Notch signaling in adipose tissue macrophages prevents diet‐induced inflammation and metabolic dysregulation

The importance of macrophages in adipose tissue (AT) homeostasis and inflammation is well established. However, the potential cues that regulate their function remain incompletely understood. To bridge this important gap, we sought to characterize novel pathways involved using a mouse model of diet‐induced obesity. By performing transcriptomics analysis of AT macrophages (ATMs), we found that late‐stage ATMs from high‐fat diet mice presented with perturbed Notch signaling accompanied by robust proinflammatory and metabolic changes. To explore the hypothesis that the deregulated Notch pathway contributes to the development of AT inflammation and diet‐induced obesity, we employed a genetic approach to abrogate myeloid Notch1 and Notch2 receptors. Our results revealed that the combined loss of Notch1 and Notch2 worsened obesity‐related metabolic dysregulation. Body and AT weight gain was higher, blood glucose levels increased and metabolic parameters were substantially worsened in deficient mice fed high‐fat diet. Moreover, serum insulin and leptin were elevated as were triglycerides. Molecular analysis of ATMs showed that deletion of Notch receptors escalated inflammation through the induction of an M1‐like pro‐inflammatory phenotype. Our findings thus support a protective role of myeloid Notch signaling in adipose tissue inflammation and metabolic dysregulation.


Introduction
Obesity is a metabolic disease characterized by abnormal and excessive accumulation of body fat and constitutes a major health Correspondence: Prof.Evangelos Andreakos e-mail: vandreakos@bioacademy.grhazard.Apart from a reduced life expectancy, obesity is associated with an increased risk for developing a wide variety of diseases including type 2 diabetes mellitus (T2DM) and cardiovascular diseases [1].Central to obesity and the associated imbalance between caloric intake and energy expenditure is a chronic lowgrade inflammation of the adipose tissue (AT) that contributes to insulin resistance (IR) and metabolic complications [2].Under obese conditions, AT undergoes a series of dynamic remodeling events, including adipocyte hypertrophy, apoptosis, infiltration of immune cells, and overproduction of inflammatory cytokines as well as extensive vascularization and ECM remodeling [3].Major drivers of chronic AT inflammation both in mice and in humans are adipose tissue macrophages (ATMs) [4] and although the sequence of events has spurred controversy, the growing consensus supports that activation of ATMs precedes the development of IR and contributes to a proinflammatory state [4,5].
Macrophages are, almost in every tissue, tightly entwined with homeostasis, immune surveillance, clearance of cellular debris as well as promotion or resolution of inflammation.The prevailing dogma for many years delineates a binary division between proinflammatory (M1) and anti-inflammatory (M2) macrophages [6,7].M1 macrophages demonstrate high glycolytic and low oxidative phosphorylation (OXPHOS) activity and express proinflammatory cytokines, whereas M2 macrophages have the opposite metabolic activities and express increased levels of antiinflammatory cytokines [8].However, it is now becoming increasingly clear that determining the phenotype of macrophages is a multifactorial process, and within a tissue microenvironment, macrophages possess a wide spectrum of activation states, that are influenced by a plethora of growth factors, cytokines, chemokines, adipokines, and hormones [9].ATMs are the most abundant immune cells in AT both under homeostatic conditions, and also in the obese state, in which their paramount importance is underlined by their prominent increase in numbers and quantity [10].Yet, ATMs are highly heterogeneous sharing markers and activities, but also exhibit significant functional differences and metabolic adaptations, the details of which remain poorly understood [10].
Notch signaling is a highly conserved and important intercellular pathway that participates in a wide range of developmental processes and controls both innate and adaptive immune cell homeostasis and function [11,12].In mammals, there are four Notch receptors (Notch1-4) and five ligands (Jagged: Jag1, Jag2 and Delta-like: DLL1, DLL3, DLL4).Ligation of Notch receptors by their ligands induces the sequential enzymatic cleavage of Notch receptors by disintegrin and metalloprotease (ADAM) family proteases and the intracellular gamma-secretase, resulting in the release of the intracellular activated domain of Notch (NICD1-4).NICD then translocates into the nucleus, inducing the transcription of Notch target genes [11].
Accumulating evidence points to a major involvement of the Notch pathway in hepatic insulin resistance, liver steatosis, and atherosclerosis [13][14][15].Only a handful of studies, however, have focused on obesity and AT-related inflammation, and these from the adipocyte point of view.Nevertheless, the role of Notch in adipocyte differentiation and homeostasis remains controversial since inconsistent findings have been reported [16,17].
Emerging lines of investigation have implicated Notch signaling in macrophage function.Evidence from in vitro experiments support enhanced M1 gene expression and proinflammatory response of macrophages upon Notch activation [18].
Notably, several studies have shown that in macrophages, Notch can also be activated by LPS-mediated TLR4 stimulation [19,20].Particularly, induction of Notch signaling through TLR has been shown to be important for the production of proinflammatory cytokines, including TNF, IL-6, IL-10, and IL-12, in TLRactivated macrophages [21,22].Notch signaling has thus risen as an important determinant of the in vitro polarization and activation of proinflammatory macrophages to M1 versus M2 regulation [23,24].Despite all that, the crosstalk between the Notch cascade and ATMs in modulating obesity-associated inflammation and metabolic disease has not yet been addressed.Here, we explore the role of Notch signaling in the pathogenesis of obesity in experimental animal models and reveal a novel potential interplay between Notch and ATMs that orchestrates obesityassociated metabolic dysfunction and metainflammation.

ATMs accumulate during diet-induced obesity and upregulate CD11c
Macrophages represent the predominant immune cell type that resides in the epididymal white adipose tissue (epiWAT) and exhibit distinct phenotypic and functional characteristics that are not fully understood [4].To address this, we employed an established mouse model of diet-induced obesity (DIO) and insulin resistance [25].Wild-type mice were fed a high-fat diet (HFD, D12492i; 20 kcal% protein, 60 kcal% fat, and 20 kcal% carbohydrate) or a normal chow diet (NCD, D12450Bi; 20 kcal% protein, 10 kcal% fat, and 70 kcal% carbohydrates) as control, and their body weight was measured weekly over a 16-wk observation period.ATMs were analyzed at an early stage, before the establishment of obesity and insulin resistance (4 weeks of diet) and at a later time point after the establishment of the metabolic phenotype (16 weeks of diet; Fig. 1A).Mice fed an HFD increased their weight as early as 3 to 4 weeks after the hypercaloric diet initiation and doubled their body mass compared with NCD-fed mice after 16 weeks (Fig. 1B).According to this model, mice also exhibited increased epiWAT mass and impaired glucose and insulin tolerance (Fig. S1A-E).In order to characterize mononuclear phagocytes of epiWAT and further discriminate macrophages within the myeloid cell pool as well as track other immune populations such as lymphocytes during homeostasis (NCD) or obesity-induced inflammation (HFD), we developed a multicolor flow cytometric analysis methodology based on a wide range of markers (Fig. S2A).Through distinct populations of epiWAT immune cells, we identified ATMs as being positive for CD45 + F4/80 + CD11b + MHCII + CD64 + markers while possessing their typical large, round shape with a centrally located nucleus and abundant cytoplasm with many vacuoles or granules characteristic of "foam macrophages" (Fig. S2A).We were also able to distinguish monocytes, different subsets of dendritic cells, eosinophils, and neutrophils along with lymphocytes.Notably, we observed that in the lean state, ATMs expressed high levels of CD301, a c-type lectin characteristic of M2-like macrophages.Analysis of the inflamed obese epiWAT, though, showed a major infiltration of macrophages in the tissue (Fig. 1C) that was accompanied by an elevation of the pro-inflammatory marker CD11c, which has been regularly used to describe phenotypically M1like macrophages.Specifically, 16-week HFD ATMs presented an eightfold increase of the CD11c marker, while they downregulated CD301 (Fig. 1D and E, Fig. S1F).

ATMs shift to a proinflammatory phenotype and dysregulate notch signaling during HFD-induced obesity and insulin resistance
To unravel the complexity of ΑΤΜs and their relation to inflammation and metabolic perturbation, we generated comprehensive RNA-seq gene expression data from sorted CD45 + F4/80 + CD11b + MHCII + CD64 + ATMs of epiWAT, isolated from WT mice fed either NCD or HFD for 4 and 16 weeks (Fig. 2A).Principal component analysis (PCA) revealed that ATMs from NCD epiWAT clustered together as did those of HFD epiWAT, and separated from each other, for both timepoints (Fig. 2B, Fig. S3A and B), indicating distinct transcriptional profiles of ATMs in homeostasis and obesity-induced inflammation.In addition, PCA showed that transcriptomes of ATMs from HFD epiWAT distinguished between the 4-and 16-week timepoints (Fig. S3B), suggesting that prolonged exposure to HFD induces major alterations to ATMs.Comparison of NCD and HFD ATMs at 4and 16 weeks revealed 322 and 5525 differentially expressed genes (DEG), respectively (Fig. 2C and D).Two hundred and twenty-three of these differentially expressed genes (DEGs) were common between 4 and 16 weeks timepoint while the rest were unique to one or the other timepoint (Fig. 2C).Further analysis demonstrated that HFD ATMs between 4-and 16 weeks of diet differentially expressed 5891 genes, 4280 of which were shared with DEGs of 16 weeks ATMs between NCD and HFD (Fig. 2C, Fig. S3C-E).Consistent with the flow cytometry data, transcriptomic data further validated that CD11c + ATMs, a subpopulation of ATMs poorly represented in animals fed a normal diet, surged under prolonged HFD feeding (Fig. 2E).Meanwhile, CD301 was markedly downregulated by HFD consumption (Fig. 2E).More importantly, among the most significantly regulated genes were many involved in inflammation, such as cytokines and chemokines (Fig. 2F).Particularly, heatmaps showed a tremendous increase in expression in a panel of cytokines such as l1a, Il1b, Il6, Il12a, Il12b, Il17c, and Il33 and multiple CCL and CXCL chemokines like Ccl2, Ccl3, Ccl4, Ccl5, Cxcl1, Cxcl2, Cxcl3, and Cxcl10 in 16-week HFD ATMs (Fig. 2F).Accordingly, the transcripts of genes encoding pattern recognition receptors and transcription factors were also greatly affected after 16 weeks of diet (Fig. S3F).Strikingly, this high proinflammatory phenotype of obese ATMs was accompanied by an altered metabolic status with various genes related to metabolic processes, cell cycle, oxidative phosphorylation, and hypoxia being severely impaired (Fig. S3G).Notably, 16-week HFD ATMs overexpressed glycolysis genes such as Hk2, Glut1, Hif1a, and Pkm while they showed reduced expression of genes related to insulin signaling such as Irs1, Irs2, Igf1, Igf1r (Fig. S3G).However, the most prominent finding was that Notch signaling emerged as a strong discriminator of NCD and HFD ATMs (Fig. 2G).Investigation of this gene set revealed a 10-fold increase in the expression levels of Jag1 ligand and a significant decrease of Notch2 receptor in the 16-week HFD ATMs (Fig. 2G).Concurrently, Rbpj, a Notchspecific transcription factor, and Hes1, one of the main target genes of the pathway, showed reduced expression, whereas the metalloprotease-disintegrin ADAM8 was highly elevated in the 16-week HFD ATMs.These results were also confirmed by qPCR analysis (Fig. S4A).In addition, we validated the high expression of Jag1 protein in the 16-wks HFD ATMs by flow cytometry (Fig. S4B-D).We also examined isolated adipocytes from wild-type mice subjected to NCD or HFD regarding their Notch molecular pattern and found significant changes in the expression of Jag1, Notch1, and Hes1 genes, indicating a complex association of Notch signaling not only within macrophages but also among the adipocyte fraction of the epididymal adipose tissue (Fig. S4E).Overall, our analyses revealed that obesity and insulin resistance altered macrophage function and led to a perturbed Notch cascade.

Double Notch1 and Notch2 deletion from macrophages aggravates diet-induced metabolic dysregulation in mice
To dissect the contribution of the Notch pathway in HFD-induced obesity and the altered macrophage transcriptome observed, we generated myeloid cell-specific knockout mice.For this purpose, we employed the LysM-Cre transgenic mice which express Crerecombinase under the control of the lysozyme 2 gene (Lyz2; lysM)) promoter and have been extensively used to achieve genetic manipulation in the myeloid compartment [26].To examine further the cellular specificity of Cre recombinase deletion within adipose tissue we generated a RosatdTomato/Cre reporter mouse model by crossing Rosa tdTomato fl/fl mice with the LysM-Cre mice (Fig. S5A).In this model, tdTomato-positive cells indicate Cre expression.Using flow cytometry, we observed that the majority of ATMs were tomato-positive and could therefore express the Cre recombinase (Fig. S5B and C).In contrast, only a small number of neutrophils in the adipose tissue were tomatopositive cells and thus expressed the Lyz2 gene (Fig. S5B and C).Taking into account that adipose tissue has very few neutrophils compared with the broad ATM population, and that even fewer of them express the Cre recombinase, we next focused on ATMs.In order to elucidate the role of Jag1 ligand, we first employed myeloid cell-specific Jag1 knockout mice (Jag1 fl/fl LysMCre +/− herein Jag1 fl CRE) generated by crossing Jag1 fl/fl mice with LysM-Cre mice (Fig. S6A).Jag1 fl CRE mice that were rendered obese by placement to a 12-week-HFD consumption (Fig. S6B) gained slightly significantly more weight than their control counterparts (Jag1 fl ) on the same diet (Fig. S6C and D).There was also a trend for higher basal glucose levels (Fig. S6E) and ATMs (Fig. S6L) but generally Jag1 fl CRE displayed normal glucose and insulin tolerance (Fig. S6F and G) along with normal metabolic parameters (Fig. S6H-J) and serum triglycerides (Fig. S6K), indicating that deletion of myeloid Jag1, despite its alarming increase in HFD ATMs, does not alter significantly the DIO phenotype in mice.On NCD, in which Jag1 expression in epiWAT macrophages is low, body weight did not display differences between the two genotypes as did not any of the other parameters assessed (data not shown), suggesting that Jag1 is not rate-limiting for epi-WAT homeostasis in the steady state.In either case, genetic ablation of myeloid Jag1 had no significant effect, raising the question of whether other Notch ligands may compensate for its absence in Jag1 deficient mice.Thus, in order to get clues into the involvement of the Notch receptors, we first sought to analyze the myeloid-specific Notch1 deficient genotype under the influence of HFD.However, when we investigated the myeloid-specific Notch1 fl CRE mice under the influence of HFD we noticed no changes (data not shown) concluding that sole Notch1 deficiency does not affect obesity or visceral adipose inflammation.Given the highly expressed transcripts of the Notch2 receptor in WT ATMs and that Notch1 and Notch2 often complement each other's action [27,28], we developed a two-gene-based transgenic mouse model, Notch1 fl/fl Notch2 fl/fl LysMcre +/− (herein N1N2 fl CRE, CRE in the figures; Fig. S7A) for the specific and concurrent depletion of the two Notch receptors from the myeloid cells and the consequent impairment of the Notch pathway, as demonstrated by the reduced read counts of Notch1 and Notch2 transcripts in RNA-Seq data obtained from ATMs of CRE vs WT animals (Fig. S7B), and the increased expression of the Notch ligand Jag1 and its target Hey1 (Fig. S7C).Based on the already established HFD protocol (Fig. 3A), we showed that HFD expedited the appearance of increased body weight in N1N2 fl CRE compared with their control counterparts N1N2 fl (WT) mice (Fig. 3B).Notably, N1N2 fl CRE mice gained 15%-20% more weight than the control mice at each timepoint assessed (Fig. 3B).This was also accompanied by an increase in their epididymal fat mass (Fig. 3C).In agreement with their phenotypic changes, N1N2 fl CRE mice developed higher basal glucose levels, higher insulin resistance, and elevated serum triglyceride levels (Fig. 3D-F) while there was also a tendency for higher cholesterol and free fatty acids serum levels (Fig. S7D and  E).These were also associated with overly high insulin and leptin levels in the blood of N1N2 fl CRE mice levels (Fig. 3G and H), indicating that knocking out myeloid Notch1 and Notch2 increases the predisposition to DIO.Consistent with the exacerbated obese phenotype, N1N2 fl CRE mice on HFD exhibited decreased O 2 consumption (VO 2 ) and CO 2 production (VCO 2 ) (Fig. S7F and G) as well as reduced metabolic rate and specifically oxidation rate of lipids while there was no difference in the oxidation rate of carbohydrates (Fig. 3I-L) and in the consumption of food (Fig. S7H).These observations suggest that an underlying perturbation in the regulation of metabolism does exist in the absence of both Notch1 and Notch2 receptors from ATMs in the DIO mode.On the other hand, Notch inactivation in the NCD background did not cause any obvious abnormalities (Fig. S8).Since dysregulation of Notch signaling befell exclusively in the WT ATMs of the 16 weeks HFD, it is possible that the Notch pathway has a selective role only in the development of obesity and insulin resistance after HFD and not in the regulation of homeostasis under NCD.

Double Notch1 and Notch2 deletion from macrophages deteriorates adipose tissue inflammation and ensuing metabolic dysregulation
To corroborate these data, we next sought to determine the contribution of deficient Notch1 and Notch2 macrophages in adipose tissue inflammation.Using flow cytometry, we assessed the relative numbers of immune cells and found that ATMs were significantly upregulated during HFD in epiWAT of N1N2 fl CRE mice (Fig. 4A and B), accounting probably for the severity of our model.More importantly, we highlighted the significant increase of CD11c + proinflammatory ATMs in the N1N2 fl CRE mice compared with the control mice.We also observed a concomitant increase in neutrophils but this was not the case for the rest of the adipose tissue immune cells that were unaltered (Fig. 4B  and C).To further characterize the N1N2 fl CRE ATMs, we sorted them to high purity and performed gene expression analysis.Remarkably, proinflammatory gene expression was elevated in HFD-fed N1N2 fl CRE ATMs compared with the controls, as Nos2, Arg1, and IFNγ were significantly upregulated, while Il10 was downregulated (Fig. 4D).These results outline once again a role for Notch1 and Notch2 in the skewed inflammatory phenotype of ATMs.To gain insight into the molecular basis of the exacerbated metabolic phenotype, we also examined genes mediating insulin resistance and glucose handling.Our results showed that Irs1 and Irs2 were diminished in sorted ATMs of N1N2 fl CRE mice (Fig. 4E) and that was in accordance with the reduction seen in the transcripts of 16 weeks HFD ATMs from WT mice (Fig. S3G) uncovering a potentially novel interplay between the impairment of Notch signaling and the induction of metabolic perturbations.In addition, we also found higher levels in the expression of Igf1r and lower levels of Hk2, further contributing to the altered metabolic profile of ATMs (Fig. 4E).To evaluate the complex network of cellular mechanisms within the epididymal adipose tissue and specifically the crosstalk between macrophages and adipocytes, we also performed a molecular analysis on isolated adipocytes from N1N2 fl CRE mice subjected to HFD.We therefore found some significant changes in the expression of various metabolic genes such as Igf1r, Furin, and Hk2 (Fig. 4F).These results suggest that deficient Notch1 and Notch2 ATMs are important drivers for transcriptional alterations in the adipocytes of their microenvironment.

exhibit transcriptome changes during DIO
To dissect further the myeloid cell landscape of deficient Notch1 and Notch2 mice subjected to HFD, we employed a comprehensive phenotyping profiling of epiWAT based on a 19-color flow cytometry analysis panel.F4/80 + ATMs, identified as being CD45 + F4/80 +/hi CD11b + MHCII + CD3 − Ly6G − represented the largest fraction of CD45 + cells (Fig. 5A, Fig. S9A) and could be further separated on the basis of CD11c expression according to the genotype, as already described (Fig. 4A, Fig. S9A and B).F4/80 + ATMs were further distinguished into CD64 + and MertK + cells and then subdivided into LYVE1 + CD16.2 + and LYVE1 − CD16.2 + populations (Fig. 5A, Fig. S9A).These are distinct populations that may largely represent embryonic and monocyte-derived macrophages, respectively, as LYVE1 has been proposed to be a marker of early-life macrophages involved in the developing embryonic lymphatic vasculature, and LYVE1 + CD16.2 + ATMs also display high expression of Trem2 and CD9, additional markers associated with an embryonic origin of the cells (Fig. 5A) [29,30].Interestingly, in both genetic backgrounds, F4/80 + ATMs analyzed as F4/80 +/hi , CD64 + MertK + F4/80 +/hi , LYVE1 + CD16.2 + F4/80 +/hi , and LYVE1 − CD16.2 + F4/80 +/hi cells did not exhibit any major differences in their ratio in the ATM pool (Fig. 5A) nor an array of cell surface markers assessed (Fig. S9A).At the same time, LYVE1 − CD16.2 + F4/80 +/hi cells expressed uniformly CD9 in accordance with the literature but only 40% of them also expressed Trem2.Similarly, in both genetic conditions, LYVE1 + CD16.2 + F4/80 +/hi ATM subsets showed a comparable phenotype, lacking Ly6C but expressing high levels of CCR2 and CD9 (Fig S9A and B).In contrast, CD11c expression was more prominent in the monocytic subset of F4/80 + ATMs of both N1N2 fl and N1N2 fl CRE mice fed HFD (Fig. S9B).Strikingly, although ATMs and specifically  CD11c + ATMs increased significantly in numbers and percentage in N1N2 fl CRE, the ratio of the two different origin ATM subsets did not change (Fig. 5A).There was also a population of monocyte-derived macrophages (CD64 − MertK − F4/80 +/lo ATMs) minimally represented and defined as being F4/80 +/lo MHCII + CD11b + CD11c + Ly6C − CD64 − MertK − CD16.2 + lacking Trem2 and CCR2 but expressing high levels of CD9 (Fig S9A and B).Ly6C + Monocytes (F4/80 − MHCII − Ly6C + CD11b + ), neutrophils, and dendritic cells (DCs) were present in lower numbers in the epiWAT of both mice (Fig S9A).As F4/80 + ATMs were the dominant leukocyte fraction in the epiWAT encompassing both LYVE1 + CD16.2 + and LYVE1 − CD16.2 + that did not change in ratio in N1N2 fl and N1N2 fl CRE mice, we decided to investigate them further as a whole.Therefore, we sorted to high purity F4/80 + ATMs from N1N2 fl and N1N2 fl CRE mice fed NCD and HFD for 12 weeks and performed RNA-sequencing.The analysis revealed that Notch1 −/− Notch2 −/− macrophages have a distinct transcriptional profile compared with wild-type control macrophages as depicted in the PCA plot (Fig. 5B).DEG analysis revealed that 316 genes were significantly altered by Notch1 and Notch2 deletion; 295 were upregulated and 91 were downregulated based on adjusted P values <0.05 (Fig. 5C).Gene Ontology analysis of the DEGs revealed ECM remodeling including genes such as Col1A1, Col4A2 Itga6, Lama2, and Lama4, and immune response-related pathways among the most affected biological processes (Fig. 5D and  E).These findings are in line with the fact that ECM remodeling in adipose tissue and inflammation are tightly interlinked biological events during obesity and metabolic diseases [31].Indeed, histology studies revealed increased collagen deposition in the adipose tissue from N1N2 fl CRE compared with control mice, further supporting the presence of increased fibrosis in Notch1 and Notch2 depleted animals (Fig S10A and B).Gene Ontology enrichment analysis also included major signaling pathways such as BMP, Ras protein, and TGF-β signaling as well as regulation of phospholipase C and small GTPase, cellular response to growth factor stimulus and ERK1 and ERK2 cascade (Fig. 5D).More importantly, significant pathways that were over-represented in N1N2 fl CRE mice were related to immune cell proliferation and differentiation and more specifically to T-cell differentiation and activation and also to leukocyte proliferation.Lipid kinase activity and hypoxia, both associated with metabolic process and tissue remodeling but also inflammation, were overrepresented in N1N2 fl CRE.Of interest, cytokine production and antigen processing and presentation were among the enriched pathways, further strengthening the idea of an exacerbated inflammatory response in the adipose tissue of N1N2 fl CRE.Kyoto Encyclopedia of Genes and Genomes pathway analysis revealed terms associated with ECM receptor interaction, protein digestion and absorption, focal adhesion, and cell adhesion molecules (Fig. 5E) suggesting that signals from the microenvironment are important drivers for transcriptional alterations in macrophages.Of paramount importance were also pathways related to lymphocyte activation like Th1, Th2, and Th17 cell differentiation.Moreover, signaling networks involving PI3K-Akt, TGF-β, and Jak-STAT were also affected and emerged as significant in the regulation of Notch1 and Notch2 deficient macrophages (Fig. 5E).Collectively, these data support a critical functional role of the Notch pathway in the proinflammatory cascade and metabolic switch of macrophages within adipose tissue following obesity-induced inflammation.

Discussion
In this manuscript, we characterize the ATM phenotype and transcriptional profile during the development of DIO and insulin resistance and explore the role of the Notch pathway in the regulation of macrophage activation.We focus on Notch signaling and its contribution to the initiation and progression of DIO-related inflammation, highlighting putative macrophage-mediated mechanisms that promote M1-like gene transcription and metabolic dysregulation.We elaborate on metabolic parameters and discuss the potential of myeloid Notch signaling as a therapeutic target in obesity-induced inflammation.
The importance of the Notch pathway in obesity and insulin resistance has been highlighted in various studies focusing on the liver.DIO in mice triggers hepatic Notch1 signaling, which through the activation of Rbpj increases mTorc1 complex stability to strongly promote fat expansion.Inversely, acute or chronic liver-specific inhibition of Notch dampens hepatic glucose production by blocking mTorc1 and therefore preventing hepatic steatosis [32,33].In addition, the pathophysiology of nonalcoholic fatty liver disease (NAFLD) has been associated with Notch pathway dysregulation.During the development of NAFLD in patients, Notch family members have been shown to exhibit altered expression patterns [34].Further studies in a cohort of women with NAFLD found a negative relationship between hepatic expression of HEY2, a Notch transcriptional repressor, and low-density lipoprotein cholesterol [35].Recent lines of investigation also shed light on the role of DLL4-Notch signaling on the pancreatic islets and showed that inhibition of this particular axis rather than complete Notch signaling rescues insulin-producing beta cells and reduces immune cell infiltration in diabetic mice [36].
Regarding adipose tissue, the relationship between Notch and adipose development and function as well as energy metabolism is still ambiguous.Existing in vitro studies support the notion that Notch signaling modulates the proliferation and differentiation of adipocyte progenitor cells [37,38].Limited research on the in vivo role of Notch within adipose tissue suggests that inhibition of Notch1 signaling in mice improves glucose tolerance and insulin sensitivity and increases the expression of brown adipose tissue-specific genes [16,39].In contrast, activation of Notch1 in adipocytes inhibits browning and transcription of Ppargc1a and Prdm16 [16].In addition, a couple of studies showed that adipocyte-specific activation of Notch signaling promotes adipocyte dedifferentiation that results in lipodystrophic phenotype and metabolic dysfunction [40,41].In this scenario, activation of Notch signaling decreased also the expression of various genes related to lipogenesis and adipogenesis [40].However, at present the impact of Notch signaling on lipid accumulation in adipose tissue has remained unclear.Thus, our study provides a mechanistic link between Notch signaling and the development of DIO syndrome through ATMs, key players orchestrating metabolic inflammation.
ATMs have been implicated in the development of insulin resistance and metabolic dysfunction associated with obesity.Yet, macrophages are highly divergent and strongly shaped by their respective tissues, possessing distinct gene expression programs characteristic of tissue residency or monocyte-derived emergence.Several studies have tried to demystify the origin of ATMs under homeostatic conditions or in disease contexts.Of note, the subset of CD9 + CD63 + Trem2 + ATMs has been found to expand significantly during obesity and originate from recruited circulating monocytes [29].
In this article, we provide a broad characterization of the different subsets of ATMs and then delve into the mechanisms by which Notch signaling can influence ATM function since recent studies have shown that the Notch pathway plays an essential role in regulating their activation and polarization, key factors that can affect the development of obesity-induced inflammation.We demonstrate that prolonged exposure to HFD induces major alterations to ATMs resulting in a pro-inflammatory and metabolic phenotype.Strikingly, this phenotype is accompanied by altered Notch signaling patterns underscoring the importance of examining myeloid Notch pathway dynamics in adipose tissue.Hence, using a genetic approach to inhibit both myeloid Notch1 and Notch2, we show that DIO results in significantly more severe effects.Specifically, when both Notch1 and Notch2 are deleted from ATMs, body weight gain and fat pad weight are increased.This is most likely related to hyperplasia rather than hypertrophy of the adipose tissue as N1N2 fl CRE mice have a tendency for higher numbers of adipocytes in their epidydimal fat (Fig. S10C-E).N1N2 fl CRE mice also exhibit raised blood glucose levels rise, and exacerbated metabolic parameters when fed an HFD.In addition, insulin and leptin levels are elevated in these mice as are triglycerides in the serum.Sole Notch1 ablation, however, failed to elicit any obvious deficiencies.This may be due to the compensatory effect of other Notch family members.Indeed, Notch2, as we show, is also expressed in ATMs, as well as Notch 3 and 4, but to a lesser extent, suggesting, though, potential redundant effects of Notch family members.Alternatively, myeloid-specific deletion of Notch1 could be surmounted by pleiotropic effects of Notch1 signaling originating from other cells of the adipose tissue or even other tissues.Another notable observation that might seem conflicting at a glance is that Jag1KO mice have no differential susceptibility to obesity despite Jag1's impressive increase in HFD ATMs.Nevertheless, Notch ligands and receptors may display rate or time-limited expression to achieve distinct activation patterns or engage different target genes between various macrophage profiles.The complexity is encoded by the diverse signaling capacity of Notch ligands and receptors, and the preferential binding of ligands to different Notch receptors.Conversely, this may be attributed to other cells within fat tissue that employ Notch signaling and can cause cell-specific effects on ATMs.
At the core of the adipose tissue inflammatory response lies the presence of macrophages.In our model, we observed an increase in the number and percentage of proinflammatory macrophages that infiltrate the adipose tissue, escalating the inflammatory milieu.This included LYVE1 + CD16.2 + and LYVE1 − CD16.2 + macrophages, most likely representing or being enriched in embryonic and monocytic origin F4/80 + ATMs, respectively, distinguished according to their Lyve1, CD16.2, Trem2, and CD9 expression profile.However, these did not change in ratio between N1N2 fl and N1N2 fl CRE mice.In view of their significant and sustained effects on DIO, the broad population of F4/80 + ATMs deficient in both Notch1 and Notch2 was analyzed in more detail at a molecular level.Our data demonstrate that abrogation of myeloid Notch1 and Notch2 induces a proinflammatory M1-like phenotype in ATMs that further promotes inflammation and regulates metabolism.This effect is accompanied by the induction of genes involved in extracellular matrix remodeling, immune proliferation and differentiation, lipid kinase activity, and hypoxia in deficient Notch1 and Notch2 ATMs.Seminal studies demonstrated that in the late stages of obesity, there is an accumulation of ECM proteins, and consequently AT fibrosis, which is associated with the proinflammatory phenotype observed in obesity [42].Macrophages have also been implicated in the repair and remodeling of ECM and can contribute directly through the production of proteases and indirectly through the modulation of fibroblast function [43,44].
It remains to be determined, though, how Notch signaling is regulated and at the same time influences the severity of the inflammatory and metabolic abnormalities of DIO ATMs.One prevailing hypothesis is that the already inflammatory milieu of the adipose tissue, favored by the uncontrollable production and/or action of proinflammatory cytokines due to prolonged HFD feeding, could impact Notch signaling.Several studies on other inflammatory conditions and autoimmune diseases have shown that TNF and IL-1β can serve as Notch activators [45,46].IFN-γ, which activates M1 macrophages, has been also proven to induce Jag1 expression rapidly, amplifying further Notch signaling [20], prompting us to speculate that this might also be the case in our context.A growing body of evidence, however, has yielded further insight into the mechanisms implicated in the activation of Notch signaling.Initially, studies have indicated TGFβ, a pathway that is severely deregulated in deficient Notch1 and Notch2 ATMs, to directly induce Hes1 expression in several cell types [47].Specifically, Hes1, a direct downstream mediator of Notch signaling that is significantly impaired in our study, has been shown to hinder inflammation by controlling the production of macrophagederived chemokines [48].Another attractive pathway candidate that could potentially synergize for the activation of Notch target genes is NF-kB signaling which is activated by both TLR ligands and inflammatory cytokines.Additional studies support that Hes1 transcription, in many cases, is dependent on inhibitors of NF-kB kinases.Interestingly, in resting cells, IκBαa was found to be present at the promoter regions of Hes1 while its expression was possible only after TNF-induced IkBαa dismissal from the promoter [49].Another group of signaling molecules, which are key determinants of inflammation and have been implicated in mediating Notch pathway activation, are mitogen-activated protein kinases, a family of serine/threonine protein kinases [50].Indeed, both inhibitors of NF-kB kinases and mitogenactivated protein kinases have been shown to interfere with inflammation-induced chromatin modifications at the Notch target gene loci [49,50].Alternatively, other findings have demonstrated that abrogation of Notch signaling in bone marrowderived macrophages suppresses their proinflammatory properties and leads to an M2-like polarization [51].Several possible reasons may explain the discrepancies between these reports outlining a different role for Notch signaling during macrophage activation.The specific outcome of Notch stimulation in macrophages is likely to involve a tightly regulated balance between positive and negative regulatory signals in a cell context and Notch ligandspecific fashion.Contradictory findings can also be attributed to the timing and dosage of Notch intervention and distinct cell types used at each experimental protocol.Last but not least, the amplitude of Notch activation can differ between physiological and pathological conditions.
Our investigation has certain constraints that must be acknowledged.First, the mechanistic relationship between Notchmediated intercellular signaling and cooperative microenvironmental cues is less clear.Notch-mediated ATM activation contributes to the onset and progression of obesity not only through cell-to-cell communication but also through direct interaction with adipocytes.Notch ligands present on the cell surface of ATMs can potentially activate Notch receptors and consequently Notch signaling in the neighboring adipocytes, exacerbating the insulin resistance of these key cell types involved in energy metabolism.Macrophage-adipocyte interaction is of significance in metabolic disorders where systemic infiltration of macrophages is prevalent.To this end caution needs to be taken when studying the metabolic phenotypes of mice with macrophage-specific deletion of Notch genes.Second, however puzzling it may seem, autoamplification of Notch signaling can be a caveat to this conclusion.Activation of Notch signaling in one cell can create a positive feedback loop and lead to the production of Notch ligands, which then activate the pathway in nearby cells, further amplifying the signal.Furthermore, chronic inflammatory responses of the adipose tissue maintain a vicious cycle of increasing tissue damage, macrophage activation, and inflammation and perturbed signaling pathways that fail to resolve even in response to therapy, rendering it difficult to uncover the causal relationship of this multifarious process.
In summary, this study provides a detailed examination of the complex interplay between Notch signaling and ATMs, demonstrating that Notch signaling in myeloid cells is critically involved in preventing adipose tissue inflammation and metabolic dysregulation in DIO.This sheds light on the underlying mechanisms that regulate metabolic and immune function in obese adipose tissue and suggests that targeting Notch signaling in ATMs or myeloid cells may be a potential therapeutic strategy for obesity-related metabolic disorders.

Mice
Wild-type (WT) and Jag1 fl/fl (B6.129S-Jag1tm2Grid/SjJ,Strain #031272) mice were purchased from Jackson Laboratories and further bred in our animal facility.Notch1 fl/fl , Notch1 fl/fl Notch2 fl/fl , B6.Cg-Gt(ROSA)26Sor tm9(CAG-tdTomato)Hze /J (RosaTomato fl/fl ) and LysMcre mice were provided by Dr. A. Klinakis (Biomedical Research Foundation, Academy of Athens).Notch1 fl/fl , Notch2 fl/fl , and LysMcre mice have been described elsewhere [52,53].Jag1 fl/fl , Notch1 fl/fl and Notch1 fl/fl Notch2 fl/fl were crossed to LysMcre mice to generate the Jag1 fl/fl LysMcre, Notch1 fl/fl LysMcre and Notch1 fl/fl Notch2 fl/fl LysMcre strains which allow the myeloidspecific ablation of ligand Jag1 and receptors Notch1 and Notch2, respectively.Notch1 fl/fl were also crossed to RosaTomato fl/fl before crossing to LysMcre mice.Genetic screening of all mice was carried out by conventional PCR on genomic DNA from mouse tail biopsies.For the detection of Cre recombinase, the following primers were used F: CCATCTGCCACCAGCCAG and R: TCGCCATCTTCCAGCAGG amplifying a 280 bp fragment.For the detection of Jag1 fl/fl , Notch1 fl/fl and Notch2 fl/fl the following primers were used: F: GGCAACAAAACTTGCATGG and R: GGGCACTAACAGAATCTTCTACA, F: CTGAGGCCTAGAGC-CTTGAA and R: CTCGGAATCCCACTGCTTAC, and F: GCTCAGC-TAGAGTGTTGTTCTTG and R: TTTGTGGCCGTAACTTTCTCATG, respectively.All sets of primers generate two fragments, one for WT, and one for floxed mice that made possible the discrimination between homozygous and heterozygous mice.All mice were housed in a controlled environment (23 ± 2°C) under a 12-h light/dark cycle with free access to water and food in full compliance with the guidelines of the FELASA recommendations at the Biomedical Research Foundation Academy of Athens.Male mice were used throughout the study as male mice are more susceptible to diet-induced metabolic dysfunction than female mice.To establish an obesity model, 6-week-old male mice were fed an HFD consisting of 60% of calories from fat (D12492i, Research Diets Inc.) or an NCD containing 10% of calories from fat (D12450i, Research Diets Inc.) as a control for various timepoints.Animal body weight and wellbeing were recorded weekly.All protocols were approved by Institutional and Regional Ethical Review Boards.

Indirect calorimetry
Metabolic measurements were performed using an Oxymax indirect calorimetry system (Columbus Instruments).In short, preweighed mice were housed individually in specifically designed Oxymax calorimeter chambers with ad libitum access to the diet and water for 72 h with a 12-h light/12-h dark cycle in an ambient temperature of 22°C.Mice were housed individually for 2 days prior to transferring into the calorimeter chamber.Rates of VO2 (mL/kg/h) and VCO2 were determined for each chamber every 20 min throughout the studies.Sensors were precalibrated with a standard gas mixture of O 2 , CO 2 , and N 2 and Oxymax system settings were as follows: airflow, 0.6 L/min and sample flow, 0.5 L/min.Metabolic rate was calculated as VO 2 × (3.815 + [1.232 × respiratory exchange ratio]) and normalized for body mass (kcal/kg/h).VO 2 , VCO 2 , respiratory exchange ratio, metabolic rate, food intake (grams), and activity (counts) were evaluated over a 72-h period.

Glucose and insulin tolerance tests
A glucose tolerance test was performed by intraperitoneal administration of 10% D-glucose (1 g/kg body weight) following overnight fasting.The insulin tolerance test was determined by intraperitoneal injection of human regular insulin (Eli Lilly and Co.) at a dose of 0.75 or 1 U/Kg body weight after a 6-h fasting.Glucose was measured for glucose tolerance test at 0, 20, 40, 60, 90, and 120 min and for insulin tolerance test at 0, 20, 40, 60, and 120 in blood drawn (5-10 μL) from a tail venesection using Bayer's Contour Next Meter (Bayer AG).

Epididymal white adipose tissue isolation
The epiWAT was isolated, after careful removal of gonads, weighed, and kept on ice in buffer I (DPBS; Gibco, ThermoFisher Scientific #14190-094) and 0.5% FBS as previously described [54].The tissue was finely minced and further digested by shaking in a solution with fresh collagenase II (1 mg/mL, Sigma-Aldrich #C6885).The mixture was incubated for 30-40 min at 37°C with gentle agitation.The last 5-10 min EDTA (10mM) was added.The tissue homogenate was filtered through a 100-μm cell strainer and centrifuged at 1400 rpm for 10 m after adding 3× FACS buffer (PBS with 0.1% FBS and 2.5 mM EDTA) to generate single-cell suspensions.The supernatant, that contained the floating tissue, separated from the stromal vascular fraction (SVF) pellet.The floating adipocytes (floating phase) were isolated further and resuspended in 1 mL TriReagent (Sigma-Aldrich) for molecular analysis.The SVF was then suspended in RBC lysis buffer (Sigma-Aldrich) to remove erythrocytes before being centrifuged at 1400 rpm for 10 min.The cell pellet was resuspended in FACS buffer and stained with the indicated antibodies.

Flow cytometry and cell sorting
EpiWAT single-cell suspensions were pre-incubated with Fc block solution to minimize unspecific staining.Then, SVF cells were resuspended in FACS buffer and stained with fluorochromeconjugated antibodies against surface markers.Cells were incubated with primary anti-mouse antibodies conjugated to fluorochromes or isotype controls for 30 min at 4°C in the dark at a concentration of 1 mg/mL for anti-mouse MHC II (IA/IE;

Serum measurements
For serum lipid analysis, mice were fasted for 16 h and weighed before blood collection.Serum was separated by centrifugation and stored at −80°C until further analysis.Serum concentrations of total cholesterol, triglyceride, and free fatty acids were determined using the commercially available cholesterol, triglyceride, and free fatty acid assay kits, respectively (Cayman Chemicals) according to the manufacturer's instructions.For cytokine/hormone analysis, sera were centrifuged for 5 min at 15,000 g and the supernatant was assessed for, C-Peptide, GIP, Glucagon, Insulin, Leptin, Resistin, PYY, and Adiponectin using the MILLIPLEX MAP Mouse Metabolic Hormone Expanded panel and Adiponectin Magnetic Bead Single Plex Kit, respectively.

RNA isolation and qPCR
For gene expression analysis, sorted macrophages (97% purity) were used for RNA isolation with the RNeasy Micro kit (QIA-GEN).RNA samples were treated with DNase I (QIAGEN) and quantified on a NanoDrop (ThermoScientific).cDNA synthesis was performed with the PrimeScript RT reagent kit (Takara).Mature adipocytes were collected in TriReagent and RNA isolation was performed through phase separation, according to standard protocols.RNA concentration and integrity were determined spectrophotometrically and electrophoretically, respectively.1 μg of the isolated RNA was treated with RQ1 DNase (Promega) and used for cDNA synthesis with the M-MLV reverse transcriptase (Promega) according to the manufacturer's instructions.Realtime quantitative PCR was performed with iTaq Universal SYBR Green Supermix (Biorad).Relative amounts of mRNA expression were normalized to Gapdh or CypA levels and calculated according to the 2-CT method.

RNA-seq analysis
Sorted macrophages from epiWAT were used for RNA isolation with the RNeasy Micro kit (QIAGEN).RNA samples were treated with DNase I (QIAGEN) and quantified on a NanoDrop (Thermo-Scientific).RNA seq libraries were prepared with the TruSeq RNA Library Prep Kit v2 (Illumina) according to the manufacturer's instructions.The quality of the libraries was validated with an Agilent DNA high-sensitivity kit run on an Agilent 2100 Bioanalyzer.Bar-coded cDNA libraries were pooled together in equal concentrations in one pool per biological tripicate and were sequenced on a HiSeq2000 (Illumina) or on a NextSeq500 (Illumina) at the Genomics Facility of the Greek Genome Center (Biomedical Research Foundation, Academy of Athens).

Transcriptomics analyses
Samples sequenced on NextSeq500 or NextSeq2000 (Illumina) were analyzed using standard protocols.Briefly, raw reads were preprocessed using FastQC v.0.11.2 and cutadapt v.1.6,and then mapped to the mouse genome (Mus musculus UCSC version mm10) using the TopHat version 2.0.13,Bowtie v.1.1.1 and Samtools version v.1.1.The read count table was produced using HTSeq v.0.6.The output computed for each sample (raw read counts) was then used as input for DESeq2 analysis and returned the log2foldchange of each comparison that was used for differential gene expression analysis.Volcano plots were constructed using the Volcano plot tool at Galaxy Main platform based on ggplot and ggrepel R packages.Heatmaps and chord diagrams were drawn using the ComplexHeatmap and circlize packages in R.

Histology
EpiWAT was harvested and fixed in 4% paraformaldehyde solution in PBS (Santa Cruz Biotechnology) overnight at 4°C, and transferred to 30% sucrose in PBS at 4°C for 3 days, prior to OCT embedding (Tissue Tek; Sakura).Tissue sections (30 μm thick) were stained with H&E to determine adipocyte diameters.Slides were washed thoroughly and coverslips were mounted with DPX mounting medium (Sigma-Aldrich).Sections were observed by microscopy on a Leica DM1000 microscope (Leica Microsystems) and image acquisition was performed using Leica Microsystems imaging software.Images were analyzed using ImageJ software (Wayne Rasband).

Picrosirius red staining
Frozen sections (18 μm) of eWAT were dipped in PBS for 1 min and were then stained in Harris for 30 min.Sections were rinsed well in distilled water, fixed in 70% ethanol (for 1 min), and washed with distilled water.Sections were stained with Picrosirius Red in a saturated aqueous solution of picric acid for 1 h in the dark and were then washed twice with 0.5% acetic acid (5 min), dehydrated with 100% ethanol, cleared with xylene, and mounted with Coverquick 2000 (VWR Chemicals).Percent Picrosirius Redpositive area was quantified using ImageJ.

Quantitation and statistical analysis
All data are presented as mean ± SEM and were analyzed on GraphPad Prism software.Statistical significance of differences was assessed using the parametric Student's two-tailed t-test for normally distributed data and the nonparametric Mann-Whitney U test for skewed data that deviate from normality.For studies with multiple parameters, two-way ANOVA with multiple comparisons was used.Differences were considered significant when p < 0.05.

Figure 1 .
Figure 1.ATMs accumulate during diet-induced obesity and upregulate CD11c.(A) Experimental scheme for diet-induced obesity model.Six-weekold male wild-type (WT) mice were fed a normal chow diet (NCD-10% calories from fat) and a high-fat diet (HFD-60% calories from fat) for 4-and 16 weeks.Body mass was measured weekly and tissue collection was performed at the indicated timepoints.(B) Body mass and representative image of WT mice fed NCD (left) and HFD (right) for 16 weeks.(C) Quantification graphs of key immune populations of the stromal vascular fraction (SVF) of epididymal white adipose tissue (epiWAT) of WT mice fed NCD and HFD as a percentage of total cells.(D) Representative flow plot of gating strategy depicting adipose tissue macrophages (ATMs) and their CD11c subpopulation from WT mice fed NCD and HFD for 16 weeks (E) Quantification of CD301 lectin and CD11c integrin of ATMs of the indicated groups using flow cytometry.Data are presented as mean values with SEM from n = 10-28 mice per group pooled from at least three independent experiments.P-values were determined by a two-tailed Mann-Whitney U test for nonparametric comparisons and two-way ANOVA.*P < 0.05, **P < 0.01, and ***P < 0.001 show significance over WT controls fed an NCD.

Figure 2 .
Figure 2. ATMs shift to a proinflammatory phenotype and dysregulate Notch signaling during HFD-induced obesity and insulin resistance.(A) Representative flow plot of ATMs and schematic depicting sorted ATMs from WT mice fed NCD or HFD for 4 and 16 weeks, respectively, used for RNA-seq analysis.(B) Principal component analysis (PCA) of gene expression data.Dots represent ATMs from mice subjected to NCD (green) or HFD (blue) for 16 weeks.(C) Venn diagram of differentially expressed genes (DEGs) in epiWAT of 4 weeks NCD vs HFD ATMs (grey circle), 4 vs. 16 weeks HFD ATMs (blue circle), and 16 weeks NCD vs. HFD ATMs (red circle).(D) Volcano plot of DEGs between ATMs of mice subjected to HFD versus NCD for 16 weeks.The red and blue points in the plot represent upregulated and downregulated DEGs respectively.(E) Box plots of transcripts per million (TPM) values in log2 scale show the distribution of CD301 and CD11c expression in the indicated groups.(F) Heatmap of immune response genes of ATMs from NCD-and HFD-fed mice for 4-and 16 weeks respectively.(G) Heatmap of genes related to Notch pathway of ATMs from NCDand HFD-fed mice for 4-and 16 weeks, respectively.RNA-seq was performed in biological triplicates for each group.Relative high gene expression is marked in red, whereas relative low gene expression is marked in blue.NCD4, NCD for 4 weeks, NCD16, NCD for 16 weeks, HFD4, HFD for 4 weeks, HFD16, HFD for 16 weeks.ATMs, adipose tissue macrophages; epiWAT, epididymal white adipose tissue; HFD, high-fat diet; NAFLD, nonalcoholic fatty liver disease; NCD, normal-chow diet.

Figure 3 .
Figure 3. Double Notch1 and Notch2 deletion from macrophages aggravates diet-induced metabolic dysregulation in mice.(A) Experimental scheme of DIO in N1N2 fl Cre mice.Male 6-week-old N1N2 fl Cre (CRE) and N1N2 fl (WT) mice were given HFD for 12 weeks and their body mass was monitored weekly.At the end of the 12 weeks mice were subjected to analysis of energy metabolism with the Oxymax system (E.M.A.) and insulin tolerance tests (ITT, and sacrificed for tissue collection and subsequent analysis.(B) Body weight of experimental mice as percentage difference to D1.The inset depicts the area under the curve (AUC) for the weight difference.Data are mean ± SEM from n = 20-28 mice per group and representative of at least five independent experiments.Quantification of (C) epiWAT weight and (D) basal blood glucose levels of CRE and WT mice at the end of the experimental protocol.(E) ITT in CRE and WT mice given HFD for 12 weeks.Percentage values of initial blood glucose concentration are presented.(F-H) Serum levels of triglycerides (F), insulin (G), and leptin (H) of the indicated groups at 12 weeks of HFD.Data are mean ± SEM from n = 8-22 mice per group pooled from 5 independent experiments.(I-L) Respiratory exchange ratio (RER) (I), metabolic (J), fat oxidation (K), and carbohydrate oxidation (L) rate of CRE and WT mice fed HFD for 12 weeks, assessed by indirect calorimetry.Data are expressed as mean ± SEM from n = 18 for WT and n = 15 for CRE mice pooled from 5 independent experiments.*P < 0.05, **P < 0.01, and ***P < 0.001 were determined by a two-tailed Mann-Whitney U test for nonparametric comparisons (C-L) and two-way ANOVA (B).epiWAT, epididymal white adipose tissue; F.C, Flow Cytometry; ITT, insulin tolerance test; E.M.A., energy metabolism analysis.

Figure 4 .
Figure 4. Double Notch1 and Notch2 deletion from macrophages deteriorates adipose tissue inflammation and ensuing metabolic dysregulation.(A) Representative flow plots of gating strategy depicting ATMs and their CD11c subpopulation from N1N2 fl Cre and N1N2 fl mice fed HFD for 12 weeks (B, C) Quantification graphs of immune cell populations of the epiWAT of N1N2 fl Cre and N1N2 fl mice given HFD for 12 weeks as a percentage of total cells by flow cytometry.Data are expressed as mean ± SEM of n = 15-27 mice pooled from three to five independent experiments.(D, E) Molecular analysis of FACS sorted ATMs from N1N2 fl Cre and N1N2 fl mice given HFD for 12 weeks.(D) Inflammatory markers Nos2, Arg1, and IFNγ are depicted on the left y-axis while IL6, TNF, and IL10 are on the right y-axis.(E) Metabolic markers Irs1, Irs2, Furin, and Igf1r are depicted on the left y-axis while Igf2r and Hk2 are on the right y-axis.All genes were assessed with qPCR.Data are expressed as mean ± SEM from n = 7 for WT and n = 5 for CRE mice.(F) Molecular analysis of isolated adipocytes from epiWAT of N1N2 fl Cre and N1N2 fl mice given HFD for 12 weeks.Data are mean ± SEM from n = 8 mice for each of the indicated groups.*P < 0.05, **P < 0.01, and ***P < 0.001 were determined by a two-tailed Mann-Whitney U test for nonparametric comparisons.ATMs, adipose tissue macrophages; epiWAT, epididymal white adipose tissue; HFD, high-fat diet.

Figure 5 .
Figure 5. Notch1 −/− Notch2 −/− macrophages exhibit transcriptome changes during DIO.(A) Representative cell density plots from spectral flow cytometry showing the strategy used for enrichment of ATM subpopulations.Numbers indicate the percentage of gated cells ± SEM.Data are mean from n = 3 mice for each of the indicated groups.(B) PCA of gene expression data.Dots represent ATMs from N1N2 fl (blue) and N1N2 fl Cre (green) mice fed an HFD for 12 weeks.(C) Volcano plot of DEGs between N1N2 fl Cre versus N1N2 fl .The red and blue points in the plot represent upregulated and downregulated DEGs, respectively.(D, E) Gene Ontology enrichment analysis regarding (C) Biological Processes (BP) or (D) Kyoto Encyclopedia of Genes and Genomes pathways of DEGs of ATMs from N1N2 fl Cre vs. N1N2 fl mice fed an HFD for 12 weeks.RNA-seq was performed in biological triplicates for each sample.ATMs, adipose tissue macrophages; DEGs, differentially expressed genes; HFD, high-fat diet; PCA, principal component analysis.
Evangelos Andreakos analyzed the data; Eleni Siouti and Ioanna-Evdokia Galani performed the RNA-seq and Luminex experiments; Apostolos Klinakis provided key conditional knockout animals; Maria Manioudaki and Eleutherios Pavlos analyzed the RNA-seq data; Eleni Siouti and Evangelos Andreakos drafted the manuscript; Evangelos Andreakos designed and supervised the study.Animals were housed in ventilated cages under specific-pathogen-free conditions with humane care in full compliance with the guidelines of the FELASA recommendations at the Biomedical Research Foundation Academy of Athens.Experiments were approved by the Directorate for Agricultural and Veterinary policy (5831/29-10-2018), authorized by the Greek Ministry of Rural Development and Food, and conducted in compliance with Greek and European regulations.