Activation of natural killer cells in the pathogenesis of preeclampsia

Background: Preeclampsia (PE) is a serious pregnancy-specific systemic inflammatory disorder. The characteristic pathological abnormality in PE is impaired uterine spiral artery (SA) remodeling. Natural cells (NK) cells have been proposed to play an important role in uterine SA remodeling, particularly in early pregnancy. Still, the definitive role of NK cells in later phases of PE is largely unstudied. In the present study, we investigated the association between NK cells and PE in late human pregnancy. Methods: Collected normal control (n = 18), late-onset (n = 28) and early-onset PE patients peripheral blood and decidua, isolated mononuclear cells, the ratio of NK cells and the level of intracellular cytokines production were evaluated using Flow cytometry. Co-culture first trimester cytotrophoblast cells with conditioned medium (CM) from decidual NK (dNK) cells, evaluated cytotrophoblast cells migration, invasion, NK cytotoxicity and soluble factors secreted by dNK cells. For multiple group comparisons, data were analyzed using one-way analysis of variance with Bonferroni post-testing when the variances were homogeneous or with Tamhane’s T2 post-testing when the variances were not homogeneous. P < 0.05 was considered significant. Results: We found that the numbers of both peripheral blood NK (pNK) cells and dNK cells and intracellular interferon (IFN)-γ, perforin and granzyme B production were significantly higher in PE compared with normal pregnancies at the time of delivery for both early-onset and late-onset disease. dNK cells from PE pregnancies not only killed primary first trimester trophoblast cells in vitro but also inhibited their migration and invasion when compared to normal controls. Using recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin (IL)-8, IFN-γ and tumor necrosis factor (TNF)-α, and neutralizing Abs against these factors, we demonstrated that GM-CSF and IL-8 in the CM from dNK cells promoted, while IFN-γ and TNF-α

and invasion.

Conclusion:
Our results suggest that activation of NK cells may play an important role in the pathogenesis of PE in late pregnancy. Elevated local levels of dNK cell-derived IFN-γ and TNF-α and decreased GM-CSF and IL-8 in preeclamptic pregnancies may directly mediate known reductions in trophoblast cell migration and invasion in deciduae from preeclamptic pregnancies. Background Preeclampsia (PE) is a serious pregnancy-specific disorder that affects approximately 3-8% of pregnancies and is defined by the development of maternal hypertension and proteinuria after 20 weeks of gestation [1]. PE is a systemic inflammatory disease that may lead to maternal multi-organ damage and fetal growth restriction [2]. The disorder is a leading cause of maternal and fetal morbidity and mortality, accounting for an estimated 14% of pregnancy-related maternal deaths and 15% of premature births worldwide [3,4].
The effects of PE can last long beyond pregnancy for both mother and child. In fact, the long-term risks of cardiovascular, cerebrovascular, and renal diseases are increased in women with a history of PE [3][4][5].
Despite well-recognized public health significance and intensive investigation, the pathogenesis of PE is still poorly understood. Delivery of the placenta remains the only effective treatment for PE, indicating that the placenta plays an important role in disease pathogenesis. One key to successful human pregnancy is a well-described and dramatic remodeling of the uterine spiral arteries (SAs) [6]. In early pregnancy, progenitor cytotrophoblast cells at the tips of anchoring villi differentiate into extravillous trophoblast cells (EVTs) that invade the uterine SAs of the decidua and myometrium and replace the endothelial layer of these uterine vessels. This transforms uterine SAs from high-resistance vessels to flaccid, high-capacitance vessels [6]. Such changes allow the 4 increased blood supply needed to sustain the growing fetus and expanding placenta throughout pregnancy. Characteristic pathological changes in PE include deficient EVT invasion and incomplete uterine SA remodeling.
Natural killer (NK) cells are a subset of lymphocytes of the innate immune system characterized by high cytolytic potential against MHC class I negative virus-infected and tumor-transformed cells [7,8]. NK cells constitute the major immune cell type in the decidua during the first trimester of human pregnancy, accounting for approximately 70% of the local lymphocytes [9], although these decidual NK (dNK) cells appear to have less cytolytic potential and more cytokine and chemokine secretary activities than their peripheral counterparts [10]. The abundance of NK cells in the decidua has suggested that these cells may play an important role in pregnancy support and placental development.
Decidual NK cells have been proposed to contribute directly to the initiation of uterine SA remodeling [11] by secreting numerous cytokines, angiogenic factors and enzymes, including matrix metalloproteinases, that disrupt vascular extracellular matrix connections [12][13][14]. In addition, a number of studies have shown that dNK cells have indirect effects on uterine vessel remodeling by modulating EVT growth, differentiation, migration and invasion [15,16]. Moreover, dNK cells from pregnancies at high risk of PE are less able to induce EVT motility and fail to induce vascular cell apoptosis when compared to dNK cells isolated from low-risk pregnancies, providing functional evidence of altered dNK cell function in PE [17]. In contrast, it has also been reported that dNK cells are a source of decidual interferon (IFN)-γ during early human pregnancy and likely participate in the inhibition of EVT invasion [18,19].
Although a possible etiological role for NK cells in the pathogenesis of PE has long aroused the interest of scientists, their role in later phases of PE is largely unstudied. In addition, there are conflicting reports regarding the numbers of dNK cells in PE deciduae. Some 5 reports show an increase in dNK cell number in PE, albeit with an altered phenotype [20,21], whereas others report decreased dNK cell numbers in preeclamptic placental samples [22,23]. Increased secretion of angiogenic factors by dNK cells from pregnancies with high uterine artery resistance has also been reported [24].
In short, we continue to search for the definitive role of NK cells in the pathogenesis of PE, particularly in late pregnancy. In the present study, we examined NK cell frequencies and function between normal and preeclamptic pregnancies. We report that in the third trimester, the numbers of both peripheral blood NK (pNK) cells and dNK cells, as well as intracellulalr IFN-γ, perforin and granzyme B production are significantly increased in PE compared with normal pregnancies. Functional analyses revealed that dNK cells from PE pregnancies at late gestation kill primary first trimester trophoblast cells in vitro. These dNK cells also inhibit primary trophoblast cell invasion and migration, and the effects may be due to increases in dNK cell IFN-γ and tumor necrosis factor (TNF)-α, and decreases in granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin (IL)-8. We hypothesize that activation of NK cells may contribute to the pathogenesis of PE in late human pregnancy.

Collection of human tissues
This study was approved by the Medical Ethics Committee of Guangzhou First People's Hospital, Guangzhou, China. Informed written consent was obtained from each participant prior to participation. Placental tissues (8-10 weeks; n = 20) from women with normal pregnancies undergoing elective terminations of pregnancy for nonmedical reasons were obtained. In addition, we collected third trimester peripheral blood and decidual tissue samples from 42 women with early-onset or late-onset PE (14 and 28, respectively) and 18 6 healthy pregnant women at the time of delivery. All participants delivered by caesarean section without prior labor. Clinical characteristics of these participants are shown in Peripheral blood was obtained just before delivery by venipuncture using heparin as an anticoagulant to isolate peripheral blood mononuclear cells (PBMCs). All samples were processed immediately following collection.

Isolation of human cytotrophoblast cells
Cytotrophoblast cells from human first trimester [25] and term placentae [26] have been successfully isolated. Here, we isolated cytotrophoblast cells from human first trimester placentae using similar methods with minor modifications. Briefly, villous tissues were

Statistical analyses
All statistical analyses were performed using SPSS 23.0 software (IBM, Armonk, NY, USA).
For multiple group comparisons, data were analyzed using one-way analysis of variance with Bonferroni post-testing when the variances were homogeneous or with Tamhane's T2 post-testing when the variances were not homogeneous. Results were expressed as means ± SDs. A P-value of < 0.05 was considered significant.

Association between PE and dNK cell frequencies and intracellular IFN-γ, perforin and granzyme B production during late pregnancy
We determined the frequencies of dNK cells and levels of intracellular IFN-γ, perforin and granzyme B production using flow cytometry in women with PE in the third trimester. As shown in Figure 1A-H, the frequencies of dNK cells and the levels of dNK cell intracellular IFN-γ, perforin and granzyme B were all significantly higher in late-onset PE when compared to normal pregnancies (P < 0.01 for all comparisons), but significantly lower than those in specimens from women with early-onset PE (P < 0.01 for all comparisons).

Association between PE and pNK cell frequencies and intracellular IFN-γ, perforin and granzyme B production during late pregnancy
To complement our comparisons of dNK cell numbers in early and late onset PE and control pregnancies, we used flow cytometry to assess and compare the number of maternal pNK cells in women with these conditions. The number of pNK cells in late-onset PE patients was significantly higher than that in normal pregnancies (P < 0.01) and significantly lower than that in early-onset PE (P < 0.01, Figure 2A and E). Similar results were noted for the levels of pNK cell intracellular IFN-γ, perforin and granzyme B (P < 0.01 for all comparisons, Figure 2B-D, F-H).

Immune activation of dNK cells in PE during late pregnancy
Since dNK cells play an important role in regulating trophoblast invasion and migration [15,16], we analyzed the effects of dNK cells on these processes. As shown in Figure 3A   13 and B, while CM from dNK cells from women with late-onset PE significantly decreased the invasion and migration of first trimester trophoblast cells when compared with normal pregnancies (P < 0.01 for both comparisons), the effect was even more robust when using CM from early-onset PE dNK cells (P < 0.01 for both comparisons). Additionally, dNK cell anti-trophoblast cytotoxicity was highest for dNK cells isolated from women with earlyonset PE, lower for those isolated from late-onset PE subjects, and lowest when the source was the deciduae of unaffected control pregnancies (P < 0.01 for all comparisons, Figure   3C).

Levels of soluble factors in dNK cell CM
Since CM from dNK cells isolated from normal and preeclamptic pregnancies demonstrated different effects on trophoblast cell migration and invasion, we hypothesized that these effects may be related to differences in the secretion of biologically active factors and that such changes might be detected by comparing dNK cell CM of normal and preeclamptic pregnancies. We simultaneously analyzed 10 candidate soluble factors in dNK cell CM using multiplex immunoassays. As demonstrated in Figure 4, levels of GM-CSF and IL-8 in dNK cell CM from women with late-onset PE were significantly lower than those in samples from normal pregnancies (P < 0.01 for both comparisons), but significantly higher than those in samples from early-onset PE (P < 0.01 for both comparisons). In contrast, concentrations of IFN-γ and TNF-α in dNK cell CM from late-onset PE were significantly higher than those in specimens from normal pregnancies (P < 0.01 for both comparisons), but significantly lower than those in specimens from early-onset PE (P < 0.01 for both comparisons). Levels of IL-1β, IL-2, IL-4, IL-5, IL-6 and IL-10 were low in all samples and no significant differences were detected among normal and preeclamptic pregnancies.

Effects of soluble factors secreted by dNK cells on trophoblast cell migration and invasion
Having In similar experiments, since dNK cells from early-onset PE secreted a considerable amount of IFN-γ and TNF-α, we examined the effects of these factors on trophoblast cell migration and invasion using recombinant human IFN-γ and TNF-α, and neutralizing Abs against IFN-γ and TNF-α. As shown in Figure 5C,

Discussion
Although great progress has been made during the past decade toward our understanding of the pathogenesis of PE, several formidable investigative challenges leave much about its etiology and pathophysiology unclear [27]. First, while the disease is associated with inadequate EVT invasion and insufficient uterine SA remodeling beginning in early pregnancy, significantly delayed symptom onset in the late second or early third trimester [28] makes etiologic studies in humans daunting. Further, even if one could definitively identify in early pregnancy those women who would ultimately develop signs and symptoms of PE, access to relevant placental or decidual tissues for single or multiple time point investigations is essentially impossible. Finally, PE is considered to be a disease specific to humans. Animal models in which the disease spontaneously occurs are lacking [29] and animal models utilizing induction techniques do not typically demonstrate all disease manifestations [30].

16
A two-stage model for PE pathogenesis has been proposed [31,32]. The first stage involves reduced EVT invasion and inadequate uterine vessel remodeling in the deciduae.
These structural abnormalities are, in turn linked to insufficient and hyperactive uteroplacental circulation that causes overall placental hypoxia with interspersed periods of intermittent hyperoxia [31]. The resulting hypoxic and oxidative damage promotes a second stage of release of a series of proinflammatory factors and vasoactive factors from the placenta into maternal circulatory system that cause the clinical syndrome in late pregnancy [33,34]. In the present study, we investigated the role of NK cells in the pathogenesis of the second stage of PE when the maternal inflammatory responses cause the majority of disease signs and symptoms.
We divided women with PE into early-onset and late-onset PE groups, a subclassification that often correlates with disease severity [35]. Using flow cytometry, we found that the numbers of NK cells in the peripheral blood of women with both early-onset and late-onset PE were significantly higher than those in normal control pregnancies and that the extent of these changes was highest in early-onset PE.
We also assessed NK cells within the deciduae of samples obtained at the time of delivery in our control and PE subjects. These cells were present in normal pregnancies at term.
Their functional role at this stage of gestation is not known and will be the topic in future investigations. Like our results for pNK cells, the percentages of NK cells in the decidual basalis of preeclamptic pregnancies was significantly higher than those in normal pregnancies, and the effect was most marked in women with early-onset PE. This supports prior immunohistochemical analyses of frozen decidual tissues [20] showing that CD56 + dNK cell numbers were higher in women with PE when compared with normal term control pregnancies. Similarly, Wilczyński et al. [21] used flow cytometry to demonstrate that preeclamptic patients had a higher percentage of CD3 − CD56 + CD16 + NK cells in the deciduae compared with age-matched term control pregnancies.
In contrast, using immunohistochemical analyses of frozen tissues, Williams et al. [22] reported a reduction in dNK cells in placental bed biopsies from women with PE compared with control third trimester deciduae. Further, using immunofluorescence and frozen tissues, Lockwood et al. [23] reported that deciduae from women with PE displayed significantly lower dNK cell numbers than term controls. Likewise, Rieger et al. [36] have shown that the number of decidual CD56 + CD16 + NK cells was lower in deciduae from women with PE than that in gestational age-matched preterm deliveries caused by preterm labour, multiple gestation, fetal distress, intrauterine fetal growth restriction, fetal abnormalities, placenta praevia or uterine rupture. The reasons for the discrepancies between our results and these latter findings are not immediately evident, but may result from differences in study design, subject selection, specimen origin (e.g. decidua basalis vs. placental bed biopsies), reagent specificity, sample size and/or analytic methods.
We next turned to more functional analyses and reported that the production of intracellular IFN-γ, perforin and granzyme B by both pNK and dNK cells is higher in pregnancies affected by PE when compared to controls and that the effect is most marked in women with early-onset PE. This suggests that both systemic NK cells and NK cells at the maternal-fetal interface are hyperactivated in PE and the degree of activation correlates positively with disease activity. More detailed mechanisms underlying NK cell activation in PE is currently under investigation. One hypothesis is that dNK cells may become activated and dysfunctional under the stress of pathological placental hypoxia [37]. To this point, a conversion of tolerogenic dNK cells to a largely cytolytic phenotype [38] was reported in affected rats exposed to reduced uterine perfusion pressure to model human PE. This shift was associated with higher mean arterial pressures, fetal intrauterine growth restriction and increased inflammation, changes that were reversed by the depletion of NK cells [38].
In the present study, activated dNK cells from subjects experiencing PE not only killed first trimester trophoblast cells at a higher level than those from unaffected women, but also inhibited trophoblast cell invasion and migration to a greater extent. All dNK samples were One limitation of our study is that we selected for study only 10 candidate soluble factors in dNK cell CM. Other factors produced by dNK cells may have effects on trophoblast cell function of equal or greater importance than those studied here and these effects will have gone undetected. As with many studies on PE at delivery, there was a gestational age difference in the time of sample collection among disease and control groups.
Unfortunately, choice of an appropriate gestational age-matched normal control population is difficult as premature deliveries are essentially never normal.

Conclusions
Activation of both dNK cells and pNK cells may contribute to the pathogenesis of symptomatic PE in late pregnancy. Inhibition of trophoblast cell migration and invasion by dNK CM from preeclamptic pregnancies may be due to elevated production of IFN-γ and TNF-α and decreased production of GM-CSF and IL-8 by dNK cells in affected pregnancies.

Ethics approval and consent to participate
The study was approved by the Medical Ethics Committee of Guangzhou First People's Hospital, School of Medicine, South China University of Technology. Written informed consent was obtained from the study participants prior to their enrollment.

Consent for publication
Not applicable

Availability of data and materials
The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.