Amyloid-like Fibril Formation in an All β-Barrel Protein

Acidic fibroblast growth factor from newt (Notopthalmus viridescens) is a ∼15-kDa, all β-sheet protein devoid of disulfide bonds. In the present study, we investigate the effects of 2,2,2-trifluoroethanol (TFE) on the structure of newt acidic fibroblast growth factor (nFGF-1). The protein aggregates maximally in 10% (v/v) TFE. Congo red and thioflavin T binding experiments suggest that the aggregates induced by TFE have properties resembling the amyloid fibrils. Transmission electron microscopy and x-ray fiber diffraction data show that the fibrils (induced by TFE) are straight, unbranched, and have a cross-β structure with an average diameter of 10–15 Å. Preformed fibrils (induced by TFE) of nFGF-1 are observed to seed amyloid-like fibril formation in solutions containing the protein (nFGF-1) in the native β-barrel conformation. Fluorescence, far-UV CD, anilino-8-napthalene sulfonate binding, multidimensional NMR, and Fourier transformed infrared spectroscopy data reveal that formation of a partially structured intermediate state(s) precedes the onset of the fibrillation process. The native β-barrel structure of nFGF-1 appears to be disrupted in the partially structured intermediate state(s). The protein in the partially structured intermediate state(s) is found to be “sticky” with a solvent-exposed non-polar surface(s). Amyloid fibril formation appears to occur due to coalescence of the protein in the partially structured intermediate state(s) through solvent-exposed non-polar surfaces and intermolecular β-sheet formation among the extended, linear β-strands in the protein.

Protein aggregation is a problem of importance not only in biotechnology but also in health-related industries (1,2). Globular proteins in aqueous solution often tend to aggregate under a variety of conditions of concentration, temperature, pH, and ionic strength (3)(4)(5)(6). The morphology of aggregates formed varies considerably and ranges from amorphous forms to highly structured fibrils (7)(8)(9). Structured fibrils formed in vitro closely resemble the highly organized amyloid fibrils found in association with a variety of human disorders, including Alzheimer's disease, Creutzfeldt-Jakob disease, Huntington's disease, and type II diabetes (10 -19). Several studies show proteins that are apparently unrelated in sequence and, in their native conformation, aggregate into fibrils that have characteristic amyloid-like structural and histological features (16, 20 -26). Recently, Bucciantini et al. (27) demonstrated that amyloid-like fibrils of SH3 domain (from bovine phosphatidylinositol 3-kinase) induced in vitro in 25% (v/v) 2,2,2-trifluoro ethanol (TFE) 1 (under appropriate conditions) are cytotoxic to fibroblast NIH3T3 cells. The amyloid fibrils generated ex vivo are observed to seed fibrillate in cultured NIH3T3 cells (23). Therefore, it is now increasingly believed that amyloid represents a generic form of polypeptide conformation, and all peptides/proteins have the potential to form amyloid-like fibrils under appropriate conditions (21)(22)(23)(24)(25)(26)(27).
The molecular mechanism underlying the amyloid fibril formation is poorly understood. Recent studies on proteins such as transthyretin (16,25), lysozyme (26), ␣-synuclein (28), ␣-microglobulin (29), immunoglobulin light chain variable domain (30) suggest that amyloid fibril formation from the native state occurs via conformational changes leading to the formation of sticky amyloid-prone, partially structured intermediate(s). The partially structured intermediate state(s) is postulated to associate into oligomers and subsequently undergo structural rearrangement to form amyloid-like fibrils (8,16). However, very little information exists on the conformational features of the "amyloid-prone" partially structured intermediate state(s).
In the present study, we investigate the TFE-induced conformational transitions in a ␤-barrel protein such as the newt acidic fibroblast growth factor (nFGF-1, Refs. 31 and 32). TFE is observed to induce amyloid-like fibrils in nFGF-1. The formation of amyloid-like fibrils is triggered by the accumulation of a partially structured intermediate state(s) in the TFEinduced unfolding pathway of nFGF-1.

MATERIALS AND METHODS
Heparin-Sepharose was purchased from Amersham Biosciences. Labeled 15 NH 4 Cl, D 2 O, and TFE-d 3 were purchased from Cambridge Isotope Laboratories. 1-Anilino-8-naphthalene sulfonate (magnesium salt), thioflavin T (ThT), and Congo red were purchased from Sigma. All other chemicals used were of high quality analytical grade. Unless otherwise mentioned, all solutions were made in 100 mM phosphate buffer (pH 7.0) containing 100 mM sodium chloride. All experiments were performed at 20°C.
Protein Expression and Purification-Recombinant nFGF-1 was prepared from transformed Escherichia coli BL21(DE3)pLysS cells. The nFGF-1 DNA construct consisting of 486 base pairs was inserted between the NdeI and BamH1 restriction sites. The expressed protein was * This work was supported by the National Science Council of Taiwan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. Fax: 886-35-711082; E-mail: cyu@mx.nthu.edu.tw. purified on a heparin-Sepharose affinity column over a NaCl gradient (0 -1.5 M). Desalting of the purified protein was achieved by ultra filtration using an Amicon set-up. The purity of the protein was assessed using SDS-PAGE. The first 22 residues of the full form of nFGF-1 were digested by subjecting the expressed full form of nFGF-1 to the action of chymotrypsin. Chymotrypsin digestion was carried out by incubating the column material (heparin-Sepharose containing the bound protein) with the enzyme (at an enzyme to protein ratio of 20:1) in 10 mM phosphate buffer (pH 7.2) containing 0.85 M NaCl. The incubated mixture was stirred mildly at room temperature for 3 h. The incubated material was repacked into a column and washed with 10 mM phosphate containing 0.85 M NaCl until the absorbance of the eluate plateaued to a steady base line. Truncated nFGF-1 was later eluted with 10 mM phosphate buffer (pH 7.2) containing 1.5 M NaCl. The homogeneity of the truncated nFGF-1 sample was checked by SDS-PAGE. The authenticity of the truncated sample was verified by electrospray ionization-mass spectrometry analysis. Protein concentration was estimated based on the extinction co-efficient value of the protein at 280 nm. It should be stated that the truncated newt FGF-1, which we label as nFGF-1, is used in all the experiments described ahead in the present work.
Preparation of Isotope Enriched nFGF-1-15 N isotope labeling was achieved using M9 minimal medium containing 15 NH 4 Cl. To maximize expression yields, the composition of the M9 medium was modified by the addition of a mixture of vitamins. The expression host strain, E. coli BL21(DE3)pLysS, is a vitamin B 1 -deficient host, and hence, the medium was supplemented with thiamine (vitamin B 1 ). Protein expression yields were in the range of 25-30 mg/liter of the isotope-enriched medium. Purification and chymotrypsin digestion methods to obtain truncated nFGF-1 were the same as described in the previous section. The extent of 15 N labeling was verified by electrospray ionization-mass spectrometry analysis.
Turbidity Measurements-Turbidity measurements were performed on a Hitachi U-3310 spectrophotometer. All measurements were made after 3 h of incubation (at 20°C) of the protein (nFGF-1) in appropriate concentrations of TFE. The concentration of the protein used in the turbidity experiments was 100 g/ml. The turbidity of the solutions was measured by absorbance at 350 nm. The path length of the sample cell used was 10 mm.
Circular Dichroism-All CD measurements were made on a Jasco J-720 spectropolarimeter. CD experiments were performed using a 0.2-cm quartz cell for far (195-250 nm) and 10-cm quartz cell for the near UV CD (250 -320) regions. A step size of 0.1 nm, an average time of 3 s, and an average of 10 scans were recorded to generate the data. The concentration of the protein inside was 0.1 mg/ml. The far and near UV CD spectra were smoothed using the noise-reducing option in the software supplied by the vendor (Jasco).
Congo Red and Thioflavin T Binding-The concentration of protein used in these experiments was 100 g/ml. Protein samples were incubated in various concentrations of TFE for 3 h before the dye (Congo red or thioflavin T) was added. The solutions were stirred upon the addition of the dye. Fluorescence spectra of ThT were measured on a Hitachi F2500 spectrofluorimeter using a quartz cell with a light path of 10 mm. The excitation wavelength was set at 440 nm, and bandwidths for excitation and emission lights were 2.5 and 10 nm, respectively. For the Congo red binding assay, absorption spectra of the samples containing 10 l of a 1 mM solution of the dye (Congo red) were recorded on a Hitachi U-3310 spectrophotometer in the range of 400 -700 nm.
Transmission Electron Microscopy-Electron micrographs of amyloid fibrils (generated in TFE) of nFGF-1 were acquired on a Hitachi H-7500 transmission electron microscope. A 2-l sample of the protein at a concentration of about 100 g was applied to Formvar-and carboncoated copper grids. The fibrils were washed by the successive addition of 3 wash steps (10-l aliquots of water) followed by drying with filter paper. The fibrils were stained by the addition of 10 l of uranyl acetate (1% w/v) and immediately dried with filter paper.
X-ray Diffraction-The fibers induced in 15% (v/v) TFE were suspended between the ends of 2 wax-filled glass capillaries and allowed to dry. The x-ray scattered image of TFE-induced fibrils were collected using a wavelength 1.11 Å on an off-line image plate detector at BL17B2 of the synchrotron radiation. The specimen was mounted on a 1-mm loop for free standing to avoid the x-ray beam touching any material other than fibrils. The image was processed by using the XPRESS software.
Seeding Experiment-Aggregation of nFGF-1 (5 mg/ml) in 15% (v/v) TFE were incubated for about 2 weeks at room temperature to obtain dense fibers. The TFE-induced fibers were repeatedly washed in 10 mM phosphate buffer (pH 7.2) containing 100 mM NaCl, filtered, and dried.
For the seeding experiments, various aliquots 100 -500 l of seeding solution (TFE-induced preformed fibrils) of the protein was added to freshly prepared solution (TFE induced preformed fibrils) of the protein. The seeding experiments were carried out with appropriate controls. The first control involved the incubation of an identical solution of nFGF-1 in the absence of the aliquot of the seeding solution, and the second control involved the dilution of an approximate aliquot of the seeding solution into phosphate buffer. The seeding (aggregation) was monitored by the increase in the absorbance at 350 nm. In addition the seeding solution was analyzed by transmission electron microscopy to obtain definite evidence for the existence of fibrils and analyzing their morphology.
Steady State Fluorescence-Fluorescence experiments were performed on a Hitachi F2500 spectrofluorimeter at 2.5-or 10-nm resolution. Intrinsic fluorescence measurements were made at a protein concentration of 100 g/ml using an excitation wavelength of 280 nm. 1-Anilino-8-napthalene sulfonate (ANS) binding experiments were carried out on nFGF-1 at various concentrations of TFE using an excitation wavelength of 390 nm. The emission was monitored between 420 and 600 nm. The concentrations of the dye (ANS) and the protein (nFGF-1) used were 200 M and 100 g/ml, respectively.
NMR Experiments-All NMR experiments were performed on Bruker Avance-600 NMR spectrometer at 20°C. A 5-mm inverse probe with a self-shielded z-gradient was used to obtain all gradient-enhanced 1 H, 15 N heteronuclear single quantum coherence (HSQC) spectra. 15 N decoupling during acquisition was accomplished using the globally optimized alternating-phase rectangular pulse sequence. 2048 complex data points were collected in the 1 H, 15 N HSQC experiments. 512 complex data points were collected in the 15 N dimension. The HSQC spectra were recorded at 32 scans at all concentrations of TFE. The concentration of the protein sample was 0.5 mM in 95% H 2 O and 5% D 2 O (containing 100 mM phosphate and 100 mM sodium chloride). 15 N chemical shifts were referenced using the consensus ratio of 0.01013291189. All spectra were processed on a Silicon Graphics work station using XWINNMR and AURELIA software.
Fourier Transform Infrared Spectroscopy (FT-IR)-The samples for the FT-IR spectral measurements were prepared as follows. Protein aggregates formed in 15% (v/v) TFE (at a protein concentration of 1 mg/ml) were centrifuged at 10,000 rpm using a desktop centrifuge for 20 min, and the supernatant was removed carefully. The precipitant was dried overnight in a vacuum desiccator. The dry powder of the aggregates was enclosed in a KBr tablet by a conventional high pressure method. FT-IR spectra of nFGF-1 at 0% (v/v) and 70% (v/v) TFE were acquired by dissolving the protein in 99% D 2 O containing appropriate concentrations of TFE-d 3 . All spectra were recorded with a wave number resolution of 2 cm Ϫ1 . For each spectrum, 64 -200 interferograms were collected and averaged, and a Happ-Genzel apodization function was applied before Fourier transformation. All processing procedures were carried out so as to optimize the quality of the spectrum in the amide I region, between 1600 and 1700 cm Ϫ1 .

RESULTS AND DISCUSSION
nFGF-1 is a ϳ15-kDa, all ␤-sheet protein with no disulfide bonds (31). The protein lacks helical segments, and the secondary structural elements include 12 ␤-strands arranged into a ␤-barrel architecture (Fig. 1).

Effect(s) of TFE on the Backbone Conformation-
The original objective of this study was to investigate the specificity/nonspecificity of the helix-inducing effect(s) of 2,2,2-trifluoroethanol. In this context, we monitored the secondary structural changes in nFGF-1 at various concentrations of TFE using far-UV circular dichroism (CD) spectroscopy. The far-UV CD spectrum of nFGF-1 (200 -250 nm) shows two prominent ellipticity bands (a positive ellipticity band at 228 nm and an intense minimum at around 205 nm) characteristic of type II ␤-barrel proteins (Fig. 2, inset). The 228-nm CD band is believed to primarily represent the arrangement of the ␤-strands in the ␤-barrel architecture of the protein. No significant change(s) occurs in the intensity of the 228-nm CD band at a TFE concentration lower than 5% (v/v) (Fig. 2). However, at a TFE concentration of 8% (v/v), the far-UV CD spectra of the protein show dramatic changes (Fig. 2). The positive CD band centered at 228 nm is lost, implying the disruption of the native ␤-barrel architecture. The loss of the positive CD band at 228 nm is paralleled by the appearance of a negative CD band at around 218 nm in 20% TFE (v/v) (Fig. 2), suggesting the formation of extended ␤-sheet conformation. The intensity of the negative ellipticity band at 218 nm is observed to steadily increase with the increase in the TFE concentration from 10 (v/v) to 50% (v/v) (Fig. 2). At higher concentrations of TFE (Ͼ50% (v/v)), the far-UV CD spectra of the protein show two negative ellipticity bands at 208 and 222 nm, indicating the induction of helical conformation (Fig. 2). Because nFGF-1 in its native conformation lacks helical segments, the induced helix conformation (at higher concentrations of TFE (Ͼ50% v/v)) appears to be essentially non-native in origin. The percentage of non-native helix estimated from the far-UV CD spectrum of the protein 70% v/v is about 20%. The majority (Ͼ60%) of the backbone of the protein (33, 34) (in 70% (v/v) TFE) is estimated to be unstructured. The results discussed above clearly suggest that the conformational transitions induced by TFE in nFGF-1 occur in two stages. In the first stage, the ␤-barrel conformation is disorganized, resulting in the formation of extended ␤-sheet conformation. In the second stage, portions of the backbone of the protein form non-native helix conformation.
Influence of Backbone Conformation on Protein Aggregation-The protein samples treated with TFE in the concentration range of 10% (v/v) to 45% (v/v) and incubated at room temperature (ϳ25°C) for more than 3 h are observed to turn turbid. Hence, the aggregation of the protein (nFGF-1) is examined systematically by monitoring the changes in the 350-nm absorbance (scattering) at various concentrations of TFE (after 3 h of incubation of the protein at 25°C in various concentrations of TFE). Interestingly, the aggregation profile shows that the protein tends to aggregate maximally between 8 and 45% (v/v) TFE, wherein the protein is observed to exist in an extended ␤-sheet conformation (Fig. 3). The extent of aggregates formed appears to increase with the increase in the formation of the extended ␤-sheet conformation in the protein.
In addition, the aggregates formed (at TFE concentrations between 10% to 50% (v/v) TFE) are long and thread-like. Beyond this range of TFE concentration (Ͼ50% (v/v) TFE), the protein solution turns clear (with low 350-nm absorbance values) even after 48 h of incubation at 25°C (Fig. 3). These results suggest that the aggregation of nFGF-1 is related to the nature of non-native secondary structural elements induced in the protein (by TFE). Formation of non-native extended ␤-sheet conformation appears to promote aggregation, and induction of non-native helix conformation correlates with the inhibition of protein aggregation. Similar observations were made by Dobson and co-workers, based on exhaustive point mutations in the muscle acylphosphatase protein (35,36). A clear-cut kinetic partitioning between aggregation and folding of proteins was observed. Protein aggregation was strongly correlated to the ␤-sheet propensity of the regions of the protein (muscle acylphosphatase) in which the point mutations were located (35). In addition, mutations that stabilize helix conformation were shown to decrease the aggregation process, and those that destabilized the helix increased the aggregation rate (35).
TFE-induced Aggregates Possess Amyloid Fibril-like Properties-The aggregates of nFGF-1 formed in TFE (in 8 -40% (v/v) TFE) have a fibrous texture, suggesting that they could possess properties characteristic of amyloid fibrils. This possibility was examined by characterizing the tinctorial and ultrastructural properties of the TFE-induced aggregates.
Congo red is a hydrophobic dye routinely used to identify amyloid fibril formation by proteins (37). The absorption maximum of the dye is known to undergo a red shift upon binding to ordered repetitive ␤-sheet structures in the amyloid fibrils (38). Although the TFE-induced aggregation is maximum at 10% (v/v), Congo red binding is observed maximally at 15% (v/v) TFE (Fig. 4A). The dye binding is not only accompanied by a significant increase in the absorbance intensity at 490 nm but also by a prominent red shift in the absorbance maximum (490 to 520 nm, Fig. 4A, inset). Beyond 15% (v/v) TFE the intensity of the dye at 490 nm decreases steadily with the increase in concentration of the fluoro alcohol. These results suggest that the fibrils formed maximally at 15% (v/v) TFE have amyloidlike characteristics.
The nature of the aggregates formed were also probed using ThT. ThT is a fluorescent dye used as a diagnostic to identify amyloid fibril formation (39). The emission intensity of ThT is known to increase significantly upon binding to the linear array of ␤-strands in the amyloid fibrils. There is no appreciable change(s) in the fluorescence intensity of the dye at TFE concentrations lower than 10% (v/v) (Fig. 4B). However, an 8-fold increase in the fluorescence intensity at 485 nm is observed when the dye binds to the protein in 15% TFE (Fig. 4B). These results provide further evidence suggesting that amyloid-like fibrils of nFGF-1 accumulate maximally in 15% (v/v) TFE.
The ultrastructure of the aggregates (of nFGF-1) formed in 15% (v/v) TFE was examined by transmission electron microscopy and x-ray fiber diffraction. Electron micrographs reveal that the aggregates formed are straight and unbranched, with an average diameter of about 10 -15 nm (Fig. 5A) and is similar to the width of the fibrils formed from other amyloidogenic proteins (39). X-ray fiber diffraction analysis of the nFGF-1 aggregates shows a dominant reflection at 4.7 Å on the meridian and associated equatorial reflection at 10.3 Å, typical of those expected for fibrils with a cross-␤ structure (Fig. 5B). The meridional reflection (at 4.7 Å), which primarily arises from the spacing between the ␤-sheet structure, is sharp and intense (Fig. 5B). The equatorial reflection (at 10.3 Å), which is generally attributed to the intersheet spacing, is relatively diffused (Fig. 5B). Thus, the results of the electron microscopy and x-ray fiber diffraction experiments authenticate that the aggregates formed in 15% (v/v) TFE possess features resembling that of the amyloid fibrils.
Seeding Fibril Formation with Preformed Fibrils-Seeding is a common phenomenon in aggregation and gelation processes, as it is in crystallization (40). It has been reported that aliquots of pre-formed fibrils formed by AB peptide when it is added to clear solutions substantially accelerated the rates of formation of amyloid-like fibrils (8). Recently, Bucciantini et al. (27) demonstrate that pre-formed amyloid fibrils (formed in vitro) of proteins including SH3 domain from phosphatidylinositol 3-kinase and the N-terminal domain of the E. coli HypF protein, which are not associated with amyloidoses, are highly cytotoxic and induced amyloid-like fibrils in cultured NIH3T3 fibroblast cells (27). In this context, we investigated the seeding potency of the amyloid-like fibrils of nFGF-1 induced in TFE. The potency of the pre-formed fibrils of nFGF-1 (induced in 15% TFE (v/v)) to seed fibril formation was examined by monitoring the time-dependent changes(s) in the 350-nm absorbance upon the addition of aliquots of pre-formed fibrils into solutions (1 mg/ml) of native nFGF-1. It could be observed from Fig. 6A that the 350-nm absorbance increased significantly upon the addition of the pre-formed fibrils (induced in TFE). Seeding experiments performed by the addition of increasing aliquots of pre-formed nFGF-1 fibrils into a fixed concentration (1 mg/ml) of native nFGF-1 solution revealed that the rate of aggregation increases with the increase in the amount of the preformed fibrils added (Fig. 6A). However, the extent of induced aggregation (as deduced from the 350-nm absorbance) appears to be independent of the amount of the pre-formed fibrils added (Fig. 6A). It should be mentioned that control experiments, with an aliquot of the pre-formed fibrils diluted ϳ20-fold into the buffer (10 mM phosphate containing 100 mM NaCl) and a solution of nFGF-1 diluted 20-fold with the buffer to a final protein concentration of 1 mg/ml under the same conditions, show no signs of increases in aggregation (as observed from the 350-nm absorbance). These experiments demonstrate qualitatively that pre-formed fibrils of nFGF-1 (formed in TFE) are capable of seeding fibril formation in solutions containing the protein (nFGF-1) in the native conformation.
We performed ThT binding experiments to verify if the aggregates seeded by the pre-formed fibrils have amyloid-like characteristics. A significant time-dependent increase in the fluorescence intensity of the dye (ThT) at 485 nm could be observed upon incubation of the native protein with pre-formed fibrils (Fig. 6B). These results suggest that the induced aggregates (upon seeding) indeed possess properties resembling the amyloid fibrils. In addition, transmission electron micrographs revealed that the induced aggregates are fibrillar with an average diameter of about 15 nm. 2 In summary, the results of the seeding experiments clearly demonstrate that fibril growth (in nFGF-1) is dominated by a nucleation mechanism similar to that observed in crystal growth and gelation process. Such in vitro seeding has also been observed in studies of human lysozyme and its amyloidogenic variants (41). Recently, fibrils generated under in vitro conditions in hen egg white lysozyme (which is not associated with amyloid diseases) have been shown to seed fibril formation of protein in the native conformation (42). Similarly, in vitro studies with the AB peptide associated with Alzheimer's diseases suggest that origin of the rapid onset of amyloid diseases and infectivity of prion-related diseases occurs by a nucleation (seeding) mechanism (8). The inset represents the emission intensity changes of the dye (ANS) at various concentrations of TFE. It could be observed that the emission intensity is maximum in 15% TFE, suggesting maximal accumulation of the "molten globule-like" intermediate. A). 100-, 200-, and 500-l aliquots (drawn from the pre-formed fibrils formed at 2 mg/ml native protein concentration) were added to solutions of nFGF-1 in its native state. It could be observed that the rate of induction of aggregation depends on the amount of the pre-formed fibrils added. However, the total aggregation induced appears to be independent of the concentration of the preformed fibrils used for seeding. Panel B shows the thioflavin T binding affinity of the induced aggregates. The control experiment (open triangles) with ThT added to nFGF-1 (in 10 mM phosphate buffer (pH 7.2) containing 100 mM NaCl (in the absence of the pre-formed fibrils)) shows no time-dependent increase in the emission intensity at 485 nm. These results suggest that the induced aggregates possess properties of the amyloid fibrils.

FIG. 6. Aggregation induced by pre-formed fibrils (formed in TFE) of nFGF-1 in a solution containing nFGF-1 in its native conformation (panel
cently, there is increased interest in understanding the molecular events leading to the formation of amyloid fibrils. Based on experimental evidence available in different proteins, it is now increasingly believed that amyloid fibril formation involves the ordered self-assembly of partially folded species that are crucial soluble precursors of fibrils (22)(23)(24)(25)(26)(35)(36)(37). In this context, we monitored the conformational changes in the protein before amyloid-like fibril formation.
The fluorescence spectrum of nFGF-1 shows an emission maximum at around 308 nm (Fig. 7, inset). The fluorescence of the lone tryptophan residue located at position 121 is significantly quenched in the native state of the protein (43)(44)(45)(46). This quenching effect is attributed to the presence of imidazole and pyrrole groups in the vicinity of the indole ring of the tryptophan (Trp-121) residue (31). However, the quenching effect is relieved upon unfolding of the protein, yielding a fluorescence spectrum with an emission maxima at around 350 nm (Fig. 7,  inset). Hence, the ratio of 350-to 308-nm fluorescence reliably describes the conformational transitions occurring during the unfolding of the protein (46). The fluorescence spectra of nFGF-1 does not appreciably change at lower concentrations of TFE (Ͻ8% (v/v) TFE). However, in the TFE concentration range of 8 to 15% (v/v) TFE, the ratio of the 350-to 308-nm fluorescence drastically increases, suggesting substantial loss of tertiary structural interactions in the protein (Fig. 7). The TFE-induced unfolding effects on the protein are maximum at 15% (v/v) TFE. Thus, analyzing the intrinsic fluorescence data in conjunction with that obtained using far-UV CD (indicating ␤-sheet segments in the protein) suggests that nFGF-1 exists in a partially structured state(s), with extended ␤-sheets and loosely packed side chains. Interestingly, at higher concentrations of TFE (Ͼ50 (v/v) TFE), the fluorescence spectra show a significant decrease in the 350-nm emission, suggesting a change in the microenvironment of the tryptophan residue upon induction of non-native helical conformation in the protein.
ANS is a popular hydrophobic dye that is known to bind to solvent-exposed hydrophobic surfaces in a manner characteristic of the formation of partially structured state(s) (47). The dye generally exhibits weak binding affinity to the native and unfolded state(s) of protein (46). The binding affinity of ANS to the protein does not significantly change below 8% (v/v) TFE (Fig. 8). However, beyond this TFE concentration (Ͼ8% (v/v) TFE), the fluorescence intensity of the dye at 485 nm increases drastically, reaching a maximum value at 15% (v/v) TFE. The intensity of the dye upon binding to the protein in 15% (v/v) TFE is about 20 times that observed when it (ANS) is bound to the native state of the protein (Fig. 8). In addition, the dye upon binding to the protein in 15% (v/v) TFE shows a prominent blue shift of about 42 nm (from 520 to 475 nm) in the wavelength of maximum emission (Fig. 8). These spectral characteristics clearly indicate that a partially structured state(s) with a solvent-exposed non-polar surface(s) accumulates (maximally at 15% (v/v) TFE) before the formation of amyloid-like fibrils. Beyond 15% (v/v) TFE, the fluorescence intensity of ANS at 485 nm decreases progressively (with the increase in the concentration of the fluoro alcohol (TFE)), indicating a depletion in the population of the partially structured intermediate (exhibiting high binding affinity to ANS) state(s) (Fig. 8). It appears that the partially structured intermediate state(s), which is accumulated maximally at 15% (v/v) TFE, is sticky and has a high tendency to aggregate through solvent-exposed hydrophobic surface(s) and intermolecular ␤-sheet formation (to be discussed below). The resultant aggregates possibly rearrange to form organized amyloid-like fibrils.
NMR spectroscopy facilitates the study at the level of individual amino acids residues during folding/unfolding of proteins (48). Heteronuclear correlation experiments have been shown to be very sensitive because of high magnetization transfer between directly bond nuclei (48). This aspect enables the use of 1 H, 15 N HSQC technique to investigate the conformational changes induced by TFE at high resolution. In general, 1 H, 15 N HSQC spectrum serves as a fingerprint of the conformational state of a protein (48). The HSQC spectrum of nFGF-1 in its native conformation (0% TFE) is well dispersed, and all the expected 126 1 H, 15 N cross-peaks in the spectrum could be unambiguously assigned (31). No discernable changes(s) could be observed in the HSQC spectrum of the protein acquired below a TFE concentration of 5% (v/v) (Fig. 9). We could not assess the structural features of the partially structured intermediate state(s) accumulated at 15% (v/v) TFE because the protein (even at 0.1 mM concentration) aggregates seriously after sample preparation (in 15% (v/v) TFE). Under these circumstances, the structural characteristics of the partially structured intermediate state(s) were predicted from the 1 H, 15 N chemical shift perturbation observed in the HSQC spectrum at 8% (v/v) TFE. 1 H, 15 N HSQC spectrum of the protein in 8% (v/v) TFE shows that many cross-peaks (in the spectrum) undergo significant chemical shift perturbation (Fig. 9B). These spectral features are indicative of gross conformational changes leading to the protein. Many cross-peaks correspond- ing to residues in the secondary structure regions show prominent chemical shifts in perturbation. Interestingly, residues (Lys-26, Lys-142, Ala-143, Leu-145, Leu-147, Leu-149, and Asp-154) that bridge the N-and C-terminal ends of the nFGF-1 molecule through hydrogen bonds also show significant chemical shift perturbation, (Ref. 31, Fig. 9B). This observation strongly suggests that the native ␤-barrel architecture is disrupted in the partially structured intermediate state(s) that accumulates before fibril formation. Although the structural features of the protein in 8% (v/v) TFE and 15% (v/v) TFE (wherein the partially structured state(s) is maximally populated) cannot be directly compared, it may not be far-fetched to infer that the conformational flexibility of the protein in the partially structured intermediate state(s) (at 15% (v/v) TFE) could be similar (if not higher) to that observed in 8% (v/v) TFE. The 1 H, 15 N HSQC spectra of the protein in 70% (v/v) TFE reveals that the chemical shift dispersion of the cross-peaks is vastly diminished, implying that large portions of the protein molecules are in a disordered conformation (Fig. 9A). The far-UV CD and NMR data (obtained in 70% (v/v) TFE) analyzed in conjunction suggests that the protein possibly exists primarily as a random coil with only small portions of the polypeptide backbone existing in non-native helix conformation. It should be mentioned that in the absence of detailed triple resonance data, the chemical shift indices of the resonances ( 1 H, 15 N) in 70% (v/v) TFE cannot be estimated. This aspect precludes the assignment of non-native helical segment formed in the protein in 70% (v/v) TFE.
Mechanism of Amyloid-like Fibril Formation-We are tempted to propose a model for the TFE-induced amyloid-like fibrils in nFGF-1. It is well known that alcohols such as TFE exert mild denaturant effects by partially disorganizing the tertiary and quaternary structures of proteins (49,50). The denaturant effects of TFE on proteins primarily stem from their low polarity, which weakens the hydrophobic interactions that stabilize the compact native structures of proteins (51). In this background, the first conformational transition induced by TFE (at concentrations greater than 10% (v/v)) appears to be the disorganization of the hydrophobic contacts stabilizing the native ␤-barrel structure leading to the formation of a sticky partially structured intermediate state(s) (Fig. 10). The presence of solvent-exposed, non-polar surface(s) in the partially structured state(s) (populated in 15% (v/v) TFE) is evident from the high binding affinity of the protein in the intermediate state(s) to ANS (Fig. 8). Similarly, the disorganization of the native ␤-barrel architecture is probably reflected in the significant chemical shift perturbation of most of the cross-peaks in the 1 H, 15 N HSQC spectrum in 8% TFE and the loss of the positive ellipticity band at 228 nm with the concomitant appearance of the negative CD band at 218 nm (Figs. 2 and 9). The loss of the native ␤-barrel architecture is also evident from the FT-IR spectrum acquired in 15% (v/v) TFE, wherein the 1618 and 1639 cm Ϫ1 amide I doublet bands characterizing the ␤-barrel structure disappear (Fig. 11). Aggregation of the protein (observed upon prolonged incubation (Ͼ3 h) of the protein in the TFE concentration range of 10 to 40% (v/v)) is probably triggered by the coalescence of the sticky extended ␤-sheet elements in the partially structured intermediate state(s) (Fig.  10). The amyloid-like fibril formation appears to arise due to the rearrangement and annealing of the extended ␤-sheet elements through intermolecular hydrogen bond formation. Formation of intermolecular ␤-sheets is corroborated by the presence of the 1688-cm Ϫ1 band in the FT-IR spectrum of the TFE-induced amyloid-like fibril (8) (Fig. 11). In addition to the intermolecular hydrogen bonds that favor aggregation, the enhanced solvent exposure of the non-polar side chains in the protein (in the non-native extended ␤-sheet conformation) also appears to provide a conducive environment for the condensation of the polypeptide chain to form a higher order aggregates, such as the amyloid-like fibrils (Fig. 10). The inhibition of aggregation observed in higher concentrations of TFE (Ͼ50% (v/v) TFE) is possibly due to weakening of the hydrophobic interactions stabilizing the intermolecular ␤-sheet structures in the aggregates (50,51). The loss of non-polar contacts (at higher concentration of TFE) appears to favor the formation of intramolecular hydrogen bonds (over intermolecular hydrogen bonds) conducive for the induction and formation of helix conformation. These results clearly suggest that protein aggregation is intricately linked to the backbone conformation (52,53). Induction of non-native ␤-sheet conformation appears to promote aggregation (subsequently to form fibrils), and formation of non-native helical segments in the backbone appears to inhibit aggregation. Therefore, these findings could provide useful information for the design of new strategies to thwart protein aggregation in vivo and in vitro. In addition, although TFE is not a physiologically relevant solvent, the ␤-sheet to ␣-helix structural transition induced by TFE could be used as a useful model to understand the conformational switch of the prion protein from the soluble ␣-helical form to the ␤-sheet conformation leading to amyloid fibril formation.
In general, the results of the present study support the proposal of Dobson and coworkers (21) that formation of amyloid fibrils is a generic property of peptides and proteins and not one restricted to the particular species observed in diseased states. Although TFE is not a physiologically relevant solvent, understanding the mechanism of TFE-induced amyloid formation in proteins (such as nFGF-1) provides valuable insights into the fibrillation process.