TRPML1-mediated Modulation of Dense-core Granules Tunes Functional Potential in NK Cells

to activation; however the intracellular basis for functional tuning by inhibitory receptors remains unclear. We found that NK cells expressing self-MHC specific inhibitory killer cell immunoglobulin -like receptors (KIR) show a greater accumulation of dense-core secretory granules, converged closer to the centrosome in resting NK cells, which are released upon recognition of target cells. This discrete morphological phenotype persists in self-KIR+ NK cells independently of transcriptional programs that regulate metabolism and granule biogenesis and indicates an intrinsic role for lysosomal homeostasis in NK cell education. Upon activation, interference of signaling from acidic Ca2+ stores reduced both target-specific Ca2+-flux, degranulation and cytokine production. Furthermore, inhibition of PI(3,5)P2 synthesis or genetic silencing of the PI(3,5)P2-regulated lysosomal Ca2+channel TRPML1 in primary NK cells led to an increase in granular load and enhanced functional potential. These results suggest a model where continuous unopposed signaling through activating receptors render NK cells hypofunctional through TRPML1-mediated modulation of acidic Ca2+ stores.

interactions both with stromal and hematopoietic cells 4 . Transfer of mature NK cells from one MHC environment to another results in reshaping of the functional potential based on the inhibitory input of the new MHC setting. 5 Alternatively, genetic knockdown of SLAM-family receptors by CRISPR/Cas9 leads to hyperfunctionality, 6 whereas deletion of the inhibitory signaling through ITIM and SHP-1 renders NK cells hypofunctional. 7,8 However, it remains unclear how and when the net signaling input from activating and inhibitory receptors during NK cell education tune the functional potential of the cell. One difficulty in establishing the cellular and molecular mechanisms that account for the calibration of NK cell function is the lack of a steady-state phenotype that defines the educated NK-cell state. Functional readouts used to distinguish self-specific NK cells from hyporesponsive NK cells do not provide information about the prior events that culminate in the development of effector potential. Apart from differences in the relative levels and distribution of NK cell receptors at the cell membrane, 9,10 transcriptional and phenotypic readouts at steady state provide scant differences between self and non-self specific NK cells. 11,12 While the confinement of activating receptors to areas that are permissive for signaling holds important clues Natural killer (NK) cells achieve specificity through unique combinations of variable germ-line encoded receptors. These receptors are critical for the development of cell-intrinsic functional potential, which enables spontaneous activation upon recognition of target cells displaying reduced class I MHC expression. 1 .Inhibitory interactions with self-MHC translate into a predictable quantitative relationship between self-recognition and effector potential, a process termed NK cell education. 2 Despite being clearly evident in different species, 3 NK cell education operates through an as yet largely unknown mechanism. Paradoxically, mature NK cells expressing self-MHC specific inhibitory receptors, receiving constitutive inhibitory input, exhibit increased levels of functionality. Mouse models have demonstrated that this functional phenotype is dynamic and dependent on the net signaling input to NK cells during cell-to-cell

TRPML1-mediated
Modulation of Dense-core Granules Tunes Functional Potential in NK Cells for the increased reactivity of self-specific NK cells, precisely how these structural changes translate into a profound difference in functionality between self and non-self specific NK cells remains to be established. The critical question is whether qualitative differences in proximal receptor signaling alone explains NK cell education or whether the functional potential of the cell can be a linked to a cellular mechanism that develops prior to target cell recognition as the cell interacts with its microenvironment. Whether inhibitory signalling leads to gain of function through an as yet unknown mechanism, a concept referred to as arming/stimulatory licensing, or that inhibitory receptors simply tune out tonic activation that would otherwise lead to erosion of function over time (e.g. disarming/ inhibitory licensing) 3,13,14 remains to be determined. Priming of NK cell effector function, in both mouse and human, depends on homeostatic cytokines, including IL-15 and type I interferon 15,16,17 , and is affected by metabolic cues 18,19 . Under germ-free conditions, animal models of acute infection result in a tidal upregulation of the transcription of effector loci in NK cells, which resolve with normal immune kinetics 20 . In humans, the full cumulative spectrum of NK cell differentiation is determined epigenetically and revealed through graded increases in the cellular content of both granzyme B and perforin, coupled to increasing capacity for both cytolysis and cytokine production 21 . Here we sought to examine how priming, granular development and receptor-mediated in- hibition are coordinated to establish and maintain a steady state functional potential in mature NK cells. Our findings suggest that inhibitory receptors oppose TRPML1-mediated disarming through modulation of the lysosomal pathway, including the secretory granules. Hence, NK cells expressing self-specific KIR effectively sequester functional potential through manipulation of the endolysosomal compartment, allowing them to subsequently mount strong, spontaneous effector responses from pre-existing granular stores. Moreover, the granular compartment itself appears to contribute to Ca 2+ -flux and distal signaling, effectively connecting homeostatic receptor input to lysosomal homeostasis in order to determine the enhanced functional potential observed in self-KIR+ NK cells.

Self-recognition is associated with increasing granular load in primary resting NK cells
In order to address the possible mechanisms involved in tuning of effector potential, the expression of cytotoxic effector molecules was monitored in resting NK cells by flow cytometry in discrete subsets corresponding to different stages of differentiation, from CD56 bright NK cells through subsets of mature CD-56 dim NK cells defined by the expression of NKG2A, KIR and CD57. 22 The expression of cytolytic effector molecules granzyme B and perforin increased gradually, whereas granulysin decreased, as NK cells became more differentiated. (Fig. 1a-c and Supplementary Fig. 1a). Next, granzyme B content in mature NK cells was stratified based on the expression of self versus non-self specific KIR. The stochastic expression of KIR in NK cells is independent of the MHC setting and provides a unique and natural equivalent of in vivo gene-silencing. 23,24 This allowed us to address the impact of reciprocal presence or absence of a self-KIR on granular content between individuals. Clustering of NK cell phenotypes using t-distributed stochastic neighbor embedding (tSNE) revealed high expression of granzyme B in NK cell subsets expressing self-specific KIR (Fig. 1d). Extended analysis of 64 healthy donors showed significantly higher expression of granzyme B in NK cells expressing one or more self-specific KIR in donors homozygous for HLA-C1/ C1 or HLA-C2/C2, ligands for KIR2DL3 (2DL3) and 2DL1, respectively, in NK cells that were NKG2A negative and CD57 negative, in order to control for the stage of differentiation (Fig. 1e, f). Similar trends were observed for perforin and granulysin that were both expressed at higher levels in self-specific NK cells ( Supplementary Fig. 1b). Corroborating the link between inhibitory input through self-KIR and granzyme B expression, donors that were heterozygous for HLA-C1/C2 had similarly high levels of granzyme B in both 2DL1 and 2DL3 single-positive NK cells (Fig.  1f). Granzyme B expression was also higher in 3DL1+ NK cells from donors positive for its cognate ligand HLA-Bw4 (Fig. 1g) and varied according to the expression level of 3DL1 (Fig. 1h). NK cells with higher levels of 3DL1 surface expression, also known to have a higher functional capacity, 25 exhibited greater expression of granzyme B. Notably, this phenomenon was only observed in the presence of its cognate ligand in HLA-Bw4+ donors (Fig. 1i). It is well established that NKG2A/HLA-E interactions contribute to education of NK cells. 26,27,28 In line with the results of single KIR+ NK cell subsets, NKG2A+KIR-CD57-NK cells expressed higher levels granzyme B (Fig. 1j). To study the effect of self KIR expression in a dynamic model, retroviral transduction was used to introduce full length 2DL1 or 2DL3 into NK cell lines YTS (HLA-C1/C1) and NKL (HLA-C2/C2). Corroborating the findings with ex vivo staining of primary NK cells, the transduced NK cell lines showed a similar accumulation of granzyme B following transfection of a self KIR (Fig. 1k) and enhanced functionality ( Supplementary Fig. 1c-d). These data show that the expression of inhibitory self-KIR or NKG2A are connected to the granular load of an NK cell, establishing a link between inhibitory input and the regulation of the core cytolytic machinery of NK cells.

Inhibitory input influences the level of granular content independently of transcriptional cell differentiation programs
To address whether the increased levels of granzyme B in educated NK cells were due to gene expression, we first examined the transcriptional regulation of effector programs within the context of NK cell differentiation. Transcriptome analysis was performed using single-cell tagged reverse transcription (STRT), a highly multiplexed method for single-cell RNA sequencing (RNA-Seq). 29 Naïve CD56 bright NK cells and five CD56 dim NK cell subsets, sorted on the expression of NKG2A and KIR were analysed. Since the exact order of transitions between intermediate stages of CD56 dim NK cell differentiation is not known, we focused our analysis on three discrete subsets representing previously defined stages of NK cell differentiation: CD56 bright , CD56 dim NKG2A-KIR-and CD56 dim NKG2A-KIR+ NK cells. 22 Analysis of transcription factors linked to the regulation of GzmB revealed that the immature CD56 bright NK cells predominantly expressed higher levels of transcription factors associated with responsiveness to acute stimulation such as AP1, STAT-1, STAT-5 and NF-kB. The more mature CD56 dim NK cell subsets, on the other hand, expressed higher levels of transcription factors associated with cellular differentiation, including TBX21 and PRDM1, and maintenance of effector expression, such as BATF, IRF4, and NFATc1 30, 31, 32 (Fig.  2a). Within the context of differentiation, levels of IRF4 correlated closely with granzyme B expression GzmB KIR2DL1 KIR2DL3 at both the transcript and protein level (Supplementary Fig. 2a-b) Although there is limited data on their role in human NK cell differentiation, TBX21, PRDM1 and IRF4 have been previously associated with transcription of GzmB, Prf1 and IFNG. 30,33 Next, we addressed whether the transcriptional changes associated with NK cell differentiation and KIR acquisition accounted for the observed differences in granzyme B content between self and non-self specific NK cell subsets. NKG2A-CD57-NK cells were sorted by FACS into 2DL3 or 2DL1 single positive populations from C1/C1 and C2/C2 donors and transcriptionally profiled using RNA-Seq. In line with previous studies in mice, 9 there was a near perfect correlation between genes expressed in educated and hyporesponsive human NK cell subsets, including genes encoding transcription factors for effector loci, granule biogenesis, and mechanistic target of rapamycin (mTOR) regulated metabolism (Fig. 2b). These results were confirmed in a panel of 20 selected qPCR targets comprising transcription factors and canonical cell surface markers linked to NK cell differentiation, GzmB regulation Distance from centrosome (um) Non-Self Self Non-Self Self Non-Self Self Granule Area (um 2 ) Non-Self Self Non-Self Self Granule Area (um 2 ) Non-Self Self Non-Self Self Non-Self Non-Self Self Self 2um 2um and granule biogenesis (Supplementary Fig. 2c and Supplementary Table 1). Notably, we found no difference in granzyme B mRNA expression between the sorted self and non-self specific NK cell subsets ( Fig.  2c) or between NKG2A+ and NKG2A-NK cells (Supplementary Fig. 2d). These data demonstrated that the increased levels of effector molecules such as granzyme B detected by flow cytometry in self-KIR+ NK cells occur independently of transcriptionally regulated programs, including differences in tonic metabolic input to the cell, at least at the level of gene expression.
In mouse NK cells, expression of granzyme B is further regulated by cytokine-induced translation from a preexisting pool of mRNA transcript. 15 Therefore, we explored the possibility that self and non-self specific NK cells may respond differentially to cytokine priming in vivo, resulting in divergent steady-state levels of expressed granzyme B. To address this possibility, NK cells exposed to IL-15 or IL-21 for various lengths of time were monitored for granzyme B content using flow cytometry (Fig. 2d). Both self and non-self KIR+ CD56 dim NK cells displayed increased levels of granzyme B in response to IL-15 and IL-21 stimulation. Notably, the relative differences in granzyme B between self and non-self specific NKG2A-CD57-NK cells were similar after stimulation with IL-15 or IL-21 ( Fig. 2d). Furthermore, blockade of STAT-5 and mTOR signaling with Pimozide and Torin-1, respectively, abolished the cytokine-induced increase in granzyme B in both self and non-self specific NK cells (Fig. 2e). Importantly, the same treatment with Pimozide and Torin-1 did not further reduce the pre-existing constitutive levels of granzyme B that preceded cytokine stimulation. Similar effects were noted with the Janus kinase inhibitor Ruxolitinib and the STAT3 inhibitor S3I-201 ( Supplementary Fig. 3a). Furthermore, no differences were observed in the global levels of LAMP1 ( Supplementary Fig. 3b), suggesting that active lysosomal biogenesis, known to be driven by TFEB, also did not account for the differences in granzyme B content. Similar to stimulation with cytokines, the rate at which lysosomal biogenesis was induced by starvation was equivalent in self and non-self specific NK cells ( Supplementary Fig 3c-d). These data indicate that the observed differences in granzyme B levels at rest are stable and refractory to interference of either STAT or mTOR signaling, suggesting that a stable pool of granzyme B is retained by NK cells independently of con-stitutive input through cytokine or metabolic signals. Accumulation of converged dense-core secretory granules in self-KIR+ NK cells The finding that self-KIR+ NK cells expressed higher levels of granzyme B independently of gene expression provided an initial insight into possible post-transcriptional mechanisms underlying the increased functional potential associated with NK cell education. Granzyme B is sequestered into granular structures within the acidic compartment of the cell. To determine whether the increased levels of granzyme B in self-specific NK cells was a result of higher density, number, or size of cytolytic granules, or a combination thereof, NKG2A-CD57-NK cell subsets were sorted ex vivo into self-or non-self specific NK cell subsets and imaged by confocal microscopy. Corroborating the difference in granzyme B expression observed using flow cytometry, self-KIR+ NK cells had a higher overall intensity of granzyme B staining, localized within granular structures ( Fig. 3a and Supplementary Fig. 4a). The average number of granzyme B containing granules, as defined by discrete points of localized staining intensity, was similar between self and non-self specific NK cells. However, self-KIR+ NK cells displayed an increased level of fluorescence intensity for granzyme B in the granular areas (Fig. 3b), which in turn correlated with proximity to the centrosome (Fig. 3c).
Optical resolution limits of confocal microscopy prevented accurate assessment of granule size and determination of areas with high granzyme B intensity. To address more precisely the size and density of individual granules, sorted self-KIR+ and non-self KIR+ NK cells were sectioned and stained for immuno-electron microscopy (Immuno-EM) using anti-granzyme B mAb and Protein A-gold (Fig. 3d). Quantification of gold particles per cellular section revealed overall greater granzyme B staining in educated NK cells, consistent with both the flow cytometry and confocal microscopy ( Fig. 3d-e). Since retention of granzyme B and re-loading of granules depends on the serglycin content in the granular matrix,34 sections of self-KIR+ and non-self KIR+ NK cells were stained for expression of Chondroitin Sulphate 4 (CS4), a predominant glycosaminoglycan side-chain associated with serglycin in cytotoxic lymphocytes. 35 Self-KIR+ NK cells had a higher overall intensity of CS4-staining ( Fig. 3d-e), suggesting that inhibitory interactions with self-MHC may influence the accumulation of the granular ma-trix. Granzyme B and CS4 staining further revealed a small increase in the average granular size, resulting in a larger total granular area, without significant difference in relative cell size ( Supplementary Fig. 4b-d).
Similarly, analysis of gold particle distribution against granular area in immuno-EM images revealed that self-specific NK cells had larger granzyme B-dense granule areas (Fig. 3f), and overall greater granule densities (Fig. 3g). These data provide a link between expression of self-specific inhibitory KIR and retention of enlarged, granzyme B-dense cytotoxic granules.
Self-KIR+ NK cells mobilize large dense-core granules and release granzyme B to levels of nonself-KIR+ following target cell stimulation Recent evidence suggests that the actin meshwork provides physical constraint on the release of secretory granules over a certain size. 36 Therefore, we examined the fate of dense-core granules relative to size following stimulation with K562 cells. Granzyme B release in self KIR+ NK cells was associated with strong granule mobilization, reflected in a higher discrete mode of CD107a expression (CD107a high ) ( Fig. 4a-b), and ac-companied by a decrease in granzyme B expression to the levels observed in non-self specific KIR+ NK cells (Fig. 4c). Immuno-EM of sorted CD107a high NK cells revealed a specific reduction of large and granzyme B-dense granules following target cell interaction ( Fig.  4d-e). Thus, the accumulation of dense-core granules during education and their effective release upon stimulation, provide a plausible explanation for the enhanced cytotoxic potential of self-KIR+ NK cells.

Role of the acidic Ca 2+ stores in determining functional potential in NK cells
Given that the secretory granules in killer cells share common features with the lysosomal compartment, including the maintenance of low pH and markers of lysosomes, we next set out to examine whether the increase in dense-core granules in educated NK cells was affected by regulation of lysosomal activity and whether this has further consequences for global responses to receptor ligation. Given the qualitative differences in the lysosomal compartments in self-KIR+ versus nonself-KIR+ NK cells with respect to granularity, lysosomal Ca 2+ signaling was examined to gauge the relative capacity of the acid-  ic compartment, including the secretory granules, in terms of uptake and release of intracellular Ca 2+ since this bears important implications for NK cell functions including exocytosis 37 and cytokine production. 38 Functional responses of primary NK cells were determined in the presence and absence of glycyl-L-phenylalanine-beta-naphthylamide (GPN), a dipeptide substrate of cathepsin-C associated with release of Ca 2+ from the lysosomes. 39 GPN causes osmotic permeablization of cathepsin C-positive compartments 40, resulting in the collapse of the pH gradient and controlled equilibration of small solutes (dictated by donnan equilibrium), including Ca 2+ , between the acidic compartment and the cytosol. 39 Treatment of resting primary NK cells with GPN dampened the detection of global Ca 2+ -flux in the cytosol in response to ligation with CD16 or a combination of DNAM-1 and 2B4 (Fig. 5a). GPN treatment alone resulted in a low level of mobilization of CD107a+ vesicles to the cell surface (Fig. 5b) but abrogated degranulation and more importantly abrogated the production of IFNg in response to K562 cells (Fig. 5b-c). Similar results were obtained using mefloquine, another lysosomotropic agent that specifically disrupts lysosomal homeostasis through buffering of the acidic pH gradient 41,42 (Fig. 5d). Thus, disruption of the lysosomal compartment not only affects the mobilization of secretory granules but also the production of cytokines, suggesting that cytokine production in response to stimuli requires lysosomal-derived signals. This opens the possibility that differential signalling in self-KIR+ and nonself-KIR+ NK cells may be influenced by lysosomal-derived Ca 2+ signals. Importantly, none of these compounds showed any general cellular toxicity at the doses tested as compared to the positive control L-Leucyl-L-leucine methyl ester (LeuLeuOMe), a lysosomotropic agent known to induce apoptosis in immune cells through induced lysis of cytotoxic granules ( Supplementary Fig. 5). Furthermore, GPN treatment did not interfere with degranulation in response to PMA/Ionomycin, which raises cytosolic free Ca 2+ by directly accessing both intra-and extra-cellular free Ca 2+ . In fact, degranulation in response to treatment with PMA/Ionomycin was inversely correlated with the granzyme B content of the cell (Supplementary Fig. 6a-c). Ionomycin does not work against an acidic pH gradient, 43 thus disruption of the lysosomal compartment with GPN rescued PMA/Ionomycin-induced degranulation in differentiated and self-KIR+ NK cells ( Supplementary Fig. 6d-e). The effects of disrupting the pH and donnan potential of the acidic compartment, which contains the cytolytic granules, support a role for the acidic compartment in modulating the flow of free-cytosolic Ca 2+ , and corroborate the notion that secretory granule may contribute to the tuning of effector responses in NK cells.

Pharmacological inhibition of PI(3,5)P 2 synthesis leads to formation of large secretory granules and enhanced NK cell functionality
The lysosomal compartment undergoes constant modulation through Ca 2+ regulated fission and fusion events. 44 Lysosomal fission is dependent on Ca 2+ release via the lysosome-specific channel, transient receptor potential mucolipin-1 (TRPML1), 46,47,48 which is activated by phosphoinositide 3,5-bisphosphate PI(3,5)P 2 . 49 PI(3,5)P 2 is a downstream derivative of PI3P generated by the PI3P 5-kinase PIKfyve 50 (outlined in Fig.  6a) involved in endosomal maturation and subsequent lysosomal recruitment and degradation. Hence it is well established that PIKfyve is activated downstream of the phosphatidylinositol 3-kinase (PI3K)/ AKT pathway. 51 we speculated that unopposed signaling from activating NK cell receptors may modulate the lysosomal compartment through constitutive activation of the PI3K/AKT/PIKfyve pathway. In support of this model, we found that pharmacological inhibition of PIKfyve in resting NK cells by three different small chemical inhibitors: vacuolin-1, apilimod and YM201636, 50 led to enlargement of the lysosomal compartment (Fig. 6b), accompanied by increased levels of granzyme B (Fig. 6c). Confocal microscopy of vacuolin-1 treated primary NK cells revealed localization of granzyme B within enlarged LAMP-1 + structures (Fig. 6d-e). Treatment of NK cells with vacuolin-1 increased global Ca 2+ flux in response to receptor-ligation (Fig. 6f) and enhanced specific degranulation (and mobilization of granzyme B) and IFNg production in response to stimulation by K562 cells (Fig. 6g-h). Furthermore, the increased granzyme B expression and degranulation following PIKfyve inhibition correlated with increased natural cytotoxicity against K562 cells (Fig.  6i). These results demonstrate that chemical blockade of PIKfyve results in enlargement of the granular compartment and enhanced NK cell functionality. In order to identify the point at which lysosomal disruption (GPN) or lysosomal enlargement (vacuolin-1) interfered with intracellular signaling pathways in NK cells, we probed signaling both proximal and distal to the plasma membrane. Vacuolin-1 had a minimal effect on upstream signaling, including ZAP70 and Lck following ligation of CD16 (Fig. 6j). However, the propagation of down-stream signals through NF-kB was increased by treatment with vacuolin-1. Conversely, disruption of lysosomal Ca 2+ -flux by GPN had the reverse effect on CD16-induced NF-kB signaling (Fig. 6j). Hence, physical modulation of the acidic Ca 2+ stores affects downstream signaling in response to receptor ligation and tunes NK cell effector responses.

TRPML1-mediated modulation of secretory granules in NK cells
PIKfyve is recruited to PI3P positive compartments where it activates the lysosomal Ca 2+ channel TRPML1 via the production of PI(3,5)P 2 . 47,48 Analysis of the transcriptional levels of TRPML1 in discrete NK cell subsets revealed TRPML1 mRNA was expressed at equal levels in all NK cell subset (Fig. 7a). Agonistic stimulation of TRPML using the chemical compound MK6-83, which activates TRPML1 and TRPML3, 52,53,54 resulted in loss of granzyme B (Fig.  7b) and decreased specific degranulation and IFNg responses to K562 cells (Fig. 7c). TRPML3 mRNA was not expressed in resting human NK cells (Data not shown). Conversely, silencing of TRPML1 by siR-NA in resting primary NK cells led to increased levels of granzyme B (Fig. 7d-e). Moreover, in concordance with the effects of pharmacological inhibition of PIKfyve, siRNA silencing of TRPML1 (Fig. 7f) led to enhanced degranulation and IFNg production in primary resting NK cells (Fig. 7g). These results demonstrate a role for TRPML1 in the modulation of granzyme B content and in tuning of effector function in NK cells. potential to self-MHC. However, it has been unclear how receptor input during NK cell education is integrated and retained in order for NK cells to remain self-tolerant, whilst also able to deliver spontaneous, well-tuned functional responses upon subsequent challenges. Our results suggest that unopposed activation signals lead to physical disarming of NK cells through TRPML1-induced modulation of the granular compartment. The accumulation of dense-core secretory granules under the influence of inhibitory self-MHC interactions provides mechanistic insights into the paradox of how inhibitory signaling is translated into a state of enhanced functional potential that persists between successive cell-to-cell contacts. A variety of models and nomenclature have been used to describe the process of NK cell education . However, regardless of whether the functional phenotype is caused by gain of function (arming/stimulatory licensing) in self-KIR+ NK cells or loss of function (e.g. disarming/inhibitory licensing) in nonself-KIR+ NK cells, 3, 55 the net outcome of these processes is a consistent difference in the intrinsic functional potential of cells carrying self-and non-self receptors at rest. A structural basis for the difference in functional responsiveness has recently been proposed, 9, 10 whereby educated NK cells display a unique compartmentalization of activating and inhibitory receptors at the nano-scale level on the plasma membrane. Complementing this pre-existing phenotype, we show that NK cell education is also tightly linked to the accumulation of large, granzyme B-rich dense-core cytotoxic granules, located closer to the centrosome in resting NK cells. While the accumulation of dense-core granules, and their specific release during the effector phase, hold important clues to the basis for the increased cytotoxic potential of self-KIR+ NK cells, several outstanding questions remained to be addressed. First, it remained unclear whether the observed accu-mulation of dense-core granules in self-KIR+ NK cells also contributed to the globally enhanced responsiveness associated with NK cell education. 56 An increasing body of evidence supports the role of the acidic compartment not only in the triggering of Ca 2+ signaling, but also in the spatiotemporal coordination of signaling cascades 57,58,59,60 and the regulation of receptor degradation. 51 In both T cells and NK cells, lysosomal Ca 2+ release plays an important role in degranulation. 37,57 Signaling from the lysosomal compartment was also recently shown to regulate the migratory behavior of dendritic cells in a TRPML1-dependent fashion. 61 Our data suggest that there is a quantitative relationship between modulation of the granular compartment under the influence of inhibitory receptors and intrinsic functional potential of NK cells. While the exact role of the acidic Ca 2+ store for the enhanced functional potential in self-KIR+ NK cells remains elusive, pharmacological inhibition of Ca 2+ release from the intracellular acidic stores, together with analysis of Ca 2+ -flux, consistently pointed to a role for the secretory granule in propagating surface receptor signaling. Notably, a correlation between the size of the lysosomes and level of Ca 2+ -flux has previously been describe in fibroblasts from patients with Parkinson disease. 62 On that note, it is tempting to speculate that the gain of natural cytotoxicity in lL15 stimulated CD56 bright NK cells may be likewise related to the associated enlargement of the granular compartment. 63 Second, we examined the molecular pathway that led to accumulation of dense-core granules in self-KIR positive NK cells, or rather the lack of accumulation of such granules in hyporesponsive NK cells. The difference in granule size and densities in the absence of active lysosomal biogenesis, led us to explore the pathways involved in the continuous modulation of the lysosomal compartment through fission and fusion events 45. Several activating receptors have been implicated in NK cell education, including NKG2D, SLAM family receptors and DNAM-1. 6, 11, 64 NKG2D and DNAM-1 signaling activates the PI3K/AKT pathway. 65,66 Engagement of inhibitory receptors and SHP-1 signaling block NK cell activation at an early stage of the activation signaling pathway, preventing actin cytoskeleton rearrangement and the recruitment and phosphorylation of activation receptors. 67,68,69 The PI3K/AKT pathway is also controlled by SHIP1, which has also been implicated in tuning the effec-tor function of NK cells during education. 70,71 Notably, PIKfyve and TRPML1 are activated downstream of the PI3K/AKT pathway. 51 Mutations in TRPML1 cause mucolipidosis type IV, which is characterized by enlarged lysosomes. 72,73 We hypothesized that continuous unopposed signaling through activating receptors during homeostatic cell-cell interactions may trigger the AKT/PIKfyve/TRPML1 axis and thereby promote lysosomal fission, leading to an inability to accumulate dense-core granules. To explore this hypothesis, we used a combination of pharmacological agonists and antagonists combined with genetic approaches to interfere with the PIKfyve/TRPML1 pathway. Pharmacological inhibition of PIKfyve by small chemical compounds, including vacuolin-1 and apilimod is known to cause enlarged lysosomes in several cell types, including mast cells and macrophages. 50,74,75 Here, we show that inhibition of PIKfyve by three different chemical compounds caused enlargement of the lysosomal compartment. Importantly, this was associated with increased granzyme B expression, increased Ca 2+ -flux, and more potent effector function. A similar functional phenotype was obtained when silencing the lysosome-specific Ca 2+ release channel TRPML1. Together these data support a model where tonic or intermittent activation signals through the PI3K/AKT pathway result in PIKfyve activation and TRPML1-induced lysosomal fission, ultimately leading to lack of dense-core granules and reduced functional potential in NK cells. In Chediak-Higashi Syndrome (CHS), mutation of the LYST gene leads to the formation of giant secretory lysosomal structures. 76 NK cells in CHS patients are hyperresponsive and hypersecretory but are unable to degranulate. 77,78 Although NK cell activation followed by granule convergence and polarization appears to be normal in LYST-deficient NK cells, the enlarged granules fail to pass through the cortical actin meshwork openings at the immunological synapse. 36 By monitoring granule size and granzyme B density prior to and following degranulation, we observed a selective loss of the pre-converged, large dense-core granules after degranulation. These large granules had an area above 0,2 mm 2 , corresponding to a diameter of around 500nm, pointing to a possible difference in the density of the actin meshwork and physical restriction of degranulation between resting primary NK cells and NK92 cells. However, another contributing factor may be LYST-mediated modulation of granules during during the effector response allowing smaller proportions of the large granules to be released, as has been demonstrated for degranulation in mast cells. 79 An outstanding question is how lysosomal fission leads to loss of the granular matrix and granzyme B in hyporesponsive NK cells. Granzyme B can be synthesized and secreted directly through the constitutive secretory pathway. 80 It is possible that an enhanced rate of lysosomal fission during weakly agonistic cell-cell interactions and the corresponding failure to accumulate dense-core granules in NK cells lacking self-specific receptors leads to loss of granzyme through the secretory route. 81 Indeed, NK cells in serglycin -/mice lack dense-core granules, retain less granzyme B which is secreted from the cell at a greater rate, and exhibit reduced degranulation in response to stimuli. 34 Another remaining challenge is to decipher when and where TRPML1-mediated physical disarming takes place. Transfer experiments in mice have established an indisputable role for cell-to-cell interactions in shaping the functionality of mature NK cells. 82,83,84 Although the detailed time-scale and spatial aspects of such cell interactions remain largely unknown, transfer of functional NK cells to MHC-deficient environments leads to induction of hyporesponsiveness. 4 SHP-1 intersects signaling of activating receptors upstream of Vav-1 69 and rapidly shuts down the process of forming an activating NK cell synapse with target cells. 85 While the inhibitory synapse and the productive cytolytic synapse have been studied in great detail, much less is known about immune synapses formed between resting immune cells during homeostasis. It is possible that cells lacking self-specific inhibitory receptors form a succession of non-cytolytic immune synapses under homeostasis leading to loss of dense-core granules and "leakage" of their functional potential. 81 It has previously been shown that trans-presentation of IL-15 to NK cells, resulting in activation of AKT is negatively regulated by inhibitory interactions with self MHC. 86 Thus, it is possible that unopposed constitutive IL-15 activation may occur in NK cells that lack self-specific inhibitory KIR, which in turn would affect granular stability and/or retention through the mechanism described here. The dose dependant induction of granzyme B expression in response to cytokine, or viral infection, is connected to activation of the metabolic check-point kinase mTOR. 19 Notably, however, we did not observe any transcriptional imprint in the mTOR pathway when we examined circulating NK cells at rest, arguing against a major role for metabolism in the persistence of the unique granular phenotype seen in circulating blood self-KIR+ NK cells. It was recently shown that mTOR activation contributed to the functional rheostat during effector responses in educated murine NK cells. 87 Interestingly, mTOR activation and function are dependent on its lysosomal localization and the vacuolar H(+) -ATPase (V-AT-Pase) activity. 88 Thus, the difference in lysosomal composition described in the present study could potentially contribute to enhanced mTOR activation observed in educated NK cells upon stimulation. 87 In conclusion, our findings suggest a mechanism by which NK cell education operates through modulation of dense-core granules under the influence of inhibitory receptor ligand interactions. Qualitative differences in the morphology of the lysosomal compartment and signaling from acidic Ca 2+ stores, allow the cytolytic machinery to operate independently of transcription during the effector response. Furthermore, the data suggest that it may be possible to boost NK cell functionality through targeted manipulation of the granular matrix and Ca 2+ homeostasis within lysosome-related organelles.

Cells
Buffy coats from random healthy blood donors were obtained from the Karolinska University Hospital and Oslo University Hospital Blood banks with informed consent. Peripheral blood mononuclear cells were separated from buffy coats by density gravity centrifugation (Lymphoprep; Axis-Shield) using fretted spin tubes (Sepmate; Stemcell Technologies). Genomic DNA was isolated from 200µl of whole blood using DNeasy Blood and Tissue Kit (Qiagen). KIR ligands were determined using the KIR HLA ligand kit (Olerup SSP) for detection of the HLA-Bw4, HLA-C1, and HLA-C2 motifs. NK cells were purified using negative selection (Miltenyi) with an AutoMACS Pro Seperator. K562 cells were maintained in RPMI +10%FCS.

Phenotyping by Flow Cytometry
Isolated PBMC were stained for flow cytometric analysis using an appropriate combination of antibodies as detailed in the supplementary experimental procedures. After surface staining, cells were fixed and permeabilized using a fixation/permeabilization kit (BD Bioscience Cytofix/Cytoperm) prior to intracellular staining with anti-granzyme B-A700 (GB11). Samples were acquired using an LSRII flow cytometer (Becton Dickinson) and data was analyzed using FlowJo V10.0.8 (TreeStar). Stochastic neighbor embedding (SNE) analysis was performed as described in supplementary experimental procedures

RNAseq and qPCR
RNASeq was performed using single-cell tagged reverse transcription (STRT), a highly multiplexed method for single-cell RNA-seq. Real time quantitative PCR was used to study the difference in the expression of 20 genes of interest in sorted differentiation and education subsets of NK cells, as detailed in supplementary experimental procedures.

Confocal fluorescence microscopy and Image analysis
Staining and preparation of cells for confocal microscopy is detailed in supplementary experimental procedures. The cells were examined with a Zeiss LSM 710 confocal microscope (Carl Zeiss MicroImaging GmbH) equipped with an Ar-Laser Multiline (458/488/514nm), a DPSS-561 10 (561nm), a Laser diode 405-30 CW (405nm), and a HeNe-laser (633nm). The objective used was a Zeiss plan-Apochromat 63xNA/1.4 oil DICII. Image processing and analysis were performed with basic software ZEN 2011 (Carl Zeiss MicroImaging GmbH) and Imaris 7.7.2 (Bitplane AG). Confocal z-stacks were deconvolved using Huygens Essential 14.06 (Scientific Volume Imaging b.v.), and visualized with Imaris.

Electron Microscopy
Preparation of cells for electron microscopy is detailed in supplementary experimental procedures. Ultrathin sections (70-90 nm) of cell pellets were cut on a Leica Ultracut (equipped with UFC cryochamber) at -110C, picked up with a 50:50 mixture of 2.3 M sucrose and 2% methyl cellulose. Sections were then labeled with antibodies against granzyme B (496B, eBioscience) or Chondroitin Sulphate 4 (2B6, AMSBIO), followed by a bridging rabbit-anti-mouse antibody (DAKO, Denmark) and protein A gold (University Medical Cen-ter, Utrecht, Netherlands). Microscopy was done at 80 kV in a JEOL_JEM1230 and images acquired with a Morada camera. Further image processing was done in Adobe Photoshop. Quantification was done according to established stereological procedures.

Ca 2+ Flux Assay
Freshly isolated NK cells were incubated with Fluo-4 for 30 min at 37C in PBS+2%FCS at the recommended dilution (Fluo-4 Imaging kit, Molecular Probes). Cells were then washed twice and incubated with biotinylated CD16 or biotinylated DNAM-1/2B4 (Miltenyi), with the addition of labeled specific antibodies for CD56, CD57, NKG2A, KIR2DL1, KIR2DL1/S1, KIR3DL1/S1 and KIR2DL2/L3/S2, for 10 min at room temperature. The cells were washed once more and placed on ice until assayed. Prior to FACS analysis, the cells were pre-warmed at 37C for 5 min in the presence or absence of GPN (50mM final concentration) or vacuolin-1 (10mM final concentration). Cells were immediately run on FACS for 30s, followed by addition of 10ug/mL Streptavidin and run for a further 4 min. Ionomycin was added at 4mM final concentration and run for a further 1 min. Ca 2+ -flux kinetics were analysed by FlowJo V10.0.8 (TreeStar).

siRNA Interference
Primary NK cells were isolated, rested for 2 hours and transfected either directly or Primed with 10ng/ mL IL15 for 72 hours and transfected. NK cells or cell lines were transfected by Amaxa nucleofection (Lonza) using 300pM of Dharmacon ON-TARGET plus SMARTpool control siRNA, or SMARTpool RNA targeting human TRPML1. Nuclefection was performed using the human macrophage kit using program Y-010. After nucleofection, cells were rested for 4 hours in OPTI-MEM, before an equal volume of culture medium with 2ng/ml IL15 was added. The cells were then cultured for 48 hours before phenotypic and functional testing. siRNA efficiency was determined using qPCR.

Statistical analysis
Comparisons of matched groups were made using paired Students T test. Single comparison of groups or populations of cells between donors was performed using Students t test or Mann-Whitney test for statistical significance. n.s. indicates not significant; ****p < 0.0001***; p < 0.001; **p < 0.01; and *p < 0.05. Analyses were performed using GraphPad Prism software.