G-Protein-Coupled Inwardly Rectifying Potassium (GIRK) Channel Activation by the p75 Neurotrophin Receptor Is Required for Amyloid β Toxicity

Alzheimer's disease is characterized by cognitive decline, neuronal degeneration, and the accumulation of amyloid-beta (Aβ). Although, the neurotoxic Aβ peptide is widely believed to trigger neuronal dysfunction and degeneration in Alzheimer's disease, the mechanism by which this occurs is poorly defined. Here we describe a novel, Aβ-triggered apoptotic pathway in which Aβ treatment leads to the upregulation of G-protein activated inwardly rectifying potassium (GIRK/Kir3) channels, causing potassium efflux from neurons and Aβ-mediated apoptosis. Although, GIRK channel activity is required for Aβ-induced neuronal degeneration, we show that it is not sufficient, with coincident signaling by the p75 neurotrophin receptor (p75NTR) also required for potassium efflux and cell death. Our results identify a novel role for GIRK channels in mediating apoptosis, and provide a previously missing mechanistic link between the excitotoxicity of Aβ and its ability to trigger cell death pathways, such as that mediated by p75NTR. We propose that this death-signaling pathway contributes to the dysfunction of neurons in Alzheimer's disease and is responsible for their eventual degeneration.


INTRODUCTION
Alzheimer's disease is a progressive neurodegenerative disorder that is characterized by deficits in memory and higher cognitive function. This cognitive decline is due to neurodegeneration, particularly of the hippocampus and entorhinal cortex, and loss of the cholinergic neurons of the basal forebrain. The brains of Alzheimer's disease patients contain deposits of aggregated amyloid β (Aβ), and a substantial body of work supports the hypothesis that Aβ underlies the etiology and pathogenesis of the disease (Hardy and Selkoe, 2002;Ballard et al., 2011). At the cellular level, the 42 amino acid form of Aβ (Aβ 42 ) readily aggregates into soluble oligomers (Haass and Selkoe, 2007) which can directly contribute to neuronal and synaptic dysfunction and degeneration (Walsh et al., 2002;Haass and Selkoe, 2007;Walsh and Selkoe, 2007;Shankar et al., 2008). These Aβ-mediated changes, which include the loss of dendritic spines and neurites (Mucke et al., 2000;Smith et al., 2009;Perez-Cruz et al., 2011), have been proposed to cause the cognitive decline, which precedes neuronal death in Alzheimer's disease (Ondrejcak et al., 2010;Palop and Mucke, 2010).
Neuronal degeneration in the context of Aβ is linked to the excitotoxic activities of the peptide, whereby an increase in intracellular calcium induced by Aβ triggers apoptosis by an ill-defined pathway (Palop et al., 2007). An alternative mechanism of Aβ-induced neurodegeneration is via the p75 neurotrophin receptor (p75 NTR ). p75 NTR initiates apoptosis in a variety of neurodegenerative conditions, including apoptotic death triggered by oligomeric forms of Aβ 42 in vitro (Yang et al., 2008;Coulson et al., 2009) and in animal models of Alzheimer's disease (Sotthibundhu et al., 2008;Knowles et al., 2009;Wang et al., 2011). p75 NTR is expressed throughout life in basal forebrain cholinergic neurons and is also ectopically upregulated in response to Aβ accumulation in other diseasevulnerable brain areas, such as the cortex and hippocampus (Mufson and Kordower, 1992;Chakravarthy et al., 2010;, and in other excitotoxic disease conditions such as motor neuron disease and epilepsy (Ibanez and Simi, 2012).
Although, the mechanism by which p75 NTR mediates cell death triggered by Aβ is unclear (Costantini et al., 2005;Coulson, 2006), we have previously demonstrated that neurotrophin signaling through p75 NTR can activate G-protein activated inwardly rectifying potassium (GIRK) channels, triggering a cell death pathway . GIRK channels are typically activated by inhibitory neurotransmission, allowing transitory potassium efflux, which lowers the neuronal resting membrane potential and dampens neuronal excitability (Dascal, 1997). Activation of GIRK channels is required for some forms of synaptic plasticity such as long-term potentiation (LTP) (Chung et al., 2009a), and GIRK expression can be regulated by enhanced glutamatergic activation (Chung et al., 2009b). As Aβ is known to cause excitotoxicity at least partially via glutamatergic processes (De Felice et al., 2007;Alberdi et al., 2010), these converging lines of evidence suggest that activation of the p75 NTR -GIRK pathway by excitotoxic Aβ 42 may play a direct role in neuronal degeneration. We therefore investigated whether Aβ treatment of neurons affects GIRK channel expression and activity, and whether the resultant changes lead to neuronal degeneration.

Primary Culture
All animal procedures were approved by the University of Queensland Animal Ethics Committee. Pregnant C57Bl6 mice were sacrificed and their E18 embryos were removed by Caesarian section. Hippocampal tissue was dissected from each embryonic brain, chopped, and digested with 0.05% trypsin (Gibco). Neurons were then dissociated by serial trituration, passed through a 40 µm filter and resuspended in medium. Neurons were plated on poly-L-lysine-coated tissue culture dishes, coverslips, or 3 cm MatTek glass-bottom dishes (MatTek Corp) and cultured in medium containing Dulbecco's modified Eagle's medium (DMEM)/Ham's F12 (Gibco), 10% fetal bovine serum (FBS; JRH Biosciences) and 2 ng/ml brain-derived neurotrophic factor (BDNF). This medium was then changed the next day to neurobasal medium (Gibco) supplemented with B27 (Gibco) and 2 ng/ml BDNF, and the cells were cultured for 10-24 days at 37 • C and 5% CO 2 with medium changes as required. For short-term, lower density cultures, neurons were plated in DMEM/Ham's F12 medium containing 10% NeuroCult (StemCell Technologies) and 2 ng/ml BDNF (Millipore) in tissue culture dishes coated with 0.1 mg/ml poly-L-lysine (Sigma).

Treatments
Aβ peptides were synthesized using t-Boc chemistry and purified using reverse phase HPLC by Dr. James I. Elliott at Yale University. To prepare Aβ peptides for treatment, both Aβ 42 and Aβ 16 were reconstituted in sterile water to 200 µM stock solution, incubated overnight at 4 • C and used at 20 µM final concentration in medium or electrophysiological bath solution. Fresh solutions were made for each assay. Co-treatments were added once to neuronal cultures at t = 0, the time of Aβ treatment, unless otherwise stated. APV (30 µM; Sigma) was used to block N-methyl-D-aspartate (NMDA) receptors. TertiapinQ (100 nM; Alomone) was used to block GIRK channel activity. To activate GABA B receptors and down-regulate GIRK channels, baclofen (Sigma) was used at 50 µM. The antagonist CGP55845 (1 µM; Tocris) was used to block GABA B receptor activity. The metalloprotease inhibitor TAPI-2 (20 µM; Calbiochem) was used to prevent cleavage of p75 NTR ; the initial treatment was done at t = 0 and a second treatment of TAPI-2 was performed 6 h later. The c29 peptide used to block p75 NTR death signaling was a 29 amino acid residue peptide of the juxtamembrane Chopper domain (Coulson et al., 2000;Matusica et al., 2013;KRWNSCKQNK QGANSRPVNQ TPPPEGEKL) fused to a non-naturally occurring protein transduction domain peptide (YARAAARNARA) based on PTD4 (Ho et al., 2001). The control used for c29 was a scrambled version of the peptide (SKGQVCRNQP GQNKPEPANK SWKETPLRN) fused to the transduction domain sequence and a fluorescent indicator (FITC; Matusica et al., 2013). These peptides were also synthesized at Yale University.

Aβ SDS-PAGE and Western Blotting
To identify the aggregation state of Aβ used in these assays, Aβ peptides were separated by SDS-PAGE and detected by western blotting. HPLC-purified Aβ solution was prepared as described above and samples were mixed with LDS sample loading buffer and loaded onto Nupage precast BisTris gels. SDS-PAGE was conducted in Nupage MES or MOPS SDS running buffer, after which proteins were transferred to an Immobilon R membrane (BioRad) in 20% methanol. Blots were incubated with primary Aβ antibody (1:500, clone 6F/3D to residues 8-17; Dako) followed by horseradish peroxidase-conjugated secondary antibodies in phosphate-buffered saline (PBS) containing 0.05% Tween 20, after which Aβ-immunoreactive bands were visualized with SuperSignal R West Pico or Femto Chemiluminescent Substrate (Thermoscientific), according to the manufacturer's instructions. Blots were exposed to film and developed by X-ray developer.

Calcium Imaging
To measure calcium flux, cultured neurons were loaded with the calcium indicator Oregon Green 488 R BAPTA-1 AM (5 µM; Invitrogen) for 1 h prior to imaging. They were then washed 3 times with medium and imaged on a Marianas TIRF/FRET/FRAP inverted high-speed imaging fluorescence microscope at 37 • C and 5% CO 2 . Neurons were imaged for ∼2 min to obtain baseline calcium fluorescence, then treated with Aβ peptides and the NMDA receptor antagonist APV as indicated, before being imaged 5 min later for ∼2 min (for APV) and/or 15 min later (for Aβ). Calcium flux was observed in both shortterm and long-term cultures. Although, the short-term cultures were spontaneously active, the synchronous firing of the longterm (21-24 days) cultures facilitated quantification of calcium flux, as a result of which long-term cultures were used for these experiments. Fluorescence values of all cells in the field of view (at least three fields of view per condition) of mature cultures were quantified using Slidebook software, and the data were pooled within each condition and analyzed using one-way analysis of variance (ANOVA).

Potassium Imaging
To quantify change in the intracellular potassium concentration, neurons that had been cultured for 3 days were incubated with 2 µM Asante Potassium Green-2 (Teflabs) indicator for 30 min. After exchange of the medium, cultures were left to recover for a minimum of 15 min in an incubator before being transferred to the Axio observer microscope chamber held at 37 • C and 5% CO 2 . Fluorescence (488 nm) and DIC (Nomarski) images were captured every 5 min for 3 h. Test compounds, including Aβ 42 , were added 15 min after imaging had commenced. The fluorescence values of all cells in the field of view (six fields of view per condition) were quantified using Imaris software from images taken 5 min before the addition of the test compounds (time 1) and 30, 110, or 160 min after treatment (time 2). The percentage change in fluorescence over time for each cell was then calculated.

In vitro Apoptosome Assay
Apoptosome components were isolated from the soluble fraction of cell lysates obtained from nerve growth factor (NGF)differentiated PC12 cells. PC12 cells were maintained at 37 • C with 10% CO 2 in DMEM supplemented with 10% horse serum (Sigma) and 5% FBS before being differentiated with 50 ng/ml NGF in DMEM supplemented with 0.1% horse serum for 2 days. They were then collected by scraping, washed with ice cold PBS, and centrifuged to form tight cell pellets which were weighed and resuspended 1:1 with hypotonic extraction buffer containing (in mM): 5 EGTA, 50 PIPES, 2 MgCl 2 , 1 DTT, 0.1 PMSF, pH 7. Cells were allowed to swell on ice for 25 min before being sheared with 100 strokes of a B-type pestle in a Dounce homogenizer (Kimble-Kontes). The solution was centrifuged at 100,000 g for 1 h at 4 • C, after which the supernatant was aliquoted and stored immediately at -80 • C. Lysate containing 100 µg protein was incubated in a 50 µl reaction mix with cytochrome C (0, 0.5, 1.0, 1.5, or 2 µM), 0.5 µM dATPs, and the fluorescent caspase substrate Ac-DEVD-AMC (100 µM), with KCl added to a concentration of 20, 50, 80, 110, or 140 mM. All conditions were assessed in triplicate. Cleavage of the substrate was measured by a Polarstar Optima plate reader every 5 min for 1 h (excitation filter, 380 ± 10 nm; emission filter, 460 nm); triplicate readings were pooled for each condition.

Neurite Outgrowth Assay
Neurons were plated at a density of 40,000 cells per well of a 24-well plate (Falcon) and grown for ∼3 weeks. They were then treated with Aβ, baclofen and/or CGP55845. After 20 h, neurons were washed, fixed with 4% paraformaldehyde, immunostained for β-III tubulin and co-stained with DAPI. Three random fields of view were imaged for each condition per experiment; images were overlaid with a grid and the neurite crossings were counted and divided by the number of neurons per field (DAPI-and β-III tubulin-positive soma) to give a ratio of neurites per neuron. Ratios were averaged across conditions and data were analyzed by one-way ANOVA.

In vitro Neuronal Survival Assays
To determine the level of neuronal survival, low density hippocampal cultures were plated at 40,000 cells per 11 mm diameter well of a 4-well plate (Cell-star R , Greiner) with a grid marked for later cell identification. Cells were cultured overnight in growth medium at 37 • C and 5% CO 2 . The number of neurons within a fixed grid quadrant was counted 24 h after plating (t = 0) and the same quadrant was counted again 20 h later (t = 20). Live cells were determined by morphology, and, in some experiments, also by exclusion of propidium iodide. Similar results were obtained when MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] was used to identify live cells in the cultures. All treatments were applied immediately after the initial cell count and remained in the culture medium until the end of the experiment. Survival was expressed as the percentage of neurons remaining alive at t = 20. Under each condition a minimum of 250 neurons to a maximum of 600 neurons were counted across 4 different gridded wells at t = 0. Each condition was replicated a minimum of 3 times (N = 3), with neurons derived from different litters on different days.

Statistical Analysis
Statistical analyses were performed using Prism 4 for Macintosh (GraphPad Software, Inc.). Two group comparisons were made using t-tests. For multiple comparisons, data were analyzed by ANOVA conducted using Newman-Keuls posttest comparisons, except for the potassium imaging data which were analyzed using Tukey's multiple comparisons post-test. All graphs are mean ± SEM.

Aβ Upregulates GIRK Channel Surface Expression and Activity
We first prepared a solution of Aβ 42 that contained predominantly low-order oligomers and monomers (Figures 1A,B). Treatment of cultured embryonic mouse hippocampal neurons with 20 µM Aβ 42, but not the control Aβ 16 peptide, raised the neuronal intracellular calcium concentration ( Figure 1C), indicating that it had excitotoxic properties. Furthermore, the calcium increase induced by Aβ 42 occurred most obvious in spontaneously active neuronal cultures (data not shown) and was inhibited by the NMDA receptor antagonist APV (Figure 1C), indicating glutamate receptor activity was required for the excitotoxicity. As it has been shown that over-activation of the NMDA receptor increases the number of GIRK channels at the surface of neurons (Chung et al., 2009b), we investigated the effect of Aβ 42 on GIRK channel expression in cultured embryonic mouse hippocampal neurons. Western blot analysis demonstrated that surface GIRK1 and 2 protein levels were significantly increased in neurons 2 h after Aβ 42 , but not control Aβ 16 , application (Figures 1D,E). Total cellular levels of GIRK1 and 2 subunits were unchanged by the treatments (Figures 1D,E). These data indicate that Aβ 42 causes a rapid redistribution of existing GIRK subunits into the plasma membrane.
hippocmapus. In the presence of picrotoxin to block GABA A receptor-mediated synaptic currents, electrical stimulation of the Schaffer collaterals generated slow IPSCs in CA1 hippocampal neurons when voltage-clamped at a holding potential of −50 mV ( Figure 2C). Bath application of Aβ 42 significantly increased the amplitude of sIPSCs (green trace; one-way ANOVA, Bonferroni test, p < 0.0001, n = 7; Figures 2D-F), which were completely blocked by bath application of the GABA Bselective antagonist CGP55845 (1 µM; Figures 2D-F). This effect on the synaptically evoked GIRK currents was observed from as early as 8 min after Aβ 42 application and was sustained for the duration of the recording (up to 40 min). These data indicate that synaptically activated GIRK channel currents are upregulated following Aβ 42 treatment, leading to enhanced slow inhibitory neurotransmission in the CAI hippocampal circuit.

GIRK Channel Activity Results in Lowered Intracellular Potassium
We next asked if significant potassium efflux occurred via potassium channels following Aβ treatment by measuring the intracellular potassium concentration using a fluorescent indicator APG-2. APG-2 fluorescence declined over time in all conditions over the course of 160 min, likely due to potassium efflux through leak channels ( Figure 3A). However, the intracellular fluorescence of Aβ 42 -treated neurons significantly decreased compared to that of Aβ 16 -treated neurons (Figure 3A), indicating a considerable loss of intracellular potassium in the . Neuronal physiological K + concentrations (>110 mM) completely suppressed caspase activity regardless of cytochrome c concentration, and maximal caspase activity occurred in the lowest (20 mM) K + concentration. Cytochrome c concentrations higher than 1 µM combined with K + concentrations lower than 80 mM allowed caspase activity to occur (data points are the average of triplicates; all non-overlapping data points at these concentrations are significantly different at p ≤ 0.01).
former condition. As it has been determined that the APG-2 fluorescence is saturated at potassium concentrations higher than ∼80 mM (Rimmele and Chatton, 2014), this result indicates that the internal potassium concentration of many of the Aβ 42 -treated neurons was falling below 80 mM. To determine whether the potassium efflux might be mediated via GIRK channels, neurons were treated with the most specific GIRK channel inhibitor, tertiapin, which significantly inhibited the Aβ 42 -induced reduction in intracellular potassium (Figures 3B,C).
Activation of caspases by the apoptosome can be regulated by potassium levels (Yu and Choi, 2000;Cain et al., 2001;Coulson et al., 2008). To examine whether the reduced level of intracellular potassium in the neurons treated with Aβ was capable of mediating caspase activation, we determined the relationship between potassium concentration and apoptosome formation. This was achieved by using a cell-free assay in which the core mitochondrial cell death machinery of the apoptosome, comprising cytochrome c, caspase 9, and Apaf1 (Riedl and Salvesen, 2007), was derived from cell lysates. The ability of these components to form an apoptosome and cleave a fluorogenic caspase 3 substrate was measured in decreasing concentrations of potassium buffer and increasing concentrations of exogenous cytochrome c. Physiological levels of potassium (110-140 mM; Yu and Choi, 2000) were able to prevent all apoptosome activity regardless of the cytochrome c concentration ( Figure 3D). By contrast, 80 mM potassium or less resulted in significant caspase activity, indicating permissiveness for apoptosome formation (Figure 3D). These in vitro assays demonstrated that a potassium concentration approaching half that typically found in healthy neurons, and equivalent to that induced in Aβ 42 -treated neurons, is necessary for activation of the apoptosome by cytochrome c, which in turn is a prerequisite for apoptosis via the mitochondrial cell death pathway (Cain et al., 2001).

GIRK Channel Activity Is Required for Aβ 42 -Induced Neuronal Degeneration
We next treated neuronal cultures with Aβ overnight, finding that Aβ 42 but not Aβ 16 treatment induced significant neurite degeneration as well as the death of more than 60% of neurons, with degenerative changes being observed after 5 h of treatment ( Figure 4A).
We then asked whether potassium efflux was required for Aβ-induced toxicity by increasing the extracellular potassium concentration in the culture medium from the normal 5-25 mM. First, we recorded from cultured hippocampal neurons and found that raising the level of extracellular potassium shifted the reversal potential of the GIRK current from −93 ± 4 to −36 ± 2.2 mV (n = 4). With the same change in potassium, the resting membrane potential depolarized from −81 ± 4 to −40 ± 2 mV. Importantly, the resting membrane potential shifted from being more positive than the equilibrium potential for potassium to being close to this equilibrium potential. Using the measured current-voltage relationships, the expected GIRK current at the resting membrane potential in low potassium was +21.2 ± 10 pA whereas in high potassium it was −9.6 ± 8.9 pA, i.e., there was virtually no outward current. We then raised the extracellular potassium concentration in the culture medium and found that it inhibited Aβ-induced death of neurons ( Figure 4B). Similarly, blocking GIRK channels with the GIRK channel inhibitor tertiapin which significantly inhibited the Aβ 42 -induced potassium efflux (Figure 3B), also inhibited neuronal death (Figures 4A,C). We therefore asked whether increased surface expression and activity of GIRK channels was a requirement for this Aβ-induced neurotoxicity. We reasoned that sustained stimulation of GABA B receptors, which are tightly coupled to GIRK channels, may lead to channel desensitization, and/or endocytosis and degradation of the entire receptor-channel complex, such as occurs for similar receptor-channel complexes (Clancy et al., 2007;Fowler et al., 2007;Raveh et al., 2010). Consistent with this idea, we found that a chronic 2 h treatment of cultured neurons with the GABA B agonist baclofen reversed the upregulation of GIRK1 and 2 subunits on the surface of neurons induced by Aβ treatment, without changing the total GIRK subunit level (Figures 5A,B). Chronic baclofen treatment also inhibited the loss of potassium from Aβ-treated cells (Figures 5C,D). Furthermore, baclofen treatment significantly inhibited neuronal degeneration in longterm cultures (Figures 5E,F) and neuronal death (Figure 5G) induced by Aβ 42 over 24 h in short-term cultures.
Taken together, these results indicate that potassium efflux through GIRK channels is a major mediator of Aβ 42 -induced neurotoxicity in hippocampal cultures and suggests that Aβ induces the upregulation of GIRK channel activity, thereby resulting in potassium efflux, reduced intracellular potassium, and apoptosis.

Death Signaling Is Mediated by p75 NTR Signals
Despite an increase in GIRK channels at the cell surface, their typical activation by neurotransmitter-mediated Gβγ (e.g., following activation of GABA B receptors) does not lead to a sustained potassium efflux or the lowering of intracellular potassium levels sufficient to promote cell death (see Section Discussion). However, we have previously demonstrated that p75 NTR can cause pathological activation of GIRK channels via upregulation of PIP 2 levels , which does not require Gβγ and which is necessary and sufficient for GIRK channel activation (Zhang et al., 1997;Huang et al., 1998). Furthermore, several groups, including ours, have reported a key role for p75 NTR in Aβ-induced neuronal degeneration (Sotthibundhu et al., 2008;Yang et al., 2008). Percentage survival of neurons cultured in the presence of Aβ and baclofen for 20 h. Down-regulation of GIRK channels by chronic baclofen treatment inhibited cell death, but the neurotoxicity of Aβ 42 was restored when neurons were co-cultured with the GABA B receptor antagonist CGP55845 (CGP; N = 5 experiments). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns, not significant. peptides over 20 h. c29 but not a scrambled peptide inhibited Aβ 42 -initiated death (N = 3 experiments). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns, not significant.
As p75 NTR -mediated PIP 2 generation and GIRK channelactivating signals were previously found to be dependent on metalloprotease cleavage of p75 NTR to its C-terminal fragment , we determined whether Aβ 42 treatment induced p75 NTR proteolysis. Aβ 42 but not Aβ 16 stimulated increased generation of the p75 NTR C-terminal fragment ( Figure 6A). Furthermore, an inhibitor of the p75 NTR metalloprotease (TAPI) blocked both Aβ-induced cleavage of p75 NTR and neuronal death (Figures 6A,B), indicating that activation of p75 NTR is necessary for Aβ-induced cell death.
Finally, we treated neurons with a cell-permeable peptide inhibitor of the p75 NTR C-terminal fragment that initiates the GIRK channel activity pathway  and which can act by a dominant-negative mechanism to prevent death signaling (Coulson et al., 2000). Treatment of neurons with this peptide, c29, had no effect on Aβ-induced cell surface expression of GIRK subunits (Figures 6C,D), but significantly FIGURE 7 | Proposed Aβ 42 -p75 NTR -GIRK channel death signaling pathway. Aβ 42 excitotoxicity increases intracellular calcium in neurons via NMDA receptors (NMDAR), which triggers an increase in the number of GIRK channels at the cell surface. Aβ 42 also results in the cleavage of p75 NTR to its C-terminal fragment, which can then activate the GIRK channels resident on the surface to promote a pathological potassium efflux. Lowered internal potassium concentration removes inhibition of apoptosome formation, thereby causing cell death. Cell death can be inhibited by raised extracellular potassium, the GIRK channel inhibitor tertiapin, chronic baclofen treatment removing GIRK channels from the cell surface, preventing p75 NTR cleavage or treatment with a p75 NTR signaling inhibitor (c29).
inhibited the enhanced potassium efflux from neurons cotreated with Aβ 42 (Figures 6E,F). Furthermore, c29 peptide treatment significantly inhibited Aβ 42 -induced neuronal death, whereas a cell-permeable scrambled control peptide had no effect ( Figure 6G). Together, these results suggest that p75 NTRmediated signals induced by Aβ 42 are concomitantly required for potassium efflux through GIRK channels and subsequent neuronal death.

DISCUSSION
Here we describe a novel, Aβ-triggered apoptotic pathway in which exposure to excitotoxic Aβ leads to upregulation and activation of GIRK channels, causing sustained potassium efflux from neurons and resulting in their apoptosis. Although, upregulation of GIRK channels is required for this death signaling, it appeared that this is not sufficient, with coincident activation of p75 NTR signaling by Aβ also being necessary for neuronal degeneration.

Surface GIRK Channels Are Upregulated by Aβ
Our first finding is that Aβ mediates the upregulation of GIRK channel expression on the neuronal plasma membrane. Although, the reason for this was not investigated, it has previously been observed that robust activation of NMDA receptors leads to an increase in cell surface recruitment of GIRK channels (Chung et al., 2009a,b). Analogous to this situation, Aβ can raise intracellular calcium levels by an NMDA receptordependent mechanism, and Aβ has also been shown to cause aberrant NMDA receptor activity and hyperexcitability (Palop et al., 2007). Consistent with this, the NMDA receptor antagonist APV was observed to block Aβ-induced calcium influx, as well as GIRK-mediated potassium efflux. It is therefore possible that GIRK channels are upregulated by such mechanisms in the current study (Chung et al., 2009b;Yao et al., 2013).
An acute effect of the upregulation of GIRK channels was enhanced slow inhibitory neurotransmission within the hippocampal CA1 circuit. Recruitment of GIRK channels to the plasma membrane is also a critical step in the depotentiation of NMDA-receptor-driven LTP in the hippocampus (Chung et al., 2009a,b). Therefore, Aβ-mediated upregulation of GIRK channels likely causes persistent depression of synaptic activity as observed in the mouse model of Down's syndrome (Ts65Dn). GIRK currents recorded in hippocampal neurons of these animals, which overexpress GIRK channels due to an additional copy of the GIRK2 gene (in addition to the amyloid precursor protein gene), are significantly more sensitive to inhibitory input through GABA B receptors, which others have shown leads to an impairment of excitatory input and cognitive deficits (Harashima et al., 2006;Best et al., 2007). Moreover, a recent report demonstrated that Aβ perfusion of hippocampal slices caused increased resistance of CA3 neurons to firing, an effect which was mediated by GABA B and GIRK channel activity (Nava-Mesa et al., 2013). Our findings provide an explanation for previous reports that, although Aβ is excitotoxic, neurons exposed to Aβ become synaptically silent prior to their death (Palop et al., 2007;Palop and Mucke, 2010;Yao et al., 2013).

Upregulated GIRK Channel Activity Is Necessary for Aβ-Induced Cell Death
The upregulation of GIRK channels by Aβ resulted in significant potassium efflux, demonstrated using electrophysiology and potassium imaging, that was required for subsequent neuronal death. In addition, blocking GIRK channels with tertiapin, or potassium efflux more generically, was sufficient to prevent this Aβ-induced potassium loss and subsequent neuronal degeneration. Furthermore, chronic baclofen treatment also blocked Aβ-induced potassium efflux, prevented neuronal degeneration, and down-regulated surface GIRK channel expression. Although, it is possible that the baclofen treatment masked the toxic effect of Aβ by an unrelated mechanism, taken together with our other data, the correlation between baclofen-induced down-regulation of GIRK channel activity and reduced Aβ-toxicity is consistent with a causal relationship. Potassium dysregulation is a feature of Alzheimer's disease brain tissue (Roberts et al., 2016), and similar to our findings, others have shown that potassium efflux from cortical neurons can be promoted by Aβ, and that this efflux is required to cause subsequent degeneration, although the mechanism remains unclear (Yu et al., 1998;Shabala et al., 2010). Our data indicate that Aβ 42 treatment causes sufficient potassium to leave the cell to reduce the intracellular potassium concentration to approximately half (80 mM), which in turn is permissive for the apoptosome to assemble and caspases to be activated, a finding that is consistent with previous reports (Cain et al., 2001;Coulson et al., 2008). In addition, our results strongly indicate that GIRK channels are key mediators of this critical apoptotic step. However, our findings do not rule out the possible involvement of other potassium channels in Aβ-mediated neurotoxicity.

Coincident Activation of p75 NTR and GIRK Channels Is Required to Induce Cell Death
Although, upregulation of membrane GIRK channels and increased potassium efflux were directly associated with Aβinduced neuronal degeneration, it remains unclear whether the pathological GIRK channel activation occurred by inhibitory neurotransmission and/or through a coincident Aβ activated pathway-such as that mediated by p75 NTR . Although, neurotransmitter receptor activation may play a contributory role in the observed potassium efflux, chronic stimulation of GABA B receptors inhibited cell death, coincident with GIRK channels being removed from the cell surface. Likewise, it is unlikely that the channels were activated by G q -coupled neurotransmitter receptor-mediated mechanisms, as these cause the hydrolysis of PIP 2 and inhibit channel activity (Raveh et al., 2010), with PIP 2 being required for GIRK channel opening. Several groups have previously reported that p75 NTR can increase PIP 2 levels via Rac1 (Gibon et al., 2015;Zeinieh et al., 2015), a necessary step for GIRK channel activation by p75 NTR . We therefore suggest that p75 NTR mediates pathological GIRK channel activity subsequent to Aβ-induced channel upregulation, nominally ∼2-3 h after Aβ application.
Regardless of whether or not p75 NTR is directly responsible for GIRK channel activity, p75 NTR signaling is required for Aβinduced cell death, as blocking this signaling, even in the context of enhanced surface GIRK expression, prevented both loss of cellular potassium and cell death. Aβ has been widely reported to activate p75 NTR either directly or indirectly, leading to neuronal degeneration; neurons with reduced p75 NTR expression or function are resistant to Aβ-induced toxicity in vitro and in vivo (Yaar et al., 1997;Ivins et al., 1998;Tsukamoto et al., 2003;Sotthibundhu et al., 2008;Yang et al., 2008;Knowles et al., 2009;Yu et al., 2012). However, because p75 NTR can regulate a range of signaling pathways, the mechanism by which it mediates apoptosis in response to Aβ has remained unclear (Coulson, 2006;Skeldal et al., 2011).
We suggest that Aβ causes the induction of GIRK channel surface expression which, when activated acutely by traditional G-protein-coupled receptors, enhances inhibitory neurotransmission (Figure 7). However, sustained exposure to Aβ coincidently results in activation of p75 NTR cleavage and signaling pathways, one of which involves the activation of GIRK channels. As activation of GIRK channels by p75 NTR occurs through increased levels of PIP 2 and independently of G-proteins , potassium efflux through GIRK channels could be sustained (Raveh et al., 2010). Sustained channel activity resulting in substantial potassium efflux then triggers apoptosis. However, alternatively, or coincident with p75 NTR -mediated GIRK channel activity, other p75 NTR -mediated signals (e.g., activation of c-jun kinase; Coulson, 2006) could facilitate the activation of the apoptosome via the mitochondrial death pathway in the context of already lowered intracellular potassium mediated via GIRK channels by other means.

Summary
In conclusion, we have demonstrated that exposure to excitotoxic Aβ drives an increase in the number of surface GIRK channels which, when activated, can directly result in increased inhibitory neurotransmission and neuronal circuit silencing. Over a longer timescale, GIRK channel activity is a key mediator of neuronal degeneration and apoptosis. However, for this novel cell death pathway to proceed, it also requires signaling by p75 NTR , which can be coincidently triggered by Aβ exposure, resulting in pathological GIRK channel activity.

ETHICS STATEMENT
All animal procedures were approved by the University of Queensland Animal Ethics Committee in accordance with the Australian code for the care and use of animals for scientific purposes (2013).

AUTHOR CONTRIBUTIONS
LM, VA, HG, PS, and EC designed the experiments and wrote the manuscript; LM, HG, VA, DM, SJ, and GK performed experiments; LM, HG, VA, DM, FM, and PS analyzed data and all authors discussed the work and commented on the manuscript.