Natural killer (NK) cells are large granular lymphocytes of the innate immune system and represent the first line in the host defence against invading pathogens.[1, 2] Unlike T cells, NK cells do not express an antigen-specific receptor but rather they express a large repertoire of activating and inhibitory receptors. Mature NK cells recirculate in the blood (pNK) where their number varies anywhere from 5 to 20% of total lymphocytes. Natural killer cells are also present in lymphoid and non-lymphoid tissues including the uterus where they are mainly CD56bright CD16neg. Given the fact that these cells are fully armed with cytolytic granules that can cause great harm, their activation is regulated by germline-encoded inhibitory or activating cell surface receptors that recognize self-ligands or pathogen-induced and stress-induced ligands. The net balance between activating and inhibitory signals would determine the outcome of NK cell responses against various threats. Activation of NK cells is inhibited mainly after interaction of inhibitory receptors with MHC class I molecules. However, the loss of MHC class I expression is not sufficient to trigger NK cell responsiveness because additional activating signals are required. The NK cells can eliminate their target through different mechanisms, including direct cell cytotoxicity or cytokine production. Besides their role as effectors of innate immunity, NK cells play a pivotal role in bridging the innate and adaptive arms of the immune system. By secreting large amounts of cytokines and chemokines, NK cells impact dendritic cell maturation[6, 7] and antigen-specific adaptive immune responses.[8, 9] During pregnancy, a special population of NK cells accumulates within the endometrium, which constitutes one of the maternal–fetal interfaces or decidua. These NK cells, referred to as decidual NK cells (dNK), play a pivotal role in the tissue homeostasis and endometrial vasculature remodelling that are necessary for embryo implantation and successful pregnancy. This review focuses on dNK cells and will discuss the latest work on their characteristics and functions.
Early phases of human pregnancy are associated with the accumulation of a unique subset of natural killer (NK) cells in the maternal decidua. Decidual NK (dNK) cells that are devoid of cytotoxicity play a pivotal role in successful pregnancy. By secreting large amounts of cytokines/chemokines and angiogenic factors, dNK cells participate in all steps of placentation including trophoblast invasion into the maternal endometrium and vascular remodelling. In this review, we summarize some of dNK cell features and discuss more recent exciting data that challenge the conventional view of these cells. Our new data demonstrate that dNK cells undergo fine tuning or even subvert their classical inhibitory machinery and turn into a real defence force in order to prevent the spread of viruses to fetal tissue. Today it is not clear how these phenotypic and functional adaptations impact cellular cross-talk at the fetal–maternal interface and tissue homeostasis. Ultimately, precise understanding of the molecular mechanisms that govern dNK cell plasticity during congenital human cytomegalovirus infection should lead to the design of more robust strategies to reverse immune escape during viral infection and cancer.
major histocompatibilty complex
natural killer cell receptors
Immune cells at the fetal–maternal interfaces
Pregnancy is a striking immunological paradox. Under normal healthy pregnancy, the conceptus carrying paternal antigens from an immunological point of view is a semi-allogenic graft that should be automatically rejected in an immune competent host.[11, 12] Yet, the fetus is completely protected from immune assault, suggesting that fine-tuning and complex adaptations from both parties would probably work together to thrust the immune system towards tolerance rather than rejection.
Although the fetus is never in contact with maternal tissues, direct contacts exist between maternally and fetally derived placental tissues. In haemochorial placentation (as in human and mouse placentation), these contacts occur through two distinct fetal–maternal interfaces (Fig. 1). The first interface is represented by the maternal decidua, which can be divided in three parts: (i) the decidua basalis (called here after decidua) located at the implantation site is composed of the decidualized endometrial stroma, which directly contacts the invasive extravillous trophoblast (EVT); (ii) the decidua parietalis lines the remainder of uterine cavity and is in direct contact with the non-invasive chorionic trophoblast; (iii) the decidua capsularis enclosing the conceptus acts as attachment for the chorion. Even if all deciduas contact fetal tissue, the decidua basalis is the only site where contact occurs on the first day of implantation.
The second interface is the intervillous space where maternal blood bathes the floating chorionic villous tree. This second interface constitutes a privileged site where fetal antigen shedding into maternal blood occurs. It is unclear whether maternal effector T cells sense these antigens, and whether specific adjustments are necessary to ensure systemic tolerance.
During the process of implantation, the decidua is populated by a large variety of leucocytes, which account for > 40% of the total cellular content. The major leucocyte population is represented by a particular subset of CD56bright CD16neg non-cytotoxic NK cells (dNK). In the first trimester of pregnancy, dNK cells represent >70% of decidual leucocytes.[15-19] The dNK cell number is very high throughout the first trimester and remains high through the second. However, it starts to decline from mid-gestation and reaches a normal endometrial number at term. Other immune cells are represented at much lower levels; human decidua contains 10% T cells, including CD8, CD4 and γδ T cells, as well as 20% monocytes/macrophages and 2% dendritic cells,[21-24] but B cells are barely detectable. The total number of T cells varies through the course of pregnancy but can reach up to 80% at term. The majority of decidual CD8pos and CD4pos T cells show features of induced regulatory T (Treg) cells.[25-28] The cellular cross-talk between decidual stroma, immune cells and fetal trophoblast is orchestrated by hormones/cytokines/chemokines/growth factors, and is a prerequisite for the development of the placenta.[29-32] The high level of CD56bright maternal dNK cells within the implantation site further highlights their importance in the immunology of pregnancy, which is far from being completely understood.
The origin of dNK cells
The origin of dNK cells is not yet clear. They could be generated in situ from early progenitors/precursors, which differentiate/proliferate in an environment enriched in steroid hormones and cytokines/chemokines to give rise to the dNK cell population.[33-35] This theory is further supported by the presence of an immature population of NK cells in the uterus, even before conception. These uterine NK cells regulate the differentiation and decidualization of the endometrium and their number varies during the menstrual cycle due to the effect of elevated levels of interleukin-15 (IL-15).[36, 37]
Similar to other lymphoid tissues, CD34pos precursors are present in the maternal decidua. These CD34pos progenitors are probably committed to the NK cell lineage as they express high levels of E4BP4 and Id2 transcription factors. They also express the common β chain receptor (CD122) and the IL-7 receptor α chain (CD127) but do not express stem cell markers (i.e. c-kit). Interactions with other decidual cells in a microenvironment enriched in IL-15 can easily drive the differentiation of these CD34pos progenitors into dNK cells.[38, 39]
It is also proposed that dNK cells derive from peripheral blood NK cells that migrate to the decidua through chemotaxis and acquire decidual phenotype within the local microenvironment.[33, 38, 40, 41] Studies demonstrating that decidual cells and invasive EVT produce large amounts of NK-attractant chemokines (CXCL10/IP-10, CXCL12/SDF-1, CCL2/MCP-1, CXCL8/IL-8, CX3CL1/fractalkine) and cytokines (IL-15) support this possibility.[38, 42-44] The dNK cells would originate from CD56bright pNK cells that are recruited to the decidua following the axis CXCR3–CXCL10 or CXCR4–CXCL12.[38, 42, 43]
However, dNK cells do not represent a homogeneous population as regards chemokine receptor expression; it is possible that they rise from several origins. Regardless of their origin as recruited or resident precursors/progenitors that mature locally, the decidual microenvironment conditions the education and the generation of dNK cells with unique phenotypical and functional properties to support healthy pregnancy. Consistent with this notion of local adaptations, exposure of pNK cells to transforming growth factor-β (TGF-β) or a combination of TGF-β/IL-15 or TGF-β/5-aza-2′-deoxycytidine promotes the conversion of pNK cells into an NK cell subset with reduced cytotoxic functions that can promote the invasion of human trophoblast cells.[41, 46] Moreover, the invasive EVT does not express the highly polymorphic MHC class I molecules but expresses HLA-C and the non-classical HLA-G and HLA-E MHC class I molecules that are recognized by NK cell inhibitory receptors [CD94/NKG2A and specific killer immunoglobulin-like receptor (KIR) receptors] acquired within the uterine microenvironment.
Phenotypic similarities and differences between dNK and pNK cells
Despite some similarities, the first-trimester pregnancy dNK cells and their pNK cell counterparts from the same donor present fairly distinct properties. Peripheral blood NK cells constitute up to 20% of circulating lymphocytes and are represented by two subsets; the CD56dim CD16pos subset constituting 95% total pNK and the CD56bright CD16neg minor subset. CD56dim pNK cells possess a high content of lytic granules and are highly cytotoxic while CD56bright pNK cells produce a large amount of cytokines and chemokines and are poorly cytotoxic. The majority of CD56dim CD16pos pNK cells express members of the KIR family. In contrast, most CD56bright CD16neg cells lack KIR expression but express high levels of the CD94/NKG2A inhibitory receptor. The expression of other activating and inhibitory receptors is also different in these two subsets.
On the other hand, dNK cells are largely composed of CD56bright CD16neg cells whereas CD56dim CD16pos subtype represents only a small fraction. The dNK cells display a unique repertoire of activating and inhibitory receptors that resembles the early differentiation stages of NK cells, distinguishing them from pNK cells.[16, 49-54] For instance, NKp30, NKG2C and ILT2 receptors are expressed on 30–50% of first-trimester dNK cells but only a few pNK cells express these receptors. Unlike resting pNK cells, dNK cells express the CD69 activation marker and a large fraction express NKp44 activating receptor. The expression levels of NKG2D and 2B4 (CD244) are similar for dNK and pNK cells. Like CD56bright CD16neg pNK cells, dNK cells express high levels of CD94/NKG2A inhibitory receptor (see Fig. 2). One striking difference concerns the granularity level. Even if they are poorly cytotoxic, dNK cells express much larger amounts of granzyme A, granzyme B and perforin enriched cytotoxic granules than CD56dim cytotoxic pNK cells.[18, 35, 51, 54]
Fine analyses of the dNK cell gene expression profile further highlighted these unique features of dNK cells with distinct properties, such as the expression of NKG2E, Ly-49L and KIR receptors, adhesion molecules, galectin-1 or some members of the tetraspan family (CD9, CD151, CD53, CD63).
Functions of dNK cells under normal pregnancy: mother's little helpers at play
The precise functions of dNK cells in vivo are not yet completely understood. Nonetheless, evidence exists for their pivotal contribution to the regulation of tissue homeostasis, a critical process for healthy pregnancy and optimal fetal development. At the same time, their endowment with huge plasticity and their susceptibility to external environmental stimuli should be taken into account for the success of pregnancy.
Decidual NK cells are devoid of cytotoxicity
Natural killer cells are named after their spontaneous and natural ability to kill tumours and virus-infected cells without previous sensitization. They belong to the group I of innate lymphoid cells because they produce large amounts of type I cytokines but not type II cytokines. They also secrete a large array of chemokines and other growth factors. In the periphery, CD56dim CD16pos pNK cells are highly cytotoxic, whereas CD56bright CD16neg NK cells are cytokine producers. In the decidua, dNK cells are devoid of cytolytic activity. The lack of cell cytotoxicity has been linked to default in the polarization of the microtubule organizing centre to the immunological synapse or to failure of the 2B4 receptor to convey activating signals.[54, 55] However, induction of dNK cell cytotoxic function by cytokines, such as IL-15 and IL-18, or ligation of specific activating receptor suggests that the lytic machinery is tightly regulated in normal pregnancy but can be triggered by the appropriate stress signal.[49, 55, 56] Our work and other's clearly suggest that cross-talk at the fetal–maternal interface upholds the cytotoxic function under strict control during healthy pregnancy. Inhibitory pathways involving the binding of the CD94/NKG2A inhibitory receptor to its natural ligand HLA-E expressed by the invasive fetal trophoblasts or the secretion of soluble factors such as HLA-G further comfort the tight regulation of dNK cell function during normal pregnancy.[49, 51, 57]
Decidual NK cells direct trophoblast invasion
The presence of dNK cells in the vicinity of invasive fetal trophoblasts and spiral arteries is suggestive of their active contribution to trophoblast attraction, which is necessary to promote decidualization and placental development. One particularity of dNK cells is their ability to produce a large range of cytokines and chemokines and angiogenic factors.[18, 50, 51, 59-61] Some of these soluble factors play a major role in the recruitment and attraction of fetal trophoblasts (i.e. CXCL10/IP-10, CXCL8/IL-8, CXCL12/SDF-1 and CCL2/MCP-1).[18, 50, 51, 59, 61] In contrast, invasive fetal trophoblasts can also help in the accumulation of dNK cells at the maternal decidua through the secretion of chemokines, such as SDF-1 and MIP-1α. Other factors, such as vascular endothelial growth factor (VEGF) C produced by dNK cells, can participate in immune tolerance by inducing TAP-1 expression, MHC class I molecule assembly and cell surface expression on trophoblasts. The fact that this secretion profile can be modulated by the ligation of a specific NK cell receptor suggests that the cross-talk between dNK cells and the invasive trophoblast expressing NKR ligands can regulate the secretion abilities of dNK cells.
Decidal NK cells direct the remodelling of spiral arteries
Evidence for the contribution of uterine NK cells in early phases of decidual angiogenesis was first provided by B.A. Croy and her colleagues using several strains of immunodeficient mice.[63-65] The picture is less clear in humans and the role of dNK cells in vascular remodelling is based on observations showing the presence of dNK cells in the vicinity of changing vessels. However, even if the role of human dNK cells in vasculature remodelling is not yet fully elucidated, these cells produce various pro-angiogenic and growth factors such as placental growth factor, VEGF A, and VEGF C, which can favour angiogenesis.[50, 60, 66] Vascular remodelling occurs in two steps that result in loss of the musculo-elastic structure and formation of breaks in the endothelial layer, which is then followed by the attraction of EVTs that become endovascular trophoblasts and replace the endothelium lining deep into the endometrium and partly into the myometrium.[67, 68] Both steps have been linked to the presence of dNK cells at the vicinity of the changing vessels. Changes of uterine arteries are crucial for the success of pregnancy because they ensure minimal vessel resistance and high blood flow of nutrients as well as oxygen to the conceptus.[14, 19] Immunohistochemical studies have demonstrated that the initial step of vasculature remodelling that takes place before the invasion of fetal trophoblasts is associated with significant accumulation of dNK cells and decidual macrophages within the vascular wall,[69, 70] and more recently R. Fraser and his colleagues confirmed the contribution of dNK cells to early phases of vascular remodelling in human pregnancy. Defaults in trophoblast invasion and/or vascular remodelling are hallmarks of pathological pregnancy, such as pre-eclampsia. Genetic studies suggested that special combinations of fetal HLA-C haplotypes and maternal dNK cell inhibitory KIRs increased the likelihood of pre-eclampsia.[47, 72-74] In a similar manner, the association of some HLA-C genotypes and maternal KIRs has been linked to defaults in trophoblast invasion. Hence, it is likely that the cross-talk between dNK cells and EVT either through ligation of activating and/or inhibitory KIR to their cognate ligands HLA-C and HLA-G or the secretion of a large panel of soluble factors by dNK cells contributes directly or indirectly to vasculature remodelling.[45, 75, 76]
Decidual NK cells direct induction of fetal tolerance
Immunotolerance must play a pivotal role in providing the immune privilege during pregnancy. Fetal trophoblasts do not express the classical HLA-A or B or MHC-II molecules that clearly favour their protection from T-cell attack at the maternal decidua. The majority of CD8pos and CD4pos T cells found in the decidua show an induced Treg cell phenotype. However, the exact mechanism responsible for the induction of Treg cells is not yet clearly defined. It is possible that dNK cells and decidual DC participate actively in generating this tolerogenic status. Cellular cross-talks between dNK cells, decidual macrophages/DC and T cells at the fetal–maternal interface[22, 77] might result in Treg cell induction. The tolerant microenvironment can be installed through active mechanisms such as the interaction between cytotoxic T lymphocyte antigen-4 and its ligand or indirect mechanisms implicating immunoregulatory molecules such as indoleamine 2, 3-dioxygenase, TGF-β or IL-10. Significantly lower numbers of dNK cells and decidual CD4 Treg cells have been linked to spontaneous abortion, further supporting the implication of these cells in fetal tolerance.[78-80]
Functional adaptations of NK cells during HCMV infection
Infection with human cytomegalovirus (HCMV), a member of the Herpesviridae family, is usually asymptomatic in healthy adults but can represent a real threat in immunocompromised patients. Primary HCMV infection is usually followed by the establishment of lifelong latency and sporadic reactivation phases. The role of pNK cells in controlling viral infections was supported by findings that NK-cell-deficient patients are highly susceptible to viral infections.[81, 82] The pNK cells are able to recognize and kill virus-infected cells through secretion of lytic granules containing TNF-related apoptosis-inducing ligand perforin and granzymes, Fas ligand and tumour necrosis factor-related apoptosis-inducing ligand. Recent work both in healthy adults and immunocompromised patients demonstrated that HCMV infection/reactivation could imprint the NK cell receptor repertoire. HCMV infection was associated with an increased CD94/NKG2C and KIR-positive pNK cell population that expresses low levels of NKp30, NKp46 activating receptors and the CD94/NKG2A inhibitory receptor.[83-88]
Congenital HCMV infection: role of dNK cells
Human cytomegalovirus infection is the commonest cause of congenital viral infection, affecting > 1% of live births. Primary maternal infection during the first trimester of pregnancy can lead to 40–50% of vertical transplacental transmission with permanent severe birth sequelae in almost 15% of congenitally infected newborns (i.e. neurodevelopmental abnormalities, deafness). The pathogenesis and mechanisms involved in vertical transmission are still not completely understood. HCMV spreads from the infected mother's decidual cells to the fetus. Sites of viral replication include cytotrophoblast progenitor cells in chorionic villi and differentiating/invading cytotrophoblasts. Until recently, the role of dNK cells in controlling viral infection was not known. However, epidemiological studies indicate that the rate of congenital HCMV infection is often low in the first trimester of pregnancy, which coincides with high numbers of dNK cells within the decidua, which suggests that dNK cells might be involved in protection against congenital HCMV infection.
Decidual NK cells express all the receptors involved in the response to HCMV and they also contain the necessary arsenal for cell cytotoxicity (Fig. 2). In a recent work, we provided the first evidence for the involvement of dNK cells in the response against congenital HCMV infection (see Fig. 3 for visual summary). Interestingly, dNK cells can be found in the vicinity of infected cells within floating chorionic villi, suggesting that the functional plasticity of dNK cells in response to invading pathogens is associated with modulation of their migratory phenotype. Deciual NK cells respond to congenital HCMV infection by lowering the secretion of several soluble factors (CCL2, CCL4, CCL5, CXCL10, granulocyte–macrophage colony-stimulating factor and CXCL8) that are involved in trophoblast invasion. By interfering with trophoblast invasion, dNK cells can participate actively in limiting viral spreading and congenital infection. Along the same lines, such changes within the microenvironment itself will not only limit trophoblast invasion but also induce inappropriate activation of other immune cells namely dendritic cells and T cells.
The ability to cross the placental barrier is one key determinant of invasive viruses and pathogens (hepatitis viruses, HIV, Plasmodium). Yet little is known about mechanisms underlying the fetal placenta tropism and the ability of dNK cells in the defence against these agents. Recent studies demonstrated that under certain conditions NK cells isolated from non-pregnant uterine mucosa and soluble factors secreted by decidual cells can control X4-tropic HIV-1 infection.[92, 93] Hence, it is conceivable that uterine NK and decidual NK cells act as local guardians against infection and their immune modulation might ensure efficient anti-viral protection.
During the first trimester of pregnancy dNK cells display unique phenotypic and functional properties that distinguish them from other peripheral blood or tissue NK cells. They orchestrate fetal trophoblast invasion and placental vasculature remodelling, which are necessary for the maintenance of a healthy pregnancy. Based on recent evidence, dNK cells may contribute to the protection of the fetus from congenital viral infections by eradicating local uterine viral infections and limiting the spreading of invading pathogens to fetal tissues. It is possible to speculate that defective dNK cell accumulation at the maternal decidua and/or impaired cross-talk within the local microenvironment may result in pregnancy failure. A major question is whether dNK cell response by itself is responsible for placental and fetal injuries observed in congenital infections. Future studies are necessary to unravel the molecular mechanisms controlling dNK cell functional plasticity and to understand the extent of dNK cell cross-talk with other immune cell populations and the global impact on the development of placenta and the outcome of pregnancy during congenital infections. A significant achievement in understanding the biology of NK cells was reached over the past decade. Today there is growing evidence indicating that NK cells may share more features with cells of the adaptive immune system including the development of memory in response to pathogens.[83, 84, 94-96] Therefore, the challenge in the field of immunology of pregnancy would be to understand whether dNK cells share some similarities with adaptive immunity such as clonal expansion associated with the development of long-lasting ‘memory-like’ populations. Nonetheless, progress in our understanding of dNK cell plasticity might have larger impacts beyond pregnancy and might help in designing future immunotherapy.
This work was supported in part by the Institut National de la Santé et de la Recherche Médicale and Centre National de Recherche Scientifique grants for the UMR5282. J.S. is supported by a French PhD fellowship from the ‘Ministère de l'enseignement Supérieur et de la Recherche’ and the ‘Fondation pour la Recherche Médicale’ France. We would like to thank Dr Reem Al-Daccak (INSERM UMRS 940, Paris, France) for helpful discussions and comments on the manuscript and Dr Lounas Ferrat for critical reading of the manuscript.
The authors declare that they have no conflict of interest.