Any discussion of the normal physiology of NK cells would be lacking without mention of their putative role in pregnancy. It has long been recognized that large granular lymphocytes of NK cell origin are the major lymphocyte populations lining the pregnant uterus.50 Uterine NK cells achieve peak numbers during the first trimester, when they represent 70% of all lymphocytes, and are not found at term.51 Uterine NK cells are different than mature circulating NK cells, yet phenotypically resemble the smaller unique blood NK cell subset, CD56+bright/CD16−/CD3−,52 and has low direct cytotoxicity. Since IL-2 is not produced in the placenta, proliferation of uterine NK cells is through production of IL-15 by placental macrophages and the IL-15 receptor on CD56+bright uterine NK cells.53 Significant progress has been made in understanding the role of NK cells by evaluating mice that lack NK cells. These mice are able to bear offspring but with 64% fetal loss.54 In addition, they have no endometrial glands, decreased placental size, decidual edema and abnormalities involving the large maternal arterioles supplying the placenta.50,55 The causal relationship between this phenotype and the role of NK cells was established by showing that T cell-deficient mice do not exhibit these abnormalities and transplantation of stem cells from severe combined immune deficient mice (which have normal NK cells but no T cells) into the NK cell-deficient mice reverses the phenotype.56 These studies show that CD56+bright NK cells occupying the mesometrial side of the pregnant uterus contribute to fetal implantation. Recent progress has been made in understanding this mechanism.
Since transgenic 26 mice have a 10-fold lower concentration of uterine IFN-γ,56 Ashkar and colleagues57 investigated mice lacking IFN-γ or IFN-γ receptor α-receptors. These latter two knockouts had an excessive number of uterine NK cells that were small and had limited numbers of cytoplasmic granules. The decidual arteries did not undergo normal gestation-induced remodeling. In contrast to the hypocellularity of the decidua in the NK cell knockout mice, decidua in the IFN-γ deficient mice progressed to overt necrosis during the second half of gestation. Further experiments used the RAG-2/common γ chain knockout mice, which also lack NK cells. As expected, these mice had very low concentrations of IFN-γ in the uterus. Engraftment of these NK-deficient mice with IFN-γ knockout marrow resulted in normal numbers of NK cells in the uterus, but an inability to initiate pregnancy-induced remodeling of spiral arteries.58 Interestingly, treatment with murine IFN-γ (100–1000 units per day x 6 days) reversed this defect and resulted in normal decidual artery remodeling. Taken together, these studies suggest that IFN-γ modulates the NK cell pool resident in the uterus, that uterine NK cell-derived IFN-γ is not necessary for initiation of decidualization, but appears essential for decidual maintenance in the second trimester.
It is intriguing that trophoblasts lack class I and class II HLA molecules but highly express nonclassical HLA-G, which is restricted to the placenta.59,60 Since NK cells have receptors for nonclassical class I molecules, some have hypothesized that HLA-G protects fetal cells from lysis by maternal NK cells through this mechanism. Using a migration assay, NK cell trafficking through the HLA-G-expressing endothelial cells was reduced.61 Recent studies show that HLA-G expression may interact with HLA-E molecules. Trophoblast cells were purified by flow cytometry and found to express HLA-E by reverse transcriptase polymerase chain reaction and immunoprecipitation.62 Staining with HLA-E tetrameric complexes refolded with the leader peptide derived from HLA-G identified 93% of decidual NK cells compared to 55% from the peripheral blood of non-pregnant women, suggesting the importance of NK cell lectin receptors in pregnancy. KIR may also be important, since uterine NK cells in the first trimester of pregnancy and essentially on all NK cells obtained from the placenta at term were found to express the p49 receptor, now termed KIR2DL4. KIR2DL4 was shown to directly bind HLA-G and may protect trophoblast cells from lysis by NK cells.60 Lanier62 has pointed out that it is still not clear whether KIR2DL4/HLA-G interactions are physiologic. These concerns are based on two possibly contradictory findings: 1) a woman found to have an HLA-Gnull mutation had normal fetal development and 2) fetal mice with β-2 microglobulin genes knocked out are not rejected by their heterozygous mother. Further studies into the role of NK cells in pregnancy will define these interactions and their physiologic relevance. Abnormal functions of NK cells in women with recurrent spontaneous abortions have been postulated. There seems to be a correlation with recurrent spontaneous abortion and increased NK cell cytotoxicity and an abnormally high decidual CD56+/CD16− ratio compared to normal pregnancies.63 A better understanding of NK cells in pregnancy may lead to important new therapies to prevent fetal loss.