain. (E and F) The decidual/myometrial border (E) and undecidualized endometrium (F) of a rechallenged mouse. Blood vessels (arrowheads) are identified by the presence of red blood cells, which appear green as an artifact of the immunostaining protocol. myo, myometrium; dec, decidua; asterisk, uterine lumen. Data are representative of three independent experiments (at least n = 23 implantation sites each per group).

Because antibodies to CD40 plus poly(I:C) induce a type 1–polarized T cell response (9), we investigated whether impaired decidual T cell infiltration was due to low expression of T helper 1 (Th1)/T cytotoxic 1 (Tc1) chemoattractants. Notably, 6 hours after the intravenous injection of antibodies to CD40, poly(I:C), and endotoxin-contaminated OVA, a regimen expected to increase blood levels of the proinflammatory cytokines tumor necrosis factor-α (TNFα) and interferon-γ (IFN-γ) (10, 11), high levels of the key Th1/Tc1–attracting chemokine CXCL9 (a CXCR3 ligand) (12) were apparent in the segments of myometrium overlying each E8.5 implantation site (Fig. 2, A and B). In contrast, much lower CXCL9 expression was apparent in the decidua. High CXCL9 expression was also induced in both the endometrium and myometrium of the undecidualized uteri of pseudopregnant females (fig. S3), with the vast majority of the expressing cells being CD45- stromal cells (Fig. 2C). Because endometrial stromal cells (ESCs) are the precursors of decidual stromal cells (DSCs), the major cell type of the decidua, these results suggested that the developmental process of decidualization reduced the cells’ capacity to produce T cell chemoattractants under inflammatory conditions.

Fig. 2

The decidua produces low levels of Th1/Tc1–attracting chemokines in response to inflammation. (A to C) E8.5 pregnant [(A) and (B)] or pseudopregnant (C) B6CBAF1/J mice were either untreated (A) or injected with CD40 antibodies, poly(I:C), and OVA [(B) and (C)] 6 hours before they were killed. [(A) and (B)] CXCL9 immunostaining (red) of implantation site cross sections. DAPI counterstain. (C) Double CXCL9 (green)/CD45 (red) immunostaining of the endometrium of an inflamed pseudopregnant uterus. CXCL9 immunoreactivity appears largely confined to the endoplasmic reticulum, giving a punctate appearance. Data are representative of at least three independent experiments. (D) Quantitative real-time fluorescence polymerase chain reaction analysis of cultured MSCs and DSCs. Cytokines were added for the last 6 hours of a 24-hour total culture period. Data show mean ± SEM of four independent experiments. n.s., not significant. (E) Migration of in vitro differentiated Th1 cells to supernatants collected from MSCs or DSCs treated as indicated over the entirety of a 24-hour culture period. Data show mean ± SEM of three independent experiments. (F) Effect of CXCR3 desensitization (via pre-incubation of the Th1 cells with CXCL9) or CCL5 neutralization on Th1 cell migration to supernatants from TNFα+IFN-γ–treated MSCs. Nonspecific rat immunoglobulin G (IgG) antibodies served as the control for CCL5 antibodies. Data show mean ± SEM of three independent experiments.

We next independently prepared highly enriched stromal cells from E7.5 artificial deciduomas and overlying myometrium (fig. S4) and evaluated their inflammatory response in vitro (Fig. 2D). As with other cell types (13, 14), myometrial stromal cells (MSCs) treated with a combination of TNFα and IFN-γ showed synergistic mRNA induction of both Cxcl9 and Cxcl10, which encodes a second CXCR3 ligand. Furthermore, Ccl5, whose product CCL5 (RANTES) has also been implicated in Th1/Tc1 recruitment to inflamed tissues (12), was up-regulated ~15-fold in TNFα-treated MSCs, with expression further augmented by IFN-γ. In contrast, Ccl5 and Cxcl9 transcript levels in DSCs remained unchanged after cytokine treatment, whereas TNFα+IFN-γ mildly induced Cxcl10 expression to levels that barely exceeded those of MSCs at baseline. The expression pattern of Cxcl11, which encodes the third known CXCR3 ligand, was similar to that of Cxcl9 and Cxcl10 (fig. S5). The inability of DSCs to produce Th1 chemoattractants was functionally confirmed using transwell migration assays, which furthermore showed that MSCs attract Th1 cells through the induction of CXCR3 ligands and CCL5 (Fig. 2, E and F). Together, these results suggested that the inability of DSCs to produce T cell–attracting chemokines under inflammatory conditions in vivo was due to a cell-intrinsic defect in their inflammatory cytokine response.

The inability of DSCs to produce CXCR3 ligands and CCL5 was not explained by decreased activation of NF-κB or STAT1, the major transcription factors mediating TNFα and IFN-γ signaling (fig. S6, A and B). Moreover, we could find examples of NF-κB and STAT1 target genes that were induced in DSCs in a relatively robust fashion (fig. S6C). These results suggested that the DSC chemokine expression defects were gene-specific and independent of inflammatory signaling per se. We therefore evaluated gene-specific chromatin configurations using chromatin immunoprecipitation (ChIP) assays. Elevated basal levels of the repressive histone H3 trimethyl lysine 27 (H3K27me3) mark (15) were present on the Cxcl9 and Cxcl10 promoters in DSCs as compared to MSCs (Fig. 3A, top row). Conversely, TNFα+IFN-γ treatment increased Cxcl9/10 promoter levels of acetylated histone H4 (H4Ac), a mark of active gene transcription, in MSCs but not DSCs (Fig. 3A, bottom row). Both cell populations showed the inverse patterns of H3K27me3 and H4Ac occupancies on the Gapdh and Cd8a promoters expected from these genes’ respectively high and low constitutive expression levels. Thus, low cytokine inducibility of Cxcl9/10 in DSCs was associated with the gene-specific presence of the repressive H3K27me3 histone mark.

Fig. 3

Chromatin configurations in uterine cells and tissue layers. (A) Ex vivo ChIP assays performed on cultured MSCs and DSCs. Cytokines were added for the last 6 hours of a 24-hour total culture period. Data show mean ± SD of five independent experiments. (B) In vivo ChIP assays performed on dissected uterine tissues and tissue layers. Deciduas were dissected free of embryos. Data show mean ± SD of three independent experiments.

In vivo ChIP assays performed directly on dissected E7.5 uterine tissue layers also revealed high levels of the H3K27me3 mark on the Cxcl9/10 promoters in whole decidua as compared with overlying myometrium, thus demonstrating that Cxcl9/10 silencing was also a feature of true, pregnancy-associated decidua in vivo (Fig. 3B). Recognizing that undecidualized uteri are comprised equally in volume by endometrium and myometrium (5), this result also meant that in vivo ChIP assays could be used to infer chromatin configurations in undecidualized ESCs, because these cells could not be sufficiently purified for ex vivo assays. Accordingly, Cxcl9/10 H3K27me3 promoter occupancies in whole nonpregnant uteri and E7.5 interimplantation sites were similar to those in segments of implantation site–associated myometrium and clearly not intermediate between the myometrium and decidua (Fig. 3B). This result strongly suggested that the H3K27me3 modification of the Cxcl9/10 promoters appears upon transformation of ESCs into DSCs.

The Ccl5 promoter also showed increased H3K27me3 occupancy in whole decidua as compared with myometrium and undecidualized uterus (Fig. 3B), suggesting a shared pathway for minimizing decidual chemokine expression. Interestingly, this increase was not readily apparent ex vivo, while H3K27me3 levels on the Cxcl9/10 promoters also appeared somewhat reduced upon TNFα+IFN-γ treatment (Fig. 3A). These results suggest some level of reversibility of the H3K27me3 mark at the locations we assessed by ChIP, possibly as a result of just isolating and culturing the cells. Because H4Ac levels on the Cxcl9/10 and Ccl5 promoters were nonetheless unchanged in TNFα+IFN-γ–treated DSCs (Fig. 3A), it is likely that the continued repressed status of these genes after 24 hours culture also involves either the presence of the H3K27me3 mark in other regions of their respective loci or the presence of other repressive modifications.

We next determined the effect of ectopic chemokine expression within the decidua by injecting artificial deciduomas in OVA-rechallenged mice with control, Cxcl9, or Ccl5-expressing lentiviruses mixed with enhanced green fluorescent protein (EGFP) reporter lentiviruses. For each mouse, decidual CD3+ T cell densities were normalized to myometrial CD3+ cell densities to account for systemic differences in the magnitude of the anti-OVA T cell response. CD3+ cell densities in the GFP+ decidual areas of control virus–injected mice were elevated as compared to uninfected, GFP decidual areas, thus revealing a nonspecific effect of viral infection per se. However, these densities were not further altered in mice injected with Cxcl9- or Ccl5-expressing viruses. In contrast, CD3+ cell densities in decidual areas infected with mixtures of Cxcl9- and Ccl5-expressing viruses were significantly elevated compared with control virus–infected areas (Fig. 4). The synergistic effect of dual CXCL9/CCL5 expression thus apparent is consistent with recent observations in a malignant melanoma model (16). Furthermore, because CXCL9 expression in Cxcl9+Ccl5-infected decidual areas was undetectable by tissue immunostaining (fig. S7), it was unlikely that T cell infiltration into these areas was due to supraphysiologic chemokine expression. Together, these results suggested that inadequate endogenous expression of CXCR3 ligands and CCL5 was limiting for decidual T cell accumulation.

Fig. 4

Effect of ectopic chemokine expression on effector T cell accumulation within the decidua. Mice were immunized with OVA 2 to 3 weeks before mating, rechallenged with OVA plus adjuvants on E4.5, injected with lentivirus on E5.5, and killed on E7.5. Viral preparations included samples of EGFP reporter lentiviruses so that T cells within transduced areas could be identified on anti-CD3 and anti-GFP immunostained serial sections. (A) CD3+ cell densities in infected (GFP+) and uninfected (GFP) decidual areas relative to myometrial CD3+ cell densities. Data show mean ± SEM of at least three independent experiments encompassing n = 4 control (empty vector) virus–infected mice (75 × 107 virus particles each); n = 5 Cxcl9+Ccl5 virus–infected mice (37.5 × 107 particles each); and n = 3 each of mice infected with Cxcl9- or Ccl5-expressing viruses alone (75 × 107 particles each). Myometrial CD3+ cell densities were not significantly different between the four groups (average: 0.037 cells per μm2). (B to K) GFP or CD3 immunostaining (as indicated, red) and DAPI counterstain (blue) of representative serial sections of decidualized uteri infected with Cxcl9+Ccl5 viruses [(B) and (C) and (H) to (K)] or control viruses [(D) to (G)]. [(B) and (C)] Asterisks denote different areas of the decidual lumen, which frequently contained GFP+ cells, CD3+ cells, and granulocytes scattered among necrotic debris; the dashed line indicates the border between the myometrium and decidua. [(D) to (K)] The dashed lines show the perimeters of infected areas used to calculate CD3+ cell densities. Asterisks show areas excluded from analysis because of their necrotic appearance. Arrow, decidual lumen. (H) to (K) are close-ups of (B) and (C).

Provocatively, T cells have been reported to be relatively scarce in the human decidua (17, 18), implicating the developmental program of decidual chemokine silencing described here as a potentially conserved mechanism of fetomaternal tolerance. Consistent with this possibility, Cxcl10 is expressed only focally in the human decidua, in association with periglandular leukocyte aggregates (19). As the presence of even low numbers of activated T cells at the maternal/fetal interface might disturb placental development or function, dysregulation of this pathway might also contribute to a variety of pregnancy complications. Conversely, altered chemokine silencing may influence the susceptibility of the decidua to infection. More generally, however, our results demonstrate that genes encoding Th1/Tc1–attracting chemokines are subject to epigenetic regulation in tissue stromal cells and that such regulation can significantly influence a tissue’s capacity for T cell accumulation. This demonstration raises questions regarding how the repressive H3K27me3 histone mark is targeted to select chemokine genes and whether related pathways control T cell access to the stroma of infected, autoimmunity-afflicted, or cancer-bearing tissues.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

References (2029)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank S. K. Dey and K. Johnson for advice, and A. Frey and J. Ernst for comments on the manuscript. The Histopathology and Vaccine and Cell Therapy core facilities of the New York University Cancer Institute provided histology services and tetramer reagents and were supported by NIH, National Cancer Institute (P30CA016087). P.N., E.T., and C.-S. T. performed experiments; P.A. and D.E.L. provided critical expertise and reagents; P.N. and A.E. analyzed data; and P.N. and A.E. designed experiments and wrote the manuscript. The data reported in the manuscript are tabulated in the main paper and in the supplementary materials. This work was supported by grants from NIH (RO1AI062980) and the American Cancer Society to A.E. (RSG-10-158-01-LIB).

Stay Connected to Science

Navigate This Article