Large Chaperone Complexes Through the Lens of Nuclear Magnetic Resonance Spectroscopy

Abstract

Molecular chaperones are the guardians of the proteome inside the cell. Chaperones recognize and bind unfolded or misfolded substrates, thereby preventing further aggregation; promoting correct protein folding; and, in some instances, even disaggregating already formed aggregates. Chaperones perform their function by means of an array of weak protein–protein interactions that take place over a wide range of timescales and are therefore invisible to structural techniques dependent upon the availability of highly homogeneous samples. Nuclear magnetic resonance (NMR) spectroscopy, however, is ideally suited to study dynamic, rapidly interconverting conformational states and protein–protein interactions in solution, even if these involve a high-molecular-weight component. In this review, we give a brief overview of the principles used by chaperones to bind their client proteins and describe NMR methods that have emerged as valuable tools to probe chaperone–substrate and chaperone–chaperone interactions. We then focus on a few systems for which the application of these methods has greatly increased our understanding of the mechanisms underlying chaperone functions.

1. INTRODUCTION

Over the past 60 years or so, the field of protein chemistry has seen major paradigm shifts in the way that the relationship between protein structure and function is viewed. During the blossoming era of X-ray crystallography, the role of protein–protein interactions in protein functionality, or even establishing new protein roles, was underestimated (1). The one gene, one enzyme, one function dogma proposed by Beadle & Tatum (20) led the field to believe that a linear combination between genes and phenotypes is sufficient to explain nature’s complexity in terms of protein function. Protein–protein interactions were thought to be artefacts and were not extensively studied until the 1970s. However, all proteins studied by X-ray crystallography at the time appeared to be static. Even one of the first structures of a protein complex to be studied, that of hen-egg lysozyme with a trisaccharide (published in 1966), showed a remarkable alignment between the side-chains of residues in the active site and the ligand (21, 22). In opposition to the common belief at the time, Monod et al. (104) and Koshland et al. (82) proposed that proteins can adopt numerous conformations in solution, and that ligand binding shifts the equilibrium to the active state (either as a result of conformational selection or induced fit, respectively). Over the years, nuclear magnetic resonance (NMR) spectroscopy (3, 7, 19, 27–30, 75, 77, 79, 108, 109, 120, 126, 154, 169), as well as X-ray crystallography (111, 112), molecular dynamics simulations (74, 84, 160), fluorescence spectroscopy (162, 163), and hydrogen–deuterium exchange studies (34–36), has played a crucial role in showing that proteins are inherently dynamic entities, shifting Fischer’s (43) lock and key static mechanism to include protein dynamics.

Currently, protein complexes are well-established as the prime drivers of cellular multifunctionality (92). Complexes involved in functions such as enzymatic activity or protein translation are typically of high affinity (with equilibrium dissociation constants in the nM range) (76), a property that ensures recognition of the specific partner. However, protein–protein interactions involved in the chaperone network, the cellular machinery that keeps the proteome in a healthy state, are weak and transient in nature, features that pose significant barriers to their investigation by either X-ray crystallography or cryo-electron microscopy (EM). NMR spectroscopy, in contrast, has emerged as a prime tool to characterize macromolecular interactions involving chaperones and their complexes with substrates (24, 63).

In what follows, we focus on a few examples in which NMR has played a pivotal role in our understanding of chaperone function with an emphasis on the arsenal of magnetic resonance techniques that are available for complexes that either lack a specific tertiary structure or present significant challenges and opportunities posed by their large molecular weight.

Proteostasis:

the capacity of healthy cells to maintain a balanced proteome

Molecular chaperone:

a protein that interacts with a client protein and promotes its folding

Protein folding:

the process by which newly synthesized polypeptide chains fold to their functional three-dimensional structures

2. MOLECULAR CHAPERONES: A SHOWCASE OF THE POWERS OF SOLUTION NUCLEAR MAGNETIC RESONANCE

Maintaining protein homeostasis (proteostasis) is crucial for cell integrity, and proteins termed molecular chaperones play an important role in this process (12, 56, 58, 91, 138, 140). Chaperones safeguard the proteome by forming an intricate network (44) in which various protein molecules cooperate to unlock specific chaperoning functions. The latter include the promotion of folding to the native state, the prevention of misfolding, and the disaggregation or sequestration of potentially toxic oligomeric assemblies (57, 58, 99, 165, 166). It is now accepted that chaperone–substrate complexes do not necessarily represent a well-defined structure in which the two proteins form a specific set of long-lived interactions (63). Instead, interactions between chaperone molecules or chaperones and their substrates can be diffuse and of weak affinity (167). Indeed, this feature of chaperone–substrate interactions represents a crucial aspect of chaperone function since chaperones normally have to release their substrate in a timely manner (either to their final destination or to another chaperone) rather than capture the substrate permanently (17, 50). Consider, for example, a chaperone that binds misfolded substrates and promotes their folding (65, 140). It is straightforward to show that the population of the properly folded native state of the substrate strongly relies on a relatively weak affinity for the chaperone, since a tight substrate–chaperone complex would result in substrate sequestration. High-affinity, stable contacts between the substrate and the chaperone would also prevent the formation of native interactions between substrate atoms and would not allow the client protein to fold on the surface of the chaperone. Similar observations have been made for other chaperone systems (167), including those that do not promote client protein folding but instead function as holdases keeping their substrates unfolded until they reach their destination (67, 68, 124). In recent years, several studies have increased our knowledge of the kinetics and binding affinities of chaperone–substrate complexes, as summarized in Table 1.

Table 1.

Chaperone–substrate and chaperone–chaperone interactions

Interaction	KD (μM)	koff (s−1)	kon (M−1s−1)	Reference(s)
Chaperone–substrate interactions
ttHsp40–PhoA (full length)	10	10	1 × 106	68
ttHsp40–PhoA (individual sites)	NA	100	NA	68
SecB–MBP (full)	0.05	0.05	1 × 106	67
SecB–MBP (individual sites)	4–30	NA	NA	67
SecB–PhoA (full)	0.2	0.1	1 × 106	67
SecB–PhoA (individual sites)	8–40	NA	NA	67
TF–PhoA (full)	~2	~1	1 × 105	124
TF–PhoA (individual sites)	2–14	~50	5 × 106–2 × 107	124
TF–MBP	0.5	~0.05	1 × 105	124
GroEL–Aβ40	70	1,400	2 × 107	86
GroEL–Aβ42	300	880	3 × 106	159
GroEL–SH3 (I)	60	200	3.5 × 106	87
GroEL–PolyQ7	1,300	2,650	2.3 × 106	156
Spy–SH3 (native)	~100	1,000	1 × 107	167
Spy–SH3 (I)	3	40	1.3 × 107	167
Spy–Im7 (native)	21	NA	NA	61
Spy–Im7 (unfolded)	0.3	NA	NA	61
Skp–α-synuclein	1.8	0.004	2,560	23
SecB–α-synunclein	0.7	0.003	2,800	23
Hsc70–α-synuclein	1.6	0.005	3,560	23
Hsp70–hTRF1	18	20	~1 × 106	127, 128
Hsp90–p53	2	NA	NA	127, 128
Skp–Omp4	0.022	NA	NA	54
Hsp90–Tau	5	NA	NA	71
Chaperone–chaperone interactions
ttHsp40–Hsp70 tail	11	NA	NA	68
DNAJB6JD–DNAJB6CTD	NA	530	34a	72
DNAKNBD–ClpBCCD	25	NA	NA	121
Hsp27 ACD dimer	0.5	NA	NA	72
Hsp27 IXI motif	125	NA	NA	2
Hsp90–p23	1	NA	NA	6
Bip–Bap	1	NA	NA	119
CHIP–Hsc70b	0.62	NA	NA	10
CHIP–Hsp90b	1	NA	NA	10
Hop–Hsc70b	2.7	NA	NA	10
Hop–Hsp90b	5.5	NA	NA	10
DNAJC7–Hsc70b	3	NA	NA	10
DNAJC7–Hsp90b	6.8	NA	NA	10

^aIn units of s⁻¹.

^bBased on the interaction of full-length cochaperones with Hsp70/90-derived C-terminal peptides.

Abbreviations: ACD, α-crystallin domain; Bap, nucleotide exchange factor; Bip, endoplasmic Hsp70; CHIP, carboxyl terminus of Hsp70 interacting protein; Hop, Hsp70/Hsp90 organizing proteins; MBP, maltose binding protein; NA, not applicable; PhoA, alkaline phosphatase; PolyQ7, Huntingtin N terminus plus 7 glutamines; SH3 (I), folding intermediate of Fyn SH3; TF, Trigger Factor; ttHsp40, Thermus thermophilus class B Hsp40 (dimer).

Since individual binding events may be weak, chaperones need to have mechanisms in place to ensure that substrates are efficiently captured. One such strategy that is often used in chaperone–substrate interactions involves the concepts of multivalency and avidity (53). Chaperones do not simply recognize a specific epitope on their client proteins. It is common for unfolded substrates to contain multiple chaperone binding sites that are recognized by multiple sites on the chaperone surface. Each individual binding event may be of weak affinity, but the overall multivalency of the system ensures that the substrate is effectively captured by the chaperone (67, 68, 124) and only released after chaperoning is complete. This strategy [which is also used in other biological systems such as antibody–antigen interactions (129)] provides several advantages, as seen in Figure 1. A monovalent substrate binding to a monovalent chaperone relies on a high affinity interaction between the interacting partners (Figure 1a, i). This is also largely the case for the binding of a monovalent substrate to a multivalent chaperone (Figure 1a, ii) although cooperativity may also play a role in sequential binding. Binding of a multivalent substrate to a multivalent chaperone (Figure 1a, iii) effectively increases the local concentration of substrate binding sites, ensuring full occupancy of the chaperone at much lower substrate concentrations. For simplicity, in Figure 1, we schematically depict the interaction of a divalent ligand with a divalent chaperone; in reality, however, up to 15 individual binding sites (not necessarily of the same affinity) (68) have been detected, shifting the binding curve further to the left. Multivalency thus results in a fuzzy complex, in which the correct or native local secondary structure can effectively form on the surface of the chaperone, leading to an increase in the folding yield of the substrate protein. Various bacterial chaperones, such as Trigger Factor, SecB, and Skp, known to prevent misfolding of their target proteins (Table 1) represent elegant examples of this binding mechanism (25, 67, 124). To further increase avidity, some chaperones may form oligomeric assemblies. Indeed, chaperone oligomerization is a key feature of chaperones that bind aggerated substrates and inhibit their further assembly; examples include members of the Hsp40 family (72, 73) (see Section 3.1) and small heat shock proteins (HSPs) (103) (see Section 3.3).

Figure 1.

Multivalency in chaperone–substrate interactions. (a) Three different binding scenarios: (i) binding of a monovalent substrate (red) to a monovalent chaperone (gray or blue), (ii) sequential binding of a monovalent substrate to a divalent chaperone, and (iii) binding of a divalent substrate to divalent chaperone. (b) Corresponding equilibrium binding curves calculated numerically by solving the appropriate differential equations to infinite time (sequential binding of a monovalent substrate to a monovalent chaperone in blue; sequential binding of a monovalent substrate to a divalent chaperone in green; binding of a divalent substrate to a divalent chaperone in red). For the divalent substrate binding to a divalent chaperone, [L] corresponds to the local concentration of substrate following the first binding event and was set to 100 μM; f is a unitless penalty factor to describe potential steric clashes when both chaperone sites are occupied and was given a value of 5. The association (k₁) and dissociation (k₋₁) rate constants were set to 2 × 10⁷ M⁻¹s⁻¹ and 50 s⁻¹, respectively, corresponding to an equilibrium dissociation constant of 2.5 μM.

The necessity of measuring affinity in a site-specific manner immediately suggests that techniques such as isothermal titration calorimetry, surface plasmon resonance, and fluorescence anisotropy are not suitable to characterize chaperone interactions. Even if deletion constructs were to be used, these interactions may be too weak to be probed by the aforementioned methods. In contrast, the ability of NMR spectroscopy to probe protein–protein interactions at a residue-specific level, even if these are transient in nature (3, 7, 29), render NMR uniquely suitable to study and probe the protein–protein interactions involved in the chaperone network.

2.1. Nuclear Magnetic Resonance Methods to Study Molecular Chaperones and Their Complexes with Substrates

Despite the advantages of NMR in terms of residue-specific information and the ability to probe weak biomolecular interactions, the molecular size limitations of conventional NMR constitute a significant drawback. Fortunately, recent progress in isotopic labeling strategies (146, 151), pulse sequence design (55, 75, 107, 113, 144, 147–149), and hardware have succeeded, to a large extent, in overcoming limitations on molecular size (120, 126). There are two main NMR-based approaches: (a) NMR investigation of an NMR-visible species in exchange with a large NMR-invisible or dark state and (b) direct observation of large assemblies using labeling schemes and pulse sequences that minimize the loss of magnetization during experiments.

NMR-invisible or dark state:

a state that is not directly detectable by NMR due to its low population or large molecular weight

Chemical exchange:

in NMR terms, the transfer of a nucleus to a different chemical environment

Transverse relaxation:

decoherence of the nuclear spin magnetization in the xy plane resulting in signal loss

Chemical shift:

the resonant frequency of a nucleus in a magnetic field in reference to a standard

2.2. Exchange-Based Nuclear Magnetic Resonance Techniques

NMR is uniquely capable of capturing molecular motions that span nanoseconds to minutes at atomic resolution (4, 7). Fast pico- to nanosecond motions describe dynamics within a particular energy well (for example, the dynamics of an α-helix around its equilibrium position), with slower (millisecond) motions usually arising from larger conformational changes between two (or more) energy wells (or states) (62). Conformational exchange on the millisecond timescale is particularly relevant for the study of chaperones, since it may provide information on exchange between various conformational states of the chaperone or substrate, as well as between the apo and bound forms of these proteins. In the past 15 years, various NMR methods have been developed to study chemical exchange processes even if the populations of states A and B are highly skewed (e.g., population of state A: p_A = 95%; population of state B: p_B = 5%) (7, 14). These methods normally assume exchange between states of similar molecular weight (and therefore with similar transverse relaxation rates, R_2,A = R_2,B, in the absence of exchange). However, this may not be the case when binding of a substrate to a large chaperone is involved. Below, we give a short description of the NMR methods designed to probe chemical exchange, with emphasis on the effect of a large R_2,B value (i.e., where one of the components of the system, B, has a large molecular weight). For a more technical description of these methods, the reader is referred to several excellent reviews (7, 98, 145).

In a system where state A exchanges with state B, the relative positions of the NMR peaks (chemical shift difference Δω_AB) depend on the overall exchange rate, k_ex (given by the sum of the forward and backward rate constants). In the fast exchange limit (Δω_AB ≪ k_ex), a single peak is observed at the population-weighted average of the chemical shifts of states A (ω_A) and B (ω_B), while in the slow exchange limit (Δω_AB > k_ex), two separate peaks are observed at frequencies ω_A and ω_B. In such systems, the observed transverse relaxation rate has contributions from both the intrinsic relaxation rate (R₂) and chemical exchange (R_ex). The latter is dependent on Δω_AB and k_ex. During a Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion experiment (101), the R_ex contribution is modulated by the frequency of refocusing 180° pulses (ν_CPMG) applied when the magnetization is in the xy plane. CPMG relaxation dispersion data collected over a range of ν_CPMG values and at two (or more) magnetic field strengths can be fitted simultaneously to obtain the values of Δω_AB, k_ex, and p_B (where p_A = 1 − p_B). For simple two-site exchange, analytical expressions that describe the evolution of magnetization during a CPMG relaxation dispersion experiment are readily derived; for more complicated schemes involving three or more states, numerical treatment of the McConnell equation (100) is necessary. Importantly, the residue-specific R_ex terms are sensitive to the R_2,B values (133). When the magnetization is placed on the excited state B through chemical exchange, the large value of R_2,B causes the magnetization of state B to relax before it has a chance to evolve. This results in a significant reduction in the magnitude of both R_ex (Figure 2a) and the exchange-induced shifts (δ_ex) (Figure 2b,c) and has to be taken into account when fitting even simple chemical shift titration data to avoid erroneous values for the exchange parameters.

Figure 2.

Nuclear magnetic resonance (NMR) methods for studying chemical exchange processes between states with different chemical shifts. A simple two-state exchanging system between states A and B is considered. (a) Simulated Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion curves with k_ex 800 s⁻¹, p_B = 2.5%, and Δω_AB = 150 Hz. The green curve indicates R_2,A = R_2,B = 10 s⁻¹; the gray curve indicates R_2,A = 10 s⁻¹ and R_2,B = 1,000 s⁻¹. (b) Simulated one-dimensional NMR spectra using the McConnell equations (161) for a two-state system. Chemical shifts and lineshapes were calculated as the imaginary and real parts of the eigenvalues of the matrix that describes the evolution of the magnetization of states A (M_A) and B (M_B) during acquisition (100 ms). The complete spectrum was calculated as a combination of Lorenzian lineshapes by summing over the two spins. The exchange parameters were the same as in panel a with p_B varied from 0 to 33% by varying the value of k₁ as indicated by the color scheme. (Left) R_2,A = R_2,B = 10 s⁻¹. (Right) R_2,A = 10 s⁻¹ and R_2,B = 1,000 s⁻¹. (c) Values of exchange-induced shifts (δ_ex) calculated for the spectra in panel b. The green curve indicates R_2,A = R_2,B = 10 s⁻¹ (corresponding to the left subpanel in panel b); the gray curve indicates R_2,A = 10 s⁻¹ and R_2,B = 1,000 s⁻¹ (corresponding to the right subpanel in panel b).

Although CPMG relaxation dispersion methods are ideal to probe exchange processes with k_ex values ranging from 10 to 3,000 s⁻¹, faster processes generally require the application of relaxation dispersion in the rotating frame (R1ρ) (98, 109). By using selective excitation techniques, weak spin-lock radiofrequency (RF) fields (<1,000 Hz) can be applied, extending the applicability of R1ρ relaxation dispersion to slower timescales (81). In these experiments, the application of a RF spin-lock field keeps the magnetization aligned in the direction of the effective field in the rotating frame. If, on the one hand, the interconverting states have different chemical shifts (Δω_AB ≠ 0), then non-flat relaxation dispersion profiles with a maximum around ω_B are measured as a function of the offset of the spin-lock field (Figure 3a). If, on the other hand, Δω_AB = 0 and R_2,B − R_2,A > k_ex, then off-resonance R1ρ profiles are shifted vertically and have an inverted appearance (175) (Figure 3a) characteristic of exchange with a large-molecular-weight species (73, 175). Another, perhaps simpler, NMR method to probe binding to a state of large molecular weight involves lifetime line broadening (40, 41). In this case, the difference in transverse relaxation rates (ΔR₂) in the presence and absence of exchange is measured, with care taken to suppress any contributions from chemical exchange by the application of an appropriate RF spin-lock field. On the one hand, when exchange is slow on the relaxation timescale (R_2,B ≪ k_ex), the magnetization decays fast once it is transferred to the excited state, and thus the measured ΔR₂ is to a good approximation equal to k_on (145) (Figure 3b). On the other hand, when exchange is fast on the relaxation timescale (R_2,B ≫ k_ex), R₂ ~ p_B (R_2,B − R_2,A), reflecting the weighted population average of the transverse relaxation rates of the two states (145) (Figure 3b). ΔR₂ measurements are generally combined with dark state exchange saturation transfer (DEST) (40, 145), which exploits the application of a relatively weak RF field and relies on large R_2,B values for the dark, or excited, state (Figure 3c). These methods are particularly important for investigating the interaction of an NMR-visible substrate with a high-molecular-weight, NMR-invisible chaperone. Thus, DEST, CPMG relaxation dispersion, and ΔR₂ measurements have been used successfully in our laboratory to probe exchange processes involving large-molecular-weight chaperones such as GroEL (86–88, 156, 159) and DNAJB6b (72, 73).

Figure 3.

Nuclear magnetic resonance (NMR) experiments designed to probe exchange between states with different transverse relaxation rates. (a) Simulated off-resonance R_1ρ profiles with k_ex = 8,000 s⁻¹, p_B = 2.5%, and a spin-lock frequency of 1,000 Hz. The green curve indicates Δω_AB = 450 Hz (6 ppm) and R_2,A = R_2,B = 10 s⁻¹; the gray curve indicates Δω_AB = 0 Hz, R_2,A = 10 s⁻¹, and R_2,B = 3,000 s⁻¹. The offset of the excited stated is shown by the dashed line. (b) Plot of R₂/k_on versus R_2,B for k_ex = 100 s⁻¹ (dashed line), p_B = 5%, R_2,A = 10 s⁻¹, and Δω_AB = 0 Hz. In the fast exchange limit (to the left of the dashed line), the measured ΔR₂ is approximately equal to the population-weighted average of R_2,A and R_2,B, while in the slow exchange limit (to the right of the dashed line), ΔR₂ ~ k_on. (c) Simulated dark state exchange saturation transfer profiles using the same parameters as in panel b with a spin-lock of 200 Hz for various R_2,B values, as indicated by the color bar.

2.3. Nuclear Magnetic Resonance Methods that Allow Direct Observation of High-Molecular-Weight Chaperones

Initially, heteronuclear NMR studies on protein dynamics largely focused on backbone atoms such as ¹⁵N using ¹H-¹⁵N correlation-based experiments (26, 78). For small proteins (<150 residues), the chemical shift dispersion in both ¹⁵N and ¹H_N dimensions is good, and backbone ¹⁵NH amide groups constitute isolated spin systems. For larger-molecular-weight systems, however, ¹H-¹⁵N correlation experiments suffer from low sensitivity and significant cross-peak overlap. Sidechain methyl groups, in contrast, display excellent sensitivity, even for high-molecular-weight systems, due to rapid rotation around the threefold symmetry axis (107). The development of metabolic precursors for specific protonation of the ¹³CH₃ moiety in an otherwise highly perdeuterated background (146, 151) popularized methyl-NMR for the study of high-molecular-weight proteins (32, 116, 143, 150). At the present time, it is common practice to use isoketobutyric or isovaleric acid precursors during the expression of recombinant proteins in Escherichia coli to specifically protonate the δ1 methyl group of isoleucine and the two methyl groups of valine and leucine; other precursors can be used for isotopic labeling of the methyl groups alanine and threonine (122). Apart from E. coli, other labeling strategies involving eukaryotic cells (insect or mammalian) and cell free synthesis are available (125).

The success of any NMR experiment relies on the amount of magnetization that survives before acquisition of the NMR signal. For methyl groups, transverse relaxation optimized spectroscopy (TROSY) is achieved by recording ¹H-¹³C heteronuclear multiple quantum correlation spectra (107) that are of excellent quality even for proteins up to 2 MDa in molecular weight. Resonance assignment can be performed in favorable cases using linear spin systems by transferring the magnetization to backbone atoms with already known assignments (150). Alternatively, nuclear Overhauser enhancement (NOE) or paramagnetic relaxation enhancement restraints can be used if the three-dimensional structure of the protein is available, a procedure that can be largely automated (105, 114, 155). For multidomain proteins, divide-and-conquer approaches for assignment purposes are common and can be aided by site-specific mutagenesis to resolve ambiguities in the assignment process (122, 125). Importantly, methyl groups are well distributed throughout protein structures and can therefore serve as valuable reporters of both structure and dynamics (47, 152, 153, 161).

The development of structure calculation protocols (83, 132) that allow the generation of reasonably accurate structural models from backbone chemical shifts, supplemented by sparse NMR-derived structural restraints such as methyl–methyl NOEs and/or residual dipolar couplings (RDCs), has allowed the NMR investigation of challenging targets that would not otherwise be possible. An AX3 spin system of a ¹³CH3 methyl group contains transitions that relax with different rates, a fact that poses challenges but also creates opportunities for sophisticated relaxation experiments to investigate methyl dynamics (80, 148, 149, 152, 153). Thus, many of the NMR experiments designed to detect protein motions described in the previous section have been adapted to methyl groups.

3. CHAPERONE MECHANISMS ELUCIDATED BY NUCLEAR MAGNETIC RESONANCE

In terms of the proteostasis network, NMR studies have been instrumental in elucidating the molecular mechanisms of various chaperones, including the Hsp40–Hsp70 molecular machine (68, 72, 73, 178), GroEL (86–88), SecB (67), Trigger Factor (124), periplasmic bacterial chaperones (23, 25, 97), and small HSPs (15, 16). In the following sections, we delve into a few chaperone systems for which the mechanisms were elucidated using NMR spectroscopy.

3.1. J-Domain Proteins Involved in Anti-Aggregation

J-domain (JD)-containing chaperones (DNAJs) are members of the Hsp40 family and generally act in concert with the Hsp70 machine to promote client protein folding (45, 51, 70, 115, 130). In addition to the conserved N-terminal, α-helical JD (110), the canonical isoform DNAJB1 contains a β-stranded C-terminal domain (CTD) (66, 168) that varies between different DNAJ subclasses. The CTD domain(s) of DNAJ proteins recognize misfolded substrates and deliver them to Hsp70, which, in turn, uses the energy from ATP hydrolysis to promote correct folding of the substrates (70). Crucially, the interaction of the JD with Hsp70 significantly increases the ATPase activity of Hsp70 (51), and thus, the JD can be considered as a catalyst of the entire Hsp70 cycle. In addition, new Hsp40 functions have been recently discovered that do not depend on interaction with Hsp70 (33). For example, the DNAJB6b isoform has been found to be a potent inhibitor of various biological processes, including amyloid formation (69, 94, 95) and virus replication (137). Furthermore, DNAJB6b has been implicated in the etiology of muscular dystrophy (174), various cancers (102), and Parkinson’s disease (9).

Hsp40–Hsp70 was one of the first chaperone–cochaperone pairs to be discovered in the 1970s (46, 179) and attracted extensive biochemical research interest in the years following its discovery (110). Structural studies have elucidated the structure of individual domains using constructs lacking the long and flexible Gly/Phe (GF) linker, rich in glycine and phenylalanine residues, that connects the JD and CTD. However, data dating back to 1999 clearly emphasize the critical importance of the linker regions for the specificity of the Hsp40–Hsp70 interaction (171) that drives the multifunctionality of the Hsp70 machine. Long and flexible regions in proteins have traditionally posed a significant challenge to structural methods, especially for those techniques that rely on the organization of the target protein in the form of a crystal (X-ray crystallography) or on the surface of a grid (cryo-EM). Even though the main processes in the Hsp40–Hsp70 machine were elucidated biochemically many years ago (70), only with recent advances in NMR spectroscopy, described in the previous section, has a detailed description of the regulatory mechanisms that are in play been possible (39, 68, 72). Such an understanding is key to rationally targeting this important chaperone system for therapeutic intervention in a substrate-specific manner.

In contrast to DNAJB1, which exists as a stable dimer, full-length DNAJB6b is an oligomeric protein (95). Using a deletion construct that preserves the structure of monomers but disrupts large DNAJB6b oligomers, Karamanos et al. (72) were able to structurally investigate, for the first time, an Hsp40 construct that contains the JD and CTD connected by the GF linker. Taking advantage of methyl-based NMR methods, including methyl NOE studies supplemented with backbone chemical shifts, backbone amide RDCs, and backbone amide paramagnetic relaxation enhancement measurements, they solved the solution NMR structure of monomeric DNAJB6b (72). With respect to the JD domain, the results were not surprising: The α-helical JD of DNAJB6b looks very similar to that of previous JD structures (110). However, part of the GF linker, previously thought to be intrinsically disordered, was found to form a stable helix (helix 5) that is docked on to the JD domain (72) (Figure 4a). This JD–GF interaction completely blocks the Hsp70 binding interface of the JD, thereby resulting in an auto-inhibited form of DNAJB6b (72). It is clear that auto-inhibition has to be released by undocking of helix 5 to allow Hsp70 binding and completion of the cycle; however, the mechanism of helix 5 dissociation from the JD remains unclear and may be class specific. Based on these data, Karamanos et al. (72) proposed that the GF linker, through helix 5, provides a previously unknown extra layer of regulation to the entire Hsp70 cycle. Given the high degree of amino acid sequence conservation around helix 5, this auto-inhibition mechanism might be a general feature across class B Hsp40s. Indeed, in 2020, Faust et al. (39) confirmed the mechanism of regulation provided by helix 5 (Figure 4a) for DNAJB1 by demonstrating, using paramagnetic NMR, that helix 5 is displaced from the JD domain upon addition of a peptide comprising the Glu-Glu-Val-asp (EEVD) tail of Hsp70.

Figure 4.

New insights into Hsp40 function and regulatory mechanism. (a) Revised Hsp40–Hsp70 cycle based on the results from recent nuclear magnetic resonance (NMR) studies (39, 68, 72). Dimeric DNAJB1 is shown; however, the regulatory mechanism involving helix 5 in the Gly/Phe (GF) linker was originally found in DNAJB6 (72) and is likely a general feature of class B Hsp40s. In free DNAJB1, helix 5 in the GF linker forms a helix that blocks access of Hsp70 to the J-domain (JD) (top left). Substrates bind to the C-terminal domain (CTD) I of Hsp40 and are competitively displaced by the C-terminal tail of Hsp70 (blue schematic, top right). Dissociation of helix 5 from the JD releases auto-inhibition, allowing the JD to interact with the ATPase domain of Hsp70, thereby increasing its ATPase activity (bottom right). The cycle is completed upon substrate handover to Hsp70 (bottom left). Different substrates can recognize either CTDI or CTDII, or both. Whether the C-terminal tail of Hsp70 binds to the CTDI or CTDII depends on the specific Hsp40–Hsp70 pair. (b) An array of interdomain or intermolecular interactions determine the oligomeric state of DNAJB6. Based on a variety of monomeric truncation constructs, NMR studies have shown that the major open DNAJB6 monomer species is in exchange with a closed form of the monomer, in which the JD and CTD are in close proximity, and a dimer. Dimer formation is dictated by a conformational change within the Ser/Thr (ST) region of the β1 strand of the CTD from a twisted form in the monomer to a straight configuration in the dimer (inset shows the ST region in red). Dimers readily associate with large oligomeric species that are 20–25 subunits in size. Figure adapted from figures in References 72 and 73, published in PNAS while the authors were US Government employees at the National Institutes of Health.

Although the main features of the regulatory interaction between the JD and helix 5 may be shared by various Hsp40 subclasses, CTD-mediated Hsp40 functions may be more class specific. In the case of DNAJB1, the CTD consists of two β-sandwich domains, CTDI and CTDII, that share high sequence conservation and similar structures (66). Based on X-ray crystallographic studies using short peptide substrates, the CTD is known to be responsible for substrate recognition (85). More recently, NMR studies using fusion constructs between the CTDs of various Hsp40s and model substrate proteins revealed that Hsp40 substrates retain a high degree of flexibility in the bound state with no significant secondary structure formed (68). Substrates can bind either the CTDI or CTDII, or both domains simultaneously, with the physicochemical properties of the substrate (e.g., hydrophobicity) determining the specificity for a particular CTD (39, 68). Depending on the specific Hsp40–Hsp70 pair, the unfolded C-terminal tail of Hsp70 can also bind to either the CTDI (e.g., in DNAJB1) or CTDII (68). This interaction displaces CTDI-bound substrates and creates a ternary complex between the client protein Hsp40 and Hsp70. Further conformational changes are then required to mediate substrate handover to Hsp70 and docking of the JD of Hsp40 to the ATPase domain of Hsp70 (39, 68). In addition to interacting with substrates, the CTD can also undergo self-association. In the case of DNAJB6, Karamanos et al. (73) demonstrated the existence of an elegant mechanism mediated by subtle conformational changes in the N-terminal edge strand of the CTD that determines the oligomeric state of this chaperone. These conformational changes appear to be important for subunit exchange in DNAJB6b oligomers, since mutations in the same region abolish the ability of DNAJB6 to inhibit protein aggregation (96) and are associated with Parkinson’s disease (9).

In summary, it is clear that Hsp40s use a complicated network of protein–protein interactions involving unfolded substrates and flexible multidomain proteins that orchestrate their motions to recognize specific substrates and maintain proteostasis.

3.2. The Chaperonin GroEL Nanomachine

The discovery of cylindrical chaperonins was instrumental to the conceptualization of the proteostasis network (42, 48, 49, 58, 90, 123, 139–142). The class I chaperonins (e.g., HSsp60) are highly conserved, large (up to 1 MDa in size) ATP-dependent chaperones that form ring-shaped assemblies that contain a folding chamber in each ring. The nanocages are often capped with a cochaperone during the course of the cycle and serve to encapsulate substrates, facilitating correct folding through an annealing mechanism (139–141). The bacterial chaperonin GroEL and its cochaperone GroES represent a paradigm of a nanomachine dedicated to protein folding (60, 139–141). GroEL consists of two heptameric rings with each 57-kDa subunit comprising an ATP-binding domain, a hinge domain, and an apical domain. The C-terminal 23 residues of the apical domain are disordered and protrude into the cavity blocking free passage between the rings (60). GroES, in contrast, is a dome-shaped ring that interacts with the apical domain of GroEL, an interaction that leads to substrate encapsulation, creating a bullet-shaped assembly. Large ATP-driven conformational changes are associated with GroES binding, leading to an expansion of the cavity by a factor of two (170). GroEL is known to recognize exposed hydrophobic segments of unfolded, misfolded, and intermediate states (134, 135) that bind to two or three apical domains (38, 86). Single-molecule Förster resonance energy transfer studies have shown that some substrates are transiently (approximately 15 ms) expanded upon GroEL binding, followed by further expansion caused by subsequent ATP-dependent conformational changes (89, 131). In one model (60), GroES binding to the trans ring leads to substrate compaction (approximately 200 ms) and transfer into the cis ring where it can properly fold; the cycle is completed by ATP binding to the trans ring, causing opening of the cis ring and substrate release. However, this model of GroES binding to each of the GroEL rings sequentially has been challenged by the observation of a fully bound, football-shaped GroEL (GroEL-GroES₂) in the presence and absence of substrate (172, 173).

Aggregate:

assemblies that contain more than one nonfunctional protein molecule

Even though many mechanistic details regarding the conformational cycle of GroEL are available, the mechanism by which GroEL actually promotes folding is still largely unknown. In the simplest scenario, GroEL acts as a passive cage that separates individual substrate molecules from the cellular milieu, where they can fold at effectively infinite dilution. Alternatively, GroEL could play an active role in promoting client protein folding by confinement of non-native intermediates (active cage model). A third possibility is that a misfolded, kinetically trapped intermediate is recognized by GroEL, which is able to unfold it before it is released into the GroEL cavity or the bulk solution to refold (iterative annealing model) (139, 141). Although single-molecule fluorescence studies suggest that reduced polypeptide mobility in the confined state leads to an increased folding rate (52), and hydrogen exchange experiments (118, 176, 177) showed that GroEL could potentially unfold misfolded states, direct evidence to support the (un)foldase function of GroEL was difficult to obtain. Using a triple mutant of Fyn-SH3, a well-defined system that exists in equilibrium between the native and a partially folded intermediate state (106), and by fitting a large number of NMR-based relaxation data, Libich et al. (87) were able to show that apo GroEl accelerates the overall exchange rate for the folded to intermediate state transition by approximately 20-fold. This is achieved by an approximately 500-fold increase in the rate constant for the conversion of the native to intermediate state in the presence of the GroEL relative to that for free SH3 (87) (Figure 5). SH3–GroEL interactions are predominantly hydrophobic in nature, with the hydrophobic surface that is exposed in the intermediate state formed by discontinuous regions of the polypeptide chain (87). Paramagnetic relaxation enhancement studies showed that SH3 not only interacts with the outer rim (mouth) of GroEL, but also is able to penetrate deep into the cavity, potentially facilitating encapsulation (157). Even though some of the states in the model in Figure 5 are as sparsely populated as 0.25%, the careful, simultaneous analysis of a plethora of exchange-based NMR experiments allows their accurate kinetic characterization and discrimination between various kinetic models. In turn, this kinetic analysis was the first direct demonstration of an intrinsic unfoldase activity of the chaperonin GroEL, clearly favoring the iterative annealing model (87). In addition to stabilizing the intermediate state of SH3 relative to the native folded state by exposure of a hydrophobic patch that favors GroEL binding (87), the unfolded state, in turn, is also destabilized relative to the intermediate state within the GroEL cavity on account of the larger radius of gyration of the unfolded state relative to the intermediate state (88).

Figure 5.

Mechanism of GroEL-assisted protein folding. A triple mutant of Fyn SH3 (shown as a ribbon diagram) is in exchange between major folded (F) and sparsely populated intermediate (I) states. GroEL is able to bind to both states (F-G and I-G, respectively) and catalyzes the folded-to-intermediate transition by approximately 20-fold. Exchange between the four states was determined quantitively by Libich et al. (87), while the rest of the figure represents a schematic representation of possible steps that may follow in the general case. These include binding of GroES, encapsulation of substrate, conformational changes in subunit orientation associated with ATP hydrolysis, and subsequent release of substrate into the enclosed cavity (60, 140). Interaction of the misfolded substrate with the chamber of GroEL may promote its proper folding and release by a process of iterative annealing to complete the cycle (140, 141). Figure adapted from Reference 87, published in PNAS while the authors were US Government employees at the National Institutes of Health.

With regard to substrates that aggregate and/or form fibrils, previous work showed that amyloid-β40 (Aβ40) binds to GroEL using two hydrophobic linear sequences (86), while Aβ42 has three such sites (159). Furthermore, interaction of GroEL with Aβ42 not only slows down the rate of aggregation and fibril formation, but also protects neuronal cells’ amyloid-related cytotoxicity (159). In the case of the prior protein Het-s, however, GroEL enhances protofibril and fibril formation by decorating the fibrils at regularly spaced intervals as a consequence of binding to exposed loops on their surface (158).

In summary, GroEL can recognize client proteins either through contiguous stretches of hydrophobic sequences present in intrinsically disordered polypeptides (e.g., amyloid β) and loops of folded proteins (e.g., Het-s fibrils) or via hydrophobic patches formed by discontinuous sequences on the surface of folding intermediates (e.g., SH3). The studies discussed above clearly demonstrate the power of NMR relaxation-based methods to extract detailed kinetic and structural information for complexes involved in the chaperonin network.

3.3. Small Heat Shock Proteins

HSPs take their name from the stress response induced by heat shock, which was used to identify the pathway that leads to the upregulation of these proteins (117). This broad definition was clarified later by further subclassifying the proteins involved based on domain architecture or function. Small HSPs are important members of the chaperone network (59) and are associated with inhibition of protein aggregation (164), cancer (136), and various neuropathies (37). Members of this class of chaperones, such as the abundant HSP27 (or HSPB1), αA-crystallin (αAC), and αB-crystallin (αBC), are known to form dynamic, heterogeneous assemblies (8). Despite the high molecular weight of the oligomeric species, the dynamic nature of small HSPs has facilitated several NMR studies that have provided crucial insights into their function (2, 5, 16, 31, 93). In general, small HSPs consist of disordered N-terminal domains and CTDs that flag a conserved α-crystallin domain (ACD). All three domains are involved in heterogeneous small HSP self-assembly into oligomers of different size and complexity, while removal of the disordered N-terminal domains and CTDs results in ACD dimers, stabilized by an intermolecular disulfide bond (5). Although different oligomeric states have been implicated as the active chaperone species, a monomeric ACD has recently been shown to have a higher chaperone activity than the oligomeric assemblies (5). Using truncation constructs that isolate the monomer–dimer equilibrium from equilibria involving higher-order oligomers and by applying NMR relaxation and high-pressure NMR methods, Alderson et al. (5) showed that the two β-strands (β6 and β7) involved in the dimerization interface become more flexible in the monomer. The exposed interface leads to HSP27 aggregation but may also increase the ability of the monomer to bind other aggregated species (5).

In addition to strands β6 and β7, which are involved in chaperone activity, small HSPs use a conserved groove formed by strands β4 and β8 in the ACD to recognize their substrates (93). In the various oligomeric forms, the β4–β8 binding groove can be occluded by a conserved tripeptide sequence, known as the IXI motif, located in the CTD (14). Transient interactions between the CTD and ACD are responsible for wiring up small HSP oligomers, providing a basis for subunit exchange in the oligomeric state (13, 16), and may also prevent substrate binding (11).

The unfolded N-terminal domain of small HSPs can also bind substrates and contains alternative IXI motifs that compete for binding to the β4–β8 groove (13, 16). Interestingly, the role of the β4–β8 groove in substrate recognition seems to be substrate specific. Some substrates, such as Aβ40, bind specifically to the β4–β8 groove (93); others, such as lysozyme, bind to the disordered N terminus (93), and other client proteins, such as tau, recognize both (18). Which of these interactions lead to efficient chaperoning seems to depend on the physicochemical properties and aggregation state of the substrate. Recent NMR studies have revealed that disease-related mutations in the C-terminal IXI motif of HSP27 weaken HSP27’s affinity for its own ACD, leading to decreased chaperone activity and more efficient recruitment of IXI-containing cochaperones (2). In addition, post-translational modifications of the C-terminal IXI motif increase the ability of HSP27 to inhibit aggregation of amyloid forming substrates such as α-synuclein and Aβ40 (11).

The picture that emerges from these results is of a complex energy landscape that is sensitive to multiple contributing factors, including dynamic motions of the IXI motif and small HSP oligomer assembly or disassembly, that control the activity of small HSPs. It is interesting to note that similar dynamic oligomer equilibria, modulated by regions of high flexibility, seem to govern the functions of other chaperones involved in inhibition of polymerization reactions, such as DNAJB6 (73) (Figure 4b). Dynamic subunit exchange in oligomeric chaperones may therefore constitute a general, conserved mechanism used by cells to combat protein misfolding or aggregation. Combining powerful NMR methods to characterize dynamic processes with techniques such as mass spectrometry that are able to directly detect large, heterogeneous assemblies (64) probably represents the only experimental approach for investigating such systems.

4. CONCLUSIONS AND PERSPECTIVES

Although biomolecular NMR is well established as one of the prime methods for solving the three-dimensional structure of proteins, the number of NMR structures deposited in the Protein Data Bank has declined in recent years. The revolution in resolution afforded by cryo-EM and the limitations in terms of protein size associated with NMR have generally led to the use of other structural methods, specifically crystallography and cryo-EM, as the go-to tools for structure determination of stably folded proteins. At the same time, the field of protein chemistry is realizing that transient interactions, sparsely populated states, and intrinsic disorder are the main drivers behind many important biological processes. Phenomena such as chaperone binding, protein aggregation, and phase separation are dictated not by the specific structure of the proteins involved but by weak biomolecular contacts, the timescale and reversibility of multibody interactions, and the local structural propensities of largely flexible proteins. As described in this review, NMR has already made significant contributions to the quest to elucidate how chaperones work, and it is clear that we are just beginning to uncover the fascinating mechanisms that govern one of the most vital cellular systems. Thus, in the years to come, it is evident that NMR will keep enriching our knowledge of the protein world by providing information that is not accessible using other structural methods.

SUMMARY POINTS.

1. Chaperones assist in maintaining the health of the proteome by forming dynamic complexes with their substrates.

2. Multivalency, combined with the low affinity of each individual binding site, permits efficient capture of substrates by chaperones and promotes substrate folding of chaperone-bound substrates.

3. NMR spectroscopy has emerged as a powerful tool for characterizing the molecular and structural biophysics of chaperones, even if some chaperones are of very high molecular weight.

4. The application of NMR methods has recently led to the discovery of novel chaperone mechanisms that resolve previous controversies.

FUTURE ISSUES.

1. Studies using model substrates may not capture the exact chaperone mechanisms used inside the cell. Specifically, in terms of anti-aggregation properties, chaperoning is often performed by binding to oligomeric or aggregated forms of the substrate that are still difficult to capture experimentally.

2. Although much of current research has focused on the interaction of specific chaperone–substrate complexes, a much more complex network of multiple chaperones and various client proteins is present in cells.

3. With regard to NMR, the vast majority of high-molecular-weight complexes are probed by methyl-based experiments. New NMR methods that make use of other sidechains are needed.

4. There is still a lot of room for improvement with regard to the integration of NMR data with structural data from other methods, such as cryo-EM, to obtain a complete kinetic and structural picture of how chaperones function.