Salicylate Downregulates 11β-HSD1 Expression in Adipose Tissue in Obese Mice and in Humans, Mediating Insulin Sensitization

Chemicals were purchased from Sigma-Aldrich (Dorset, U.K.) unless otherwise stated.

Simpson-Golabi-Behmel syndrome (SGBS) cells are a human preadipocyte cell line characterized by a high capacity for in vitro differentiation (39). Cells were cultured in Dulbecco’s modified Eagle’s medium–F12, supplemented with 10% (v/v) FCS, 50 units/mL penicillin per 50 units/mL streptomycin, biotin (33 μmol/L), and pantothenic acid (17 μmol/L) at 37°C in 5% CO₂. Adipogenic differentiation was induced after reaching near confluence as described previously (39). At day 14, cells were incubated for 24 h in stripped serum medium prior to treatment with salicylic acid (10, 30, and 100 μmol/L) or vehicle (0.1% ethanol) for 24 h, and RNA was extracted for analysis of 11β-HSD1 transcript. To assess the interaction between salicylate and 11β-HSD1 in determining adiponectin (AdiQ) transcript levels, day 14 cells were incubated for 24 h in cortisol-free, stripped serum medium prior to treatment with cortisol (0.1 μmol/L), cortisone (0.1 μmol/L), or vehicle (0.1% ethanol) with or without salicylic acid alone (100 μmol/L), or vehicle (0.1% ethanol) for 24 h. RNA was extracted for analysis.

Experiments were performed under license from the U.K. Home Office. Mice were maintained under controlled conditions of light (0700–1900 h) and temperature (18–20°C) and allowed access to food and drinking water ad libitum. Adult male C57Bl/6 mice were obtained (Harlan Olac, Oxfordshire, U.K.) at age 12 weeks. C57Bl/6 Lean animals were maintained on normal chow diet (2.27% fat, 4.05% sucrose; 801151; Special Diet Services). Diet-induced obese C57Bl/6 mice (C57Bl/6 DIO) were given 10 weeks of high-fat diet (58% fat, 12% sucrose; D12331; Research Diets, New Brunswick, NJ) before treatment. Homozygous male 11β-HSD1–deficient (HSD1KO) mice were bred from an in-house colony on a C57Bl/6 background (>10 generations backcross) (21) and fed a high-fat diet until weight matched with C57Bl/6 DIO mice (∼38 g).

Sodium salicylate (salicylate) (120 mg/kg/day) or distilled water (vehicle) was administered from 1 week after arriving (C57Bl/6 Lean), after 10 weeks of high-fat feeding (C57Bl/6 DIO), or after achieving target weight (HSD1KO-DIO) for 4 weeks to groups of n = 8 via osmotic minipumps implanted subcutaneously between the scapulae.

For glucose tolerance tests, mice were fasted for 6 h before administration of glucose (2 g/kg) by intraperitoneal injection and blood sampling by tail nicks for 90 min thereafter.

Animals were killed by decapitation and tissues dissected onto dry ice and stored at −80°C until analysis.

A total of 16 otherwise healthy men, selected to represent a wide range of BMIs, from 20 to 50 kg/m², took part in a double-blind balanced randomized crossover study to compare salsalate (Disalcid; 3M Pharmaceuticals, Northridge, CA; 1 g p.o. every 8 h for 2 weeks) with placebo. There was a 2-week washout period between phases, consistent with the short half-life of salsalate (<8 h). Inclusion criteria included normal thyroid, renal, and hepatic function; alcohol intake <28 units/week; no glucocorticoid use in the previous 6 months; no medication, including nonsteroidal anti-inflammatory medication in the previous month or during the study; and no history of dyspepsia or peptic ulcer disease. Local ethical committee approval and written informed consent were obtained.

At the end of each phase, volunteers attended the clinical research facility on 2 consecutive days. On day 1, they completed a 24-h urine collection and attended at 0830 h after fasting overnight. Blood was taken for glucose, insulin, lipid profile, and plasma cortisol. Height, weight, waist, and hip circumferences; percent fat by bioimpedance (Omron BF302, Milton Keynes, U.K.); and blood pressure by mercury sphygmomanometer were recorded. An adipose biopsy was obtained from periumbilical subcutaneous adipose tissue under local anesthesia (lignocaine 1%; Braun; Sheffield, U.K.) using a 14G biopsy needle (Sterican; B. Braun Melsungen AG) attached to a sterile 60 mL syringe. The tissue was then washed with diethyl pyrocarbonate–treated water, frozen immediately on dry ice, and stored at −80°C.

On day 2, volunteers took oral dexamethasone 1 mg (Merck Sharp & Dohme, Hertsfordshire, U.K.) at 2300 h the evening before, fasted overnight, and attended at 0900 h. A venous cannula was inserted and oral cortisone acetate 25 mg (Pantheon Inc., Swindon, U.K.) was administered with water. Plasma samples were collected at intervals shown in Fig. 5 in lithium heparin and stored at −20°C.

Mouse glucose and insulin were measured by hexokinase assay (Thermo Electron, Victoria, Australia) and ELISA (Crystal Chem Inc., Downers Grove, IL), respectively.

Human plasma cortisol was measured using the ImmunChem cortisol ¹²⁵I radioimmunoassay coated tube kit (MP Biomedicals, High Wycombe, U.K.). Enzyme immunoassays (Eurogenetics Tasah Corp. UK Ltd., Hampton, U.K.) were used to measure plasma insulin and C-peptide. Electrolytes and salicylate were measured by dry-slide chemistry with a Vitros 950 (Ortho Clinical Diagnostics, Raritan, NJ), and glucose was determined on a Cabas Mira Plus (Roche, Mannheim, Germany). Triglycerides, total cholesterol, and HDL cholesterol were measured using ELISA kits (TG, CHOL, and HDLC-plus, respectively; Roche).

Human urinary cortisol and its metabolites (5β-tetrahydrocortisol [THF], 5α-THF, 5β-tetrahydrocortisone [THE], cortols, cortolones, and cortisone) were measured in 24-h samples by electron impact gas chromatography/mass spectrometry as previously described (40). The balance between 11β-HSD activities in all tissues was assessed as the ratio of (5β-THF + 5α-THF) to THE.

For cultured cells and mouse tissue samples, total RNA was extracted using TRIzol reagent (Invitrogen, Paisley, U.K.) and RNeasy Mini Kit (QIAGEN, West Sussex, U.K.). RNA integrity was verified on an agarose gel (1.2% w/v in 0.5× Tris/borate/EDTA). RNA (0.25 μg) was reverse transcribed with random primers using the QuantiTect Reverse Transcription Kit (QIAGEN). Real-time PCR was performed using the Roche LightCycler 480 (Roche Applied Science, Burgess Hill, U.K.). Primers (Invitrogen) were designed for use with intron-spanning probes within the Roche Universal Probe Library. Results from adipose tissue from lean and obese mice were corrected for abundance of 18S and/or TATA binding protein mRNA, which were not affected by treatment. Results from SGBS cells were corrected for abundance of TATA-binding protein mRNA, which was not affected by treatment. All samples were analyzed in triplicate and amplification curves plotted (y-axis fluorescence, x-axis cycle number). Triplicates were deemed acceptable if the SD of the crossing point was <0.5 cycles. A standard curve (y-axis crossing point, x-axis log concentration) for each gene was generated by serial dilution of cDNAs pooled from different samples, fitted with a straight line, and deemed acceptable if reaction efficiency was between 1.7 and 2.1.

For human adipose tissue, 100 mg tissue was homogenized and RNA extracted using the QIAGEN RNeasy Lipid Tissue Mini kit (West Sussex, U.K.) and quantified using the Ribogreen quantitation kit (Molecular Probes, Eugene, OR). RNA integrity was verified on a 1.2% agarose/Tris borate/EDTA gel. Oligo dT-primed cDNA was synthesized from 0.5 μg RNA using Promega Reverse Transcription System (Promega, Southampton, U.K.), together with controls from which reverse transcriptase was omitted. Transcript quantification was performed as above for mouse samples. Primer-probe sets for cyclophilin A and 11β-HSD1 were purchased from Applied Biosystems (Cheshire, U.K.). Random primers were used for 18S. The results are expressed as a ratio to the mean of cyclophilin A and 18S, as internal controls that did not differ between groups.

11β-HSD1 functions as a reductase in vivo, converting inactive 11-dehydrocorticosterone to corticosterone. However, in vitro, dehydrogenase activity predominates, so velocity of 11β-HSD1 activity in adipose was quantified through conversion of corticosterone to 11-dehydrocorticosterone. Enzyme kinetics were determined from whole-adipose tissue homogenate. Standardized amounts of protein for each tissue (mesenteric adipose 300 μg/mL; subcutaneous adipose 100 μg/mL) were incubated (37°C) in Krebs-Ringer buffer containing glucose (1% w/v), NADP (2 mmol/L), [³H]₄-corticosterone (10 nmol/L), and unlabeled corticosterone (9.99 μmol/L). Steroids were extracted after incubation (mesenteric adipose 24 h; subcutaneous adipose 22 h) with ethyl acetate, the organic phase was evaporated under nitrogen, and extracts were solubilized in the mobile phase (water to acetonitrile to methanol; 60:15:25, 1.5 mL/min). Steroids were separated by high-performance liquid chromatography using a C₁₈ Sunfire column (Waters, Hertfordshire, U.K.) at 42°C and quantified by on-line liquid scintillation counting.

Results are presented as mean ± SEM and analyzed by Student t tests or two-way ANOVA with Bonferroni post hoc tests, as indicated in the figure legends. Incremental changes induced by salsalate (i.e., value after salsalate minus value after placebo) were correlated with indices of obesity and plasma salsalate concentrations using Pearson correlation tests.

Ten weeks of high-fat feeding caused diet-induced obesity in male C57Bl/6 DIO mice, with elevated fasting plasma insulin levels compared with C57Bl/6 Lean mice (Table 1). On glucose tolerance testing, C57Bl/6 DIO mice were hyperglycemic and hyperinsulinemic compared with C57Bl/6 Lean mice, indicating development of insulin resistance (Table 1 and Fig. 1).

Four weeks of treatment with salicylate (120 mg/kg/day) had little measurable effect in C57Bl/6 Lean mice, except for elevated fasting plasma insulin. In C57Bl/6 DIO mice, salicylate decreased both fasting and postprandial plasma glucose levels (Table 1 and Fig. 1). Furthermore, there was a trend to reduced plasma triglyceride levels after salicylate treatment in C57Bl/6 DIO mice (P = 0.059) (Table 1).

Salicylate significantly reduced 11β-HSD1 mRNA in omental adipose tissue in C57Bl/6 DIO mice, with a similar trend in mesenteric adipose (P = 0.057) (Fig. 2A). In mesenteric adipose of C57Bl/6 DIO mice, salicylate also reduced 11β-HSD1 enzyme activity (Fig. 2B).

HSD1KO-DIO mice were protected against fasting hyperinsulinemia compared with weight-matched C57Bl/6 DIO controls. However, HSD1KO-DIO mice were still hyperinsulinemic, both fasting and postprandial, compared with C57Bl/6 Lean mice (Table 1 and Fig. 1). In marked contrast to C57Bl/6 DIO mice, salicylate had no effects on biochemical indices in HSD1KO-DIO mice (Table 1 and Fig. 1).

In C57Bl/6 DIO mice (Fig. 3), salicylate altered mRNA levels of several genes. In mesenteric adipose (Fig. 3B), salicylate increased AdiQ and decreased TNF-α, monocyte chemoattractant protein-1 (MCP-1), adipose triglyceride lipase (ATGL), and angiotensinogen (AGT) mRNA levels. This pattern of salicylate-induced changes was similar in omental adipose (Fig. 3A). With the exception of increasing AdiQ mRNA levels, these effects were absent in subcutaneous adipose tissue (Fig. 3C).

Compared with salicylate-treated C57Bl/6 DIO mice, HSD1KO-DIO mice had a similar pattern of transcript changes in visceral adipose, with increased AdiQ and reduced TNF-α and ATGL mRNA levels (Fig. 3). However, in HSD1KO-DIO mice, salicylate treatment did not alter mRNA levels of any of the genes affected in C57Bl/6 DIO mice (Fig. 3).

Characteristics of the 16 male participants are shown in Table 2. Salicylate levels in plasma averaged 134 ± 33 mg/L during active treatment and were undetectable during placebo. Only 1 of the participants reported tinnitus during salsalate therapy, and there were no changes in indices of insulin sensitivity or lipid profile. Nevertheless, salsalate reduced 11β-HSD1 mRNA levels in subcutaneous adipose, an effect which was unrelated to BMI (Fig. 4) or plasma salicylate levels (data not shown). This was accompanied by reduced transcript levels for MCP-1 (33.6 ± 5.6% suppression, P < 0.02), but not TNF-α or AdiQ. The effect of salsalate on 11β-HSD1 activity appeared restricted to adipose tissue since there was no change in first pass conversion of cortisone to cortisol in liver or in the ratio of cortisol to cortisone metabolites in urine (Fig. 4).

In fully differentiated human SGBS adipocytes, 24-h incubation with salicylic acid in the absence of steroid dose dependently reduced 11β-HSD1 mRNA levels (Fig. 5A) but had no effect on AdiQ mRNA (Fig. 5B). Both cortisol and cortisone treatment (0.1 μmol/L for 24 h) reduced mRNA levels of AdiQ. In the presence of cortisone, coincubation with salicylic acid attenuated the suppression of AdiQ, whereas salicylate had no effect on suppression of AdiQ by cortisol (Fig. 5B).

These results identify a novel mechanism by which salicylates may enhance insulin action in diet-induced obesity, involving downregulation of adipose 11β-HSD1 expression. This downregulation occurs selectively in adipose tissue depots in mice and in humans in vivo, is likely mediated directly within adipose, and is mimicked in human adipocytes in culture. It is crucial that mice deficient in 11β-HSD1 display similar changes in in vivo metabolism and intra-adipose gene expression as salicylate-treated wild-type mice, and salicylate has no additional effects on these parameters in the absence of 11β-HSD1, implicating 11β-HSD1 as a mediator of the beneficial effects of salicylate.

The magnitude and components of the metabolic effects of salicylate remain somewhat contentious. For example, in humans, in addition to reports of improved insulin sensitivity (10,6,41,7), several reports describe worsening of insulin resistance (42–44) or attribute improved metabolism to an increase in insulin concentrations (9). Here, we confirm variable efficacy of salicylate in different settings. In lean mice, salicylate modestly increased plasma insulin concentration but had no other metabolic effects. In DIO mice, salicylate lowered fasting plasma glucose concentrations and improved glucose tolerance, as well as reducing 11β-HSD1 in visceral adipose (omental and mesenteric) but not subcutaneous adipose tissue. These effects, together with the lack of change in plasma insulin concentrations, are consistent with insulin sensitization primarily in visceral adipose tissue due to reduced local glucocorticoid regeneration by 11β-HSD1 (45,46), with secondary improvements in insulin sensitivity in other organs potentially mediated by altered adipokine secretion. The HSD1KO-DIO mice were <2 weeks younger than the C57Bl/6 DIO mice, so it is highly unlikely that age-dependent effects are responsible for the discrepancies in responses to salicylate between groups. In contrast, in the human study, although we did not undertake sensitive measurements of insulin sensitivity, such as euglycemic clamps, there were no measured effects of salsalate on fasting insulin or lipid profile; this was the case even if we examined only obese participants (not shown). It may be that high doses are required to induce robust insulin sensitization in humans: the salicylate concentrations achieved in this study (mean 134 mg/L, ∼1 mmol/L) (Table 2) are lower than in some other studies (e.g., 270 mg/L) (41). This inconsistency is in keeping with a mechanism of action mediated by 11β-HSD1 downregulation, since recent phase 2 clinical trials of selective 11β-HSD1 inhibitors demonstrate that metabolic effects are of relatively small magnitude and not always statistically significant (26,47). Previous studies demonstrating efficacy of salicylates in rodents use genetic models of obesity, including Zucker obese rats and ob/ob mice (32,2,18,3). To our knowledge, this is the first study to demonstrate the insulin-sensitizing effects of salicylate in a nongenetic model of obesity. Indeed, diet-induced obesity was chosen because it is a more relevant model to the obesity seen in humans. Furthermore, salicylates improve insulin sensitivity during lipid-induced insulin resistance (2,7). 11β-HSD1 activity is increased acutely in adipose tissue by intralipid infusion (48), potentially contributing to lipid-induced insulin resistance. Blockade by salicylates of the effective increase in 11β-HSD1 activity could therefore explain its relatively selective effect on lipid-induced insulin resistance.

Protective metabolic effects of 11β-HSD1 deficiency in mice receiving a high-fat diet have been described previously (22) and are shown here to bear remarkable similarity to the pattern of effects of salicylates on glucose tolerance and adipose lipase expression. It might be argued that the lack of effect of salicylates in HSD1KO mice is confounded by the improved metabolic “baseline” in these animals. However, to avoid this, we administered high-fat diet for long enough to achieve similar body weight in HSD1KO mice to that observed in C57Bl/6 DIO mice and, thus, ensured these HSD1KO-DIO mice were markedly hyperinsulinemic compared with lean mice. The striking lack of efficacy of salicylate treatment in HSD1KO mice is therefore consistent with downregulation of 11β-HSD1 playing a key role in its mechanism of action.

Downregulation of 11β-HSD1 by salicylates is likely to reflect a direct mechanism in adipose tissue, given its replication in an adipocyte cell line. Furthermore, we have demonstrated in vitro that salicylic acid reverses cortisone-mediated suppression of the insulin-sensitizing adipokine AdiQ; since salicylic acid had no effect when the contribution of 11β-HSD1 was negated by removal of steroid or by addition of cortisol, these data indicate that inhibition of 11β-HSD1 by salicylates mediates the altered AdiQ expression also observed in vivo. The downregulation of 11β-HSD1 was selective for the visceral adipose depots in mice but was evident in subcutaneous adipose in humans. Central adipose depots exhibit greater variation in lipolytic activity and more intense proinflammatory responses during high-fat feeding and are more glucocorticoid responsive (49). However, in humans, recent evidence suggests that 11β-HSD1 regenerates more cortisol in subcutaneous than in visceral adipose tissue (50), suggesting species-specific differences. Moreover, we did not find evidence that changes in proinflammatory cytokines, such as TNF-α, mediate the effects of salicylates on 11β-HSD1 expression; the downregulation of TNF-α by salicylates in C57Bl/6 DIO mice, but resistance to this effect in 11β-HSD1KO mice, suggests that effects of salicylates on intra-adipose inflammation may be a consequence rather than cause of downregulation of 11β-HSD1. This is further supported by the downregulation of TNF-α in the visceral adipose of 11β-HSD1KO mice compared with C57Bl/6 mice. In human adipose, TNF-α mRNA did not change in association with altered 11β-HSD1 mRNA. In the absence of cytokine-mediated regulation, more detailed investigation will be required to dissect the molecular mechanisms behind the salicylate-mediated regulation of 11β-HSD1 expression.

In conclusion, these findings suggest that the anti-inflammatory agent salsalate alters glucocorticoid metabolism in mice and humans in a pattern that differs between liver and subcutaneous adipose tissue. Downregulation of intra-adipose 11β-HSD1 may contribute to the insulin-sensitizing effect of salicylates.

This work was supported by grants from the British Heart Foundation, the Wellcome Trust, and the Medical Research Council.

Within the past 2 years, J.R.S. and B.R.W. have consulted for AstraZeneca, Incyte, Roche, and Wyeth; received lecture fees from Merck; and received research funding from AstraZeneca and Wyeth. J.R.S. and B.R.W. are inventors on relevant patents held by the University of Edinburgh. No other potential conflicts of interest relevant to this article were reported.

M.N. conducted the experiments, analyzed data in mice and in vitro, and wrote the manuscript. D.E.L. and C.L.E. conducted the experiments and analyzed data in mice and in vitro. D.J.W., R.H.S., and R.A. conducted the experiments and analyzed data in humans. J.R.S. and K.E.C. contributed to discussion. B.R.W. wrote the manuscript, which was reviewed by all authors. B.R.W. 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.

Parts of this study were presented at ENDO 2010 93rd Annual Meeting and Expo of The Endocrine Society, San Diego, California, 19–22 June 2010, and at the British Endocrine Society Meeting, Manchester, U.K., 15–18 March 2010.

The authors thank the Wellcome Trust Clinical Research Facility staff for facilitating the study; Jill Harrison, Alison McNeilly, Carolynn Cairns, Lynne Ramage, andAsh Taylor (all of University of Edinburgh) and the staff of the Clinical Biochemistry Laboratory, Western General Hospital, for technical assistance; and the staff of the Biomedical Research Facility, University of Edinburgh, for help with animal care. SGBS cells were kindly provided by Prof. Martin Wabitsch, University of Ulm, Germany, and Dr. Kerry McInnes, University of Edinburgh.