P2Y6 receptor signaling in natural killer cells impairs insulin sensitivity in obesity

Natural killer (NK) cells contribute to the development of obesity-associated insulin resistance and have previously been shown to up-regulate the expression of the P2Y purinoreceptor 6 (P2Y6R) upon high fat diet (HFD)-induced obesity. Here, we reveal that NK cell-specific inactivation of the P2Y6R gene improves insulin sensitivity in obese mice and reduces the expression of chemokines in adipose tissue infiltrating NK-cells. Obese mice lacking P2Y6R specifically in NK cells exhibited a reduction in adipose tissue inflammation, exhibited improved insulin-stimulated suppression of lipolysis in adipose tissue and a reduction in hepatic glucose production, leading to an overall improvement of systemic insulin sensitivity. In contrast, myeloid lineage specific P2Y6R inactivation does not affect energy or glucose homeostasis in obesity. Collectively, we show that P2Y6R signaling in NK cells contributes to the development of obesity-associated insulin resistance and thus might be a future target for the treatment of obesity-associated insulin resistance and type 2 diabetes.

NK cells isolated from adipose tissue of mice fed a HFD when compared to lean mice 119 on a control diet (Theurich et al., 2017). NK cells are part of the innate immune system 120 responsible for the recognition and elimination of virus-infected and cancerous cells 121 (Vivier et al., 2008). They are able to destroy target cells via the release and joined 122 action of the cytotoxic mediators perforin and granzyme (Pardo et al., 2002) or through 123 the exocytosis of cytokines such as interferon gamma (IFNγ) and tumor necrosis factor 124 alpha (TNFα) (Girart et al., 2007;Wang et al., 2012;Jewett et al., 1996). NK cell activity 125 infiltrating the adipose tissue of obese animals and humans are arranged around dead 134 adipocytes, forming characteristic crown-like structures (CLS) (Murano et al., 2008;135 Cinti et al., 2005). These necrotic adipocytes might release UDP as danger signal 136 leading to an increased infiltration or local proliferation of immune cells, thus promoting 137 metaflammation and development of insulin resistance. 138 Given the improvement of insulin action in conventional P2Y6R knockout mice 139 (Steculorum et al., 2017;Jain et al., 2020), and the role of P2Y6R-signaling in immune 140 cells, we aimed to investigate the role of this receptor in NK cells and macrophages in 141 the development of obesity-associated metaflammation and insulin resistance. To this 142 end, we generated mice lacking P2Y6R specifically in these immune compartments. 143 While ablation of P2Y6R in myeloid lineage cells had no effect on metaflammation and 144 obesity-associated insulin resistance, obese mice lacking P2Y6R specifically in NK 145 cells exhibited a reduction in adipose tissue inflammation, exhibited improved insulin-146 stimulated suppression of lipolysis in adipose tissue and a reduction in hepatic glucose 147 production, leading to an overall improvement of systemic insulin sensitivity. 148 Collectively, these experiments extend the tissue-specific actions of P2Y6R signaling 149 in obesity, further characterizing P2Y6R inhibition as a novel target for metabolic 150

Deletion of the P2Y6R gene in NK cells does not protect from HFD-induced 153 obesity 154
Among other genes, P2Y6R expression contributes to the discrimination of NK cells 155 from lean and obese mice (Theurich et al., 2017). Furthermore, it has been reported 156 that circulating uridine levels are increased in obesity (Steculorum et al., 2015;Deng 157 et al., 2017). To further define potential changes in uridine and UDP in different 158 organismal compartments upon development of HFD-induced obesity and insulin 159 resistance, we compared uridine and UDP levels in plasma, liver, adipocytes and the 160 adipose tissue stromal vascular fraction (SVF) of mice that had been kept on normal 161 chow diet (NCD) or had been exposed to HFD-feeding for 6 or 12 weeks. This analysis 162 revealed increased levels of uridine in plasma of obese compared to lean mice after 163 12 weeks of HFD feeding (Suppl. Fig. S1A), while uridine levels in liver and SVF 164 increased already at 6 weeks of HFD feeding and remained elevated until 12 weeks of 165 HFD feeding (Suppl. Fig. S1B, C), while adipocyte uridine concentration increased 166 after 6 weeks of HFD feeding, but reached comparable levels to NCD-fed mice after 167 12 weeks (Suppl. Fig. S1D). While UDP levels in plasma paralleled the increase of 168 circulating uridine 12 weeks after HFD-feeding (Suppl. Fig. S1E) as well as the 169 increases in adipose tissue SVF 6 and 12 weeks after high fat diet feeding (Suppl. Fig.  170 S1G), hepatic UDP levels remained unaltered upon HFD feeding (Suppl. Fig. S1F), 171 and adipocyte UDP levels increased only after prolonged HFD feeding (Suppl.

177
In light of the coordinated regulation of uridine and UDP in plasma and adipose 178 tissue SVF, compartments which directly interact with immune cells, we aimed to 179 investigate the impact of P2Y6R signaling in NK cells on body weight development, 180 glucose metabolism, and obesity-associated metaflammation in vivo. In order to 181 specifically delete the P2Y6R in NK cells, we crossed mice carrying a loxP-flanked 182 P2Y6R gene (Jain et al., 2020) with Ncr1-Cre mice (Eckelhardt et al., 2011). PCR 183 analysis of DNA isolated from NK cells of control (P2Y6 fl/fl ) or NK cell-specific P2Y6-184 deleted (P2Y6 ΔNcr1 ) mice confirmed successful and specific disruption of P2Y6R 185 expression in NK cells (Suppl. Fig. S2). 186 Starting at the age of 6 weeks, P2Y6 ΔNcr1 mice and their respective littermate controls 187 were fed a HFD or control diet (CD). HFD feeding did not result in any differences in 188 body weight between genotypes in male mice (Fig.1A), while female P2Y6 ΔNcr1 mice 189 exhibited a mild protection from HFD-induced weight gain upon prolonged HFD-190 feeding (Suppl. Fig. S3A). Consistent with unaltered body weight development, the 191 relative amount of lean and fat mass of male mice after 15 weeks on HFD was not 192 different between genotypes (Fig. 1B). In line with this, male P2Y6 ΔNcr1 mice did not 193 exhibit any differences in food intake (Suppl. Fig. S4A), water consumption (Suppl. Fig.  194 S4B), energy expenditure ( Fig. S4C) or locomotor activity (Fig. S4D) compared with 195 control mice on HFD. However, after 14 weeks of HFD feeding we could observe a 196 slight reduction in liver weights in male (Fig. 1C) and female (Suppl. Fig. S3B) 197 P2Y6 ΔNcr1 mice, as well as a decrease in fasted leptin levels in male P2Y6 ΔNcr1 mice 198 compared to their control littermates (Fig. 1D). Collectively, NK-cell P2Y6R signaling 199 has no major impact on energy homeostasis and body weight control in obesity, while 200 it appears to affect liver weight in obesity. 201 202 Fig. S2: P2Y6R is selectively deleted in NK cells. Splenic cells from P2Y6 fl/fl (n=2) and P2Y6 ΔNcr1 (n=2) mice were separated with magnetic beads resulting in two fractions: NK cells and nonNK cells. Genomic DNA was extracted from each group. Genotyping for P2Y6R was performed by competitive PCR using the primers listed in table 1 which resulted in a 305 bp PCR-product (floxed) and a 370 bp PCR-product (P2Y6R deletion) in NK cells isolated from P2Y6 ΔNcr1 (Ncr1-Cre positive mice (TG band 450 bp)). This deletion band was neither detected in NK cells from P2Y6 fl/fl , (Ncr1-Cre negative mice (WT band 650 bp)) nor in nonNK cells of either genotype.

P2Y6R signaling in natural killer cells impairs insulin sensitivity in HFD mice 206
Next, we aimed to determine the impact of P2Y6R signaling in NK cells on glucose 207 metabolism and insulin sensitivity. To this end, we performed insulin tolerance tests 208 (ITT) and glucose tolerance tests (GTT) at various timepoints upon exposure to CD or 209 HFD. Compared to controls, deletion of the P2Y6R from NK cells improved insulin 210 sensitivity after 13 and 20 weeks of HFD feeding in both genders ( Fig Fig. S5F-H). While we did not 214 observe any differences in fasted blood glucose levels between P2Y6 ΔNcr1 mice and 215 their control littermates on HFD (Fig. 2G), fasted insulin concentrations were reduced 216 in P2Y6 ΔNcr1 mice as early as after 10 weeks of HFD feeding (Fig. 2H) concomitant with 217 a reduction of the homeostasis model assessment index for insulin resistance (HOMA-218 IR) (Fig. 2I). These effects were HFD-specific and could not be observed in mice on 219 control diet (Suppl. Fig. S5I-J). Taken together, abrogation of P2Y6R-signaling from NK 220 cells, partially protects from the development of obesity-associated insulin resistance. 221

NK cell-specific deletion of P2Y6R improves insulin-induced suppression of 226 hepatic glucose production in obesity 227
To further dissect tissue-specific effects of P2Y6R signaling in NK cells on insulin 228 sensitivity and glucose homeostasis, we performed hyperinsulinemic-euglycemic 229 clamp studies in HFD-fed P2Y6 ΔNcr1 mice and their littermate controls. Mice with NK 230 cell-specific depletion of the P2Y6R required significantly higher glucose infusion rates 231 (GIR) compared to their littermate controls (Fig. 3A) in order to maintain euglycemic 232 blood glucose levels during the steady state of the clamp experiment (Fig. 3B), while 233 exhibiting comparable levels of circulating human insulin concentrations (Fig. 3C). 234 Quantitative assessment of insulin's ability to suppress gluconeogenesis in both 235 groups of mice revealed that P2Y6 ΔNcr1 mice exhibited a clear improvement in insulin´s 236 ability to suppress hepatic glucose production (HGP) (Fig. 3D). Since hepatic glucose 237 production is controlled both through insulin-induced changes in the expression of key 238 enzymes of gluconeogenesis as well as through gluconeogenic substrate availability, 239 such as fatty acids released from adipose tissue through lipolysis, we next compared in adipose tissue (Nielsen et al., 2014). We found that insulin inhibits the 248 phosphorylation of HSL at serine 563 in P2Y6 ΔNcr1 mice (Fig. 3F), which is a main 249 phosphorylation site known to regulate HSL-driven lipolysis (Nielsen et al., 2014), while 250 phosphorylation of serine 565 only showed a mild increase in P2Y6 ΔNcr1 mice (Fig. 3F). 251 However, in contrast to an enhanced ability of insulin to suppress HGP and 252 adipose tissue lipolysis, insulin-stimulated uptake of [ 14 C]-D-glucose into perigonodal 253 adipose tissue (PGAT), brown adipose tissue (BAT), brain, and skeletal muscle (SKM) 254 remained unaltered in P2Y6 ΔNcr1 mice compared to their controls (Fig. 3G). Consistent 255 with unaltered insulin-stimulated glucose uptake in these organs, also assessment of 256 insulin-stimulated protein kinase B (AKT) phosphorylation during the clamp failed to 257 reveal significant differences in P2Y6 ΔNcr1 mice compared to their controls (Fig. 4). 258 Moreover, insulin's ability to activate AKT phosphorylation in liver did not differ between 259 the two groups of animals (Fig. 4). Thus, deletion of P2Y6R from NK-cells in obesity 260 appears to enhance insulin-stimulated suppression of lipolysis in PGAT and thus 261 gluconeogenic substrate supply to liver. 262  Organs of clamped P2Y6 fl/fl (n=9) and P2Y6 ΔNcr1 (n=12) mice on HFD (9 weeks) were further investigated by western blot analysis. Representative immuno-blots for phosphorylated AKT (Ser473), total AKT and Calnexin in PGAT, BAT, liver, brain and SKM of clamped mice. Quantification of pAKT Ser473 levels was performed after normalization to Calnexin and total AKT as loading controls. Statistics: Unpaired twotailed t-test, *p≤0.05. Data are represented as mean ± SEM. ○ denotes individual mice.

NK cell-specific deletion of P2Y6R upregulates genes associated with 265 triglyceride homeostasis and increases mitochondrial function in liver 266
To further investigate the effects of NK cell-specific P2Y6R signaling on liver and 267 adipose tissue, we compared gene expression profiles of liver and PGAT from 268 Interestingly, analysis of the overlap of commonly regulated genes in liver and 283 PGAT revealed that 10 genes (Lilra5, Rora, Herc1, Bnip2, Lactb, Ciao2a, Rps27l, Ice2, 284 Anxa2, Vps13c) were similarly differentially expressed in both organs (Suppl. Fig. S6). 285 Importantly, some of them are known to be specifically expressed in immune cells and 286 NK cells, thus raising the possibility that some of the commonly observed differences 287 in both tissues might result from their coordinated expression changes in tissue-288 infiltrating NK cells. Here, only Lilra5 expression was found to be up-regulated, 289 whereas the expression of Rora and Herc1 was down-regulated (Suppl. Fig. S6). As triglyceride (TG) homeostasis was affected by NK cell-specific deletion of 296 P2Y6R, we conducted lipidomic analyses of liver samples from mice fed a HFD for 9 297 weeks. These analyses revealed no significant overall changes in hepatic TG (Fig. 5G) 298 and diacylglyceride (DAG) contents (Fig. 5H), however some specific lipid species 299 showed a trend towards reduction in liver of P2Y6 ΔNcr1 compared to controls (C16:0, 300 C18:0 TGs, C16:0-C18:1, C18:1-C18:1 DAGs). 301 Since hepatic lipid homeostasis is closely linked to mitochondrial function in liver, 302 we next investigated the rates of oxidative capacity and extracellular acidification rates 303 (ECAR) in micro-punches from liver of HFD-fed P2Y6 ΔNcr1 mice and their control 304 littermates. This analysis revealed a significant upregulation of the ECAR (Fig. 5I), 305 indicating an increase in glycolysis, while the oxygen consumption rate (OCR) was 306 only slightly but not significantly improved in liver of HFD-fed P2Y6 ΔNcr1 mice compared 307 to their control littermates (Fig. 5I). 308 Collectively, NK cell-specific P2Y6R signaling appears to affect triglyceride 309 metabolism and mitochondrial function in liver, while coordinately regulated differences 310 in gene expression between liver and adipose tissue possibly arise from NK cell 311 intrinsic changes in gene expression in tissue-infiltrating NK cells. , RAR-related orphan receptor alpha (Rora), Perilipin 2 (Plin2) and Leukocyte immunoglobulin-like receptor subfamily A member 5 (Lilra5) was quantified in clamped liver and PGAT of P2Y6 fl/fl (n=5) and P2Y6 ΔNcr1 (n=5) mice. Quantification was determined by the 2 -ΔΔCt method using the housekeeping gene TATA-binding protein (Tbp) for normalization. (G-H) Lipidomic analysis of livers from P2Y6 fl/fl (n=8) and P2Y6 ΔNcr1 (n=4) mice on HFD (9 weeks). (I) Seahorse analysis in liver punches of P2Y6 fl/fl (n=5) and P2Y6 ΔNcr1 (n=4) mice on HFD (9 weeks). Baseline oxygen consumption rate (OCR) as measure for respiration was plotted against the extracellular acidification rate (ECAR) as measure for glycolysis. Statistics: (C-I) Unpaired two-tailed t-test, *p≤0.05, **p≤0.01, ***p≤0.001, ****p≤0.0001. Data are represented as mean ± SEM. ○ denotes individual mice. of lymphocyte chemotaxis, lymphocyte migration and neutrophil migration (Fig. 6C), 331 whereas pathways of protein folding, and chaperone cofactor-dependent protein 332 folding were detected in liver-derived NK cells (Fig. 6D). 333

334
Consistent with our previous findings in liver and adipose tissue, we also found 335 genes of the presumably NK cell-specific gene set to be differentially expressed in 336 isolated NK cells of both tissue origins (Fig. 6E). This supports our hypothesis that 337 similar differences in gene expression between liver and adipose tissue possibly arise 338 from NK cell intrinsic changes in gene expression of tissue-infiltrating NK cells. Besides 339 this first NK cell-specific gene set, we found a second set of overlapping genes in 340 isolated NK cells derived from adipose tissue and liver. This gene set includes the up-341 regulated genes Trip4, Stip1, Gtf2a2 and Hsph1 and the down-regulated genes Rab8b, 342 Usp3, Ppib, F2rl3 and Snx1 (Fig. 6F). Usp3, Ppib and Snx1 have also been found to 343 be differentially expressed in liver and Rab8b in adipose tissue of clamped P2Y6 ΔNcr1 344 mice when compared to their littermate controls (data not shown). 345 Notably, in adipose tissue-derived NK cells the gene expression of the 346 chemokines Ccl3, Ccl4, Xcl1 and Cxcl16 (Fig. 6G) was found to be down-regulated in 347 parallel with the down-regulation of Csf1 expression in hepatic NK cells (Fig. 6B). As 348 these chemokines are involved in the recruitment and activation of immune cells, we 349 aimed to elucidate the impact of this P2Y6R-dependent and NK cell-specific down-350 regulation on the infiltration of immune cells into different organs. In line with this 351 reduced expression of chemokines in NK cells, we observed a significant reduction of 352 NK cell numbers in liver and spleen, a decrease in NKT cell numbers in PGAT and a 353 reduction of macrophages in spleen of P2Y6 ΔNcr1 mice compared to control (Fig. 7A). 354 In addition, infiltration of MAC2 positive macrophages into adipose tissue was reduced 355 in mice with NK cell-specific P2Y6R deletion after prolonged HFD-feeding (Suppl. Fig.  356 S8A), concomitant with a mild but non-significant reduction in the expression of 357 inflammatory cytokines and chemokines (Suppl. Fig. S8B). On the contrary, expression 358 of inflammatory markers in liver of mice fed a HFD for 22 weeks were not significantly 359 altered between genotypes (Suppl. Fig. S8C). NK cells could be found (Fig. 7B). However, abrogation of P2Y6R signaling did not 369 affect the overall maturation state in NK cells (Fig. 7B).  expression. Here, we found a more than three-fold upregulation of P2Y6R expression 394 in M2-polarized compared to M1-polarized bone marrow-derived macrophages (Suppl. 395 Fig. S9B). 396 We then investigated the impact of macrophage-specific P2Y6R signaling on 397 the metabolic phenotype of HFD-fed mice. Mice with macrophage-specific deletion of 398 P2Y6R showed the same body weight development as their control littermates (Fig.  399 8A) as well as no differences in body fat content (Fig. 8B). In parallel, investigations of 400 insulin sensitivity and glucose tolerance after 10, 14 or 20 weeks of HFD-feeding did 401 not reveal any significant differences between P2Y6 ΔLysM and P2Y6 fl/fl mice (Fig. 8C-H). 402 This was in line with results from indirect calorimetry in mice fed a HFD for 18 weeks. 403 Analyses of food intake, water consumption, energy expenditure, total activity or 404 respiratory exchange rate (RER) did not reveal any differences between P2Y6 ΔLysM and 405 P2Y6 fl/fl mice (Suppl. Fig. S9C-G). Also, insulin and leptin levels in plasma of 16 hours-406 fasted mice did not differ between genotypes concomitant with unaltered HOMA-IR 407 ( Fig. 8I-K). 408 Since UDP levels are increased in adipocytes (Suppl. Fig. S1H) where it can 409 serve as a "find me" signal triggering macrophages to eliminate necrotic cells, such as 410 hypertrophic and necrotic adipocytes (Cinti et al., 2005), we examined the infiltration 411 of macrophages into adipose tissue of HFD-fed mice by immunohistochemistry. 412 However, we did not find any differences in the amount of MAC2 positive macrophages 413 between in P2Y6 ΔLysM and P2Y6 fl/fl mice (Suppl. Fig. S10). In line with the overall 414 inconspicuous phenotype of P2Y6 ΔLysM mice, we also did not observe any differences 415 in lipid accumulation in liver between genotypes (Suppl. Fig. S10). 416 In conclusion, we demonstrated here that P2Y6R-signaling in macrophages 417 does not play a major role in the context of obesity-induced insulin resistance and 418 metaflammation. 419     (Crinier et al., 2018;Zhao et al., 2020). Despite this tissue-486 specificity, we found a common set of ten genes that were similarly differentially 487 expressed in adipose tissue and liver of clamped, thus insulin-stimulated mice. As we 488 assumed that this gene set reflects that of tissue-infiltrating NK cells, rather than being 489 an insulin-regulated gene set, we further investigated the transcriptome of isolated NK 490 cells from adipose tissue and liver of HFD-fed P2Y6 fl/fl and P2Y6 ΔNcr1 mice. Also here, 491 transcripts of this gene set were found to be differentially expressed in NK cells of both 492 origins, supporting our hypothesis that this gene set is NK cell-specific and that the 493 of our NK cell-specific gene cluster to any of these available gene sets with the 510 exception of Rora, whose expression has been reported to be higher in liver-resident 511 DX5 -NK cells (Peng et al., 2013). 512 The only gene that was consistently up-regulated in all our transcriptomic data 513 sets was Lilra5. This receptor belongs to the family of leukocyte immunoglobin-like 514 receptors which are known to regulate leukocyte activation. Lilra5 has been reported 515 to exist as transmembrane protein or as secreted molecule (Borges et al., 2003). The In light of macrophages' roles as drivers for obesity-induced metaflammation (Xu et al., 534 2003) and an increase in UDP levels in adipose tissue upon HFD-feeding, it was an 535 obvious question whether macrophage-specific P2Y6R-signaling might contribute to 536 obesity-associated metaflammation and development of insulin resistance. However, 537 here we showed that mice with macrophage-specific abrogation of P2Y6R-signaling 538 did not present in any conspicuous phenotype in regard to bodyweight development, 539 insulin sensitivity, food intake or energy expenditure. In addition, we did not observe 540 any alterations in adipose tissue inflammation. As obesity development is 541 accompanied by a macrophage switch from M2-to M1-polarized macrophages in 542 adipose tissue (Castoldi et al., 2016), it was interesting to note that the expression of 543 P2Y6R is 3-fold up-regulated in M2-polarized macrophages compared to unpolarized 544 or M1-polarized macrophages in vitro. Comparison of M1 and M2 polarized 545 macrophages derived from human blood only reported a difference in the expression 546 of other P2Y receptors such as P2Y1R, P2Y12R and P2Y13R that were found to be 547 down-regulated in M1 compared to M2 macrophages and P2Y2R, P2Y8R, P2Y10R 548 and P2Y14R being up-regulated (Gerrick et al., 2018). Another study on murine 549 BMDMs found P2Y1R to be up-regulated in M2-polarized macrophages when 550 compared to unpolarized cells (Jablonski et al., 2015). Nevertheless, our data clearly 551 indicate that P2Y6R signaling in macrophages appears to be dispensable for the 552 manifestation of obesity-induced metaflammation. 553 Taken together, our study demonstrates that insulin sensitivity and 554 metaflammation depend on P2Y6R signaling in NK cells in the context of obesity, which 555 may provide a target for the treatment of obesity-associated pathologies and diabetes. 556 557 558

Materials and Methods 559
Animals 560 All mouse experiments were approved by the local authorities (Bezirksregierung Köln, 561 Germany) and conducted in accordance with NIH guidelines. Mice were housed in 562 groups of 3-5 animals at 22-24°C and a 12-hour light/dark cycle. Mice had ad libitum 563 access to food and water at all times, and food was only withdrawn if required for an 564 experiment. Mouse models were generated by crossing previously published mouse 565 strains, all of which have been backcrossed to C57BL/6 mice for at least ten   and hepatic glucose production (HGP) was calculated as previously described (Könner 651 et al., 2007). In vivo glucose uptake into perigonodal and brown adipose tissue, brain, 652 liver and skeletal muscle was calculated based on the accumulation of 2DG6P in the 653 respective tissue and the disappearance rate of 2DG from plasma as described 654 previously (Könner et al., 2007).   Hercules, CA, USA). Following, membranes were blocked in western blotting reagent 751 (Roche, Basel, Switzerland) diluted 1:10 in TBS-T buffer (Tris-buffered saline with 1% 752 (v/v) Tween 20) for 1-2 h at room temperature. Incubation with primary antibodies was 753 conducted over night at 4°C. Subsequently, membranes were washed three times for 754 5-7 min in TBS-T followed by the incubation in horse radish peroxidase-labeled 755 secondary antibody over night at 4°C. All primary and secondary antibodies were 756 diluted in 5% (v/v) western blotting reagent (Roche, Basel, Switzerland). After washing 757 for 3 times in TBS-T, membranes were incubated with SuperSignal ECL Western 758 Blotting Substrate (Thermo Fisher Scientific, Waltham, MA, USA) and luminescence 759 was detected by the Fusion Solo Vilber Lourmat system (Vilber Lourmat GmbH, 760 Eberhardzell, Germany). If necessary, membranes were stripped in stripping buffer 761 (6% SDS (v/v), 188 mM Tris pH 6.8, 2% (v/v) β-Mercaptoethanol) for 30 min at 56°C 762 in a shaking water bath. Afterwards, membranes were washed three times in TBS-T 763 followed by blocking in 10% (v/v) western blotting reagent (Roche, Basel, Switzerland) 764 for 1-2 h at room temperature. Then, the incubation with primary and secondary 765 antibodies was conducted as described above. Band intensities were quantified using 766 ImageJ (NIH). The intensity of protein bands was normalized to the intensity of calnexin 767 or β-Actin bands which served as loading control. Band intensities of protein bands 768 from phospho-targets such as pAKT and pHSL were additionally normalized to bands 769 of total protein, AKT and HSL, respectively. For detailed information of individual 770 primary and secondary antibodies see table 3. 771 772

RNA Isolation and RNASeq of Bulk Sorted NK Cells 815
NK cells from murine organs and blood were purified from single cell suspensions 816 using FACSAria IIIu or FACSAria Fusion cell sorters (BD Biosciences, San Jose, CA, 817 USA) after immunostainings as described above. NK cells were sorted from HFD-fed 818 animals identified as single/viable/CD45 + /CD3 -/NK1.1 + /Ncr1 + cells using a 70 µm 819 nozzle at 70 psi pressure. During sorting, sample and collection tubes were chilled. 820 Purified cells were directly sorted into FACS buffer (autoMACS Running Buffer, Miltenyi 821 Biotec, Bergisch Gladbach, Germany) and pelleted by centrifugation for 5 min at 822 400 g and 4°C. Total RNA was extracted using the Arcturus RNA picopure kit (Thermo 823 Fisher Scientific, Waltham, MA, USA) following the manufacturer's instructions. Due to 824 the low amount of input material, pre-amplification using the Ovation RNASeq System 825 V2 (NuGEN, Redwood, CA, USA), was performed as described previously (Paeger et 826 al., 2017). For library preparation, the Illumina Nextera XT DNA sample preparation 827 protocol (Illumina, San Diego, CA, USA) was used, with 1 ng cDNA input. After 828