Synergistic Binding of DnaJ and DnaK Chaperones to Heat Shock Transcription Factor σ32 Ensures Its Characteristic High Metabolic Instability

Background: Interaction between chaperone and co-chaperone is crucial for chaperone binding to substrates. Results: The reduced affinity of mutant σ32 for DnaJ reduces complex formation with DnaK. Conclusion: The stable binding of σ32 to DnaK requires the exposure of σ32 sites to bind DnaJ. Significance: This model can be applied to chaperone action in helping newly synthesized polypeptides to fold. Escherichia coli heat shock transcription factor σ32 is rapidly degraded by ATP-dependent proteases, such as FtsH and ClpYQ. Although the DnaK chaperone system (DnaK, DnaJ, and GrpE) promotes σ32 degradation in vivo, the precise mechanism that is involved remains unknown. Our previous results indicated that σ32 mutants containing amino acid substitution in the N-terminal half of Region 2.1 are markedly stabilized in vivo. Here, we report the further characterization of these mutants by examining purified σ32 mutants in vitro. Surprisingly, I54A σ32, a very stable mutant, is more susceptible to ClpYQ and FtsH proteases than wild-type σ32, indicating that the stability of σ32 does not always reflect its susceptibility to proteases. Co-precipitation and gel filtration analyses show that purified σ32 mutants exhibit a reduced affinity for DnaJ, leading to a marked decrease in forming a complex with DnaK in the presence of DnaJ and ATP. Other mutants with modestly increased stability (A50S σ32 and K51E σ32) show an intermediate efficiency of complex formation with DnaK, suggesting that defects in binding to DnaK and DnaJ are well correlated with metabolic stability; effective interaction with DnaK promotes σ32 degradation in vivo. We argue that the stable and effective interaction of heat shock protein 70 (Hsp70) with a substrate polypeptide may generally require the simultaneous binding of heat shock protein 40 (Hsp40) to distinct sites on the substrate.

Various cellular processes are controlled by the regulated degradation of key protein factors. Escherichia coli heat shock transcription factor 32 (encoded by the rpoH gene), which is required for the heat shock response, is rapidly degraded with a half-life of 1-2 min during steady-state growth at 30°C (1,2). When E. coli cells are shifted from 30 to 42°C, 32 is transiently stabilized, and the translation of rpoH mRNA increases, which lead to much higher levels of 32 and heat shock proteins (HSPs), 2 including molecular chaperones and ATP-dependent proteases. After alleviating stress-induced damage caused by misfolded or unfolded proteins, 32 becomes unstable and is rapidly degraded. However, because excess amounts of HSPs are toxic to the cell, the tight regulation of 32 levels by the degradation machinery is crucial for sustaining growth under any circumstances.
Among the five ATP-dependent proteases (Lon, ClpAP, ClpXP, FtsH (HflB), and ClpYQ (HslUV)) known in E. coli (3), FtsH is the major protease involved in 32 degradation (4,5). FtsH is a member of the AAA proteases (6) and is a membranebound metalloprotease with an active site that is exposed to the cytoplasm. It is known to degrade some cytoplasmic proteins as well as membrane proteins (7). Although FtsH is essential for growth, a ⌬ftsH strain can be isolated if the cell simultaneously contains an unusually high activity of R-3-hydroxyacyl-ACP dehydrase (encoded by the fabZ gene) (8). The major function of FtsH is the maintenance of the proper lipopolysaccharide/ phospholipids ratio by the degradation of LpxC. Other proteases are cytosolic and generally target abnormal proteins as well as some other specific substrates. 32 is stabilized in a mutant that is multiply deficient in Lon, ClpXP, and ClpYQ (9). Purified ClpYQ directly degrades 32 ; 32 degradation by both FtsH and ClpYQ in vitro is promoted by higher temperatures (10).
In addition to proteases, the DnaK chaperone system consisting of DnaK (Hsp70), DnaJ (Hsp40), and GrpE (nucleotide exchange factor) is required for the rapid degradation of 32 in vivo because the half-life of 32 is much longer in dnaK, dnaJ, and grpE mutants (11,12). 32 directly interacts with DnaK or DnaJ and forms a stable ternary complex in the presence of ATP (13,14,15,16,17). Although this interaction with DnaK and DnaJ has been thought to induce a conformational change in 32 and promote degradation by proteases, such DnaK chaper-□ S This article contains supplemental Experimental Procedures and Figs. S1-S12. 1 To whom correspondence may be addressed. Tel.: 81-76-264-6229; Fax: 81-76-264-6230; E-mail: mkanemo@staff.kanazawa-u.ac.jp.
one system effects on 32 degradation have never been observed in vitro (18,19). Hsp70 chaperone systems in most organisms are known to be involved in the folding of newly synthesized polypeptides, the refolding of denatured proteins, the dissociation of proteins from complexes, and the degradation of abnormal proteins (20 -23). A canonical Hsp70 chaperone system consists of Hsp70, Hsp40, and a nucleotide exchange factor (Hsp70 has an ATPase activity). Although the ATP-bound form of Hsp70 has a low affinity for substrates due to its high substrate exchange rate, the ADP-bound form has a higher affinity for substrates with a low substrate exchange rate. Hsp40 acts as a co-chaperone and activates the Hsp70 ATPase, although it also acts as a chaperone and binds to unfolded polypeptides to prevent aggregation. Through an interaction with Hsp40, Hsp70 becomes an ADP-bound form and tightly binds substrates. A nucleotide exchange factor promotes the dissociation of ADP bound to Hsp70, by which Hsp70 returns to its ATP-bound form and releases its substrates. After dissociation from Hsp70, the polypeptide chains fold into a functional three-dimensional architecture.
A model has been proposed in which Hsp40 first recognizes and binds substrate polypeptides and then transfers them to Hsp70. However, in some cases, DnaK itself recognizes and binds to substrate polypeptides independent of DnaJ, as in the case of 32 (24). Because most of the substrates of the Hsp70 chaperone system are unfolded polypeptides that are structurally heterogeneous, a native form of 32 that can directly interact with DnaK and DnaJ should provide a useful model to elucidate the function and mechanism of the Hsp70 chaperone system. Studies with another native substrate, bacteriophage P1 RepA protein, have located DnaK-and DnaJ-binding sites in RepA (25).
To gain further insight into chaperone functions in 32 degradation, we isolated many 32 mutants that are stable in vivo (26). These mutants contain one or two amino acid substitutions in the N-terminal half of Region 2.1. In this study, we examined the affinity of some mutant 32 for proteases and chaperones and obtained results indicating that mutations generally decrease the capacity of 32 to form complexes with DnaK and DnaJ.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-KY1603 (MC4100 (F Ϫ araD ⌬(argF-lac) U169 rpsL relA flbB deoC ptsF rbsR) ⌬rpoH30::kan zhf50::Tn10 suhX401 (pF13-PrpoD hs -lacZ)) (27), which is a MC4100 derivative that lacks the rpoH gene but is able to grow at 37°C due to the overproduction of GroEL and GroES, was used for the purification of 32 . AD202 (MC4100 ompT::kan) (28) was used for the purification of FtsH. HB101 (supE ⌬(mcrC-mrr) recA ara-14 proA lacY galK rpsL xyl-5 mtl-1 leuB thi-1) and MC4100 were used for the purification of DnaK and DnaJ, respectively. KY1459 (MC4100 ⌬dnaK52::cat) (29) and KY1456 (MC4100 dnaJ::Tn10-42) (30) were used for the purification of DnaK-His and DnaJ-His, respectively. The plasmid pKV1142 carries the isopropyl-␤-D-thiogalactopyranoside (IPTG)-dependent trc promoter (trcp) (26). The ftsH, dnaK, and dnaJ genes that lack authentic promoters were amplified by PCR and cloned under the trc promoter. In the case of DnaK-His and DnaJ-His, six histidine codons were added to the 3Ј-end of the corresponding genes on the pKV1142 derivatives. Both DnaK-His and DnaJ-His have two extra amino acids, arginine and serine, between the last amino acid residue of the intact protein and the His tag, which is derived from the BglII site that is used to construct the plasmids. The plasmid pET3a (Novagen) carries the T7 promoter that is transcribed by T7 RNA polymerase and was used to construct various rpoH expression plasmids. Promoterless wild-type and mutant rpoH genes were cloned under the T7 promoter from pKV1142 and pKV1585 derivatives (26). The nucleotide sequences of all of the PCRamplified and cloned fragments were confirmed by DNA sequence analysis.
DnaK was purified from HB101 cells harboring pKV1957 (pKV1142 trcp-dnaK). The cells were grown until the late log phase in L broth containing 50 g/ml ampicillin at 30°C, and DnaK synthesis was induced by 1 mM IPTG. Cells were harvested after 3 h, treated with lysozyme and sodium deoxycholate, and disrupted, and the resulting lysate was centrifuged as in the purification of 32 . The supernatant was treated with ammonium sulfate. Proteins that were precipitated in a range of ammonium sulfate concentrations between 0.24 and 0.31 g/ml were dissolved, dialyzed against Buffer A, and loaded onto a HiTrap heparin column (GE Healthcare). The flow-through fraction was repeatedly loaded onto an ATP-agarose (Sigma-Aldrich) column that was equilibrated with Buffer A containing 100 mM NaCl and 5 mM MgCl 2 . Proteins were eluted with Buffer A containing 100 mM NaCl and 5 mM ATP. The eluate was loaded onto a HiTrap Q-Sepharose column (GE Healthcare) equilibrated with Buffer A, and proteins were eluted with a linear gradient of NaCl. The fractions containing DnaK were dialyzed against Buffer B (25 mM HEPES-NaOH (pH 8.0), 1 mM EDTA, 1 mM dithiothreitol, 10% (v/v) glycerol, 100 mM NaCl) and applied to a HiPrep Sephacryl S-300 column (GE Healthcare). DnaK was concentrated and stored at Ϫ70°C.
DnaJ was purified from MC4100 cells harboring pKV1961 (pKV1142 trcp-dnaJ). The cells were grown to late log phase in L broth containing 50 g/ml ampicillin at 30°C, and DnaJ synthesis was induced by 1 mM IPTG. Cells were harvested after 1 h, treated with lysozyme and sodium deoxycholate, and disrupted, and the resulting lysate was centrifuged as in the puri-fication of 32 . The supernatant was treated with ammonium sulfate. Proteins that were precipitated in a range of ammonium sulfate concentrations between 0.21 and 0.28 g/ml were dissolved, dialyzed against Buffer A, and loaded onto a HiTrap Q-Sepharose column. Proteins were eluted with a linear gradient of NaCl. The fractions containing DnaJ were dialyzed against Buffer A and applied to a HiTrap heparin column. Proteins were eluted with a linear gradient of KCl. DnaJ was concentrated and stored at Ϫ70°C.
DnaK-His was purified from KY1459 cells harboring pKV2231 (pKV1142 trcp-dnaK-his; the dnaK gene with six histidine codons at the 3Ј-end). The cells were grown to late log phase in L broth containing 50 g/ml ampicillin at 30°C, and DnaK-His was induced by 1 mM IPTG. Cells were harvested after 3 h, resuspended in Buffer C (50 mM Tris-HCl (pH 7.5), 10% (v/v) glycerol) containing 0.1% lysozyme, and kept on ice for 30 min. The cells were disrupted by sonication, and the resulting lysate was centrifuged at 75,000 ϫ g for 90 min. The supernatant was loaded onto a Ni 2ϩ -NTA-agarose (Qiagen) column equilibrated with Buffer C. Proteins were eluted with Buffer C containing 250 mM imidazole. The eluate was loaded onto a HiTrap Q-Sepharose column and a HiPrep Sephacryl S-300 column as in the purification of DnaK. DnaK-His was concentrated and stored at Ϫ70°C. DnaJ-His was purified from KY1456 cells harboring pKV2237 (pKV1142 trcp-dnaJ-his; the dnaJ gene with six histidine codons at the 3Ј-end) essentially as described (33).
All of the purified proteins, except for DnaJ, were greater than 90% pure as estimated by SDS-PAGE followed by staining with Coomassie Brilliant Blue (CBB) and were quantified by Bradford protein assays (Bio-Rad). The purity of DnaJ was 80 -90%. Because the functional form of DnaJ is the dimer, the concentrations of DnaJ indicated are in the dimer form.
Co-immunoprecipitation and Pull-down Assay-The interaction of 32 with DnaK and/or DnaJ was examined using a co-immunoprecipitation assay according to the method described previously (25), except that anti-32 serum was used. The 32 -antibody complex was precipitated by adding protein A-Sepharose. After subjecting the precipitate to SDS-PAGE, 32 and DnaK were detected by CBB staining, and DnaJ was detected by immunoblotting with anti-DnaJ serum.
Pull-down assays for the interaction between 32 and DnaK-His or DnaJ-His were performed using Ni 2ϩ -NTA-agarose. DnaK-His was preincubated in Buffer D (20 mM Tris-HCl (pH 7.5), 100 mM KCl, 10% glycerol, 5 mM Mg(CH 3 COO) 2 , 1 mg/ml BSA) at 30°C for 2 h to convert oligomer forms into monomer form (17). After adding 32 without or with a nucleotide (ATP or ADP) (total volume, 50 l), the mixture was incubated at 30°C for 30 min. Next, 400 l of Buffer E (50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.1% Tween 20) and 10 l of a 50% slurry of Ni 2ϩ -NTA-agarose were then added, and the mixture was rotated at 4°C for 1 h. Agarose beads were washed with Buffer F (50 mM Tris-HCl (pH 7.5), 300 mM NaCl, 0.2% Tween 20) five times, and DnaK-His and 32 were dissociated from the agarose beads by treating the precipitate with 250 mM imidazole. Proteins were subjected to SDS-PAGE and detected by CBB staining. Experiments using DnaJ-His were performed in a similar fashion, except that the preincubation and addition of nucleo-tides were omitted. In all of the assays, gel images were taken with an LAS3000 image analyzer (GE Healthcare), and the quantitation of the protein bands was performed with Multi Gauge software (GE Healthcare).
Gel Filtration-A gel filtration analysis was performed as described previously (17). DnaK or DnaK-His was preincubated in Buffer G (20 mM Tris-HCl (pH 7.5), 200 mM KCl, 1 mM DTT, 0.1 mM EDTA, 10% glycerol, 5 mM Mg(CH 3 COO) 2 ) at 30°C for 2 h. DnaJ-His, 32 , and ATP (1 mM) were added, and the mixture was incubated at 30°C for 30 min. After 5 min on ice, molecular mass standards (␤-amylase, 200 kDa; apo-transferrin, 81 kDa; carbonic anhydrase, 29 kDa) were added, and the mixture was loaded on a Superdex 200 gel filtration column (GE Healthcare) equilibrated with Buffer G containing 200 M ATP at 4°C. Aliquots from each eluted fraction (0.5 ml) were subjected to SDS-PAGE. The three standard proteins were detected by CBB staining, and DnaK (or DnaK-His), DnaJ-His, and 32 were detected by immunoblotting. Gel images were taken with LAS3000 image analyzer, and the quantitation of protein bands was performed using Multi Gauge software. The fraction numbers of the eluted proteins were as follows: ␤-amylase (200 kDa (perhaps due to self-aggregation), the fraction numbers corresponding to a free DnaJ dimer (DnaJ 2 ) were not precisely determined.

RESULTS
Purified Stable 32 Mutants Are Not Necessarily More Resistant to Proteases-Among the 32 mutants that are stabilized in vivo (26), the two most stable variants, I54A 32 and L47Q/ L55Q 32 , whose half-lives are 9-fold and more than 10-fold longer than that of the wild type, respectively, were examined in an in vitro degradation system using ClpYQ and FtsH proteases. As expected from high 32 activity in vivo (26), the purified 32 mutants were co-fractionated with core RNA polymerase like wild-type 32 using gel filtration (supplemental Fig.  S1, A-C), indicating that large conformational changes did not occur due to amino acid substitution or during purification. When L47Q/L55Q 32 was incubated with ClpYQ at 37 or 42°C, it was degraded much more slowly than wild-type 32 in accordance with its stability in vivo (Fig. 1, supplemental Fig.  S2). In contrast, I54A 32 was degraded slightly faster than wild type at 37°C and even faster at 42°C ( Fig. 1 and supplemental  Fig. S2).
Similar results were obtained when the same 32 mutants were examined with FtsH protease: L47Q/L55Q 32 degraded more slowly than wild type, whereas I54A 32 degraded much faster than wild-type at all temperatures (30, 37, and 42°C) ( Fig.  1 and supplemental Fig. S2). These results indicate that stability of 32 in vivo does not always reflect its direct susceptibility to proteases and that the high stability of I54A 32 in vivo is attributed to some intracellular state of 32 .
Stable 32 Mutants Show Reduced Affinity for DnaJ-The above results showed that other factors in addition to proteases determine 32 stability in vivo. Because the DnaK chaperone system has been thought to be required for both inactivation and rapid degradation of 32 in vivo (11,12,35,36,37), and wild-type 32 is known to form complexes with DnaK and/or DnaJ (13,14,15,16,17), stable 32 mutants were strongly expected to have decreased affinity for DnaK and/or DnaJ chaperones. To test this hypothesis, co-immunoprecipitation experiments were performed using anti-32 serum. Because the DnaK-32 complex was not detected in the presence of ATP, and DnaK was precipitated even without 32 in the presence of ADP, these experiments were performed without adding nucleotide. When I54A 32 or L47Q/L55Q 32 was mixed with DnaK and immunoprecipitated with anti-32 serum, the amounts of DnaK co-precipitated were similar to that obtained with wildtype 32 ( Fig. 2A and supplemental Fig. S3A). In contrast, very little DnaJ was co-precipitated with mutant 32 compared with wild type (Fig. 2B and supplemental Fig. S3B), suggesting that the 32 mutants are specifically deficient in binding to DnaJ. Although an appreciable amount of DnaJ was precipitated with anti-32 serum even without the 32 addition, as shown in Fig.  2B (control (Ϫ)), the effects of mutation on 32 -DnaJ interaction were clearly observed.
To otherwise examine the effects of mutation on 32 -DnaK or 32 -DnaJ interaction, we constructed His-tagged DnaK and DnaJ at the C terminus, and their activities were confirmed both by complementation of the dnaK/dnaJ mutant phenotype in vivo and by refolding activity of the unfolded luciferase in vitro (data not shown). In pull-down assays using Ni 2ϩ -NTA-agarose, more than 90% of the DnaK-His or DnaJ-His was recovered in the presence of ATP or ADP (data not shown). When wild-type 32 was mixed with DnaK-His, a significantly higher amount of 32 was co-eluted with DnaK-His in the presence of ADP than in the presence of ATP or in a control without nucleotides ( Fig. 3A  and supplemental Fig. S4A). Two 32 mutants showed no detectable difference in the amount of 32 co-eluted with DnaK-His compared with wild-type 32 (Fig. 3A and supplemental Fig. S4A). In contrast, similar pull-down assays using DnaJ-His revealed much less interaction with mutant 32 than with wild-type 32 (Fig. 3B and supplemental Fig. S4B). Taken together with the above results of the co-immunoprecipitation experiments (Fig. 2, A and B), these results clearly indicate that amino acid substitutions in Region 2.1 resulted in a reduced affinity for DnaJ specifically.  , and immunoprecipitates were analyzed as in A, except that DnaJ was detected by immunoblotting with anti-DnaJ serum (supplemental Fig. S3B). DnaJ and the rabbit immunoglobulin heavy chain were quantified, and the ratios of DnaJ to immunoglobulin heavy chain were calculated. No 32 was added to the mixture for control (Ϫ). C and D, 32

Stable 32 Mutants Cannot Efficiently Form Complexes with
DnaK in the Presence of DnaJ and ATP-Although the above results show that each chaperone can bind to 32 independently, DnaJ is known to activate DnaK to form a stable complex with 32 in the presence of ATP (16,17). Therefore, we examined the 32 -DnaK interaction in the presence of DnaJ and ATP. The co-immunoprecipitation experiments first showed that the amount of DnaJ co-precipitated with mutant 32 was much less than that with wild-type 32 (Fig. 2C, supplemental  Fig. S3C). This is consistent with the result obtained when 32 interacted with DnaJ alone (Fig. 2B). However, unlike the results seen when 32 interacted with DnaK alone (Fig. 2A), the amounts of DnaK co-precipitated with mutant 32 clearly decreased compared with the wild type in the presence of DnaJ ( Fig. 2D and supplemental Fig. S3D). To examine the effects of mutation on 32 -DnaK interaction in the presence of DnaJ and ADP, we constructed His-tagged 32 at the C terminus. In pulldown assays using Ni 2ϩ -NTA-agarose, 60 -80% of the 32 -His was recovered. When wild-type 32 -His was mixed with DnaK and DnaJ, a significantly higher amount of DnaK was co-eluted with 32 -His in the presence of ATP than in the presence of ADP (supplemental Fig. S5). In the presence of ADP, there was no significant difference in the DnaK-32 complex formation between wild-type 32 -His and I54A 32 -His, regardless of whether DnaJ was included or not (supplemental Fig. S5), suggesting that DnaJ has no effect on the 32 -DnaK interaction in the presence of ADP. In the presence of DnaJ and ATP, a higher amount of DnaK was co-eluted with wild-type 32 -His than with I54A 32 -His (supplemental Fig. S5). These results suggest that the low affinity of mutant 32 to DnaJ leads to an unstable interaction between the altered 32 protein and DnaK. It is thus conceivable that the binding of DnaJ to wild-type 32 induces a stable interaction with DnaK.
To further evaluate the extent of 32 -DnaK (or 32 -DnaJ) interaction, mixtures of 32 , DnaK, and DnaJ were analyzed by gel filtration. DnaJ-His was used instead of DnaJ because of its higher purity. DnaK concentration was varied from 0.8 to 3.2 M, whereas 32 and DnaJ-His were kept constant at 0.4 M. As shown in Fig. 4 and supplemental Fig. S6, 32 (Fig. 4, A-D, circles). Although DnaJ was eluted over a wide range, distinctive peaks of eluted DnaJ were detected and corresponded with those of 32 in the case of wild-type 32 , whereas in the case of mutant 32 , no significant DnaJ peak was observed, and most DnaJ molecules were eluted in Fractions 17-19 (aggregated DnaJ) (supplemental Fig. S8). DnaK-His was precipitated with Ni 2ϩ -NTA-agarose beads, and proteins that were eluted by imidazole were subjected to SDS-PAGE. After staining with CBB (supplemental Fig. S4A), the ratios of 32 to DnaK-His were calculated. The amount of wild-type 32 Fig. S10). Because it was difficult to analyze an effect of even higher amounts of DnaJ on complex formation by gel filtration due to its self-aggregation, pull-down assays were performed with Histagged 32 . Even when a higher amount of DnaJ was used (in this case, authentic DnaJ was used), the level of DnaK co-eluted with I54A 32 -His did not change (supplemental Fig. S11). All of these results together with those of co-immunoprecipitation and pulldown experiments (Figs. 2 and 3) suggest that much higher concentrations of DnaJ are required to overcome the low affinity of mutant 32 for DnaJ in forming DnaK-32 -(DnaJ 2 ) ternary complex under these conditions.
Moderately Stabilized 32 Mutants in Vivo Show Intermediate Affinity for Chaperones-To clarify the relationship between the in vivo stability of 32 and the affinity for chaperones, we examined two other 32 mutants, A50S and K51E, that exhibit moderately increased in vivo stability (half-life approximately 4 times longer than that of wild type) (26). The purified 32 mutants were co-fractionated with core RNA polymerase like wild-type 32 using gel filtration (supplemental Fig. S1, D and E). When these 32 mutants were incubated with FtsH protease at 42°C, K51E 32 was slightly more stable, but A50S 32 was more susceptible to FtsH than the wild-type, much like I54A 32 (Fig. 5A and supplemental Fig. S12A). This again indicates that in vivo stability does not directly reflect susceptibility to FtsH protease. We next analyzed the interaction of these 32 mutants with DnaK-His and DnaJ-His by gel filtration. Although DnaK-His was used in this experiment, the results were similar to those with authentic DnaK; in other words, the percentage of wild-type 32 and I54A 32 in its free form (Fractions 31-35) at 1.6 M DnaK-His was 10.8 and 62.3%, respectively ( Fig. 5B and supplemental Fig. S12B), and 9.2 and 38.6% at 2.4 M DnaK-His ( Fig. 5C and supplemental Fig. S12C), indicating that the His tag does not appreciably affect the efficiency of complex formation (Table 1). In the case of A50S 32 and K51E 32 , the percentage of free 32 was 40.4 and 59.4% at 1.6 M DnaK-His and 26.5 and 18.5% at 2.4 M DnaK-His, respectively (Fig. 5, B and C, and supplemental Fig. S12, B and C). When the percentage of free 32 was plotted against relative half-lives in vivo, a clear correlation was evident between these    Fig. 4, B and C. b Data from the experiment shown in Fig. 5, B and C.
two parameters (Fig. 6), strongly suggesting that the effective interaction of DnaK and DnaJ with 32 is critical for sustaining the characteristic instability of 32 in vivo.

DISCUSSION
Several specific amino acid residues in the N-terminal half of Region 2.1 of 32 are intimately involved in the rapid degradation of 32 (26). Here, we demonstrate that amino acid substitutions in this region lead to reduced affinity of 32 for DnaJ (Figs. 2B and 3B), which in turn leads to unstable interaction of 32 with DnaK in the presence of DnaJ and ATP (Figs. 2D and 4). The close correlation found between the efficiency of complex formation of 32 with DnaK/DnaJ and in vivo stability (Fig.  6) strongly suggests that the defective interaction of 32 with DnaK and DnaJ causes stabilization of 32 in vivo. Thus, binding of DnaK and DnaJ to 32 most probably exerts conformational change on 32 , which then promotes degradation by the proteases.
Obrist et al. selected 32 mutants (L47Q, A50V, I54F, and I54T) based on their resistance to proteases (38). They argue that these mutations do not affect their affinity for DnaK or DnaJ and that Region 2.1 is involved in the interaction with FtsH protease (39,40). The present finding that I54A 32 and A50S 32 are more sensitive to ClpYQ and FtsH, whereas L47Q/L55Q 32 and K51E 32 are less sensitive (Figs. 1 and 5A), also suggests that Region 2.1 could be recognized directly by these proteases. Yura et al. (41) isolated 32 mutants (A50D, K51E, I54N, and I54T) based on resistance to feedback inhibition of 32 activity by DnaK and DnaJ or simply on increased 32 activity. They showed that these mutants are defective in chaperone-mediated feedback control, in addition to having increased 32 stability. However, analysis of the most defective mutant (I54N) exhibits chaperone-mediated inactivation in vitro and only a 2-fold higher dissociation constant (K D ) value for DnaJ using surface plasmon resonance. This is insufficient to account for the strong in vivo phenotype. Although the reasons for these differences remain unclear, our gel filtration results examining the interaction of 32 with DnaK in the presence of DnaJ and ATP led to the finding that the 32 mutants tested are clearly defective in chaperone-32 complex formation (Figs. 4 and 5). Analyses of I54N 32 gave results very similar to those obtained for I54A 32 (data not shown). Although the reduced affinity of the 32 mutants for DnaK and DnaJ can account for their stability in vivo, detailed mechanisms of chaperone-promoted 32 degradation remain unsolved. In this connection, we note that recent results suggest that the bacterial signal recognition particle plays a critical role in the chaperonemediated feedback inhibition/degradation of 32 . 3 As for the role of the N-terminal half of Region 2.1 of 32 in DnaJ binding, neighboring residues 57-66 of 32 were recently reported to be a DnaJ-binding site (42). A 32 mutant with three amino acid substitutions at positions 60, 62, and 63 showed reduced affinity for DnaJ. However, the fact that various amino acid substitutions at positions 47, 50, 51, 54, and 55 lead to 32 stabilization primarily due to defects in DnaJ binding suggests 3 T. Yura, personal communication. The thickness of an arrow in each reaction shows a relative reaction rate. When the arrow is thicker, the rate is higher. A, 32 degradation model. In 32 , the DnaJ-binding site(s) remains on the surface even after the completion of all folding steps. In this case, the simultaneous binding of DnaK and DnaJ on the same 32 molecule promotes the ATPase activity of DnaK (Reaction 1), and ADP-bound DnaK can tightly bind to 32 regardless of whether DnaJ associates or dissociates (Reaction 2). After GrpE (E. coli nucleotide exchange factor (NEF)) induces the exchange of nucleotides (Reaction 3), DnaK molecules dissociate from the substrate. In this process, a conformational change of 32 may be induced (not drawn in this figure), and released 32 is degraded by proteases (Reaction 4). B, Hsp70 chaperone action to an unfolded polypeptide. Some ATP-bound Hsp70 molecules rapidly associate with and dissociate from an unfolded polypeptide (such as newly synthesized polypeptides) independent of Hsp40. When they interact with Hsp40 on the same polypeptide, ATP bound to Hsp70 is hydrolyzed (Reaction 5). The resulting ADP-bound Hsp70 can stably bind to the substrate polypeptide due to its slow dissociation rate even after Hsp40 dissociates (Reaction 6). A nucleotide exchange factor promotes the exchange of ADP for ATP on Hsp70 molecules (Reaction 7). After Hsp70 dissociation, the polypeptide can fold into its functional shape. In this process, the Hsp40-binding site is sequestered inside the polypeptide (Reaction 8). The polypeptide losing affinity for Hsp40 can no longer form a stable complex with Hsp70, although it retains high affinity for Hsp70 (Reaction 9). the importance of this region in DnaJ binding. We propose that an expanded region of 32 , residues 47-66, affects DnaJ-binding directly.
Judging from the prevailing thought that Hsp40 first recognizes substrate polypeptides and then transfers them to Hsp70, the low affinity of mutant 32 for DnaK could be explained by assuming that the 32 mutant with low affinity for DnaJ is not recognized directly by DnaK. However, DnaK alone can recognize and bind 32 without the help of DnaJ ( Figs. 2A and 3A) (13,14,16,17), and DnaK appears to bind to 32 at a site different from that which binds to DnaJ (42). A truncated DnaJ mutant lacking the substrate-binding domain can induce the stable interaction of DnaK with 32 by stimulating ATPase activity of DnaK, although excess amounts are required compared with wild-type DnaJ (16). Furthermore, 32 and DnaJ synergistically stimulate the ATPase activity of DnaK (43). So far, chaperone mutants have been used in most previous experiments to examine chaperone-substrate interaction, whereas substrate mutants were used in the present study. These data complement each other and suggest that the major function of DnaJ is stimulation of DnaK ATPase activity in the DnaK chaperone cycle and that the most efficient stimulation occurs when both DnaJ and DnaK simultaneously bind to the same 32 molecule. We propose the following model for 32 -chaperone interaction (Fig. 7A). DnaJ interacts with 32 at DnaJ-binding sites, and ATP-bound DnaK independently interacts with 32 at DnaK-binding sites. When DnaK-binding sites are located near the DnaJ-binding sites, ATP bound to DnaK is promptly hydrolyzed through a transient DnaK-DnaJ interaction. The resulting ADP-bound DnaK could stably and effectively bind 32 . Chaperone binding would induce conformational change on 32 , which then promotes degradation by the proteases. Without the proper interaction of DnaJ with DnaK on the same 32 molecule, the hydrolysis of ATP bound to DnaK would be slow, and the affinity of DnaK for 32 would remain low. As a result, weak interaction between mutant 32 and DnaJ could lead to high amounts of free 32 .
It has been demonstrated that DnaK and DnaJ can independently bind to other substrates (oligopeptides and denatured proteins) and that DnaJ efficiently stimulates the ATPase activity of DnaK when DnaJ and DnaK bind to different segments of the same polypeptide chain (24). In a mammalian Hsp70 chaperone system, it was also recently shown that Hsp70 and Hsp40 bind independently to an unfolded protein (44). Thus, we can expand the above model to explain the general function of the Hsp70 chaperone system (Fig. 7B). The present results suggest that a polypeptide with low affinity for Hsp40 cannot be an active substrate for Hsp70. This scheme can be applied to the folding processes of unfolded polypeptides, such as newly synthesized polypeptide chains. In the process of repeated association with and dissociation from an unfolded polypeptide, Hsp70 molecules interacting with Hsp40 on the same polypeptide could stably bind to the substrate polypeptide. Only this stable binding by ADP-bound Hsp70 could effectively modulate the folding process of substrate polypeptides. Given that DnaJ catalytically activates DnaK functions (16,24,43), Hsp40 appears to promptly dissociate from Hsp70-substrate complexes. After Hsp70 dissociation, if Hsp40-binding sites on the substrate are sequestered inside the polypeptide chain, the polypeptide can no longer form a stable complex with Hsp70, although it retains a high affinity for Hsp70. Hsp40-binding sites appear to play an important role in stable Hsp70-substrate interaction. To effectively and stringently regulate Hsp70-substrate interaction, the number of Hsp40-binding sites per substrate polypeptide may be inherently fewer than that of the Hsp70-binding sites (44), or Hsp40-binding sites may be folded into the inside of the substrate polypeptide early in the folding process. In at least some proteins, such as 32 and RepA, Hsp40binding site(s) remain on their surface even after completing the folding steps and can be used as regulatory domains (Fig.  7A). An important question is why Hsp40 binding dominates Hsp70 binding. Major Hsp40s function in a dimer form. The binding of Hsp70 to a substrate through an interaction with Hsp40 is expected not only to raise substrate specificity but also to raise cooperativity among Hsp70 molecules for binding to a substrate.