Structural and functional analysis of the Hsp70/Hsp40 chaperone system

2.1. Domain organization and essential allosteric coupling

The sequences and structures of Hsp70s are highly conserved across all species examined.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 All Hsp70s contain two functional domains: a nucleotide‐binding domain (NBD) at the N‐terminus and a substrate‐binding domain (SBD) at the C‐terminus (Figure 1a), corresponding to the two essential intrinsic activities. Connecting these two functional domains is a short liner, the interdomain linker. The NBD is about 44 kDa, and it is the energy component. It is this domain that binds and hydrolyzes ATP to ADP to provide energy to power the chaperone activity. The SBD is the site to which polypeptide substrates bind so that Hsp70s can help them fold, assemble, transport across membranes, or degrade. In general, Hsp70s recognize short stretches of polypeptides of 5–7 residues in length with no sequence specificity but with a strong preference for hydrophobic residues, especially aliphatic amino acids, and residues with positive charges.32, 33, 34, 35, 36, 37, 38

Although each domain can bind their respective substrates tightly, the chaperone activity absolutely requires the functional coupling between these two domains upon ATP binding, the ATP‐induced allosteric coupling.1, 8, 39, 40, 41, 42, 43, 44, 45, 46 The nucleotide‐bound states of Hsp70s regulate the polypeptide‐substrate binding affinity and kinetics. In the absence of ATP, Hsp70s are in either the ADP‐bound form or the nucleotide‐free form (apo form). In these forms, there is little contact between the two domains.37, 39, 40, 43, 47, 48, 49, 50 The substrate‐binding properties behave as if the domains are isolated, characterized by relatively high affinity and slow kinetics.39, 40, 47, 49 In sharp contrast, upon ATP binding, extensive contacts form between the domains and these intimate contacts modulate both domains' activities. ATP binding to NBD reduces the peptide‐substrate binding affinity of the SBD by 2–3 orders of magnitudes through drastically accelerating the binding and release rates, especially for release.32, 51 At the same time, peptide‐substrate binding to SBD significantly speeds up the ATP hydrolysis rate of the NBD.5, 32 This ATP‐induced allosteric coupling ensures that the energy from ATP hydrolysis is efficiently used to power the chaperone activity. Recent advances in structural and biophysical studies have revolutionized our understanding of this allosteric coupling and the molecular mechanism of Hsp70 chaperone activity.8, 44, 52, 53, 54, 55, 56, 57 We will discuss these advances in detail below.

2.2. NBD structures

The first Hsp70 structure to be determined was the crystal structure of the isolated NBD from bovine Hsc70 in complex with ADP, published in 1990.58 Now, more than a dozen structures of the isolated NBD from a number of Hsp70s have been published or deposited to the Protein Data Bank (PDB).59, 60, 61, 62, 63, 64, 65, 66 Although bound with various nucleotides including ATP, all these structures are virtually identical in conformation. They all represent the ADP‐bound state regardless of what nucleotide is bound, or what ATP‐hydrolysis deficiency mutation is present. This suggests that the ADP‐bound conformation is much more stable for the isolated NBD than the active ATP‐bound conformation, and the NBD‐SBD contacts are required to maintain the active ATP‐bound conformation. As shown in Figure 1b, the NBD is composed of two large lobes: I and II. Between the two lobes is a deep cleft where the nucleotide binds. Each lobe is further divided into two subdomains: A and B. The bound nucleotide forms extensive contacts with all four subdomains, thus stabilizing this unique ADP‐bound conformation for the NBD.

NEFs are a class of well‐established cochaperones for Hsp70s. They were proposed to facilitate the chaperone activity by speeding up the exchange of ADP for ATP after ATP hydrolysis. A number of structural studies on the complexes of NEFs with isolated NBDs suggested that NEFs stabilize the nucleotide‐free state of NBD by prying open the nucleotide‐binding pocket.26, 27, 67, 68, 69, 70 The conformation of the nucleotide‐free state and mechanism of nucleotide exchange have been reviewed previously.8, 26, 27, 71

2.3. SBD structures

The SBD structure is represented by the crystal structure of the isolated SBD from DnaK, a major Hsp70 in E.coli.36 Since its identification as an Hsp70, DnaK has been widely used as a model Hsp70. This isolated DnaK SBD structure is the first SBD structure from an Hsp70. As shown in Figure 1c, the SBD is composed of two subdomains: SBDβ and SBDα. In this structure, a model peptide‐substrate NR (sequence NRLLLTG) was cocrystalized and bound to a narrow pocket formed in SBDβ. The SBDβ consists of eight β strands organized into two β sheets. The peptide‐binding pocket is formed between two loops, L_1,2 and L_3,4, and two β strands, β3 and β4. The SBDα covers the peptide‐binding pocket as a lid. All the following NMR and crystal structures on the isolated SBD revealed essentially the same conformation regardless of the presence or absence of peptide substrates.30, 36, 49, 52, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81 Except for one NMR study, all these structures have peptide substrates bound in the peptide‐binding pocket. Even the NMR study on SBD without peptide substrate reported virtually identical conformation but with more conformational dynamics.75 The bound peptide substrates were either added peptides, or the interdomain linker mimicking the peptide substrate. The bound peptide substrates could be the key determining factor for these crystallographic and NMR studies. Like the full‐length Hsp70s, the isolated SBDs tend to form poorly ordered oligomers.82, 83, 84, 85, 86, 87 Different from the functional dimers for Hsp70s,88, 89, 90 it is generally believed that these oligomers are formed through the binding of the interdomain linker to the peptide‐binding pocket of another molecule.17, 30, 49 This oligomerization may have some functional importance,91 but has presented challenges for structural studies.

2.4. Structures of full‐length Hsp70s in the ADP‐bound and nucleotide‐free states

After the publication of the isolated domain structures, the structures of full‐length Hsp70s, especially in the ATP‐bound state, the allosteric active state, were highly sought to understand the molecular mechanism of Hsp70 chaperone activity. However, this turned out to be challenging. Various biochemical and NMR studies have suggested that in the ADP‐bound and nucleotide‐free state, there is little contact between the two domains and each domain behaves as if isolated.37, 39, 40, 43, 47, 48, 49, 50 A NMR study on DnaK generated a positioning of the two functional domains in the ADP‐bound state (Figure 2a), providing a general picture of the overall conformation of this state.47 This positioning supports the idea that the two domains are relatively independent and there is no single fixed overall conformation for the ADP‐bound state. Two crystal structures of Hsp70s in complex with ADP have provided further support for this conformation.30, 49 The nucleotide‐free state structure of the bovine Hsc70 is the first crystal structure of a close‐ to full‐length Hsp70.73 In this structure, some small contacts between NBD and SBDα are revealed (Figure 2e); however, these contacts have been considered artifactual. One important feature for all these full‐length or close‐ to full‐length structures is that the SBD conformation is essentially identical to that of the isolated SBD. Although peptide substrates were not included in the crystallization conditions, either the interdomain linker or the truncated and unfolded C‐terminal of SBDα is bound to the peptide‐binding pocket, mimicking peptide‐substrate binding. In summary, these structural analyses are consistent with the biochemical studies and revealed a single conformation for the peptide‐binding pocket: closed on the bound peptide substrate (Figures 1c and 3a).

2.5. Structures of full‐length Hsp70s in the ATP‐bound state and the mechanism of allosteric coupling

In combination with recent NMR and FRET analysis, a number of structures of full‐length Hsp70s in the ATP‐bound state have provided unprecedented insights into allosteric coupling and have revolutionized our understanding of the molecular mechanism of Hsp70 chaperone activity.

Obtaining a crystal structure of a full‐length Hsp70 in the ATP‐bound state was challenging. There were three main hurdles to overcome: (a) the ATP‐bound state is short lived due to constant ATP hydrolysis; (b) the interaction between the two domains are too transient to be captured by structural analysis; and (c) purified Hsp70 proteins tend to form low‐ordered, flexible oligomers, which is incompatible with structural analysis. Moreover, all nonhydrolyzable ATP analogs tested so far have failed to induce allosteric coupling in Hsp70s in a way that ATP does. Stabilizing a functional ATP‐bound state that is compatible with structural analysis is the key.

The first full‐length Hsp70‐ATP structure determined was for Sse1 (Figure 2c).56 Sse1 is an Hsp110 chaperone from yeast. Hsp110s are distant homologs of Hsp70s.92, 93, 94 Sse1 has the advantage for crystallographic study that it does not have the aforementioned hurdles. This Sse1‐ATP structure revealed extensive domain contacts. More importantly, the domain contacts and relative orientation of the domains agree with many biochemical and NMR results on classic Hsp70s. One striking feature of this structure is the position of SBDα: it peels away from interacting with SBDβ in the isolated SBD structure and now docks on the side of NBD (compare Figure 2a,c). Without the SBDα lid covering, the peptide‐binding pocket is more accessible than the isolated SBD structure. This conformation is consistent with the fast kinetics and low affinity in the ATP‐bound state. However, this SBDα conformation alone is not sufficient to explain the drastically accelerated kinetics and reduced affinity in the ATP‐bound state since deletion of the SBDα only results in moderate decrease in affinity and increase in kinetics.75 Conformational changes in SBDβ, especially around the peptide‐binding pocket, must be essential for the allosteric coupling. However, the relatively low sequence conservation, especially in the SBDβ region, has limited the applicability of this Sse1‐ATP structure to Hsp70s.

Inspired by this Sse1‐ATP structure, two groups have published the first two intact Hsp70‐ATP structure using E. coli Hsp70 DnaK.53, 54 These two structures both took advantage of an ATPase‐deficient mutant T199A in the NBD.95 This mutation significantly reduced the ATPase activity but maintained allosteric coupling.40, 95 However, the T199A mutation alone is not sufficient for crystallization. The two groups used different strategies. The DnaK‐ATP structure solved by our group took advantage of a SBD mutation, L_3,4′, which mimics the Sse1 sequence.96 The DnaK‐ATP structure solved by Mayer's group introduced a disulfide bond between NBD and SBDα based on the Sse1‐ATP structure, thus stabilizing the ATP‐bound state. Despite this difference, these two structures are almost identical. The overall conformations of these two DnaK‐ATP structures including the NBD conformation and location of the interdomain linker are highly similar to that of the Sse1‐ATP structure (Figure 2d), supporting the homology between Hsp70s and Hsp110s. Compared to the isolated SBD structures, which represent the ADP‐bound and nucleotide‐free states, there are two major conformational changes, which are dramatic. The first one is the position of SBDα, which was previously revealed by the Sse1‐ATP structure. The second major change is with the peptide‐binding pocket. In the isolated SBD structure, the two peptide‐binding loops close onto the bound peptide with favorable and extensive contacts (closed conformation, Figure 3a). In contrast, the L_3,4 side of the peptide‐binding pocket, including L_3,4 and its supporting loop L_5,6, opens up in both the DnaK‐ATP structures although differing in the extend of opening. In the structure solved by our group, the L_3,4 side is flipped wide open (open conformation, Figure 3b); whereas in Mayer's structure, the L_3,4 side is only partially open (partially open confirmation, Figure 3c). Besides these loops, the bottom of the peptide‐binding pocket, formed by β3 and β4, is also flipped open at the L_3,4 side. This occurs for both structures. Moreover, the L_3,4 side in the DnaK‐ATP structure solved by Mayer's group has high B‐factors (constantly above 100 Ų, and some atoms are even over 200 Ų), and L_5,6 in two of the four protomers are missing, suggesting high flexibility in these regions. Thus, the peptide‐binding pocket of the ATP‐bound state is dynamic and the two open conformations are most likely representatives of the conformational ensembles of the ATP‐bound state.

Based on the DnaK‐ATP structure solved by Mayer's group, structures of two specialized Hsp70s, Ssb1 and Ssz1, have been solved.57, 97 Our group solved a structure of human BiP, the major Hsp70 in ER, using an analogous construct as in the DnaK‐ATP structure.52 Almost identical conformation was observed for the peptide‐binding pocket for all these structures except for Ssz1 which has a nonclassical peptide‐binding pocket.

The open conformations of the peptide‐binding pocket are consistent with the fast kinetics and reduced affinity in the ATP‐bound state. However, with only these open conformations, it is hard to explain how Hsp70s decide when to bind and when to release substrates for productive chaperone cycles. Is binding and releasing substrate random or tightly controlled? Moreover, both the DnaK‐ATP structures depend on the ATPase‐deficient T199A mutation. The T199A mutation not only knocks out chaperone activity but also completely abolishes Hsp40 interaction.95, 98 It is still unclear which step of the ATP‐bound state that this T199A mutation represents. Interestingly, a new structure of human BiP, BiP‐ATP2, with a different conformation was solved in which the peptide‐binding pocket is completely closed (Figure 3d).55 Different from all the other published Hsp70‐ATP structures, this BiP‐ATP2 structure has a completely wild‐type (WT) NBD, that is, it does not carry the T199A mutation, although it contains the L_3,4′ mutation. Thus, it most likely represents a different step in the ATP‐bound state. Solution studies further suggested that this fully closed conformation is the major conformation in solution for the ATP‐bound state. This conformation is incompatible with binding substrate. As the dominant conformation in solution for the ATP‐bound state, this conformation supports an active release of bound substrate upon ATP binding.

In summary, up to now, no WT Hsp70 structure in complex with ATP has been solved except the Sse1‐ATP structure. These structural studies revealed three conformations of the peptide‐binding pocket for the ATP‐bound state: open, partially open, and fully closed (Figure 3). A transition state from the ATP‐bound to the ADP‐bound has been proposed by an NMR study44 and is waiting for high‐resolution structural analysis.

2.6. Implications in chaperone activity and chaperone cycles

The classic paradigm for the Hsp70 chaperone cycle had evolved mainly based on essential allosteric coupling. In a simplified view, allosteric coupling is about how different nucleotide‐bound states control peptide‐substrate binding. Due to the fast kinetics in the ATP‐bound state, peptide substrate was proposed to bind in this state to start the chaperone cycle. Once the substrate binds, ATP hydrolysis is stimulated. With the high affinity and slow kinetics for peptide substrates, the ADP‐bound state holds the bound peptide for a while to prevent misfolding. Upon nucleotide exchange, Hsp70 reverts back to the ATP‐bound state. Due to the fast release rate, the bound peptide is released to fold on its own. The Hsp70‐ATP is available to bind other peptide substrates to restart a new chaperone cycle.

In combination with recent NMR and FRET studies, these new full‐length Hsp70‐ATP structures and their associated biochemical analysis have provided support for a more active chaperone cycle with tightly controlled substrate binding and release (Figure 4). The open conformations are for binding substrates to start chaperone cycles while the fully closed conformation is responsible for releasing bound substrates at the end of chaperone cycles after ATP rebinding. Moreover, Hsp40s were revealed to be the key to convert the fully closed conformation to the open conformations. In the ATP‐bound state, Hsp70 is mainly in the full‐closed conformation, and incompatible with substrate binding. Hsp40s not only bring substrates to Hsp70s but also convert Hsp70s to the open conformations to initial substrate binding. Once substrate binds, both substrate and Hsp40s stimulate the ATPase activity drastically. Now, Hsp70s are in ADP‐bound state, and Hsp40s are dissociated. Once ATP rebinds, Hsp70s revert back to the fully closed conformation, and the substrate is squeezed out of Hsp70s. This active release of bound substrate and Hsp40 regulation gives the chaperone cycle direction instead of random oscillation between the two nucleotide‐bound states.

3.1. Classification and functional diversity of Hsp40s

As a key class of cochaperones for Hsp70s, Hsp40s are more diverse than Hsp70s in both sequence and function. The conserved biochemical activity for all Hsp40s is to specifically interact with Hsp70‐ATP and stimulate ATP hydrolysis by the Hsp70.16, 19, 20, 23, 24, 99 Some Hsp40s can bind polypeptide substrates directly and show chaperone activity in preventing aggregation.17, 100, 101 Since the substrate binding preference of Hsp40s is different from that of Hsp70s, it was proposed that Hsp40s also serve as a substrate scanning factor for Hsp70s and bring polypeptide substrates to Hsp70s.102 Although Hsp40s are less conserved than Hsp70s, all share a conserved J‐domain named after the well‐studied E.coli Hsp40, DnaJ. The signature J‐domain is about 70 amino acids in length and has been shown to be critical for stimulating the ATPase activity of the Hsp70s.103, 104

Based on domain organizations, Hsp40s are classified into three classes: I, II, and III (Figure 5a). The class I and class II Hsp40s are relatively more conserved than the class III Hsp40 in both structure and function. This review mainly focuses on the first two classes. DnaJ, the founding member of class I, represents the domain organization of class I. Class I has five domains: J‐domain, glycine/phenylalanine rich (G/F) domain, carboxyl‐terminal domain 1 (CTD1) which contains a zinc finger‐like region (ZFLR), carboxyl‐terminal domain 2 (CTD2), and dimerization (D) domain. 16, 19, 20, 22 Class II has the same domain organization as DnaJ but with a CTD1 that does not contain ZFLR. CTD1 and CTD2 in both class I and class II Hsp40s are conserved in structure and are involved in binding polypeptide substrates. The role of ZFLR is still unclear although it seems essential for chaperone activity.105, 106 Class III is much more divergent with only the J‐domain conserved. Unlike the N‐terminal location in class I and class II Hsp40s, the J‐domain in class III Hsp40s can be in any location.

The sequence and function diversity of Hsp40s was proposed to dictate the chaperone activity of Hsp70s and enable the diverse functions of Hsp70s.16, 19, 20, 23, 24, 99 Usually, there are more Hsp40 members than those of Hsp70s in any organism or a given cellular compartment. Various studies have suggested that multiple Hsp40s work with a single Hsp70 to specify the chaperone activity of the Hsp70. One of the well‐established roles of class I Hsp40s is assisting Hsp70s in general protein folding. Consistent with the sequence and domain diversity, many class III Hsp40s have specialized roles in assisting Hsp70s in proteostasis. A beautiful example is BiP, the single classic Hsp70 in ER, and its seven known Hsp40 cochaperones, ERdjs.25, 99, 107, 108 These ERdjs enable BiP to perform diverse functions in the ER: ERdj1 and ERdj2 are for protein import into ER,109, 110, 111 ERdj3 and ERdj6 are for protein folding,112, 113, 114, 115 and ERdj4 and ERdj5 are for protein degradation116 whereas the function of ERdj7 is unclear.99, 107 Moreover, a recent study suggested that class I and class II Hsp40s form complexes to assist Hsp70s in efficiently dismantling protein disaggregation.117 It is possible that different Hsp40s regulate the chaperone cycle of Hsp70s differently in order to achieve their diverse functions besides recognizing different substrates.

3.2. Isolated domain structures of Hsp40s

The NMR structures of the isolated J‐domains from two Hsp40s (E. coli Hsp40 DnaJ and human Hsp40 Hdj1) provided the first structural information for Hsp40s.118, 119, 120 These two Hsp40s belong to class I and class II Hsp40s, respectively. Following these structures, a number of crystal and NMR structures from all three classes were solved, and all revealed essentially the same fold for the J‐domain.121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131 The J‐domain is composed of four α‐helices: I to IV (Figure 5b). Between the helices II and III is the highly conserved HPD (histidine–proline–aspartate) motif. The HPD motif is essential for interacting with Hsp70s and stimulating the ATPase activity of Hsp70s.

The CTD including CTD1 and CTD2 is relatively conserved in class I and class II Hsp40s except for the presence of ZFLD in CTD1 of class I Hsp40s. The structure of CTD is represented by the crystal structure of the isolated CTD from Ydj1, a class I Hsp40 from yeast (Figure 5c).132 Except for the ZFLD, CTD1 and CTD2 have a similar fold. Each is made of two layers of β sheets. The CTD1 and CTD2 from class II hsp40 share a similar fold.128, 133, 134, 135 Interestingly, a peptide substrate was cocrystallized with the CTD of Ydj1 and revealed the structural basis of peptide binding. The peptide substrate (sequence: GWLYEIS) is bound to CTD1 by adding an antiparallel β‐strand to one of the two layers of the β‐sheets. The central Leu fits in a hydrophobic pocket formed on a conserved surface of CTD1. This Leu forms the most extensive and significant contacts with Ydj1, providing an explanation for the substrate preference of Hsp40s for hydrophobic residues. However, the shallow binding site for the peptide substrate raises concerns for the functional meaning. Thus, the precise mechanism of how class I and class II Hsp40s bind substrates is still under debate.

3.3. Full‐length structures of Hsp40s

At the present time, there is no high‐resolution full‐length structure known for either class I or class II Hsp40s. This could be due to the high conformational dynamics of these multidomain Hsp40s.22, 128 Both class I and class II Hsp40s are known to form dimers16, 20, 21, 23, 24, 136 although a recent study suggested that ERdj3, a class I Hsp40 in ER, functions as a tetramer.137 A structural model of full‐length DnaJ in dimeric form from T. thermophilus was built based on EPR and crystallographic results (Figure 5d).128 This DnaJ belongs to class II Hsp40s. The overall structure is a V‐shaped dimer with a large cleft formed between the two monomers. The two monomers dimerize through the extreme C‐terminal D domains. The J‐domains are at the end of each stalk formed by the two CTDs. Full‐length structures are available for a number of class III Hsp40s that facilitate the understanding of their unique chaperone activities.121, 133