PDI augments kainic acid-induced seizure activity and neuronal death by inhibiting PP2A-GluA2-PICK1-mediated AMPA receptor internalization in the mouse hippocampus

Protein disulfide isomerase (PDI) is a redox-active enzyme and also serves as a nitric oxide donor causing S-nitrosylation of cysteine residues in various proteins. Although PDI knockdown reduces α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR)-mediated neuronal activity, the underlying mechanisms are largely unknown. In the present study, we found that under physiological condition PDI knockdown increased CaMKII activity (phosphorylation) in the mouse hippocampus. However, PDI siRNA inhibited protein phosphatase (PP) 2A-mediated GluA2 S880 dephosphorylation by increasing PP2A oxidation, independent of S-nitrosylation. PDI siRNA also enhanced glutamate ionotropic receptor AMPA type subunit 1 (GluA1) S831 and GluA2 S880, but not GluA1 S845 and GluA2 Y869/Y873/Y876 phosphorylations, concomitant with the enhanced protein interacting with C kinase 1 (PICK1)-mediated AMPAR internalization. Furthermore, PDI knockdown attenuated seizure activity and neuronal damage in response to kainic acid (a non-desensitizing agonist of AMPAR). Therefore, these findings suggest that PDI may regulate surface AMPAR expression through PP2A-GluA2-PICK1 signaling pathway, and that PDI may be one of the therapeutic targets for epilepsy via AMPAR internalization without altering basal neurotransmission.

surface 7 .Interestingly, PDI inhibition protects neurons from glutamate-induced oxidative cytotoxicity 8 .Similar to DTNB also known as PDI inhibitor 9 , PDI knockdown reduces the amounts of free thiols on NMDAR subunits (GluN1 and GluN2A) and attenuates acute seizure activity in response to pilocarpine and NMDA as well as spontaneous seizures in chronic epilepsy rats independent of S-nitrosylation, while it does not induce alterations in basal neurotransmission (paired-pulse response) and ER stress under physiological condition 6,10 .Thus, we have reported that PDI might be one of the reductive enzymes regulating the redox-mediated NMDAR activity.
On the other hand, activity-dependent alterations in synaptic strength are mediated by α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR) in response to NMDAR activity.Indeed, NMDAR-mediated Ca 2+ entry, and the subsequent activations of Ca 2+ /calmodulin (CaM)-dependent kinase II (CaMKII) and protein kinase C (PKC) increase AMPAR channel conductance by increasing serine (S) 831 phosphorylation of glutamate ionotropic receptor AMPA type subunit 1 (GluA1).Furthermore, NMDAR activation enhances protein kinase A (PKA)-mediated GluA1 S845 phosphorylation, which regulates AMPAR peak response open probability and synaptic trafficking of AMPARs 11,12 .However, NMDAR activation also facilitates AMPAR internalization by enhancing PKC-and CaMKII-mediated GluA2 S880 phosphorylation 13,14 .On note, DTNB-induced NMDAR oxidation cannot affect AMPAR-and γ-aminobutyric acid receptor (GABAR)-mediated synaptic responses, although it attenuates seizure activity 15,16 .Therefore, it is likely that NMDAR redox would not involve the activitydependent regulation of AMPAR functionality.However, we have found that PDI knockdown reduces the amplitude and frequency of neuronal discharges in response to AMPA as well as NMDA without altering GABAergic inhibitions 6,10 .These discrepancies would be consequences from properties of DTNB and PDI siRNA: DTNB oxidizes C residues on proteins exposed to the extracellular milieu, because of poor membrane permeability 17 .Unlike DTNB, PDI siRNA regulates the thiol modification of proteins in ER, cytosol and cell surface 6,10,18,19 .Furthermore, PDI serves as a nitric oxide (NO) donor causing S-nitrosylation of C residues that is a posttranslational protein modification 20,21 .Therefore, it is noteworthy exploring the underlying mechanisms of the blockade of AMPAR-mediated neuronal activity induced by PDI knockdown, which has been largely unknown.
Here, we found that under physiological condition PDI knockdown inhibited PKC, but increased CaMKII activity (phosphorylation) in the mouse hippocampus without affecting PKA activity.However, PDI siRNA enhanced GluA1 S831 and GluA2 S880, but not GluA1 S845 and GluA2 Y869/Y873/Y876 phosphorylations, and diminished surface AMPAR expression.These phenomena were relevant to the inhibition of protein phosphatase (PP) 2A by enhancing disulfide bond formation (oxidation), independent of S-nitrosylation.Furthermore, PDI knockdown attenuated seizure activity and neuronal damage in response to KA, a non-desensitizing agonist of AMPAR 22,23 , by rapidly increasing protein interacting with C kinase 1 (PICK1)-mediated AMPAR internalization.Therefore, these findings indicate that PDI may regulate surface AMPAR expression via PP2A-mediated GluA2 S880 dephosphorylation, and suggest that PDI may be one of the therapeutic targets for epilepsy via AMPAR internalization without altering basal neurotransmission.
PDI knockdown increases GluA2:PICK1 binding and AMPAR internalization.Finally, we evaluated the effect of PDI knockdown on AMPAR internalization in response to KA. KA did not affect surface expressions of GluA1 and GluA2 in control siRNA-treated animals (Fig. 6a-c).As compared to saline, however,    www.nature.com/scientificreports/rectification pattern 23 .Thus, GluA2 subunit dominantly inhibits AMPAR-dependent Ca 2+ permeability as well as channel conductance.Furthermore, GluA2 homotetramer generates steady-state currents when the receptors are desensitized 40 .In the present study, PDI knockdown increased GluA2 S880 phosphorylation concomitant with the reduced surface GluA2 expression.In addition, PDI knockdown enhances GluA1 S831 phosphorylation that increases channel conductance 24,25 .Thus, it is presumable that PDI knockdown would increase the surface expression of Ca 2+ -permeable GluA2-lacing AMPAR or the preponderance of homomeric GluA1 tetramers that would result in much larger single-channel conductance than heteromeric GluA1-GluA2 tetramers through GluA1 S831 phosphorylation 26,27 .However, the present study also shows that PDI siRNA decreased surface GluA1 expression and neuronal activity in response to KA without altering the GluA1/GluA2 ratio.Furthermore, PDI knockdown did not affect GluA1 S845 phosphorylation that drives surface trafficking of GluA1-containing AMPAR and facilitates AMPAR insertion into synaptic membrane to potentiate the peak AMPAR current 11,41,42 .In contrast to S845 phosphorylation, GluA1 S831 phosphorylation is not required for AMPAR trafficking, although it directly enhances the function of AMPAR 26,27 .Given that the GluA1/2 heterotetramer is the most dominant AMPAR subtype in the hippocampal pyramidal cells and > 95% of AMPARs contain the GluA2 subunit under physiological condition 39,43 , our findings indicate that PDI knockdown may diminish neuronal activity in response to KA by reducing surface heteromeric GluA1-GluA2 tetramers through GluA2 S880 phosphorylation without increasing the proportion of homomeric GluR1 tetramers, and that the increased GluA1 S831 phosphorylation may be an adaptive response to the decrease in the total number of AMPAR tetramers.Surface GluA2 expression is reversely regulated by interactions with glutamate receptor interacting protein 1 (GRIP1) and PICK1.GRIP1 increases surface GluA2 expression, while PICK1 diminishes it 44,45 .GluA2 Y869/ Y873/Y876 phosphorylation destabilizes GluA2:GRIP1 interaction and promote its internalization 46,47 .Phosphorylation of GluA2 S880 site also causes dissociation of GRIP1 and subsequently binds to PICK1, then leads to the endocytosis of AMPAR 35,48 .When GluA2 S880 site is dephosphorylated by PP2A, PICK1 dissociates from GluA2 and GRIP1 binds to GluA2 36,37,49 .Thus, the phosphorylation of S880 on GluA2 is an essential step for AMPAR internalization, which is inhibited by Y876 phosphorylation 50 .Compatible with these reports, PDI siRNA increased GluA2 S880 phosphorylation and GluA2:PICK1 binding without affecting GluA2 Y869/Y873/ Y876 phosphorylation, which was caused by oxidation-mediated PP2A inhibition under physiological condition, and facilitated GluA2:PICK1 interaction following KA treatment.Regarding that tyrosine phosphorylations of GluA2 is are required for AMPAR internalization 46 and GluA2-PICK1 interaction is not influenced by tyrosine phosphorylation of GluA2 47 , these findings indicate that PDI knockdown may facilitate GluA2:PICK1 binding and AMPAR internalization independent of tyrosine GluA2 phosphorylation.
Both GluA1 S831 and GluA2 S880 sites are substrate of PKC and CaMKII [12][13][14] .NMDAR activation initially increases Ca 2+ -dependent PKC activity followed by persistent CaMKII activation 51 .Furthermore, NMDARstimulated Ca 2+ entry activates CaMKII, and subsequent phosphorylates GluA1 at S831 site to increase channel conductance 12,52 .Since PDI enhances NMDAR-mediated neuronal activity through the sulfhydration (reduction) of cysteine residues on the NMDAR redox site 6,10 , it is plausible that PDI siRNA would result in the decrease in GluA1 S831 and GluA2 S880 phosphorylation, accompanied by the reduced PKC T497 and CaMKII T286 phosphorylations (Fig. 8d).However, the present study shows that PDI knockdown reduced PKC T497 phosphorylation, but increased CaMKII T286 phosphorylation concomitant with the enhanced GluA1 S831 and GluA2 S880 phosphorylation.CaMKII is autophosphorylated at T286 site and converted to a Ca 2+ -independent or autonomous species, when Ca 2+ -dependent PKC activation is inhibited 53 .Therefore, our findings suggest that PDI knockdown may lead to CaMKII T286 autophosphorylation, which would increase GluA1 S831 and GluA2 S880 phosphorylations, independent of NMDAR inhibition.
PP1 and PP2A, but not PP2B, dephosphorylates GluA2 S880 site 30 .PP1 and PP2A are also responsible for CaMKII T286 dephosphorylation.In particular, PP2A is the major phosphatase of Ca 2+ -independent T286-autophosphorylated CaMKII 54 .Of interest, PP2A is susceptible to oxidation with an inhibition of phosphatase activity.Indeed, intermolecular disulfide formation (oxidation) between C266 and C269 in PP2A is an inhibitory redox switch in its phosphatase activity 32,33 .Furthermore, PDI has the reductive activity at the cell surface, although it acts predominantly as an oxidase in the ER [55][56][57] .In the present study, we found that PDI bound to PP2A, but not PP1, and PDI siRNA decreased PDI:PP2A interaction and the amount of free thiols (oxidation) without altering that of SNO-thiols under physiological condition.These findings indicate that PDI downregulation induced by siRNA may decrease the PDI:PP2A binding, which would subsequently diminish the amount of total thiol on PP2A, and suggest that PDI may act as a reductase rather than NO transporter of PP2A.The present data also show that KA increased PDI:PP2A and GluA2:PP2A bindings without altering surface GluA1/2 expression and GluA2:PICK1 interaction in control siRNA-infused animals.Although we could not provide the underlying mechanisms in the present study, it is likely that these phenomena may be maladaptive responses to KA-induced seizures: KA may augment PDI-mediated sulfhydration (reduction) of PP2A and the subsequent PP2A-mediated GluA2 S880 dephosphorylation, which would prolong seizure activity in response to KA.Since GluA2 S880 phosphorylation facilitates AMPAR internalization, in turn, PP2A-mediated GluA2 S880 dephosphorylation may result in the unchanged GluA2 surface expression upon KA treatment by abrogating PICK1-mediated AMPAR internalization.These are compatible with previous studies demonstrating that KA is a non-desensitizing agonist of AMPAR 22,23 .In contrast, PDI siRNA attenuated seizure activity in response to KA and further decreased surface GluA1/2 expression concomitant with augmented GluA2:PICK1 interaction following KA injection as compared to saline treatment, although it did not change GluA2:PP2A binding.Regarding that KA activates PP2A and decreases CaMKII activity and its autophosphorylation in the mouse hippocampus at 2.5-6 h after injection 58,59 , these findings also indicate that PDI may abolish AMPAR internalization by activating PP2A.Given that the clear loss of PP2A activity correlates with the cysteine oxidation 60 , our findings suggest that PDI siRNA may increase GluA1 S831, GluA2A S880 and CaMKII T286 phosphorylations www.nature.com/scientificreports/by inducing oxidation-mediated PP2A inhibition under physiological condition, which would facilitate PICK1mediated AMPAR internalization following KA treatment.On the other hand, KA shows a biphasic effect on presynaptic transmission in the hippocampus and other brain regions.Low KA concentrations (50-100 nM) increase glutamate release, while higher concentrations decrease it 61 .In the amygdala, KA receptor (KAR)-activation induces a depression of NMDA and AMPA-mediated evoked excitatory postsynaptic current (eEPSC) amplitude, which is prevented by PKA, but not PKC, inhibitors 62 .In the hippocampus, KAR activation also negatively modulates glutamate release by convergence with group II metabotropic glutamate receptors (mGluRII, inhibitory autoreceptors in presynaptic region) through PKA 63,64 .In contrast, KAR activation facilitates glutamate release through Ca 2+ -CAM-mediated PKA activation in the hippocampus and thalamocortical synapses 65,66 .Thus, it is presumable that PDI would regulate seizure susceptibility to KA by affecting presynaptic KAR.However, the present study shows that PDI knockdown inhibited PKC, but increased CaMKII activity without affecting PKA activity under physiological condition.Considering that activation of PKA, but not PKC, is required for KAR-mediated regulation of presynaptic glutamate release 62,63,65,66 , our findings suggest that PDI may not influence presynaptic glutamate release in response to KA.
In conclusion, the present data indicate that PDI played a crucial role in PP2A activation by reducing disulfide bonds, which abolished AMPAR internalization by dephosphorylating GluA2 S880 and CaMKII T286 sites and abrogating GluA2:PICK1 interaction (Fig. 8d).Therefore, our findings provide a novel mechanism to modify AMPAR-mediated responses under physiological and pathological conditions.

Methods
Ethics and guidelines.All experimental animal protocols were approved by the Animal Care and Use Committee of Hallym University (#Hallym 2021-30, approval date: May 17, 2021).This study was also carried out in compliance with the ARRIVE guidelines.
Experimental animals and chemicals.Male C57BL/6J mice (8 weeks old) were used in the present study.Animals 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).All reagents were obtained from Sigma-Aldrich (USA), except as noted.
Seizure induction and electroencephalogram recording.Three days after surgery, mice were given KA (25 mg/kg, i.p.) or an equal volume of normal saline instead of KA.Saline-treated mice were used as controls.Diazepam (Valium; Roche, France; 10 mg/kg, i.p.) was administered 2 h after KA injection.Electrode-implanted animals were also given KA (25 mg/kg, i.p.) after baseline recording for at least 30 min.Electroencephalographic (EEG) signals were recorded with a DAM 80 differential amplifier (0.1-1000 Hz bandpass; World Precision Instruments, USA) and digitized (1000 Hz) using LabChart Pro v7 software (AD Instruments, Australia).EEG total power was obtained from integration of EEG amplitude in the frequency bands from 0 to 80 Hz.Frequency-power spectral temporal maps were generated with filtrations of the frequency domain (0-80 Hz) and the amplitude domain (0-50 mV).Total power was measured during the 2 h recording session and spectrograms were automatically calculated using a Hanning sliding window with 50% overlap by LabChart Pro v7 software (AD Instruments, Australia).Total EEG power was normalized by the baseline power obtained from each animal.Latency of seizure on-set was defined as the time point showing more than 3 s and consisting of a rhythmic discharge between 4 and 10 Hz with amplitude of at least two times higher than the baseline EEG.After recording, animals were used for biochemical experiments 6,10,18,19,21 .
Co-immunoprecipitation and membrane fraction.The tissues were lysed in radioimmune precipitation buffer (RIPA) with protease and phosphatase inhibitor cocktails (Roche Applied Sciences) and 1 mM sodium orthovanadate.After quantification of total protein concentration, equal amounts of protein (amount in 150 μg) were precipitated with the PDI, PP1, PP2A or PICK1 antibody and subsequently incubated with protein

Figure 2 .
Figure 2. Effects of PDI knockdown on GluA2 phosphorylation and surface AMPAR expression in the hippocampus under physiological condition.As compared to control siRNA, PDI siRNA enhances GluA2 S880, but not Y869/873/876, phosphorylation in the mouse hippocampus.GluA2 S880 upregulations are observed in the dendrites and cell bodies of hippocampal neurons.However, PDI siRNA reduces surface expression of GluA1 and GluA2 without affecting surface GluA1/GluA2 ratio.(a) Representative Western blot images for GluA2, p-GluA2 Y869/873/876 and p-GluA2 S880 levels.(b-d) Quantitative analyses of the effects of PDI siRNA on GluA2 (b), p-GluA2 Y869/873/876 (c) and p-GluA2 S880 levels (d) based on the Western blot data (*p < 0.05 vs. control siRNA; n = 7, respectively; Mann-Whitney test).(e) Representative double immunostaining for NeuN and GluA2 S880 in CA1 and CA3 pyramidal neurons.(f) Representative Western blot images for surface expressions of GluA1 and GluA2.(g-i) Quantitative analyses of the effects of PDI siRNA on surface GluA1 (g) and GluA2 (h) levels and surface GluA1/GluA2 ratio (i) based on the Western blot data (*p < 0.05 vs. control siRNA; n = 7, respectively; Mann-Whitney test).

Figure 3 .
Figure 3. Effects of PDI knockdown on PKC, CaMKII and PKA phosphorylations in the hippocampus under physiological condition.As compared to control siRNA, PDI siRNA reduces PKC T497 phosphorylation, while it enhances CaMKII T286 phosphorylation.PDI knockdown does not affect PKA T197 phosphorylation in the mouse hippocampus.(a) Representative Western blot images for PKC, p-PKC T497, CaMKII, p-CaMKII T286, PKA and p-PKA T197 levels.(b-d) Quantitative analyses of the effects of PDI siRNA on p-PKC T497 (b), p-CaMKII T286 (c) and p-PKA T197 levels (d) based on the Western blot data (*p < 0.05 vs. control siRNA; n = 7, respectively; Mann-Whitney test).

Figure 4 .Figure 5 .
Figure 4. Effects of PDI knockdown on PDI bindings to PP1 and PP2A, and the thiolization and S-nitrosylation and on PP2A in the hippocampus under physiological condition.As compared to control siRNA, PDI siRNA reduces PDI:PP2A bindings without affecting PDI:PP1 bindings.In addition, PDI knockdown reduced the amount of total thiols on PP2A, while it does not affect that of SNO-thiols in the mouse hippocampus.(a) Representative Western blot images for the PDI:PP1 bindings.(b-c) Quantitative analyses of the effects of PDI siRNA on PP1 level (b) and PDI:PP1 binding (c) based on the Western blot data (n = 7, respectively).(d) Representative Western blot images for the PDI:PP2A bindings.(e-f) Quantitative analyses of the effects of PDI siRNA on PP2A level (e) and PDI:PP2A binding (f) based on the Western blot data (*p < 0.05 vs. control siRNA; n = 7, respectively; Mann-Whitney test).(g) Representative Western blot images for the thiolization and S-nitrosylation on PP2A.(h-i) Quantitative analyses of the effects of PDI siRNA on thiolization (h) and S-nitrosylation on PP2A (i) based on the Western blot data (*p < 0.05 vs. control siRNA; n = 7, respectively; Mann-Whitney test).

Figure 6 .
Figure 6.Effects of PDI knockdown on surface AMPAR expression in the hippocampus following KA injection.Control siRNA does not affect surface AMPAR expression in both saline-and KA-treated animals.PDI knockdown decreases surface expression of GluA1 and GluA2 without altering surface GluA1/GluA2 ratio in both saline-and KA-treated animals.PDI knockdown reduces surface AMPAR expression in KA-treated animals more than saline-treated animals.(a) Representative Western blot images for surface expressions of GluA1 and GluA2.(b-d) Quantitative analyses of the effects of PDI siRNA on surface GluA1 (b) and GluA2 (c) levels and surface GluA1/GluA2 ratio (d) following KA injection (*, # p < 0.05 vs. control siRNA vs. saline; n = 7, respectively; Kruskal-Wallis test followed by Tukey post-hoc test).

Figure 8 .
Figure 8. Effects of PDI knockdown on PICK1 bindings to GluA2 in the hippocampus following KA injection.PDI siRNA does not influence PICK1 level in both saline-and KA-treated groups.As compared to control siRNA, PDI siRNA increases GluA2A:PICK1 binding in saline-treated group.Although KA does not affect GluA2:PICK1 binding in control siRNA-infused group, it increases it in PDI siRNA-infused group.(a) Representative Western blot images for the GluA2:PICK1 binding.(b-c) Quantitative analyses of the effects of PDI siRNA on PICK1 level (b) and GluA2:PICK1 binding (c) following KA injection (*, # p < 0.05 vs. control siRNA vs. saline; n = 7, respectively; Kruskal-Wallis test followed by Tukey post-hoc test).(d) Scheme of the role of PDI in AMPAR internalization.PDI may reduce lead to reduction-induced PP2A activation, which would abolish PICK1-mediated AMPAR internalization by dephosphorylating GluA2 S880 and CaMKII T286 sites. https://doi.org/10.1038/s41598-023-41014-7