Decidual T Cells Exhibit a Highly Differentiated Phenotype and Demonstrate Potential Fetal Specificity and a Strong Transcriptional Response to IFN

Successful pregnancy in eutherian placental mammals requires maternal immunological tolerance of the developing semiallogeneic fetus (1). Improved understanding of the physiological mechanisms that underlie this immune regulation is likely to provide insights into the etiology of pregnancy complications and may impact studies of immunological tolerance in the setting of transplantation and cancer.

The classical model of immune regulation during human pregnancy has been based on a relative shift in the maternal immune response from an inflammatory Th1 cytokine pattern to a Th2 profile (2). However, it is now recognized that the complex immunological interactions at the maternal interface cannot be explained with this simple binary classification. Indeed, villous implantation and invasion (3) are facilitated by an inflammatory environment, and maternal T cell function exhibits great diversity and plasticity (4).

The formation of the human hemochorial placenta involves the invasion of fetal extravillous trophoblast (EVT) cells, which remodels maternal spiral arteries and reduces their resistance to blood flow. Prior to and during this process, the maternal uterine endometrium is transformed into a layer termed the decidua (5). In early pregnancy, the decidua accumulates large numbers of specialized uterine NK cells (CD56^brightCD16⁻), and their interaction with EVT plays a key role in successful placentation (6, 7). However, as pregnancy progresses the number of uterine NK cells declines and by the third trimester T lymphocytes become the predominant leukocyte population (8). Despite this, the role that decidual T cells play in modulating the uterine environment and their potential recognition of the fetus remain controversial questions.

Maternal T cells within decidua are likely to make direct anatomical interactions with EVT and would therefore become exposed to fetal Ag. However, EVT does not express HLA-A, HLA-B, or HLA class II alleles, although it does retain HLA-C expression (9), and significantly higher levels of activated T cells and T regulatory cells are induced within decidua of HLA-C mismatched pregnancies (10). Murine models suggest that fetal proteins are presented to the maternal immune system indirectly by maternal APCs (11). Indeed, large numbers of fetal trophoblast cells (and fragments) are shed into the maternal circulation during normal pregnancy and provide a rich supply of fetal and placental Ags to the maternal immune system. Conversely, maternal dendritic cells appear to be limited in their ability to migrate from the pregnant uterus (12), and epigenetic silencing of key chemokines in the decidual stroma may limit T cell access to the decidua (13).

In human pregnancy, maternal CD8⁺ T cells with specificity for fetal Ags are detectable in maternal peripheral blood both during (14) and after (15–17) pregnancy. Studies of T cell biology directly within human decidua are more limited and the antigenic specificity of these cells is unclear. Effector memory CD8⁺ cells have been demonstrated in this setting and shown to express low levels of perforin and granzyme (18).

The mechanisms by which decidual T cells are regulated are unclear and may depend on the potential of these cells to gain anatomical access to fetal tissue. Large numbers of T regulatory cells are generated during pregnancy (19, 20), and murine studies show that these are essential for fetal survival (21). Inhibitory checkpoint proteins such as PD-1 and T cell Ig and mucin domain containing 3 (Tim-3) play a critical role in murine models of transplantation and pregnancy (22). Ab-mediated blockade results in allograft and fetal rejection (23, 24). Expression of these inhibitory checkpoint proteins has been demonstrated on human decidual CD8⁺ T cells (25) whereas programmed death ligand 1 (PD-L1) is also expressed in human decidua (22, 26).

The objectives of this study were to perform a detailed phenotypic and transcriptional assessment of both decidua-derived CD4⁺ and CD8⁺ T cells to comprehensively characterize these tissue-derived subsets and their antigenic specificity. Our findings show that decidua is enriched with activated effector T cells that demonstrate an unusual transcriptional profile dominated by IFN-responsive pathways. Moreover, effector T cells within decidua have the potential to recognize fetal tissue, and this functional response is actively suppressed by local T regulatory cells.

In total, 90 placentae were obtained from healthy mothers undergoing third trimester elective cesarean section. Peripheral blood and cord blood samples were also obtained at this time. Decidua basalis was dissected and dissociated via gentleMACS (Miltenyi Biotec) and enzymatically digested using 1 mg/ml collagenase (Sigma-Aldrich) and 200 U/ml DNase (Sigma-Aldrich) at 37°C for 30 min. Digested tissue was then filtered, washed using RPMI 1640 (Sigma-Aldrich), and mononuclear cells were isolated using density gradient medium Lymphoprep (Alere). Maternal and cord blood mononuclear cells were similarly separated using density gradient.

To phenotype T cell memory status, 1 × 10⁶ fresh maternal or decidual mononuclear cells (n = 45) were stained with: CD3 AmCyan (clone SK7; Becton Dickinson); CD4 PE (clone RPA-T4; Cambridge Bioscience); CD8 eFluor 450 (clone OKT8; eBioscience); CD45RA Alexa Fluor 700 (clone HI100; BioLegend); CCR7 FITC (clone 150503, Becton Dickinson); CD27 allophycocyanin–eFluor 780 (clone 0323; eBioscience); CD28 PerCP-Cy5.5 (clone L293; Becton Dickinson); CD57 allophycocyanin (clone HCD57; BioLegend); CD14 ECD (clone RM052; Beckman Coulter); and CD19 ECD (clone J3-119; Beckman Coulter). Samples were then analyzed using LSR II (Becton Dickinson).

To phenotype intracellular cytokines and T regulatory cells, prior to Ab staining, fresh maternal and decidual monocular cells (n = 35) were incubated in RPMI 1640 supplemented with l-glutamine (Sigma-Aldrich) and bovine FCS (Sigma-Aldrich), and 50 ng/ml PMA (Sigma-Aldrich), 1 μg/ml ionomycin (Sigma-Aldrich), and 1.25 μg/ml monensin (Sigma-Aldrich) were added for 4 h. Unstimulated samples were set up in parallel. Samples were surface stained with: CXCR3 PE-Cy7 (clone IC6; Becton Dickinson); CD25 FITC (clone IM0478U; Beckman Coulter); CD127 eFluor 780 (clone ebioRDR5, eBioscience); CD8 Brilliant Violet 510 (clone RPA-T; BioLegend); CD4 PerCP-Cy5.5 (clone RPA-T8; BioLegend); CD3 Brilliant Violet 605 (Becton Dickinson); CD14 ECD (clone RM052; Beckman Coulter); CD19 ECD (clone J3-119; Beckman Coulter); and Live/Dead fixable red stain dye (Life Technologies). Following surface staining, samples were permeabilized and fixed using Fix/Perm solution and buffer (eBioscience). Samples were intracellularly stained with: IL-4 allophycocyanin (clone 8D4-8; BioLlegend); IL-17 Pacific Blue (clone BL168; BioLegend); FOXP3 PE (clone PCH101; eBioscience); and IFN-γ Alexa Fluor 700 (clone 4S.B3; BioLegend).

Characterization of checkpoint marker phenotypes in maternal and decidual samples (n = 8) was out carried by staining with: CD3 AmCyan (clone SK7; Becton Dickinson); CD8 allophycocyanin-Cy7 (clone SK1; BioLegend); CCR7 FITC (clone 150503; Becton Dickinson); CD45RA Alexa Fluor 700 (clone HI100; BioLegend); CD14 ECD (clone RM052; Beckman Coulter); CD19 ECD (clone J3-119; Beckman Coulter); PD-1 Brilliant Violet 421 (clone EH2.2H7; BioLegend); CTLA-4 PE-Cy5 (clone BNI3; Becton Dickinson); Tim-3 PE-Cy7 (clone F382E2; BioLegend); and lymphocyte activation gene 3 (Lag-3; clone FAB2319P; R&D Systems).

Fresh maternal and decidual mononuclear cells (n = 10) were prepared as previously described. T -cell enrichment was performed using Easy Sep T cell enrichment Ab mix, magnetic beads and Easy Sep Purple Magnet separation (StemCell Technologies). The T cell enriched suspensions were then split between non CD4⁺ T regulatory depleted and T regulatory depleted samples. Suspensions that were retained for T regulatory depletion were depleted using CD25 MicroBeads (Miltenyi Biotec) with magnetic MACS microcolumn separation. To trace proliferation of cells in the assay, 1 μl of CellTrace Violet dye (Invitrogen) was added per 10⁶ cells and incubated at 37°C for 20 min.

Cord blood mononuclear cells were isolated using a method previously described. Five milliliters of RBC lysis buffer was added to each sample and cell suspensions were then centrifuged and resuspended in 20 ml of enriched media. Cord blood lymphocytes were then irradiated at 3000 rad.

Irradiated cord blood lymphocytes were added at a 2:1 ratio to each maternal and decidual sample in a 96-well round-bottom plate in enriched media. Each sample was run in duplicate and T regulatory cell–depleted samples were also run in parallel. Positive control samples were established with the addition 5 μl of CD3 Dynabeads (Becton Dickinson) instead of cord blood. Negative controls were established in only enriched media. To each sample, 10 U of IL-2 cytokine was added. Samples were incubated for 4 d. At day 3, 100 μl of media per well was replaced with fresh media along with 10 U of IL-2.

Prior to harvesting on day 4, 1.25 μg of monensin was added to each sample and incubated for a further 4 h. Each sample was harvested and washed with PBS. Surface staining was performed using: CD3 Brilliant Violet 605 (clone SK7; BioLegend); CD8 Brilliant Violet 510 (clone RPAT-8; BioLegend); CD4 PerCP-Cy5.5 (clone RPAT-4; BioLegend); and Live/Dead fixable red stain dye (Life Technologies). Samples were permeabilized and fixed using 100 μl of 4% PFA (Sigma-Aldrich) and 10 μl of 4% saponin (Sigma-Aldrich). Intracellular staining was performed using: IL-10 PE (clone JES3-9D7; eBioscience); IL-4 allophycocyanin (clone 8D4-8; BioLegend); IL-17 allophycocyanin-Cy7 (clone BL168; BioLegend); IFN-γ Alexa Fluor 700 (clone 4S.B3; BioLegend); granzyme B FITC (clone GB11; BioLegend); and perforin PE-Cy7 (clone dG9; eBioscience). Samples were washed then resuspended in MACS before flow cytometry was performed.

Mononuclear cells from the decidua basalis and maternal peripheral blood were prepared as previously described (n = 9) from pregnancies that resulted in male offspring. Additional HY control samples (n = 5) were obtained from pregnancies that resulted in female offspring. Patient samples were selected based on MHC A2 and B7 positivity. Samples were incubated with HY PE dextramer (Immundex, Denmark), either A2 (FIDSYICQV) or B7 (SPSVDKARAEL) specific, for 20 min prior to a full flow cytometer staining protocol. Abs used were: Live/Dead fixable red stain dye (Life Technologies); CD8 eFluor 450 (clone OKT8; eBioscience); CD3 AmCyan (clone SK7; Becton Dickinson); PD-1 PE-Cy7 (clone EH12.2H7; BioLegend); CD69 Alexa Fluor 647 (clone B176236; BioLegend); CD45RA Alexa Fluor 700 (clone HI100; BioLegend); and CCR7 FITC (clone 150503; Becton Dickinson).

Twelve samples selected for the microarray consisted of six matched maternal and decidual mononuclear cells prepared using previously stated methods. Prior to flow cytometric cell sorting, the following Abs were applied: CD3 allophycocyanin (clone UCHT1; Invitrogen); Live/Dead fixable red stain dye (Life Technologies); CD4 PE-Cy7 (clone RPAT4; eBioscience); CD8 PerCP-Cy5.5 (clone RPAT8; eBioscience); CD45RA PE (clone HI100; BioLegend); and CCR7 FITC (clone 150503; Becton Dickinson). Per maternal and decidual samples, cells were sorted into separate CD4 and CD8 effector memory samples. RNA was immediately isolated and extracted using Qiagen RNeasy Plus mini and micro kits. RNA samples were immediately stored at −80°C. In total, 24 RNA samples were sent to Eurofins Genomics services. An Affymetrix Pico labeling kit was used to amplify RNA, and the Affymetrix human gene 2.0 ST array was used to carry out the transcriptome analysis. Microarray data were deposited in ArrayExpress under accession number E-MTAB-5517 (http://www.ebi.ac.uk/arrayexpress).

Flow cytometry data were analyzed using FACSDiva (Becton Dickinson), and statistical analysis was performed using Prism (GraphPad Software). Non-Gaussian distribution was applied to all samples, and therefore a Wilcoxon matched pairs signed rank test was used to identify significance between two sets of data, whereas a Friedman test with a Dunn multiple comparison was used for multiple sets. Microarray data were analyzed with the R Limma package (Bioconductor). Normalization was performed with the Loess (intra-array) and quantile (interarray) methods. An adjusted p value (Benjamini and Hochberg’s method) of ≤0.05 was taken as significant for differences in gene expression. The empirical Bayesian shrinkage of moderated t statistics function (eBayes) was used to generate Bayes factors for significantly upregulated or downregulated genes in decidua compared with maternal peripheral blood.

Volcano plots were generated using R and genes annotated based on significant Bayes value (>5). Gene set enrichment analysis was performed using GSEA and MSigDB (Broad Institute).

Written informed consent was obtained from all women recruited into the study. Matched maternal blood, cord blood, and placenta were collected with the approval of Health Research Authority–West Midlands, Edgbaston Research Ethics Committee (RG_14-194).

To investigate the nature of the decidual T cell infiltrate at term (>37 wk gestation) we obtained decidual tissue collected at the time of elective cesarean section from 45 women who had an uncomplicated pregnancy. A detailed analysis of the phenotype of lymphocytes obtained from the decidua was undertaken and compared with matched maternal peripheral blood. Flow cytometry was used to determine the T cell memory status according to the pattern of CD45RA and CCR7 coexpression (CD45RA⁺CCR7⁺ [naive], CD45RA⁻CCR7⁺ [central memory], CD45RA⁻CCR7⁻ [effector memory], and CD45RA⁺CCR7⁻ [revertant memory]) (Fig. 1A, 1B). Strikingly, the dominant population of T cells within the decidua basalis (and parietalis; Supplemental Fig. 1A, 1B) was the effector memory subset, which comprised 53% of CD4⁺ and 51% of CD8⁺ T cells, respectively. These values were markedly increased compared with maternal peripheral blood where the proportions were 30 and 26% respectively (both p < 0.0001) (Fig. 1C). To investigate the differentiation status of these effector memory cells in more detail, we also examined the pattern of CD27 and CD28 coexpression, which subclassifies effector memory T cells within four subsets (CD27⁺CD28⁺ [EM1], CD27⁺CD28⁻ [EM2], CD27⁻CD28⁻ [EM3], and CD27⁻CD28⁺ [EM4]) (Supplemental Fig. 1C, 1D) (27). Decidual effector memory cells were found to be markedly more differentiated than those within maternal blood, with a particularly high proportion of the EM3 subset (CD27⁻ and CD28⁻). In particular, this subset represented 29 and 33% of decidua-derived CD4⁺ and CD8⁺ effector memory cells, respectively, compared with 1.8 and 14% in peripheral blood (both p < 0.0001). Additionally, expression of CD57, a marker of late differentiation, was also markedly increased on decidual CD4⁺ and CD8⁺ T cells in comparison with maternal peripheral blood (both p = 0.0001) (Supplemental Fig. 2A, 2B).

We then examined the functional capacity of T cells within the two compartments by analysis of intracellular expression of IFN-γ and IL-4 in paired blood and decidual samples (Fig. 1D, 1E). Expression of both of these cytokines was markedly increased in both the CD4⁺ and CD8⁺ decidual T cells compared with peripheral blood. As anticipated, CD4⁺ T cells showed a lower level of IFN-γ expression and a higher degree of IL-4 production compared with the CD8⁺ T cells; an average of 25 and 5.4% of CD4⁺ T cells produced IFN-γ and IL-4 following mitogenic stimulation compared with 60 and 1.2% of CD8⁺ T cells. However, decidual T cells consistently demonstrated increased expression of both cytokines compared with peripheral blood subsets (p = 0.0004 and <0.0001 for IFN-γ and p = 0.0010 and 0.0006 for IL-4, respectively) (Fig. 1F, 1G). To ensure that these results were not due to the increased proportion of effector memory cells residing in the decidua, CD4 and CD8 effector memory T cells were analyzed for IFN-γ and IL-4 expression following mitogen stimulation. Both IFN-γ and IL-4 were significantly increased in CD8 and CD4 effector memory T cells (p = 0.0039 and 0.0078 for IFN-γ and p = 0.039 and 0.0039 for IL-4, respectively). The proportion of T cells that expressed the chemokine receptor CXCR3 was also measured in non–mitogen-stimulated samples (Supplemental Fig. 2C). It was found that CD4 T cells expressing CXCR3 was significantly higher in the decidua, 16.5%, compared with that of maternal PBMCs, 5.6% (p = 0.0001).

The activation of effector T cells is attenuated by the action of T regulatory cells, and the relative balance of these two compartments is important for the maintenance of immune homeostasis. T regulatory cells were identified through a CD3⁺CD4⁺CD25^highCD127⁻ and FOXP3⁺ phenotype (Fig. 2A), and consistent with previous reports (28), increased numbers of T regulatory cells were measured in the decidua, 8.1% of CD4⁺ T cells compared with 3.2% of maternal peripheral blood T cells (p ≤ 0.0001) (Fig. 2B).

Additionally, we also went on to examine potential mechanisms of intrinsic regulation of effector T cell activation. As such, we determined the expression of inhibitory checkpoint proteins on decidual CD4⁺ and CD8⁺ T cells and compared this to maternal peripheral blood (Fig. 2C, 2D). The expression of PD-1, Tim-3, CTLA-4, and Lag-3 was markedly increased on decidual effector memory T cells as compared with effector memory T cells within peripheral blood. This pattern was seen both for CD4⁺ and CD8⁺ T cells and was most marked in the case of PD-1, expressed on 45 and 56% of CD4⁺ and CD8⁺ T cells, respectively. Tim-3 and CTLA-4 expression was very low on peripheral blood CD4⁺ T cells, whereas in the decidua expression approached 10%. Lag-3 also significantly increased from 10 to 17% in decidual CD4⁺ cells. In CD8⁺ T cells, Tim-3 and CTLA-4 were expressed at somewhat higher levels on CD8⁺ T cells within blood compared with CD4⁺ cells. However, again the percentage expression of Tim-3, CTLA-4, and LAG-3 significantly increased from 15, 4, and 8% to 28, 14, and 16%, respectively, on CD8⁺ decidual cells compared with maternal peripheral blood. Boolean analysis was used to examine the pattern of coexpression of inhibitory proteins. Expression of two markers was the most frequent coexpression pattern and was dominated by PD-1. PD-1 was commonly found coexpressed with either Tim-3 (47 and 37%), CTLA-4 (32 and 21%), or Lag-3 (35 and 41%) in CD4⁺ and CD8⁺ T cells, respectively (Fig. 2C, 2D).

Having established that decidual T cells contain a high proportion of highly functional effector cells, we sought to understand whether this pattern is driven by recognition of fetal Ag. Initially, HLA-peptide multimers containing immunodominant peptides derived from HY and restricted by either HLA-A*0201 or HLA-B*0702 were used to identify HY-specific CD8⁺ cells in women during pregnancies with male fetuses (Fig. 3, Supplemental Fig. 3A). As previously reported (14, 15, 16), HY-specific T cells were identifiable within maternal peripheral blood in gender-mismatched pregnancies, although present at very low frequency (0.06%) of the total CD8⁺ T cell population. In contrast, HY-specific CD8 T cells were markedly increased within decidua increasing 10-fold to represent 0.6% of CD8⁺ T cells (p = 0.0313) (Fig. 3B). Additionally, HY-specific CD8 T cells were not detected in the decidua or peripheral blood of pregnancies that resulted in female offspring, which suggests that any HY-specific response in the decidua is due to HY minor histocompatibility Ag present on placental tissue. HY-specific CD8 T cells were predominantly effector memory with very high levels of PD-1 and the activation marker CD69 (52.02 and 65% of total effector memory, respectively) compared with maternal peripheral blood (Fig. 3C, 3D).

We assessed the functional response of the decidual T cell population following exposure to matched, irradiated fetal blood cells using an MLR. The functional response of the decidual T cells was assessed by measurement of proliferation and expression of the cytokines IFN-γ IL-4, IL-10, and the cytotoxicity markers perforin and granzyme (n = 10). Decidual CD4⁺ and CD8⁺ T cells both demonstrated significantly stronger proliferation than did maternal peripheral blood T cells in response to fetal Ag exposure (p = 0.0039 and 0.027, respectively) (Fig. 4B, 4C). To investigate the regulation of this functional response to fetal cells, we went on to deplete CD4⁺ T regulatory cells prior to the MLR assay. CD25^high T cell depletion (Supplemental Fig. 2D) led to a substantial increase in both CD4⁺ and CD8⁺ T cell proliferation in response to stimulation with cord blood cells (p = 0.0020 and 0.0039). Notably, proliferation of both CD4⁺ and CD8⁺ T cells increased by 1.59- and 1.9-fold in comparison with non–T regulatory cell-depleted cultures.

Granzyme production in response to fetal Ag exposure during the MLR significantly increased from 2.79 and 17.15% in peripheral blood CD4⁺ and CD8⁺ T cells to 8.08 and 32% of decidual CD4⁺ and CD8⁺ T cells, respectively (Fig. 4D, 4E). Levels of IL-10 and IL-4 were also markedly increased in decidual CD4⁺ T cells (3.77 and 11.32%, respectively) in comparison with peripheral blood CD4⁺ T cells (1.27 and 3.42%, respectively) following the MLR (Fig. 4F, Supplemental Fig. 3). Following T regulatory cell depletion in MLRs, granzyme was significantly increased in decidual CD4⁺ and CD8⁺ T cells (p = 0.0068 and 0.0010, respectively) and IL-4 was also increased in CD4⁺ decidual T cells in comparison with maternal peripheral blood T cells (p = 0.0039). Overall, these data indicate that decidual CD4⁺ and CD8⁺ T cells include T cells with alloreactive specificity against fetal Ag and that these are controlled both intrinsically by high levels of expression of PD-1 and also through the activity of resident regulatory T cells.

To comprehensively describe the differences between the decidual and peripheral effector memory T cell population, we went on to perform a transcriptional analysis using a microarray of CD4⁺ and CD8⁺ effector memory T cells purified from decidual tissue or matched peripheral blood. Twenty-six and 31 genes were significantly upregulated in the decidual CD4⁺ and CD8⁺ T cells, respectively. Additionally, six genes were shown to be downregulated within decidual CD4⁺ effector T cells (Fig. 5A, 5B).

Eight genes, RSAD2, IFI6, IFI27, IFI44L, IFIH1, EIF2AK2, DDX60, and USP41, were upregulated in both CD4⁺ and CD8⁺ from decidua. Interestingly, these are all associated with the cellular response to IFN signaling. RSAD2, also known as viperin, is an IFN-inducible antiviral protein induced by type I and type II IFN. It also facilitates TCR-mediated GATA3 activation and optimal Th2 cytokine production (29, 30). IFI6, IFI27, IFI44L, and IFIH1 are all IFN-inducible genes that are associated with a type 1 IFN signature, and EIF2AK2 is a dsRNA-dependent protein kinase R that is among the first proteins to become activated following viral infection or stress signaling (31). DDX60, also known as DEAD (Asp-Glu-Ala-Asp) box polypeptide 60, is a cytosolic DNA sensor and RNA helicase that is involved in RLR-mediated type I IFN production after viral infection and mediates viral RNA degradation via the RNA exosome (32). The function of USP41 is unknown but it has high sequence homology with USP18, a ubiquitin-specific peptidase that acts as a negative regulator of IFN-α/β responses (33). Several differences were observed in the transcriptional profile between CD4⁺ and CD8⁺ T cells within decidua, including the OAS1, OAS2, and OAS3 genes, which encode members of the 2-5A synthetize family involved in the innate immune response to viral infection and that were upregulated by ∼2- to 5-fold in CD8⁺ decidual T cells. A specific finding of interest within CD4⁺ T cells was a marked increase in expression of the CLIC1 intracellular chloride channel, which has not previously been reported to be expressed in decidual T cells.

Of the six significantly downregulated genes in CD4⁺ effector memory decidual cells, three are described as noncoding RNAs. Despite an absence of documented functional relevance in T cell immunology, these noncoding RNAs may have a role in epigenetic regulation of gene expression, and the downregulation of these genes may be distinct in T cell activation in the decidua. Indeed, a recent study has suggested several upregulated and downregulated microRNAs have a potential role in the decidua of patients of recurrent miscarriage (34).

The genes encoding Galectin-1 and Galectin-9 demonstrated increased expression in decidual CD4⁺ and CD8⁺ effector memory cells, respectively. Galectin-1 expression has been previously observed on decidual NK cells, and galectin-9 is also present on T regulatory cells in pregnancy (35, 36). As such, this particularly high expression by effector T cells within decidua is further evidence that these proteins are important immune-regulatory molecules during pregnancy (37). Further interrogation of the role of the differentially expressed genes using the Database for Annotation, Visualization, and Integrated Discovery identified several pathways that were significantly upregulated, most notably “immune responses to virus” and “regulation of apoptosis” (Fig. 5C, 5D).

We also examined the combined pattern of gene expression within specific immune response and regulatory pathways (Supplemental Table I). Gene set enrichment analysis demonstrated that gene sets and pathways associated with apoptosis, negative regulation of activation of T cells, and IFN-γ and IFN-α were enriched in decidual T cells. Consistent with our earlier findings, genes encoding checkpoint proteins Lag-3, Tim-3, CTLA-4, and additionally TIGIT were differentially expressed in decidual effector CD8⁺ and CD4⁺ T cells, in comparison with peripheral blood, and contributed to the core enrichment of negative regulation of T cell activation gene sets and pathways within decidual T cells. Several molecules such as TRAIL and Casp3 and genes encoding CCR5, granzyme B, CXCR3, and IFN-γ were also differentially expressed in decidual effector T cells and contributed to the enrichment of apoptosis and allograft response gene sets within decidual T cells, respectively. These findings indicate that CD4⁺ and CD8⁺ T cells at the maternal–fetal interface demonstrate a unique transcriptional profile including a dominant response to IFN signaling and activation of inflammatory pathways, but they also express a wide range of proteins associated with immune regulation.

It is established that lymphoid cells play a pivotal role in eutherian reproduction. However, the role of T lymphocytes (which become the major lymphoid subset within the decidua toward term) is much less well characterized but is clearly important given that T cell dysregulation is associated with adverse pregnancy outcomes, especially disorders of malplacentation such as pre-eclampsia and fetal growth restriction. Understanding T cell function at the maternal–fetal interface (within the decidua) is therefore very important for targeted medical advancements in these conditions. In this study, we have therefore examined the phenotype, function, transcriptome, and specificity of circulating and decidual CD4⁺ and CD8⁺ T cells at third trimester.

It has been previously shown that there is a high concentration of differentiated CD8⁺ T cells in the human decidua during late pregnancy (18) along with T regulatory cells (28, 38). In this study, we have analyzed both CD8⁺ and CD4⁺ subsets, with a comparison with matched maternal blood samples, and had the opportunity to combine this with further functional analysis and novel transcriptional analysis of decidual effector T cells. Increased proportions of both CD4⁺ and CD8⁺ effector T cells were seen in decidual tissue compared with blood, although, importantly, note that this may at least partly reflect the observation that naive T cells are relatively excluded from peripheral tissue. Despite this, several differences were observed in the transcriptional and functional profile of effector cells between blood and tissue. These included increased representation of the CD27⁻CD28⁻ subset with increased cytolytic properties and shortened telomere length (27), suggesting that these T cells may potentially be activated by local alloantigen (18). We also demonstrated that decidual effector memory T cells express high levels of IFN-γ and IL-4, an unusual profile that may potentially relate to local effects of progesterone (39). The immunological balance that is maintained at the maternal–fetal interface is regulated not only by cellular processes distinct to maternal immune cells and fetal tissue, but that of maternal hormonal changes at the maternal–fetal interface. Progesterone in particular is known to induce increased levels of IL-4 in maternal CD4⁺ and CD8⁺ T cells and limit their cytotoxicity (39, 40).

Given this strong functional capacity, we were particularly interested in the mechanisms that might regulate decidual T cell activation and focused on mechanisms of extrinsic and intrinsic suppression. Several groups have reported an increase in T regulatory cells within decidua (41), and we were able to confirm this with a 2.5-fold increase in the proportion of T regulatory cells compared with blood. We were further interested to study the expression of inhibitory checkpoint proteins on decidual T cells, as these proteins mediate intracellular signaling pathways that limit the magnitude and outcome of T cell responses by negatively regulating T cell activation and function. Indeed, Abs that block inhibitory receptors are having a transformative impact on the treatment of many forms of cancer (42). We observed that the expression of four of the major checkpoint proteins, PD-1, Tim-3, CTLA-4, and LAG-3, increased substantially on effector memory T cells within decidua. A similar profile has been observed on decidual T cells in the first trimester and indicates that this profile is established early within pregnancy and is maintained throughout gestation (43). PD-L1 expression has been demonstrated on syncytiotrophoblasts and intermediate trophoblastic cells located in the chorion laeve and implantation site, which provides further support for the importance of this inhibitory pathway in reproduction (44).

In this context, it is important to consider that the pattern of expression of checkpoint proteins on decidual T cells is comparable to the levels that are expressed on tumor-infiltrating lymphocytes. The use of checkpoint inhibitors for cancer treatment might therefore have a profound impact on the outcome of human pregnancy and should be used with caution in this setting. Some murine models have suggested that interruption of the PD-1/PD-L1 axis can lead to fetal loss, but this observation requires further investigation (22, 43, 45).

Combinatorial staining showed that checkpoint proteins were often expressed together on decidual T cell populations. Again, PD-1 was the dominant partner in each case, but 35% of both CD4⁺ and CD8⁺ PD-1⁺ T cells expressed an additional checkpoint protein. This indicates that checkpoint expression is not redundant, but the physiological importance of these patterns of expression, and their relevance in relationship to checkpoint blockade in oncology, will need further investigation.

Given the evidence of local activation of T cells within decidua, it was important to assess whether decidual lymphocytes could recognize proteins expressed on fetal tissue. Interestingly, although both CD4⁺ and CD8⁺ decidual T cells proliferated in response to cord blood cells, this effect was more marked for CD8⁺ T cells. The profile of paternally derived Ags presented on cord blood cells will represent only a subset of those expressed on all fetal tissues, and as such we think that this assay would have underestimated the total magnitude of the fetal-specific immune response. Nonetheless, importantly, we were able to use it to demonstrate a role for T regulatory cells in the suppression of the decidual T cell response to fetal Ag. T regulatory cell depletion led to a 1.59- and 1.9-fold increase in decidual CD4⁺ and CD8⁺ T cell proliferation in response to cord blood and demonstrates the importance of this regulatory population in suppressing maternal response to fetal protein.

To investigate the potential importance of inhibitory protein expression on fetal-specific decidual T cells we were able to combine HLA-peptide tetramer staining of HY-specific T cells with assessment of PD-1 expression. This revealed a high level of expression of PD-1 on HY-specific cells, further implicating this checkpoint protein in an important role in regulating the activity of fetal=specific T cells within the decidual microenvironment. Unfortunately, the low frequency of such cells prevented us from assessing the functional impact of PD-1 inhibition. The overall picture that emerges from these results is that both intrinsic and extrinsic regulation are used in control of fetal-specific T cell function within decidua.

Microarray analysis revealed a wide range of genes whose transcriptional activity was differentially expressed in decidual effector memory CD4⁺ and CD8⁺ T cells compared with peripheral blood. Significantly, genes associated with IFN signaling receptivity were markedly increased in both CD4⁺ and CD8⁺ decidual T cells. Interestingly, high levels of IFN-γ have been reported within decidual tissue (46) and this appears to be localized primarily within cells of the monocyte linage (47). IFN production is driven most strongly by viral infection, and the observations raise the question as to whether localized viral infection is found within decidual tissue. Indeed, we would suggest that endogenous retroviruses are the likely trigger for IFN production within decidua. Endogenous retroviruses represent >8% of the human genome (48, 49) and became integrated into the mammalian genome between 0.1 and 40 million years ago via retroviral infection of germ cells (50). Notably, a protein encoded from the env gene of ERVW-1 called Syncytin-1 has an essential role in placentogenesis (51, 52) through formation of the syncytiotrophoblast. Furthermore, syncytium 1 has been shown to be released into the periphery via placental microvesicles that are able to illicit a T cell response (53), and substantial expression of proteins from the endogenous retrovirus HERV-K has been demonstrated in villous and extravillous cytotrophoblast (54). In the context of the functional relevance to this type 1 IFN signaling, IFN-α is known to be a significant antiangiogenic factor, resulting in downregulation of proangiogenic factors, including VEGF (55). This can result in impairment of vascular remodeling in the case of lupus patients, and it has been suggested that this IFN response can be localized to the decidua, resulting in malplacentation and pregnancy complication (56). Taken together, these data may implicate decidual T cells in a role that modulates further vascularization as pregnancy progresses.

Interestingly, our microarray also revealed substantial upregulation of the CLIC1 gene in decidual CD4⁺ effector T cells. This chloride channel protein has garnered much attention recently within solid tumors where its expression is associated with metastasis and invasion (57–59). CLIC1 is a metamorphic protein that can shift between two or more stable conformations, and its expression is increased by hypoxia where it acts to regulate cell function through the reactive oxygen species–mediated p38 MAPK signaling pathway (58, 60). It is known that the placenta develops in a hypoxic environment, and that hypoxia inducible factors play a key role in modulating placental cell function under low oxygen conditions (61).

Our work identifies, to our knowledge, a new role for CLIC1 in reproductive T cell immunology, and more research will be required to establish its physiological function in decidual CD4⁺ effector memory T cells. Of note CLIC3, which is another member of the CLIC family, has been shown to be important in pre-eclampsia, where increased levels at the syncytiotrophoblast layer result in disruption of membrane potential and downstream apoptosis (62).

Transcription of the LGALS1 gene, which encodes Galectin-1, was upregulated in CD4⁺ T cells. Galectin-1 is expressed by both endometrial and uterine NK cells during pregnancy (35), and it is involved in generation of tolerogenic dendritic cells (63). Intracellular expression of galectin-1 is also observed in activated T cells (64, 65) where it acts to increase sensitivity to the action of extracellular Galectin-1, in which it activates downstream apoptosis (37). In contrast, decidual CD8⁺ T cells upregulate galectin-9, a ligand of Tim-3, which is involved in both chemoattraction and apoptosis (66) as well as inhibition of chronic inflammation (67). Expression of Galectin-9 by T regulatory cells promotes stability of an immunosuppressive phenotype (68), and its expression by decidual CD8⁺ T cells may therefore increase effector cell regulation following interaction with Tim-3 (37).

The observation that highly differentiated fetal-reactive effector T cells are found within the decidua in healthy pregnancy raises questions regarding their potential role. HY dextramer staining was consistent with previous findings using MHC-peptide multimers when identifying human T cell responses to low-affinity minor histocompatibility Ags (14, 15). These maternal immune responses are likely to have been primed following systemic transfer of fetal cells and placental debris into the maternal circulation, which may involve either local or systemic presentation within lymphoid tissue. Note that the HLA-peptide multimers used in the study included the HLA alleles of HLA-A2 and HLA-B7 and these proteins are not expressed on fetal syncytiotrophoblast or EVT. As such it remains possible that anatomical limitations may serve to largely limit the potential interaction between fetal-specific T cells and fetal tissue. In this regard, it will be important in future studies to investigate the presence and function of fetal-specific cells restricted by HLA-C and HLA-G, which are the only polymorphic HLA molecules expressed on EVT.

Indeed, the finding that human decidua contains high numbers of T cells, at least some of which respond to fetal Ags, is provocative in relationship to the potential role of this response in disorders of pregnancy. Pre-eclampsia is associated with poor trophoblast invasion and impaired remodeling of maternal arteries, and strong evidence for a role for suboptimal NK activation has been established (69). The potential contribution of alloreactive T cells to this disorder has been more difficult to determine (13, 70, 71) and this may reflect the observation that the initial features of pre-eclampsia are established early in gestation, prior to the accumulation of a significant T cell infiltrate (71). Despite this, T cell–mediated immunopathology may be implicated in specific subtypes of pre-eclampsia and transcriptional analysis of placental tissue has recently identified a subset of disease associated with immune infiltration and inflammation (72). Moreover, it is now recognized that villitis of unknown etiology may result from recognition of fetal tissue by the humoral and cellular components of the maternal adaptive immune system (73). The clinical expression of villitis of unknown etiology includes fetal growth restriction, preterm birth, and recurrent pregnancy loss, and the histology is characterized by a heavy maternal inflammatory cell infiltrate of T cells and macrophages with a predominant CD8⁺ profile (74–76). This is associated with focal inflammation of fetal villi, and it is noteworthy that this profile is present to a modest extent in many normal pregnancies. The evolutionary balance of fetal-specific T cell responses may be delicately poised to permit successful reproduction in the great majority of cases while avoiding intensive local or systemic maternal immune suppression.

Our study reinforces the emerging concept that T cells play an increasingly important role at the maternal–fetal interface and that fetal-specific recognition is carefully modulated by multiple checkpoint proteins and T regulatory cells. Improved understanding of immune homeostasis within this unique microenvironment is likely to have substantial implications for many areas of clinical immunology.

The authors have no financial conflicts of interest.