Prolonged seizure activity causes caspase dependent cleavage and dysfunction of G-protein activated inwardly rectifying potassium channels

Recurrent high-frequency epileptic seizures cause progressive hippocampal sclerosis, which is associated with caspase-3 activation and NMDA receptor-dependent excitotoxicity. However, the identity of caspase-3 substrates that contribute to seizure-induced hippocampal atrophy remains largely unknown. Here, we show that prolonged high-frequency epileptiform discharges in cultured hippocampal neurons leads to caspase-dependent cleavage of GIRK1 and GIRK2, the major subunits of neuronal G protein-activated inwardly rectifying potassium (GIRK) channels that mediate membrane hyperpolarization and synaptic inhibition in the brain. We have identified caspase-3 cleavage sites in GIRK1 (387ECLD390) and GIRK2 (349YEVD352). The YEVD motif is highly conserved in GIRK2-4, and located within their C-terminal binding sites for Gβγ proteins that mediate membrane-delimited GIRK activation. Indeed, the cleaved GIRK2 displays reduced binding to Gβγ and cannot coassemble with GIRK1. Loss of an ER export motif upon cleavage of GIRK2 abolishes surface and current expression of GIRK2 homotetramic channels. Lastly, kainate-induced status epilepticus causes GIRK1 and GIRK2 cleavage in the hippocampus in vivo. Our findings are the first to show direct cleavage of GIRK1 and GIRK2 subunits by caspase-3, and suggest the possible role of caspase-3 mediated down-regulation of GIRK channel function and expression in hippocampal neuronal injury during prolonged epileptic seizures.

. Prolonged seizure activity induces C-terminal cleavage of GIRK channels in cultured hippocampal neurons. (a) Whole-cell patch clamp recording of spontaneous action potentials in cultured hippocampal neurons (12-13 DIV, pretreated with 200 μM DL-APV for 2-3 days at 10-11 DIV) before and after APV control (left representative trace) and APV withdrawal (right representative trace). Activation of synaptic NMDAR upon APV withdrawal induced high frequency burst firing of action potentials and sustained depolarization. (b) Surface biotinylation of cultured hippocampal neurons after APV control (ctl) or APV withdrawal (wd) was analyzed by immunoblotting with antibodies recognizing intracellular N-termini of GIRK1 and GIRK2. Prolonged APV withdrawal for 30-90 min resulted in the C-terminal cleavage of GIRK1 and GIRK2 proteins that were biotinylated (Surface) and in the lysates (Total). (c,d) Quantitative immunoblot analyses of total GIRK2 (c) and GIRK1 (d) expression in cultured hippocampal neurons after APV control for 90 min (ctl, n = 4-5) or APV withdrawal (wd) for 30-90 min (n = 3-4 each time point). GAPDH served as a loading control. Data shown represent the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.005. The cropped gray-scale blots are displayed. Full-length blots are included in the Supplementary Fig. S3. (Fig. 3d). Similarly, D347E mutation in 344 YEVD 347 motif of GIRK4 abolished caspase-3 mediated cleavage of GIRK4 (Fig. 3e). Although recent publication has shown that caspases can cleave at glutamate residue with slower kinetics in synthetic peptide substrates 31 , a complete absence of caspase-3 mediated cleavage of GIRK1-D390E, GIRK2-D352E, and GIRK4-D347E in 90 min reaction indicates that 387 ECLD 390 of GIRK1, 349 YEVD 352 in GIRK2, and 344 YEVD 347 in GIRK4 are caspase-3 cleavage motifs.
subunits that contribute to formation of the extended cytoplasmic pore 34 . Specifically, the Gβγ contact site in GIRK2 is formed by secondary β sheet structure elements βK, βL, βM and βN from one subunit and by elements βD and βE from an adjacent subunit 34 . Sequence alignment of the distal C-terminal tails of all GIRK subunits revealed that caspase-3 cleavage site 387 ECLD 390 of GIRK1 is located distal to its Gβγ contact site whereas caspase-3 cleavage sites (YEVD) of GIRK2, GIRK3, and GIRK4 are located at their secondary βM sheet structure elements within the Gβγ contact site (Fig. 4a,b).
Since 349 YEVD 352 motif of GIRK2 falls within and near the Gβγ contact sites (Fig. 4c,d), caspase-3 cleavage of GIRK2 may disrupt GIRK2 interaction with Gβγ. To test this, co-immunoprecipitation of Gβγ was performed with wild type GIRK2A or truncated GIRK2A-Y353X in which Y353 residue right after 349 YEVD 352 motif was mutated to a stop codon. Thus, GIRK2A-Y353X mimics GIRK2A cleaved by caspase-3. We first found that the level of GIRK2A-Y353X proteins was consistently lower compared to wild type GIRK2A proteins although an equal amount of each plasmid was transfected in HEK293T cells (Fig. 5a). To increase the expression level of GIRK2A-Y353X to a similar extent as wild type GIRK2A for coimmunoprecipitation, we doubled the amount of GIRK2A-Y353X plasmid for transfection compared to wild type GIRK2A plasmid (Fig. 5b). Immunoprecipitation of Gβ 1 γ 2 tagged with yellow fluorescent proteins (YFP) resulted in coimmunoprecipitation of wild type GIRK2A (a) Sequence alignment of GIRK subunits with caspase-3 cleavage sites of known substrates that resemble consensus EXXD and XEXD motifs. (b-e) In vitro transcription and translation of GIRK subunits with 35 S-methionine followed by in vitro cleavage assay with purified caspase-1 or caspase-3. The cleavage reaction products were separated by SDS-PAGE gel and visualized by autoradiograph. (b) Caspase-3 but not caspase-1 cleaved wild type (WT) GIRK1. (c) Caspase-3 also cleaved mutant GIRK1-D393E but not mutant GIRK1-D390E, indicating that D390 within 387 ECLD 390 motif is the caspase-3 cleavage site. (d) Caspase-3 but not caspase-1 cleaved GIRK2 WT. Caspase-3 also cleaved mutant GIRK2-D346E but not mutant GIRK2-D352E, indicating that D352 within 349 YEVD 352 motif is the caspase-3 cleavage site. (e) While caspase-1 had no effect, caspase-3 cleaved GIRK4 WT but not mutant GIRK4-D347E, indicating that D347 within 344 YEVD 347 motif is the caspase-3 cleavage site.
We also noticed a significant decrease in GIRK1 expression when this subunit is coexpressed with GIRK2A-Y353X compared to wild type GIRK2A (Fig. 5a). Since GIRK1 heteromerization with GIRK2 stabilizes GIRK1 expression 36,37 , we hypothesized that caspase-3 cleavage of GIRK2 may disrupt GIRK2 interaction with GIRK1. To test this, we first doubled the amount of GIRK2A-Y353X plasmid for transfection compared to wild type GIRK2A plasmids, and performed immunoprecipitation using the same amount of anti-GIRK2 N-terminal antibodies that could immunoprecipitate a fraction but not all of transfected GIRK2A subunits from the lysate. Although the same amount of wild type GIRK2A and GIRK2A-Y353X proteins were immunoprecipitated, GIRK1 co-immunoprecipitated only with wild type GIRK2A but not GIRK2A-Y353X (Fig. 5c), indicating that truncated GIRK2A-Y353X could not coassemble with GIRK1.  WT alone, or GIRK2A-Y353X alone (n = 6-7 each). GAPDH served as a loading control. GIRK2A-Y353X expression was significantly lower than GIRK2A WT expression. Coexpression of GIRK2A-Y353X decreased HA-GIRK1 expression compared to coexpression of GIRK2A WT. (b) Immunoprecipitation (IP) with anti-GFP antibodies was performed from untransfected HEK293T cells (none, n = 4), or the cells transfected with YFP-Gβ 1 and YFP-Gγ 2 alone (n = 4) or together with GIRK2A WT (n = 6) or GIRK2A-Y353X (n = 6). To increase GIRK2A-Y353X expression to a similar extent as GIRK2A-WT expression, we doubled the amount of GIRK2A-Y353X plasmid for transfection compared to GIRK2A-WT plasmids. Coimmunoprecipitation of GIRK2-Y353X with Gβ 1 γ 2 was decreased compared to GIRK2A WT. (c) IP with anti-GIRK2 N-terminal antibodies was performed from untransfected HEK293T cells (none), or the cells expressing HA-GIRK1 alone or together with GIRK2A WT or GIRK2A-Y353X (n = 3 each). To increase GIRK2A-Y353X expression to a similar extent as GIRK2AWT expression, we doubled the amount of GIRK2A-Y353X plasmid for transfection compared to GIRK2A-WT plasmids. Coimmunoprecipitation of HA-GIRK1 with GIRK2A-Y353X was reduced compared to GIRK2A WT. In (b,c), *points at IgG bands. Data shown represent the mean ± SEM (**p < 0.01, ***p < 0.005 against GIRK2A WT; # p < 0.05, ### p < 0.005 against HA-GIRK1 + GIRK2A WT). The cropped gray-scale blots are displayed. Full-length blots are included in the Supplementary Fig. S5.
Scientific RepoRts | 7: 12313 | DOI:10.1038/s41598-017-12508-y GIRK2A truncated at 349 YEVD 352 motif does not express at the plasma membrane. Trafficking of GIRK channels is tightly regulated by multiple amino acid sequence motifs within their subunits that control their forward trafficking from the endoplasmic reticulum (ER) as well as post-ER endocytic trafficking 18,[36][37][38] . GIRK1 subunits are retained in the ER when expressed alone, but GIRK1 assembly with GIRK2 containing ER export motif 396 ELETEEE 403 allows efficient surface expression of heterotetrameric channels 36,37 . The C-terminal tail of GIRK2 distal to its 349 YEVD 352 motif contains this ER export motif (Figs 4b and 6) 36 . To test if caspase-3 cleavage of GIRK2 and subsequent loss of its ER export motif disrupts surface expression of GIRK2 channels or GIRK1/ GIRK2 channels, surface immunostaining was performed in COS7 cells transfected with GIRK1 or GIRK2A, which were tagged with an extracellular hemagglutinin (HA) epitope ( Fig. 6a,b,e). While wild type HA-GIRK2A channels and HA-GIRK1/GIRK2A channels displayed robust surface expression, truncated HA-GIRK2A-Y353X channels or HA-GIRK1/GIRK2A-Y353X channels failed to express on the plasma membrane ( Fig. 6c,d,f-g). Lower total expression of HA-GIRK1 was observed in the cells cotransfected with GIRK2A-Y353X compared to the cells expressing wild type GIRK2A (Fig. 6f), similar to our western blot analyses in HEK293T cells (Fig. 5a).
GIRK2A channels truncated at 349 YEVD 352 motif do not express K + current. To test if truncated GIRK2A-Y353X channels are functional, two-electrode voltage clamp recording was performed to examine macroscopic K + currents of wild type GIRK2A or truncated GIRK2A-Y353X homotetrameric channels from Xenopus oocytes coexpressing Gβ 1 γ 2 subunits (Fig. 7). Wild type GIRK2A channels displayed very low level of basal K + currents in the absence of Gβ 1 γ 2 , but produced significantly larger K + currents upon coexpression of Gβ 1 γ 2 (Fig. 7a,c), consistent with previous studies reporting Gβγ-dependent activation of GIRK channels 35,39 . In contrast, GIRK2A-Y353X channels produced negligible K + currents whether Gβ 1 γ 2 proteins were coexpressed or not (Fig. 7a,c). Because this lack of K + currents (Fig. 7a,c) could be due to very low protein expression of GIRK2A-Y353X compared to wild type GIRK2A (Fig. 7b), we repeated the recording in the oocytes injected with 20 ng of GIRK2A-Y353X cRNA. Although these oocytes expressed increased level of GIRK2A-Y353X especially in the presence of Gβ 1 γ 2 coexpression compared to the oocytes injected with 5 ng of cRNA (Fig. 7b), they failed to produce K + currents in the presence or absence of Gβ 1 γ 2 (Fig. 7a,c). None of the tested oocytes produced Na + currents, indicating that K + selectivity is intact in wild type and truncated GIRK2A channels.

Kainate-induced status epilepticus induces C-terminal cleavage of GIRK channels in the hippocampus.
To test if prolonged epileptic seizures in vivo (i.e. status epilepticus) lead to the C-terminal cleavage of GIRK channels, we used a well-established rodent model of TLE in which status epilepticus was induced by intraperitoneal (i.p.) injection of kainate (Fig. 8a), a potent agonist for ionotropic glutamate receptors 4,40 . We chose the kainate model of TLE due to hippocampus-restricted injuries and histopathological correlates of hippocampal sclerosis 4,40 associated with increased level and activation of caspase-3 41,42 . Furthermore, kainate-induced seizures have been shown to induce expression of an immediate early gene c-Fos which is involved in dampening excitability and promoting survival of hippocampal neurons 43 , as well as another immediate early gene c-Jun which is a primary substrate of c-Jun N-terminal kinases (JNKs) important for excitotoxic neuronal apoptosis in the hippocampus following status epilepticus 44 .
Injection of kainate (9 mg/kg) but not vehicle control (H 2 O) induced behavioral hyperactivity within 20 min and recurrent stage 4-5 seizures (Racine scale) and status epilepticus within 1 hour (h) in Sprague Dawley rats. Kainate-induced status epilepticus significantly increased the level of C-terminally cleaved GIRK2 subunits in rat hippocampi at 8 h post injection (Fig. 8b,c). There was an increasing trend in c-Jun expression at 8 h post kainate injection, but this trend did not reach statistical significance (Fig. 8b,c). In separate experiments, CD001 rat hippocampal lysate was prepared at 3 h post injection with vehicle control (saline) or kainate (20 and 30 mg/ kg) (Fig. 8d,e). There was a low level of cleaved GIRK1 and GIRK2 proteins in the hippocampi of saline-treated CD001 rats (Fig. 8d). Injection of 20 mg/kg kainate caused stage 2-3 seizures which did not last after 1 h post injection, and did not further induce C-terminal cleavage of GIRK1 and GIRK2 at 3 h post injection (Fig. 8d). However, injection of 30 mg/kg kainate induced recurrent stage 4-5 seizures in CD001 rats which significantly increased the levels of c-Fos and C-terminally cleaved GIRK1 proteins but not the levels of C-terminally cleaved GIRK2 proteins within 3 h post injection (Fig. 8d,e). Similar to rats, C57BL/6 J mice injected with 15 mg/kg and 30 mg/kg kainate displayed recurrent stage 3-4 seizures and 4-5 seizures respectively, and displayed enhanced c-Fos and c-Jun expression and increased C-terminal cleavage of GIRK2 ( Supplementary Fig. S8).

Discussion
The intracellular N-and C-terminal tails of GIRK channel subunits provide multi-functional regions that form the cytoplasmic pore of the channels and the interacting domains for Gβγ and other signaling proteins that regulate their function and trafficking 6,45 . Here, we demonstrate a novel mode of GIRK channel regulation: caspase-3 cleavage of neuronal GIRK channel subunits GIRK1 and GIRK2 which occurs during prolonged seizure activity (Figs 1-2). Our surface biotinylation experiment showed that prolonged seizure activity induced by APV withdrawal resulted in caspase-3 dependent C-terminal cleavage of endogenous GIRK1 and GIRK2 subunits at the plasma membrane (Figs 1 and 2). Furthermore, such cleavage of GIRK subunits was also observed in total neuronal lysate which contained both surface and intracellular GIRK subunits (Figs 1 and 2). Since our previous report has shown that 7% total GIRK1 proteins and 19% total GIRK2 proteins were on the cell membrane of cultured hippocampal neurons 18 , cleavage of almost 50% total GIRK1 and GIRK2 proteins (Fig. 2) suggests that intracellular GIRK1 and GIRK2 were also cleaved by caspase-3.
To date, our studies are the first to identify caspase-3 cleavage sites in GIRK subunits (Fig. 3). The "YEVD" caspase-3 cleavage motifs and the surrounding amino acid sequences are highly conserved in the intracellular C-terminal tails of GIRK2-4 (Fig. 4) and located upstream of specific motifs that control their surface expression, intracellular trafficking, and protein-protein interaction 18,36-38 . Indeed, GIRK2A-Y353X, which mimics GIRK2A or biotin-conjugated secondary antibodies followed by Alexa594-conjugated streptavidin (Exp-2, n = 10 cells per transfection). Following fixation and permeabilization, total HA-GIRK1 and GIRK2A proteins were labeled with mouse anti-HA and rabbit anti-GIRK2 N-term antibodies, respectively, followed by Alexa488-and Alexa-680-conjugated secondary antibodies. (f) Representative images showing low level of HA-GIRK1/GIRK2A Y353X channels at the plasma membrane compared to WT channels. Scale bars are 15 μm. (g) Background subtracted, mean fluorescence intensities of surface HA-GIRK1 in untransfected COS7 cells (none) or cells coexpressing GIRK2A WT, or Y353X. Data shown represent the mean ± SEM. **p < 0.01, ***p < 0.005. cleaved by caspase-3, failed to express on the plasma membrane of COS7 cells due to the loss of its ER export motif 396 ELETEEE 403 (Fig. 6). Although we cannot completely exclude the possibility that caspase cleavage at YEVD motif may directly destroy the channel activity, impairment in surface expression most likely underlies the inability of truncated GIRK2A-Y353X channels to produce current in oocytes (Fig. 7, 8f). Similarly, the loss of forward trafficking motif in GIRK4 by caspase-3 cleavage of intracellular GIRK4 (Fig. 4b) is expected to reduce GIRK4 surface expression. Since GIRK3 lacks an ER export motif and therefore requires GIRK2 for its surface expression 8,36 , caspase-3 cleavage of GIRK2 and GIRK3 subunits at the ER is expected to block the ER export of GIRK2/GIRK3 channels.
Interestingly, we observed a significantly lower expression of truncated GIRK2A-Y353X than wild-type GIRK2A (Figs 5a, 6b and 7b), suggesting the degradation of cleaved GIRK2 proteins retained in the ER. Furthermore, a marked decrease in GIRK1 surface expression was also observed in cells expressing GIRK2A-Y353X (Fig. 6) which were unable to interact with GIRK1 (Fig. 5), consistent with the report that surface expression of GIRK1 requires its interaction with GIRK2 8,36 . Reduced total expression of GIRK1 (Fig. 5) and the ladder of cleaved GIRK1 proteins in the immunoblots of cultured hippocampal neurons (Figs 1 and 2) points to the possibility of ubiquitination-mediated degradation of truncated GIRK1. These observations suggest the instability of cleaved channel portions of GIRK1 and GIRK2, although the fate of much shorter C-terminal fragments of GIRK1 and GIRK2 remains unclear.
Importantly, caspase-3 cleavage of GIRK2-4 at "YEVD" motif removes the secondary βM and βN sheet structural elements from the channel contact surface for Gβγ proteins (Fig. 4), which mediate membrane-delimited activation of GIRK channels 45 . Since truncated GIRK2A-Y353X coimmunoprecipiated with Gβγ significantly less than the wild type GIRK2A (Fig. 5), caspase-3 mediated cleavage of GIRK2 is predicted to decrease Gβγ binding to GIRK1/GIRK2 channels at the plasma membrane that constitute most neuronal GIRK channels 7 , as well as GIRK2/GIRK3 or GIRK4 channels that are expressed in a subset of neurons [8][9][10] . Caspase-3 cleavage site in GIRK1 387 ECLD 390 (Fig. 3) is located far distal to the Gβγ contact surface of GIRK1 (Fig. 4), and thus cleavage would have minimal effect on Gβγ binding to GIRK1. However, inability of truncated GIRK2A-Y353X to interact with   (Fig. 5) suggests that caspase-3 cleavage of GIRK2 would cause disassembly of GIRK1/GIRK2 channels, further contributing to their dysfunction (Fig. 8f).
There is emerging evidence for a macromolecular complex on the plasma membrane consisting of GIRK channels, GPCRs, G-protein subunits, and signaling proteins such as regulators of G protein signaling 11,46-48 that support membrane-delimited opening of GIRK channels upon activation of GPCRs 45 . Furthermore, Gβγ binding is shown to strengthen the interaction between GIRK channels and phosphatidylinositol-4,5-bisphosphate (PIP 2 ), an essential co-factor for channel gating 32,49 . We speculate that caspase-3 cleavage of GIRK2-4 on the plasma membrane and subsequent reduction in Gβγ and GIRK1 binding would destabilize this macromolecular complex and PIP 2 interaction, decoupling the activation of GPCRs from the membrane-delimited gating of GIRK channels.
What are the physiologic consequences of caspase-3 mediated cleavage of GIRK subunits? GIRK channels regulate resting membrane potential and excitability in hippocampal neurons by mediating inhibitory effects of GPCRs for neurotransmitters and neuromodulators, including adenosine A1 receptors and GABA B receptors 6 . Our current study has observed that surface expression of GIRK1 and GIRK2 is increased within 15-30 min of seizure activity (Figs 1 and 2). These results are consistent with our previous studies 18,19 , which report that this regulation involves enhanced recycling of GIRK channels to the plasma membrane, and is associated with increased basal GIRK current and GIRK channel activation induced by adenosine A1 receptors but not GABA B receptors 18,19 . Such initial upregulation of GIRK surface density may likely provide homeostatic defense to reduce neuronal excitability against seizure activity.
However, when this seizure activity persisted for >30 min, we discovered a novel finding that surface and total GIRK1 and GIRK2 proteins are cleaved by caspases sensitive to potent caspase-3 inhibitor DEVD-fmk with Ki of 0.2 nM (Fig. 2). Since DEVD-based peptide inhibitor also blocks other apoptosis executioner caspase-7 (Ki = 1.6 nM), as well as upstream apoptosis initiators caspase-8 (Ki = 0.9 nM) and caspase-10 (Ki = 12 nM) 50 , the identity of caspases that cleave GIRK1 and GIRK2 during prolonged seizure activity in hippocampal neurons remains unclear. Nonetheless, we demonstrated that GIRK2A-Y353X, which mimics GIRK2 cleaved at YEVD motif by caspase-3, had decreased binding to Gβγ and GIRK1 and failed to display surface and current expression (Figs 4-7). Based on our findings and the reports of increased spontaneous sporadic and lethal seizures in GIRK2 knock-out mice 16 , we propose that caspase-mediated cleavage and subsequent down-regulation of GIRK channels may decrease their basal and GPCR-activated K + current and disrupt their ability to dampen excitability (Fig. 8f). Future studies are needed to determine the role of caspase-mediated cleavage of GIRK in neuronal excitability.
The possible involvement of apoptosis initiators caspase-8 and 10 that can also cleave at DEVD and XEXD 29,30 motifs is intriguing, since earlier cleavage of GIRK1 and GIRK2 by these caspases could potentially alter neuronal physiology before cell death. Furthermore, in addition to a key role of caspase-3 in terminal apoptosis execution events 51 , recent studies have revealed apoptosis-independent roles of caspase-3 in neurons such as synaptic plasticity 52,53 including NMDAR-dependent long-term depression of excitatory synapses 54,55 . Interestingly, we observed a low level of cleaved GIRK1 and GIRK2 subunits in cultured hippocampal neurons under APV control condition (Figs 1 and 2) and in the hippocampi of vehicle-treated control rats (Fig. 8), which could be caused by caspase-8 and 10 or apoptosis-independent caspase-3 activity. GIRK channels and GABA B receptors are detected in dendritic spines that harbor the majority of excitatory synapses in hippocampal neurons 11 where their activation provides slow inhibitory postsynaptic currents 15 . In addition to postsynaptic roles, activation of GIRK channels by GABA B receptors in presynaptic terminals has been shown to inhibit neurotransmitter release 56,57 . Furthermore, both GIRK channels and A 1 receptors reside on the dendritic spines and shafts 11,15,58,59 where GIRK activation by adenosine attenuates the excitatory postsynaptic potentials 14 and contributes to depotentiation of NMDAR-dependent long-term potentiation 19 . Therefore, it is tempting to speculate that caspase-mediated cleavage of GIRK1 and GIRK2 may disrupt GIRK-mediated synaptic inhibition and/or plasticity before neurons are fully committed to apoptosis.
We have also discovered that kainate-induced status epilepticus in rats leads to C-terminal cleavage of GIRK1 and GIRK2 in their hippocampi (Fig. 8) where GIRK1-3 subunits are expressed with overlapping distribution patterns 60 . Studies in human TLE and kainate rodent models of TLE indicate that excessive glutamate release during status epilepticus causes hippocampal neuronal death by a combination of necrosis and apoptosis depending on the seizure intensity and the cellular energy levels 3,5 . In addition, activation of caspase-2, 3, 6, 7, and 8 has been observed in the hippocampi of human TLE patients and rodent models of TLE where activity of caspase-3 and 7 is highly correlated with hippocampal sclerosis 42 . In cultured hippocampal neurons, apoptosis induced by prolonged high-frequency epileptiform discharges is evident by DNA fragmentation and activation of caspase-3 family proteins 20,21 , and is blocked by pharmacological inhibition of NMDARs 20 . Indeed, one important culprit for glutamate-induced neuronal death has been shown to be massive accumulation of intracellular Ca 2+ upon overstimulation of NMDARs 61 . Though highly speculative, caspase-mediated cleavage and down regulation of neuronal GIRK channels may cause sustained depolarization and overstimulation of NMDARs, creating pathologic positive feedback mechanisms that amplify Ca 2+ overload and sensitize the neurons to Ca 2+ -induced excitotoxic death. This drastic measure could allow the brain to eliminate the neurons that persistently produce epileptiform discharges when initial homeostatic mechanisms fail to dampen their excitability, and may in part the mean ± SEM. *p < 0.05, ***p < 0.005. (f) A working model for the regulation of GIRK channels by seizure activity. Induction of seizure activity increases surface expression and current of GIRK1/GIRK2 channels within 15-20 min by promoting their recycling possibly as a homeostatic defense to heightened excitability 18,19 . In contrast, prolonged seizure activity (>30 min) leads to caspase-3 dependent cleavage of GIRK1/GIRK2 channels in the ER and plasma membrane, which may likely decrease their surface and current expression, heteromerization, and/or interaction with Gβγ. underlie sclerosis at or close to the seizure foci in the hippocampi of human TLE patients 3 . Testing this hypothesis warrants future studies. Lastly, GIRK channels may likely represent one of many caspase-3 targets, and therefore future research shall include a more comprehensive identification of other caspase-3 substrates involved in seizure-induced hippocampal neuronal injury.

Methods
Experimental animals. All procedures involving animals were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Illinois Urbana-Champaign and the University of California San Francisco (UCSF) in accordance with the guidelines of the U.S National Institutes of Health (NIH).
In vitro cleavage reaction. The pcDNA3 plasmids containing wild type or mutant GIRK cDNA (1 μg) were incubated with the TnT coupled Reticulocyte Lysate System (50 μL, Promega) and L-[ 35 S]-Methionine (>1,000 Ci/ mmol at 10 mCi/ml) (Amersham Biosciences) for 90 min at 30 °C. This system utilizes crude reticulocyte lysate which could contain microsomes, and has been widely used to synthesize membrane proteins in vitro in a cell-free system 62 . Translated proteins were subjected to in vitro cleavage reaction at 37 °C for 2 h with 0.2 μg caspase-1 (Calbiochem) or caspase-3 (BD Biosciences) in a buffer containing (in mM): 20 PIPES, 100 NaCl, 10 DTT, 1 EDTA, 0.1% CHAPS, and 10% sucrose. Products were separated by electrophoresis and visualized by exposure to autoradiographic BioMax MS films (Kodak).
Visual molecular dynamics. Crystal structures of GIRK2 alone or GIRK2 complexed with Gβγ (protein data bank ID code 4kfm) 34 were rendered with Visual Molecular Dynamics software 63 to create molecular surface representation of GIRK2 in blue and Gβγ in green.