Rescue of homeostatic regulation of striatalexcitability and locomotor activity in a mousemodel of Huntington’s diseaseYumei Cao1, David Bartolomé-Martín1, Naama Rotem, Carlos Rozas, Shlomo S. Dellal, Marcelo A. Chacon,Bashkim Kadriu, Maria Gulinello, Kamran Khodakhah, and Donald S. Faber2

Department of Neuroscience, Albert Einstein College of Medicine, Bronx, NY 10461

Edited* by Michael V. L. Bennett, Albert Einstein College of Medicine, Bronx, NY, and approved January 7, 2015 (received for review March 28, 2014)

We describe a fast activity-dependent homeostatic regulation ofintrinsic excitability of identified neurons in mouse dorsal stria-tum, the striatal output neurons. It can be induced by brief burstsof activity, is expressed on a time scale of seconds, limits repetitivefiring, and can convert regular firing patterns to irregular ones. Weshow it is due to progressive recruitment of the KCNQ2/3 chan-nels that generate the M current. This homeostatic mechanism issignificantly reduced in striatal output neurons of the R6/2 trans-genic mouse model of Huntington’s disease, at an age when theneurons are hyperactive in vivo and the mice begin to exhibit loco-motor impairment. Furthermore, it can be rescued by bath perfusionwith retigabine, a KCNQ channel activator, and chronic treatmentimproves locomotor performance. Thus, M-current dysfunction maycontribute to the hyperactivity and network dysregulation charac-teristic of this neurodegenerative disease, and KCNQ2/3 channelregulation may be a target for therapeutic intervention.

M current | intrinsic excitability | Huntington’s disease | KCNQ channels |homeostasis

Huntington’s disease (HD) is a fatal inherited autosomalneurodegenerative disorder, with its primary symptoms be-

ing progressive development of motor and cognitive dysfunction(1). The mutated gene, huntingtin (HTT), and its mutation, anexpansion of the number of CAG repeats, were identified 20 yago. However, the mechanism(s) underlying the pathologicalchanges that culminate in the degeneration of striatal outputneurons (SONs) remain unknown. Early animal models (2)generated a number of testable hypotheses, most notable beingthat the neurons degenerate because of a hyperactivity that leadsto a build-up of excitotoxic molecules. However, more recentstudies implicate alternative pathologies, such as altered tran-scriptional activity, calcium regulation and mitochondrial func-tion, or disruptions in normal neuronal patterns of activity (3)and show that neuronal dysfunction and behavioral and motorsymptoms of HD precede neurodegeneration (2). These studieshave been facilitated by access to transgenic mice models, in-cluding R6/1 and R6/2 mice, which express a truncated region ofthe mutant human HTT gene with expanded CAG repeats (4). Invitro recordings in both lines revealed that SONs are depolarizedand have higher input resistances than do wild-type (WT) con-trols, at a stage where deficits in locomotor activity begin to bemanifest (5–7). Furthermore, in vivo recordings indicate that at5–9 wk of age, when the mice exhibit overt motor deficits, R6/2SONs have higher firing rates and more regular discharge pat-terns compared with WT (8, 9). In contrast, neurodegenerationand death occur later (2). Hence, we asked whether cellularmechanisms that influence excitability might be altered in theearly stages of HD and might serve as targets for alleviatingassociated behavioral symptoms.Hyperactivity and related changes in neuronal firing patterns

could reflect alterations in synaptic transmission and its activity-dependent modifications or in intrinsic membrane properties

governing neuronal excitability (10–12). The latter can also bemodulated by activity (13) and have homeostatic roles (14). Wedescribe here a fast activity-dependent homeostatic control ofexcitability (fADH) in SONs. In WT mice, fADH can be inducedby brief trains of impulses and is expressed on a time scale ofseconds. It modifies firing rate and timing of evoked spikes,converting regular firing patterns to irregular ones, with thelatter mode resembling the accommodation attributed to voltage-and time-dependent activation of the M current mediated byKCNQ [or voltage-gated potassium channel (Kv) subfamily 7or Kv7] channels (14). Indeed, increasing activation of KCNQchannels on successive trials underlies fADH. Strikingly, wefound that fADH is reduced in R6/2 SONs, that two KCNQactivators (15, 16) rescued fADH in R6/2 SONs, thereby re-storing WT firing patterns, and that the locomotor signs of HDin the R6/2 mouse were ameliorated by chronic treatment withone of the activators.

ResultsBasic Membrane Properties of WT and R6/2 SONs. Comparison ofSON basic membrane properties for 4- to 6-wk-old mice revealedan increased input resistance in the transgenics, from 43.7 ± 3.8MΩ to 68.2 ± 7.3 MΩ, with a corresponding decrease in rheo-base, the threshold current for a long pulse, from 201.3 ± 15.4

Significance

Neurons typically regulate their intrinsic excitability to preventexcessive excitation and to gate information transfer. Thispaper describes an activity-dependent decrease in intrinsicexcitability following brief bursts of nerve impulses. This ho-meostatic mechanism, due to the recruitment, or sensitization,of voltage-gated potassium channels, the KCNQ2/3 channels, isreduced in striatal neurons of two transgenic mouse models ofHuntington’s disease at an age when these neurons are hy-peractive and motor symptoms begin to appear. Pharmaco-logical activation of these channels restores homeostasis intransgenic neurons, in vitro, and reduces motor impairment inbehaving mice, consistent with the hypothesis that hyperac-tivity enables establishment of dysfunctional neural circuitsand that KCNQ channels could serve as therapeutic targets forthe treatment of HD.

Author contributions: Y.C., D.B.-M., M.G., K.K., and D.S.F. designed research; Y.C., D.B.-M.,N.R., C.R., S.S.D., M.A.C., B.K., M.G., and D.S.F. performed research; Y.C., D.B.-M., N.R.,C.R., S.S.D., M.A.C., B.K., M.G., and D.S.F. analyzed data; and Y.C., D.B.-M., and D.S.F.wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Freely available online through the PNAS open access option.1Y.C. and D.B-M. contributed equally to this work.2To whom correspondence should be addressed. Email: dfaber@aecom.yu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1405748112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1405748112 PNAS | February 17, 2015 | vol. 112 | no. 7 | 2239–2244

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1

pA to 138.3 ± 13.0 pA, and without a detectable change in thethreshold voltage (Table S1). Comparable results were foundwhen the data analysis was restricted to recordings with a seriesresistance < 25 MΩ, although the decrease in rheobase did notreach significance with the smaller sample sizes (Table S1).These observations are similar to those reported for R6/2 SONsat an overlapping range of ages, 5–7 wk (5, 17). However, therewas no difference in resting membrane potential (Table S1), incontrast to the depolarization reported for the slightly older R6/2neurons. Because WT SONs tend to discharge in brief bursts invivo (18, 19), we asked whether the relationship between stim-ulus current (I) and impulse frequency (f) was altered, using 300-mssuprathreshold depolarizing pulses to evoke action potentialtrains, with a stimulus frequency (<0.5 Hz) that does not inducefADH. The f-I relations were essentially the same, after nor-malizing stimulus strength with respect to rheobase, and the cellshad the same maximum discharge rates (Fig. 1A). Thus, a changein input resistance per se would not necessarily account for dif-ferences in firing patterns observed in vivo.

Characterization of fADH.We noted that the responsiveness of WTSONs to a suprathreshold 300-ms depolarizing pulse repeated at1-s intervals gradually became weaker; the stimulus initiallyevoked a train of ∼10–13 impulses (Fig. 1B), but the firing fre-quency steadily decreased with successive stimuli and was re-duced by ∼50% after 10–20 trials. Fig. 1B contrasts this fADHin the WT with the minimal accommodation, or adaptation,exhibited by R6/2 SONs. For quantification, we defined theAdaptation Index, In = 1 − (no. of spikes, nth trial/no. of spikes,first trial); theoretically, In ranges from 0 to 1.0 (maximal ad-aptation to no evoked spikes). In Fig. 1B, for the WT SON, I10and I20 = 0.25 and 0.50, respectively, whereas for the R6/2 ex-ample, both indices = 0.14. Overall, the transgenics exhibited asignificant shift to the left in the cumulative distribution for I20(Fig. 1C), which averaged 0.26 ± 0.02, compared with 0.49 ± 0.03(n = 62 each) for the WT neurons. The differences between themedians and between the distributions are all highly significant(P < 0.001).There are two populations of SONs, those expressing D1- or

D2-dopamine receptors, also known as the direct and indirect

pathway neurons, with reported differences in electrophysiolog-ical properties, namely that D1 SONs apparently have a higherrheobase and a lower input resistance and fire at lower fre-quencies to the same magnitude depolarizing current pulse (20).These correlations were characteristic of our results, as shown byseparately sorting the WT and R6/2 datasets into subpopulationsdefined by those neurons with high and low rheobase values(upper and lower 25% or 50%), and there were no differences inI20 (Fig. S1A). In contrast, when the data were sorted accordingto high or low values of I20, rheobase, input resistance, and evokedspike rate were the same, in WT or in R6/2 (Fig. S1B). Thesefindings, coupled with the unimodal character of the cumulativedistributions for I20, suggest fADH is expressed in both neuronalsubtypes in WT and is diminished in both in the HD model.fADH was manifest within seconds and was cumulative. With

the standard protocol, it developed with a time constant of ∼10.5 sand decayed with a time constant of 16.3 s, such that recoverywas complete with 1- to 2-min rest (Fig. 1 B and D). It was notassociated with changes in input resistance, spike height or half-width, as measured for the first spike in an evoked train, orresting membrane potential. Also, it could be induced with otherconditioning frequencies, in the range of 0.5–2.0 Hz, with a ten-dency to increase as a function of both stimulus strength andfrequency (Fig. 2A). For example, on average, I20 increased by78% and 83% when conditioning frequency was doubled from 1to 2 Hz while maintaining the strength constant at 2.0 or 2.8times rheobase, respectively, and by 28% and 31% when thestrength was increased while holding frequency constant at 1 or2 Hz, respectively. These results suggest the magnitude of fADHdepends on the time integral of the conditioning depolarization;however, subthreshold depolarizations did not induce fADH.Finally, the magnitude of the Adaptation Index grew progres-sively as a function of the cumulative number of spike dischargesin the preceding 20 s (Fig. 2B), as might be expected of a spike-counting mechanism characterized by interplay between activa-tion and recovery processes with similar kinetics.We asked whether fADH was a consequence of dialysis of the

neuronal cytoplasm, by recording first in the cell attached modeand stimulating extracellularly through the same electrode, al-lowing direct comparison with subsequent recordings from the

Fig. 1. Impairment of fADH in R6/2 SONs. (A) WT and R6/2SONs have similar f-I relations and asymptotic firing rates.Plot of mean impulse frequency during a 300-ms depola-rizing current pulse versus current amplitude, normalizedwith respect to rheobase. Also shown are exponential fitsof data from four WT (black) and seven R6/2 (red) SONs. (B)fADH is prominent in WT (Left) but not in R6/2 (Right)SONs. Top three traces in each column are first, 10th and20th spike trains evoked by the current pulse repeated at1 Hz. Stimulus strength, 2× rheobase. Bottom traces, re-covery after 1-min rest and sample current pulse. (C and D)Characterization of fADH. (C) Cumulative distributions ofI20 for WT (black) and R6/2 (red) SONs (n = 62), MW U test,***P < 0.001, KS test, ***P < 0.001. (D) Progressive build-up of adaptation, and recovery kinetics, plotted as Adap-tation Index (In, ordinate), versus time (abscissa). Recoverydata obtained by following conditioning series with restperiods of 2, 5, 10, 30, and 60 s (one test pulse per condi-tioning series). Time constants of development, τdev, andrecovery, τrec, calculated from monoexponential fits ofpooled data (7 SONs).

2240 | www.pnas.org/cgi/doi/10.1073/pnas.1405748112 Cao et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1

same neuron in whole cell mode. fADH was observed in bothmodes, with the Adaptation Index being greater, on average, inthe latter (Fig. S2A). Also, the magnitude of fADH was the sameafter warming the slices from 26.0 ± 0.6 to 30.7 ± 0.4 °C, sug-gesting it is operational at physiological temperatures (Fig. S2B).

fADH Alters Structure of Impulse Trains. fADH introduces botha progressive slowing of impulse activity within an evoked burstof spikes and an increased variability in the interspike interval,whereas impulse trains in R6/2 SONs are more regular andpredictable (Fig. 2 C1, C2, and D). It could be evoked with otherstimulating paradigms, such as ones that mimicked repetitivesynaptic input, and in that case, the evoked discharge patternbecame irregular in WT but not in the transgenic neurons (Fig. 2E1 and E2). Thus, the stimulus protocols used here evokeddischarge patterns similar to those of WT SONs recorded in vivoat a similar age, namely an overall low frequency firing ratecharacterized by brief bursts separated by periods of quiescence(18, 19). In contrast, R6/2 SONs discharge at higher frequenciesand are more regular (9), consistent with the hypothesis thatthese differences reflect the differential expression of fADH.Therefore, we investigated the mechanism underlying fADH.

fADH Is Due to Increased M-Current Activation. Potential mecha-nisms underlying fADH include Ca2+- and voltage-dependentmodulation of ion channels that shape neuronal activity patterns,such as a persistent subthreshold Na+ current (21) or various K+

channels (11, 22). Because neuronal activity is typically associ-ated with an increase in intracellular [Ca2+], we first askedwhether chelating this cation with 25 mM BAPTA in the re-cording pipette affected fADH. However, BAPTA had no effecton fADH, as seen by comparing the cumulative distributionsof I20 in its presence or absence (Fig. S3). We then focused onK+ channels, specifically the KCNQ channels that generate the

voltage-dependent M-current that contributes to accommodationin a number of cell types (23, 24) and is expressed by SONs (25).The KCNQ channel blocker XE-991 (6 μM) was bath-applied

while recording from SONs, and it significantly reduced adap-tation (Fig. 3A). In the illustrated case, I20 decreased from 0.6 to0.18. The mean adaptation indices I10 and I20 from 7 SONs (31–37 d postnatal) in which the drug effect was followed by suc-cessful wash out (recovery to within 10% of control) were bothreduced by 50% or more (Fig. 3C).Finally, Fig. 3B demonstrates directly, with voltage-clamp

recordings, that the induction protocol and fADH are associatedwith an enhanced outward current response to a 70-mV, 300-msdepolarizing step from −90 mV (Upper) and that in the presenceof XE-991, there is no additional current induced by the condi-tioning paradigm (Middle). The magnitude of the XE-991 sen-sitive current is increased by approximately 60% after inductionof fADH (Lower). In contrast, XE-991 had no effect on theresponses of the R6/2 neurons. An alternative method forquantifying M current is as the current that deactivates whenmembrane potential is stepped back from −20 to −50 mV (25).The amplitude of the deactivating current was enhanced afterfADH induction, and this increase decayed back toward base-line over a period of 60–90 s, comparable to the time course ofthe decay of fADH (Fig. S4). Also, comparison of normalizedI-V plots before conditioning did not reveal any differences in

Fig. 2. Characterization of fADH. (A) Adaptation Indices after 10, 20, and 30trials, plotted as functions of stimulus frequency (1 and 2 Hz) and strength (2,2.4, 2.8, and 3.2 times rheobase, Rb). (B) Adaptation Index versus cumulativenumber of spikes in the preceding 20 trials (n = 4 experiments) with expo-nential fit (red). (C1, C2, and D) Influence of fADH on spike firing patterns.(C1 and C2) Spike raster plots from the experiments of Fig. 1B, contrastingexpression of fADH in WT (1) with that in R6/2 (2). Timing of successiveevoked spikes in each train is marked by color-coded symbols, with trialnumber increasing from top to bottom. (D) Deficit in fADH in R6/2 minimizesvariability in spike timing. Mean interspike interval (ISI) for the first andfourth intervals in the trains and mean first spike latency plotted versus tracenumber for WT (black) and R6/2 (red) (7 WT and 14 R6/2 SONs). (E1 and E2)fADH evoked by brief stimulus trains. Same format as in Fig. 1B but stimu-lation was with a train of 10 suprathreshold pulses (2x Rb, 10-ms duration,30-ms cycle time). Evoked spike train in R6/2 (2) is at a higher frequency andis more regular than that in WT (1).

Fig. 3. fADH is due to activity-dependent enhancement of M current. (A)fADH in control (Left) and in the presence of 6 μM XE-991 (Right). Traces are,from top to bottom, first and 20th trials, and recovery. Stimulus: 1.8x Rb. (B)Voltage clamp recordings of current responses to a 70-mV step from −90 mV,in control (Upper) and after adding XE-991 (Middle). Responses obtained be-fore (black) and after (red) inducing fADH. (Lower) XE-991 sensitive currents inthe two conditions. (C) Bar plots of effects of 6 μM XE-991 (Upper), 10 μMMcN-A-343 (Middle), and 10 μM Wortmannin (Lower), on I10 and I20. n = 7, 7,and 5, for Control, XE-991 and washout, respectively; n = 8, 8, and 5, forControl, McN-A-343, and washout, respectively; n = 6, 6, and 4, for Control,Wortmannin, and washout, respectively; error bars are SEM; Student’s t testandWelch’s t test, not significant (n.s.) P > 0.05, *P < 0.05, **P < 0.01. (D and E)Same format as in A, but for 10 μM McN-A-343 and 10 μM Wortmannin,respectively.

Cao et al. PNAS | February 17, 2015 | vol. 112 | no. 7 | 2241

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1

voltage dependence of M-current activation between WT andR6/2 SONs (Fig. S4).The M-current magnitude is sensitive to the level of phos-

phatidylinositol 4,5-bisphosphate (PIP2) in the plasma mem-brane, through mechanisms that may alter channel activation, forinstance, by regulating channel open probability, voltage de-pendence, or number (26). PIP2 hydrolysis is increased by acti-vation of M1 muscarinic receptors, through a G protein-coupledreceptor mediated cascade, and SON excitability is enhanced bycholinergic inhibition of KCNQ channels (26). As predicted, theM1 agonist, McN-A-343 (10 μM) reversibly reduced fADH (Fig.3 C and D). In the illustrated example, I20 was reduced by 50%,and overall I10 and I20 were reduced by 47 and 40%, respectively,from 0.38 ± 0.04 to 0.20 ± 0.03 (P < 0.01) and from 0.42 ± 0.04 to0.25 ± 0.03 (P < 0.01, n = 8; in 5 experiments, the block was re-versed by washout). Also, the activity of PI-4 kinase, which catalyzesthe synthesis of PIP in the plasma membrane, can be enhanced bydepolarization, favoring increased downstream synthesis of PIP2(27–29). This first step in the synthetic cascade can be blocked byhigh concentrations (10 μM) of Wortmannin, and indeed, thiscompound significantly reduces fADH (Fig. 3 C and E). Takentogether, these results are consistent with the hypothesis that fADHis due to activity-dependent sensitization of the M current.We performed Western blot analysis of KCNQ2 expression in

dorsal striatum and prefrontal cortex of 5-wk-old WT and R6/2mice. Expression of two inward rectifying channel proteins,Kir2.1 and Kir2.3 (30), was also quantified. Although KCNQ2abundance in cortex appeared to be reduced in the transgenic,the effect was not significant, and we found no significant dif-ference in its level in striatum (Fig. S5). Also, Kir2.1 expressionwas reduced only in striatum, whereas Kir2.3 was reduced inboth structures. The maintained expression of KCNQ2 is con-sistent with the demonstration that these channels can be acti-vated in R6/2 SONs, as described below.

M-Current Activators Rescue fADH in R6/2 Neurons. We next testedthe possibility that the reduction in fADH in R6/2 SONs reflectsaltered channel regulation by asking whether M-current activa-tors could reverse this functional deficit. Fig. 4 A and C and Fig.4 B and D illustrate results obtained from R6/2 SONs duringbath superfusion of 25 μM diclofenac (15) and 10 μM retigabine(16), respectively. In both, the activators, which shift the con-ductance-voltage relation to the left on the voltage axis, re-versibly enabled fADH, increasing I20 by a factor of 2–3. Onaverage, diclofenac increased I20 from 0.22 ± 0.02 to 0.50 ± 0.06(n = 8, P < 0.01) whereas the control and washout data werestatistically the same (P > 0.45). The corresponding values forretigabine are 0.24 ± 0.03 vs. 0.54 ± 0.05 (P < 0.001). This effectis in contrast to the observation that the magnitude of fADH inWT SONs was unaffected by retigabine (Fig. S6). Finally, in-creasing intracellular PIP2 concentration by including diC8-PIP2(31) in the recording pipette solution also rescued fADH in R6/2SONs (Fig. S6). These findings indicate that the transgenic SONshave latent activatable KCNQ channels.Multiple transgenic mouse models are used to study HD, and

the temporal progression of the signs and symptoms differ,depending on the construct and the number of CAG repeats. Forexample, motor signs seen at 4 wk of age in the truncated R6/2model (32), appear later, at approximately 2 mo, in the BACHDmouse, which has the full-length mutated human gene (33, 34).We thus asked whether fADH could be induced in the WT miceused to generate the BACHD and whether there is a deficit inthis adaptive mechanism at 2 mo in the BACHD mice. SONsfrom both sets of mice exhibited fADH, and it was quantitativelysmaller in the BACHD SONs (Fig. S7). Specifically, the meanAdaptation Index was decreased by 34% and 31% after 10 and20 trials, respectively. This result indicates there is a similardeficit in fADH at quite different ages, but at a comparable stage

of the pathology. We also confirmed that fADH could be res-cued in the affected BACHD neurons by applying retigabine; in8 SONs, it increased I10 from 0.13 ± 0.019 to 0.24 ± 0.044 (P <0.05) and I20 from 0.19 ± 0.02 to 0.31 ± 0.052 (P < 0.05).

Chronic Treatment with an M-Current Activator Ameliorates LocomotorDeficits in R6/2 Mice. As noted, R6/2 mice begin to exhibit loco-motor deficits as early as 4–5 wk of age (32). Because ourelectrophysiological results indicate there is a deficit in fADH atthe same age, we asked whether chronic treatment with retiga-bine would reduce or slow the development of these deficits, asmeasured by performance in the open field test, and positiveresults were obtained across a 3-wk testing period (Fig. 5). Micewere treated with daily i.p. injections of retigabine (10 mg/kg),starting at 4 wk of age and were tested once per week from week5 through week 7. In this age range, the R6/2 controls did notexhibit seizures, and thus any effects of retigabine would not bedue to its antiepileptic properties. General locomotor activitywas assessed as track length, and exploration as the number ofrearings, in a 9-min period. For both measures, the WT controlsthat were untreated, vehicle-treated, and drug-treated were sta-tistically equivalent, and the first two groups are combined in theanalysis and graphs. The same was the case for the R6/2 un-treated and vehicle-treated controls, and they are referred to ascontrol R6/2. The figure illustrates that the control R6/2 miceare significantly impaired in both measures of general activityand exploration, compared with WT, across the 3-wk testingperiod. More importantly, administration of retigabine signifi-cantly improved locomotor activity in treated R6/2 mice com-pared with control R6/2 mice across the 3-wk test period. Thedata also demonstrate a trend, albeit not statistically significant,toward improvement in the number of rearings in treated R6/2mice compared with control R6/2 mice. Thus, the M-currentactivator rescued both electrophysiological and behavioral phe-notypes of the transgenic HD model. Finally, because the age ofthe mice used for the behavioral studies encompassed 5–7 wk, weconfirmed that fADH was still differentially expressed in WTand R6/2 at 7–8 wk (Fig. S8).

DiscussionThe major findings of this study are that SONs exhibit activity-dependent homeostatic regulation of excitability that is asso-ciated with enhanced activation of the M current (11) and is

Fig. 4. fADH is expressed by R6/2 SONs after superfusion with M-currentactivators. Data from two activators, 25 μM diclofenac (A and C) and 10 μMretigabine (B and D) A and B are the same format as in Fig. 2A. Left, control;Right, activator. (C) Bar plots of I10 and I20 for control (n = 8), diclofenac (n =8), and washout (n = 2). (D) Bar plots for retigabine (n = 7 for control andactivator, 3 for washout). (C and D) Error bars are SEM; Student’s t test andWelch’s t test, n.s. P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.

2242 | www.pnas.org/cgi/doi/10.1073/pnas.1405748112 Cao et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1

depressed in SONs in two transgenic mouse models of HD, at ageswhen motor symptoms are just emerging (32–34). Moreover, thisdeficit is at least partially reversed by acute exposure to M-currentactivators. Finally, daily treatment of the R6/2 mice with an acti-vator improves motor performance in the affected mice.We propose that fADH is due to a progressive activity-

dependent enhanced activation, or sensitization, of KCNQchannels, such that after one brief train of impulses, the nextexcitation can generate a larger K+ conductance and slow neu-ronal spiking. Presumably, deactivation of the conductance be-tween test stimuli and the relatively slow time constant ofactivation of the M current by depolarization, 80–170 ms, inSONs (25) and in CHO cells (35), explains why this modulationhas a minimal effect on input resistance or the timing of the firstimpulse within a burst. We suggest that the observed increase inthe outward XE-991–sensitive currents after fADH inductionmost likely reflects a shift in open probability or an increase inthe number of available channels (27), consistent with the effectof PIP2. Although a shift in the voltage dependence of the chan-nels, as occurs with retigabine (16), cannot be ruled out, it is lesslikely, given that the normalized I-V plots of WT and R6/2 SONsare the same for depolarizations up to −20 mV. Regardless, ourresults are consistent with the previous demonstration thatM-current suppression increases SON excitability (25). The in-creased input resistance, which could reflect down-regulation ofa second channel, would also contribute to increased excitability.The temporal domain of fADH is suggestive of real-time

adjustments in intrinsic excitability, and distinguishes it fromslower developing long-term forms of intrinsic plasticity, whichalso reflect K+ channel regulation (13, 14). Thus, fADH is a formof metaplasticity, and it is due to time-dependent M-currentrecruitment in response to prolonged or repeated excitation,presumably by increasing the level of PIP2 in the membrane, andit can be expected to have a braking effect on neuronal activity.

The effects of Wortmannin and PIP2 are consistent with thehypothesis that fADH is the consequence of a depolarization-induced increased synthesis of PIP2, through a membrane-boundenzymatic cascade. Huntingtin inserts in the plasma membranethrough interactions with phosphoinositides, and N-terminalhuntingtin fragments target membrane regions enriched in PIP2(36). Furthermore, these interactions are altered by poly-glutamine expansions of the N-terminal fragments, including a shiftin targeting toward intracellular regions (37, 38). This aberrantmolecular organization might well underlie the deficit in fADHin the HD mice.A slower and more persistent form of activity-dependent ho-

meostatic intrinsic plasticity, exhibited by hippocampal pyrami-dal neurons, is also mediated by KCNQ channels (14). However,in that case, the induction mechanism is Ca2+-dependent, lead-ing to the suggestion that complementary homeostatic mecha-nisms operating through separate but convergent signaling path-ways may function to place an upper limit on neuronal firing, andto guarantee a degree of variability in impulse pattern.There is a striking agreement among the ages at which R6/2

mice first become symptomatic, demonstrate increased activityand altered firing patterns in vivo, and exhibit a deficit in fADH.This correlation is strengthened by the parallel observations withthe BACHD mice. In addition, evidence that KCNQ channelscan be recruited in the transgenic neurons, coupled with theeffects, at the same age, of the M-current activator retigabineon motor performance, identifies a potential for reducing bothneuronal hyperactivity and motor impairment. Its effect on lo-comotor activity presumably is not due to a nonspecific reductionin neuronal activity, because it had no effect on WT mice.Homeostatic mechanisms can have significant effects on net-

work development and function (39, 40). As one example, sup-pression of KCNQ2 expression increases the excitability andreduces spike-frequency adaptation of hippocampal CA1 pyra-midal cells and alters hippocampal morphology, with associatedbehavioral changes, and the manifestation of these effects de-pends on the time window of suppression (41). Given these con-siderations, we suggest that fADH may have a critical role in theformation and stability of the circuitry in dorsal striatum, for ex-ample, by preventing the aberrant hyperactivity that has beenimplicated in the etiology of a number of neurodegenerative dis-orders (42).

Experimental ProceduresElectrophysiology. Parasagittal brain slices (300 μm thick) were prepared, asdescribed (43), from transgenic R6/2 mice and their WT (CBA × C57BL/6J)littermates (postnatal days 25–57) as described (43), and from BACHD miceand their WT littermates (FVB/N). Animals were bred and genotyped atJackson Labs and provided by the CHDI Foundation. Animal handling anduse followed a protocol approved by the Institutional Animal Care and UseCommittee of Albert Einstein College of Medicine. Briefly, animals weredecapitated under isofluorane anesthesia. The brain was removed quicklyand transferred into ice-cold saline that contained the following (in mM):125 NaCl, 4 KCl, 10 glucose, 1.25 NaH2PO4, 25 NaHCO3, 0.5 CaCl2, and2.5 MgCl2, equilibrated with a 5% CO2-95% O2 mixture (pH 7.3). The twosagittal brain hemispheres were separated and adhered to the stage ofa Vibratome Series 1000 Classic vibroslicer, and slices were cut at an angle of10 ± 2° (44). Slices were allowed to recover for at least 1 h at room tem-perature in artificial cerebrospinal fluid (ACSF) consisting of (in mM): 125NaCl, 4 KCl, 10 glucose, 1.25 NaH2PO4, 25 NaHCO3, 2 CaCl2 and 1 MgCl2 (pHbuffered to 7.3 with 5% CO2-95% O2).

Whole-cell voltage- and current-clamp recordings were performed, usingIR-differential interference contrast (DIC) video microscopy to visualize andphotograph individual neurons, and cell types were distinguished on thebasis of characteristic morphology, action potential waveforms and re-sponses to 300–400 ms depolarizing and hyperpolarizing current pulses. Toquantify input resistance, we used the slope of first part in the hyper-polarizing quadrant of the I-V function for small hyperpolarizing and de-polarizing current pulses, thereby avoiding the region of pronouncedinward rectification.

Fig. 5. Deficits in open field behavior evident in R6/2 mice were partiallyameliorated by retigabine, assessed as track length when tested once perweek across a 3-wk test period. (A) R6/2 mice had significantly lower activity,than WT mice + = P < 0.01 R6/2 compared with control WT, repeatedmeasures ANOVA [F(1,57) = 30.4]. Retigabine-treated R6/2 mice had signifi-cantly increased activity across the 3-wk test period compared with un-treated R6/2 mice. *P < 0.05 [F(1,29) = 4.9]. Untreated and retigabine-treatedWT mice were statistically indistinguishable. The numbers of mice in eachgroup were as follows: WT Control, 27; Treated WT, 12; R6/2 Control, 20;Treated R6/2, 11. (B) R6/2 mice had significantly less exploration, assessed asnumber of rears, than WT mice + = P < 0.01, repeated measures ANOVA,[F(1,57) = 50.5]. The apparent improvement in rears in R6/2 mice treated withretigabine compared with untreated R6/2 mice did not reach statistical sig-nificance [F(1,29) = 3.3, P < 0.08].

Cao et al. PNAS | February 17, 2015 | vol. 112 | no. 7 | 2243

NEU

ROSC

IENCE

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1

Recordings were at room temperature, unless stated otherwise, withpatch-type pipette electrodes (4–8 MΩ) filled with (in mM) 100 potassiumgluconate, 50 KCl, 5 NaCl, 0.5 CaCl2, 5 EGTA, 25 Hepes, 2 MgATP, 0.3 GTP (pH7.2). Series resistance (<40 MΩ) was monitored throughout each experiment,and data from cells with more than 20% changes in series resistance werediscarded. Although 40 MΩ is commonly used as a cutoff in recordings fromSONs (14, 25), when the data analysis was restricted to values <25 MΩ,results were essentially the same (Table S1). Electrophysiological data weredigitized at 20 kHz with a National Instruments BNC 2090 A-D convertorconnected to a Macintosh personal computer and analyzed with customizedsoftware, as well as in IGOR and Origin. All data are from neurons withresting membrane potentials of magnitude greater than −70 mV.

Statistical Analysis. Values are expressed as mean ± SEM (n = number of cells).Two-tailed Student’s and Welch’s t tests were used for statistical compar-isons at the P < 0.05 significance level. Cumulative probability plots werecompared by using the Kolmogorov–Smirnov (KS) test and the Mann–Whitney (MW) U test. Behavioral data were analyzed by using a repeated-measures ANOVA.

Drugs. XE-991 and Wortmannin (Ascent Scientific), diclofenac and McN-A-343(Sigma-Aldrich), and retigabine (provided by CHDI Foundation) were added toACSF after obtaining control data, and at least 15 min was allowed to lapsebefore recording in the presenceof the drug. To test for involvement of activity-dependent rises in intracellular Ca2+ concentration, 25 mM BAPTA was

added directly to the pipette solution. To test the PIP2 hypothesis, cellswere loaded with a water soluble form of PIP2, diC8-PIP2 (Cayman Chem-ical) via the patch pipette. diC8-PIP2 was dissolved in deionized water at1 μg/mL concentration and stored in aliquots at −20 °C (31). An aliquot wasthawed, diluted in internal solution at 30 μM final concentration, andused immediately.

Open Field. There were three groups each of WT and R6/2 mice: untreatedcontrols, those treatedwith 10mg/kg (i.p.) retigabine daily and those injectedwith vehicle. Treatment started in week 4 and testing in week 5. Mice wereplaced in an opaque Plexiglas arena (16 square inches) and allowed to freelyexplore the arena for 9 min, during which time voluntary locomotion (totaldistance traveled in centimeters), exploration (number of rears, defined aslifting of the upper body and forepaws off the ground, whisking and sniffing)were scored while being recorded digitally. Longer test durations were notused, because they are often associated with prolonged periods of inactivity,i.e., habituation. Track length was scored automatically with Viewer trackingsoftware (Biobserve) and rears were scoredmanually. All measurements weredone blind to genotype and treatment condition.

ACKNOWLEDGMENTS. We thank R. Grantyn for comments on the manuscriptand for suggestions about experimental design. This work was supported byCHDI Foundation (A-3674), NIH (NS050808, NS079750, and RR0277888), andFondecyt (11140430).

1. The Huntington’s Disease Collaborative Research Group (1993) A novel gene con-taining a trinucleotide repeat that is expanded and unstable on Huntington’s diseasechromosomes. Cell 72(6):971–983.

2. Cepeda C, Cummings DM, André VM, Holley SM, Levine MS (2010) Genetic mousemodels of Huntington’s disease: Focus on electrophysiological mechanisms. ASNNeuro 2(2):e00033, 10.1042/AN20090058.

3. Zuccato C, Valenza M, Cattaneo E (2010) Molecular mechanisms and potential ther-apeutical targets in Huntington’s disease. Physiol Rev 90(3):905–981.

4. Mangiarini L, et al. (1996) Exon 1 of the HD genewith an expanded CAG repeat is sufficientto cause a progressive neurological phenotype in transgenic mice. Cell 87(3):493–506.

5. Ariano MA, et al. (2005) Striatal potassium channel dysfunction in Huntington’s dis-ease transgenic mice. J Neurophysiol 93(5):2565–2574.

6. Cummings DM, et al. (2006) Aberrant cortical synaptic plasticity and dopaminergic dys-function in a mouse model of Huntington’s disease. Hum Mol Genet 15(19):2856–2868.

7. Cummings DM, et al. (2009) Alterations in cortical excitation and inhibition in geneticmouse models of Huntington’s disease. J Neurosci 29(33):10371–10386.

8. Rebec GV, Conroy SK, Barton SJ (2006) Hyperactive striatal neurons in symptomaticHuntington R6/2 mice: Variations with behavioral state and repeated ascorbatetreatment. Neuroscience 137(1):327–336.

9. Miller BR, Walker AG, Barton SJ, Rebec GV (2011) Dysregulated neuronal activitypatterns implicate corticostriatal circuit dysfunction in multiple rodent models ofHuntington’s disease. Front Syst Neurosci 5:26.

10. Turrigiano G (2011) Too many cooks? Intrinsic and synaptic homeostatic mechanismsin cortical circuit refinement. Annu Rev Neurosci 34:89–103.

11. Misonou H (2010) Homeostatic regulation of neuronal excitability by K(+) channels innormal and diseased brains. Neuroscientist 16(1):51–64.

12. Frick A, Magee J, Johnston D (2004) LTP is accompanied by an enhanced local excit-ability of pyramidal neuron dendrites. Nat Neurosci 7(2):126–135.

13. Xu J, Kang N, Jiang L, Nedergaard M, Kang J (2005) Activity-dependent long-termpotentiation of intrinsic excitability in hippocampal CA1 pyramidal neurons. J Neurosci25(7):1750–1760.

14. WuWW, Chan CS, Surmeier DJ, Disterhoft JF (2008) Coupling of L-type Ca2+ channelsto KV7/KCNQ channels creates a novel, activity-dependent, homeostatic intrinsicplasticity. J Neurophysiol 100(4):1897–1908.

15. Peretz A, et al. (2005) Meclofenamic acid and diclofenac, novel templates of KCNQ2/Q3 potassium channel openers, depress cortical neuron activity and exhibit anticon-vulsant properties. Mol Pharmacol 67(4):1053–1066.

16. Rundfeldt C, Netzer R (2000) The novel anticonvulsant retigabine activates M-currentsin Chinese hamster ovary-cells tranfected with human KCNQ2/3 subunits. NeurosciLett 282(1-2):73–76.

17. Klapstein GJ, et al. (2001) Electrophysiological and morphological changes in striatalspiny neurons in R6/2 Huntington’s disease transgenic mice. J Neurophysiol 86(6):2667–2677.

18. Wilson CJ, Groves PM (1981) Spontaneous firing patterns of identified spiny neuronsin the rat neostriatum. Brain Res 220(1):67–80.

19. Wilson CJ (1993) The generation of natural firing patterns in neostriatal neurons.Prog Brain Res 99:277–297.

20. Gertler TS, Chan CS, Surmeier DJ (2008) Dichotomous anatomical properties of adultstriatal medium spiny neurons. J Neurosci 28(43):10814–10824.

21. D’Ascenzo M, et al. (2009) Activation of mGluR5 induces spike afterdepolarizationand enhanced excitability in medium spiny neurons of the nucleus accumbens bymodulating persistent Na+ currents. J Physiol 587(Pt 13):3233–3250.

22. Jentsch TJ (2000) Neuronal KCNQ potassium channels: Physiology and role in disease.Nat Rev Neurosci 1(1):21–30.

23. Brown DA, Adams PR (1980) Muscarinic suppression of a novel voltage-sensitive K+current in a vertebrate neurone. Nature 283(5748):673–676.

24. Aiken SP, Lampe BJ, Murphy PA, Brown BS (1995) Reduction of spike frequency ad-aptation and blockade of M-current in rat CA1 pyramidal neurones by linopirdine(DuP 996), a neurotransmitter release enhancer. Br J Pharmacol 115(7):1163–1168.

25. Shen W, Hamilton SE, Nathanson NM, Surmeier DJ (2005) Cholinergic suppression ofKCNQ channel currents enhances excitability of striatal medium spiny neurons.J Neurosci 25(32):7449–7458.

26. Suh BC, Inoue T, Meyer T, Hille B (2006) Rapid chemically induced changes of PtdIns(4,5)P2 gate KCNQ ion channels. Science 314(5804):1454–1457.

27. Li Y, Gamper N, Hilgemann DW, Shapiro MS (2005) Regulation of Kv7 (KCNQ) K+channel open probability by phosphatidylinositol 4,5-bisphosphate. J Neurosci 25(43):9825–9835.

28. Zhang X, et al. (2010) Depolarization increases phosphatidylinositol (PI) 4,5-bisphos-phate level and KCNQ currents through PI 4-kinase mechanisms. J Biol Chem 285(13):9402–9409.

29. Chen X, et al. (2011) Membrane depolarization increases membrane PtdIns(4,5)P2levels through mechanisms involving PKC βII and PI4 kinase. J Biol Chem 286(46):39760–39767.

30. Kubo Y, et al. (2005) International Union of Pharmacology. LIV. Nomenclature andmolecular relationships of inwardly rectifying potassium channels. Pharmacol Rev57(4):509–526.

31. Sohn JW, Lim A, Lee SH, Ho WK (2007) Decrease in PIP(2) channel interactions is thefinal common mechanism involved in PKC- and arachidonic acid-mediated inhibitionsof GABA(B)-activated K+ current. J Physiol 582(Pt 3):1037–1046.

32. Hickey MA, Gallant K, Gross GG, LevineMS, Chesselet MF (2005) Early behavioral deficitsin R6/2 mice suitable for use in preclinical drug testing. Neurobiol Dis 20(1):1–11.

33. Gray M, et al. (2008) Full-length human mutant huntingtin with a stable poly-glutamine repeat can elicit progressive and selective neuropathogenesis in BACHDmice. J Neurosci 28(24):6182–6195.

34. Estrada-Sánchez AM, Rebec GV (2013) Role of cerebral cortex in the neuropathologyof Huntington’s disease. Front Neural Circuits 7:19.

35. Gamper N, Stockand JD, Shapiro MS (2003) Subunit-specific modulation of KCNQpotassium channels by Src tyrosine kinase. J Neurosci 23(1):84–95.

36. Kegel KB, et al. (2005) Huntingtin associates with acidic phospholipids at the plasmamembrane. J Biol Chem 280(43):36464–36473.

37. Kegel KB, et al. (2009) Polyglutamine expansion in huntingtin alters its interactionwith phospholipids. J Neurochem 110(5):1585–1597.

38. Kegel KB, et al. (2009) Polyglutamine expansion in huntingtin increases its insertioninto lipid bilayers. Biochem Biophys Res Commun 387(3):472–475.

39. Marder E, Goaillard JM (2006) Variability, compensation and homeostasis in neuronand network function. Nat Rev Neurosci 7(7):563–574.

40. Turrigiano GG, Nelson SB (2004) Homeostatic plasticity in the developing nervoussystem. Nat Rev Neurosci 5(2):97–107.

41. Peters HC, Hu H, Pongs O, Storm JF, Isbrandt D (2005) Conditional transgenic sup-pression of M channels in mouse brain reveals functions in neuronal excitability,resonance and behavior. Nat Neurosci 8(1):51–60.

42. Palop JJ, Mucke L (2010) Synaptic depression and aberrant excitatory network activityin Alzheimer’s disease: Two faces of the same coin? Neuromolecular Med 12(1):48–55.

43. Kirmse K, Dvorzhak A, Kirischuk S, Grantyn R (2008) GABA transporter 1 tunesGABAergic synaptic transmission at output neurons of the mouse neostriatum.J Physiol 586(Pt 23):5665–5678.

44. Beurrier C, Ben-Ari Y, Hammond C (2006) Preservation of the direct and indirect pathwaysin an in vitro preparation of the mouse basal ganglia. Neuroscience 140(1):77–86.

2244 | www.pnas.org/cgi/doi/10.1073/pnas.1405748112 Cao et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 22

, 202

1