Early dysfunction and progressive degeneration of the subthalamic nucleus in mouse models of Huntington's disease

The subthalamic nucleus (STN) is an element of cortico-basal ganglia-thalamo-cortical circuitry critical for action suppression. In Huntington's disease (HD) action suppression is impaired, resembling the effects of STN lesioning or inactivation. To explore this potential linkage, the STN was studied in BAC transgenic and Q175 knock-in mouse models of HD. At <2 and 6 months of age autonomous STN activity was impaired due to activation of KATP channels. STN neurons exhibited prolonged NMDA receptor-mediated synaptic currents, caused by a deficit in glutamate uptake, and elevated mitochondrial oxidant stress, which was ameliorated by NMDA receptor antagonism. STN activity was rescued by NMDA receptor antagonism or the break down of hydrogen peroxide. At 12 months of age approximately 30% of STN neurons had been lost, as in HD. Together, these data argue that dysfunction within the STN is an early feature of HD that may contribute to its expression and course. DOI: http://dx.doi.org/10.7554/eLife.21616.001


Introduction
The basal ganglia are a network of subcortical brain nuclei that are critical for action selection and central to the expression of several psychomotor disorders (Albin et al., 1989;Wichmann and DeLong, 1996). Information flow from the cortex to the output nuclei of the basal ganglia occurs via three major pathways. The so-called direct pathway through the striatum promotes movement and 'rewarding' behavior through inhibition of GABAergic basal ganglia output (Chevalier and Deniau, 1990;Kravitz et al., 2010;Kravitz and Kreitzer, 2012). In contrast, the indirect pathway via the striatum, external globus pallidus and subthalamic nucleus (STN) and the hyperdirect pathway through the STN suppress the same processes through elevation of basal ganglia output (Maurice et al., 1999;Tachibana et al., 2008;Kravitz et al., 2010;Kravitz and Kreitzer, 2012). Indeed, interruption of the indirect and hyperdirect pathways through lesion or inactivation of the STN is associated with elevated/involuntary movement, impulsivity and psychiatric disturbances such as hypomania and hyper-sexuality (Crossman et al., 1988;Hamada and DeLong, 1992;Baunez and Robbins, 1997;Bickel et al., 2010;Jahanshahi et al., 2015).
Huntington's disease (HD) is an autosomal dominant, neurodegenerative disorder caused by an expansion of CAG repeats in the gene (HTT) encoding huntingtin (HTT), a protein involved in vesicle dynamics and intracellular transport (Huntington's Disease Collaborative Research Group, 1993;Saudou and Humbert, 2016). Early symptoms of HD include involuntary movement, compulsive behavior, paranoia, irritability and aggression (Anderson and Marder, 2001;Kirkwood et al., 2001). These symptoms have traditionally been linked to cortico-striatal degeneration, however a role for the STN is suggested by their similarity to those caused by STN inactivation or lesion. The hypoactivity of the STN in HD models in vivo Abercrombie, 2015a, 2015b) and the susceptibility of the STN to degeneration in HD (Lange et al., 1976;Guo et al., 2012) are also consistent with STN dysfunction.
Several mouse models of HD have been generated, which vary by length and species origin of HTT/Htt, CAG repeat length, and method of genome insertion. For example, one line expresses fulllength human HTT with 97 mixed CAA-CAG repeats in a bacterial artificial chromosome (BAC; Gray et al., 2008), whereas Q175 knock-in (KI) mice have an inserted chimeric human/mouse exon one with a human polyproline region and~188 CAG repeats in the native Htt (Menalled et al., 2012).
Although HD models exhibit pathogenic processes seen in HD, they do not exhibit similar levels and spatiotemporal patterns of cortico-striatal neurodegeneration. Striatal neuronal loss in aggressive Htt fragment models such as R6/2 mice does occur but only close to death (Stack et al., 2005), whereas full-length models exhibit minimal loss (Gray et al., 2008;Smith et al., 2014). Despite the loss and hypoactivity of STN neurons in HD and the similarity of HD symptoms to those arising from STN lesion or inactivation, the role of the STN in HD remains poorly understood. We hypothesized that the abnormal activity and progressive loss of STN neurons in HD may reflect abnormalities within the STN itself. This hypothesis was addressed in the BAC and Q175 KI HD models using a combination of cellular and synaptic electrophysiology, optogenetic interrogation, two-photon imaging and stereological cell counting.

Results
Data are reported as median [interquartile range]. Unpaired and paired statistical comparisons were made with non-parametric Mann-Whitney U and Wilcoxon Signed-Rank tests, respectively. Fisher's exact test was used for categorical data. p < 0.05 was considered statistically significant; where multiple comparisons were performed this p-value was adjusted using the Holm-Bonferroni method (adjusted p-values are denoted p h ; Holm, 1979). Box plots show median (central line), interquartile range (box) and 10-90% range (whiskers).

NMDAR-mediated EPSCs are prolonged in BACHD STN neurons
As described above, the majority of studies report that astrocytic glutamate uptake is diminished in the striatum in HD and its models. To test whether the STN of BACHD mice exhibits a similar deficit, EPSCs arising from the optogenetic stimulation of motor cortical inputs to the STN (as described by Chu et al., 2015) were compared in WT and BACHD mice before and after inhibition of GLT-1 and GLAST with 100 nM TFB-TBOA. STN neurons were recorded in ex vivo brain slices in the whole-cell voltage-clamp configuration using a cesium-based, QX-314-containing internal solution to maximize voltage control. Neurons were held at À40 mV and recorded in the presence of low (0.1 mM) external Mg 2+ and the AMPAR antagonist DNQX (20 mM) to maximize and pharmacologically isolate the evoked NMDAR-mediated excitatory postsynaptic current (EPSC); analysis was performed on average EPSCs from 5 trials with 30 s recovery between trials ( Figure 1D-H). The amplitude weighted decay time constant of the NMDAR EPSC was moderately but significantly prolonged in BACHD compared to n = 12;] ms; n = 11; p = 0.0455; Figure 1E-H). Subsequent application of TFB-TBOA increased the decay time constant of the NMDAR EPSC in STN neurons derived from WT .0] ms; ] ms; n = 9; p = 0.0039; Figure 1E,H) but had no effect in BACHD neurons ] ms; BACHD TFB-TBOA: 44.9 [34.7-52.2] ms; n = 10; p = 0.3223; Figure 1F,H). In control conditions, the amplitudes of EPSCs recorded from WT and BACHD neurons were similar (WT: 50.1 [34.7-61.0] pA; n = 12;  pA; n = 11; p h = 0.7399; Figure 2A) and there was no correlation between EPSC amplitude and the decay time constant in either group (WT: r 2 = 0.16; n = 12; p h = 0.5871; BACHD: r 2 = 0.10; n = 12; p h = 0.6686; Figure 2B). In order to increase spillover of glutamate from synaptic release sites, cortico-STN inputs were optogenetically stimulated 5 times at 50 Hz and the resulting compound NMDAR-mediated EPSC was compared in WT and BACHD STN neurons. Interestingly, the decay of compound NMDAR EPSCs under control conditions or following inhibition of glutamate uptake were not different in WT and BACHD mice   Figure 2C-E). Together, these data demonstrate that individual, but not compound, NMDAR-mediated cortico-STN synaptic EPSCs are prolonged in the BACHD model. after (gray) inhibition of astrocytic glutamate uptake with 100 nM TFB-TBOA. Inset, the same EPSCs scaled to the same amplitude. (F) Examples of optogenetically stimulated NMDAR EPSCs from a BACHD STN neuron before (green) and after (gray) inhibition of astrocytic glutamate uptake with 100 nM TFB-TBOA. (G) WT (black, same as in E) and BACHD (green, same as in F) optogenetically stimulated NMDAR EPSCs overlaid and scaled to the same amplitude. (H) Boxplots of amplitude weighted decay show slowed decay kinetics of NMDAR EPSCs in BACHD STN neurons compared to WT, and that TFB-TBOA increased weighted decay in WT but not BACHD mice. *p < 0.05. ns, not significant. Data for panels B-C provided in Figure 1source data 1; data for panel H provided in Figure 1-source data 2. DOI: 10.7554/eLife.21616.002 The following source data is available for figure 1: Source data 1. Autonomous firing frequency and CV for BACHD and WT STN neurons in Figure 1B-C. DOI: 10.7554/eLife.21616.003 Source data 2. Amplitude weighted decay of NMDAR-mediated EPSCs in Figure 1H.

Blockade of NMDARs rescues the autonomous activity of BACHD STN neurons
To test whether disrupted autonomous firing in BACHD is linked to NMDAR activation, brain slices from BACHD mice were incubated in control media or media containing the NMDAR antagonist D-AP5 (50 mM) for 3-5 hr prior to loose-seal, cell-attached recordings from STN neurons ( Figure 3). D-AP5 treatment rescued autonomous firing in slices derived from 5-7 month old BACHD mice compared to untreated control slices ( Figure 3A Figure 3C) and incidence of firing (untreated: 24/27 (89%); D-AP5 treated 22/27 (81%); p = 0.7040; Figure 3D) were unaltered by D-AP5 treatment. Thus, prolonged blockade of NMDARs rescued autonomous firing in BACHD STN neurons but had no effect on autonomous activity in WT STN neurons. Together, these data demonstrate that NMDAR activation contributes to the disruption of autonomous activity in BACHD STN neurons.
The mitochondria of BACHD STN neurons are subject to elevated oxidant stress NMDAR activation can elevate mitochondrial oxidant stress (Dugan et al., 1995;Moncada and Bolaños, 2006;Brennan et al., 2009;Nakamura and Lipton, 2011). To test whether STN neurons from BACHD mice exhibit increased mitochondrial oxidant stress, a mitochondria-targeted redox probe MTS-roGFP (Hanson et al., 2004) was virally expressed in 5-7-month-old BACHD mice and WT littermates ( Figure 4A). 1-2 weeks later MTS-roGFP was imaged in brain slices under two-photon excitation with 890 nm light. Oxidant stress was estimated from the fluorescence of MTS-roGFP in individual neurons under baseline conditions relative to the fluorescence of MTS-roGFP under conditions of full reduction and oxidation in the presence of 2 mM dithiothreitol and 200 mM aldrithiol-4, respectively (Sanchez-Padilla et al., 2014). STN neurons were selected for analysis based on their appearance under two-photon, Dodt contrast imaging and were distinguishable from STN glial cells by their relatively large diameter ( Figure 4A). STN neurons are reliably recorded when this selection strategy is employed to guide patch clamp recording (Atherton et al., 2008(Atherton et al., , 2010. MTS-roGFP imaging revealed that relative oxidant stress in BACHD STN neurons was elevated compared optogenetic stimulation from a WT STN neuron (C) before (black) and after (gray) inhibition of astrocytic glutamate uptake with 100 nM TFB-TBOA and a BACHD STN neuron (D) before (green) and after (gray) TFB-TBOA application. (E) Line segment plots of amplitude weighted decay of compound NMDAR EPSCs before and following TFB-TBOA. The decays of compound NMDAR ESPCs were similar in WT and BACHD before TFB-TBOA application. In addition, inhibition of astrocytic glutamate uptake prolonged the decay of compound NMDAR ESPCs in all neurons tested. ns, not significant. Data for panels A-B provided in Figure 2-source data 1; data for panel E provided in Figure 2-source data 2. DOI: 10.7554/eLife.21616.005 The following source data is available for figure 2: Source data 1. Amplitude and amplitude weighted decay of NMDAR-mediated EPSCs in Figure 2A Figure 4B). In a separate experiment (performed 15 months later) to test whether NMDAR activation ex vivo contributed to elevated mitochondrial oxidant stress, brain slices from a different cohort of BACHD mice were treated for >3 hr in 50 mM D-AP5 prior to imaging. MTS-roGFP imaging confirmed that D-AP5treated slices exhibited lower oxidant stress compared to untreated slices from the same mice (untreated: 0.39 [0.35-0.48]; n = 13; D-AP5-treated: 0.32 [0.24-0.42]; n = 17; p = 0.0445; Figure 4C). Thus, STN neurons from BACHD mice exhibit elevated mitochondrial oxidant stress, which can be reduced by antagonism of NMDARs.
Impaired autonomous activity of STN neurons in BACHD mice is due to increased activation of K ATP channels NMDAR receptor-generated mitochondrial oxidant stress in BACHD may lead to the activation of K ATP channels, which act as metabolic sensors and homeostatic regulators of excitability in multiple cell types (Nichols, 2006). STN neurons abundantly express all the molecular components of K ATP channels including the Kir6.2 pore-forming subunit of the K ATP channel (Thomzig et al., 2005) and the SUR1, SUR2A and SUR2B regulatory subunits (Karschin et al., 1997;Zhou et al., 2012). To determine whether K ATP channels are responsible for impaired firing in BACHD mice, the effects of the K ATP channel inhibitor glibenclamide (100 nM) on WT and phenotypic BACHD autonomous firing ex vivo were compared. Glibenclamide application increased both the rate (BACHD control: 5.7 To further examine the effects of K ATP channels on autonomous firing, whole-cell current clamp recordings were obtained from 5-7-month-old BACHD mice and WT littermates ( Figure 5C,D). Consistent with the hyperpolarizing and shunting effects of K ATP channels, the interspike voltage trajectory was shallower in BACHD neurons compared to WT ( Examples of whole-cell recordings from WT and BACHD STN neurons before (left) and after (right) inhibition of K ATP channels with 100 nM glibenclamide. (D) Population data (from 5-7-month-old mice). The interspike voltage trajectory was lower in BACHD neurons compared to WT. K ATP channel inhibition increased the interspike voltage trajectory in BACHD neurons but had no effect in WT. As with cell-attached recordings, inhibition of K ATP channels had mixed effects on firing in WT neurons, whereas in BACHD mice inhibition of K ATP channels increased the frequency and regularity of firing. *p < 0.05. ns, not significant. Data for panel B provided in Figure 5-source data 1; data for panel D provided in Figure 5-source data 2. DOI: 10.7554/eLife.21616.013 The following source data is available for figure 5: Source data 1. Autonomous firing frequency and CV for WT and BACHD STN neurons under control conditions and following glibenclamide application in Figure 5B. DOI: 10.7554/eLife.21616.014 Source data 2. Autonomous interspike voltage trajectory, firing frequency and CV for whole-cell recordings from WT and BACHD STN neurons in Figure 5D. DOI: 10.7554/eLife.21616.015 mV/s; n = 7; WT glibenclamide interspike voltage trajectory: 361.9 [216.7-522.9] mV/s; n = 7; p h = 0.3750; Figure 5C,D).

NMDAR activation produces a persistent K ATP channel-mediated disruption of autonomous activity in WT STN neurons
To further examine whether elevated NMDAR activation can trigger a homeostatic K ATP channelmediated reduction in autonomous firing in WT STN, brain slices from 2-month-old C57BL/6 mice were incubated in control media or media containing 25 mM NMDA for 1 hr prior to recording (Figure 7).  Figure 6. The abnormal autonomous activity of STN neurons in BACHD mice is rescued by inhibition of K ATP channels with gliclazide. (A) Examples of loose-seal cell-attached recordings of a STN neuron from a 6-month-old BACHD mouse before (upper) and after (lower) inhibition of K ATP channels with 10 mM gliclazide. (B) Population data (5-7-month-old). In BACHD STN neurons inhibition of K ATP channels with gliclazide increased the frequency and regularity of firing. *p < 0.05. Data for panel B provided in Figure 6-source data 1. DOI: 10.7554/eLife.21616.016 The following source data is available for figure 6: Source data 1. Autonomous firing frequency and CV for WT and BACHD STN neurons under control conditions and following gliclazide application in Figure 6B. DOI: 10.7554/eLife.21616.017 66; NMDA CV: 0.24 [0.10-0.72]; n = 65; p h = 0.0150; Figure 7A-C) of autonomous activity relative to control slices. In a subset of neurons glibenclamide was applied to inhibit K ATP channels. In neurons from untreated slices glibenclamide had no effect on firing rate (control: 16.6 [10.9-31.3] Hz; glibenclamide: 25.0 [16.3-32.8] Hz; n = 6; p h = 0.2188; Figure 7A-D) or regularity (control CV: 0.08 [0.07-0.37]; glibenclamide CV: 0.08 [0.06-0.09]; n = 6; p h = 0.3125; Figure 7A-D). However, in neurons from NMDA pre-treated slices glibenclamide application elevated firing rate (control: 3.3 [2.3-5.1] Hz; glibenclamide: 11.4 [10.8-24.4] Hz; n = 10; p h = 0.0078; Figure 7A-D) and regularity (control CV: 0.83 [0.25-1.03]; glibenclamide CV: 0.12 [0.07-0.16]; n = 8, p h = 0.0208; Figure 7A-D) to levels similar to that seen in untreated slices. Together, these data demonstrate that increased activation of STN NMDARs can lead to a persistent K ATP channel-mediated homeostatic reduction in autonomous activity in STN neurons. (D) Inhibition of K ATP channels with 100 nM glibenclamide restored firing in STN cells from NMDA pre-treated slices but had no effect on firing from naïve slices. *p < 0.05. ns, not significant. Data for panel C provided in Figure 7-source data 1; data for panel D provided in Figure 7-source data 2. DOI: 10.7554/eLife.21616.018 The following source data is available for figure 7: Source data 1. Autonomous firing frequency and CV for control and NMDA pre-treated C57BL/6 STN neurons in Figure 7C. DOI: 10.7554/eLife.21616.019 Source data 2. Autonomous firing frequency and CV for control and NMDA pre-treated C57BL/6 STN neurons in control conditions and following glibenclamide application in Figure 7D.  -Vizi, 2005). Superoxide and hydrogen peroxide can also be produced by NADPH oxidase (Brennan et al., 2009). Because K ATP channels are activated by H 2 O 2 (Ichinari et al., 1996;Avshalumov et al., 2005), we tested whether H 2 O 2 contributes to K ATP channel-mediated disruption of ex vivo autonomous activity in the BACHD model. The effect of a membrane permeable form of the enzyme catalase (polyethylene glycol-catalase), which breaks down H 2 O 2 , on the autonomous firing of STN neurons from 4-6-month-old BACHD mice was examined ( Figure 8). To test if the actions of H 2 O 2 on autonomous firing are confined to its effects on K ATP channels, these experiments were repeated in the presence of glibenclamide. As seen previously, application of glibenclamide increased firing rate (BACHD control: 5.2 [1.0-6.7] Hz; BACHD glibenclamide: 8.5 [7.2-11.6] Hz; n = 8, p h = 0.0156; Figure 8D) and regularity (BACHD control CV: 0.83 [0.27-1.30]; BACHD glibenclamide CV: 0.23 [0.15-0.58]; n = 8; p h = 0.0156; Figure 8D). Subsequent application of catalase had no additional effect on firing rate (8.7 [7.2-14.1] Hz; n = 8; p h = 0.6406; Figure 8D) but did produce a small but statistically significant increase in regularity (CV: 0.14 [0.11-0.21]; n = 8; p h = 0.0156; Figure 8D). In WT mice, catalase application did not change firing rate (WT control: 11.0 [10.5-14.2] Hz; WT catalase: 14.3 [11.3-18.3] Hz; n = 7; p = 0.0781; Figure 9) but lead to a small but statistically significant increase in regularity (WT control CV: 0.12 [0.10-0.23]; WT catalase CV: 0.11 [0.07-0.13]; n = 7; p = 0.0469; Figure 9). The effects of catalase on the frequency and regularity of firing in BACHD mice were greater than those observed in WT mice (frequency: p = 0.0154; CV: p = 0.0007; Figure 9). Together, these data suggest that suppression of autonomous activity of STN neurons in BACHD mice is largely mediated by the modulatory effect of H 2 O 2 on K ATP channels.
To test if the elevation of oxidant stress can result in K ATP channel activation in WT STN neurons, glutathione peroxidase was inhibited with mercaptosuccinic acid (MCS) (Avshalumov et al., 2005).  Figure 10). These data are also consistent with an action of H 2 O 2 on STN K ATP channels.

The STN degenerates progressively in BACHD mice
HD patients exhibit 20-30% STN neuron loss (Lange et al., 1976;Guo et al., 2012). Because mitochondrial oxidant stress and reactive oxygen species can trigger apoptotic pathways leading to cell death (Green and Reed, 1998;Bossy-Wetzel et al., 2008), the number of STN neurons in 12month-old BACHD and WT mice were compared ( Figure 11). The brains of BACHD mice and WT littermates were first fixed by transcardial perfusion of formaldehyde, sectioned into 70 mm coronal slices and immunohistochemically labeled for neuronal nuclear protein (NeuN). The total number of NeuN-immunoreactive STN neurons and the volume of the STN were then estimated using unbiased stereological techniques. Both the total number of STN neurons (WT: 10,793 [9,070-11,545]; n = 7; BACHD: 7,307 [7,047-9,285]; n = 7; p = 0.0262) and the volume of the STN (WT: 0.087 [0.084-0.095] mm 3 ; n = 7; BACHD: 0.078 [0.059-0.081] mm 3 ; n = 7; p = 0.0111; Figure 11A,B) were reduced in 12-month-old BACHD versus WT mice. The density of STN neurons was not different in BACHD and WT mice (WT: 121,248 [107,139] neurons/mm 3 ; n = 7; BACHD: 115,273 [90,765] neurons/mm 3 ; n = 7; p = 0.8048; Figure 11A,B). To determine whether the difference in cell number represents an early developmental abnormality or a progressive loss of adult neurons, the numbers of neurons in 2-month-old BACHD and WT mice were also compared. At 2months-old, the total number of STN neurons (WT: 10,373 [9,341-14,414]; n = 7; BACHD: 10,638 [10,513-13,877]; n = 7; p = 0.7104; Figure 11C), the volume of the STN (WT: 0.098 [0.090-0.125] mm 3 ; n = 7; BACHD: 0.085 [0.080-0.111] mm 3 ; n = 7; p = 0.1649; Figure 11C) and STN neuronal density (106,880 [98,100-115,985] neurons/mm 3 ; n = 7; BACHD: 124,844 [115,711] neurons/mm 3 ; n = 7; p = 0.1282; Figure 11C) were not different in WT and BACHD mice. Together, these data demonstrate that between the ages of 2 months and 12 months BACHD mice lose approximately one third of their STN neurons compared to WT littermates.  The STN of Q175 KI mice exhibits similar abnormalities to those observed in the BACHD model STN neurons from BACHD mice exhibit perturbed autonomous firing that is caused by NMDAR activation/signaling leading to mitochondrial oxidant stress, H 2 O 2 generation and K ATP channel activation. Furthermore, STN neurons are progressively lost in BACHD mice. To determine whether these features are specific to the BACHD model or a more general feature of HD models, a subset of experiments were repeated in heterozygous Q175 KI mice ( Figure 12). STN neurons from 6-monthold Q175 mice exhibited a severely reduced rate of autonomous activity (WT: 7.8 [1.9-14.7] Hz; n = 90; Q175: 0.0 [0.0-6.3] Hz; n = 90; p < 0.0001; Figure 12A Figure 12E,F) and proportion of active neurons (Q175 control: 30/45 (67%); Q175 D-AP5 treated: 33/45 (73%); p = 0.6460; Figure 12E,F) were unaltered. The 12-month-old Q175 STN (n = 7) exhibited a median 26% reduction in the total number of STN neurons with no effect on other parameters (WT: 8,661 [7,376] Figure 8C). Break down of H 2 O 2 elevated autonomous firing in BACHD STN neurons only. The boxplot confirms that the elevation of firing due to catalase application was greater in BACHD mice. (B) Line plots illustrating a small but statistically significant effect of catalase on the regularity of autonomous action potential generation in STN neurons from WT mice (black) compared to a larger increase in regularity following catalase application in BACHD neurons (green; BACHD data same as in Figure 8C). The boxplot confirms that the increase in regularity due to catalase was greater in BACHD mice. *p < 0.05. ns, not significant. Data provided in Figure 9-source data 1. DOI: 10.7554/eLife.21616.023 The following source data is available for figure 9: Source data 1. Autonomous firing frequency and CV for WT and BACHD STN neurons under control conditions and following catalase application in Figure 9. DOI: 10.7554/eLife.21616.024 [63,624-103,020] neurons/mm 3 ; p = 0.2086; Figure 12G,H). Taken together, these data show that the STN exhibits similar dysfunction and neuronal loss in both the transgenic BACHD and Q175 KI mouse models of HD.

Discussion
Dysfunction of the striatum and cortex has been extensively characterized in HD models, but relatively few studies have examined the extra-striatal basal ganglia. Here, we report early NMDAR, mitochondrial and firing abnormalities together with progressive loss of STN neurons in two HD mouse models. Furthermore, dysfunction was present in HD mice prior to the onset of major symptoms, implying that it occurs early in the disease process (Gray et al., 2008;Menalled et al., 2012). Cell death in the STN also preceded that in the striatum, as STN neuronal loss was observed at 12 months of age in both BACHD and Q175 mice, a time point at which striatal neuronal loss is absent but psychomotor dysfunction is manifest (Gray et al., 2008;Heikkinen et al., 2012;Smith et al., 2014;Mantovani et al., 2016). Together these findings argue that dysfunction within the STN contributes to the pathogenesis of HD. Astrocytes appear to play a pivotal role in HD. Expression of mutant huntingtin in astrocytes alone is sufficient to recapitulate neuronal and neurological abnormalities observed in HD and its models (Bradford et al., 2009;Faideau et al., 2010). Furthermore, astrocyte-specific rescue approaches ameliorate some of the abnormalities observed in HD models (Tong et al., 2014;Oliveira et al., 2016). In the STN, inhibition of GLT-1 (and GLAST) slowed individual NMDAR EPSCs in WT but not BACHD mice and eliminated the differences in their decay kinetics, arguing that impaired uptake of glutamate by astrocytes contributed to the relative prolongation of NMDARmediated EPSCs in BACHD STN neurons. Interestingly, and in contrast to the striatum (Milnerwood et al., 2010), when spillover of glutamate onto extrasynaptic receptors was increased by train stimulation and inhibition of astrocytic glutamate uptake, the resulting compound NMDAR EPSC and its prolongation by uptake inhibition were similar in BACHD and WT mice, arguing against The following source data is available for figure 10: Source data 1. Autonomous firing frequency and CV for WT and BACHD STN neurons under control conditions and following MCS and glibenclamide application in Figure 10B. DOI: 10.7554/eLife.21616.026 an increase in extrasynaptic STN NMDAR expression/function in BACHD mice. Slowing of astrocytic glutamate uptake has recently been shown to occur during increased presynaptic activity (Armbruster et al., 2016). Thus, train stimulation may have slowed glutamate uptake sufficiently to occlude/eliminate the differences in uptake that were observed in BACHD and WT STN neurons during single stimulation. Whether the modest differences in glutamate uptake that The following source data is available for figure 11: Source data 1. BACHD STN neuron counts, density and STN volume in Figure 11B were observed here are sufficient to promote NMDAR-mediated dysfunction in HD STN neurons remains to be determined. NMDARs play a key role in the abnormal activity of STN neurons in HD models. Antagonism of STN NMDARs in BACHD and Q175 brain slices rescued autonomous STN firing. Conversely, acute activation of STN NMDARs persistently disrupted STN firing in WT brain slices. If the relatively low level of glutamatergic transmission present ex vivo is sufficient to impair firing then this impairment is likely to be more severe in vivo where STN neurons are powerfully patterned by glutamatergic transmission arising from the cortex, thalamus, pedunculopontine nucleus and superior colliculus (reviewed by Bevan, 2017). Non-synaptic sources of extracellular glutamate, such as diffusion/ release from astrocytes (Cavelier and Attwell, 2005;Lee et al., 2013) may also contribute to excessive NMDAR activation in HD mice.
Extended antagonism of NMDARs in BACHD slices also reduced mitochondrial oxidant stress in STN neurons. NMDAR activation can elevate ROS through a variety of Ca 2+ -and nitric oxide-associated signaling pathways and their actions on mitochondria, NADPH oxidase and antioxidant expression (Dugan et al., 1995;Moncada and Bolaños, 2006;Brennan et al., 2009;Nakamura and Lipton, 2011;Valencia et al., 2013). Although we saw no evidence of basal mitochondrial dysfunction that was not attributable to increased NMDAR function, there is considerable evidence that mutant huntingtin causes transcriptional dysregulation, which leads to defective mitochondrial quality control, an increase in the proportion of defective, ROS generating mitochondria and an increase in opening of the permeability transition pore (Milakovic and Johnson, 2005;Panov et al., 2002;Fernandes et al., 2007;Song et al., 2011;Chaturvedi et al., 2013;Johri et al., 2013;Martin et al., 2015). Thus, basal mitochondrial dysfunction could render HD STN neurons especially sensitive to NMDAR-mediated transmission and signaling.
Catalase rapidly restored autonomous firing in the BACHD model, an effect occluded by inhibition of K ATP channels, arguing that H 2 O 2 , through its action on K ATP channels is the major cause of firing disruption. H 2 O 2 can act on K ATP channels by decreasing their sensitivity to ATP (Ichinari et al., 1996), reducing the ratio of ATP to ADP (Krippeit-Drews et al., 1999), and/or modulating channel gating through a sGC-cGMP-PKG-ROS(H 2 O 2 )-ERK1/2-calmodulin-CaMKII signaling pathway (Zhang et al., 2014). H 2 O 2 is likely to directly modulate STN K ATP channels in HD mice because disrupted firing was also observed when STN neurons were recorded in the whole-cell configuration with patch pipettes containing exogenous ATP. Furthermore, H 2 O 2 break down rapidly rescued activity, consistent with a direct action on K ATP channels. H 2 O 2 -dependent modulation of K ATP channels has been extensively characterized in midbrain dopamine neurons where it powerfully suppresses cellular excitability and synaptic transmission (Avshalumov et al., 2005;Bao et al., 2009). The activation of K ATP channels in STN neurons may represent a form of homeostasis that suppresses firing when mitochondrial oxidant stress is high, limiting the possibility of oxidant damage and bioenergetic failure (Ray et al., 2012;Sena and Chandel, 2012).  (center), and neuron density (right) in 12-month-old Q175 mice. The number of STN neurons was lower in Q175 mice compared to WT. *p < 0.05. ns, not significant. Data for panel B provided in Figure 12-source data 1; data for panel D provided in Figure 12-source data 2; data for panel F provided in Figure 12-source data 3; data for panel H provided in Figure 12-source data 4. DOI: 10.7554/eLife.21616.029 The following source data is available for figure 12: Source data 1. Autonomous firing frequency and CV for Q175 and WT STN neurons in Figure 12B. DOI: 10.7554/eLife.21616.030 Source data 2. Autonomous firing frequency and CV for Q175 in control conditions and following glibenclamide application Figure 12D. DOI: 10.7554/eLife.21616.031 Source data 3. Autonomous firing frequency and CV for control and D-AP5 pre-treated Q175 STN neurons in Figure 12F. DOI: 10.7554/eLife.21616.032 Source data 4. Q175 STN neuron counts, density and STN volume in Figure 12H. DOI: 10.7554/eLife.21616.033 In HD, chronic oxidant stress can lead to damage, such as lipid and protein peroxidation and nuclear/mitochondrial DNA damage, which profoundly impair cellular function and promote cell death (Perluigi et al., 2005;Browne and Beal, 2006;Acevedo-Torres et al., 2009). Consistent with the negative effects of such processes on neuronal viability, we observed progressive loss of STN neurons in both the BACHD and Q175 models. Furthermore, the level of neuronal loss at 12 months in the BACHD and Q175 models was similar to that observed in HD patients (Lange et al., 1976;Guo et al., 2012). The absence of neuronal loss in the cortex and striatum in the same models at an equivalent time point suggests that STN dysfunction and degeneration may be particularly influential in the early disease process. Although the STN is known to degenerate in HD, it is not clear why neuronal loss is ultimately less than that observed in the striatum at the end stage of the disease, despite the fact that dysfunction and degeneration occur earlier (at least in HD models). Future research will be required to determine whether subtypes of STN neurons exhibit selective vulnerability and/or whether the processes promoting their degeneration, e.g. cortical activation of STN NMDARs, ultimately wane.
As a key component of the hyperdirect and indirect pathways, the STN is critical for constraining cortico-striatal activity underlying action selection (Albin et al., 1989;Oldenburg and Sabatini, 2015). In the 'classical' model of basal ganglia function, degeneration of indirect pathway striatal projection neurons is proposed to underlie the symptoms of early stage HD (Albin et al., 1989).
Here we show for the first time that STN dysfunction and neuronal loss precede cortico-striatal abnormalities in HD models. Thus, dysfunction and degeneration of cortical and striatal neurons occurs in concert with profound changes in other elements of the basal ganglia. Therapeutic strategies that target the STN may therefore be useful not only for treating the psychomotor symptoms of early-to mid-stage HD but also for influencing dysfunction and degeneration throughout the cortico-basal ganglia-thalamo-cortical circuit.

Animals
All animal procedures were performed in accordance with the policies of the Society for Neuroscience and the National Institutes of Health, and approved by the Institutional Animal Care and Use Committee of Northwestern University. Adult male hemizygous BACHD mice (RRID:IMSR_JAX: 008197) and heterozygous Q175 mice (RRID:IMSR_JAX:027410), their WT litter mates, and C57BL/6 mice (Charles River Laboratories International, Inc., Wilmington, MA, USA) were used in this study.

Two-photon imaging
MTS-roGFP-expressing neurons were imaged at 890 nm with 76 MHz pulse repetition and~250 fs pulse duration at the sample plane. Two-photon excitation was provided by a G8 OPSL pumped Mira 900 F laser (Coherent, Santa Clara, CA, USA) and sample power was regulated by a Pockels cell electro-optic modulator (model M350-50-02-BK, Con Optics, Danbury, CT, USA). Images were acquired using an Ultima 2 P system running PrairieView 5 (Bruker Nano Fluorescence Microscopy, Middleton, WI, USA) and a BX51WI microscope (Olympus, Tokyo, Japan) with a 60 Â 0.9 NA objective (UIS1 LUMPFL; Olympus). After baseline fluorescence had been measured, the maximum and minimum fluorescence were determined by the application of 2 mM dithiothreitol and then 200 mM aldrithiol-4 to fully reduce and oxidize the tissue, respectively. The relative oxidation at baseline, a measure of oxidative stress, was then calculated (Sanchez-Padilla et al., 2014).

Immunohistochemistry and stereology
Mice were lightly anesthetized with isoflurane, deeply anesthetized with ketamine/xylazine (87/13 mg/kg i.p.) and then perfused transcardially with~5 ml of phosphate buffered saline (PBS) followed by 30 ml of 4% formaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were removed and postfixed for 2 hr in 4% formaldehyde, then washed in PBS. Brains were blocked and 70 mm thick coronal sections containing the STN were cut using a vibratome (VT1000S; Leica). Sections were washed in PBS and incubated for 48 hr at 4˚C in anti-NeuN (clone A60; MilliporeSigma, Darmstadt, Germany; RRID:AB_2298772) at 1:200 in PBS with 0.2% Triton X-100 (MilliporeSigma) and 2% normal donkey serum. Sections were then washed in PBS and incubated for 90 min at room temperature in Alexa Fluor 488 donkey anti-mouse IgG (1:250; Jackson Immunoresearch, West Grove, PA, USA; RRID:AB_ 2340846) in 0.2% Triton X-100 and 2% normal donkey serum. Then the sections were washed in PBS and mounted on glass slides in Prolong Gold anti-fade medium (Thermo Fisher Scientific, Waltham, MA, USA).
NeuN labeled sections were imaged using an Axioskop two microscope (Carl Zeiss) with a 100 Â 1.3 NA oil immersion objective (Plan-Neofluar 1018-595; Carl Zeiss). Unbiased stereological counting of STN neurons in a single hemisphere was performed using the optical fractionator technique (West et al., 1991) as implemented in Stereo Investigator (MBF Bioscience, Williston, VT, USA; RRID:SCR_002526), using a counting frame of 50 mm Â 50 mm Â 8 mm and a grid size of 150 mm Â 150 mm; all sections containing the STN were used for counting (~8 sections). STN volume was calculated from the sum of the areal extent of the STN on each section multiplied by the section thickness (70 mm). For all individual counts the Gundersen Coefficient of Error (CE) (Gundersen et al., 1999) was less than 0.1 (0.080 [0.075-0.090]), and the investigator performing the counting was blinded to the genotype of the mouse.

Data analysis and statistics
Electrophysiological data were analyzed using routines running in Igor Pro 6 and 7 (Wavemetrics, Portland, OR, USA; RRID:SCR_000325) or matplotlib (Hunter, 2007;RRID:SCR_008624). The firing rate of STN neurons was calculated from 1 min of recording or 100 action potentials (whichever covered the longer time period). Imaging data were analyzed using FIJI (Schindelin et al., 2012;RRID: SCR_002285). Statistical analyses were performed in Prism 5 (GraphPad Software, San Diego, CA, USA; RRID:SCR_002798) or R (http://www.r-project.org/; RRID:SCR_001905, RRID:SCR_000432). In order to make no assumptions about the distribution of the data, non-parametric statistical tests were used, and data are reported as median [interquartile range]; outliers were not excluded from the analysis. An a-level of 0.05 was used for two-way statistical comparisons performed with the Mann-Whitney U test for unpaired data (represented with box plots), the Wilcoxon signed rank test for paired data (tilted line segment plots), Fisher's exact test for categorical data (bar plots) or the F-test for linear regression. Where datasets were used in multiple comparisons the p-value was adjusted to maintain the family-wise error rate at 0.05 using the Holm-Bonferroni method (Holm, 1979); adjusted p-values are denoted p h . Box plots show median (central line), interquartile range (box) and 10-90% range (whiskers). For the primary findings reported in the manuscript, sample sizes for Mann-Whitney and Wilcoxon tests were estimated to achieve a minimum of 80% power using formulae described by Noether (1987). The effect sizes used in these power calculations were estimated using data randomly drawn from uniform distributions (runif() function in R stats package). For Mann-Whitney tests, with a 50 percentile change in median between groups X and Y (the interquartile ranges of the groups don't overlap) P(Y > X) » 0.88 giving an estimation that at least 10 observations per group would be needed to achieve 80% power; for a 25 percentile change (the median of Y falls outside the interquartile range of X) P(Y > X) » 0.72 and the estimated requirement is at least 27 observations per group. For Wilcoxon tests, if all pairs of observations show the same direction of change, P(X + X' > 0)= 1 giving an estimation that at least 10 observations would be needed to achieve 80% power (note though that it is possible to show empirically that 6 observations gives 100% power in this case); if 90% of observations show the same direction of change, P(X + X' > 0) » 0.98 and the estimated requirement is at least 12 pairs of observations. Ethics Animal experimentation: This study was performed in accordance with the policies of the Society for Neuroscience and the National Institutes of Health. All animals were handled according to approved Institutional Animal Care and Use Committee protocols (IS00001185) of Northwestern University. All procedures were performed under isoflurane or ketamine/xylazine anesthesia, and every effort was made to minimize suffering.