Allosteric signal transmission in the nucleotide-binding domain of 70-kDa heat shock protein(Hsp70) molecular chaperonesAnastasia Zhuravleva and Lila M. Gierasch1

Department of Biochemistry and Molecular Biology and Department of Chemistry, University of Massachusetts, Amherst, MA 01003

Edited* by Wayne A. Hendrickson, Columbia University, New York, NY, and approved March 10, 2011 (received for review September 27, 2010)

The 70-kDa heat shock protein (Hsp70) chaperones perform a widearray of cellular functions that all derive from the ability of theirN-terminal nucleotide-binding domains (NBDs) to allosterically reg-ulate the substrate affinity of their C-terminal substrate-bindingdomains in a nucleotide-dependent mechanism. To explore thestructural origins of Hsp70 allostery, we performed NMR analysison the NBD of DnaK, the Escherichia coli Hsp70, in six differentstates (ligand-bound or apo) and in two constructs, one that retainsthe conserved and functionally crucial portion of the interdomainlinker (residues 389VLLL392) and another that lacks the linker. Che-mical-shift perturbation patterns identify residues at subdomaininterfaces that constitute allosteric networks and enable theNBD to act as a nucleotide-modulated switch. Nucleotide bindingresults in changes in subdomain orientations and long-range per-turbations along subdomain interfaces. In particular, our findingsprovide structural details for a key mechanism of Hsp70 allostery,by which information is conveyed from the nucleotide-binding siteto the interdomain linker. In the presence of ATP, the linker binds tothe edge of the IIA β-sheet, which structurally connects the linkerand the nucleotide-binding site. Thus, a pathway of allostericcommunication leads from the NBD nucleotide-binding site to thesubstrate-binding domain via the interdomain linker.

conformational ensemble ∣ NMR chemical-shift analysis ∣ subdomainreorientations ∣ actin fold ∣ signal propagation

The 70-kDa heat shock proteins (Hsp70s) compose one of themost well studied and ubiquitously distributed families of

allosteric proteins (1). Hsp70s assist in an extraordinarily broadspectrum of cellular processes, including protein folding, disag-gregation, and translocation. All chaperone activities of Hsp70sare based on their ability to interact with short hydrophobic pep-tide segments of protein substrate in an ATP-dependent fashion.Hsp70s contain two domains—a 44-kDa N-terminal nucleotide-binding domain (NBD) and a 15-kDa C-terminal substrate-bind-ing domain (SBD)—connected by a highly conserved hydropho-bic linker. The allosteric cycle of Hsp70s involves an alternationbetween the ATP-bound state with low affinity and fast exchangerates for substrates, and the ADP-bound state with high affinityand low exchange rates for substrates. In turn, substrate bindingto the SBD results in about 10-fold stimulation of ATPase activityof the NBD. However, the same ATPase stimulation can beachieved for the isolated NBD in the presence of the conservedinterdomain linker sequence motif (389VLLL392) (2–4), indicat-ing that the linker plays a key role in NBD function and allostery.

The Hsp70 NBD belongs to the Actin/Hexokinase/Hsp70superfamily, members of which share a number of commonfeatures (5, 6). The NBD is composed of two lobes, I and II; eachlobe consists, in turn, of two subdomains: IA and IB for lobe I,and IIA and IIB for lobe II. Nucleotide binds at the bottom of thedeep central cleft at the interface between subdomains IB andIIB, and all four subdomains are involved in nucleotide coordi-nation. It has been suggested that nucleotide-dependent confor-mational changes due to subdomain reorientations are an

intrinsic property of all NBDs, crucial for their functions (6).Significant subdomain movements have been directly observedbetween different X-ray structures of hexokinase (6); by contrast,there are no significant conformational changes seen in X-raystructures of Hsp70 NBDs (2, 7) and actin (8). Nonetheless,NMR data (3, 9, 10) and molecular dynamics (MD) calculations(11) for the Hsp70 NBD demonstrated significant conforma-tional flexibility related to subdomain reorientations, which sug-gests that allosteric coupling in Hsp70s NBD occurs via a networkof NBD subdomain motions (10). Despite the progress to date,experimental characterization of the nature of these conforma-tional changes for Hsp70 NBDs and a mechanism by whichthese subdomain motions are exploited during the allosteric cycleremain elusive.

Some insight into the nature of the conformational landscapeof Hsp70 NBDs comes from X-ray structures of their complexeswith nucleotide exchange factors (NEFs), which show an openingof the nucleotide-binding pocket relative to other NBD structuresconformation (12–15). Additionally, the recently determinedX-ray structure of ATP-bound yeast Hsp110, Sse1, a distant re-lative of the Hsp70s, emerges as a plausible model for the Hsp70ATP-bound conformation. It shows the interdomain linker andSBD directly interacting with the NBD (16), a likely structuralroute by which the linker directly influences conformationalchanges in the Hsp70 NBD. The three structures (i.e., NEF-freeand NEF-bound Hsp70 NBD, and Hsp110 NBD) retain similarstructural organization within NBD subdomains yet have differ-ent relative subdomain orientations. Together they serve as anexcellent starting point for modeling the Hsp70 NBD allostericensemble.

To explore the allosteric landscape of the Hsp70 NBD, wehave examined the relationship between allosteric modulatorsand conformational changes in the NBD of the Escherichia coliHsp70, DnaK. Global chemical-shift analysis of six differentstates, apo or bound to alternative ligands, of two NBD con-structs—one with and one without the conserved 389VLLL392

linker sequence—reveals that the NBD exists as an ensemble ofseveral conformations interconverting via rigid-body subdomainmotions. Residues located on subdomain interfaces constitutean allosteric network in the NBD, providing multiple pathwaysfor signal transduction. Pairwise comparisons of chemical-shiftperturbations in turn reveal how nucleotide binding coordinatessubdomain rotations, and consequently enables transmission of

Author contributions: A.Z. and L.M.G. designed research; A.Z. performed research; A.Z.and L.M.G. analyzed data; and A.Z. and L.M.G. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: 1HN, 15N, 13CO, 13Cα, and 13Cβ chemical shifts for ADP.NBD388 ,apo.NBD388 , and ATP.NBD392 have been deposited in the BioMagResBank, www.bmrb.wisc.edu (accession nos. 17208, 17209, and 17210).1To whom correspondence should be addressed. E-mail: gierasch@biochem.umass.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1014448108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1014448108 PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6987–6992

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0

signals through the NBD, providing communication with theSBD and Hsp70 cochaperones.

ResultsChemical Shifts as Signposts on the Hsp70 NBD Allosteric Landscape.The two powerful NMR measurements used most often tomonitor protein structural rearrangements—nuclear Overhausereffects and residual dipolar couplings (RDCs)—may not be sen-sitive in all cases to small to moderate domain reorientations thatoccur in large proteins with little change in secondary structure,both because experimental data are too sparse and because mea-surement errors may be comparable to or larger than the overallstructural perturbations. By contrast, chemical-shift perturba-tions, which are also easier to obtain, are extremely accurate andexquisitely sensitive to protein local and global structure. Thus,chemical shifts are excellent reporters to monitor even smallchanges in a protein conformational ensemble (17–19). To iden-tify residues affected by conformational changes in the isolated44-kDa NBD of the E. coliHsp70, DnaK, we compared chemicalshifts for backbone amide 1HN, 15N, carbonyl 13C, 13Cα, and 13Cβatoms in several functionally important states, including thenucleotide-free, ADP-, ADP.Pi(phosphate)-, and ATP-boundstates (Table S1). Nucleotide-induced conformational changes inDnaK and other Hsp70s require the correct positioning of theMg2þ and Kþ ions in the nucleotide-binding site. Moreover, thepresence of ATP is essential: Functional conformational changesdo not occur with several ATP analogs (e.g., ATPγS, AMP-PNP)(1, 20, 21). To understand how these small perturbations in thenucleotide-binding site affect the global NBD conformation, weincluded in our analysis the ADP.noMg (in the absence of Mg2þions) and the ATPγS-bound states. To obtain data for the ATP-bound state over longer experimental times without interferencefrom ATP hydrolysis, we incorporated the T199A mutation,which blocks ATP hydrolysis but still allows functional ATP bind-ing and ATP-induced conformational changes (22).

As noted above, previous biochemical studies revealed theimportance of the interdomain linker for NBD conformationalchanges (2–4). In addition to these data, we found that thepresence of the 389VLLL392 linker motif changes the energeticsof nucleotide binding to the NBD. NMR titrations of the ADP-bound NBD with ATP (Fig. S1) revealed that binding affinitiesfor ATP and ADP.Pi differ significantly for the 1–388 and 1–392NBD constructs. By NMR, the shorter construct has a higheraffinity for ADP than ATP, which agrees with Kd values of 170and 610 nM for ADP and ATP binding, respectively, as measuredby isothermal calorimetry (23). On the contrary, when the

389VLLL392 motif is present on the isolated NBD, ATP bindsmore tightly than ADP, mimicking behavior of the full-lengthprotein with Kd values of 250 and 160 nM for ADP and ATP,respectively (23). Moreover, NMR evidence gathered using theNBD392 construct (residues 1–392) pointed to a coupling of theeffects of the linker and the nucleotide on NBD conformation(3). To understand the role of the linker 389VLLL392 motif, che-mical shifts of NBD were analyzed for two NBD constructs:NBD388 (residues 1–388) and NBD392 (residues 1–392, includingthe conserved linker 389VLLL392 motif).

Backbone amide 1HN, 15N, carbonyl 13C, 13Cα, and 13Cβassignments for the six different states of NBD388 and NBD392

(a total of 12 NBD states) were obtained for about 85% of allresidues. Their chemical shifts were compared pairwise betweenindividual states and as a group.

NBD Nucleotide Cleft Opens by Rotation of Subdomain IIB. To explorethe structural features that enable the NBD to bind and releasenucleotide, as required in its allosteric cycle, we analyzed chemi-cal-shift perturbations between the nucleotide-free and ADP-bound states of the NBD; the data are presented as a functionof residue position in Fig. 1A. Previous RDC analysis demon-strated that in solution the ADP-bound NBD exists in the“closed” conformation (10). X-ray structures of Hsp70 NBDsbound to their respective NEFs (12–15, 24), which enhance therate of nucleotide release and binding, show subdomain IIB to berotated about 10 to 30° with respect to the rest of the NBD andthus opening of the nucleotide-binding cleft (Fig. 2A). For mostresidues in our analysis, chemical-shift changes between the apoand ADP-bound states were small and moderate (not exceeding0.3 ppm) (Fig. 1A). As expected, large perturbations (shown inred in Fig. 1A and, on the structure, in Fig. 2B) were observedfor residues directly affected by nucleotide binding (i.e., thoselocated in the nucleotide-binding site) and, notably, for someresidues relatively distant from the nucleotide-binding site—inparticular the two α-helices of subdomain IIB (Fig. 2B, green).These α-helices are located at the interface between subdomainsIB and IIB, and consequently would experience large environ-mental perturbations upon opening of the nucleotide-bindingcleft, as seen in the crystal structures of NEF-bound Hsp70 or70–kDa heat shock cognate proteins (Hsc70s) (Fig. 2A). Thus,our results are consistent with a model where nucleotide disso-ciation favors rotation of subdomain IIB and opening of thenucleotide-binding cleft. Interestingly, the rest of subdomain IIB(i.e., the double-stranded β-sheet and the loop connecting thetwo strands, which are distant from the IB–IIB interface) shows

0.0

0.2

0.4

0.6

0.8

1.0

20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380Residue number

B20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380

0.0

0.2

0.4

0.6

0.8

1.0 IA IAIB IIA IIAIIBX X LN N N N N N N N N L

∆δ

(pp

m)

tot

A IA IAIB IIA IIAIIBX X LN N N N N N N Napo vs. ADP

ADP vs. ATP

N

Fig. 1. Nucleotide-induced conformational changes in the DnaK NBD. (A and B) Histograms showing chemical-shift differences,Δδtot ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðΔδHÞ2 þ ð0.154ΔδNÞ2 þ ð0.341ΔδCOÞ2

p, for backbone atoms as a function of residue number, where ΔδH, ΔδN, or ΔδCO are 1HN, 15N, and 13CO che-

mical-shift differences between the apo and ADP-bound states of NBD388 (A), and between the ADP- and ATP-bound states of NBD392 (B). Residues with largeΔδtot (>0.3 ppm) and significant chemical-shift perturbation (at least one ΔδH, ΔδN, or ΔδCO value is larger than two corresponding chemical-shift errors; i.e.,0.06, 0.6, and 0.6 ppm for 1HN and 15N, and 13CO atoms, respectively) are colored red and yellow, respectively; the rest are shown as cyan. The green back-ground highlights regions that are highly affected by nucleotide binding, and the top bar shows NBD subdomains: IA (dark green), IB (light green), IIA (darkblue), IIB (light blue), crossing α-helices (red, X), the 389VLLL392 linker motif (yellow, L), and the nucleotide-binding site (black, N).

6988 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014448108 Zhuravleva and Gierasch

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0

almost no chemical-shift perturbation but displays enhancedmobility on the nanosecond timescale, as indicated by the highpeak intensities for this region (Fig. S2A). This finding fits re-markably well with elastic network modeling (ENM) predictionsshowing that the same region enjoys the largest mobility andserves as an adjustable NEF recognition path in the NBD (25).These ENM calculations and previous MD simulations (11) pre-dicted that the open conformation is a state present on the Hsp70NBD conformational landscape and not induced by NEF binding.Our chemical-shift perturbation data are entirely consistent withthis prediction.

ATP and the Interdomain linker Cooperatively Bind the NBD. Toexplore the role of the interdomain linker in Hsp70 allosteryand understand how the linker and the nucleotide-binding sitecommunicate with each other, we examined how the presenceof the 389VLLL392 linker motif affects the NBD conformationalensemble by comparing nucleotide-free, ADP-, and ATP-boundstates between the two NBD constructs: NBD392 and NBD388

(i.e., with and without the 389VLLL392 motif).The last five residues arising from the C-terminus of NBD388

display narrow intense peaks in NMR spectra, and their chemicalshifts and peak intensities are identical whether the NBD isbound to ATP or ADP, or no nucleotide is bound (Fig. S2A). Thishigh peak-intensity suggests that these residues are flexible on thenanosecond timescale. We conclude that the NBD388 C-terminusis a flexible, water-exposed segment and that its behavior is not

linked to nucleotide-dependent conformational changes of theNBD. In striking contrast, the C-terminal residues of NBD392

experience chemical-shift and mobility changes upon even smallchanges in the nucleotide-binding site (Fig. 3 A and B). In nucleo-tide-free NBD392, the C-terminus is highly flexible, and in turn,minimal long-range chemical-shift perturbations were observedwhen comparing this state to nucleotide-free NBD388 (Fig. 3C).In the ADP- and ATP-bound states, peaks from the C-terminalresidues shift and broaden (Fig. 3 A and B), and moreover,the pairwise comparison between NBD388 and NBD392 showssignificant effects in the rest of the NBD (Fig. 3C), indicatingthat there is communication between the linker and the NBD.Previously, it was proposed that a solvent-accessible hydrophobiccleft, which is formed by the crossing helices on the interfacebetween subdomain IA and IIA, is the most probable binding sitefor the linker (3). Indeed, we found HN-HN NOESY cross-peaksbetween the linker and residues from the hydrophobic cleft in theATP-bound state, but not in apo.NBD (Fig. S3A). Upon ADPbinding, the 389VLLL392 linker motif causes chemical-shift per-turbations in the hydrophobic cleft (Fig. S4). These appear toarise from weak binding of the linker to this cleft, as the changesare small and local. Strikingly, ATP binding dramatically en-hances the linker effect on the NBD conformational ensemble,causing chemical-shift changes throughout the NBD (Fig. 3C).Our data show that in addition to direct effects of linker bindingon the hydrophobic cleft, the entire interface between the NBDlobes is perturbed, suggesting allosteric linker-induced lobereorientation. In turn, in the ATP-bound state the conformationof the C-terminal segment dramatically changes compared to theflexible solvent-exposed conformation it adopts in the nucleotide-free state.

It was previously reported that NBD393, a construct that in-cludes one more residue (D393), had even higher ATPase thanNBD392 (4). In our chemical-shift analysis, the addition of D393to the linker motif (i.e., in the NBD393 construct) results in smalllong-range perturbations around the nucleotide-binding site andin subdomain IIA (Fig. S5B). However, the overall structure ofthe ATP-bound state of NBD392 and NBD393 remains very simi-lar. Only a few residues showed significant chemical-shift differ-ences. We suggest that these small perturbations fine-tune theATPase activity and illustrate how sensitive the catalytic effi-ciency is to small perturbations.

Taken together, our observations support a two-way couplingmechanism between binding of ATP and the interdomain linker;we conclude that cooperative binding of the 389VLLL392 linkersequence and ATP results in an NBD conformation favorableto ATPase hydrolysis and ATP binding and poised for interdo-main allostery.

IIB

IIA

IA

IBIIB

IIA

IA

IB

open

closedA B

Fig. 2. Opening of the nucleotide-binding cleft upon nucleotide dissocia-tion. (A) Comparison of alternative conformations of subdomain IIB in X-raystructures of DnaK homologues: in green, the closed form as seen in theisolated Bos taurus Hsc70 NBD [Protein Data Bank (PDB) ID code 1KAX]; inyellow, the open form as seen in the complex of yeast Sse1 with the Bostaurus Hsc70 NBD (PDB ID code 3C7N); in red, the open form as seen inthe complex of yeast Sse1 with human Hsp70 NBD (PDB ID code 3D2F).(B) Mapping of the chemical-shift differences from Fig. 1А onto the structureof the DnaK NBD (PDB ID code 1DKG:D). Residues with large chemical-shiftperturbations (red) and the highly affected subdomain IIB α-helices (green)are colored as in Fig. 1А.

H (ppm)

N (

ppm

)

8.2 8.0 7.8

130

131 noM

g

ADP

Pi

ATP

apo

1

15

ATPATPγS

ADP

noM

g

apo

1

2

3

4

5

Intre

l

ATPγS

A

B

C IA IAIB IIA IIAIIBX X LN N N N N N N N N

0.0

0.2

0.4

0.6

0.8

0.2

0.2

∆δ

(pp

m)

tot

Residue number280 300 320 340 360 38020 40 60 80 100 120 140 160 180 200 220 240 260

apo

ADP

ATP

IA IAIB IIA IIAIIBX X LN N N N N N N N N

Fig. 3. The cooperative effect of linker and ATP binding. (A) Relative peak intensities (Intrel) of the C-terminal residue for Leu392 for different NBD392 states.The Intrel value is the ratio of the Leu392 peak height to an average peak height in an HNCO spectrum in a corresponding NBD state. (B) Blow-up of the regionof the 15N TROSY spectra showing resonances corresponding to the C-terminal Leu392 for different NBD392 states. (C) Histograms showing combined chemical-shift differences (Δδtot) as a function of residue number for backbone 1HN and 15N, and 13CO atoms (as in Fig. 1) between NBD388 and NBD392 in the apo,ADP-, and ATP-bound states. Yellow and cyan colors highlight significant and insignificant chemical-shift perturbations (as defined in Fig. 1), respectively. Graybackground highlights the interfaces between two lobes. The top bar is the same as for Fig. 1.

Zhuravleva and Gierasch PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6989

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0

Binding of ATP and Linker Leads to Widespread NBD Structural Reor-ganization. A previous RDC analysis of the AMP-PNP–boundstate of a construct of Thermus thermophilus DnaK NBD lackingthe interdomain linker revealed reorientations of the subdomainsIA and IIA relative to each other, such that the hydrophobic cleftbecame more accessible (9). Our pairwise chemical-shift pertur-bation analysis of the ADP- and ATP-bound NBD388 constructalso showed long-range perturbations of the hydrophobic cleft.Thus, binding of ATP shifts the NBD to a conformation favorablefor linker binding.

The consequence of the ATP-induced conformational changein the NBD and resulting linker binding is clearly illustrated byour pairwise chemical-shift perturbation analysis of ADP- andATP-bound NBD392. In this longer construct, which retains a keystructural participant in the allosteric function of Hsp70s, the in-terfaces between NBD subdomains experience much more exten-sive perturbations between the ADP- and ATP-bound states thanseen for NBD388. The chemical-shift perturbations (shown as afunction of residue number in Fig. 1B), when mapped onto thestructure (Fig. 4A), revealed the largest changes for the C-termi-nus, the IIA β-sheet, and the N-terminal crossing α-helix.

To interpret our chemical-shift data structurally, and in sodoing describe conformational changes, we generated two modelstructures representing the ADP- and ATP-bound state of theNBD (SI Text). For the ATP-bound conformation (ATP:NBDM),we built a homology model of the DnaK NBD based on thecrystal structure of its distant relative, yeast Hsp110 (Sse1) boundto ATP, which has been suggested to represent the ATP-boundstate of Hsp70 (16). The ADP-bound conformation of DnaK(ADP:NBDM) was obtained from the X-ray structure of theNBD of its close homologue, the Bos taurus Hsc70 NBD (26).Whereas the structure within the NBD subdomains for the twomodels (ADP:NBDM and ATP:NBDM) is quite similar, lobe IIis rotated significantly with respect to lobe I, and subdomainsIIA and IIB rotate relative to each other (Fig. S6). The resultingrotation of the IA–IIA subdomains against one another exposesbinding sites for the interdomain linker along the edge of the IIAβ-sheet (Fig. 4B). Moreover, these structural changes correlatewell with shear movements of the crossing α-helices relative toeach other and to the core IA and IIA β-sheet that have beendirectly observed for another Hsp70 homolog, hexokinase (6).Consequently, the modeled NBD structures predict that thelinker and IIA β-sheet should experience the largest structuralperturbations. This prediction is fully confirmed by our experi-mental data, which indeed reveal the largest chemical-shift

changes for these regions (Fig. 4A). In addition, large chemicalshifts for the N-terminal crossing α-helix hint at its possible rolein domain reorientations, as was predicted from hexokinase struc-tures (6).

In order to obtain more detailed structural information aboutthe ATP-induced NBD conformational changes and specificallythe linker conformation, we measured NH RDCs in the apoand ATPγS-bound states of the NBD392 (Fig. S3C). However, theprecision of our RDC data did not enable accurate and unambig-uous characterization of structural changes between these twostates (see SI Text). Nonetheless, RDCs confirmed that the linkerundergoes dramatic changes in its conformation upon ATP bind-ing. In the apo state, RDC values for the linker are close tozero, which agrees with its flexible, solvent-exposed conforma-tion, suggested by chemical-shift analysis. On the contrary, in theATPγS-bound state [which mimics ATP-NBD (Fig. S5A)], thelinker displays large positive RDCs, indicative of significantlymore structure. Consistent with this observation, Cα and Cβ che-mical shifts of the linker show significant β-sheet propensity(Fig. S2B). Moreover, RDC values indicate that NHs of the linkerbecame parallel to those in the IIA β-sheet (Fig. S3C), whichagrees well with the homology model derived from the Hsp110structure (Fig. 4B). We believe these results provide a usefulmodel for the structure of the linker, in general, but nonethelesswe caution that the detailed structure of the linker seen forNBD392 may not recapitulate its structure when attached to therest of the SBD.

An Intramolecular Allosteric Network in the NBD. In order to gaina holistic picture of the NBD conformational ensemble and inso doing explore the parts of this molecular machine that respondto ligands, we compared chemical shifts of the six different statesfor both the NBD388 and NBD392 constructs and compiled thegreatest shifts, site-by-site. We found that the secondary structureof the protein is largely unchanged: The Cα and Cβ chemicalshifts do not vary significantly between different NBD states(Fig. S7); in all cases, these shifts are consistent with the second-ary structure derived from the DnaK NBD X-ray structure(Fig. S2B). By contrast, the backbone amide (1HN and 15N) andcarbonyl 13C chemical shifts of most residues are modulated bynucleotide binding and/or by the presence of the linker (Fig. 5).Even small changes in the nucleotide-binding site (e.g., betweenADP-, ADP.Pi-, and ADP.noMg-bound states), caused long-rangeperturbations (Fig. S8). We found that about 60% of the NBDresidues showed chemical-shift perturbations of backbone amide(1HN and 15N) and carbonyl 13C atoms when all 12 differentNBD states were compared; i.e., for these residues chemical shiftswere significantly different between at least two states (Fig. 5B).These residues represent “hot spots” in the NBD structure, asthey respond to the binding of ligands or presence of the linker.Not surprisingly, the majority of these hot spots map to the inter-faces between NBD subdomains (Fig. 6A), whereas the rest of theprotein is largely unperturbed. These results fully agree with amodel previously proposed based on analysis of RDCs, whichshowed significant NBD subdomain reorientations in solutioncompared to X-ray structure and suggested that allosteric cou-pling in the NBD occurs via a network of subdomain motions(10). Importantly, our chemical-shift analysis explores NBD con-formational space more extensively, using NBD388 and NBD392 insix different states, all of which can be explained by an ensembleof conformations with similar overall architecture, the same foldswithin subdomains, but with different subdomain orientations. Aquestion remains whether nucleotide binding results in a newstructure on the NBD energy landscape, or rather shifts a preex-isting equilibrium between two or more already populated NBDconformations. Although most NBD residues were representedby single resonances in NMR spectra, significant broadeningof a number of peaks (Fig. S2A) suggests that local or global con-

IIB

IIA

IA

IBIIB

IIA

IA

IB

linker on/off

ATP

A B

Fig. 4. The linker binds to the hydrophobic cleft between the subdomains IAand IIA. (A) Mapping of the chemical-shift differences from Fig. 1B onto thehomology model of the DnaK ATP-bound NBD structure built from the Sse1structure (PDB ID code 2QXL:B) (Fig. S6). The regions highlighted in Fig. 1B bygreen background are shown in green on the structure. Residues with largechemical-shift perturbations (highlighted in red in Fig. 1B) are shown as redspheres. (B) Superposition of the ADP- (green) and ATP- (yellow) bound con-formations, derived from homology models as described in the text. For theATP-bound conformation, the C-terminal 12 residues are shown in red, andthe interdomain linker (389VLLL392) bound to the hydrophobic cleft is shownas red spheres.

6990 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014448108 Zhuravleva and Gierasch

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0

formational exchange processes are occurring at a rate that is in-termediate on the NMR timescale. Moreover, for many residuesdifferent chemical shifts were observed for all 12 states examined(Fig. 5A). It is very unlikely that a unique “single structure” existsfor each state; instead, the single resonances observed most likelyrepresent averages of multiple conformations rapidly inter-converting on the NMR timescale. The “peak walking” patternobserved (Fig. 5A) shows that most resonances do not move ina simple manner in response to ligands, and thus suggests thatligands adjust the populations of more than two NBD conforma-tions. The conformational heterogeneity for the Hsp70 NBDensemble implied by the chemical-shift behavior is consistentwith computational predictions (11) and previous experiments(3, 9, 10, 27). These observations also account for previous resultsshowing that changes in the nucleotide-binding site, such as singlepoint mutations or the absence of inorganic ions, cause disruptionof Hsp70 allostery without significant changes in structure byX-ray crystallography (1, 20, 21, 26, 28, 29).

DiscussionOur results provide detailed molecular insights into intramolecu-lar signal transduction in Hsp70 molecular chaperones. Our che-mical-shift data lead us to conclude that binding of a nucleotideligand and the docking of the interdomain linker result in changesin subdomain and lobe orientations in the Hsp70 NBD, providinglong-range perturbations along lobe and subdomain interfaces(Fig. 6A). Importantly, by interacting with all four subdomains,nucleotide binding selects preferred features on the protein land-scape and consequently coordinates further allosteric events. Anallosteric signal can propagate bidirectionally. Therefore, notonly do perturbations in the nucleotide-binding site affect subdo-main interfaces, but also a signal can propagate from subdomain

interfaces to the nucleotide-binding site, and in so doing regulatenucleotide binding and ATPase activity. Our results are entirelyconsistent with a previous hypothesis that allosteric signal trans-duction occurs via a network of motions of protein modules (e.g.,subdomains) (30, 31). According to this model, residues at theinterfaces between modules form an allosteric network betweenprotein active sites. More specifically, our results argue thatnucleotide dissociation favors rotation of subdomain IIB andopening of the nucleotide-binding cleft, which had been pre-dicted early by ENM (25) and MD (11) calculations. Addition-ally, NEF binding, which accelerates nucleotide exchange,appears to act by selection of preexisting conformations on thelandscape (25).

Because our study included a comparison of NBD constructswith and without the functionally critical first segment of theinterdomain linker, we can provide a structural model for two-way coupling pathway between binding of ATP and the linker,which helps understanding of NBD to SBD communication. Viasubdomain reorientations, ATP binding results in exposure of abinding surface for the interdomain linker. Linker binding, inturn, stabilizes the ATPase-active conformation of the NBD,which is similar to that seen in the Hsp110 NBD crystal structure(16). As a result, the β-sheet of subdomain IIA structurallyconnects the interdomain linker and the nucleotide-binding site(Fig. 6B). Indeed, linker binding to an edge of the IIA β-sheetresults in rotation of subdomain IIA relative to subdomain IAand the crossing α-helices, and consequently changes the orienta-tion of the β1–β2 turn in subdomain IIA that is responsible forcoordination of the ATP γ-phosphate (6). Consequently, thepresence of the ATP γ-phosphate and the linker is essential forthe ATPase active conformation. The activation of the ATPaseactivity upon extension of the construct [e.g., in NBD393 (4)] isevidence that additional structural adjustments can occur andenable further regulation of ATPase activity by Hsp70 cochaper-ones and the SBD.

Thus, we offer an explanation for how the interdomain linkerworks as a switch: By binding to the hydrophobic cleft in a nu-cleotide-dependent fashion, the linker directly affects the NBDconformation and regulates interdomain communication, bring-ing the SBD close to the NBD. Indeed, single point mutationsof the linker and hydrophobic cleft residues perturb the NBDATPase activity and disturb its regulation by the SBD and cocha-perones (2, 16, 32, 33). The fact that the hydrophobic cleft is thebinding site for the Hsp70 cochaperone DnaJ, which regulatesthe rate of ATP hydrolysis (2, 34, 35), indicates that cochaperoneregulation of ATPase activity most likely occurs through the sameallosteric networks (i.e., via the linker and the hydrophobic cleft).Intriguingly, the hydrophobic cleft comprises a sector of coe-volved residues in the NBD, which is important for stabilizingthe interdomain interfaces and mediating allosteric communica-tion between the NBD and SBD (36).

Evolution has capitalized on the ability of the Actin/Hexoki-nase/Hsp70 family NBD structure to utilize ATP as a signal for

B

Residue number

E310

L312

L227A58

F146A258

I140

K122

7.5 7.0

121

125

123

127

H (ppm)1

N (

ppm

)15

A

280 300 320 340 360 38020 40 60 80 100 120 140 160 180 200 220 240 2600.0

0.2

0.4

0.6

0.8

1.0

∆δ

(pp

m)

tot

IA IAIB IIA IIAIIBX X LN N N N N N N N N

Fig. 5. An intramolecular allosteric network in the NBD. (A) Representative regions from overlaid 1H-15N TROSY spectra of the 12 NBD states. Several examplesof nonoverlapping resonances for residues with significant conformational changes are labeled to highlight their peak-walking patterns. The color codefor the 12 spectra is: red (ATP.NBD392), magenta (ATPγS.NBD392), green (ADP∕ADP:Pi∕ADP:noMg:NBD392), light blue (apo:NBD392), orange (ATP:NBD388),yellow (ATPγS:NBD388), light green (ADP∕ADP:Pi∕ADP:noMg:NBD388), blue (apo:NBD388). (B) Histograms showing chemical-shift differences, Δδtot ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ðΔδHÞ2 þ ð0.154ΔδNÞ2 þ ð0.341ΔδCOÞ2p

, for backbone atoms as a function of residue number; where ΔδH, ΔδN, or ΔδCO are the largest 1HN, 15N, and 13COchemical-shift differences between any 2 of the 12 states of NBD388 and NBD392. Coloring and the top bar are the same as for Fig. 1.

IIB

IIAIA

IB

linker

γP

IIA

linker

αX1αX2

ATP

IA

A B

Fig. 6. Mechanism of intramolecular allostery in the Hsp70 NBD. (A) Map-ping of allosteric hot spots onto the structure of the DnaK NBD (PDB ID code1DKG:D): Residues with large and significant chemical-shift differences fromFig. 5B are shown in red and yellow, respectively. Cyan and gray indicate in-significant changes and residues with no data, respectively. (B) Structuralmodel for two-way coupling pathway between the nucleotide-binding siteand the interdomain linker: superposition of the IIA β-sheet and the crossingα-helices for the ATP- (yellow) and ADP- (green) bound conformations of theNBD. The ATP γ-phosphate and the linker are in red, and residues involved innucleotide binding are shown as spheres.

Zhuravleva and Gierasch PNAS ∣ April 26, 2011 ∣ vol. 108 ∣ no. 17 ∣ 6991

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0

downstream functions, and the present study using chemical-shift perturbation has shed light on the structural origins ofthis conserved allosteric machine. Chemical-shift perturbationanalysis has proven itself to be a rich source of information in acomplex conformational landscape with shifting populations.

Materials and MethodsProtein Expression and Purification. Plasmids encoding DnaK NBD(1–388)T199A, NBD(1–392)T199A, and NBD(1–393)T199A were prepared previouslyin our laboratory; they were expressed in E. coli and purified as described(3, 37). 2H-, 13C-, and 15N-labeled NMR samples were prepared using estab-lished protocols (38), as described previously (37) (see SI Text).

NMR Spectroscopy. To obtain backbone and Cβ resonance assignments, werecorded standard sets of transverse-relaxation optimized spectroscopy(TROSY)-modified triple-resonance experiments, developed for 15N-, 13C-,and 2H-labeled proteins. Details about NMR experiments and sample condi-tions are summarized in Table S1. All spectra were obtained at 26 °C on a600-MHz Bruker Avance spectrometer using a TXI cryoprobe, processedusing nmrPipe (39), and analyzed using Cara (40).

Chemical-Shift Analysis. Backbone 1HN, 15N, 13CO , 13Cα, and 13Cβ chemicalshifts for individual states were obtained with the program AutoAssign(41) and/or manually (see SI Text).

The 1HN∕15N∕13CO combined chemical-shift change of a particularresidue between different conformations was calculated as described pre-viously (42):

Δδtot ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðΔδHÞ2 þ ð0.154ΔδNÞ2 þ ð0.341ΔδCOÞ2

q;

where 0.154 and 0.341 are the weighting factors for 15N and 13CO, respec-tively. Errors in chemical-shift values, 0.03, 0.3, 0.3, 0.3, and 0.5 ppm for 1HN,15N, 13CO, 13Cα, and Cβ atoms, respectively, were obtained as average line-widths for 1H dimensions and spectral resolution in corresponding 3D spectraand 2D 15N-TROSY for 13C and 15N dimensions, respectively.

ACKNOWLEDGMENTS. This work was supported by National Institutes ofHealth Grant GM027616.

1. Mayer MP, Brehmer D, Gassler CS, Bukau B (2001) Hsp70 chaperone machines. AdvProtein Chem 59:1–44.

2. Jiang JW, et al. (2007) Structural basis of J cochaperone binding and regulation ofHsp70. Mol Cell 28:422–433.

3. Swain JF, et al. (2007) Hsp70 chaperone ligands control domain association via anallosteric mechanism mediated by the interdomain linker. Mol Cell 26:27–39.

4. Vogel M,MayerMP, Bukau B (2006) Allosteric regulation of Hsp70 chaperones involvesa conserved interdomain linker. J Biol Chem 281:38705–38711.

5. Bork P, Sander C, Valencia A (1992) An ATPase domain common to prokaryotic cellcycle proteins, sugar kinases, actin, and hsp70 heat shock proteins. Proc Natl AcadSci USA 89:7290–7294.

6. Hurley JH (1996) The sugar kinase heat shock protein 70 actin super family: Implica-tions of conserved structure for mechanism. Annu Rev Biophys Biomol Struct25:137–162.

7. Flaherty KM, Wilbanks SM, DeLuca-Flaherty C, McKay DB (1994) Structural basis of the70-kilodalton heat-shock cognate protein ATP hydrolytic activity. II. Structure of theactive-site with ADP or ATP bound to wild-type and mutant ATPase fragment. J BiolChem 269:12899–12907.

8. Reisler E, Egelman EH (2007) Actin structure and function: What we still do notunderstand. J Biol Chem 282:36133–36137.

9. Bhattacharya A, et al. (2009) Allostery in Hsp70 chaperones is transduced bysubdomain rotations. J Mol Biol 388:475–490.

10. Zhang YB, Zuiderweg ERP (2004) The 70-kDa heat shock protein chaperone nucleo-tide-binding domain in solution unveiled as a molecular machine that can reorientits functional subdomains. Proc Natl Acad Sci USA 101:10272–10277.

11. Woo HJ, Jiang JW, Lafer EM, Sousa R (2009) ATP-induced conformational changesin Hsp70: Molecular dynamics and experimental validation of an in silico predictedconformation. Biochemistry 48:11470–11477.

12. Arakawa A, et al. (2010) The C-terminal BAG domain of BAG5 induces conformationalchanges of the Hsp70 nucleotide-binding domain for ADP-ATP exchange. Structure18:309–319.

13. Polier S, Dragovic Z, Hartl FU, Bracher A (2008) Structural basis for the cooperation ofHsp70 and Hsp110 chaperones in protein folding. Cell 133:1068–1079.

14. Schuermann JP, et al. (2008) Structure of the Hsp110 : Hsc70 nucleotide exchangemachine. Mol Cell 31:232–243.

15. Sondermann H, et al. (2001) Structure of a Bag/Hsc70 complex: Convergent functionalevolution of Hsp70 nucleotide exchange factors. Science 291:1553–1557.

16. Liu QL, Hendrickson WA (2007) Insights into Hsp70 chaperone activity from a crystalstructure of the yeast Hsp110 Sse1. Cell 131:106–120.

17. Cavalli A, Salvatella X, Dobson CM, Vendruscolo M (2007) Protein structure determi-nation from NMR chemical shifts. Proc Natl Acad Sci USA 104:9615–9620.

18. Mulder FAA, Filatov M (2010) NMR chemical shift data and ab initio shielding calcula-tions: Emerging tools for protein structure determination. Chem Soc Rev 39:578–590.

19. Shen Y, et al. (2008) Consistent blind protein structure generation from NMR chemicalshift data. Proc Natl Acad Sci USA 105:4685–4690.

20. Bukau B, Horwich AL (1998) The Hsp70 and Hsp60 chaperone machines. Cell92:351–366.

21. Palleros DR, Reid KL, Shi L, Welch WJ, Fink AL (1993) ATP-induced protein Hsp70complex dissociation requires Kþ but not ATP hydrolysis. Nature 365:664–666.

22. McCarty JS, Walker GC (1991) DnaK as a thermometer: Threonine-199 is site of autop-hosphorylation and is critical for ATPase activity. Proc Natl Acad Sci USA 88:9513–9517.

23. Taneva SG, Moro F, Velazquez-Campoy A, Muga A (2010) Energetics of nucleotide-Induced DnaK conformational states. Biochemistry 49:1338–1345.

24. Harrison CJ, HayerHartl M, DiLiberto M, Hartl FU, Kuriyan J (1997) Crystal structure ofthe nucleotide exchange factor GrpE bound to the ATPase domain of the molecularchaperone DnaK. Science 276:431–435.

25. Liu Y, Gierasch LM, Bahar I (2010) Role of the Hsp70 ATPase domain intrinsic dynamicsand sequence evolution in enabling its function interactions with NEFs. PLoS ComputBiol 6:1–15.

26. O’BrienMC, Flaherty KM,McKay DB (1996) Lysine 71 of the chaperone protein Hsc70 isessential for ATP hydrolysis. J Biol Chem 271:15874–15878.

27. Mapa K, et al. (2010) The conformational dynamics of the mitochondrial Hsp70chaperone. Mol Cell 38:89–100.

28. Johnson ER, McKay DB (1999) Mapping the role of active site residues for transducingan ATP-induced conformational change in the bovine 70-kDa heat shock cognateprotein. Biochemistry 38:10823–10830.

29. Sousa MC, McKay DB (1998) The hydroxyl of threonine 13 of the bovine 70-kDa heatshock cognate protein is essential for transducing the ATP-induced conformationalchange. Biochemistry 37:15392–15399.

30. Daily MD, Gray JJ (2009) Allosteric communication occurs via networks of tertiary andquaternary motions in proteins. PloS Comput Biol 5:1–14.

31. del Sol A, Tsai CJ, Ma BY, Nussinov R (2009) The origin of allosteric functional modula-tion: Multiple pre-existing pathways. Structure 17:1042–1050.

32. Davis JE, Voisine C, Craig EA (1999) Intragenic suppressors of Hsp70 mutants: Interplaybetween the ATPase- and peptide-binding domains. Proc Natl Acad Sci USA96:9269–9276.

33. Vogel M, Bukau B, Mayer MP (2006) Allosteric regulation of Hsp70 chaperones by aproline switch. Mol Cell 21:359–367.

34. Gassler CS, et al. (1998) Mutations in the DnaK chaperone affecting interaction withthe DnaJ cochaperone. Proc Natl Acad Sci USA 95:15229–15234.

35. Suh WC, et al. (1998) Interaction of the Hsp70 molecular chaperone, DnaK, with itscochaperone DnaJ. Proc Natl Acad Sci USA 95:15223–15228.

36. Smock RG, et al. (2010) An interdomain sector mediating allostery in Hsp70 molecularchaperones. Mol Syst Biol 6:414.

37. Clerico EM, Zhuravleva A, Smock RG, Gierasch LM (2010) Segmental isotopic labelingof the Hsp70 molecular chaperone DnaK using expressed protein ligation. Biopoly-mers 94:742–752.

38. Tugarinov V, Kanelis V, Kay LE (2006) Isotope labeling strategies for the study ofhigh-molecular-weight proteins by solution NMR spectroscopy. Nat Protoc 1:749–754.

39. Delaglio F, et al. (1995) NMRpipe—A multidimensional spectral processing systembased on Unix pipes. J Biomol NMR 6:277–293.

40. Keller R (2004) Optimizing the process of nuclear magnetic resonance spectrumanalysis and computer aided resonance assignment.. (Eidgenössiche TechnischeHochschule, Zurich, Switzerland) Doctoral thesis.

41. Zimmerman DE, et al. (1997) Automated analysis of protein NMR assignments usingmethods from artificial intelligence. J Mol Biol 269:592–610.

42. Evenas J, et al. (2001) Ligand-induced structural changes to maltodextrin-bindingprotein as studied by solution NMR spectroscopy. J Mol Biol 309:961–974.

6992 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1014448108 Zhuravleva and Gierasch

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 8,

202

0