Glutathione peroxidase activity is neuroprotective in models of Huntington’s disease

Abstract

Huntington’s disease (HD) is a fatal neurodegenerative disorder caused by a CAG repeat expansion encoding a polyglutamine tract in the huntingtin (htt) protein¹. Here, we report a genome-wide overexpression suppressor screen which identified 317 open reading frames that ameliorate toxicity of a mutant htt fragment in yeast, and play roles in diverse cellular processes, including mitochondrial import and copper metabolism. Two suppressors encode glutathione peroxidases (GPxs), conserved antioxidant enzymes that reduce hydrogen peroxide and lipid hydroperoxides². Using genetic and pharmacological approaches in yeast, mammalian cells and Drosophila, we find that GPx activity robustly ameliorates HD-relevant metrics and is more protective than other antioxidant approaches tested. Importantly, we observe that GPx activity – unlike many antioxidant treatments – does not inhibit autophagy, an important mechanism for clearing mutant htt. As previous clinical trials indicate GPx mimetics are well-tolerated in humans, this study may have important implications for the treatment of HD.

Genetic and chemical screens have employed yeast models of Huntington’s disease (HD) to isolate gene deletion suppressors, gene deletion enhancers and chemical modifiers of mutant htt toxicity^3-6, elucidating cellular pathways involved in mutant htt toxicity and identifying several promising therapeutic targets and compounds. Here we have complemented this work via a systematic screen of the Yeast ORF Collection (Open Biosystems)⁷, which contains the vast majority of the fungus’ open reading frames (ORFs >5500/6607) in expression vectors maintained in the parental Y258 yeast strain. We transformed a construct encoding a mutant htt fragment (Htt103Q) which forms aggregates and causes cellular toxicity⁸ into each ORF strain, and tested for suppression of Htt103Q-dependent toxicity via growth assays (Figure 1a,b).

Figure 1.

Suppression of Htt103Q toxicity in yeast by ORF overexpression. (a) Schematic representation of genome-wide systematic ORF suppressor screen in yeast. (b) Suppression of Htt103Q toxicity in yeast by GPX1, GPX3, and mGpx1. Viability of the parental strain, Y258, and cells overexpressing yeast GPxs was determined using spotting assays. Equal numbers of cells were serially diluted five-fold and plated on media containing glucose to assess cell numbers and galactose to induce expression of Htt103Q and the indicated ORF. (c) Network analysis shows a high degree of interconnectivity amongst the ORF suppressors. The network data was analyzed with an iterative minimum filter of 2. Functional groups are color-coded as indicated. A total of 57 nodes with 115 edges define this network.

We identified 317 ORFs that suppressed Htt103Q toxicity in yeast (Supplementary Table 1), 172 (~54%) of which are functionally annotated in the Saccharomyces Genome Database, with 41 having one-to-one human orthologs (Table 1). Analysis of the functionally annotated genes with the DAVID bioinformatics package⁹ revealed a number of enriched groups (Table 2), including mitochondrial import, copper chaperone activity, and purine biosynthesis. Interestingly, genes encoding for components of the mitochondrial protein import machinery – critical for import of nuclear-encoded proteins to the mitochondrial membranes and matrix¹⁰ – were the largest significantly enriched group (P<0.001) (Table 2, Supplementary Figure 1). Impaired mitochondrial protein import due to the amyloid beta peptide (Aβ) has been observed in Alzheimer’s disease (AD)¹¹, suggesting that mutant htt may cause a similar dysfunction in yeast.

Table 1.

Yeast ORF suppressors of Htt103Q toxicity with one-to-one human orthologs.

ORF	Human Ortholog	Description
ADE13	ADSL	Adenylosuccinate lyase
ADE4	PPAT	Amidophosphoribosyltransferase
ADK1	AK2	Adenylate kinase cytosolic
AIM2	CMBL	Cysteine hydrolase of the dienelactone hydrolase family
AIM29	C2ORF76	Unknown function
AIM45	ETFA	Probable electron transfer flavoprotein subunit alpha
BTS1	GGPS1	Geranylgeranyl pyrophosphate synthase
BUD17	PDXK	Bud site selection protein 17
BUD23	WBSCR22	Putative methyltransferase
BUD32	TP53RK	Serine/threonine-protein kinase
BUR6	DRAP1	Negative cofactor 2 complex subunit alpha
CAB4	COASY	Unknown function
CAF40	RQCD1	40 kDa CCR4-associated factor
CDC123	CDC123	Cell division cycle protein 123
COX3	J01415.32	Cytochrome c oxidase subunit 3
DMC1	DMC1	Meiotic recombination protein 13
ENT3	CLINT1	Epsin-3
FAU1	MTHFS	5-methyltetrahydrofolate cyclo-ligase
FUR1	UPRT	Uracil phosphoribosyltransferase
GAB1	PIGU	GPI trnsamidase component GAB1
GAL7	GALT	Galactose-1-phosphate uridylyltransferase
GCS1	ARFGAP1	ADP-ribosylation factor GTPase-activating protein GCS1
GDT1	TMEM165	Unknown function
HMF1	HRSP12	High dosage growth inhibitor
IRC24	SPR	Uncharacterised oxidoreductase
MET7	FPGS	Folylpolyglutamate synthase
MGS1	WRNIP1	DNA-dependant ATPase
NSA2	TINP1	Unknown function
RAD10	ERCC1	DNA repair protein RAD10
RPL33B	RPL35A	60S ribosomal subunit
RPL37B	RPL37	60S ribosomal subunit
RRP46	EXOSC5	Exosome complex exonuclease
SEE1	METTL10	Uncharacterised methyltransferase
SMX3	SNRPF	Small nuclear ribonucleoprotein F
TAD2	ADAT2	tRNA-specific adenosine deaminase subunit
TIM10	TIMM10	Mitochondrial import inner membrane translocase subunit
TIM22	TIMM22	Mitochondrial import inner membrane translocase subunit
TIM9	TIMM9	Mitochondrial import inner membrane translocase subunit
YER134c	MDP1	Magnesium-dependent phosphatase 1
YOL114c	ICT1	Unknown function
YOR283w	C12ORF5	Probable phosphoglycerate mutase

Human orthologs determined using g:Profiler, a web-based tool for functional interpretation of gene lists⁴⁴.

Table 2.

Annotation clusters significantly enriched amongst ORF suppressors of Htt103Q toxicity.

Annotation Cluster	Genes	P-value	Fold Enrichment
Disulfide bond	10	< 0.001	5.6
Protein localisation in mitochondrion	12	< 0.001	6.8
Mitochondrial membrane organization	7	< 0.001	6.2
Cytosol	10	< 0.001	5.0
Copper chaperone activity	4	< 0.001	23.9
Purine biosythesis	5	0.001	9.5
Metal-binding	32	0.001	1.8
Regulation of hydrolase activity	6	0.006	4.9
Cytosolic small ribosomal subunit	6	0.015	4.0
Clathrin coated vesicle	5	0.013	5.2
Lipid moiety-binding region: S-palmitoyl cysteine	4	0.030	5.8

Annotation clusters determined by the DAVID Functional Annotation tool. Count refers to number of genes within input gene list that fall into the specified cluster. P-value represents the threshold of EASE Score, a modified Fisher Exact P-Value, for gene-enrichment analysis. Enrichment category denotes the overall enrichment of the particular cluster over background.

To explore functional connectivity amongst the suppressor ORFs we performed network analysis using the Osprey Network Visualization System (Version 1.2.0)¹². Osprey employs the BioGRID database¹³ to build networks by inserting all known interactions for each gene of interest (“node”). The interactions (or “edges”) are derived from several sources, including affinity capture experiments and synthetic lethality analyses. Using this approach we observed a total of 166 interactions among 111 nodes (data not shown). To select genes with higher level relationships, the network data was processed with an iterative minimum filter of two, which identified all of the nodes which have a minimum of two interactions within the group (Figure 1c). This analysis uncovered a highly interconnected network of 57 nodes with 115 edges which reinforced and expanded the functionally enriched groups amongst the suppressors, including: protein localization in mitochondria (P=1.9 × 10⁻⁵), purine ribonucleotide biosynthetic process (P=6.4 × 10⁻³), copper ion transport (P=1.3 × 10⁻²), and regulation of GTPase activity (P=3.2 × 10⁻²). Intriguingly, this analysis suggests that while the suppressors fall into several functional categories, many interact in a common network.

We next compared the list of ORF suppressors identified here to the published modifier screens in yeast^3,6 which identified 28 gene deletion suppressors and 52 enhancers of mutant htt toxicity. As genes which enhance mutant htt toxicity when deleted might be expected to suppress toxicity when overexpressed, we were surprised to find only one gene in common between the ORF overexpression suppressors and the gene deletion enhancers – MMS4, which encodes a subunit of the Mms4-Mus81 endonuclease involved in DNA repair¹⁴ – and no overlap with the deletion suppressors. This might be explained by the earlier screens not being systematic, but instead representing random screens of pooled gene deletion strains. However, this observation may also suggest that protective genes need not necessarily play roles in the pathways contributing to mutant htt toxicity.

In addition to providing mechanistic insight into the nature of mutant htt toxicity, we have previously found that genetic screens in yeast can uncover novel candidate therapeutic targets for HD^3,15,16. Two of the most robust suppressors identified were GPX1 and GPX3, which encode antioxidant GPx enzymes (Figure 1b). Many studies with HD models and patient samples indicate that oxidative stress may play an important role in disease pathogenesis^3,17,18. Indeed, markers of oxidative stress are increased in the plasma of both symptomatic and pre-manifest HD patients¹⁹. We have also previously observed increased levels of ROS in yeast expressing Htt103Q³. Interestingly, GPX1 and GPX3 were the only antioxidant genes which emerged from our screen, suggesting that this class of antioxidant enzymes is particularly protective against mutant htt toxicity. While overall levels of oxidative stress are increased in HD patients, total GPx activity remains unchanged^20,21; thus supplementation of GPx activity may have therapeutic value.

We therefore investigated the efficacy of GPx activity in a mammalian cell model of mutant htt toxicity. First we expressed murine Gpx1 (mGpx1) – the most abundant mammalian GPx protein – in yeast and found it also suppresses Htt103Q toxicity (Figure 1b). We then tested mGpx1 in a neuron-like rat pheochromocytoma (PC12) cell line that inducibly expresses a mutant htt fragment (Htt103Q)²². Htt103Q expression in these cells results in several HD-relevant phenotypes, including formation of mutant htt aggregates and increased levels of caspase-3 activation, a marker for apoptosis²³. Htt103Q expression in PC12 cells did not significantly alter GPx activity, analogous to HD patient data (Figure 2a). We found that cells stably overexpressing mGpx1 exhibited a ~35% increase in total cellular GPx activity (P<0.0001; Figure 2a) and a ~50% reduction in caspase 3/7 activation in response to mutant htt expression (P<0.0001; Figure 2b). Thus, as in yeast cells, expression of GPx activity is protective in this mammalian cell model of HD.

Figure 2.

mGpx1 and ebselen improve HD-relevant phenotypes in PC12 cells. (a) Total GPx activity was measured in mGpx1 overexpressing or control cell lines, with (induced) or without (uninduced) Htt103Q expression. Overexpression of mGpx1 increases total GPx activity in PC12 cells (****P<0.0001 compared to uninduced control). N = 7 per condition. (b) Quantification of caspase 3/7 activation in response to Htt103Q expression in cells overexpressing mGpx1. Overexpression of mGpx1 reduces Htt103Q mediated caspase 3/7 activation (****P<0.0001). Values represent the ratio of caspase 3/7 activity in cells expressing Htt103Q for 72 hours versus cells not expressing Htt103Q. N = 12 (WT); N = 9 (mRFP); N = 8 (Gpx1) per condition. (c) Effect of the GPx mimetic ebselen on Htt103Q-mediated caspase activation. A single 10 μM dose of ebselen significantly reduced caspase 3/7 activation in Htt103Q expressing cells (****P<0.0001). Values represent the ratio of caspase 3/7 activity in cells expressing Htt103Q for 72 hours versus cells not expressing Htt103Q. Analysis performed via unpaired, two tailed Mann-Whitney test. N = 12 per condition. (d) mGpx1 overexpression (****P<0.0001) and ebselen treatment (****P<0.0001) significantly reduce Htt103Q-mediated ROS production (N = 9 per condition). Unless stated statistical analyses performed via one-way ANOVA with post-hoc tests for data in all panels (ns = not significant; **P<0.01). Data are shown as means ± standard errors of the mean (SEMs).

The availability of small selenocysteine-containing molecules that mimic GPx make supplementation of GPx activity therapeutically viable in HD²⁴. Ebselen, the best characterized of these compounds, is highly bio-available, readily crosses the blood-brain barrier²⁵, and has produced promising results in clinical trials for stroke²⁴ and noise induced hearing loss²⁶. We thus tested ebselen in HD cells and found a significant ~66% reduction of caspase 3/7 activation compared to control cells (P<0.0001)(Figure 2c).

We next examined if this rescue of mutant htt toxicity was due to the scavenging of reactive oxygen species (ROS). To ensure that any changes in ROS levels were unrelated to apoptosis we assayed cells 6 h after mutant htt induction, prior to significant increases in caspase activation (data not shown). Levels of ROS were 1.9-fold higher (P<0.0001) in PC12 cells expressing Htt103Q versus uninduced controls, and either ebselen treatment (P<0.0001) or expression of mGpx1 (P<0.0001) dramatically abrogated this increase (Figure 2d).

Consequently, we analysed mGpx1 in vivo by taking advantage of a widely-used Drosophila HD model expressing an exon 1 fragment of mutant htt (Htt93Qex1)²⁷. We drove expression of Htt93Qex1 using the UAS/GAL4 system either pan-neuronally with the elavGAL4 driver, or in circadian clock neurons using the PdfGAL4 driver. Expression of Htt93Qex1 with elavGAL4 generates several disease-relevant phenotypes, including neurodegeneration of photoreceptor cells (rhabdomeres) in the fly eye (Figure 3a,b). mGpx1 overexpression resulted in a ~35% neuroprotection (P<0.001) in the eye of 10 day old adult flies, as well as rescue at day 1 (P<0.05) and day 7 (P<0.001) (Figure 3b). Importantly, a significant genotype × time interaction (F_2,50=7.76, P=0.0012) was observed, reflecting a slowed progression of neurodegeneration in the eye when mGPx1 was co-expressed (Figure 3b). We found that HD flies (elavGAL4-driven) exhibit a ~57% reduction in locomotor activity compared to controls (P<0.001), and that strikingly, overexpression of mGpx1 completely restored activity levels (P<0.01)(Figure 3c). Furthermore, we found that expression of mGpx1 was able to ameliorate circadian arrhythmicity generated by expression of Htt93Qex1 in the neuropeptide pigment-dispersing factor (PDF) expressing clock neurons²⁸, resulting in an ~1.8-fold increase in the fraction of rhythmic flies (Supplementary Table 2). We reasoned that this arrythmicity was due to the neurodegeneration of the small lateral ventral neurons (s-LN_vs)²⁸, which are particularly important for self-sustained circadian behavioural rhythmicity in constant darkness²⁹. Indeed, we observed a reduced number of s-LN_vs (Figure 3d, Supplementary Figure 2), while the large lateral ventral neurons – the other PDF-expressing clock neurons – were not affected (data not shown)²⁸. Strikingly, mGpx1 expression in HD flies completely rescued neurodegeneration of s-LN_vs (Figure 3d, Supplementary Figure 2).

Figure 3.

mGpx1 and ebselen ameliorate phenotypes in HD flies. (a) Pseudopupil images from wild-type, HD flies, and HD flies overexpressing mGpx1 at day 10. In wild-type flies, seven rhabdomeres are visible per ommatidium. Scale bar = ~10 μM. (b) Quantification of average rhabdomeres per ommatidum in HD flies with and without pan-neural mGpx1 overexpression at day 1 [N = 14 (Htt93Q); N = 9 (Htt93Q Gpx1)], day 7 [N = 7 (Htt93Q); N = 9 (Htt93Q Gpx1)], and day 10 [N = 8 (Htt93Q); N = 9 (Htt93Q Gpx1)] after eclosion. Significant genotype (F_1,50=82.3, P=3.82x10⁻¹²) and Age (F_2,50=94.3, P=1.06 ×10⁻¹⁷), effects were observed, as well as a significant interaction between genotype and age (F_2,50=7.76, P=0.00116), indicating that mGpx1 reduces the rate of neurodegeneration in Htt93Q flies. Post-hoc Newman-Keuls comparisons showed significant rescue of neurodegeneration in mGpx1-expressing flies at day 1 (*P<0.05), day 7 (***P<0.001), and day 10 (***P<0.001). Browne-Forsythe analyses for equality of variance revealed a significant inequality of variance (P=0.02) due to a single group (Htt93Q mGpx1; day 1). The non-parametric Mann-Whitney test confirmed a difference at day 1 between the two groups (P=0.03), and between day 1 and 7 Htt93Q mGpx1flies (P<0.0001). (c) Total locomotor activity was assessed as the average number of beam crossings per 30 min bin per fly. Htt93Qex1 expressing flies (N = 29) exhibit reduced locomotor activity compared to controls [***P<0.001; N = 30 (WT – w118); N = 32 (mGpx1)], which is restored by mGpx1 (**P<0.01; N = 32). Comparisons performed by Kruskal-Wallis one-way analysis of variance. (d) Neurodegeneration of PDF clock neurons was assessed using confocal microscopy in day 7 adult flies. Flies expressing Htt93Qex1 driven by Pdf-GAL4 (N = 16 brain hemispheres) exhibited loss of s-LNV neurons compared to Canton S controls (N = 13), which was ameliorated by co-expression of mGpx1 (****P<0.0001; N = 16). Comparisons performed by Kruskal-Wallis one-way analysis of variance. (e) Quantification of average rhabdomeres per ommatidum in HD flies with ebselen treatment or vehicle (DMSO; N = 10) at day 7 after eclosion, and in newly emerged Day 0 adult flies (untreated control; N = 10). Significant treatment effects were observed, both with 100 μM (**P<0.01; N = 9) and 300 μM (***P<0.001; N = 9) ebselen doses. Data are shown as means ± standard errors of the mean (SEMs).

To test whether pharmacological GPx activity ameliorates neurodegeneration, newly emerged adult HD flies were transferred to vials of food containing ebselen or vehicle (DMSO). We found that ebselen was able to rescue up to ~60% of rhabdomere degeneration at day 7 (P<0.001)(Figure 3e). Thus, our in vivo analyses in Drosophila provide further support for our observations in yeast and mammalian cells.

An outstanding question is why GPx activity is robustly protective in these models, when in general, antioxidant strategies have had limited efficacy in HD models and patients, despite strong evidence that oxidative stress contributes to pathology¹⁸. As other classical antioxidant genes were not identified in our screen, we individually tested a panel of antioxidant genes in HD model yeast. We found that unlike GPx activity, the overexpression of superoxide dismutases, catalases, and glutathione reductases in yeast did not consistently confer protection against Htt103Q-mediated toxicity (Supplementary Table 3). We thus analyzed PC12 cells and found that overexpression of catalase or superoxide dismutase 1 (SOD1), while increasing enzyme activity, did not ameliorate mutant htt-induced caspase activation (Supplementary Figure 3a-c). Furthermore, pan-neuronal expression of ectopic catalase in the mitochondrial matrix of fruit flies, which has been shown to dramatically decrease mitochondrial hydrogen peroxide release³⁰, was not neuroprotective (Supplementary Figure 3d), and past work has found that overexpression of either cytosolic or mitochondrial forms of superoxidase dismutase does not reduce neurodegeneration in a fly model of HD³¹. Thus we conclude that GPx activity is a particularly robust antioxidant strategy in the HD models tested.

A possible explanation for these results comes from the observation that many antioxidants inhibit basal and induced levels of autophagy, a process known to be protective in several models of HD³¹. Indeed, treatment with several antioxidants, as well as overexpression of superoxide dismutase, can enhance neurodegeneration in a Drosophila model of HD. We thus tested whether ebselen modulated autophagy in PC12 cells via measurement of the autophagy marker LC3-II (Figure 4a). We found that ebselen did not alter induction of autophagy by rapamycin, whereas a control antioxidant (N-acetylcysteine) strongly inhibited autophagy (Figure 4b). These observations indirectly support the notion that GPx activity acts as a protective antioxidant strategy for HD by reducing ROS levels without adverse effects on autophagy.

Figure 4.

Ebselen does not inhibit basal or induced autophagy. PC12 cells were treated with either 10 μM ebselen or 10 mM NAc for 24 hours, with the addition of 400 nM bafilomycin A1 for the final 4 hours. Autophagy was induced in the indicated samples by the addition of 200 nM rapamycin for 24 hours. (A) Levels of autophagy were determined by measuring LC3-II levels by immunoblot analysis. (B) Densitometric analysis of bands was performed using ImageJ and the amounts of LC3-II are expressed relative to the untreated control. Statistical analyses performed via one-way ANOVA with post-hoc tests (****P<0.0001). Data are shown as means ± standard errors of the mean (SEMs). N = 7 per condition.

We have shown that increasing total cellular GPx activity by genetic or pharmacological means is protective in multiple HD model systems, likely by decreased oxidative damage. Indeed, we find that mutant htt-dependent increases in ROS can be blocked by both ebselen treatment and overexpression of mGpx1. This resonates with previous work showing that modulation of mGpx1 activity influences toxicity in cell-based and animal models of oxidative stress³². As ROS play an important role in apoptotic signalling, elevating GPx activity not only decreases apoptosis, but protects mice and primary neurons from apoptotic agents^33,34. Therefore, it is likely that some of the GPx protection observed in our study is occurring through a similar mechanism. Though not addressed in our study, increased levels of GPx activity also decrease neuroinflammation in response to ischemia/reperfusion³³. Recent work supports a role for neuoinflammation in HD pathology³⁵, thus it is possible that GPx treatment in HD patients may have protective effects via dampening of this inflammatory response.

An important feature of our study is the availability of several efficacious GPx mimetics which are well-tolerated by humans²⁴. Critically we have shown that treatment with the well-characterized GPx mimetic ebselen is strongly protective in both mammalian cell and fly models of HD. In addition to directly scavenging ROS³⁶, ebselen is able to decrease ROS levels by inhibiting the activity of several ROS-producing enzymes, including lipoxygeneases³⁷, nitric oxide synthase³⁸, and NADPH³⁹ oxidase, which contributes to ROS production in HD⁴⁰. These additional modes of ROS elimination may in part explain why we found ebselen to be more neuroprotective than mGpx1 (Figures 2 and 3). It is critical to underscore the finding that ebselen does not inhibit basal or induced autophagy in mammalian cells. As recent work has found that autophagy is regulated by ROS signalling⁴¹, our observations suggest that the peroxides targeted by GPx activity are not required for this regulation. In summary, our study shows for the first time that GPx mimetics may hold considerable promise for the treatment of HD and possibly other neurodegenerative disorders where oxidative stress plays a significant role^24,42,43.

Methods

Overexpression suppressor screen

The Yeast ORF Collection⁷ (Open Biosystems) was used to identify overexpression suppressors of Htt103Q toxicity. The collection contains plasmids for the overexpression of over 5500 individual S. cerevisiase ORFs in the MATa strain Y258 (MATa, pep4-3, his4-580, ura3-53, leu2-3,112) arrayed in a 96-well plate format. The collection was transformed with p425-Htt103Q-GFP using a high-throughput lithium acetate method based method as previously described⁴ and plated onto selective media containing glucose. The plasmids p425GALL-Htt25Q-GFP and p425GALL-Htt103Q-GFP were generated by amplifying the huntingtin constructs from pYES2-Htt25Q-GFP and pYES2-Htt103Q-GFP⁸ and cloning them between the SacI and XbaI sites of p425GALL⁴. Transformants were grown overnight in 96-well plates containing selective media supplemented with glucose, serially diluted 100-fold in distilled water, and spotted onto selective media containing glucose to assess cell numbers or 2 % galactose and 2 % raffinose to induce expression of both Htt103Q and the yeast ORF constructs. Growth was scored after 3 days growth at 30°C on glucose or 5 days on galactose and raffinose. At the 1/100 dilution no growth was observed in control yeast expressing Htt103Q while carrying an empty vector construct. Experimental strains were scored as suppressors if growth was similar to a 1/25 dilution of the control Htt103Q-expressing strain, representing at least a 4-fold improvement in growth. The transformation and growth tests were performed twice and strains able to grow on galactose and raffinose containing media both times were considered to contain ORF suppressors of Htt103Q mediated toxicity. Gene ontology searches were performed with the Functional Annotation Tool of the DAVID bioinformatics package⁹ (http://david.abcc.ncifcrf.gov/home.jsp). Network visualization was performed using the Osprey Network Visualization System (Version 1.2.0)¹².

Cell culture

The cell line Htt14A2.5²² used throughout this study was a kind gift from Leslie Thompson (University of California, Irvine) and was mycoplasma free. Cells were routinely cultured at 37 °C with 5 % CO₂ in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 100 U/l penicillin, 100 μg/l streptomycin, 2 mM L-glutamine, 5 % FBS, 10 % heat-inactivated horse serum and 200 μg/ml G418. Expression of mutant Htt was induced by the addition of 5 μM ponesterone (PA) for the indicated time. All assays were performed on actively growing sub-confluent cultures. Cells were transfected with an Amaxa® nucleofector® device using solution V and program U-029. Stable transformants were selected with 200 μg.ml⁻¹ hygromycin. The vector pRM1 for the overexpression of mGpx1 was constructed by amplifying the mGpx1 coding sequence and its SECIS element (-19 to +769 relative to the ATG) from murine cDNA and inserting it into the NheI site of pIRES-hyg3 (Clontech). The complete mCat and mSod1 ORFs were amplified and inserted into the BsrGI and NheI sites of pIRES-hyg3 (Clontech).

Antioxidant Enzyme Assays

Commercially available enzyme assay kits were used to determine the activity of glutathione peroxidase (Assay Designs), superoxide dismutase (Cayman) and catalase (Cayman). Cells were grown in either the presence or absence of PA for 72 hours and lysates prepared from 2 × 10⁶ cells according to the manufacturer’s instructions. Enzyme activity was determined using a Fluorostar Omega microplate reader (BMG Labtech) and the results normalised to total protein concentration.

Determination of caspase activation

Caspase-3/7 activation was assayed using Caspase-Glo® 3/7 reagent (Promega). Cells were seeded at a density of 5,000 cells/well in white 96-well plates, cultured in either the presence or absence of PA for 72 hours, and caspase 3/7 activity determined according to the manufacturer’s instructions. The luminescent output was detected using a Fluorostar Omega microplate reader (BMG Labtech). Results are expressed as the ratio of the luminescent output in PA-treated cells versus untreated cells.

Measurement of ROS production

ROS production was determined in cells using the ROS sensitive fluorescent dye dihydroethidium (DHE). Cells were grown in the presence or absence of PA for 6 hours, washed twice with HBSS, and stained with 2 μM DHE in HBSS for 15 minutes at 37 °C with 5 % CO₂. The cells were then washed twice with HBSS, trypsinized, harvested by centrifugation at 700 rpm for 5 minutes and suspended in 1 ml of HBSS. Fluorescence was then determined by FACS analysis using a FACSDiva.

Determination of LC3-II levels

LC3-II levels were determined using a modification of a previously described protocol³¹. In brief, cells were seeded in a 6-well plate at a density of 2 × 10⁵ cells per well and allowed to attach to the plate overnight. Samples were then treated with 200 nM rapamycin, 10 μM ebselen or 10 mM N-acetyl cysteine for 24 hours as indicated in the legend. Bafilomycin A1 (400 nM) was added to all samples 4 hours prior to harvest. Cells were washed with PBS and lysed in Cell Lysis-M buffer supplemented with protease inhibitors (Roche diagnostics) for 5 minutes at room temperature. Protein concentrations were determined using a Nanophotometer Pearl (Implen) and 50 μg of each sample separated on a 15 % polyacrylamide gel by SDS-PAGE. Samples were transferred to 0.2 μM PVDF membranes (Biorad) and probed with primary antibodies against LC3 (Novus Biologicals, NB100-2220, 1:5,000) and tubulin (Santa Cruz Biotechnology, sc-8035, 1:5,000). They were visualised using EZ-ECL detection reagent (Biological Industries) and densitometry was performed on X-ray films using Image J. Levels of LC3-II were normalised to tubulin and amounts are expressed relative to the untreated control.

Fly stocks

Flies were raised on maize media, in LD12:12 at 25 °C. The × chromosomal elav-GAL4 [c155] driver was obtained from the Bloomington Stock Center, Indiana. The Pdf-GAL4 driver⁴⁵ was a kind gift of Paul Taghert. The w;+;UAS-Htt exon1-Q93 flies have been previously described (line P463)²⁷, and were kindly provided by J. Lawrence Marsh and Leslie Thompson. Transgenic flies expressing ectopic catalase in the mitochondrial matrix, as well as appropriate pCasper4 empty vector controls, were a generous gift of Robin Mockett³⁰. A full-length cDNA of murine Gpx1 (mGpx1), including SECIS element in the 3′ UTR, was cloned into the pUAST vector by standard cloning methods, and transgenic UAS lines for mGpx1 expression were generated by BestGene Inc. and transgene insertion sites were mapped using hiTAIL PCR (data not shown). The w;UAS-mGpx1(2-5M);+ line was used for the analyses described in this study. A second mGpx1 line [w;UAS-mGpx1(2-4M);+] was also assayed via these metrics, and showed similar protective phenotypes (data not shown). Expression of mGpx1 was confirmed in both lines by QPCR (data not shown).

Pseudopupil analysis

The number of visible rhabdomeres per ommatidium was scored for >100 ommatidia per fly, and at least 7 flies were examined per genotype. For compound treatment, ebselen (Sigma) dissolved in DMSO (0.1 % final) was added to the maize media prior to setting at the required doses. Flies were added to the media upon eclosion, moved daily to fresh vials of food with the appropriate experimental treatment, and rhabdomeres scored after 7 days of treatment. Percent rhabdomere rescue was calculated for each genotype as follows: [(X-Y)/(7 – Y)] × 100, where × = average rhabdomeres for a “protective” genotype, Y = average rhabdomeres for control Htt93Qex1 flies. Scale bar determined by comparison of light microscopy images to phalloidin-stained confocal images of control flies (data not shown).

Immunohistochemistry and confocal imaging

Brains of 10-12 day 7 males were dissected in ice cold PBST (1 X PBS with 0.5 % Triton-X100) after being fixed overnight in 4 % formaldehyde-PBS. Further brain permeabilization was carried out by three 20 minutes washes with PBS containing 1 % Triton X-100 prior to blocking with 10% goat serum in PBST for at least 1 hour. Mouse anti-PDF antibody⁴⁶ (PDF C7, Developmental Studies Hybridoma Bank, 1:600) was diluted in fresh blocking solution and incubated overnight at 4 °C. After three 20 minute washes with PBST secondary fluorescent antibodies were added. Anti-mouse Cy5 antibody diluted in PBST (Abcam, ab6563, 1:500) was incubated with samples for 3 hours at room temperature. After three 20 minutes washes with PBST, brains were mounted in a solution of 3 % n-propylgallate and 80 % glycerol in PBS. Finally, brains were visualised on an Olympus FV1000 confocal microscope and processed with Olympus software.

Locomotor activity and circadian rhythm analyses

For locomotor activity analysis, flies with elav-driven Htt93Qex1 expression were monitored using Trikinetics DAM2 activity monitors. 32 flies per genotype were individually analysed in activity tubes for 2 days in light:dark 12:12 h cycles (LD12:12), followed by 7 days in constant darkness (DD) during which the circadian period was determined⁴⁷. Locomotor activity levels were assayed as the average number of infrared beam crossings per 30 minutes bin during the LD regime over the two days. Only flies surviving for at least 5 days in the activity tubes were analysed. For circadian rhythm analyses, flies with Pdf-driven Htt93Qex1 expression were monitored as above. The data recorded during DD was subjected to CLEAN spectral analysis⁴⁷ using BeFly!, an in-house behavioral analysis software package⁴⁸.

Statistical analyses

Data was analyzed by ANOVA with Newman-Keuls post-hoc tests using Statistica (StatSoft Ltd.) or Prism 5 (GraphPad Software). When assumptions of normality and equal variances were violated (using Browne-Forsythe or D’Agostino-Pearson/Shapiro-Wilk tests, respectively), non-parametric tests were employed (see corresponding figure legends). Sample numbers were determined empirically (¹⁵ and data not shown). Relative data values were log transformed prior to statistical analyses. χ² analysis was used to assess differences in the proportions of rhythmic/arrhythmic flies among groups.