Article

Olivier Genest

0022-2836/Published by E

Hsp70 and Hsp90 of E. coli Directly Interactfor Collaboration in Protein Remodeling

†, Joel R. Hoskins †, Andre

a N. Kravats †,Shannon M. Doyle and Sue Wickner

Laboratory of Molecular Biology, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA

Correspondence to Shannon M. Doyle and Sue Wickner: 37 Convent Drive, Room 5144, National Institutes of Health,Bethesda, MD 20892, USA. doyles@mail.nih.gov; wickners@mail.nih.govhttp://dx.doi.org/10.1016/j.jmb.2015.10.010Edited by J. Buchner

Abstract

Hsp90 is a highly conserved molecular chaperone that remodels hundreds of client proteins, many involved inthe progression of cancer and other diseases. It functions with the Hsp70 chaperone and numerouscochaperones. The bacterial Hsp90 functions with an Hsp70 chaperone, DnaK, but is independent of Hsp90cochaperones. We explored the collaboration between Escherichia coli Hsp90 and DnaK and found that thetwo chaperones form a complex that is stabilized by client protein binding. A J-domain protein, CbpA,facilitates assembly of the Hsp90Ec–DnaK–client complex. We identified E. coli Hsp90 mutants defective inDnaK interaction in vivo and show that the purified mutant proteins are defective in physical and functionalinteraction with DnaK. Understanding how Hsp90 and Hsp70 collaborate in protein remodeling will provide thegroundwork for the development of new therapeutic strategies targeting multiple chaperones andcochaperones.

Published by Elsevier Ltd.

Introduction

Proteins belonging to the Hsp90 family are presentin nearly all organisms and comprise a highlyconserved class of ATP-dependent molecular chap-erones [1–4]. In eukaryotes, Hsp90 is essential forcell viability. It is required for remodeling andactivation of hundreds of client proteins involved inmany crucial cell processes, such as cell signalingand response to stress. Protein remodeling byHsp90 requires the assistance of the Hsp70 chap-erone and numerous Hsp90 cochaperones.Hsp90 is a homodimer with each protomer

containing an N-terminal ATP binding domain[nucleotide-binding domain (NBD)], a middle domain(M-domain) that participates in client binding [5–7]and a C-terminal domain (C-domain) that is involvedin dimerization and client binding [4,5]. EukaryoticHsp90 also contains a linker region of about 50amino acids between the NBD and the M-domainand a C-terminal extension of 35 amino acids thatinteracts with several cochaperones. Hsp90 un-dergoes multiple conformational changes in re-

lsevier Ltd.

sponse to ATP binding and hydrolysis [8–14]. Inthe absence of ATP, the Hsp90 dimer adopts anopen V-shaped structure with the protomers inter-acting via the C-terminal dimerization domain [13].When ATP is bound, the protein adopts a closedconformation with the two N-domains of the dimerinteracting and a portion of the N-domain, the “lid”,closing over the nucleotide in each protomer [15].Additional conformational changes occur upon ATPhydrolysis [8–12], and after ADP release, Hsp90reverts back to the open conformation [2,11].Additionally, client protein binding and cochaperoneinteractions cause changes in the conformation ofHsp90 and affect the residence time in the variousconformations [1,14,16,17].The Hsp70 chaperone and more than 20 cocha-

perones, including Hop/Sti1, Aha1/Hch1, p23/Sba1,Cdc37 and Sgt1, collaborate with Hsp90 to remodeland activate the diverse group of client proteinsin the eukaryotic cytosol [1]. The cochaperonesregulate the Hsp90 ATPase activity and recruitspecific client proteins. Some cochaperonesdirect the chaperone cycle by stabilizing specific

J Mol Biol (2015) 427, 3877–3889

3878 Interaction between Hsp70 and Hsp90 of E. coli

conformations such as the open or the closed stateof Hsp90. For example, Hop/Sti1 interacts simulta-neously with Hsp70 and Hsp90 through its multipletetratricopeptide repeat domains and facilitatessubstrate transfer from Hsp70 to Hsp90 by stabiliz-ing the open conformation of Hsp90 [18,19].The bacterial homolog of Hsp90 in Escherichia

coli, the product of htpG and referred to as Hsp90Ec,is a very abundant protein under nonstress condi-tions and is further induced under stress condi-tions [1]. It shares about 50% sequence similaritywith human Hsp90. Hsp90Ec is not an essentialprotein under laboratory conditions [20]. However,when cells carry mutations in Hsp90Ec they growmore slowly at high temperature [20], exhibit a slightincrease in aggregated proteins at high temperature[21], lose adaptive immunity conferred by theCRISPR system [22] and show a subtle defect inmotility [23]. Additionally, when Hsp90Ec is over-expressed in E. coli, cells filament and becomesensitive to SDS [5].Both eukaryotic Hsp90 and Hsp90Ec have been

shown to remodel proteins in vitro. For example,eukaryotic Hsp90 reactivates denatured luciferase inconjunction with Hop/Sti1, Hsp70 and Hsp40[19,24,25]. Similarly, Hsp90Ec has the ability toreactivate heat-denatured luciferase in vitro [26]. Thisreaction requires ATP hydrolysis by Hsp90Ec and alsorequires DnaK, the E. coli homolog of Hsp70, and theDnaK cochaperone, DnaJ (or a DnaJ homolog, CbpA)[26]. GrpE, the prokaryotic nucleotide exchange factor,stimulates the rate of reactivation, although it is notessential [26]. Hsp90Ec and DnaK physically interact tomediate protein reactivation independent of a Hop/Sti1-like cochaperone [26]. E. coli Hsp90 is not unique[3,4,27]; recently, it has been reported that Hsp90 andHsp70 contact one another directly in complexescontaining Hop and client protein [28–30]. Moreover,Hsp90 from Synechococcus elongatus, Neurosporacrassa and Plasmodium falciparum have also beenshown or suggested to interact with their cognateHsp70 system [31–34]. Biochemical experiments usingE. coli proteins suggest that DnaK and DnaJ/CbpA actfirst on the client protein and then Hsp90Ec and theDnaK system collaborate synergistically to completeremodeling of the client protein [26].In this paper, we explored the mechanism of

collaboration between Hsp90Ec and DnaK both invivo and in vitro. We show that Hsp90Ec and DnaKform a binary complex and Hsp90Ec, DnaK and clientprotein form a ternary complex. CbpA promotesassembly of the ternary complex. We identifiedHsp90Ec mutants defective in DnaK interaction invivo and show that the purified mutant proteins aredefective in DnaK interaction in vitro and impaired inprotein reactivation with DnaK and its cochaperones.Together, these findings provide a better under-standing of how these two important chaperonescollaborate in client remodeling.

Results

Formation of an Hsp90Ec–DnaK–client proteincomplex is facilitated by CbpA

We previously showed that Hsp90Ec functionssynergistically with DnaK and its cochaperones, aJ-domain protein (CbpA or DnaJ) and GrpE in clientprotein reactivation in vitro [26]. In addition, weshowed that Hsp90Ec and DnaK interact in vivo in abacterial two-hybrid assay and in vitro using purifiedproteins [26]. To shed light on the mechanismof protein remodeling by Hsp90Ec and DnaK, wesought to dissect the multiprotein reaction pathwayinto intermediates and partial reactions.We explored the interaction between Hsp90Ec and

DnaK by testing whether binding of client protein orDnaK cochaperones affects the stability of thepreviously observed Hsp90Ec–DnaK complex [26].We used an in vitro protein–protein interaction assay(pull-down assay) in which DnaK was labeled withbiotin, DnaK D45C-biotin, and incubated with variouspure proteins in the presence of ATP (Fig. 1 andSupplemental Fig. S1a). Biotinylated DnaK, along withDnaK-associated proteins, was then captured onneutravidin agarose beads, the beads were washed,proteins were eluted and the eluted proteins wereanalyzed bySDS-PAGE.When biotinylatedDnaK andHsp90Ec were incubated together, Hsp90Ec weaklyassociated with DnaK (Fig. 1a, lane 2), as observedpreviously by anultrafiltration assay [26].When the twochaperones were incubated with ribosomal protein L2,a client protein known to interact with Hsp90Ec [5,16],we observed significantly more Hsp90Ec associatedwith biotinylatedDnaK and L2, suggesting formation ofa more stable ternary Hsp90Ec–DnaK–L2 complex(Fig. 1a, lane 3). CbpA further stimulated assembly orstabilization of the DnaK–Hsp90Ec–L2 complex(Fig. 1a, lane 4, and Supplemental Fig. S1b) and thestimulatory effect of CbpA required L2 (Fig. 1a, lane 6).Additional experiments indicated that DnaK couldassociate with L2 and CbpA alone, as well as L2 andCbpA together (Supplemental Fig. S1c). In contrast,DnaJ did not affect assembly of the DnaK–Hsp90Ec–L2 complex (Supplemental Fig. S1b). We do notunderstandwhyCbpAandDnaJ behaved differently inthese experiments and in protein reactivation [26].However, they differ in that DnaJ contains a cystei-ne-rich Zn2+ binding region (Type I J-domain protein)and CbpA lacks this region (Type II J-domainprotein). GrpE had no detectable effect on theassociation of Hsp90Ec with DnaK and L2 in thepresence of CbpA and was not detected inassociation with the complex (Fig. 1a, lane 5). In acontrol experiment, Hsp90Ec, L2, CbpA and GrpEdid not bind detectably to the neutravidin agarose(Fig. 1a, lane 8). Together, these results suggestthat the DnaK–Hsp90Ec complex is strengthened

(a)

Hsp90Ec

L2

Hsp90Ec

DnaK

DnaK(D45C-bio)Hsp90Ec

L2CbpAGrpE

CbpA

GrpE

Hsp90Ec

Hsp90Ec

DnaK

L2CbpA

DnaK(D45C-bio) Hsp90Ec

DnaK(NH3-bio) DnaK V436F(NH3-bio) DnaK T199A(NH3-bio)

L2, CbpA

(b)

Anti-Hsp90Ec

Anti-Hsp90Ec

Fig. 1. Stabilization of DnaK–Hsp90Ec complex in the presence of client protein and cochaperone. (a and b) Interactionbetween DnaK and Hsp90Ec in the presence or absence of L2, CbpA and GrpE. Pull-down assays were carried out asdescribed in Materials and Methods using biotinylated DnaK. Proteins associated with DnaK-biotin were analyzed bySDS-PAGE. Proteins were visualized by Coomassie blue staining; Hsp90Ec was monitored by immunoblot analysis usingHsp90Ec antiserum. (a) Pull-down assays contained biotinylated DnaK D45C, Hsp90Ec, L2, CbpA and GrpE whereindicated. (b) Assays contained biotinylated DnaK D45C, biotinylated DnaK wild-type or biotinylated DnaK mutant asindicated; Hsp90Ec wild-type or mutant as indicated; CbpA and L2. In (a) and (b), 2.4 μM DnaK, 3.6 μM Hsp90Ec, 2.3 μML2, 0.4 μM CbpA and 0.13 μM GrpE were used. M indicates DnaK, Hsp90Ec, L2, CbpA and GrpE as markers. In (a) and(b), representative gels from three or more independent experiments are shown.

3879Interaction between Hsp70 and Hsp90 of E. coli

upon binding of a client protein and CbpA promotesformation of the ternary complex.Since L2 independently binds Hsp90Ec [16] and

DnaK (Supplemental Fig. S1c) and also stabilizesthe Hsp90Ec–DnaK complex, we tested if clientbinding by Hsp90Ec, DnaK or both chaperones was

necessary for complex stabilization. When we mixedHsp90Ec W467R, a client-binding-defective mutant[5], with DnaK D45C-biotin in the presence of L2 andCbpA, we were unable to detect the mutant Hsp90Ecin association with DnaK (Fig. 1b, lane 3). We nexttested a DnaK mutant defective in client binding,

DnaKHtp

G

DnaK, H

tpG

Additive

0.00

0.05

0.10

0.15

0.20

0.25

(c)

(b) (a)

[L2] ( M)

AT

P h

ydro

lysi

s (n

mol

/min

)

0 1 2 3

1.5

1.0

0.5

0.0

DnaK

Hsp90Ec

Hsp90Ec + DnaK

WT

E34A

W467R

Controls

WT Hsp90Ec

+ L2

+ DnaK

+ L2 + DnaK

+ L2 + DnaK + GA

E34A

+ L2

+ DnaK

+ L2 + DnaK

W467R

+ L2

+ DnaK

+ L2 + DnaK

DnaK

L2

DnaK + L2

ATP hydrolysis (nmol/min)

0.0 0.2 0.4 0.6 0.8 1.0

Hsp90Ec:

DnaK:

0.0 0.2 0.4 0.6 0.8 1.0

WT

WT DnaK

+ L2

+ Hsp90Ec

+ L2 + Hsp90Ec

T199A

+ L2

+ Hsp90Ec

+ L2 + Hsp90Ec

V436F

+ L2

+ Hsp90Ec

+ L2 + Hsp90Ec

ATP hydrolysis (nmol/min)

Controls

T199A

V436F

Hsp90Ec

L2 + Hsp90Ec

0.2

0.1

0.0AT

P h

ydro

lysi

s (n

mol

/min

)

(d)

Fig. 2. Hsp90Ec and DnaK function synergistically in ATP hydrolysis. (a) ATP hydrolysis by DnaK, Hsp90Ec or thecombination of DnaK and Hsp90Ec. Data from seven replicates are presented as mean ± SEM (standard error of themean). The additive value of hydrolysis by DnaK alone and Hsp90Ec alone is shown as a black hatched bar and is meant toaid the reader. (b) ATP hydrolysis by Hsp90Ec, DnaK or Hsp90Ec and DnaK in the presence of increasing concentrations ofL2. (c) ATPase activity by the combination of wild-type or mutant Hsp90Ec and DnaK in the presence or absence of 1 μML2. Geldanamycin (30 μM) was added where indicated. (d) ATPase activity by the combination of wild-type or mutantDnaK and Hsp90Ec in the presence or absence of L2. In (a) to (d), ATPase was measured as described in Materials andMethods using 1 μM wild-type or mutant DnaK, 1 μM wild-type or mutant Hsp90Ec and 1 μM L2. In (b) and (c), data fromthree or more replicates are presented as mean ± SEM. In (c) and (d), broken lines indicate the additive ATPase activity ofwild-type Hsp90Ec with L2 and wild-type DnaK with L2 and are meant to aid the eye.

3880 Interaction between Hsp70 and Hsp90 of E. coli

V436F [35,36], in pull-down experiments usingbiotinylated V436F. Hsp90Ec was not detected inassociation with DnaK V436F in the presence of L2and CbpA (Fig. 1b, lane 6). Together, these resultsdemonstrate that the ternary DnaK–Hsp90Ec–L2complex requires client binding by both Hsp90Ecand DnaK.To assess whether ATP hydrolysis by Hsp90Ec

and/or DnaK were required for formation of theternary complex, we tested ATP-hydrolysis-defec-tive variants of the two chaperones. We found thatHsp90Ec E34A, a mutant that binds but does nothydrolyze ATP [8], associated much more weaklywith DnaK D45C-biotin than wild-type Hsp90Ec inthe presence of L2 and CbpA (Fig. 1b, lane 4).Similarly, a DnaKmutant, T199A, which can bind butnot hydrolyze ATP [37], was biotinylated and found

to bind Hsp90Ec more weakly than a similarlybiotinylated wild-type DnaK in the presence of L2and CbpA (Fig. 1b, lane 7). Together, these resultsindicate that ATP hydrolysis and/or ATP-drivenconformational changes by both chaperones areimportant for formation of a stable Hsp90Ec–DnaK–L2complex.

ATP hydrolysis by Hsp90Ec and DnaK issynergistically stimulated

The demonstration of an Hsp90Ec–DnaK–clientcomplex prompted us to examine the functionalsignificance of the complex by monitoring the effectof a client protein, L2, on ATP hydrolysis by thecombined action of Hsp90Ec and DnaK. In theabsence of L2, the rate of hydrolysis by the mixture

E269

R267

Q358

D366

R355

K354

G270/K271

K238

(a)T18-Hsp90Ec

T25-DnaK

(b)

(c)

1000

750

500

250

0

-gal

acto

sida

seac

tivity

(M

iller

uni

ts)

T18-Hsp90Ec:

T25-DnaK T25-empty

T18-R267CT25-DnaK

T18-E269CT25-DnaK

T18-G270A/K271AT25-DnaK

T18-K354CT25-DnaK

T18-R355CT25-DnaK

T18-R355LT25-DnaK

T18-Q358CT25-DnaK

T18-D366CT25-DnaK

T18-Hsp90EcT25-empty

T18-emptyT25-DnaK

T18-ZipT25-Zip

T18-Hsp90EcT25-empty

T18-ZipT25-Zip

T18-Hsp90EcT25-DnaK

T18-K238CT25-DnaK

Fig. 3. Identification of Hsp90Ec amino acid residues involved in DnaK interaction in vivo. (a and b) Interaction betweenDnaK and Hsp90Ec wild-type or mutant in a bacterial two-hybrid system in vivo, as described in Materials and Methods.DnaK was fused to one domain of B. pertussis adenylate cyclase, T25. Hsp90Ec wild-type and mutants were each fused tothe other domain, T18. T25-DnaK was coexpressed with each of the T18-Hsp90Ec mutants separately in cya− ΔhtpG cellsand the interaction between DnaK and Hsp90Ec wild-type or mutant was monitored by the expression of a reporter gene,β-galactosidase, on MacConkey indicator plates (a) and in liquid assays (b). In (a), a representative plate from threeindependent experiments is shown. In (b), β-galactosidase activity is shown as mean ± SEM (n = 3). (c) Surface-renderedmodel of the crystal structure of the Hsp90Ec dimer in the apo form (PDB ID: 2ioq) [13] with the C-terminal domains alignedto the crystal structure of the isolated C-terminal domain (PDB ID: 1sf8) [39] using PyMOL (www.pymol.org). One protomeris gray. The NBD, M-domain and C-terminal domain of the other protomer are colored pale blue, wheat and pale green,respectively. The mutated residues are labeled and colored.

3881Interaction between Hsp70 and Hsp90 of E. coli

of Hsp90Ec and DnaK was only slightly greater thanthe sum of hydrolysis by each chaperone separately,suggesting the possibility of a synergistic stimulationof ATP hydrolysis (Fig. 2a). As previously shown,ATP hydrolysis by Hsp90Ec alone was stimulatedby L2 [5,16] (Fig. 2b–d), while ATP hydrolysis byDnaK was unaffected (Fig. 2b–d). In the presence of1 μM L2, the rate of hydrolysis by the pair ofchaperones was ~1.6-fold higher than the additive

rates of hydrolysis of each chaperone separately inthe presence of L2 (Fig. 2b–d). These observationssuggest that the physical interaction betweenHsp90Ec and DnaK is reflected in a functionalcollaboration between the two chaperones in stim-ulating ATP hydrolysis.To determine if Hsp90Ec and DnaK act synergis-

tically or if one chaperone activates the ATPaseof the other in the presence of L2, we tested ATP-

3882 Interaction between Hsp70 and Hsp90 of E. coli

hydrolysis-defective Hsp90Ec and DnaK mutantproteins. The synergistic activity was preventedwhen Hsp90Ec E34A [26] was substituted for wild-type (Fig. 2c). Geldanamycin, a specific inhibitor of

a)WT Hsp90Ec E269C

R355L

R355C

Dna

K R

etai

ned

(%)

60

40

20

0

G270A/K271A

[Hsp90Ec] ( M)0 2 4 6 8

b)50

40

30

20

10

0

Dna

K R

etai

ned

(%)

c)

Hsp90Ec

Hsp90Ec

DnaK

L2CbpA

Anti-Hsp90Ec

Fig. 4. Hsp90Ec M-domain mu-tants are defective in DnaK interac-tions in vitro. (a) Interaction betweenDnaKandHsp90Ecwasmonitored bymeasuring retention of [3H]DnaK oncellulose filters with a 100-kDaexclusion limit in the presence oincreasing concentrations of wild-type ormutant Hsp90Ec, as describedin Materials and Methods. (b) Inter-action between DnaK and Hsp90Ecwild-type or mutant or BSA wasmeasured by ultrafiltration as in (a)using 1 μM Hsp90Ec. (c) Interactionbetween DnaK D45C-biotin andwild-type or mutant Hsp90Ec in thepresence of L2 and CbpA wasdetermined as described in Materialsand Methods. DnaK-associatedproteins were analyzed by immuno-blot analysis and Coomassie bluestaining following SDS-PAGE. In (a)and (b), data from at least threereplicates are presented as mean ±SEM. In (a), the apparent Kd valuesfor the wild-type, R355C, R355LG270A/K271A and E269C were0.42, 1.49, 1.79, 1.53 and 0.85 μMrespectively. In (c), 2.4 μM DnaK3.6 μM Hsp90Ec, 2.3 μM L2 and0.4 μM CbpA were used and arepresentative gel from three inde-pendent experiments is shown.

(

(

(

Hsp90 [38], also blocked the synergistic stimulationof ATPase activity (Fig. 2c). In addition, DnaK T199Awas unable to stimulate ATPase activity withHsp90Ec in the presence of L2 (Fig. 2d). Thus, ATP

f

,

,,

3883Interaction between Hsp70 and Hsp90 of E. coli

hydrolysis and/or the associated ATP-dependentconformational changes by both Hsp90Ec and DnaKare essential for the collaborative activity of the twochaperones.As seen for Hsp90Ec–DnaK–L2 complex forma-

tion, client binding by both chaperones was essentialfor synergistic ATPase stimulation in the presence ofL2. ATP hydrolysis was not stimulated aboveadditive when a client-binding-defective mutant ofHsp90Ec, W467R [5], or DnaK, V436F [35,36], wassubstituted for wild-type in the ATPase assay with L2(Fig. 2c and d). These results show that bothchaperones must have functional client bindingsites to collaborate in synergistic ATP hydrolysis.In contrast to ternary complex formation, we saw nodetectable effect of CbpA and GrpE on the syner-gistic stimulation of ATP hydrolysis by the twochaperones in the presence of L2 (SupplementalFig. S2a). DnaJ and GrpE slightly inhibited thestimulation of ATP hydrolysis by DnaK and Hsp90Ecin the presence of L2 (Supplemental Fig. S2b).Together, these results demonstrate that Hsp90Ec

and DnaK form a physical and functional complex inthe presence or absence of a client protein.

Mutations in the M-domain of Hsp90Ec causedefective interaction with DnaK in vivo

We developed a screen for the isolation ofHsp90Ec mutants potentially defective in DnaKinteraction with the aim of determining the region ofHsp90Ec that interacts with DnaK. The screen tookadvantage of our previous observation that Hsp90Ecand DnaK interact in vivo in a bacterial two-hybridassay [26]. In our earlier work, plasmids carryingfusions between the gene coding for Hsp90Ec andone fragment of the Bordetella pertussis adenylatecyclase gene, T18, and between DnaK and the otherfragment of the cyclase gene, T25, were construct-ed. When the two plasmids were coexpressed in anE. coli strain carrying a mutation in the geneencoding adenylate cyclase (cya−) and a deletionof the gene encoding Hsp90Ec (ΔhtpG), the cAMPreporter gene, β-galactosidase, was expressed andcolonies appeared red on indicator plates [26](Fig. 3a and b). These results indicate that DnaKand Hsp90Ec interact in vivo but do not exclude thepossibility that other cellular proteins are involved inthe interaction, such as client proteins. However,Hsp90Ec W467R and other client-binding-defectivemutants interacted with DnaK similarly to wild-typeHsp90Ec in the two-hybrid assay (Supplemental Fig.S3a).We constructed a T18-Hsp90Ec plasmid library

containing randomly mutagenized htpG. We thencoexpressed the T18-Hsp90Ec mutagenized plasmidlibrary with the T25-DnaK plasmid in E. coli cya−

ΔhtpG and screened for white colonies on indicatorplates (see Supplemental Methods). Among the

Hsp90Ec mutants obtained, two contained R355substitutions; one had a substitution to Cys andthe other one had a substitution to Leu. We re-constructed the R355C and R355L mutations inthe T18-Hsp90Ec plasmid since the mutated genesalso coded for several additional amino acidchanges. When T18-Hsp90Ec R355C or T18-Hsp90Ec R355L was coexpressed with T25-DnaK inthe cya− ΔhtpG strain, we observed that the colonieswere verypale pink on indicator plates (Fig. 3a) and theβ-galactosidase level was ~10% that of wild-type(Fig. 3b). These results suggest that residue R355 ofHsp90Ec is important for the in vivo interaction withDnaK.Hsp90Ec R355 is located in a long α-helix in the

M-domain and is surface exposed in both the openand the closed conformations of Hsp90Ec [13](Fig. 3c and Supplemental Fig. S3b and S3c). Toidentify additional Hsp90Ec residues important forDnaK interaction in vivo, we substituted amino acidsin other surface-exposed residues in the vicinity ofR355 using site-directed mutagenesis. Three mu-tants were constructed in the same α-helix as R355,including K354C, Q358C and D366C (Fig. 3c andSupplemental Fig. S3b and S3c). Other mutants inthe region, including K238C, R267C, E269C andG270A/K271A, were also constructed (Fig. 3c andSupplemental Fig. S3b and S3c). The mutantproteins were then tested for the ability to interactwith DnaK in the bacterial two-hybrid assay. Weobserved that strains expressing T25-DnaK andT18-Hsp90Ec K238C, E269C or G270A/K271A werepink on indicator plates (Fig. 3a) while thoseexpressing T25-DnaK and T18-Hsp90Ec R267C,K354C, Q358C or D366C were red (Fig. 3a). Thelevel of β-galactosidase produced by the cellsreflected the color of the colonies: β-galactosidaselevels in cells expressing T25-DnaK andT18-Hsp90Ec K238C, G270A/K271A or E269Cwere ~16% that of wild-type (Fig. 3b). Cellsexpressing the other four mutants, T18-Hsp90EcR267C, K354C, Q358C or D366C with T25-DnaK,produced levels of β-galactosidase slightly lowerthan wild-type (~80% of wild-type) (Fig. 3b). Incontrol experiments, the steady-state levels of all ofthe mutant fusion proteins were similar to wild-type(Supplemental Fig. S3d).Thus, these mutants define a region in the Hsp90Ec

M-domain that is important for interaction with DnaKin vivo. However, the two-hybrid assay does notdistinguish between a binary interaction and aninteraction involving other cellular components.

DnaK-interaction-defective Hsp90Ec mutantproteins identified in vivo are defective in DnaKbinding in vitro

To determine if the Hsp90Ec mutant proteins weidentified were defective in direct interaction with

AT

P h

ydro

lysi

s (n

mol

/min

)

1.0

0.8

0.6

0.4

0.2

0.0

0.4

0.2

0.0

AT

P h

ydro

lysi

s (n

mol

/min

)

- L2+ L2

(a)

(b)

Additive value: (Hsp90Ec, L2) + (DnaK, L2)Hsp90Ec, DnaK, L2

(c)

0 10 20 30 40

R267CK354CE269CG270A/K271AR355LR355CK238C

Luci

fera

se r

eact

ivat

ion

(%)

Time (min)

WT Hsp90Ec

Luc alone

DnaK, CbpA, GrpE

12

9

6

3

0

+ DnaK,CbpA,GrpE

Fig. 5. Hsp90Ec mutants defective in DnaK interaction are defective in functional collaboration with DnaK. (a) ATPaseactivity of wild-type or mutant Hsp90Ec in the absence or presence of L2 was measured as described in Materials andMethods. (b) ATPase activity was determined in the presence of wild-type or mutant Hsp90Ec, DnaK and L2. The additivevalue of ATP hydrolysis by DnaK in the presence of L2 and Hsp90Ec wild-type or mutant in the presence of L2 is shown bygray bars to aid the reader. (c) Reactivation of heat-denatured luciferase was monitored over time as described inMaterials and Methods using Hsp90Ec wild-type or mutant in combination with DnaK, CbpA and GrpE. Wild-type, black;R267C, orange; K354C, purple; E269C, blue; G270A/K271A, pink; R355L, red; R355C, green; K238C, light blue; DnaK,CbpA, GrpE alone, black with open circles; heat-denatured luciferase alone, gray. In (a) to (c), data from at least threereplicates are presented as mean ± SEM.

3884 Interaction between Hsp70 and Hsp90 of E. coli

3885Interaction between Hsp70 and Hsp90 of E. coli

DnaK, we purified the mutant proteins and testedthem for their ability to bind DnaK in vitro in theabsence of other proteins. Using an ultrafiltrationassay, we previously showed that labeled DnaK wasspecifically retained on 100-kDa molecular weightcutoff filters in the presence of wild-type Hsp90Ec[26] (Fig. 4a). When we tested the Hsp90Ec mutantproteins in this assay, we observed that Hsp90EcR355L, R355C, G270A/K271A, E269C, R267C andK354C were partially defective in DnaK bindingcompared to wild-type (Fig. 4a and b). Hsp90EcQ358C and D366C were like the wild-type (Fig. 4b).These results show that the Hsp90Ec mutants thatwere most defective in DnaK interaction in vivo aredefective in DnaK interaction in vitro.We next tested the mutant proteins in the pull-down

assay and again found that themutants that were verydefective in the two-hybrid assay, Hsp90Ec R355C,R355L, K238C, G270A/K271A and E269C, were alsodefective in the ability to associate with biotinylatedDnaK in the presence of L2 with or without CbpA(Fig. 4c andSupplemental Fig. S4a). Themutants thatwere partially defective in the two-hybrid assay,R267C and K354C, showed decreased ability tointeract with DnaK in the presence of L2 (Fig. 4c).Altogether, our results suggest that a region of theHsp90EcM-domain nearR355 is important for a binaryinteraction with DnaK and for a ternary interaction withDnaK and L2, in the presence or absence of CbpA.

DnaK-binding-defective Hsp90Ec mutants areimpaired in functional collaboration with DnaKin vitro

Based on the in vivo and in vitro protein–proteininteraction results, we predicted that the Hsp90mutants defective in DnaK interaction would also bedefective in the synergistic stimulation of ATPhydrolysis with DnaK. In control experiments, weobserved that the mutant Hsp90Ec proteins hadbasal rates of ATP hydrolysis similar to wild-typeshowing that the mutations do not cause defects inATP hydrolysis (Fig. 5a). Moreover, in the presenceof the client protein, L2, ATPase activity of theHsp90Ec mutants was stimulated similar to wild-type(Fig. 5a). Since ATP hydrolysis by client-binding-de-fective Hsp90Ec mutants is not stimulated by L2 [5](Fig. 2c), these observations suggest that Hsp90Ecmutants defective in DnaK interaction are notdefective in client binding.We then tested the mutants for their ability to

synergistically stimulate ATP hydrolysis in combi-nation with DnaK in the presence of L2. WhenHsp90Ec R355C, R355L, K238C, E269C or K354Cwas substituted for wild-type, synergistic stimulationof ATPase was not observed; instead, hydrolysiswas similar to the sum of hydrolysis by eachchaperone alone with L2 (Fig. 5b). ATP hydrolysisby Hsp90Ec G270A/K271A or R267C and DnaK in

the presence of L2 was greater than additive, but thestimulation was less than wild-type Hsp90Ec(Fig. 5b). ATP hydrolysis by Hsp90Ec Q358C orD366C with DnaK and L2 was similar to or greaterthan the synergistic stimulation seen with wild-typeHsp90Ec with DnaK and L2 (Fig. 5b). These resultsprovide in vitro evidence to suggest that the regionsurrounding R355 in the M-domain is important forthe functional collaboration between Hsp90Ec andDnaK.

Hsp90Ec mutant proteins defective in DnaKinteraction are defective in protein reactivationin collaboration with the DnaK system in vitro

To determine if the Hsp90Ec residues important forinteraction with DnaK are also important for proteinreactivation by Hsp90Ec, we tested the mutantproteins in an in vitro protein reactivation assay.We previously showed that the combination ofwild-type Hsp90Ec and the DnaK system, composedeither of DnaK, CbpA and GrpE or of DnaK, DnaJand GrpE, function together to reactivate heat-inactivated luciferase [26] (Fig. 5c). Hsp90Ec aloneis unable to reactivate luciferase and the DnaKsystem reactivates luciferase poorly [26] (Fig. 5c).Four mutants, Hsp90Ec R355C, R355L, K238C andG270A/K271A, were defective in luciferase reacti-vation and exhibited rates of reactivation indistin-guishable from the DnaK system alone (Fig. 5c).Hsp90Ec E269C, K354C and R267C were partiallydefective in luciferase reactivation, displaying ratesof reactivation ~50% of wild-type Hsp90Ec with theDnaK system (Fig. 5c). Hsp90Ec Q358C and D366Creactivated luciferase at rates comparable to wild-type Hsp90Ec (Supplemental Fig. S5). Thus, theHsp90Ec residues that are important for the interac-tion with DnaK both in vivo and in vitro and for thesynergistic stimulation of ATPase with DnaK are alsoimportant in client remodeling with the DnaK system.Altogether, these results suggest that an

Hsp90Ec–DnaK complex and a larger complex ofHsp90Ec–DnaK–client protein are intermediates inthe pathway of protein remodeling. Additionally, theysuggest that DnaK interacts directly with Hsp90Ecthrough residues in the middle domain of Hsp90Ec.Collectively, these results provide insight into themechanism of collaboration during protein remodel-ing between two important molecular chaperones.

Discussion

In this work, we explored the mechanism ofcollaboration between Hsp90 of E. coli and theDnaK chaperone system. Unlike client activationand remodeling by eukaryotic Hsp90 and the Hsp70system, the E. coli system is independent of Hsp90cochaperones and thus provides a much simpler

3886 Interaction between Hsp70 and Hsp90 of E. coli

system to study. It is possible that cochaperones,which have not yet been identified in E. coli, areinvolved in organizing a complex of Hsp90Ec and theDnaK chaperone system. However, it is alsopossible that the DnaK system by interacting withHsp90Ec directly bypasses the need for additionalcochaperones and provides some of the regulatoryroles of the eukaryotic Hsp90 cochaperones.Our results show that Hsp90Ec and DnaK form a

weak complex that is stabilized by client binding.Ternary complex formation is facilitated by a J-protein,CbpA. Whether CbpA is associated with the DnaK–Hsp90Ec–client complex remains to be unequivocallydetermined. ATP hydrolysis or conformational chang-es associated with hydrolysis by both Hsp90Ec andDnaK is required for this collaboration. Moreover, ATPhydrolysis is synergistically stimulated by the twochaperones in the presence of client protein, suggest-ing the functionality of the Hsp90Ec–DnaK complex.However, further study is necessary to clarify thesynergistic action of the two chaperones and deter-mine whether DnaK stimulates the activity of Hsp90Ec,Hsp90Ec stimulates the activity of DnaK or theystimulate each other.In Eukaryotes, recent studies have identified

multiprotein assemblies of Hsp90, Hsp70 and Hopwith client protein [28–30]. Although Hsp40 wasnot observed in the chaperone complexes, it wasincluded in the reaction mixtures for formation of thecomplexes. The presence of dimeric Hsp90, Hop anda client (a fragment of glucocorticoid receptor) wascommon to all of the ternary complexes. Thequaternary form of Hsp70 observed differed betweenthe studies. In one study, Hsp70 formed an antiparalleldimer in the complex [30]; in the other studies, itwas a monomer [28,29]. In all three studies, aninteraction was observed between Hsp90 andHsp70, although the regions of interaction suggestedby the various studies differed [28–30]. Moreover,posttranslational modifications have been shown toaffect formation of the complexes [30]. Together,these observations suggest the existence of multipleconformations for the Hsp90–Hop–Hsp70–substratecomplex, but the details of the interactions be-tween the proteins in the complexes remain to beelucidated.The recent visualization of the eukaryotic Hsp90–

Hsp70–Hop–client complex and earlier biochemicalwork from many groups suggest a mechanism forsubstrate transfer from Hsp70 to Hsp90 [19,28–30,40,41]. In the current model, the Hsp70 cochaper-one, Hsp40, likely presents the client to Hsp70 andstabilizes the interaction between the client and Hsp70prior to the binding of Hsp70 to Hop and Hop to Hsp90[28–30]. The formation of a quaternary Hsp90–Hsp70–Hop–client complex then positions theHsp70-bound client near the Hsp90 client-bindingregion [28–30]. Coordination of the ATP hydrolysiscycles by the two chaperones, in ways that are not fully

understood, promotes conformational changes in thechaperones and leads to substrate transfer andmaturation [28–30].Our previous work [26] and that presented here to

elucidate the mechanism of protein remodeling bythe combined action of DnaK and Hsp90Ec suggest asimilar but simpler mechanism of client remodelingby the prokaryotic chaperones than by the eukary-otic chaperones [26]. First, like remodeling by theDnaK system alone [35,42], DnaJ/CbpA targets theclient for recognition by DnaK and some initialprotein remodeling is performed by DnaK andDnaJ/CbpA alone. This step requires ATP hydroly-sis. Next, DnaK recruits Hsp90Ec to the client via adirect interaction between DnaK and Hsp90Ec. Thisinteraction is further stabilized by an additionalinteraction between the client and Hsp90Ec and isfacilitated by a J-domain protein. Then, Hsp90Ec andDnaK act synergistically in a reaction requiring ATPhydrolysis by both chaperones. Likely coordinated,ATP-hydrolysis-driven conformational changes inHsp90Ec and DnaK promote remodeling and releaseof the client [26].In an analogous reaction to the collaboration

between DnaK and Hsp90Ec in prokaryotes andamong Hsp70, Hsp90 and Hop in eukaryotes, DnaKand Hsp70 have been shown to function collabora-tively with the ClpB and Hsp104 AAA+ disaggre-gases, respectively, in protein disaggregation invivo and in vitro [43–45]. The collaboration betweenHsp104/ClpB by Hsp70/DnaK involves a directprotein–protein interaction between the two chap-erones [46–50]. Additionally, ATP hydrolysis by theDnaK system and ClpB is synergistically stimulatedin the presence of aggregated substrate [44]. Thesimilarities in these two bichaperone protein remod-eling reactions suggest a new role for DnaK/Hsp70as a regulator of the activity of multiple ATP-depen-dent molecular chaperones.In summary, this study illuminates how E. coli

Hsp90 and Hsp70 interact and how they collaboratein protein remodeling. Understanding this interac-tion is of universal importance and is critical todeveloping cancer therapies, possibly drug combi-nations that simultaneously target Hsp90 andHsp70 or a cochaperone.

Materials and Methods

Plasmids and strains

Single-substitution mutations of Hsp90Ec and DnaKwere made with the QuikChange mutagenesis system(Stratagene) using pET-HtpG [26], pRE-DnaK [51],pET-DnaK [52] or pT18-Hsp90Ec [26]. All mutationswere verified by DNA sequencing. The Hsp90Ecrandom mutagenesis library and Bth101ΔhtpG wereconstructed as described in Supplementary Methods.

3887Interaction between Hsp70 and Hsp90 of E. coli

Proteins

Hsp90Ec wild-type and mutants [26], DnaK wild-type andmutants [51], DnaJ [51], CbpA [53], GrpE [51] and His-tagged L2 [16] were isolated as previously described. Allproteins were N95% pure as determined by SDS-PAGE.Mutant Hsp90Ec proteins exhibited ATPase activity andgel-filtration chromatograms similar to wild-type (Fig. 5aand Supplemental Fig. S4b). Luciferase and luciferin werefrom Promega and Roche, respectively. Concentrationsgiven are for Hsp90Ec, DnaJ, CbpA and GrpE dimers andfor DnaK, L2 and luciferase monomers. DnaK D45C waslabeled using a 20-fold excess of Maleimide-PEG11-Biotin(PEG, polyethylene glycol) (Thermo, Life Technologies)and DnaK wild-type, T199A and V436F were labeled usinga 1.5-fold excess of NHS-PEG4-Biotin (Thermo, LifeTechnologies) as recommended by the manufacturer.Excess biotin reagent was removed by dialysis. DnaKD45C-biotin was similar to wild-type in luciferase reactiva-tion (Supplemental Fig. S1a). DnaK was labeled with 3H aspreviously described (740 cpm/pmol) [54].

Luciferase reactivation

Luciferase reactivation was performed as previ-ously described with modifications [26]. We incubated40 nM heat-denatured luciferase at 24 °C in reactionmixtures (75 μl) containing 25 mM Hepes (pH 7.5),50 mM KCl, 0.1 mM ethylenediaminetetraacetic acid,2 mM DTT, 10 mM MgCl2, 50 μg/ml bovine serumalbumin (BSA), 3 mM ATP, an ATP regenerating system(25 mM creatine phosphate and 6 μg creatine kinase),0.95 μMDnaK, 0.15 μMCbpA, 0.05 μMGrpE and 0.5 μMHsp90Ec wild-type or mutant. Aliquots were removedat the indicated times and light output was measuredusing a Tecan Infinite M200Pro in luminescence modewith an integration time of 1000 ms. Reactivation wasdetermined compared to a nondenatured luciferasecontrol.

ATPase activity

Steady-state ATP hydrolysis was measured as previ-ously described using 1 μM Hsp90Ec wild-type or mutant[5,8] with L2 (1 μM), DnaK wild-type or mutant (1 μM) andgeldanamycin (30 μM) where indicated.

Ultrafiltration assay

Association of Hsp90Ec and DnaK was measuredusing an ultrafiltration assay as previously describedwith modifications [26]. We incubated 0.13 μM [3H]DnaKat 24 °C for 5 min in reaction mixtures (100 μl) contain-ing 20 mM Tris–HCl (pH 7.5), 75 mM KCl, 10% glyc-erol (vol/vol), 0.05% Triton X-100 (vol/vol), 5 mM DTT andHsp90Ec wild-type or mutant or BSA as indicated. Reactionswere filtered through Microcon DNA Fast Flow filters(Millipore) by centrifugation at 3200g for 10 min. Retainedproteins were recovered with 10% SDS and radioactivitywas measured. Background corrections were made bysubtracting the percentage of [3H]DnaK retained in theabsence of Hsp90Ec (b10%).

Protein–protein interaction assay

A pull-down assay was used to measure interaction ofHsp90Ec with DnaK. We incubated 2.4 μM DnaK D45C-biotin for 5 min at 23 °C in reaction mixtures (50 μl)containing PD buffer [20 mM Tris–HCl (pH 7.5), 75 mMKCl, 10% glycerol (vol/vol), 0.01% Triton X-100 (vol/vol),2 mM DTT, 10 mM MgCl2 and 2 mM ATP] with 3.6 μMHsp90Ec wild-type or mutant, 2.3 μM L2, 0.4 μM CbpA and0.13 μM GrpE. We then added 20 μl of neutravidinagarose (Thermo, Pierce) and incubated it for 5 min at23 °C with mixing. The reactions were diluted with 0.4 mlPD buffer, centrifuged for 1 min at 1000g, and therecovered agarose beads were washed twice with 0.4 mlPDbuffer. Bound proteins were eluted with buffer containing2 M NaCl and analyzed by immunoblot analysis orCoomassie blue staining following SDS-PAGE. Whereindicated, biotinylated DnaK wild-type, DnaK T199A orDnaK V436F was used.

Bacterial two-hybrid assay

Bacterial two-hybrid assays were performed as previ-ously described [26,55].

Acknowledgements

We thank Dan Masison, Michael Reidy and AureliaBattesti for many helpful discussions. This researchwas supported by the Intramural Research Programof the National Institutes of Health, National CancerInstitute, Center for Cancer Research.Author Contributions: O.G., J.R.H., A.N.K.,

S.M.D. and S.W. designed the experiments. O.G.,J.R.H., A.N.K. and S.M.D. performed the experi-ments. All authors were involved in data interpreta-tion and discussion. S.M.D. and S.W. wrote themanuscript with contributions from all other authors.The authors declare no competing financial interests.

Appendix A. Supplementary data

Supplementary data to this article can be foundonline at http://dx.doi.org/10.1016/j.jmb.2015.10.010.

Received 18 August 2015;Received in revised form 30 September 2015;

Accepted 9 October 2015Available online 23 October 2015

Keywords:Hsp40;CbpA;DnaJ;

molecular chaperone;protein remodeling

3888 Interaction between Hsp70 and Hsp90 of E. coli

Present address: O. Genest, Laboratoire de Bioénergé-tique et Ingénierie des Protéines, Aix Marseille Université,

13400 Marseille, France.

†O.G., J.R.H. and A.N.K. contributed equally to this work.

Abbreviations used:BSA, bovine serum albumin; NBD, nucleotide-binding

domain.

References

[1] J.L. Johnson, Evolution and function of diverse Hsp90homologs and cochaperone proteins, Biochim. Biophys.Acta 1823 (2012) 607–613.

[2] M.P. Mayer, Gymnastics of molecular chaperones, Mol. Cell39 (2010) 321–331.

[3] L.H. Pearl, C. Prodromou, Structure and mechanism ofthe Hsp90 molecular chaperone machinery, Annu. Rev.Biochem. 75 (2006) 271–294.

[4] A. Rohl, J. Rohrberg, J. Buchner, The chaperone Hsp90:Changing partners for demanding clients, Trends Biochem.Sci. 38 (2013) 253–262.

[5] O. Genest, et al., Uncovering a region of heat shock protein90 important for client binding in E. coli and chaperonefunction in yeast, Mol. Cell 49 (2013) 464–473.

[6] C. Prodromou, The “active life” of Hsp90 complexes,Biochim. Biophys. Acta 1823 (2012) 614–623.

[7] D.R. Southworth, D.A. Agard, Species-dependent ensem-bles of conserved conformational states define the Hsp90chaperone ATPase cycle, Mol. Cell 32 (2008) 631–640.

[8] C. Graf, M. Stankiewicz, G. Kramer, M.P. Mayer, Spatiallyand kinetically resolved changes in the conformationaldynamics of the Hsp90 chaperone machine, EMBO J. 28(2009) 602–613.

[9] K.A. Krukenberg, F. Forster, L.M. Rice, A. Sali, D.A. Agard,Multiple conformations of E. coli Hsp90 in solution: Insightsinto the conformational dynamics of Hsp90, Structure 16(2008) 755–765.

[10] K.A. Krukenberg, D.R. Southworth, T.O. Street, D.A. Agard,pH-dependent conformational changes in bacterial Hsp90reveal a Grp94-like conformation at pH 6 that is highly activein suppression of citrate synthase aggregation, J. Mol. Biol.390 (2009) 278–291.

[11] K.A. Krukenberg, T.O. Street, L.A. Lavery, D.A. Agard,Conformational dynamics of the molecular chaperoneHsp90, Q. Rev. Biophys. 44 (2011) 229–255.

[12] C. Ratzke, M. Mickler, B. Hellenkamp, J. Buchner, T. Hugel,Dynamics of heat shock protein 90 C-terminal dimerization isan important part of its conformational cycle, Proc. Natl.Acad. Sci. U. S. A. 107 (2010) 16101–16106.

[13] A.K. Shiau, S.F. Harris, D.R. Southworth, D.A. Agard,Structural analysis of E. coli hsp90 reveals dramaticnucleotide-dependent conformational rearrangements, Cell127 (2006) 329–340.

[14] M. Taipale, D.F. Jarosz, S. Lindquist, HSP90 at the hub ofprotein homeostasis: Emerging mechanistic insights, Nat.Rev. Mol. Cell Biol. 11 (2010) 515–528.

[15] M.M. Ali, et al., Crystal structure of an Hsp90-nucleotide-p23/Sba1 closed chaperone complex, Nature 440 (2006)1013–1017.

[16] Y. Motojima-Miyazaki, M. Yoshida, F. Motojima, Ribosomalprotein L2 associates with E. coli HtpG and activates itsATPase activity, Biochem. Biophys. Res. Commun. 400(2010) 241–245.

[17] T.O. Street, L.A. Lavery, D.A. Agard, Substrate binding driveslarge-scale conformational changes in the Hsp90 molecularchaperone, Mol. Cell 42 (2011) 96–105.

[18] H. Wegele, L. Muller, J. Buchner, Hsp70 and Hsp90—A relayteam for protein folding, Rev. Physiol. Biochem. Pharmacol.151 (2004) 1–44.

[19] H. Wegele, S.K. Wandinger, A.B. Schmid, J. Reinstein, J.Buchner, Substrate transfer from the chaperone Hsp70 toHsp90, J. Mol. Biol. 356 (2006) 802–811.

[20] J.C. Bardwell, E.A. Craig, Eukaryotic Mr 83,000 heat shockprotein has a homologue in Escherichia coli, Proc. Natl.Acad. Sci. U. S. A. 84 (1987) 5177–5181.

[21] J.G. Thomas, F. Baneyx, ClpB and HtpG facilitate de novoprotein folding in stressed Escherichia coli cells, Mol.Microbiol. 36 (2000) 1360–1370.

[22] I. Yosef, M.G. Goren, R. Kiro, R. Edgar, U. Qimron, High-temperature protein G is essential for activity of theEscherichia coli clustered regularly interspaced short palin-dromic repeats (CRISPR)/Cas system, Proc. Natl. Acad. Sci.U. S. A. 108 (2011) 20136–20141.

[23] M.O. Press, et al., Genome-scale co-evolutionary inferenceidentifies functions and clients of bacterial Hsp90, PLoSGenet. 9 (2013) e1003631.

[24] J.P. Grenert, B.D. Johnson, D.O. Toft, The importance ofATP binding and hydrolysis by hsp90 in formation andfunction of protein heterocomplexes, J. Biol. Chem. 274(1999) 17525–17533.

[25] B.D. Johnson, R.J. Schumacher, E.D. Ross, D.O. Toft, Hopmodulates Hsp70/Hsp90 interactions in protein folding, J.Biol. Chem. 273 (1998) 3679–3686.

[26] O.Genest, J.R. Hoskins, J.L. Camberg, S.M.Doyle, S.Wickner,Heat shock protein 90 from Escherichia coli collaborates withthe DnaK chaperone system in client protein remodeling, Proc.Natl. Acad. Sci. U. S. A. 108 (2011) 8206–8211.

[27] P.C. Echeverria, A. Bernthaler, P. Dupuis, B. Mayer, D.Picard, An interaction network predicted from public data as adiscovery tool: Application to the Hsp90 molecular chaper-one machine, PLoS One 6 (2011) e26044.

[28] S. Alvira, et al., Structural characterization of the substratetransfer mechanism in Hsp70/Hsp90 folding machinerymediated by Hop, Nat. Commun. 5 (2014) 5484.

[29] E. Kirschke, D. Goswami, D. Southworth, P.R. Griffin, D.A.Agard, Glucocorticoid receptor function regulated by coordi-nated action of the Hsp90 and Hsp70 chaperone cycles, Cell157 (2014) 1685–1697.

[30] N. Morgner, et al., Hsp70 forms antiparallel dimers stabilizedby post-translational modifications to position clients fortransfer to Hsp90, Cell Rep. 11 (2015) 759–769.

[31] G. Banumathy, V. Singh, S.R. Pavithra, U. Tatu, Heat shockprotein 90 function is essential for Plasmodium falciparumgrowth in human erythrocytes, J. Biol. Chem. 278 (2003)18336–18345.

[32] D.G. Freitag, P.M. Ouimet, T.L. Girvitz, M. Kapoor, Heatshock protein 80 of Neurospora crassa, a cytosolic molecularchaperone of the eukaryotic stress 90 family, interactsdirectly with heat shock protein 70, Biochemistry 36 (1997)10221–10229.

[33] P.J. Murphy, K.C. Kanelakis, M.D. Galigniana, Y. Morishima,W.B. Pratt, Stoichiometry, abundance, and functional signif-icance of the hsp90/hsp70-based multiprotein chaperone

3889Interaction between Hsp70 and Hsp90 of E. coli

machinery in reticulocyte lysate, J. Biol. Chem. 276 (2001)30092–30098.

[34] H. Nakamoto, et al., Physical interaction between bacterialheat shock protein (Hsp) 90 and Hsp70 chaperonesmediates their cooperative action to refold denaturedproteins, J. Biol. Chem. 289 (2014) 6110–6119.

[35] T. Laufen, et al., Mechanism of regulation of hsp70chaperones by DnaJ cochaperones, Proc. Natl. Acad. Sci.U. S. A. 96 (1999) 5452–5457.

[36] S. Rudiger, M.P. Mayer, J. Schneider-Mergener, B. Bukau,Modulation of substrate specificity of the DnaK chaperone byalteration of a hydrophobic arch, J. Mol. Biol. 304 (2000)245–251.

[37] T.K. Barthel, J. Zhang, G.C. Walker, ATPase-defectivederivatives of Escherichia coli DnaK that behave differentlywith respect to ATP-induced conformational change andpeptide release, J. Bacteriol. 183 (2001) 5482–5490.

[38] B. Panaretou, et al., ATP binding and hydrolysis are essentialto the function of the Hsp90 molecular chaperone in vivo,EMBO J. 17 (1998) 4829–4836.

[39] S.F. Harris, A.K. Shiau, D.A. Agard, The crystal structure ofthe carboxy-terminal dimerization domain of htpG, theEscherichia coli Hsp90, reveals a potential substrate bindingsite, Structure 12 (2004) 1087–1097.

[40] A. Rohl, et al., Hsp90 regulates the dynamics of itscochaperone Sti1 and the transfer of Hsp70 betweenmodules, Nat. Commun. 6 (2015) 6655.

[41] A.B. Schmid, et al., The architecture of functional modules inthe Hsp90 co-chaperone Sti1/Hop, EMBO J. 31 (2012)1506–1517.

[42] B. Bukau, A.L. Horwich, The Hsp70 and Hsp60 chaperonemachines, Cell 92 (1998) 351–366.

[43] S.M. Doyle, O. Genest, S. Wickner, Protein rescue fromaggregates by powerful molecular chaperone machines, Nat.Rev. Mol. Cell Biol. 14 (2013) 617–629.

[44] S.M. Doyle, J.R. Hoskins, S. Wickner, Collaboration betweenthe ClpB AAA+ remodeling protein and the DnaK chaperonesystem, Proc. Natl. Acad. Sci. U. S. A. 104 (2007)11138–11144.

[45] J.R. Glover, S. Lindquist, Hsp104, Hsp70, and Hsp40: Anovel chaperone system that rescues previously aggregatedproteins, Cell 94 (1998) 73–82.

[46] M. Carroni, et al., Head-to-tail interactions of the coiled-coildomains regulate ClpB activity and cooperation with Hsp70 inprotein disaggregation, Elife 3 (2014) e02481.

[47] S.M. Doyle, et al., Interplay between E. coli DnaK, ClpB andGrpE during protein disaggregation, J. Mol. Biol. 427 (2015)312–327.

[48] J. Lee, et al., Heat shock protein (Hsp) 70 is an activator ofthe Hsp104 motor, Proc. Natl. Acad. Sci. U. S. A. 110 (2013)8513–8518.

[49] R. Rosenzweig, S. Moradi, A. Zarrine-Afsar, J.R. Glover, L.E.Kay, Unraveling the mechanism of protein disaggregationthrough a ClpB-DnaK interaction, Science 339 (2013)1080–1083.

[50] B. Sielaff, F.T. Tsai, The M-domain controls Hsp104 proteinremodeling activity in an Hsp70/Hsp40-dependent manner,J. Mol. Biol. 402 (2010) 30–37.

[51] D. Skowyra, S. Wickner, The interplay of the GrpE heat shockprotein and Mg2+ in RepA monomerization by DnaJ andDnaK, J. Biol. Chem. 268 (1993) 25296–25301.

[52] M. Miot, et al., Species-specific collaboration of heat shockproteins (Hsp) 70 and 100 in thermotolerance and proteindisaggregation, Proc. Natl. Acad. Sci. U. S. A. 108 (2011)6915–6920.

[53] C. Ueguchi, M. Kakeda, H. Yamada, T. Mizuno, An analogueof the DnaJ molecular chaperone in Escherichia coli, Proc.Natl. Acad. Sci. U. S. A. 91 (1994) 1054–1058.

[54] J.R. Hoskins, M. Pak, M.R. Maurizi, S. Wickner, The role ofthe ClpA chaperone in proteolysis by ClpAP, Proc. Natl.Acad. Sci. U. S. A. 95 (1998) 12135–12140.

[55] G. Karimova, J. Pidoux, A. Ullmann, D. Ladant, A bacterialtwo-hybrid system based on a reconstituted signal transduc-tion pathway, Proc. Natl. Acad. Sci. U. S. A. 95 (1998)5752–5756.