Portal de Periódicos da CAPES

Molecular analyses of Machado-Joseph disease

2003; Karger Publishers; Volume: 100; Issue: 1-4 Linguagem: Inglês

10.1159/000072862

1424-8581

Taeko Kobayashi, Akira Kakizuka,

Neurological disorders and treatments

Machado-Joseph disease (MJD) is an inherited neurodegenerative disorder with an autosomal dominant inheritance. The brain regions affected by MJD include the Clarke column and spinocerebellar tracts, pons, dentate nuclei, substantia nigra, anterior horn cells, red nuclei, and peripheral nerves, but MJD is distinguished from other inherited cerebellar ataxias because it spares the olives and cerebellar cortex. The established phenotypes of MJD include gait and limb ataxia, ophthalmoplegia, dystonia, amyotrophy, dysarthria, rigidity, pyramidal signs, facial and lingual fasciculations, and bulging eyes (the most characteristic feature).In 1977 Romanul proposed the existence of a new disease entity based on apparent common clinical symptoms in four family pedigrees, including the Machado family and Joseph family (Nakano et al., 1972; Rosenberg et al., 1976). Since these pedigrees were all of Portuguese Azorean origin, the new disease was referred to as "Azorean disease" at the time (Romanul et al., 1997; Nakamoto et al., 1998). Later, the existence of several families with similar clinicopathological phenotypes of non-Azorean origin was reported, and therefore this disease has come to be more preferably known as Machado-Joseph disease (Healton et al., 1980; Lima and Coutinho, 1980; Sakai et al., 1983; Suite et al., 1986; Yuasa et al., 1986; Takiyama et al., 1989).In 1993, a linkage study of large Japanese pedigrees identified the MJD locus on a distal region of the long arm of chromosome 14 (Takiyama et al., 1993). Almost simultaneously, a locus covering a similar region on chromosome 14 was reported as the thirdly identified locus of spinocerebellar ataxia (Stevanin et al., 1994). In this review, we introduce our long-term studies of MJD, covering the identification of MJD1 (the gene responsible for MJD) to the recent molecular analysis of MJD, and from the data obtained from MJD analyses, we discuss the potential general molecular mechanisms that underlie a broad spectrum of neurodegenerative disorders, if not all of them.Neurodegenerative disorders have a considerable range of signs and symptoms: dementia, ataxia, movement disorders, and so forth. However, from a more general point of view, they share several common features, e.g., the inheritances that result in these disorders are usually autosomal dominant, in many cases disease symptoms appear after middle age, and these symptoms progress after appearing. The most prominent pathology is neuronal cell loss and degeneration, although each disorder has its own regions of the central nervous system that are susceptible. Furthermore, for several inherited diseases, such as Huntington's disease (HD) and dentatorubral-pallidoluysian atrophy (DRPLA), clinical symptoms worsen and the age of onset becomes earlier in each succeeding generation. This phenomenon is called "anticipation." These features suggest that common mechanisms underlie these inherited neurodegenerative disorders, but the real molecular bases of these mechanisms have long remained unknown.In retrospect, the first breakthrough was made in 1991 by La Spada et al. (1991). They reported the androgen receptor (AR) gene as the gene responsible for spinobulbar muscular atrophy (SBMA), an X-linked recessive neurodegenerative disorder. The gene responsible for HD, the most typical autosomal dominant neurodegenerative disease, was identified in 1993 (The Huntington's Disease Collaborative Research Group, 1993). Both disorders have totally different symptoms, inheritances, and chromosomal loci (on chromosomes X and 4 in SBMA and HD, respectively), and they both are caused by genes encoding nonhomologous proteins. However, patient-related mutations were expansions (more than 40 repeats) of polyglutamine (polyQ)-coding CAG repeats. The polyglutamines were located near the N-terminal portion in the resulting proteins. Surprisingly, in both diseases, patients with longer expansions appeared to have more severe clinical symptoms and to manifest an earlier disease onset. Even more surprisingly, the highly expanded CAG repeats in HD appeared to be unstably transmitted to succeeding generations, with a tendency toward further elongation, typically in paternal transmission (Duyao et al., 1993). This observation provided a simple but convincing molecular basis of anticipation. Thus, suddenly the possibility was raised that expansions of CAG repeats or other triplet repeats are also responsible for other inherited neurodegenerative disorders, especially those with anticipation.On the assumption that several other inherited neurodegenerative disorders are also caused by CAG triplet expansions in yet unknown genes, we first searched for novel CAG-containing genes that are expressed in the human brain. Using an oligonucleotide with a (CTG)13 repeat as a probe, which we expected to hybridize strongly to CAG repeats and weakly to CGG repeats, we screened a human brain cDNA library. We succeeded in isolating about 30 clones containing CAG repeats, but not CGG repeats (Kawaguchi et al., 1994). One of them consisted of 1,776 bp with a long open reading frame encoding a soluble cytoplasmic protein with an amino acid length of 360 (Kawaguchi et al., 1994) (Fig. 1A). The CAG repeat was predicted to be translated into a glutamine tract, as observed in SBMA and HD, but, instead, was located near the C-terminal portion. Northern blot analysis revealed that transcripts were faintly detectable in all of the tissues examined, with the exception of testis, in which strong expression was observed. Reverse transcriptase–polymerase chain reaction (RT-PCR) analysis of human brain mRNA demonstrated expression of both alleles with different lengths of CAG repeats, suggesting that each allele had lengths of CAG repeats that were polymorphic in nature. In addition, the originally identified stop codon was later found to be polymorphically changed to a tyrosine-coding TAC in about 60% of the alleles in Japanese (Fig. 1B) (Goto et al., 1997). A splice variant with a different C-terminal end (Fig. 1B) was also identified (Nakamoto et al., 1998).Despite our success with RT-PCR, we were unable to amplify genomic fragments containing the CAG repeats by PCR using any PCR primers that were constructed from the coding sequence. We assumed the presence of an intron close to the CAG repeat and thus screened a human genomic library to isolate the corresponding genomic clones; in the process, we obtained four related genomic fragments. We gave the symbol MJD1 to the gene corresponding to the cDNA and the symbols MJD2, MJD3, and MJD4 to the other three putatively related genes (Kawaguchi et al., 1994). MJD1 and the other three genomic fragments were mapped at 14q32.1, 8q23, 14q21, and Xp22.1, respectively (Kawaguchi et al., 1994). The physiological functions of MJD1 remain to be clarified. However, MJD1 has two ubiquitin-interacting motifs in the N-terminal side of polyQ and carries another motif, named "Josephin," at the N-terminus (Fig. 1C). A recent report shows that MJD1 is able to bind to HHR23A and HHR23B (the human homologs of the yeast DNA repair protein RAD23) through their ubiquitin-like (UBL) domains (Wang et al., 2000).As predicted, we identified an intron just upstream of the CAG repeat in MJD1. This intron sequence allowed us to amplify by PCR the MJD1 genomic sequence surrounding the CAG repeat. Then, we compared the CAG repeat numbers in MJD1 of healthy individuals with those of clinically diagnosed MJD patients (Fig. 2A) (Kawaguchi et al., 1994). Consistent with our RT-PCR analysis on human brain mRNA, CAG repeat numbers in MJD1 were found to be polymorphic in healthy individuals, from 13 to 36 repeats. The most frequent allele was found to contain 14 repeats, representing about 30% of the total, followed by those with 27, 28, and 21 repeats, representing approximately 15% each. In contrast, MJD patients were found to have longer CAG repeats, from 62 to 84 repeats, and a clear reverse correlation was observed between the repeat numbers and the age of disease onset (Fig. 2B). One repeat elongation was estimated to accelerate onset by approximately 2.6 yr. A recent report shows that rare intermediate repeat lengths, such as 53 and 54 repeats, are also pathogenic (van Alfen et al., 2001). In patients with CAG expansions in both MJD1 genes, onset occurred much earlier than predicted from the repeat numbers, indicating a gene-dosage effect.At almost the same time, an additional two inherited neurodegenerative diseases, spinocerebellar ataxia type 1 (SCA1) and DRPLA, were found to be caused by the expansion of CAG repeats in different responsible genes (Banfi et al., 1993; Koide et al., 1994; Nagafuchi et al., 1994). It was predicted that all of the CAG repeats that had been identified as causing these five inherited neurodegenerations by expansion were transcribed and translated into polyQ stretches as portions of apparent intracellular proteins. Thus, the simplest and easiest idea to test was that expanded polyQ (ex-polyQ) itself plays a toxic function that is causally related to the neurodegeneration.To test this idea, expression vectors for MJD cDNAs with various CAG repeat lengths were transfected in cultured COS7 cells and neuronal PC12 cells, and their phenotypes were examined (Fig. 3A) (Ikeda et al., 1996; Yasuda et al., 1999). All of these proteins were expressed with a hemagglutinin epitope (HA) being tagged in their N- terminals. No obvious morphological changes were observed by the expression of full-length MJD proteins with normal (MJD27 and MJD35) and even diseased (MJD79) lengths of polyQs, nor were they observed in a C-terminal portion of the MJD protein with a normal polyQ length (Q27C). Conversely, cell death was induced when the C-terminal portion with ex-polyQ (Q79C) and an ex-polyQ itself (Q79) were expressed (Fig. 3B) (Yasuda et al., 1999). An immunocytochemical analysis using an anti-HA antibody showed that Q79C and Q79 had stained punctate patterns, suggesting that they aggregated in the cells (Fig. 3C) (Yasuda et al., 1999). In cells with such ex-polyQ aggregates, nuclear condensation and fragmentation of DNA—phenotypes related to apoptosis—were observed. The strength of the aggregates and cell death were parallel, both depending on the length and amount of ex-polyQ. Interestingly, as the protein portions surrounding polyQ became shorter, the strength of the aggregates and the amount of induced cell death increased (see below).To investigate whether ex-polyQ could also induce neurodegeneration or neuronal cell death in vivo, we created transgenic mice in which a similar set of MJD proteins (Q79C, Q79, Q35C, and MJD79) were expressed specifically in mouse Purkinje cells using the L7 gene cassette (Ikeda et al., 1996). All of the Q79C transgenic mice, their F1 offspring, and two of six Q79 transgenic mice were found to have an ataxic posture and gait disturbance. The affected cerebellums were extremely atrophic, which could be seen at a glance. Their volumes were approximately one eighth those of the cerebellums of their non- transgenic littermates (Fig. 3D) (Ikeda et al., 1996). The target Purkinje cells were scarce. However, we obtained clear evidence of cell death. In contrast, none of the Q35C and MJD79 proteins in the transgenic mice showed any abnormal phenotypes.These results clearly show that the ex-polyQ is capable of causing cell death in cultured cell and neurodegeneration in transgenic mice. In contrast, the full-length MJD proteins, even those with ex-polyQs, appeared to have a trivial amount of cellular toxicity, if toxicity was present at all, and the toxicity only became evident when certain C- terminal portions with ex- polyQs were cut off and released in the cells. Based on these results, we suggested a "processing model" that proposes that cleavage or processing is a crucial step in the pathogenesis of the polyQ diseases, namely, that the brain region-specific cleavage of the diseased proteins with ex-polyQs leads to the degeneration of specific brain regions for each disease. This model is able to account for the apparent discrepancy between the affected brain regions and expression patterns of the genes responsible for the polyQ diseases.To find further evidence to substantiate the processing model, we attempted to identify cultured cells in which there is MJD protein processing activity. For this purpose, we expressed MJD79 in a variety of cell lines and examined these MJD79 proteins immunocytochemically. Initially, we were unable to find any evidence suggesting cleavage of the MJD protein in any of the cultured cells examined, which were all dividing cells. However, when neuronally differentiated postmitotic PC12 cells were examined several days after transfection of MJD79, we noticed that ex-polyQ–mediated aggregates were made in approximately one in a few hundred of the transfected PC12 cells; these aggregates could be detected only with an antibody that recognized an amino acid stretch just downstream of the polyglutamine portion of the MJD protein, and not with an antibody to a tag attached to the N-terminus of the MJD protein (Yasuda et al., 1999).We then designed a "cleavage-dependent cell selection system" in which weak protein processing activity is converted into a strong transcriptional activity, thus allowing the expression of drug-resistant proteins required for selection. This system will provide a novel cell selection system for isolating cells containing processing activities for any protein concerned (Fig. 4A) (Yamamoto et al., 2001). The full-length MJD35 protein fused with Ga14VP16, a strong transcriptional activator, at the C- terminus and with the transmembrane region of the Fas receptor at the N- terminus, is expressed together with a Gal4 reporter plasmid, which drives the expression of a drug-resistant marker protein, in this case for Zeocin resistance. More specifically, this system should work in the following way: when the MJD protein, in this case MJD35, is cleaved by its processing enzyme, the cleaved portion with Gal4VP16 will translocate to the nucleus and transactivate the reporter gene, thereby resulting in Zeocin resistance in dividing pre-differentiated PC12 cells. The Zeocin-resistant cells obtained from this system should contain PC12 subclones with enhanced processing activity of the MJD protein. By repeating this selection process, cells with much higher processing activity for the MJD protein should be obtained. We used MJD35, which contains a normal length of polyglutamine repeats in the selection process, to avoid aggregate formation of the cleaved fragments and consequent cell death. We were indeed able to obtain several PC12 subclones after sequential selection with 100 µg/ml and 400 µg/ml Zeocin. Among these subclones, two representative subclones (B16 and C10) were analyzed by Western blots and were confirmed to have increased MJD processing activity (Fig. 4B) (Yamamoto et al., 2001). A rough estimation of the cleavage site of the MJD protein, based on the sizes of the cleaved products, suggests that the MJD protein is cleaved in these cells around the 250th amino acid residue (Yamamoto et al., 2001).This processing model appears to be applicable not only for MJD but also for several other polyQ diseases. Apparent processed fragments were found in the nuclear inclusion (NI) in patients' brain samples of HD, MJD, SBMA, and DRPLA as truncated proteins without the epitopes covering the entire proteins (DiFiglia et al., 1997; Li et al., 1998; Schmidt et al., 1998; Schilling et al., 1999). Recently, it has been reported that an aspartic protease inhibitor inhibits the HD protein cleavage in cultured cells, but the responsible protease still remains to be identified (DiFiglia, 2002; Lunkes et al., 2002).What is the significance of the processing in polyQ diseases? In vitro analysis using GST-fused ex-polyQ and several cellular models clearly demonstrated that polypeptides containing long polyQ aggregate easily when surrounding protein portions become short (Ikeda et al., 1996; Scherzinger et al., 1997). In addition, the truncated fragment of the MJD protein was shown in a cellular model to recruit full-length MJD proteins into insoluble aggregates (Perez et al., 1998). These results suggest that, by steric hindrance, surrounding portions of ex-polyQs inhibit the seed formation in a critical initial step of aggregate formation. In the case of the MJD protein, it possesses a higher-ordered structure in its N-terminal portion and a relatively free structure around the C- terminus where the polyQ stretch resides (unpublished observations), which fits well with the observation that the C-terminal portion of MJD tends to make an aggregate (Yasuda et al., 1999). Another report showed that the overexpression of polyQ-expanded full-length MJD proteins is capable of causing the formation of intranuclear inclusions and cell death, suggesting that a high concentration of the MJD protein may overcome the steric hindrance (Evert et al., 1999).Why does aggregate formation correlate with cell death in neural cells? To analyze cellular response to aggregate formation toward cell death, we constructed PC12 cell lines expressing flag-tagged polyQ (Q14) or ex-polyQ (Q79) under control of the tet-off promoter (Yasuda et al., 1999). These cells were named PC12-Q14 and PC12-Q79, respectively. PC12 cells are a well-characterized cell line with the ability to differentiate into postmitotic neuron-like cells by the simple addition of nerve growth factor (NGF). NGF was added when tetracycline was removed from the medium in this experiment. When Q79 was expressed by the tet-off, the PC12-Q79 cells died gradually, accomplishing differentiation; almost all cells became extinct 96–120 h after tet-off. We analyzed the subcellular localization of ex-polyQ by an immunocytochemical analysis using an anti-flag antibody (Fig. 3C) (Yasuda et al., 1999). Q79 was diffusely distributed in the cytoplasm 24 h after tet-off. At 48 h, most cells contained one or more cytoplasmic ex-polyQ aggregates, together with many small aggregates in the nucleus. At 72 h, large ex-polyQ aggregates were observed in the nucleus, as well as in the cytoplasm. Then the cells started dying, indicating that cytoplasmic and nuclear ex-polyQ aggregate formations are events that precede cell death. Such aggregate formation and cell death were not observed in PC12-Q14 cells.During this cell-death assay, we occasionally found colonies of cells that appeared to be resistant to ex-polyQ–induced cell death. These cells contained large ex-polyQ aggregates exclusively in the cytoplasm. We unsuccessfully attempted to clone these cells. Any subtle treatment, such as simple passage, caused cell death. These results suggest that nuclear aggregates may cause stronger cell death signals than cytoplasmic aggregates and may correlate with the evidence that NIs are more frequently observed than cytoplasmic aggregates in the affected brain regions of patients with MJD, SCA1, DRPLA, HD, and other polyglutamine diseases (Burright et al., 1995; Davies et al., 1997; Paulson et al., 1997; Shimohata et al., 2000).Using these cells, cell death-related factors were examined during the course of ex-polyQ–induced cell death. First, we examined the potential involvement of caspases in cell death. Contrary to our expectation, only marginal activation of caspase 3 was observed. In addition, several caspase inhibitors (e.g., DEVD, Z-VAD) were unable to inhibit cell death, although some apoptotic phenotypes, such as chromosomal condensation, were abolished. These results suggest that caspases were involved in the manifestation of certain apoptotic phenotypes, but not in the commitment of cell death. Next, we examined the activation of other apoptosis-related signal transduction factors and found that SEK1-JNK, stress-activated protein kinases (Sánchez et al., 1994), were clearly activated. Twenty-four hours after tet-off, the small population of cells showed small aggregates in the nucleus in which SEK1 was activated. In the cells with a large NI at 72 h, SEK1 was activated only in the place that was in agreement with the inclusion. Moreover, we found that c-Jun, the target protein of JNK, had been activated in the same place. Consistently, the dominant-negative SEK1 mutant inhibited ex-polyQ–induced cell death. These results suggest the possibility that the accumulation of ex-polyQ aggregates in the nucleus provides a platform for the death signal kinase cascade. Then, what activates SEK1? Our recent analysis found that apoptosis-regulating signal kinase 1 (ASK1), an MAPKKK (mitogen-activated protein kinase kinase kinase), is responsible for SEK1 activation in ex-polyQ–mediated cell death (Nishitoh et al., 2002).Several reports published at around the same time have shown that many cellular factors, including transcription factors, were incorporated into the ex-polyQ aggregates, thereby leading to the hypothesis that NIs may sequester important nuclear factors, which, in turn, alter gene expression and then induce cellular abnormality, most typically cell death. Indeed, it has been shown that ex-polyQ aggregates can trap transcriptional factors, such as the TATA-binding protein (TBP), CREB binding protein (CBP), p53, mSin3A, and TAFII130 (Shimohata et al., 2000). In addition, a different transcriptional repression mechanism has been proposed as being a potential pathological mechanism of polyQ disease, in which the histone acetyltransferase activity of coactivators, such as CBP/p300, are inhibited by direct interaction with diseased proteins (Steffan et al., 2001; Li et al., 2002).There is another argument, this one proposing that NIs do not cause cell death but, rather, function in a protective manner (Saudou et al., 1998). In the case of Alzheimer's disease, small soluble Aβ fibril, rather than aggregated Aβ fiber itself, has been argued to be toxic to the cell (Ellis and Pinheiro, 2002; Walsh et al., 2002). However, it is not yet known whether this scenario is applicable to ex-polyQ or other disease-related aggregates.Ex-polyQ NIs are highly ubiquitinated in the brains of MJD patients and in animal and cellular models of MJD. Furthermore, it has recently been reported that several forms of proteasome complexes, as well as Hsp70 and Hsp40 (Hdj1 and Hdj2), are co-immunostained on NIs in the brains of MJD patients (Chai et al., 1999a, b). Ex-polyQ aggregates are a common feature of polyQ disease. These lines of evidence have raised another possibility that ex-polyQ aggregates sequester cellular factors that are involved in protein quality control and inhibit their functions. In cultured cells, ex-polyQ aggregates have indeed been shown to decrease proteasome activity, which, in turn, is predicted to result in further promotion of ex-polyQ aggregates (Bence et al., 2001; Nishitoh et al., 2002). Similar weakened protein degradation has been also reported in other neurodegenerative disorders with cytoplasmic protein aggregates (Leroy et al., 1998; Lam et al., 2000; Shimura et al., 2000). These observations suggest a potential link between neuronal degeneration and dysfunction of protein degradation pathways, most likely via the ubiquitin-proteasome system.Since the accumulation of abnormal proteins definitely puts stress on cells, especially on non-dividing neurons, we assumed the existence of a cellular factor, namely, a sensor protein, which recognizes and monitors the accumulation of abnormal proteins in the cells. To identify such a potential sensor protein, we performed a pull-down experiment using GST-fused MJD79 protein as a representative of an abnormal protein. All mammalian cultured cells subsequently examined were found to contain a protein with a molecular weight of approximately 100 kDa that interacted with GST-MJD79 much more strongly than GST-MJD35 (Hirabayashi et al., 2001). This protein was then purified from HeLa cells and COS cells by affinity purification using GST-MJD79. Mass analysis and protein sequencing revealed that the 100-kDa proteins purified from both cells were identical and were a previously reported protein named p97/VCP, a member of the AAA ATPase family. Northern blot analysis on human RNA showed that p97/VCP is expressed not only throughout brain but all other tissues as well. p97/VCP is reportedly one of the most abundant intracellular proteins and has been shown to be involved in postmitotic membrane fusion processes in endoplasmic reticulum, the Golgi apparatus, and nuclear envelopes (Latterich et al., 1995; Rabouille et al., 1995; Kondo et al., 1997; Hetzer et al., 2001). In these processes, p97/ VCP differentially utilizes its partners, such as p47 and the Ufd1/Npl4 complex.Next, using deletion mutants of MJD79, we investigated which portion of MJD79 is recognized by p97/VCP. As expected, the ex-polyQ portion was found to be indispensable for binding with p97/VCP (Fig. 5A) (Hirabayashi et al., 2001). Moreover, deletion analysis of p97/VCP revealed that the portion containing amino acid residues 143–199, located near the N-terminus, is responsible for binding to MJD79 (Fig. 5B) (Hirabayashi et al., 2001). p97/VCP consists of three domains: N, D1, and D2 (Zhang et al., 2000). The D1 and D2 domains are ATPase domains, whereas the N domain is flexible and moves dynamically via nucleotide binding (Rouiller et al., 2000). Interestingly, the N domain is predicted to be the binding domain to its cofactor p47 and multi-ubiquitin (Rouiller et al., 2000). We then tried to determine whether p97/VCP acts as a chaperon toward ex-polyQs as observed by Hsp104, another member of the AAA family (Krobitsch and Lindquist, 2000; Kimura et al., 2001). Although we have not yet succeeded in using mammalian cells to obtain results supporting this idea, p97/VCP can indeed function to enhance relatively short polyQs to aggregate in yeast cells (Kimura et al., 2002; Y. Kimura, personal communication).p97/VCP was observed in normal condition throughout mammalian cells, such as neuronally differentiated PC12 cells. Interestingly, p97/VCP was found to co-localize not only with ex-polyQ aggregates but also with several other protein aggregates in neuronal cells. We observed p97/VCP immunostaining in ex-polyQ–mediated nuclear and cytoplasmic inclusions in cell culture models (Fig. 6A) (Hirabayashi et al., 2001). p97/VCP co-staining with nuclear inclusions was also confirmed in our transgenic rat model and in brain samples from HD patients (Fig. 6B) (Hirabayashi et al., 2001). Consistent with the idea that p97/VCP can recognize a broad range of abnormal proteins, p97/VCP was found to bind to α-synuclein and the C-terminal portion of β-cleaved APP containing Aβ (unpublished observations). In addition, p97/VCP was found to co-localize with Lewy bodies in brain samples of patients with sporadic Parkinson's disease or dementia with Lewy bodies (Fig. 6B) (Hirabayashi et al., 2001). p97/VCP also co-localized with aggresomes, which were induced at a perinuclear position by proteasome inhibitor treatments in PC12 cells (Fig. 6C) (Hirabayashi et al., 2001). It is notable that in these cells with aggresome or ex-polyQ aggregates, many large cytoplasmic vacuoles were concomitantly observed, followed by cell death.At the same time, we took another approach, this time using Drosophila genetics for identifying genes involved in ex-polyQ–mediated neurodegeneration. We first created a Drosophila model of polyQ diseases (Higashiyama et al., 2002). Transgenes encoding a flag epitope-tagged Q22, Q79, and Q92 were expressed in the compound eyes of Drosophila under the control of an eye-specific glass (GMR) promoter. Only pathogenic lengths of polyQs (Q79 and Q92) induced cell death, showing eye-degeneration phenotypes with loss of red pigmentation, loss of internal eye integrity, and severe degeneration of the photoreceptor neurons. The degenerative phenotypes were dependent on ex-polyQ lengths and transgene dosages and were progressive (Fig. 7) (Higashiyama et al., 2002).We then performed genetic screening to search for dominant suppressors and enhancers of the eye degeneration. Q79 flies were crossed with about 180 fly lines from a stock that had chromosomal deficiencies. Our survey revealed that one chromosomally deficient fly line, Df(2R)X1, acts as a dominant suppressor. To identify the gene responsible for this suppression, available P-element-inserted mutants located within the deleted region were tested, and two weak dominant suppressors were found, the independent P-element-inserted mutants I(2)k15502 and I(2)03775, both having slightly suppressed eye abnormality (Fig. 8A) (Higashiyama et al., 2002). Then, using a plasmid rescue experiment, the locations of both P-elements were identified to be within the 5′-noncoding region of the ter94 gene. Surprisingly, Ter94 is the Drosophila homolog of p97/VCP. To determine if the ter94 mutation is responsible for the recovery of ex-polyQ-induced eye degeneration, we crossed GMR-Q92 flies with several ter94 mutants, such as the reportedly strong (ter9426-8) and weak (ter9422-26) loss of function mutants (Ruden et al., 2000). In flies having ter9426-8 in one allele (ter9426-8/+), eye degeneration was dramatically suppressed (Fig. 8A) (Higashiyama et al., 2002). In flies having ter9422-26 (ter9422-26/+), degeneration was also suppressed, but much less so than in the observed ter9426-8/+ flies. These results indicate that the ter94 protein plays a necessary function in eye degeneration, namely, in ex-polyQ–induced cell death.Examination of ex-polyQ expression revealed that ex-polyQ aggregates were formed in the nucleus of the eye disc of third instar larvae and in the compound eye in GMR-Q92 fly pupae. Similar nuclear aggregates were also formed in ter9426-8/+ flies (Higashiyama et al., 2002) (Fig. 8B). The total amounts and sizes of ex-po