Inhibitors of Oxidative Phosphorylation Modulate Astrocyte Inflammatory Responses through AMPK-Dependent Ptgs2 mRNA Stabilization

Inflammatory activation of astroglia adds to the pathology of various neurological diseases. Astrocytes respond to microglia-derived cytokines such as interleukin-1α (IL-1α) with enhanced inflammatory signaling. This provokes pro-inflammatory gene expression of, among others, the eicosanoid-generating enzyme prostaglandin endoperoxide synthase 2 (Ptgs2). Whereas metabolic regulation of innate immune cell inflammatory responses is intensely studied, pathways related to how metabolism modulates inflammatory signaling in astrocytes are underexplored. Here, we examined how mitochondrial oxidative phosphorylation affects inflammatory responses towards IL-1α and tumor necrosis factor α in neonatal rat astrocytes. Blocking respiratory complex I and III or adenosine triphosphate (ATP) synthase did not affect activation of inflammatory signaling by IL-1α, but did elicit differential effects on inflammatory gene mRNA expression. Remarkably, mRNA and protein expression of Ptgs2 by IL-1α was consistently up-regulated when oxidative phosphorylation was inhibited. The increase of Ptgs2 resulted from mRNA stabilization. Mitochondrial inhibitors also increased IL-1α-triggered secretion of eicosanoids, such as prostaglandin E2, prostaglandin F2α, and 6-keto-prostaglandin F1α, as assessed by liquid chromatography/mass spectrometry. Mechanistically, attenuating oxidative phosphorylation elevated adenosine monophosphate (AMP) and activated AMP-activated protein kinase (AMPK). AMPK silencing prevented Ptgs2 up-regulation by mitochondrial inhibitors, while AMPK activators recapitulated Ptgs2 mRNA stability regulation. Our data indicate modulation of astrocyte inflammatory responses by oxidative metabolism, with relevance towards eicosanoid production.


Introduction
Inflammatory processes within the central nervous system (CNS) underlie the pathology of neurologic disorders such as multiple sclerosis, and contribute to other neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease, and amyotrophic lateral sclerosis [1]. Whereas cells of the mononuclear phagocyte system, in particular brain-resident microglia, are major perpetuators of neuroinflammation, the contribution of other CNS-resident cells is being increasingly recognized. Astrocytes are CNS-resident cells with important homeostatic functions, but which also have roles in the pathogenesis of neurologic disorders [2]. In response to pro-inflammatory cytokines released by activated microglia, in particular interleukin (IL)-1α and tumor necrosis factor-α (TNFα), astrocytes acquire an activated, neurotoxic phenotype (A1 polarization) [3,4] and upregulate pro-inflammatory genes, which may be influenced by environmental insults [5]. In addition to typical pro-inflammatory and neurotoxic genes, part of the gene expression program may be devoted to resolution of inflammation, with gene products serving context-dependent pro-inflammatory or pro-resolving functions [6]. A prominent example is the inducible prostaglandin endoperoxide synthase (Ptgs2), a key enzyme of eicosanoid biosynthesis. Its products, prostaglandins, show both pro-inflammatory as well as pro-resolving qualities [6]. Our recent studies revealed complex regulation of Ptgs2 expression at both the transcriptional and post-transcriptional levels in primary rat astrocytes [7][8][9].
Cellular metabolism controls inflammatory gene expression in cells of the innate and adaptive immune systems [10,11]. Although glycolytic metabolism has established roles in pro-inflammatory immune cell activation [10], mitochondrial activity significantly contributes to regulation of immune responses through a variety of mechanisms [12]. Astrocytes are considered to rely mostly on glycolytic metabolism, providing lactate to fuel neuronal oxidative metabolism [13]. However, astrocytes also display robust mitochondrial activity and are able to consume different metabolic substrates, including fatty acids and glutamate [14,15]. How mitochondria in astrocytes regulate inflammatory responses is poorly understood.
We evaluated how mitochondrial activity regulates inflammatory gene expression in primary rat cortical astrocytes.
Whereas mitochondrial activity is dispensable for the induction of several inflammatory genes, some genes, including Ptgs2, are modulated. Mechanistically, deficient mitochondrial adenosine triphosphate (ATP) generation triggers adenosine monophosphate (AMP)-activated protein kinase (AMPK)-dependent Ptgs2 mRNA stabilization, resulting in increased prostanoid generation.

RNA Isolation and Analysis
Total RNA from astrocytes was isolated using PeqGold RNAPure kit (PeqLab, Erlangen, Germany) and reverse transcribed using Maxima First Strand cDNA Synthesis kit (Thermo Scientific, Waltham, MA, USA). Quantitative real-time polymerase chain reaction (PCR) was carried out on a CFX96 system from Bio-Rad using iQ SYBR TM green (Bio-Rad, Hercules, CA, USA). Primer sequences are available upon request. Target gene expression was normalized to the expression of actin.

ATP Determination
ATP levels in cellular lysates were determined using CellTiter-Glo kit (Promega, Mannheim, Germany) according to manufacturer instructions.

AMP Determination by LC-MS
The concentrations of AMP, ADP, and ATP in the samples were analyzed by liquid chromatography-electrospray ionization-tandem mass spectrometry, as previously described [16]. Briefly, methanolic cell extracts (approximately 750 µl samples) were spiked with 20 µl internal standard solution ( 13 C 10 , 15 N 5 -AMP and 13 C 10 , 15 N 5 -ATP, 100 ng/ml each in methanol), centrifuged, and the clear supernatant was evaporated at 45 • C under a gentle stream of nitrogen. Adenosine phosphates were separated using an anion exchange high-performance liquid chromatography (HPLC) column (BioBasic AX, 150x2.1 mm, Thermo Fisher, Waltham, MA, USA) within 15 minutes. A mass spectrometer 5500 QTrap (Sciex, Darmstadt, Germany) was used as an analyzer, operating as triple quadrupole in positive multiple reaction monitoring mode. The analysis of ATP was performed as previously described [16]. Additionally, AMP and ADP were quantified using 13 C 10 , 15 N 5 -AMP as internal standard. The precursor-to-product ion transitions used as quantifiers were m/z 348.2 → 136.1 for AMP and m/z 427.8 → 136.1 for ADP. Data acquisition and quantification were performed using Analyst software version 1.6.3 and MultiQuant software version 3.0.2 (both Sciex, Darmstadt, Germany), respectively, employing the internal standard method (isotope dilution mass spectrometry). Calibration curves were calculated by linear regression with 1/x weighting.

Statistical Analysis
Data are presented as means ± SE of at least three independent experiments. Normality of data sets was tested using Shapiro-Wilk test. Data were analyzed by one-way ANOVA analysis of variance with Bonferroni post hoc means comparison using GraphPad Prism version 5.03, GraphPad, San Diego, CA, USA). Differences were considered statistically significant at p < 0.05.

Results
Whereas the ability of astrocytes to respond to pro-inflammatory stimuli with increased mRNA expression of various pro-inflammatory genes is well documented, the question of how astrocyte metabolism affects these responses has not been addressed so far. To analyze the impact of mitochondrial respiration on inflammatory responses of astrocytes, we exposed primary rat cortical astrocytes to pro-inflammatory stimuli typically released by activated microglia (TNFα and IL-1α) [4] for 3 h. This was done in the absence or presence of inhibitors of mitochondrial respiratory complexes I (rotenone) and III (antimycin), as well as mitochondrial ATP synthase (oligomycin), followed by analysis of mRNA expression of genes associated with pro-inflammatory astrocyte polarization. Preliminary experiments verified that the inhibitors, at concentrations used in this study, suppressed oxygen consumption in astrocytes. As seen in Figure 1A, IL-1α robustly induced the expression of typical pro-inflammatory mRNAs. We observed heterogeneous modulation of IL1α-induced mRNA expression by mitochondrial inhibitors. Whereas antimycin and oligomycin did not influence the induction of Cxcl10, Ccl5, and Il6 mRNAs, they decreased IL-1α-stimulated levels of Tnf mRNA, and increased mRNA expression of Ptgs2, as well as inducible nitric oxide synthase (Nos2). We noticed that all inhibitors affected Ptgs2 expression to a similar extent, while rotenone had a strong inhibitory effect on Tnf and Cxcl10 expression but failed to potentiate IL-1α-induced Nos2 mRNA expression. Divergent effects of mitochondrial inhibitors were observed using TNFα as a stimulus ( Figure 1B). Whereas Ptgs2, Tnf, and Nos2 expression followed the same pattern of dependency as for IL-1α stimulation, expression of Cxcl10 and Ccl5 was suppressed by inhibitors of oxidative phosphorylation (except for the lack of the effect of oligomycin on Ccl5). Il6 induction remained intact in the presence of rotenone and oligomycin, but was potentiated by antimycin. Together, these data suggest that the influence of mitochondrial inhibitors on inflammatory mRNA expression is stimulus-and gene-dependent. In further experiments, we focused on the effect of mitochondrial inhibitors on Ptgs2 expression, since it was up-regulated by all inhibitors of oxidative phosphorylation, regardless of stimulus. Next, we confirmed the potentiating effect of mitochondrial inhibitors on the expression of Ptgs2 at the protein level ( Figure 1C). Furthermore, several prostaglandin downstream products of the Ptgs2 enzymatic activity accumulated in cell supernatants of IL-1α-treated astrocytes after co-incubation with rotenone, antimycin, and oligomycin ( Figure 1D). PGH 2 , which is generated upon activation of Ptgs2, is metabolized by various secondary enzymes into eicosanoids with different biological activities [18]. Therefore, we estimated the eicosanoid spectrum using UPLC-MS/MS. Several metabolites of the cyclooxygenase pathway were detectable in the culture medium of control and IL1α-stimulated cells: 6-keto-PGF 1α (a stable derivative of PGI 2 ), PGF 2α , PGE 2 , PGD 2 , TXB 2 (a stable derivative of TxA 2 ), and 12-HHT (12-Hydroxyheptadecatrienoic acid), which may be produced from PGH 2 through thromboxane synthase-dependent and -independent pathways [19]. IL-1α did not alter levels of TXB 2 and PGD 2 , but markedly induced the release of 6-keto-PGF 1α , PGF 2α , PGE 2 , and 12-HHT ( Figure 1D). In accordance with their effects on Ptgs2 protein ( Figure 1C), antimycin and oligomycin potentiated IL-1α-stimulated release of 6-keto-PGF 1α , PGF 2α , and PGE 2 , but rotenone suppressed secretion of these eicosanoids, indicating that rotenone may inhibit cyclooxygenase activity. Next, we addressed mechanisms of how mitochondrial inhibitors potentiate Ptgs2 mRNA expression. Analyzing early signaling by IL-1α, we found no effect of mitochondrial inhibitors on activation of the nuclear factor κB (NFκB) signaling (phosphorylation of IκB kinase (IKK)α/β), or the activities of c-Jun N-terminal kinase (JNK) and p38 mitogen-activated protein kinase (MAPK) (Figure 2A), suggesting that early pro-inflammatory signaling cascades underlying cytokine-induced transcriptional activation do not require mitochondrial activity. Since levels of Ptgs2 mRNA are regulated both by transcriptional and RNA stability mechanisms, we then assessed stability of Ptgs2 mRNA by analyzing its half-life in IL-1α-treated cells. Therefore, we pre-treated astrocytes with IL-1α in the presence or absence of antimycin and oligomycin, followed by the addition of the transcription inhibitor actinomycin D. In actinomycin-treated cells, the levels of Ptgs2 mRNA rapidly decay, a reflection of the low stability of Ptgs2 mRNA. Interestingly, mitochondrial inhibitors delayed the decay of Ptgs2 transcripts, indicating that increased Ptgs2 mRNA stability may underlie the stimulatory effect of mitochondrial inhibitors on Ptgs2 gene expression ( Figure 2B). We then questioned how mitochondrial inhibition causes mRNA stabilization. Respiratory inhibition decreased ATP production, thus, increasing the AMP/ATP ratio and activating the major energy sensor AMP-activated protein kinase (AMPK). At the same time, mitochondrial inhibition may also increase levels of mitochondrial reactive oxygen species (ROS), thereby eliciting ROS-dependent signaling. However, the mitochondrial ROS blocker mito-TEMPO did not influence IL-1α-stimulated Ptgs2 mRNA expression in the presence or absence of antimycin or oligomycin ( Figure 3A). On the other hand, mitochondrial inhibitors reduced cellular ATP content ( Figure 3B) and increased AMP concentration in cell lysates, as assessed by LC-MS ( Figure 3C). Expectedly, these alterations activated AMPK, as assessed by increased phosphorylation of AMPK and its target acetyl-CoA carboxylase (ACC) ( Figure 3D). Next, we explored pharmacological AMPK activation towards Ptgs2 gene expression in IL-1α-treated astrocytes. As seen in Figure 3E, the pharmacological AMPK activators phenformin and A-769662 potentiated IL-1α-induced mRNA expression of Ptgs2. To confirm the role of AMPK in potentiation of IL-1α-elicited Ptgs2 expression by mitochondrial inhibitors, we silenced expression of the predominant α1-isoform of the catalytic AMPK subunit (Prkaa1). As shown in Figure 3F,G, Prkaa1 silencing effectively suppressed Prkaa1 mRNA and protein expression, and largely attenuated Ptgs2 mRNA and protein in astrocytes treated with IL-1α and oligomycin. Finally, we observed increased mRNA stability of Ptgs2 in the presence of the AMPK activator phenformin ( Figure 3H). Taken together, our data suggest that the potentiating effect of mitochondrial inhibitors on cytokine-stimulated Ptgs2 gene expression in astrocytes is AMPK-mediated.

Discussion
The novel finding of this study is a gene-specific susceptibility of the inflammatory response of primary rat astrocytes to inhibition of mitochondrial oxidative phosphorylation. Using several pharmacological inhibitors targeting different components of the mitochondrial respiratory chain, we show that cytosolic kinase signaling cascades induced by pro-inflammatory cytokines remain largely intact in the absence of functional oxidative phosphorylation. This may reflect the reliance of astrocytes on glycolytic metabolism [13]. Similarly, inhibitors of oxidative phosphorylation do not generally suppress inflammatory responses of professional phagocytes [20], showing nuanced regulation of selected cytokine responses dependent on e.g. ROS generation by individual respiratory complexes [21,22]. Interestingly, our data also show cell-type-specific differences of individual cytokine induction depending on intact respiratory chains. In contrast to observations of macrophages where increased Tnf expression in response to TLR agonists was insensitive to mitochondrial inhibitors [22,23], in astrocytes, all drugs targeting oxidative phosphorylation (especially complex I inhibitor rotenone) attenuated cytokine-induced Tnf mRNA. We also noticed stimulus-dependent effects of mitochondrial inhibitors on the expression of chemokines Cxcl10 and Ccl5, with only TNFα-induced expression being sensitive. The causes of these discrepant effects remain to be investigated.
Mechanistically, we observed an AMPK-dependent Ptgs2 mRNA stabilization to drive increased Ptgs2 expression and activity in cytokine-treated astrocytes upon inhibition of oxidative phosphorylation. Although some reports suggested ROS-dependent AMPK activation due to redox modification of AMPK [24,25], our data support a classical model of AMPK activation due to an elevated AMP/ATP ratio in astrocytes treated with mitochondrial inhibitors, consistent with other detailed studies showing that ROS-dependent activation of AMPK proceeds through alteration of cellular AMP/ATP [26,27]. Remarkably, despite being considered a glycolytic cell type, astrocytes display substantial ATP losses upon inhibition of oxidative phosphorylation. This may be explained by the reliance of the first enzyme of glycolytic cascade, hexokinase, on mitochondria-derived ATP, facilitated by the mitochondrial location of hexokinase in neurons and astroglia [28,29]. Therefore, astrocytes may not be able to rapidly compensate for the mitochondrial ATP production loss by rapid upregulation of glycolysis. We note that although inhibiting mitochondrial ROS did not alter Ptgs2 upregulation by inhibitors of oxidative phosphorylation, expression of other inflammatory genes, such as IL-1β, is known to undergo regulation by mitochondrial ROS, as well as by mitochondrial membrane potential in innate immune cells [21]. Furthermore, AMPK activation following mitochondrial dysfunction is known to cause mitochondrial fragmentation [30], which also may contribute to altering inflammatory gene expression, as we have recently shown [31]. Whether these mechanisms contribute to modulating inflammatory gene expression in astrocytes remains to be addressed in future research. It should also be explored whether Ptgs2 induction by other cytokines present in inflammatory milieu of activated microglia, such as IL-6, is similarly affected by inhibiting mitochondrial activity How does AMPK stabilize Ptgs2 mRNA? Several proteins are involved in the regulation of Ptgs2 mRNA stability [32]. Most prominent are the RNA-binding proteins HuR (stabilizing) [33] and tristetraprolin (TTP) (destabilizing) [34]. HuR is an attractive candidate AMPK target, since AMPK has been reported to increase cytoplasmic levels of HuR through phosphorylation of nuclear importin [35]. In contrast, Ptgs2 stability may be negatively affected by AMPK due to TTP up-regulation by mTORC1 signaling [36], considering that AMPK is a well-established inhibitor of mTORC1 [37,38]. Although our previous research showed the relevance of TTP and HuR regulation for inflammatory gene regulation in astrocytes [39], we could not find evidence for altered cytosolic levels of HuR or TTP after mitochondrial inhibition in our system (data not shown). At least 16 proteins may bind Ptgs2 3'untranslated region [32]. Furthermore, deregulated expression of miRNAs may contribute to altered stability of Ptgs2 transcript under our conditions. Considering this complexity, the link between AMPK and Ptgs2 stability surely warrants further research.
An important finding of our work is that mitochondrial activity differentially affects the spectra of synthesized eicosanoids. The functions of astrocytes are sensitive to modulation by various eicosanoids [17]. In this work, we observed that astrocytes responded to stimulation by IL-1α released 6-keto-PGF 1α , PGF 2α , PGE 2 , and 12-HHT, but not PGD 2 and TXB 2 . This is different from astrocyte responses to stimulation by LPS, where increased production of PGE 2 , TXB 2 , 6-keto-PGF 1α , and PGD 2 was documented [40]. Apparently, stimulus-dependent effects were operating on the arachidonic acid metabolism system. Traditionally, changes in the expression of Ptgs correlate well with changes in the level of secreted PGE 2 , since Ptgs is considered to be a limiting enzyme in the bienzyme system for eicosanoid synthesis from arachidonic acid. Our data on the spectrum of eicosanoids during stimulation with IL-1α and modulation by mitochondrial inhibition indicate a currently unobserved relationship between mitochondrial activity and the enzymes of downstream terminal eicosanoid synthesis. This opens the possibility of changing the ratio between eicosanoids, not only through the influence on the enzymes of their synthesis, but also through the regulation of mitochondrial activity. We also noticed that rotenone exerted differential effects, attenuating secretion of multiple prostanoids while increasing Ptgs2 expression. These observations indicate additional effects due either to complex I inhibition or to other off-target actions of rotenone, either suppressing Ptgs2 enzymatic activity or inhibiting the delivery of the Ptgs2 substrate, arachidonic acid. This may also be specific to cell culture conditions, as rotenone did not affect kidney PGE 2 levels when applied in a whole animal model [41].
In conclusion, our work shows that mitochondria contribute to inflammatory responses of astroglial cells. Apparently, AMPK activation through energy loss is only one of several mitochondrial-dependent mechanisms to regulate inflammatory gene expression. Whereas Ptgs2 expression in neuronal and glial cells is usually connected to increased pain, we previously connected increased Ptgs2 stability to an anti-inflammatory spectrum of activities of the PPARγ-activating drug rosiglitazone [9]. Whether documented activation of AMPK by rosiglitazone contributes to this effect remains an interesting, unexplored possibility [42].