Huntingtin is a cytoplasmic protein; its functions are as yet undetermined. In mice, deletion of the huntingtin gene results in early embryonic lethality, whereas later deletion of huntingtin by conditional mutagenesis causes neuronal degeneration (Dragatsis, 2000; Duyao, 1995; Nasir, 1995; Zeitlin, 1995). In rat sciatic nerve axons, huntingtin is transported in both anterograde and retrograde pathways (Block-Galarza, 1997). Immunohistochemical studies in human and rat brains reveal cytoplasmic huntingtin within neurons, and biochemical analysis indicates that huntingtin is enriched in compartments containing vesicle-associated proteins (DiFiglia, 1995; DiFiglia, 1997). Huntingtin interacts with many proteins, including nuclear, transcriptional, and signaling proteins (Cattaneo, 2001; Freiman, 2002). One protein of particular interest is the huntingtin-associated protein 1 (HAP1; Li, 1996). Although its function is currently unknown, HAP1 is also transported in both anterograde and retrograde pathways (Block-Galarza, 1997) and is found associated with vesicle membranes in synaptosomal fractions, indicating that the HAP1 interaction with huntingtin may occur within axons (Engelender, 1997). Additionally, HAP1 strongly associates with p150Glued, a critical component of the dynein-based transport system (Engelender, 1997; Li, 1998). Recent work suggests a role for a Drosophila HAP1-like protein in kinesin-dependent transport of mitochondria (Stowers, 2002). Together, these findings lead to the still untested suggestion that huntingtin has an important function in the axonal transport machinery itself (Gunawardena, 2003 and references therein).

In Huntington's disease (HD), effecting the brain, aggregates of mutant huntingtin are observed in nuclear inclusions and in dystrophic neurites (DiFiglia, 1997; Becher, 1998). In HD transgenic mice, N-terminal huntingtin fragments and their aggregates initially accumulate in striatal neurons, and later these neurons form aggregates in axonal processes and terminals (Li, 2000). Neuropil aggregates have been observed in the striatum in the lateral globus pallidus (LGP), a region into which medium spiny neurons project. How a function that may normally be associated with cytoplasmic vesicles can contribute to nuclear dysfunction and whether this reflects a normal nuclear signaling role of huntingtin is unknown. A testable possibility is that (1) normal huntingtin has a role in axonal transport and (2) mutant huntingtin causes neuronal dysfunction by poisoning vesicular transport within neurons, which can ultimately contribute to neurodegeneration. This hypothesis has now been tested in vivo in Drosophila (Gunawardena, 2003 and references therein).

Huntington's disease (HD) is one of nine neurodegenerative diseases that result from the expansion of CAG repeats leading to proteins containing abnormally long polyQ tracts. Although little is known about the mechanism by which polyQ expansion leads to pathogenesis, it has been proposed that misfolding of the mutant protein triggers a cascade of events, ultimately causing disease. The misfolded protein may undergo proteolytic cleavage, interact with other proteins, self-aggregate, and in many cases, translocate into the nucleus. Indeed, a common characteristic of all polyQ diseases is the formation of nuclear or cytoplasmic and axonal or dendritic inclusions of the disease protein (Li, 2000; Li, 2001; Piccioni, 2002; Paulson, 1997; Becher, 1998; Ishikawa, 1999). In the nucleus, aggregated polyQ proteins have been suggested to recruit transcription factors, caspases, and molecular chaperones and other proteins, which may stimulate apoptosis (Gunawardena, 2003 and references therein).

This study finds that wild-type version of Drosophila huntingtin (htt) is needed for normal axonal transport. In addition, pathogenic polyQ proteins alone, or in the context of human htt exon 1 or in the context of another polyQ disease protein Machado-Joseph disease (MJD), also known as spinocerebellar ataxia 3 (SCA3), can interfere with the axonal transport machinery and cause neuronal apoptosis and organismal death. Thus, these findings demonstrate that pathogenic polyQ proteins can poison the axonal transport system and support the proposal that defects in axonal transport may contribute to neuronal failure in HD and other polyQ expansion diseases (Gunawardena, 2003).

An important concern in interpreting these data is whether the transport failures observed are caused by direct poisoning of the transport machinery by pathogenic polyQ proteins, or if it is an indirect consequence in neurons that have become sick or are dying from other causes. There are four strong arguments that support the interpretation that pathogenic polyQ proteins themselves poison the transport machinery leading to neuronal failure. (1) Two different screens carried out for axonal transport mutants suggest that this phenotype is relatively rare. In one screen, only 4 out of 12,000 mutagenized chromosomes exhibited a phenotype diagnostic of transport abnormalities, namely abnormal larval motility combined with organelle accumulations in axons (Bowman, 1999; Bowman, 2000). In another screen, 446 out of 13,000 mutants exhibited the larval motility phenotype. 114 of these were tested for organelle accumulations in axons, but only 3 were found to exhibit this phenotype. In this context, among mutations where the gene product is known, only mutations in genes encoding motor protein subunits, or genes encoding proteins where there is strong evidence to support a role in the transport machinery, cause this phenotype. (2) An overexpression screen using EP elements yielded 36 lines that had the larval motility phenotype out of 2300 tested when driven with 179Y-GAL4; only 3 of these had organelle accumulations within their axons. (3) Overexpression of GFP, or of any protein lacking the C terminus of APP relatives (that is thought to interact with the transport machinery), did not cause a transport phenotype (Gunawardena, 2001). In the current analysis of proteins implicated in polyQ expansion diseases, similar phenotypic selectivity was observed. Proteins with short polyQ regions do not cause transport failures unless motor protein dosage is reduced. (4) Induction of neuronal apoptosis by excess expression of the cell death gene reaper failed to cause axonal transport problems. Expression of MJD-65QNLS, which primarily targets to the nucleus and induces apoptosis, also did not cause transport failures. Thus, cell death or sick cells do not generally cause axonal transport defects (Gunawardena, 2003).

If expression of pathogenic polyQ proteins directly causes transport failures, what might be the mechanism? One possibility is that aggregation of pathogenic polyQ proteins in narrow axons can physically impair transport of large organelles or vesicles. While attractive, this possibility on its own does not easily account for the observation that polyQ proteins with repeat lengths that are not pathogenic can cause transport failures when combined with reductions in motor protein gene dose (e.g., MJD-27Q). In addition, recent work argues that aggregation per se may not be required for neuronal toxicity (Klement, 1998; Yoo, 2003). A second possibility is that titration of motor proteins from other critical pathways induces vesicle stalling during transport in narrow axons, which can nucleate organelle accumulations that block subsequent transport. While this possibility accounts for the observed ability of pathogenic polyQ proteins to significantly reduce the soluble pool of motor proteins, it does not easily account for the ability of short polyQ repeats to titrate motor proteins without causing transport failure. The third possibility, which is favored, is that motor protein titration and a propensity to aggregate and physically block transport in narrow axons act in concert to poison axonal transport. This mechanism, while more complex than the others, best accounts for a number of important observations including those that are not easily explained by the simplest models. For example, this mechanism accounts for the observed ability of short polyQ repeats to titrate soluble motor proteins but to poison the transport machinery only when motor protein gene dosage is further reduced. It is also consistent with the observation that motor protein gene dosage generally needs to be reduced by more than 50% to cause significant transport phenotypes. This mechanism also accounts for observations (Piccioni, 2002; Li, 2000; Li, 2001) that expression of polyQ proteins in neurons causes axonal inclusions that contain the polyQ proteins themselves. EM examination of the ultrastructure of these inclusions in a mouse model of Huntington's disease reveals a morphology of accumulated vesicles, organelles, and distended axons that is virtually identical to what was observe in axonal blockages formed in the Drosophila system. Similarly, axonal abnormalities, perinuclear and nuclear accumulations, with severe dysfunction in mechanosensory neurons were observed in C. elegans expressing pathogenic polyQ repeats in the context of htt (Parker, 2001). In addition, axonopathies are prominent in a number of polyQ expansion diseases (Li, 2000), and this view also accounts for the beneficial effects of chaperone increases upon transport phenotypes. Live analysis also shows the accumulation of YFP-tagged vesicles into nonmotile aggregates in axons of larvae expressing expanded polyQ repeats. This mechanism would also suggest that there is a phenotypic continuum caused by motor protein reductions or physical aggregation where the burden of either one alone, if substantial enough, can cause phenotypes. However, the relative extent to which axonal blockages contribute to transport failure as compared to motor protein depletion needs to be investigated further. It is also noted that this proposal has the virtue of providing a plausible explanation for the otherwise puzzling ability of broadly expressed proteins to cause neuron-specific toxicity in human disease. Thus, it is conceivable that defects in axonal pathways may contribute to early disease neuropathology (Gunawardena, 2003).

This study provides direct evidence that Htt is required for normal axonal transport. Neuronal depletion of Drosophila htt using RNAi caused an axonal blockage phenotype, which is characteristic of mutations not only in cytoskeletal motor proteins that are required for axonal transport, but also in proteins that function as receptors for motors (Bowman, 2000; Gunawardena, 2001). These axonal blockages were enhanced by a 50% reduction in kinesin. Loss of htt in the eye caused a distinct degenerative phenotype similar to what has been observed in weak mutations of DLC (Bowman, 1999), DIC, and heterozygous dominant mutations of p150^Glued (Boylan, 2000). Until now, previous work has only hinted at a possible transport function for Htt but provided no direct evidence in support of this important proposal. For example, mouse models of HD and conditional Htt knockout mice all exhibited degeneration of axon fibers, compatible with, but not establishing, a function for Htt in axonal transport (Dragatsis, 2000; DiFiglia, 1997; Hodgson, 1999; Li, 2000). Thus, the current data together with Htt localization data strongly support a functional role for htt in fast axonal transport. Although how Htt associates with the axonal transport machinery is still unclear, it is proposed that a subclass of vesicles containing Htt may associate with motor proteins via HAP1, or a similar protein, which establishes the link between Htt and p150^Glued, thereby enabling transport. Although a true Drosophila homolog of HAP1 has yet to be identified, Drosophila Milton is related to HAP1 and has been suggested to be required for kinesin-dependent transport of mitochrondria (Stowers, 2002). In addition, many coiled-coil linker proteins exist that could facilitate this connection (Gunawardena, 2003).

The proposal that Htt is required for axonal transport explains why both reduction and gain of function cause similar phenotypes, since both can lead to failures of vesicle transport that might physically and biochemically cause organelle blockages in axons. For example, in mouse models, both loss (Dragatsis, 2000) and gain of function of htt causes neurodegeneration (Mangiarini, 1996; DiFiglia, 1997; Li, 2000) and axonal pathology (Sapp, 1999; Li, 2000; Li, 2001). It is striking that in the Drosophila system too, both loss of htt function and polyQ-induced gain of function cause similar axonal blockage phenotypes including neurodegeneration in the adult eye, which may result due to disruption of a specialized neuronal pathway. Both processes could contribute to the observed reduced trafficking of the neurotrophic factor BDNF in mouse HD brains (Zuccato, 2001). Indeed, HD is a dominantly inherited disease with both homozygous and heterozygous individuals affected similarly by a gained toxic function (Gunawardena, 2003).

An important point of controversy is whether neuronal toxicity in HD and other polyQ diseases results from nuclear or cytoplasmic events. PolyQ-induced disease pathogenesis can occur via two mechanisms -- one that induces apoptosis by nuclear accumulation, and the other that induces neuronal dysfunction by disrupting axonal transport -- although these two pathways may not be mutually exclusive. PolyQ-induced neuronal death did not result upon expression of a pathogenic polyQ protein restricted to the cytoplasm by the addition of a nuclear export signal (NES), although axonal blockages formed and organismal death resulted. In contrast, expression of a nuclear-targeted polyQ protein (NLS) caused nuclear accumulations, apoptosis, and lethality. These findings suggest that translocation of polyQ protein into the nucleus is required for cell death and that cytoplasmic polyQ proteins can cause axonal blockage. In fact, similar to the situation of polyQ proteins without an NES, a 50% reduction in kinesin-enhanced organelle blockages and a 50% reduction in dynein caused early organismal lethality of polyQ-NES expressing animals. Taken together, these data support two pathways for pathogenesis by polyQ proteins. In the first, polyQ-induced cytoplasmic perturbations in axonal transport pathways could directly instigate neuronal failure and organismal death. In the second, accumulation of pathogenic polyQ proteins within the nucleus (perhaps enhanced by axonal blockages) and events triggered by the nuclear presence of pathogenic polyQ protein could trigger neuronal death, neuronal degeneration, and finally organismal death. These ideas are consistent with recent findings on the androgen receptor, which, when expanded by a polyQ stretch in the N-terminal A/B domain, causes spinal and bulbar muscular atrophy (SBMA), an X-linked, adult-onset neurodegenerative disorder. While expression of expanded polyQ repeats in the context of the androgen receptor also causes neuropil aggregates and alters the distribution of kinesin (Piccioni, 2002), abnormal binding of the ligand, androgen, to polyQ-expanded human androgen receptor causes neurodegeneration due to ligand-dependent structural alteration that promotes nuclear translocation (Takeyama, 2002). Thus, it is possible that pathogenic polyQ proteins cause polyQ-induced cytoplasmic accumulations and these accumulations may promote abnormal protein-protein interactions that could trigger a cascade of toxic events, ultimately leading to neurodegeneration and organismal death. Indeed, a recent study (Kayed, 2003) suggests that soluble oligomers, which are common to most aggregate forming diseases, may be cytotoxic (Gunawardena, 2003).

Both polyQ toxicity (Warrick, 1999; Kazemi-Esfarjani, 2000) and alpha-synuclein toxicity (Auluck, 2002) observed in the Drosophila adult eye and brain are dramatically modulated by excess chaperones. It is conceivable that polyQ toxicity within axons is also modulated by chaperones. Indeed, axonal blockages and neuronal cell death are completely suppressed by excess HSC70 together in transgenic lines expressing expanded polyQ repeats, although organismal lethality still persists. The chaperone interaction with misfolded mutant polyQ protein may prevent abnormal interactions with motor proteins and other proteins, thereby preventing organelle blockages within axons and neuronal death. Consistently, cytoplasmic axonal aggregations caused by excess polyQ repeats with NES are also suppressed by chaperone expression, although organismal lethality still occurs. Lethality, however, may result due to the fact that although chaperones are modulating abnormal or misfolded proteins, they may be unable to completely prevent the toxic activity of abnormal aggregations of disease protein in time for normal development to proceed (Gunawardena, 2003).

Protein aggregation appears to be a common manifestation in many neurodegenerative diseases, and increasing evidence suggests that such accumulations can be a major trigger of cellular stress and neuronal death (Wyss-Coray, 2002). In Alzheimer's disease, accumulation of the 4 kDa Aß fragment in amyloid plaques and aggregation of phosphorylated Tau in neurofibillary tangles is observed surrounded by degenerating neurites; deposits of aggregated prion proteins with amyloid-like structures are observed in mad cow or Creutzfeld-Jacobs disease; in Parkinson's disease, abnormal alpha-synuclein accumulations known as Lewy bodies are seen; as well, in HD and other expanded polyQ diseases, abnormal accumulations of mutant protein are observed as nuclear and sometimes axonal inclusions. The widespread occurrence of axonal (or dendritic) inclusions leads to the proposal that perturbations in transport could be a common pathway in neurodegenerative disease. In support of this idea, recent findings indicate that dynein (Hafezparast, 2003) and dynactin (Puls, 2003) mutations can induce motor neuron degeneration in mice and humans (Gunawardena, 2003).

In this context, the strongest evidence comes from HD, which is characterized by the preferential loss of striatal neurons. Strikingly, htt accumulations are found in axons of striatal projection neurons (Li, 2001), and it has been argued that these striatal axonal inclusions are better correlated with striatal neuron loss than the presence of nuclear inclusions. Expression of expanded polyQ repeats in the context of the androgen receptor also forms neuropil aggregates and alters the distribution of kinesin (Piccioni, 2002), further supporting the idea that early pathology can occur within axonal processes together with axonal inclusions. It is possible that wild-type htt is required for efficient vesicle trafficking of cortical BDNF, since mutant htt interfers with its anterograde transport, contributing to BDNF depletion in the striatum (Cattaneo, 2001). The importance in transport of neurotrophic factors is also evident in Alzheimer's disease, where one of the earliest detectable signs of disease is the loss of synapses and retrograde degeneration of neurons, accompanied by the decay of intracellular traffic (Terry, 2000). In addition, excess of APP proteins containing the toxic Aß region perturbs axonal transport pathways and causes neuronal cell death (Gunawardena, 2001). Thus, it is proposed that perturbation in axonal transport can contribute to early disease pathology owing to disruption in proper transport of essential neuronal components, triggering a cascade of events leading to neuronal failure and death (Gunawardena, 2003).

The Huntington's disease (HD) gene encodes a protein, huntingtin, with no known function and no detectable sequence similarity to other proteins in current databases. To gain insight into the normal biological role of huntingtin, a cDNA encoding a protein that is a likely homolog of the HD gene product in Drosophila melanogaster was isolated and sequenced. The complete sequence of 43,125 contiguous base pairs of genomic DNA was determined that encompass the Drosophila HD gene, allowing the intron-exon structure and 5'- and 3'-flanking regions to be delineated. The predicted Drosophila Huntingtin protein has 3583 amino acids, which is several hundred amino acids larger than any other previously characterized member of the HD family. Analysis of the genomic and cDNA sequences indicates that Drosophila HD has 29 exons, compared with the 67 exons present in vertebrate HD genes, and that Drosophila Huntingtin lacks the polyglutamine and polyproline stretches present in its mammalian counterparts. The Drosophila HD mRNA is expressed in a broad range of developmental stages and in the adult, a temporal pattern of expression similar to that observed for mammalian HD transcripts. Five regions of high similarity can be discerned from multiple sequence alignments between Drosophila and vertebrate huntingtins. These regions may define functionally important domains within the protein (Li, 1999).

Human huntingtin is notable for contributing to the definition of HEAT (Huntingtin, Elongation factor 3, protein phosphatase 2A, TOR1) repeats, degenerate ~38 amino acid motifs that normally appear in tandem arrays with each unit consisting of two helical domains separated by a non-helical region (Andrade, 1995). The crystal structures of the PP2A PR65/A subunit and ß-importin show these proteins to be composed almost entirely of 15 and 19 HEAT repeats, respectively, which form a flexible solenoid-like structure as they cooperate to mediate a number of different protein-protein interactions central to the proteins' functions. Despite its large size, the 10 HEAT repeats of human huntingtin located in approximately the amino-terminal one-half of the protein were the only distinct functional motifs recognized, providing limited clues to the protein's function. However, the amino-terminal segment of huntingtin has been reported to interact with more than two dozen proteins that implicate it in such diverse processes as signal transduction, transcriptional regulation, RNA splicing, intracellular trafficking and cytoskeletal function (Takano, 2002).

The presence of conserved functional motifs in vertebrate and Drosophila Huntingtin was sought by cross-species comparison in the carboxyl-terminal region and throughout the proteins. Comparison with the Prosite HEAT profile revealed only sub-significant matches to two HEAT motifs (amino acids 58-95 and 1326-1364) because of the extensive sequence diversity in these repeats. To capture this diversity, MEME (Multiple EM for Motif Elicitation) was employed to construct ungapped motifs (represented as position-dependent probability matrices) from known HEAT repeats and the GenBank nr protein database was searched with MAST (Takano, 2002).

MEME motifs, trained with known huntingtin HEAT repeats from the amino-terminal segment, or with non-huntingtin HEAT repeats, efficiently detected the corresponding huntingtin HEAT repeats and other HEAT-containing proteins. However, they also frequently revealed significant matches to additional HEAT-like sequences in the original proteins, including some in Huntingtin's carboxyl-terminal region. Consequently, an iterative analysis was carried out of vertebrate huntingtin HEAT repeats, since after each round, newly implicated HEAT-like sequences from huntingtin were incorpored into species-specific and cross-species MEME motifs. This iterative approach culminated in the detection of 36 HEAT-like sequences in the vertebrate huntingtins, extending across the entire protein. The consensus secondary structure prediction for these HEAT-like sequences is a pair of helical domains, separated by a non-helical spacer. These findings are consistent with Andrade (2001) who recently reported a few additional HEAT repeats in the vertebrate huntingtins, detected by a homology-based iterative gapped alignment procedure (REP). The data suggest that huntingtin is largely made up of HEAT-like sequences, perhaps explaining the lack of other recognizable functional motifs (Takano, 2002).

Neither the MEME motif representing all vertebrate huntingtin HEAT-like sequences nor the REP program efficiently detects equivalent sequences in Drosophila huntingtin. However, an iterative process using MEME motifs based on Drosophila HEAT repeats, sometimes in concert with fish HEAT-like huntingtin sequences, revealed 28 HEAT-like sequences in Drosophila huntingtin, again spanning the entire protein. Four of these (D1, D13, D16, and D28), located in appropriate order and position within the protein, could be related directly to corresponding vertebrate HEAT-like sequences (H2, H12, H16, and H35) using cross-species MEME motifs representing individual vertebrate HEAT-like sequences. The consensus secondary structure prediction for the 28 Drosophila HEAT-like sequences indicates a conservation of structure despite divergence in primary sequence (Takano, 2002).