Polyglutamine Expansion in Ataxin-3 Does Not Affect Protein Stability

Polyglutamine proteins that cause neurodegenerative disease are known to form proteinaceous aggregates, such as nuclear inclusions, in the neurons of affected patients. Although polyglutamine proteins have been shown to form fibrillar aggregates in a variety of con-texts, the mechanisms underlying the aberrant conformational changes and aggregation are still not well un-derstood. In this study, we have investigated the hypothesis that polyglutamine expansion in the protein ataxin-3 destabilizes the native protein, leading to the accumulation of a partially unfolded, aggregation-prone intermediate. To examine the relationship between polyglutamine length and native state stability, we produced and analyzed three ataxin-3 variants containing 15, 28, and 50 residues in their respective glutamine tracts. At pH 7.4 and 37 °C, Atax3(Q50), which lies within the pathological range, formed fibrils significantly faster than the other proteins. Somewhat surprisingly, we observed no difference in the acid-induced equilibrium and kinetic un/folding transitions of all three proteins, which indicates that the stability of the native conformation was not affected by polyglutamine tract extension. This has led us to reconsider the mechanisms and factors involved in ataxin-3 misfolding, and we have developed a new model for the aggregation process in which the pathways of un/folding and misfolding are distinct and separate. Furthermore, given that native state stability is unaffected by polyglutamine length, we consider the possible role and influence of other factors in the fibrillization of ataxin-3. The polyglutamine diseases are now well Experiments were performed on an Applied Photophysics SF.18MV stopped flow apparatus. Unfolding was monitored by measuring the changes in fluorescence intensity at 360 nm using an excitation wave- length of 278 nm. Unfolding at pH 1.9 was performed by rapidly mixing one volume of 0.1 mg/ml protein solution in PBS, pH 7.4, with an equal volume of 50 m M HCl buffer, pH 1.3, at 25 °C. Refolding experiments were performed by initially preincubating the protein for 30 min at 25 °C at a concentration of 2.5 (cid:2) M in 50 m M HCl at pH 1.9. The protein was refolded by rapidly mixing one volume of the unfolded reaction mixture with an equal part of 20 (cid:4) PBS, pH 7.4; the final pH of refolding was 7.2. The data collected from unfolding and refolding experiments were collected from 10 experiments, which were averaged and fitted to either single or double exponential functions as appropriate.

The polyglutamine diseases are now well established as a group of genetic neurodegenerative disorders arising from the expansion of a repeated glutamine tract in specific proteins (1,2). Each of the nine currently identified polyglutamine diseases is associated with a particular protein, with the manifestation of disease generally observed when the polyglutamine tract of the relevant protein is elongated to over a threshold of ϳ40 residues. The protein ataxin-3, which has recently been associated with ubiquitin binding and deubiquitinating functions (3)(4)(5)(6), is responsible for the condition spinocerebellar ataxia type 3, also known as Machado-Joseph Disease (7), in which the polyglutamine tract of ataxin-3 is expanded to over 45 residues. 1 The different polyglutamine proteins share little or no structural or functional homology outside of the glutamine tract, yet fundamental similarities are consistently observed in the various disease states, with progressive dysfunction and death of neurons resulting in neurodegeneration and eventual death. The expanded polyglutamine tract is generally accepted to be the key causal element of the disease process, especially because increasingly longer polyglutamine proteins are associated with earlier onset and more severe manifestation of the disease state (2). Another striking feature of polyglutamine diseases is the formation of aggregates such as nuclear inclusions (NIs), 2 which contain the expanded form of disease-associated protein, within the neurons of affected patients (8 -10). The exact role and significance of NIs is as yet uncertain; however, they are inseparably associated with the manifestation of polyglutamine disease.
The mechanisms of polyglutamine toxicity are still largely unelucidated; however, an increasing number of reports indicate that the polyglutamine disorders belong to a wider range of diseases that associated with protein misfolding, including Alzheimer's disease, Parkinson's disease, and a range of amyloidoses (11)(12)(13)(14). All of these disorders involve the formation and deposition of protein aggregates within diseased tissues and/or cells. One of the most commonly observed forms of such aggregates is that of amyloid plaques, or fibrils, which have a characteristic cross-␤-sheet structure (15). In the case of polyglutamine diseases, the formation of NIs and the presence of proteasomal proteins in aggregates suggest that similar protein misfolding processes take place. In vivo, NIs display a range of morphologies, including granular, fibrillar, and amorphous forms (8,10,(16)(17)(18). In vitro, expanded forms of polyglutamine-containing proteins and peptides have been shown to form fibrils very readily (17,(19)(20)(21)(22). It has also been shown that nonpathological length proteins can also form fibrils, albeit generally at slower rates or under specific solution conditions (23)(24)(25)(26).
Recent work from our laboratory and others have shown that destabilization of the native state of nonpathological variants of ataxin-3 by various stresses such as chemical denaturation and heat can result in the formation of fibrillar aggregates (24,25,27). Based on various results, we and others have proposed that in the pathological state, the elongation of the polyglutamine tract disrupts the stability of the native conformation of the ataxin-3 (19,24). We have hypothesized that the loss of native state stability leads to the formation and accumulation of a partially unfolded, aggregation-prone species, resulting in fibrillization. This is supported by an earlier study showing that expansion of a polyglutamine tract inserted into myoglobin results in a progressive loss of conformational stability of the protein (22) and is also concordant with the generalized model of protein misfolding associated with a wider variety of conformational diseases (11,12,28,29).
Although expanded polyglutamine proteins have been shown to form fibrillar aggregates much more rapidly than shorter length proteins (22,23,26), to date, no study has examined the relative stabilities of variants ranging into the pathological length of ataxin-3 or any other polyglutamine protein. We have successfully produced milligram quantities of a pathological length variant, Atax3(Q50), which contains 50 glutamine repeats. This has allowed us to perform a comparative study with two other nonpathological variants, containing 15 and 28 residues in their polyglutamine tract (Atax3(Q15) and Atax3(Q28)) (see Fig. 1A). Building on previous results (24), we have performed a comprehensive investigation into the effects of polyglutamine length on the conformational changes involved in folding, unfolding, and native state stability of ataxin-3. Interestingly, our results show that upon polyglutamine expansion, the unfolding and folding transitions of ataxin-3 are essentially unchanged, indicating that the hypothesized mechanism of native state destabilization upon polyglutamine expansion in fact does not apply to ataxin-3. This has led us to consider other alternative pathways, mechanisms, and factors that may be significant in the pathogenic processes of aggregation and fibrillization.

Cloning of Atax3(Q15) and Atax3(Q50)
Atax3(Q15)-A construct encoding Atax3(Q15) was engineered by modification of the pQE30 Atax3(Q28) vector that had been created previously (24) from a vector encoding the MJD1a gene variant (7,30). EcoN1 sites flanking the polyglutamine-encoding region were introduced using the QuikChange technique sites coupled with a cassette mutagenesis approach allowing for the manipulation of the length of the polyglutamine tract without introducing or substituting amino acids within the protein. The following oligonucleotides were used: Nterminal end of the polyglutamine-coding region, sense, 5Ј-gaagcctactttaggaaacagcagcag-3Ј, and antisense, 3Ј-cttcggatgaaatcctttgtcgtcgtc-5Ј; and C-terminal of the polyglutamine-coding region, sense, 5Ј-cagcagcagcagcacctgcagcagggggacctatca-3Ј, and antisense, 3Ј-gtcgtcgtcgtcgtggacgtcgtccccctggatagt-5Ј. A Q15 cassette was then created with EcoNI restriction sites at either end using the following complementary oligonucleotides: 5Ј-tttgaaaaacagcagcaaaagcagcagcagcagcagcagcagcagcagc-3Ј and 3Ј-aactttttgtcgtcgttttcgtcgtcgtcgtcgtcgtcgtcgtcgtcgt-5Ј. The oligonucleotides were allowed to anneal slowly prior to their ligation into the EcoN1 digested vector. The nature of the sites allowed for directional cloning of the cassette, with the disappearance of the EcoNI sites as proof of successful ligation. DNA sequencing was used to verify the integrity of the construct.
Atax3(Q50)-The cDNA encoding human Atax3(Q50) was a gift from Henry Paulson. A HindIII digestion site was engineered into the 3Ј end of the gene. By exploiting this HindIII site and the 5Ј BamHI site, the cDNA was then subcloned into the pQE30 expression vector (Qiagen).

Expression and Purification of Ataxin-3 Variants
All of the variants were expressed and purified as described previously (24) with two modifications as outlined below. Firstly, an extra purification step was performed to aid removal of contaminating proteases. Following centrifugation of the sonicated cells, prior to binding to nickel-nitrilotriacetic acid affinity resin, the supernatant was mixed for 2 h at 4°C with 2 ml of p-aminobenzamidine-agarose previously equilibrated with lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, pH 8.0). The unbound protein, containing the relevant ataxin-3 variant, was mixed with nickel-nitrilotriacetic acid resin. All other stages of the purification were the same as previously described, except that glycerol (10% v/v) and Triton X-100 (0.1%) were also present in all of the buffers prior to the gel filtration chromatography stage.

Spectroscopic Methods
All of the fluorescence measurements were performed in PBS (137 mM NaCl, 2.68 mM KCl, 10.1 mM NaH 2 PO 4 , 1.76 mM KH 2 PO 4 ), pH 7.4. Fluorescence emission spectra were recorded on a PerkinElmer LS50B spectrofluorometer with a thermostated cuvette holder at 25°C, using a 1-cm-path length quartz cuvette. An excitation wavelength of 280 nm was used, with measurement of spectra from 300 to 400 nm. Emission and excitation slit widths were set at 4.0 nm, and a scan speed of 25 nm/min was used.
ThioT fluorescence was measured using an excitation wavelength of 445 nm, with the emission recorded at 480 nm. The excitation slit width used was 5.0 nm, and the emission slit width was 10.0 nm. The assay buffer contained PBS, pH 7.4. 15 l of the relevant protein sample was mixed with 485 l of assay buffer; these samples were analyzed immediately, with the fluorescence emission signal averaged over a 10-s period.
CD spectra were measured on a Jasco-810 spectropolarimeter at 25°C using a thermostatted cuvette with a path length of 0.1 cm. The spectra were recorded from 190 to 250 nm, using a scan speed of 20 nm/min, with 5 s/point signal averaging, and 222 measurements were recorded with the signal averaged over 30 s. The CD spectra were analyzed by spectral deconvolution using the CONTINLL algorithm (31,32) as provided by the DICHROWEB on-line facility (33,34).

Electron Microscopy
Transmission electron microscopy images were obtained using a Jeol JEM-200CX transmission electron microscope. The acceleration voltage was 100 kV. The samples were adsorbed onto a carbon-coated grid and stained with 1% (w/v) uranyl acetate.

Equilibrium Acid Denaturation and Fibrillogenesis
Different stock buffers, passed through a 0.22-m filter, were used at different pH ranges in equilibrium unfolding experiments. 50 mM Tris-HCl was used between pH 7 and 8, 50 mM sodium phosphate was used between pH 6 and 7, 50 mM sodium acetate was used between pH 4 and 6, 50 mM sodium citrate was used between pH 2 and 4, and 50 mM HCl was used below pH 2. The protein, in PBS buffer, was diluted with equal part stock buffers of the desired pH. The reaction mixture was allowed to equilibrate for 5 min at 25°C before fluorescence analysis, and the final pH was measured. More extensive equilibration for longer time periods up to 12 h showed no difference in results following spectroscopic analysis. The experiments were performed using final protein concentrations ranging from 1 to 5 M. The samples were analyzed by recording the fluorescence emission spectra between 300 and 400 nm ( ex ϭ 280 nm) or monitoring the change in far-UV CD signal at 222 nm. For each sample, the absorbance of all solutions at the excitation wavelength was also measured and recorded.
For the pH-induced fibrillogenesis assay, samples across a range of different pH values were prepared as for the equilibrium folding analysis, with a final protein concentration of 10.5 M. Following overnight incubation at 25°C, each sample was vortexed, and 50 l was removed and incubated with ThioT; the change in fluorescence signal was then monitored as previously described (24).

Unfolding/Refolding Kinetics
Experiments were performed on an Applied Photophysics SF.18MV stopped flow apparatus. Unfolding was monitored by measuring the changes in fluorescence intensity at 360 nm using an excitation wavelength of 278 nm. Unfolding at pH 1.9 was performed by rapidly mixing one volume of 0.1 mg/ml protein solution in PBS, pH 7.4, with an equal volume of 50 mM HCl buffer, pH 1.3, at 25°C. Refolding experiments were performed by initially preincubating the protein for 30 min at 25°C at a concentration of 2.5 M in 50 mM HCl at pH 1.9. The protein was refolded by rapidly mixing one volume of the unfolded reaction mixture with an equal part of 20ϫ PBS, pH 7.4; the final pH of refolding was 7.2. The data collected from unfolding and refolding experiments were collected from 10 experiments, which were averaged and fitted to either single or double exponential functions as appropriate.

Characterization of Ataxin-3 Variants-
We have successfully purified milligram quantities of ataxin-3 variants containing 15, 28, and 50 glutamine repeats, as shown in Fig. 1. In accordance with previous studies (22,25,35,36), we observed preferential binding of the polyglutamine-specific antibody 1C2 for the longer variant, which lies at the cusp of toxicity in Machado-Joseph disease, although nonpathological proteins could also be recognized (Fig. 1C). All three purified proteins were in a soluble monomeric form as judged by gel filtration analysis, and mass spectrometry showed that they contained the correct number of glutamine residues within their respective polyglutamine tracts (data not shown). The far-UV CD spectrum of Atax3(Q28) was concordant with previously reported data (24) (Fig. 2), and the spectra of all three variants were very similar to each other. This suggested that there were no major differences in secondary structure, an observation that was confirmed by spectral deconvolution of the far-UV spectra (Fig. 2, inset).
Rapid Fibrillization of Ataxin-3 Containing an Expanded Glutamine Repeat-A key behavioral feature of polyglutamineexpanded proteins is the ready formation of fibrils, as shown for a range of different proteins (17, 19 -22). In the case of ataxin-3, it has been shown that a variant containing 78 glutamine residues will readily fibrillize under native conditions  (19), whereas shorter variants are also capable of forming fibrillar aggregates under partially denaturing conditions (24,25). In the current study, we examined the behavior and propensity of ataxin-3 of different lengths to form fibrils. The threshold of disease length polyglutamine tracts for ataxin-3 is ϳ45 glutamine residues, 1 and as such, it is not surprising to observe that under native conditions, Atax3(Q50) forms fibrils much more quickly than both Atax3(Q15) and Atax3(Q28). When incubated at 37°C at 65 M, following a lag phase of 3-4 h, Atax3(Q50) exhibits a significant increase in ThioT fluorescence, which continues to increase for up to and beyond 24 h. In contrast, Atax3(Q15) and Atax3(Q28) show no increase in ThioT fluorescence for up to 24 h (Fig. 3A). The faster fibrillization of longer variants is consistent with previously reported results for other polyglutamine-containing proteins and peptides (22,23,26). Examination of the samples by electron microscopy after 24 h of incubation revealed that Atax3(Q50) formed filamentous aggregates with a fibrillar-like morphology, ranging from 20 to 45 nm in diameter (Fig. 3, B and C). Closer inspection of these fibrils suggested that they had a banded texture, suggestive of protofilament species that have assembled and aligned in a lateral manner (Fig. 3C). Inspection of Atax3(Q15) and Atax3(Q28) by electron microscopy at the same time point showed no formation of aggregates (data not shown).
Acid Denaturation of Ataxin-3-The propensity of nonpathological length variants of ataxin-3 to aggregate under denaturing conditions has suggested that native state destabilization by an expanded polyglutamine tract plays a key role in polyglutamine aggregation and fibrillization (19,24). Given our previous results with Atax3(Q28) (24), we extended our investigations into the pathological range of polyglutamine length using Atax3(Q50). Our initial guanidine equilibrium unfolding experiments with Atax3(Q15) and Atax3(Q28) indicated that there was no difference in the unfolding transitions between these two variants (data not shown). However, when we tried to extend these experiments to Atax3(Q50), we found that the pathological variant aggregated extremely rapidly in low guanidine concentrations within 2 h. Electron microscopy showed that these aggregates were fibrillar in nature (data not shown). This prevented us from performing a comparative study on the three variants using guanidine equilibrium denaturation.
Given the prohibitive difficulties of using guanidine, alternative methods of denaturation were explored. We found that upon incubation at pH 1.9, the acid-denatured state of Atax3(Q28), or what may be termed the "A state," displayed spectral properties similar to partially folded conformational species induced by pressure, temperature, and denaturant (24,27). Specifically, the A state showed an increased fluorescence signal, increased bis-ANS binding, and a red shift in fluorescence emission maximum ( max ) from 337 nm in the native state to 343 nm in the A state (Figs. 4A and 5C). Although the A state is not a fully unfolded conformation compared with the fully denatured species formed in 6 M guanidine (24), it is a non-native species with increased hydrophobic exposure. Furthermore, upon incubation at pH 1.9, all three variants showed no formation of fibrils over 24 h, as judged by ThioT fluorescence (Fig. 4B) and SDS-PAGE (data not shown). Therefore, it was possible to observe the effects of polyglutamine expansion on native state stability by acid denaturation.
Acid-induced Equilibrium Denaturation of Ataxin-3-The equilibrium acid denaturation curves of all three variants as monitored by fluorescence spectroscopy and far-UV CD are shown in Fig. 5. The transitions for all three proteins are very similar as judged by spectral peak, fluorescence intensity, bis-ANS binding, and CD signal. However, when the absorbance of the samples at 280 nm was measured, a significant increase in light scatter was observed between pH 4 and 5, indicative of aggregation (Fig. 5, shaded bars). These aggregates also bound ThioT, suggesting the formation of fibrils (Fig. 5E).
Kinetic Analysis of Acid-induced Denaturation of Ataxin-3-The aggregation of ataxin-3 between pH 4 and 5 precluded us from performing thermodynamic analysis of the transition to the A state. Therefore, we explored the effects of polyglutamine length on native state stability by investigating the kinetics of acid denaturation. Using stopped flow fluorescence, we were able to follow both the unfolding and refolding transitions of all three variants to and from pH 1.9 (Fig. 6). The rates of both unfolding and refolding for all three proteins were found to be not significantly different over a 10-fold concentration range. We found that the denaturation of all three variants to the A state occurred over 20 s via a single exponential transition. The refolding transition took place over a much shorter overall time scale (5 s) and was best described by a double exponential function, which indicates the presence of an intermediate upon the refolding pathway. There was an ϳ10-fold difference between the rates of the initial fast phase and the second, slower transition (Table I). Importantly, the refolding and unfolding rates for all three proteins were found to be independent of protein concentration, indicating that aggregation was not occurring during the experiments. Most interestingly, the results in Table I indicate that although there is some slight variation, between all three proteins both the unfolding and refolding rates were largely similar, indicating that the kinetic stability of the native state is not affected by polyglutamine expansion with respect to the un/folding pathway. DISCUSSION The results presented here describe a comprehensive investigation of the relationship between polyglutamine length and protein stability in proteins from both the pathological and nonpathological range. We began our investigation with the hypothesis that polyglutamine expansion may compromise the conformational stability of the native state of ataxin-3. Contrary to this expectation, all of our data suggest that the stability of the native state is unaffected by polyglutamine length. Although protein aggregation hindered our ability to glean quantitative data from equilibrium experiments, the kinetic data convincingly indicated that the propensity of the protein to unfold and refold to and from the A state is largely unaffected by polyglutamine length (Table I).
It has been shown previously in a number of different contexts that proteins and peptides containing elongated polyglutamine tracts fibrillize much faster than nonexpanded variants (22,23,26), and our current findings with ataxin-3 reflect this trend (Fig. 3). These observations are also consistent with the aggregation and toxicity of expanded polyglutamine proteins in cell and animal models (16,17,37,38), as well as disease tissue (8 -10, 36). In previous work, it has been suggested by ourselves and others that expansion of the polyglutamine tract may destabilize the native protein, leading to aggregation and fibrillization (19,22,24). This follows the generally accepted paradigm of protein misfolding diseases, in which the native conformation of a protein is destabilized by mutation, resulting in aberrant folding pathways and accumulation of a partially folded, aggregation-prone intermediate through which fibrils are formed (11,12,28,29) (Fig. 7A). In light of the greatly increased propensity of pathological length proteins to fibrillize, the finding that polyglutamine expansion has little effect on the native state stability of ataxin-3 is somewhat surprising. Our current results do not fit the "classic" model of protein misfolding, and thus we have had to rethink the possible mechanisms and factors involved in polyglutamine misfolding and pathogenesis.
As mentioned earlier, previous work indicates that expansion of a polyglutamine tract inserted into myoglobin increases destabilization of the protein (22), which is clearly contrary to the results presented here. However, it must be noted that myoglobin does not naturally contain a polyglutamine tract and was specifically targeted and studied as a model protein.
Therefore, the insertion of extraneous glutamine residues into the carrier protein could well be expected to have a destabilizing effect, whereas a naturally occurring polyglutamine protein such as ataxin-3 has evolved to maintain its native state stability in the face of polyglutamine expansion. In contrast to myoglobin, the polyglutamine expansion within a different model protein, CI2, was found to have no significant effect on the equilibrium and kinetic folding and unfolding transitions of the protein (39). Interestingly, the engineered polyglutamine tract in CI2 was located in a mobile loop linking two domains, which could be roughly likened to the case of ataxin-3, in which the polyglutamine-containing portion of the protein is believed to be a flexible solvent-exposed domain, in contrast to the folded N-terminal Josephin domain (40,41). On the other hand, in the myoglobin model (22), although a flexible loop was also targeted for polyglutamine insertion, the entire protein functions as a single domain. Thus, it is conceivable that in the case of ataxin-3 and CI2, the positioning of the polyglutamine tract in a mobile position completely external to folded domains may be a contributing factor to the minimal impact of glutamine expansion on the overall stability of the protein.
Given that polyglutamine elongation to a pathological range does not affect the stability of the native state of ataxin-3 relative to its unfolding pathway yet dramatically increases the rate of fibrillogenesis, it seems that the pathway of aggregation and fibrillization by ataxin-3 is quite distinct from the normal folding and unfolding transitions. It has been proposed that the fibril is an especially stable structure that any given protein can adopt (15,42); however, the kinetic barrier for fibril formation is generally too large to be overcome (42). Thus, in a nonpathological scenario, although the native state may be less thermodynamically stable than the fibril, the activation energy required for fibril formation is prohibitively high, and therefore the protein preferentially adopts an equilibrium between the native and unfolded conformations. In the generic protein misfolding model, mutations destabilize the native state, thereby reducing the kinetic barrier to unfolding and resulting in the accumulation of partially folded, intermediate ensembles that are more easily able to overcome the activation energy barrier to fibril formation (Fig. 7A). In the case of ataxin-3, we observe a separation of the kinetic pathways of fibrillogenesis and unfolding (Fig. 7B). Therefore, upon polyglutamine expansion, the change in free energy required to overcome the activation barrier between the native and unfolded conformations of ataxin-3 remains unaffected, but the activation barrier along the fibrillization pathway is lowered (Fig. 7C).
At first glance such behavior seems highly unconventional; however, it is not necessarily unprecedented. Disease-associated, aggregation-prone mutants of the prion protein have pre- viously been found to display very little or no difference in thermodynamic stability compared with the wild-type protein (43,44). Furthermore, a recent report also suggests that the map of fibrillization pathways may be more complex than previously thought (45). In their system, Plakoutsi et al. (45) found that aggregation and fibril formation does not necessarily involve a passage through the unfolding pathways and partially unfolded forms but rather may also occur via native or nearly native conformations. It would seem that a similar process applies to ataxin-3, whereby conformational changes caused by the expanded polyglutamine tract do not perturb the dynamics of un/folding but are sufficient to lower the activation barrier to the fibrillization pathway and thus increase the propensity of the protein to form fibrils from a native or nearly native conformation (Fig. 7, B and C).
The processes involved in polyglutamine-induced fibrillization are no doubt complex, and in the current field of knowledge our understanding of the important factors influencing conformational events remains limited. The observation that polyglutamine length does not compromise the native state stability of ataxin-3 raises the question of what other factors might influence the propensity of the expanded protein to fibrillize. A combination of recent studies elegantly indicate that with a change in pI of a model protein, conditions of varied pH have relatively little effect on native state stability and enzyme function, whereas protein solubility and the propensity to form fibrils are highly correlated (46,47). This scenario is mirrored by our results, where fibrillization occurs concomitantly with the loss of solubility of all variants between pH 4 and 5. The insolubility in this range is not entirely unexpected; proteins are generally most insoluble near their pI, and theoretical pI of ataxin-3 with 15, 28, and 50 glutamine residues was calculated to be 5.06. In light of this, and given that ataxin-3 fibrillization does not necessarily proceed via a partially unfolded state, it is possible that the solubility of the protein may be a key element involved in the process. This is also consistent with difficulties experienced by ourselves and others in producing satisfactory quantities of expanded ataxin-3 variants in soluble form (19,25) and also the well documented insolubility of polyglutamine peptides alone and their rapid propensity to form fibrils even at nonpathological lengths (20,23).
Alternatively, in light of structural models of polyglutamine aggregation that involve the length-dependent self-association of glutamine tracts in ␤-strand conformations (35,48,49), the addition of extra residues may affect the exposure, conformation and/or flexibility of the polyglutamine tract, independently of the rest of the protein. Therefore, it is possible that while not influencing the kinetic or thermodynamic stability of the protein, polyglutamine elongation in the exposed, mobile region may simply increase the probability of inter-or intramolecular interactions being formed by the glutamine residues, thus leading to fibrillogenesis. This possibility is further supported by recent work suggesting that in an amyloidogenic conformation, aggregation of a protein is led by the most solvent-exposed parts of a protein (50). Thus, an exposed, expanded polyglutamine tract with increased molecular contacts may be the starting point for a fibrillogenic reaction, independently of the stability of the overall protein.
Related to this latter point, another possible factor contributing to ataxin-3 fibrillogenesis without influencing stability might be the intrinsic conformational activities of the polyglutamine tract itself. A kinetic analysis of fibrillization by polyglutamine peptides has suggested that fibrillogenesis is initiated by a transition of a random coil polyglutamine tract to a compact, misfolded conformer, which then acts as a nucleus for the addition and conversion of further molecules, resulting in FIG. 6. Representative stopped flow traces for ataxin-3 variants undergoing pH-induced unfolding (A) and refolding (B). Unfolding was initiated by a 1:1 dilution of the native protein into 50 mM HCl, resulting in a final pH of 1.9. Refolding was performed by 1:2 dilution of acid-unfolded protein with 20ϫ PBS to produce a final pH 7.2. Upon refolding, the fluorescence emission spectrum of the refolded product was observed to be the same as the original native protein (data not shown). All un/folding transitions were monitored by the emitted fluorescence at 360 nm at 25°C. the elongation of the fibril (51). Under nonpathological circumstances this structural transition is energetically unfavorable; however, upon expansion of the polyglutamine tract the equilibrium between the native random coil and the misfolded conformer shifts toward the latter state, thus lowering the activation barrier to the fibrillogenesis pathway. In this way, fibrillization is driven by the changes within the polyglutamine tract itself rather than global changes in the overall protein. Rather, misfolding from the native conformation occurs via a separate pathway. C, schematic diagram of activation barriers involved in ataxin-3 un/folding and misfolding. The activation barrier for un/folding is unaffected by polyglutamine (polyQ) length, and therefore there is no change in the rates of un/folding. In contrast, the activation barrier to the fibrillar form is polyQ length-dependent, and therefore differences are observed in the kinetics of fibrillogenesis by different variants.
This does not preclude the possibility that ataxin-3 can form fibrils via a partially folded species; however, such an intermediate would most likely lie off the normal folding pathway (Fig. 7B).
We have previously suggested that domains external to the polyglutamine tract may be important in maintaining the solubility of the protein in the context of protecting against native state destabilization (24). Although our current findings dispute the original hypothesis, this does not preclude a role for nonpolyglutamine regions in preventing the misfolding reaction in the framework of the model of fibrillogenesis being driven by conformational change of the polyglutamine tract. Nonpathological length polyglutamine peptides are inherently capable of forming fibrils on their own, albeit slowly (20), yet this behavior generally is not observed in the context of an entire protein. Furthermore, sequences of the regions immediately flanking the polyglutamine tract have been shown to influence the rate of aggregation (52), and some polyglutamine proteins undergo necessary proteolysis prior to aggregation, with the release of a polyglutamine-containing fragment that subsequently forms fibrils (53)(54)(55)(56). Therefore, it is conceivable that domains such as the Josephin domain in ataxin-3 may play a role in maintaining the solubility of the protein, or exert a restraining force such that the conformational shift to the misfolded form is impeded or is energetically less favorable. However, upon expansion, the native random coil conformation of the polyglutamine tract becomes a more prominent feature of the overall protein structure, and more importantly, the intrinsic propensity of the polyglutamine tract to adopt an inappropriate conformation may be strengthened, and the counteractive influence of the external regions to this transition may become less effective. Similarly, the disruption of nonpolyglutamine domains and regions under denaturing conditions (24,25) or their removal by cleavage (53)(54)(55)(56) would also result in their loss of ability to prevent the transition to the misfolded conformer.
In this study, the somewhat unanticipated finding that polyglutamine length does not affect ataxin-3 stability or unfolding has necessitated a reconsideration of the process by which the protein misfolds. In light of recent evidence suggesting that the traditionally conceived pathways of aggregation may not necessarily apply to all proteins, it would appear that factors other than native state stability, such as glutamine exposure, solubility, integrity of external domains, or conformational flexibility, may also be key determinants in influencing the misfolding behavior of a protein. Thus, in the case of ataxin-3, the separation of normal un/folding and fibrillogenic pathways may represent an alternative mechanism of aberrant conformational change in proteins.