Sulforaphane increases metabolism to ameliorate obesity and insulin resistance

Study published here: Glucoraphanin ameliorates obesity and insulin resistance through adipose tissue browning and reduction of metabolic endotoxemia in mice

Low-grade sustained inflammation, triggered by chronically high levels of proinflammatory cytokines and gut microbiota-derived circulatory lipopolysaccharide (LPS), links obesity with comorbidities such as insulin resistance and nonalcoholic fatty liver disease (NAFLD) (1,2).

Although a number of pharmacological treatments for obesity and NAFLD have been tested, few drugs are clinically available owing to the lack of long- term efficacy and safety concerns (3,4). Thus, a novel therapeutic approach that would improve energy metabolism and reduce chronic inflammation in obesity is sorely needed.

Nuclear factor (erythroid-derived 2)-like 2 (Nrf2), a basic leucine zipper transcription factor, is widely expressed in human and mouse tissues, and serves as a defense response against extrinsic and intrinsic stressors (5). Upon exposure to electrophilic and oxidative stress, Nrf2 detaches from its repressor, Kelch-like ECH-associated protein 1-nuclear factor (Keap1), and is translocated from the cytoplasm into the nucleus.

This translocation leads to the transcriptional activation of genes encoding phase 2 detoxifying and antioxidant enzymes (6).

In addition to the ubiquitous induction of cytoprotective genes, Nrf2 regulates a large number of genes involved in glucose and lipid metabolism. In the liver, the constitutive activation of Nrf2 via Keap1 knockdown represses the expression of genes involved in gluconeogenesis (7) and lipogenesis (8), thereby alleviating obesity, diabetes, and hepatic steatosis.

Accordingly, synthetic Nrf2 inducers such as synthetic triterpenoid 2-cyano-3,12-dioxoolean-1,9-dien-28-oic acid (CDDO)-imidazolide (9), CDDO-methyl ester (known as bardoxolone methyl) (10), and dithiolethione analog, oltipraz (11), have been shown to ameliorate high-fat diet (HFD)-induced obesity and diabetes.

These synthetic Nrf2 inducers also decrease liver and adipose tissue lipogenesis, and enhance glucose uptake in skeletal muscles. However, the mechanisms by which Nrf2 enhances energy metabolism in response to a HFD remain largely unknown.

Although enhanced Nrf2 signaling has shown promising results in several animal studies, the synthetic Nrf2 inducers have caused adverse cardiac events and gastrointestinal toxicities in clinical trials (12,13).

These observations prompted us to explore a safer Nrf2 inducer for the treatment of obesity, insulin resistance, and NAFLD.

Sulforaphane, an isothiocyanate derived from cruciferous vegetables, is one of the most potent naturally occurring Nrf2 inducers; this compound exhibits anticancer activity in cancer cell lines and in carcinogen-induced rodent models (14).

Among the cruciferous vegetables, broccoli sprouts are the best source of glucoraphanin, a stable glucosinolate precursor of sulforaphane (15). In both rodents and humans, glucoraphanin is hydrolyzed by gut microbiota-derived myrosinase into bioactive sulforaphane prior to intestinal absorption (16). A recent clinical study demonstrated the safety of orally administered glucoraphanin (17).

In the present study, we examined the dietary glucoraphanin-mediated modulation of systemic energy balance and the mitigation of chronic inflammation, insulin resistance, and NAFLD in diet-induced obese mice.

Research Design and Methods Glucoraphanin preparation
The sulforaphane precursor, glucoraphanin, was prepared as previously described (17) with minor modifications. Briefly, one day after germination from broccoli seeds (Caudill Seed Company, Louisville, KY), sprouts were boiled in water for 30 min.

The water extract was mixed with dextrinized cornstarch and subsequently spray-dried to yield an extract powder containing 135 mg of glucoraphanin per gram (0.31 mmol/g) (18). The total glucoraphanin titer in the resulting powder was determined by HPLC as previously reported (19).

Mice and diets
Male C57BL/6JSlc mice were purchased from Japan SLC (Hamamatsu, Japan) at 7 weeks of age. Nrf2-knockout (Nrf2-/-) mouse strain (RBRC01390; C57BL/6J background) was provided by RIKEN BRC (Tsukuba, Japan) (6). After a week of acclimation, mice were fed normal chow (NC; containing 2.2% dextrinized cornstarch, 10% kcal from fat, Research Diets, New Brunswick, NJ), NC containing 0.3% glucoraphanin (NC- GR; containing 2.2% extract powder), a high-fat diet (HFD; containing 2.2% dextrinized cornstarch, 60% kcal from fat, #D12492; Research diets), or a HFD containing 0.3% glucoraphanin (HFD-GR; containing 2.2% extract powder) for 14 weeks.

Both the NC and the HFD containing cornstarch or glucoraphanin were prepared by Research Diets. All mice studied were maintained on a 12 h light/dark cycle at 24–26°C with free access to water and food. All animal procedures were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals at Kanazawa University, Japan.

Indirect calorimetry
After 3 weeks of feeding, mice were individually housed in an indirect calorimeter chamber at 24–26°C (Oxymax; Columbus Instruments, Columbus, OH). Calorimetry, daily body weight, and daily food intake data were acquired during a 3-day acclimation period, followed by a 2-day experimental period. Oxygen consumption (VO2) and carbon dioxide production (VCO2) were measured in each chamber every 20 minutes. The respiratory exchange ratio (RER = VCO2/VO2) was calculated using Oxymax software. Energy expenditure was calculated as shown below and normalized to the body mass of each subject.
Energy expenditure = VO2 × [3.815 + (1.232 ×RER)]

Metabolic measurements and biochemical analyses
Metabolic parameters, body fat composition, insulin sensitivity, and glucose tolerance were assessed as previously described (20). Plasma LPS levels were analyzed using a Limulus amebocyte lysate assay kit (QCL-1000; Lonza, Allendale, NJ). Plasma LPS binding protein (LBP) levels were determined using an ELISA kit (Enzo Life Sciences, Farmingdale, NY).

Immunoblotting was performed with primary antibodies (Supplementary Table 1) as previously described (20). mRNA expression levels were determined by quantitative real-time PCR using SYBR Green with the primers (Supplementary Table 2) as previously described (20).

Isolation and differentiation of inguinal white adipose tissue-derived primary beige adipocytes
Stromal vascular fractions (SVFs) from inguinal white adipose tissue (WAT) of 7- week-old wild-type and Nrf2-/- mice were prepared as previously reported (21). At confluence, SVF cells were induced for 2 days with differentiation medium containing DMEM/F-12 supplemented with 10% FBS, 20 nM insulin, 1 nM T3, 5 μM dexamethasone, 500 μM isobutylmethylxanthine, 125 μM indomethacin, and 0.5 μM rosiglitazone (all from Sigma-Aldrich, St. Louis, MO).

Induced cells were subsequently cultured in maintenance medium (DMEM/F-12 containing 10% FBS, 20 nM insulin, and 1 nM T3) for 5 days and treated with DMSO or sulforaphane (Toronto Research Chemicals, Toronto, Canada) at the indicated concentrations for 48 h.

Fluorescence-activated cell sorting (FACS)
Cells from the liver and epididymal WAT were prepared as previously described (22). Isolated cells were incubated with Fc-Block (BD Bioscience, San Jose, CA), and subsequently incubated with fluorochrome-conjugated antibodies (Supplementary Table 1).

Flow cytometry was performed using a FACSAria II (BD Bioscience), and the data were analyzed using FlowJo software (Tree Star, Ashland, OR).
Analysis of gut microbiota via pyrosequencing of the 16S rRNA gene
Metagenomic DNA was extracted from mouse cecal content with a QIAamp DNA stool kit (Qiagen, Hilden, Germany). The V1–V2 region of the 16S rRNA gene was amplified using primer sets as previously reported (23).

Mixed samples were prepared by pooling approximately equal amounts of PCR amplicons from each sample and subjected to a GS Junior System (Roche Diagnostics, Basel, Switzerland) for subsequent 454 sequencing.

Pre-processing and taxonomic assignment of sequencing reads were conducted as described previously (23) and separated by unique barcodes. The 16S rRNA sequence database was constructed by retrieving 16S sequences of bacterial isolates (1200–2384 bases in length) from the Ribosomal Database Project Release 10.27.

We used 4000 filter- passed reads of 16S sequences for the operational taxonomic unit (OTU) analysis of each sample. Clustering of 16S sequence reads with identity scores > 96% into OTUs was performed using UCLUST (

Representative sequences with identity scores > 96% for each OTU were assigned to bacterial species using BLAST. Principal component analysis using EZR software ( sct/SaitamaHP.files/statmedEN.html) was applied for assessment of alterations of cecal bacterial phylum associated with diets.

Statistical analyses
Data were expressed as mean ± SEM. P < 0.05 was considered statistically significant. Statistical differences between pairs of groups were determined by a two-tailed Student’s t-test. An overall difference between more than two groups was determined using a one-way ANOVA.

If one-way ANOVAs were significant, differences between individual groups were estimated using a Bonferroni post hoc test All calculations were performed using SPSS Statistics (v19.0, IBM, Armonk, NY).


Glucoraphanin decreases weight gain and adiposity, and increases energy expenditure in high-fat diet (HFD)-fed mice
To investigate the effects of glucoraphanin on systemic energy balance, we examined the body weight of wild-type mice fed normal chow (NC) or a high-fat diet (HFD), supplemented with glucoraphanin or vehicle (i.e., cornstarch only). Glucoraphanin reduced weight gain only in HFD-fed mice without affecting food intake (Fig. 1A and Supplementary Fig. 1A).

This reduction was not accompanied by evidence of gross toxicity. We determined the plasma concentration of sulforaphane in NC-GR and HFD-GR mice, but not in NC or HFD mice, indicating that glucoraphanin was absorbed as a sulforaphane following food consumption (Supplementary Fig. 1B).

The reduction of weight gain in glucoraphanin-treated HFD-fed mice was largely attributed to the decreased fat mass but not lean mass (Fig. 1B). To assess energy expenditure, we placed the mice in indirect calorimetry cages after 3 weeks of feeding, before an evident change in the body mass of HFD-fed mice was observed (HFD: 29.9 ± 0.5 g vs. HFD-GR: 28.5 ± 0.5 g).

Glucoraphanin-treated mice fed the HFD exhibited consistently higher VO2 and VCO2 than vehicle-treated HFD-fed controls (Fig. 1C and D), leading to increased energy expenditure (Fig. 1E); however, they displayed similar RER (Fig. 1F), suggesting that glucoraphanin supplementation enhanced sugar and fat use under HFD conditions. On NC-fed mice, glucoraphanin did not affect these parameters of energy balance (Fig. 1B–F).

Consistent with increased energy expenditure, glucoraphanin increased the core body temperature of HFD-fed mice by approximately 0.5°C (Fig. 1G).

Glucoraphanin improves diet-induced insulin resistance and glucose tolerance
After 14 weeks of feeding, glucoraphanin supplementation did not affect plasma triglycerides, total cholesterol, and free fatty acid (FFA) levels in either NC- or HFD-fed mice (Table 1). On NC, blood glucose levels were not altered by glucoraphanin, but on the HFD, glucoraphanin-treated mice exhibited significantly lower fasted blood glucose compared with vehicle-treated controls (Table 1).

Additionally, glucoraphanin significantly decreased plasma insulin concentrations in HFD-fed mice under both fasted and fed conditions, resulting in lower homeostatic model assessment-insulin resistance (HOMA-IR; Table 1).

During the insulin tolerance test (ITT), glucoraphanin significantly enhanced the reduction in blood glucose levels in HFD-fed mice, but not in NC-fed mice compared with the vehicle-treated controls (Fig. 2A). Glucoraphanin improved glucose tolerance in HFD- fed mice during the glucose tolerance test (GTT), but had no effect on NC-fed mice (Fig. 2B).

Insulin secretion during the GTT was not affected by glucoraphanin (data not shown). In line with increased insulin sensitivity, insulin-stimulated Akt phosphorylation on Ser473 was enhanced by glucoraphanin in the liver, muscle, and epididymal WAT of mice fed the HFD (Fig. 2C).

Glucoraphanin does not exert anti-obesity and insulin-sensitizing effects in Nrf2-/- mice
Although the Keap1-Nrf2 pathway is a well-known target of sulforaphane, this isothiocyanate has also been reported to modulate different biological pathways independent of the Keap1-Nrf2 pathway (24,25).

To determine whether the anti-obesity and insulin-sensitizing effects of glucoraphanin are mediated through Nrf2, the effects of glucoraphanin on energy balance and glucose metabolism were assessed in NC- and HFD- fed Nrf2-/- mice.

Although food intake and plasma concentration of sulforaphane in NC-GR or HFD-GR diet-fed Nrf2-/- nice were comparable that detected in wild-type mice fed NC- GR or HFD-GR diet (Supplementary Fig. 1C and D), the effects of glucoraphanin following HFD feeding on weight gain (Fig. 3A), VO2 (Fig. 3B), VCO2 (Fig. 3C), energy expenditure (Fig. 3D), RER (Fig. 3E), rectal temperature (Fig. 3F), insulin sensitivity (Fig. 3G), and glucose tolerance (Fig. 3H) were abolished by the Nrf2 deficiency.

These data are consistent with comparable plasma metabolic parameters between glucoraphanin-treated and vehicle-treated mice on the HFD. These metabolic parameters include lipids, blood glucose, insulin, HOMA-IR, and liver enzymes such as alanine transaminase (ALT) and aspartate transaminase (AST) (Supplementary Table 3).

Glucoraphanin blocks HFD-induced reduction of Ucp1 expression in WAT of wild- type mice but not in Nrf2-/- mice

The increased energy expenditure and body temperature of glucoraphanin-treated HFD-fed mice suggest an increase in adaptive thermogenesis. However, glucoraphanin supplementation had little effect on the size and number of lipid droplets in the intrascapular brown adipose tissue (BAT) of HFD-fed wild-type mice (Supplementary Fig. 2A).

In addition, the mRNA expression of uncoupling proteins (Ucps), PGC-1α, and deiodinase 2 in BAT and of Ucps in skeletal muscle was not altered by glucoraphanin supplementation in both NC- and HFD-fed wild-type mice (Supplementary Fig. 2B and C). In BAT, HFD increased Ucp1 protein expression, but glucoraphanin did not alter the expression in both wild-type and Nrf2-/- mice (Fig. 4A).

Brown-like adipocytes expressing Ucp1, also known as beige cells, exist in various WAT depots and can contribute to thermogenesis (26). Compared with NC, HFD significantly decreased Ucp1 protein levels in epididymal and inguinal WAT of both wild-type and Nrf2-/- mice (Fig. 4A).

Glucoraphanin supplementation restored HFD-induced reduction in Ucp1 protein levels in epididymal and inguinal WAT of wild-type mice but not those in Nrf2-/- mice.

To examine whether the effects of glucoraphanin were fat cell-autonomous and Nrf2-mediated, we tested the effects of sulforaphane, an active metabolite of glucoraphanin, on the expression of brown fat-selective genes in primary beige adipocytes obtained from inguinal WAT of wild-type and Nrf2-/- mice.

In beige adipocytes derived from wild-type mice, treatment with sulforaphane induced the Nrf2 target gene, NAD(P)H:quinone oxidoreductase 1 (Nqo1; Fig. 4B) and antioxidant genes (Supplementary Fig. 3A).

Concurrently, sulforaphane significantly increased the mRNA expression of brown-fat selective genes, including Ucp1, Prdm16, Cidea, and Elovl3 (Fig. 4B).

In contrast, in Nrf2-deficient beige adipocytes, sulforaphane failed to activate Nrf2 as judged by unaltered mRNA expression of the target genes and to promote the expression of brown-fat selective genes (Supplementary Fig. 3B and Fig. 4C).

Importantly, Nrf2-deficient beige adipocytes exhibited less differentiation levels associated with attenuated lipid accumulation (Supplementary Fig. 3C) and lower mRNA expression of fatty acid binding protein 4 (Supplementary Fig. 3C) and brown-fat selective genes compared with wild-type beige adipocytes (Fig. 4C).

Glucoraphanin reduces hepatic steatosis and oxidative stress in HFD-fed mice
The HFD causes hepatic steatosis and inflammation, eventually leading to steatohepatitis. As shown in Fig. 5A, the increase in liver weight caused by the 14-week- HFD was alleviated by glucoraphanin supplementation.

Glucoraphanin also attenuated HFD-induced hepatic steatosis (Fig. 5B). Additionally, compared with the HFD group, the lower levels of plasma ALT, plasma AST, liver triglycerides, and liver FFAs in the HFD- GR group indicate that glucoraphanin alleviated HFD-induced liver damage (Fig. 5C and D).

The reduction in hepatic steatosis was accompanied by the decreased expression of the following lipogenic genes: sterol regulatory element binding protein-1c (Srebf1), fatty acid synthase (Fasn), and peroxisome proliferator-activated receptor gamma (Pparγ) (Fig. 5E).

Additionally, hepatic levels of malondialdehyde, a marker of lipid peroxidation, were increased by the HFD. Glucoraphanin attenuated lipid peroxidation (Fig. 5F) and decreased gene expression of the NADPH oxidase subunits gp91phox, p22phox, p47phox, and p67phox (Fig. 5E).

The HFD led to a compensatory increase in the expression of genes involved in fatty acid β-oxidation (Ppara and Cpt1a) and anti-oxidative stress (Cat, Gpx1, and Sod1) in the liver. However, glucoraphanin did not increase the expression of these genes in the liver of HFD-fed mice further (Fig. 5E).

Glucoraphanin suppresses HFD-induced proinflammatory activation of macrophages in liver and adipose tissue
In response to the HFD, liver-resident macrophages (Kupffer cells) increase the production of proinflammatory cytokines that promote insulin resistance and NAFLD in mice (27). In particular, chemokine (C-C motif) ligand 2 (Ccl2) promotes the recruitment of chemokine (C-C motif) receptor 2 (Ccr2)-positive monocytic lineages of myeloid cells into the liver (28). These recruited cells produce a large amount of proinflammatory mediators and activate a lipogenic program (28).

Here, we found a prominent induction of tumor necrosis factor-α (Tnf-α), Ccl2, and Ccr2 in the liver of HFD-fed mice, which was markedly reduced in glucoraphanin-treated mice (Fig. 6A).

Glucoraphanin significantly suppressed HFD-induced inflammatory pathways such as c-Jun N-terminal kinase (JNK) and extracellular signal-regulated kinase (Erk) (Fig. 6B).

Glucoraphanin tended to decrease levels of p-NF-κB p65 (Ser536) in HFD-fed mice, although this decrease was not statistically significant (Fig. 6B). Of note, in the liver of Nrf2-/- mice, glucoraphanin failed to suppress HFD-induced inflammatory signal pathways (Supplementary Fig. 4).

In addition, glucoraphanin significantly decreased the HFD-induced hepatic expression of macrophage markers, including F4/80, Cd11b, and Cd68 (Fig. 6C). Tissue macrophages are phenotypically heterogeneous and have been characterized according to their activation/polarization state as M1-like proinflammatory macrophages or M2-like anti- inflammatory macrophages (29).

Consistent with the decreased expression of macrophage markers, glucoraphanin prevented macrophage (F4/80+CD11b+ cell) accumulation in the liver of HFD-fed mice (Fig. 6D).

Additionally, glucoraphanin decreased the number of M1- like liver macrophages expressing surface markers (F4/80+CD11b+CD11c+CD206−) (Fig. 6E).

In contrast, glucoraphanin increased the number of M2-like liver macrophages (F4/80+CD11b+CD11c−CD206+), resulting in a predominantly M2-like macrophage population (Fig. 6E).

Moreover, glucoraphanin decreased the mRNA expression of Tnf-α and NADPH oxidase in the epididymal WAT of HFD-fed mice (Supplementary Fig. 5A).

Although the HFD-induced expression of macrophage markers and macrophage accumulation in epididymal WAT were not altered by glucoraphanin (Supplementary Fig. 5B and C), the number of M1-like macrophages was significantly decreased in the epididymal WAT of glucoraphanin-treated HFD-fed mice (Supplementary Fig. 5D).

Glucoraphanin decreases circulating LPS and the relative abundance of Proteobacteria in the gut microbiomes of HFD-fed mice

Gut microbiota-derived LPS induces chronic inflammation that eventually leads to insulin resistance in obesity, termed metabolic endotoxemia (1,2).

Based on our observation that glucoraphanin alleviates inflammation in the liver and epididymal WAT of HFD-fed mice, we subsequently investigated the effects of glucoraphanin on metabolic endotoxemia and gut microbiota.

In accordance with previous studies (1,2), the HFD induced a 2-fold increase in circulatory LPS levels, which was reduced by glucoraphanin supplementation (Fig. 7A). Furthermore, plasma and hepatic levels of the LPS marker, LBP, were significantly elevated by the HFD and were reduced by glucoraphanin supplementation (Fig. 7B).

A principal component analysis distinguished cecal microbial communities based on diet and treatment, revealing that the metagenomes of HFD-fed mice formed a cluster distinct from the cluster formed by NC-fed mice (Fig. 7C). However, samples from glucoraphanin-treated HFD-fed mice formed a cluster that was indistinguishable from that of vehicle- or glucoraphanin-treated NC-fed mice (Fig. 7C).

Importantly, consistent with previous reports (30,31), further analysis at the phylum level demonstrated that the proportion of Gram-negative Proteobacteria was significantly elevated in the gut microbiomes of HFD-fed mice, which was suppressed by glucoraphanin supplementation (Fig. 7D and E).

The increase in the relative abundance of Proteobacteria in HFD-fed mice is mostly explained by an increase in the relative abundance of bacteria from the family Desulfovibrionaceae (Fig. 7F), key producers of endotoxins in animal models of obesity (30).

In fact, the relative abundance of Desulfovibrionaceae was positively correlated with plasma LPS levels (Fig. 7G). Furthermore, plasma LPS levels were significantly and positively correlated with the hepatic mRNA levels of Tnf-α, gp91phox, and F4/80 (Fig. 7H).

Similarly, the liver expression of other marker genes was both significantly and positively correlated with plasma LPS levels and with one another (Supplementary Table 4).

In the present study, we demonstrated that glucoraphanin, a stable precursor of the Nrf2 inducer sulforaphane, mitigated HFD-induced weight gain, insulin resistance, hepatic steatosis, oxidative stress, and chronic inflammation in mice.

The weight-reducing and insulin-sensitizing effects of glucoraphanin were abolished in Nrf2-/- mice. Additionally, glucoraphanin lowered plasma LPS levels in HFD-fed mice, and decreased the relative abundance of Desulfovibrionaceae.

At the molecular level, glucoraphanin increased Ucp1 protein expression in WAT depots, while suppressing the hepatic mRNA expression of genes involved in lipogenesis, NADPH oxidase, and inflammatory cytokines.

Our data suggest that in diet-induced obese mice, glucoraphanin restores energy expenditure and limits gut-derived metabolic endotoxemia, thereby preventing hepatic steatosis, insulin resistance, and chronic inflammation.

Consistent with previous reports demonstrating the anti-obesity effects of synthetic Nrf2 inducers (9–11), we show that the oral administration of glucoraphanin mitigates HFD-induced weight gain (Fig. 1A). The dose of glucoraphanin used in the current study (about 12 μmol/mouse/day) is similar to that used in other experiments investigating its antitumor effects in mice (14,32,33).

Here, we show that the effect of glucoraphanin on whole-body energy expenditure and the protein expression of Ucp1 in WAT were abolished in Nrf2-/- mice (Fig. 3D and 4A).

A recent study using adipocyte-specific PRDM16- deficient mice indicated that adaptive thermogenesis in beige fat also contributes to systemic energy expenditure (26). The mutant mice in the aforementioned study, which exhibited markedly reduced Ucp1 mRNA expression in inguinal WAT and minimal effects on BAT, developed obesity and insulin resistance in response to a HFD.

Thus, we believe that the increased energy expenditure in glucoraphanin-treated HFD-fed mice stems from, at least in part, an increase in beige fat, even though the expression of Ucps in BAT and skeletal muscle is not altered. Further analysis using Ucp1-knockout mice will elucidate the relative contribution of beige fat to the Nrf2-mediated metabolic effects elicited by glucoraphanin.

Several studies indicated that Nrf2-/- mice are partially protected from HFD-induced obesity and associated with milder insulin resistance compared with wild-type counterparts (9,34,35). Recently, Schneider et al. demonstrated that mitigation of HFD-induced obesity in Nrf2-/- mice, which was 25% less body weight than that of wild-type mice after 6 weeks of feeding (35).

They also found that HFD-fed Nrf2-/- mice exhibit a 20–30% increase in energy expenditure that is associated with an approximately 3-fold up-regulation of Ucp1 protein expression in abdominal WAT (35). In the present study, Nrf2-/- mice gained less weight after 6 weeks of HFD feeding compared with HFD-fed wild-type mice (Figure 1A and 3A; Nrf2-/-: 35.9 ± 0.9 g vs. wild-type: 39.0 ± 0.8 g, P < 0.05).

The lower body mass of Nrf2-/- mice raises the possibility that anti-obesity effect of glucoraphanin was completely phenocopied by Nrf2 gene deficiency. However, several observations suggest that Nrf2-/- mice only partially phenocopy the effect of glucoraphanin on weight gain reduction. 1)

In the present study, the weight difference between Nrf2-/- and wild-type mice was only 8%, which is much less than that in the previous study (35). 2) In Nrf2-/- mice, compared with NC, HFD induced significant weight gain (Fig. 3A), glucose intolerance (Fig. 3H), and insulin resistance as judged by increased HOMA-IR (Supplementary Table 3). 3)

Metabolic rate and energy expenditure of Nrf2-/- mice were comparable with those in wild-type mice (Fig. 1C–E and 3B–D). 4) Ucp1 protein levels in both epididymal WAT and inguinal WAT of HFD-fed Nrf2-/- mice were lower than those in HFD-GR-fed wild-type mice (Fig. 4A).

Taken together, these findings suggest that Nrf2 gene deficiency is not sufficient to block HFD-induced obesity by increasing energy expenditure and Ucp1 expression in WAT depots, and to mask the effect of glucoraphanin.

However, we cannot fully exclude the possibility that the effects of glucoraphanin are mediated by Nrf2 independent-mechanisms. Possible reasons for the discordance in metabolic phenotypes of Nrf2-/- between the previous study (35) and ours may be due to differences in knockout mouse lines and experimental conditions (e.g. age of mice at beginning of HFD feeding, composition of HFD, and temperature in the metabolic chamber).

Our in vitro study of primary beige adipocytes revealed that sulforaphane promotes the expression of brown-fat selective genes (Fig. 4B). Importantly, the concentration of sulforaphane used in cell culture (0.2–5 μM) is comparable with that detected in mice fed NC-GR and HFD-GR diet (Supplementary Fig. 1B). Moreover, we determined that Nrf2 acts as a positive regulator of beige adipocyte differentiation (Fig. 4C and Supplementary Fig. 3C).

The less differentiation levels in Nrf2-deficient beige adipocytes are in agreement with previous reports demonstrating Nrf2 induces white adipocyte differentiation through increasing the gene expression of Pparγ (34) and Cebpβ (36), common transcription factors regulating the differentiation of brown, beige, and white adipocytes.

Furthermore, it is noteworthy that Nrf2 has been reported to bind NF-E2-binding sites in the 5′-flanking region of the human and rodent Ucp1 genes (37).

However, we cannot exclude the possibilities that glucoraphanin affects sympathetic nervous activity or that hormonal factors are regulating fat browning (38). In addition, mitochondrial reactive oxidative species facilitate Ucp1-dependent respiration in BAT and whole-body energy expenditure by promoting the sulfenylation of a specific cysteine residue (Cys253) in Ucp1 (39).

The molecular mechanism by which Nrf2 regulates the expression and thermogenic activity of Ucp1 in beige adipocytes requires further investigation.

Glucoraphanin supplementation improved the systemic glucose tolerance and insulin sensitivity of HFD-fed mice. Although the molecular mechanism by which synthetic Nrf2 inducers enhance glucose uptake is unclear, AMP-activated protein kinase (AMPK) activation may mediate this enhancement in mouse skeletal muscle and adipose tissue (9– 11).

The phosphorylation levels of AMPK (Thr172) and acetyl-CoA carboxylase (Ser79) in peripheral insulin target tissues were comparable between glucoraphanin-treated NC- and HFD-fed mice and vehicle-treated controls (Supplementary Fig. 6).

These data suggest that AMPK activation is not necessary for glucoraphanin to exert its insulin-sensitizing effect on HFD-fed mice. Additional studies using the hyperinsulinemic-euglycemic clamp technique are needed to determine which tissues contribute to the insulin-sensitizing effects of glucoraphanin.

The beneficial effects of glucoraphanin on hepatic lipid metabolism were not accompanied by AMPK activation or the increased expression of fatty acid β-oxidation genes (Fig. 5E).

Instead, glucoraphanin mitigated HFD-induced oxidative stress and inflammation in the liver. In obesity, hepatic inflammation mediated by macrophage/monocyte-derived proinflammatory cytokines promotes lipogenesis through the inhibition of insulin signaling and SREBP activation (40,41). In fact, the depletion of Kupffer cells by clodronate liposomes ameliorates hepatic steatosis and insulin sensitivity in HFD-fed mice (27).

Furthermore, Ccl2- or Ccr2-deficient mice are protected from diet- induced hepatic steatosis even though they still become obese (42,43). Moreover, the specific ablation of M1-like macrophages restores insulin sensitivity in diet-induced obese mice (44), while the deletion of Pparδ, which promotes M2 activation, predisposes lean mice to develop insulin resistance (45).

Therefore, decreased hepatic macrophage accumulation and M2-dominant polarization of hepatic and adipose macrophages account, at least in part, for the protection from hepatic steatosis and insulin resistance in glucoraphanin-treated HFD-fed mice.

One of the most important findings of this study is that glucoraphanin decreases the relative abundance of Gram-negative Proteobacteria, particularly family Desulfovibrionaceae, while reducing circulatory LPS levels (Fig. 7).

Recent studies demonstrated that a significant increase in Desulfovibrionaceae, potential endotoxin producers, in the gut microbiomes of both HFD-induced obese mice and obese human subjects compared to lean individuals (30,31,46). We cannot exclude the possibility that other microbiota-derived products, such as bile acids and short chain fatty acids, also mediate the metabolic action of glucoraphanin.

Whether the interaction between sulforaphane and gut microbiota is affected directly or indirectly by altered host physiology remains to be determined.

However, several studies have suggested that sulforaphane can alter the gut microbiota directly, as isothiocyanates (including sulforaphane) have been shown to exhibit antibacterial activity against Proteobacteria (47,48).

This activity may proceed via redox disruption and enzyme denaturation reactions involving the isothiocyanate reactive group, -N=C=S, the thiol group (-SH) of glutathione and proteobacterial proteins (47,48). Additionally, sulforaphane exhibits antibacterial activity against Helicobacter pylori, a member of the phylum Proteobacteria (49).

We are unaware of any previous reports demonstrating that isothiocyanates inhibit the proliferation of Desulfovibrionaceae. The mechanistic underpinnings of this antibacterial activity require elucidation.

In conclusion, the results of the present study indicate that glucoraphanin may be effective in preventing obesity and related metabolic disorders such as NAFLD and type 2 diabetes.

A recent clinical study demonstrated that supplementation with a dietary dose of glucoraphanin (69 μmol/day) for two months significantly decreased the plasma liver enzymes, ALT and AST, although body mass did not change (18).

Long-term treatment with a higher dose of glucoraphanin (800 μmol/day), which can be safely administered without any harmful side effects (17), may be required to achieve an anti-obesity effect in humans.