Loss of Canonical Notch Signaling Affects Multiple Steps in NK Cell Development in Mice

This article is featured in In This Issue, p.3145

Natural killer cells represent a distinct group of innate lymphocytes important for controlling viral infections as well as cancer (1) and were first defined as cytotoxic effector cells able to kill target cells without specific immunization. NK cells also produce proinflammatory cytokines, such as TNF and IFN, and chemokines that regulate adaptive immune responses as well as other innate populations, including dendritic cells, macrophages, and neutrophils (1). NK cell deficiency results in an increased susceptibility to infection and correlates with elevated cancer incidence (2, 3). NK cell malignancies, although infrequent, are very aggressive and difficult to treat (4). Thus, understanding NK cell development has both basic biological and clinical significance.

Mouse conventional cytotoxic NK cells are defined phenotypically as CD3⁻NK1.1⁺CD49b⁺ cells (5) and are mainly derived from progenitors in the bone marrow (BM) but can be found in multiple different tissues, including the spleen, lymph nodes, and liver (6).

NK cell differentiation from hematopoietic stem cells (HSC) and their developmental niches as well as the cellular and regulatory pathways governing NK cell development are not fully understood. The current model of mouse NK cell development comprises several major cellular stages (Fig. 1A) with the earliest step involving generation of a common lymphoid progenitor (CLP), giving rise to T, B, and NK cells (5). The transition from CLP to an early NK cell progenitor (pre-NKP) is marked by the downregulation of FLT3 (CD135) (pre-NKP) (7), followed by upregulation of IL-2/IL-15R β-chain (CD122) (5) in fully restricted NKPs (rNKPs) (7). NKPs express lower levels of IL-7R (CD127) compared with CLPs; however, the dynamics of IL-7R expression during early NK lymphopoiesis are not well understood (8). From the NKP stage, cells upregulate NK1.1 receptor [immature NK (iNK) cell 2], and as they further mature, they acquire the expression of Ly49 receptors (iNK3) and integrin α2 (CD49b) [mature NK (mNK) cell 4] (5), upregulate CD11b (Mac-1) (mNK5) and CD43 molecules (9), and downregulate CD27 (mNK6) (5, 10, 11) (Fig. 1A). Thymic NK cells express IL-7R, lack Ly49 receptor expression, and show low levels of CD43 and CD11b (12).

The identification of bipotent T/NK progenitors in multiple fetal tissues (13–15), as well as in adult BM (16), and the fact that the early T cell progenitors in adult thymus maintain the ability to generate NK cells (17, 18) suggest a close developmental relationship between T and NK cells (18). Multiple transcription factors such as TCF-1, T-BET, eomesodermin (Eomes), and ID2 are required for the development of both T and NK cells, whereas Bc11b upregulation inhibits NK lineage potential (5, 19, 20).

Notch signaling is essential during T cell commitment in the thymus, including suppressing alternative lineage fates of developing lympho-myeloid progenitors (21) with B cell, myeloid, and dendritic cell potentials (22–24). In contrast, several in vitro studies have suggested that NK cell development might not be restricted by Notch signaling (16, 17) and that Notch signaling may be permissive or even supportive for NK lymphopoiesis (reviewed by Kling et al. 25). In agreement with this, mice reconstituted with Rbpj-deleted BM cells showed no defect in mNK cell generation (26). However, the physiological role of canonical Notch signaling during early NK cell development remains unexplored.

Rbp-Jk^fl/flVav-Cre^tg/+ mice were generated by cross-breeding of Rbp-Jk^fl/fl (27) and Vav-Cre^tg/+ (28) mice; wild-type (WT) C57BL/6 CD45.1 (JBmsd) mice were obtained from The Jackson Laboratory. All mice were maintained under specific pathogen-free conditions at Lund University Animal Facility. The Ethical Committee at Lund University approved all performed experiments.

The BM, spleen, and thymus were obtained as previously described (16). BM cells were extracted using mortar. Single-cell suspensions were prepared by breaking up the tissues in PBS containing 5% FCS (Sigma-Aldrich) and filtering through a 70-μm cell strainer (BD Biosciences) (16). Cells were counted with the Sysmex (KX-21N) Hematology analyzer or in a Neubauer chamber. WBCs were isolated from peripheral blood (PB) on a dextran gradient (16).

Cells were incubated with 2.4G2 (anti-FcRIII) Ab to block Fc receptors and then stained with specific mAbs (listed in Supplemental Table I). 7-aminoactinomycin D (7-AAD) (Sigma-Aldrich) was used to exclude dead cells from the analysis. Fluorescence minus one (FMO) and isotype controls were used to determine the positive signal. Intracellular staining to detect Eomes and IFN-γ was performed using Foxp3 Staining Buffer Set (eBioscience) following the manufacturer’s instructions. Samples were analyzed on LSR II (BD Biosciences), and analysis was done with FlowJo software (TreeStar).

All cell sorting was performed on FACSAria IIu (BD Biosciences) using Purity precision mode on 96-well plates.

Spleen cells were plated at 2 × 10⁶ cells/ml in round-bottom 96-well plates in RPMI 1640 medium (PAA Laboratories) supplemented with 10% FCS (Sigma-Aldrich), 1% penicillin/streptomycin (Sigma-Aldrich), 1% L-glutamine (Sigma-Aldrich), and 10⁻⁴ M 2-ME (Sigma-Aldrich). NK cells were specifically activated with a 4-h incubation with a purified anti-NK1.1 Ab at 37°C. Cells incubated in medium alone were used as a negative control, whereas those activated with PMA (Sigma-Aldrich) and ionomycin (IO) (Sigma-Aldrich) were a positive control. Afterwards, cells were harvested and the surface expression of CD107a (lysosomal-associated membrane protein 1; LAMP-1) (29) was detected by flow cytometry. To measure IFN-γ production, spleen cells were plated at 5 × 10⁶ cells/ml in round-bottom 96-well plates in RPMI 1640 (PAA Laboratories), supplemented with 10% FCS (Sigma-Aldrich), 1% of penicillin/streptomycin (Sigma-Aldrich), and 10⁻⁴ M 2-ME (Sigma-Aldrich), and cultured for 4 h in the presence of 60 ng/ml of recombinant mouse IL-12 (BioLegend) and 30 ng/ml of recombinant mouse IL-18 (BioLegend) with brefeldin A and monensin (BioLegend). Negative and positive control were treated as in the degranulation assay. Intracellular IFN-γ production within different NK cell populations was detected by flow cytometry.

Lethally irradiated (900 cGy) 10–23-wk-old C57BL/6 CD45.1 WT recipient mice were transplanted with 2.5 × 10⁶ unfractionated BM cells from 10-wk-old CD45.2 Rbp-Jk^fl/flVav-Cre^tg/+ or Rbp-Jk^fl/flVav-Cre^+/+ littermate controls together with 2.5 × 10⁶ unfractionated BM cells from 10-wk-old WT CD45.1 mice. At 8–10 wk after transplantation, mice were analyzed for multilineage donor-derived reconstitution in PB, spleen, thymus, and BM.

Cells were sorted in duplicates at 25 cells per well into 96-well plates into 4 μl of lysis buffer containing 0.4% NP40, deoxynucleoside triphosphates, DTT, and RNase OUT (Invitrogen) and were snap frozen. RT-PCR and 18 cycles of preamplification with a mix of TaqMan probes (at 0.4×) were performed in each well following the One-Step qRT/PCR with ROX protocol from Invitrogen. The product (diluted 1:5) and TaqMan probes were loaded into a 48.48 Dynamic Array IFC (Fluidigm) following the manufacturer’s protocol. The following genes were detected using respective probes (Applied Biosystems): Blimp1 Mm00476128_m1; Elf4 Mm01321797_m1; Eomes Mm01351985_m1; Ets1 Mm01175819_m1; Gata3 Mm00484683_m1; Gzmb Mm00442834_m1; Hes1 Mm01342805_m1; Hes5 Mm00439311_g1; Hprt Mm00446968_m1; Id2 Mm00711781_m1; Ikzf1 Mm00456421_m1; Il15ra Mm04336046_m1; Il2ra Mm00434261_m1; Il2rb Mm00434268_m1; Il2rg Mm00442885_m1; Il7r Mm00434295_m1; Irf2 Mm00515206_m1; Nfil3 Mm00600292_s1; Notch1 Mm00435245_m1; Notch2 Mm00803077_m1; Prf1 Mm00812512_m1; RBP-Jk Mm01217627_g1; Stat4 Mm00448890_m1; Stat5a Mm03053818_s1; Tbx21 Mm00450960_m1; Tcf1 Mm00493445_m1; Tox Mm00455231_m1; and Ubc Mm01201237_m1. Chips were run in Biomark (Fluidigm), and data were analyzed through Real-Time PCR Analysis Software (Fluidigm). Ubiquitin C and Hprt were both used as reference genes for double normalization.

All results are expressed as means (±SEM). Graphs and statistical analysis were performed using GraphPad Prism 6 (GraphPad Software, La Jolla, CA). The statistical significances between groups were determined by the Mann–Whitney U test for unpaired samples.

To address the potential involvement of the Notch pathway during NK cell development, we first investigated the expression pattern of the essential mediator of canonical Notch signaling Rbp-Jk and other genes related to the Notch pathway, within different NK cell compartments (Fig. 1A). mRNA for Rbp-Jk, as well as Notch1 and Notch2 receptors, and the transcription factor Tcf7, known to be induced by Notch signaling (30), were expressed at all NK cell developmental stages (Fig. 1B). The expression of the Notch direct target gene Hes-1 (31) was also detected in nearly all NK cell compartments, whereas Hes-5 mRNA was found only in pre-NKP and rNKP but not at later stages (Fig. 1B).

To elucidate the role of the Notch pathway during NK cell development, we generated Rbp-J-k^fl/flVav-Cre^tg/+ mice in which Rbp-Jk is specifically deleted in all hematopoietic cells after crossing Rbp-Jk^fl/fl with Vav-Cre^tg/+ mice (27, 28).

The distribution of mature cells in PB in Rbp-Jk^fl/flVav-Cre^tg/+ mice was significantly altered compared with Rbp-Jk^fl/flVav-Cre^+/+ littermate controls (Fig. 1C, 1D); however, the total WBC counts were not changed (Fig. 1E). Interestingly, and not previously reported, Rbp-Jk–deleted mice showed a 67% decrease in the frequency of conventional NK1.1⁺CD49b⁺ NK cells in PB (Fig. 1C, 1D). As previously shown (22, 32), T cells were almost undetectable, whereas B cells were slightly increased in PB of Rbp-Jk^fl/flVav-Cre^tg/+ mice (Fig. 1C, 1D). Cellularity was increased 1.8-fold in the spleen, unchanged in the BM, and showed a reduction of 87% in the thymus (Fig. 1F). As previously reported, T cell development was impaired after Notch deletion (32), including a loss of double negative thymocytes (DN) 2, DN3, and double positive thymocytes as well as a significant decline in DN1/early thymic progenitors (33) (Supplemental Fig. 1A–D). Also, single positive CD4⁺ and single positive CD8⁺ thymocytes were severely reduced in Rbp-Jk–deleted mice (Supplemental Fig. 1A–D). Although, as reported in previous studies (26), the proportion of IL-7R⁺ thymic NK cells among thymocytes was not altered after Rbp-Jk deletion (data not shown). When taking into account reduced thymic cellularity (Fig. 1F), the total number of thymic NK cells was decreased (Supplemental Fig. 1E), whereas both the frequency (data not shown) and the total number of IL-7R⁻ conventional NK cells was increased (Supplemental Fig. 1F). Also, in the absence of Notch signaling, the total number of CD19⁺ B cells was expanded in the thymus (32), and the frequency of CD11b⁺CD11c⁺MHC class II⁺ dendritic cells was increased (data not shown) (34), whereas the total dendritic cell number was not significantly altered (Supplemental Fig. 1G, 1H).

Next, we investigated whether the loss of canonical Notch signaling has an impact on NK cell differentiation at different hierarchical developmental stages (Fig. 1A).

The proportion of the earliest BM lymphoid progenitors are as follows: lymphoid-primed multipotent progenitor (LMPP) and CLP were not changed in Rbp-Jk^fl/flVav-Cre^tg/+ mice, in sharp contrast to pre-NKP (Lin⁻CD27⁺CD244⁺CD127⁺FLT3⁻CD122⁻) and rNKP (Lin⁻CD27⁺CD244⁺CD127⁺FLT3⁻CD122⁺) (7) that were significantly reduced (50 and 92.2% respectively; Fig. 2A, 2B).

Conventional NK cells undergo their final maturation and acquire cytotoxic functionality in the spleen (10, 11, 35–37); therefore, we next examined whether Rbp-Jk deletion affects different splenic NK cell populations (Fig. 1A). Notably, the total number of iNK2 (NK1.1⁺Ly49⁻CD49b⁻), mNK4 (NK1.1⁺Ly49⁺CD49b⁺CD27⁺CD11b⁻), and mNK5 (NK1.1⁺Ly49⁺ CD49b⁺CD11b⁺CD27⁺CD43⁺) cells was significantly increased, whereas the most mature cytotoxic mNK6 (NK1.1⁺Ly49⁺CD49b⁺CD11b⁺CD27⁻CD43⁺) cells were reduced in Rbp-Jk^fl/flVav-Cre^tg/+ mice (Fig. 2C, 2D), together suggesting a developmental defect resulting in severe reductions at the final maturation stage of mNK6.

Collectively, these results support the involvement of canonical Notch signaling in regulation of the earliest NK cell developmental stages in the BM as well as their final maturation steps in the spleen.

To better understand the perturbed differentiation of Rbp-Jk–deleted NK cells, we investigated the expression of specific receptors important for NK cell maturation and function (5).

Functional cytotoxic NK cells express a diverse repertoire of Ly49 receptors required for NK cell tolerance and function. Ly49 receptors start to be expressed at the iNK3 stage and Ly49 upregulation, and expression is regulated by transcription factor Eomes (38).

The number of NK cells expressing diverse Ly49 receptors at the early maturation stages of iNK and mNK4 was not significantly altered in Rbp-Jk^fl/flVav-Cre^tg/+ mice (Fig. 3A), in line with no significant changes in the level of expression and in the number of Eomes⁺ cells (Fig. 3B, 3C). However, in the absence of Rbp-Jk, the number of Ly49-expressing NK cells was reduced within the more differentiated mNK5 and mNK6 populations (Fig. 3A), indicating a delay in their functional maturation.

TRAIL is a key NK cell effector molecule important for anti-tumor activity. iNK cells upregulate TRAIL at the iNK2 stage (5), and the constitutive TRAIL expression represents a hallmark of cytotoxic iNK cells (39). Importantly, after Rbp-Jk deletion, both the total number of NK cells expressing TRAIL and the level of receptor expression were clearly increased at all of the differentiation stages, further supporting their delayed maturation (Fig. 3B, 3C).

Killer cell lectin-like receptor G1 (KLRG1) is an inhibitory receptor for MHC class I, and the fraction of KLRG1⁺ NK cells increases during their maturation (40). In Rbp-Jk–deleted mice, the level of KLRG1 expression and the number of KLRG1⁺ cells were significantly reduced within the iNK pool and the mNK6 population (Fig. 3B, 3C).

These results together suggest that maturation of NK cells in Rbp-Jk–deleted mice is delayed, and their phenotype is altered.

Next, we investigated whether and how Rbp-Jk deletion affects the expression of genes controlling NK cell development (41).

Transcription factors and cytokines govern NK lymphopoiesis, starting from the earliest progenitors. Although many are shared with the other lineages, some are unique to NK cells, and their specific deletion results in NK cell deficiencies (41). Transcription factors ID2 (42, 43), STAT5 and STAT4 (44, 45), IRF-2 (46, 47), ETS1 (48), NFIL3 (also called E4BP4) (49), and IKZF1 (50) are known to drive early NK cell developmental stages, whereas Eomes and T-BET (encoded by Tbx21) (38, 51, 52), ELF4 (also known as MEF) (53), TCF1 (8), GATA3 (12, 54), BLIMP1 (55), and TOX (56, 57) are acting at the later stages involving differentiation, expansion, and functional maturation.

Consistent with the reduction in pre-NKPs and rNKPs in Rbp-Jk–deleted mice (Fig. 2A, 2B), the expression of Ets-1 and Ikzf1 mRNA encoding the transcription factor IKAROS was significantly reduced in Rbp-Jk–deleted pre-NKPs (Fig. 4A). Also, the levels of Stat5b, Stat4, and Irf2 mRNA were reduced in Rbp-Jk^fl/flVav-Cre^tg/+ pre-NKPs (Fig. 4A). In the absence of canonical Notch signaling, Id2 expression was reduced in pre-NKPs but increased in iNK2 and iNK3 cells, whereas it was not changed at the later stages (Fig. 4A). Nfil3 expression was not altered within any of the NK cell compartments after Rbp-Jk deletion (Fig. 4A). Because Nfil3 is required only during early NK cell development (58), the NK cell impairment in Rbp-Jk–deleted mice is likely due to Notch signaling affecting other important NK cell regulatory pathways.

IL-7 is a cytokine important during lymphoid development, and IL-7R is expressed within pre-NKP and rNKP populations (7, 43). Interestingly, the expression of Il-7r was reduced in pre-NKPs after Rbp-Jk deletion (Fig. 4A).

In line with the perturbed and delayed maturation of Rbp-Jk–deleted NK cells, the expression of several late-acting regulators was changed. Eomes mRNA levels were reduced in Rbp-Jk–deleted iNK2 and iNK3 cells, whereas the gene expression of Tbx21 as well as Tox and Blimp1 was not altered (Fig. 4B). The levels of Elf4 were lower in all of the NK cell developmental stages in Rbp-Jk^fl/flVav-Cre^tg/+ mice, whereas the expression of Tcf1 was reduced in both iNK and mNK cells (Fig. 4B).

GATA3, which is directly regulated by Notch (21), is critical for thymic NK cells; however, Gata3-deficient mice showed no major defects in conventional NK cells (12, 54). Upon Rbp-Jk deletion, Gata3 was clearly expressed within the different NK cell compartments with reduced levels in pre-NKP and increased expression in iNK3 cells (Fig. 4B).

Collectively, these data suggest that the loss of canonical Notch signaling has an impact on the expression of multiple key genes important for early and late NK cell developmental stages.

To establish whether the NK cells generated in the absence of Rbp-Jk are functionally affected, we analyzed the expression of specific receptors involved in NK cell–mediated killing, cytotoxic activity using the degranulation assay, and cytokine release after activation.

Despite the significant reduction in the mature cytotoxic NK cell pool, Rbp-Jk–deleted NK cells maintained their ability to release cytolytic granules after specific activation with anti-NK1.1 Ab (59) that is marked by the upregulation of CD107a (LAMP-1) molecule at the cell surface (Fig. 5A). Also, the expression of NKp46, NK1.1, and IL-2Rβ (CD122) receptors, known to be engaged in NK cell–mediated killing, was not changed after RbpJ deletion (Supplemental Fig. 2). However, the levels of CD107a were significantly higher within mNK4, mNK5, and mNK6 populations, not only in NK cells activated with anti-NK1.1 Ab but even in nonactivated Rbp-Jk–deleted NK cells compared with controls (Fig. 5A, Supplemental Fig.3A).

Cytokine production represents another functional NK cell activity (1); therefore, we next investigated whether Rbp-Jk deletion has an impact on IFN-γ production. The proportion of NK cells that released IFN-γ within iNK3, mNK5, and mNK6 populations was significantly higher in Rbp-Jk–deleted mice compared with littermate controls, even without activation (Fig.5B, Supplemental Fig. 3B).

To further characterize NK cell functionality and spontaneous activation, we studied the expression of CD69 and IL-2Rα (CD25), known NK cell activation markers (60), as well as DNAM-1 adhesion molecule triggering NK-dependent anti-tumor activity (61). Indeed, in Rbp-Jk^fl/flVav-Cre^tg/+ mice, both the level of expression and absolute number of iNK and mNK cells expressing CD69 were higher (Fig. 6A–C), and although the expression level was not altered, the number of CD25 (Fig. 6B, 6C) and DNAM-1 (Fig. 6B, 6C) positive cells was increased, together indicating a hyperactivation of NK cells.

However, the expression of Gzmb and Prf1, encoding cytolytic proteins GRANZYME B and PERFORIN, released by NK cells during cytotoxic killing (1) was not altered after Rbp-Jk deletion (Fig. 6D).

Normally, liver-resident NK cells, unlike the conventional spleen NK cells, express CD49a while lacking CD49b expression (6). Interestingly, a significant fraction of splenic NK cells expressed CD49a in Rbp-Jk–deleted mice (Fig. 6B, 6C).

IL-15 is critical for the development, function, and survival of NK cells (5); however, prolonged stimulation with IL-15 leads to chronic activation and phenotypic changes (62). The expression of Il15ra, Il2rb, and Il2rg mRNA encoding IL-15Rα, IL-2Rβ, and IL-2Rγ subunits, all involved in IL-15 signaling, was elevated in iNK cells in Rbp-Jk^fl/flVav-Cre^tg/+ mice (Fig. 6D).

Taken together, these results suggest that canonical Notch signaling, in addition to an impact on the final NK cell maturation stages, plays a role in limiting the activation status of NK cells.

To determine whether the effects of the Rbp-Jk deletion are cell intrinsic, we performed competitive transplantations using unfractionated BM cells from Rbp-Jk^fl/flVav-Cre^tg/+ mice or Rbp-Jk^fl/flVav-Cre^+/+ littermate controls. The hematopoietic reconstitution from Rbp-Jk–deleted BM was significantly lower in all the tissues compared with controls (Fig. 7A). Notably, Rbp-Jk–deleted cells failed to generate not only T cells, but also NK cells, whereas B and myeloid cell regeneration was not impaired (Fig. 7B). Furthermore, although donor-derived reconstitution of CLPs was not altered, pre-NKPs were reduced by 23% (although not reaching statistical significance; p = 0.071) and rNKPs were not produced from transplanted Rbp-Jk–deleted BM cells (Fig. 7C), in agreement with the findings in the Rbp-Jk^fl/flVav-Cre^tg/+ mice themselves. The population of donor-derived mature splenic cytotoxic mNK6 cells was significantly reduced, whereas the pool of mNK4/5 cells was increased (Fig. 7D), corroborating the maturation defect seen in the adult Rbp-Jk^fl/flVav-Cre^tg/+ nontransplanted mice (Fig. 2C, 2D). Importantly, and in further support for cell-intrinsic defects induced by Rbp-Jk deletion, regeneration of NK cell compartments (both NK progenitors and mNK cells derived from WT CD45.1 BM cells cotransplanted with Rbp-Jk–deleted BM cells) was not impaired (Fig.7E–G).

To establish whether the changes in NK cell compartments induced by the loss of Notch occur also at early postnatal stages, 2-wk-old Rbp-Jk^fl/flVav-Cre^tg/+ mice were investigated. Similar to adult mice, Rbp-Jk deletion resulted in a severe reduction of thymic cellularity, whereas the spleen and BM cellularities were not changed (Supplemental Fig. 4A). The number of rNKPs in the BM was decreased in Rbp-Jk–deleted mice (Supplemental Fig. 4B, 4C), and the transition through mNK4–mNK6 stages was perturbed as illustrated by the increased number of mNK4 cells, although the NK cell phenotype in 2-wk-old mice was less severe compared with the adult.

Collectively, these results support that the effects of the deletion of canonical Notch signaling are cell intrinsic, and the impact of the loss of Notch on NK cell development is already manifested early after birth.

For decades after their discovery, conventional NK cells were thought to be the only innate effector lymphocytes. Recent studies identifying new populations of innate lymphoid cells (ILCs) placed NK cells within the ILC1 group. Conventional NK cells and noncytotoxic ILC1 cells represent different lineages (63); however, their developmental origin partially overlaps (64).

Within the hematopoietic system, Notch is critical for promoting T cell development (32). The Notch pathway has also been implicated in regulation of innate lymphocytes and in generation, differentiation, and maturation of ILC2 and other ILCs cells (25, 63). However, it has remained unclear to what degree canonical Notch signaling has an impact on conventional NKPs and maturation, because most experiments addressing the impact of Notch were limited to NK cell differentiation in vitro (16, 17, 25). Previous in vivo studies with chimera mice generated after transplantation of Rbpj-deleted BM cells showed no significant alterations in donor-derived conventional NK cell pool in the spleen and no changes in thymic-dependent IL-7R⁺ NK cell population, whereas, as expected, T cells were absent (21, 26).

Interestingly and not previously reported, Rbp-Jk–deleted mice had a severely reduced pool of conventional NK cells in PB and an altered phenotype of mature cytotoxic splenic NK cells showing a sign of spontaneous hyperactivation with elevated levels of CD69, CD25, DNAM-1, and CD107a molecules as well as increased IFN-γ production. The increased gene expression level of Il15ra, Il2ra, Il2rb, and IL2rg, all part of the IL-15 signaling cascade, in Rbp-Jk–deleted NK cells suggests that this hyperactivation could result from the altered response to IL-15, similar as in Ets1^−/− mice (48) or after chronic stimulation with IL-15/IL-15α complexes (62).

CD49a integrin is expressed on the surface of liver NK cells (6) and ILC1 cells (64) but not on steady-state conventional splenic CD49b⁺ NK cells. Because liver-resident NK cells do not migrate to other tissues (6) and ILC1 cells lack CD49b, PERFORIN, and GRANZYME B expression, CD49a⁺ NK cells in the spleen in Rbp-Jk–deleted mice most likely represent the conventional splenic NK cells with an activated phenotype, similar to superactivated CD49a⁺ T lymphocytes (65). Furthermore, human PB cytotoxic NK cells have been shown to change their phenotype and upregulate CD49a after exposure to hypoxia, TGF-β1, and a demethylating agent (66).

To better depict the impairment in generation of mNK cells and their altered phenotype, we investigated in detail different NK cell compartments. The loss of canonical Notch signaling had no effect on the earliest LMPP or CLP BM progenitors. In contrast, the pool of NK lineage-restricted progenitors pre-NKPs and in particular rNKPs (7) was dramatically decreased, whereas the iNK cells were increased, suggesting a partial block at the early NK cell developmental stages. Furthermore, compatible with perturbed NK cell maturation, the number of cells expressing KLRG1 receptors within the splenic iNK cells in Rbp-Jk–deleted mice was reduced, whereas the number of TRAIL-positive cells increased and, similar to mature cytotoxic NK cells, displayed an activated phenotype as indicated by the increased number of DNAM1- and CD69-positive cells.

The expression of IL-7R, CD69, DNAM1, TRAIL, and T-BET within CD3⁻NK1.1⁺CD49b⁻ population defines a heterogeneous pool that includes some subsets of ILC1. However, ILC1 cells are rarely found in the spleen, and the CD49b⁻ splenic NK cells found in RbpJ-deleted mice did not show an increase in IL-7R expression compared with controls but clearly expressed PERFORIN and GRANZYME B, cytolytic proteins not typically produced by ILC1 cells. The increased levels of TRAIL detected in the CD49b⁻ cells could be misleading; however, given the fact that the expression of TRAIL was maintained in more mature CD49b⁺ NK cells, these results collectively suggest that Rbp-Jk–deleted splenic iNK cells have a preserved NK lineage identity (67).

The final NK cell maturation stages were also affected by the loss of Notch, leading to increased mNK4 and mNK5 compartments and reduced pool of cytotoxic mNK6 cells.

These findings are in agreement with the previous work implicating the involvement of the Notch pathway in promoting generation of human NK cells from hematopoietic progenitors, regulating production of IFN-γ by NK cells, enhancing expression of KIR receptors, and functional NK cell maturation during sequential differentiation stages of human NK cells in vitro (68, 69).

The requirements for Notch signaling were cell autonomous, as Rbp-Jk–deleted BM cells reconstituted the phenotype of donor Rbp-Jk^fl/flVav-Cre^tg/+ mice when competitively transplanted into WT recipients, and Rbp-Jk–deleted cells in these chimera mice had no effect on generation of NK cell compartments from cotransplanted WT BM donor cells. Although previous studies reported (26) that chimera mice generated by transplanting Rbp-Jk–deleted BM cells showed no changes in splenic CD3⁻CD127⁻NKp46⁺ NK cells, the different NK cell developmental stages were not investigated. In agreement with that work (26), the frequency of CD127⁺ thymic NK cells was not altered in Rbp-Jk–deleted mice in our studies, although the observed increase in the CD127⁻ conventional NK cells in the thymus, after Rbp-Jk deletion, in this study was not previously reported.

Despite the almost undetectable rNKP pool in Rbp-Jk–deleted BM, functional NK cells were generated. This, together with the high expression of NK lineage-specific genes in Rbp-Jk–deleted pre-NKP and iNK cells, suggests that the CLP-rNKP path might not be the only NK lineage pathway and that potentially LMPPs and CLPs in the BM, and early thymic progenitors in the thymus could support NK cell generation independently of rNKPs.

To better understand the impact of canonical Notch signaling on the sequential NK cell differentiation stages, we investigated the expression of genes involved in the NK cell regulatory circuits and found that Rbp-Jk deletion affected the expression of multiple genes critical for NK cell development and function. In line with the severe reduction in pre-NKPs and rNKPs in Rbp-Jk–deleted mice, the expression of Ikzf1, Stat5b, and Ets1, known to be important for regulation of the very early NK developmental stages (41), was significantly reduced.

The perturbation in NK cell maturation in the absence of canonical Notch signaling can be, in part, due to the reduced level of Eomes that controls different aspects of NK (38, 41, 70) and T cell differentiation (70). The Eomes expression in CD8⁺ T cells is directly regulated by Notch signaling (71) and, in this study, we show that Notch deletion affects the Eomes expression in iNK cells. Consistent with Eomes being required for the expression of Ly49 and CD49b in differentiating NK cells (38, 67), the loss of Notch led to a perturbed NK cell maturation. This is further supported by studies showing that the reduced levels of Eomes in patients after BM transplantation affect NK cell differentiation and function (72). Reduced expression of transcription factor MEF, which is known to be important for NK cell maturation (53), could also contribute to the iNK cell status in Rbp-Jk–deleted mice.

T-BET regulates NK cell cytotoxic activity (51) and, together with EOMES, is essential for generation of functional NK cells (38). The reduced Eomes expression and unchanged levels of T-bet in RbpJk-deleted NK cells, most likely, contribute to their immature phenotype. Consistent with the unaltered expression of T-bet, NK cell cytotoxic activity seems not to be affected by the absence of Notch signaling.

In agreement with the finding that TCF-1 regulated by canonical Notch signaling (30, 73) is required for generation of NKPs and mNK cells (8), Tcf1 expression was significantly reduced after Rbp-Jk deletion, leading to a decrease in the number of NKPs as well as mNK cells.

Identifying how exactly the loss of Rbp-Jk impacts NK cell development at the molecular level will require further studies that are beyond the scope of this work.

Collectively, our data support a critical role of canonical Notch signaling at multiple stages during NK cell lineage development and for NK cell functionality.

We thank Barbara L. Kee, Kevin Ramirez, and Natalija Buza-Vidas for advice and helpful discussions and Olga Kotova for technical assistance. We are thankful to Barbara L. Kee for critically reading and commenting on the manuscript.

The authors have no financial conflicts of interest.