Transcription factor Zhx2 restricts NK cell maturation and suppresses their antitumor immunity

Natural killer (NK) cells contribute to the host’s first line of tumor immunosurveillance with the ability to identify and eliminate transformed cells through the release of cytotoxic granules containing perforin and granzymes (Huntington et al., 2007b; Vivier et al., 2008). Harnessing NK cell effector function represents a critical immunotherapeutic approach to cancer (Guillerey et al., 2016). Although NK cells are set up throughout development to be potent killers, their roles can be subverted at the tumor site (Mamessier et al., 2011; Sconocchia et al., 2012). Tumor-infiltrating NK (TINK) cells frequently exhibit a dysfunctional phenotype, express decreased levels of activating receptors and increased levels of inhibitory receptors, produce lesser amounts of cytokines such as IFN-γ, and exert lower cytotoxicity. Both animal data and clinical evidence suggest the loss of functional NK cells and the accumulation of inactive and immature NK cells in the tumor microenvironment (Jacobs et al., 2001; Platonova et al., 2011; Richards et al., 2006). A better understanding of the molecular regulatory mechanisms that control NK cell maturation and survival is crucial for the development of NK cell–based immunotherapies.

NK cells develop mainly in the bone marrow (BM). After the acquisition of the IL-15 receptor β chain (CD122), followed by the expression of NK1.1, NK cells continue a late maturation program that can be further classified based on the surface expression of CD11b and CD27: CD11b⁻CD27⁺, CD11b⁺CD27⁺, and CD11b⁺CD27⁻ (Chiossone et al., 2009; Kim et al., 2002). During maturation, NK cells gradually increase CD11b expression, decrease CD27 expression, maintain a balance between the expression of activating and inhibitory receptors, progressively increase their cytotoxic capacity but decrease their potential for homeostatic expansion, and become prone to apoptosis (White et al., 2017). NK cell development and functional maturation are triggered by plenty of extracellular signals (i.e., cytokines), among which IL-15 is critical for both NK cell lineage commitment and terminal maturation (Huntington et al., 2007a). In addition, NK cell development is dictated by a series of transcription factors (TFs) that promote the expression of genes coding for effector molecules and surface markers of maturation (Brillantes and Beaulieu, 2019). For instance, Id2 (Constantinides et al., 2014; Delconte et al., 2016), STAT5 (Marçais et al., 2014; Vargas-Hernández et al., 2020), Tox (Vong et al., 2014), Ets1 (Barton et al., 1998; Taveirne et al., 2020), and Nfil3 (Seillet et al., 2014) regulate early stages of NK development, whereas concerted actions of T-bet and Zeb2 determine terminal NK cell maturation and survival (Gordon et al., 2012; van Helden et al., 2015). Although huge advances have been achieved in transcriptional regulation of NK development, negative regulators in the process of NK cell maturation are still largely unknown.

Zinc fingers and homeoboxes 2 (Zhx2), one member of the ZHX family, is identified as a ubiquitous transcriptional repressor (Liu et al., 2015). Zhx2 has attracted considerable attention due to its involvement in tumorigenesis, cell differentiation, and metabolism-related diseases (Wu et al., 2020b; You et al., 2020; Yue et al., 2012), whereas its function in the immune system is mostly overlooked. Zhx2 is abundantly expressed in the thymus and spleen and has been reported to play a role in B cell development and macrophage polarization (Nagel et al., 2015; Wang et al., 2020). By manipulating macrophage activation and differentiation, Zhx2 participates in inflammation-related diseases such as atherosclerosis and sepsis (Erbilgin et al., 2018; Wang et al., 2020). These findings indicate Zhx2 as a potentially interesting gene in immune cell differentiation. However, its function in NK cells is unclear.

Here, we identified Zhx2 as a novel regulator of NK cells that restricts NK cell maturation, survival programs, and effector functions. Mechanically, Zhx2 lowered NK cell responses to IL-15 and transcriptionally repressed Zeb2 expression. Deletion of Zhx2 in NK cells markedly enhanced the formation of mature NK cells and promoted curative antitumor immunity of NK cell adoptive immunotherapy. These data provide new insight into the regulatory mechanisms of NK cell maturation and reveal novel opportunities for NK cell–based immunotherapy.

We first detailed the Zhx2 expression pattern in immune cells using a publicly available database. Notably, among members of the ZHX family, ZHX2 was highly expressed in lymphocytes (Zhao et al., 2020; Fig. S1 A) and NK cells isolated from different tissues (Dogra et al., 2020; Fig. S1 B), indicating its role in NK cells.

To study the role of Zhx2 in NK cells, we crossed Ncr1-Cre mice (gift from Prof. Zhongjun Dong, Institute for Immunology, Tsinghua University, Beijing, China) with mice carrying floxed Zhx2 alleles (gift from Prof. Brett T. Spear, Department of Microbiology, Immunology and Molecular Genetics, University of Kentucky, Lexington, KY) to generate an NK cell–specific knockout mouse referred to as the Zhx2^Δ/Δ mouse. Western blot analysis confirmed the deletion of Zhx2 in splenic NK cells (Fig. S1 C). Compared with WT (Zhx2^+/+) mice, a two- to threefold increase of CD3⁻NK1.1⁺CD49b⁺ NK cells was observed in Zhx2^Δ/Δ mice in the liver, spleen, and blood in terms of both relative ratios and absolute cell counts (Fig. 1, A and B). Similar results were seen when NK cells were gated as CD3⁻NK1.1⁺ cells (Fig. S1, D and E). We then examined the role of Zhx2 in NK cell maturation. As shown in Fig. 1, C–E, and Fig. S1, F and G, NK cell–specific Zhx2 depletion led to an overall increase of the most mature CD27⁻CD11b⁺ NK cell population (CD27^low) in the liver, spleen, blood, and BM in both relative proportions and absolute cell numbers. The frequency of the KLRG1⁺ mature NK subsets (pregated on CD3⁻NK1.1⁺CD49b⁺ cells) was also significantly higher in the liver and spleen from Zhx2^Δ/Δ mice than in those from Zhx2^+/+ mice (Fig. 1 F), suggesting a more mature phenotype of NK cells (Huntington et al., 2007a). To further verify the role of Zhx2 in NK cell maturation, an in vitro NK cell differentiation system was established with NK cells from tamoxifen-induced Zhx2 knockout mice (Cag-Zhx2^Δ/Δ, obtained by crossing Zhx2^fl/fl mice to the Cag^cre strain) and control Cag-Zhx2^+/+ mice (Fig. 1 G). Briefly, spleen NK cells were purified from Cag-Zhx2^+/+ and Cag-Zhx2^Δ/Δ mice that received a 3-d i.p. injection of tamoxifen (T5648; Sigma-Aldrich) for in vitro culture for 5 d with 20 ng/ml IL-2 and 20 ng/ml IL-15. Western blot analysis confirmed the tamoxifen-induced deletion of Zhx2 in splenic NK cells of Cag-Zhx2^Δ/Δ mice (Fig. S1 H). Flow cytometry (FCM) results showed that, although the input NK cells from Cag-Zhx2^+/+ and Cag-Zhx2^Δ/Δ mice at day 0 showed no difference in NK cell maturation (Fig. S1 I), Zhx2 depletion in NK cells induced a greater proportion of the matured CD27^low NK population after 5-d IL-15 stimulation (Fig. 1 H). Moreover, NK precursor (NKP) cells, which were characterized as Lin⁻CD244⁺CD27⁺CD122⁺NK1.1⁻ (Tang et al., 2012), were unaltered in Cag-Zhx2^Δ/Δ mice, excluding the role of Zhx2 in early steps of NK cell development (Fig. S1, J and K). Collectively, all these results suggest that Zhx2 restricts NK cell terminal maturation.

To explore whether Zhx2 is cell intrinsically required for NK cell maturation, mixed BM chimera experiments were performed (Fig. 2 A). Congenic BM cells (CD45.2⁺Zhx2^Δ/Δ and CD45.1⁺Zhx2^+/+) were transferred into irradiated WT CD45.1⁺CD45.2⁺ recipients at a 1:1 ratio, and CD3⁻NK1.1⁺CD49b⁺ NK cell chimerism was assessed in the spleen 8 wk after transfer. As displayed in Fig. 2 B, CD45.2⁺Zhx2^Δ/Δ BM cells reconstituted the NK cell compartment in the spleen better than those of CD45.1⁺Zhx2^+/+ mice. Further FCM analysis showed that the increase of NK cells from Zhx2^Δ/Δ BM cells was predominately in the CD27^low NK population, which was coupled with a relative decrease of Zhx2^Δ/Δ-derived NK cells in double-positive (DP) stage (Fig. 2 C). Noncompetitive transplant, whereby CD45.2⁺Zhx2^+/+ or CD45.2⁺Zhx2^Δ/Δ BM was transferred into CD45.1⁺ WT congenic irradiated recipients (Fig. 2 D), demonstrated a similar increase in CD45.2⁺CD3⁻NK1.1⁺CD49b⁺ cells developing from the transferred Zhx2^Δ/Δ BM after 8 wk (Fig. 2 E). The percentage of CD27^low was also higher in the Zhx2^Δ/Δ group (Fig. 2 F). Thus, we conclude that Zhx2 regulates NK cell homeostasis in a cell-intrinsic manner.

NK cell maturation accompanies an enhancement of effector functions. Thus, we next assessed the functional effects of Zhx2^Δ/Δ NK cells in vitro and in vivo. FCM analyses showed significantly reduced expression of inhibitory receptor NKG2A as well as the obviously increased expression of activating receptor NKG2D on splenic Zhx2^Δ/Δ NK cells (Fig. 3 A). Upon stimulation of anti-NK1.1, hepatic and splenic CD3⁻NK1.1⁺ NK cells from Zhx2^Δ/Δ mice displayed markedly enhanced production of IFN-γ, TNF-α, CD107a, and granzyme than those from Zhx2^+/+ mice (Fig. 3 B). Real-time quantitative PCR (qPCR) with purified murine splenic NK cells verified the augmented production of effector molecules in Zhx2^Δ/Δ NK cells (Fig. 3 C). The increased effector molecules were also detected in CD27^low NK cells from Zhx2^+/+ and Zhx2^Δ/Δ mice (Fig. S2, A and B). Furthermore, splenic NK cells from control and Zhx2^Δ/Δ mice were tested for their ability to kill mouse lymphoma Yac-1 cells. Supporting the notion that Zhx2 inhibits NK functions, NK cells from Zhx2^Δ/Δ mice showed significantly higher killing activities against Yac-1 cells than NK cells from Zhx2^+/+ mice in both different effector/target (E:T) ratio and real-time cytotoxicity assays (Fig. 3, D and E). This was further validated with an in vivo killing experiment in which CFSE labeling Yac-1 cells were i.p. injected in either Zhx2^Δ/Δ or Zhx2^+/+ mice. As shown in Fig. 3 F, there were fewer remaining Yac-1 cells in Zhx2^Δ/Δ mice.

Next, we evaluated Zhx2^Δ/Δ NK cell–mediated tumor elimination in vivo by s.c. injection of mouse hepatoma cell line Hepa1-6 cells into Zhx2^Δ/Δ and Zhx2^+/+ mice, respectively. As shown in Fig. 3, G and H, larger tumors had evolved in Zhx2^+/+ mice, whereas a pronounced reduction of tumor mass was observed in Zhx2^Δ/Δ mice.

To validate the role of ZHX2 in human NK cells, ZHX2 overexpression and knockdown were performed in human NK cell line NK92 cells. As expected, overexpression of ZHX2 decreased, whereas interference of ZHX2 expression enhanced expression of functional molecules IFN-γ, TNF-α, granzyme, and perforin (Fig. S2, C and D). Also, ZHX2 inhibited the cytotoxicity of NK92 cells against target K562 cells (Fig. S2 E). Altogether, these data demonstrate that Zhx2 inhibits NK cell cytotoxicity and cytokine production.

Our above data suggest that loss of Zhx2 results in an increased number of terminal mature NK cells and subsequently augmented NK cell functions. We next investigated the mechanism of the increment of CD27^low NK cells in Zhx2^Δ/Δ mice. To address that, splenic NK cells were purified from both Zhx2^+/+ and Zhx2^Δ/Δ mice, and cell proliferation and survival were assessed. As shown in Fig. 4, A and C, both an ex vivo 5-ethynyl-2′-deoxyuridine (EdU) labeling assay and Ki67 staining did not detect any significant difference between NK cells from the two groups, indicating that Zhx2 does not influence NK cell proliferation. Also, all four stages of NK cells, including the mature CD27^low NK cells, showed no difference in EdU and Ki67 staining between Zhx2^+/+ and Zhx2^Δ/Δ mice (Fig. 4, B and D). In contrast, an annexin V/7-AAD staining assay showed an obviously reduced number of apoptotic NK cells isolated from Zhx2^Δ/Δ mice compared with those from control Zhx2^+/+ mice (Fig. 4 E). To confirm the apoptotic resistance of NK cells from Zhx2^Δ/Δ mice, the apoptosis inducer cycloheximide (CHX) was added to NK cell culture. As shown in Fig. 4 F, upon CHX treatment, in vitro cultured Zhx2^Δ/ΔNK cells underwent less apoptosis than control cells at all detected time points. In accordance, RNA sequencing (RNA-seq) and gene set enrichment analysis (GSEA) using the Broad Hallmark gene set collection demonstrated significant enrichments for a gene set encoding apoptosis-related molecules in Zhx2^+/+ NK cells relative to their expression in the Zhx2^Δ/Δ NK cell subset (Fig. 4 G).

We further asked whether Zhx2 contributes to the maintenance of NK cells in vivo. To this end, congenitally marked NK cells from Zhx2^+/+ or Zhx2^Δ/Δ mice were labeled with CellTrace Violet (CTV) or CFSE, respectively, and then mixed in a 1:1 ratio and cotransferred into either C57BL/6 WT recipient mice (Fig. 4 H) or nonobese diabetic Prkdc^em26IL2rg^em26/Gpt (NSG; NOD-Prkdc^em26IL2rg^em26/Gpt) recipient mice (Fig. 4 I), which lack all lymphocytes but have abundant IL-15. FCM results showed a selective loss of Zhx2^+/+ NK cells in recipient mice. Concurrently, Zhx2^Δ/Δ NK cells displayed increased viability compared with Zhx2^+/+ NK cells in the CD27^low subset (Fig. 4 J). These data underline the importance of Zhx2 in NK cell survival.

IL-15 plays a crucial role in NK cell biology, including cell survival, maturation, and function (Marçais et al., 2014; Waldmann, 2015). We then examined whether Zhx2 participates in IL-15 signaling in NK cells. GSEA was performed comparing RNA-seq data of splenic NK cells purified from Zhx2^+/+ and Zhx2^Δ/Δ mice to address gene expression related to IL-15 signaling. As shown in Fig. 5 A, Zhx2^Δ/Δ NK cells demonstrated enrichment for genes associated with IL-15 signaling. IL-15 accesses its downstream signaling via a common γ subunit (CD132) and an IL-15Rβ chain (CD122; Mishra et al., 2014; Waldmann, 2015). FCM did not detect any difference in the expression of CD122 in NK cells with Zhx2 deficiency (Fig. 5 B). We thus hypothesized that Zhx2 deficiency might influence IL-15–initiated phosphorylation of STAT5 and the survival kinase AKT, two major signaling events stimulated by IL-15 (Marçais et al., 2014; Waldmann, 2015; Lin et al., 2017; Wang et al., 2019). Supporting this hypothesis, after in vitro stimulation of IL-15, NK cells isolated from Zhx2^Δ/Δ mice displayed obvious increases in phosphorylated STAT5 and AKT compared with those from Zhx2^+/+ mice (Fig. 5 C). Furthermore, with the IL-15 treatment, Zhx2^Δ/Δ NK cells showed significantly enhanced effector functions displaying as increased levels of CD107a and IFN-γ. The augmented responses of Zhx2^Δ/Δ NK cells were particularly evident when IL-15 was present at a high concentration (50 ng/ml; Fig. 5 D). Consistent with the notion that IL-15 boosts bioenergetic metabolism and promotes the survival of NK cells (Kobayashi and Mattarollo, 2019), the Seahorse XF Analyzer detected higher maximum respiration in IL-15–stimulated Zhx2^Δ/Δ NK cells (Fig. 5 E), whereas FCM showed a decrease of apoptosis in IL-15–stimulated NK cells from Zhx2^Δ/Δ mice compared with those from Zhx2^+/+ mice (Fig. 5 F).

To further verify the depressing function of Zhx2 in the IL-15 pathway, the selective IL-15 signaling STAT5 inhibitor (STAT5-Inh, HY-102048; MedChemExpress) was employed. As previously reported (Juen et al., 2017), STAT5-Inh inhibited the phosphorylation of STAT5 (Fig. S3 A) and enhanced apoptosis in NK cells (Fig. 5 G). More important, treatment with STAT5-Inh not only largely rescued the IL-15–triggered up-regulation of the phosphate STAT5 level in Zhx2^Δ/Δ NK cells (Fig. S3 A) but also eliminated the difference of IL-15–induced cell apoptosis between Zhx2^+/+ and Zhx2^Δ/Δ NK cells (Fig. 5 G). Accordingly, STAT5-Inh almost completely destroyed the enhanced production of CD107a and IFN-γ expression in Zhx2^Δ/Δ NK cells (Fig. S3 B). Taken together, phenotypic, functional, and metabolic analyses concordantly demonstrate that Zhx2 restrains the NK cell response to IL-15.

Zhx2 has been demonstrated as a TF (Yue et al., 2012). To further identify the direct target of Zhx2 and elucidate the underlying mechanism by which Zhx2 regulates NK cells, RNA-seq and assay for transposase-accessible chromatin using sequencing (ATAC-seq) were performed using Zhx2^+/+ and Zhx2^Δ/Δ splenic NK cells. The volcano plot derived from RNA-seq showed that 1,642 genes were significantly altered in Zhx2^Δ/Δ NK cells compared with Zhx2^+/+ NK cells (1,325 up-regulated and 317 down-regulated; P ≤ 0.05; Fig. 6 A). To obtain more in-depth insight into the impact of Zhx2 loss on NK cell development, we used GSEA comparing transcriptional profiles of Zhx2^+/+ NK cells and Zhx2^Δ/Δ NK cells against previously defined signature genes of DP or CD27^low NK cells (Meinhardt et al., 2015). Fig. 6 B shows that the set of Zhx2^Δ/Δ NK cells was preferentially active in the CD27^lowsubset, whereas the set of Zhx2^+/+ NK cells was preferentially active for the immature DP subset, supporting the role of Zhx2 in restricting NK cell maturation.

A total of 6,426 peak genes were differentially detected in ATAC-seq of Zhx2^Δ/Δ NK cells compared with Zhx2^+/+ NK cells. We centered all the differential ATAC peaks according to the transcription start sites (TSSs) of the closest genes, and the resulting heatmap of the average ATAC-seq signals shows an increased openness in the Zhx2^Δ/Δ strains around the TSS with a peak just upstream of the TSS (Fig. S4 A). In gene ontology (GO) analyses, immune response and inflammatory response were significantly overrepresented, further supporting the involvement of Zhx2 in NK cell responses (Fig. S4 B). These data indicated that the lack of Zhx2 prompted an altered transcriptional network to become operational in NK cells.

To identify the direct target of Zhx2, we performed a cluster analysis of sharing genes in ATAC-seq and RNA-seq data from Zhx2^+/+ and Zhx2^Δ/Δ splenic NK cells, as well as reported CD27^low/DP NK represented genes (Meinhardt et al., 2015). As shown in Fig. 6 C, most genes were specific to only one or two clusters. However, nine genes changed in all three clusters, which are believed to be Zhx2-regulated genes in NK cell maturation. The Zhx2-mediated regulation of these nine genes was further verified with real-time qPCR with purified splenic CD3⁻NK1.1⁺ NK cells from Zhx2^Δ/Δ and Zhx2^+/+ mice (Fig. S4 C). We then compared the expression levels of these nine genes during NK cell development using reported data (Meinhardt et al., 2015). Very interestingly, consistent with the role of Zhx2 in limiting NK cell maturation, two (Tcf7, Slamf6) of three genes, which were decreased in Zhx2-deficient cells, were gradually decreasing during NK cell development, whereas Zhx2 down-regulated genes, including Itgam, the encoding gene of well-known maturation marker CD11b, as well as Emilin2, Fbxl2, and Zeb2, showed their highest expression in the matured CD27^low NK cells (Fig. S4 D). Notably, among six of the Zhx2 down-regulated targets, Zeb2 was the most suppressed TF (Fig. S4 C).

Researchers in a previous study reported that Zeb2 is indispensable for the survival of mature NK cells and essential for CD27^low NK cell maturation (van Helden et al., 2015), which coincides with the role of Zhx2 in NK cells. To verify Zeb2 as the direct target of Zhx2, several molecular biology methods were employed. Real-time qPCR showed the negative correlation of Zeb2 and Zhx2 mRNA expression in NK cells isolated from Zhx2^Δ/Δ mice and Zhx2^+/+ mice (Fig. 6 D). In accordance with this, cotransfection and a dual luciferase assay demonstrated that ZHX2 overexpression markedly inhibited ZEB2 promoter activity in NK92 cells (Fig. 6 E). Furthermore, a chromatin immunoprecipitation (ChIP) assay performed with anti-ZHX2 and NK92 cell lysate demonstrated that ZHX2 significantly occupied with ZEB2 promoter in NK92 cells (Fig. 6 F). The Zhx2-mediated transcription inhibition of Zeb2 was further confirmed with GSEA comparing our RNA-seq data and reported data (Dominguez et al., 2015). As shown in Fig. 6 G, Zeb2–up-regulated transcripts were greatly enriched in Zhx2^Δ/Δ subsets (Fig. 6 G). Consistently, ATAC-seq data showed a higher accessibility of the Zeb2 promoter region in Zhx2^Δ/Δ NK cells than that in Zhx2^+/+ NK cells (Fig. 6 H). All these results suggested that Zhx2 is capable of binding to the Zeb2 locus, thus regulating Zeb2 transcription.

To validate the involvement of Zeb2 in Zhx2-mediated regulation of NK cell maturation and function, Zeb2 interference was performed by using a lentivirus in splenic NK cells from Zhx2^+/+ and Zhx2^Δ/Δ mice, and the splenic NK cells were then transferred to irradiated host C57BL/6 WT mice. As shown in Fig. 6, I and J, Zeb2 knockdown significantly dampened the enrichment of the CD27^low NK subset and augmented CD107a expression in the Zhx2^Δ/Δ group. The interference efficiency of Zeb2 was detected by Western blot analysis (Fig. S4 E). All these data illustrate that Zhx2 inhibits NK cell maturation and function by binding to the Zeb2 promoter, thus leading transcriptional repression of Zeb2.

It has been well established in previous studies that blockade of NK cell maturation impairs NK cells’ antitumor potential (Platonova et al., 2011; Richards et al., 2006). We wondered whether ZHX2 participates in this process. To address this, we analyzed The Cancer Genome Atlas (TCGA) hepatocellular carcinoma (HCC) dataset for NK cell abundance using specific gene sets, including GZMH, GZMB, KLRD1, PRF1, NCR1, NMUR1, GNLY, CD160, and TBX21, for identifying NK cells (https://xcell.ucsf.edu/; Aran et al., 2017). As shown in Fig. 7 A, high ZHX2 expression in NK subsets correlated with a low frequency of TINK cells, and, vice versa, tumors with more TINK cells displayed less ZHX2 expression. Furthermore, patients with low ZHX2 expression showed better survival than those with high ZHX2 in NK cells (Fig. 7 B).

To define whether Zhx2 participates in the regulation of TINK cells, a murine model of orthotopically HCC was created, and NK cells were harvested for FCM (Fig. S5 A). As shown in Fig. S5 B, the percentage of CD3⁻NK1.1⁺ TINK cells from Zhx2^Δ/Δ mice was significantly higher than that in the Zhx2^+/+ group. Moreover, terminally mature NK cell subsets (CD27^low) displayed an increase in tumors of Zhx2^Δ/Δ mice (Fig. S5 C). Also, TINK cells from the Zhx2^Δ/Δ mice had a marked increase in expression of the effector molecule CD107a (Fig. S5 D). These findings demonstrated that Zhx2 negatively regulates the function and maturation of NK cells in the tumor microenvironment, which is strictly related to tumor immune evasion.

We then sought to determine whether the deletion of Zhx2 would generate more functional and long-survival NK cells to protect against cancer. To address that, both an s.c. hepatoma homograft and a lung metastasis model were performed to assess the antitumor capacity of Zhx2^Δ/Δ NK cells. As shown in Fig. 7, C–E, transfer of splenic NK cells from Zhx2^Δ/Δ mice greatly suppressed the growth of Hepa1-6 s.c. homograft. Also, Zhx2-deficient NK cells significantly reduced lung metastasis of B16-F10 melanoma (Fig. 7, F–H). The number of metastatic nodules in the lung was significantly reduced in mice that received transfer of Zhx2^Δ/Δ splenic NK cells (Fig. 7, F and G), and the overall survival of the mice treated with Zhx2^Δ/Δ NK cells was also significantly longer (Fig. 7 H).

To verify whether ZHX2 could be a potential intervention target with the aim of enhancing NK cell immunosurveillance in tumor therapy, two human HCC models were set up in NSG mice. First, HepG2-luciferase cells (5 × 10⁶) were i.p. injected, and, in the meantime, 2 × 10⁶ NK92 cells infected with lentiviral vector expressing negative control shRNA (LV-shNC) or lentiviral vector expressing shRNA agaist ZHX2 (LV-shZHX2) were transferred (Fig. 7 I). The results of a D-luciferin–based bioluminescence assay showed that transfer of NK92 cells infected with LV-shZHX2 greatly suppressed the growth of peritoneal hepatoma xenografts across time (Fig. 7 J). The overall survival of the LV-shZHX2 group was also significantly longer (Fig. 7 K). Similar results were obtained with an s.c. HCC xenograft model. Briefly, human HCC cell line Huh7 cells (5 × 10⁶) were s.c. injected into the axilla of NSG mice, and, 1 wk afterward, 2 × 10⁶ indicated NK92 cells were transfused via the caudal vein (Fig. 7 L). As shown in Fig. 7, L and M, significantly smaller tumors were found in the mice that received LV-shZHX2 NK92 cell treatment. These findings highlight the therapeutic potential of antitumor immunity of ZHX2-deficient NK cells. ZHX2 may be a new target based on the adoptive transfer of NK cells.

NK cells are innate immune cells with immunosurveillance and immunoregulatory functions. The development and maturation of NK cells are harmoniously controlled by a complex transcriptional regulatory network (Brillantes and Beaulieu, 2019). In the present study, we identified Zhx2 as a novel negative regulator of NK cell maturation. In the absence of Zhx2, NK cells are driven toward terminal differentiation and are resistant to apoptosis. Mechanically, Zhx2 controls IL-15 signaling and transcription of Zeb2, a TF identified as a major driver of CD27^low NK cell differentiation (Dominguez et al., 2015). Our data add a novel mechanism to TF-mediated NK cell maturation.

Zhx2 is an understudied protein that has been ascribed to have a transcriptional function on the basis of tumor suppressor (Liu et al., 2015; Yue et al., 2012). Literature has shown Zhx2 functions in the immune system, including macrophage regulation and B cell development (Hystad et al., 2007; Nagel et al., 2015; Nagel et al., 2014; Wang et al., 2020; Wu et al., 2020a). Our results highlight the role of Zhx2 in the restriction of NK cell maturation. We found that Zhx2 was abundantly expressed in NK cells. With the absence of Zhx2, NK cells exhibited an intrinsic enhanced maturation mainly affecting the number of the most mature subsets. Concurrently, Zhx2 loss promoted the NK cell receptor repertoire and enhanced NK cell cytotoxicity and cytokine production. Although Zhx2 has been well recognized as a TF controlling cell proliferation–related genes in tumor cells (Shen et al., 2008; Yue et al., 2012), our ex vivo results and RNA-seq data showed that Zhx2 did not affect NK cell proliferation. Notably, our results showed that Zhx2 knockout promotes NK cell sensitivity to IL-15, the crucial cytokine maintaining NK cell proliferation (Marçais et al., 2014); however, we did not find differences in IL-15–stimulated proliferation of Zhx2^+/+ and Zhx2^Δ/Δ NK cells. In fact, Zhx2-decifient NK cells displayed reduced apoptosis both in vivo and in vitro in the presence of IL-15 stimulation, leading to prolonged survival. This is consistent with a previous report showing increased apoptosis in Zhx2-null macrophages (Erbilgin et al., 2018). Both in vitro and in vivo data demonstrated that Zhx2 mainly restricts the terminal maturation of NK cells. In accordance with this, Zhx2 expression level gradually decreased during NK cell maturation, whereas mature CD27^low NK cells displayed a minimum level of Zhx2 (data not shown), indicating the requirement of Zhx2 suppression for NK maturation.

We have defined two pathways whereby Zhx2 controls NK cell maturation and function. One is that Zhx2 down-regulates NK cell responses to IL-15, the cytokine crucial for NK cell development and survival (Wang et al., 2019; Wang and Zhao, 2021). Deletion of Zhx2 in NK cells resulted in increased activity of AKT and STAT5, two major signaling pathways of IL-15. Treatment with STAT5-Inh eliminated the prolonged survival and enhanced cytotoxicity in Zhx2-deficient NK cells. Interestingly, consistent with the role of IL-15 in regulating NK cell metabolism (Marçais et al., 2014; Wang et al., 2018) and with recently reported data showing enrichment of glycolysis-related genes in Zhx2-expressing macrophages (Wang et al., 2020), our data demonstrated that Zhx2 loss increased the metabolic rate in NK cells and that the increased metabolic rate negatively affected NK cell survival. This highlights the importance of Zhx2 in maintaining NK cell metabolism, which should contribute to its role in controlling NK cell development and maintenance of function.

Second, we identified that the TF Zeb2 is required for Zhx2-mediated restriction of NK cell maturation. Zeb2 is expressed almost exclusively in the terminally differentiated effector and effector memory populations of CD8⁺ T cells (Omilusik et al., 2015). This correlation of Zeb2 expression with the degree of differentiation has also been reported in NK cells (Dominguez et al., 2015). Zhx2 and Zeb2 expression inversely correlated with each other in NK cells. ATAC-seq, a ChIP assay, and dual luciferase report assays revealed that Zhx2 occupies a binding site upstream of the TSS of Zeb2 and represses Zeb2 transcription in NK cells. Interference of Zeb2 destroys Zhx2-mediated repression of NK cell maturation and function. Therefore, we place Zeb2 as a downstream target of Zhx2 in the regulatory network that promotes the formation of the terminally differentiated subset of NK cells. It has been reported that T-bet is necessary for inducing Zeb2 expression (Hu et al., 2020). We have thus analyzed T-bet expression in Zhx2-deficient NK cells. However, T-bet mRNA and protein levels were not different between Zhx2^+/+ and Zhx2^Δ/Δ NK cells (data not shown), indicating that Zhx2 regulates Zeb2 expression in a T-bet–independent manner. Moreover, because Zeb2 has been directly or indirectly implicated in modulating IL-15–mediated survival of developing NK cells (Wang et al., 2019; Wang and Zhao, 2021), it might also link Zhx2 with IL-15 signaling. As a ubiquitous TF, Zhx2 transcriptionally regulates a variety of targets, leading to diverse functions. Here, our work proves Zeb2 as a novel target of Zhx2 in NK cells, although some other mechanisms may also exist.

The critical role of Zhx2 in restricting NK cell maturation was also observed in the tumor microenvironment. Data from TCGA showed a negative correlation of level of ZHX2 expression in NK cells with the number of TINK cells, whereas orthotopic transplant of a liver tumor in Zhx2^Δ/Δ mice had an increased number of TINK cells with enhanced maturation and fortified effector functions. Accumulation of immature NK cells has been reported in different tumors (Krneta et al., 2016). Our data strengthen the involvement of Zhx2 in the dysregulation of NK cells in the tumor microenvironment. A computational network study suggested Zhx2 as one of the most regulated TFs in myeloid cells (Espinal-Enríquez et al., 2017). Defining the key molecules causing the sustained expression of Zhx2 in TINKs or interfering Zhx2 expression in NK cells should identify the potential targets and approaches for reversing dysfunction of TINK cells, which is definitely beneficial for tumor control.

In conclusion, our results advance the mechanistic understanding of Zhx2 control of NK cell maturation and effector functions. Deletion of Zhx2 not only promotes IL-15–mediated NK cell activation and maturation but also allows better cell viability, resulting in better antitumor activities. Our results reveal Zhx2 as a previously unknown NK cell checkpoint and suggest that therapeutic inhibition of Zhx2 in adoptively transferred NK cells could improve the efficacy of cancer treatments.

Ncr1^icre mice were a gift from Prof. Zhongjun Dong (Tsinghua University, Beijing, China). Cag^cre mice were purchased from The Jackson Laboratory. Zhx2^floxed mice were a gift from Prof. Brett T. Spear (University of Kentucky, Lexington, KY). CD45.1 mice were kindly provided by Dr. Xiaolong Liu (Center for Excellence in Molecular Cell Science, Chinese Academy of Sciences, Shanghai, China). C57BL/6 mice (6–8 wk of age) were purchased from the Animal Research Center of Shandong University. NSG mice were purchased from Jiangsu GemPharmatech. To generate Ncr1^icre-Zhx2^+/+(Zhx2^+/+) and Ncr1^icre-Zhx2^Δ/Δ (Zhx2^Δ/Δ) mice, Zhx2^fl/fl mice were crossed to the Ncr1^icre strain. To generate Cag^cre-Zhx2^+/+(Cag-Zhx2^+/+) and Cag^cre-Zhx2^Δ/Δ (Cag-Zhx2^Δ/Δ) mice, Zhx2^fl/fl mice were crossed to the Cag^cre strain. To activate Cre in Cag^cre mice, tamoxifen (T5648; Sigma-Aldrich) was injected i.p. at 2 mg for 3 d consecutively. The CD45.1⁺ strain was crossed to the CD45.2⁺ strain to generate the CD45.1⁺CD45.2⁺ strain. All mice were maintained under specific pathogen–free conditions, and experiments were performed with the Shandong University Laboratory Animal Center’s approval.

Human NK cell line NK92 cells were cultured in α-MEM containing 12.5% FBS, 12.5% horse serum, 0.2 mM inositol, 0.1 mM 2-ME, 0.02 mM folic acid, and 100 U/ml recombinant human IL-2. The mouse cell line Yac-1 and human chronic myelocytic leukemia K562 cells were cultured in RPMI 1640 medium plus 10% FBS. Mouse melanoma B16-F10 cells, mouse hepatoma Hepa1-6 cells, human liver cancer cell lines Huh7 and HepG2, and human embryonic kidney HEK293T cells were cultured in DMEM plus 10% FBS. Cells were purchased from the American Type Culture Collection. The cell lines were authenticated in 2017, and Mycoplasma contamination was routinely tested.

For isolation of mononuclear cells from the mouse liver, tissues were harvested, teased apart, and mashed through a nylon mesh, and then the cell suspensions were centrifuged over 40% Percoll gradient medium (GE Healthcare). Mononuclear cells were obtained with fluorescent antibodies for 30 min at 4°C. NK cells were gated as CD3⁻NK1.1⁺CD49b⁺ cells. NK cells applied to microarray analyses, transfer, cytotoxicity assays, immunofluorescence assays, and metabolic assays were purified by a magnetic-activated cell sorter (MACS) kit (Miltenyi Biotec). The purity of the NK cells was >90% for each assay.

Cell suspensions were surface labeled for human or mouse antibodies for 30 min at 4°C. FCM was performed on a Cytoflex S (Beckman Coulter) or Gallios flow cytometer and analyzed by FlowJo 10.6.2 or CytExpert 2.3.0.

For intracellular staining, freshly isolated cells were cultured in the medium at 37°C in a 5% CO₂ incubator and stimulated with 20 ng/ml mIL-15 (R&D Systems) for 2 h, then with brefeldin A (BioLegend) at a final concentration of 10 µg/ml for 4 h. After surface staining, cells were fixed with intracellular fixation buffer for 20 min, then permeabilized with permeabilization buffer for 10 min. Intracellular staining was performed with antibodies diluted into permeabilization buffer, and cells were fixed and permeabilized using a Fixation/Permeabilization Concentrate and Diluent Buffer Set following the manufacturer’s guidelines (eBioscience). For CD107a and intracellular staining, cells were incubated with anti-NK1.1 and CD107a for 4 h at 37°C in a 5% CO₂ incubator, followed by staining for extracellular markers.

Intracellular staining for p-STAT5 (Tyr694) and p-AKT (Tyr308) was performed with the Fixation/Permeabilization Concentrate and Diluent Buffer Set according to the manufacturer’s guidelines (eBioscience). Briefly, after surface staining, cells were fixed with intracellular fixation buffer for 20 min, then permeabilized with permeabilization buffer for 10 min. Intracellular phosphor staining was performed for 1 h at 4°C in the dark.

Apoptosis was analyzed by annexin V and 7-AAD double staining according to the manufacturer’s instructions (BD Biosciences).

Purified NK cells or NK92 cells were lysed in radioimmunoprecipitation assay buffer (Beyotime) on ice for 30 min. The cell lysis solution was centrifuged at 12,000 g for 10 min, and supernatants were collected. Samples were then run on 10% Bis-Tris protein gels. Separated proteins were then transferred to polyvinylidene difluoride membranes and blocked with 5% wt/vol milk at room temperature for 2 h. Membranes were then incubated with primary antibodies in 5% wt/vol BSA in Tris-buffered saline containing 0.1% Tween-20 overnight at 4°C, then incubated with HRP-conjugated secondary antibodies (Proteintech) for 1 h at room temperature.

All antibodies for immunoblot staining are shown in Table S1.

BM from sex-matched donor mice was obtained by flushing femurs and tibias. For competitive transfers, the two different BM cells were mixed at a 1:1 ratio, and a total of 1 × 10⁷ BM cells were injected i.v. into recipient mice, which had been lethally irradiated with 10 Gy.

Purified NK cells labeled with 5 µM CFSE were cultured with IL-2 (100 U/ml) and IL-15 (50 ng/ml) at 37°C in a 5% CO₂ incubator. After 3 d, NK cells were harvested and analyzed by FCM.

Cultured NK cells from Zhx2^+/+ and Zhx2^Δ/Δ mice were incubated with 0.1 nM CFSE (Beyotime) and CTV (Invitrogen), respectively. Labeled NK cells were then cocultured (1 × 10⁵ of each) in vitro with IL-2 for 5 d before analysis by FCM.

CFSE-labeled target cells were cocultured with purified NK cells with IL-2 (100 U/ml) at different E:T ratios (10:1, 5:1, 2.5:1) at 37°C in a 5% CO₂ incubator for 4 h. Then 7-AAD was added, and FCM identified lysed cells (CFSE⁺7-AAD⁺). For real-time cytotoxicity assays, the function of NK cells was monitored using a Real-Time Cell Analyzer-Multiple Plate system (PerkinElmer). This platform measures live target cells in real time. We cocultured Yac-1 cells (1 × 10⁴ cells/well) as target cells and NK cells (1–10 × 10⁴ cells/well) as effectors in 200 µl culture medium at 37°C with 5% CO₂. The cell index represented changes in electrical impedance and reflected the number of unkilled target cells. For real-time monitoring, the cell index was read automatically every 20 min. To determine NK cell cytotoxic activity, 2 × 10⁶ CFSE-labeled Yac-1 cells were i.p. injected into WT mice and Zhx2 knockout mice. Cells in the peritoneal cavity were harvested, and CFSE-labeled Yac-1 cells were measured 24 h after injection. NK cells were purified using an MACS kit.

For the extracellular assay, 2 × 10⁵ purified NK cells were pretreated with 20 ng/ml IL-15 for 12 h, then seeded in a Seahorse Bioscience culture plate coated with Cell-Tak solution (Corning) in XF Base Medium Minimal DMEM (Agilent) with 10 mM glucose and 2 mM glutamine in a non-CO₂ incubator for 1 h. The basal oxygen consumption rate and maximum respiration were measured by an XF96 Seahorse Extracellular Flux Analyzer (Agilent) following the manufacturer’s instructions.

Male mice 6–8 wk old were used for all experiments. In the s.c. transplanted tumor model, Hepa1-6 cells without and with purified CD3-NK1.1⁺/Dx5⁺ cells (ratio to Hepa1-6 cells 1:1) were inoculated in the inguinal area. The tumor sizes in each mouse were monitored using a Vernier caliper. Tumor volume was calculated by the formula 1/2ab² (a stands for the major axis and b for the minor axis). For the orthotopic liver tumor model, the s.c. Hepa1-6 tumors were dissected, cut into pieces of ∼1 mm³, and transplanted into the liver parcel of recipient mice. NSG mice were used to establish peritoneal hepatoma xenografts to assess the antitumor capacity of NK cells. To assess the killing capacity of NK cells with or without Zhx2, we injected human HepG2-luciferase cells (5 × 10⁶) i.p. and, meanwhile, transferred NK92 cells (2 × 10⁶) with LV-shNC or LV-shZhx2 lentivirus infection before treatment. IL-2 (1 × 10⁵ U; Jiangsu Kingsley Pharmaceuticals) was injected i.p. every 2 d to support NK cell survival in vivo. After 1 or 2 wk, for tumor imaging in vivo, luciferin was injected i.p. and imaged for 1 min using an IVIS Spectrum imaging system (PerkinElmer). NK cells were purified using an MACS kit. For the B16-F10 melanoma lung metastasis model, mice were injected with 1 × 10⁶ B16-F10 melanoma cells via the caudal vein, and, 7 d later, 2 × 10⁵ purified Zhx2^+/+ or Zhx2^Δ/Δ NK cells were transferred, and IL-2 (1 × 10⁵ U; Jiangsu Kingsley Pharmaceuticals) was injected i.p. every 2 d to support NK cell survival in vivo. For human HCC tumor models, we injected s.c. human HCC cell line Huh7 cells (5 × 10⁶) into the groin to establish a s.c. xenograft model. After 1 wk, we transferred NK cells (2 × 10⁶) with LV-shNC or LV-shZhx2 lentivirus infection before treatment. IL-2 (1 × 10⁵ U; Jiangsu Kingsley Pharmaceuticals) was injected i.p. every 2 d to support NK cell survival in vivo.

Total RNA was isolated using TRIzol reagent, and cDNA synthesis was performed using the RevertAid First Strand cDNA Synthesis Kit and random primers according to the manufacturer’s instructions. Real-time qPCR and PCR were performed using SYBR Green Real-Time qPCR Master Mix (TOYOBO) and 2× Taq PCR Master Mix (Tiangen). Primer pairs for target genes are shown in Table S2.

The sequencing data were filtered with SOAPnuke (v1.5.2) by (1) removing reads containing sequencing adapter, (2) removing reads whose low-quality base ratio (base quality ≤5) is >20%, and (3) removing reads whose unknown base (“N” base) ratio is >5%. Afterward, clean reads were obtained and stored in FASTQ format. Raw data were uploaded into the Sequence Read Archive (accession nos. PRJNA726006, PRJNA726357). The clean reads were mapped to the reference genome using HISAT2 (v2.0.4). Bowtie2 (v2.2.5) was applied to align the clean reads to the reference coding gene set, then the expression level of gene was calculated by RSEM (v1.2.12). Essentially, differential expression analysis was performed using the PoissonDis algorithm with false discovery rate (FDR) ≤0.001 and |Log₂Ratio| ≥ 1. To gain insight into the change of phenotype, GO (http://www.geneontology.org/) and Kyoto Encyclopedia of Gene and Genomes (https://www.kegg.jp/) enrichment analysis of annotated differently expressed genes was performed by phyper (https://stat.ethz.ch/R-manual/R-devel/library/stats/html/Hypergeometric.html) on the basis of a hypergeometric test. The significance levels of terms and pathways were corrected by Q value with a rigorous threshold (Q ≤ 0.05) by the Bonferroni correction.

Statistical significance was determined using Prism 8.3.0 (GraphPad Software). Two-tailed unpaired or paired Student’s t tests between two groups and two-way ANOVA across multiple groups were used to determine significance. The difference in overall survival was tested using log-rank tests. Data are presented as mean ± SEM. Statistical significance was reported as *, P < 0.05; **, P < 0.01; ***, P < 0.001; and n.s., no significance.

Fig. S1 shows Zhx2 cell and subcell profiling. Fig. S2 explains how Zhx2 inhibits NK cell functions. Fig. S3 illustrates how Zhx2 weakens IL-15–promoted NK cell function. Fig. S4 shows distinct transcriptome profiles and ATAC peaks in Zhx2-deficient NK cells. Fig. S5 shows that Zhx2 inhibits NK cell maturation and function in the tumor microenvironment. Table S1 lists antibodies used for FCM and Western blot analysis. Table S2 lists sequences of primers used for RT-qPCR.

The RNA-seq data and ATAC-seq data in this publication have been deposited in the National Center for Biotechnology Information Sequence Read Archive and are accessible through accession nos. PRJNA726006 and PRJNA726357.

The graphical abstract was prepared using BioRender software.

This study was supported in part by the National Natural Science Foundation of China (key program 81830017 and nos. 81902433, 81902051), the National Outstanding Youth Science Fund Project of National Natural Science Foundation of China (81425012), the Taishan Scholar Foundation of Shandong Province (tspd20181201), the Collaborative Innovation Center of Technology and Equipment for Biological Diagnosis and Therapy in Universities of Shandong, the Key Research and Development Program of Shandong (2018YFJH0503), and the National Key Research and Development Program (2018YFE0126500).

Author contributions: C. Ma and S. Tan formulated the study concept. C. Ma and S. Tan designed the studies. S. Tan, X. Guo, M. Li, T. Wang, Z. Wang, Z. Wu, and C. Li performed the experiments. S. Tan, X. Guo, and C. Ma analyzed the results. X. Liang, L. Gao, N. Li, and C. Ma interpreted the results. S. Tan, X. Liang, L. Gao, N. Li, and C. Ma wrote and edited the manuscript.