Ablation of AMP-activated protein kinase α1 and α2 from mouse pancreatic beta cells and RIP2.Cre neurons suppresses insulin release in vivo
- First Online:
- Cite this article as:
- Sun, G., Tarasov, A.I., McGinty, J. et al. Diabetologia (2010) 53: 924. doi:10.1007/s00125-010-1692-1
AMP-activated protein kinase (AMPK) is an evolutionarily conserved enzyme and a target of glucose-lowering agents, including metformin. However, the precise role or roles of the enzyme in controlling insulin secretion remain uncertain.
The catalytic α1 and α2 subunits of AMPK were ablated selectively in mouse pancreatic beta cells and hypothalamic neurons by breeding Ampkα1 [also known as Prkaa1]-knockout mice, bearing floxed Ampkα2 [also known as Prkaa2] alleles (Ampkα1 −/− ,α2 fl/fl ,), with mice expressing Cre recombinase under the rat insulin promoter (RIP2). RIP2 was used to express constitutively activated AMPK selectively in beta cells in transgenic mice. Food intake, body weight and urinary catecholamines were measured using metabolic cages. Glucose and insulin tolerance were determined after intraperitoneal injection. Beta cell mass and morphology were analysed by optical projection tomography and confocal immunofluorescence microscopy, respectively. Granule docking, insulin secretion, membrane potential and intracellular free Ca2+ were measured with standard techniques.
Trigenic Ampkα1 −/− ,α2 fl/fl expressing Cre recombinase and lacking both AMPKα subunits in the beta cell, displayed normal body weight and increased insulin sensitivity, but were profoundly insulin-deficient. Secreted catecholamine levels were unchanged. Total beta cell mass was unaltered, while mean islet and beta cell volume were reduced. AMPK-deficient beta cells displayed normal glucose-induced changes in membrane potential and intracellular free Ca2+, while granule docking and insulin secretion were enhanced. Conversely, βAMPK transgenic mice were glucose-intolerant and displayed defective insulin secretion.
Inhibition of AMPK activity within the beta cell is necessary, but not sufficient for stimulation of insulin secretion by glucose to occur. AMPK activation in extrapancreatic RIP2.Cre-expressing cells might also influence insulin secretion in vivo.
KeywordsAMPK Beta cell Insulin secretion Islet Knockout RIP.Cre
AMP-activated protein kinase
Ampkα1 −/− ,α2 fl/fl mice expressing Cre recombinase (Ampkα1 −/− ,α2 fl/fl ,Cre +)
Liver kinase B1
Mechanistic target of rapamycin
Rat insulin promoter
AMP-activated protein kinase (AMPK) is an evolutionarily conserved fuel-sensitive protein kinase implicated in the control of glucose homeostasis and playing roles in insulin-sensitive tissues [1–3] and in the pancreatic beta cell [4–7]. While the stimulation of AMPK activity in muscle and liver is now seen as a likely mechanism through which glucose-lowering agents, including metformin and thiazolidinediones, act to improve insulin sensitivity , the long-term effects of these agents on pancreatic beta cell survival and insulin release are less clear .
Mammalian AMPK is a trimeric protein comprising a catalytic α-subunit, encoded by one of two separate genes (Ampkα1 [also known as Prkaa1] and Ampkα2 [also known as Prkaa2]), a scaffold β- (β1 or β2) subunit and a regulatory γ-subunit (γ1, γ2 or γ3) [9, 10]. The existence of two separate AMPKα subunit genes has so far hindered investigations of the role AMPK activity plays in controlling glucose homeostasis in mammals, since the unconditional deletion of both isoforms leads to early embryonic lethality in mice . By contrast, animals with global inactivation of the AMPKα1 isoform do not display significant metabolic abnormalities [11, 12]. Deletion of the Ampkα2 gene leads to insulin resistance and glucose intolerance, in part due to increased parasympathetic tone . Whereas insulin secretion was normal in islets isolated from whole-body Ampkα2 knockout mice, insulin release in vivo appeared to be diminished when measured at a single time point during oral glucose tolerance tests. However, no measurements were made during intraperitoneal glucose tolerance tests, where the complicating effects of potentially altered incretin release could be excluded. Importantly, since complexes containing the AMPKα1 isoform are substantially (>tenfold) more abundant in beta cell lines than AMPKα2 complexes , increases in expression of the former may also, in part, have compensated for the loss of AMPKα2.
The cell-permeant AMPK activator, 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR), diminishes glucose-stimulated insulin secretion from clonal beta cells and islets [4, 6], an effect mimicked by the biguanide metformin  or by the expression of a constitutively active form of AMPK . Moreover, overexpression of constitutively active AMPK diminished the performance of islets transplanted into streptozotocin-induced diabetic mice . These actions may be due in part to blockade of secretory granule transport to the cell surface  as a result of kinesin-1 light chain phosphorylation . In addition, AMPK activation decreases beta cell viability [18, 19] possibly by phosphorylating the cell cycle regulator, p53 . By contrast, inhibition of AMPK activity with dominant-negative forms of the enzyme tends to increase insulin secretion at low glucose concentrations  without affecting release at elevated glucose concentrations , a finding that is consistent with inactivation of the enzyme as a result of AMP depletion under these conditions.
AMPK is also thought to play an important role in the central control of feeding and glucose homeostasis. Thus, forced changes in AMPK in the ventromedial hypothalamus achieved by stereotactic injection of viral vectors led to marked changes in food intake and body weight  and in hypoglycaemia sensing [23, 24]. Moreover, the deletion of Ampkα2 in neurons expressing pro-opiomelanocortin or agouti-related peptide leads respectively to increased or decreased food intake in mice . Finally, manipulation of AMPK activity ex vivo was found to affect glucose-induced changes in the electrical activity of Ca2+ transients in isolated agouti-related peptide (glucose-inhibited), but not in pro-opiomelanocortin (glucose-responsive) neurons in culture [26, 27].
We have previously demonstrated distinct roles in the pancreatic beta cell for AMPK complexes containing differing catalytic subunits . Thus, the α2 subunit, which displayed substantial nuclear localisation [4, 5], was implicated in control of gene expression, whereas α1-containing complexes, whose total activity exceeded that of α2-containing complexes by five to ten fold , were almost exclusively cytosolic. The role of the latter remains unclear, but may include regulation of plasma membrane ion channels, as proposed in oxygen-sensing cells in the carotid body .
To assess the role of AMPK in insulin-expressing cells, we generated trigenic mice that were globally inactivated for Ampkα1 and had Ampkα2 deleted selectively in pancreatic beta cells and in a small population of hypothalamic neurons, using the rat insulin promoter (RIP)2  and transgenic mice expressing activated AMPK selectively in beta cells. Using these models we highlight multiple novel mechanisms through which AMPK controls insulin production and glucose homeostasis in mammals.
Generation of mutant mice lacking Ampkα1 globally and selectively lacking Ampkα2 in pancreatic beta cells and RIP2.Cre neurons
Ampkα1 +/− ,α2 fl/fl (fl/fl) mice were first crossed with wild-type C57BL/6 mice to generate double heterozygous Ampkα1 +/− ,α2 fl/+ mice. Offspring were then crossed with heterozygous RIP2.Cre + transgenic mice (expressing Cre recombinase under the RIP2; Jackson Laboratory, Bar Harbor, ME, USA). The resulting triple heterozygous AMPKα1 +/− ,α2 fl/+ ,Cre + mice were interbred with their siblings. Since RIP2.Cre transgenic mice have been suggested to show glucose intolerance and impaired insulin secretion [30, 31], Cre + mice were always used as a negative control. Due to the low probability (1:64) of obtaining double knockout mice through heterozygote crossing, two different breeding strategies were used to obtain double AMPK knockout mice and their littermate controls. First, to generate Ampkα1-knockout mice, bearing floxed Ampkα2 alleles (Ampkα1 −/− ,α2 fl/fl ) and expressing Cre recombinase (βAMPKdKO) and their heterozygous Ampkα1 +/− ,α2 fl/+ ,Cre + mice controls, Ampkα1 −/− ,α2 fl/+ ,Cre − and Ampkα1 +/− ,α2 fl/fl ,Cre + mice were crossed. Second, to produce heterozygous Ampkα1 +/−,α2 fl/+,Cre + and their wild-type Ampkα1 +/+,α2 +/+,Cre + littermate controls, Ampkα1 +/− ,α2 fl/+ ,Cre + and wild-type mice were crossed. All mice were kept on a C57BL/6 background and offspring genotypes were obtained in the expected Mendelian ratios.
Generation of mutant mice selectively overexpressing AMPK.CA or -DN in pancreatic beta cells and RIP.Cre neurons
An expression vector containing the RIP2 promoter fragment (600 bp), c-myc-tagged rat AMPKα1312.T172D (CA) or AMPKα2.D157A (DN) cDNA, and an SV40 poly (A) cassette was excised with BssHII and microinjected into the male pronucleus of fertilised C57BL/6 oocytes. The injected zygotes were re-implanted into pseudo-pregnant female C57BL/6 mice (GenOway, Lyon Cedex, France). We obtained three AMPK.CA (named C1, C2 and C10) and two AMPK.DN (D1 and D2) founder mice that stably transferred the corresponding transgenes to their offspring. Founder mice were crossed with wild-type C57BL/6 mice to achieve F1 generation. Distributions of genotypes in the offspring followed a Mendelian pattern. All AMPK transgenic mice were kept heterozygous. F3 and later generations, and their littermate wild-type controls were used for experiments. All lines were maintained on a pure C57BL/6 background.
Mouse maintenance and diet
Mice were housed two to five animals per cage in a pathogen-free facility with 12 h light/dark cycle and had free access to standard mouse chow diet or a high-fat diet (60% [wt/wt] fat content; Research Diet, New Brunswick, NJ, USA). As indicated, 6-week-old mice were transferred to a high-fat diet for a maximum of 18 weeks. All in vivo procedures described here were performed at the Imperial College Central Biomedical Service and approved by the UK Home Office Animals Scientific Procedures Act, 1986.
Body weight and food intake
Fed mouse weights were monitored weekly after 6 weeks of age. Food intake was measured daily for three consecutive days using a metabolic cage.
In vivo physiological studies
Intraperitoneal glucose tolerance test
Mice fasted for 15 h (water allowed) were intraperitoneally injected with 1 g glucose/kg mouse weight. Blood from the tail vein was obtained at 0, 15, 30, 60, 90 and 120 min after injection. Blood glucose levels were measured with an automatic glucometer (Accuchek; Roche, Burgess Hill, UK). To study the effect of α-adrenergic antagonist on glucose tolerance, mice were intraperitoneally injected with 10 mg/kg mouse weight phentolamine 30 min before glucose challenge.
Plasma insulin measurement
Mice fasted for 15 h were intraperitoneally injected with 3 g glucose/kg mouse weight. Blood from mice tail veins was collected into a heparin coated tube (Sarstedt, Beaumont Leys, UK) at 0, 15 and 30 min after injection. Plasma was separated by centrifuging the blood at 2,000 g for 5 min. Plasma insulin levels were measured using an ultrasensitive mouse insulin ELISA kit (Mercodia, Uppsala, Sweden). Normal fed plasma insulin levels were measured from blood collected from tail veins of 12-week-old mice between 10:00 and 11:00 hours.
Insulin tolerance test
Bovine insulin (0.75 U/kg; Sigma, Dorset, UK) was intraperitoneally injected into fed mice between 13:00 and 14:00 hours. Blood glucose levels were measured at 0, 15, 30 and 60 min after injection.
Urine collection and catecholamine measurement
Daily urine collection from each mouse for a period of 3 days was performed using metabolic cages. Catecholamine levels in urine were determined by reverse-phase HPLC.
Details of islet isolation and insulin secretion, electron microscopy, electrophysiology, optical projection tomography, Ca2+ imaging, RNA extraction and RT-PCR, AMPK assay, antibodies and immunohistochemistry are provided in Electronic supplementary material (ESM 1).
Data are expressed as means ± SEM. Significance was tested by two samples unpaired or paired Student’s t test using Excel, or by ANOVA using Graphpad 4.0. A value of p < 0.05 was considered significant.
βAMPKdKO mice have normal body weight but are hyperglycaemic
βAMPKdKO mice have normal beta cell mass but smaller beta cells and islets
AMPK is a known regulator of the mechanistic target of rapamycin (mTOR) complex, acting to phosphorylate the mTORC1 components Raptor and the upstream regulator tuberous sclerosis complex-2 . Since mTOR is involved in the regulation of cell size , we assessed the size of individual beta cells, using anti-E-cadherin (Fig. 5d, e) or anti-GLUT2 (not shown) antibodies to label the plasma membrane. This revealed an ∼18% decrease in the average area of beta cells, corresponding to a decrease in volume of ∼36%, closely in line with the reduction in average islet size. However, no changes in the phosphorylation state of the downstream targets of mTOR, ribosomal S6 subunits, were detected (not shown), arguing against changes in the activity of the latter pathway being responsible for the decrease in beta cell volume. Unexpectedly, beta cell proliferation was also substantially (more than twofold) increased, as assessed by Ki67 staining (Fig. 5f, g). No changes in the low level of apoptosis could be detected by caspase-3 or in situ TUNEL staining (not shown), possibly reflecting rapid clearance of apoptotic cells.
Glucose-induced insulin secretion is enhanced in islets from βAMPKdKO mice but beta cell electrical activity and intracellular free Ca2+ changes are normal
To determine whether beta cells from βAMPKdKO mouse islets show altered glucose sensing, as observed after LKB1 deletion (G. Sun, A. Tarasov, J. McGinty, P.M. French, A. McDonald, I. Leclerc and G. Rutter, unpublished results), we used electrophysiological approaches. While glucose-induced changes in the conductance of ATP-sensitive K+ channels revealed lower conductance of these channels at low (3 mmol/l) but not elevated (16.7 mmol/) glucose in βAMPKdKO beta cells compared with heterozygous controls, this did not translate into a difference in membrane potential changes (Fig. 6b, c). Likewise, no differences were apparent in the extent of glucose or depolarisation- (KCl-) induced increases in intracellular free Ca2+ ions in heterozygous compared with βAMPKdKO beta cells (Fig. 6d–g). Correspondingly, again in contrast to the impact of LKB1 deletion (G. Sun, A. Tarasov, J. McGinty, P.M. French, A. McDonald, I. Leclerc and G. Rutter, unpublished results), no differences were observed in the levels or plasma membrane association of the liver/beta cell glucose transporter Glut2 (Fig. 6h) . By contrast, the number of morphologically docked granules was significantly increased in βAMPKdKO beta cells (Fig. 6i, j).
Overexpression of constitutively active AMPK in beta cells causes glucose intolerance
The above results suggested that, at extrapancreatic sites of RIP2.Cre expression, notably in the mediobasal hypothalamus, AMPK activity is permissive of insulin secretion in vivo.
Islets from transgenic mice overexpressing AMPK.CA (AMPK.CA transgenic) displayed a significant increase in total (α1 + α2 complex) AMPK activity at elevated (16.7 mmol/l), but not at low (2.8 mmol/l) glucose concentrations, where the endogenous enzyme was strongly activated (Fig. 8 g, h) . Conversely, AMPK activity was reduced at low glucose concentrations in islets from AMPK.DN transgenic mice, generated in parallel with AMPK.CA mice (Fig. 8g, h). Assuming approximately equal endogenous AMPK levels in beta compared with islet non-beta cells (ESM Fig. 1) and an islet beta cell content of 60% , AMPK activity was increased by 36% and reduced by 50%, in beta cells by overexpression of AMPK.CA or AMPK.DN respectively.
To determine whether the alterations in insulin secretion in AMPK.CA mice also result from defects at the level of the beta cell, we performed studies with isolated islets. Insulin release stimulated by 16.7 mmol/l glucose was decreased by >60% (Fig. 9k), a similar change to that observed after maintenance of mice for 18 weeks on high-fat diet. Indeed, islets from AMPK.CA transgenic mice maintained on high-fat diet displayed no further diminution in glucose-stimulated insulin secretion compared with wild-type littermates on the same diet (Fig. 9k). Conversely, glucose-stimulated insulin release was significantly enhanced in islets from AMPK.DN transgenic mice (Fig. 9l), reminiscent of the effect of deleting AMPKα1 and -α2 subunits (Fig. 6a), although in this case the improved secretion was eliminated in islets from animals fed a high-fat diet.
The principal aim of this study was to determine the physiological impact of complete and selective loss of both catalytic isoforms of AMPK from pancreatic beta cells. This is an important question, given the likely role of AMPK as a target for glucose-lowering drugs and uncertainties surrounding the role of AMPK in the beta cell [5, 6, 15, 18, 21, 45, 46].
Cell autonomous roles of AMPK within the beta cell
Examined here in isolated islets and beta cells, total inactivation of AMPK led to a potentiation of glucose-induced insulin secretion. We also noted an increase in ATP-sensitive K+ (KATP) channel activity at low glucose concentrations (Fig. 4b, c), perhaps reflecting altered trafficking of channel subunits . However, glucose-induced KATP channel closure and increases in intracellular Ca2+ were unaltered, suggesting that the loss of an inhibitory effect of AMPK-mediated phosphorylation of kinesin light chains [16, 17] and increased granule translocation to the plasma membrane (Fig. 6i, j) may underlie the enhanced secretion. Conversely, selective activation of AMPK in beta cells in βAMPK.CA transgenic mice decreased glucose-stimulated insulin secretion. These findings are consistent with previous results involving the overexpression in islets from of AMPK.CA animals in vitro , as well as with the effects of pharmacological activation of AMPK in clonal beta cells and isolated islets using AICAR or metformin [4, 6, 14, 48]. Furthermore, a diminished number of docked granules was observed in MIN6 cells overexpressing AMPK.CA .
We also show here that AMPK deletion leads to a decrease in cell volume, an effect strikingly different from that of deleting the upstream kinase, LKB1. Indeed, beta cells lacking LKB1 as a result of excision mediated either by RIP2.Cre (G. Sun, A. Tarasov, J. McGinty) or pancreas duodenum homeobox-1 promoter/Cre recombinase-oestrogen receptor (PDX-1-CreER) [38, 49] are larger than wild-type beta cells. LKB1 and AMPK appear therefore to engage substantially distinct downstream signalling pathways in beta cells, the former acting at least to a large extent via polarity-regulating kinase partitioning-defective 1/microtubule affinity-regulating kinase 2 (PAR1B/MARK2) [38, 49].
In the present study, inactivation of AMPK in RIP2.Cre neurons exerted no apparent effect on body mass or food intake. However, αAMPKdKO mice displayed defective glucose homeostasis due to abnormalities in the capacity of a preserved beta cell mass to secrete insulin in response to hyperglycaemia. This deficiency was only partially compensated for by an increase in insulin sensitivity, probably due to increased insulin receptor levels or enhanced downstream signalling in target tissues . Importantly, we used intraperitoneal injection of glucose, rather than oral administration to achieve increases in blood glucose concentration in the absence of a substantial release of incretins including glucagon-like peptide-1. In this way, we sought to compare the effects of an increase in circulating glucose in vivo with changes imposed on isolated islets. Nonetheless, we noted a dramatic decrease in insulin release in βAMPK.dKO mice in vivo, which was not apparent in vitro.
We considered the possibility that decreases in beta cell and islet size contribute to this deficiency in insulin secretion in vivo. However, the absence of impaired insulin release from islets in vitro (in fact, the opposite was observed) makes it unlikely that these changes are responsible for the drastically impaired insulin release observed in vivo in βAMPKdKO mice. It seems that the absence of the ‘opposite’ in vivo phenotype (i.e. improved glucose intolerance) in mice overexpressing activated AMPK under the same promoter (Fig. 9c, e, g, i) is most likely to reflect the predominant expression of the transgene in the pancreatic beta cell (rather than the hypothalamus) in adult mice. One possible explanation for these data is that a signal or signals emanating from RIP2.Cre neurons controls the activity of beta cells in vivo. Indeed, central administration of leptin has previously been shown to inhibit insulin secretion in vivo  and it is conceivable that this involves RIP2.Cre neurons and intracellular signalling pathways modulated by AMPK. Nevertheless, further and more definitive studies are needed to ascertain whether and by what means RIP.Cre neurons may influence insulin secretion.
A further interesting finding of the present studies is that mice lacking AMPK in the beta cell are somewhat less susceptible to the deleterious effects of high-fat diet on glucose metabolism and insulin secretion in vitro. These results are consistent with the previously demonstrated role of AMPK  in the actions of cytokines  on beta cell function and mass.
The results presented here suggest that activation of AMPK in hypothalamic neurons and in pancreatic beta cells play distinct roles in the control of insulin release in vivo. Our findings should inform the use and development of agents that act through AMPK to control glycaemia.
This work was supported by grants to G. A. Rutter from the Wellcome Trust (Programme Grant 081958/2/07/Z), The European Union (FP6 “Save Beta”), the Medical Research Council (G0401641) and National Institutes of Health (RO1 DK071962-01), as well as by a JDRFI Post-Doctoral Fellowship to A. I. Tarasov. We thank B. Kola (Queen Mary, University of London) for useful discussion, and P. Meda (University of Geneva) and B. Thorens (University of Lausanne) respectively for the kind provision of anti-ZO-1 and anti-GLUT2 antibodies. L. Lawrence is thanked for the preparation of pancreatic slices.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.