Abstract

   Spliced X-box–binding protein 1 (XBP1s) is an essential transcription
   factor downstream of interleukin-15 (IL-15) and AKT signaling, which
   controls cell survival and effector functions of human natural killer
   (NK) cells. However, the precise mechanisms, especially the downstream
   targets of XBP1s, remain unknown. In this study, by using XBP1
   conditional knockout mice, we found that XBP1s is critical for
   IL-15–mediated NK cell survival but not proliferation in vitro and in
   vivo. Mechanistically, XBP1s regulates homeostatic NK cell survival by
   targeting PIM-2, a critical anti-apoptotic gene, which in turn
   stabilizes XBP1s protein by phosphorylating it at Thr^58. In addition,
   XBP1s enhances the effector functions and antitumor immunity of NK
   cells by recruiting T-bet to the promoter region of Ifng. Collectively,
   our findings identify a previously unknown mechanism by which
   IL-15–XBP1s signaling regulates the survival and effector functions of
   NK cells.

INTRODUCTION

   Natural killer (NK) cells are innate immune cells that directly
   eliminate tumor cells and virus-infected cells and elicit other
   adaptive immune responses ([45]1, [46]2). Upon activation, they produce
   a distinct set of cytokines such as interferon-γ (IFN-γ), tumor
   necrosis factor–α (TNF-α), granzyme B, and perforin. NK cells also
   secrete chemokines such as C-C motif chemokine ligand 1 (CCL1), CCL2,
   CCL3, and CCL4. In immunotherapy, NK cells have emerged as safe and
   promising treatments for patients with certain cancers ([47]3, [48]4).
   For example, chimeric antigen receptors (CARs) can be used to direct NK
   cells toward tumor cells carrying the corresponding antigens ([49]5,
   [50]6). However, it remains challenging to enhance NK cell cytotoxicity
   and persistence in vivo in NK cell–directed therapies ([51]4).

   The endoplasmic reticulum (ER) is the main cellular compartment
   committed to protein folding and maturation within eukaryotes, which is
   critical for maintaining cell homeostasis ([52]7). However, the
   accumulation of unfolded or misfolded proteins that are beyond ER
   protein-folding capacity may result in a stress condition termed “ER
   stress” ([53]7). Cells will activate the unfolded protein response
   (UPR) to alleviate the unfolded protein load, which is controlled by
   three ER stress sensors: protein kinase RNA-like ER kinase (PERK)
   ([54]8), inositol-requiring enzyme 1 (IRE1) ([55]9), and activating
   transcription factor 6 (ATF6) ([56]10). IRE1α is maintained in a
   repressed state under nonstress conditions through an association with
   ER chaperone-binding immunoglobulin protein (BiP). Upon ER stress, BiP
   dissociates and binds misfolded proteins, leading to partial IRE1α
   phosphorylation and dimerization ([57]11). Activation of IRE1α triggers
   the splicing of X-box–binding protein 1 (XBP1) mRNA to produce an
   active transcriptional activator, spliced XBP1 (XBP1s) ([58]12,
   [59]13). XBP1s regulates the transcriptional programs of UPR target
   genes encoding ER chaperones (Hspa5, Ddit3, ERdj4, Pdia3, and Dnajc3)
   ([60]14), ER-associated protein degradation components (Edem1, Herpud1,
   and Hrd1) ([61]15, [62]16), folding enzyme (Pdia6) ([63]14), ER
   translocon (Sec61a1) ([64]17), and other transcription factors (Klf9
   and Foxo1) ([65]18, [66]19). The IRE1α-XBP1 signaling pathway is
   implicated in a wide range of biological processes, including
   tumorigenesis, lipid and glucose metabolism, immune cell development,
   and immune responses ([67]17, [68]20-[69]25).

   Our group previously reported that XBP1s is an essential transcription
   factor downstream of interleukin-15 (IL-15) and AKT signaling that
   controls cell survival and effector functions in human NK cells
   ([70]26). XBP1s also facilitates NK cell proliferation during mouse
   cytomegalovirus (MCMV) infection by regulating c-Myc ([71]27),
   indicating that it plays a fundamental role in controlling NK cell
   immunity. However, the target genes downstream of XBP1s in NK cells
   have not been identified in the cytokine activation setting, nor is it
   known how XBP1s regulates NK cell survival and effector functions.
   Endogenous XBP1s has a short half-life because it is rapidly destroyed
   by proteasomes ([72]28). We previously showed that, in human NK cells,
   AKT maintains XBP1s stability by inhibiting its ubiquitination and
   therefore degradation ([73]26). It has also been reported that XBP1s is
   stabilized when it is phosphorylated by p38 mitogen-activated protein
   kinase (MAPK) and IκB kinase β (IKKβ) ([74]28, [75]29). However,
   whether there are other factors that contribute to the stability of
   XBP1s in NK cells is unknown.

   In this study, we used XBP1 conditional knockout (cKO) mice to explore
   how XBP1s regulates NK cell survival and effector functions. Our
   findings reveal that XBP1s is critical for IL-15–mediated NK cell
   survival. Mechanistically, XBP1s regulates NK cell survival by
   targeting PIM-2, a critical anti-apoptotic gene, in a positive feedback
   loop. Moreover, XBP1s enhances NK cells’ effector functions and
   antitumor immunity by interacting with T-bet to promote IFN-γ
   production. Collectively, our findings uncover detailed mechanisms by
   which IL-15–XBP1s signaling controls the survival and effector
   functions of NK cells.

RESULTS

XBP1s is required for NK cell homeostasis under ER stress

   To define the role of XBP1s in NK cell homeostasis in vivo, we
   generated NK cell–specific cKO mice by crossing Xbp1^f/f mice ([76]30)
   with Ncr1-iCre mice ([77]31) (hereafter called Xbp1^cKO mice). Xbp1^f/f
   mice were used as littermate controls. XBP1s deletion in NK cells was
   verified by quantitative polymerase chain reaction (qPCR) and
   immunoblotting ([78]fig. S1A). Phenotypic analyses revealed that the
   frequencies and absolute numbers of NK cells were similar in the bone
   marrow (BM), spleen, and lung ([79]fig. S1B) in Xbp1^cKO mice compared
   with Xbp1^f/f mice, indicating that XBP1s does not affect NK cell
   homeostasis at steady state. In addition, XBP1s deficiency did not
   affect NK cell maturation, as shown by CD11b and CD27 staining
   ([80]fig. S1C). These results are consistent with a prior study showing
   that XBP1 is dispensable for NK cell basal homeostasis ([81]27).

   We explored the role of XBP1 in NK cell homeostasis in different ER
   stress conditions including during ectopic IL-15 expression, a tumor
   challenge model, and an experimental autoimmune encephalomyelitis (EAE)
   model. We first treated mice with IL-15, as we previously did for human
   NK cells ([82]26). The frequencies and absolute numbers of NK cells
   were significantly lower in Xbp1^cKO mice than in Xbp1^f/f mice
   ([83]Fig. 1A). To further explore the role of XBP1s in the presence of
   abundant IL-15 protein in vivo, we crossed Xbp1^cKO mice with our IL-15
   transgenic (Tg) mice ([84]32, [85]33) to generate Il15^TgXbp1^cKO mice
   ([86]fig. S1D). Il15^TgXbp1^cKO mice showed a significantly reduced
   frequency and absolute number of NK cells compared with the
   Il15^TgXbp1^f/f mice ([87]Fig. 1B), similar to mice treated with IL-15.
   However, NK cell maturation did not change substantially in the mice
   treated with IL-15 or in the Il15^Tg mice ([88]fig. S1, E and [89]F).

Fig. 1. XBP1s contributes to NK cell homeostasis in mice under ER stress.

   [90]Fig. 1.
   [91]Open in a new tab

   (A) Representative plots (left) and summary data (right) of the
   percentages and absolute numbers of NK cells in the spleens from
   Xbp1^f/f and Xbp1^cKO mice treated with IL-15 for 5 days (n = 6 per
   group). (B) Representative plots (left) and summary data (right) of the
   percentages and the absolute numbers of NK cells in the spleens from
   Il15^TgXbp1^f/f and Il15^TgXbp1^cKO mice (n = 8 per group). (C)
   Representative plots (left) and summary data (right) of the percentages
   and the absolute numbers of NK cells in the lungs from B16F10
   tumor-bearing Xbp1^f/f and Xbp1^cKO mice (n = 6 per group). (D)
   Representative plots (left) and summary data (right) of the percentages
   and the absolute numbers of NK cells in the brains from Xbp1^f/f and
   Xbp1^cKO mice at the onset of the EAE (n = 5 per group). Data are shown
   as means ± SD and were analyzed with an unpaired two-tailed t test.
   Data are representative of at least two independent experiments. *P <
   0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

   Tumor microenvironment factors, such as hypoxia and nutrient
   deprivation, can activate ER stress and UPR responses ([92]34). We
   found that Xbp1^cKO mice had significantly higher B16-F10 melanoma
   tumor colonies than Xbp1^f/f mice ([93]fig. S1G). Furthermore, the
   percentage and the absolute number of tumor-infiltrating NK cells were
   significantly lower in Xbp1^cKO mice compared with those in Xbp1^f/f
   mice ([94]Fig. 1C). We also performed experiments using the mouse EAE
   model to induce another ER stress condition ([95]35, [96]36). Xbp1^cKO
   mice displayed a reduced EAE score compared with Xbp1^f/f mice
   ([97]fig. S1H). We found a significantly reduced percentage and
   absolute number of NK cells in the brains of Xbp1^cKO mice compared
   with those from Xbp1^f/f mice at the onset of EAE ([98]Fig. 1D).
   Together, these results indicate that XBP1s is required for NK cell
   homeostasis under conditions of stress defined by excessive IL-15,
   cancer, and autoimmunity.

XBP1s controls IL-15–mediated NK cell survival but not proliferation

   To investigate the mechanism by which XBP1s signaling regulates NK cell
   homeostasis, we examined the proliferation and survival of
   XBP1-deficient NK cells at steady state and under the three
   aforementioned conditions. NK cells from Xbp1^f/f and Xbp1^cKO mice had
   similar proliferation and survival at steady state, as evidenced by
   Ki67 staining and annexin V staining, respectively ([99]fig. S2, A and
   [100]B). IL-15–treated Xbp1^cKO mice and Xbp1^f/f mice had comparable
   percentages of Ki67^+ proliferating NK cells, as didIl15^Tg− Xbp1^cKO
   mice compared with Il15^TgXbp1^f/f mice ([101]fig. S2, C and [102]D).
   In addition, a 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay
   further showed that NK cell proliferation rate was similar between
   Xbp1^f/f and Xbp1^cKO mice in the BM, spleen, and liver ([103]fig. S2,
   E to [104]G). However, the viability of NK cells was significantly
   lower in the IL-15–treated Xbp1^cKO mice or Il15^TgXbp1^cKO mice than
   in the IL-15–treated Xbp1^f/f or Il15^TgXbp1^f/f mice, respectively, as
   evidenced by annexin V staining ([105]Fig. 2, [106]A and [107]B).
   Similarly, decreased survival but unchanged proliferation was observed
   in NK cells from Xbp1^cKO mice compared with those from Xbp1^f/f mice
   with tumor challenge or at the onset of EAE disease ([108]fig. S2, H
   and [109]I). These results suggest that the defects of NK cell
   homeostasis by XBP1s ablation at steady or tumor challenge states may
   be due to disrupted cell survival but not proliferation.

Fig. 2. XBP1s controls IL-15–mediated NK cell survival in vitro and in vivo.

   [110]Fig. 2.
   [111]Open in a new tab

   (A) Percentages of annexin V^+ NK cells in the spleens from Xbp1^f/f
   and Xbp1^cKO mice treated with IL-15 (n = 6 per group). (B) Percentages
   of annexin V^+ NK cells in the spleens from Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO mice (n = 8 per group). (C) Splenic NK cells isolated
   from Il15^TgXbp1^f/f and Il15^Tg Xbp1^cKO mice were cultured in vitro
   in the presence of IL-15 (50 ng/ml), followed by cell counting with a
   trypan blue exclusion assay. (D) Live NK cells were counted at 4, 8,
   12, and 24 hours after IL-15 withdrawal. Data shown are the fold change
   from the data collected at 0 hours. (E) Representative plots (left) and
   summary data (right) of apoptotic (annexin V^+SYTOX Blue^+/−) NK cells
   from Xbp1^f/f and Xbp1^cKO mice 5 days after in vitro culture with
   IL-15 (50 ng/ml). (F) Representative plots (left) and summary data
   (right) of apoptotic (active caspase-3^+) NK cells from Xbp1^f/f and
   Xbp1^cKO mice 5 days after in vitro culture with IL-15 (50 ng/ml). (G)
   Representative plots (left) and summary data (right) of apoptotic
   (FAM-FLICA^+) NK cells from Xbp1^f/f and Xbp1^cKO mice 5 days after in
   vitro culture with IL-15 (50 ng/ml). Data are shown as means ± SD and
   were analyzed with an unpaired two-tailed t test (A, B, and E to G) or
   two-way ANOVA with the Holm-Šídák post-test (C and D). Data are
   representative of at least two independent experiments. *P < 0.05; **P
   < 0.01; ***P < 0.001; ****P < 0.0001.

   We then used the ectopic IL-15 model to further investigate the
   detailed mechanism by which XBP1s signaling controls NK cell survival.
   We isolated splenic NK cells from Xbp1^cKO mice and Xbp1^f/f mice and
   cultured them in the presence of IL-15 in vitro. The NK cells from the
   Xbp1^cKO mice showed significantly less growth than those from the
   Xbp1^f/f mice ([112]Fig. 2C). NK cells from the Xbp1^cKO mice that were
   expanded in vitro with IL-15 also exhibited significantly more rapid
   cell death after IL-15 was withdrawn from the environment than those
   from the Xbp1^f/f mice ([113]Fig. 2D). A CellTrace Violet (CTV)
   labeling assay and a 5-bromo-2′-deoxyuridine (BrdU)/7-aminoactinomycin
   D (7-AAD) incorporation assay showed that both the proliferation and
   division of NK cells were unimpaired by XBP1 deficiency during IL-15
   stimulation ([114]fig. S2, J and [115]K). However, NK cells from the
   Xbp1^cKO mice exhibited significantly more apoptosis than those from
   the Xbp1^f/f mice, as evidenced by annexin V staining ([116]Fig. 2E),
   active caspase-3 staining ([117]Fig. 2F), and the fluorochrome-labeled
   inhibitors of caspases (FLICA) assay ([118]Fig. 2G).

   Activated NK cells undergo activation-induced cell death mediated by
   the Fas-FasL pathway ([119]37, [120]38). NK cells also express
   TNF-related apoptosis-inducing ligand (TRAIL), which triggers cell
   death by receptor-mediated apoptosis ([121]39). To investigate whether
   the NK cell survival defect mediated by XBP1s deficiency is caused by
   FasL or TRAIL-induced cell apoptosis, we analyzed the expression levels
   of FasL and TRAIL in NK cells from Xbp1^f/f and Xbp1^cKO mice before
   and after stimulating with IL-15. We found that the expression levels
   of FasL and TRAIL were similar between NK cells from Xbp1^f/f and those
   from Xbp1^cKO mice with or without IL-15 stimulation ([122]fig. S3, A
   and [123]B). Using neutralizing antibodies to block FasL or TRAIL could
   not reduce NK cell apoptosis mediated by XBP1 deficiency in response to
   IL-15 ([124]fig. S3C). These results suggest that XBP1s-mediated NK
   cell survival is independent of FasL and TRAIL. Collectively, these
   murine studies demonstrate that XBP1s controls IL-15–mediated NK cell
   survival but not proliferation. This conclusion is consistent with our
   prior finding that XBP1s positively regulates IL-15–mediated cell
   survival in human NK cells ([125]26).

XBP1s maintains NK cell survival by targeting PIM-2 in mice

   To identify the downstream target genes of XBP1s in NK cells, we
   performed RNA sequencing (RNA-seq) analysis in NK cells purified from
   the spleens of Il15^TgXbp1^cKO mice and Il15^TgXbp1^f/f mice. We
   focused on splenic NK cells because NK cells isolated from the spleen
   had the highest expression levels of XBP1s when stimulated with IL-15
   compared with NK cells from the BM, liver, and lung ([126]fig. S4A). We
   also performed chromatin immunoprecipitation coupled with
   high-throughput DNA sequencing (ChIP-seq) in splenic NK cells purified
   from Il15^TgXbp1^f/f mice. Two independent biological replicates were
   performed. Overlapping analysis using BEDTools ([127]40) showed that
   4890 peaks were present in both replicates ([128]fig. S4B). The
   majority of the XBP1 binding peaks were enriched in the promoter
   regions (65.56%) and intergenic regions (21.35%) ([129]fig. S4, C and
   [130]D), indicating that it binds both promoters and potential
   enhancers. The search for enriched motifs using HOMER revealed a
   sequence element, 5′-CACGTCA-3′, as the highest score motif with P = 1
   × 10^−172, which is identical to the XBP1 binding motif defined from
   the JASPAR database and from mouse T helper 2 (T[H]2) cells ([131]fig.
   S4E). In addition, the Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis revealed that the peaks were mainly
   enriched in terms of “protein processing in endoplasmic reticulum,”
   “protein export,” “cell cycle,” and “apoptosis” ([132]fig. S4F), which
   is consistent with the known biological role of XBP1. Consistent with a
   prior study in T[H]2 cells, we also found strong binding peaks around
   the promoter of XBP1 itself in NK cells ([133]fig. S4G), suggesting
   that XBP1 has autoregulation ability.

   Gene Ontology analysis of RNA-seq data demonstrated that differentially
   expressed genes (DEGs) related mainly to apoptosis and the ER UPR
   ([134]Fig. 3A). Integration of RNA-seq data and ChIP-seq data showed 33
   target genes of XBP1s in NK cells ([135]fig. S4H). Among them, Pim2 was
   markedly decreased in NK cells from Il15^TgXbp1^cKO mice compared with
   those from Il15^TgXbp1^f/f mice ([136]Fig. 3B). Pim2 encodes a
   proto-oncogene that acts as a serine/threonine protein kinase
   ([137]41). Prior studies have shown that PIM-2 inhibits apoptosis and
   promotes cell survival in tumor cells and immune cells
   ([138]42-[139]44). However, it was not previously recognized as a
   target of XBP1s. We found significantly lower transcript and protein
   levels of PIM-2 in NK cells from Il15^TgXbp1^cKO mice compared with
   those from Il15^TgXbp1^f/f mice ([140]Fig. 3, [141]C and [142]D),
   suggesting that XBP1s may transactivate PIM-2 expression.

Fig. 3. XBP1s maintains NK cell survival by targeting PIM-2.

   [143]Fig. 3.
   [144]Open in a new tab

   (A) Top 10 Gene Ontology clusters of DEGs in RNA-seq data. (B) Heat
   maps of top 20 genes expressed differentially between Il15^TgXbp1^f/f
   and Il15^TgXbp1^cKO mice. (C and D) qPCR (C) and immunoblotting (D)
   showing expression levels of PIM-2 in NK cells from Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO mice (n = 4 per group). (E) Integrative Genomics Viewer
   tracks displaying the distribution of XBP1s-binding peaks across the
   Pim2 transcript. (F and G) Luciferase reporter assay showing that XBP1s
   activates Pim2 gene transcription in HEK-293T cells (F) and mouse NK
   cells (G). (H and I) Binding of XBP1s to the Pim2 promoter in
   IL-15–treated Xbp1^f/f NK cells (H) or IL-15–treated Xbp1^cKO NK cells
   (I) was determined by ChIP-qPCR. (J) Scheme for retroviral transduction
   and BM chimeras. BM cells from Il15^TgXbp1^f/f and Il15^TgXbp1^cKO mice
   were infected with retrovirus encoding Pim2 or empty vector (EV). The
   infected cells were transferred into lethally irradiated CD45.1
   recipient mice. After 8 weeks, mice were intraperitoneally injected
   with 2 μg of recombinant human IL-15 for 5 days. NK cells were isolated
   from the spleens of the chimeric mice and subjected to flow cytometry.
   (K) Summary data (left) and representative plots (right) of NK cells in
   chimeric mice (n = 3 per group). Data are shown as means ± SD and were
   analyzed with an unpaired two-tailed t test (C and F to I) or two-way
   ANOVA with the Holm-Šídák post-test (K). Data are representative of at
   least two independent experiments. ns, not significant; *P < 0.05; ***P
   < 0.001; ****P < 0.0001.

   When we analyzed the promoter region of Pim2, we identified a potential
   binding site for XBP1s ([145]fig. S4I). Moreover, a strong binding peak
   of XBP1s at the promoter region of Pim2 was observed ([146]Fig. 3E). A
   luciferase reporter assay in both human embryonic kidney (HEK)–293T
   cells and mouse NK cells showed that XBP1s activated Pim2 gene
   transcription ([147]Fig. 3, [148]F and [149]G). Moreover, ChIP-qPCR
   revealed that the binding of XBP1s to the promoter region of Pim2 was
   robustly enriched compared with a normal immunoglobulin G (IgG) control
   in IL-15–treated Xbp1^f/f NK cells, but no enrichment was observed in
   IL-15–treated Xbp1^cKO NK cells ([150]Fig. 3, [151]H and [152]I, and
   [153]fig. S4J). This suggests that XBP1s positively regulates Pim2
   transcription in mouse NK cells by directly binding to the Pim2
   promoter. To confirm that PIM-2 is a functional target of XBP1s in NK
   cells, we first collected BM cells from Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO mice and spin-infected them with retrovirus encoding
   Pim2. The infected cells were then transferred into lethally irradiated
   CD45.1 recipient mice. After 8 weeks, NK cells from recipients were
   assessed by flow cytometry ([154]Fig. 3J). As predicted, we found that
   ectopic expression of PIM-2 ([155]fig. S4K) compensated for XBP1
   deficiency in NK cells in vivo ([156]Fig. 3K).

An XBP1s–PIM-2 axis controls NK cell survival in humans

   To determine whether XBP1s also maintains the survival of human NK
   cells by targeting PIM-2, we first overexpressed XBP1s in human NK
   cells. This overexpression of XBP1s significantly increased PIM-2
   expression ([157]Fig. 4, [158]A to [159]C). We next found two binding
   sites for XBP1s in the promoter region of PIM2 ([160]Fig. 4D). Then, a
   luciferase reporter assay in HEK-293T cells and human NK cells showed
   that XBP1s activated PIM2 gene transcription in the human cells
   ([161]Fig. 4, [162]E and [163]F). ChIP-qPCR also showed that human
   XBP1s bound to the PIM2 promoter ([164]Fig. 4G and [165]fig. S4L).
   Moreover, overexpression of PIM-2 partly rescued NK cells after
   treatment with the XBP1s inhibitor 4μ8C in the presence of IL-15,
   increasing their survival ([166]Fig. 4, [167]H and [168]I).
   Collectively, our findings reveal a previously unknown XBP1s–PIM-2 axis
   that regulates NK cell survival in both humans and mice.

Fig. 4. XBP1s maintains cell survival by targeting PIM-2 in human NK cells.

   [169]Fig. 4.
   [170]Open in a new tab

   (A to C) qPCR (A) and immunoblotting (B and C) showing PIM-2 levels in
   human NK cells transduced with XBP1s lentivirus (n = 3 per group). (D)
   Scheme denoting putative XBP1s-binding sites in the PIM2 promoter. (E
   and F) Luciferase reporter assay showing that XBP1s activates PIM2 gene
   transcription in HEK-293T cells (E) and human NK cells (F). (G) Binding
   of XBP1s to the PIM2 promoter in human NK cells was determined by
   ChIP-qPCR. (H and I) Scheme for lentiviral transduction of PIM-2 into
   human NK cells and treatment with an inhibitor of IRE1α 4μ8C (50 μM)
   and IL-15 (50 ng/ml) (H). Live NK cell numbers were counted at the
   indicated time using a trypan blue exclusion assay [(I), n = 3 per
   group]. Data are shown as means ± SD and were analyzed with an unpaired
   two-tailed t test (A, C, and E to G) or one-way ANOVA with the
   Holm-Šídák post-test (I). Data are representative of at least two
   independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****P <
   0.0001.

PIM-2 stabilizes XBP1s through phosphorylation at Thr^58

   PIM-2 belongs to a family of serine/threonine protein kinases, and it
   functions as an oncogene by phosphorylating a wide range of cellular
   proteins ([171]45, [172]46). This led us to hypothesize that PIM-2
   might phosphorylate and then stabilize XBP1s, the process of which
   would create a positive feedback loop between PIM-2 and XBP1s in NK
   cells. We transiently overexpressed FLAG-tagged XBP1s and different
   doses of hemagglutinin (HA)–tagged PIM-2 in HEK-293T cells.
   Overexpression of PIM-2 increased protein levels of XBP1s in a
   dose-dependent manner ([173]Fig. 5A). However, PIM-2 did not enhance
   endogenous Xbp1s mRNA expression ([174]fig. S5A). Given that PIM-2
   increases protein levels of XBP1s without increasing its transcription,
   we examined whether PIM-2 affects the stability of the XBP1s protein.
   First, we transiently overexpressed FLAG-tagged XBP1s or FLAG-tagged
   XBP1s plus HA-tagged PIM-2 in HEK-293T cells. Then, we treated the
   cells with cycloheximide (CHX) to inhibit global protein translation.
   The presence of PIM-2 significantly delayed the degradation of XBP1s
   ([175]Fig. 5, [176]B and [177]C), indicating that PIM-2 increases the
   amount of XBP1s protein by enhancing stability. Because XBP1s
   degradation is mediated by ubiquitination and proteasomes ([178]12), we
   then examined the effect of coexpressing PIM-2 and XBP1s.
   Ubiquitination of XBP1s was reduced by the coexpression compared with
   that in control cells that expressed vector instead of PIM-2 ([179]Fig.
   5D). This finding indicates that PIM-2 inhibits the ubiquitination of
   XBP1s, thereby suppressing proteasomal degradation.

Fig. 5. PIM-2 stabilizes XBP1s protein through phosphorylation at Thr^58.

   [180]Fig. 5.
   [181]Open in a new tab

   (A) FLAG-tagged XBP1s and different doses of HA-tagged PIM-2 were
   transiently transfected into HEK-293T cells. XBP1s levels were analyzed
   by immunoblotting. (B and C) HEK-293T cells were transiently
   transfected with FLAG-tagged XBP1s or together with HA-tagged PIM-2 and
   further treated with CHX for the indicated time periods. XBP1s levels
   were analyzed by immunoblotting. Representative blots (B) and
   quantification of immunoblotting signals of XBP1s to actin levels (C)
   are shown. (D) HEK-293T cells were transiently transfected with the
   FLAG-tagged XBP1s and Ubb plasmids with or without the HA-tagged PIM-2
   plasmid. Ubiquitinylated XBP1s levels were analyzed by immunoblotting.
   (E) Immunoblotting for HA-tagged PIM-2 and FLAG-tagged XBP1s proteins
   after immunoprecipitation of FLAG from HEK-293T cells. (F)
   Immunoblotting for HA-tagged PIM-2 and FLAG-tagged XBP1s proteins after
   immunoprecipitation of HA from HEK-293T cells. (G) IL-2–expanded mouse
   NK cells were treated with IL-15 (50 ng/ml) for 24 hours. Cell lysates
   were immunoprecipitated with rabbit anti-XBP1s or IgG isotype control
   and immunoblotted with rabbit anti–PIM-2 or mouse anti-XBP1s. (H)
   Primary human NK cells were treated with IL-15 (50 ng/ml) for 24 hours.
   Cell lysates were immunoprecipitated with rabbit anti-XBP1s or IgG
   isotype control and immunoblotted with rabbit anti–PIM-2 or mouse
   anti-XBP1s. (I) Immunoblotting using a phospho-Thr/Ser–specific
   antibody to detect phosphorylated XBP1s levels after
   immunoprecipitation of FLAG from HEK-293T cells. (J and K) FLAG-tagged
   XBP1s and HA-tagged PIM-2 (WT or KD PIM-2, K61A) were transfected into
   HEK-293T cells. Immunoblotting using phospho-Thr–specific (J) or
   phospho-Ser–specific antibody (K) for detecting phosphorylated XBP1s
   levels after immunoprecipitation of FLAG from HEK-293T cells. (L)
   FLAG-tagged WT XBP1s or Thr^58 mutant (T58A) XBP1s and HA-tagged PIM-2
   were transfected into HEK-293T cells. Immunoblotting using
   phospho-Thr–specific antibody to detect phosphorylated XBP1s levels
   after immunoprecipitation of FLAG from HEK-293T cells. (M) HEK-293T
   cells were transiently transfected with FLAG-tagged XBP1s or T58A
   mutant XBP1s together with HA-tagged PIM-2 and further treated with CHX
   for 1 hour. XBP1s levels were analyzed by immunoblotting. Data are
   shown as means ± SD and were analyzed with two-way ANOVA with the
   Holm-Šídák post-test (C). Data are representative of at least two
   independent experiments. *P < 0.05; ****P < 0.0001.

   This reduced ubiquitination prompted us to hypothesize that PIM-2
   stabilizes XBP1s by phosphorylation. To test this idea, we transfected
   FLAG-tagged XBP1s and HA-tagged PIM-2 into HEK-293T cells. Reciprocal
   coimmunoprecipitation (co-IP) data revealed an interaction between
   XBP1s and PIM-2, because PIM-2 was coprecipitated with XBP1s ([182]Fig.
   5E) and vice versa ([183]Fig. 5F). In addition, XBP1s also precipitated
   endogenous PIM-2 in IL-15–treated mouse and human NK cells ([184]Fig.
   5, [185]G and [186]H), indicating that PIM-2 stabilizes XBP1s by direct
   modification. We then determined whether PIM-2 could phosphorylate
   XBP1s. Compared with the empty vector control, transfection of PIM-2
   increased serine/threonine phosphorylation levels of XBP1s, as detected
   by immunoblotting with phospho-Thr/Ser–specific antibodies ([187]Fig.
   5I). Thus, PIM-2 enhanced Ser/Thr phosphorylation of XBP1s and
   increased its stability.

   To determine which residues on XBP1s are phosphorylated by PIM-2, we
   cotransfected FLAG-tagged XBP1s and HA-tagged PIM-2 [wild type (WT) or
   kinase-dead (KD) PIM-2, K61A] into HEK-293T cells ([188]45). Compared
   with the PIM-2–KD or vector control, transfection of WT PIM-2 increased
   XBP1s phosphorylation on threonine residues, as detected by
   immunoblotting with a phospho-Thr–specific antibody ([189]Fig. 5J).
   However, PIM-2 did not affect phosphorylation levels of serine residues
   in XBP1s ([190]Fig. 5K). Because prior studies had identified a
   consensus substrate motif for PIM-2, R-X-R-H-X-pS/T ([191]47), we
   performed in silico analyses for potential PIM-2 substrate motifs in
   XBP1s. We identified one putative PIM-2 phosphorylation site (Thr^58)
   in XBP1s ([192]fig. S5B) and then found that mutating Thr^58 (T58A)
   significantly reduced PIM-2–induced XBP1s phosphorylation ([193]Fig.
   5L). To determine whether PIM-2 stabilizes XBP1s by phosphorylating it
   at Thr^58, we cotransfected XBP1s-WT or XBP1s-T58A with HA-tagged PIM-2
   and then treated with CHX. PIM-2 was not able to stabilize this mutant
   XBP1s as effectively as it stabilized WT XBP1s ([194]Fig. 5M). This
   suggests that Thr^58 is a critical phosphorylation site that
   contributes to the stabilization of XBP1s by PIM-2 kinase. Together,
   our data suggest that PIM-2 stabilizes XBP1s by phosphorylating it at
   Thr^58.

PIM-2 is essential for NK cell development and homeostasis in mice

   PIM-2 plays a critical role in regulating cell survival and
   proliferation in cancer cells ([195]44, [196]48, [197]49). Prior
   studies reported that PIM-2 negatively regulates Foxp3^+ regulatory T
   cells and T cell responses in transplantation and tumor immunity
   ([198]42, [199]50). We evaluated the role of PIM-2 in NK cell
   development and homeostasis using PIM-2 KO mice ([200]42, [201]51). We
   found that the percentages and absolute numbers of NK cells were
   significantly decreased in the spleens and BM from PIM-2 KO mice
   compared with those from WT mice ([202]fig. S5, C and [203]D), which is
   consistent with a prior report ([204]42), indicating that PIM-2
   positively affects NK cell homeostasis at steady state. To further
   explore whether PIM-2 affects NK cell development, we examined the
   hematopoietic stem cells (HSCs) and NK cell progenitors in the BM.
   PIM-2 deficiency significantly reduced the percentages and absolute
   numbers of HSCs, pre-NK cell progenitors (preNKPs), and refined NKPs
   (rNKPs) but markedly increased percentages and absolute numbers of
   common lymphoid progenitors (CLPs) in the BM ([205]fig. S5E),
   indicating that PIM-2 regulates NK cell development likely at the
   transition from CLP to preNKP stage. However, further studies are
   warranted to reveal the detailed mechanism by which PIM-2 regulates NK
   cell development and homeostasis.

XBP1 deficiency impairs NK cell effector functions and antitumor immunity

   We have shown that Xbp1^cKO mice displayed more tumor metastasis
   compared with Xbp1^f/f mice in association with decreased
   tumor-infiltrating NK cells ([206]Fig. 1C and [207]fig. S1C). However,
   whether XBP1 regulates NK cell effector functions remains undetermined
   in mice. We found that XBP1 deficiency impairs IFN-γ production by NK
   cells in the settings of cytokine costimulation with IL-12 plus IL-18
   or in the presence of a tumor cell line (YAC-1) ([208]fig. S6, A and
   [209]B). In contrast, NK cells from Xbp1^f/f and Xbp1^cKO mice have
   similar low baseline levels of IFN-γ production in the absence of this
   stimulation ([210]fig. S6C). We then assessed the functional capacity
   of XBP1-deficient NK cells on the Il15^Tg background. NK cells from
   Il15^TgXbp1^cKO mice produced less granzyme B and perforin compared
   with those from Il15^TgXbp1^f/f mice ([211]Fig. 6, [212]A and [213]B),
   and IFN-γ production was significantly decreased ([214]Fig. 6C). After
   activation with IL-15, moreover, NK cells from Il15^TgXbp1^cKO mice had
   less cytotoxicity against YAC-1 cells than those from Il15^TgXbp1^f/f
   mice ([215]Fig. 6D). Also, we found a larger number of B16-F10 melanoma
   tumor metastases in Il15^TgXbp1^cKO mice than in Il15^TgXbp1^f/f mice
   ([216]Fig. 6E). Last, an NK depletion assay showed that the antitumor
   effect of XBP1s was NK cell–dependent ([217]Fig. 6F). Collectively, our
   findings indicate that XBP1 also controls NK cell effector functions
   upon certain types of activation.

Fig. 6. XBP1s deficiency impairs NK cell effector functions and antitumor
immunity.

   [218]Fig. 6.
   [219]Open in a new tab

   (A to C) Expression levels of granzyme B (A), perforin (B), and IFN-γ
   (C) in NK cells from Il15^TgXbp1^f/f and Il15^TgXbp1^cKO mice (n = 8
   per group). (D) Poly(I:C)-activated NK cells were isolated from the
   spleen and cocultured with YAC-1 cells at a ratio of 5:1, 2.5:1, and
   1.25:1 for 4 hours. Cytotoxicity of NK cells was evaluated by standard
   ^51Cr release assays (n = 3 per group). (E) Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO mice were intravenously injected with B16F10 cells (2 ×
   10^5). Fourteen days after injection, mice were euthanized for
   postmortem analysis. Quantification of total metastatic nodules in the
   lung and the gross morphology of individual lung lobes are shown (n = 5
   per group). (F) Il15^TgXbp1^f/f and Il15^TgXbp1^cKO mice were
   intravenously injected with B16F10 cells (2 × 10^5) together with
   intraperitoneal injection of anti-NK1.1 antibody or IgG antibody (200
   μg per mouse) on days 0 and 7 after tumor inoculation. The data show
   quantification of total metastatic nodules in the lung on day 14 (n = 5
   per group). Data are shown as means ± SD and were analyzed with an
   unpaired two-tailed t test (A to E) or two-way ANOVA with the
   Holm-Šídák post-test (F). Data are representative of at least two
   independent experiments. **P < 0.01; ***P < 0.001.

XBP1s promotes IFN-γ production by recruiting T-bet to the promoter region of
Ifng

   We previously showed that XBP1s positively regulates human NK cell
   effector functions by enhancing the expression of IFN-γ and granzyme B
   ([220]26). However, the mechanism has not been fully unraveled. We
   therefore analyzed the expression of genes encoding molecules related
   to NK cell effector functions, including Ifng (IFN-γ), Gzmb (granzyme
   B), and Prf1 (perforin), in NK cells from Il15^TgXbp1^f/f or
   Il15^TgXbp1^cKO tumor-bearing mice. mRNA levels of Ifng and Gzmb, but
   not Prf1, were significantly lower in NK cells from Il15^TgXbp1^cKO
   tumor-bearing mice than in those fromIl15^Tg− Xbp1^f/f tumor-bearing
   mice ([221]Fig. 7, [222]A to [223]C). These murine data are consistent
   with our findings with human NK cells, indicating that XBP1s controls
   NK cell effector functions, most likely by positively regulating IFN-γ
   and granzyme B.

Fig. 7. XBP1s promotes IFN-γ production by recruiting T-bet to the promoter
region of Ifng.

   [224]Fig. 7.
   [225]Open in a new tab

   (A to C) qPCR analysis of mRNA levels of Ifng (A), Gzmb (B), and Prf1
   (C) in NK cells isolated from tumor tissues of Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO tumor-bearing mice. (D) Luciferase reporter assay
   showing transcriptional activity of Gzmb after XBP1s overexpression.
   (E) Mouse XBP1s and T-bet were transfected alone or together into
   HEK-293T cells. The transcriptional activity of Ifng was examined by
   luciferase reporter assays. (F) Human XBP1s and T-BET were transfected
   alone or together into HEK-293T cells. The transcriptional activity of
   IFNG was examined by luciferase reporter assays. (G) Mouse XBP1s and
   T-bet were transfected alone or together into mouse NK cells. The
   transcriptional activity of Ifng was examined by luciferase reporter
   assays. (H) Human XBP1s and T-BET were transfected alone or together
   into human NK cells. The transcriptional activity of IFNG was examined
   by luciferase reporter assays. (I) IL-2–expanded mouse NK cells were
   treated with IL-15 (50 ng/ml) for 24 hours. Cell lysates were
   immunoprecipitated with rabbit anti–T-bet or IgG isotype control and
   immunoblotted with mouse anti-XBP1s or rabbit anti–T-bet. (J) Primary
   human NK cells were treated with IL-15 (50 ng/ml) for 24 hours. Cell
   lysates were immunoprecipitated with rabbit anti–T-bet or IgG isotype
   control and immunoblotted with mouse anti-XBP1s or rabbit anti–T-bet.
   (K) The cross-linked DNA-protein complexes in IL-15–activated mouse NK
   cells were separately immunoprecipitated using antibodies against XBP1s
   or T-bet. The binding of XBP1/T-bet complex to the Ifng promoter was
   determined by ChIP-qPCR. (L) The cross-linked DNA-protein complexes in
   IL-15–activated human NK cells were separately immunoprecipitated using
   antibodies against XBP1s or T-BET. The binding of XBP1/T-BET complex to
   the IFNG promoter was determined by ChIP-qPCR. (M and N) Scheme for
   lentiviral transduction of T-bet into mouse NK cells (M). Data shown
   are summary data (left) and representative plots (right) that came from
   quantifying IFN-γ production by transduced NK cells [(N), n = 4 per
   group]. Data are shown as means ± SD and were analyzed with an unpaired
   two-tailed t test (A to D) or one-way ANOVA with the Holm-Šídák
   post-test (E to H, K, and L) or two-way ANOVA with the Holm-Šídák
   post-test (N). Data are representative of at least two independent
   experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****p < 0.0001.

   Our previous study showed that XBP1s binds directly to the GZMB
   promoter to enhance the transcription of GZMB via T-BET in human NK
   cells ([226]26). To determine whether XBP1s behaves similarly in mice,
   we performed a luciferase reporter assay, finding that mouse XBP1s
   activated Gzmb gene transcription ([227]Fig. 7D), as in human NK cells.
   We also found a strong binding peak of XBP1 in the promoter region of
   Gzmb in mouse NK cells ([228]fig. S7A). Within 2 kb of the promoter
   upstream of the mouse or human Ifng/IFNG gene transcription start site,
   11 and 4 XBP1 binding sites were identified, respectively. However,
   very weak binding peaks were found around the promoter region of Ifng
   in mouse NK cells ([229]fig. S7B). Using a luciferase reporter assay,
   we found that XBP1s did not activate Ifng/IFNG transcription in either
   mouse or human cells ([230]fig. S7, C and [231]D), suggesting that
   XBP1s does not directly regulate Ifng/IFNG expression.

   We therefore turned our attention to the transcription factor T-bet
   (encoded by Tbx21), the master regulator of IFN-γ expression in T cells
   and NK cells ([232]52, [233]53). We previously showed that XBP1s can
   interact with T-BET in human NK cells and that it regulates the
   transcription of GZMB by binding to that gene′s promoter ([234]26). To
   investigate whether mouse XBP1s could directly promote the
   transcription of Tbx21, we used luciferase reporter assays and found
   that XBP1s does not promote transcription by interacting with the
   Tbx21/TBX21 promoter in either mouse or human cells ([235]fig. S7, E
   and [236]F). ChIP-seq data also demonstrated that XBP1 has no binding
   peaks around the promoter region of Tbx21 in mouse NK cells ([237]fig.
   S7G). Moreover, XBP1s also did not bind to the Tbx21 enhancer element
   ([238]fig. S7H), a conserved sequence 12 kb upstream of the
   transcription start site that regulates Tbx21 expression ([239]54).
   Therefore, XBP1s does not appear to promote IFN-γ expression by
   increasing T-bet/T-BET expression.

   We therefore proposed that XBP1s might promote IFN-γ expression via the
   same mechanism that it uses to recruit T-bet to the promoter of Ifng as
   it regulates the expression of human GZMB ([240]26). After we
   cotransfected mouse XBP1s and T-bet into HEK-293T cells, we found
   significantly higher transcriptional activity of Ifng than in cells
   transfected with XBP1s alone or T-bet alone ([241]Fig. 7E). A similar
   difference was observed between human XBP1s and T-BET ([242]Fig. 7F).
   Moreover, we performed luciferase reporter assays in mouse and human NK
   cells and observed similar results ([243]Fig. 7, [244]G and [245]H).
   Co-IP followed by immunoblotting using antibody to XBP1s showed that
   XBP1s interacts with T-bet/T-BET in IL-15–activated mouse/human NK
   cells ([246]Fig. 7, [247]I and [248]J). To further prove our hypothesis
   that XBP1 promotes IFN-γ production by recruiting T-bet/T-BET to the
   promoter region of Ifng/IFNG, we used antibodies against XBP1s or
   T-bet/T-BET to separately precipitate the cross-linked DNA-protein
   complexes in IL-15–activated mouse and human NK cells and analyzed the
   binding of the XBP1s/T-bet complex to the Ifng/IFNG promoter by
   ChIP-qPCR. Our results revealed that the DNA immunoprecipitated by
   anti-XBP1s or anti–T-bet/T-BET contained the Ifng/IFNG promoter. The
   Ifng/IFNG promoter DNA amount immunoprecipitated by anti-XBP1s or
   anti–T-bet/T-BET was significantly higher than that immunoprecipitated
   by an IgG control antibody in both mouse/human NK cells ([249]Fig. 7,
   [250]K and [251]L). The binding of T-bet and T-BET to the Ifng or IFNG
   promoter, respectively, served as control ([252]Fig. 7, [253]K and
   [254]L). Last, to confirm that T-bet is a functional target of XBP1s in
   mouse NK cells, we spin-infected NK cells from Xbp1^cKO mice with
   lentivirus encoding T-bet ([255]Fig. 7M). As predicted, we found that
   ectopic expression of T-bet partly compensated for the defect in IFN-γ
   production when XBP1s-deficient NK cells were treated with IL-12 plus
   IL-15 ([256]Fig. 7N). Together, our data suggest that XBP1s promotes
   IFN-γ production by recruiting T-bet to the promoter region of Ifng.

DISCUSSION

   IL-15 is a dominant regulator of NK cell survival and effector
   functions ([257]4, [258]33, [259]55, [260]56). However, the molecular
   mechanisms that underlie this remain poorly understood. In this study,
   we provide genetic evidence that IL-15–XBP1s signaling controls NK cell
   survival and effector functions through two targets: PIM-2 and T-bet.
   XBP1s binds to the promoter of Pim2 and induces its expression to
   maintain NK cell survival. In turn, PIM-2 stabilizes XBP1s protein
   through phosphorylation, creating a positive feedback loop. This
   permits the downstream interaction between XBP1s and T-bet, thereby
   promoting granzyme B and IFN-γ production ([261]fig. S8). Thus, our
   present study, together with our prior report ([262]26), confirms that
   XBP1s plays an indispensable role in IL-15–mediated cell survival and
   effector functions in NK cells.

   XBP1s has a short half-life because it is rapidly degraded by
   proteasomes. In human NK cells, IL-15 stabilizes XBP1s by
   phosphorylating AKT, which is upstream of XBP1s. Phosphorylated AKT
   then inhibits the ubiquitination of XBP1s, maintaining its stability
   ([263]26). Here, we found that the stability of XBP1s can also be
   regulated by its downstream target, PIM-2, which creates a positive
   feedback loop. In particular, PIM-2 interacts with XBP1s physically and
   phosphorylates the latter at the Thr^58 residue. Collectively, our
   findings reveal that NK cells have at least two mechanisms for
   stabilizing XBP1s after the cells are activated by IL-15. These
   mechanisms are responsible for the survival and functioning of
   IL-15–activated NK cells.

   In resting cells, in the absence of ER stress, XBP1 is maintained in an
   inactive state. It has been reported that XBP1 is not required for
   homeostasis, development, or effector function of T cells or dendritic
   cells (DCs) at steady state ([264]21, [265]57) but is essential to
   control the function of T cells or tumor-associated DCs under
   conditions of tumor challenge ([266]21, [267]22). We and others have
   shown that XBP1s is inactivated in resting NK cells but can be
   activated by IL-15 and MCMV infection ([268]26, [269]27). In the
   current study, although we confirm that XBP1 does not affect NK cell
   homeostasis and maturation at steady state, our findings reveal that
   the endogenous XBP1 is important for NK cell survival and effector
   functions under conditions of excessive IL-15, tumor challenge, and
   inflammation.

   It was previously reported that XBP1s drives NK cells to proliferate
   and expand after MCMV infection by transactivating c-MYC ([270]27).
   However, our current study demonstrates that XBP1s uses PIM-2 signaling
   to control NK cell survival but not proliferation in response to IL-15.
   This indicates that activation of IRE1α-XBP1 signaling not only is
   regulated by different upstream regulators but also has different
   targets depending on the initial stimulus. Therefore, it is necessary
   to identify and distinguish the exact target genes of XBP1s in NK cells
   under specific conditions. Furthermore, overexpression of PIM-2 cannot
   completely rescue the loss of NK cells due to deficiency of XBP1; we
   acknowledge that other mechanisms beyond PIM-2 may be involved in
   XBP1s-mediated survival of NK cells. Last, our study indicates that
   PIM-2 is essential for NK cell homeostasis in mice likely by regulating
   NK cell development at the transition from the CLP stage to the preNKP
   stage. Further studies are warranted to reveal the associated
   mechanism.

   Recently, NK cell–based therapies, such as CAR-NK, have attracted
   attention because of their allogeneic, off-the-shelf potential for
   treating cancer ([271]5, [272]58). Although there are no direct binding
   sites for XBP1s on the Ifng promoter, we found that XBP1s interacts
   with T-bet to enhance T-bet’s binding to that promoter, thereby
   promoting NK cell activity. We therefore believe that future studies
   are warranted to evaluate whether manipulating NK cells to overexpress
   XBP1 could be a practical and effective approach for treating certain
   forms of cancer.

   In conclusion, we show here that XBP1s is a crucial transcription
   factor that acts downstream of IL-15 signaling to control two of NK
   cells’ most important features—survival and effector functions—in
   animal models and human cells. Thus, targeting the IL-15–XBP1s–PIM-2
   signaling pathway may be a promising strategy for advancing NK
   cell–based therapies against tumors.

MATERIALS AND METHODS

Study design

   The aim of this study was to explore the role and detailed mechanisms
   by which IL-15–XBP1s signaling controls survival and effector functions
   of NK cells. This was achieved using Xbp1^cKO mice and Il15^TgXbp1^cKO
   compound mice; a tumor melanoma model; an EAE model; BM chimeras; and
   several biochemistry experiments, such as RNA-seq, ChIP assay and co-IP
   assay, flow cytometric analysis, quantitative real-time reverse
   transcription PCR (RT-PCR) analysis, ex vivo cytotoxicity assay, etc.
   Outcomes were presented as NK cell percentages, numbers, survival, and
   effector functions, etc. All data points generated were included in the
   analysis. For in vitro studies, as least three independent experiments
   were performed. For in vivo studies, mice were randomly assigned for
   experiments. The sample size is specified in each figure legend. The
   experiments were performed unblinded.

Mice

   Xbp1^fl/fl mice were generated in the laboratory of L. H. Glimcher
   (Harvard Medical School) as previously described ([273]30). Ncrl-iCre
   mice were provided by E. Vivier (Centre d′Immunologie de
   Marseille-Luminy) ([274]31). IL-15 Tg (Il15^Tg) mice were generated by
   our group and backcrossed into the C57BL/6 background ([275]32,
   [276]33). Il15^TgXbp1^cKO mice were generated by crossing Il15^Tg mice
   with Xbp1^cKO mice in the animal facility at City of Hope. NCI
   B6-Ly5.1/Cr (CD45.1) mice were purchased from Charles River
   Laboratories. PIM-2 KO mice were provided by X. Yu (Medical College of
   Wisconsin) and A. Berns (Netherlands Cancer Institute) ([277]42). All
   experimental mice were bred and maintained under specific pathogen–free
   conditions. Six- to 12-week-old male or female mice were used for the
   experiments. Age- and sex-matched Cre-negative littermates were used as
   controls. All animal experiments were approved by the City of Hope
   Institutional Animal Care and Use Committee.

Human peripheral blood samples

   Peripheral blood samples from de-identified healthy donors were
   obtained from the Michael Amini Transfusion Medicine Center of City of
   Hope National Medical Center under institutional review board–approved
   protocols.

Lung metastatic melanoma model

   Cells from the murine melanoma cell line B16F10 were cultured in
   Dulbecco’s modified Eagle′s medium (Gibco) supplemented with 10% fetal
   bovine serum (FBS; Gibco). B16F10 cells (2 × 10^5) were injected
   intravenously into mice. Fourteen days after injection, the mice were
   euthanized for postmortem analysis. Metastatic nodules in the lung were
   analyzed macroscopically and counted. For NK cell depletion, mice were
   injected intraperitoneally with 200 μg of anti-NK1.1 antibody (clone
   PK136, Bio X Cell) on days 0 and 7 after tumor inoculation.

EAE model

   EAE was induced in 12-week-old female Xbp1^f/f and Xbp1^cKO mice using
   the Hooke Kit MOG[35–55]/CFA Emulsion PTX kit (catalog no. EK-2110,
   Hooke Laboratories) according to the manufacturer’s protocol. In brief,
   mice were immunized with an emulsion of myelin oligodendrocyte
   glycoprotein peptide (MOG[35–55]) in complete Freund’s adjuvant (CFA)
   by subcutaneous injection into two different sites (100 μl of emulsion
   per site) on each hind flank, followed by intraperitoneal
   administration of 100 μl of pertussis toxin (PTX) in phosphate-buffered
   saline (PBS) after 2 hours of MOG[35–55] immunization. PTX (100 μl) was
   given again 24 hours later. The mice were monitored for clinical
   symptoms of EAE daily from 9 to 18 days after immunization. EAE was
   scored on a scale of 0 to 5 based on the grading system for clinical
   assessment of EAE as described previously ([278]59).

Flow cytometry

   Single-cell suspensions were prepared from the BM, spleens, lungs, and
   tumor tissues of mice as described previously ([279]33). The following
   fluorescence dye–labeled antibodies were used in this study: BV510-CD3ε
   (145-2C11, BioLegend), phycoerythrin (PE)–CD3ε (145-2C11, BioLegend),
   PE/Cy7-CD3ε (145-2C11, BioLegend), BV510-NK1.1 (PK136, BioLegend),
   allophycocyanin (APC)–Cy7-NK1.1(PK136, BioLegend), APC/Fire 750-NK1.1
   (PK136, BioLegend), fluorescein isothiocyanate (FITC)–NKp46 (29A1.4,
   BioLegend), BV421-CD11b (M1/70, BioLegend), APC/Cy7-CD27 (LG.3A10,
   BioLegend), PE–IFN-γ (XMG1.2, BioLegend), PE/Cy7–granzyme B (QA16A02,
   BioLegend), APC-perforin (S16009A, BioLegend), PE–annexin V
   (BioLegend), Alexa Fluor 488–Ki67 (11F6, BioLegend), BUV395-Ki67 (B56,
   BD Biosciences), PE-FasL (MFL3, BioLegend), PE-TRAIL (N2B2, BioLegend),
   FITC–c-kit (2B8, BioLegend), BV510–Sca-1 (D7, BioLegend), PE-FLT3
   (A2F10.1, BD Biosciences), and BV-786-CD122 (5H4, BD Biosciences). To
   evaluate mouse NK cell proliferation, we labeled cells with 5 μM
   CellTrace Violet (Invitrogen) according to the manufacturer’s protocol.
   Intracellular staining of Ki67 was performed by fixing and
   permeabilizing with the Foxp3/Transcription Factor Staining Kit
   (eBioscience). In vitro BrdU labeling was performed using the FITC BrdU
   Flow Kit (BD Biosciences) according to the manufacturer’s protocol. In
   vivo EdU labeling was performed using the Click-iT EdU Alexa Fluor 488
   Flow Cytometry Assay Kit (Invitrogen) according to the manufacturer’s
   protocol. In brief, mice were intraperitoneally injected with 2 mg of
   EdU in PBS, and EdU labeling was examined 16 hours later using the
   Click-iT EdU Assay kit. For IFN-γ staining, cells were stimulated with
   IL-12 (10 ng/ml) plus IL-18 (10 ng/ml) or YAC-1 cells at the
   effector-to-target ratio of 10:1 for 16 hours (BioLegend). Cells were
   stained with the indicated cell surface markers and fixed/permeabilized
   using the Fixation/Permeabilization Kit (BD Biosciences). Flow
   cytometry analysis was performed on BD LSRFortessa X-20 (BD
   Biosciences). Data were analyzed using NovoExpress Software (Agilent
   Technologies). The antibodies used are listed in [280]table S1.

NK cell isolation, culture, transduction, and treatment

   Mouse NK cells were isolated from the spleens of mice using the EasySep
   Mouse NK Cell Isolation Kit (STEMCELL Technologies) as previously
   described ([281]33). Human NK cells were enriched using RosetteSep
   Human NK Cell Enrichment Cocktail (STEMCELL Technologies) and
   Ficoll-Paque (GE Healthcare). Both mouse and human NK cells were
   cultured in RPMI 1640 with 10% heat-inactivated FBS (Gibco) at 37°C in
   a 5% CO[2] humidified incubator in the presence of IL-2 (100 U/ml;
   catalog no. Bulk Ro 23-6019, National Cancer Institute). For human NK
   cell transduction, human XBP1s cDNA or PIM2 cDNA was cloned into the
   pCDH-CMV-MCS-EF1-copGFP lentivirus vector (System Biosciences). Empty,
   XBP1s-expressing, or PIM-2–expressing lentiviruses were produced in
   HEK-293T cells with psPAX2 packaging (Addgene) and pMD2.G envelope
   plasmid DNA (Addgene) using PEI MAX (Polysciences) according to the
   manufacturer’s instructions. Concentrated lentivirus was added to NK
   cells cultured in RPMI medium 1640 supplemented with 10% FBS and
   polybrene (10 μg/ml) with 2000g centrifugation for 2 hours. Cells were
   cultured for an additional 48 hours. Green fluorescent protein
   (GFP)–positive cells were sorted using BD FACSAria Fusion (BD
   Biosciences). Sorted GFP-positive NK cells were pretreated with 4μ8C
   (50 μM/ml) for 1 hour, washed twice with RPMI 1640, and then cultured
   with IL-15 (50 ng/ml). Live cells were counted on days 3, 5, and 7
   after IL-15 treatment. For mouse NK cell transduction, mouse Tbx21 cDNA
   was cloned into the pCDH-CMV-MCS-EF1-copGFP lentivirus vector (System
   Biosciences). Empty or T-bet–expressing lentiviruses were produced as
   described above. Mouse NK cells were transduced with concentrated
   lentivirus through centrifugation. GFP-positive NK cells were sorted 48
   hours after transduction and treated with IL-12 (10 ng/ml) plus IL-15
   (10 ng/ml) for 24 hours. The cells were then collected for flow
   cytometry analysis.

In vivo mouse treatment with IL-15

   Mice were intraperitoneally injected with 2 μg of recombinant human
   IL-15 (catalog no. 745101, National Cancer Institute) for 5 days. They
   were then euthanized for flow cytometric analysis.

Quantitative real-time RT-PCR

   RNA was isolated using the RNeasy Mini Kit (QIAGEN) and then
   reverse-transcribed to cDNA with the PrimeScript RT Reagent Kit with
   gDNA Eraser (Takara Bio) following the manufacturer’s instructions.
   mRNA expression levels were analyzed using SYBR Green PCR Master Mix
   and the QuantStudio 12K Flex Real-Time PCR System (both from Thermo
   Fisher Scientific). Primer sequences are listed in [282]table S2.

Immunoblotting

   Immunoblotting was performed according to standard procedures as
   previously described ([283]33). The following antibodies were used:
   Xbp1s (ab220783, Abcam), Xbp1s [catalog no. 40435, Cell Signaling
   Technology (CST)], XBP1s (647502, BioLegend), Pim-2 (4730, CST), Pim-2
   (ab129057, Abcam), T-bet (13232, CST), anti-FLAG M2-peroxidase
   [horseradish peroxidase (HRP)] (A8592, Sigma-Aldrich),
   anti-HA–peroxidase antibody (H6533, Sigma-Aldrich), ubiquitin (58395,
   CST), phosphoserine/threonine-phenylalanine antibody (9631, CST),
   phosphothreonine (42H4) (9386, CST), phosphoserine antibody (PSR-45)
   (NB600-405, Novus Biologicals), β-actin (catalog no. 66009-1-Ig,
   Proteintech), and β-tubulin (2128, CST). The protein bands were
   detected using a SuperSignal West Pico Plus chemiluminescence ECL kit
   (Thermo Fisher Scientific) and imaged with FluorChem E system (Protein
   Simple) or detected with IRDye secondary antibody (Li-COR Biosciences)
   and visualized with Odyssey CLx Imager (Li-COR Biosciences).
   Densitometric analysis was performed to quantify intensity of
   immunoblot bands using ImageJ. The antibodies used are listed in
   [284]table S1.

RNA sequencing

   Total RNA was isolated from sorted purified splenic NK cells from
   Il15^TgXbp1^f/f and Il15^TgXbp1^cKO mice using the RNeasy Mini Kit
   (QIAGEN). RNA libraries were generated using the KAPA Hyper Prep Kit
   (KAPA Biosystems). Sequencing was performed at the City of Hope
   Genomics Facility on an Illumina HiSeq2500 machine with single-read
   50–base pair (bp) mode. Sequencing reads were mapped to the mouse
   genome using HISAT2 (version v101). Transcript counts were generated
   using HTSeq, and differential gene analysis was performed using DESeq2.
   DEGs from RNA-seq are listed in [285]table S3.

Plasmids

   An Xbp1s cDNA open reading frame clone with FLAG-tag at the C terminus
   was purchased from GenScript (clone ID: OMu19236). Amplified XBP1s cDNA
   was cloned into the pCDH-CMV-MCS-EF1α-copGFP lentivirus vector (System
   Biosciences) to generate pCDH-XBP1s-FLAG. XBP1s (T58A) with a
   C-terminal FLAG-tag was synthesized by Twist Bioscience and cloned into
   a pCDH vector. Mouse Pim2 was PCR-amplified from mouse NK cells to add
   an HA-tag at the N terminus. Then, it was inserted into the pCDH
   vector. Human PIM2 and mouse Tbx21 were cloned into the pCDH vector. KD
   PIM-2 (K61A) with an N-terminal HA-tag was synthesized by Twist
   Bioscience and cloned into the pCDH vector. The pCMV-8xHis-Ub plasmid
   was purchased from Addgene (#107392).

Luciferase reporter assay

   Mouse Pim2, Gzmb, and Ifng and human PIM2 and IFNG promoter regions
   ranging from −2000 to +100 bp of the transcription start site were
   cloned into a pGL4-basic luciferase reporter vector (Promega) to
   generate the pGL4 reporter plasmid. The human and mouse TBX21/Tbx21
   reporter plasmids and the Tbx21 enhancer reporter plasmid were
   generated as previously described ([286]60). HEK-293T cells purchased
   from the American Type Culture Collection (ATCC) were cotransfected
   with the reporter plasmids and with XBP1s or T-bet overexpression
   plasmids or empty vector together with a pRL-TK renilla reporter
   plasmid (Promega) to normalize transfection efficiency. The cells were
   harvested for lysis 24 hours after transfection, and luciferase
   activity was quantified fluorimetrically with the Dual-Luciferase
   System (Promega). Luciferase reporter assay in primary NK cells was
   performed as previously described ([287]61, [288]62). IL-2–expanded
   mouse NK cells were starved for 12 hours before transfection. The
   promoter and/or expression plasmids (2 μg each) and pRL-TK (500 ng)
   were electroporated into 5 × 10^6 starved mouse NK cells or primary
   human NK cells with the Amaxa Human NK Cell Nucleofector Kit (Lonza)
   and an Amaxa nucleofector device (Amaxa) with program U-001 according
   to the manufacturer’s instructions. Cells were harvested at 4 hours
   after transfection, and lysates were analyzed for luciferase activity
   using the Dual-Luciferase System (Promega).

ChIP-seq and ChIP-qPCR

   ChIP assays were carried out using a Pierce Magnetic ChIP Kit (catalog
   no.26157, Thermo Fisher Scientific) according to the manufacturer’s
   instructions. Briefly, 5 × 10^6 purified mouse primary NK cells or
   human NK cells were pretreated with IL-15 (50 ng/ml) for 1 hour. The
   cells were then cross-linked with fresh 1% formaldehyde (Thermo Fisher
   Scientific) for 10 min at room temperature (RT) and quenched with 125
   mM glycine for 5 min at RT. Cells were washed three times in cold PBS
   containing protease inhibitors. Nuclei were isolated using the
   SimpleChIP Enzymatic Chromatin IP Kit (#91820, CST), and chromatin was
   sheared on ice using a Fisher sonic dismembrator with the setting of 10
   cycles of 4 s “on” and 8 s “off.” The input fraction (2%) was saved,
   and the remaining sheared chromatin was used for ChIP with a rabbit
   anti-Xbp1s antibody (catalog no. 40435, CST) or rabbit anti–T-bet
   antibody (97135, CST) or an IgG isotype control (Millipore) in 1× ChIP
   buffer with rotation overnight at 4°C. Protein G magnetic beads were
   added to each immunoprecipitation and incubated for 2 hours at 4°C with
   rotation. The chromatin bead antibody complexes were then washed three
   times with low-salt wash buffer and one-time with high-salt wash
   buffer. Chromatin was then eluted using 1% SDS in Tris-EDTA.
   Cross-linking was reversed by incubation with 5 M NaCl and proteinase K
   at 65°C overnight. DNA was purified with a ChIP DNA Clean &
   Concentrator kit (Zymo Research). ChIP-seq libraries were prepared
   using a KAPA Hyper DNA kit (Roche) and sequenced on an Illumina NovaSeq
   6000 platform with 100-bp paired-end reads at the Translational
   Genomics Research Institute (TGen).

   Paired-end reads were obtained for the two replicates. All reads were
   trimmed with Trim Galore v0.6.7 and subjected to quality control with
   FastQC v0.11.9 before and after adaptor trimming ([289]63). After
   trimming, the reads were then mapped to the mouse genome (GRCm38/mm10)
   using Bowtie2 v2.4.5 ([290]64). Peak calling for each ChIP replicate to
   the input DNA was performed using MACS2 v2.2.7.1 ([291]65). The peaks
   were annotated to mm10 genomic features using the annotatePeaks.pl
   function from HOMER v4.11.1, and the motifs enriched in ChIP peaks were
   analyzed with the findMotifsGenome.pl function from HOMER v4.11
   ([292]http://homer.ucsd.edu/homer/motif/). ChIP peaks were visualized
   using Integrative Genomics Viewer software ([293]www.igv.org). KEGG
   pathway enrichment analysis of peaks was performed using DAVID
   ([294]https://david.ncifcrf.gov/). The DNA immunoprecipitated by the
   indicated antibodies was analyzed with qPCR. The sequences of all
   primers are listed in [295]table S1.

Coimmunoprecipitation

   Cells were lysed in lysis buffer [10% glycerol, 50 mM Hepes-KOH (pH
   7.5), 100 mM KCl, 2 mM EDTA, 0.1% NP-40, 10 mM NaF, 0.25 mM Na[3]VO[4],
   50 mM β-glycerophosphate, 2 mM dithiothreitol, and 1× protease and
   phosphatase inhibitor cocktail] and were incubated on ice for 15 min.
   Cell lysates were centrifuged at 14,000g for 20 min at 4°C, and
   supernatants were collected. Cell lysates (about 5 mg per sample) were
   incubated with antibody (5 μg per sample) or 20 μl of Pierce Anti-HA
   Magnetic Beads (88836, Thermo Fisher Scientific) or Anti-FLAG M2
   Magnetic Beads (M8832, Sigma-Aldrich) overnight at 4°C with gentle
   rotation. Dynabeads Protein A (20 μl) (10001D, Invitrogen) was added to
   the tubes, which were rotated at RT for 1 hour. Beads were separated
   out with a magnet and washed three times with cold lysis buffer. Fifty
   microliters of 50 mM glycine (pH 2.8) were added to the beads, which
   were incubated at RT on a rotator for 10 min to elute the protein. The
   beads were magnetically separated, and the supernatants containing the
   target antigen were saved. Ten microliters of neutralization buffer
   [0.5 M Tris (pH 7.5) and 1.5 M NaCl (pH 8.0) solution] were added for
   each 50 μl of eluate. The eluate was resuspended in 4× LDS Sample
   Buffer (Invitrogen) and incubated at 70°C for 10 min. The supernatants
   were then collected and used for immunoblotting.

Protein degradation assay

   HEK-293T cells were transfected with XBP1s-FLAG alone or with PIM-2–HA.
   After 24 hours of transfection, the cells were treated with the
   translation inhibitor CHX (20 μg/ml, Sigma-Aldrich). Cells were
   collected at 0.25, 0.5, 1, 2, and 4 hours after CHX incubation. Cells
   without treatment were dubbed 0 hours. Protein levels of XBP1s were
   determined by immunoblotting and subsequently quantified with ImageJ
   software.

Ex vivo cytotoxicity assay

   Ex vivo cytotoxicity of NK cells was evaluated by standard ^51Cr
   release assays. The mouse T cell lymphoma cell line YAC-1 (ATCC) was
   used for target cells. Mice were intraperitoneally injected with
   polyinosinic:polycytidylic acid [poly(I:C); 200 μg/mice] for 18 hours.
   Poly(I:C)-activated NK cells were isolated from the spleen using the
   EasySep Mouse NK Cell Isolation Kit (STEMCELL Technologies). Purified
   NK cells were cocultured with target cells at a ratio of 5:1, 2.5:1,
   and 1.25:1 in the presence of IL-2 (50 U/ml).

Retroviral transduction and BM chimeras

   Mouse Pim2 cDNA was cloned into the MSCV-IRES-GFP plasmid. Empty or
   PIM-2–expressing retroviruses were produced in GP2-293 cells with
   pCL-Eco packaging plasmid (Addgene), using PEI MAX (Polysciences)
   according to the manufacturer’s instructions. BM for transduction was
   isolated by flushing femurs and tibias from Il15^TgXbp1^f/f and
   Il15^TgXbp1^cKO mice 4 days after treatment with 5 mg of 5-fluorouracil
   (Sigma-Aldrich). BM cells (2 × 10^6/ml) were seeded in a 24-well plate
   in 1 ml of culture medium and 1 ml of retroviral supernatants
   supplemented with polybrene (10 μg/ml; EMD Millipore). Cells were
   spin-infected at 230g for 1.5 hours at 32°C and then incubated at 37°C
   for 4 to 6 hours. Transduced BM cells were washed twice and cultured in
   Iscove’s modified Dulbecco’s medium + 15% FBS + 1×
   penicillin/streptomycin + stem cell factor (50 ng/ml) + IL-3 (20 ng/ml)
   + IL-6 (10 ng/ml) (PeproTech). Three days later, cells were washed and
   transferred to a lethally irradiated CD45.1 congenic recipient. Eight
   weeks later, NK cells were isolated from the spleen for flow cytometry
   analysis.

Statistical analysis

   Data are presented by descriptive statistics such as means ± SD. For
   continuous end points such as flow cytometry data or RT-PCR data, a
   Student’s t test or paired t test was used to compare two independent
   or matched conditions/groups, and one-way analysis of variance (ANOVA)
   models were used to compare three or more independent
   conditions/groups. Two-way ANOVA models were used to assess the
   interaction effects between two factors (2 × 2), followed by post hoc
   group comparisons. Linear mixed models (i.e., two-way ANOVA with
   mixed-effects models in PRISM) were used to account for the
   variance-covariance structure due to repeated measures such as tumor
   volume. All tests were two-sided. P values were adjusted for multiple
   comparisons by the Holm-Šídák procedure. P < 0.05 was considered
   statistically significant. *P < 0.05, **P < 0.01, ***P < 0.001, and
   ****P < 0.0001. Prism software v.9.0 (GraphPad) and SAS v.9.4 (SAS
   Institute) were used to perform statistical analyses.

Supplementary Material

   Figs. S1 to S8; Tables S1 to S3
   [296]NIHMS1894608-supplement-Figs__S1_to_S8__Tables_S1_to_S3.pdf^
   (2.1MB, pdf)
   Data files S1 and S2
   [297]NIHMS1894608-supplement-Data_files_S1_and_S2.zip^ (808.9KB, zip)
   MDAR Reproducibility Checklist
   [298]NIHMS1894608-supplement-MDAR_Reproducibility_Checklist.pdf^
   (246.3KB, pdf)

Acknowledgments: