|Home | About | Journals | Submit | Contact Us | Français|
Hsp90 is a ubiquitous molecular chaperone. Previous structural analysis demonstrated that Hsp90 can adopt a large number of structurally distinct conformations, however the functional role of this flexibility is not understood. Here we investigate the structural consequences of substrate binding with a model system in which Hsp90 interacts with a partially folded protein (Δ131Δ), a well-studied fragment of staphylococcal nuclease. SAXS measurements reveal that under apo conditions Hsp90 partially closes aroundΔ131Δ and in the presence of AMPPNP Δ131Δ binds with increased affinity to Hsp90’s fully closed state. FRET measurement show that Δ131Δ accelerates the nucleotide-driven open/closed transition and stimulates ATP hydrolysis by Hsp90. NMR measurements reveal that Hsp90 binds to a specific, highly structured, region of Δ131Δ. These results suggest that Hsp90 preferentially binds a locally structured region in a globally unfolded protein and this binding drives functional changes in the chaperone by lowering a rate-limiting conformational barrier.
Hsp90 is a ubiquitous molecular chaperone. Originally identified in the heat shock response, Hsp90 plays important regulatory roles under non-stress conditions by its interactions with specific classes of substrates such as kinases and nuclear receptors (Young et al., 2001). Consistent with being upregulated upon heat shock, Hsp90 can suppress thermal aggregation (Jakob et al., 1995; Wiech et al., 1992) and facilitate protein folding by reducing misfolding via interactions with aggregation-prone unfolding intermediates (Schneider et al., 1996). In vivo, Hsp90 receives unfolded proteins transferred from Hsp40/70 and receives specific classes of native or near-native substrates that are recruited by dedicated cochaperones. In eukaryotes, Hsp90’s substrate interactions are highly regulated by cochaperones that modulate the Hsp90 ATPase activity, aid in substrate recruitment, maturation or release, and target substrates for degradation or intracellular trafficking. These important and multifaceted roles are reflected in Hsp90’s high cellular abundance (2–5% of cytosolic protein under non-stressed conditions). The in vivo characterized substrates (see http://www.picard.ch/) have not been found to share a common sequence or structure motif and span an exceptionally wide range of sizes from α-synuclein to telomerase (14–290 kD (Falsone et al., 2009; Forsythe et al., 2001)).
Recent structural work demonstrated that while Hsp90 is a dimer where each monomer has three well-folded and stable domains (N-terminal, middle, C-terminal), the overall molecule can adopt radically different conformations (Figure 1) in response to nucleotide and conditions. For example, under apo conditions the Hsp90 from E. coli, HtpG, crystallized in a ‘V’-shaped conformation (Shiau et al., 2006) (ribbon structure, left panel Figure 1) while in solution a highly open conformation (blue surface structure) and a more closed conformation (red surface) are populated in a pH-dependent manner (Krukenberg et al., 2008; Krukenberg et al., 2009b). The more closed state is similar to a crystal structure of the Hsp90 homolog specific to the endoplasmic reticulum, Grp94 (Dollins et al., 2007). The apo conformations largely differ by rigid-body rotation at the interface between the middle domain (MD) and C-terminal domain (CTD), creating a variable sized cleft between the monomer arms; there is also NTD-MD rotation, changing the NTD orientation. Addition of non-hydrolysable ATP analogs such as AMPPNP results in closure to an N-terminal domain (NTD) dimerized conformation (Ali et al., 2006) (Figure 1, central panel), while ADP transiently stabilizes a very compact state (Figure 1, right panel)(Shiau et al., 2006) (Southworth and Agard, 2008). Small angle x-ray scattering (SAXS) and single particle electron microscopy measurements have shown that multiple Hsp90 conformations coexist in a delicate equilibrium that can be shifted not only by nucleotide binding, but also by osmolyte and pH conditions (Krukenberg et al., 2009b; Southworth and Agard, 2008; Street et al., 2009). SAXS has been a powerful tool for characterizing the Hsp90 conformational ensemble (Bron et al., 2008; Krukenberg et al., 2009a; Krukenberg et al., 2008; Krukenberg et al., 2009b; Onuoha et al., 2008; Zhang et al., 2004), and has shown that the apo-state flexibility is universal in all homologs that have been examined (HtpG, Hsc82, hHsp90α, Grp94 and TRAP, (Krukenberg et al., 2009a) and unpublished observations). However, the functional role of Hsp90 flexibility and conformational dynamics is not understood.
Indeed, despite the ubiquity of Hsp90 and the long list of in-vivo identified substrates (known as client proteins) little is known about how Hsp90 makes these interactions. There is some evidence suggesting that substrate folding stability is linked to Hsp90 binding. Hsp90 has a much stronger in-vivo interaction with the highly destabilized v-Src versus c-Src (otherwise having 98% sequence identity) (Taipale et al., 2010), NMR studies of p53 indicate that human Hsp90 only binds after substrate unfolding(Rudiger et al., 2002), and HtpG has been found to bind an unfolded ribosomal protein L2 (Motojima-Miyazaki et al., 2010). Indeed, the ability of Hsp90 to interact with and shift the equilibrium between metastable conformations is thought to play a role in Hsp90 in transitioning the glucocorticoid receptor between apo and ligand-bound states, which requires a large conformational change. These observations suggest that Hsp90/substrate interactions may be enhanced by reducing substrate stability to favor partially structured or metastable conformations. This approach is technically challenging because for most proteins partially folded states are difficult to populate and are prone to misfolding and aggregation.
One protein system that is amenable to this approach is the well-characterized staphylococcal nuclease (SN). Extensive studies have shown that a 131-residue fragment of SN (Δ131Δ; full length is 149 residues) is globally unfolded but remains compact with residual structured regions(Shortle, 2002) (Alexandrescu et al., 1994; Alexandrescu and Shortle, 1994; Wang and Shortle, 1995). Indeed, Δ131Δ and other similarly destabilized SN variants are close in free energy to the native state, as indicated by the fact they can be effectively refolded with tight binding inhibitors and stabilizing osmolytes (Baskakov and Bolen, 1998; Wang et al., 1995). Δ131Δ is monomeric at high concentrations, stable under a wide variety of conditions, and amenable to NMR, all of which has made it an ideal model system to investigate structural properties of unfolded proteins (Shortle, 2002). Here we test Δ131Δ as a model system to investigate Hsp90/substrate interactions. Using a combination of SAXS, FRET, binding anisotropy and NMR, we find that Hsp90 binds a structured region of Δ131Δ, which results in conformational and functional changes in the chaperone.
To determine Hsp90’s binding affinity for Δ131Δ, we labeled a cysteine variant of Δ131Δ with the IAEDANS fluorophore to measure fluorescence polarization anisotropy. The Perrin equation estimates of the rotational correlation times for Δ131Δ and HtpG (5 and 55 ns) span the IAEDANS excited state lifetime (10–15 ns), suggesting that binding will significantly increase polarization anisotropy. Indeed, upon addition of the bacterial Hsp90, HtpG, the fluorescence anisotropy of IAEDANS-labeled Δ131Δ increases substantially (Figure 2a). The concentration series is well-fit by single-site non-cooperative binding (solid lines), resulting in a Kd of 9 μM. Similar binding curves were measured for the yeast Hsp90 homolog (Kd of 6 μM) whereas addition of BSA resulted in minimal anisotropy changes (not shown). Anisotropy titration measurements show saturation near a 1:1 stoichiometry (Hsp90 dimer: Δ131Δ, Supplemental Figure 1a).
Previous studies have shown that unfolded SN fragments can be refolded in the presence of CaCl2 and a tight binding inhibitor, thymidine 3′,5′-bisphosphate (pdTp) (Shortle and Meeker, 1989). We used these stabilizers to refold IAEDANS-labeledΔ131Δ and found that HtpG no longer significantly increases fluorescence anisotropy, demonstrating that HtpG is binding a globally unfolded state. However, as discussed later HtpG selectively interacts with a highly structured region within the unfolded protein.
The micromolar concentration range for binding precludes structural analysis by electron microscopy, which for Hsp90 is performed at ~ 100 nM, whereas SAXS measurements are ideally suited for this concentration regime. Previous studies have shown that SAXS measurements can be used to determine the multi-state conformational equilibrium of Hsp90. For these experiments, X-ray scattering intensity was measured between Q values (4πsinθ/λ, where 2θ is the scattering angle) of 0.01 to 0.3Å−1, radially averaged and buffer subtracted. The resulting data were transformed to an interatomic distance distribution, P(r), using the GNOM program (Svergun, 1991). The P(r) distribution of HtpG alone has significant magnitude at large distances (Figure 2b), indicative of a substantial open state population, very similar to other measurements of HtpG at pH 7.5 (Krukenberg et al., 2009b). Subsequent addition of Δ131Δ shows a conformational change; a concentration series with increasing Δ131Δ results in a systematic contraction in the P(r) distribution (Figure 2b). These changes reflect a transition to more compact Hsp90 conformations that saturates near a 1:1 stoichiometry (Hsp90 dimer: Δ131Δ). The Guinier analysis of low-Q scattering is linear for all concentrations measured (data not shown), indicating no significant aggregation.
Similar Δ131Δ-induced contractions are observed in the P(r) distributions for the yeast and human Hsp90 homologs (Supplemental Figure 1b, c). Their contractions are smaller than for HtpG, which is consistent with previous SAXS and electron microscopy measurements that show the yeast and human homologs more strongly favor the open state relative to HtpG in the absence of nucleotide (Southworth and Agard, 2008). The substrate-induced conformational change is specific to unfolded staphylococcal nuclease (SN), addition of folded wild-type SN results in no significant conformational change and the scattering of SN and HtpG are independent and additive (Supplemental Figure 2a, b).
The measured SAXS change in response to Δ131Δ (Figure 2b) has two contributions: (1) additional scattering from bound Δ131Δ and (2) altered scattering from chaperone contraction. As discussed below, we separated out these contributions by first determining the location of Δ131Δ on HtpG subdomains using ab-initio reconstructions and then applying structure-based fitting and rigid-body analysis to determine the chaperone conformation.
To locate Δ131Δ on HtpG we first examined a subdomain containing the N-terminal and middle domains (NM, residues 1-495). This monomeric fragment behaves as a rigid object and is significantly smaller than the HtpG dimer, increasing the relative signal from Δ131Δ scattering in the complex. We used the DAMMIN and MONSA programs(Svergun, 1999) to generate ab-initio molecular envelopes from the NM/Δ131Δ scattering curve (Supplemental Figure 2c), for use as a starting point for modeling scattering from the full length HtpG dimer. DAMMIN determines an envelope from exhaustive rearrangement and Monte-Carlo minimization of dummy atoms to match the experimentally-determined distance distribution. MONSA simultaneously fits scattering data of complexes and individual sub-components to determine the location of members within a complex. No initial model was used and yet the resulting DAMMIN volume for the NM domain alone matches well with the expected structure (upper left panel, Figure 3a).
The MONSA reconstruction shows that Δ131Δ (cyan volume, Figure 3a upper right) binds predominantly at the middle domain. Multiple repeated MONSA runs were very similar (Supplemental Figure 2d). The DAMMIN and MONSA envelopes for the NM/Δ131Δ complex are broadly consistent with small local differences (Supplemental Figure 2e); the additional information from simultaneous fitting used in MONSA provides higher resolution information versus DAMMIN. The conclusion that Δ131Δ binds at a terminal region of the NM domain is evident from the primary scattering data, which shows that the addition of Δ131Δ results in a relative increase in long range scattering distances (Supplemental Figure 2c). When the MONSA reconstruction is mapped back on the full-length HtpG structure, Δ131Δ is found to project between the monomer arms (Figure 3a). Although these reconstructions provide a starting point for analyzing the SAXS data from the full-length HtpG/Δ131Δ complex, it should be noted that the reconstructions are too low resolution to conclude whether Δ131Δ binding is restricted to the MD or whether additional contacts are made to the NTD.
To analyze the SAXS data from the full-length HtpG/Δ131Δ, we first examined whether any of the four dominant HtpG conformations (open, ATP, Grp94, and ‘V’-shaped) could satisfactorily describe the chaperone conformation upon binding Δ131Δ. For reference, previous studies showed that HtpG’s conformational ensemble could be determined by linear combination fitting of SAXS data with different structural states (Krukenberg et al., 2009a). In the absence of Δ131Δ and at pH7.5, HtpG adopts an 81/19% open/Grp94 equilibrium resulting in a good fit to the chaperone alone data (blue squares, Figure 3b) and is quantified by a low R-factor of 2.2% (see equation in Methods). Single HtpG conformations fit the HtpG/Δ131Δ scattering poorly (Supplemental Figure 3a), with R-factors of 23, 24, 17, and 17% for the open, ATP, Grp94, and ‘V’-shaped conformations, respectively. Inspection of Supplemental Figure 3a shows that combining only ATP, Grp94, and ‘V’-shaped conformations would fit the experimental data poorly due to the lack of long-range scattering distances above 120 Å. By contrast a 58%:42% linear combination of open:‘V’-shape provides a better description of the SAXS data (Supplemental Figure 3a) indicating that HtpG remains in a conformational equilibrium upon Δ131Δ binding, but now with the V-shaped conformer.
To account for scattering from bound Δ131Δ we generated structural models of HtpG: Δ131Δ complexes with Δ131Δ residues of SN attached to the HtpG middle domain, based on the MONSA reconstruction (see Methods). For the open and crystallographic ‘V’-shaped conformations this addition significantly decreased R-factors (from 23 to 13% for the open state and from 17 to 11% for the ‘V’-shaped conformation), whereas for the Grp94 and ATP conformations this addition resulted in higher R-factors. Indeed, a 51/49 linear combination of the open: Δ131Δ/’V’: Δ131Δ states fit the scattering data well (Figure 3b, R-factor of 3.8%). The relative contributions from the open and ‘V’-shaped states to the P(r) fit are shown in Supplemental Figure 3b. All SAXS fitting statistics are summarized in Supplemental Table 1. Although the MONSA analysis suggests Δ131Δ binds predominantly at the MD, as an additional check we generated structural models with Δ131Δ located at different NTD positions and confirmed with linear combination fitting with the open state that this positioning resulted in a poorer fit to the data (Supplemental Figure 3c).
The above results indicate that HtpG adapts its conformation by partially closing around Δ131Δ. To best determine the chaperone conformation in this state, we performed rigid-body analysis. Since the open/’V’-shaped/Grp94 conformations of HtpG differ primarily by the opening angle between the middle and C-terminal (MC) domains (Figure 1), the Grp94 conformation can be used as a starting point in a rigid-body minimization where the MC angle is systematically explored for an optimal fit to the scattering data. To ensure robustness we performed this analysis by simultaneously fitting four data sets with varying ratios of open and Δ131Δ-induced closure (see Methods). This analysis confirms that a ‘V’-shaped structure is favored, very similar in opening angle to the crystallographically-determined structure (comparison shown in Supplemental Figure 4c). The rigid-body model (RBM) in equilibrium with the open state (Figure 3c) results in a good fit with an R-factor of 3.4% (Figure 3b). The remaining three data sets used in the simultaneous fitting have similarly low R-factors (Supplemental Table 1; 1.9, 1.9, and 3.3%). Since the RBM does not allow for the NTD rotation observed in the ATP conformation, we also confirmed that an ATP: Δ131Δ/open: Δ131Δ combination does not fit the data (R-factor of 7.3%).
In the presence of AMPPNP Hsp90 undergoes a dramatic closure involving N-terminal dimerization (Figure 1). For HtpG at pH 7.5 previous studies showed that saturating AMPPNP only drives a partial population shift to the ATP conformation (Krukenberg et al., 2009a). Since the ATP conformation is marginally populated it is sensitive to whether Δ131Δ binds favorably; if Δ131Δ favors/disfavors binding the closed state then an increase/decrease in closure is expected. Indeed, in the presence of 10 mM AMPPNP we find that Δ131Δ increases closure, which results in a P(r) distribution very different from that observed under apo conditions (Figure 4a). To quantify the relative populations of different states, we again used linear combination fitting, utilizing the structural models based on the previous MONSA analysis (Figure 4b). This fitting shows an approximately two-fold increase in the ATP state population (from 39% to 70%, Supplemental Table 1).
Consistent with the above observations we find the binding affinity between HtpG and Δ131Δ is enhanced two-fold by AMPPNP (Figure 4c). This enhancement is blocked by a competitive inhibitor geldanamycin (GE); the small increase in Kd from GE is from the DMSO storage buffer, which alone increases the Kd. These results predict that the effect of AMPPNP on the binding affinity for Δ131Δ should be most pronounced at pH 9, where HtpG undergoes a full population shift from the most open state to the ATP conformation (Krukenberg et al., 2009a). Indeed, at pH 9 there is a 5-fold increase in binding affinity for Δ131Δ upon addition of AMPPNP. Under both apo and AMPPNP conditions binding is pH dependent over the neutral range (pH 6–9) suggesting the involvement of a histidine. There is a strong salt dependence to Δ131Δ binding Kd and the resulting conformational equilibrium in HtpG: Δ131Δ (Supplemental Figure 4a, b), suggesting an electrostatic contribution. This conformational variation was useful in performing the rigid body minimization of multiple data sets as discussed earlier (see Methods).
In contrast to AMPPNP, there is a negligible influence of ADP on binding. Previous electron microscopy measurements have identified a compact HtpG conformation in the presence of ADP (Shiau et al., 2006; Southworth and Agard, 2008), yet this state is only transiently populated under SAXS experimental conditions (Krukenberg et al., 2008). Our results therefore suggest that under our experimental conditions, the ADP state is insufficiently populated to affect the bulk binding of Δ131Δ.
Since Δ131Δ affects the apo/ATP equilibrium it either accelerates closure or slows reopening. We tested this prediction with FRET since the Buchner and Hugel labs have showed that the open/ATP kinetics can be monitored this way (Hessling et al., 2009; Mickler et al., 2009). Following their work, we generated heterodimers of HtpG labeled with Alexafluor 647 and Alexafluor 555 at positions 62 and 341 on opposite monomers (see Methods). FRET measurements were performed at pH 9 because under these condition HtpG undergoes a complete open/ATP population shift (Krukenberg et al., 2009a). Indeed, under apo conditions there is minimal FRET whereas after an extended incubation with AMPPNP, there is significant FRET as indicated by an increase/decrease in acceptor/donor fluorescence (Figure 5a).
The large change in acceptor fluorescence at 664 nm provides a sensitive assay for closure kinetics. As shown in Figure 5b (red circles), upon addition of AMPPNP there is a slow increase in acceptor fluorescence, with single exponential kinetics similar to those measured with the yeast Hsp90 homolog (Hessling et al., 2009). When Δ131Δ is incubated with HtpG prior to addition of AMPPNP, nucleotide-driven closure is accelerated five-fold (Figure 5b, green squares), similar to the affinity enhancement measured by anisotropy at pH 9. As a control, we tested the influence of BSA and the folded wild-type SN (both at 50 μM, the same concentration that was used with Δ131Δ) on closure kinetics and found no significant change (not shown). In contrast to closure kinetics, Δ131Δ has no effect on reopening from the closed state (Supplemental Figure 5a). The closure acceleration by Δ131Δ implies that hydrolysis should also be accelerated, which we tested with a previously described assay (see Methods). At 50 μM Δ131Δ the hydrolysis rate is increased four-fold (Supplemental Figure 5b). The increase in hydrolysis is specific to Δ131Δ’s influence on HtpG, as this increase can be abolished by 200 μM radicicol.
The SAXS modeling and anisotropy data suggest a single bound Δ131Δ per HtpG dimer. This predicts a non-cooperative concentration dependence of the Δ131Δ-induced closure acceleration, in contrast to a two-site cooperative model in which an initial quadratic dependence on Δ131Δ concentration would be expected. Indeed, the Δ131Δ-induced closure acceleration (Figure 5c) has a simple rectangular hyperbolic concentration dependence (solid line).
A significant advantage of using Δ131Δ as a model system is that it is amenable to NMR measurements. To gain a higher resolution picture of the interaction site on Δ131Δ, we measured the Δ131Δ HSQC with and without HtpG. The original assignment of Δ131Δ was performed at low pH, high temperature and the absence of buffer, salt and magnesium chloride (Alexandrescu and Shortle, 1994), under which conditions HtpG is not stable. However, the majority of the peaks remain and at similar chemical shifts at pH 6.0, 25 mM MES, 25 mM KCl and 5 mM MgCl2 (Figure 6a, left panel). The transferred Δ131Δ assignments under these conditions were confirmed with HNCA and CBCA(CO)NH measurements on double labeled sample. Of the Δ131Δ residues of Δ131Δ, 41 residues could be unambiguously assigned allowing us to monitor their response to HtpG.
A non-specific binding interaction between to the entire Δ131Δ and HtpG would result in a loss of signal uniformly across the molecule, however at stoichiometic concentrations of HtpG and Δ131Δ we observe that a subset of peaks disappear while others have reduced intensity (Figure 6a, right panel). The fractional loss of peak height across different positions of the Δ131Δ sequence is HtpG concentration dependent and shows a clear trend where locations near Δ131Δ residue 100 are highly impacted by the addition of HtpG whereas more distant positions are less affected (Figure 6b). The region from residues 85–110 is completely broadened at stoichiometric concentrations of Hsp90. There is a roughly monotonic increase in peak height at increasing sequence distance away from the region that loses complete peak intensity, which suggests that chain mobility restricted by HtpG binding is relieved with increasing chain length from the binding site due to lack of structure in Δ131Δ. No new peaks were observed upon addition of HtpG, as expected for their complete broadening due to the slow tumbling of the interface region of Δ131Δ in complex with HtpG.
Significantly, the region with maximal peak loss (residues 85–110) has been identified as the most structured region within the globally unfolded protein. In particular, the helix between residues 97–107 has been shown to have significant structure as well as two turns located at residues 83–86 and 94–97 (Alexandrescu et al., 1994; Alexandrescu and Shortle, 1994; Wang and Shortle, 1995). Relaxation measurements have shown that this region has high order parameters and positive NOEs again indicative of significant structure (Alexandrescu and Shortle, 1994; Ohnishi and Shortle, 2003).
Hsp90:client interactions have proven difficult to study in-vitro likely because the chaperone favors interactions with partially folded or metastable client states that are only transiently populated. Here our aim has been to test the use of a model system of non-native states for probing Hsp90:substrate interactions. Our choice of system was guided by the fact that although globally unfolded, Δ131Δ has significant residual structure, is only marginally unstable, but is soluble and non-aggregating even at high concentrations. Using a combination of SAXS, FRET, fluorescence anisotropy and NMR, we have found that HtpG binds a specific region of Δ131Δ and this binding results in large-scale conformational and functional changes to the chaperone. These findings reveal basic steps in the Hsp90: Δ131Δ nucleotide cycle (Figure 7).
Our SAXS measurements and modeling suggest that under apo conditions HtpG adapts its conformation to Δ131Δ by a partial closure (Figure 3c). This structural analysis was aided by localizing Δ131Δ with ab-initio reconstructions prior to structure-based fitting and rigid-body analysis. The SAXS fitting on full-length HtpG shows that upon binding Δ131Δ the chaperone adopts an equilibrium between a ‘V’-shaped conformation and a fully open state (Figure 3c, Supplemental Table 1), indicating that Hsp90 maintains significant flexibility even after substrate loading. The residual flexibility suggests that Hsp90 could accommodate other cochaperones or binding partners within the loaded conformation. Also, to advance through the nucleotide cycle, Hsp90 must undergo large conformational changes requiring significant structural plasticity to reach the ATP state. While our SAXS modeling suggests that Δ131Δ remains bound to roughly the same region on HtpG, it is likely that concomitant with HtpG closure, there is some alteration in Δ131Δ:HtpG interactions and perhaps in Δ131Δ structure.
The apo, substrate-bound conformation of Hsp90 has a significant impact on the kinetics of the nucleotide cycle. Following previous work (Hessling et al., 2009; Mickler et al., 2009), we used kinetic FRET measurements to show that closure to the ATP state is significantly accelerated by Δ131Δ (Figure 5). This closure acceleration is paralleled by an ATPase acceleration (Supplemental Figure 5b), similar to reports of ATPase stimulation of human Hsp90 by the ligand binding domain of the glucocorticoid receptor (McLaughlin et al., 2002). Our findings suggest that closure is rate limiting in ATP hydrolysis by Hsp90, and that client binding activates the chaperone by lowering this rate-limiting conformational barrier. The coupling between client binding, Hsp90 conformational changes and subsequent ATP hydrolysis, suggests a simple mechanism by which Hsp90 restricts unnecessary ATP utilization while maximizing efficiency of client activation.
The ATP state transition involves numerous structural changes: (i) ATP binding restructures an N-terminal helical region that makes cross-monomer contacts, (ii) dramatically changes the NTD/MD orientation leading to an interaction between a highly conserved arginine (residue 336 in HtpG) and the ATP γ-phosphate, and (iii) is associated with the release of a β-strand that is swapped across monomers stabilizing N-terminal dimerization, however it is not known which of these processes (or others) are rate-limiting. Indeed, our SAXS measurements are too low resolution to conclude whether Δ131Δ contacts are restricted to a single monomer or whether Δ131Δ-induced closure is driven by cross monomer contacts. A detailed study is needed to address these points.
The results with Δ131Δ suggest that Hsp90’s conformational plasticity is functionally important and may allow it to adapt to structurally diverse substrates, which catalyze further structural changes that lead to ATP hydrolysis. This implies that Hsp90’s flexibility should be conserved, which is indeed true for the homologs investigated by SAXS (HtpG, Hsc82, hHsp90a, Grp94 and TRAP, (Krukenberg et al., 2009a) and unpublished observations). For bacterial, yeast and human Hsp90, detailed electron microscopy measurements have shown that a three-state apo-ATP-ADP conformational cycle is conserved but that the equilibria between states is species-specific (Southworth and Agard, 2008). This result suggested that the Hsp90 conformational equilibrium is tuned to the specific substrate/cochaperone requirements of each organism. Indeed, Δ131Δ binds to the bacterial, yeast, and human Hsp90 homologs and affects their conformations however the relative magnitude of these structural changes are species-specific (Supplemental Figure 1).
The conformational diversity and large structure of Hsp90 suggests that it can provide a combinatorial set of binding surfaces and conformations for interacting with structurally diverse substrates. An electron microscopy reconstruction of an Hsp90-Cdc37-Cdk4 complex (Vaughan et al., 2006) shows that Hsp90 adopts a more closed conformation in complex with the kinase-cochaperone complex than we observe with Δ131Δ. This observation suggests that different substrates and cochaperone complexes can be accommodated by different Hsp90 conformations and possibly have different nucleotide cycle dependences.
Our NMR measurements suggest that Hsp90 selectively interacts with a region of Δ131Δ (residues ~ 85–110) that has been shown to have significant structure despite the fact that Δ131Δ is globally unfolded (Alexandrescu et al., 1994; Alexandrescu and Shortle, 1994; Ohnishi and Shortle, 2003; Wang and Shortle, 1995). Hsp90 is often referred to as operating in later stages of client folding, consistent with this finding. Future studies will be required to reveal whether binding changes the structure within this region, which specific elements of Hsp90 are involved in binding, and the impact of nucleotide-induced conformational changes in Hsp90. Inspection of the fractional peak height distribution in Figure 6b also shows decreased peak heights towards the C-terminus of Δ131Δ, possibly indicating a second binding site. Clearly, although SAXS measurements are ideal for characterizing the flexible Hsp90 conformation and the influence of Δ131Δ, high-resolution measurements are needed to elucidate these molecular details.
The expression and purification of Hsp90 homologs and variants as well as Δ131Δ has been described previously (Alexandrescu et al., 1994; Cunningham et al., 2008; Krukenberg et al., 2008; Southworth and Agard, 2008). SAXS measurements were performed at the SIBYLS beamline (12.3.1) at the Advanced Light Source in Berkeley. Scattering was measured with 0.5, 2, and 5 second integrations and buffer subtracted. The radial intensity was converted to the P(r) representation with the GNOM program (Svergun, 1991). Dmax cut-off values were selected manually to achieve smooth tails at high distance values.
DAMMIN (Svergun, 1999) reconstructions used spherical starting models with no symmetry constraints. MONSA (Svergun, 1999) reconstructions used simultaneous fitting on multiple data sets (the NM domain and the NM/Δ131Δ complex) keeping the Δ131Δ Rg fixed at 18 Å. The SUBCOMP program was used to align structures into DAMMIN and MONSA reconstructions. Δ131Δ residues of an SN crystal structure (1STN) was used to approximate scattering from Δ131Δ by appending it to the middle domains of the crystallographically solved ‘V’-shaped conformation (2IOQ) as well as the open, Grp94, and ATP conformations. This was achieved by aligning HtpG conformations to the NM/Δ131Δ MONSA envelope and positioning SN to avoid steric clash with HtpG. All HtpG conformations are available for download (http://www.msg.ucsf.edu/agard/).
Structure-based fitting and rigid-body analysis of SAXS data was performed with the PRFIT program, described previously (Krukenberg et al., 2009b). The quality of fit is quantified by an R-factor
For rigid-body analysis, the N-terminal and middle domains of HtpG were allowed to pivot against the C-terminal domain at residue 500 using rigid-body analysis described previously (Krukenberg et al., 2009a; Krukenberg et al., 2008; Krukenberg et al., 2009b). General three-axis rotational motion around this pivot started with a course step size of 10° and then was refined with 1° steps. The attached Δ131Δ residues of SN were included on a single monomer arm. The rigid-body fitting was simultaneously performed on multiple data sets with varying levels of Δ131Δ-induced closure. Specifically, we used a salt series that titrates the binding affinity and degree of closure (Supplemental Figures 4a, b) and a previously characterized HtpG variant (H446K) that favors the Grp94 conformation (Krukenberg et al., 2009a). If the ‘V’-shaped conformation is a robust solution, then the rigid-body search will independently identify this state in these diverse data sets. To determine the robustness of the SAXS modeling, limiting cases with zero or two bound Δ131Δ were investigating and indeed only modestly affect fitted parameters (range shown in brackets in Supplemental Table 1) as expected given the small size of Δ131Δ relative to HtpG.
Fluorescence anisotropy was measured on a Jobin Yvon fluorometer with excitation and emission monochromator slits both set to 5 nm, an integration time of 2 seconds, and excitation/emission wavelengths of 340/480 nm. A five-fold molar excess of IAEDANS (Invitrogen) was used to label a cysteine variant of Δ131Δ, K16C, at room temperature for one hour and free dye was removed by extensive dialysis. Measurements with nucleotides and geldanamycin were pre-equilibrated for at least 20 minutes. Binding stoichiometry measurements were performed at 50 μM Δ131Δ with only a small concentration, 500 nM, labeled with IAEDANS.
FRET measurements were performed on the same fluorometer. A five-fold molar excess of dye (Alexafluor 647 and Alexafluor 555, Invitrogen) was incubated with HtpG variants E62C and D341C for three hours at room temperature. HtpG has no native cysteines. The reaction was quenched with 2-fold excess of BME over dye. Free dye was separated by extensive dialysis and G50 size exclusion chromatography (Roche). HtpG heterodimers (250 nM at a 1:1 stoichiometry) were formed by incubation for 30 minutes at 30 °C. FRET measurements had 250 nM of heterodimer at a 1:1 stiochiometry. Closure was initiated by manual mixing of 5 mM AMPPNP and reopening was initiated by subsequent addition of 50 mM ADP. The excitation and emission slits were 2 and 3 nm, with an integration time of 0.3 s. To control for photobleaching and burst phase fluorescence, separate reactions without nucleotide were monitored and subtracted from nucleotide-containing samples.
ATP hydrolysis was measured by a phosphate release assay described previously (Cunningham et al., 2008). Briefly, 1 μM of HtpG dimer was incubated with 50 μM Δ131Δ at pH 9, 50 mM KCl, 5 mM MgCl2 at 22 °C and assayed for radiolabeled 32Pi release. 5mM ATP was used with 0.8 pM of radiolabled ATP. Negligible background ATPase was confirmed by addition of 200 μM radicicol. Free Pi was separated from ATP and ADP by thin layer chromatography. Spot imaging was performed on a Typhoon Imager (GE Healthcare) and was quantified with the ImageQuant program (GE Healthcare).
Isotopically labeled Δ131Δ was produced by a 10 mL overnight starter culture, which was spun down and washed in M9 minimal growth medium, and then resuspended in 1 L of M9 media supplemented with 1 g/L 15N ammonium chloride and 0.5 g/L isogrow growth supplement (Sigma). Uniformly labeled 13C, 15N samples were derived similarly with 1 g of 13C glucose per liter of media. HSQC measurements were performed on a Bruker Avance800 with Δ131Δ concentration of 200 μM. HNCA and CBCA(CO)NH measurements were performed on an Avance500. Data was processed with NMRPipe (Delaglio et al., 1995) and peak height analysis was performed with the ccpn analysis software (http://www.ccpn.ac.uk). The Δ131Δ assignment was transferred from a previously published study of Δ131Δ (Alexandrescu and Shortle, 1994) and confirmed with the HNCA and CBCA(CO)NH experiments.
We thank David Shortle for kindly proving the Δ131Δ construct and Greg Hura for help with SAXS data collection and Mark Kelly for help with NMR. Funding for this project was provided by the Howard Hughes Medical Institute. TOS was supported by a Damon Runyon Cancer Research Foundation fellowship. Many thanks to members of the Agard lab for helpful discussions.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.