NgR1 is an NK cell inhibitory receptor that destabilizes the immunological synapse

The formation of an immunological synapse (IS) is essential for natural killer (NK) cells to eliminate target cells. Despite an advanced understanding of the characteristics of the IS and its formation processes, the mechanisms that regulate its stability via the cytoskeleton are unclear. Here, we show that Nogo receptor 1 (NgR1) has an important function in modulating NK cell-mediated killing by destabilization of IS formation. NgR1 deficiency or blockade resulted in improved tumor control of NK cells by enhancing NK-to-target cell contact stability and regulating F-actin dynamics during IS formation. Patients with tumors expressing abundant NgR1 ligand had poor prognosis despite high levels of NK cell infiltration. Thus, our study identifies NgR1 as an immune checkpoint in IS formation and indicates a potential approach to improve the cytolytic function of NK cells in cancer immunotherapy. The formation of an immunological synapse is central to the ability of NK cells to lyse target cells. Here the authors show that Nogo receptor 1 (NgR1) might be a good target for cancer immunotherapy as it destabilizes the NK synapse, resulting in defective killing of tumor cells.

The formation of an immunological synapse (IS) is essential for natural killer (NK) cells to eliminate target cells. Despite an advanced understanding of the characteristics of the IS and its formation processes, the mechanisms that regulate its stability via the cytoskeleton are unclear. Here, we show that Nogo receptor 1 (NgR1) has an important function in modulating NK cell-mediated killing by destabilization of IS formation. NgR1 deficiency or blockade resulted in improved tumor control of NK cells by enhancing NK-to-target cell contact stability and regulating F-actin dynamics during IS formation. Patients with tumors expressing abundant NgR1 ligand had poor prognosis despite high levels of NK cell infiltration. Thus, our study identifies NgR1 as an immune checkpoint in IS formation and indicates a potential approach to improve the cytolytic function of NK cells in cancer immunotherapy.
Cancer immunotherapy has been used for decades to eliminate tumors, and various types of immune cells have been used for the purpose, especially those that can lyse target cells via cell-to-cell contact 1,2 . NK cells are used in cancer immunotherapy owing to their natural properties; for instance, there are advantages of using allogeneic NK cells for adoptive transfer, and NK cells confer safety against cytokine release syndrome and can serve as first-line defenders for target cells 3,4 . NK cells form an IS, based on cell-to-cell contact, to recognize and eliminate virus-infected and transformed cancer cells [5][6][7][8] . The IS is a dynamic supramolecular structure, where spatiotemporal organization of cytoskeleton, surface receptors and intracellular signaling proteins occurs for signal integration and directed secretion of effector molecules 7,9 . NK cell IS formation is a multistep sequential event comprising initiation of a transient contact for target cell surveillance, establishment of firm adhesion mediated by adhesion molecules and cytoskeleton remodeling, and polarization and excretion of lytic granules for target cell lysis 10 . Activating and inhibitory signals, integrated through the IS, determine target cell fates. Inhibitory signals typically work in the early stages of IS to interfere with activating signals that form and stabilize the IS through cytoskeleton remodeling, leading to NK-target interactions, thereby acting as immune checkpoints 8,11 . Existing immune checkpoint inhibitors do not ensure complete satisfactory outcomes in patients 12,13 ; therefore, the discovery of novel inhibitory receptors would have notable scientific and clinical implications, as such receptors could be ideal targets for cancer immunotherapy.
The term IS originated from synapses of the nervous system, which have similar properties of cell-to-cell contact and signal transmission 14 . Guidance cues, including attractive and repulsive cues, are the key molecules regulating axonal outgrowth and synapse formation in the central nervous system (CNS) 15,16 . Among the many Article https://doi.org/10.1038/s41590-022-01394-w in neuroinflammation 22 . However, studies are limited to the nervous system, providing insufficient information about molecular mechanisms, and do not address the role of NgR1 in tumor control, which is a crucial function of immune cells. Therefore, defining the role of NgR1 in IS formation and the subsequent killing effects of NK cells is necessary to improve our understanding of the antitumor mechanism of NK cells.
Here, we report a modulatory function of NgR1 in the killing ability of NK cells against NogoA-expressing target cells via interference with repulsive cues that function as inhibitors for axon growth and synaptic function, NgR1 has a well-established role in neurophysiopathology 17,18 . NgR1 consists of domains including leucine-rich repeat, stalk and glycosylphosphatidylinositol (GPI) and has high homology between humans and mice 19 . NgR1 recognizes its ligand, NogoA, and induces neuronal degeneration by RhoA signals that regulate actin cytoskeletal dynamics in damaged neurons 20,21 . NgR1 is also expressed in various immune cells and regulates their adhesion to myelin expressing NogoA, suggesting an immunomodulatory role  Article https://doi.org/10.1038/s41590-022-01394-w contact stability and complete IS formation. Moreover, the relationship between NogoA expression in tumors and NK cell infiltration indicates unfavorable clinical outcomes in patients. Taken together, these findings reveal an underlying mechanism of regulation of IS formation by NgR1, suggesting a potential target that could be used to improve cancer immunotherapy.

NgR1 interferes with the antitumor function of NK cells
To assess the role of NgR1 in NK cells, we first investigated the cell surface expression of NgR1. We found NgR1 to be expressed in mouse primary NK cells, CD8 T cells and the EL4 cell line ( Fig. 1a and Extended Data Fig. 1a). As the GPI-anchored receptor NgR1 lacks an endo-domain for intracellular signaling, it requires the participation of a complex of coreceptors, including immunoglobulin-like domain-containing protein 1 (LINGO1), tumor necrosis factor receptor superfamily member 19 (TROY) and neutrophin receptor (p75NTR) 23,24 . We found the coreceptors of NgR1 to be expressed in mouse primary splenic NK cells and the EL4 cell line (Extended Data Fig. 1b). As NK cells recognize the ligand and lyse cancer cells 3,6 , we confirmed the expression of NogoA, a ligand of NgR1, in various cancer cell lines (Extended Data Fig. 1c). Next, we measured the cell-mediated cytotoxicity of mouse splenocytes to investigate the involvement of NgR1 in the cytolytic function of immune cells. After treatment with NEP1-40, which is an antagonistic peptide of NgR1 (ref. 25), splenocytes of wild-type (WT) mice showed higher cytotoxicity than those of controls ( Fig. 1b and Extended Data Fig. 1d). Moreover, we found that splenocytes from NgR1-knockout (KO) mice exhibited higher cytolytic effects than those from WT mice (Fig. 1c).
To verify the specificity of NEP1-40 and determine whether the function of NgR1 is restricted to NK cells, we investigated the cytotoxicity of NK cells isolated from the spleens of WT and KO mice with or without NEP1-40 treatment. NK cells from KO mice showed higher cytotoxicity than those from WT mice, and only NK cells from WT mice showed increased cytotoxicity with NEP1-40 treatment (Fig. 1d).
As NK cytotoxicity improved upon NgR1 deficiency, we hypothesized that WT and KO mice would exhibit different resistance to tumors. To verify this hypothesis, a syngeneic mouse model of lung metastasis was established for WT and KO mice (Fig. 1e). After intravenous administration of B16F10 cells, fewer metastatic nodules remained in the lungs of KO mice than in those of WT mice, but no difference was seen in the NK cell-depleted group, suggesting that the animal model was NK cell-dependent ( Fig. 1f,g). As NK cell infiltration in tumors is an indicator of tumor resistance 26,27 , we investigated the population of intrapulmonary NK cells. The NK cell population (CD3 − NK1.1 + ) that infiltrated the lungs bearing tumor nodules was greater in KO mice than in WT mice (Fig. 1h). These data suggest that NgR1 in NK cells contributes negatively to tumor control. To investigate whether the improved antitumor effect caused by NgR1 deficiency was due to intrinsic alterations in immune composition, we analyzed the populations of immune cells from WT and KO mice. Total NK cells (CD3 − NK1.1 + ) were classified into four maturation stages according to the expression of CD11b and CD27, namely immature NK cells (CD27 − CD11b − ), early mature NK cells (mNK; CD27 + CD11b − ), mNK cells (CD27 + CD11b + ) and late mNK cells (CD27 − CD11b + ) 28 . We found no difference between WT and KO mice in resting and IL-2-stimulated total or classified NK cell populations, or in IFN-γ expression (Extended Data Fig. 2a-c). Populations of CD8 T cells (CD3 + CD8 + ), CD4 T cells (CD3 + CD4 + ), B cells (B220 + ), myeloid cells (CD11b + Gr1 + ), neutrophils (CD11b + Gr1 high ), monocytes (CD11b + Gr1 low ) and macrophages (CD11b + F4/80 high ) also showed no differences between WT and KO mice (Extended Data Fig. 2d-f). These data collectively indicate that NgR1 deficiency does not affect the composition of immune cells, implying that NgR1 is mainly involved in the effector function of NK cells.

NgR1 regulates NK cell actin cytoskeleton dynamics
As NgR1 is involved in the tumor control of mouse NK cells, it might have a prominent role in human NK cells as well. To verify this, we investigated the expression of NgR1 and its signals in human NK cells. NgR1 was found to be expressed in human NK cells, including umbilical cord blood (UCB)-derived NK cells (UCB-NK), peripheral blood (PB)-derived NK cells (PB-NK), cytokine-induced mNK cells, NK92 cells, UCB-CD8 T cells and Jurkat cells ( Fig. 2a and Extended Data Fig. 3a). We also found that the coreceptors of NgR1 were expressed in human UCB-NK, PB-NK, mNK, NK92 and Jurkat cell lines (Extended Data Fig. 3b). NgR1 is well known to recognize NogoA and activates cytoskeleton-regulatory signals 20,29 . With RhoA activation by NgR1 stimulation, Rho-associated coiled-coil-containing protein kinase (ROCK) phosphorylates LIM domain kinase 1 (LIMK1) and LIMK2 to phosphorylate and inactivate cofilin, which severs filamentous (F)-actin into globular (G)-actin 18,20 . To investigate the signaling of NgR1, we stimulated NgR1 in NK cells. Following treatment with Nogo-P4, an agonistic peptide of NgR1 (ref. 30), both RhoA and LIMK were activated and cofilin was inactivated in NK92 cells and UCB-NK cells ( Fig. 2b and Extended Data Fig. 3c). RhoA activation promotes stress fiber formation through actomyosin-based contraction and cofilin inactivation, leading to actin cytoskeleton reorganization 31,32 . Accumulation of F-actin causes the formation of cell membrane protrusion, which affects cell migration and adhesion 33 . As NgR1 stimulation activates RhoA and inactivates cofilin, we hypothesized that NgR1 would affect the actin regulation of NK cells. To verify this, F-actin was directly visualized in NK92 cells expressing Lifeact-GFP, and the effects of Nogo-P4 treatment on F-actin dynamics were assessed by video microscopy. Nogo-P4-treated NK92 cells exhibited significantly increased F-actin intensity (Fig. 2c,d and Supplementary Video 1) and membrane protrusion frequency (Fig. 2c,e and Supplementary Video 1) compared with untreated (Ctrl) or scrambled peptide (Scram)-treated NK cells. Phalloidin staining of F-actin (Extended Data Fig. 4a,b,e,f) and membrane protrusion frequency measurements from bright field images (Extended Data Fig. 4c,d and Supplementary Video 2) using WT or Lifeact-GFP-expressing NK92 cells demonstrated identical results, indicating a minimal Lifeact-mediated artifact of F-actin dynamics 34 in NK92 cells. These data suggest that NgR1 in NK cells regulates actin cytoskeleton dynamics through RhoA signals.

NgR1 regulates NogoA-mediated NK cell killing
To assess the NogoA-specificity of NgR1 in NK cell killing, we investigated NK cell-mediated cytotoxicity by regulating the function and expression of NgR1 in NK cells or NogoA in target cells. First, we confirmed the expression of NogoA in cancer cell lines. NogoA was expressed at different levels on the surfaces of the target cells (Extended Data Fig. 5a). Killing of NK92 cells relative to K562 cells, which hardly express NogoA, showed no difference with or without blocking of NgR1 with NEP1-40 (Fig. 3a). However, NK-mediated killing was significantly reduced for NogoA-overexpressing K562 cells and was rescued by NEP1-40 treatment, similar to the results in HEK293T cells (Fig. 3b,c and Extended Data Fig. 5b,c). After treatment with NEP1-40, there was increased NK cytotoxicity against U87MG cells, a glioma cell line from brain tumors known to express high levels of NogoA 35 (Fig. 3d). We further found that NK cytotoxicity increased upon inhibition of NogoA expression in U87MG cells or NgR1 expression in NK92 cells (Fig. 3e,f and Extended Data Fig. 5d-f). As NK cell activity depends on the balance between activating and inhibitory signals to kill the target 3,5,11 , we questioned whether NgR1 could be involved in killing NK cell-resistant targets. To verify this, we investigated the expression of activating and inhibitory ligands in several cell lines. AU565 cells expressed NogoA while expressing little or no activating ligand such as ULBP1, ULBP2, ULBP3 or MIC-A/B (Extended Data Fig. 5a,g). The cytotoxicity of NK cells against AU565 cells significantly increased owing to the blocking of NgR1 by NEP1-40 treatment (Fig. 3g). Similar to NK92 cells, human   Fig. 5h-l). Therefore, the data collectively suggest that NgR1 acts as an inhibitor of the cytolytic function of NK cells, specifically in the presence of NogoA.

NK-to-target cell contact is destabilized by NgR1
To further gain insight into how NgR1 signaling inhibits NK cell cytotoxicity, we directly observed interactions between NK and target cells by live-cell imaging. Most of the control NK92 cells transiently Step I Transient <10 min Step II Stable >10 min

Killing and detachment
Step III   Fig. 4b) and then form stable synapses (step II in Fig. 4b) that direct the polarized secretion of lytic granules to perform target cell lysis (step III in Fig. 4b). We analyzed the percentage of NK92 cells transiently interacting with target cells with a contact duration <10 min and measured the contact duration between NK92 cells and target cells forming stable synapses. NEP1-40-treatment was found to significantly reduce transient interactions and increase contact duration, resulting in enhanced cytotoxicity (Fig. 4c-e). Similar results were obtained with human PB-NK or mouse NK cells when NgR1 was either blocked by NEP1-40 (Extended Data Fig. 6a-h and Supplementary Videos 4 and 5) or genetically knocked out (Extended Data Fig. 6i-l and Supplementary Video 6). Together, these results indicate that NgR1 signaling reduces NK cell cytotoxicity by interfering with stable synapse formation.

F-actin dynamics are regulated by NgR1 during IS formation
NgR1 signaling is responsible for altering the actin cytoskeletal balance 20,29 . To investigate the molecular mechanism of NgR1 signals regulating NK-to-target contact, we regulated ROCK or LIMK activity or cofilin expression (Fig. 5a). As cofilin directly regulates F-actin dynamics 20,31,32 , we hypothesized that inhibition of cofilin expression would affect F-actin turnover, resulting in impairment of NK cell contact and subsequent killing. To verify this, we suppressed the expression of cofilin in NK92 cells with small interfering RNA (siRNA) (Extended Data Fig. 7a) and found that NK92-U87MG conjugation was reduced and NK killing was impaired (Fig. 5b,c). Next, we investigated NK-to-target contact and NK killing by treating NK cells with LIMKi3 (a LIMK1 and LIMK2 inhibitor). Nogo-P4-treated NK92 cells increased phosphorylation of LIMK and cofilin, whereas LIMKi3 treatment inhibited both LIMK phosphorylation and subsequent cofilin phosphorylation (Extended Data Fig. 7b). Moreover, we found that inhibition of LIMK and cofilin phosphorylation occurred through LIMK inhibition under NogoA-expressing U87MG cell-mediated NgR1 stimulation of NK92 cells (Fig. 5d) Fig. 5j,k). The cytotoxicity of NK92 cells against U87MG increased under NEP1-40 treatment. Although the cytotoxicity of NK92 cells was decreased by treatment with wiskostatin or CK-666, it was restored by cotreatment with NEP1-40 (Fig. 5l,m). These data suggest that both activation and NgR1 signals of NK cells, which regulate IS formation with actin dynamics, influence NK cytotoxicity. Next, we visualized F-actin dynamics in the context of IS using NK92 cells expressing Lifeact-GFP. F-actin polymerization at the IS is among the early events that are critical for stable IS formation 7,8 . In the control group, NK92 cells contacting U87MG cells mainly polarized F-actin outward from NK-to-target contacts and frequently detached from the target cells (top panel of Fig. 6a and Supplementary Video 8). By contrast, NEP1-40-treated NK92 cells polarized F-actin toward the NK-to-target contacts and maintained stable IS formation (bottom panel of Fig. 6a and Supplementary Video 8). F-actin distributions with respect to NK-to-target contacts were classified into four cases and were plotted for NK92-U87MG conjugates after the initiation of coculture (Fig. 6b). In the control group, the majority of NK92 cells polarized F-actin outward from the NK-to-target contacts, whereas in the NEP1-40-treated group, most NK92 cells polarized F-actin toward the NK-target contacts (Fig. 6b). NgR1 signaling also influenced lytic granule convergence and polarization toward the NK-target contacts, both of which are critical events for cytotoxicity (Fig. 6c,d and Supplementary Video 9); the majority of NK92 cells in the control group exhibited diffuse patterns of granules, indicating their failure to converge the granules. By contrast, most NK92 cells treated with NEP1-40 converged the lytic granules to the distal pole (approximately 50% in 30 min) and then polarized them toward the IS (approximately 80% in 120 min). Together, these results indicate that suppressed F-actin dynamics, mediated by NgR1 signaling at NK-to-target contacts, promote F-actin polymerization outward from the cell-to-cell contacts, resulting in the detachment of NK92 cells before lytic granule polarization toward IS.

NgR1 functions as an immune checkpoint in NK cells
Considering that NK cells perform immune surveillance and NgR1 inhibits the cytolytic function of NK cells, the therapeutic effect of NgR1 blockade was investigated using a xenograft mouse model. Following subcutaneous injection of U87MG cells, NSIG mice were intravenously injected with NK92 cells on days 10 and 14, and scrambled peptide or NEP1-40 was injected intratumorally on days 11 and 15 (Fig. 7a). Tumor size was decreased in the group where NK92 cells and NEP1-40 were administered, compared with the group injected with phosphate-buffered saline (PBS; control) and NK92 cells with scrambled peptide (NK92) (Fig. 7b,c). Survival rates of tumor-bearing mice were improved in the group receiving NK92 cells with NEP1-40 (Fig. 7d). We next assessed whether NgR1 could be linked to clinical outcomes along with a relationship with RTN4, the gene encoding Nogo, in patients with cancer. First, rich-or poor-NK-infiltration groups were divided using CIBERSORT, and then clinical prognosis was deduced by grouping of patients with high or low RTN4 expression using The Cancer Genome Atlas (TCGA) pan-cancer data (Fig.  7e). In all cancer patients, regardless of the quantity of infiltrated NK cells, we confirmed that RTN4 was a risk factor for mortality and that high RTN4 led to poor clinical outcomes (Fig. 7f,g, Extended Data Fig.  8a and Supplementary Table 1). In particular, the RTN4 expression level of NK-rich patients represented a higher risk for overall survival   Table 2). Collectively, these data suggest that NgR1 expression in cytolytic immune cells serves as an immune checkpoint to inhibit IS formation and could represent a novel therapeutic target for controlling tumors (Fig. 7h).

Discussion
In this study, we identified NgR1 as a novel NK cell inhibitory receptor prohibiting stable IS formation by LIMK-cofilin-mediated alteration of actin dynamics. NgR1 deficiency or blockade promoted the stable formation of IS, thereby increasing NK cell killing and tumor control. Even when NK cells were infiltrated, patients with cancer expressing abundant Nogo still had poor prognosis. This revealed the underlying mechanism of improvement of NK cell function by NgR1 and highlighted the clinical finding that NgR1 acts as an immune checkpoint.
In the CNS, axonal growth and synapse formation are critical for signal transmission of neurons. In contrast to attractive cues promoting axon outgrowth and synapse formation by Rac1, and CDC42 signals promoting cytoskeleton dynamics, inhibitory cues interfere with axonal growth and synaptic function via RhoA signals that suppress cytoskeleton remodeling 15,16 . Neuronal growth inhibitors interfere with synapse formation by inducing axon growth cone collapse and retraction 15,16,18 . Among the inhibitory cues, NogoA-NgR1 interaction, which belongs to the myelin-associated inhibitor category, inhibits axonal growth and synaptic function through RhoA-mediated cytoskeleton regulation and is required to prevent abnormal neuronal sprouting in healthy brain and for neuronal degeneration upon CNS damage 18,21,29 . NgR1 acts as an inhibitory receptor that directly regulates neurons rather than blocking signals of attractive cues 38,39 . Recently, NgR1 has been reported to inhibit the adhesion of immune cells 40 ; however, this is restricted to the nervous system and detailed mechanisms are missing. Elucidating the role of NgR1, which is otherwise well defined in the nervous system, in tumor control by immune cells is necessary to obtain a better understanding of the mechanism of action of immune cells. This, in turn, could contribute to improving the performance of cancer immunotherapy.
NK cells form IS with target cells to identify infected and/or transformed cells and exert cytotoxicity. Synapse-mediated cytotoxicity is accomplished via stepwise processes, including tethering, F-actin accumulation, firm adhesion, granule convergence to microtubule organizing center (MTOC), MTOC polarization to IS and granule exocytosis 8 . Signaling mediated by inhibitory receptors, such as NKG2A and KIR, acts primarily in the early stages of IS to prevent activating signals from occuring 8,11 . Functionally, inhibitory receptor signals destabilize the IS and promote NK cell detachment and migration 41 . The characteristics of NgR1-mediated NK cell inhibition observed in this study were similar to those of inhibitory receptors; in the presence of NgR1-NogoA interactions, NK cells exhibited transient interactions with target cells and failed to polarize F-actin and lytic granules toward the IS, resulting in impaired cytotoxicity. However, NgR1 is distinct from other inhibitory receptors owing to the downstream signals that it triggers. Typically, inhibitory receptors recognize either major histocompatibility complex class I (MHC I) or non-MHC I ligands and signal through immunoreceptor tyrosine-based inhibitory motifs (ITIMs) located at their cytoplasmic tails, thereby blocking the activating signals 11,13 . NgR1, on the other hand, interacts with NogoA and signals through coreceptors, as it is a GPI-anchored receptor with no cytoplasmic domain 29 . In addition, NgR1 inhibits actin reorganization through RhoA GTPase, in contrast to activation signals such as NKG2D, which activate GTPases including CDC24, Rac1, and Vav1 that are favorable   Although the role of actin cytoskeleton remodeling mediated by the RhoA-ROCK-LIMK-cofilin axis, downstream of NgR1 signaling, has been well established in neural synapse 18,21,29 , it has not been investigated in IS. Although RhoA-mediated inhibition of T cell and NK cell cytotoxicity 43,44 and cofilin-mediated T cell synapse formation for T cell activation 44,45 have previously been reported, upstream signals activating RhoA and deactivating cofilin have not yet been identified. For ITIM-mediated actin remodeling in NK cells that leads to synaptic destabilization, dephosphorylation of Vav1 (ref. 46), a guanine nucleotide exchange factor important for immune cell activation, and phosphorylation of the adapter protein Crk 47 have primarily been considered 10 . It would be interesting to see whether the pathway identified in this study is also triggered in ITIM-mediated NK cell suppression. In addition, as RhoA and Rac1 interfere with each other, the detailed relationship of these signals 48 , which are spatiotemporally involved in NK killing, should also be elucidated. Overall, our preliminary results reveal that CD8 + T cells express NgR1, and that NgR1-mediated signaling inhibits T cell-mediated cytotoxicity, implying that NgR1 could be a crucial inhibitory checkpoint for our immune system.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41590-022-01394-w.

Mice
All mice were randomly bred and/or maintained in a specific-pathogen-free animal facility at 22-26 °C and 40-60% humidity on a 12-h dark-light cycle, with food (Harlan diet, 2018S) and water as needed in the Laboratory Animal Resource Center at KRIBB. C57BL/6N mice (WT) purchased from DooYeol Biotech, NgR1 KO mice (C57BL/6-Rtn4r tm1cyagen ) purchased from Cyagen Biosciences Inc. and NSIG mice purchased from GHBIO were used for experiments at 6-8 weeks of age. Male mice were randomly used for in vitro studies and experiments on antitumor effects in vivo. All animal experiments were performed in agreement with the Animal Experimental Ethics Committee of KRIBB.

Primary NK cell preparation
WT and KO mice were used for vitro experiments with splenocyte harvesting and NK cell isolation. Splenocytes were recovered by grinding the spleens of mice with a 70-μm cell strainer (SPL Life Sciences). NK cells were isolated from splenocytes using an mouse NK Cell Isolation Kit (Miltenyi Biotec). Isolated NK cells were cultured in RPMI 1640 (Gibco) containing 10% fetal bovine serum (FBS), 1% penicillin-streptomycin and 10 ng ml −1 recombinant human IL-2 (hIL-2) (PeproTech). Human UCB was used for NK cell isolation. UCB-NK were isolated using CD3 + cell depletion with Rosette Sep (StemCell Technologies), density separation with Lymphoprep (StemCell Technologies) and enrichment with a human NK Cell Isolation Kit (Miltenyi Biotec). Human PB was used for NK cell isolation. Blood depleted of platelet-rich plasma was isolated by density gradient centrifugation using Ficoll-Paque (GE Healthcare) to collect PB mononuclear cells, which were collected from the interface layer after centrifugation. PB-NK cells were isolated with a human NK Cell Isolation Kit (Miltenyi Biotec). Cytokine-induced mNK cells were differentiated from CD34 + hematopoietic stem cells (HSCs). Human CD34 + HSCs were isolated from UCB using Rosette Sep, Lymphoprep and a human CD34 MicroBead Kit (Miltenyi Biotec). NK cell precursors were differentiated from CD34 + HSCs in Myelocult H5100 (StemCell Technologies) supplemented with hydrocortisol (10 −6 M), SCF (30 ng ml −1 ), Flt3 (50 ng ml −1 ) ligand and IL-7 (5 ng ml −1 ) for 14 days. mNK cells were differentiated from NK cell precursors by stimulation with hydrocortisol (10 −6 M), IL-21 (30 ng ml −1 ) and IL-15 (30 ng ml −1 ) for 14 days. Mouse and human primary NK cells were maintained under 37 °C and humidified 5% CO 2 conditions.

Flow cytometry analysis
Cells were washed and stained with antibodies in PBS containing 1%  Table 3). Antibody-labeled cells were analyzed by fluorescence-activated cell sorting (Canto II, BD Biosciences) and data were collected using FlowJo software (Tree Star).

Cytotoxicity assay
NK cell-mediated cytotoxicity was measured using calcein-AM release assay. Target cells were incubated with calcein-AM (Invitrogen) for 1 h under 37 °C and humidified 5% CO 2 conditions. Calcein-labeled target cells were plated in 96-well round-bottomed plates and then cocultured with serially diluted effector cells at the desired effector:target ratio for the desired incubation time. The calcein released from lysed target cells into the supernatant by effector cells was measured using a multimode microplate reader (Molecular Devices). Maximal and spontaneous release of calcein was simulated by adding 2% Triton X-100 and complete medium to calcein-labeled target cells. The percentage of specific lysis was calculated using the following formula: (experimental release - spontaneous release) / (maximum release - spontaneous release) × 100%. NEP1-40 or anti-NgR1 was used as a treatment during coculture of effector and target cells to block NgR1.

Lung metastatic syngeneic mouse model
A lung metastasis model using the B16F10 melanoma cell line was constructed in WT and KO mice. For depletion of NK cells in mice, 100 μg of anti-NK1.1 or isotype control in 200 μl Dulbecco's PBS (DPBS) was injected intravenously through tails of mice 4 days and 1 day before and 2 days after intravenous injection of B16F10. The CD3 − NK1.1 + populations were measured by flow cytometry analysis 5 days before and 4 and 13 days after B16F10 injection. Each mouse was injected intravenously with 2 × 10 5 B16F10 cells on day 0. The body weights of the mice were measured on days 0, 9 and 14 after injection of B16F10. Mice were euthanized 14 days after injection of cancer cells, and their lungs were removed for analysis of metastatic melanoma nodules. The removed lungs were ground with a 70-μm cell strainer to analyze the intrapulmonary CD3 − NK1.1 + population by flow cytometry.

Cell-line-derived tumor xenograft mouse model
A xenograft model using the human U87MG glioblastoma cell line was established in NSIG mice lacking T cells, B cells and NK cells. For the development of solid tumors, 4 × 10 6 U87MG cells were implanted subcutaneously into mice on day 0. NK92 cells (2.5 × 10 5 ) were injected intravenously through the tails of mice 10 and 14 days after U87MG injection, and 300 μg of Scram or NEP1-40 in 100 μl DPBS were injected intratumorally on days 11 and 15. In the Ctrl group inoculated with only U87MG cells, without NK92 cells, intratumoral injection of 100 μl DPBS was performed. From 10 days after plantation of the tumor cells, the width and length of the developed tumor were measured every 4 days with a caliper to calculate the tumor size in mm 3 using the following formula: (width + length) 2 / 2, and body weights were measured at the same time. The tumor size did not exceed 20 mm 2 according to the guidelines of the KRIBB Animal Experimental Ethics Committee. After tumor development, survival of mice was monitored every day during the remainder of the experimental period.

Immunoblotting
Cells were washed with ice-cold PBS and lysed using cOmplete Lysis-M (Roche), and protein concentrations were measured with a Pierce BCA Protein Assay Kit (Thermo Scientific). Cell lysates containing 10-20 μg of proteins were reduced by boiling for 10-15 min using sodium dodecyl sulfate (SDS) sample buffer. Samples were Article https://doi.org/10.1038/s41590-022-01394-w separated by 8-12% SDS polyacrylamide gel electrophoresis and transferred to a polyvinylidene fluoride membrane (Millipore). Membranes were blocked with 1-5% nonfat milk or BSA in PBS containing 0.05% Tween-20 (Duchefa Biochemie) for 40-60 min at room temperature and then incubated with primary antibodies overnight at 4 °C. Membranes were then washed and incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit or anti-mouse IgG secondary antibody for 40-60 min at room temperature. After washing, the membranes were developed using SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Scientific), and immunoblot images were obtained using WSE-6100 LuminoGraph (ATTO). Data were collected using CASanalyzer (ATTO) software. For endogenous RhoA-GTP detection, cells were lysed using cOmplete Lysis-M, and cell lysates were pulled down using a Rho Activation Assay Biochem Kit (Cytoskeleton) according to the manufacturer's protocol. Cell lysates were incubated with Rhotekin, a Rho effector protein that has a Rho-binding domain and is tagged with GST, in a rocker at 4 °C for 1 h. Centrifuged and washed samples were reduced by boiling for 10-15 min using SDS sample buffer, followed by separation and transfer. The following antibodies were used: anti-pLIMK1/2, anti-LIMK2, anti-pCofilin, anti-cofilin, anti-GAPDH, anti-NogoA, anti-NgR1, anti-rabbit IgG-HRP and anti-mouse IgG-HRP (Supplementary Table 3).

Live-cell imaging
A modified Olympus IX 83 epi-fluorescence microscope with a 40X (UPlanFLN, NA = 1.30) objective lens and an ANDOR Zyla 4.2 sCOMS camera was used for imaging experiments. The microscope was automatically controlled by Micro-manager. For live-cell imaging, the microscope stage was equipped with a Chamlide TC incubator system (Live Cell Instrument) maintaining cell culture conditions (37 °C, CO 2 5%). Acquired images were processed using ImageJ. To observe F-actin dynamics in NK92 cells, NK92 cells expressing Lifeact-GFP were seeded on a glass coverslip, and time-lapse imaging acquiring differential interference contrast and GFP fluorescence was initiated about 15 min after cell seeding. To observe interactions between NK cells and cancer cells, cancer cells were first plated on gelatin-coated coverslips and incubated for 12 h in a tissue culture incubator (37 °C with 5% CO2) so that cancer cells could adhere and spread on the substrates. Then, the coverslip containing cancer cells was mounted in a magnetic chamber (Chamlide CF, Live Cell Instrument), and the chamber was loaded on the microscope stage equipped with the incubator system. NK cells were added to the chamber, and timelapse imaging was initiated 15 min later to allow NK cells to sediment and initiate interactions with the cancer cells. NK cells transfected with Lifeact-GFP and labeled with lysosensor (Thermo Fisher) were used to visualize F-actin and lytic granules, respectively. Killing probability was measured by time-lapse imaging. Total interaction events and killed target cells were directly counted in each field of view.

CIBERSORT
CIBERSORT was used for mathematical analysis of pan-cancer gene expression data (https://cibersortx.stanford.edu) 49 . We used log 2 -scaled FPKM values for deconvolution analysis of available gene expression levels 50 . The gene expression file created based on 9,574 cases was uploaded to CIBERSORT as a mixture file, and CIBERSORT Nature Immunology Article https://doi.org/10.1038/s41590-022-01394-w was run with options for LM22 reference file, 500 permutations and quantile normalization disabled. The quantity of NK cells was calculated as the sum of the quantities of activated NK cells and resting NK cells in the LM22 reference file. The samples for which the quantities of NK cells or CD8 T cells were in the top 10% to 50% with statistical significance (P < 0.05) were included in the NK or CD8 T-rich group, and the others were included in the poor group.

Statistical and survival analysis
One-way or two-way analysis of variance (ANOVA) with Tukey's or Sidak's multiple comparisons test or unpaired two-tailed Student's t test was used to analyze the significance of findings using GraphPad Prism. The log-rank test and Wald test with 95% confidence intervals were performed to estimate the significance of differences in overall survival. Kaplan-Meier plots and forest plots were used to visualize the results. All survival analyses of clinical information were carried out using R software packages (v.3.6.3). Statistical values are indicated in each figure. Data are represented as mean ± s.d. or s.e.m. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications 28,36,42,46 . Data distribution was assumed to be normal, but this was not formally tested. Data collection and analysis were not performed blind to the conditions of the experiments. No data were excluded from the analysis.

Ethics
This study complies with all applicable codes of ethics. Human studies were approved by the Korea Research Institute of Bioscience and Biotechnology (KRIBB) Institutional Review Board (P01-201610-31-002) and SNU Institutional Review Board (E2102-003-006). For the use of human UCB or PB, informed consent was obtained from the donor through the Cord Blood Bank of Korea or Korea Red Cross Blood Services. Animal studies were approved by the Animal Experimental Ethics Committee (AEC-21016, −21017) and conformed to the Regulations on the management and use of laboratory animals of KRIBB.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The data supporting the findings of this study are available within the paper and from the corresponding authors upon request. Patient samples, including gene expression and clinical information, were accessed from the GDC data portal (https://portal.gdc.cancer.gov). Primary tumor data for survival analysis were from TCGA. Mathematical data for pan-cancer gene expression were analyzed using CIBER-SORT (https://cibersortx.stanford.edu). Source data are provided with this paper.