Uterine natural killer cell biology and role in early pregnancy establishment and outcomes.

Changes to the maternal environment even before conception can affect early placentation. For example, obese patients are known to have an increased risk of preeclampsia, gestational hypertension, preterm birth, and perinatal death ( 99 ). Animal studies utilizing embryo transfer have demonstrated that in mice, the maternal environment is responsible for many of the placental changes associated with obesity ( 16 , 100 ). In addition, the supraphysiologic hormonal milieu created by ARTs leads to an increased risk of placental abnormalities as seen in a number of mouse models ( 14 – 16 , 101 ) as well as demonstrated adverse sequelae in humans including low birth weight ( 13 ), preeclampsia ( 10 ), placenta previa ( 11 ), and placental abruption ( 10 ). Although the mechanisms contributing to these outcomes are poorly understood, recent evidence suggests that the maternal uterine immune cell population is a significant contributor to adverse perinatal outcomes. Immune cell population disturbances have been found in multiple adverse pregnancy outcomes including fetal growth restriction and preeclampsia ( 22 ). In addition, these complications have been associated with abnormal placentation ( 1 – 4 ), suggesting a role for maternal immune cells in processes significant for the establishment of the placenta. In this section, we examine how uNK cells could impact three processes essential for placentation—decidualization, trophoblast invasion, and spiral artery remodeling. Endometrial decidualization is an imperative process that begins before the arrival of the embryo, altering the structure of the endometrium to allow for embryo implantation. These predominantly hormone-mediated structural changes involve transformation of endometrial stromal fibroblasts into secretory, epithelioid-like decidual cells, differentiation of glandular elements, and increased tortuosity of the spiral arteries ( 102 – 104 ). With massive uNK cell influx into the environment at this time, it is suspected that they may play a critical role in this process. Although current literature demonstrates that uNK cells are present in abundance at time of decidualization, their role in decidualization of endometrial stromal cells is not yet clear. The initiation of an increase in uNK cells begins 3 days after the luteinizing hormone surge ( 41 ), before arrival of the embryo. The first stromal changes, however, are not noted until approximately 8 days after the luteinizing hormone surge ( 41 ). This temporal appearance suggests that prepregnancy NK cells play a role in the changes to the endometrial stroma that are necessary for successful implantation. This is supported by data showing that adding supernatant from cultured uNK cells to endometrial stromal cells cocultured with uterine epithelial cells led to significant changes in expression of genes known to be markers of decidualization ( 105 ). In addition, this coculture system demonstrated the presence of a positive feedback loop where uNK cells promoted stromal cell production of factors such as IL-15 involved in uNK cell proliferation and localization to the endometrium. In addition, another study reinforced the presence of this positive feedback loop, demonstrating that pregnancy uNK cells additionally secreted factors such as IL-25 required for decidualized stromal cell proliferation ( 106 ). In addition, this work tested the functional influence of pregnancy uNK cells on stromal cells and found that decidualization was accelerated in the presence of uNK cells, as measured by prolactin secretion and upregulation of IL-25 production ( 106 ). While these studies provide compelling evidence for a role for uNK cells in decidualization, both utilized uNK cells isolated from first trimester tissue after the onset of pregnancy ( Table 2 ). In humans, stromal cell decidualization takes place before pregnancy, and examining how prepregnancy uNK cells interact with stromal cells is, thus, imperative. However, these data including temporal relationship of uNK abundance increase before decidualization as well as evidence of crosstalk between uNK and stromal cells with functional evidence of uNK importance in endometrial decidualization ( Fig. 2A ), integral to human pregnancy establishment. During the process of implantation and establishment of the placenta, there are two routes of EVT invasion. The first occurs through the decidua, giving rise to interstitial EVTs ( 61 ). The second is up the lumen of the spiral arteries, giving rise to endovascular EVTs ( 61 ). This second group of EVTs takes on characteristics of endothelial cells and replaces maternal endothelial cells within the dilated spiral arteries. Extravillous trophoblast invasion is essential for establishment of the placental bed for the purpose of oxygen and nutrient exchange between the mother and developing fetus. Uterine NK cells have been shown to assist in regulation of this process. Uterine NK cells produce cytokines that regulate EVT invasion. This has been demonstrated through several mechanisms. Supernatant from cultured uNK cells obtained from late first trimester abortion tissue stimulates EVT invasion from placental explants through a Matrigel layer ( 107 , 108 ). In addition, uterine NK cell conditioned medium is able to promote trophoblast differentiation and invasion of cultured primary cytotrophoblasts and an EVT cell line through Matrigel. Additionally, this conditioned medium enables trophoblasts to acquire endothelial cell features ( 109 ). Specific agents of chemoattraction have additionally been proffered as a mechanism of influence via uNK release of hepatocyte growth factors, IL-8, CXCL10, and IL-6 ( 110 , 111 ). Further investigation of the role of IL-8 in this process demonstrates high levels of IL-8 protein and mRNA in uNK cells, stimulation of EVT invasion with exogenous IL-8, upregulation of matrix metalloproteinase (MMP)-2 with exposure to IL-8, and finally significant reduction in EVT invasion with exposure to IL-8 neutralizing antibody ( 108 ). In vitro, granulocyte macrophage colony-stimulating factor is known to enhance primary trophoblast invasion. Granulocyte macrophage colony-stimulating factor production by pregnancy NK cells increases when activated by HLA-C2 (expressed by trophoblasts), thereby demonstrating positive regulation of invasion by the NK cells ( 112 ). In addition, vascular endothelial growth factor expression by uNK cells appears to play a role as coculture of uNK cells treated with a sphingosine-I-phosphate analog and receptor modulator, leading to decreased VEGF expression results in decreased EVT migration compared with untreated uNK cell coculture ( 113 ). Other studies have demonstrated uNK cell inhibition of EVT invasion. Exposure of first trimester trophoblasts to concordant first trimester uNK cells demonstrates contact-independent inhibition of EVT invasion. On further investigation, changes in protease expression (MMP-2 and MMP-9) and E-cadherin expression were documented in the EVTs. Finally, interferon gamma (IFN-gamma) has been shown as a potential mediator of this process with demonstrated partial reversal of EVT invasion inhibition with introduction of anti-IFN-gamma ( 114 ). These data demonstrate the functional effect of various secreted products of uNK cells on EVT invasion, thereby supporting the role of uNK cells in the regulation of trophoblast invasion ( Fig. 2B ). This multimodal influence of uNK cells on trophoblast invasion may again have implications for dysfunctional placentation and subsequent adverse perinatal outcomes. The spiral arteries of the endometrium are a transient component of the uterine architecture, providing blood supply to the endometrium during the luteal phase of the menstrual cycle. Remodeling of these arteries must occur to allow successful implantation of a human embryo and integration with the maternal vascular supply. This process involves loss of the musculoelastic structure of the arteries with fibrinoid material replacement of the vascular media ( 61 , 110 , 115 ). During this time, the endothelium swells with resultant desquamation and sloughing into the lumen ( 116 ). In addition, the arteries dilate to decrease resistance to blood flow ( 61 ). The process continues into early pregnancy with completion of the remodeling of the decidual portion of the spiral arteries by 10–12 weeks’ gestation and the myometrial portion of these vessels by 14–16 weeks’ gestation ( 61 ). These structural changes have both trophoblast-independent and trophoblast-dependent stages ( 62 ). This has been demonstrated in the setting of ectopic pregnancy with notable structural changes to the endometrium without the presence of an intrauterine gestation ( 117 ). This complex process is integral for successful placentation in early pregnancy. There are a number of influences at play in guiding this process. Uterine NK cells can be found in the vicinity of the spiral arteries during spiral artery remodeling, suggesting a role in this process ( 115 ). Culture media from uNK cells can induce disruption of vascular smooth muscle cells and breakdown of extracellular matrix components as required for spiral artery remodeling ( 62 ). Additionally, uNK cells are known to secrete MMPs, which are required for disruption and loss of cohesion between layers of vascular smooth muscle ( 115 ). In addition to MMPs, uNK cells secrete cytokines such as tumor necrosis factor alpha, IFN-gamma, and IL-10; growth factors such as granulocyte macrophage colony-stimulating factor and placental growth factor; and angiogenic factors such as VEGF-C, angiopoietins 1 and 2, and transforming growth factor beta 1, all of which have been implicated in spiral artery remodeling ( 33 , 61 , 62 , 110 – 112 ). Notably, the production of angiogenic growth factors by uNK cells during pregnancy decreases with increasing gestational age, further supporting a unique temporal role of these cells during spiral artery remodeling ( 61 ). To further explore how uNK cells could functionally impact early pregnancy, several knockout mouse models have been developed that are deficient in NK cells. Such studies demonstrate lack of spiral artery remodeling as well as vacuolated decidua. This results in no development of metrial glands, imperative for ongoing pregnancy in the mouse, as well as markedly decreased placental surface areas compared with normal controls. Pregnancies in these mice demonstrate adverse outcomes including decreased litter size, increased rates of intrauterine demise, intrauterine growth restriction, and offspring with persistent weight deficit into adulthood with severity of such outcomes dependent on the knockout approach ( 118 , 119 ). In another model of mice severely deficient in NK cells, <1% of normal frequency additionally demonstrated implantation site histopathological abnormalities with subsequent fetal death or growth restriction ( 120 ). A subsequent study then transplanted bone marrow from mice with preserved NK cells and found that spiral artery remodeling was restored leading to improvement of fetal viability as well as increase in placenta size ( 121 ). Later literature supported these findings utilizing a mouse NK-depleted model to demonstrate delayed spiral artery development ( 122 ). This group further linked this finding back to trophoblast invasion with demonstration of decreased oxygen tension at site of placentation with subsequent effect on trophoblast lineage decisions ( 122 ). This literature supports the role of uNK cells in spiral artery remodeling through human studies demonstrating not only their location adjacent to spiral arteries but also a number of soluble factors that regulate this process ( Fig. 2C ), while murine studies suggest functional implications.

Natural killer cells are innate lymphocytes well known to be critical for host defense against viruses and tumors ( 31 ). They express an array of activating and inhibitory receptors, the balance of which governs production of cytokines and cytolysis by NK cells. Natural cytotoxicity receptors and killer immunoglobulin-like receptors expressed by NK cells are germline-encoded and recognize certain activating ligands and major histocompatibility complexes, respectively. In contrast, adaptive immune lymphocytes express individually rearranged T and B cell receptors that allow the host to react against virtually any antigen. Further, while adaptive immune lymphocytes predominate in such traditional lymphoid tissues as lymph nodes and the spleen, NK cells are overrepresented in the uterus. Natural killer cells present in uterine tissue both before and during pregnancy are unique cell populations with divergent characteristics compared with their peripheral blood counterparts. Peripheral blood NK (pb NK) cells are cytotoxic in nature and play an integral role in the clearance of virally infected and tumor cells ( 32 ). Uterine NK cells display distinct differences in surface marker expression—while pb NK cells are predominantly CD56 dim and CD16 bright , uNK cells are CD56 bright and CD16 dim ( 33 ). While these cells contain cytolytic molecules, their cytotoxic effect is not applied to classical target cells in vitro, playing a permissive, rather than a defensive, role ( 27 , 34 , 35 ). Most data exploring uNK cells have been generated utilizing pregnancy uNK cells isolated from first trimester termination tissue ( Tables 1 and 2 ). Fewer studies have examined the number and function of prepregnancy uNK cells, present in the endometrium before trophoblast arrival ( 36 ). This is a significant distinction as the immune cells that regulate early EVT invasion and placental establishment are likely those that accumulate in the endometrium during the secretory phase and may differ from those present in the first trimester. Supporting this, a study by Kopcow et al. ( 37 ) showed distinct gene expression differences between NK cells from cycling endometrium and first trimester NK cells, and both of these cell types differed from circulating peripheral NK cells. Another study by Manaster et al. ( 35 ) demonstrated phenotypic and functional differences between the prepregnancy and pregnancy uNK cell populations: prepregnancy NK cells possessed a different repertoire of chemokine receptors and were not cytotoxic until activated by interleukin (IL)-15. These investigators go on to speculate that prepregnancy NK cells are nonfunctional but poised to act after pregnancy. However, the investigators in this study focused specifically on cytotoxicity and the cytokines known to be expressed by NK cells obtained during pregnancy. Current data suggest a much broader secretome from these NK cells, and a wider study of the function of prepregnancy NK cells is still needed. Uterine NK cells are not a static population of cells but undergo cyclic shifts in abundance during the menstrual cycle, with the population enriching during the window of implantation and during early pregnancy, when trophoblast invasion occurs. In nonpregnant endometrium, uNK cells increase drastically between the proliferative and secretory phases, going from 38.9% of all lymphocytes to 83.2% of all lymphocytes by the late secretory phase ( 38 ). If conception occurs, this increase in uNK cell number lasts through the first trimester of pregnancy and then decreases in the late second and early third trimester after cessation of trophoblast invasion around 20 weeks ( 23 – 26 , 38 , 39 ) ( Fig. 1 ). Enrichment of prepregnancy uNK cells is independent of the presence of an embryo, as evidenced by biopsies in nonpregnant decidualized endometrium as well as in sampling of decidualized endometrium from women with ectopic pregnancy ( 40 ). Notably, if pregnancy is not established during a given cycle, the uNK cells simply apoptose ( 41 ). These cyclic changes in population have led to the study of their function in endometrial preparation for pregnancy as well as early pregnancy implantation. The origin of uNK cells is an area of intense scrutiny. Although no clear consensus currently exists, it remains unclear whether uNK cells originate from pb NK cells that migrate into the uterus or whether they originate in situ. In support of recruitment from the peripheral circulation, cytokines and chemokines secreted by endometrial cells have been suggested to function in recruitment of peripheral NK cells. Specific factors such as CXCL12, CXCL9, CXCL10, CCL3, and CCL4 are produced by decidual stromal cells, uterine endothelial cells, and fetal trophoblast cells that induce chemotactic responses in pb NK cells ( 42 – 44 ). In vitro, exposure of pb NK cells to factors such as transforming growth factor beta 1, hypoxia, or stromal cell conditioned medium leads to pb NK cell acquisition of phenotypic and functional characteristics of uNK cells ( 45 , 46 ). Examination of bone marrow transplant recipients additionally demonstrated the presence of donor-derived cells capable of generating endometrium ( 47 ). More recent literature from uterus transplant recipients demonstrates uNK cells after transplantation identified as recipient in origin on the basis of human leukocyte antigen (HLA) expression, indicating that these cells may be replenished from peripheral circulation over time ( 48 ). While the aforementioned data support peripheral immune cell recruitment, an elegant study by Matsuura-Sawada et al. ( 49 ) showed an increase in human uNK cells when human proliferative endometrium was transplanted subcutaneously into immunodeficient mice lacking NK cells. This suggests that uNK cells could arise in the endometrium and proliferate from in situ hematopoietic precursors. Studies show that nonpregnant endometrium and first trimester decidua both contain these precursors ( 50 – 52 ), and NK cell differentiation factors are produced by the endometrium in a cyclic manner that coincides with uNK cell abundance ( 53 ). Regardless of uNK cell origin, it seems clear that hormonal regulation plays a significant role in NK cell accumulation. In vitro, pregnancy uNK cells obtained during the first trimester demonstrate increased migration in response to estrogen ( 54 ). Using in vitro cultured human endometrial tissue samples, estradiol and progesterone promote chemotactic marker expression, supporting a role for these hormones in uNK cell accumulation ( 55 ). Complicating the distinction between prepregnancy and pregnancy uNK cells even further, a recent study profiling first trimester maternal decidua discovered multiple coexisting uNK cell subtypes. Using single-cell RNA sequencing, Vento-Tormo et al. ( 56 ) defined dNK1, dNK2, and dNK3 populations in the decidua on the basis of unique surface marker expression, with each cell type demonstrating a unique transcriptome. While the function of each subset is unknown, dNK1 cells express transcripts that suggest a role in EVT recognition and interaction; dNK2 cells potentially have anti-inflammatory functions; and dNK3 cells could play a role in EVT regulation. A second study that similarly profiled the transcriptome of maternal–fetal interface separated uNK cells present in the first trimester into “resting” NK cells, which represented most uNK cells present, and “proliferative” uNK cells that were nearly sevenfold fewer and were enriched in genes involved in cell cycle processes and cell proliferation ( 57 ). The earlier described subsets were corroborated by a recent study that distinguished these populations using protein surface markers via mass cytometry ( 58 ). In addition, other groups have described heterogeneity in the first trimester uNK cell population: pregnancy trained decidual NK cells that confer “trained memory” were recently described, existing only in multigravid women ( 59 ). These cells produce increased levels of factors supporting vascularization, and the investigators speculate that they originate from precursors present in the endometrium during the menstrual cycle. Fu et al. ( 60 ) identified a subset of first trimester uNK cells that express CD49a and Eomes and produce growth-promoting factors such as pleiotrophin, osteoglycin, and osteopontin that play a role in fetal development. Patients with recurrent pregnancy loss (RPL) contain fewer growth-promoting factor+ uNK cells, and lack of these cells results in fetal growth restriction in mice. Another significant characteristic of uNK cells is the impact of gestational age on their function, although few studies have explored this phenomenon, possibly owing to the difficulty in obtaining human tissue. Angiogenic factors such as Ang1, Ang2, and vascular endothelial growth factor (VEGF)-C are secreted at higher levels by uNK cells obtained between 8 and 10 weeks of pregnancy compared with 12–14 weeks ( 61 ). This is accompanied by corresponding functional differences: vascular smooth muscle cell disorganization, a hallmark of spiral artery remodeling, was only observed using supernatant from 8–10-week uNK cells and not 12–14-week uNK cells in a chorionic plate artery in vitro model ( 62 ). Interestingly, examining first trimester uNK cell gestational age effects on EVT invasion and apoptosis yielded the opposite conclusion: 12–14-week uNK cell supernatant increased EVT invasion compared with 8–10-week uNK cell supernatant. These data suggest that uNK cell function during pregnancy changes throughout gestation. This could reflect changing requirements for extent of trophoblast invasion and/or artery remodeling required at different stages of pregnancy and is a significant variable to consider in analyzing data and future studies. The heterogeneity identified by the aforementioned studies could truly represent distinct uNK cell subpopulations or depict progressive stages in uNK cell development, represent multiple functions of a single population, or arise from regional differences in localization within the decidua. It remains to be seen if these subsets exist before conception.

Uterine NK cells dominate the maternal uterine immune cell environment and accumulate in the uterus at timepoints critical for embryo implantation and placentation. As discussed in this review, emerging studies point to an active involvement of these cells in processes necessary for appropriate placentation: stromal cell decidualization, trophoblast invasion, and spiral artery remodeling. In addition, there has been considerable evidence suggesting that disturbances in uNK cell number and/or function play a role in disorders of placentation. Prematurely, clinical tests have already emerged to measure NK cell abundance as a marker for successful pregnancy outcome ( 123 , 124 ). It is critical that we first understand the potentially changing role of NK cells throughout pregnancy to thoroughly comprehend the biologic bases of these pathologies and eventually translate these to clinical practice. Uterine NK cells are one of many significant maternal cell types that interact with and potentially regulate EVTs during early pregnancy ( 125 ). To properly investigate their function in a physiologically relevant manner, we must use systems that incorporate other maternal as well as fetal cell types. These studies, however, prove challenging because of a lack of animal and in vitro models that accurately replicate embryo implantation and early human placentation as well as ethical considerations around in vivo human studies. Animal models currently used, including mice, rats, guinea pigs, and primates, are unable to recapitulate the depth of invasion and cellular complexity seen in human placentation ( 126 – 131 ). In vitro studies most often examine uNK cells with just one other fetal or maternal cell type, failing to mimic the complex interplay between EVTs and the multicellular maternal environment ( 132 – 134 ). Human tissue studies often use the term placenta, which has a significant cellular difference; placental bed biopsies, which do not include all relevant cell types; or hysterectomized uteri, which often have underlying pathology ( 2 , 135 ). It is critical for us to develop novel methods for studying early placentation to expand our understanding of the cells and factors that regulate this critical process. While these issues are common to other studies examining implantation and placentation, the current work studying uNK cells involves specific limitations as well. Foremost is the extrapolation of behavior from peripheral immune cells despite unique characteristics of uNK cell populations ( 136 , 137 ). Additionally, uNK cells from animal models show distinct developmental, phenotypic, and functional differences compared with human uNK cells, reducing the translatability of findings in animal models ( 137 ). In addition, functional readouts when studying uNK cells are unclear, since cytotoxicity assays such as chromium release assay, considered the “gold standard” in testing NK cell function, may not be relevant to the function of uNK cells ( 138 ). Finally, obtaining nonpregnant human tissue during the window of implantation and first trimester termination tissue is challenging. Without human tissue, the study of secretory-phase NK cells becomes extremely limited because mouse uteri, the most commonly used animal model in the study of early pregnancy, do not contain mature NK cells before implantation ( 139 ). Current studies should be interpreted keeping in mind these difficulties, and developing a physiologically relevant in vitro alternative to study uNK cell function is a critical next step. Although this review brings to light a body of evidence indicating a significant role of uNK cells in early pregnancy establishment, the remainder of the immune cell environment merits attention. Comparison of the immune environment between late proliferative and midsecretory endometrium demonstrates upregulation of immune-related genes. This includes genes related to not only proliferation of uNK cells but also inhibition of T cell growth and the classical complement pathway ( 140 ). Studies have demonstrated that, like NK cells, T regulatory cells additionally increase in numbers within the endometrium before embryo implantation and during pregnancy ( 141 ). Others have shown that the overall number of T regulatory cells is decreased in women with recurrent miscarriage and numbers may in addition be linked to the development of preeclampsia ( 142 ). These findings remind us of the complexity of crosstalk required for normal pregnancy establishment and ongoing pregnancy, with a need for expansion to study the entirety of the immune cell environment within the uterus. In conclusion, early pregnancy implantation is a key element of healthy pregnancy establishment, which remains incompletely understood. Literature points to a likely essential role of the immune system in this process, particularly as seen here with the unique subpopulation of uNK cells. Evidence suggests that these uNK cells are involved in endometrial decidualization, spiral artery remodeling, and trophoblast invasion, all of which are essential for normal placentation. Shortcomings in each of these essential steps for implantation can lead to adverse perinatal outcomes such as hypertensive disorders of pregnancy, intrauterine growth restriction, and RIF. Therefore, determining failures with regard to uNK influence on these processes may provide new targets for medical intervention. Further research is needed to clarify the mechanism of action of uNK cells in this domain to guide such intervention. To accomplish this, special care must be taken to not generalize uNK cells as one homogenous population and distinguish between NK cell subtypes in any future investigation. Finally, this field would greatly benefit from in vitro model systems that accurately capture the complexities of the maternal–fetal interface.

The clinical relevance of uNK cell abundance and function to infertility and adverse pregnancy outcomes remains in dispute. This may be due largely in part to heterogeneity of the available studies ( Table 3 ). For example, some studies demonstrate a decrease in uNK cells on examination of endometrial biopsies in the secretory phase of women with infertility ( 63 – 65 ) and a decrease in CD56 bright uNK cells in women with miscarriage after assisted reproductive technology (ART) compared with those with live birth after ART ( 66 ). However, other data demonstrate no difference in uNK abundance ( 67 , 68 ) or increased uNK progenitor cells in the endometrium of infertile women ( 50 ) on the basis of endometrial biopsy at time of elective surgery. For those studies focusing on recurrent implantation failure (RIF) after in vitro fertilization, most studies demonstrate no difference in abundance of uNK cells in women with RIF compared with those with history of RIF with subsequent pregnancy ( 69 ), historical controls ( 70 ), or fertile women ( 71 , 72 ). However, one recent analysis of a relatively small cohort of women demonstrated a lower number of secretory uNK cells when obtained by uterine lavage in women with RIF compared with infertile women without RIF ( 73 ). To add further confusion to the subject, two more studies demonstrated an increase in uNK abundance in women with RIF ( 74 , 75 ) with another demonstrating an increase in peripheral-type NK cells within the uterus ( 76 ). The greatest volume of literature with regard to evaluating uNK cell abundance and early pregnancy outcomes exists in the literature studying RPL. Although most of these studies are relatively small, with 20 or fewer subjects per arm, there are two larger studies investigating 71 and 87 women with RPL compared with 20 and 10 controls, respectively ( 77 , 78 ). The first of these larger studies demonstrates an increase in uNK cells, but this was quantified in placental tissue utilizing immunohistochemistry at the implantation site after the demise of the pregnancy as opposed to prepregnancy uNK cells ( 77 ). The second of these studies evaluated prepregnancy uNK cells by midsecretory endometrial biopsy and similarly found an increase in uNK abundance in women with RPL vs. normal controls ( 78 ). In examining smaller studies evaluating this population, most conclude that there is no difference in uNK cell abundance in RPL ( 68 , 79 – 83 ). Of those studies remaining, most demonstrate an increase in uNK abundance in RPL ( 84 – 87 ) although all but one of these is on the basis of placental tissue as opposed to endometrial tissue. Two additional studies demonstrate lower uNK cells in this population, although one of these studies evaluated uNK abundance in menstrual blood ( 64 , 88 ). Finally, none of these studies examined uNK cell subtypes, and it remains unknown whether alterations in uNK cell abundance can be predictably correlated to early adverse pregnancy outcomes. Aberrant NK cell number and function have been associated not only with pregnancy establishment but also with disorders of placentation that may manifest later on in pregnancy such as preeclampsia, intrauterine growth restriction, and placenta accreta. Among these, the most well studied is preeclampsia. Although the symptoms of preeclampsia develop toward the end of gestation, it has been suggested that abnormal trophoblast invasion during the first trimester is responsible for this phenotype. Histologic examination of placental bed biopsies demonstrates spiral artery changes of the placental bed in hypertensive pregnancies ( 89 , 90 ). Placental bed biopsies performed at time of cesarean section in normotensive patients as well as those with preeclampsia with severe features demonstrated trophoblast invasion in 100% of the decidual spiral arteries in the normotensive patients but only 44% in those with severe preeclampsia ( 91 ). Several studies have attempted to correlate these changes with immune cell number and function. Late first trimester pregnancies with higher uterine artery resistance indices, implicated in the pathophysiology of preeclampsia, show uNK malfunction and decreased expression of HLA-binding cell-surface receptors integral to normal EVT invasion ( 92 , 93 ). Although it is difficult to demonstrate a causative relationship between immune cell function and hypertensive disorders, animal data has supported this hypothesis. Data from BPH/5 mice, which develop late gestational hypertension and proteinuria that resolve on delivery, among other features of preeclampsia, showed a causative relationship between regulatory factors that controlled immune cell abundance, NK cell abundance, and the development of placentation abnormalities ( 94 ). Others have demonstrated that decreasing pregnancy-related factors known to promote recruitment and activation of uNK in the mouse leads to failure of spiral artery remodeling and reendothelialization, hallmarks of the pathophysiology of preeclampsia, in the setting of reduced uNK numbers ( 95 ). Other disorders of placentation, such as placenta accreta, which is associated with excessive invasion of trophoblasts, have additionally been linked with disordered trophoblast invasion and immune cell function. Placenta accreta specimens demonstrate a multitude of abnormalities including imbalance of cytotrophoblasts and decidua, failure of spiral artery remodeling, and predominance of immature-appearing EVTs ( 96 ). A study of gravid hysterectomy cases found that accreta, increta, and percreta cases demonstrated fewer remodeled spiral arteries as well as many with incomplete remodeling ( 4 ). In addition, this study noted increased depth of vascular remodeling as well as deeper invasion of interstitial trophoblasts in the uterine wall ( 4 ). Early human data demonstrates a possible quantitative difference in pregnancy NK cells in placenta accreta spectrum compared with normal placentation ( 97 ). More recent evidence demonstrates a link between uNK dysfunction and disordered placentation via Grb2-associated binding protein 3 (Gab3). Gab3 has known function in NK cell priming and expansion; therefore, Sliz et al. ( 98 ) utilized a Gab3-deficient mouse model to evaluate pregnancy implications. The resultant impairment in uNK expansion in these mice led not only to increased trophoblast invasion but also to adverse pregnancy outcomes including retained placenta, maternal hemorrhage, and undelivered fetoplacental units at term ( 98 ).