The metabolic response to inflammation in astrocytes is regulated by nuclear factor‐kappa B signaling

Inflammation and metabolism are intrinsically linked with inflammatory stimuli inducing metabolic changes in cells and, in turn, metabolic capacity determining cellular inflammatory responses. Although well characterized in peripheral immune cells there is comparatively less known about these “immunometabolic” responses in astrocytes. In this study, we tested the hypothesis that the astrocytic inflammatory response driven by nuclear factor‐kappa B (NF‐κB) signaling is dependent on glycolytic metabolism. Using mouse primary cortical astrocyte cultures, we assessed changes in cellular metabolism after exposure to lipopolysaccharide (LPS), with cytokine ELISAs and immunoblotting being used to measure inflammatory responses. Results indicate temporally distinct metabolic adaptations to pro‐inflammatory stimulation in astrocytes: 3 hr LPS treatment increased glycolysis but did not alter mitochondrial metabolism, while following 24 hr of LPS treatment we observed increased oxidative phosphorylation, and decreased glycolytic capacity and glucose uptake, partly due to reduced glucose transporter 1 expression. Inhibition of NF‐κB signaling with the IKK‐beta inhibitor TPCA‐1 prevented the LPS induced changes to glycolysis and oxidative phosphorylation. Furthermore, TPCA‐1 treatment altered both glycolysis and oxidative phosphorylation independently from inflammatory stimulation, indicating a role for NF‐κB signaling in regulation of basal metabolism in astrocytes. Inhibition of glycolysis with 2‐deoxyglucose significantly attenuated LPS‐induced cytokine release and NF‐κB phosphorylation, indicating that intact glycolysis is required for the full inflammatory response to LPS. Together our data indicate that astrocytes display immunometabolic responses to acute LPS stimulation which may represent a potential therapeutic target for neuroinflammatory disorders.

with cytokine ELISAs and immunoblotting being used to measure inflammatory responses. Results indicate temporally distinct metabolic adaptations to proinflammatory stimulation in astrocytes: 3 hr LPS treatment increased glycolysis but did not alter mitochondrial metabolism, while following 24 hr of LPS treatment we observed increased oxidative phosphorylation, and decreased glycolytic capacity and glucose uptake, partly due to reduced glucose transporter 1 expression. Inhibition of NF-κB signaling with the IKK-beta inhibitor TPCA-1 prevented the LPS induced changes to glycolysis and oxidative phosphorylation. Furthermore, TPCA-1 treatment altered both glycolysis and oxidative phosphorylation independently from inflammatory stimulation, indicating a role for NF-κB signaling in regulation of basal metabolism in astrocytes. Inhibition of glycolysis with 2-deoxyglucose significantly attenuated LPS-induced cytokine release and NF-κB phosphorylation, indicating that intact glycolysis is required for the full inflammatory response to LPS.
Together our data indicate that astrocytes display immunometabolic responses to acute LPS stimulation which may represent a potential therapeutic target for neuroinflammatory disorders.

K E Y W O R D S
astrocyte, glycolysis, inflammation, metabolism, nuclear factor-kappa B 1 | BACKGROUND In peripheral immune cells of the myeloid and lymphoid lineage, inflammatory stimuli differentially modulate metabolic pathways including glycolysis and oxidative phosphorylation, with the full immune response being dependent on the activity of these processes (O'Neill, Kishton, & Rathmell, 2016;Russell, Huang, & VanderVen, 2019). This bidirectional relationship between immune and metabolic function has become known as "immunometabolism." Immunometabolism is emerging as a therapeutic target for disease, leading to the repurposing of pharmaceuticals targeting cellular metabolism (e.g., Metformin) for conditions including cancers (Singer, Cheng, Kreutz, Ho, & Siska, 2018), Alzheimer's disease (Campbell et al., 2018;Koenig et al., 2017), and autoimmune diseases such as multiple sclerosis and lupus (Norata et al., 2015). Anti-inflammatory drugs are also being investigated to treat metabolic diseases including Type-2 diabetes and obesity (Goldfine et al., 2008;Goldfine et al., 2010;Lee, Wollam, & Olefsky, 2018). Despite the therapeutic potential of this approach, there is less understanding of whether the same immunometabolic changes are also seen in the central nervous system (CNS), most notably in astrocytes.
Astrocytes are important for the maintenance of the CNS microenvironment and are key regulators of neuronal activity and behavior, resulting in physiological changes in the whole organism (Liddelow et al., 2017). Astrocytes become inflamed in vivo in response to a broad range of inflammatory conditions, including metabolic stress or infection, and undergo a series of morphological and functional adaptations termed astrogliosis (Sofroniew & Vinters, 2010;Zamanian et al., 2012).
Preventing inflammation in astrocytes through inhibition of the nuclear factor-kappa B (NF-κB) signaling pathway can alter the pathophysiological response to disease. For example, inhibition of NF-κB activity in glialfibrillary acidic protein (GFAP) expressing astrocytes promotes functional recovery of spinal cord injury in mice, resulting in reduced glial scarring and increased locomotor activity (Brambilla et al., 2005). Additionally, in the context of metabolic disease, inhibition of astrocyte NF-κB signaling in the mouse, by deleting IκB kinase β from GFAP expressing cells, improves glucose tolerance and reduces obesity by promoting energy expenditure (Douglass, Dorfman, Fasnacht, Shaffer, & Thaler, 2017). Some existing evidence already links inflammation with changes in cellular metabolism in astrocytes. In aging mice, increased NF-κB activity in astrocytes correlates with elevated aerobic metabolism (Jiang & Cadenas, 2014). Furthermore, in vitro studies in mouse primary astrocytes demonstrate that treatment with inflammatory cytokines increases their glucose uptake, reduces intracellular glycogen stores, and increases flux through the tricarboxylic acid (TCA) cycle and pentose phosphate pathway (Bélanger, Allaman, & Magistretti, 2011;Gavillet, Allaman, & Magistretti, 2008). In a related study, Meares et al. showed that decreasing substrate availability to astrocytes in vitro, by limiting glucose and pyruvate, reduces interferon-gamma-induced inflammatory responses (Meares, Qin, Liu, Holdbrooks, & Benveniste, 2013). The antiinflammatory effect of reduced glucose or pyruvate availability was replicated by pharmacological activation of the cellular energy sensor AMP-activated protein kinase (AMPK) (Meares et al., 2013). Exposure to nitric oxide, which can be an inflammatory factor in high concentrations (Sharma, Al-Omran, & Parvathy, 2007), also enhances the activity of signaling pathways which regulate glycolysis in astrocytes: hypoxia inducible factor 1α and AMPK (Almeida, Moncada, & Bolaños, 2004;Brix, Mesters, Pellerin, & Jöhren, 2012). Together these data suggest that astrocytes rely on glycolysis to achieve full inflammatory activation, but to our knowledge, this important question has never been directly examined.
We tested the hypothesis that the astrocytic inflammatory response driven by NF-κB signaling is dependent on glycolytic metabolism. The impact of inflammation on energy metabolism in astrocytes was investigated, including an examination of whether NF-κB signaling plays a role in the regulation of astrocyte metabolism during inflammation. Further understanding of the role of astrocyte metabolism in the inflammatory response may open the opportunity for new therapies for CNS disease.

| Reagents
Unless otherwise stated, all reagents were purchased from Sigma-Aldrich (part of Merck, UK) or ThermoFisher Scientific (UK).

| Animals and ethics
All animal studies were conducted in accordance with the UK Animals in Scientific Procedures Act 1986 (ASPA) and study plans were approved by the institutional Animal Welfare and Ethical Review Body at the University of Exeter. Mouse pups (postnatal Days 1-5) from breeding pairs of C57BL6/J mice (Charles River, UK) were used to produce cultures of primary astrocytes from the cortex; offspring of both sexes were used. Mice were group housed on a 12:12 light-dark cycle at 22 ± 2 C, with unlimited access to standard laboratory rodent diet (EURodent diet [5LF2], LabDiet, UK) and water.

| Cell viability assay
Cells were seeded at a density of 3.5 × 10 5 cells/well in six well plates. After 24 hr of treatment (described above in Section 2.4) cells were dissociated with 0.05% trypsin-EDTA and transferred to flow cytometry tubes where trypsin was neutralized with fluorescence activated cell sorting (FACS) buffer (2% vol/vol FBS in 1X PBS). Cells were centrifuged at 100g for 5 min and the supernatant was aspirated. Cells were resuspended in FACS buffer and stained with propidium iodide (2 μg/ml), which is taken into cells with disrupted membranes, for 10 min at 4 C. Uptake of dye was determined using flow cytometry, with increased uptake representing cell death (BD Accuri C6; BD Systems; 488 nm; 633/22 nm). Gating used is shown in supplementary Figure S2a. Region of interest P1 in Plot 2 was used to select events classified as dead cells.

| Immunoblotting
Cells were plated at a density of 4.5 × 10 5 cells per 60 mm diameter dish. After treatment, cells were washed with 1X PBS to remove excess BSA. Cells were lysed and protein isolated with modified RIPA buffer (supplementary methods S1) and centrifugation at 1.3 × 10 4 g.
Protein content was estimated using the Bradford method according to the manufacturer's instructions (BioRad; UK). Proteins were separated on a 12% vol/vol acrylamide gel using SDS-PAGE and transferred to nitrocellulose membranes. Precision plus protein Dual color standards (BioRad) were run on all gels and used to estimate protein size. Membranes were blocked for 1 hr in Odyssey Blocking Buffer GAPDH expression was used as a loading control. Antibodies against GAPDH were incubated with membranes for 1 hr at room temperature (see supplementary methods S2 for further details). Expression of total NF-κB was estimated by stripping (1X Re-Blot Plus Strong Solution; Millipore), blocking and reprobing the same membranes used to assess phosphorylated NF-κB expression. Between antibody binding steps, blots were washed three times for 5 min with TBS-T. Bound antibody was detected with species specific fluorescent antibodies, incubated with membranes for 1 hr at room temperature (see supplementary methods S2 for further details). Protein was visualized using the Li-Cor Odyssey scanner. Fluorescence from detected protein was estimated with Image Studio software (V5.2; Licor). For NF-κB immunoblots, expression of phosphorylated NF-κB (normalized to GAPDH) was represented as a ratio of phosphorylated NF-κB (normalized to GAPDH):total NF-κB (normalized to GAPDH). Data are shown as fold change from the mean of control groups.

| Media cytokine concentrations
Cells were plated at a density of 4.5 × 10 5 cells per 60 mm diameter dish. Media was collected from cells after treatment and cytokine concentrations determined using DuoSet ELISAs (BioTechne; UK) against mouse TNF-α, IL-6, and IL-10 as per the manufacturer's instructions.
Cytokine concentrations were normalized to protein content, collected and estimated using the Bradford method (BioRad) as described above.

| Metabolic analyses
The metabolic rate of cells was determined by measuring the extracellular acidification rate (ECAR) as a measure of glycolysis, and oxygen consumption rate (OCR) as a measure of mitochondrial function using the Seahorse Metabolic Bioanalyzer XF e 96 (Agilent; UK). The day before treatment cells were plated in 96 well Agilent Seahorse Culture plates (Agilent) at 4 × 10 4 cells per well in supplemented DMEM (10% vol/vol FBS, 200 U/ml penicillin-streptomycin, 8 mM L-glutamine, and 7.5 mM glucose). After treatment, and 1 hr prior to experimentation, media was replaced with Seahorse XF DMEM medium (pH 7.4; Agilent) supplemented with either L-glutamine (2 mM; Glycolytic Stress Test) or L-glutamine (2 mM), glucose (2.5 mM), and sodium pyruvate (2.5 mM; Mito Stress Test). Cells were incubated for 1 hr at 37 C in a humidified non-CO 2 incubator. Glycolytic rate and oxidative metabolism were determined using Glycolytic Stress Tests and Mito Stress Tests, respectively, as per the manufacturer's instructions (Agilent).

| Glycolytic Stress Test
Cells were treated sequentially with glucose (10 mM), oligomycin (1 μM), and 2-DG (50 mM). All treatments were rehydrated in Seahorse XF DMEM medium (pH 7.4; Agilent) supplemented with 2 mM L-glutamine. Glycolytic rate describes the highest ECAR measurement after addition of glucose. Glycolytic capacity describes the maximum ECAR measurement after addition of oligomycin. Glycolytic reserve was calculated as the difference between glycolytic capacity and glycolytic rate.

| Mito Stress Test
Cells were treated sequentially with oligomycin (0.5 μM), carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone (FCCP; 1 μM), and rotenone with antimycin A (R/A; 0.5 μM). All treatments were rehydrated in Seahorse XF DMEM medium (pH 7.4; Agilent) supplemented with L-glutamine (2 mM), glucose (2.5 mM), and sodium pyruvate (2.5 mM). Basal respiration was calculated as OCR immediately prior to addition of oligomycin minus the lowest measurement after addition of R/A. ATP production was calculated as the difference between basal respiration and the lowest OCR measurement after addition of oligomycin. Proton leak was calculated as the difference between the lowest OCR measurement (after addition of oligomycin) and the lowest OCR measurement after addition of R/A. Maximal respiration was calculated as the maximum OCR measurement after FCCP addition, minus the lowest OCR measurement after R/A addition. Spare capacity was calculated as the difference between basal respiration and maximal respiration. Coupling efficiency was ATP production as a percentage of basal respiration.

| Glucose uptake, ATP, and glycogen content measurements
Glucose uptake was determined using Glucose Uptake-Glo kit (Promega) following the manufacturer's instructions. Cells were plated at 4 × 10 4 cells/well in 96 well plates. A total of 10 μM 2-DG was used, while omission of 2-DG used as a negative control. Luminescence was measured after 30 min incubation in detection reagent. Data are shown as fold change from the mean of contemporaneous control groups.
Intracellular ATP (iATP) concentrations were quantified using the ATPLite Two Step kit (PerkinElmer) following the manufacturer's instructions with modifications as previously described (Vlachaki Walker et al., 2017). Data are shown as fold change from the mean of contemporaneous control groups.
Glycogen content was measured using the Glycogen Assay Kit (Fluorometric; Cell Biolabs; UK). Cells were plated at 1× 10 6 cells/dish in 100 mm Petri dishes. After treatment, cells were washed twice with PBS to remove exogenous glucose, collected in PBS and snap frozen.
Cells were lysed through sonication at 4 C (LOW, 10 cycles, 10 on/10 off) and centrifuged at 1.3 × 10 4 g for 20 min at 4 C. Supernatants were used for glycogen quantification. The assay was carried out following the manufacturer's instructions in black 96 well plates, and fluorescence was measured after 45 min incubation in the reaction mix. Values below detection are reported as zero.
All plate-based assays were quantified using a PheraStar plate reader (BMG).

| Mitochondrial membrane potential
Mitochondrial membrane potential was determined using tetramethylrhodamine ethyl ester (TMRE) dye uptake. Cells were plated in six well plates at 3.5 × 10 5 cells/well. After treatment (as appropriate), cells were dissociated using 0.05% trypsin-EDTA.
Cells were transferred to flow cytometry tubes, and trypsin was neutralized with FACS buffer (2% vol/vol FBS in 1X PBS). Cells were centrifuged at 100g for 5 min and the supernatant was aspirated. F I G U R E 1 Mouse primary cortical astrocytes displayed an inflammatory response to lipopolysaccharide (LPS). (a) Representative antiphospho-nuclear factor-kappa B (NF-κB) p65 (Ser536) (top) and anti-total NF-κB p65 (bottom) immunoblot after 3 hr treatment with 0.1 μg/ml LPS with anti-GAPDH loading control. (b) Densitometric analysis of (a): anti-phospho-NF-κB p65 (Ser536) fluorescence as a ratio of anti-total NF-κB p65 normalized to anti-GAPDH fluorescence and represented as fold change in fluorescence from the mean of vehicle-treated controls (n = 8).

| Data presentation and statistics
Data were processed using Microsoft Excel 2013. GraphPad Prism 8 was used to present data and for statistical analyses. For comparisons between two groups, unpaired t tests were used. For multiple comparisons, oneway analysis of variance with post hoc Tukey's tests were used. A minimum of three independent replicates were used for each study. Data are presented as mean ± SEM. Statistical significance was taken as p < .05.

| RESULTS
3.1 | LPS stimulated NF-κB signaling and cytokine secretion in mouse primary astrocytes Compared to vehicle-treated control cells, NF-κB (p65) phosphorylation in astrocytes was increased 3 and 24 hr after LPS treatment (113.8%, p = .0031 and 93.9%, p = .0058, respectively; Figure 1a,b F I G U R E 2 Lipopolysaccharide (LPS) increased glycolytic rate in mouse primary astrocytes within 30 min posttreatment. (a) Oxidative phosphorylation rate, represented as oxygen consumption rate (OCR) over time, of cortical primary mouse astrocytes (CRTAS). Treatment point indicated by vertical dotted line. LPS treatment compared with vehicle-treated control (0.017 mM fatty acid free bovine serum albumin; n = 18, two independent plates). (b) Area under the curve (AUC) of OCR after treatment with LPS (unpaired t test, p > .05; n = 18, two independent plates). (c) Glycolysis represented as extracellular acidification rate (ECAR) over time of CRTAS after treatment with LPS (n = 18, two independent plates). (d) AUC of ECAR after treatment with LPS (unpaired t test, **** p < .0001; n = 18, two independent plates).

| LPS induced an acute increase in glycolytic rate in mouse primary astrocytes
Following measurement of baseline, the astrocytes were treated with LPS or vehicle and OCR (Figure 2a,b) and ECAR (Figure 2c,d) measured every 6 min for 1 hr using the Seahorse Bioanalyzer XF e 96.
No acute change in OCR was observed (Figure 2a,b), indicating that oxidative phosphorylation was not altered. However, there was a rapid 20% increase in ECAR indicating increased glycolytic rate (Figure 2c,d). As a complementary measure of mitochondrial activity, mitochondrial membrane potential was assessed 15 min after treatment with LPS using TMRE dye uptake (Figure 2e). There was no change in mitochondrial membrane potential in LPS-treated astrocytes compared to vehicle-treated controls (p = .58), while the positive control, the mitochondrial oxidative phosphorylation uncoupler FCCP, significantly reduced TMRE uptake by 72% (p < .0001; Figure 2e). This suggests that the integrity of the mitochondrial membrane was not immediately altered by LPS treatment, and rate of proton flow across the membrane was unaltered. Astrocyte iATP concentrations were also unchanged after 15 min of treatment with LPS in comparison to vehicletreated controls (p = .31; Figure 2f). There was a 25% increase in F I G U R E 3 Glycolytic rate and capacity were increased in mouse primary cortical astrocytes after 3 hr treatment with lipopolysaccharide (LPS). (a) Extracellular acidification rate (ECAR) during Glycolytic Stress Test of mouse primary cortical astrocytes 3 hr after treatment with LPS compared with vehicletreated control (0.017 mM fatty acid free bovine serum albumin; n = 30, two independent plates). Glu = 10 mM glucose; Oligo = 1 μM oligomycin; 2-DG = 50 mM 2-deoxyglucose. (b) Glycolytic rate: change in ECAR after treatment with glucose (unpaired t test, * p < .05; n = 30, two independent plates). (c) Glycolytic capacity: change in ECAR after treatment with oligomycin (unpaired t test, * p < .05; n = 30, two independent plates). (d) Glycolytic reserve: difference between glycolytic rate and glycolytic capacity (unpaired t test, p > .05; n = 30, two independent plates). (e) Glucose uptake after 3 hr treatment with LPS as fold change in 2DG6P luminescence (unpaired t test, p > .05; n = 6). Data are expressed as mean ± SEM [Color figure can be viewed at wileyonlinelibrary.com] glucose uptake rate in the cells after 15 min treatment with LPS in comparison to vehicle-treated controls; however, this did not reach statistical significance (p = .24; Figure 2g).

| Changes to glycolytic metabolism were maintained 3 hr after treatment with LPS in mouse primary astrocytes
To further interrogate the changes in cellular metabolism related to inflammation, mouse primary astrocytes were treated with LPS for 3 hr and subjected to a Glycolytic Stress Test. Glycolytic rate and capacity were both significantly increased by 26% (p = .0138) and 20% (p = .025), respectively (Figure 3a-c). This change in extracellular acidification was unlikely to be due to dissolved CO 2 from oxidative phosphorylation, as OCR was not increased during the same treatment protocol (supplementary Figure S3). However, while the glucose uptake assay indicated a 10% increase (Figure 3e) in cells treated with LPS for 3 hr compared to vehicle-treated controls, this was not statistically significant (p = .46).
The glycolytic reserve was unchanged between LPS-treated astrocytes and vehicle-treated controls (Figure 3d).
The Mito Stress Test was used to evaluate mitochondrial function after 3 hr of LPS treatment and no changes were observed compared to vehicle-treated controls (supplementary Figure S3a,d-h). Furthermore, 3 hr of LPS treatment did not impact TMRE uptake compared to vehicle-treated astrocytes (supplementary Figure S3b), in contrast F I G U R E 4 Glycolytic capacity in mouse primary astrocytes was reduced after 24 hr lipopolysaccharide (LPS) treatment. (a) Extracellular acidification rate (ECAR) during Glycolytic Stress Test of mouse primary cortical astrocytes 24 hr after treatment with LPS compared with vehicle-treated control (0.017 mM fatty acid free bovine serum albumin; n = 44, three independent plates). Glu = 10 mM glucose; Oligo = 1 μM oligomycin; 2-DG = 50 mM 2-deoxyglucose. (b) Glycolytic rate: change in ECAR after treatment with glucose (unpaired t test, p > .05; n = 44, three independent plates). (c) Glycolytic capacity: change in ECAR after treatment with oligomycin (unpaired t test, *** p < .001; n = 44, three independent plates). (d) Glycolytic reserve: difference between glycolytic rate and glycolytic capacity (unpaired t test, **** p < .0001; n = 44, three independent plates). (e) Intracellular glycogen content after 24 hr treatment with LPS (unpaired t test, * p < .05; n = 4). (f) Glucose uptake after 24 hr treatment with LPS as fold change in 2DG6P luminescence compared to the mean of vehicletreated controls (unpaired t test, ** p < .01; n = 6). Data are expressed as mean ± SEM [Color figure can be viewed at wileyonlinelibrary.com] to the FCCP-treated (positive control) cells which showed a 70% reduction in TMRE uptake (p < .0001). This suggests that astrocyte mitochondrial membrane integrity was unaltered, and coupling was not impacted 3 hr after LPS treatment. iATP concentrations were also unchanged (supplementary Figure S3). Together these data suggest that while there were adaptive changes in glycolysis, mitochondrial metabolism was unchanged by 3 hr of treatment with LPS.
F I G U R E 5 Twenty-four hours exposure to lipopolysaccharide (LPS) increased mitochondrial respiration in mouse primary astrocytes. (a) Oxygen consumption rate (OCR) during Mito Stress Test of mouse primary cortical astrocytes 24 hr after treatment with LPS compared with vehicle-treated control (0.017 mM fatty acid free bovine serum albumin; n = 28-30, 2 independent plates). Oligo = 0.5 μM oligomycin; FCCP = 1 μM; R/A = 0.5 μM rotenone/antimycin. (b) Mitochondrial membrane potential after 24 hr treatment with LPS as fold change from the means of vehicle-treated controls measured by tetramethylrhodamine ethyl ester (TMRE) dye uptake. A total of 10 μM FCCP was used as a positive depolarization control (one-way analysis of variance (ANOVA) with post hoc Tukey, **** p < .0001; n = 6). (c) Intracellular ATP (iATP) after 24 hr treatment with LPS as fold change from the mean of vehicle-treated controls (unpaired t test, p > .05; n = 4). (d) Basal respiration: difference in OCR prior to oligomycin injection and after rotenone/antimycin injection (unpaired t test, **** p < .0001; n = 28-30, two independent plates). (e) Maximal respiration: difference in OCR after injection of FCCP and after rotenone/antimycin injection (unpaired t test, *** p < .001; n = 28-30, two independent plates). (f) Spare capacity: difference in OCR between basal respiration (d) and maximal respiration (e; unpaired t test, **** p < .0001; n = 28-30, two independent plates). (g) Proton leak: difference in OCR after oligomycin injection and after rotenone/antimycin injection (unpaired t test, ** p < .01; n = 28-30, two independent plates). (h) ATP production: difference in OCR prior to oligomycin injection and after oligomycin injection (unpaired t test, **** p < .0001; n = 28-30, two independent plates). (i) Coupling efficiency: percentage of basal respiration (d) used for ATP production (h; unpaired t test, p > .05; n = 28-30, two independent plates). Data are presented as mean ± SEM [Color figure can be viewed at wileyonlinelibrary.com] 3.4 | The glycolytic capacity of mouse primary astrocytes was reduced after 24 hr treatment with LPS Mouse primary astrocytes were treated with LPS for 24 hr prior to evaluation using the Glycolytic Stress Test. In contrast to the acute (3 hr) treatment, after 24 hr of LPS treatment, there was no significant difference in glycolytic rate between LPS-and vehicle-treated cells (Figure 4a,b; p = .29). However, glycolytic capacity was reduced by 47% in the LPS-treated astrocytes in comparison to vehicle-treated controls ( Figure 4c; p = .0002), resulting in a 97% decrease in glycolytic reserve ( Figure 4d; p < .0001). This was reflected in a 82% depletion of intracellular glycogen (Figure 4e; p = .023). This reduction in glycogen stores and glycolytic capacity may have been due, at least in part, to the 37% reduction in the rate of glucose uptake seen in the 24 hr LPS-treated cells, compared to vehicle-treated controls (Figure 4f; p = .0091). These data show that after 24 hr of exposure to LPS, astrocytes undergo pronounced changes in their cellular metabolism, characterized by a reduction in both the glucose uptake and glycolytic capacity.

| Mitochondrial respiration in mouse primary astrocytes was altered after 24 hr of LPS treatment
Given the adaptive changes to glycolysis observed after 24 hr of LPS treatment, the impact on mitochondrial metabolism was also investigated. Treatment with LPS for 24 hr increased basal mitochondrial respiration in astrocytes by 53% in comparison to vehicle-treated controls (Figure 5a,d; p < .0001). Conversely, maximal mitochondrial respiration was reduced by 27% (Figure 5e; p = .0003), resulting in a 79% reduction in spare capacity (Figure 5f; p < .0001). Proton leak was increased by 54% (Figure 5g; p = .0019). Mitochondrial ATP production was increased by 52% (Figure 5h; p < .0001) and total iATP levels were unchanged after 24 hr treatment with LPS compared to vehicle-treated cells (Figure 5c; p = .48). The mitochondrial coupling efficiency was unchanged (Figure 5i) and there was no difference in TMRE dye uptake between cells treated with LPS for 24 hr and vehicle-treated controls (Figure 5b). This corroborates the observed absence of change in coupling efficiency. Together these data suggest after 24 hr of LPS treatment basal mitochondrial activity of mouse primary astrocytes is increased, despite reduced mitochondrial capacity. Combined with the results of the Glycolytic Stress Test, these data suggest that astrocytes alter their metabolism to maintain iATP through altered substrate preference.

| Glucose transporter 1 expression was reduced by 24 hr of LPS treatment in mouse primary astrocytes
To investigate potential causes of the reduced glycolytic capacity observed after 24 hr LPS treatment, glucose transporter 1 (GLUT1) glucose transporter expression was quantified using immunoblotting.
After 3 hr of LPS treatment GLUT1 expression was not altered com-

| NF-κB inhibition reduced baseline glycolytic metabolism and prevented LPS induced increases in glycolytic rate
In other cell types, NF-κB signaling is implicated in the regulation of cellular metabolism (Goetzman & Prochownik, 2018; F I G U R E 6 Glucose transporter glucose transporter 1 (GLUT1) expression in mouse primary astrocytes was reduced after 24 hr treatment with lipopolysaccharide (LPS). Astrocytes receiving the combined TPCA-1 and LPS (3 hr) treatment displayed a 27% reduction in glycolytic rate compared to those treated with LPS alone (Figure 7a,b; p < .0001) suggesting that NF-κB signaling plays an instrumental role in the glycolytic response to LPS. Importantly, treatment with TPCA-1 alone reduced glycolytic rate by 28% compared to vehicle-treated cells (Figure 7a,b; p < .0001). This suggests that NF-κB signaling plays an important role in regulating basal metabolism in astrocytes which may impact the extent of the LPS-induced response. Glycolytic capacity was also significantly reduced in TPCA-1 cells in the presence (29%; p < .0001) or absence of LPS (36%; p < .0001) compared to controls (Figure 7c,d). This resulted in a reduction of glycolytic reserve in TPCA-1 groups (32% with LPS and 45% without LPS treatment; both p < .0001).

| Inhibition of the metabolic response to LPS attenuated cytokine release from mouse primary astrocytes
To investigate the necessity of glycolytic metabolism for the astrocyte inflammatory response, cells were pretreated with the nonmetabolizable glucose analogue, 2-DG (10 mM) prior to stimulation with LPS. Confirming the inhibitory effect of 2-DG on glycolysis, mouse primary astrocytes treated with 2-DG showed a 36% reduction in ECAR (p < .0001) but no impact was observed on OCR (supplementary Figure S6). Then, 24 hr treatment with 2-DG caused a 2% decrease in cell viability which, while statistically significant

| DISCUSSION
In common with the extensive published studies from peripheral immune cells (Bantug, Galluzzi, Kroemer, & Hess, 2018;O'Neill et al., 2016;Russell et al., 2019), our data indicate that astrocytes also display immunometabolic responses. Increases in glucose uptake and glycolytic metabolism were evident in astrocytes within minutes of LPS treatment and by 3 hr post-LPS, the Glycolytic Stress Test revealed an increased rate of glycolysis and glycolytic capacity. Due to the speed of the response, these changes are likely driven by increased activity of glucose transporters and the rate limiting glycolytic enzymes. Inhibition of NF-κB signaling with TPCA-1 abolished the LPS-induced glycolytic response in astrocytes at 3 hr suggesting a key role for this pathway in the observed response. Translocation of GLUT transporters into the plasma membrane is regulated by NF-κB signaling in other cell types (Sommermann et al., 2011). Although no change in total GLUT1 expression was seen at 3 hr after LPS treatment our studies, the subcellular localization of GLUT1 was not assessed; therefore, increased localization of GLUT1 at the plasma membrane cannot be ruled out. NF-κB signaling mediates increases in expression of c-Myc, a transcriptional regulator of glycolytic genes (Goetzman & Prochownik, 2018;La Rosa, Pierce, & Sonenshein, 1994), in fibroblasts and of hexokinase II in sarcoma cells (Londhe et al., 2018). Our results suggest that this pathway may also contribute to the changes seen in astrocytes.
Chronic LPS exposure (24 hr) resulted in metabolic adaptations in astrocytes, probably to maintain ATP production in the face of prolonged insult: Glycolytic capacity and reserve were reduced, but were accompanied by increased mitochondrial metabolism. As the experimental medium used for the Mito Stress Test was supplemented with pyruvate, the increases in mitochondrial metabolism were unlikely to be directly caused by the change in glycolysis reported at this time point. The increase in mitochondrial metabolism was likely driven by increased TCA cycling to enable utilization of alternate fuel sources as it has previously been shown that astrocytes increase TCA cycle rate after 48 hr treatment with IL-1β or TNF-α (Gavillet et al., 2008). We did not measure consumption of alternative carbon sources in this study so we cannot reliably ascertain how the TCA cycle is being supplied in the absence of glucose or glycogen derived carbon in our model. Amino acid availability has been demonstrated to be important for immune responses in macrophages (Viola, Munari, Sánchez-Rodríguez, Scolaro, & Castegna, 2019). Indeed, data from other groups suggests that astrocytes increase metabolism of glutamate during hypoglycemia (Hertz & Hertz, 2003;Nissen, Pajęcka, Stridh, Skytt, & Waagepetersen, 2015). In our model, the astrocytes were supplemented with glutamine; this may enable them to use amino acids to maintain oxidative metabolism during inflammation or absence of glucose, and this could be an interesting avenue for further studies.
In our study, TPCA-1 reduced the LPS-induced increase in mitochondrial activity, suggesting a regulatory role for NF-κB in astrocyte mitochondrial metabolism. In other cell types, NF-κB promotes mitochondrial respiration through increased expression and stabilization of p53 (Aleyasin et al., 2004;Fujioka et al., 2004;Mauro et al., 2011), which may also be the case in astrocytes. p53 activity reduces transcription of GLUT1 (Johnson & Perkins, 2012;Mauro et al., 2011;Schwartzenberg-Bar-Yoseph, Armoni, & Karnieli, 2004), which in our study was significantly reduced by chronic LPS treatment in a NF-κBdependent manner. This reduced GLUT1 expression may explain the attenuated glycolytic capacity and glucose uptake we observed in astrocytes exposed to LPS for 24 hr. We also reported a depletion of intracellular glycogen stores after 24 hr of LPS treatment. This data are consistent with data from other groups in astrocytes treated with pro-inflammatory cytokines for 48 hr (Bélanger et al., 2011;Gavillet et al., 2008). It is possible that restricting glucose uptake, through downregulating GLUT1 expression, results in depletion of glycogen stores and may be a mechanism by which astrocytes attempt to reduce glycolytic activity to limit the inflammatory response. This conclusion is further supported by the increased concentrations of the anti-inflammatory cytokine IL-10 measured at 24 hr, demonstrating that astrocytes are attempting to mitigate the pro-inflammatory environment by this timepoint.
A reduction in autocrine/paracrine actions of LPS-induced astrocyte cytokine production on metabolism is likely partially responsible for the metabolic differences seen downstream of TPCA-1 inhibition.
However, our data indicating reduced mitochondrial metabolism in astrocytes treated with TPCA-1 for 24 hr also suggests an important regulatory role for NF-κB signaling in basal astrocyte metabolism, probably driven through the mechanism(s) discussed above. This is supported by the observation that treatment with TPCA-1 alone reduced glycolytic rate after 3 hr, suggesting NF-κB plays a role in maintaining basal glycolytic flux in astrocytes. This decrease in glycolytic rate was only evident after addition of glucose, suggesting that after inhibition of NF-κB signaling astrocytes likely have reduced ability to take up exogenous glucose rather than impaired utilization of internal glycogen stores. In addition to inhibiting NF-κB signaling, TPCA-1 has also been reported to inhibit STAT3 signaling (Nan et al., 2014), which can also regulate metabolic function during inflammation (Poli & Camporeale, 2015). As such, a role for STAT3 signaling in mediating the effects seen cannot be ruled out.
Blocking glycolytic function in astrocytes with 2-DG reduced cytokine release and NF-κB (p65) phosphorylation after 24 hr LPS treatment, suggesting reduced activity of the canonical NF-κB pathway. However, NF-κB phosphorylation was still evident 3 hr after LPS treatment despite glycolytic inhibition, perhaps suggesting a role for autocrine/paracrine actions of residual cytokine production in perpetuating the inflammatory response. The initial production of cytokines may be reduced due to 2-DG associated suppression of other inflammatory pathways. For example, the glycolytic enzyme pyruvate kinase isoenzyme M2 acts as a kinase for STAT3 (Gao, Wang, Yang, Liu, & Liu, 2012). Reduced glycolytic flux may lead to lower expression of this enzyme, reducing STAT3 signaling. This would lead to a reduced inflammatory response. Alternatively, the reduced glycolytic rate may prevent the production of carbon intermediates required for cytokine production, through limiting the repurposing of TCA cycle enzymes (Menk et al., 2018;Williams, O'Neill, & O'Neill, 2018). This would result in a negative feedback loop reducing the inflammatory response (Caslin et al., 2018;McGettrick & O'Neill, 2013). It has been reported that in macrophages 2-DG reduces mitochondrial function independently of glycolytic function (Wang et al., 2018); however, in our data there was no change in oxidative rate as a result of 2-DG treatment.
This suggests that attenuation of the LPS-induced inflammatory response in the presence of 2-DG was likely a result of glycolytic inhibition rather than modulation of mitochondrial metabolism, but further investigation would be required to confirm this.
In conclusion, we have shown that astrocytes have a rapid and flexible metabolic response to LPS-induced inflammatory insult. This metabolic response was abolished through inhibition of canonical NF-κB signaling using the pharmacological inhibitor TPCA-1, implicating this pathway as a major regulator of inflammatory metabolism in astrocytes. Furthermore, pharmacological inhibition of NF-κB signaling was sufficient to impact metabolism in astrocytes, independent of inflammation, suggesting an integral role for NF-κB in basal astrocyte metabolism. In support of our hypothesis, inhibition of glycolysis attenuated cytokine release from astrocytes, and resulted in inhibition of LPS-induced NF-kB phosphorylation, suggesting that glycolysis in astrocytes is essential for maintaining an inflammatory response.
Together these data indicate that in astrocytes metabolic adaptations are necessary for the full inflammatory response to LPS and, in turn, inhibiting NF-κB signaling is sufficient to impact basal metabolism.

ACKNOWLEDGMENTS
This work was supported by grants from the Medical Research Council (MR/N012763/1 K. L. J. E. and C. B., which supported NMH), Diabetes UK (RD Lawrence Fellowship to C. B.; 13/0004647) and the European Foundation for the Study of Diabetes (to C. B. and KLJE, which supports PGWP), and internal funding from the University of Exeter Medical School (which supports J. L. R.).