Drosophila Models of Neurodegenerative Diseases

Pathological Features and Genes Involved in the Pathogenesis of Alzheimer’s Disease and Tauopathy

AD is the most common neurodegenerative disease and is the leading cause of dementia worldwide. Clinically, AD manifests as a gradual decline of cognitive functions such as learning and memory. Pathologically, the disease is characterized by selective atrophy of the hippocampus and the frontal cerebral cortex. Two microscopic brain lesions, neuritic or amyloid plaque and neurofibrillary tangles (NFT), are diagnostic of AD in the demented patient when present in abnormally high densities for a given age. The main components of the amyloid plaque are the Aβ-40 and Aβ-42 peptides, which are generated by endoproteolysis of the amyloid precursor protein (APP) via the sequential action of β- and γ-secretases. β-secretase activity is provided by the β-site APP-cleaving enzyme (BACE), whereas γ-secretase activity depends on a protein complex consisting of presenilin (Psn), nicastrin, aph-1, and pen-2 (15). Autosomal dominant mutations in APP, Psn-1, and Psn-2 can accelerate disease onset and progression in familial AD (FAD) cases. At the molecular level, these mutations promote the generation of amyloidogenic Aβ peptides (16). Also, defective trafficking of APP into recycling pathways may underlie the pathological accumulation of Aβ in some late-onset FAD cases (17). Taken together, these results have all but confirmed the amyloid hypothesis, which postulates that the initiating event in AD is the accumulation of Aβ peptide that may underlie synaptic failure, thereby resulting in the remarkably pure impairment of cognitive function. Excesses in Aβ accumulation trigger neuronal dysfunction, cell death, and irreversible dementia (18).

The major component of the NFT is tau protein, a highly soluble, microtubule-binding protein normally enriched in axons (19). In AD it becomes abnormally phosphorylated and is insoluble in NFT (20–22). This observation has provided circumstantial evidence that tau abnormalities may lead to neuronal dysfunction and degeneration. Direct evidence of the pathogenicity of tau was provided by the discovery that mutations in tau are associated with frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17) (23–26). NFT is also a pathological hallmark of AD, where it is detectable in the abnormal neuritic processes associated with amyloid plaques (27, 28).

The mechanistic relationship between NFT and amyloid plaque in disease pathogenesis remains a matter of some controversy (18, 29). Studies in transgenic mice have shown that expression of mutant human tau (h-tau) alone can result in tangle pathology and neuronal loss without accompanying plaque pathology (30, 31). Induction of amyloid deposits in these animals can further promote NFT formation (32–34). These observations suggest that the two lesions may be causally linked, although the molecular mechanism involved is not understood. The notion that amyloid plaque and tau NFT pathologies are causally associated is further supported by recent studies demonstrating that blocking Aβ peptide production in transgenic mouse models could remove amyloid pathology as well as early-stage tau lesions (35). Together, these studies appear to lend further support to the amyloid hypothesis. Note that NFT pathology is also evident in non-AD dementias as well as in a wide variety of other conditions that do not include any amyloid plaques, suggesting that events other than Aβ peptide accumulation can also trigger the formation of NFT (36). The sequence of molecular events leading to NFT formation is poorly understood.

Modeling Alzheimer’s Disease in Drosophila

Most of the genes implicated in AD pathogenesis have clear fly homologs. There is a fly homolog of APP (known as APP-like or APPL). Flies deficient for APPL exhibit a behavioral abnormality that can be rescued by a human APP transgene, indicating functional conservation between APPL and human APP (37). However, APPL lacks the amyloidogenic Aβ peptide sequence at the C terminus, and there is no evidence that it is processed in vivo as is human APP. The γ-secretase complex components are conserved in Drosophila and are known to be involved in the processing of Notch (38, 39). Although the fly does not appear to contain a clear homolog of mammalian BACE, it is possible that other proteases may provide a similar function, as it has been reported that overexpression of human APP causes axonal transport defects and increases cell death in the larval brain in an Aβ-dependent manner (40). Furthermore, one study showed that Aβ peptide can be generated in Drosophila from a modified human APP transgene (41).

Fly models of Aβ peptide–induced amyloid formation and neurodegeneration were recently generated. In one study (42), a human BACE transgene was coexpressed with a human APP transgene and a fly Psn transgene, both of which contained FAD-linked mutations. Transgenic flies expressing all three transgenes developed β-amyloid plaques and age-dependent neurodegeneration. Genetic and pharmacological manipulation of γ-secretase activity can modulate the severity of the disease phenotype in this model (42). In studies employing another approach (43, 44), Aβ-40 and Aβ-42 peptides were directly fused to signal peptides and expressed as transgenes. Transgenic flies expressing Aβ-42 developed (a) diffuse amyloid deposits, (b) age-dependent learning defects, detected using a Pavlovian olfactory learning assay, and (c) neurodegeneration, indicated by neuronal loss and vacuole formation in various brain regions. Aβ-40-expressing flies developed only age-dependent learning defects but no amyloid formation or neurodegeneration. These studies thus emphasize the importance of Aβ-42-associated toxicity in disease pathogenesis. The development of these Aβ-toxicity models should pave the way for future studies aimed at understanding the molecular events in the pathogenic cascade of AD and at screening for drug targets and potential therapeutics.

Modeling Tauopathy in Drosophila

Tau-induced neurodegeneration has also been successfully modeled in Drosophila. The first evidence of tau-induced neurotoxicity in flies was provided by transgenic studies of a bovine tau–green fluorescent protein reporter that caused axonal degeneration in sensory neurons (45). Subsequently, fly models of tauopathy were created by expressing wild-type or FTDP-17-linked mutant forms of human tau (46, 47). The fly genome includes a fly homolog of tau (known as dtau). Overexpression studies showed that directed expression of wild-type vertebrate tau or dtau in adult mushroom body neurons, which control olfactory learning and memory in Drosophila, strongly compromised associative olfactory learning and memory in the absence of detectable neurodegeneration (48). This result suggests that behavioral plasticity decrements, possibly reflecting synaptic dysfunction, may represent the earliest neuropathological manifestation of tauopathy.

Phosphorylation Control of Tau Toxicity

Tau phosphorylation and dephosphorylation have attracted considerable interest in tauopathy research, fueled by the detection of abnormally phosphorylated tau in NFT. A number of protein kinases and phosphatases have been found to regulate tau phosphorylation in vitro. These include glycogen synthase kinase (GSK)-3β, mitogen-activated protein kinase (MAPK), Cdk2/5, protein kinases A and B (PKA and PKB, respectively), calmodulin-dependent protein kinase II (CaMKII), microtubule affinity–regulating kinase (MARK), and protein phosphatases 1, 2A, 2B, and 2C (49). Despite a large body of biochemical studies, surprisingly little is known about the in vivo roles of these enzymes in regulating tau accumulation and presumed toxicity. The fact that many of these genes play essential roles during development, as well as the existence of considerable genetic redundancy, may have impeded in vivo genetic analyses of these genes in tau phosphorylation and toxicity in mammalian systems.

The compactness of the genome and the existence of an arsenal of molecular genetic tools in Drosophila make it feasible to evaluate the importance of phosphorylation in tau-mediated neurotoxicity. Through loss-of-function and overexpression genetic studies and biochemical analysis, partitioning defective–1 (PAR-1) was found to be a physiological tau kinase that plays a central role in determining tau phosphorylation and toxicity in Drosophila (50). PAR-1 is a serine/threonine kinase originally identified in Caenorhabditis elegans for its role in regulating cell polarity and asymmetric cell division (51). PAR-1 homologs have been found in eukaryotes ranging from yeast to mammals, where they possess essential cellular and developmental functions. MARKs, the mammalian homologs of PAR-1, regulate microtubule dynamics, epithelial cell polarity, and neuronal differentiation under normal conditions and are bound to NFT in AD brain (52–54). In the fly tauopathy model, overexpression of PAR-1 leads to elevated tau phosphorylation at S262 and S356 sites and to enhanced toxicity, whereas removing PAR-1 function or mutating PAR-1 phosphorylation sites in tau abolishes tau toxicity (50). Importantly, PAR-1 has been found to initiate a multisite phosphorylation process that generates pathogenic tau. In this process, phosphorylation by PAR-1 at the S262/S356 sites precedes and is obligatory for downstream phosphorylation events, including the T208 site that is phosphorylated by GSK-3β and Cdk5 (50). These results, together with the findings from a large-scale genetic screen that uncovered protein kinases and phosphatases as major modifiers of tau toxicity (55), clearly established the importance of phosphorylation in determining tau toxicity in vivo.

Although the presence of protein aggregates is a common feature of neurodegenerative diseases, these aggregates’ role in disease remains uncertain. However, studies of fly tauopathy models have shed some light on the role of pathogenic protein aggregation. In some disease models, there is a correlation between the extent of macroscopically visible protein aggregate formation and disease severity, whereas in others, disease symptoms can occur without formation of large protein aggregates. In certain cases, genetic manipulations that interfere with protein aggregate formation actually exacerbate disease symptoms (56). In fly tauopathy models, neurodegeneration can occur in the absence of NFT formation, suggesting that some soluble, nontangled form of phosphorylated tau is intrinsically toxic (46). Interestingly, when fly GSK-3β is coexpressed with tau, large NFT-like aggregates form in photoreceptor neurons, and neurodegeneration is enhanced (47). Although this observation suggests that NFT-like aggregates may contribute to tau toxicity, the evidence is also compatible with the idea that some soluble, nontangled forms of phosphorylated tau are the primary neurotoxic species and that these species also accumulate under GSK-3β-overexpression conditions. Further studies in Drosophila and vertebrate models are needed to elucidate the relationship between NFT formation and tau toxicity.

Postmortem analysis indicates that in the AD patient brain tau is hyperphosphorylated in more than 20 serine/threonine sites (57). The importance of individual phosphorylation sites in determining tau toxicity in vivo has not been clearly defined. In theory, these phosphorylation sites could act in concert to confer tau toxicity. Alternatively, a small subset of sites may play a dominant role in determining tau toxicity. Analysis of the PAR-1-triggered sequential tau phosphorylation has emphasized the importance of certain phosphorylation events in determining tau toxicity (50). The relative ease of making transgenic flies express mutant tau proteins enables us to pinpoint the key phosphorylation events that play critical roles in conferring tau toxicity. Such information will help identify the neurotoxic tau species and will reveal critical steps in the pathogenic cascade of tauopathy.

Signaling Mechanisms That Link Amyloid and Tau Lesions

The upstream signaling mechanisms that regulate tau phosphorylation under normal and disease conditions are not well understood. The identification of PAR-1 as a physiological tau kinase enables dissection of the signaling mechanisms. Recent studies show that phosphorylation of PAR-1 by the tumor-suppressor protein LKB1 is required for its activation; this, in turn, promotes tau phosphorylation in Drosophila. Diverse stress stimuli, such as high osmolarity and human APP–induced neurotoxicity (Figure 1), can promote PAR-1 activation and tau phosphorylation in an LKB1-dependent manner (58). These results reveal a previously unknown function for the tumor-suppressor protein LKB1 in a signaling cascade, through which the phosphorylation and function of tau are regulated by diverse AD-related toxic stimuli. Hyperphosphorylation of tau at the PAR-1 target site has also been observed in human AD patient brain and in APP/tau double-transgenic mouse AD models (59, 60), suggesting that aberrant activation of PAR-1-like kinases occurs in AD conditions in mammals. Together, these studies implicate the LKB1/PAR-1/tau-phosphorylation cascade as a missing molecular link between amyloid and tau lesions in AD pathogenesis (Figure 2). Because each kinase is likely to have more than one substrate, further studies are needed to identify additional substrates of LKB1 and PAR-1 and to assess their contribution to AD-related neurodegeneration.

Role of Synaptic Dysfunction in Disease Pathogenesis

Prior to the onset of plaque and tangle pathology and frank neuronal death in susceptible AD brain regions, mild cognitive impairment is observed. Subsequent studies have confirmed that loss of synaptic function is one of the earliest signs of the disease process in AD. This has led to the proposal that AD is primarily a synaptic failure (61). Studies in animal models have supported the role of synaptic dysfunction in disease pathogenesis. In both vertebrates and invertebrates, APP and APP-like proteins have been shown to exert normal physiological function at the synapse (37, 62). The mouse APP-transgenic model exhibits learning and memory deficits in the absence of obvious neuronal degeneration, supporting a primary role for synaptic dysfunction in APP pathogenesis (63). Interestingly, removal of endogenous tau can suppress most of the cognitive deficits in the APP model (64), indicating a critical role for tau in mediating the toxic effects of APP at the synapse. This is consistent with findings that synaptic toxicity is also observed in Drosophila and mouse transgenic tau models (65, 66).

It appears that tau is not the only synaptic protein whose abnormality may contribute to synaptic dysfunction in AD. A recent study found that overactivation of PAR-1 kinase in Drosophila causes the delocalization of Dlg, the fly homolog of the key postsynaptic scaffold protein Postsynaptic Density (PSD)-95, from the postsynaptic membrane at the neuromuscular junction (Figure 3). This effect is mediated through the direct phosphorylation of Dlg by PAR-1 (67). Dlg’s synaptic localization has been found to be decreased in AD patients and in mouse AD models (68–70). Given that aberrant activation of PAR-1-like kinases has been implicated in AD conditions, this study offers a new approach to understanding the molecular basis of synaptic dysfunction in AD.

Pathological Features and Molecular Mechanisms Involved in Parkinson’s Disease Pathogenesis

PD is the most common movement disorder and the second most common neurodegenerative disease. The classical form of the disease is clinically characterized by muscle rigidity, resting tremor, bradykinesia, and postural instability. The movement abnormality in PD is largely attributable to the deficiency of brain dopamine content, which is caused by degeneration of dopaminergic neurons in the substantia nigra. A pathological hallmark of the disease is the formation of Lewy bodies, intra-cytoplasmic inclusions that are composed of α-Synuclein (α-Syn) and ubiquitin, among other proteins. The most common forms of PD are sporadic with no known cause. However, postmortem studies of sporadic cases have identified defective mitochondrial complex I function and oxidative damage in nigrostriatal dopaminergic neurons (71, 72). The oxidative damage is thought to reflect both the increased burden of free radical production resulting from dopamine metabolism and mitochondrial dysfunction and the limited scavenging ability of dopaminergic neurons (73, 74). The identification of environmental toxins such as 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) and rotenone, which cause PD-like symptoms by inhibiting complex I function and promoting oxidative stress, further supports the mitochondrial dysfunction and oxidative stress models of PD and provides pharmacological tools for disease studies (14, 75).

Epidemiological studies also point to the contribution of genetic factors in the pathogenesis of PD. At least 10 distinct loci (PARK1–11) have been linked to rare familial forms of PD (FPD) (76, 77). As with FAD, it is anticipated that understanding the molecular lesions associated with these FPD genes will shed light on the pathogenesis of the more common forms of the disease. To date, six FPD genes have been molecularly cloned, including α-Syn, Parkin, Ubiquitin C-terminal hydrolase–1 (Uchl-1), DJ-1, phosphatase and tensin homolog (PTEN)-induced kinase 1 (PINK1), and leucine-rich repeat kinase 2 (LRRK2) (78–84). Of these six genes, α-Syn and Parkin have been the most intensively studied. Although initial biochemical and biophysical studies linked mutations of both genes to aberrant protein folding and handling, more recent in vitro cell culture studies and in vivo genetic studies have revealed their connection to mitochondrial impairment and oxidative stress (85).

DJ-1, PINK1, and LRRK2 are three recently cloned FPD genes whose molecular functions (as they pertain to PD pathogenesis) remain poorly defined. PINK1 is a putative serine/threonine kinase localized to mitochondria (82, 86). PINK1 was initially isolated as a gene induced by overexpression of the tumor suppressor PTEN (87), but its functional relationship with PTEN and whether it participates in phosphatidylinositol 3–kinase (PI3K)/PTEN signaling are not known. Nevertheless, the subcellular localization of PINK1 protein emphasizes mitochondria as its primary site of activity. DJ-1 encodes a highly conserved protein belonging to the ThiJ/PfpI superfamily. LRRK2 encodes a large protein that contains a leucine-rich repeat (LRR) domain, a kinase domain, a RAS-like GTPase domain, and a WD-40 domain. Cell culture studies have shown that overexpression of LRRK2 can cause neuronal death in a kinase activity–dependent manner and that some of the pathogenic mutations may possess enhanced kinase activity (88).

To investigate the in vivo physiological roles of the FPD genes and to model dopaminergic degeneration caused by dysfunction of these genes, transgenic or knockout mouse models for α-Syn, Parkin, DJ-1, and Pink1 have been generated (89–96). Although these models exhibit various pathological and behavioral phenotypes, the cardinal feature of PD, namely selective degeneration of dopaminergic neurons, has not been faithfully reproduced. It is possible that the short lifespan of mice may have precluded observation of the late-onset dopaminergic degeneration phenotype. Alternatively, certain fundamental differences in dopaminergic neuron physiology between mice and humans could make mice less vulnerable to the dysfunction of the individual FPD genes. Construction of double- or triple-mutant combinations or use of genetic or pharmacological manipulations to alter the aging process may help reveal a dopaminergic degeneration phenotype.

Modeling Parkinson’s Disease in Drosophila

Role of Molecular Chaperones in Disease Pathogenesis

Using a candidate gene approach, investigators have shown that directed expression of the molecular chaperone HSP70, which was previously shown to be effective in ameliorating neurodegeneration in polyglutamine (polyQ) models (described in detail below), prevents dopaminergic neuronal loss associated with α-Syn expression in Drosophila (97) (Figure 6). Further, interference with endogenous chaperone activity exacerbates α-Syn toxicity. The authors’ observation that Lewy bodies in human postmortem tissue are immunopositive for molecular chaperones demonstrates the relevance of these findings to human diseases. These findings suggest that abnormal protein folding and aggregation may be common mechanisms in the pathogenesis of polyQ- and α-Syn-associated neurodegenerative diseases.

A recent human genetics study indicated that polymorphisms in the human Hsp70 genes may affect susceptibility to PD (111). The potential of Hsp70 gene therapy in a mouse model of MPTP-induced PD was recently tested. Hsp70 gene transfer to dopamine neurons by a recombinant adeno-associated virus significantly protects the mouse dopaminergic system against MPTP-induced dopaminergic toxicity (112). The effectiveness of HSP70 in preventing dopaminergic degeneration in an acute pharmacological model and in a chronic genetic model suggests that increasing chaperone activity may offer a new avenue for the treatment of PD.

Regulation of α-Syn Toxicity by Phosphorylation

It was previously shown in mammalian systems that α-Syn protein found in Lewy bodies is phosphorylated at serine 129 (Ser129) (113). To assess the role of phosphorylation of Ser129 in α-Syn toxicity and inclusion formation, Chen & Feany (114) performed mutagenesis studies in the Drosophila α-Syn model of PD. Mutating Ser129 to nonphosphorylatable alanine completely suppressed dopaminergic neuronal loss caused by human α-Syn, whereas changing Ser129 to the phosphomimetic aspartate significantly enhanced α-Syn toxicity. The authors further showed that G protein–coupled receptor kinase 2 (Gprk2) phosphorylated Ser129 and enhanced α-Syn toxicity in vivo (Figure 6). Furthermore, blocking phosphorylation at Ser129 substantially increased aggregate formation. Ser129 phosphorylation thus appears to be important in regulating α-Syn neurotoxicity and inclusion formation (114). Given the broad involvement of α-Syn abnormality in familial as well as sporadic PD, additional studies investigating upstream signaling mechanisms that impinge on Gprk2 to regulate α-Syn phosphorylation and toxicity will lead to further insights into PD pathogenesis in general.

Neuroprotective Function of Parkin

The candidate gene approach was also used to study the relationship between Parkin and α-Syn in dopaminergic neuron degeneration. There are indications that these two genes may functionally interact in the disease process. Parkin colocalizes with α-Syn in Lewy bodies, and an O-glycosylated form of α-Syn is reported to be a substrate of Parkin (115). In Drosophila, overexpression of Parkin effectively suppresses α-Syn-induced degeneration of dopaminergic neurons (11). This suppression is not associated with significant changes in total cellular α-Syn protein levels, suggesting that the toxic species of α-Syn, if targeted by Parkin, may represent only a small portion of total α-Syn protein. Alternatively, Parkin suppression of α-Syn toxicity may act through a mechanism unrelated to α-Syn degradation by the ubiquitin-proteasome pathway (Figure 6). Using two other assays of α-Syn-induced toxicity, premature decline of climbing ability, and degeneration of retinal cells, another group also reported protective effects of Parkin against α-Syn toxicity in Drosophila (116).

The neuroprotective effect of Parkin has also been demonstrated in mammalian systems. In cell culture systems, overexpression of mutant α-Syn decreases proteasome function, which can be counteracted by overexpression of Parkin (117). In an in vivo study carried out in a rat lentiviral model of PD, overexpression of wild-type rat Parkin protected against the toxicity of disease-associated human α-Syn A30P. Animals overexpressing Parkin show significant reductions in α-Syn-induced neuropathology, leading to preservation of TH-positive cell bodies in the substantia nigra and to sparing of TH-positive nerve terminals in the striatum. Surprisingly, Parkin-mediated neuroprotection has been associated with an increase of hyper-phosphorylated α-Syn inclusions, supporting a role for Parkin in the genesis of Lewy bodies and implicating an intriguing neuroprotective role for α-Syn inclusion formation in the disease process (118). These studies suggest that compounds capable of upregulating Parkin activity could offer a promising treatment for PD.

DJ-1 and PI3K/AKT Signaling

Using DJ-1A RNAi-induced photoreceptor degeneration as an assay, a systematic genetic interaction study was carried out to identify genes and pathways that can modify DJ-1A-associated neurodegeneration. This search led to the isolation of components of the PI3K/Akt signaling pathway as specific genetic modifiers (106) (Figure 7). Reduction of PI3K/Akt signaling enhanced DJ-1A RNAi phenotypes, whereas activation of PI3K/Akt signaling significantly rescued DJ-1A RNAi phenotypes. The modifying effects of PI3K/Akt signaling on DJ-1A RNAi phenotypes was also manifested in dopaminergic neurons. Consistent with the genetic interaction studies, PI3K/Akt signaling was impaired in DJ-1A RNAi animals, as indicated by the reduction of p-Akt levels. An independent study (119) identified DJ-1A as a genetic modifier of PTEN-induced small-eye phenotype in Drosophila. The authors performed further biochemical studies in mammalian cells and found that DJ-1 RNAi results in decreased phosphorylation of PKB/Akt, whereas DJ-1 overexpression leads to hyperphosphorylation of PKB/Akt and increased cell survival. Furthermore, in primary breast cancer samples, DJ-1 expression correlates negatively with PTEN immunoreactivity and positively with Akt hyperphosphorylation (119). DJ-1 thus appears to be a key negative regulator of PTEN and may serve as a useful prognostic marker for cancer. These studies raise the interesting possibility that cancer and PD, two seemingly disparate diseases, may share certain underlying biochemical pathways or risk factors. Supporting a connection between cancer and PD, epidemiological studies found that incidences of tumorigenesis, especially the non-malignant tumors, were significantly lower in PD patients (120). One way to think about the connection between PD and cancer is that tumors can result from the imbalance of a common signaling pathway that controls cell proliferation and survival. Tipping the balance can result in overproliferation and cancer in one direction, and cell death and PD in the other.

Genetic Interaction Between Pink1 and Parkin and Their Physiological Roles in Mitochondria

The indirect flight muscle degeneration, male sterility, dopaminergic neuron degeneration, and mitochondrial morphological defects observed in dPink1-mutant or dPink1 RNAi flies are strikingly similar to those seen in dParkin mutants. This observation prompted genetic epistasis studies to investigate the relationship between Pink1 and Parkin. Overexpression of Parkin rescues the male sterility and mitochondrial morphology defects of dPink1 mutants, whereas double mutants lacking both Pink1 and Parkin function show phenotypes identical to those observed in either mutant alone. Overexpression of Pink1 has no effect on dParkin-mutant phenotypes. These observations suggest that Pink1 and Parkin function in the same pathway, with Pink1 acting upstream of Parkin (107–109) (Figure 6). Consistent with this notion, Parkin protein level is reduced in dPink1 RNAi animals (109). Future studies are needed to test whether Pink1 can directly modify Parkin.

The exact mitochondrial process regulated by Pink1 and Parkin is completely unknown. An understanding of this process will provide significant insight into the disease mechanism and could suggest novel therapeutic targets. Given the prominent mitochondrial morphological defects seen in dPink1 and dParkin mutants, one particular area that warrants future investigation is the mitochondrial fission/fusion pathway, a conserved mitochondria-remodeling process that controls the dynamic distribution and morphology of mitochondria in all eukaryotes (121). Mitochondrial fission/fusion has recently been shown to be important for regulating synaptic structure and plasticity (122), and imbalance of mitochondrial fission/fusion can lead to neurodegeneration (123). Future studies will ascertain whether Pink1/Parkin regulate mitochondrial morphology through interaction with components of the fission/fusion machinery.

It is intriguing that the indirect flight muscle and dopaminergic neurons are particularly vulnerable to Pink1 and Parkin mutations. It is possible that high endogenous levels of oxidative stress in these tissues may confer this vulnerability. Oxidative stress has been recognized as a major cause of aging and is considered a major pathogenic factor in PD (124). As the main cellular sites where oxygen is consumed, mitochondria produce most of the ROS in a cell. Thus it is not surprising that mitochondrial dysfunction is closely associated with oxidative stress. As one of the most metabolically active tissues, fly indirect flight muscle probably produces more ROS than any other tissue and is presumably under constant oxidative stress (125). Dopaminergic neurons are also under constant oxidative stress due to the oxidative metabolism of dopamine. These conditions may confer particular vulnerability of these tissues to additional oxidative insults caused by mitochondrial dysfunction in dPink1 and dParkin mutants.

Gene Expression Changes That Predate Neurodegeneration

Whole-genome analysis of gene expression changes represents an alternative approach to defining disease mechanism and identifying new players. To investigate the transcriptional program encoding the molecular machineries involved in α-Syn-induced neuropathology, the expression of the entire Drosophila genome at presymptomatic, early, and advanced disease stages was characterized by DNA microarray analysis (126). Signature transcripts including lipid-, energy-, and membrane-transport mRNAs were found to be closely linked with disease pathogenesis. The advantage of the DNA microarray approach is that it allows detection of gene expression changes at the presymptomatic stage. Indeed, changes in the expression of genes involved in catecholamine biosynthesis, mitochondrial function, and lipid binding were observed in histologically and behaviorally normal 1-day-old α-Syn transgenic flies that became symptomatic at 20 days. In an age-matched tauopathy fly model, the transcription of α-Syn-associated genes was normal. This result is consistent with the notion that the pathways of neurodegeneration may be highly specific for different diseases (55, 127).

Genes Involved in Polyglutamine Diseases

The molecular cloning of the gene responsible for Huntington’s disease (HD) more than a decade ago revealed that the disease-causing mutation is a CAG trinucleotide repeat that encodes polyQs in the Huntingtin (Htt) protein. Later studies showed that diverse dominantly inherited neurodegenerative diseases are caused by CAG repeat expansion within disease-specific proteins. These include spinobulbar muscular atrophy (SBMA), spinocerebellar ataxia (SCA) 1, 2, 3, 6, 7, and 17, and dentatorubralpallidoluysian atrophy. These polyQ expansions confer toxic properties to the resident proteins. There is a good correlation between repeat length and disease severity (128). These polyQ repeats form intranuclear inclusions in neurons and cultured cells and are intrinsically toxic. However, when the polyQ sequence was placed in other proteins, its toxicity was altered (129). This suggested that the flanking protein sequence in the endogenous protein can modulate polyQ toxicity, leading to the suggestion that the protein context-dependent toxicity of polyQ may explain the neuronal subtype specificity of the different diseases. As is discussed below, protein cofactors binding to the nonpolyQ regions of the disease gene products could have profound effects on the toxicity of polyQ-containing proteins.

Fly Models of Polyglutamine Diseases

A number of polyQ disease models have been established in Drosophila. These include models for SCA-3, SCA-1, HD, and SBMA. In all four cases, expression of truncated, expanded polyQ-containing protein fragments leads to retinal degeneration. In the case of SCA-1, overexpression of the nonexpanded wild-type ataxin-1 protein also results in retinal degeneration, indicating that the ataxin-1 protein is intrinsically toxic at high levels (8). In the SBMA model, ingestion of androgen and its known antagonists promote neurodegeneration in flies expressing full-length mutant human androgen receptor (hAR), which is otherwise nontoxic (130). This suggests that hormone binding, subsequent structural alteration, and nuclear localization of mutant hAR protein are critical steps in the pathogenesis of SMBA. Because the fly eye is composed of ~800 eye units known as ommatidia, which are highly organized on an epithelial surface, subtle changes in gene function are amplified in this system. This feature, plus the fact that the eye is dispensable for animal viability, has enabled large-scale genetic screens to identify second-site genetic modifiers of the polyQ models. These models have also been widely used in testing the therapeutic effects of candidate compounds.

Forward Genetic Screens Identify Novel Mechanisms and Pathways Involved in Spinocerebellar Ataxia–1 Pathogenesis

Significant insights into the pathogenic pathways of polyQ diseases were obtained from the SCA-1 model. In loss-of-function and overexpression screens for genes that can modify expanded ataxin-1-induced neurodegeneration, genes involved in a number of biological processes were identified (8). Not surprisingly, molecular chaperones, which have previously been implicated in regulating polyQ toxicity, were recovered in the SCA-1 screen. This screen also identified modifier genes involved in RNA processing, nuclear transport, transcriptional regulation (histone deacetylation), and cellular detoxification. Such processes had not been implicated in polyQ pathogenesis prior to this report. The identification of these new genes and mechanisms demonstrates the power of fly genetics in dissecting disease mechanisms. Further analysis of these modifier gene products in vertebrate models will help validate the findings from the fly models.

Regulation of Ataxin-1 Toxicity by Phosphorylation and 14-3-3 Binding

Using the SCA-1 model, investigators found that the 14-3-3 protein, an abundant and multifunctional protein, can bind to ataxin-1 (Atx-1), leading to the stabilization and accumulation of that toxic protein (131). 14-3-3 proteins are known to bind other proteins in a phosphorylation-dependent manner. Indeed, the interaction between Atx-1 and 14-3-3 is phosphorylation dependent. The PI3K/AKT pathway was found to regulate the association between Atx-1 and 14-3-3. AKT phosphorylates Atx-1 at a conserved Ser776 residue, and this phosphorylation event provides a binding site for 14-3-3 (Figure 8). In support of the role of AKT and 14-3-3 in SCA-1 pathogenesis, in vivo genetic manipulation of AKT and 14-3-3 levels modulates Atx-1-induced neurodegeneration. These results suggest that manipulating PI3K/AKT-pathway activity or interfering with the interaction between 14-3-3 and Atx-1 may provide new strategies for therapeutic intervention. Significantly, phosphorylation of Atx-1 also appeared to be important in regulating disease pathogenesis in a mouse SCA-1 model (132). As discussed above, activation of the PI3K/Akt pathway can suppress the DJ-1A-associated photoreceptor and dopaminergic neuron degeneration, but the same genetic manipulation enhances SCA-1 pathogenesis. This reinforces the notion that pathways of neurodegeneration are highly specific for different diseases.

Endogenous Factors and Flanking Protein Sequences Contribute to the Selectivity of Polyglutamine Toxicity

The role of endogenous factors in influencing polyQ toxicity is best illustrated by studies of Atx-1 in fly and mouse. The Drosophila Atx-1 (dAtx-1) has a conserved AXH domain but lacks the polyQ tract found in human Atx-1 (hAtx-1). Overexpression of hAtx-1 in flies produces phenotypes similar to those of dAtx-1 but different from the polyQ peptide alone. The transcription factors Senseless and Gfi-1 have been found to interact with Atx-1’s AXH domain. In flies, overexpression of Atx-1 inhibits sensory organ development by decreasing the level of Senseless protein. Similarly, overexpression of wild-type and polyQ-expanded hAtx-1 reduces Gfi-1 levels in mouse Purkinje cells. Deletion of the AXH domain abolishes the effects of glutamine-expanded hAtx-1 on Senseless/Gfi-1. Furthermore, loss of Gfi-1 mimics SCA-1 phenotypes in Purkinje cells (133). These results suggest that endogenous factors such as Senseless/Gfi-1 contribute to the selective Purkinje cell–degeneration phenotype in SCA-1 through protein-protein interaction with the nonpolyQ portion of the Atx-1 protein (Figure 8).

The observation that the toxicity of polyQ repeats can be altered when placed in the context of other proteins provides further insights into the specificity of the different polyQ diseases (129). It has been found that polyQ peptide is intrinsically toxic in a polyQ length-dependent and progressive manner. However, when polyQ was expanded in Drosophila Dishevelled (a protein that contains a naturally occurring polyQ tract but that is not associated with disease), the toxic effect of expanded polyQ was neutralized by this new protein context. The dependence of polyQ toxicity on flanking sequence and on endogenous trans-acting factors may explain why selective neuronal populations are affected in each disease, despite the broader expression domains of the disease-gene products.

Modulation of Polyglutamine Toxicity by Sumoylation and Ubiquitination

Studies of a fly model of SBMA have emphasized the roles of the ubiquitin-proteasome pathway and the sumoylation pathway in polyQ pathogenesis. Expression of a pathogenic hAR protein with an expanded polyQ repeat results in nuclear and cytoplasmic inclusion formation and in cellular degeneration (134). The influence of ubiquitin-dependent modification and the proteasome pathway on neural degeneration was studied in this model. Compromising the ubiquitin/proteasome pathway enhanced degeneration and decreased polyQ protein solubility. Further, overexpression of a mutant form of the SUMO-1-activating enzyme Uba2 showed that polyQ-induced degeneration was intensified when the cellular SUMO-1 protein conjugation pathway was compromised, although the exact mechanism of mutant Uba2 action on polyQ was not addressed in this study. This study nevertheless implicates an interplay between the ubiquitin/proteasome and the SUMO pathways in modulating polyQ toxicity (134).

Another study of expanded Htt protein provided more conclusive evidence for the involvement of sumoylation in regulating polyQ toxicity (135). The authors found that sumoylation and ubiquitination compete for the same Lys residue in Htt. Sumoylation stabilized Htt, prevented aggregate formation, and enhanced the protein’s transcriptional repressor activity. In a fly HD model, sumoylation enhanced neurodegeneration, whereas ubiquitination had the opposite effect. Mutating the Lys residue prevented both sumoylation and ubiquitination and abrogated neurodegeneration. Sumoylation thus appears to be an essential step in polyQ pathogenesis and may represent a unique opportunity for therapeutic intervention.

Role of Specific Cell Death Regulators in Polyglutamine Pathogenesis

As in other neurodegenerative diseases, the role of programmed cell death in polyQ disease pathogenesis is a matter of debate. In fly models, expression of the antiapoptotic protein p35 has little effect on polyQ toxicity (136). Further, apoptosis-related genes have not been recovered as modifiers from the various genetic screens. However, testing on candidate cell death genes has revealed a common role for Dark in polyQ pathogenesis; Dark is the fly homolog of human Apaf-1, a key regulator of apoptosis in mammals. PolyQ-induced cell death was dramatically suppressed in flies lacking Dark. This was accompanied by suppression of caspase activation and aggregate formation. Dark colocalizes with ubiquitinated protein aggregates. Further, mammalian Apaf-1 colocalizes with Htt-containing aggregates in a murine HD model and in HD brain, suggesting a common role for Dark/Apaf-1 in polyQ pathogenesis (137). The molecular pathway mediating Dark/Apaf-1 function remains to be defined. It is possible that Dark/Apaf-1 regulates polyQ aggregation and toxicity in a cell death–independent manner.

Axonal Transport and Polyglutamine Toxicity

Aggregates of polyQ-containing protein are found in the nucleus and cytoplasm of diseased neurons. Although there is considerable evidence for transcriptional deregulation in polyQ pathogenesis, presumably caused by nuclear aggregates, it is possible that the cytoplasmic aggregates also contribute to disease pathogenesis. Several studies have indicated that defects in axonal transport may contribute to polyQ toxicity (138–140), presumably due to the physical blockage of axonal transport by the large protein aggregates. Intriguingly, inhibition of the activity of the Drosophila homolog of Htt also leads to axonal transport defects (138), suggesting that polyQ may act by interfering with endogenous Htt gene function.

Noncoding Trinucleotide Repeat Disease

In addition to the polyQ diseases, where trinucleotide repeat expansion leads to toxic protein production, there are other diseases caused by trinucleotide repeat expansion within noncoding regions of mRNAs. Examples include SCA-8, -10, -12, myotonic dystrophy types 1 and 2, and fragile X mental retardation (FMR). In contrast to the polyQ diseases, the mechanism by which the noncoding region trinucleotide repeats cause pathology is not clear, and there are no good mammalian models for studying these diseases.

To understand how the rCGG repeats in the 5′ UTR of the FMR1 gene cause neurodegeneration in premutation carriers, 90 rCGG repeats—a number that is intermediate between the numbers of repeats found in patients (>200) and in normal individuals (<60)—were expressed in the fly eye. These 90 noncoding rCGG repeats were sufficient to cause degeneration of photoreceptors (141). Although no mutant protein was produced from this repeat sequence, the degenerating neurons formed HSP70- and ubiquitin-positive neuronal inclusion bodies. Furthermore, overexpression of HSP70 suppressed rCGG repeat–induced neuronal death. This result demonstrates that abnormal RNAs can also cause neurodegeneration. It is possible that the expanded RNAs cause toxicity by sequestering essential RNA-binding proteins and molecular chaperones.

In another fly model of RNA-induced neurodegeneration, the human SCA-8 gene, which encodes a noncoding RNA, was expressed in fly eye (142). It induced a late-onset, progressive neurodegeneration phenotype. Interestingly, both wild-type and trinucleotide-expanded SCA-8 induced neurodegeneration phenotypes. Using these SCA-8 models, a genetic screen was carried out; it resulted in the identification of four neuronally expressed RNA-binding proteins as modifiers of SCA-8-induced neurodegeneration. Further characterization of the biochemical interaction between these RNA-binding proteins and SCA-8 RNAs, both wild type and trinucleotide expanded, and the functional significance of these interactions will help us understand the mechanism of noncoding-region trinucleotide repeats–induced neurodegeneration.

In an effort to model the pathogenesis of myotonic dystrophy using Drosophila, transgenic flies incorporating CTG repeats in the 3′ UTR of a reporter gene were generated (143). In this study, expanded CUG repeats formed discrete ribonuclear foci in muscle cells. However, flies expressing expanded CUG repeats had no detectable pathological phenotype, suggesting that in contrast to expanded polyQ-containing proteins, neither the expanded CUG-repeat RNA nor the ribonuclear foci are directly toxic in muscle cells (143).

Amyotrophic Lateral Sclerosis and Prion Diseases

Two other prominent types of neurodegenerative disease, amyotrophic lateral sclerosis and prion diseases, have been successfully modeled in mice. Prion diseases are unique among neurodegenerative diseases in that the causative agents are thought to be proteins with abnormal confirmation that can alter the conformation of normal cellular proteins. A recent study (144) suggests that Drosophila may be a useful model for studying prion diseases. Transgenic flies heterologously expressing disease-associated (P101L) mouse prion protein (PrP) in cholinergic neurons exhibited severe locomotor dysfunction and premature death as larvae and as adults. Clinical abnormalities were accompanied by age-dependent accumulation of misfolded PrP molecules, intracellular PrP aggregates, and neuronal vacuoles. These transgenic flies thus exhibited several hallmark features of human Gerstmann–Straussler–Scheinker (GSS) syndrome (144). This model should enable future genetic and pharmacologic studies of GSS.