KCC2 downregulation facilitates epileptic seizures

GABAA receptor-mediated inhibition depends on the maintenance of low level intracellular [Cl−] concentration, which in adult depends on neuron specific K+-Cl− cotransporter-2 (KCC2). Previous studies have shown that KCC2 was downregulated in both epileptic patients and various epileptic animal models. However, the temporal relationship between KCC2 downregulation and seizure induction is unclear yet. In this study, we explored the temporal relationship and the influence of KCC2 downregulation on seizure induction. Significant downregulation of plasma membrane KCC2 was directly associated with severe (Racine Score III and above) behavioral seizures in vivo, and occurred before epileptiform bursting activities in vitro induced by convulsant. Overexpression of KCC2 using KCC2 plasmid effectively enhanced resistance to convulsant-induced epileptiform bursting activities in vitro. Furthermore, suppression of membrane KCC2 expression, using shRNAKCC2 plasmid in vitro and shRNAKCC2 containing lentivirus in vivo, induced spontaneous epileptiform bursting activities in vitro and Racine III seizure behaviors accompanied by epileptic EEG in vivo. Our findings novelly demonstrated that altered expression of KCC2 is not the consequence of seizure occurrence but likely is the contributing factor.

GABA A receptor is a Cl − permeable ionotropic receptor, which can mediate either excitatory or inhibitory synaptic responses depending on the intracellular Cl − concentration ([Cl − ] i ) 1,2 . The [Cl − ] i is largely determined by the SLC12A cation-chloride cotransporters, including the sodium-potassium-chloride cotransporter 1 (NKCC1) and the neuron-specific potassium-chloride cotransporter 2 (KCC2). In immature neurons, robust expression of NKCC1 is regarded as a major source of Cl − inner transporter to maintain a high [Cl − ] i . Due to developmental upregulation of KCC2 expression but downregulation of NKCC1 in mature neurons, KCC2 is the primary determinant in maintaining [Cl − ] i below electrochemical equilibrium by extruding Cl − 3-5 . Robust KCC2 expression is essential to promote the switch from excitatory to fast inhibitory action of GABA during development 3 . Decrease of KCC2 expression level and subsequent diminish of GABA inhibitory effect has been found in several neurological disorders, including epilepsy [6][7][8][9][10][11][12][13] . Epilepsy is known as the recurrence of seizures, and seizure is defined as abnormal excessive or synchronous neuronal network activities 14 . The decrease of GABA inhibition leading to the imbalance of excitation and inhibition will generate the seizure 15,16 . Subicular pyramidal neurons from patients with refractory temporal lobe epilepsy exhibit positive shift of GABA responses 17 , which are attributable to reduction of KCC2 expression 17,18 . And KCC2 downregulation is also detected in the hippocampus of temporal lobe epilepsy animal models, giving rise to inhibitory efficacy reduction and neuron excitability enhancement 12,13 .
Given the crucial role of KCC2 in maintaining GABA A R inhibitory function, understanding the casual link between KCC2 downregulation and seizure is particularly important. In previous studies KCC2 downregulation was observed after epilepsy. Several researches have found that KCC2 downregulation is triggered by abnormal excitatory activities of neuronal network 19 . Elevated activity induced by glutamate application caused downregulation of KCC2 and this downregulation is dependent on Ca 2+ -permeable NMDA receptor activity 20 . However, it is still unclear whether KCC2 was downregulated before seizure or as a consequence of seizure.
To further figure out the temporal relationship between the downregulation of KCC2 and the seizure, we combined electrophysiology and molecular biology approaches to demonstrate that KCC2 downregulation preceded convulsant induced epileptiform bursting activities both in vitro and in vivo. Intervening endogenous KCC2 expression level in cultured hippocampal neurons directly resulted in increase of spontaneous epileptiform bursting activity, while overexpression of KCC2 increased the resistance to convulsant induced epileptiform bursting activity. Furthermore, seizure-like behaviors and epileptiform EEG have been observed in individual rat, in which  Supplementary Fig. 2. Bar histogram showing quantification of KCC2 expression in different seizure level animal groups. CTZ sub-groups were all normalized to DMSO group. KCC2 was significantly downregulated in Racine III (***P < 0.001) and IV-VII score animals (***P < 0.001).
These in vivo experiment results indicated that the epileptiform bursting activities were directly associated with the higher level seizure behaviors, and positively correlated to a significant downregulation of plasma membrane KCC2 level.
KCC2 reduction occurred before the epileptiform bursting activities. To understand the role of KCC2 in the seizure induction, we further studied the temporal relationship between the epileptiform bursting activities and the downregulation of KCC2 by using in vitro hippocampal slice preparation. Bath application of either CTZ (50 μM) induced a progressive increase of neuronal network activities in dentate gyrus layer, which gradually developed to recurrent interictal-like spike activities and later transformed to highly synchronized ictal-like HFHA bursting discharges ( Fig. 2A). Among 6 slices showed epileptiform bursting firing induced by CTZ, the average latency of the first interictal-like spike and ictal-like bursting activity was 114.7 ± 16.3 min (range: 71 to 180 min) and 154.7 ± 17.4 min (range: 102 to 206 min), respectively (Fig. 2B). In association with this activity increase, we found the immunoreactivity of KCC2 labeling has already been significantly decreased in the dendritic area of the granule cells in DG (63 ± 7%, normalized to DMSO, n = 3; one sample T test, P < 0.05) at 120 min (2 h) after CTZ treatment (Fig. 2C). In order to further reveal the temporal relationship between the epileptiform bursting activities and the downregulation of the membrane KCC2 expression, WB experiments were performed to detect the membrane KCC2 expression change at the time point of 0.5, 1 and 2 hr after CTZ treatment. The results in CTZ model showed that there was a pronounced fall of the plasma membrane KCC2 protein level started between 0.5 to 1 hr time point (55.1 ± 7.2%, n = 4; P < 0.01), through out to the 2 hr time point (50.8 ± 4.2%, n = 4; P < 0.01) (Fig. 2D). In addition, in acute 0-Mg 2+ model, a well acknowledged in vitro epilepsy model, similar epileptiform activities were evoked (Supplementary Fig. 1A) with the average latency of the first interictal-like spike and the ictal-like HFHA bursting discharge was 44.8 ± 12.2 min (range: 21 to 76 min, n = 5) and 145.7 ± 23.0 min (range: 86 to 228 min, n = 5), respectively ( Supplementary Fig. 1B). However, the significant decrease of the plasma membrane KCC2 level was detected as early as at 1 hr time point (68.1 ± 6.2%, n = 4; P < 0.05) and through out to 2 hr time point (56.5 ± 2.9%, n = 3; P < 0.01) of time-matched normal ACSF control ( Supplementary Fig. 1C). Compared to the time point between the bursting activities and the membrane KCC2 downregulation in the same preparation from either CTZ or 0-Mg 2+ model, it was interesting to find that the membrane KCC2 downregulation occurred along with the increased hippocampal activity but much earlier than the happening time point of the epileptiform bursting activities.
Phosphorylation of KCC2 residue S940 plays an important role in maintaining surface KCC2 stability and activity 20 . Our WB results further revealed that the ratio of the membrane pS940 KCC2/total membrane KCC2 protein was significantly declined as early as at 1 hr (55.9 ± 5.8%, n = 3; P < 0.05) and with further reduction (27.2 ± 5.8%, n = 3; P < 0.01) at 2 hr time point, indicating a significant dephosphorylation of S940 occurred as early as at 1 hr after CTZ treatment in hippocampal slices (Fig. 2E). Besides KCC2, the membrane expression of NKCC1 in hippocampus was also examined in the in vitro CTZ model. Our results showed that the expression level of NKCC1 was rather very low, almost undetectable in both normal control and CTZ treatment condition. In this experiment condition, the quantification of the expression level of NKCC1 after CTZ treatment at either 1 h or 2 hr time points showed no significant change (1 h: 109.9 ± 17.4%, n = 3; 2 h: 93.8 ± 4.7%, n = 3) compared to DMSO control treatment (Fig. 2F).
These results showed that the downregulation of the membrane KCC2, as well as the reduction of the ratio of pS940 membrane KCC2, the stable phase of the membrane KCC2, during seizure induction preceded the epileptiform bursting activities, suggesting that it might be the cause for epileptiform bursting activity generation.
Dysfunction of KCC2 occurred before the epileptiform bursting activities. Above results indicated that the significant loss of membrane KCC2 and its S940 residue phosphorylation had occurred between 0.5 to 1 hr and lasted to at least 2 hr after CTZ treatment. To test whether CTZ-induced membrane KCC2 loss also affected its function, the direct measurement of Cl − extrusion capacity and E GABA was performed in CA1 pyramidal neurons in hippocampal slices.
First we tested whether CTZ-induced membrane KCC2 loss directly affected its sustained Cl − regulatory capacity. Repetitive stimulation of Schaffer collateral evoked frequency dependent synaptic depression on recorded GABAergic IPSCs in CA1 pyramidal neurons under gramicidin perforation patch clamp, similar as previous reported 27 . At the holding potential of −80 mV, which is favored for Cl − efflux through GABA A receptors, frequency dependent IPSC depression induced by a train of stimulation (24 pulses at 20 Hz) had no significant difference between DMSO control and 2 hr CTZ treatment groups (n = 8; P > 0.05; Fig. 3A). This result suggested that no matter whether membrane KCC2 was downregulated or not, the inward GABA current was irreverent with KCC2 function. In contrast, when membrane potential was held at −40 mV, large amount of Cl − intrudes when GABA A receptors are activated and KCC2 is required to function to maintain the intracellular low level of Cl − by continuously extruding Cl − . Indeed, in this condition, CTZ treated neurons showed significantly enhanced synaptic depression compared to the DMSO control (n = 8; P < 0.01; Fig. 3B). It indicated that the loss of membrane KCC2 due to convulsant treatment apparently impaired its function to extrude chloride and attenuated the inhibition of GABA A R.
Then, the change of the E GABA , which was determined by the measurement of the reversal potential of evoked IPSCs at the different holding potential in voltage clamp mode of perforation patch clamp, at the different time point after CTZ treatment was measured to study the temporal relation of the KCC2 function impairment. After continuous CTZ treatment for either 0.5 hr or 1 hr, there was a significant positive shift of the E GABA after 1 hr CTZ treatment (CTZ 0 hr, −91.9 ± 2.9 mV vs. CTZ 1 h, −70.5 ± 7.4 mV, n = 5; P < 0.05; Fig. 3D), but no obvious change of the E GABA after only 0.5 hr CTZ treatment (CTZ 0 hr, −83.1 ± 2.8 mV vs. CTZ 0.5 hr, −85.4 ± 4.9 mV, n = 5; P = 0.70; Fig. 3C), which is in consistent with our WB data. In addition, we also detected that CTZ incubation remarkably led to a depolarizing shift of E GABA at 2 hr time point compared to DMSO control (DMSO, −70.8 ± 1.1 mV, n = 9 vs. CTZ, −60.9 ± 1.9 mV, n = 8; P < 0.001; Fig. 3E,F).
Since positive shift of E GABA could be induced by either a decrease of KCC2 or an increase of NKCC1, we performed pharmacological experiments by using either furosemide (100 µM), a KCC2 preferable blocker, or bumetanide (10 µM), a selective NKCC1 inhibitor, to test their contribution to this E GABA change. In accordance with the perforated patch recordings, small tip whole-cell recordings also revealed a significant depolarizing shifted of E GABA after 2 hr CTZ treatment (DMSO, −62.3 ± 0.7 mV, n = 13 vs. CTZ, −56.2 ± 0.9 mV, n = 11; P < 0.001; Fig. 3G,H). According to that, following experiments were conducted under small tip whole-cell recordings. Consistent with the previous reported 27 , when furosemide was added into the ACSF to block KCC2 function, a significant depolarizing shift of the E GABA was detected (FUR, −58.1 ± 1.6 mV, n = 7, P < 0.05; Fig. 3G and H), which mimicked CTZ effect. However, there was no effect on E GABA was detected in inhibiting NKCC1 with bumetanide (BUM, −64.0 ± 1.2 mV, n = 9, P = 0.2; Fig. 3G,H). Thus, this result further indicated that the convulsant stimulation-induced depolarizing shift of E GABA was likely largely dependent on the loss of KCC2 function but not due to the involvement of NKCC1.
In conclusion, our current results indicated that the functional reduction of KCC2 occurred before the formation of highly synchronized bursting discharges, suggesting that the lack of KCC2 might act as an important contributing factor in generating seizure related epileptiform bursting activities.
This result indicated that suppression of endogenous KCC2 expression and impairment of the KCC2 function alone are sufficient to induce neurons to generate spontaneous epileptiform bursting activities.

Knocking down of KCC2 expression level facilitated epileptic seizure in vivo.
Since knocking down of KCC2 expression in cultured hippocampal neurons could facilitate the seizure related epileptiform bursting activities, we further investigated whether downregulation of KCC2 in vivo could be prone to generate spontaneous seizure activity. We used lenti-virus containing with either shKCC2-EGFP or the scrambled shRNA vector with EGFP as control to knock down KCC2 expression in hippocampal dentate gyrus area (DG). Confocal images showed that both shKCC2-EGFP and scram-EGFP virus successfully infected DG cells. Brain sections immunostaining with KCC2 antibody showed GFP positive cells were co-labeled with KCC2 signal in scrambled shRNA control rat, while none or little co-labeling of GFP with KCC2 signal was observed in shKCC2 rat (Fig. 6A). Therefore, data indicated that the membrane KCC2 protein expression was efficiently knocked down by shKCC2-EGFP in DG cells in vivo. In three rats, more than two injection positions, bilaterally, were observed to be successfully infected with shKCC2-EGFP virus. Short lasting abnormal EEG events with corresponding spontaneous Racine III seizure behaviors were caught by EEG/video monitoring. The power analysis of the representative EEG traces revealed that the frequency peak of the ictal like EEG activities was at around 3-5 Hz (Fig. 6B), which was in agree with above result that the mean frequency for Racine III was at 3.6 ± 0.3 Hz in CTZ-induced seizure animals (see Fig. 1F). No abnormal activity was detected in two virus infection control rats, although the recording time was similar. Above results indicated that knocking down endogenous KCC2 alone could induce to generate epileptiform bursting EEG and spontaneous seizure in freely moving rats.

Discussion
Our main findings are: 1) ictal-like epileptiform bursting activities in EEG could be directly correlated to the seizure behavior in freely moving rats in CTZ seizure model, and significant KCC2 downregulation only occurred in Racine score III and above seizure behavior rats with ictal-like epileptiform bursting activities; 2) the downregulation of the membrane KCC2 expression and Cl − extrusion function occurred much earlier than ictal-like epileptiform bursting activities in hippocampal slices; 3) knocking down of KCC2 expression induced the occurrence of spontaneous epileptiform bursting activities in vitro, and seizure behavior with ictal-like epileptiform bursting EEG activities in vivo; and 4) upregulation of KCC2 expression inhibited convulsant induced epileptiform bursting activities in vitro. These findings showed that KCC2 downregulation during epileptogenesis played a facilitation role in generating ictal-like epileptiform bursting activities in neurons, hence to evoke seizure.
Researches have found that KCC2 was downregulated in various seizure animal models. Hippocampal KCC2 immunostaining was reduced 6 hr after kindling-induced seizures 30 and plasma membrane KCC2 in dentate gyrus was significantly downregulated for 1 and 2 weeks after pilocarpine-induced status epilepticus 13 . Consistent with those previous findings in vivo, reduction of hippocampal membrane KCC2 have also been observed in our current study in CTZ-induced seizure model. MRS IV or above behaviors are used to estimate whether acute seizure model is well established. In this study, MRS IV or above seizure behaviors were noticed to correspond with the synchronized HFHA bursting events of EEG by monitoring behavior and EEG recording simultaneously. Interestingly, we found that characteristic epileptiform discharges appeared in MRS III animals as well. Moreover, KCC2 reduction was also detected only in MRS III or above, but not in MRS II or below animals. These results suggested that epileptiform bursting activities were closely linked to seizure behaviors, and could be regarded as an indicator to investigate the temporal relation between seizure behavior and protein (such as KCC2) expression change. In this study, development of neuronal network activities from interictal-like activity to ictal-like bursting discharges could be induced in both Mg 2+ free ACSF and CTZ model in hippocampal slices preparation. Similar to previous studies, interictal-like activity induced by Mg 2+ depletion led to reduced KCC2 expression 4,31 . Previous research has reported that spontaneous interictal-like, but not ictal-like activities, induced downregulation of KCC2 in a subpopulation of subicular principal neurons from adult patients with temporal lobe epilepsy 32 . In agree with the human result, we found that downregulation of membrane KCC2 had occurred (<1 hr) before ictal-like bursting discharge occurrence, but along with the interictal-like spike discharges in the hippocampal neurons. Furthermore, we also found the occurrence of depolarizing shift of E GABA was earlier than the epileptiform bursting discharge generation. Since the change in E GABA is attributed to the altered KCC2-mediated Cl − extrusion capacity, It suggested functional KCC2 reduction was in advance of epileptiform bursting discharges. In addition, NKCC1, another chloride cotransporter mainly expressed in the neuron in neonatal stage, have been reported to be upregulated during seizure. However, our WB blot results from the same sample as detected the KCC2 downregulation showed no change of the NKCC1 expression in the hippocampal neuronal membrane. Taken together, our findings by first time sorted out the temporal relationship between the KCC2 downregulation and the generating of ictal-like epileptiform bursting activities, which indicating that the impairment of KCC2 was not just the consequence of the epileptic seizure, but might be a key factor to facilitate the epileptiform ictal-like bursting activity happening, and hence the seizure behavior.
For the purpose to figure out the functional role of KCC2, increasing researches take advantage of molecular approaches to manipulate endogenous KCC2 expression both in vitro and in vivo. In vitro, silencing of endogenous KCC2 in cultured neurons reduces neurotoxic resistance 33 . In vivo, KCC2 homozygous mutant mice could not survive for long time after birth 34 . Frequent seizure episodes presented in those mice might result in the postnatal morality 35 . Dysfunction of KCC2 by biallelic mutations might induce migrating focal seizures 36 . Besides, although no abnormal phenotype was showed in heterozygote mice, a higher susceptibility for seizure activity was observed in comparison with wild-type ones 35 . In this study, knocking down of the endogenous KCC2 using shRNA KCC2 increased the risk to generate spontaneous bursting activities. Instead of that, overexpression of KCC2 even suppressed the ratio of neurons showing bursting activities induced by convulsant. But, it is not known whether reduction of functional KCC2 would lead to the occurrence of the epileptiform activities or seizure behaviors in animal. To work out the question without affecting neuronal maturation, shKCC2 lentivirus was adopted to knock down KCC2 in dentate gyrus in adult rats. As expected, spontaneous seizure episodes and associated HFHA EEG were detected in shKCC2 group rats with well infected neurons bilaterally in DG area, but not in virus infection control group rats. This result strongly supported our view that hippocampal KCC2 expression downregulation could induce spontaneous seizure occurrence.
Thus, the remaining question was how the KCC2 was downregulated at the time point (<1 hr) earlier than the epileptiform bursting activities occurrence after CTZ treatment in hippocampal slices. The membrane stability and the function of KCC2 is subjected to be regulated by residues phosphorylation, such as serine 940 (S940) and tyrosine (Tyr-1087) [37][38][39] . Since phosphorylation of S940 maintains the stability of KCC2 on the cell surface and the co-transporter activity, most of the studies were focused on S940 residues of KCC2 20,32,37 . Dephosphorylation of S940 was reported to involve in glutamate induced activity dependent downregulation of membrane KCC2 and E GABA positive shifts 20 . Recent research even demonstrated that S940 mutant accelerated the latency and lethality of status seizure induced by KA in vivo 40 . Here, we discovered that there was a significant dephosphorylation of S940 (reduced ratio of membrane pS940 KCC2/total membrane KCC2) occurred as early as at 1 hr after CTZ treatment, along with the downregulation of the membrane KCC2 level, earlier than the first occurrence time point for ictal-like bursting activities of the neurons. Considering the importance of pS940 on surface KCC2 stability and the significant reduction of the pS940 at 1 hr CTZ treatment before epileptiform bursting activities, we hypothesized that the dephosphorylation of the pS940 induced the instability of the membrane KCC2 and caused the downregulation of the membrane KCC2 in the early stage of seizure induction. Thus, maintaining of the stability of the membrane KCC2, such as preserving the pS940 residues of KCC2, likely would play a crucial role to limit the epileptogenesis under pathological condition 40 .
Taking together, this study suggested that the reduction of membrane KCC2 during increased neuronal activity after pathologic, such as convulsant stimulation, condition, further interrupted the imbalance of excitation and inhibition by suppressing the GABA inhibitory function, drove the neuronal network to develop bursting activity, and finally led to the formation of ictal-like epileptiform bursting neuronal activities and seizure. It provides a new avenue for interrupting seizure occurrence by either blocking of KCC2 downregulation or increasing the upregulation during the early phase of pathologic conditions. Thus, protection of KCC2 downregulation could serve as a new target for developing novel anti-epileptic drugs. Plasmid constructs. KCC2 cloned into pIRES2-EGFP and small hairpin RNA plasmid interfering with the expression of KCC2 (shRNA KCC2 ), were designed by Shanghai Genechem Co. Ltd. The target sequences (GCCATTTCCATGAGCGCAA) were inserted into pGCsi-U6-RFP vector to generate the shRNA KCC2 construct and a non-silencing scrambled sequence as a control shRNA vector.