NK Cell–Specific Gata3 Ablation Identifies the Maturation Program Required for Bone Marrow Exit and Control of Proliferation Free

Natural killer cells are an important component of the innate immune system. They have gained much attention because of their ability to kill transformed cells without prior sensitization to Ag (1, 2). NK cells can induce the activation of other immune cells by producing proinflammatory cytokines such as IFN-γ and TNF-α (3). In addition, NK cells mediate the direct killing of target cells by secreting cytotoxic granules that contain perforin and granzymes (4, 5). Because NK cells play a major role in the early control of viral infection and cancer, understanding the molecular mechanism by which NK cells fully develop and become functionally mature is of great importance. NK cells primarily undergo the development and maturation program in the bone marrow (BM). When they become mature, NK cells migrate through the bloodstream and seed into peripheral organs such as the spleen, liver, and lungs, where they can provide early protection against infection (6, 7). In the BM, NK precursor cells (NKPs) express the IL-2R β-chain (CD122), and this receptor is maintained in all of the stages of NK cell development. CD122 in combination with the common γ-chain (CD132) form the IL-2/15R, allowing NK cells to be responsive to IL-15. NK cell development depends on IL-15 and the downstream signaling molecules JAK1/3 and STAT5 (8–10). Immature NK (iNK) cells can be identified by the expression of NK1.1, and these cells further develop into mature NK (mNK) cells as they acquire CD11b. Notably, NK cells in the BM differentiate through four stages of maturation, which are associated with the acquisition of the surface markers CD27 and CD11b (11, 12). In the first stage, NK cells lack both CD27 and CD11b receptors. In the second and third stages, NK cells acquire CD27 and CD11b respectively. In the last stage of maturation, NK cells become negative for the expression of CD27. Therefore, NK cells develop as follows: CD27⁻CD11b⁻, CD27⁺CD11b⁻, CD27⁺CD11b⁺, and CD27⁻CD11b⁺.

Numerous elegant studies using immune cell-specific deficient mice or in vitro NK cell differentiation have defined factors responsible for the developmental process of NK cells (7, 13, 14). Examples of those transcription factors include Id2, E4bp4 (also known as Nfil3), T-bet, and Eomes. Id2 is an antagonist of the transcription factor TCF3, which promotes the development of B cells and myeloid cells (15). Eomes was known to bind the CD122 promoter region and significantly lower CD122 expression, and reduced responsiveness to IL-15 was observed in NK cells from Eomes-deficient mice (16). E4bp4 seems to function upstream of Eomes, so that E4bp4 deficiency caused severe defects in NK cell development (17, 18). A recent paper demonstrated that NK cells deficient in 3-phosphoinositide-dependent protein kinase-1 (PDPK1), a kinase downstream of PI3K and upstream of mammalian target of rapamycin, are unable to differentiate into NK cells in vivo (19). Rescue of the defect of PDPK1 by ectopic expression of E4bp4 or Eomes suggests that PDPK1 signaling is critical for NK cell development via induction of E4bp4 and Eomes. Lastly, NK cell development is altered in the absence of T-bet, resulting in increased numbers of NK cells in the BM and reduced peripheral NK cell numbers. T-bet and Eomes seem to play roles in regulating CD122 in NK cells (16, 20).

Several tissue-specific NK cell subsets have been identified, and their origins and differentiation programs are current topics of interest. Not only critical for T cell development, the thymus also generates a unique NK cell population called thymic NK cells. In contrast to the dispensable role of IL-7 for conventional NK (cNK) cells, thymic NK cells express IL-7Rα and are highly dependent on IL-7 (21). More recently, liver-resident NK cells defined as CD49b⁻ and CD49a⁺ were identified. They appeared phenotypically immature by expressing TRAIL (22), but showed relatively low proliferative capacity at the steady state (23). Adoptive transfer of hepatic mononuclear cells rescued the liver-resident NK cell population in lethally irradiated recipient mice, demonstrating that liver-specific hematopoietic progenitor cells are the origin of hepatic CD49b⁻CD49a⁺ NK cells (24). The liver-resident NK cells produce large amounts of TNF-α and IL-2 and less IFN-γ, suggesting their unique function (25). Notably, these cells have been demonstrated to exhibit memory potential in a model of hapten-specific contact hypersensitivity (24). Furthermore, a recent study of transcriptomic analysis indicated that cNK cells, thymic NK cells, and liver-resident NK cells are all generated as distinct lineages (25).

From a family of six transcription factors termed GATA binding protein, GATA3 contains two zinc-finger motifs that bind the consensus sequence [A/T]GATA[A/G] promoter elements (26). GATA3 is known to play crucial roles in a variety of immune cells in a developmental stage– and cell-lineage–specific fashion (27). For instance, GATA3 is important in the development and maintenance of T cells (27–29). Notably, GATA3 plays a critical role in the differentiation of Th2 cells and their ability to secrete cytokines such as IL-4, IL-5, and IL-13. GATA3 is also expressed in the NK cell lineage, and its role is restricted to the NK cell maturation, as GATA3 deficiency did not result in a major defect in NK cell frequencies but led to reduced T-bet expression accompanied by defective IFN-γ production and liver-specific homing of NK cells (30). Similar to T cells’ developmental requirement for GATA3, thymic NK cells expressing IL-7Rα require GATA3 (21) and are absent in Vav-Cre-Gata3^fl/fl mice (25). However, cNK cells and liver-resident NK cells are still present in the Vav-Cre-Gata3^fl/fl mice (25, 31) in which Gata3 deficiency is induced in all hematopoietic cells at the early stage of development, indicating the dispensable role of GATA3 in their early development. Therefore, the lineage- and stage-specific roles of GATA3 during NK cell development without the influence from other cell types are warranted. In this study, we generated a conditional Gata3-deficient mouse using NKp46-Cre mice to dissect the role of Gata3 at the late stage of NK cell development. We demonstrate that GATA3 is required for NK cell maturation beyond the acquisition of CD27. The incomplete maturation of NK cells due to GATA3 deficiency resulted in a defect in BM egress, and this was associated with decreased NK cell frequencies in peripheral organs. Surprisingly, despite the intrinsic defect in NK cell maturation and accompanying decrease of NK cell frequencies in the periphery, comparable protection between NK cell-specific Gata3 knockout mice and control littermates was observed during murine CMV (MCMV) infection. This protection was due to enhanced expansion of NK cells in the periphery of NKp46-Cre-Gata3^fl/fl mice and associated with enhanced upregulation of CD25 expression. Therefore, our result suggests that iNK cells found in peripheral organs can proliferate vigorously and spread and could provide a reservoir of efficient effector NK cells during a virus infection.

NKp46^iCre knockin (32), Gata3^flox mice (33, 34), and CD45.1⁺ Ly49H^−/− mice (35) were housed in a specific pathogen-free environment. The Ly49h-intron primers were D6Ott151-forward (F): 5′-GTGCTACCACTGAAAACCATTG-3′ and D6Ott151-reverse (R): 5′-CTGTCTCTTGAGTCACCTGCAC-3′ (36). The Gata3 floxed and Gata3 deleted alleles were genotyped using the following primers: F, 5′-GTCAGGGCACTAAGGGTTGTT-3′; floxed-R, 5′-TGGTAGAGTCCGCAGGCATTG-3′; and deleted-R, 5′-TATCAGCGGTTCATCTACAGC-3′ (33). To identify the specificity of Gata3 deletion by Cre recombinase, T cells, B cells, and NK cells from spleen and NKT cells from liver of NKp46-Cre-Gata3^fl/fl mice were isolated and sorted using MoFlo Astrios (Beckman Coulter). Purities of sorted cell populations were >96%. For PCR, genomic DNA was extracted from each purified cell population. Unless otherwise stated, all experiments were carried out using littermate controls from the NKp46-Cre^+/− × Gata3flox^+/+ × NKp46-Cre^−/− × Gata3flox^+/+ mating. All mice used for experiments were between 6 and 12 wk old. All procedures were approved by and conducted in accordance with the institution’s animal guidelines of the University of Ottawa.

Spleen and blood were harvested, and a single-cell suspension was obtained following RBC lysis and filtration through a 70-μm filter. Leukocytes were isolated from liver using Percoll gradient centrifugation. BM cells were extracted from tibia and femur.

The following mAbs were used: anti-GATA3 (TWAJ), anti-NK1.1 (PK136), anti-CD3 (17A2 and 145-2C11), anti-CD27 (LG.7F9), anti-CD11b (M1/70), anti-Ly49D (4E5), anti-Ly49G2 (4D11), anti-Ly49H (3D10), anti-NKG2D (CX5), anti-NKG2A–B6 (16a11), anti-CD62L (MEL-14), anti–killer cell lectin–like receptor subfamily G, member 1 (KLRG1; 2F1), anti-TCRβ (H57-597), anti-CXCR4 (2B11), anti–IFN-γ (XMG1.2), anti-CD45.2 (104), anti-CD25 (PC61.5), anti-CD49b (DX5), and anti-CD107a (1D4B) from eBioscience; anti-GATA3 (L50-823), anti-Ly49C/I (5E6), anti-CXCR3 (CXCR3-173), anti-STAT4 (8/Stat4), anti-IgG1 (A85-1), anti–p-STAT4 (38/p-Stat4), anti-CD49a (Ha31/8), anti-CD19 (1D3), anti–Ki-67 (B56), and anti-BrdU (3D4) from BD Biosciences; anti–IL-12Rβ2 (305719) and anti–IL-18Rα (112614) from R&D Systems; mouse IgG1, κ (MOPC-21) and rat IgG2b, κ (RTK4530) from BioLegend; and Fixable Far Red Live/Dead from Invitrogen. For intracellular IFN-γ measurements in vivo, leukocytes from BM, spleen, and liver were harvested postinfection at day 1.5 and incubated in RPMI 1640, 10% FBS, 1× penicillin/streptomycin, 1% L-glutamine, 10 mmol HEPES, 50 μmol 2-ME (RP-10) media containing 5 μg/ml brefeldin A for 4 h, followed by staining for intracellular IFN-γ using BD Cytofix/Cytoperm protocols (BD Biosciences). Intracellular staining of GATA3 and Ki-67 was carried out using a Foxp3 staining kit (eBioscience). Surface staining of CXCR4 was performed at 37°C for 10 min. Cells were acquired using FACS Cyan ADP and analyzed using Kaluza software v2 (Beckman Coulter) and FlowJo (Tree Star).

NK cell exit from BM and labeling of sinusoidal lymphocytes were achieved as described previously (37). Briefly, AMD3100 (Sigma-Aldrich) was given i.p. (150 μg/mouse) 1 h before sacrifice. To label sinusoidal lymphocytes, mice were injected i.v. with 1 μg anti-CD45.2 mAb conjugated with biotin (eBioscience). Mice were sacrificed 2 min after Ab injection. Then, cells were isolated from the blood and BM and stained with Streptavidin-PE before being analyzed using FACS Cyan ADP and Kaluza software v2 (Beckman Coulter).

Transwell assay was performed as previously described (37, 38). Briefly, 5-μm pore transwell inserts were used to analyze the migratory response of BM cells to CXCL12. BM cells were pretreated with or without 50 μmol AMD3100 for 90 min in RPMI 1640 medium supplemented with 1% FBS. Cells were then washed and counted. Using the same media, 50 ng/ml CXCL12 was prepared, and 500 μl/well was added into a Transwell-24 well plate (Corning). The inserts were placed inside the wells, and 5 × 10⁵ cells was loaded into each insert. After incubating the plate at 37°C for 3 h, the migrated cells were collected. The cells were counted, and the frequency of NK cells was determined by flow cytometry. Cell migration was calculated as percentage of input NK cells.

Spleens were harvested, and a single-cell suspension of splenic leukocytes was enriched by negative selection using the EasySep Mouse NK cell enrichment kit (Stemcell Technologies). Cells were then seeded in triplicates in RP-10 media on Ab-coated plates or cultured with different combinations of recombinant human (rh)IL-2 (1000 U/ml; National Institutes of Health/National Cancer Institute at Frederick Biological Resources Branch Preclinical Repository), recombinant murine (rm)IL-12 (50 ng/ml; eBioscience), and rmIL-18 (100 ng/ml; PeproTech). After 1 h of stimulation, brefeldin A was added to a final concentration of 5 μg/ml, and cells were incubated for 4 h followed by staining for intracellular IFN-γ using BD Cytofix/Cytoperm protocols (BD Biosciences).

Smith strain MCMV stocks were generated in our laboratory from infected salivary glands of BALB/c mice and viral titers determined by standard plaque assays (36). Mice were infected with 5000 PFUs MCMV i.p. To measure viral burdens in organs, spleens and livers from infected mice were homogenized by MagNA Lyser (Roche Applied Science), and the lysates were appropriately diluted and overlaid on mouse embryonic fibroblast cells in duplicates in 2% DMEM (2% FBS, 1× penicillin/streptomycin, 1% L-glutamine, and 10 mmol HEPES). Plaques were counted 3 d later and represented as log PFU per gram organ. Blood serum from naive or infected mice were appropriately diluted and analyzed for the production of mouse IFN-γ and IL-12 using a Cytometric Bead Array kit (BD Biosciences). Samples were prepared according to the manufacturer’s instructions, acquired on FACS Cyan ADP (Beckman Coulter), and analyzed using the BD FCAP Array Software (BD Biosciences).

NK cell cytotoxicity was measured by analyzing the antitumor activity of NK cells against YAC-1 target cells. Splenic leukocytes were isolated from Gata3^fl/fl or NKp46-Cre-Gata3^fl/fl mice that were either naive or infected with MCMV. To ensure equal numbers of NK cells were used, spleen populations were surface stained for NK and T cells and each sample adjusted to appropriate concentrations in sterile NK media (RPMI 1640, 10% FBS, 1× penicillin/streptomycin, 1% L-glutamine, 10 mmol HEPES, 50 μmol 2-ME, 1% nonessential amino acids, and 1 mmol sodium pyruvate). YAC-1 cells were cultured in sterile RP-10 media and labeled with 100 μCi [⁵¹Cr] for 1 h at 37°C, washed three times with PBS, and made to a concentration of 5 × 10⁴ cells/ml in NK medium. In V-bottom 96-well plates, 100 μl YAC-1 cells were added in triplicates and incubated for 4 h with 100 μl effector cells with E:T ratios ranging from 0.3:1 to 10:1. Supernatants were then used to quantify the amount of [⁵¹Cr] released due to specific lysis of YAC-1 targets by NK cells, and counted using a 2470 WIZARD² Automatic γ Counter (PerkinElmer). NK cytotoxicity was calculated according to the formula: % specific [⁵¹Cr] release = ([experimental release − minimum release]/[maximum release − minimum release]) × 100.

Spleens were harvested under sterile conditions, and a single-cell suspension of splenic leukocytes was generated after RBC lysis and filtration through 70-μm nylon mesh. Cells were enriched by negative selection using the EasySep Mouse NK cell enrichment kit (Stemcell Technologies) and labeled with Cell Proliferation Dye eFluor 450 (eBioscience). Cells were then cultured in 96-well plates in triplicates in RP-10 media with different concentrations of rhIL-2 or rmIL-15 (eBioscience). To measure BrdU incorporation, BrdU was added to the culture at a final concentration of 10 μmol, 24 h before harvesting the cells.

To evaluate CD25 induction on NK cells, 5E5 total splenocytes were stimulated in a 96-well plate with rmIL-12 (10 ng/ml; eBioscience) and/or rmIL-18 (10 ng/ml; PeproTech) for 24 h. To measure total STAT4 levels, cells were first stained using purified anti-STAT4 Ab followed by a secondary staining using anti-IgG1 conjugated to PE. To detect p-STAT4, cells were rested at 37°C for 4 h. Cells were then stimulated with rmIL-12 (10 ng/ml) and rmIL-18 (50 ng/ml) for 30 min before staining. To measure STAT4 and p-STAT4, cells were fixed with BioLegend Fix/Perm buffer and permeabilized with prechilled 100% methanol before intracellular staining.

Spleen cells from Gata3^fl/fl or NKp46-Cre-Gata3^fl/fl mice (CD45.2⁺) were isolated, and equal numbers of Ly49H⁺ NK cells were transferred into Ly49H^−/− recipients (CD45.1⁺). Ly49H^−/− recipients were either uninfected or given 3000 PFU MCMV i.p. 1 d following adoptive transfer. Leukocytes from spleen and liver of recipient mice were harvested on day 5 postinfection.

NK cells from Gata3^fl/fl or NKp46-Cre-Gata3^fl/fl mice were enriched by negative selection using the EasySep Mouse NK cell enrichment kit (Stemcell Technologies) and sorted using MoFlo Astrios (Beckman Coulter). Purities of sorted NK cells were >98%. Total RNA was extracted from the sorted NK cells using TRIzol reagent (Life Technologies) according to the manufacturer’s instructions. Using 35 ng total RNA, cDNA was reverse transcribed in a 20-μl reaction using the First Strand cDNA Synthesis kit (Thermo Scientific). For quantification of Gapdh and target genes by real-time PCR, cDNA was added to a 20-μl reaction of FastStart Universal SYBR Green Master (Rox) (Roche) and amplified using ViiA 7 Dx Real-Time PCR Instrument (Applied Biosystems). Expression of target genes was normalized to Gapdh levels. The primer sequences for Gapdh are: F, 5′-ACCACAGTCCATGCCATCAC-3′ and R, 5′-TCCACCACCCTGTTGCTGTA-3′; for Id2 are F, 5′-ATGAAAGCCTTCAGTCCGGTG-3′ and R, 5′-AGCAGACTCATCGGGTCGT-3′; for T-bet are F, 5′-CTAAGCAAGGACGGCGAATGT-3′ and R, 5′-GGCTGGGAACAGGATACTGG-3′; for E4bp4 are F, 5′-GAACTCTGCCTTAGCTGAGGT-3′ and R, 5′-ATTCCCGTTTTCTCCGACACG-3′; for Ets1 are F, 5′-TGGAATGTGCAGATGTCCCG-3′ and R, 5′-TGAGCATGCTCGATACCGTA-3′; for Irf2 are F, 5′-CGATTATTCAACTGACGGGCTTTC-3′ and R, 5′-GCATTCGCATCCGTTCCAC-3′; for Eomes are F, 5′-GATGTACGTTCACCCAGAAT-3′ and R, 5′-ATCGTAGTTGTCCCGGAAGC-3′; and for Tox are F, 5′-CTACTGCAGTAGTCGGTACTCC-3′ and R, 5′-CTGAAGCAGCTAGCAGCATAC-3′.

Significance of results was determined by two-tailed unpaired Student t tests (*p < 0.05, **p < 0.01, ***p < 0.001) and graphed using Prism Version 5 (GraphPad Software).

The complete ablation of Gata3 leads to embryonic lethality due to noradrenaline deficiency and improper kidney development (33, 39). Developmental stage–specific roles of GATA3 in diverse immune cells have been demonstrated (27). To investigate the role of GATA3 at the late stage of development of NK cells, we generated NKp46-Cre-Gata3^fl/fl mice using a mouse expressing the Cre recombinase under the NKp46 promoter. Compared to Vav-Cre-Gata3^fl/fl mice, NKp46-Cre expression is induced at a stage of NK cell development that follows the acquisition of NK1.1, therefore allowing us to study the role of GATA3 at a late stage of development. In a previous report, Gata3^fl/fl mice that carry loxP sites flanking exon 4 of the Gata3 locus have been successfully used to generate conditional Gata3-deficient mice (Fig. 1A) (33, 40). To determine whether Cre-mediated recombination occurred only in cNK cells in the spleens of NKp46-Cre-Gata3^fl/fl mice, we extracted genomic DNA from sorted T, B, NK, and NKT cells isolated from NKp46-Cre-Gata3^fl/fl mice and determined the Cre-mediated recombination by genomic DNA PCR. The Gata3-floxed allele was present among sorted T cells, B cells, and NKT cells but only weakly amplified from sorted NK cells (Fig. 1B). Contrarily, the Gata3-deleted allele was predominately detected from sorted NK cells but not in other populations. Intron 1 of Ly49h gene was used as control and showed similar genomic abundance between the different sorted populations. These results indicate that the genomic deletion of Gata3-exon 4 occurred predominantly in splenic NK cells of NKp46-Cre-Gata3^fl/fl mice.

We also evaluated GATA3 protein expression in NK cells from NKp46-Cre-Gata3^fl/fl mice and their control littermates that carry loxP-flanked Gata3 alleles but lack NKp46-Cre (Gata3^fl/fl mice). GATA3 is known to be expressed in mature T cells but not in B cells or myeloid cells (40). Accordingly, T cells expressed high levels of GATA3, and this expression was at a similar intensity as NK cells from control mice (Fig. 1C). Non-T/non-NK cells that are composed of B cells and myeloid cells did not express GATA3 as expected. Notably, NK cells from NKp46-Cre-Gata3^fl/fl showed reduced expression of GATA3 compared with NK cells from control mice. To determine at which stage of NK cell differentiation GATA3 expression is abrogated in NKp46-Cre-Gata3^fl/fl mice, we measured the expression of GATA3 in NKPs, iNK cells, and mNK cells in the BM. GATA3 was not expressed in NKPs, but was induced at the iNK stage and retained its expression in mNK cells (Fig. 1D). However, in NKp46-Cre-Gata3^fl/fl mice, GATA3 expression was reduced in both iNK and mNK cells, suggesting that NKp46-Cre–induced Gata3 downmodulation occurs as early as the iNK cell stage. Taken together, the NKp46-Cre-Gata3^fl/fl mouse is an adequate model to study the role of GATA3 in the development and maturation of NK cells.

Previously, GATA3 deficiency at the hematopoietic stem cell stage was shown to promote the homing of NK cells into the liver (30). To determine whether GATA3 influences the distribution of NK cells within different compartments, we examined the NK cell frequencies in the BM, spleen, and liver of NKp46-Cre-Gata3^fl/fl mice. Compared to their control littermates, NK cell frequencies and numbers in NKp46-Cre-Gata3^fl/fl mice were found higher in the BM but lower in the spleen and liver (Fig. 2A, 2B), suggesting a defect in the egress of NK cells from the BM to peripheral organs in the absence of GATA3.

To investigate the role of Gata3 in the NK cell maturation process, we assessed the maturation of NK cells in NKp46-Cre-Gata3^fl/fl mice. NK cells from NKp46-Cre-Gata3^fl/fl mice displayed an immature phenotype, and this was characterized by an increase in double-negative (DN) and CD27 single-positive (SP) populations and a decrease in double-positive (DP) and CD11b SP subsets. This was shown in all three organs tested (Fig. 2C, 2D). Therefore, GATA3 is important for the differentiation of NK cells from CD27 SP to DP.

The newly identified CD49b⁻CD49a⁺ liver-resident NK cells are known to be generated from hepatic hematopoietic progenitors (24). Because a defect in the homing of NK cells into the liver has been identified in chimeric mice using Gata3-deficient fetal liver cells (30), the dramatic reduction of NK cells in the livers shown in the previous report might not be due to the lack of hepatic progenitor cells. Alternatively, GATA3 is required for the development of liver-resident NK cells from hepatic progenitor cells. To test this possibility, both cNK cells and liver-resident NK cells, which express CD49b and CD49a, respectively, were analyzed in NKp46-Cre-Gata3^fl/fl mice. A reduction in the numbers of cNK cells was found in NKp46-Cre-Gata3^fl/fl mice, which can be explained by the inability of these cells to home into the liver (Fig. 3A, 3B). Interestingly, the numbers of liver-resident NK cells were also reduced in NKp46-Cre-Gata3^fl/fl mice, indicating that GATA3 is required for the development of liver-resident NK cells. The impaired ability of cNK cells to migrate from the BM to the liver in combination with the defective development of liver-resident NK cells in the absence of GATA3 is consistent with the dramatic reduction of NK1.1⁺CD3⁻ cell frequencies in the livers of NKp46-Cre-Gata3^fl/fl mice (Fig. 2A, 2B).

The critical roles of GATA3 in the development of diverse immune cells suggest that GATA3 might function indirectly by positively or negatively modulating other transcription factors. To gain insight into the mechanism underlying the GATA3-governed regulation of transcription factors, we measured the mRNA level of transcription factors known to drive NK cell development. Gata3-deficient NK cells showed reduced expression of Id2, T-bet, and E4bp4 (Supplemental Fig. 1A). Downmodulation of these transcription factors could be in line with developmental defects observed in NKp46-Cre-Gata3^fl/fl mice. Notably, defects in NK cell maturation in Id2-deficient or E4bp4-deficient NK cell progenitor cells are the most severe and result in a maturation blockade occurring before NK1.1 acquisition in the BM (17, 41, 42), different from the accumulation of NK1.1⁺ iNK cells in NKp46-Cre-Gata3^fl/fl mice, suggesting GATA3 involvement in multiple stages of NK cell development. The NK cell developmental defect in T-bet–deficient mice showed increased numbers of NK cells in the BM and reduced peripheral NK cell numbers, similar to NKp46-Cre-Gata3^fl/fl mice (43). Paradoxically, mRNA expression of Tox, which is required for NK cell maturation, was found abundant in NK cells from NKp46-Cre-Gata3^fl/fl mice (44). Taken together, this result suggests a complex interplay between GATA3 and these transcription factors during the development of NK cells.

To understand GATA3-mediated control in NK cell development, we decided to further characterize the expression of several surface receptors on NK cells from NKp46-Cre-Gata3^fl/fl mice and their control littermates. In NKp46-Cre-Gata3^fl/fl mice, the expression of inhibitory and activating receptors was slightly reduced in the spleen and liver (Fig. 4A, 4B). Surprisingly, NKp46-Cre-Gata3^fl/fl mice exerted dramatically perturbed expression of molecules involved in cell adhesion such as CD11b and CD62L on NK cells (Fig. 4A, 4C). Known to associate with CD18 to form an integrin (45), CD11b was poorly expressed on NK cells in all three compartments analyzed from NKp46-Cre-Gata3^fl/fl mice. The proportion of NK cells expressing l-selectin (CD62L) was also decreased in these mice. Interestingly, NK cells from NKp46-Cre-Gata3^fl/fl mice showed increased expression of the chemokine receptor CXCR3. Although the role of CXCR3 on NK cell migration at the steady state is unclear, it has been shown to be important for the recruitment of NK cells into tumors (46). Thus, the altered expression of molecules involved in cell adhesion may explain the defective ability of NK cells to exit the BM in NKp46-Cre-Gata3^fl/fl mice.

Because CD27⁺CD11b⁻ iNK cells are overrepresented in total NK cells, we decided to compare the phenotypes in subsets of distinct maturation stages based on the expression of CD27 and CD11b. The expression pattern of most surface receptors was similar in NK cell subsets of equivalent maturation stages when comparing NKp46-Cre-Gata3^fl/fl mice with their control littermates (Supplemental Fig. 1B, 1C), indicating that the differential expression can be explained by higher proportions of CD27⁺CD11b⁻ iNK cells in NKp46-Cre-Gata3^fl/fl mice. In addition, the expression of the inhibitory receptor KLRG1 was almost completely abrogated on NK cells from NKp46-Cre-Gata3^fl/fl mice (Fig. 4A, 4C). However, the expression of KLRG1 was drastically different in DP and CD11b SP subsets, suggesting that KLRG1 expression is directly GATA3 regulated.

Among various chemokine receptors, the critical role of CXCR4 on the retention of NK cells in the BM was suggested (37). The expression of CXCR4 is decreased in mature NK cells, making these cells less responsive to CXCL12, a CXCR4 ligand highly expressed in the BM environment (37, 47–49). Injection of mice with the CXCR4 antagonist AMD3100 resulted in the recruitment of iNK cells from the BM to the blood (37). Given that NK cells from NKp46-Cre-Gata3^fl/fl mice are immature and that these NK cells show a defect in egress from the BM to the periphery, we sought to determine whether the use of AMD3100 would allow NK cells to exit the BM in NKp46-Cre-Gata3^fl/fl mice. Mice were injected i.p. with AMD3100 for 1 h and then injected i.v. with anti-CD45.2 mAb 2 min before sacrifice to label sinusoidal lymphocytes. PBMCs and BM were analyzed. In control littermates, the proportion of NK cells in BM sinusoids was not altered following injection of AMD3100 (Fig. 5A, 5B). However, NK cell frequencies within BM sinusoids were increased in AMD3100-treated NKp46-Cre-Gata3^fl/fl mice. Similarly, the fold change of the absolute NK cell number in BM sinusoids upon AMD3100 treatment was heightened in NKp46-Cre-Gata3^fl/fl mice (Fig. 5C). This was associated with increased number of NK cells in the blood (Fig. 5D), even though the effect was not as dramatic as in BM sinusoids. As shown previously (37), the immature CD27 SP NK cells were selectively recruited to the BM sinusoids and the blood upon CXCR4 antagonist AMD3100 treatment (Fig. 5E), suggesting that antagonizing CXCR4 rescues the defective egress of immature Gata3-deficient NK cells from the BM.

The ability of BM NK cells to migrate in response to the CXCR4 ligand, CXCL12, was assessed in vitro using the transwell migration assay. Pretreatment with AMD3100 reduced the chemotaxis of NK cells from NKp46-Cre-Gata3^fl/fl mice but did not affect the migration of NK cells from control littermates (Fig. 5F), suggesting that chemotaxis to CXCL12 is responsible for the retention of iNK cells of NKp46-Cre-Gata3^fl/fl mice in the BM. Notably, NK cells in the BM of NKp46-Cre-Gata3^fl/fl mice showed higher expression of CXCR4 (Fig. 5G, 5H). Taken together, GATA3 is required for the CXCR4 downmodulation associated with mature NK cell phenotype, and CXCR4-mediated retention is a major mechanism for the abundant NK cell proportion in the BM of NKp46-Cre-Gata3^fl/fl mice.

It was previously shown that in the absence of GATA3, NK cells are defective in producing IFN-γ when stimulated through the surface receptor CD11b or using IL-12 and IL-18 (30). The defective IFN-γ production might be due to the reduced expression of surface CD11b or cytokine receptors. In addition, the ability of Gata3-deficient NK cells to produce IFN-γ has not been thoroughly shown via stimulation of NK-activating receptors. Thus, we tested whether cross-linking of the activating receptors NK1.1 and Ly49H would result in reduced production of IFN-γ in the absence of GATA3. The expression of surface NK1.1 was comparable between NKp46-Cre-Gata3^fl/fl and control mice. In vitro stimulation through NK1.1 resulted in similar frequencies of NK cells producing IFN-γ for both NKp46-Cre-Gata3^fl/fl and control mice (Fig. 6A, 6B), indicating that stimulation through an activating receptor can lead to normal cytokine production by NK cells in the absence of GATA3. When stimulated through Ly49H, slightly less NK cells were able to produce IFN-γ in the absence of GATA3, but this might be explained by the moderate reduction of Ly49H⁺ NK cell proportions in NKp46-Cre-Gata3^fl/fl mice (Fig. 4A, 4B). The role of GATA3 on IFN-γ production was also investigated upon cytokine stimulation. When NK cells were stimulated with different combinations of cytokines such as IL-2, IL-12, and IL-18, Gata3-deficient splenic NK cells exhibited a reduced ability to produce IFN-γ. NK cell degranulation evaluated by the LAMP-1 expression was largely comparable between NK cells from NKp46-Cre-Gata3^fl/fl and control mice. Taken together, GATA3 is largely dispensable for induction of effector functions in response to stimulation of activating receptors, but is required for efficient IFN-γ production in response to cytokines.

During MCMV infection, cytokine levels in the serum peak at day 1.5, whereas the proportion of proliferating NK cells reaches its peak at day 3 (50, 51). To determine whether the defect in cytokine production observed in vitro can also be seen in vivo, we measured the production of IFN-γ by NK cells from NKp46-Cre-Gata3^fl/fl mice upon challenge with MCMV. Following infection with MCMV, increased NK cell frequencies were found in the livers, presumably due to cell migration to the liver (52). Fewer NK cells in the spleens and livers of NKp46-Cre-Gata3^fl/fl mice produced IFN-γ in vivo at day 1.5 post–MCMV infection (Fig. 7A, 7B). IFN-γ production at day 1.5 of MCMV infection is largely mediated by IL-12 (53, 54). However, IFN-γ and IL-12 levels in the serum of NKp46-Cre-Gata3^fl/fl mice were similar to their control littermates (Fig. 7C). When we compared IFN-γ expression in subsets of distinct maturation stages, CD27 SP and DP NK cells were the most efficient at producing IFN-γ (Supplemental Fig. 2A, 2B).

We next investigated whether this defect in IFN-γ production in combination with the reduced NK cell frequencies in the periphery would render the NKp46-Cre-Gata3^fl/fl mice highly susceptible to viral infection. Interestingly, we observed only a slightly higher viral burden in the spleens and similar viral burden in the livers of NKp46-Cre-Gata3^fl/fl mice at day 3 following infection (Fig. 7D). In order to explain this unexpected resistance in NKp46-Cre-Gata3^fl/fl mice, we analyzed the NK cell populations in these mice. Surprisingly, despite their reduced numbers in the spleen and liver at the steady state, following MCMV infection, NK cells from NKp46-Cre-Gata3^fl/fl mice were able to expand to similar frequencies as NK cells from control littermates (Fig. 7E, 7F). In particular, the proportions of NK cells in the liver were greatly increased and associated with increased proliferation of liver NK cells at day 2.5 postinfection based on Ki-67 staining (Fig. 7G, 7H). CD27 SP and DP NK cells showed the highest levels of proliferation (Supplemental Fig. 2E, 2F). When we compared Ki-67 expression in subsets of distinct maturation stages, the expression was similar in NK cell subsets of equivalent maturation stages, indicating that the higher proportion of iNK cells contained in total NK cells of NKp46-Cre-Gata3^fl/fl mice explain the increased Ki-67 expression. To exclude the possibility that these expanded cells originate from residual GATA3⁺ NK cells in which the Cre recombination failed to occur, we measured GATA3 expression in the expanded NK cells from day 3 post–MCMV infection. The expanded NK cells in NKp46-Cre-Gata3^fl/fl mice did not express GATA3 (Fig. 7I). Not only were NK cells from NKp46-Cre-Gata3^fl/fl mice able to expand during MCMV infection, but they were also able to execute their cytotoxic functions, as shown by killing assay (Fig. 7J).

In order to be able to bind to IL-2 with high affinity and therefore to respond to low levels of physiological IL-2 concentrations in vivo, NK cells need to express IL-2Rα (CD25) to complete the heterotrimeric high-affinity IL-2R composed of IL-2Rα, β, and the common γ-chain (55). During MCMV infection, IL-12 and IL-18 can induce the upregulation of CD25 (56). Following MCMV infection, NK cells from NKp46-Cre-Gata3^fl/fl mice expressed higher levels of CD25 than NK cells from control mice (Fig. 8A, 8B, Supplemental Fig. 2C, 2D), which could explain the enhanced proliferation of these cells. To exclude the effect of differential inflammation during MCMV infection, we cultured naive splenic NK cells from NKp46-Cre-Gata3^fl/fl mice and control littermates in the presence of IL-12 and IL-18 to measure the upregulation of CD25 in vitro. NK cells from NKp46-Cre-Gata3^fl/fl mice were more efficient at upregulating CD25 as compared with NK cells from control littermates (Fig. 8C, 8D), and this was not due to increased levels of IL-12 in the serum (Fig. 7C). Interestingly, the enhanced CD25 expression was observed despite reduced IL-18Rα expression on GATA3-deficient NK cells (Fig. 8E, 8F). IL-12Rβ2 expression was similar between the NKp46-Cre-Gata3^fl/fl and control mice. CD27 SP and DP NK cells populations were the most efficient at upregulating CD25 upon stimulation with IL-12 and IL-18, respectively (Supplemental Fig. 3A, 3B). Similar to the Ki-67 expression, CD25 expression in all subsets in both NKp46-Cre-Gata3^fl/fl and control littermates was equivalent, indicating that the higher proportion of immature NK cells in NKp46-Cre-Gata3^fl/fl mice explain the increased CD25 expression. The reduced expression of IL-18Rα might explain the decreased CD25 expression upon IL-18 stimulation on NK cells of NKp46-Cre-Gata3^fl/fl mice (Supplemental Fig. 3C, 3D).

STAT4 is required for CD25 induction upon IL-12 stimulation during MCMV infection (56). Expression of total STAT4 was similar between the two groups (Fig. 8G), and stimulation with IL-12 and IL-18 led to similar phosphorylation levels of STAT4 in NK cells from NKp46-Cre-Gata3^fl/fl mice. Because NK cells from NKp46-Cre-Gata3^fl/fl mice expand robustly during MCMV infection in vivo, we next sought to determine whether naive splenic NK cells are more responsive than control NK cells when stimulated with IL-2 in vitro. NK cells from NKp46-Cre-Gata3^fl/fl showed enhanced proliferation upon stimulation with low doses of IL-2 (Fig. 8H, 8I, Supplemental Fig. 3E, 3F). It could be explained by increased upregulation of CD25 on NK cells from NKp46-Cre-Gata3^fl/fl mice compared with those from control littermates. Indeed, enriched NK cells from NKp46-Cre-Gata3^fl/fl mice displayed higher expression of CD25 when cultured for 5 d at low doses of IL-2 (Fig. 8H, 8I). Therefore, splenic NK cells from NKp46-Cre-Gata3^fl/fl mice exhibited intrinsically enhanced CD25 expression and proliferative ability. Similar hyperresponsiveness of NK cells from NKp46-Cre-Gata3^fl/fl mice was observed in response to IL-15 (Supplemental Fig. 3G, 3H).

During MCMV infection, the increased proliferation of NK cells from NKp46-Cre-Gata3^fl/fl mice may result from higher viral load and/or higher inflammation compared with control littermates. Therefore, to investigate whether the enhanced proliferative potential of NK cells from NKp46-Cre-Gata3^fl/fl mice is cell intrinsic, we adoptively transferred equal numbers of Ly49H⁺ NK cells from either control or NKp46-Cre-Gata3^fl/fl (CD45.2⁺) spleen into Ly49H-deficient (CD45.1⁺) recipient mice. The recipient mice were either infected with MCMV 1 d later or left uninfected before being sacrificed 5 d postinfection (Fig. 9A). In both spleen and liver, we observed greater expansion of donor NK cells when these cells originated from NKp46-Cre-Gata3^fl/fl mice (Fig. 9B, 9C), indicating that NK cells in NKp46-Cre-Gata3^fl/fl mice show an intrinsically enhanced proliferative capability.

In the study, we investigated the NKp46-Cre–driven Gata3-deficient mice. Compared to Vav-Cre-Gata3^fl/fl mice for which Gata3 deficiency is induced in all hematopoietic cells at the early stage of development (31), NKp46-Cre-Gata3^fl/fl mice provide a mouse model that allows us to study the roles of GATA3 at a late stage of NK development in a cell-intrinsic manner. We identified a critical role of GATA3 in the NK cell maturation stage associated with seeding of peripheral organs, highlighted by increased numbers of NK cell in the BM and decreased numbers in the peripheral organs of NKp46-Cre-Gata3^fl/fl mice. Interestingly, intact frequencies of NK1.1⁺ cNK cells were found in the spleens of Vav-Cre-Gata3^fl/fl mice, suggesting that Gata3 deficiency at the early stage of NK cell development might allow an alternative pathway of BM egress (31). Otherwise, the developmental defect in other immune cells in the absence of GATA3 might alleviate the defect in NK cell maturation.

Liver-resident NK cells are different from cNK but share many characteristics with ILC1s, as they express NK1.1, NKp46, and CD49a and require T-bet for their development (23–25, 57, 58). In contrast to its dispensable role in the development of cNK precursor cells, GATA3 is important in the development of multiple ILCs. GATA3 was initially identified to be essential in ILC2 maintenance in mice and humans (59–61). More recently, GATA3 was shown to be essential for the generation of the common progenitor cells for IL-7Rα–expressing ILCs. In the studies using Vav-Cre-Gata3^fl/fl mice or chimeric mice generated with Gata3-deficient fetal liver hematopoietic precursor cells, several ILC subsets including ILC2s and ILC3s were severely affected in the absence of GATA3 during their development (31, 62). Higher proportions of liver-resident NK cells were present in Vav-Cre-Gata3^fl/fl mice compared with controls; however, the absolute numbers of those cells were not provided (25). Therefore, our data showing reduced numbers of liver-resident NK cells in NKp46-Cre-Gata3^fl/fl mice are rather surprising. It is possible that deletion of other immune cells due to the early deletion of Gata3 might provide an advantageous environment for the generation of liver-resident NK cells. Similarly, frequencies of ILC1s were found decreased in the absence of GATA3 (58).

NK cells developing in the absence of GATA3 showed a defect in exiting the BM at the steady state. NK cells displayed an immature phenotype characterized by an accumulation of the CD27⁺CD11b⁻ subset in the BM, spleens, and livers, demonstrating that GATA3 is required for the NK cell maturation beyond the CD27 SP stage. More recently, Ikzf3 was reported to be required for the maturation in the spleen of CD27⁻CD11b⁺ NK cells, suggesting that IKZF3 regulates the maturation of NK cells subsequent to GATA3 (63). Interestingly, the iNK cells exhibited lower expression of adhesion molecules such as CD11b and CD62L. It is has been shown that NK cells that enter the BM sinusoids are mainly mature NK cells that express CD11b (37). During development, the BM egress of NK cells is mainly regulated by two independent events, CXCR4 desensitization and sphingosine-1-phosphate receptor 5 (S1P5) engagement (37). The chemokine receptor CXCR4 was shown to retain iNK cells in the BM, and CXCR4 expression is negatively correlated with NK cell maturation. In this study, we tested the ability of NK cells from NKp46-Cre-Gata3^fl/fl mice to exit the BM by using AMD3100, a CXCR4 antagonist. Treatment of NKp46-Cre-Gata3^fl/fl with AMD3100 resulted in the egress of the immature CD27 SP NK cells from the BM to the blood. Moreover, our transwell migration assay indicated that NK cells from NKp46-Cre-Gata3^fl/fl mice exhibit reduced chemotaxis in response to CXCL12 upon ADM3100 treatment. Therefore, sustained CXCR4 expression on NK cells in the BM of NKp46-Cre-Gata3^fl/fl mice blocks the migration of NK from the BM to the periphery. Antagonizing CXCR4 by AMD3100 could be used in clinical settings for patients who carry genetic defects that lead to low numbers of peripheral NK cells due to their incomplete maturation.

The migratory defects observed in NK cells from NKp46-Cre-Gata3^fl/fl mice are similar to those in S1P5-deficient, Prmd1-deficient, and T-bet–deficient mice (38, 64, 65), showing higher proportions of NK cells in the BM and reduced proportions of NK cells in the periphery. T-bet acts upstream of Blimp1 in mouse NK cells (65), and T-bet expression is regulated by GATA3 (30) (Supplemental Fig. 1). More importantly, S1P5 is a direct target of T-bet and its expression is significantly reduced on NK cells from T-bet–deficient mice (64). Thus, the reduced expression of T-bet observed in NKp46-Cre-Gata3^fl/fl mice might be responsible for the migratory defect of NK cells. Similarly, Foxo1, a negative regulator in the late stage of NK cell maturation, represses T-bet (66). Unfortunately, we could not evaluate the expression of S1P5 on NK cells due to the lack of available good Abs.

Our data indicated that NK cells from NKp46-Cre-Gata3^fl/fl mice exhibit defective IFN-γ production when stimulated by cytokines, consistent with a previous report (30). At least the defective IFN-γ production in response to IL-12 is not likely explained by the levels of cytokine receptor expression because the expression of the inducible IL-12Rβ2 chain is comparable (67). The defect seems to be narrowed to the response to cytokines, because stimulation through activating receptors such as NK1.1 or Ly49H showed comparable IFN-γ production with control littermates. In addition, the cytotoxic function of NK cells from MCMV-infected NKp46-Cre-Gata3^fl/fl mice was intact, suggesting that stimulatory signals through activating receptors on NK cells from Nkp46-Cre-Gata3^fl/fl mice are largely intact and capable of inducing cytokine production and cytotoxicity.

Despite having iNK cells with reduced frequencies in the peripheral organs, NKp46-Cre-Gata3^fl/fl mice controlled the virus to a similar extent as their control littermates on day 3 postinfection. Remarkably, the frequencies of NK cells in spleens and livers of NKp46-Cre-Gata3^fl/fl mice reached similar numbers to NK cells from control littermates. The robust NK cell expansion can be explained by the enhanced proliferative capability of NK cells from NKp46-Cre-Gata3^fl/fl mice. The increased proliferative capability was cell intrinsic, as NK cells from NKp46-Cre-Gata3^fl/fl mice divided more robustly in response to IL-2 or IL-15 than NK cells from control mice in vitro. Adoptive transfer experiments supported that the heightened proliferative ability of NK cells from NKp46-Cre-Gata3^fl/fl mice was cell intrinsic. Finally, the enhanced proliferation of NK cells from NKp46-Cre-Gata3^fl/fl mice was associated with increased CD25 expression, which allowed NK cells to efficiently respond to low doses of IL-2. Similar to the aforementioned defective IFN-γ response to IL-12, the increased expression of CD25 is not explained by dysregulated IL-12 and/or IL-18 cytokine receptors. Despite the reduced IL-18 receptor expression, IL-18 stimulation induced comparable levels of CD25 expression. Moreover, higher CD25 induction was observed on NK cells in NKp46-Cre-Gata3^fl/fl mice, whereas the level of STAT4, which is a critical transcription factor for inducing CD25 upon IL-12, was comparable, and STAT4 was phosphorylated at a similar degree upon IL-12/IL-18 stimulation. There have been indications that iNK cells exhibit great proliferative capacity regardless of stimulation. In the BM, DX5⁺CD27⁺CD11b⁻ NK cells actively proliferate under steady-state conditions (12, 68). During MCMV infection, CD11b^low NK cells proliferate more than CD11b^high NK cells at day 2 postinfection (68). In addition, CD11b⁻ NK cells show enhanced proliferation as compared with CD11b⁺ NK cells when stimulated in vitro with IL-15 (69). Similarly, NK cells from Ets1-deficient mice showed reduced maturation and were hyperresponsive to IL-15 (70). Notably, a study also demonstrated cell-specific gene regulation by Gata3 by revealing a list of genes that are differentially regulated in ILC2 and Th2 cells (31). In contrast to CD25 upregulation on NK cells in our model during viral infection, GATA3 positively regulates CD25 expression on ILC2s, but not ILC3s (31). Similarly, Gata3 deletion did not reduce the expression of Klrg1 and Id2 in ILC2s. Altogether, these results indicated the stage- and cell-specific roles of GATA3 during ILC development. Therefore, it might be interesting to know whether CD25 upregulation on NK cells in NKp46-Cre-Gata3^fl/fl mice is regulated by developmental stage-specific epigenetic modifications in the Cd25 locus. The differential binding of GATA3 on the NKG2A promoter region of human NK cells but not Jurkat T cells supports the influence of epigenetic regulations on the roles of GATA3 (71).

Taken together, we dissected the role of GATA3 in regulating the development of NK cells. Even though GATA3 is largely dispensable for the generation of cNK cells, it has a significant impact on the maturation of NK cells, resulting in a migratory defect in BM egress in the absence of GATA3. Notably, peripheral iNK cells of NKp46-Cre-Gata3^fl/fl mice expand more than mNK cells from control littermates during MCMV infection, allowing rapid accumulation of NK cells with intact cytotoxicity. Our data demonstrated that iNK cells hold a higher intrinsic proliferative capability and can serve as a reservoir of cytotoxic NK cells upon infection and therefore will advance our understanding of the antiviral NK cell response in individuals who have defects in NK cell maturation.

We thank Almohanad Alkayyal in Dr. Rebecca Auer’s laboratory at Ottawa Hospital Research Institute for help with the transwell migration assay. We also thank the National Institutes of Health/National Cancer Institute at Frederick Biological Resources Branch Preclinical Repository for providing the rhIL-2 cytokine.

E.V. is a cofounder and shareholder of Innate–Pharma. The other authors have no financial conflicts of interest.