Blockade of AMPA Receptor Regulates Mitochondrial Dynamics by Modulating ERK1/2 and PP1/PP2A-Mediated DRP1-S616 Phosphorylations in the Normal Rat Hippocampus

N-Methyl-D-aspartate receptor (NMDAR) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) activations induce fast and transient mitochondrial fragmentation under pathophysiological conditions. However, it is still unknown whether NMDAR or AMPAR activity contributes to mitochondrial dynamics under physiological conditions. In the present study, MK801 (a non-competitive NMDAR antagonist) did not affect mitochondrial length in hippocampal neurons as well as phosphorylation levels of dynamin-related protein 1 (DRP1)-serine (S) 616, extracellular-signal-regulated kinase 1/2 (ERK1/2), c-Jun N-terminal kinase (JNK), p38 mitogen-activated protein kinase (p38 MAPK) and AMPAR. In contrast, perampanel (a non-competitive AMPAR antagonist) elongated mitochondrial length in neurons concomitant with diminishing phosphorylations of DRP1-S616, ERK1/2, and JNK, but not p38 MAPK. Perampanel also reduced protein phosphatase (PP) 1, PP2A and PP2B phosphorylations, indicating activations of these PPs which were unaffected by MK801. U0126 (an ERK1/2 inhibitor) elongated mitochondrial length, accompanied by the reduced DRP1-S616 phosphorylation. SP600125 (a JNK inhibitor) did not influence mitochondrial length and DRP1 phosphorylations. Okadaic acid (a PP1/PP2A inhibitor) reduced mitochondrial length with the up-regulated DRP1-S616 phosphorylation, while CsA (a PP2B inhibitor) increased it with the elevated DRP1-S637 phosphorylation. Co-treatment of okadaic acid or CsA with perampanel attenuated the reductions in DRP1-S616 and -S637 phosphorylation without changing DRP1 expression level, respectively. GYKI 52466 (another non-competitive AMPAR antagonist) showed the similar effects of perampanel on phosphorylations of DRP1, ERK1/2, JNK, PPs, and GluR1 AMPAR subunits. Taken together, our findings suggest that a blockade of AMPAR may regulate the cooperation of ERK1/2- and PP1/PP2A for the modulation of DRP1 phosphorylations, which facilitate mitochondrial fusion.


INTRODUCTION
Mitochondria are dynamic organelles responsible for the generation of ATP via oxidative phosphorylation, which is essential for cell viability. To maintain mitochondrial homeostasis, mitochondria change their shapes by mitochondrial dynamics (fusion and fission; Li et al., 2004;Verstreken et al., 2005). Mitochondrial fission (fragmentation) is necessary for the biogenesis of mitochondria and their quality control by mitophagic elimination (Youle and Van der Bliek, 2012). Mitochondrial fusion (elongation) is also involved in mitochondrial quality control by repartition of lipids, proteins and mitochondrial DNA Ono et al., 2001). Therefore, the aberrant mitochondrial dynamics result in the impaired bioenergetics, excessive generation of reactive oxygen species (ROS), loss of mitochondrial membrane potential, dysfunction of endogenous respiration and the release of pro-apoptotic factors from mitochondria, which trigger cell death (Morris and Hollenbeck, 1993;Ligon and Steward, 2000;Weiss et al., 2000;Frank et al., 2001;Olichon et al., 2003;Chen et al., 2005;Youle and Van der Bliek, 2012).
During the course of this study, we validated the effects of a non-competitive NMDAR antagonist (MK801) and non-competitive AMPAR antagonists (perampanel and GYKI 52466) on mitochondrial length in the rat hippocampal neurons to elucidate the receptor-mediated mechanisms underlying the mitochondrial dynamics in vivo. Here, we demonstrate that the blockade of AMPAR led to mitochondrial elongations via regulating ERK1/2-and PP1/PP2A-mediated DRP1-S616 phosphorylations. Therefore, our findings suggest that AMPAR activity may contribute to mitochondrial dynamics under physiological condition.

Experimental Animals and Chemicals
Male Sprague-Dawley (SD) rats (7 weeks old, Daehan Biolink, South Korea) were used in the present study. Animals were given a commercial diet and water ad libitum under controlled conditions (22 ± 2 • C, 55 ± 5% and a 12:12 light/dark cycle with lights). Animal protocols were approved by the Institutional Animal Care and Use Committee of Hallym University (Chunchon, South Korea). The number of animals used, and their suffering were minimized in all cases. All reagents were obtained from Sigma-Aldrich (St. Louis, MO, United States), except as noted.

Western Blot
After animals (n = 7 in each group, respectively) were sacrificed via decapitation, the hippocampi were obtained. The hippocampal tissues were homogenized and protein concentration determined using a Micro BCA Protein Assay Kit (Pierce Chemical, Rockford, IL, United States). Western blot was performed by the standard protocol. The primary antibodies used in the present study were listed in Table 1. The bands were detected and quantified on an ImageQuant LAS4000 system (GE Healthcare, United States). Since OPA1, ERK1/2 and pERK1/2 antibodies, but not others, clearly showed two bands and were changed to the same degree, we quantified both bands. As an internal reference, rabbit anti-β-actin primary antibody (1:5000) was used. The values of each sample were normalized with the corresponding amount of β-actin. The ratio of phosphoprotein to total protein was described as phosphorylation level.

Immunohistochemistry and Measurement of Mitochondrial Length
Animals (n = 7 in each group, respectively) were anesthetized with urethane (1.5 g/kg, i.p.) and then transcardially perfused with 4% paraformaldehyde (pH 7.4). The brains were removed and post-fixed overnight in the same solution, then sequentially placed in 30% sucrose at 4 • C. Coronal sections were cut at a thickness of 30 µm on a cryostat. During sections, we confirmed the intracerebroventricular location of a brain infusion kit. Free-floating sections were first incubated with 10% normal goat serum (Vector, Burlingame, CA, United States) in PBS for 30 min at room temperature. Sections were then incubated in a mitochondrial marker (Mitochondrial complex IV subunit 1, MTCO1, Abcam, United Kingdom) as the primary antibody (in PBS containing 0.3% Triton X-100) at room temperature overnight. After washing in PBS, sections were incubated for 1 h in a Cy3-conjugated secondary antiserum. For negative control, the hippocampal tissues were incubated with pre-immune serum instead of primary antibody. As the result of the negative control test, no immunoreactive structure was observed. Images were captured using an AxioImage M2 microscope or a confocal laser scanning microscope (LSM 710, Carl Zeiss, Inc., Oberkochen, Germany). Individual mitochondrion length in PV cells and CA1 neurons (n = 20/section) was measured by using ZEN lite software (Blue Edition, Carl Zeiss, Inc., Oberkochen, Germany) following 3D-reconstruction: based on our previous study (Kim and Kang, 2017;Ko and Kang, 2017), 25 serial images (z-stack, 1 µm) were obtained from each hippocampal section. Serial images were stacked, aligned, visualized and converted into 3D images using ZEN lite program. Thereafter, individual mitochondrial length (long axis) was measured. Two different investigators who were blind to the classification of tissues performed the measurement of mitochondrial length.

Data Analysis
One-way ANOVA was used to determine statistical significance of data. For post hoc multiple comparisons, Bonferroni's test was applied. A p-value below 0.05 was considered statistically significant.

Effects of MK801 and Perampanel on Mitochondrial Dynamics
First, we investigated whether MK801 or perampanel affects expressions or phosphorylations of mitochondrial dynamics-related molecules. Figure 1 shows that both MK801 and perampanel did not influence DRP1, OPA1, and MFN1/2 expression levels. However, perampanel, but not MK801, reduced DRP1-S616 phosphorylation to 0.4-fold of vehicle level in the hippocampus (p < 0.05, n = 7, respectively; Figures  The average mitochondrial length in CA1 neurons was 1.39, 1.41, and 2.1 µm in vehicle-, MK801-and perampanel-treated animals, respectively. In DGCs, the average mitochondrial length was 1.37, 1.5, and 2.78 µm in vehicle-, MK801-and perampanel-treated animals, respectively. Therefore, these findings indicate that blockade of AMPAR, but not NMDAR, may facilitate mitochondrial fusion in the hippocampus under physiological condition.

Effects of MK801 and Perampanel on Protein Phosphatases
Since DRP1 activity is also regulated by various protein phosphatase activities (Campello and Scorrano, 2010), protein phosphatases (PPs) would be involved in the different effects of MK801 and perampanel on mitochondrial dynamics. Thus, we tested whether MK801 and perampanel affect the activities of PPs in the hippocampus. MK801 did not influence the expression and phosphorylation of PP1, PP2A, and PP2B (Figures 4A-G and Supplementary Figure 4). Consistent with our previous study (Kim et al., 2019), perampanel reduced PP1, PP2A, and PP2B phosphorylations to 0.66-, 0.6-, and 0.46-fold of vehicle levels, although it did not affect their expression levels (p < 0.05 vs. vehicle, n = 7, respectively; Figures 4A-G and Supplementary Figure 4). Since the phosphorylations inhibit the activities of protein phosphates (Hashimoto et al., 1988;MacDonnell et al., 2009), our findings indicate that perampanel, but not MK801, may activate PP1, PP2A, and PP2B in the rat hippocampus.

Effects of Various Inhibitors of Kinases and PPs on Mitochondrial Dynamics
Based on the present data aforementioned, we validated the effects of U0126 (an ERK1/2 inhibitor), SP600125 (a JNK inhibitor), okadaic acid (a PP1/PP2A inhibitor) and CsA (a PP2B inhibitor) on DRP1 phosphorylation and mitochondrial length to confirm the underlying mechanism of perampanel for mitochondrial dynamics. All chemicals did not affect DRP1 expression in the hippocampus (Figures 5A,B and Supplementary Figure 5). However, U0126 reduced DRP1-S616, not -S637, phosphorylation (p < 0.05 vs. vehicle, n = 7, respectively; Figures 5A,C,D and Supplementary Figure 5). Okadaic acid and CsA increased DRP1-S616 and -S637 phosphorylation, respectively   Figure 5), while co-treatment of CsA did not affect it. These findings indicate that both ERK1/2 and PP1/PP2A may coordinately regulate DRP1-S616 phosphorylation. U0126 and SP600152 reduced ERK1/2 and JNK phosphorylations without altering their expression levels (p < 0.05 vs. vehicle, n = 7, respectively; Figures 5A,E-H and Supplementary Figure 5). Consistent with previous studies (Liu et al., 2008;Hou et al., 2013), okadaic acid and CsA also enhanced increased PP1/PP2A and PP2B phosphorylation levels without affecting their expression levels, respectively (p < 0.05 vs. vehicle, n = 7, respectively; Figures 6A-G and Supplementary  Figure 6). In addition, co-treatment of U0126 did not influence the effects of okadaic acid and CsA on PP phosphorylations (Figures 6A-G and Supplementary Figure 6). These findings also suggest that each inhibitor may not show off-target effects in the present study.
Similar to DRP1 phosphorylations, U0126 and CsA elongated mitochondrial length in CA1 neurons and DGCs (p < 0.05 vs. vehicle, n = 7, respectively; Figures 7A-D), while okadaic acid resulted in mitochondrial fragmentation in both neuronal subpopulations (p < 0.05 vs. vehicle, n = 7, respectively; Figures 7A-D). SP600125 did not affect mitochondrial length, as compared to vehicle (Figures 7A-D). Together with the data obtained from perampanel treatment, our findings suggest that blockade of AMPAR induced by perampanel may elongate mitochondrial length via inhibiting ERK1/2 activity as well as enhancing PP1/PP2A activities.

Effects of Okadaic Acid and CsA on DRP1, ERK1/2, JNK, and PP Phosphorylations Induced by Perampanel
Since perampanel effectively activated (reduced phosphorylation level) PP1/PP2A and PP2B, we also applied co-treatment of okadaic acid, or CsA, with perampanel to investigate the role of PPs on DRP1 phosphorylation induced by perampanel. Co-treatment of okadaic acid and CsA with perampanel significantly attenuated the reductions in DRP1-S616 and -S637 phosphorylations without changing DRP1 expression level, respectively (p < 0.05 vs. vehicle and perampanel, respectively; n = 7, respectively; Figures 8A-D and Supplementary Figure 7). Co-treatments of okadaic acid and CsA did not affect the reductions in ERK1/2 and JNK phosphorylations induced by perampanel (Figures 8E-H  and Supplementary Figure 7). In addition, co-treatment of okadaic acid and CsA significantly alleviated the reductions in PP1/PP2A and PP2B phosphorylations without altering PPs expression levels, respectively (p < 0.05 vs. vehicle, n = 7, respectively; Figures 9A-G and Supplementary Figure 8).

Effects of GYKI 52466 on DRP1, ERK1/2, JNK, and PP Phosphorylations
To confirm the role of blockade of AMPAR in DRP1 phosphorylations, we also investigated the effects of GYKI 52466, another allosteric AMPAR inhibitor (a non-competitive AMPAR antagonist), on DRP phosphorylations and activities of kinases as well as PPs. Similarly to perampanel, GYKI 52466 reduced DRP1-S616 phosphorylation but not S637 phosphorylation level,   Figure 8). Therefore, our findings indicate that allosteric AMPAR inhibition may commonly affect ERK1/2, JNK, PP1, PP2A, and PP2B activities in the normal rat hippocampus.

Effects of MK801 and Perampanel on GluR1 Phosphorylation
N-Methyl-D-aspartate receptor activity affects AMPAR functionality. Briefly, GluR1, a subunit of AMPAR, is phosphorylated at serine (S) 831 site when NMDAR activation, while GluR1-S845 phosphorylation was decreased (Ai et al., 2011). Therefore, the effects of MK801 and perampanel on AMPAR phosphorylations are noteworthy, which would lead to the distinct mitochondrial dynamics. In the present study, MK801 did not affect GluR1 expression and its S831 and S845 phosphorylations (Figure 10 and Supplementary  Figure 9). In contrast to MK801, perampanel, and GYKI 52466 reduced GluR1 expression to 0.7-and 0.77-fold of vehicle level, respectively (p < 0.05 vs. vehicle, n = 7, respectively; Figures 10A,B and Supplementary Figure 9). However, GluR1-S831 phosphorylation was increased to 1.38-and 1.35-fold of vehicle level, respectively (p < 0.05 vs. vehicle, n = 7, respectively; Figures 10A,C and Supplementary Figure 9). GluR1-S845 phosphorylation was unaffected by MK801, perampanel or GYKI 52466 (Figures 10A,D and Supplementary  Figure 9). These findings indicate that MK801 may not affect AMPAR activity, and that the selective AMPA inhibition by perampanel or GYKI 52466 may evoke mitochondrial elongation under physiological condition.

DISCUSSION
The major findings in the present study are that blockade of AMPAR elongated mitochondrial length by regulating ERK1/2-FIGURE 8 | Effects of co-treatment of perampanel (PER) with okadaic acid (OA) or CsA, and GYKI 52466 on expressions and phosphorylations of DRP1, ERK1/2, and JNK. Co-treatment of okadaic acid and CsA with perampanel significantly attenuates the reductions in DRP1-S616 and -S637 phosphorylations without changing DRP1 expression level, but not ERK1/2 and JNK phosphorylations, induced by perampanel. Similar to perampanel, GYKI 52466 reduces DRP1-S616 phosphorylation, but not S637 phosphorylation level, without affecting DRP1 expression. GYKI 52466 also abolishes ERK1/2 and JNK phosphorylation levels.  N-Methyl-D-aspartate receptor-and AMPAR-mediated intracellular Ca 2+ overloads evoke necrosis or delayed apoptosis of neurons via bioenergetic collapse, activation of calpains, oxidative stress, and mitochondrial dysfunctions (Choi, 1992;Arundine and Tymianski, 2004;Li and Wang, 2016). The impaired mitochondrial dynamics also contribute to neuronal death in various neurological diseases (Rintoul et al., 2003;Grohm et al., 2012;Kim et al., 2014;Ruiz et al., 2018). Interestingly, the roles of excessive mitochondrial fission in neuronal death show a receptor specific manner, since Mdivi-1 attenuates mitochondrial fission induced by NMDA, but not AMPA (Ruiz et al., 2018). However, it is still unknown whether NMDAR or AMPAR activity contributes to mitochondrial dynamics under physiological conditions. In the present study, MK801 increased ERK1/2 phosphorylation, but reduced PKC phosphorylation. These findings are consistent with previous reports demonstrating the effect of MK801 on ERK1/2, JNK, and p38 MAPK (Chen et al., 2003;Seo et al., 2007). However, we found that MK801 did not affect mitochondrial length in both CA1 neurons and DGCs as well as phosphorylation levels of DRP1, JNK, and p38 MAPK. Thus, these findings indicate that blockade of NMDAR may not contribute to mitochondrial dynamics under physiological condition.
Unlike MK801, the present data demonstrate that perampanel elongated mitochondrial length in both CA1 neurons and DGCs. Furthermore, perampanel and GYKI 52466 reduced DRP1-S616 phosphorylation accompanied by diminishing phosphorylations (activations) of ERK1/2, PKC and JNK, but not p38 MAPK. Based on these results, it is likely that the blockade of AMPAR would inhibit mitochondrial fission by decreasing ERK1/2 and JNK activities. However, our other studies reveal that U0126 (an ERK1/2 inhibitor), but not SP600125 (a JNK inhibitor), abrogated DRP1-S616 phosphorylation and evoked mitochondrial elongation. Since AMPAR activation increases ERK1/2 phosphorylation (Mao et al., 2004), therefore, our findings suggest that AMPAR inhibition may lead to mitochondrial elongation by inhibiting ERK1/2 activity.
In the present study, co-treatment of okadaic acid, but not CsA, abrogated U0126-induced reduction in DRP1-S616 phosphorylation, although co-treatment of U0126 did not influence the effects of okadaic acid and CsA on PP phosphorylations. Furthermore, co-treatments of okadaic acid and CsA significantly alleviated the reductions in PP1/PP2A and PP2B phosphorylations induced by perampanel, respectively. Co-treatment of okadaic acid and CsA with perampanel also attenuated the reductions in DRP1-S616 and -S637 phosphorylations without changing DRP1 expression level, respectively. However, co-treatments of okadaic acid and CsA did not affect perampanel-mediated reductions in ERK1/2 and JNK phosphorylations. These findings indicate that both ERK1/2 and PP1/PP2A may coordinately regulate DRP1-S616 phosphorylation, and that each inhibitor may not show off-target effects in the present study. Thus, our findings suggest that blockade of AMPAR may regulate the cooperation of ERK1/2and PP1/PP2A for the modulation of DRP1 phosphorylations.
It is well-known that the phosphorylations of GluR1 subunit of AMPAR at S831 and S845 sites increase the conductance of AMPAR and potentiate rapid excitatory neurotransmission (Derkach et al., 1999;Snyder et al., 2000). Interestingly, AMPAR activity is distinctly regulated by NMDAR functionality: NMDAR activation phosphorylates GluR1 subunit at S831 site, but decreases GluR1-S845 phosphorylation (Ai et al., 2011). Therefore, it is considerable that the discrepancies of the effects of MK801 and perampanel on mitochondrial dynamics would be relevant to an indirect regulation of NMDAR to GluR1 phosphorylation. Consistent with a previous study (Zhang et al., 2014), the present study reveals that MK801 did not affect GluR1 expression and its phosphorylations. However, perampanel decreased GluR1 expression, while it increased GluR1-S831 phosphorylation ratio with the unaltered GluR1-S845 phosphorylation ratio. Since GluR1 phosphorylations represent the enhanced AMPAR-mediated synaptic currents (Zhang et al., 2016), our findings indicate that the increase in S831 phosphorylation ratio of GluR1 may be one of the adaptive responses for the reductions in AMPAR functionality or GluR1 expression level by perampanel. These findings also support that AMPAR rather than NMDAR may regulate mitochondrial dynamics under physiological condition.
On the other hand, the effects of AMPAR antagonist on GluR1 expression have been controversial. NBQX, a competitive AMPAR antagonist, increases GluR1 expression (Thiagarajan et al., 2005) or not (Maeng et al., 2008). Similarly to our recent study (Kim et al., 2019), both perampanel and GYKI 52466 down-regulated GluR1 expression, but increased GluR1-S831 phosphorylation. Although we could not explain the underlying mechanisms concerning these phenomena, it is plausible that the properties of non-competitive (allosteric) AMPAR antagonists, such as peramapenl and GYKI 52466, would result in the distinct effect on GluR1 expression as compared to competitive AMPAR antagonists. In addition, the serine residue phosphorylation plays an important role in regulating the conductance and trafficking of AMPAR (Lee et al., 1998(Lee et al., , 2000Shepherd and Huganir, 2007). Therefore, our findings suggest that the elevated GluR1-S831 phosphorylation by perampanel and GYKI 52466 may be one of the adaptive responses for the diminished GluR1 expression or AMPAR-mediated currents. This is because AMPAR activity in response to AMPA is regulated in negative feedback manners (Snyder et al., 2000). Further studies are needed to elucidate the regulatory mechanisms of GluR1 expression by allosteric AMPAR antagonists.

CONCLUSION
To the best of our knowledge, the present data demonstrate the previously unreported AMPAR-mediated underlying mechanism for mitochondrial dynamics under physiological condition. Furthermore, our findings suggest that the regulation of AMPAR functionality may be one of the therapeutic targets for neurological diseases related to aberrant mitochondrial dynamics.