The nucleotide‐bound/substrate‐bound conformation of the Mycoplasma genitalium DnaK chaperone

Abstract

Hsp70 chaperones keep protein homeostasis facilitating the response of organisms to changes in external and internal conditions. Hsp70s have two domains—nucleotide binding domain (NBD) and substrate binding domain (SBD)—connected by a conserved hydrophobic linker. Functioning of Hsp70s depend on tightly regulated cycles of ATP hydrolysis allosterically coupled, often together with cochaperones, to the binding/release of peptide substrates. Here we describe the crystal structure of the Mycoplasma genitalium DnaK (MgDnaK) protein, an Hsp70 homolog, in the noncompact, nucleotide‐bound/substrate‐bound conformation. The MgDnaK structure resembles the one from the thermophilic eubacteria DnaK trapped in the same state. However, in MgDnaK the NBD and SBD domains remain close to each other despite the lack of direct interaction between them and with the linker contacting the two subdomains of SBD. These observations suggest that the structures might represent an intermediate of the protein where the conserved linker binds to the SBD to favor the noncompact state of the protein by stabilizing the SBDβ‐SBDα subdomains interaction, promoting the capacity of the protein to sample different conformations, which is critical for proper functioning of the molecular chaperone allosteric mechanism.

Comparison of the solved structures indicates that the NBD remains essentially invariant in presence or absence of nucleotide.

Short abstract

http://imolecules3d.wiley.com/imolecules3d/review/vOYMnEAmHM3p6wxYgulJNC3xv0P4Zo7iUujMANFeHj0yq36fAFGzP4asF8kjFW9E825/1620, http://imolecules3d.wiley.com/imolecules3d/review/vOYMnEAmHM3p6wxYgulJNC3xv0P4Zo7iUujMANFeHj0yq36fAFGzP4asF8kjFW9E825/1621

Abbreviations

NBD

nucleotide binding domain

PMF

peptide mass fingerprinting

SBD

substrate binding domain.

Introduction

The Heat shock protein 70 kDa (Hsp70) family of molecular chaperones are present in all domains of life and expressed in response to a variety of cellular stresses, having roles in protein translocation, protein degradation, and assembly and disassembly of protein complexes among others.1, 2 Besides, in unstressed cells, Hsp70s can interact with members of other families of molecular chaperones, such as Hsp40, to promote and regulate protein folding.3, 4 Structurally, members of the Hsp70 family share the same domain architecture (Fig. S1, Supporting Information): a N‐terminal nucleotide binding domain (NBD), where ATP is hydrolyzed, connected by a highly conserved linker to a C‐terminal substrate binding domain (SBD), where unfolded peptide substrates or protein substrates bind. NBD has a U‐shape given by two lobules (I and II) with two subdomains (A and B) each. SBD is composed by two subdomains: a β‐sandwich (SBDβ) and a bundle of α‐helices known as lid (SBDα). Due to the emerging role of Hsp70s in several diseases there is an increasing need to understand its mechanism of action at the molecular level.5

DnaK, a member of the Hsp70 family of proteins in prokaryotes, has been intensively studied and its functioning is now known to imply cycles of binding and release of peptide substrates promoted by allosteric coupling driven by the binding and hydrolysis of ATP.5, 6

The apo and nucleotide‐bound states of a number of DnaK‐NBD crystal structures have been reported as well as several DnaK‐SBD structures.7, 8 These studies revealed that each domain can bind their substrates separately, though the linkage of the two domains is required for the allosteric functioning of DnaK. NBD and SBD appear to behave mostly independently from each other,9, 10 in ATP‐free and/or substrate‐bound DnaK states. Instead, in the crystal structure of the yeast Sse1:ATP complex (Sse1 belongs to the Hsp110 heterogeneous family of abundant molecular chaperones related to Hsp70), the first determined corresponding to a DnaK ATP‐bound/substrate‐free state, SBD interacts closely with NBD, which also forms a network of interactions with the linker, giving a compact structure.11 In agreement with this, more recent reports on the structure of the intact Escherichia coli DnaK showed that upon ATP‐binding, the affinity of SBD for peptide substrates is reduced and the close interaction between NBD and SBD occurs.12, 13

The studies here described are centered on two constructs of the Mycoplasma genitalium DnaK (MgDnaK) protein in complex with different nucleotides or in their absence. All structures correspond to the noncompact, nucleotide‐/substrate‐bound state of the protein. In this particular case, NBD and SBD do not interact directly with each other although the hydrophobic linker interacts with SBD subdomains β and α. The existence of such interaction had been proposed by NMR studies performed on EcDnaK14 and also observed on the EcDnaK:ADP structure.10 Furthermore, comparison of structures of MgDnaK and of the isolated MgDnaK‐NBD confirms that the domain remains essentially invariant in presence or absence of nucleotide.

Results and Discussion

The structure of MgDnaKΔCt in complex with ADP·Pi or AMP‐PNP

The structure of the MgDnaKΔCt construct, which lacks the protein C‐terminus after the first helix of SBDα (Fig. S1, Supporting Information and Materials and Methods section), has been determined from crystals obtained in the presence of either ATP or AMP‐PNP. MgDnaK is composed, similarly to other DnaK proteins, of the NBD and SBD domains, which are interconnected by a linker with a sequence (D₃₆₉‐VLLLDVT‐P₃₇₇) that is in agreement with the consensus sequence: D/E‐V/I/L‐V/L‐L‐L‐D‐V‐*‐P.15 The structure of the NBD from MgDnaKΔCt corresponds to a typical N‐terminal ∼45 kDa actin‐like NBD and the SBD is also similar to the known structures of other DnaKs [Fig. 1(A), S1, Supporting Information], with the SBDβ formed by a β‐sandwich containing the peptide substrate binding pocket, which is exclusively composed by the hydrophobic residues Ile382, Phe407, Val419, Ile453, and Ile455. The two MgDnaKΔCt structures determined were very similar with a rmsd for the superposition of Cα atoms of only 0.32 Å. Both structures span from Asn5 to His527 although residues for the short loop connecting the SBD subdomains β and α (residues 484–488), were disordered. Even despite the truncation of SBDα, the first α‐helix of the subdomain that remain in the MgDnaKΔCt construct interacts with SBDβ, similarly to what is observed in intact SBD domains.7, 8 In the MgDnaKΔCt structures the substrate binding site is occupied by the extended C‐tail of the construct that is mimicking a peptide substrate [Fig. 1(B, C)]. In the compact (ATP‐bound/substrate‐free) state of DnaK, SBDα interacts extensively with NBD.12, 13 Therefore, truncation of SBDα in the MgDnaKΔCt construct was expected to destabilize the compact state enhancing the formation of noncompact states such as the one determined. Truncation of SBDα also appears to facilitate the formation of a noncompact, substrate‐bound state, by placing the C‐tail of the MgDnaKΔCt construct, including a hexa‐histidine tag, in a position that allows its binding to the substrate binding site. In these nucleotide‐/substrate‐bound MgDnaKΔCt structures, NBD and SBD do not interact directly with each other, although both domains remain close even despite the lack of interactions of the linker with symmetry‐related molecules within the crystal (Fig. S2, Supporting Information). In turn, SBD subdomains β and α establish some interactions with the linker [Fig. 1(B,D)], namely a hydrogen bond between the carbonyl oxygen of the linker residue Asp374 with the backbone amide of Thr398 from SDBβ and hydrophobic interactions between side chains of Leu371 and Leu373 from the linker and residues Pro377 and Leu378 from SDBβ and Leu488 and Ile493 from SBDα. An ADP molecule and an inorganic phosphate were found in the crystals prepared with ATP, indicating that the added ATP was fully hydrolyzed when bound to MgDnaKΔCt (accordingly, this structure will be referred to as MgDnaKΔCt:ADP·Pi hereafter). Instead, an AMPPNP molecule and a magnesium ion were clearly defined bound to MgDnaKΔCt in the structure from crystals prepared with AMPPNP.

Figure 1.

Overall structure of the M. genitalium DnaK protein. (A) Stereoview of the NDB and SBD domains represented as ribbons of MgDnaKΔCt connected by a highly conserved hydrophobic linker. The structure is depicted in rainbow coloring and the magnesium ion is represented as a red sphere. (B) View of the SBD domain 90° apart from the view on (A) with the C‐terminal tail shown in green. (C) Close‐up view of the substrate binding pocket where the construct C‐terminal tail is bound, mimicking a peptide substrate. The hydrophobic residues forming the substrate binding pocket are shown as sticks and follow the rainbow coloring on (A). The residues from the C‐terminal tail are shown as green atom‐colored sticks. (D) Stereoview of the linker region of MgDnaKΔCt:ADP·Pi (green ribbon) showing its interaction with residues from the SBD subdomains β and α. Residues from the linker and from subdomains β and α are depicted in stick mode and colored yellow, red and orange, respectively. http://imolecules3d.wiley.com/imolecules3d/review/vOYMnEAmHM3p6wxYgulJNC3xv0P4Zo7iUujMANFeHj0yq36fAFGzP4asF8kjFW9E825/1620

Comparison of the ADP‐bound/substrate‐bound structures MgDnaKΔCt and GkDnaKΔCt

The MgDnaKΔCt:ADP·Pi and GkDnaKΔCt:ADP·Pi16 crystal structures correspond both to DnaKs with truncated SBDαs. Superposition of individual domains indicates a high structural similarity with r.m.s.d. of 0.83 and 0.97 Å for NBDs and SBDs, respectively [Fig. 2(A,B)]. Binding of the nucleotide to NBD is identical in both structures with the ADP molecule and an inorganic phosphate clearly defined. Binding of an extended peptide into the substrate binding pocket of SBD is also similar in both structures. However, the hydrophobic linkers, the bound peptides and the relative positions of domains are remarkably different [Fig. 2(A)]. The linker of MgDnaKΔCt:ADP·Pi interacts, as indicated before, with some residues from the SBD subdomains β and α [Fig. 1(D,A)], while in GkDnaKΔCt:ADP·Pi the linker is fully extended [Fig. 2(A)] interacting with the peptide binding pocket of a neighbor molecule in the crystal. Moreover, in the MgDnaKΔCt:ADP·Pi structure NBD and SBD remain close to each other and with the hydrophobic linker not participating in crystal contacts (Fig. S2, Supporting Information). Therefore, the structures trapped in the MgDnaKΔCt crystals correspond to a noncompact, nucleotide‐bound/substrate‐bound conformation of DnaK but might be a representative of an intermediate state of the protein where the conserved linker binds to the SBD suggesting that this interaction favors the noncompact state of the protein by stabilizing the SBDβ‐SBDα subdomains interaction.

Figure 2.

Structural comparison of MgDnaKΔCt (blue ribbon) with the ADP‐bound/substrate‐bound GkDnaKΔCt structure (green ribbon, PDB ID 2V7Y). Superposition of the NDB domains (A) and the SBD domains (B) of MgDnaKΔCt:ADP·Pi and GkDnaKΔCt:ADP·Pi. http://imolecules3d.wiley.com/imolecules3d/review/vOYMnEAmHM3p6wxYgulJNC3xv0P4Zo7iUujMANFeHj0yq36fAFGzP4asF8kjFW9E825/1621

Crystal structure of the MgDnaK‐NBD domain in absence or presence of nucleotide

Given the impossibility of obtaining crystals of MgDnaKΔCt in the absence of nucleotides, the crystal structure of the isolated MgDnaK‐NBD domain was determined both in its apo form and in complex with ATP or AMP‐PNP [Fig. 3(A–C)]. Again, as in the MgDnaKΔCt structures an ADP molecule and an inorganic phosphate were found in the crystals prepared with ATP, indicating that the added ATP was fully hydrolyzed. A magnesium ion was clearly identified bridging the phosphate β and the inorganic phosphate [Fig. 3(B)]. The nucleotide binds to NBD with the adenosine and the deoxyribose moieties placed at the interface of subdomains IIA–IIB and with nucleotide phosphates α and β interacting with residues from subdomain IIA, while the inorganic phosphate interacts mainly with subdomain IA. In apoMgDnaK‐NBD two sulfate ions occupy the sites corresponding to the nucleotide phosphate β and the inorganic phosphate [Fig. 3(A)]. Superposition of the three NBD structures gives an average rmsd of only 1.02 Å. Superpositions of these structures with NBD from MgDnaKΔCt:AMP‐PNP give rmsds below 0.85 Å. Therefore, the NBD domain remains essentially unchanged both in presence or absence of a bound nucleotide and also when isolated or while being part of the intact protein. On the contrary, substrate binding results in very significant structural changes in SBD. The absence of clear structural changes in NBD that could correlate with either ADP/ATP exchange or ATP hydrolysis suggests that AMPPNP does not induce conformational changes in MgDnaK similarly to what was observed for the ATP analog ATPγS in EcDnaK.17 However, protein constructs were not purified or crystallized in the presence of K⁺ ions, condition that is essential for ATP‐induced allosteric transition from the extended to the compact conformation. Therefore, nucleotide exchange in MgDnaK is likely promoted by the interaction with cochaperones, mainly with GrpE, which would be in agreement with previous proposals.18

Figure 3.

Comparison of the nucleotide binding site from the MgDnaK‐NBD structures. (A) Two sulfate ions and a glycerol molecule are found in the nucleotide binding pocket of apoMgDnaK‐NBD. (B) An ADP molecule and an inorganic phosphate together with four water molecules (red spheres) are shown octahedrally coordinating a magnesium ion (green sphere) in MgDnaK‐NBD:ADP·Pi. The nucleotide, added as ATP, was fully hydrolyzed and the magnesium ion bridges the phosphate β and the inorganic phosphate. (C) An AMP‐PNP molecule was clearly defined in the MgDnaK‐NBD:AMPPNP structure. Phosphate β coordinates with a magnesium ion (green sphere), which presents only partial occupancy. In all three panels the 2Fo‐Fc electron density corresponding to the ligands is shown contoured at 1σ. Ligands are represented as sticks following an atom‐colored code, while blue is used to color the protein.

Conclusions

The crystal structures of the Mycoplasma genitalium chaperone DnaK in complex with ADP·Pi or AMP‐PNP correspond to a noncompact, nucleotide‐bound/substrate‐bound conformation of DnaK. In this state, none of the interactions found in the compact state between NBD and SBD are maintained although the hydrophobic linker is neither fully extended nor participating in crystal packing interactions but interacting instead with the two subdomains of SBD. These observations suggest that the MgDnaKΔCt structures described here‐in can be representative of an intermediate conformation of the protein where the conserved hydrophobic linker binds to the SBD to favor the noncompact state of the protein by stabilizing the SBDβ‐SBDα subdomains interaction. This conformation emphasizes the importance of the tunability of the DnaK structure that must be sampled in order to promote signal transmission between domains and that is critical for the proper functioning of the molecular chaperone allosteric mechanism.

Materials and Methods

Cloning and protein production of different variants of MgDnaK

M. genitalium DnaK, coded by mg305, was cloned between the NdeI and XhoI (Fermentas) restriction sites of pET21d (Novagen, Madison, WI), which adds an hexa‐histidine tag at the C‐terminus of proteins, as well as the variants MgDnaK‐NBD (residue 1–366) and MgDnaKΔCt (residue 1–521). These constructs were transformed in E. coli BL21(DE3) competent cells. Generally, cell cultures were grown in LB medium supplemented with 100 µg mL⁻¹ ampicillin to an OD₆₀₀ of ∼0.6–0.8 and induced with 0.5 mM IPTG for 16 h at 20°C. Cells collected by centrifugation at 4500g for 20 min were suspended in 0.02M Tris‐HCl (pH 8.0), 0.5M NaCl, cOmplete™ EDTA‐free protease inhibitor (Roche Diagnostics, Mannheim, Germany) buffer. The suspensions were sonicated and the supernatants recovered by centrifugation at 48,000g for 20 min.

Target proteins isolation consisted in an affinity chromatography performed through a HisTrap 5 mL column (GE Healthcare Life Sciences, Uppsala, Sweden) in purification buffer (0.02M Tris‐HCl (pH8.0), 0.5M NaCl) using a gradient of imidazole to 0.5M, followed by a gel filtration performed on a Superdex 200 10/300 GL column (GE Healthcare Life Sciences, Uppsala, Sweden) equilibrated in the previously mentioned purification buffer. Proteins were then concentrated by ultrafiltration using a 10 kDa cut‐off Amicon device (Darmstadt, Germany). The major band present in a denaturing polyacrylamide gel of MgDnaK had an apparent molecular weight smaller than expected for the entire protein and its analysis by Peptide Mass Fingerprinting (PMF) revealed that the band corresponded to a truncated form of the protein. As the full‐length protein was unstable in solution two variants were produced: (i) DnaK‐NBD, corresponding to the well conserved nucleotide binding domain, and (ii) DnaKΔCt, containing NDB, SDBβ, and the first α‐helix of SBDα. The latter was designed based on the PMF data from MgDnaK.

Crystallization and X‐ray diffraction data collection

MgDnaK‐NBD and MgDnaKΔCt, taken respectively to concentrations of 11 and 10 mg mL⁻¹, were dispensed on crystallization plates prefilled with commercially available crystallization screens after 1 h incubation at 4°C with 1 mM MgCl₂ and in absence or presence of 1 mM ATP or AMPPNP. Crystals grew within 3–5 days at 20°C using the hanging drop vapor‐diffusion method by mixing equal quantities of protein and precipitant solutions. MgDnaK‐NBD and MgDnaKΔCt crystallized in 0.1M Tris‐HCl (pH 7.5–8.0), 2.2–2.6M ammonium sulfate and 0.02M potassium thiocyanate, or 0.1M Sodium citrate (pH 5.5), 28.75% (v/v) PEG MME 5000, and 5.5% (v/v) Butanol. MgDnaK‐NBD and MgDnaKΔCt crystals were flash‐cooled in liquid nitrogen in presence of 20% (v/v) glycerol for cryoprotection and X‐ray data collection was performed at 100 K respectively at BL13‐XALOC19 (ALBA synchrotron, Barcelona, Spain) and ID23eh120 (ESRF, Grenoble, France) beamlines. Data was indexed and integrated with XDS21 and SCALA22 (Table 1). MgDnaK‐NBD crystals belong to space group H3 while the MgDnaKΔCt crystals belong to P4₃2₁2, both containing one molecule per asymmetric unit.

Table 1.

X‐ray Data Processing and Refinement Statistics

	MgDnaKΔCt: AMP‐PNP	MgDnaKΔCt: ADP·Pi	MgDnaK‐NBD: AMP‐PNP	MgDnaK‐NBD: ADP·Pi	MgDnaK‐NBD apo
Data collection
Beamline	ID23eh1, ESRF	ID23eh1, ESRF	XALOC, ALBA	XALOC, ALBA	XALOC, ALBA
PDB ID	5OBU	5OBV	5OBY	5OBW	5OBX
Wavelength (Å)	0.9763	0.9763	0.9795	0.9795	0.9795
Space group	P43212	P43212	H3	H3	H3
Unit cell
a, b, c (Å)	a = b = 133.7, c = 64.1	a = b = 133.6, c = 64.5	a = b = 157.1, c = 56.8	a = b = 157.9, c = 57.5	a = b = 157.9, c = 64.7
α, β, γ (°)	α = β = γ = 90	α = β = γ = 90	α = β = 90, γ = 120	α = β = 90, γ = 120	α = β = 90, γ = 120
Resolution (Å)	47.3–2.0 (2.1–2.0)a	47.2–2.5 (2.6–2.5)	78.5–1.3 (1.4–1.3)	79.0–1.4 (1.44–1.40)	79.0–1.8 (1.83–1.78)
R merge (%)	7.0 (16.0)	5.0 (36.0)	5.0 (58.0)	5.7 (69.0)	3.0 (51.0)
R pim (%)	2.6 (5.7)	1.6 (11.9)	3.0 (36.0)	2.8 (32.2)	2.0 (36.0)
Mean I/σ(I)	5.6 (4.0)	17.2 (1.2)	12.0 (1.9)	11.3 (1.5)	16.0 (1.7)
Completeness (%)	99.6 (97.3)	99.7 (98.0)	99.5 (97.0)	99.1 (92.0)	99.4 (98.0)
No. of reflections	41 211 (5 787)	20 864 (2 917)	127 585 (18 174)	98 799 (6849)	57 432 (8 302)
Redundancy	12.0 (3.9)	7.0 (6.5)	7.6 (2.0)	5.5 (1.8)	14.5 (5.3)
Refinement
R factor/R free (%)	17.2/21.2	21.9/26.6	14.1/16.1	17.7/19.8	17.5/20.3
Average B‐factors (Å2)	27.1	53.7	23.4	23.1	20.1
Protein	26.1	53.8	23.5	22.5	18.5
Ligand/ion	24.9	43.9	30.3	32.5	63.6
Water	38.4	48.7	35.0	36.0	46.7
R.m.s. deviations
Bonds (Å)	0.0153	0.0123	0.0190	0.0182	0.0143
Angles (°)	1.7599	1.5776	2.1618	2.0730	1.7352
Ramachandran favored (%)	98.45	96.90	98.92	99.46	98.92
Ramachandran outliers (%)	0.00	0.19	0.00	0.27	0.00

^aData for highest resolution shell are shown in parenthesis.

Crystal structure determination and refinement

The crystal structure of MgDnaKΔCt:AMPPNP was determined by molecular replacement using the program MolRep23 and the NBD and SBD domains of GkDnaK (PPB ID 2V7Y)16 as search models. The structure solved to 2.0 Å resolution was refined to a R _factor/R _free of 17.2/21.1% (Table 1). The MgDnaKΔCt:ATP structure was solved to 2.5 Å resolution using Phaser24 and the MgDnaKΔCt:AMPPNP (PDB ID 5OBU) structure as search model. The structure was refined to a R _factor/R _free of 21.9/26.6%.

The structure of the NBD in presence of ATP was solved using Phaser and the NBD domain of the MgDnaKΔCt:AMPPNP (PDB ID 5OBU). The structures of the apo form of the NBD and of this domain in presence of AMPPNP (PDB IDs 5OBX and 5OBY, respectively) were solved by using MgDnaK‐NBD:ADP·Pi (PDB ID 5OBW) as search model. The programs Coot25 and Refmac526, 27 were used for model building and refinement of all the determined structures and the Molprobity server28 was used to evaluate the quality of the different structures (Table 1). To avoid overfitting of the structures, PDB_REDO29 server (https://xtal.nki.nl/PDB_REDO/) was used to determine the best relative weight for the X‐ray target function and the geometry or the B‐factor restraints for the last cycle of refinement. Figures were prepared with PyMOL (The PyMOL Molecular Graphics System, Version 1.7 Schrodinger, LLC).

Supporting information

Supporting Information

Supporting Information

Supporting Information