AMP-Activated Protein Kinase and Glycogen Synthase Kinase 3β Modulate the Severity of Sepsis-induced Lung injury
- 38 Downloads
Alterations in metabolic and bioenergetic homeostasis contribute to sepsis-mediated organ injury. However, how AMP-activated protein kinase (AMPK), a major sensor and regulator of energy expenditure and production, affects development of organ injury and loss of innate capacity during polymicrobial sepsis remains unclear. In the present experiments, we found that cross-talk between the AMPK and GSK3β signaling pathways controls chemotaxis and the ability of neutrophils and macrophages to kill bacteria ex vivo. In mice with polymicrobial abdominal sepsis or more severe sepsis induced by the combination of hemorrhage and intraabdominal infection, administration of the AMPK activator metformin or the GSK3β inhibitor SB216763 reduced the severity of acute lung injury (ALI). Improved survival in metformin-treated septic mice was correlated with preservation of mitochondrial complex V (ATP synthase) function and increased amounts of ETC complex III and IV. Although immunosuppression is a consequence of sepsis, metformin effectively increased innate immune capacity to eradicate P. aeruginosa in the lungs of septic mice. We also found that AMPK activation diminished accumulation of the immunosuppressive transcriptional factor HIF-1α as well as the development of endotoxin tolerance in LPS-treated macrophages. Furthermore, AMPK-dependent preservation of mitochondrial membrane potential also prevented LPS-mediated dysfunction of neutrophil chemotaxis. These results indicate that AMPK activation reduces the severity of polymicrobial sepsis-induced lung injury and prevents the development of sepsis-associated immunosuppression.
Severe infection accompanied initially by an overly exuberant inflammatory response and later by immunosuppression is frequently associated with dysfunction of vital organs and has a direct impact on morbidity and mortality in critically ill patients (1). Sepsis is the most frequent cause of death in hospitalized patients (2). Sterile inflammatory conditions linked to hemorrhage, trauma or burns worsen organ dysfunction in polymicrobial sepsis (3,4). Acute respiratory distress syndrome (ARDS) (5,6) frequently accompanies sepsis, and is associated with higher mortality rates in this setting (7). Effective pharmacologic interventions are not available for sepsis, a condition that affects more than a million patients each year in the United States (8). Similarly, there is no available pharmacologic intervention that improves the outcome from ARDS (9).
Excessive production of inflammatory mediators, including cytokines such as IL-1β and IL-17, disruption of endothelial and epithelial barriers with increased permeability, along with alterations in cellular bioenergetics and immunosup-pression appear to contribute to organ dysfunction and mortality in sepsis (1,10, 11, 12). While innate immune cells play a central role in host response to infection, exaggerated macrophage and neutrophil proinflammatory activation is also implicated in increased severity of sepsis-induced organ injury (13, 14, 15). The late or adaptive phase of sepsis is associated with apoptosis of lymphocytes and with epithelial and endothelial cell dysfunction as well as with diminished activation of neutrophils, macrophages and other cell populations involved in innate immunity. Such late phase immunosuppression appears to contribute to enhanced susceptibility to secondary infections that result in increased mortality (16,17). Previous experiments have shown that loss of mitochondrial structure and function in immune cells is implicated in organ failure in sepsis (18,19). Of note, the extent of mitochondrial dysfunction in the lungs has been shown to correlate with mortality in sepsis (19,20). Approaches to prevent mitochondrial dysfunction or to restore mitochondrial bioenergetics may diminish the severity of sepsis-associated lung injury (21, 22, 23).
The ability of the AMP-activated protein kinase (AMPK) to detect metabolic alterations and to modulate cellular bioenergetic and redox states appears to contribute to mortality and organ dysfunction in sepsis as well as to recovery from this life-threatening condition (24, 25, 26). AMPK is a heterotrimer that consists of one catalytic α and two regulatory β and γ subunits (27). This serine/threonine kinase has a unique mechanism of activation that is coupled to increases in energy demand, typically either due to excessive energy expenditure and/or deficient energy production. Such situations are associated with increased AMP-to-ATP ratios followed by AMP-dependent binding to the AMPKγ subunit, allosteric domain rearrangement and phosphorylation of T172-AMPKα by upstream kinases (28). Both AMP binding and phosphorylation of T172 are required for optimal AMPK activation. Activated AMPK participates in limiting energy expenditure while promoting pathways of energy production, including fatty acid oxidation, glycolysis and enhanced oxidative phosphorylation (24). Although enhanced AMPK activation induced by pharmacologic agents, such as metformin, is an important therapeutic approach to type 2 diabetes, recent studies also show that administration of metformin can retard aging in experimental models, and has been suggested to be associated with an increased lifespan of diabetic patients (29,30). In addition to the effects of AMPK activation on glucose and lipid metabolism, studies, including those from our laboratory, indicate that activated AMPK has antiinflammatory effects in TLR4-activated neutrophils and macrophages, and also diminishes the severity of endotoxin-induced lung injury in preclinical models (31, 32, 33).
Although sepsis is accompanied by alterations in bioenergetics of immune and parenchymal cells, as well as an increase in generation reactive oxygen and nitrogen species (ROS/RNS), that should result in AMPK activation, activation of AMPK is often not found in such settings (34, 35, 36, 37). More recently, we have shown that the Iκ B kinase beta (IKKβ)/glycogen synthase kinase beta (GSK3β) signaling axis contributes to preventing AMPK activation both after TLR4 engagement in neutrophils and macrophages, and in the lungs of mice subjected to sterile inflammatory conditions, including endotoxemia (38). However, it is not known whether this mechanism is operational during polymicrobial interabdominal infection, a clinically relevant issue in sepsis-induced ARDS. Because polymicrobial sepsis is linked to harmful inflammation and diminished capacity of the innate system for bacterial eradication, we also determined if AMPK activation contributes to subsequent development of immunosuppression. In particular, we examined whether AMPK activation affects clearance of P. aeruginosa lung infection following polymicrobial abdominal sepsis.
Materials and Methods
Male C57BL/6 mice were purchased from the National Cancer Institute in Frederick, Maryland. Mice 10 to 12 wks of age were used for experiments. The mice were housed in the animal facility at the University of Alabama at Birmingham. All experiments were conducted in accordance with protocols approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee.
Reagents and Antibodies
The GSK3β inhibitor BIO (6-bromoin-dirubin-3′-oxime) was purchased from R&D Systems whereas SB216763 was from Sigma-Aldrich. PS-1145 (IKK1/2 inhibitor), metformin (AMPK activator), LPS (TLR4 agonist) and brewer thioglycollate medium were obtained from Sigma-Aldrich. W-peptide (chemoattractant) was purchased from Phoenix Pharmaceuticals. Antibodies for phospho Thr172-AMPK, phospho Ser485-AMPK, total AMPK, phospho Ser79-ACC and total ACC were purchased from Cell Signaling Technology. HRP-conjugated β-actin antibody was from Santa Cruz Biotechnology. Anti-phospho-Thr479-AMPK antibody was generated as described previously (39) and was a gift from Ken Inoki of the University of Michigan. Anti-HMGB1 antibody was purchased from R&D Systems. Total OXPHOS Rodent WB Antibody Cocktail was obtained from Abcam. Hoechst dye and JC-1 probe were from Life Technologies. siRNA to the AMPKα1 subunit, scrambled siRNA and Accell medium were purchased from Thermo Scientific/Dharmacon.
Neutrophil and Peritoneal Macrophage Isolation
Bone marrow neutrophils were purified using a negative selection column (31,38). In brief, bone marrow cell suspensions were isolated from the femur and tibia of mice by flushing with RPMI 1640 medium with 5% fetal bovine serum (FBS). The cell suspension was passed through a glass wool column and collected by washing with phosphate buffer solution (PBS) containing 5% FBS. Negative selection to purify neutrophils was performed by incubation of the cell suspension with biotinylated primary Abs specific for the cell-surface markers F4/80, cluster of differentiation 4 (CD4), CD45R, CD5 and TER119 (Stem Cell Technologies) for 15 min at 4°C followed by subsequent incubation with antibiotin tetrameric Abs (100 µL; Stem Cell Technologies) for 15 min. The complex of antitetrameric Abs and cells was then incubated with colloidal magnetic dextran iron particles (60 µL; Stem Cell Technologies) for an additional 15 min at 4°C. The T cells, B cells, RBCs, monocytes and macrophages were captured in a column surrounded by a magnet, allowing the neutrophils to pass through. Neutrophil purity, as determined by Wright-Giemsa-stained cytospin preparations, was consistently >98%.
Peritoneal macrophages were isolated as described previously (38). Macrophages were elicited in 10- to 12-wk-old mice by use of Brewer thioglycollate. Cells were collected 4 d after intraperitoneal (IP) injection of thioglycollate, cultured for 3 d ex vivo and then treated as described in the figure legends.
Transwell Migration Assay
Bone marrow neutrophils (106 cells/well) were incubated with PS1145 (0 or 10 µmol/L), BIO (0 or 5 µmol/L), SB216763 (0 or 30 µmol/L) or metformin (0 or 500 µmol/L) for 2 h followed by exposure to LPS (300 ng/mL) for an additional 60 min. Transwell migration assay was performed using 24-well cell plate BD Falcon cell culture inserts (pore size 3 µm) (Translucent PET Membrane, BD Biosciences). Briefly, bone marrow neutrophils (106/mL) were placed into the upper reservoir, and chemotaxis initiated by inclusion of W-peptide (50 nmol/L) in the lower reservoir of transmigration chamber. Chemotaxis was determined after neutrophils were allowed to migrate for 60 min in RPMI media supplemented with FBS (5%). Cells in capillary structures of transmigrating chamber were subjected to Wright-Giemsa-staining followed by image acquisition using light microscopy. Each condition was tested three or more times using independent cell populations.
In Vitro Killing Assay
Neutrophils (5 × 106 cells/mL) cultured in RPMI 1640 (0.5% FBS) were pretreated with PS1145 (0 or 10 µmol/L) or SB216763 (0 or 30 µmol/L) for 60 min or metformin (0 or 500 µmol/L) for 2 h, followed by inclusion of Pseudomonas aeruginosa (PAK; 5 × 107/mL; 1:10 ratio neutrophil/PAK) for an additional 90 min. Similar to neutrophils, PS1145-,SB216763- or metformin-treated peritoneal macrophages (5 × 106 cells/mL) were incubated with PAK (macrophage/PAK; 1:10 ratio) for 90 min. The cell/bacterial solutions were centrifuged at 375g for 5 min, and then the cell pellets were lysed by adding 100 µL of Triton-X 100 (0.1%). The lysates were then plated on agar plates with ampicillin and incubated overnight at 37°C. Colony-forming units (CFUs) were calculated by counting bacterial colonies grown on agar plates using colony counter software (Bio-Rad) and expressed as a percentage of colonies obtained from untreated neutrophils or macrophages.
siRNA Knockdown of AMPKα1
Peritoneal macrophages were incubated with scramble (1 µmol/L) or AMPKα1-specific siRNA (1 µmol/L), as described previously (40). Briefly, cells (5 × 105/well) in 12-well plates were incubated in Accell medium (serum free) containing siRNA (1 µmol/L) for AMPKα1 for 72 h. Cells were then subjected to AMPK Western Blot analysis or exposure to GSK3β inhibitor followed by TNF-α enzyme-linked immunosorbent assay (ELISA).
Cecal Ligation and Puncture (CLP)-Induced Sepsis
CLP was performed in 10- to 12-wk-old male C57BL/6 mice as described before (41). Briefly, the cecum was ligated 1.0 cm from the tip of cecum, which was an approximately 50% cecum ligation. A through-and-through puncture was performed with a 21-gauge needle and then a drop of feces was extruded to the peritoneal cavity. Saline (0.9%; 500 µL) was then applied into the peritoneal cavity and the abdominal wall incision was closed in two layers. The control group of mice (sham) underwent surgery without CLP.
A Mouse Model of Hemorrhage and Resuscitation
Hemorrhage was performed using the previously described method (42). C57BL/6 male mice were anesthetized by inhalation of isoflurane (5%), and then both femoral arteries were cannulated with catheters (Braintree Scientific). The systemic arterial pressure line was continuously measured, independently from the hemorrhage/resuscitation catheter line. Blood withdrawal was performed for 60 minutes with a 25 ± 5 mmHg mean arterial pressure (MAP), typically a resultant of nearly 60% (~800 µL) blood loss. Next, mice were fully resuscitated with Hanks’ Balanced Salt solution (HBSS; Sigma-Aldrich) for 30 min. CLP procedure was conducted within 24 h, as described above.
Application of Metformin or GSK3β inhibitor SB216763 in Mice Subjected to Sepsis-Induced Lung injury
Mice were treated with metformin (100 mg/kg) or control (saline) IP applications in three doses; 48 h, 24 h and 30 min prior to CLP. In selected experiments, the second dose of metformin was given before hemorrhage. Mice were given the GSK3β inhibitor SB216763 (20 mg/kg) dissolved in 500 µL of DMSO/saline (1:40) or control vehicle (DMSO/saline 1:40) IP three times, that is, 48 h, 24 h and 30 min prior to CLP. Mice were euthanized 24 h after CLP, followed by preparation of lung homogenates for Western blot analysis, lung sections for H&E staining and collection of BAL fluids for cytokine ELISA. In particular, BAL fluids were collected by lavaging the lungs three times with 1 mL of PBS followed by measurement of inflammatory cytokines and protein content. Independent groups of mice were used to measure wet-to-dry ratios to determine the extent of pulmonary edema. In particular, after measuring the weight of freshly harvested (wet) lungs, the lungs were kept in an incubator for 7 d at 80°C. Next, the weight of dry lungs was measured followed by calculation of wet-to-dry ratio. Independent groups of mice were used to prepare lung homogenates in RIPA buffer (Sigma-Aldrich) followed by Western blot analysis of phosphorylated and total amounts of AMPK.
ELISA was used to measure cytokines in bronchoalveolar lavage (BAL) fluids. The amounts of tumor necrosis factor alpha (TNF-α), MIP-2, IL-6 and KC were determined by using commercially available ELISA kits (R&D Systems) according to the manufacturer’s instructions and as previously described (38,43).
Macrophage Endotoxin Tolerance Assay
Peritoneal macrophages (3 × 105/well) were first treated with LPS (0 or 10 ng/mL) for 24 h then media washed three times followed by incubation for an additional 60 min. Next, cells were exposed to a second stimulation with LPS (10 ng/mL) for 4 h. In selected experiments, macrophages were also treated with metformin (1 mmol/L), AICAR (500 µmol/L) or SB216763 (30 µmol/L) for 60 min followed by incubation with LPS (first stimulation) for an additional 24 h.
Measurement of Mitochondrial Membrane Potential (mΔΨ)
Bone marrow neutrophils were seeded 80% (confluent) in a 4-well chambered coverslip coated with fibronectin (40 µg/mL). The cells were left unaltered or treated with AICAR (250 µmol/L), metformin (500 µmol/L) or BIO (20 µmol/L) for 60 min followed by inclusion of LPS (300 ng/mL) for an additional 60 min. The JC-1 probe (100 ng/mL) and Hoechst (1 µg/mL) were applied 30 min before image acquisition. Microscopy was performed using a confocal laser scanning microscope (model LSM 710 confocal microscope; Carl Zeiss MicroImaging). Quantitative fluorescent intensity (red/green pixel intensity) of the images was processed using IPLab Spectrum. In an additional experiment, mitochondrial membrane potential was also measured after mitochondrial depolarization with FCCP (100 nmol/L).
In Vivo Bacterial Killing Assay
Briefly, mice were anesthetized with isoflurane and then suspended by their upper incisors on a 60° incline board. The tongue was gently extended and wildtype PAK strain of Pseudomonas aeruginosa (2.5 × 107/mouse) suspension in PBS (50 µL) or PBS alone (control; 50 µL) was deposited into the pharynx followed by bacterial aspiration into the lungs, similar to the method that was described previously (32). Lung homogenates were prepared 4 h after P. aeruginosa instillation and serial dilutions used to determine CFUs/mL.
Number of bacterial colonies grown on agar plates (CFUs) were measured using colony counter software (Bio-Rad).
Protein Concentration and Cell Counts in BAL Fluid
Briefly, protein concentration in BAL fluid was determined by Bradford method with Bio-Rad protein assay dye reagent concentrate (Bio-Rad). The numbers of neutrophils in BAL fluid were determined after cytospin and Wright-Giemsa staining followed by image acquisition using light microscopy.
Western Blot Analysis
Western blot analysis was performed as described previously (34,38). Each experiment was carried out three or more times with peritoneal macrophages or lung homogenates obtained from separate groups of mice. In selected experiments, BAL fluids (30 µL) were mixed with Laemmli sample buffer and boiled for 5 min followed by Western blot analysis of HMGB1.
Multigroup comparisons were performed using one-way analysis of variance (ANOVA) with Tukey post hoc test. Statistical significance was determined by the Student t test for comparisons between two groups. A value of P < 0.05 was considered significant. Analyses were performed on SPSS version 16.0 (IBM) for Windows (Microsoft).
All supplementary materials are available online at https://doi.org/www.molmed.org .
Participation of AMPK and GSK3β Signaling Pathways in Neutrophil- and Macrophage-Dependent Bacterial Killing
AMPK Activation by Metformin or through GSK3β Inhibition Decreased the Severity of Lung Injury following Polymicrobial Sepsis
AMPK Activation Diminishes the Severity of Lung injury following Hemorrhage and intraabdominal Sepsis
AMPK Activation Prevents Dissipation of Mitochondrial ATP Synthase (Complex V) and increases the Amounts of ETC Complexes III and IV in the Lungs of Septic Mice
GSK3β Regulates AMPK Activity in the Lungs of Septic Mice
These results suggest that metformin activates a specific AMPK pool not affected by GSK3β-dependent inhibition in the lungs of septic mice (Figure 5G).
AMPK Activation Prevents Macrophage Reprogramming into an Endotoxin Tolerogenic Phenotype
AMPK Activation Diminishes HIF-1α Production in Macrophages
Recent studies have shown that enhanced expression of hypoxia-inducible factor 1α (HIF-1α) participates in sepsis-induced immunosuppression (50). Thus, we investigated whether AMPK activation affects HIF-1α accumulation in LPS-treated macrophages. As shown in Figures 6C and D, incubation of macrophages with metformin, AICAR or the GSK3β inhibitor SB216763 resulted in diminished LPS-induced accumulation of HIF-1α. These results suggest that cross-talk between AMPK and GSK3β may be implicated in regulating development of macrophage immunosuppressive phenotypes.
AMPK and GSK3β Signaling Pathways Participate in Modulating Neutrophil Chemotaxis
AMPK Activation Diminished the Onset of Immunosuppression in Mice with Sepsis
In this study, we found that crosstalk between AMPK and GSK3β was involved in regulating lung inflammation and development of lung injury in experimental models of polymicrobial abdominal sepsis or by the more severe combination of hemorrhage and abdominal sepsis. The AMPK activator metformin and the GSK3β inhibitor SB216763 prevented the decrease in neutrophil chemotaxis induced by LPS, and also enhanced the ability of neutrophils and macrophages to kill bacteria. In vivo treatment with metformin improved survival of mice with polymicrobial abdominal sepsis, stabilized mitochondrial complex V and increased the amounts of mitochondrial complexes III and IV. Although activated AMPK diminished production of proinflammatory mediators in LPS-treated macrophages, this event was not associated with diminished bacterial killing. Indeed, metformin or SB216763 effectively prevented development of LPS-induced macrophage immunosuppressive phenotypes. Similarly, metformin increased bacterial clearance in the lungs of mice with sepsis.
In spite of increases in AMP-to-ATP ratios and ROS formation, which normally result in AMPK activation, there was no increase in AMPK activity in the experimental models of ARDS or in critically ill patients (35, 36, 37). Recent studies demonstrated that interactions between PI3K/AKT and GSK3β as well as between IKKβ and GSK3β promoted direct phosphorylation and inactivation of AMPK, thereby suggesting potential mechanisms for the lack of AMPK activation in preclinical models of sepsis and in patients with critical illness (38,39). Such findings are consistent with the ability of GSK3β inhibitors to diminish monocyte proinflammatory activation and to reduce mortality in experimental sepsis, ischemic organ injury and endotoxin-induced lung injury (38,55,56).
Previous studies, including results obtained in our laboratory, have shown that enhanced AMPK activation diminished the severity of lung injury in experimental models of sterile inflammation, such as after exposure of the lungs to LPS or peptidoglycan (PGN) (31,32,36). Our new data indicate that AMPK activation also decreases pulmonary injury in the setting of polymicrobial sepsis. Of note, despite the inhibitory activity of AMPK activation on proinflammatory cytokine release in the lungs, bacterial clearance was increased in the lungs of septic mice treated with metformin and in mice with Pseudomonas aeruginosa pneumonia. It is important to note that while moderate inflammation is necessary to initiate the accumulation of immune cells to sites of infection, diminishing exaggerated and deleterious proinflammatory activation is not necessarily associated with loss of innate immune function (1,8,17). For example, a recent study has shown that mice treated with the specific NF-κB inhibitor BMS-345541 had reduced severity of lung injury following CLP-induced sepsis (57). Similarly, we found that treatment with an IKK1/2 inhibitor or activation of AMPK by metformin or GSK-3β inhibitors had no adverse effects on bacterial killing by neutrophils or macrophages ex vivo. Of note, activation of AMPK has been shown to increase phagocytosis in neutrophils and macrophages (58, 59, 60, 61).
Our results show that AMPK activation provided substantial protection against sepsis-induced lung injury. However, the exact role of AMPK that plays during recovery of immune and peripheral tissue homeostasis needs to be further examined. A possible mechanism by which AMPK activation may modulate acute inflammatory responses, such as sepsis-induced lung injury, is linked to cellular bioenergetics. Previous studies have shown that sepsis-mediated organ injury was associated with alterations in mitochondrial structure and function (19). Mitochondrial impairment in peripheral tissues and in immune cells has been correlated with morbidity and mortality associated with sepsis (18,62). For example, significant loss of ATP synthase (complex V) has been found in circulating monocytes in patients with sepsis, and is likely to participate in disrupted immune bio-energetic homeostasis (63). Our results showed that AMPK activation effectively prevented loss of ATP synthase (ETC complex V) in the lungs of septic mice. Of note, besides direct immunoregulatory action, AMPK activation is also involved in preservation of epithelial and endothelial bioenergetics and in recovery of lung tissue homeostasis, including restoration of intercellular connections (33,36). A recent study indicates that AMPK activation in the brain also diminished LPS-mediated development of sepsis-ALI, evidence for more diverse mechanism of AMPK action (64). Of note, while AMPK is an established metabolic sensor and regulator, MKK3 signaling axis has been also shown to affect mitochondrial function in sepsis/ALI (65).
Diminished macrophage and neutrophil proinflammatory activation, as well as T-cell exhaustion, is characteristic of the immunosuppression described in late sepsis (16,66). While engagement of TLR4 in macrophages and neutrophils diminishes AMPK activity (36,38), HIF-1α has been shown to promote an immunosuppressive status in monocytes during human sepsis (50). Our results indicate that AMPK activation prevented both accumulation of HIF-1α and development of endotoxin tolerance in LPS-treated macrophages. These new findings are similar to the ability of AMPK activation to inhibit HIF-1α expression in cancer cells and insulin- and IGF-1-induced expression of HIF-1α in endothelial cells (67,68).
Our data suggest that therapeutic interventions that induce AMPK activation may be beneficial in diminishing organ dysfunction, enhancing bacterial clearance and improving survival in severe polymicrobial sepsis. Although this hypothesis is primarily supported through the use of metformin as an AMPK activator, similar results were found using other AMPK activators, including GSK-3β inhibitors, as performed in the present experiments. Previous studies have suggested that metformin can be used safely in patients with critical illness, COPD or asthma (48, 49, 50). More than 50 million type 2 diabetics are taking metformin daily worldwide, and recently metformin was selected for a clinical trial to evaluate its effect on human longevity (https://doi.org/ClinicalTrials.gov NCT02432287) (69).
Given the safety of metformin in many humans with diverse pathophysiologic conditions, and the suggestion that metformin may have beneficial effects in diminishing inflammation-associated organ dysfunction, it may be appropriate to consider its use in clinical trials enrolling severely ill septic patients with organ dysfunction.
The authors declare that they have no competing interests as defined by Molecular Medicine, or other interests that might be perceived to influence the results and discussion reported in this paper.
We thank Ken Inoki from the University of Michigan for the anti-phospho-Thr479-AMPK antibody. Funding was provided by National Institutes of Health Grant HL107585 to JW Zmijewski.
- 6.Force ADT, et al. (2012) Acute respiratory distress syndrome: the Berlin Definition. JAMA. 307:2526–33.Google Scholar
- 61.Labuzek K, Liber S, Gabryel B, Adamczyk J, Okopien B. Metformin increases phagocytosis and acidifies lysosomal/endosomal compartments in AMPK-dependent manner in rat primary microglia. Naunyn Schmiedebergs Arch. Pharmacol. 381:171–86.Google Scholar
- 69.Metformin in Longevity Study (MILES) [Internet]. (2016) [updated 2015 Dec 8; cited 2016 Jan 29]. Available from: https://doi.org/clinicaltrials.gov/ct2/show/record/NCT02432287ClinicalTrials.gov identifier: NCT02432287.