EPO has multiple positive effects on astrocytes in an experimental model of ischemia

Erythropoietin (EPO) has neuroprotective effects in central nervous system injury models. In clinical trials EPO has shown beneficial effects in traumatic brain injury (TBI) as well as in ischemic stroke. We have previously shown that EPO has short-term effects on astrocyte glutamatergic signaling in vitro and that administration of EPO after experimental TBI decreases early cytotoxic brain edema and preserves structural and functional properties of the blood-brain barrier. These effects have been attributed to preserved or restored astrocyte function. Here we explored the effects of EPO on astrocytes undergoing oxygen-glucose-deprivation, an in vitro model of ischemia. Measurements of glutamate uptake, intracellular pH, intrinsic NADH fluorescence, Na,K- ATPase activity, and lactate release were performed. We found that EPO within minutes caused a Na,K-ATPase-dependent increase in glutamate uptake, restored intracellular acidification caused by glutamate and increased lactate release. The effects on intracellular pH were dependent on the sodium/hydrogen exchanger NHE. In neuron-astrocyte co-cultures, EPO increased NADH production both in astrocytes and neurons, however the increase was greater in astrocytes. We suggest that EPO preserves astrocyte function under ischemic conditions and thus may contribute to neuroprotection in ischemic stroke and brain ischemia secondary to TBI.


Introduction
Astrocytes are the most abundant cell within the central nervous system (CNS). More than 10 subtypes have been characterized, possessing heterogeneous functions depending on localization and situation. The functions comprise, but are not limited to, synapse development and functional maintenance, neurotrophic support, homeostasis of ions and neurotransmitter function and repair of the bloodbrain barrier (BBB) as compiled in several reviews (Castejón, 2015;Ferrarelli, 2017;Magaki et al., 2018;Song et al., 2020;Cohen-Salmon et al., 2021;Hart and Karimi-Abdolrezaee, 2021;Hasel and Liddelow, 2021). The astrocytic end feet lining the BBB participates in the homeostasis of fluid and solutes in the CNS and therefore have a pivotal role in the development of brain edema.
During conditions with deprived supply of energy and oxygen in the CNS, as seen both in ischemia and secondary injury in traumatic brain injury (TBI), astrocyte glycolysis may benefit neuronal function. Via anaerobic metabolism and production of lactate, which function as a valuable energy substrate to neurons, neuronal activity can be sustained and thus contribute to neuroprotection (Pellerin and Magistretti, 1994;Ros et al., 2001;Cater et al., 2003;Machler et al., 2016;Muraleedharan et al., 2020;Roumes et al., 2021). When ischemia overrides astrocyte tolerance, astrocyte functions can be impeded causing BBB disintegration, edema formation and loss of cell integrity due to energy dependent pump failure and decreased glutamate uptake. Eventually, cell death of both astrocytes and neurons will occur (Choi and Rothman, 1990;Arundine and Tymianski, 2004;Nicholls et al., 2007;Rossi et al., 2007;Beard et al., 2021). Avoidance of these secondary injury mechanisms in clinical conditions could have a positive impact on patient outcome. Treatment options of secondary brain injuries as seen after both ischemia, TBI and other brain injuries are still limited, despite several preclinical and clinical studies on potential neuroprotective agents. EPO has shown efficient neuroprotective effects in a plethora of preclinical cellular and animal ischemia and trauma models (Pei et al., 2014;Ratilal et al., 2014;Jeong et al., 2017;Wang et al., 2017;Yoo et al., 2017;Fernando et al., 2018;Zhang et al., 2018;Jacobs et al., 2021).
Although astrocytes express EPO receptors and have multiple functions in acute ischemic brain injury, data regarding effects of EPO on astrocytes under ischemic conditions are still sparse.
In clinical trials, EPO has shown various beneficial effects in patients. EPO was shown to reduce 6-month mortality in TBI and to improve long term neurological outcome, but not mortality, in ischemic stroke patients (Tsai et al., 2015;Benoit et al., 2019;Katiyar et al., 2020;Liu et al., 2020a,b). Though there has also been concerns regarding negative effects (Lee et al., 2019;Vittori et al., 2021;Ma et al., 2022).
We hypothesized that EPO may have beneficial effects on astrocyte metabolism and decrease glutamate toxicity in ischemia. To explore this, we studied the effects of EPO on glutamate homeostasis in astrocytes undergoing oxygen-glucose deprivation (OGD), an in vitro model of ischemia. Studies were performed on rat primary astrocytes and astrocyte-neuron co-cultures. We found that EPO had potential positive effects on astrocyte glutamate uptake and metabolism in ischemic conditions.

EPO increases astrocyte glutamate uptake in OGD
To explore the effect of EPO on glutamate homeostasis, we measured the effect of EPO on astrocyte glutamate uptake. D-Aspartate has shown to be a mixed-type noncompetitive inhibitor of L-glutamate uptake in astrocytes (Drejer et al., 1982(Drejer et al., , 1983, why D-aspartate was used for stimulation in these experiments. Treatment with 5 nM EPO for 5 min increased [ 3 H] D-aspartate uptake by 22 % in cultured cortical astrocytes exposed to OGD (22 % ± 4.2 %, p < 0.05) (Fig. 1A).
We then explored whether the effect of EPO on glutamate uptake was dependent on Na,K-ATPase activity. Due to robust metabolism of glutamate, [ 3 H] D-aspartate uptake was used as an indicator of glutamate transporter function. When Na,K-ATPase was inhibited by 2 mM ouabain, EPO no longer caused any increase in [ 3 H] D-aspartate uptake, regardless of whether excessive aspartate was present or not (Fig. 1C). The findings support coupling between astrocyte Na,K-ATPase function and glutamate uptake, and that the positive effect of EPO on glutamate uptake is dependent on Na,K-ATPase activity.

The effect of EPO on pH is dependent on NHE function
To further investigate the effect of EPO on intracellular pH (pH i ) in astrocytes exposed to OGD, we explored the role of the sodium/ hydrogen exchanger (NHE). With concomitant 20 µM 5-(N,N-Dimethyl) amiloride hydrochloride (DMA), a selective blocker of NHE, the restoration of pH i by EPO was attenuated (0.27 ± 0.07 versus 0.05 ± 0.008, p < 0.001) (Fig. 2C). The change in pH i during DMA exposure was not significantly different from baseline (Fig. 2B).
EPO by itself cause intracellular alkalinization in primary astrocytes undergoing OGD, with a pH i increase of 0.18 units (7.09 ± 0.017-7.27 ± 0.018, p < 0.001) ( Fig. 2A). The pH i increased gradually and reached a plateau within 5-10 min. After application of DMA, the alkalinization Fig. 1. EPO increases astrocyte glutamate uptake and stimulates Na,K-ATPase in astrocytes undergoing OGD. The increased glutamate uptake is dependent on Na,K-ATPase. [ 3 H] D-aspartate uptake was used as an indicator of glutamate uptake. (A) [ 3 H] D-aspartate uptake measured in primary astrocytes during OGD. Incubation with 5 nM EPO for 5 min significantly increased [ 3 H] Daspartate uptake (*p < 0.05). (B) To assess Na, K-ATPase activity in primary astrocytes exposed to OGD, ouabain dependent 86 rubidium uptake was measured. Both without glutamate (left panel) and with glutamate (right panel), 5 min of EPO exposure increased ouabain dependent rubidium uptake (*p < 0.05). (C) [ 3 H] D-aspartate uptake measured in primary astrocytes during OGD with and without EPO. When Na,K-ATPase was inhibited by 2 mM ouabain, the increased [ 3 H]D-aspartate uptake induced by EPO was abolished both without (left panel) and with (right panel) aspartate. Values are means ± SD. by EPO was progressively attenuated (7.27 ± 0.018-7.14 ± 0.017, p < 0.001) ( Fig. 2A).

Effect of EPO on astrocyte glutamate uptake under OGD is dependent on NHE function
Next, we investigated the role of NHE for the effect of EPO on glutamate uptake in OGD. When 200 µM aspartate was applied to stimulate glutamate uptake, EPO increased [ 3 H] D-aspartate uptake by 31 % in astrocytes exposed to ODG (31 % ± 5.2 %, p < 0.05) (Fig. 3B). When cells were co-incubated with DMA, the EPO effect on astrocyte [ 3 H] D-aspartate uptake was abolished (Fig. 3B). DMA alone had no effect on [ 3 H] D-aspartate-uptake (Fig. 3A), indicating that the EPO stimulation of glutamate uptake in OGD is dependent on NHE function.

EPO enhances astrocyte lactate release in OGD
Astrocyte lactate release has been suggested to be tightly coupled to astrocyte glycolysis (Magistretti, 2006), which in turn is dependent on pH i under hypoxic conditions (Swanson and Benington, 1996). After 5 min incubation, EPO caused a 35 % increase in lactate release in primary astrocytes undergoing OGD (35 % ± 4.6 %, p < 0.01). When NHE was inhibited by DMA the effect on lactate release was abolished. DMA alone did not attenuate lactate release (Fig. 4A).

EPO increases intrinsic NADH in astrocytes exposed to OGD
Further we explored the effects of EPO on astrocyte glycolysis and energy metabolism in astrocyte-neuron co-cultures undergoing OGD. Real time intrinsic fluorescence of NADH was used as a sensitive Values are means ± SD, n; number of cells. (B) Summarized data on pH i restoration caused by EPO in OGD after cellular acidification (decrease in pH i ) induced by glutamate. When NHE was inhibited by DMA, the pH i recovery was significantly reduced (*p < 0.001). (C) The left panel shows a representative curve of calibrated pH i in one individual cell. Astrocytes exposed to OGD were incubated with DMA, glutamate and EPO as indicated in the top bar and with horizontal bars. The restoration of pH i (i.e. pH i increase) caused by EPO was impeded when DMA was present (20 µM, ~10 min-15 min). The right panel shows the summarized pH i data (*p < 0.001, n = 31). indicator of astrocyte glycolytic energy metabolism (Kasischke et al., 2004). Astrocytes were identified as SR101 positive (SR101+) cells and neurons as SR101 negative (SR101− ) cells after the experiments. EPO induced a transient NADH increase in both SR101+ and SR101− cells, suggesting enhanced glycolysis in both cell types (Fig. 4B). The maximal fluorescence changes (representing the peaks of transient NADH increase) were significantly increased compared to baseline. The increase in NADH was significantly higher in SR101+ cells compared to SR101− cells (4.6 % versus 2.1 %, p < 0.05), suggesting that the EPO stimulation of NADH production is more robust in astrocytes than in neurons.

Discussion
Research on neuroprotection at the cellular level in ischemic brain injury has traditionally focused on neurons. With increasing understanding of astrocyte-neuron metabolic coupling (Pellerin, 2005) and glutamate dynamics during ischemia (Ottersen et al., 1996), astrocyte function in ischemia has attracted justified increasing attention.
As an efficient tissue-protective agent (Brines and Cerami, 2005) EPO has been shown to offer neuroprotection in various animal models of brain ischemia (Bernaudin et al., 1999;Wang et al., 2004). One randomized clinical trial in ischemic stroke patients showed higher mortality in the EPO group attributed to intracerebral hemorrhage, brain edema, and thromboembolic events (Ehrenreich et al., 2009). This trial used extremely high doses of EPO and showed that enhanced delivery did not improve the therapeutic efficacy but raised safety concerns. Other studies have shown thromboembolic events, delayed ischemic hemorrhage and tumor growth (Lee et al., 2019;Vittori et al., 2021;Ma et al., 2022). BBB hinders 98 % of therapeutic small molecules and nearly 100 % of therapeutic macromolecules from entering the CNS, and EPO-derivates with better BBB permeability and more specific desired activation of the EPOR might increase neuroprotection potency and alleviate adverse effects Chiu et al., 2020;Peng et al., 2020;Li et al., 2022). In research using in vivo models with BBB disruption or in vitro models without BBB, EPO can still be used and provide valuable data on effects.
The mechanisms of the desired effects are probably several, however not fully understood. We previously showed that EPO could attenuate the glutamate effect on water permeability in astrocytes and decrease neurological symptoms in a model of primary brain edema (Gunnarson et al., 2009). Further, we found that administration of EPO following TBI in an animal model decreased early cytotoxic brain edema and preserved structural and functional properties of the BBB, leading to attenuation of the vasogenic edema response (Blixt et al., 2018).
The aim of the present study was to evaluate short-term effects of EPO on astrocytes under ischemic conditions, and to study mechanisms involved.
Using the OGD-model, we found that EPO within minutes caused a Na,K-ATPase dependent increase in glutamate uptake by 22 % in astrocytes. Glutamate uptake was further increased, to 31 %, when glutamate was added to the cultures. As expected, glutamate caused a decrease in pHi due to the concomitant uptake of protons along with glutamate. We found that EPO restored this intracellular acidification and normalized pH i . EPO also increased astrocyte lactate release by 35 % in cells undergoing OGD. The sodium/hydrogen exchanger NHE is involved in the regulation of intracellular pH by enabling exchange of sodium and hydrogen. When NHE was blocked by the selective blocker DMA, the EPO effect on pH i and lactate were abolished. In neuronastrocyte co-cultures, EPO increased NADH production significantly in both cell types, with a significant larger increase by 4,6% in astrocytes (SR101+ cells), compared to 2,1%in neurons (SR101− cells) (Fig. 5).
The extracellular glutamate concentration is controlled by astrocytes and must be fine-tuned to allow for neurotransmission and to avoid cytotoxicity (Hara and Snyder, 2007;Yin et al., 2020). In complete ischemia, glutamate homeostasis can be compromised due to acute energy depletion resulting in failing energy-dependent pump activity (Dallas et al., 2007;Passlick et al., 2021). It has been suggested that astrocytes can delay energy depletion through anaerobic glycolysis (Rossi et al., 2007;Díaz-García et al., 2017). However, during ischemia, pronounced intracellular acidification in astrocytes has been observed due to the massive glutamate uptake (Danbolt, 2001;Azarias et al., 2011;Levy et al., 1998). Cellular acidification can inhibit glycolysis in astrocytes (Peak et al., 1992;Swanson and Benington, 1996), even if it also has been shown that glutamate uptake itself can stimulate astrocyte aerobic glycolysis (Pellerin and Magistretti, 1994).
Here we found that EPO increased glutamate uptake and restored the acidification associated with glutamate uptake in astrocytes undergoing OGD, i.e. a potential beneficial effect to avoid cytotoxicity. Since we found that EPO restored astrocyte pH i , we speculate that this effect may be one of the mechanisms for the enhanced glutamate uptake caused by EPO in astrocytes (Swanson et al., 1995).
EPO has been shown to initiate proliferative responses in hematopoietic cells by activation of the sodium-hydrogen exchanger (NHE) (Rich et al., 1998). NHE is also one of the major pH i regulators for cortical astrocytes (Kintner et al., 2004). In accordance with this, we found that the EPO-induced pH i increase (i.e. restoration of lowered pH) in astrocytes undergoing OGD was attenuated by blocking NHE. The EPO-induced pH i increase was observed regardless of the presence of glutamate or not, suggesting that the effect is due to NHE activation rather than modulation of glutamate transporter function.

Fig. 4. EPO enhances lactate release and increases astrocyte intrinsic NADH fluorescence in OGD. (A)
Lactate release was significantly increased by 5 min of EPO treatment in astrocytes undergoing OGD. Co-incubation with the NHE inhibitor DMA abolished the increased lactate release caused by EPO (*p < 0.01). (B) To evaluate NADH production astrocyte-neuron co-cultures were exposed to OGD. The left upper image shows intrinsic NADH fluorescence in astrocyte-neuron co-culture exposed to OGD. The left lower image shows the merged channels of intrinsic NADH fluorescence and cultures loaded with SR101. The SR101 positive cells (SR101+/astrocytes) and SR101 negative cells (SR101− /neurons) are indicated by arrows. The right panel shows the summarized data of the maximal change in NADH fluorescence after EPO application. The NADH fluorescence increase was significant in both SR101+ and SR101− cells, but the relative difference was higher in SR101+ cells (*p < 0.05, n = 13). Values are means ± SD, n; number of cells.
Lactate production, export and metabolism in astrocytes is vital for survival and function in both astrocytes and neurons. Restoration of astrocyte pH i can enhance glycogen breakdown and promote glycolysis in aglycemia (Choi et al., 2012). The stimulation of glycolysis can in turn result in excessive production and externalization of lactate (Pellerin and Magistretti, 1994;Pellerin et al., 2007). It can therefore be proposed that the EPO-induced restoration of the lowered pH i caused by glutamate, may be beneficial for astrocyte function by preserving or restoring cellular glycolysis. Indeed, we found that EPO caused a pH i -coupled cellular lactate release, indicating that enhanced glycolysis was induced by EPO via intracellular pH regulation during OGD, and that EPO could increase NADH production in both astrocytes and neurons, adding support to a stimulation of metabolism (Singhal et al., 2018). The NADH production was significantly greater in astrocytes than in neurons. Lactate has, as an energy substrate, been shown to have a neuroprotective role in ischemia both in vitro and in vivo (Cater et al., 2003;Berthet et al., 2009;Ichai et al., 2009Ichai et al., , 2013Bisri et al., 2016). Therefore, the observed lactate release could theoretically constitute one component of the neuroprotective effect induced by EPO in ischemic stroke (Sargin et al., 2010). For neurons to use lactate as an energy source, sufficient oxygen supply is required (Pellerin, 2005), and very high levels of lactate can be toxic (Henriksen et al., 1992;Berthet et al., 2009). Our results suggest that EPO can modulate astrocyte function to maintain local homeostasis via pH i regulation under ischemic stress. The contribution of this effect to neuroprotection in vivo remains to be determined.
After observing the EPO enhanced pH i dependent glycolysis and glutamate uptake we investigated the metabolic coupling between these effects. When NHE was inhibited, the increase in glutamate uptake caused by EPO was abolished. Glutamate uptake is highly dependent on sodium gradient across the plasma membrane (Danbolt, 2001;Belov Kirdajova et al., 2020;Schousboe, 2020). Ischemia may, by lower intracellular ATP, impede this by decreasing Na,K-ATPase activity (Silver and Erecinska, 1997). We found that the EPOincreased astrocyte glutamate uptake was dependent on Na,K-ATPase function. The cellular energy state correlates with Na,K-ATPase pump function via ATP production (Silver and Erecinska, 1997). We suggest that EPO enhanced glycolysis EPO contributes to ATP production and provide energy to the crucial Na,K-ATPase.
In summary, insights into the effects and mechanisms of neuroprotective pharmacological agents at the cellular level should provide valuable information regarding their therapeutic potential. Based on the present study, we suggest that EPO attenuates the injurious effects of ischemia by enhancement of astrocyte function and metabolism, and that EPO thus may maintain several astrocyte homeostatic preserving mechanisms that in turn can be beneficial for neurons at risk. Whether this will limit progression of injury remains to be studied. If ischemia persists, glycogen depletion will eventually occur, indicating a time point where neuronal damage is inevitable.

Limitations
There is an abundance of uncertainty related to transference of results from cellular models to human appliances and clinical outcome in trials. Conducting experiments on isolated cells instead of on functional tissue may, due to absence of cell-cell interaction, both over-and underestimate the relevance of acquired data. Parts of the knowledge gap may however, to some extent, be enlightened by results from research on isolated astrocytes, as here. The present study has however not explored intracellular signaling pathways or further analyzed potential direct and/or indirect effects and this should be further explored.

Conclusion
In conclusion, our findings in this study on cultured astrocytes indicate that EPO has positive effects on astrocyte glutamate uptake and metabolism in a model of cerebral ischemia. The effects are, at least partly, mediated by preservation of NHE function and Na,K-ATPase activity (Fig. 5). EPO may therefore, by these mechanisms, contribute to neuroprotection in ischemic conditions such as TBI with secondary ischemic injury and in primary brain ischemia. The findings suggest a potential neuroprotective role of EPO via astrocytes in ischemia.

Fig. 5.
Summary of results; EPO effects on astrocytes undergoing OGD. EPO increases intracellular pH, externalization of lactate, and increases NADH production in both astrocytes and neurons. EPO also stimulates NKA and enhances glutamate uptake. EPOs effect on glutamate uptake is dependent on both NKA and NHE, and the effect on pH i and lactate is dependent on NHE. NKA: Na,K-ATPase; EAAT: excitatory amino acid transporter; NHE: Na/H-exchanger.

Ethics statement
Animal care and experimental procedures were conducted in accordance with EC directive 86/609/EEC. All experiments were approved by the Stockholm Northern Regional Ethical Committee, Stockholm, Sweden, and carried out in accordance with the Swedish National Law on ethical use of animals.

Cell cultures and oxygen-glucose deprivation
To obtain neuron-conditioned astrocytes in primary culture, prefrontal cortex was dissected from male Sprague-Dawley rat pups (B&K Universal) at embryonic day 18. The cells were tryptinized with 0.25 % trypsin (Gibco, Invitrogen) for 15 min at 37 C. Cells were seeded onto 90 mm plates precoated with Poly-Ornithine (80 µg/mL, Sigma) and grown in Neurobasal® Medium (Gibco, Invitrogen) containing 50 U/mL penicillin and 50 µg/mL streptomycin supplemented with 2 mM L-glutamine and 10 % FBS. The medium was changed twice a week. Cells were replated on uncoated glass after 2 weeks to obtain pure astrocyte cultures and cultured for another week. To obtain neuron-conditioned medium, as described above, cells were tryptinized and 2 × 10 6 cells were seeded onto 90 mm plates precoated with Poly-Ornithine and grown in MEM containing 2 mM L-glutamine and 10 % horse serum. The medium was replaced by Neurobasal® Medium containing 50 U/mL penicillin and 50 µg/mL streptomycin supplemented with 2 mM Lglutamine and B27 (1×) (Gibco, Invitrogen) after 4 h. Half of the medium was changed twice a week with medium without glutamine. Cells were cultured for 20 days. Neuron-conditioned medium was harvested and added to primary astrocyte cultures with 20 % total medium volume 1 to 2 days before experiments. To obtain astrocyte-neuron co-cultures, 0.1 × 10 6 cells were seeded onto 30 mm plates precoated with Poly-Ornithine and grown in Neurobasal® Medium containing 50 U/mL penicillin and 50 µg/mL streptomycin supplemented with 2 mM Lglutamine, B27 and 1 % FBS. Half of the medium was changed twice a week with medium without glutamine. The cells were cultured for 3 weeks.
To achieve OGD, glucose-free artificial CSF (aCSF: in mM, NaCl 125, KCl 2.5 or MgCl 2 1, NaH 2 PO 4 1.25, CaCl 2 2, NaHCO 3 25, Mannitol 25, pH 7.3-7.4) was saturated with a 95 % N 2 /5% CO 2 mixture for 30 min. The cells were washed twice with OGD buffer and incubated with OGD buffer for 15 min before all experiments. Fresh cells were used in all experiments. All drugs in treatment were purchased from Sigma unless otherwise noted.

Glutamate transporter assay
Glutamate uptake in astrocytes is essential to avoid glutamate excitotoxicity. Due to robust metabolism of glutamate, [ 3 H] D-aspartate uptake was used as an indicator of glutamate transporter function. Daspartate is known to be a mixed-type noncompetitive inhibitor of Lglutamate uptake in astrocytes, why D-aspartate was used for stimulation in these experiments. Astrocyte primary cultures were washed twice and incubated for 15 min at RT with OGD buffer or same buffer with 20 µM 5-(N,N-Dimethyl)amiloride hydrochloride (DMA, Sigma) before experiments. For Na,K-ATPase dependent assay, a 15 min preincubation of 2 mM ouabain together with OGD buffer was performed. Together with treatment, 0.1 µCi [ 3 H] D-aspartate (PerkinElmer Life) and 200 µM unlabeled D-aspartate (Sigma) were added to each sample. Rat recombinant EPO (5 nM, Sigma) was used in the experiments and the reaction was allowed to proceed for 5 min at RT. The buffer was then removed, and the reaction stopped by adding ice-cold buffer. The cells were washed twice more with cold buffer and 1 ml 0.1 M NaOH was added for cell lysis. The lysate was added to 5 ml scintillation cocktail for counting. All results were adjusted for background by subtracting counts obtained from samples excluding protein from the assay. The [ 3 H] Daspartate uptake was normalized to protein concentration. Results are reported as percentage of controls (basal [ 3 H] D-aspartate without EPO or DMA) where control was set to 100 %. Each sample was analyzed as the mean value of duplicates. Experiments were repeated 3 times.

Rubidium-86 uptake
Na,K-ATPase is essential for intracellular sodium homeostasis and the driving force for sodium-coupled glutamate uptake. Ouabain dependent Rubidium-86 uptake was measured to evaluate Na,K-ATPase activity in primary cultured astrocytes. Cells were washed twice and incubated for 15 min at room temperature (RT) with OGD buffer with or without 2 mM ouabain. Thereafter 2.5 µCi of rubidium-86 (PerkinElmer Life) was added to each sample. After 5 min at RT the buffer was removed, and reaction was stopped by adding ice cold buffer containing 5 mM BaCl 2 . Cells were washed twice more with cold buffer and 1 ml 0.1 M NaOH was added for cell lysis and counted in a liquid scintillation counter. The rubidium-86 uptake was normalized to protein concentration and each sample was analyzed as the mean value of duplicates. Ouabain dependent rubidium-86 uptake was calculated as the difference between ouabain-free counts and ouabain incubation counts. Experiments were repeated 3 times. Results are reported as percentage of controls where the value of the controls was set to 100 %.

Intracellular pH measurement
Intracellular pH (pH i ) is of importance for intracellular and mitochondrial function, such as glycolysis and mitochondrial energy production. To assess pH i in astrocytes, primary astrocytes were loaded with the BCECF-AM (1 µM, Invitrogen) for 20 min in serum-free medium at 37 • C. Radiometric imaging (F 490nm /F 440nm ) was performed using a closed chamber (POCmini-2, PeCon GmbH) mounted on a Zeiss Axioskop 2 microscope with a 40×/1.3NA epifluorescent oil-immersion objective. Emission fluorescence was detected with a CCD camera (Hamamatsu ORCA-ER C4742-95) via an image-intensifier unit (Hamamatsu C9016). BCECF-AM loaded cells were excited at wavelength 440 and 490 nm and emission fluorescence was recorded with a BP510-540 nm filter every 6 s. All experiments were performed at 37 • C using glucose-free aCSF continuously bubbled by 95 % N 2 /5 %CO 2 (i.e. OGD). At the end of each experiment, cells were permeabilized with 2 µM Nigericin and a calibration curve was obtained with a series of calibration solutions of known pH values. All devices were controlled, and data were analyzed using meta-Fluor software (Molecular Devices, Downingtown, PA).

Lactate assay
Although lactate production can reflect energy failure, lactate is produced by astrocytes and is regarded as a substrate for neurons under ischemic energy-depletion. Extracellular lactate was measured by using a lactate assay kit (BioVision, Catalog #K607-100). Astrocyte primary cultures were washed twice and incubated for 15 min at room temperature (RT) with OGD buffer before experiments. After 5 min treatment, the OGD buffer was collected, and lactate was measured by fluorometric assay according to the manufacturer's instructions. Protein contents in the supernatants were determined using a RC DC Protein Assay (Bio-Rad Laboratories). The lactate production was normalized to protein concentration. Results are reported as percentage of controls where control treatment was set to 100 %. Each sample was analyzed as duplicates and repeated 5 times of which the mean value was calculated.

Intrinsic NADH fluorescence measurement
To explore effects on intracellular energy production, we further measured nicotinamide adenine dinucleotide (NADH). It has been shown that dynamics of astrocyte intrinsic fluorescence is dominated by NADH (Kasischke et al., 2004). Dynamic NADH production in live cells was obtained by using closed POCmini-2 chamber with 30 mm glass. The chamber was mounted on inverted Zeiss Axiovert 200 with a 40×/ 1.3NA epifluorescent oil-immersion objective. Emission fluorescence was detected with Zeiss AxioCam HRm. Cells were excited by HBO 100 arc-lamp with BP355-375 nm excitation filter and emission fluorescence was recorded with 455 nm LP every 10 s. The experiments were performed using glucose-free aCSF continuously bubbled by 95 % N 2 /5 % CO 2 (i.e. OGD) in astrocyte-neuron co-cultures. Sulforhodamine 101 (SR101, Invitrogen) was used as a specific marker of astrocytes (Nimmerjahn et al., 2004). Following every experiment, astrocytes were identified by SR101 positive staining (0.5 µM, 5 min; Excitation 587/25, Emission 605 LP) (Fig. 4B).

Statistical analysis
Statistical analysis was performed using a two-sided Student's t test or when appropriate ANOVA followed by Tukey. Calculations were performed using SPSS 20 (IBM SPSS Statistics, Armonk, NY). All data were normally distributed. P values <0.05 were considered statistically significant. Data are expressed as mean ± SD. Graphs were made using Prism 6.02 (GraphPad Software, La Jolla, CA).

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.