Huntington's disease (HD) is a devastating autosomal dominant neurodegenerative disease caused by a CAG trinucleotide repeat expansion encoding an abnormally long polyglutamine tract in the huntingtin protein. Much has been learnt since the mutation was identified in 1993. We review the functions of wild-type huntingtin. Mutant huntingtin may cause toxicity via a range of different mechanisms. The primary consequence of the mutation is to confer a toxic gain of function on the mutant protein and this may be modified by certain normal activities that are impaired by the mutation. It is likely that the toxicity of mutant huntingtin is revealed after a series of cleavage events leading to the production of N-terminal huntingtin fragment(s) containing the expanded polyglutamine tract. Although aggregation of the mutant protein is a hallmark of the disease, the role of aggregation is complex and the arguments for protective roles of inclusions are discussed. Mutant huntingtin may mediate some of its toxicity in the nucleus by perturbing specific transcriptional pathways. HD may also inhibit mitochondrial function and proteasome activity. Importantly, not all of the effects of mutant huntingtin may be cell-autonomous, and it is possible that abnormalities in neighbouring neurons and glia may also have an impact on connected cells. It is likely that there is still much to learn about mutant huntingtin toxicity, and important insights have already come and may still come from chemical and genetic screens. Importantly, basic biological studies in HD have led to numerous potential therapeutic strategies.
Huntington's disease (HD) is a devastating autosomal dominant neurodegenerative disorder named after George Huntington, who provided a classic account of the condition in 1872 in The Medical and Surgical Reporter . However, the first definite description of HD by Charles Oscar Waters in 1841 provides a lucid picture of one of its main clinical features, chorea, and its hereditary nature : “It consists essentially in a spasmodic action of all the voluntary muscles of the system, of involuntary and more or less irregular motions of the extremities, face and trunk… The disease is markedly hereditary… The first indications of its appearance are spasmodic twitching of the extremities, generally of the fingers which gradually extend and involve all the involuntary muscles. This derangement of muscular action is by no means uniform; in some cases it exists to a greater, in others to a lesser, extent, but in all cases gradually induces a state of more or less perfect dementia. When speaking of the manifestly hereditary nature of the disease, I should perhaps have remarked that I have never known a case of it to occur in a patient, one or both of whose ancestors were not, within the third generation at farthest, the subject of this distressing malady…”
Although Waters stated that “the singular disease rarely, very rarely indeed, makes its appearance before adult life, and attacks after the age of 45 are also very rare”, this reflects the peak of the incidence distribution, since HD can present at any age.
The pathology of HD reveals striking neurodegeneration in the corpus striatum and shrinkage of the brain. These features were initially described by Meynert (1877)  and Jelgersma (1907) . The obvious loss of the caudate and putamen (corpus striatum) has led to the widely held belief that these neurons are most vulnerable to the mutation and also that loss of these specific neuronal populations can account for the motor, psychiatric and cognitive features of disease. More recent studies suggest that there is also widespread cortical loss/dysfunction in early HD . This raises the possibility that some of the features of HD may be driven by cortical dysfunction and the speculation that some of the striatal loss may be a secondary consequence of perturbations to cortico–striatal pathways.
GENETICS OF HD
The gene responsible for HD (HTT) was discovered in 1993 and encodes a 350 kDa ubiquitously expressed protein called huntingtin . The causative mutation is an abnormal expansion of a tract of uninterrupted CAG trinucleotide repeats within the coding sequence of the gene, 17 codons downstream of the initiator ATG codon in exon 1. CAG is a codon for glutamine, and the mutation leads to an abnormally expanded polyglutamine tract in huntingtin . There are now nine diseases that are known to be caused by expanded CAG-encoded polyglutamine tracts, including many of the dominant SCAs (spinocerebellar ataxias): SCA1, 2, 3, 6, 7 and 17.
In normal individuals, the number of CAG repeats is 35 or fewer, with 17–20 repeats found most commonly . Repeats between 27 and 35 are rare and are not associated with disease, but are meiotically unstable and can expand into the disease range of 36 and above, when transmitted through the paternal line. Most adult-onset cases have 40–50 CAGs, whereas expansions of 50 and more repeats generally cause the juvenile form of the disease. Incomplete penetrance has been observed in individuals with 36–41 repeats, but the estimates of penetrance for this group are imprecise [8,9].
There is a strong inverse relationship between the age of onset of HD and the number of CAG repeats. Longer repeats are correlated with an earlier age of onset . However, there is a wide variation in the age of onset with a given CAG repeat number, and the CAG repeat number itself has poor predictive power on the age of onset for any given individual. Only approx. 70% of the variance in the age of onset of HD can be accounted for by the number of CAG repeats. The residual variance is represented by other modifying genes and environmental factors [11–18].
Many trinucleotide-repeat disorders, including HD, are characterized by the phenomenon of anticipation, where the age of onset decreases and the disease severity increases in successive generations. This phenomenon can be explained by meiotic instability (which increases the number of CAG repeats) that appears to be greater in spermatogenesis than oogenesis; anticipation is mainly observed when the mutation is inherited through the paternal line [19–21].
In contrast with some of the inherited dominant ataxias where the clinical course is more severe in homozygotes , HD was previously believed to be one of the rare genetic diseases which demonstrated ‘complete dominance’, i.e. heterozygotes were as badly affected as homozygotes. However, more recent clinical and molecular studies have suggested that, although homozygosity for the HD mutation does not influence the age of onset of symptoms, homozygosity is associated with a more aggressive disease course [23,24].
GENETIC MECHANISM OF ACTION OF THE CAG MUTATION
Genetic data in humans and transgenic animal models suggest that polyglutamine mutations confer a deleterious gain-of-function on the target proteins [10,25,26]. HD is an autosomal dominant condition: one mutated gene is sufficient to cause the disease, in spite of the presence of a normal gene inherited from the other parent. In humans, loss of one of the two HTT genes occurs in Wolf–Hirschorn syndrome as a result of a terminal deletion of one chromosome 4, involving the loss of one HTT gene , and has also occurred with a balanced translocation with a breakpoint between exons 40 and 41 which physically disrupts the HTT gene . Hemizygous inactivation of huntingtin does not cause an abnormal HD-like phenotype. In addition, mice with only one functioning Htt gene do not show features of the disease [29–31].
Gain-of-function is also suggested by studies showing that the expanded CAG repeat is toxic itself. Expression of expanded polyglutamine peptides alone in Drosophila models has been shown to cause neurodegeneration . Ordway et al.  created a mouse model where a 146 CAG repeat sequence was inserted into the hypoxanthine phosphoribosyltransferase (Hprt) gene, which is not involved in any CAG-repeat disorders, and inactivation of the Hprt gene alone does not have any deleterious effects in mice. These mutant mice produced a polyglutamine-expanded form of the hypoxanthine phosphoribosyltransferase protein and developed a late-onset neurological phenotype that progressed to premature death . Transgenic overexpression of polyglutamine expansions, either in the context of the full-length huntingtin protein or only exon 1 of Htt, also produce neurodegeneration in mice and Drosophila . Although the primary consequence of the HD mutation is to confer gain-of-function, this does not preclude the possibility that disease severity may be modified by certain loss-of-function effects .
Wild-type huntingtin is found mostly in the cytoplasm, although a small proportion of the protein is intranuclear . The protein is known to be associated with the plasma membrane, endocytic (both clathrin-coated and non-coated) and autophagic vesicles, endosomal compartments, the ER (endoplasmic reticulum), the Golgi apparatus, mitochondria and microtubules [37–42].
Although the polyglutamine repeat in huntingtin has received attention for its pathogenic properties when expanded, it is possibly not essential for normal function [43–45]. Another feature of huntingtin protein structure is the presence of multiple HEAT (huntingtin, elongation factor 3, the PR65/A subunit of protein phosphatase 2A and the lipid kinase Tor) repeat sequences; 28–36 of these motifs are predicted to be distributed along the entire length of the huntingtin protein [46,47] (Figure 1). A HEAT repeat is a degenerate ∼50-amino-acid sequence comprising two anti-parallel α-helices forming a hairpin. HEAT motifs are usually involved in protein–protein interactions and are found in proteins that often play roles in intracellular transport (including nucleocytoplasmic shuttling), microtubule dynamics and chromosome segregation. These proteins are also characterized by high helical content (>50%) and frequently form superhelical structures with continuous hydrophobic cores [48–50]. Indeed, characterization of full-length huntingtin by biophysical methods suggests that the protein is an elongated superhelical solenoid with a diameter of ∼200 Å (1 Å=0.1 nm) .
Huntingtin and its normal cellular roles
It is unknown whether huntingtin contains NLSs (nuclear localization signals). However, a conserved NES (nuclear export signal) is found near its C-terminus . In addition, the N-terminal 17-amino-acid sequence of huntingtin has been suggested to act as a NES owing to its binding to the nuclear exporter Tpr (translocated promoter region). Expansion of the polyglutamine repeat interferes with this interaction causing accumulation of mutant huntingtin in the nucleus [51,52]. The 17-amino-acid N-terminal stretch of huntingtin has been recognized as playing an important role, together with a cluster of the first three HEAT repeats flanked by positively charged regions (amino acids residues 172–372), in targeting huntingtin to various intracellular membrane-bound organelles, including the plasma membrane, mitochondria, endosomal/autophagic vesicles, the Golgi apparatus and the ER [38,41,42]. Finally, several lysine residues within the same 17-amino-acid sequence immediately before the polyglutamine repeat, appear to compete for SUMOylation and ubiquitination, post-translational modifications that could regulate the half-life, localization and nuclear export of wild-type huntingtin, as well as modifying the toxicity of the mutant protein [53–55].
Cys214 of huntingtin is palmitoylated by Hip (huntingtin-interacting protein) 14, a palmitoyltransferase that regulates trafficking and function of huntingtin as well as several other neuronal proteins [56,57]. Palmitoylation could potentially play a role in the development of pathology as huntingtin with an expanded polyglutamine repeat is a much poorer Hip14 substrate compared with wild-type protein .
Both wild-type and mutant huntingtin are cleaved by various intracellular proteases, including caspases 1, 3, 6, 7 and 8, calpain and an unidentified aspartic protease [58–62]. Although the importance of huntingtin proteolysis for its physiological function remains to be elucidated, the role of mutant huntingtin cleavage in the progress of the disease is well established, as the full-length mutant protein is less toxic than its N-terminal fragments [63,64] (see below).
HUNTINGTIN, A PLAYER OF MANY GAMES
Despite substantial efforts directed towards understanding huntingtin function during last 14 years, its normal cellular roles remain poorly defined. This is predominantly due to the large size of the protein that makes isolation and analysis difficult, the lack of obvious homology with other proteins, ubiquitous localization within the cell and promiscuous interactions with more than 200 partners identified to date [43,44,65,66].
We will briefly describe several of the most well-studied cellular roles attributed to huntingtin (also see Figure 1). Although these diverse functions are currently considered to be relatively independent of each other, this view may change with an advancement of our knowledge about this interesting protein. Already, huntingtin is beginning to emerge as a scaffold protein orchestrating converging intracellular trafficking and signalling pathways.
Huntingtin is essential for normal embryonic development, as the loss of protein causes increased apoptosis and disrupted transport of maternal nutrients into the fetus, leading to lethality of mouse embryos around day 8.5 [29–31]. Knockdown of huntingtin expression in zebrafish embryos also produces a variety of developmental defects, including disruption of iron homoeostasis . In addition, the protein is required in adult neurons and testis for cellular viability . The high levels of cell death in HTT-knockout animals suggests that the protein may have an anti-apoptotic role . This idea is supported by the observations that overexpression of wild-type huntingtin protects against various apoptotic insults including those caused by starvation, mitochondrial toxins and overexpression of mutant huntingtin with an expanded polyglutamine repeat [69–71]. One possible molecular explanation for the anti-apoptotic capability of wild-type huntingtin is that it binds and sequesters the pro-apoptotic protein Hip1 that together with HIPPI (Hip1 protein interactor) can activate pro-caspase 8 . Also, huntingtin may also inhibit caspase 3 directly .
Huntingtin is involved in transcription regulation by interacting with an array of transcriptional factors and other proteins involved in the regulation of mRNA production [43,44,65,66]. Huntingtin has also been shown to interact with tryptophan (WW) domain-containing proteins implicated in non-receptor signalling and pre-mRNA splicing . On the basis of this, and by analogy with other HEAT domain-containing proteins (e.g. importins) that interact with transcriptional regulatory proteins and facilitate their transport between cytoplasm and nucleus, huntingtin has a proposed role in nucleocytoplasmic shuttling of transcriptional regulators and mRNA [47,51,75]. However, this function of wild-type huntingtin remains largely speculative, as most research has focused so far on the perturbations of transcriptional activity by mutant huntingtin. The most well-established example of huntingtin functioning as a transcriptional regulator is its role in the production of BDNF (brain-derived neurotrophic factor), which does not require the nuclear translocation of huntingtin and is performed in the cytosol. In this case, huntingtin binds and sequesters REST (repressor element-1 silencing transcription factor)/NRSF (neuron-restrictive silencer factor), a transcription factor that binds to NRSE (neuron-restrictive silencer element), an upstream DNA element found in ∼2000 genes including BDNF. Thus huntingtin acts as a positive transcriptional regulator of NRSE-regulated genes such as BDNF .
The role of huntingtin in vesicle trafficking was originally proposed on the basis of its localization to endocytic/endosomal vesicles in axons and synaptic terminals and from its interaction with a number of endocytic/trafficking proteins, including α-adaptin, Hip1, Hip14, Hap (huntingtin-associated protein) 1, Hap40, PACSIN1 (protein kinase C and casein kinase substrate in neurons-1) and SH3GL3 [SH3 (Src homology 3)-domain Grb2-like 3] (endophilin 3) (reviewed in [43,44,66,77]). Recently, this list has been extended to bona fide endocytic proteins, such as clathrin and dynamin . However, the role of huntingtin in endocytosis, despite being proposed by many, remains elusive, as the effect of huntingtin knockdown on the endocytosis of plasma membrane receptors has yet to be shown.
In contrast, the function of huntingtin as a facilitator of long- and short-range transport along microtubules is documented in mammalian cells, Drosophila and mouse models [39,78–81]. Huntingtin interacts directly, as well as via its binding partner Hap1, with the dynein/dynactin microtubule-based motor complex responsible for retrograde cellular trafficking. In addition, Hap1 binds another molecular motor, kinesin, and thus could play a role (independently or as part of a complex with huntingtin) in anterograde axonal transport. Huntingtin in complex with another partner Hap40 was shown to be important for movement of Rab5-positive endosomes along microtubules. As a result, knockdown of huntingtin inhibits movement of vesicles and mitochondria along neuronal projections [39,78–82]. Importantly, in addition to its role in the transcriptional regulation of BDNF, huntingtin is essential for efficient axonal transport of vesicles containing this pro-survival factor and thus controls neurotrophic support and endurance of neuronal cells .
As discussed above, huntingtin is highly expressed presynaptically, where it interacts with many proteins involved in synaptic vesicle exocytosis and recycling [43,44,66,77]. However, with the exception of the above examples, the role of huntingtin itself in these processes remains to be established. Huntingtin also seems to be important for normal synaptic transmission as part of the protein machinery localized to PSD (postsynaptic density), an electron-dense dendritic part of the synapse. Here, huntingtin interacts directly with the SH3 domain of a key regulator of postsynaptic activity, PSD-95, which in turn forms complexes with NMDA (N-methyl-D-aspartate) and kainate receptors belonging to the family of ionotropic glutamate receptors . Huntingtin was shown to negatively regulate the activity of glutamate receptors . As PSD-95 is also involved in relaying the signal from glutamate receptors to proteins such as SynGAP (synaptic GTPase-activating protein), huntingtin could potentially also modulate this signal transduction thus regulating synaptic plasticity [84,85]. Finally, huntingtin has been implicated in mGluR1 (metabotropic glutamate receptor 1) signalling via its interaction with optineurin .
MUTANT HUNTINGTIN AND HD PATHOGENESIS
Huntingtin cleavage: a probable rate-limiting step
There is now strong support for the idea that mutant huntingtin cleavage resulting in an N-terminal fragment containing the polyglutamine expansion is a key step in pathogenesis (Figure 2). N-terminal mutant huntingtin fragments are sufficient to produce HD-like abnormal clinical syndromes in model animals [87–89] and intranuclear inclusions [88,89]. Mutant huntingtin may be cleaved into a repertoire of different fragments by different proteases, including caspases, calpains and an as yet uncharacterized aspartic endopeptidase. Two cleavage sites at residues 513 and 552 are susceptible to caspase 3, producing N-terminal fragments of polyglutamine huntingtin of approx. 70 and 75 kDa respectively . The cleavage site at residue 552 is also susceptible to caspase 2 . A slightly larger peptide fragment, 80 kDa in size, is produced by caspase 6 cleavage at residue 586. The proteolysis and subsequent toxicity of the mutant protein can be modified (usually suppressed) as a result of phosphorylation of huntingtin by protein kinases, including Akt, Cdk5 (cyclin-dependent kinase 5) and ERK1 (extracellular-signal-regulated kinase 1) [91–93]. Importantly, recent data strongly suggest that inhibition of caspase 6 cleavage of mutant huntingtin rescues both the behavioural and neuropathological HD phenotype in mice expressing full-length mutant Htt transgenes . Thus the cleavage events may be rate-limiting steps in pathogenesis allowing conversion of comparatively non-toxic or benign full-length mutant huntingtin into toxic fragments.
Mutant huntingtin induces many different toxic pathways, some of which may interlink
In addition to caspases, calpains can also cleave huntingtin. The most N-terminal calpain cleavage site is at residue 536, which would lead to the formation of a 72 kDa N-terminal fragment of huntingtin as an intermediate product. This fragment may be cleaved further to generate a 47 kDa product, which is small enough in size to shuttle in and out of the nucleus . Huntingtin has also been reported to be a substrate for unidentified aspartic endopeptidases . This protease generates smaller N-terminal fragments of huntingtin compared with caspases and calpains and may be a crucial factor for the formation of intranuclear inclusions [61,94].
Huntingtin aggregation and its relationship with toxicity
The formation of neuronal intranuclear and intracytoplasmic inclusions of mutant huntingtin are pathological hallmarks of HD , and aggregates are a feature of all the known polyglutamine diseases. There has been considerable debate whether these represent toxic or protective species, or epiphenomena. In human brains, the density of inclusions in the cortex correlates with repeat length [95,96], consistent with in vitro data. However, there is little correlation between inclusion burden and the areas of the brain most affected in HD [97,98].
In mammalian cell culture systems, there is a strong correlation between aggregate formation and cellular toxicity, and cell death follows the formation of aggregates in many cases [99–101]. In HD mice expressing mutant Htt exon 1, intranuclear neuronal inclusions were detected before or near the onset of behavioural changes [87,102]. Overexpression of molecular chaperone(s) including hsps (heat-shock proteins) 70, 40, 104 and the chaperonin TRiC (tail-less complex polypeptide 1 ring complex) reduced both aggregation and cell death in HD cellular and/or mouse models [100,103–113]. However, these chaperones may be reducing the number of large inclusions by preventing oligomer formation, and it may be the oligomeric precursors that are the most toxic species.
Certain studies have reported a dissociation between aggregate formation and toxicity. When R6/1 HD exon 1 mice were crossed with tissue transglutaminase-knockout mice, this resulted in partial rescue of the brain and body weight loss and early mortality of the phenotype, but an increase in intranuclear inclusions . Overexpression of CA150, a transcription factor, rescued neuronal toxicity, while it increased neuritic aggregation without reducing nuclear inclusions . Furthermore, promotion of inclusion formation with a small molecule in a cell culture model of HD rescued huntingtin-mediated proteasome dysfunction . The most striking dissociation between aggregate formation and toxicity was the demonstration that cells that formed huntingtin inclusions had an improved survival compared with those that did not form inclusions . This suggests that cells with large inclusions are less compromised than cells with diffuse mutant huntingtin. However, the study did not compare the toxicity in cells with aggregates with wild-type cells. Also, cells with diffuse huntingtin are likely to contain oligomeric structures. Such oligomeric forms may be highly reactive because of their larger surface area to volume ratios, compared with large inclusions, and this may correlate with toxicity. However, in vivo, certain large inclusions may exert toxic effects if they block neuronal processes, as this may impair anterograde and retrograde transport.
PERTURBATION OF TRANSCRIPTION AS A POSSIBLE DISEASE MECHANISM
The idea that the nucleus may be an important site for huntingtin toxicity was suggested by studies proposing that the full-length wild-type protein was mainly localized to the cytosol, whereas the cleaved mutated molecule redistributed to the nuclear compartment . Furthermore, nuclear localization of mutant huntingtin appeared to enhance its toxicity both in cell culture and in mice [119–121]. A number of transcriptional regulators contain glutamine-rich activating domains, important for the interaction between transcription factors and transcription regulators. This led to the possibility that proteins carrying polyglutamine stretches could associate with transcription factors, leading to transcriptional alterations [122,123].
CBP [CREB (cAMP-response-element-binding protein)-binding protein] and Sp1 (specificity protein 1) have been identified as two major transcriptional regulators affected by polyglutamine proteins. CBP is an important transcription co-activator and is a major mediator of survival signals in neurons. It has histone acetyltransferase activity, which is important for allowing transcription factors access to DNA. The C-terminal glutamine-rich domain of CBP can mediate its interaction with mutated huntingtin. The interaction causes cellular toxicity and CBP relocalization from the nucleus into huntingtin aggregates [124–126]. Interestingly, further studies showed that CBP–huntingtin binding is primarily mediated by CBP's acetyltransferase domain and that the interaction depends on huntingtin's polyglutamine tract and proline-rich region. Moreover, huntingtin can bind to other proteins with acetyltransferase domains, modulating their activity .
The connection between histone acetylation and neurodegeneration led to further investigations testing the potential of HDAC (histone deacetylase) inhibitors for therapy. In a Drosophila polyglutamine model, treatment with butyrate or SAHA (suberoylanilide hydroxamic acid) rescued neurodegeneration . Similar results obtained in yeast and in cell culture provided evidence for the role of histone acetylation in neurodegeneration [128,129]. Further studies have reported beneficial effects of HDAC inhibitors in HD mouse models [130,131].
Sp1 is a sequence-specific transcription activator which binds to CG-rich regions of DNA. It contains a glutamine-rich activation domain, through which it binds to and regulates molecules of the transcriptional machinery, such as TF (transcription factor) IID, a multiprotein complex composed of TBP (TATA-box-binding protein) and multiple TAFIIs (TBP-associated factors). Moreover, Sp1 interacts with different molecules of the TFIID complex, particularly binding TAFII130 through the glutamine-rich domain, supporting the idea that the glutamine interface plays a fundamental role in recruiting components of the transcriptional machinery and subsequently RNA polymerase II . A specific interaction occurs between the N-terminus of mutant huntingtin and Sp1 , which interferes with Sp1-driven gene regulation. Indeed, mutant huntingtin interacts with Sp1, disrupting the specific Sp1–TAFII130 complex and altering the expression of certain Sp1 neuronal target genes, including the dopamine D2 receptor . Sp1–TAFII130 overexpression is able to counterbalance mutant huntingtin toxicity and suppression of dopamine D2 receptor regulation in cell culture. However, this situation may be more complex, as recent studies using cellular and transgenic HD models demonstrated that reduction of Sp1 could be neuroprotective . In particular, Qui et al.  showed an increase in Sp1 expression levels in different experimental models of HD, suggesting that suppression of Sp1 could be beneficial for HD pathology, while an increase in Sp1 levels may enhance mutant huntingtin toxicity.
Recent studies have shown that huntingtin also interacts with members of the core transcriptional machinery other than TFIID and TFIIF, affecting gene transcription in a polyglutamine-dependent manner .
Another pathway by which huntingtin affects transcription regulation involves the transcriptional regulation of BDNF, which is important for the survival of striatal neurons and for the activity of cortico–striatal synapses. Studies using in vitro and in vivo models showed that wild-type huntingtin, but not the mutated form, modulates BDNF expression in the cortex by regulating its transcription. [35,137]. The expression of BDNF is regulated by REST/NRSF, which recognizes and binds to the NRSE within the BDNF promoter [138–140]. Wild-type huntingtin is able to bind and sequester the cytosolic REST/NRSF, limiting its translocation to the nucleus and allowing BDNF transcription. Mutated huntingtin does not bind REST/NRSF effectively, leading to its accumulation in the nucleus. This leads to transcriptional repression of NRSE-sensitive genes, such as BDNF .
A new insight is emerging connecting impaired energy metabolism and transcription as contributors to HD pathology via PGC-1α (peroxisome-proliferator-activated receptor γ co-activator-1α). PGC-1α  is a transcriptional co-activator involved in different metabolic programmes that mainly acts as a fundamental regulator of mitochondrial biogenesis and respiration . Mice lacking PGC-1α show defects in brown adipose tissue as well as a pattern of neurodegeneration not unlike that seen in HD [143,144]. The possibility that PGC-1α could have a role in HD was suggested by observations of reduced levels of PGC-1α mRNA expression in human and mouse HD brains, and experiments showing that overexpression of PGC-1α reversed the effects of mutant huntingtin in cell models and in HD mice. PGC-1α expression is regulated directly by the CREB–TAF4 complex, which is impaired by mutant huntingtin, abolishing its ability to bind to the PGC-1α promoter . An alternative possibility is that mutated huntingtin binds directly to PGC-1α, affecting its ability to up-regulate expression of its downstream targets [145,146].
ELEVATED ROS (REACTIVE OXYGEN SPECIES) AND MITOCHONDRIAL DYSFUNCTION IN HD
ROS and metabolic mitochondrial dysfunction have been implicated in many neurodegenerative diseases [147–149]. Mitochondria are the major source of ROS production  and, at the same time, are also a key target for ROS damage. The respiratory chain (especially complex I and the Q cycle operating in complex III) generates superoxide, which is converted into hydrogen peroxide by MnSOD (manganese superoxide dismutase) [151,152]. Hydrogen peroxide can react with the available iron to produce the extremely reactive hydroxyl radical . Superoxide also reacts with nitric oxide to produce the dangerous peroxynitrite , which inhibits the respiratory chain  and inactivates aconitase and MnSOD [156,157]. Superoxide can also directly inactivate certain Fe–S proteins such as aconitase [158,159]. To protect the cell against ROS damage, mitochondria contain a variety of antioxidant systems. These include non-enzymatic components, such as α-tocopherol, coenzyme Q10 or glutathione, as well as enzymatic components such as MnSOD, catalase and glutathione peroxidase [160,161]. However, excessive production of ROS or a disruption of the antioxidant mechanisms can lead to oxidative damage to mitochondrial protein, lipid and DNA .
Evidence from post-mortem brains of HD patients and transgenic mouse models suggests that mitochondrial metabolic dysfunction could play a role in HD pathogenesis [163–165]. Mitochondrial impairment and oxidative stress have even been detected in asymptomatic HD carriers , indicating that this may be an early step in disease development. It is, however, not clear whether metabolic mitochondrial dysfunction is a primary cause in HD or a secondary consequence underlying neuronal loss . One possible mechanism of how mutant huntingtin could lead to mitochondrial impairment is through direct association with the outer mitochondrial membrane, which was shown in brain mitochondria from transgenic mice expressing a pathological CAG-repeat and isolated mitochondria from lymphoblasts of HD patients [168–171].
In addition to energy production and metabolism, mitochondria also play an important role in cellular calcium homoeostasis and apoptosis [172,173], and isolated mitochondria from HD mice also showed decreased membrane potential, depolarized at lower calcium loads compared with controls  and were more sensitive to calcium-induced cytochrome c release . These effects could be reproduced by incubating normal mitochondria with mutant huntingtin in vitro.
A relationship between the number of CAG repeats and mitochondrial ATP production has been reported . In huntingtin striatal cells, the ATP/ADP production decreases as repeat numbers increase, whether in the normal or the disease-causing range. The decreased ATP/ADP ratio was linked to enhanced calcium influx through NMDA receptors. Impaired energy metabolism probably leads to reduced ATP production, with a concomitant reduced mitochondrial membrane potential and a higher vulnerability to NMDA-mediated calcium influx and excitotoxicity [175,176]. Calcium influx could trigger further free radical production, exacerbating cell damage. There is also a potentiating effect of mutant huntingtin on NMDA receptor activity as NMDA-evoked currents and NMDA-mediated calcium transients were significantly increased in striatal neurons from YAC72 transgenic mice compared with wild-type controls, which could lead to an increased vulnerability to excitotoxicity [177,178]. Also, calcium influx through the NMDA receptor results in impaired mitochondrial function and increased oxidative stress [179,180].
Brain mitochondria have a higher concentration of lipids with polyunsaturated acyls, which are more sensitive to oxidative damage than other lipids . An increase in striatal lipid peroxidation was observed in HD transgenic mice which paralleled the worsening of the neurological phenotype . The overall effects of lipid peroxidation probably decrease membrane fluidity, making it easier for phospholipids to exchange between the two halves of the bilayer. This would increase the leakiness of the membrane to substances that do not normally cross it other than through specific channels, and also cause damage and inactivation of membrane proteins, receptors, enzymes and ion channels . Products of the lipid peroxidation process, such as 4-hydroxyhexenal and 4-hydroxynonenal, have also been shown to facilitate the induction of mitochondrial permeability transition , which could lead to cell death by release of apoptogenic factors.
Mutant huntingtin has been shown to directly impair the motility of mitochondria, with aggregates probably acting as ‘physical roadblocks’ for mitochondrial transport [81,185]. Aggregates may impair the passage of mitochondria along neuronal processes, causing them to accumulate adjacent to aggregates and become immobilized . This may heighten glutamate excitotoxicity and alter calcium handling owing to the inability to transverse the neurite.
ROS can cause direct damage to DNA, and an enhanced ROS production may lead to accumulation of somatic mutations . 8-OHdG (8-hydroxy-2′-deoxyguanosine) is a biomarker for oxidative DNA damage and increased levels of this ROS-damaged guanine nucleotide were found in mtDNA (mitochondrial DNA) from HD post-mortem parietal cortex  as well as in R6/2 HD transgenic mice . Also, increased oxidative damage to total DNA was found in caudate and frontal cortex of HD post-mortem brain . 8-OHdG can cause nucleotide base mispairing, resulting in DNA point mutations, probably leading to respiratory dysfunction, higher rates of ROS production and higher susceptibility to apoptotic stimuli [190–192].
In addition to showing meiotic instability, the HD mutation also shows somatic instability. Different CAG repeat lengths are seen in different neurons. Whether or not this has an impact on disease severity is not certain, but this phenomenon is certainly an attractive contributor to pathology. A recent study showed that the age-dependent CAG somatic mutation events associated with HD occur in the process of removing oxidized base lesions, and are largely mediated by the single base excision repair enzyme, OGG1 (7,8-dihydro-8-oxoguanine-DNA glycosylase) . OGG1 is activated in response to oxidative DNA lesions and results in somatic instability. This initiates a potential positive-feedback loop, since longer CAG stretches will lead to even more oxidative damage and hence more OGG1 activity .
ROS may also result in the formation of protein carbonyls; oxidatively modified proteins and enhanced protein carbonyl levels have been found in the striatum of R6/2 mice [194–196]. These modified proteins are generally dysfunctional owing to loss of catalytic or structural integrity, which may lead to decreased activities of key metabolic enzymes and disturbed cellular signalling systems [195,197,198].
It is interesting to consider that many of the pathogenic process proposed in HD pathogenesis may interact, and this potential cross-talk may lead to various types of positive-feedback loops (see Figure 2 for examples).
EVIDENCE FOR AND AGAINST THE IMPAIRMENT OF THE UPS (UBIQUITIN–PROTEASOME SYSTEM) IN HD
It has been proposed that the UPS is impaired in HD and that this contributes to the disease mechanism. However, this is controversial, and conflicting results have been obtained from different assays performed in a variety of different HD model systems. Some groups have demonstrated decreased proteasome activity [199,200], some have shown no change in activity [201,202] and others have even demonstrated an increase in proteasome activity [203,204] in response to mutant huntingtin expression.
The UPS consists of multiple components and is not only important for protein turnover, but also essential for normal cellular and physiological function [205,206]. At the centre of the UPS is the 20S catalytic core of the proteasome. This is a barrel-shaped multisubunit complex that has three main proteolytic activities: chymotrypsin, trypsin and peptidyl-glutamyl that cleave after hydrophobic, basic and acidic residues respectively . 19S regulatory particles (also termed PA700) bind either side of the 20S core proteasome to form the 26S proteasome . A cascade of enzymes act to covalently attach multiple ubiquitin molecules to target proteins, which mark them for degradation . Polyubiquitin chains are recognized by the 19S regulatory particle, which facilitates protein degradation by ATP-dependent de-ubiquitination and unfolding of the target protein, and opening the outer rings of the 20S core proteasome. The activity of the proteasome can be altered by its association with a number of regulatory molecules and complexes such as the PA28 family of proteasome activators, which enhance the degradation of short peptides  and Rad23, which is thought to shuttle ubiquitinated proteins to the proteasome . In addition, the catalytic activity of the proteasome can also be modulated by alterations in subunit composition in response to cellular stimuli [e.g. IFNγ (interferon γ) induction of immunoproteasome subunits LMP (low-molecular-mass polypeptide) 2 and LMP7 to bias proteolysis in favour of producing short peptides suitable for MHC-1 presentation at the cell surface] .
Although responsible for the degradation of short-lived and damaged proteins, the UPS indirectly regulates other cellular activities. The UPS also has a role in cell signalling through the degradation of many key regulatory proteins, protein subunits and transcription factors such as p53 and IκB (inhibitor of nuclear factor κB). Recently, it has been proposed that the proteasome has a role in normal synaptic function and plasticity, and is involved in the NMDA-dependent remodelling of the protein composition of synapses . Thus impairment of the UPS is likely to have a detrimental effect on the function of the cell and indeed the whole organism.
The idea that the UPS may be impaired in polyglutamine expansion disorders initially came from studies showing the labelling of polyglutamine aggregates with antibodies raised against ubiquitin and proteasome subunits in cell models [100,214], transgenic mice  and human post-mortem samples . From these observations, the sequestration hypothesis was proposed. It suggested that the sequestration of UPS components in aggregates and the altered subcellular localization of proteasomes might affect UPS activity. However, in contrast with the sequestration hypothesis, inhibition of the proteasome has been demonstrated in cells co-expressing a GFP (green fluorescent protein)–degron (GFP tagged to ubiquitin) construct and pathogenic Htt exon 1 constructs in the absence of visible aggregates , and there is some evidence to suggest that some molecules are not sequestered tightly into aggregates, but are only loosely associated and can diffuse freely . Another model to account for the impairment of the proteasome in HD came from both in vitro and cell model data suggesting that expanded polyglutamine-containing proteins are not easily degraded by the eukaryotic proteasome, which can only accommodate unfolded proteins [217,218]. As these studies show that the proteasome cannot cleave between successive glutamine residues in a polyglutamine tract, the choking hypothesis proposes that proteins containing expanded polyglutamine tracts may get ‘stuck’ in the proteasome and block the entry of other substrates into the barrel of the 20S catalytic core. Although it has been shown that synthetically generated polyglutamine aggregates do not inhibit 26S proteasome function in vitro , it has recently been shown that fibrillar species purified from HD transgenic mouse and human HD post-mortem brains do decrease proteasome activity in vitro .
The first study to measure proteasome activity in HD cell models directly used a fluorigenic substrate specific for the chymotrypsin activity of the proteasome . A shift in chymotryspin activity was demonstrated from cytosolic fractions to aggregate-containing precipitated fractions derived from lysates from both a stable HD cell model (expressing huntingtin exon 1 with a 150 polyglutamine repeat) and brain lysates derived from R6/1 mice . Chymotrypsin activity was reduced in the cytosolic fraction and increased in precipitated fractions derived from lysates of polyglutamine-expressing cells compared with control cells . This suggested the altered localization of proteasomes to aggregates. The authors also demonstrated reduced degradation of the endogenous proteasome substrate, p53 . This study strongly suggested the impairment of the UPS in HD. Subsequently, these data were supported further by a study in cells using a reporter molecule comprising EGFP (enhanced GFP) fused to a short sequence that targets the protein for proteasome degradation (termed degrons) . When this EGFP–degron reporter was co-expressed with mutant huntingtin in cells, EGFP fluorescence was increased more than 2-fold compared with cells expressing wild-type huntingtin. This observation implicates a major impairment of the proteasome function because >50% decrease of chymotrypsin-like activity is required to obtain a 50% increase of GFP fluorescence . Similar results were found with the ΔF508 mutant cystic fibrosis membrane conductance regulator, an unrelated protein sharing only the propensity to aggregate, suggesting that proteasome impairment is caused by aggregate formation . Consistent with these data, a reduction of chymotrypsin and peptidyl-glutamyl activities has been demonstrated in lysates from human HD post-mortem brains and HD patient skin fibroblasts .
Using the co-expression of NLS- or NES-tagged EGFP–degrons and NES or NLS mutant polyglutamine constructs, a global impairment of the UPS was demonstrated, regardless of the intracellular locations of the proteins containing the expanded polyglutamine tracts or the degron reporters . The authors also examined two hypotheses proposing mechanisms for the inhibition of proteasome activity. Contrary to the sequestration hypothesis being the only mechanism, they demonstrated proteasome inhibition in the absence of visible aggregates. They also showed that synthetic protein aggregates do not inhibit activity of the 26S proteasome function in vitro, suggesting that UPS impairment is unlikely to be caused solely by blocking the proteasome. Indeed, the decreases in nuclear proteasome function by extranuclear mutant polyglutamine and vice versa, argued that the observed effects were independent of interactions between mutant protein and the proteasome. Nevertheless, one cannot discount either of these models, as fibrillar forms of huntingtin purified from transgenic mouse and human post-mortem brains do inhibit the 26S proteasome in vitro . Furthermore, an accumulation of proteasome substrates may occur in the presence of normal proteasome function, owing to abnormalities in ubiquitination, de-ubiquitination or compromise of the activities of various shuttling proteins that may be required to traffic ubiquitinated proteins to the proteasome.
Data contrary to the above, suggesting that the proteasome is not impaired in polyglutamine expansion disorders, come from a variety of sources. SH-SY5Y cells stably expressing mutant huntingtin did not show a difference in the degradation of fluorigenic peptides, compared with cells expressing wild-type huntingtin . Also, an increase (not decrease, as expected) in the chymotrypsin and trypsin activities of the proteasome was observed in lysates derived from the cortex and striatum of the HD94 conditional mouse model of HD . This was attributed to an increase in the levels of the proteasome subunits LMP2 and LMP7 and the induction of the immunoproteasome. Increased proteasomal chymotrypsin-like activity has also been observed in brain lysates from the R6/2 model of HD compared with non-transgenic littermates . However, this study found no change in overall 26S proteasome activity and showed that the nuclear proteasome activator PA28 (also termed REGγ) is not involved in polyglutamine pathology . This is in contrast with data demonstrating the reversal of proteasome dysfunction in mutant-huntingtin-expressing striatal neurons and rescue of cell death by PA28 overexpression .
One of the caveats of studies in cell culture and many transgenic models is that they express artificially high levels of mutant proteins, which may induce proteasome dysfunction either directly or indirectly. The role of the proteasome in vivo has recently been tested using the knockin mouse model of SCA7 (a polyglutamine expansion disorder caused by mutations in ataxin 7) crossed with a transgenic mouse expressing an EGFP–degron reporter . The authors observed an increase in levels of the reporter in neurons at late stages of the disease. However, this was not due to inhibition of proteasome activity, but instead correlated with an increase in mRNA coding the EGFP–degron reporter .
Thus it still remains unclear whether the UPS is impaired in HD. Conflicting data are likely to occur as many of the experimental approaches used to assess UPS function have caveats. For instance, studies of UPS function have been performed in many different models of HD (stable, inducible and transient cell models, transgenic Drosophila, transgenic mice and human post-mortem samples). These models may represent different stages of the human disease and express HTT transgenes of different sizes (e.g. full-length huntingtin or smaller exon 1 fragments) at different levels under the control of different promoters. In addition, the various reporters used are likely to be assessing the activity of different components of the UPS. One problem with assays of isolated proteasome activity using small fluorigenic peptides is that modest changes in proteasome number/activity may not be rate-limiting for substrate clearance. It is likely that ubiquitin conjugation, and, in some situations, transport of ubiquitinated proteins to the proteasome, may be more important physiological regulators, and these will not be measured using these substrates. This has been partially circumvented by using EGFP–degron reporters and measuring levels of endogenous UPS substrates such as p53. However, levels of artificial EGFP–degron reporters may be affected by changes in mRNA encoding the reporter . Likewise, protein levels of endogenous proteasome substrates such as p53 are likely to be affected not only by UPS activity, but also by changes in transcription elicited by mutant huntingtin . In order to try to overcome these problems, UPS function has recently been assessed using polyubiquitin chains as an endogenous biomarker . The amount of polyubiquitin chains within a cell was shown be a faithful indicator of UPS function and, using this approach, impairment of the UPS was demonstrated in brain lysates from R6/2 HD transgenic mice, the HdhQ150/Q150 knockin model of HD and human HD post-mortem brains . One question raised by this study is whether the increased numbers of ubiquitin chains are necessarily due to proteasome dysfunction, or an increase in the ubiquitination rate of substrates. In other words, although the amounts of polyubiquitin chains will increase when proteasome function is disrupted, they can also increase in the context of normal proteasome activity or under conditions where substrate degradation is enhanced (for instance, if induction of ubiquitination exceeds substrate clearance). Thus, although this study is consistent with data suggesting impaired proteasome function in HD, it is still not definitive, and the conflicting results of previous studies must be properly resolved before it is proposed that the UPS is truly impaired in HD.
THE UPS AND DEGRADATION OF MUTANT HUNTINGTIN
One strategy for the treatment of polyglutamine expansion disorders is to decrease levels of the toxic mutant protein. This could be achieved by increasing the clearance of the mutant protein. Indeed, induction of autophagy by treatment with the mTOR (mammalian target of rapamycin) inhibitor rapamycin has been demonstrated to reduce aggregation and attenuate toxicity in HD cell and mouse models .
It is unclear whether proteins with an expanded polyglutamine tract are good proteasome substrates. Huntingtin interacts with the human ubiquitin-conjugating enzyme E2-25K, which requires the polyglutamine domain . Parkin, an E3-ubiquitin ligase, also co-localizes with mutant huntingtin aggregates in HD mouse and human brains, and overexpression of parkin enhances the clearance of the mutant protein . These data suggest that huntingtin may be a proteasome substrate. Consistent with this, proteasome inhibitors such as lactacystin and epoxomycin prevent mutant huntingtin clearance in a conditional HD mouse or cell models after its expression is stopped . Proteasome inhibition also increases mutant huntingtin aggregation and toxicity in HD cell models [100,200,225–227]. As the proteasome is unable to cleave between glutamine residues within polyglutamine tracts [217,218], up-regulation of proteasome activity would possibly reduce the levels of proteins with polyglutamine expansions and associated flanking sequences, producing increased levels of long isolated polyglutamine tracts. Such products are predicted to be more toxic than the inputs that have flanking sequences. However, such products have been shown to be degraded by puromycin-sensitive aminopeptidase, albeit very slowly and inefficiently . It is unclear whether the substrate capacity of puromycinsensitive aminopeptidase could be overwhelmed if proteasome activity were increased. In addition, modulation of the proteasome may not be a good therapeutic strategy. The proteasome has a key regulatory role, and altering its rate of degradation may have many side effects. One may be able to use chemical chaperones such as trehalose or Congo Red to increase the degradation of polyglutamine-containing proteins, as these agents shift the equilibrium towards increasing the levels of soluble monomeric proteasome-accessible species and away from aggregates [229,230]. This may make the polyglutamine proteins more accessible to the proteasome.
NON-CELL-AUTONOMOUS PROCESSES IN HD
Much of the focus on pathogenic mechanisms in HD has focused on cell-autonomous mechanisms. Although less attention has been focused on the pathological role of huntingtin in cell–cell interactions, a number of studies suggest that non-cell-autonomous mechanisms may also contribute to disease pathogenesis.
MSNs (medium spiny neurons), which are particularly vulnerable to the HD mutation, are innervated by glutamatergic axons, and overstimulation of glutamate receptors can induce cell death via excitotoxicity . HD transgenic mouse models show increased NMDA receptor activity in neurons [178,232]. The abundant glutamatergic afferents to MSNs and the NMDA receptor subunit composition in MSNs [233,234] may contribute to their preferential vulnerability in HD, especially when the glutamatergic input is increased or the clearance of extracellular glutamate is decreased. Clearance of extracellular excitatory neurotransmitters is largely performed by glutamate transporters [GLT-1 (glutamate transporter-1) and GLAST (glutamate aspartate transporter)] in astrocytes, the major subtype of glia . Huntingtin can reduce the expression level of GLT-1 in the brains of HD transgenic mice and Drosophila [236,237]. A decreased activity of GLT-1-dependent glutamate uptake in astrocytes leads to an increase of glutamate concentration extracellularly. This may then lead to increased and deleterious calcium entry in striatal neurons. All of these events may contribute to the activation of ATP- and calcium-dependent deleterious enzymes such as calpains, caspases and endonucleases .
Elegant glial–neuron co-culture experiments showed that N-terminal huntingtin in glia promoted the death of cultured neurons that did not express huntingtin . Huntingtin may affect various functions of glial cells, including their production of chemokines and neurotrophic factors. The neurodegeneration caused by overexpressed N-terminal huntingtin in neurons in vitro is reduced in the presence of glial cells. The conclusion might be that glial dysfunction contributes more to pathology than glial degeneration itself.
Another type of non-cell autonomous mechanism in HD comes from studies in mouse models expressing toxic mutant huntingtin fragments either in all neurons in the brain or only in cortical pyramidal neurons, which are vulnerable in HD . Restriction of huntingtin expression to cortical pyramidal neurons was sufficient to produce nuclear accumulation, but insufficient to produce neuropathology or motor deficits. However, expression in all of the neurons showed both progressive motor deficits and HD cortical pathology via nuclear accumulation, aggregation, reactive gliosis, dysmorphic neurites and dark neuron degeneration. One attractive possibility is that mutant huntingtin in cortical interneurons attenuates the ability of these neurons to mediate inhibition on their target pyramidal neurons and that this loss of inhibition contributes to pathology. However, these data do not preclude a combination of cell-autonomous and non-cell-autonomous mechanisms working in concert.
GENETIC SCREENS REVEAL MANY NOVEL PATHWAYS IN HD PATHOGENESIS
HD pathology may be a result of the cumulative effect of a variety of pathway perturbations. Many candidate-based approaches for HD treatment have identified target pathways, but searches for novel modifiers of HD pathology may allow us to identify further targets for therapeutic intervention as well as gain a better understanding of the pathology. One way of identifying such targets is through genetic or chemical screens.
Yeast two-hybrid screens have been used extensively for identifying huntingtin interactors [55,241–247]. Hap1, Hip2, Hip1 and CBS (cystathionine β-synthase), as well as many other proteins, have been identified as mutant huntingtin protein interactors. A study based on a vast network of 186 protein–protein interactions , identified GIT1 [GPCR (G-protein-coupled receptor) kinase-interacting protein 1], as a novel interactor. Interestingly, GIT1 associates with wild-type huntingtin in mammalian cells, but, in pathogenic circumstances, GIT1 localizes to mutant huntingtin aggregates and is required for their formation. In HD brains, GIT1 is cleaved, resulting in altered function.
Recently, Kaltenbach et al.  performed a yeast two-hybrid screen, as well as affinity pull-downs with MS, to identify huntingtin interactors. Out of 234 novel protein targets identified, an arbitrary set of 60 genes encoding interacting proteins were tested for their ability to behave as genetic modifiers of neurodegeneration in a Drosophila model of HD. This high-content validation assay showed that 27 of 60 orthologues tested were high-confidence genetic modifiers involved in a variety of pathways such as synaptic transmission, signal transduction, transcription and cytoskeletal organization. This study provides powerful evidence that huntingtin interactors are a particularly enriched source of HD modifiers.
Many screens have attempted to identify genetic and pharmacological modifiers of aggregate formation and clearance. A cell-free filter retardation assay identified benzothiazole derivatives as inhibitors of the formation of HD-insoluble aggregates , whereas a cell-based screen identified a lead compound that was able to specifically clear mutant huntingtin protein, but not normal huntingtin . Yamamoto et al.  used a gene array in a cell line that established transcriptional changes induced by the mutant huntingtin protein. Genes that were up-regulated were targeted using siRNA (short interfering RNA) molecules to decrease expression. Of these up-regulated genes, 23 were required for clearance of the mutant huntingtin protein. Activation of IRS-2 (insulin receptor substrate 2), which is involved in signalling from insulin and IGF-1 (insulin-like growth factor 1), enhanced the clearance of aggregate-prone proteins.
A genome-wide RNAi (RNA interference) screen has been used to identify loss-of-function enhancers of aggregation in a Caenorhabditis elegans model of polyglutamine disease . The 186 genes that enhanced aggregation were involved in a variety of pathways, such as RNA metabolism, protein synthesis, protein folding, protein degradation and protein trafficking, reflecting a diversity of potential cellular pathways that may have an impact on polyglutamine disease pathogenesis.
A number of cell-based assays have aimed to identify modulators of polyglutamine toxicity [254,255]. A screen of 4850 haploid mutants in a yeast (Saccharomyces cerevisiae) model identified 52 enhancers and 28 suppressors of mutant Htt exon 1-Q53-induced toxicity [256,257]. The enhancers are involved in cellular processes such as protein folding, response to stress and the UPS . Suppressors of toxicity are involved in transcription, protein aggregation, vesicle transport, vacuolar degradation and the kynurenine pathway, which is involved in tryptophan degradation . These studies suggest that yeast models may provide important insights into the biology of HD.
Rescue of cellular toxicity has also been studied in the nematode worm. A C. elegans model expressing N-terminal huntingtin carrying a stretch of 150 glutamine residues in the glutamatergic ASH sensory neurons leads to their degeneration by day 8 . In a genetic screen, for the enhancement of neurodegeneration, loss of pqe-1 (polyglutamine enhancer 1) gene function enhances neurodegeneration and pqe-1 overexpression rescues cellular toxicity. This gene encodes a putative glutamine-rich RNA exonuclease, which may rescue toxicity by sequestrating polyglutamine-expanded proteins.
Drosophila models of neurodegenerative diseases have recently been established and genetic screens in Drosophila have provided a number of modifiers. These include hsp40-like J domains proteins (dHDJ1 and dTPR2) . Molecular chaperones have been implicated in the rescue of other polyglutamine-induced pathogeneses in Drosophila, as well as in other models [260–262]. A novel modifier has been identified, dMLF1, the Drosophila homologue of human MLF1 (myeloid leukaemia factor 1), which suppresses toxicity by potentially inhibiting mutant polyglutamine protein aggregation . A SCA1 Drosophila model expressing ataxin-1(82Q) under the gmr driver, leads to severe external eye abnormality and reduced retinal thickness . Genetic screens with this model identified genes involved in protein-folding/heat-shock response, cellular detoxification, nuclear transport, RNA processing and transcriptional cofactors as modifiers of polyglutamine pathogenesis. Recently, a screen of a Drosophila model of SCA3 identified the miRNA bantam a suppressor of toxicity , revealing a further possible modifier pathway.
THERAPY FOR HD
HD has a number of features which make it a comparatively tractable problem, compared with neurodegenerative diseases which do not have Mendelian inheritance. Its autosomal dominant nature and single type of mutation allows most people at risk to be potentially identified before symptoms develop, making pre-symptomatic treatment a feasible possibility. This is important because a significant amount of neuronal loss has already occurred by the time most neurodegenerative diseases present clinically, lowering the rate of loss is potentially easier than repairing damage after it has occurred. The development of therapeutic strategies for HD may have wider relevance, most obviously for the eight other polyglutamine diseases, but even possibly for other neurodegenerative conditions characterized by abnormalities of protein conformation.
There are no disease-modifying treatments for HD in routine clinical use, and current treatment is therefore symptomatic. Although many trials have concentrated on mechanisms and outcomes associated with movement disorder, patients report that their quality of life is more often decreased by psychiatric manifestations of their condition, including depression, irritability and apathy . Rates of depression may be as high as 40%, and suicide may occur in as many as 10% [267,268]. Obsessive compulsive symptoms are also common . Given their frequency and impact, it is surprising that the evidence base for treatment of psychiatric disturbance is limited to case studies. SSRIs [selective serotonin (5-hydroxytryptamine)-re-uptake inhibitor] or mirtazapine may be preferred for depression as they have a more favourable anticholinergic profile compared with some other antidepressants . Improvement in depression and obsessional thinking has also been reported with olanzapine and sertraline [271,272]. Risperidone and amisulpiride may have a role in the treatment of psychosis in HD [273,274], while there are reports of quetiapine helping with behavioural disturbance .
A major neurological symptom associated with HD is chorea. Tetrabenazine depletes dopamine from central neurons. The first relatively large randomized control trial of its use in HD  found that tetrabenazine did significantly improve the UHDRS (unified HD rating scale) and global improvement assessments, but was associated with an increased incidence of adverse effects, including one suicide. Atypical antipsychotics are often used in the clinic, although the evidence from a larger trial of clozapine showed disappointing efficacy with significant side effects . More encouraging results are reported for olanzapine, but from small open-label studies . Strikingly, there are limited data providing conclusive support for any cognitive-enhancing therapies in HD. A study in a small number of HD patients treated with rivastigmine suggested possible motor and cognitive benefit . These results were not supported by a further small trial of donepezil . Given the limited proven efficacy of symptomatic treatments, an important part of treatment is co-ordinating appropriate social, paramedical and palliative care for HD patients.
Although further work towards developing and validating symptomatic treatments is clearly justified, the steadily increasing knowledge base around potential pathways leading to neurodegeneration in HD provides the possibility to develop rational mechanism-based therapies that may slow the neurodegeneration and neurological dysfunction at the core of this disease. We have selected a few possible examples of such therapeutic strategies that have been initiated largely in cell and animal models, before conducting studies in humans (Figure 3).
Potential disease-modifying strategies for HD
Preventing mutant gene expression is an attractive strategy, as it aims to remove the primary culprit: the toxic mutant protein. Human individuals with only one working copy of the HTT gene suffer no obvious adverse consequences. It has been possible to use siRNA in mouse models of HD  and other polyglutamine diseases  to decrease mutant protein expression and aggregation and prolong survival. These trials used direct intraventricular injection of the siRNA, a technique that may be relatively less acceptable to human sufferers. There are safety issues that will need to be addressed for this therapeutic approach. First, the knockdown must be specific to the mutant form of the protein and the wild-type should ideally be unaffected (as it may have anti-apoptotic functions). Although it may be possible to target the mutant allele to some extent using single nucleotide polymorphisms, these are likely to show interindividual variation and therefore require a library of siRNAs . Other safety concerns centre on both off-target effects (decreasing expression of genes other than the HTT gene) and inactivation of tumour-suppressor genes. These current technical difficulties and safety concerns mean that, although these techniques are potentially exciting, their use in clinical trials may be somewhat more distant.
If it is not possible to silence production of the mutant protein, then enhancing its clearance may be an alternative (or adjunct). Mutant huntingtin is cleared by autophagy, a process involving the formation of double-membrane structures called autophagosomes around a portion of cytoplasm. These autophagosomes ultimately fuse with lysosomes, where their contents are degraded. The strategy of up-regulating autophagy to increase clearance of mutant protein has shown promise in cell, Drosophila and mouse models of HD where the increase in clearance shows a preference for the mutant form of the protein . Drugs which can be used to up-regulate autophagy (rapamycin, carbamazepine, sodium valproate) are already U.S. FDA (Food and Drug Administration)-approved and some have a long track record in treating human CNS (central nervous system) disease. This approach is also one where the principle has been shown to work in animal and cell models expressing other aggregate-prone proteins that lead to human disease, including tau and mutant α-synuclein .
Ideally, we would like to develop drugs that target putative pathological mechanisms. One process that may be rate-limiting is huntingtin cleavage by caspases and other proteolytic enzymes. Indeed, the importance of cleavage of huntingtin and the role of apoptosis in HD have been described above. Inhibitors of the enzymes which cleave mutant huntingtin to its toxic N-terminal fragment(s) (including caspases 2, 3 and 6 and calpain) may provide potential targets for therapeutic intervention. Given the role of caspases in induction of apoptosis their inhibition may be doubly attractive. Broad-spectrum caspase inhibitors increased survival in an exon 1 mouse model, but required intracerebral administration . However, there may be cancer risks with long-term caspase inhibition. Minocycline may prevent apoptosis by inhibiting the mitochondrial permeability transition and is a caspase inhibitor (although it probably has a range of properties). Despite mixed results in mouse models [286,287], its history of long-term use in humans as an antibiotic has encouraged human trials. A small trial in humans has suggested benefit  and good tolerability has been reported in larger trials, the final outcome of which is awaited .
The potential importance of mitochondrial dysfunction and its implication in cell death in HD has already been described. Compounds which enhance mitochondrial stability have been investigated. Improvements in survival, neuropathology and motor performance have been reported in mouse models treated with creatine , and early results from its use in human sufferers suggest good tolerability .
Other potential treatments that have been investigated include transglutaminase inhibitors. Transglutaminases belong to a family of closely related proteins that catalyse the cross-linking of a glutamine residue of a protein/peptide substrate to a lysine residue of a protein/peptide co-substrate with the formation of a GGEL [Nϵ-(γ-L-glutamyl)-L-lysine] cross-link. These bonds may be important in the formation of aggregates and the toxicity of mutant huntingtin. Cystamine is a transglutaminase inhibitor which improves survival, motor phenotype and neuropathology in mouse models , and preliminary dose-finding and tolerability trials in human sufferers have been completed .
Chaperones help proteins to adopt more stable conformations and prevent aggregation, and their production is increased in heat-shock responses. Chaperones may be induced by chemical initiators of heat-shock responses such as geldanamycin which is protective in cell models of HD . Small-molecule ‘chemical chaperones’ such as trehalose have a similar effect and have shown beneficial effects in mouse models of HD .
Mutant huntingtin binds a number of transcription factors, of which the best described is CBP . CBP acetylates histones and thereby exposes the DNA sequence to allow transcription. Given that the mutant huntingtin therefore results in decreased histone acetylation, HDAC inhibitors have been investigated as a therapeutic strategy. These drugs have shown efficacy in both Drosophila  and mouse models  and one (phenylbutyrate) has long-term safety data in humans in the treatment of ornithine transcarbamylase deficiency . However, current HDAC inhibitors have major side effects.
Excitotoxic cell death in HD is implicated by the reproduction of HD pathology by NMDA agonists and the increased sensitivity of NMDA receptors in the presence of mitochondrial dysfunction. As a result, there has been interest in NMDA receptor antagonists as a potential therapeutic strategy. Murine models show increased survival with the NMDA receptor antagonists riluzole  and remacemide (alone and an additive effect with coenzyme Q10) . Amantadine, riluzole, lamotragine and remacemide have all been studied in human randomized controlled trials. The results for both amantadine and riluzole have been mixed, although, in both cases, the results from the largest and best-designed trials have been disappointing [299,300]. A relatively large (n=55) double-blind randomized control trial of lamotragine for 30 months found no significant difference in primary or secondary response variables . In one of the largest trials in HD to date, 347 patients were randomized to either coenzyme Q10 or remacemide, both or neither for 30 months. Sadly, the encouraging results from the mouse study were not replicated, with no treatment arm showing advantage over placebo . Memantine (an NMDA antagonist licensed in the U.K. for Alzheimer's disease) has shown some promise in an open-label trial, but this remains to be confirmed using more rigorous methodology . It should be noted that many of the compounds described act via more than one mechanism. For example, creatine is a transglutaminase inhibitor as well as having mitochondrial effects, and coenzyme Q10 may have effects on mitochondrial stability and amelioration of excitotoxicity as well as being an antioxidant.
One way to repair neuronal loss in HD may be with transplantation. Studies in HD mouse models transplanting either fetal striatal grafts  or wild-type cortex  showed the potential for graft survival and some modest improvements in phenotype. Trials have been carried out in humans, mostly using tissue from human fetuses [305–309], although one used porcine material . Results from these trials have been encouraging in terms of graft survival and largely also in terms of safety (although, in one study, of the seven patients transplanted, three developed subdural haematomas ). Although graft survival and function has been promising using proxy measures of glucose uptake  and post-mortem examination , the symptomatic benefits have been less clear-cut (although the trials have been largely associated with stability or slowed decline of motor symptoms). Although this approach may have potential, there is still much to do in terms of determining the optimal source and storage method of graft tissue (including the possibility of stem cell sources), choice of recipient, location of graft, graft susceptibility to disease process, rating of outcome, use of immunosuppression and study design (including the possible use of sham surgery to provide control). The results of longer-term follow-up trials currently underway are awaited. Also, grafting striatum may have potential value in alleviating motor symptoms, but is unlikely to be able to have a significant impact on the cognitive features of this disease which are likely to be due to the extensive cortical damage.
Neuronal loss in HD results in decreased availability of neurotrophic factors to adjacent neurons and subsequently further neuronal loss. The relationship between growth factors and mutant huntingtin appears to be antagonistic, with each being down-regulated by the presence of the other [35,314]. Difficulties crossing the blood–brain barrier and side effects associated with parenteral administration have resulted in trials of centrally administered neurotrophic factors. Cells engineered to express neurotrophins can be encased in capsules which protect them from host immune defences, but allow the release of neurotrophins. This approach has been used successfully in primate toxin models  (which may bear only distant resemblance to human HD) and Phase I trials have been conducted in humans . Safety results from this trial were reassuring, and, although the clinical benefits were not significant, the results were encouraging and further data are awaited.
In the 14 years since the causative gene in HD was discovered, huge advances have been made in understanding the biology of the disease and in designing rational therapeutic strategies which work well in animal models. Despite this, there remains no disease-modifying treatment. The challenge for the next decade will be translating laboratory data into clinical treatments, but, given the progress of the past, it is a challenge which can be viewed with optimism. Furthermore, it remains a challenge to conduct clinical trials in human HD, given its insidious onset and slow multifaceted progression. A further challenge will be to learn how to conduct powerful and cost-efficient trials to explore the growing repertoire of strategies emerging from more basic research studies. These may be considerably simplified by the identification of suitable biomarkers for disease progression .
We thank Michael Jardine and James Tweedley for help with illustrations. Work in D. C. R.'s laboratory on HD is funded by the MRC (Medial Research Council), the Wellcome Trust (Senior Fellowship to D. C. R.), Action Medical Research (Research Training Fellowship to B. R. U.), Sackler scholarship (B. R. U.), EU (European Union) [EUROSCA (European Integrated Project on Spinocerebellar Ataxias) and TAMAHUD (Identification of Early Disease Markers, Novel Pharmacologically Tractable Targets and Small Molecule Phenotypic Modulators in Huntington's Disease)] and NIHR (National Institute of Health Research) Biomedical Research Centre (Addenbrooke's Hospital).
brain-derived neurotrophic factor
enhanced green fluorescent protein
GPCR (G-protein-coupled receptor) kinase-interacting protein 1
huntingtin, elongation factor 3, the PR65/A subunit of protein phosphatase 2A and the lipid kinase Tor
manganese superoxide dismutase
medium spiny neuron
nuclear export signal
nuclear localization signal
neuron-restrictive silencer element
neuron-restrictive silencer factor
peroxisome-proliferator-activated receptor γ co-activator-1α
repressor element-1 silencing transcription factor
reactive oxygen species
Src homology 3
short interfering RNA
specificity protein 1