Heptameric ring structure of the heat-shock protein ClpB, a protein-activated ATPase in Escherichia coli1

doi:10.1006/jmbi.2000.4165 available online at http://www.idealibrary.com on

J. Mol. Biol. (2000) 303, 655±666

Heptameric Ring Structure of the Heat-shock Protein ClpB, a Protein-activated ATPase in Escherichia coli Keun I. Kim1, Gang-Won Cheong2, Seong-Cheol Park2, Jung-Sook Ha2, Kee M. Woo3, Soo J. Choi1 and Chin H. Chung1* 1

School of Biological Sciences Seoul National University Seoul 151-742, Korea 2 Department of Biochemistry Gyeongsang National University, Chinju 660701, Korea 3

Department of Biochemistry College of Medicine Soonchunhyang University Cheonan, Choongnam 330-090, Korea

The heat-shock protein ClpB is a protein-activated ATPase that is essential for survival of Escherichia coli at high temperatures. ClpB has also recently been suggested to function as a chaperone in reactivation of aggregated proteins. In addition, the clpB gene has been shown to contain two translational initiation sites and therefore encode two polypeptides of different size. To determine the structural organization of ClpB, the ClpB proteins were subjected to chemical cross-linking analysis and electron microscopy. The average images of the ClpB proteins with endon orientation revealed a seven-membered, ring-shaped structure with a central cavity. Their side-on view showed a two-layered structure with an equal distribution of mass across the equatorial plane of the complex. Since the ClpB subunit has two large regions containing consensus sequences for nucleotide binding, each layer of the ClpB heptamer appears to represent the side projection of one of the major domains arranged on a ring. In the absence of salt and ATP, the ClpB proteins showed a high tendency to form a heptamer. However, they dissociated into various species of oligomers with smaller sizes, depending on salt concentration. Above 0.2 M NaCl, the ClpB proteins behaved most likely as a monomer in the absence of ATP, but assembled into a heptamer in its presence. Furthermore, mutations of the ®rst ATP-binding site, but not the second site, prevented the ATP-dependent oligomerization of the ClpB proteins in the presence of 0.3 M NaCl. These results indicate that ClpB has a heptameric ring-shaped structure with a central cavity and this structural organization requires ATP binding to the ®rst nucleotidebinding site localized to the N-terminal half of the ATPase. # 2000 Academic Press

*Corresponding author

Keywords: heat shock protein; protein-activated ATPase; ClpB; chaperone; Clp/Hsp104 family

Introduction Protease Ti (ClpAP) in Escherichia coli is an ATP-dependent protease consisting of two different polypeptide subunits, both of which are required for proteolysis (Goldberg, 1992; Gottesman & Maurizi, 1992; Chung et al., 1996). ClpA has two ATP-binding sites, while ClpP contains a serine active site for proteolysis. The isolated ClpA shows protein-activated ATPase activity, which in the reconstituted ClpAP complex is linked to protein breakdown (Hwang et al., The ®rst two authors contributed equally to this work. E-mail address of the corresponding author: [email protected] 0022-2836/00/050655±12 $35.00/0

1988; Katayama et al., 1988). ClpA is a member of the Clp/Hsp100 family of highly conserved proteins that have one or two regions of particularly high similarity, which contain a consensus sequence for ATP-binding (Gottesman et al., 1990; Schirmer et al., 1996). Members of the family include a second E. coli gene, called clpB, and genes from other bacteria, trypanosome, yeast and plants (Schirmer et al., 1996). Because of the sequence similarity of ClpB with that of ClpA, particularly in the ATP-binding regions (Gottesman et al., 1990), ClpB has been suggested to play an important role in ATP-dependent proteolytic processes in E. coli (Squires et al., 1991). However, it has been shown that ClpB cannot replace ClpA in supporting the ATP-dependent proteolysis by ClpP (Woo et al., 1992). # 2000 Academic Press

656 ClpB is a heat shock protein (Squires et al., 1991; Kitagawa et al., 1991) and has a protein-activated ATPase activity (Woo et al., 1992). The null mutant strain of the clpB gene showed a slower growth rate at 44  C and a higher rate of death above 50  C (Squires et al., 1991; Kim et al., 1998). In addition, it has been shown that the ``S fraction'' consisting of thermally aggregated, endogenous proteins can be stabilized in clpB null mutant cells (Laskowska et al., 1996). Hsp104, the yeast homolog of ClpB, is also a heat shock protein and plays an important role in thermotolerance (Sanchez & Lindquist, 1990). Unlike other chaperones, Hsp104 does not prevent the aggregation of denatured proteins (Glover & Lindquist, 1998). Instead, Hsp104 together with Hsp70 and Hsp40 can reactivate aggregated proteins generated upon thermal or chemical denaturation. Interestingly in thermophilic eubacteria, Thermus thermophilus, the clpB gene is a member of the dnaK gene cluster in the order dnaK-grpE-dnaJ-dafA-clpB (Motohashi et al., 1996). In this bacterium, TClpB also reactivates thermally aggregated proteins in cooperation with TDnaK, TDnaJ and TGrpE proteins (Motohashi et al., 1999). Recently, it has been reported that E. coli ClpB protein also shows chaperone activity (Zolkiewski, 1999; Goloubinoff et al., 1999). Similar to Hsp104 and TClpB, E. coli ClpB reactivates denatured and aggregated proteins in the presence of ATP and DnaK/DnaJ/GrpE. The clpB gene in E. coli has been shown to contain dual initiation sites for translation and therefore encode two polypeptides with different sizes, the 93 and 79 kDa ClpB proteins (referred to as ClpB93 and ClpB79, respectively) (Park et al., 1993). Similar to the wild-type ClpB protein, the puri®ed ClpB93 has an inherent ATPase activity that can be stimulated by proteins, such as casein and insulin. However, ClpB79 cannot be activated by the proteins, despite the fact that ClpB79 retains the two consensus sequences for ATP binding and also has inherent ATPase activity (Park et al., 1993). Thus, the N-terminal portion of ClpB93 has been suggested to contain one or more sites or domains responsible for protein binding. Although the biochemical properties of ClpB have been studied in detail, relatively little is known about the oligomeric structure of the ATPase. We have initially reported that the puri®ed ClpB protein behaves as a tetramer under conditions that gel ®ltration was performed in the presence of low salt concentrations (Woo et al., 1992). On the other hand, it has recently been reported that under conditions with 0.2 M KCl ClpB behaves as a monomer in the absence of ATP but can form a hexamer in its presence (Zolkiewski et al., 1999). In addition, it has been reported that ClpB79 inhibits the protein-activated ATPase activity of ClpB93 without any effect on the basal ATPase activity, and this inhibition may be mediated by formation of mixed oligomeric complexes (Park et al., 1993). Therefore, the present studies were carried out in order to investigate in

Heptameric Ring Structure of ClpB

detail the oligomeric structure of the ClpB proteins using both chemical cross-linking and electron microscopic analyses, the effect of ATP on oligomerization of their subunits and the interaction between ClpB93 and ClpB79 for possible formation of mixed oligomers.

Results Heptameric ring structure of ClpB In order to determine the oligomeric nature of the ClpB protein, ClpB93 and ClpB79 were incubated with glutaraldehyde for various periods, followed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and Coomassie blue staining as described in Materials and Methods. Figure 1 shows that the subunits of the ClpB proteins are cross-linked to various intermediates corresponding to the sizes from monomers to hexamers or heptamers. However, densitometric scanning of the gels generated by 0.25 minutes incubation revealed that the top bands were doublets (data not shown), suggesting that the gels contain in total seven bands. Upon prolonged incubation, no additional band appeared. These results suggest

Figure 1. Cross-linking of ClpB93 and ClpB79 with glutaraldehyde. The puri®ed ClpB93 (a) and ClpB79 proteins (b) (0.5 mg/ml) were cross-linked with 0.1 % glutaraldehyde for the indicated period of time at 25  C. After incubation, the samples were treated with one-fourth volume of 1 M Tris-HCl, and subjected to SDS-PAGE followed by staining with Coomassie blue R-250 as described in Materials and Methods. The size markers used were the cross-linked phosphorylase b proteins (from the bottom to the top: monomers to hexamers).

Heptameric Ring Structure of ClpB

that both ClpB93 and ClpB79 consist of seven subunits. To analyze further the oligomeric structure of ClpB, we performed electron microscopy using the puri®ed ClpB93 and poly-His-tagged ClpB93 (His-ClpB93: see below). The electron micrograph (Figure 2(a)) of the negatively stained ClpB93 showed two different views, ring-shaped and striated structures, depending on orientation on the grid. Most of the particles with end-on orientation showed ring-shaped structure, although a few particles showed a different aspect, either due to a different orientation of cylinder-shaped particle or due to stain distribution. On the other hand, the striated structure with side-on

657 orientation was less frequently observed, perhaps due to a tendency of the ClpB93 particles with top-view orientation to absorb to the carbon ®lm. A total of 849 top-views of well-stained particles were translationally aligned, and was subjected to multivariate statistical analysis (van Heel & Frank, 1981). The resulting eigenimages represent all important structural feature of the original data set. The averages of translationally, but not rotationally, aligned images and ten most signi®cant eigenimages are shown in Figure 2(b). The eigenimages showed the 7-fold rotational symmetry (Figure 2(b), panels 7 and 8) and less clearly the 6-fold symmetry (Figure 2(b), panel 10).

Figure 2. Electron micrograph of ClpB93 and multivariate statistical analysis for end-on oriented ClpB93. (a) Electron micrographs were obtained by negative staining of ClpB93 with 2 % uranyl acetate as described in Materials and Methods. The arrowhead indicates a representative of side-on view of ClpB93. The scale bar represents 100 nm. (b) Average (Av) of translationally, but not rotationally, aligned images and ten most signi®cant eigenimages (1-10) are shown. (c) Class averages (1-16) were derived from the rotationally aligned images using ten most signi®cant eigenvectors but without application of any symmetrization. The numerals shown below each panel are the number of particles seen in each class.

658 If the images have different rotational symmetries in the original data set, they could be separated according to the similarity of features based on eigenimages. For classi®cation, we selected ten most signi®cant eigenvectors. By this approach, 16 classes were discriminated according to the similarity of features after rotational alignment but without application of any symmetrization (Figure 2(c)). The class averages showed seven centers of mass arranged in a ring, which contained low or some density in its center. However, minor deviations from 7-fold symmetry are evident in almost all the 16 class averages, most likely due to an unintentional inclination during preparation or microscopy. Particularly, in two class averages shown in panels 3 and 13 of Figure 2(c), the subunits showed heterogeneous images, i.e. one subunit to be much larger than the others. This could occur due to a particle with incomplete stain embedding or to an unintentional inclination, thus two subunits producing one. In order to analyze further the rotational symmetry of top-on view images, the same data set was separated into many classes (10 to 30) using different eigenimages (10 to 20). Also the data set aligned with an arbitrarily chosen reference was separated according to the similarity of feature based on eigenimages. From

Figure 3. Correlation averages of the ClpB oligomers. (a) Correlation average of end-on views after symmetrization. The number of particles used for averaging were 849 and 230 for ClpB93 and His-ClpB93, respectively. Averages of the complexes show seven centers of masses arranged on a ring with some density in the center. (b) Average of side-on views after symmetrization. These averages show a two-layered structure with a dyad across the equatorial plane. The number of particles used for averaging were 278 and 120 for ClpB93 and His-ClpB93, respectively. The scale bar represents 10 nm.

Heptameric Ring Structure of ClpB

the resulting class averages, we could ®nd that these preparations of ClpB contain predominantly heptamers. The average of 849 top-on view revealed the 7fold symmetry (Figure 3(a)) with low density in its center. The diameters of the ring and the central hole were approximately 15 and 4 nm, respectively. Figure 3(b) shows the averaged image of 278 side-on view of the ClpB93 oligomer. This average revealed a two-layered structure with an equal distribution of mass across the equatorial plane of the complex. ClpA, the regulatory ATPase component of the ClpAP protease, has two large regions, each containing consensus sequences for nucleotide binding domains (Gottesman et al., 1990; Schirmer et al., 1996). Moreover, the electron microscopic structure of ClpA has shown that its subunits consist of the respective domains of similar size forming two stacked rings (Kessel et al., 1995; Beuron et al., 1998). In addition, the architecture and dimension of ClpB93 (15 nm  10 nm) are similar to the average of the ClpA hexamer (Kessel et al., 1995). Thus, it appears that the ClpB monomer also consists of two major domains of approximately equal size and each layer of the ClpB heptamer represents the side projection of one of the major domains arranged on a ring. The averaged images of His-ClpB93 were obtained after rotational and translational alignment. Like ClpB93, the average of top-view oriented His-ClpB93 revealed 7-fold center of mass arranged on a ring with a signi®cant density in the center (Figure 3(a)). The average showed a similar feature but with a minor deviation in the dimension of His-ClpB93 (14 nm  9 nm). In addition, the diameter of the central hole of His-ClpB93 (5 nm) appeared to be slightly larger than that of ClpB93 (4 nm). However, the nature of the density particularly in the center of His-Clp93 rings and the reason(s) for the minor deviation in its dimension are unknown at present. The average images of side-on view of His-ClpB93 also showed the two-layered structure with an approximately equal distribution of mass between the two layers (Figure 3(b)). These projected images as well as the cross-linked species (Figure 2) indicate that the ClpB proteins have predominantly a heptameric ring-shaped structure, in contrast to the recent report by Zolkiewski et al. (1999) suggesting a hexameric structure. Effects of salt and ATP on oligomerization of ClpB We have previously reported that, under typical gel-®ltration conditions with the proteins at 1 mg/ml in the presence of 50-100 mM NaCl, ClpB93 runs as an oligomer of 400-450 kDa whether or not ATP is present (Woo et al., 1992; Park et al., 1993). On the other hand, Zolkiewski et al. (1999) have recently reported that, under conditions with 0.2 M NaCl, ClpB behaves as a monomer in the absence of ATP, but can form a

Heptameric Ring Structure of ClpB

Figure 4. Effect of NaCl and ATP on oligomerization of ClpB93. (a) The puri®ed ClpB93 protein was chromatographed using a Shodex PROTEIN KW-804 column (1 cm  45 cm) in the presence and absence of the indicated amounts of NaCl or/and ATP. Fractions of 0.25 ml were collected at a ¯ow rate of 0.5 ml/minute, and aliquots (50 ml) of them were subjected to SDSPAGE. The gels were then stained with Coomassie blue R-250. The size markers used for the gel ®ltration were as follows: 1, thyroglobulin (669 kDa); 2, apoferritin (440 kDa); 3, b-amylase (200 kDa); 4, bovine serum albumin (66 kDa), which are indicated by the arrows. The numerals just above the gels indicate the fraction numbers from the gel ®ltration. (b) Electron micrographs of ClpB93 obtained as in Figure 2, but in the absence (panel 1) or presence of 0.3 M NaCl (panel 2) and both 0.3 M NaCl and 2 mM ATP (panel 3).

hexamer in its presence. In order to clarify this discrepancy, we ®rst examined the effects of increasing concentrations of NaCl in the absence of ATP on the oligomeric nature of ClpB. The puri®ed ClpB93 was subjected to gel ®ltration on a Shodex PROTEIN KW-804 column as described in Materials and Methods, followed by SDS-PAGE (Figure 4(a)). In accord with our earlier report (Woo et al., 1992; Kim et al., 1998), ClpB93 eluted as a broad peak corresponding to the size of 400450 kDa in the presence of 0.1 M NaCl. In the absence of the salt, ClpB93 also eluted as a broad peak but with a larger size (500-550 kDa). However, upon increasing the salt concentration, the ClpB93 protein further dissociated into oligomers with smaller molecular masses. In the presence of 0.3 M NaCl, ClpB93 ran as a sharp peak corresponding to a size of about 150 kDa. No further

659 change in the elution pattern of ClpB93 was observed with NaCl concentrations up to 0.5 M. In addition, we obtained exactly the same data when NaCl was replaced by KCl (data not shown). Similar data were obtained when the same experiments were performed with ClpB79 and the poly-Histagged form of the ClpB proteins (data not shown). These results suggest that ionic interaction between subunits is primarily responsible for ClpB oligomerization. We then examined the effect of ATP on oligomerization of ClpB93 in the presence of 0.3 M NaCl. ClpB93 was incubated with 0.3 M NaCl for ®ve minutes and together with 2 mM ATP for the next ®ve minutes at room temperature. The samples were then separated by gel ®ltration in the presence of the same concentrations of NaCl and ATP, followed by SDS-PAGE. As shown in Figure 4(a) (last panel), ATP shifted the elution of the ClpB93 protein in the fractions corresponding to the size of 500-550 kDa. These results indicate that ATP binding is required for oligomerization of ClpB in the presence of high salt concentrations. In order to determine whether ClpB93 indeed forms a heptamer in the presence of both ATP and NaCl, electron microscopy was performed under the same conditions. As shown in Figure 4(b), ClpB93 formed a heptamer in the presence of both 0.3 M NaCl and ATP (panel 3) as in their absence (panel 1). On the other hand, little or no oligomeric complex of ClpB93 could be seen in the presence of 0.3 M NaCl alone (panel 2). These results con®rm that ATP binding is crucial for assembly of the ClpB heptamer in the presence of high concentrations of salt but not in its absence. Effects of mutations in the ATP-binding sites on ClpB93 oligomerization In our previous study (Kim et al., 1998), we could not de®ne the role of any of the two ATPbinding sites in oligomerization of ClpB93 because the effects of high salt concentrations in the presence of ATP had not been examined. However, it is now clear that ATP binding plays a crucial role in formation of the ClpB heptamer under conditions of high ionic strength. Therefore, we examined whether any of the two ATP-binding sites involves in oligomerization of ClpB. The puri®ed ClpB93 and its mutant forms were subjected to gel ®ltration on a Shodex PROTEIN KW-804 column equilibrated with 0.3 M NaCl in the presence or absence 2 mM ATP. As shown in Figure 5, all of ClpB93 and its mutant forms ran as monomers or dimers in the absence of ATP (indicated by the dotted lines). In the presence of ATP, ClpB93 and ClpB93/K611T carrying the mutation in the second ATP-binding site (i.e. Lys661 was replaced by Thr) behaved as a heptamer. In contrast, ClpB93/K212T carrying the ®rst site mutation as well as ClpB93/ K212T/K611T carrying the mutation in both sites ran as monomers or dimers, whether or not ATP was present. On the other hand, in the absence of

660

Heptameric Ring Structure of ClpB

activity in a concentration-dependent fashion (Figure 6). In contrast, the most dramatic inhibition of ATP hydrolysis by ClpB93/K212T was observed with NaCl at the concentrations below 0.1 M. However, in the presence of 2 mM ATP, which is the same concentration used for both gel ®ltration and ATPase assay, ClpB93 and ClpB93/K611T retained their heptameric structure even in the presence of 0.3 M NaCl, although ClpB93/K212T could not (see Figure 5). In addition, removal of NaCl by dialysis could fully restore the ATPase activity of all of ClpB93 and its mutant forms (data not shown). These results suggest that the inhibitory effect of salt on ATP hydrolysis may be due to interference of high ionic strength with the interaction of the adenine nucleotide with the catalytic site of the ATPase. In the case of ClpB93/K212T, the mutation in the ®rst ATP-binding site may increase the sensitivity to ionic strength by an unknown mechanism, thus resulting in a strong inhibition of its ATPase activity at low salt concentrations. Mixed oligomer formation between ClpB93 and ClpB79

Figure 5. Effects of mutations in the ATP-binding sites on oligomerization of ClpB93. The puri®ed ClpB93 (Wt) and its mutant forms, ClpB93/K212T (K212T), ClpB93/K611T (K611T) and ClpB93/K212T/K611T (K212T/K611T), were subjected to gel ®ltration analysis in the absence (dotted lines) and presence of 2 mM ATP (continuous lines). The buffer used for the gel ®ltration contained 0.3 M NaCl. The size makers used are the same as in Figure 4.

both NaCl and ATP, all of the mutant proteins ran on the column in nearly the same fashion as the parental ClpB93 did (data not shown; see the top panel of Figure 4(a)). These results indicate that the ®rst ATP-binding site is responsible for assembly of ClpB93 into a heptamer in the presence of 0.3 M NaCl. We then examined the effects of increasing concentrations of NaCl on the ATPase activity of ClpB93 and its ATP-binding site mutants. NaCl at concentrations below 0.1 M showed relatively little effect on either of the basal or casein-activated ATP hydrolysis by ClpB93 and ClpB93/K611T, whereas the salt at above 0.1 M inhibited the ATPase

ClpB79 has been shown to inhibit the proteinactivated ATPase activity of ClpB93 but not its basal activity (Park et al., 1993). Therefore, it has been suggested that ClpB93 and ClpB79 form a mixed oligomeric complex, in which ClpB79 may block the interaction of ClpB93 with proteins. To examine whether the ClpB proteins can indeed form a mixed oligomer, we ®rst constructed the plasmids that can generate poly-His-tagged ClpB proteins, His-ClpB93 and His-ClpB79, and puri®ed them using Ni2‡-nitrilo-triacetic acid (NTA) agarose column as described in Materials and Methods. Upon analysis by SDS-PAGE, both of the Histagged proteins were found to be puri®ed to apparent homogeneity (Figure 7(a), lanes 2 and 4). The puri®ed ClpB93 and ClpB79 proteins were also shown as a control (lanes 1 and 3). We also examined whether the N-terminal modi®cation by poly-His might in¯uence the ATPase activity of the ClpB proteins. The ATPase activity of His-ClpB93 could no longer be activated by casein, but markedly increased to a similar extent seen with the parental ClpB93 in the presence of casein (Figure 7(b)). On the other hand, the same N-terminal modi®cation showed little or no effect on ATP hydrolysis by ClpB79 whether or not casein was present. We have previously shown that ATP hydrolysis by ClpB79 lacking the N-terminal 148 amino acid residues is not activated by casein, unlike ClpB93, and therefore suggested that the N-terminal region represents a domain responsible for binding with proteins (Park et al., 1993). Therefore, it appears that the N-terminal modi®cation by poly-His may mimic the binding of proteins to ClpB93. We then examined whether the ClpB proteins could indeed form mixed oligomers. A ®xed

Heptameric Ring Structure of ClpB

661

Figure 6. Effect of NaCl on the ATPase activity of ClpB93 and its ATP-binding site mutants. Reaction mixtures contained 2 mM ATP, 3 mg of ClpB93, ClpB93/K212T or ClpB93/K611T and increasing amounts of NaCl. They were incubated in the absence (open symbols) and presence of 10 mg of casein (closed symbols) for 30 minutes at 37  C. ATP hydrolysis was then assayed as described in Materials and Methods.

amount of His-ClpB93 was incubated for ®ve minutes with increasing amounts of ClpB79 in the presence of 0.3 M NaCl, at which concentration all of the ClpB proteins should be monomers or dimers. After incubation, the samples were treated with 2 mM ATP and further incubated for an additional ®ve minutes at 4  C. The samples were then loaded onto Ni2‡-NTA columns equilibrated with 50 mM imidazole. After collecting the ¯ow-through fractions, the bound proteins were eluted with 250 mM imidazole. Both the protein fractions were then subjected to SDS-PAGE. Upon raising the amounts of the ClpB79 protein incubated,

increased amounts of ClpB79 were recovered together with His-ClpB93 in the bound fractions (Figure 8), despite the fact that ClpB79 without the modi®cation of its N terminus by poly-His cannot bind to the af®nity column equilibrated with 50 mM imidazole. On the other hand, the ClpB79 protein that did not interact with His-ClpB93 was recovered in the ¯ow-through fractions. In addition, the recovery of ClpB79 in the bound fractions did not increase any further upon further increasing the amount of the protein in the incubation mixtures (data not shown), likely due to simultaneous rise in the probability of interaction

Figure 7. SDS-PAGE of poly-His-tagged ClpB proteins and their ATPase activity. (a) His-ClpB93 (lane 2) and HisClpB79 (lane 4) were puri®ed as descried in Materials and Methods, and aliquots of them (5 mg) were subjected to SDS-PAGE. ClpB93 (lane 1) and ClpB79 (lane 3) are also shown as a control. Lane M indicates the size marker proteins. (b) The ATPase activity of the ClpB proteins were assayed in the absence (open bars) and presence of 10 mg of casein (®lled bars). Reaction mixtures containing 3 mg of the puri®ed proteins and 2 mM ATP were incubated for 30 minutes at 37  C, and the inorganic phosphates released were determined. The results presented are the means of S.E. for triplicate experiments.

662

Heptameric Ring Structure of ClpB

Figure 8. Mixed oligomer formation between ClpB93 and ClpB79. In a ®nal volume of 0.2 ml, 15 mg of His-ClpB93 was incubated for ®ve minutes at 4  C with increasing amounts of ClpB79 in the presence of 0.3 M NaCl. After incubation, the samples were treated with 2 mM ATP and incubated again for ®ve minutes. The proteins were then loaded onto NTA-columns (total resin volume of 0.2 ml) equilibrated with 20 mM Tris-HCl (pH 7.8) containing 5 mM MgCl2, 2 mM ATP, 0.3 M NaCl and 50 mM imidazole. After collecting 0.4 ml of the ¯ow-through fraction (indicated by the letter F), the bound proteins (B) were eluted with the same volume of the equilibration buffer containing 250 mM imidazole. Aliquots of the fractions (70 ml) were electrophoresed and stained with Coomassie blue R250. As a control, 15 mg of His-ClpB93 or ClpB79 by itself was also treated as above. The numerals shown above the gel indicate the mass ratio of His-ClpB93 to ClpB79.

between the ClpB79 subunits by themselves. Similar data were obtained when His-ClpB79 and ClpB93 were used for the same experiments (data not shown). These results strongly suggest that ClpB79 forms mixed oligomeric complexes with ClpB93.

Discussion ClpB is a member of the Clp/Hsp100 family of highly conserved proteins that have one or two regions of ATP-binding sequences (Gottesman et al., 1990; Schirmer et al., 1996). Some of the family members, including the ClpA and ClpX ATPases, play a role not only as the regulatory components in the ATP-dependent proteolysis (Hwang et al., 1988; Katayama et al., 1988; Wojtkowiak et al., 1993; Gottesman et al., 1993) but also as molecular chaperones (Wickner et al., 1994; Wawrzynow et al., 1995). The other members, such as Hsp104 and ClpB, appear to function exclusively as a molecular chaperone that reactivates already aggregated proteins in the presence of DnaK/DnaJ/GrpE (Glover & Lindquist, 1998; Zolkiewski, 1999; Goloubinoff et al., 1999), thus being crucial for the survival of cells at high temperatures (Sanchez & Lindquist, 1990; Squires et al., 1991). All of these oligomeric proteins, except ClpB, have been shown to have similar structural organization. They dissociate into monomers, dimers, and/or trimers in the absence of ATP, but assemble into hexamers in its presence (Parsell et al., 1994; Singh & Maurizi, 1994; Seol et al., 1995; Grimaud et al., 1998). On the other hand, HslU (also called ClpY), which is the regulatory component of the ATP-dependent HslVU protease, forms both hexameric and heptameric ring structures in the presence of ATP and behaves as monomers and dimers in its absence (Shin et al., 1996; Rohrwild et al., 1997). In the present studies, we have demonstrated that ClpB has a heptameric ring-shaped structure. Chemical cross-linking analysis using glutaralde-

hyde revealed that ClpB is composed of seven subunits. Moreover, upon electron microscopic analysis, both ClpB93 and His-ClpB93 with end-on orientation were found to have predominantly a seven-membered ring-shaped structure with an opening of about 4 nm. Thus, it appears that ClpB only has a heptameric ring structure among the Clp/Hsp100 family member. With side-on orientation, both ClpB93 and His-ClpB93 showed a twolayer striated structure with a similar size separated from the equatorial plane of the heptamer. Signi®cantly, this structural feature of the ClpB proteins is highly similar to that of ClpA. Like ClpA, ClpB has two large regions, each containing consensus sequences for nucleotide binding (Gottesman et al., 1990; Schirmer et al., 1996). In addition, it has been shown that the ClpA subunits consist of the respective domains of similar size forming two stacked rings (Kessel et al., 1995; Beuron et al., 1998). Thus, it appears that the ClpB monomer also consists of two major domains of approximately equal size and each layer of the ClpB heptamer represents the side projection of one of the major domains arranged on a ring. Despite the similarity in the amino acid sequence and structural organization between ClpA and ClpB, the ClpB ATPase cannot interact with the ClpP tetradecamer. On the geometrical basis, the 7-fold symmetry of ClpA should be more favorable for its interaction with the 7-fold symmetric ClpP than 6-fold symmetry. Nevertheless, ClpA and ClpX, both having 6-fold symmetry (Beuron et al., 1998; Grimaud et al., 1998), interact with the 7-fold ClpP protease and stimulate degradation of proteins by ClpP in an ATP-dependent dependent manner. In addition, one form of HslVU (ClpYQ) exhibits a seven to six mismatch in the complex, although a 6-fold form of the HslU (ClpY) has also been observed (Rohrwild et al., 1997; Bochtler et al., 2000). Therefore, it has been suggested that a small relative rotation at the interface of 6-fold ATPase and 7-fold protease brings them into a pseudo-equivalent position and may

663

Heptameric Ring Structure of ClpB

facilitate unfolding or translocation of target proteins (Beuron et al., 1998; Voges et al., 1999). Noteworthy is the observation that the ClpB93 heptamer runs as a smaller protein complex (i.e. with the size of 500-550 kDa) on gel ®ltration (see Figures 4 and 5), compared to its calculated molecular mass (about 650 kDa). It is also noteworthy that the size of ClpB93 estimated in the presence of 0.3 M NaCl alone is about 150 kDa, which does not correspond to either the size of a monomer (93 kDa) or dimer (186 kDa). In the heptameric complex, the ClpB subunits may interact tightly with each other and form a compact structure, thus running faster in the gel ®ltration column than anticipated. On the other hand, the ClpB93 monomer might have behaved as a larger protein on the column than its actual size, probably due to its bi-lobed structure consisting of two regions of ATP-binding domains. Furthermore, upon the cross-linking analysis in the presence of 0.3 M NaCl alone, nearly all of the ClpB93 proteins remained as a monomer even after incubation for a prolonged period (e.g. 30 minutes) with glutaraldehyde (data not shown). Thus, it appears likely that ClpB behaves as a monomer in the presence of high concentrations of salts. In our previous report, we misinterpreted the effect of ATP and ATP-binding site mutations on the oligomerization of ClpB because the effects of increasing concentrations of salt had not been studied (Woo et al., 1992; Kim et al., 1998). In the present studies, however, we have clearly demonstrated that ClpB93 dissociates into various species of oligomers consisting of different numbers of subunits, depending on salt concentrations. At above 0.2 M NaCl, all of the ClpB93 proteins dissociated most likely into a monomer but could be reassembled into a heptameric complex upon ATP treatment. It is also noteworthy that ClpB93 forms a heptameric complex in the absence of both salt and ATP, unlike the other Clp/Hsp100 family members, such as ClpA and HslU, which do not form a hexamer without ATP whether or not salt is present (Seol et al., 1995; Shin et al., 1996). In addition, we have shown that the ®rst ATP-binding site, but not the second site, plays an essential role in the heptamer formation of ClpB, using the ClpB93 proteins containing mutations in either or both of the two highly conserved ATP-binding sites. Thus, it is clear that ClpB forms a heptameric complex in an ATP-dependent fashion, particularly under conditions with high salt concentrations, such as 0.3 M. Since E. coli cells contain a similar concentration of salt and an ATP concentration much higher than 2 mM, ClpB in vivo is likely to retain the heptameric ring-shaped structure for its cellular functions, such as in reactivation of already aggregated proteins. We have previously shown that the ®rst ATP-binding site of ClpA is responsible for ATP-dependent oligomerization and the second site is for its ATPase activity (Seol et al., 1995). In addition, it has been demonstrated that the ®rst

ATP-binding site is essential for the ATPase activity of Hsp104 while the second site is responsible for its oligomerization (Parsell et al., 1994). Thus, it appears that the ATP-dependent oligomerization property is common to all members of the Clp/Hsp100 family, although the functions of the ATP-binding sites are reversed in the case of Hsp104 and the role of the second ATP-binding site of ClpB remains unknown.

Materials and Methods Materials The recombinant plasmids, pBS/ClpB93, pBS/ClpB79, pBS/ClpB93/K212T, pBS/ClpB93/K611T and pBS/ ClpB93/K212T/K611T, were constructed as described previously (Park et al., 1993; Kim et al., 1998). Restriction endonucleases and other DNA-modifying enzymes were purchased from New England BioLabs and Takara (Japan). NTA-agarose was obtained from Qiagen. Oligonucleotide primers were synthesized using an automated DNA synthesizer (Applied Biosystem, model 384A). All other reagents were purchased from Sigma, unless otherwise indicated. In order to generate His-ClpB93, a SphI restriction site was introduced immediately upstream of the ATG start codon of the clpB93 gene by PCR-based site-directed mutagenesis. The SphI/PstI fragment containing the clpB93 gene was ligated into the pQE32 vector (Qiagen), which had been cut with the same restriction enzymes. The resulting His-ClpB93 contains additional 13 amino acid residues (MRGSHHHHHHGIR) to the N terminus of ClpB93. To construct His-ClpB79, the HpaI/HindIII fragment was cut out from pClpB93 (Park et al., 1993), and ligated into pQE32, which had been treated with SmaI/HindIII. The resulting His-ClpB79 contains additional 20 amino acid residues (MRGSHHHHHHGIRMRARYPG) to the N terminus of ClpB79. Protein purification E. coli cells overproducing the ClpB proteins were grown in Luria broth at 35  C to reach an absorbance (600 nm) of 0.7, and then shifted to 45  C. For induction of His-ClpB93, the cells were treated with 1 mM isopropyl-thio-b-D-galactoside instead of temperature shift. After culturing the cells for two hours under each condition, they were harvested and resuspended in 50 ml of 20 mM Tris-HCl (pH 7.8) containing 5 mM MgCl2. The cells were then disrupted with a French press at 14,000 psi and centrifuged for three hours at 100,000 g. The supernatants were dialyzed against the same buffer, and subjected to puri®cation of ClpB93, ClpB79 and the mutant forms of ClpB93 as described previously (Woo et al., 1992), but with an additional gel ®ltration step using a Superdex-200 column. His-ClpB93 and HisClpB79 were isolated using Ni2‡-NTA resin by following the standard procedure supplied by the manufacturer. The ClpB proteins were puri®ed to apparent homogeneity upon analysis by SDS-PAGE (Laemmli, 1970). Protein cross-linking The puri®ed ClpB proteins were dialyzed against 20 mM Hepes buffer (pH 7.8) containing 5 mM MgCl2.

664 The proteins were diluted with the same buffer to a ®nal concentration of 1 mg/ml, and divided into aliquots of 20 ml. Cross-linking reactions were initiated by treatment with 20 ml of 0.2 % (v/v) glutaraldehyde. After appropriate time intervals at 25  C, the reaction was stopped by addition of 40 ml of 1 M Tris-HCl (pH 7.0). The samples were concentrated to 10 ml using a vacuum dryer (Speed-vac, Servant), and treated with 10 ml of a phosphate buffer (pH 7.0) consisting of 28 mM NaH2PO4, 72 mM Na2HPO4, 1 % (w/v) SDS, 1 % (v/v) 2-mercaptoethanol, 0.1 % (w/v) bromophenol blue and 6 M urea. They were boiled for ®ve minutes, and subjected to electrophoresis using 3 % (w/v) polyacrylamide tube gels under denaturing conditions as described by Webber et al. (1972). The gels were stained with Coomassie brilliant blue R-250. Electron microscopy The puri®ed ClpB proteins were applied to glowdischarged carbon-coated copper grids. After allowing the proteins to absorb for one to two minutes, the grids were rinsed on droplets of deionized water, and stained with 2 % (w/v) uranyl acetate. Specimens were examined in the Jeol 2010 at an accelerating voltage of 120 kV using low-dose unit. Electron micrographs were recorded on Kodak ®lm (SO163) at a magni®cation of 38,700 (nominal magni®cation of 40,000). Image processing Light-optical diffractograms were used to select the micrographs, to examine the defocus and to verify that no drift or astigmastism was present. Suitable areas were digitized as arrays of 1024  1024 pixels with Leaf Scan 45 at pixel size of 20 mm corresponding to 0.52 nm at specimen level. For image processing, the SEMPER (Saxton et al., 1979) and EM (Hergel, 1996) software packages were used. From digitized micrographs, smaller frames of 64  64 pixels containing individual particles were extracted interactively. The side-view images were aligned translationally and rotationally using standard correlation methods (Baumeister et al., 1988; Phipps et al., 1991). For analyzing the rotational symmetry of top-on view images, the aligned images were subjected to multivariate statistical analysis (van Heel & Frank, 1981). The individual images were aligned translationally but not rotationally as described by Marco et al. (1994). The resulting eigenimages represent all important structural features of the original data set. If the images have different rotational symmetries in the original data set, the eigenimages reveal the different symmetry axes. Moreover, these images can be distinguished and subsequently separated based on eigenimages. The rotationally aligned images were classi®ed based on eigenvectoreigenvalue data analysis, and subsequent averaging was performed for each class separately. Gel filtration analysis Gel ®ltration experiments were performed at room temperature using a Shodex PROTEIN KW-804 column (Showa Denko, Japan) with Waters HPLC system equipped with Waters 996 photodiode array detector. Aliquots (100 ml) of the puri®ed ClpB proteins at a concentration of 0.5 mg/ml were chromatographed on the column equilibrated with 20 mM Tris-HCl (pH 7.8) and

Heptameric Ring Structure of ClpB 5 mM MgCl2 in the presence or absence of various concentrations of NaCl and/or ATP. Assays ATP hydrolysis was assayed by incubation of the reaction mixtures (0.1 ml) containing appropriate amounts of the ClpB proteins and 2 mM [g-32P]ATP in 100 mM TrisHCl (pH 8.0) buffer containing 10 mM MgCl2, 1 mM DTT and 1 mM EDTA. After incubation at 37  C for appropriate periods, the reaction was terminated by sequential addition of 0.1 ml of 0.1 M HCl, 0.1 ml of 2 mg/ml bovine serum albumin and 0.25 ml of 10 % (w/v) suspension of activated charcoal. The samples, were stirred on a vortex mixer, incubated for 15 minutes on ice and centrifuged at 15,000 g for ®ve minutes. The radioactivity of free inorganic phosphates released into the supernatant fractions were then determined using a liquid-scintillation counter. Proteins were assayed by their absorbance at 280 nm or by the dye-binding method (Bradford, 1976) using bovine serum albumin as a standard.

Acknowledgments We are grateful to Dr W. Baumeister for critical reading of this manuscript. We thank Drs R. Hegerl and D. Typke (Max-Planck-Institute for Biochemistry, Germany) for their assistance in using the EM software package. Mr Soo J. Choi is a recipient of a BK21 fellowship. This work was supported by grants from Korea Science and Engineering Foundation through Research Center for Proteineous Materials, The Korea Ministry of Education (Non-directed Research Fund), and Lotte Foundation.

References Baumeister, W., Dahlmann, B., Hegerl, R., Kopp, F., Luehn, L. & Pfeifer, G. (1988). Electron microscopy and image analysis of the multicatalytic proteinase. FEBS Letters, 241, 239-245. Beuron, F., Maurizi, M. R., Belnap, D. M., Kocsis, E., Booy, F. P., Kessel, M. & Steven, A. C. (1998). At sixes and sevens: characterization of the symmetry mismatch of the ClpAP chaperone-assisted protease. J. Struct. Biol. 123, 248-259. Bochtler, M., Hartmann, C., Song, H. K., Bourenkov, G. P., Bartunik, H. D. & Huber, R. (2000). The structures of HslU and the ATP-dependent protease HslU-HslV. Nature, 403, 800-805. Bradford, M. M. (1976). A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72, 248254. Chung, C. H., Seol, J. H. & Kang, M. S. (1996). Protease Ti (Clp), a multi-component ATP-dependent protease in Escherichia coli. Biol. Chem. 377, 549-554. Glover, J. R. & Lindquist, S. (1998). Hsp104, Hsp70, and Hsp40: a novel chaperone system that rescues previously aggregated proteins. Cell, 94, 73-82. Goldberg, A. L. (1992). The mechanism and functions of ATP-dependent proteases in bacterial and animal cells. Eur. J. Biochem. 203, 9-23. Goloubinoff, P., Mogk, A., Zvi, A. P. B., Tomoyasu, T. & Bukau, B. (1999). Sequential mechanism of solubil-

Heptameric Ring Structure of ClpB ization and refolding of stable protein aggregates by a bichaperone network. Proc. Natl Acad. Sci. USA, 96, 13732-13737. Gottesman, S. & Maurizi, M. R. (1992). Regulation by proteolysis: energy-dependent proteases and their targets. Microbiol. Rev. 56, 592-621. Gottesman, S., Squires, C., Pichersky, E., Carrington, M., Hobbs, M., Mattick, J. S., Dalrymple, B., Kuramitsu, H., Shioza, T., Foster, T., Clark, W. C., Ross, B., Squires, C. L. & Maurizi, M. R. (1990). Conservation of the regulatory subunit for the Clp ATP-dependent protease in prokaryotes and eukaryotes. Proc. Natl Acad. Sci. USA, 87, 3513-3517. Gottesman, S., Clark, W. P., de Crecy-Lagard, V. & Maurizi, M. R. (1993). ClpX, an alternative subunit for the ATP-dependent Clp protease of Escherichia coli. Sequence and in vivo activities. J. Biol. Chem. 268, 22618-22626. Grimaud, R., Kessel, M., Beuron, F., Steven, A. C. & Maurizi, M. R. (1998). Enzymatic and structural similarities between the Escherichia coli ATP-dependent proteases, ClpXP and ClpAP. J. Biol. Chem. 273, 12476-12481. Hergel, R. (1996). The EM program package: a platform for image processing in biological electron microscopy. J. Struct. Biol. 116, 30-34. Hwang, B. J., Woo, K. M., Goldberg, A. L. & Chung, C. H. (1988). Protease Ti, a new ATP-dependent protease in Escherichia coli, contains proteinactivated ATPase and proteolytic functions in distinct subunits. J. Biol. Chem. 263, 8727-8734. Katayama, Y., Gottesman, S., Pumphrey, J., Rudikoff, S., Clark, W. P. & Maurizi, M. R. (1988). The two-component, ATP-dependent Clp protease of Escherichia coli. Puri®cation, cloning, and mutational analysis of the ATP-binding component. J. Biol. Chem. 263, 15226-15236. Kessel, M., Maurizi, M. R., Kim, B., Kocsis, E., Trus, B. L., Singh, S. K. & Steven, A. C. (1995). Homology in structural organization between E. coli ClpAP protease and the eukaryotic 26 S proteasome. J. Mol. Biol. 250, 587-594. Kim, K. I., Woo, K. M., Seong, I. S., Lee, Z. W., Baek, S. H. & Chung, C. H. (1998). Mutational analysis of the two ATP-binding sites in ClpB, a heat shock protein with protein-activated ATPase activity in Escherichia coli. Biochem. J. 333, 671-676. Kitagawa, M., Wada, C., Yoshioka, S. & Yura, T. (1991). Expression of ClpB, an analog of the ATP-dependent protease regulatory subunit in Escherichia coli, is controlled by a heat shock sigma factor (sigma 32). J. Bacteriol. 173, 4247-4253. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature, 227, 680-685. Laskowska, E., Kuczynska-Wisnik, D., Skorko-Glonek, J. & Taylor, A. (1996). Degradation by proteases Lon, Clp and HtrA, of Escherichia coli proteins aggregated in vivo by heat shock; HtrA protease action in vivo and in vitro. Mol. Microbiol. 22, 555-571. Marco, S., Urena, D., Carrascosa, J. L., Waldmann, T., Peter, J., Hegerl, R., Pfeifer, G., Sack-Kongehl, H. & Baumeister, W. (1994). The molecular chaperone TF55: assessment of symmetry. FEBS Letters, 341, 152-155. Motohashi, K., Yohda, M., Endo, I. & Yoshida, M. (1996). A novel factor required for the assembly of the DnaK and DnaJ chaperones of Thermus thermophilus. J. Biol. Chem. 271, 17343-17348.

665 Motohashi, K., Watanabe, Y., Yohda, M. & Yoshida, M. (1999). Heat-inactivated proteins are rescued by the DnaKJ-GrpE set and ClpB chaperones. Proc. Natl Acad. Sci. USA, 96, 7184-7189. Park, S. K., Kim, K. I., Woo, K. M., Seol, J. H., Tanaka, K., Ichihara, A., Ha, D. B. & Chung, C. H. (1993). Site-directed mutagenesis of the dual translational initiation sites of the clpB gene of Escherichia coli and characterization of its gene products. J. Biol. Chem. 268, 20170-2017. Parsell, D. A., Kowal, A. S. & Linquist, S. (1994). Saccharomyces cerevisiae Hsp104 protein. Puri®cation and characterization of ATP-induced structural changes. J. Biol. Chem. 269, 4480-4487. Phipps, B. M., Hoffmann, A., Stetter, K. O. & Baumeister, W. (1991). A novel ATPase complex selectively accumulated upon heat shock is a major cellular component of thermophilic archaebacteria. EMBO J. 10, 1711-1722. Rohrwild, M., Pfeifer, G., Santarius, U., Muller, S. A., Huang, H. C., Engel, A., Baumeister, W. & Goldberg, A. L. (1997). The ATP-dependent HslVU protease from Escherichia coli is a four-ring structure resembling the proteasome. Nature Biol. 4, 133-139. Sanchez, Y. & Lindquist, S. (1990). HSP104 required for induced thermotolerance. Science, 248, 1112-1115. Saxton, W. O., Pitt, T. J. & Horner, M. (1979). Digital image processing: the Semper system. Ultramicroscopy, 4, 343-354. Schirmer, E. C., Glover, J. R., Singer, M. A. & Lindquist, S. (1996). HSP100/Clp proteins: a common mechanism explains diverse functions. Trends Biochem. Sci. 21, 289-296. Seol, J. H., Baek, S. H., Kang, M. S., Ha, D. B. & Chung, C. H. (1995). Distinctive roles of the two ATPbinding sites in ClpA, the ATPase component of protease Ti in Escherichia coli. J. Biol. Chem. 270, 8087-8092. Shin, D. H., Yoo, S. J., Shim, Y. K., Seol, J. H., Kang, M. S. & Chung, C. H. (1996). Mutational analysis of the ATP-binding site in HslU, the ATPase component of HslVU protease in Escherichia coli. FEBS Letters, 398, 151-154. Singh, S. K. & Maurizi, M. R. (1994). Mutational analysis demonstrates different functional roles for the two ATP-binding sites in ClpAP protease from Escherichia coli. J. Biol. Chem. 269, 29537-29545. Squires, C. L., Pedersen, S., Ross, B. M. & Squires, C. (1991). ClpB is the Escherichia coli heat shock protein F84.1. J. Bacteriol. 173, 4254-4262. van Heel, M. & Frank, J. (1981). Use of multivariate statistics in analyzing the images of biological macromolecules. Ultramicroscopy, 6, 187-194. Voges, D., Zwickl, P. & Baumeister, W. (1999). The 26S proteasome: a molecular machine designed for controlled proteolysis. Annu. Rev. Biochem. 68, 10151068. Wawrzynow, A., Wojtkowiak, D., Marszalek, J., Banecki, B., Jonsen, M., Graves, B., Georgopoulos, C. & Zylicz, M. (1995). The ClpX heat-shock protein of Escherichia coli, the ATP-dependent substrate speci®city component of the ClpP-ClpX protease, is a novel molecular chaperone. EMBO J. 14, 18671877. Weber, K., Pringle, J. R. & Osborn, M. (1972). Measurement of molecular weights by electrophoresis on SDS-acrylamide gel. Methods Enzymol. 26, 3-27. Wickner, S., Gottesman, S., Skowyra, D., Hoskins, J., McKenney, K. & Maurizi, M. R. (1994). A molecular

666

Heptameric Ring Structure of ClpB

chaperone, ClpA, functions like DnaK and DnaJ. Proc. Natl Acad. Sci. USA, 91, 12218-12222. Wojtkowiak, D., Georgopoulos, C. & Zylicz, M. (1993). Isolation and characterization of ClpX, a new ATP-dependent speci®city component of the Clp protease of Escherichia coli. J. Biol. Chem. 268, 2260922617. Woo, K. M., Kim, K. I., Goldberg, A. L., Ha, D. B. & Chung, C. H. (1992). The heat-shock protein ClpB

in Escherichia coli is a protein-activated ATPase. J. Biol. Chem. 267, 20429-20434. Zolkiewski, M. (1999). ClpB cooperates with DnaK, DnaJ, and GrpE in suppressing protein aggregation. A novel multi-chaperone system from Escherichia coli. J. Biol. Chem. 274, 28083-28086. Zolkiewski, M., Kessel, M., Ginsburg, A. & Maurizi, M. R. (1999). Nucleotide-dependent oligomerization of ClpB from Escherichia coli. Protein Sci. 8, 1899-1903.

Edited by W. Baumeister (Received 12 July 2000; received in revised form 4 September 2000; accepted 5 September 2000)