A novel N-terminal extension in mitochondrial TRAP1 serves as a thermal regulator of chaperone activity.

Hsp90 is a conserved chaperone that facilitates protein homeostasis. Our crystal structure of the mitochondrial Hsp90, TRAP1, revealed an extension of the N-terminal β-strand previously shown to cross between protomers in the closed state. In this study, we address the regulatory function of this extension or 'strap' and demonstrate its responsibility for an unusual temperature dependence in ATPase rates. This dependence is a consequence of a thermally sensitive kinetic barrier between the apo 'open' and ATP-bound 'closed' conformations. The strap stabilizes the closed state through trans-protomer interactions. Displacement of cis-protomer contacts from the apo state is rate-limiting for closure and ATP hydrolysis. Strap release is coupled to rotation of the N-terminal domain and dynamics of the nucleotide binding pocket lid. The strap is conserved in higher eukaryotes but absent from yeast and prokaryotes suggesting its role as a thermal and kinetic regulator, adapting Hsp90s to the demands of unique cellular and organismal environments.

Introduction

Hsp90 is a highly conserved molecular chaperone essential for protein and cellular homeostasis. Although molecular chaperones generally promote protein folding and prevent aggregation, Hsp90 is unique in that it interacts with substrate (‘client’) proteins that are already in a semi-folded state to facilitate downstream protein–protein interactions and promote client function in diverse biological pathways (Jakob et al., 1995; Taipale et al., 2012). Hsp90 interacts with nearly 10% of the eukaryotic proteome (Zhao et al., 2005), and its client proteins vary significantly in sequence, structure, and size (Echeverria et al., 2011). In most eukaryotes, there are four different Hsp90 homologs: Hsp90α and Hsp90β in the cytoplasm, Grp94 in the endoplasmic reticulum (ER), and TRAP1 in mitochondria, with each homolog contributing unique biological functions (Chen et al., 2006; Johnson, 2012). Deregulation of Hsp90 protein levels and function has been linked to multiple human diseases and for this reason Hsp90 is a target for biochemical characterization, structural studies, and drug discovery (Luo et al., 2010; Taipale et al., 2010). Despite such importance, little is known about the biochemical characteristics that regulate client interaction and specificity.

Hsp90 exists as a homodimer, with each protomer consisting of three major domains. The N-terminal domain (NTD) binds to ATP, the C-terminal domain (CTD) provides a dimerization interface between protomers, and the middle domain (MD) provides a stabilizing γ-phosphate contact to help facilitate ATP hydrolysis (Cunningham et al., 2012). Together with the CTD, the MD has been shown to aid in the formation of client interactions (Street et al., 2011, 2012; Genest et al., 2013). Large, rigid body motions about each of the domain interfaces give rise to an ensemble of remarkably diverse conformational states that dictate the functional Hsp90 cycle (Ali et al., 2006; Shiau et al., 2006; Dollins et al., 2007; Southworth and Agard, 2008; Lavery et al., 2014) (Video 1) and are linked to client maturation in vivo (Panaretou et al., 1998). Work from numerous labs has demonstrated conservation of the underlying conformational cycle and mechanism; however, each Hsp90 homolog has a distinct conformational equilibrium and catalytic rate (Panaretou et al., 1998; Richter et al., 2008; Southworth and Agard, 2008). Binding of ATP to the NTD nucleotide-binding pocket ultimately leads to stabilization of an NTD-dimerized state. Key steps in this transformation include ATP binding, closure of a mobile structure (lid) over the nucleotide, and a subsequent 90o rotation of the NTD relative to the MD (Krukenberg et al., 2011). Dimer closure is the rate-limiting step for Hsp90 ATPase activity and mutations that either subtly increase or decrease ATPase rates compromise viability in yeast (Nathan and Lindquist, 1995; Hessling et al., 2009). However, our understanding of the sequence of events that regulate these structural rearrangements is limited.

Video 1.

Conformational dynamics of the Hsp90 cycle.

A morph between known conformations throughout the activity cycle of Hsp90 (PDB codes with no order dictated: 2O1V, 2CG9, 2IOP, 2IOQ, 4IPE, 4IVG).

DOI: http://dx.doi.org/10.7554/eLife.03487.003

Recently, we solved a series of full-length crystal structures of TRAP1 bound to different ATP analogs (Lavery et al., 2014), providing new insights into the structure, dynamics, and mechanism of Hsp90. Of particular note was the marked asymmetry between protomers of the homodimer, primarily at the interface between the MD and CTD. This asymmetry was sampled in solution, proved essential for catalytic turnover, and provided a new model for coupling the energy of ATP hydrolysis to client remodeling. A second feature highlighted by the TRAP1 crystal structure was an ordered 14-residue extension (out of 26 total additional residues) of the N-terminal β-strand previously shown to cross over (‘swap’) between protomers in the closed state (Ali et al., 2006). While absent in yeast and bacteria, this extension, or ‘strap,’ is found in most eukaryotic Hsp90 proteins including the cytosolic and organellar forms (Chen et al., 2006) and can extend for as many as 122 residues as recently found in a splice variant of Hsp90α in higher eukaryotes (Tripathi and Obermann, 2013). Structure based point mutations and complete removal of the TRAP1 strap (Δstrap) resulted in a sixfold increase in ATPase activity in zebrafish TRAP1 (zTRAP1) (Lavery et al., 2014), evidence that the strap plays a regulatory role. Similarly, deletion of the strap in Grp94 (residues 22-72, referred to as the ‘pre-N domain’) resulted in a fivefold increase in ATPase (Dollins et al., 2007), indicative of a conserved regulatory role, although the mechanism remains unclear.

In this study, we explore the conformational cycle of TRAP1 and demonstrate that the strap is responsible for a large thermal barrier between the apo (open) and ATP bound (closed) states. Using negative-stain electron microscopy (EM) and Small-Angle X-ray Scattering (SAXS), we demonstrate that removal of the strap results in a profound reduction in the temperature sensitivity observed in multiple TRAP1 homologs, indicating that the strap is responsible for this unique behavior. Additionally, we develop fluorescence resonance energy transfer (FRET) and continuous-wave EPR (CW-EPR) assays to show that the strap regulates the rate-limiting conformational transitions that precede NTD dimerization, including NTD rotation and lid closure over the ATP-binding pocket. These results indicate that the strap must stabilize both the apo state and the closed state, providing a unique evolutionary strategy for modulating different phases of the kinetic landscape and optimizing in vivo function of diverse Hsp90s.

Results

A temperature-sensitive kinetic barrier limits the conformational transition from apo to the closed state in TRAP1

With all previously studied Hsp90s, incubation with slowly- or non-hydrolyzable ATP analogs favors accumulation of a closed, NTD-dimerized state. However, the extent of closed-state accumulated and the rate of closure differentiated the Hsp90s with the individual values positively correlating with the ATP hydrolysis rate (Richter et al., 2008; Southworth and Agard, 2008; Hessling et al., 2009). Specifically, we used EM to demonstrate that the large variability in observed ATPase rates of cytosolic bacterial (bHsp90), yeast (yHsp90), and human Hsp90 (hHsp90), directly correlated with the ability of each homolog to reach a closed conformation in the presence of non-hydrolyzable ATP (AMPPNP) (Southworth and Agard, 2008). Here, negative-stain EM was again used to monitor the ability of human TRAP1 (hTRAP1) to transition from the apo conformation to the closed conformation in the presence of AMPPNP. While hTRAP1 has an ATPase rate similar to the E. coli Hsp90 (∼0.5 min−1) (Cunningham et al., 2012), surprisingly hTRAP1 remained in the open conformation despite incubation with saturating AMPPNP (Figure 1A). Noting that the discrepancy might be related to different incubation temperatures between the two experiments, we monitored the ability of hTRAP1:AMPPNP to close as a function of temperature. After 1 hr (Figure 1A) or overnight (Figure 1B) incubation at room temperature (RT, ∼23°C) hTRAP1 remained in the open state. However, after a single hour of incubation at increasing temperatures, the closed state was increasingly populated (Figure 1A). These results correlate well with the temperature sensitive steady-state hydrolysis rates of hTRAP1 that increases by nearly 200-fold between 25°C and 55°C (Leskovar et al., 2008) and are consistent with closure being rate-limiting for hydrolysis. Importantly, the equilibrium reached at each temperature (Figure 1A) remains fixed after subsequent incubation at RT overnight (Figure 1B). These data suggest both a large, unusually thermally sensitive kinetic barrier to closure and a highly stable closed state.

Figure 1.

A temperature-dependent barrier separates the apo and closed state of TRAP1.

(A) Negative stain electron microscopy (EM) images of hTRAP1 in the presence of AMPPNP at increasing temperatures for 1 hr. While the population at equilibrium appears to remain in an apo conformation at room temperature (RT), conversion to the closed state appears to be intermediate at 30°C and nearly complete at 37°C and 42°C. (B) Negative stain EM images of reactions incubated at 23°C and 37°C from A after returning the sample to RT and incubating overnight. Both populations remain apo and closed (respectively) demonstrating the large kinetic barrier that limits the conformational transition from apo to the closed state. Scale bar is 100 nm.

DOI: http://dx.doi.org/10.7554/eLife.03487.004

The N-terminal strap is responsible for TRAP1 thermal sensitivity

To better measure the equilibrium between conformational states as a function of temperature, we used SAXS, which can directly quantify the solution distribution of open and closed states (Krukenberg et al., 2008). As demonstrated by a shift towards a more compact pair-wise inter-atomic distance distribution, P(r), there was a strong correlation between temperature and dimer closure, (Figure 2A). By fitting the distributions as a linear combination of open and closed states, the fraction of closed state can be accurately estimated (‘Materials and methods’). After 1 hr at 20°C only 0.4% of the molecules have reached the closed conformation, while at 43°C roughly 84% of the molecules are closed (Figure 2C and Table 1). In agreement with our EM data, the equilibrium does not revert back to the apo state when the temperature is lowered (Figure 2D). Interestingly, TRAP1 from zebrafish (zTRAP1) displays a shifted temperature-dependent conformational equilibrium that correlates with its higher basal ATPase rate (Figure 2 and Table 2) and the lower physiological temperature of zebrafish (∼29°C).

Figure 2.

A large energy barrier to the closed state is modulated by the NTD-strap.

(A) SAXS distributions at equilibrium for hTRAP1 (left) and zTRAP1 (right) (84% identical to hTRAP1) in apo and in the presence of saturating AMPPNP at indicated temperatures for 1 hr. The closed-state population substantially increases at and above 36°C for hTRAP1 while zTRAP1 maintains a higher level of % closed at even lower temperatures, consistent with the differences in physiological temperatures of the two species. (B) SAXS distributions of Δstrap in matching conditions from A showing that removal of the strap mitigates the temperature-dependent barrier between the apo and closed states. (C) Quantification of percent closed for both TRAP1 species ± the strap region. Apparent is the different temperature dependence of hTRAP1 and zTRAP1 and the loss of temperature response of the chaperone in the case of Δstrap. (D) A plot of percent closed state verses temperature of WT hTRAP1 (left) and Δstrap hTRAP1 after closure has completed at each given temperature (solid bars as in (C). These samples were then cooled for 2 hr at 20°C (stripped bars). The data suggest a highly stable closed state.

DOI: http://dx.doi.org/10.7554/eLife.03487.005

Table 1.

Quantification of percent closed using SAXS data for both TRAP1 species ± the strap

DOI: http://dx.doi.org/10.7554/eLife.03487.006

Temperature (°C)	Protein	% Closed state	R
20	WT hTRAP1	0.5	0.042
23	WT hTRAP1	2	0.042
30	WT hTRAP1	31	0.024
32	WT hTRAP1	41	0.019
36	WT hTRAP1	74	0.011
40	WT hTRAP1	83	0.010
43	WT hTRAP1	84	0.012
20	WT zTRAP1	36	0.015
23	WT zTRAP1	48	0.011
30	WT zTRAP1	75	0.027
32	WT zTRAP1	81	0.033
36	WT zTRAP1	80	0.030
43	WT zTRAP1	69	0.016
20	hTRAP1 Δstrap	66	0.015
23	hTRAP1 Δstrap	68	0.014
30	hTRAP1 Δstrap	69	0.014
32	hTRAP1 Δstrap	68	0.015
36	hTRAP1 Δstrap	67	0.018
40	hTRAP1 Δstrap	69	0.016
43	hTRAP1 Δstrap	71	0.015
20	zTRAP1 Δstrap	60	0.016
23	zTRAP1 Δstrap	64	0.014
30	zTRAP1 Δstrap	61	0.015
32	zTRAP1 Δstrap	62	0.014
36	zTRAP1 Δstrap	55	0.016
43	zTRAP1 Δstrap	76	0.011

Table 2.

Steady-state ATP hydrolysis rates at temperatures and buffer conditions of assay specified (i.e., EPR is under EPR buffer and temperature conditions). If not noted (top four reactions), conditions are the same as reference (Lavery et al., 2014)

DOI: http://dx.doi.org/10.7554/eLife.03487.007

Protein	zTRAP1 ATPase (min−1)	hTRAP1 ATPase (min−1)
WT (30°C)	1.36 ± 0.12	0.463 ± 0.003
salt bridge point mutants (30°C) (Lavery et al., 2014)	(E-A) 3.57 ± 0.62 (H-A) 5.08 ± 0.90
Δstrap (30°C)	5.84 ± 0.47	13.3 ± 0.5
Δ60-69 (30°C)		0.47 ± 0.02
WT FRET (30°C)		0.21 ± 0.0.01
Δstrap FRET (30°C)		11.9 ± 2.1
CFree WT EPR (23°C)	0.88 ± 0.05
CFree Δstrap EPR (23°C)	5.65 ± 0.22
CFree WT (Inter FRET) (30°C)		0.79 ± 0.0.03
CFree Δstrap (Inter FRET) (30oC)		7.6 ± 0.36
CFree WT (Intra FRET) (30°C)		0.35 ± 0.0.01

Red text indicates WT or strap truncated protein with native cysteine and label free, while Blue indicates labeled protein used in FRET and EPR experiments (each in indicated buffer conditions). EPR samples are cysteine free except for the desired probe position and are spin-labeled. Inter FRET and Intra FRET samples are cysteine free except for the desired probe position and are labeled with Alexa Fluor dyes (Life Technologies, see ‘Materials and methods’). Note that ‘Intra FRET’ refers to both probe positions on the same promoter, whereas ‘Inter FRET’ refers to one probe position per promoter. Errors represent the standard deviation of triplicate experiments.

Previous studies by Richter et al. had demonstrated that removal of the initial β-strand in yHsp90 (corresponding to post-strap residues in TRAP1) increased ATPase activity and facilitated N-terminal dimerization (Frey et al., 2007). The ordered strap extension observed in the zTRAP1 structure is also kinetically important, as Δstrap (deletion of zTRAP1 residues 73–100) and a single point mutant aimed at disrupting a conserved, stabilizing salt bridge at the beginning of the strap, accelerated hydrolysis by sixfold and fourfold, respectively. Together, these raised the possibility that the strap might also be responsible for the unusual temperature-regulated energy landscape observed in TRAP1 homologs.

As a first step, we show that in hTRAP1, strap removal (lacking residues 60–85) has an even more profound impact on ATPase activity (∼30-fold) than on zTRAP1 (Figure 3, Table 2). The larger increase in ATPase activity for hTRAP1 correlates with the more significant temperature dependence (Figure 2C) and thus a higher kinetic barrier for hTRAP1 at the experimental temperature of 30°C. Notably, a smaller truncation lacking residues 60–69, that preserved the conserved His71:Glu142 salt-bridge in hTRAP1, did not have an effect on ATPase activity (Figure 3, Table 2).

Figure 3.

NTD-strap regulates ATP hydrolysis rates.

WT and strap mutants for hTRAP1. Removal of the strap (Δstrap) results in a ∼30-fold increase in ATPase rate, while truncations before the previously reported salt bridge contact Δ60–69 (Lavery et al., 2014) show no change in activity. Average steady-state hydrolysis rates (min−1) above each bar, standard deviation of triplicate measurements can be found in Table 2.

DOI: http://dx.doi.org/10.7554/eLife.03487.008

Strikingly, SAXS revealed that at every temperature examined in both hTRAP1- and zTRAP1-Δstrap immediately equilibrated to a closed conformation after adding AMPPNP, indicating a loss of temperature sensitivity (at least at temperatures ≥20°C) (Figure 2B). This indicates that beyond a role in stabilizing the closed conformation through trans-protomer interactions, the strap must also be involved in apo interactions that inhibit a transition towards the closed state. These data together with previously solved crystal structures of other Hsp90 N-terminal domains displaying cis-contacts of the initial β-strand suggest that the strap likely makes equivalent contacts with the same NTD (cis) in the apo state that it forms with the trans-NTD in the closed state (Shiau et al., 2006; Dollins et al., 2007; Li et al., 2012).

NTD-strap limits closure rate by regulating NTD rotation and lid dynamics

The above results predict that the rate of closure should be proportional to temperature. Fluorescence Resonance Energy Transfer (FRET) provides a more convenient method than SAXS to directly measure the rate of closure (Hessling et al., 2009; Mickler et al., 2009; Street et al., 2011). Two FRET constructs were designed to probe distinct aspects of the closure reaction, relying on a Cys-free version of hTRAP1 (Lavery et al., 2014). The first construct modeled from FRET positions previously designed for yHsp90 placed a single Cys residue on each protomer (E140C and E407C, ‘Inter FRET’) so as to give an increase in FRET upon closure (Hessling et al., 2009). The second construct modeled on previous work with bHsp90 (Street et al., 2012), adds two Cys residues to a single protomer (S133C and E407C, ‘Intra FRET’), and is designed to track the ∼90o NTD rotation (relative to the MD) that occurs upon closure. After forming heterodimers, closure reactions were initiated with AMPPNP over a temperature range mirroring our SAXS experiments and the change in FRET was monitored. Pre- and post-reaction fluorescent scans showed a predicted FRET change indicative of closure for each FRET construct (Figure 4A). As expected, the rate of closure correlated with increasing temperature (Figure 4B, Figure 4—figure supplement 1, Table 3) for both dimer closure and NTD:MD rotation measurements. To measure the contribution of the strap to the kinetics of closure, we truncated the strap region of either one or both protomers in each FRET construct (although the dimeric Δstrap construct used to measure NTD rotation proved too unstable to obtain reliable data). In both cases, a large acceleration of closure was apparent (Figure 4C) with the largest acceleration (16-fold) observed for the double-strap deletion.

Figure 4.

The NTD-strap regulates closure rate of TRAP1.

(A) Steady-state FRET scans at 23°C for apo and AMPPNP reactions after closure with AMPPNP reached completion illustrating the anti-correlated change in FRET upon closure as measured by ‘dimer closure’ between protomers (left, Inter FRET) and rotation of the NTD from apo to the closed state within one protomer ‘NTD:MD Rotation’ (right, Intra FRET). (B) Temperature-dependent closure rates for WT hTRAP1 measured by both the dimer closure and NTD rotation FRET probes from A. Closure rates are comparable between these two sets of FRET probes as indicated in the table to the right. The predicted increase in rate at higher temperatures is apparent. (C) Closure at 30°C of WT compared to heterodimers lacking one or both NTD strap residues measured by dimer closure FRET (left) and NTD rotation FRET (right). Closure rates are found in the table for each experiment. (D) Temperature-dependent closure rates of Δstrap protein measured using the dimer closure probes from A (Inter FRET) illustrating both a rate acceleration and a dramatic loss of temperature dependence compared to WT (B, left panel). (E) Arrhenius plot of WT and Δstrap plotted using data from panels (B) (left) and (D). From the difference in activation energies Ea between WT and Δstrap, the strap contributes approximately 60% of the measured Ea for WT hTRAP1 (48.8 kcal/mol Ea for WT; 29 kcal/mol Δstrap). These data are consistent with the steady-state SAXS and ATPase and show that removal of the strap region lowers the energy barrier between apo and the closed state.

DOI: http://dx.doi.org/10.7554/eLife.03487.009

(A) Kinetics of FRET closure at lower temperatures (23°C, 30°C, 32°C) with fits shown for full measured curve. All reactions were taken to completion. (B) Data from Figure 4D plotted to focus on the smaller difference in closure rate for Δstrap at increasing temperatures.

DOI: http://dx.doi.org/10.7554/eLife.03487.010

Calculated Ea for each homolog is listed in figure legend parentheses.

DOI: http://dx.doi.org/10.7554/eLife.03487.011

Figure 4—figure supplement 1.

Alternative view of curve fits for Figure 4B,D.

(A) Kinetics of FRET closure at lower temperatures (23°C, 30°C, 32°C) with fits shown for full measured curve. All reactions were taken to completion. (B) Data from Figure 4D plotted to focus on the smaller difference in closure rate for Δstrap at increasing temperatures.

DOI: http://dx.doi.org/10.7554/eLife.03487.010

Table 3.

Kinetics of conformational changes as measured by FRET. Errors represent the standard deviation of triplicate experiments

DOI: http://dx.doi.org/10.7554/eLife.03487.012

Protein	Temperature (°C)	FRET probe position	Kclose (min−1)	Kreopen (min−1)	Khyd (min−1)
CFree WT hTRAP1 (Intra FRET)	23	S133C.E407C	0.0011 ± 0.00006
	30	S133C.E407C	0.0066 ± 0.00007
	32	S133C.E407C	0.021 ± 0.001
	36	S133C.E407C	0.073 ± 0.002
	42	S133C.E407C	0.36 ± 0.011
	30	S133C.E407C	0.073 ± 0.005
CFree WT hTRAP1 (Inter FRET)	23	E140C/ E407C	0.003
	30	E140C/ E407C	0.02 ± 0.002	0.00210 ± 0.00003
	32	E140C/ E407C	0.039
	36	E140C/ E407C	0.118
	42	E140C/ E407C	0.431
CFree Δstrap single (Inter FRET)	30	E140C/ E407C	0.21 ± 0.013
CFree Δstrap double (Inter FRET)	23	E140C/ E407C	0.073
	30	E140C/ E407C	0.31 ± 0.024	0.016 ± 0.003
	32	E140C/ E407C	0.456
	36	E140C/ E407C	0.853
	42	E140C/ E407C	1.335
*CFree WT hTRAP1 (Inter FRET) *ATP used	30	E140C/ E407C	0.42 ± 0.01		7.56 ± 0.99
	25	E140C/ E407C	0.19 ± 0.01 (0.003 ± 0.0001 s−1)		4.3 ± 0.2 (0.071 ± 0.004 s−1)
*CFree Δstrap double (Inter FRET) *ATP used	30	E140C/ E407C	1.5 ± 0.01		10.6 ± 0.4
	25	E140C/ E407C	0.92		7.57

Denotes ATP was used for closure. Relates to Figures 4,6 and Figure 6—figure supplement 1. All other reactions used AMPPNP as the ATP analog. Note that ‘Intra FRET’ in red refers to both probe positions on the same promoter, whereas ‘Inter FRET’ in blue refers to one probe position per promoter.

Figure 6—figure supplement 1.

ATP Hydrolysis at 25°C.

(A–B) Kinetic experiment designed to decouple ATP hydrolysis from the preceding closure step at 25°C outlined in Figure 6C–E. The measured rates for WT agree well with previous single turnover measurements (Leskovar et al., 2008), however, we find that decoupling the closure rate from hydrolysis results in a reassignment of the previous rates with closure being the slowest step. Kinetic rates for each are listed in the table inset.

DOI: http://dx.doi.org/10.7554/eLife.03487.017

A good way to quantitate the contribution of the strap to the thermal barrier is to measure closure rates as a function of temperature with and without the strap (Figure 4B,D and Figure 4—figure supplement 1) and to calculate the activation energy (Ea) towards closure (i.e., the temperature dependent barrier height). At every temperature sampled removing the strap results in an acceleration of closure compared to WT and an overall loss in temperature dependence (Figure 4D). Comparing the fold changes in closure rates (Table 3), we see the largest fold change at lower temperatures (23°C: 24-fold, 30°C: 16-fold, 32°C: 12-fold, 36°C: sevenfold, and 42°C: threefold). This increased impact at lower temperatures is readily evident in an Arrhenius plot calculated from the inter FRET experiments (Figure 4E). The resultant activation energies (Ea) taken from the slopes of these curves are 48.8 kcal/mol and 29 kcal/mol, for WT and Δstrap respectively. From the difference, the strap appears to be contributing ∼20 kcal/mol towards the Ea of WT hTRAP1, which we interpret as ∼ 20 kcal/mol of enthalpic stabilization of the open state. Our Ea for WT hTRAP1 is consistent with that measured previously under slightly different conditions (Leskovar et al., 2008), but is considerably higher than that calculated for other Hsp90 homologs (Figure 4—figure supplement 2) (Frey et al., 2007). As a control, we also measured steady-state ATPase rates on the labeled protein used for the FRET experiments. While these showed differences in absolute ATPase rates between 1.5 and fourfold compared with their unlabeled counterparts (Tables 2 and 3), the relative impact of strap deletion was consistent across experiments. Together, these data support a model in which the N-terminal strap limits closure by inhibiting the rotational movement of the NTD that is necessary to form the catalytically active closed state.

Figure 4—figure supplement 2.

Arrhenius plots for Hsp90 homologs plotted using data from reference (Frey et al., 2007).

Calculated Ea for each homolog is listed in figure legend parentheses.

DOI: http://dx.doi.org/10.7554/eLife.03487.011

To probe the underlying mechanism of the NTD-strap in the closure reaction, we sought to examine the relationship of the strap to the dynamics of the NTD lid (zTRAP1 residues 191–217) that closes over the ATP binding pocket; a mechanism conserved in many ATPases. Previous studies with yHsp90 have suggested a correlation between the ‘β-strand swap’ and dynamics of the NTD lid (Richter et al., 2006). In an open conformation and prior to nucleotide binding, the lid makes contacts with helix 1 (H1) (Richter et al., 2006; Shiau et al., 2006; Dollins et al., 2007; Li et al., 2012), while in the closed state the lid rotates to secure nucleotide via interactions at conserved sidechains (Ser193 and Ser195 in zTRAP1) inside the nucleotide binding pocket (Ali et al., 2006; Lavery et al., 2014) (Videos 2 and 3). This closed state lid conformation is incompatible with the NTD:MD apo state conformation as it would clash with the MD (Shiau et al., 2006; Dollins et al., 2007).

NTD-strap anti-correlated lid conformational changes.

A morph between two conformations of Hsp90, from the Apo state with cis-protomer interactions between NTD and strap, to the nucleotide bound closed state where the strap makes trans-protomer interactions. This morph demonstrates the significant number of contacts that are lost and then reformed to accommodate movement of the NTD to form the NTD-dimerization interface. (PDB codes 2IOQ, 4IVG, 4IPE).

DOI: http://dx.doi.org/10.7554/eLife.03487.013

Movement of the lid to accommodate the NTD-dimerization interface.

A morph between two conformations of yHsp90 NTD, either in the APO state or nucleotide bound closed conformation. This morph demonstrates the coordinated movement and changing contacts between both the β-strand (pink) and NTD lid (red) to facilitate the NTD-dimerization interface of a dimerized Hsp90 molecule. (PDB codes 4AS9, 2CG9).

DOI: http://dx.doi.org/10.7554/eLife.03487.014

To test whether the strap has a role in lid stabilization, we developed an electron paramagnetic resonance (EPR) spectroscopy assay to track lid mobility in the apo and closed states (‘Materials and methods’). A cys-free version of zTRAP1 with an Ala201Cys mutation allowed labeling of a fully accessible cysteine residue in the lid with N-(1-oxyl- 2,2,6,6-tetramethyl-4-piperidinyl)maleimide (MSL). We observed a small difference in ATPase activity with the MSL-labeled TRAP1 compared with ATPase rates measured using WT TRAP1 suggesting a minor labeling effect on steady-state catalytic turnover (∼1.3 fold). EPR spectra are sensitive to the rotational mobility of the attached MSL probe making it a useful reporter for changes in local conformational dynamics (Hubbell et al., 2000). EPR spectra of full-length zTRAP1 were recorded at 23°C and shown to be more mobile in the nucleotide bound state compared to the apo state (Figure 5A). Mobility of the lid as measured with EPR is consistent with apo structures showing low B-factors in this region due to significant contacts with H1 of the cis protomer (Richter et al., 2006; Shiau et al., 2006). Conversely, crystal structures of TRAP1 and other Hsp90 homologs bound to ATP analogs in the closed and dimerized conformation show that the lid folds over the nucleotide, has increased B-factors and lacks many of the stabilizing contacts with the N-terminal domain found in the apo state (Ali et al., 2006; Lavery et al., 2014). This is consistent with the mobile signature in the EPR observed for the closed conformation. Comparing apo state equilibrium measurements for WT and Δstrap shows little change upon strap deletion (Figure 5A). Fortunately EPR is sufficiently sensitive and the closure kinetics for TRAP1 are sufficiently slow, that it is possible to directly monitor changes in lid state over time. By plotting the change in normalized peak heights over time (‘Materials and methods’, Figure 5B), it is apparent that the amplitude changes for both the mobile and immobile peaks are well fit by a single exponential curve for each sample. From this, it is clear that the rate of change between states as monitored by lid mobility is much faster for the Δstrap sample than for WT. The fold difference between rates is on the order of changes in ATPase rates under conditions used in the EPR experiment (Table 2). Altogether, these data suggest that the local conformational changes of lid closure and NTD-rotation are part of the rate-limiting barrier to the closed state and are regulated by N-terminal residues of the strand swap and extended strap in TRAP1.

Figure 5.

Lid Closure rate is regulated by the NTD-strap.

(A) Continuous Wave (CW) EPR scans of cysteine Free WT (top) and Δstrap zTRAP1 (bottom) labeled with a spin-probe on the NTD-lid (green) in order to observe changes to the lid in the apo and closed states (‘Materials and methods’). In the apo state the lid probe shows signal for both mobile and immobile states, although crystallographic data indicate that even in the mobile state, the majority of the lid is still reasonably well ordered. After addition of AMPPNP, the observed signal shifts indicating a predominantly mobile state of the lid, which corresponds to changes in lid dynamics that accompany NTD rotation and dimerization. Only subtle differences are seen in the mobile:immobile peak ratio upon strap deletion. (B) CW-EPR scans at ∼23°C taken for the cysteine-free WT (red) and Δstrap zTRAP1 (blue) over time after addition of AMPPNP. The percent change in peak height (final vs start) over time is plotted for both the immobile (squares) and mobile (circles) components, showing a clear anti-correlation. The mobile and immobile populations were jointly fit with a single exponential process (‘Materials and methods’) having a rate constant of 0.014 min−1 for WT and 0.075 min−1 for Δstrap, demonstrating a strong coupling between the strap and the NTD-lid.

DOI: http://dx.doi.org/10.7554/eLife.03487.015

Dissecting further regulatory functions of the NTD-strap

The experiments above collectively suggest a major role for the N-terminal strap as a direct modulator of the kinetic barrier separating the apo and closed states for TRAP1. Moreover, it appears to be also responsible for the pronounced temperature-sensitivity. Although the experiments above indicate a strong role in modulating the forward closure rate, the TRAP1 crystal structure would suggest that deleting the strap might also compromise the stability of the closed state, thereby enhancing reopening rate and shifting the equilibrium towards the open state. To measure the re-opening rate, inter FRET-labeled hTRAP1 was pre-closed with AMPPNP. After closure was complete a 20-fold excess ADP was added such that upon re-opening of the NTD dimer interface, ADP would exchange resulting in a decreased FRET signal (Street et al., 2011). Previous studies found apo state nucleotide on and off-rates to be fast (Leskovar et al., 2008), thus the above experiment provides a good approximation of the uni-molecular reopening rate. Monitoring FRET kinetics revealed that strap removal accelerated re-opening of the NTD dimer interface by ∼eightfold (0.0021 min−1 → 0.016 min−1; Figure 6A, Table 3). These data suggest that the strap contacts observed in the closed state (Lavery et al., 2014) do in fact impact closed-state stability, by about 1.2 kcal/mol, however, the larger effect (∼16-fold, 0.02 min-1 → 0.31 min−1, 1.7 kcal/mol) is on the kinetic barrier corresponding to release of the strap from the apo state.

Figure 6.

The NTD-strap plays a smaller role in additional steps of the ATPase cycle.

(A) Schematic of dimer closure and re-opening upon addition of AMPPNP (PNP) using the dimer closure FRET probe (left). Re-opening of WT hTRAP1 and Δstrap was induced by 20-fold excess ADP after closure with AMPPNP. Re-opening was accelerated by ∼eightfold upon removal of the strap as determined by the ratio of the rates (table inset). (B) Steady-state FRET scans of dimer closure FRET in apo and plus ATP in the absence of Mg2+. Without Mg2+ a closed state accumulates, whereas subsequent addition of Mg2+ (‘+ATP & Mg2+’) allows hydrolysis to proceed thereby shifting the population to the apo state. (C) Schematic of a kinetic experiment using the Mg2+ dependence to separate the rate of hydrolysis from rate of closure. By omitting Mg2+, the population can be synchronized in a closed state that is unable to hydrolyze ATP. Subsequent rapid addition of Mg2+ leads to ATP hydrolysis, which has now been decoupled from the closure step. (D) Kinetic experiments measuring closure and (E) ATP hydrolysis. No closed state accumulates if Mg2+ is included in the closure reaction. Again we observe that removal of strap residues leads to an accelerated closure rate, whereas the difference in ATP hydrolysis is small. Kinetic rates for each are listed in the table insets.

DOI: http://dx.doi.org/10.7554/eLife.03487.016

(A–B) Kinetic experiment designed to decouple ATP hydrolysis from the preceding closure step at 25°C outlined in Figure 6C–E. The measured rates for WT agree well with previous single turnover measurements (Leskovar et al., 2008), however, we find that decoupling the closure rate from hydrolysis results in a reassignment of the previous rates with closure being the slowest step. Kinetic rates for each are listed in the table inset.

DOI: http://dx.doi.org/10.7554/eLife.03487.017

Because the strap could also play a role in the hydrolysis reaction, we needed a method to decouple closure from ATP hydrolysis. As closure is rate-limiting, even single-turnover experiments would provide an aggregate rate made up of the closure and hydrolysis steps. During the course of our FRET experiments, we discovered that omitting Mg2+ from the reaction buffer results in an accumulation of the closed-state in the presence of ATP without ATP hydrolysis. By contrast, in the presence of ATP and Mg2+, TRAP1 is predominantly in the apo state as a consequence of hydrolysis (Figure 6B,D). The latter is consistent with previous observations with yHsp90 (Hessling et al., 2009) and bHsp90. These observations allowed us to decouple the closure and hydrolysis steps by pre-incubating TRAP1 with excess ATP without Mg2+, thereby stalling the reaction in the closed state (illustrated in Figure 6C). Upon addition of Mg2+, ATP is hydrolyzed and the equilibrium shifts predominantly to the apo state as seen by the loss of FRET (Figure 6B). Testing WT and Δstrap in this assay revealed an acceleration of closure with removal of the strap, consistent with experiments using AMPPNP (Figure 6D). Interestingly, the closure rate measured by FRET is significantly faster with ATP than AMPPNP suggesting a significant difference in energetics between the nucleotide analogs (Table 3). The use of ATP for FRET-based closure measurements better matches our ATPase measurements and points to a correlation between closure rates and ATPase activity (0.79 min−1 ATPase vs 0.42 min−1 FRET Closure, both measurements with Inter FRET probe protein), though we do still observe a difference perhaps representing a small Mg2+ contribution. Addition of excess Mg2+ showed the predicted drop in FRET and revealed a minor difference in hydrolysis rate (∼1.4-fold) (Figure 6E, Table 3), suggesting that the strap may also subtly alter lid dynamics in the closed state. The acceleration effects observed for the Δstrap protein are greater at 25°C, where the temperature dependent difference is more pronounced (Figure 6—figure supplement 1, Table 3). Notably, our measured closure and hydrolysis rates matched previously reported values for these steps modeled using a global fitting procedure (Leskovar et al., 2008). However, the closure and hydrolysis rates measured here (0.003 s−1 and 0.07 s−1, respectively) were somewhat arbitrarily assigned to the reverse order in the previously reported model. Since our experiments independently measure both reactions, we can now assign closure to be the slowest and hence rate-limiting step. This model is in good agreement with the other data presented in this study.

Our combined data better define the kinetic cycle for TRAP1 and support a model where the strap regulates multiple steps with the largest contribution being to the thermal sensitive rate-limiting kinetic barrier between the apo and nucleotide-bound closed states.

Discussion

The conservation of Hsp90 has been established from bacteria to humans, giving rise to homologs in different species and distinct versions in different cellular compartments (Johnson, 2012). Though biochemical and structural studies have identified key differences in the thermodynamic and kinetic properties amongst the homologs, the underlying set of conformations and the overall ATP hydrolysis cycle appear conserved and essential for client maturation in vivo (Panaretou et al., 1998; Southworth and Agard, 2008).

Here, we identify and characterize unique kinetic and thermodynamic properties of the mitochondrial Hsp90 (TRAP1) and use a combination of biophysical and biochemical techniques to consistently show that a 26-residue N-terminal extension or ‘strap’ (compared to yHsp90) (Lavery et al., 2014), kinetically regulates the formation of the active closed conformation and is responsible for the surprising temperature-dependence of closure. This extension is elaborated to varying degrees in the different Hsp90 isoforms; absent in yeast and bacterial Hsp90s, shortest in the dominantly expressed mammalian cytosolic Hsp90s and longest in the mammalian organellar Hsp90s (Figure 7). Below we propose that extensions and variability in the N-terminal sequence serve to fine-tune the activity of Hsp90 homologs in diverse species or compartments in response to functional demands and environmental factors, with temperature playing an important role in TRAP1.

Figure 7.

Evolution of Hsp90 NTD-strap sequences.

Alignments were generated individually for each Hsp90 isoform using a conserved portion of the N-terminal domain and the NTD-strap region. The variable signal sequences for TRAP1 and Grp94 were removed before aligning the 10 divergent sequences. Helix one (H1) of the NTD is annotated above the alignments and begins just after the strictly conserved Phe residue that structurally appears to separate the β-strand region of the NTD from H1. This alignment clearly shows the divergence of both length and sequence within the NTD-strap region and also reveals that residues are more conserved amongst Hsp90 isoforms within H1 and the region following H1. TRAP1 has a much longer strap region than cytosolic Hsp90 and conservation does not pick up until the structural region, as made evident in the TRAP1 crystal structure (Lavery et al., 2014). Both yHsp90 and bHsp90 lack a significant strap sequence and Grp94 clearly has an extended and well-conserved strap region.

DOI: http://dx.doi.org/10.7554/eLife.03487.018

N-terminal residues and kinetic regulation of Hsp90

While the crystal structure of the TRAP1 closed state revealed that the strap made stabilizing interactions with the trans protomer, we show here that its dominant role in modulating ATPase activity is to limit the closure kinetics, presumably though analogous cis-protomer interactions in the apo state. Removal of the strap leads to a ∼30-fold increase in ATPase rate and faster closure kinetics that include the smaller conformational steps of NTD-rotation and lid closure, as well as loss of thermal regulation of dimer closure.

The strap extension in TRAP1 appears to continue and expand upon the kinetic regulatory affects observed previously for the first eight residues in yHsp90, which makes contacts on the trans-protomer in the closed state (Ali et al., 2006; Lavery et al., 2014). Deletion of these residues was shown to accelerate the ATPase rate by ∼1.5-fold by allowing H1 and the lid to undergo conformational changes necessary to form trans-protomer contacts at the NTD-dimer interface (Richter et al., 2002, 2006). These effects are understood in the light of numerous apo NTD structures showing this strand makes analogous contacts with its own NTD in the apo state (Shiau et al., 2006; Dollins et al., 2007; Li et al., 2012). In the TRAP1 closed-state structure, the 14 ordered residues wrap around the side of the NTD and add an additional 757 Å2, as calculated with PISA (Krissinel and Henrick, 2007), of buried surface area and several new trans-protomer contacts (Lavery et al., 2014). We propose that similar additional contacts are made in the apo state (Videos 2 and 3), which is supported by our own data showing that truncations up the first major contact (salt bridge) have no effect on ATPase (Figure 3). Given the similar accelerating effects on ATPase and dynamics as studied in multiple organisms, it is likely that the β-strand and the strap are acting on the same barrier.

Distilling the available information, we outline a model that defines kinetic steps in the Hsp90 ATPase cycle and consequently determine the rates of ATP hydrolysis (Figure 8A). Specifically, after ATP is bound, release of cis contacts of the β-strand/strap is coupled to lid closure and NTD rotation, presenting surfaces that form and stabilize an NTD-dimerized state. Due to closure-induced strain this ultimately results in an asymmetric conformation (Lavery et al., 2014). Hydrolysis of one of the two ATPs leads to rearrangement of client binding residues (red) between the MD:CTD thus coupling the first ATP to client remodeling when clients are bound to this region. The actual conformational state post hydrolysis of the first ATP is currently unknown, but is here schematized as the symmetric state identified in the yHsp90 crystal structure (Ali et al., 2006). After the second ATP is hydrolyzed the chaperone assumes the previously observed compact ADP conformation before resetting the cycle to the apo state.

Figure 8.

Model for the conformational cycle and unique energy landscape of TRAP1.

(A) In the absence of nucleotide the chaperone is in equilibrium between various open conformations (for simplicity we only show the most open) with the strap folded back onto the cis protomer. Upon binding of ATP, conformational changes necessary for the transition to the closed state are initiated. Here, we propose that the cis contacts of the strap are broken allowing the lid and NTD to undergo conformational changes towards the closed state. After the slow closure step the chaperone assumes the previously reported asymmetric conformation (Lavery et al., 2014). Sequential hydrolysis leads to changes in symmetry rearranging the unique MD:CTD interfaces and client binding residues (red) before sampling the ADP conformation and resetting the cycle to the apo state equilibrium. (B) Model for the unique energy landscape of TRAP1. Solid lines illustrate the energy landscape of WT TRAP1, and the dashed lines depict the change in landscape upon the loss of the extended N-terminal strap sequence in TRAP1. By stabilizing both the apo and closed states, the strap increases the effective height of the energy barrier. This modulates the conformational landscape, and in the case of hTRAP1 provides pronounced temperature sensitivity.

DOI: http://dx.doi.org/10.7554/eLife.03487.019

Specific regulation of the energetic landscape imparted by the TRAP1 strap is depicted in Figure 8B. Here, we propose the effect of the strap ultimately impacts the kinetic barrier height as the strap stabilizes both the apo and closed states, although the apo state stabilization is dominant. Thus, addition of a structural element that makes analogous interactions in both the apo (cis) and closed (trans) states provides a novel strategy for kinetic regulation by accentuating the barrier between the apo and closed conformations.

Functional implications for the evolution of an N-terminal strap

While Hsp90 is very highly conserved across species, there are several regions such as the N-terminus, the charge linker and the very C-terminus that have diverged significantly during evolution. As highlighted in Figure 7, the different classes of Hsp90s segregate quite clearly according to the length of their N-termini, with the bacterial and yeast Hsp90s being the shortest, followed in turn by the metazoan cytosolic Hsp90s, the mitochondrial TRAP1s, and the ER Grp94s. One exception is a recently discovered Hsp90α alternative splice variant that creates a very large N-terminal extension of 122 residues. In keeping with observations here, biochemical analysis revealed that this extension is a negative regulator of ATPase activity (Tripathi and Obermann, 2013).

In the TRAP1 family, conservation of the strap is strong through the known structured region (His87 in zTRAP1), but decreases towards the N-terminus, and is greatly reduced for TRAP1s from blood fluke, insects, and the sea urchin. Cytosolic Hsp90 has the same drop in conservation and a much shorter strap. By contrast, Grp94 has a very long and very well conserved strap region, with a somewhat variable, but very acidic N-terminus. Despite its long size, deleting the analogous strap region in Grp94 accelerates ATP hydrolysis by only fivefold, although temperature modulation was not investigated (Dollins et al., 2007). However, its extreme length, the strong conservation, and the modest effect of deletion on ATPase rates, suggest a possible regulatory role that could couple other phenomena beyond temperature to the rate-limiting conformational changes required for ATP hydrolysis.

The observation that the catalytic efficiency of different Hsp90s vary by ∼15-fold (Richter et al., 2008) suggest that regulation of the rate-limiting step has been highly tuned through evolution for functional importance. In support of this, yHsp90 mutations that accelerate or decelerate ATPase rates result in significant growth defects and loss of client protein folding in vivo (Nathan and Lindquist, 1995; Prodromou et al., 2000). The evolution of additional residues at the N-terminus of the Hsp90 gene provides a convenient way to adapt the chaperone's conformational cycle to function with diverse clients encountered by the different homologs or under stressed environmental conditions. Additionally, while the cytosolic Hsp90s are highly regulated by several co-chaperones (Zuehlke and Johnson, 2010), only one co-chaperone has been identified for the organellar homologs (Liu et al., 2010). This brings forth the possibility that the more extended strap in these homologs could directly or indirectly perform some of the regulation that co-chaperones provide to Hsp90 in the cytosol.

The marked temperature sensitivity observed with TRAP1 raises the intriguing possibility that it represents a homeostatic response in mitochondria where heat is generated through uncoupling of the electron transport chain (Rousset et al., 2004). In keeping with the physiological relevance, we demonstrate that the thermal sensitive kinetic barrier is measurably different between zebrafish and human TRAP1, which have significantly different physiological temperatures and environments. Additionally, added contacts that the strap provides could be a target for post-translation modifications or even provide a novel binding site for ions, metabolites, or other factors that could modulate the regulatory functions of this element. These observations provide an example of how evolved extensions at the Hsp90 N-terminus can be used to fine-tune chaperone activity to match organism-specific environmental conditions or unique subcellular demands required for optimal function.

Materials and methods

Protein production and purification

Full-length and mutant versions of TNF receptor-associated protein 1 (TRAP1) from Homo sapiens and Danio rerio (hTRAP1 and zTRAP1, respectively) were purified using our previously described protocol (Lavery et al., 2014). The coding sequence of proteins used in this study were cloned into the pET151/D-TOPO bacterial expression plasmid (Life Technologies, Grand Island, NY) and mutant versions of were generated by standard PCR based methods. Cysteine-free hTRAP1 with encoded cysteine positions (Glu140Cys or Glu407Cys) on each or (Ser133Cys and Glu407Cys) on a single protomer, allowed for site-specific labeling with maleimide derivative Alexa Fluor 555/647 dyes (Life Technologies) for FRET experiments. These constructs were also purified as previously described (Lavery et al., 2014),with a final size exclusion chromatography storage buffer of 50 mM Hepes pH 7.5, 100 mM KCl, 500 μM TCEP. Aliquots of stored protein were labeled with fluorescent dyes as described below.

Negative-stain electron microscopy

WT hTRAP1 was initially diluted to 0.1 mg/ml in a buffer containing 20 mM NaH2PO4 pH 7, 50 mM KCl, and 2 mM MgCl2, 0.02% n-octyl-β-D-glucoside + 2 mM AMPPNP. Reactions were incubated at various temperatures for 1 hr (or overnight), followed by dilution to 0.01 mg/ml in the buffer above including 2 mM AMPPNP to maintain nucleotide concentration. 5 µl of the resulting reactions was then incubated for ∼1 min on 400 mesh Cu grids (Pelco, Redding, CA) coated with a thin carbon layer (∼50–100 Å). Following sample incubation, the grid was washed 3× with miliQ water, and lastly stained 3× with uranyl formate pH 6. The final stain was removed by vacuum until the surface of the grid was dry. Prepared grids were imaged with a TECNAI 12 (FEI, Hillsboro, OR) operated at 120 kV. Images were recorded using a 4k × 4k CCD camera (Gatan, Pleasanton, CA) at 52,000 magnification, at −1.5 μm defocus. Representative closed state particles were selected in EMAN (Ludtke et al., 1999).

SAXS data collection and analysis

TRAP1 homologs and mutant proteins were buffer exchanged into 20 mM Hepes pH 7.5, 50 mM KCl, 2 mM MgCl2, 1 mM DTT. 75 μM protein (monomer concentration) was used as the final concentration for all reactions, and 2 mM AMPPNP was added to initiate closure. Reactions were incubated at various temperatures for 1 hr followed by a spin at max speed in a tabletop centrifuge for 10 min immediately prior to data collection to remove any trace aggregation.

Data were collected at the Advanced Light Source (ALS) at beamline 12.3.1 with sequential exposure times of 0.5, 1, and 0.5 s. Each sample collected was subsequently buffer subtracted and time points were averaged using scripts provided at beamline 12.3.1 and our own in-house software ‘saxs_multiavg.py’. The scattering data were transformed to P(r) vs r using the program GNOM (Svergun, 1992) and Dmax was optimized. The resulting distributions were fit using an in-house least squares fitting program ‘saxs_combine.py’ in the region where non-zero data were present for the target data and closed state model. For the fitting we chose theoretical scattering data for our TRAP1 closed-state model (Lavery et al., 2014) and the WT apo data for each TRAP1 homolog. The WT apo data were chosen as the best representation of apo for two reasons. (1) The apo state of Hsp90 proteins consist of a mix of conformations (Southworth and Agard, 2008) of which the various conformations and percent of each remains to be elucidated for TRAP1, and (2) removal of the strap (particularly in hTRAP1) induces a shift of the apo distribution towards the closed state as observed for hTRAP1 by SAXS (data not shown), which would result in a value of percent closed for the Δstrap protein that would under represent the true value relative to WT. The theoretical scattering curve for the TRAP1 crystal structure was generated in the program CRYSOL (Svergun et al., 1995). The percent of components utilized in the fit and an R factor (R_merge) that is similar to a crystallography R factor in nature is output from our least-squares fitting program and values reported in Table 1. R_merge is defined as the equation belowwhere Pobs(r) is the observed probability distribution and Pcalc(r) is the calculated modeled fit. Both pieces of in-house software used for SAXS data analysis, ‘saxs_multiavg.py’ and ‘saxs_combine.py’, have been deposited at GitHub.com (https://github.com/agardd/saxs_codes).

Steady-state ATPase measurements

Steady-state kinetic measurements for various Hsp90 homolog and mutants were carried out in previously described conditions unless otherwise indicated (Lavery et al., 2014). Specific buffer conditions used to measure kinetic rates for cysteine free zTRAP1 proteins used in EPR were 20 mM Hepes pH 7.4, 150 mM NaCl, 2 mM MgCl2 at 23°C with 2 mM ATP (see EPR method description). Buffer conditions used to measure kinetic rates for cysteine free hTRAP1 (WT and ΔStrap) used in FRET experiments were 50 mM Hepes pH 7.5, 50 mM KCl, 5 mM MgCl2 with 2 mM ATP (see FRET method description) measured at 30°C. Results were plotted using the program R (R Development Core Team, 2010).

Fluorescence Resonance Energy Transfer (FRET) measurements

Purified protein was labeled with maleimide derivative AlexaFluor 555 (Donor) and 647 (Acceptor) (Life Technologies) at fivefold excess over protein (pre-mixed at 2.5-fold concentration each dye for dual labeled sample) overnight at 4°C. Labeling reactions were then quenched with twofold β-mercaptoethanol over dye concentration and free dye was removed with desalting columns containing Sephadex G-50 resin (illustra Nick Columns, GE Healthcare, Pittsburgh, PA).

For FRET measurements using probes that monitor closure across the dimer (Glu140Cys, Glu407Cys, ‘Inter FRET’), labeled protein was mixed at a 1:1 ratio with a final concentration of 250 nM. For measurements with the probe that measures NTD rotation (Ser133Cys and Glu407Cys, ‘Intra FRET’), WT hTRAP1 was mixed in 20-fold excess over labeled protein (250 nM labeled protein:5 μM WT). Heterodimers for experiments with all FRET probes were formed at 30°C for 30 min in a reaction buffer consisting of 50 mM Hepes pH 7.5, 50 mM KCl, 5 mM MgCl2. Following heterodimer formation, closure was initiated by addition of 2 mM AMPPNP at various temperatures (Figure 4). To measure re-opening, 40 mM ADP was rapidly mixed with pre-closed reactions (closed as in Figure 4). For ATP hydrolysis experiments, closure was initiated with 2 mM ATP in a reaction buffer consisting of 50 mM Hepes pH 7.5, 50 mM KCl. After closure was complete, hydrolysis was initiated by rapid addition of 5 mM MgCl2.

Closure and ATP hydrolysis experiments (Figures 4 and 6B) were carried out using a Jobin Yvon fluorometer with excitation and emission monochromator slits set to 2 nm/3 nm (respectively), an integration time of 0.3 s, and excitation/emission wavelengths of 532/567 nm (donor) and 532/667 nm (acceptor). Re-opening experiments (Figure 6A) was measured at 30°C using a SpectraMax5 plate reader with excitation and emission wavelengths as above and with a 540 nm emission cutoff. Kinetic measurements were taken at a time interval to minimize photobleaching.

The change in FRET (ratio of Donor and Acceptor fluorescence—division done to graph positive changes and normalized for visual comparison) was well fit with a single exponential (fit in KaleidaGraph, Synergy Software, Reading, PA) to obtain the rate of closure and NTD rotation (Fit 1), as well as re-opening and ATP hydrolysis rates (Fit 2).where m1 is the time zero value, m2 is the amplitude, m3 is the rate constant, and x is time in seconds. For steady-state FRET scans (taken before and after kinetic measurements), reactions were excited at 532 nm and emission was collected from 550–750 nm. FRET scans were normalized such that the area under the curve is 1.

The activation energy was calculated by fitting a plot of the natural log (ln) of the observed closure rate (y-axis) verses inverse temperature (x-axis) using the equation below (Fit 3)where Ea is the activation energy, T is temperature (kelvin, K), R is the gas constant (kcal K−1 mol−1), and A is a pre-exponential factor. ATPase rates used to calculate Ea for Hsp90 homologs were taken from reference (Frey et al., 2007).

Continuous-wave electron paramagnetic resonance (EPR)

Cysteine-free zTRAP1 with a Ala201Cys mutation on the lid was exchanged into non-reducing EPR buffer (20 mM Hepes pH 7.4, 150 mM NaCl) at 100 μM (monomer concentration) and labeled by the addition of N-(1-oxyl- 2,2,6,6-tetramethyl-4-piperidinyl)maleimide (MSL, Sigma, St. Louis, MO) to 2.5× concentration of protein overnight at 4°C. The labeled protein was then run through a Micro Bio-Spin column P-30 (Bio-Rad, Hercules, CA) to eliminate free probe. EPR spectra were obtained at ∼100 μM labeled protein ± 2 mM AMPPNP in the buffer above with addition of 2 mM MgCl2 and after heating at 30°C for 30 min to ensure closure has completed (Figure 5A). For the time course (Figure 5B), protein (apo) was spiked with 2 mM AMPPNP and EPR scans recorded overtime at room temperature (∼23°C).

EPR measurements were performed with a Bruker EMX EPR spectrometer (Bruker, Billerica, MA) in a 50-μl glass capillary. First derivative X-band spectra were recorded in a high-sensitivity microwave cavity using 50-s, 10-mT-wide magnetic field sweeps. The instrument settings were as follows: microwave power, 25 mW; time constant, 164 ms; frequency, 9.83 GHz; modulation, 0.1 mT at a frequency of 100 kHz. Each spectrum used in the steady-state data analysis was an average of 10–20 sweeps from an individual experimental preparation, with one sweep used for kinetic measurements.

Analysis of the raw peak heights indicated that both the mobile and immobile fractions were changing as a concerted single exponential process. As a consequence, to determine the rate constant, it was unnecessary to account for peak overlaps or the starting fraction in each state. To quantify, the raw peak heights at each time point were determined using the Bruker EMX EPR spectrometer software (Bruker, Billerica, MA) and converted to a percent change over the time course. The rates of lid closure for WT and Δstrap were estimated by fitting the normalized peak heights for each sample to a single exponential decay process with the same rate constant for the mobile and immobile peaks (done as a constrained non-linear fit in Prism v6, GraphPad software, La Jolla, CA).

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for sending your work entitled “A novel N-terminal extension in mitochondrial Hsp90 (TRAP1) serves as a thermal regulator of chaperone activity” for consideration at eLife. Your article has been favorably evaluated by John Kuriyan (Senior editor), a Reviewing editor, and 2 reviewers.

The Reviewing editor and the reviewers have commented positively on your manuscript. Both reviewers have identified one point that needs additional support. This issue concerns the fact that the physiological relevance of the work is not so clear, since the major effects for human TRAP1 happen at temperatures below 37°C. Both reviewers were interested in seeing data recorded at higher temperatures. Please address this important issue in the revised manuscript and as many of the comments in the two reviews as possible.

Reviewer #1

Hsp90 is a molecular chaperone found in prokaryotes and in the cytosol as well as organelles of the eukaryotic cell. Its function is coupled to ATP hydrolysis along with associated large conformational changes. While the key steps of the Hsp90 ATPase cycle are understood, the regulation of these structural rearrangements is still elusive. Previously, the authors solved the crystal structure of mitochondrial Hsp90, TRAP1, bound to AMPPNP.

They identified a 14-residue extension of the N-terminal beta-strand which crosses over between protomers in the closed state. This 'strap' is found in higher eukaryotes but is absent in yeast and bacteria. The authors also showed that point mutations or the deletion of the strap in TRAP1 (of zebra fish) or endoplasmic Grp94 results in an increase in ATPase activity. In this study, they show that a temperature-dependent kinetic barrier limits the conformational changes from the apo to the closed form of TRAP1. At lower temperature (23°C) TRAP1is predominantly open, even in the presence of AMPPNP. Local conformational changes associated with lid closure are a part of the rate limiting step to the closed state and are regulated by the strap in TRAP1.

This is an interesting observation adding to our understanding of aspects of the conformational cycle of Hsp90 and species-specific differences.

However, the biological function of the temperature regulation remains unclear to me, assuming that 37°C may be the resting state of human mitochondria and higher temperature present during energy generation or under thermal stress conditions. Thus the regulation should be active at higher temperatures, such as 42°C. Experiments addressing this issue would be of interest.

Specific points:

1) The authors state that the kinetic barrier for closure is large and unusually sensitive to temperature changes. Examples should be included to allow for comparison.

2) The kinetic data should include fits and resulting rate constants (and also which equation was used) to judge quality of the kinetic model.

3) If the measured conformational change is indeed rate limiting then its temperature dependence should be the same as that of the ATPase activity measured before. Is this indeed the case? How is the relationship of the shortened variant?

4) Figure 1: Do other Hsp90 isoforms also show a similar trend over a temperature range or is this special just for TRAP1?

5) Experiments in Figure 1 and 2 do not show kinetics, just a shift of equilibria.

6) EM images for delta strap should be included.

7) In an Agard publication from 2008 (Southworth & Agard, Cell 2008) EM images of HtpG, yeast and human Hsp90 are shown in the apo state, with AMPPNP, and with ADP. The human Hsp90α is in an open conformation with AMPPNP at 37°C (according to them TRAP1 is fully closed at 37°C).

8) Figure 2–figure supplement 1: Isn't this the same figure as Figure 2C just without the delta-strap?

9) Is it known that the trans contacts that strap forms in the closed conformation stabilize the closed conformation?

10) Figure 4A: For the NTD:MD rotation, the FRET probe does not show a significant change in the FRET signal.

11) Figure 4B: Why was saturation not reached for the 42°C sample?

12) Data for the dimer closure FRET construct should be included and it would be important to see how delta-strap acts in the FRET assay at different temperatures to compare with the SAXS data.

13) Why was the double strap mutant too unstable for NTD rotation FRET and not for inter-protomer FRET? Isn't it the same construct just with a different Cys label?

14) Figure 5A: The legend says that in the apo state the lid probe is in equilibrium between mobile and immobile states. In the Results section it is stated that the apo form is predominantly immobile.

15) According to the authors, deleting strap compromises the stability of the closed structure and hence enhances the reopening rate and shifts the equilibrium towards the open state. If deleting strap shifts the equilibrium towards the open state, why is the delta-strap construct predominantly closed?

In this context, the authors mention that the effect of deleting the strap on the opening of the NTD interface is smaller than the effect on the kinetic barrier corresponding to the release of strap from the apo state. The respective numbers seem to be missing from Table 3.

16) Are the ATPase activities of the cys-free version of hTRAP1 and zebrafish TRAP1 mutant identical to the respective wild type proteins?

17) Figure 6. The legend says delta strap is ∼7 fold faster; text says ∼8 fold faster.

18) Figure 6C: a trace showing that in presence of ATP without Mg2+, there is no ATPase activity should be added. Not adding Mg2+ is not necessarily equivalent to not having (ambient) Mg2+ present in the solution. The ATP induced changes in FRET signals should be measured also in the presence of EDTA.

19) The authors mention that in the absence of Mg2+, ATP and AMPPNP show pronounced differences in kinetics of FRET signal, the difference with/without Mg2+ should be even more pronounced, that is the kinetics of o/c may be substantially faster in the presence of Mg2+. Measuring the FRET kinetics upon addition of Mg-ATP and Mg-AMPPNP is crucial to show that ATP induced closing kinetics in absence of Mg2+ are indeed representative for the ATPase cycle.

20) How does one know that Mg2+ can actually bind to the closed form; and that a re-open is not necessary for this to happen?

Reviewer #2

In this study, Partridge and colleagues investigate the role of the N-terminal extension (“strap”) of TRAP1, the mitochondrial Hsp90 isoform, which in the previously solved crystal structure of the TRAP1 dimer wraps around the N-domain of the opposite protomer. They characterized the effects of the strap on the dimer closure kinetics, the rotation of the N-domain relative to the M-domain, the ATP hydrolysis and movement of the ATP lid (N-domain) using negative stain electron microscopy (EM), small angle x-ray scattering (SAXS), fluorescence resonance energy transfer (FRET) and electron paramagnetic resonance (EPR) measurements. They demonstrate that the strap region is responsible for a temperature-dependent increase in the rate of TRAP1 closure, as well as the increase in the ATPase activity.

The data presented here is convincing and interesting. The physiological relevance of the observed phenomenon is not so clear since the major effects for human TRAP1 happen at temperatures below 37°C. Nevertheless, the story could be published after the authors addressed the raised issues.

Major comments

1) Figure 1 shows negative stain EM images of TRAP1 in the presence of AMPPNP pre-incubated at different temperatures. Few representative samples are picked from each grid to show the transition from open to closed conformation with the increase in temperature. The authors should quantify the open and closed structures from a representative square of the electron micrograph.

2) Using SAXS the authors measured AMPPNP-induced transition of human and zebrafish TRAP1 to the closed conformation between 20 and 36°C (Figure 2). To determine the physiological relevance of their observations the authors could have measured the AMPPNP induced transition of human TRAP1 at 37 to 42°C. Does human TRAP1 become more active at heat shock temperatures?

3) Figure 4: The authors investigate AMPPNP-induced changes in TRAP1 conformation using FRET. Control experiments with only the acceptor dye need to be shown especially as the changes in fluorescence for the NTD:MD rotation seems to be very small. The temperature at which the experiment of Figure 4(A) has been performed should be mentioned in the figure legend.

4) The dimer closure FRET experiments have been performed only at 30°C. As the paper deals with effect of temperature on ATPase rates of TRAP1, it would be very important to see the change in rate of dimer closure at different temperatures. It would also be interesting whether the rates of NTD-MD-rotation and dimer closure are similar to each other.

5) In Table 2 the authors write that the steady state ATPase rate for human TRAP1 was 0.463 min-1 and in Table 3 the write that the closing rate for human TRAP1 at 30°C was 0.02 min-1. These values do not fit together and contradict the claims of the authors. The authors should indicate at which temperature the ATPase assays were performed and correlate the closing rate with the ATP hydrolysis rates to substantiate their claims. Maybe the authors will have to measure the closing rate upon addition of ATP instead of AMPPNP. This seems possible since omission of Mg prevents hydrolysis as the authors have shown.

The Reviewing editor and the other reviewers have commented positively on your manuscript. Both reviewers have identified one point that needs additional support. This issue concerns the fact that the physiological relevance of the work is not so clear, since the major effects for human TRAP1 happen at temperatures below 37°C. Both reviewers were interested in seeing data recorded at higher temperatures. Please address this important issue in the revised manuscript and as many of the comments in the two reviews as possible.

The major question echoed by all was whether the experiments done in this paper could be done at higher temperatures to better reflect how the observed temperature dependent activity of TRAP1 would be beneficial above homeostatic temperatures of the organism for the homologs tested. We have attempted to address these concerns by including new data recorded at temperatures above 37°C. 37°C data were already included in the original manuscript. This includes new electron micrographs of samples incubated at 42°C, SAXS data up to 43°C, and better explanation of previously recorded data demonstrating that the rate of ATPase activity in TRAP1 will continue to increase until 60°C, at which point TRAP1 begins to denature. The new data taken at temperatures above 37°C has been added to already existing figures.

The new EM data has been incorporated into Figure 1, panel A. The new SAXS data has been incorporated into Figure 2, panels A, B, C, D. Table 1 has also been updated to include data above 37°C, reflecting the changes made to Figure 2. We also responded to the reviewer’s requests by including FRET data for both types of probes, intra and now inter FRET (Figure 4B). The increasing temperature series for both sets of FRET probes behaves similarly and increases as a response to temperature with each assay having a comparable fold increase. Further we added extensive temperature series of closure kinetics measured using the inter FRET probe between 23 and 42°C for the delta strap variant of human TRAP1. Two new panels have been added with this data (Figure 4D, E). Importantly this data, together with that previously shown in Figure 4B allows calculation of an Arrhenius activation energy ±strap. This shows that the strap contributes ∼60% of the activation energy measured for WT TRAP1 (Figure 4E, Figure 4 legend and within the main text). Further, we have made every attempt to improve the manuscript as suggested.

These edits include changes to main text figures, as well as further edits to the text. We have additionally pointed out examples of data recorded at temperatures above 37°C that were included in the original submission, such as the FRET experiment measuring the temperature dependence of NTD rotation. These changes and our responses to individual comments by reviewers are discussed in more detail below.

Reviewer #1

Hsp90 is a molecular chaperone found in prokaryotes and in the cytosol as well as organelles of the eukaryotic cell. Its function is coupled to ATP hydrolysis along with associated large conformational changes. While the key steps of the Hsp90 ATPase cycle are understood, the regulation of these structural rearrangements is still elusive. Previously, the authors solved the crystal structure of mitochondrial Hsp90, TRAP1, bound to AMPPNP.

They identified a 14-residue extension of the N-terminal beta-strand which crosses over between protomers in the closed state. This 'strap' is found in higher eukaryotes but is absent in yeast and bacteria. The authors also showed that point mutations or the deletion of the strap in TRAP1 (of zebra fish) or endoplasmic Grp94 results in an increase in ATPase activity. In this study, they show that a temperature-dependent kinetic barrier limits the conformational changes from the apo to the closed form of TRAP1. At lower temperature (23⁰C) TRAP1is predominantly open, even in the presence of AMPPNP. Local conformational changes associated with lid closure are a part of the rate limiting step to the closed state and are regulated by the strap in TRAP1.

This is an interesting observation adding to our understanding of aspects of the conformational cycle of Hsp90 and species-specific differences.

However, the biological function of the temperature regulation remains unclear to me, assuming that 37⁰C may be the resting state of human mitochondria and higher temperature present during energy generation or under thermal stress conditions. Thus the regulation should be active at higher temperatures, such as 42⁰C. Experiments addressing this issue would be of interest.

We have attempted to address the concern of all reviewers by now including

SAXS and EM data taken at temperatures above 37°C. Both high temperature datasets agree with the basic observation that a compact or “closed” conformation dominates the population at equilibrium.

With EM we see that the predominant form is a closed conformation at higher temperatures as demonstrated in Figure 1. Additionally there continues to be an increase in the % closed population as measured with SAXS, Figure 2. This increase in the % closed population in both species has been tabulated in Table 1. Looking at both Figure 2 and Table 1 you can see that the % closed does continue to increase beyond 37 °C, although there is one outlier in all this data and that is WT zTRAP1, which shows a decrease in % closed. Presumably, temperatures above 37° C are physiologically irrelevant for zebrafish. That said, Δstrap zTRAP1 did continue to show an increase in the % closed population. Concerning steady-state ATP hydrolysis measurements at temperatures above 37 °C we also modified the text to make it more obvious that temperature dependence of ATPase had previously been characterized for TRAP1 by Johannes Buchner’s lab in manuscripts referenced in the text.

Our original submitted manuscript did include some FRET measurements taken at 42°C with WT hTRAP1 showing a dramatic increase in the rate of closure compared with 36°C, Figure 4B and Table 3. A 42°C closure rate of the Δstrap variant is also included in Figure 4D and Table 3.

Specific points:

1) The authors state that the kinetic barrier for closure is large and unusually sensitive to temperature changes. Examples should be included to allow for comparison.

We thank the reviewer for pointing this out and we have tried to better emphasize this point in the first section of the Results. hHsp90 has no change while yHsp90 and TRAP1 do have temperature sensitivity, however TRAP1 appears to be more extreme. By collecting a new series of Δstrap temperature data (Figure 4D), we can now calculate an Arrhenius activation energy for both the WT and Δstrap variants of TRAP1. Arrhenius fits are now included in Figure 4E and with modifications included in the main text. We have also included a quantification of the activation energy for Hsp90 homologs (yHsp90 and Grp94) in Figure 4–figure supplement 2, utilizing previously reported rates found in Frey et al 2007. Comparing our calculated activation energy as well as a previously reported value (Leskovar et al 2008), we find that the other homologs have significantly lower activation energies. These data clearly show that TRAP1 has unusually large response to temperature changes.

2) The kinetic data should include fits and resulting rate constants (and also which equation was used) to judge quality of the kinetic model.

Agreed. The kinetic data in the manuscript now all include a plot of the fit used to determine the rate constant. The equations are now included in the methods section as well.

3) If the measured conformational change is indeed rate limiting then its temperature dependence should be the same as that of the ATPase activity measured before. Is this indeed the case? How is the relationship of the shortened variant?

We have better highlighted in the text our observation of the difference in closure rate (measured by FRET) with AMPPNP and ATP, and the steady-state ATPase rates. We find that the closure rate measured with ATP better matches the ATPase rate measured for the fully labelled Inter FRET probe used in experiments shown in Figure 6D and in the same buffer and temperature conditions (see Methods). Though we have not measured the temperature dependence of closure with ATP at the full range of temperatures as AMPPNP, we do show that the closure rate is slower at 25 °C and that the fold difference in closure rate is greater between WT and Δstrap protein at 25 °C (Table 3). The difference between the ATP analogs suggests that the energetics differ which could shift (but not mitigate) the observed temperature dependence of closure depending on the analog used. Importantly, we point out that despite experimental differences that could arise due to cysteine removal, labelling, or usage of varying nucleotide, our observation of the unique temperature dependent closure (also supported by Leskovar et al 2008) and the role of the strap as a structural element responsible for regulating this observation remains constant across all experiments in our manuscript.

Our proposal that closure is rate limiting is supported by previous studies (Hessling et al 2009 and Leskovar et al 2008) as well as our measurements in Figures 4 and 6, which allow us to evaluate the rate of closure, re-opening and hydrolysis in matching conditions for our WT and Δstrap FRET probes. We were able to decouple the closure and hydrolysis steps by removing MgCl2 from the closure reactions in the presence of nucleotide and find that closure is much slower than hydrolysis, with both WT and Δstrap hTRAP1 (Figure 6D and 6E, expounded upon within our manuscript). These experiments also show that removal of the strap has the greatest impact on the closure rate, a significant but smaller effect on re-opening, and a minor effect on hydrolysis.

We have additionally included FRET experiments with the Inter FRET probes to monitor the rates of closure ± the strap (Figure 4D). From this data an Arrhenius plot of both the WT and Δstrap proteins has been added in Figure 4E. Calculating the difference in Ea between WT and Δstrap we assign the contribution of the strap to Ea at approximately 60% of the measured Ea for WT hTRAP1 (48.8 kcal/mol Ea for WT; 29 kcal/mol for Δstrap). These data are consistent with the steady-state SAXS and ATPase, and show that removal of the strap region lowers the energy barrier between apo and the closed state.

4) : Do other Hsp90 isoforms also show a similar trend over a temperature range or is this special just for TRAP1?

This trend is not just specific for TRAP1 and we would like to point the reviewers to Leskovar et al 2008 and Krukenberg et al 2008. The sensitive temperature range and specific rates do vary dramatically among Hsp90 homologs, with TRAP1 displaying particularly heightened sensitivity (also see specific point 1 author response above). Regulation via the strap is the focus of our manuscript and the strap does not exist in the yHsp90 and bHsp90 homologs. In addition to TRAP1 and Grp94, the temperature sensitivity has also been recently reported for an Hsp90α alternative splice variant (Tripathi et al 2013) that importantly is imparted by a long N-terminal extension of ∼122 amino acids. We have made an effort to highlight these points in our discussion section.

5) Experiments in do not show kinetics, just a shift of equilibria.

We have changed the titles and Figure descriptions to better highlight these experiments as equilibrium experiments. These experiments initially suggested to us that TRAP1 might have a unique energy landscape, which we set out to elucidate the underlying mechanism of the phenomena.

6) EM images for delta strap should be included.

We have not collected EM images for Δstrap. After taking the initial EM images of the WT at various temperatures we moved to measure the % closed state by SAXS, which is a much more quantitative measure of conformational states and has previously be used to measure Hsp90 conformational equilibrium by our lab and others (Frey et al 2007).

7) In an Agard publication from 2008 (Southworth & Agard, Cell 2008) EM images of HtpG, yeast and human Hsp90 are shown in the apo state, with AMPPNP, and with ADP. The human Hsp90α is in an open conformation with AMPPNP at 37⁰C (according to them TRAP1 is fully closed at 37⁰C).

TRAP1 is the mitochondrial variant that shows noticeably different behavior from the cytosolic form of hHsp90 that was shown in the (Southworth & Agard, Cell 2008) manuscript. In this study we sought to bootstrap from our recent TRAP1 crystal structure (none are available for Hsp90α) to gain possible molecular insights into this difference. Cytosolic hHsp90 does not significantly close with AMPPNP at 37 °C suggesting a different energetic landscape and consequently a much lower ATPase activity, as highlighted in Southworth et al. It is important to note that although the energetics are different between Hsp90 homologs (likely due to different physiological environments, specific clients, and different requirements for co-chaperones) the underlying conformational states and mechanism of protein folding are conserved as also highlighted in Southworth et al.

8) Figure 2–figure supplement 1: Isn't this the same figure as just without the delta-strap?

Figure 2–figure supplement 1 (now Figure 2D) is meant to demonstrate that TRAP1 will remain closed even after cooling the sample back to 20°C for two hours. By this observation TRAP1 is kinetically trapped in the closed state after heating and will remain closed even when cooled. These data support a large temperature dependent barrier to the closed state that is overcome upon increasing temperature and a stable closed state once the transition has occurred rather than a pronounced temperature dependence of the equilibrium states. As mentioned above, the supplemental figure has now been combined with Figure 2, panel D, in an attempt to make this less confusing.

9) Is it known that the trans contacts that strap forms in the closed conformation stabilize the closed conformation?

From our previously published crystal structure of TRAP1 (Lavery et al 2014), cocrystallized with AMPPNP, we know that the strap makes substantial contacts with the trans-NTD while in the closed conformation, suggesting a role in stabilization. However, this is most directly shown by our new observation that deleting the strap accelerates reopening (Figure 6A).

10) : For the NTD:MD rotation, the FRET probe does not show a significant change in the FRET signal.

The change in FRET signal with the NTD:MD rotation probe is on par with previously published results using the same probes but with bHsp90 (Street et al, 2011). The measurements are quite reliable even though the delta for this set of probe positions is less than that for the cross protomer set.

11) : Why was saturation not reached for the 42⁰C sample?

It does. We have now included the full dataset in Figure 4B to better depict saturation at 42°C.

12) Data for the dimer closure FRET construct should be included and it would be important to see how delta-strap acts in the FRET assay at different temperatures to compare with the SAXS data.

We now show temperature dependent closure as measured with FRET using both the NTD rotation and dimer closure probes (Figure 4B). We observe that the dimer closure probes (Inter FRET) display comparable temperature dependent closure as the NTD rotation probe (Intra FRET). Most directly, we have now included a new temperature dependent closure series for the Δstrap variant (Figure 4D).

These measurements show a dramatic loss of temperature dependent closure and quantification of the activation energy difference (Arrhenius plot using WT and Δstrap FRET data) shows that the strap contributes over half of the WT Ea (Figure 4E).

The FRET data is in good agreement with the SAXS data, which shows an increase in closed state at higher temperatures measured after 1 hour at the respective temperature. By estimating the % closed state at 1 hour from the FRET data (Author response image 1) we get 9%, 42%, 66%, 90% and 87% (compared to 2%, 31%, 41%, 74% and 84% at the respective temperatures, Table 1). Considering the 1 hour time point for the Δstrap, the reaction has reached completion according to our FRET and SAXS measurements in Figure 3B and 4C/D, respectively. Although our normalized FRET data is only an estimate of the percent closed state molecules and shouldn’t be taken as a quantitate number, estimated percent closed state compared to the SAXS data show the same trend.

Author response image 1.

Adapted from Figure 4B (Intra FRET probe data set).

13) Why was the double strap mutant too unstable for NTD rotation FRET and not for inter-protomer FRET? Isn't it the same construct just with a different Cys label?

No, it is a different construct having two incorporated cysteine’s (S133C.E407C), within one protomer, whereas the inter-protomer FRET uses only 1 Cys per protomer (either E407C or E140C). We do not have a detailed explanation for why removing the strap in combination with the 2 Cys mutations in one protomer with the NTD rotation FRET pair destabilizes the protein. During purification of the Δstrap NTD FRET protein most was lost to cleavage products even when protease inhibitors were included.

The small amount of full-length protein that was purified did not appear unstable during experimental measurements, but given the unusual behavior during purification we did not have enough confidence in our measurements to report quantitative rates and draw conclusions from these measurements.

14) : The legend says that in the apo state the lid probe is in equilibrium between mobile and immobile states. In the Results section it is stated that the apo form is predominantly immobile.

We have reworded the legend to state that the apo state is in equilibrium between mobile and immobile as measured with EPR. After addition of AMPPNP there is a substantial change such that the mobile population dominates.

15) According to the authors, deleting strap compromises the stability of the closed structure and hence enhances the reopening rate and shifts the equilibrium towards the open state. If deleting strap shifts the equilibrium towards the open state, why is the delta-strap construct predominantly closed?

In this context, the authors mention that the effect of deleting the strap on the opening of the NTD interface is smaller than the effect on the kinetic barrier corresponding to the release of strap from the apo state. The respective numbers seem to be missing from .

The respective numbers that should be considered are listed in Table 3 under the Kclose and Kreopen heading for the CysFree hTRAP1 (E140C/E407C) and the Δstrap double (E140C/E407C) constructs. By taking the ratio of the rates, Kclose has a 16-fold difference while Kreopen has an 8-fold difference. This is further described in the main text in the FRET based “Dissecting further regulatory functions of the NTD-strap” section of the Results section in this manuscript. Thus deleting the strap destabilizes the open state more than it destabilizes the closed state. However, the overall equilibrium is still in favor of the open state in the absence of ATP.

16) Are the ATPase activities of the cys-free version of hTRAP1 and zebrafish TRAP1 mutant identical to the respective wild type proteins?

No. The cys-free versions of hTRAP1 and zTRAP1 are faster than WT in ATPase rates by 1.5–2 fold. This is presented in our previous manuscript (Lavery et al 2014). The molecular basis for the increased ATPase in the Cysteine Free TRAP1 constructs is unknown; however, all of our measurements are done by comparing WT to mutant (Δstrap, or strap point mutants) in matching conditions. We are interested and drawing conclusions from the fold change between mutant protein and the respective WT protein. As mentioned above we also point out that despite any differences that come about due to cysteine removal, labelling or nucleotide used in the experiments, our observation of the unique temperature dependent closure and the role of the strap as the structural piece responsible for these observation remains constant across all experiments in our manuscript.

17) . The legend says delta strap is ∼7 fold faster; text says ∼8 fold faster.

Thank you for pointing this out. This was a mistake and the difference has now been corrected. The legend now matches the text with “8-fold”.

18) : a trace showing that in presence of ATP without Mg2+, there is no ATPase activity should be added. Not adding Mg2+ is not necessarily equivalent to not having (ambient) Mg2+ present in the solution. The ATP induced changes in FRET signals should be measured also in the presence of EDTA.

The coupled NADH reaction used to measure ATPase rates is dependent on

Mg2+ so we are unable to do the comparable ATPase experiment in absence of Mg2+.

19) The authors mention that in the absence of Mg2+, ATP and AMPPNP show pronounced differences in kinetics of FRET signal, the difference with/without Mg2+ should be even more pronounced, that is the kinetics of o/c may be substantially faster in the presence of Mg2+. Measuring the FRET kinetics upon addition of Mg-ATP and Mg-AMPPNP is crucial to show that ATP induced closing kinetics in absence of Mg2+ are indeed representative for the ATPase cycle.

While the closure rate with ATP-Mg2+ would be ideal to compare to AMPPNPMg2+ (all FRET data in Figure 4 is done with AMPPNP-Mg2+), we are unable to measure closure as hydrolysis is faster than closure, hence the closed state does not build up in the presence of Mg2+. Rather, we measure a FRET signal that is always equivalent to Apo (no change from time zero). This was also seen in FRET studies with yeast Hsp90 (yHsp90) in Hessling et al, 2009. We noted the differences in the manuscript as an observation, however we feel uncovering the molecular reason for variability of rates and affinity between the two ATP analogs is beyond the scope of this manuscript. While interesting, here our focus is on the changes in these rates that are connected with the strap.

20) How does one know that Mg2+ can actually bind to the closed form; and that a re-open is not necessary for this to happen?

ATP hydrolysis upon addition of Mg2+ is relatively fast compared to the reopening rate 0.463 vs. 0.002 for WT protein or 13.3 vs. 0.016 for Δstrap protein. This strongly suggests that Mg2+ can bind to the closed state.

Reviewer #2

In this study, Partridge and colleagues investigate the role of the N-terminal extension (“strap”) of TRAP1, the mitochondrial Hsp90 isoform, which in the previously solved crystal structure of the TRAP1 dimer wraps around the N-domain of the opposite protomer. They characterized the effects of the strap on the dimer closure kinetics, the rotation of the N-domain relative to the M-domain, the ATP hydrolysis and movement of the ATP lid (N-domain) using negative stain electron microscopy (EM), small angle x-ray scattering (SAXS), fluorescence resonance energy transfer (FRET) and electron paramagnetic resonance (EPR) measurements. They demonstrate that the strap region is responsible for a temperature-dependent increase in the rate of TRAP1 closure, as well as the increase in the ATPase activity.

The data presented here is convincing and interesting. The physiological relevance of the observed phenomenon is not so clear since the major effects for human TRAP1 happen at temperatures below 37⁰C. Nevertheless, the story could be published after the authors addressed the raised issues.

We thank reviewer for the suggested experiments to strengthen our manuscript. We have attempted to address this concern, echoed by both the editor and Reviewer 1, by including new EM and SAXS data recorded above 37°C. It is clear from the SAXS data that the rate of closure continues to increase as temperatures increase above 37 °C. This is also clear from the FRET closure data (Figure 4B) which shows a significant increase in closure rate at 42°C. A more detailed description of the additional data and references has been described in the comments above for Reviewer 1.

Major comments

1) shows negative stain EM images of TRAP1 in the presence of AMPPNP pre-incubated at different temperatures. Few representative samples are picked from each grid to show the transition from open to closed conformation with the increase in temperature. The authors should quantify the open and closed structures from a representative square of the electron micrograph.

We very much agree that quantification of the %closed verses Apo state at each temperature is quite important and must be included. While we have done this in the past (Southworth, 2008), in practice, this is a somewhat painful and laborious procedure to do rigorously and instead here we chose to quantify the %closed state by SAXS. This is a significantly more quantitative assay for measuring equilibrium of states and has previously be used to measure Hsp90 conformational equilibrium by our lab and others (Frey et al, 2007). The quantification of %closed state as measured by SAXS can be seen in Figure 3 and Table 1.

2) Using SAXS the authors measured AMPPNP-induced transition of human and zebrafish TRAP1 to the closed conformation between 20 and 36⁰C (). To determine the physiological relevance of their observations the authors could have measured the AMPPNP induced transition of human TRAP1 at 37 to 42⁰C. Does human TRAP1 become more active at heat shock temperatures?

We now include SAXS data above 37°C and up to 43°C. Interestingly, TRAP1 from humans seems to have the largest jump in activity around 36-40°C, just where it might be most physiologically relevant. Analogously, the largest jump in % closed for TRAP1 from zebrafish seems to be in the 20-30°C range. Both species loose temperature sensitivity without the strap. However both species do show a jump in activity in Δstrap when going up to 43°C.

3) : The authors investigate AMPPNP-induced changes in TRAP1 conformation using FRET. Control experiments with only the acceptor dye need to be shown especially as the changes in fluorescence for the NTD:MD rotation seems to be very small. The temperature at which the experiment of has been performed should be mentioned in the figure legend.

Addition of nucleotide showed an appropriate anti-correlation of the donor/acceptor FRET signals and the anticipated direction of the change in FRET (increase in FRET for Inter FRET probe, and decrease in FRET for Intra FRET probe). The steady-state scans shown in Figure 4A were taken after the closure reaction was complete. Here, to avoid any temperature dependence on the dyes, etc., closure was induced by heat shock for 1hr and then the samples actually measured at room temperature. The temperature at which the scans were done has been added to the figure legend.

4) The dimer closure FRET experiments have been performed only at 30⁰C. As the paper deals with effect of temperature on ATPase rates of TRAP1, it would be very important to see the change in rate of dimer closure at different temperatures. It would also be interesting whether the rates of NTD-MD-rotation and dimer closure are similar to each other.

Because NTD-MD rotation is tightly coupled to closure, the closure rates measured by either probe set are similar; we had initially hoped to be able to tease apart these individual steps, but in practice, they seem kinetically inseparable (differences due to probe locations). We have now included the temperature dependent closure experiment (Figure 4B) and note that both probe sets are comparable in matching experiments. We chose to do the dimer closure FRET experiment +/- strap at 30 °C with SAXS, EM and ATPase as this is the temperature where we see the largest differences between WT and Δstrap in the temperature range assayed. Additionally, we have now included a temperature dependent closure series for the Δstrap variant (Figure 4D). As predicted, comparing the fold changes of closure rates (Table 3) at each temperature we see the largest fold change at lower temperatures (23 °C: 24-fold, 30 °C: 16-fold, 32 °C: 12-fold, 36 °C: 7-fold, and 42 °C: 3-fold). The clear impact of the strap is revealed from a comparison of the Arrhenius plots in Figure 4E.

5) In the authors write that the steady state ATPase rate for human TRAP1 was 0.463 min-1 and in the write that the closing rate for human TRAP1 at 30⁰C was 0.02 min-1. These values do not fit together and contradict the claims of the authors. The authors should indicate at which temperature the ATPase assays were performed and correlate the closing rate with the ATP hydrolysis rates to substantiate their claims. Maybe the authors will have to measure the closing rate upon addition of ATP instead of AMPPNP. This seems possible since omission of Mg prevents hydrolysis as the authors have shown.

We do recognize the significant difference between closure rates determined with AMPPNP and the ATPase measurements taken with ATP. For reference the best comparison should be done with the same protein and buffer conditions used in the FRET experiments with ATP as indicated in Table 2 and Table 3 (0.79 min-1 ATPase- Table 2 vs. 0.42 min-1 closure- Table 3). There is a strong nucleotide dependence on the closure rates, and the closed state never builds up with ATP plus Mg2+ as hydrolysis is faster than closure. However, for this manuscript we would like to focus on a more broadened and molecular aspect of the strap mediating temperature sensitivity to regulate closure and thereby activity of TRAP1 in both zebrafish and human. We have done matched experiments for WT and Δstrap for each assay and have drawn our conclusions from the differences between these matched experiments.

We show the temperature dependence of the closure reaction and the loss of temperature dependence upon deletion of the strap is robustly observed between multiple biophysical and biochemical of experiments. We ask that the reviewers please excuse our reluctance to dive even further into molecular differences between ATP analogs as we feel this is beyond the scope of our study.