SCA1 to 3, 6, and 7 are caused by the expansion of an unstable trinucleotide CAG repeat that encodes a tract of polyglutamines in the mutant protein. Even though there are no sequence similarities among the proteins encoded by these SCA genes, the diseases are thought to share a common pathogenesis at the protein level. Thus, models to explain the pathogenic mechanism behind each disease focus on the protein encoded by the disease-causing allele and, in particular, on the expanded polyglutamine tract within the protein. The models all posit that the mutation confers a toxic “gain-of-function”; that is, disease develops because of the expression of a mutant form of the protein, not because the protein loses its normal function.
Studies on SCA Patient Material
One of the most important pieces of information regarding the mechanism of pathogenesis for the SCAs is that each of these disorders is inherited as a dominant disease. Several pathogenic mechanisms may result in a dominant disorder; one common mechanism is the expression of a mutant protein (rather than loss of function of the normal protein, as occurs in haploinsufficiency). In the case of SCA1, this point is illustrated by the fact that disease develops only as a consequence of the expansion of a polyglutamine tract within the SCA1 gene on chromosome 6p23.106 A heterozygous deletion of the SCA1 gene does not result in SCA1, but may contribute to mental retardation and seizures in 6p deletion syndromes.152,153
A second important observation from patient studies is that for each of the polyglutamine-based SCAs examined, the mutant protein is detected in patient material and varies in size according to the length of the CAG repeat.97,117,126,154,155 This clearly demonstrates that expansion of the CAG tract causes an increased number of glutamine residues. Importantly, it is also consistent with a model of pathogenesis based on a toxic gain-of-function by the protein with an expanded number of glutamines.
Immunohistochemical studies using patient material have been performed to determine the subcellular localization of both wild-type and mutant proteins encoded by the SCA mutant alleles. For SCA1, both the normal and mutant ataxin-1 are localized to neuronal nuclei throughout the brain in normal as well as affected individuals.97 In Purkinje cells, a primary cellular site of SCA1 pathology, cytoplasmic protein was detected to a lesser extent. Peripheral tissues expressing ataxin-1 have a cytoplasmic localization. The protein encoded by the SCA7 gene is found predominantly within the nucleus in all cell types examined.117 In contrast, ataxin-2, ataxin-3, and atrophin-1 appear to normally be predominantly cytoplasmic.126,146,154,155 SCA6 encodes the α1A-voltage-dependent calcium channel, which has four transmembrane domains, and thus localizes to the plasma membrane.61
One aspect of the subcellular distribution for four of the SCA polyglutamine proteins is the localization of mutant proteins to ubiquitinated microscopic nuclear inclusions or aggregates within neuronal nuclei (Figs. 226-3 and 226-4). Nuclear inclusions containing the polyglutamine protein have been detected in brains of SCA1,112 MJD/SCA3,156 SCA7,139 and DRPLA 157 patients. In some diseases, nuclear inclusions are found only in neurons in regions of the brain affected by the disease, for example, SCA1, MJD/SCA3, and DRPLA. Interestingly, in SCA7 patient material, ataxin-7-containing nuclear inclusions were not restricted to sites of severe neuronal loss associated with disease. Nuclear inclusions are not seen in SCA2 brain samples, but there is a clear increase in ataxin-2 immunostaining in Purkinje cells and dentate neurons, suggesting that there might be cytoplasmic accumulation of the mutant protein.155 The dramatic presence of nuclear inclusions in brain material from SCA patients has led to the suggestion that protein aggregation is a critical molecular component of disease.152,158 Furthermore, the presence of ubiquitin, a molecule that normally modifies and tags proteins targeted for proteolysis,159 in the nuclear inclusions suggests that the formation of these aggregates is due to protein misfolding, and this misfolding may alter cellular degradation of the polyglutamine protein and possibly other cellular proteins.
To ascertain directly whether nuclear aggregates of a polyglutamine protein altered the cellular distribution of proteasomes, brain tissue from an SCA1 patient was examined immunohistochemically.114 In brain regions affected by SCA1, ataxin-1 nuclear aggregates were intensely stained by antisera to the 20S subunit of the proteasome. Because alterations in proteasome function are usually associated with increased expression of the stress-response or heat-shock chaperone proteins, the subcellular distribution of the Hsp40 and 70-kDa heat-shock proteins, respectively (Hsp70) families of chaperones were also assessed in an SCA1 brain. Ataxin-1 nuclear aggregates in neurons in the nucleus points centralis, a region affected in SCA1, were found to be positive for Hsp40, HDJ-2/HSDJ. In control human brain, Hsp40 immunohistochemical staining was cytoplasmic. Only minimal immunostaining for the constitutive form of Hsp70 was detectable. The localization of Hsp40 to the nuclear ataxin-1 aggregates supports the hypothesis that protein misfolding is responsible for the presence of the ataxin-1 aggregates in SCA1 patient brains. Furthermore, the ubiquitination of the aggregates and the association of the 20S proteasomes with the nuclear aggregates imply that nuclear proteasome-dependent protein degradation may be altered in SCA1.114 These observations regarding SCA1 aggregates have recently been corroborated by SCA3 studies in which a fraction of the SCA3 inclusions stained positively with antibodies corresponding to Hsp40, Hsp70, and the 20S proteasome.160 One important conclusion to be drawn from analyses of the subcellular distribution of the polyglutamine proteins in SCA patient material is that the nucleus is the primary site of pathogenesis in most of these diseases. Only in SCA2 and SCA6 does the pathogenesis seem to occur primarily in the cytoplasm.
Cellular Models of SCA Pathogenesis
Examining polyglutamine proteins in cultured cell lines is proving to be an important strategy for assessing the behavior and properties of SCA mutant proteins. Analysis of the subcellular pattern of ataxin-1 expression upon transfection of the SCA1 gene into COS cells also revealed a predominately nuclear localization.112 Ataxin-1 with an expanded number of glutamines formed nuclear aggregates that were larger than those aggregates seen in COS cells transfected with a wild-type allele of SCA1. In transfected kidney epithelial 293T cells, normal ataxin-3 with 27 repeats, as well as expanded ataxin-3 with 78 repeats, distributed diffusely throughout the cytoplasm.156 Interestingly, transfected 293T cells expressing a C-terminal fragment of ataxin-3 containing the glutamine repeat localized to large cytoplasmic and occasionally nuclear aggregates. Furthermore, the expression of a truncated form of ataxin-3 with an expanded number of glutamines induced cell death in 293T cells. COS-7 cells expressing truncated forms of atrophin-1 with expanded polyglutamine stretches had filamentous peri- and intranuclear aggregates and eventually underwent apoptosis.157 It is interesting that the cell death was partially suppressed by the addition of transglutaminase inhibitors, suggesting that the transglutaminase reaction may be involved in atrophin-1 aggregation and cell death.
Tait et al. found that full-length ataxin-3 localized primarily to the nucleus in transfected COS-7 and neuroblastoma cells.127 These investigators also suggested that ataxin-3 associated with the nuclear matrix of COS-7 cells. Ataxin-1 has also been found associated with the nuclear matrix in transfected COS-1 cells.112 In the case of ataxin-1, colocalization studies showed that mutant ataxin-1 causes a specific redistribution of the nuclear matrix-associated domain, promyelocytic oncogenic domain, containing the promyelocytic leukemia protein in transfected COS-1 cells. These data for ataxin-1 and ataxin-3 further support the concept that expression of a mutant SCA protein perturbs the ability of the nucleus to function properly.
Using transfected HeLa cells, Cummings et al. demonstrated that overexpression of the HDJ-2/HSDJ chaperone decreased the size and frequency of mutant ataxin-1 nuclear aggregates.114 This result strongly indicated that misfolding of mutant ataxin-1 is involved in aggregation and that the cellular levels of a chaperone may directly regulate the propensity of an expanded polyglutamine protein to aggregate. Similarly, overexpression of HDJ-2/HSDJ in cells transfected with mutant ataxin-3 reduced the frequency of aggregate formation.160 Thus, the studies performed to date on cultured cells expressing mutant forms of ataxin-1 and ataxin-3 demonstrate that SCA pathogenesis involves misfolding of the mutant protein in the nucleus that could then alter proteolytic processing of the mutant protein, thereby disrupting nuclear function. Central to this model are the ubiquitinated nuclear aggregates of the mutant SCA proteins. However, the precise role of these nuclear aggregates in pathogenesis is unclear from either the patient or cultured cell studies.
A key aspect of SCA pathogenesis that must be addressed by any model is the selective pathology seen in only a subset of neurons despite the broad expression pattern of each gene. One explanation for this might be that cell-specific proteins mediate pathogenesis through an interaction with each disease protein. For SCA1, the yeast two-hybrid approach has identified proteins that interact with ataxin-1.161–163 Yeast expressing either a wild-type or mutant allele of ataxin-1 were transformed with human and mouse brain cDNA libraries. From the yeast two-hybrid screen using an adult human brain cDNA library, over 100 clones that encoded a protein capable of interacting equally well with wild-type and mutant ataxin-1 were identified.161 Of these, a substantial number were found to encode ataxin-1. Thus, this approach led to the conclusion that the ataxin-1 protein is able to form multimers with itself. Further analysis of deletion mutants of ataxin-1 in the yeast two-hybrid system mapped this self-association region to a stretch between amino acids 495 and 605, beyond the polyglutamine stretch.
In a second yeast two-hybrid screen using a mouse brain cDNA library, the leucine-rich acidic nuclear protein (LANP) was identified as an ataxin-1 interacting protein.162 Previously, in situ hybridization studies demonstrated expression of murine Lanp mRNA in cerebellar Purkinje cells and granule neurons, and that the highest level of Lanp expression occurs in Purkinje cells around postnatal day 14,164 the same time of the transient burst of Sca1 expression in the mouse.111 LANP has been previously isolated from peripheral tissues using a variety of independent strategies. Vaesen et al. purified it as a putative HLA DR-associated protein I (PHAPI) from lymphoblastoid cells based on its association with a peptide fragment from HLA-DR2.165 The same protein was also isolated as phosphoprotein 32 (pp32) and found to inhibit the formation of transformed foci in rat embryo fibroblasts when cotransfected with a variety of oncogenes.166 Li et al. suggested that LANP is a phosphatase 2A inhibitor.167 More recently, mapmodulin, a protein proposed to modulate the binding of microtubule-associated proteins to microtubules, was shown to have the identical peptide sequence as PHAPI and LANP.168
Immunohistochemical staining for LANP in mouse cerebellar cortex confirmed its predominant expression by Purkinje cells and localized it to the nuclei of these cells.162 Given the very high level of LANP expression within the nuclei of cerebellar Purkinje cells, this protein must have some role in Purkinje cell function that is probably distinct from some of the cytoplasmic functions proposed in peripheral dividing cells. Immunofluorescence studies revealed that in transfected COS cells, LANP colocalized to the nuclear aggregates containing either wild-type or mutant ataxin-1.162
The yeast two-hybrid approach was used to define the regions of ataxin-1 that interact with LANP and to determine whether the strength of this interaction is influenced by the number of glutamines within ataxin-1.162 From these studies, it appears that ataxin-1 interacts with LANP through the polyglutamine tract and via sequences located between residues 570 to 816. Furthermore, the strength of interaction between LANP and ataxin-1 containing 82 glutamines (82Q) was significantly higher than with ataxin-1 containing 30 glutamines (30Q), suggesting that LANP may also be involved in SCA1 pathogenesis. The temporal and subcellular pattern of LANP and ataxin-1 expression, combined with the cellular distribution of LANP expression, argue strongly that this interaction is an important component of the selective neuronal degeneration in SCA1.
An interaction between the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was also demonstrated using the yeast two-hybrid approach.163 Burke et. al. demonstrated that GAPDH interacts with huntingtin and atrophin-1.169 Because GAPDH is a regulatory enzyme in glycolysis, it is tempting to propose that the binding of mutant ataxin-1 or atrophin-1 to GAPDH may have deleterious effects on its function, decreasing ATP production, and subsequently leading to neuronal degeneration.169 However, several important considerations render this hypothesis less likely. First, the cellular levels of GAPDH are much higher than the cellular levels of any of the polyglutamine proteins. Thus, it is unlikely that a stoichiometric interaction between SCA proteins and GAPDH would impair the function of GAPDH to a physiologically relevant degree. In addition, GAPDH, a regulatory enzyme in the glycolytic conversion of glucose to pyruvic acid, is widely expressed throughout the brain. Yet, each SCA is characterized by a unique cellular pathologic pattern. It is not clear how the interaction of a polyglutamine protein with the ubiquitously expressed GAPDH would affect only a specific subset of neurons. Therefore, it seems that if the interaction of ataxin-1 with GAPDH is involved in SCA1 pathogenesis, it is most likely one component of a complex pathway that occurs well after pathogenesis is initiated.
Animal Models of the SCAs
The first transgenic mouse model of a SCA utilized the strong Purkinje cell-specific regulatory elements from the Purkinje cell protein 2 (Pcp2/L7) gene to direct expression of the human SCA1 cDNA transgenes encoding the entire ataxin-1 protein.170 Transgenic lines were established expressing high levels of either a wild-type allele of SCA1 with 30 repeats or an expanded allele with 82 repeats. Only animals from transgenic lines expressing the expanded 82 CAG repeat SCA1 transgene developed severe ataxia and pathologic changes within their cerebellar Purkinje cells. These studies demonstrated that pathologic changes are induced specifically by the expression of an intact, expanded SCA1 allele. In contrast, although they manifested hippocampal and cerebellar learning difficulties, mice homozygous for a null mutation within their Sca1 gene failed to develop the ataxia and Purkinje cell pathology seen in transgenic mice expressing a mutant SCA1 allele.115 This clearly demonstrated that SCA1 pathogenesis requires the expression a mutant allele of SCA1 and does not ensue with the loss of ataxin-1 function.
In SCA1 transgenic mice, normal ataxin-1 localized to several ≈0.5 μM nuclear aggregates. Expanded ataxin-1, on the other hand, localized to a single ≈2 μM ubiquitinated nuclear aggregate, as it does in patient material.112 These aggregates also stained positive for the 20S proteasome and the HDJ-2/HSDJ (Hsp40) chaperone protein.114 The appearance of single nuclear aggregates in SCA1 transgenic mice carrying the expanded ataxin-1 allele preceded the onset of ataxia by approximately 6 weeks. Thus, the SCA1 transgenic mice manifest the cellular pathologic features found in patients.
To gain insight into disease progression in the SCA1 transgenic mice, animals were examined at various ages for neurologic alterations and morphologic changes at the cellular level.171 SCA1 transgenic animals expressing an expanded allele with 82 repeats progressively lost cerebellar function. Cerebellar impairment in 5-week-old mice was limited to decreased ability to improve motor performance over 4 days of successive trials on a rotating rod; the mice showed no gait abnormalities, and were normal in motor activity, balance, and coordination. But by 12 weeks, this slight motor skill impairment became full-blown ataxia.
No histologic alterations were detectable during the first 3 weeks of postnatal life in the SCA1 mice. At the time of impaired motor learning on the rotating rod apparatus (5 to 6 weeks of age), the Purkinje cells showed cytoplasmic vacuoles, a loss of proximal dendritic branches, and a decrease in the number of dendritic spines, indicating that mutant ataxin-1 may impair the maintenance of dendritic arborization. By the time ataxia was severe (12 to 15 weeks), there were numerous morphologic abnormalities in the Purkinje cells, but little evidence of cell loss. Expression of full-length mutant ataxin-1 in cells vulnerable to the human disease evidently can disrupt cellular function enough to cause ataxia without killing the affected neuronal population. This is an important finding, as the neurologic phenotype in SCA patients had long been assumed to result from apoptosis. The Purkinje cell loss that does occur in late-stage disease most likely results from the dysfunction already begun in earlier stages.
One striking neuropathologic feature of SCA1 transgenic mice is the movement of the Purkinje cell bodies away from their typical location at the interface of the molecular and granule cell layer into the intermediate level of the molecular layer. This heterotopia is not due to a developmental abnormality: histologic examination of young animals revealed no heterotopia in the early stages of the disease, but it is frequent by the time of overt ataxia.
A more likely explanation for the occurrence of heterotopic Purkinje cells is based on the appearance of the dendritic changes earlier in the course of the disease. Postmortem studies of the cerebellum in patients with SCA1 have shown structural abnormalities of Purkinje cells as well as cell loss. Using Golgi techniques and immunohistochemical methods, Ferrer et al.172 and Koeppen et al.173 described dendritic simplification with loss of spines similar to that found in the SCA1 transgenic mice. A hallmark of SCA1 pathology is the occurrence of frequent proximal axonal dilatations (torpedoes), a feature not found in the SCA1 mice. It is clear from autopsy studies of SCA1 patients that morphologic alterations antedate cell death in at least some of the Purkinje cells of these patients.
Having demonstrated the characteristics of disease in transgenic mouse models of SCA1,170,171 subsequent studies examined the molecular and cellular aspects of ataxin-1 that are critical for pathogenesis. The first series of experiments addressed the importance of the subcellular localization of ataxin-1. Histologic analyses revealed that ataxin-1 is predominantly a nuclear protein.97 Typically, the cellular transport machinery recognizes a target protein by its nuclear localization signal (NLS), and then moves the protein from the cytoplasm to the nucleus through pores in the nuclear envelope.174 Although NLSs can be quite variable in appearance, many of these peptide motifs conform to the consensus cluster of arginines and lysines found in SV40 T antigen.175 Two possible NLSs of this type were found in the ataxin-1 amino acid sequence: one near the N-terminus at lysine 15, and one near the C-terminus at lysine 772.176 Analysis of the role of these putative NLSs in ataxin-1 nuclear transport was carried out in transfected COS cells. The C-terminal motif at lysine 772, but not the N-terminal element, is, indeed, a functional NLS.
Klement et al. studied transgenic mice expressing ataxin-1(82Q) with a mutated nuclear localization sequence, ataxin-1K772T. These mice failed to develop Purkinje cell pathology and associated motor dysfunction, and no ataxin-1 aggregates formed in the cytoplasm. Thus, nuclear localization is critical for pathogenesis and ataxin-1 aggregation. Furthermore, these data demonstrate that the mutation exerts its toxic effects at the protein, not the RNA, level.176 In a second series of transgenic mice, ataxin-1 (77Q) containing a deletion of amino acids within the self-association region was expressed within the nuclei of Purkinje cells.176 These mice developed ataxia and Purkinje cell pathology similar to the original SCA1 mice—with no evidence of nuclear ataxin-1 aggregation. Thus, although nuclear localization of ataxin-1 is necessary, nuclear aggregation of ataxin-1 is not required to initiate pathogenesis in transgenic mice.
The best available model of SCA1 pathogenesis proposes that sequences within the ataxin-1 protein, in addition to the polyglutamine tract, are critical in specifying the site and course of disease. The deleterious effects leading to SCA1 symptoms are initiated by the localization of ataxin-1 to the nucleus. Once there, the mutant form of ataxin-1 misfolds and is altered in its distribution. The nuclear distribution of ataxin-1 may depend on residues within the self-association region as well as the length of the polyglutamine tract, because deletion of ataxin-1's self-association domain compromises its ability to form nuclear aggregates in COS cells despite its nuclear localization. In the nucleus, ataxin-1 interacts with LANP and probably other proteins. It is likely that changes in nuclear architecture and the interaction of mutant ataxin-1 with other nuclear proteins alter gene expression, all of which may well lead to neuronal dysfunction, symptomatology, and, eventually, neuronal loss. Thus, while the expanded polyglutamine tract is clearly critical for pathogenesis, additional segments of ataxin-1 are just as significant. The extent to which this model of SCA1 pathogenesis may be applicable to other SCAs and polyglutamine disorders remains to be seen.
Ikeda and colleagues established transgenic mice expressing full-length and truncated versions of the MJD/SCA3 protein also using the Pcp-2/L7 promoter region.177 In independent transgenic lines, they obtained expression of MJD1 cDNA transgenes encoding full-length and truncated proteins with 79 glutamine residues (designated MJD79 and Q79C, respectively, by the authors) as well as a truncated protein with 35 glutamine residues (Q35C). The Q35C and Q79C proteins contained a fragment comprising the polyglutamine tract and 42 residues C-terminal to glutamine repeat (hence the C designation). In addition, transgenic animals expressing only a 79-glutamine residue tract (Q79) were produced using the Pcp-2/L7 promoter region. Animals from both the Q79C and Q79 lines developed ataxia rapidly, by 4 weeks of age. None of the animals containing the full-length MJD79 or the truncated Q35C constructs was ataxic by 23 or 32 weeks of age, respectively.
Histologic examination of an 8-week-old Q79C transgenic animal revealed that although the cerebellum maintained its fundamental structure, it occupied only about 15 percent of its normal volume. All three layers of the cerebellum were affected. The molecular layer showed substantial thinning; the Purkinje cells were shrunken, had attenuated dendrites, and showed reduced calbindin immunoreactivity; the granule cell layer showed a significant reduction in cell number, with many of the existing cells displaying an altered, shrunken morphology.
The expanded polyglutamine tract can thus, on its own, induce Purkinje cell death and ataxia. However, in contrast to the SCA1 transgenic animals, the SCA3 transgenic mice developed ataxia as a direct result of Purkinje cell loss. Ikeda and colleagues proposed that a truncated form of MJD1—and perhaps other polyglutamine repeat-containing proteins—is more potent at inducing cell death than its full-length counterpart. Furthermore, they hypothesized that a cell-specific proteolytic cleavage of the mutant protein exposes a subset of cells to the detrimental effects of an elongated polyglutamine tract. The finding that ataxin-3 can be cleaved by a caspase,178 a member of a family of cysteine proteases involved in apoptosis, is consistent with the possibility that ataxin-3 exerts its toxicity after cleavage and release of a polyglutamine tract. Further support for this model was provided by the report that a C-terminal fragment of ataxin-3 containing an expanded polyglutamine tract formed cytoplasmic and nuclear aggregates and was cytotoxic in transfected 293T cells.156 However, to date there is no direct evidence that cleavage of ataxin-3 is critical for MJD/SCA3 pathogenesis.