Histone deacetylase inhibitors arrest polyglutamine-dependent neurodegeneration in Drosophila

previously found that a fragment of mutant Htt interacts directly with CBP, which contains an acetyltransferase domain and is a co-activator of numerous promoters 3–5 . Mutant Htt represses transcription from CBP/p300 co-activated promoters in cell culture 5,6 , and CBP and other transcriptional regulatory proteins are sequestered in cytoplasmic and nuclear aggregates in both transgenic mice and patient brains 5,6 . In addition, ectopic over-expression of CBP reduces polyglutamine-mediated death of cultured cells 6,7 . These observations prompted us to investigate whether Htt interacts with other acetyltransferase-containing enzymes, to

Proteins with expanded polyglutamine repeats cause Huntington's disease and other neurodegenerative diseases. Transcriptional dysregulation and loss of function of transcriptional coactivator proteins have been implicated in the pathogenesis of these diseases 1 . Huntington's disease is caused by expansion of a repeated sequence of the amino acid glutamine in the abnormal protein huntingtin (Htt). Here we show that the polyglutamine-containing domain of Htt, Htt exon 1 protein (Httex1p), directly binds the acetyltransferase domains of two distinct proteins: CREB-binding protein (CBP) and p300/CBP-associated factor (P/CAF). In cell-free assays, Httex1p also inhibits the acetyltransferase activity of at least three enzymes: p300, P/CAF and CBP. Expression of Httex1p in cultured cells reduces the level of the acetylated histones H3 and H4, and this reduction can be reversed by administering inhibitors of histone deacetylase (HDAC). In vivo, HDAC inhibitors arrest ongoing progressive neuronal degeneration induced by polyglutamine repeat expansion, and they reduce lethality in two Drosophila models of polyglutamine disease. These ®ndings raise the possibility that therapy with HDAC inhibitors may slow or prevent the progressive neurodegeneration seen in Huntington's disease and other polyglutamine-repeat diseases, even after the onset of symptoms.
Huntington's disease is a late-onset neurodegenerative disease characterized by a movement disorder, neuropsychiatric symptoms and cognitive de®cits caused by expansion of a glutamine repeat in the Htt protein. Currently, no cure or effective treatment for this agonizing, lethal disease exists. The aetiology of several other neurodegenerative diseases, including spinocerebellar ataxia I and Kennedy's disease, also involves expansion of a polyglutamine repeat 2 . The pathology of these diseases may involve transcriptional dysregulation 1 . We previously found that a fragment of mutant Htt interacts directly with CBP, which contains an acetyltransferase domain and is a co-activator of numerous promoters 3±5 . Mutant Htt represses transcription from CBP/p300 co-activated promoters in cell culture 5,6 , and CBP and other transcriptional regulatory proteins are sequestered in cytoplasmic and nuclear aggregates in both transgenic mice and patient brains 5,6 . In addition, ectopic overexpression of CBP reduces polyglutamine-mediated death of cultured cells 6,7 . These observations prompted us to investigate whether Htt interacts with other acetyltransferase-containing enzymes, to  a, Schematic representations of CBP and P/CAF. RID, nuclear hormone receptor interacting domain; CH1, CH2, CH3, cysteine±histidine-rich regions 1, 2 and 3; Br, bromodomain; KIX, CREB-binding domain. The amino-acid residues used as 35 S-labelled probes for GST pull-down assays with GST±Htt proteins are designated below each protein diagram. b, Results of GST pull-down assays using radioactive domains of CBP and P/CAF with GST±51QP Htt. The AT domains of both CBP and P/CAF interact with GST±51QP. c, Interaction of GST±Htt fusion proteins, containing wild-type and expanded polyglutamine stretches, with and without the proline-rich domain, with radioactive probes containing the AT domains of CBP and P/CAF. A representative autoradiogram of the middle panel is shown. Slight alteration in the pattern of CBP (1,459±1,877) in the GST±51QP lane as compared with the other lanes is due to co-migration of GST±51QP with labelled CBP(1,459±1,877).
investigate the biochemical basis of the interaction and, ®nally, to ask whether effects on the levels of acetylation activity are important in polyglutamine-induced neurodegeneration.
The exact nature of the pathogenic entity of Htt is not known, although evidence suggests that cleavage products of mutant Htt containing the expanded polyglutamine domain are the pathogenic agents 8,9 . Httex1p, a polypeptide encoded by the ®rst exon of Htt that contains the polyglutamines plus 65 additional amino acids, has been shown to cause pathology in mice similar to that of Huntington's disease 10 . That this truncated Htt with an expanded polyglutamine domain interacts with full-length CBP in vitro 5 and co-aggregates in cell culture 5,6,11 suggests two hypotheses for Htt action that are not mutually exclusive. First, it is possible that the available activity of cellular proteins containing natural polyglutamine domains could be reduced by trapping them in aggregates through polyglutamine±polyglutamine interactions. Alternatively, soluble interactions between Htt and cellular targets could directly affect enzymatic activities.
To distinguish the contributions of polyglutamine±polyglutamine interactions from those of other interactions, we investigated which functional domains of CBP are targeted for binding by Httex1p, using glutathione S-transferase (GST) pull-down assays. 35 S-labelled protein probes spanning the length of CBP (Fig. 1a) were incubated with glutathione-agarose beads coupled to Httex1p with 51 glutamine repeats and the proline-rich domain. Signi®cant binding was observed to probes containing the CH2 domain and the amino-terminal portion of the acetyltransferase domain (residues 1,069±1,459), to probes containing the carboxy-terminal portion of the acetyltransferase domain (residues 1,459±1,759), and to probes containing part of the acetyltransferase domain as well as the CH3 domain (residues 1,459±1,877) (Fig. 1b). A CBP probe containing part of the CH3 domain and the C-terminal polyglutamine stretch (residues 1,742±2,441) binds to GST±51QP, whereas deletion of the polyglutamine stretch (residues 1,742±2,441DQ) has a negligible effect on binding. A CBP fragment containing the C-terminal polyglutamine domain of CBP (residues 2,161±2,441) did not interact in vitro with GST±51QP. Our data indicate that the direct interaction between the 51QP Htt fragment and CBP is not mediated by glutamine±glutamine interactions. Rather, they indicate that 51QP interacts primarily with the region of CBP containing the acetyltransferase and CH3 domains.
Having found that the expanded repeat domain of Htt interacts with both CBP and P/CAF primarily through the acetyltransferase and neighbouring domains, we sought to determine the features of Httex1p that are important for binding. We tested whether the length of the polyglutamine tract or the presence of the proline-rich region affected this interaction. Binding of GST±Httex1p to the acetyltransferase/CH3 domain of CBP is enhanced when the polyglutamine repeat expands from the normal range (GST±25QP or GST±25Q) to a pathogenic range (GST±51QP or GST±51Q) (Fig. 1c). GST fusions that contain the Htt proline-rich domain (GST±20QP and GST±51QP) interact better with the acetyltransferase-domain probes than do their counterparts without the proline-rich domain (GST±20Q and GST±51Q). Thus, the prolinerich region enhances, but is not essential for, the in vitro interaction. A representative autoradiogram and quanti®cation by phosphorimager analysis show binding of CBP probe 1,459±1,877 to the different GST±Htt constructs (Fig. 1c).
Because Httex1p interacts in vitro with the acetyltransferase domains of P/CAF and CBP, we measured the effect of Httex1p on the acetyltransferase activity of several widely expressed transcriptional co-activator proteins in vitro: CBP, p300 and P/CAF. Equimolar amounts of GST, GST±20QP, GST±20Q, GST±51QP and GST±51Q proteins were incubated with GST±CBP-AT (acetyltransferase) (residues 1,099±1,877), GST±p300-AT (residues 1,195±1,707), or GST±P/CAF-AT (residues 87±832), containing the acetyltransferase domains. Following a short incubation, the acetylation of an H4 peptide was monitored. The presence of the Htt GST fusion protein, GST±51Q, as well as GST±20QP and GST± 51QP, which contained the proline-rich region, signi®cantly reduced acetyltransferase activity of CBP, p300 and P/CAF, whereas GST and GST±20Q did not (Fig. 2). We conclude that the direct interaction of Htt with acetyltransferase domains (Fig. 1c) inhibits acetyltransferase function and that long polyglutamine stretches are more potent than short normal-range repeats. The presence of the proline-rich region increases the inhibitory potential of polyglutamine polypeptides. Surprisingly, GST±51Q inhibits acetyltransferase activity almost as well as GST±51QP, demonstrating that expanded polyglutamine stretches can interact with acetyltransferase domains and inhibit function without the proline-rich region.
Because Httex1p interacts directly with acetyltransferase domains and inhibits their function in vitro, we analysed acetylation levels of H3 and H4 in two PC12 cell lines stably transfected with inducible Httex1p constructs (our own unpublished data and E. Schweitzer, unpublished data). Expression of Httex1p with a normal-range repeat (25QP) reduces acetylation of both H3 and H4 compared with uninduced cells (Fig. 3a). Expression of an expanded repeat Httex1p (103QP) shows an even greater reduction of acetylation (Fig. 3a). To verify these results, we used a second transformed cell line, generated in a different strain of PC12 cells. Induction of an expanded polyglutamine repeat Httex1p (103QP) in the second PC12 line also leads to reduction of H4 acetylation in whole-cell extracts compared with uninduced cells (Fig. 3b). When these cells are treated with the HDAC inhibitor sodium butyrate, levels of H4 acetylation in both uninduced and induced cells are increased (Fig. 3b)  acid (SAHA) also increased acetylation of H3 and H4, reversing the decrease induced by Httex1p (data not shown). Although acetylation levels are altered, total histone levels in whole-cell extracts, determined by Coomassie blue staining of total cell lysates, are unchanged by induction of Htt proteins. Thus, expression of Httex1p causes a global reduction in acetylation of histones H3 and H4 that is reversed in the presence of HDAC inhibitors. We found that even normal repeat Htt can bind to and inhibit acetyltransferase activity (Figs 1 and 2), so why is pathology associated with only expanded repeat Htt? Mutant Htt can be proteolytically processed; the pathogenic fragment localizes to the nucleus, where it is capable of inhibiting acetyltransferase activity 8,12,13 . However, unexpanded repeat Htt does not normally localize to the nucleus 14 and therefore is not present in the cellular compartment appropriate to inhibit nuclear CBP activity. Indeed, unexpanded polyglutamine repeats can cause pathology. For instance, unexpanded human ataxin-1 protein, which contains 30 glutamines, is normally localized to the nucleus 15 . If it is expressed at suf®ciently high levels there, it can produce neurodegenerative phenotypes similar to expanded 82-glutamine ataxin-1 in either Drosophila or mice 15 . Thus, polyglutamine pathogenesis depends heavily on both level and location.
The reduced acetylation of histones observed in the presence of expanded repeat Httex1p in vitro (with or without prolines), and the subsequent reversal of this effect with HDAC inhibitors in cell culture, suggested that reduced acetyltransferase activity may be an important component of polyglutamine pathogenesis in vivo. Expanded polyglutamine peptides alone 16 as well as expanded repeat Httex1p (this report) are intrinsically cytotoxic and cause reduced viability and neuronal degeneration when expressed in Drosophila neurons. If polyglutamine pathology involves suppression of histone acetylation, then one would predict that inhibition of the deacetylation process by either of two completely independent mechanisms (for example, pharmacologically or genetically) would slow or reduce polyglutamine pathogenesis in vivo. To test this hypothesis, transgenic¯ies were engineered to express either Httex1p or just polyglutamine peptides 16 in neurons. We monitored the effects of feeding¯ies with the HDAC inhibitors butyrate and SAHA, and of genetically reducing the activity of their HDAC, on both lethality and degeneration of photoreceptor neurons.
Neurodegeneration is most readily observed in the¯y compound eye, which is composed of a regular trapezoidal arrangement of seven visible rhabdomeres (subcellular light-gathering structures) produced by the photoreceptor neurons of each Drosophila ommatidium (Fig. 4j). We found that expression of Httex1p with 93 glutamines (Q93) led to a progressive loss of rhabdomeres (Fig. 4a). Rather than the normal seven visible rhabdomeres, the number of rhabdomeres seen in¯ies expressing Httex1p Q93 progressively declined from an average of 6.35 at day 1 to 5.13 and 4.66 at days 6 and 12 after eclosion, respectively (that is, following emergence from the pupal case as an adult). Rearing larvae that expressed Httex1p Q93 on SAHA-or butyrate-containing food reduced the level of degeneration observed (Fig. 4b, c). Expression of the Httex1p Q93 transgene results in ,70% lethality (data not shown) and early adult death (Fig. 4d). In contrast, animals fed the HDAC inhibitor SAHA show increased viability (10 mM SAHA suppresses lethality to 45%; data not shown), and early adult death is markedly repressed in a concentration-dependent manner (Fig.  4d).
The effects of HDAC inhibitors on transgenic¯ies expressing extended polyglutamine peptides alone (Q48) were similar to those described above: Q48¯ies fed butyrate or SAHA had the same distribution of rhadomeres by day 6 as 1-day-old¯ies (data not shown), whereas their siblings that did not receive HDAC inhibitors showed a signi®cant degeneration of rhabdomere number over time (average of 5.47 at day 1 versus 3.92 at day 6; Fig. 4g, j). Even when fed HDAC inhibitors only after emerging from the pupal case as adults, progressive degeneration of photoreceptor neurons was still prevented (Fig. 4h, i). Therefore, even when administered to animals already exhibiting neurodegeneration, HDAC inhibitors markedly retard (or arrest) further neuronal degeneration. Thus, HDAC inhibitors rescue pathological effects of both polyglutamine peptides and Htt exon 1 polypeptides in vivo.
It is possible that HDAC inhibitors might affect cellular processes other than the deacetylase pathways. As an independent test of the signi®cance of acetylation levels in the pathogenic process, we manipulated acetylation levels genetically and examined pathology. The Drosophila Sin3A locus encodes a co-repressor protein that is a component of HDAC complexes 17 . We found that reducing the levels of HDAC by a partial loss of function mutant, Sin3A 08269 , in heterozygotes increased the viability of Httex1p Q93¯ies from 29% to 65% and led to a reduction in the extent and rate of neurodegeneration (Fig. 4e). Because this mutant allele represents a partial loss of Sin3A function, the effect on rescue of neurodegeneration may be less than that observed in the presence of HDAC inhibitors. Thus, both genetic and pharmacological reductions in the activity of HDAC reduce the rate and extent of polyglutamine-induced pathology.
To rule out the possibility that the rescue of degeneration and lethality by HDAC inhibitors was simply due to altered expression of the polyglutamine transgenes in the presence of HDAC inhibitors, we used western blotting to compare the level of transgene expression in larvae expressing exon 1 polypeptides in neurons in the presence and absence of butyrate or SAHA (Fig. 4f). Transgene expression was unaltered by the presence of HDAC inhibitors.
Although we have shown functional interactions of Htt with CBP, P/CAF and p300, these results do not exclude the possibility that other acetyltransferases may be targeted, but they do suggest that treatments that raise global levels of acetylation may be effective in ameliorating the effects of Huntington's disease and other neurodegenerative processes, even after the onset of symptoms. The above results raise the possibility that Htt peptides can lead to reduced levels of acetylation and transcription, both by binding to acetyltransferase domains and inhibiting soluble activity, and by sequestering polyglutamine-containing transcription factors (such as CBP and others) by trapping them into aggregates. When tested in vivo in Drosophila models of polyglutamine pathogenesis, inhibition of the deacetylation process by two independent mechan-ismsÐpharmacological (HDAC inhibitors) and genetic (reduction of Sin3A activity)Ðreduced degeneration of photoreceptor neurons and lethality. Other investigators have observed a reduction in acetylation induced by expanded polyglutamine-repeat proteins,     GST pull-down assays GST fusion proteins were puri®ed and experiments were performed as described 5 . Domains of CBP were ampli®ed by polymerase chain reaction (PCR) and the fragments cloned into pcDNA3.1. Binding and activity data were compared by an analysis of variance with StatView.

Acetyltransferase assays
The effect of Htt proteins on the acetyltransferase activity was assayed in vitro by a modi®ed technique 22 . GST fusion proteins were washed in 1´HAT buffer (10 mM butyrate at pH 7.5, 10% glycerol, 50 mM Tris HCl at pH 8.0, 0.5 mM dithiothreitol (DTT), 0.1 mM EDTA, 0.1 mM phenylmethyl sulphonyl¯uoride (PMSF)), then eluted with 15 mM glutathione in 1´HAT buffer containing 50 mM NaCl. Quanti®ed by Coomassie stain, 0.6 nmol of Htt GST fusion proteins or GSTalone were incubated for 10 min at room temperature with 10 pmol of GST±CBP (amino acids 1,099±1,877), 4 pmol of GST±p300 (amino acids 1,195±1,707) (purchased from Upstate Biotechnology), or 360 pmol of GST±P/CAF (amino acids 87±832), as estimated by Bradford analysis, in a total volume of 55 ml. We then added 10 nCi 14 C-acetyl coenzyme A (52 mCi mmol -1 , NEN) and 2 mg of biotinylated N-terminal H4 (amino acids 1±21) peptide (Upstate Biotechnology) in a volume of 4 ml, and the mixture was incubated for 45 min at 30 8C. We added 500 ml of 1´HAT buffer with 30 ml of a 50% slurry of streptavidin-agarose (Upstate Biotechnology) pre-equilibrated in 1´HAT buffer. This mixture was rotated at 4 8C for 20 min, then spun in a microfuge at 2,600 g for 2 min. The supernatant was removed, the pellet washed twice in RIPA buffer (50 mM Tris at pH 7.5, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA and 0.1 mM PMSF) and counted in a liquid scintillation counter. Assays were done in quadruplicate and the standard error of the mean calculated.
Histone acetylation analysis of PC12 cells PC12 cells, stably transfected with ecdysone-inducible (Invitrogen) F103QP-EGFP (ref. 5 and our own unpublished data) were plated, induced with 5 mM ponasterone (Invitrogen) for 48 h, and treated for the last 24 h with 5 mM sodium butyrate. Controls included uninduced and/or non-butyrate-treated cells. Cells were lysed in 1´HAT buffer with 50 mM NaCl, 0.3% NP40 and a protease-inhibitor cocktail on ice for 10 min. Fifty micrograms of whole-cell extract were analysed by western blot. Separate, independently derived, ecdysone-inducible PC12 cell lines were also analysed. These cells, denoted PC12/pBWN:Htt ex1Q103-EGFP cells and PC12/pWN:Httex1Q25-EGFP, were a gift of E. S. Schweitzer. These clonal cells contain a modi®ed exon1 of Htt 11 inserted into an expression vector containing the Bombyx ecdysone-regulated element 23 . Twenty-®ve micrograms of whole-cell extracts from PC12 cells stably transfected with plasmids encoding inducible 25QP-EGFP and 103QP-EGFP were analysed for acetyltransferase activity. Control cells (uninduced) and cells induced with 1 mM tebufenizide for 12 h were analysed by western blot. Anti-acetylated histone H4, anti-histone H3 and anti-acetylated histone H3 (all from Upstate Biotechnology) were used to determine levels of acetylated histones H3 and H4 relative to levels of total H3.

Drosophila stocks and crosses
Expression of polyglutamine-containing peptides is driven by the bipartite expression system upstream activator sequence (UAS)-GAL4 in transgenic¯ies 16 . Injection of plasmids expressing Httex1p with 20 or 93 glutamines produced nine and ten lines, respectively. Two lines with the most severe neuronal phenotype (P463 and P465) were chosen for the studies reported here. Polyglutamine stocks are w; P(w +mC = UAS-Q93httexon1) 4F1 and w; Pw mC UAS À Q48 myc=flag) 20 . Constructs under the control of a yeast UAS were crossed to¯ies expressing the yeast GAL4 transcriptional activator 21 driven by the neuron-speci®c promoter elav (chromosome-2 driver for Q48 lines 24 and X-chromosome driver for Htt exon 1 lines 25 ) that is expressed in all neurons from embryogenesis onwards: w; Pw mC ; w; elav À GAL4=CyO; Pw mC Act À GFPJMR1CyO actin-GFP or w; P(w +mW .hs = GawB)elavC155. The inhibitor concentration ranges tested were based on cell culture experiments (SAHA, a gift from Calbiochem) or published position effect variegation studies (butyrate) 26 . For testing the effect of Sin3A on polyglutamine phenotypes, w; P(w +mC = UAS-Q93httexon1) 4F1 virgins were crossed to w; P(w +mC w Pw; elav-GAL4)/Y; Sin3A 08269 /Bc Gla males.

Pseudopupil analysis and transgene expression
Pseudopupil analysis allows visualization of the arrangement of rhabdomeres in the ommatidia of the compound eye 27 . Adult¯ies were decapitated, one eye was dipped in a drop of nail polish and the head was mounted on a microscope slide. Eyes were analysed with a Nikon EFD-3/Optiphot-2 scope using a´50 objective, and photographed with a Spot camera. At least 200 ommatidia were scored for each condition. For western analysis of transgene expression, 20 larvae from each sample were ground in a buffer containing 0.1 M phosphate at pH7.1, 0.3 M sucrose, 0.02 mM phenylthiourea, protease cocktail and 0.1 mM PMSF, and 200 mg of total lysate were loaded per lane.