Marked Instability of the ς32 Heat Shock Transcription Factor at High Temperature

The heat shock response inEscherichia coli depends on a transient increase in the intracellular level of ς32 that results from both increased synthesis and transient stabilization of normally unstable ς32. Although the membrane-bound ATP-dependent protease FtsH (HflB) plays an important role in degradation of ς32, our previous results suggested that several cytosolic ATP-dependent proteases including HslVU (ClpQY) are also involved in ς32 degradation (Kanemori, M., Nishihara, K., Yanagi, H., and Yura, T. (1997) J. Bacteriol. 179, 7219–7225). We now report on the ATP-dependent proteolysis of ς32 by purified HslVU protease and its unusual dependence on high temperature: ς32 was rapidly degraded at 44 °C, but with much slower rates (∼15-fold) at 35 °C. FtsH-dependent degradation of ς32 also gave similar results. In agreement with these results in vitro, the turnover of ς32 in normally growing cells at high temperature (42 °C) was much faster than at low temperature (30 °C). Taken together with other evidence, these results suggest that the ς32 level during normal growth is primarily determined by the stability (susceptibility to proteases) and synthesis rate of ς32 set by ambient temperature, whereas fine adjustment such as transient stabilization of ς32 observed upon heat shock is brought about through monitoring changes in the cellular state of protein folding.

When cells and organisms are exposed to high temperature, synthesis of a set of heat shock proteins is rapidly induced. The induction generally occurs at the transcriptional level mediated by specific transcription factors. In Escherichia coli, the level of the heat shock transcription factor 32 (encoded by the rpoH gene) rapidly and transiently increases upon temperature upshift and directs RNA polymerase to transcribe the heat shock genes encoding molecular chaperones (GroEL, DnaK, etc.), ATP-dependent proteases (Lon (La), ClpXP, HslVU (ClpQY), and FtsH (HflB)), and other heat shock proteins (1)(2)(3)(4). The increase in the 32 level depends on both increased synthesis and stabilization of normally unstable 32 (half-life of ϳ1 min) (5). Production of abnormal proteins such as those containing amino acid analogs or heterologous proteins can also induce heat shock proteins and mimic the heat shock response (6 -8). However, in the latter cases, the 32 level increases as the result of stabilization and not of increased synthesis of 32 (9,10). Thus, stabilization and enhanced synthesis of 32 observed upon temperature upshift represent two distinct events that presumably involve different signaling pathways (3,4).
Upon temperature upshift from 30 to 42°C, the rate of 32 synthesis increases ϳ10-fold within 3-4 min (5). The induction occurs at the translational level mediated by the secondary structure of the 5Ј-portion of rpoH mRNA (11)(12)(13). Recent work showed that high temperature directly disrupts the mRNA secondary structure, perhaps without involvement of cellular factors, leading to enhanced ribosome entry and initiation of translation (14). Besides translational induction, marked stabilization of 32 (8-fold) occurs for the first 4 -5 min upon heat shock, followed by rapid destabilization (5). The DnaK chaperone team (DnaK, DnaJ, and GrpE) is required for the rapid 32 degradation because 32 is markedly stabilized in dnaK/dnaJ/ grpE mutants (15,16). Although the exact roles of these chaperones in 32 turnover remain unknown, transient stabilization of 32 has been thought to occur by titrating the chaperones away from 32 by unfolded proteins accumulating at high temperature since these chaperones are also involved in dealing with unfolded or misfolded proteins (17,18). Accordingly, free DnaK and/or DnaJ chaperones that can engage in the turnover of 32 have been proposed to act as a "cellular thermometer" that should monitor changes in the folding state of the cell (19,20). In agreement with these proposals, DnaK/DnaJ chaperones were shown to be normally limiting in vivo: a slight increase or decrease in the chaperones causes a decrease or increase in 32 (and consequently, heat shock proteins), respectively (21). A membrane-bound ATP-dependent metalloprotease (FtsH) was the first protease shown to degrade 32 in vivo and in vitro (22,23). More recent data suggested that a cytosolic ATP-dependent protease (HslVU) along with other proteases (ClpAP (Ti) and Lon) can also participate in the turnover of 32 in vivo (24). HslVU is a two-component ATP-dependent protease consisting of a catalytic subunit (homododecamer of HslV (ClpQ) and an ATPase subunit (homohexa-or heptamer of HslU (ClpY)); one HslV complex is flanked by HslU complexes to form structures resembling the ClpAP protease and the eukaryotic 26 S proteasome (25,26).
In this report, we examined degradation of 32 in vitro by the HslVU protease and found that purified HslVU can degrade 32 in an ATP-dependent manner. Furthermore, the susceptibility of 32 to HslVU-or FtsH-mediated proteolysis in vitro markedly increased at high temperature. The marked instability of 32 at high temperature was also demonstrated in vivo, suggesting that the susceptibility of 32 to proteases reflecting ambient temperature plays a critical role in the regulation of 32 and the heat shock response in E. coli.
Plasmids pKV1025 and pKV1022 (27) carry the hslV and hslU genes, respectively, under control of the trc promoter on the pTrc99A vector (Amersham Pharmacia Biotech). A promoterless intact rpoH gene (EcoRV-NsiI fragment) was placed under control of the trc promoter on pTrc99A to obtain pKV1278. pMAL-c-sulA carrying a malE-sulA fusion gene was kindly donated by Y. Ishii and Y. Kato (Kyushu Institute of  Technology).
Media, Chemicals, and Reagents-L broth (29) was used for generally growing cells; ampicillin (50 g/ml) was added when necessary. The synthetic medium used was medium E (29) supplemented with 0.5% glucose and 2 g/ml thiamine; ampicillin was used at 10 g/ml. IPTG (final concentration, 1 mM) was used to induce synthesis of HslV, HslU, 32 , and MBP-SulA directed by the trc or tac promoter. ␣-Casein (bovine, 70% purity) was purchased from Sigma. L-[ 35 S]Methionine (29.6 TBq/mmol) was obtained from American Radiolabeled Chemicals. Other chemicals were obtained from Nacalai Tesque, Kyoto, or Wako Pure Chemicals (Osaka, Japan). Antiserum against FtsH was kindly supplied by Y. Akiyama (Kyoto University).
RNA polymerase core enzyme consisting of ␤, ␤Ј, and two ␣-subunits was purified from strain KY1603 lacking 32 . Cells were grown to late-log phase in L broth at 37°C, harvested by centrifugation, suspended in buffer A, and kept on ice for 30 min. After addition of sodium deoxycholate to 0.05%, cells were sonicated and centrifuged at 75,000 ϫ g for 90 min, and proteins were precipitated by ammonium sulfate (0.35 g/ml). The pellet was resuspended and dialyzed against buffer B and loaded onto a HiLoad Q-Sepharose column (Amersham Pharmacia Biotech). Proteins were eluted with a linear gradient of NaCl, and the fractions containing RNA polymerase were precipitated by ammonium sulfate (0.35 g/ml). The pellet was resuspended and dialyzed against buffer I and loaded onto a HiTrap heparin column, and proteins were eluted with a linear gradient of KCl. The fractions containing RNA polymerase were dialyzed against buffer E and run through a HiPrep Sephacryl S-300 column (Amersham Pharmacia Biotech), and RNA polymerase core enzyme was concentrated and stored at Ϫ70°C.
MBP-SulA fusion protein was purified from KY2691 cells harboring pMAL-c-sulA. Cells were grown to mid-log phase in L broth containing ampicillin at 37°C, and production of MBP-SulA was induced by IPTG. After 1 h of incubation, cells were harvested by centrifugation, resuspended in buffer E, and sonicated. MBP-SulA fusion protein was puri-fied according to the instruction manual (New England Biolabs Inc.), except that buffer E was used as the column buffer. All purified proteins were Ͼ90% pure as estimated by SDS-PAGE followed by staining with Coomassie Brilliant Blue, and their concentration was determined by Bradford protein assays (Bio-Rad) (32).

ATP-dependent Degradation of 32 by the HslVU Protease-
Our previous data in vivo suggested that besides FtsH, the cytosolic ATP-dependent proteases including HslVU participate in the turnover of 32 (24). To test whether HslVU can directly degrade 32 , we purified HslV and HslU separately from a pair of multiprotease-deficient KY2691 mutants lacking all known cytosolic ATP-dependent proteases, but harboring an expression plasmid for hslV or hslU. This permitted us to obtain HslV and HslU preparations with no contamination of other cytosolic ATP-dependent proteases. Although these strains carried the wild-type ftsH gene, no detectable amount of FtsH was found in either the HslV or HslU preparation used as judged by immunoblotting (data not shown). When purified 32 was incubated with HslV, HslU, and ATP under the standard assay conditions at 42°C, marked degradation of 32 occurred, whereas no detectable degradation occurred when any of these components was omitted (Fig. 1A). These results support our previous observations in vivo, suggesting that HslVU and other cytoplasmic proteases directly contribute to the turnover of 32 to appreciable extents.
To determine the stoichiometry of HslV (19 kDa) and HslU (50 kDa) for the maximal proteolytic activity, the reaction was carried out by increasing the amount of HslV with a fixed amount of HslU (40 g/ml, 0.8 M protomer) or by increasing the amount of HslU with a fixed amount of HslV (16 g/ml, 0.84 M protomer). In both cases, the amount of 32 hydrolyzed increased with increasing amounts of the other component and reached the maximal activity when HslV and HslU were present at an approximate molar ratio of 1:1 protomers (Fig. 1,B and C). These results are consistent with the notion that the HslVU protease consists of one HslV dodecamer and two HslU hexa-or heptamers (25,26).
Inhibition of 32 Degradation by RNA Polymerase (Core Enzyme)-To examine the effect of RNA polymerase on 32 degradation by the HslVU protease, 32 was preincubated with RNA polymerase at 44°C before adding HslVU. 32 was hardly degraded by HslVU under these conditions (Fig. 2, lane 5), whereas active degradation occurred in the control without RNA polymerase (lane 2), indicating that the prior interaction with RNA polymerase protected 32 from proteolytic attack by HslVU. By contrast, RNA polymerase did not affect degradation of the cell division inhibitor SulA, another substrate for HslVU (27,35), fused to the maltose-binding protein (MBP-SulA) (lanes 8 and 11). When RNA polymerase was incubated with MBP-SulA fusion protein at 44°C, about half of the RNA polymerase became insoluble and disappeared from the soluble fraction (lane 12). When preincubated with 32 , however, most of the RNA polymerase remained soluble (lane 6), indicating that most of the 32 binds to the polymerase to form RNA polymerase holoenzyme.
Marked Temperature Dependence of 32 Degradation-When proteolysis of 32 by HslVU was examined at four different temperatures (35,38,41, and 44°C), a striking temperature dependence was observed ( Fig. 3 and Table I). Only limited degradation occurred at lower temperatures (35 and 38°C), whereas rapid degradation was observed at higher temperatures (41 and 44°C). The ratio of activities at 44 and 35°C based on the initial rates of proteolysis was estimated to be ϳ15 (Table I). These results indicate that 32 can be degraded by HslVU at the physiological temperature range, but particularly efficiently at higher temperatures.
To address the question of whether 32 or HslVU is primarily responsible for the observed high temperature-dependent 32 degradation, two other substrates (MBP-SulA and ␣-casein) were tested and compared with 32 for the temperature dependence of proteolysis. When MBP-SulA fusion protein was incubated with HslVU and ATP at various temperatures, only a weak temperature dependence was observed, unlike with 32 : the fusion protein was degraded at 35°C with appreciable efficiency, ϳ40% of that at 44°C (Fig. 4A and Table I). Similar results were obtained when ␣-casein was digested with HslVU ( Fig. 4B and Table I). Taken together, these results suggest that 32 rather than HslVU is primarily responsible for the observed dependence of proteolytic activities on high temperature.
FtsH Catalyzes the Highly Temperature-dependent Proteolysis of 32 -To further confirm the unusual temperature dependence of 32 degradation, we carried out similar experiments with FtsH, known to degrade 32 in vivo and in vitro. Accordingly, 32 was incubated with FtsH at 35 or 44°C in the presence of ATP under the conditions used by Tomoyasu et al. (23). As shown in Fig. 5A, 32 degradation was much faster (ϳ4-fold) at 44°C than at 35°C. By contrast, ␣-casein used as a control substrate was degraded with similar efficiencies at both temperatures (Fig. 5B), indicating that 32 is also susceptible to FtsH-mediated proteolysis particularly at high temperature. These results strongly suggest that the unusual high temperature dependence of 32 degradation is ascribed to certain structural features of 32 , although the possibility that both HslVU and FtsH proteases exhibit an increased affinity for 32 at the elevated temperature was not excluded.
Digestion of 32 by Chymotrypsin-We then compared the susceptibility of 32 to chymotrypsin, known to exhibit a low  6 g) or MBP-SulA fusion protein (1.6 g) and ATP (4 mM) was kept at 44°C for 30 min; RNA polymerase (RNA pol; 19 g) or buffer was added; and after 6 min, HslU (1.6 g) and HslV (0.48 g) were added. A portion (10 l) was taken before adding HslV (lanes 1, 4, 7, and 10). The mixtures were incubated at 44°C for 30 min (lanes 2, 5, 8, and 11). Part of the samples were centrifuged at 18,500 ϫ g for 10 min, and the resulting supernatants (lanes 3, 6, 9, and 12) along with the others were analyzed by SDS-PAGE followed by Coomassie Brilliant Blue staining. Molecular mass markers (in kilodaltons; Nacalai Tesque) are indicated to the left. substrate specificity. When 32 was incubated with chymotrypsin at 35 or 44°C, it was promptly degraded at both temperatures, and several major degradation products appeared that were tentatively referred to as peptides 1-5 (Fig. 6). Peptides 1 and 5 rapidly appeared and gradually disappeared with similar time courses at both temperatures, indicating that chymotrypsin exerted quantitatively similar activities at both temperatures. On the other hand, peptide 2 gradually accumulated and remained for some time at 35°C, but accumulated much less at 44°C. In contrast, peptides 3 and 4 appeared more rapidly at 44°C than at 35°C. These results suggest that peptide 1 represents a primary intermediate in the degradation pathways at both temperatures and that further degradation (to yield peptides 2-4) follows disparate kinetics (or cleavage patterns) at 35 and 44°C. The conformation of 32 might differ significantly at the two temperatures.
Reversible Temperature-dependent Change in the Susceptibility of 32 to HslVU-We next examined the reversibility of the change in sensitivity of 32 to HslVU proteolysis by comparing the results from four distinct incubation protocols (Fig.  7). The first two protocols (I and II) served as controls, in which both the initial 2-h preincubation of 32 alone and the subsequent incubation with HslVU (1 h) were done at 44 and 35°C, respectively. In the third and fourth protocols, 32 was kept at 44°C for the first hour, shifted to 35°C for another hour, and finally incubated with the HslVU protease (1 h) at 44°C (protocol III) or at 35°C (protocol IV). The results showed that Ͼ70% of 32 was degraded at 44°C (protocol III), whereas only ϳ10% was degraded at 35°C (protocol IV), clearly indicating the reversibility of the temperature-dependent change in susceptibility to the protease. In other words, 32 probably assumes a state more sensitive or refractory to the attack by the HslVU protease depending on the temperature (44 or 35°C) to which it was last exposed.
Degradation of 32 in Vivo-In view of the above results obtained in vitro with two proteases that can degrade 32 , we addressed the question of whether the in vivo turnover of 32 reflects the observed high susceptibility of 32 to proteases at high temperature. Thus, the half-life of 32 was determined with exponentially growing cells at different temperatures by pulse-chase experiments, followed by immunoprecipitation with specific antiserum. When wild-type cells (MG1655) grown at 30, 37, and 42°C were pulse-labeled with [ 35 S]methionine and chased with excess unlabeled methionine, the half-life of 32 showed a clear dependence on temperature: 1 min at 30°C, 20 s at 37°C, and 10 -15 s at 42°C (Fig. 8A), indicating that 32 is extremely unstable at high temperature.
To further analyze the temperature dependence of 32 degradation in vivo, we also examined the stability of 32 by using the 32 overexpression system. IPTG was added to the exponentially growing cells at 30°C to induce 32 synthesis directed by the trc promoter on the expression plasmid (pKV1278). After 30 min, chloramphenicol was added to stop further protein synthesis, and the culture was divided into two, one kept at 30°C and the other shifted to 42°C. The half-life of 32 was determined by following the remaining levels of 32 by immu-   15 g; B) was incubated with FtsH (1 g) in the reaction mixture for FtsH (25 l) at 35°C (E) or 44°C (q) (upper panels). Samples (4 l) withdrawn at intervals were analyzed by SDS-PAGE followed by Coomassie Brilliant Blue staining as described in the legend to Fig. 3. Bands of 32 and ␣-casein in A and B, respectively, were quantified, and the amount of proteins degraded is presented as percent of the total substrate used (lower panels). noblotting at both temperatures (Fig. 8B). The half-life of 32 turned out to be ϳ20 min at 30°C under these conditions. When shifted to 42°C, 32 was stable for 3-4 min, followed by rapid degradation with a half-life of 1 min or less. The latter observation also suggests that the destabilization of 32 at high temperature does not require de novo protein synthesis. These results seem to be quite consistent with those obtained in vitro. The initial time lag consistent with the temporary stabilization previously observed by Straus et al. (5) is probably explained by assuming that nonnative proteins accumulated upon tempera-ture upshift transiently titrate chaperones and proteases away from 32 .

DISCUSSION
The present experiments demonstrated that purified HslVU protease can catalyze the ATP-dependent proteolysis of 32 , as predicted by our previous genetic data (24). As was the case with FtsH (23) and ClpAP (36) proteases, the HslVU protease did not require the presence of DnaK/DnaJ chaperones for activity, although these chaperones are needed for the rapid turnover of 32 in vivo (15,16). Since the Lon protease can also degrade 32 in vitro, 2 most of the known ATP-dependent proteases (possibly except for ClpXP) appear to be capable of hydrolyzing 32 as their common substrate. Such an overlapping specificity is not unique to 32 because several other proteins are now known to be degraded by more than one proteases: this includes Xis of phage (37), the SsrA-tagged repressor (38,39), and the SulA cell division inhibitor (27,35). Special interest is attached to the case of 32 , however, because the above proteases and their subunits (except ClpA) are heat shock proteins under 32 control (3). This means that these proteases can autogenously regulate their own intracellular levels by modulating the turnover of 32 as their substrate. Furthermore, since Lon, ClpAP, and HslVU proteases are known to recognize and degrade nonnative proteins in vivo (40), they are expected to play roles similar to those of the DnaK/DnaJ chaperones in modulating the heat shock response, i.e. the free proteases may also serve as a cellular thermometer. In addition, the membrane-associated FtsH protease was quite active in degrading ␣-casein (Fig. 5), a widely used substrate for cytosolic ATP-dependent proteases. FtsH may also turn out to be involved in degradation of various nonnative proteins in vivo and consequently in modulation of the heat shock response. The previous results suggested a nearly equal contribution of the membrane-bound protease FtsH and the sum of three cytosolic proteases (HslVU, ClpAP, and Lon) in the turnover of 32 in vivo (24), although it was difficult to assess the relative contribution of individual proteases accurately in normal E. coli cells.
The proteolysis of 32 by HslVU or FtsH showed a marked dependence on high temperature as compared with that of other substrates (Figs. [3][4][5]. The results of chymotrypsin digestion of 32 suggested a conformational change of 32 as a basis for the observed differences in susceptibility to the proteases between 35 and 44°C. Although the exact nature of the change(s) remains obscure, it seems likely that a conformational change occurs at a restricted region of 32 since 32 evidently retains much of the activities at 44°C in vivo (41) and in vitro (Fig. 2) (42). 32 may exhibit a unique intrinsic instability based on a conformational change at high temperature. Alternatively, the proteases might show increased affinities for 32 at high temperature, although these possibilities are not mutually exclusive.
In agreement with the results obtained in vitro, the half-life of 32 in vivo as determined by pulse-chase experiments with normally growing cells turned out to be much shorter at higher temperature than at lower temperature, ϳ1 min at 30°C and 10 -15 s at 42°C (Fig. 8A). Tilly et al. (15) reported previously, using an overexpression system of 32 , that the half-life of 32 varied with temperature: 0.7 min at 42°C, 1.3 min at 37°C, 1.7 min at 30°C, and 15 min at 22°C, quite consistent with our results. As shown in Fig. 8B, the overproduced 32 that was markedly stabilized at 30°C was rapidly degraded upon a shift to 42°C in the absence of protein synthesis, indicating that destabilization of 32 7. Reversibility of the change in sensitivity of 32 to HslVU. 32 was incubated for 3 h in a reaction mixture (50 l) according to four different protocols. I and II, 32 was incubated at constant temperature (control); III and IV, temperatures were changed at 1-h intervals as indicated (left to right). In all protocols, HslV and HslU were added at the end of a 2-h incubation. Samples (10 l) were taken before (lanes 1, 3, 5, and 7) and after (lanes 2, 4, 6, and 8) addition of the protease and were subjected to SDS-PAGE followed by Coomassie Brilliant Blue staining (upper panel). Bands of 32 were quantified, and the amount of 32 remaining is presented as percent of the total 32 used (lower panel).
FIG. 8. Turnover of 32 in exponentially growing cells at different temperatures. A, MG1655 cells were grown in synthetic medium at 30, 37, and 42°C to mid-log phase, pulse-labeled with [ 35 S]methionine (100 Ci/ml) for 30 s, and chased with excess unlabeled methionine (200 g/ml). Samples were taken as indicated, starting with the first sample taken at 30 s (set as time 0), and labeled 32 was analyzed by immunoprecipitation followed by SDS-PAGE. Analysis of the labeled bands was done with a Fuji BAS2000 bioimaging analyzer. B, the stability of 32 overproduced from the expression plasmid was examined. MG1655 cells harboring pKV1278 were grown in synthetic medium at 30°C to mid-log phase; IPTG (1 mM) was added; and after 30 min, chloramphenicol (100 g/ml) was added. The culture was divided into two parts; one was kept at 30°C, whereas the other was shifted to 42°C (time 0). Samples were taken at 2-min intervals and analyzed for the remaining 32 levels by immunoblotting.
novo synthesis of degradation machinery including proteases and chaperones. As to the requirement of the DnaK chaperone team for the turnover of 32 in vivo, it should be noted that the dependence of 32 degradation on these chaperones (particularly DnaJ and GrpE) was much less striking at high temperature (42°C) than at low temperature (30°C) (16). These results could be related to the high susceptibility of 32 to proteases at high temperature. The DnaK/DnaJ chaperones might make 32 more susceptible to proteases. However, the presence of any of the DnaK/DnaJ/GrpE chaperones failed to affect the FtsH-mediated 32 proteolysis under the conditions used by Liberek and co-workers (43). In any event, the present results combined with other evidence suggest that the intracellular stability of 32 is modulated by its own susceptibility to proteases reflecting ambient temperature, as well as through interaction with the chaperones.
Finally, the present data raise the intriguing question of the potential physiological significance of the unusually high instability of 32 at high temperature. Even when the difference in cellular growth rate at 30 and 42°C (ϳ2-fold) was taken into consideration, the in vivo turnover of 32 seemed to be appreciably faster at 42°C. On the other hand, the synthesis rate of 32 at 42°C appears to be severalfold higher than that at 30°C (5). 3 The finding that cells synthesizing 32 at a higher rate simultaneously exhibit increased degradation of 32 at high temperature might seem paradoxical. However, the increased synthesis primarily determined by the stability of rpoH mRNA secondary structure (14) would be absolutely needed to rapidly respond to temperature upshift, whereas the higher rate of degradation may be required to counteract the higher rate of synthesis to avoid excessive accumulation of 32 . Although higher levels of 32 are required for cell growth at higher temperature (44), excessive 32 is deleterious and inhibits cell growth (15,45). We suggest that the basal level of 32 is determined by its synthesis rate and intrinsic stability, both directly reflecting the ambient temperature. In addition, the temporary adjustment or fine-tuning of the 32 level such as the transient stabilization of 32 observed immediately following heat shock is accomplished most effectively by modulating its stability in accordance with the state of protein folding in the cell.