Effects of obesity on NK cells in a mouse model of postmenopausal breast cancer

Obesity is a widely spread disease and a crucial risk factor for malign disorders, including breast cancer of women in the postmenopause. Studies demonstrated that in case of obesity crucial natural killer (NK) cell functions like combating tumor cells are affected. This study aims to analyze NK cells and NK cell receptor expression of obese mice in a model for postmenopausal breast cancer. Therefore, female BALB/c mice were fed either a high fat or a standard diet. Thereafter, ovaries were ectomized and a syngeneic and orthotopical injection of 4T1-luc2 mouse mammary tumor cells into the mammary adipose tissue pad was performed. Obese mice showed increased body weights and visceral fat mass as well as increased levels of leptin and IL-6 in plasma. Moreover, compared to the lean littermates, tumor growth was increased and the NKp46-expression on circulating NK cells was decreased. Furthermore, the activating NK cell receptor NKG2D ligand (MULT1) expression was enhanced in adipose tissue of obese tumor bearing mice. The present study gives novel insights into gene expression of NK cell receptors in obesity and aims to promote possible links of the obesity-impaired NK cell physiology and the elevated breast cancer risk in obese women.

like leptin obviously have an impact on the immune system 22 . Taking a closer look, the functions of natural killer (NK) cells, which are an essential part of the innate immune system, represent 10 to 15% of the peripheral lymphocytes. It is known, that they are impaired by leptin, which is secreted by the adipocytes and increases relative to the gain of body fat mass 23,24 . NK cells play an important role in the early defense by identifying and killing virally infected and tumor cells without prior sensitization and restriction by major histocompatibility (MHC) antigens 25 . The identification of targets by NK cells is regulated by the expression of activating and inhibitory receptors and in particular by binding of ligands expressed by the target cell, followed by the lysis of target cells 26 . It has been shown that obese individuals show a dysfunction of NK cells 23,24,[27][28][29] . Moreover, a prospective study demonstrated that impaired NK cell functions are associated with an increased cancer incidence 30 . Interestingly, obese patients who lost body weight and body fat mass by bariatric surgery or by exercise training and nutrition counseling could reverse their impaired NK cell activity and NK cell-mediated cytokine synthesis 31,32 . In addition, a recent study showed that lipid accumulation in NK cells from obese individuals is associated with the loss of NK cell cytotoxicity against tumor cells and could be restored by metabolic reprogramming 33 . This indicates that there might be a promising link between impaired NK cell functions in obese individuals and the increased risk for cancer under obesity.
The aim of the present study was to mimic postmenopausal breast cancer under diet-induced obese (DIO) conditions in a mouse model to investigate immune cell frequencies and associations between functional NKcell marker and tumor development in obesity.
Hence, to induce obesity female BALB/c mice received a high-fat diet and to achieve a postmenopausal status ovariectomy was conducted. To induce a mammary carcinoma, triple negative 4T1-luc2 cells, that do not express the estrogen receptor (ER), progesterone receptor (PgR) or the gene for human epidermal growth factor receptor 2 (HER2), were injected orthotopically in the mammary fat pad. This enabled us to investigate the isolated effect of extragonadal derived estrogen on NK cells, independently of the known proliferating effect of estrogens on breast cancer cells. To investigate initial NK cell profiles, we analyzed NK cell numbers, NK cell receptor expression and cytokine-levels 20 h after application of the tumor cells (short-term experiment). Therefore, effects of diet-induced obesity on early tumor development in a mouse model for postmenopausal breast cancer could be studied. Furthermore, in the long-term experiment, tumor growth was monitored by bioluminescence imaging and NK cell receptor natural killer group 2 member D (NKG2D) as well as the ligands UL16-binding protein-like transcript 1 (MULT1) and retinoic acid early inducible-1 gene (Rae-1) were determined in adipose tissue of mice.

Results
Body and organ weights. Mice fed the DIO diet gained significantly more weight compared to mice fed the control diet (Fig. 1a-c). In the long-term experiment significant weight gain of DIO-fed animals compared to control diet fed animals even continued after 4T1-luc2 tumor-cell-injection and during tumor growth (Fig. 1b,c). In contrast to control diet fed animals, mice of the DIO-fed group lost weight under tumor development, though maintaining a higher body weight compared to mice of control diet group. Appropriately, in both experiments DIO-fed animals gained significantly more visceral fat mass compared to animals fed the control Figure 1. Experimental setting (a,b) and anthropometric data (a-g) of mice in short-term and long-term experiment. Weight gain of mice in response to high fat diet (DIO) compared to control diet (Co) and sodium chloride (NaCl) or 4T1-luc2 (Tumor) over a period of 17 weeks (short-term experiment, (a) or 21 weeks (long-term experiment, (b). Representative pictures for weight gain of mice of the Control or DIO group (c) and visceral fat mass of mice in the short-term experiment (d,e). Spleen weight of mice in the long-term experiment (f) and representative pictures (g) of spleen of 4T1-luc2-tumor-bearing mice compared to sodium chloride-injected mice (NaCl). Values represent means ± SEM, n = 7 mice/group. Different superscript letters (a,b) indicate significant differences between individual experimental groups analyzed by Tukey´s multiple comparison test (p ≤ 0.05).  Table S2).
In the short-term experiment, mice of the DIO-fed group showed significantly increased spleen weights compared to control diet-fed animals (Supplementary Table S1). Tumor growth during the long-term experiment led to significantly higher spleen weights compared to animals receiving sodium chloride (NaCl) injection as a control (Fig. 1f,g; Supplementary Table S2). Liver weights of mice did not differ between groups of the short-term experiment (Supplementary Table S1). In the long-term experiment, mice of the DIO group and mice receiving tumor cells independent of the diet showed significantly increased liver weights compared to control diet-fed mice receiving NaCl injection (Supplementary Table S2).
Plasma levels of leptin, IL-6, IFN-γ and TNF-α in short-term experiment. Diet-induced obesity led to significantly increased leptin levels in the plasma of mice, compared to their corresponding control dietfed groups (Fig. 2a). Plasma levels of interleukin (IL)-6 were increased in DIO-fed mice compared to mice fed the control diet. DIO-fed mice, which received NaCl-injection showed the highest IL-6 levels in the plasma and differed significantly to mice fed the control diet (Fig. 2b). Interferon (IFN)-γ-and tumor necrosis factor (TNF)α-plasma levels did not differ between the four groups (Fig. 2c,d).
Monitoring of tumor development by bioluminescence imaging and caliper measurement. Monitoring of tumor growth by bioluminescence imaging in the long-term experiment revealed that all mice, who received 4T1Luc2 cells developed a tumor and that mice fed the high-fat diet showed a significantly increased tumor area compared to mice fed the control diet from week two after tumor challenge on ( Fig. 3a,b,d). Bioluminescence intensity and caliper measurement of the tumors were significantly higher in DIO-mice compared to control mice after three weeks of tumor growth (Fig. 3a-e). Nevertheless, tumor weight did not differ between control-and DIO-mice at the end of the long-term experiment (Fig. 3f).
Flow cytometry analysis of peripheral blood immune cells. In the short term experiment two-way ANOVA analysis demonstrated that the frequency of lymphocytes in groups fed the high-fat diet was significantly decreased compared to mice fed the control diet independently of the injection of tumor cells (Fig. 4a,c). and DIO diet fed mice (DIO) receiving either sodium chloride (NaCl) or 4T1-luc2 breast cancer cells (Tumor) in short term experiment. Values represent means ± SEM, n = 7 mice/group. * indicates significant differences of means between mice receiving DIO diet (DIO) compared to mice receiving control diet (Co) analyzed by two-way ANOVA (** p ≤ 0.01; *** p ≤ 0.001). Different superscript letters (a,b) indicate significant differences between individual experimental groups analyzed by Tukey´s multiple comparison test (p ≤ 0.05).

Scientific Reports
| (2020) 10:20606 | https://doi.org/10.1038/s41598-020-76906-5 www.nature.com/scientificreports/ In contrast, frequency of granulocytes increased in groups fed the high-fat diet compared to the control diet and significantly differed from control diet-fed animals without tumor cell challenge (Fig. 4b,c). While frequency of Ly6C high monocytes was not altered, frequency of Ly6C low monocytes was significantly decreased in the DIO groups ( Fig. 4d-f). Although the frequency of NK cells, defined by the expression of CD335 and the lack of CD3, did not differ, median fluorescence of CD335/natural cytotoxicity triggering receptor (NCR)1/NKp46 was significantly decreased in high-fat diet groups compared to control diet groups ( Fig. 4 g-i). Numbers of T cells, T cell subpopulations as well as B cells remained unchanged between the different groups ( Fig. 4j-m).
In the long-term experiment mice of tumor groups showed significantly decreased frequency in lymphocytes and Ly6C low , while granulocytes were increased significantly ( Fig. 5a-f). Moreover, tumor-bearing mice showed a depletion of Ly6C high monocytes in peripheral blood independent of the fed diet (Fig. 5e,f). Percentage of NK cells from lymphocytes were the highest in tumor challenged animals fed the high-fat diet and differed significantly to all other groups (5g, i). Additionally, mice receiving the control diet and the NaCl injection showed the significantly lowest NK cell frequency ( Fig. 5 g,i). Two-way ANOVA analysis revealed that the median fluorescence intensity of CD335/NKp46 on NK cells was significantly increased in tumor challenged groups independent of the diet (Fig. 5 h; p ≤ 0.0001). T cell frequency remained unchanged, but two-way ANOVA revealed that tumor growth significantly decreased the frequency of the subpopulation of T helper cells (CD4 + /CD8 − ; p ≤ 0.05) and increased the frequency of cytotoxic T cells (CD4 − /CD8 + ; p ≤ 0.05) most pronounced in the DIO/Tumor group ( Fig. 5j-l). Frequency of B cells was decreased in tumor challenged mice and differed significantly compared to mice receiving the control diet and NaCl injection (Fig. 5m).
Immunohistochemically staining and relative mRNA concentrations of NKG2D-receptor and -ligands in adipose tissues of mice. As expected DIO mice showed enlarged, adipocytes compared to control diet-fed groups as demonstrated by representative pictures of adipose tissue of mice in the long-term experiment (Fig. 6b,e,h). To investigate NK cells in adipose tissue, immunohistochemically staining and analysis of the relative mRNA concentration of NKG2D-receptor and the NKG2D-ligands MULT1 and Rae-1 were conducted. Mice fed the high-fat diet showed significantly higher expression of the NKG2D-ligand MULT1 in adipose tissue compared to animals fed the control diet independent of tumor growth ( Fig. 6d-f). Expression of NKG2D as well as the corresponding ligand Rae-1 in adipose tissue did not differ between the groups (Fig. 6a-c and g-i).

Expression of functional NK cell markers on isolated splenic NK cells.
Analysis of the relative mRNA expression of estrogen receptor (ESR) 1 in isolated splenic NK cells of the short-term experiment showed a significantly decreased expression in control diet-fed animals receiving tumor cell injection compared to NaClinjection. In DIO-fed animals, relative mRNA expression of ESR1 did not differ although they were as low as www.nature.com/scientificreports/ in mice of the Co/Tumor-group (Fig. 7a). NCR1/NKp46 expression in NK cells was significantly influenced by DIO-diet and tumor cell injection (Fig. 7d). Comparing single groups relative mRNA concentrations of NCR1/ NKp46 were significantly decreased in the DIO/NaCl and the DIO/Tumor group compared to animals receiving the control diet and a NaCl-injection (Fig. 7d). Expression of ESR2, G protein-coupled estrogen receptor (GPER) 1, programmed cell death protein 1 (PDCD1/PD-1), killer cell lectin-like receptor subfamily B member (Klrb) 1c/CD161 and Klrk1/NKG2D did not differ between the four experimental groups (Fig. 7b,c,e-g).

Discussion
Breast cancer is the most common cancer in women worldwide. Approximately 2.1 million diagnoses of breast cancer were estimated for 2018, which accounts for 11.6% of the total cancer incidence burden and results in the leading cause of cancer death in woman 34 . Several risk factors for example aging, nulliparity, family history, menopausal hormone therapy and genetic mutations, have been identified 35 . Nevertheless, also lifestyle factors like diet, nutrition and physical activity, seem to have a great impact on the incidence of breast cancer 36 .
In detail, there is a strong evidence that increased body fatness throughout adulthood and adult weight gain are two important risk factors for postmenopausal breast cancer, which is the most diagnosed breast cancer in woman 36 . Mechanisms underlying the risk factor obesity are continuously investigated and the immune system seems to play a central role in the interplay of obesity and the impaired cancer defense 37 . Obesity is associated with a state of chronic low grade inflammation which is represented by a an increased inflammatory profile and an impaired immune regulation 33,[38][39][40][41][42][43] . As NK cells are a central component of the innate immune system and are capable to directly destroy tumor cells 25,44 , the present study aimed to investigate the NK cell phenotype in www.nature.com/scientificreports/ a diet-induced obese postmenopausal breast cancer mouse model (short-term experiment) complemented by monitoring of tumor-growth (long-term experiment). As expected, diet-induced obese animals showed significantly higher body weights and visceral fat mass compared to control diet-fed animals. Although tumor development led to loss of body weight in DIO-fed animals and visceral fat mass in both diet groups, significant differences between the diet groups still preserved. Plasma concentrations of leptin and IL-6 were increased in DIO-fed mice after 17 weeks, representing the chronic low grade inflammation of the obese status 45 . This inflammatory condition was underlined by significantly increased frequencies of granulocytes in the mice of the DIO groups, which is in line with results from other studies comparing normal weight and obese subjects [46][47][48] . To mimic postmenopausal breast cancer, mice were ovariectomized and 4T1-luc2 cells were injected into the fat pad of the mammary gland. This orthotopic and syngeneic breast cancer model in BALB/c mice shares multiple features of the progressive forms of estrogen-insensitive human metastatic breast cancer 49 . Interestingly the final tumor weight did not differ between control and DIO-fed mice. This may be caused by higher appearance of necrosis observed in the tumors of obese mice, resulting in a different density of the solid tumors. Necrosis occurs in fast growing tumors, is a result of inadequate vascularization and is associated with a poor prognosis 50 . Bioluminescence and caliper-based monitoring of tumor growth revealed an increased tumor burden for DIO-fed mice. A study of Kim et al. reported similar results, although they used non-ovariectomized obese-resistant BALB/c mice 51 . Hence, it is still a subject of discussion, if the state of obesity per se is causal for the increased cancer incidence or if dietary components like fat are mediating the observed effects 52 .
Interestingly tumor burden of mice in the long-term experiment led to significantly higher spleen and liver weights independent of the diet, which is in accordance with other studies using the 4T1-luc2 cell line for the induction of breast cancer in mice and is based on the elevated hematopoiesis 49,53,54 . This was also seen in markedly increased frequency of granulocytes and decreased frequency of lymphocytes and monocytes in tumor bearing mice of the long-term experiment. In contrast to Trottier et al., who showed an enhanced hematopoiesis www.nature.com/scientificreports/ in HFD fed mice, we observed significantly decreased lymphocytes in the blood of obese mice compared to control group in the short-term experiment 55 . The chronic low grade inflammation in obesity is associated with an increased infiltration and migration of immune cells into adipose tissue 56 . As higher numbers of lymphocytes and macrophages are found in adipose tissue of obese individuals, it could be assumed that this is causal for the observed decreased frequency of circulating lymphocytes in the present study 57 . T cells express CD3 and can be further classified by their expression of CD4, representing T-helper cells and CD8, representing cytotoxic T cells. In the long-term experiment, a slight decrease in T-helper cells (CD3 + / CD4 + /CD8 − ) and a slight increase in cytotoxic T cells (CD3 + /CD4 − /CD8 + ) in peripheral blood was observed in the tumor mice compared to the normal weight mice without tumor cell injection. To our knowledge, until now no data of the frequency of T-cell subsets on DIO-fed tumor-bearing mice exist. One study in rats reported significantly decreased frequencies of CD4 + T cells and significantly increased frequencies of CD8 + T cells in DIO animals 58 . However, it is known, that in DIO mice and obese humans CD4 + and CD8 + T cells show an enhanced recruitment and infiltration into the adipose tissue [59][60][61] . Supposable, the increase of cytotoxic T cell frequency observed in the present study is due to a regulatory response on the increased tumor burden of DIO-fed mice.
High-fat diet was seen to impair B cell-lymphopoiesis of the bone marrow niche and tumor-infiltrating B cells are recognized as a new hallmark in breast cancer 62,63 . Interestingly, in the present study B cell frequency in the peripheral blood was downregulated in mice of the tumor group as well as by DIO and was the lowest in DIO/Tumor mice. Although not analyzed in the present study, this could be due to the described decreased B cell lymphopoiesies and an increased infiltration of B cells into tumor tissue.
Obesity is associated with increased circulating monocytes 64 , which is found to be linked mechanistically to leptin 65,66 . Mouse monocyte subsets are classified as Ly6C high and Ly6C low . Ly6C high monocytes differentiate into Ly6C low monocytes in the circulation. Under obese conditions, Ly6C high monocytes, who represent the inflammatory monocytes, migrate to the side of inflammation and differentiate into adipose tissue macrophages. Circulating Ly6C low monocytes represent the anti-inflammatory monocytes and are resident, patrolling in the blood vessels 67,68 . Interestingly, flow cytometry analyzes in the short-term experiment of the present study revealed a decrease of Ly6C low monocytes in the peripheral blood of DIO mice, while Ly6C high monocytes remained unchanged. However, this effect was diminished in tumor bearing mice as frequency of Ly6C low monocytes was decreased compared to NaCl-injected mice. Ly6C high monocytes even diminished, suggesting that this may be triggered by a higher infiltration of macrophages into the tumor tissue induced by monocyte chemoattractant proteins 69 .
For frequency of NK cells in the peripheral blood of obese individuals controversial data are reported 23,29,33,58,[70][71][72] . In this study NK cell frequencies in the blood of mice were not altered in the short-term experiment. However, in the long-term experiment, DIO led to significantly increased NK cell frequencies and was most pronounced in tumor-challenged DIO-fed mice, probably as a regulatory feedback mechanism. This was accompanied by higher median fluorescence intensity for the activating NK cell receptor CD335/NKp46. www.nature.com/scientificreports/ Concerning control diet-fed mice and DIO mice bearing the tumor, it shows that although NK cell frequency is higher in the DIO group, the median fluorescence intensity does not differ, suggesting decreased CD335/ NKp46-receptor expression per single cell in obese mice. Moreover, this seems to be independent of tumor burden, as this is also true for mice lacking the tumor. This is supported by the fact that relative mRNA expression of CD335/NKp46 in splenic NK cells is significantly decreased by DIO-feeding in the short-term experiment. The NK cell effector function is mediated by inhibitory and activating surface receptors as well as by adhesion and cytokine receptors 73 . They recognize cell-surface and extracellular-secreted ligands, e.g. major histocompatibility complex class I (MHC-I) and MHC-I-related molecules and non-MHC molecules 74 . MHC-I molecules (human leukocyte antigens (HLAs) in humans and H-2 in mice) are constitutively expressed by nucleated healthy cells and are identified by inhibitory NK cell receptors. The loss of MHC-I upon infection or malignant transformation leads to a predominance of activating signals and therefore to NK cell activation and target cell lysis 75 . Additionally, the expression of activating ligands is necessary for target cell killing and cytokine production by NK cells. Activating ligands are expressed by stressed cells upon infection or cellular transformation 73,76 . Thus, the balance and imbalance between activating and inhibitory signals dynamically regulate the effector function of the NK cell 25,77 . Interestingly, functional studies demonstrated that an increase in co-inhibitory receptor PD-1 expression on NK cells is associated with a decline in NK cell-mediated tumor defense and with a poorer prognosis in digestive cancers 78 . Another study reported higher levels of PD-1 expression upon stimulation in obese children compared to normal weight counterparts, accompanied by a reduced NK cell-mediated killing of target cells 70   www.nature.com/scientificreports/ expression for NKp46 and NKG2D 58,[70][71][72]79 . As already discussed, in the present study the expression of NKp46 on NK cells was significantly decreased in obese mice. Nevertheless, this was not true for NKG2D and Klrb1c representing additional activating NK cell receptors. Immunohistochemically staining of NKG2D receptor in adipose tissue also revealed no differences between control-fed or DIO-fed mice. As recently reviewed by O´Shea et al. NK cells are discussed to regulate adipose tissue homeostasis by killing of inflammatory macrophages after an NK-cell-mediated recruiting to adipose tissue 80 . Wensveen and colleagues hypothesized, that an upregulation of NKp46 ligands in obese adipose tissue leads to an activation of NK cells, which is followed by an IFNγ-induced differentiation of M2 macrophages into inflammatory M1-macrophages 81 . However, in contrast to the present study and results from Chung et al. 82 no upregulation of NKG2D ligands in obese adipose tissue was found. In the present study we cannot clearly proof whether the observed upregulation of NKG2D ligand MULT1 in obese adipose tissue is due to "stressed" fat cells or caused by inflammatory macrophages, which also express NKG2D ligands, therefore making them a target for NK cell lysis 83 . Nevertheless, it can be hypothesized that NK cells are collected by the inflammatory adipose tissue under obesity by upregulation of NKG2D ligand MULT1 and therefore not available for cancer cell killing in tumor tissue. This may have at least in part contributed to the observed increased tumor development in obese mice. Data about expression of estrogen receptors on NK cells are very limited, which is nicely reviewed by Kovats et al. 84,85 . In obesity, ESR on breast cancer cells is downregulated 86,87 . In the present study we chose 4T1-luc2 cells, which are ESR-negative, to induce breast cancer in mice. In postmenopausal obesity estrogen levels are known to be elevated due to higher aromatase activity in peripheral tissues, mainly in adipose tissue 88 . We therefore generated a model to study the effect of higher estrogen levels in a postmenopausal status on NK cells, independent from the direct estrogen-effect on the tumor. It is well established that estrogens decrease NK cell cytotoxicity 84,89-95 and high estrogen levels lead, as a feedback mechanism, to a downregulation of estrogen receptor expression in breast cancer cells 96,97 . Interestingly in our study, DIO-fed mice tended and tumor bearing control diet-fed mice had significantly decreased expression of ESR1 on NK cells, suggesting a direct regulation of NK cells via increased estrogen levels in postmenopausal breast cancer and obesity. Nevertheless, estrogens are also known to affect tumor growth by the influence of tumor microenvironment and angiogenesis of tumors 98,99 . Therefore, we can only speculate and future studies have to confirm these results, especially based on protein expression.
To conclude, results of the present study describe an altered expression of NK cell receptors and ligands and an increased tumor burden in postmenopausal obese mice. As obesity is a preventable risk factor, it is important to analyze mechanisms underlying the altered NK cell defense of tumor cells by future studies. Recent studies on postmenopausal breast cancer survivors investigating the impact of weight reduction by physical activity or dietetic interventions on the progression of breast cancer 100-103 highlight the importance of this research field.

Methods
Animals and experimental setup. Fifty-six female, ten weeks old, BALB/c mice were purchased from Charles River (Sulzfeld, Germany) and were housed in groups of two in standard type II polycarbonate individual ventilated cages (IVC). Cages were provided with embedding material (LINOCREL FS 14, Altromin, Lage, Germany) and a polycarbonate based mouse igloo (Zoonlab, Castrop-Rauxel, Germany). Mice were maintained at an ambient temperature of 22 ± 2 °C and relative humidity of 65 ± 5% on a 12:12 h light dark cycle with lights on from 6 am to 6 pm. Rodent standard diet (Altromin 1324) and water were available ad libitum. After one week of settling in and randomly assignment by body weight, twenty-eight mice received a high-fat diet (50% fat) ad libitum to initiate diet-induced obesity (DIO). Body weight of the mice was detected once per week by an appropriate balance (FTB-BA-d_0720, Kern, Balingen-Frommern, Germany). The federal authorities for animal research in Halle (Germany) approved the experimental protocol. The principles of laboratory animal care were followed according to the guidelines of the European (FELASA) and German Society of Laboratory Animal Sciences (GV-SOLAS). Mice were fed 13-14 weeks with the specific diet until ovariectomy was performed (for details see below). In the following three weeks, mice were allowed to recover from surgery. To induce a mammary carcinoma, half of the animals per diet group were injected 4T1-luc2 cells, which is described in detail below. Sodiumchlorid (NaCl) served as a control. Consequently, four experimental groups (n = 7) resulted: Co/ NaCl; Co/Tumor; DIO/NaCl and DIO/Tumor. To enable observation of a short tumor cell challenge (20 h) vs. a long tumor cell challenge (four weeks), two time points of scarification were chosen (Fig. 1a,b).
Ovariectomy. For anesthesia during ovariectomy a combination of ketamine (Ketavet 100 mg/mL, Zoetis Germany GmbH, Berlin, Germany) and medetomidine (Dorbene vet 1 mg/mL, Zoetis Germany GmbH) was used. Ketamine and medetomidine were suspended in physiological saline solution and injected i.p. in a final concentration of 100 mg/kg body weight for ketamine and 1.2 mg/kg body weight for medetomidine. To protect eyes of the mice against dehydration during the surgery, they were moisten with dexpanthenole (Bepanthen 10 g, Bayer AG, Leverkusen, Germany). To further support analgesia, ten minutes after beginning of anesthesia metamizol (Novaminsulfon-ratiopharm 1 g/2 mL, Ratiopharm AG, Ulm, Germany) was applied subcutaneously (lean mice: 10 mg; obese mice: 20 mg). Surgical intervention and awaking period was performed on a warming blanket to prevent mice from cooling. An approximately 0.5 cm long incision through the skin and two incisions through muscle and peritoneum bilaterally and parallel to the backbone were made while mice were placed in prone position. Ovaries were positioned outside the body via these incisions and removed by thermal cautery. Hereafter, incisions were closed using an sterile absorbable thread for in situ suture (V2130H, Ethicon, Norderstedt, Germany) and a sterile synthetic non-absorbable thread for skin suture (EH7147H; Ethicon). For pain management drinking water of the mice was supplemented with 1.6 mg/mL metamizole (Novaminsulfonratiopharm 500 mg/mL, Ratiopharm AG) 10 days and the drinking volume was controlled daily. www.nature.com/scientificreports/ Culture and application of tumor cells. Tumor induction in mice was achieved by 4T1-luc2 cell application (Perkin-Elmer; Berlin, Germany) cultured before in vitro as recommended by the manufacturers. This cell line is derived from BALB/cfC3H and expresses luciferase encoded by luc2 gene. It is an established animal model to mimic human breast cancer stage IV and allows bioluminescence monitoring of tumor growth 49 . After induction of DIO and ovariectomy (see above), animals of the tumor groups were injected with 10 6 4T1-luc2 cells dissolved in 0.1 ml cell culture medium (RPMI 1640; Gibco, Invitrogen, Carlsbad, California, USA) orthotopically in the subcutaneous fat pad of the fourth right mammary gland under light isoflurane anesthesia as described above. Control animals received 0.1 ml 0.9% NaCl in the same anatomical site.
Bioluminescence imaging and caliper measurement of tumor growth in the long-term experiment. In the long-term experiment, tumor volume was recorded once per week by caliper measurement using the following formula: length × width 2 × 0.52. Bioluminescence images were taken on day 7, 14 and 21 after injection of 4T1-luc2 tumor cells. Mice were anaesthetized by light isoflurane inhalation and images were taken exactly 30 min after i.p. injection of 150 µl luciferin (PerkinElmer, Massachusetts, USA) and by using VisiLuxII (Visitron Systems GmbH, Puchheim, Germany). Images were analyzed using the FIJII-software 104 .
Blood and organ sampling. Mice were sacrificed under general isoflurane anesthesia either after 20 h (short-term experiment) or four weeks (long-term experiment) by puncture of the right heart ventricle. Blood was withdrawn and spleen, visceral adipose tissue, liver and tumors were resected. Organs were weighed and immediately frozen in liquid nitrogen and stored at − 80 °C for RNA isolation and lipid analysis. Adipose tissue was formalin fixed and paraffin embedded. Blood samples were either used for plasma collection or stored on crushed ice for cytometry analyzes.
Immunhistochemical stainings of adipose tissue. Paraffin sections (5 μm) of visceral adipose tissue were prepared and stained with the mAbs anti-NKG2D (ab203353, Abcam, Cambridge, United Kingdom) and anti-UL16 Binding Protein 1 (ULBP1/MULT1, ABIN966609, antibodies online, Aachen, Germany) using the EnVision + System (horseradish peroxidase (HRP) labelled polymer anti-rabbit, Dako North America Inc., Carpinteria, USA). For Rae-1 staining mAB Rae-1 pan specific antibody (AF1136, R&D Systems, Minneapolis, USA) and anti-Goat HRP-DAB Cell & Tissue Staining Kit (CTS008, R&D Systems) were used. For both staining protocols, 3,3′-diaminobenzidine (DAB) was used as a chromogen for visualization of the respective proteins. Finally, sections were counterstained with hematoxylin and covered with Eukitt mounting medium (Sigma Aldrich, St. Louis, Missouri, USA). In the following, five tissue sections per mouse (n = 5) were analyzed with unknown group identity using a microscope (BZ9000 Fluorescence Microscope, Keyence, Neu-Isenburg, Germany) and FIJII-software with colour deconvolution plugin 104 . Per random offset function a grid of squares with the size of 12 × 10 8 µm 2 was created and eight individual fields were used to determine DAB-positive and hematoxylin-positive pixels, excluding the following criteria: a) blood vessel; b) artefacts, e.g. air bubbles, blur, spots; c) damaged tissue; d) neighboring fields. The relative expression of the respective proteins was calculated by the formula: pixel(DAB)/(pixel(DAB) + pixel (hematoxylin)), according to Shindhu et al. 105 . Hereby, the influence of obesity-related hypertrophy of adipose tissue could be taken into account and was equalized.
Flow cytometry analysis. For flow cytometry analysis, whole blood samples were stained following the almost identical protocol already described elsewhere 72  Realtime RT-PCR analysis. Splenic NK cells were directly isolated by magnetic activated cell sorting (MACS) using an NK cell isolation kit (Miltenyi, Bergisch Gladbach, Germany). Thereafter, mRNA was isolated by using Dynabeads oligo (dT) (Invitrogen, Thermo Fisher Scientific Inc.) following manufacturer´s instructions. Synthesis of cDNA was performed by reverse transcriptase reaction according to the supplier's instruction as already described (Thermo Fisher Scientific Inc.) as already described 72  www.nature.com/scientificreports/ dilutions 106 . The mRNA concentrations of genes were measured by realtime detection reverse transcriptase-PCR (iQ5, BioRad) using SYBR Green MIX (BioRad). For determination of mRNA concentration, a threshold cycle (Ct) was obtained from each amplification curve using the software qPCRsoft 3.4 (Analytik Jena AG).
Statistical analysis. Data analysis was performed using Graph Pad Prism software V7 (GraphPad Inc., La Jolla, CA, USA). Analyzes were performed as two-way ANOVA for the two main factors "diet" and "tumor" followed by Tukey´s multiple comparison test, for individual comparison of all groups. Results are presented as means ± standard error of the mean (SEM). Means were considered as significantly different at p ≤ 0.05. * indicates significant differences between mice receiving the DIO diet compared to control diet. Different superscript letters (a,b,c) indicate significant differences between individual experimental groups.

Data availability
All data generated or analyzed during this study are included in this published article and its additional files. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.