Crystal and Solution Structures of an HslUV Protease–Chaperone Complex

Cell, Vol. 103, 633–643, November 10, 2000, Copyright 2000 by Cell Press

Crystal and Solution Structures of an HslUV Protease–Chaperone Complex Marcelo C. Sousa,* Christine B. Trame,* Hiro Tsuruta,† Sigurd M. Wilbanks,*k Vijay S. Reddy,‡ and David B. McKay*§ * Department of Structural Biology Stanford University School of Medicine Stanford, California 94305 † Stanford Synchrotron Radiation Laboratory SLAC Stanford University Stanford, California 94309 ‡ Department of Molecular Biology The Scripps Research Institute 10550 North Torrey Pines Road La Jolla, California 92037

Summary HslUV is a “prokaryotic proteasome” composed of the HslV protease and the HslU ATPase, a chaperone of the Clp/Hsp100 family. The 3.4 A˚ crystal structure of an HslUV complex is presented here. Two hexameric ATP binding rings of HslU bind intimately to opposite sides of the HslV protease; the HslU “intermediate domains” extend outward from the complex. The solution structure of HslUV, derived from small angle X-ray scattering data under conditions where the complex is assembled and active, agrees with this crystallographic structure. When the complex forms, the carboxy-terminal helices of HslU distend and bind between subunits of HslV, and the apical helices of HslV shift substantially, transmitting a conformational change to the active site region of the protease. Introduction The HslUV complex (for heat shock locus gene products U and V; Chuang et al., 1993) is a bacterial homolog of the eukaryotic proteasome (Missiakas et al., 1996; Rohrwild et al., 1996; Yoo et al., 1996). It degrades specific target proteins in an ATP-dependent, chaperoneassisted manner. The protease component, HslV, is a “double donut” of hexameric rings, with the peptidase catalytic sites located in an interior cavity of the complex (Kessel et al., 1996; Rohrwild et al., 1997). Structures of the E. coli HslV protease (Bochtler et al., 1997) and the structurally related archaeal Thermoplasma acidophilum (Lo¨we et al., 1995) and eukaryotic yeast (Groll et al., 1997) proteasomes are known. By itself, HslV has a relatively low peptidase activity on short peptide substrates and negligible “polypeptidase” activity on protein substrates; the proteolytic activity of HslV is enhanced one to two orders of magnitude by HslU, § To whom correspondence should be addressed (e-mail: dave.

[email protected]). k Present address: Biochemistry Department, University of Otago, Dunedin, New Zealand.

hexameric rings of which bind HslV to form the active HslUV complex (Huang and Goldberg, 1997; Seol et al., 1997; Yoo et al., 1997). HslU is a member of the “Clp/Hsp100” family of molecular chaperones (Schirmer et al., 1996). Functions that have been attributed to members of this family include facilitating the degradation of target proteins by cognate proteases, as well as disassembly of oligomeric protein assemblies. For example, ClpX and ClpA target ssrAtagged polypeptides to the ClpP protease for degradation (Keiler et al., 1996; Gottesman et al., 1998); in an additional and contrasting activity, ClpX can disassemble a tight tetrameric, DNA bound complex of the MuA transposase (Levchenko et al., 1995). Recent studies using ssrA-tagged green fluorescent protein (GFP) as a substrate have shown that ClpA and ClpX unfold GFP in an ATP-dependent process, supporting a model whereby ClpX or ClpA would actively denature a polypeptide substrate and translocate it to the catalytic compartment of the ClpP protease (Weber-Ban et al., 1999; Hoskins et al., 2000; Kim et al., 2000; Singh et al., 2000). HslU has functional parallels with ClpX and ClpA. Degradation of polypeptides by the HslUV complex has an ATP requirement (although ATP hydrolysis does not seem to be mandatory (Huang and Goldberg, 1997)), and for at least one substrate, the SulA protein, HslU recognizes a carboxy-terminal “tag” sequence (Kanemori et al., 1999; Seong et al., 1999), reminiscent of recognition of the carboxy-terminal ssrA tag by ClpX and ClpA. One can anticipate that the mechanisms by which the ClpX and ClpA proteins, on the one hand, and the homologous HslU protein on the other, unfold and translocate polypeptides to their respective proteases will share many similarities, despite the dissimilarities of the ClpP and HslV proteases. Proteins in the Clp/Hsp100 family have either one (e.g., HslU and ClpX) or two (e.g., ClpA) ATP binding domains. Sequence analysis suggested initially (Neuwald et al., 1998), and the crystallographic structure of the E. coli HslU protein confirmed unambiguously (Bochtler et al., 2000), that the Clp/Hsp100 proteins are members of the extended AAA (for ATPases associated with a variety of cellular activities; Neuwald et al., 1998) family, a group of proteins whose diverse activities often require ATP-modulated assembly of oligomeric (often hexameric) ring structures. The HslU protomer has three structural domains. The amino terminal, ATP binding domain (“N domain”), and the carboxy-terminal domain are similar in structure to the “D2” ATP binding fragment of a distantly related AAA protein, the NEM-sensitive fusion protein (NSF) which participates in vesicle fusion, and which also assembles into a hexameric ring when ATP binds (Lenzen et al., 1998; Yu et al., 1998). A feature of HslU that is not found in other AAA proteins is an intermediate domain (“I domain”) of approximately 130 amino acids that interrupts the polypeptide sequence of the N domain and extends outward in a tentacle-like manner from the HslU ring. Thus, HslU shows structural similarity with other AAA proteins, which are presumably mandated to some extent by functional parallels, partic-

Cell 634

Table 1. Crystallographic Data Collection and Refinement Statistics Data Collection Cell parameters a (A˚) b (A˚) c (A˚) Observations Unique reflections Resolution range (A˚) (last shell) Completeness Av. Redundancy Average I/␴ Rsyma Refinement Final resolution in refinement Rcrystb (30–3.4) Rfreeb (30–3.4) Number of reflections (working set) Number of reflections (test set) Protein ⫹ ATP atoms rmsc bond deviation (A˚) rmsc angle deviation (deg)

209.22 220.58 241.07 297,864 134,912 30.0–3.40 (3.52–3.40) 0.891 (0.806) 2.2 (1.8) 11.5 (3.0) 0.064 (0.285) 30.0–3.4 0.241 0.285 120,431 6,383 45,556 0.010 1.42

Numbers in parentheses correspond to the last resolution shell. a Rsym ⫽ ⌺|Ii ⫺ ⬍I⬎|/⌺⬍I⬎ where I ⫽ diffraction intensity; ⬍I⬎ ⫽ mean measured intensity. b Rcryst ⫽ ⌺|Fc ⫺ Fo|/⌺Fo where Fo ⫽ observed structure factor amplitude in working set ⫽ |I1/2|; Fc ⫽ structure factor amplitude calculated from model; Rfree same, using Fo in test set (5.0% of reflections). c rms, root mean square.

ularly in the mechanism by which ATP modulates the oligomeric state of the proteins; at the same time, the I domains are a unique structural feature which one expects to be HslU-specific in function. A structure for an E. coli HslUV complex has been reported (Bochtler et al., 2000). The quaternary structure places the ATP binding rings of HslU distal to the HslV protease, with the I domains bridging the interstitial space between them. This model has been difficult to reconcile with electron microscopy data that envision a more compact structure (Missiakas et al., 1996; Rohrwild et al., 1997), and how this complex participates in the ATP-dependent proteolytic reaction cycle has not become apparent. We have crystallized and solved the structure of an HslUV complex from H. influenzae, whose HslU and HslV proteins share approximately 80% amino acid sequence identity with their E. coli counterparts. In addition, we have carried out small angle X-ray scattering (SAXS) experiments on HslUV to determine its structure in solution. These studies reveal an alternative quaternary structure for the HslUV complex under conditions where peptidase assays confirm that it is active. Results Quaternary Structure of HslUV in Crystals We have solved the structure of the H. influenzae HslUV complex to 3.4 A˚ using molecular replacement and subsequent rebuilding with models for HslV and HslU that have been refined independently to 1.9 A˚ and 2.3 A˚, respectively (Table 1). Representative Fo–Fc omit maps are shown in Figure 1; Figure 1A shows a well-ordered region of the structure, specifically an ATP bound to

one HslU subunit; Figure 1B shows the carboxy-terminal segment of HslU, which, in contrast, is one of the more poorly ordered regions of the structure. The HslV protomers are generally well-defined, as are the N and C domains of the HslU protomers. Two short loops in the HslU subunits, residues 88–94 and 266–269, could not be traced unambiguously. Additionally, the distal segments of the I domains, approximately residues 120–230 (with minor differences from one subunit to another) are disordered and cannot be traced. (The I domains are characteristically disordered, as revealed in three other crystallographic structures (Bochtler et al., 2000) as well as in this one.) Figure 2A illustrates the crystal packing; there is one U12V12 complex per asymmetric unit, held in place primarily by side-to-side interactions between the HslU hexameric rings. This results in alternating packing “layers” formed by the HslU rings perpendicular to the crystallographic c axis; fortuitously, one layer is well-ordered (the “blue” layer in Figure 2A, in which the atoms are colored by B factor), and the other is more disordered (the “yellow/red” layer in Figure 2A), with the consequence that one end of the HslUV complex (which we will refer to colloquially as the “upper” or “top” half) is more well-ordered in the crystal, and hence more precisely defined in the structure, than the other (which we refer to as the “lower” half). The average B factors for mainchain atoms in the final model are 38.6 A˚2, 57.6 A˚2, 89.6 A˚2, and 159.2 A˚2 for the upper HslU ring (computed from the N and C domains, excluding the I domain), upper HslV hexamer, lower HslV hexamer, and lower HslU ring, respectively. When the upper half of the final model is superimposed on the lower half, the root mean square (rms) difference in mainchain atomic positions is 0.93 A˚; we do not see any major differences in structure between the top and bottom halves of the complex in this crystal form. In the HslU hexamer, there is a cleft between the N and C domains of the protomers. In HslV, a pair of helices protrude from each protomer on the top and bottom (apical) surfaces of the dodecamer. In our HslUV complex, the apical helices of HslV protomers bind in the cleft between the N and C domains of HslU protomers to form a double-capped complex; the ATP binding “rings” of the HslU hexamers are in intimate contact with the HslV protease, and the I domains extend outward from the complex (Figure 2B). The overall dimensions of the complex are: the HslV dodecamer is ⵑ75 A˚ high and ⵑ110 A˚ in diameter. The rings of the HslU hexamers are ⵑ35–40 A˚ high and ⵑ125 A˚ in diameter, giving a net height of ⵑ150 A˚ for the complex before the I domains are included. The portion of the I domains that we have been able to build in these crystals extend ⵑ15–20 A˚ above the HslU rings. The overall shape of the HslUV complex is similar to that of side views that are seen with negative staining by electron microscopy; the overall dimensions agree to within 10% with those reported from electron microscopy images (Kessel et al., 1996; Rohrwild et al., 1997). Quaternary Structure of HslUV in Solution In the course of developing a purification protocol for HslUV, we found that in the presence of 1 M KCl, HslU and HslV proteins coelute as a large complex (⬎600

HslUV Structure 635

Figure 1. Representative Electron Density Maps Stereo views of Fo ⫺ Fc simulated annealing omit maps, computed with phases calculated from models in which the atoms of interest were deleted from the model used in refinement. (A) the ATP binding site of HslU, contoured at 5␴. Protein is shown as a ribbon diagram; ATP from the final HslUV model (average B factor 29.3) is shown as a ball and stick representation. (B) Carboxy-terminal segment of HslU (average B factor 119.1), contoured at 3␴ (magenta) and 6␴ (cyan). Residues of HslU which were omitted are shown in green, oriented with the carboxy-terminal Leu-444 at the bottom of the figure; neighboring residues of HslV are shown in standard colors (oxygen, red; nitrogen, blue; carbon, gray). Figure was prepared with BOBSCRIPT (Esnouf, 1997, 1999). The rendering and stereo pair generation of all figures was done with RASTER3D (Merritt and Bacon, 1997) and IMAGEMAGIK (http://www.wizards.dupont.com/cristy/ ImageMagick.html).

kDa, the exclusion size of the column) by gel filtration on Superdex 200. The crystals of the HslUV complex used in this work were, therefore, grown in the presence of 1 M KCl. We have also found that HslUV chromatographs on Superose 6 as a complex of molecular weight ⬎ 660 kDa in low ionic strength, as long as a significant concentration of Mg2⫹ is present. We have carried out solution small angle X-ray scattering (SAXS) experiments on HslUV under conditions where it forms a stable complex in solution. Scattering curves were measured under steady-state conditions, both with and without the addition of 1 mM ATP (Figure 3). Indirect Fourier transforms of the experimental scattering curves were computed to give the pair distribution function (P(r) function, which is a histogram of the lengths of all interatomic (electron) vectors of the molecule); values for the radius of gyration (Rg) were computed from the P(r) function as described in Experimental Procedures. In the presence of ATP, the experimental Rg is 71.9 ⫾ 0.5 A˚; without addition of ATP, it is 72.8 ⫾ 0.4 A˚; in both cases, the maximum dimension of the

protein complex (dmax) estimated from the experimental data is 235 ⫾ 5 A˚. We have calculated theoretical scattering curves and P(r) functions from atomic models and compared them to the experimental results (Figure 3). Using the crystallographic HslUV model reported here, the computed Rg is 63.8 A˚. Comparison of the model and experimental P(r) functions shows that the model is underrepresented in the longer interatomic distances. The fit of the calculated scattering curve to the experimental curve has a ␹2 value of 1.09. Since the I domains will contribute significantly to the solution scattering curves, but are poorly defined in the HslUV model reported here, we have spliced additional segments of the I domains from crystallographic models of HslU alone (Bochtler et al., 2000) onto the HslUV model. The result of adding approximately 50 of the 90 missing residues in the I domain is shown in Figure 3 (middle panel); the calculated Rg is 69.0 A˚ and the ␹2 value for the fit to the experimental scattering curve is 0.96. The agreement to the experimental P(r) function

Cell 636

Figure 2. Structure of the HslUV Complex (A) Crystal packing in orthorhombic HslUV crystals. Coil diagram of five HslUV complexes related by crystallographic symmetry, colored by temperature factor from blue (lowest) to red (highest). Crystallographic cell axes are shown in red; the c axis is vertical; the a–b plane is horizontal and perpendicular to the plane of the page. (B) Ribbon diagram of the HslUV complex; helices are shown as cylinders. HslU is colored magenta except for a single protomer from each ring, which is highlighted in blue. HslV is colored yellow except for a single protomer in each of the upper and lower rings, which is highlighted in red. Figures 2 and 4–6 were prepared with MOLSCRIPT (Kraulis, 1991).

is significantly better. Since the I domains in the model are not complete, they only partially account for the discrepancy between experimental and calculated scattering curves and Rg values; however, the calculated Rg value is in closer agreement with the measured value. Using the model reported for the E. coli HslUV complex (Bochtler et al., 2000), the fit of the calculated scattering curve to the experimental curve has a ␹2 value of 2.26. The calculated Rg is 91.5 A˚, the calculated dmax is ⬎250 A˚ and the P(r) function disagrees substantially from the experimental curve, particularly at large interatomic distances (ⱖ200 A˚), where there is a large peak due primarily to the interatomic vectors between the ATP binding rings of HslU on opposite ends of the complex. Peptide Hydrolysis Activity of HslUV in Solution and in Crystals In order to verify that the H. influenzae HslUV protein used in these studies was enzymatically active, material purified by recrystallization was redissolved and assayed for peptidase activity using a small chromogenic peptide as a substrate (described in Experimental Procedures). The specific peptidase activity of our HslUV was 0.7 (nmol min⫺1 ␮g⫺1), which compares well with the reported values for the activity of E. coli HslUV, 0.27 (nmol min⫺1 ␮g⫺1) (Yoo et al., 1996). To verify that the HslUV complex in the crystals was active, peptidase

activity was carried out on a suspension of microcrystals; Figure 4 illustrates the activity of crystalline HslUV. Changes in Structure when the HslUV Complex Forms Comparison of the structures of the HslU and HslV protomers in the HslUV complex with their structures in crystals of uncomplexed H. influenzae HslU (C. B. T. and D. B. M., unpublished data) and HslV (M. C. S. and D. B. M., unpublished data) reveals significant conformational changes, suggesting a mechanism of allosteric activation of HslV by HslU. These changes include (1) the N- and C-terminal domains of HslU flex slightly around the hinge between them, changing the orientation of one relative to the other by ⵑ3⬚, (2) the carboxyterminal 14 residues of HslU, which form a short, wellordered loop and helix that bind to the C domain in uncomplexed HslU (red segment in Figure 5A) distend to give an extended linker and a short helix that binds between HslV protomers (blue segment extended from HslU in Figure 5A), and (3) the pair of apical helices on the HslV protomer that bind to HslU are shifted substantially, in a manner that propagates a conformational change from the apical surfaces to the active site clefts of the HslV protease (yellow and blue segments of HslV in Figure 3A). (It should be borne in mind that the description of interactions and conformational changes in the HslUV complex at the intermediate resolution of this

HslUV Structure 637

Figure 3. Solution Small Angle Scattering Results on HslUV Columns from left to right: (left) worm diagram of crystallographic models used to calculate scattering curves and P(r) function, with HslV yellow, HslU N and C domains green, and HslU I domains magenta; (middle) plot of the experimental and calculated P(r) functions, with experimental curves, dotted line; calculated, solid line; (right) plot of experimental scattering points with error bars and calculated curves (solid lines). (A) H. influenzae HslUV crystal structure as presented in this paper. (B) H. influenzae HslUV crystal structure but with I domains extended, as described in text. (C) E. coli HslUV crystal structure (PDB ID code 1DOO; Bochtler et al., 2000).

structure relies on the fact that, with the exception of the carboxy-terminal 14 residues of the HslU protomers, the structure was built by making adjustments to wellrefined models of the HslV and HslU protomers, which yields a substantially more accurate model than de novo model building into a 3.4 A˚ map.) It became apparent in the early stages of model rebuilding that the carboxy-terminal 14 residues of the HslU model used in molecular replacement did not agree with the electron density maps, and that the penultimate helix of HslU emerges into an extended segment of polypeptide that terminates in a short helix sequestered

between subunits of HslV, as shown in Figures 5A and 5B. The extended linker appears to be flexible; its continuity is clear, but it differs from one subunit to the next, and in many subunits, it appears to have multiple conformations. The carboxy-terminal Leu-444 residue of HslU is clearly distended from the helix (data shown in Figure 1B); the terminal carboxyl group forms a salt bridge with Lys-28 of one HslV protomer, while the guanidinium group of Arg-35 of the same HslV protomer hydrogen bonds to the peptide backbone. A hydrophobic surface of the short terminal helix of HslU binds a hydrophobic pocket of a neighboring HslV protomer. Notably, within

Cell 638

Figure 4. Peptidase Activity of HslUV Crystals Plot of measured fluorescence versus time for a suspension of HslUV crystals (solid line), compared to supernatant (dashed line), assayed as described in Experimental Procedures.

the HslU and HslV sequences that are currently available, Lys-28 and Arg-35 of HslV are conserved, and both the specific length and the hydrophobic nature of the terminal short helix of HslU are also conserved. It also became apparent that the apical helices of the HslV model used in molecular replacement did not agree with the maps, and that residues 47–92 needed to be shifted substantially (Figure 5A). These helices bind in the cleft between the N and C domains of HslU, in close proximity to the ATP binding site; the interface is shown in more detail in Figure 5C. The side chains of Gln-68 and His-70 of HslV (which are well-ordered in electron density maps) hydrogen bond with the polypeptide backbone of the residue segment 386–390 of the C domain of HslU. The peptide segment, residues 391–393, is in contact with the ATP binding loop of the N domain (the “P loop”, or Walker “A” motif; Walker et al., 1982), and Arg-394 is in a position where its side chain could interact with ADP or ATP. The equivalent arginine of E. coli HslU binds the terminal phosphate of AMPPNP (Bochtler et al., 2000). Sequence alignments show that an arginine is found at this position in the majority of AAA proteins, suggesting it may have a role in coupling ATP binding or hydrolysis in the N domain to a conformational change in the C domain. Again referencing Figure 5C, it can be seen that the left side of the apical helices interface to a protruding loop from the N domain of HslU. This loop, of which residues Gln-312 and Val-313 are in contact with HslV, lies between the ATP binding P loop of HslU and the second Walker motif, the “DExx” motif, the aspartic acid (Asp-257 of HslU) of which binds Mg2⫹ of a Mg2⫹ATP complex and the glutamic acid of which is a catalytic residue for ATP hydrolysis. The segment of peptide spanning residues 301–313 in HslU, sandwiched between the Walker A and B motifs (colored yellow in Figure 5C), is more generally referred to as the “sensor I” motif of AAA proteins (Koonin, 1993). Many AAA proteins have a threonine or asparagine at a position corresponding to residue 310 of HslU in this motif; it has been suggested that the side chain of this residue may interact directly with the terminal phosphate of ATP, thereby “sensing” the hydrolysis state of the nucleotide (Neuwald et al., 1998). In HslU, an alanine is found at this

position (while in E. coli ClpA, threonine is found in one ATP binding domain and asparagine in the other), presenting an ambiguity regarding the role of this motif in sensing and modulating ATP hydrolysis in this case. However, it is provocative to consider that the “clustering” and the interdomain interactions of the P loop and sensor I motif of the N domain of HslU, a segment of polypeptide in the C domain, and the apical helices of HslV, may be mandated by an imperative for coupling ATP binding and hydrolysis to assembly of the complex and activation of the protease; determining whether or not this is the case will require further study. The shift in the apical helices of HslV propagates to the active site region. Specifically, the segment of polypeptide that includes residues 46–50 shift ⵑ3 A˚ from their position in the protomer of HslV alone (Figure 6). This segment includes a glycine residue, G48, which is conserved in known HslV sequences, as well as in ␤-type subunits of proteasomes (Seemu¨ller et al., 1995). It is probable that such a large shift in the active site cleft is correlated with the mechanism of HslU-dependent activation of proteolytic activity. Relevant to this suggestion is the observation that a class of vinyl sulfone inhibitors which covalently modify HslV will only do so in the presence of HslU and ATP, consistent with such an allosteric mechanism of HslV activation (Bogyo et al., 1997). The consensus hydrolytic mechanism for proteasome homologs invokes nucleophilic attack by the hydroxyl of Thr-1 on the carbonyl of the scissile peptide bond of the substrate, with simultaneous abstraction of the hydroxyl proton, possibly by the terminal amino group or (directly or indirectly) by Lys-33 (Seemu¨ller et al., 1995). For this mechanism to work efficiently, the carbonyl of the substrate needs to be positioned in line with the hydroxyl of Thr-1 (directly above it in the orientation of Figure 6). Structural information on inhibitors bound to proteasomal subunits (Lo¨we et al., 1995; Bochtler et al., 1997; Groll et al., 1997) suggests that the peptide backbone of residues 47–48 may act as a “guide” segment for positioning the scissile bond for nucleophilic attack by the threonine hydroxyl; if this is the case, then an ⵑ3 A˚ shift of the guide segment would be expected to dramatically affect hydrolytic activity. Understanding precisely how this is accomplished will require further structural studies with inhibitors bound to HslV. Discussion The chromatographic sizing column data, enzymatic assay data, solution SAXS data, and crystallographic data provide a self-consistent set of results in support of the HslUV structure presented here being that of an active complex. Consistent with the requirement that (and suggestive of a mechanism whereby) the proteolytic activity is stimulated by HslU, assembly of the HslUV complex induces a shift in the apical helices of HslV which propagates to the active site cleft of the protease. Interactions between the apical helices of HslV and the cleft between domains of HslU do not, by themselves, reveal the obligatory mechanism of ATP-dependent assembly and disassembly of the complex. How-

HslUV Structure 639

ever, invoking “guilt by association”, the interactions at this interface suggest participation of a segment of polypeptide in the C domain of HslU, the P loop and “sensor I loop” motifs of the N domain of HslU, and the apical helices of HslV in the mechanism of coupling ATP binding and hydrolysis with assembly and activation of the HslUV complex. The orientation of the HslU hexamers in the complex suggests a partitioning of activities within the domains of HslU: the N and C domains may be responsible for assembly of the complex and for activation of the HslV protease; the I domains (which sequence alignments identify as unique to the HslU protein family; see Pfam family #4; Bateman et al., 2000) may be responsible for selecting polypeptide substrates for degradation and for the initial steps of channeling them to the interior cavity of the complex. How this latter process is accomplished remains a mystery. The interior cavity of the HslUV complex forms a steep “funnel” with a diameter of ⵑ20 A˚ at the narrowest constriction, which is at the interface of the HslU and HslV rings; in the absence of dramatic “swelling” of the complex when polypeptide substrates are bound, a substrate would need to be denatured and “threaded” through the channel to the catalytic site of the protease. This would be congruent with the mechanism that has been proposed for the related ClpAP and ClpXP chaperones (Weber-Ban et al., 1999; Hoskins et al., 2000; Kim et al., 2000; Singh et al., 2000). Comparison of the peptidase activity of the E. coli HslUV on small substrates with the “polypeptidase” activity on proteins has revealed differences in the dependence of the two activities on ATP (Rohrwild et al., 1996; Huang and Goldberg, 1997; Seol et al., 1997; Yoo et al., 1997). Specifically, the Kact(ATP) (i.e., the ATP concentration for half-maximal activation of activity) for peptidase activitity is ⵑ3–4 ␮M, while the Kact(ATP) for polypeptidase activity is approximately two orders of magnitude higher, ⵑ200–300 ␮M. Nonhydrolyzable ATP analogs can substitute for ATP in the peptidase activity, albeit at a higher concentration (e.g., Kact(AMPPNP) ⵑ300 ␮M) and with a more complex time course and dependence on monovalent ions. ATP hydrolysis facilitates the polypeptidase activity, although it is apparently not an absolute requirement; AMPPNP can also induce polypeptidase activity in the presence of 50 mM KCl (Huang and Goldberg, 1997; Yoo et al., 1997). The peptidase activity on small substrates and the polypeptidase activity on proteins differ in that the first requires only activation of the HslV protease, while the second requires, in addition, denaturation of the polypeptide substrate and channeling of it to the interior cavity of HslV. The first activity, i.e., activation of the HslV protease by HslU with nucleoside triphosphate bound, can arguably be accounted for by the HslUV structure reported here. The polypeptide denaturation activity presumably requires HslUV to cycle through at least one additional conformational state; the second state is apparently readily accessed by hydrolysis of ATP (Yoo et al., 1996; Huang and Goldberg, 1997). The minimal scheme this suggests for HslUV activity is a two-state model, in which an “ATP bound” state (which can also be induced by nonhydrolyzable ATP analogs) activates peptidase activity, and a second state (which

may occur with ADP bound) additionally facilitates denaturation of a polypeptide substrate. The nature of the second state can only be a subject of speculation at this time; in the context of the structure presented here, we note that subtle structural rearrangements around the nucleotide binding site of HslU have the potential to translate into significant “flexing” of the N and C domains about their hinge point, thereby altering the interactions at the HslU-HslV interface, as well as into substantial conformational changes at the extremities of the I domains, in view of their long “lever arm” relative to the hinge point. In a hypothetical scenario, if the I domains had a substrate polypeptide bound, a large conformational rearrangement of their relative orientation could serve to facilitate denaturation of the polypeptide. Whatever the mechanism of cycling through alternative conformational states of HslU may be, many aspects of it are likely to generalize to other Clp/Hsp100 proteins, given the relatively high level of conservation of sequence and structure within the family. In this context, HslUV can be viewed as a prototype system for elucidating the molecular chaperone mechanism of a representative of the Clp/Hsp100 protein family. Experimental Procedures Cloning, Expression, Purification, and Analysis of Proteins An expression vector for the hslVU operon was constructed from a parent pRSET-B plasmid (Invitrogen) and plasmids GHIJE71 and GHIBP32 (American Type Culture Collection) from the Haemophilus influenzae genome sequencing project (Fleischmann et al., 1995). Briefly, (1) the BamHI–EcoRI fragment of GHIJE71, encompassing the complete hslV coding sequence and a 5⬘ fragment of the hslU coding sequence, was ligated into a BamHI, EcoRI-digested pRSET vector; (2) the resulting vector was digested with ClaI and KpnI, and the ClaI–KpnI fragment of GHIBP32, which includes the 3⬘ remainder of the hslU coding sequence through the “stop” codon, was ligated into it; (3) the resulting vector was then digested with NdeI and BsiWI and a short synthetic duplex oligonucleotide was ligated into it to align the NdeI site of the pRSET vector with the “start” codon of the hslVU operon. Protein was expressed from this plasmid in E. coli BL21(DE3) by induction of T7 polymerase expression with IPTG (Studier et al., 1990). The HslUV complex was purified by gravity-flow chromatography on Q-sepharose (condition: 20 mM phosphate buffer, pH 7.0, gradient 0 → 1 M KCl), on which HslU and HslV coeluted, followed by FPLC gel filtration on Superdex 200 (Pharmacia; conditions: 20 mM phosphate buffer, pH 7.0, 1 M KCl), on which the HslUV complex (MW ⵑ820 kDa) separated from excess HslV (MW ⵑ230 kDa) and contaminant proteins. In some cases, HslUV was purified further by recrystallization using the conditions described below. Expression and purification of native and SeMet-labeled HslU protein is described elsewhere. The identities of the purified recombinant Haemophilus HslU and HslV proteins were confirmed by aminoterminal amino acid sequencing. Size exclusion chromatography was performed at room temperature on a Superose 6 column (HR 10/30; Pharmacia) equilibrated with 150 mM KCl, 5 mM MgCl2, 25 mM MOPS, pH 7.0. In a typical experiment, 50 ␮l of HslUV protein at a concentration of ⵑ2 mg/ml was loaded on the column and eluted at a flow rate of 0.4 ml/min.

Assay of Peptide Hydrolysis Activity Assays of peptide hydrolysis activity in solution were carried out at 25⬚C in mixtures containing 4 ␮g/ml HslUV, 0.1 mM Z-Gly-Gly-LeuAMC (Bachem), 150 mM KCl, 5 mM MgCl2, 25 mM Tris-HCl, pH 7.8, and 1 mM ATP in a total volume of 3 ml. The release of AMC produced by peptide hydrolysis was followed by continuously moni-

Cell 640

Figure 5. Conformational Changes that Occur when the HslUV Complex Assembles (A) Stereo view of protomers from uncomplexed HslU and HslV superimposed on one pair of HslU–HslV protomers from the HslUV complex (chain ID C and I in PDB file). For HslU, rmsd ⫽ 0.76 A˚ for superposition of residues 2–86, 248–265, and 270–335; for HslV, rmsd ⫽ 0.40 A˚ for superposition of residues 1–45 and 112–172. The HslU N domain is on the upper left; the C domain is on the upper right; the I domain extends beyond the top of the drawing. The ATP molecule as seen in the HslUV crystals is shown as a ball and stick representation. Where the structures have minimal difference, the HslU protomer is magenta; the HslV protomer is green. Where the structures in the complex differ substantially from the uncomplexed structures, protomers from uncomplexed HslU and HslV are colored red and yellow, respectively, while

HslUV Structure 641

Figure 6. Conformational Changes around the Catalytic Site of HslV Stereo ribbon drawing of the active site region. The HslUV structure is colored green. The segment of uncomplexed HslV that differs substantially from the complex (see Figure 3A) is colored magenta. Selected residue side chains and polypeptide backbone are shown in the ball and stick representation.

toring the fluorescence with an excitation wavelength of 380 nm and an emission wavelength of 430 nm. Assays of peptide hydrolysis activity in crystals were carried out on suspensions of microcrystals (ⵑ1–10 microns in linear dimension) produced by bulk recrystallization as a final step of purification of HslUV. Crystals were washed five times in a mother liquor consisting of 6% polyethylene glycol monomethyl ether of average molecular weight 2000 (PEG-MME2K), 1 M KCl, 10 mM Mg(OAc)2, 50 mM citrate, pH 6.0, then incubated for 30 min at room temperature and separated from supernatant. Crystals and supernatant were assayed as described above. Solution Small Angle X-Ray Scattering (SAXS) Small angle X-ray scattering data were measured on beamline 4–2 of the Stanford Synchrotron Radiation Laboratory (SSRL) with a linear detector (one-dimensional position sensitive proportional counter; BioLogic-EMBL, Grenoble, France) as described previously (Wilbanks et al., 1995). The X-ray wavelength was 1.38 A˚. Samples were measured in a solution containing 25 mM MOPS pH 7.0, 150 mM KCl, 5 mM MgCl2, 5mM DTT. Where indicated, ATP was added to a final concentration of 1 mM. Protein concentrations ranged 1–2 mg/ml. Beam exposure was 20 min per sample; data were collected and analyzed in 1 min frames; no deterioration of the scattering curve that would indicate radiation damage was seen over 20 successive frames collected during the 20 min period. X-ray scattering data were processed with the program SAPOKO (programs of D. I. Svergun and M. Koch; see http://www.emblhamburg.de/ExternalInfo/Research/Sax/index.html). Electron pair distance distribution functions (P(r) functions) and radii of gyration (Rg) were computed from the experimental scattering curves using the program GNOM (Semenyuk and Svergun, 1991; Svergun, 1991). Theoretical scattering curves and radii of gyration were computed from the coordinates of crystallographic structures using the program CRYSOL, which represents a protein molecule with a series of spherical harmonics and includes a single hydration layer of density

0.334 e⫺/A˚3 on the surface of the molecule, and also CRYDAM, which uses a number of uniform-density spheres to represent a protein shape (Svergun et al., 1995). Theoretical P(r) functions were computed from crystallographic models, using only the ␣ carbon coordinates in order to reduce computation time. Crystallographic Structure Determination HslUV protein was concentrated to 5–20 mg/ml in 10 mM MOPS, pH 7.0, 1 mM Mg(OAc)2, 1 mM ATP, and crystals were grown at 4⬚C in hanging or sitting drops. Precipitant was optimized to: 3%–6% PEG-MME2K, 1 M KCl, 10 mM Mg(OAc)2, 50 mM citrate, pH 6.0. Crystals with SeMet-labeled HslU were grown by mixing labeled HslU with unlabeled HslV and proceeding as described above. Crystallographic data were collected on synchrotron beamlines 9–1 of SSRL and 5.0.2 of the Lawrence Berkeley Laboratory Advanced Light Source (ALS). For final data collection, crystals were first adapted stepwise over a period of hours to a stabilization solution: 20% PEG-MME2K, 1 M KCl, 50 mM citrate, 10 mM Mg(OAc)2, 1 mM ATP, pH 6.0. Then, the 1 M KCl of the stabilization solution was replaced stepwise with 1 M Mg(OAc)2, which acted as a cryoprotectant, allowing the crystals to be flash-frozen in a stream of nitrogen gas at ⵑ100 K. Crystals are orthorhombic, space group P21212, a ⫽ 209.22 A˚, b ⫽ 220.58 A˚, c ⫽ 241.07 A˚ after freezing, with one U12V12 complex per asymmetric unit. Diffraction was anisotropic, with the weakest diffraction along the c* reciprocal space axis. Data were collected on an ADSC CCD detector and processed with DENZO; for native HslUV crystals, ␭ ⫽ 1.10000 A˚; for HslUV crystals incorporating SeMet-HslU, ␭ ⫽ 0.97992 A˚, the absorption peak of Se in the crystals. Native data collection statistics are summarized in Table 1. The HslUV structure was solved by molecular replacement with structure of uncomplexed H. influenzae HslV and HslU. The H. influenzae HslV structure has been solved by molecular replacement with the 3.8 A˚ E. coli HslV structure (Bochtler et al., 1997) and refined with data to 1.9 A˚ resolution; current refinement statistics are:

protomers from the HslUV complex are colored blue. For clarity, small differences arising from the flexing of HslU N and C domains are not highlighted. (B) Stereo ribbon diagram of the interaction between the carboxy-terminal residues of HslU (magenta; chain ID A) and two subunits of HslV (green and yellow; chain ID G and H). Selected side chains shown as stick models with carbon atoms colored the same as the subunit. Hydrogen bonds and salt bridges represented as red dotted lines. (C) Stereo ribbon diagram of interaction between apical helices of HslV (green; chain id A) and the cleft of HslU (purple; chain ID G). The orientation is approximately the same as Figure 5A. The Walker A motif, Walker B motif, and sensor I motif of HslU are colored blue, red and yellow, respectively.

Cell 642

Rcryst ⫽ 0.200, Rfree ⫽ 0.230, rmsd of bond lengths ⫽ 0.010 A˚, rmsd of bond angles ⫽ 1.60⬚ (M. C. S. and D. B. M., unpublished data). The H. influenzae HslU structure has been solved by molecular replacement with the 3.0 A˚ E. coli HslU structure (Bochtler et al., 2000) and refined to 2.3 A˚ resolution. The HslU crystals have an unusual twinning, or “one dimensional disorder”, which will be described elsewhere (C. B. T. and D. B. M., unpublished data); in addition, the I domains of the HslU protomer, constituting ⵑ27% of the molecule, are disordered and cannot be traced; both of these factors have affected the refinement statistics: Rcryst ⫽ 0.266, Rfree ⫽ 0.288, rmsd of bond lengths ⫽ 0.006 A˚, rmsd of bond angles ⫽ 1.14⬚. The Haemophilus HslV and HslU protomers show no major differences in tertiary structure from the published E. coli structures; we have used them rather than the E. coli structures for molecular replacement since (1) the protein source and amino acid sequences are identical to those of the HslUV complex, and (2) the H. influenzae structures provide more precise initial models since they are refined to higher resolution. A self-rotation function computed with native data was consistent with 622 noncrystallographic symmetry (NCS) for the U12V12 complex, with the 6-fold axis nearly parallel to the crystallographic c axis. Systematic molecular replacement searches with the program AMORE (Navaza, 1994), using a HslU hexamer as a search model, yielded two sets of solutions, each with the 6-fold axis of the hexamer tilted ⵑ10⬚ from the crystallographic c axis. The two solutions were related by a 2-fold axis perpendicular to the 6-fold, i.e., one hexamer was nearly parallel to the c axis and the other was antiparallel. Translation searches positioned two HslU hexamers in the unit cell. Rotation and translation searches with a HslV dodecamer yielded a solution with its 6-fold axis tilted 11⬚ from the c axis and with the dodecamer sandwiched between the two HslU hexamers. This starting model for the U12V12 complex was used in subsequent refinement with the program package CNS (Brunger et al., 1998) and rebuilding with the molecular modeling package O (Jones, 1978). Rigid body refinement of all protomers in the U12V12 complex resulted in Rcryst ⫽ 0.497, Rfree ⫽ 0.489. This model was used to generate 12 NCS operators by taking a HslU–HslV pair of protomers as the NCS asymmetric unit of the molecular complex. Using data to 3.4 A˚ and imposing strict 12-fold NCS constraints, a cycle of torsional simulated annealing followed by positional refinement and B factor minimization reduced the R factors to Rcryst ⫽ 0.415, Rfree ⫽ 0.427. A Foexp(␣calc) map was calculated with model phases and averaged using the RAVE package (Kleywegt and Read, 1997), imposing the 6-fold NCS of the complex but not requiring 2-fold symmetry perpendicular to it. The correlation coefficient of the averaged map was 0.90. The phases derived from the averaged map were used to compute an anomalous difference Fourier map with data to 4 A˚ resolution, which was ⵑ75% complete in Bijvoet pairs, collected on crystals having SeMet-labeled HslU; several positive peaks in the resulting map coincided or were in close proximity to methionines in the model, lending credibility to the molecular replacement solution. The averaged map indicated that much of the model required only minor adjustments; however, some parts of the model deviated substantially from the map and required rebuilding. Several rounds of refinement in which strict 6-fold NCS restraints were imposed on the model, followed by 6-fold averaging of the map and model rebuilding, were done until no improvement of the free R factor was observed (Rcryst ⬇ 0.35, Rfree ⬇ 0.36). At this point, the strict NCS constraints were replaced with tight (300 kcal mol⫺1 A˚⫺2) NCS restraints, and operators which related the protomers within each hexameric ring of HslU and HslV were computed; protomers were restrained toward being identical within a ring but not between different rings. Torsional simulated annealing followed by positional and individual B factor refinement reduced the R factors to Rcryst ⫽ 0.300, Rfree ⫽ 0.318. Iterative cycles of rebuilding and refinement identified additional features, including unambiguous density for an ATP molecule in each HslU protomer. In the upper HslU protomers, it is clear that ATP binds with the base in the anti conformation (Figure 1A). In the lower HslU protomers, the nucleotide electron density was substantially noisier; it clearly encompassed an ATP molecule, but the conformation of the base was less definitive; since we see no significant differences in the nucleotide binding environment between the top and bottom halves, the base of the ATP was built in the anti conformation for consistency with the upper subunits.

At the point Rcryst ⫽ 0.285, Rfree ⫽ 0.302, the NCS restraints were relaxed to 50 kcal mol⫺1 A˚⫺2 for the upper half of the complex (as defined in discussion of Results; protomer chains A–F and G–L for HslU and HslV, respectively) and 200 kcal mol⫺1 A˚⫺2 for the lower half, (chains M–R and S–X for HslV and HslU, respectively), and cycles of positional and group B factor refinement followed by rebuilding were performed until the Rfree showed no further improvement. Final model statistics are summarized in Table 1. A final check of the model was made with the 4.0 A˚ data from the HslUV crystals grown with SeMet-labeled HslU. The model was subjected to rigid body and positional refinement against the SeMet data—necessitated because of the slightly different unit cell of the SeMet-labeled crystals—and an anomalous difference Fourier map was computed. Of the 54 methionines modeled in the upper HslU hexamer, the sulfur atoms of 45 of them overlapped with peaks of height 3␴ or greater in the map, and six more were within ⵑ1 A˚ of a peak 2.5␴ or higher, confirming that the model is correct. Not surprisingly, fewer peaks were apparent in the lower HslU hexamer, due to its greater static disorder in the crystal; six methionines coincided with peaks 3␴ or higher, and ten more were near peaks 2.5␴ or higher. Supplemental Data Additional data showing simulated annealing omit maps of portions of the HslUV model that are disscussed in the text and shown in Figures 5 and 6, as well as refinement statistics for the HslV model used in molecular replacement, are available online (http://www.cell. com/cgi/content/full/103/4/633/DC1). Acknowledgments We thank Lushen Li for excellent technical assistance in biochemistry, members of the SSRL and ALS staff for excellent beamline assistance, and Dimitri Svergun for helpful discussion on solution scattering analysis and the use of his programs Gnom, Crysol, and Crydam. This work was supported by award GM-39928 from the National Institutes of Health (NIH) to D. B. M.; V. R. is supported by the MMTSB research resource RR12255 of NIH to TSRI. This work is based upon research conducted at the Stanford Synchrotron Radiation Laboratory (SSRL), which is funded by the Department of Energy (BES, BER) and the National Institutes of Health (NCRR, NIGMS), and at the Advanced Light Source (ALS), which is funded by the Department of Energy. Received September 5, 2000; revised October 4, 2000. References Bateman, A., Birney, E., Durbin, R., Eddy, S.R., Howe, K.L., and Sonnhammer, E.L. (2000). The Pfam protein families database. Nucleic Acids Res. 28, 263–266. Bochtler, M., Ditzel, L., Groll, M., and Huber, R. (1997). Crystal structure of heat shock locus V (HslV) from Escherichia coli. Proc. Natl. Acad. Sci. USA 94, 6070–6074. Bochtler, M., Hartmann, C., Song, H.K., Bourenkov, G.P., Bartunik, H.D., and Huber, R. (2000). The structures of HsIU and the ATPdependent protease HsIU-HsIV. Nature 403, 800–805. Bogyo, M., McMaster, J.S., Gaczynska, M., Tortorella, D., Goldberg, A.L., and Ploegh, H. (1997). Covalent modification of the active site threonine of proteasomal beta subunits and the Escherichia coli homolog HslV by a new class of inhibitors. Proc. Natl. Acad. Sci. USA 94, 6629–6634. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: A new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. Chuang, S.E., Burland, V., Plunkett, G. 3d, Daniels, D.L., and Blattner, F.R. (1993). Sequence analysis of four new heat-shock genes constituting the hslTS/ibpAB and hslVU operons in Escherichia coli. Gene 134, 1–6. Esnouf, R.M. (1997). An extensively modified version of MolScript

HslUV Structure 643

that includes greatly enhanced coloring capabilities. J. Mol. Graph. Model. 15, 132–134. Esnouf, R.M. (1999). Further additions to MolScript version 1.4, including reading and contouring of electron-density maps. Acta Crystallogr. D Biol. Crystallogr. 55, 938–940.

Rohrwild, M., Coux, O., Huang, H.C., Moerschell, R.P., Yoo, S.J., Seol, J.H., Chung, C.H., and Goldberg, A.L. (1996). HslV-HslU: A novel ATP-dependent protease complex in Escherichia coli related to the eukaryotic proteasome. Proc. Natl. Acad. Sci. USA 93, 5808– 5813.

Fleischmann, R.D., Adams, M.D., White, O., Clayton, R.A., Kirkness, E.F., Kerlavage, A.R., Bult, C.J., Tomb, J.F., Dougherty, B.A., Merrick, J.M., et al. (1995). Whole-genome random sequencing and assembly of Haemophilus influenzae Rd. Science 269, 496–512.

Rohrwild, M., Pfeifer, G., Santarius, U., Muller, S.A., Huang, H.C., Engel, A., Baumeister, W., and Goldberg, A.L. (1997). The ATPdependent HslVU protease from Escherichia coli is a four-ring structure resembling the proteasome. Nat. Struct. Biol. 4, 133–139.

Gottesman, S., Roche, E., Zhou, Y., and Sauer, R.T. (1998). The ClpXP and ClpAP proteases degrade proteins with carboxy-terminal peptide tails added by the SsrA-tagging system. Genes Dev. 12, 1338–1347.

Schirmer, E.C., Glover, J.R., Singer, M.A., and Lindquist, S. (1996). HSP100/Clp proteins: a common mechanism explains diverse functions. Trends Biochem. Sci. 21, 289–295.

Groll, M., Ditzel, L., Lo¨we, J., Stock, D., Bochtler, M., Bartunik, H.D., and Huber, R. (1997). Structure of 20S proteasome from yeast at 2.4 A resolution. Nature 386, 463–471. Hoskins, J.R., Singh, S.K., Maurizi, M.R., and Wickner, S. (2000). Protein binding and unfolding by the chaperone ClpA and degradation by the protease ClpAP. Proc. Natl. Acad. Sci. USA 97, 8892– 8897. Huang, H., and Goldberg, A.L. (1997). Proteolytic activity of the ATPdependent protease HslVU can be uncoupled from ATP hydrolysis. J. Biol. Chem. 272, 21364–21372. Jones, A. (1978). A Graphics Model Building and Refinement System for Macromolecules. J. Appl. Crystallogr. 11, 268–272.

Seemu¨ller, E., Lupas, A., Stock, D., Lo¨we, J., Huber, R., and Baumeister, W. (1995). Proteasome from Thermoplasma acidophilum: a threonine protease. Science 268, 579–582. Semenyuk, A.V., and Svergun, D.I. (1991). GNOM—a program package for small-angle scattering data processing. J. Appl. Crystallogr. 24, 537–540. Seol, J.H., Yoo, S.J., Shin, D.H., Shim, Y.K., Kang, M.S., Goldberg, A.L., and Chung, C.H. (1997). The heat-shock protein HslVU from Escherichia coli is a protein-activated ATPase as well as an ATPdependent proteinase. Eur. J. Biochem. 247, 1143–1150. Seong, I.S., Oh, J.Y., Yoo, S.J., Seol, J.H., and Chung, C.H. (1999). ATP-dependent degradation of SulA, a cell division inhibitor, by the HslVU protease in Escherichia coli. FEBS Lett. 456, 211–214.

Kanemori, M., Yanagi, H., and Yura, T. (1999). The ATP-dependent HslVU/ClpQY protease participates in turnover of cell division inhibitor SulA in Escherichia coli. J. Bacteriol. 181, 3674–3680.

Singh, S.K., Grimaud, R., Hoskins, J.R., Wickner, S., and Maurizi, M.R. (2000). Unfolding and internalization of proteins by the ATPdependent proteases ClpXP and ClpAP. Proc. Natl. Acad. Sci. USA 97, 8898–8903.

Keiler, K.C., Waller, P.R., and Sauer, R.T. (1996). Role of a peptide tagging system in degradation of proteins synthesized from damaged messenger RNA. Science 271, 990–993.

Studier, F.W., Rosenberg, A.H., Dunn, J.J., and Dubendorff, J.W. (1990). Use of T7 RNA polymerase to direct expression of cloned genes. Methods Enzymol. 185, 60–89.

Kessel, M., Wu, W., Gottesman, S., Kocsis, E., Steven, A.C., and Maurizi, M.R. (1996). Six-fold rotational symmetry of ClpQ, the E. coli homolog of the 20S proteasome, and its ATP-dependent activator, ClpY. FEBS Lett. 398, 274–278.

Svergun, D.I. (1991). Mathematical methods in small-angle scattering data analysis. J. Appl. Crystallogr. 24, 485–492.

Kim, Y.I., Burton, R.E., Burton, B.M., Sauer, R.T., and Baker, T.A. (2000). Dynamics of substrate denaturation and translocation by the ClpXP degradation machine. Mol. Cell 5, 639–648. Kleywegt, G.J., and Read, R.J. (1997). Not your average density. Structure 5, 1557–1569. Koonin, E.V. (1993). A common set of conserved motifs in a vast variety of putative nucleic acid-dependent ATPases including MCM proteins involved in the initiation of eukaryotic DNA replication. Nucleic Acids Res. 21, 2541–2547. Kraulis, P. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946–950. Lenzen, C.U., Steinmann, D., Whiteheart, S.W., and Weis, W.I. (1998). Crystal structure of the hexamerization domain of N-ethylmaleimidesensitive fusion protein. Cell 94, 525–536. Levchenko, I., Luo, L., and Baker, T.A. (1995). Disassembly of the Mu transposase tetramer by the ClpX chaperone. Genes Dev. 9, 2399–2408. Lo¨we, J., Stock, D., Jap, B., Zwickl, P., Baumeister, W., and Huber, R. (1995). Crystal structure of the 20S proteasome from the archaeon T. acidophilum at 3.4 A resolution. Science 268, 533–539. Merritt, E.A., and Bacon, D.J. (1997). Raster3D. Photorealistic molecular graphics. Methods Enzymol. 277, 505–524. Missiakas, D., Schwager, F., Betton, J.M., Georgopoulos, C., and Raina, S. (1996). Identification and characterization of HsIV HsIU (ClpQ ClpY) proteins involved in overall proteolysis of misfolded proteins in Escherichia coli. EMBO J. 15, 6899–6909. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallogr. A50, 157–163. Neuwald, A.F., Aravind, L., Spouge, J.L., and Koonin, E.V. (1998). AAA⫹: A class of chaperone-like ATPases associated with the assembly, operation, and disassembly of protein complexes. Genome Res. 27, 27–43.

Svergun, D., Barberato, C., and Koch, M.H.J. (1995). Crysol: a program to evaluate X-ray solution scattering of biological macromolecules from atomic coordinates. J. Appl. Crystallogr. 28, 768–773. Walker, J.E., Saraste, M., Runswick, M.J., and Gay, N.J. (1982). Distantly related sequences in the alpha- and beta-subunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold. EMBO J. 1, 945–951. Weber-Ban, E.U., Reid, B.G., Miranker, A.D., and Horwich, A.L. (1999). Global unfolding of a substrate protein by the Hsp100 chaperone ClpA. Nature 401, 90–93. Wilbanks, S.M., Chen, L., Tsuruta, H., Hodgson, K.O., and McKay, D.B. (1995). Solution small-angle X-ray scattering study of the molecular chaperone Hsc70 and its subfragments. Biochemistry 34, 12095–12106. Yoo, S.J., Seol, J.H., Shin, D.H., Rohrwild, M., Kang, M.S., Tanaka, K., Goldberg, A.L., and Chung, C.H. (1996). Purification and characterization of the heat shock proteins HslV and HslU that form a new ATP-dependent protease in Escherichia coli. J. Biol. Chem. 271, 14035–14040. Yoo, S.J., Seol, J.H., Seong, I.S., Kang, M.S., and Chung, C.H. (1997). ATP binding, but not its hydrolysis, is required for assembly and proteolytic activity of the HslVU protease in Escherichia coli. Biochem. Biophys. Res. Comm. 238, 581–585. Yu, R.C., Hanson, P.I., Jahn, R., and Bru¨nger, A.T. (1998). Structure of the ATP-dependent oligomerization domain of N-ethylmaleimide sensitive factor complexed with ATP. Nat. Struct. Biol. 5, 803–811. Protein Data Bank Accession Numbers Coordinates of the HslUV complex, as well as the HslV and HslU models used for molecular replacement, have been deposited in the Protein Data Bank, accession id 1G3I, 1G3K, and 1G41, respectively.