F U L L - L E NG TH OR I G I N A L R E S E ARCH

Mechanisms of increased hippocampal excitability in theMashl+/� mouse model of Na+/K+-ATPase dysfunction

Arsen S. Hunanyan1 | Ashley R. Helseth1 | Elie Abdelnour1 | Bassil Kherallah1 |

Monisha Sachdev1 | Leeyup Chung2 | Melanie Masoud1 | Jordan Richardson1 |

Qiang Li3,4,5 | J. Victor Nadler6 | Scott D. Moore3,4,5 | Mohamad A. Mikati1,2

1Division of Pediatric Neurology,Department of Pediatrics, Duke UniversityMedical Center, Durham, NC, USA2Department of Neurobiology, DukeUniversity School of Medicine, Durham,NC, USA3Durham Veterans Affairs MedicalCenter, Durham, NC, USA4Department of Psychiatry andBehavioral Sciences, Duke UniversityMedical Center, Durham, NC, USA5Veterans Affairs Mid-Atlantic RegionMental Illness Research, Education, andClinical Center, Durham, NC, USA6Department of Pharmacology andCancer Biology, Duke UniversityMedical Center, Durham, NC, USA

CorrespondenceMohamad A. Mikati, Department ofPediatrics, Children Health Center, DukeUniversity Medical Center, Durham, NC,USA.Email: mohamad.mikati@duke.edu

Funding informationThe Duke Institute for Brain Sciences,Grant/Award Number: 4510874; CureAHC Foundation, Donation Number:383-3697; Duke University, Grant/AwardNumber: 3912247, 4410161; DukeTranslational Research Institute, Grant/Award Number: SOMVP-2012

SummaryObjective: Na+/K+-ATPase dysfunction, primary (mutation) or secondary (energy

crisis, neurodegenerative disease) increases neuronal excitability in the brain. To

evaluate the mechanisms underlying such increased excitability we studied mice

carrying the D801N mutation, the most common mutation causing human disease,

specifically alternating hemiplegia of childhood (AHC) including epilepsy.

Because the gene is expressed in all neurons, particularly c-aminobutyric acid

(GABA)ergic interneurons, we hypothesized that the pathophysiology would

involve both pyramidal cells and interneurons and that fast-spiking interneurons,

which have increased firing rates, would be most vulnerable.

Methods: We performed extracellular recordings, as well as whole-cell patch

clamp recordings from pyramidal cells and interneurons, in the CA1 region on

hippocampal slices. We also performed immunohistochemistry from hippocampal

sections to count CA1 pyramidal cells as well as parvalbumin-positive interneu-

rons. In addition, we performed video—electroencephalography (EEG) recordings

from the dorsal hippocampal CA1 region.

Results: We observed that juvenile knock-in mice carrying the above mutation

reproduce the human phenotype of AHC. We then demonstrated in the CA1

region of these mice the following findings as compared to wild type: (1)

Increased number of spikes evoked by electrical stimulation of Schaffer collat-

erals; (2) equalization by bicuculline of the number of spikes induced by

Schaffer collateral stimulation; (3) reduced miniature, spontaneous, and evoked

inhibitory postsynaptic currents, but no change in excitatory postsynaptic cur-

rents; (4) robust action potential frequency adaptation in response to depolariz-

ing current injection in CA1 fast-spiking interneurons; and (5) no change in

the number of pyramidal cells, but reduced number of parvalbumin positive

interneurons.

Significance: Our data indicate that, in our genetic model of Atp1a3 mutation,

there is increased excitability and marked dysfunction in GABAergic inhibition.

This supports the performance of further investigations to determine if selective

expression of the mutation in GABAergic and or glutamatergic neurons is neces-

sary and sufficient to result in the behavioral phenotype.

Accepted: 10 May 2018

DOI: 10.1111/epi.14441

Epilepsia. 2018;1–14. wileyonlinelibrary.com/journal/epi Wiley Periodicals, Inc.© 2018 International League Against Epilepsy

| 1

KEYWORD S

alternating hemiplegia of childhood, ATP1A3, D801N, epilepsy, parvalbumin

1 | INTRODUCTION

Major functions of the Na+/K+-ATPase pump are to regu-late membrane excitability, Na+/K+ ionic gradients, synap-tic transmission, neurotransmitter reuptake, and cellvolume.1 Dysfunction of this pump is involved in manyneurologic disorders including Alzheimer’s disease, mooddisorders, rapid-onset dystonia-parkinsonism, spongiformencephalopathy, a-synucleinopathies, schizophrenia, epi-lepsy in animal models, epilepsy in humans, and alternatinghemiplegia of childhood (AHC).2–10

The ATP1A3 gene encodes for the Na+/K+ pump a3subunit, which is found exclusively in neurons, mostprominently in c-aminobutyric acid (GABA)ergic interneu-rons.7,11,12 One role of the a3 isoform is rapid reversal ofthe large transient increase of [Na+]i that occurs during sus-tained action potentials of cells. Thus, it is likely that dys-function of the a3 subunit affects primarily the function ofrapidly firing GABA interneurons.

Autosomal dominant mutations in the a3 isoform areassociated with a number of neurologic disorders, includingAHC.8,9 AHC manifests episodes of hemiplegia, dystonia,epileptic seizures, and sudden unexpected death in epi-lepsy.10,13 Epilepsy usually starts in childhood, and is oftenfocal and drug resistant.2,10,13–15 The most commonATP1A3 mutation that causes AHC in about one-third ofpatients is the D801N mutation.8–10,15

Despite the development of excellent mouse models ofAHC,1,4,5,16–19 the mechanisms leading to excitation/inhibi-tion imbalance at the cellular level due to Atp1a3 dysfunc-tion are still not fully understood. Recently, we created aknock-in mouse model carrying the D801N mutation anddemonstrated that adult heterozygous mice (Mashl+/�) strik-ingly manifest a phenotype recapitulating essentially all ofthe human AHC-related symptoms including epilepsy.5

Because epilepsy in AHC is first observed in child-hood,2,10,13–15 our aim in this study was to investigate theneural mechanisms of increased excitability in juvenileMashl+/� mice. Our hypothesis was that hippocampal sliceswill manifest increased excitability with reduced GABAer-gic inhibition. We addressed this question by performingelectrophysiology experiments in CA1 hippocampal neu-rons in P21-30 Mashl+/� mice.

2 | METHODS

All procedures were approved by Duke University Institu-tional Animal Care and Use Committee.

2.1 | Mice

Generation of Mashl+/� mutant mice, as described previ-ously, is detailed in the Supporting Information.

2.2 | Video-EEG recording

Video-electroencephalography (EEG) recordings were per-formed in Mashl+/� and wild-type (WT) age- and sex-matched littermates from both sexes at P30 (see Data S1for detailed description of recordings). No differences werefound in these or the other experiments described belowbetween the 2 sexes, so the data were combined.

2.3 | Slice electrophysiology

Recordings were performed from P21-P30 mice. Prepara-tion of hippocampal slices were done as described previ-ously5 with some modifications. Mice were deeplypreanesthetized with 3% isoflurane in an induction chamberand kept in a face mask with 2-4% isoflurane. After tran-scranial perfusion the brains were removed and hippocam-pal sagittal slices were cut and used for both extracellularand whole cell patch clamp recordings (see SupplementaryMaterial for detailed description of procedures). Werecorded from CA1 pyramidal neurons located in a deeplayer, for example, on the border of the stratum pyramidale

Key points

• D801N knock-in mutation of Na/K pumpinduced spontaneous seizures in juvenile mice

• Mashl+/� manifest increased excitability of CA1pyramidal cells

• Mashl+/� mice manifest reduced spontaneousinhibitory postsynaptic currents (sIPSCs), minia-ture inhibitory postsynaptic currents (mIPSCs),and evoked inhibitory postsynaptic currents(eIPSCs) but not excitatory postsynaptic currents(EPSCs) to CA1 pyramidal neurons compared towild-type littermates

• Mashl+/� mice CA1 PV+ interneurons showaction potential frequency adaptation comparedto wild-type littermates

• Mashl+/� mice have reduced number of PV+

interneurons in the CA1 region compared towild-type littermates

2 | HUNANYAN ET AL.

(SP) and stratum oriens (SO) as well as from interneuronslocated in CA1 stratum pyramidale-stratum oriens-alveus(SP-SO-alveus). We chose deep layer pyramidal cellsbecause it was recently found that these cells have more inhi-bitory connections from fast-spiking interneurons than thosein superficial layers.20 Fast-spiking and non–fast-spikinginterneurons were identified as previously defined accordingto their electrophysiological and firing properties.21–24

2.4 | Immunohistochemistry studies

We performed co-localization experiments by injecting bio-cytin and then staining for parvalbumin (PV). We also per-formed stereological cell counts of glutamatergic and PV+

neurons in CA1. We describe these methods in detail in theSupporting Information.

2.5 | Statistical analysis

For cumulative probability data analysis, we used the Kol-mogorov-Smirnov test. All other data analyses were per-formed using Student’s t-test (Excel, Microsoft, Redmond,WA, USA), one-way ANOVA, two-way ANOVA, or two-way RM ANOVA following the Tukey test (SigmaPlot 11.0,Systat Software, San Jose, CA, USA). Nominal data wereanalyzed using the chi-square or Fisher’s exact tests. Resultsare presented as mean � standard error of the mean (SEM)and considered statistically significant if P < .05.

3 | RESULTS

3.1 | In vivo model shows spontaneousrecurrent epileptic seizures, risk of suddenunexpected death of epilepsy, and other AHCmanifestations

In our previous study, we demonstrated that adultMashl+/� micemanifested the clinical features of AHC.5 In this study we docu-mented the occurrence of the same phenotype in juvenileMashl+/� mice including spontaneous recurrent seizures, dysto-nias, hemiplegias, andmyoclonus. These occurred spontaneouslyor upon stimulation. Inducing stimuli included exposure to roomtemperature water, vestibular stimulation, or even simple manip-ulations (see Video S1 for spontaneous seizure and Video S2 forhemiplegia). Upon observation of an average time of 38 hours, 7of 38Mashl+/�mice (15 female, 23 male) mice had seizures withan average duration of 1.04 minutes (range 5 seconds to 4 min-utes) and estimated seizure frequency was 0.42 seizures/animal/day. In addition, we observed that we could precipitate seizuresin any mutant mouse if we applied triggering stimuli such asmanipulation with vestibular stimulation or exposure to water. Of10 consecutive observed seizures precipitated by manipulation in10 different mice, 4 mice died during the seizures. Of 6 mutants

age P30 and older recorded with depth electrodes and video-monitoring, 4 had electrographic seizures, most of which (54/60) had epileptic behavioral correlates. None of the 5 wild-typehad any seizures. Forty-one of 60 observed seizures (meanduration 288.9 � 140 s/seizure, documented during total of222 hours), started either exclusively or concurrently from thehippocampus (Figure S1, Video S3). Eleven of those startedexclusively from the hippocampus and 10 others started exclu-sively from frontal cortex. Finally, we demonstrated that themortality rate in these mutant mice increased significantly ascompared to wild-type (P < .05) (Figure S2).

3.2 | Increased excitability of the CA1subregion of the hippocampus and effect ofbicuculline in Mashl+/� as compared towild-type

We found increased number of population spikes recorded fromCA1 stratum pyramidale to 1 Hz single-pulse (wild-type:1.5 � 0.18, n = 3 mice/8 slices vsMashl+/�: 2.5 � 0.24, n = 3mice/9 slices, P = .004, Figure 1A,E) as well as paired-pulsestimulus (100 msec interval, at first stimulation, wild-type:1.87 � 0.22, n = 3 mice/8 slices vs Mashl+/�: 3.33 � 0.33,n = 3 mice/9 slices, P = .003, at second stimulation, wild-type:2.5 � 0.32 vs Mashl+/�: 3.77 � 0.14, responses analyzed at30th pulse, P = .002, Figure 1C,F), but not to low frequency0.1 Hz (data not shown) as compared to wild-type. BecauseGABAA receptor activation has been shown to play a major rolein the regulation of network excitability, we then measured theabove responses in the presence of the GABAA receptor blockerbicuculline (10 lmol/L) and found that the number of spikes wasequalized in both wild-type and Mashl+/� genotypes in the pres-ence of bicuculline in both stimulation paradigm (for single pulse,P = .133, Figure 1B,E, for paired-pulse, P = .148, Figure 1D,F).

3.3 | No change in fiber volley amplitude andin the input-output curves between genotypes

Increased excitability in Mashl+/� vs wild-type mice hip-pocampal slices could be because of the difference of the fibervolley amplitude or field excitatory postsynaptic potential(fEPSP) slope in a given stimulus intensity. However, therewere no significant differences in the slopes of the fiber volleyvs stimulus intensity curves between genotypes nor were thereany significant differences in the input–output curves betweenthe 2 genotypes (Figure S3A-D).

3.4 | Pyramidal cells show someabnormalities in membrane properties inMashl+/� mice

To determine whether increased excitability in the hip-pocampus of Mashl+/� mice could be due to abnormal

HUNANYAN ET AL. | 3

membrane properties of pyramidal cells, we performedwhole-cell current clamp recordings. CA1 pyramidal cellswere identified by their pyramidal shape as well as by theiraction potential phenotype during increasing depolarization-

induced current pulses. We found that there were someelectrophysiologic differences between the 2 genotypes(Table S1). In Mashl+/� mice, pyramidal cells, restingmembrane potential was depolarized as compared to wild-

FIGURE 1 Increased excitability in Mashl+/� mice CA1 hippocampal slices. Representative traces of population spikes recorded from thehippocampal slices in the stratum pyramidale of the CA1 region evoked by electrical stimulation (1 Hz frequency) of Schaffer collaterals. A,Single-pulse stimulation of 1 Hz in wild-type and Mashl+/� mice. Note the presence of additional spike in mutant mouse marked by arrow. B,Same slices as in (A) in the presence of GABAA receptor inhibitor bicuculline (10 lmol/L). C, Paired-pulse stimulation (interstimulus interval:100 msec) induced increased number of spiking in Mashl+/� mice with both first and second stimulation. D, In the presence of GABAA receptorblocker bicuculline, the number of spikes was equal. E, Quantification of the number of spikes with single-pulse and paired-pulse (F). P1 and P2represent first and second stimulation induced responses, respectively. Arrowheads show stimulus artifact and arrows indicate additional spiking.Diagram shows the cut between CA3/CA1 and the position of the stimulation (stim.) and recording (rec.) electrodes. The traces are averageresponses from 30 traces. Data presented as mean � SEM, ** P < .01

4 | HUNANYAN ET AL.

type littermates (Mashl+/�: �63.6 � 0.6 mV, n = 16 mice/50 cells/37 slices; wild-type: �66.1 � 0.4 mV, n = 16mice/53 cells/33 slices; P = .002; Table S1). There was nostatistically significant difference between genotypes in theresponse to increasing currents of 10 pA increments, 1 sec-ond duration, 0 to +200 pA (Figure 2A; two-way RMANOVA, P = .995). On the other hand, afterhyperpolariza-tion amplitude (AHP) in Mashl+/� mice was significantlysmaller than that in wild-type mice (P = .005; Figure 2Ainset, 2B). It is important to note that Mashl+/� pyramidalcells showed double action potentials in 11 cells of 50 (6of 16 mice, Figure 2D1), whereas in wild-type mice nosuch cells were observed in 53 cells (16 mice, P = < .003Chi-square test, Figure 2C). Moreover, in contrast to wild-type mice 4 cells in mutants showed bursting action poten-tial (P = .037, Fisher’s exact test). Representative examplesof rheobase current induced normal in wild-type, doubleand bursting action potentials in Mashl+/� mice pyramidalcells, are shown in Figure 2C, D1, and D2, respectively.

3.5 | Reduced threshold for action potentialin CA1 pyramidal neurons in response toSchaffer collateral stimulation in Mashl+/� ascompared to wild-type

Because population spikes can be affected by the propertiesof pyramidal cells and the amplitude of the EPSP, in thenext experiments, we performed current clump recordingfrom pyramidal cells during stimulation of Schaffer collat-erals. Of interest, the threshold of stimulus intensity toinduce action potential was significantly less in Mashl+/�

mice (wild-type: 375 � 22.9 lA, n = 4 mice/8 cells vsMashl+/�: 289.1 � 32.2 lA, n = 4 mice/8 cells, P = .047,Figure 2E,F).

3.6 | Fast-spiking CA1 interneurons showreduced excitability due to marked actionpotential frequency adaptation in Mashl+/�

mice

CA1 interneurons are known to be important in the regula-tion of hippocampal excitability, and dysfunction of theseinterneurons has been involved in facilitation of seizures intemporal lobe epilepsy. GABAergic interneurons in the CA1area are highly diverse and they differ in their firing pat-terns.22 Because Atp1a3 is expressed highly in interneurons,we next performed whole cell recordings from CA1 SP-SO-alveus interneurons in hippocampal slices from wild-type lit-termate and Mashl+/� mice. Because our goal was to studyphysiologically defined types of interneurons that were clas-sified after patching based on their firing characteristics,21–24

we studied 146 interneurons (73 mutants and 73 in wild-type). Of these, 16 in mutants and 27 in wild-type were non–

fast-spiking nonaccommodating, 39 in mutants and 26 inwild-type were non–fast-spiking accommodating, and 18 inmutants and 20 in wild-type were fast-spiking. We founddecreased excitability due to marked accommodation in onlythe fast-spiking interneurons in mutants as compared to wild-type (Figures 3 and 4). Because we found differences in thefast-spiking interneurons we then sought to determine ifthose are PV positive (PV+) or not. So we injected biocytinand then stained for PV in a subgroup of slices and foundthat 9/9 (4 mice) mutant and 11/11 (5 mice) of wild-typefast-spiking interneurons were PV+. None of the non–fast-spiking interneurons stained with PV. This is consistent witha previously reported study that found that PV+ interneuronsare fast-spiking.25 To further validate our findings, werecorded from CA1 PV interneurons specifically expressingtdTomato as reporter in mice expressing tdTomato and theMashl+/� mutations. Consistent with above results we foundadaptation of fast-spiking PVtdTomato interneurons inMashl+/� mice hippocampal slices but not in wild-type tdTo-mato interneurons (Figure 4, WT n = 3 mice/5 cells/4 slices,Mashl+/� n = 2 mice/4 cells/4 slices). We also recordednon–fast-spiking SO interneurons (wild-type: n = 30 mice/53 cells; Mashl+/�: n = 26 mice/55 cells, Figure 3), and posthoc immunohistochemistry using fast-spiking marker PV didnot show any co-staining with biocytin. There were no sig-nificant differences in the number of action potentials duringrising stimulus current in those non-PV+ interneuronsbetween genotypes (Figure 3, Table S1). However, in non–fast-spiking accommodating interneurons the mean ampli-tude of action potential, measured at rheobase current, inMashl+/� mice (52.6 � 2 mV) was significantly smallercompared with wild-type (60.2 � 2 mV, P = .015). We didnot use other interneuron markers for identification of thesubtypes of the non-PV+ interneurons because in the hip-pocampal CA1 area there are at least 21 types of interneuronsand because doing so would constitute a separate study byitself.22–26

Taken together, these results demonstrate that inMashl+/� mice excitability of fast-spiking PV+ and non–fast-spiking accommodating interneurons is dramaticallydecreased.

3.7 | Excitatory postsynaptic inputs to CA1pyramidal neurons are not increased inMashl+/� mice

Increased excitability of the CA1 hippocampal region couldbe due to increased excitatory synaptic inputs on to theCA1 pyramidal neurons. To assess this hypothesis, werecorded spontaneous excitatory postsynaptic currents(sEPSC) and miniature excitatory postsynaptic currents(mEPSC) from the hippocampal CA1 pyramidal neurons ina voltage clamp mode. We found that neither amplitude

HUNANYAN ET AL. | 5

nor frequency of sEPSCs (amplitude, wild-type:14.9 � 0.9 pA, n = 6 mice/15 cells/15 slices vs Mashl+/�:13.6 � 0.7 pA, n = 5 mice/10 cells/10 slices, P = .949;frequency, wild-type: 0.72 � 0.09 Hz vs Mashl+/�:0.73 � 0.17 Hz, P = .964) and mEPSCs (amplitude, wild-type: 11 � 0.6 pA, n = 6 mice/15 cells/15 slices vsMashl+/�: 11.2 � 0.6 pA, n = 5 mice/10 cells/10 slices,P = .845; frequency, wild-type: 0.35 � 0.05 Hz vsMashl+/�: 0.32 � 0.1 Hz, P = .767) were increased inMashl+/� mice as compared to wild-type littermates

(Figure 5A-F). These results demonstrate that basal gluta-matergic neurotransmission to CA1 pyramidal cells are notaugmented in mutant mice.

3.8 | Inhibitory postsynaptic currents in CA1pyramidal neurons are markedly reduced inMashl+/� mice

To investigate if observed increases in excitability of areaCA1 could be due to decrease in GABA synaptic

FIGURE 2 Comparison of voltage responses to current steps, and of EPSPs in response to Schaffer Collaterals stimulation at 1 Hz, of CA1pyramidal cells in wild-type (WT) and Mashl+/� mice hippocampal slices. A, Depolarization (+ 200 pA) and hyperpolarization (�200 pA, 1 sduration) injected currents induced membrane potential changes. Note a similar pattern of action potential firing in both genotypes. Restingmembrane potential in both cells was �65 mV. Inset shows representative truncated single action potentials recorded from WT (blue trace) andMashl+/� (red trace) mice, showing a decrease in afterhyperpolarization amplitude (AHP) in Mashl+/� mice. Blue arrow shows the amplitude ofAHP in WT, whereas red arrow shows Mashl+/� AHP amplitude. The amplitude of AHP is calculated as the distance from dotted line (whereaction potential amplitude was calculated) to the deepest negative potential. B, Graph showing the difference in AHP between genotypes. Datapresented as mean � SEM, * P < .05. C, Action potential induced by depolarizing current injection in WT mice at threshold current. Arrowshows single action potential in WT pyramidal cell. D1, ~38% of Mashl+/� mice (11 cells of 50) show doublet firing induced by depolarizingcurrent injection. Arrows show dual action potentials in Mashl+/� pyramidal cells. D2, In Mashl+/� mice (4 cells of 50) show bursting firinginduced by depolarizing current injection. Arrowhead shows bursting action potential in Mashl+/� pyramidal cells. Traces in (C) and (D1, D2)obtained at rheobase currents. WT (n = 16 mice/53 cells), Mashl+/� (n = 16 mice/50 cells). E, Superimposed EPSPs during 50 lA and 400 lAstimulus intensity in WT mice and superimposed EPSPs during 50 lA and 150 lA stimulus intensity in Mashl+/� mice. F, Note that actionpotential was generated with lower stimulus intensity in Mashl+/� mice as compared to WT mice (*P < .05)

6 | HUNANYAN ET AL.

transmission, we recorded inhibitory postsynaptic currents(IPSCs) from CA1 pyramidal cells. The Kolmogorov-Smir-nov test revealed a significant difference in peak current ofevents (Figure 5G-I; P = .001). The mean of spontaneousinhibitory postsynaptic current (sIPSC) amplitude (Mashl+/�:38.8 � 3.9 pA, n = 12 mice /22 cells/22 slices vs wild-type: 53.6 � 4.3 pA, n = 10 mice/19 cells/19 slices;P = .016; Tukey test, Figure 5G-I) was significantly smal-ler than in wild-type mice, but the frequency of events wasnot statistically different (Mashl+/�: 4.7 � 0.3 Hz vs WT:5.5 � 0.3 Hz, P = .135, Figure 5I). There was no statisti-cally significant difference in sIPSCs rise time (10-90%)and decay time between genotypes (wild-type: rise time3.3 � 0.6 msec, Mashl+/�: rise time 3.4 � 0.4 msec,P = .83; wild-type: decay time 24 � 1.8 msec, Mashl+/�:decay time 22 � 2.1 msec, P = .34). Next, we recordedminiature inhibitory postsynaptic currents (mIPSCs) and

found that in Mashl+/� mice hippocampal slices, the meanof mIPSC amplitude (Mashl+/�: 25.2 � 2.4 pA, n = 7mice/13 cells/13 slices vs wild-type: 40.8 � 4.5 pA, n = 5mice/13 cells/13 slices; P = .006; Figure 5J-L) and the fre-quency of events (Mashl+/�: 3.9 � 0.3 Hz vs WT:4.7 � 0.2 Hz, P = .048) were significantly smaller than inwild-type mice. To ensure that mIPSCs were induced byactivation of GABAA receptors, bicuculline (10 lmol/L)was added to the recording chamber. Bicuculline blockedall recorded events (data not presented).

We further examined the disinhibition of CA1 pyramidalneurons by electrically eliciting GABA input onto the CA1pyramidal neurons. The stimulation electrode was positionedin the stratum oriens 100 lm away from the recorded CA1cell and monosynaptic GABA receptor mediated IPSCs wereisolated in the presence of CNQX (20 lmol/L) and AP5(50 lmol/L). Stimulation of stratum oriens induced evoked

FIGURE 3 Comparison of voltage responses to current steps of CA1 stratum oriens non–fast-spiking (FS) accommodating and non-FSnonaccommodating interneurons in wild-type littermate (WT) and Mashl+/� mice hippocampal slices. A, Representative traces showing actionpotential firing pattern in a non-FS accommodating interneuron in WT (n = 14 mice/26 cells blue traces) and Mashl+/� mice (n = 15 mice/39cells, red traces). B, No difference found in the number of action potentials in response to current injections recorded in non-FS accommodatinginterneurons between genotype (P = .475, two-way RM ANOVA, Tukey test). C, Representative traces showing action potential firing pattern ina non-FS nonaccommodating interneuron in WT (n = 16 mice/27 cells, blue traces) and Mashl+/� mice (n = 11 mice/16 cells, red traces). D,Number of action potentials in response to current injections recorded in non-FS nonaccommodating interneurons were not significantly differentbetween genotype (P = .810, two-way RM ANOVA, Tukey test). Presented representative responses were evoked by hyperpolarization anddepolarization (�200 pA, +100 pA, 200 pA, 450 pA with 1 s duration) injected currents. Insets are superimposed action potentials recorded fromWT (blue traces) and from Mashl+/� mice (red traces). E, Confocal image of PV immunostaining (green) of resectioned hippocampal slice afterelectrophysiologic recordings of 3 neurons in CA1 region loaded with biocytin (red), which showed non–fast-spiking phenotype. F, Highermagnification of selected region of the image in (E). Scale bar in both (E and F) is 50 lm. Data presented as mean � SEM

HUNANYAN ET AL. | 7

inhibitory postsynaptic currents (eIPSCs) with threshold cur-rents that were usually 20-30 lA and were not differentbetween genotypes (n = 6 mice for each genotype, P > .05).The mean amplitude of eIPSCs in Mashl+/� mice wasdepressed starting from the 30th pulse of 1 Hz stimulation(Figure 5M,N). These results show decreased GABA trans-mission from SO interneurons to CA1 pyramidal neuronsduring repetitive stimulation, and that there is abnormalshort-term plasticity mediated by inhibitory synaptic trans-mission in the hippocampus.

3.9 | Immunohistochemistry

Immunohistochemistry was performed on dorsal hippocam-pal slices from Mashl+/� and WT littermates to count thenumber of CA1 pyramidal cells using neurogranin (NG) asa marker. The number of NG+ cells was not differentbetween genotypes (wild-type: 20937 � 1537, n = 5 micevs Mashl+/�: 24122 � 1633, n = 3 mice, P = .22, Fig-ure 6A,C). We also quantitated the number of these PV+

interneurons in area CA1. A majority of PV+ cells were

FIGURE 4 Comparison of voltage responses to current steps of CA1 parvalbumin positive (PV+) fast-spiking interneurons in wild-typelittermate (WT) and Mashl+/� mice hippocampal slices using whole-cell current-clamp recordings. A, Representative traces showing actionpotential firing pattern in a PV+ fast-spiking interneuron in WT (n = 14 mice/20 cells, blue traces) and (B) Mashl+/� mice (n = 15 mice/18 cells,red traces) in response to incremental steps to current. Presented responses were evoked by depolarization (100 pA, 200 pA, 300 pA and 450pA, 1 s duration) injected currents. Note that in contrast to WT, fast-spiking interneurons in Mashl+/� mice show a strong firing frequencyadaptation with depolarizing current injection. Inset shows superimposed action potentials recorded from WT (blue traces) and from Mashl+/�

mice (red traces). C, The mean number of action potentials in response to each current step. D, Confocal image shows double immunostainingwith biocytin (red) and parvalbumin (PV, green) after electrophysiologic recording. Scale bar, 20 lm. E, Comparison of voltage responses tocurrent steps of CA1 parvalbumin-tdTomato interneurons in wild-type (WT: n = 3 mice/5 cells/4 slices) and Mashl+/� mice (n = 2 mice/4 cells/4slices) hippocampal slices (F). Note that, in contrast to WT, PVtdTomato mouse Mashl+/�, PVtdTomato mouse interneuron shows strongadaptation of action potential frequency. Inset shows example of fluorescence image of expression of tdTomato in CA1 PV interneurons takenfrom WT mouse. SO - stratum oriens, SP - stratum pyramidale, SR-stratum radiatum. Data presented as mean � SEM, ** P < .01, *** P < .001

8 | HUNANYAN ET AL.

located in the stratum oriens and stratum pyramidale in hip-pocampal slices prepared from both mutant mice and wild-type littermates. Stereological quantification revealedreduced number of PV+ cells in Mashl+/� compared towild-type littermates (Mashl+/�: 3826 � 347, n = 6 mice;wild-type: 6231 � 285; n = 6 mice, P < = .001; Fig-ure 6B,D). These results suggest that the Atp1a3 mutationalso caused a reduction in PV+ interneurons in the dorsalCA1 area.

4 | DISCUSSION

In this study, we found that the D801N mutation results inspontaneous recurrent seizures, increased excitability inCA1 hippocampal slices, marked impairment in GABAinhibition, PV+ cell loss, and action potential frequencyadaptation in fast-spiking PV+ interneurons.

4.1 | Increased excitability in extracellularrecordings and role of GABA

Our findings of increased excitability in juvenile Mashl+/�

mice extend our prior findings in Mashl+/� adult mice5 andthe findings of increased excitability in other AHC mod-els.4,18 As in adults, we found in this study on juvenilesincreased spiking as a result of fast, 1 Hz, but not slow,0.1 Hz stimulation. In the present study, we also investi-gated the effects of GABA receptor blockade and found asimilar degree of spiking in juvenile mutants and wild-typemice in the presence of the GABAA receptor antagonistbicuculline. This is consistent with the hypothesis thatGABA dysfunction is a major cause of increased excitabil-ity in our model. Of note is that in mice carrying theI810N mutation, valproic acid reduces the severity of sei-zures,4 and in those with the D801Y mutation, clonazepamcorrects some of the behavioral changes.18

4.2 | Increased excitability in thehippocampal CA1 pyramidal cells

Depolarization of resting membrane potential as well asdoublets and bursting action potentials in some pyramidalcells can significantly contribute to overall increasedexcitability in Mashl+/� mice hippocampal slices. Theseobserved abnormalities are most probably not due to adirect effect of a3 mutation on pyramidal cells becauseexpression of the a3 subunit is lower in pyramidal neu-rons12 than in GABAergic interneurons where we did notfind similar abnormalities. Given our findings of reducedinhibition, some of the observed changes in pyramidal cellproperties, thus, may be secondary to changes in GABAer-gic tonic inhibition onto these neurons. Other explanations

of the changes in pyramidal cells could be due to changesin [K+]o or K+ conductance. In contrast to ourmouse model, Holm et al. found that in D801Y mice,which have a less severe phenotype of AHC than ourmodel, there are only minor changes in CA1 pyramidal cellelectrophysiology.18

No change in EPSP amplitude and reduction in IPSPssuggest that GABAergic inhibition to pyramidal neurons isparticularly compromised in the mutants. CA1 pyramidalcells receive massive GABAergic inputs from fast-spikingPV+ interneurons, which control the output of local net-works by inhibiting action potential generation in principalcells.26,27

4.3 | Markedly reduced GABAergicinhibition from fast-spiking interneurons

To our surprise, in Mashl+/� mice, fast-spiking CA1interneurons exhibited marked action potential frequencyadaptation in response to depolarizing stimuli, indicatingreduced excitability of these interneurons. CA1 interneu-rons are activated by recurrent collaterals of pyramidal cellsand therefore have a significant role in controllingexcitability in area CA1.28 Because fast-spiking interneu-rons fire action potentials with higher frequency, these cellswould require rapid restoration of Na+ and K+ ion gradi-ents across the plasma membrane, which is one of themajor roles of Atp1a3. Dysfunction of the Na/K pump dueto mutation and/or due to secondary causes, such as ATPdepletion resulting from ongoing seizure activity, may leadto the loss of ionic gradient across the membrane, and canultimately lead to failure of action potential generation.Indeed, the Na+/K+-ATPase current density (or its activity)is known to be 3-to 7-fold larger in fast-spiking interneu-rons than in cortical pyramidal neurons.29 In addition,increases in K+ concentration due to seizure activity couldfurther impair mutated pump function resulting depolariza-tion block of fast-spiking interneurons.30,31 An additionalpotential could involve upregulation of calcium-dependentpotassium currents in these interneurons.

Our recordings also revealed marked reductions in IPSCfrequency and amplitude. This suggests both pre- and post-synaptic changes in GABA synapses. The Na+ gradientgenerated by Na+/K+-ATPase is important as a drivingforce for neurotransmitter uptake.1 The Atp1a3 mutationcould increase GABA concentration in the synaptic cleftby reducing uptake. This sustained increase of GABAcould potentially lead to internalization or desensitizationof postsynaptic GABAA receptors. Consistent with ourfindings of abnormal firing in GABAergic neurons is that2 other models of Atp1a3 mutations manifest abnormalitiesin cerebellar GABAergic neuronal firing.16,30 Whereas thetypes of observed abnormalities differ among the 3 models,

HUNANYAN ET AL. | 9

this is likely to be due to differences in the type of cellsstudied in the ages investigated, and to the likely differenteffects of different mutations. For example, the D801Nmutation shows a dominant negative effect when expressed

in Xenopus laevis oocytes,32 whereas knocking out thegene is not expected to do so.

We also found that the eIPSC amplitudes weredepressed during repetitive stimulation at SO and only after

10 | HUNANYAN ET AL.

the 30th pulse. This could be because in the resting statethere is a not much change in the balance of excitatory andinhibitory synaptic responses from SO, whereas duringrepetitive stimulation there is robust excitation of SOinterneurons, which leads to depression of action potentialfiring and ultimately an increased excitability of pyramidalcells. This is also consistent with the function of the a3subunit being important under conditions of increased fir-ing. Other histological subtypes of interneurons not investi-gated in this study may be involved too and should be thesubject of future investigations.

We found reduction of PV+ cell numbers in ourmice. The cause may be prior seizure activity, as loss ofPV+ cells has been reported in chronic epilepsy includ-ing in the hippocampus.33 Alternatively, the reductionmay be related to other effects of the Atp1a3 mutation.Dysfunction of the Na/K pump can lead to secondarycalcium overload through reversed Na+/Ca2+ exchange.17

Interneurons that normally highly express the a3 subunitand that are fast-spiking, such as PV+ interneurons,would then be the most vulnerable to potential Ca2+-mediated toxicity. In addition, Na+/K+-ATPase is also asignal transducer that modulates cell growth, adhesion,and apoptotic threshold.34 Consequently, dysfunction inthis molecule can potentially result in cell death andimpaired neuronal migration.35,36

Our findings of reduced GABAA inhibition in a geneticmodel of epilepsy are similar to findings in some other epi-lepsy models. Dravet syndrome epileptic mice exhibitaction potential frequency adaptation in PV+ interneu-rons,37 impaired GABAergic neurotransmission,38 andreduction in IPSC frequency, but also have increased AP

threshold and rheobase not manifested in our mice.23,37,38

Finally, PV+ interneurons are important in other types ofinduced epilepsy models.39,40

4.4 | Potential implications of our findings

Our findings raise the possibility that neurophysiologic dys-function in other areas of the brain, similar to what wefound in the hippocampus, may be responsible for the otherAHC manifestations.16,31,41,42 Although spreading depres-sion, the presumed mechanism of hemiplegias in AHC,5

has been traditionally related to glutamatergic mechanisms;there is also strong evidence that loss of GABAergic inhi-bition is important too.43 Our findings are also in line withemerging evidence to support the “interneuron energyhypothesis.”44 Taking into consideration the complexity ofneuronal circuits and especially the great heterogeneity ofGABAergic interneurons, experiments using conditionalmice allowing the introduction of the Atp1a3 mutation inspecific types of neurons should be a useful approach todissect further the AHC- and Atp1a3-related epilepsypathophysiology.

ACKNOWLEDGMENTS

This work was supported by grants from The Duke Insti-tute for Brain Sciences (DIBS) Fund number 4510874,Duke Fund numbers 4410161 and 3912247 the DukeTranslational Research Institute (DTRI) Grant SOMVP-2012 (MAM) and donation from the CureAHC Foundation.We would like to deeply thank all members of Dr. McNa-mara’s laboratory, including Dr. James McNamara, Wei-

FIGURE 5 Comparison of excitatory and inhibitory postsynaptic currents recorded from CA1 pyramidal neurons in wild-type littermates(WT) and Mashl+/� mice hippocampal slices. A, Example traces of spontaneous excitatory postsynaptic current (sEPSC) recorded in Mashl+/�

and in WT mice slices. B, Average values of sEPSC amplitude and frequency. C, Example traces of miniature excitatory postsynaptic current(mEPSC) recorded in Mashl+/� and WT mice. D, Average values of mEPSC amplitude and frequency. WT (n = 6 mice/15 cells/15 slices),Mashl+/� (n = 5 mice/10 cells/ 10 slices). E, Cumulative probability curve of sEPSC current amplitude (P = .58, Kolmogorov-Smirnov test). F,Cumulative probability curve of mEPSC current (P = .99, Kolmogorov-Smirnov test). G, Representative spontaneous inhibitory postsynapticcurrent (sIPSC) recordings in WT (n = 10 mice/19 cells/19 slices) and Mashl+/� mice (n = 12 mice/22 cells/22 slices) recorded in a whole-cellvoltage-clamp mode. H, Cumulative probability curve of current of sIPSC. Kolmogorov-Smirnov test revealed significant difference betweengenotypes (P = .001). The mean amplitude in Mashl+/� mice (38.8 � 3.9 pA) was significantly smaller than in WT (53.6 � 4.3 pA, P = .016).I, Cumulative probability curve of interevent intervals of sIPSCs. The mean frequency was not different between genotypes (WT: 5.5 � 0.3 Hz;Mashl+/�: 4.7 � 0.3 Hz, P = .13). J, Representative miniature inhibitory postsynaptic current (mIPSC) recordings in WT (n = 5 mice/13 cells/13slices) and Mashl+/� mice (n = 7 mice/13 cells/13 slices) recorded in a whole-cell voltage-clamp mode. K, Cumulative probability curve ofmIPSC current is shifted to the left in Mashl+/� mice (P < .001, Kolmogorov-Smirnov test). The mean amplitude in Mashl+/� mice (25.2 � 2.4pA) was significantly smaller than in WT (40.8 � 4.5 pA, P = .006). L, Cumulative probability curve of mIPSC interevent intervals is shifted tothe right in Mashl+/� mice (P < .001, Kolmogorov-Smirnov test). The mean frequency was significantly smaller in Mashl+/� mice (WT:4.7 � 0.2 Hz; Mashl+/�: 3.9 � 0.3 Hz, P = .048). Data are presented as mean � SEM, * P < .05, ** P < .01. M, Evoked inhibitorypostsynaptic currents (eIPSCs) during 1 Hz frequency stimulation of stratum oriens with increasing stimulus intensity. N, Representative eIPSCtraces from WT (blue trace) and Mashl+/� mice (red trace). Diagram shows the position of stimulation (stim.) and recording (rec.) electrodes in ahippocampal slice. Arrows on the traces indicate stimulus artifacts. Solid traces are responses evoked by first pulse while dotted traces areresponses evoked by 60th pulse. Data presented as mean � SEM, * P < .05

HUNANYAN ET AL. | 11

Hua Qian, Dr. Ram S. Puranam, Dr. Enhui Pan, and Dr.Xiao-Ping He for their technical expertise and support. Inparticular, we would like to thank Dr. Ram S. Puranamand Dr. Enhui Pan for discussion and their suggestionsregarding aspects of the experimental design. We deeplythank Drs. James McNamara and Dwight Koeberl for theirvaluable discussions and support. We deeply thank Dr.Daryl W. Hochman for his help and design of experimentsand other generous support. We also thank Dr. Hana N.Dawson for support and training for stereological quantifi-cation of cell numbers. We would also like to thank Dr.David Goldstein and his lab for the performance of Sangersequencing in their laboratory.

DISCLOSURE

None of authors has any conflict of interest to disclose. Weconfirm that we have read the Journal’s position on issues

involved in ethical publication and affirm that this report isconsistent with those guidelines.

ORCID

Mohamad A. Mikati http://orcid.org/0000-0003-0363-8715

REFERENCES

1. Moseley AE, Williams MT, Schaefer TL, et al. Deficiency in Na,K-ATPase alpha isoform genes alters spatial learning, motoractivity, and anxiety in mice. J Neurosci. 2007;27:616–26.

2. Masoud M, Prange L, Wuchich J, et al. Diagnosis and treatmentof alternating hemiplegia of childhood. Curr Treat Options Neu-rol. 2017;19:8.

3. Haglund MM, Schwartzkroin PA. Role of Na-K pump potassiumregulation and IPSPs in seizures and spreading depression in

A

B

C D

FIGURE 6 Representative images of CA1 hippocampal pyramidal cells and parvalbumin GABAergic interneurons immunoreactivity. A,Immunohistochemical staining with pyramidal cell marker neurogranin (NG) in WT (n = 5 mice) and Mashl+/� (n = 3 mice) hippocampalsections. B, Immunohistochemical staining with parvalbumin (PV) in WT and Mashl+/� hippocampal sections (n = 6 mice per genotype). PartsA and B photomicrographs on left were taken with 49 magnification (black scale bar = 500 lm); photomicrographs on right were taken with609 (for A) and 209 (for B) magnification (white scale bar = 60 lm). C, D, Quantification of total cell count with each marker in CA1 regionusing stereological analysis. Note that Mashl+/� mice have ~39% less PV+ cells than WT (***P < = .001). Error bars represent mean � SEM

12 | HUNANYAN ET AL.

immature rabbit hippocampal slices. J Neurophysiol.1990;63:225–39.

4. Clapcote SJ, Duffy S, Xie G, et al. Mutation I810N in the alpha3 iso-form of Na+, K + -ATPase causes impairments in the sodium pumpandhyperexcitability in theCNS.PNAS.2009;106:14085–90.

5. Hunanyan AS, Fainberg NA, Linabarger M, et al. Knock-inmouse model of alternating hemiplegia of childhood: behavioraland electrophysiologic characterization. Epilepsia. 2015;56:82–93.

6. Grisar T, Guillaume D, Delgado-Escueta AV. Contribution ofNa+, K(+)-ATPase to focal epilepsy: a brief review. EpilepsyRes. 1992;12:141–9.

7. Paciorkowski AR, McDaniel SS, Jansen LA, et al. Novel muta-tions in ATP1A3 associated with catastrophic early life epilepsy,episodic prolonged apnea, and postnatal microcephaly. Epilepsia.2015;56:422–30.

8. Heinzen EL, Swoboda KJ, Hitomi Y, et al. De novo mutations inATP1A3 cause alternating hemiplegia of childhood. Nat Genet.2012;44:1030–4.

9. Rosewich H, Thiele H, Ohlenbusch A, et al. Heterozygous de-novo mutations in ATP1A3 in patients with alternating hemiple-gia of childhood: a whole-exome sequencing gene-identificationstudy. Lancet Neurol. 2012;11:764–73.

10. Panagiotakaki E, De Grandis E, Stagnaro M, et al. Clinical pro-file of patients with ATP1A3 mutations in Alternating Hemiple-gia of Childhood-a study of 155 patients. Orphanet J Rare Dis.2015;10:123.

11. McGrail KM, Phillips JM, Sweadner KJ. Immunofluorescentlocalization of three Na, K-ATPase isozymes in the rat centralnervous system: both neurons and glia can express more than oneNa, K ATPase. J Neurosci. 1991;11:381–91.

12. Bøttger P, Tracz Z, Heuck A, et al. Distribution of Na/K-ATPasealpha 3 isoform, a sodium potassium P-type pump associatedwith rapid-onset of dystonia parkinsonism (RDP) in the adultmouse brain. J Comp Neurol. 2011;519:376–404.

13. Mikati MA, Maguire H, Barlow CF, et al. A syndrome of autoso-mal dominant alternating hemiplegia: clinical presentation mim-icking intractable epilepsy; chromosomal studies; and physiologicinvestigations. Neurology. 1992;42:2251–7.

14. Mikati MA, Kramer U, Zupanc ML, et al. Alternating hemiplegiaof childhood: clinical manifestations and long-term outcome.Pediatr Neurol. 2000;23:134–41.

15. Sasaki M, Ishii A, Saito Y, et al. Genotype-phenotype correla-tions in alternating hemiplegia of childhood. Neurology.2014;82:482–90.

16. Ikeda K, Satake S, Onaka T, et al. Enhanced inhibitory neuro-transmission in the cerebellar cortex of Atp1a3-deficient heterozy-gous mice. J Physiol. 2013;591:3433–49.

17. Kirshenbaum GS, Clapcote SJ, Duffy S, et al. Mania-like behav-ior induced by genetic dysfunction of the neuron-specific Na+,K + -ATPase alpha3 sodium pump. PNAS. 2011;108:18144–9.

18. Holm TH, Isaksen TJ, Glerup S, et al. Cognitive deficits causedby a disease-mutation in the alpha3 Na(+)/K(+)-ATPase isoform.Sci Rep. 2016;6:31972.

19. Holm TH, Lykke-Hartmann K. Insights into the pathology of thealpha3 Na(+)/K(+)-ATPase Ion pump in neurological disorders;lessons from animal models. Front Physiol. 2016;7:209.

20. Lee SH, Marchionni I, Bezaire M, et al. Parvalbumin-positivebasket cells differentiate among hippocampal pyramidal cells.Neuron. 2014;82:1129–44.

21. Kawaguchi Y, Katsumaru H, Kosaka T, et al. Fast spiking cellsin rat hippocampus (CA1 region) contain the calcium-bindingprotein parvalbumin. Brain Res. 1987;416:369–74.

22. Freund TF, Buzs�aki G. Interneurons of the hippocampus. Hip-pocampus. 1996;6:347–470.

23. Rubinstein M, Han S, Tai C, et al. Dissecting the phenotypes ofDravet syndrome by gene deletion. Brain. 2013;138:2219–33.

24. Lacaille JC, Williams S. Membrane properties of interneurons instratum oriens-alveus of the CA1 region of rat hippocampusin vitro. Neuroscience. 1990;36:349–59.

25. Kaiser T, Ting JT, Monteiro P, et al. Transgenic labeling of par-valbumin-expressing neurons with tdTomato. Neuroscience.2016;321:236–45.

26. Klausberger T, Somogyi P. Neuronal diversity and temporaldynamics: the unity of hippocampal circuit operations. Science.2008;321:53–7.

27. Bezaire MJ, Soltesz I. Quantitative assessment of CA1 local cir-cuits: knowledge base for interneuron-pyramidal cell connectivity.Hippocampus. 2013;23:751–85.

28. Blasco-Ibanez JM, Freund TF. Synaptic input of horizontal interneu-rons in stratum oriens of the hippocampal CA1 subfield: structuralbasisof feed-backactivation.Eur JNeurosci. 1995;7:2170–80.

29. Anderson TR, Huguenard JR, Prince DA. Differential effects ofNa+-K+ ATPase blockade on cortical layer V neurons. J Physiol.2010;588:4401–14.

30. Isaksen TJ, Kros L, Vedovato N, et al. Hypothermia-induced dys-tonia and abnormal cerebellar activity in a mouse model with asingle disease-mutation in the sodium pump. PLoS Genet.2017;13:e1006763.

31. Shin DS, Yu W, Fawcett A, et al. Characterizing the persistentCA3 interneuronal spiking activity in elevated extracellular potas-sium in the young rat hippocampus. Brain Res. 2010;1331:39–50.

32. Li M, Jazayeri D, Corry B, et al. A functional correlate of sever-ity in alternating hemiplegia of childhood. Neurobiol Dis.2015;77:88–93.

33. Andrioli A, Alonso-Nanclares L, Arellano JI, et al. Quantitativeanalysis of parvalbumin immunoreactive cells in the humanepileptic hippocampus. Neuroscience. 2007;149:131–43.

34. Aperia A. New roles for an old enzyme: Na,K-ATPase emergesas an interesting drug target. J Intern Med. 2007;261:44–52.

35. Kulikov A, Eva A, Kirch U, et al. Ouabain activates signalingpathways associated with cell death in human neuroblastoma.Biochim Biophys Acta. 2007;1768:1691–702.

36. Shin HK, Ryu BJ, Choi SW, et al. Inactivation of Src-to-ezrinpathway: a possible mechanism in the ouabain-mediated inhibi-tion of A549 cell migration. Biomed Res Int. 2015;2015:537136.

37. Han S, Tai C, Westenbroek RE, et al. Autistic-like behaviour inScn1a+/- mice and rescue by enhanced GABA-mediated neuro-transmission. Nature. 2012;489:385–90.

38. Tai C, Abe Y, Westenbroek RE, et al. Impaired excitability ofsomatostatin- and parvalbumin expressing cortical interneurons ina mouse model of Dravet syndrome. Proc Natl Acad Sci USA.2014;111:E3139–48.

39. Cammarota M, Losi G, Chiavegato A, et al. Fast spikinginterneuron control of seizure propagation in a cortical slicemodel of focal epilepsy. J Physiol. 2013;591:807–22.

40. Khoshkhoo S, Vogt D, Sohal VS. Dynamic, Cell-Type-SpecificRoles for GABAergic Interneurons in a Mouse Model of Optoge-netically Inducible Seizures. Neuron. 2017;93:291–8.

HUNANYAN ET AL. | 13

41. Gittis AH, Leventhal DK, Fensterheim BA, et al. Selective inhi-bition of striatal fast-spiking interneurons causes dyskinesias. JNeurosci. 2011;31:15727–31.

42. Calderon DP, Fremont R, Kraenzlin F, et al. The neural sub-strates of rapid-onset Dystonia Parkinsonism. Nat Neurosci.2011;14:357–65.

43. Wang M, Li Y, Lin Y. GABAA receptor a2 subtype activationsuppresses retinal spreading depression. Neuroscience.2015;298:137–44.

44. Kann O, Papageorgiou IE, Draguhn A. Highly energized inhibi-tory interneurons are a central element for information processingin cortical networks. J Cereb Blood Flow Metab. 2014;34:1270–82.

SUPPORTING INFORMATION

Additional supporting information may be found online inthe Supporting Information section at the end of the article.

How to cite this article: Hunanyan AS, Helseth AR,Abdelnour E, et al. Mechanisms of increasedhippocampal excitability in the Mashl+/� mousemodel of Na+/K+-ATPase dysfunction. Epilepsia.2018;00:1–14. https://doi.org/10.1111/epi.14441

14 | HUNANYAN ET AL.