Development and function of murine B220⁺CD11c⁺NK1.1⁺ cells identify them as a subset of NK cells

To investigate the nature of B220⁺CD11c⁺ cells in lymphoid organs, we analyzed Siglec-H and NK1.1 expression within these cells. Siglec-H is a PDC marker that, in contrast to bone marrow stromal antigen 2 (120G8, 927, and PDCA1) and Ly49Q, maintains specificity after viral infection or stimulation with TLR ligands (8, 10–12). NK1.1 is expressed on conventional NK cells and a recently described hybrid cell type, termed IKDC, which appears to share functional features with NK cells, PDCs, and DCs (2, 3). B220⁺CD11c⁺ cells in the spleen, lymph nodes, and bone marrow included three distinct subsets: (a) a Siglec-H⁺NK1.1⁻ subset that corresponds to PDC; (b) a Siglec-H⁻NK1.1⁺ subset that fulfills the definition of IKDC (B220⁺CD11c⁺NK1.1⁺ cells); and (c) a Siglec-H⁻NK1.1⁻ subset (Fig. 1 A). This latter subset was highly represented in the spleen and lymph node, where it mostly included mature CD19⁺IgM⁺ B cells (not depicted). In bone marrow, this subset was less abundant and exhibited further heterogeneity, as ∼40–50% of the cells expressed CD19 and IgM (Fig. 1 B). To determine which cell subset is capable of producing IFN-α in response to TLR9 stimulation, we sorted the three cell subsets and measured IFN-α released in culture supernatant after stimulation with CpG oligodeoxyribonucleotide (ODN). Siglec-H⁺NK1.1⁻ cells were the only cells capable of producing copious amounts of IFN-α (up to 10 ng/ml from 10⁵ cells). In contrast, Siglec-H⁻NK1.1⁺ and Siglec-H⁻NK1.1⁻ cells did not produce detectable levels of IFN-α. Collectively, these results provide evidence of phenotypic and functional heterogeneity within B220⁺CD11c⁺ cells in lymphoid organs and define the Siglec-H⁺NK1.1⁻ cell subset as the professional IFN-α producer cell type.

To compare B220⁺CD11c⁺ NK1.1⁺ cells with conventional NK cells, we enriched DX5⁺ cells from spleens by magnetic positive selection and sorted B220⁺CD11c⁺ and B220⁻CD11c^low/− cells, excluding contaminating CD3⁺, CD19⁺, and Siglec-H⁺ cells (Fig. 2 A). By using CD11c and B220 antibodies conjugated to allophycocyanin and PerCP-Cy5.5 for sorting, we were able to further characterize the two cell populations using FITC- and RPE-conjugated antibodies. B220⁺CD11c⁺ cells were highly enriched in cells coexpressing the TNF receptor family member CD27, which is the receptor for CD70, and the integrin αM (CD11b) (Fig. 2 B). They also included high percentages of cells expressing the growth factor receptor c-kit (CD117), but not the IL-7Rα chain (CD127), which was recently reported as a marker for NK cells of thymic origin (14). In contrast, B220⁻CD11c^low/− cells contained a high percentage of CD11b⁺CD27⁻ NK cells, few CD117⁺ cells (5–8% in different experiments), and some CD127⁺ cells. B220⁺CD11c⁺ cells also exhibited a unique repertoire of activating and inhibitory receptors. As compared with B220⁻CD11c^low/−, they contained two- to threefold more cells expressing the activating receptor Ly49D, which recognizes both allogeneic and xenogenic MHC class I (15, 16), and expressed slightly higher levels of 2B4 (Fig. 2 C). B220⁺CD11c⁺ cells also expressed higher level of the lymphocyte homing molecule L-selectin (CD62L) (Fig. 2 D). Several other Ly49 family receptors were similarly expressed by the two subsets, with the exception of a slight but reproducible increase in the percentage of Ly49C/I⁺ cells within the B220⁻CD11c^low/− population (Fig. S1). The receptors NKG2A/C/E and the integrin CD49 (DX5) were also equally expressed (Fig. S1 and Fig. 2).

Comparison of the cytolytic capacity of B220⁺CD11c⁺ and B220⁻CD11c^low/− NK cells revealed that these NK cell subsets killed the classical target YAC-1 ex vivo equally well (Fig. 2 E). Reduction of YAC-1 killing by an anti-NKG2D antibody demonstrated that cytotoxicity was partly mediated by the activating receptor NKG2D in both subsets. We also compared the ability of the two cell subsets to secrete cytokines in response to different activating stimuli. Both subsets produced IFN-γ in response to PMA/ionomycin and IL-12/IL-18 (Fig. 2 F), although B220⁺CD11c⁺ cells produced two- to fourfold more IFN-γ than B220⁻CD11c^low/− cells. In addition, both cell subsets produced IFN-γ and TNF-α in response to CpG stimulation, although B220⁺CD11c⁺ cells were again much more efficient than B220⁻CD11c^low/− cells (Fig. 2, F and E, and Fig. S1). Collectively, these results demonstrate that B220⁺CD11c⁺ NK cells are enriched in CD27⁺CD11b⁺ NK cells that secrete cytokines more effectively than conventional B220⁻CD11c^low/− NK cells, although they are equivalent in lytic capacity.

It has been proposed that B220⁺CD11c⁺NK1.1⁺ cells are related to DCs because they express MHC class II (2, 3). To determine whether MHC class II is a unique feature of B220⁺CD11c⁺NK1.1⁺ cells, we assessed MHC class II expression on B220⁺CD11c⁺ and B220⁻CD11c^low/− NK cells sorted from the spleen. An antibody specific for I-A and I-E detected MHC class II on few B220⁺CD11c⁺ NK cells (between 1 and 5%) (Fig. 3 A and Fig. S2). Treatment of sorted cells with IFN-γ, a known inducer of MHC class II, did not result in a marked increase of MHC expression (Fig. 3 A). However, stimulation of sorted cells with CpG 1668 induced approximately a twofold increase in the percentage of MHC class II⁺ cells (Fig. 3 B). We conclude that very few spleen NK1.1⁺B220⁺CD11c⁺ cells express MHC class II, and this expression appears to be controlled by an IFN-γ–independent TLR-dependent mechanism. In peripheral lymph nodes many more NK1.1⁺ NK cells expressed cell surface MHC class II than in the spleen. However, MHC class II expression did not correlate with CD11c and B220 expression (Fig. 3 C).

To determine whether B220⁺CD11c⁺NK1.1⁺ cells follow DC or NK cell developmental pathways, we assessed which cytokines are required in vivo for B220⁺CD11c⁺NK1.1⁺ development. FLT3L is a cytokine essential for the development of classical DC subsets and PDC (17, 18), whereas NK cells are less FLT3L dependent (19). Analysis of heterogeneity of B220⁺CD11c⁺ cells in FLT3L-deficient mice showed that Siglec-H⁺NK1.1⁻ PDCs were selectively reduced in the spleen and lymph nodes of these mice (Fig. 4 A and Fig. S3). On the contrary, the presence of NK1.1⁺Siglec-H⁻ was minimally affected, suggesting that B220⁺CD11c⁺NK1.1⁺ cells do not follow a PDC/DC developmental pathway. Although in the bone marrow the relative percentage of Siglec-H⁺NK1.1⁻ was only partly reduced, and that of NK1.1⁺Siglec-H⁻ cells was conserved, all B220⁺CD11c⁺ cells were severely reduced (∼10–15-fold in different mice). Therefore, the total numbers of bone marrow Siglec-H⁺NK1.1⁻ and NK1.1⁺Siglec-H⁻ cells are also reduced (Fig. S3).

It has been shown that IL-2Rβ^−/− mice lack B220⁺CD11c⁺NK1.1⁺ cells (3). IL-2Rβ is the receptor for IL-15, a cytokine required for NK cell development (20). Thus, we examined B220⁺CD11c⁺ heterogeneity in IL-15^−/− mice and found that B220⁺CD11c⁺NK1.1⁺ cells were severely reduced, whereas PDCs and Siglec-H⁻NK1.1⁻ cells were not particularly affected (Fig. 4 B and Fig. S3). This result corroborated the hypothesis that the development of B220⁺CD11c⁺NK1.1⁺ cells depends on IL-15/IL-15R signaling. This conclusion was at odds with a previous report showing that B220⁺CD11c⁺NK1.1⁺ cells do not develop in mice lacking γc, which is required for IL-15R signaling. To address this discrepancy, we analyzed RAG2^−/− and RAG2^−/−γc^−/− mice. PDCs were present in both RAG2^−/− and RAG2^−/−γc^−/− mice (Fig. 5 and Fig. S3). Siglec-H⁻NK1.1⁻ cells were almost absent in the spleens and lymph nodes of RAG2^−/− and RAG2^−/−γc^−/− mice, corroborating the notion that this subset contains mainly B cells. In contrast, B220⁺CD11c⁺NK1.1⁺ cells were dramatically reduced in the lymphoid organs of RAG2^−/−γc^−/− but not RAG2^−/− mice (Fig. 5 and Fig. S3).

Notably, a few residual NK1.1⁺ cells were present in the spleens of RAG2^−/−γc^−/− mice. To define the nature of these cells, we enriched DX5⁺ cells from RAG2^−/−γc^−/− spleens by positive selection. Few live cells (0.5–1 × 10⁴/spleen) copurified with cellular debris (Fig. S4 A). Only 5–10% of these cells were CD3⁻CD19⁻NK1.1⁺ and coexpressed CD11c and B220, and to some extent also MHC class II ex vivo (Fig. S4 B). Collectively, our results indicate that most B220⁺CD11c⁺NK1.1⁺ cells require IL-15/IL-15R signaling for their development, supporting the notion that these cells belong to NK cell lineage. However, very few B220⁺CD11c⁺NK1.1⁺ cells do develop in the absence of γc. The number of cells recovered from RAG2^−/−γc^−/− spleens was ∼1% of the B220⁺CD11c⁺NK1.1⁺ cells purified from WT mice. Whether these cells are a novel cell type or a rare precursor cell and what the function of these cells is in vivo remains to be determined.

Dissection of the heterogeneous B220⁺CD11c⁺ cell population has allowed us to address important questions concerning the nature of B220⁺CD11c⁺NK1.1⁺ cells as well as possible functional overlaps with PDCs and DCs. We found that PDCs but not B220⁺CD11c⁺NK1.1⁺ cells produce IFN-α upon stimulation with CpG. We also demonstrated that FLT3L is required for normal PDC numbers in the bone marrow and peripheral lymphoid organs, but it is not essential for B220⁺CD11c⁺NK1.1⁺ cell development. Therefore, PDCs and B220⁺CD11c⁺NK1.1⁺ cells are distinct with respect to secretion of type I IFN and developmental requirements.

Are B220⁺CD11c⁺ NK cells any different from other NK cells? We found that most B220⁺CD11c⁺ are CD27⁺CD11b⁺. A previous study showed that CD27 is a marker for immature NK cells as well as a subset of mature NK cells that home to secondary lymphoid organs and have highly cytolytic and IFN-γ secretion capacity. In contrast, mature CD11b⁺CD27⁻ NK cells home to peripheral tissues and have limited effector functions (21). Our results demonstrate that CD27⁺CD11b⁺ NK cells are highly enriched within the B220⁺CD11c⁺ NK population. Moreover, their phenotypes and functions are reminiscent of those of human CD56^bright NK cells, which are preferentially found in lymphoid organs (22). Both human CD56^bright and murine CD27⁺CD11b⁺ NK cells express high levels of the lymph node homing receptor CD62L, the growth factor receptor c-kit, have unique MHC class I receptor repertoires, and secrete IFN-γ more effectively than other NK cells. Because the counterligand of CD27, CD70, is highly expressed by DCs in the T cell area of lymphoid organs (23), B220⁺CD11c⁺ NK cells may represent a subset of NK cells that preferentially interact with DCs.

B220⁺CD11c⁺NK1.1⁺ cells were proposed as a novel DC type partially on the basis of their phenotype, particularly the expression of CD11c, B220, and MHC class II. We found that MHC class II is not expressed on most B220⁺CD11c⁺ NK cells in the spleen, and it is not induced upon exposure to IFN-γ. However, we did find few MHC class II⁺ cells within the B220⁺CD11c⁺ NK subset, and the percentage of these MHC class II⁺ cells increased to two- to threefold upon stimulation with CpG. Although peripheral lymph nodes harbored many more MHC class II⁺ NK1.1⁺ cells, MHC class II expression did not correlate with that of CD11c and B220. Thus, MHC class II expression may reflect an activation stage of NK cells induced by the lymph node microenvironment. Interestingly, activated human NK cells have been shown to express MHC class II (24) and T cell co-stimulatory molecules such as OX40L (25, 26) and to function as nonprofessional antigen-presenting cells (24). The impact of activated MHC class II⁺ NK cells on immune responses in vivo remains to be determined. Because NK cells express the low affinity Fc receptor for Ig, CD16, it is tempting to speculate that MHC class II⁺ NK cells may capture antigen–antibody complexes through CD16 and present processed antigens onto MHC class II. We observed that MHC class II can be induced by CpG but not IFN-γ. Thus, MHC class II expression on some NK cells appears to be uniquely induced by TLR9. Interestingly, human NK cells also express TLR9 and are activated by CpG (27). Thus, NK cell–mediated antigen presentation may be particularly relevant in autoimmune diseases such systemic lupus erithematosus, where cellular DNA in complex with antinuclear antibodies can be captured through CD16 and directed to an endosomal compartment to trigger TLR9 signaling.

B220⁺CD11c⁺NK1.1⁺ cells were designated IKDC also because of their presence in mice that lack γc, which is essential for NK cell but not DC development. However, we have demonstrated that most B220⁺CD11c⁺NK1.1⁺ cells require IL-15 and γc signaling for development, as do conventional NK cells. Consistent with this result, a recent study demonstrated that IKDC development, like NK cell development, requires the Id-2 transcriptional inhibitor (7). Thus, although B220⁺CD11c⁺NK1.1⁺ cells may originate from a Lin⁻Sca-1⁺c-Kit^HiThy1.1⁻L-selectin⁺ lymphoid progenitor that may be distinct from the common NK cell progenitor (7), most B220⁺CD11c⁺NK1.1⁺ cells appear to belong to the NK cell lineage and have no obvious developmental overlap with DCs.

C57BL/6 mice were purchased from The Jackson Laboratory. Mice lacking FLT3L, IL-15, or RAG-2 were purchased from Taconic Farms. Mice doubly deficient in RAG-2 and common cytokine receptor γc mice were provided by W.M. Yokoyama and R.D. Schreiber (Washington University, St. Louis, MO). All animal studies were approved by the Washington University animal studies committee. Antibodies against CD11b, CD11c, CD19, NK1.1, B220, CD3, Ly49-A, Ly49-D, Ly49-C/I, CD27, 2B4, CD62L, DX5, NKG2A/C/E, and I-A/I-E were purchased from BD Biosciences. Antibodies against Ly49C/I/F/H and CD127 were purchased from eBioscience. The anti–Siglec-H antibody was described previously (13). Sorting was performed on a MoFlo cell sorter (DakoCytomation) or a FACSVantage (BD Biosciences). Flow cytometry analysis was performed on a FACSCalibur or FACScan (BD Biosciences). Data were analyzed using the CELLQuest Pro software (BD Biosciences).

Mouse spleen and lymph nodes were mechanically disrupted and digested with collagenase D (Sigma-Aldrich). Splenocytes and bone marrow were treated with red blood cell lysing buffer (Sigma-Aldrich). Cell suspensions were filtered and treated with FcR blocking antibody before staining. To assay for type I IFN production, sorted cell populations were cultured for 18 h at 10⁶ cells/ml and stimulated with CpG ODN2216 at 10 μg/ml. Cell culture supernatants were assayed for IFN-α production by ELISA (PBL Interferonsource). To FACS sort NK cell subsets, DX5⁺ cells were preenriched by magnetic positive selection with DX5 microbeads (Miltenyi Biotec). Cytotoxicity was measured by standard chromium release assay. To assay for IFN-γ and TNF-α production, 2 × 10⁴ cells were stimulated overnight with IL-2, IL-12 (1 ng/ml; PeproTech), IL-18 (10 ng/ml; PeproTech), PMA (10⁻⁷ M; Sigma-Aldrich), ionomycin (0.5 μg/ml; Sigma-Aldrich), and CpG ODN1826 and CpG ODN1668 (12 μg/ml; QIAGEN) in 100 μl of complete medium. Supernatants were collected and cytokines measured by BD Cytometric Bead Arrays (mouse inflammation kit and mouse Th1/Th2 kit).

Fig. S1 shows expression of NK cell receptors and CpG-induced IFN-γ production in B220⁺CD11c⁺ and B220⁻CD11c^low/− NK cells. Fig. S2 shows expression of MHC class II and CD122 on B220⁺CD11c⁺ and B220⁻CD11c^low/− NK cells. Fig. S3 shows absolute numbers of B220⁺CD11c⁺ cell subsets in the spleen and bone marrow of WT, FLT3-L^−/−, IL-15^−/−, RAG2^−/−, and RAG2^−/−γc^−/− mice. Fig. S4 shows characterization of DX5⁺ cells from the spleen of RAG2^−/−γc^−/− mice.

We would like to thank Suzanne Schloemann in the Flow Cytometry Core of the Department of Pathology and Immunology; William Eades, Jacqueline Hughes, and Chris Holley in the Siteman Cancer Center High Speed Sorter Core Facility for flow cytometry, excellent cell sorting, and advice; Susan Gilfillan for reading the manuscript and advice; Wayne M. Yokoyama and Robert D. Schreiber for generous supply of reagents; and Jason Mills, Victoria Ramsey, and Anja Fuchs for advice and technical assistance in performing gene expression profiling.

A.L. Blasius was supported by the Cancer Research Institute training grant T32 CA009547. This work was supported by National Institutes of Health grants AI056139 and CA109673-02.

The authors have no conflicting financial interests.