Highlights
- •Transcriptomic analyses identify uNK cell-restricted cytokines with receptors on EVTs
- •Trophoblast organoids are exposed to the cytokines to model uNK cell-EVT interactions
- •uNK cell cytokines enhance EVT differentiation by promoting subtypes deep in decidua
- •uNK cell cytokines affect diverse genetic programs important for human placentation
Summary
In humans, balanced invasion of trophoblast cells into the uterine mucosa, the decidua, is critical for successful pregnancy. Evidence suggests that this process is regulated by uterine natural killer (uNK) cells, but how they influence reproductive outcomes is unclear. Here, we used our trophoblast organoids and primary tissue samples to determine how uNK cells affect placentation. By locating potential interaction axes between trophoblast and uNK cells using single-cell transcriptomics and in vitro modeling of these interactions in organoids, we identify a uNK cell-derived cytokine signal that promotes trophoblast differentiation at the late stage of the invasive pathway. Moreover, it affects transcriptional programs involved in regulating blood flow, nutrients, and inflammatory and adaptive immune responses, as well as gene signatures associated with disorders of pregnancy such as pre-eclampsia. Our findings suggest mechanisms on how optimal immunological interactions between uNK cells and trophoblast enhance reproductive success.
Introduction
The placenta supports the fetus throughout pregnancy, and abnormal placental development is a major cause of maternal and fetal mortality.1 In humans, trophoblast cells develop from the trophectoderm surrounding the blastocyst and differentiate into two main lineages.2 Cytotrophoblast cells cover the placental villi and form an overlying syncytium, which mediates nutrient and gas exchange with maternal blood, whereas extravillous trophoblast cells (EVTs) invade into the maternal decidual stroma, differentiating into interstitial EVTs (iEVTs) that eventually fuse to form placental bed giant cells (GCs). iEVTs migrate toward maternal spiral arteries and, in combination with endovascular EVTs that move down inside the arteries, remodel them to become high conductance vessels that can deliver a sufficient blood supply to the developing fetus.3 Defects in this process can cause disordered blood flow into the intervillous space, damage to the placental villous tree, and, thus, fetal growth restriction.4 Abnormal placentation underlies other common diseases of pregnancy including pre-eclampsia, stillbirth, and recurrent miscarriage.5,6
It has been difficult to study the detailed mechanisms underpinning these processes in normal and abnormal placentation due to the challenges of performing experiments that investigate the early stages of human pregnancy when EVTs interact with decidual cells. We have developed a trophoblast organoid system, where cytotrophoblast cells can be differentiated to EVTs that has now allowed us to address this question.7 Single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics of first trimester pregnant hysterectomy specimens show that our organoid model reliably recapitulates the in vivo EVT states.8 As EVTs invade the decidua, they encounter maternal immune cells. There is a long-standing notion that decidual leukocytes play a role in the pathogenesis of pre-eclampsia because there are immunological features of memory and specificity, demonstrated through epidemiological, genetic, and functional studies.6 The main population of immune cells that EVTs will interact with are uterine natural killer (uNK) cells that are spatially and temporally associated with placentation.9,10,11
Studies in animal models have demonstrated roles for uNK cells and invasive trophoblast in establishing the maternal-fetal interface.12,13,14,15,16,17 In rodents, uNK cells affect arterial remodeling and trophoblast invasion,15,18 and studies in humans suggest a similar role.9,19,20,21,22,23,24,25,26,27,28 Further evidence that uNK cells and trophoblast influence reproductive outcomes has come from previous histological and genetic studies in normal and abnormal pregnancies in humans.5,29,30,31,32,33,34 In pre-eclampsia and fetal growth restriction, defective trophoblast transformation of uterine spiral arteries by EVTs is detected by Doppler ultrasound in women early in pregnancy.35,36,37 Here, we use trophoblast organoids to investigate how uNK cells interact with EVTs and affect placentation during the early stages of pregnancy.
In humans, uNK cells express killer immunoglobulin-like receptors (KIRs) that bind to their human leukocyte antigen (HLA)-C ligands expressed by EVTs.33 All HLA-C alleles can be assigned to C1 or C2 groups based on dimorphism in the α1 domain. This means that EVTs can express C1 and/or C2 epitopes. The highly polymorphic KIR family includes the activating KIR2DS1 and the inhibitory KIR2DL1 receptors that can recognize C2 epitopes on EVTs.34,38 Thus, each pregnancy is characterized by different genetic combinations of maternal KIR and fetal HLA-C, resulting in a variable response of uNK cells when they encounter EVTs. Virtually, all individuals have the KIR2DL1 receptor, but only ∼45% Europeans have KIR2DS1. Genetic studies suggest that mothers lacking KIR2DS1, combined with a C2+HLA-C fetus, are at increased risk of pre-eclampsia,39,40 whereas mothers with KIR2DS1 and fetal C2+HLA-C are protected41,42; this combination is also associated with larger birth weights of babies.43 Experimentally, activating KIR2DS1 with its C2+HLA-C ligand stimulates secretion of CSF2 (granulocyte-macrophage colony-stimulating factor) by uNK cells, which influences trophoblast migration in vitro,42 illustrating the functional relevance of uNK cell cytokines during placentation.
In this study, we use trophoblast organoids to investigate how uNK cell-derived cytokines influence EVT behavior. First, by identifying cytokine-receptor pairs that mediate potential interactions between uNK cells and EVTs from in vitro modeling of the KIR-HLA combination (maternal KIR2DS1/fetal C2+HLA-C) that protects against pre-eclampsia and in vivo scRNA-seq profiling of maternal-fetal interface, we define a cocktail of cytokines restricted to uNK cells. Their functional consequences on EVTs are investigated using trophoblast organoids followed by comparisons with primary tissue samples. This reductionist approach, by focusing on specific uNK cell-derived cytokines, overcomes the considerable intrinsic genetic and experimental difficulties of co-culturing first trimester primary uNK cells with EVTs. Our findings show that uNK cell cytokines enhance EVT differentiation and regulate other diverse processes important during early pregnancy.
Results
Identification of cytokine-receptor pairs that mediate interactions between uNK cells and EVTs
NK cells in peripheral blood function by killing target cells or secreting cytokines/chemokines.44 In contrast, uNK cells are poorly cytolytic but do produce cytokines that are likely to affect trophoblast behavior.9 Our previous studies using scRNA-seq and mass cytometry defined three uNK cell subsets, uNK1, uNK2, and uNK3.45,46 Among these three subsets, uNK1 and some uNK2 cells express different combinations of activating or inhibitory KIR that bind the C1+ and C2+HLA-C groups expressed by EVTs45,46 (Figure 1A). Increased secretion of cytokines by uNK cells is seen after stimulation of activating KIR that are protective against pre-eclampsia.42 To test this further, we refined our earlier experiment and modeled KIR-HLA interactions by co-culturing uNK cells from women possessing the KIR2DS1 gene (n = 3 donors; Table S1) with the HLA-null cell line, 721.221 (221), transfected to express either C1+ or C2+HLA-C allotypes. After stimulation using C1+ or C2+221 cells, uNK cells were separated using flow cytometry into three subsets as follows: activating KIR2DS1 single positive (sp), inhibitory KIR2DL1sp, and KIR2DS1/KIR2DL1 double positive (dp) (Figure 1B). By transcriptional profiling of each subset using RNA sequencing, KIR2DS1sp uNK cells (the protective activating KIR for C2+HLA-C) show distinct responses to C2+HLA-C compared with those seen with KIR2DL1sp and KIR2DS1/L1dp from the same donor (Figures S1A–S1C). Genes exclusively upregulated in KIR2DS1sp uNK cells are particularly enriched for cytokines such as XCL1 and CSF2 (Figures 1C and S1D; Table S2).
To validate whether the cytokines are released by uNK cells in vivo, we examined the expression of 74 common cytokines/chemokines in different cell types at the maternal-fetal interface using our previous scRNA-seq data45 (Figure S1E). This revealed four cytokines that are restricted to uNK cells in comparison with other decidual cell types: XCL1, CSF2, CSF1, and CCL5 (Figure 1D). Among them, XCL1 and CSF2 are upregulated in KIR2DS1sp uNK cells after interaction with C2+HLA-C, confirmed by intracellular fluorescence-activated cell sorting (FACS) (n = 6 donors) (Figures 1E and S1F). Although the other two cytokines, CSF1 and CCL5, do not alter in response to KIR-HLA interactions (Figure S1G), they are major products of uNK cells45,47,48,49 (Figure 1D). Protein expression of all four cytokines is further confirmed by intracellular FACS in the uNK cell subsets (Figures 1F and S1H). CSF1, CCL5, and XCL1 were predicted to interact with EVTs in our previous scRNA-seq and spatial transcriptomic analysis of the maternal-fetal interface (CSF2 mRNA levels are low in in vivo scRNA-seq data).8,45 These interactions are further reinforced by the presence of the cognate receptors for all four cytokines in EVTs, as demonstrated by immunohistochemistry of EVTs in vivo,50,51,52,53 our scRNA-seq data45 (Figure S1I), and by flow cytometry of primary trophoblast cells (Figure 1G). We have therefore identified four cytokines—CSF1, CSF2, XCL1, and CCL5—with restricted production by uNK cells whose receptors are expressed by EVTs.
Modeling interactions between uNK cells and EVTs using trophoblast organoids
To determine the effect of this uNK cell cytokine cocktail on EVT behavior, we used our trophoblast organoid system.7 We performed flow cytometry to confirm the expression of the cognate receptors for these cytokines on organoids differentiated to EVTs (Figure 2A), demonstrating the feasibility of modeling interactions between uNK cells and EVTs using organoids. To mimic the in vivo decidual microenvironment, we exposed the organoids to the uNK cell cytokine cocktail during the induction of trophoblast cells to EVTs in EVT differentiation medium (EVTM) (Figure 2B; Table S3). As a control, the uNK cell cytokines were also added to organoids cultured in trophoblast organoid medium (TOM) without any EVT differentiation (Figure 2B; Table S3). We noticed that in some differentiating organoids cultured with these cytokines, more cells appeared to be invading the Matrigel droplet from the organoid (Figure 2C). We therefore checked the expression of genes defining the different trophoblast sub-populations by reverse transcription polymerase chain reaction (RT-PCR) (Figure 2D; Table S4). There was reduction in the expression of genes characteristic of villous cytotrophoblast (VCT), EPCAM, ITGA6, and MKI67 (EVTs no longer proliferate after beginning the invasive process), and upregulation of HLA-G, the definitive EVT marker, confirming differentiation to EVTs.54,55,56,57 ITGA2 is found in cells in a niche in the extravillous cytotrophoblast cell columns (CCCs),58 and its expression is increased after addition of cytokines for 96 h, suggesting that EVT differentiation is enhanced by the uNK cell cytokine cocktail.
Cytokines derived from uNK cells enhance EVT differentiation
To investigate further the cellular and molecular changes induced in EVTs by the uNK cell cytokines, we performed scRNA-seq of organoids at different time points during EVT differentiation treated with and without cytokines (Figure 2B; Table S1). After integration across organoids derived from different donors, followed by stringent quality control (Figures S2A–S2F), we obtained 67,996 cells from both control and cytokine-treated organoids which, based on canonical marker genes, covered the three main trophoblast populations: VCT, syncytiotrophoblast (SCT), and EVTs (Figure 3A). EVTs were further subdivided into three early, two intermediate, and three late subtypes based on their gradually increasing expression of established EVT marker genes (Figure S2G). VCT and SCT are detected in both TOM and EVTM, whereas, as expected, EVTs are only present in the latter (Figure 3A). To delineate the course of EVT differentiation, we next performed trajectory analysis by using the transcriptomic vector field in scTour.59 This recapitulates the known bidirectional differentiation pathways from VCT toward either SCT or EVTs, with EVTs undergoing further continuous progression from early, intermediate, to late stages (Figure 3A). At later time points, there is a trend for an increased proportion of late EVT subtypes in cytokine-treated organoids (Figures 3B and S2H). For instance, EVT_late_3 is mainly detected at 96 h from the organoids treated with cytokines in two donors (Figures 3B and S2F). This is reinforced by a differential abundance analysis, which demonstrates that late EVT subtypes, especially EVT_late_3 are significantly enriched for cytokine-treated versus untreated cells (Figure 3C)….
