|Home | About | Journals | Submit | Contact Us | Français|
Understanding of heterogeneous NK subsets is important for the study of NK cell biology and development, and for the application of NK cell-based therapies in the treatment of disease. Here we demonstrate that the surface expression of CD94 can distinctively divide mouse NK cells into two approximately even CD94low and CD94high subsets in all tested organs and tissues. The CD94high NK subset has significantly greater capacity to proliferate, produce IFN-γ, and lyse target cells than does the CD94low subset. The CD94high subset has exclusive expression of NKG2A/C/E, higher expression of CD117 and CD69, and lower expression of Ly49D (activating) and Ly49G2 (inhibitory). In vivo, purified mouse CD94low NK cells become CD94high NK cells, but not vice versa. Collectively, our data suggest that CD94 is an Ag that can be used to identify functionally distinct NK cell subsets in mice and could also be relevant to late-stage mouse NK cell development.
The NK cells are large granular lymphocytes that constitute one critical component of the innate immune system, and thus part of the body’s essential first line of defense against infectious pathogens and malignant transformation. At least two major functions of mature NK cells, cytotoxicity and cytokine (e.g., IFN-γ and TNF-α) production, have been well characterized and contribute toward this first line of defense (1–3). In particular, NK cells produce large amounts of IFN-γ, a cytokine essential for the clearance of infectious pathogens as well as for effective tumor surveillance (4–6).
NK cells are not a homogeneous population. In humans, CD56dim NK cells exhibit higher natural and Ab-dependent cell-mediated cytotoxicity and less monokine-stimulated cytokine production, whereas CD56bright NK cells produce abundant cytokines but are distinctly less cytotoxic (7). Mouse NK cells do not express CD56; thus, there are no mouse NK subsets comparable with those defined in humans by density of CD56 surface expression. Therefore, our initial goal of this study was to identify a surface marker that could separate mouse NK cells into two clearly distinct and discontinuous populations and then analyze the functional and potential developmental relationship of the two subsets. Among surface markers that can be found on both human and mouse NK cells, CD94 is of particular interest, given that we have previously demonstrated that CD94 is useful to distinguish functionally discrete developmental stages of human NK cells (8), but the relevance of its heterogeneous expression in mouse NK cells is not completely defined.
CD94 is a type II integral membrane protein which is related to the C-type lectin superfamily (9). It covalently associates with five different members of the NKG2 family (NKG2A, B, C, E, and H) but not with NKG2D (10). The natural ligand for these CD94/NKG2 heterodimers is the nonclassical MHC class I molecule HLA-E in humans, and its mouse functional homolog Qa-1b (10). In this study, we show that CD94 expression can segregate mouse NK cells into two approximately even subsets with functional and phenotypical distinction in varied organs and tissues. CD94high mouse NK cells have higher proliferative and cytotoxic ability and produce more monokine-stimulated IFN-γ than CD94low NK cells. In vivo, purified mouse CD94low NK cells become CD94high NK cells, but not vice versa.
The following anti-mouse mAbs were purchased from BD Pharmingen: CD3 (145-2C11); NK1.1 (PK136); IFN-γ (XMG1.2); CD11b (M1/70); NKG2A/C/E (20d5); Ly49D (4E5); Ly49G2 (4D11); CD117 (2B8); CD2 (LAF-2); CD16/32 (2.4G2); CD27 (LG.3A10); CD43 (1B11); CD44 (IM7); B220 (RA3-6B2); CD62L (MEL-14); CD69 (H1.2F3); CD122 (TM-β1); CD49b (DX5); 2B4 (2B4); CD127 (SB/199); KLRG1 (2F1); Ly49I (YL1-90). The anti-mouse mAb CD94 (18d3), NKp46 (29A1.4), NKG2D (CX5), CCR7 (4B12), and Ly49C/I/F/H (14B11) were purchased from eBioscience.
Single-cell suspensions were prepared using mononuclear cells from mouse bone marrow, liver, lymph nodes, lung, and spleen before Ab staining. To extract mononuclear cells from lung and liver, the sliced lung and liver tissues were digested with collagenase IV and DNase I (Sigma-Aldrich). The isolated mononuclear cells were stained with the Abs listed above, followed by analysis with a FACSCalibur or LSRII (BD Bioscience) to detect surface expression for each Ag. NK cells were gated by NK1.1+CD3−.
Cytotoxic activity for mouse NK cells was assessed against Yac-1 using a standard 51Cr release assay. Effector cells were CD94high or CD94low NK cells purified from mouse spleen or pooled peripheral blood. The splenic CD94 NK cell subsets were isolated by cell sorting (>95% purity) after being enriched by a mouse NK cell isolation kit (Miltenyi Biotec), and the blood NK cell subsets were directly sorted from PBMC. Effector cells of splenic CD94 NK cell subsets were serially diluted for multiple E:T ratios, whereas one E:T ratio (10:1) was used for blood CD94 NK cell subsets due to the limited numbers of cells that were recovered. A constant number of 51Cr-labeled target cells (5 × 103/well) was used. The experiments were performed in triplicate wells for each E:T ratio in 96-well V-bottom plates, which were briefly centrifuged before a 4-h incubation. After the incubation, cytotoxicity was assessed by 51Cr release, which was measured in supernatants using a gamma counter. The standard formula for calculation of percent target cell lysis was used.
Cell-free supernatants from purified mouse splenic or blood CD94high and CD94low NK cells were analyzed by ELISA after in vitro stimulation with monokines, as previously described (11).
Purified CD94high or CD94low NK cells (1 × 105 cells for splenic subsets and 1 × 104 cells for blood subsets) were plated in triplicate wells in 96-well plates in 200 μl of RPMI 1640 with 10% FBS and 2-ME (50 μM) in the presence of high-dose human IL-2 (1000 U/ml) for assessment of in vitro cell growth. Viable cells were enumerated 2–10 days after plating using the trypan blue (Invitrogen) exclusion assay and a standard hemocytometer.
For 7 consecutive days before sacrifice, wild-type (WT)3 C57BL/6 mice (8–11 wk old) were given drinking water containing BrdU at a concentration of 0.8 mg/ml (changed daily). To determine BrdU incorporation of CD94high and CD94low NK cell subsets, after sacrifice of the mice, total spleen cells were isolated and stained for FACS analysis with NK1.1-allophycocyanin, CD94-PE, CD3-PerCP, and BrdU-FITC according to the manufacturer’s instructions (BD Biosciences).
C57BL/6 mice were female littermates (6–11 wk old) and were obtained from The Jackson Laboratory. All animal work was approved by The Ohio State University Animal Care and Use Committee, and mice were treated in accordance with the institutional guidelines for animal care.
Populations of CD94lowCD11bhigh, CD94highCD11bhigh, CD94low CD11blow/−, and CD94highCD11blow/− mouse splenic NK cells were sorted from C57BL/6 CD45.1 congenic mice by a FACSAria II cytometer (BD Biosciences) to >98% purity and transferred to sublethally irradiated (6.5 Gy from a gamma source) WT CD45.2 mice through tail vein injections (3.0–7.5 × 105 donor cells/mouse). Flow analysis on total blood or splenic lymphocyte cells was performed using a FACSCalibur cytometer (BD Biosciences) to determine the change in CD94 and CD11b surface expression within the above four NK populations of NK cells at time points of 1 or 2 wk after the injection.
C57BL/6 mice were injected i.v. with 1 μg of murine IL-12 (mIL-12; R&D Systems) and 0.5 μg of mIL-18 (R & D Systems) per mouse. Spleens were harvested from the mice 20 h after the injection. Splenocytes were immediately processed and cultured ex vivo for 4 h with brefeldin A (BD Pharmingen), before harvest. Surface staining was performed with anti-NK1.1-allophycocyanin, anti-CD3-PerCP, and anti-CD94-FITC mAbs (BD Pharmingen), and the cells were fixed and permeabilized using Cytofix/Cytoperm reagent (BD Biosciences). Cells then underwent intracellular staining with an anti-mouse IFN-γ-PE mAb or isotype control-PE mAb (BD Pharmingen). Cells were assessed on a FACSCalibur cytometer (BD Biosciences), and analyses were performed using the CellQuest (BD Biosciences) software program to detect IFN-γ production.
Data were compared using Student’s two-tailed t test. A p value of <0.05 was considered statistically significant. Unless otherwise specified, the error bars represent the SEM in all figures.
We first tested whether mouse NK1.1+CD3− NK cells can be distinguished phenotypically by CD94 surface Ag expression. Lymphocytes collected from blood, bone marrow, liver, lymph nodes, lung, and spleen were stained with CD3, NK1.1, CD11b, and CD94. Flow analysis indicated that after gating on the NK1.1+CD3− population, NK cells can be clearly separated into CD94high and CD94low subsets in all of the above tested organs or tissues (Fig. 1). The distribution pattern of these two subsets is similar, although not identical, among these different organs or tissues, with the CD94highNK1.1+CD3− NK cell subset ranging from 38% in bone marrow to 58% in lymph nodes. This dichotomous pattern of CD94 surface expression can be observed both in the mature CD11bhigh NK cells and in the less mature CD11blow/− NK cells (Fig. 1 and Ref. 12).
NK cell receptors are tightly linked to NK cell development and function (8, 12, 13). We therefore examined the expression patterns of a number of activating and inhibitory receptors in mouse CD94high and CD94low NK cell subsets by flow analysis. We found that nearly all (99.8%) CD94high NK cells from spleen co-express NKG2A/C/E (Fig. 2), consistent with the previous finding that CD94 associates with NKG2A/C/E (10). Surface expression of NKG2A/C/E is nearly absent from NK cells expressing low levels of CD94 (Fig. 2). These two subsets are also divergent in expression of Ly49 because CD94high NK cells have a lower expression of both the inhibitory molecule Ly49G2 (CD94high = 43.4% vs CD94low = 61%; p < 0.05; n = 5) and the activating molecule Ly49D (CD94high = 24.4% vs CD94low = 62.1%; p < 0.01; n = 5); however, there is little if any difference in expression of other tested Ly49 markers such as Ly49C/I/F/H. These data suggest that murine CD94high and CD94low NK cell subsets may have balanced expression of activating and inhibitory Ly49 receptors. The CD94high NK cell subset has higher expression of CD117 and CD69 compared with the CD94low subset (p < 0.05, n = 5), without substantial difference in expression of other surface markers including CD11b and NKp46, which are two important Ags expressed on mature NK cells. Therefore, it appears that the CD94high and CD94low NK cell subsets are likely close to each other in the developmental process. Likewise, when CD94high and CD94low NK cell subsets were isolated from bone marrow, surface expression patterns of all tested receptors were similar to those seen in the corresponding populations of splenic NK cells (supplemental Fig. 1).4
To measure the proliferative capacity of CD94high and CD94low NK cell subsets in vitro, sorted splenic cells belonging to each of these populations were cultured in the presence of IL-2. Fig. 3A demonstrates that CD94high NK cells showed significantly higher proliferation compared with CD94low NK cells at day 4 (p < 0.01, n = 3), and also at day 6 (p < 0.05, n = 3). The higher proliferative capacity of the CD94high NK cell subset compared with the CD94low NK cell subset was also seen in NK cells from blood (Fig. 2B). To compare their proliferative potential in vivo, WT mice received BrdU-containing drinking water for 7 days, and splenic NK cells were analyzed for BrdU incorporation by FACS. Fig. 3C shows that CD94high NK cells had a modest but significantly higher rate of BrdU incorporation compared with CD94low NK cells (p < 0.01; n = 11). However, the difference of in vivo proliferation rates of these two subsets within bone marrow was not statistically significant (data not shown).
We next assessed IFN-γ production in CD94high and CD94low subsets of mouse NK cells. When an equal number of each of the two FACS-purified splenic NK cell subsets (≥98% purity) were co-stimulated by IL-12 and IL-18 in vitro, CD94high cells secreted larger quantities of IFN-γ than CD94low cells (Fig. 4A). We repeated this experiment using FACS-purified blood CD94high and CD94low NK cell subsets, and similar data were obtained (Fig. 4B). We next compared IFN-γ production in vivo by injecting IL-12 and IL-18 i.v. into mice, followed 20 h later by sacrifice and quantification of IFN-γ production using intracellular staining. Both CD94high and CD94low NK cell populations were discernable in bone marrow and spleen as determined by the intensity of CD94 surface staining on CD3−NK1.1+ NK cells (Fig. 4C, left). The ability of the two populations to produce IFN-γ correlated with the extent of CD94 surface expression in both spleen and bone marrow, i.e., CD94high > CD94low, whether judged by the percentage of IFN-γ+ cells, or by the mean fluorescence intensity of total cells from each subset (Fig. 4C; p < 0.01; n = 3).
To compare the cytotoxic capacities of the mouse CD94high and CD94low NK cell subsets, sorted CD94high and CD94low splenic NK cells were separately incubated with MHC class I-mismatched Yac-1 target cells. The 51Cr release assay indicated that splenic CD94high NK cells have significantly higher natural cytotoxicity than CD94low splenic NK cells (p < 0.05, n = 3; Fig. 5A). Similar results were obtained for the CD94high and CD94low NK cell subsets purified from blood (Fig. 5B). To exclude the possibility that the higher cytotoxicity observed in CD94high cells may be caused by stimulation of CD94high cells by the CD94 Ab used for cell sorting, the CD94 Ab and its isotype control IgG Ab were separately incubated with purified total NK1.1+CD3− splenic NK cells for 6 h, a time period equivalent to that used for cell preparation and cell sorting. After this incubation, cells were harvested, washed, and used as effectors in a 51Cr release assay which indicated that cytotoxicity after incubation with the CD94 Ab was nearly identical with that observed with the IgG control Ab (Fig. 5C). Therefore, the observed difference in cytotoxicity of CD94high and CD94low NK cell subsets is indeed attributed to an intrinsic difference between the two subsets, rather than potential activation induced by the CD94 Ab binding during sample preparation.
Previous work from Hayakawa and Smyth (14) shows that the combination of CD11b (Mac-1) and CD27 is able to define three phenotypically and functionally distinct NK subsets: CD11blow CD27high, CD11bhighCD27high, and CD11bhighCD27low. Supplemental Fig. 2 shows that CD11blowCD27high, CD11bhighCD27high, and CD11bhighCD27low have relatively high, intermediate, and low levels of CD117 expression, respectively, consistent with the previous report which also demonstrates that CD11blowCD27high and CD11bhighCD27high may represent an early and an intermediate stage of mouse NK development, respectively, whereas CD11bhighCD27low may represent a relatively late stage (14). We noticed that these NK subsets have an approximately equal percentage of CD94high and CD94low cells with slightly higher percentage of CD94high cells in the CD11blowCD27high subset (supplemental Fig. 2). These data suggest that composition of the NK subsets characterized by differential expression of CD94 and those identified using CD27 do not completely overlap. To confirm this, we made use of the combined expression of CD94 and CD27 surface markers to sort out four populations (CD94highCD27high, CD94highCD27low, CD94lowCD27high, CD94lowCD27low) of NK1.1+CD3− mouse NK cells for a cytotoxicity assay, because CD27high mature NK cells have been shown to be more cytotoxic than CD27low cells (14). We found that double-high-positive cells (CD94highCD27high) had the highest cytotoxic activity, whereas single-high-positive cells (CD94highCD27low and CD94lowCD27high) had moderate cytotoxic activity, and double low positive cells (CD94lowCD27low) had the lowest cytotoxic activity (Fig. 5D). These data not only support our phenotypic and functional characterization of mouse NK subsets based on CD94 expression but also are consistent with the previously described properties of mouse CD27 NK subsets (14).
To determine the potential developmental fate of CD94low and CD94high NK cells in vivo, we purified these two subsets from C57BL/6 CD45.1 mice by sorting and then adoptively transferring them into congenic CD45.2 mice. Because CD94 is expressed on both CD11bhigh and CD11blow/− populations (Fig. 1) and because CD11b has also been reported as a marker of mature mouse NK cells (12), CD11b expression was also considered in these adoptive transfer experiments. The following four splenic NK cell subsets from CD45.1 C57BL/6 mice were sorted from enriched total NK cells and adoptively transferred into recipient mice (Fig. 6, top): 1) CD94lowCD11bhigh; 2) CD94highCD11bhigh; 3) CD94low CD11blow/−; and 4) CD94highCD11blow/− (Fig. 6, lower left panels). After i.v. injection of cells from each of the four sorted populations into individual CD45.2 congenic sex-matched littermate mice, we found that by 1–2 wk after adoptive transfer, in both spleen and blood, CD94low NK cells (subsets 1 and 3 in Fig. 6 and supplemental Fig. 3) were able to become CD94high NK cells, but not vice versa (subsets 2 and 4). Similarly, by 1–2 wk, transplanted CD11blow/− NK (subsets 3 and 4 in Fig. 6 and supplemental Fig. 3) had become CD11bhigh but not vice versa (subsets 1 and 2), an in vivo finding supportive of the concept that CD11b is a marker for murine NK maturation (12).
On the basis of human studies (13), we hypothesized that the acquisition of CD94 might be associated with distinct functional attributes in mouse NK cells. In this study, we therefore examined CD94 surface expression in mouse NK cells and found that surface density of CD94 divides mouse NK cells into distinct CD94high and CD94low subsets in a variety of organs and tissues. A majority of the surface markers that were examined are present at similar density in both of these mouse NK cell subsets. The exceptions noted were that CD94high NK cells are exclusively NKG2A/C/E+, express lower Ly49D (activating) and Ly49G2 (inhibitory) Ly49 molecules, and higher CD117 and CD69 than do CD94low NK cells. CD94high NK cells are more functionally activated, having higher monokine-stimulated IFN-γ production and possessing higher cytotoxicity and proliferative capacity than CD94low NK cells. In both blood and spleen, CD94high NK cells remain CD94high after adoptive transfer into a WT recipient mouse, whereas the CD94low subset becomes CD94high in vivo.
NK receptors are tightly linked to NK development and function (8, 12, 13). The similar expression levels of most surface markers including CD11b and NKp46 in CD94high and CD94low NK cells suggest that these two subsets might be very close in the developmental process. However, the differential expression of several other surface markers (e.g., CD117 and CD69) noted in this study also suggests that the two subsets are not identical and may have some functional distinctions. CD117 is the receptor for the cytokine stem cell factor, also known as steel factor or c-kit ligand, which is expressed on the surface of hemopoietic stem cells as well as other cell types (15). CD117 is usually expressed on cells with a high proliferative potential. We most recently found that c-kit ligand and IL-2/15 combine to promote NK cell proliferation through MAPK pathways (16). In this study, we discovered higher expression of CD117 in the murine CD94high NK cell subset than in the CD94low subset, which is congruent with the high proliferative capacity of the former subset. CD69 belongs to the family of c-type lectin receptors the genes of which are clustered in the NK gene complex (17, 18). Its cross-linking induces NK cell proliferation, cytotoxic activity, and cytokine production (19–21). The relatively higher CD69 surface expression on the murine CD94high NK cell subset is associated with higher functional activities in terms of IFN-γ production, cytotoxicity, and cell proliferation.
The lack of CD56bright and CD56dim NK cell subsets in mouse has resulted in some elegant studies to functionally and phenotypically separate mouse NK cells into two subsets. Smyth and colleagues (14) have carefully studied CD27 surface expression in mice and applied it to dissect mouse NK cell subsets. They found that the immature CD11blow NK cells have only high CD27 expression (CD27high), whereas the mature CD11bhigh NK cells have two subsets, CD27high and CD27low (14, 22). The two subsets of mature mouse NK cells (CD11bhighCD27high and CD11bhigh CD27low) are described as being functionally distinct, with the CD11bhighCD27high NK cells producing more IFN-γ and possessing higher cytotoxicity than CD11bhighCD27low NK cells. The CD11bhighCD27high NK cell subset proliferates faster and represents an earlier stage than the CD11bhighCD27low subset.
The composite of the mouse NK cell subsets identified by using CD94 surface expression is different from those previously identified by using CD11b and CD27 surface expression (14). Indeed, the three distinct subsets separated by CD11b and CD27 surface expression (CD11blowCD27high, CD11bhighCD27high, and CD11bhigh CD27low) each contains approximately equivalent proportions of CD94high and CD94low NK cells (supplemental Fig. 2). The previous study shows CD27 expression to be a continuous gradient on mouse NK cells (14), and we confirm this in our study. We go on to show that the intensity of CD94 surface expression separates mouse NK cells into two distinct and relatively discontinuous populations in all tested organs and tissues.
In addition to providing an alternative phenotypic and functional definition of mouse NK subsets based on the density of CD94 surface expression, our in vivo study showed that the density of CD94 expression is also likely associated with distinct NK cell developmental stages, although this will require further study due to the complexity of NK cell development. Our adoptive transfer experiment, performed in WT mice and using a CD45 (Ly5) congenic marker, indicated that mature mouse NK cells remain CD94highCD11bhigh, whereas NK cells with the low expression of either CD94 or CD11b enhance their expression in vivo, which is in support of the model proposed in a previous characterization of mouse NK cell development (12). The in vivo data presented in this study suggest that a mouse NK precursor cell may first acquire low-density expression of CD94 and then increase its expression while acquiring surface expression of NKG2A/C/E+ and undergoing further functional maturation. We could not find any expression of NKG2A/C/E in the CD94low NK cell subset in spleen, bone marrow, or blood, suggesting that a high level of CD94 surface expression may first be required for its covalent association with NKG2A/C/E.
Although CD94/NKG2A is an inhibitory molecule, its engagement with MHC class I peptides (10) may induce functional maturation, just as engagement of the Ly49-inhibitory molecules with MHC class I peptides is thought to result in NK cell licensing (23). Therefore, it is conceivable that CD94/NKG2A receptors are also involved in this process. This notion in mice has been previously discussed (24) and is supported by human data showing that a subpopulation of blood NK cells lacking both NKG2A and killer Ig-related receptor is developmentally immature, as evidenced by poor cytotoxicity and IFN-γ production (25). The role of CD94/NKG2A engagement of MHC class I during NK development, as specifically relates to the process of NK licensing or education should likely be pursued in future studies.
In summary, in this study we demonstrate that the CD94 surface marker can be used to phenotypically and functionally define subsets of mouse NK cells. How CD94 acquisition governs NK cell development and effector functions requires further study. Because CD94high NK cells have a higher capacity to produce IFN-γ and to lyse target cells, they may offer new opportunities for manipulation in experimental mouse models of sepsis, infection, autoimmune diseases, or even immunosuppression. This could be even more promising if combined with the previously identified CD27 surface marker given that the mouse CD94highCD27high NK subset shows the greatest capacity for lysis and cytokine production among the subsets identified by these two markers.
We are grateful for support from the Flow Cytometry, Nucleic Acid, Microarray and Mouse Phenotyping Shared Resources within The Ohio State University Comprehensive Cancer Center.
1This work was supported by National Cancer Institute Grants CA95426 and CA68458 (to M.A.C).
3Abbreviations used in this paper: WT, wild type; mIL-12, murine IL-12.
4The online version of this article contains supplemental material.
The authors have no financial conflict of interest.