REVIEW ARTICLE: The Unique Properties of Uterine NK Cells


Ofer Mandelboim, The Lautenberg Center for General and Tumor Immunology, The Hebrew University, Hadassah Medical School, IMRIC, Jerusalem, Israel.


Citation Manaster I, Mandelboim O. The unique properties of uterine NK cells. Am J Reprod Immunol 2009


Natural killer (NK) cells are lymphocytes of the innate immunity system that are able to kill various hazardous pathogens and tumors. However, it is now widely accepted that NK cells also possess non-destructive functions, as has been demonstrated for uterine NK cells. Here, we review the unique properties of the NK cells in the uterine mucosa, prior to and during pregnancy. We discuss the phenotype and function of mouse and human endometrial and decidual NK cells and suggest that the major function of decidual NK cells is to assist in fetal development. We further discuss the origin of decidual NK cells and suggest several possibilities that might explain their accumulation in the decidua during pregnancy.


Natural killer (NK) cells comprise approximately 5–15% of peripheral blood lymphocytes. They originate in the bone marrow from CD34+ hematopoietic progenitor cells,1 although recent studies suggest that NK cell development also occurs in secondary lymphoid tissues2 and in the thymus.3

NK cells populate different peripheral lymphoid and non-lymphoid organs, including lymph nodes, thymus, tonsils, spleen, and uterus.3,4 These innate effector cells specialize in killing tumor and virally infected cells and are able to secrete a variety of cytokines.5,6 In the peripheral blood, there are two NK subpopulations. The CD56dim CD16+ NK cells, which comprise ∼90% of the NK population, are considered to be more cytotoxic than the CD56bright CD16 NK cells, which comprise only ∼10% of peripheral blood NK cells and are the primary source of NK-derived immunoregulatory cytokines, such as interferon-γ (IFN-γ), tumor necrosis factor (TNF)-β, interleukin (IL)-10, IL-13, and granulocyte–macrophage colony-stimulating factor (GM-CSF).7 Although, a recent report suggests that even the CD56dim CD16+ NK population could secrete a large amount of cytokines, especially when interacting with target cells.8 These two NK subsets also differ in the expression of NK receptors, chemokine receptors and adhesion molecules, and in their proliferative response to IL-2. For example, CD56dim NK cells express high levels of the killer cell Ig-like receptors (KIRs) and CD57,9 whereas most of the CD56bright NK cells do not express KIRs and CD57, but express high levels of CD94/NKG2 receptors.10 The differential expression of chemokine receptors and adhesion molecules can also account for the functional differences between these NK subsets. For example, CD56bright NK cells express high levels of CCR7, CXCR3, and CXCR4.7,11 In addition, they express high levels of the adhesion molecule l-selectin.7 The expression of these molecules implies that CD56bright NK cells can migrate to secondary lymphoid organs, as well as to non-lymphoid organs. Indeed, it was shown that the T-cell regions of lymph nodes are enriched with CD56bright NK cells.12 It was also demonstrated that non-lymphoid tissues, such as the decidua, are enriched with this NK subset,11 which will be discussed later.

CD56dim NK cells, on the other hand, lack the expression of these molecules, but express high levels of the chemokine receptors CXCR1 and CX3CR1, which are linked to lymphocyte migration to peripheral inflammatory sites. One of the most prominent differences between the CD56bright and the CD56dim NK subsets is their intrinsic cytotoxic capabilities. As mentioned above, resting CD56dim NK cells are much more cytotoxic than resting CD56bright NK cells.7 The molecular mechanisms responsible for this are not fully understood. CD56dim NK cells are more granular than CD56bright NK cells13 and differences in the intracellular signaling pathways between the two subsets may also account for their cytotoxic capabilities. Indeed, it was demonstrated by gene expression profiling that compared with CD56dim NK cells, CD56bright NK cells express lower levels of the CD3ζ adaptor molecule, which mediates some of the natural cytotoxicity receptor signaling.14

Importantly, CD56dim NK cells exhibit high expression levels of FcγRIII (CD16), whereas CD56bright NK cells do not express CD16 or express only low levels of it and therefore, cannot perform antibody-dependent cellular cytotoxicity (ADCC). CD16 is a unique receptor not only because of its late function when the adaptive immune response is already activated, but also because among almost all NK cell receptors tested, it is the only receptor that could function independently without the help of other NK cell receptors.8

Uterine NK cells

It is now well established that NK cells can act as major regulators of the immune response, in addition to their ‘classical’ role of killing hazardous cells. The CD56bright CD16 NK subset is considered as the regulatory subset and a prominent example for its regulatory role is the function of these NK cells in the uterine mucosa prior to and during pregnancy, in the endometrium and decidua tissues, respectively.

NK cells in the non-pregnant uterus: endometrial NK cells

Murine Endometrial NK Cells

The data on mouse endometrial NK (eNK) cells are quite limited. It is known that mouse eNK cells are first found in 2-week-old mice as small and agranular cells.15 Recently, it has been suggested that B220+CD11c+NK1.1+ cells may be analogous to human CD56bright NK cells16 and a recent study indeed identified these cells in the uterus of virgin mice.17 In this study, the phenotype of mouse eNK cells was examined and it was demonstrated that eNK cells are B220+CD11c+NK1.1+ DX5+ (a phenotype that is similar to that of mouse peripheral blood and spleen NK cells18). These eNK cells also express CD122 (the IL-2/IL-15 receptor common β subunit), NKp46 (which is considered the most specific NK marker across species), CD11b (an integrin subunit), CD27 (TNF receptor family member), and CD69 (an activation marker which is also expressed on human eNK cells). It is important to note that mouse eNK cells do not stain for DBA,17–19 which binds N-acetyl-d-glalctosamine conjugates and is considered a selective marker of mouse uterine NK cells.19 However, it was suggested that the reactivity to DBA is acquired by uterine NK cells only during pregnancy, at gestation day (gd) 6,18 and therefore is not suitable as a specific marker for mouse eNK cells. Thus, from the little information available about eNK cells, it seems as if they represent a unique population of NK cells.

Human eNK Cells

Human eNK cells have been extensively studied in recent years. Immunohistochemistry studies showed that the absolute numbers of eNK cells increase dramatically from the proliferative to the late secretory phase of the menstrual cycle.20 Studies also indicated that eNK cells are proliferative, especially in the secretory phase of the menstrual cycle, as they were positive for the proliferation marker Ki67.21 However, as other lymphocyte populations can also increase in numbers during this period, the important parameter that should be considered when evaluating the importance of eNK cells during the menstrual cycle is that of lymphocyte percentage. Indeed, we have recently demonstrated that the percentage of human eNK cells actually remains constant during the menstrual cycle and only 30% of the endometrial lymphocytes are NK cells. Furthermore, the major lymphocyte population in the endometrium is that of T cells and not NK cells.20 Earlier studies support these findings.22,23

Phenotype of Human eNK Cells

Few studies have characterized the phenotype of eNK. Eriksson et al.9 showed that on the one hand, eNK cells share a similar expression profile of CD56, CD57, CD94, and CD16 with peripheral blood CD56bright NK cells. On the other hand, eNK cells share a similar expression profile of KIR receptors CD158b and NKB1 with CD56dim NK cells and they also lack the expression of l-selectin.24 Furthermore, eNK cells were shown to express the activation markers HLA-DR and CD69.22 We have recently characterized the expression pattern of the NK-activating receptors on eNK cells (isolated from endometrial tissues from women undergoing Pipelle biopsy before IVF treatments because of male infertility problems) and demonstrated that eNK cells lack the expression of CD16, but express relatively high levels of NKp46 and NKG2D [as do human decidual NK (dNK) cells]. However, in contrast to dNK cells, eNK cells also lack the expression of NKp30 and NKp44.20 This unusual repertoire of activating receptors and other cell surface markers makes eNK cells unique among other known NK subsets. The lack of expression of NKp30 and NKp44 could hypothetically be a result of sustained activation of the receptors by their unknown ligands, which are expressed in tissue,20 as was previously shown regarding NKG2D.25

CD9, a member of the tetraspanin family of proteins that has various cellular and physiological functions,26 was suggested as a specific marker for uterine NK cells (both eNK and dNK cells) as it was shown to be highly expressed on these cells,27 but not on peripheral blood NK cells.9 However, we have recently found that freshly isolated peripheral blood NK cells also express CD9 (Manaster I, Mandelboim O, unpublished data) and therefore we claim that CD9 cannot be considered as a unique marker for human uterine NK cells. Similar results were obtained for mouse uterine NK cells, which also do not uniquely express CD9.18

Functional Properties of Human eNK Cells

eNK cells were shown to express perforin and although Jones et al.28 determined that eNK cells are cytotoxic (with the exception of early proliferative phase eNK cells), their cytotoxic activity was extremely low (<20%). We have recently demonstrated that freshly isolated eNK cells exhibit extremely low levels of cytotoxicity and fail to produce cytokines such as IFN-γ, interferon-inducible protein-10 (IP-10), vascular endothelial growth factor (VEGF), and placenta growth factor (PLGF), without additional cytokine stimulation.20 This lack of NK function was observed in both proliferative and secretory phase eNK cells. Importantly, following activation with IL-15 (a cytokine that is important for NK cell differentiation,29,30 is known to be important during pregnancy31,32 and whose receptor is expressed on eNK cells33) eNK cell cytotoxicity and their secretion of IFN-γ and IP-10 was up-regulated.20 Therefore, our results suggest that eNK cells are inert lymphocytes in the endometrium that are unable to kill target cells or to secrete NK known cytokines and growth factors, before IL-15 activation. Supporting these results, Eriksson et al.9 have also shown that eNK cells were able to produce IFN-γ and IL-10 following activation with IL-12 and IL-15. Recently it was demonstrated that eNK clones are able to secrete VEGF-A and VEGF-C and thereby support the endovascular process;34 however, these eNK cells were grown in culture in the presence of IL-2, a cytokine that was shown not to be expressed in the tissue and therefore is less suitable for in vitro activation of eNK cells.35 As stated above, we determined that freshly isolated eNK cells do not secrete VEGF and also do not contain VEGF transcripts.20

NK cells in the pregnant uterus: decidual NK cells

Mouse Decidual NK Cells

In the mouse uterus, decidualization and implantation of the blastocyst occur at gd 4. At gd 6, dNK can be detected in the decidua basalis, as they stain positive for DBA.19 From gd 8, dNK cells proliferate in the mesometrial lymphoid aggregate of pregnancy (MLAp), a transient lymphoid structure that forms between the two layers of myometrial smooth muscle.36 In these lymphoid structures, dNK cells surround the uterine artery branches that enter the implantation sites. These cells peak in number at mid-gestation (gd 9–10) and their numbers decline afterwards, at gd 10–12.36

Phenotype of Mouse dNK Cells

The receptor repertoire of mouse dNK cells has only recently been defined. Yadi et al.18 found that there are two distinct subsets of CD122+ CD3 dNK cells within the mouse uterus at mid-gestation. The smaller subset that was identified was similar in phenotype to peripheral blood mouse NK cells, expressing both NK1.1 and DX5. The second, larger subset displayed a unique phenotype: these dNK cells did not express the common markers of mature NK cells (NK1.1 and DX5) nor did they express the differentiation markers CD27 and CD43. However, they did express the NK specific marker; NKp46 and also NKG2D and CD16 and therefore these cells are part of the NK lineage. The unique receptor repertoire of dNK cells further includes the expression of several Ly49 receptors, the expression of activation markers such as CD69 and KLRG1 (which is considered as a marker for active NK cell proliferation37) and the expression of CD117 (the c-kit receptor). Another study, by Mallidi et al.17 described the phenotype of NK1.1+ dNK cells as DX5+ NKp46+ CD27+ CD11b+ CD11c+ CD69+. Interestingly, the NK1.1+ dNK cells expressed more B220 and CD69 than NK1.1+ eNK cells and also expressed ICOS (which is expressed on activated NK cells38), whereas eNK cells did not express ICOS at all.

Human Decidual NK Cells

In the fetal-maternal interface, the maternal uterine tissue is in close contact with the fetal-derived trophoblast cells. This interface contains immune cells, which constitute 40% of the cells in the human decidua.39 Analysis of this immune population has revealed that, unlike any other tissues or mucosal surfaces, 50–70% of the human decidual lymphocytes are NK cells, while the remainder are CD14+ macrophages, dendritic cells, CD4+ T cells, a few CD8+ T cells, γδ T cells, and NKT cells.35 dNK cell numbers are the highest in the first trimester of pregnancy and their numbers decline during the second trimester. As in mice, only few dNK cells are present in the human decidua at term.36

Phenotype of Human dNK Cells

The majority of dNK cells are CD56bright CD16 (as opposed to mouse dNK cells which express high levels of CD1618). Indeed, dNK cells resemble peripheral blood CD56bright CD16 NK cells also in the high expression levels of CD94/NKG2.40 However, similar to eNK cells, dNK cells resemble CD56dim CD16+ NK cells in the expression of KIRs41 and in their granules cell content. In fact, dNK cells differ from peripheral blood NK cells both in phenotype and in function. Comparison analysis of the gene expression in dNK cells versus peripheral blood NK cells showed that dNK cells should be considered as a unique NK subset.27 dNK cells over-expressed several genes, compared with the two peripheral blood NK subsets and several genes were exclusively expressed in dNK cells. For example, granzyme A was significantly over-expressed in dNK cells, as were the C-type lectin-like receptors NKG2C and NKG2E.

dNK cells have been shown to express several activating receptors, including NKp46, NKp30, NKp44 (in contrast to human eNK cells which lack NKp30 and NKp44 expression, as discussed above), NKG2D, and 2B4.42–44 The expression of NKp44 (which is not expressed on non-activated peripheral blood NK cells) and the expression of the activation marker CD6945 (which is also expressed on mouse dNK cells) suggest that dNK cells might already be activated in the local environment of the decidua. Considering this, it would be interesting to test in the future whether the activation of only one activating receptor on dNK cells by targets expressing individual ligands, would result in enhanced cytotoxicity, unlike NK cells in the peripheral blood.8

The expression of inhibitory receptors includes NKG2A, KIR2DL4, KIR2DL1, KIR2DL2/L3, and ILT-242,45–47 which might function to inhibit the cytotoxic potential of dNK cells, as discussed below.

Decidual NK cells do not execute their cytotoxic potential toward trophoblasts

Although dNK cells are in close contact with fetal-derived trophoblasts they do not exert cytolytic functions against trophoblast cells.48 Several studies have shown that the general cytotoxicity of dNK cells is reduced compared with peripheral blood NK cells,42,49 despite the fact that they express several activating receptors (as mentioned above), as well as high levels of perforin and granzyme A and B.27,42,50 The cytotoxic activity of dNK cells, although potentially low, is still preserved, as engagement of NKp46 (but not NKp30) in freshly isolated dNK cells induced intracalcium mobilization, perforin polarization, granule exocytosis and triggered apoptosis in target cells.45 Such existing killing potential of dNK cells might be important in case of uterine viral infection.

Several different explanations for the lack of cytotoxicity toward trophoblast cells have been proposed. First, this phenomenon could be a result of inhibitory interactions between the non-classical class I MHC- molecules HLA-G and HLA-E and the inhibitory receptors expressed on dNK cells, e.g. ILT-2, KIR2DL4 [32], and CD94/NKG2A.45,51 However, ILT-2, the most dominant HLA-G binding NK inhibitory receptor is only expressed on ∼20% of dNK cells, and whether KIR2DL4 could interact with HLA-G and inhibit NK cell activity is still controversial.52 Second, it has been suggested by Kopcow et al.44 that dNK cells are unable to form mature activating synapses and to polarize perforin. This might also not be the only explanation, because as mentioned above, NKp46 is cytotoxic in dNK cells.45

Vacca et al.42 provided another possible explanation according to which, the cytotoxic activity of dNK cells is inhibited by the receptor 2B4, which delivers inhibitory signals that correlate with low or absent signaling lymphocyte activation molecule-associated protein (SAP) expression in dNK cells.

Finally, it seems, of course, reasonable that interactions of dNK cells with neighboring immune and non-immune cells at the decidua further inhibit their ability to damage the local tissue.

Tissue remodeling properties of dNK cells

The decidual microenvironment probably encourages dNK cells to exert their constructive functions. The landmark studies of Croy’s group demonstrated the novel concept of constructive functions for mouse dNK cells in vivo at the fetal-maternal interface and their involvement in tissue homeostasis.53 Their work demonstrated that depletion of dNK cells in the mouse decidua resulted in abnormal implantation sites and inadequate remodeling of the decidual spiral arteries. Furthermore, they showed that these abnormalities were a result of dNK-derived IFN-γ, which positively regulates the diameter of the lumen of the spiral arteries during decidualization.53,54 Thus, dNK cells are crucial for normal vascular modifications at the decidua, a process that is vital for the development of a normal pregnancy. Besides IFN-γ, mouse dNK cells also secrete colony stimulating factor-1 (CSF-1), IL-1, leukemia inhibitory factor (LIF),55 TNF-α, and VEGF.56

Studies of human dNK cells have shown that dNK cells produce a variety of cytokines and growth factors. At the mRNA level, it was shown that human dNK cells produce transcripts of GM-CSF, CSF-1, TNF-α, LIF, and IFN-γ.57 A recent study has shown that the engagement of NKp30, but not NKp46, induces the secretion of TNF-α, MIP1-α, MIP1-β, GM-CSF, and IFN-γ by human dNK cells that were shortly activated with IL-2 or IL-15 for 48 hr.45 Human dNK cells can also secrete IL-8 and IP-10 and it was demonstrated that these chemokines bind to their receptors on invasive trophoblasts causing trophoblast migration.43

Human dNK cells can also produce a variety of angiogenic factors, including several members of the VEGF family, PLGF, angiopoetin-2 (Ang-2), and NKG5.43 These findings further support the function of dNK cells as major regulators of vascular remodeling during the early stage of pregnancy.

The ability of dNK cells to secrete a variety of cytokines that support these developmental processes during pregnancy suggests that dNK cells must be activated in the tissue, rather than inhibited, to exert their constructive roles at the fetal-maternal interface and establish a normal pregnancy. Indeed, it has been shown that dNK clones expressing the activating receptor KIR2DS4 generated higher amounts of IL-8, IP-10, VEGF, and PLGF than clones expressing inhibitory receptors, such as KIR2DL1. This suggests that activation of dNK cells reduces the risk of pre-eclampsia, through the production of sufficient amounts of growth factors and chemokines by dNK cells.43 These factors contribute to trophoblast invasion and vascular modifications, as discussed above. The study of Moffet’s group strongly supports this notion as well. Their study suggested that strong inhibition of dNK cells, as a result of interactions between certain KIR alleles on dNK cells and certain HLA-C allels on extravillous trophoblasts, increases the likelihood of pre-eclampsia. Furthermore, interactions between KIRs and HLA-C, which induce activation of dNK cells, result in a better trophoblast invasion.58

Decidual NK cells and their decidual neighborhood

The unique properties of dNK cells probably result from intense communication between these cells and their neighboring decidual cells, local cytokines, other immune cells, and secreted hormones that create the special microenvironment of this tissue.

dNK cells are in close contact with invasive trophoblasts and local decidual cells49 and therefore, there is probably a constant exposure of dNK receptors to their ligands. Indeed, decidual stromal cells express unknown ligands for the dNK-activating receptors NKp30 and NKp44.43 In addition, purified HLA-G+ trophoblasts express unknown ligands for NKp44.43 Thus, chronic stimulation of dNK receptors could affect the function of dNK cells. A support for this hypothesis comes from a mouse in vivo model in which NK cells, which were chronically exposed to the NKG2D ligand, were impaired in their NKG2D-dependent cytotoxicity, but constitutively produced IFN-γ.59 It is therefore possible that chronic stimulation of dNK-activating receptors by their ligands could be responsible for their lack of cytotoxicity toward fetal cells and their enhanced ability to produce growth factors.

Soluble factors produced by neighboring decidual, immune or trophoblast cells can also influence dNK cells. These soluble factors could be cytokines, such as IL-1531 or other proteins, such as trophoblast-derived soluble HLA-G.60,61 Another possibility is hypoxic stress within the decidua that might influence the expression of the ligands for the dNK receptors. Indeed, tissue stress, such as genotoxic stress, was shown to up-regulate the expression of NKG2D-ligands that stimulate NK cells.62 Further study is needed to support this hypothesis.

The origin of decidual NK cells

The mechanisms controlling the accumulation of CD56bright CD16 NK cells in the decidua are still being investigated.

Several possibilities for the origin of dNK cells have been proposed. One possibility is that NK cells are recruited from other organs or from the peripheral blood to the decidua, where they undergo further tissue specific differentiation. Alternatively, it was suggested that self renewal from local progenitor cells is the mechanism responsible for the accumulation of NK cells in the decidua, as will be discussed later. It is also possible that dNK cells originate in eNK cells that already reside in the tissue and undergo further differentiation into dNK cells in the new environment that pregnancy creates. Our suggestion (as discuss below) is that dNK cells are probably a heterogeneous population that encompasses all of the above.

NK cell migration to the uterus

Several studies support the notion that dNK cells originate in peripheral blood NK cells.43,63

Keskin et al.64 suggested that dNK cells might originate from the CD56dim CD16+ peripheral blood NK cells that migrate to the decidua and differentiate locally to dNK cells under the influence of tissue-derived TGF-β and other factors. However, other studies support the hypothesis that the CD56bright CD16 dNK cells originate rather in the CD56bright CD16 NK subset.

The recruitment of NK cells from the blood to the decidua involves adhesion molecules. l-selectin is highly expressed on CD56bright CD16 NK cells, as opposed to CD56dim CD16+ NK cells, and was shown to be involved in the initial adhesion to lymph node high endothelial venules, therefore giving the CD56bright CD16 NK cells an advantage in extravasation to tissues.65 Interestingly, CSPG-2, the ligand of l-selectin, was shown to be highly expressed in the tissue, during the secretory phase of the menstrual cycle. Thus, it was suggested that l-selectin is involved in the recruitment of CD56bright CD16 NK cells to this tissue,66 although it is not clear yet whether l-selectin ligands are expressed in the decidua, during pregnancy.

CXCL12 was shown to play an important role in NK cell migration to the decidua.11,67 CXCR4, which is highly expressed on both peripheral blood CD56bright CD16 and dNK cells seems to be essential for CD56bright CD16 migration, through its interactions with its ligand CXCL12, which is expressed by invasive trophoblasts.11 The CD56bright CD16 peripheral blood NK cells that were attracted to the decidua by the invasive trophoblasts further differentiate in the decidual microenvironment and acquire dNK characteristics.

Other chemokines were also shown to participate in the attraction of peripheral NK cells to the decidua. For example, it was suggested that cytotrophoblasts can attract CD56bright CD16 NK cells by producing MIP1-α.68

In mice, the origin of dNK cells is also not clear. Murine studies indicate that dNK cells do not self-renew in the uterus, but are rather derived from secondary lymphoid tissue.69 Indeed, it was recently suggested that mouse dNK cells do not originate in the thymus, as they are negative for CD127,18 which was suggested as a molecular marker of a pathway of mouse NK cell development that originates in the thymus.3 It is possible that mouse dNK cells originate in the small population of NK1.1+DX5+ NK cells that are found in the mouse decidua and resemble peripheral blood mouse NK cells.18

The involvement of chemotaxis in the control of dNK accumulation is still not clear. Studies of CCR2−/−, CCR5−/−, MIP1-α−/− or MIP1-α−/−CCR2−/− null mice did not detect any changes in the localization or activation of NK cells.70

Self renewal from local progenitor cells

dNK cells might alternatively originate in hematopoietic progenitor cells that reside in the endometrium, proliferate and differentiate into dNK cells during early pregnancy.

The presence of hematopoietic stem cells (HSC) in the human endometrium was demonstrated by Lynch et al.71 who showed the existence of a relatively mature HSC population in the endometrium that does not express lineage-committed markers.

Indeed it was shown that when human endometrium was transplanted into NOD/SCID/γcnull mice, there was an increase in NK cell levels by day 28 of the menstrual cycle.72 NOD/SCID/γcnull mice lack T and B lymphocytes, and have extremely low levels of NK cells. Therefore, migration of NK cells from the peripheral blood to the tissue cannot account for the observed increase in NK cell numbers, which was determined by the expression of CD56, which is expressed in human, but not in murine NK cells.

Another finding that could support the concept that dNK cells might originate from local stem cells is the expression profile of chemokine receptors in dNK cells. dNK cells express high levels of CXCR3 and intermediate levels of CXCR4.11 However, as these receptors and others are not expressed on all dNK cells, it is possible that some dNK cells, which do not express chemokine receptors at all, originated in local progenitor cells and did not migrate to the decidua from the peripheral blood via their chemokine receptors.

Differentiation of eNK cells into dNK cells

It is conceivable that if NK-progenitor cells reside in the endometrium, they differentiate into eNK cells rather than dNK cells. Indeed, we have recently observed that human eNK cells do not express any of the chemokine receptors tested (including CXCR1, 2, 3, and 4 and CCR1, 2, 3, 5, and 7), therefore suggesting that eNK cells do not migrate to the endometrium from other tissues or from the blood, but rather originate from local hematopoietic progenitor cells.20 Furthermore, we found that eNK cells display an immature form: they possess no apparent functional activity (no cytotoxicity and no cytokine secretion) and do not express the major activating receptors NKp30 and NKp44. However, we observed that following IL-15 activation, eNK cell cytotoxicity and cytokine secretion were up-regulated and they acquired a phenotype similar to that of dNK cells, as NKp30 and NKp44 activating receptors were up-regulated as well.20 Therefore, we suggested a hypothesis according to which, after conception, the levels of IL-15 rise in the decidua31 and promote the differentiation of eNK cells toward dNK cells. Therefore, eNK cells might be part of the progenitor cells of dNK cells.20

A similar idea was recently suggested in the mouse model: mouse NK1.1+ eNK cells express low levels of B220 and do not express ICOS, whereas dNK cells express high levels of B220 and ICOS. Interestingly, following IL-15 activation, the authors observed an up-regulation of B220 and ICOS expression on eNK cells, suggesting that in the mouse, eNK cells might be an early, undifferentiated form of dNK cells.17 It should be noted, however, that in their experiment, the authors could not determine whether the observed eNK differentiation was indeed a direct effect of IL-15, as their culture contained other uterine cells as well.


The two NK subsets of the uterine mucosa are intensely investigated. The eNK cells seem inactive relatively to dNK cells, which are probably their mature, fully differentiated form. However, more research is needed to establish the exact role of eNK cells in the cycling endometrium, the origin of dNK cells (although it is probably a combination of migration to the tissue as well as differentiation of local cells) and their relationship with their surrounding decidual environment.


This work was supported by the Israel Science Foundation, the European consortium LSHC-CT-2005-518178, the European consortium MRTN-CT-2005, the ICRF, and the BSF. We thank our long-term collaborators, Prof. Simcha Yagel and his team.