BI83CH13-Kay
ARI
3 May 2014
ANNUAL REVIEWS
11:18
Further
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
Click here for quick links to Annual Reviews content online, including: • Other articles in this volume • Top cited articles • Top downloaded articles • Our comprehensive search
Bringing Dynamic Molecular Machines into Focus by Methyl-TROSY NMR Rina Rosenzweig1,2,3 and Lewis E. Kay1,2,3,4 Departments of 1 Biochemistry and 3 Molecular Genetics, University of Toronto, Toronto, Ontario M5S 1A8, Canada 2
Department of Chemistry, University of Toronto, Toronto, Ontario M5S 3H6, Canada
4
Program in Molecular Structure and Function, Hospital for Sick Children, Toronto, Ontario M5G 1X8, Canada
Annu. Rev. Biochem. 2014. 83:291–315 The Annual Review of Biochemistry is online at biochem.annualreviews.org This article’s doi: 10.1146/annurev-biochem-060713-035829 c 2014 by Annual Reviews. Copyright All rights reserved
Keywords high–molecular weight complexes, protein NMR, allostery, protein–substrate interactions, methyl labeling, deuteration
Abstract Large macromolecular assemblies, so-called molecular machines, are critical to ensuring proper cellular function. Understanding how proper function is achieved at the atomic level is crucial to advancing multiple avenues of biomedical research. Biophysical studies often include X-ray diffraction and cryo–electron microscopy, providing detailed structural descriptions of these machines. However, their inherent flexibility has complicated an understanding of the relation between structure and function. Solution NMR spectroscopy is well suited to the study of such dynamic complexes, and continued developments have increased size boundaries; insights into function have been obtained for complexes with masses as large as 1 MDa. We highlight methyl-TROSY (transverse relaxation optimized spectroscopy) NMR, which enables the study of such large systems, and include examples of applications to several cellular machines. We show how this emerging technique contributes to an understanding of cellular function and the role of molecular plasticity in regulating an array of biochemical activities.
291
BI83CH13-Kay
ARI
3 May 2014
11:18
Contents
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
INTRODUCTION . . . . . . . . . . . . . . . . . . NMR STUDIES OF SUPRAMOLECULAR SYSTEMS: TECHNICAL ASPECTS . . . . . . . . . BINDING AND MECHANISM . . . . . Protein Disaggregation . . . . . . . . . . . . . Molecular Basis of Synaptic Vesicle Priming . . . . . . . . . . . . . . . . . . . . . . . . Binding Mechanism of the HMGN2 Protein to the Nucleosome Core Particle . . . . . . . . . . . . . . . . . . . . . . . . . Measurement of Active-Site pKa Values in the Proteasome. . . . . . . . Other Systems . . . . . . . . . . . . . . . . . . . . . DYNAMICS AND ALLOSTERY . . . . Proteasome Degradation Machinery . . . . . . . . . . . . . . . . . . . . . . Mapping the Energy Landscape of the Allosteric Hsp70 Chaperone . . . . . . . . . . . . . . . . . . . . . KcsA Potassium-Channel Dynamics . . . . . . . . . . . . . . . . . . . . . . Exosome Dynamics . . . . . . . . . . . . . . . . CONCLUDING REMARKS . . . . . . . . .
292
293 293 294 295
297 298 300 301 301
305 307 308 309
INTRODUCTION Solution NMR spectroscopy has become a powerful tool for the study of protein interactions and dynamics; it provides atomicresolution structural information under near-physiological conditions (1–3). Although most applications during the past two decades have focused on relatively small protein systems with molecular weights within ∼50 kDa, in the past several years there has been significant interest in the use of solution NMR spectroscopy for studies of high–molecular weight protein complexes. This interest has been fueled by the tremendous potential of NMR in studies of molecular dynamics and by the realization of the importance of such dynamics to biochemical function (4–6).
292
Rosenzweig
·
Kay
Accordingly, investigators have expended great effort in developing new isotope labeling strategies (7–9) and complementary NMR experiments that enable the preservation of NMR signals that would otherwise decay rapidly for large-protein systems (10, 11). Studies of protein complexes by solution NMR are, in fact, not new at all. Examples from as early as the 1970s and 1980s include hemoglobin (12, 13), alkaline phosphatase (14, 15), dehydrogenases (16), serine proteases (17, 18), aspartate transcarbamylase (19–23), aminotransferases (24, 25), and various muscle proteins (26–30). In these systems, NMR was used to learn about mechanisms, and the questions addressed often could be answered either by the addition of 19 F, 13 C, or various metal probes or by focusing on shifted proton resonances in cases of paramagnetic proteins. Current research efforts are building on the same fundamentals as these early pioneering studies, with a continued emphasis on understanding biochemical function. Today, however, it is possible to focus on much larger systems with molecular weights reaching 1 MDa and to provide a more atomistic view based on a larger number of probes. The objective of these studies is to address specific questions within the broader context of existing biophysical and biochemical data while using the unique strengths of NMR spectroscopy. These advantages include sensitivity to molecular interactions over a broad spectrum of affinities, including those in the millimolar-tomicromolar range that often are not quantifiable using other methods (31), and the ability to probe functional molecular dynamics over many orders of magnitude in timescale (4, 32). In what follows, we briefly discuss both the methyl-labeling approach that has been developed for studies of supramolecular systems and the experiments that are designed to take advantage of the labeling (33). We illustrate the utility of the methodology through discussion of several examples that focus on important molecular machines and demonstrate the significance of structural plasticity to function.
BI83CH13-Kay
ARI
3 May 2014
11:18
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
NMR STUDIES OF SUPRAMOLECULAR SYSTEMS: TECHNICAL ASPECTS Central to the success of any NMR study is the appropriate choice of labeling. For applications to high–molecular weight protein systems, an important and general approach involves the use of methyl group probes (13 CH3 ) in an otherwise highly deuterated environment (8, 34–45). Selection of this labeling scheme is motivated not only by the fact that methyl groups are prevalent in hydrophobic cores and at molecular interfaces, thereby serving as reporters of internal dynamics and structure (46), but also by the high quality of the spectra obtained for such groups. Excellent spectral sensitivity results from the three equivalent protons in each methyl, coupled with the rapid rotation of the methyl moiety about its threefold symmetry axis, and its localization to flexible ends of side chains. Moreover, the inherent spin physics of a methyl group facilitates the preservation of NMR signals, even in high–molecular weight complexes, via a so-called methyl-TROSY (transverse relaxation optimized spectroscopy) effect that manifests in 13 C–1 H heteronuclear multiple quantum correlation (HMQC) spectra (11). Another important factor is that it is now possible to purchase precursors for 13 CH3 labeling of any methyl position in the six methylcontaining amino acids in a protein that is otherwise highly deuterated. Labeling of methyl groups as 13 CH2 D and 13 CHD2 is also possible (47), although applications involving these isotopomers are not described here. The requisite precursors are simply added to the growth media, prior to the induction of protein overexpression, to efficiently produce molecules with the desired methyl labeling (35, 36). Note that in the recording of optimal 13 C–1 H correlation spectra, which form the basis of any analysis (see the examples provided throughout this review), one must consider the trade-off between (a) using a sufficiently large number of probes so as to achieve good coverage over the complete protein sequence and (b) obtaining high-resolution and high-sensitivity spectra. As the number of
protons increases, spectra quality decreases because the magnetic fields produced by protons from one methyl group decrease the lifetimes of signals from adjacent methyl probes, leading to line broadening. With this trade-off in mind, labeling of Ileδ1, Leu, and Val methyl groups is often a good compromise (8). In this case, the prochiral methyl groups of a given Leu or Val side chain are labeled 13 CH3 ,12 CD3 by use of either precursors that are stereospecifically pro-R or proS 13 CH3 labeled (43) or precursors in which one methyl is labeled 13 CH3 and the other 12 CD3 in a random manner (8). Finally, in cases in which methyl groups are not localized to regions of interest, they can be “engineered in” through mutation of a residue to cysteine, followed by labeling with 13 CH3 −S−, producing a Met-like side chain (48). Details about protein labeling and expression (49, 50), as well as a discussion of the spin physics of a 13 CH3 methyl group attached to a large macromolecule that is exploited in methyl-TROSY NMR (11, 51, 52), can be found in several previous publications.
BINDING AND MECHANISM One of the most widespread applications of solution bio-NMR spectroscopy involves the study of proteins with their binding partners. This reflects that the binding of macromolecules to their targets is critically important for biological function and that NMR is a powerful technique for characterizing interactions structurally and for obtaining the thermodynamic and kinetic parameters that describe the binding process. Moreover, NMR can be extremely sensitive to binding events over a millimolar-to-nanomolar range of affinities and for submicrosecond-to-second timescales (31, 32). Not surprisingly, the literature is full of examples (3); the vast majority of them focus on protein systems with molecular masses lower than 50 kDa, for which traditional NMR methods are most effective (3, 31, 53– 55). However, although the average molecular www.annualreviews.org • Dynamic Molecular Machines
293
ARI
3 May 2014
11:18
mass of a eukaryotic protein is approximately 40 kDa (56), still within the range of traditional NMR, 70–80% of functional proteins are assembled into much larger oligomers that comprise several monomeric subunits (57, 58). For these systems, the preparation of highly deuterated protein(s) with methyl protonation and the use of methyl-TROSY approaches are necessary. We describe numerous applications below to illustrate how detailed mechanistic information may be obtained through simple methyl correlation spectra, thereby increasing our understanding of how important supramolecular complexes function.
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
Protein Disaggregation The aggregation of misfolded proteins can perturb protein homeostasis and cellular function, sometimes causing disease (59). Not surprisingly, cells have developed robust protein quality–control systems to counteract protein aggregation, both under normal physiological conditions and during stress. The heat-shock protein ClpB (or Hsp100) is the major protein disaggregase in yeast, bacteria, plants, and mitochondria of all eukaryotic cells, and it is essential for cell survival during severe stress (60–63). Although ClpB is an active ATPase in which ATP hydrolysis is critical for protein disaggregation, the recovery of functional proteins from aggregates requires synergistic interaction with a second ATP-dependent molecular chaperone, DnaK (60, 64–67); together, they form a powerful bichaperone system. Structural characterization of the ClpB/DnaK complex has been hindered, however, both by the low affinity of the complex and by the dynamic nature of its components. Thus, despite increasing biochemical data (59), only a partial understanding of the overall mechanism of protein disaggregation has been obtained. Rosenzweig et al. (68) used methyl-TROSY NMR methods to study the interaction between the 580-kDa ClpB hexamer and the 70-kDa DnaK chaperone. The binding interface for each of the two proteins in the 294
Rosenzweig
·
Kay
complex was elucidated using a series of NMR titrations wherein one of the components was selectively 13 CH3 labeled at the Ile-δ1, Leu-δ, Val-γ, and Met-ε positions in an otherwise fully deuterated background (referred to as 2 H,ILVM–13 CH3 labeling). Figure 1a shows a selected region from a series of 13 C–1 H HMQC spectra of labeled DnaK recorded with increasing concentrations of a fully deuterated (NMR-invisible) ClpB hexamer. The figure illustrates changes in chemical shifts of DnaK residues L280 and L48 upon complex formation; all residues whose methyl groups showed spectral changes were subsequently mapped onto the structure of DnaK to determine the binding interface for ClpB (Figure 1b). Similarly, the authors obtained the corresponding DnaK binding interface on ClpB by repeating the experiment with the labeling scheme reversed. In addition to mapping of the ClpB/DnaK complex, changes in chemical shifts upon binding were also used to calculate the dissociation constant (Kd ) for the complex. Chemical-shift changes for residues in the DnaK binding surface of ClpB were measured at each titration point and fitted to a one-to-one binding model (Figure 1c) to obtain a Kd of 25 ± 3 μM for the ClpB/DnaK association (68). Although such low affinities are typically very difficult to quantify by other biophysical techniques, NMR spectroscopy remains a powerful way to study complexes involving weak and/or transient interactions. Beyond localizing binding sites, detailed structural information about the complex can be obtained when structural models of the individual molecular players are available. Often the simplest approach is to use paramagnetic relaxation enhancement (PRE) (69–71), whereby nitroxide spin labels are attached to one of the components and intermolecular distances between the unpaired electron of the spin label and the NMR probe, in this case a methyl group of interest, are measured (Figure 1d ). Rosenzweig et al. (68) measured the distances between the spin labels attached to ClpB (one at a time) and the methyl
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
ARI
3 May 2014
11:18
group protons of 2 H,ILVM–13 CH3 –labeled DnaK by taking differences in 1 H transverse relaxation rates obtained with either oxidized (PRE-active) or reduced (PRE-inactive) spin labels (Figure 1d ) through the use of methyl 1 H transverse relaxation experiments that employed the methyl-TROSY approach. The distances so obtained, along with chemicalshift perturbation data (Figure 1a,c), were used as restraints in the HADDOCK molecular docking program (72, 73) to generate the first structural model of the ClpB/DnaK complex (Figure 1e). On the basis of this structure, which showed that ClpB binds to a similar region of DnaK [the nucleotide binding domain (NBD) (Figure 1b)] as the DnaK cochaperone, GrpE, the authors performed additional methyl-TROSY binding experiments. These studies revealed that ClpB and GrpE do indeed compete for DnaK binding. Biochemical assays motivated by the structural analysis described above showed that GrpE is required only for substrate release downstream of ClpB/DnaKmediated disaggregation and that the presence of GrpE is detrimental to the disaggregation reaction. Subsequently, the authors found that ClpB by itself can release aggregated substrates from DnaK and that the binding of DnaK to ClpB activates the process. This research emphasizes the synergy between a combined biochemical, methyl-TROSY NMR approach, in this case to elucidate the mechanism for aggregate reactivation in the cell (Figure 1f ).
Molecular Basis of Synaptic Vesicle Priming Neuronal communication is mediated through the regulated release of neurotransmitters, facilitating communication between adjacent neurons and ultimately controlling the function of the nervous system. At the terminal end of the nerve, neurotransmitter-filled synaptic vesicles dock to a specialized region of the presynaptic plasma membrane and then undergo a maturation process known as priming, after which they become fusion competent. In response to a Ca+2 influx, primed vesicles undergo rapid
exocytosis with the plasma membrane, thereby releasing neurotransmitters that lead to signal propagation (77, 78). In elegant studies, Rizo and colleagues (79) elucidated the molecular details of part of this process, in part through the use of methyl-TROSY-based NMR methods. Synaptic vesicle exocytosis requires three neuronal soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins: syntaxin-1, synaptobrevin, and the 25kDa synaptosome-associated protein (SNAP25). Prior to fusion, these SNAREs assemble into a complex through the formation of a stable four-helix bundle involving each of their C-terminal SNARE motifs in a process that brings the vesicle and the plasma membranes together in preparation for fusion (78). In addition to SNAREs, another obligatory component of the fusion machinery, Munc18, interacts selectively with syntaxin-1 in the closed conformation (Figure 2a) (80–82), a default configuration of syntaxin-1 that prevents formation of the SNARE complex. Opening of syntaxin-1 and the SNARE complex assembly, which are critical for vesicle priming, may be promoted by another protein named Munc13 (83, 84). To understand the mechanism of syntaxin-1 opening and the function of Munc13 in this process, Rizo and colleagues (79) conducted a series of binding experiments between the 72-kDa Munc13-1 MUN domain and components of the exocytosis machinery. Similarly to studies of disaggregation described above, these authors monitored the interactions in this system by recording 13 C–1 H HMQC spectra, with one or more of the involved proteins labeled by 2 H,ILV–13 CH3 . MUN interacted weakly with Munc18-1 (Kd ≥ 150 μM), the SNARE complex (Kd = 35 ± 10 μM), and the ∼130-kDa SNARE-Munc18-1 complex (Kd = 13 ± 4 μM) (79), corroborating prior observations that Munc18 can bind to the assembled SNARE complex containing open syntaxin-1 (85, 86). Moreover, because NMR provides site-specific information even for such weak interactions, the authors mapped the www.annualreviews.org • Dynamic Molecular Machines
295
3 May 2014
11:18
a
b 23.0 23.2 23.4 23.6 23.8
c 13C (ppm)
L48 L280
L106 1.0
0.9
Nucleotide binding Substrate binding domain domain
1H (ppm)
d
DnaK + ClpB reduced spin label ClpB 502C spin labeled ClpB 479C spin labeled ClpB 222C spin labeled
Intensity
i
ii
24.5
0.08
L486
25.0 25.5 1.2
1.1
1.0
0.9
1H (ppm)
V473
0.06 0.04
L486
0.02
Kd = 24.9 μM ± 3.4 μM
0
0
200
400
DnaK (μM)
e
13CH –ILVM–labeled 3
1.2 1 0.8 0.6 0.4 0.2 0 0
priming was proposed in which the weak MUN/syntaxin-1 SNARE motif interaction (Kd = 46 ± 20 μM) enhances the opening of syntaxin-1, thereby accelerating SNARE complex formation. By exploiting the fact that many of the syntaxin-1 methyl residues have different chemical shifts when syntaxin-1 is bound to Munc18-1 (closed syntaxin-1) or when incorporated into the SNARE complex (open syntaxin-1) (Figure 2b), the rate of transition between these two states was obtained both with and without the MUN domain by monitoring the decrease of syntaxin-1 resonances from the closed state and the concomitant buildup of cross-peaks derived from the open
Δδ (ppm)
ARI
binding of MUN to the C-terminal SNARE motif of syntaxin-1, and not to the N-terminal domain (Figure 2a), which had previously been thought to be the binding site. Note that in extensive studies that included gel filtration, isothermal titration calorimetry, and pull-down assays, the authors did not observe interactions between MUN and the SNAREs, emphasizing the important role that solution NMR can play in studies of large complexes, now that the appropriate tools are becoming available. On the basis of the results described above and additional experiments, a revised mechanism for Munc13-1 function in vesicle
13C (ppm)
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
DnaK ClpB 479
DnaK L288
DnaK L288
ClpB 502
0.01 0.02 0.03 0.04 0.05
ClpB
Time (s)
f
i Aggregate-bound DnaK
Rosenzweig
iii
ATP ADP Pi
ATP ADP Pi
DnaK refolding system Native substrate
ClpB
296
ii
·
Kay
600
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
ARI
3 May 2014
11:18
state (Figure 2c,d). The acceleration due to the MUN domain was significant. Additional experiments monitoring the transition between the Munc18-1/closed syntaxin-1 complex and the open SNARE complex were performed in real time by use of a series of 10-min 13 C–1 H methyl-HMQC spectra (Figure 2e). That the Munc13-1 MUN domain accelerates the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex (Figure 2f ) is notable, especially because it binds relatively weakly to either Munc18-1 or to the SNAREs, coupled with the strong (∼1-nM) interaction between syntaxin-1 and Munc18-1. As Rizo and colleagues make clear, tight protein complexes may provide ideal starting and final points, but weak interactions can often be critical in generating transitions between them that are necessary to regulate biochemical processes.
Binding Mechanism of the HMGN2 Protein to the Nucleosome Core Particle Chromatin structure and function are regulated by numerous proteins through specific binding to nucleosomes. Among them are the high-mobility group nucleosomal (HMGN) proteins, a group of chromatin factors that modulate transcription (87). A recent methylTROSY NMR study of the interaction between the HMGN2 NBD and the 210-kDa nucleosome core particle (NCP) has shed light on the binding mechanism (88). A series of 13 C–1 H HMQC experiments were recorded on NCP samples in which one of the four histones was deuterated and ILV–13 CH3 labeled and the remaining proteins were fully deuterated. Through the monitoring of chemical-shift changes, one histone at a time, the binding interface between the HMGN2 NBD and the
←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 1 NMR studies of the ClpB/DnaK complex. (a) NMR titration of DnaK (70 kDa) that is highly deuterated and selectively 13 CH3 labeled at the Ile-δ1, Leu-δ, Val-γ, and Met-ε positions (2 H,ILVM–13 CH3 ), with 2 H and hexameric ClpB (580 kDa). A series of 13 C–1 H heteronuclear multiple quantum correlation data sets were recorded in which a fixed concentration of DnaK was titrated with increasing concentrations of ClpB. In the selected region of the resulting 13 C–1 H correlation spectra, ClpB concentration–dependent changes (arrows) are highlighted for L48 and L280 of DnaK. Blue spectra indicate no ClpB, and cyan to red spectra indicate 0.2 to 2.0 ClpB in molar equivalents. (b) DnaK structure. The nucleotide binding domain is shown in cyan, the substrate binding domain in gray, and the lid section of the substrate binding domain in green. Residues with concentration-dependent chemical-shift changes due to ClpB binding are colored orange. (c) Chemical-shift titration profiles can be used to calculate dissociation constants, as illustrated for (i ) 2 H,ILVM–13 CH3 –labeled ClpB L486, color-coded as in panel a as a function of increasing concentrations of added 2 H DnaK. (ii ) Measured chemical-shift changes at each titration point are plotted as a function of DnaK concentration (ClpB L486 and V473) and fitted to a one ClpB hexamer–to–one DnaK binding model (solid line), resulting in a dissociation constant (Kd ) of 24.9 ± 3.4 μM for the ClpB/DnaK complex. (d ) Intermolecular ClpB/DnaK long-range distances from paramagnetic relaxation enhancement (PRE) data. A complex between 2 H,ILVM–13 CH3 –labeled DnaK and 2 H ClpB, with a nitroxide spin label attached to one of several selected residues, was produced. Distances were calculated from differences in methyl 1 H spin–spin relaxation rates measured from samples in which the nitroxide group is oxidized or reduced, as described elsewhere (69–71). The 1 H relaxation decay profiles for DnaK L288 complexed with ClpB, to which is attached either a reduced (inactive) spin label (red curve) or spin labels at position 502 (blue), 222 (orange), or 479 ( green). The red, orange, and green curves decay at the same rate, indicating that the distances between DnaK L288 and the nitroxide groups localized to positions 222 and 479 of ClpB are greater than 25 A˚. In contrast, the DnaK L288 relaxation rates significantly increase when a spin label is attached to ClpB residue 502 (red curve versus blue curve), and the distance between ClpB L502C and DnaK L288 is ∼19 A˚. (e) Distances measured in PRE experiments for ClpB/DnaK complexes. In the ClpB/DnaK structure, ClpB is colored yellow, DnaK light purple (spin-label positions 479 and 502 on ClpB are colored green and dark blue, respectively), and DnaK L288 red. The orange dashed lines denote the intermolecular distances (
ARI
3 May 2014
ANNUAL REVIEWS
11:18
Further
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
Click here for quick links to Annual Reviews content online, including: • Other articles in this volume • Top cited articles • Top downloaded articles • Our comprehensive search
Bringing Dynamic Molecular Machines into Focus by Methyl-TROSY NMR Rina Rosenzweig1,2,3 and Lewis E. Kay1,2,3,4 Departments of 1 Biochemistry and 3 Molecular Genetics, University of Toronto, Toronto, Ontario M5S 1A8, Canada 2
Department of Chemistry, University of Toronto, Toronto, Ontario M5S 3H6, Canada
4
Program in Molecular Structure and Function, Hospital for Sick Children, Toronto, Ontario M5G 1X8, Canada
Annu. Rev. Biochem. 2014. 83:291–315 The Annual Review of Biochemistry is online at biochem.annualreviews.org This article’s doi: 10.1146/annurev-biochem-060713-035829 c 2014 by Annual Reviews. Copyright All rights reserved
Keywords high–molecular weight complexes, protein NMR, allostery, protein–substrate interactions, methyl labeling, deuteration
Abstract Large macromolecular assemblies, so-called molecular machines, are critical to ensuring proper cellular function. Understanding how proper function is achieved at the atomic level is crucial to advancing multiple avenues of biomedical research. Biophysical studies often include X-ray diffraction and cryo–electron microscopy, providing detailed structural descriptions of these machines. However, their inherent flexibility has complicated an understanding of the relation between structure and function. Solution NMR spectroscopy is well suited to the study of such dynamic complexes, and continued developments have increased size boundaries; insights into function have been obtained for complexes with masses as large as 1 MDa. We highlight methyl-TROSY (transverse relaxation optimized spectroscopy) NMR, which enables the study of such large systems, and include examples of applications to several cellular machines. We show how this emerging technique contributes to an understanding of cellular function and the role of molecular plasticity in regulating an array of biochemical activities.
291
BI83CH13-Kay
ARI
3 May 2014
11:18
Contents
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
INTRODUCTION . . . . . . . . . . . . . . . . . . NMR STUDIES OF SUPRAMOLECULAR SYSTEMS: TECHNICAL ASPECTS . . . . . . . . . BINDING AND MECHANISM . . . . . Protein Disaggregation . . . . . . . . . . . . . Molecular Basis of Synaptic Vesicle Priming . . . . . . . . . . . . . . . . . . . . . . . . Binding Mechanism of the HMGN2 Protein to the Nucleosome Core Particle . . . . . . . . . . . . . . . . . . . . . . . . . Measurement of Active-Site pKa Values in the Proteasome. . . . . . . . Other Systems . . . . . . . . . . . . . . . . . . . . . DYNAMICS AND ALLOSTERY . . . . Proteasome Degradation Machinery . . . . . . . . . . . . . . . . . . . . . . Mapping the Energy Landscape of the Allosteric Hsp70 Chaperone . . . . . . . . . . . . . . . . . . . . . KcsA Potassium-Channel Dynamics . . . . . . . . . . . . . . . . . . . . . . Exosome Dynamics . . . . . . . . . . . . . . . . CONCLUDING REMARKS . . . . . . . . .
292
293 293 294 295
297 298 300 301 301
305 307 308 309
INTRODUCTION Solution NMR spectroscopy has become a powerful tool for the study of protein interactions and dynamics; it provides atomicresolution structural information under near-physiological conditions (1–3). Although most applications during the past two decades have focused on relatively small protein systems with molecular weights within ∼50 kDa, in the past several years there has been significant interest in the use of solution NMR spectroscopy for studies of high–molecular weight protein complexes. This interest has been fueled by the tremendous potential of NMR in studies of molecular dynamics and by the realization of the importance of such dynamics to biochemical function (4–6).
292
Rosenzweig
·
Kay
Accordingly, investigators have expended great effort in developing new isotope labeling strategies (7–9) and complementary NMR experiments that enable the preservation of NMR signals that would otherwise decay rapidly for large-protein systems (10, 11). Studies of protein complexes by solution NMR are, in fact, not new at all. Examples from as early as the 1970s and 1980s include hemoglobin (12, 13), alkaline phosphatase (14, 15), dehydrogenases (16), serine proteases (17, 18), aspartate transcarbamylase (19–23), aminotransferases (24, 25), and various muscle proteins (26–30). In these systems, NMR was used to learn about mechanisms, and the questions addressed often could be answered either by the addition of 19 F, 13 C, or various metal probes or by focusing on shifted proton resonances in cases of paramagnetic proteins. Current research efforts are building on the same fundamentals as these early pioneering studies, with a continued emphasis on understanding biochemical function. Today, however, it is possible to focus on much larger systems with molecular weights reaching 1 MDa and to provide a more atomistic view based on a larger number of probes. The objective of these studies is to address specific questions within the broader context of existing biophysical and biochemical data while using the unique strengths of NMR spectroscopy. These advantages include sensitivity to molecular interactions over a broad spectrum of affinities, including those in the millimolar-tomicromolar range that often are not quantifiable using other methods (31), and the ability to probe functional molecular dynamics over many orders of magnitude in timescale (4, 32). In what follows, we briefly discuss both the methyl-labeling approach that has been developed for studies of supramolecular systems and the experiments that are designed to take advantage of the labeling (33). We illustrate the utility of the methodology through discussion of several examples that focus on important molecular machines and demonstrate the significance of structural plasticity to function.
BI83CH13-Kay
ARI
3 May 2014
11:18
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
NMR STUDIES OF SUPRAMOLECULAR SYSTEMS: TECHNICAL ASPECTS Central to the success of any NMR study is the appropriate choice of labeling. For applications to high–molecular weight protein systems, an important and general approach involves the use of methyl group probes (13 CH3 ) in an otherwise highly deuterated environment (8, 34–45). Selection of this labeling scheme is motivated not only by the fact that methyl groups are prevalent in hydrophobic cores and at molecular interfaces, thereby serving as reporters of internal dynamics and structure (46), but also by the high quality of the spectra obtained for such groups. Excellent spectral sensitivity results from the three equivalent protons in each methyl, coupled with the rapid rotation of the methyl moiety about its threefold symmetry axis, and its localization to flexible ends of side chains. Moreover, the inherent spin physics of a methyl group facilitates the preservation of NMR signals, even in high–molecular weight complexes, via a so-called methyl-TROSY (transverse relaxation optimized spectroscopy) effect that manifests in 13 C–1 H heteronuclear multiple quantum correlation (HMQC) spectra (11). Another important factor is that it is now possible to purchase precursors for 13 CH3 labeling of any methyl position in the six methylcontaining amino acids in a protein that is otherwise highly deuterated. Labeling of methyl groups as 13 CH2 D and 13 CHD2 is also possible (47), although applications involving these isotopomers are not described here. The requisite precursors are simply added to the growth media, prior to the induction of protein overexpression, to efficiently produce molecules with the desired methyl labeling (35, 36). Note that in the recording of optimal 13 C–1 H correlation spectra, which form the basis of any analysis (see the examples provided throughout this review), one must consider the trade-off between (a) using a sufficiently large number of probes so as to achieve good coverage over the complete protein sequence and (b) obtaining high-resolution and high-sensitivity spectra. As the number of
protons increases, spectra quality decreases because the magnetic fields produced by protons from one methyl group decrease the lifetimes of signals from adjacent methyl probes, leading to line broadening. With this trade-off in mind, labeling of Ileδ1, Leu, and Val methyl groups is often a good compromise (8). In this case, the prochiral methyl groups of a given Leu or Val side chain are labeled 13 CH3 ,12 CD3 by use of either precursors that are stereospecifically pro-R or proS 13 CH3 labeled (43) or precursors in which one methyl is labeled 13 CH3 and the other 12 CD3 in a random manner (8). Finally, in cases in which methyl groups are not localized to regions of interest, they can be “engineered in” through mutation of a residue to cysteine, followed by labeling with 13 CH3 −S−, producing a Met-like side chain (48). Details about protein labeling and expression (49, 50), as well as a discussion of the spin physics of a 13 CH3 methyl group attached to a large macromolecule that is exploited in methyl-TROSY NMR (11, 51, 52), can be found in several previous publications.
BINDING AND MECHANISM One of the most widespread applications of solution bio-NMR spectroscopy involves the study of proteins with their binding partners. This reflects that the binding of macromolecules to their targets is critically important for biological function and that NMR is a powerful technique for characterizing interactions structurally and for obtaining the thermodynamic and kinetic parameters that describe the binding process. Moreover, NMR can be extremely sensitive to binding events over a millimolar-to-nanomolar range of affinities and for submicrosecond-to-second timescales (31, 32). Not surprisingly, the literature is full of examples (3); the vast majority of them focus on protein systems with molecular masses lower than 50 kDa, for which traditional NMR methods are most effective (3, 31, 53– 55). However, although the average molecular www.annualreviews.org • Dynamic Molecular Machines
293
ARI
3 May 2014
11:18
mass of a eukaryotic protein is approximately 40 kDa (56), still within the range of traditional NMR, 70–80% of functional proteins are assembled into much larger oligomers that comprise several monomeric subunits (57, 58). For these systems, the preparation of highly deuterated protein(s) with methyl protonation and the use of methyl-TROSY approaches are necessary. We describe numerous applications below to illustrate how detailed mechanistic information may be obtained through simple methyl correlation spectra, thereby increasing our understanding of how important supramolecular complexes function.
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
Protein Disaggregation The aggregation of misfolded proteins can perturb protein homeostasis and cellular function, sometimes causing disease (59). Not surprisingly, cells have developed robust protein quality–control systems to counteract protein aggregation, both under normal physiological conditions and during stress. The heat-shock protein ClpB (or Hsp100) is the major protein disaggregase in yeast, bacteria, plants, and mitochondria of all eukaryotic cells, and it is essential for cell survival during severe stress (60–63). Although ClpB is an active ATPase in which ATP hydrolysis is critical for protein disaggregation, the recovery of functional proteins from aggregates requires synergistic interaction with a second ATP-dependent molecular chaperone, DnaK (60, 64–67); together, they form a powerful bichaperone system. Structural characterization of the ClpB/DnaK complex has been hindered, however, both by the low affinity of the complex and by the dynamic nature of its components. Thus, despite increasing biochemical data (59), only a partial understanding of the overall mechanism of protein disaggregation has been obtained. Rosenzweig et al. (68) used methyl-TROSY NMR methods to study the interaction between the 580-kDa ClpB hexamer and the 70-kDa DnaK chaperone. The binding interface for each of the two proteins in the 294
Rosenzweig
·
Kay
complex was elucidated using a series of NMR titrations wherein one of the components was selectively 13 CH3 labeled at the Ile-δ1, Leu-δ, Val-γ, and Met-ε positions in an otherwise fully deuterated background (referred to as 2 H,ILVM–13 CH3 labeling). Figure 1a shows a selected region from a series of 13 C–1 H HMQC spectra of labeled DnaK recorded with increasing concentrations of a fully deuterated (NMR-invisible) ClpB hexamer. The figure illustrates changes in chemical shifts of DnaK residues L280 and L48 upon complex formation; all residues whose methyl groups showed spectral changes were subsequently mapped onto the structure of DnaK to determine the binding interface for ClpB (Figure 1b). Similarly, the authors obtained the corresponding DnaK binding interface on ClpB by repeating the experiment with the labeling scheme reversed. In addition to mapping of the ClpB/DnaK complex, changes in chemical shifts upon binding were also used to calculate the dissociation constant (Kd ) for the complex. Chemical-shift changes for residues in the DnaK binding surface of ClpB were measured at each titration point and fitted to a one-to-one binding model (Figure 1c) to obtain a Kd of 25 ± 3 μM for the ClpB/DnaK association (68). Although such low affinities are typically very difficult to quantify by other biophysical techniques, NMR spectroscopy remains a powerful way to study complexes involving weak and/or transient interactions. Beyond localizing binding sites, detailed structural information about the complex can be obtained when structural models of the individual molecular players are available. Often the simplest approach is to use paramagnetic relaxation enhancement (PRE) (69–71), whereby nitroxide spin labels are attached to one of the components and intermolecular distances between the unpaired electron of the spin label and the NMR probe, in this case a methyl group of interest, are measured (Figure 1d ). Rosenzweig et al. (68) measured the distances between the spin labels attached to ClpB (one at a time) and the methyl
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
ARI
3 May 2014
11:18
group protons of 2 H,ILVM–13 CH3 –labeled DnaK by taking differences in 1 H transverse relaxation rates obtained with either oxidized (PRE-active) or reduced (PRE-inactive) spin labels (Figure 1d ) through the use of methyl 1 H transverse relaxation experiments that employed the methyl-TROSY approach. The distances so obtained, along with chemicalshift perturbation data (Figure 1a,c), were used as restraints in the HADDOCK molecular docking program (72, 73) to generate the first structural model of the ClpB/DnaK complex (Figure 1e). On the basis of this structure, which showed that ClpB binds to a similar region of DnaK [the nucleotide binding domain (NBD) (Figure 1b)] as the DnaK cochaperone, GrpE, the authors performed additional methyl-TROSY binding experiments. These studies revealed that ClpB and GrpE do indeed compete for DnaK binding. Biochemical assays motivated by the structural analysis described above showed that GrpE is required only for substrate release downstream of ClpB/DnaKmediated disaggregation and that the presence of GrpE is detrimental to the disaggregation reaction. Subsequently, the authors found that ClpB by itself can release aggregated substrates from DnaK and that the binding of DnaK to ClpB activates the process. This research emphasizes the synergy between a combined biochemical, methyl-TROSY NMR approach, in this case to elucidate the mechanism for aggregate reactivation in the cell (Figure 1f ).
Molecular Basis of Synaptic Vesicle Priming Neuronal communication is mediated through the regulated release of neurotransmitters, facilitating communication between adjacent neurons and ultimately controlling the function of the nervous system. At the terminal end of the nerve, neurotransmitter-filled synaptic vesicles dock to a specialized region of the presynaptic plasma membrane and then undergo a maturation process known as priming, after which they become fusion competent. In response to a Ca+2 influx, primed vesicles undergo rapid
exocytosis with the plasma membrane, thereby releasing neurotransmitters that lead to signal propagation (77, 78). In elegant studies, Rizo and colleagues (79) elucidated the molecular details of part of this process, in part through the use of methyl-TROSY-based NMR methods. Synaptic vesicle exocytosis requires three neuronal soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins: syntaxin-1, synaptobrevin, and the 25kDa synaptosome-associated protein (SNAP25). Prior to fusion, these SNAREs assemble into a complex through the formation of a stable four-helix bundle involving each of their C-terminal SNARE motifs in a process that brings the vesicle and the plasma membranes together in preparation for fusion (78). In addition to SNAREs, another obligatory component of the fusion machinery, Munc18, interacts selectively with syntaxin-1 in the closed conformation (Figure 2a) (80–82), a default configuration of syntaxin-1 that prevents formation of the SNARE complex. Opening of syntaxin-1 and the SNARE complex assembly, which are critical for vesicle priming, may be promoted by another protein named Munc13 (83, 84). To understand the mechanism of syntaxin-1 opening and the function of Munc13 in this process, Rizo and colleagues (79) conducted a series of binding experiments between the 72-kDa Munc13-1 MUN domain and components of the exocytosis machinery. Similarly to studies of disaggregation described above, these authors monitored the interactions in this system by recording 13 C–1 H HMQC spectra, with one or more of the involved proteins labeled by 2 H,ILV–13 CH3 . MUN interacted weakly with Munc18-1 (Kd ≥ 150 μM), the SNARE complex (Kd = 35 ± 10 μM), and the ∼130-kDa SNARE-Munc18-1 complex (Kd = 13 ± 4 μM) (79), corroborating prior observations that Munc18 can bind to the assembled SNARE complex containing open syntaxin-1 (85, 86). Moreover, because NMR provides site-specific information even for such weak interactions, the authors mapped the www.annualreviews.org • Dynamic Molecular Machines
295
3 May 2014
11:18
a
b 23.0 23.2 23.4 23.6 23.8
c 13C (ppm)
L48 L280
L106 1.0
0.9
Nucleotide binding Substrate binding domain domain
1H (ppm)
d
DnaK + ClpB reduced spin label ClpB 502C spin labeled ClpB 479C spin labeled ClpB 222C spin labeled
Intensity
i
ii
24.5
0.08
L486
25.0 25.5 1.2
1.1
1.0
0.9
1H (ppm)
V473
0.06 0.04
L486
0.02
Kd = 24.9 μM ± 3.4 μM
0
0
200
400
DnaK (μM)
e
13CH –ILVM–labeled 3
1.2 1 0.8 0.6 0.4 0.2 0 0
priming was proposed in which the weak MUN/syntaxin-1 SNARE motif interaction (Kd = 46 ± 20 μM) enhances the opening of syntaxin-1, thereby accelerating SNARE complex formation. By exploiting the fact that many of the syntaxin-1 methyl residues have different chemical shifts when syntaxin-1 is bound to Munc18-1 (closed syntaxin-1) or when incorporated into the SNARE complex (open syntaxin-1) (Figure 2b), the rate of transition between these two states was obtained both with and without the MUN domain by monitoring the decrease of syntaxin-1 resonances from the closed state and the concomitant buildup of cross-peaks derived from the open
Δδ (ppm)
ARI
binding of MUN to the C-terminal SNARE motif of syntaxin-1, and not to the N-terminal domain (Figure 2a), which had previously been thought to be the binding site. Note that in extensive studies that included gel filtration, isothermal titration calorimetry, and pull-down assays, the authors did not observe interactions between MUN and the SNAREs, emphasizing the important role that solution NMR can play in studies of large complexes, now that the appropriate tools are becoming available. On the basis of the results described above and additional experiments, a revised mechanism for Munc13-1 function in vesicle
13C (ppm)
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
DnaK ClpB 479
DnaK L288
DnaK L288
ClpB 502
0.01 0.02 0.03 0.04 0.05
ClpB
Time (s)
f
i Aggregate-bound DnaK
Rosenzweig
iii
ATP ADP Pi
ATP ADP Pi
DnaK refolding system Native substrate
ClpB
296
ii
·
Kay
600
Annu. Rev. Biochem. 2014.83:291-315. Downloaded from www.annualreviews.org by University of Laval on 07/15/14. For personal use only.
BI83CH13-Kay
ARI
3 May 2014
11:18
state (Figure 2c,d). The acceleration due to the MUN domain was significant. Additional experiments monitoring the transition between the Munc18-1/closed syntaxin-1 complex and the open SNARE complex were performed in real time by use of a series of 10-min 13 C–1 H methyl-HMQC spectra (Figure 2e). That the Munc13-1 MUN domain accelerates the transition from the syntaxin-1/Munc18-1 complex to the SNARE complex (Figure 2f ) is notable, especially because it binds relatively weakly to either Munc18-1 or to the SNAREs, coupled with the strong (∼1-nM) interaction between syntaxin-1 and Munc18-1. As Rizo and colleagues make clear, tight protein complexes may provide ideal starting and final points, but weak interactions can often be critical in generating transitions between them that are necessary to regulate biochemical processes.
Binding Mechanism of the HMGN2 Protein to the Nucleosome Core Particle Chromatin structure and function are regulated by numerous proteins through specific binding to nucleosomes. Among them are the high-mobility group nucleosomal (HMGN) proteins, a group of chromatin factors that modulate transcription (87). A recent methylTROSY NMR study of the interaction between the HMGN2 NBD and the 210-kDa nucleosome core particle (NCP) has shed light on the binding mechanism (88). A series of 13 C–1 H HMQC experiments were recorded on NCP samples in which one of the four histones was deuterated and ILV–13 CH3 labeled and the remaining proteins were fully deuterated. Through the monitoring of chemical-shift changes, one histone at a time, the binding interface between the HMGN2 NBD and the
←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− Figure 1 NMR studies of the ClpB/DnaK complex. (a) NMR titration of DnaK (70 kDa) that is highly deuterated and selectively 13 CH3 labeled at the Ile-δ1, Leu-δ, Val-γ, and Met-ε positions (2 H,ILVM–13 CH3 ), with 2 H and hexameric ClpB (580 kDa). A series of 13 C–1 H heteronuclear multiple quantum correlation data sets were recorded in which a fixed concentration of DnaK was titrated with increasing concentrations of ClpB. In the selected region of the resulting 13 C–1 H correlation spectra, ClpB concentration–dependent changes (arrows) are highlighted for L48 and L280 of DnaK. Blue spectra indicate no ClpB, and cyan to red spectra indicate 0.2 to 2.0 ClpB in molar equivalents. (b) DnaK structure. The nucleotide binding domain is shown in cyan, the substrate binding domain in gray, and the lid section of the substrate binding domain in green. Residues with concentration-dependent chemical-shift changes due to ClpB binding are colored orange. (c) Chemical-shift titration profiles can be used to calculate dissociation constants, as illustrated for (i ) 2 H,ILVM–13 CH3 –labeled ClpB L486, color-coded as in panel a as a function of increasing concentrations of added 2 H DnaK. (ii ) Measured chemical-shift changes at each titration point are plotted as a function of DnaK concentration (ClpB L486 and V473) and fitted to a one ClpB hexamer–to–one DnaK binding model (solid line), resulting in a dissociation constant (Kd ) of 24.9 ± 3.4 μM for the ClpB/DnaK complex. (d ) Intermolecular ClpB/DnaK long-range distances from paramagnetic relaxation enhancement (PRE) data. A complex between 2 H,ILVM–13 CH3 –labeled DnaK and 2 H ClpB, with a nitroxide spin label attached to one of several selected residues, was produced. Distances were calculated from differences in methyl 1 H spin–spin relaxation rates measured from samples in which the nitroxide group is oxidized or reduced, as described elsewhere (69–71). The 1 H relaxation decay profiles for DnaK L288 complexed with ClpB, to which is attached either a reduced (inactive) spin label (red curve) or spin labels at position 502 (blue), 222 (orange), or 479 ( green). The red, orange, and green curves decay at the same rate, indicating that the distances between DnaK L288 and the nitroxide groups localized to positions 222 and 479 of ClpB are greater than 25 A˚. In contrast, the DnaK L288 relaxation rates significantly increase when a spin label is attached to ClpB residue 502 (red curve versus blue curve), and the distance between ClpB L502C and DnaK L288 is ∼19 A˚. (e) Distances measured in PRE experiments for ClpB/DnaK complexes. In the ClpB/DnaK structure, ClpB is colored yellow, DnaK light purple (spin-label positions 479 and 502 on ClpB are colored green and dark blue, respectively), and DnaK L288 red. The orange dashed lines denote the intermolecular distances (
