Open Access

The architecture of functional modules in the Hsp90 co‐chaperone Sti1/Hop

Andreas B Schmid, Stephan Lagleder, Melissa Ann Gräwert, Alina Röhl, Franz Hagn, Sebastian K Wandinger, Marc B Cox, Oliver Demmer, Klaus Richter, Michael Groll, Horst Kessler, Johannes Buchner

Author Affiliations

  1. Andreas B Schmid1,,
  2. Stephan Lagleder1,2,3,,
  3. Melissa Ann Gräwert1,,
  4. Alina Röhl1,,
  5. Franz Hagn1,,
  6. Sebastian K Wandinger1,§,
  7. Marc B Cox4,
  8. Oliver Demmer1,2,3,
  9. Klaus Richter1,
  10. Michael Groll*,1,
  11. Horst Kessler*,1,2,3 and
  12. Johannes Buchner*,1
  1. 1 Department Chemie, Center for Integrated Protein Science (CIPSM), Technische Universität München, Garching, Germany
  2. 2 Department Chemie, Institute for Advanced Study (IAS), Technische Universität München, Garching, Germany
  3. 3 Chemistry Department, Faculty of Science, King Abdulaziz University, Jeddah, Saudi Arabia
  4. 4 Department of Biological Sciences, Border Biomedical Research Center, University of Texas, El Paso, TX, USA
  1. *Corresponding authors: Department Chemie, Technische Universität München, Lichtenbergstrasse 4, D‐85747 Garching, Germany. Tel.: +49 89 289 13361; Fax: +49 89 289 13363; E-mail: michael.groll{at} or Tel.: +49 89 289 13300; Fax: +49 89 289 13210; E-mail: kessler{at} or Tel.: +49 89 289 13340; Fax: +49 89 289 13345; E-mail: johannes.buchner{at}
  1. These authors contributed equally to this work

  • Present address: Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA

  • § Present address: Kinaxo Biotechnologies GmbH, Am Klopferspitz 19a, 82152 Martinried, Germany


Sti1/Hop is a modular protein required for the transfer of client proteins from the Hsp70 to the Hsp90 chaperone system in eukaryotes. It binds Hsp70 and Hsp90 simultaneously via TPR (tetratricopeptide repeat) domains. Sti1/Hop contains three TPR domains (TPR1, TPR2A and TPR2B) and two domains of unknown structure (DP1 and DP2). We show that TPR2A is the high affinity Hsp90‐binding site and TPR1 and TPR2B bind Hsp70 with moderate affinity. The DP domains exhibit highly homologous α‐helical folds as determined by NMR. These, and especially DP2, are important for client activation in vivo. The core module of Sti1 for Hsp90 inhibition is the TPR2A–TPR2B segment. In the crystal structure, the two TPR domains are connected via a rigid linker orienting their peptide‐binding sites in opposite directions and allowing the simultaneous binding of TPR2A to the Hsp90 C‐terminal domain and of TPR2B to Hsp70. Both domains also interact with the Hsp90 middle domain. The accessory TPR1–DP1 module may serve as an Hsp70–client delivery system for the TPR2A–TPR2B–DP2 segment, which is required for client activation in vivo.


Hsp90 is an abundant and essential molecular chaperone in the eukaryotic cell required for the maturation and activation of diverse client proteins (Young et al, 2004; Picard, 2006; Wong and Houry, 2006; Pearl et al, 2008). These include transcription factors such as steroid hormone receptors (SHRs) and p53 (Smith et al, 1990; Pratt and Toft, 2003; Muller et al, 2004; Walerych et al, 2004; Zhao et al, 2005; Romer et al, 2006; McClellan et al, 2007), kinases and other proteins (Geller et al, 2007; Wandinger et al, 2008; Gong et al, 2009; Taipale et al, 2010). Co‐chaperones play an important role in this process.

For SHRs, the activation cascade was shown to start with the binding of SHRs to Hsp40 and to Hsp70 (Hernandez et al, 2002a). SHRs are then transferred to Hsp90 in a process that is facilitated by the adaptor co‐chaperone Sti1/Hop (Chen and Smith, 1998; Johnson et al, 1998; Morishima et al, 2000; Song and Masison, 2005). After transfer, and concomitant with conformational changes in Hsp90 to a closed conformation, Sti1/Hop and Hsp70 dissociate while co‐chaperones like Sba1/p23 and prolyl isomerases (PPIases) such as Fkbp51/52 or Cyp40 are recruited (Pratt and Toft, 2003; Riggs et al, 2004; Li et al, 2011). Finally, Sba1/p23 and PPIases are released, resulting in the completion of SHR activation (Freeman et al, 2000; Johnson et al, 1996; Young and Hartl, 2000).

Of special importance is the co‐chaperone Sti1/Hop. It connects and regulates the Hsp90 and Hsp70 chaperone machineries (Scheufler et al, 2000; Wegele et al, 2003, 2006; Carrigan et al, 2004; Shaner et al, 2005; Flom et al, 2006, 2007), has no chaperone activity itself (Bose et al, 1996; Freeman et al, 1996), and provides a scaffold for the substrate transfer process. In this ordered sequence of events, Sti1/Hop inhibits the Hsp90 ATPase in a non‐competitive manner arresting the conformational transitions in Hsp90 (Richter et al, 2003; Hessling et al, 2009). However, it is still unclear how Sti1/Hop mediates its different functions. This is in part due to the complexity of the protein.

In yeast and mammals, Sti1/Hop is a monomeric protein (Yi et al, 2010; Li et al, 2011) composed of three tetratricopeptide repeat (TPR) domains and two domains rich in aspartate and proline (DP domains) (Odunuga et al, 2004; Kajander et al, 2009) (Figure 1A). The DP domains are located C‐terminally of the TPR1 domain (DP1) and the TPR2B domain (DP2) (Nelson et al, 2003). Their three‐dimensional structures as well as their functions are unknown.

Figure 1.

TPR2A–TPR2B is the central element for Hsp90 inhibition. (A) Domain architecture of Sti1 and Sti1 fragments. (B) Effects of Sti1 fragments on the yHsp90 ATPase. The ATPase activity of 2 μM yHsp90 was determined using an ATP‐regenerative ATPase assay at 30°C. Fragments of Sti1 were tested for inhibition at an equimolar concentration of 2 μM. Data are presented as mean values±s.e. of three independent experiments. (C) Titration of inhibitory fragments (square: Sti1‐FL; circle: TPR2A–TPR2B–DP2; triangle: TPR2A–TPR2B). Mean values of three independent experiments are shown. (D) Titration of the Ssa1 peptide PTVEEVD (circles) or the yHsp90 peptide TEMEEVD (squares) to a preformed complex of 2 μM yHsp90 and 2 μM TPR2A–TPR2B.

TPR domains are protein–protein interaction modules (Blatch and Lassle, 1999). They are formed by a tandem array of two antiparallel α‐helices (the TPR motif), which generate a right‐handed helical structure with an amphipathic channel. In Sti1/Hop, they bind to C‐terminal stretches of Hsp90 and Hsp70 (Scheufler et al, 2000). Previously, the crystal structures of TPR1 and TPR2A of human Hop were solved in the presence of peptides (Scheufler et al, 2000). The TPR2A and the TPR2B domain have been shown to be involved in binding to Hsp90 (Chen and Smith, 1998; Scheufler et al, 2000; Carrigan et al, 2004; Song and Masison, 2005; Flom et al, 2007). A role in Hsp70 binding could be shown for the TPR1 and TPR2B domain (Carrigan et al, 2004; Flom et al, 2007).

The aim of this study was to gain further insight into the structure and function of Hop/Sti1. By a combination of in vivo and in vitro experiments, we were able to define the function of basic modules of Sti1 and to determine their structures.


Dissection of Sti1 peptide binding

In order to elucidate the structural organization of yeast Sti1, we dissected the modular protein into its constituent domains (Figure 1A). A striking conundrum of Sti1/Hop is the presence of three TPR domains for two different binding partners, Hsp70 and Hsp90. To clarify the peptide‐binding specificity of the Sti1 TPR domains, peptides of different length (pentapeptide to octapeptides) corresponding to either the C‐terminal residues of Hsp70 (Ssa1) or Hsp90 (yHsp90) were tested for binding to the TPR domains by isothermal calorimetry (ITC) (Table I).

View this table:
Table 1. Affinity of Sti1 fragments for C‐terminal peptide fragments of yHsp90 and Ssa1

Sti1 TPR1 preferentially binds Hsp70 peptides and TPR2A binds Hsp90 peptides, similar to mammalian Hop (Scheufler et al, 2000). The TPR2B domain of Sti1, however, does not seem to have a preferred binding partner based on its peptide‐binding properties. It bound the Hsp70 and Hsp90 peptides with relatively low affinity (Kd≈4 μM). In contrast to TPR1 and TPR2A, maximum binding to TPR2B was already obtained with the heptapeptides, possibly explaining the low binding specificity of this domain. Overall, these peptide‐binding data imply that Sti1 has one high affinity Hsp90 peptide‐binding site (TPR2A), one preferential Hsp70 peptide‐binding site with slightly lower binding affinity (TPR1) and a third, less selective Hsp70/Hsp90 peptide‐binding site with even lower affinity compared to the other TPR domains (TPR2B).

The TPR2A–TPR2B module mediates ATPase inhibition

Binding of Sti1 to Hsp90 is accompanied by an inhibition of the Hsp90 ATPase activity (Prodromou et al, 1999; Richter et al, 2003; Li et al, 2011). To elucidate the minimum element of Sti1 responsible for this effect, we analysed the Hsp90 ATPase activity in the presence of different Sti1 constructs. It turned out that the TPR2A–TPR2B segment is sufficient to completely inhibit the Hsp90 ATPase (Figure 1B). However, the isolated TPR2A or TPR2B domains as well as both domains added together had no effect. Interestingly, full‐length Sti1, TPR2A–TPR2B–DP2 or TPR2A–TPR2B were equally effective in ATPase inhibition (Figure 1C). The inhibition mediated by TPR2A–TPR2B could be abrogated by addition of the Hsp90 peptide TEMEEVD whereas the Hsp70 peptide PTVEEVD showed no effect (Figure 1D). Taken together, these results show that TPR2A–TPR2B is the core unit of Sti1 for inhibiting Hsp90 activity and that the peptide‐binding groove of TPR2A is important for this function.

A rigid linker region defines the orientation of the TPR2A and TPR2B domains

We crystallized the TPR2A–TPR2B segment in the presence of the pentapeptide MEEVD (C‐terminal end of yHsp90) and solved the structure to a resolution of 2.6 Å (Supplementary Table SI). In agreement with the architecture of other TPR domains, both TPR2A and TPR2B consist of three TPR motifs, each comprising two helices (helices 1/2, 3/4 and 5/6), plus an additional helix (helix 7) at the C‐terminus (Figure 2A). The seven helices generate a right‐handed helical structure with an amphipathic channel. Additional electron density between the two domains revealed that the linker region is structurally well defined. The architecture of the linker orients the two TPR domains in an S‐shaped form with their binding grooves pointing in opposite directions. This orientation is achieved nearly exclusively by cation‐π packing of the Arg 425 side chain against the aromatic ring of Tyr 390 (Figure 2B). These residues are further stabilized via hydrogen bond formation between Arg 425, and Glu 421, as well as between Tyr 390 and Glu 421. These residues are strictly conserved in various Sti1/Hop orthologues. Additional experiments confirmed that this domain architecture is indeed physiologically relevant (see below).

Figure 2.

Structural characterization of the TPR2A–TPR2B module. (A) Ribbon representation of the TPR2A–TPR2B domain. The backbone of the TPR2A–TPR2B fragment is shown in grey. The bound pHsp90 peptides are shown in stick representation in green. The three amino acids (Y390, E421 and R425) involved in the stabilization of the rigid double construct are shown in stick representation in cyan. (B) Zoom of the linker domain as shown in (A). Possible hydrogen bonds are indicated as dashed lines. (C) The TPR2B domain in the presence of the C‐terminal Hsp70 heptapeptide (pHsp70). The backbone is shown as ribbon representation in grey and the bound peptide as stick and balls representation in yellow. The electron density of the FoFc map is shown in red at a contour of δ=2. The interacting residues are shown as sticks in grey. Additionally, the residues of the carboxylated clamp, which are not engaged in electrostatic interactions with the peptide, are shown in magenta. (D) Overlay of bound peptides to different TPR domains. Yellow: the pHsp70 heptapeptide bound to Sti1 TPR2B domain (C), green: the pHsp90 pentapeptide bound to the groove of the TPR2A domain in the TPR2A–TPR2B fragment (B), black: the pHsp70 heptapeptide bound to Hop TPR1 domain (pdb entry code: 1ELW) and grey: pHsp90 non‐apeptide bound to CHIP (pdb entry code: 2C2L). RMSD for Cα atoms of the last three residues is 0.33, 0.49 and 1.63 for TPR1:pHsp70, TPR2A:pHsp90 and CHIP:pHsp90, respectively. (E, F) Representation of the electrostatic potential modelled onto the accessible molecular surface as calculated and visualized with GRASP and the respective peptide. For clarity, the C‐terminal Asp residue of the peptide is referred to as Asp 0 and the preceding residues are numbered in descending order as Val (−1), Glu (−2). (E) Binding of pHsp90 to the TPR2A domain of the TPR2A–TPR2B fragment (same orientation and colour coding as in (A)). (F) Binding of pHsp70 to the TPR2B domain (same orientation and colour coding as in (C)).

Structural basis for the weaker peptide affinity of the Sti1 TPR2B domain

Investigation of the difference Fourier map in the amphipathic channels of the TPR2A–TPR2B fragment showed clear density for the Hsp90 pentapeptide in the groove of TPR2A (Figure 2A and E; Supplementary Figure S1A). In contrast, only weak additional electron density was visible in the TPR2B domain corresponding to the last three residues (EVD) of the peptide (Figure 2A; Supplementary Figure S1B). Whereas the peptide‐binding behaviour of the TPR1 and TPR2A domains have been structurally analysed before (Scheufler et al, 2000; Kajander et al, 2009), no structure of a TPR2B:peptide complex has been published so far. Thus, we crystallized TPR2B with the C‐terminal Hsp70 peptide PTVEEVD (pHsp70). In this case, the FoFc map displayed strong electron density, which accommodates the complete Hsp70 heptapeptide (Figure 2C). Consistent with the relative similar side chain interactions at the carboxylate clamp, the terminal three residues (EVD) align remarkably well with the structures of peptides bound to other TPR domains (Figure 2D). Closer analysis of the electrostatic surface representations of the binding groove of the TPR2B domain (Figure 2F) explains the weaker binding affinities of the pHsp70 as well as pHsp90 (Table I). On the one hand, the TPR2B domain lacks a hydrophobic pocket to accommodate the methionine of pHsp90 as observed for the TPR2A domain in the structure of the TPR2A–TPR2B fragment (Figure 2E). On the other hand, TPR2B contains a selective binding cavity for the threonine residue of pHsp70 (Figure 2F). However, the peptide backbone has to adopt an energetically unfavourable helical turn to be anchored at this position (Figure 2D).

The DP domains represent a novel α‐helical fold

Since no structures of the Sti1/Hop DP domains have been reported so far, we determined their folds by NMR spectroscopy. Our analysis of the domain boundaries showed that DP1 comprises residues 127–197 and that residues 198 to ∼260 of Sti1 are unstructured. This implies that the TPR1–DP1 segment is connected to the rest of Sti1 via a linker of about 60 residues. DP1 and DP2 represent exclusively helical folds with six helices in DP1 and five in DP2 located between the conserved aspartate–proline (DP) motifs of Sti1 (Figure 3A). Packing of the secondary structure elements in DP2 is rather loose with little long‐range NOE contacts between the helices (Supplementary Table SII). Important hydrophobic contacts are mediated mainly by methionine ε‐methyl groups. In DP1, an additional small helix comprising residues 133–137 stabilizes the arrangement of the secondary structure elements by providing aromatic contacts via Phe 136. Consequently, the DP1 RMSD of the structure calculation is lower than of DP2 (backbone RMSD DP1: 0.11 Å, DP2: 0.66 Å; Figure 3B; Supplementary Table SII). In DP2, the residues corresponding to the respective DP1 helix (520–526) are not structured and probably serve as a linker region connecting TPR2B and DP2. The helices of DP2 form a groove, which could serve as a binding site for ligands. In contrast, the groove in DP1 is occupied by the additional N‐terminal helix. Electrostatic potentials calculated for DP1 and DP2 show a slightly positive potential within these grooves (Figure 3C).

Figure 3.

Structures of the Sti1 DP1 and DP2 domain. (A) NMR structures of DP1 (left) and DP2 (right) representing the average structure of the 20 lowest‐energy structures obtained from simulated annealing calculations after refinement. The additional N‐terminal helix of DP1 is coloured in red. (B) Superposition of the 20 lowest‐energy structures of DP1 (left) and DP2 (right) obtained from simulated annealing calculations after refinement. (C) Calculated electrostatic potentials for DP1 (left) and DP2 (right). Red indicates negative and blue indicates positive potential.

Sti1 TPR2A–TPR2B interacts with the Hsp90 middle domain

Since the affinity of Sti1 for Hsp90 (KD of 40 nM, Richter et al, 2003) is much higher than the affinity for the Hsp90 C‐terminal peptide (KD of 300 nM), there seems to be an additional binding site for Sti1 in Hsp90. Furthermore, the ATPase inhibition by Sti1 also requires an additional interaction site. To identify this binding site, we tested individual Hsp90 domains for interaction with Sti1 by NMR spectroscopy. For the 15N‐labelled Hsp90‐N domain (residues 1–210) and full‐length Sti1 we could not detect complex formation. For the 15N‐labelled Hsp90 middle (M) domain and the TPR2A–TPR2B fragment, we observed chemical shift perturbation (CSP) and line broadening in Hsp90‐M (Figure 4A). Mainly residues around position 456 of Hsp90‐M were affected (Figure 4B and C). The isolated TPR2A and TPR2B domains also bound to Hsp90‐M but the shifts were smaller and higher concentrations were required to observe complex formation compared to the TPR2A–TPR2B fragment (data not shown). This indicates that TPR2A and TPR2B form a joint binding site for Hsp90‐M. To test for an interaction of the TPR2A–TPR2B fragment with Hsp90‐C, in addition to the binding of the C‐terminal Hsp90 peptide, we used an Hsp90‐C construct lacking the 20 C‐terminal residues (Hsp90‐CΔ20). Adding the TPR2A–TPR2B fragment to the 15N‐labelled Hsp90‐CΔ20 construct resulted in strong line broadening. Only signals corresponding to flexible residues were still present in the spectrum indicating binding (data not shown). However, an unspecific interaction cannot be ruled out completely since the deletion construct still contained a stretch of highly charged unfolded residues at the C‐terminal end that might be recognized by the TPR peptide‐binding grooves.

Figure 4.

Binding of TPR2A–TPR2B to the Hsp90‐M domain monitored by NMR. (A) (Left) Overlay of the 15N‐TROSY spectra for free TPR2A–TPR2B (black) and in complex with the Hsp90‐M domain (red). (Right) Overlay of the 15N‐TROSY spectra for free Hsp90‐M (black) and in complex with TPR2A–TPR2B (red). (B) CSP plots for binding of Hsp90‐M to 15N TPR2A–TPR2B (left) and binding of TPR2A–TPR2B to 15N Hsp90‐M (right). Significance levels (average CSP+2x standard deviation) are indicated by a dashed line. (C) Mapping of CSP data from (B) onto the crystal structures of TPR2A–TPR2B (left, pdb: 3uq3) and Hsp90‐M (right, pdb: 1hk7). (D) (Left) PRE data for the complex between 15N‐labelled TPR2A–TPR2B and different Proxyl‐modified Hsp90‐M variants. Signal intensity ratios for the spectra before and after addition of ascorbic acid are plotted against the residue number for each mutant. The domain boundary within the TPR2A–TPR2B fragment is indicated. (Right) Model of the complex between TPR2A–TPR2B and Hsp90‐MC obtained by HADDOCK. The unstructured C‐terminal ends of Hsp90 not resolved in the crystal structure are drawn in red and the MEEVD motif is represented by red dots.

The NMR spectra of the isolated 15N‐labelled TPR2A and TPR2B domains overlaid well with the spectrum of the two‐domain construct (Supplementary Figure S2A). Therefore, the assignments of the isolated TPR domains could be transferred. Only signals for residues located at the domain boundaries were shifted in the two‐domain construct and signals around R425, which is in good agreement with the crystal structure of TPR2A–TPR2B (Figure 2B). Adding Hsp90‐M to 15N‐labelled TPR2A–TPR2B resulted in line broadening and specific shifts (Figure 4A). Residues affected by the binding were located in both TPR2A and TPR2B (Figure 4B and C). Significant shifts within TPR2A occurred around residue 368 and within TPR2B around residue 446. This supports the assumption that TPR2A and TPR2B form a joint binding site for Hsp90‐M. To gain further information on the binding sites on Hsp90‐M and TPR2A–TPR2B, we performed paramagnetic relaxation enhancement (PRE) experiments. Residues 411, 422 and 456 in Hsp90‐M, which are in proximity to the supposed main binding site of TPR2A–TPR2B were mutated to Cys and modified with a Proxyl‐label. After complex formation with the respective Hsp90‐M mutants, reduction of signal intensities was measured on the 15N‐labelled TPR2A–TPR2B fragment (Figure 4D). Especially for the S411C and S456C variants, which are located on the outside of the Hsp90 dimer, strong effects could be observed, mainly for TPR2A residues around position 370. For S422C, which is located more at the inner side of the Hsp90 dimer, PRE effects were only visible on TPR2B residues. At K484C, which is further away from the putative binding site, no signal reduction could be observed. We then performed HADDOCK docking runs (de Vries et al, 2010) to obtain a model for the complex between Hsp90 and the TPR2A–TPR2B fragment. The docking was first conducted with CSP data only. The PRE data were used afterwards to select the docking solution that fits the experimental data best (Supplementary Figure S2B). In the complex, TPR2A is oriented towards the C‐terminal part of Hsp90‐M directly contacting the outside of Hsp90‐M only with residues of helix 7 (368–374). The peptide‐binding groove of TPR2A is pointing to Hsp90‐C. TPR2B is oriented to the N‐terminal part of Hsp90‐M, slightly wrapped around Hsp90‐M with the backside of its peptide‐binding groove pointing to Hsp90. Thus, in agreement with the experimental data, the TPR2B peptide‐binding site is still accessible for Hsp70 binding.

Contribution of Sti1 domains to the maturation of v‐src and glucocorticoid receptor in yeast

The activity of the Hsp90–client protein v‐src was identified to be dependent on Sti1 in yeast (Chang et al, 1997). In wild‐type yeast cells, expression of v‐src leads to cell death due to aberrant phosphorylation of the proteome, whereas Δsti1 yeast are viable, allowing us to test which of the Sti1 domains are necessary for v‐src folding (Supplementary Figure S3). In fact, it turned out that most fragments did not support v‐src folding. Yet, expressing ΔTPR1 restored lethality like full‐length Sti1, indicating that TPR1 does not display a significant role for v‐src folding. Importantly, deletion of DP2 was sufficient to completely abolish v‐src activity.

It was previously shown that yeast cells lacking Sti1 have considerably lower glucocorticoid receptor (GR) activity compared to wild‐type yeast cells (Chang et al, 1997; Carrigan et al, 2005; Flom et al, 2006). To determine whether domains of Sti1 can substitute for the full‐length protein, we expressed Sti1 fragments in Δsti1 cells together with human GR and a β‐galactosidase reporter plasmid, which allowed for the quantitation of hormone‐induced GR activity (Figure 5A–C).

Figure 5.

Influence of Sti1 variants on client activity and complex formation with Hsp90. (AC) Δsti1 yeast cells expressing Sti1 fragments containing a human GR expression plasmid and a β‐galactosidase reporter vector were induced with DOX during exponential growth. Cells were lysed 8–10 h after induction and assayed for β‐galactosidase activity and for lysate concentration. Experiments were performed in duplicate with three values each. Error bars indicate standard error. (D) Formation of complexes between Ssa1 and Sti1. Complex formation between Ssa1 and Sti1 variants was analysed by analytical ultracentrifugation. Sedimentation profiles were converted into dc/dt plots according to standard procedures. In all, 0.5 μM of fluorescein‐labelled Ssa1 was incubated in the absence of Sti1 fragments (black) or complexed with either 3 μM full‐length Sti1 (blue) or TPR2A–TPR2B (green) in 10 mM potassium phosphate, 1 mM TCEP, pH 7.5. Centrifugation was performed at 20°C and 42 000 r.p.m. (E) Formation of ternary complexes between Ssa1 and Sti1–yHsp90. Complex formation between Ssa1 and Sti1 variants was analysed by analytical ultracentrifugation in the presence of access yHsp90. Sedimentation profiles were converted into dc/dt plots according to the standard procedures. In all, 0.5 μM of fluorescein‐labelled Ssa1 was incubated in the absence of Sti1 fragments (not shown) or complexed with either 3 μM full‐length Sti1 (black), Sti1–N435A (red), TPR2A–TPR2B (green) or TPR2A–TPR2B–N435A (blue) in 10 mM potassium phosphate, 1 mM TCEP, pH 7.5. Centrifugation was performed at 20°C and 42 000 r.p.m.

In agreement with the literature, Δsti1 cells showed 20% of the GR activity observed in wild‐type cells (Figure 5A). Expression of full‐length Sti1 in the deletion strain restored GR activity. Interestingly, the same dependence of Sti1 fragments on GR activation could be observed as in the case of v‐src folding: Most fragments did not support GR activation. Only the fragment ΔTPR1 activated GR like full‐length Sti1 (Figure 5A), implying that the TPR1 domain of Sti1 is dispensable for GR activation. The further deletion of DP1 in the fragment TPR2A–TPR2B–DP2 led to a GR activation of 77%. Interestingly, again, the deletion of the DP2 domain completely abrogated Sti1 function in GR activation in vivo (Figure 5A).

Both in vivo assays suggest a special function of DP2 in client activation. Since DP1 and DP2 are structurally very similar, as revealed in this study, we were interested whether DP1 was able to replace DP2 functionally. However, the construct in which DP2 was replaced by DP1 (‘DP swap’) did not support GR activation above background levels (Figure 5B). This shows that, despite their structural homology, the DP domains have different effects on GR activity. To obtain information on which parts of DP2 are involved in GR activation, we tested DP2 point mutants in full‐length Sti1 (V540A, L553A, Q557A, Q564A, T578A and I584A) in vivo (Supplementary Figure S4B). While the mutant L553A displayed only background levels of GR activity, the activity of I584A and V540A was only moderately reduced (Figure 5B). Surprisingly, the mutant Q564A showed enhanced activity in GR activation (around 130% of wild type). Taken together, this result indicates a special role of helices 3 and 4 of DP2 for GR activation.

To elucidate the functional importance of the unexpected S‐shaped orientation of the TPR2A and TPR2B domains, we mutated R425 that positions the rigid linker (see Figure 2B) and assayed GR activation in yeast (Figure 5B). Strikingly, GR activity was reduced to about 55%, demonstrating the importance of the orientation of TPR2A and TPR2B for Sti1 function.

To investigate the contribution of peptide binding to TPR2B for GR activity, we mutated residues in the peptide‐binding site (N435A, R465A, R469A, N435A/R465A and N435A/R469A; Supplementary Figure S4A). Interestingly, all mutants tested did not support GR activation, strongly suggesting that TPR1 cannot compensate for the loss of peptide binding in TPR2B (Figure 5C).

Formation of ternary Hsp70–Sti1–Hsp90 complexes

We asked if the inability to support GR activity observed for TPR2B variants in vivo was due to the loss of Hsp70 binding to this domain. Therefore, we labelled Hsp70 (Ssa1) with Fluorescein (Hsp70*) and subjected it to analytical ultracentrifugation in either the absence or presence of Sti1 variants.

Hsp70* alone sedimented with 4 S. After addition of Sti1 or TPR2A–TPR2B, the sedimentation coefficient of Hsp70* increased to 5.5 S or 5 S, respectively, indicating complex formation (Figure 5D). Addition of Hsp90 (yHsp90) to the preformed Sti1–Hsp70* or TPR2A–TPR2B–Hsp70* complexes resulted in s‐values of 9 S or 8.5 S (Figure 5E), which represent the ternary complexes Hsp70*–Sti1–Hsp90 and Hsp70*–TPR2A–TPR2B–Hsp90. This result confirms that TPR2A–TPR2B binds Hsp70 and Hsp90 simultaneously. However, compared to full‐length Sti1, the amount of Hsp70* in the ternary complexes was reduced (Figure 5E). Interestingly, Sti1 with a mutation in the TPR2B peptide‐binding groove (N435A) still supported ternary complex formation, which indicates that the two chaperones can also be bound via TPR1 and TPR2A. The fragment TRP2A–TPR2B comprising this mutation, however, did not support complex formation with Hsp90 and Hsp70 (Figure 5E).

Remarkably, while simultaneous binding of both Hsp90 and Hsp70 can be achieved via TPR2A and TPR2B or via TPR2A and TPR1, GR activation in vivo requires an interaction where both chaperones are bound via the TPR2A–TPR2B module.


Sti1/Hop is a modular protein composed of three TPR and two DP domains. This protein fulfils several functions in the context of the Hsp70 and Hsp90 machinery. It brings both chaperones in physical contact in a ternary complex and inhibits the ATPase activity of Hsp90 (Prodromou et al, 1999) as a non‐competitive inhibitor (Richter et al, 2003) by restricting its conformational flexibility (Hessling et al, 2009). Together, this results in an efficient transfer of clients to Hsp90. Although aspects of this reaction had been addressed before, the underlying mechanism such as the structural basis for the inhibition of the Hsp90 ATPase and the structure of the DP domains remained elusive. By combining X‐ray crystallography and NMR spectroscopy, we were able to identify a 60‐amino‐acid long unstructured segment connecting DP1 and TPR2A and to solve the structures of four Sti1 domains and that of the central building block of Sti1, the TPR2A–TPR2B module. We reveal that this module is required and sufficient to completely block the Hsp90 ATPase. Interestingly, the TPR2A and TPR2B domains alone or in combination had no comparable effect on the Hsp90 ATPase, implying that both domains do not work independently but have to be located on one polypeptide chain. When the rigid linker connecting TPR2A and TPR2B was made flexible, Sti1 function was compromised in vivo. This demonstrates the importance of the fixed domain orientation in the TPR2A–TPR2B module. Thus, for these consecutive domains, the ‘pearls on a string’ view of independently functioning segments does not apply.

For mammalian Hop, it had been proposed earlier that there are interactions with Hsp90 beyond the C‐terminal peptide (Onuoha et al, 2008). In this study, we could define by NMR spectroscopy that within the TPR2A–TPR2B module TPR2B binds to Hsp90‐M in addition to the binding of TPR2A to Hsp90‐C. The additional interactions of Hsp90 with TPR2B are independent of the canonical TPR peptide‐binding groove. Molecular docking simulations of this complex using NMR data suggest that TPR2A–TPR2B is bound to Hsp90 in an orientation leaving the TPR2B peptide‐binding site accessible for Hsp70. Furthermore, TPR2B occupies a position between the two Hsp90 subunits, thereby sterically preventing the conformational rearrangement of the Hsp90‐M domains during the ATPase cycle (Figure 6). In this context, it is noteworthy that inhibition of one of the two Hsp90 subunits is sufficient for completely blocking the Hsp90 ATPase (Li et al, 2011). In accordance with this, a recent investigation of the organization of the Hop–Hsp90 complex by cryo‐EM suggests that Hop is placed between the two subunits of an Hsp90 dimer (Southworth and Agard, 2011). Such a model is consistent with the essential requirement of the TPR2B domain for the inhibition of Hsp90 while allowing the simultaneous binding of Hsp90 and Hsp70 to TPR2A–TPR2B.

The peptide specificity of TPR2B is determined by two factors. On the one hand, the affinity of the Hsp70/Hsp90‐C‐terminal peptides to TPR2B is reduced due to the difference in the binding pockets. Whereas TPR2B lacks a defined specificity pocket for binding of Hsp90, the selectivity pocket for the threonine residue of pHsp70 requires the peptide to adopt a helical turn to be anchored at this site. On the other hand, the electrostatic interactions involving the conserved C‐terminal aspartate residue are altered in the TPR2B domain. Structures previously solved for the Hop TPR1 and TPR2A domains in complex with the C‐terminal peptide of Hsp70 and Hsp90, respectively, revealed that a conserved set of five amino acids in the central grooves of the TPR domains forms a carboxylate clamp for ligand binding (Scheufler et al, 2000; Kajander et al, 2009). This motif serves as a socket for the anchoring of the peptide and interacts tightly with both carboxylate functions of the conserved C‐terminal aspartate residue. The Sti1 TPR2B sequence, however, represents one of the rare exceptions where these five amino acids are not conserved. In this case, two lysines are replaced by arginines (R400 and R465, respectively) and an asparagine is exchanged by a lysine (L404). Nevertheless, strong electrostatic interactions involving the residues R400, N435 and R465 are observed in the TPR2B:pHsp70 complex (Figure 2C and F). Yet, the remaining two residues of the carboxylate clamp, L404 and R431, have no influence on the binding of the peptide, which explains the observed weaker affinity. Nevertheless, the importance of this binding site was demonstrated by mutations of residues in the peptide‐binding site, which did not support GR activation in vivo (Figure 5C).

Surprisingly, although the TPR2A–TPR2B module retains the full inhibitory capacity for the Hsp90 ATPase activity and can bind both Hsp70 and Hsp90, it is unable to support the activation of clients in vivo. For this reaction, the DP2 domain is essential, as the fragment TPR2A–TPR2B–DP2 exhibited most of the activity of wild‐type Sti1. The DP domains exhibit homologous helical folds. Helices 3 and 4 in DP2 (four and five in DP1) form a wedge‐shaped surface making little contacts to neighbouring secondary structure elements. This leaves a groove that might serve as binding site for other proteins. In agreement with this notion, our mutational analysis suggests that helices 3 and 4 of the DP2 domain have a special function in GR activation. In contrast, the groove in DP1 is occupied by an additional small helix, which might hamper access to a potential binding site. Accordingly, DP1 cannot replace DP2 in GR activation. The important role of DP2 for client activation in vivo may be explained by a direct interaction with the substrate protein. Binding of DP2 to Hsp70 or Hsp90 could not be detected. However, at least for Hop, no chaperone activity has been detected in standard in vitro assays (Bose et al, 1996; Freeman et al, 1996). An alternative explanation is that DP2 is required to promote conformational rearrangements associated with client transfer to Hsp90 or release of Sti1 from Hsp90 and Hsp70. DP1 also seems to play a role in this context. Of note, it has a positive effect on GR activity only when present together with DP2. In this context, it is interesting that in some organisms, Sti1/Hop lacks the DP1 domain (Drosophila melanogaster) or the TPR1–DP1 segment (Caenorhabditis elegans) (Flom et al, 2007; Gaiser et al, 2009). Sti1 devoid of TPR1 supported GR and v‐src activation comparably to full‐length Sti1. On the one hand, this finding demonstrates that the TPR1 domain of Sti1 is dispensable for the function of Sti1 in the context of client activation. On the other hand, it implies that Hsp70 binds to TPR2B. Our data show that Hsp90 and Hsp70 can interact simultaneously with TPR2A–TPR2B or TPR2A–TPR1 to form a ternary Hsp90–Sti1–Hsp70 complex. Thus, Hsp70 can bind to either TPR1 or TPR2B but binding of Hsp70 to TPR1 does not seem to be productive in the activation process. For activation, Hsp70 has to bind to the TPR2B domain. This proposed change of binding sites for Hsp70 from TPR1 to TRP2B may occur coupled to the binding of Hsp90, which seems reasonable since the affinity of Hop for Hsp70 was shown to increase in the presence of Hsp90 (Hernandez et al, 2002b). When Sti1 is bound to Hsp90, TPR2A is occupied by the C‐terminal end of Hsp90 and TPR2B is accessible for Hsp70. The importance of DP2 for client activation further supports the notion that the adjacent TPR2B domain is the Hsp70‐binding site in the context of Hsp90.

Taken together, our data suggest a scheme of events (Figure 6), in which Hsp70 first binds to an Hsp90–client protein and C‐terminally interacts with the TPR1 site of Sti1. This complex associates with Hsp90 via the Sti1 TPR2A site and inhibits the conformational changes in Hsp90 via interaction of TPR2A with Hsp90‐M. This triggers the transfer of Hsp70 and the client to the TPR2B–DP2 module. Only this seems to allow the loading of a client to Hsp90, which is required for their activation in vivo. Thus, the molecular architecture of Sti1 allows integrating several functions required to coordinate two chaperone machineries.

Figure 6.

Model of Sti1 interaction with Hsp90 and Hsp70. Sti1 consists of TPR1 (1–126), DP1 (127–197), TPR2A (257–390), TPR2B (393–520) and DP2 (528–589). The TPR1–DP1 fragment is connected flexibly to the rigid TPR2A–TPR2B block via an unstructured linker region of about 60 residues. The DP2 domain is linked to TPR2B via six unstructured residues. The Hsp90 C‐domain binds TPR2A and the Hsp90‐M domain TPR2B. This leads to the inhibition of the Hsp90 ATPase. Due to higher affinity, Hsp70 is initially bound to TPR1. DP1 may stabilize the bound client. Subsequently, Hsp70 and client are transferred to TPR2B–DP2. From this position, the client is then transferred to Hsp90.

Materials and methods

Protein purification

pET28 vectors carrying the genes for the various Sti1 and Ssa1 fragments plus an N‐terminal 6xHis‐SUMO‐tag or a Thrombin cleavable 6xHis‐tag were transformed into the Escherichia coli strain BL21 (DE3) Codon Plus. Protein expression was induced by addition of 1 mM isopropyl‐β‐d‐thiogalactopyranosid (IPTG) at OD600 of 0.5 for 4 h at 30°C. In the case of full‐length Ssa1, Pichia pastoris strain KM71H‐Ssa1 (aox1∷ARG4; arg4; 6xHis‐SSA1 gene genomically inserted at AOX1 locus) was used for expression, which was induced with 0.5% (v/v) methanol.

Proteins were first purified by a 5‐ml Hi‐Trap column (GE Healthcare, München, Germany). After cleavage of the His6‐SUMO tag or His‐tag, respectively, gel filtration chromatography was performed with a Superdex 200 PrepGrade column (GE Healthcare) preequilibrated in 40 mM Hepes (pH 7.5), 150 mM KCl, 5 mM MgCl2. For isotope labelling of proteins for NMR spectroscopy, standard M9 medium supplemented with 1 g/l 15N ammonium chloride and 2 g/l 13C d‐glucose (Eurisotope, Saarbrücken, Germany) was used. In the case of deuterated proteins 2H, 13C d‐glucose and D2O (60–80%) (Euriso‐Top GmbH, Saarbrücken, Germany) were used.

Single point mutations of Sti1 or yHsp90 domains were generated using the QuikChange II Site‐Directed Mutagenesis Kit (Agilent, La Jolla, USA).

Yeast strains and plasmids

The Δsti1 yeast strain YOR027w (BY4741; Mat a; his3Δ1; leu2Δ0; met15Δ0; ura3Δ0; YOR027w∷kanMX4, from Euroscarf) was used to evaluate the activity of GR and v‐src in yeast in dependence of various fragments of Sti1. The human GR (hGR) expression vector (p413GPD‐hGR) was created by inserting the hGR gene in the vector pSPUTK‐hGR (kind gift of DF Smith) including the yeast Kozak consensus sequence AACAAAATG. The reporter plasmid with GR response elements used in the experiments was pUCΔSS‐26X (Louvion et al, 1996). Expression plasmids for Sti1 fragments were constructed by cloning each fragment in a p425GPD expression vector.

Peptide synthesis

Peptides for ITC analysis were synthesized using Fmoc chemistry on a TCP resin. Synthesis was analysed by ESI mass spectrometry and HPLC.

Isothermal titration calorimetry

Affinities of TPR domains for Ssa1 and yHsp90 peptides were measured by ITC using a MicroCal VP‐ITC titration calorimeter (MicroCal Inc., Northhampton, USA). In all, 40–50 aliquots of a 1.5‐mM solution of the respective peptide were titrated into a 150‐μM solution of the TPR domain in 10 mM potassium phosphate, 1 mM TCEP pH 7.5 at a temperature of 20°C. Peptide was injected in 5 μl aliquots to the protein until saturation was reached. Binding curves were corrected for dilution heats and fitted to a one‐site binding model using the software provided by the manufacturer (Origin software package, GE Healthcare).

ATPase assay

ATPase activities were measured using a regenerating ATPase assay as described before (Richter et al, 2003). Assays were performed in 50 mM Hepes (pH 7.5), 50 mM KCl, 5 mM MgCl2 and 2 mM ATP at 30°C. Protein concentrations were 2.0 μM for yHsp90 and up to 4.0 μM for the Sti1 fragments.

Analytical ultracentrifugation

Yeast Hsp70 (Ssa1) was covalently coupled to the amine‐reactive dye 5‐(and ‐6)‐carboxyfluorescein (Invitrogen, La Jolla, USA) as recommended by the manufacturer. Analytical ultracentrifugation was carried out in a Beckman ProteomeLab XL‐A (Beckman, Fullerton, USA) equipped with a fluorescence detection system (Aviv Biomedical, Lakewood, USA) using labelled Ssa1 at concentrations of 500 nM and unlabelled proteins at concentrations of 3 μM if not indicated differently. Sedimentation analysis was carried out at 42 000 r.p.m. in a TI‐50 Beckman rotor (Beckman, Fullerton) at 20°C in a 10‐mM potassium phosphate, pH 7.5, 1 mM TCEP buffer. To determine the size of the complexes, the raw data were converted to dc/dt profiles as described before (Gaiser et al, 2010) and then analysed by bi‐Gaussian functions.

GR activity assay in yeast cells

The Δsti1 yeast strain was transformed with the GR expression vector, the reporter vector, an empty p425GPD vector or a p425GPD vector containing one of the designed Sti1 fragments, respectively. Single clones were grown at 30°C in minimal medium to stationary phase. Then, cells were diluted to an OD600 of 0.5, induced with 10 μM deoxycorticosterone (DOX; Sigma Aldrich, St Louis, USA) and grown for 8 h at 30°C. β‐Galactosidase assays were performed as described elsewhere (Flom et al, 2007) using 2‐nitrophenyl β‐d‐galactopyranosid (ONPG; Sigma Aldrich, St Louis, USA) in Z‐buffer as substrate. The absolute activity of β‐galactosidase can be calculated with the following formula: (4000 × OD420)/(incubation time × OD595) nmol/min/mg. The relative activity values were obtained by setting the absolute value of the full‐length signal to 100%. Results are the mean of three independent experiments. Error bars indicate standard error.

V‐src assay in yeast cells

The Δsti1 yeast strain was transformed with a galactose‐inducible v‐src expression plasmid pRS316‐v‐src and an empty p425GPD vector or a p425GPD vector containing one of the designed Sti1 fragments, respectively. Single clones were grown at 30°C in minimal medium with 2% (w/v) raffinose to stationary phase. Dilution series of these cultures were spotted onto solid minimal growth medium containing either 2% (w/v) galactose or 2% (w/v) glucose and growth was assayed.

Crystallization, structure determination and refinement

Crystallization conditions for the TPR2B domain in complex with the pHsp70 (Ssa1) peptide as well as the fragment TPR2A–TPR2B in presence of the pHsp90 (yHsp90) peptide (Biomatik, Wilmington, USA) were identified utilizing sitting drop vapour diffusion with different crystallization suites (Qiagen, Hilden, Germany) at 20°C. Using a Phoenix crystallization robot (Art Robbins Instruments, Sunnyvale, USA), drops containing 100 nl protein solution and 100 nl crystallization solution were equilibrated against reservoir containing 45 μl crystallization solution. Generally, the crystals grew within 48 h and were soaked for ∼1–2 min in appropriate cryoprotectant before being amorphously super‐cooled in a stream of liquid nitrogen gas at 100 K.

Crystals containing Sti1 TPR2B (70 mg/ml) with the bound heptapeptide PTVEEVD were obtained with 0.2 M lithium chloride and 2.2 M (NH4)2SO4 and crystals of the two‐domain fragment, TPR2A–TPR2B (120 mg/ml) preincubated with the pentapeptide MEEVD grew in buffer containing 0.2 M TMAO, 0.1 M Tris pH 8.8, and 20% PEG 2000 MME.

Datasets were collected either using synchrotron radiation at the X06SA‐beamline, SLS (Villigen, Switzerland) or on a Bruker Microstar/X8 Proteum with our in‐house Cu rotating anode (Bruker AXS GmbH; Supplementary Table SI). Synchrotron datasets were processed using the program package XDS (Kabsch, 1993) and in‐house datasets with the Proteum software suite from Bruker (Proteum 2 Software Suite; Bruker AXS, Inc.). For structure determination of the various Sti1 domains, molecular replacement was performed in Phaser within the CCP4i GUI (Collaborative Computational Project, 1994) by using the coordinates of the TPR2A domain of human Hop (pdb entry code: 1ELR). The models were completed either using the interactive three‐dimensional graphic program MAIN or COOT. Stereochemically restrained refinement of the models was carried out in REFMAC5 using maximum‐likelihood targets. Water molecules were located with ARP/wARP solvent and verified manually. Molecular and electron density illustrations were prepared in PyMOL (Delano Scientific LLC). For references see Supplementary Table SI.

NMR spectroscopy

Backbone assignments for the single domains of TPR2A, TPR2B, DP1 and DP2 were obtained by recording a set of standard triple resonance experiments including HNCO, HNCA, HNCACB, HN(CO)CA, HN(CO)CACB and HN(CA)CO on a Bruker DMX600 or DMX750 spectrometer (Bruker Biospin, Rheinstetten, Germany) at a temperature of 293 K in 50 mM potassium phosphate, 50 mM KCL, 1 mM TCEP, pH 7.5. Sample concentrations were around 1 mM in each case. For DP1 and DP2, U13C,15N‐labelled samples were used whereas for TPR2A and TPR2B U13C,15N‐labelled samples with additional partial deuteration (∼60%) were used. Sequence‐specific resonance assignments were made using the semi‐automated program PASTA (Leutner et al, 1998) and in‐house written software. To confirm the assignments obtained, additional 3D 15N‐edited NOESY spectra were recorded (Jahnke et al, 1995). Processing of spectra was done with Topspin 1.3 (Bruker Biospin) and for analysis the program SPARKY (TD Goddard and DG Kneller, SPARKY 3, University of California, San Francisco) was used.

Aliphatic side chain assignments for DP1 and DP2 domains were made by using HNHB, CC(CO)NH, CCH‐TOCSY, HCCH‐TOCSY and HCH‐COSY experiments. Stereospecific assignment of prochiral H methylene and valine methyl groups and the resulting rotamer assignment were made using 3JNHβ couplings observed in the HNHB experiment and from NOESY patterns. For structure determination, a set of the following NOESY experiments was recorded for DP1 and DP2 on an Avance900 spectrometer (Bruker Biospin): 15N‐HSQC‐NOESY, NNH‐NOESY and 13C‐HSQC‐NOESY, heteronuclear edited 3D‐CCH NOESY and 3D‐CNH NOESY (Diercks et al, 1999). Cross‐peaks from the NOESY spectra were converted into distance restraints by dividing them into four classes: Strong, medium, weak and very weak corresponding to restraints on upper distance of 2.7, 3.2, 4.0 and 5.0 Å. Lower distance restraints were also applied for very weak and absent sequential HN–HN cross‐peaks using a minimum distance of 3.2 Å and for medium intensity or weaker sequential and intraresidual HN–Hα cross‐peaks using a minimum distance of 2.7 Å. Additionally, pseudo atom corrections (using r−6 averaging) were used on upper distances for methyl groups (+0.8 Å for one methyl and +1.5 Å for two methyls) and non‐stereospecifically assigned methylene groups (+0.7 Å). Dihedral angle restraints could be obtained for backbone φ and ψ angles based on Cα, Cβ, C′ and Hα chemical shifts measured using the program TALOS. Restraints were applied for predictions of the program consistent with the input chemical shifts, 3JHNHα coupling constants measured from an HNHA experiment and the observed NOE pattern within sequential residues. Predictions were used for the structure calculation with the tolerance calculated by the program ±5°. Directly measured 3JHNHα couplings were also applied as restraints on the backbone φ angles. Structure calculation was performed with XPLOR‐NIH (Schwieters et al, 2003) using standard protocols. For refinement of the structures, backbone NH residual dipolar couplings (RDCs) were included in the calculation. For the measurement of HN RDCs, the TROSY and semi‐TROSY components of the HN correlation signal were compared in the isotropic and anisotropic case. Alignment of the DP domains was achieved by using Pf1‐Phages in a concentration of 8 mg/ml (Hyglos GmbH, Regensburg, Germany).

Resonance assignments for the DP domains have been deposited in the BMRB ( with the accession codes 18090 (DP1) and 18091 (DP2). The corresponding coordinates have been deposited in the PDB ( with the accession codes 2llv (DP1) and 2llw (DP2).

NMR measurements of the interaction of Sti1 domains with yHsp90 domains were carried out in 10 mM potassium phosphate, 1 mM TCEP, pH 7.5. About 200 μM of the isotope labelled part of the complex was used with the unlabelled part added to a 1:1 or 2:1 ratio.

Spin labelling was done with single cysteine variants of the yHsp90‐M domain. Mutagenesis was done by standard quickchange reactions (QuikChange II Site‐Directed Mutagenesis Kit; Agilent, La Jolla, USA). Modification of the cysteine residue was achieved by adding a two‐fold excess of (iodoacetamido)‐proxyl (PROXYL; Sigma Aldrich, Taufkirchen, Germany). Free label was removed by passing the protein solution twice over a desalting column (GE Healthcare). For reduction of the spin‐label, a 5–10‐fold molar excess of ascorbic acid was added to the sample and incubated for 2 h at room temperature.

Molecular docking

Structure files for the docking were obtained from the protein data bank (Hsp90 MC: 2cge). A proxyl‐modified template of the Hsp90 MC fragment was first built with Xplor‐NIH (Schwieters et al, 2003), where NOE restraints have been back calculated from the crystal structure. Docking was done using the HADDOCK webserver (de Vries et al, 2010). Chemical perturbation data for 15N‐labelled TPR2A–TPR2B and Hsp90‐M were used as input by defining active and passive residues. In all, 1000 rigid body docking runs, 200 structure calculation runs with torsion angle dynamics and 200 refinements in explicit solvent were carried out. The resulting structures were clustered according to intermolecular energy terms and RMSD from the lowest‐energy structure. The best‐scored structures after docking were validated by back calculation of PRE effects and comparison with the CSP data.

Accession codes

The coordinates of the structures have been deposited in the PDB ( with the accession codes 3upv (TPR2B:pHsp70), 3uq3 (TPR2A–TPR2B:pHsp90), 2llv (DP1) and 2llw (DP2).

Supplementary data

Supplementary data are available at The EMBO Journal Online (

Conflict of Interest

The authors declare that they have no conflict of interest.

Supplementary Information

Supplementary Data [emboj2011472-sup-0001.pdf]


We thank the staff of the Swiss Light Source (Villigen, Switzerland) for their assistance in X‐ray data collection. We thank Jing Li for the help with the artwork. The v‐src expression plasmid pRS316‐v‐src was a kind gift of S Lindquist. ABS and FH were supported by the CompInt program of the Elitenetzwerk Bayern. This work was supported by grants from the DFG (SFB 594) and the Fonds der Chemischen Industrie to JB, HK and MG. MAG was supported by the Peter und Traudl Engelhorn Stiftung. AR gratefully acknowledges the support of the TUM Graduate School's Faculty Graduate Center of Chemistry at the Technische Universität München. MBC was in part supported by the Grant Number 5G12RR008124 (to the Border Biomedical Research Center (BBRC)/University of Texas at El Paso) from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH).

Author contributions: MAG and MG performed the crystallization and X‐ray studies. SL and FH conducted the NMR experiments. ABS, SL, AR and SKW carried out the biochemical experiments. ABS, AR and MBC performed the in vivo studies. KR collected the analytical ultracentrifugation data. OD synthesized the peptides. JB, HK and MG designed the experiments. ABS, SL, MAG, AR, HK, MG and JB wrote the manuscript.