Oxidative Stress Inhibits Healthy Adipose Expansion Through Suppression of SREBF1-Mediated Lipogenic Pathway

The prevalence of obesity and type 2 diabetes mellitus (T2DM) is increasing worldwide. T2DM is characterized by peripheral insulin resistance and insufficient insulin release from pancreatic β-cells. Among the drugs used for the treatment of T2DM, two agents, metformin and peroxisome proliferator–activated receptor-γ (PPARγ) agonists, can clinically improve peripheral insulin resistance. For better treatment of T2DM, it is important to elucidate the mechanisms of insulin resistance in obesity and to develop insulin-sensitizing drugs.

Under excess calorie intake or insufficient energy expenditure, the white adipose tissue (WAT) stores lipids and becomes hypertrophic. Hypertrophic adipocytes impair insulin signaling not only in these cells but also in other insulin-sensitive organs through dysregulated secretion of adipocytokines, such as adiponectin (1), tumor necrosis factor (2), interleukin 6 (3), and resistin x(4). Recent studies have demonstrated that such changes are preceded by various changes in WAT, such as infiltration of inflammatory cells (5), endoplasmic reticulum stress (6), hypoxia (7), and oxidative stress (8).

We reported that the presence of high levels of pro-oxidants, such as NADPH oxidases, and low levels of antioxidant enzymes, such as Cat and Sod, in obese WAT resulted in high oxidative stress, conceptualized as adipose oxidative stress (Fat reactive oxygen species [ROS]), in mice (8) and humans (9–11). Many studies have suggested that Fat ROS are involved in various pathways in metabolic syndrome, including adipocyte insulin signaling (12), adipose inflammation (8), endoplasmic reticulum stress (13), mitochondrial function (14), cellular senescence (15), gut microbiota (16), adiponectin (Adipoq) (8), adiponectin receptor (AdipoR) (17), and angiotensin II signaling (18). Systemic or skeletal muscle-specific manipulation of oxidative stress in vivo, including administration of apocynin (8) or manganese­–5,10,15,20-tetrakis (4-benzoic acid) porphyrin (MnTBAP) (19), overexpression of Sod (19,20) or Cat (21,22), and glutathione depletion (23–25), showed various effects on insulin sensitivity. However, the causal role of Fat ROS in obesity in vivo still remains unclear because no mouse model has been available in which oxidative stress targeting adipocytes is manipulated.

We describe here the generation of mice with genetically manipulated ROS specifically in adipocytes. Elimination of Fat ROS potentiated healthy adipose expansion with decreased ectopic lipid deposition and improved insulin sensitivity. Conversely, augmented Fat ROS inhibited healthy adipose expansion with ectopic lipid accumulation and deteriorated insulin sensitivity. As a mechanism, Fat ROS accelerated adipose inflammation and adipose tissue fibrosis and inhibited de novo lipogenesis with the suppression of sterol-regulatory element-binding transcription factor 1 (SREBF1) transcriptional activity through a reduction in KDM1A protein expression.

To generate aP2-Cat/SOD1 double-transgenic mice, the rat Cat gene and human SOD1 gene were amplified by PCR and were inserted into an aP2 promoter cassette. These plasmids were microinjected into fertilized eggs from C57BL/6J mice, generating an aP2-Cat transgenic mouse and an aP2-SOD1 transgenic mouse. The expression of transgenes was checked in several lines of aP2-Cat transgenic and aP2-SOD1 transgenic mice, and the line with the highest expression of each transgene in WAT was selected. In the next step, the aP2-Cat transgenic and aP2-SOD1 transgenic mice were crossed, and the resultant wild-type mice and aP2-Cat/SOD1 double-transgenic mice were analyzed as littermates.

Adipoq-Cre mice were provided by Rosen and colleagues (26). The LacZ-Neo cassette was removed from Gclc mutant mice (05085, The European Mouse Mutant Archive) by crossing them with CAG-FLPe mice (Riken). The resultant Gclc floxed mice were crossed with Adipoq-Cre mice to generate Adipoq-Gclc–knockout (AKO) mice. The AKO mice and Gclc floxed mice were crossed, and the resultant AKO mice and Gclc floxed mice were analyzed as littermates. To distinguish the floxed allele from the wild-type allele, the following primers were used: 5′-AGGAGGAGGGTGGGAAATTAC-3′ and 5′-GCATAACATCAAAGCCTGCAG-3′. To distinguish null alleles from floxed alleles, the following primers were used: 5′-GTCAGCATCTCTGCTGTGGATC-3′ and 5′-AGGAGGAGGGTGGGAAATTAC-3′.

In experiments with a high fat/high sucrose (HF/HS) diet, the mice were fed F2HFHSD (Oriental Yeast) from 6 weeks of age. All mice were maintained under specific pathogen-free conditions and had free access to water and chow.

The mice were anesthetized, and blood samples were collected from the inferior vena cava. Tissues were carefully removed and snap frozen in nitrogen.

Total RNA was isolated with Sepasol-RNA I Super G (Nacalai Tesque, Kyoto, Japan) or TRI Reagent (Sigma-Aldrich), according to the protocol provided by the manufacturer. First-strand cDNA was synthesized from total RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche) and was subjected to real-time RT-PCR using a Light Cycler (Roche) according to the instructions provided by the manufacturer. Values were expressed relative to the mRNA level of Rplp0. The primers used are provided in Supplementary Table 1.

Adipose tissues were fractionated, and adipose tissue macrophages (ATMs) were collected by FACS Aria II (BD Biosciences) as described previously (27). M1 and M2 macrophages were analyzed by FACS Aria II as described previously (28).

Catalase (CAT) and Sod activities in WAT were measured using the Catalase Assay Kit (Cayman) and the Superoxide Dismutase Assay Kit (Cayman), respectively, according to the instructions supplied by the manufacturer. Values were adjusted to the level of protein measured by the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific).

Tissue hydrogen peroxide was measured using the Hydrogen Peroxide Assay Kit (Abcam), according to the instructions supplied by the manufacturer. Values were adjusted to the level of protein measured by the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific).

Tissue 8-isoprostane was extracted and measured using the 8-Isoprostane EIA Kit (Cayman) according to the protocol recommended by the manufacturer. Briefly, tissue was lysed, incubated with potassium hydroxide, trapped with C18 Sep-Pak (Waters), washed with water and hexane, eluted with ethyl acetate, evaporated, and reconstituted with enzyme immunoassay buffer. Values were adjusted to the DNA amounts measured by the CellTox Green Cytotoxicity Assay (Promega).

Tissue total glutathione was measured using the GSSG/GSH Quantification Kit (Dojindo Laboratories, Kumamoto, Japan) according to the protocol recommended by the manufacturer. Values were adjusted to the level of protein measured by the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific).

Tissue carbonylated protein was detected using the OxyBlot Protein Oxidation Detection Kit (Millipore) according to the protocol recommended by the manufacturer. Blots were quantified using the ImageQuant LAS 4000 (GE Healthcare).

Paraffin sections of WAT or brown adipose tissue (BAT) were processed for hematoxylin-eosin (H-E) staining and were examined under a light microscope. Adipocyte areas were measured using the BZ-II Analyzer (Keyence). Paraffin sections of liver were processed for Oil Red O staining. Paraffin sections of WAT were stained using the Picro-Sirius Red Stain Kit (ScyTek Laboratories) according to the protocol recommended by the manufacturer. Stained signals were measured using the BZ-X Analyzer (Keyence).

Liver triglycerides and glycogen were extracted and measured using the Triglyceride E-test (Wako Pure Chemical Industries) and Glycogen Assay Kit (Cayman), respectively, using the procedures recommended by the manufacturers. Values were adjusted to the DNA amounts measured by the CellTox Green Cytotoxicity Assay (Promega).

Plasma insulin, triglycerides, and nonesterified fatty acids (NEFAs) were measured using the insulin ELISA kit (Morinaga), triglyceride E-test (Wako Pure Chemical Industries), and NEFA C-test (Wako Pure Chemical Industries), respectively, according to the protocols supplied by the manufacturers.

Food was withheld for 4 h before glucose (1 g/kg) or insulin (1.3 units/kg) was administered intraperitoneally. Blood samples were collected from the tail vein at the indicated times after the injection. Blood glucose was immediately determined using the Glutest Sensor (Sanwa Kagaku Kenkyusho, Nagoya, Japan).

Food was withheld for 6 h before insulin (2.6 units/kg) was administered by intraperitoneal injection. Mice were sacrificed 10 minutes after the injection, and tissues were rapidly removed and frozen in nitrogen. Protein levels were determined by blotting with anti-Akt antibody (Cell Signaling) and anti–phosphorylated-Akt antibody (Ser 473) (Cell Signaling) and were then quantified using the ImageQuant LAS 4000 (GE Healthcare).

3T3-L1 mouse fibroblasts were maintained in DMEM (high glucose)/10% FBS and were differentiated into adipocytes by treatment with dexamethasone, insulin, and 3-isobutyl-1-methylxanthine for 2 days.

Doxycycline hydrochloride was purchased from Nacalai Tesque. Tert-butyl hydroperoxide (TBHP) solution in water and N-acetyl-L-cysteine (NAC) were purchased from Sigma-Aldrich.

Platinum-E cells were transfected with pRetroX-Tet-On Advanced (TaKaRa), pRetroX-Tight-Hyg (TaKaRa) harboring the mouse Cat gene, and pRetroX-Tight-Pur (TaKaRa) harboring the mouse Sod1 gene. The media containing the retroviruses were harvested 48 h after transfection, filtered, and transferred to 3T3-L1 cells. Infected cells were selected with 200 μg/mL G418, 400 μg/mL hygromycin, and 1 μg/mL puromycin.

Intracellular ROS were measured using the DCFDA Cellular ROS Detection Assay Kit (Abcam) with the instructions provided by the manufacturer.

3xSRE-Luc, cytomegalovirus (CMV) 7, CMV-nuclear Srebp1a, CMV-nuclear Srebp1c, and CMV-nuclear Srebp2 were provided by Shimano and colleagues (29). 3xChoRE-Luc, CMVS, CMVS–carbohydrate response element-binding protein (ChREBP), and CMVS-Mlxγ were provided by Towle and colleagues (30). 3T3-L1 adipocytes were transfected with these plasmids together with the pRL-CMV vector (Promega) as an internal control. The cells were harvested 24 h later and subjected to a luciferase assay using the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s instruction.

Mature 3T3-L1 adipocytes were harvested and small interfering (si)RNA (Qiagen) were introduced by reverse transfection using the Lipofectamine RNAiMAX Reagent (Thermo Fisher Scientific) according to the instructions provided by manufacturer.

Cells or adipose tissues were lysed with TNE buffer (10 mmol/L Tris-HCl [pH 7.8], 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP40), and the lysates were subjected to Western blotting with antibodies to KDM1A (ab17721; Abcam) and ACTB (A5441; Sigma-Aldrich).

Data are presented as the mean ± SEM. Differences between groups were analyzed by Student two-tailed t tests. Statistical significance was set at P < 0.05.

The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Osaka University, Graduate School of Medicine. All animal experiments were performed in accordance with the Osaka University Institutional Animal Care and Use Committee Guidelines.

To establish the Fat ROS–eliminated mice, we generated aP2-Cat/SOD1 double-transgenic mice (aP2-dTg), which expressed rat Cat and human SOD1 under control of the aP2 promoter. Superoxide dismutase 1 (SOD1) and CAT act synergistically in scavenging ROS; SOD1 converts the superoxide radical into hydrogen peroxide, which is then catalyzed into water and oxygen by CAT (Fig. 1A). In accordance with the known expression of the aP2 gene (31), these mice expressed transgenes abundantly in the WAT, BAT (Fig. 1B), and ATMs (Fig. 1C). In the gonadal WAT of aP2-dTg mice fed the HF/HS diet, the activities of CAT and SOD were 1.7- and 7.7-fold higher, respectively, than those in their wild-type littermates (Fig. 1D). Tissue hydrogen peroxide was significantly reduced in subcutaneous and gonadal WAT but not in mesenteric WAT of obese aP2-dTg mice (Fig. 1E). 8-Isoprostane, a well-defined marker of tissue oxidative damage (20,32), was also decreased in gonadal WAT but not in the liver of obese aP2-dTg mice (Fig. 1F). Fractionation of adipose tissue showed reduced 8-isoprostane in the adipocyte fraction but not in stromal vascular cells (SVC) (Fig. 1F). Based on these data, we established the aP2-dTg mice as Fat ROS–eliminated mice.

Generation of Fat ROS–augmented mice was achieved by depleting glutathione specifically in adipocytes, which reduced oxygen radicals to hydrogen peroxide or hydrogen peroxide to water and oxygen, together with glutathione peroxidase (Fig. 1A). For this aim, floxed mice of the catalytic subunit of glutamate-cysteine ligase (Gclc), which is a rate-limiting enzyme for glutathione synthesis (Fig. 1A), were bred with adipocyte-specific Adipoq-Cre mice to generate AKO mice. As expected, DNA (see Supplementary Fig. 1) and mRNA expression levels (Fig. 2A) of the Gclc gene were knocked out specifically in WAT and BAT. As expected, tissue total glutathione was significantly decreased in the WAT of AKO mice (Fig. 2B), and tissue hydrogen peroxide was significantly increased in the WAT of AKO mice (Fig. 2C). Tissue 8-isoprostane was increased in subcutaneous and gonadal WAT but not in the liver (Fig. 2D). Tissue 8-isoprostane did not change in BAT (Fig. 2D) despite reduction of Gclc gene expression (Fig. 2A), the reason of which is currently unclear; however, glutathione might not play major roles in the antioxidant system of BAT, or whitening of BAT (presented later) attenuated generation of ROS. Protein oxidation was also increased in the WAT of AKO mice (Fig. 2E). On the basis of these data, we established AKO mice as Fat ROS–augmented mice.

The aP2-dTg mice fed the HF/HS diet gained similar body weights to their wild-type littermates (Fig. 3A). However, after loading with the HF/HS diet for 24 weeks, these mice exhibited significant changes in tissue weight. While the weights of subcutaneous and gonadal WAT increased, those of mesenteric WAT and the liver decreased, with no changes in skeletal muscle or BAT weight (Fig. 3B). Histological analyses showed significantly larger adipocytes in subcutaneous (Fig. 3C) and gonadal WAT (see Supplementary Fig. 2A) and smaller adipocytes in the mesenteric WAT of aP2-dTg mice (see Supplementary Fig. 2B), indicating that the alteration of WAT weight in each depot resulted from the alteration of lipid accumulation in adipocytes. The expansion of subcutaneous WAT was associated with increased expression levels of Acly, Scd1, Fasn, and Acaca, which catalyze key steps in de novo lipogenesis, in aP2-dTg mice compared with those in their wild-type littermates (Fig. 3D). Gene expression of Srebf1 was not changed (Fig. 3D). The gene expression levels of these lipogenic genes were not induced in the mesenteric WAT of aP2-dTg mice (see Supplementary Fig. 3A), possibly explaining the distinct tendency toward lipid accumulation between the subcutaneous/gonadal and mesenteric WAT in Fat ROS–eliminated mice (Fig. 3B). The expression levels of lipolysis-related genes—Lpl, Lipe, and Pnpla2—in the subcutaneous WAT of aP2-dTg mice were similar to those in wild-type mice (see Supplementary Fig. 3B). Oil Red O staining revealed improved steatosis in the liver of aP2-dTg mice compared with that of wild-type littermates (Fig. 3E), which was further confirmed by decreased triglyceride content, reduced glycogen content, and downregulated expression levels of Pparg, a marker of steatosis (Fig. 3F). The expression levels of lipogenic genes, Fasn and Scd1, were significantly lower in the liver, possibly explaining the improved steatosis in aP2-dTg mice (see Supplementary Fig. 3C). From these data we speculated that removal of Fat ROS shifted the lipid accumulation toward subcutaneous and gonadal WAT from mesenteric WAT and the liver in aP2-dTg mice.

Next, we investigated the metabolic phenotypes of Fat ROS–eliminated mice fed the HF/HS diet for 24 weeks. The aP2-dTg mice showed similar profiles of fasting plasma glucose, insulin, triglyceride, NEFA, and glucose tolerance (see Supplementary Fig. 4) to their wild-type littermates. Whole-body insulin sensitivity was significantly improved in aP2-dTg mice compared with that in their wild-type littermates (Fig. 4A), and aP2-dTg mice exhibited enhanced insulin-dependent AKT (Ser473) phosphorylation in the liver but not in WAT or skeletal muscle (Fig. 4B and see Supplementary Fig. 5), consistent with improved steatosis. These data indicated that subcutaneous/gonadal WAT in these mice underwent adipose expansion associated with improved systemic and hepatic insulin sensitivity.

Recently, two types of adipose expansion—healthy expansion and pathological expansion—have begun to be recognized (33). Adipose healthy expansion has reportedly been associated with less adipose fibrosis (34–37) or enhanced de novo lipogenesis (38). In addition to enhanced de novo lipogenesis in WAT of aP2-dTg mice (Fig. 3D), we analyzed tissue fibrosis and macrophage polarization in these mice. Picrosirius staining showed less adipose tissue fibrosis in subcutaneous and gonadal WAT of aP2-dTg mice compared with wild-type littermates (Fig. 4C). Some of the fibrosis-related genes (Col1a1 and Timp1) were downregulated and angiogenesis-related genes (Vegfa and Fgf2) were upregulated in the subcutaneous WAT of aP2-dTg mice (Fig. 4D). The population of M2 macrophages was decreased, whereas M1 macrophages were increased in the WAT of Fat ROS–eliminated mice (Fig. 4E). Taken together, these data suggested that beneficial changes in tissue fibrosis, macrophage polarization, and angiogenesis, as well as increased de novo lipogenesis, might contribute to the healthy adipose expansion in Fat ROS–eliminated mice.

AKO mice fed the HF/HS diet showed a trend to gain less body weight than their control littermates (Fig. 5A). After loading on the HF/HS diet for 6 weeks, these mice exhibited significant changes in tissue weights. Although the weight of WAT decreased, the weights of BAT and the liver increased (Fig. 5B). Histological analyses showed significantly smaller adipocytes in gonadal WAT of AKO mice (Fig. 5C). Smaller adipocytes in AKO mice were associated with decreased expression levels of Acly, Scd1, Fasn, Acaca, and Srebf1 compared with those in their wild-type littermates (Fig. 5D). In the liver (Fig. 5E) and BAT (Fig. 5F), lipid accumulation was greater in AKO mice than in their control littermates. Consistent with the lipid accumulation in BAT, thermogenic function was impaired as assessed by diminished expressions of Ucp1 and Ppargc1a in the BAT of AKO mice (Fig. 5G).

Insulin resistance was more prominent in AKO mice than that in their control littermates (Fig. 6A). Picrosirius staining showed enhanced adipose tissue fibrosis in WAT of AKO mice compared with control littermates (Fig. 6B). Fibrosis-related genes (Col1a1, Tgfb1, and Timp1) were upregulated, whereas angiogenesis-related genes (Vegfa, Vegfb, and Fgf2) were downregulated in WAT of AKO mice (Fig. 6C). The population of M1 macrophages was increased, whereas that of M2 macrophages was decreased (Fig. 6D), consistent with increased inflammatory genes in WAT of AKO mice (Fig. 6E). These phenotypes in AKO mice provided a sharp contrast to those in aP2-dTg mice, further supporting the roles of Fat ROS in healthy adipose expansion, insulin resistance, and adipose de novo lipogenesis.

Next, we investigated whether oxidative stress altered the expression of lipogenic genes in cultured adipocytes. Treatment of 3T3-L1 adipocytes with TBHP, a superoxide radical generator, resulted in significant downregulation of Acly and Scd1 (Fig. 7A). Conversely, treatment with NAC, a ROS-eliminating agent, resulted in significant upregulation of these genes (Fig. 7B), indicating that oxidative stress suppressed the gene expression levels of lipogenic genes in adipocytes. To genetically confirm this hypothesis, we introduced 3T3-L1 cells with stable expression of pRetroX-Tet-On, pRetroX-Tight-Pur-Cat, and pRetroX-Tight-Hyg-Sod1, and thus established 3T3-L1-TetON-Cat/Sod1 cells. These cells conditionally expressed ectopic Cat and Sod1 when treated with doxycycline (see Supplementary Fig. 6). As expected, pretreatment with doxycycline eliminated TBHP-induced ROS (Fig. 7C) and partially reversed the suppression of lipogenic genes by TBHP in 3T3L1-TetON-Cat/Sod1 cells (Fig. 7D). The gene expression levels of lipogenic genes are predominantly regulated by Srebfs and the Chrebp/MAX-like protein X (Mlx) complex. TBHP dose-dependently suppressed the transcriptional activities of the nuclear forms of SREBF1A and 1C but not of SREBF2 in 3T3-L1 adipocytes (Fig. 7E). In contrast, TBHP exerted no effects on the transcriptional activity of the CHREBP/MLX complex (Fig. 7F). These results suggested that ROS downregulated the expression of lipogenic genes through the suppression of SREBF1 transcriptional activities.

To gain mechanistic insight into how oxidative stress inhibited SREBF1 transcriptional activities, we focused on Kdm1a (also known as Lsd1), which was recently reported to be necessary for the transcriptional activity of SREBF1 in hepatocytes (39). In 3T3-L1 adipocytes, Kdm1a was also necessary for full activation of nuclear SREBF1A (Fig. 8A and B). We found that treatment with TBHP decreased KDM1A protein abundance (Fig. 8C), with no change in the mRNA expression levels (see Supplementary Fig. 7) in 3T3-L1 adipocytes. Knockdown of Kdm1a dampened the suppressive effects of TBHP on SREBF1A transcriptional activity (Fig. 8D). KDM1A protein expression was significantly decreased in WAT of AKO mice (Fig. 8E and see Supplementary Fig. 8A) and showed the trend to increase in WAT of aP2-dTg mice (see Supplementary Fig. 8B and C) compared with that in their control littermates, without any change in mRNA expression (see Supplementary Fig. 9). Taken together, Fat ROS suppressed SREBF1A transcriptional activities, at least in part, through the reduction of KDM1A protein abundance.

In the current study, we established Fat ROS–eliminated and –augmented mice through the genetic manipulation of antioxidant-related genes. Under diet-induced obesity, these models showed highly consistent phenotypes with each other. Fat ROS–eliminated mice exhibited white adipose expansion with beneficial macrophage polarization, fibrosis, and de novo lipogenesis, accompanied by decreased ectopic lipid deposition and improved insulin sensitivity. Conversely, Fat ROS–augmented mice exhibited restricted adipose expansion with unfavorable adipose inflammation, fibrosis, and de novo lipogenesis, accompanied by increased ectopic lipid accumulation and accelerated insulin resistance.

On one hand, there is sufficient evidence demonstrating that adipocyte hypertrophy with adipose inflammation is a cause of metabolic disease. On the other hand, the adipose tissue can act as a buffer to store excess energy, preventing ectopic lipid deposition and insulin resistance. For example, lipodystrophic mice, which have sparse adipose tissue, exhibit severe insulin resistance with lipid accumulation in the liver (40,41). In agreement with these beneficial roles of adipose tissue, several recent reports have documented adipose expansion associated with improvement in insulin sensitivity and decreased ectopic lipid accumulation, which is conceptualized as healthy adipose expansion (14,34–38). Adipose expansion in Fat ROS–eliminated mice could be regarded as healthy expansion, whereas restricted adipose expansion in Fat ROS–augmented mice indicated the inhibition of healthy expansion.

Several reports have documented adipose tissue fibrosis as an underlying mechanism of healthy adipose expansion. Collagen VI deficiency in adipose tissues (35) and Irf5 (34) or Mincle (36) deficiency in macrophages have been associated with healthy adipose expansion through decreased adipose fibrosis. Adipose tissue inflammation was associated with restricted adipose expansion and insulin resistance through altered adipose fibrosis (37). In this study, adipose expansion in Fat ROS–eliminated mice was associated with altered macrophage polarization from M1 to M2 and attenuated adipose fibrosis, whereas adipose restriction in Fat ROS–augmented mice was associated with increased adipose inflammation and accelerated fibrosis. Oxidative stress is known to activate proinflammatory transcription factors, such as nuclear factor-κB, STAT3, hypoxia-inducible factor 1α, activator protein 1, and Nrf2 (42). Oxidative stress also induces inflammatory-related gene expressions in adipocytes in vitro, such as Il6, Ccl2, and Serpine1 (PAI-1) (8). So, it seemed reasonable that removal of Fat ROS prevented adipose inflammation and associated adipose fibrosis, whereas augmented Fat ROS potentiated inflammatory changes and resulting fibrosis in adipose tissues.

Another study reported the involvement of de novo lipogenesis in adipose healthy expansion (38). Likewise, in Fat ROS–modified mice, altered adipose expansion was associated with altered de novo lipogenesis as expressions of lipogenic genes were upregulated in Fat ROS–eliminated mice and were downregulated in Fat ROS–augmented mice. These suggested that oxidative stress might directly inhibit de novo lipogenesis in adipocytes. However, few studies have examined the relationships between oxidative stress and de novo lipogenesis in adipocytes. Guo et al. (43) reported that octanoate inhibited adipocyte lipogenesis through oxidative stress, although the underlying mechanism was unclear. In the current study, we confirmed that oxidative stress directly suppressed the expression of lipogenic genes in 3T3-L1 adipocytes, which was accompanied by the suppression of the transcriptional activities of SREBF1. We further identified Kdm1a as responsible for these actions because oxidative stress decreased KDM1A protein expression both in vitro and in vivo, and knockdown of Kdm1a dampened the suppressive effects of oxidative stress on the expression of lipogenic genes. KDM1A is reportedly destabilized by Jade2 (44) and is stabilized by Usp28 (45). Thus, oxidative stress might decrease the stability of KDM1A through these ubiquitin-related proteins, although further studies will be required to confirm this hypothesis.

In Fat ROS–eliminated mice, mesenteric WAT was paradoxically decreased in contrast to the expansion of subcutaneous and gonadal WAT. This finding is important, because expansion of mesenteric WAT is clinically more correlated with metabolic syndrome (46). These differences were probably mediated by distinct sensitivity to antioxidants among fat depots, because overexpression of Cat and Sod1 efficiently eliminated Fat ROS in subcutaneous and gonadal WAT but not in mesenteric WAT (Fig. 1G). The highest expression of NADPH oxidase in mesenteric WAT (see Supplementary Fig. 10) might explain these different sensitivities to antioxidants, but further analyses would be required.

The concept of healthy adipose expansion is also applicable to humans. Many studies have reported that 10–25% of obese subjects do not present metabolic disturbances, which is categorized as “metabolically healthy but obese (MHO) phenotype” (47–49). MHO subjects were characterized by predominant fat accumulation toward subcutaneous WAT (50), lower liver fat content (51), and lower number of adipose tissue macrophages (52), which are highly consistent with the phenotypes of Fat ROS–eliminated mice. Although the underlying mechanisms that arise in the MHO phenotype are still poorly understood, lower Fat ROS might lead to MHO phenotypes in human subjects as well as in mice. Also, PPARγ agonists reduce the intraabdominal fat mass without affecting total fat and improve insulin sensitivity in human subjects (53), similar to the phenotypes seen in Fat ROS–eliminated mice. Considering that Cat and Sod1 are direct PPARγ target genes in adipocytes (54–56), it is fascinating to speculate that the beneficial effects of PPARγ agonists could be mediated, at least in part, by a reduction of Fat ROS. Further studies will be required whether oxidative stress–related genes, including CAT and SOD, are regulated in the WAT of MHO subjects or patients treated with PPARγ agonists.

In this study, we used the aP2 promoter to specifically drive Cat and Sod1 in adipocytes. However, expression of the aP2 gene is not known to be strictly specific in adipocytes. In fact, ectopic Cat and Sod1 were to some extent expressed in tissues other than adipose tissues and expressed in ATM with comparable level to adipocytes (Fig. 1B and C). Although we evaluated oxidative stress and found no change in the liver or SVC, it did not completely exclude the involvement of ROS in tissues other than adipocytes. Nevertheless, we believe that healthy adipose expansion in aP2-dTg mice was a consequence of ROS elimination in adipocytes, because the AKO mice, in which highly adipocyte-specific Adipoq-Cre was used, showed a nearly completely opposite phenotype to that of aP2-dTg mice. Further studies using models with more adipocyte-specific deletion of ROS would be warranted.

In conclusion, we demonstrated, through loss- and gain-of-function studies in vivo, that obesity-induced Fat ROS led to accelerated adipose inflammation, augmented fibrosis, and attenuated de novo lipogenesis in WAT, resulting in restricted healthy adipose expansion with increased ectopic lipid accumulation, eventually leading to deteriorated insulin sensitivity. Attenuation of de novo lipogenesis by oxidative stress was mediated by inhibition of SREBF1 transcriptional activities through the suppression of KDM1A protein expression (Fig. 8F). Because these unfavorable effects of Fat ROS could be reversed by the enrichment of antioxidants, such as Cat and Sod1, the development of agents that could increase antioxidants specifically in adipocytes could potentially be useful in the treatment of metabolic syndrome.

Acknowledgments. The authors thank Haruyo Sakamoto (Department of Metabolic Medicine, Osaka University Graduate School of Medicine) for technical support.

Funding. This work was supported by Japan Society for the Promotion of Science KAKENHI grant #24591329, a Japan Foundation for Applied Enzymology grant, and a Japan Health Foundation grant.

Duality of Interest. This work was supported by grants from Astellas Pharma, Inc.; Shionogi & Co., Ltd.; Kyowa Hakko Kirin Co., Ltd.; and Merck Sharp & Dohme. A.F. belongs to endowed department by Takeda Pharmaceutical Co.; Sanwa Kagaku Kenkyusho Co., Ltd.; Rohto Pharmaceutical Co., Ltd.; Fuji Oil Holdings, Inc.; and Roche DC Japan. S.K. belongs to endowed department by Kowa Pharmaceutical Co., Ltd.

The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author Contributions. Y.O. designed and performed the experiments, analyzed the data, and wrote the manuscript. A.F. interpreted the data and wrote the manuscript. E.H., H.K., and S.K. generated and maintained the mice. M.O. and I.S. supervised the project and wrote the manuscript. A.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.