Spinal and bulbar muscular atrophy (SBMA) was first described in a paper entitled, ‘Progressive bulbar palsy’, in 1897 by Kawahara in Japan [1,2]. The author reported the clinical characteristics of two brothers and their uncle on the maternal side with progressive atrophy of the tongue, dysarthria, dysphagia and gait disturbance. SBMA is also known as Kennedy's disease, named after William R. Kennedy, whose study on 11 patients from two families depicted the clinical, genetic and pathological features of this disorder . As SBMA mainly affects males and SBMA is frequently associated with gynaecomastia, testicular failure and other feminized signs, this disease had been thought to be caused by abnormality of the androgen receptor (AR). The underlying genetic abnormality was determined to be abnormal expansion of a CAG repeat in the AR gene in 1991 . This was the first discovery of a polyglutamine (polyQ)-mediated neurodegenerative disease caused by expansion of a trinucleotide CAG repeat encoding glutamine in the causative gene. To date, nine polyQ-mediated neurodegenerative disordershave been identified , and many molecular biological studies have been undertaken to elucidate the pathogenesis of these diseases and to develop treatment methods. Therapies for SBMA can be broadly classified into two categories: (i) disease-modifying therapies and (ii) symptom-relief therapies. The ideal therapy appears to be a combination of these two potential therapeutic strategies. As to disease-modifying therapies, long-term clinical trials are required to verify that these potential therapies can delay the clinical onset and progression of SBMA by targeting certain clinical events. On the other hand, for symptom-relief therapies such as L-dihydroxyphenylalanine (L-DOPA) for Parkinson's disease, the duration of clinical trials tends to be short. Here we highlight the research findings from which the concept of the ligand- and chaperone-dependent pathophysiology of SBMA has emerged, and discuss disease-modifying therapeutic approaches.
Clinical features of SBMA
Spinal and bulbar muscular atrophy is characterized by premature muscular exhaustion, slowly progressive muscular weakness, atrophy, and fasciculation in bulbar and limb muscles . For instance, bilateral facial and masseter muscle weakness, poor uvula and soft palatal movements, and atrophy of the tongue with fasciculation are observed. The muscle weakness and atrophy in the limbs are either generalized or prominent in the proximal muscles, and are usually symmetrical. Patients also present occasional painful muscle cramps mainly in the lower legs and trunk, and hand tremor [7,8]. The motor involvement is slowly progressive and eventually confines some patients to a wheelchair. SBMA patients may also have mild sensory impairment, although it usually remains subclinical. In most cases, the vibration sense is slightly diminished in the distal lower extremities, but occasionally all of the sensory modalities are slightly disturbed . Deep tendon reflex is diminished or absent with no pathological reflex. Patients with SBMA do not have cerebellar symptoms, dysautonomia, or cognitive impairment. Patients occasionally show signs of androgen insensitivity such as gynaecomastia, impaired spermatogenesis, testicular atrophy, impotence and decreased fertility, some of which are detected before the onset of motor symptoms [7,9–12]. Serum testosterone levels are usually normal or elevated. Abdominal obesity is common, whereas male pattern baldness is rare. Testosterone treatment has been administered to some patients, although it does not affect the progression of SBMA [13–15].
In patients with SBMA, needle electromyography shows neurogenic abnormalities, and distal motor latencies are often prolonged in nerve conduction studies. Both the sensory nerve action potential and sensory evoked potential are reduced or absent in some cases . The serum creatine kinase level is elevated in the majority of patients . Hyperlipidemia, slight hepatic dysfunction, and impaired glucose tolerance or diabetes mellitus are also detected in some patients . Female heterozygous and homozygous carriers are usually asymptomatic, although some have subclinical phenotypes such as high amplitude motor unit potentials or an elevated serum creatine kinase level [19–22]. The two major components of muscle weakness in patients with SBMA are motor neurone degeneration and myopathic degeneration. SBMA patients display myopathic symptoms including elevated serum creatine kinase level and frequent muscle cramps prior to the onset of muscle weakness [7,8,23]. Elucidation of the basis of the neuronal or muscle-specific pathogenesis will help clarify the origin of muscle weakness in patients with SBMA.
Androgen receptor, the causative protein of SBMA, is a 110-kDa nuclear receptor that belongs to the steroid/thyroid hormone receptor family . Upon binding of an androgen, that is, testosterone or dihydrotestosterone, with AR, the AR binds to an androgen response element in the target gene to regulate its expression. AR is essential for the major effects of androgens including normal male sexual differentiation and pubertal sexual development, although AR-independent, nongenomic functions of androgens have been reported [25–27]. AR is expressed not only in primary and secondary sexual organs, but also in nonreproductive organs including the kidney, skeletal muscle, adrenal gland, skin and nervous system, suggesting the far-reaching influence of androgens on a variety of mammalian tissues. In the central nervous system, AR is expressed at relatively high levels in spinal and brainstem motor neurones, which are the same cells that are vulnerable in SBMA. The AR gene is located on chromosome Xq11-12. This approximately 90-kb-long gene contains eight exons coding for the functional domains specific to the nuclear receptor family. The first exon encodes the N-terminal transactivating domain. Exons 2 and 3 encode the DNA-binding domain, whereas exons 4 through 8 encode the ligand-binding domain. The N-terminal transactivating domain, which contains the polyQ region, possesses a major transactivation function (AF-1) that is maintained by interaction with general transcriptional coactivators such as the c-AMP response element binding protein, TAFII130 and steroid receptor coactivator-1. The CAG repeat begins at codon 58 in the first exon of AR. The number of CAG repeats is highly variable due to slippage of DNA polymerase upon DNA replication. Whereas abnormal elongation of CAG repeats causes SBMA, an abnormally low number of CAG repeats increases the risk of prostate cancer . The number of polymorphic CAG repeats in the AR gene normally ranges between 14 and 32, but it ranges between 40 and 62 in SBMA patients . It also shows somatic mosaicism . An inverse correlation has been reported between the CAG repeat size and the age at onset of SBMA [22,31–33]. There was also an inverse correlation between the CAG repeat size and the degree of muscular weakness adjusted by the age of the patient at examination . Intergenerational CAG repeat expansion is observed via predominantly paternal transmission rather than maternal transmission, suggesting that particular instability of the CAG repeat occurs during spermatogenesis [31,34]. The severity of the disease differs in each male member of the same family.
Spinal and bulbar muscular atrophy has been considered to be an X-linked disease, whereas other polyQ diseases show autosomal dominant inheritance. In fact, female SBMA patients have few, if any, clinical manifestations, even though they possess a similar number of CAG repeats in the disease allele as their siblings with SBMA. Reduced mutant AR expression due to X-inactivation may prevent disease manifestation in females. However, hormonal intervention studies in mouse and fly models strongly suggest that a reduced testosterone level prevents nuclear accumulation of the mutant AR protein, resulting in absence of a neurological phenotype in females [35–37]. This view is strongly supported by the observation that homozygous female carriers manifest few symptoms . Therefore, it seems to be inappropriate to regard SBMA as an X-linked recessive inherited disease, but rather its neurological phenotype likely depends on the serum testosterone concentration.
In patients with SBMA, lower motor neurones are markedly depleted through all spinal segments and in brainstem motor nuclei except for the third, fourth and sixth cranial nerves (Figure 1A) [6,38]. The number of nerve fibres in the ventral spinal nerve root is reduced, reflecting motor neuronopathy. Sensory neurones in the dorsal root ganglia were less severely affected, and large myelinated fibres demonstrate a distally accentuated sensory axonopathy in the peripheral nervous system [39,40]. The neurones in Onufrowicz nuclei, intermediolateral columns and Clarke's columns of the spinal cord are generally well preserved. Muscle histopathology shows both neurogenic and myogenic findings; there are groups of atrophic fibres with small angular fibres, fibre type grouping and clumps of pyknotic nuclei as well as variability in fibre size, hypertrophic fibres, scattered basophilic regenerating fibres and central nuclei (Figure 1B,C) .
A pathologic hallmark of most polyQ diseases is the presence of nuclear inclusions (NIs). In SBMA patients, NIs containing the mutant AR are detected in the residual motor neurones in the brainstem and spinal cord  as well as in the skin, testis and some other visceral organs . These NIs are detectable by antibodies that recognize (i) a small portion of the N-terminus of the AR protein; (ii) the expanded polyQ (1C2); (iii) many components of the ubiquitin–proteasome system (UPS); and (iv) molecular chaperone pathways, but not by antibodies against the C-terminus of the protein. This observation suggests that the C-terminus of AR is truncated or masked upon formation of NIs [43–46]. Although NIs are important histopathological findings, their role in the pathogenesis of polyQ diseases has been debated [47–51]. Several studies suggested that inclusion formation might be a cellular response against the toxicity of abnormal polyQ proteins [52–54]. On the other hand, nuclear localization or accumulation of the abnormal proteins has been considered to be decisive for inducing neuronal cell dysfunction and degeneration in polyQ diseases including SBMA [55–59]. Immunohistochemical studies on autopsied SBMA patients and scrotal skin biopsy specimens from SBMA patients using 1C2 antibody showed that diffuse nuclear accumulation of the mutant AR was far more frequently observed than NIs, being distributed in a wide array of central nervous system nuclei and in a greater number of visceral organs than thus far believed (Figure 1D–F) . In neural systems, diffuse nuclear mutant AR accumulation also occurs in unaffected tissues including the basal ganglia, thalamus, hypothalamus, various midbrain, pontine and medullary nuclei, posterior horn, intermediolateral and Clarke's nuclei of the spinal cord, and sensory and sympathetic ganglion neurones, as well as in the affected brainstem and spinal cord motor neurones. The distribution of NIs was similar to the distribution of diffuse nuclear accumulation of mutant AR among neural and nonneural tissues, although the frequency of NIs in each tissue was far less than the frequency of diffuse nuclear accumulation (Figure 1d) . It is of note that the extent of diffuse nuclear accumulation of mutant AR in motor and sensory neurones of the spinal cord was strongly correlated with the CAG repeat length, but not with the number of NIs in the spinal motor and sensory neurones . Accumulating evidence suggests that NIs are not the toxic polyQ species, but that an oligomeric form of mutant AR may be the major pathogenic species. In view of the time course of the disease, diffuse nuclear accumulation of mutant proteins with an expanded polyQ tract might be an early event prior to NI formation, which is closely related to manifestation of neuronal dysfunction [35,60–62]. However, the molecular pathogenetic process by which diffuse nuclear mutant AR accumulation induces neuronal dysfunction and death still remains unclear. One possibility is that the interaction of transcriptional regulatory proteins with polyQ-expanded proteins results in aberrant transcriptional regulation which may result in neuronal dysfunction and cell death [63,64].
Another important observation is the presence of cytoplasmic mutant AR inclusions in neural and nonneural tissues . In neural tissues, cytoplasmic accumulation is restricted to certain neuronal populations including dorsal root ganglia neurones, the mammillary body, hypothalamus, facial motor nucleus, and anterior and posterior horn neurones (Figure 1e). Among nonneural tissues, cytoplasmic inclusions occur in certain organs. Cytoplasmic inclusions are detectable by antibodies that recognize the Golgi apparatus . Colocalization of a polyQ-expanded mutant protein with the Golgi apparatus has also been reported for ataxin-2 , although the significance of this colocalization remains unclear. Expression of polyQ-expanded mutant ataxin-2 disrupted the normal morphology of the Golgi complex and increased cell death . On the other hand, lysosomal occurrence of other mutant proteins with an expanded polyQ tract in neurones has been reported in dentatorubral-pallidoluysian atrophy  and Huntington's disease (HD) . Lysosomal localization of polyQ-expanded mutant proteins suggests a lysosomal autophagic degradation process acting independently of the ubiquitin–proteasome pathway in polyQ diseases . Additionally, the reason why particular tissues have predominantly nuclear or cytoplasmic accumulation of polyQ-expanded mutant AR is not known. Differences in the predominant pathway of degradation of the mutant AR could influence the intracellular site of accumulation of mutant AR and eventual cell toxicity. One important question is whether cytoplasmic accumulation of mutant AR exerts cytotoxicity in neural and nonneural tissues. Cytoplasmic accumulation of mutant AR  as well as the accumulation of other mutant proteins involving an expanded polyQ tract [65,68,69] in Golgi apparatus and lysosomes indeed has been found to induce cytotoxicity. The accumulation of mutant protein with expanded polyQ tract in Golgi apparatus or lysosomes increases death of cultured cells via activation of apoptosis-related effectors such as caspase-3 [65,69,70]. It should be noted that histologically or immunohistochemically evident accumulation of a mutant protein is not necessarily cytotoxic, while microaggregates at the molecular level that are histologically undetectable may exert cytotoxicity. Indeed, excessive accumulation of mutant AR in aggresomes was found to protect cells from a cytotoxic form of mutant AR . However, an immunohistochemical study on autopsied SBMA patients strongly suggested that cytoplasmic accumulation of mutant AR is related to mutant AR-mediated cytotoxicity and eventual symptom manifestation . For instance, the pancreas showed cytoplasmic accumulation of mutant AR without obvious nuclear accumulation . SBMA patients have an elevated serum glucose level and impaired glucose tolerance, suggesting islet cell dysfunction in the pancreas . Although further studies on the significance of cytoplasmic accumulation of mutant AR are needed, nuclear accumulation of the mutant AR protein in motor neurones appears to cause motor neurone dysfunction, while cytoplasmic accumulation may underlie some visceral and possibly some neuronal dysfunction in patients with SBMA. The pathologic process is likely to differ in different tissues, being more prominent in the nuclei of motor neurones, but mainly cytoplasmic in certain neuronal populations and visceral organs. This cytoplasmic mutant AR is not ubiquitinated in contrast to nuclear-accumulated mutant AR, particularly the heavily ubiquitinated NIs, suggesting that modification of mutant AR differs in the nucleus and cytoplasm. We also need to further clarify which degradation process affecting mutant AR is most active in a given tissue, for example, lysosomal in certain viscera vs. via the ubiquitination pathway in most neural tissues. Taken together, diffuse nuclear accumulation of mutant AR is apparently a cardinal pathogenetic process underlying neuronal dysfunction and eventual cell death, while cytoplasmic accumulation may also contribute to the pathophysiology of SBMA.
Two mechanisms of the involvement of polyQ-expanded mutant AR in the pathogenesis of SBMA have been proposed: mutant AR acquires a toxic property, damaging motor neurones; or loss of normal AR function induces neuronal degeneration . As androgens have trophic effects on neuronal cells, one can assume that loss of AR function may play a role in the pathogenesis of SBMA. Expansion of the polyQ tract mildly suppresses the transcriptional activities of AR, probably because it disrupts the interaction between the N-terminal transactivating domain of AR and transcriptional coactivators. Although this loss of function of AR may contribute to androgen insensitivity, a gain of toxic function of the mutant AR due to the expanded polyQ tract has been believed to play a pivotal role in the pathogenesis of SBMA. This hypothesis is supported by the observations that patients with severe testicular feminization lacking AR function and AR knockout mice do not have motor impairment . Moreover, a transgenic mouse model expressing a protein composed of expanded polyQ driven by the human AR promoter demonstrated motor impairment, suggesting that the expanded polyQ protein is sufficient to induce the pathogenic process , whereas the male AR knock-in mouse model of SBMA has signs of androgen insensitivity such as decreased fertility, progressive abnormalities of germ cell maturation and the Sertoli cell cytoskeleton, and testicular atrophy . However, a recent study revealed that the absence of endogenous normal AR protein in SBMA transgenic mice had deteriorative effects on neuromuscular and endocrine-reproductive features of these mice, although this mouse model expressing AR with expanded polyQ tract does not display signs of androgen insensitivity in the presence of the normal endogenous mouse AR gene . Collectively, both the gain-of-function mutant protein toxicity and loss of normal AR protein function are the basis for motor neurone degeneration in SBMA, whereas impairment of AR function possibly causes signs of androgen insensitivity.
The fact that AR has a specific ligand, that is, testosterone, renders the pathogenesis of SBMA unique among polyQ diseases. An in vitro study using transfected COS cells showed that AR localized in the nucleus in the presence of testosterone, while AR remained largely in the cytoplasm in the absence of the hormone . The AR is normally confined to a multiheteromeric inactive complex in the cell cytoplasm, and translocates into the nucleus in a ligand-dependent manner. Moreover, the half-life of AR is prolonged in the presence of its ligand, suggesting ligand-dependent stabilization of AR [76,77]. This intracellular trafficking and stabilization of AR appear to play important roles in the pathogenesis of SBMA.
We previously generated transgenic mice expressing the full-length human AR gene containing either 24 or 97 CAG repeats under the control of a cytomegalovirus enhancer and a chicken β-actin promoter . This model recapitulated not only the neurologic disorder but also the phenotypic difference with gender which is a specific feature of SBMA. Mice that expressed AR with 97 CAG repeats (AR-97Q) exhibited progressive motor impairment, while none of the mice that expressed AR with 24 CAG repeats (AR-24Q) showed abnormal phenotypes . The AR-97Q mice demonstrated small body size, short lifespan, progressive muscle atrophy, and weakness as well as reduced cage activity, all of which were markedly pronounced and accelerated in the male AR-97Q mice, but were either not observed or far less severe in the female AR-97Q mice. Western blot analysis revealed the transgenic AR protein smearing from the top of the gel in proteins isolated from the spinal cord, cerebrum, heart, muscle and pancreas. Although the male AR-97Q mice had greater amounts of smearing protein than their female counterparts, the female AR-97Q mice had a greater amount of monomeric AR protein. Diffuse nuclear staining of AR-97Q and less frequent NIs as detected by 1C2, were demonstrated in neurones of the spinal cord, cerebrum, cerebellum, brainstem, and dorsal root ganglia as well as in nonneuronal tissues such as the heart, muscles and pancreas. Male AR-97Q mice showed markedly more abundant diffuse nuclear staining and NIs than females, in agreement with the gender differences in symptoms and Western blot profile. Despite the profound gender difference in pathogenic AR protein expression, there was no significant difference in the mRNA level of transgene expression between the male and female AR-97Q mice. Other laboratories have also generated various animal models expressing the full-length human AR with expanded polyQ tract, almost all of which display phenotypic expression of motor dysfunction with gender effects . Ligand-dependent neurodegeneration has also been observed in a fruit fly model of SBMA . These observations indicate that the ligand plays important roles in the gender difference of phenotypes, especially with regard to its interactions with mutant AR in the posttranscriptional stage.
The dramatic gender difference of phenotypes led us to attempt hormonal interventions as a treatment for SBMA. First, we castrated male AR-97Q mice in order to reduce their testosterone level . Next, leuprorelin, a potent luteinizing hormone-releasing hormone (LHRH) analogue, was administered subcutaneously to noncastrated AR-97Q mice . Leuprorelin suppresses the release of the gonadotrophins, luteinizing hormone and follicle-stimulating hormone. Two to 4 weeks after the start of leuprorelin administration, the serum testosterone level decreased to the level achieved by surgical castration. Leuprorelin-treated male AR-97Q mice showed profound improvement of symptoms and histopathologic findings, and reduced nuclear localization of the mutant AR compared with the untreated, noncastrated male AR-97Q mice [35,37]. The body weight, motor function, and lifespan of male AR-97Q mice significantly improved by castration or leuprorelin administration. Western blot analysis and histopathologic studies revealed diminished nuclear accumulation of mutant AR in the male AR-97Q mice that had undergone castration or leuprorelin administration compared with that in untreated, noncastrated male AR-97Q mice. These results suggest that testosterone has toxic effects in AR-97Q mice by accelerating nuclear translocation of the mutant AR and promoting stabilization of the mutant AR protein. On the contrary, castration and leuprorelin administration each prevented nuclear localization and stabilization of the mutant AR by reducing the testosterone level. Nuclear localization of the mutant protein with expanded polyQ tract is likely to be important for inducing neuronal cell dysfunction and degeneration in the majority of polyQ diseases. It thus appears logical that a reduction in testosterone level improved the phenotypic expression of SBMA by preventing nuclear localization of the mutant AR. Testosterone deprivation reversed motor dysfunction in another transgenic mouse model of SBMA .
When leuprorelin was subcutaneously administered to male AR-97Q mice every 2 weeks starting at 5 weeks of age, leuprorelin initially increased the serum testosterone level by acting as an agonist at the LHRH receptor, but subsequently reduced it to an undetectable level. Leuprorelin-treated AR-97Q mice showed deterioration of body weight, gait and performance of the rotarod task at 8–9 weeks of age, during the time when the serum testosterone level initially increased through the agonistic effect of leuprorelin . This elevation in testosterone level was transient and was followed by suppression of testosterone production and sustained reversibility of polyQ pathogenesis. Intriguingly, immunostaining of tail specimens, sampled from the same individual mouse, demonstrated an increase in the number of muscle fibres with nuclear 1C2 staining at 4 weeks of leuprorelin administration, although the number of muscle fibres with nuclear 1C2 staining decreased after another 4 weeks of treatment . These results indicate that testosterone deprivation is sufficient to reverse both the symptomatic and pathologic phenotypes in AR-97Q mice.
Successful treatment of AR-97Q mice with leuprorelin led us to perform testosterone blockade therapies in SBMA patients . In a preliminary open trial, treatment with leuprorelin for 6 months significantly reduced nuclear accumulation of mutant AR in the scrotal skin of SBMA patients, suggesting that androgen deprivation interrupts the pathogenic process of human SBMA . Another trial on a larger scale is currently underway to verify the clinical benefits of leuprorelin in SBMA patients.
Heat shock protein (Hsp) 90, a molecular chaperone, is essential for the function and stability of the AR, the C-terminus of which has a high affinity for Hsp90. Hsp90 induces a conformational change in AR that is required for its nuclear translocation after ligand activation [24,81,82]. Hsp90 functions in multicomponent complexes of chaperone proteins including Hsp70, Hop, Cdc37 and p23, leading to the folding, activation and assembly of Hsp90 client proteins . Two main Hsp90 complexes are thought to exist: one complex is a stabilizing form with Cdc37 and p23 and this complex stabilizes Hsp90 client proteins including AR, while the other complex is a proteasome-targeting form with Hsp70 and Hop and it directs Hsp90 client proteins to proteasome degradation (Figure 2) [84–87]. p23 is thought to modulate Hsp90 activity during the last stages of the chaperoning pathway, leading to stabilization of Hsp90 client proteins in an ATP-dependent manner . Hop is known to independently bind with both Hsp90 and Hsp70, thereby promoting the Hsp90/Hsp70 linkage, and is thought to direct the triage decision for client proteins by bridging the Hsp90–Hsp70 interaction . Hsp90 inhibitors inhibit the ATP-dependent progression of the Hsp90 complex towards the stabilizing form and shift it to the proteasome-targeting form, resulting in proteasomal degradation of the Hsp90 client protein (Figure 2) [89,90]. Steroid receptors, including the progesterone receptor and the glucocorticoid receptor, were the first Hsp90 client proteins to be identified [91,92]. As for AR, Hsp90 is essential for maintaining its high ligand-binding affinity and for its stabilization [81,93]. In practice, Hsp90 inhibitors reduce the binding affinity of AR for androgens, and induce the degradation of AR [93,94]. Numerous oncoproteins belonging to the Hsp90 client protein family are selectively degraded in the UPS by Hsp90 inhibitors, and 17-allylamino-17-demethoxygeldanamycin (17-AAG), a first-in-class Hsp90 inhibitor, is now under clinical trials as a novel molecular-targeted agent for a wide range of malignancies . In addition, Hsp90 inhibitors have been shown to have some neuroprotective effects against various stresses such as drug-induced toxicity, oxidative stress, and oxygen glucose deprivation [96–99]. AR also belongs to the Hsp90 client protein family, and is degraded in the presence of Hsp90 inhibitors [85,93,94]. Therefore, we explored the possibility of using 17-AAG as a therapeutic agent for neurodegenerative diseases by examining its effects on mutant AR in cultured cells and in a mouse model of SBMA. We found that 17-AAG inhibits nuclear accumulation of this protein in cultured cells, leading to marked amelioration of the motor phenotype of AR-97Q mice without detectable toxicity . Of interest is the finding that 17-AAG preferentially targeted mutant AR rather than wild-type AR to proteasomal degradation. A high association between p23 and AR containing expanded polyQ tract renders the mutant AR more sensitive to Hsp90 inhibitors than wild-type AR . Western blot and filter trap analyses in AR-97Q mice both showed that 17-AAG significantly reduced the amount of the insoluble high-molecular-weight complex of mutant AR as well as the amount of soluble monomer of mutant AR in the spinal cord and skeletal muscle. Moreover, in an immunostaining study of nervous tissue from AR-97Q mice, 17-AAG significantly reduced the amount of diffuse nuclear-accumulated AR, suggesting that 17-AAG had a curative effect on SBMA by reducing the amounts of oligomers and the soluble monomeric form of the mutant AR. Alternatively, 17-AAG may inhibit aggregation of mutant AR by inducing Hsp70 and Hsp40 expression. Hsp90 inhibitors cause disassociation of heat shock factor-1 (HSF-1) from the Hsp90 complex and trimerization of the HSF-1, thereby resulting in HSP activation . The Hsp90 inhibitor, geldanamycin, induced Hsp70 and Hsp40 expression in transfected COS-1 cells, thereby inhibiting polyQ-induced abnormal aggregation of huntingtin protein . However, as 17-AAG displayed only a limited ability to induce Hsp70 and Hsp40 expression in mouse tissue , the large decrease in the amount of AR seen in the insoluble fraction in vivo, rather than being a result of HSP induction, may be due to the potent ability of 17-AAG to degrade the soluble monomeric form of the mutant protein, thereby preventing its aggregation . Furthermore, assembly of Hsp90 and its cofactors into complexes is required for retrograde, dynein-dependent movement of several steroid receptors [102,103]. Thus, in cells that do not express HSF-1, Hsp90 inhibitors inhibit the translocation of AR to the nucleus and prevent ligand-dependent aggregation of the polyQ-expanded AR by inhibiting dynein-dependent AR trafficking .
Among the other proposed therapeutic approaches we previously studied [35,37,105,106], the efficacy of 17-AAG most closely approximated the successful hormonal therapy using the LH-RH analogue, leuprorelin . However, unlike leuprorelin, the Hsp90 inhibitor 17-AAG holds enormous potential for application to a wide range of neurodegenerative diseases in addition to SBMA as previously reported [107–109]. We regard this general versatility as very important for the development of Hsp90 inhibitors as a treatment for neurological disorders. The strategy behind Hsp90 inhibitors differs from previous strategies employed against polyQ diseases, which unavoidably allowed abnormal protein to remain and placed much value mainly on inhibition of protein aggregation. We consider that the ability to facilitate degradation of disease-causing proteins by modulation of Hsp90 function would be of value when applied to SBMA and other neurodegenerative diseases. There is no doubt that reduction of the amount of the main culprit protein would have a curative effect against various neurodegenerative diseases. In fact, one therapeutic approach that directly reduces the level of abnormal protein by RNA interference has already proven beneficial in various mouse models of polyQ diseases and amyotrophic lateral sclerosis [110–112].
17-AAG-induced degradation requires a well-preserved proteasome function [89,90,100,113]. However, a question as to whether the UPS is impaired or not in patients with SBMA has been raised concerning this UPS-dependent therapy . It is generally accepted that the UPS is involved in the pathology of polyQ diseases, as many components of the UPS and molecular chaperones are known to colocalize with polyQ-containing NIs [114,115]. Previous studies performed in cultured cell models suggested that the UPS is impaired in patients with polyQ diseases [54,116–118]. If this hypothesis were true, 17-AAG would not be able to exert its pharmacological effect on polyQ diseases. In this regard, recent studies using in vivo proteasome assays have raised important questions as to whether patients with polyQ diseases have an impaired UPS [53,119,120]. It has been reported that neuronal dysfunction developed without significant impairment of the UPS in a mouse model of SCA7 . Consistent with this, it was also demonstrated that proteasome impairment did not contribute to the pathogenesis of HD in a mouse model . Furthermore, in conditional mouse models of polyQ disease, genetic loss of the abnormal gene product led to rapid clearance of pre-existing polyQ-mediated NIs and reversible improvement of the abnormal phenotypes [121,122]. If the UPS were irreversibly damaged in patients and animal models of polyQ diseases, then the amount of pre-existing NIs would not decrease. We therefore consider that treatment with 17-AAG, which takes advantage of a self-clearing system to target disease-causing proteins, is a reasonable therapeutic strategy against polyQ-related and other neurodegenerative diseases.
Induction of heat shock proteins
Many components of the UPS and molecular chaperones are known to colocalize with polyQ-containing NIs, implying that these proteins are involved in neurodegeneration in polyQ diseases. HSPs are classified into different families according to molecular size: Hsp100, Hsp90, Hsp70, Hsp60, Hsp40 and small HSPs . These HSPs are either constitutively expressed or inducibly synthesized after cellular stress. HSPs play important roles in maintaining correct folding, assembly, and intracellular transport of newly synthesized proteins. For example, Hsp70 and Hsp90, which are essential components of the AR–chaperone complex in the cell cytoplasm, regulate the function, nuclear translocation, and degradation of AR . Under toxic conditions, the synthesis of HSPs is rapidly up-regulated, and nonnative proteins are refolded as a consequence. Therefore, forced overexpression of HSPs resulted in acquisition of tolerance against various types of stresses, and protection against apoptosis in various disease models . In various polyQ disease models, both genetic and pharmacological overexpression of HSPs have been shown to suppress aggregate formation and cellular toxicity [109,114,126–128]. Hsp70 cooperates with Hsp40 in functioning as a molecular chaperone. These HSPs have been proposed to prevent the initial conformation conversion of mutant polyQ-containing protein from a random coil to a β-sheet, leading to attenuation of toxic oligomer formation . Overexpression of Hsp70, together with Hsp40, inhibited the toxic accumulation of abnormal polyQ-containing protein and suppressed cell death in a variety of cellular models of polyQ diseases including SBMA [114,129]. Hsp70 has also been shown to facilitate proteasomal degradation of abnormal AR protein in a cell culture model of SBMA . The favourable effects of Hsp70 have been verified in studies using mouse models of polyQ diseases. Overexpression of the inducible form of human Hsp70 markedly ameliorated the symptomatic and histopathological phenotypes of the SCA1 mouse model  and AR-97Q mice . These beneficial effects were dependent on the dose of the Hsp70 gene and correlated with the reduction in the amount of nuclear-accumulated mutant AR protein . It should be noted that Hsp70 overexpression also significantly reduced the amount of the soluble form of mutant AR, suggesting that Hsp70 overexpression accelerated the degradation of mutant AR in AR-97Q mice.
Favourable effects obtained by genetic modulation of HSPs suggest that pharmacological induction of molecular chaperones might be a promising approach for the treatment of SBMA and other polyQ diseases. Many studies have shown that HSP induction by Hsp90 inhibitors exerted potentially neuroprotective effects in models of HD [101,132,133], tauopathies [134–136], Parkinson's disease [137–139], stroke [140,141] and autoimmune encephalomyelitis . As for polyQ diseases, Sittler et al.  first showed that geldanamycin significantly suppressed aggregation of mutant huntingtin in a cultured cell model of HD via induction of Hsp70 and Hsp40 heat shock response. Geranylgeranylacetone (GGA), an acyclic isoprenoid compound with a retinoid skeleton, has been shown to strongly induce HSP expression in various tissues . This compound has been used as an oral anti-ulcer drug. Oral administration of GGA up-regulated the levels of Hsp70, Hsp90 and Hsp105 via activation of HSF-1 in the central nervous system and inhibited nuclear accumulation of the pathogenic AR protein, resulting in amelioration of polyQ-dependent neuromuscular phenotypes of the AR-97Q mice . Thus, enhancement of cellular defences using Hsp90 inhibitors and GGA is a reasonable clinical approach for the treatment of neurodegenerative diseases. The ability of Hsp90 inhibitors to significantly induce the expression of HSPs has been demonstrated in cultured cells and a fly model, but not in mammals. As 17-AAG had only a limited ability to induce Hsp70 expression in mouse tissue , further studies should be performed to address to what extent Hsp90 inhibitors can induce the expression of HSPs in mouse models of neurodegenerative disorders other than SBMA.
On the other hand, several studies suggest that polyQ elongation interferes with the protective cellular responses against cytotoxic stress . Transfected cells expressing truncated AR composed of the first N-terminal 127 amino acids of human AR with an expanded polyQ tract showed delayed induction of Hsp70 after heat shock compared with that in transfected cells expressing the full-length AR . Bates and colleagues reported progressive decreases in the expression of Hsp70 and Hsp40 in the brain lesion of an animal model of HD , and such decreases were also observed in AR-97Q mice . The threshold of HSP induction is known to be relatively high in spinal motor neurones . Taken together, impairment of the capability of HSP induction is implicated in the pathogenesis of motor neurone degeneration in SBMA as HSPs are potent suppressors of polyQ toxicity.
Biomarkers for clinical trials
As SBMA is a slowly progressive disease and its precise natural history has not been well elucidated, long-term clinical trials are necessary to assess whether certain drugs can alter the natural progression of the disease. Suitable biomarkers that reflect the pathogenesis and severity of SBMA, are necessary to be able to assess the therapeutic efficacy and to improve the power and cost-effectiveness of longitudinal drug trials. Punch biopsy of the scrotal skin is safe and easy to perform on patients, whereas it is not practical to obtain a biopsy specimen from the central nervous system. We studied a biomarker of SBMA, that is, the degree of nuclear accumulation of mutant AR in epithelial cells of scrotal skin biopsy samples, which can be used as a surrogate endpoint in therapeutic trials . In that study, the degree of nuclear accumulation of mutant AR in scrotal skin biopsy samples from 13 SBMA patients was assessed by 1C2 staining (Figure 1f). The percentage of cells with nuclear 1C2 staining among the scrotal skin epithelial cells tended to be correlated with that in spinal motor neurones among five autopsied SBMA cases, and it was positively correlated with the CAG repeat length and inversely correlated with the functional Activities of Daily Living scale as assessed by the Norris score on limbs among the 13 SBMA patients . The results demonstrate that the percentage of cells with nuclear 1C2 staining in scrotal skin biopsy samples could predict the pathogenic process in the motor neurones of patients with SBMA. Moreover, subcutaneous injections of leuprorelin in SBMA patients reduced both the intensity and frequency of diffuse nuclear 1C2 staining in scrotal skin epithelial cells during the first 4 weeks of therapy and this effect was markedly enhanced after 12-week treatment, suggesting that 1C2 immunostaining of scrotal skin biopsy samples for nuclear mutant AR is a practical procedure for estimating the severity of SBMA pathogenesis in the nervous system . Based on the observations described above, the degree of 1C2-stained nuclear mutant AR accumulation in biopsied scrotal skin is likely to be a potent biomarker reflecting the pathogenic process of SBMA. Since the degree of nuclear accumulation of mutant AR in biopsied scrotal skin appears to be a promising surrogate endpoint, further trials are necessary to evaluate this biomarker by confirming that the change in scrotal skin findings correctly predicts the true clinical outcome event such as becoming wheelchair-bound, aspiration pneumonia or death.