The pathogenic mechanisms of polyglutamine diseases and current therapeutic strategies

Proteolytic cleavage of polyglutamine proteins

Pathogenesis of several polyQ disorders including HD, SBMA, and SCA3, appears linked to proteolytic cleavage resulting in production of toxic polyQ-containing fragments. In other diseases, such as SCA1, the proteolysis of the mutant protein has been not confirmed. The conformational changes in the proteins with expanded polyQ lead to misfolding and proteolytic cleavage into smaller toxic fragments (Paulson 1999, 2000; Lunkes et al. 2002; Gardian et al. 2005). The use of several antibodies directed against N- and C-terminal parts of the protein showed that only a truncated version, including the polyQ expansion, aggregates in the vast majority of the polyQ diseases (Wellington and Hayden 2000). Caspase-mediated cleavage sites were identified or predicted in htt, atrophin-1, ataxin-3, and AR (Wellington et al. 1998). As a consequence of the cleavage, the resulting fragment could have an enhanced toxic effect (Ikeda et al. 1996; Ellerby et al. 1999) and/or could more easily enter the nucleus (Igarashi et al. 1998; Hackam et al. 1999) and/or more rapidly aggregate (Igarashi et al. 1998; Cooper et al. 1998). In several studies, truncated proteins with polyQ expansions appeared more prone than full-length proteins to form inclusions or cause cell death by apoptosis (Paulson et al. 1997; Ikeda et al. 1996; Martindale et al. 1998; Merry et al. 1998). Inclusions found in the brains of HD patients were stained by antibodies against sequences located close to polyQ stretch but not by those against more C-terminal parts of htt (Sieradzan et al. 1999; DiFiglia et al. 1997). The N-terminal fragments in the patients’ brains have been identified as generated by caspases (Wellington et al. 2002). On the other hand, the presence of full-length mutant htt in the sodium dodecyl sulfate and urea insoluble material from HD patients’ brains was reported (Dyer and McMurray 2001). A more recent study has shown that the inclusions isolated from HD brains contain broad ranges of N-terminal fragments of expanded htt with sizes of 50–150 kDa rather than intact mutant protein (Hoffner et al. 2005). These fragments may result from multiple proteolytic activities with little specificity. It is however possible, that the smaller products of specific cleavage seed the inclusions followed by recruitment of larger non-specific fragments. This is supported by the fact that the cleavage sites for caspases 2, 3, and 6 have been identified in human htt at amino acids 552, 513 (and 552), and 586, respectively, while the cleavage at amino acid 552 is an early event in the course of the disease (Wellington et al. 2002). The cleavage of mutant htt at the 586 amino acid by caspase 6 has been shown to be crucial in the HD pathogenesis as it was essential for the HD-related behavioral phenotype and selective neuropathology in yeast artificial chromosome HD mouse model expressing the whole human htt with 128Q (YAC128) (Graham et al. 2006). Modulating the mutant htt cleavage by caspase 3 at the amino acids 513 and 552, on the other hand, had no effect on disease progression and neurodegeneration in these mice. Caspase 6- but not caspase 3-resistant mutant htt protected neurons from excitotoxic stress suggesting that not all htt fragments contribute to excitotoxic neuronal death equally. Moreover, nuclear translocation of expanded htt was delayed in mice expressing mutant htt resistant to caspase 6 cleavage (Graham et al. 2006). This study strongly supports the hypothesis that generation of a specific htt fragment may represent an initial event in HD pathogenesis.

It has been reported that ataxin-3 is the target of caspase 1 at 241, 244, and 248 amino acids, and that the brains from SCA3 transgenic mice contained a C-terminal fragment (Goti et al. 2004; Berke et al. 2004; Colomer Gould et al. 2007). It has also been shown that normal ataxin-3 was extensively proteolysed in COS-7 cells with cleavage occurring at six other sites throughout the protein. Interestingly, mutant ataxin-3 underwent proteolysis to a much lesser extent with expanded polyQ masking the C-terminal cleavage sites (Pozzi et al. 2008) suggesting that SCA3 pathology can arise also through the accumulation of uncleaved protein. On the other hand, toxic fragment hypothesis has been introduced by Haacke et al. (2006), stating that expanded polyQ fragment must be released from the protective sequence context of full-length protein to be able to trigger the aggregation process. It is therefore possible that a toxic fragment initiates the aggregation process and later, as the intracellular levels of uncleaved expanded ataxin-3 increase, this process becomes independent of the presence of toxic fragments (Pozzi et al. 2008).

Mutant htt alters calcium signaling and increases Ca²⁺ levels in different cells (Bezprozvanny and Hayden 2004). Consequently, a calcium-dependent protease, calpain is activated in the brains of HD patients, where calpain co-localizes with htt aggregates and calpain cleavage products are present (Gafni and Ellerby 2002). Another cleavage of mutant htt is performed by an aspartyl protease between amino acids 104 and 114 (Lunkes et al. 2002). It is not known yet whether this cleavage is specific to mutant htt.

Proteolytic cleavage may be not an essential step in pathogenesis of all polyQ diseases. For polyQ protein normally localized in the cytoplasm, such as htt, ataxin-3, or atrophin-1, proteolysis appears to facilitate their translocation into nucleus enhancing their toxicities. For proteins already localized in the nucleus, for example, TATA-binding protein (TBP) or ataxin-7, proteolytic cleavage may not have so important role.

Phosphorylation of polyglutamine proteins

Phosphorylation may affect proteolytic cleavage, the initial conversion of mutant polyQ proteins to pathogenic conformations and nuclear transport (Warby et al. 2009). The impact of htt phosphorylation on HD pathogenesis has been demonstrated in several studies. Phosphorylation of htt at S421 by protein kinase Akt or serum and glucocorticoid-induced kinase SGK is neuroprotective against HD cellular toxicity (Humbert et al. 2002; Rangone et al. 2004). Recently, it has been reported that S421 phosphorylation of htt decreased the accumulation of both full-length htt and htt fragments (including those generated by caspase 6) in the nucleus (Warby et al. 2009). This observation is consistent with a previous study showing the effect phosphorylation status of S421 on vesicular transport in neurons. When phosphorylated, htt recruited kinesin-1 to the dynactin complex on vesicles and microtubules and promoted the anterograde transport of vesicles away from nucleus. Conversely, when htt was not phosphorylated, kinesin-1 was released and vesicles were more likely to undergo retrograde transport back to nucleus (Colin et al. 2008). Calcineurin (calcium/calmodulin-regulated serine/threonine protein phosphatase) dephosphorylates S421 in vitro and in cells. Using a rat model of HD with lentiviral-mediated expression of a polyQ htt fragment in the striatum, it has been shown that inhibition of calcineurin activity either by over-expression of a dominant-interfering form, by RNA interference, or by the specific inhibitor FK506, leads to an increased phosphorylation of S421 and prevents mutant htt-mediated death of striatal neurons (Pardo et al. 2006).

Phosphorylation of mutant htt at S434 by serine-threonine kinase Cdk5 reduced htt cleavage at 513 amino acid by caspase 3 and inhibited polyQ aggregation and cytotoxicity (Luo et al. 2005). The aggregation and cellular toxicity of mutant htt has also been shown to decrease upon phosphorylation at S536, which inhibits the cleavage of htt by calpain at this amino acid (Gafni et al. 2004; Schilling et al. 2006). Many phopshorylation sites were identified throughout whole htt sequence and several kinases have been proposed to be involved including MAPK/ERK kinase 1/2 or extracellular signal regulated kinase 1 (Schilling et al. 2006). Further studies are needed to understand how all the phosphorylation events affect the function and regulation of htt and their role in HD pathogenesis.

As mentioned in the section `Phosphorylation of polyglutamine proteins', htt phosphorylation by Akt has been shown neuroprotective. In contrary, phosphorylation of mutant ataxin-1 by Akt at S776 promotes its binding to 14-3-3, which in turn leads to ataxin-1 accumulation and neurodegeneration (Chen et al. 2003). Abnormal nuclear stabilization of mutant ataxin-1 by Akt was shown to have critical role in SCA1 pathogenesis in ataxin-1[82Q]-A776 transgenic mice (Emamian et al. 2003). In a recent study, however, inhibition of Akt either in vivo or in cerebellar extract did not decrease the phosphorylation of ataxin-1 at S776 arguing against involvement of Akt as the key kinase in SCA1 pathogenesis. The same study suggested that cAMP-dependent protein kinase is responsible for the S776 phosphorylation in the cerebellum (Jorgensen et al. 2009). Other phosphorylation sites that may play a role in ataxin-1 localization and interactions in SCA1 are S239 and probably T236. S239 is within the consensus sequences targeted by casein kinase 1 (CK1), Cdc2/Cdk5, and extracellular signal regulated kinase (Vierra-Green et al. 2005).

Ataxin-3 has been reported to undergo phosphorylation at S256 by glycogen synthase kinase 3β and this phosphorylation resulted in reduced mutant ataxin-3 aggregation in vitro (Fei et al. 2007). Another kinase, CK2 has been shown to be a regulator of nuclear localization of ataxin-3. CK2-dependent phosphorylation of ataxin-3 at S236 and S340/S352 decreased the appearance of nuclear inclusions and controlled the nuclear translocation of ataxin-3 providing a reasonable therapeutic approach for SCA3 (Mueller et al. 2009).

The polyQ expansion in AR activates major mitogen-activated protein kinase pathways, from which the p44/42 pathway in turn phosphorylates at S514. This phosphorylation has been shown to promote the mutant AR cytotoxicity (LaFevre-Bernt and Ellerby 2003). In atrophin-1, S734 has been identified as a phosphoacceptor in cell cultures and rat brain. Both normal and mutant protein forms are phosphorylated by c-Jun N-terminal kinase (JNK), but precise analyses revealed a reduced affinity of c-Jun N-terminal kinase (JNK) to expanded AR (Okamura-Oho et al. 2003). Biological significance of atrophin-1 phosphorylation needs to be evaluated.

With the development of specific inhibitors of kinases and proteases, modulating phosphorylation and proteolytic processing of mutant polyQ proteins could be a promising treatment strategy.

Sumoylation of polyglutamine proteins

Small ubiquitin-like modifier (SUMO) proteins bind covalently to specific lysine residues in target proteins regulating their cellular localization, protein–protein interactions, and transcription factors transactivation (Sarge and Park-Sarge 2009). Antagonistic relationship between sumoylation and ubiquitination has been proposed as they share common consensus sequence (Muller et al. 2001). Enhanced immunoreactivity of SUMO1 has been observed in affected neurons of HD, SCA1, SCA3, and DRPLA patients (Ueda et al. 2002).

Sumoylation of N-terminal fragment htt at K6, K9, and K15 was suggested to enhance the stability and reduce the aggregation of mutant htt, which however seemed to result in toxic intermediate polyQ oligomers accumulation and transcriptional repression (Steffan et al. 2004). Importantly, the same lysines are also ubiquitinated, however generally, sumoylation made pathology substantially worse while ubiquitination made it modestly better in Drosophila model of HD suggesting that (Steffan et al. 2004).

In ataxin-1, five major sumoylation sites have been identified (K16, K194, K610, K697, and K746). Sumoylation of ataxin-1 was dependent on its nuclear localization, phosphorylation at S776, self-association domain and polyQ length, all having role in the subcellular distribution of ataxin-1 in COS-1 cells. Nuclear localization, on the other hand, did not appear dependent on sumoylation, although it could have role in the efficiency of nucleocytoplasmic shuttling of ataxin-1. The lack of nucleus-to-cytoplasmic traffic of mutant ataxin-1 with disrupted interactions of ataxin-1 with other proteins leading to transcription dysregulation could be attributed to the decreased sumoylation (Riley et al. 2005).

Neuronal intranuclear inclusions in brains of DRPLA patients and mutant atrophin-1 aggregates in DRPLA cellular model were highly sumoylated. The results of this study showed active role of sumoylation in the pathogenesis of DRPLA accelerating aggregate formation and cell death (Terashima et al. 2002).

Androgen receptor is another target of sumoylation and this modification inhibits AR activity while significantly reducing mutant AR aggregation without affecting the levels of the receptor. Interestingly, SUMO inhibited AR aggregation through a unique mechanism that does not depend on critical features essential for its interaction with canonical SUMO-binding motifs (Mukherjee et al. 2009). Precise role of sumoylation in the polyQ-related neurodegenration is still not completely understood. Interestingly, while sumoylation appears neuroprotective in SBMA and SCA1, in case of HD and DRPLA it seems to accentuate the pathological process.

Fragment transgenic Huntington’s disease mouse models

As mentioned in the section `Mouse models', there is a problem with inverse correlation of genetic accuracy of the model and the phenotype or the extent of neuropathological lesions observed in the animals. A study comparing the knock-in Hdh^Q150/Q150 mice and N-terminal exon 1 fragment model, R6/2, revealed that both models had similar phenotypic and molecular characteristics, but they appeared much later in the knock-in mice (Woodman et al. 2007). Robust phenotype and pathology display model mice expressing the fragment of the gene containing the expanded polyQ. In case of HD, three mouse models with an N-terminal segment of mutant htt have been used in HD research, including R6/2, R6/1, and N171-82Q. The first and best characterized is the R6/2 line, which has an N-terminal end of exon 1 originally with ∼150Q (Mangiarini et al. 1996). This model revealed that expression of only exon 1 containing expanded polyQ is sufficient to produce mice with many neuropathological features of HD. It is one of the most commonly used genetic models of HD because of a progressive and homogenous phenotype. The short survival of about 3 months and well-quantifiable phenotype makes it a great tool for experimental therapeutic interventions. Moreover, the behavioral impairment and neuropathological findings suggested that the R6/2 model corresponds to human HD to a large extent (Stack et al. 2005). The effects on lifespan extension, body weight, and improvements in motor performance have been used as indicators of the treatment outcome in many studies. Other tests, such as limb clasping score, grip strength, or general activity became also very useful in evaluating and comparing different therapeutic approaches. It has been found that the variability in survival and the amelioration of the phenotype with increased CAG repeat size reduces the utility in therapeutic trials and may corrupt the results (Stack et al. 2005). In our experience, R6/2 mice with repeat size between 125 and 145 CAG were very appropriate for testing different treatments.

The R6/1 mouse model harbors exon 1 of the human HD gene with ∼116 CAG repeats (Mangiarini et al. 1996). This model has been not studied so extensively as R6/2. The disease starts later and has milder course than that in R6/2 with which it shares the hindlimb clasping, gait abnormalities and other motor deterioration. The later disease onset correlates with delayed appearance of htt aggregates and brain atrophy. Also, R6/1 mice usually live more than 1 year (Naver et al. 2003). In several treatment studies, N171-82Q mice expressing N-terminal fragment of htt gene containing exons 1 and 2 with 82 CAG was used (Schilling et al. 1999a). These mice have similar but less severe and more variable phenotype than R6/2. The lifespan varies between ∼4 and 6 months. Interestingly, the exposure of R6/1and R6/2 mice to enriched environment delayed the disease onset and ameliorated the clasping phenotype and the loss of peristriatal cerebral volume (Hockly et al. 2002; van Dellen et al. 2000; Glass et al. 2004), probably because of restoration of the neurogenic process (Lazic et al. 2006).

More recently, HD190Q and HD150Q mice were generated expressing exon 1 htt with 190 and 150 CAG under htt promotor, respectively. The transgene also contains enhanced green fluorescent protein, which makes this model a simple and sensitive tool for in vivo testing of therapeutic molecules for inhibiting aggregate formation that can be visualized by a fluorescent imager. Down-regulation of mRNAs for a number of hypothalamic peptides that are known to be involved in feeding behavior and energy homeostasis was observed in these mice. The median survival of HD190Q is ∼21 weeks and that of HD150Q mice 32 weeks (Kotliarova et al. 2005). This model however will need further phenotypic analysis.

Full-length transgenic Huntington’s disease mouse models

An HD mouse model with full-length human htt with 48 or 89Q manifested progressive behavioral and motor dysfunction with neuronal loss and gliosis (Reddy et al. 1998). A yeast artificial chromosome HD mouse model expresses the whole human htt with 128Q (YAC128) and replicates the neuropathology in HD patients (Hodgson et al. 1999). First symptoms, such as hindlimb clasping or gait abnormalities occur at about 3 months (Slow et al. 2003). Apoptosis is activated and mitochondrial dysfunction occurs in these mice as well (Fei et al. 2007). YAC128 appears to be a very useful model for therapeutic studies (Tang et al. 2009). A newer model, BAC transgenic mouse model of HD, was generated by introduction of BAC consisting of whole human htt locus of 170 kb with an expansion of 97 CAG (Gray et al. 2008). BAC transgenic mouse model of HD revealed that the progressive course of HD and the selective pathology of mouse brains might occur without early nuclear accumulation of aggregated htt. Synaptic pathology was observed prior to neuronal degeneration in these mice with first signs of synaptic dysfunction detectable at 3 months and a progressive synaptic pathology at 6 months (Spampanato et al. 2008).

Knock-in Huntington’s disease mouse models

Four knock-in HD mouse models have been generated and are one of the most precise genetic replicas of conditions in humans and were expected to mimic the pathogenesis of HD better than the transgenic models. The symptoms and neuropathological findings of these mice turned out to be quite subtle with no cerebral atrophy and only rare nuclear inclusions occurring later than in transgenic mice. No shortening of the lifespan was observed in knock-in models (Shelbourne et al. 1999; Wheeler et al. 2000; Lin et al. 2001; Levine et al. 1999), therefore survival is not suitable as an endpoint in therapeutic studies. On the other hand, these mice display measurable neuropathological features and behavioral symptoms that can be used for validation of experimental treatments.

The HdhQ111 line is a knock-in model with 111 CAG repeats inserted into the mouse HD gene (Wheeler et al. 2000). Nuclear htt immunoreactivity in ventral striatal neurons at about 6 weeks of age and later, 4.5 months, formation of N-terminal htt intranuclear inclusions in medium spiny neurons were observed in these mice. Late-onset motor deficits comprising mild gait disturbances were detected by footprint analysis at 24 months of age accompanied by reactive gliosis in affected areas of the brain. Lack of neuronal apoptosis in the striatum at this age indicates slow progression of the disease (Wheeler et al. 2002).

Similarly to HdhQ111, in HdhQ94 and HdhQ140 knock-in mice, exon 1 of the mouse htt is replaced by mouse/human chimeric exon 1, in these cases with 94 and 140 CAG repeats, respectively. Mice expressing 94Q protein show increased repetitive rearing at night at 2 months followed by decreased locomotion activity at 4 and 6 months of age. Intranuclear microaggregates were observed at 4 months together with reduced mRNA levels in striatum. The neuronal intranuclear inclusions are widely distributed throughout striatum at 6 months of age (Menalled et al. 2002).

HdhQ140 mice develop very similar phenotype as HdhQ94, however the disease starts earlier. The increased rearing occurs at 1 month, hypoactivity at 4 months and gait deficits at 12 months of age. Htt inclusions were observed from 1 month of age and there was a more widespread distribution than in HdhQ94 mice with intranuclear aggregates in several brain regions at age of 6 months (Menalled et al. 2003). Neuronal loss with reactive gliosis was found at 20–26 months of age with as much as 38% reduction in striatal volume. Alterations in dopaminergic signaling in HD mice could be at least partly attributed to the decrease in 32 kDa dopamine and cAMP-regulated phosphoprotein expression observed in the striata of HdhQ140 mice from 12 months of age (Heng et al. 2008).

Another knock-in mouse model is Hdh^(CAG)150 with insertion of 150 CAG expansion into exon 1 of the mouse htt homolog. Hdh^(CAG)150 mice have late-onset phenotype with neuronal intranuclear inclusions present predominantly in the striatum (Lin et al. 2001; Tallaksen-Greene et al. 2005; Heng et al. 2007). Significant behavioral phenotype does not appear within first year and only after 70–100 weeks of age in homozygous mice, weight loss, decreased activity, diminished rotarod performance, and clasping was observed. At age of 100 weeks, both homozygous and heterozygous Hdh^(CAG)150 mice exhibited resting tremor, unsteadiness, and staggering gait. Gliosis was significantly increased by 14 months. By 27 weeks, nuclear htt immunoreactivity and ubiquitin-positive inclusions were initially present within the matrix compartment and later, at 70–100 weeks of age, intranuclear inclusions were observed in most striatal neurons. Striatal dopamine D₁ and D₂ receptor binding sites were reduced at 100 weeks of age in both homozygous and heterozygous mice with a 50% neuronal loss and a 40% reduction in striatal volume. The D₁ and D₂ receptors were, however, already diminished at age of 70 weeks with no loss of striatal neurons suggesting that neuronal dysfunction precedes neurodegeneration (Heng et al. 2007).

Mouse models of other polyglutamine diseases

Transgenic mouse models bearing either truncated or full-length cDNA of respective genes for SCA1–3 and 7, DRPLA, and SBMA, YAC mouse models for SCA3 and SBMA and knock-in mouse lines for SCA1 and SCA7 have been generated (Yamada et al. 2008; Marsh et al. 2009). Models of all polyQ diseases showed accumulation of mutant proteins in neuronal nuclei except SCA2, where the expanded ataxin-2 was localized in the cytoplasm (Huynh et al. 2000).

Polyglutamine mouse models and preclinical therapeutic testing

Recently, concerns about applicability of different mouse models for preclinical therapeutic testing have been raised. Especially, the suitability of the models with truncated forms of mutant polyQ proteins have been discussed because they might represent very artificial systems not reflecting the real conditions observed in human diseases. Usually, these truncated transgenes are expressed from more than one copy of the cDNA per cell resulting in an over-expression and intensified toxicity. However, this should not be a prerequisite for avoiding the fragment models for initial experimental treatments. It has been shown that in the full-length HD mouse models, the cleaved form of mutant htt is the mediator of cytotoxicity (Graham et al. 2006). If a therapy is able to ameliorate the neuropathological and behavioral phenotype of an ‘exaggerated’ state, the more subtle damages observed in full-length or knock-in mouse models would be expected to be manageable with the same therapy and in an optimal situation with more pronounced beneficial outcome or with lower doses of the drug reducing the risk of unwanted side effects. Moreover, some of the fragment models, such as R6/2, are so well characterized and have been employed in many studies that they are the best choice for at least initial in vivo treatment testing and comparing the outcomes with preceding trials. Although this is not true in some therapeutic strategies, such as those targeting the proteolytic cleavage of the intact proteins. For other treatments, such as those targeting polyQ protein aggregation or clearance, the fragments models are very useful because of the robust expression of the transgenes and constant appearance of neuronal inclusions. For an eventual translation of the particular treatment into the clinic, however, the results from a fragment model should be validated in a second mouse model with a full-length transgene or in a knock-in model.

Gene silencing

Decreasing the levels of the mutant protein and thus preventing the downstream deteriorating effects appears to be one of the best strategies. The therapeutic potential of down-regulating abnormal gene expression has been demonstrated in a tetracycline-regulated mouse model of HD (Yamamoto et al. 2000) and a doxycycline-regulated SCA1 mouse model (Zu et al. 2004). The nuclear inclusions, which formed after induction of the mutant htt expression, disappeared when the expression was shut down. Also the behavioral phenotype was ameliorated, suggesting that therapeutic approaches aimed either at inhibition of mutant htt expression or its degradation might be effective. Several techniques targeting polyQ protein expression have been explored including siRNA or short-hairpin RNA (Bonini and La Spada 2005). This strategy however appeared to be not feasible in humans because of the non-specific ablation of the normal gene copy. On the other hand, this problem could be overcome as single nucleotide polymorphisms have been identified with some of them being suitable to siRNA targeting (Lombardi et al. 2009; Pfister et al. 2009). Recently, microRNAs suppressing the expression of ataxin-1 mRNA by specific binding to its 3′-untranslated region and decreasing the levels of the protein were identified and successfully tested in cells (Lee et al. 2008). For diseases with pathology at limited localizations such as retina in SCA7 this approach is suitable. On the other hand, targeting wide CNS pathology in other polyQ diseases would be more challenging.

Enhancement of protein degradation

Enhancing the degradation of mutant proteins is another therapeutic approach and the attempts to increase the autophagic clearance of mutant htt resulted in reduced htt toxicity in N171-82Q transgenic mice and in the Drosophila and the zebrafish HD models (Sarkar et al. 2007b; Williams et al. 2008). The activations of both mammalian target of rapamycin-dependent (e.g. by rapamycin analog CCI-779) or -independent (e.g. by lithium, calpain inhibitors, etc.) pathways modulation were beneficial and the combination treatment resulted in additive protection against polyQ-related neurodegeneration (Sarkar et al. 2008). Lithium also improved the neurological and pathological findings in SCA1(154/2Q) knock-in mice (Watase et al. 2007). There is not much knowledge about regulation of the enzymatic UPS activity. While a relatively large-scale of UPS inhibitors exists, no chemical compound actually activating the UPS activity has been available. Recently, amiloride, a well-known potassium-sparing diuretic drug widely used in clinics, and its derivative, benzamil, have been reported to reduce the polyQ aggregation and toxicity in HD models. Benzamil ameliorated brain pathology, motor deficits, and increased the lifespan of R6/2 mice (Wong et al. 2008). Another compound shown to increase enzymatic UPS activity is Y-27632, a rho-associated kinases inhibitor, which already has been used in clinical trials because of its anti-ischemic, antivasospastic, and antihypertensive effects (Lai and Frishman 2005). Interestingly, this compound also enhanced the macroautophagy activity, and this unique effect of modulating both main cellular degradation pathways led to reduced levels and reduced aggregation of mutant htt, ataxin-3, AR, and atrophin-1 in cell systems (Bauer et al. 2009). Although Y-27632 was previously shown to reduce the polyQ toxicity in Drosophila model of HD (Pollitt et al. 2003), this treatment needs to be tested further in mouse models.

Inhibition of aggregation

One of the first therapeutic approaches in polyQ diseases has been aimed at the prevention of aggregation. The therapeutic potential of small molecules able to prevent directly the formation of polyQ aggregates has been shown in several studies. For example treatment of R6/2 mice with Congo Red or trehalose significantly increased the mice survival by 16.4% and 11.3%, respectively (Sanchez et al. 2003; Tanaka et al. 2004). In a recent study, however, chronic administration of Congo Red did not improve the phenotype and lifespan significantly, but this could have been a consequence of limited ability of this compound to cross the blood–brain barrier (Wood et al. 2007). Trehalose has been reported to activate macroautophagy in an mammalian target of rapamycin-independent manner (Sarkar et al. 2007a) so the observed beneficial effect in mice might result from the stabilization of the expanded polyQ protein and its enhanced degradation. Conserving mutant htt in the native non-toxic conformation has been thought to prevent the cytotoxic effect of the mutant protein. PolyQ-binding peptide 1 was shown to prevent conversion of the expanded polyQ into aggregation-prone β-sheet-rich conformation and prevent neurodegeneration in a Drosophila model of HD (Nagai et al. 2003, 2007).

Cystamine may reduce expanded polyQ aggregation by inhibition of TG that is thought to crosslink expanded polyQ proteins and facilitate their aggregation. In HD brains, increased TG activity was observed (Karpuj et al. 1999). In one study, where the treatment of R6/2 mice started at age of 7 weeks, cystamine did not affect aggregate formation, while improving the mice phenotype (Karpuj et al. 2002). This could have been a result of caspase inhibition and antioxidant activity of cystamine (Lesort et al. 2003). In another study though, where the treatment started at 3 weeks, the htt aggregation decreased by 68% in striatum and by 47% in neocortex (Dedeoglu et al. 2002). The survival improvement was high in both studies with 12% and 19.5% increase, respectively.

Another strategy to reduce the misfolding, oligomerization, and aggregation of polyQ proteins is to increase the cellular levels of molecular chaperones such as Hsp70. The up-regulation not only interferes with the aggregation process but may also enhance the degradation of the mutant protein through UPS. Beside over-expression of HSPs, several promising compounds have been introduced which induce the expression of HSPs. Geranylgeranylacetone is an acyclic isoprenoid compound, used as an oral anti-ulcer drug, strongly induced HSPs expression in different tissues (Hirakawa et al. 1996). Oral administration of geranylgeranylacetone enhanced the expression of Hsp70, Hsp90, and Hsp105 through induction of heat-shock factor 1 in CNS, where it inhibited nuclear accumulation of mutant AR leading to improved motor behavior and increased survival of transgenic SBMA mice (Katsuno et al. 2005). An Hsp90 inhibitor, 17-(allylamino)-17-demethoxygeldanamycin (17-AAG) has also been demonstrated to ameliorate neurodegeneration in SBMA mice (Waza et al. 2005). Treatment with 17-AAG dissociated p23 (Hsp90 cochaperone) from the Hsp90–AR complex and facilitated the proteasomal degradation of the mutant AR. Recently, it has been reported that 17-AAG suppressed polyQ-induced neurodegeneration in Drosophila models of SCA3 and HD via activation of heat-shock factor 1 which in turn up-regulated molecular chaperones Hsp40, Hsp70, and Hsp90. 17-AAG reduced the lethality of SCA3 and HD Drosophila models by 74.1% and 46.3%, respectively (Friedman et al. 2008). An analog of 17-AAG, 17-(dimethylaminoethylamino)-17-demethoxygeldanamycin has recently been reported to induce expression of Hsp40 and Hsp70, degrade the pathogenic AR protein, and to ameliorate the phenotype in SBMA transgenic mice (Tokui et al. 2009).

Modification of transcription

Several compounds modulating the deregulated gene transcription have been tested in different models of polyQ diseases. Especially, targeting histone methylation and acetylation is thought to achieve normalization of transcription disrupted by expanded polyQ proteins. Histone deacetylase inhibitors such as suberoylanilide hydroxamic acid (SAHA) and sodium butyrate (SB) have been demonstrated to ameliorate the phenotype in polyQ disease models. SAHA has been tested in R6/2 mice in which it improved the motor phenotype however no amelioration in body weight was observed and higher doses of SAHA were toxic (Hockly et al. 2003). The effect of SB was investigated in R6/2 mice and besides the improvement in motor phenotype it also delayed the body weight loss and extended the lifespan by more than 20%. Moreover, SB increased not only the acetylation of histones H3 and H4 but also of SP1 resulting in improved transcriptional regulation in R6/2 brains (Ferrante et al. 2003). In DRPLA transgenic mice expressing full-length atrophin-1 with 118Q (Atro-118Q14), SB ameliorated the motor impairments and extended the lifespan (Ying et al. 2006). Treating AR-97Q SBMA mice expressing full-length AR with SB mitigated the neuromuscular phenotype and increased survival (Minamiyama et al. 2004). Another histone deacetylase inhibitor, phenylbutyrate, improved survival, and attenuated gross brain atrophy and ventricular enlargement in N171-82Q HD transgenic mice (Gardian et al. 2005).

The inhibition of histone H3 methylation with mithramycin resulted in a great extension of lifespan in R6/2 mice accompanied by improvement in motor performance and striatal pathology (Ferrante et al. 2004). The improvement of H3 and H4 modification profile after chromomycin and mithramycin treatment was observed also in N171-82Q mouse model (Stack et al. 2007). Another drug shown to mitigate the transcriptional dysbalance is the phosphodiesterase type IV inhibitor rolipram, which increases the phosphorylation and activity of CREB. Rolipram ameliorated the neuropathological findings, slowed the progression of neurological phenotype and increased the survival in R6/2 mice. Moreover, BDNF, whose expression is impaired in HD (Zuccato et al. 2003), was induced in treated mice through restored function of CREB (DeMarch et al. 2008). In another study, it was shown that rolipram also prevented the sequestration of CBP into nuclear aggregates (Giampa et al. 2009).

In the AR-97Q SBMA mouse model, a drug specifically disrupting the interaction of mutant AR with CBP, ASC-J9 (dimethylcurcumin), reduced the SBMA symptoms and reversed muscular atrophy in treated mice. This effect was accompanied by restoration of vascular endothelial growth factor 164 expression (Yang et al. 2007).

In a C. elegans HD transgenic model expressing N-terminal htt with 128Q, treatment with resveratrol resulted in amelioration of mutant polyQ toxicity and it also reduced cell death in neuronal cells derived from HdhQ111 knock-in mice. The underlying mechanism was shown to be an enhanced activation of silent mating type information regulation-2 protein through activation of daf-16, a member of the forkhead box family of Forkhead transcription factors (Parker et al. 2005). Resveratrol treatment appears to be very promising candidate for support therapy of polyQ diseases.

Mitochondria, energy metabolism, and antioxidants

Different compounds improving energy metabolism defects or reducing oxidative stress in polyQ diseases have been successfully tested in mouse models. Reduced concentrations of creatine and phosphocreatine were observed in basal ganglia of HD patients (Sanchez-Pernaute et al. 1999). Creatine administration stabilized the MPT, prevented ATP depletion, and increased the protein synthesis. In R6/2 mice, the treatment with 2% creatine ameliorated the brain pathology, improved the phenotype and increased the lifespan by 17.4% (Ferrante et al. 2000). Similarly, in N171-82Q HD mice, the survival extended by 19.3% with 2% dietary creatine treatment (Andreassen et al. 2001a). Other drugs preventing the depletion of ATP, or increasing the activities of mitochondrial complexes II–III seen in HD, such as dichloroacetate and triacetyluridine were successfully tested in R6/2 mice with moderate but significant effects (Andreassen et al. 2001b; Saydoff et al. 2006).

Antioxidants such as α-lipoic acid, coenzyme Q₁₀, clioquinol, tauroursodeoxycholic acid (also antiapoptotic effect), or BN82451 have been proven effective in R6/2 mouse lines (Giampa et al. 2009). Especially coenzyme Q₁₀, an antioxidant, a cofactor of the mitochondrial electron transport chain and inhibitor of MPT, extended the survival of R6/2 mice by up to 26.3% while greatly improving the motor perfomance and reducing weight loss and nuclear inclusions (Smith et al. 2006). Clioquinol administration reduced the striatal atrophy, decreased nuclear inclusion formation, ameliorated the HD phenotype development, and enhanced the lifespan in R6/2 mice by 20% (Nguyen et al. 2005).

Excitotoxicity

The overactivation of NMDA glutamate receptors and subsequent excitotoxicity in striatal neurons leading to their death has been thought to have a role in pathogenesis of HD and electrophysiological studies in brain slices from R6/2 mice indeed suggested complex changes in glutamatergic transmission (Cepeda et al. 2003). Therefore, compounds intercepting excessive glutamate release have been tested for therapeutic purposes. Treatment of R6/2 mice with a glutamate antagonist, riluzole, decreased the body weight loss, affected the initial locomotor activity, significantly increased the mean survival (10.2%), and appeared to modulate the aggregation process (Schiefer et al. 2002). Although the beneficial effects of riluzole were not confirmed in a recent study (Hockly et al. 2006), clinical trials were carried out in HD patients. Unfortunately, the latest 3-year trial showed no beneficial effect (Landwehrmeyer et al. 2007). Two drugs affecting the levels of glutamate in synaptic cleft through stimulation of the pre-synaptic metabotropic glutamate receptor 2 (mGluR2) or inhibition of the post-synaptic mGluR5, LY379268, and 2-methyl-6-(phenylethynyl)-pyridine, respectively were also tested for their ability to combat HD in R6/2 mice (Schiefer et al. 2004). Similar to riluzole, both compounds significantly increased the lifespan of R6/2 mice. In one study, the administration of remacemide (NMDA antagonist) or coenzyme Q₁₀ prolonged the survival of R6/2 mice by 15.5% and 14.5%, respectively, and reduced motor deficits, weight loss, and aggregate formation. When the drugs were combined, the survival extended by 31.8% (Ferrante et al. 2002). This combination was not so effective in N171-82Q mice but the lifespan still improved by more than 20% (Schilling et al. 2001). The clinical trial using either of these drugs or in combination, however displayed no significant effects in HD patients (Huntington Study Group 2001). Another NMDA antagonist, memantine, has been reported to retard the disease progression in HD patients (Beister et al. 2004).

Apoptosis

From antiapoptotic drugs, minocycline appeared to be a good candidate for therapeutic trials. Minocycline was shown to inhibit the mitochondrial release of apoptosis inducing factor, proapoptotic protein Smac/Diablo, and cytochrome c, while decreasing the cleavage of proapoptotic factor Bid and activating caspases 1, 3, 8, and 9 (Wang et al. 2003). Intraperitoneal administration of minocycline to R6/2 mice extended the lifespan by 13.5% (Chen et al. 2000). When administered perorally, minocycline failed to improve the symptoms in R6/2 mice (Smith et al. 2003). In another study, minocycline and coenzyme Q₁₀ improved survival of R6/2 mice 11.2% and 14.6%, respectively, and when administered together, the neuropathology and phenotype outcome improved when compared with separate treatments and the lifespan extension was 18.2% (Stack et al. 2006). The results of the clinical studies using minocycline produced promising data (Bonelli et al. 2004) but a long-term clinical trial should be conducted to definitely evaluate the benefits of minocycline in HD patients.

Inhibition of caspases could comprise double beneficial effect in polyQ pathogenesis. First, decreased generation of caspase-cleaved fragments of mutant proteins may result in reduced toxicity and second, execution of apoptosis may be blocked. A broad caspase inhibitor, z-Val-Ala-Asp-fluoromethylketone, improved the rotarod performance of R6/2 mice and extended the lifespan by 25% (Ona et al. 1999). The combined administration of caspases 1 and 3 inhibitors, Tyr-Val-Ala-Asp-chloromethylketone and Asp-Glu-Val-Asp-aldehyde-fluoromethylketone, respectively, increased the survival by 17.3% (Chen et al. 2000).

Other treatment possibilities

Selective serotonin reuptake inhibitors are a group of drugs widely used for the treatment of patients with depression and severe anxiety disorders (Blier and de Montigny 1994). In addition to restoration of serotonin levels in striatum, the treatment of N171-82Q mice with paroxetine reduced brain atrophy, delayed the onset of motor dysfunction, attenuated weight loss, and increased the survival (Duan et al. 2004). Another selective serotonin reuptake inhibitor, sertraline, had similar effects in N171-82Q mice. Upon this treatment, increased BDNF levels, preserved levels of Hsp70 and Bcl-2 in the brains, and enhanced neurogenesis were observed (Duan et al. 2008). The enhanced neurogenesis and increased BDNF levels were also reported in R6/2 mice treated with sertraline (Peng et al. 2008). Importantly, the effective levels of sertraline in mice were comparable to those achievable in humans (Duan et al. 2008).

Control of Ca²⁺ homeostasis seems to be a promising therapeutic strategy for polyQ diseases and particularly the modulation of the IP3R1 activity in ER membrane could provide a good target (Bezprozvanny 2007; Chen et al. 2008b). Inhibiting the excessive Ca²⁺ release from intracellular storages would potentially result in reduced deteriorating events such as calpain activation or mitochondrial dysbalance leading to caspase and apoptosis activation. Although there has been a focus on identifying downstream effectors to address the particular pathogenic event without disturbing other physiological or protective cellular processes, in this case, the relatively proximal component of the intracellular Ca²⁺-dependent network, down-regulation of IP3R1 activity, could be evaluated more extensively.

Another strategy to advance toward an effective therapy of polyQ diseases is the combination of drugs targeting different aspects of these disorders. As mentioned in the section `Excitotoxicity', the synchronous utilization of some compounds resulted in increased therapeutic effects in some cases (coenzyme Q₁₀ with remacemide or minocycline). Combinations of a treatment targeting the polyQ protein abnormal processing with another treatment aimed at a downstream pathogenic process are of particular interest. For example, the concomitant treatment of R6/2 mice with cystamine (prevents polyQ aggregation by TG inhibition) and mithramycin (restores the polyQ-mediated transcriptional dysregulation) extended the mean survival by 40% and markedly improved the neuropathological findings more than a single treatment by any of these two drugs (Ryu et al. 2006).

As described in the section `Role of nuclear localization of mutant polyglutamine proteins', testosterone binds to AR and promotes its uptake by nucleus, the primary site of SBMA pathogenesis (Stenoien et al. 1999; Katsuno et al. 2002). Treatment aimed at testosterone blockage by castration has been shown to be beneficial in SBMA transgenic mice (Katsuno et al. 2002). Chronic adminitration of a luteinizing hormone-releasing hormone analog, leuprorelin, caused decreased serum levels of testosterone in male transgenic AR-97Q mice resulting in amelioration of neuromuscular phenotype and increased lifespan (Katsuno et al. 2003). Importantly, phase 2 clinical trial showed beneficial effects of leuprorelin in SBMA patients and is a good premise for large-scale clinical trials of androgen deprivation (Banno et al. 2009).

In last decade, a great progress has been achieved in understanding the pathomechanism of polyQ diseases, but, unfortunately, the therapy development does not keep up. The challenge for the next decade will be the translation of promising therapeutic strategies from preclinical trials to clinical use.