Oxidative Stress Underlies the Ischemia/Reperfusion-Induced Internalization and Degradation of AMPA Receptors

Stroke is the fifth leading cause of death annually in the United States. Ischemic stroke occurs when a blood vessel supplying the brain is occluded. The hippocampus is particularly susceptible to AMPA receptor-mediated delayed neuronal death as a result of ischemic/reperfusion injury. AMPA receptors composed of a GluA2 subunit are impermeable to calcium due to a post-transcriptional modification in the channel pore of the GluA2 subunit. GluA2 undergoes internalization and is subsequently degraded following ischemia/reperfusion. The subsequent increase in the expression of GluA2-lacking, Ca2+-permeable AMPARs results in excitotoxicity and eventually delayed neuronal death. Following ischemia/reperfusion, there is increased production of superoxide radicals. This study describes how the internalization and degradation of GluA1 and GluA2 AMPAR subunits following ischemia/reperfusion is mediated through an oxidative stress signaling cascade. U251-MG cells were transiently transfected with fluorescently tagged GluA1 and GluA2, and different Rab proteins to observe AMPAR endocytic trafficking following oxygen glucose-deprivation/reperfusion (OGD/R), an in vitro model for ischemia/reperfusion. Pretreatment with Mn(III)tetrakis(1-methyl-4-pyridyl)porphyrin (MnTMPyP), a superoxide dismutase mimetic, ameliorated the OGD/R-induced, but not agonist-induced, internalization and degradation of GluA1 and GluA2 AMPAR subunits. Specifically, MnTMPyP prevented the increased colocalization of GluA1 and GluA2 with Rab5, an early endosomal marker, and with Rab7, a late endosomal marker, but did not affect the colocalization of GluA1 with Rab11, a marker for recycling endosomes. These data indicate that oxidative stress may play a vital role in AMPAR-mediated cell death following ischemic/reperfusion injury.


Introduction
Stroke is responsible for the death of nearly 140,000 people in the United States annually. Of the estimated 800,000 strokes that occur in the United States each year, approximately 87% are ischemic [1]. Ischemic stroke occurs when a blood vessel in the brain is blocked, hindering the vessel's ability to provide oxygen and nutrients to brain tissue. While it is necessary to reintroduce blood flow to the infarcted area, this act also results in further damage by inflammation, oxidative stress, and delayed neuronal death (DND) within vulnerable neuronal populations, including CA1 hippocampal pyramidal neurons [2,3]. During ischemia, the lack of energy available disrupts ATP-dependent processes that maintain ionic gradients, which are critical to cellular survival. Disrupting the ionic balance leads to excessive release of neurotransmitters, including glutamate, which are unable to be effectively cleared from synapses [4]. Excessive stimulation of N-methyl-D-aspartate receptors (NMDARs) by glutamate is a contributing factor to DND [5][6][7][8], but methyl-4-pyridyl)porphyrin (MnTMPyP), a cell-permeable superoxide dismutase mimetic, and examined the endocytic trafficking of GluA2 and GluA2 AMPAR subunits following OGD/R. Interestingly, we identified oxidative stress-dependent differential trafficking patterns between GluA1 and GluA2 AMPAR subunits following OGD/R.

OGD/R Did Not Alter Cell Viability in U251-MG Cells
We first sought to determine whether our model of OGD/R resulted in increased cell death compared to normoxic controls. There was no significant difference in cell viability by trypan blue exclusion in the OGD/R-exposed cells compared to the normoxic control ( Figure 1), confirming that our in vitro model of ischemia/reperfusion is not lethal and any differences in protein levels are not a result of cell death. Oxygen-glucose deprivation/reperfusion (OGD/R) does not alter the viability of U251-MG cells. U251-MG cells underwent OGD for 20 min, followed by reperfusion for 60 min. There was no significant difference between the viability of normoxic control U251-MG cells and OGD/R-treated U251-MG cells (n = 3; unpaired student t-test). ns denotes no significance.

MnTMPyP Pretreatment Did Not Prevent Agonist-Induced Internalization of AMPARs
To examine whether our superoxide dismutase mimetic, MnTMPyP, prevents agonist-induced internalization, we first exposed U251-MG cells to 3 min of agonist stimulation (100 µM AMPA; 100 µM cyclothiazide) in the presence or absence of MnTMPyP (100 µM). Following the 3 min of AMPA/cyclothiazide stimulation, cells were fixed at 5-or 10min time-points. There was an increase in colocalization between GluA1 and Rab5, and GluA2 and Rab5 following AMPA stimulation ( Figure 2). This increased colocalization indicates that GluA1 and GluA2 AMPAR subunits were internalized and presented within Rab5-positive early endosomes. Agonist-induced internalization of GluA1 and GluA2 in Rab5-positive endosomes was not antagonized by MnTMPyP. Figure 1. Oxygen-glucose deprivation/reperfusion (OGD/R) does not alter the viability of U251-MG cells. U251-MG cells underwent OGD for 20 min, followed by reperfusion for 60 min. There was no significant difference between the viability of normoxic control U251-MG cells and OGD/R-treated U251-MG cells (n = 3; unpaired student t-test). ns denotes no significance.

MnTMPyP Pretreatment Did Not Prevent Agonist-Induced Internalization of AMPARs
To examine whether our superoxide dismutase mimetic, MnTMPyP, prevents agonistinduced internalization, we first exposed U251-MG cells to 3 min of agonist stimulation (100 µM AMPA; 100 µM cyclothiazide) in the presence or absence of MnTMPyP (100 µM). Following the 3 min of AMPA/cyclothiazide stimulation, cells were fixed at 5-or 10-min time-points. There was an increase in colocalization between GluA1 and Rab5, and GluA2 and Rab5 following AMPA stimulation ( Figure 2). This increased colocalization indicates that GluA1 and GluA2 AMPAR subunits were internalized and presented within Rab5positive early endosomes. Agonist-induced internalization of GluA1 and GluA2 in Rab5positive endosomes was not antagonized by MnTMPyP.
OGD/R has been previously shown to produce reactive oxygen species (ROS) in hippocampal and cortical neurons [22], and we wanted to examine whether we observe a similar effect in our model system. Utilizing a nitroblue tetrazolium (NBT) assay, we determined that U251-MG cells produce ROS in a time-dependent manner with OGD/R exposure. Phorbol 12-myristate 13-acetate (PMA) activates NADPH oxidase resulting in increased ROS production [38] and was utilized as a positive control for the NBT assay. ROS production following OGD/R was first observed at reperfusion time-point 15 min, and maximally produced with 60 min of reperfusion. Pretreatment with MnTMPyP ameliorated this OGD/R-induced increase in ROS (Figure 3). Based upon previous results [39,40] within our lab, we sought to determine whether superoxide production had a direct effect on AMPAR subunit endocytic trafficking following OGD/R.   and Rab5, and GluA2 and Rab5, respectively. This indicates internalization following agonist stimulation, which was unaffected by pretreatment with MnTMPyP, a superoxide scavenger. (C,D) are quantifications of (A,B), respectively (n = 30). **** p < 0.0001; ANOVA with Tukey's post hoc test comparing agonist conditions to corresponding controls (no agonist stimulation ± MnTMPyP). Data are expressed as mean ± SEM. Scale bar is 100 µM. Arrows indicate colocalization.
creased ROS production [38] and was utilized as a positive control for the NBT assay. ROS production following OGD/R was first observed at reperfusion time-point 15 min, and maximally produced with 60 min of reperfusion. Pretreatment with MnTMPyP ameliorated this OGD/R-induced increase in ROS (Figure 3). Based upon previous results [39,40] within our lab, we sought to determine whether superoxide production had a direct effect on AMPAR subunit endocytic trafficking following OGD/R.

Pretreatment with MnTMPyP Attenuated the OGD/R-Induced Internalization of GluA1 and GluA2 AMPAR Subunits
Since our model system, U251-MG cells, produced ROS during OGD/R, we wanted to determine whether pretreatment with a superoxide scavenger, MnTMPyP, affected the endocytic trafficking of AMPAR subunits, GluA1 and GluA2. Following OGD/R, GluA1 and GluA2 subunits were both highly colocalized with Rab5-positive early endosomes. Interestingly, the time-points of GluA1 and GluA2 internalization differed with OGD/R. GluA1 was highly colocalized with Rab5 at OGD/R-5 and OGD/R-15-min time-points. GluA2 colocalized with Rab5-positive early endosomes solely at the OGD/R-5-min timepoint, indicating that GluA2 moved quickly within the endocytic pathway from an early endosome to a sorting endosome, or a Rab7-positive late endosome. GluA1 and GluA2 transiently transfected in U251-MG cells internalized following OGD/R exposure. To determine whether oxidative stress played a role in this internalization, we pretreated the U251-MG cells with MnTMPyP. Pretreatment with MnTMPyP attenuated the OGD/R-induced colocalization of both GluA1 and GluA2 with Rab5-positive early endosomes ( Figure 4). . Pretreatment of OGD/R-exposed U251-MG cells with MnTMPyP antagonized reactive oxygen species (ROS) generation. Spectrophotometric quantification of ROS following either OGD/R or normoxic conditions with or without pre-treatment of MnTMPyP (100 µM) utilizing nitroblue tetrazolium chloride (NBT), to examine ROS production (n = 3). Phorbol 12-myristate 13-acetate (PMA; 1 µM for 15 min) was used as a positive control. * p < 0.05; ** p < 0.01; **** p < 0.0001; ANOVA with Tukey's post hoc test comparing OGD/R conditions to normoxic control ± MnTMPyP). Data are expressed as mean ± SEM.

Pretreatment with MnTMPyP Attenuated the OGD/R-Induced Internalization of GluA1 and GluA2 AMPAR Subunits
Since our model system, U251-MG cells, produced ROS during OGD/R, we wanted to determine whether pretreatment with a superoxide scavenger, MnTMPyP, affected the endocytic trafficking of AMPAR subunits, GluA1 and GluA2. Following OGD/R, GluA1 and GluA2 subunits were both highly colocalized with Rab5-positive early endosomes. Interestingly, the time-points of GluA1 and GluA2 internalization differed with OGD/R. GluA1 was highly colocalized with Rab5 at OGD/R-5 and OGD/R-15-min time-points. GluA2 colocalized with Rab5-positive early endosomes solely at the OGD/R-5-min time-point, indicating that GluA2 moved quickly within the endocytic pathway from an early endosome to a sorting endosome, or a Rab7-positive late endosome. GluA1 and GluA2 transiently transfected in U251-MG cells internalized following OGD/R exposure. To determine whether oxidative stress played a role in this internalization, we pretreated the U251-MG cells with MnTMPyP. Pretreatment with MnTMPyP attenuated the OGD/R-induced colocalization of both GluA1 and GluA2 with Rab5-positive early endosomes ( Figure 4).

GluA1 and GluA2 AMPAR Subunits Did Not Undergo OGD/R-Induced Degradation in the Presence of MnTMPyP
Following GluA1 and GluA2 AMPAR subunit internalization to Rab5-positive endosomes, experiments were performed to examine whether GluA1 or GluA2 AMPAR subunits are trafficked to Rab11-positive recycling endosomes in cells subjected to OGD/R. GluA1 remained highly colocalized with Rab11-positive recycling endosomes under both normoxic and OGD/R conditions. Treatment with MnTMPyP had no effect on this colocalization under normoxic nor OGD/R conditions ( Figure 5A,C). Conversely, GluA2 was not highly colocalized with Rab11-positive recycling endosomes under any conditions, with or without MnTMPyP treatment. (Figure 6A,C). It is likely that there was a recycling pool of GluA1 homomeric AMPARs under both basal and pathologic conditions, but not GluA2-containing AMPARs.   To examine whether AMPAR subunits were being trafficked to the degradative pathway following OGD/R, we examined their colocalization with Rab7-positive late endosomes destined for lysosomal degradation. Both GluA1 and GluA2 subunits were highly colocalized with Rab7-positive late endosomes following OGD/R, likely indicating they were fated for degradation ( Figure 5B,D and Figure 6B,D). Pretreatment with MnTMPyP, scavenging superoxide, prevented GluA1 and GluA2 AMPAR subunits being sorted to Rab7-positive late endosomes. This may occur because MnTMPyP treatment prevents the AMPAR subunits from internalizing with OGD/R or it prevents GluA1 and GluA2 subunits from trafficking to Rab7-positive endosomes with OGD/R exposure.
To further confirm that GluA1 and GluA2 subunits are degraded with OGD/R, we performed Western blot analysis to examine their respective protein levels. As expected, both GluA1 and GluA2 are degraded in a time-dependent manner with OGD/R. GluA1 protein levels decreased at OGD/R-30-min and maximally at OGD/R-60-min time-points. GluA2 protein levels decreased at the OGD/R-60-min time-point. Pretreatment with MnTMPyP prevented the OGD/R-induced degradation of GluA1 partially, and GluA2 completely ( Figure 7). nternalization, maximally at OGD/R-5 min; the effect was ameliorated with pretreatment of MnTMPyP. (D) Quantificaion of (C) microscopy results (n = 30). ** p < 0.01; **** p < 0.0001; ANOVA with Tukey's post hoc test comparing OGD/R onditions to the corresponding control (normoxic ± MnTMPyP). Data are expressed as mean ± SEM. Scale bar is 100 µm. rrows indicate colocalization.

GluA1 and GluA2 AMPAR Subunits Did not Undergo OGD/R-Induced Degradation in the Presence of MnTMPyP
Following GluA1 and GluA2 AMPAR subunit internalization to Rab5-positive endosomes, experiments were performed to examine whether GluA1 or GluA2 AMPAR subunits are trafficked to Rab11-positive recycling endosomes in cells subjected to OGD/R. GluA1 remained highly colocalized with Rab11-positive recycling endosomes under both normoxic and OGD/R conditions. Treatment with MnTMPyP had no effect on this colocalization under normoxic nor OGD/R conditions ( Figure 5A,C). Conversely, GluA2 was not highly colocalized with Rab11-positive recycling endosomes under any conditions, with or without MnTMPyP treatment. (Figure 6A,C). It is likely that there was a recycling pool of GluA1 homomeric AMPARs under both basal and pathologic conditions, but not GluA2-containing AMPARs. To examine whether AMPAR subunits were being trafficked to the degradative pathway following OGD/R, we examined their colocalization with Rab7-positive late endosomes destined for lysosomal degradation. Both GluA1 and GluA2 subunits were highly colocalized with Rab7-positive late endosomes following OGD/R, likely indicating they were fated for degradation ( Figures 5B,D and 6B,D). Pretreatment with MnTMPyP, scavenging superoxide, prevented GluA1 and GluA2 AMPAR subunits being sorted to Rab7positive late endosomes. This may occur because MnTMPyP treatment prevents the AM-PAR subunits from internalizing with OGD/R or it prevents GluA1 and GluA2 subunits from trafficking to Rab7-positive endosomes with OGD/R exposure. To further confirm that GluA1 and GluA2 subunits are degraded with OGD/R, we performed Western blot analysis to examine their respective protein levels. As expected, both GluA1 and GluA2 are degraded in a time-dependent manner with OGD/R. GluA1 protein levels decreased at OGD/R-30-min and maximally at OGD/R-60-min time-points. GluA2 protein levels decreased at the OGD/R-60-min time-point. Pretreatment with MnTMPyP prevented the OGD/R-induced degradation of GluA1 partially, and GluA2 completely (Figure 7).

Discussion
In this study, we have demonstrated that ischemic/reperfusion-induced internalization and subsequent endocytic trafficking of GluA1 and GluA2 AMPAR subunits to late endosomes is mediated in an oxidative stress signaling pathway. The loss of GluA2-containing AMPARs at the plasma membrane following ischemia/reperfusion occurs in vulnerable areas of the brain, such as the hippocampus, leading to delayed neuronal death

Discussion
In this study, we have demonstrated that ischemic/reperfusion-induced internalization and subsequent endocytic trafficking of GluA1 and GluA2 AMPAR subunits to late endosomes is mediated in an oxidative stress signaling pathway. The loss of GluA2containing AMPARs at the plasma membrane following ischemia/reperfusion occurs in vulnerable areas of the brain, such as the hippocampus, leading to delayed neuronal death [9]. The subunit composition switch from GluA2-containing, Ca 2+ -impermeable AMPARs to GluA2-lacking, Ca 2+ -permeable AMPARs occurs via the internalization and degradation of the GluA2 subunit [41], transcriptional and translational downregulation of GluA2 levels [16], and the increase in GluA2-lacking AMPARs at the plasma membrane [14]. In this study, we have examined the endocytic trafficking pathways of both GluA1 and GluA2 AMPAR subunits following OGD/R and the role of oxidative stress in mediating these processes.
It is well-studied that oxidative stress exacerbates cell death following ischemia/ reperfusion [39,40,[42][43][44][45]. There are three distinct temporal oxidative stress mechanisms that contribute to neuronal injury following ischemia/reperfusion. With hypoxic onset, mitochondria generate an initial burst of ROS followed closely by xanthine oxidase activation. Once reperfusion occurs, NADPH oxidase produces a large burst of ROS in a calciumdependent manner [22]. In this study, we observed a robust increase in ROS production during the reperfusion phase of OGD/R in U251-MG cells. Astrocytes do express NADPH oxidase [46], so it is possible that the large increase in ROS is due to activation of NADPH oxidase in U251-MG cells. Alternatively, there could be delayed ROS production from mitochondria and/or xanthine oxidase. Further studies using pharmacological inhibitors and/or genetic tools are needed to identify the source of ROS in our U251-MG OGD/R model system.
This study is the first to indicate that endocytic trafficking of GluA1 and GluA2 subunits can be modulated with ROS following OGD/R. Utilizing MnTMPyP, a superoxide dismutase mimetic, we identified ROS-mediated trafficking of GluA1 and GluA2 with OGD/R that is distinct from agonist-induced internalization. MnTMPyP treatment had no effect on agonist-induced internalization of GluA1 and GluA2 subunits, but attenuated both the internalization and degradation of GluA1 and GluA2 subunits with OGD/R. Under physiological conditions, endocytosis of GluA2 is mediated by protein kinase C alpha (PKCα)-dependent phosphorylation of GluA2 Ser880 residue. PKCα is activated by increased intracellular calcium and is redox-sensitive [47][48][49]. Further studies are needed to examine the relationship between ROS and PKCα activation. Upon activation, PKCα translocates the plasma membrane by protein interacting with C kinase 1 (PICK1) [50], where it phosphorylates GluA2 [51]. The phosphorylation of GluA2 at the Ser880 residue, and increased association with PICK1, increases the internalization of GluA2-containing AMPARs [19,52] thus reducing the surface population of AMPARs. This event allows for a transient increase in calcium permeable AMPARs at the plasma membrane, which is a critical component for LTP under physiologic conditions [53], but is harmful when uncontrolled during pathophysiological conditions like ischemia/reperfusion [13] (Figure 8). It has also been reported that ubiquitination of GluA1 facilitates agonist-induced endocytosis [57] and lysosomal targeting [58]. GluA1-containing, GluA2-lacking AMPARs are recruited to the plasma membrane surface during LTP by palmitoylation of A-kinase anchoring protein 150 (AKAP150), a scaffold that binds to and regulates GluA1 phosphorylation and trafficking [59]. Upon AMPA agonist stimulation, myosin Vb captures and mobilizes Rab11-positive, recycling endosomes for AMPAR insertion [60]. The mechanisms for GluA1 trafficking during LTP are well understood, but whether all these mechanisms are conserved during pathologic ischemic/reperfusion injury remains to be determined. It is particularly interesting that GluA1 homomeric AMPAR subunits are present within Rab11-positive endosomes, but GluA2 homomeric AMPAR subunits are not during OGD/R. This suggests that there is an unknown differential trafficking mechanism between GluA1 and GluA2 with regards to receptor recycling during OGD/R. We utilized an immortalized cell line in this study, which is a study limitation. It is unclear whether these results are reproducible in different cell lines, in primary cultures of neuronal cells, or in vivo, and further studies will need to be conducted to confirm these results.
This study is the first to examine how oxidative stress mediates endocytic trafficking of GluA1 and GluA2 AMPAR subunits following OGD/R, but further studies are needed to fully characterize the role of superoxide in mediating the endocytic trafficking and degradation of the GluA1 and GluA2 AMPAR subunits with OGD/R.

Cell Culture
U251-MG cells, an immortalized human astrocytoma cell line, were maintained in The present study examined the effect of superoxide on the internalization, degradation, and recycling of GluA1 and GluA2 AMPAR subunits with OGD/R in U251-MG cells. The superoxide scavenger, MnTMPyP, prevented both the internalization of both GluA1 and GluA2 receptor subunits following OGD/R. In the absence of superoxide, the degradation of GluA1 was partially rescued with OGD/R, whereas the degradation of GluA2 was completely prevented with MnTMPyP treatment. Interestingly, GluA1 was present within Rab11-positive recycling endosomes under both normoxic and OGD/R conditions, with MnTMPyP having no effect. This suggests that there is a pool of GluA1 homomeric AMPARs that is unique to GluA1, as GluA2 was not present in Rab11-positive endosomes under any conditions studied.
The recycling of GluA1 homomeric AMPARs to the plasma membrane, a critical component of LTP [53], is a tightly regulated process. Activity-dependent endocytic sorting of GluA1 depends on phosphorylation [54] at the Ser845 residue by PKA, which promotes receptor insertion and decreases receptor endocytosis [55]. The fate of internalized GluA1containing AMPARs to either Rab7-containing endosomes, fated for degradation, or Rab11containing endosomes, fated for recycling, depends upon phosphorylation [56]. It has also been reported that ubiquitination of GluA1 facilitates agonist-induced endocytosis [57] and lysosomal targeting [58]. GluA1-containing, GluA2-lacking AMPARs are recruited to the plasma membrane surface during LTP by palmitoylation of A-kinase anchoring protein 150 (AKAP150), a scaffold that binds to and regulates GluA1 phosphorylation and trafficking [59]. Upon AMPA agonist stimulation, myosin Vb captures and mobilizes Rab11-positive, recycling endosomes for AMPAR insertion [60]. The mechanisms for GluA1 trafficking during LTP are well understood, but whether all these mechanisms are conserved during pathologic ischemic/reperfusion injury remains to be determined. It is particularly interesting that GluA1 homomeric AMPAR subunits are present within Rab11positive endosomes, but GluA2 homomeric AMPAR subunits are not during OGD/R. This suggests that there is an unknown differential trafficking mechanism between GluA1 and GluA2 with regards to receptor recycling during OGD/R.
We utilized an immortalized cell line in this study, which is a study limitation. It is unclear whether these results are reproducible in different cell lines, in primary cultures of neuronal cells, or in vivo, and further studies will need to be conducted to confirm these results.
This study is the first to examine how oxidative stress mediates endocytic trafficking of GluA1 and GluA2 AMPAR subunits following OGD/R, but further studies are needed to fully characterize the role of superoxide in mediating the endocytic trafficking and degradation of the GluA1 and GluA2 AMPAR subunits with OGD/R.

Microscopy
Following OGD/R or AMPA stimulation experiments, transfected U251-MG cells were imaged via confocal microscopy (60X objective), Olympus FluoView1000/IX81 confocal microscope system (Olympus Corporations of the Americas Headquarters, Center Valley, PA, USA) and colocalization coefficients (Pearson's Correlation Coefficient) were obtained using FIJI software (freely available from National Institutes of Health, Bethesda, MD, USA). All microscopy and data analyses were conducted in a blind manner.

Western Blotting
Protein concentration was determined using a bicinchoninic acid assay (BCA; Thermo Fisher Scientific, Waltham, MA, USA) and samples were heated at 100 • C in LDS/reducing agent buffer (Thermo Fisher Scientific, Waltham, MA, USA) and resolved via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Samples were then transferred to a nitrocellulose membrane (Bio-Rad, Berkeley, CA, USA). Blots were blocked for one hour at room temperature with 5% (w/v) non-fat dry milk in tris buffered saline, 0.1% (v/v) tween-20, pH 7.5 (TBS-T). After blocking, blots were incubated with the primary antibody overnight at 4 • C at the concentration indicated. Blots were incubated with the secondary antibody for 1 h at room temperature. The following antibodies were used: anti-RFP (1:500; Invitrogen, Waltham, MA, USA), anti-β-Actin (1:10,000; Cell Signaling, Danvers, MA, USA), anti-mouse (1:2000; Cell Signaling, Danvers, MA, USA), and anti-rabbit (1:2000; Cell Signaling, Danvers, MA, USA). Immunoreactive bands were visualized and captured with a Fuji imaging system using enhanced chemiluminescence. Bands were analyzed using Fuji Image-Gauge software (Fujifilm North America Corporation, Valhalla, NY, USA).

Trypan Blue Exclusion
Following OGD/R, U251-MG cells were incubated in trypsin/EDTA (0.05%) and centrifuged at 1000× g for 10 min. The cell pellet was resuspended in complete medium and trypan blue (1:1 dilution). Cell viability was determined using trypan blue exclusion on Countess II Cell Counter (Thermo Fisher Scientific, Waltham, MA, USA).

NBT Assay
U251-MG cells were plated at 5 × 10 4 in 6 well plates and, once they reached 70% confluency, were used for experiments. NBT assay was adapted from [65,66]. Briefly, after 4 h of serum starvation, 0.5 mg/mL NBT was added to cells and incubated for one hour. Excess NBT was washed away 3 times with warmed PBS. U251-MG cells underwent 20 min of OGD and various reperfusion time points: 0, 15, 30, or 60 min with or without MnTMPyP treatment. Phorbol 12-myristate 13-acetate (PMA; 1 µM; Sigma Aldrich, St. Louis, MO, USA) was used as a positive control. Following treatment, cells were fixed with absolute methanol, air dried, and formazan deposits were dissolved in 100% dimethyl sulfoxide and KOH [2M]. Absorbance was measured at 620 nm with Spectra Max Gemini M2 plate reader (Molecular Devices, Sunnyvale, CA, USA).

Data Analysis and Scientific Rigor
Either Student's t-test or one-way ANOVA with Tukey's post hoc test were conducted using GraphPad Prism 8 software to determine statistical significance. All imaging and analyses were performed blindly.  Acknowledgments: U251-MG cells were generously gifted by Richard Bridges, University of Montana. GluA1 and GluA2 fluorescently tagged plasmids were generously gifted from Richard Huganir, Johns Hopkins. We acknowledge the WSU Microscopy Core for use of their facilities and equipment. We acknowledge the University of Montana Molecular Histology & Fluorescence Imaging Core Facility for use of their equipment.

Conflicts of Interest:
The authors declare no conflict of interest.