Background
Microglia are resident macrophages of the central nervous system (CNS) including the neuronal retina [
1,
2]. As innate immune cells, microglia can initiate immune protection under pathophysiological conditions to remove metabolic wastes, cellular debris and pathogens [
2‐
4]. Under physiological conditions, the debris in the CNS microenvironment is largely derived from active neurons, including apoptotic bodies and synaptic debris [
5]. Microglia are actively involved in synaptic pruning and synapse maturation [
6], therefore play an important role in the regulation of synaptic plasticity [
7,
8]. During infection, microglia can phagocytize pathogens and remove dead cells to maintain homeostasis [
9]. Uncontrolled microglial activation has been implicated in neurodegenerative diseases such as Huntington disease, multiple sclerosis, Alzheimer’s disease [
10], and retinal degeneration [
2].
A dynamic cell body and flexibility to rapid change shape are essential for microglia to carry out their protective roles. The protrusions of “resting” microglia are continuously moving around, scanning the CNS [
11,
12]. After activation, microglia rapidly extend processes towards the damaged site and retract processes facing the other direction [
13]. The processes are employed to phagocytize cellular debris and damaged neurons [
14]. Once engaged with pathogens, they become fully activated, obtain an “ameboid” shape and can move around rapidly. In addition, microglia release various cytokines and chemokines, and can proliferate in situ in response to increased demand in immune protection [
15]. Morphogenesis, phagocytosis, and migration require dynamic reorganization of the actin cytoskeleton, for which vast amount of energy in the form of adenosine triphosphate (ATP) are necessary [
16]; whereas, proliferation and the production of inflammatory mediators need essential “building materials” such as amino acids, nucleotides, and fatty acids. How microglia meet their metabolic demands is not fully understood.
Glucose, amino acids and fatty acids are the three main energy substrates. Each cell type has its own unique metabolic profile and is fueled by specific energy substrates under physiological conditions, and changes in cellular activity can induce a switch of this metabolic profile. A previous study using RNA-sequencing technique has shown that microglia express transporters for all these substrates, indicating a flexible use of available energy resources [
17]. However, which metabolic pathway dominates the energy resource in microglia under pathophysiological conditions remains unknown.
Recent studies have shown that LPS-stimulated inflammatory microglia switch from mitochondrial oxidative phosphorylation (OXPHOX) in resting conditions to anaerobic glycolysis [
18], a phenomenon that is similar to active macrophages [
19]. The mammalian brain utilizes glucose as its main source of energy [
20], and the majority of CNS glucose is consumed by neurons through the glucose transporter 3 (GLUT3) [
20]. During CNS inflammation, the high level of glycolysis in active microglia not only supports the production of toxic inflammatory mediators, but also consumes vast amount of glucose that is desperately needed by neurons. Further understanding the metabolic pathways of active microglia will be important to better control neuroinflammation and improve the management of neurodegeneration.
In this study, we show that microglia express high levels of glucose transporter 1 (GLUT1), which critically controls glucose uptake under pathophysiological conditions. Blocking GLUT1 suppressed microglial activation and reduced neurodegeneration in a mouse model of light-induced retinal degeneration.
Materials and methods
Animals
Four-week old wild-type C57BL/6 J mice were used for primary microglial culture. Twelve-week old C57BL/6J and CX3CR1gfp/+ mice (C57BL/6 J background) were used in the light-induced retinal degeneration study. All mice were maintained in the Biological Services Unit of Queen’s University Belfast (QUB) with 12-h light-dark cycle. All procedures were conducted under the regulation of the UK Home Office Animals (Scientific Procedures) Act 1986, and were in compliance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmology and Vision Research. The procedures were approved by the Animal Welfare & Ethical Review Body of QUB.
Microglial culture
BV2 cells were cultured in DMEM (Gibco, BRL, Paisley, UK) supplemented with 10% fetal calf serum (FCS) and 100 μg/ml primocin (Invivogen, San Diego, California, USA). Primary murine microglia were isolated and cultured from C57BL/6 J mice using a previously reported protocol with slight modifications [
21]. Briefly, brain tissues were dissected from 4-week old mice and digested with 0.05% trypsin at 37 °C for 10 min. The single cell suspension was cultured in DMEM with 10% FCS and 20% L929 conditioned media [
22]. After 5 days, floating cells were removed, and attached cells (microglia) were cultured for a further 5 days. The cells were passaged once reached confluency, and cells from passages 2–5 were used in the study. Immunocytochemistry showed that > 95% of cells were CD11b
+F4/80
+. During primary microglia culture, one cell line was spontaneously immortalized and subsequently cloned. One of the clones, B6M7 was also used in this study.
Cell stimulation
BV2, B6M7 and primary microglia were treated with LPS (100 ng/ml, E. coli, serotype 055: B5) + IFNγ (100 ng/ml) or IL-4 (20 ng/ml) (all from Sigma-Aldrich, Dorset, UK) for 24 h. The cells were termed as M(LPS + IFNγ) and M(IL-4) respectively. The supernatants were collected for cytokine/chemokine measurements (see below). Cells were collected for gene and protein expression studies.
Quantitative real-time reverse transcription PCR
Total RNA was extracted from BV2, B6M7 or primary microglia using TRI Reagent (Sigma-Aldrich). cDNA synthesis was performed with a random primer using the SuperScript™ II Reverse Transcriptase kit (Invitrogen) in a reaction of 1 μg of total RNA per reaction. Murine mRNA expression levels were quantified by qRT-PCR using SYBR Green I Master and the LightCycler 480 system (Roche Diagnostics GmbH, Mannheim, Germany) with specific primers (Table
1). GLUT1–13 primers were purchased from Primer Design (Southampton, UK). Other primers were self-designed using the Primer-Blast (
https://www.ncbi.nlm.nih.gov/tools/primer-blast/) and purchased from Integrated DNA Technologies, Inc. (Leuven, Belgium).
Table 1
Primers used for qRT-PCR
β-actin | 58 | F CCTTCCTTCTTGGGTATG |
R TGTAAAACGCAGCTCAGTAA |
GLUT1 | 60 | F CCCCGTCCTGCTGCTATTG |
R GCACCGTGAAGATGATGAAGAC |
GLUT3 | 60 | F TGTCCAGGAACCGATCTATGC |
R CCACCAGGAACAGAGAAACTAC |
GLUT4 | 60 | F CCAGTATGTTGCGGATGCTAT |
R TTTTAGGAAGGTGAAGATGAAGAAG |
GLUT5 | 60 | F ACCTCAGCGCAGGCGTGAAA |
R AGCAGGCTATGAGGCAGGTGGA |
GLUT6 | 60 | F CGTCCAGTTTGTGCCAAGG |
R CAGCAGGGGTATCAGGGTAAG |
GLUT8 | 60 | F GCTCACCGCTGCCTTCTG |
R CGAAATGGGCTGTGACTTGTT |
GLUT9 | 60 | F ACTATGTGGACTCAATGCGATCT |
R TCAATGACCAAGCCAGAGAAGA |
GLUT10 | 60 | F CTCTGTGGCGTTCATCAAGTG |
R ACGAGGAAGGCGGAGACT |
GLUT12 | 60 | F TGGCAGCAAAACCTTCCTCT |
R AGGGACTGGTTAAGAAGACTATGG |
GLUT13 | 60 | F CTGGAATAAACTGGATTTTCAACG |
R TCAGGGAGGCAGCCATAGA |
Arginase 1 | 61 | F TTATCGGAGCGCCTTTCTCAA |
R TGGTCTCTCACGTCATACTCTGT |
CCL2 | 58 | F AGGTCCCTGTCATGCTTCTG |
R TCTGGACCCATTCCTTCTT |
IL-1β | 61 | F TCCTTGTGCAAGTGTCTGAAGC |
R ATGAGTGATACTGCCTGCCTGA |
iNOS | 60 | F GGCAAACCCAAGGTCTACGTT |
R TCGCTCAAGTTCAGCTTGGT |
TNF-α | 58 | F GCCTCTTCTCATTCCTGCTT |
R CTCCTCCACTTGGTGGTTTG |
IL-12P40 | 60 | F GACATCATCAAACCAGACCCGCC |
R GCCTTTGCATTGGACTTCGGT |
Cytokine and chemokine measurement
Cytokines and chemokines including MCP-1/CCL2, IL-6, TNF-α and CCL5/RANTES in supernatants were measured by Cytometric Bead Array (CBA) assay (BD Biosciences, Oxford, UK) according to the manufacturer’s instructions. Briefly, 25 μl sample and 25 μl capture beads were mixed and incubated for 1 h followed by the addition of 25 μl PE-conjugated detection reagent for a further hour. Samples were washed and were acquired using FACS Canto II (BD Biosciences). Data were analyzed using FlowJo software. The fold change in cytokine levels relative to control groups was calculated.
Produciton of Reactive Oxygen Species (ROS)
Intracellular ROS was measured using CellROX Green kit; mitochondrial ROS was measured using MitoSOX (MitoSOX Red), and nitrite oxide (NO) concentration in the supernatants was measured using the Griess Reagent Kit according to the manufacturers’ instructions. All reagents were from Thermo Fisher Scientific (Winsford, UK). Fluorescent signaling for CellROX and MitoSOX and absorbance for NO was immediately measured using FLUOstar Omega plate reader (BMG Labtech Ltd., Bucks, UK).
Glucose uptake
Cells were cultured in a 96-well plate until 80% confluence. Prior to the assay, cells were washed with PBS twice and incubated in serum-free DMEM overnight. Glucose uptake by different microglial cells was measured using the Glucose Uptake Assay Kit (ab136955, Abcam, Cambridge, UK) following manufacturer’s instructions. Glucose levels were was quantified using the FLUOstar Omega plate reader (BMG Labtech).
Cell bioenergy tests
BV2 and B6M7 cells were seeded in XFe 96-well microplates (6000 cells/well) (Agilent Technologies, Sana Clara, USA) with/without 5 μM STF31 (C23H25N3O3S, Millipore, Watford, UK) for 24 h, followed by LPS + IFNγ (both 100 ng/ml) or IL-4 (20 ng/ml) stimulation for a further 24 h. Cells were washed and incubated in base medium (Agilent Technologies) at 37 °C for 1 h. Extracellular Acidification Rate (ECAR) and Oxygen Consumption Rate (OCR) were measured in real-time with Glycolysis Stress Test Kit and Mito Stress Test Kit respectively using the Seahorse XFe96 Analyser (Agilent Technologies) following manufacturer’s instructions. Data were normalized by cell numbers that was measured by the YO-PRO®-1 Assay (Thermo Fisher Scientific).
GLUT1 inhibition by STF31 in vitro
Microglial cells were seeded into 96-well plates and treated with different concentrations of STF31 (Millipore) ranging from 0.01 μM to 50 μM for 24 h to 48 h. Cells were further processed for glucose uptake assay (see above) or cell viability using AlamarBlue® Assay (Thermo Fisher Scientific) following the manufacturer’s instructions.
Phagocytosis assay
B6M7 microglial cells in 96-well plates (10,000/well) were treated with different concentrations of STF31for 24 h. The pHrodo® Green S. aureus BioParticles® kit (Thermo Fisher Scientific) was used to determine phagocytosis of microglia following the manufacture’s instructions. The fluorescence intensity was quantified using a plate reader (FLUOstar Omega microplate reader) at 485 nm (Excitation)/530 nm(Emission).
TUNEL staining
Cells and tissues were fixed in 2% paraformaldehyde (PFA, Agar Scientific Ltd., Cambridge, UK) for 10 min or 2 h, respectively. Eyes were embedded in optimal cutting temperature compound (OCT, Thermo Fisher Scientific) and cryosectioned (Leica CM1950 cryostat, UK). 14 μm thick cryosections were treated with 0.1% Triton X-100 (Millipore) for 5 min at room temperature followed by 3 washes with PBS. Samples were incubated with TUNEL-MIX, followed by 5% TUNEL-Enzyme in TUNEL Label (Roche). DNase I (50 U/μl, Sigma-Aldrich) treated samples were used as positive control. Negative controls were incubated with TUNEL label only. The samples were mounted using Vectashield medium with DAPI (Vector Laboratories Ltd., Peterborough, UK).
Light-induced retinal degeneration and STF31 administration in mice
C57BL/6 J (12-week old) mice were treated with 10 mg/kg STF31 twice daily for 2 days, followed by once daily for another 3 days. The vehicle dimethyl sulfoxide (DMSO) treated mice were used as controls. Body weight and electroretinography were measured on day 6, and eyes collected for immunohistochemistry. Four mice were used in each group.
The CX3CR1gfp/+ mice were treated as above with the first injection received 1 day before light exposure. Mice were dark-adapted for 16 h and pupils were dilated with 1% phenylephrine and 2.5% tropicamide (Chauvin, Essex, UK) under dim light. Mice were anesthetized with ketamine (Vetoquinol UK Ltd., Buckingham, UK) and Rompun (Bayer HealthCare, Kiel, Germany) and exposed to a focal white light (50,000 lx) delivered by an otoscope (1218AA, Karl Storz, Tuttlingen, Germany) for 10 min. At the end of the STF31 treatment, clinical and immunohistochemical investigations were performed. Five mice were used in each group.
Spectral domain optical coherence tomography
Mice were anesthetized and pupils dilated as described above. Viscotears Liquid Gel (Novartis Pharmaceuticals Ltd., Surrey, UK) was used to moisture the cornea. OCT images (30° field of view) were collected using Spectral Domain Optical Coherence Tomography (SD-OCT, Heidelberg Engineering Ltd., Hertfordshire, UK). Neuroretinal thickness was measured in the area approximately 1000 μm away from the edge of the optical disc.
Electroretinography
Scotopic electroretinography (ERG) was conducted in 12-week old WT C57BL/6 J mice before and after STF31 treatment using Espion Visual Electrophysiology system (Diagnosys LLC, Cambridge, UK) as described previously [
23]. Briefly, mice were anesthetized and pupils dilated as above. The mouse was placed on the heat-pad (38 °C). ERG was recorded using mouse corneal ERG electrodes in response to single white light flash, delivered by a standard Ganzfeld Stimulator (Diagnosys LLC). ERG signals were bandpass filtered between 0.3–500 Hz (without notch filtering), amplified 500-fold and digitized at 2 kHz using the CMGS-1 electrophysiology system (Diagnosys LLC).
Retinal flat mount staining
Mouse eyes were fixed in 2% PFA at room temperature for 2 h. The retinas were dissected as previously described [
24]. Samples were incubated with 1% Triton X-100 in PBS for 4 h, and blocked by incubation with 1% BSA. Retinal flatmounts were incubated with rabbit anti-cone arrestin (Millipore, Watford, UK) at 4 °C overnight, followed by Alexa Fluor 594-conjugated donkey anti-rabbit IgG (Life Technologies Ltd., Paisley, UK) for 2 h. Images were obtained by confocal microscopy (Eclipse TE200-U, Nikon UK Ltd., Surry, UK). The Fiji Image-J Software was used for image analysis.
Immunoblotting
Microglia were lysed in RIPA buffer with protease inhibitor cocktails and phosphatase inhibitors (Sigma-Aldrich). The protein concentration was determined by BCA kit (Thermo Fisher Scientific). Fifty microgram proteins were loaded on a 10% SDS-PAGE gel and transferred to an Immobilon polyvinylidene difluoride membrane (Millipore). The membranes were incubated sequentially with 3–5 ml primary antibodies (Table
2) in 5% dried milk solution in TBS/T overnight at 4 °C. The membranes were washed and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (Agilent) for 1 h in the dark at room temperature. The signal was detected by autoradiography using enhanced chemiluminescence (Clarity™ Western ECL Blotting Substrates, BioRad). Densitometric analysis was performed using ImageJ software.
Table 2
Antibodies used for western blot
β-actin | Mouse Monoclonal (Santa Cruz, 1:10000) | Polyclonal Rabbit anti-mouse HPR (1:5000) |
GLUT1 | Rabbit Polyclonal (Thermo Fisher, UK; 1:1000) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
GLUT4 | Rabbit Polyclonal (Thermo Fisher, UK; 1:1000) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
GLUT5 | Rabbit Polyclonal (Thermo Fisher, UK; 1:1000) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
GLUT9 | Rabbit Polyclonal (Millipore Merck, UK; 1:5000) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
GLUT10 | Rabbit Polyclonal (Thermo Fisher, UK; 1:800) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
GLUT12 | Rabbit Polyclonal (Antibodies-online; 1:2000) | Polyclonal Goat anti-Rabbit HPR (1:5000) |
Statistical analyses
Data were processed and analyzed by GraphPad Prism Software and presented as mean ± SEM. Comparisons between more than two groups were conducted using one-way ANOVA followed Dunnett’s Multiple Comparison Tests. Probability values of < 0.05 were considered significant.
Discussion
Glucose is essential for microglial survival and function. A previous study has shown that glucose depletion led to 74% microglia death within 30 min and 98% after 48 h [
27]. Here, we show that glucose uptake in microglia is mediated predominately by GLUT1. Importantly, we show that glycolysis in inflammatory M(LPS + IFNγ) microglia also relays on GLUT1-mediated glucose uptake, and blocking GLUT1 re-programmed proinflammatory microglia from glycolysis to OXPHOX, suppressed microglial activation and reduced light-induced retinal degeneration. Our data also show that GLUT1-mediated glucose metabolism is critically involved in mitochondrial respiration in M(IL-4) microglia. Our results highlight the importance of GLUT1-mediated glycolysis in microglial function under pathophysiological conditions.
Naïve microglia are fueled by both glycolysis and mitochondrial oxidative phosphorylation. Previous studies have shown that microglia express GLUT3, GLUT4 [
28], and GLUT5 [
29,
30]. GLUT1 was reported to be expressed in oligodendrocytes and astrocytes but not microglia [
31]. In the current study, we detected 13 GLUTs in murine primary microglia, BV2, and a newly established murine microglial cell line B6M7 cells. Surprisingly, we found that GLUT1 was expressed at the highest level among other GLUTs, and the expression of GLUT1 was further enhanced following LPS + IFNγ stimulation. This is in line with the metabolic re-programming of the cells towards anaerobic glycolysis. The expression of other GLUTs was not affected by LPS + IFNγ treatment. Our results suggest that GLUT1 critically controls glucose uptake and the glycolytic pathway in microglial cells, particularly under inflammatory conditions.
Phagocytosis is important for microglia to maintain CNS homeostasis. We show that blocking GLUT1-mediated glucose uptake reduced glycolysis (Fig.
6) and suppressed phagocytosis (Fig.
8a). Our results suggest that GLUT1-mediated glycolysis is critically involved in microglial phagocytosis. This is in line with our recent observation in the phagocytic function of peritoneal macrophages [
32]. Blocking GLUT1 in naïve microglia did not affect mitochondrial basal respiration and ATP production, but significantly reduced SRC (Fig.
7f). SRC is important for cells to deal with stress conditions when additional energy is required. Our results suggest that GLUT1-dependent glycolysis may provide reserved energy to microglia under stress conditions for survival and function.
Microglial activation is a necessary and beneficial response to CNS insults. However, uncontrolled microglial activation contributes critically to the pathogenesis of various neurodegenerative diseases [
33] through the release of excessive amount of inflammatory mediators. Recent studies have shown that M1-type proinflammatory microglia are fueled by anaerobic glycolysis [
18]. Previous studies have shown that the glycolysis pathway provides essential building blocks such as amino acids and fatty acids for the assembling of inflammatory mediators in macrophages [
34‐
36] and effector T helper cells [
37‐
39]. Our data and that of others [
18] suggest that inflammatory microglia are also fueled by the glycolytic pathway.
A recent study using dual tracer positron emission tomography has shown that following cerebral ischemia, neurons and astrocytes died from oxygen deficiency but immune cells metabolized glucose non-oxidatively [
40], suggesting that they are likely supported by anaerobic glycolysis at the ischemic inflammatory site. The immune cells at the lesion sites may consist of resident microglia and infiltrating leukocytes. Shifting immune cell metabolism towards anaerobic glycolysis in an environment in which the supply of oxygen and nutrients is restricted requires the upregulation of the facilitators of glucose transport. Our results suggest that inflammatory microglia upregulate GLUT1 to facilitate glucose uptake.
Using STF31, we show that when GLUT1-mediated glucose uptake is blocked in inflammatory microglia, the metabolic pathway can be further re-programmed towards OXPHOX. Consequently, the expression of inflammatory genes and the production of inflammatory mediators are reduced. It is important to note that although STF31 dose-dependently reduced glucose uptake in microglia, it did not completely block glucose uptake (Fig.
6a). 5 μm STF31 reduced glucose uptake from 85 pmol to 25 pmol in M(LPS + IFNγ) microglia (Fig.
6b), and glycolysis was not diminished (Fig.
6c-e). The results suggest that apart from GLUT1, other GLUTs also facilitate glucose uptake in microglia. This may explain why microglia did not undergo cell death after STF31 treatment as seen in some tumor cells [
41].
Another important observation of this study is the requirement of glucose in mitochondrial OXPHOX in alternatively active M(IL-4) microglia. Previous studies have shown that the M(IL-4) microglia are supported by an intact tricarboxylic acid (TCA) cycle and enhanced mitochondrial OXPHOX [
18]. However, the fuel source of OXPHOX in M(IL-4) microglia remains elusive. We show that when glucose uptake is restricted by STF31, mitochondrial basal respiration, ATP production and SRC were all significantly reduced, whereas mitochondrial superoxidate is increased in M(IL-4) microglia. Previously, fatty acids were shown to support mitochondrial oxidative metabolism in M(IL-4) macrophages [
42]. Recent studies suggest that glucose can also fuel OXPHOX in M(IL-4) macrophages [
43‐
45]. Our results suggest that mitochondrial respiration in M(IL-4) microglia is at least partially fueled by glucose.
The newly established microglial cell line B6M7 cells were used in this study. The cells express high levels of TMEM119 (a microglial-specific marker) but low levels of F4/80 (a typical macrophage marker). ATP is known to induce chemotaxis in microglia but not macrophage through Gi/o-coupled P2Y receptors [
46]. The B6M7 cells, but not peritoneal macrophages had a strong chemotactic response to ATP stimulation. Our results suggest that B6M7 cells phenotypically and functionally resemble microglia but not macrophages. Lineage tracing study will help to confirm the true identity of the cells. The B6M7 cells and BV2 cells have some functional differences. For example, ATP-induced chemotaxis is significantly higher in B6M7 cells than that in BV2 cells. The LPS + IFNγ-induced inflammatory gene expression was also significantly higher in B6M7 cells than that in BV2 cells. The B6M7 cells could be an useful additional resource in neuroimmunology studies.
Conclusion
Glucose uptake in microglia is dominated by GLUT1. Under inflammatory conditions, microglia further upregulate GLUT1 to facilitate glucose uptake and promote anaerobic glycolysis. Blocking GLUT1 is an effective approach to re-program the metabolic pathway and control microglial activation and neurodegeneration. As glucose uptake in neurons is mediated predominately by GLUT3 [
20], blocking GLUT1 may not only suppress microglial activation, but also increase available nutrients to neurons, particularly in the inflamed ischemic CNS when glucose supply is restricted. Since GLUT1 is expressed by many other cells (e.g., astrocytes, Muller cells, photoreceptors, endothelial cells) in the retina and it critically controls glucose uptake to CNS parenchyma at the blood-brain-barrier [
20], cell-specific inhibition of GLUT1 (e.g., via CX3CR1 or CFSR promoters) would be an ideal approach to control microglial activation and neuroinflammation.