Lgals9 deficiency ameliorates obesity by modulating redox state of PRDX2

The adipose tissue is regarded as an endocrine organ and secretes bioactive adipokines modulating chronic inflammation and oxidative stress in obesity. Gal-9 is secreted out upon cell injuries, interacts with T-cell immunoglobulin-3 (Tim-3) and induces apoptosis in activated Th1 cells. Gal-9 also binds to protein disulfide isomerase (PDI), maintains PDI on surface of T cells, and increases free thiols in the disulfide/thiol cycles. To explore the molecular mechanism of obesity, we investigated Gal-9−/− and Gal-9wt/wt C57BL/6J mice fed with high fat-high sucrose (HFHS) chow. Gal-9−/− mice were resistant to diet-induced obesity associated with reduction of epididymal and mesenteric fat tissues and improved glucose tolerance compared with Gal-9wt/wt mice. However, the number of M1, M2 macrophages, and M1/M2 ratio in epididymal fat were unaltered. Under HFHS chow, Gal-9−/− mice receiving Gal-9−/− or Gal-9wt/wt bone marrow-derived cells (BMCs) demonstrated significantly lower body weight compared with Gal-9wt/wt mice receiving Gal-9−/− BMCs. We identified the binding between Gal-9 and peroxiredoxin-2 (PRDX2) in sugar chain-independent manner by nanoLC-MS/MS, immunoprecipitation, and pull-down assay. In 3T3L1 adipocytes, Gal-9 knockdown shifts PRDX2 monomer (reduced form) dominant from PRDX2 dimer (oxidized form) under oxidative stress with H2O2. The inhibition of Gal-9 in adipocytes may be a new therapeutic approach targeting the oxidative stress and subsequent glucose intolerance in obesity.

The obesity is now pandemic in worldwide and the data from WHO demonstrated that more than 1.9 billion adults were overweight and over 650 million obese in 2016 (https:// www. who. int/ news-room/ fact-sheets/ detail/ obesi ty-and-overw eight). It substantially increases the risk of diseases such as type 2 diabetes, fatty liver disease, dyslipidemia and hypertension 1 . The adipose tissue has been regarded as an endocrine organ and it secretes bioactive adipokines leading to low grade chronic inflammation and amplification of oxidative stress 2 . In the obese patients, the chronic inflammation is demonstrated by the elevation of high-sensitive C-reactive protein 3 , interleukin (IL)-6 3 , IL-10, and tumor necrosis factor α (TNF-α) 4 , while the oxidative stress revealed by decreased antioxidant enzymes (catalase, glutathione peroxidase, superoxide dismutase) 5 or increased levels of malondialdehyde 3,5 , urinary 8-hydroxy-2′-deoxyguanosine 6 and lipid peroxidation markers (8-iso-prostaglandin F2α) 7 . By targeting inflammation and oxidative stress in obese patients, the interventions with foods and supplements in a randomized controlled trial have been conducted, such as Mediterranean diet with extra virgin olive oil 6 , pigmented rice 8 , Baru almonds 5 , and curcumin 3 . Although life-style modifications and intervention with foods and supplements have been vigorously attempted, the mechanism for the inflammation and oxidative stress in obesity is not fully explored.
The identification of Gal-9 interacting proteins in adipocytes. Bone marrow transplantation experiments suggested that the absence of Gal-9 in adipocytes may contribute the resistance to obesity and insulin resistance in Gal-9 −/− mice. To identify the Gal-9 interacting proteins in adipocytes, the immunoprecipitated protein complexes by monoclonal anti-Gal-9 antibody or isotype control isolated from epididymal fat pads in Gal-9 wt/wt mice were subjected to nanoLC-MS/MS analysis and analyzed by Mascot. Anti-Gal-9 antibody-and  Table 1d, e, respectively. Since the absence of intracellular Gal-9 may be involved in the amelioration of obesity, the intracellular proteins were firstly screened. The proteins found only in anti-Gal-9 antibody-precipitated sample were subjected to gene ontology analysis by PANTHER 15.0 (Supplementary Table 2). Gal-9, peroxiredoxin 2 (PRDX2), serine/threonine-protein kinase RIO2, ATP-citrate synthase, cytosolic phospholipase A2 zeta, and serine/threonine-protein kinase WNK1 were identified as cytosolic protein (Supplementary Table 2). The normalized emPAI of PRDX2 is highest and ranked top in these cytosolic proteins. In functional aspects,  www.nature.com/scientificreports/ mitochondrial oxidative stress in adipocytes is known to cause insulin resistance, and 6 mitochondrial proteins, electron transfer flavoprotein subunit alpha (ETFA), ferredoxin-2, 3-ketoacyl-CoA thiolase (ACAA2), longchain specific acyl-CoA dehydrogenase (ACADVL), acyl-CoA dehydrogenase family member 9 (ACAD9), and carnitine O-palmitoyltransferase 2 (CTP2), were also listed in the proteins found only in anti-Gal-9 antibodyprecipitated sample (Supplementary Table 1d and 2).
Gal-9 binds to PRDX2 and its dimer/monomer ratio is reduced by Gal-9 siRNA. Mitochondria are major source of oxidative stress in adipocytes under obesity and insulin resistance state, and peroxiredoxins are important for antioxidant responses against oxidative stress, we further investigated whether Gal-9 interacts with peroxiredoxins and mitochondrial proteins. Among peroxiredoxins 1, 2, 3, 4, and 5, Gal-9 was detected in the protein complexes immunoprecipitated by anti-PRDX2 antibody in the lysates isolated from epididymal adipose tissues of Gal-9 wt/wt mice ( Supplementary Fig. 2). Total lysate of 3T3L1 cells without plasmid transfection was applied to Anti-HA tag Beads and Gal-9 was detected by Western blot analysis ( Supplementary Fig. 3a, lane 4). However, the binding between Gal-9 and Beads was completely inhibited by the addition of 0.2 M lactose (Fig. 5a, lane 4, Supplementary Fig. 4a). Thus, following experiments were performed in the presence of 0.2 M lactose. PRDX2-FLAG-HA-pcDNA3.1 (PRDX2) and FLAG-HA-pcDNA3.1 (HA) were transfected into 3T3L1 cells, and total cell lysates were subjected to Anti-HA tag Beads isolation and Western blot analysis. Gal-9 and HA-tagged PRDX2 and were detected in protein complexes isolated by Anti-HA tag Beads ( Fig. 5b and Supplementary Fig. 4b). Although anti-thioredoxin (TRX) antibody cross-reacted to HA-tagged PRDX2, native TRX was not detected in the HA-tag purified protein complexes ( Fig. 5c and Supplementary Fig. 4c). The binding between Gal-9 and PRDX2 is sugar chain-independent, since the addition of 0.2 M lactose, which interrupts the binding between Gal-9 and β-galactoside sugar, did not inhibit the complex formation between Gal-9 and HA-tagged PRDX2. Two Cysteine residues of PRDX2 form disulfide-linked homodimers upon oxidation by hydroperoxides such as H 2 O 2 , and they were slowly reduced to their monomers by thioredoxin and thioredoxin oxidase system. Thus, PRDX2 dimer/monomer ratio well-reflects cytosolic redox status and higher ratios represent oxidative state in the cytosol. In 3T3L1 cells, Gal-9 siRNA reduced the protein expression of Gal-9 to 42.1 ± 3.2% compared with NC siRNA treated cells ( Fig. 5d and Supplementary Fig. 4d). Under the basal states without H 2 O 2 , PRDX2 dimer/monomer ratios were not altered in NC siRNA (2.12 ± 0.23) and Gal-9 siRNA (2.27 ± 0.02) treatments. However, in the presence of 10 μM H 2 O 2 , the dimer/monomer ratios were reduced from 2.38 ± 0.04 to 1.88 ± 0.10 by the treatments with Gal-9 siRNA (Fig. 5e and Supplementary Fig. 4e), suggesting that Gal-9 knockdown shifts the redox state to reducing conditions. We also investigated the oxidative stress signaling pathway such as p38 mitogen-activated protein kinase (p38-MAPK), stress-activated protein kinase (SAPK)/Jun amino terminal kinase (JNK), and p42/p44 mitogen-activated protein kinases [MAPK; extracellular signal-regulated kinase 2/1 (ERK2/1)]. Although the treatment with Gal9-siRNA partially reversed increased phospho/total SAPK/JNK and phospho/total ERK2/1 ratios by the oxidative stress under 10 μM H 2 O 2 , they did not reach the statistical differences ( Supplementary Fig. 5).

Discussion
Adipose tissue macrophages (ATMs) play key roles in the inflammation of adipose tissues and link to the development of insulin resistance in obesity. The cell surface differentiation markers for M1 macrophages are CD11c, CD44, CD163, CD172, while those for M2 are arginase 1, CD206, and CD301 28 . In the initial studies, the lean mice demonstrated M2 dominant phenotype in ATMs and the obese mice demonstrated M1 dominant phenotype, although mixed M1/M2 phenotypes were reported in later studies 29 . M1 ATMs were localized to crown-like structures surrounding adipocytes and released high levels of proinflammatory cytokines such as IL-6, IL-8, and TNF-α. Th1 cells are induced by IL-12 and IFN-γ, and secrete the proinflammatory cytokines IFN-γ and TNF-α, which promote the differentiation of M1 macrophages 28 . Gal-9 is abundantly expressed and remained in cytoplasm in steady states; however, it is secreted out as DAMPs or PAMPs and induces prominent apoptosis of Th1 and Th17 cells by binding to Tim-3 as a ligand. Thus, we initially hypothesized that the deficiency of Gal-9 in diet-induced obesity (DIO) mice would enhance the inflammation process in visceral adipose tissues associated with deteriorated obesity phenotype and insulin resistance. Unexpectedly, Gal-9 −/− DIO mice demonstrate improved obesity and insulin resistance phenotypes without alterations in adipose tissue inflammation and M1/M2 polarization. Similarly, unexpected results were observed in a pristane-induced lupus model in Gal-9 −/− BALB/c mice 30 . Since the injection of recombinant Gal-9 induced the apoptosis of T cells and ameliorated systemic lupus erythematosus (SLE) in MRL-lpr lupus-prone mice 20 , we thought that Gal-9 −/− BALB/c mice would demonstrate the aggravation of the disease. However, pristane treated Gal-9 −/− BALB/c mice were protected from nephritis, arthritis and peritoneal lipogranuloma formation 30 . The commercially available ELISA kits for Gal-9 detected the degradation products and measured concentrations were higher than actual concentrations of intact Gal-9. By the development of specific ELISA for whole and intact Gal-9, the concentration in healthy human subjects was 110 pg/mL (67-154 pg/mL) 31 . Although Gal-9 is abundantly and ubiquitously expressed in various organs, the physiological plasma concentration of Gal-9 is very low. The apoptotic potential of Gal-9 was mainly demonstrated by the application of recombinant Gal-9 protein and it required much higher doses at pharmacological range above the physiological concentrations. Furthermore, the cross-link and lattice formation of the cell-surface glycoproteins, and the subsequent intracellular signaling by Gal-9 may be brought by the optimum concentrations of Gal-9, but not by excess amount of Gal-9 32,33 . Thus, the Gal-9 released from adipocytes in obesity may not be enough to demonstrate the potential to induce the apoptosis Th1 cells in adipose tissues.
In adipose tissues in DIO mice, the expression of Gal-1, Gal-9 in subcutaneous adipose tissues (SAT), and Gal-3 in SAT and visceral adipose tissues (VAT) were progressively increased. In contrast, Gal-12 declined  34 . Although these galectins were expressed in both mature adipocytes and SVF, Gal-1 increased in adipocytes, Gal-3 and Gal-9 increased in SVF, whereas Gal-12 was dominant in adipocytes 34 . Gal-1 34 ablation resulted in increased adiposity with impaired glucose metabolism and systemic inflammation. Gal-3 deficiency was reported to link to increased adiposity and dysregulated glucose metabolism 35 in the initial study; however, Gal-3 deletion resulted in the improvement of insulin resistance and hematopoietic-derived Gal-3 was shown to cause cellular and systemic insulin resistance 36 in the later study. The ablation of Gal-12 known as a negative regulator of lipolysis resulted in increased mitochondrial respiration, reduced adiposity and increased insulin sensitivity 37 , while the ablation of Gal-12 was also characterized by the M2 polarization associated with reduced insulin sensitivity in cultured macophages 38 . The role of secreted Gal-9 in extracellular milieu has been extensively investigated by focusing on the modulation of immune function, i.e. the regulation of apoptosis and immune checkpoint; however, the cytosolic function of Gal-9 remains unexplored especially in the aspect of redox state of the cells 39 . It is well known that galectins rapidly lose sugar-chain binding activities if they are not kept in reducing buffers because of cross-linking and oxidation of cysteine residues or tryptophan residue in the carbohydrate recognition domain. The labile galectins * * * * * * *   www.nature.com/scientificreports/ are Gal-1 and Gal-2, while most others, such as Gal-3 and Gal-4, are more stable in the absence of reducing conditions. Since galectins are synthesized in the cytosol, which is highly reducing environment, sugar-chain binding activities are maintain in the cytosol. The oxidation of Gal-1 promoted the formation of the Cys16-Cys88 disulfide bond, and multimers through Cys2 and oxidized Gal-1 did not bind to lactose 40 . Interestingly, oxidized form of Gal-1 without lectin activity promotes the axonal regeneration, not in the reduced form 41 . Similarly, the oxidation of Gal-2 by H 2 O 2 resulted in the loss of lectin activity, while treatment of Gal-2 with S-nitrosocysteine prevented H 2 O 2 -induced inactivation 42 . In contrast to Gal-1 and Gal-2, Gal-3 is actively involved in the regulation of redox state in the patients with aortic stenosis (AS) and animal model of cardiac damage. Gal-3 was up-regulated in myocardial biopsy from AS patients and negatively correlated with the expression of PRDX4 43 . The inhibition Gal-3 with modified citrus pectin (MCP) exhibited dramatic improvement in cardiac function of the doxorubicin-treated rats by upregulating anti-oxidant PRDX4 44 .
Although the involvement of Gal-9 in redox regulation had been totally unknown, the role Gal-9 in cell surface redox control of T cells was reported 25 . Thioredoxin and protein disulfide isomerase (PDI) are members of PDI family. PDI cleaves disulfide bonds by thiol/disulfide exchange on the plasma membrane, while PDI in the endoplasmic reticulum operates as an oxidase, creating disulfide bonds 25 . Exogenous Gal-9 binds to PDI via O-glycan, maintains the retention of PDI on surface of T cells, and increases free thiols in the disulfide/thiol cycles 45,46 . In our experiments, we firstly demonstrated that Gal-9 binds to PRDX2 independent of sugar-binding activity of Gal-9, since the binding was not cancelled by the addition of lactose. Two pairs of cysteine residues in PRDX2 form disulfide-linked homodimers in the presence of H 2 O 2 , and they return to their monomers by thioredoxin and thioredoxin oxidase system 47 . Higher PRDX2 dimer/monomer ratio indicates the oxidative stress and is a sensitive sensor for redox states in the cytosol. In 3T3L1 adipocytes, the knockdown of Gal-9 resulted in lower PRDX2 dimer/monomer ratio in the presence of H 2 O 2 , suggesting the amelioration of oxidative stress in cytosol. Furthermore, it can be speculated that the disruption of the binding between Gal-9 and PRDX2 by knocking down Gal-9 resulted in the facilitation of dimer to monomer conversion by thioredoxin ( Supplementary Fig. 6). The oxidative stress induces the insulin resistance associated with increased levels of inflammatory cytokines, such as leptin, MCP-1, IL-6, and TNF-α, and recued levels of adiponectin 48 . In Gal-9 −/− C57BL/6J mice, they were resistant to diet-induced obesity and the absence of Gal-9 ameliorated the oxidative stress by shifting lower dimer/monomer ratio of PRDX2 in 3T3L1 adipocytes.
In conclusion, Gal-9 −/− C57BL/6J mice fed with HFHS chow were resistant to DIO without the alterations in production of inflammatory cytokines, M1/M2 macrophage polarization, and formation of crown-like structure. Although recombinant Gal-9 is known to induce the apoptosis of Th1 cells and reduction of M1 macrophages, the reduction of body weight was independent of BMCs revealed by BMT experiments. Instead, we found that Gal-9 binds to PRDX2 and demonstrated that Gal-9 knockdown ameliorated the oxidative stress and reduced dimer/monomer ratio of PRDX2 in adipocytes. The inhibition of Gal-9 in adipocytes may be a new therapeutic approach targeting the oxidative stress and subsequent insulin resistance in obesity.
Gal-9 −/− BALB/c mice were crossed to C57BL/6JJcl mice (CLEA Japan, Tokyo, Japan) for 10 generations. Gal-9was detected by PCR primers 5′-GCG AGG CCA GAG GCC ACT TGT GTA GC-3′ and GTG ACA ATA CTG TTC CTC TGC AGG -3′, while Lgals9 wt (Gal-9 wt ) by 5′-TGG GGT GTC CTG CAG ACA GCA CAT AA-3′ and 5′-CCA GTG CTA CGG CGA CAT AGC CTC -3′. By crossing Gal-9 wt/-C57BL/6JJcl mice, we produced Gal-9 wt/wt , Gal-9 wt/-, and Gal-9 −/− littermates by standard breeding techniques. Gal-9 wt/wt and Gal-9 −/− mice were used for the following experiments. The 6-week-old male mice were fed with standard chow diet (STD) (MF, Oriental Yeast, Japan) or high fat-high sucrose diet (HFHS) (D12331, Research Diets, New Brunswick, NJ), and they were euthanized at 28 weeks of age. VȮ 2 and RQ were continuously monitored for 24 h by using O 2 /CO 2 metabolism measuring system (MK-5000, Muromachi, Kyoto, Japan). The locomotor activity was recorded for 24 h by the frequency of interrupting an infrared sensor (ACTIMO-100 N, SHINFACTORY, Fukuoka, Japan). IPGTT was performed at 20-22 weeks of age after 12 h fasting, blood glucose and plasma insulin levels were measured (Skylight Biotech, Tokyo, Japan). At 28 weeks of age, sera and tissue samples were collected and weighed after 12 h fasting. Serum leptin, adiponectin (ELISA), cholesterol, triglyceride, non-esterified free fatty acid (HPLC), liver cholesterol and liver triglyceride (Folch's method) were measured by Skylight Biotech, Tokyo. All animal experiments were approved by the Animal Care and Use Committee of the Department of Animal Resources, Advanced Science Research Center, Okayama University. All animal experiments were performed in accordance with relevant guideline and regulations.
Immunofluorescence. Epididymal adipose tissues were first stained with rat anti-mouse F4/80 and rabbit anti-mouse perilipin, and subsequently stained with goat anti-rat IgG (Alexa Fluor 488) and donkey anti-rabbit IgG (Alexa Fluor 555), respectively (Invitrogen). Nuclear stain was performed by DAPI. www.nature.com/scientificreports/ Flow cytometry analysis. Mature adipocytes and stromal vascular fractions (SVF) were isolated as described below. Epididymal fat tissues were minced and incubated in a fresh digesting media of Krebs Ringer HEPES (KRH) buffer for 60 min at 37 °C and were separated into adipocytes and SVF by using mesh with a grid diameter 300 µm. The SVF cells were subjected flow cytometry analysis as previously described 30  . Dead cells were excluded from analysis using 7-aminoactinomycin D staining (BD pharmigen). All data were acquired with FACSAria I flow cytometer (BD Biosciences) and analysed using FlowJo software (TreeStar, Ashland, OR).

3T3L1 culture and siRNA experiments. 3T3L1 cells (ATCC) were cultured in Dulbecco's Modified
Eagle's Medium (Thermo Fisher Scientific) containing 10% fetal bovine serum and they were transfected with Silencer select Pre-designed siRNA Lgals9 (Gal-9 siRNA) and silencer select negative control siRNA (NC siRNA) (Thermo Fisher Scientific, Cat#s69189) by Lipofectamine RNAi MAX (Cat#13,778,030). After 40 h, the 3T3L1 cells were further cultured for 20 minutes in the absence and presence of 10 μM H 2 O 2 .

Statistical analysis.
All results are expressed as means ± standard deviation (SD). Normal distribution was confirmed by Shapiro-Wilk test. Only plasma insulin concentration and percentage of M1 and M2 macrophage did not follow the normal distribution. For parametric analyses, the multiple comparisons were performed by one-way ANOVA with Tukey-Kramer method and two-pair comparisons by Student's t test using SPSS software (IBM, Chicago, IL). For non-parametric analyses, the multiple comparisons were performed by Kruskal-Wallis test with Bonferroni correction. A value of p < 0.05 was regarded as statistically significant.