Exercise activates AMPK in mouse and human pancreatic islets to decrease senescence

Abstract

Beta (β)-cell senescence contributes to type 2 diabetes mellitus (T2DM). While exercise is vital for T2DM management and significantly affects cellular ageing markers, its effect on β-cell senescence remains unexplored. Here, we show that short-term endurance exercise training (treadmill running, 1 h per day for 10 days) in two male and female mouse models of insulin resistance decreases β-cell senescence. In vivo and in vitro experiments revealed that this effect is mediated, at least in part, by training-induced increases in serum glucagon, leading to activation of 5'-AMP-activated protein kinase (AMPK) signalling in β-cells. AMPK activation resulted in the nuclear translocation of NRF2 and decreased expression of senescence markers and effectors. Remarkably, human islets from male and female donors with T2DM treated with serum collected after a 10-week endurance exercise training programme showed a significant decrease in the levels of senescence markers. These findings indicate that exercise training decreases senescence in pancreatic islets, offering promising therapeutic implications for T2DM.

Extended Data Fig. 1 ∣. Distance run per day.

Extended Data Figure 1 refers to Fig. 1. Distance (kilometers) run per day for young 6-weeks-old C57BL/6 females (a, n=5-11/group) and for retired breeders, male 6–9 months-old (b, n=5-10/group) in the wheel-running group (trained). Each line represents an individual mouse.

Extended Data Fig. 2 ∣. Body composition parameters.

Extended Data Figure 2 refers to Figs. 1 and 3. (a) Body weight in grams of female trained mice (6 weeks of age, n=5-11/group) from control and trained (wheel) groups. Relates to Fig. 1a and b. (b) Body weight in grams of sedentary and trained (treadmill) at the end of the protocol for 12 -weeks-old (n=10 animals/group, males), relates to Fig. 1d-h; (c) Representative pictures from DEXA Fat and lean mass were measured by dual-energy X-ray absorptiometry (DEXA) (d-g). (h) Body weight in grams of sedentary and trained (treadmill) animals sedentary and control with insulin resistance induced by S961 (males, n=5/group, 6–9 months-old), relates to Fig. 1i-l; (i) Body weight in grams of sedentary and trained (treadmill) animals sedentary and control (males, n=5/group, 6–9 months-old); relates to Fig. 3a-c.N= 5 animals/group, males, 6–9 months old. Two sided-test and Welch’s correction. Mean ± SD. *P<0.05.

Extended Data Fig. 3 ∣. Glucose and insulin tolerance tests.

Extended Data Figure 3 refer to Fig. 1d-h. AUC for ITT (a) in 12-weeks-old C57BL/6 males after treadmill exercise protocol; n=10/group (a), serum levels from S961 treated mice, n=4-5/group (b), AUC and blood glucose levels for IPGTT (c, e) and ITT (d, f) for trained and sedentary 6–9 months-old animals (n=5/group) from reversibility prevention (c and d) and reversibility groups (e and f). N=4-10/group. Two sided-test and Welch’s correction. Mean ± SD. ***P<0.001.

Extended Data Fig. 4 ∣. Glucose, IL6 and insulin response.

Extended Data Figure 4 refers to Figs. 4 and 5. Glucose levels (a) and concentration of IL-6 (b) after exercise-protocol (6–9-month-old C57BL/6 male mice; n=10/group in A and n=5/group in B). (c) insulin from glucose stimulated insulin secretion performed in islets from male mice (C57BL6, 3–6 months-old, cultured in presence of glucagon and trained serum. N=5 animals/group. Two sided-test and Welch’s correction. Mean ± SD. *P<0.05, **P<0.01, ****P<0.0001.

Extended Data Fig. 5 ∣. AMPK activation and P21CIP1 levels in islets treated with human from exercise participants without T2D.

Extended Data Figure 5 refers to Fig. 8. Experiment conducted in individuals without T2D. The results showed increased AMPK phosphorylation (A and B) but without changes in senescence marker P21CIP1 (A and C). Samples from individuals ranging from 22 to 57 years old, male, n=6/group. Two sided-test and Welch’s correction. Mean ± SD. *P<0.05.

Fig. 1 ∣. Exercise training reduces β-cell senescence in mice.

a, Islets from female trained mice (6 weeks of age, n = 3–7 per group) fed an HFD showed decreased β-gal⁺ cells compared with the control–sedentary group (range of absolute values, 46–88). b, Exercise training restored insulin secretion during GSIS in both the control and HFD groups. c, Secretion index (insulin secreted at 16.8 mM glucose/insulin secreted at 2.6 mM glucose) from the islets. Statistical significance refers to a value of 1, which indicates glucose unresponsiveness, by one-sample t and Wilcoxon tests. d–h, Male C57BL/6 mice (12 weeks of age, n = 5 per group) trained on a treadmill for 10 days presented an increase in glucose clearance during IPGTT (d), decreased fasting basal insulin concentration (e), lower fold change in β-gal⁺ cells (f; range of absolute values, 48–91) and decreased gene expression of p21Cip1 (g) without changes in p16Ink4a (h). i–l, Male mice (6–9 months old, n = 5 per group) were infused with the insulin receptor antagonist S961 through an osmotic pump for 14 days. Exercise training during this time restored GSIS (i) without changes in β-gal⁺ cells (j; range of absolute values, 43–81) or p21Cip1 expression (k) but with a significant decrease in p16Ink4a mRNA expression (l). Data were analysed using one-way Brown–Forsythe ANOVA, and a two-sided t test with Welch’s correction was used to compare two groups. Blood glucose levels over time were analysed using two-way ANOVA. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ****P < 0.0001. AUC, area under the curve; FC, fold change; NS, no significance.

Fig. 2 ∣. Prevention of β-cell senescence outweighs its reversal.

a, Workflow of placing 6- to 9-month-old male C57BL/6J mice into prevention and reversibility models. For prevention studies, animals were first exercised for 10 days and then administered S961 through an osmotic pump for 14 days. For reversibility studies, animals received S961 for 14 days, followed by exercise for 10 days. b–e, Islets from male mice (6–9 months of age, n = 3–5 per group) trained as a prevention strategy displayed restored β-cell responsiveness to glucose (b), no changes in the percentage of β-gal⁺ cells (c; range of absolute values, 77.7–95.2), decreased expression of p21Cip1 (d) and no changes in p16Ink4a expression (e). f–i, In the reversibility study in male mice (6–9 months old, n = 4–5 per group), no differences were observed in GSIS (f), percentage of β-gal⁺ cells (g; range of absolute values, 5.78–46.3) or p21Cip1 gene expression (h); however, there was a substantial decrease in p16Ink4a levels (i). j,k, Violin plot (j) and heatmap (k) of RNAseq results showing reduced expression of the SenMayo senescence gene set in pancreatic islets from male C57BL/6 mice (6–9 months old, n = 5 per group) trained by treadmill running for 10 days compared with sedentary mice. l–n, Male and female INK-ATTAC (4–14 months old, n = 5 per group; seven males and three females) transgenic mice matched for age were trained by treadmill running for 10 days. They showed a lower basal insulin concentration (l) and increased glucose clearance during an ITT (m) with no changes during an IPGTT (n). o,p, Immunohistochemical detection of FLAG in islets from sedentary and trained INK-ATTAC animals showed that exercise training decreased p16Ink4a in β-cells (o), as reflected by a higher percentage of islets with no or low staining for FLAG in the pancreas of trained animals, indicating lower cellular senescence (male and female mice, 4–14 months old, islets from three to five mice per group) (p). A two-sided t test with Welch’s correction was used to compare two groups. Blood glucose levels over time were analysed using two-way ANOVA. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Panel a created with BioRender.com.

Fig. 3 ∣. Exercise training activates AMPK in pancreatic islets.

a–c, Immunostaining for pAMPKα (Thr172) (a) in the pancreas of mice (6–9 months old, male, n = 5 per group), sedentary or trained by treadmill running for 10 days. Training increased the percentage of the staining intensity (b) and the positively stained islet area (c). Owing to high inter-islet heterogeneity, results are shown per islet from five individual animals. d–h, Islets were isolated from male mice (6–9 months old, n = 5 per group) less than 12 h after the last bout of exercise. Western blotting (d) showed increased phosphorylation of AMPK (e) and ACC (f) but no increase in the phosphorylation of Raptor (g) in the trained group compared with the sedentary group. LKB1 levels were increased in the pancreatic islets of trained animals compared with sedentary ones (h). i–k, Mouse (6–9 months old, male, n = 7 for qPCR analysis and n = 10 for β-gal expression) and human (nondiabetic, 37–62 years old, male and female, n = 8 for qPCR and n = 11 for β-gal expression) islets were treated in vitro with the AMPK activator C991. Significance was calculated based on untreated mouse or human islets (C, control). The results revealed decreased β-gal⁺ cells only in human islets after 4 days in culture (i; absolute value range 3.78–46.3); p21Cip1 gene expression was decreased in both mouse and human islets (j), and p16Ink4a expression was decreased only in mouse islets (k). l–q, Mouse (6–9 months old, male, n = 4 per group) islets were incubated with adenovirus (Ad) carrying AMPK-DN or control adenovirus-GFP at an MOI of 100 viral particles per cell. The effectiveness of AMPK-DN was tested through western blotting (l), in which the knockdown of AMPK (m) was associated with increased P21CIP1 (n). Those islets were tested in the presence of glucagon (20 pM) or 10% trained serum in medium, and qPCR results showed that the introduction of AMPK-DN did not influence alterations in p21Cip1 mRNA expression (o). Nevertheless, it notably increased the expression levels of Trp53 (p) and Hmgb1 (q) across all conditions. Data were analysed using one-way Brown–Forsythe ANOVA, and a two-sided t test with Welch’s correction was used to compare two groups. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 4 ∣. Glucagon activates AMPK in mouse and human islets.

a, Serum from sedentary and trained male mice was used to treat naive islets from male C57BL/6 mice (6–9 months old, n = 5 per group) in culture medium for 4 days. b, Naive islets that received serum from trained animals had significantly decreased p21Cip1 mRNA expression compared with islets that received serum from sedentary animals (‘sedentary serum’). c,d, Heatmap (c) and violin plot (d) of RNAseq results demonstrating upregulation of genes related to GCGR activation (C57BL/6 mice, 6–9 months old, male, from sedentary and trained groups, n = 5 per group). e, Naive islets that received serum from trained animals had significantly higher intracellular cAMP than islets that received serum from sedentary animals. f, Acute exercise bouts in male mice exercised by treadmill running (6–9 months old, n = 4–5 per group) increased glucagon secretion after exercise training irrespective of insulin resistance status (green circles indicate serum samples from animals with S961-induced insulin resistance, whereas blue circles indicate serum samples from vehicle-treated animals). g–k, Western blot analysis of SDS–PAGE-separated proteins from mouse islets (6–9 months old, male, n = 6 per group) cultured with glucagon for 2 days (g) presented increased AMPK phosphorylation normalized to total AMPK protein expression (h), increased total AMPK (i), decreased P21CIP1 normalized to β-actin (j) and no changes in P16INK4A (k). l–p, Western blot analysis of SDS–PAGE-separated proteins from human islets (37–62 years old, male and female, n = 6 per group) cultured with glucagon for 4 days (l) presented increased AMPK phosphorylation (m) and total expression (n), decreased P21CIP1 normalized to β-actin (o) and no changes in P16INK4A (p). Analysis was performed using a two-sided t test with Welch’s correction. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001. Panel a created with BioRender.com.

Fig. 5 ∣. GLP-1R and GCGR activate AMPK and decrease senescence.

a–d, Analysis of circulating factors elevated during exercise training in serum from C57BL/6 mice (6–9 months old, male, n = 5 per group). Exercise acutely increased adrenaline (a), but no training effect was detected (b) when compared with the sedentary group. Elevation of GLP-1 was a training effect (c), and circulating glucagon levels remained elevated 12 h after a bout of exercise (d). e–h, Islets from male mice (6–9 months old, n = 4 per group) were incubated with either the GLP-1R agonist exendin-4 (10 nM) or glucagon (20 pM) for 48 h before protein analysis (e). Glucagon increased the activation of AMPK (f), and both GLP-1 and glucagon diminished the expression of the senescence marker P21CIP1 (g) with no changes in P16INK4A (h). i, Primary mouse islets (6–9 months old, male, n = 4 per group) were cultured in medium supplemented with trained serum in the presence of the GLP-1R antagonist exendin(9–39) (1 μM) or the GCGR antagonist MK0893 (1 μM) for 48 h before western blot analysis. j–l, Knockdown of Glp-1 and Gcgr diminished AMPK activation (j) and increased the expression of P21CIP1 (k) and P16INK4A (l). Data were analysed using one-way Brown–Forsythe ANOVA, and a two-sided t test with Welch’s correction was used to compare two groups. Values are presented as mean ± s.d. *P < 0.05, ***P < 0.001.

Fig. 6 ∣. Gcgr knockdown partially impairs exercise training effects.

a–g, Male C57BL/6 mice aged 6–9 months were given adomeglivant, a GCGR antagonist, during 2 weeks of treadmill training. Islets were collected and compared with those from trained mice without treatment with the antagonist (a). Blocking the GCGR with adomeglivant impaired AMPK activation during exercise training (b) without affecting total AMPK (c). The phosphorylation of the AMPK targets ACC (d) and Raptor (e) decreased. Blocking the GCGR reversed the effects of exercise training on pancreatic senescence markers, leading to a significant increase in P21CIP1 (f) and P16INK4A (g) expression. h–k, Islets isolated from male C57BL/6 mice aged 6–9 months were dispersed into single cells and transfected with lentivirus targeting Gcgr. A total of 600 islet equivalents were transplanted into the kidney capsule of male C57BL/6 mice (12 weeks old), categorizing them into sedentary and trained groups receiving either control (sh-control) or sh-GCGR islets (n = 5 per group). Following 2 weeks of treadmill exercise, we analysed gene expression in the grafts (h). Transfection efficacy was confirmed by bright field and green fluorescence microscopy (i) and by qPCR, showing a 69% decrease in Gcgr expression (j). Exercise training decreased p21Cip1 expression (k) regardless of Gcgr status. However, p16Ink4a expression (l) was reduced only in sh-control grafts, indicating the role of Gcgr in exercise-induced reduction in senescence. Data were analysed using one-way Brown–Forsythe ANOVA, and a two-sided t test with Welch’s correction was used to compare two groups. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ***P < 0.001. Panel h created with BioRender.com.

Fig. 7 ∣. Exercise training promotes the nuclear translocation of NRF2.

a,b, Western blot analysis (a) of SDS–PAGE-separated proteins from mouse islets (male, 6–9 months, n = 5 per group) showed that AMPK activation after exercise training was followed by increased NRF2 expression (b). c,d, Sedentary and trained animals (male, 6–9 months, n = 5 per group) were perfused immediately after the last bout of exercise, and the pancreas was immunostained for NRF2 (colocalization of DAPI and NRF2) and mounted with DAPI (c). ImageJ quantification showed increased nuclear intensity of NRF2 in trained animals compared with sedentary ones (d). e,f, Dispersed mouse islets (male, 6–9 months, n = 8 per group) were treated with siNRF2 to knock down the expression of Nrf2. qPCR analysis showed a 61% decrease in the expression of NRF2 in cells treated with siNRF2 compared with controls (e). Knockdown of Nrf2 led to a significant increase in p16Ink4 expression (n = 6 per group) (f). g–i, Primary islets from 6-month-old male mice (n = 4 per group) were cultured under various conditions: control medium (RPMI 1640 + 10% FBS), medium with glucagon (20 pM) or medium with trained serum (RPMI 1640 + 10% trained serum), with or without siRNA targeting Nrf2 for knockdown. Islets cultured with glucagon and trained serum showed decreased p21Cip1 expression (g). However, knocking down NRF2 hindered this decrease. Additionally, the expression of Trp53 (h) and Hmgb1 (i), other senescence markers, increased even in postexercise serum. Data were analysed using one-way Brown–Forsythe ANOVA, and a two-sided t test with Welch’s correction was used to compare two groups. Values are presented as mean ± s.d. *P < 0.05, **P < 0.01, ****P < 0.0001.

Fig. 8 ∣. Postexercise serum reduces senescence markers in T2DM islets.

a, Workflow diagram of the 10-week moderate-intensity training programme. Serum was collected before and after training, added to the culture medium with human islets and incubated for 4 days. b, Analysis of post-training serum demonstrated an increase in glucagon levels when compared with pretraining serum. c–g, Western blot analysis of SDS–PAGE-separated proteins (c) from human islets treated with pre- and post-training serum from volunteers revealed increased AMPK phosphorylation (d), with a decrease in P21CIP1 (e) and P16INK4A (f) protein content and an increase in LKB1 levels (g). h,i, qPCR analysis showed a decrease in P21CIP1 gene expression in the post-training group (h) and in P16INK4A expression (i). Experiments were performed once using multiple samples and in triplicate, with consistent results observed across replicates. j–l, Linear correlation between ΔHbA1c due to training and P21CIP1 mRNA levels (j), P16INK4A protein levels (k) and P21CIP1 protein levels (l). m,n, Linear correlation between changes in circulating glucagon levels due to training and P21CIP1 mRNA levels (m) and P21CIP1 protein levels (n). Data were analysed using a two-sided t test with Welch’s correction. Values are presented as mean ± s.d. *P < 0.05. Panel a created with BioRender.com.