Glycodelin-A stimulates the conversion of human peripheral blood CD16⁻CD56^bright NK cell to a decidual NK cell-like phenotype

Abstract

STUDY QUESTION

Does glycodelin-A (GdA) induce conversion of human peripheral blood CD16⁻CD56^bright natural killer (NK) cells to decidual NK (dNK) cells to facilitate placentation?

SUMMARY ANSWER

GdA binds to blood CD16⁻CD56^bright NK cells via its sialylated glycans and converts them to a dNK-like cells, which in turn regulate endothelial cell angiogenesis and trophoblast invasion via vascular endothelial growth factor (VEGF) and insulin-like growth factor-binding protein 1 (IGFBP-1) secretion, respectively.

WHAT IS KNOWN ALREADY

dNK cells are the most abundant leucocyte population in the decidua. These cells express CD16⁻CD56^bright phenotype. Peripheral blood CD16⁻CD56^bright NK cells and hematopoietic precursors have been suggested to be capable of differentiating towards dNK cells upon exposure to the decidual microenvironment. These cells regulate trophoblast invasion during spiral arteries remodelling and mediate homoeostasis and functions of the endothelial cells. GdA is an abundant glycoprotein in the human decidua with peak expression between the 6th and 12th week of gestation, suggesting a role in early pregnancy. Indeed, GdA interacts with and modulates functions and differentiation of trophoblast and immune cells in the human feto-maternal interface. Aberrant GdA expression during pregnancy is associated with unexplained infertility, pregnancy loss and pre-eclampsia.

STUDY DESIGN, SIZE, DURATION

CD16⁺CD56^dim, CD16⁻CD56^bright and dNK cells were isolated from human peripheral blood and decidua tissue, respectively, by immuno-magnetic beads or fluorescence-activated cell sorting. Human extravillous trophoblasts were isolated from first trimester placental tissue after termination of pregnancy. Biological activities of the cells were studied after treatment with GdA at a physiological dose of 5 μg/mL. GdA was purified from human amniotic fluid by immuno-affinity chromatography.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Expression of VEGF, CD9, CD49a, CD151 and CD158a in the cells were determined by flow cytometry. Angiogenic proteins in the spent media of NK cells were determined by cytokine array and ELISA. Blocking antibodies were used to study the functions of the identified angiogenic proteins. Endothelial cell angiogenesis was determined by tube formation and trans-well migration assays. Cell invasion and migration were determined by trans-well invasion/migration assay. Binding of normal and de-sialylated GdA, and expression of L-selectin and siglec-7 on the NK cells were analysed by flow cytometry. The association between GdA and L-selectin on NK cells was confirmed by immunoprecipitation. Extracellular signal-regulated protein kinases (ERK) activation was determined by Western blotting and functional assays.

MAIN RESULTS AND THE ROLE OF CHANCE

GdA treatment enhanced the expression of dNK cell markers CD9 and CD49a and the production of the functional dNK secretory product VEGF in the peripheral blood CD16⁻CD56^bright NK cells. The spent media of GdA-treated CD16⁻CD56^bright NK cells promoted tube formation of human umbilical vein endothelial cells and invasiveness of trophoblasts. These stimulatory effects were mediated by the stimulatory activities of GdA on an ERK-activation dependent production of VEGF and IGFBP-1 by the NK cells. GdA had a stronger binding affinity to the CD16⁻CD56^bright NK cells as compared to the CD16⁺CD56^dim NK cells. This GdA-NK cell interaction was reduced by de-sialylation. GdA interacted with L-selectin, expressed only in the CD16⁻CD56^bright NK cells, but not in the CD16⁺CD56^dim NK cells. Anti-L-selectin functional blocking antibody suppressed the binding and biological activities of GdA on the NK cells.

LARGE SCALE DATA

N/A.

LIMITATIONS, REASONS FOR CAUTION

Some of the above findings are based on a small sample size of peripheral blood CD16⁻CD56^bright NK cells. These results need to be confirmed with human primary dNK cells.

WIDER IMPLICATIONS OF THE FINDINGS

This is the first study on the biological role of GdA on conversion of CD16⁻CD56^bright NK cells to dNK-like cells. Further investigation on the glycosylation and functions of GdA will enhance our understanding on human placentation and placenta-associated complications with altered NK cell biology.

STUDY FUNDING/COMPETING INTEREST(S)

This work was supported by the Hong Kong Research Grant Council Grant 17122415, Sanming Project of Medicine in Shenzhen, the Finnish Cancer Foundation, Sigrid Jusélius Foundation and the Finnish Society of Clinical Chemistry. The authors have no competing interests to declare.

glycodelin, natural killer cell, trophoblast, endothelial cell, L-selectin

Introduction

Natural killer (NK) cells are the most abundant (~70%) leucocyte population at the maternal–foetal interface (Erlebacher, 2013) in early pregnancy. Unlike the majority of NK cells in the peripheral blood (CD16⁺CD56^dim), most of the resident NK cells in the decidua have a phenotype of CD9⁺CD16⁻CD49a⁺CD56^brightCD151^brightKIR⁺, lower cytotoxicity and a higher cytokine secretion (Moffett-King, 2002; Koopman et al., 2003; Dosiou and Giudice, 2005). It has been suggested that peripheral blood CD16⁻CD56^bright NK cells and hematopoietic precursors are recruited into the decidua during pregnancy and differentiate into the decidual NK (dNK) cells upon exposure to the local microenvironment (van den Heuvel et al., 2005; Carlino et al., 2008; Poli et al., 2009; Erlebacher, 2013; Montaldo et al., 2015; Bjorkstrom et al., 2016).

dNK cells lie in close proximity with the uterine spiral arteries and the invading extravillous trophoblasts (Moffett-King, 2002; Erlebacher, 2013). They gradually disappear after mid-trimester and are nearly absent at term. Studies utilizing animal models deficient of dNK cells show that the primary role of dNK cells is to promote placental vascular remodelling via secretion of vascular and trophoblast regulatory factors (Moffett-King, 2002; Andraweera et al., 2012; Erlebacher, 2013). Aberrant levels and activity of dNK cells are associated with pathogenesis of pre-eclampsia, foetal loss, intrauterine growth restriction, preterm birth and congenital infection (Erlebacher, 2013).

Glycodelin is a glycoprotein belonging to the lipocalin family with four isoforms (Seppala et al., 2002). One of the isoforms, glycodelin-A (GdA), is expressed in high amounts in secretory and decidualised endometrium and in amniotic fluid. Its concentration in the human decidua rises in early pregnancy, and peaks between the 6th and 12th week of gestation (Seppala et al., 2002). Aberrant levels of GdA in serum, endometrial tissue and/or uterine flushing is associated with recurrent miscarriage, early pregnancy loss and unexplained infertility (Seppala et al., 2002, 2007). The most well studied role of GdA is its regulatory role of immune cell function at the feto-maternal interface for maintenance of pregnancy (Lee et al., 2011c, 2016). GdA suppresses proliferation and induces apoptosis of T-cells (Lee et al., 2009), induces a tolerogenic phenotype in dendritic cells (Scholz et al., 2008), modulates cytokine production in peripheral blood NK cells (Lee et al., 2010) and macrophages (Lee et al., 2012), and induces a Th-2–type shift in cytokine profile (Lee et al., 2011a). Glycosylation, in particular the terminal sialic acids of the GdA-derived glycans, mediate the binding of GdA to its glycan receptors, such as L-selectin (Lee et al., 2012), and, thus, is important for the biological activities of GdA (Mukhopadhyay et al., 2004; Poornima and Karande, 2007; Lee et al., 2009, 2011b).

Despite the importance of dNK cells in pregnancy, little is known about the role of decidua-derived factors, in regulating peripheral blood CD16⁻CD56^bright NK cells conversion to dNK cells. Previously studies suggested that TGFβ (Keskin et al., 2007; Allan et al., 2010), in combination with demethylating agents under hypoxic conditions (Cerdeira et al., 2013), results in the formation of dNK-like cells. In this study, we hypothesized that GdA facilitates the generation of dNK cells and thereby modulates the placentation process.

Materials and Methods

Purification of glycodelin-A from human amniotic fluid

The Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster approved this study. Amniotic fluid was collected from women who underwent amniocentesis for prenatal diagnosis. GdA was purified from the amniotic fluid by affinity chromatography using anti-glycodelin monoclonal antibody clone F43-7F9 (Riittinen et al., 1989) as described (Lee et al., 2009, 2010, 2011a, 2011b, 2012, 2014). Bacteria and insoluble debris were filtered from the extract by passing it through a 0.22 μm filter (Millipore, USA). Bacterial endotoxin was removed by Pierce^® high capacity endotoxin removal spin column (Thermo Fisher Scientific, USA). The identity of the purified GdA was confirmed by Western blotting and mass spectrometry (Supplementary Fig. S1). De-sialylated GdA was prepared by incubation of GdA with sialidase-coated agarose beads as described (Lee et al., 2012). The success of de-sialylation was verified by lectin immunoassay (Lee et al., 2009, 2011b, 2012).

Isolation of human natural killer cells

Human female peripheral blood was obtained from the blood donation samples of the Hong Kong Red Cross. Peripheral blood mononuclear cells (PBMCs) were obtained by Ficoll-Paque (GE Healthcare Life Sciences) density gradient centrifugation. Blood NK cells were enriched from PBMCs by negative immuno-magnetic NK cell isolation kit (Miltenyi Biotech, Germany). CD16⁺CD56^dim and CD16⁻CD56^bright NK cells were further purified by two methods. For the angiogenesis antibody array, CD16⁺CD56^dim and CD16⁻CD56^bright NK cells were isolated by fluorescence activated cell sorting using BD FACSAria SORP cell sorter (BD Biosciences, USA) (Supplementary Fig. S2). Blood NK cells after negative immuno-magnetic separation were stained with allophycocyanin (APC)-conjugated anti-human CD3, Alexa Fluor^® (AF)−488 conjugated anti-human CD56, phycoerythrin (PE) conjugated anti-human CD16 and 4′,6-diamidino-2-phenylindole (DAPI). The purity of the blood NK, CD16⁺CD56^dim and CD16⁻CD56^bright NK cells was >95% (Supplementary Fig. S3). For other studies, CD16⁻CD56^bright NK cells were enriched by CD56 MicroBeads (Miltenyi Biotech) after negative immuno-magnetic NK cell isolation steps. The purity of the isolated CD16⁻CD56^bright NK cell was > 95% as determined by flow cytometry (Supplementary Fig. S3). Antibodies used for flow cytometry analysis in this study are listed in Supplementary Table S1.

dNK cells were isolated as described previously (Male et al., 2012). In brief, first trimester decidual tissues were collected, with written consent, from patients who had undergone surgical termination of pregnancy due to psychosocial reasons. The tissues were minced and digested with DNase I (50 μg/mL) and collagenase (300 U/mL) and the cells were passed through 100 and 40 μm filters followed by Ficoll-Paque density gradient centrifugation. The cells were then cultured on a plastic plate for 2 h. The non-adherent cells were collected and dNK cells were further enriched using CD56 MicroBeads as described above. The purity of the isolated dNK cells using this method was > 90% (Supplementary Fig. S3), however, the preparation may include some NKT cell population.

Conversion of blood CD16⁻CD56^bright NK cells to decidual NK-like cells

NK cells (1 × 10⁶) were cultured in the Roswell Park Memorial Institute (RPMI-1640) medium supplemented with 10% foetal bovine serum (FBS), 500 IU/mL of recombinant human interleukin (IL)−2 (Peprotech), 5 ng/mL of recombinant human IL-15 (Peprotech) and 5 μg/mL of GdA for 5 days. NK cells cultured under the same conditions without GdA were used as the control. Similar culture condition with the inclusion of both IL-2 and IL-15 has also been used in other studies for NK cells from endometrium (Eriksson et al., 2004; Mselle et al., 2009). We included both cytokines to study the transition of the NK precursor cells from the blood stream to the decidua. The GdA concentration used was within the physiological range of decidual GdA during first and second trimester of pregnancy (Seppala et al., 2002). For the preparation of NK cell spent medium, the NK cells (control, GdA-treated or dNK cells) were washed twice by PBS followed by culture in serum-free RPMI-1640 medium without cytokine or GdA supplement. The spent medium was then collected after 24 h for the profiling of angiogenesis-related soluble proteins and bioassays. GdA content in the spent medium was <0.01 μg/mL as determined by the ELISA (Lifespan bioscience, Inc, Seattle, WA).

Analysis of NK cell phenotype

NK cells (1×10⁶) were stained for dNK cell surface markers with AF488-conjugated anti-CD56, PE-conjugated anti-CD16, and APC-conjugated anti-CD9, CD49a, CD151 and CD158a antibodies (Supplementary Table S1) for 1 h at 4°C. The cells were washed in PBS twice and resuspended in blocking buffer for flow cytometry analysis. To determine the intracellular levels of vascular endothelial growth factor (VEGF), the cells were fixed in fixation buffer (eBioscience) and stained with anti-CD56-AF488 and anti-CD16-PE antibodies. The stained cells were permeabilized with permeabilization buffer (eBioscience), washed and resuspended in 80 μL of permeabilisation buffer containing 20 μL of APC-conjugated anti-VEGF antibody (Clone 23410, R&D Systems) for 20 min, before resuspension for flow cytometric analysis. The data were analysed by the software FlowJo 10.0.7 (Tree Star Inc., Ashland, USA). The phenotype expression in CD16⁻CD56^bright and CD16⁺CD56^dim NK cells were analysed after differential gating.

Profiling of angiogenesis-related soluble proteins

The profile of angiogenesis-related soluble proteins in the GdA-treated NK cell serum-free spent medium was determined semi-quantitatively by antibody array (R&D Systems) according to the manufacturer’s protocol. The density of the protein spots was analysed by densitometry using Quantity One software (Bio-Rad). The levels of VEGF and insulin-like growth factor-binding protein 1 (IGFBP-1) in the spent serum-free culture medium were determined by ELISA (VEGF ELISA, DY293B-05; IGFBP-1 ELISA, DY871-05, R&D Systems) according to the manufacturer’s protocol.

Extravillous trophoblast and endothelial cell culture

First trimester placental villi were obtained with written consent from women undergoing surgical termination of pregnancy due to psychosocial reasons. Primary trophoblast was isolated as described previously (Male et al., 2012; Lee et al., 2013a; 2013b). Primary extravillous trophoblast (EVT) was obtained by culturing the isolated trophoblast on 2% fibronectin (1 mg/mL, Sigma) coated plate for 12 h (Male et al., 2012). The isolated cells were stained with AF488-conjugated anti-cytokeratin-7 (Clone EPR17078, Abcam) and HLA-G (Clone MEM-G/9, Abcam) and were analysed by flow cytometry (Supplementary Fig. S4). The JEG-3 choriocarcinoma cell line (American Type Culture Collection) was also used. These cells express a repertoire of human leucocyte antigen molecules similar to EVT (Apps et al., 2009) and are frequently used to study EVT functions (Wolf et al., 2010). Primary EVTs and JEG-3 cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) medium containing 10% FBS.

Human umbilical vein endothelial cells (HUVEC) were purchased from Lonza (Basel, Switzerland) and were cultured in DMEM/F12 medium supplemented with 1% endothelial cell growth supplement (2 mg/mL, Sigma, USA), 0.1% heparin (0.1 g/mL, Sigma) and 10% FBS, on 1% gelatin (Sigma) coated plate.

Endothelial cell functions

The effect of spent serum-free medium from CD16⁻CD56^bright NK cells on angiogenesis was determined by the tube formation assay. In brief, a 24-well plate or chamber slide was coated with 300 μL of phenol red-free Matrigel^® Basement Membrane Matrix (Corning). HUVEC (5 × 10⁴ cell/well for 24 well plate and 0.5×10⁴ cell/well for chamber slide) in serum-free culture medium were seeded into the coated wells. The spent serum-free medium from CD16⁻CD56^bright NK cell culture was then added at a ratio of 1:1 (v/v). After 15 h of incubation, images of the endothelial cell network were captured under an inverted microscope and analysed using the ImageJ Pro Plus software (Media cybernetics, Inc. Washington) with an angiogenesis analyser plugin.

Endothelial cell migration was determined as described (Lee et al., 2013a) using the CytoSelect™ cell migration assay. In brief, HUVEC (5×10⁴ cell/well) in serum-free culture medium were mixed with the spent serum-free medium from CD16⁻CD56^bright NK cells at a ratio of 1:1 (v/v). The cells were transferred to the cell culture insert and allowed to migrate through a basement membrane for 12 h. The migrated cells at the bottom of the membrane were stained with the CyQuant GR Dye and quantified by a fluorescence plate reader with an excitation of 485 nm and an emission of 530 nm. The role of NK cell-derived VEGF or IGFBP-1 on endothelial cell function was also studied by inclusion of 10 μg/mL anti-VEGF (V6627, Sigma) and anti-IGFBP-1 (Clone 33627, R&D Systems) blocking antibodies in the culture.

Trophoblast functions

Trophoblast invasion was determined by the trans-well invasion assay (BD BioCoat^TM Matrigel^TM Invasion Assay, BD Biosciences). Trophoblast (1×10⁵ cells/well) suspended in serum-free medium were seeded into a cell culture insert along with the spent serum-free medium from CD16⁻CD56^bright NK cells (1:1). Medium with 20% FBS was added to the tissue culture well, and trophoblast was allowed to invade through the Matrigel basement membrane matrix for 24 h. Cells that had invaded through the matrix were stained with crystal violet (Millipore) and lysed with 10% acetic acid for quantification at an absorbance of 595 nm, using a microplate reader.

Flow cytometric analysis of binding of GdA and de-sialylated GdA on NK cells

GdA and de-sialylated GdA were labelled with AF647 fluorescence dye as described (Thermo Fisher Scientific) (Lee et al., 2011a, 2011b, 2012). NK cells (0.5×10⁶) were incubated with 1 μg/mL labelled GdA for 1.5 h. The cells were then washed twice with PBS and analysed by flow cytometry. The data were analysed by the FlowJo software.

Detection of L-selectin on CD16⁻CD56^bright NK cells

The expression of two sialic acid-dependent glycan receptors, L-selectin and siglec-7, on NK cells was determined by flow cytometry. NK cells (1×10⁶) were stained with AF488-conjugated anti-CD56, PE-conjugated anti-CD16 and allophycocyanin (APC)-conjugated anti-L-selectin (Clone DREG-56, 1 μg/mL, eBioscience)/AF647-conjugated anti-siglec-7 (Clone 194212, 1 μg/mL, R&D Systems) antibodies. Isotypic antibodies were used as the control. The data were analysed by the FlowJo software.

Interaction between GdA and L-selectin on NK cells

The role of L-selectin/siglec-7 on GdA-NK cells interaction was confirmed by treating the NK cells with 1 μg/mL of AF647-conjugated normal or de-sialylated GdA in the presence of anti-L-selectin (ab18959, Abcam) or anti-siglec-7 (Clone 194212, R&D Systems) blocking antibodies in a molar ratio of 1:5. The cell-bound fluorescence was measured by flow cytometer after incubation for 2 h.

Co-immunoprecipitation experiments were performed to study the interaction between GdA and L-selectin. In brief, the membrane proteins of NK cells (1×10⁷) were isolated by the Proteo-Extract Transmembrane Protein Extract Kit (Merck, Darmstadt, Germany) as described (Huang et al., 2013, 2015) and incubated with 5 μg/mL of GdA with gentle shaking overnight at 4°C. Monoclonal anti-glycodelin antibody (Clone F43-7F9) immobilized on sepharose beads was then used to precipitate the GdA-associated protein complexes, which were analysed by Western blot using the anti-L-selectin (Clone DREG-56) and anti-L-siglec-7 (Clone 194212) antibodies. Membrane protein extracts without GdA treatment were used as the control. The GdA-mediated biological activities on NK cells were studied as described above after anti-L-selectin blocking antibody treatment.

Effect of GdA on extracellular signal-regulated protein kinases (ERKs) levels in NK cells

NK cells (1×10⁶ cells) were cultured with GdA (5 μg/mL) in the presence or absence of anti-L-selectin blocking antibody. The cells were lysed by the Cytobuster protein extraction reagent (Merck, Darmstadt, Germany) and resolved in 10% polyacrylamide gel for Western blot analyses using polyclonal antibodies against ERK (#9101 S, Cell Signalling Technology), phosphorylated ERK (#9102 S, Cell Signalling Technology) or β-actin (Sigma-Aldrich) for normalization. The control cells were treated with ERK inhibitors PD98059 (10 μM) and U0126 (1 μM).

Statistical analysis

All values were expressed as mean ± SD. Sample size (N) refers to the number of NK cell samples. Non-parametric ANOVA on Rank test was used to test the statistical differences between groups. Parametric Student’s t-test or non-parametric Mann Whitney U test were used where appropriate as the post-test. The data were analysed by SigmaStat 2.03 (Jandel Scientific). A P-value of <0.05 was considered as significant.

Results

GdA induces the expression of CD9, CD49a and VEGF in CD16⁻CD56^bright NK cells

GdA treatment at the concentration used did not affect the viability of NK cells as demonstrated by XTT viability assay (data not shown), consistent with our previous findings (Lee et al., 2010). Flow cytometry was employed to determine the GdA action on the expression of dNK cell markers CD9, CD49a, CD151 and CD158a, and the functional dNK secretory product VEGF (Koopman et al., 2003; Cerdeira et al., 2013). GdA treatment significantly (P < 0.05) enhanced the expression of CD9, CD49a and VEGF in the CD16⁻CD56^bright NK cells as compared to the control (Fig. 1). The VEGF up-regulatory effect was specific for CD16⁻CD56^bright NK cells, as the effect could not be observed with CD16⁺CD56^dim NK cells.

Figure 1

GdA influences the expression of dNK cell phenotypic markers. CD16⁺CD56^dim or CD16⁻CD56^bright NK cells were incubated with or without GdA (5 μg/mL). The expression of dNK cell intercellular marker VEGF and surface marker CD9, CD49a, CD151 and CD158a, and the functional dNK secretory product VEGF was measured by flow cytometry. Isotype IgG was used as a negative control. Data are mean ± SD, N = 5. *P < 0.05 vs. control NK cell.

GdA-treated CD16⁻CD56^bright NK cells produce angiogenesis-related proteins

The GdA treatment upregulated the expression of VEGF and IGFBP-1 (P < 0.05) in the CD16⁻CD56^bright NK cells as detected by membrane-based angiogenesis antibody array (Fig. 2A) and confirmed by ELISAs (Fig. 2B). dNK cells also had a significantly (P < 0.05) higher level of VEGF and IGFBP-1 secretion when compared to the control.

Figure 2

The angiogenesis-related soluble proteins profile of GdA-treated CD16⁻CD56^bright NK cells and dNK cells. CD16⁻CD56^bright NK cells were cultured with or without GdA (5 μg/mL). (A) Secretion of angiogenesis-related soluble proteins was measured by human angiogenesis antibody array (N = 3). The intensities of the spots were analysed by densitometry. Results are expressed as fold-changes relative to control = (Density of GdA group – Density of negative control)/(Density of control group – density of negative control). (B) Concentrations of VEGF and IGFBP-1 in NK cells were measured by ELISAs (N = 5). The samples were run in duplicate. Data are expressed as mean ± SD, *P < 0.05 vs. control. + P < 0.05 vs. GdA group.

GdA enhances angiogenetic activity of CD16⁻CD56^bright NK cells via VEGF

The spent medium of GdA-treated CD16⁻CD56^bright NK cell promoted tube formation (Fig. 3A) and cell migration (Figure 3B) of the HUVEC cells significantly (P < 0.05) when compared to the control. The inclusion of anti-VEGF antibody, but not of anti-IGFBP-1 antibody, suppressed the stimulatory activity of GdA on tube formation. dNK cells spent medium also promoted the tube formation and cell migration of the HUVEC cells when compared to the control. In addition, culture medium of the CD16⁻CD56^bright NK cells had no effect on proliferation of HUVEC cells as determined by Cyquant NF proliferation assay (Invitrogen, Carlsbad) (data not shown).

Figure 3

VEGF secreted by GdA-treated CD16⁻56^bright NK cells regulates the angiogenic capacity of HUVEC. (A) HUVECs were treated with the immunomagnetically separated CD16⁻CD56^bright NK and dNK cell conditioned medium with isotype, anti-IGFBP-1 or anti-VEGF antibodies for the tube formation assay (N = 5). Number of meshes formed were analysed by ImageJ (Angiogenesis analyser). Tube formation (%) = (No. of meshes of treatment group)/(No. of mesh of control group) × 100%. (B) HUVEC were treated with immunomagnetically separated CD16⁻CD56^bright NK and dNK cell conditioned medium with isotype, anti-IGFBP-1 or anti-VEGF antibodies (10 μg/mL) for the migration assay (N = 5). Cell migration (%) = (Absorbance of treatment group)/(Absorbance of control group) × 100%. Data are mean ± SD. *P < 0.05 vs. CD16⁻56^bright NK cell medium + isotype antibody group. + P < 0.05 vs. GdA-treated CD16⁻56^bright NK cell medium + isotype antibody group.

GdA enhances the invasion-promoting activity of CD16⁻CD56^bright NK cells via IGFBP-1

GdA treatment significantly (P < 0.05) enhanced the invasion-promoting activity of CD16⁻CD56^bright NK cell spent medium on primary EVTs and JEG-3 cells (Fig. 4). Anti-IGFBP-1 antibody, but not anti-VEGF antibody, supplementation nullified the stimulatory effect of GdA (Fig. 4). dNK cells spent medium also promoted the invasion of the JEG-3 cells when compared to the control. Spent medium obtained from GdA-treated CD16⁻CD56^bright NK cell failed to exhibit any significant effect on migration (Fig. 4) and proliferation (data not shown) of JEG-3 cells.

Figure 4

IGFBP-1 secreted by GdA-treated CD16⁻56^bright NK cells regulates trophoblast invasion. Primary extravillous trophoblasts and JEG-3 cells were treated with the immunomagnetically separated CD16⁻CD56^bright NK and dNK cell conditioned medium with isotype, anti-IGFBP-1 or anti-VEGF antibodies for invasion and migration assays (N = 5). Cell invasion and cell migration are expressed as a ratio (%) between the absorbance of treatment group and absorbance of control group. Data are mean ± SD. *P < 0.05 vs. Control CD16⁻56^bright NK cell medium + isotype antibody group. + P < 0.05 vs. GdA-treated CD16⁻56^bright NK cell medium + isotype antibody group.

Interaction between GdA and L-selectin on NK cells

CD16⁻CD56^bright NK cells had a higher binding capacity for fluorescently labelled GdA as compared to the CD16⁺CD56^dim NK cells (Fig. 5A). Interestingly, after de-sialylation GdA failed to bind to both subsets of NK cells (Fig. 5B). Therefore, we studied the expression of two sialic acid-dependent glycodelin receptors, L-selectin and siglec-7, on NK cells (Fig. 5C). While siglec-7 expression was found on both NK subsets, L-selectin was specifically expressed on the CD16⁻CD56^bright NK cells only. The inclusion of antibodies against L-selectin, but not siglec-7, significantly (P < 0.05) suppressed the binding of GdA to the CD16⁻CD56^bright NK cells (Fig. 5D). The results were confirmed by immunoprecipitation, demonstrating the interaction between GdA and L-selectin in the membrane protein fraction of the CD16⁻CD56^bright NK cells (Fig. 5E).

Figure 5

GdA interacts with L-selectin found on CD16⁻CD56^bright NK cells. (A) Immunomagnetically separated CD16⁻CD56^bright and CD16⁺CD56^dim NK cells were incubated with fluorescence-conjugated GdA (N = 4) and (B) fluorescence-conjugated normal/de-sialylated GdA (N = 4). (C) Expression of L-selectin and siglec-7 in CD16⁻CD56^bright and CD16⁺CD56^dim NK cells were analysed by flow cytometry. (D) NK cells were incubated with fluorescence-conjugated GdA in the presence of anti-L-selectin, anti-siglec-7 or isotype antibodies in a molar ratio of 1:5 (N = 4). (E) Immunoprecipitation experiment showing the interaction between GdA and L-selectin in the membrane protein fraction of NK cells (lanes 1 and2) (N = 3). NK cell membrane lysate (lane 3) and GdA + anti-Gd column were used as positive and negative controls (lane 4), respectively. Data are expressed as mean ± SD. *P < 0.05 vs. control. + P < 0.05 vs GdA or GdA + isotype antibody group.

L-selectin mediates the GdA actions on CD16⁻CD56^bright NK cells in an ERK dependent manner

Anti-L-selectin antibody treatment inhibited the GdA-induced production of VEGF and IGFBP-1 (Fig. 6A) and GdA-induced angiogenesis and invasion-promoting activities (Fig. 6B) of the CD16⁻CD56^bright NK cells.

Figure 6

L-selectin mediates binding and activities of GdA on CD16⁻CD56^bright NK cells. (A) Immunomagnetically separated CD16−CD56^bright NK cells were cultured with or without GdA (5 μg/mL) in the presence or absence of L-selectin antibody. VEGF and IGFBP-1 were measured by ELISAs (N = 5). The samples were run in duplicate. (B) Control and GdA-treated CD16⁻56^bright NK cells were cultured with serum-free medium in the presence or absence of L-selectin antibody for 24 h. HUVEC and JEG-3 were treated with the CD16^-56^bright NK cell-conditioned medium for the tube formation assay (N = 5) and invasion assay (N = 5), respectively. Tube formation (%) = (No. of meshes of treatment group)/(No. of mesh of control group) × 100%. Cell invasion (%) = (Absorbance of treatment group)/(Absorbance of control group) × 100%. Data are expressed as mean ± SD. *P < 0.05 vs. control + isotype antibody group. ⁺P < 0.05 vs. GdA + isotype antibody group.

Treatment with GdA for 30 minutes significantly (P < 0.05) increased the levels of phosphorylated ERKs (ERK1, 44 kDa; and ERK2, 42 kDa) in the CD16⁻CD56^bright NK cells (Fig. 7A). The stimulatory effect of GdA was reverted by the addition of anti-L-selectin antibody (Fig. 7B). The stimulatory effect of GdA on VEGF and IGFBP-1 secretions are ERK dependent, as ERK kinase inhibitors (PD98059 and U0126) abolished such effects in the CD16⁻CD56^bright NK cells (Fig. 7C). The inhibitors alone did not affect VEGF and IGFBP-1 production (Fig. 7C) or viability (data not shown) of the CD16⁻CD56^bright NK cells.

Figure 7

GdA triggers phosphorylation of ERK in CD16⁻CD56^bright NK cells by binding to L-selectin. (A) Expression of total and phosphorylated ERK on immunomagnetically separated CD16⁻CD56^bright NK cells with GdA (5 μg/mL) treatment for different times were measured by Western blot analysis (N = 4). (B) Expression of total and phosphorylated ERK (30 min) by GdA-treated CD16⁻CD56^bright NK cells in the presence or absence of anti-L-selectin antibody (N = 4). CD16⁻CD56^bright NK cells without GdA treatment were used as control (control NK cells). (C) Effect of ERK inhibitors (PD98059 or U0126) on VEGF and IGFBP secretions from GdA-treated CD16⁻CD56^bright NK cells (N = 5). CD16⁻CD56^bright NK cells without GdA treatment were used as control (control NK cells). Data are expressed as mean ± SD. *P < 0.05 vs. control or control + isotype antibody group. ⁺P < 0.05 vs. GdA + isotype antibody group.

Discussion

In the first trimester uterine decidua, ~70% of the leucocytes are NK cells, which are phenotypically different from those of peripheral blood (Erlebacher, 2013). The mechanisms controlling the accumulation of CD16⁻CD56^bright NK cells in the decidua are still not known. Several possibilities for the origin of the dNK cells have been proposed, one of which is that CD16⁻CD56^bright NK cells are recruited from the peripheral blood to the decidua, where they undergo tissue specific differentiation (van den Heuvel et al., 2005; Poli et al., 2009; Montaldo et al., 2015; Bjorkstrom et al., 2016). In the pre-implantation period and the first trimester of pregnancy, NK cells reside in the GdA rich decidua (Seppala et al., 2007; Lee et al., 2011c). Therefore, we investigated whether GdA could induce dNK cell conversion and we determined the mechanism involved.

Our data show that GdA polarizes peripheral blood NK cells towards the dNK cell phenotype; GdA treatment increased the expression of CD9, CD49a and VEGF of CD16⁻CD56^bright NK cells. CD9 and CD49a distinguish dNK cells from peripheral blood NK cells, while VEGF is an important functional secretory product of dNK cells (Koopman et al., 2003; Cerdeira et al., 2013). CD9 is a member of the tetraspanin family and is characterized by its extracellular and transmembrane-spanning domains. It forms complexes with integrins and other tetraspanins. CD49a is an integrin-α1 subunit. It is an adhesion molecule and it regulates the migration, proliferation and cytokine production of leucocytes. Apart from GdA, TGF-β1 transforms CD16⁺CD9⁻ peripheral blood NK cells into CD16⁻CD9⁺ NK cells, in vitro, resembling dNK cells (Keskin et al., 2007; Allan et al., 2010). Additionally, TGF-β1 along with demethylating agent 5-aza-2′-deoxycytidine treatment under hypoxic conditions also results in the formation of dNK-like cells expressing CD9, CD49a and VEGF (Cerdeira et al., 2013). Importantly, only CD16⁻CD56^bright NK cells, but not CD16⁺CD56^dim NK cells, responded to GdA treatment leading to up-regulation of VEGF expression. It has been proposed that dNK cell-derived VEGF maintains functional integrity of the decidual arteries (Lash et al., 2006). GdA also induces VEGF production in other cells, including endometrial epithelial cells (Park et al., 2006) and cumulus cells (Hayes et al., 2006) in vitro. In congruence with the increased VEGF secretion, spent GdA-treated CD16⁻CD56^bright NK cell culture medium improved the tube formation capacity of HUVEC cells. Neutralizing VEGF in the medium abolished this effect, indicating that the GdA-induced VEGF secretion was essential for the CD16⁻CD56^bright NK cell-mediated angiogenesis. Another study, using a synthetic glycodelin-derived peptide and glycodelin-rich amniotic fluid, also suggested that VEGF mediates the action of glycodelin on angiogenesis during embryogenesis and endometrial cancer (Song et al., 2001). The critical role of VEGF in vascular remodelling during pregnancy is well documented (Koopman et al., 2003; Hanna et al., 2006; Fraser et al., 2012; Kim et al., 2013).

dNK cells are located near the spiral arteries and close to the invaded EVT (Moffett-King, 2002; Erlebacher, 2013). We found that GdA-induced production of a pro-invasive factor IGFBP-1 from the CD16⁻CD56^bright NK cells. Neutralization of IGFBP-1 by blocking antibodies reduced the invasive capacity, hinting at a modulatory role in EVT invasion. The pro-invasive nature of the GdA-treated CD16⁻CD56^bright NK cells contradict our previous observations showing a direct inhibitory effect of GdA on EVT invasion (Lam et al., 2009, 2011). Unlike cancer cells, trophoblast invasion needs to be tightly controlled. Thus, the seemingly opposite actions of GdA may represent a negative feedback mechanism controlling the extent of EVT invasion. Similarly, GdA or its N-glycans stimulate hCG production in villous cytotrophoblast cells (Jeschke et al., 2003) and hCG also induces EVT invasion (Lee et al., 2013a).

GdA interacts via its carbohydrate chains with surface receptors of various cell types in the reproductive tract, modulating fertilization (Yeung et al., 2009) and pregnancy (Lee et al., 2016). The sialylation of GdA is crucial for its biological activities (Jayachandran et al., 2004, 2006; Poornima and Karande, 2007; Lee et al., 2009, 2011b, 2012; Lam et al., 2011). Reduced sialylation may have clinical consequences in women with gestational diabetes (Lee et al., 2009, 2011b). In this study, de-sialylated GdA failed to bind to both NK cell subsets, suggesting that the binding of GdA to NK cells is sialic acid-dependent.

L-selectin is one of the major sialic-acid-recognizing receptors in human NK cells. L-selectin is crucial for tethering and rolling of CD16⁻CD56^bright NK cells on endometrial endothelial cells and for recruitment of NK cells to the endometrial tissue (Lotze and Thomson, 2010). Immunoprecipitation and blocking antibody experiments confirmed the interaction of GdA with L-selectin on the cell membrane of NK cells. L-selectin is highly expressed in CD16⁻CD56^bright NK cells, but not in CD16⁺CD56^dim NK cells, and thus GdA has a stronger binding to former, than to the latter, NK cells. The GdA-L-selectin interaction mediates the actions of GdA on the NK cells production of VEGF/IGFBP-1, which in turn affects tube formation and trophoblast invasion.

Since L-selectin antibody could not completely block the GdA binding, and GdA also bound weakly to the CD16⁺CD56^dim NK cells in a sialic acid-dependent manner, we sought to study siglec-7 as another GdA receptor in NK cells. Although siglec-7 was expressed in both studied NK cell populations, siglec-7 functional blocking antibody had no effect on GdA binding, suggesting that siglec-7 did not bind to GdA. There are other known glycan-dependent receptors of GdA including fucosyltransferase-5 (Sperm) (Chiu et al., 2007), E-selectin (Sperm) (Chiu et al., 2004), siglec-6 (Trophoblast) (Lam et al., 2011), CD22 (B-cell) (Dell et al., 1995), CD7 (T-cell) (SundarRaj et al., 2009) and asialoglycoprotein receptor (Park et al., 2005). It remains to be determined whether these molecules participate in the binding of GdA to the NK cells.

dNK cell differentiation involves a cascade of events and they acquire different markers at different stages of the differentiation process (Chiossone et al., 2014). L-selectin-mediated interactions contribute to the trafficking of peripheral blood CD16⁻CD56^bright NK cells to the decidua (van den Heuvel et al., 2005). However, most tissue-resident CD16⁻CD56^bright NK cells, including dNK (Koopman et al., 2003) either have lost or partially down-regulated their L-selectin expression at the final stages of NK cell development (Melsen et al., 2016). Therefore, it is highly suggestive that the GdA-L-selectin interaction may only promote the dNK cell differentiation at the early stage and the successful differentiation into dNK cells requires other pregnancy-associated factors. Consistently, apart from GdA, a panel of pregnancy-associated factors, such as TGFβ (Keskin et al., 2007; Allan et al., 2010) and the hypoxic environment (Cerdeira et al., 2013), are also involved in the differentiation of dNK. The lack of these factors during differentiation in our study may explain the difference in the expression of some reported markers for dNK cells, including CD151 and CD158a, between GdA-treated CD16⁻CD56^bright NK and dNK cells. All these results indicated that GdA treatment alone is not enough to induce the formation of a true dNK cell. Further investigation is required to elucidate the combined role of these factors on dNK cell differentiation.

In the present study, ERK signalling mediated the GdA-induced VEGF and IGFBP-1 production from the NK cells; inhibitors of the ERK pathway PD98059 and U126 abolished the GdA-induced VEGF and IGFBP-1 production. This is in line with two previous observations. First, ERK signalling pathways are associated with regulation of IL-6 production in monocytes and macrophages via L-selectin (Lee et al., 2012). Second, GdA also activates ERK in spermatozoa (Chiu et al., 2010), trophoblast (Lam et al., 2012) and monocytes/macrophages (Lee et al., 2012). Nevertheless, other signalling pathway(s) may also be involved in the regulation of GdA-induced functions in NK cells.

A major limitation of this study is that the trophoblast and the NK cells in the co-culture experiments are not isolated from the same donor and there was a small sample size. Therefore, an indirect co-culture system was employed to minimize direct cell contact. In addition, the origin of human dNK cells is poorly understood. There are two major possibilities: (i) dNK cells are derived from peripheral blood NK cells and differentiated into dNK upon stimulation by the decidual microenvironment; and (ii) dNK cells are derived from hematopoietic precursors present in the deciduas. Even though some studies have suggested that the CD34⁺ hematopoietic precursors present in the decidual tissues may develop to dNK cells (Vacca et al., 2011; Chiossone et al., 2014), the first possibility cannot be excluded and both explanations do not exclude each other (Vacca et al., 2011). For example, the resemblance of dNK cells to the CD56^bright peripheral blood NK subset in terms of their surface CD56^brightCD16⁻ phenotype, has suggested that dNK cells may possibly be derived from CD56^bright pNK cells that are seeded in the uterus (Koopman et al., 2003).

In sum, this study shows that GdA stimulates the partial conversion of peripheral blood CD16⁻CD56^bright NK cells to cells with a dNK-like phenotype and enhances the biological activities of the transformed cells on angiogenesis and trophoblast invasion via binding to L-selectin receptor, activation of the ERK signalling and subsequent production of VEGF and IGFBP-1.

Acknowledgements

The authors thank the Hong Kong Red Cross Blood Transfusion Service for their help in obtaining the buffy coat samples, and staff of the Faculty Core Facility, LKS Faculty of Medicine, The University of Hong Kong, for their technical support.

Authors’ roles

C.L.L.: design the study, acquisition of data, analysis of data and drafting the article. M.V., C.X.W., K.K.W.L.: acquisition of data. H.K., M.S., R.H.W.L. and E.H.Y.Ng.: analysis of data and critical revision of the article for important intellectual content. W.S.B.Y. and P.C.N.C.: design the study, analysis of data and drafting the article.

Funding

Hong Kong Research Grant Council (Grant 17122415), Sanming Project of Medicine in Shenzhen and the Finnish Cancer Foundation, Sigrid Jusélius Foundation and the Finnish Society of Clinical Chemistry.

Conflict of interest

The authors have no competing interests to declare.

Author notes

Cheuk-Lun Lee and Madhavi Vijayan authors contributed equally to the work.