Cognitive deficits caused by a disease-mutation in the α3 Na+/K+-ATPase isoform

The Na+/K+-ATPases maintain Na+ and K+ electrochemical gradients across the plasma membrane, a prerequisite for electrical excitability and secondary transport in neurons. Autosomal dominant mutations in the human ATP1A3 gene encoding the neuron-specific Na+/K+-ATPase α3 isoform cause different neurological diseases, including rapid-onset dystonia-parkinsonism (RDP) and alternating hemiplegia of childhood (AHC) with overlapping symptoms, including hemiplegia, dystonia, ataxia, hyperactivity, epileptic seizures, and cognitive deficits. Position D801 in the α3 isoform is a mutational hotspot, with the D801N, D801E and D801V mutations causing AHC and the D801Y mutation causing RDP or mild AHC. Despite intensive research, mechanisms underlying these disorders remain largely unknown. To study the genotype-to-phenotype relationship, a heterozygous knock-in mouse harboring the D801Y mutation (α3+/D801Y) was generated. The α3+/D801Y mice displayed hyperactivity, increased sensitivity to chemically induced epileptic seizures and cognitive deficits. Interestingly, no change in the excitability of CA1 pyramidal neurons in the α3+/D801Y mice was observed. The cognitive deficits were rescued by administration of the benzodiazepine, clonazepam, a GABA positive allosteric modulator. Our findings reveal the functional significance of the Na+/K+-ATPase α3 isoform in the control of spatial learning and memory and suggest a link to GABA transmission.

Scientific RepoRts | 6:31972 | DOI: 10.1038/srep31972 currently four different mutations are known; D801Y causes RDP 1,11 or AHC 12 , and D801N, D801E and D801V cause or AHC 2,3,13-15 . In the central nervous system (CNS), the α 3 isoform is expressed in neurons whereas the α 2 subunit is expressed in glia, and the α 1 subunit appears to be expressed ubiquitously 16,17 . Together, these Na + /K + -ATPase isoforms are responsible for maintaining the Na + and K + electrochemical gradients that determine cell resting membrane potentials, and support the electrical activity of excitable cells, as well as the transport of other ions and metabolites and driving neurotransmitter reuptake 18 . Although the role of the Na + /K + -ATPases in the etiology of neurological diseases is poorly understood, reduced Na + /K + -ATPase activity has been linked to conditions such as epileptic seizures and schizophrenia 19,20 . The distinguishing feature of α 3 Na + /K + -ATPases is their several-fold lower affinity for activation by cytoplasmic Na + compared to that of α 1 Na + /K + -ATPases 21 . In rapidly firing neurons, therefore, when action potentials increase the intracellular Na + concentration, [Na + ] i , beyond levels saturating the "housekeeping" α 1 Na + /K + -ATPases, activation of α 3 Na + /K + -ATPases continues to increase as [Na + ] i rises. As [Na + ] i is linked to [Ca 2+ ] i through the Na + /Ca 2+ exchanger, the α 3 isoform thus protects neurons against catastrophic elevation of [Na + ] i and [Ca 2+ ] i 22 and general loss of the Na + electrochemical gradient. Atp1a3 mouse models have provided insights into the role of the α 3 isoform in neurological diseases. Currently, two knock-out and two knock-in mouse models have been reported. The heterozygous knock-out α 3 +/KOI4 mouse (Atp1a3 tm1Ling ) displayed spatial learning and memory deficits, hyperlocomotion and increased locomotor response to methamphetamine 23 . The α 3 +/ΔE2-6 mice knock-out mouse (Atp1a3 tm1.1Kwk ) showed increased sensitivity to kainate-induced dystonia and enhanced inhibitory neurotransmission of molecular-layer interneuron-Purkinje cell synapses in the cerebellar cortex 24 . The Myshkin mouse harbors the heterozygous I810N disease mutation in the Atp1a3 gene (Myshkin, α 3 , Atp1a3 Myk ), which in humans causes AHC 25 . The Myshkin mice are characterized by seizure activity 26 and mania-like behavior, and showed increased response to amphetamine, similar to what has been reported for bipolar patients 27 as well as motor dysfunction and cognitive impairments related to compromised thalamocortical functionality 28 . The D801N mutation is found in more than one third of AHC patients 3 . A recent study reported that heterozygous D801N knock-in mice (Mashl +/− , Mashlool, α 3 +/I801N ) 29 manifested several AHC-like symptoms including neuromuscular deficits, spontaneous recurrent seizures, and predispositions to kindling, to flurothyl-induced seizures and to Sudden Unexpected Death in Epilepsy (SUDEP) 29 .
To further address the complex genotype-phenotype relationship particular to the D801 amino acid position (the D801N/E/V mutations are associated with AHC whereas D801Y is associated with both RDP 1,11 and AHC 12 ) in the α 3 isoform, the α 3 +/D801Y mouse (Atp1a3 tm1Klh ) was generated and the general behavior and cognitive functions were explored. We found that α 3 +/D801Y mice display ATP1A3-related symptoms such as hyperactivity, lower threshold for PTZ-induced epileptic seizures, and reduced hippocampus-dependent cognitive performance.
The hippocampus has been suggested to be the primary brain structure for spatial memory acquisition, memory storage and consolidation 30 . The hippocampal formation comprises dentate gyrus, the hippocampus proper and the adjacent parahippocampal cortices. The major excitatory input to the hippocampus arises from the entorhinal cortex via the perforant path that primarily terminates in the dentate gyrus. The dentate axons project to the CA3 region and from there the Schaffer collaterals convey the processed input to the CA1 area.
In spite of this, we did not observe any change in the excitability of CA1 pyramidal neurons in the α 3 mice. The cognitive deficits were rescued by administration of the GABA positive allosteric modulator, the benzodiazepine clonazepam. The α 3 +/D801Y model demonstrates a role in cognition comparable to the D801Y AHC manifestation, and will be suitable for investigations of disease mechanisms and development of therapeutic interventions.

Results
Non-Mendelian ratio and reduced α 3 protein. Upon generation of the α 3 +/D801Y mouse model ( Fig. 1), the line was back-crossed for > 7 generations before testing. Analysis of the Mendelian distribution among 200 offspring, at 3 weeks of age, showed that 35% were genotyped as α 3 +/D801Y (Fig. 2a), suggesting neonatal absorption and/or perinatal death. It should be noted that homozygous α 3 D801Y/D801Y mice died at birth. The introduction of the D801Y mutation caused a 15% reduction in total α 3 protein levels, but a 33% increase of the α 1 protein levels, as tested by Western blotting (WB) analysis of whole brain, cortex, hippocampus, and cerebellum lysates from adult α 3 +/D801Y mice relative to WT levels (Fig. 2b,c). α 3 +/D801Y caused hyperactivity, but not anxiety. In the open field test (OF), the α 3 +/D801Y mice displayed hyperlocomotion relative to WT mice (Fig. 3a). After an initial habituation period of 8-10 minutes, WT mice showed a typical increase in horizontal rotation and meander. In contrast, α 3 +/D801Y mice showed minimal habituation and almost exclusively changed direction when reaching the walls of the enclosure (Fig. 3b,d). Track plot analysis revealed a significant increase in time spent in the OF periphery (Fig. 3c). We hypothesized this to be a consequence of the low meandering rather than an indication of anxiety. In support, when tested in the elevated plus maze (EPM), the α 3 +/D801Y mice did not discriminate between entering open and closed arms (Fig. 3e,f) and spent 240% more time in the EPM open arms compared to WT mice (Fig. 3g).
Thus, the α 3 +/D801Y mice appears to reflect hyperactivtity and arousal to some degree, as they became highly agitated and displayed hyperactivity in response to handling and novel environments (described below in the Barnes Maze test), related to symptoms described for AHC patients.
Reduced seizure threshold in the α 3 +/D801Y mice. Corresponding to the high rate of seizures reported for AHC patients, reduced Na + /K + -ATPase activity was shown to influence seizure activity in the Myshkin mouse model 26 and contribute to SUDEP in the Mashl +/− mouse model 29  intraperitoneally with the non-competitive GABA antagonist, pentylenetetrazole (PTZ). PTZ induced a significantly stronger effect in the α 3 +/D801Y mice as shown by the increased lethality in the α 3 +/D801Y mice relative to WT mice (Fig. 4).
The excitability of CA1 pyramidal neurons is not changed in the α 3 +/D801Y mice. The hippocampal CA1 region is one of the brain areas in which PTZ induces highly synchronized epileptiform burst activity 31 . We therefore hypothesized that a reduced PTZ seizure threshold would be reflected in an increased excitability of CA1 pyramidal neurons. Using intracellular recordings in acute brain slices, no major difference in the basic membrane properties of CA1 pyramidal neurons in α 3 +/D801Y and WT mice was found. The resting membrane potential (RMP) and input resistance (R in ) were similar in α 3 +/D801Y and WT mice (Supplementary Table 1). The threshold for induction of action potentials (APs) (Supplementary Table 1) as well as the overall composition of the APs were also similar ( Fig. 5a) and typical of CA1 pyramidal neurons 32 . The only difference found with respect to the AP was a slight, but significant, reduction of the rate of decay (rate of repolarization) in α 3 +/D801Y mice compared to WT (Fig. 5a). Depolarizing current pulses (1 s) induced trains of APs displaying frequency accommodation in WT and α 3 +/D801Y mice (Fig. 5b). However, the amount of accommodation was less in α 3 mice compared to WT mice, primarily due to a slower initial discharge rate in α 3 +/D801Y mice compared to WT mice (Fig. 5c). No significant difference in steady-state electroresponsive behavior, measured as the frequency vs. current (f-I) relationship, was found between α 3 +/D801Y and WT mice (Fig. 5d). The α 3 Na + /K + -ATPase has been suggested to serve as a "reserve" transporter activated when [Na + ] i is high, such as following prolonged high frequency discharge (reviewed in 22 ). High frequency firing was evoked by a 20 s suprathreshold depolarizing current pulse. The assumption being that a reduced "reserve" capacity in α 3 mice would lead to a higher rise in the [Na + ] i concentration, resulting in a more pronounced activity-dependent reduction of the amplitude of aPs. However, the activity-dependent reduction was similar between the two genotypes ( Fig. 5e), as was the amplitude of the post-pulse slow afterhyperpolarization (Supplementary Table 1).

Spatial learning and memory is reduced in a 3
+/D801Y compared to WT littermates. More than 90% of AHC patients show signs of developmental delay or mental retardation 7,8 . Similarly, a recent care report showed that 90% of RDP patients with onset at or after 18 years had trouble learning in school 33 . The α 3 mice were tested for spatial learning and memory performance using the Barnes Maze (Hippocampus-based spatial reference memory) and passive avoidance (amygdala-and hippocampus-based memory) 34 .
In the Barnes Maze test, the mice were subjected to 4 training sessions per day for 4 days and subsequently a single test on day 5 and day 12. WT mice showed a significant reduction in latency to enter the escape tunnel (total latency). A similar learning curve was not observed for the α 3 +/D801Y mice (Fig. 6a). Interestingly, the WT and α 3 +/D801Y mice would reach the escape tunnel at the same time (primary latency) (Fig. 6b). However, once at the escape tunnel, the WT mice quickly entered, whereas the α 3 +/D801Y mice would walk past multiple times before entering (Fig. 6c). Further analysis of zone occupancy confirmed that both genotypes after training spent the majority of time investigating the target area. Whereas WT mice specifically occupied the target zone on testing day 5 and 12, the α 3 +/D801Y mice showed equal interest in the adjacent zone (− 1) and remained outside for extended periods (Fig. 6d).
To assess the learning performance further, the strategy used to locate the target zone was analysed. Strategies were categorised as 1) Direct, where the mouse located the target tunnel or an adjacent hole using external cues 2) Serial, where the mouse seemingly chose a random hole and subsequently searched adjacent holes in a serial manner in a clockwise or counterclockwise direction or 3) Mixed, where the mouse displayed a more random search pattern and occasionally crosses the center of the circular platform.
During the first 4 days of training, it was evident that both genotypes shifted from employing the mixed search strategy to using either the serial or direct approach (Fig. 6e). At day 5, both genotypes showed a 50/50 utilisation of the serial and direct strategies. When testing the mice one week later on day 12, none of the α 3 +/D801Y used the direct strategy and had reverted to using the serial or the mixed strategy. In contrast, the WT mice still utilised the direct and serial strategies (Fig. 6e).
Fear memory is reduced in a 3 +/D801Y compared to WT littermates. We used passive avoidance test to assess fear memory in the mice. Compared to WT mice, the α 3 +/D801Y mice showed a significantly reduced latency to re-enter the dark compartment (Fig. 6f).
Alterations in inhibitory interneurons contribute to cognitive deficits associated with several psychiatric and neurological diseases. Inhibition by GABA receptors regulating neuronal activity helps to establish the appropriate network dynamics that support normal cognition 35 . To investigate if GABA transmission might be involved in the cognitive deficits observed in the mice, α 3 +/D801Y and WT littermate mice were injected with the benzodiazepine, clonazepam and subsequently tested in the passive avoidance test. Clonazepam administrated intraperitoneally at 0.0625 mg/kg has previously been shown not to cause significant sedation or anxiolytic effect in the OF and EPM 36 . At this concentration, clonazepam completely normalized the performance of α 3 +/D801Y mice in the passive avoidance test (Fig. 6f). After an initial habituation period, WT mice displayed increased meander defined as horizontal rotation per distance traveled. This behavior was completely absent in the α 3 +/D801Y mice, here shown as the average score of 14 α 3 +/D801Y and 14 WT mice (d). Representative track plots of elevated plus maze exploration are shown in (g). WT mice (N = 14) preferred to enter the closed arms whereas α 3 +/D801Y mice (N = 15) showed no arm preference (e,f). The α 3 +/D801Y mice occupied the open arms significantly longer than WT mice (g). All data shown are means ± SD *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Reduced number of hippocampal dentate gyrus granule cells in the α 3 +/D801Y mice. Since both memory tests pointed towards hippocampal defects, we performed histological examination of this brain region (Fig. 7a). Subsequent stereological counting showed a significantly reduced number of granule layer neurons in the dentate gyrus of the α 3 +/D801Y mice (Fig. 7b). . The overall composition of the APs was similar and in both strains the APs were typically followed by a fast (f) and medium (m) afterhyperpolarization (AHP). There was no significant difference in the amplitude, half-width, or rate-of-rise between the WT and α 3 +/D801Y mice; however, the rate of decay of APs was significantly slower in α 3 Representative examples of discharge behavior in response to a 1 s depolarizing pulses in WT (N = 13) and α 3 +/D801Y (N = 11) mice. During repetitive firing, neurons from both strains displayed frequency accommodation (b). Plot of the averaged instantaneous discharge frequency vs. interspike interval number during repetitive firing of 16 Aps (c). Plot of discharge frequency vs. current intensity (d). Representative examples of responses to 20 s long suprathreshold current pulses in WT (N = 13) and α 3 +/D801Y (N = 11) mice. In both strains there was relatively fast decline of the AP amplitude to a maintained level, and termination of the current pulse was followed by a pronounced slowly decaying afterhyperpolarization (AHP s ). No significant difference was found in the average extent of AP amplitude decline (left) or in the magnitude of the AHP s (right). The pre-pulse baseline potential was − 67.4 ± 0.7 mV and − 66.6 ± 1 mV in WT (N = 12) and α 3 +/D801Y (N = 12) mice, respectively (e). All data shown are means ± SEM. *P < 0.05. Scientific RepoRts | 6:31972 | DOI: 10.1038/srep31972 Hippocampal brain sections from the α 3 +/D801Y mice revealed a large number of pyknotic nuclei within the dentate gyrus granule cell layer compared to WT littermates ( Fig. 7c-f), suggesting that the reduced number of granule layer neurons in the dentate gyrus of the α 3 +/D801Y mice was partly due to this.

Discussion
ATP1A3 mutations have been recognized in infants and children presenting with diverse neurological symptoms 5,6 . In AHC patients, some of the most devastating symptoms include bouts of hyperactivity that may cause patients to injure themselves accidently and epileptic seizures that are associated with SUDEP and worsening of cognitive impairments. A recent case reports suggest that RDP patients also suffer from cognitive impairments, albeit to a lesser degree as many patients are able to attend and finish high school 33 .
The highly variable nature of ATP1A3 diseases even for patients carrying the same mutation, has poised the theory that other factors such as genetic background, epigenetic modifications and environmental triggers influence the disease course.
Several studies show that α 1 expression is influenced by changes in [Na + ] 37 and [K + ] 38 , both are likely affected by reduced α 3 Na + /K + -ATPase activity. The compensatory upregulation of α 1 protein in response to the reduced α 3 protein expression in whole α 3 +/D801Y brain lysates is therefore to be expected. Similar observations have previously been reported for the α 3 +/KOI4 mice 23 . Homozygous α 3 D801Y/D801Y mice died at birth, suggesting that α 1 upregulation could not compensate for α 3 loss at this stage, This is in accordance with suggestions that α 1 and α 3 differ not only in substrate affinity but also in localisation 39 and that the CNS shifts from predominantly using the α 1 and α 2 isoforms during early development to α 1 and α 3 during post-natal development 26 .
In exploration-based tests for anxiety-like behavior, such as the OF and EPM, it can be difficult to dissociate symptoms of hyperactivity and attention deficits from anxiety-like behavior, as they may interfere with spontaneous exploratory locomotion. Although the α 3 +/D801Y mice spent significantly more time in the OF periphery, we propose this to be a direct consequence of the lack of meander rather than a trait of anxiety. In support, the α 3 +/D801Y did not discriminate between entering open and closed arms in the EPM and spent significantly more time than WT mice exploring the EPM open arms. In further support of this hypothesis, similar behavior was described for an attention deficit mouse model 40,41 . The hyperactive phenotype observed in the OF was induced response to handling and novelty in general. This was particularly evident during the memory tests, where repeated handling and the stressful environment caused some α 3 +/D801Y mice to become so agitated that they would jump off the testing platforms repeatedly and hurt themselves. These mice were omitted from the study. We believe this behavior could reflect in some degree the hyperactive and manic episodes of AHC patients.
Reduced Na + /K + -ATPase activity has previously been described in human epilepsy patients 20 . There is a growing appreciation that genetic factors contribute to the etiology of seizures 42 . With the recent case report of D801Y patient diagnosed with late onset mild AHC, it is possible that other symptoms of ATP1A3-related disorders are further affected by the genetic background.
The Myshkin and Mashl +/− mice were originally maintained in the 129S1/SvImJ and 129 SV background, respectively, and developed spontaneous tonic-clonic seizures and epileptic discharges as well as a SUDEP-like phenotype 26,29 .
Despite a reduced PTZ seizure threshold, we did not observe spontaneous seizures in the α 3 +/D801Y mice. Spontaneous seizures have also not been reported for the α 3 +/KOI4 and α 3 +/ΔE2-6 mice 23,24 . Supporting a strong contribution from genetic background, the latter two Atp1a3 mouse models were maintained in the seizure resistant C57BL6/J strain 43 . Given the close relationship to the C57BL/6JRJ strain, it is likely that some of the same genetic modifiers play a role in the seizure phenotype of the α 3 +/D801Y mice. In further support of this theory, the Myshkin mice became resistant to stress-induced seizures once maintained in the seizure-resistant C57BL/6NCr strain for 20 generations 27 .
Electrophysiological measurements in acute brain slices from naïve animals showed only minor differences between the genotypes. This cannot account for the decreased seizure threshold. It therefore seems unlikely that the increased excitability of the α 3 +/D801Y mice can be explained by changes in the basic membrane properties of the CA1 pyramidal neurons.
The hippocampus regulates the generation of long term memory 44 and spatial learning 45 and has been shown to be critical for spatial memory in human subjects with hippocampal damage (reviewed in 46 ). Together, the presented memory tests suggest an essential role of hippocampal α 3 Na+ /K+ -ATPase in consolidating particularly long-term spatial memory and fear-dependent memory, whereas short-term memory seemed less affected.
The Barnes maze test showed that the α 3 +/D801Y mice took significantly longer time to enter the target tunnel. This difference was not caused by an inability of the α 3 +/D801Y mice to locate the tunnel, but rather the fact that the α 3 +/D801Y mice spent upwards of four times longer at the entrance before entering. The Barnes maze test relies on the instinct of mice to escape a brightly lit area and to seek protection in a tunnel. The mouse is guided towards the tunnel via external cues mounted on the surrounding walls. The lack of discrimination between open and closed arms in the EPM would suggest that the instinct to seek cover is suppressed in the α 3 +/D801Y mice. Similar behavior has previously been described in the dopamine transporter knockout mouse, a mouse model for attention-deficit/hyperactivity disorder (ADHD) and schizophrenia-like behavior 40,41 .
We propose that the increased total latency is a consequence of failure to process the stressful surroundings of the Barnes maze, originally created to drive the mice into the tunnel.
Barnes maze track plot analysis revealed that the mice irrespective of genotype employed similar search strategies during the first 4 days of training: At day 5, both genotypes used the serial and direct strategies to an equal extent, suggesting that the α 3 +/D801Y mice have functional short-term spatial memory. The direct approach requires that the mice navigate using external cues, confirming that the vision of the α 3 +/D801Y mice was intact. At day 12, the α 3 +/D801Y mice no longer used the direct strategy and reverted to predominantly using the serial and to a lesser extent the mixed approach. This suggests that their long-term spatial memory is affected. Primary latencies were not affected by this change in tactics. We initially expected a compensatory effect of the hyperlocomotion, previously observed in OF, but average speed for both genotypes was similar (not shown).
The α 3 +/D801Y showed reduced fear memory in the passive avoidance test. It was clear that given the right environment, the mice were fully capable of entering a dark compartment. A likely explanation for poor performance of the α 3 +/D801Y in the Barnes maze is therefore that the stressful environment interfered with the decision-making of the α 3 +/D801Y mice. The reduced spatial learning and memory abilities in the α 3 +/D801Y mice suggested dysfunction of the amygdala and hippocampal brain regions. Histological examination revealed a large number of pyknotic nuclei within the granule layer of the dentate gyrus. Similar nuclear morphology has previously been described in ouabain-treated cultured cortical neurons undergoing hybrid cell death 47 . Furthermore, similar hippocampal pathology has previously been reported in rats injected with ouabain into the hippocampus 48 and in a rat model of pilocarpine-induced chronic epilepsy 49 , thus strengthening the link between Na + /K + -ATPases perturbations and seizure.
The α 3 isoform is highly expressed in the GABAergic basket cells in the subgranular zone 17 that are responsible for proliferation of granule cells during early development. It is therefore likely that the reduced number of granule cells in the dentate gyrus is the result of skewed apoptosis/proliferation in this region and that this is directly affected by the reduced α 3 activity.
Memory deficits were noted in the heterozygous knock-in mouse models Myshkin 50 and Mashl +/−29 as well as the heterozygous knock-out model α 3 +/KOI4 mice 23 , strongly supporting the role of the α 3 in hippocampus-dependent cognition. The α 3 +/KOI4 mice showed reduced expression of the N-methyl-D-aspartic acid receptor (NMDR) 51 . The NMDR has a well-documented role in the formation of several memories, including spatial, olfactory and contextual memory 52 . NMDA receptor expression was reduced by applying ouabain to cerebellar neurons 53 . Reduced NMDA receptor NR1 expression was described in homozygous E18 Myshkin mice, but not in heterozygous E18 and adult Myshkin mice 26 .
The co-expression of the α 3 isoform in GABAergic neurons suggests an association between the Na + / K + -ATPase and GABA transmission. The role of GABA A receptors in learning and memory and neurological Scientific RepoRts | 6:31972 | DOI: 10.1038/srep31972 disorders is well documented (recently reviewed 54 ). In particular, GABA regulates oscillations implicated in learning and memory, by generating synchronized inhibitory postsynaptic potentials. Dysfunctions caused by the α 3 isoform has previously been linked to GABA transmissions in the Myshkin mouse 26 and aberrant cerebellar function in the α 3 +/ΔE2-6 mice 24 . To test if increasing GABAergic transmission could improve learning, the α 3 mice was treated with the benzodiazepine, clonazepam. A single injection rescued passive avoidance performance and thus fear-dependent memory. Interestingly, similar effects were recently reported for the Scn1a +/− mouse model for Dravet's syndrome 36 , a disease where GABAergic neurotransmission is specifically impaired by a mutation in the SCN1A gene encoding voltage-gated sodium channel Na V 1.1. Interestingly, the Scn1a +/− mice also exhibited hyperactivity, and impaired context-dependent spatial memory. Supporting the hypothesis that GABA is indeed a major contributor towards ATP1A3-related diseases is the fact that GABA A receptors are implicated in childhood epilepsy 55 , and patients with temporal lobe epilepsy exhibit altered expression of the mRNA encoding the GABA A receptor in several hippocampal sub regions [56][57][58] . It was clear that PTZ, as a temporal lobe epileptic inducer, lowered the seizure threshold in the α 3 +/D801Y mice compared to WT mice, supporting the role of GABA. The effects of clonazepam are associated with allosteric activation of the ligand-gated GABA A receptor 59 . The current note of GABA A receptor complex subunits is that the GABA A α 5 subunit might be implicated in memory and learning, however, the GABA A subtype alteration in the α 3 +/D801Y mice remains to be elucidated. The results presented here strengthen the ongoing debate of the complexity of the ATP1A3-related diseases: Why some mutations are specific to RDP, AHC or CAPOS, and why the same mutations may produce intermediate symptoms could give rise to so very different disease courses. Given that the D801Y mutation has been shown to cause RDP and AHC in human patients, the α 3 +/D801Y mice may present as a unique platform to investigate this further. ATP1A3-related diseases have no effective treatments 5 . It is therefore interesting that the cognitive deficits in the α 3 +/D801Y mice could be reverted by a single low dose of clonazepam. This novel mouse model could be helpful for future developments of targeted treatments in neuropharmacology and memory functions 60 .
Animal ethics and conditions. All in vivo studies were performed using α 3 +/D801Y mice and WT obtained by crossing α 3 +/D801Y mice (generation N ≥ 8) with C57BL/6JRj mice (Janvier). Mice were kept at a daily 12 h light/dark cycle. Tests were performed during the light cycle. The mice were maintained as heterozygotes (α 3 +/D801Y ) since homozygotes were neonatally lethal. Mice used for all experimental procedures are between 10-20 weeks of age. In vitro electrophysiology. Preparation of brain slices. Male mice were anesthetized with isoflurane and decapitated. The brain was removed and quickly placed in dissection medium (in mM; 120 NaCl, 2 KCl, 1.25 KH 2 PO 4 , 6.6 HEPES acid, 2.6 NaHEPES, 20 NaHCO 3 , 2 CaCl 2 , 2 MgSO 4 and 10 D-glucose, bubbled with 95% O 2 and 5% CO 2 ) at 4 °C. The hippocampus was dissected free, and 400 μ m slices were cut using a McIlwain tissue chopper. One slice was immediately transferred to the recording chamber, where it was placed on a nylon-mesh grid at the interface between warm (31-32 °C) aCSF (in mM; 124 NaCl, 3.25 KCl, 1.25 NaH 2 PO 4 , 20 NaHCO 3 , 2 CaCl 2 , 2 MgSO 4 and 10 D-glucose, bubbled with 95% O 2 and 5% CO 2 , pH 7.3) and warm humidified gas (95% O 2 , 5% CO 2 ). Perfusion flow rate was 1 ml/min. The slice rested for at least one hour before electrophysiological recordings were started. The remaining slices were stored in dissection medium bubbled with 95% O 2 and 5% CO 2 at room temperature.
Electrophysiological recordings. Intracellular recordings were obtained using borosilicate glass electrodes (1.2 mm OD; Clark Electromedical, Pangbourne, UK) filled with 4 M K + acetate and placed in stratum pyramidale in area CA1. Conventional recording techniques were employed, using a high-input impedance amplifier (Axoclamp 2A, Molecular Devices, USA) with bridge balance and current injection facilities. Signals were digitized online using a Digidata 1440 interface and transferred to a computer for analysis employing pCLAMP (version 10, Molecular Devices). Inclusion criteria were a stable resting membrane potential (RMP) ≤ − 50 mV, a membrane input resistance (R in ) ≥ 10 MΩ and an action potential amplitude ≥ 70 mV.
Once an intracellular recording was established, a series of stimulation protocols were employed both at RMP and after the membrane potential was clamped at − 65 and − 70 mV.
Analysis. R in was evaluated from the current-voltage relationship close to RMP; the AP threshold was measured using short (4 ms) depolarizing current pulses of increasing intensity; the rates of rise and decay of the AP were taken as the maximal slopes; the frequency vs. current (f-I) relationship was estimated with 500 ms depolarizing current pulses of increasing intensity. Frequency accommodation was estimated as the variance of interspike duration during repetitive firing of 16-19 APs evoked by a 1 s depolarizing current pulse from a baseline potential of − 65 mV. The time-dependent decay in AP amplitude during 20 s repetitive firing was estimated using the following formula: (1 st AP -last AP)/1 st AP.
Unless otherwise indicated, values are given as mean ± S.E.M, and the unpaired Student's t-test or Mann-Whitney rank sum test were used for statistical evaluation. For multiple comparisons the two-way ANOVA was used. The level of significance was set at 5%.
Behavioral paradigms. Experiments were conducted blinded using α 3 +/D801Y mice and age-matched WT littermates. Mice were transferred to the test room one hour prior to testing for acclimation and tests were performed 1-2 days after last cage change. Behavioral apparatuses were cleaned between tests in 70% EtOH and only one gender was tested per experiment. Elevated plus maze. Entries into the open and closed arms of the elevated-plus maze (Stoelting Europe; Dublin, Ireland), time spent in these arms, as well as distance traveled was recorded for 10 minutes using the ANY-maze software (Stoelting Company).
Barnes Maze and passive avoidance were performed as recently described. Trials were recorded by using computerised 34 tracking/image analyser system and analysed using the ANY-maze tracking system (Stoelting Company). The following parameters were recorded: errors, distance from tunnel, search strategy and time that the mouse took to escape into the tunnel i.e. total latency. Errors were defined as nose pokes and head deflections over any hole that did not have the tunnel. The search strategies were determined by examining each mouse's daily session and defined in to three categories: (1) Direct (spatial): Moving directly to target hole or to an adjacent hole before visiting the target. (2) Mixed: Hole searches separated by crossing through the center of the maze or unorganised search. (3) Serial: The first visit to the target hole was preceded by visit at least two adjacent holes in serial manner, clockwise or counter clockwise direction 63 .
The passive avoidance test was initiated on the acquisition day (A). The mouse was placed in a brightly lit compartment with an electronically controlled door leading into a dark compartment. The latency (s) was recorded for the mouse to enter the dark compartment. Once in the dark compartment, the door closed and the mouse received an electric shock (0.42 mA for 1 s). Twenty-four hours later (retention day, R), the mouse was reintroduced to the same brightly lit compartment and the latency to enter the dark compartment was recorded as an indicator of memory of the shock.
Clonazepam passive avoidance rescue. Thirty minutes prior to passive avoidance training, the mice received 0.0625 mg/kg clonazepam intraperitoneally (Roche, Hvidovre, Denmark) dissolved in 0.9% sterile saline (vehicle) or vehicle alone.
PTZ seizure threshold. Mice were given 75 mg/kg pentylenetetrazole (Sigma-Aldrich; Schnelldorf, Germany) or 0.9% NaCl vehicle IP and monitored and video-recorded for 30 minutes after which they were euthanized.
Brain sampling and immunohistochemistry. Mice were deeply anesthetized with an overdose of pentobarbital. Approximately 0.05 mL per 10 g body weight pentobarbital was given intraperitonally (50 mg/mL pentobarbital, Aarhus University Hospital). When sedated, the mice were fixed upon a polystyrene board, and their chests were cut open with a blunt pair of scissors. The mice were perfused with 20 mL ice cold phosphate buffered solution (PBS) transcardially and subsequently with 20 mL ice cold 4% paraformaldehyde (PFA) in PBS. The brains were carefully dissected out and post-fixed in 4% PFA, PBS 4 °C over night (ON). The brains were cut in half following the midline and the olfactory bulb and cerebellum were dissected from the right hemispheres, and these halves were used for sampling. The left hemispheres were saved for later studies.
The tissues were infiltrated in paraffin using a Shandon Citadel TM Tissue Processor (Thermo Scientific). The right hemispheres were coronally sliced on a microtome at a thickness of 30 μ m. A stereotactic atlas of the mouse brain (Paxinos and Franklin, 2003, second edition) was used to identify a region juxtaposed to the hippocampus in order to have a visual guideline for initiating the collection of the sections. The point of reference chosen was the dorsal third ventricle at approximately Bregma − 0.22 mm that was approximately situated 0.70 mm frontally to the hippocampus. The hippocampus stretches from Bregma − 0.94 mm to − 3.88 mm according to the atlas. Every second section was collected on Superfrost ® Plus Microscope Slides (Thermo Scientific) spanning 4 series. One of the series was used for stereological analysis. Microscopic examination ensured that the hippocampus had been fully sectioned. After collection of the sections, the slides were put in the oven at 65 °C for 30 min. Sections were then deparaffinised Xylene (2 × 15 min), 99% EtOH (3 × 5 min), 96% EtOH (3 × 5 min) and 70% EtOH (2 × 5 min). Sections for stereological analysis were stained with toluidine blue, subsequently dehydrated and cover slipped. Sections for histological assessment were stained with Hoechst and cover slipped. Stereological analysis. The optical fractionator was used as counting methodology, and quantitative stereological analysis was performed by the same person who was blinded to the phenotype of the mice.
Equipment. Counting was done on a computerized optical microscope (Olympus BX50) equipped with a motorized stage and focus control system (Prior Scientific, ProScan ™ III). A highly specialized software program (Visiopharm integrator system version 4.5.1.324) was used for counting.
Sampling. Every second section was collected spanning 4 series in order to have at least six sections of the hippocampus in every series. Counting was only done on one of them giving a section sampling fraction, ssf, of 1/8. The final tissue thickness after shrinkage was determined to be approximately 25 μ m, which allowed the height of the disector to be 15 μ m with safeguard zones of 5 μ m at the top and 5 μ m at the bottom. The top of the tissue is excluded from the counting as sectioning can extract parts of cells close the sectioning plane 64 . The height of the disector is used to calculate the height sampling fraction, = − hsf h t / , Q where − t Q is the Q − -weighted mean section thickness which can be calculated as: An unbiased counting frame with an area, a = 76 μ m 2 , was superimposed on each field of view within the area of the granular layer of dentate gyrus. The step length was 100 μ m in both the x and y plane and hence the sampling area fraction was: In order to achieve a sample estimate, N, with a CE less than 0.1, the number of sections, the step-length and the area of the counting frame were dimensioned so that ~200 cells were counted per series.
Counting. The GrDG of each hippocampus present on a slide was delineated using a 4× objective at a final magnification of 135× . Subsequently, meander counting was performed using the 100× objective at a final magnification of 3366× . At each field of view, the microscope was slowly focused down from top to bottom of the section. All non-pyknotic neurons sampled by the 2D unbiased counting frame and located within the height of the optical disector were counted. Pyknotic cells were counted as a separate population of cells.
Statistical analysis. All data were shown as mean ± s.d. (or SEM) and statistical analyses were done using Graphpad Prism version 5.01 or 6.03 (GraphPad Software Inc, La Jolla, CA, USA or the R software (R Foundation for Statistical Computing, Vienna, Austria) 66 (Supplementary Table 1). The obtained male and female mice data were only pooled when this was statistically validated (P < 0.05), and all statistical tests and the P-values obtained are presented in Supplementary Table 1.
Data availability. The α 3 +/D801Y mouse model is available through a Material Transfer Agreement (MTA).