Inhibition of AKT/GSK3β/CREB Pathway Improves the Responsiveness to AMPA Receptor Antagonists by Regulating GRIA1 Surface Expression in Chronic Epilepsy Rats

α-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR) has been reported as one of the targets for treatment of epilepsy. Although maladaptive regulation of surface expression of glutamate ionotropic receptor AMPA type subunit 1 (GRIA1) subunit is relevant to the responsiveness to AMPAR antagonists (perampanel and GYKI 52466) in LiCl-pilocarpine-induced chronic epilepsy rats, the underlying mechanisms of refractory seizures to AMPAR antagonists have yet been unclear. In the present study, we found that both AMPAR antagonists restored the up-regulations of GRIA1 surface expression and Src family-mediated glycogen synthase kinase 3β (GSK3β)-Ca2+/cAMP response element-binding protein (CREB) phosphorylations to control levels in responders (whose seizure activities were responsive to AMPAR) but not non-responders (whose seizure activities were uncontrolled by AMPAR antagonists). In addition, 3-chloroacetyl indole (3CAI, an AKT inhibitor) co-treatment attenuated spontaneous seizure activities in non-responders, accompanied by reductions in AKT/GSK3β/CREB phosphorylations and GRIA1 surface expression. Although AMPAR antagonists reduced GRIA2 tyrosine (Y) phosphorylations in responders, they did not affect GRIA2 surface expression and protein interacting with C kinase 1 (PICK1) protein level in both responders and non-responders. Therefore, our findings suggest that dysregulation of AKT/GSK3β/CREB-mediated GRIA1 surface expression may be responsible for refractory seizures in non-responders, and that this pathway may be a potential target to improve the responsiveness to AMPAR antagonists.


Introduction
Epilepsy is a common clinical neurological disease characterized by spontaneous seizures due to abnormal neuronal discharges. Over decades, the incidence of epilepsy presents an increasing tendency with a reported number of 65 million patients worldwide. About 30% temporal lobe epilepsy (TLE) patients show refractory seizures that are uncontrolled by standard medications with antiepileptic drugs (AEDs). In addition, patients with intractable seizures have a high mortality rate [1,2]. Given the inefficacy of AEDs and the unrelenting nature of intractable epilepsy, therefore, the necessity to explore the underlying mechanisms of generation of refractory seizures is undoubted.
Although the pathogenesis of epilepsy remains to be fully elucidated, hyper-activation of the glutamate receptor is one of the important causes of epilepsy. A massive release of glutamate and the subsequent over-activation of glutamate receptors cause aberrant neuronal hyper-excitability, followed by delayed neuronal death and secondary injury [3,4]. Recently, α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR) has been focused on as one of the major therapeutic targets for treatment of epilepsy [5] since AMPAR antagonists, but not N-methyl-D-aspartate receptor (NMDAR) antagonists, terminate status epilepticus (SE, a prolonged seizure activity) and seizure activity in animal models and preferentially suppress seizure-related long-term consequences [6][7][8].
AMPAR is one of the ionotropic glutamate receptors, which exists as a homomeric or heteromeric assembly of four subunits (glutamate ionotropic receptor AMPA type subunit (GRIA) 1-4) encoded by different genes [9]. GRIA subunits preferentially dimerize soon after the translocation in the endoplasmic reticulum (ER) and tetramerize prior to ER exit [10]. Activity-dependent phosphorylation/dephosphorylation of intracellular C-terminal sites on the AMPAR subunits GRIA1 and GRIA2 regulate their synaptic trafficking and functional properties [11,12]: Phosphorylation of GRIA1 at serine (S) 831 and/or S845 sites increases both open channel conductance and its surface expression [13,14]. In contrast, GRIA2 S880 phosphorylation facilitates its internalization [15]. Interestingly, GRIA2 subunit regulates the ion-permeable properties of AMPAR. GRIA2-containing AMPAR is permeable to Na + and K + but not Ca 2+ . In contrast, GRIA2-lacking AMPAR is highly permeable to Ca 2+ , which activates a series of intracellular protein kinases or immediate early genes. In the hippocampus of epileptic animals, GRIA2 surface expression is decreased [16,17], while >95% of AMPARs contain the GRIA2 subunit under physiological conditions [18]. Therefore, it is likely that increased membrane GRIA1/GRIA2 ratio may be an important component of the ictogenesis and the seizure-associated consequences by leading to a high Ca 2+ permeability of AMPAR in the epileptic brain. Indeed, we have reported that membrane GRIA1/GRIA2 ratio is significantly higher in chronic epilepsy rats than that in control animals, indicating the existing preponderance of Ca 2+ -permeable AMPAR, although both GRIA1 and GRIA2 levels are lower in the hippocampus. In addition, AMPAR antagonists (perampanel and GYKI 52466) decrease membrane GRIA1/GRIA2 ratio by reducing GRIA1, but not GRIA2, surface expression [19,20]. These effects of AMPAR antagonists are relevant to Src family/casein kinase 2 (CK2)/phosphatase and tensin homolog deleted on chromosome ten (PTEN)-mediated Ca 2+ /cAMP response element-binding protein (CREB) S133 phosphorylation, which are observed in responders (whose seizure activities are responsive to AMPA antagonists), but not non-responders (whose seizure activities are not uncontrolled) [19]. However, the up-stream signaling pathways regulating membrane GRIA1/GRIA2 ratio in the epileptic hippocampus have yet been elusive.
On the other hand, GRIA2 phosphorylations regulate their surface expression through interactions with glutamate receptor interacting protein 1 (GRIP1) and protein interacting with C kinase 1 (PICK1) that contains the postsynaptic density-95/Discs large/zona occludens-1 (PDZ) domain. GRIP1 enhances GRIA2 surface expression, whereas PICK1 facilitates its internalization [21,22]. Src family-mediated GRIA2 phosphorylations on tyrosine (Y) 869, Y873, and Y876 sites destabilize its interaction with GRIP1 and allow more GRIA2-PICK1 binding to promote its internalization [23,24]. In addition, protein kinase C (PKC)-mediated S880 phosphorylation also decreases GRIA2 binding to GRIP1, but not to PICK1, and results in rapid GRIA2 internalization [21]. However, the regulations of the GRIA2 subunit phosphorylations in response to AMPAR antagonists in the epileptic hippocampus are largely unknown. In an effort to understand the pathogenesis of intractable epilepsy, therefore, it will be noteworthy to more specifically investigate the underlying mechanisms of refractory seizures to AMPAR antagonists.
Here, we demonstrated that the anti-convulsive effects of AMPAR antagonists were closely relevant to the regulation of AKT/glycogen synthase kinase 3β (GSK3β)/CREBmediated GRIA1 surface expression rather than the modulation of PICK1-GRIA2 internalization. In addition, impairment of this signaling pathway resulted in refractory seizures in response to AMPAR antagonists, which were improved by 3-chloroacetyl indole (3CAI, an AKT inhibitor) co-treatment. Therefore, our findings suggest that AKT/GSK3β/CREB pathway may be a potential therapeutic strategy to improve the treatment of intractable TLE in response to AMPAR antagonists.

Experimental Animals and Chemicals
Male Sprague-Dawley (SD) rats (7 weeks old) were provided with a commercial diet and water ad libitum under controlled temperature, humidity, and lighting 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 (Hallym 2018-2, 26 April 2018 and Hallym 2018-21, 8 June 2018). All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA), except as noted. Figure 1 illustrates the design of the drug trial methodology, which was a modified protocol based on Ko and Kang [25]. Animals were intraperitoneally (i.p.) given LiCl (127 mg/kg) 24 h before the pilocarpine treatment. Animals were treated with pilocarpine (30 mg/kg, i.p.) 20 min after atropine methylbromide (5 mg/kg i.p.). Two hours after SE on-set, animals were administered diazepam (Valium; Hoffman la Roche, Neuilly sur-Seine, France; 10 mg/kg, i.p.) and dosage was repeated, as needed. Control animals received saline in place of pilocarpine. Animals were video-monitored 8 h a day for general behavior and occurrence of spontaneous seizures by 4 weeks after SE induction. Behavioral seizure severity was evaluated according to Racine's scale [26]: 1, immobility, eye closure, twitching of vibrissae, sniffing, facial clonus; 2, head nodding associated with more severe facial clonus; 3, clonus of one forelimb; 4, rearing, often accompanied by bilateral forelimb clonus; and 5, rearing with loss of balance and falling accompanied by generalized clonic seizures. We classified chronic epilepsy rats that showed behavioral seizures with seizure score ≥3 more than once.

Surgery
Control and epilepsy rats were implanted with monopolar stainless steel electrodes (Plastics One, Roanoke, VA, USA) in the right hippocampus (stereotaxic coordinates were −3.8 mm posterior; 2.0 mm lateral; −2.6 mm depth to bregma) under Isoflurane anesthesia (3% induction, 1.5-2% for surgery, and 1.5% maintenance in a 65:35 mixture of N 2 O:O 2 ). Animals were also implanted with a brain infusion kit 1 (Alzet, Cupertino, CA, USA) to infuse with 3CAI (an AKT inhibitor, 25 µM) into the right lateral ventricle (1 mm posterior; 1.5 mm lateral; −3.5 mm depth to the bregma, see below). Throughout surgery, core temperature of each rat was maintained at 37-38 • C. Electrode was secured to the exposed skull with dental acrylic.

Experiment II
Some non-responders in experiment I were given saline (i.p.) over a 7-day period. Thereafter, animals were daily given perampanel or GYKI 52466 by the aforementioned method. Rats were also connected with Alzet 1007D osmotic pump (Alzet, Cupertino, CA, USA) containing vehicle or 3CAI (25 µM) [28]. The pump was placed in a subcutaneous pocket in the dorsal region. In a pilot study and our previous study [28], this dosage of 3CAI did not show behavioral and neurological defects and could not change the seizure susceptibility and seizure severity in response to pilocarpine. After recording (18 h after the last drug treatment), animals were used for Western blot.

Western Blot
Animals were sacrificed by decapitation, and their hippocampi were obtained and homogenized in lysis buffer containing protease inhibitor cocktail (Roche Applied Sciences, Branford, CT, USA) and phosphatase inhibitor cocktail (PhosSTOP ® , Roche Applied Science, Branford, CT, USA). Thereafter, total protein concentration calibrated using a Micro BCA Protein Assay Kit (Pierce Chemical, Rockford, IL, USA). To analyze membrane GRIA subunit expressions, the bilateral hippocampal tissues (~100 mg) were washed gently with ice-cold PBS and chopped and homogenized for 10 strokes using a Dounce tissue homogenizer. Thereafter, membrane proteins were extracted with a subcellular Protein Fractionation Kit for Tissues (Thermo Scientific Korea, Seoul, South Korea), according to the manufacturer's instructions. Western blot was performed by the standard protocol: Sample proteins (10 µg) were separated on a Bis-Tris sodium dodecyl sulfate-poly-acrylamide gel (SDS-PAGE) and transferred to membranes. Membranes were incubated with 2% bovine serum albumin (BSA) in Tris-buffered saline (TBS; in mM 10 Tris, 150 NaCl, pH 7.5, and 0.05% Tween 20), and then reacted with primary antibodies (Table 1) overnight at 4 • C. After washing, membranes were incubated in a solution containing horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h at room temperature. Immunoblots were detected and quantified using ImageQuant LAS4000 system (GE Healthcare Korea, Seoul, South Korea). Optical densities of proteins were calculated with the corresponding amount of β-actin or N-cadherin.

Data Analysis
Seizure parameters (frequency, duration and Racine scores) were assessed by different investigators who were blind to the classification of animal groups and treatments. Shapiro-Wilk W-test was used to evaluate the values on normality. Mann-Whitney U-test, Wilcoxon signed rank test, Student's t-test, and paired Student's t-test were applied to determine statistical significance of data. Comparisons among groups were also performed using repeated measures ANOVA, Friedman test, and one-way ANOVA followed by Bonferroni's post hoc comparisons. A p-value less than 0.05 was considered to be significant.

Effects of AMPAR Antagonists on PICK1 Expression
PICK1 negatively regulates GRIA2 surface expression [21,22]. Src family and/or PKC-mediated GRIA2 phosphorylations facilitate GRIA2-PICK1 bindings, which lead to GRIA2 internalization and a relative increase of GRIA2-lacking, Ca 2+ -permeable AM-PARs [21,23,24]. However, both perampanel and GYKI 52466 decrease GRIA1/GRIA2 ratio by reducing GRIA1, but not GRIA2, surface expression [19,20]. Considering GRIA2 phosphorylations, it is likely that altered PICK1 expression may be involved in unaffected GRIA2 surface expression by AMPAR antagonists. To confirm this, we evaluated the effects of perampanel and GYKI 52466 on PICK1 expression in the epileptic hippocampus. In the present study, PICK1 protein level was lower in chronic epilepsy rats than that in controls (t (12) = 10.4, p < 0.001, Student t-test; Figure 6A,B and Figure S4), which was unaffected by AMPAR antagonists (F (4,27) = 0.2, p = 0.93, one-way ANOVA; Figure 6A,B). These findings indicate that reduced PICK1 expression may decrease its binding to phosphorylated GRIA2, which would serve to maintain GRIA2 surface expression as an adaptive response. In addition, it is unlikely that PICK1-mediated GRIA2 internalization may be a contributor to generation of refractory seizures in response to AMPAR antagonists.

Discussion
The principal findings of the present study were that AMPAR antagonists ameliorated spontaneous seizure activity by regulating surface expression of GRIA1 but not GRIA2. In addition, impaired AKT/GSK3β-mediated CREB S133 activation was involved in refractory seizures in response to AMPAR antagonists, which suggests that the regulation of these pathways may be a potential therapeutic strategy to improve the medication of intractable TLE.
Since the activation and desensitization kinetics of AMPAR are much faster than those of NMDAR [38], AMPAR antagonists have been focused to prevent ictogenesis and steadily progressive seizure-related brain pathologic plasticity in the epileptic brain [5]. Indeed, AMPAR antagonists terminate SE and seizure activity that are uncontrolled by NMDAR antagonists [6][7][8]. Among them, only perampanel is marketed for the treatment of focal epilepsy, which is a novel non-competitive AMPAR antagonist without affecting the NMDAR or kainate receptors [5], although its effects on signaling pathways remain to be elucidated. In the present study, both perampanel and GYKI 52466 ameliorated spontaneous seizure activity in 54% and 50% of animals (responders) in each group, concomitant with the reduced membrane expression of GRIA1 but not GRIA2. Furthermore, both AMPAR antagonists restored upregulated membrane/total GRIA1 ratio and membrane GRIA1/GRIA2 ratio to control levels. These effects of AMPAR antagonists were relevant to inhibition of AKT/GSK3β/CREB signaling pathway in responders. Because both AMPAR subunits are also located in endomembrane systems, such as the Golgi complex, endosomes, and ER [39,40], the GRIA1 and GRIA2 concentrations in membrane fraction may not directly represent the "pure surface expressions" in the present study. Considering that AMPAR trafficking is tightly regulated from the synthesis in ER to surface expression [41,42] and CREB regulates surface GRIA1 expression [37], however, the decreases in membrane/total GRIA1 ratio and membrane GRIA1/GRIA2 ratio indicate the diminished surface GRIA expression [19,20]. CREB activity (phosphorylation) is increased in human patients and animal models of epilepsy [35,36]. CREB phosphorylation is regulated by the activation of multiple signaling cascades, including GSK3β, PKC, protein kinase A (PKA), and Ca 2+ -calmodulin-dependent protein kinase II (CAMKII) [32,33,43,44]. In the present study, we found that GSK3β S9 (but not PKC) and CREB S133 phosphorylations were higher in the epileptic hippocampus than those in normal one, which were abolished by both AMPAR antagonists in responders, but not non-responders. Since the increased GSK3β-mediated CREB S133 phosphorylation participates in ictogenesis in human patients and animal models of epilepsy [32,35,36], these findings indicate that AMPAR antagonists may negatively regulate surface GRIA expression via GSK3β/CREB signaling pathway in responders.
PI3K/AKT-mediated GSK3β phosphorylation inhibits its activity, which is conversely regulated by PTEN [31,33]. Recently, we have reported that PTEN expression is decreased in the epileptic hippocampus, concomitant with the reduced Src/CK2-mediated phosphorylations [19]. Regarding phosphorylation-mediated inhibition of PTEN activity [45], we have postulated that the reduced PTEN phosphorylation (the increased activity) may be an adaptive response to the decreased expression in the epileptic hippocampus, while this alteration may be insufficient to suppress spontaneous seizure activities. This is because AMPAR antagonists further diminish PTEN phosphorylation only in responders, accompanied by increasing its expression, which is abrogated by dipotassium bisperoxovanadium(pic) dihydrate (BpV(pic), a PTEN inhibitor) [19,20]. In the present study, 3CAI-induced AKT inhibition improved the responsiveness to AMPAR antagonists in non-responders through GSK3β-mediated CREB regulation. Furthermore, 3CAI reduced surface GRIA1 expression in non-responders to AMPAR antagonists. Therefore, it is likely that the regulation of the PTEN/AKT/GSK3β/CREB pathway be essentially required for the anti-epileptic effects of AMPAR antagonists. Indeed, the hippocampi of animal models and the surgically resected hippocampal tissues from drug-resistant TLE patients show upregulated AKT activity (phosphorylation) [46][47][48]. Similar to the present data, furthermore, topiramate (TPM, an AED) modulates AMPAR-mediated AKT/GSK3β/CREB signaling pathway [49]. Therefore, our findings suggest that maladaptive regulation of AKT/GSK3β/CREB pathway may be one of the important causes in pharmacoresistant seizures in response to AMPAR antagonists. PICK1 contributes to a shift in the AMPAR composition to GRIA2-lacking (Ca 2+permeable) by facilitating GRIA2 internalization [21,23,24]. In the present study, GRIA2 and PICK1 expression levels in the epileptic hippocampus were lower than those in the control (normal) hippocampus. These findings are consistent with a previous study demonstrating downregulation of GRIA2 and PICK1 expression in kainic acid-induced chronic epilepsy rats [17]. The present study also demonstrates that both AMPAR antagonists effectively reduced GRIA2 Y phosphorylations in responders, but not in non-responders, without affecting its S880 phosphorylation. Furthermore, GRIA2 Y phosphorylations showed direct proportional relationships with seizure parameters. Phosphorylations regulate surface GRIA2 expression through interactions with GRIP1 and PICK1: PKC-mediated S880 phosphorylation reinforces its internalization by inhibiting GRIP1-GRIA2 bindings [15,21]. In contrast, Src family-mediated Y phosphorylations facilitate PICK1-GRIA2 interaction to promote GRIA2 internalization [23,24]. Consistent with our previous studies [19,50], both AMPAR antagonists reduced Src family, but not PKC, activity. Considering the roles of GRIA2 phosphorylations in its interactions with GRIP1/PICK1 [15,21,23,24], it is plausible that both AMPAR antagonists would attenuate seizure activity by inhibiting PICK1-mediated GRIA2 internalization. However, AMPAR antagonists cannot influence surface GRIA2 expression [20]. In addition, the present data show that AMPAR antagonists did not influence the reduced PICK1 expression in the hippocampi of both responders and non-responders, which was unaffected by 3CAI co-treatment in non-responders, although PICK1 is a substrate of AKT/GSK3β pathway [34]. Therefore, it is likely that the reduced GRIA2 Y phosphorylations may reflect the decreased Src family activity induced by AMPAR antagonists rather than the inhibition of GRIA2-lacking AMPAR assembly. Furthermore, it is hypothesized that reduced PICK1 expression in the hippocampus would be involved in ictogenesis in chronic epilepsy rats via other pathways rather than the regulation of surface GRIA2 expression since PICK1 interacts with type III metabotropic glutamate receptor (mGlu7R) in the presynaptic active zone and regulates glutamate release [50]. Indeed, mGlu7R deletion and a polypeptide interfering PICK1 and mGlu7R show typical epileptic seizures [51,52]. Therefore, our findings indicate that PICK1-mediated regulation of surface GRIA2 expression may not be involved in the anti-epileptic effects and responsiveness of AMPAR antagonists, although both perampanel and GYKI 52466 inhibited Src-mediated GRIA2 Y phosphorylations.
On the other hand, SE and spontaneous seizure activity lead to vasogenic edema formation (serum extravasation) induced by brain-blood barrier (BBB) disruption, which deteriorates seizure activity [53][54][55]. During recovery of vasogenic edema, multidrug efflux transporter expressions, such as p-glycoprotein (p-GP), breast cancer resistance protein (BCRP), and multidrug resistance protein-4 (MRP4), are increased in the hippocampus, which reduce AED concentrations in the brain and cause pharmacoresistant epilepsy [53]. Since we did not compare the pharmacokinetics of AMPAR antagonists between responders and non-responders in the present study, the possibility could not be excluded that the decreased responsiveness of AMPAR antagonists in non-responders would be a consequence from the lower concentration of these compounds induced by over-expression or hyperactivation of drug efflux transporters. In addition, AKT is one of the common down-stream molecules during seizure-induced BBB leakage. Indeed, 3CAI effectively attenuates SEinduced vasogenic edema formation, although it does not show anti-epileptic effects [28]. Thus, it is likely that 3CAI would also increase the efficacies of AMPAR antagonist by inhibiting AKT-mediated serum leakage or upregulation of multidrug efflux transporter. On the other hand, CREB activation plays a protective role against BBB disruption [56,57]. Considering the inhibitory effect of 3CAI on AKT/GSK3β/CREB signaling pathway in the present study, however, it is unlikely that 3CAI may be involved in the CREB-mediated regulation of vascular permeability. Further studies are needed to elucidate whether 3CAI enhances the responsiveness to AMPAR antagonists by inhibiting serum extravasation and/or multidrug efflux systems.

Conclusions
The present study revealed that AMPAR antagonists ameliorated spontaneous seizure activity by affecting the Src-mediated AKT/GSK3β/CREB signaling pathway, which was relevant to the regulation of surface expression of GRIA1 rather than GRIA2. In addition, the dysregulation of this pathway was one of the causes of refractory seizures to AMPAR antagonists. Therefore, our findings suggest that the Src/AKT/GSK3β/CREB pathway may be one of the potential therapeutic targets for the treatment of intractable TLE.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/biomedicines9040425/s1, Figure S1: Representative full-gel images of Western blots in Figure  3A, Figure S2: Representative full-gel images of Western blots in Figure 4A, Figure S3: Representative full-gel images of Western blots in Figure 5A, Figure S4: Representative full-gel images of Western blots in Figure 6A, Figure S5: Representative full-gel images of Western blots in Figure 6B, Figure  S6: Representative full-gel images of Western blots in Figure 7A, Figure S7: Representative full-gel images of Western blots in Figure 9A, Figure S8: Representative full-gel images of Western blots in Figure 10A.

Conflicts of Interest:
The authors declared that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.