Regulation of Fasted Blood Glucose by Resistin

See allHide authors and affiliations

Science  20 Feb 2004:
Vol. 303, Issue 5661, pp. 1195-1198
DOI: 10.1126/science.1092341


The association between obesity and diabetes supports an endocrine role for the adipocyte in maintaining glucose homeostasis. Here we report that mice lacking the adipocyte hormone resistin exhibit low blood glucose levels after fasting, due to reduced hepatic glucose production. This is partly mediated by activation of adenosine monophosphate–activated protein kinase and decreased expression of gluconeogenic enzymes in the liver. The data thus support a physiological function for resistin in the maintenance of blood glucose during fasting. Remarkably, lack of resistin diminishes the increase in post-fast blood glucose normally associated with increased weight, suggesting a role for resistin in mediating hyperglycemia associated with obesity.

The parallel epidemics of obesity and type 2 diabetes suggest a relation between increased adipose mass and insulin resistance (1). Adipocytes secrete several signaling molecules that affect glucose homeostasis, such as fatty acids, adiponectin, leptin, interleukin-6, and tumor necrosis factor–α (2). Resistin is an adipocyte-secreted hormone that has been linked to diabetes (3, 4), a view supported by increased blood glucose and increased hepatic glucose production when resistin is administered acutely in rodents (5, 6). However, the role of resistin in glucose metabolism is controversial (7), and the normal physiological function of resistin is unknown.

We generated mice deficient in resistin by replacing the coding exons of the resistin gene (rstn) with the reporter gene lacZ (fig. S1, A and B) (8). Mating of heterozygous animals resulted in normal-sized litters of all expected genotypes at the rstn locus (+/+, +/–, and –/–) in the expected Mendelian ratios (fig. S1C). LacZ expression was confined to white adipose tissue (WAT) (fig. S2), indicating that the WAT-specific expression of resistin is determined by regulatory elements at the rstn gene locus (9). Resistin messenger RNA (mRNA) and protein expression were completely ablated in WAT of null mice (fig. S3A). Absence of circulating resistin was verified by immunoblot analysis and radioimmunoassay of mouse serum (fig. S3, B and C). The serum resistin concentration in rstn (+/–) mice was more than half that in (+/+) mice (20.6 ± 2.0 ng/ml versus 26.3 ± 1.7 ng/ml), suggesting that the (+/–) mouse compensates for its reduced gene dosage by altering resistin production and/or secretion. The (–/–) mice appeared grossly normal, were fertile upon reaching sexual maturity, and had normal size and distribution of adipose depots (10).

Rstn (–/–) and (+/+) mice gained weight similarly both on normal chow (10) and when fed a high-fat diet (Fig. 1A). Remarkably, however, blood glucose levels after 4 to 6 hours of fasting were 20 to 30% lower in the (–/–) mice on normal chow (Fig. 1B) or in the setting of diet-induced obesity (Fig. 1C), without significant differences in body weight or serum levels of insulin, leptin, adiponectin, and triglycerides (table S1). Blood glucose was also significantly lower in the (–/–) mice after overnight fasting (Fig. 1D). Administration of recombinant resistin to (–/–) mice fed high-fat diet to restore circulating levels near to those of the similarly fed (+/+) mice was sufficient to raise their fasted blood glucose levels by ∼25% (Fig. 1E).

Fig. 1.

Metabolic phenotype of rstn (–/–) mice. (A) Induction of diet-induced obesity. Mice were placed on high-fat diet at 6 weeks of age. (B) Blood glucose (after a 6-hour fast) of 13-week-old mice on normal chow. *P < 0.001. (C) Blood glucose (after a 4-hour fast) of 14-week-old mice that had been on high-fat diet for 8 weeks. *P < 0.02. (D) Blood glucose of 14-week-old mice [after an overnight (16 hour) fast] that had been on high-fat diet for 8 weeks. *P < 0.02. [(A) to (D)] n = 19 (+/+) mice; n = 16 (–/–) mice. (E) Blood glucose levels after intraperitoneal administration of resistin (1.5 μg) or vehicle to (–/–) mice. Mice fasted for 5 hours before treatment, and the fast continued for 3 hours thereafter. Serum level of resistin in (–/–) mice 2 hours after administration was 64.6 ± 16.1 ng/ml (n = 3 mice). For comparison, serum resistin level in (+/+) littermates on high-fat diet 2 hours after treatment with vehicle was 78.1 ± 4.9 ng/ml (n = 5 mice, P = 0.35 versus (–/–) mice 2 hours after resistin injection). n = 6 mice for both vehicle and resistin treatment. *P ≤ 0.03.

On normal chow, the glucose tolerance of the (+/+) and (–/–) mice was similar (10). However, when the animals were placed on a high-fat diet, glucose tolerance was significantly better in the (–/–) mice (Fig. 2A; fig. S4B). Fasting insulin levels and insulin tolerance tests were similar in the two groups (10). Because insulin tolerance testing assesses glucose disposal after an acute intraperitoneal injection of insulin, we assessed whole-body glucose homeostasis in euglycemic glucose clamp studies. During constant hyperinsulinemia (3.6 mU/kg/min) in mice with diet-induced obesity, a higher glucose infusion rate was required to maintain normal glucose level in (–/–) mice than in (+/+) mice (Fig. 2B). This difference was primarily accounted for by a dramatic reduction in glucose production in (–/–) mice, though whole-body glucose disposal was similar in (+/+) and (–/–) mice (Fig. 2, C and D). Administration of recombinant resistin to achieve levels comparable to those in (+/+) mice on a high-fat diet reversed the alterations in glucose metabolism in (–/–) mice fed the same diet (Fig. 2, B and D).

Fig. 2.

Dynamic metabolic studies of rstn (–/–) mice. (A) Glucose tolerance test of 22-week-old mice that had been on high-fat diet for 12 weeks. n = 8 (+/+) mice; n = 5 (–/–) mice. *P < 0.07, **P < 0.05, ***P < 0.002. (B to D) Hyperinsulinemic euglycemic glucose clamp studies of 20-week-old mice that had been on high-fat diet for 14 weeks. n = 9 (+/+) mice; n = 7 (–/–) mice; n = 9 (–/–) mice treated with resistin. (B) GIR, glucose infusion rate. *P < 0.0005 versus (+/+) mice and P < 0.01 versus (–/–) mice treated with resistin. (C) Rd, rate of glucose disposal. (D) GP, glucose production. *P < 0.0005 versus both other groups. Mean resistin levels of (–/–) and (+/+) mice in this studywere 1.1 ± 2.7 ng/ml, 61 ± 7 ng/ml, respectively. Mice receiving resistin (2.5 or 5 μg bolus then 3 μg per hour for 90 min) had a mean serum level of 84 ± 7 ng/ml (8).

Gluconeogenesis is a key metabolic pathway involved in hepatic glucose production (11). Administration of the gluconeogenic substrate pyruvate increased blood glucose levels to a significantly greater extent in (+/+) versus (–/–) mice (Fig. 3A), suggesting that lack of resistin impairs gluconeogenesis. Consistent with this hypothesis, gene expression of the key gluconeogenic enzymes glucose 6-phosphatase (G6Pase) and phosphoenolpyruvate carboxykinase (PEPCK) were markedly decreased in the livers of the clamped (–/–) mice (Fig. 3, B and C; fig. S5A). The levels of two factors that promote gluconeogenesis, PGC-1α (12) and HNF4α (13), and two factors that inhibit gluconeogenesis, phosphorylated forms of FoxO1 (14) and protein kinase B/Akt (15), were similar in the (+/+) and (–/–) mice (fig. S5B).

Fig. 3.

Gluconeogenic factors in rstn (–/–) mice. (A) Blood glucose levels after pyruvate tolerance test. n = 9 (+/+) mice; n = 7 (–/–) mice *P < 0.02. (B to E) Studies of hepatic extracts from mice (18 weeks old, on high-fat diet for 12 weeks) clamped as in Fig. 2. n = 5 (+/+) mice; n = 4 (–/–) mice. (B) G6Pase gene expression *P < 0.04. (C) PEPCK gene expression. *P < 0.05. (D) Phospho-AMPK. *P < 0.05. (E) Phospho-ACC *P < 0.002. (F) Protein extracts from livers of the clamped mice were immunoprecipitated with antibodyto AMPK, then incubated with substrate for AMPK (SAMS) peptide in the presence of γ-32P-adenosine triphosphate before SDS–polyacrylamide gel electrophoresis and autoradiography (top). Each sample contained equal amounts of AMPK as shown by immunoprecipitation (bottom).

The G6Pase and PEPCK genes are both down-regulated by activation of adenosine monophosphate–activated protein kinase (AMPK) (16), which regulates several metabolic pathways, including gluconeogenesis (17). The active, phosphorylated form of AMPK was increased by more than a factor of two in the rstn (–/–) livers without significant alteration of total AMPK level (Fig. 3D; fig. S5C). Moreover, phosphorylation of the AMPK substrate acetyl CoA carboxylase (ACC) was increased in the livers of (–/–) mice, consistent with activation of AMPK (Fig. 3E; fig. S5C). Measurement of AMPK activity in vitro confirmed this result (Fig. 3F). The increase in hepatic phospho-AMPK was abrogated by resistin treatment of the (–/–) mice (fig. S6). Hepatic triglyceride content tended to be lower in the (–/–) mice [29.1 ± 1.6 mg/g (n = 6 mice) versus 34.0 ± 2.1 mg/g (n = 9)], consistent with AMPK activation (18), although the difference did not reach statistical significance. Serum levels of leptin and adiponectin, which suppress hepatic glucose production in part via activation of AMPK (19, 20), were not different in the resistin-null mice (table S1). These results suggest that resistin normally acts on the liver to inhibit AMPK, and this response is impaired in the resistin-null mouse.

Diet-induced obesity is a major cause of obesity-related insulin resistance and type 2 diabetes (21). As expected, rstn (+/+) mice on a high-fat diet exhibited a strong positive correlation between body weight and blood glucose (Fig. 4A, open squares, r2 = 0.45, P < 0.002). Remarkably, this relation was weaker in the (–/–) mice (Fig. 2A, solid triangles, r2 = 0.24, P < 0.06). The statistical association between weight and glucose was significantly less in the rstn (–/–) mice (P < 0.02). Moreover, fasting blood glucose was not significantly different between rstn (+/+) and (–/–) mice weighing ≤32 g (Fig. 4B). By contrast, heavier rstn (+/+) mice (>32 g) had markedly higher fasting glucose than (–/–) mice of similar weight (Fig. 4C), indicating that absence of resistin protects against fasting hyperglycemia associated with obesity.

Fig. 4.

Relation between glucose and body weight is altered in rstn (–/–) mice. (A) Fasted blood glucose of 14-week-old mice that had been placed on high-fat diet for 8 weeks, plotted as a function of bodyweight. The mean and SEM of these glucose values are shown in Fig. 1C. (B) Blood glucose levels in mice weighing 32 g or less. (C) Blood glucose levels in mice weighing more than 32 g.

The phenotype of rstn (–/–) mice implicates resistin in normal glucose homeostasis and as a factor in the positive correlation between weight and fasted glucose levels. Although resistin has been shown to impair glucose transport into cultured adipocytes (5, 22) and muscle (23, 24), the major effect of resistin deficiency on whole-body glucose metabolism was impairment of hepatic glucose output. This is consistent with the observation that resistin infusion augments glucose production (5, 6). Reduced hepatic expression of gluconeogenic enzymes in the absence of resistin are likely related, at least in part, to activation of AMPK, suggesting that resistin signaling opposes that of two adipocyte hormones, leptin and adiponectin, which activate AMPK and stimulate gluconeogenesis (2527). Thus, targeting the AMPK pathway may provide a novel therapeutic strategy for type 2 diabetes.

These studies demonstrate that resistin is an important regulator of glucose metabolism in mouse models that are commonly used to study obesity, insulin resistance, and type 2 diabetes. The role of resistin in human insulin resistance is controversial, in part because human resistin appears to be expressed at highest levels in macrophages (28, 29). However, antidiabetic thiazolidinedione drugs inhibit resistin gene expression in human macrophages (30) as well as in mouse adipocytes (5, 31, 32). Resistin expression is positively correlated with insulin resistance in humans (33, 34), and serum resistin levels are elevated in human obesity (35). Furthermore, polymorphisms of the human rstn gene have been linked to insulin resistance in certain populations (3638). It is, therefore, tempting to speculate that human and mouse resistin have similar metabolic functions despite their divergent sites of production. This would be consistent with the emerging concept that adipocytes and macrophages share the property of secreting inflammatory hormones and cytokines that regulate metabolic function (39, 40).

Supporting Online Material

Materials and Methods

Figs. S1 to S6

Table S1


References and Notes

View Abstract

Navigate This Article