Folding and Fibril Formation of the Cell Cycle Protein Cks1*

The Saccharomyces cerevisiae Cks protein Cks1 has a COOH-terminal glutamine-rich sequence not present in other homologues. Cks proteins domain swap to form dimers but unique to Cks1 is the anti-parallel arrangement of protomers within the dimer. Despite the differences in Cks1 compared with other Cks proteins, we find the domain swapping properties are very similar. However, aggregation of Cks1 occurs by a route distinct from the other Cks proteins studied to date. Cks1 formed fibrillar aggregates at room temperature and neutral pH. During this process, Cks1 underwent proteolytic cleavage at a trypsin-like site into two fragments, the globular Cks domain and the glutamine-rich COOH terminus. At high protein concentrations, the rate of fibril formation was the same as the rate of proteolysis. The dominant species present within the fibrils was the glutamine-rich sequence. Consistent with this result, fibril formation was enhanced by addition of trypsin. Moreover, a truncated variant lacking the glutamine-rich sequence did not form fibrils under the same conditions. A lag phase at low protein concentrations indicates that fibril formation occurs through a nucleation and growth mechanism. The aggregates appear to resemble amyloid fibrils, in that they show the typical cross-β x-ray diffraction pattern. Moreover, infrared spectroscopy data indicate that the glutamine side chains are hydrogen-bonded along the axis of the fibril. Our results indicate that the proteolytic reaction is the crucial step initiating aggregation and demonstrate that Cks1 is a simple, tunable model system for exploring aggregation mechanisms associated with polyglutamine deposition diseases.

The Cks (Cyclin-dependent kinase subunit) 8 family of proteins is essential for control of the cell cycle in eukaryotes (1). The proteins are known to regulate the activity of cyclin-dependent kinases, although their underlying mechanism of action is not fully understood (2). The Cks structure consists of a four-stranded ␤-sheet capped at one end by two or three short ␣-helices (3) (Fig. 1A). Cks proteins undergo domainswapped dimerization via the reciprocal exchange of the COOH-terminal ␤-strand, both in solution (4,5) and in crystals (6,7), although the very high dissociation constants observed suggest that these proteins do not domain swap in vivo. Domain swapping in Cks proteins is mediated by a highly conserved sequence, the so-called "hinge loop " (8) that links the exchanged ␤-strand with the rest of the protein (Fig. 1A). This part of the protein is the only region where there are differences in conformation between the monomer and dimer forms. The Cks homologue from Saccharomyces cerevisiae, Cks1, is the largest member of the Cks family and has a unique sequence at the COOH terminus containing 16 consecutive glutamine residues followed by 17 residues rich in glutamine, proline, and serine (Fig. 1B). Also unique is an anti-parallel arrangement of monomer subunits in the domain-swapped dimer form of the protein and a disulfide link between the two subunits as a result of a cysteine residue in the hinge loop (3) (Fig. 1A).
The expansion of CAG trinucleotide repeats coding for polyglutamine (poly(Q)) leads to neurodegenerative disorders such as Huntington disease that are characterized by the formation of insoluble, fibrillar aggregates in the brain (reviewed in Refs. 9 and 10). A threshold number of 35-45 repeats, varying somewhat among the diseases, is needed for pathogenicity. This threshold number is in good agreement with the length dependence of the aggregation kinetics of poly(Q) sequences in vitro (11)(12)(13). Aggregates are thought to cause disease either by direct toxicity (14) or through abnormal interactions with other cellular proteins by interfering with the regulation of gene transcription (15,16). For some amyloid diseases it has been proposed that oligomeric intermediates formed early in the aggregation process may be even more harmful than mature fibrils (17) and it is speculated that this could also apply to poly(Q)-related disorders (18). At least nine inherited diseases are caused by CAG repeat expansions, but the proteins involved have little in common, either structurally or functionally, other than the poly(Q) region. Difficulties in working with the full-length proteins, on the one hand, and the low solubility of poly(Q) model peptides, on the other, mean that many details of the poly(Q) aggregation mechanism remain to be resolved. In particular, atomic information on the poly(Q) fibril structure has yet to be obtained and the role played by the nonpoly(Q) regions of the proteins in the aggregation process is not yet defined. Poly(Q) aggregates are ␤-sheet rich and exhibit many of the defining features of amyloid fibrils (11, 13, 19 -21). The aggregation of poly(Q) sequences is a nucleation-dependent process (20,22,23) similar to that seen for amyloidogenic peptides and proteins such as the A␤ peptide associated with Alzheimer disease (24). In these respects, expanded CAG repeat diseases appear to conform to the unified model of protein misfolding disorders (25,26).
The two distinct features of Cks1, its glutamine-rich COOH terminus and the anti-parallel orientation of the monomers within the domainswapped dimer, prompted us to investigate Cks1 folding and aggregation and to compare it with that of other Cks proteins that we have previously analyzed. In particular, we were interested in determining whether aggregation occurs through the poly(Q) region or through the globular region (either by domain swapping or otherwise). We show here that, despite the sequence and structural differences from the other Cks proteins, the domain swapping properties of Cks1 are conserved within members of this family (27,28); by contrast, aggregation of Cks1 occurs by a route distinct from that of the Schizosaccharomyces pombe Cks protein, Suc1. Suc1 forms domain-swapped aggregates at neutral pH and room temperature, which subsequently convert to fibrillar aggregates. 9 For Cks1, we observe fibril formation within a few days at neutral pH and room temperature. These fibrils, however, consist solely of the COOH-terminal glutamine-rich sequence of Cks1, residues 112-150. Fibril formation is dependent on the proteolytic cleavage of the COOH terminus and its rate is greatly increased by the addition of a protease that cleaves the protein at a point in the sequence just before the glutamine-rich region. The results are consistent with a model for the origin of poly(Q) disorders and a subset of other protein misfolding diseases, in which the critical step is the production of a fibrillogenic sequence by proteolysis.

MATERIALS AND METHODS
High purity urea was obtained from Rose Chemicals Ltd. (United Kingdom). All other chemicals were obtained from Sigma unless stated otherwise.
Protein Expression and Purification-The coding sequence for Cks1 was amplified from a S. cerevisiae cDNA library by PCR with primers introducing 5Ј BamHI and 3Ј EcoRI restriction sites. The PCR product was cloned into a pRSET(A) vector (Novagen) modified for the expression of a His 6 NH 2 -terminal fusion protein with a thrombin cleavage site. Correct clones were identified by DNA sequencing (Oswell, University of Southampton). Site-directed mutagenesis (mutants C90V, C90S, P93A, E94P, P95A, and T 118 , the truncated variant at residue 118) was performed using the QuikChange kit (Stratagene) and was confirmed by DNA sequencing. C41(DE3) cells were transformed with the plasmid and a small number of colonies were picked from the plate into flasks containing 1 liter of FIGURE 1. Structure of domain-swapped Cks proteins. A, the relative orientation of the monomeric subunits is anti-parallel in the domainswapped dimer of the S. cerevisiae Cks1 (left) (3) and parallel in the domain-swapped dimer of human Cks2 (right) (6). The nature of the domain swap is also depicted schematically below the structure. B, the amino acid sequence alignment of the Cks1 proteins from human and yeast sources.
2TY and ampicillin media. The cells were grown at 37°C to A 600 ϳ 0.6 and then induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside overnight. Cells were harvested by centrifugation at 5000 ϫ g at 4°C for 15 min, and the pellet was resuspended in 50 mM sodium phosphate buffer (pH 7.5) containing 300 mM NaCl and 1 mM DTT. As Cks1 from S. cerevisiae contains one cysteine residue, 1 mM DTT was added to all buffers to prevent the formation of intermolecular disulfide bonds. Cks1 was purified by nickel-nitrilotriacetic acid affinity chromatography followed by size-exclusion chromatography. Protein samples were pure as judged by SDS-PAGE and mass spectrometry. The monomer and domain-swapped dimer forms of Cks1 were found to be well separated by gel filtration and were collected separately, as described previously for Suc1 (5).
Determination of Dissociation Constants for Domain Swapping-Determination of dissociation constants for the domain-swapped dimer of Cks1 was carried out as described previously (5). Briefly, wild-type or mutant Cks1 was incubated at 40°C at different protein concentrations until the fraction of dimer remained constant with time. Samples were quickly cooled on ice for 5 min and the proportions of monomer and dimer were determined by analytical gel filtration (Amersham Biosciences, S75 Sepharose) at room temperature in 50 mM Tris-HCl (pH 7.5), 300 mM, 1 mM EDTA, 1 mM DTT. The data reflect the monomerdimer equilibrium at 40°C as the samples do not re-equilibrate significantly at room temperature over the course of the experiment. A plot of the square of the monomer concentration versus the dimer concentration was fitted to a straight line, the slope of which gives the dissociation constant.
Transmission Electron Microscopy-Protein samples (at 1.8 mM concentration) were incubated for varying lengths of time at 25°C in 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mM DTT. For electron microscopy, samples were diluted 100 times and applied to Formvar-coated nickel grids. The blotted sample was negatively stained with 2% (w/v) aqueous uranyl acetate solution and examined in a Philips CM100 electron microscope operated at 80 kV.
Experiments with Protease and Protease Inhibitors-1.7 mM Cks1 was incubated either with 0.1 mg/ml trypsin or a protease inhibitor mixture (1 Complete tablet/100 ml, Roche Diagnostics) at room temperature. The samples were then analyzed by mass spectrometry and electron microscopy after 1 and 2 days, respectively.
Mass Spectrometry-Mass spectra were collected using a quadrupole time-of-flight-2 mass spectrometer equipped with a nanoflow Z-spray source (Micromass, Manchester, UK) as described previously (29).
Amino Acid Analysis-High-order oligomers of larger than 8 S formed during aggregation of Cks1 were sedimented in a Beckman Optima TLX ultracentrifuge. The supernatant was carefully removed from each sample and the pellet resuspended in water. The two protein fractions were hydrolyzed and analyzed using the standard ion-exchange ninhydrin procedure (30).
X-ray Fiber Diffraction-Fibrils were purified by sedimentation (as described above) and resuspended at 5-10 mg/ml. X-ray specimens were aligned using a stretch frame by air-drying these samples between two wax-filled capillary ends. Fiber diffraction images were collected on a R-AXIS IV image plate x-ray detector mounted on a Rigaku RU200 rotating anode source of CuK ␣ (1.5418 Å). Images were analyzed by the Java application "Crystal Clear" software.
Fourier Transform Infrared Spectroscopy-Samples were placed in a Bio-ATRCell TM II equipped with a silicon crystal. The cell was equilibrated at 25°C by using a water bath and infrared spectra were recorded with a Bruker Equinox 55 Fourier transform infrared spectrometer equipped with a liquid nitrogen-cooled mercury cadmium telluride detector. The sample compartment was continuously purged with dry air and 256 interferograms were co-added after registration with a resolution of 2 cm Ϫ1 . The buffer spectrum was recorded and subsequently subtracted from a protein sample using Protein Dynamics software (Bruker). Corrections for water vapor in the atmosphere were applied where necessary.

RESULTS
Domain Swapping Behavior of Cks1-To determine whether or not domain swapping leads to aggregation of Cks1, we first quantified the domain swapping propensity of the wild-type protein and of a number of mutants including a variant lacking the glutamine-rich COOH-terminal region. All measurements of domain swapping behavior were carried out in the presence of reducing agent so that the intrinsic propensity of the protein to domain swap could be investigated without the complicating factor of intermolecular disulfide bond formation. The interconversion between the monomeric form of Cks1 and its domainswapped dimer was found to be very slow at room temperature (several hours, depending on the protein concentration), as observed previously for the S. pombe Cks protein Suc1 (5) and therefore the K d was determined at 40°C to enable the equilibrium distribution of the two forms of the protein to be reached on a reasonable time scale. Wild-type Cks1 is fully folded at this temperature as it is well below its midpoint of denaturation (58°C from thermal unfolding monitored by CD and fluorescence, data not shown). The same situation holds for the mutants studied here (Table 1) as only small changes (1-2°C) in the midpoint of denaturation were observed. The proportion of monomer and dimer in each sample was determined by analytical gel filtration, carried out at room temperature so that the distribution of the two species is not perturbed during the time required for these experiments. The monomer and dimer are well separated in the elution profile ( Fig. 2A), and the K d was calculated by measuring the monomer and dimer content of samples at different protein concentrations. A plot of the dimer concentration versus the square of the monomer concentration is linear, indicating that equilibrium was effectively achieved for all protein concentrations investigated (Fig. 2B), and the slope of the line gives the K d values for the wild-type and mutant Cks1 dimer (Table 1).
To probe the domain swapping properties further, we mutated residues in the hinge loop (residues 91 to 99, sequence HAPEPHILL), the only region of the protein that adopts a different structure in the monomer compared with the dimer. We also investigated whether the glutamine-rich COOH-terminal region modulates the domain swapping propensity of Cks1. Dissociation constants of the domain-swapped dimer for wild-type and mutant Cks1 are given in Table 1 together with those previously measured for Suc1 for comparison. Substitution of the

Fibril Formation in Cks1
glutamate residue in Cks1 at position 94 by proline, thereby creating three sequential proline residues in the hinge loop, greatly increases the propensity for dimer formation. Moreover, substitution of one or the other of the proline residues at positions 93 and 95 by alanine also has a large effect on the dimerization propensity. The K d of P93A is much higher, 90 mM, and of P95A much lower, 27 M, than that of the wildtype protein (380 M). Mutants in which the cysteine residue, at position 90, was substituted by valine or serine show values of K d that are only slightly higher than the wild-type value. Most interesting is the observation that truncation of the protein at residue 118 preceding the poly(Q) region (T 118 ) destabilizes the dimer substantially relative to the monomer, increasing the K d to almost 1 mM. The Poly(Q) Sequence of Cks1 Is the Origin of Fibril Formation-The Cks1 monomer and dimer were incubated for several days at room temperature and under near physiological conditions (pH 7.5, 1 mM EDTA, 1 mM DTT) and a protein concentration of 1.8 mM. Samples from the solution were taken at regular intervals and examined using negative stain transmission electron microscopy (EM). The samples were found to form clear gels after several days of incubation, a phenomenon that is often indicative of aggregation. The EM images show that both monomer and dimer samples initially form amorphous aggregates, but long, unbranched fibrils appear after a few days (Fig. 3). The fibrils are typically 1-5 m in length and 20 nm in width. The morphology appears very similar whether the fibrils are formed from the monomeric form or the dimeric form of Cks1 (Fig. 3). Neither the monomeric nor the dimeric form of the truncated variant of Cks1 lacking the poly(Q) region (T 118 ) formed observable fibrils under the conditions used here, although amorphous aggregates were evident.
Proteolytic Release of the Poly(Q) Sequence Initiates Fibril Formation-The oligomerization state of Cks1 during the fibril incubation period was probed by analytical gel filtration. The chromatogram obtained for monomeric wild-type Cks1 at the beginning of the incubation period (Fig. 4A) shows a single peak with an elution volume of 16 ml. After 4 days of incubation, two further peaks appear at 14.5 and 17.6 ml. The former peak corresponds to the domain-swapped dimer. The latter peak was subsequently shown by mass spectrometry to correspond to a truncated monomer (M*) (see below). Some of the dimer molecules are also truncated, as is evident from the additional peak at 15.4 ml (D*) observed in the sample that was incubated for 10 days. The elution volumes of the truncated monomer and dimer correspond closely to those of the monomer and dimer of the T 118 variant (17.1 and 15.9 ml, respectively). Analysis by mass spectrometry (Fig. 6A) shows that cleavage of the protein takes place at residue Arg 111 , just prior to the poly(Q) region, where the most COOH-terminal trypsin site in the sequence is located (Fig. 1B). The fact that cleavage occurs in the samples may be attributed to trace quantities of proteases remaining after purification. No cleavage was observed for T 118 , even though the trypsin site at position Arg 111 is still present in this truncated variant. Possible explanations for this protease resistance are given under "Discussion." The mass spectrometry data also reveal that heterodimers composed of fulllength and truncated protein molecules were also present in the solutions after 1 day, as were oligomers of the poly(Q) fragment produced by proteolysis. Fig. 4B shows a plot of the total areas of the elution peaks from the truncated monomeric and dimeric species as a function of incubation time. The time courses of the proteolytic cleavage of the wild-type and mutants of Cks1 (Fig. 4B) reflect those observed for gel formation; the wild-type and P93A monomer form gels after 7 days of incubation, the E94P monomer after 12 days and the E94P dimer after 15 days. The results are consistent with the fibrils being composed of the cleaved poly(Q) region, and they indicate that the rate-limiting step of fibril formation under the conditions used here is the proteolytic reaction through which this region is generated.
Fibrils Formed in the Presence of Trypsin Consist Solely of the Poly(Q) Sequence-Fibril formation at a concentration of 1.7 mM Cks1 was monitored both in the presence of protease inhibitors and presence of   JULY 7, 2006 • VOLUME 281 • NUMBER 27 trypsin at 0.1 mg/ml. TEM analysis was carried out on samples incubated for 2 days at room temperature. Fibril formation was dramatically accelerated in the presence of trypsin, consistent with the hypothesis that proteolysis is the rate-limiting step in the aggregation process (Fig.  5). Heterogeneous mixtures of amorphous aggregates and fibrils are evident in the EM images taken of samples incubated in the presence of the protease inhibitors for 2 days (Fig. 5A). By contrast, protein precipitation occurred within a few hours in the presence of trypsin, and the EM images taken after 2 days showed a mass of well defined fibrils in the absence of amorphous aggregates (Fig. 5B). Mass spectrometric analysis of aliquots of the solutions confirmed that the globular Cks1 domain was fully digested by trypsin; peaks corresponding to the predicted m/z values of the COOH-terminal poly(Q) fragment (residues 112-150) were, however, absent (Fig. 6B). The latter would be consistent with a scenario in which the poly(Q) region aggregates rapidly and is not detectable in the mass spectrometer as a consequence of the highly heterogeneous molecular weight distribution that is associated with such aggregates. Amino acid analysis (Fig. 6C) of ultracentrifugation fractions shows that there is essentially complete separation of the digested Cks1 fragments lacking the poly(Q) region and the poly(Q) fragment itself into supernatant and pellet fractions, respectively. This finding shows clearly that the long and unbranched fibrils seen in the EM images (Fig. 5B) are formed from the poly(Q) region of Cks1.

Fibril Formation in Cks1
The dependence on protein concentration of the rate of poly(Q) fibril formation was monitored in the presence of 0.1 mg/ml trypsin. At a protein concentration of 1 mM, the solution was turbid after 2 days and the quantity of fibrils visible by EM did not increase further with time. By contrast, no fibrils could be detected by EM after 2 days of incubation of wild-type monomeric Cks1 at either 0.1 or 0.5 mM concentrations, although well defined fibrils were present in both samples after incubation for 12 days. Because cleavage by trypsin is complete within hours, we can conclude that there is a pronounced lag of several days in the formation of the fibrils from the poly(Q) fragments at submillimolar protein concentrations.
Biophysical Characterization of Poly(Q) Fibrils-Fibrils produced by incubation of Cks1 with trypsin at pH 7.5 remained intact and in solution below pH 3.9 and further biophysical analysis of the fibrils was carried out at pH 2.0. The x-ray diffraction pattern of the fibrils, measured following their alignment by means of a stretch frame, at this pH contained two main reflections (Fig. 7A). One is a 4.7-Å reflection on the meridian, a reflection commonly observed for amyloid fibrils and attributed to the spacing between strands in a ␤-sheet structure where the strands are oriented perpendicular to the fibril axis (31). This reflection is dominant and sharp, as expected because this spacing is essentially sequence independent and there is an effectively infinite array of well oriented strands in each fibril (32).
As well as the 4.7-Å meridional reflection, classical amyloid fibrils have an equatorial reflection at 10 -12 Å, corresponding to the distance between the arrays of ␤-sheets in the fibrils. The position of this reflection is dependent on the steric bulk of the component amino acid side chains (33), and it has been found to be weak in many samples of amyloid fibrils and absent in the diffraction patterns of poly(Q) and poly(N) fibers examined by Perutz et al. (34). The intensity of a reflection of this type depends on the number of ␤-sheets that are stacked together in each fibril (34) and it can also be weak because the fibrils are rotationally averaged around their long axis (32). In the diffraction pattern of the fibrils from the poly(Q)-rich COOH-terminal fragment of Cks1, this equatorial reflection, expected at 10 -12 Å is absent. Instead, we observe a weak equatorial reflection at 8.3 Å, which corresponds to a similar reflection reported by Perutz et al. (34,35). Fig. 7B shows the CD spectrum of full-length Cks1 under native conditions at pH 7.5 (dotted line) and poly(Q) fibrils from Cks1 at pH 2.0 (solid line). The CD spectrum of the fibrils from Cks1 at pH 2.0 is typical of that observed for ␤-sheet structures, and has no apparent contributions from any other type of secondary structure (Fig. 7B). The same observation has been reported previously for a poly(N) peptide (20) and the poly(Q) and poly(N)-rich regions of the Sup35 protein of yeast prions (20). The CD data are supported by Fourier transform infrared spectroscopy measurements that show a band at ϳ1645 cm Ϫ1 for fulllength Cks1 at pH 7.5 (Fig. 7C, dotted line). By contrast, a band at ϳ1622 cm Ϫ1 , typical of ␤-sheet structure, is obtained for the poly(Q) fibrils from Cks1 at pH 7.5 and 2.0 (Fig. 7C, solid lines). The amide I band in the fibril spectra obtained at pH 2.0 is further characterized by two well resolved bands at ϳ1608 and ϳ1656 cm Ϫ1 , which can be assigned to the C ϭ O stretching and NH 2 deformation vibrations of the glutamine side chains (36,37). The positions of these bands are slightly shifted to lower and higher wave numbers, respectively, compared with the positions observed in spectra of free glutamine (36,37). The shifts and high intensities of the absorption bands of the side chains indicate that the latter are involved in hydrogen bonding interactions. This conclusion is in agreement with previous findings by Krull et al. (38). Also in Fig. 7C is the amide I band at pH 7.5, which shows a higher absorbance at ϳ1640 cm Ϫ1 than at pH 2.0 in addition to the three peaks observed at pH 2.0. The ϳ1640 cm Ϫ1 peak is probably an absorption band of the peptide -NH 3 ϩ end groups and its increased intensity at higher pH values reflects the pH dependence of its extinction coefficient (36,37). Furthermore, absorption in this region of the IR spectrum is generally attributed to the disordered structure, although it is possible that non-fibrillar species contribute to this absorbance and are removed by lowering the pH and further purification of the sample. The smaller bandwidth at pH 2.0 is suggestive of an increased homogeneity of the vibrations, which is likely to reflect the fact that the fibrils remain in solution only at low pH.  Fig. 3). The elution peaks of monomer (M) and dimer (D) and their respective cleavage products (M* and D*) (confirmed subsequently by mass spectrometry) are labeled. B, cleavage time course shown as the total intensity of the elution peaks of the cleaved monomer M* plus 2/3 of the peak intensity of the cleaved dimer D* versus time. This takes into account that the elution peak D* consists of a mixture of fully cleaved dimers and partly cleaved heterodimers and assumes that they occur in a ratio of 1:2, respectively. Cytidine (elution volume 21 ml) was added to the protein solution at the start of the experiment as an internal standard to normalize the measured intensities.
A similar spectrum to that reported here at pH 7.5 has been observed for fibrils of ataxin-3 that contain a 78-residue glutamine repeat region (39).

Cks1 and Other Cks Proteins Show Similar Domain Swapping
Behavior-The orientation of the monomers in the crystal structure of the domain-swapped dimer of Cks1 is different from that observed for the dimers of the other Cks proteins (3,6,7,40,41) (Fig. 1). The Cks1 dimer is also unique in that it has a disulfide bond linking the two monomers. Despite these differences, mutations in the hinge loop alter the domain swapping propensity in similar ways in all three of the Cks proteins studied so far (5,28). Mutation to alanine of the first proline residue in the hinge loop (position 93 in Cks1) increases the K d of the dimer formation by several orders of magnitude, whereas mutation to alanine of the second proline residue in the hinge loop (position 95 in Cks1) decreases the K d significantly; it has been shown previously that the first proline residue creates strain in the hinge loop conformation of the monomer, and that the second proline residue creates strain in the dimer (5). Substitution of the glutamate residue in the hinge loop (position 94 in Cks1) by a proline, resulting in a repeat of three proline residues, decreases the K d of the dimer by several orders of magnitude relative to the wild-type protein. This effect has been rationalized previously in terms of the additional strain likely to be generated in the hinge loop of the monomer by such a sequence of residues (5). By contrast, mutation of the cysteine residue at position 90 has only a small effect on the K d , and the equivalent position in the other Cks proteins is occupied by a small hydrophobic residue, such as isoleucine or valine.
Interestingly, although the presence of the poly(Q) sequence does not alter significantly the free energy of unfolding of the protein (data not shown) it does decrease the domain swapping propensity by almost 3-fold. Poly(Q) sequences have been inserted in two proteins, chymotrypsin inhibitor 2 (42) and RNase A (43). Unlike Cks1, the insertions were in the middle of the proteins, in the hinge loop region, and the result was that domain swapping was promoted with poly(Q) sequences assembling from different molecules assembling together. In Cks proteins the swapped domain is the most COOH-terminal ␤-strand and the poly(Q) sequence is located COOH-terminal to this strand in Cks1; consequently, the poly(Q) sequence should not be conformationally constrained in either monomer or dimer forms and so should not affect the relative stability of the two forms. Moreover, the monomers adopt an anti-parallel arrangement according to the crystal structure of the Cks1 dimer and so the two poly(Q) regions extend from the globular dimer in opposite directions and would not be expected to interact. It is therefore not clear why the poly(Q) sequence has an effect on the domain swapping propensity of Cks1. . ESI-MS spectra and amino acid analysis of wild-type Cks1 after incubation at room temperature. A, Cks1 at a concentration of 1.7 mM was incubated for 2 days at pH 7.5. The spectrum is dominated by the charge series corresponding to full-length Cks1 (A series) and the cleaved Cks1 (B series) species. In addition, a heterodimer consisting of cleaved and uncleaved Cks1 and an oligomer consisting of 11 poly(Q)-rich COOHterminal peptides of Cks1 can be detected (series C and D, respectively). B, Cks1 was incubated in the presence of 0.1 mg/ml of trypsin for 24 h. All fragments predicted for the fully digested Cks1 protein were identified, and are indicated, except those consisting of the COOH-terminal poly(Q) region; by contrast, peaks corresponding to the molecular mass of the COOH-terminal poly(Q) peptide fragment (residues 112-150) are absent, consistent with their incorporation into large aggregates. C, aggregates Ͼ8 S from the trypsin-treated Cks1 solution as described in B were sedimented by ultracentrifugation. The compositions of the supernatant (gray columns) and pellet (black columns) were determined by amino acid analysis. The observed values fit well with the expected mole ratios (mole amino acid/mol protein) for each amino acid (expected values are indicated by black bars), assuming that the supernatant and the pellet consist of the digested fragments from the globular Cks1 domain (lacking the poly(Q) fragment) and the poly(Q)-rich region, respectively. The Poly(Q) Sequence of Cks1 Forms Fibrillar Aggregates following Proteolytic Cleavage-For Cks1, two mechanisms of self-association suggest themselves: the first involves the globular region and takes place via domain swapping and the second involves the glutamine-rich COOH-terminal region. The results in this study show that aggregation of the glutamine-rich residues can generate well ordered fibril structures at neutral pH and room temperature, whereas domain swapping in the globular region does not appear to generate ready formation of aggregates under the conditions explored in this study. At millimolar protein concentrations, proteolysis is the rate-limiting step in fibril for-mation and the rate of fibril formation can be greatly increased by the addition of trypsin. At submillimolar protein concentrations, a significant lag phase in fibril formation is observed, consistent with the nucleation mechanism that has been proposed for the formation of fibrils from poly(Q) sequences (13,44), and indeed more generally for amyloid fibrils (24).
In a study of poly(Q) peptides containing between 15 and 44 glutamine residues flanked by 2 lysines on each side, both the nucleation lag time and the critical protein concentration required for aggregation were found to depend strongly on the glutamine repeat length (13). At 10 M, the Gln 15 peptide exhibited a lag time of more than 10 days, whereas aggregation of the Gln 44 peptide occurred after only a few hours. Here we report that a 39-residue COOH-terminal fragment of Cks1, produced upon proteolytic cleavage and containing a stretch of 16 consecutive glutamine residues, does not form fibrils for at least 2 days when incubated at submillimolar concentrations. By contrast, at millimolar protein concentrations precipitation occurs within a few hours with well defined fibrils becoming visible after only 2 days. The poly(Q) stretch appears to remain in solution at millimolar protein concentrations, however, as long as it is attached to the globular domain in the context of the full-length Cks1. Proteolytic cleavage is therefore the critical first step in the overall aggregation process described here, as it results in a high local concentrations of a relatively short poly(Q) stretch that would otherwise have little tendency to aggregate. One question that arises from our results concerns the origin of the proteolytic activity that is observed to recognize the trypsin-like target site on Cks1 but is not affected by trypsin inhibitors (compare Fig. 3 showing EM of Cks1 alone with Fig. 5A showing EM of Cks1 incubated with inhibitor). Interesting in this regard is the observation that the truncated variant T 118 lacking the poly(Q) sequence is resistant to proteolysis although it still contains the cleavage site at position Arg 111 . There are two possible explanations. The first is that Cks1 itself, in combination with a poly(Q)rich COOH-terminal extension, has some intramolecular cleavage activity. This possibility is rather unlikely because Cks1 is not known to possess such activity against other proteins, nor is any other Cks protein. The second explanation is that a protease co-purifies only with the full-length protein because of an affinity for the poly(Q) region. Such an observation may be of significance with respect to poly(Q) diseases. Proteolytic cleavage of a poly(Q) sequence is a common first step in these diseases and it is possible that these positions within the respective proteins are susceptible to proteolysis because the poly(Q) sequence acts in some way as a recognition site for such a protease. Similar behavior may occur in the case of Cks1.
The Structure of the Poly(Q) Fibrils-Fibrils of proteins containing an elongated polyglutamine sequence have been shown to bind Congo red and exhibit green birefringence when examined under polarized light (11,39); these structures have been classified as amyloid fibrils on the basis of these properties. However, the molecular structure is unknown and several models have been proposed for the fibrillar structures formed by many different types of protein, including cross-␤ and ␤-helix structures. Perutz and co-workers (34) suggested the latter as a model for poly(Q) aggregates and its possible relevance to other amyloid fibrils has been discussed elsewhere (45,46). The ␤-helix model is based on an analysis of the x-ray diffraction pattern of Asp 2 -Gln 15 -Lys 2 in which the 10 -12-Å equatorial reflection, characteristic of the cross-␤ structure (31), is absent but an additional equatorial reflection at 8.3 Å can be observed; the origin of this latter reflection is still a matter of debate (35,47). In the present work we have obtained a diffraction pattern from fibrils derived from Cks1 that is similar to that of Asp 2 -Gln 15 -Lys 2 , and in addition, the infrared spectrum has extra features FIGURE 7. Characterization of poly(Q) fibrils from Cks1. Fibrils were produced at pH 7.5 after digestion of Cks1 with 0.1 mg/ml trypsin. The insoluble species were washed 3 times by centrifugation and resuspended in water (pH 7). To solubilize the fibrils, the pH was adjusted to a value of 2.0 by addition of HCl. A, x-ray fiber diffraction pattern. The meridional reflections originating from the inter-strand spacing (4.7 Å) and the unassigned equatorial reflection at 8.3 Å, previously observed specifically for poly(Q) and poly(N) fibrils, are labeled. B, CD spectra of poly(Q) fibrils from Cks1 (solid line) and 10 M full-length Cks1 under native conditions for comparison (dotted line). Fibrils were produced as described above at 10 mg/ml and were diluted 100-fold in 5 mM HCl. C, Fourier transform infrared spectra of poly(Q) fibrils (solid lines) and full-length Cks1 (dotted line) showing the amide I region at pH 2.0 (a) and 7.5 (b). The spectra have been baselinecorrected using a linear function only.

Fibril Formation in Cks1
compared with those of typical cross-␤ amyloid structures, such as found with insulin fibrils (48). Most amyloid fibrils contain mainly the ␤-sheet structure as indicated by a strong component in the Fourier transform infrared spectrum between 1638 and 1622 cm Ϫ1 (49), without significant absorption that is typical of turns or loops. By contrast, in the case of Cks1 fibrils the ␤-sheet band at 1622 cm Ϫ1 is accompanied by two almost equally intense bands at 1608 and 1656 cm Ϫ1 , which arise from the glutamine side chains. We have interpreted the intensity and position of the glutamine side chain bands in the amide I region of the IR spectrum in terms of hydrogen bonding. Two recent models for poly(Q) fibrils and the Q/N-rich Sup35 peptide (35,50), based on x-ray diffraction data, offer an explanation of how these interactions might arise. In these models the glutamine and asparagine side chains are interdigitated allowing tight packing. Moreover, these side chains form interamide hydrogen bonds along the axis of the fibril, and we suggest that these features could be the origin of the additional IR bands; they may also account for the 8.3 Å reflection and the higher order broad equational reflection at roughly 3.9 Å .
Cks1 as a Model System for Studying Mechanisms of Aggregation Associated with Disease-Naturally occurring poly(Q) sequences are found in a number of proteins, including several transcription factors. They are usually less than 25 residues in length, i.e. well below the human threshold of about 35 residues that is associated with human pathologies. Cks1 from S. cerevisiae is unique among the Cks family of proteins in having a glutamine-rich sequence at its COOH terminus. The remainder of the sequence, constituting the globular domain, is very highly conserved and it has been shown that the COOH-terminal region of Cks1 is not required for its normal function (3). It therefore seems unlikely that the glutamine-rich sequence plays any functional role. However, to date no studies have addressed whether proteolytic cleavage and/or aggregation of Cks1 occur in vivo.
Several studies indicate that proteolytic release of a fibrillogenic peptide is an important aspect of a number of protein misfolding diseases. Caspases and calpains, proteases important in apoptosis, have been shown to release the poly(Q)-containing fragment of the huntingtin protein and the Alzheimer A␤ peptide is released after proteolysis of the amyloid precursor protein by secretase enzymes (51). Also, it has been consistently observed that antibodies selective for the NH 2 -terminal region of huntingtin, which includes the poly(Q) region, but not those that recognize the middle or COOH-terminal regions of the sequence, bind to the proteinaceous inclusions in Huntington disease. There are multiple ways in which a poly(Q) peptide fragment might convert into a pathogenic form; for example, as well as self-associating, the cleaved peptide might co-aggregate with the uncleaved protein or with other poly(Q)-containing proteins. This latter effect could perhaps explain the toxicity associated with the aggregation phenomenon, for example, by transcriptional dysregulation (16,52), and indeed poly(Q)-containing transcription factors have been detected in neuronal inclusions (15,(53)(54)(55). Poly(Q) fragments may also interfere with other cellular processes such as ubiquitin-mediated protein degradation by the proteasome (56). It has recently been shown that the proteasome cannot cleave long sequences of poly(Q) residues (57), suggesting that the occasional failure of partially degraded products to exit the proteasome could perhaps obstruct its central channel, causing malfunction. The proteasome may also contribute to the aggregation process by releasing poly(Q) fragments with a greater tendency to aggregate than the fulllength poly(Q) proteins (57).
There are several problems inherent in characterizing the assembly mechanism of poly(Q) sequences: on the one hand, the peptides in isolation are intractable because they are insoluble even at low concen-trations; on the other hand, most of the proteins implicated in poly(Q) diseases are large and consequently make difficult subjects for biophysical analysis. More tractable systems are therefore needed to investigate the underlying aggregation events associated with poly(Q) sequences. Strategies suggested so far include fusing poly(Q)-rich peptides to a solubilizing tag or inserting the peptide within a small soluble protein (58 -60). In the present paper, however, we have demonstrated the controlled and tunable release of a poly(Q) fragment of a naturally occurring protein that subsequently associates into fibrils. We show in addition that highly ordered fibrils can be obtained from the relatively short poly(Q) sequence found in Cks1. The findings suggest that Cks1 is therefore an interesting system for unraveling further details of the molecular mechanism of poly(Q) aggregation.