Involvement of AMP-activated protein kinase in neuroinflammation and neurodegeneration in the adult and developing brain

Abstract

Microglial activation followed by neuroinflammation is a defense mechanism of the brain to eliminate harmful endogenous and exogenous materials including pathogens and damaged tissues, while excessive or chronic neuroinflammation may cause or exacerbate neurodegeneration observed in brain injuries and neurodegenerative diseases. Depending on conditions/environments during activation, microglia acquire distinct phenotypes, such as pro-inflammatory, anti-inflammatory, and disease-associated phenotypes, and show their ability to phagocytose various objects and produce pro-and anti-inflammatory mediators. Prevention of excessive inflammation by regulating the microglia’s pro/anti-inflammatory balance is important for alleviating progression of brain injuries and diseases. Among many factors involved in the regulation of microglial phenotypes, cellular energy status plays an important role. Adenosine monophosphate-activated protein kinase (AMPK), which serves as a master sensor and regulator of energy balance, is considered a candidate molecule. Accumulating evidence from adult rodent studies indicates that AMPK activation promotes anti-inflammatory responses in microglia exposed to danger signals or various stressors mainly through inhibition of the nuclear factor κB (NF-κB) signaling and activation of the nuclear factor erythroid-2-related factor-2 (Nrf2) pathway. However, AMPK activation in neurons exposed to stressors/insults may exacerbate neuronal damage if AMPK activation is excessive or prolonged. While AMPK affects microglial activation states and neuronal cell survival rates in both the adult and the developing brain, studies in the developing brain are still scarce, even though activated AMPK is highly expressed especially in the neonatal brain. More in depth studies in the developing brain are important, because neuroinflammation/neurodegeneration occurred during development can result in long-lasting brain damage.

Introduction

Microglial activation and neuroinflammation in the brain are the innate immune processes to defend the brain against harmful agents, such as bacterial and viral pathogens and dead cells and injured tissues (Rivest, 2009). In the healthy state, microglia, the primary innate immune cells of the brain, survey danger signals derived from those harmful agents (Lenz and Nelson, 2018). Depending on the conditions under which microglia are exposed to the danger signals/stressors, activated microglia acquire different phenotypes: some are more phagocytic and some secrete higher amounts of pro- or anti-inflammatory mediators than the others. These microglia with diverse characteristics are often classified using nomenclatures of macrophages: classical proinflammatory (M1) and alternate anti-inflammatory (M2) phenotypes (Orihuela et al., 2016;Cherry et al., 2014b;Hu et al., 2015;Jha et al., 2016), although this classification is only one of the designation systems for characterizing functions of activated microglia (Cherry et al., 2014b) and is not precisely grouped by distinct molecular patterns (Ransohoff, 2016).

Pro- and anti-inflammatory states of microglia affect the progression of brain injuries, neuroimmune disorders, neurodegenerative diseases, and other inflammation-related disorders. Therefore, factors which promote pro- or anti-inflammatory microglia may serve as therapeutic targets for treatment of such disorders (Song and Suk, 2017;Block et al., 2007;Colonna and Butovsky, 2017;Norden et al., 2015). While many factors are involved in microglial activation, recent literature highlights the importance of cellular energy status of microglia. It is indicated that mitochondrial function, glucose availability, and glycolytic rate in microglia influence pro-inflammatory gene expression at both transcriptional and post translational levels (Ghosh et al., 2018). Some molecules which regulate energy metabolism seem to influence microglial inflammatory status. For instance, silent information regulator 1 (SIRT1), a major regulator of energy metabolism, is a key regulator of NF-κB, a transcription factor inducing various pro-inflammatory factors (Chen et al., 2005;Cho et al., 2015;Ge et al., 2015), and peroxisome proliferator-activated receptors γ (PPARγ), that is also highly involved in energy metabolism, is suggested to be a key regulator of anti-inflammatory functions of microglia (Flores et al., 2016;Savage et al., 2015;Yamanaka et al., 2012;Zhao et al., 2015b). Nuclear factor erythroid-2-related factor-2 (Nrf2), which regulates expression of antioxidant proteins, and triggering receptor expressed on myeloid cells 2 (TREM2), which is necessary for phagocytic activity of microglia, are also involved in both energy metabolism and inflammation in microglia (Zhang et al., 2017;Vilhardt et al., 2017;Ulland et al., 2017). All these molecules show tight relationship with adenosine monophosphate-activated protein kinase (AMPK). AMPK, a highly conserved serine/threonine protein kinase, serves as a master sensor and regulator of energy balance and controls many cellular processes, such as autophagy, cell death, and inflammation to maintain cellular homeostasis (Peixoto et al., 2017). Once activated, it acts to restore energy homeostasis by activating catabolic pathways that generate ATP and by inhibiting anabolic pathways that consume ATP (Hardie et al., 2012;Hardie et al., 2016). While AMPK in the hypothalamus is known to be a central integrator of energy status and a master controller of feeding behavior through circulating orexigenic and anorexigenic hormones, AMPK also plays an important role in cellular energy homeostasis in the brain especially after insults, such as ischemia and hypoxia (Ramamurthy and Ronnett, 2006), and may affect neuroinflammation. In this review, roles of AMPK in microglial inflammation will be discussed based on studies using in vitro microglial culture and in vivo rodent models. Since microglial inflammation is tightly linked to neurodegeneration in brain injuries, neuroimmune disorders, neurodegenerative diseases, involvement of AMPK in neurodegeneration is also discussed to explore the possible anti-inflammatory/neuroprotective effects of AMPK activation in the adult and the developing brain.

1. Pro- and anti-inflammatory states of microglia

Microglia, which are developed from myeloid precursor cells in the embryonic yolk sac, are the primary innate immune cells in the brain, and respond to danger signals such as damage-associated molecular patterns (DAMPS), pathogen-associated molecular patterns (PAMPS), and dead cells/debris, although they have many other functions including synaptic organization and control of neuronal excitability. Once activated by danger signals and other stressors, microglia acquire diverse phenotypes. Traditionally, activated microglia are distinguished from homeostatic microglia (M0) and classified into pro-inflammatory (M1-like), anti-inflammatory (M2-like), or the mixture of both, primarily based on whether pro- or anti-inflammatory mediators and other cellular markers are differentially expressed, or how microglia are activated [e.g. via lipopolysaccharides (LPS), IL4, or IL13] (Lenz and Nelson, 2018;Kabba et al., 2018;Mosser et al., 2017;Hamzei et al., 2018). A typical pro-inflammatory state of microglia can be seen in animals and in cultured cells treated with LPS, a well-known PAMP derived from Gram-negative bacteria. These LPS-induced neuroinflammation models have been extensively studied (Lehnardt et al., 2003;Lehnardt et al., 2002;Orihuela et al., 2016). LPS trigger Toll-like receptor (TLR) 4 on microglia, activate a transcription factor NF-ĸB, and enhance expression of proinflammatory cytokines, chemokines, and cytotoxic factors, such as tumor necrosis factor α (TNF-α), interleukin-1 β (IL-1β), TLRs, cytokine receptors, nitric oxide, and reactive oxygen species (ROS) (Lehnardt et al., 2003;Lehnardt et al., 2002;Akira and Takeda, 2004;Orihuela et al., 2016). In contrast, IL-4 and IL-13, which are T-helper type 2 (Th2) cytokines, are often used to induce anti-inflammatory functions of microglia (Mori et al., 2016;Hamzei et al., 2018). These cytokines promote expression of Arg1 (Arginase-1), Ym-1 (chitinase-like protein), CD200R, IL-10, TGF-β (transforming growth factor), and Fizz-1 (found in inflammatory zone 1), which are considered specific markers for anti-inflammatory (M2-like) microglia (Song and Suk, 2017). M2-like microglial states may be subclassified into M2a, M2b, and M2c, as shown in macrophages (Subramaniam and Federoff, 2017). Furthermore, recent studies using single-cell RNA sequencing have led to findings of more diversified microglial populations during development or in response to injury (Mosser et al., 2017;Hammond et al., 2018). It has been reported that the profile of LPS-induced activated microglia (M1-like) is different from that of a microglial population uniquely associated with neurodegenerative diseases (Sousa et al., 2018) and called disease associated microglia (DAM) or microglial neurodegenerative (MGnD) phenotypes (Keren-Shaul et al., 2017;Krasemann et al., 2017;Shi and Holtzman, 2018). Also, when microglia are activated by multiple signals in the brain injury or neurodegenerative diseases, microglia may show more complex phenotypes: pro-inflammatory, anti-inflammatory, disease-associated, or mixture of these phenotypes, depending on the severity of triggers and cellular and environmental conditions of microglia including the effects from other cell types (e.g., neurons and astrocytes). Such inflammatory states of microglia may show dynamic changes during the process of inflammation. For instance, after ischemic stroke, microglia are activated by receiving danger signals from damaged neurons through CX3CL1 (a chemokine, also called fractalkine)/CX3CR1 (a receptor of CX3CL1) signaling, release pro-inflammatory factors, and aggravate ischemia-mediated brain damage (Tang et al., 2014). However, in response to sub lethal ischemia, neurons may express IL-4, which polarizes microglia to an anti-inflammatory phenotype. Such microglia initiate phagocytosis for clearance of debris and produce trophic factors that may promote brain repair (Zhao et al., 2015a;Wang et al., 2018a). Thus, pro- and anti-inflammatory states of microglia are involved in the progression of neuroimmune disorders, neurodegenerative diseases, and other inflammation-related disorders. Therefore, factors which regulate microglial activation states, especially, factors which shift microglial polarization from pro- to anti-inflammatory (repair-promoting) phenotypes, may serve as therapeutic agents for treatment of such disorders (Song et al., 2015).

2. Factors affecting neuroinflammation and energy metabolism

While many factors are involved in pro- and anti-inflammatory pathways in microglia, recent literature indicates the importance of cellular energy status of microglia. Microglia has a capacity to generate ATP by both glycolytic and oxidative pathways, and while quiescent microglia primarily rely on oxidative phosphorylation for ATP production, activated microglia seem to increase reliance on glycolysis (Cherry et al., 2014a;Ghosh et al., 2018;Orihuela et al., 2016;Gimeno-Bayon et al., 2014;Voloboueva et al., 2013). In macrophages, activated pro-inflammatory cells mainly generate energy through glycolysis, whereas anti-inflammatory cells predominantly generate energy through oxidative phosphorylation (O’Neill and Hardie, 2013). In microglia, it has been reported that mitochondrial function, glucose availability, and glycolytic rate influence pro-inflammatory gene expression at both transcriptional and post translational levels (Ghosh et al., 2018). When glucose utilization is suppressed by caloric restriction, ketogenic diet, and a glycolytic inhibitor [e.g., 2-deoxyglucose (2DG)], brain injury-induced neuroinflammation (e.g., microglial activation, induction of TNF-α), neurodegeneration, tissue loss, and functional impairment are reduced, while hypoxia and hyperglycemia, which promote glucose utilization, exacerbate inflammation and worsen brain injury (Shen et al., 2017;Robbins and Swanson, 2014;Loncarevic-Vasiljkovic et al., 2012). , It has also been reported that hyperglycolysis and induction of the key glycolytic enzyme hexokinase 2 (HK2) are essential for microglia-mediated neuroinflammation under hypoxia (Li et al., 2018b).

It is indicated that such relationship between inflammatory reactions and energy metabolism are related to the coordinated actions of NF-κB, which drives pro-inflammatory phenotype with glycolytic metabolism, and SIRT1, a major regulator of energy metabolism and tissue survival (Kauppinen et al., 2013). SIRT1, a member of the sirtuins (a family of NAD⁺-dependent lysine deacetylase), activates or inhibits several protein targets via deacetylation of histones and non-histone proteins, and SIRT1 deacetylates and inactivates the p65 component of the NF-κB pathway, thus limiting the expression of NF-κB-dependent genes and suppressing inflammatory responses (Yeung et al., 2004). Although these studies are mainly performed in peripheral tissues, studies in the brain indicate similar functions of SIRT1 on NF-κB. Resveratrol, a potent SIRT1 activator, ameliorates LPS-induced NF-κB activation in mouse hippocampus (Ge et al., 2015), and activation of SIRT1 markedly reduces NF-κB signaling and shows strong neuroprotection against amyloid β toxicity in primary cortical cultures (Chen et al., 2005). On the other hand, SIRT1 deficiency in microglia leads to enhanced expression of IL-1β and causes cognitive decline (Cho et al., 2015). These studies suggest that SIRT1 is a key regulator of NF-κB signaling. PPARs are members of a type II nuclear hormone receptor superfamily of ligand-activated nuclear transcription factors. Among them PPARγ regulates the expression of genes involved in inflammation, redox equilibrium, trophic factor production, insulin sensitivity, and the metabolism of lipids and glucose (Cai et al., 2018), thus representing another key molecule to connect inflammation and energy metabolism. PPARγ is abundant in immune-related cell types, particularly in adipocytes, macrophages, dendritic cells, and microglia (Yuan et al., 2015;Villapol, 2018), and is indicated as a key regulator of anti-inflammatory functions of microglia, because PPARγ activation increases anti-inflammatory-related gene expression and down-regulates pro-inflammatory mediators in activated microglia/macrophages and promotes phagocytosis of apoptotic cells, contributing to the resolution of inflammation (Flores et al., 2016;Savage et al., 2015;Villapol, 2018;Yamanaka et al., 2012;Zhao et al., 2015b), although there is a report showing that an antagonist of PPARγ promotes the pro-inflammatory to anti-inflammatory phenotypic shift in LPS-treated microglia (Ji et al., 2018). The Nrf2 transcription factor is a master regulator of antioxidant responses. Target genes of Nrf2 include NADPH quinone oxidoreductase-1 (NQO-1), heme oxygenase-1 (HO-1) and glutathione S-transferase (GST), and Nrf2 is also considered one of the anti-inflammatory molecules (Vilhardt et al., 2017). TREM2 is another molecule involved in energy metabolism and inflammation. TREM2 has been postulated to be a phagocytic receptor on myeloid cells including microglia, and the expression is associated with suppression of inflammatory gene expression by attenuating PI3K/NF-κB signaling. TREM2 overexpression inhibits LPS-induced microglial activation and pro-inflammatory reaction (Zhang et al., 2017). TREM2 sustains cell energetic and biosynthetic metabolism through mTOR signaling in microglia, and its deficiency disturbs the mTOR pathway, anabolic and energetic metabolism, leading to compensatory autophagy (Ulland et al., 2017).

Thus, SIRT1, PPARγ, Nrf2, and TREM2 are related to energy metabolism and enhance pro- to anti-inflammatory polarization of microglia, and these molecules are tightly linked to AMPK, which is also a key molecule for cellular energy balance. It has been shown that SIRT1 activates AMPK via deacetylation of the tumor suppressor liver kinase B1 (LKB1), an upstream kinase of AMPK, and AMPK promotes the synthesis of cellular NAD⁺, which is required for SIRT activity (Kauppinen et al., 2013). Many studies have supported the notion that the Nrf2 antioxidant pathway is downstream of AMPK (Yaku et al., 2013;Iwasaki et al., 2013). PPARγ phosphorylation by AMPK represses both the ligand-dependent and -independent trans activating function of the receptor (Burns and Vanden Heuvel, 2007). Also, TREM2 deficiency causes AMPK phosphorylation (activation), impaired mTOR signaling, and enhancement of compensatory autophagy (Ulland et al., 2017). As described below, AMPK regulate microglial inflammation.

3. AMPK as a regulator of inflammation

AMPK seems to play an important role in promoting microglial polarization from the pro- to anti-inflammatory phenotype. AMPK, a highly conserved serine/threonine protein kinase, serves as a master sensor and regulator of energy balance and controls various cellular processes, such as autophagy, cell death, and inflammation to maintain cellular homeostasis (Peixoto et al., 2017) and can be activated by many natural and synthesized agonists and activators (Fig.1). AMPK is a heterotrimer consisting of a catalytic subunit (α1 or α2) coupled to two regulatory subunits (β1 or β2; γ1, γ2, or γ3), which is ubiquitously expressed in mammalian cells (Hardie et al., 2016;Stapleton et al., 1996). The activity is controlled by: 1) phosphorylation at threonine 172 of the α subunit via up-stream protein kinases; LKB1 that is constitutively active (Hawley et al., 2003;Woods et al., 2003;Shaw et al., 2004), calmodulin kinase kinase β (CaMKKβ) (Hawley et al., 2005;Oakhill et al., 2012), and the transforming growth factor-β activated kinase (TAK1) (Momcilovic et al., 2006), 2) allosteric activation by AMP binding to γ subunit, and 3) activation of LKB1-induced Thr172 phosphorylation by AMP binding to γ subunit and inhibition of protein phosphatase-2Cα (PP2C-α)-induced Thr172 dephosphorylation by AMP or ADP binding to γ subunit (Hardie et al., 2016;O’Neill and Hardie, 2013;Xiao et al., 2007;Xiao et al., 2011;Gowans et al., 2013). Since ATP antagonizes these effects of AMP and ADP, AMPK acts as a sensor of cellular AMP/ATP and ADP/ATP ratios, which increase during cellular energy stress, although AMPK can also be activated by phosphorylation at threonine 172 by CaMKKβ activation through an increase in cellular Ca²⁺ levels, which is independent of cellular ADP/ATP ratios (Hawley et al., 2005). Activated AMPK acts to restore energy homeostasis by activating catabolic pathways that generate ATP, such as fatty acid oxidation, mitochondrial oxidative phosphorylation, glucose uptake through the phosphorylation of Akt and other enzymes necessary to the translocation of glucose transporter type 4 (GLUT4), and by inhibiting anabolic pathways that consume ATP, such as protein and lipid synthesis (Hardie et al., 2016;Hardie et al., 2012). AMPK also upregulates several antioxidant genes by activating Nrf2, which regulates the expression of antioxidant proteins that protect against oxidative damage triggered by injury and inflammation (Iwasaki et al., 2013;Yaku et al., 2013). AMPK also reduces protein synthesis and stimulates autophagic pathways via inhibition of the mechanistic target of rapamycin (mTOR) (Peixoto et al., 2017). In the brain, AMPK is predominantly expressed in neurons but also found in astrocytes, oligodendrocytes, and microglia (Culmsee et al., 2001;Lu et al., 2010;Ramamurthy and Ronnett, 2006;Turnley et al., 1999), and AMPK affects microglial pro- and anti-inflammatory phenotypes. Accumulating data from experiments using LPS as a trigger for pro-inflammatory reactions indicate that AMPK activation induces a microglial phenotypic shift from pro-inflammatory to anti-inflammatory in AMPK activation-dependent manner. In primary cultured microglia or BV2, LPS-induced pro-inflammatory reaction is associated with reduction in phospho (p)-AMPK (activated AMP) (Li et al., 2018b;Park et al., 2018;Xu et al., 2015;Zhou et al., 2014), and AMPK activators including metformin and 5-aminoimidazole-4-carboxamide 1-β-D-ribofuranoside (AICAR) suppress pro-inflammatory and enhance anti-inflammatory reactions (Giri et al., 2004;Lee et al., 2015;Lu et al., 2010;Li et al., 2018a;Xu et al., 2015;Park et al., 2018;Wang et al., 2018d;Zhou et al., 2014;Velagapudi et al., 2017;Chen et al., 2014). For example, AICAR (an analogue of AMP) attenuates LPS-induced activation of NF-κB and inhibits expression of proinflammatory cytokines (TNF-α, IL1β, IL-6) and iNOS, and these effects of AICAR are inhibited by the dominant negative form of AMPK and anti-sense AMPK treatment in primary rat astrocytes, microglia, and peritoneal macrophages (Giri et al., 2004). Betulinic acid (BA), a naturally occurring pentacyclic triterpenoid, attenuates LPS-induced pro-inflammation reaction in BV-2 microglial cells via CaMKKβ-dependent AMPK activation, and compound C (an AMPK inhibitor) and AMPK siRNA abolish the effects, although BA does not affect phagocytic activation induced by LPS (Li et al., 2018a). ENERGI-F704, an AMPK agonist, decreases the protein level and nuclear translocation of NF-κB, leading to reduction of pro-inflammatory mediators, such as IL-6, TNF-α, iNOs, and COX-3 in LPS-stimulated microglia (BV2) (Chen et al., 2014). Also, AMPK/Nrf2 signaling is involved in anti-inflammatory action of Petatewalide B against LPS in microglia (Park et al., 2018). Similar to the results shown in microglial culture, LPS decreases p-AMPK levels (Zhou et al., 2014), and AMPK activation by various activators suppresses pro-inflammatory reactions and enhances anti-inflammatory reactions in LPS-induced rodent neuroinflammation models (Lu et al., 2010;Wang et al., 2018d;Xu et al., 2015;Zhou et al., 2014;Li et al., 2018a). For instance, while LPS [intracerebroventricular (icv) injection] decreases p-AMPK levels and induces expression of M1 signature genes (iNOS, IL-1β, TNFα, and IL-6) in the lateral septal complex area, administration of hydrogen sulfide (H2S) donors enhances AMPK activation, attenuates M1 signature gene expression, and enhances expression of M2 signature genes (Arg-1, YM1/2, and IL-10) (Zhou et al., 2014). Treatment with BA significantly increases p-CaMKKβ and p-AMPK in the cerebral cortex and suppresses LPS [intraperitoneal (i.p.) injection]-induced mRNA expression of TNF-α and iNOS and promotes M2 gene CD206 and Arg-1 mRNA expression (Li et al., 2018a). Balasubramide derivative 3C ameliorates depressive behaviors of LPS (i.p.)-induced neuroinflammatory mice by promoting the anti-inflammatory activation of microglia through the CaMKKβ-dependent AMPK/peroxisome proliferator-activated receptor-γ coactivator (PGC-1α) signaling pathway (Wang et al., 2018d). Salvanoic acid C inhibits LPS-induced inflammatory response and NF-ĸB activation through the activation of AMPK/Nrf2 signaling in vivo and in vitro (Song et al., 2018). Thus, AMPK activation seems to attenuate proinflammatory reaction induced by LPS and enhance anti-inflammatory reaction both in in vivo and in vitro. Although metformin and other AMPK activators used in these experiments may not be specific to AMPK, experiments using AMPK siRNAs or other inhibitors and genetic modifications of AMPK indicate the importance of AMPK activation in the microglial shift from pro- to anti-inflammatory phenotypes (Giri et al., 2004;Li et al., 2018a;Park et al., 2018;Song et al., 2018;Wang et al., 2018d;Zhou et al., 2014). These experiments also show that LPS treatment decreases microglial p-AMPK levels, which are restored by AMPK activators.

Fig. 1.

AMPK-signaling and its functions regulated by agonists and activators. Arrows indicate activation or elevation of AMPK.

In addition to LPS-induced inflammatory models, the microglial shift from pro- to anti-inflammatory phenotypes by AMPK activation has been also reported in several animal models of inflammation-related diseases. In a MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) plus probenecid (MPTPp) mouse model of Parkinson’s, metformin enhances AMPK phosphorylation and autophagy and inhibits NF-ĸBp65 nuclear translocation and activation of NLRP3 (NOD-, LRR-and pyrin domain-containing 3) inflammasome in the middle brain. Metformin also reduces the transcription of pro-inflammatory cytokines (TNF-α and IL6) and restores the transcription of anti-inflammatory cytokines (Lu et al., 2016). In the mouse model of diabetic encephalopathy induced by streptozotocin (STZ), elevation of AMPK by metformin reduces expression of astrocytic and microglial markers and inflammation mediators (Oliveira et al., 2016). It has also been reported that resveratrol reduces TNF-α and IL-6 transcripts, the expression of NF-ĸB, p38, and p-ERK1/2, as well as astrocytic activation in STZ-induced diabetic rats (Jing et al., 2013), which show lower hippocampal p-AMPK levels compared to the control rats. In a rat model of global cerebral ischemia, activation of AMPK by metformin suppresses pro-inflammatory responses (such as production of TNF-α, IL-6, iNOS, and ROS) and confers anti-inflammatory effects (such as production of IL-10) in the AMPK dependent manner (Ashabi et al., 2015). Metformin treatment for 3 weeks before permanent middle cerebral artery occulsion suppresses inflammation reaction (Zhu et al., 2015). Metformin treatment for 2 weeks before subjecting to 9 min asphyxia cardiac arrest (CA) ameliorates CA-induced glial activation in the hippocampal CA1 region and cortex, which is accompanied by augmented AMPK phosphorylation and autophagy activation (Zhu et al., 2018). Not only metformin, but also other compounds which activate AMPK supress inflammation. The compound 3C (Balasubramide derivative 3C)-mediated activation of CaMKKβ, AMPK, and PGC-1α is involved in the anti-inflammatory effects of 3C in the brain of LPS-treated mice and ischemic rats (Wang et al., 2018d). A dual AMPK/Nrf2 activator (HP-1c) reduces brain inflammation after focal ischemic stroke (Wang et al., 2018c).

Thus, in both cultured microglia and the brain, AMPK activators inhibit inflammation caused by LPS and other stressors, and the inhibition is attenuated by blocking AMPK activity genetically or pharmacologically, indicating that AMPK is involved in the pro- to anti-inflammatory shift of microglia. While multiple mechanisms behind the action of AMPK have been suggested, the aforementioned studies indicate that inhibition of NF-ĸB and activation of Nrf2 may be important (Fig. 2). Activated AMPK, which acts to restore energy homeostasis by activating catabolic pathways that generate ATP and by inhibiting anabolic pathways that consume ATP (Hardie et al., 2016;Hardie et al., 2012), indirectly inhibits NF-κB activation through multiple downstream pathways including SIRT1, Folkhead box O (FOXO), and PGC1α (Peixoto et al., 2017;Jeon, 2016). AMPK activates SIRT1, which directly inhibits NF-κB, or AMPK phosphorylates p53 and FOXO, which inhibit NF-κB signaling (Kauppinen et al., 2013;Peixoto et al., 2017). PGC-1α also promotes microglial anti-inflammatory activation by inhibiting NF-κB activity (Yang et al., 2017). AMPK can also modulate the NF-κB pathway by phosphorylating eNOS at Ser177 (Peixoto et al., 2017). In addition, AMPK may induce anti-inflammatory microglia by Nrf2 activation (Wang et al., 2018b; Song et al., 2018; Park et al., 2018).

Fig.2.

Possible mechanisms behind the pro- to anti-inflammatory phenotypic shift by AMPK activation in microglia activated by LPS and other cellular stressors.

Such anti-inflammatory polarization of microglia induced by AMPK activation is often associated with neuroprotection in animal models of brain disorders, such as ischemic, Parkinson’s, and streptozotocin-induced diabetic models (Jing et al., 2013;Lu et al., 2016;Oliveira et al., 2016;Wang et al., 2018d;Wang et al., 2018c). However, in Parkinson’s (MPTP) model, while metformin reduces microglial morphological changes as well as TNF-α, IL-1β and iNOS mRNAs levels, it exacerbates dopaminergic neuron damage (Ismaiel et al., 2016). It seems that AMPK activation may or may not exert neuroprotective effects depending on degrees and length of AMPK activation as described below.

4. AMPK and neurodegeneration

Dysregulated inflammation of the CNS may induce neurodegeneration. Once microglia are excessively activated, they can damage neurons by secreting neurotoxic molecules, such as proinflammatory cytokines, metalloproteinases, nitric oxide (NO), and ROS (Block et al., 2007;Norden et al., 2015;Colonna and Butovsky, 2017). As described above, the anti-inflammatory effects of AMPK activation are consistent across different models and different AMPK activators (Curry et al., 2018). Whether this anti-inflammatory action of AMPK via regulation of microglia results in the survival of neurons and other cell types is an important question when we consider the AMPK pathways as a therapeutic target for brain disorders related to neuroinflammation. If the neurodegeneration is induced or enhanced by neuroinflammation as predicted in some neurodegenerative diseases and brain injuries (Block et al., 2007;Norden et al., 2015;Colonna and Butovsky, 2017), AMPK activation may exert neuroprotection by attenuating neuroinflammation. However, AMPK, a master regulator of cellular energy balance in various cell types, may provide protective or harmful effects directly on neurons, which show high metabolic demands. In the brain, AMPK is activated in response to a number of pathologically relevant stresses, such as hypoxia (Gusarova et al., 2011;Mungai et al., 2011) and excitotoxicity (Weisova et al., 2013;Concannon et al., 2010), when AMP/ATP or ADP/ATP ratios increase. Therefore, AMPK activation is recognized as the development of neurological disorders, although it is unclear whether this activation is beneficial or deleterious (McCullough et al., 2005). Such AMPK activation may be necessary to restore AMP/ATP balance, but excessive or prolonged AMPK activation may be harmful. In cultured hippocampal neurons, AICAR, an AMPK activator, attenuates neuronal death induced by glucose deprivation, chemical hypoxia, glutamate, and amyloid β peptide (Culmsee et al., 2001). Also, metformin reduces neuronal apoptosis in cultured cortical neurons exposed to ethanol (Ullah et al., 2012). However, prolonged AMPK activation by kainic acid seems to induce apoptosis in cultured hippocampal neurons (Ullah et al., 2014). Also, zinc-induced neuronal death, which is thought to be a key mechanism of ischemic neuronal death, is mediated by enhanced AMPK via up-regulation of Bim (Bcl-2 homology domain 2 (BH3)-only protein) and activation of caspase-3 in cultured neurons (Eom et al., 2016).

In vivo, a number of studies show that metformin provides neuroprotection against cerebral ischemia induced by middle cerebral artery occlusion (Arbelaez-Quintero and Palacios, 2017). For example, metformin treatment for 2 weeks before global cerebral ischemia inhibits inflammatory responses and attenuates cell death in the rat hippocampus in AMPK-dependent manner (Ashabi et al., 2014), and metformin treatment for 3 weeks before permanent middle cerebral artery occulsion suppresses inflammation, reduces infarct volume, and improves neurological deficits in rats (Zhu et al., 2015). Further, acute metformin preconditioning activates AMPK and autophagy in the rat brain and reduces infarct bolume, neurological deficits and cell apoptosis during the focal cerebral ischaemia (Jiang et al., 2014). Also, Metformin treatment for 2 weeks before subjecting to asphyxia cardiac arrest (CA) ameliorates CA-induced neuronal degeneration as well as glial activation in the rat hippocampal CA1 and cortex, which is accompanied by augmented AMPK phosphorylation and autophagy activation (Zhu et al., 2018). These metformin’s effects are attenuated by an AMPK inhibitor (compound C) or an autophagy inhibitor (3-methladenine) (Jiang et al., 2014). Not only metformin, but also other compounds which activate AMPK show neuroprotection. The compound 3C (Balasubramide derivative 3C)-mediated activation of CaMKKβ, AMPK, and PGC-1α is involved in the anti-neuroinflammatory and neuroprotective effects of 3C in the brain of LPS-treated mice and ischemic rats (Wang et al., 2018d). A dual AMPK/Nrf2 activator (HP-1c) reduces rat brain inflammation and provides neuroprotection after focal ischemic stroke by enhacing anti-inflammatory microglia population and by anti-oxidative effects (Wang et al., 2018c).

However, AMPK activation induced by focal stroke in adult mice enhances metabolic dysfunction, and AMPK inhibition or AMPKα−2 deletion is neuroprotective and induces smaller infarct volumes compared with wild-type littermates (Li et al., 2007). Similarly, pharmacological inhibition of AMPK by C75 or compound C provides neuroprotection in stroke (middle cerebral artery occlusion) in mice (McCullough et al., 2005). Since acute metformin treatment exacerbates stroke injury in mice, while chronic metformin given pre-stroke is neuroprotective, the timing, duration and amount of AMPK activation may be factors which determine the effects of AMPK on the ischemic brain (Li et al., 2010).

Neuroprotection by AMPK activation is also reported in inflammatioin-related diseases other than stroke/ischemia. In mouse models of Parkinson’s, ghrelin, a gut peptide hormone, mediates the neuroprotective effects of calorie restriction by elevating p-AMPK levels and reducing MPTP-induced loss of substantia nigra dopamine neurons (Bayliss et al., 2016). Also, metformin reduces pro-inflammatory cytokines and dopaminergic neurodegeneration via AMPK-mediated autophagy and mitochondrial ROS clearance in a MPTP+probenecid (MPTPp) mouse model (Lu et al., 2016). In these studies, MPTP treatment increases p-AMPK levels as shown in ischemia models. However, not all stress-inducers activate cellular AMPK. As described above, LPS treatment, which induces inflammation, decreases p-AMPK levels in cultured microglia (Li et al., 2018a;Xu et al., 2015;Park et al., 2018;Zhou et al., 2014) and also reduces p-AMPK in a mouse neuroinflammation model (Zhou et al., 2014), although it is not clear which cell types are responsible for the reduction in vivo. In relevant to this, in the mouse model of experimental autoimmune encephalomyelitis (EAE), AMPK levels are low in immune cells including microglia (Nath et al., 2009) and treatment of mice with AICAR attenuates EAE by reducing the inflammatory response and protecting neurons (Nath et al., 2005). In STZ-induced diabetic rats, which show lower hippocampal p-AMPK levels compared to the control rats, resveratrol treatment increases hippocampal p-AMPK, diminishes neurodegeneration and astrocytic activation, and reduces TNF-α and IL-6 transcripts as well as the expression of NF-κB, p38, and ERK1/2 phosphorylation (Jing et al., 2013). In the mouse model of diabetic encephalopathy produced by STZ, metformin reduces neuroinflammation and decreases the loss of neurons in the hippocampus (Oliveira et al., 2016). In this study, STZ does not increase p-AMPK levels. Traumatic adult brain injury decreases AMPK activity, and AICAR/metformin improves cognitive outcome (Hill et al., 2016). In Alzheimer disease (AD) mouse models, resveratrol improves proteostasis and enhances cognitive functions, and these effects are mainly mediated by increased activation of the SIRT-1/AMPK pathway (Porquet et al., 2014;Corpas et al., 2018). In these studies, activation levels of AMPK seem to be no difference between AD and wild-type mice. Also, in a AD mouse model, metformin atenuates spatial memory deficit and neuron loss, and decreases amyloid-β plaque and chronic inflammation through AMPK activation (Ou et al., 2018). In SOD(G93A) amyotrophic lateral sclerosis (ALS) mice, which do not show the elevation of AMPK, resveratrol exerts neuroprotective effects on motoneurons by activating the SIRT1/AMPK pathway (Mancuso et al., 2014).

Thus, AMPK activation not only induces anti-inflammatory reactions in microglia but also exerts neuroprotection, especially when initial AMPK activation levels by insults are not excessive. AMPK activation may provide neuroprotection through anti-inflammatory action, enhanced autophagy, anti-oxidant activation through the Nrf2 pathway, and restoration of energy levels. It is also suggested that activation of pro-survival Akt is mediated by AMPK-induced insulin-like growth factor1 receptor (IGF-1R) activation via phosphorylation of the insulin receptor substrate-1 (IRS-1), and inhibition of AMPK by compound-C results in decreased p-Akt , suggesting a central role for AMPK in Akt activation (Leclerc et al., 2010). However, when energy levels/metabolic states are deeply disturbed by stressors, such as severe excitotoxicity or oxygen/glucose deprivation-induced injury, and AMPK is highly activated, further AMPK activation may be neurotoxic. It is suggested that within a short period of energy deficiency, AMPK activation enhances astrocytic glycolysis and ketosis to compensate energy deficiency in ischemic neurons. However, prolonged glycolysis in astrocytes leads to progressive acidosis and inhibits the ability of neurons to use lactate as energy sources, leading to neuronal death (Li and McCullough, 2010). Also, activation of AMPK may enhance the level of oxidative stress (Ju et al., 2014) leading to neurodegeneration. It is also indicated that activation of AMPK reduces Akt activation through inhibition of mTOR and leads to dephosphorylation and nuclear translocation of FOXO3, and FOXO3 phosphorylation by AMPK in the nucleus induces the pro-apoptotic Bcl-2 family protein Bim elevation during excitotoxic injury (Davila et al., 2012). It is shown that rotenone activates AMPK and inactivates Akt, causing neuronal cell death in culture (Xu et al., 2014), while metformin ameliorates sepsis-induced brain injury by inhibiting apoptosis, oxidative stress, and neuroinflammation via Akt activation (Tang et al., 2017). Thus, AMPK activation may lead to Akt activation or inhibition depending on the conditions of cells under stress and may regulate neurodegeneration/neuroprotection.

These studies imply that cellular AMPK levels modified by both stressors (e.g., ischemia, excitotoxicity) and AMPK activators (e.g., AICAR, metformin) may determine subsequent cell fates. AMPK activators may exert more efficient anti-inflammatory and neuroprotective actions against stressors which lower cellular AMPK levels, such as LPS, while AMPK activators induce or enhance neurotoxicity against stressors which exceedingly elevate cellular AMPK levels, such as severe ischemia. Also, cellular levels of p-Akt may affect the AMPK activation-induced cell fate, because activated Akt in microglia may enhance inflammation while activated Akt is generally neuroprotective.

Neuroprotection by AMPK observed in some of the in vivo disease/injury models may block microglial proinflammatory activation, which can be triggered by receiving danger signals from damaged neurons (Roh and Sohn, 2018;Venegas and Heneka, 2017;Schindler et al., 2018). Therefore, the AMPK-induced microglial shift to the proinflammatory phenotype may be caused not only by the direct effects of AMPK on microglia described in the previous section but also by the neuroprotective effects of AMPK.

5. Microglia in the developing brain

Microglia is originated from myeloid precursors born in the yolk sac around E7 in mice. Then immature microglia invade the CNS from E8.5 and follow tangential and radial migration pathways to colonize all CNS regions. In the CNS, proliferation and differentiation of immature microglia continue during early postnatal period (Mosser et al., 2017). It is found that microglia express distinct sets of genes that can divide microglia into three distinct groups: early (E10.5–14), pre-microglia (E14-P9], and adult microglia (P28 and on) (Matcovitch-Natan et al., 2016). Morphology of some pre-microglia are amoeboid and similar to activated phagocytic microglia in the adult brain although gene expression profiles are different (Lenz and Nelson, 2018). In the developing brain, microglia play a critical role both in normal brain development and in response to infection and other early life stresses. During the development of healthy CNS, microglia support a variety of processes, such as neurogenesis, synaptogenesis, synaptic elimination, myelination, cell death, cell survival, and angiogenesis/vascularization via functions such as phagocytosis and release of diffusible factors (Pierre et al., 2017;Lenz and Nelson, 2018). For example, at P5, microglia start engulfing inappropriate pre-synaptic elements via interactions between microglia and unwanted synapses through classical complement cascade with complement proteins C1q and C3 as identification signals (Stephan et al., 2012;Schafer et al., 2012;Bialas and Stevens, 2013). Also, TREM2 is indicated to be essential for microglia-mediated synaptic refinement during the early stages of brain development (Filipello et al., 2018). A subset of microglia in the early postnatal subventricular zone seems to enhance neurogenesis and oligodendrogenesis through production of several cytokines (Shigemoto-Mogami et al., 2014), and the survival of layer V cortical neurons requires microglia, which are accumulated close to the subcerebral region and secrete IGF1 (Ueno et al., 2013). It has been also indicated that a CD11c-positive microglial subset, which is the major source of IGF-1, displays myelinogenic and neurogenic phenotypes (Wlodarczyk et al., 2017). Mice lacking the colony stimulating factor 1 receptor (CSF1R) for CSF1 or IL34, which is critical for microglial proliferation, differentiation and survival, show abnormal brain development including behavioral defects (Erblich et al., 2011;Rosin et al., 2018), while microglial loss in adulthood seems to have little impact on behavioral outcomes (Lenz and Nelson, 2018). Thus, microglia play critical roles in normal neurodevelopment. Therefore, if microglia are activated and disturbed by early life stress including infection, normal neurodevelopmental trajectories can be disrupted, and neurodevelopmental disorders, neurodegenerative diseases, and brain injuries may be triggered or worsened. Microglia-involved neuroinflammation is also important in the pathogenesis of injury to the immature brain (Pierre et al., 2017;Mottahedin et al., 2017). Especially, complications during preterm birth, such as neonatal infection, meningitis, neonatal hypoxia-ischemia, and neonatal stroke, in which inflammation is a common factor, impose a risk of injury to the developing brain (Mottahedin et al., 2017). While the peripheral immune response is indicated to be suppressed in neonates, it is unclear if immune response is less active in the developing CNS (Christensen et al., 2014). In rodent infection models, it has been shown that LPS and CpG oligodeoxynucleotides (CpG ODN, PAMP present in B cells) trigger to produce higher levels of cytokines in postnatal day 2 (P2) mouse neonates compared to weanlings (Christensen et al., 2014). The hippocampus of P4 rats, which has a high number of microglia with activated/amoeboid morphology, produces pro-inflammatory cytokines such as IL-1β, IL6, and TNF-α upon LPS exposure (Schwarz and Bilbo, 2011), and stimulation of TLRs by LPS or Poly(I:C) during the prenatal or neonatal period results in robust inflammatory responses in the fetal and newborn brain, including marked microglial proliferation and increased cytokine expression (Hagberg et al., 2015). Systemic injection of LPS in P5 mouse pups induces an acute and transient increase in reactive microglia with both M1 and M2 phenotypes and dysregulation of the ongoing process of neurogenesis in the developing hippocampal germinal niche (Smith et al., 2014). In neonatal rodent hypoxia-ischemia/stroke models, microglial activation and widespread neuronal apoptosis are observed, while necrosis is a dominant type of neuronal cell death in adult brain (Vexler and Yenari, 2009). In a mouse model of neonatal excitotoxic brain damage, injection of ibotenate induces excitotoxic lesions along with microglial activation, and the inhibition of microglial and/or blood-derived monocytes activation is accompanied by a significant reduction in severity of ibotenate-induced brain lesions (Dommergues et al., 2003). Prenatal alcohol exposure at E14/15 triggers higher expression of proinflammatory cytokines in embryonic brain than maternal brains, which may be due to less developed anti-oxidant systems in the developing brain (Akhtar et al., 2017). Also, alcohol injection into P4 mice triggers MCP-1 (monocyte chemoattractant protein-1)/CCR2 (C-C Motif Chemokine Receptor 2) signaling-mediated microglial activation and neuroinflammation (Zhang et al., 2018a). Cultured microglia isolated from neonatal brains show higher production of proinflammatory cytokines after being activated by ATP compared to other ages (Lai et al., 2013). Also, it has been shown that the acute phase of neonatal stroke, excitotoxic injury, or systemic infection, microglia are major cells for inflammation responses, and monocyte infiltration is low compared to adult brain injury (Denker et al., 2007;Smith et al., 2014;Dommergues et al., 2003).

Thus, embryonic and neonatal microglia in the brain seem to show prominent pro-inflammatory phenotypes after activation and may be prone to activation following an immune challenge (Pierre et al., 2017). Furthermore, the inflammatory activation of microglia during embryonic and neonatal periods seems to induce neurodevelopmental abnormalities: Prenatal exposure of rodents to influenza, Poly(I:C), or LPS increases cytokines such as IL-1β, TNF-α, IL6 and induces neurodevelopmental abnormalities in the offspring (Meyer, 2014). Acute exposure to LPS in prenatal mice induces inflammation and alters glial cytoarchitecture in the P40 brain (O’Loughlin et al., 2017). Alcohol intake during gestation and lactation increases levels of several cytokines/chemokines in the maternal sera, amniotic fluid, and brains of fetuses and offspring through the TLR4 response, leading to long-term behavioral abnormalities (Pascual et al., 2017). Neonatal hyperglycemia induces CXCL (C-X-C Motif Chemokine Ligand) 10/CXCR3 (C-X-C Motif Chemokine Receptor) signaling, microglial activation, and astrocytosis in the rat hippocampus and alters long-term synaptogenesis and behavior (Shi et al., 2017;Satrom et al., 2018).

Anti-inflammatory functions of microglia have been also reported in the developing brain. Pharmacological depletion of microglia prior to neonatal stroke exacerbates the release of inflammatory cytokines, suggesting that at least a subpopulation of microglia shows anti-inflammatory responses (Hagberg et al., 2015;Fernandez-Lopez et al., 2016;Faustino et al., 2011). During the acute phase of neonatal ischemia, genetic deletion of the scavenger receptor CD36 reduces phagocytosis of neuronal debris by microglia and enhances neuroinflammation (Li et al., 2015;Woo et al., 2012). Our studies indicate that alcohol injection into P7 mice induces acute microglial activation along with apoptotic neurodegeneration (Saito et al., 2010;Saito et al., 2012;Saito et al., 2015). These activated microglia are localized in the same area where apoptotic neurodegeneration is detected by Fluoro-Jade stain, and the Fluoro-Jade stain is frequently observed in Iba1-positive microglia (Saito et al., 2015). In addition, the rod- or amoeba-shaped microglia are strongly labeled with antibody against CD68, a marker for phagocytes (Saito et al., 2015), and labeled with antibody against cleaved tau (tau cleaved by activated caspase-3) present in degenerating neurons (Saito et al., 2010), suggesting that these microglia are phagocytes engulfing degenerating neurons. Ahlers and his colleagues (Ahlers et al., 2015) also show that P7 alcohol activates microglia which contain markers of late-stage apoptotic neurons, and apoptotic bodies and are deactivated within 1–2 days. The authors show further that alcohol-induced microglial activation and transient elevation of TNF-α and IL-1β are largely abolished in BAX (Bcl-2-associated X protein) null mice lacking apoptotic neurodegeneration. Thus, neonatal microglia have high capacity of phagocytosis and eliminate apoptotic neurons without prolonged inflammation, probably because one of the physiological roles of microglia in the neonatal brain is to eliminate neurons dying by programmed cell death. However, it has been also shown that minocycline, which inhibits microglial activation and alleviates neuroinflammation, protects neurons from ethanol-induced apoptosis in the P4 mouse brain (Wang et al., 2018b), suggesting that the initial phase of pro-inflammatory microglia may contribute to ethanol-induced neurodegeneration.

These studies indicate that although there are differences in the process of microglial activation/neuroinflammation between the developing and the adult brain, microglial polarization from the pro-to anti-inflammatory phenotypes seems to be also beneficial in the developing brain for alleviating acute neurodegeneration as well as the long-lasing adverse effects on brain development.

6. AMPK and neuroinflammation/neurodegeneration in the developing brain

AMPK is present in the developing brain. It has been shown that catalytic (α1 and α2) and noncatalytic (β2 and γ1) subunits of AMPK are abundant in the embryonic rat brain (Culmsee et al., 2001). Mouse brain mRNA levels for α2 and β2 subunits of AMPK are higher between E14 and P8 compared to adults, whereas expression of α1, β1, and γ1 subunits is consistent at all ages examined, and immunostaining shows a mainly neuronal distribution of all isoforms (Turnley et al., 1999). Our observation also indicates that AMPK is highly expressed in neurons during the early neonatal period in the developing mouse brain. Specifically, the content of p-AMPK (activated form) and p-acetyl-CoA carboxylase (p-ACC) is high at P4 and P7 and decreases to 20% of the amount of P4 by P19, indicating that AMPK in the neonatal brain is more activated and phosphorylates its substrate ACC more strongly than AMPK in the P19 brain (Saito et al., 2009). It has been established that AMPK regulates lipogenesis and β-oxidation through modifying ACC and SREBP-1 transcriptionally and post-transcriptionally in the peripheral organs (Hardie and Pan, 2002). The regulation of ACC by AMPK has been observed not only in the peripheral tissues but also in hypothalamic neurons, cortical neurons, and Neuro2a cells (Dasgupta and Milbrandt, 2007;Landree et al., 2004). Our previous studies (Saito et al., 2007) have shown that ethanol decreases the content of p-AMPK and p-ACC, and increases triglyceride (TG) synthesis, which is considered a marker for lipogenesis, in P7 mouse brains. Conversely, nutrient deprivation increases p-AMPK and decreases TG levels in these brains. However, neither ethanol exposure nor nutrient deprivation affects levels of p-AMPK and TG in P19 brains. These observations, combined with the neuronal localization of AMPK, ACC, and fatty acid synthase (FAS) and the high content of pAMPK and pACC in early neonatal brains, support the notion that AMPK regulates lipogenesis in neurons during the early neonatal period (Saito et al., 2009). Higher AMPK activity or higher phosphorylated levels of AMPK during the mouse neonatal period is also observed in other studies (Rousset et al., 2015;Williams et al., 2011).

Functions of AMPK during brain development have not been well clarified. It has been reported that knockout of AMPK β1 (AMPK β−/−) subunit in mice leads to severe loss of neurons and oligodendrocytes as well as abnormal astrocyte proliferation (Dasgupta and Milbrandt, 2009). However, it has also been shown that AMPK β−/− mice display no visible brain developmental deficits (Dzamko et al., 2010), and using transgenic mice that are ubiquitously inactivated for the AMPKα1 gene and conditionally inactivated for the AMPKα2 gene, it has been indicated that AMPK is not required for neurogenesis, neuronal differentiation, or neuronal survival in vivo under normal conditions, while overactivation of AMPK during metabolic stress impairs neuronal polarization in a mTOR-dependent manner (Williams et al., 2011). Similarly, in differentiating hippocampal neurons, activation of AMPK by energetic stress suppresses both mTOR and Akt signaling pathways and inhibits neuronal development during axon outgrowth, dendrite growth and arborization (Ramamurthy et al., 2014).

Althlough studies on the relatioship between AMPK activation and microglial activation/neurodegeneration in the developing brain are still scarce, some studies show that AMPK activation is associated with anti-inflammatory action of microglia in the developing brain. Chemerin (a chemokine, also classified as a adipokine) suppresses inflammation and improves neurological recovery via ChemR23 (the receptor for chemerin)/CaMKK2/AMPK/Nrf2 pathway after germinal matrix hemorrhage (GMH) in neonatal rats (Zhang et al., 2018b). Experiments using Lipo-Alpha-NETA (ChemR23 inhibitor) and Lipo-Dorsomorphin (AMPK inhibitor) indicate that alleviation of GMH-induced inflammation by chemerin may be primarily mediated through AMPK/Nrf2 signaling in microglia. In a model of neonatal hypoxic-ischemic brain injury, metformin treatment attenuates brain infarct volumes and brain edema 24 h after hypoxic-ischemic injury, and the neuroprotection by metformin is associated with suppression of the neuroinflammation assessed by decreases in expression levels of proinflammatory mediators, such as TNF-α, IL-6, IL-Iβ, and iNOS, and also decreases in levels of TLR4 and NF-B, although the direct involvement of AMPK has not been proved in these studies (Fang et al., 2017). Also, it has been shown that pioglitazone, a PPARγ agonist, effectively blocks ethanol-induced neuroinflammation manifested by microglial morphological changes and production of cytokines and chemokines in neonatal mice (Drew et al., 2015). Although the effects of AMPK are not examined in this study, previous studies indicating that AMPK activation is linked to increased expression of PPARγ (Li et al., 2017a;Chen et al., 2018;Zhong et al., 2018;Sun et al., 2017) suggest the involvement of AMPK.

Neuroprotective effects of AMPK in the developing brain are also indicated in several studies. Intranasal administration of adiponectin attenuates neuronal apoptosis induced by hypoxia-ischemia in P10 rats at least in part by activating adipoR1/APPL1 (an adaptor protein mediating adiponectin signaling)/LKB1/AMPK signaling pathway (Xu et al., 2018). Similarly, osmotin (a homolog of adiponectin) and adiponectin attenuate alcohol-induced apoptosis in P7 rats partially via AMPK activation (Naseer et al., 2014). Sestrin 2, a highly conserved stress-inducible protein, provides neuroprotection against rat neonatal hypoxic-ischemic encephalopathy by AMPK signaling activation (Shi et al., 2017). It has been also indicated that up-regulation of nicotinamide mononucleotide adenylyltransferase 1 (NMNAT1) induces neuroprotection against ischemic injury through AMPK activation (Liang et al., 2015). However, AMPK activation also induces apoptosis under some conditions as observed in the adult brain. It has been shown that AMPK activates FOXO3a and promotes neuronal apoptosis in the P7 rat brain during the early phase after hypoxia-ischemia. Activation of AMPK is accompanied by the decrease of p-mTOR, p-Akt and p-FOXO3a, which induces FOXO3a translocation into the nucleus and up-regulates the expression of Bim and cleaved caspase 3 (Li et al., 2017b). High dose of glutamate significantly increases brain glutamate levels and activates AMPK, leading to cell death in the hippocampus of P7 rats, SH-SY5Y, and BV2 cells probably through the enhanced level of oxidative stress (Shah et al., 2016), as observed in kainic acid-induced hippocampal neurons (Ullah et al., 2014).

Thus, although the effects of AMPK on neuroinflammation and neurodegeneration have not been extensively studied in the developing brain, available data indicate that AMPK affects brain development especially after the metabolic stress. As observed in the adult brain, AMPK activation in the neonatal brain appears to promote anti-inflammatory responses in microglia and may reduce or enhance neuroapoptosis, depending on the AMPK activation conditions.

Highlights.

• AMPK activation shifts microglia from a pro- to anti-inflammatory phenotype.

• NF-κB and Nrf2 are involved in anti-inflammatory functions of AMPK.

• AMPK activation, if not excessive or prolonged, exerts neuroprotection.

• High levels of activated AMPK are present in the neonatal mouse brain.

• Anti-inflammatory/neuroprotective effects of AMPK are seen in adults and neonates.

Summary.

Accumulating evidence indicates that regulation of metabolic homeostasis by the AMPK signaling pathway plays important roles in determining phenotypes of microglia activated by stressors such as pathogens, environmental toxins, and ischemia in the adult brain. Activation of AMPK has been shown to polarize microglia to anti-inflammatory phenotypes in various rodent models. It is suggested that Inhibition of the NF-ĸB pathway and activation of the Nrf2 pathway are the major mechanisms behind the AMPK-induced anti-inflammatory reactions in microglia. Also, AMPK activation often exerts neuroprotection, while excess or prolonged elevation of AMPK induces neuronal cell death. The degree of AMPK activation as well as the status of Akt activation seem to be important factors which determine the fates of neurons under stress. Similarly, AMPK activation appears to promote anti-inflammatory microglial polarization in the developing brain and may attenuate or enhance neuroapoptosis depending on the AMPK activation conditions. However, studies in the developing brain are still scarce and need to be explored further, because neuroinflammation /neurodegeneration in the developing brain may cause the long-term adverse effects, which may lead to neurodevelopmental or neurodegenerative diseases.