|Home | About | Journals | Submit | Contact Us | Français|
Uterine natural killer (uNK) cells accumulate at the maternal-fetal interface during gestation and are thought to have an important role during pregnancy in both mice and humans. While the cell surface phenotype of human uNK cells is increasingly well defined, less is known regarding the cell surface expression profile of murine uNK cells both before and during gestation. Herein, we demonstrate that murine NK1.1+ (KLRB1C) endometrial NK (eNK) cells, derived from virgin mice, and NK1.1+ decidual NK (dNK) cells, obtained from pregnant mice, belong to the B220+ (PTPRC) CD11c+ (ITGAX) subset of NK cells. While B220 expression was low on NK1.1+ eNK cells, it was increased on a subset of NK1.1+ dNK cells at Embryonic Day 10.5. Endometrial NK and dNK cells also differed somewhat in their expression patterns of two activation markers, namely, CD69 and inducible costimulator (ICOS). The eNK cells acquired a B220hiICOS+ dNK cell surface phenotype when cultured in vitro in the presence of uterine cells and murine interleukin 15. Thus, the cell surface profiles generated for both NK1.1+ eNK cells and dNK cells demonstrate that they belong to the recently described B220+CD11c+ subset of NK cells, which are potent cytokine producers.
In addition to their fundamental role in host defense, natural killer (NK) cells are thought to have an important role during pregnancy [1–3]. Natural killer cells are present in the uterus both before and during gestation in mice and humans [3, 4]. In mice, it is thought that NK progenitor cells migrate mainly from secondary lymphoid organs to the uterus, at which point they undergo further differentiation . In humans, it has been proposed that, in addition to the recruitment of NK cells from the periphery, some uterine NK (uNK) cells may arise from hematopoietic stem cells present in the endometrium [6–12]. In mice, decidual NK (dNK) cell numbers peak near midgestation and begin to decline shortly thereafter [13, 14]. Few NK cells can be detected in the mouse uterus at the end of pregnancy . In humans, dNK cell numbers are highest in the first trimester of pregnancy, when they are thought to comprise the majority of decidual lymphocytes [16–18]. Decidual NK cell numbers begin to decline during the second trimester, and, similar to mice, few human NK cells are present in the uterus at term [1, 19].
Studies [20–22] utilizing NK cell-deficient mouse models demonstrated that mice deficient in NK cells display implantation site anomalies, incomplete uterine spiral artery remodeling, and impaired decidualization of uterine stromal cells. Subsequent studies [23, 24] revealed that interferon γ (IFNγ, IFNG) was the key cytokine produced by murine dNK cells that supported pregnancy-associated vascularization of the uterus and the process of decidualization. Human dNK cells have been shown to secrete cytokines, chemokines, and angiogenic factors and are believed to regulate trophoblast invasion [10, 25–30]. Thus, uNK cells are thought to help establish a successful pregnancy through constructive effects on vascularization and placentation.
Human NK cells can be divided into two functionally distinct subsets based on levels of CD56 (NCAM1) expression [31, 32]. CD56bright NK cells are thought to be effective cytokine producers but poorly cytotoxic . In contrast, CD56dim NK cells are thought to be poor cytokine producers yet highly cytotoxic . During a normal human pregnancy, the vast majority of uNK cells display a CD56brightCD16 (FCGR3)− cell surface phenotype [16, 18, 34]. A functional homologue of human CD56 does not exist in mice; thus, it has been difficult to compare NK cell subsets between the two species.
Recently, several studies [35–37] have demonstrated that peripheral B220+CD11c+NK1.1+ cells belong to the NK cell lineage. B220+CD11c+NK1.1+ NK cells were shown to produce greater amounts of IFNγ than conventional NK cells in response to certain stimuli [35, 37]. While it has been suggested that B220+CD11c+NK1.1+ cells correspond to activated NK cells , it has also been proposed that these cells may be analogous to human CD56bright NK cells . Given that CD56bright cells represent the vast majority of NK cells found in human decidua during pregnancy and in an attempt to delineate a potential physiologic role for these cells, we examined whether murine uNK cells belong to the B220+CD11c+NK1.1+ subset of NK cells.
Six- to seven-wk-old female C57BL/6NCr mice were purchased from the National Cancer Institute (Frederick, MD). Mice in estrus were mated with males of proven fertility overnight. The morning on which the vaginal plug was detected is referred to as Embryonic Day (E) 0.5. All procedures described herein were reviewed and approved by the animal studies committee at Washington University and were performed in accord with institutional animal care and use committee approval.
Virgin or pregnant mice at E10.5 were euthanized and perfused with 20 ml of PBS (Sigma-Aldrich, St. Louis, MO) containing 1 μg/ml heparin (Sigma-Aldrich), and the uteri were subsequently isolated. The mesometrium was removed from the uteri, which were cut directly below the oviduct and above the cervix. The uteri were subsequently placed in dishes containing 4°C Hanks balanced salt solution (HBSS; Invitrogen, Carlsbad, CA). Uteri derived from virgin mice were cut longitudinally, minced into small pieces of approximately 1 mm3, and transferred to 50-ml conical tubes containing 4°C HBSS. Interimplantation sites were removed from uteri derived from pregnant mice. The remaining implantation sites were cut along the antimesometrial edge. The fetus and placenta were removed from each site, and the placenta was retained in a separate dish. The uteri, including the myometrium and the mesometrial lymphoid aggregate of pregnancy, were then minced into small pieces as already described and transferred to a 50-ml conical tube containing 4°C HBSS. The decidua basalis was subsequently removed from the placenta, minced into small pieces, and added to the uterine tissue already dissected from the pregnant mice.
Dissected uteri derived from virgin and pregnant mice were washed twice in 4°C HBSS to remove blood contamination. The tissue was then digested in 10 ml of HBSS containing 2 mg/ml collagenase type I (Invitrogen). Uteri were digested at 37°C for 60 min with intermittent vortexing. After digestion, the cells were pelleted and resuspended in 5 ml of ice-cold fetal bovine serum (FBS) (Sigma-Aldrich). They were then pelleted again and resuspended in PBS containing 10% FBS. The cells were passed through a 70-μm cell strainer (BD Falcon; Fisher Scientific, Pittsburgh, PA), counted, and stained for analysis by multiparameter flow cytometry. Staining of each sample with 7-amino-actinomycin D (7-AAD) demonstrated that cell viability after collagenase digestion was routinely >75% live cells.
As a control for collagenase treatment, murine splenocytes were either conventionally isolated or isolated by collagenase digestion. Briefly, splenocytes were isolated from adult C57BL/6 mice by mechanically disrupting the spleen and passing the splenocytes through a 70-μm cell filter. Alternately, spleens were minced, and collagenase digested as already described. The red blood cells from both cell preparations were lysed using Tris-buffered ammonium chloride (0.14 M NH4Cl and 0.017 M Tris [pH 7.2]). The cells were washed and stained with the indicated antibody as described herein, and expression was examined by flow cytometry. Live gates were based on 7-AAD fluorescence. The percentages of positive cells for the indicated antibodies are largely similar between conventionally isolated and collagenase-treated splenocytes (Supplemental Fig. S1 available at www.biolreprod.org).
The following antibodies and staining reagents were purchased from eBioscience (San Diego, CA): fluorescein isothiocyanate (FITC)-conjugated mouse IgG2a, FITC-conjugated rat IgG2b, FITC-conjugated Armenian hamster IgG, FITC-conjugated anti-mouse NK1.1 (KLRB1C), FITC-conjugated anti-mouse major histocompatibility complex (MHC) class II (I-A/I-E), FITC-conjugated anti-mouse CD69, FITC-conjugated anti-mouse/rat inducible costimulator (ICOS), FITC-conjugated anti-mouse CD45 (LCA), phycoerythrin (PE)-conjugated rat IgG2a, PE-conjugated mouse IgG2a, PE-conjugated rat IgG2b, PE-conjugated rat IgM, PE-conjugated Armenian hamster IgG, PE-conjugated anti-mouse NK1.1, PE-conjugated anti-mouse NKp46 (NCR1), PE-conjugated anti-mouse pan-NK cells (CD49b, ITGA2, and DX5), PE-conjugated anti-mouse CD69, PE-conjugated anti-mouse ICOS, PE-conjugated anti-mouse CD11c (ITGAX), PE-conjugated anti-mouse CD11b (ITGAM), PE-conjugated anti-mouse/human/rat CD27, PE-conjugated anti-mouse CD45 (LCA), PE-conjugated streptavidin, biotin-conjugated rat IgG2a, biotin-conjugated anti-mouse/human CD45R (B220), PerCP-Cy5.5-conjugated Armenian hamster IgG, PerCP-Cy5.5-conjugated anti-mouse CD3 (CD3E), and allophycocyanin (APC)-conjugated anti-mouse CD45 (LCA). The PE-conjugated anti-mouse CD122, anti-mouse CD16 (FCGR3)/CD32 (FCGR2B)-2.4G2, and 7-AAD were purchased from BD Biosciences (San Jose, CA). The FITC-conjugated Dolichos biflorus agglutinin (DBA) was purchased from Sigma-Aldrich.
A total of 3.5 × 105 uterine cells were incubated in PBS/10% FBS containing 5% normal mouse serum (Sigma-Aldrich), 5% normal rat serum (Sigma-Aldrich), and 2 μg/ml anti-mouse CD16/CD32 (2.4G2) antibody for 15 min at 4°C to block nonspecific antibody binding. The cells were pelleted and resuspended in 100 μl of PBS/10% FBS containing the indicated antibodies or FITC-conjugated DBA lectin for 30 min at 4°C. All antibodies and the FITC-DBA lectin staining reagent were titered to determine optimal concentrations. Each staining cocktail also contained an APC-conjugated anti-mouse CD45 (LCA) antibody and a PerCP-Cy5.5-conjugated anti-mouse CD3 antibody. The cells were washed three times with PBS. Cells stained with a biotinylated primary antibody were then incubated in PE-conjugated streptavidin for 30 min at 4°C. The cells were washed three times in PBS and ultimately resuspended in PBS/10% FBS for analysis. To demonstrate the specificity of DBA lectin binding, the DBA reagent was preincubated with either 100 mM N-acetyl-d-galactosamine or 100 mM of an irrelevant sugar (d-glucose) for 15 min at 4°C before adding the lectin to the uterine cells. Data were collected using a BD FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA) and analyzed with FlowJo software (TreeSTar, Inc., Ashland, OR). The data shown in all figures are gated to include only CD45+ cells and to exclude all CD3+ cells within the CD45+ gate. Thus, the cells shown in all figures are CD45+ and CD3−. All samples in each experiment were run in duplicate. To address cell viability in the duplicate sample, 7-AAD was substituted for the PerCPCy5.5 anti-mouse CD3 antibody. The data obtained from each sample in the pair were compared and yielded identical results (data not shown). At least 15 implantation sites were pooled for each experiment in which dNK cells were analyzed. In addition, uteri from at least 10 virgin female mice were pooled for each experiment in which endometrial NK (eNK) cells were analyzed. All flow cytometry experiments were conducted at least three times each, and the results shown are from one representative experiment.
Uterine cell suspensions were generated from virgin mice as already described. A total of 8 × 105 cells were plated per well in a six-well plate in 2 ml of RPMI-1640 media containing 10% FBS, 2 mM l-glutamine (Cambrex Bio Science, Walkerville, MD), 1 mM sodium pyruvate (Cambrex Bio Science), 0.5% sodium bicarbonate (Cambrex Bio Science), 50 μg/ml penicillin/streptomycin (Cambrex Bio Science), 100 μM β-mercaptoethanol (Sigma-Aldrich), and 100 ng/ml murine interleukin (IL) 15 (eBioscience). Cells were cultured for 48 h at 37°C and 5% CO2. The cells were then harvested using cell dissociation solution (Sigma-Aldrich), stained, and analyzed by flow cytometry as already described. On the day the in vitro uterine cell cultures were harvested for analysis, virgin mice were euthanized to use as controls in the flow cytometry experiments. This experiment was conducted at least three times, and the results shown are from one representative experiment.
To determine whether NK1.1+ eNK and dNK cells display a B220+CD11c+ cell surface phenotype, we analyzed uterine cells isolated from either virgin mice or pregnant mice at E10.5 by multiparametric flow cytometry. First, a forward- and side-scatter gate was placed around the uterine cell population (Fig. 1, panel A1). Next, we gated on CD45+ cells within the forward- and side-scatter gate as shown in Figure 1, panel A2. Finally, although NK1.1 has been shown to be a useful NK cell marker in C57BL/6 mice [38, 39], it is also expressed on NKT cells . However, these two cell populations can be distinguished by expression of the CD3 (CD3E) molecule, which is present on NKT cells but absent on conventional NK cells. Thus, a final gate was drawn within the CD45+ gate to exclude CD3+ cells (Fig. 1, panel A3). The data presented herein represent CD45+CD3− cells.
We were able to detect the NK1.1 marker on a population of uterine cells derived from both virgin and pregnant mice (Fig. 1, panels B2 and C2, respectively). As expected, the number of NK1.1+ cells was greater in the pregnant uterus than the virgin uterus. Thus, a population of eNK and dNK cells expressed the NK1.1 marker, and we subsequently examined the expression of other peripheral NK cell surface markers on the NK1.1+ uNK cells.
We examined B220 protein expression on NK1.1+ eNK cells and dNK cells at E10.5. B220 was detected on NK1.1+ eNK and dNK cells (Fig. 1, panels B2, B3, and B4 and panels C2, C3, and C4, respectively). The majority of NK1.1+ cells present in both the virgin and pregnant mouse uteri were B220+, and, conversely, the majority of B220+ cells were NK1.1+. B220 expression increased on a subset of NK1.1+ dNK cells compared with eNK cells. Thus, the majority of NK1.1+ eNK cells and dNK cells at E10.5 express the B220 protein.
We also detected the integrin subunit CD11c on NK1.1+ eNK and dNK cells (Fig. 1, panels B6, B7, and B8 and panels C6, C7, and C8, respectively). Again, the vast majority of NK1.1+ eNK and dNK cells expressed CD11c protein. Thus, NK1.1+ eNK cells and dNK cells display a B220+CD11c+ cell surface phenotype.
As previously mentioned, the cell surface profile of human uNK cells is better defined than that of murine uNK cells both before and during pregnancy. Thus, we profiled eNK cells and dNK cells at E10.5 for the presence of a panel of cell surface markers expressed on peripheral NK cells. In mice, cells committed to the NK cell lineage express CD122, the IL2/IL15 receptor common β subunit [41, 42]. Virtually all NK1.1+ eNK and dNK cells expressed CD122 (Fig. 2, panels 1 and 2, and Fig. 3, panels 1 and 2, respectively). For Figures 2, ,3,3, and 4, the isotype control data and data depicting cells stained with an anti-NK1.1 antibody alone are presented as corresponding Supplemental Figures S2, S3, and S4. It has been suggested that the natural cytotoxicity receptor NKp46 (NCR1), which is specifically expressed on NK cells, may be used as an NK cell marker across species [43–46]. We examined NKp46 expression and found that NK1.1+ eNK and dNK cells also expressed this protein (Fig. 2, panels 3 and 4, and Fig. 3, panels 3 and 4, respectively). Thus, NKp46 also defines NK1.1+ uNK cells.
Dolichos biflorus agglutinin binds N-acetyl-d-galactosamine glycoconjugates and has been shown to specifically stain murine uNK cells [47, 48]. We examined the ability of both NK1.1+ eNK and dNK cells to bind DBA lectin. Endometrial NK cells did not react with DBA lectin (Fig. 2, panels 5 and 6). In contrast, we were able to detect DBA lectin reactivity on a subset of dNK cells (Fig. 3, panels 5 and 6). To demonstrate binding specificity, the DBA reagent was preincubated with either 100 mM N-acetyl-d-galactosamine (Fig. 3, panel 7) or 100 mM of an irrelevant sugar (d-glucose) (Fig. 3, panel 8) before adding the DBA to the uterine cells. The presence of N-acetyl-d-galactosamine completely blocked DBA lectin binding to the NK1.1+ dNK cells, whereas d-glucose had no effect. Thus, roughly 20% of the NK1.1+ dNK cells isolated demonstrated DBA lectin reactivity.
The B220+CD11c+NK1.1+ peripheral NK cell population recently described in mice was shown to express CD11b (ITGAM) and CD27 [35, 37]. The integrin subunit CD11b and the tumor necrosis factor (TNF) receptor family member CD27 were both expressed on a portion of NK1.1+ eNK and dNK cells (Fig. 2, panels 7 and 8 and panels 9 and 10, respectively, and Fig. 3, panels 9 and 10 and panels 11 and 12, respectively). Not all NK1.1+ eNK and dNK cells appeared to express these proteins.
While MHC II protein expression was detected in both the virgin and pregnant mouse uteri, few NK1.1+ NK cells expressed MHC II molecules (Fig. 2, panels 11 and 12, and Fig. 3, panels 13 and 14, respectively). Thus, few NK1.1+ cells derived from virgin uteri expressed MHC II protein, and pregnancy did not lead to a significant increase in the number of NK1.1+ dNK cells that express MHC II. CD49b (recognized by the monoclonal antibody DX5) is an α2 integrin that is expressed on most NK1.1+CD3− NK cells. CD49b was detected on a subset of both eNK and dNK cells (Fig. 2, panels 13 and 14, and Fig. 3, panels 15 and 16, respectively). It appeared that some NK1.1+ eNK and dNK cells do not express CD49b.
We next examined markers associated with NK cell activation. CD69 (very early activation antigen) is expressed on leukocytes during activation, and its expression has been detected on both eNK and dNK cells in humans [18, 49–51]. Similarly, we found that CD69 is expressed on the majority of murine NK1.1+ eNK and dNK cells (Fig. 2, panels 15 and 16, and Fig. 3, panels 17 and 18, respectively). In addition, CD69 expression increased in the uterus as the result of pregnancy. We also examined the expression of ICOS on uNK cells. ICOS has been shown to support NK cell function, and its expression is increased on activated NK cells . Low levels of ICOS were detected in the virgin uterus (Fig. 2, panel 17). Very few NK1.1+ eNK cells appeared to express this protein (Fig. 2, panel 18). In contrast, uterine ICOS expression increased during gestation, and a subset of NK1.1+ dNK cells were ICOS+. Moreover, the majority of ICOS+ cells were NK1.1+ (Fig. 3, panels 19 and 20).
The data presented thus far demonstrate that the cell surface phenotypes of NK1.1+ eNK cells and dNK cells at E10.5 differ primarily by their levels of B220 and ICOS expression. In addition, CD69 expression is somewhat higher on a subset of dNK cells compared with eNK cells. Interleukin 15 is required for NK cell development and proliferation and is present in the uterus of both mice and humans during gestation [53–58]. Moreover, IL15 has been shown in vitro to expand the B220+CD11b+ population of NK cells . To determine if eNK cells possessed the ability to adopt a dNK cell surface phenotype, we generated a uterine single-cell suspension from virgin mice and cultured the uterine population in vitro in the presence of murine IL15 and uterine cells for 48 h. Virgin female mice were euthanized on the same day the in vitro cell culture was harvested to act as a control in the flow cytometry experiment. B220 expression increased on in vitro-cultured NK1.1+ eNK cells compared with freshly isolated eNK cells (Fig. 4, panels A1 and A2 vs. panels B1 and B2). CD69 expression levels were similar on in vitro-cultured NK1.1+ eNK cells compared with controls (Fig. 4, panels A3 and A4 vs. panels B3 and B4). Finally, ICOS expression was induced on the in vitro-cultured NK1.1+ eNK cells compared with controls, which expressed little to no ICOS protein (Fig. 4, panels A5 and A6 vs. panels B5 and B6). NK1.1 protein expression was lower on eNK cells cultured in vitro compared with freshly isolated eNK cells. This reasons for this phenomenon were unclear.
Human uNK cells are becoming better defined with respect to cell surface markers expressed and effector functions elicited. Less is known regarding the cell surface profile of murine uNK cells and more specifically the cell surface phenotype of peripheral NK cells that migrate to the mouse uterus. However, a recent study by Yadi et al.  has provided insight into the distinct receptor profile of murine uNK cells.
Herein, we demonstrate that both murine NK1.1+ eNK and dNK cells display a B220+CD11c+ cell surface phenotype. This profile is similar to the recently described subset of peripheral NK cells in mice, which are potent IFNγ producers [35–37]. The peripheral B220+CD11c+NK1.1+ NK cell subset was also shown to express CD27 and CD11b [35, 37]. We demonstrated that a portion of B220+CD11c+NK1.1+ eNK and dNK cells also expressed CD27 and CD11b. CD27 is found on immature NK cells and has been used as a marker to define functional subsets within the mature murine NK cell population . CD27+ NK cells were shown to have enhanced effector functions compared with CD27− cells . Therefore, the cell surface phenotype of NK1.1+ eNK and dNK cells generated thus far is similar to that described for peripheral NK cells that are effective cytokine producers. Most important, the positive effects of uNK cells on placentation and vascularization are largely mediated by growth factors, cytokines, chemokines, and angiogenic factors secreted by these cells [1, 3].
The NK cell precursors and immature NK cells express little to no B220, whereas approximately 10%–30% of mature NK1.1+ CD49b+ NK cells in the bone marrow and spleen express this protein . In contrast to what is seen in the periphery, we found that the majority of NK1.1+ eNK and dNK cells expressed B220. Thus, the uterus appears to be enriched in B220+NK1.1+ NK cells both before and during gestation. We also demonstrated that B220 expression dramatically increased on a subpopulation of NK1.1+ NK cells as the result of pregnancy. Previous investigations have demonstrated increased B220 expression on activated NKT cells . In addition, B220 protein expression was reported to be upregulated on splenocytes cultured in the presence of IL2 and in response to TLR stimulation . It was suggested that this protein acts as an NK cell activation marker. At the time of blastocyst implantation and uterine decidualization, uNK cells become activated . We speculate that the increase in B220 expression on NK1.1+ NK cells at E10.5 reflects the activation status of dNK cells.
We demonstrated that NK1.1+ eNK and dNK cells in the mouse express CD122 and NKp46 and that a large subset of these cells also express CD49b. In contrast, very few NK1.1+ eNK cells expressed MHC II molecules. Moreover, pregnancy and thus uNK cell activation did not lead to a significant increase in the number of dNK cells expressing MHC II. In humans, activated NK cells can express MHC II proteins and act as antigen-presenting cells . Recently, activated murine NK cells were also shown to upregulate MHC II expression [35, 37]. It was suggested that MHC II acts as an NK cell activation marker, with specific stimuli leading to the upregulation of this protein . While NK cell activation may lead to MHC II upregulation on peripheral NK cells, we did not find a significant increase in the number of NK1.1+ dNK cells that express MHC II.
Histologically, DBA lectin has been used to identify uNK cells in implantation sites, and more recently it has been used to identify these cells in flow cytometry studies [48, 60, 64–66]. This reagent specifically stains uNK cells but largely does not stain peripheral NK cells [48, 67]. Histological and flow cytometry studies [48, 60] demonstrated that uNK cells derived from virgin mice do not bind DBA lectin, and our data reaffirm this observation. The DBA-reactive uNK cells appear early during gestation at approximately Day 5 [15, 48, 67]. Yadi et al.  recently identified two uNK cell populations in the uterus at E9.5. The larger subset displayed a distinct cell surface phenotype in part because the uNK cells (CD122+CD3−) were NK1.1−DX5−DBA+. The majority of mature NK cells express NK1.1 and DX5 [42, 68, 69]. The smaller subset resembled peripheral NK cells, as they were DBA− NK1.1+DX5+. Although we isolated DBA+ uNK cells, our enzymatic procedure did not allow for the isolation of the uNK cell population described by Yadi et al. , as we found that virtually all NK1.1+ dNK cells were also CD122+. It appears that we have mainly isolated immature uNK cells given that the dNK cells were NK1.1+CD122+ and that approximately 80% of the dNK cells were DBA−. The larger more fragile DBA+ uNK cells do not appear to survive this isolation procedure. The origin of DBA+ uNK cells is unknown. In mice, it is thought that the self-renewing progenitors of uNK cells traffic to the uterus from the periphery . It is tempting to speculate that NK1.1+ uNK cells, specifically those of the B220+CD11c+ subset, migrate from the periphery to the uterus, where they may further differentiate into DBA+ uNK cells.
A recent study  demonstrated differential expression of certain NK cell receptors on human eNK and dNK cells. Specifically, eNK cells were shown to lack NKp30 and chemokine receptors, whereas dNK cells express these proteins. The eNK cells stimulated with IL15 demonstrated a marked increase in NKp30 and NKp44 protein expression. Moreover, it was shown that eNK cells were functionally static until stimulated with IL15, at which point they acquired the ability to mediate NK cell effector functions . Analogous results were obtained in a separate study  using IL12 and IL15. Similarly, we demonstrated that murine NK1.1+ eNK cells and dNK cells differ in their expression of certain cell surface proteins. NK1.1+ eNK cells expressed low levels of B220 and little to no ICOS, whereas a subset of NK1.1+ dNK cells expressed high levels of B220 and were ICOS+. In vitro culture of eNK cells with IL15 and uterine cells led to a B220hiICOS+ cell surface phenotype on the NK1.1+ NK cells. Thus, eNK cells may be an early form of certain dNK cells. We cannot conclude that the upregulation of the aforementioned cell surface proteins was a direct effect of IL15. The eNK cells were not purified; they were cultured in the presence of uterine cells. Thus, the effects of IL15 on eNK cells may in part be mediated by effects of this cytokine on the uterine cells present in the culture.
Uterine stromal cells produce IL15, and its expression is upregulated by progesterone, which is important in the decidualization process [56–58, 72, 73]. In addition, immune cells present in the uterus have been shown to secrete this cytokine [56, 74, 75]. Other potential stimulators of eNK cells are also present in the nonpregnant endometrium. MICA, ULBP2, and ULBP3, which are NKG2D (KLRK1) ligands, have been detected on human endometrial cells [70, 76]. We and others have demonstrated that eNK cells express the activation marker CD69 . If, similar to humans, murine endometrial cells in the nonpregnant uterus express IL15 and NKG2D ligands, in theory a minimal amount of eNK cell stimulation could occur, an amount that potentially could lead to low levels of B220 and CD69 protein expression on CD11c+NK1.1+ eNK cells. In humans, findings suggest that eNK cells mediate low levels of cytotoxicity . Although, in mice, eNK cells have been shown to be immature and nongranulated [15, 79, 80]. Alternately, CD69 could act to keep NK cells in the uterus, as it has been shown to retain lymphocytes in lymphoid tissue . We cannot rule out the possibility that the cell isolation procedure led to the expression of the CD69 marker on eNK cells. However, ICOS was not upregulated under the same purification conditions.
Thus, the cell surface phenotype of NK1.1+ eNK and dNK cells, shown in Figure 5, is similar to the recently identified B220+CD11c+ peripheral NK cell population found to secrete higher amounts of IFNγ than conventional NK cells. Given that the majority of NK1.1+ uNK cells present in the uterus are also B220+CD11c+, it is tempting to speculate that the peripheral pool of NK cells with this cell surface phenotype are those that migrate to the uterus. However as previously mentioned, if B220 acts as a uNK cell activation marker, then the presence of IL15 or NKG2D ligands in the virgin uterus could potentially lead to subthreshold stimulation of the uNK cells, with induction of low levels of B220 protein expression at the cell surface. Further studies are warranted to distinguish between these two possibilities.
We thank Dr. Anne Croy and Dr. Leonidas Carayannopoulos for their invaluable advice and thoughtful discussions. We also thank Sarah Frazier for technical assistance.
1Supported by NIH grants 5K12HD00145908 and 2P60DK02057931 (to J.K.R.).