Formation of Critical Oligomers is a Key Event During Conformational Transition of Recombinant Syrian Hamster Prion Protein

We have investigated the conformational transition and aggregation process of recombinant Syrian hamster prion protein (SHaPrP 90–232 ) by Fourier transform infrared spectroscopy, circular dichroism spectroscopy, light scattering, and electron microscopy under equilibrium and kinetic conditions. SHaPrP 90–232 showed an infrared absorbance spectrum typical of proteins with a predominant (cid:302) -helical structure both at pH 7.0 and at pH 4.2 in the absence of guanidine hydrochloride. At pH 4.2 and destabilizing conditions (0.3–2 M guanidine hydrochloride), the secondary structure of SHaPrP 90–232 was transformed to a strongly H-bonded, most probably intermolecularly arranged antiparallel (cid:533) -sheet structure as indicated by dominant amide I band components at 1620 and 1691 cm –1 . Kinetic analysis of the transition process showed that the decrease in (cid:302) -helical and the increase in (cid:533) -sheet structures occurred concomitantly according to a bimolecular reaction. However, the concentration dependence of the corresponding rate constant pointed to an apparent third order reaction. No (cid:533) -sheet structure was formed within the dead time (190 ms) of the infrared experiments. Light scattering measurements revealed that the structural transition of SHaPrP 90–232 was accompanied by formation of oligomers, whose size was linearly dependent on protein concentration. Extrapolation to zero protein concentration yielded octamers as smallest oligomers, which are considered as “critical oligomers”. The small oligomers showed spherical and annular shapes in electron micrographs. Critical oligomers seem to play a key role during the transition and aggregation process of SHaPrP 90–232 . A new model for the structural transition and aggregation process of the prion protein is described.


Introduction
The prion protein (PrP) 1 is -following the protein-only hypothesis -the sole agent causing a group of neurodegenerative disorders (1,2), the so-called prion diseases or prionoses (3). The most important ones among them are bovine spongiform encephalopathy (BSE) in cattle, scrapie in sheep, and Creutzfeldt-Jakob disease (CJD) in humans.
The crucial step in transmission and manifestation of prion diseases is the conversion of benign monomeric cellular prion protein (PrP C ), which has a mainly -helical secondary structure, to pathogenic multimeric scrapie prion protein (PrP Sc ), which is predominantly folded into -sheets (4,5). Noteworthy, PrP C and PrP Sc do not differ in their amino acid sequence.
Similar mechanisms play an essential role in a number of other neurodegenerative disorders including Alzheimer's, Parkinson's and Huntington's diseases. Therefore, the coupled processes of protein misfolding and aggregation, the kinetics of these processes, and the molecular species involved are of fundamental interest.
Late products of the conversion are amyloid fibrils and amyloid plaques, which are widely considered to be direct effectors of the above-mentioned disorders. However, evidence is accumulating that intermediates or by-products of the assembly process could be the pathogenic form of PrP (6,7) and other disease-related proteins (8). 5 isopropyl--D-thiogalactoside. After additional 5 h at 37 °C the cells were harvested by centrifugation at 2350 ´ g. The cell pellet was resuspended in 50 ml lysis buffer (100 mM Tris, 200 mM NaCl, 10 mM Na-EDTA, 0.2 % Triton-X-100, pH 7.2) and stirred for 30 min. 1 mmol MgCl 2 (final concentration: 19 mM) and 500 U Benzonase (Roche, Germany) were added. After additional 5 min the lysate was centrifuged at 10240 ´ g. The pellet was washed with 50 mM Tris, pH 7.4.
For purification PrP was extracted from cell lysate equivalent to 2 l of cell culture and reduced by resuspending the pellet in 40 ml reduction buffer (50 mM Tris, 6 M GuHCl, 50 mM DTT, pH 7.4).
After centrifugation (30 min at 25280 ´ g) and vacuum filtration through a 0.45 µm filter the protein solution was applied to 10 ml nickel-nitrilotriacetic acid Agarose (Qiagen, Germany), which was pre-washed in guanidine buffer (50 mM Tris, 6 M GuHCl, pH 7.4), and incubated for 30 min under shaking. Non-specifically bound proteins were removed by washing 3 times with guanidine buffer and once with guanidine buffer plus 1 mM Imidazol. The Agarose with bound PrP was filled into an "Econo-Pac" column (Biorad, USA). PrP was eluted by 20 ml guanidine buffer plus 250 mM Imidazol. To prevent formation of oligomers, eluted PrP was collected in 50 ml guanidine buffer. The concentration of PrP was determined by HPLC analysis (ET 250/4 6 During dialysis 10 U thrombin per mg PrP were added. Cleavage of the His 6 tag was allowed to take place overnight at room temperature. Completion of cleavage was checked by SDS PAGE. Due to the vector construction, the expressed and cleaved PrP contained N-terminally four additional amino acids (Gly-Ser-His-Met). The PrP solution was filtrated, 10 µl of protease inhibitor ('Complete, Mini, EDTA-free', Roche, Germany) were added, and the solution was concentrated to a final volume of 25 ml. After dialysis overnight against phosphate buffer (20 mM NaH 2 PO 4 , pH 7.0), PrP was analyzed by HPLC, sterile filtrated, and stored at -20 °C until final purification.
For final purification GuHCl was added to the pre-purified PrP samples to obtain a concentration of 6 M. 10-15 mg PrP were loaded on a EP 250/16 Nucleosil 300-7 C8 column (Macherey-Nagel). A solvent port had to be used to load the column with the sample. To prevent precipitation of the protein in the pump, the HPLC system was pre-washed thoroughly with 6 M GuHCl at a flow rate of 8 ml/min. Elution was performed by a linear gradient of 30-40 % eluent B in eluent A over 20 min at a flow rate of 10 ml/min. PrP containing fractions were pooled, lyophilized, resuspended in guanidine/phosphate buffer (20 mM NaH 2 PO 4 , 6 M GuHCl, pH 7.0) and dialyzed against the same buffer to remove residual TFA, which has a very intense C=O stretching band at 1680 cm -1 in the amide I region. Afterwards PrP was refolded by diluting 1:10 with phosphate buffer. After 1-2 h precipitated protein was removed by filtrating. The solution was dialyzed against phosphate buffer. For examination at pH 4.2, PrP was furthermore dialyzed against acetate buffer (20 mM sodium acetate, pH 4.2). Purity of the protein was determined by HPLC and SDS PAGE; identity of the protein was verified by mass spectroscopic analysis.

Stopped flow device
All measurements were performed using a novel stopped flow device, linked to an IFS 28/B FTIR spectrometer (Bruker Optics, Germany) equipped with a rapid scan option. The stopped flow device was designed specifically for high-precision FTIR kinetic and difference spectroscopy of biological macromolecules in 1 H 2 O, and is described in detail elsewhere (17).
Briefly, the principle elements of the system are a two-channel high pressure syringe pump, a microstructured diffusional mixer, and a specially designed thin-layer infrared flow cell. HPLC tubing and fittings are used throughout. Samples are injected into a continuous flow of distilled water via HPLC sample injection valves under computer control, and the measurement is triggered after the flow is stopped when the samples have filled the flow cell. For the measurements described in the present work, a flow rate of 3 ml/min and a sample loop volume of 15 µl were used. The dead time of the experiment, i.e. the time delay between mixing the samples and obtaining the first spectrum, was an average value of 190 ms. Kinetic measurements of the transition of PrP were performed by mixing GuHCl and PrP solutions in the stopped flow device, and the appropriate control experiments were performed with GuHCl or PrP alone. The particular advantages of this stopped flow device for the present work were the extremely precise and reproducible cell pathlength, the short time between sample and reference data acquisitions, and the very gentle mixing achieved with the diffusion micromixer. These benefits resulted in the unprecedentedly high quality of the FTIR spectra presented below.

FTIR parameters
The interferograms were recorded double sided (forward-backward) at a mirror frequency of 200 kHz. The upper and lower frequency folding limits were 7899 and 0 cm -1 , respectively. A Blackman-Harris 3 term function and a zero filling factor of 4 were used for Fourier 8 transformation resulting in spectra encoding approximately 1 data point per 1 cm -1 . The pathlength of the flow cell was approximately 8 µm. The non-linearity of the MCT detector was corrected prior to Fourier transformation.

Acquisition of protein spectra
The initial PrP concentration was 8 mg/ml. Since all PrP solutions were diluted 1:2 in the stopped flow device, measurements were performed at a concentration of 4 mg/ml (» 0.24 mM). The GuHCl-containing solutions were buffered by either 20 mM NaH 2 PO 4 or 20 mM sodium acetate to ensure proper pH values of 7.0 or 4.2, respectively. Steady state spectra of either buffer or PrP + buffer were measured independently 5 times with 256 scans in each case and finally averaged.
Buffer absorbance was subtracted from PrP + buffer spectra.
Time-resolved spectra of buffered GuHCl solutions and buffered PrP + GuHCl mixtures were measured in the rapid scan mode. All spectra were recorded continuously; the time lag between two spectra was determined by the number of scans per spectrum (see below). The acquisition processor of the FTIR spectrometer had a capacity to store 60 spectra. The early ones were recorded with a low number of scans to get spectra with high time resolution but low signal to noise ratio, the later ones were recorded with more scans to get spectra with a higher signal to noise ratio. Since the reaction kinetics of the -to--transition was a function of GuHCl concentration, the distribution pattern of scans per spectrum was adjusted as needed (Table 1).

Table 1, near here
The pure PrP spectra were obtained by subtracting GuHCl and buffer spectra. To obtain the timedependent changes, the difference spectra between each of the 60 single spectra and the last spectrum (i.e. the 60 th spectrum) were calculated. Due to small instabilities of the FTIR spectrometer, baseline shifts in the range of ± 0.0002 AU were observed between subsequent spectra and were offset corrected using the region from 1900 to 1750 cm -1 , which is free of 9 spectral features, as a reference. Second derivatives were calculated by applying the Savitzky and Golay algorithm with 9 smoothing points (18).
For evaluation of structural changes of PrP taking place within the experimental dead time, the difference between the first of the 60 spectra obtained after mixing PrP with GuHCl and a steady state spectrum of PrP in the absence of GuHCl was calculated.

Light scattering
SLS and DLS were measured simultaneously with one and the same instrument at a scattering angle of 90°. The laboratory-built apparatus, presently equipped with a diode-pumped, cw laser Millennia IIs (Spectra-Physics, USA) and a high quantum yield avalanche photodiode, has been described in detail (19). Apparent molar masses were estimated from the relative scattering intensities using toluene as a reference sample and applying a refractive index increment ( ¶n/ ¶c) = 0.19 ml/g. The translational diffusion coefficients D were obtained from the measured autocorrelation functions using either the program CONTIN (20) or applying the method of cumulants (21). The diffusion coefficients were converted into Stokes radii via the Stokes-Einstein equation R S = k B T/(6 0 D), where k B is Boltzmann's constant, T is the temperature in K, and 0 is the solvent viscosity. For kinetic light scattering experiments two experimental schemes were used. At high protein concentrations, prefiltered (100 nm pore-size) solutions and solvents were rapidly mixed in 100 µl fluorescence cells. At low concentrations, mixing was done before filtration into 30 µl flow-through cells.

CD measurements
CD measurements were carried out at protein concentrations between 0.2 and 1.2 mg/ml on a J-720 (JASCO, Japan) CD spectrometer, which was calibrated with (+)-camphorsulfonic acid at

Electron microscopy
Before preparation for electron microscopy, samples were diluted to a protein concentration of about 30 µg/ml with the corresponding solvent. Specimens were visualized by negative staining with 1 % uranyl formate using a double-carbon film technique. 400 mesh copper grids covered with a carbon-coated Triafol microgrid were used as support. Micrographs were taken with an EM910 electron microscope (LEO, Germany) at 80 kV and a magnification of 63,000.

Analysis of kinetic data
The time series of FTIR spectra between 1200 and 1900 cm -1 was analyzed by singular value decomposition (SVD) using the implemented SVD routine of the program package "Mathematica" (version 4.2, Wolfram Research, USA). A fundamental question concerning the evaluation of the kinetic data, particularly in the present case, is the reaction order of the process under study. Since aggregation is involved in the transition process, an apparent higher order reaction must be taken into consideration. First order reactions follow exponential functions , while higher order reactions are characterized by a power law of the form (23). This plot is strictly linear for n = 2, allowing to estimate k from the slope of the data, and shows characteristic curvatures for n = 1 and higher order reactions. Both FTIR and CD kinetics were treated in this way. In special cases (n > 1) the kinetics were fitted directly by a power law function with arbitrary n.
This enables one to calculate the reaction order n from the concentration dependence of the apparent reaction rate

Results
Large quantities of recombinant PrP (10-20 mg per liter of E. coli culture) were obtained in > 99 % purity as proven by SDS PAGE, HPLC, and MS (data not shown) due to the two-step purification procedure. The purified PrP was quantitatively oxidized as proven by HPLC (data not shown). Only very slight differences were observed between the FTIR spectra of hamster and human PrP at pH 7 (data not shown). This is not surprising, because both proteins have a high sequence homology (87 % of the amino acids are identical) and exhibit almost identical three dimensional by guest on March 23, 2020

Steady-state FTIR spectroscopy
http://www.jbc.org/ Downloaded from 13 structures as was proven by NMR spectroscopy (29,30). All experiments presented in the following were carried out using hamster PrP.
We have reexamined unfolding of PrP C at room temperature under the influence of chaotropic agents using FTIR spectroscopy. Previous investigations by CD spectroscopy (see (31)   This species turned out to be a multimeric state of PrP, which could be unfolded by high concentrations of denaturant, but which was stable in the absence and at low concentrations of denaturant. The kinetics of formation and the structural properties of this -rich, multimeric state were the subject of the investigations described below.

Kinetic FTIR spectroscopy
Since 1.0 M GuHCl at pH 4.2 appeared to be optimal to study the transition process, we investigated the structural changes of PrP under these conditions in greater detail. To kinetically detect these changes, difference spectroscopy was applied as described in the Experimental Procedures section.
This time dependence is consistent with a second order reaction. A further helpful visual test concerning the order of the apparent reaction rate is to plot Fig. 3B). Such a plot must be linear for a second order reaction, while remarkable and characteristic deviations from linearity are expected for both first and higher order reactions (see and 2, respectively. The nearly identical spectra confirmed the FTIR results (Fig. 1B) that the secondary structure was essentially unchanged on the transition from pH 7 to pH 4.2. The spectra had a pattern that is typical of a protein that has predominantly helical secondary structure.
After a GuHCl concentration jump from 0 to 1 M at pH 4.2, the CD spectrum changed with a strongly concentration-dependent rate. The spectrum after complete transition (Fig. 4, curve 4) had the characteristic features of proteins with -rich secondary structure. It is an important question, whether this -to-transition proceeded already in the monomeric state or was entirely coupled with the aggregation process detected, e.g., by light scattering (see below). Therefore the first CD spectrum measured 1 min after mixing with GuHCl at low protein concentration (0.27 mg/ml) is also shown (Fig. 4, curve 3). The influence of aggregation should be negligible at this time, because SHaPrP was expected to be still in the monomeric state. This was confirmed by size exclusion chromatography (SEC, data not shown). This spectrum (Fig. 4, curve 3) was clearly distinguishable from both the spectrum in the absence of GuHCl (curve 2) and that after transition has been completed (curve 4). The pattern of the spectrum was, however, more similar to that of the initial helical state. where the second smaller peak originated from the presence of small amounts of aggregates. The ratio of the peak areas could be used to separate the contributions of monomers and aggregates to the static light scattering intensity. Extrapolating light scattering from monomers to zero protein concentration yielded molar masses M of (16,000 ± 2,000) g/mol and (18,000 ± 2,000) g/mol at

Fig. 8, near here
In order to determine the morphology of the growing particles, small amounts of this sample were withdrawn 25 days (3.6×10 4 min) and 55 days (7.9×10 4 min) after starting the transition and subjected to electron microscopy (Fig. 8). Fig. 8A demonstrates that the particles formed during the first kinetic phase had mostly spherical or annular structures with diameters of 10-15 nm (arrows and inset in Fig. 8A). Moreover, some larger spherical particles with diameters of about 19 20 nm and a few fibrillar structures can be seen. Comparison of Fig. 8A and Fig. 8B shows that the increases in mass and Stokes radius at later times can be attributed to the length growth of curved fibrillar structures.

Discussion
The transition of different recombinant prion protein sequences into -rich, aggregated structures has been studied by several groups (9)(10)(11)31,37,38). The consistent finding of the latest studies is the appearance of a particular -rich oligomer during the assembly process. At present it is not fully established, whether the oligomer itself is a toxic species, an on-pathway precursor of later stages of aggregated PrP, or an off-pathway by-product of the assembly process. Since the oligomer was formed by the full-length-protein (11) as well as by truncated sequences (37)  more detailed information about these secondary structure changes, the size distribution and morphology of the relatively stable oligomers, the kinetics of the transition reaction, particularly its concentration dependence, and the sequence of steps involved in this process.

FTIR spectroscopy
To elucidate the transition process of recombinant PrP on the level of secondary structure in more detail, the use of FTIR spectroscopy was a very appropriate technical approach. In this study FTIR spectra of high quality as shown in Figs. 1 and 2 (33,(39)(40)(41). The use of the specially designed micro stopped-flow device in combination with FTIR rapid scan techniques enabled us to monitor the --transition (which is relatively fast at protein concentrations used in this study) in real time with sufficiently high time-resolution and excellent reproducibility. The small structural changes which occurred within the dead time of the experiments will be discussed later.
Up to now the --transition of recombinant PrP was studied mainly by CD spectroscopy (13,34,37,38,42). The advantage of FTIR spectroscopy is that bands specific for -helices, - The strong tyrosine band at 1517 cm -1 did not shift during the partial unfolding or transition process of PrP. However, when PrP was unfolded completely by 6 M GuHCl at pH 7, the band shifted to 1516 cm -1 (data not shown). A shift in this range is significant and typical for un-and refolding processes of proteins (46,47). This indicates that the alterations of the secondary structure during the transition process did not change the micro-environments of the tyrosine residues as strongly as a complete unfolding procedure did. This observation is especially

Aggregation process
The aggregation process accompanying the secondary structure transition was studied by SLS and DLS at concentrations between 0.2 and 2.9 mg/ml. At the highest concentration, which is comparable to that used for the FTIR experiments (4 mg/ml), the initial aggregation process was 1. The average mass of the oligomers at the transient maximum was found to depend linearly on protein concentration (Fig. 7B). This result possibly explains why different sizes of the oligomer have been reported in the literature (10,11).
2. Linear extrapolation of the transient maximum value to zero protein concentration (Fig. 7B) yielded a relative mass of 7.9 ± 1.0. This leads to a stoichiometry of 8 monomers forming a stable misfolded oligomer. Accordingly, mostly octamers were formed at low protein concentrations.
Additionally, the larger oligomers formed at higher concentration showed a tendency to disaggregate at later times ("overshoot", Fig. 7A). Thus, the octamers appear to be the most stable species and will be termed "critical oligomers" according to similar observations during the assembly process of other proteins towards fibrillar structures (48). The oligomers appeared to be stabilized by intermolecular -sheets with strong hydrogen bonds according to our FTIR data. Particularly, the detection of these octamers stabilized by strongly hydrogen bonded - were not possible due to the overshoot phenomenon at high concentrations (Fig. 7A). The slope s = 2.0 ± 0.2 in Fig. 7D is consistent with an apparent reaction order of 3. Combining this result with those obtained from the kinetic spectroscopic data, our estimations of the reaction order fall into the range between 2 and 3. In any case, we find a reaction order smaller than the apparent fifth order obtained by Baskakov et al. (10), when the -oligomer was formed at 5 M urea starting from the unfolded state at 10 M urea.
5. The spherical shape of the rapidly formed aggregates was also confirmed by electron microscopy (Fig. 8). Closer inspection of the morphology reveals annular shapes, particularly among the smaller particles. It is conceivable that the most stable octamer consists of a ring structure of 8 monomers.
6. The rapidly formed oligomers were able to assemble further during a second, much slower process (Fig. 7E) in a linear fashion leading to curved protofibrillar structures, as can be judged from electron micrographs (Fig. 8). Protofibril formation seems to be a side reaction of a possible subpopulation of the oligomers, since the amount of oligomers was only partly reduced even after observation times of about 2 months. Furthermore, the growth rate of a sample with fourfold higher concentration was only slightly higher as compared to that shown in Fig. 7E. Straight mature fibrils were not observed under the conditions used in this study. Therefore we suppose that other structures larger than the protofibrils observed cannot be obtained without further rearrangement of the basic internal structure of the oligomers. However, the oligomers may provide an ideal pool of pre-aggregated material, enabling further cooperative structural transitions under modified external conditions, e.g. at elevated temperatures. 26 7. Up to now we failed to detect any smaller oligomers than the critical octamers, particularly by measuring SEC chromatograms at different times using protein concentration of 0.08 mg/ml (data not shown). Similar conclusions can be drawn from the SEC data reported by others (10,38). On the other hand, the measured apparent reaction order 3 and the apparent bimolecular step involved in the changes of the spectroscopic properties demand the transient population of smaller oligomers such as dimers or tetramers.
A possible model for the aggregation process

Fig. 9, near here
In the following, we will summarize the most important data of our study and previous findings in a model shown in Fig. 9. Upon the change from native environmental conditions to those favoring the transition (1 M GuHCl, pH 4.2) PrP C is transferred into the helical state '. In the 'state, PrP C is destabilized and has a somewhat less helical structure and an enhanced tendency to aggregate. The subsequent steps are presently still not resolved very well and are buried within the "gray box" on the pathway from ' to the oligomer. However, they must fulfil some kinetic requirements. Furthermore, we have indications from spectroscopic data and SEC that the intermediate states involved in these steps are only extremely weakly populated. Presumably, all reactions steps on the left side of B in Fig. 9 are reversible. However, once a critical oligomer size with its specific stabilizing interactions is reached, the backward reaction constant becomes very small, making the transition on the right hand side of B in Fig. 9 to an irreversible one. The critical oligomer size is 8 monomers in our case. The linear concentration dependence of the oligomer size is indicative of additional bimolecular steps leading to multimers of octamers. The slow disintegration of larger oligomers ("overshoot" phenomenon) cannot be explained adequately at present.
Though we cannot resolve the details within the "gray box" in Fig. 9, our kinetic data and the observed stoichiometry pose some constraints on the possible mechanisms. The apparent bimolecular reaction (n = 2) revealed by the kinetic traces of our optical data together with the apparent third order reaction consistent with the concentration dependence of the reaction rates (2.6 < n < 3) can be explained consistently by assuming two consecutive bimolecular steps with a weakly populated intermediate. At least one additional reaction step is required to fulfil the stoichiometric constraints to build up the critical octamer. A plausible scheme is shown in Fig. 9, however until now no clear signature of this particular scheme was found in our kinetic data.
The observed kinetic features of the reaction leading to misfolded -octamers are in marked contrast to proposed mechanisms of misfolding and amyloid formation (49)(50)(51). The involvement of oligomers larger than dimers in conformational transition rules out the heterodimer mechanism (49). The secondary structure transition according to a bimolecular reaction contrasts with the postulated unimolecular reaction of conformational transition of unstructured oligomers in the socalled nucleated conformational conversion mechanism (50). The involvement of the -sheet rich octamer in a sequence of bimolecular steps leading to larger multimers cannot be explained assuming a nucleated polymerization mechanism (51). The exclusion of a nucleated polymerization mechanism is further supported by the absence of any lag phase. These considerations lead to the conclusion that a distinctive, in its details so far unknown mechanism must be responsible for the observed transition process.

Toxicity of Oligomers
Whether the oligomers and particularly the critical octamers are themselves a toxic species, is still an open question. Such a role could be inferred from observations that small oligomers of the A peptide are toxic at early stages of Alzheimer's disease (52). A possible toxic role was also postulated for small aggregates of -synuclein. Volles and Lansbury (53) have shown that small 28 annular aggregates of -synuclein enhance the permeability of model membranes by a pore-like mechanism. Note that annular structures have been observed for the critical oligomers of PrP         Tables   Table 1: Distribution pattern of scans per spectrum for all measured GuHCl concentrations. Since no more than 60 spectra could be stored for each rapid scan measurement, the number of scans and thus the time of measurement (larger numbers of averaged scans result in longer times of measurement) for each of the 60 spectra had to be chosen in order to observe the whole reaction with an appropriate time resolution.