Endometrial natural killer (NK) cells reveal a tissue-specific receptor repertoire

Abstract

STUDY QUESTION

Is the natural killer (NK) cell receptor repertoire of endometrial NK (eNK) cells tissue-specific?

SUMMARY ANSWER

The NK cell receptor (NKR) expression profile in pre-pregnancy endometrium appears to have a unique tissue-specific phenotype, different from that found in NK cells in peripheral blood, suggesting that these cells are finely tuned towards the reception of an allogeneic fetus.

WHAT IS KNOWN ALREADY

NK cells are important for successful pregnancy. After implantation, NK cells encounter extravillous trophoblast cells and regulate trophoblast invasion. NK cell activity is amongst others regulated by C-type lectin heterodimer (CD94/NKG2) and killer cell immunoglobulin-like (KIR) receptors. KIR expression on decidual NK cells is affected by the presence of maternal HLA-C and biased towards KIR2D expression. However, little is known about NKR expression on eNK cells prior to pregnancy.

STUDY DESIGN SIZE, DURATION

In this study, matched peripheral and menstrual blood (a source of endometrial cells) was obtained from 25 healthy females with regular menstrual cycles. Menstrual blood was collected during the first 36 h of menstruation using a menstrual cup, a non-invasive technique to obtain endometrial cells.

PARTICIPANTS/MATERIALS, SETTING, METHODS

KIR and NKG2 receptor expression on eNK cells was characterized by 10-color flow cytometry, and compared to matched pbNK cells of the same female. KIR and HLA-C genotypes were determined by PCR-SSOP techniques. Anti-CMV IgG antibodies in plasma were measured by chemiluminescence immunoassay.

MAIN RESULTS AND THE ROLE OF CHANCE

KIR expression patterns of eNK cells collected from the same female do not differ over consecutive menstrual cycles. The percentage of NK cells expressing KIR2DL2/L3/S2, KIR2DL3, KIR2DL1, LILRB1 and/or NKG2A was significantly higher in eNK cells compared to pbNK cells, while no significant difference was observed for NKG2C, KIR2DL1/S1, and KIR3DL1. The NKR repertoire of eNK cells was clearly different from pbNK cells, with eNK cells co-expressing more than three NKR simultaneously. In addition, outlier analysis revealed 8 and 15 NKR subpopulation expansions in eNK and pbNK cells, respectively. In contrast to the pbNK cell population, the expansions present in the eNK cell population were independent of CMV status and HLA-C genotype. Moreover, the typical NKG2C imprint induced by CMV infection on pbNK cells was not observed on eNK cells from the same female, suggesting a rapid local turnover of eNK cells and/or a distinct licensing process.

LIMITATIONS REASONS FOR CAUTION

Based on our previous work and the parameters studied here, menstrual blood-derived eNK cells closely resemble biopsy-derived eNK cells. However, sampling is not done at the exact same time during the menstrual cycle, and therefore we cannot exclude some, as yet undetected, differences.

WIDER IMPLICATIONS OF THE FINDINGS

Our data reveals that NK cells in the pre-implantation endometrium appear to have a dedicated tissue-specific phenotype, different from NK cells in peripheral blood. This may indicate that eNK cells are finely tuned to receive an allogeneic fetus. Studying the endometrial NKR repertoire of women with pregnancy related problems could provide clues to understand the pathogenesis of pregnancy complications.

STUDY FUNDING/COMPETING INTEREST(S)

No external funding was obtained for the present study. None of the authors has any conflict of interest to declare.

TRIAL REGISTRATION NUMBER

NA

NK cells, KIR, killer immunoglobulin receptor, endometrium, menstrual blood, HLA-C, NK cell receptors, NKG2A, NKG2C, CMV

Introduction

Natural killer (NK) cells have an important function in immune defense and reproduction. They are specialized in killing virus-infected and tumor transformed target cells by the balanced action of both activating and inhibitory receptors (Farag and Caligiuri, 2006). The main families of NK cell receptors (NKR) are the natural cytotoxicity receptors (NCR), the immunoglobulin-like transcripts (LILRB), the C-type lectin heterodimer family (CD94:NKGs) and the killer-immunoglobuline-like receptors (KIRs). The inhibitory NKG2A and activating NKG2C receptor, members of the CD94:NKGs family, can recognize and bind HLA-E (Lanier, 2005). The inhibitory receptor LILRB1 can recognize HLA-G (Shiroishi et al., 2003). KIRs are classified into inhibitory (L) and activating (S) receptors. Members of the KIR2D subfamily recognize and bind HLA-C. Based on a polymorphism at the α1 domain of the HLA-C protein, HLA-C is divided into HLA-C1 and HLA-C2. KIR2DL1 and KIR2DS1 can recognize and bind to HLA-C2, while KIR2DL2/L3 can recognize and bind mainly to HLA-C1 (Chazara et al., 2011).

During reproduction and pregnancy, NK cells play an important role in ensuring correct placentation, and normal development and growth of the fetus (Moffett and Loke, 2006). In first trimester decidua, 70% of all lymphocytes are NK cells, which are in close contact with extravillous trophoblast cells (EVT) at the site of implantation. Together with macrophages and stromal cells, these decidual NK (dNK) cells, phenotypically and functionally different from peripheral blood NK (pbNK) cells (Koopman et al., 2003), will promote EVT invasion and spiral artery remodeling, thereby ensuring an adequate blood supply to the fetus (Hanna et al., 2006). NKR on dNK cell are capable of recognizing the unique array of ligands on EVT cells, i.e. HLA-C, HLA-E, and HLA-G (Hanna et al., 2006). Since both KIR and HLA-C haplotypes are highly polymorphic, each pregnancy will be characterized by their own specific combination of KIR and HLA-C. The combination of maternal KIR AA genotype and fetal HLA-C has been associated with recurrent miscarriage and pre-eclampsia (Hiby et al., 2004, 2008).

Up till now, most research focused on dNK cells using first trimester abortion material, showing that dNK cells express various KIRs and are biased towards KIR2D expression (Sharkey et al., 2008; Male et al., 2011). However, little data is available on expression and composition of NKR on endometrial NK (eNK) cells, with one study investigating the composition of uterine NK cell KIR repertoire in menstrual blood (Ivarsson et al., 2017). It has been suggested that eNK cells are immature precursors that differentiate and proliferate into dNK cells after implantation of an embryo (Manaster et al., 2008; Yagel, 2009). As NK cell reactivity is particularly relevant at the time of and early after implantation, analysis of NKR expression around this time may yield insight in the role of NKR⁺ eNK cells for successful pregnancy.

Here, eNK cells, obtained from menstrual blood (van der Molen et al., 2014; Feyaerts et al., 2017), were immunophenotyped for NKR expression with 10-color flow cytometry, and compared to pbNK cells of the same female. Our aim was to determine differences in NKR expression between eNK and matched pbNK cells. In addition, we investigated the stability of NKR expression on eNK cells over different menstrual cycles, and the influence of CMV seropositivity and HLA-C genotype on NKR expression. We hypothesize that, in order to be prepared for successful pregnancy, eNK cells will have a tissue-specific NKR expression pattern, different from pbNK cells.

Materials and Methods

Human subjects

Twenty-five healthy female volunteers (Table I), with regular menstrual cycle were included upon written informed consent with regard to scientific use, according to the Dutch Medical Research Involving Human Subject Act (WMO). Exclusion criteria were known autoimmune diseases, smoking and current use of hormonal contraceptives or a copper intra uterine device. Menstrual blood was collected during the first 36 hours of menstruation, in three 12-h intervals, using a menstrual cup (Femmecup Ltd, London, UK). Collected menstrual blood was stored in a 30 ml tube, containing 10 ml of RPMI 1640 medium, supplemented with pyruvate (1 mM), glutamax (2 mM), penicillin (100 U/ml), streptomycin (100 μg/ml) (Thermo Fisher Scientific, Waltham, USA), 10% v/v human pooled serum (HPS, manufactured in-house), and 0.3% v/v sodium citrate (Merck, Darmstadt, Germany) at room temperature. Isolation of mononuclear cells was performed within 24 h of collection. Peripheral blood was obtained from all women participating.

Table I

Donor characteristics.

	Donors (N = 25)
Age, median (range)	29 years (20–46)
Sampled longitudinally (3 menstrual cycles)	4/25 (16%)
Previous pregnancy	11/25 (44%)
Previous miscarriage	5/25 (20%)
Natural conception	10/11 (91%)
Previous unprotected heterosexual coitus	23/25 (92%)
Menstruation duration, median (range)	28 days (23–35)
Length of menstrual cycle, median (range)	5 days (4–7)
CMV positive	10/21 (48%)1
KIR and HLA-C genotyped	22/25 (88%)2

¹Information not available for four donors.

²Information not available for three donors.

Isolation of lymphocytes

Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation using Lymphoprep (1.077 ± 0.001 g/ml, 290 ± 15 mOsm, Axis-Shield PoC AS, Oslo, Norway). Menstrual blood was first washed with PBS (Braun, Melsungen, Germany), and mucus and blood clots were removed using a 70 μm cell strainer (Falcon®, Corning Inc., NY, USA). The human granulocyte depletion cocktail RosetteSep™ (Stemcell technologies Inc, Vancouver, Canada) was used, according to manufacturer’s instructions, to deplete granulocytes. After density gradient centrifugation (Lymphoprep), menstrual blood mononuclear cells (MMC) were isolated. The remaining cell pellet after isolation of MMC, i.e. red blood cells and granulocytes, was used for DNA isolation. Typically 94% and 96% of respectively PBMC and MMC were viable cells (Supplementary Fig. S1).

Flow cytometric analysis

The following fluorochrome-conjugated monoclonal antibodies were used to phenotypically characterize MMC and PBMC samples: CD3-APC-AF750, CD16-FITC, CD45-KO, CD56-ECD (Beckman Coulter, Fullerton, CA, USA), CD85j-PC5.5 (LILRB1;Beckman Coulter, custom made), CD158a-AF700 (KIR2DL1; R&D systems, Abingdon, UK), CD158a/h-PC5.5/APC-AF700 (KIR2DL1/S1; Beckman Coulter, custom made), CD158b2-FITC (KIR2DL3; R&D systems), CD158b1/b2-PC7 (KIR2DL2/L3/S2; Beckman Coulter), CD158e1-BV421 (KIR3DL1; BioLegend, San Diego, USA), CD158e1/e2-APC (KIR3DL1/S1; Beckman Coulter), CD159a-APC/PB (NKG2A; Beckman Coulter, custom made), CD159c-PE (NKG2C; R&D systems), and Fixable Viability Dye-eFluor780 (eBioscience, San Diego, USA). Custom made antibodies were kindly provided by Beckman Coulter. Briefly, cells were washed twice with PBS-bovine serum albumin (BSA; Sigma-Aldrich, St. Louis, USA) (0.2% v/v), followed by staining with the antibodies for 20 min at room temperature in the dark. After washing twice with PBS-BSA (0.2% v/v), cells were analyzed using the 10-color flow cytometer Navios™ (480 nm argon blue laser, 405 nm solid state violet laser, 636 nm solid state laser, Beckman Coulter). A minimum of 250 000 cells per staining was applied. Unstained samples and samples without the presence of the specific KIR gene were used for gate settings. Data analysis was performed using Kaluza 1.5a software (Beckman Coulter). A typical gating strategy used for analysis of NK cells in menstrual and peripheral blood is depicted in Supplementary Fig. S1.

KIR and HLA genotyping

Genomic DNA was isolated from the remaining cell pellet after MMC isolation using silica gel membrane technology (Qiagen, Hilden, Germany). KIR genotype and HLA-C allelic variants were typed by PCR-SSOP, using Luminex 100 (Sanbio, Uden, Netherlands) for the following genes: KIR2DL1/S1, KIR2DL1/L2/L3/L4/L5, KIR3DL1/L2/L2, KIR2DS1/S2/S3/S4/S5, KIR3DS1, KIR2DP1, KIR3DP1, HLA-C1 and HLA-C2.

CMV seropositivity

Plasma was obtained from peripheral blood to determine anti-CMV IgG antibodies by chemiluminescence immunoassay (CMV IgG II, Liaison XL, DiaSorin, Saluggia, Italy).

SPADE analysis

Unsupervised analysis of all PBMC and MMC flow cytometry data was performed with the SPADE algorithm (stand-alone version 3.0; Qiu et al., 2011). After compensation and preliminary gating with Kaluza 1.3, CD45⁺CD56⁺CD3^– NK cells (gating strategy Supplementary Fig. S1) were isolated from each individual data file by processing with R software (www.r-project.org, last accessed 9 January 2018) using scripts, as developed and kindly provided by S. Schlickeiser and B. Sawitzki (Institute of Medical Immunology, Charité – University Medicine Berlin, Berlin, Germany). These NK cells were used as input during SPADE analysis, while CD16, CD56, NKG2C, NKG2A, LILRB1, KIR2DL2/L3/S2, KIR2DL1/S1 and KIR3DL1/S1 were used as clustering markers.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 5 (GraphPad software Inc., La Jolla, CA, USA). For comparison of paired observations a non-parametric Wilcoxon-Signed Ranked test was used. Non-parametric Mann–Whitney test was used for unpaired observations. Results were shown as mean ± SD and P < 0.05 was considered significant. Co-expression was calculated with the ‘Tree’ function in Kaluza software. A Tukey’s range test was used to identify outliers in NK cell populations with a distinct NKR expression pattern. As described before (Beziat et al., 2013), two additional criteria were used: outliers should represent at least 5% of the total NK cell population and at least 20% of the NKG2A⁺NKG2C^+/–, NKG2A^–NKG2C^–, or NKG2A^–NKG2C⁺ expressing cells.

Results

Endometrial NK cells contain more NKR expressing cells compared to peripheral blood NK cells

We first examined the NK cell receptor (NKR) expression patterns of endometrial NK (eNK) cells from four females over three consecutive menstrual cycles. NK cells were defined as CD45⁺CD56⁺CD3^– lymphocytes (gating strategy Supplementary Fig. S1). Analysis showed that single NKR expression and KIR expression patterns on eNK cells were similar between the different menstrual cycles, even when sampled 15 months apart. This indicates that NKR expression on eNK cells does not differ over different menstrual cycles (Supplementary Fig. S2).

Subsequently, we compared NKR expression of eNK cells from 25 females with matched peripheral blood NK (pbNK) cells. We observed that the percentage of eNK cells expressing NKG2A (86.6% ± 7.6% versus 39.5% ± 12.2%), LILRB1 (65.6% ± 13.2% versus 38.2% ± 20.6%), KIR2DL2/L3/S2 (64.9% ± 14.5% versus 32.7% ± 11.9%), KIR2DL3 (36.3% ± 16.9% versus 12.4% ± 10.1%), and KIR2DL1 (23.4% ± 12.1% versus 15.5% ± 8.5%) was significantly higher compared to matched pbNK cells (Fig. 1). No significant difference was observed in the frequency of NK cells expressing NKG2C (18.3% ± 9.4% versus 19.2% ± 20.3%), KIR2DL1/S1 (37.8% ± 14.5% versus 30.5% ± 14.9%) and KIR3DL1 (19.7% ± 13.9% versus 16.0% ± 11.4%). Overall, eNK cells contained more NKR expressing NK cells compared to pbNK cells, which might suggest that there is skewing of expression by eNK cells towards HLA recognition.

Figure 1

Differential expression of natural killer (NK) cell receptors on NK cells found in peripheral and menstrual blood samples (N = 25). (A) Representative staining for KIR2DL2/L3/S2, KIR2DL3, KIR2DL1/S1, KIR2DL1, KIR3DL1, NKG2A, NKG2C and LILRB1 on endometrial (eNK) and peripheral blood (pbNK) NK cells. (B) Percentages of NK cells expressing NKG2A, NKG2C, LILRB1, KIR2DL2/L3/S2, KIR2DL1/S1, KIR3DL1, KIR2DL3 and KIR2DL1 are shown for paired pbNK and eNK cell samples. *P < 0.05, ***P < 0.001.

During differentiation, pbNK cells will first acquire KIR2DL2/L3/S2 expression before acquiring KIR2DL1/S1 (Fischer et al., 2007; Schonberg et al., 2011). Our data showed that the percentage of eNK cells expressing KIR2DL2/L3/S2 alone, and expressing both KIR2DL1/S1 and KIR2DL2/L3/S2 was significantly higher compared to pbNK cells, while very few eNK cells were positive for KIR2DL1/S1 alone (Supplementary Fig. S3A). This phenomenon may suggest that eNK cells are only recently matured.

Uterine NK cells can recognize HLA-E on trophoblast cells by the activating NKG2C and inhibitory NKG2A receptor (King et al., 2000). The percentage of NKG2A⁺NKG2C^– and NKG2A⁺NKG2C⁺ eNK cells was significantly higher compared to pbNK, while the frequency of NKG2A^–NKG2C⁺ and NKG2A^–NKG2C^– eNK cells was significantly lower (Supplementary Fig. S3B). Remarkably, the majority of eNK cells were NKG2A⁺NKG2C^–, which might suggest that recognition of HLA-E on trophoblast cells by an inhibitory receptor is preferred.

Endometrial NK cells have a distinct NKR expression pattern and co-express multiple NKR

By combining the expression of KIR2DL2/L3/S2, KIR2DL1/S1, KIR3DL1/S1, LILRB1, NKG2A and NKG2C, we identified 64 phenotypically distinct populations of NK cells (hereafter referred to as subpopulations, although this term does not infer that the phenotypes of these populations are fixed). We observed that the NKR repertoire of eNK was different from pbNK cells and has distinct features (Fig. 2A and Supplementary Fig. S4). Although the majority of eNK cell subpopulations were NKG2A⁺, the percentage of NKG2A single positive eNK cells appeared lower compared to pbNK cells. The frequency of NK cells negative for any of the analyzed NKR was lower for eNK compared to pbNK cells (Fig. 2B). Interestingly, the majority of eNK cells co-expressed NKG2A and KIR2DL2/L3/S2 together with other NKR. This may suggest that eNK cells are biased towards HLA-E and HLA-C1 recognition. The NKR repertoire on eNK cells is more diverse than on pbNK cells, as is visible by the high frequency of eNK cells co-expressing three or more NKR (Fig. 2C). Analysis of matched pbNK and eNK cells of the same female gave similar results (Supplementary Fig. S4). A Tukey’s outlier analysis of the 64 NKR subpopulations revealed expansions of eight subpopulations within the eNK cell pool and 15 expansions within the pbNK cell pool (Table II and Fig. 2A and B, outliers marked as diamond shapes). Only 4 out of 12 females showed expansions both in eNK and pbNK cells. These expansions were tissue-specific, since different expansions were observed in eNK versus pbNK cells of the same female.

Table II

Summary table of natural killer receptor (NKR) subpopulation expansions of NK cells in (A) endometrium and (B) peripheral blood.

Donor	Expansion			CMV seropositivity	HLA-C	Self-specific KIR expansion?
A: Endometrium
MBC14	NKG2A+LILRB1+ KIR2DL2/L3/S2+			–	C2C2	No
MBC31	NKG2A+LILRB1+ KIR2DL2/L3/S2+ KIR2DL1/S1+			–	C1C2	Yes
MBC33	LILRB1+KIR2DL1/S1+ NKG2C+			+	C2C2	Yes
MBC34	NKG2C+			+	C1C2	No
MBC40	null			–	C1C1	No
MBC46	NKG2A+LILRB1+			–	C1C1	No
MBC50	Null	KIR2DL2/L3/S2+		–	C2C2	No
B: Peripheral blood
MBC14	LILRB1+	LILRB1+ KIR2DL2/L3/S2+		–	C2C2	No
MBC16	NKG2A+NKG2C+	LILRB1+ KIR2DL2/L3/S2+ NKG2C+	NKG2A+LILRB1+ NKG2C+	+	C1C1	Yes
MBC19	LILRB1+ KIR2DL1/S1+ NKG2C+			+	C2C2	Yes
MBC27	KIR2DL2/L3/S2+ NKG2C+	LILRB1+ KIR2DL2/L3/S2+ NKG2C+		?	C1C2	Yes
MBC33	LILRB1+ KIR2DL1/S1+ NKG2C+			+	C2C2	Yes
MBC34	LILRB1+ KIR2DL1/S1+ NKG2C+			+	C1C2	Yes
MBC35	KIR2DL2/L3/S2+ KIR2DL1/S1+			+	C1C2	Yes
MBC48	KIR2DL2/L3/S2+ KIR2DL1/S1+ KIR3DL1/S1+NKG2C+	LILRB1+ KIR2DL2/L3/S2+ KIR2DL1/S1+NKG2C+	LILRB1+KIR2DL2/L3/S2+ KIR2DL1/S1+ KIR3DL1/S1+ NKG2C+	+	C2C2	Yes
MBC50	KIR2DL2/L3/S2+ KIR2DL1/S1+			–	C2C2	Yes

Figure 2

Natural killer (NK) cell receptor (NKR) expression patterns on NK cells are different in menstrual blood compared to peripheral blood (N = 25). (A, B) Frequency of each NK cell subpopulation (KIR2DL2/L3/S2^+/–KIR2DL1/S1^+/–KIR3DL1/S1^+/–LILRB1^+/–NKG2C^+/–NKG2A^+/–) is shown as a percentage of total NK cells for paired endometrial NK (eNK) and peripheral blood (pbNK) cells. (A) NKG2A positive NK cell subpopulations and (B) NKG2A negative NK cells are shown separately. Lines indicate mean. Outliers are marked with a diamond shape. (C) Cumulative percentage of NK cells expressing 0, 1, 2, 3, 4, 5, or six NKR (mean ± SD, paired eNK and pbNK cells). *P < 0.05, **P < 0.01, ***P < 0.001.

Béziat et al. showed that CMV infection induced expansion and differentiation of self-specific inhibitory KIR-expressing pbNK cells (Beziat et al., 2013). Our data indeed showed that in pbNK cells 60% of CMV^pos females (6/10) displayed NKR subpopulation expansions, while only 18% of CMV^neg females (2/11) (Tables I and II; CMV status unknown for four females). However, in eNK cells, only 20% of CMV^pos (2/10) and 45% of CMV^neg (5/11) females displayed expansions. Interestingly, in pbNK cells, 73% of expansions (11/15) expressed self-specific KIR, while this was only the case for 25% (2/8) of eNK cells (Table II). Hence, the data suggests that, in contrast to peripheral blood, NKR subpopulation expansions in endometrium are not associated with CMV seropositivity nor express self-KIR.

In addition, unsupervised clustering of our dataset with SPADE analysis showed a higher diversity within the eNK cell population, with 19 additional subpopulations not present in the pbNK cell population. In contrast, pbNK cells revealed only one subpopulation that was not found in the eNK cell population (missing subpopulations indicated with arrows and node number in CD56 plot of Fig. 3). The majority of the 19 additional subpopulations present in the eNK cell population are CD56^bright NK cells with high expression of NKG2A and KIR (Supplementary Fig. S5). The one subpopulation absent in eNK cells is CD16⁺, and LILRB1⁺ with low CD56, KIR and NKG2 expression.

Figure 3

SPADE analysis of natural killer cell receptor (NKR) expression pattern on endometrial NK (eNK) cells and peripheral blood NK (pbNK) cells (N = 25). Each tree depicts the expression levels of one marker. SPADE is a visualization method that organizes data in a 2D-plot based on similarities in expression of the selected markers. The nodes of the tree represent clusters of cells with similar marker expression and the size of the node is a representation of the amount of cells with these characteristics. The color red and blue represents, respectively, high or low expression of a particular marker. The orientation of a node is inferred from cellular hierarchies, i.e. more similar subpopulations are adjacent to each other. Arrows and representative node numbers in CD56 figure indicate which subpopulations are absent in endometrium or periperhal blood while still present in the other.

In contrast to peripheral blood NK cells, CMV has no imprint on the endometrial NK cell phenotype

Guma et al. reported that CMV^pos individuals have higher frequencies of NKG2C⁺ pbNK cells compared to CMV^neg individuals (Guma et al., 2004). Our data showed that the CMV imprint on pbNK cells, i.e. higher percentage of NKG2C⁺ NK cells, is not observed on eNK cells of the same female (Fig. 4). This suggests that, in contrast to pbNK cells, CMV seropositivity does not affect NKG2C expression of eNK cells.

Figure 4

No increased percentage of NKG2C⁺ natural killer (NK) cells in menstrual blood of cytomegalovirus (CMV) positive females. Frequencies of NK cells expressing NKG2C are shown for paired endometrial NK (eNK) and peripheral blood NK (pbNK) cells from females seropositive (CMV+; N = 10) or seronegative for CMV (CMV−; N = 11). Lines indicate mean. *P < 0.05, **P < 0.01.

HLA-C genotype does not influence KIR expression on endometrial NK cells

HLA-C genotype has been shown to influence expression of cognate KIR receptors on pbNK cells (Schonberg et al., 2011). In order to study the influence of KIR ligands on self-KIR expression by eNK cells, HLA-C and KIR genotypes of 22 females were determined. In pbNK cells, the presence of the HLA-C2 epitope resulted in increased percentage of KIR2DL1/S1⁺ NK cells (Fig. 5A and B). This effect was not observed in eNK cells. No significant effect of HLA-C genotype was observed for KIR2DL1⁺ and KIR2DS1⁺ pbNK and eNK cells. The HLA-C1 epitope did not affect the percentage of KIR2DL2/L3/S2⁺ pbNK and eNK cells. In addition, independent of the HLA-C genotype, more KIR2DL2/L3/S2⁺ eNK cells could be observed compared to the pbNK cell population.

Figure 5

No influence of HLA-C genotype on the expression of cognate KIR2D receptors on menstrual blood-derived natural killer (NK) cells (N = 22). (A) The frequency of NK cells expressing KIR2DL1/S1, KIR2DL1, and KIR2DS1 is shown as mean ± SD for peripheral blood NK (pbNK) and endometrial NK (eNK) cells of females with presence (C2Cx, HLA-C2C2 and HLA-C1C2 genotype) or absence (C1C1, HLA-C1C1 genotype) of C2 epitope. (B) The frequency of NK cells expressing KIR2DL2/L3/S2 is shown as mean ± SD for pbNK and eNK cells of females with presence (C1Cx, HLA-C1C1 and HLA-C1C2 genotypes) or absence (C2C2, HLA-C2C2 genotype) of C1 epitope. (C, D) Frequency of each NK cell subset, expressing different KIR combinations (KIR2DL2/L3/S2^+/–KIR2DL1/S1^+/–KIR3DL1/S1^+/–) is shown as a percentage of total NK cells. (C) eNK and pbNK cells of four females with the same KIR genotype and the same HLA-C genotype. (D) eNK and pbNK cells of five females with the same KIR genotype but a different HLA-C genotype. Lines indicate mean. *P < 0.05.

Based on the combined expression of KIR2DL2/L3/S2, KIR2DL1/S1 and KIR3DL1/S1, we identified eight different NK cell subpopulations. Four females with the same KIR and HLA-C genotype showed similar KIR expression patterns on eNK and pbNK cells (Fig. 5C). Moreover, five females with the same KIR but different HLA-C genotype also had similar KIR expression patterns on eNK cells, while the pattern on pbNK cells was more diverse (Fig. 5D). This suggests that the HLA-C genotype does not have an influence on KIR expression of eNK cells.

Discussion

NK cells play an important role in ensuring correct placentation and normal development and growth of the fetus (Moffett and Loke, 2006). It has been implied that eNK cells are immature precursors that differentiate into dNK cells after implantation of an embryo (Manaster et al., 2008; Yagel, 2009). In order to be prepared for implantation, we hypothesize that eNK cells will have a unique NK cell receptor (NKR) expression pattern compared to pbNK cells. In this study, we directly compared the NKR receptor repertoire of NK cells present in menstrual (a source of endometrial cells) and matched peripheral blood of the same females using 10-color flow cytometry. We showed that, within the same female, the NKR expression profile of eNK cells does not differ between consecutive menstrual cycles. The NKR expression pattern of eNK cells clearly differs from pbNK cells, with most eNK cells expressing ≥3 NKR simultaneously, while pbNK cells mainly express <3 different NKR. In addition, the observed NKR subpopulation expansions of eNK cells were tissue-specific, and in contrast to pbNK cells, independent of HLA-C genotype or CMV status. This unique, tissue-specific NKR repertoire of eNK cells might be important to prepare eNK cells towards the reception of an allogeneic fetus.

The frequency of NKG2A, LILRB1, KIR2DL2/L3/S2, KIR2DL3 and KIR2DL1 positive NK cells was significantly higher for eNK cells, similar to previously reported data (Ivarsson et al., 2017). The higher frequency of NKR⁺ eNK cells implies that the bias towards HLA recognition is already present before pregnancy (Lukassen et al., 2004; Ivarsson et al., 2017), and is not only a feature of pregnancy (Sharkey et al., 2008; Male et al., 2011; Xiong et al., 2013). In addition, while the HLA-C genotype of an individual is of influence on the expression of cognate KIR receptors on pbNK cells (Schonberg et al., 2011; Sharkey et al., 2015), our data showed that maternal HLA-C did not seem to affect cognate KIR2D expression on eNK cells. A role for maternal HLA-C in selecting self-KIR expression on dNK and eNK cells has only been found by Sharkey et al. (2015) but not by others (Male et al., 2011; Xiong et al., 2013; Ivarsson et al., 2017).

eNK cells can recognize HLA-E on trophoblast cells by the activating NKG2C and inhibitory NKG2A receptor (King et al., 2000). We observed that almost all eNK cells are NKG2A⁺ (~90%), a feature of uterine NK cells also observed by many others (Lukassen et al., 2004; Kusumi et al., 2006; Male et al., 2011; Sharkey et al., 2015). Moreover, the majority of eNK cells expressed NKG2A alone or together with NKG2C, while very few eNK cells expressed NKG2C alone. In peripheral blood, NKG2A regulates the response of NKG2C⁺ NK cells against target cells expressing HLA-E (Saez-Borderias et al., 2009; Beziat et al., 2010). This might suggest that recognition of HLA-E on trophoblast cells by NKG2A is preferred to avoid uterine NK cell mediated cytotoxicity towards trophoblast cells.

By combining the expression of KIR2DL2/L3/S2, KIR2DL1/S1, KIR3DL1/S1, LILRB1, NKG2A and NKG2C we identified 64 phenotypically distinct populations of NK cells. We observed that the NKR repertoire of eNK and pbNK cells are different and have distinct features. Interestingly, the majority of eNK cells co-expressed NKG2A and KIR2DL2/L3/S2 together with other NKR, a feature also observed on dNK cells (Sharkey et al., 2015). It may suggest that eNK and dNK cells are biased towards HLA-E and HLA-C1 recognition. Both KIR2DL2 and KIR2DL1 can recognize HLA-C2 but during KIR acquisition, NK cells will first acquire KIR2DL2/L3 before they can acquire KIR2DL1 (Schonberg et al., 2011). Our results showed that the percentage of eNK cells positive for KIR2DL2/L3/S2 alone, and expressing both KIR2DL1/S1 and KIR2DL2/L3/S2 was significantly higher compared to pbNK cells, while the percentage of eNK cells expressing KIR2DL1/S1 alone was significantly lower, which is in line with previous results on eNK and dNK cells (Male et al., 2011). This may suggest that eNK cells are only recently matured. In addition, preference for KIR2DL2/L3/S2 on eNK cells may decrease the influence of KIR2DL1, which could be a protective mechanism for successful pregnancy since the combination of the KIR2DL1 and HLA-C2 genotype has been associated with pre-eclampsia and recurrent miscarriage (Hiby et al., 2004, 2008).

pbNK cells were shown to mainly express up to three receptors simultaneously, while eNK cells also co-expressed more receptors, suggesting that the NKR repertoire on eNK cells is more diverse. KIR repertoire acquisition is strongly modulated by HLA class I molecules. TAP-deficient patients, whose surface expression of HLA class I is significantly reduced, show a higher frequency of NKG2A⁺KIR⁺ NK cells and KIR co-expression (Sleiman et al., 2014). A more diverse NKR co-expression on eNK cells could suggest that eNK cells still need to mature, possibly this will take place only after implantation and might depend on fetal HLA. This diverse NKR repertoire could be necessary for eNK cells to successfully encounter an embryo which could have any possible fetal HLA phenotype.

A Tukey’s outlier analysis of our 64 NKR subpopulations identified eight subpopulation expansions in eNK cells, and 15 expansions in pbNK cells. These expansions were tissue-specific since different expansions were observed in eNK and pbNK cells. In accordance with Béziat et al., we showed that CMV infection leads to expansion of self-specific KIR-expressing pbNK cells (Beziat et al., 2013). In endometrium, the observed expansions were not associated with CMV infection and not biased towards self-KIR expression. This is in contrast with Ivarsson et al., who reported NKG2C⁺self-KIR⁺ NK cell expansions in endometrium (Ivarsson et al., 2017). However, in their study it was not possible to investigate the association between CMV and self-KIR due to a low percentage of CMV^neg females. CMV viral infection has also been correlated with a higher proportion of NKG2C⁺ pbNK cells (Guma et al., 2004), which was also observed in our study for pbNK cells but not for eNK cells. Overall, this suggests that the imprinting observed on pbNK cells is not found on eNK cells. Since the origin of eNK cells is still unclear, there are several possibilities for this discrepancy between expansions in endometrium and peripheral blood. It may reflect selective recruitment of phenotypically distinct populations from peripheral blood to endometrium, induction of distinct populations under influence of the uterine microenvironment after recruitment of immature NK cells, or local proliferation from progenitor cells which are not influenced in the same way by viral infection as pbNK cells (Manaster and Mandelboim, 2010).

Although we previously showed that menstrual blood lymphocytes are very similar to endometrial biopsy-derived lymphocytes (van der Molen et al., 2014), and show similar expression patterns over consecutive cycles, it is unknown if NKR expression will change over the course of the menstrual cycle or how expression of menstrual blood NK cells will relate to eNK cells at time of implantation. Nevertheless, it could be a valuable tool to investigate the immune cell composition of women with pregnancy complications and fertility issues. A limitation of our study is that no additional markers were used to identify innate lymphoid cells type3 (ILC3s), which can also be CD56⁺ (Hazenberg and Spits, 2014). However, in our hands ILC3s constitute only 0.04% and 0.002% of total lymphocytes in menstrual and peripheral blood, respectively (data not shown). Therefore, we consider the contribution of ILC3s to the NK cell pool to be limited. In addition, in our previous work we showed that eNK cells are from mucosal origin (Feyaerts et al., 2017). Therefore, we did not use a tissue resident marker to differentiate pbNK from eNK cells.

In summary, this study shows that the NKR expression profile of eNK cells appears truly unique, co-expressing multiple NKR. The repertoire does not differ over consecutive menstrual cycles, and is independent of CMV status or HLA-C genotype. This unique, tissue-specific NKR repertoire might be important to prepare the endometrium towards reception of an allogeneic fetus and successful pregnancy. Therefore, it would be interesting to study the NKR repertoire of eNK cells in more detail in the future. Studying the endometrial NKR repertoire of women with pregnancy related problems could provide clues to understand the pathogenesis of pregnancy complications.

Supplementary data

Supplementary data are available at Human Reproduction online.

Authors’ roles

D.F. and T.K. coordinated and designed study, analyzed data and wrote manuscript. R.G.M. and I.J. conceived and designed study, analyzed data and wrote manuscript. B.C. and S.Z. participated in data collection, and assisted in interpretation of data. O.W.H.H. and A.M. participated in design of study, and provided critical discussion. All authors evaluated the manuscript and contributed to its content.

Funding

No external funding was obtained for this study.

Conflict of interest

None of the authors have a conflict of interest.

Acknowledgements

The authors wish to thank all the women who participated in this study, Konstantina Nikolakopoulou who assisted with sample processing, and Marion Dinnissen-van Poppel and Paul Daemen from the department of Medical Microbiology (Radboud University Medical Center) for anti-CMV IgG determination.

Author notes

†

The authors consider that the first two authors should be regarded as joint First Authors.