Acid-denatured small heat shock protein HdeA from Escherichia coli forms reversible fibrils with an atypical secondary structure

The periplasmic small heat shock protein HdeA from Escherichia coli is inactive under normal growth conditions (at pH 7) and activated only when E. coli cells are subjected to a sudden decrease in pH, converting HdeA into an acid-denatured active state. Here, using in vitro fibrillation assays, transmission EM, atomic-force microscopy, and CD analyses, we found that when HdeA is active as a molecular chaperone, it is also capable of forming inactive aggregates that, at first glance, resemble amyloid fibrils. We noted that the molecular chaperone activity of HdeA takes precedence over fibrillogenesis under acidic conditions, as the presence of denatured substrate protein was sufficient to suppress HdeA fibril formation. Further experiments suggested that the secondary structure of HdeA fibrils deviates somewhat from typical amyloid fibrils and contains α-helices. Strikingly, HdeA fibrils that formed at pH 2 were immediately resolubilized by a simple shift to pH 7 and from there could regain molecular chaperone activity upon a return to pH 1. HdeA, therefore, provides an unusual example of a “reversible” form of protein fibrillation with an atypical secondary structure composition. The competition between active assistance of denatured polypeptides (its “molecular chaperone” activity) and the formation of inactive fibrillary deposits (its “fibrillogenic” activity) provides a unique opportunity to probe the relationship among protein function, structure, and aggregation in detail.

Pathogenic Escherichia coli strains must be resistant to the highly acidic environment of the stomach to successfully infect their human host (1). A major contributor to this acid resistance is expressed in the form of two structurally homologous but functionally distinct periplasmic small heat shock proteins, HdeA and HdeB (2)(3)(4)(5)(6)(7). The two chaperones are initially translated as precursors (HdeA, 110 amino acids; HdeB, 108 amino acids) (8) that are transported to the periplasm and then processed to mature forms (HdeA, 89 residues; HdeB, 79 residues). At neutral pH, both HdeA and HdeB form inert, ␣-helix-rich dimers (exemplified for HdeA in Fig. 7) (4). However, HdeA and HdeB both undergo acid-induced denaturation at low pH (ϽpH 3 for HdeA; ϽpH 4 for HdeB), upon which these two proteins become active molecular chaperones, capable of binding to a broad range of polypeptides that are denatured and are therefore susceptible to aggregation (2,6,9). Due to differences in structure and molecular interactions with targets, the conditions under which HdeA and HdeB are optimally functional are different; HdeA functions best at pH 2, whereas HdeB is optimally active at pH 4 (7). This difference translates to a separation of roles between the two small heat shock proteins that are reflected in their chaperone-client specificities (10). Apparently, the mechanism by which HdeA and HdeB prevents aggregation of periplasmic targets is through simple binding and release, as no additional molecular mechanisms, such as ATPase activity or specific cofactor-binding sites, have been implicated so far (7,(11)(12)(13).
In an effort analogous to our previous studies to use bacterial chaperones for protein aggregation control (14,15), we set out to probe the possibilities of using HdeA and HdeB in the control of irreversible aggregation and fibrillation of proteins, specifically at low pH. We began our experiments using human ␣-Synuclein because previous studies by Uversky et al. (16) demonstrated that this protein forms fibrils more readily at lower pH (pH Ͻ5) than at neutral pH ( Fig. 1). Surprisingly, from these preliminary experiments we found that HdeA itself, in its aciddenatured active form, was capable of forming fibril-like deposits that shared numerous characteristics with pathological amyloid fibrils, such as binding to the Thioflavin-T (Thio-T) 2 fluorescent probe (17) and regular twisted fibril morphologies that were observable by transmission EM (TEM) and atomicforce microscopy (AFM).
We describe here our findings regarding the fibril-forming abilities of HdeA and the relationship between this ability and the function of HdeA as an acid-induced molecular chaperone. A very surprising finding was that once formed, these HdeA fibrils could be reconstituted through a simple shift to neutral pH and regain molecular chaperone activity. Our findings highlight the remarkable reversibility of the structural transitions of HdeA and the suitability of using HdeA as an attractive exper-This work was partially supported by Tottori University, Japan Agency for Medical Research and Development (AMED) Grant JP18dm0107073 and the Skylark Food Science Institute. The authors declare that they have no conflicts of interest with the contents of this article. This article contains Figs. S1-S3 and Table S1. 1 To whom correspondence should be addressed: Dept. of Engineering, Tottori University, 4-101 Koyama-Minami, Tottori City, Tottori 680-8552 Japan. Tel./Fax: 81-857-31-5691; E-mail: mizobata@tottori-u.ac.jp. 2 The abbreviations used are: Thio-T, Thioflavin-T; ADH, alcohol dehydrogenase; AFM, atomic-force microscopy; ATR, attenuated total reflection; TEM, transmission electron microscopy. cro ARTICLE imental system to probe the relationship between protein structural dynamism and activity.

Results
Our original intent was to see whether HdeA was capable of suppressing fibrillation of ␣-Synuclein (16). However, as shown in Fig. 1A, contrary to our expectations the addition of HdeA to a sample of ␣-Synuclein incubated at pH 2.0 caused a large increase in Thio-T-binding fluorescence compared with samples containing ␣-Synuclein only. This prompted us to probe in more detail the behavior of HdeA at low pH in isolation, which revealed this protein's ability to form similar insoluble fibrils. Fig. 1B shows a typical fibril-forming reaction of HdeA at pH 2.0, monitored by Thio-T fluorescence. The reaction proceeded according to a mechanism with an initial lag phase where no increase in fluorescence was observed followed by a rapid increase in fluorescence that reflected the extension phase. The Thio-T fluorescence intensities attributed to the formation of fibrils were stronger in the presence of 150 mM NaCl, and experiments in the absence of NaCl showed no increases in Thio-T fluorescence intensity (Fig. 1B, red). However, subsequent experiments showed that this lack of fluorescence signal increase was not due to the absence of fibrils but to differences in fibril morphology that altered the sensitivity to Thio-T, as probed in detail in Fig. 3A.
Each of the samples in Fig. 1, A and B, that displayed increases in Thio-T fluorescence intensity were confirmed for the presence of actual fibrillar deposits by negatively stained TEM analysis (Fig. 1C). Fibrils observed for the ␣-Synuclein ϩ HdeA samples consisted mainly of short, densely stained fibrils mixed with longer, thinner twisted fibrils. In contrast, fibrils of ␣-Synuclein formed at pH 2.0 without HdeA consisted mainly of short fibril clumps. Finally, fibrils of HdeA were short, rodlike fibrils that showed a tendency to aggregate into larger deposits.
Having established that HdeA was capable of forming fibrillar aggregates at low pH, we next probed the sensitivity of this reaction to various experimental parameters. We first observed Reversible fibrillation of small heat shock protein HdeA a strong dependence of HdeA fibrillation on the concentration of NaCl ( Fig. 2A). At low NaCl concentrations, HdeA fibrils were formed through an extended lag phase followed by a modest increase in Thio-T fluorescence. In the presence of higher concentrations of NaCl, HdeA fibrils formed much more rapidly, and a very strong increase in fluorescence intensity exemplified the extension phase.
Because the results suggested that HdeA formed fibrils from its active acid-unfolded state, we wished to determine whether this fibril transformation was indeed correlated to the acid unfolding of HdeA. Previous experiments have shown that the native structure of HdeA undergoes denaturation at solvent conditions between pH 2.2 and pH 4.0, with transition midpoint pH values slightly lower than 3 (for secondary structure monitored by CD (9)) or 3.5 (for dimer dissociation monitored by FRET (11)). Therefore, we performed fibril-forming assays at various pH conditions centered around this pH region. As shown in Fig. 2B, HdeA tended to form fibrils rapidly at pH conditions below 3.0. An extension of the lag phase of fibrilla-tion and a decrease in the final Thio-T fluorescence value were both observed at pH conditions higher than 3.0. We also noted that the average trace of fibril formation at pH 3.0 became rather erratic and apparently multistage in nature. However, we found that this was caused by the fact that individually prepared samples of HdeA displayed large variances in lag-phase intervals (Fig. S1). These variances in lag phase caused an artifact in the averaged trace.
The results in Fig. 2B demonstrated that HdeA formed fibrils under conditions where this protein was active as a molecular chaperone, e.g. at pH values lesser than 3. We were therefore curious to determine whether these two tendencies, fibrillogenesis and molecular chaperone activity, were mutually exclusive, so we performed fibril extension experiments in the presence of a model unfolded target protein, yeast alcohol dehydrogenase (ADH) at pH 1.0. As seen in Fig. 2C, the addition of increasing concentrations of acid-denatured ADH served to strongly suppress the fibril formation of HdeA at this pH, suggesting that the molecular chaperone activity of HdeA took precedence Reversible fibrillation of small heat shock protein HdeA over the formation of fibrils when target proteins were present. This suppression of fibrils was also reflected in TEM images of HdeA samples formed in the presence of denatured ADH (Fig.  2D) where we generally observed less fibrillar structures when the concentration of denatured ADH was increased.
To characterize the morphological characteristics of HdeA fibrils under the conditions probed in Fig. 2, we next performed TEM and AFM characterization of HdeA fibrils. We found that the fibril morphology of HdeA was quite sensitive to NaCl concentration. Fibrils of HdeA formed in low concentrations of NaCl tended to form long, thin filaments, whereas in the presence of high concentrations of NaCl, HdeA fibrils tended toward shorter and thicker rodlike forms (Fig. 3A). In contrast, HdeA fibrils that were formed at a certain NaCl concentration were relatively indifferent to the pH conditions where fibrillation occurred; for example, as seen in Fig. 3B, we could not detect any large differences in morphology in fibrils formed in the absence of NaCl at various pH values. This visual impression was confirmed in more quantitative analyses that utilized AFM. As shown in Fig. 3C, the thicknesses (as measured by the vertical height difference of fibrils probed on mica by AFM; individual AFM images are shown in Fig. S2) of fibrils formed at a given NaCl concentration was statistically indistinguishable regardless of the pH in which the fibrils were formed. However, when we compared the dimensions of fibrils formed in the presence of 150 mM NaCl with those formed in the absence of NaCl,  Fig. S2. The number of measurements taken were n ϭ 18 for samples formed at pH 2.0 and n ϭ 9 for samples formed at other pH conditions. Measurements of samples formed at a given pH in the presence and absence of NaCl were compared using Welch's two-tailed t test (solid brackets) to determine whether the height of the fibrils differed when formed at different NaCl concentrations (p values are given next to the main figure). Additionally, measurements for the three samples formed in the presence of a given NaCl concentration were compared using one-way analysis of variance (ANOVA) (dashed brackets) to determine whether the average height of fibrils formed in the presence of a given NaCl concentration was statistically indistinguishable in thickness regardless of the pH. Taken together, both statistical analyses supported our notion that the height (thickness) of the HdeA fibrils formed depended on the NaCl concentration during the experiment but not on the pH.

Reversible fibrillation of small heat shock protein HdeA
we found that the former were significantly thicker at each pH that we tested.
From the results in Figs. 2 and 3, we may conclude that HdeA tends to form fibrils when the original structure is perturbed to expose a putative interface where fibril seed nucleation and subsequent fibril extension may proceed. The presence of NaCl served to shorten the initial lag time required for fibril nucleus formation, suggesting that HdeA fibrillation was dependent partially on either ionic shielding or enhancement of hydrophobic effects caused by an increase in ionic strength. NaCl may act to change the interactions between HdeA molecules during fibrillation, which are translated to differences in fibril morphology.
Because the molecular chaperone activity of HdeA depends on the acid denaturation of its native structure, we were curious to determine whether sudden changes in pH would affect the fibril-forming reaction of HdeA in a similar manner. To this end, we probed the effects of shifting the pH of HdeA fibril samples that had been formed at pH 2.0 to higher values. Interestingly, incubating HdeA fibrils that were formed at pH 2 at higher pH resulted in decreases in Thio-T fluorescence (Fig.  4A). The degree of signal decrease depended on the incubation pH; fluorescence intensities were unchanged after a pH shift to 4.6, but a partial decrease in intensity occurred upon a shift to pH 5.6, and complete loss of signal was observed at pH values higher than 6.0. The loss of fibrillar forms after incubation at pH 7 was confirmed in TEM images as shown in Fig. 4B. SDS-PAGE analysis of the reconstituted fibrils (Fig. 4C) showed that the insoluble fibrils formed at pH 2 (Fig. 4C, lane group c) were largely shifted to the soluble fraction by the shift in pH (Fig. 4C, lane group d). Strikingly, the HdeA samples that were recovered after incubation at pH 7 were active as molecular chaperones when reintroduced to low pH conditions as assayed using ADH (Fig. 4D, compare lane groups b and d). The data in Fig. 4 suggested strongly that the formation of HdeA fibrils at low pH was reversible and that HdeA was capable of converting freely between inactive dimer (neutral pH) and two very different low-pH structural forms, active monomer and insoluble fibril.
To probe the details of the structural transitions between native dimer, inactive fibril, and resolubilized HdeA, we performed spectral analysis aimed at determining the secondary structural characteristics of each HdeA form. CD spectra of representative fibril samples (Fig. 5A) revealed that the CD spectra of HdeA fibrils formed at pH 2.0 diverged somewhat from typical amyloid fibrils, such as fibrils of ␣-Synuclein (Fig.  5A, magenta). An analysis of the CD spectra using an online tool (BeStSel) (18) suggested that the HdeA fibrils contained significant fractions of ␣-helical content ( Fig. S3 and Table S1). HdeA fibrils that had been resolubilized by a shift in pH to 7.0 ( Fig. 5B) closely resembled the spectra for the native HdeA dimer regardless of the initial salt conditions used for the fibrillation reaction, suggesting that HdeA was indeed capable of reversibly returning to a state resembling its initial native conformation.
As we were interested in determining the process by which HdeA fibrils were formed at low pH, we next probed the changes in CD spectra that occur during formation of fibrils (Fig. 5C) both in the presence (150 mM) and absence of NaCl. In the absence of NaCl, we observed a slow change in the CD spectra of HdeA over the course of several days. No intermediate states during this transition were immediately apparent. In contrast, the CD spectra of HdeA in the presence of 150 mM NaCl initially transitioned from a spectrum resembling an aciddenatured state to one with a strong negative absorbance at 216 nm, which subsequently shifted again to a spectrum with a smaller negative absorbance with a minimum at 218 nm. This result suggested that formation of HdeA fibrils in the presence of 150 mM NaCl involved at least one intermediate conformation.
To obtain additional information regarding the secondary structure of HdeA fibrils at pH 2 and pH 7, we used attenuated total reflection (ATR)-FTIR spectroscopy to monitor the Amide I band of lyophilized HdeA samples hydrated with deuterium oxide (D 2 O). As seen in Fig. 6A, the IR absorbance spectra of native HdeA showed a peak in the vicinity of 1640 cm Ϫ1 and resembled the spectra of the ␣-helical-rich protein lysozyme, whereas spectra of the fibrils displayed a peak at ϳ1625 cm Ϫ1 and resembled the spectra of ␣-chymotrypsin, a protein with high ␤ content (19). A closer look using second derivative analysis of the absorbance spectra provided more details. All of the absorbance spectra measured showed absorption bands at 1652 and 1642 cm Ϫ1 , which are both characteristic absorption bands of ␣-helix (19). The spectra of native HdeA displayed a band at 1637 cm Ϫ1 that was not seen in the other samples, which may be attributed to random structure; in contrast, all of the samples with the exception of native HdeA and the fibril form of HdeA formed in 0 mM NaCl displayed a prominent band at 1634 cm Ϫ1 , which is characteristic of ␤-sheet (19). These results support the idea that fibrils of HdeA contain significant amounts of both ␣-helical and ␤-sheet secondary structure.
This hybrid nature of HdeA fibrils was highlighted further when compared with the second derivative spectra of native HdeA and ␣-Synuclein fibrils (Fig. 6B). Unlike native HdeA, with prominent absorption bands at 1660 -1640 cm Ϫ1 , or ␣-Synuclein, with strong absorption bands observed between 1640 and 1620 cm Ϫ1 , the second derivative spectra of HdeA fibrils formed in the presence of 150 mM NaCl showed observable peaks in both of these regions.
Interestingly, the IR spectra of HdeA fibril samples formed at pH 2 and subsequently reconstituted at pH 7.0 resembled neither the spectra of native HdeA nor the spectra of the original fibril form (Fig. 6C). Notably, we observed significant differences mainly in the 1640 -1620 cm Ϫ1 region that may be attributed to incubation at pH 7. These differences may be suggestive of a soluble state that is distinct from the original native structure of HdeA that is formed by fibrillation at pH 2 and subsequent reconstitution at pH 7, which is in apparent conflict with the results seen in the CD spectra (Fig. 5B) that suggested that HdeA fully recovered the secondary structure of its native form.

Discussion
The small heat shock proteins HdeA and HdeB are an important component of the acid-resistance mechanism in Gramnegative bacteria (20). E. coli strains that do not express these molecular chaperones have been demonstrated to be highly Reversible fibrillation of small heat shock protein HdeA sensitive to acidic conditions (2,4). The inherent "environmentally sensitive" nature of these small heat shock proteins is a novel and interesting characteristic in the molecular chaperone mechanism of HdeA/HdeB, and similar "conditionally active" molecular chaperones are currently an interesting subject of study.
In this study we have found that, in addition to becoming a promiscuous molecular chaperone upon denaturation at low pH, HdeA displays a tendency to form insoluble fibrillar aggre-gates that were morphologically similar to amyloid fibrils, the irreversible protein deposits that are implicated in many neurological disorders. The kinetics of this reaction is sensitive to the pH of the experiment, and the morphological aspects of the reaction are sensitive to the ionic strength. Generally, the tendency toward rapid fibril formation seems to be stronger at lower pH and higher salt concentrations; however, the resultant fibrils are relatively short and rodlike. Contrariwise, although both the rate and lag time of fibrillation at higher pH and lower

Reversible fibrillation of small heat shock protein HdeA
NaCl concentrations reflect a more gradual fibrillation process, the resultant HdeA fibrils are longer, filament-like structures.
Although morphologically similar and detectable with Thio-T, an amyloid-specific fluorescence probe, we have found that fibrils formed from acid-denatured HdeA deviated somewhat from common characteristics of other, typical amyloid fibrils in their CD spectra (Fig. 5). Compared with spectra of ␣-Synuclein fibrils, which displayed a clear absorbance minimum at 218 nm, the CD spectra of HdeA fibrils all displayed minor absorbances at shorter wavelengths. Estimation of the secondary structural content (Fig. S3 and Table S1) suggested that the fibrils might contain significant amounts of ␣-helical content. This is an unusual characteristic to detect in fibrils, although not entirely unknown, as seen previously for helixturn-helix peptides derived from apolipoprotein A-I (21) as well as the bacterially secreted phenol-soluble modulin PSM␣3, recently determined in crystallography analysis as a fibril with amphipathic cross-␣ structure (22). Considering the activity of HdeA as molecular chaperone under these conditions, it is quite possible that the denatured HdeA polypeptide retains a considerable amount of ␣-helical structure even when incorporated into fibrils, and only a crucial localized segment of the amino acid sequence is required to form the Thio-T-reactive fibril interface. We would like to add, however, that the estimation of ␣-helical content of HdeA fibrils by BeStSel was abnormally high, exceeding 50%. Secondary structure estimation of proteins by BeStSel is sensitive to small differences in protein concentration and cell pathlength (per the "Information" section of the BeStSel website), and the values shown in Table S1 may be overestimating the actual secondary structure content of the samples. ATR-FTIR spectra of HdeA fibrils (Fig. 6) in D 2 O displayed a characteristic IR absorption band centered at 1637 cm Ϫ1 , indicative of ␤-sheet structures, so the question of the accurate secondary structural content of HdeA fibrils remains open for further experimentation.
The differences in fibril morphology that could be attained by modulating the NaCl concentration of the sample suggested that there may be multiple HdeA forms that are fibrillogenic and lead to different structural forms. This idea was supported by results in Fig. 5C, which demonstrated that, at least in the presence of 150 mM NaCl, HdeA formed fibrils via at least one intermediate state that was distinguishable by CD spectral analysis. The idea that HdeA forms morphologically diverse fibrils through distinct intermediate states that are formed according to different experimental conditions needs to be probed further.
The path toward HdeA fibril formation was demonstrated to be in competition with the role of molecular chaperone, and we have found that the two pathways are mutually exclusive; HdeA is incapable of acting as a molecular chaperone in the fibril state, and the presence of unfolded proteins such as ADH prevents this chaperone from attaining the inactive fibril form. A

Reversible fibrillation of small heat shock protein HdeA
simple explanation for this result would be that the molecular interfaces for both molecular chaperone activity and fibrillogenesis either overlap or interact with each other so that simultaneous attainment of both states is impossible. However, prediction algorithms for amyloid core regions also highlight the possibilities of a more complex mechanism. When we submitted the amino acid sequence of HdeA to an online site that utilizes and compares multiple algorithms to predict potential amyloidogenic sequence regions (http://aias. biol.uoa.gr/AMYLPRED2/) 3 (23), we found a total of three sequence "hot spots" that were flagged by a plurality of prediction algorithms as potential amyloid core regions (sequences are highlighted when more than 5 of the 11 algorithms provided flag the sequence region as "amyloidogenic") (Fig. 7). The sequence region with the most "hits," corresponding to residues 52-59 (VQGIATVT) in the HdeA amino acid sequence, includes two valine residues that have been identified as residues that are necessary for the molecular chaperone activity of HdeA (Val 52 and Val 58 (24); diverting this sequence region to fibrils would likely serve to prevent HdeA from interacting with denatured protein targets. The presence of two other, lesser flagged sequence regions ( 32 AVGFA 36 and 61 AIVQA(C)T 67 ; the C is not flagged) is interesting when we consider the multiple morphologies that HdeA fibrils have displayed in our experiments. We note also with interest that the three putative fibrillogenic sequence regions may be mapped to two diverse locations in the native HdeA structure at pH 7; 32 AVGFA 36 is located at the dimer interface, whereas the other two sequences are exposed sections at the outer sides of the dimer (Fig. 7). The sequence region 32 AVGFA 36 is potentially very interesting when we consider that our experiments suggest that the formation of HdeA fibrils depends on a structural transition that occurs at pH ϳ3 (Fig. 2B), whereas the dissolution of this fibril structure occurs at pH values greater than 4.6 ( Fig.  4A). A recent study by Salmon et al. (25) explored the detailed process of HdeA acid denaturation as monitored by intrinsic tryptophan fluorescence and NMR. The authors' results using intrinsic fluorescence suggested that the acid denaturation of HdeA proceeded through two distinct transitions, the first transition occurring at a midpoint pH of 5.56 and a second transition at pH 2.81. These two transition midpoints correlate rather strongly to the dissolution and formation of HdeA fibrils, respectively, in our experiments. It should be noted that Salmon et al. (25) determined that the transition occurring at a midpoint pH of 5.56 involved the Reversible fibrillation of small heat shock protein HdeA protonation of Glu 37 , a residue immediately adjacent to the 32 AVGFA 36 sequence region and a residue closely involved in the activation of the HdeA molecular chaperone activity. By performing mutation analysis of this residue, we could conceivably probe in future experiments the relationship between fibrillation of HdeA and the amino acid sequence of this protein in more detail.
The reversible fibrillation of HdeA is an example of a relatively rare phenomenon; however, very prominent examples of such reversibility exist, such as the p53 tetramerization domain (26), the repeat domain of Pmel17 (27), and the two reversible amyloid cores RAC1 and RAC2 in the LC domain of fused in sarcoma (FUS), a protein integral to the formation of RNAenriched cellular granules (28). Interestingly, in many of the cases above, the reversibility of the fibrillogenic reaction was utilized as a biologically relevant modulable switch that lends itself to a unique function, which has resulted in the concept of "functional amyloids." It would be intriguing if E. coli HdeA was also utilizing this ability to form fibrils in a presently unknown function in the bacterial periplasm or if other components of the periplasm modulate this behavior. Fig. 8 summarizes the results of our present study. After denaturation triggered by low-pH conditions, HdeA is converted to an active chaperone that binds to acid-denatured proteins in the periplasm ("chaperone"). Under the same conditions, HdeA is also converted readily to an atypical, insoluble fibrillar form ("ϽFϾ") that is reverted to a soluble form upon another shift to neutral pH ( "ϽNЈϾ"). At present, we are not able to determine conclusively that HdeA is capable of returning completely to the original native structure (if ϽNϾ ϭ ϽNЈϾ) or is assuming another soluble form; the finding that this recovered form is capable of being reactivated as a molecular chaperone at low pH seems to argue for complete recovery of ϽNϾ, whereas the results of secondary structural analysis (Figs. 5 and 6) presently give conflicting data; the CD spectra argue for almost complete reversibility, whereas the ATR-FTIR experiments suggest that the reconstituted species is markedly distinct from native HdeA. Although we intend to probe this discrepancy in much more detail in further experiments, a possible explanation for the discrepancies seen in the ATR-FTIR experiments in Fig. 6 may lie in the fact that we used D 2 Ohydrated lyophilized samples in these measurements; the ϽNЈϾ state may be metastable and prone to reversion to the ϽFϾ state upon concentration. This outstanding question, among many other open questions, will need to be addressed in further studies.

Expression vectors and protein purification
A pET23a(ϩ)-derived expression vector for E. coli HdeA was constructed by replacing the ORF of an E. coli GroEL expres- Regions that were flagged as amyloidogenic by more than five of the 11 provided algorithms are color shaded and represented in red lettering within the sequence. The specific algorithms that scored a "hit" on a given amino acid residue are listed below the amino acid sequence and denoted by the hash symbol (#), and the specific algorithms that contributed to the consensus of the highlighted regions are shown in red letters to the left of the figure. Each shaded sequence segment is mapped onto the ribbon model structure of HdeA with the corresponding color. In the amino acid sequence, the two underlined valine residues (Val 52 and Val 58 ) are essential for the molecular chaperone activity of HdeA as determined by Wu et al. (24), and the green lettered residues Asp 20 and Asp 51 were characterized as "acid sensor" residues that dictate the denaturing transition of HdeA as determined by Foit et al. (30). The Glu 37 residue denoted in cyan is another residue characterized by Foit et al. (30) as a residue that stabilizes the dimer interface of native HdeA. Structural images of HdeA (derived from Protein Data Bank (PDB) code 1dj8 (4)) were produced using UCSF Chimera (31). The algorithms used by AMYLPRED2 to evaluate the sequence are as follows: AGGRESCAN (32), Amy-loidMutants (33), Amyloidogenic Pattern (34)  Reversible fibrillation of small heat shock protein HdeA sion vector (pETEL) previously developed in our lab (29) with the ORF of the HdeA precursor sequence. This plasmid (pET23-HdeA) was used to transform JM109(DE3) cells, and transformants were further cultivated in LB medium for use in purification of HdeA.
Purification of HdeA was performed according to a protocol based upon an original procedure described by Kern et al. (2) where initial preparation of periplasmic extracts of E. coli was performed using osmotic shock followed by sequential chromatography steps using DEAE-Sephacel and hydroxyapatite. However, in our procedure, we substituted the DEAE-Sephacel anion-exchange step with an anion-exchange column chromatography step using Resource Q, developed with a linear NaCl gradient of 0 -0.5 M, and we substituted the hydroxyapatite step with a cation-exchange chromatography step using Resource S based upon a modification introduced by Tapley et al. (11) that utilized HiTrap SP Sepharose for the second chromatography step. The buffer used in the Resource S chromatography was 20 mM sodium acetate, pH 4.0, and the column was developed using a 0 -0.7 M NaCl linear gradient. All chromatography steps were performed on an ÄKTA FPLC system.

In vitro fibril formation
Monitoring of protein samples for Thio-T-positive fibril formation was performed in 20 mM glycine-HCl buffer adjusted to the relevant pH and NaCl concentrations. Protein concentrations were set to 1 mg/ml for both ␣-Synuclein and HdeA. Samples also contain 20 M Thio-T to allow real-time monitoring of fibril extension by specific fluorescence. Reactions were prepared in 96-well plates, and sample incubation and sample measurements were performed using a PerkinElmer Life Sciences ARVO X4 fluorescence plate reader with agitation abilities. Samples were agitated for 15-min intervals interspersed with 10 s of fluorescence measurement at an excitation wavelength of 450 nm using an emission cutoff filter at 486 nm.
In experiments that involved a shift in the experimental pH from pH 2 to higher values (specifically Figs. 4 and 5A), samples were adjusted to the new pH by adding aliquots of Tris base solution to achieve the required pH. Typically, 1.6 l of 2 M Tris base solution was sufficient to shift a 150-l sample initially incubated at pH 2 to pH 7. Prior to the experiments, mock samples were used to estimate the concentration and amount of Tris base solution required to attain a given pH, and the actual pH of each sample was estimated visually by spotting small aliquots onto Phenol Purple (pH 3.4 -6.4) or Methyl Red (pH 5.4 -7.0) pH test paper (Advantec Toyo Roshi, Japan).

Transmission EM
Samples that had been monitored for Thio-T fluorescence increase as described above were diluted (10 -30-fold) and applied to collodion-covered carbon mesh disks for 90 s. Excess samples were blotted off, and the sample disks were briefly rinsed by applying 5 l of Milli-Q water and immediate blotting. Samples were next stained by application of a 15-folddiluted solution of EM Stainer (Nisshin EM Co., Japan) for 30 s followed by blotting and air drying of the completed sample. TEM micrographs were taken on a JEOL 1400Plus electron microscope operating at 80 kV.

Gel electrophoresis of HdeA species and monitoring aggregation rescue assays of ADH
The partitioning of various HdeA samples into soluble and insoluble fractions was observed by separating the soluble and insoluble species using centrifugation for 20 min at 15,000 ϫ g and subsequent analysis by 15% Laemmli SDS-PAGE. Insoluble protein was solubilized using 8 M urea when necessary. It should be noted that the results in Fig. 4C are not a quantitative representation of the amount of HdeA in the soluble and insoluble species as differences in concentration ratio are incorporated as a consequence of the protocol.
The ability of HdeA to prevent the acid-induced aggregation of ADH was tested by adding 10 M ADH samples to 100 M HdeA samples and lowering the pH of the mixture to pH 1.0 using concentrated TCA. The lower pH value in this experiment was necessary to aggregate the target ADH under these conditions. As seen in Fig. 2C, the ability of HdeA to form fibrils was unchanged at pH 1. Samples adjusted to pH 1.0 were incubated for 1 h at 25°C and then observed using SDS-PAGE as outlined above.

CD spectra
For the results in Fig. 5, A and B, HdeA samples (1 mg/ml) were prepared initially at 25°C in either 30 mM Tris-HCl buffer, pH 7.0, or 20 mM glycine-HCl buffer adjusted to the respective pH and NaCl concentrations. Before measurement, each sample was diluted 20-fold to a concentration of 50 g/ml with Milli-Q water. Dilution using water lowered the overall absorbance of each sample and allowed reliable measurements to be made up to 190 nm. Prior to incorporating this protocol, we confirmed that the dilution step would not result in pH shifts to values higher than 4, and the results in Fig. 4A suggest that changes in fibril structure and morphology as a result of this shift would be unlikely. Measurements were taken in 1-mmpath quartz cells on a Jasco J-820 spectropolarimeter. Spectra are averages of eight individual scans. The secondary structural composition of each sample as suggested by the CD spectra was estimated by using an online spectrum analysis tool, BeStSel (http://bestsel.elte.hu/index.php) 3 (18,43).
For the time-lapse scans shown in Fig. 5C, samples were prepared in either 150-l aliquots in a 96-well plate (for the 150 mM NaCl samples) or in 1-ml aliquots in 13 (inner diameter) ϫ 100-mm glass test tube (for the 0 mM NaCl samples; P-type test tube, Nichiden-Rika Glass). Test tube samples were agitated in a constant-temperature water bath. For both samples, the initial concentration of HdeA was 1 mg/ml. Before measurement, aliquots were diluted with the respective original buffers (50fold for the 150 mM samples (HdeA concentration during measurement, 20 g/ml) and 20-fold for the 0 mM samples (HdeA concentration during measurement, 50 g/ml)).

AFM
AFM measurements of HdeA fibrils were performed on a Digital Instruments Nanoscope IVa at room temperature. Samples were prepared by applying 10 -20 g/ml aliquots onto freshly cleaved mica stages and incubating at room temperature for 1 h. The samples were washed by application of Milli-Q water to the stage and samples were air-dried overnight before Reversible fibrillation of small heat shock protein HdeA measurement. Measurements were performed using a Bruker NCHV-10V cantilever in tapping mode.

ATR-FTIR
ATR-FTIR spectra were measured on a PerkinElmer Life Sciences Spectrum 65 FT-IR spectrophotometer equipped with the universal ATR sampling accessory (diamond/ZnSe crystal). Samples of HdeA in various configurations and a reference sample of ␣-Synuclein fibril form (ϳ20 g) in their respective buffers were lyophilized for ϳ25 h to remove water and first measured directly on the apparatus to estimate the absorbance of the sample. Eight scans (4 cm Ϫ1 resolution) with CO 2 /H 2 O atmospheric suppression applied were averaged. Next, 2 l of 99.9% D 2 O (Cambridge Isotope Laboratories) was added directly onto the samples, and pressure was reapplied. After a 5-min incubation, the samples were remeasured to obtain the spectra shown in Fig. 6. The analysis functions of the software package (Spectrum version 10.03.09) were used to obtain second derivative spectra (number of data points used as analysis window: 5).