Entrapping intermediates of thermal aggregation in alpha-helical proteins with low concentration of guanidine hydrochloride.

Aggregation of proteins is a problem with serious medical implications and economic importance. To develop strategies for preventing aggregation, the mechanism(s) and pathways by which proteins aggregate must be characterized. In this study, the thermally induced aggregation processes of three alpha-helix proteins (myoglobin, cytochrome c, and lysozyme) in the presence and absence of 1.0 m guanidine hydrochloride (GdnHCl) were investigated by means of infrared spectroscopy. In the absence of GdnHCl, intensities of the alpha-helix bands (approximately 1656 cm(-1)) decrease as a function of temperature at above 50 degrees C. With myoglobin and cytochrome c, the loss of helix bands was accompanied by the appearance of two new bands at 1694 and 1623 cm(-1), indicative of the formation of intermolecular beta-sheet aggregates. For lysozyme, bands indicative of intermolecular beta-sheet aggregates did not appear in any significant intensity. In the presence of 1.0 m GdnHCl, two major intermediate states rich in 3(10)-helix (represented by the band at 1663 cm(-1)) and beta-turn structure (represented by the band at 1667 cm(-1)), respectively, were observed. These findings demonstrated that IR spectroscopic studies of protein aggregation using a combination of thermal and chemical denaturing factors could provide a means to populate and characterize aggregation intermediates.

Protein folding is arguably the most important process studied in biophysics and structural biology because it converts linear polypeptide chains into three-dimensional structures that endow proteins with all their vital activities (1)(2)(3). Studies of protein folding are often plagued competing, off-pathway aggregation processes. Aggregation of proteins is also a problem with serious medical implications, e.g. in human disease states like Alzheimers disease (4), Parkinsons disease (5,6), and monoclonal immunoglobulin amyloidosis (7,8).
Furthermore, protein aggregation during production, shipping, storage and delivery of therapeutic proteins is a problem of significant economic importance (9)(10)(11). To develop strategies for preventing protein aggregation, the mechanism(s) and pathways by which proteins aggregate must be characterized. Such characterization is complicated because light scattering interferes with many of the optical techniques that are now standard for examining protein folding pathways, e.g., fluorescence and circular dichroism spectroscopies.
In contrast, infrared spectroscopy (IR) 1 is insensitive to light scattering, and thus provides a valuable method for studying protein aggregation. IR spectroscopy can be used to study not only the secondary structure of proteins in the soluble, native state (12-15), but also in precipitated states, both native (e.g. salted out) and denatured (e.g, thermally, chemically, or mechanically induced aggregates) (16)(17)(18)(19)(20). However, a shortcoming of IR spectroscopy is that relatively high protein concentrations (e.g. 20 mg/ml in H 2 O) are needed to obtain high quality spectra. With such high concentrations of protein, perturbations used to induce structural transitions leading to aggregation (e.g., high temperature) cause a rapid conversion of native protein to insoluble aggregates that are rich in intermolecular β-sheet (16-18, 21, 22). This conversion is so rapid with respect to typical IR spectral acquisition (e.g. 5 minutes) that intermediates in the aggregation pathway usually cannot be detected. Yet, current theoretical and experimental mechanisms for protein aggregation suggest that aggregates are formed from partially folded intermediates (e.g., 23). Thus, what is needed to capitalize on the advantages of IR spectroscopy for the study of protein aggregation is a means of populating these folding intermediates sufficiently for IR spectroscopic characterization.
To address this issue, in the current study we combined thermal and chemical approaches to allow population of aggregation pathway intermediates. With this approach, the model protein pool for the unfolding studies may be expanded to include proteins with a wide range of secondary structural compositions and various degree of resistance to chemical denaturants.
Here, we present the results of studies in which we compared unfolding and aggregation of three α-helix predominant proteins (myoglobin, cytochrome c, and lysozyme) in the presence or absence of 1.0 M gdnHCl using IR spectroscopy. A 1.0 M gdnHCl concentration is chosen because earlier studies have shown that at this concentration of gdnHCl and room temperature, both myoglobin and cytochrome c are not unfolded (24,25). Here we will show that elevation of the temperature above 25 °C allows partial unfolding to occur, while the presence of gdnHCl both lowers the temperature at which transitions occur and inhibits the conversion of these partially folded intermediates to aggregates.

MATERIALS AND METHODS
Protein Sources and Preparations. Cytochrome c (type VI, horse heart), myoglobin (horse heart), and lysozyme (chicken egg white) were purchased from Sigma Chemical Co. at which a spectrum was acquired was controlled within 0.5°C using a custom-built Peltier IR cell temperature controller. Spectral acquisition at a given temperature required approximately 6 minutes (i.e., dwell time at the given temperature). The average heating rate between spectral acquisition temperatures was 1.5 °C/min. Protein spectra were obtained according to previously established criteria and double-subtraction procedure (15,27). For the best result the spectra of buffer and 1.0 M gdnHCl/buffer were subtracted from the spectrum of protein separately. The second-derivative spectra were obtained with a 7-point Savitsky-Golay derivative function and then baseline corrected as previously described (17).

RESULTS
Thermally-Induced Structural Transitions. Figure 1 shows the infrared absorbance spectra of myoglobin, cytochrome c, and lysozyme in 50 mM potassium phosphate buffer (pH 7.2) as a function of temperature. The two strong bands centered near the 1656 and 1548 cm -1 are the so-called the amide I and II bands, respectively. The amide I band arises primarily from the C=O stretching vibration of the peptide linkages that constitute the backbone structure of proteins and is known to be sensitive to the secondary structural composition and conformational changes of proeins (12, 15, 28). The amide II band arises mainly from an out-of-phase combination of N-H in-plane bending and C-N stretching vibrations of peptide linkages (28) and is less useful in protein structural analysis. At 25 °C all three proteins exhibited the amide I band maximum near the 1656 cm -1 , a frequency characteristic to proteins containing predominantly α-helical structures (27). Elevation of temperature to ca. 80 °C resulted similar spectral changes in all three proteins. The amide I absorbance maximum near 1656 cm -1 decreased as a function of temperature, accompanied by intensity increase at 1623 and 1694 cm -1 . In addition, a temperature-dependent intensity decrease and frequency red-shift at the amide II bands were observed for all three proteins. The changes at the amide II region seem to be nonspecific to thermally-induced protein unfolding and aggregation, because similar changes were also reported for the recombinant human factor XIII, a predominantly β-sheet protein (29).  (17,31,32). The latter become a predominant spectral feature of thermally aggregated states of myoglobin and cytochrome c.
When cooled from 80 °C to 25 °C, the two β-sheet aggregate bands remain unchanged in their intensity, but shifted ca. 2 cm -1 to a lower wavenumber (data not shown). In addition to the appearance of IR bands assigned to intermolecular beta sheet, the presence of aggregated protein in the samples was documented by the observation that the cooled samples formed gels, which are characteristic extensive intermolecular interactions in protein samples.
For lysozyme, bands indicative of intermolecular β-sheet do not appear in any significant intensity, even though the pronounced loss of native α-helix is irreversible (data not shown).
The lack of well-resolved bands suggest that the structure of thermally perturbed lysozyme is comprised of a heterogeneous ensemble of proteins with non-native β-turns, extended strands, and residual α-helix.   (27). Differing from the α-helix in that the hydrogen bond is of the 4→1 type rather than the 5→1 type, 3 10 -helix is also less common in proteins than αhelix (34). Nevertheless, 3 10 -helix structure has been found by X-ray crystallographic analysis, for example, in peptides containing α-aminoisobutyric acid residues (35,36) and in globular proteins such as α-lactalbumin (31% α-helix and 20% 3 10 -helix) (37). By examining the IR spectra of synthetic peptides containing α-aminoisobutyric acid residues, Kennedy and colleagues (33) showed that the peptides have an amide I band maximum between the 1666 and 1662 cm -1 . Prestrelski and co-workers (38) reported that the deconvoluted spectrum of α-

DISCUSSION
The thermally-induced aggregation processes of the majority of the proteins studied by FT-IR spectroscopy can be described with a two-state model. The predominant secondary structural element (α-helix or β-sheet) decreases as a function of temperature and is concomitantly replaced by intermolecular β-sheet due to protein aggregation (16-18, 22, 39). The latter is evident by the appearance of a strong band near 1624±8 cm -1 (low-wavenumber β-sheet component) accompanied by a weak band near 1693±5 cm -1 (high-wavenumber β-sheet component). Intermolecular β-sheet structure is a common secondary structural element in the aggregated state of proteins (e.g., 16-18 22, 31, 32, 39). The formation of intermolecular βsheet structure in thermally-induced protein aggregates was clearly demonstrated by Clark and colleague (40) using small-angle x-ray scattering. Later, they reported a close relationship between the thermally-induced intermolecular β-sheet aggregates and appearance of a new, well-defined amide I band component near 1620 cm -1 (31). More recently, Damaschun et al.
(32) studied fibrils of phosphoglycerate kinase with small-angle x-ray scattering and provided similar corroboration for assignment to the high and low wavenumber IR bands noted above to intermolecular β-sheet. Our data for the thermally-induced structural transitions in myoglobin and cytochrome c in the absence of gdnHCl suggest a direct transition from the native conformation to intermolecular β-sheet aggregates, evidenced by the isobestic points visible for both proteins (Figures 1, 2).
Lysozyme in the absence of gdnHCl also shows isobestic evidence of a two-state transition (Figures 1 &2). The amount of β-sheet aggregate present in the thermally perturbed state of lysozyme is negligible, however, judging by the weak band near 1625 cm -1 (lowwavenumber β-sheet component). The major secondary structural elements in the thermally perturbed state of lysozyme are represented by two broad bands at 1657 and 1685 cm -1 , assignable to the residual α-helix and β-turn structure, respectively (15).
It is noteworthy that the random (unordered) structure observed in the thermallyaggregated state of other proteins (16-18, 22, 39) is not a major element in the thermally perturbed state of the three α-helix proteins studied here. Unordered structure is generally associated with an amide I component at 1645±4 cm -1 for proteins in D 2 O solution (12) and 1648±2 cm -1 for proteins in H 2 O solution (15). The latter assignment is supported by the IR spectrum of the model compound poly-L-lysine in H 2 O solution at neutral pH (38), in which the polypeptide is known to be a random coil (41)(42)(43). The lack of random coil as a major structural element in the thermally-aggregated state has also been reported for the β-sheet predominant protein, recombinant human Factor XIII (29). We should point out that the so-called loop structure, which may be classified as part of random structure, could also contribute to the band intensity near 1658 cm -1 (14,44,45). Previous IR spectroscopic studies on superoxide dismutase have also shown that the random/loop structure is represented by a part of the 1658 cm -1 band (accounts for 1/3 of random/loop), in combination with a more prominent band at 1647 cm -1 (accounts for 2/3 of random/loop) (44). Furthermore, with existing IR spectroscopic data on a large number of proteins in H 2 O solution (30), a protein with a significant amount of random/loop structure that does not exhibit a major band in the 1648±2 cm -1 has not been observed. The spectra of the thermally perturbed states of proteins in the current study do not have a band in 1648±2 cm -1 region of their IR spectra. Thus, the remaining bands near 1656 cm -1 in their spectra are more likely associated with the residual α-helix structure than with newly formed unordered/loop structure.
Addition of 1.0 M gdnHCl in conjunction with elevated temperatures results in significant accumulation of non-native species rich in 3 10 -helix and β-turn structures, which we suggest are intermediates between the native protein and aggregated states for myoglobin and cytochrome c. Similar structures are seen in lysozyme, although intermolecular β-sheet aggregates are not formed in the temperature range we tested. It is worth noting that similar spectral features have been previously observed in gdnHCl-induced unfolded state of iso-1cytochrome c (46). Bowler and co-workers (46) reported that as concentration of gdnHCl increased, the intensity of the 1657 cm -1 band of iso-1-cytochrome c due to α-helix structure was lost gradually and replaced by three main features at 1687, 1666 and 1660 cm -1 , ascribable to β-turn, β-turn, and 3 10 -helix, respectively. Furthermore, Millhauser (47) has proposed a thermodynamic folding pathway for helical peptide: random coil ⇔ nascent helix ⇔3 10 -helix ⇔ α-helix, after examining the equilibrium of 3 10 -helix/α-helix from the perspective of crystallographic studies and spectroscopic data from double label electron spin resonance, nuclear magnetic resonance, and circular dichroism spectroscopies. In addition, molecular dynamics simulations have suggested that 3 10 -helice exist as a kinetic folding intermediate of analogous α-helical proteins (48,49). Aggregation pathway intermediates containing 3 10 -helix and β-turn structures that we observe may also be the intermediates on a folding pathway. For many proteins, it has been found that aggregates are formed from intermediates on the folding pathway (reviewed in 23).
We have demonstrated in the present study that IR spectroscopic investigations of protein aggregation using a combination of thermal and chemical denaturing factors can provide a means to populate and characterize aggregation intermediates. This method should be valuable for studying the aggregation processes of a wide range of proteins. We speculate that identification and characterization of aggregation intermediates may lead to new interdiction strategies for amyloidogenic human diseases, as well as to improvements in industrial processing, storage and delivery of therapeutic proteins.