Themis2 regulates natural killer cell memory function and formation

Immunological memory is a hallmark of the adaptive immune system. Although natural killer (NK) cells are innate immune cells important for the immediate host defence, they can differentiate into memory NK cells. The molecular mechanisms controlling this differentiation are yet to be fully elucidated. Here we identify the scaffold protein Themis2 as a critical regulator of memory NK cell differentiation and function. Themis2-deficient NK cells expressing Ly49H, an activating NK receptor for the mouse cytomegalovirus (MCMV) antigen m157, show enhanced differentiation into memory NK cells and augment host protection against MCMV infection. Themis2 inhibits the effector function of NK cells after stimulation of Ly49H and multiple activating NK receptors, though not specific to memory NK cells. Mechanistically, Themis2 suppresses Ly49H signalling by attenuating ZAP70/Syk phosphorylation, and it also translocates to the nucleus, where it promotes Zfp740-mediated repression to regulate the persistence of memory NK cells. Zfp740 deficiency increases the number of memory NK cells and enhances the effector function of memory NK cells, which further supports the relevance of the Themis2-Zfp740 pathway. In conclusion, our study shows that Themis2 quantitatively and qualitatively regulates NK cell memory formation.

as cytoplasmic events and the transcriptional and epigenetic changes as nuclear events [10][11][12] , the molecular machinery that regulates the differentiation into memory cells is not yet fully understood.
NK cells play an important role in anti-viral and anti-tumor immunity 13 .They recognize unhealthy cells by utilizing a repertoire of activating NK receptors and exert their cytotoxic activity and production of chemokines and cytokines such as interferon (IFN)-γ for the eradication of these cells 13,14 .NK cells have traditionally been classified as innate immune cells; however, emerging evidence has shown that NK cells can differentiate into memory cells with enhanced effector functions 5,6,9 .An activating NK receptor, Ly49H, expressed on mouse NK cells specifically recognizes the mouse cytomegalovirus (MCMV) protein m157 on infected cells and transmits an activation signal via an adaptor molecule, DAP12, and the proximal phosphorylation of ZAP70 and Syk [15][16][17] .In addition to the essential role of Ly49H + NK cells in the control of MCMV infection 18 , activated Ly49H + NK cells expand as effector NK cells, undergo apoptosis after the peak of NK cell response, or differentiate into long-lived memory NK cells during MCMV infection 5,19 .Memory NK cells show several enhanced functional properties: (1) Long-term persistence, (2) the capability of undergo secondary expansion, (3) augmented cytotoxicity and IFN-γ production after stimulation of activating NK receptors in vitro, (4) provision of improved host defence against MCMV infection, and (5) enhanced anti-tumor activity in vivo 5,6,19 .Similarly, human NK cells expressing the activating NK receptor NKG2C show robust proliferation, persistence, recall response, and augmented cytotoxicity in response to cytomegalovirus infection, demonstrating the existence of memory NK cells in humans 9,20,21 .
Previous studies have demonstrated the role of NK receptors, cytokines, and transcription factors in the differentiation of memory NK cells in mice [22][23][24][25][26][27][28][29] .However, these studies have primarily focused on the quantity of memory NK cells rather than their qualitative traits.Recent studies have revealed unique transcriptional and epigenetic landscapes in memory NK cells 30,31 , suggesting that the qualitative and quantitative traits of memory NK cells are intricately regulated by transcriptional regulation, epigenetic modification, and signaling through NK receptors and cytokine receptors.Furthermore, a limited number of studies have demonstrated negative regulation of memory NK cell differentiation 23,29 .However, the molecular mechanisms that qualitatively and quantitatively regulate the differentiation of memory NK cells are yet to be fully understood.
In this work, we identify Themis2 as a critical negative regulator of the differentiation of memory NK cells and investigate the role of Themis2 in NK cell memory formation and function.We show that Themis2 in the cytoplasm suppresses Ly49H signaling by attenuating ZAP70/Syk phosphorylation.Moreover, Themis2 translocates into the nucleus and promotes Zfp740-mediated repression to regulate the persistence of memory NK cells.Our study shows that Themis2 quantitatively and qualitatively regulates NK cell memory formation.

Themis2 negatively regulates NK cell memory formation
To identify critical regulators for memory NK cell differentiation, we used NKp46-CreERT2 Tg mice with Rosa26-yellow fluorescent protein (YFP)/YFP alleles (hereinafter, NK-CreERT2 mice) 19 .NKp46 + cells in NK-CreERT2 mice express YFP upon CreERT2-mediated excision of the loxP-floxed STOP sequence in the Rosa26 loci after tamoxifen administration.Thus, NK-CreERT2 mice allowed us to track MCMV-primed NK cells, irrespective of NK cell subsets expressing or not expressing Ly49H, as YFP + NK cells after tamoxifen administration on days 0, 1, and 2, following MCMV infection on day 0 (Fig. 1a).Considering that the vast majority of NKp46 + NK1.1 + TCRβ -lymphocytes in the spleen are conventional NK cells 32,33 , NKp46 + type 1 and type 3 innate lymphoid cells in the spleen would be negligible in this study on naïve and MCMV-primed NK cells in the spleen.On day 10 post-infection (pi), we purified YFP + cells from the spleen, which included MCMV-primed effector Ly49H + KLRG1 high or Ly49H -KLRG1 high NK cells.On day 28 pi, we further purified MCMV-primed long-lived memory NK cells with the stringent phenotypic definition of Ly49H + KLRG1 high Ly6C + DNAM-1 -to low and cytokine-activated Ly49H -KLRG1 high NK cells (Fig. 1a and Supplementary Fig. 1).We then subjected these cell subsets to RNA-seq and obtained 51,826 unique transcripts (feature IDs), 44 of which had expressions that were upregulated in effector Ly49H + NK cells and maintained in memory NK cells, as compared with their expressions in naïve Ly49H + or Ly49H -NK cells, effector Ly49H -NK cells, or cytokineactivated NK cells (Fig. 1b).Because memory NK cells display transcriptomic and epigenetic profiles distinct from those of naïve and effector NK cells 31 , we further narrowed down the candidates to three genes -Themis2, Trp73, and Trib1by their nuclear localization signal (NLS) or subcellular location by using the UniProt knowledgebase (https://www.uniprot.org)(Fig. 1b).Among three genes, we chose Themis2 as the focus of the present study because we confirmed expression in naïve NK cells at the protein level (Supplementary Fig. 2a).
A Themis family member Themis2 is known as a signaling scaffold protein that regulates activation signaling through Toll-like receptor 4 and B cell receptor in macrophages and B cells, respectively [34][35][36] .Themis2 mRNA expression was upregulated in effector Ly49H + NK cells after MCMV infection, and this upregulation was maintained in the subsequent memory NK cells, whereas its expression in Ly49H -NK cells remained unchanged after MCMV infection (Supplementary Fig. 2b), which is consistent with the expression kinetics identified in an RNA-seq analysis (Supplementary Fig. 2c).Consistent with the kinetics of Themis2 mRNA, Themis2 protein expression was increased in effector Ly49H + NK cells after MCMV infection and maintained in memory NK cells (Supplementary Fig. 2d).A functional protein association network analysis using the STRING database v10.5 (https://string-db.org)demonstrated enrichment of mouse Themis2 and human THEMIS2 in the KEGG Pathway "Natural killer cell-mediated cytotoxicity" pathways (mmu04650 and hsa04650, respectively) (Supplementary Fig. 2e), indicating that Themis2 may be involved in NK cell function.NK cells in naïve wild-type (WT) and Themis2deficient (Themis2 -/-) mice 34 did not show significant differences in the percentage and development of NK cells (Supplementary Fig. 2f, g).
Activation signaling through Ly49H is required for MCMV-primed expansion and differentiation of Ly49H + NK cells into memory NK cells 5,19 .We investigated the role of Themis2 in activation signaling through Ly49H.Themis2 -/-NK cells showed greater degranulation and phosphorylation of ZAP70 and Syk, both of which are downstream molecules of the Ly49H-DAP12 adaptor complex 16,17 , after stimulation with the anti-Ly49H mAb, as compared with that in WT NK cells (Supplementary Fig. 3a, b).Together, these results indicate that The-mis2 inhibits NK cell activation via Ly49H.Furthermore, Themis2 -/-NK cells showed increased degranulation after stimulation of NKG2D, NKp46, and NK1.1 (NKR-P1C) (Supplementary Fig. 3c).
To further investigate mechanistic details of Themis2-mediated inhibition of Ly49H signaling, we established a transfectant of a human NK cell line NKL, which expresses the endogenous DAP12 and ZAP70, but not SYK, expressing transfected Ly49H and FLAG-tagged THEMIS2.Biochemical analysis demonstrated that THEMIS2 was coimmunoprecipitated with DAP12 and ZAP70 after stimulation with anti-Ly49H mAb when we used DTSSP for chemical crosslinking cytosolic proteins (Supplementary Fig. 3d, e).However, the coimmunoprecipitation was not detected when we did not use DTSSP (Supplementary Fig. 3d, e), suggesting that the Ly49H-mediated signal induces a protein complex formation consisting of THEMIS2, DAP12, and ZAP70 with a low affinity.
To examine the role of Themis2 in NK cells in vivo, WT and Themis2 -/-mice were infected with MCMV.Themis2 -/-Ly49H + NK cells displayed a more activated phenotype than did WT Ly49H + NK cells on day 1.5 pi, as demonstrated by increased an activation marker CD69 and 7 mice (Themis2 -/-)).g Expression of active Caspase 3/7 in donor WT and Themis2 -/-Ly49H + KLRG1 + and high NK cells in the spleen during MCMV infection.Data are pooled from 2 experiments (n = 3 mice (day 0), 6 mice (days 7 and 17 pi), and 8 mice (day 28 pi)).h Schematic representation of the evaluation of NK cell differentiation by using Themis2 +/+ and Themis2 -/-NK-CreERT2 mice, MCMV infection, and tamoxifen administration.i Percentages of Themis2 +/+ and Themis2 -/-yellow fluorescent protein (YFP) + Ly49H + KLRG1 + and high NK cells in the blood during MCMV infection.Data are pooled from 2 experiments (n = 6 mice in each group).j Percentages and number of memory Themis2 +/+ and Themis2 -/-NK cells in the spleen on day 28 pi.Data are pooled from 2 experiments (n = 5 mice (Themis2 +/+ ) and 6 mice (Themis2 -/-)).Statistical analysis was performed using one-way ANOVA (d, e, g, i) and two-sided Student's t test (f, j).Data are presented as mean values SD (d-g, i, j).Source data are provided as a Source Data file.
To clarify the effects of Themis2 on cell division, survival, and apoptosis of Ly49H + NK cells during MCMV infection, we examined the expression of Ki-67 (a cell cycle marker), Bcl-2 (a survival factor), and Bim and active Caspase 3/7 (apoptotic factors) in WT and Themis2 -/- Ly49H + NK cells during MCMV infection.After the peak of NK cell response, the proportion of active Caspase 3/7 + Ly49H + NK cells was smaller in Themis2 -/-mice than in WT mice on days 7, 17, and 28 postinfection (Fig. 1g).By contrast, Themis2 deficiency had little impact on the frequency of Ki-67 + NK cells and expression of Bim and Bcl-2 during MCMV infection (Supplementary Fig. 4d-f).These results indicate that Themis2 negatively regulates Caspase 3/7-mediated apoptosis in the contraction and memory phases.

Themis2 negatively regulates the memory NK cell function
To examine whether Themis2 regulates the effector function of memory NK cells, naïve WT and Themis2 -/-Ly49H + NK cells and memory WT and Themis2 -/-NK cells derived from the spleen of recipient Tyrobp -/-mice on 28 pi as shown in Fig. 1c were co-cultured with m157-expressing RMA cells.Memory Themis2 -/-NK cells showed greater degranulation and IFN-γ production after co-culture with m157-expressing RMA cells compared with memory WT NK cells and naïve Themis2 -/-Ly49H + NK cells (Fig. 2a).To examine secondary expansion of these memory NK cells, Tyrobp -/-mice received memory WT and Themis2 -/-NK cells at a ratio of one to one and were infected with MCMV.Memory Themis2 -/-NK cells showed a more robust secondary expansion than did memory WT NK cells (Fig. 2b).
To analyze the function of memory Themis2 +/+ and Themis2 -/- NK cells in the eradication of MCMV, we isolated naïve or memory Ly49H + NK cells from uninfected and MCMV-infected Themis2 +/+ and Themis2 -/-NK-CreERT2 mice on day 28 pi, transferred these NK cells separately into naïve Tyrobp -/-mice, and then challenged these mice with MCMV.The memory Themis2 -/-NK cells conferred greater improved protection from MCMV infection to the recipient mice than did memory Themis2 +/+ NK cells and naïve Themis2 +/+ and Themis2 -/-NK cells (Fig. 2c).Together, these results indicate that Themis2 negatively regulates the quality, as well as quantity, of memory NK cells.

Themis2 is translocated into the nucleus of NK cells
Themis2 was predicted to have a bipartite NLS.Since Themis2 -/-NK cells showed enhanced memory formation after MCMV infection, it was likely that the activation signal via Ly49H induces translocation of Themis2 in the nucleus.To address the issue, we expressed GFP-fused THEMIS2 in a human NK cell line NKL expressing Ly49H.Immunofluorescence microscopy analysis showed that THEMIS2 was localized predominantly in the cytoplasm; however, stimulation with an anti-Ly49H mAb induced translocation of THEMIS2 into the nucleus, with concentrations in the nucleus peaking at 24 to 48 h after stimulation (Fig. 3a, b).To examine the subcellular localization of Themis2 in primary mouse NK cells, naïve Ly49H + NK cells, effector Ly49H + NK cells, and memory NK cells were purified from the spleen of naïve and MCMV-infected and tamoxifen-administered Themis2 +/+ NK-CreERT2 mice.Whereas Themis2 was predominantly detected in the cytoplasm of naïve Ly49H + NK cells, an equivalent amount of Themis2 was observed in the nucleus of effector Ly49H + NK cells and memory NK cells (Fig. 3c, d).Taken together, these results demonstrate that activation signal via Ly49H induces the nuclear translocation of Themis2 in NK cells and suggest that Themis2 is involved in transcriptional or epigenetic events for the differentiation into effector and memory NK cells after MCMV infection.

Themis2 regulates transcriptional and epigenetic changes
To analyze the role of Themis2 in the nucleus, we performed RNA-seq and chromatin immunoprecipitation sequencing (ChIP-seq) of trimethylation at the 4 th lysine residue of the histone H3 (H3K4me3) and acetylation at the 27 th lysine residue of histone H3 (H3K27ac), epigenetic markers of active promoters and regulatory regions correlated with active transcription, respectively, by using naïve and memory NK cells from naive and MCMV-infected Themis2 +/+ and Themis2 -/-NK-CreERT2 mice (Fig. 4a).A gene ontology (GO) analysis of differentially expressed genes (DEG) between memory Themis2 +/+ and Themis2 -/-NK cells (hereinafter, memory DEGs, Supplementary Fig. 6a-c) yielded enriched signatures associated with the apoptosis, transcription, and cell cycle (Fig. 4b and Supplementary Fig. 6d).A gene set enrichment analysis (GSEA) revealed that memory Themis2 +/+ NK cells showed enrichment of gene sets found in activated and effector T cells, whereas memory Themis2 -/-NK cells showed those found in memory T cells (Supplementary Fig. 6e), suggesting that the loss of Themis2 promotes differentiation into memory NK cells by changing the transcriptional profile.Moreover, memory DEG promoters showed significant enrichment of binding motifs for transcription factors (Fig. 4c and Supplementary Fig. 6f).These findings raised the possibility that Themis2 inhibits the differentiation of memory NK cells by participating in transcriptional regulation after nuclear translocation.
Themis2 +/+ and Themis2 -/-NK cells did not show any obvious differences in their epigenome-wide H3K4me3 and H3K27ac landscapes, irrespective of whether they were naïve and memory NK cells (Supplementary Fig. 7a).However, pairwise intersection heatmaps of the H3K4me3 and H3K27ac regions in naïve and memory NK cells revealed that memory Themis2 +/+ and Themis2 -/-NK cells had distinct epigenetic profiles, particularly in the H3K4me3 region, as indicated by the observed negative correlation (Fig. 4d).To examine the function of the differentially accessible regions (DAR) between memory Themis2 +/+ and Themis2 -/-NK cells, the H3K4me3 DARs uniquely present in either memory WT or Themis2 -/-NK cells were extracted (Supplementary Fig. 7b), and their single nearest genes (hereafter, memory H3K4me3 DAR genes) were identified (Supplementary Fig. 7c).GO analysis of the memory H3K4me3 DAR genes revealed signatures associated with the cell cycle, apoptosis, and transcription (Fig. 4e).Memory H3K4me3 DAR gene promoters showed significant enrichment of binding motifs for members of the KLF, C2H2 zinc finger, and the ETS families (Fig. 4f and Supplementary Fig. 7d).These results overlapped with those of the GO analysis and motif analysis of memory DEGs (Fig. 4c and Supplementary Fig. 6f).
To examine the non-promoter H3K27ac-marked regions, H3K27ac DARs uniquely present in either memory Themis2 +/+ or Themis2 -/-NK cells were extracted; H3K27ac DARs overlapping with CCCTC-binding factor (CTCF)-bound distal and proximal enhancer-like signatures (ELS) available in ENCODE (https://www.encodeproject.org/)were further extracted to highlight non-promoter H3K27ac regions; and then their single nearest genes (hereinafter, memory H3K27ac DAR distal or proximal ELS genes) were identified (Supplementary Fig. 7e).Unlike memory H3K4me3 DAR genes, memory H3K27ac DAR distal ELS genes showed enriched signatures associated with the activation and differentiation of lymphocytes and leukocytes (Supplementary Fig. 7f).In contrast, memory H3K27ac DAR proximal ELS genes showed enriched signatures associated with the cell cycle (Supplementary Fig. 7g).
We assessed the correlation of H3K4me3 and H3K27ac landscapes in naïve and memory Themis2 +/+ or Themis2 -/-NK cells by using the unified atlas of these H3K4me3 and H3K27ac regions.The pairwise intersection heatmap revealed that Themis2 has little impact on the overall correlations in H3K4me3 and H3K27ac regions in all NK cell subsets (Supplementary Fig. 7h).We also examined the correlation of gene expression of memory DEGs and signals of memory H3K4me3 DARs.The scatter plots suggested the possibility that Themis2 might be involved in gene expression and promoter accessibility of transcription-associated genes (Supplementary Fig. 7i).Together, these findings support the hypothesis that Themis2 negatively regulates NK cell memory formation through the transcriptional regulation of apoptosis-associated genes in the nucleus, in addition to the negative regulation of activation signaling via Ly49H in the cytoplasm.

Themis2 promotes Zfp740-mediated transcriptional repression
To identify Themis2-binding nuclear proteins, the nuclear proteins of MCMV-primed effector Ly49H + NK cells from the spleen of MCMVinfected and tamoxifen-administered Themis2 -/-NK-CreERT2 mice were incubated with recombinant FLAG-tagged Themis2 protein (Themis2-FLAG); the Themis2-bound proteins were then immunoprecipitated with an anti-FLAG mAb (Supplementary Fig. 8a-c and Fig. 5a), and analyzed by matrix-assisted laser desorption-ionization time-offlight (MALDI-TOF) mass spectrometry.A MASCOT search identified nine transcription factors and one epigenetic enzyme expressed in NK cells (Supplementary Fig. 8d), implying that Themis2 preferentially interacts with transcription factors in the nucleus of NK cells.
To identify the key transcription factor for memory NK cell differentiation, we analyzed the lists of transcription factors predicted by motif analyses of memory DEG promoters and memory H3K4me3 DAR gene promoters (Fig. 4c, f, and Supplementary Fig. 6f, 7d) and compared those lists with a list of Themis2-bound transcription factors detected by MADLI-TOF mass spectrometry (Fig. 5b).Consequently, Zfp740, a zinc finger protein with unknown function, was found to be common among the three lists of transcription factors (Fig. 5b).Mouse Zfp740 (UniProt Q6NZQ6) is a C2H2 zinc finger transcription factor with three zinc finger domains and a low complexity disordered region.Mouse Zfp740 and human ZNF740 (UniProt Q8NDX6) have high homology (Blastp E-value, 3 × 10− 114 ), and their C-terminus amino acid sequences containing zinc finger domains are identical.Although no studies have reported the role of mouse Zfp740, Ingenuity Pathway Analysis revealed that human ZNF740 has an RNA polymerase IIspecific transcription factor with DNA-binding activity and that it interacts with Polycomb-group proteins BMI1 (aka PCGF4) and RNF2 (aka RING1B) (Supplementary Fig. 8e), both of which are core components of Polycomb repressive complex 1 37 .These results suggest that Zfp740 might be involved in transcriptional repression by interacting with transcriptional repressor complexes.Zfp740 was located in the nucleus in 293 T cells expressing MYC-tagged Zfp740 protein (Zfp740-MYC), as determined by immunoblotting analysis with anti-MYC mAb (Supplementary Fig. 8f).Moreover, nuclear Themis2 protein was physically associated with the Zfp740 protein in 293 T cells coexpressing Zfp740-MYC and Themis2-FLAG with an additional NLS (referred to as NLS-Themis2-FLAG) for efficient nuclear localization (Fig. 5c and Supplementary Fig. 8f).
To determine the mode of action of Zfp740 as a transcription factor, we designed luciferase reporter constructs driven by a promoter harboring either the Zfp740-binding motif predicted by the motif analyses of memory DEG promoters and memory H3K4me3 DAR gene promoters (Fig. 4c, f) or a random motif, and transiently expressed with Zfp740 in 293 T cells.Zfp740 reduced the luciferase activity in a Zfp740-binding motif-dependent manner (Fig. 5d).Next, we addressed the effect of nuclear Themis2 on the transcriptional activity of Zfp740 by co-expression with or without NLS-Themis2.
Although NLS-Themis2 alone had little impact on luciferase activity, co-expression of NLS-Themis2 and Zfp740 significantly reduced it, as compared with Zfp740 alone (Fig. 5e).These results indicate that Zfp740 is a transcriptional repressor and Themis2 promotes Zfp740mediated transcriptional repression.
Memory Themis2 -/-NK cells highly expressed anti-apoptosisassociated genes with GO terms related to the negative regulation of apoptosis, such as Niban2, Bcl10, Map4k4, and Sirt1 [38][39][40][41] , whose promoters had putative Zfp740-binding sequences, as compared with memory WT NK cells (Supplementary Fig. 9k).Consistent with these results, memory Zfp740 -/-NK cells highly expressed the anti-apoptosis-associated genes with promoters bearing putative Zfp740binding sequences, compared with memory WT NK cells (Fig. 6d).These findings support the possibility that Themis2 and Zfp740 inhibit the differentiation and persistence of memory NK cells through transcriptional repression of anti-apoptosis-associated genes during MCMV infection.
To investigate whether Zfp740 impacts the effector function of NK cells, naïve and memory WT and Zfp740 -/-Ly49H + NK cells were cocultured with m157-expressing RMA cells.Although naïve WT and Zfp740 -/-Ly49H + NK cells degranulated and produced IFN-γ against m157-expressing RMA cells to the same degree, memory Zfp740 -/-NK cells showed more efficient degranulation and IFN-γ production against m157-expressing RMA cells compared with memory WT NK cells and naïve WT and Zfp740 -/-Ly49H + NK cells (Fig. 6e).These findings demonstrate that Zfp740 suppresses the effector function of memory, but not of naïve, NK cells.Thus, Zfp740 also negatively regulates the quantity and quality of NK cell memory formation.

Discussion
In this study, we identified Themis2 as a critical regulator of the differentiation and function of memory NK cells during MCMV infection.Themis2 inhibited activation signaling via Ly49H by attenuating the activation of ZAP70 and/or Syk, which restricted effector function and expansion of memory Ly49H + NK cells.Moreover, Themis2 was translocated into the nucleus in Ly49H + NK cells following MCMV infection.In the nucleus, Themis2 promoted Zfp740-mediated transcriptional repression of anti-apoptosis-associated genes to regulate the persistence of memory NK cells.Thus, Themis2 qualitatively and quantitatively limits the formation of NK cell memory (Supplementary Fig. 10).
There is limited literature describing the function of Themis2 in immune cells.In contrast to an inhibitory effect of Themis2 on Ly49H signaling, previous reports have demonstrated that Themis2 promotes activation of ERK and p38, augmenting Toll-like and B cell receptor signals in macrophages and B cells, respectively 34,35 .These findings imply that Themis2 exerts an activating or inhibitory function in a stimulus-or cell-type-dependent manner.Themis1 is a well-studied Themis family protein with similar domain architectures to those of Themis2 36 .Although Themis1 is required for the activation of thymocytes through modulation of T cell receptor signaling 42,43 , more recent studies have revealed that Themis1 forms a complex with SHP-1 and increases phosphatase activity through stabilization of phosphorylated SHP-1 44,45 .Considering the functional interchangeability of Themis1 and Themis2 in thymocyte development, as well as the similarity of the activation through T cell receptor and Ly49H, both of which transmit activation signals via signaling immunoreceptor tyrosine-based activation motif-bearing adaptor molecules (CD3ζ and DAP12, respectively) and ZAP70 14,46,47 , Themis2 might bind to SHP-1 and inhibit activation signaling in NK cells through increased phosphatase activity 14,48 .The phosphorylation of ZAP70/SYK, which are signaling molecules downstream of DAP12, triggers the activation of multiple signaling adaptors and enzymes, including PLCγ1/2 48 .PLCγ1/2 catalyzes the production of the second messengers diacylglycerol and inositol(1,4,5)-trisphosphate, which triggers PKC activation and an increased Ca 2+ release from the endoplasmic reticulum, respectively, and the subsequent Ca 2+ -calcineurin signaling, resulting in nuclear translocation of NFAT 48 .Thus, Themis2 might promote SHP-1mediated dephosphorylation of the protein substrates of the tyrosine kinases linked to activating NK receptors, including Ly49H, which contributes to the downregulation of the Ca 2+ influx.Notably, SHP-1 limits the proliferation of Ly49H + NK cells and host protection against MCMV infection 49 .Thus, Themis2 likely forms a complex with SHP-1 and presumably with DAP12 and or ZAP70 (Supplementary Fig. 3d and 3e), and attenuates Ly49H signaling, resulting in restricted effector function of naïve and memory NK cells.Furthermore, naïve Themis2 -/-NK cells showed augmented effector functions after stimulation of NKG2D, NKp46, and NK1.1 (NKR-P1C), as well as Ly49H, suggesting that the cytoplasmic Themis2 may have an inhibitory role in activation signaling through a diverse group of activating NK receptors, irrespective of naïve or memory NK cells.Interestingly, Zfp740 also inhibited the effector function of memory NK cells.Thus, the enhanced effector function of memory Themis2 -/-NK cells may be attributable to the loss of Themis2-mediated inhibition of activation signaling in the cytoplasm and the loss of Themis2-Zfp740 complexmediated repression in the nucleus.Intriguingly, Themis2 regulated Caspase 3/7-mediated apoptosis of MCMV-primed effector and memory NK cells in the contraction and memory phases, whereas Themis2 had little impact on the apoptosis in naïve NK cells.In addition, Themis2 inhibited effector functions and proliferation of naïve and memory Ly49H + NK cells.These results suggest that Themis2 regulates the effector function and or survival of naïve and differentiated NK cells.On the other hand, although Zfp740 regulated the survival and the effector function of memory NK cells, the effector function of naïve NK cells was not affected by Zfp740.It is presumed that the Zfp740 function is specific, particularly to regulating the expression of anti-apoptosis-associated genes in memory NK cells.
The comprehensive understanding of Themis2-bound nuclear proteins in naïve, effector, and memory NK cells would be important, because differential Themis2-bound proteins may regulate NK cell memory.However, there is a technical limitation: The quantity of nuclear protein obtained from an NK cell.Mouse NK cells per spleen yield smaller than 100 μg nuclear proteins.Further, the frequency of memory NK cells is approximately 10% in splenic NK cells in Themis2 -/- NK-CreERT2 mice (Fig. 1j).To overcome the issue, we used effector Ly49H + NK cells to identify Themis2-binding transcription factors, because they are increased in number on days 7-10 pi.However, it seems technically impossible to obtain an adequate amount of nuclear proteins from memory NK cells.Although we were interested in Themis2-bound nuclear proteins in naïve, effector, and memory NK cells comprehensively, effector NK cells were realistic for the pulldown and the subsequent mass spectrometry.Future studies using other methods might clarify the roles of differential Themis2-bound nuclear proteins in controlling NK cell memory.
Transcriptional repression is achieved by the combinatorial mechanisms of action of a variety of repressor complexes consisting of transcriptional repressors, co-factors, and epigenetic regulators, and often through enzymatic histone modifications for epigenetic silencing 50 .Polycomb repressive complexes 1 and 2 (PRC1 and PRC2) constitute a conserved gene silencing system that plays pivotal roles in embryogenesis, development, and differentiation processes 37 .Zfp740 is a member of the C2H2 zinc finger transcription factors that activate or repress the expression of genes involved in development and differentiation 51 .Human ZNF740 is physically associated with BMI1 and RNF2 (Supplementary Fig. 8e) 52 , both of which are core components of PRC1 37 .Given the conserved function of mouse Zfp740 and human ZNF740 with their identical zinc finger domain amino acid sequences, one possible scenario is that the Themis2-Zfp740 complex participates in PRC1-mediated transcriptional repression; the complex binds to Zfp740-binding sequences of the regulatory elements and might recruit PRC1 for stable silencing of Zfp740 target genes in memory NK cells.A recent study has reported that an ETS family transcriptional factor, Fli1, restricts the formation of memory precursor Ly6C -Ly49H + NK cells with increased protein expression of Bcl-2 at the early stage of MCMV infection 29 .In the present study, the populations of WT and Themis2 -/-Ly6C + Ly49H + NK cells showed comparable each other on day 7 pi (Supplementary Fig. 4a, 5a), and WT and Themis2 -/-Ly49H + NK cells equivalently expressed Bcl-2 during MCMV infection (Supplementary Fig. 4e), suggesting that Themis2 may not be involved in the regulation of the size of the memory precursor NK cell pool.We also identified Fli1 in our motif analyses of memory DEG promoters and memory H3K4me3 DAR gene promoters (Fig. 5b).It is possible that multiple transcriptional factors, including Zfp740 and Fli1, redundantly and independently control the quantity and quality of memory NK cells with different modes of action at the same or distinct regulatory elements.The cooperative interactions of Zfp740, Fli1, and Polycomb group proteins in forming NK cell memory have not yet been examined.Further studies are needed to elucidate the mechanistic role of Zfp740 and Themis2 in Zfp740-mediated repression, i.e., identification of Zfp740 target genes and their biological processes regulated by Zfp740 repressor complex with or without Themis2.Such investigations might reveal the core molecular identity of NK cell memory, although a current technical limitation is the lack of a chromatin immunoprecipitation-validated antibody against Zfp740.
Here, we demonstrated that Themis2 deficiency reinforces expansion, persistence, effector functions, and memory differentiation of MCMV-primed Ly49H + NK cells and that it confers improved host protection against MCMV infection.All these features of Themis2deficient NK cells appear beneficial for anti-viral and anti-tumor immunity.We also found that Themis2 deficiency did not disadvantage NK cells; hence, the physiological relevance of Themis2 is still unknown.The functional properties of memory NK cells are attractive for developing NK cell-based immunotherapies for incurable infectious diseases and malignancies.Thus, further studies of the detailed molecular mechanisms of Themis2-mediated inhibition of NK cell function are needed to gain important insights into the identity of NK cell memory; such studies might provide opportunities for developing therapeutic interventions to selectively unleash NK cells to combat viral infections and cancers.

Methods
This study was approved by the Laboratory Animal Ethics Committee of the University of Tsukuba (approval number 22-154) and approved by the Ethics Committee for Medical Sciences at the University of Tsukuba (approval number 234-2).
Mice between 6 and 24 weeks of age were used for experiments in a gender-matched manner.All mice were housed and maintained under the specific-pathogen-free conditions and at 23.5 °C ± 2.5 °C ambient temperature and 52.5% ± 12.5% humidity on a 14-h light (5 am to 7 pm): 10-h dark cycle (7 pm to 5 am).The mice were fed a MF diet (MF, Oriental Yeast, Tokyo, Japan).All strains of mice were bred separately.Mice were euthanized at experimental endpoint by inhalation of 70 to 100% CO 2 for 30 to 60 seconds, followed by cervical dislocation.All procedures were approved by the Laboratory Animal Ly49H + NK cells were enriched on days 7, 17, and 28 pi, as described above, and then active Caspase 3 and or 7 were stained by using a FAM FLICA Caspase-3/7 Kit according to the manufacturer's instructions (Bio-Rad, Hercules, California, U.S.A.).Cells were washed and extracellular molecules were stained with fluorochrome-conjugated mAbs.
In all experiments, doublet cells and dead cells were excluded by FSC-A-H gating, followed by SSC-A-W gating, and dead cells were excluded by using 510/50 or 525/50 nm filter of flow cytometers.
Samples were run on LSRFortessa and FACSAria III (BD Biosciences), data were obtained by using BD FACSDiva V8, and the data were analyzed with FlowJo V10 (FlowJo, Ashland, Oregon, U.S.A.).
3x FLAG tag was added at the C-terminus of the full-length THE-MIS2 ORF and constructed into pMXs-IRES-GFP retroviral vector (Cell Biolabs, San Diego, California, U.S.A) using a NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs, Ipswich, Massachusetts, U.S.A.).A retroviral vector pMXs-IRES-GFP encoding FLAG-tagged THEMIS2 (THEMIS2-FLAG) was prepared by transfection of the retroviral construct and pcDNA3.4encoding vesicular stomatitis virus G envelope glycoprotein were transfected into packaging 293gp cells (RIKEN BioResource Research Center, Tsukuba, Ibaraki, Japan) using Lipofectamine 3000 (Thermo Fischer Scientific).The retroviral supernatant was added into the culture of NKL-Ly49H cells in RetroNectin-coated wells (Takara Bio) and centrifuged at 700 x g rpm and 32 °C for 30 min.GFP + NKL-Ly49H cells were sorted with BD FACS Aria III (BD Biosciences) and NKL-Ly49H cells expressing THEMIS2-FLAG was established.These procedures were approved by the Ethics Committee for Medical Sciences at the University of Tsukuba (approval number 234-2).All participants underwent an informed consent process.
Cells were cytospun by using a Shandon Cytospin 3 Centrifuge (Thermo Fisher Scientific), mounted with VECTASHIELD Vibrance Antifade Mounting Medium with DAPI (Vector Laboratories, Burlingame, California, U.S.A.), and observed under a confocal laser microscopy FluoView FV10i (Olympus, Tokyo, Japan).Percentages of nuclear Themis2 per cell were calculated by FLUOVIEW Ver4.1 (Olympus) and represented as a proportion of Themis2 signals overlapping DAPI relative to all Themis2 signals in a cell.

RNA-seq
Naïve Ly49H + and Ly49H -NK cells, effector Ly49H + and Ly49H -NK cells, memory NK cells, and cytokine-activated Ly49H -NK cells were purified by double sorting with FACSAria III (BD Biosciences) as described above.RNA in sorted NK cells (1.2 × 10 3 to 1.8 × 10 5 ) was isolated using a TRIzol reagent (Thermo Fisher Scientific), the yield was monitored using an Agilent Bioanalyzer RNA 6000 Pico Kit (Agilent Technologies, Santa Clara, California, U.S.A.), and processed to sequencing library using a SMART-seq Stranded Kit (Takara Bio).Sequencing was performed by NextSeq 500 (Illumina, Santa Monica, California, U.S.A.).FASTQ files were imported to CLC Genomics Workbench 12.0 (Qiagen, Hilden, Germany) and mapped to mm10.Normalized expression values were obtained by quantile normalization for the total count.DEGs were defined as having normalized expression >10 and log 2 (fold change) >1 with false discovery rate (FDR) < 0.05.The pairwise heatmap of the number of DEGs and the heatmap of memory DEGs with relative expression with Z-score was drawn by using Heatmapper 1.0.0 (http://www.heatmapper.ca/).
H3K4me3 DARs uniquely present in either memory Themis2 +/+ or Themis2 -/-NK cells were extracted by bedtools on Galaxy v21.09 57 and these unique H3K4me3 DAR-associated single nearest genes (memory H3K4me3 DAR genes) were identified by using GREAT v4.0.4 58 .H3K27ac DARs uniquely present in either memory Themis2 +/+ or Themis2 -/-NK cells were extracted by Galaxy.To highlight non-promoter H3K27ac-marked regions, these unique H3K27ac DARs were overlapped with CTCF-bound enhancer-like signatures (ELS) available in ENCODE (https://www.encodeproject.org/) 59 , because CTCF binding correlates with enhancer activity and enhancer-promoter interaction to maintain gene expression 60 .These unique H3K27ac DAR-associated single nearest genes with distal and proximal ELS (memory H3K27ac DAR distal and proximal ELS genes, respectively) were identified by using GREAT.The heatmap of memory H3K4me3 DARs with the relative read count was drawn by Morpheus (https://software.broadinstitute.org/morpheus/).H3K4me3 and K3K27ac reads of naïve and memory NK cells versus input controls were normalized and scaled by Galaxy.These epigenome-wide heatmaps of the read density in 5 kb from transcription start sites (TSS) were drawn by deepTools on Galaxy.The heatmaps of pairwise intersections of individual and unified atlases of H3K4me3 peaks and H3K27ac peaks of naïve and memory NK cells was drawn with their Spearman's correlations for pairwise Jaccard statistics by deepTools on Galaxy.The distance from individual memory DARs to TSS was plotted with the number of DARs within each distance range by GREAT.

Bioinformatics
GO analysis of memory DEGs was performed by DAVID v2021q4 61 and ShinyGO v0.6.1 62 .GO analysis of memory DAR genes were performed by GREAT.
Functional protein association networks of mouse Themis2 and human THEMIS2 were constructed and KEGG Pathways of their networks were implemented by STRING 63 .
GSEA V4.2.3 65 was performed by using normalized expression in memory WT and Themis2 -/-NK cells and normalized enrichment score (NES) and FDR were computed.
For an integrative analysis of memory DEGs and memory H3K4me3 DARs, the absolute values of the log 2 -transformed fold change of normalized expression of memory DEGs and the fold change of signal per million reads of memory H3K4me3 DARs were represented as scatter plots.Spearman's correlation coefficients with their p values of all memory DEGs/ memory H3K4me3 DARs and these genes annotated with the GO term DNA-templated transcription were calculated by using cor.test()function in R (https://www.r-project.org/).A mouse gene list annotated with DNA-templated transcription was obtained by using biomaRt package in R.
The knowledge-based network analysis on ZNF740 was performed by Ingenuity Pathway Analysis 20.0 66 and the protein-protein interactions in the nucleus were represented.

Motif analysis
Transcription factor binding motifs in promoters of memory DEGs and memory H3K4me DAR genes were analyzed by the MEME Suite 5.3.0 67 .Enriched motifs in promoter sequences (200 bp upstream and 100 bp downstream of TSS) were discovered by MEME STREME.Transcription factors that can bind to these motifs were ranked by MEME Tomtom.The p value for each motif was calculated as the probability that a random motif of the same width as the target would have an optimal alignment with a match score as good or better than the target's.MEME Tomtom estimates the p value using a null model consisting of sampling motif columns from all the columns in the set of target motifs 68 .
To prepare Themis2-FLAG, pcDNA3.4-Themis2-FLAGwas transfected into 293 F cells using an ExpiFectamine 293 Transfection Kit (Thermo Fisher Scientific).On day 3 post-transfection, 293 F cells were homogenized in hypotonic buffer.Lysates were incubated with anti-FLAG M2 affinity gel (Sigma) in HEPES-buffered saline supplemented with protease inhibitor cocktail (Merck Millipore) for 2 h at 4 °C with agitation and eluted with 3x FLAG peptides (Sigma-Aldrich) by incubating 30 min at 4 °C.The purity and quality (e.g.monomer, aggregation) was confirmed by SDS-PAGE, followed by silver staining using a Pierce Silver Stain Kit (Thermo Fisher Scientific).
Two-milligram nuclear proteins of effector Themis2 -/-NK cells were incubated with 200 μg Themis2-FLAG for 3 h at 4 °C with agitation.The mixtures were chemically crosslinked with 1 mM 3,3'-dithiobis(sulfosuccinimidyl propionate, DTSSP) (Thermo Fisher Scientific) for 30 min on ice.After gel filtration of the mixtures by using a HiTrap Desalting column (Cytiva, Marlborough, Massachusetts, U.S.A.), Themis2-FLAG-bound proteins were immunoprecipitated with anti-FLAG M2 affinity gel (Sigma-Aldrich) and eluted with 3x FLAG peptides (Sigma-Aldrich).The immunoprecipitants were separated by SDS-PAGE by using Mini-PROTEAN TGX polyacrylamide (PA) gels (Bio-Rad), followed by silver staining using a Pierce Silver Stain for Mass Spectrometry (Thermo Fisher Scientific).Selected protein bands that appeared only in immunoprecipitants with Themis2-FLAG and blank gels with the same molecular weight positions were excised and proceeded previously as described 69 .The resulting peptide peaks uniquely present in selected gel pieces, but not in blank gel pieces, were picked up and proteins were identified through peptide mass fingerprint matching by using flexControl 3.4 and MASCOT Server 2.7 with UniProtKB/Swiss-Prot database (https://www.uniprot.org/).Transcription factors and epigenetic enzymes among Themis2-bound nuclear proteins were filtered; those expressed in NK cells were extracted by utilizing RNA-seq data of NK cell subsets and were represented.

Biochemistry
NKL-Ly49H cells expressing THEMIS2-FLAG were stimulated with anti-Ly49 mAb for 2 or 5 min, as described above.Stimulated cells were homogenized in hypotonic buffer and the supernatants were collected as cytosolic proteins, as described above.Cytosolic proteins were treated with or without DTSSP and THEMIS2-FLAG-bound proteins were immunoprecipitated with anti-FLAG M2 affinity gel (Sigma-Aldrich) and eluted with 3x FLAG peptides (Sigma-Aldrich).The immunoprecipitants were separated by SDS-PAGE, transferred into polyvinylidene difluoride membranes (Merck Millipore), and immunoblotted with rabbit anti-DAP12 mAb (D7G1X, Cell Signaling Technology) or rabbit anti-ZAP70 mAb (99F2, Cell Signaling Technology) for 12-16 h at 4 °C, followed by HRP-conjugated anti-rabbit IgG (Bio-Legend).The chemiluminescence was developed with a SuperSignal West Pico PLUS Chemiluminescence Substrate (Thermo Fisher Scientific) and detected with ImageQuant LAS 4000 mini and ImageQuant TL software ver8 (Cytiva).The membrane was reblotted with biotinylated anti-FLAG mAb (M2, Sigma-Aldrich), followed by streptavidin-HRP (Cytiva).
Themis2-FLAG-binding nuclear proteins were immunoprecipitated by incubation with anti-FLAG M2 affinity gel (Sigma-Aldrich) and eluted with 3x FLAG peptides (Sigma-Aldrich).The immunoprecipitants were separated by SDS-PAGE, transferred into polyvinylidene difluoride membranes (Merck Millipore), and immunoblotted with mouse anti-MYC-tag mAb (9B11, Cell Signaling Technology) for 16 h at 4 °C, followed by HRP-conjugated anti-mouse IgG (Cytiva).The chemiluminescence was developed with a SuperSignal West Pico PLUS Chemiluminescence Substrate (Thermo Fisher Scientific) and detected with ImageQuant LAS 4000 mini (Cytiva).The membrane was reblotted with biotinylated anti-FLAG mAb (M2, Sigma-Aldrich), followed by streptavidin-HRP (Cytiva).The nuclear translocation efficiency of Themis2-FLAG and NLS-Themis2-FLAG was compared by calculating the ratio of nuclear to cytosolic FLAG signals.
The dilution factors for these antibodies for biochemistry were 1000 to 2000 or these antibodies were diluted and used according to the manufacturers' instructions.

3 )Fig. 1 |
Fig. 1 | Themis2 negatively regulates NK cell memory formation.a Schematic representation of the approach used to obtain the natural killer (NK) cell subsets that were used to identify critical regulators of memory NK cell differentiation by RNA-seq (n = 3 in each group).b Criteria used to select critical regulators of the differentiation of memory NK cells.c Schematic representation of the evaluation of NK cell differentiation by co-transfer of WT and Themis2 -/-Ly49H + NK cells into Tyrobp -/-mice and mouse cytomegalovirus (MCMV) infection.d Percentages of donor WT and Themis2 -/-Ly49H + KLRG1 + and high NK cells in the blood during MCMV infection.Data are pooled from 2 experiments (n = 8 mice in each group).e Ratio of donor WT and Themis2 -/-Ly49H + KLRG1 + and high NK cells in the blood during MCMV infection.Flow cytometry plots are representative of 2 experiments (n = 4 mice in each group).Data are pooled from 2 experiments (n = 8 mice in each group).f Percentages and numbers of donor memory WT and Themis2 -/-NK cells in the spleen on day 28 pi.Data are pooled from 2 experiments (n = 8 mice (WT)

FDRFig. 4 |
Fig. 4 | Themis2 regulates transcriptional and epigenetic changes in memory NK cells.a Schematic representation of the preparation of naïve and memory Themis2 +/+ and Themis2 -/-natural killer (NK) cells for RNA sequencing (b, c) and chromatin immunoprecipitation sequencing of tri-methylation at the 4th lysine residue of the histone H3 (H3K4me3) regions (d-f) and acetylation at the 27th lysine residue of histone H3 (H3K27ac) regions (n = 3 in each group).Naïve Themis2 +/+ and Themis2 -/-yellow fluorescent protein (YFP) + Ly49H + NK cells were purified from the spleen of uninfected and tamoxifen-administered Themis2 +/+ and Themis2 -/-NK-CreERT2 mice on day 10.Memory Themis2 +/+ and Themis2 -/- YFP + Ly49H + KLRG1 high NK cells were purified from the spleen of mouse cytomegalovirus (MCMV)-infected and tamoxifen-administered Themis2 +/+ and Themis2 -/- NK-CreERT2 mice on day 28 pi.Red boxes represent the gating of naïve and memory NK cells.b Gene ontology (GO) analysis of memory differentially expressed genes (DEG).GO terms of Biological Process are shown with p values.c Motif analysis of memory DEGs.Transcription factors that can bind to these binding sequences enriched in memory DEG promoters are shown with p values.d Pairwise intersections of individual atlases of H3K4me3 and H3K27ac in naïve and memory Themis2 +/+ and Themis2 -/-NK cells.Pairwise intersection heatmaps are shown with Spearman's correlation values.e GO analysis of memory H3K4me3 differentially accessible region-associated (DAR) genes.GO terms of Biological Process are shown with false discovery rate (FDR).f Motif analysis of memory H3K4me3 DAR genes.Transcription factors that can bind to these binding sequences enriched in memory H3K4me3 DAR gene promoters are shown with p values.The p value for each motif was calculated by a one-sided statistical test of MEME Tomtom (see the "Methods" section).Source data are provided as a Source Data file.

Fig. 5 |
Fig. 5 | Themis2 promotes Zfp740-mediated transcriptional repression.Nuclear proteins of effector Themis2 -/-natural killer (NK) cells were incubated with The-mis2-FLAG, and Themis2-bound proteins were immunoprecipitated.Unique protein bands of the immunoprecipitated (IP) products were analyzed by mass spectrometry.a Silver staining of IP products.Proteins bands uniquely present in IP products incubated with Themis2-FLAG are shown by white arrows.These images are representative of 5 experiments.b Venn diagram of transcription factors predicted by motif analyses of memory differentially expressed gene (DEG) promoters, and memory tri-methylation at the 4 th lysine residue of the histone H3 (H3K4me3) differentially accessible region-associated (DAR) gene promoters, and Themis2bound transcription factors detected by mass spectrometry.c Interaction of The-mis2 with Zfp740 in the nucleus.Themis2-FLAG in the nuclear proteins of 293 T coexpressing nuclear localization signal (NLS)-Themis2-FLAG and Zfp740-MYC was immunoprecipitated with anti-FLAG mAb and immunoblotted (IB) with anti-MYC mAb, followed by reblotting with anti-FLAG mAb.The image is representative of 3 experiments.d Luciferase reporter assay with Zfp740.Luciferase reporter constructs with Zfp740-binding or random motifs were transfected with or without Zfp740 into 293 T cells.The transcriptional activity is represented as relative light unit.Data are pooled from 4 experiments (n = 8 wells (no promoter), 12 wells (Zfp740 motif (+)), and 13 wells (Zfp740 motif (-))).e Luciferase reporter assay with Zfp740 in the presence or absence of Themis2.Luciferase reporter constructs with Zfp740-binding or random motifs were transfected with or without Zfp740 and or NLS-Themis2 into 293 T cells.The transcriptional activity is represented as relative light unit.Data are pooled from 3 experiments (n = 6) wells (no promoter), 9 wells (Zfp740 motif (+);-Zfp740-Themis2, + Zfp740-Themis2), and 10 wells (Zfp740 motif (+);-Zfp740 + Themis2, + Zfp740 + Themis2, and Zfp740 motif (-)).Statistical analysis was performed using two-sided Student's t test (d) and one-way ANOVA (e).Data are presented as mean values ±SD (d, e).Source data are provided as a Source Data file.