1. Biochemistry and Chemical Biology

Controlling protein function by fine-tuning conformational flexibility

  1. Sonja Schmid  Is a corresponding author
  2. Thorsten Hugel  Is a corresponding author
  1. Institute of Physical Chemistry, University of Freiburg, Germany
  2. Signalling research centers BIOSS and CIBSS, Albert Ludwigs University, Germany
https://doi.org/10.7554/eLife.57180
Share this article
Cite this article
  1. Sonja Schmid
  2. Thorsten Hugel
(2020)
Controlling protein function by fine-tuning conformational flexibility
eLife 9:e57180.
https://doi.org/10.7554/eLife.57180
  2. Download BibTeX
  3. Download .RIS

Abstract

In a living cell, protein function is regulated in several ways, including post-translational modifications (PTMs), protein-protein interaction, or by the global environment (e.g. crowding or phase separation). While site-specific PTMs act very locally on the protein, specific protein interactions typically affect larger (sub-)domains, and global changes affect the whole protein non-specifically. Herein, we directly observe protein regulation under three different degrees of localization, and present the effects on the Hsp90 chaperone system at the levels of conformational steady states, kinetics and protein function. Interestingly using single-molecule FRET, we find that similar functional and conformational steady states are caused by completely different underlying kinetics. We disentangle specific and non-specific effects that control Hsp90’s ATPase function, which has remained a puzzle up to now. Lastly, we introduce a new mechanistic concept: functional stimulation through conformational confinement. Our results demonstrate how cellular protein regulation works by fine-tuning the conformational state space of proteins.

eLife digest

Proteins play a wide variety of roles in the cell and interact with many other molecules. The behavior of proteins depends on their structure; yet, proteins are often flexible and will change shape, much like a tree in the wind. Nevertheless, for some of the activities that it performs, a protein must adopt one specific shape. Therefore, the likelihood that the protein will take on this specific shape directly determines how efficiently that protein can perform a specific job.

The shape of a protein can be regulated by changes at several levels; these could include modifying one of the amino acid building blocks that make up that protein, binding to another protein, or by placing the protein in a part of the cell that is crowded with other large molecules. Schmid and Hugel wanted to understand how these three different types of regulation affect the structure of a protein and how they relate to its activities.

The protein Hsp90 was used as a test case. It typically exists with two copies of the protein bound together, either in a parallel or a V-shape. Hsp90 plays several important roles in metabolism and can break down molecules of ATP, the so-called energy currency of the cell. All three types of regulation favored the Hsp90 pairs taking the parallel structure and increased its breakdown of ATP. The results suggest that the Hsp90 pair has a flexible structure, and that reducing this flexibility can improve Hsp90’s efficiency in carrying out its role.

It was particularly unexpected that the large-scale, unspecific effect of placing the protein in a crowded environment could have such similar results to a small-scale, precise change of a single amino acid within the protein. While all three forms of regulation help to stabilize the parallel structure for Hsp90, they do this through different mechanisms, which influence the speed and the way that the protein transitions between the two structures. Schmid and Hugel believe that these results offer a new perspective on how diversely the shape and function of proteins is controlled at the molecular level, which could have wider implications for medical diagnostics and treatment.

Introduction

Protein function is essential for life as we know it. It is largely encoded in a protein’s amino-acid chain that dictates not only the specific 3D structure, but also the conformational flexibility and dynamics of a protein in a given environment. Precise regulation of protein function is vital for every living cell to cope with an ever-changing environment, and occurs on many levels pre- and post-translationally (Vogel and Marcotte, 2012; Hirano et al., 2016). After translation by the ribosome, protein function depends strongly on post-translational modifications (PTMs) (Deribe et al., 2010), but also on binding of nucleotides (Hodge and Ridley, 2016), cofactors (Weikum et al., 2017), various protein-protein interactions (PPIs) (Scott et al., 2016), and global effects, such as temperature (Danielsson et al., 2015), macro-molecular crowding and phase separation (Uversky, 2017), redox conditions (Levin et al., 2017), osmolarity (de Nadal and Posas, 2015) etc. Importantly, this regulation occurs on very diverse levels of localization. Global effects affect the whole protein non-specifically, PPIs act at a given interface and site-specific modifications are very localized. Nevertheless, all of them influence the molecular properties that determine the 3D conformation, the conformational dynamics, and thereby also the function of a protein (Schwenkert et al., 2014; Lewandowski et al., 2015; Schummel et al., 2016; Wei et al., 2016; Csizmok and Forman-Kay, 2018). The chaperone protein Hsp90 (Schopf et al., 2017) is an excellent test system to investigate diverse regulation mechanisms (Stetz et al., 2018). It was recently discussed that a single PTM can functionally mimic a specific co-chaperone interaction in human Hsp90 (Zuehlke et al., 2017). Here we take a next step and disentangle how a PTM-related point mutation, a co-chaperone interaction, and macro-molecular crowding affect the function, kinetics, and thermodynamics of this multi-domain protein.

Hsp90 is an important metabolic hub. Assisted by about twenty known cochaperones, yeast Hsp90 is involved in the maturation of 20% of the entire proteome (Taipale et al., 2010). Among its substrates (referred to as ‘clients’) are many kinases involved in signal transduction, hormone receptors, the 'guardian of the genome p53' (Lane, 1992), and also cytoskeletal proteins such as actin, tubulin, and many more (Pearl and Prodromou, 2006; Khandelwal et al., 2016). Cancer cells were found to be ‘addicted’ to Hsp90 (Trepel et al., 2010), which is therefor also a prominent drug target in cancer research. Hsp90 is a homo-dimer where each monomer consists of three domains (Ali et al., 2006): the N-terminal domain (N) with a slow ATPase function, the middle domain (M) believed to be the primary client interaction site (Verba et al., 2016), and the C-terminal domain providing the main dimerization contacts. Apart from closed conformations, where the three domains align in parallel, Hsp90 exists primarily in V-shaped, open conformations with dissociated N-terminal and middle domains (Krukenberg et al., 2008; Hellenkamp et al., 2017). Both global arrangements are semi-stable at room temperature. As a consequence, Hsp90 alternates constantly between open and closed conformations - even in the absence of the chemical energy source, ATP (Mickler et al., 2009; Schmid et al., 2016; Lee et al., 2019). Surprisingly, the characteristic conformational changes of Hsp90 are only little affected by e.g. anti-cancer drug candidates (Schmid et al., 2018) or natural nucleotides (Schmid et al., 2016). In addition, to the stress-induced isoform discussed herein (Hsp82), there is also a cognate isoform (Hsc82) in yeast, which differs in unfolding stability, client range and more, despite 97% sequence identity (Girstmair et al., 2019).

Here, we present three orthogonal ways to modulate Hsp90's conformational state space, illustrated in Figure 1. The investigated point mutation A577I is located in the C-terminal hinge region of Hsp90. Residue A577 is the equivalent of the post-translational S-nitrosylation site C598 in human Hsp90 (Martínez-Ruiz et al., 2005). While nitrosylation of that residue has a two-fold inhibitory effect – on the ATPase function and the client stimulation by human Hsp90 – the A577I mutation caused a nearly 4-fold amplification of the ATPase rate (Retzlaff et al., 2009). The fact that the point mutation is located in the C-terminal domain and the ATP binding site in the N-domain indicates a long-range communication, offering valuable mechanistic insight in Hsp90’s intra-molecular plasticity. Second, we consider the protein-protein interactions between Hsp90 and the activating co-chaperone Aha1, which is a well-known stimulator of Hsp90's inherently slow ATPase activity. It makes contacts to the middle and N-terminal domain, which rearranges the ATP lid (Schulze et al., 2016), and the catalytic loop (including R380) (Pearl, 2016) in a favorable way for ATP hydrolysis. The affinity of Aha1 for Hsp90 itself is also markedly enhanced by PTMs (Xu et al., 2019). The third way of modulation mimics the crowding encountered in the cell, which is full of proteins, nucleic acids, vesicles and organelles. We mimic cellular macro-molecular crowding using the common crowding agent Ficoll400, that is branched polymeric sucrose. In contrast to the previous two modulations, crowding represents a completely non-specific, physical interference.

Figure 1
Download asset Open asset
Protein regulation uses different degrees of localization.

Mutations or PTMs act most locally, protein-protein interactions (PPIs) act on the protein domain level, and changes in the global environment, such as crowding or phase separation, act non-specifically and globally on the protein. Each of them affects conformational thermodynamics and kinetics to fine-tune the protein conformational state space and thereby protein function.

At first sight, all three modulations provoke a similar steady-state behavior in Hsp90. But our single-molecule experiments allow us to disentangle the different underlying causes thereof.

Results

Mutation, cochaperone and crowding show similar thermodynamics

First we follow yeast Hsp90's conformational kinetics in real-time using single-molecule Förster resonance energy transfer (smFRET) measured on a total-internal reflection fluorescence (TIRF) microscope (Figure 2a). The FRET pair configuration displayed in Figure 2a (top) results in low FRET efficiency (little acceptor fluorescence) for V-shaped, open conformations of Hsp90, and high FRET efficiency (intense acceptor fluorescence) for more compact, closed conformations. This allows us to obtain steady-state information, like the population of closed conformations, and also the kinetics of conformational changes (Schmid et al., 2016; Roy et al., 2008; Lerner et al., 2018; Mazal and Haran, 2019). Example traces obtained from three Hsp90 molecules under different conditions are displayed in Figure 2b: First for the point mutant A577I, then in the presence of the cochaperone Aha1, and lastly under macro-molecular crowding by Ficoll400. The observed transitions between the low- and high-FRET states reflect global opening or closing. Figure 2c shows the steady-state (equilibrium) population of open and closed conformations. In all three cases, a shift toward closed conformations is observed with respect to the corresponding reference distribution obtained under equivalent experimental conditions (see caption of Figure 2 and Materials and methods for details). The optimal buffer conditions for each of the three cases differ slightly, which explains the differences between the three reference datasets displayed in Figure 2c (black lines). Relative populations and their uncertainties were obtained by bootstrapping (see Materials and methods and Supplementary file 1).

Figure 2 with 4 supplements see all
Download asset Open asset
Mutation, cochaperone interaction, and crowding show similar thermodynamic effects.

(a) Illustration of the single-molecule FRET experiment using an objective-type TIRF microscope (bottom): cross section through the objective, and flow chamber (both gray) and the dichroic mirror (black) separating the laser excitation (green) from the collected fluorescence (yellow). The zoom view (top) shows the fluorescently labeled Hsp90 (FRET donor, orange; acceptor red), which is immobilized on a PEG-passivated (dark blue) coverslip (light blue) using biotin-neutravidin coupling (black and gray). (b) Example time traces obtained from individual Hsp90 molecules for the point mutant A577I (top), in the presence of 3.5 µM cochaperone Aha1 (center), or under macro-molecular crowding by 20wt% Ficoll400 (bottom). Depicted are the FRET efficiency E (black), the fluorescence of the FRET donor (green) and acceptor (red) and the directly excited acceptor (gray). White and colored overlays denote low- and high-FRET dwells, respectively, as obtained using a hidden Markov model and the Viterbi algorithm. (c) FRET histograms compiled from many single-molecule trajectories as indicated, and normalized to unity (wt: wild type). Reference data (black) were measured under the specific conditions of each of the three experiment series (see also Materials and methods). Example traces of the reference data are provided in Figure 2—figure supplement 1. All fits and fit coefficients for 2 c are shown in Figure 2—figure supplement 2. Experimental conditions for A577I or wt Hsp90: 2 mM ATP in 40 mM Hepes, 150 mM KCl, 10 mM MgCl2, pH7.5. Conditions with and without 3.5 µM Aha1: wt Hsp90, 2 mM ATP in 40 mM Hepes, 20 mM KCl, 5 mM MgCl2, pH 7.5. Conditions with and without crowding by 20wt% Ficoll400: wt Hsp90 in 40 mM Hepes, 150 mM KCl, 10 mM MgCl2, pH7.5. The data were recorded in 5 to 13 videos per dataset, on one or more days. The number of individual molecules included per histogram are: A577I, 154; wt, 163; +Aha1, 366; -Aha1, 231; +crowding, 50; -crowding, 81. All smFRET traces are available from Figure 2—source data 1.

Figure 2—source data 1

smFRETdatasets.zip.

https://cdn.elifesciences.org/articles/57180/elife-57180-fig2-data1-v1.zip

The A577I mutation increased the closed population from (13 ± 3.5)% to (51 ± 5.5)%. This is a large effect considering that no big change in hydrophobic nor charged interactions is expected to result from this hydrophobic-to-hydrophobic point mutation. In particular, as the slightly bulkier isoleucine side chain points outward in the crystal structure of Hsp90’s closed conformation (Ali et al., 2006). In addition to the A577I homodimer, already the A577I/wild type (wt) hetero-dimer shows a considerably larger population of closed conformations, especially in the presence of ATP (Figure 2—figure supplement 3 (left)). Under ADP conditions the additive effect is also observed, but weaker (Figure 2—figure supplement 3 (right)). Under both conditions, the second A577I in the homodimer leads to a further shift toward closed conformations. The slight but consistent shift of the corresponding low-FRET peak in Figure 2c (left) could be explained by a sterical hindrance of the farthest opening in the A577I homodimer. Furthermore, very fast transitions at the temporal resolution limit (200 ms) occurred more frequently, which can be seen by the increased overlap between the two populations. Both, sterical hindrance and faster transitions, can be interpreted as a global stiffening of Hsp90’s structural core formed by the C- and middle domain. The interaction with Aha1 (Figure 2c, center) forms inter-domain (N–M) and inter-monomer contacts. The latter increase the affinity for N-N binding, and thus cause Hsp90 to shift from (24 ± 4.0)% closed to (48 ± 5.0)% closed population, which is in line with previous qualitative reports (Hessling et al., 2009; Wortmann et al., 2017). Lastly, macro-molecular crowding increased the closed population from (11 ± 4.5)% to (50 ± 10.5)%, in agreement with previous ensemble findings (Halpin et al., 2016). In contrast, to the A577I mutant, crowding slowed down fast fluctuation at the resolution limit, which creates well-separated populations in Figure 2c). The induced shift toward closed conformations appears in a concentration-dependent manner, as can be seen in Figure 2—figure supplement 4. In contrast to polymeric, branched sucrose (Ficoll400), monomeric sucrose had only negligible effect on the steady-state populations. This proves that macro-molecular crowding is the cause of the observed population shift, and a biochemical glucose-associated reason can be dismissed.

In all three cases, a clear shift toward closed conformations is observed, although to slightly different extents. Importantly, based on these distributions alone, the energetic origin of the population shift remains unclear. That is whether the observations arise from a stabilization of closed conformations, or a destabilization of the open conformations, or even a combination of both. To answer these questions, we solved the full kinetic rate model, presented below.

Same thermodynamics, different conformational kinetics

Figure 3a shows the kinetic rate models describing Hsp90's global opening and closing dynamics. The significant changes caused by each type of modulation are highlighted in red and green as indicated. The corresponding quantitative rate changes and confidence intervals are displayed in Figure 3b. We used the Single-Molecule Analysis of Complex Kinetic Sequences (SMACKS [Schmid et al., 2016]) to quantify rate constants and uncertainties directly from the smFRET raw data. For Hsp90’s global conformational changes, we consistently infer 4-state models (Schmid et al., 2016; Schmid et al., 2018; Schmid and Hugel, 2018): two low-FRET states (open conformations) and two high-FRET states (closed conformations). Although only two different FRET efficiencies can be resolved, at least four kinetic states are needed to describe the observed kinetic heterogeneity. Based on recent results (Hellenkamp et al., 2017), we expect an entire ensemble of open sub-conformations that - on the timescale of the experiment - are sufficiently well described by two kinetically different low-FRET states. The two closed states, one short-lived (state 2) and one longer-lived (state 3), likely differ in local conformational elements (Giannoulis et al., 2020; Huang et al., 2019). The well-known N-terminal beta-sheet with or without its cross-monomer contacts (as observed in the closed crystal structure [Ali et al., 2006]) could explain the additional stabilization of state 3 with respect to state 2. All states 0, 1, 2, 3 of this highly dynamic Hsp90 system represent conformational ensembles that are defined by their FRET efficiency and their kinetics (rather than discrete or ‘frozen’ conformations).

Figure 3 with 3 supplements see all
Download asset Open asset
Different conformational kinetics cause similar thermodynamics.

(a) Kinetic rate models observed for the point mutant A577I, the cochaperone Aha1, the macro-molecular crowding agent Ficoll400 - each compared to the reference data: wild-type, no Aha1, no crowding, respectively. Significant differences to the reference are highlighted in red and green. Conformational kinetics are described by four states: states 0,1 represent open conformations, and 2,3 are closed conformations. Large and small arrows and circles indicate the size of rates and populations, respectively. For crowding, only 3 links are found. (b) The relative rate change under the three conditions in (a) with respect to the reference. Gray boxes show the 95% confidence interval of the reference data. Transition names and color code as in (a). All values are listed in Supplementary file 1C. The molecule counts are the same as for Figure 2.

In the case of the cochaperone Aha1, the shift in the FRET efficiency histogram originates from opposing changes of the fast rates between states 1 and 2 (Figure 3—figure supplement 1). This is in contrast to the effect of the C-terminal point mutation A577I, which collectively accelerates the pathway from state 0 via state 1 and state 2 to state 3. Note that a kinetic model with only three links - similar to the model for crowding in Figure 3a, right - is statistically sufficient (according to likelihood ratio testing detailed in Schmid et al., 2016 SI point 1.2) to describe the observed kinetics of the A577I homodimer, implying less kinetically heterogeneous fluctuations (Figure 3—figure supplement 2).

Under macro-molecular crowding, changes of the conformational dynamics are visible already from the stationary distributions: as shown in Figure 2a, the low and high FRET populations are most separated in this case. This is indicative of fast fluctuations, at or below the timescale of the sampling rate (5 Hz) that are slowed down at higher viscosity. Still, transitions between open and closed conformations are regularly observed in the experiment (Figure 3—figure supplement 3). For the fully resolved kinetics, the main difference is observed for the rates between the closed states 2 and 3. This agrees with the increase of the closed population under macro-molecular – but not small molecular – crowding. Altogether, Figure 3 shows three completely different kinetic effects that underlie seemingly analogous ensemble behavior.

Based on the complete kinetic rate models, we can now deduce the impact on free energies along a specific spatial reaction coordinate, here the N-terminal extension (Figure 4). The point mutation A577I causes an asymmetric destabilization of open conformations. Aha1 leads to simultaneous destabilization of open conformations and stabilization of closed conformations, whereas macro-molecular crowding only stabilizes closed conformations. Importantly, this information is not accessible from the steady-state distributions in Figure 2c.

Figure 4
Download asset Open asset
Three contrasting effects on Hsp90's conformational energy landscape.

(a) The open conformation is destabilized by the A577I mutation. (b) Aha1 inversely affects both equilibria. (c) Macro-molecular crowding stabilizes the closed conformation. The dashed black line indicates the reference. This mechanistic information was obtained from all six rate models represented in Figure 3, it is not accessible from Figure 2c alone. To illustrate the changes on the global energy landscapes, we chose the transition state as a reference for the measured relative energies.

ATPase stimulation – to varying degrees

The collective shift toward closed conformations is accompanied by an overall increase in ATPase activity under all three conditions (Figure 5): 7-fold for A577I, 17-fold for Aha1, 4-fold for crowding, respectively. Thus, the increase in ATP hydrolysis rate does not reflect the increase in the closed population observed in Figure 2c. This is a first indication of causalities that involve more than just the occurrence of closed conformations. In the following, we dissect the molecular origins of the increased ATPase activity. The effect of macro-molecular crowding can serve as an estimate of the ATPase stimulation caused exclusively by the relative stabilization of closed conformations, because a biochemical interaction of sucrose was excluded in control experiments (see above). Remarkably, the entirely non-specific interaction leads already to a considerable ATPase acceleration of a factor 4. This supports the wide-spread notion that the closed conformation represents Hsp90's active state (Prodromou, 2012). But, in comparison to the biochemical effect of the cochaperone Aha1, the stimulation by crowding is still modest, despite the much larger closed population. Specifically, the 4.5-fold increased closed population, comes with a 4-fold increased ATPase activity, whereas in the presence of Aha1 already a 2-fold increased closed population is accompanied by a 17-fold ATPase stimulation. The fact that Aha1 induces the smallest increase in closed population, but under the same conditions the largest ATPase stimulation, highlights the functional importance of specific contacts between Aha1 and Hsp90, which are responsible for 88% of the ATPase stimulation by Aha1.

Figure 5
Download asset Open asset
ATPase stimulation by local and global modulations.

Namely by the point mutation A577I (Retzlaff et al., 2009), in the presence of 3 µM Aha1 (Jahn et al., 2014) and 20wt% Ficoll400, respectively. The ATPase stimulations are normalized by the activity of the wild type, without Aha1 and without crowding, respectively.

In the case of the A577I mutant a 3.9-fold increased closed population comes with a 7-fold ATPase amplification. This could result from the mentioned hindrance of extremely open states indicating a conformational stiffening, restricting Hsp90's native flexibility. In other words, not only changes in the equilibrium, but also changes in the kinetics affect the ATPase activity as discussed below.

A comparison of the observed changes on the conformational steady-state and the ATPase rate allows us to estimate the contribution of conformational confinement in each case (Supplementary file 1B). We find that conformational confinement accounts for 100% of the ATPase stimulation induced by macro-molecular crowding, for 60% of the stimulation by the point mutation A577I, and for only 10% of the stimulation by cochaperone Aha1. The contribution of conformational confinement is thus largest for purely non-specific interactions, while additional allosteric effects may increase the stimulation by A577I, and specific rearrangements of the active site dominate the stimulating action of cochaperone Aha1. Our data do not indicate additional combinatorial effects of these three stimulations, which can be investigated in a future study.

Discussion

Herein we compare three types of Hsp90 modulations, spanning a wide range from a site-specific point mutation, via cochaperone binding, to completely non-specific macro-molecular crowding. All three modulations provoke a similar steady-state (equilibrium) behavior, namely an increase in Hsp90's closed conformation and in Hsp90's ATPase rate. In contrast, they show significant differences in their kinetics, which could be revealed by single-molecule FRET. This can be rationalized by the emerging picture of yeast Hsp90, as a very flexible dimer that relies critically on external assistance (e.g. by cochaperones) to control its non-productive flexibility. For example, Hsp90’s ATPase function requires the concerted action of the N-terminal nucleotide binding pocket with the ATP lid and distant elements such as the catalytic loop of the middle domain and parts of the opposite N-domain (the N-terminal β1-α1 segment). These elements – also called the catalytic unit (Prodromou, 2012) - however, are very flexible, such as the entire multi-domain dimer. Consequently, anything that constrains this flexibility and confines Hsp90 in a more compact conformation, has a high potential to increase the combined probability for such a concerted action - be it by specific or even non-specific interaction.

Figure 6 shows that the functional impact of conformational confinement can be understood as a direct result of combinatorics: in a flexible protein such as Hsp90, the catalytically active elements have many translational and rotational degrees of freedom. Thus, the probability for a certain hydrolysis-competent conformation is very small. It is however increased dramatically by conditions that constrain these degrees of freedom – even non-specifically – and localize the catalytically active elements. This notion can be further extended to cochaperone binding, and it also explains mutual effects upon client interaction. We conclude, while Hsp90’s flexibility may facilitate its numerous interactions with diverse clients and cochaperones, the flexibility itself has substantial off-state character regarding the ATPase function of Hsp90. In turn, by directly affecting the lifetime of the nucleotide/active site contacts and the resulting structural stabilization, ATP hydrolysis controls also longer-range allosteric effects linked to these residues.

Figure 6
Download asset Open asset
Stimulation of protein function by conformational confinement - a result of combinatorics.

Enzymatic binding pockets of proteins are often composed of flexible elements with many degrees of freedom, (a) left. All of these have to adopt a specific 3D configuration to form the catalytically active state, (a) right. For example in Hsp90, the hydrolysis competent state includes direct contacts with residues from different domains: G100, T171, D79, N37, E33, R380, and G121-F124 (counter-clockwise from top, in the left panel). As a result, the probability to be in the catalytically active state is very small, compared to the combined probability for all other accessible conformations (left), and the catalytic activity is low. Intriguingly, specific or non-specific confinement reduces the non-productive degrees of freedom in Hsp90, leading to a relative stabilization of the hydrolysis-competent state and therefore stimulation of the catalytic activity, as illustrated in (b). Here we have shown that this can occur through allosteric effects of mutations, protein-protein interaction, or macromolecular crowding. Structures were visualized with Pymol, based on pdb entry 2cg9.

A closer look at the regulation of Hsp90's conformational energy landscape by single-molecule FRET shows the many ways to reach similar ensemble results. The point mutation A557I destabilizes the open conformation, most probably by preventing access to a subset of the conformational ensemble of open states. This leads to an overall stiffened structure (more confined) of the A577I homodimer and therefore less extensive random walks, that is a more streamlined conformational transition toward the closed conformations. Macro-molecular crowding stabilizes the closed conformation by simple, sterical confinement. The cochaperone Aha1 combines both mechanisms with additional specific rearrangements. In theory, all three modulations can lead to the exactly same thermodynamic observation, and in fact we observe very similar steady-state distributions. Nevertheless, the kinetics may still vary significantly – as experimentally demonstrated herein. This is an important mathematical fact that holds true for all protein systems.

Moreover, our findings indicate that regulation by cochaperones - and protein-protein interactions in general - can have far-reaching thermodynamic, kinetic, and functional consequences. Some of them can possibly be mimicked by individual point mutations, but other consequences might be missed. Lastly, as shown in this work, already non-specific, purely physical macro-molecular crowding has strong effects on thermodynamics, kinetics and function, therefore caution is advised when relating in vitro findings to in vivo function. Certain mutations that show little or no ATPase activity in vitro, may be ‘rescued’ by the conformational confinement in the crowded cellular environment, and still be ATPase active in vivo. As demonstrated herein, in many in vitro experiments macro-molecular crowding could easily be included.

In conclusion, we demonstrated herein three ways of protein regulation, ranging from site-specific localized to global modulations. All three ways show very similar thermodynamic observations, which are, however, caused by clearly different conformational kinetics. This is direct evidence for the importance of kinetics, and of the dynamic structure–function relationship in proteins. The reduction of non-productive structural flexibility stimulates Hsp90’s ATPase function even by entirely non-specific means. Our findings demonstrate that functional stimulation as a result of conformational combinatorics plays an important role in protein regulation. We anticipate that such conformational confinement - by localized or global modulations - is an important mechanistic concept with wide-spread implications for protein function in diverse systems.

Materials and methods

Protein construct preparation

Request a detailed protocol

Yeast Hsp90 dimers (UniProtKB: P02829) with a C-terminal coiled-coil motif (kinesin neck region of D. melanogaster) were used to avoid dissociation at low concentrations (Mickler et al., 2009). Previously published cysteine positions (Hessling et al., 2009) allowed for specific labeling with donor (61C) or acceptor (385C) fluorophores (see below). Point mutation A577I was introduced using QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies). The constructs were cloned into a pET28b vector (Novagen, Merck Biosciences, Billerica, MA). They include an N-terminal His-tag followed by a SUMO-domain for later tag cleavage. The QuickChange Lightning kit (Agilent Technologies, Santa Clara, CA) was used to insert an Avitag for specific in vivo biotinylation at the C-terminus of the acceptor construct. Escherichia coli BL21star cells (Invitrogen, Carlsbad, CA) were cotransformed with pET28b and pBirAcm (Avidity Nanomedicines, La Jolla, CA) by electroporation (Peqlab, Erlangen, Germany) and expressed according to Avidity’s in vivo biotinylation protocol. The donor construct was expressed in E. coli BL21(DE3)cod+ (Stratagene, San Diego, CA) for 3 hr at 37°C after induction with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at OD600 = 0.7 in LBKana. A cell disruptor (Constant Systems, Daventry, United Kingdom) was used for lysis in both cases. Proteins were purified as published (Jahn et al., 2014) (Ni-NTA, tag cleavage, anion exchange, size exclusion chromatography). 95% purity was confirmed by SDS-PAGE. Fluorescent labels (Atto550- and Atto647N-maleimide) were purchased from Atto-tec (Siegen, Germany) and coupled to cysteins according to the supplied protocol. If not stated differently, all chemicals were purchased from Sigma Aldrich, St. Louis, MO.

Single-molecule FRET (smFRET) was measured as previously detailed using a home built TIRF setup (Schmid et al., 2016). Hetero-dimers (acceptor + donor) were obtained by 20 min incubation of 1 µM donor and 0.1 µM biotinylated acceptor homodimers in measurement buffer (40 mM Hepes, 150 mM KCl, and 10 mM MgCl2, pH7.5) at 47°C. This favors biotinylated heterodimers to bind to the polyethylene glycol (PEG, Rapp Polymere, Tuebingen, Germany) passivated and neutravidin (Thermo Fisher Scientific, Waltham, MA) coated fluid chamber. Residual homodimers are recognized using alternating laser excitation (ALEX) of donor and acceptor dyes (Lee et al., 2005; Hellenkamp et al., 2018) and excluded from analysis. For optimal interaction affinity with Aha1, measurements were performed in low salt buffer (40 mM Hepes, 20 mM KCl, 5 mM MgCl2, pH 7.5 with 3.5 µM Aha1 and 2 mM ATP). For comparison, data without Aha1 were measured accordingly. Notably, significant binding was previously found for Aha1 with labeled Hsp90-385C at much lower concentration of 0.3 µM (Li et al., 2013), which is exactly the dissociation constant reported for unlabeled Hsp90 (Panaretou et al., 2002). This implies that, although not directly detectable in the experiment, Hsp90 exists predominantly in complex with Aha1 under the used conditions. A577I/wt constructs were created through monomer exchange (see above). They are distinguished from both kinds of homodimers through the fluorescence signal (donor+acceptor). Measurements were performed in measurement buffer plus 2 mM ATP if not stated differently. Macro-molecular crowding was mimicked by 20wt% polymeric sucrose, known as Ficoll400 (Sigma Aldrich) in measurement buffer if not stated differently.

Single-Molecule FRET data analysis

Request a detailed protocol

Single-molecule FRET trajectories were selected, and corrected for background, donor leakage, acceptor direct excitation, as well as different fluorescence quantum yields, detection efficiencies, excitation efficiencies, and laser powers, as described before (Hellenkamp et al., 2018).

The uncertainties of steady-state populations were estimated by bootstrapping: out of each dataset, random subsets (or samples) containing two thirds of the total number of traces were drawn with replacement. A histogram was compiled for each subset, and fitted with a double-Gaussian fit function and fixed peak positions from Figure 3—figure supplement 2. Relative populations were obtained for each histogram. This was repeated for 1000 subsets. We report the means of 1000 such populations, and as an uncertainty measure their standard deviations (rounded upwards to 0.5). See also Supplementary file 1A.

SMACKS is based on two-dimensional semi-ensemble Hidden Markov models (HMMs), as described in detail earlier (Schmid et al., 2016). In brief, smFRET donor and acceptor fluorescence traces serve as input data. SMACKS comes in two steps. First, individual HMMs are optimized for each molecule separately, to allow for the experimentally observed intensity differences between different molecules. The resulting emission parameters for each molecule – that is Gaussian mean and variance representing the fluorescence intensity levels - are then used for the second step: the semi-ensemble HMM optimization. Now the kinetic parameters – so-called start and transition probabilities – are ensemble-optimized based on an entire dataset, using the individual emission parameters of the first step. The transition probabilities are converted into rate constants by considering the sampling rate of the experiment (frames per second). Several such HMM optimizations with increasing model size (i.e. numbers of states) are calculated, and compared using the Bayesian Information Criterion (BIC), which reproducibly yields four states for Hsp90’s characteristic conformational dynamics: two open, two closed. Within the four-state model, we determine the most likely number of links using likelihood ratio tests following the hierarchical search for simplified models described in Bruno et al., 2005.

Activity assay

Request a detailed protocol

ATPase activity was measured at 37°C coupled to NADH oxidation, which was followed as a decrease in absorption at 340 nm using an ATP regenerative assay similar to Tamura and Gellert, 1990: 0.2 mM NADH Di-Na (Roche, Basel, CH); 2 mM phosphoenol pyruvate K-salt (Bachem, Bubendorf, CH); 2 U/ml pyruvate kinase, Roche; 10 U/ml lactate dehydrogenase, Roche; in 40 mM Hepes, 150 mM KCl, 10 mM MgCl2, pH 7.5.

Data availability

All single-molecule FRET traces for each dataset are provided as source data (smFRETdatasets.zip). The source code of SMACKS is available for download here: https://www.singlemolecule.uni-freiburg.de/SMACKS.

References

    1. Prodromou C
    (2012) The ‘active life’ of Hsp90 complexes
    Biochimica Et Biophysica Acta (BBA) - Molecular Cell Research 1823:614–623.
    https://doi.org/10.1016/j.bbamcr.2011.07.020
    1. Tamura JK
    2. Gellert M
    (1990)
    Characterization of the ATP binding site on Escherichia coli DNA gyrase affinity labeling of Lys-103 and Lys-110 of the B subunit by pyridoxal 5'-diphospho-5'-adenosine
    The Journal of Biological Chemistry 265:21342–21349.

Decision letter

  1. Maria Spies
    Reviewing Editor; University of Iowa, United States
  2. José D Faraldo-Gómez
    Senior Editor; National Heart, Lung and Blood Institute, National Institutes of Health, United States
  3. Michael Schlierf
    Reviewer; TU Dresden, Germany
  4. Timothy Craggs
    Reviewer; University of Sheffield, United Kingdom

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Controlling protein function across scales – from point mutation to crowding" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and José Faraldo-Gómez as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Michael Schlierf (Reviewer #2); Timothy Craggs (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data. In the case of your paper, the manuscript can benefit from some combinatorial experiments (see suggestions below).

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

Schmid and Hugel present a detailed smFRET study on HSP90 conformational kinetics as it transitions between four conformational states. The authors investigate the effects of mutations, co-chaperones and crowding on state population and kinetics. What add significance to the study is a demonstration of the effects of molecular crowding, which is used to simulate conditions resembling cellular environment. Most prominent finding here is a demonstration that variations in experimental conditions (e.g. mutations or crowding) may have similar thermodynamic effects as revealed via population experiments, while very different kinetic effects. Thus, the study highlight the importance to discriminate steady-state thermodynamic outputs from kinetic effects and the ability of smFRET approach to extract both parameters using the same set of data.

All reviewers and the editor agree that the study is important and well executed. The results are interesting and novel, and the concepts discussed are applicable to many conformationally dynamic protein systems.

Essential revisions:

While the reviewers agree that there is no need for essential additional data, we felt that if you have some combinatorial experiments, e.g. Aha1 and crowding to support their claims of combinatorial effects (Figure 6). Otherwise, you could weaken some of the concluding claims.

Please address the following questions regarding the interpretation and data analysis:

1) Figure 2 and data therein:

Having the conditions for the three different experiments in the legend would be helpful. Does this explain the difference in the reference data (i.e. the ionic strength effects the overall conformational profile)? It would be useful to provide some quantitation of the mid-point of the FRET populations (or present them in a vertical stack to aid comparison.) Are the reference data for Figure 2C plots 1 and 3 measured under identical conditions? If so, is the difference between these data significant?

2) In Figure 2B, please show representative traces for WT sample without cochaperone or crowding factor; change the format of all the traces so that the idealization quality can be checked visually by reader. The fitting of time traces to determine the kinetics of transitions is central to the presented work. Please provide some detail as to how the SMACKS algorithm works. Does it involve a global fit to ALL the data? From the results presented, it is not clear how you rule out transitions between, for example state 0 and state 2. Is this just what falls out of the data analysis?

3) 4 states are proposed to fit the kinetic model best. You assume that HSP90 is adapting the same identical molecular conformations, meaning state 0, 1, 2, and 3 are the same states in presence or absence of e.g. the co-chaperone. However, the peaks in the FRET efficiency histograms, especially of the high FRET efficiency state, but also the low FRET efficiency state (in A577I) are strongly shifted in presence of all modifications, which indicate a slightly different molecular conformation. For a strict comparison of free energy landscapes as suggested in Figure 4 no common reference state is therefore available. Please discuss whether the kinetics states are necessarily different conformations to which your current signal is blind?

4) As a common reference state, the transition state between the N-terminal compact and N-terminal expanded state id used (Figure 4). Is there any evidence that this is possible, why not the compact or expanded state?

5) Figure 4: From your kinetics you have some idea of the rates of the transitions between the 2 open conformations and the two closed conformations which you could show in a supplementary figure? This is potentially helpful because for example Figure 4C it is really state 3 that is stabilized with respect to state 2.

6) Each measured reference should be identical. Yet, the population of the high FRET efficiency state varies from 14 to 29%. This is a two-fold difference, which is also what kinetically is named as a significant difference in transition rates. Can you explain this difference, explain how the state occupancies in the subsection “Mutation, cochaperone and crowding show similar thermodynamics” have been determined and what the relative error is? It would further be good to also show example traces of the reference experiments with wt in the figure supplements to give the reader a chance for comparison beyond the state histogram and rate summaries.

7) Please expand the following discussion: What does the current work allow the authors to conclude about the function of the ATPase activity in HSP90? Especially delineate the physico-chemical conclusions from the functional conclusion.

8) Figure 6 suggests combinatorial effects. By extension, this suggests that crowding and the point-mutation would have an even more enhanced effects of transition rates or ATPase activity. Can you show evidence by a combination of any two of their conditions and support the combinatorial suggestion from Figure 6?

9) A clearer description of the data analysis needs to be placed in the Materials and methods section.

10) Statistical errors are reported for the interconversion rates between the various conformational states detected. Please also specify whether the results of each experimental condition were obtained from a single experiment (i.e. video) recorded in a single day, or that the results of each experimental condition were obtained from a set of videos recorded on different days to check for reproducibility? Regarding the state populations an error of the extracted percentages would be helpful. This can be done e.g. by bootstrapping.

https://doi.org/10.7554/eLife.57180.sa1

Author response

Essential revisions:

While the reviewers agree that there is no need for essential additional data, we felt that if you have some combinatorial experiments, e.g. Aha1 and crowding to support their claims of combinatorial effects (Figure 6). Otherwise, you could weaken some of the concluding claims.

Thank you for pointing out that Figure 6 could be misunderstood as claiming combinatorial effects, e.g. of Aha1 and crowding. This was not our intention: please note the ‘or’ in the legend of Figure 6: “Intriguingly, specific or non-specific confinement reduces the non-productive degrees of freedom in Hsp90,…” and also: ”Here we have shown that this can occur through allosteric effects of mutations, protein-protein interaction, or macromolecular crowding.”

Since this is indeed an important point, we added the following text to clarify that herein we do not claim combinatorial effects of point mutation, cochaperone interaction, and macro-molecular crowding. In contrast, we describe how conformational confinement plays a role in all three cases:

“A comparison of the observed changes on the conformational steady-state and the ATPase rate allows us to estimate the contribution of conformational confinement in each case (see Table 2 in Supplementary file 1). […] Our current data does not indicate additional combinatorial effects of these three stimulations, which can be investigated in a future study.”

Lastly, we found a mistake in Figure 6, which depicted the representation of the point mutation A577I in place of the co-chaperone Aha1 – and vice versa. We corrected this mistake, such that Figure 6 is now in line with all the data, and the text.

Please address the following questions regarding the interpretation and data analysis:

1) Figure 2 and data therein:

Having the conditions for the three different experiments in the legend would be helpful.

We thank the reviewers for their suggestion, and added the experimental conditions now also to the caption of Figure 2, which will help the reader to find them.

Does this explain the difference in the reference data (i.e. the ionic strength effects the overall conformational profile)?

Yes, the different measurement conditions explain the differences between the three reference datasets. E.g. the lower salt concentration that is necessary for optimal Aha1 binding, leads to a slightly larger high-FRET (closed) population (as reported previously in Schmid and Hugel, 2018; Figure 2C). We added the following sentence to the main text to clarify this:

“The optimal buffer conditions for each of the three cases differ slightly, which explains the differences between the three reference datasets displayed in Figure 2C (black lines).”

It would be useful to provide some quantitation of the mid-point of the FRET populations (or present them in a vertical stack to aid comparison.)

We thank the reviewers for this suggestion. We present the panels vertically now, to better visualize the mid-points of the FRET population. In addition, we added a new Figure 2—figure supplement 2 showing all double-Gaussian fits and their fit coefficients.

Are the reference data for Figure 2C plots 1 and 3 measured under identical conditions? If so, is the difference between these data significant?

No, Figure 2C plot 1 includes 2 mM ATP, while plot 3 does not, as described above. We thank the reviewers for pointing out that this was not clear enough. This important information is now not only found in the Materials and methods section, but also in the legend of Figure 2.

2) In Figure 2B, please show representative traces for WT sample without cochaperone or crowding factor.

We added a new Figure 2—figure supplement 1 with representative traces of all three reference datasets.

Change the format of all the traces so that the idealization quality can be checked visually by reader.

We are not sure what the reviewers mean by ‘idealization quality’. It could mean a likelihood score for each datapoint’s assignment to a certain state (0 to 3). The γ probabilities (after Rabbiner’s nomenclature for Hidden-Markov models; Rabiner (1989) Proc IEEE) could provide a measure for this. However:

i) SMACKS does currently not output this information. We agree such a feature could give a relative rating of state assignment reliability per datapoint, and we gratefully keep this suggestion in mind for future versions of SMACKS.

ii) It is not the standard in the field: we know exactly one smFRET kinetics publication, which does report such ‘idealization quality’ albeit for a single-trace analysis: Greenfeld et al. (2012) Plos ONE. Recent smFRET kinetics papers do not report such a quantity, e.g.: Barth,.…, Lamb (2020) JACS;

Fitzgerald,.…, Blanchard (2019) Nature; Zhou,.…, Ha (2019) Nat Chem Biol; Schärfen, Schlierf (2019) Methods; Dulin,.…, Kapanidis (2018) Nat Comm; Kilic,.…, Seidel, Fierz (2018) Nat Comm; van de Meent,.…, Gonzalez (2014) Biophys J.

iii) The uncertainty measures for the kinetic analysis are already provided in the form of confidence intervals for each rate constant specified in Table 1 in Supplementary file 1.

The fitting of time traces to determine the kinetics of transitions is central to the presented work. Please provide some detail as to how the SMACKS algorithm works.

We agree that adding some more detail improves the readability of this manuscript, although SMACKS has been described before in great detail (Schmid, Götz and Hugel, 2016). We added the following paragraph to the Materials and methods section:

“SMACKS is based on two-dimensional semi-ensemble HMM, as described in detail earlier [Schmid, Götz and Hugel, 2016]. In brief, smFRET donor and acceptor fluorescence traces serve as input data. ][…] Within the four-state model, we determine the most likely number of links using likelihood ratio tests following the hierarchical search for simplified models described in [Bruno et al., 2005].”

Does it involve a global fit to ALL the data?

Yes, SMACKS involves a global HMM optimization of the transition probabilities based on all molecules in a given dataset (however not on all six datasets at once).

From the results presented, it is not clear how you rule out transitions between, for example state 0 and state 2. Is this just what falls out of the data analysis?

For two open (o), and two closed (c) states, the maximum number of mathematically identifiable links is four (eight rate constants) – independent of data quality. For four links, there are two mathematically equivalent cyclic models (-o-o-c-c- or -o-c-o-c-); in the case of three links, there are three mathematically equivalent models, as discussed in Schmid and Hugel, 2018, in Figure 5A, and previously for ion channel data by Bruno et al., 2005. To select one of the mathematically equivalent models, we used prior knowledge about Hsp90’s structure, which led to the presented models with the fewest open-to-closed links, for three reasons:

i) The long-distance transition path of ca. 4nm between the observed open and closed conformations renders the lowest number of open to close transitions most likely.

ii) From confocal smFRET results (Hellenkamp et al., 2018), we know that the two open states represent a much larger conformational ensemble with shallow energy barriers. Hence, transitions within the conformational ensemble of open states are to be expected.

iii) The long-lived / short-lived closed states are likely the result of added / missing cross-monomer contacts of Hsp90’s most N-terminal β-sheet, which are resolved in the crystal structure by Ali et al., 2006, pdb:2cg9.

Altogether, we dismiss ‘cross transitions’ (0 <-> 2, 1 <-> 3) based on both statistical criteria, and prior knowledge on Hsp90.

3) 4 states are proposed to fit the kinetic model best. You assume that HSP90 is adapting the same identical molecular conformations, meaning state 0, 1, 2, and 3 are the same states in presence or absence of e.g. the co-chaperone. However, the peaks in the FRET efficiency histograms, especially of the high FRET efficiency state, but also the low FRET efficiency state (in A577I) are strongly shifted in presence of all modifications, which indicate a slightly different molecular conformation. For a strict comparison of free energy landscapes as suggested in Figure 4 no common reference state is therefore available. Please discuss whether the kinetics states are necessarily different conformations to which your current signal is blind?

The reviewers are right: there can be small structural differences between states 0,1,2,3. However, such a ‘strict comparison of free energy landscapes’ as mentioned by the reviewers would require that all atom positions are known. This is not feasible experimentally, and zero uncertainty is even theoretically impossible. Therefore, state definitions are used in order to interpret experimental data. This is particularly important to gain insight into dynamic systems, such as Hsp90, where the energy landscape shows many shallow minima leading to conformational ensembles. It is a unique capability of smFRET trajectories that conformational dynamics can be resolved experimentally along a specific reaction coordinate and as a function of time, i.e. not only as fluctuations without direction. We set here the reaction coordinate along the most characteristic open-close transition of Hsp90. And as in all smFRET studies this means that we cannot observe dynamics that are orthogonal to that coordinate. But we can rule out the appearance of a new third FRET population that would represent a qualitatively new state. In fact, under all conditions, we separately and consistently find two FRET efficiency populations that split into four kinetic states (whose structural interpretation is described in the subsection “Same thermodynamics, different conformational kinetics”), which makes this global definition of four states the most reasonable one – despite slight shifts of mean positions on a locally shallow energy landscape. Please, note also that small shifts in FRET efficiency arise in dynamic systems already through changes of the kinetics on the order of the time resolution of the experiment.

In summary, we do not claim that states 0,1,2,3 are discrete nor 100% ‘identical molecular conformations’. Instead, they are conformational ensembles defined by their FRET efficiency and their kinetics, which allows us to compare their relative energy levels.

We added the following sentence to the main text:

“All states 0, 1, 2, 3 of this highly dynamic Hsp90 system represent conformational ensembles that are defined by their FRET efficiency and their kinetics (rather than discrete or ‘frozen’ conformations).”

4) As a common reference state, the transition state between the N-terminal compact and N-terminal expanded state id used (Figure 4). Is there any evidence that this is possible, why not the compact or expanded state?

The reviewers are right, only relative energies can be determined. However, if we consider e.g. the case of macro-molecular crowding, the main difference in the observed kinetics leads to a reduced off-rate out of the closed conformations towards the open conformations. In terms of energy landscapes, one can realize this change either by raising two energy levels (the transition state and the global open minimum), or by just lowering one state (the global closed minimum). In line with parsimony arguments (Okham’s razor), we deemed the latter more likely. Analogous arguments apply for the point mutation A577I. Lastly, in case of the binary effect observed with the cochaperone Aha1, we chose to keep the transition state as the reference, for consistency. Importantly, this reference choice does not affect the relative energies nor our conclusions.

We added the following sentence to the legend of Figure 4 to clarify this point:

“To illustrate the changes on the global energy landscapes, we chose the transition state as a reference for the measured relative energies.”

5) Figure 4: From your kinetics you have some idea of the rates of the transitions between the 2 open conformations and the two closed conformations which you could show in a supplementary figure? This is potentially helpful because for example Figure 4C it is really state 3 that is stabilized with respect to state 2.

The reviewers are right, we do resolve the transition rates between the 2 open conformations and the two closed conformations. They are already displayed in Figure 3 along with all other rate constants. Also the absolute values of these rate constants and corresponding confidence intervals are provided in Table 1 in Supplementary file 1.

Lastly, the steady-state changes are shown in the histograms in Figure 2 and population sizes are specified in the main text, and in Table 1 in Supplementary file 1.

The reviewers are also right that under macro-molecular crowding, the kinetic state 3 is stabilized with respect to state 2, which is shown in Figure 3. This results in a smaller open population, since state 2 is the reservoir for any opening transition. The data are fully self-consistent in this way, and no net fluxes or net particle rates can arise in the crowding experiment, which is an equilibrium experiment.

6) Each measured reference should be identical.

The three reference datasets are not expected to be identical, as explained under point 1). We are sorry for the confusion.

Yet, the population of the high FRET efficiency state varies from 14 to 29%. This is a two-fold difference, which is also what kinetically is named as a significant difference in transition rates. Can you explain this difference, explain how the state occupancies in the subsection “Mutation, cochaperone and crowding show similar thermodynamics” have been determined and what the relative error is?

Yes, we can explain this difference by the different experimental conditions explained under point 1). The state occupancies (or population sizes) were determined using double-Gaussian fits. They agree very well with the state occupancy determined by SMACKS. We performed additional bootstrapping to estimate the uncertainties of the population sizes, and report now the bootstrapped means and standard deviations both in the main text, and in the new Table 1 in Supplementary file 1. The following description of the bootstrapping procedure was added to the Materials and methods section:

“The uncertainties of steady-state populations were estimated by bootstrapping: out of each dataset, random subsets (or samples) containing two thirds of the total number of traces were drawn with replacement. […] We report the standard deviation of 1000 such populations as an uncertainty measure.”

It would further be good to also show example traces of the reference experiments with wt in the figure supplements to give the reader a chance for comparison beyond the state histogram and rate summaries.

We agree and added the example traces to the new Figure 2—figure supplement 1.

7) Please expand the following discussion: What does the current work allow the authors to conclude about the function of the ATPase activity in HSP90? Especially delineate the physico-chemical conclusions from the functional conclusion.

Our work demonstrates that conformational confinement stimulates the ATPase activity of highly dynamic Hsp90. We link this to the fact that the active site consists of several flexible residues that have to be in a specific arrangement to form the ATPase competent state (Figure 6). Concerning the ATPase activity, we conclude in turn that by directly affecting the lifetime of these contacts and the resulting structural stabilization, ATP hydrolysis itself controls also longer-range allosteric effects linked to these residues.

Moreover, we conclude that Hsp90’s ATPase activity in vivo differs from the typically reported in vitro rate constants. Also, certain mutations (e.g. mimicking PTMs) that show little to no ATPase activity in vitro, may be ‘rescued’ by the conformational confinement in the crowded cellular environment, and still be ATPase active in vivo. This can be tested by ATPase activity assays performed under macromolecular crowding.

We added the following two sentences to the Discussion section:

“In turn, by directly affecting the lifetime of active site contacts and the resulting structural stabilization, ATP hydrolysis itself controls also longer-range allosteric effects linked to these residues.”

“Certain mutations that show little or no ATPase activity in vitro, may be ‘rescued’ by the conformational confinement in the crowded cellular environment, and still be ATPase active in vivo.”

8) Figure 6 suggests combinatorial effects. By extension, this suggests that crowding and the point-mutation would have an even more enhanced effects of transition rates or ATPase activity. Can you show evidence by a combination of any two of their conditions and support the combinatorial suggestion from Figure 6?

We did not study, nor do we claim combinatorial effects as discussed in our first response above.

9) A clearer description of the data analysis needs to be placed in the Materials and methods section.

We thank the reviewers for their suggestion, and have added an additional paragraph to the Materials and methods section. (See main point 2.)

10) Statistical errors are reported for the interconversion rates between the various conformational states detected. Please also specify whether the results of each experimental condition were obtained from a single experiment (i.e. video) recorded in a single day, or that the results of each experimental condition were obtained from a set of videos recorded on different days to check for reproducibility?

We added the following sentence to the caption of Figure 2:

“The data was recorded in 5 to 13 videos per dataset on one or more days.”

Regarding the state populations an error of the extracted percentages would be helpful. This can be done e.g. by bootstrapping.

We agree with the reviewers, and report now mean populations and standard deviations obtained by bootstrapping, both in the main text, and in Table 1 in Supplementary file 1. (See also main point 6.)

https://doi.org/10.7554/eLife.57180.sa2

Article and author information

Author details

  1. Sonja Schmid

    Institute of Physical Chemistry, University of Freiburg, Freiburg, Germany
    Present address
    Department of Bionanoscience, Kavli Institute of Nanoscience Delft, Delft University of Technology, Delft, Netherlands
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    s.schmid@tudelft.nl
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3710-5602

  2. Thorsten Hugel

    1. Institute of Physical Chemistry, University of Freiburg, Freiburg, Germany
    2. Signalling research centers BIOSS and CIBSS, Albert Ludwigs University, Freiburg, Germany
    Contribution
    Conceptualization, Supervision, Funding acquisition, Validation, Visualization, Project administration, Writing - review and editing
    For correspondence
    thorsten.hugel@physchem.uni-freiburg.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3292-4569

Funding

European Commission (681891)

  • Thorsten Hugel

Deutsche Forschungsgemeinschaft (SFB 1381)

  • Thorsten Hugel

Swiss National Science Foundation (​Postdoc.Mobility fellowships)

  • Sonja Schmid

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The cochaperone Aha1 was a kind gift of Dr. Markus Jahn. We thank Dr. Markus Götz for previous contributions to data analysis, and Dr. Eli van der Sluis for helpful comments on the manuscript. This work has been funded in part by the European Research Council through ERC grant agreement no. 681891 and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project-ID 403222702 – SFB 1381. SS acknowledges the Postdoc.Mobility fellowship no. P400PB_180889 by the Swiss National Science Foundation.

Senior Editor

  1. José D Faraldo-Gómez, National Heart, Lung and Blood Institute, National Institutes of Health, United States

Reviewing Editor

  1. Maria Spies, University of Iowa, United States

Reviewers

  1. Michael Schlierf, TU Dresden, Germany
  2. Timothy Craggs, University of Sheffield, United Kingdom

Version history

  1. Received: March 24, 2020
  2. Accepted: June 28, 2020
  3. Version of Record published: July 22, 2020 (version 1)

Copyright

© 2020, Schmid and Hugel

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 3,049
    Page views
  • 383
    Downloads
  • 25
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Sonja Schmid
  2. Thorsten Hugel
(2020)
Controlling protein function by fine-tuning conformational flexibility
eLife 9:e57180.
https://doi.org/10.7554/eLife.57180

Categories and tags

Research organism

Further reading

    1. Biochemistry and Chemical Biology
    2. Epidemiology and Global Health

    Status and physiological significance of circulating adiponectin in the very old and centenarians: an observational study

    Takashi Sasaki, Yoshinori Nishimoto ... Yasumichi Arai
    Research Article

    Background: High levels of circulating adiponectin are associated with increased insulin sensitivity, low prevalence of diabetes, and low body mass index (BMI); however, high levels of circulating adiponectin are also associated with increased mortality in the 60-70 age group. In this study, we aimed to clarify factors associated with circulating high-molecular-weight (cHMW) adiponectin levels and their association with mortality in the very old (85-89 years old) and centenarians.

    Methods: The study included 812 (women: 84.4%) for centenarians and 1,498 (women: 51.7%) for the very old. The genomic DNA sequence data were obtained by whole genome sequencing or DNA microarray-imputation methods. LASSO and multivariate regression analyses were used to evaluate cHMW adiponectin characteristics and associated factors. All-cause mortality was analyzed in three quantile groups of cHMW adiponectin levels using Cox regression.

    Results: The cHMW adiponectin levels were increased significantly beyond 100 years of age, were negatively associated with diabetes prevalence, and were associated with SNVs in CDH13 (p = 2.21 × 10-22) and ADIPOQ (p = 5.72 × 10-7). Multivariate regression analysis revealed that genetic variants, BMI, and high-density lipoprotein cholesterol (HDLC) were the main factors associated with cHMW adiponectin levels in the very old, whereas the BMI showed no association in centenarians. The hazard ratios for all-cause mortality in the intermediate and high cHMW adiponectin groups in very old men were significantly higher rather than those for all-cause mortality in the low level cHMW adiponectin group, even after adjustment with BMI. In contrast, the hazard ratios for all-cause mortality were significantly higher for high cHMW adiponectin groups in very old women, but were not significant after adjustment with BMI.

    Conclusions: cHMW adiponectin levels increased with age until centenarians, and the contribution of known major factors associated with cHMW adiponectin levels, including BMI and HDLC, varies with age, suggesting that its physiological significance also varies with age in the oldest old.

    Funding: This study was supported by grants from the Ministry of Health, Welfare, and Labour for the Scientific Research Projects for Longevity; a Grant-in-Aid for Scientific Research (No 21590775, 24590898, 15KT0009, 18H03055, 20K20409, 20K07792, 23H03337) from the Japan Society for the Promotion of Science; Keio University Global Research Institute (KGRI), Kanagawa Institute of Industrial Science and Technology (KISTEC), Japan Science and Technology Agency (JST) Research Complex Program 'Tonomachi Research Complex' Wellbeing Research Campus: Creating new values through technological and social innovation (JP15667051), the Program for an Integrated Database of Clinical and Genomic Information from the Japan Agency for Medical Research and Development (No. 16kk0205009h001, 17jm0210051h0001, 19dk0207045h0001); the medical-welfare-food-agriculture collaborative consortium project from the Japan Ministry of Agriculture, Forestry, and Fisheries; and the Biobank Japan Program from the Ministry of Education, Culture, Sports, and Technology.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics

    Vibrio cholerae's ToxRS bile sensing system

    Nina Gubensäk, Theo Sagmeister ... Tea Pavkov-Keller
    Research Article

    The seventh pandemic of the diarrheal cholera disease, which began in 1960, is caused by the Gram-negative bacterium Vibrio cholerae. Its environmental persistence provoking recurring sudden outbreaks is enabled by V. cholerae's rapid adaption to changing environments involving sensory proteins like ToxR and ToxS. Located at the inner membrane, ToxR and ToxS react to environmental stimuli like bile acid, thereby inducing survival strategies e.g. bile resistance and virulence regulation. The presented crystal structure of the sensory domains of ToxR and ToxS in combination with multiple bile acid interaction studies, reveals that a bile binding pocket of ToxS is only properly folded upon binding to ToxR. Our data proposes an interdependent functionality between ToxR transcriptional activity and ToxS sensory function. These findings support the previously suggested link between ToxRS and VtrAC-like co-component systems. Besides VtrAC, ToxRS is now the only experimentally determined structure within this recently defined superfamily, further emphasizing its significance. In-depth analysis of the ToxRS complex reveals its remarkable conservation across various Vibrio species, underlining the significance of conserved residues in the ToxS barrel and the more diverse ToxR sensory domain. Unravelling the intricate mechanisms governing ToxRS's environmental sensing capabilities, provides a promising tool for disruption of this vital interaction, ultimately inhibiting Vibrio's survival and virulence. Our findings hold far-reaching implications for all Vibrio strains that rely on the ToxRS system as a shared sensory cornerstone for adapting to their surroundings.

    1. Biochemistry and Chemical Biology
    2. Microbiology and Infectious Disease

    A tRNA modification in Mycobacterium tuberculosis facilitates optimal intracellular growth

    Francesca G Tomasi, Satoshi Kimura ... Matthew K Waldor
    Research Article

    Diverse chemical modifications fine-tune the function and metabolism of tRNA. Although tRNA modification is universal in all kingdoms of life, profiles of modifications, their functions, and physiological roles have not been elucidated in most organisms including the human pathogen, Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis. To identify physiologically important modifications, we surveyed the tRNA of Mtb, using tRNA sequencing (tRNA-seq) and genome-mining. Homology searches identified 23 candidate tRNA modifying enzymes that are predicted to create 16 tRNA modifications across all tRNA species. Reverse transcription-derived error signatures in tRNA-seq predicted the sites and presence of nine modifications. Several chemical treatments prior to tRNA-seq expanded the number of predictable modifications. Deletion of Mtb genes encoding two modifying enzymes, TruB and MnmA, eliminated their respective tRNA modifications, validating the presence of modified sites in tRNA species. Furthermore, the absence of mnmA attenuated Mtb growth in macrophages, suggesting that MnmA-dependent tRNA uridine sulfation contributes to Mtb intracellular growth. Our results lay the foundation for unveiling the roles of tRNA modifications in Mtb pathogenesis and developing new therapeutics against tuberculosis.