Mapping Temperature-induced Conformational Changes in the Escherichia coli Heat Shock Transcription Factor σ32 by Amide Hydrogen Exchange*

Stress conditions such as heat shock alter the transcriptional profile in all organisms. In Escherichia coli the heat shock transcription factor, σ32, out-competes upon temperature up-shift the housekeeping σ-factor, σ70, for binding to core RNA polymerase and initiates heat shock gene transcription. To investigate possible heat-induced conformational changes in σ32 we performed amide hydrogen (H/D) exchange experiments under optimal growth and heat shock conditions combined with mass spectrometry. We found a rapid exchange of around 220 of the 294 amide hydrogens at 37 °C, indicating that σ32 adopts a highly flexible structure. At 42 °C we observed a slow correlated exchange of 30 additional amide hydrogens and localized it to a helix-loop-helix motif within domain σ2 that is responsible for the recognition of the -10 region in heat shock promoters. The correlated exchange is shown to constitute a reversible unfolding with a half-life of about 30 min due to a temperature-dependent decrease in stabilization energy. We propose that this gradual decrease in stabilization energy of domain σ2 with increasing temperatures facilitates the unfolding of σ32 by the AAA+ protease FtsH thereby decreasing its half-life. Taken together our data show that the σ2 domain of σ32 can act as a thermosensor, which might be important for the heat shock regulation.

The heat shock response is a ubiquitously conserved protective mechanism of cells to cope with stress-induced damage in proteins. In Escherichia coli heat shock to 42°C induces a 10 -20-fold transient increase in the expression of about 20 heat shock genes through the stress-dependent increase in the level and activity of the heat shock transcription factor 32 (1). In vivo 32 is very unstable with a half-life of approximately 1 min at 30°C (2). Degradation of 32 in vivo requires the AAAϩ protease FtsH and the DnaK chaperone system (3,4). Imme-diately after temperature up-shift the synthesis as well as the half-life of 32 transiently increase by 10-and 8-fold, respectively. The transient stabilization of 32 is believed to be caused by the competition of unfolded proteins with 32 for binding to DnaK and FtsH (1,5,6). Under steady state conditions at 42°C the half-life of 32 decreases to 10 -15 s, probably due to an increase in the levels of available DnaK and FtsH.
However, it remains an important open question whether 32 itself can act as a thermosensor, similar to the heat shock transcription factor HSF1 of Drosophila (7), by undergoing conformational alterations in response to heat shock. To investigate this question we determined the folding status of 32 under temperatures of optimal growth (37°C) and under heatstress conditions (42°C).
In recent years amide hydrogen exchange combined with mass spectrometry has become an important method for studying the conformational properties of proteins (8 -11). In combination with pepsin digestion this method can resolve conformational changes down to the peptide level (8). We therefore built a high performance liquid chromatography mass spectrometry (HPLC-MS) 1 setup that allowed us to monitor the incorporation of deuterons into full-length 32 . Using an in-line column packed with immobilized pepsin for rapid reproducible digestion we were able to localize slow and fast exchanging regions within the entire sequence of 32 . The results indicate that 32 has an unusually high degree of flexibility at 37°C, and that heat treatment to 42°C leads to unfolding of a stable subdomain. Our data support a role for 32 as a thermosensor.
Amide Hydrogen Exchange Measurements-Amide hydrogen exchange was initiated by a 50-fold dilution of 100 pmol 32 into D 2 O containing 25 mM HEPES, pD 7.6, 50 mM KCl, 5 mM MgCl 2 , and 5% glycerol at 37 or 42°C. After various times (0.2-90 min), the exchange reaction was quenched by decreasing the temperature to 0°C and the pH to 2.2 with 500 mM K x H y PO 4 .
In-line Peptic Digestion/Rapid Desalting HPLC Setup-The setup consisted of two HPLC pumps (Agilent 1100 series; Waldbronn, Germany), a Rheodyne injection valve (Model 7725i; Rheodyne, Rohnert Park, CA) with a 200-l stainless steel sample loop, and a 2-position/ 10-port valve with microelectric actuator (Valco C2-1000EP6; Schenkon, Switzerland). A schematic drawing of the setup is shown in Fig. 1. Pump A delivered the solvent for desalting (200 l/min, 0.05% trifluoroacetic acid) and pump B for elution (20 l/min, 70% acetonitrile, * This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB388, SFB352, Leibnizprogramm) (to B. B. and M. P. M.) and the Fonds der Chemischen Industrie (Kékulé scholarship) and Marie Curie Short Term Fellowship (to W. R.), and DABIC (Danish Biotechnology Instrument Center) (to P. R. and T. J. D. J.). 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.
¶ To whom correspondence may be addressed: Zentrum fü r Molekulare Biologie Heidelberg, Im Neuenheimer Feld 282, 69120 Heidelberg, Germany. Tel.: 49-6221-546829; Fax: 49-6221-545894; E-mail: M. Mayer @zmbh.uni-heidelberg.de or bukau@zmbh.uni-heidelberg.de. 0.05% trifluoroacetic acid). Quenched samples were loaded via the injection valve and pushed through the pepsin column (1 ϫ 20 mm) by pump A. The resulting peptic fragments were immediately trapped on a self-packed reversed-phase column (0.8 ϫ 2 mm, Poros R1) for desalting and concentration. After 2 min the 10-port valve was switched to elute the peptides with organic solvent (pump B) directly into the electrospray ion source. At the same time the pepsin column was back-flushed by pump A. The digestion, desalting, and elution required less than 3 min. The whole setup was immersed in an ice bath to minimize back-exchange. For kinetic measurements of the full-length protein the pepsin column was omitted.
Mass Spectrometry and Data Treatment-Electrospray ionization mass spectra were acquired on a quadrupole time-of-flight instrument (Q-TOF Ultima, Micromass, Manchester, UK, for full-length 32 , and QSTAR Pulsar, ABI SCIEX, Toronto, Canada, for peptic peptides). Protein mass spectra were deconvoluted with the MaxEnt software (Micromass). Peptic peptides of 32 were identified on the basis of their MS/MS spectra. The deuterium content of the peptides was calculated by using the average mass difference between the isotopic envelopes of the deuterated and the undeuterated peptides. The average masses were determined with the MagTran software (13).

Measurement of the Kinetics of Amide Hydrogen Exchange-
To determine the solvent-accessible sites, temperature stability, and conformational changes of 32 we wanted to probe the overall kinetics of amide hydrogen exchange and then, under identical conditions, localize the fast and slow exchanging regions. Therefore, we needed to efficiently measure the masses of the full-length protein and of peptic peptides generated after the exchange reaction is quenched. To achieve this and to minimize back-exchange during digestion and desalting we built up an on-line "rapid desalting"-HPLC setup consisting of two HPLC pumps, a 2-position/10-port valve, and a 1 l reversed-phase trap column. This system was coupled on-line to an electrospray ionization quadrupole time-of-flight mass spectrometer ( Fig. 1). For the full-length protein, amide hydrogen exchange was initiated by diluting 32 into D 2 O-buffer. The exchange reaction was stopped by pH shift to 2.2 and temperature shift to 0°C. Under these conditions amide hydrogens exchange on average with half-lives of 60 -90 min (14), whereas deuterons incorporated into the side chains and amino/carboxyl termini rapidly exchange back to protium. Thus, the mass increase observed after deuteration and desalting reflect primarily deuterium incorporation at the backbone amide groups. For the localization of the fast and slow exchanging regions a 16-l column packed with immobilized pepsin was used to generate fragments that were subsequently desalted on the trap column and analyzed by MS. The identity of the peptides was established separately by digestion of fully protiated 32 under identical conditions with subsequent collision-induced dissociation tandem mass spectrometry. The very small volumes of the pepsin column (16 l) and the trap column (1 l) reduced the time for digestion and desalting to less than 3 min, thereby minimizing the time during which back-exchange could occur.
Full-length 32 Protein Amide Hydrogen Exchange Kinetics-At 37°C 32 exchanged about 190 of its 286 amide protons for deuterons within 1 min (Fig. 2a). The degree of deuterium incorporation increased over time leaving maximally 60 -70 amide hydrogens (Ͻ25% not counting back-exchange) that did not exchange under the experimental conditions within 60 min. The overall kinetics could be fitted to a biexponential equation (Fig. 2b and Table I 32 . After a 30 min exchange at 42°C the bimodal distribution is clearly visible for both the original and the deconvoluted data. b, kinetics of deuterium incorporation at 37 (open squares) and 42°C (closed triangle, lower mass peak; closed inverted triangle, higher mass peak). The curves are fits of a biexponential equation to the data (see Table I for parameters). c, correlated deuterium incorporation; the deconvoluted peaks at 42°C were quantified using MagTran software, and percent of the higher mass peak is plotted against the time; the curve is a fit of a first order rate equation to the data. b and c, representative data of at least 4 independent experiments are shown. exchanging hydrogens participate in a stable structure of specific hydrogen bonds, which is a surprisingly small number. For well folded globular proteins about 50% of the amide hydrogens are found to be protected against exchange after 60 min (examples in Ref. 15). 2 These data can be interpreted in two ways. 32 either is a loosely folded protein with a low secondary structure content, or its structure has a high degree of flexibility.
At 42°C the initial exchange kinetics was similar albeit slightly faster (205 hydrogens in the order of ϭ 7 s, 22 hydrogens with ϭ 9.5 min). However, after about 10 min the appearance of a second species was observed that had incorporated an additional 30 deuterons. The amount of this second species increased at the cost of the first species with a half-life of about 30 min (Fig. 2, a and c). In the second species less than 30 hydrogens (10% not counting back-exchange) were protected against exchange.
This behavior could indicate a slow loss of secondary structure due to thermal denaturation. To test this possibility we pre-incubated 32 at 42°C for one hour before diluting the protein into D 2 O. The exchange kinetics under these conditions was, however, identical to the kinetics of the protein measured immediately after temperature shift (data not shown). These data indicate that the second species is not due to denaturation but to a reversible unfolding of a small structural motif. Because no intermediates between the first and the second species were observed, the exchange kinetics for the second species occurred in a correlated fashion, i.e. the refolding rate is much slower than the exchange rate (see "Discussion" for explanation).
Localization of the Slow and Fast Exchanging Regions by Peptic Digestion-After deuteration, the 32 protein was digested under quenched conditions using the in-line pepsin column. Fig. 3 shows representative segments of the mass spectra of a fully protiated sample (bottom) and samples deuterated for various periods at 37 and 42°C. In the fully protiated state each peptide is represented by several peaks on the mass-tocharge (m/z) scale due to the isotopic natural abundance. After incubation in D 2 O the isotope envelope of most peptide peaks was shifted to higher m/z values due to incorporation of deuterons. For example, the peptides B, C, and D shown in Fig. 3 (m/z 867.69 5ϩ , aa 121-155; m/z 877.46 5ϩ , aa 257-294; m/z 880.45 2ϩ , aa 2-18) exchanged all amide hydrogens within 12 s under the conditions used. In contrast, the peptides A and E (m/z 866.16 3ϩ , aa 79-101; m/z 885.94 4ϩ , aa 19-49) exchanged hydrogens much slower, leaving after 90 min 11 and 9 hydrogens unexchanged ( Fig. 3 and data not shown). At 42°C the exchange kinetics for peptide E was similar to that observed at 37°C. In contrast, for peptide A the exchange kinetics is very different at 42°C as compared with 37°C. After 10 min deu-teration at 42°C a bimodal isotope distribution appears. As exchange proceeds, the abundance of the lower mass isotope distribution decreases while a corresponding increase is observed for the higher mass isotope distribution. Such a bimodal isotope distribution is the result of a local unfolding with a lifetime sufficiently long to allow complete exchange of several amide hydrogens within a single opening event. The lower mass isotope distribution represents a population of 32 molecules that has not yet unfolded whereas the higher mass population has unfolded and undergone the correlated exchange. Close inspection of the data revealed that such a correlated exchange in the part of the protein represented by peptide A also occurs at 37°C albeit with much slower kinetics (see 60 min trace, Fig. 2a, left panel: shoulder to the higher mass range; 45 and 60 min traces, Fig. 3: peaks between peptide A and B).
Using our HPLC-MS setup we were able to map the kinetic exchange profile for peptic peptides of 32 covering 99% of the sequence of the entire protein (Fig. 4a). Overlapping peptides that differed pairwise only by extensions at the C terminus were used to improve the resolution of the deuteron incorporation map (16,17). For example, peptides 79 -89 did not incorporate any deuterons at 37°C within 10 s, whereas peptides 79 -101 incorporated three deuterons under the same conditions. These deuterons can then be localized to the segment spanning residues 90 -101. For all pairs of peptides the intrinsic chemical exchange rate for the C-terminal amide hydrogen in the short peptide at quench conditions was found to be comparable with that of the corresponding residue in the extended peptide. In this way it was ensured that the deuterium occupancy for the non-overlapping C-terminal sequence extension was correctly assigned. Fig. 4b shows a plot of the relative deuteron incorporation per amide after 10 s, 2 min, and 90 min at 37°C and after 90 min at 42°C against the primary sequence of 32 . The N terminus (residues 2-18) and the Cterminal half of the protein (residues 121-293) incorporate deuterons rapidly. The small decrease in relative exchange at the C terminus is due to the more rapid back-exchange of the amide hydrogens within the His 6 -tag as calculated according to Englander's algorithm (14). In contrast to the exchange behavior of the N terminus and the C-terminal half of the 32 protein, two regions, residues 79 -89 and 104 -115, were relatively well protected at 37°C and showed correlated exchange at 42°C (Fig. 4b, arrowheads).
Homology Modeling of the 32 Structure-To visualize fast and slow exchanging regions of 32   a The centroids of the lower mass peak (see Fig. 2a) were fitted to the biexponential equation. b The centroids of the higher mass peak were fitted to the biexponential equation. Because the higher mass peak was not visible before ten minutes no value for the fast rate is given.
c The amounts of higher and lower mass peaks at 42°C were determined by integration of the peak areas, and the change of the percentage of the higher mass peak over time was fitted to a first order rate equation.
tural model of 32 residues 48 -284, not covering the N terminus, although the homology between 32 and 70 extends over the entire sequence of 32 including the N-terminal part. The reason for the inability to model the N-terminal part in this initial approach was found, by multiple sequence alignment, to be an insertion of 69 residues in T. thermophilus 70 that does not exist in E. coli 32 . In the 70 structure this insertion forms a separate part of the 2 domain whereby the N and C termini of the insertion are close together in space. We therefore removed the coordinates of the intervening residues 111 to 179 from the Protein Data Bank file, connected the C and N termini of residues 110 and 180, and minimized the free energy of this structure using the Discover module of the Insight II program (Accelrys Inc., San Diego, CA). The resulting Protein Data Bank file was used as a template for the modeling of 32 via the SWISS-MODEL program and yielded a structure model including residues 9 -284. The amino acids corresponding to residues 1-8 were not present in the 70 structure. Based on the secondary structure prediction program PHD, these first amino acids were unstructured, and according to our amide hydrogen exchange data solvent exposed and not involved in any secondary structure stabilizing hydrogen bonds. We therefore added them manually in an extended conformation. Fig. 4c shows the structural model colored according to the degree of amide hydrogen exchange at different time points and temperatures with blue indicating 0 -20%, green 20 -45%, yellow 45-70%, and red 70 -95% deuterium incorporation. 95% was the maximal amount of deuterium incorporation measured due to proton back-exchange during the desalting step. The N-terminal 2 domain is well structured and shows little deuteron incorporation at 37°C, whereas the C-terminal half of the model contains very little secondary structure and rapid deuteron incorporation is observed. Although this model of 32 represents the RNA polymerase-bound form, our data are fully consistent with the model showing fast exchange in unstructured regions and slow exchange in the tightly folded ␣-helices of domain 2 responsible for promoter recognition at the Ϫ10 region. The precision with which this model fits our amide hydrogen exchange data is demonstrated by a helix-loop-helix element in domain 2 where deuteron incorporation into the loop region (Fig. 4, b and c, open arrow head) was significantly faster than into the ␣-helices (Fig. 4, b and c, closed arrow  heads). The domain 4 , which recognizes the heat shock promoter at the Ϫ35 region, forms a three helix bundle in our model but does not seem to be very tightly packed. Our exchange data show that the helix stabilizing H-bonds in this domain break and reform rapidly allowing deuterium incorporation. After 90 min at 42°C even in domain 2 significant deuteron incorporation is observed, demonstrating an increased flexibility and loosening-up of this domain.

Mass Spectrometry and Amide
Hydrogen Exchange-In this study we developed an on-line rapid desalting-HPLC-electrospray ionization mass spectrometry setup that allows the rapid and efficient analysis of the kinetics of amide hydrogen exchange of full-length proteins and the on-line generation of peptic fragments for the localization of fast and slow exchanging regions.
In recent years amide hydrogen exchange is being increasingly used for probing protein structure and stability (8 -10) because it does not depend on the presence of proteolytic cleavage sites or fluorescent markers. Several methods are used to detect amide hydrogen exchange including NMR, matrix-assisted laser desorption/ionization time-of-flight-mass spectrometry, and electrospray ionization mass spectrometry. Our rapid desalting HPLC-MS setup has several advantages over other methods. First, low protein concentrations in the M range can be used as compared with the mM range used in NMR. This was essential in the case of 32 because the protein tended to aggregate at higher concentrations. Second, several coexisting conformations can be observed at the same time, which allowed the detection of the correlated exchange at 42°C. In NMR an average exchange rate is determined. Third, sample handling under quenched conditions was fully automated, which guaranteed an exact timing of the digestion and desalting and therefore easily reproducible results. In the case of matrixassisted laser desorption/ionization-MS pipetting must be very accurate to obtain comparable results. Finally, the information covering close to 100% of the entire protein sequence can be acquired with just one spectrum. Our setup generated highly reproducible peptic digests of the 32 protein as clearly visible in Fig. 3 and minimized back-exchange to generally less than 10%. Parallel to our work, Smith and co-workers (22) developed a similar setup. In our study, no adjustment was made for deuterium loss during quenched conditions as described previously (17), because we are mainly interested in the difference between the exchange kinetics of 32 at 37 and 42°C and not an exact determination of exchange rate constants of individual peptides.
Overall Deuteron Incorporation into 32 -We analyzed the exchange kinetics of full-length 32 and the propensity for deuterium incorporation within 99% of its primary sequence. Our data indicate that 32 is a relatively loosely folded or highly flexible protein already at the regular growth temperature of E. coli (37°C). The rapid incorporation of deuterons into the full-length protein as well as the localization of slow and fast exchanging regions showed that only domain 2 is protected against exchange indicating a stable domain. Except for the part that exchanged in a correlated fashion, the stability of 32 at 42°C was similar to its stability at 37°C as indicated by similar overall exchange kinetics. Heat shock conditions (42°C) caused a correlated exchange of several amide hydrogens in domain 2 with a mean half-life of about 30 min that is indicative for a local, reversible unfolding event.
These data fit well to our model of the 32 structure despite the fact that this model was derived by homology modeling to the RNA polymerase-bound form of 70 (18) and not to the uncomplexed structure of this protein. Due to the high sequence coverage and to the precision with which deuteron incorporation could be determined (see helix-loop-helix element), our setup can be used to verify structural models of proteins where no crystal structure is available and to analyze the dynamics of protein structures. A surprising result was that domain 4 , which is responsible for recognition of the Ϫ35 region of heat shock promoters, is not better protected against deuteron incorporation and therefore not tightly folded. Maybe the structure of this domain is stabilized upon binding to RNA polymerase and/or DNA.
Using circular dichroism and fluorescence spectroscopy on 32 , Chattopadhyay and Roy (23) recently observed that the molar elipticity between 200 and 230 nm decreases between 35 and 42°C, whereas the intrinsic tryptophan fluorescence stayed constant. They speculated that the region between residues 181 and 208 unfolds upon temperature up-shift. In contrast, we could now map precisely the unfolding region at the peptide level to residues 79 -105 and demonstrate that the  32 . a, amino acid sequence of 32 and peptic peptides used for the analysis (underlined). Sequence coverage was close to 100%. b, the propensity for amide hydrogen exchange at different times and temperatures is plotted against the primary sequence of 32 . Overlapping peptides were used to increase the resolution of the deuteron incorporation map. Correlated exchange at 42°C occurred mainly in two regions indicated with closed arrowheads. The intervening loop region is indicated with an open arrowhead. c, secondary structure representation of the model of the 32 structure colored according to their exchange behavior at 37°C after 10 s, unfolding event occurs in a cooperative and reversible manner. This clearly shows the power of amide hydrogen exchange as compared with global structural methods as circular dichroism spectroscopy.
In uncomplexed 32 the domains 2 and 4 may pack against each other as suggested by the recent structure of E in complex with a fragment of its anti--factor RseA (24). However, such a packing does not stabilize the domains 3 and 4 because our data clearly demonstrate that 32 adopts loosely folded and highly flexible conformations allowing rapid exchange of the majority of its amide hydrogens. These data and our model show that only the N-terminal 2 domain is tightly folded and the C-terminal half resembles an unfolded protein. This is consistent with the short in vivo half-life of 32 at all temperatures and with the fact that 32 is prone to aggregation at higher concentration. Furthermore, our data explain why 32 constitutes a substrate for Hsp70 chaperones, which otherwise recognize unfolded proteins.
Correlated Exchange (EX1)-To understand the physiological implications of the observed correlated exchange in the helix-loop-helix element of domain 2 the exchange mechanisms must be considered. For amide hydrogen exchange to occur a structural-specific H-bond between an amide hydrogen and an H-bond acceptor has to open, thereby allowing the catalyst (hydroxide ion) to attack. Such an opening can be a global or local unfolding of a secondary structural element within the native structure of a protein according to Equation 1, with F and U representing the folded and unfolded conformation of the structural element and k 1 , k Ϫ1 , and k 2 being the unfolding, refolding, and intrinsic chemical exchange rate constants. There are two limiting cases for the observed exchange rates, called EX1 and EX2. In the EX1 case, the refolding is rate-limiting for the overall exchange kinetics (k Ϫ1 Ͻ Ͻ k 2 ), and the amide hydrogens exchange in a correlated manner in each opening event. The observed exchange rate (k obs ) is then equal to the unfolding rate (k obs ϭ k 1 ). The signature for EX1 kinetics is a bimodal isotope distribution in peptide mass spectra and the occurrence of two separate, interconverting mass peaks in protein mass spectra. Under native state conditions, however, EX2 is most commonly observed because the rates for unfolding/refolding are generally much greater than the intrinsic chemical exchange rate (k Ϫ1 Ͼ Ͼ k 2 ; k obs ϭ k 2 ϫ k 1 /k Ϫ1 ) (25). The occurrence of two distinct interconverting mass peaks observed in the 32 protein mass spectra (Fig. 2, 42°C, 30 min) as well as the bimodal isotope distributions in the mass spectra of some of its peptic peptides (peptide A in Fig. 3 and data not shown) clearly indicate a correlated exchange according to the EX1 mechanism. This correlated exchange observed here in 32 is interesting from the following structural and physiological points of view.
Usually EX1 kinetics are observed in the presence of chemical denaturants because they decrease the refolding rate k Ϫ1 without affecting the intrinsic chemical exchange rate k 2 . Refolding of a protein implicates the formation of H-bonds and a hydrophobic core, which usually occurs very fast in the absence of denaturants (ns to s; see Ref. 26). Because the intrinsic chemical exchange rates are in the ms time scale, EX1 is rarely observed under native state conditions (27,28). Therefore, some property of the domain 2 must impose a slow ratelimiting step on the refolding reaction. Despite extensive analysis of the existing structures of -factors and our structural model of 32 , it remained unclear which property of domain 2 and in particular the highly conserved helix-loop-helix element is responsible for the correlated exchange behavior. The stability of this structural element, however, could be estimated. Because k Ϫ1 is at least five-times smaller than the intrinsic chemical exchange rate k 2 (for 95% exchange to happen before refolding occurs), k 1 /k Ϫ1 Յ k 1 /(k 2 /5) and the resulting ͉⌬G͉ Յ ͉R ϫ T ϫ ln(k 1 /(k 2 /5))͉ with R being the universal gas constant and T the absolute temperature. With the values for k 1 (3.6 ϫ 10 Ϫ4 s Ϫ1 ), as determined from the kinetics of the correlated exchange, and k 2 (50 s Ϫ1 ), as calculated according to Englander's algorithm using the HXPep-program (Ref. 14, courtesy Z. Zhang), ⌬G at 42°C can be calculated to ͉⌬G͉ Յ ͉ Ϫ 27 kJ/mol͉. At 37°C the unfolding rate k 1 was much smaller than at 42°C, and the helix-loop-helix element was therefore more stable. At intermediate temperatures the correlated exchange occurred with intermediate rates, and the stability of the helix-loophelix element was between the stability at 37°C and 42°C. This structural element seems to be able to "measure" temperature by a gradual increase in unfolding rate, which is equivalent to a gradual decrease in activation energy necessary for its unfolding. Because refolding is comparatively slow, the helix-loop-helix element could act like a metastable switch that becomes more labile with increasing temperatures.
What are the physiological implications of these observations? 32 is a very unstable protein with a half-life of 1 min at 30°C and 10 -15 s at 42°C in vivo. Recently, it was shown that the membrane AAAϩ protease FtsH, which is mainly responsible for the degradation of 32 , does not have a robust unfolding activity and that the thermodynamic stability of the substrate protein is rate-limiting for degradation by FtsH (29). In vitro, FtsH degrades 32 with a half-life of Ͼ60 min at 35°C and 40 min at 44°C (30). These data are reflected very closely by our exchange kinetics, suggesting that in vitro the unfolding kinetics of the helix-loop-helix element may be rate-limiting for the degradation of 32 by FtsH. Consistent with this view is the fact that FtsH degrades proteins processively from one terminus, whereby it needs 5-20 aa for an initial interaction (31)(32)(33). For 32 it has been shown that the C terminus is not responsible for degradation by FtsH suggesting that degradation proceeds from the N terminus (12). According to our data and our model of 32 , the N terminus contains a short unfolded stretch, which could be sufficient for recognition by FtsH, followed by the only stable domain 2 .
In contrast to in vitro degradation data, in vivo the half-life of 32 is 10 to 15 s at 42°C suggesting additional factors may aid degradation of 32 by promoting the unfolding of the Nterminal domain. In fact, the lack of significant deuteron incorporation at 37°C within 60 min and the slow in vitro unfolding half-life of 30 min at 42°C indicate a relatively high stability of this structural element. Furthermore, high stability is very often interrelated to a high refolding rate. The fact that the helix-loop-helix element combines a low unfolding rate with a low refolding rate may facilitate regulation of the heat shock response because once unfolded, degradation of 32 by FtsH is not hampered by rapid refolding.
It has been shown that 32 production after temperature up-shift is increased due to the melting of the secondary structure of its encoding mRNA and consequently increased translation rates (34,35). The 32 encoding mRNA therefore has a built-in thermosensor. Our data now suggest that the 32 protein also has a built-in thermosensor that counteracts the 32 encoding mRNA. This mechanism of temperature-dependent synthesis and degradation rates may contribute to the stabilization of the 32 steady-state levels at all temperatures.