|Home | About | Journals | Submit | Contact Us | Français|
Although replete with cytotoxic machinery, uterine NK (uNK) cells remain tolerant at the maternal-fetal interface. The mechanisms that facilitate the uNK cell tolerance are largely unknown. Here we demonstrate that VEGF C, a pro-angiogenic factor produced by uNK cells, is responsible for their non-cytotoxic activity. VEGF C-producing uNK cells support endovascular processes as demonstrated in a three dimensional co-culture model of capillary tube formation on matrigel. Peripheral blood NK cells fail to produce VEGF C and remain cytotoxic. This response can be reversed by exogenous VEGF C. We show that cytoprotection by VEGF C can be related to induction of the “transporter associated with antigen processing (TAP)-1” expression and MHC class I assembly in target cells. siRNA-mediated silencing of TAP-1 expression abolished the VEGF C-imparted protection. Overall, these results demonstrate that empowerment of uNK cells with angiogenic factors keeps them non-cytotoxic. This phenotype is critical to their pregnancy compatible immuno-vascular role during placentation and fetal development.
Maternal immune tolerance of the semi-allograft fetus is a subject of intense debate. Although the decidualized uterus is replete with immune cells, fetal development ensues unhindered and results in normal term delivery. An important aspect of successful pregnancy outcome is that they invoke immune tolerance and de novo angiogenesis, two highly intriguing, indispensable and well choreographed processes (1–3). These processes are central to optimal placentation and involve interactions between immune cells, endothelial cells, and invading trophoblasts. With respect to immune cells, T cell-mediated rejection of the fetus has been shown in mice, but not in humans (4). Moreover, uterine T cells or macrophages may not promote immuno-vascular cross talk with the invading trophoblast lacking expression of HLA (HLA-A, HLA-B, or HLA-D) antigens (1, 5). Thus, it is reasonable to propose that the predominant immune tolerance cross-talk is between uterine NK (uNK) cells and the trophoblast. Interestingly, uNK cells have been uniquely detected in the proximity of endothelial cells and invading trophoblasts associated with spiral arteries undergoing physiological transformation (6, 7).
How uNK cells induce tolerance to the fetal tissue is an important unresolved question, with implications for our understanding of the role of not only uNK cells but peripheral blood NK (pNK) cells in adverse pregnancy outcomes, tumor surveillance, autoimmunity and immunity against infections. In humans, unlike peripheral blood NK cells of cytotoxic CD56dimCD16+ phenotype, the pregnant uterus (decidua) is populated with specialized uNK cells that are of non-cytotoxic CD56brightCD16− phenotype (8–10). These cells are present in the endometrium during the secretory phase of the menstrual cycle and reach peak numbers (60–70%) during first trimester of pregnancy. Thereafter, their numbers decline rapidly (11–13). Although the physiological role of uNK cells remains poorly understood, several groundbreaking observations and views have implicated these cells in decidualization, invasion of the trophoblast, and production of angiogenic factors and chemokines (14–18). Major mechanistic findings on uNK cells have come from animal studies suggesting their role in spiral artery remodeling. NK cell-deficient mice display abnormalities in decidual artery remodeling and trophoblast invasion, possibly due to lack of uNK cell-derived IFN-γ (3, 14). In humans, lack of “physiological transformation” of decidual spiral arteries is thought to be associated with preeclampsia, preterm birth and intrauterine growth restriction (7, 19, 20). In this regard, evidence has been presented for defective interactions between uNK cells and the invading trophoblast in a cohort of preeclampsia patients (7, 21).
Despite their pregnancy compatible role, uNK cells at the maternal-fetal interface harbor a repertoire of natural cytotoxicity receptors (NCRs), cytolytic granules, activating and inhibitory killer-IgG receptors (KIR) (16,22–24). This suggests that these innate cells retain their primordial role as sentinels with an ability to attack target cells in response to “danger signals” in their proximate milieu. Indeed, recent studies from our laboratory and others have shown that uNK cells “switch on” their cytolytic machinery leading to fetal resorption or preterm birth in mice exposed to lipopolysaccharide (LPS) (25, 26) or dsRNA (27). In addition, increased cytolytic CD16+ NK cells have been reported in the endometrium of patients with a history of recurrent spontaneous abortion and implantation failure (28–30). Other studies have shown that uNK cells cultured in vitro can be readily activated with IL-2 to lyse target cells including trophoblasts (31, 32). Thus, the mechanisms that facilitate pregnancy compatible, non-cytotoxic characteristics of uNK cells need to be delineated. We hypothesized that the signature angiogenic machinery of uNK cells was responsible for their non-cytotoxic behavior during normal pregnancy. We show that this is accomplished through uNK cell-produced VEGF C which enhances resistance to lysis of trophoblasts and endothelial cells through induction of the transporter associated with antigen processing (TAP-1) protein.TAP-1 is associated with transport of processed peptides to the endoplasmic reticulum to ascertain functional MHC class I assembly (33).
All human studies were done with the approval of the Institutional Review Board of Dartmouth Medical School (Hanover, NH) and Women and Infants Hospital (Providence, RI). Informed consent was obtained from all tissue donors.
Mononuclear cells from blood were collected from non-pregnant healthy women 35 ± 5 years of age separated from heparinized blood samples by Ficoll gradient centrifugation as previously described (51). NK cells were further purified using human NK cell isolation kit (Milteyni Biotec, USA) based on depletion of non-NK cells (negative selection) as per the manufacturer's instructions.
First trimester decidua were obtained from pregnant women who underwent elective termination. All subjects were between 7 and 11 weeks gestation. Gestational age was calculated based on last menstrual period or ultrasound at the time of procedure. Decidual tissue was processed for NK cell isolation immediately after extraction as previously described (51). The cells were suspended in RPMI supplemented with 10% fetal calf serum and antibiotics and adjusted to a concentration of 1 × 106 cells/ml. dNK cells were further separated using human NK cell isolation kit (Milteyni Biotec, USA) as described above. Cell viability was confirmed to be >90% by trypan blue staining. The purity of the preparation (>95%) were confirmed by FACS by gating on forward scatter and side scatter dot plots of CD45+ cells to exclude non-cellular debris (51).
CD56brightCD3− uNK cells were isolated from endometrial tissue specimens obtained from women undergoing hysterectomy for various gynecological disorders. Tissue samples used were distal to any pathological changes and were processed to isolate NK cell population as described (38). To generate uNK cell clones, enriched NK cells were cultured in 500 U IL-2/ml for 2–3 days to allow cell expansion in complete media [RPMI 1640 supplemented with 2-mercaptoethanol (50 μM), penicillin (100 U/ml), streptomycin (100 μg/ml), sodium pyruvate (1 mM), nonessential amino acids (0.1 mM), and 5% human serum]. After 2–3 days in culture, uNK cells were cloned using limiting dilution method as described (36). Actively growing cells from wells were analyzed to confirm their NK cell phenotype (CD56+, CD3−), and these uNK clones were expanded and maintained in minimum amounts of IL-2 for the remainder of their culture.
Immortalized first trimester trophoblast cell line HTR8 with properties of invasive extravillous cytotrophoblasts was established and kindly provided by Dr. Charles Graham (52). HTR8 cells were grown to ~80% confluence in RPMI standard growth medium and used only during eight passages. Human umbilical cord endothelial cells (HUVEC) and human uterine endothelial cells (HUtEC) were obtained from Cambrex (East Rutherford, NJ, USA) and cultured in EBM-2 medium (Cambrex, East Rutherford, NJ, USA). All cells were maintained in standard culture conditions of 5% CO2 at 37°C.
We have recently established a three dimensional dual culture system to study endovascular activity involving trophoblasts and endothelial cells (34). This method has been now modified to examine the role of uNK cells in angiogenic processes. Briefly, growth factor-reduced Matrigel (BD Biosciences, San Diego, CA, USA) was thawed overnight at 4°C and mixed to homogeneity. Culture plates (48-well) were coated with 0.1 ml of Matrigel and allowed to gelatinize at 37°C for 30 min. Trophoblasts or endothelial cells (2.5 × 104), labeled with cell tracker green CMFDA or cell tracker red CMTMR (Molecular Probes, Eugene, OR, USA) respectively were co-cultured with pNK or uNK cells (1:1:1) on matrigel coated plates. The spontaneous interaction and endothelial cell-directed tube formation by trophoblasts was monitored and recorded 12–14 hrs after incubation under standard culture conditions using florescence microscopy (Nikon Eclipse TS 100 coupled with CCD camera). The average number of tubes/vacuoles formed was quantified by counting the number of tube like structures formed by connected capillary bridge in four different fields by two independent investigators (34).
We determined the cytotoxic ability of uNK cells and pNK cells on target endothelial (HUVEC, HUtEC) or trophoblast (HTR8) cells using calcein-AM retention assay (38). Briefly, adherent confluent target cells (HUVEC, HUtEC or HTR8; 2×103) were plated in a 96 well plate and labeled with 8 mM calcein-AM (Molecular Probes) in serum- and phenol red-free medium (GIBCO) for 40 min at 37°C. Based on pilot experiments, effector cells (pNK or uNK cells) were added at 25:1 ratio to calcein-AM labeled target cells in quadruplicate. Maximal lysis was determined by solubilizing target cells in lysis buffer (0.1% Triton X-100 in PBS, pH 9.0). After the indicated time (2–3 hr) of incubation at 37°C, the assays were terminated by washing the plates twice and the remaining fluorescence was read using a 96 well fluorescence plate reader (SpectraMax Gemini EM, Molecular Devices) with excitation and emission at 485 and 538nm, respectively. Percent specific cytotoxicity was calculated as described (37). To monitor the cytoprotective activity of VEGF C or VEGF A, the target cells were pre-incubated overnight with cytokines (100 ng/ml) and subjected to cytotoxicity assay.
We cultured 1×106 uNK cell clones (n=22) individually for 48 hrs in RPMI media containing 5 U/ml recombinant IL-15. Measurement of the concentrations of VEGF A, VEGF C, PlGF, IFN-γ, and TNF-α, in supernatants was completed in triplicate using respective ELISA kits (Quantikine Kits, R&D System, Minneapolis, MN, USA) according to the manufacturers' instructions.
We evaluated the surface expression by FACS analysis using flurochrome-conjugated mAb. Cells were harvested and incubated in dark at 4°C for 30 min with antibodies. The antibodies used were anti-class-I HLA (clone W6/32, Abcam), anti-HLA E (MEM-E/08, Abcam), anti-HLA G (MEM-G/9, Abcam), anti-human CD16 (clone 3G8), anti-human CD56, CD45, CD3, NKP46 (BD Biosciences Pharmingen), NKp30, NKp44, NKG2D (BioLegend), VEGF R1 (49560), VEGF R2 (89106), VEGF R3 (54733) (R&D Systems). For HLA-E staining we used the anti-human HLA-E (MEM-E/08) as primary antibody, followed by FITC labeled goat anti-mouse secondary antibody. TAP-1 expression in response to VEGF C (100 ng/ml), IFN-γ (100 ng/ml), VEGF A (100 ng/ml), PlGF (100 ng/ml) or combination of VEGF C and IFN-γ (100 ng/ml each) was carried out by intracellular staining using specific antibody (clone TAP1.28, MBL laboratory, Japan) after fixing and permeablizing the cells as recommended by the manufacturer. Negative controls were performed for each cell type by incubating the cells with isotype-matched antibodies. The stained cells were acquired (10,000 cells) and analyzed by FACS Calibur™ (Becton Dickinson).
We transfected HTR8 trophoblast cells in a six well plate with control siRNA (scrambled) or siRNA against TAP-1 (Santa Cruz Biotechnology, Inc.) using Lipofectamine 2000 (Invitrogen) for 12 hours in OptiMEM I reduced serum medium (Invitrogen). After the transfection period, the cells were allowed to recover in complete RPMI media for another 12 hours. Cells were then harvested for FACS analysis, immunobloting or used as targets for cytotoxicity assay.
We determined the protein expression of TAP-1 in HTR8 trophoblast cells treated with recombinant human IFN-γ (100 ng/ml), VEGF C (100 ng/ml R&D System) or combination of both for 24hrs. The cell lysates were prepared in lysis buffer (10 mM Tris HCl, pH 7.6, 50 mM NaCl, 5 mM EDTA, 50 mM NaF, 0.1 mM sodium orthovanadate, 30 mM sodium pyrophosphate, 1% Triton X-100, 10 μg/ml leupeptin, and 10 μg/ml aprotinin) on ice for 30 min. The cell lysates were separated on 12% SDS–polyacrylamide gels and blotted onto PVDF membranes and probed with anti-TAP-1 antibody (clone TAP1.28, MBL laboratory, Japan) (1:200 dilution) in 1% BSA in PBST. Actin was probed using anti-actin antibodies (Chemicon, USA) and used as an internal loading control. The bands were visualized using horseradish peroxidase conjugated secondary antibodies followed by ECL chemiluminescence (Amersham Biosciences, Piscataway, NJ). The protein bands were recorded using Konica SRX 101A developer.
P values for a pair of data sets were obtained with the two-tailed Student's t-test.
Recent observations have suggested that human decidual NK cells chemoattract trophoblasts and regulate their invasion and participation in vascular remodeling (16). A central question that still remains to be addressed is what factors regulate the immune and vascular activities of non-cytotoxic uNK cells. To address this, we have taken advantage of primary uNK and pNK cells, and short term cultured clones of these cells to assess their ability to influence vascular remodeling. NK cell clones from peripheral blood or endometrial tissue were established and propagated as described in Methods. We utilized a co-culture model of vascular remodeling that involves trophoblast-endothelial cell-NK cell interactions on matrigel and mimics trophoblast invasion of the spiral arteries during pregnancy (34). Using this approach, we demonstrate that HTR8 cells, representing first trimester extravillous trophoblasts, footprint the endothelial cell-guided capillary like tube structures in response to growth factors from endothelial basal media (EBM2) (Fig.1A). Primary pNK cells disrupt the interaction (tube structure formation) between HTR8 cells and HUVEC endothelial cells (Fig. 1B). On the other hand, decidual NK (dNK) cells facilitate formation of definitive tube structures with intense yellow color, a sign of trophoblast-endothelial cell co-localization (Fig. 1C). NK cells are not seen in these panels as they were not marked with a cell tracker.
To further delineate the mechanism(s) for differences between pNK and uNK cells, we employed NK cell clones maintained in the presence of low doses of IL-2. uNK and pNK cell clones were established from 4 different individuals and further characterized as described in Methods. We tested a total of 22 uNK cell clones, 18 of which mimicked primary pNK cells with tube disruption phenotype (data with a representative clone are shown in Fig. 1D), whereas 4 clones facilitated tube formation similar to dNK cells (data with a representative clone are shown in Fig. 1E). A total of 5 pNK cell clones were also tested (data not shown) and the observations were similar to their primary counterparts (Fig. 1B). The average numbers of tubes/vacuoles formed in the presence or absence of “disruptive” or “facilitative” uNK clones were quantified as described in Methods and are presented in Fig. 1H.
Next, we examined production of growth factors (VEGF A, VEGF C, PlGF) and cytokines (IFN-γ, TNF-α) secreted by disruptive and facilitative uNK cell clones. For simplicity, data are shown only for VEGF A, VEGF C and IFN-γ as PlGF and TNF-α were minimally produced as detected by ELISA (Fig. 1I). Although uNK cell clones produced variable amounts of VEGF A and VEGF C, a striking observation was that facilitative uNK cell clones produced 3–4 fold higher levels of VEGF C compared to disruptive clones. Although VEGF C mRNA and protein have previously been shown to be expressed in NK cells from endometrium and decidua (35, 36), the immuno-regulatory role of VEGFs in the uNK cell biology has not been described. On the other hand, no VEGF C mRNA was detected in the human placenta (37), suggesting a unique role of uNK cell-produced VEGF C at the maternal-fetal interface. This prompted us to evaluate exogenous VEGF C and VEGF A for their ability to rescue tube formation in the presence of disruptive uNK cell clones. Interestingly, VEGF C (Fig. 1F), not VEGF A (Fig. 1G), was partially able to rescue tube formation. Quantification of VEGF C-mediated rescue of formation of tube structures is shown in Fig. 1H.
Cytolytic functions are generally associated with CD56dimCD16+ NK cells (8). To rule out disparate behavior of uNK cell clones due to heterogeneity in phenotypic characteristics, we performed FACS analysis for the surface expression of CD16 and CD56 on all uNK cell clones and compared them with pNK cell clones. CD56 expression on uNK and pNK cell clones was found to be bright and dim, respectively (Fig. 2B), in agreement with published observations (38). All the uNK cell clones (n=22) used in the study were found predominantly to be CD16 null (CD56+CD16−), whereas pNK cell clones (n=5) were mostly CD56+CD16+ (Fig. 2A). In both groups, NK cells were CD45+ and CD3− (data not shown). Since the expression of NCRs is a distinguishing feature of NK cells, we monitored the surface expression of NKP30, NKp44, NKp46 and NKG2D. uNK cell clones not only expressed all NCRs irrespective of their ability to disrupt or support tube formation, but the intensity of their surface expression was significantly higher compared to pNK cell clones (Fig. 2C). In addition, freshly isolated pNK cells lacked expression of NKp44. Thus, it is possible that NKp44 expression is regulated by activation stimuli. Since addition of VEGF C-producing uNK cells or exogenous VEGF C was able to restore capillary tube formation (Fig. 1F), we evaluated the expression of VEGF receptors R1, R2 and R3 on uNK cells. As shown in Fig. 2D, none of the VEGF receptors were expressed on either pNK cells or uNK cell clones, suggesting that angiogenic factors secreted by uNK cells act in a paracrine manner on cells expressing these receptors. For a positive control, HTR8 trophoblast cells were used as these cells are known to express all VEGF receptors (Fig. 2D).
Disruption of tube formation by uNK cell clones could result from the cytolytic activity of these cells. This prompted us to assess cytolytic activity of uNK cell clones. Using calcien-AM retention assay (39), we evaluated the cytotoxic effects of uNK and pNK cell clones on HUVEC cells, human uterine endothelial cells (HUtEC), and HTR8 cells. Although non-pregnant endometrium does not contain trophoblasts, the use of extravillous HTR8 cells is justified as the same endometrial NK cells are further amplified in the decidua where they encounter invading trophoblasts. Further, since addition of VEGF C was able to restore tube formation (Fig. 1F), we investigated whether VEGF C or VEGF A would be able to rescue target cells from NK cell-mediated killing. To establish a suitable effector:target (E:T) cell ratio, we performed the assay using a wide range of E:T ratios involving a disruptive NK cell clone and target HUVEC or HTR8 cells (Fig. 3A). Based on these data, we used 25:1 E:T ratio for all subsequent experiments involving all disruptive uNK cell clones (n=18) and pNK cell clones (n=5). Fig. 3, B–G shows data for representative uNK or pNK cell clones. Disruptive uNK cell clones demonstrated robust cytolytic activity towards all target cells, HUVEC (Fig. 3B), HUtEC (Fig. 3C) or HTR8 (Fig. 3D), albeit at varying levels. Similarly, pNK cells were able to robustly lyse HUVEC (Fig. 3E) and to lesser extent HTR8 cells (Fig. 3F). Importantly, pre-incubation of target cells overnight with VEGF C (Fig. 3, B–F) but not VEGF A (Fig. 3G) inhibited NK cell-mediated lysis.
Since both cytotoxic and non-cytotoxic uNK cell clones express a similar profile of cytotoxic machinery (Fig. 2C), it is plausible that VEGF C modifies target cells to escape NK cell-mediated killing. In this regard, induction of TAP-1 expression has been shown to be associated with cytoprotection of microvascular endothelial cells from activated NK cells (39, 40). We found that VEGF C up-regulated the expression of TAP-1 in HTR8 cells and HUVEC cells as confirmed by FACS analysis (Fig. 4, A–B) and by Western blotting (Fig. 5B). This induction of TAP-1 was found to be synergistic with IFN-γ on HTR8 cells (Fig. 4A and Fig. 5B). Quantification of percent TAP-1 positive HTR8 or HUVEC cells in response to VEGF C, VEGF A, PlGF and IFN-γ by flow cytometry is presented in Fig. 4B. Both VEGF C and IFN-γ induced TAP-1 expression in HUVEC and HTR8 although IFN-γ was found to be more potent in HUVEC. On the other hand, IFN-γ elicited less activity in trophoblast HTR8 cells (Fig. 4A and Fig. 5B). Importantly, VEGF A and PlGF failed to induce any significant amount of TAP-1 in either cell type (Fig. 4B).
To further confirm the role of TAP-1 in the VEGF C-induced non-cytotoxicity profile of uNK cells, we used gene specific siRNA to knock-down TAP-1 expression in HTR8 cells induced by IFN-γ, VEGF C or combination of both. As demonstrated by FACS analysis (Fig. 5A) and Western blotting (Fig. 5B), siRNA treatment of cells significantly blocked induction of TAP-1 by VEGF C, IFN-γ or combination of both. Next, we tested whether siRNA-mediated inhibition of TAP-1 expression would reverse the protection profile of target cells. siRNA treatment abolished the protection imparted by VEGF C or VEGF C+IFN-γ in HTR8 cells against NK cell-mediated lysis as indicated by the calcien-AM cytotoxicity assay (Fig. 5C). This clearly confirms both mechanistic and functional importance of VEGF C in priming targets that otherwise would be susceptible to killing by cytotoxic NK cells.
Activation or inhibition of the cytolytic machinery in NK cells is determined by the extent of MHC class I complex on target cells. Because we observed up-regulated TAP-1 expression in HUVEC and HTR8 cells in response to VEGF C, we propose that TAP-1 induction results in greater amounts of MHC class I on treated cells. In this regard, we analyzed HLA-E expression which engages the CD94-NKG2A inhibitory or CD94-NKG2C activating receptors on NK cells (41, 42). Basal expression of HLA-E was low in both HUVEC and HTR8 cells. Interestingly, VEGF C but not VEGF A induced HLA-E expression on HUVECs, and HTR8 trophoblast cells (Fig. 6A) which was consistent with TAP-1 induction (Fig. 4B). Next, we evaluated the surface expression of MHC class I molecules using a pan-class I antibody on HUVEC and HTR8 cells (Fig. 6B). Both cell types exhibited significant MHC class I expression in response to VEGF C but not VEGF A (Fig. 6B). As trophoblasts are not known to express classical MHC class I molecules, the observed expression with pan-antibody possibly reflects binding to induced HLA-C or HLA-G. Thus, VEGF C-induced TAP-1 expression and MHC class I assembly in trophoblasts and endothelial cells could facilitate survival by providing inhibitory signals to uNK cells.
Host defense, self-tolerance, and tissue transplantation are major health issues that invoke both innate and adaptive immune responses. However, the immune defense mechanisms of the fetus have challenged the concepts embedded in the traditionally defined pathways of immunity. This is further complicated by “immunologically tolerant” phenotype of the predominant NK cell population in the pregnant uterus and their perceived role in placentation-associated endovascular activity (16, 24). In this study we describe novel mechanisms by which uNK cells maintain non-cytotoxic and pro-angiogenesis status.
Our results provide evidence for a central link between production of VEGF C, a pro-angiogenic factor, by uNK cells and their non-cytotoxic phenotype. Using primary human peripheral blood or endometrial NK cells and panel of clones derived from primary cells, we demonstrate that VEGF C production is a hallmark property of non-cytotoxic uNK cells. We further demonstrate that VEGF C, not VEGF A or PlGF, can protect target endothelial and trophoblast cells from killing by cytotoxic pNK cells. Lack of killing is directly associated with VEGF C-mediated induction of TAP-1 in target cells, a key molecule in the process of MHC class I assembly. This is directly demonstrated by siRNA-mediated knockdown of VEGF C-mediated TAP-1 expression which results in reversal of non-cytotoxic characteristics of uNK cells.
Using a three-dimensional culture system on matrigel to evaluate the endovascular activity of uNK cells, we demonstrate that there are differences in the ability of activated uNK cell clones to facilitate or disrupt the endothelial-trophoblast dual cell interactions mimicking vascular remodeling. While CD56brightCD16− uNK cell clones consistently expressed NCRs, they differed in their cytotoxic profile. This is not surprising because uNK cell clones were propagated in vitro in the presence of activating stimuli. In vivo, most uNK cells are likely to be of non-cytotoxic phenotype. Despite in vitro conditions, several uNK cell clones maintained their non-cytotoxic phenotype. In these clones, the non-cytotoxic phenotype correlated with the production of high amounts of VEGF C. Importantly, exogenous VEGF C could rescue trophoblasts and endothelial cells from traditionally cytotoxic NK cells. Dysregulation in VEGF C production at the maternal-fetal interface could be a signal for poor angiogenesis and pregnancy complications. Reduced expression of VEGF C has been reported in pregnancies experiencing intrauterine growth restriction (IUGR) and preeclampsia (43). Involvement of VEGF C may thus explain non-killer phenotype of uNK cells despite possessing toxic granules and expressing cytotoxicity receptors.
Since our results show that uNK cells do not express receptors for VEGFs, it is tempting to argue that angiogenic factors are produced for a paracrine action on uterine endothelial cells and invading trophoblasts. Our results point to this activity of uNK cell derived VEGF C. VEGF C acts by binding to VEGF R2 and VEGF R3 and triggers intracellular survival signals in tumor cells and promote their invasion (44). In addition to maintaining uNK cell non-cytotoxicity, VEGF C may support survival of endothelial cells (45). Although during pregnancy, intrauterine hormones and factors that upregulate secretions of VEGF C in uNK cells are not known, in vitro studies suggest involvement of cytokines like IL-15 (36).
The non-cytotoxic capacity of NK cells is based on its ability to recognize surface major histocompatibility complex (MHC) class I molecules on target cells that deliver signals to suppress NK cell functions. A lack of engagement of such MHC-specific receptors leads to NK cell-mediated killing (8, 18, 23). Interestingly, Bender and colleagues (39) have shown that endothelial cells exhibited sensitivity to activated peripheral blood NK cells in the absence of expression of TAP-1. IFN-γ was shown to induce TAP-1 in endothelial cells and protect them from peripheral blood NK cell-mediated killing (39). However, our data suggest that VEGF C is robustly produced by non-cytotoxic uNK cells compared to IFN-γ. It is thus proposed that VEGF C is the predominant regulator of TAP-1 expression in the uterus, although it is likely that VEGF C and IFN-γ act in tandem as supported by our observations (Fig. 5B). Interestingly VEGF A or PLGF failed to induce TAP-1 expression (Fig. 4B). TAP-1 is a key factor essential for peptide loading for MHC class I assembly and antigen presentation (33, 46). Silencing TAP-1 expression using siRNA in trophoblasts abolished the cyto-protective activity of VEGF C, confirming the mechanism of action of VEGF C.
Trophoblasts cells do not express classical MHC class I molecules but express HLA-C and unconventional HLA-G and HLA-E molecules (5). By using HLA-E as a molecule of choice, we show that VEGF C enhanced surface expression of HLA-E on first trimester HTR8 trophoblasts. Surface expression of HLA-E normally depends on the recruitment and binding of TAP-dependent classical nonameric peptides with anchoring residues derived from HLA class I signal sequence (47, 48). Interestingly, TAP-1 expression has been shown to be significantly higher in extravillous trophoblasts that are positive for non-polymorphic HLA class I molecules (5, 49, 50). We also detected significant presence of pan-MHC class I on both HTR8 trophoblast cells and HUVECs in response to VEGF C, suggesting that this angiogenic factor is a potent immunoregulator in the uterine microenvironment.
Taken together, these findings support a dual role of VEGF C in immune tolerance and promoting active angiogenesis by uNK cells. Our findings for the first time provide evidence that non-cytotoxicity of uNK cells is directly coupled to their angiogenic translational machinery. uNK cells apparently prepare trophoblasts and endothelial cells to express modest levels of MHC class I molecules to evade killing by a major cytotoxic cell population at the maternal-fetal interface.
We thank James Padbury for his thoughtful critique of the manuscript.
Disclosures The authors have no financial conflict of interest.