research papers
Elements in nucleotide sensing and hydrolysis of the AAA+ disaggregation machine ClpB: a structure-based mechanistic dissection of a molecular motor
aDepartment of Biomolecular Mechanisms, Max Planck Institute for Medical Research, Jahnstrasse 29, 69120 Heidelberg, Germany
*Correspondence e-mail: cathleen.zeymer@mpimf-heidelberg.mpg.de, jochen.reinstein@mpimf-heidelberg.mpg.de
ATPases of the AAA+ superfamily are large oligomeric molecular machines that remodel their substrates by converting the energy from ATP hydrolysis into mechanical force. This study focuses on the molecular chaperone ClpB, the bacterial homologue of Hsp104, which reactivates aggregated proteins under cellular stress conditions. Based on high-resolution crystal structures in different nucleotide states, mutational analysis and nucleotide-binding kinetics experiments, the ATPase cycle of the C-terminal nucleotide-binding domain (NBD2), one of the motor subunits of this AAA+ disaggregation machine, is dissected mechanistically. The results provide insights into nucleotide sensing, explaining how the conserved sensor 2 motif contributes to the discrimination between ADP and ATP binding. Furthermore, the role of a conserved active-site arginine (Arg621), which controls binding of the essential Mg2+ ion, is described. Finally, a hypothesis is presented as to how the ATPase activity is regulated by a conformational switch that involves the essential Walker A lysine. In the proposed model, an unusual side-chain conformation of this highly conserved residue stabilizes a catalytically inactive state, thereby avoiding unnecessary ATP hydrolysis.
Keywords: ClpB; AAA+ protein; molecular chaperones; molecular motors; transient kinetics; enzyme mechanisms; nucleotide sensing.
3D view: 4lj4,4lj5,4lj6,4lj7,4lj8,4lj9,4lja
1. Introduction
The molecular chaperone ClpB, the bacterial homologue of Hsp104, is a disaggregation machine that provides thermotolerance by resolubilizing and reactivating aggregated proteins in concert with the DnaK (Hsp70) chaperone system (Sanchez & Lindquist, 1990; Doyle & Wickner, 2009). ClpB belongs to the family of AAA+ proteins (ATPases associated with various cellular activities) that comprises a wide range of molecular machines which utilize the energy from ATP hydrolysis to remodel their respective substrates (Tucker & Sallai, 2007; Hanson & Whiteheart, 2005). Like other members of the AAA+ superfamily, ClpB forms hexameric ring structures (Lee et al., 2007). The disaggregation process is highly complex and involves the translocation of polypeptide chains through the central pore of the hexameric assembly (Weibezahn et al., 2004), which is achieved by highly coordinated action of the different ATPase sites within the complex (Werbeck et al., 2008; Franzmann et al., 2011). Recent studies have revealed mechanistic details regarding the interplay between ClpB and its co-chaperone DnaK (Seyffer et al., 2012; Rosenzweig et al., 2013), which is believed to recruit ClpB to the aggregates (Winkler et al., 2012).
The nucleotide-binding domains (NBDs) of AAA+ proteins constitute the motor subunits of the molecular machine. They share highly conserved structural features (Wendler et al., 2012). The overall fold of an NBD consists of an α/β core domain, which possesses a central, five-stranded, parallel β-sheet surrounded by five α-helices, and a C-terminal helical bundle. A unifying feature that is shared by all AAA+ proteins, but is also more generally found amongst various families of ATPases and GTPases, is the phosphate-binding loop (P-loop) or Walker A motif characterized by the consensus sequence GXXXXGKT/S (Walker et al., 1982; Saraste et al., 1990). The P-loop residues form interactions with the phosphates of the bound nucleotide. In particular, the highly conserved Walker A lysine was found to be absolutely essential for nucleotide binding and catalysis. The second common feature in AAA+ proteins is the Walker B motif hhhhDE, where h is a hydrophobic amino acid (Walker et al., 1982). The highly conserved Walker B glutamate is believed to activate the attacking water molecule during ATP hydrolysis. AAA+ proteins are active as oligomers, with the catalytic site being located at the interface between two NBD subunits in the oligomer. Highly conserved arginine residues, termed arginine fingers, are located in this interface and are believed to contact the nucleotide of the neighbouring subunit (Ogura et al., 2004). Furthermore, conserved sensor motifs have been identified and proposed to be involved in nucleotide-state sensing, allosteric communication or discrimination between the ADP and ATP states, which is a prerequisite for a functioning ATPase motor (Guenther et al., 1997). The sensor 1 motif contains a conserved polar residue (N/T) that is essential for ATP hydrolysis. The sensor 2 motif (GAR) located on the C-terminal helical bundle contains a conserved arginine residue that can interact directly with the nucleotide. A model of the typical AAA+ catalytic site with the essential residues in their active conformation is depicted in Fig. 1.
The AAA+ disaggregation machines Hsp104 and ClpB possess two NBDs per monomer, thus forming double-ring structures with 12 potential ATPase sites in the active hexameric complex. In order to dissect interfering processes in the oligomeric context of the full-length protein, the N-terminal nucleotide-binding domain (NBD1) and the C-terminal nucleotide-binding domain (NBD2) of ClpB from Thermus thermophilus were studied as separately expressed constructs with regard to nucleotide binding, and ATPase activity (Beinker et al., 2005; Werbeck et al., 2009; Zeymer et al., 2013). On the basis of this detailed analysis, we are now able to tackle open mechanistic questions concerning nucleotide-state sensing and the conversion of chemical energy from ATP hydrolysis into mechanical motion. We determined high-resolution X-ray structures in different nucleotide states for one of the motor subunits, namely NBD2, of ClpB and combined the structural analysis with nucleotide-binding kinetics studies. Our results provide a better understanding of how the ATPase cycle of the molecular motor is regulated and modulated, which is essential in order to gain further insights into the functioning of the complete disaggregation machine.
2. Materials and methods
2.1. Construct design and mutagenesis
The design of the ClpB NBD2 construct, comprising amino acids 520–854 of ClpB from T. thermophilus, has been described previously (Beinker et al., 2005). The active-site mutations K601Q, R621Q, N709A and R806A were introduced by PCR according to the QuikChange protocol (Agilent Technologies, Santa Clara, California, USA). DNA sequencing was performed by Eurofins MGW Operon (Ebersberg, Germany).
2.2. Protein expression and purification
All ClpB NBD2 variants were expressed recombinantly in Escherichia coli BL21 (DE3) RIL cells and purified as described previously (Beinker et al., 2005; Werbeck et al., 2009). The proteins were stored in buffer A (50 mM Tris–HCl pH 7.5, 50 mM KCl, 5 mM MgCl2, 2 mM EDTA).
2.3. Crystallization and structure determination
Crystals of all ClpB NBD2 variants were grown in a hanging-drop vapour-diffusion setup at 20°C. The protein solution (∼8 mg ml−1 in buffer A) was either nucleotide-free or was supplemented with 2 mM ADP, 2 mM MANT-dADP or 2 mM AMPPCP and was mixed in a 1:1 ratio with the reservoir solutions. Crystals of wild-type ClpB NBD2 (space group P65) in the nucleotide-free state and in complex with ADP or with AMPPCP grew in 0.2 M sodium citrate, 0.1 M HEPES–NaOH pH 7.0, 20% 2-propanol. Adding 15%(v/v) PEG 400 afforded sufficient cryoprotection for flash-cooling in liquid nitrogen. The reservoir solution for the crystallization of ClpB NBD2 K601Q in complex with MANT-dADP (space group P212121) consisted of 0.1 M HEPES–NaOH pH 7.5, 12% PEG 6000, 50 mM MgCl2. The cryoprotectant solution was additionally supplemented with 25% glycerol and 2 mM MANT-dADP. Crystals of ClpB NBD2 R621Q in complex with ADP or AMPPCP (space group P65) grew in 0.1 M Tris–HCl pH 7.5, 15–20% 2-propanol, 10–100 mM MgCl2. The cryoprotectant solution was additionally supplemented with 15% PEG 400 and 2 mM of the respective nucleotide. Crystals were passed quickly through the cryoprotectant solutions and were flash-cooled in liquid nitrogen. Where indicated, crystals of ClpB NBD2 R621Q were soaked in cryoprotectant solution supplemented with 100 mM GdmCl for 60 min prior to cooling. Diffraction data collection was performed on beamline X10SA at the Swiss Light Source (Villigen, Switzerland) with the crystals maintained at 100 K. The XDS program package was used for data processing (Kabsch, 2010). Phasing of the ADP-bound wild-type structure was performed by using CNS with residues 520–854 of the full-length ClpB structure (PDB entry 1qvr ; Lee et al., 2003) as the search model. Patterson correlation (Brünger, 1990) was required owing to the different domain arrangement of the search model. Subsequent structures were solved using Phaser (McCoy et al., 2007). Model building and was performed in iterative cycles using Xfit (McRee, 1999) or Coot (Emsley & Cowtan, 2004) and REFMAC (Murshudov et al., 2011) including TLS (Winn et al., 2001), respectively. The model quality was validated with MolProbity (Chen et al., 2010). Structure illustrations were generated using PyMOL (Schrödinger). The coordinates and structure factors have been deposited in the Protein Data Bank (PDB entries 4lj4 , 4lj5 , 4lj6 , 4lj7 , 4lj8 , 4lj9 and 4lja ).
2.4. Stopped-flow experiments and global fitting analysis of nucleotide-binding data
Nucleotide-binding kinetics measurements of ClpB NBD2 variants were performed using a BioLogic SFM-400 stopped-flow instrument (BioLogic Science Instruments, Claix, France) in single mixing configuration at 25°C, essentially as described before for the wild-type protein (Werbeck et al., 2009). The fluorescently labelled MANT-dADP and MANT-dATP were purchased from BIOLOG (Bremen, Germany). The excitation wavelength was 296 nm and the fluorescence signal was observed using a 400 nm longpass filter (400FG03-25, LOT Oriel Group). This setup was used to selectively excite protein-bound MANT-nucleotides via fluorescence resonance (FRET) from the initially excited tryptophan residues of the protein. Measurements were performed in buffer consisting of 50 mM Tris–HCl pH 7.5, 200 mM KCl, 5 mM MgCl2, 2 mM EDTA. A set of direct mixing, displacement and competition experiments was performed to determine the association rate constants kon and the dissociation rate constants koff for both labelled (MANT-dADP and MANT-dATP) and unlabelled (ADP and ATP). The nucleotide-binding affinity Kd was calculated from the ratio koff/kon. ATP-binding experiments were performed in the presence of 0.01 mg ml−1 pyruvate kinase and 200 µM phosphoenol pyruvate as an ATP-regenerating system (MANT-nucleotides are unaffected by pyruvate kinase). Firstly, direct mixing experiments were performed in which 2 µM ClpB NBD2 was mixed with different concentrations (5–25 µM) of MANT-dADP or MANT-dATP. In displacement experiments, 2 µM ClpB NBD2 was incubated with 5 µM MANT-dADP or MANT-dATP and then mixed with different concentrations of ADP (0.5–5 mM). In competition experiments, ClpB NBD2 was mixed simultaneously with both MANT-dADP and unlabelled nucleotide (ADP or ATP), with final mixing concentrations of 0.5 µM protein, 2.5 µM MANT-dADP and 2.5–100 µM ADP or 25–300 µM ATP. Additional experimental data were needed to unambiguously determine the ATP-binding parameters. We therefore performed another displacement experiment in which 1 µM ClpB NBD2 was incubated with 200 µM ATP (in the presence of phosphoenol pyruvate and pyruvate kinase as an ATP-regenerating system) and then mixed with 10–100 µM MANT-dADP. The different experimental data sets were combined in a global fitting analysis using the program DYNAFIT (Kuzmic, 1996). The kinetic model given in the scheme below (where Nt = ADP or ATP) was applied for the fitting procedure, which utilizes numerical integration of the corresponding differential equations and nonlinear least-squares minimization.
The rate constants were treated as global parameters that were shared by all data sets included; the response coefficients that reflect the amplitude information were treated as semi-global parameters shared among one experimental series.
2.5. Steady-state ATPase assay
The steady-state ATPase activity of ClpB NBD2 variants was measured in a coupled colorimetric assay essentially as described previously (Schlee et al., 2001; Beinker et al., 2005). The measurements were performed at 25°C using a Jasco V-650 spectrophotometer (Jasco Germany GmbH). 10 µM protein was incubated in assay buffer (50 mM Tris–HCl pH 7.5, 100 mM KCl, 2 mM EDTA, 2 mM DTE, 0.8 mM phosphoenolpyruvate, 0.4 mM NADH, 0.1 g l−1 BSA, 4 U ml−1 pyruvate kinase, 6 U ml−1 lactate dehydrogenase, 5 mM MgCl2). The reaction was started by adding 1 mM ATP. The decrease in NADH absorption was monitored at 340 nm over time and the maximum slope was used to calculate the ATPase turnover rate per molecule.
3. Results and discussion
3.1. High-resolution structures of ClpB NBD2 in different nucleotide states reveal an inactive, open conformation of the catalytic site
In order to understand the mechanism behind the force generation upon ATP hydrolysis in the disaggregation machine ClpB, we aimed at obtaining high-resolution crystal structures of one of its two nucleotide-binding domains, NBD2, in different nucleotide states. We sought to determine structural differences between snapshots of different stages during the ATPase cycle in order to reveal the essential conformational changes that drive this molecular motor. We therefore crystallized ClpB NBD2 from T. thermophilus under nucleotide-free conditions as well as in the presence of ADP and the nonhydrolysable ATP analogue AMPPCP. Crystallization experiments using 2-propanol as the precipitant yielded well diffracting diamond-shaped crystals for all three states. Structures were determined by to resolutions of 2.8 Å (nucleotide-free), 2.4 Å (ADP) and 1.9 Å (AMPPCP) (Table 1). The ClpB NBD2 molecules exhibit a helical arrangement with sixfold symmetry owing to the hexagonal screw axis of the Therefore, the interface between neighbouring NBD2 molecules is slightly shifted compared with a planar hexameric complex, which to date could only be modelled based on electron-microscopy data (Lee et al., 2007; Biter et al., 2012) and not observed in X-ray structures of ClpB. In our structures, the proposed arginine finger Arg747 located in this interface is about 11 Å away from the nucleotide bound to the neighbouring NBD2 molecule. The overall fold of our structures is very similar to recently published structures of ClpB NBD2 (Biter et al., 2012) and the corresponding part of the structure of full-length ClpB (Lee et al., 2003). However, we observed a deviation in the positioning of the helix–loop–helix motif comprising residues 710–740, which is most likely to be caused by alternative crystal contacts in our crystal form. Most importantly, our structures of the nucleotide-free and AMPPCP-bound states are of significantly higher resolution than those available previously, allowing a more detailed structural analysis.
‡Rwork = . §Rfree is calculated like Rwork, using 5% of the data that were excluded from ¶Ideal values according to Engh & Huber (1991). ††Structure validation by MolProbity (Chen et al., 2010). |
We focused on analyzing the arrangement of the mechanistically relevant residues in the catalytic site (Fig. 2). The nucleotide-bound structures contained one molecule of ADP or AMPPCP positioned by the highly conserved hydrogen-bonding network consisting of nucleobase interactions as well as hydrogen bonds between the phosphate groups and the P-loop residues. The nucleotide-free structure contained a phosphate ion bound at the same position as the α-phosphate in the nucleotide-bound structures. This single phosphate ion forms a salt bridge (d = 3.0 Å) with the positively charged side chain of the sensor 2 residue Arg806, which is not observed in the nucleotide-bound states, in which Arg806 is involved in a crystal contact with Glu520 of a neighbouring molecule. Apart from this, the structures of the different nucleotide states are highly similar. Strikingly, in all ClpB NBD2 structures determined to date the side chain of the Walker A lysine Lys601, which is crucial for nucleotide binding and hydrolysis, is present in a completely stretched conformation that prevents the expected interaction with the β/γ-phosphates of the bound nucleotide (Fig. 2). Instead, the stretched Lys601 side chain forms hydrogen bonds to the carbonyl backbone groups of Gly595 and Ser708. These unusual interactions directly influence both the conformation of the P-loop (Gly595–Thr602) and the essential sensor 1 motif (Asn709). They furthermore stabilize an open, inactive conformation of the catalytic site in which the Walker B motif consisting of Asp667 and Glu668, which is believed to activate the attacking water molecule for ATP hydrolysis, cannot come close enough to fulfil its task and therefore rests in an inactive position. The second remarkable feature observed in all ClpB NBD2 structures is the lack of the essential Mg2+ ion usually coordinated by the phosphates of the bound nucleotide. Despite extensive efforts including co-crystallization with up to 100 mM MgCl2, it was not possible to incorporate the crucial Mg2+ ion into the nucleotide-bound structures of wild-type ClpB NBD2. This further supports the hypothesis that the catalytic site arrangements obtained to date reflect an inactive resting state with regard to ATP hydrolysis of ClpB NBD2. Importantly, this inactive picture characterized by a stretched Walker A lysine conformation and the absence of Mg2+ from the catalytic site is not only observed in our structures but also in all other ClpB NBD2 structures published to date, independent of the nucleotide state (Biter et al., 2012; Lee et al., 2003).
3.2. Implications from a ClpB NBD2 Walker A mutant (K601Q) structure in complex with the fluorescently labelled nucleotide MANT-dADP
In order to perturb the open, inactive arrangement of the ClpB NBD2 catalytic site, we replaced the Walker A lysine with a glutamine. It has previously been shown that this mutation causes severe defects in both nucleotide binding and ATPase activity (Schlee et al., 2001). To crystallize the mutant ClpB variant NBD2 K601Q, we supplemented the crystallization cocktail with the fluorescently labelled nucleotide MANT-dADP, which has been routinely used for nucleotide-binding studies of ClpB and is known to bind more than 100-fold more strongly to NBD2 than unlabelled ADP (Werbeck et al., 2009). This compensated for the loss in binding affinity owing to the Walker A mutation and allowed us to obtain a nucleotide-bound structure of ClpB NBD2 K601Q (Fig. 3).
The complex of ClpB NBD2 K601Q and MANT-dADP crystallized from a condition containing PEG 6000 and yielded a different crystal form to that observed previously for the wild-type protein in conditions containing 2-propanol (P212121 instead of P65). The structure was determined by to a resolution of 2.8 Å (Table 1), showing well defined electron density for the previously unstructured pore-loop region (residues 640–650). The fluorescently labelled MANT-dADP binds at the same position as the unlabelled in the structures obtained previously. The aromatic MANT group attached to the 3′-OH of the ribose fits well into a mainly hydrophobic cavity (Fig. 3c), which also explains the preferential binding of 3′-MANT-ADP compared with 2′-MANT-ADP owing to steric restrictions (Werbeck et al., 2009). The catalytic site shows a more closed conformation, characterized by a narrowing of the cleft between the large α/β core domain and the C-terminal small domain (Fig. 3a). A domain-motion analysis using DynDom (Hayward & Lee, 2002) comparing the MANT-dADP-bound form with the ADP-bound form showed that the angle between the domains decreases by 14°. The essential sensor 2 residue Arg806 located on the C-terminal small domain forms a salt bridge (d = 2.8 Å) with the β-phosphate of the nucleotide. An additional phosphate ion is present in the active site at a distance of 5 Å from the β-phosphate of the nucleotide, suggesting that we may have captured an ADP + Pi (product-bound) state here. However, this structure also does not reveal the active arrangement of the essential Walker B motif that would explain its mode of action. Furthermore, we were again not able to observe the crucial Mg2+ ion in the catalytic site. Interestingly, another highly conserved active-site residue, Arg621, adopts a different conformation to that in the structures described previously. This arginine points towards the nucleotide and is positioned directly between the Walker A residue Thr602 and the Walker B residue Asp667, thereby causing an arrangement that might block the incorporation of Mg2+ because it places a positive charge close to where the positively charged Mg2+ ion is expected to bind (Fig. 3b). With the intention to facilitate the incorporation of Mg2+ and thus obtain an active structure, we next generated the mutant variant ClpB NBD2 R621Q and performed both biochemical characterization and crystallization experiments, which are described in the following sections.
3.3. The active-site mutations R806A and R621Q modify the discrimination between ADP and ATP binding and influence the ATPase activity of ClpB NBD2
Changes in nucleotide-binding properties owing to active-site mutations provide very useful insights regarding the mechanistic roles of the respective residues in ATP hydrolysis and nucleotide-state sensing. The nucleotide-binding parameters of ClpB NBD2 have been characterized previously using transient kinetics experiments with fluorescently labelled MANT-nucleotides (Werbeck et al., 2009). A set of stopped-flow experiments including direct mixing, displacement and competition experiments in combination with global fitting analysis allowed the determination of the binding parameters for both MANT-labelled as well as unlabelled Interestingly, the binding affinity of ClpB NBD2 for ADP is almost 200-fold higher than for ATP. This strong discrimination is mainly owing to a lower dissociation koff for ADP (Werbeck et al., 2009). However, the X-ray structures available to date could not explain this phenomenon.
The residues Asn709 and Arg806 are part of the conserved sensor 1 and sensor 2 motifs, respectively, which have been proposed to be involved in nucleotide-state sensing (Wendler et al., 2012). We therefore characterized the nucleotide-binding properties and the steady-state ATPase activity of the two sensor-motif mutants NBD2 N709A and NBD2 R806A as well as the mutant NBD2 R621Q that was used for crystallization (Fig. 4 and Table 2). Mutating the sensor 1 residue Asn709 caused only very minor changes in nucleotide binding, whereas the ATPase activity was reduced by about 30-fold compared with the wild type. This indicates that Asn709 is of catalytic importance, but is not necessarily directly involved in nucleotide binding, which is in agreement with a previous study on the sensor 1 motif in Hsp104, the yeast homologue of ClpB (Hattendorf & Lindquist, 2002a).
|
Despite being nucleotide-binding competent, the ATPase activity of the sensor 2 mutant NBD2 R806A was almost completely abolished. This is in agreement with previous studies on other AAA+ proteins, which showed that the sensor 2 arginines are catalytically essential, most likely by stabilizing intermediate states of the hydrolysis cycle (Wendler et al., 2012). Our nucleotide-binding kinetics measurements revealed a significantly reduced discrimination between ADP and ATP binding for the sensor 2 mutant NBD2 R806A, with only a very minor influence on ATP binding arising from the mutation, but significantly impaired ADP binding (Fig. 4d). This result hints towards a direct interaction of Arg806 with the nucleotide that contributes to the binding energy in the ADP state but not in the ATP state, which is discussed in further detail in the following section with respect to the structural data available. Previous results from equilibrium titrations with Hsp104 by Lindquist and coworkers suggested similarly impaired nucleotide binding for both ADP and ATP in the presence of the sensor 2 mutation R826M in the oligomeric context of the full-length protein (Hattendorf & Lindquist, 2002b). Our approach using only the isolated nucleotide-binding domain 2 (NBD2) of ClpB allows a more detailed kinetic analysis of the nucleotide-binding behaviour that is able to capture the differences in nucleotide discrimination.
The mutant variant NBD2 R621Q is nucleotide-binding competent, but shows an approximately 30-fold reduced ATPase activity. It is highly likely that Arg621 is involved in stabilizing negative charges, especially in the transition state. The R621Q mutation significantly reduces the discrimination between ATP and ADP, mainly by impaired ADP binding, but also owing to a slight increase in ATP binding (Fig. 4d and Table 2). As discussed before for NBD2 R806A, it is possible that Arg621 functions as a nucleotide sensor, adopting alternative conformations dependent on the nucleotide-binding state. Arg621 is highly conserved in ClpB and Hsp104 homologues from different organisms, as well as related AAA+ proteins that possess a second C-terminal nucleotide-binding domain, such as ClpA and ClpC. A potential mechanistic role of Arg621 is discussed further in the next section, which introduces the high-resolution X-ray structure of ClpB NBD2 R621Q in complex with the nonhydrolysable ATP analogue AMPPCP.
3.4. Mechanistic insights from the structure of ClpB NBD2 R621Q bound to an ATP analogue with the essential Mg2+ ion incorporated into the catalytic site
3.4.1. The role of Arg621 in Mg2+ incorporation
Analysis of the ClpB NBD2 K601Q structure (see Fig. 3b) led to the hypothesis that the highly conserved residue Arg621 might interfere with the binding of Mg2+ to the phosphates of the bound nucleotide in the catalytic site. In at least one of its possible conformations, the Arg621 side chain positions a positive charge close to the expected Mg2+-binding site and thereby competes with Mg2+ incorporation. The mutation R621Q was therefore introduced to facilitate the incorporation of the essential Mg2+ into the ClpB NBD2 structure. Crystallization of ClpB NBD2 R621Q in the presence of the nonhydrolysable ATP analogue AMPPCP and MgCl2 yielded well diffracting diamond-shaped crystals in P65 (Table 1). The structure was determined to a resolution of 1.7 Å and for the first time we could unambiguously locate the essential Mg2+ ion between the β- and γ-phosphates of the bound AMPPCP (Fig. 5a). The Mg2+ ion is coordinated octahedrally by the two phosphate groups, the hydroxyl group of the Walker A residue Thr602 and three water molecules. The Walker B residue Asp667 forms hydrogen bonds to two of the hydration-shell water molecules. Additionally, we observe electron density for a second Mg2+ ion coordinated octahedrally by six water molecules at a distance of 4.5 Å to the γ-phosphate of the bound AMPPCP. We assume that this additional Mg2+ ion is not of mechanistic relevance but rather is incorporated owing to the R621Q mutation and the nonphysiologically high MgCl2 concentration of 75 mM used in the crystallization cocktail. Indeed, when comparing the Arg621 side-chain conformations in the different structures available, two alternative positions are observed. The positively charged side-chain guanidinium group is either located close to the nucleotide, where it blocks the incorporation of the essential Mg2+ ion, or close to where the second (most likely nonphysiological) Mg2+ ion is found in the NBD2 R621Q structure (Fig. 5c). The negatively charged environment that surrounds this second position (formed by Asp623, Asp667 and Glu668) can easily stabilize a positive charge, be it the Arg621 side chain in the wild-type protein or the second Mg2+ ion incorporated into this position owing to the R621Q mutation. To verify this hypothesis, we repeated the crystallization experiment in the presence of only 10 mM MgCl2 and additionally soaked with 100 mM guanidinium chloride (GdmCl), which was supposed to function as a mobile substitute for the arginine side chain that was removed owing to the R621Q mutation. Indeed, in the structure obtained from crystals grown in this condition, only one Mg2+ ion is present, which occupies the expected position between the β- and γ-phosphates of the bound AMPPCP. Additionally, a Gdm+ ion is localized where the second Mg2+ ion was bound before, suggesting that this position would usually be occupied by the Arg621 side chain (Fig. 5b).
Based on this structural analysis and the kinetic data showing a reduced discrimination between ADP and ATP binding in ClpB NBD2 R621Q (Fig. 4d and Table 2), we propose that Arg621 is involved in nucleotide sensing and the incorporation of the essential Mg2+ ion into the catalytic site which is crucial for efficient ATP hydrolysis. Arg621 adopts two different conformations, which are presumably dependent on the stage of the ATPase cycle (Fig. 5c). Since the R621Q mutation selectively impairs ADP binding, we conclude that in the ADP state Arg621 contributes significantly to the nucleotide-binding energy, presumably by mimicking the nucleotide-associated Mg2+. In the ATP state, Arg621 is most likely to adopt the position further away from the nucleotide, thereby not contributing to the binding energy but allowing the incorporation of the essential Mg2+ ion. Regulatory metal switches have also been found in other ATPases, such as the DNA mismatch-repair protein MutS, in which the essential Walker B motif is rearranged in the absence of Mg2+ (Lebbink et al., 2010), which is reminiscent of the situation that we observe for ClpB NBD2. We therefore hypothesize that a variation of a metal-switch mechanism with Arg621 as a key regulatory feature is implemented in ClpB NBD2.
3.4.2. Nucleotide state-sensing mechanisms
The sensor 1 and 2 motif residues Asn709 and Arg806, respectively, are catalytically essential and have been proposed to be involved in sensing and communicating different nucleotide states (Guenther et al., 1997). The structural basis of how this function is fulfilled, however, is poorly understood. Owing to the high resolution of 1.7 Å, we were able to assign alternative conformations for certain residues in the structure of ClpB NBD2 R621Q in complex with AMPPCP. We identified a hydrogen-bonding network between sensor 1 residue Asn709, P-loop residue Thr597 and the γ-phosphate of the bound nucleotide. Depending on its orientation, Thr597 can either form a hydrogen bond to the nucleotide or to the side chain of Asn709 (Fig. 6a). One can speculate that in the ATP state Thr597 preferentially interacts with the γ-phosphate. In the nucleotide-free or ADP state, when no γ-phosphate is present, it rotates and interacts with Asn709. This putative switch might be of regulatory importance. Even though the sensor 1 mutation N709A causes no significant changes in the nucleotide-binding properties, it severely impairs ATP hydrolysis. This is in agreement with the hypothesis that the sensor 1 residue works in concert with the Walker B motif in activating the attacking water molecule during hydrolysis (Gai et al., 2004). A nucleotide-dependent conformational switch of Asn709 might be the regulating element for this function. However, the structural data available are not sufficient to unambiguously confirm this hypothesis. Based on the kinetic data, we can state that the essential sensor 1 motif (Asn709) is not directly involved in nucleotide binding, which is inconsistent with the role of a classic nucleotide sensor. This result is in agreement with a previous study by Hattendorf & Lindquist (2002a).
In contrast, our characterization of the sensor 2 motif (Arg806) provides clear evidence for direct nucleotide sensing and discrimination. In addition to the AMPPCP-bound structure of ClpB NBD2 R621Q in the presence of MgCl2 and GdmCl, we also obtained the structure of this mutant variant in the ADP-bound state to a resolution of 2.1 Å (Table 1). Since the sensor 2 residue Arg806 was no longer trapped in a crystal contact in these structures, we could identify two different conformations of Arg806 that are adopted depending on the nucleotide state (Fig. 6b). In the ADP state, the side chain of Arg806 interacts directly with the β-phosphate of the bound nucleotide (d < 3.0 Å), which is also observed in both our MANT-dADP-bound structure and the recently published NBD2 structure (PDB entry 4fcw ; Biter et al., 2012). In the ATP state, the Arg806 side chain must adopt an alternative conformation because otherwise it would clash with the γ-phosphate. Indeed, Arg806 is located about 4 Å away from the nucleotide in the AMPPCP-bound structure. This is in agreement with our nucleotide-binding data, which show that the sensor 2 mutation R806A causes significantly impaired ADP binding but almost no change in ATP binding. It can be concluded that the direct interaction with the nucleotide in the ADP state contributes to the binding energy, whereas in the ATP state no interaction between Arg806 and the nucleotide takes place. This, together with the function of residue Arg621 described in the previous sections, provides a veritable explanation for the observed discrimination between ADP and ATP binding in ClpB NBD2. The combination of structural and kinetic analysis further defines the role of the sensor 2 residue Arg806 in nucleotide-state sensing and identifies one clear difference between the ADP and ATP state, which is crucial for understanding the mode of action of this ATPase motor domain.
3.5. Comparison with other AAA+ protein structures regarding the hexameric context
To evaluate the conclusions drawn from the ClpB NBD2 structures presented in this work, it is crucial to discuss the results in the light of the oligomeric context of the holoprotein. Even though smaller oligomeric species, as present for the separate NBD2 construct, hydrolyze ATP efficiently, full disaggregation activity is most likely to require a hexameric assembly. There are several structural studies available that show hexameric states of AAA+ proteins, for example for p97, ClpX, SV40 LTag and ClpC. Despite a common overall architecture, the essential interactions and resulting structural changes in the oligomeric arrangement upon ATP hydrolysis vary remarkably for the different AAA+ proteins. For ClpC, which shares high sequence identity with ClpB, a planar hexameric state in complex with the adaptor protein MecA has been observed in both X-ray structures as well as cryo-EM studies (Liu et al., 2013; Wang et al., 2011). Unfortunately, the resolution is not sufficient to discuss detailed side-chain interactions. For the AAA+ helicase LTag from SV40, structures of the planar hexamer are available for different nucleotide states at very high resolution (up to 1.9 Å), which allowed a detailed mechanistic analysis (Gai et al., 2004). Here, a conserved lysine from the neighbouring subunit works in concert with the classic arginine finger. Furthermore, extensive mechanistic and structural studies regarding the hexameric assembly have been performed for ClpX, another prominent AAA+ unfoldase (Glynn et al., 2009, 2012; Martin et al., 2005; Stinson et al., 2013), and for p97, an ERAD-associated AAA+ protein (Davies et al., 2008; Li et al., 2012; Nishikori et al., 2011).
To evaluate our results, we compared the average distances between the nucleotide and the catalytically relevant Walker and sensor residues in our structures with those observed in the available high-resolution hexameric AAA+ structures, using the analysis presented by Wendler and coworkers as a starting point (Wendler et al., 2012). For instance, the distance between the side-chain carbonyl O atom of the Walker B aspartate and the γ-phosphate O atom of the nucleotide ranges from 3.0 to 5.5 Å in published hexameric structures and is 4.5 Å in the case of ClpB NBD2 R621Q + AMPPCP. Similarly, the distance between the nucleotide and the sensor 1 asparagine side chain ranges from 3.1 Å for SV40 LTag (PDB entry 1svm ; Gai et al., 2004) to 6.0 Å for p97 (PBD entry 3cf1 ; Davies et al., 2008), and is 4.6 Å in the case of ClpB NBD2. Sensor 2 motifs are only present in a subset of AAA+ proteins. We compared ClpB NBD2 with ClpX (PDB entry 3hws ; Glynn et al., 2009) and also found similar distances and orientations regarding the sensor 2 arginine. Even though we cannot exclude that our single NBD2 structures somewhat differ from their counterparts in the hexamer, this comparison indicates that they may very well represent snapshots of relevant states that are also adopted in the hexameric context.
It is remarkable that all of the structures of ClpB (full-length and NBD2) available to date show a helical arrangement of subunits representing a spiral-like oligomeric state, which is also observed in the available high-resolution structure of ClpA (Guo et al., 2002), a closely related unfoldase that also bears two NBDs per molecule. Even though cryo-EM data suggest that the planar hexameric ring is the predominant active form, it cannot be excluded that spirals are present in solution and are of certain mechanistic relevance. A possible biological role of a spiral has actually been discussed for other AAA+ proteins (Erzberger & Berger, 2006).
Even though the exact arrangement of the functionally relevant oligomeric state in ClpB remains elusive, our NBD2 structures contribute to the overall mechanistic understanding by providing structural details at high resolution for one separate motor building block of an AAA+ machine in different states. A summary of the different states captured for ClpB NBD2 is presented in Fig. 7.
3.6. Various P-loop proteins exhibit similar inactive conformations: a potential regulatory function of the Walker A lysine as a molecular switch
Even though we were finally able to incorporate the essential Mg2+ ion into the structure of ClpB NBD2, the overall conformation of the catalytic site remained open and therefore inactive. The Walker A lysine Lys601 is still stretched and contacts P-loop backbone carbonyl groups but not the nucleotide. We therefore analyzed the structures of other P-loop proteins available at sufficiently high resolutions and found more examples of such open inactive catalytic site conformations characterized by a stretched Walker A lysine. In the structure of a closely related AAA+ protein, ClpA, the situation is almost identical (PDB entry 1r6b ; Xia et al., 2004). Whereas in NBD1 of both of the proteins ClpA and ClpB the Walker A lysine directly interacts with the nucleotide in the conventional way, in NBD2 the stretched Walker A lysine conformation is observed (Fig. 8a). However, more distant relatives in the large family of P-loop proteins were also found to adopt this open inactive conformation of the catalytic site in their respective crystal structures. Three examples are shown in Fig. 8(b): the kinesin family member Kin10 in complex with ADP (PDB entry 3pxn ; Cochran et al., 2011), a human GTPase controlling apoptosis in lymphocytes, GIMAP2, in complex with GTP (PDB entry 2xtn ; Schwefel et al., 2010) and the DNA mismatch-repair protein MutS in complex with ADP (PDB entry 3k0s ; Lebbink et al., 2010). Interestingly, in the latter example the inactive catalytic site arrangement is also coupled to the absence of Mg2+. We hypothesize that the open catalytic site conformations represent an inactive resting state during the ATPase cycle, which is often captured in X-ray structures and may be of regulatory importance. Zhang and Wigley proposed previously that a so-called `glutamate switch' directly controls the conformation of the essential Walker B glutamate residue in AAA+ domains, keeping it in an inactive state by hydrogen bonding to a polar residue located on the tip of β-strand β2 (Zhang & Wigley, 2008). In ClpB NBD2 from T. thermophilus, this residue corresponds to a threonine (Thr625). However, we do not observe a direct hydrogen-bonding interaction with the Walker B glutamate Glu668 in our structures. Instead, we suggest a different mechanism that forces the catalytically essential residues, including the Walker B glutamate, into an inactive position. We propose that the Walker A lysine in its stretched conformation acts like a `molecular doorstop' that keeps the catalytic site open and thereby inactive in order to avoid unnecessary ATP turnover, for example in the absence of a secondary substrate. Many ATPases have a low basal ATP-hydrolysis activity which is significantly stimulated upon addition of substrate, with the highly conserved Walker A lysine having a catalytic role. We propose a second, regulatory function for the Walker A lysine, where it is involved in switching between an inactivated, `fuel-saving' state and the conventional active state of a molecular motor.
The question arises as to why this regulatory mechanism is only implemented in a subset of AAA+ domains. In ClpB, the stretched Walker A lysine conformation is observed exclusively in the C-terminal nucleotide-binding domain NBD2, but not in the N-terminal NBD1. In the current mechanistic view, NBD2 is the actual motor of the ClpB disaggregation machine, whereas NBD1 is mainly of regulatory importance, interacting with the co-chaperone DnaK via the unique coiled-coil middle domain, which turns out to be a highly sophisticated regulatory element of ClpB (Rosenzweig et al., 2013; Oguchi et al., 2012). Thus, NBD1 is already highly regulated, whereas the strong motor NBD2 needs efficient regulation, which we propose is provided by a conformational switch of the Walker A lysine that is likely to be induced by binding of the substrate protein.
4. Conclusions
Our structure-based analysis of ClpB NBD2, a representative of the AAA+ motor domains, reveals mechanistic insights into nucleotide sensing and the regulation of the ATPase activity. We show how the conserved sensor 2 motif directly contributes to the discrimination between ADP and ATP binding. Furthermore, we suggest that a conserved active-site arginine regulates the incorporation of the essential Mg2+ ion and is therefore essential for the functioning of the ATPase cycle. In addition, we propose a regulatory switch where the conserved Walker A lysine, when present in a fully stretched conformation, causes an inactive arrangement of the catalytic site to avoid unnecessary ATPase turnover.
Acknowledgements
We thank Sabine Zimmermann and Jessica Eschenbach for excellent technical assistance and Anton Meinhart for help with in-house X-ray data collection. We are indebted to Anna Scherer, Victor Kulik and Philipp Beinker for contributions during the initial stages of the project. We are grateful to the X-ray data-collection teams from the Max Planck Institutes in Heidelberg and Dortmund, who performed the measurements at the Swiss Light Source (SLS). We also thank the SLS staff for their support in setting up the beamline and Ingrid Vetter and Chris Roome for excellent support with the crystallographic software. The project was funded by the Max Planck Society and a PhD scholarship to CZ from the German National Academic Foundation.
References
Beinker, P., Schlee, S., Auvula, R. & Reinstein, J. (2005). J. Biol. Chem. 280, 37965–37973. Web of Science CrossRef PubMed CAS Google Scholar
Biter, A. B., Lee, S., Sung, N. & Tsai, F. T. (2012). Proc. Natl Acad. Sci. USA, 109, 12515–12520. Web of Science CrossRef CAS PubMed Google Scholar
Brünger, A. T. (1990). Acta Cryst. A46, 46–57. CrossRef Web of Science IUCr Journals Google Scholar
Chen, V. B., Arendall, W. B., Headd, J. J., Keedy, D. A., Immormino, R. M., Kapral, G. J., Murray, L. W., Richardson, J. S. & Richardson, D. C. (2010). Acta Cryst. D66, 12–21. Web of Science CrossRef CAS IUCr Journals Google Scholar
Cochran, J. C., Zhao, Y. C., Wilcox, D. E. & Kull, F. J. (2011). Nature Struct. Mol. Biol. 19, 122–127. Web of Science CrossRef Google Scholar
Davies, J. M., Brunger, A. T. & Weis, W. I. (2008). Structure, 16, 715–726. Web of Science CrossRef PubMed CAS Google Scholar
Doyle, S. M. & Wickner, S. (2009). Trends Biochem. Sci. 34, 40–48. Web of Science CrossRef PubMed CAS Google Scholar
Emsley, P. & Cowtan, K. (2004). Acta Cryst. D60, 2126–2132. Web of Science CrossRef CAS IUCr Journals Google Scholar
Engh, R. A. & Huber, R. (1991). Acta Cryst. A47, 392–400. CrossRef CAS Web of Science IUCr Journals Google Scholar
Erzberger, J. P. & Berger, J. M. (2006). Annu. Rev. Biophys. Biomol. Struct. 35, 93–114. Web of Science CrossRef PubMed CAS Google Scholar
Franzmann, T. M., Czekalla, A. & Walter, S. G. (2011). J. Biol. Chem. 286, 17992–18001. Web of Science CrossRef CAS PubMed Google Scholar
Gai, D., Zhao, R., Li, D., Finkielstein, C. V. & Chen, X. S. (2004). Cell, 119, 47–60. Web of Science CrossRef PubMed CAS Google Scholar
Glynn, S. E., Martin, A., Nager, A. R., Baker, T. A. & Sauer, R. T. (2009). Cell, 139, 744–756. Web of Science CrossRef PubMed CAS Google Scholar
Glynn, S. E., Nager, A. R., Baker, T. A. & Sauer, R. T. (2012). Nature Struct. Mol. Biol. 19, 616–622. Web of Science CrossRef CAS Google Scholar
Guenther, B., Onrust, R., Sali, A., O'Donnell, M. & Kuriyan, J. (1997). Cell, 91, 335–345. CrossRef CAS PubMed Web of Science Google Scholar
Guo, F., Maurizi, M. R., Esser, L. & Xia, D. (2002). J. Biol. Chem. 277, 46743–46752. Web of Science CrossRef PubMed CAS Google Scholar
Hanson, P. I. & Whiteheart, S. W. (2005). Nature Rev. Mol. Cell Biol. 6, 519–529. Web of Science CrossRef CAS Google Scholar
Hattendorf, D. A. & Lindquist, S. L. (2002a). EMBO J. 21, 12–21. Web of Science CrossRef PubMed CAS Google Scholar
Hattendorf, D. A. & Lindquist, S. L. (2002b). Proc. Natl Acad. Sci. USA, 99, 2732–2737. Web of Science CrossRef PubMed CAS Google Scholar
Hayward, S. & Lee, R. A. (2002). J. Mol. Graph. Model. 21, 181–183. Web of Science CrossRef PubMed CAS Google Scholar
Kabsch, W. (2010). Acta Cryst. D66, 125–132. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kuzmic, P. (1996). Anal. Biochem. 237, 260–273. CrossRef CAS PubMed Web of Science Google Scholar
Lebbink, J. H., Fish, A., Reumer, A., Natrajan, G., Winterwerp, H. H. & Sixma, T. K. (2010). J. Biol. Chem. 285, 13131–13141. Web of Science CrossRef CAS PubMed Google Scholar
Lee, S., Choi, J.-M. & Tsai, F. T. (2007). Mol. Cell, 25, 261–271. Web of Science CrossRef PubMed CAS Google Scholar
Lee, S., Sowa, M. E., Watanabe, Y.-H., Sigler, P. B., Chiu, W., Yoshida, M. & Tsai, F. T. F. (2003). Cell, 115, 229–240. Web of Science CrossRef PubMed CAS Google Scholar
Li, G., Huang, C., Zhao, G. & Lennarz, W. J. (2012). Proc. Natl Acad. Sci. USA, 109, 3737–3741. Web of Science CrossRef CAS PubMed Google Scholar
Liu, J., Mei, Z., Li, N., Qi, Y., Xu, Y., Shi, Y., Wang, F., Lei, J. & Gao, N. (2013). J. Biol. Chem. 288, 17597–17608. Web of Science CrossRef CAS PubMed Google Scholar
Martin, A., Baker, T. A. & Sauer, R. T. (2005). Nature (London), 437, 1115–1120. Web of Science CrossRef PubMed CAS Google Scholar
McCoy, A. J., Grosse-Kunstleve, R. W., Adams, P. D., Winn, M. D., Storoni, L. C. & Read, R. J. (2007). J. Appl. Cryst. 40, 658–674. Web of Science CrossRef CAS IUCr Journals Google Scholar
McRee, D. E. (1999). J. Struct. Biol. 125, 156–165. Web of Science CrossRef PubMed CAS Google Scholar
Murshudov, G. N., Skubák, P., Lebedev, A. A., Pannu, N. S., Steiner, R. A., Nicholls, R. A., Winn, M. D., Long, F. & Vagin, A. A. (2011). Acta Cryst. D67, 355–367. Web of Science CrossRef CAS IUCr Journals Google Scholar
Nishikori, S., Esaki, M., Yamanaka, K., Sugimoto, S. & Ogura, T. (2011). J. Biol. Chem. 286, 15815–15820. Web of Science CrossRef CAS PubMed Google Scholar
Oguchi, Y., Kummer, E., Seyffer, F., Berynskyy, M., Anstett, B., Zahn, R., Wade, R. C., Mogk, A. & Bukau, B. (2012). Nature Struct. Mol. Biol. 19, 1338–1346. Web of Science CrossRef CAS Google Scholar
Ogura, T., Whiteheart, S. W. & Wilkinson, A. J. (2004). J. Struct. Biol. 146, 106–112. Web of Science CrossRef PubMed CAS Google Scholar
Rosenzweig, R., Moradi, S., Zarrine-Afsar, A., Glover, J. R. & Kay, L. E. (2013). Science, 339, 1080–1083. Web of Science CrossRef CAS PubMed Google Scholar
Sanchez, Y. & Lindquist, S. L. (1990). Science, 248, 1112–1115. CrossRef CAS PubMed Web of Science Google Scholar
Saraste, M., Sibbald, P. R. & Wittinghofer, A. (1990). Trends Biochem. Sci. 15, 430–434. CrossRef PubMed Web of Science Google Scholar
Schlee, S., Groemping, Y., Herde, P., Seidel, R. & Reinstein, J. (2001). J. Mol. Biol. 306, 889–899. Web of Science CrossRef PubMed CAS Google Scholar
Schwefel, D., Fröhlich, C., Eichhorst, J., Wiesner, B., Behlke, J., Aravind, L. & Daumke, O. (2010). Proc. Natl Acad. Sci. USA, 107, 20299–20304. Web of Science CrossRef CAS PubMed Google Scholar
Seyffer, F., Kummer, E., Oguchi, Y., Winkler, J., Kumar, M., Zahn, R., Sourjik, V., Bukau, B. & Mogk, A. (2012). Nature Struct. Mol. Biol. 19, 1347–1355. Web of Science CrossRef CAS Google Scholar
Simonetta, K. R., Kazmirski, S. L., Goedken, E. R., Cantor, A. J., Kelch, B. A., McNally, R., Seyedin, S. N., Makino, D. L., O'Donnell, M. & Kuriyan, J. (2009). Cell, 137, 659–671. Web of Science CrossRef PubMed CAS Google Scholar
Stinson, B. M., Nager, A. R., Glynn, S. E., Schmitz, K. R., Baker, T. A. & Sauer, R. T. (2013). Cell, 153, 628–639. Web of Science CrossRef CAS PubMed Google Scholar
Tucker, P. A. & Sallai, L. (2007). Curr. Opin. Struct. Biol. 17, 641–652. Web of Science CrossRef PubMed CAS Google Scholar
Walker, J. E., Saraste, M., Runswick, M. J. & Gay, N. J. (1982). EMBO J. 1, 945–951. CAS PubMed Web of Science Google Scholar
Wang, F., Mei, Z., Qi, Y., Yan, C., Hu, Q., Wang, J. & Shi, Y. (2011). Nature (London), 471, 331–335. Web of Science CrossRef CAS PubMed Google Scholar
Weibezahn, J., Tessarz, P., Schlieker, C., Zahn, R., Maglica, Z., Lee, S., Zentgraf, H., Weber-Ban, E. U., Dougan, D. A., Tsai, F. T., Mogk, A. & Bukau, B. (2004). Cell, 119, 653–665. Web of Science CrossRef PubMed CAS Google Scholar
Wendler, P., Ciniawsky, S., Kock, M. & Kube, S. (2012). Biochim. Biophys. Acta, 1823, 2–14. Web of Science CrossRef CAS PubMed Google Scholar
Werbeck, N. D., Kellner, J. N., Barends, T. R. M. & Reinstein, J. (2009). Biochemistry, 48, 7240–7250. Web of Science CrossRef PubMed CAS Google Scholar
Werbeck, N. D., Schlee, S. & Reinstein, J. (2008). J. Mol. Biol. 378, 178–190. Web of Science CrossRef PubMed CAS Google Scholar
Winkler, J., Tyedmers, J., Bukau, B. & Mogk, A. (2012). J. Cell Biol. 198, 387–404. Web of Science CrossRef CAS PubMed Google Scholar
Winn, M. D., Isupov, M. N. & Murshudov, G. N. (2001). Acta Cryst. D57, 122–133. Web of Science CrossRef CAS IUCr Journals Google Scholar
Xia, D., Esser, L., Singh, S. K., Guo, F. & Maurizi, M. R. (2004). J. Struct. Biol. 146, 166–179. Web of Science CrossRef PubMed CAS Google Scholar
Zeymer, C., Werbeck, N. D., Schlichting, I. & Reinstein, J. (2013). J. Biol. Chem. 288, 7065–7076. Web of Science CrossRef CAS PubMed Google Scholar
Zhang, X. & Wigley, D. B. (2008). Nature Struct. Mol. Biol. 15, 1223–1227. Web of Science CrossRef CAS Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.