Eomesodermin spatiotemporally orchestrates the early and late stages of NK cell development by targeting KLF2 and T-bet, respectively

Abstract

Eomesodermin (Eomes) is a critical factor in the development of natural killer (NK) cells, but its precise role in temporal and spatial coordination during this process remains unclear. Our study revealed that Eomes plays distinct roles during the early and late stages of NK cell development. Specifically, the early deletion of Eomes via the CD122-Cre transgene resulted in significant blockade at the progenitor stage due to the downregulation of KLF2, another important transcription factor. ChIP-seq revealed direct binding of Eomes to the conserved noncoding sequence (CNS) of Klf2. Utilizing the CHimeric IMmune Editing (CHIME) technique, we found that deletion of the CNS region of Klf2 via CRISPRi led to a reduction in the NK cell population and developmental arrest. Moreover, constitutive activation of this specific CNS region through CRISPRa significantly reversed the severe defects in NK cell development caused by Eomes deficiency. Conversely, Ncr1-Cre-mediated terminal deletion of Eomes expedited the transition of NK cell subsets from the CD27⁺CD11b⁺ phenotype to the CD27⁻CD11b⁺ phenotype. Late-stage deficiency of Eomes led to a significant increase in T-bet expression, which subsequently increased the expression of the transcription factor Zeb2. Genetic deletion of one allele of Tbx21, encoding T-bet, effectively reversed the aberrant differentiation of Eomes-deficient NK cells. In summary, we utilized two innovative genetic models to elucidate the intricate mechanisms underlying Eomes-mediated NK cell commitment and differentiation.

Introduction

Natural killer (NK) cells play a crucial role in the elimination of tumor cells and pathogen-infected cells [1–3]. NK cells originate from common lymphoid progenitors (CLPs) and undergo development in the bone marrow before circulating to peripheral tissues. The development of NK cells involves the progressive acquisition of specific receptors at different stages. CLPs differentiate into NK cell progenitors (NKps), characterized by increased expression of CD122, an IL-15 receptor subunit that confers sensitivity to IL-15, a crucial determinant of NK cell commitment. Immature NK cells undergo cytokine-driven development, particularly IL-15-mediated maturation, during which they acquire the expression of functional receptors, the capability for cytokine secretion, and cytotoxic activity. Based on their expression of CD27 and CD11b, immature NK cells can be further classified into subsets, such as the DN (CD27⁻CD11b⁻) and CD27 SP (CD11b⁻CD27⁺) subsets. Mature NK cells comprise two subsets, DP (CD11b⁺CD27⁺) and CD11b SP (CD27⁻CD11b⁺) [4, 5].

The developmental processes of NK cells are regulated by a complex network of transcription factors [6, 7]. Among these regulators, T-box transcription factors, including Eomesodermin (Eomes) and T-bet, exert broad effects on directing lymphocyte development and function. T-box transcription factors possess a conserved sequence-specific DNA-binding motif known as the T-box domain, which spans 180-200 amino acid residues. Specifically, T-bet has been identified as an indispensable driver of Th1 differentiation and IFN-γ production in CD4⁺ T cells [8]. Furthermore, T-bet is the predominant transcription factor in mature NK cells, increasing their responsiveness to IL-12 while concurrently suppressing cell cycle progression through the downstream transcription factor Zeb2 [9]. Eomes has been demonstrated to collaborate with T-bet to facilitate the expression of CD122 [10]. Consequently, it is postulated that Eomes and T-bet redundantly regulate the differentiation of CD8⁺ effector T cells [11, 12]. Eomes and T-bet govern distinct sets of genes involved in NK cell development. Eomes is expressed predominantly in immature NK cells and plays a pivotal role in early lineage specification [9]. However, the precise temporal and spatial regulatory role of Eomes during NK cell development, encompassing the commitment, differentiation, and maturation stages, remains largely elusive.

In this study, we employed an innovative Eomes-deficient mouse model to elucidate the role of Eomes in regulating NK cell commitment and maturation. Our findings demonstrated that deletion of Eomes at the NKp stage inhibited NK cell maturation, while the absence of Eomes at the iNK stage promoted the terminal maturation of NK cells. Remarkably, during the early stage, Eomes directly targeted KLF2 to facilitate NK cell commitment and maturation. Furthermore, during the terminal maturation of NK cells, Eomes suppressed T-bet and Zeb2 expression, thereby maintaining the rejuvenation of NK cells. Collectively, our findings underscore the intricate role played by Eomes in orchestrating the development of NK cells through the modulation of different target genes.

Results

Deletion of Eomes during the NKp stage results in significant impairment of NK cell development

To obtain a comprehensive understanding of the regulatory role of Eomes in NK cell development, we conducted an analysis of Eomes expression profiles in NK cells across different subsets and developmental stages. Our findings revealed a dynamic pattern of Eomes expression, with particularly heightened levels observed during the early phase of NK cell development. Specifically, NKp and CD27⁺CD11b^– (CD27 SP) immature NK cells exhibited higher expression levels of Eomes than CD27⁺CD11b⁺ (DP) and CD27^–CD11b⁺ (CD11b SP) mature NK cells (Fig. S1A, B). Further investigation into these observations may provide valuable insights into the specific roles of Eomes during different phases of NK cell development. To clarify the sequential involvement of Eomes during the commitment stage of NK cell development, we performed a breeding experiment involving Eomes^f/f mice and CD122-Cre mice (Fig. S1C). The use of Eomes^f/f/CD122^Cre/+ mice enabled us to selectively inactivate Eomes beginning at the NKp stage. We selected Eomes^f/f mice as the appropriate control group for subsequent comparisons with Eomes^f/f/CD122^Cre/+ mice because CD122-Cre transgenic mice and Eomes^f/f mice have similar percentages and absolute numbers of NK cells (Fig. S1D). The deletion efficiency was validated through intracellular staining of Eomes in relevant NK cell subpopulations. Eomes was effectively depleted at the translational level beginning at the NKp stage was observed in Eomes^f/f/CD122^Cre/+ mice, allowing specific investigation of the impact of Eomes at this particular stage of NK cell development (Fig. S1E). The depletion of Eomes via the CD122-Cre transgene did not significantly impact the number of T or B cells in the examined tissues compared to the corresponding value in their wild-type littermate controls (Fig. S2A, B). Despite the expression of CD122 in NKT cells, conditional deletion of Eomes in Eomes^f/f/CD122^Cre/+ mice did not affect the proportion or absolute number of NKT cells in the spleen or liver compared to the corresponding value in control mice (Fig. S2C, D). However, compared with their WT counterparts, Eomes^f/f/CD122^Cre/+ mice exhibited a significant decrease in both the proportion and absolute number of CD3⁻NK1.1⁺ NK cells across various tissues, including spleen, bone marrow (BM), lymph nodes (LNs), liver, and lung tissues (Fig. 1A).

Fig. 1.

Deletion of Eomes during the NKp stage impairs NK cell homeostasis and development. A Representative flow cytometry plots (left) and absolute numbers (right) of CD3⁻NK1.1⁺ NK cells in the indicated organs from Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice (n = 8). B Representative flow cytometric profiles (left) and the percentages (right) of NK subsets in the spleen and bone marrow (BM) of Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice. NKp (CD3⁻CD122⁺NK1.1⁻CD11b⁻), iNK (CD3⁻CD122⁺NK1.1⁺CD11b⁻) and mNK (CD3⁻CD122⁺NK1.1⁺CD11b⁺) (n = 4). C Representative flow cytometric profiles (left) and the percentages (right) of NK cell subsets in the spleen and BM of Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice. DN (CD3⁻NK1.1⁺CD27⁻CD11b⁻), CD27 SP (CD3⁻NK1.1⁺CD27⁺CD11b⁻), DP (CD3⁻NK1.1⁺CD27⁺CD11b⁺), and CD11b SP (CD3⁻NK1.1⁺CD27⁻CD11b⁺) (n = 5). D The proportions of development-related receptors expressed on NK cells in the spleen and BM of Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice (n = 5). E Representative plot (left) and rejection rates (right) of β2m^–/– splenocytes in the spleen and LNs of Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice. R1 refers to CFSE^low splenocytes from WT mice, while R2 refers to CFSE^high splenocytes from β2m-deficient mice (n = 6). F Representative flow cytometry plots (left) and percentages (right) of RMA-S cell rejection in Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice (n = 5). G The gross morphology of lung lobes (left), quantification of lung weights (middle), and total number of metastatic nodules in the lungs (right) in Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice (n = 5). Each symbol represents an individual mouse. The data are representative of at least three independent experiments. The data are presented as the means ± SDs

We subsequently investigated the impact of Eomes deletion on the developmental subsets of NK cells (Fig. S3). Compared to the corresponding values in WT control mice, a significant increase in the proportion of NKp cells (CD3⁻CD122⁺NK1.1⁻CD11b⁻) was observed in both the spleen and BM of Eomes^f/f/CD122^Cre/+ mice. However, the proportion of the mature subset (CD3⁻CD122⁺NK1.1⁺CD11b⁺) was notably decreased (Fig. 1B). Furthermore, we observed a significant decrease in the proportion of mature NK cells with the CD27⁺CD11b⁺ or CD27⁻CD11b⁺ phenotype and an increase in the proportion of immature NK cells with the CD27⁻CD11b⁻ or CD27⁺CD11b⁻ phenotype in Eomes^f/f/CD122^Cre/+ mice (Fig. 1C). We further analyzed the effect of Eomes deletion on NKp cells in detail. There was a significant increase in the number of NKp cells in both the spleen and bone marrow of the Eomes^f/f/CD122^Cre/+ mice compared to the mice in the control group (Fig. S4A). Additionally, NKp cell proliferation and apoptosis were also elevated in Eomes^f/f/CD122^Cre/+ mice (Fig. S4B, C). Further experiments revealed that NKp cells in Eomes^f/f/CD122^Cre/+ mice exhibited a reduced expression level of the chemokine receptor CCR5, while the level of the chemokine receptor CXCR3 remained unaffected (Fig. S4D). These findings suggest that conditional deletion of Eomes during the NKp phase impedes NKp differentiation.

During the maturation process, NK cells sequentially acquire the expression of various receptors that are crucial for their development and function. A detailed analysis of these markers revealed a significant reduction in the proportion of Ly49 family-expressing NK cells in both the spleen and BM of Eomes^f/f/CD122^Cre/+ mice. Additionally, a greater proportion of NK cells in these mice displayed the immature marker CD127, indicating delayed maturation in these mice compared to the WT controls (Fig. 1D). Furthermore, Eomes-deficient NK cells exhibited significantly decreased expression of terminal maturation markers such as KLRG1 and CD43 (Fig. 1D). Collectively, these findings suggest that conditional deletion of Eomes during the NKp stage results in impairment of the NKp-to-iNK transition, thereby impeding the maturation of NK cells.

Considering the expression of CD122 in innate lymphoid cell 1 s (ILC1s), we evaluated the impact of CD122-Cre-mediated Eomes deletion on ILC1 homeostasis in the BM and liver. Our findings revealed a significant decrease in the population of conventional natural killer (cNK) cells within the BM and liver of Eomes^f/f/CD122^Cre/+ mice. However, no notable effect on the number of ILC1s was observed (Fig. S5A–D). This could be attributed to the lack of Eomes expression in ILC precursors in the BM and liver (Fig. S5E-J).

To address whether the reductions in the NK cell pools observed in mice with Eomes deletion is a result of cell-autonomous defects rather than potential defects in hematopoietic stem cell (HSC) generation, we conducted competitive bone marrow chimera experiments (Fig. S6A). The findings revealed that NKp cells derived from WT mice demonstrated the typical capacity to differentiate into NK cells. Conversely, NKp cells with Eomes deletion exhibited notable impairment of their differentiation into NK cells, especially mature NK cells (Fig. S6B–E). Therefore, the critical requirement of Eomes for NK cell development is cell intrinsic.

Deletion of Eomes at the NKp stage results in diminished NK cell-mediated immunosurveillance

To investigate the impact of Eomes deletion during the NKp stage on NK cell function, we conducted experiments utilizing three different in vivo tumor models. Compared with control mice, Eomes^f/f/CD122^Cre/+ mice had significantly lower levels of MHC-I^−/− splenocyte rejection activity (Fig. 1E). Additionally, Eomes^f/f/CD122^Cre/+ mice had a decreased ability to clear RMA-S cancer cells from the peritoneal cavity (Fig. 1F). Furthermore, the Eomes^f/f/CD122^Cre/+ mice exhibited notably increased lung weights and tumor colony numbers in the melanoma lung metastasis model (Fig. 1G). Collectively, these findings suggest that the absence of Eomes during NKp development impairs the immune surveillance capacity of NK cells.

Eomes regulates early NK cell development by controlling the expression of KLF2

To elucidate the underlying mechanism by which Eomes regulates early NK cell development, we performed RNA-Seq analysis on splenic NK cells from Eomes^f/f/CD122^Cre/+ and control mice. By comparison of the global gene expression profiles, significant differences were identified between NK cells derived from Eomes^f/f/CD122^Cre/+ mice and those derived from control mice. The analysis revealed 1014 downregulated genes and 193 upregulated genes (identified as genes with a fold decrease or increase, respectively, of greater than 2) in Eomes-deficient NK cells. To further investigate the underlying biological processes, we performed pathway analysis specifically on the downregulated genes. Notably, there was marked enrichment of genes associated with immune cell differentiation in NK cells from Eomes^f/f/CD122^Cre/+ mice compared to those from control mice. These findings suggest that Eomes is a critical regulator of the development of diverse cell lineages, including both lymphoid and myeloid cells (Fig. 2A). Furthermore, downregulation of key transcription factors involved in NK cell development, including Myb, Klf2, and Gata3, was observed in Eomes-deficient NK cells (Fig. 2B). These findings were validated through RT‒PCR analysis, which confirmed that the expression levels of Myb and Klf2 were lower in NK cells from Eomes^f/f/CD122^Cre/+ mice than in those from control mice (Fig. 2C). ChIP-seq (chromatin immunoprecipitation and sequencing) was performed to investigate whether Eomes directly regulates Klf2 and Myb expression to modulate NK cell development. The results showed that Eomes directly bound to the promoter region of Klf2 (Fig. 2D). However, no significant binding of Eomes to the promoter region of the Myb gene was detected (Fig. 2E). Consistent with previous reports [13], we also demonstrated the dynamic expression of KLF2 in NK cells (Fig. S7). Moreover, the deletion of KLF2 has been shown to inhibit NK cell maturation, consistent with the phenotype associated with Eomes deletion [13]. Therefore, we speculate that Eomes potentially contributes to the development of NK cells through the regulation of KLF2.

Fig. 2.

Eomes regulates early NK cell development by controlling the expression of KLF2. A RNA sequencing was performed on NK cells isolated from the indicated mice. Pathway enrichment analysis results are shown. The expression of transcription factors in NK cells was analyzed by RNA sequencing (B) and validated by RT‒PCR (C) (n = 4). D, E ChIP assays were performed with the sorted splenic NK cells. The expression levels of Klf2 (D) and Myb (E) were visualized at their gene loci. F–H BM reconstitution assay. Representative flow cytometry plots (F) and the percentages (G, left), absolute numbers (G, right), and developmental stages (H) of NK cells in the recipient mice (n = 5)

Next, we aimed to investigate whether the developmental deficiency caused by Eomes deletion can be rescued by introducing KLF2. To answer this question, we utilized a retroviral system to ectopically express KLF2 in hematopoietic stem cells (HSCs) derived from Eomes^f/f/CD122^Cre/+ mice and subsequently transplanted these modified HSCs into recipient mice. After an eight-week period, we observed substantial increases in both the proportion and absolute number of NK cells, with values approximately 70% of those in the control group. This suggests that in the absence of Eomes, the ectopic expression of KLF2 can effectively restore NK cell development in the spleen and bone marrow (Fig. 2F, G). A comprehensive analysis of different NK cell subsets revealed that the overexpression of KLF2 facilitated the transition from NKp cells to iNK cells and then to mature NK cells in Eomes^f/f/CD122^Cre/+ mice (Fig. 2H). Given the crucial role played by KLF2 in maintaining the homeostasis of lymphocytes, including NK cells [13–18], it is likely that Eomes regulates NK cell development by modulating KLF2 expression.

Eomes promotes the early differentiation of NK cells by directly binding to the CNS region of KLF2

To validate the binding of Eomes to KLF2 gene promoters, ChIP‒PCR assays were conducted, and the results confirmed the presence of Eomes binding (Fig. 3A). To investigate the functional significance of Eomes-mediated regulation of KLF2, we utilized the CHimeric IMmune Editing (CHIME) system [19–21] and selected three single guide RNAs (sgRNAs) for targeted deletion of specific regions within the Klf2 promoter. Deletion of the Klf2 promoter region resulted in phenotypes similar to those observed in Eomes^f/f/CD122^Cre/+ mice, including a decreased NK cell percentage and impaired NK cell maturation compared to those of control mice (Fig. 3B–E). In contrast, constitutive activation of the Klf2 CNS region using CRISPRa led to enhanced Klf2 transcription and largely rescued the severe defect in NK cell counts observed in Eomes-deficient mice (Fig. 3F). These findings further validate that Eomes governs NK cell development by directly regulating its target gene, Klf2.

Fig. 3.

Eomes promotes the early differentiation of NK cells by directly binding to the CNS region of KLF2. A Schematic graphs showing the Eomes binding sites in the promoter region of the Klf2 gene (left). The enrichment of Eomes in the promoter region of the Klf2 gene was quantified by RT‒PCR (right) (n = 5). Representative flow cytometry plots (B) and the percentages (C) of NK cells in the recipient mice (n = 4). Representative flow cytometric profiles (D) and the quantification of NK cell developmental subsets in the spleen (E, left) and BM (E, right) of the recipient mice (n = 3). F Representative flow cytometric plots (left) and the absolute numbers (right) of NK cells after specific activation of KLF2 in Eomes^f/f/CD122^Cre/+ mice (n = 4). The values are expressed as the means ± SDs

The absence of Eomes during the iNK stage expedites the terminal maturation of NK cells

Subsequently, we investigated the impact of Eomes on NK cell maturation in later stages by generating Eomes^f/f/eYFP^stop/Ncr1^Cre/+mice through crossing of Eomes^f/f/Ncr1^Cre/+ mice and Rosa26^stopYFP mice (Fig. S1A). Expression of the Ncr1-Cre transgene resulted in efficient deletion of the target genes at the iNK stage (Fig. S8). In these mice, yellow fluorescent protein (YFP) was utilized to monitor Cre recombinase activity. The efficacy of Eomes deletion was validated through intracellular staining of relevant NK cell subpopulations. Specifically, efficient deletion of Eomes was observed in NK cells after the immature stage of development (Fig. S1C). Both the proportion and number of YFP-labeled cells were slightly lower in the spleen, BM, LN, liver, and lungs of the Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice than in the corresponding tissues of the control mice (Fig. 4A). However, the early development of NK cells in Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice, as determined by CD11b and NK1.1 staining, was comparable to that observed in wild-type mice (Fig. 4B). However, the proportions of CD27⁺CD11b⁻ and CD27⁺CD11b⁺ cells were significantly reduced in the Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice, while the proportion of terminally mature NK cells (CD27⁻CD11b⁺) was substantially greater in both the spleen and BM in these mice compared to the control mice (Fig. 4C). In contrast to those from Eomes^f/f/CD122^Cre/+ mice, NK cells from Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice exhibited a decrease in the immature subset and an increase in the mature subset, accompanied by elevated KLRG1 expression and slightly reduced CD117 and CD127 levels, compared to NK cells from control mice (Fig. 4D). The increased population of mature NK cells in Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice suggests that the expression of Eomes is beneficial for maintaining NK cell rejuvenation.

Fig. 4.

The absence of Eomes during the iNK stage expedites the terminal maturation of NK cells. A Representative flow cytometry plots (left) and the absolute numbers (right) of CD3⁻YFP⁺ NK cells in the indicated organs from Eomes^f/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 12). B Representative flow cytometric profiles (left) and percentages (right) of NKp, iNK, and mNK subsets in the spleen (upper) and BM (bottom) from Eomes^f/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 4). C Representative flow cytometric profiles (left) and the percentages (right) of NK cell subsets, including the DN, CD27 SP, DP, and CD11b SP subsets, in the spleen (upper) and BM (bottom) of Eomes^f/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 5). D The percentages of NK cells expressing developmental markers in the spleen (left) and BM (right) of Eomes^f/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 5). The data represent one of three independent experiments and are expressed as the means ± SDs

Eomes suppresses T-bet expression during NK cell terminal differentiation

To further investigate the role of Eomes in regulating late-stage NK cell development, we conducted RNA-Seq analysis on splenic NK cells from Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice and control mice. Comparison of the global gene expression profiles revealed that in Eomes^f/f/Ncr1^Cre/+ mice compared to control mice, 583 genes were upregulated, while 456 genes were downregulated. Subsequently, pathway analysis was conducted on the upregulated genes specific to Eomes deficiency. We identified significant enrichment of pathways related to leukocyte differentiation and generic transcription in Eomes-deficient NK cells by KEGG pathway enrichment analysis (Fig. 5A). Moreover, upregulation of several crucial transcription factors for NK cell development, including Tbx21 and Zeb2, was observed in Eomes-deficient NK cells (Fig. 5B). The RT‒PCR results further confirmed the elevated expression of Zeb2 and Tbx21 in NK cells from Eomes^f/f/eYFP^stop/Ncr1^Cre/+ mice (Fig. 5C). ChIP-seq analysis revealed specific binding of Eomes to the Tbx21 promoter region but not to the Zeb2 promoter region in NK cells (Fig. 5D, E). Consistent with previous reports, T-bet directly activated Zeb2 expression by binding to its promoter, while it did not directly bind to Eomes (Fig. 5F, G). The presence of H3K4me3 and H3K27ac modifications as positive controls indicated that the binding loci were located specifically at promoter regions. The ChIP‒PCR results confirmed the binding of Eomes to the promoter region of Tbx21 (Fig. 5H). T-bet expression showed a dynamic trend during the development of NK cells, increasing with the maturation of NK cells (Fig. S9). Furthermore, an increase in the T-bet protein level was observed in Eomes-deficient NK cells (Fig. 5I, J). Collectively, these findings suggest that Eomes regulates NK cell terminal differentiation by suppressing the expression of T-bet.

Fig. 5. Eomes suppresses T-bet expression during NK cell terminal differentiation.

A Results of pathway enrichment analysis of the differentially expressed genes in the indicated mice. B Heatmap visualization of genes with differential expression between WT and Eomes-deficient NK cells. C RT‒PCR analysis of the expression of the indicated transcription factors in NK cells from the indicated mice (n = 5). Genome Browser tracks of the NK cell-associated transcription factors Tbx21 (D) and Zeb2 (E) are shown for NK cell Eomes ChIP-seq vs. input DNA, H3K4me3 ChIP-seq and H3K27ac ChIP-seq. Genome Browser tracks of the NK cell-associated transcription factors Zeb2 (F) and Eomes (G) are shown for NK cell T-bet ChIP-seq vs. input DNA, H3K4me3 ChIP-seq and H3K27ac ChIP-seq. H Enrichment of Eomes binding to the Tbx21 gene was quantified by ChIP‒PCR (n = 5). Representative flow cytometric profiles (I, J, left) and MFI values (I, J, right) of T-bet expression in the total NK cell population (I) (n = 4) and the indicated NK cell subsets (J) (n = 4). All the data are presented as the means ± SDs and are representative of at least three independent experiments

Eomes suppresses T-bet expression during NK cell terminal differentiation

Subsequently, we examined the impact of T-bet on NK cell differentiation in Tbx21^f/f/eYFP^stop/Ncr1^Cre/+ mice and found significant reductions in both the proportion and number of NK cells across the evaluated tissues compared to those in control mice (Fig. 6A). The early development of NK cells in Tbx21^f/f/eYFP^stop/Ncr1^Cre/+ mice was comparable to that in wild-type mice, as evidenced by CD11b and NK1.1 expression (Fig. 6B). Additionally, we observed a significantly increased proportion of CD27⁺CD11b⁺ cells and a decreased proportion of CD27⁻CD11b⁺ mature NK cells in Tbx21^f/f/eYFP^stop/Ncr1^Cre/+ mice compared to control mice (Fig. 6C). Consistent with the observed developmental impairment, a significant decrease in the expression of the terminal mature receptors KLRG1 and CD43 was detected in NK cells from Tbx21^f/f/eYFP^stop/Ncr1^Cre/+ mice (Fig. 6D). Notably, deletion of one allele of Tbx21 resulted in a significant increase in the proportion of CD27⁺CD11b⁺ NK cells in Eomes-deficient mice (Fig. 6E), suggesting that Eomes may enhance the enrichment of this subset by antagonizing T-bet expression. Therefore, these two transcription factors may play opposite roles in maintaining the CD27⁺CD11b⁺ NK cell subset. Zeb2 is considered a downstream effector of T-bet in the terminal differentiation of NK cells. Additional evidence showed that Zeb2 exhibited reduced expression in T-bet-deficient NK cells (Fig. 6F) but was highly expressed in Eomes-deficient NK cells (Fig. 5C). Taken together, these findings suggest that Eomes may facilitate and maintain the development of CD27⁺CD11b⁺ NK cells by inhibiting the T-bet–Zeb2 axis.

Fig. 6.

Eomes inhibits NK cell terminal differentiation by inhibiting the T-bet–Zeb2 axis. A Representative flow cytometry plots (left) and the absolute numbers (right) of CD3⁻NK1.1⁺ NK cells in the indicated organs from Tbx21^/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 12). B Representative flow cytometric profiles (left) and the percentages (right) of the NKp, iNK, and mNK subsets in the spleen (upper) and BM (bottom) of Tbx21^/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 4). C Representative flow cytometric profiles (left) and the percentages (right) of NK cell subsets in the spleen (upper) and BM (bottom) of Tbx21^/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 5). D Expression of development-related receptors on NK cells in the spleen and BM of Tbx21^/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 5). E Representative flow cytometric profiles (left) and the percentages (right) of NK cell subsets in the spleen and BM of the indicated mice (n = 5). F RT‒PCR analysis of Zeb2 mRNA expression in NK cells from Tbx21^/f/eYFP^stop/Ncr1^Cre/+ and control mice (n = 4). The data are presented as the means ± SDs and are representative of at least two independent experiments

Discussion

Although the transcription factor Eomes is recognized primarily for its critical functions in T cells, its biological role extends beyond T cells to other immune cell populations, such as innate lymphoid cells (ILCs) and natural killer (NK) cells [8, 10, 22–26]. In this study, we demonstrated that Eomes plays a crucial role in regulating both the early and late stages of NK cell development by targeting key transcription factors, namely, KLF2 and T-bet, respectively. By employing two distinct genetic mouse models to specifically delete Eomes at the NKp or the iNK stage, we elucidated the stage-specific functions of Eomes during NK cell development. In all mouse models, we observed a significant reduction in the NK cell count and impaired NK cell development. Remarkably, the deletion of Eomes at the NKp stage inhibited NK cell maturation, whereas the deletion of Eomes at the iNK stage promoted NK cell maturation. Through comprehensive analysis of RNA-seq and ChIP-seq data, we revealed that Eomes directly targets KLF2 during early NK cell development and interacts with the T-bet–Zeb2 axis during late NK cell development. Thus, our study provides novel insights into the mechanism governing the regulation of NK cell development by Eomes.

During the development of certain innate immune cells, for example, both early ILCs and NK progenitors, Eomes is upregulated and plays a crucial role in promoting cell fate commitment and differentiation [22, 23, 27]. Recent research has demonstrated that Eomes is localized in different subcellular compartments within distinct subsets of human CD8 + T cells [28]. Interestingly, we found that Eomes deficiency at the NKp stage arrests NK cell maturation at the CD27 SP stage, consistent with the effects of Eomes deletion in hematopoietic stem cells via the Vav1-Cre transgene [22], suggesting that the expression of Eomes in NK cells at the early developmental stage is critical for their further development and maturation. In contrast, deletion of Eomes in NK cells at the immature stage resulted in an increase in the percentage of mature NK cells, similar to the phenotype observed when a tamoxifen-inducible, type-1-ILC-specific (Ncr1-targeted) Cre transgene was used to delete Eomes in mature NK cells [29]. However, this observation contradicts the findings reported by Zhang et al. [9], who demonstrated that the deletion of Eomes in NCR1-iCre^+/− mice resulted in significant blockade of early development and a marked reduction in the population of mature NK cells. We hypothesize that these differences may arise from the distinct promoters used in the two different lines of Ncr1-Cre mice employed in these studies. Specifically, in the Ncr1-Cre mouse line used in our laboratory, the mini-Ncr1 promoter is utilized, primarily targeting late-stage NK cells, particularly CD11b⁺ NK cells, for genetic deletion (Fig. S8). In contrast, Zhang et al. utilized knock-in mice in which gene deletion was initiated during the early stages of NK cell development. Thus, the expression of Eomes in NK cells plays distinct regulatory roles at distinct developmental stages. During or before the NK progenitor stage, the expression of Eomes facilitates the commitment to the NK cell lineage and subsequent maturation processes. However, once the NK cell lineage is established, the expression of Eomes in both immature and mature NK cells serves to prevent senescence and maintain rejuvenation. Therefore, precise control of the amplitude and timing of the expression of Eomes is crucial for dynamically modulating NK cell maturation.

We also confirmed that there is reciprocal regulation of Eomes and T-bet expression at the translational level. In contrast to the deletion of Eomes, the deletion of T-bet at the immature natural killer (iNK) cell stage led to the accumulation of double-positive (DP) NK cells, suggesting that T-bet expression tends to promote the maturation of NK cells. Hence, the spatiotemporal balance between Eomes and T-bet expression serves as a crucial rheostat for controlling NK cell maturation. Eomes and T-bet synergistically regulate the development of NK cells, and even subtle alterations in this balance can exert significant effects on NK cell development and maturation.

Although the importance of Eomes in NK cell homeostasis and function is well recognized, its direct target genes remain to be identified. Through a comprehensive analysis of RNA-seq and ChIP-seq data, we discovered that KLF2 is a target gene induced by Eomes. Selective disruption of Eomes resulted in a reduction in the transcript level of KLF2. Notably, KLF2 was previously reported to play a crucial role in maintaining homeostasis across various lymphocyte compartments, including quiescent B cells [14–16], T cells [17, 18], and NK cells [13]. Eomes-deficient NK cells and KLF2-deficient NK cells have similar phenotypes. Analysis of ChIP-seq data revealed that Eomes binds to the promoter of Klf2, thereby promoting its expression. Furthermore, the overexpression of KLF2 partially compensated for the deficiency of Eomes in NK cells. In line with previous reports, we also observed that T-bet can induce Zeb2 expression. The depletion of either T-bet or Zeb2 in NK cells resulted in the arrest of NK cell maturation at the CD27⁺CD11b⁺ DP stage. Interestingly, transgenic expression of Zeb2 in Tbx21-deficient NK cells partially restored the normal maturation process [30]. Our RNA-seq analysis revealed a reduction in Zeb2 expression upon T-bet deletion, whereas deletion of Eomes resulted in the induction of Zeb2 expression. Moreover, ChIP-seq analysis demonstrated that Eomes can bind to the Tbx21 promoter region, while T-bet binds to highly conserved T-box sequences in the Zeb2 gene, as observed in CD8⁺ T cells [31]. However, Eomes did not directly bind to the Zeb2 promoter region. Collectively, these results indicate that T-bet regulates NK cell development by controlling the expression of its direct target gene, Zeb2. Additionally, Eomes inhibits Zeb2 expression by directly targeting Tbx21 and inhibiting its expression. Due to the limited amounts of NK cell subsets at each developmental stage, we utilized the total NK cell population in the ChIP-seq experiment mentioned above, precluding the use of individual NK cell subsets for RNA-seq analysis. Hence, we are currently unable to determine whether the binding of Eomes to Klf2 and Tbx21 is stage specific, and further experimental verification is required.

In summary, our study elucidates the stage-specific role of Eomes in the specification of the NK cell lineage. These findings further underscore the coordination of NK cell development by Eomes and T-bet, despite their binding to largely overlapping genomic sites. Interestingly, these transcription factors appear to regulate distinct gene sets, a phenomenon that may be attributed to their distinct temporal activity profiles. Additionally, our study highlights the importance of employing appropriate tools and methodologies to fully reveal the spatiotemporal expression patterns of genes for understanding the homeostasis and functional aspects of immune cells.

Materials and methods

Mice

C57BL/6 (B6) mice and CRISPR–dCas9-activator transgenic mice were obtained from the animal facilities of Tsinghua University. CD122^Cre/+ and Ncr1^Cre/+ mice were generated in our laboratory. β2m-Deficient mice, Rag1^-/-γc^- mice, R26^stop-YFP mice, Eomes^f/f mice, CD45.1 mice and Tbx21^f/f mice were purchased from The Jackson Laboratory. Cas9 transgenic mice were a gift from Min Peng (Tsinghua University, Beijing, China) [20]. All 8-12-week-old mice used in the experiments were on a C57BL/6 background and were housed under specific pathogen-free conditions on a 12/12-hour light-dark cycle at an ambient temperature of 20-22 °C. Ethical approval for all experimental procedures involving mice was obtained from the Animal Ethics Committee of Tsinghua University.

Cell lines

B16F10 cells were cultured in DMEM (HyClone) supplemented with 10% FBS (Gibco). RMA-S cells and RMA cells were cultured in RPMI-1640 medium (HyClone) supplemented with 10% FBS. All cells were maintained at 37 °C in a 5% CO2 incubator.

Flow cytometry

Flow cytometric analysis was performed using a BD Fortessa, a five-laser cell analyzer manufactured by BD Biosciences. Monoclonal antibodies specific for various mouse markers, including CD3e (eBio500A2), NK1.1 (PK136), CD117 (2B8), CD127 (A7R34), Ly49A (A1), Ly49C/I (5E6), Ly49H (3D10), Ly49G2 (eBio4D11), Ly49D (eBio4E5), CD11b (M1/70), CD27 (LG.7F9), NKG2A (16a11), NKG2D (CX5), CD43 (1B11), Eomes (Dan11mag), T-bet (eBio4B10), KLRG1 (2F1), and CD122 (TM-betal), and the appropriate isotype controls were utilized in this study. The antibodies were procured from eBioscience (San Diego, CA) or BD Biosciences (Mississauga, Ontario, Canada). To analyze surface markers, cells were stained with the indicated antibodies diluted in PBS containing 2% fetal bovine serum. Transcription factor analysis was performed by fixing and permeabilizing the cells using a Foxp3/Transcription Factor Staining Buffer Set, followed by staining for intracellular molecules. FlowJo software (Tree Star) was used for data analysis. The expression level of each marker is presented as a proportion or as the net mean fluorescence intensity (ΔMFI), which was determined by subtracting the mean fluorescence intensity of the respective isotype control from the mean fluorescence intensity of the marker.

In vivo splenocyte rejection assay

Splenocytes derived from WT or β2m^–/– mice were subjected to red blood cell depletion using Ficoll-Hypaque density gradient centrifugation. Subsequently, the splenocytes obtained from β2m^–/– mice were labeled with 5 µM CFSE, while the splenocytes obtained from WT mice were labeled with 0.5 µM CFSE (a concentration tenfold lower than that used to label β2m^–/– splenocytes) obtained from Molecular Probes. These two types of CFSE-labeled splenocytes were mixed at a 1:1 ratio. Aliquots of the resulting mixture containing 2 × 10⁶ splenocytes were intravenously injected into mice pretreated with poly(I:C) at a dose of 10 µg/mg. After eighteen hours, flow cytometry was used to assess the presence of CFSE-positive cells in the spleen and lymph nodes.

In vivo RMA-S clearance assay

Mice were pretreated with poly(I:C) at a dose of 10 µg/mg and then intraperitoneally injected with a mixture of target cells consisting of NK-sensitive RMA-S cells expressing GFP (10⁶ cells) and NK-insensitive RMA cells expressing Ds-Red (10⁶ cells). Eighteen hours post-injection, the mice were euthanized, and cells in the peritoneal cavity were collected through repeated washes with PBS supplemented with 2 mM EDTA. Flow cytometry was used to determine the relative percentages of RMA-S and RMA cells.

Mouse model of B16 melanoma lung metastasis

In this study, B16F10-luciferase melanoma cells in the logarithmic growth phase were resuspended in 1× Hank’s balanced salt solution (HBSS) and intravenously injected into mice (2 × 10⁵ cells/mouse). After a two-week incubation period, the mice were euthanized, and their lungs were excised and weighed. The surface nodules on the lungs were then meticulously counted using a dissecting microscope.

RNA-seq analysis

NK cells were isolated from the spleen through an enrichment process using a PE-conjugated anti-CD49b antibody and magnetic sorting (Miltenyi), and the enriched splenic NK cells were subsequently sorted using flow cytometry to obtain the CD3⁻NK1.1⁺ cell population. A total of 2×10⁴ purified NK cells were then transferred to RNeasy lysis buffer, and total RNA was extracted using an RNeasy Micro Kit (QIAGEN) following the manufacturer’s instructions. To construct libraries, the SMART-seq Kit V2 Pico Input RNA Kit (Takara) was utilized, and the resulting libraries were subjected to 76 cycles of sequencing on the NextSeq 500 platform. The sequencing data were processed using a previously described methodology [32].

Bone marrow reconstitution

CD45.1⁺ WT, Eomes^f/f and Eomes^f/f/CD122^Cre/+ mice were treated with 5-fluorouracil (5-FU, 150 mg/kg), and bone marrow cells were collected after four days. For competitive transfer, a 1:1 mixture of 2 × 10⁵ Lin⁻CD122⁺NK1.1⁻ NKp cells from Eomes^f/f or Eomes^f/f/CD122^Cre/+ CD45.2 mice and cells from CD45.1 mice was transferred into irradiated Rag1^-/-γc^- recipient mice. After eight weeks, the development of NK cells was evaluated by flow cytometry. During the process of bone marrow reconstitution, the recipient mice were administered water containing an antibiotic (neomycin, 1 mg/ml) twice a week for two weeks.

Chromatin immunoprecipitation and sequencing

CD3⁻NK1.1⁺ cells were isolated from the spleens of WT mice through cell sorting. Subsequently, ChIP (chromatin immunoprecipitation) was conducted following the ChIPmentation protocol [33]. Briefly, cells were fixed in 1% formaldehyde at room temperature for 15 min, followed by quenching of the crosslinking reaction with 0.125 M glycine and sonication. The chromatin was then incubated overnight at 4 °C with 5 µg of a rabbit monoclonal anti-Eomes antibody (sc-69269, Santa Cruz), rabbit monoclonal anti-T-bet antibody (sc-21749 X, Santa Cruz), rabbit polyclonal anti-H3K4me3 antibody (ab8580, Abcam) or rabbit polyclonal anti-H3K27ac antibody (ab4729, Abcam). Immune complexes were isolated using protein A agarose beads (Thermo Fisher Scientific, Waltham, MA). Crosslinking was reversed by heating to 65 °C. After purification, the DNA was sent for sequencing (Novogene), and quantitative PCR of the ChIP products was performed using gene-specific primers and SYBR® Premix Ex Taq (RR420A, Takara). The primers utilized were as follows: Klf2 site 1 (forward: TCGTCACCCGTGTTCCCCGC; reverse: CCGCCCCCGCGGCCGGGCCCCGCC), Klf2 Site2 (forward: GGGGCGGGGCCCGGCCGCGGGGGCGG; reverse: GGGACGAGC TCCGGGCTCAGCCTAA), Klf2 site 3 (forward: GCGCAGTCCGGGCTCCCGCAGTA GCCGC; reverse: GGCACAGAGGGCCGGGCTAGGAGG), Tbx21 site 1 (forward: GTGTAAAGTCACAAAGCCTGGGC; reverse: AGTCCTGGAAGGCACAGCAAAGGC), Tbx21 site 2 (forward: GGCGGAACTTCCTGGGGGAGAGA; reverse: GAATTCGCT TTTGGTGAGGACTGAA), and Tbx21 site 3 (forward: ACACTTAGGAGTGGGGGTCGG; reverse: GGCACAGAGGGCCGGGCTAGGAGG). Fold differences in expression were calculated using the 2^-ΔΔCT method.

Retrovirus production and transduction

Virus packaging was performed in HEK293T cells cultured in DMEM (HyClone) supplemented with 10% FBS (Gibco) and penicillin and streptomycin (Sigma‒Aldrich) except during transfection. HEK293T cells were transfected with the indicated plasmid and the pCL-Eco helper plasmid (Addgene #12371) at 50–60% confluence using a polyethyleneimine (PEI)-based (Sigma‒Aldrich) transfection system. After 24 h, the medium was changed to regular medium containing antibiotics. Forty-eight hours post-transfection, the virus-containing cell culture medium was harvested and filtered through a 0.45 μM filter. Bone marrow cells were subjected to three rounds of transduction at 12-h intervals with virus-containing medium diluted 1:1 with DMEM containing 10% FBS, penicillin, streptomycin, and polybrene (4.5 μg/mL).

Bone marrow reconstitution with KLF2 overexpression

The bone marrow reconstitution assay was conducted following a previously established protocol, with certain modifications [34]. Briefly, the indicated mice were intravenously injected with 5-fluorouracil (5-FU). Four days later, mouse bone marrow cells were collected and then infected with the retrovirus MSCV expressing the mouse KLF2 gene. Subsequently, the infected bone marrow cells were intravenously injected into recipient mice that had received sublethal irradiation. After eight weeks, bone marrow reconstitution in the recipients was evaluated using flow cytometry.

Bone marrow reconstitution with CRISPR-Cas9-based knockout of KLF2

To construct a retroviral vector capable of expressing a single-guide RNA (sgRNA) along with the Thy1.1 marker, the hU6-sgRNA-EF1α-Cas9-P2A-puro expression cassette derived from lentiCRISPRv2 (Addgene #52961) was inserted into the pMSCV backbone (Addgene #74056). Subsequently, the Cas9-P2A-puro cassette was replaced with the Thy1.1 cassette, resulting in the construction of a modified vector designated pMSCV-sgRNA-Thy1.1, as previously described [20, 34]. The specific sequences of the CRISPR sgRNAs used were as follows: sgNontargeting (sgControl), TTCGCACGATTGCACCTTGG; sgRNA targeting Klf2, TTGTTGACGCCGTCGCCATA; sgRNA targeting Klf2, AACAGCATAGTTAGCGCGCG; and sgRNA targeting Klf2, ACCGCCAGGCTTATATACCG. These sgRNA sequences were inserted into the pMSCV-sgRNA-Thy1.1 plasmid to facilitate targeted gene knockout. For the experiment, first, Cas9 transgenic mice were treated with 5-FU, and bone marrow cells were collected four days later. These cells were subjected to spin infection with MSCV retroviruses expressing the indicated sgRNAs. The infected bone marrow cells were subsequently transferred into recipient mice that had undergone sublethal irradiation. Flow cytometric analysis was conducted eight weeks after the transfer to evaluate the success of bone marrow reconstitution in the recipient mice.

Bone marrow reconstitution with CRISPR-dCas9-mediated activation of KLF2 expression

For construction of the dCas9 transcriptional activation system, the pMSCV-sgRNA-Thy1.1 vector served as the backbone, with Thy1.1 substituted with the BFP-P2A-Cre cassette. This modified construct was denoted pMSCV-sgRNA-BFP, as previously described [21, 34]. The following sgRNA sequences were used: sgRNA targeting Klf2, CGTCGCCATAGGGACGGTCG; sgRNA targeting Klf2, ACAGCATAGTTAGCGCGCGC; and sgRNA targeting Klf2, ACCGCCAGGCTTATATACCG. The pMSCV-sgRNA-BFP vector containing the corresponding sgRNA sequence was utilized for targeted gene activation. For the experiment, first, CRISPR-dCas9-activator transgenic mice were treated with 5-FU, and bone marrow cells were collected four days later. These cells were then subjected to spin infection with MSCV retroviruses expressing the indicated sgRNAs. The infected bone marrow cells were subsequently transferred into recipient mice that had undergone sublethal irradiation. After eight weeks, bone marrow reconstitution in the recipients was assessed by flow cytometry.

Statistical analyses

The statistical analyses were conducted with GraphPad Prism 8 software. To evaluate the significance of differences, a two-tailed t test was utilized after confirmation of equal variance. A significance level of <0.05 was used to establish statistical significance. The following notations were used to indicate the level of significance: *P < 0.05, **P < 0.01, and ***P < 0.001.

Author contributions

JH designed and conducted the experiments, analyzed the data, and wrote the manuscript. DC and WX performed most of the bioinformatics analyses. XH and YQ performed the flow cytometry experiments. MY and ZD designed the study, supervised the research, and revised the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon request.

Competing interests

The authors declare no competing interests. ZD is an editorial board member of Cellular & Molecular Immunology, but he has not been involved in the peer review of or decision-making regarding the article.

Ethics

The animal study was reviewed and approved by Tsinghua University.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-024-01164-8.