Main

Neurodegeneration restricts neuron numbers during development but can become a serious issue in disease conditions such as temporal lobe epilepsy (TLE).1 GABA-activated Cl− channels contribute to activity-dependent refinement of neural networks by triggering the so-called giant depolarizing potentials providing developing neurons with a sense of activity essential for neuronal survival and co-regulation of excitatory glutamatergic and (inhibitory) GABAergic synapses.2 By regulating transmembrane Cl− gradients KCC2 plays a vital role in development and disease.3 In addition, KCC2 plays a protein structural role in spine formation through its C-terminal protein domain (CTD).4, 5 Hence, regulation of KCC2 expression and function is relevant for development and disease-specific plasticity of neural networks.6, 7, 8, 9

GlyR α3K RNA editing leads to proline-to-leucine substitution (P185L) in the ligand-binding domain and generates gain-of-function neurotransmitter receptors.10, 11, 12, 13 GlyR RNA editing is upregulated in the hippocampus of patients with TLE and leads to GlyR α3K185L-dependent tonic inhibition of neuronal excitability associated with neurodegeneration.14 KCC2 expression promotes neuroprotection14, 15 but whether this involves regulation of transmembrane Cl− gradient or protein structural role is a matter of debate.14, 15

Here, we assessed neuroprotection through several KCC2 variants in two different models of neurodegeneration including chronic neuronal silencing (α3K185L model) and acute neuronal overexcitation (NMDA model).14, 15 The results identify a threonine- and tyrosine-phosphorylation-resistant KCC2 variant with increased Cl− transport activity, but they also demonstrate that the N-terminal KCC2 protein domain (NTD) is sufficient for neuroprotection.

Results

We first investigated the mechanisms underlying the neurotoxic effects of the RNA-edited GlyR α3K variant (α3K185L) on primary hippocampal neurons as described.14 Neurons were transfected at day in vitro (DIV) 6 either with non-edited (185P; control) or RNA-edited GlyR α3K (185L) and maintained for 3 days in 10 μM glycine, a concentration that should selectively activate GlyR α3K185L.10 As shown previously,14 ~50% of the GlyR α3K185L-expressing neurons exhibited fragmented dendrites and pyknotic nuclei (Figures 1a and c; Supplementary Table 1), both clear indicators of neurodegeneration, whereas overexpression of non-edited GlyR α3K185P in 10 μM glycine had little effect on neuronal survival (Figures 1b and c). Appearance of fragmented dendrites was not due to rapid internalization of surface-labeled GlyR α3K185L because comparable fluorescent signals were obtained upon cell permeabilization (not shown). To further investigate the possibility that GlyR activation is responsible for neurodegeneration, GlyR α3K185P-expressing neurons were maintained under GlyR-activating conditions (400 μM glycine), and GlyR α3K185L-expressing neurons under non-activating (0 μM glycine) conditions.10 In agreement with different apparent glycine affinities of non-edited GlyR α3K185P and edited GlyR α3K185L10 the data in Figure 1c clearly show that glycine-dependent activation of GlyR α3K triggers neurodegeneration.

Figure 1
figure 1

GlyR activation-dependent neurodegeneration and KCC2-dependent neuroprotection. Images of neurons with surface-stained GlyR α3K185L (a) and GlyR α3K185P (b). According to fragmented dendrites and pyknotic nuclei, neurodegeneration is obvious in neurons with continuous GlyR activation (a). (c) Quantification of the fraction of surviving neurons in each condition reveals that neurodegeneration occurs only under GlyR α3K-activating conditions and that co-expression of KCC2 (either splice variant KCC2a or KCC2b) protects neurons with continuous GlyR α3K185L activation. Numbers in the bar graphs indicate the number of cultures analyzed. For KCC2 protein structure see Supplementary Figure 1. For values, see Supplementary Table 1. ***P <0.001

Full size image

Expression of KCC2 is inversely correlated to GlyR RNA-editing in TLE hippocampi.14 Two KCC2 RNA splice variants with different N-termini are known16 (Supplementary Figure 1A), but neuroprotective effects were demonstrated for KCC2b only.14, 15 We therefore tested whether KCC2a and KCC2b protect neurons against GlyR α3K185L-mediated degeneration (Figure 1c). KCC2a- and KCC2b-positive neurons were identified with 2A self-processing peptide-coupled EGFP (Supplementary Figure 1B). Co-expression of KCC2a and KCC2b rescued the survival of neurons cultured under GlyR α3K185L-activating conditions (Figure 1c; Supplementary Table 1). This result confirms our previous data14 and furthermore reveals that KCC2-dependent neuroprotection does not depend on alternative splicing of exon 1.16 KCC2b was henceforth used and referred to as ‘KCC2wt’.

Role of spontaneous neural network activity for GlyR-dependent neurodegeneration

Spontaneous neuronal activity at the beginning of experimental GlyR α3K185L expression period is mainly dependent on GABAAR activation (Supplementary Figure 2), suggesting that inability of chronically silent neurons with enduring GlyR α3K185L activation14 to participate in spontaneous neural network activity underlies neurodegeneration. To address this possibility, we blocked either synaptic or synaptic and non-synaptic GABAARs, as described,17 using 0.2 μM or 3 μM of the competitive GABAAR antagonist GABAzine. However, neither 0.2 μM nor 3 μM GABAzine rescued chronically silent neurons, and KCC2wt-dependent neuroprotection did not require GABAAR activation (Figure 2a; Supplementary Table 1). Furthermore, neurodegeneration/neuroprotection did not require action-potential-dependent synaptic transmission because tetrodotoxin (TTX) had no effect on survival (Figure 2b; Supplementary Table 1). These results suggest that intrinsic cellular mechanisms are responsible for neurodegeneration rather than inability of chronically silent neurons to participate in spontaneous neural network activity.

Figure 2
figure 2

Spontaneous neuronal network activity does not play a role in GlyR-dependent neurodegeneration. Quantification of the effects of GABAzine (a) and TTX (1 μM, b) on the survival of neurons with activated GlyR α3K185L. Two different GABAzine concentrations (0.2 and 3 μM) were used to block synaptic or synaptic and non-synaptic GABAAR activation. Numbers in the bar graphs indicate the number of cultures analyzed. For values, see Supplementary Table 1. Statistical significance is indicated with ***P<0.001

Full size image

Role of intracellular Ca2+ concentration for GlyR-dependent neurodegeneration

Activation of non-synaptic NMDA receptors (NMDARs) and Ca2+-activated Cl− channel Clca1 were implicated in neurodegeneration,18 and changes in resting intracellular Ca2+ concentration ([Ca2+]i) are involved in excitotoxicity.19 Thus, GlyR α3K185L-dependent neuronal intrinsic mechanisms of neurodegeneration may involve changes in intracellular Ca2+ homeostasis. To investigate this possibility, [Ca2+]i of neurons expressing GlyR α3K185L or GlyR α3K185P in 10 μM glycine (i.e., receptor-activating and non-activating conditions) were determined using fura-2, as described by Jung et al.20 (Figure 3a). However, [Ca2+]i was comparable in all conditions as [Ca2+]i in neurons with activated GlyR α3K185L (78.9±19.2 nM, n=13) were similar to those in neurons with non-activated GlyR α3K185P (69.4±11.1 nM, n=10; P=0.7125) or untransfected neurons (80.6±6.2 nM, n=194, P=0.9280; Figure 3b). These results rule out increased [Ca2+]i as a cause of GlyR α3K185L activation-dependent neurodegeneration.

Figure 3
figure 3

Continuous GlyR α3K185L activation does not affect resting [Ca2+]i. Intracellular Ca2+ concentration at rest ([Ca2+]i) was determined according to the procedure described by Jung et al.20 (a) Neurons were loaded 2–3 days after transfection with fura-2. Red and gray traces correspond to signals obtained from a GlyR α3K185L-positive neuron (red) and non-transfected neurons (gray) in the neighborhood of the transfected neuron in the same viewfield. A solution with 50 mM KCl was applied to monitor the viability of the cells according to their response with regard to fura-2 signals. Cells that did not respond to 50 mM KCl with changes in the F340/F380 ratio were not included in the determination of resting [Ca2+]i. Resting [Ca2+]i was determined within the 10 s (marked with a yellow bar) prior to the application of 50 mM KCl. To obtain minimal (0 mM) and maximal (10 mM) Ca2+ signals for calibration, cells were permeabilized with either 10 μM ionomycin or 10 μM 4-bromo-antibiotic A23187. 2 mM Mn2+ were applied to quench the signal at the end of each experiment and to obtain background fluorescence that was subtracted from all F340 and F380 values. (b) GlyR α3K185L expression and activation in the presence of 10 μM glycine does not affect resting [Ca2+]i. Numbers in the bar graphs indicate the number of neurons analyzed

Full size image

Enduring GlyR α3K185L activation persistently changes neuronal intrinsic membrane properties

We next investigated neuronal intrinsic membrane properties including membrane resistance (RN) and the reversal potential of all membrane conductances, henceforth termed ‘membrane reversal potential’ (Vrev), a value of transmembrane voltage at which ionic diffusive and electrical forces counterbalance and no net transmembrane current is observed. We used gramicidin-perforated patch clamp because under this condition the intracellular milieu is not dependent on the recording pipette solution but on ion channels open at a given time of recording. We used voltage-clamp configuration and voltage pulses (−5 mV) starting from the holding potential (−50 mV) to determine RN. Voltage ramps (from −100 to −30 mV, 0.5 mV/ms; 140 ms total duration) were used to determine Vrev as exactly as possible according to reversal of current–voltage (IV) relationships. To determine Vrev of GABAAR-dependent currents (EGABA) IV curves measured in the presence and absence of GABA were subtracted. The fluorescent dye lucifer yellow and high Cl− concentration in the pipette solution allowed monitoring stability of the perforated patch configuration (Supplementary Figure 3A). Neurons with GlyR α3K185L expression were identified according to co-expression of mCherry fluorescent protein (Supplementary Figure 3A).

We first analyzed IV relationships in the absence of GABA, that is, under basal conditions (Figure 4a, ‘IVbas’), in control neurons (mCherry only) and in neurons with activated GlyR α3K185L and determined Vrev. In neurons with enduring GlyR α3K185L activation, Vrev was significantly more depolarized than in control neurons (Figures 4a and c; −52.1±1.1 mV, n=12 versus −63.6±2.7 mV, n=12; P=0.0005). Furthermore, RN of neurons with enduring GlyR α3K185L activation was significantly decreased compared to control neurons (Figure 4d; 108±31 MΩ, n=12 versus 429±65 MΩ, n=12; P=0.0012), indicating presence of open GlyR α3K185L channels in the plasma membranes of neurons exposed to 10 μM glycine.

Figure 4
figure 4

Intrinsic membrane properties of neurons with enduring GlyR α3K185L activation. (a) Primary hippocampal neurons were either transfected with mCherry alone or with mCherry and GlyR α3K185L (185L) and cultured for 2–3 days under receptor-activating conditions (10 μM glycine). Neurons were investigated by perforated patch clamp (gramicidin) and voltage clamped at −50 mV. Under these conditions, the application of 100 μM GABA lead to hyperpolarizing currents in most of the control neurons (a, left). GABA-evoked currents at the holding potential of −50 mV in GlyR α3K185L-expressing neurons were strongly diminished (a, right). Current–voltage (IV) relationships were obtained by applying voltage ramps from −100 to −30 mV every 5 s. Note that current traces during the voltage ramps are not shown. The time points for determination of IVGABA and IVbas are indicated. (b) The IVs of GABA-evoked currents were determined by subtracting the IVs in the presence and absence of GABA. Traces from individual cells are shown. The yellow-boxed high-power view between the two IV plots shows magnified IV curves of the depicted areas from recordings of the control cell left hand (black trace) and the GlyR α3K185L-positive cell right hand (red trace) in the presence of 10 μM glycine. (c–e) The perforated patch-clamp experiments revealed that RN, Vrev and EGABA were decreased in neurons with continuous GlyR α3K185L activation. Note that washout of glycine from the extracellular solution only partially rescued RN, Vrev and EGABA. Numbers in the bar graphs indicate the number of neurons analyzed. Statistical significance is indicated with *P<0.05, **P<0.01 and ***P<0.001

Full size image

We next determined EGABA by subtracting IV curves measured in the presence (IVGABA) and absence (IVbas) of 100 μM GABA (Figures 4a, b and e21). EGABA values should be a measure of Vrev of Cl− currents because GABAARs are like GlyRs permeable for Cl−. GABAARs are also permeable for HCO3− (PHCO3/PCl∼0.5–0.622, 23), but as our solutions were not based on CO2/HCO3−, the determined EGABA values primarily reflect Vrev of Cl−, that is ECl. Most of the control neurons displayed hyperpolarizing currents at −50 mV in 100 μM GABA (Figure 4a, left hand), which reversed at −58.7±2.3 mV (n=12; Figures 4b). In contrast, in neurons with enduring GlyR α3K185L activation, GABA responses at the holding potential (−50 mV) were shunted (Figure 4a right hand; Supplementary Figure 3B right hand), and IVGABA reversed close to that holding potential (−52.5 ± 1.1 mV, n=12, P=0.0195 compared to control, Figure 4b right hand, Figure 4e). Consequently, the calculated driving force of GABA responses was 4.9±2.9 mV in control neurons (n=12; Figure 5d), whereas it was close to 0 mV (−0.4±0.9 mV, n=12; Figure 5e) in neurons with enduring GlyR α3K185L activation. Together with the strongly decreased RN (Figure 4d), we conclude that enduring GlyR α3K185L activation leads to large permanent Cl− conductance that shunts GABA-dependent currents due to adaptation of transmembrane Cl− gradient according to imposed membrane potential.

Figure 5
figure 5

KCC2 does not rescue neuronal intrinsic membrane properties of neurons with enduring GlyR α3K185L activation. Primary hippocampal neurons were either co-transfected with GlyR α3K185L and mCherry, KCC2wt-2A-mCherry or KCC2pr-2A-mCherry (Supplementary Figure 1B) and cultured for 2–3 days in the presence of 10 μM glycine (Gly). Hatched bars correspond to conditions with KCC2 co-expression, and for the purpose of direct comparison red and faded red bars shown in Figures 4c–e are shown here again. RN (a), Vrev (b) and EGABA (c) were determined with gramicidin-perforated patch-clamp experiments in the voltage-clamp mode. Note that co-expression of KCC2wt or KCC2pr had no significant effect on any of the investigated parameters that reflect neuronal intrinsic membrane properties. (d and e) Calculated driving forces for GABAAR responses. (d) Summary of the driving forces for GABAAR responses of cultured hippocampal neurons transfected with either mCherry, KCC2wt-2A-mCherry or KCC2pr-2A-mCherry, calculated by subtraction of Vrev from EGABA values. See Supplementary Figure 1 for constructs and Supplementary Figure 4 for sample traces from KCC2-2A-mCherry-expressing neurons. Compared to mCherry-transfected neurons, KCC2wt expression had no significant effect on the GABA driving forces of DIV 8–10 hippocampal neurons. In contrast, overexpression of KCC2pr significantly shifted the driving forces for GABAAR responses toward more negative values. The change in the driving forces is due to a negative shift of EGABA (Supplementary Figure 4), not of Vrev, indicating that the intracellular Cl− concentration was changed by KCC2pr, not by KCC2wt, after 2–4 days of overexpression. Note that the calculated GABAAR driving forces were not dependent on the extracellular glycine concentration (0 or 10 μM) during the perforated patch-clamp experiments. (e) Same as d, except that GlyR α3K185L was co-expressed with mCherry, KCC2wt-2A-mCherry or KCC2pr-2A-mCherry. Owing to the large GlyR α3K185L-dependent Cl− shunt all calculated driving forces were close to 0 mV. Numbers in the bar graphs indicate the number of neurons analyzed. Statistical significance is indicated with *P<0.05, **P<0.01 and ***P<0.001

Full size image

We next asked whether acute glycine washout during recording of neurons with activated GlyR α3K185L would have effects on Vrev and EGABA (Figures 4c–e). However, Vrev and EGABA remained at more depolarized potentials compared to control neurons (−49.2±2.4 mV for Vrev, −50.8±1.5 mV for EGABA, n=11), and RN recovered only partially (Figure 4d; 107.8±30.5 MΩ, n=12 in 10 μM glycine versus 203.9±47.2 MΩ, n=11 after glycine washout; P=0.0006). As GABAAR-dependent currents were observed upon glycine washout (Supplementary Figure 3B, right hand) GABAAR downregulation as a reason for observed effects regarding EGABA can be ruled out. Thus, enduring GlyR α3K185L activation during 3 days induced long-lasting changes in neuronal intrinsic membrane properties, which might involve changes in permeability for different ions including, besides Cl−, also K+, a well known major driving force of resting membrane potential generation.

Enduring GlyR α3K185L activation persistently changes resting membrane potential

We finally investigated resting membrane potential (Vm) in current-clamp configuration (Supplementary Figures 3C and D). We clamped current at 0 pA and determined apparent Vm of neurons with continuous GlyR α3K185L activation in 10 μM glycine. Actually, Vm of GlyR α3K185L-expressing neurons in 10 μM glycine required 2–3 min to stabilize after switching from voltage- to current-clamp configuration, and Vm slowly depolarized under these conditions from values around −50 mV (i.e., the formerly imposed holding potential) to values determined in current-clamp configuration reflecting apparent Vm (Supplementary Figure 3C). These slow adaptations of membrane potential consistently occurred only in 10 μM glycine and in GlyR α3K185L-expressing neurons indicating that outwardly directed Cl− currents through GlyR α3K185L slowly decrease [Cl−]i and thereby depolarize Vm. Hence, these experiments revealed that Vm was significantly more depolarized in neurons with activated GlyR α3K185L compared to control neurons (Supplementary Figure 3D; −35.9±2.6 mV, n=5 versus −51.2±2.4 mV, n=9; P=0.0104). As acute glycine washout significantly influenced Vm (Supplementary Figure 3D; −35.9±2.6 mV, n=5 in 10 μM glycine versus −44.8±2.2 mV, n=10 upon glycine washout; P=0.0026), while no differences were observed in control neurons (Supplementary Figure 3D; −52.6±2.4 mV, n=8 in 10 μM glycine versus −51.2±2.4 mV, n=7 upon glycine washout; P=0.6953), these data suggest that enduring GlyR α3K185L activation shifts Vm toward ECl. After glycine washout, Vm remained more depolarized in GlyR α3K185L-expressing neurons compared to control neurons (P=0.0306), and hence as discussed above, these results consolidate conclusion that enduring GlyR α3K185L activation induced long-lasting changes of neuronal intrinsic membrane properties. A priori, these results suggest that KCC2 expression may prevent persistent changes in intrinsic membrane properties and possibly involve Cl− transport in neuroprotection, as suggested earlier.14, 15

KCC2 Cl− transport activity is not relevant for neuroprotection

Gramicidin-perforated patch clamp was used again to analyze the effects of co-expression of KCC2wt on membrane properties of neurons with enduring GlyR α3K185L activation. KCC2-positive neurons were identified according to 2A-self-cleaving peptide-coupled mCherry (Supplementary Figure 1B). KCC2wt was not able to substantially change neuronal intrinsic membrane properties of GlyR α3K185L-expressing neurons (Figure 5). Similar to mCherry/GlyR α3K185L-expressing neurons in 10 μM glycine (Figure 4), RN remained significantly decreased (77.6±12.8 MΩ, n=9, Figure 5a), and Vrev and EGABA also shifted toward imposed holding potential of −50 mV when KCC2wt was co-expressed (−53.4±1.4 mV and −53.7±2.5 mV, respectively, n=8, Figures 5b and c; calculated driving forces for GABAAR responses, Figure 5e). This suggests that KCC2wt Cl− transport activity was not sufficient to overcome GlyR α3K185L-mediated Cl− conductance. Moreover, upon acute glycine washout, Vrev and EGABA remained largely unchanged (Figures 5b and c), and RN recovered only partially (Figure 5a), similar to neurons with enduring GlyR α3K185L activation in the absence of KCC2wt (Figure 4d). Thus, co-expression of KCC2wt was not able to prevent persistent changes in intrinsic membrane properties of neurons with enduring GlyR α3K185L activation. For control purpose, we checked KCC2 Cl− transport functionality and recorded KCC2wt-positive neurons in the absence of GlyR α3K185L expression (Supplementary Figure 4). Surprisingly, no significant differences in EGABA were found between control neurons (EGABA: −58.7±2.3 mV, n=12) and those expressing KCC2wt (EGABA: −55.4±2.8 mV, n=7, P=0.3920; Supplementary Figure 4B and Figure 5d). However, a mutant variant of KCC2 (‘KCC2pr’; Supplementary Figure 1A), which should be resistant to phosphorylation-dependent downregulation of Cl− transport activity24, 25 shifted EGABA in neurons without GlyR α3K185L co-expression (KCC2pr: −65.8±1.5 mV, n=8; P=0.0325 versus control: −58.7±2.3 mV, n=12; P=0.0049 versus KCC2wt: −55.4±2.8 mV, n=7; Supplementary Figures 4A and B; Figure 5d). These results were confirmed using recording of intracellular Ca2+ dynamics with Oregon Green 488 (Life Technologies, Darmstadt, Germany) in response to GABA application in more immature hippocampal neurons at DIV 6–7 (Supplementary Figure 5). Again, KCC2wt was not able to prevent GABA-induced increases in intracellular Ca2+ signals in hippocampal neurons without GlyR α3K185L co-expression (FGABA/FKCl: 0.21±0.04, n=29 for KCC2wt versus 0.23±0.02, n=114 for untransfected neurons, P=0.7666; Supplementary Figure 5C), whereas the KCC2pr mutant was effective (FGABA/FKCl: 0.12±0.03, n=29 for KCC2pr versus 0.22±0.02, n=110 for untransfected neurons, P=0.0107; Supplementary Figure 5C). Importantly, KCC2wt and KCC2pr Cl− transport activities were apparent in primary cortical neurons under these conditions (FGABA/FKCl: 0.11±0.03, n=36 for KCC2wt versus 0.40±0.03, n=158 for untransfected neurons, P<0.001; FGABA/FKCl: 0.08±0.03, n=44 for KCC2pr versus 0.50±0.03, n=188 for untransfected neurons, P<0.001; Supplementary Figure 5D). Hence, KCC2wt is a functional Cl− transporter, and cortical and hippocampal neurons reveal phosphorylation-dependent differences in the regulation of KCC2 Cl− transport activity. However, because KCC2wt protected hippocampal neurons against GlyR α3K185L activation-dependent neurodegeneration (Figure 1), these results also suggest that Cl− extrusion is not relevant to KCC2-dependent neuroprotection. Indeed, block of Cl− import activity through NKCC1 using bumetanide (10 μM) also failed to rescue neuronal survival (Figure 6a).

Figure 6
figure 6

Protein structural role of KCC2 in neuroprotection. (a) The antagonist bumetanide (10 μM) of the Cl− transporter NKCC1 has no effect on neuronal survival irrespectively of whether hippocampal neurons were challenged with tonic GlyR α3K185L activation or not. (b) The Cl− transport-deficient KCC2wt/pr-C568A as well as the C-terminally truncated KCC2-ΔCTD were neuroprotective, whereas KCC2 constructs lacking the N-terminal domain (NTD) or containing only the isolated KCC2-CTD were not neuroprotective. Note that the isolated KCC2-NTD was sufficient to mediate full rescue of neuronal survival. Numbers in the bar graphs indicate the number of cultures analyzed. For values, see Supplementary Table 1. Statistical significance is indicated with ***P<0.001

Full size image

The KCC2-NTD mediates neuroprotection in the GlyR α3K185L model of neurodegeneration

KCC2 was previously shown to have a protein structural role for synaptogenesis by interaction with cytoskeleton-associated protein 4.1N.5, 26 This finding encouraged us to clarify if cytoskeletal signaling is involved in neuroprotection, but KCC2-C568A mutant which is unable to interact with protein 4.1N and Cl− transport-deficient5, 26 rescued survival of neurons with continuous GlyR α3K185L activation as well as KCC2wt (Figure 6b; Supplementary Table 1). Furthermore, KCC2pr-C568A rescued neuronal survival indicating that KCC2-dependent neuroprotection is independent of phosphorylation of Y903, T906, T1007 and Y1087 in the KCC2-CTD (Supplementary Figure 1A). These results reveal that KCC2-dependent neuroprotection is independent of protein 4.1N-dependent cytoskeletal signaling and further strengthen our conclusion that KCC2-dependent neuroprotection does not rely on Cl− transport. They also suggest that it is not the KCC2-CTD, which mediates neuroprotection. Indeed, deletion of NTD, not of CTD, abolished KCC2-dependent neuroprotection (Figure 6b; Supplementary Table 1). Reciprocally, co-expression of the KCC2-NTD, not of KCC2-CTD, rescued survival of neurons with enduring GlyR α3K185L activation (Figure 6b). That KCC2-ΔNTD or KCC2-CTD failed to protect neurons was not due to poor protein expression (Supplementary Figure 6). These results identify KCC2-NTD as relevant neuroprotective signaling domain.

KCC2-NTD mediates neuroprotection in the NMDA-dependent model of excitotoxicity

To sustain the finding that KCC2-NTD plays a pivotal role in neuroprotection, we investigated whether neuroprotective function of KCC2-NTD holds in another model of neurodegeneration. In contrast to our GlyR α3K185L-dependent model of neurodegeneration, which uses chronic silencing of neuronal activity, the NMDA-dependent excitotoxicity model relies on neuronal overexcitation. Actually, KCC2 Cl− transport activity was recently postulated to mediate neuroprotection in the NMDA model of neurodegeneration.15 Therefore, we tested KCC2wt and mutant KCC2 variants coupled to EGFP via 2A peptides (Supplementary Figure 1) in the NMDA-dependent model of neurodegeneration.15 Neurons were transfected at DIV 6 with non-edited GlyR α3K185P (to assess neuronal morphology in addition to appearance of pyknotic nuclei for quantification of neurodegeneration, Figures 7a and b), kept in culture for two days (in non-GlyR-activating conditions), and then incubated for 30 min in 40 μM NMDA in the absence or presence (10 μM) of glycine before maintaining the culture for 24 h as described by Pellegrino et al.15 Glycine is a co-agonist of NMDARs and therefore is expected to enhance the effect of NMDA with regard to neurodegeneration, but NMDAR internalization27 may interfere with this process. However, NMDA effects on neuronal survival in the absence and presence (10 μM) of glycine were comparable, though slightly more pronounced in the presence of glycine (Figures 7c and d; Supplementary Table 1). Co-expression of KCC2wt and KCC2pr protected neurons against NMDA-induced excitotoxicity (Figure 7c), and the Cl− transport-deficient KCC2wt/pr-C568A mutants also succeeded (Figure 7c; Supplementary Table 1). Most importantly, KCC2-ΔNTD and KCC2-CTD failed to protect neurons against NMDA-induced excitotoxicity, whereas KCC2-ΔCTD and KCC2-NTD rescued neuronal survival (Figures 7c and d), similar to our GlyR model of neurodegeneration. Thus, in contrast to previous findings,15 our results identify a common mechanism of KCC2-dependent neuroprotection in different models of neurodegeneration, which is independent of Cl− transport activity but involves the KCC2-NTD.

Figure 7
figure 7

Protein structural aspects of KCC2-dependent neuroprotection in the NMDA model of neurodegeneration. (a and b) Images of control neurons (a) and neurons treated for 30 min with 40 μM NMDA 24 h prior to fixation (b) are shown. GlyR α3K185P was expressed under non-receptor-activating conditions and surface stained in order to assess neurodegeneration in the same way as in the GlyR-dependent model of neurodegeneration (Figure 1). Degenerated neurons could be identified by their fragmented dendrites and pyknotic nuclei (b). (c) Quantification of the fraction of surviving neurons revealed that co-expression of KCC2wt or KCC2pr also protects NMDA-treated neurons. Again, Cl− transport-deficient KCC2wt/pr-C568A also rescued neuronal survival. N-terminally truncated KCC2-ΔNTD or the isolated KCC2-CTD was not able to mediate neuroprotection. Note that the isolated KCC2-NTD was sufficient to mediate full rescue of neuronal survival. (d) Quantification of the effects of glycine (10 μM) on NMDA-dependent excitotoxicity and KCC2-dependent neuroprotection. Numbers in the bar graphs indicate the number of cultures analyzed. For values, see Supplementary Table 1. Statistical significance is indicated with ***P<0.001

Full size image

Discussion

We elucidated the mechanisms of KCC2-dependent neuroprotection and identify a protein structural role of KCC2-NTD in neuroprotection in the glycine-dependent model of chronic neuronal silencing14 and the NMDA-dependent model of excitotoxicity.15

Mechanisms of neurodegeneration

Our study ruled out the possibility that inability of chronically silent neurons to become spontaneously activated and participate in spontaneous neural activity in the developing network is a reason for GlyR α3K185L-dependent neurodegeneration because block of GABAAR activation, which provides a major driving force for spontaneous neuronal activity at this developmental stage in the culture dish (Supplementary Figure 2) or action potential-dependent synaptic transmission did not influence survival of neurons with enduring GlyR α3K185L activation. Rather, all results synergistically point to enduring GlyR α3K185L activation dependent, long-lasting changes of neuronal intrinsic physiological properties involving decreased RN, Vrev and Vm as reasons for GlyR α3K185L-dependent neurodegeneration. Actually, plasma membrane perforation and resulting impairment of intracellular ion homeostasis were recently associated with the pathophysiology of Alzheimer's disease28 but GlyR α3K185L-dependent neurodegeneration could not be associated with increased intracellular resting [Ca2+] as it may be the case in patients with Alzheimer’s disease.28 This is possibly due to inactivation of, or changes in, expression of voltage-gated Ca2+ channels as a consequence of decreased Vm in neurons with enduring GlyR α3K185L activation.

In the hippocampus, RNA-edited GlyR α3K185L contributes to tonic inhibition of cells with low GlyR β subunit protein expression.14, 29, 30, 31, 32, 33, 34 RNA splicing generates the long variant GlyR α3L185L, which is preponderantly expressed in the hippocampus and operates at presynaptic sites by contributing to regulation of neurotransmitter release.12, 29, 35, 36 However, as the ratio between α3K and α3L is increased in TLE patients with hippocampal sclerosis (i.e., neurodegeneration),29 our results regarding mechanisms of GlyR α3K185L-dependent neurodegeneration and KCC2-mediated neuroprotection are relevant to the understanding of the pathophysiology of TLE. RNA splicing and its regulatory impact on subcellular distribution of ligand-gated GlyRs and GABAARs indeed recently emerged as a critical determinant of neuronal dysfunction in TLE,29, 37, 38, 39 and non-synaptic Cl− channels have also been implicated by other groups in neuronal cell death. A recent study even provides a link between non-synaptic NMDAR activation and excitotoxicity through Ca2+-activated Cl− channel Clca1.18 With regard to neurodegeneration, alteration of non-synaptic Cl− channel function thus represents an important determinant of cellular programs that elicit cell death, and for this reason, care should be taken when considering the use of glycine for neuroprotective purposes.40, 41, 42

Identification of the neuroprotective capacity of the KCC2-NTD in different models of neurodegeneration

KCC2 is a developmentally regulated gene product, which can undergo functional downregulation in the diseased brain.7, 8, 9, 43, 44 Functional downregulation involves phosphorylation of amino acids in the KCC2-CTD.24, 25 In our study, KCC2wt-dependent Cl− extrusion was apparent only in cortical neurons (Supplementary Figure 5), which demonstrates functionality of KCC2wt-dependent Cl− extrusion in general, but also identifies neuron type-specific (cortical versus hippocampal neurons) differences in the apparent efficacy of Cl− extrusion through KCC2wt. The discrepancy to other studies in which KCC2wt overexpression was reported to significantly shift EGABA to hyperpolarized potentials in cultured hippocampal neurons (e.g., Li et al.5and Chudotvorova et al.45) is probably due to differences in the cell culture preparation (e.g., cell density, duration and strength of KCC2 expression, time point of investigation and percentage of GABAergic interneurons in the culture dish). However, hippocampal neurons in our culture preparations develop normally as they are excitable and spontaneously active due to depolarizing GABAAR signaling (Supplementary Figure 2). Thus, the discrepancy to other studies might involve culture-specific changes in the phosphorylation of the S940 site, which was shown to enhance Cl− extrusion capacity of KCC2wt.46, 47 Nevertheless, as Cl− extrusion through the KCC2pr variant (which cannot be phosphorylated at Y903, T906, T1007 and Y1087) was apparent in our hippocampal neuron preparations (Supplementary Figures 4 and 5, and Figure 5d), our results clearly identify a role for these sites in phosphorylation-dependent regulation of KCC2 Cl− extrusion, which is in agreement with mounting evidence for the role of threonine phosphorylation in the downregulation of KCC2-dependent Cl− transport (for review see Kahle et al.48). Our results furthermore make a clarifying contribution to the controversial discussion of the role of tyrosine phosphorylation in the regulation of KCC2 Cl− extrusion capacity48 as they show that KCC2pr with unphosphorylatable threonine and tyrosine residues has an increased Cl− extrusion capacity. However, more detailed study is necessary to fully clarify the role of threonine and tyrosine phosphorylation.

Beyond its ‘classical’ function as Cl− transporter that contributes to the developmental switch of GABA action from de- to hyperpolarization,3 KCC2 fulfills a protein structural function, which contributes to co-regulation of glutamatergic and GABAergic synapses during synaptogenesis in development.4, 5 The neuroprotective effect of KCC2 was demonstrated already in 2008,14 but whether KCC2-dependent neuroprotection involves a protein structural function or regulation of intracellular [Cl−] is a matter of debate.11, 14, 15 Although our initial study with GlyR-dependent neuronal hypoactivity-induced neurodegeneration14 did not address this question, a follow-up study showed that KCC2-dependent neuroprotection depends on its Cl− transport activity in the NMDA-dependent hyperexcitation model of neurodegeneration.15

By challenging primary hippocampal neurons with these two different experimental neurotoxic strategies including enduring GlyR α3K185L activation associated with chronic inhibition of neuronal activity as it may occur in the hippocampus of TLE patients14 and NMDA-dependent excitotoxicity,15 our study provides compelling evidence for a protein structural role of KCC2 in neuroprotection. This conclusion is based on the fact that both KCC2wt (which did not effectively show Cl− extrusion in primary hippocampal neurons; Supplementary Figures 4 and 5, and Figures 5d and e) and Cl− transport-deficient KCC2-C568A rescued neuronal survival in both neurotoxic conditions. Most strikingly, the KCC2-NTD was neuroprotective in both experimental models of neurodegeneration. All these results provide evidence for a protein structural neuroprotective role of KCC2 in neuroprotection, and consistently, our study identifies the KCC2-NTD as the relevant KCC2 protein domain sufficient for neuroprotection. In contrast to the rather large (299 amino acid encompassing) KCC2-CTD with its well-established protein structural role in synapse formation,4, 5 ectopic expression of the relatively compact, 100 amino acid spanning, KCC2-NTD (Supplementary Figure 1A) is a novel suitable but non-exclusive strategy to prevent neurodegeneration in TLE and other neurodegenerative diseases as it would work independently of the KCC2-CTD (involved in regulation of synapse formation) or full-length KCC2-dependent regulation of chloride transport. Thus, KCC2 is a moonlighting protein49 as it is not only a protein with Cl− transport activity but also harbors several protein domains with synaptogenic and neuroprotective activities (Figure 8).

Figure 8
figure 8

Different functional roles of KCC2. The scheme summarizes previous results and findings presented here. (a) Full-length KCC2 is a Cl− transporter that can be positively (S940; green) or negatively (Y903, T906, T1007 and Y1087; red, this study) regulated by phosphorylation. (b) The synaptogenic function of the KCC2-CTD (blue) is established and described elsewhere.5, 26 (c) Neuroprotective function of the isolated KCC2 N-terminal domain (KCC2-NTD, this study)

Full size image

Materials and Methods

Molecular cloning

A cDNA coding for hKCC2b wildtype (designated ‘wt’) was kindly provided by the laboratory of Kai Kaila (Helsinki, Finland) and recently used in one of our studies.14 The KCC2a cDNA was cloned from human postmortem hippocampus RNA (pool of 20 healthy Caucasians, Clontech (Palo Alto, CA, USA). By using the 2A-self-processing peptide EGRGSLLTCGDVEENPGP derived from Thosea asigna,50 we generated constructs for expression of KCC2 and mCherry or EGFP from polycistronic mRNA (Supplementary Figure 1). To generate the phosphorylation-resistant variant KCC2pr, which cannot be functionally downregulated, we substituted by site-directed mutagenesis (GeneEditor, Promega, Mannheim, Germany) alanine and phenylalanine for threonine and tyrosine, respectively, at the relevant positions in the CTD;24, 25 see Supplementary Figure 1 for sequence details. The amino acid substitution C568A was generated using QuickChange Lightning site-directed mutagenesis kit (Agilent Technologies, Waldbronn, Germany). Truncations were generated using the full-length KCC2 constructs and appropriate primer sets. All the KCC2 constructs for neuronal expression are equipped with a cytomegalovirus (CMV) enhancer-human synapsin-1 promoter (hSyn1). All the constructs contain the woodchuck posttranscriptional regulatory element (WPRE), which was derived from a lentiviral vector (Clontech). All expression constructs were verified with DNA sequencing. Molecular cloning of GlyR α3K185P and GlyR α3K185L constructs is described in former publications of our laboratory.10

Cell culture and transfection

All animals were killed according to the permit given by the Office for Health Protection and Technical Safety of the regional government of Berlin (LaGeSo, 0122/07) and in compliance with regulations laid down in the European Community Council Directive. Hippocampal cells were isolated from E19 Wistar rats and kept in B27-supplemented Neurobasal medium (Life Technologies) in 24-well plates as described.14 Transfection was carried out on DIV 6 using 300 ng DNA per well in combination with Effectene transfection reagent (Qiagen, Hilden, Germany), following the manufacturer's protocol as described.14 Either low (185P) or high (185L) affinity receptor types of the short (K) splice variant of GlyR α310 were expressed for the duration of 3 days in the presence or absence of added glycine (0, 10 or 400 μM). For this purpose, a glycine-free minimal essential medium was used.14 In some experiments, neurons were exposed to NMDA (40 μM) for 30 min in the absence or presence of 10 μM glycine and analyzed 24 h after application. For co-transfection, 60 ng of KCC2-coding plasmids were mixed with 240 ng GlyR α3K185L. For Oregon Green imaging of GABA-elicited Ca2+ signals, neurons were investigated at an earlier time point in cell culture, at DIV 6–7.

Chemicals

HEPES (4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid) and all inorganic salts as well as glycine (G7126), GABA (A2129), NMDA (M3262), gramicidin (G5002), lucifer yellow (L0144) and bicuculline methiodide (14343) were purchased from Sigma-Aldrich (Steinheim, Germany). GABAzine (1262) came from Tocris (Bristol, UK), TTX (6973.1) from Carl Roth (Karlsruhe, Germany), 4-bromo-A-23187 (BML-CA101-0001) and ionomycin (LKT-I5753-M001) from ENZO Life Sciences (Lörrach, Germany). Fura-2/AM and Oregon Green 488/AM were purchased from Life Technologies. Stock solutions were made with H2O in the case of TTX (1 mM), glycine and GABA (1 M each), and with DMSO (dimethyl sulfoxide) (SERVA Electrophoresis GmbH, Heidelberg, Germany) in the case of gramicidin, 4-bromo-A-23187 and ionomycin (10 mM each).

Immunochemistry

A polyclonal chicken antibody was used to visualize HA-tagged surface GlyR (1 : 200, Bethyl Laboratories, Montgomery, TX, USA). The secondary antibody was made in donkey, conjugated with indocarbocyanine (Cy3) and purchased from Jackson ImmunoResearch Laboratories (Suffolk, UK). Surface staining of HA-tagged GlyR α3K was performed for 5 min at 37 °C in cell culture medium, as described.14, 29 Prior to fixation with paraformaldehyde (PFA) cells were washed three times with cell culture medium. For fixation with PFA, cells were incubated for 15 min at room temperature (RT) with ice-cold PBS containing 4% PFA and 4% sucrose, followed by three wash steps in PBS. Incubation of fixed cells in freshly made 50 mM NH4Cl for 15 min at RT was used to quench free aldehyde groups from the PFA fixation. Cells were then again washed three times in PBS and blocked with PBS/gelatine (0.1%) prior to permeabilization with 0.12% Triton X-100(Sigma-Aldrich) in PBS/gelatine for 4 min at RT. Before incubation with first antibodies, coverslips were again washed with PBS/gelatine. In some experiments, KCC2 was stained for 1 h at RT using a rabbit polyclonal antibody (#07-432; Merck Millipore, Germany) diluted 1 : 200 in PBS/gelatine. Prior to incubation with secondary antibodies cells were washed three times with PBS/gelatine. For KCC2 stainings, a secondary antibody made in donkey and conjugated with indocarbocyanine (Cy5, Jackson ImmunoResearch Laboratories) was used. After 45 min of incubation at RT cells were again washed three times with PBS/gelatine followed by additional two wash steps in PBS. Stained cell preparations were finally mounted on microscope slides using DAPI-containing vectashield medium (Vector Laboratories, Peterborough, UK).

Electrophysiology

An EPC-7 amplifier and Patchmaster software (HEKA, Lambrecht, Germany) were used for patch-clamp recordings. Patch pipettes, made from borosilicate glass (Science Products, Hofheim, Germany), had resistances of 2–6 MΩ when filled with the intracellular solution containing (in mM) KCl (130), NaCl (5), CaCl2 (0.5), MgCl2 (1), EGTA (5) and HEPES (30). Application of substances was gravity driven. The tip (360 μM) of a perfusion pencil (AutoMate Scientific, Berkeley, CA, USA) was placed close (ca 100 μM) to the recorded neuron to ensure relatively rapid application of substances. Under these conditions, the wash-in duration of lucifer yellow-containing test solution was <500 ms. For the perforation of the patch, the pipette solution contained additionally 50–100 μM gramicidin and 100 μM lucifer yellow, which allowed monitoring the stability of the perforated patch (Supplementary Figure 3A). In addition, a strong shift of the baseline current due to the high Cl− concentration in the pipette solution indicated membrane rupture upon transition to whole-cell configuration. In this case, the recording was stopped. The standard extracellular solution (E1; pH 7.4) contained (in mM) NaCl (140), KCl (5), MgCl2 (1), CaCl2 (2), HEPES-NaOH (10) and glucose (10). In the voltage-clamp mode, neurons were clamped at a potential of −50 mV. IV relationships were obtained from voltage ramps from −100 to −30 mV with a duration of 140 ms applied every 5 s. Series and input resistances were checked throughout the whole duration of each experiment by applying −5 mV pulses prior to the voltage ramps. All data were acquired with a sampling rate of 10 kHz after filtering at 2.8 kHz. All experiments were performed at RT (20–25 °C).

Calcium imaging

Prior to ratiometric and non-ratiometric Ca2+ imaging experiments cells were loaded with fura-2/AM or Oregon Green 488/AM, respectively, by incubating the cells in E1 buffer (see above) supplemented with 1–5 μM of fura-2 or Oregon Green for 20 min at 37 °C. Subsequently, cells were incubated for further 20 min in E1 to ensure deesterification. Glass coverslips with the dye-loaded neurons were placed into a recording chamber (ca 1 ml volume) on the stage of an Axiovert 10 or an Axio Lab.A1 micoroscope (both Zeiss, Oberkochen, Germany). Cells were submerged with a constant flow of E1 through an infusion pipette, which was placed in close vicinity (ca 200 μm) to the recorded cells to ensure short wash-in/washout durations. Transfected cells were identified by the fluorescence of mCherry, which served as control or was co-expressed using the 2A-self-cleaving peptide in KCC2 constructs (see Supplementary Figure 1B for constructs). Ratiometric measurements were performed with the Polychrome V, a Clara Interline CCD camera (Andor Technology, Belfast, UK) and Live Acquisition software (Till Photonics, Martinsried, Germany) using 340 and 380-nm excitation wavelengths. Excitation and emission light were separated by a 510-nm dichroic mirror. The emitted light was filtered using a 530-nm longpass filter. Exposure times were 20 ms (340 nm) and 5 ms (380 nm), and the rate for [Ca2+] measurement was set to one pair of images per 1 sec. 50 mM KCl were applied to check the viability of the cells. To obtain minimal (0 mM) and maximal (10 mM) Ca2+ signals for calibration,20 cells were permeabilized with either 10 μM ionomycin or 10 μM 4-bromo-antibiotic A23187. 2 mM Mn2+ were applied at the end of each experiment to quench the signal and thus to obtain background fluorescence that was subtracted from all F340 and F380 values used for [Ca2+]i calculation according to:

where KD is the dissociation constant of the fura-2-Ca2+complex (225 nM51). F380,min and F380,max are the fluorescence values of free (0 mM, min) and Ca2+-bound (10 mM, max) fura-2. R, Rmin, Rmax are the fluorescence ratios (F340/F380) at the beginning of the experiment (R) and after permeabilization in 0 mM (Rmin) and 10 mM (Rmax) Ca2+, respectively. Non-ratiometric experiments were performed with an HBO-100 lamp (OSRAM GmbH, Munich, Germany), an electronic shutter, a SPOT pursuit camera and Metamorph software (Visitron Systems, Puchheim, Germany). Oregon Green fluorescence was excited and detected using an appropriate filter set (XF22, Omega Opticals, Olching, Germany) using 100 ms shutter open times at a frequency of three per second. Oregon Green fluorescence signals are expressed as FGABA/FKCl, where FKCl and FGABA are the fluorescence values in the presence of E1 supplemented with 50 mM KCl (90 mM instead of 140 mM NaCl) and 100 μM GABA, respectively.

Data analysis and statistics

All numerical data are reported as mean±S.E.M. Statistical analysis (ANOVA and post hoc Tukey's test) was performed using the software IGOR Pro 6.3 (WaveMetrics, Lake Oswego, OR, USA). Significance levels are indicated as *P<0.05, **P<0.01 and ***P<0.001.

Quantitative data of neuronal survival is presented as values that are normalized to the number of vital neurons in control conditions. Degenerated and vital neurons were counted from at least three independent hippocampal cell cultures. The number of experiments is indicated in brackets in the bar graphs. The conditions GlyR α3K185P and α3K185L in the presence of 10 μM glycine were included in each experiment for normalization purpose.

Vrev was calculated by determining the voltage at which no net current was observed (zero point of the basal IV curve in the absence of GABA, IVbas). The values presented in figures and text are not corrected for the liquid junction potential (+3.75 mV). EGABA was obtained by determining the zero point of the GABA IV relationship that was calculated by subtraction of IVbas from IVGABA (Figure 4). Analysis was performed by a homemade procedure written using IGOR Pro 6.3 (WaveMetrics).