Brain IRS2 Signaling Coordinates Life Span and Nutrient Homeostasis

See allHide authors and affiliations

Science  20 Jul 2007:
Vol. 317, Issue 5836, pp. 369-372
DOI: 10.1126/science.1142179


Reduced insulin-like signaling extends the life span of Caenorhabditis elegans and Drosophila. Here, we show that, in mice, less insulin receptor substrate–2 (Irs2) signaling throughout the body or just in the brain extended life span up to 18%. At 22 months of age, brain-specific Irs2 knockout mice were overweight, hyperinsulinemic, and glucose intolerant; however, compared with control mice, they were more active and displayed greater glucose oxidation, and during meals they displayed stable superoxide dismutase–2 concentrations in the hypothalamus. Thus, less Irs2 signaling in aging brains can promote healthy metabolism, attenuate meal-induced oxidative stress, and extend the life span of overweight and insulin-resistant mice.

Reaching old age in good health is not just good luck but the result of a favorable balance between hundreds of disease-causing and longevity-promoting genes; regardless, some common mechanisms that influence life span have emerged (1). First, calorie restriction reliably increases animal longevity, and second, reduced insulin-like signaling extends life span in Caenorhabditis elegans and Drosophila melanogaster (2, 3). Calorie restriction and reduced insulin-like signaling might be linked because fasting reduces the intensity and duration of insulin secretion required for glucose homeostasis, and reduced insulin-like signaling promotes the expression of antioxidant enzymes that are associated with longevity (35). Adapting these principles to humans is challenging because calorie restriction is difficult and because reduced insulin-like signaling can be associated with small stature, metabolic disease, and diabetes.

Insulin and insulin-like growth factor-1 (IGF1) bind to receptors on the surface of all cells that phosphorylate tyrosyl residues on the insulin receptor substrates (IRSs)—chico in Drosophila and Irs1, -2, -3, and -4 in mammals. This signaling cascade activates the phosphoinositide-3-kinase (Pik3C) and the thymoma viral proto-oncogene Akt, which regulates many cellular processes, including the inactivation of forkhead box O1 (FoxO1) transcription factor (6). Reduced chico expression decreases brain and body growth while increasing life span up to 50%, which is related to the increased activity of dFOXO in Drosophila (7, 8). In mice, the deletion of Irs1 reduces body growth and causes hyperinsulinemia, whereas the deletion of Irs2 (Irs2–/– mice) reduces brain growth and causes fatal diabetes by 3 months of age because of pancreatic β cell failure (9). By comparison, young (2-months-old) Irs2+/– mice display normal metabolic phenotypes (10). Old (22½-months-old) Irs2+/– mice were slightly heavier than wild-type (WT) mice, although young and old WT and Irs2+/– mice consumed the same amount of food each day (Fig. 1, A and B). In addition, old Irs2+/– mice were more insulin sensitive than old WT mice (Fig. 1C) because their fasting insulin and glucose concentrations were lower (fig. S1, A and B).

Fig. 1.

Metabolism and life span of Irs2+/– mice. (A) Food intake [average (g/24 hour) ± SEM, n = 6 mice] in male or female mice at 2 and 22 months of age. (B) Body weight [average (g) ± SEM, n = 6] at 22½ months. (C) HOMA2 (homeostatic model assessment) of insulin sensitivity (IS) (average ± SEM, n = 8, *P < 0.05). (D) Survival probability was determined by Cox regression for each WT (Embedded Image) and Irs2 +/– (◼, blue) mouse. Solid lines indicate the Cox survival probability of WT and Irs2+/– mice controlled for other covariates (see table S3 for details). Hatched lines correspond to Kaplan-Meier (KM) estimates obtained by parametric regression (see table S4 for details).

Because insulin sensitivity is associated with longevity (3), we compared the life spans of WT and Irs2+/– mice. Inspection of the results suggested that the date of birth (DOB), paternal (PID) and maternal (MID) identity, and sex influenced life span (table S1); therefore, we used semiparametric (Cox) and parametric regression to control the covariates and evaluate the effect of Irs2 (11). Cox regression revealed a 48-fold (P <10–9) reduced risk of death for Irs2+/– mice compared with WT controls (Fig. 1D and table S3). By using parametric analysis, we found that the median life span for Irs2+/– mice was 17% longer [for WT, median = 789 days and 95% confidence interval (CI) 755 to 769; for Irs2+/–, median = 925 days and 95% CI 887 to 940; P = 0.01] (table S4). The maximum life span, estimated at the 90th percentile, increased similarly (for WT, 837 days and 95% CI 801 to 845; for Irs2+/–, 982 days and 95% CI 935 to 998) (Fig. 1D and table S4).

Irs2 is expressed throughout the body (Fig. 2A) and many regions of the brain, including the cerebrum, the cerebellum, and the arcuate and paraventricular nuclei of the hypothalamus (fig. S2) (12, 13). Reduced insulin-like signaling in neurons increases the life span of C. elegans and Drosophila, so it is possible that reduced neuronal Irs2 could extend mouse life span (3, 14, 15). To test this hypothesis, we deleted one (bIrs2+/–) or both (bIrs2–/–) loxP-flanked Irs2 alleles (fIrs2-alleles) in the brain by intercrossing fIrs2 mice with nestin-cre transgenic mice (13). Polymerase chain reaction (PCR) analysis confirmed that Irs2 RNA was retained in all tested tissues of the bIrs2–/– mice except for the brain (Fig. 2A). Quantitative reverse transcription PCR (RT-PCR) confirmed that Irs2 RNAwas reduced about 50% in bIrs2+/– brains and more than 90% in bIrs2–/– brains compared with that of control fIrs2 mice (Fig. 2A). By contrast, Irs2 RNA increased in the pancreas of bIrs2+/– and bIrs2–/– mice, confirming that nestin-cre expression did not take place in pancreatic β cells (Fig. 2A).

Fig. 2.

Reduced neuronal Irs2 causes peripheral insulin resistance. (A) (Top) RT-PCR of Irs2 RNA in mouse tissues and (bottom) quantitative RT-PCR (normalized ± SEM; n = 5; *P < 0.05) in brain and pancreas (Irs2–/– indicates systemic Irs2–/– mice). (B) Food intake [average (g/24 hour) ± SEM; n = 6; *P < 0.05] at 2 months and 22 months of age. (C) HOMA2 of IS (average ± SEM, n = 8; *P < 0.05). (D) Average area (±SEM; n = 6; *P < 0.05) under the blood glucose clearance curve. (E) Fasted insulin concentrations (±SEM, n = 14 to 16; *P < 0.05). (F) (Left) Pancreas sections from 22-month-old mice were immunostained with antibody against insulin (green) or glucagon (red), and (right) the amount of islet β cell area (average ± SEM, n = 6 sections; *P < 0.05).

Crosses between bIrs2+/– mice produced offspring at a normal frequency, whereas crosses between bIrs2+/– and bIrs2–/– mice produced fewer offspring; offspring were never produced by crossing bIrs2–/– mice (fig. S3A). Old male and female bIrs2–/– mice consumed the most food each day by comparison to the other mice (Fig. 2B). However, by 22 months the bIrs2+/– and bIrs2–/– mice were about 10 g heavier than controls, owing in part to increased adiposity (fig. S3, B and C). The bIrs2–/– mice were 10% longer, and their brains were 30% smaller than those of fIrs2 or bIrs2+/– mice (fig. S3, D and E). These results support previous conclusions that brain Irs2 signaling promotes embryonic brain growth, central nutrient homeostasis, melanocortin 4 receptor signaling (body length), and fertility (13, 1618).

Control fIrs2 mice developed insulin resistance between 2 and 22 months of age (Fig. 2C). Unlike old Irs2+/– mice (Fig. 1), old bIrs2+/– mice and young and old bIrs2–/– mice were insulin resistant (Fig. 2C). All the insulin-resistant mice displayed mild glucose intolerance (Fig. 2D). However, diabetes did not develop in these mice because insulin concentrations increased to compensate for peripheral resistance, reaching the highest amount in old male bIrs2+/– and bIrs2–/– mice (Fig. 2E). Consistent with the observed hyperinsulinemia, pancreatic islets were larger in old male bIrs2+/– and bIrs2–/– mice than in old controls (Fig. 2F).

Next, we compared the life spans of fIrs2, bIrs2+/–, and bIrs2–/– mice by using semiparametric (Cox) and parametric regression to control for the covariates (table S2). Although bIrs2+/– and bIrs2–/– mice displayed metabolic changes usually associated with a shorter life span, their risk of death determined by Cox regression was significantly reduced compared with that of controls (for bIrs2+/–, a 14-fold reduction, P <10–11; for bIrs2–/–, a sixfold reduction, P <10–5)(Fig. 3A and table S3). With parametric regression, we found that the median life spans for bIrs2+/– and bIrs2–/– mice were 18% and 14% longer, respectively, than life spans for controls (for fIrs2, median = 791 days and 95% CI 731 to 796; for bIrs2+/–, median = 936 days and 95% CI 923 to 945; for bIrs2–/–, median = 901 days and 95% CI 888 to 919); the maximum life spans (90th percentile) increased similarly (Fig. 3A and table S4).

Fig. 3.

Life span and energy metabolism. (A) Observed survival probabilities—fIrs2 (⚫), bIrs2+/– (◼, blue), and bIrs2–/– (▲, red)—calculated from the Cox regression. Solid lines show the indicated gene effects controlled for other covariates (table S3). Hatched lines are the corresponding KM estimates based on the parametric regression (table S4). (B) Median O2 consumption (solid horizontal line), 99% CI (notch), and the lower and upper quartiles (n = 425 for each genotype, **P < 0.0001) by male mice in dark (shaded boxes) and light cycles (open boxes). (C) Median voluntary movement (±95% CI) for male mice measured during a 5-s interval in the dark cycle (n = 425 for each genotype, *P < 0.02, **P < 0.0001). (D) Average Rq (VCO2/VO2) determined during 48 hours from six male mice of the indicated age and genotype. Median Rq determined in dark (DC, black bar) and light (LC) cycles (Kruskal-Wallis nonparametric test: *P < 0.0001; ×P = 0.03).

To determine whether brain Irs2 affects systemic metabolism, we studied young and old mice in a comprehensive lab animal monitoring system (CLAMS). Young control (fIrs2) mice were more active and consumed more oxygen than did young bIrs2+/– or bIrs2–/– mice (Fig. 3, B and C). Oxygen consumption by old fIrs2 and bIrs2+/– mice declined to the same amount, whereas the bIrs2–/– mice consumed slightly less oxygen (Fig. 3B). All the mice were less active at 22 months; however, the old bIrs2+/– and bIrs2–/– mice were about twice as active as the old controls (Fig. 3C).

Next we determined the respiratory quotient (Rq = VCO2/VO2, where V is volume) to estimate the daily transition between fat (Rq = 0.7) and carbohydrate (Rq = 1) oxidation (19). The Rq for all the young mice displayed the usual diurnal rhythm, which approached the maximum value during the dark cycle (Fig. 3D). By contrast, old fIrs2 mice lost the diurnal rhythm; the Rq was indistinguishable between light and dark cycles (Fig. 3D). However, the Rq for old bIrs2+/– and bIrs2–/– mice increased significantly during the dark cycle, revealing a more youthful transition between fat and carbohydrate oxidation (Fig. 3D). Indeed, healthy long-lived humans also display a higher Rq that is closer to the value of healthy middle-aged adults (20).

Oxidative stress is associated with a reduced life span, and many enzymes protect cells from oxidative stress, especially superoxide dismutase (Sod) (3, 4, 21). In mice, FoxO1 promotes the expression of Sod2, so we investigated whether reduced neuronal Irs2 might help maintain brain Sod2 concentrations during feeding. Sod2 and FoxO1 protein was measured by immunoblotting hypothalamic lysates from young and old mice before and 2 hours after feeding (Fig. 4A). The Sod2 concentrations were not changed by feeding the young mice; however, Sod2 decreased at least 50% in the old fed control mice (Fig. 4, A and B). By contrast, Sod2 was not reduced by feeding old bIrs2+/– and bIrs2–/– mice (Fig. 4, A and B). A similar pattern was observed for FoxO1 levels in young and old hypothalamic tissues (Fig. 4, A and C). Thus, hypothalamic Sod2 reveals a more youthful response to feeding in mice with reduced brain Irs2.

Fig. 4.

Loss of brain Irs2 stabilizes Sod2 in the postprandial brain. (A) Hypothalamic lysates were prepared from pairs of male siblings of the indicated genotype before (Fast) or after 2-hour feeding (Fed), resolved by SDS polyacrylamide gel electrophoresis, and immunoblotted with antibodies against Sod2 or FoxO1 (two independent experiments are shown). β-tubulin (shown for one experiment) was immunoblotted for all the experiments to confirm equivalent loading. Autoradiographs were quantified, and the ratio of intensities (Fed/Fast) for (B) Sod2 (n = 5) or (C) FoxO1 (n = 4) was calculated. Boxes show the median ratio (solid horizontal line) and the lower and upper quartiles; the Kruskal-Wallis nonparametric test was used to compare the groups across all genotypes (*P < 0.05).

Together, our results show that reduced Irs2 signaling in all tissues (Irs2+/–) or just in the brain (bIrs2+/–) increases the life spans of mice maintained on a high-energy diet about 5 months and about 4 months in bIrs2–/– mice (table S4). Some studies show that calorie restriction, reduced body size, and increased peripheral insulin sensitivity extend mammalian life span (5, 22). However, our long-lived mice are slightly larger and consume about the same or slightly more food than the short-lived controls. Indeed, long-lived systemic Irs2+/– mice are more insulin sensitive and glucose tolerant than WT mice; however, long-lived brain-specific bIrs2+/– and bIrs2–/– mice are insulin resistant, hyperinsulinemic, and glucose intolerant. The mechanism responsible for this disparity is unknown. Regardless, our results point to the brain as the site where reduced insulin-like signaling can have a consistent effect to extend mammalian life span—as it does in C. elegans and D. melanogaster (1, 3).

As mammals age, compensatory hyperinsulinemia usually develops to maintain glucose homeostasis and prevent the progression toward life-threatening type 2 diabetes (6); however, increased circulating insulin might have negative effects on the brain that can reduce life span (4, 21, 23). By directly attenuating brain Irs2 signaling, an aging brain can be shielded from the negative effects of hyperinsulinemia that ordinarily develop with overweight and advancing age. Consistent with this hypothesis, moderate daily exercise, calorie restriction, and weight loss—which reduce circulating insulin—might increase life span by attenuating Irs2 signaling in the brain. Other strategies that improve peripheral insulin sensitivity, such as reduced growth hormone signaling, could have the same effect (5). Indeed, human centenarians display increased peripheral insulin sensitivity and reduced circulating insulin concentrations (23). Hence, we suggest that the Irs2 signaling cascade in the brain integrates the effects of peripheral nutrient homeostasis with life span.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 to S4


References and Notes

View Abstract

Navigate This Article