Atomic-Level Description of Protein Folding inside the GroEL Cavity

Subscriber access provided by University of Sunderland

Article

Atomic-Level Description of Protein Folding inside the GroEL Cavity Stefano Piana, and David E. Shaw J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.8b07366 • Publication Date (Web): 02 Oct 2018 Downloaded from http://pubs.acs.org on October 4, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Atomic-Level Description of Protein Folding inside the GroEL Cavity

Stefano Piana1,* and David E. Shaw1,2,*

1

2

D. E. Shaw Research, New York, NY 10036, USA

Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY 10032, USA

* To whom correspondence should be addressed. Stefano Piana E-mail:

[email protected]

Phone:

(212) 403-8165

Fax:

(646) 873-2165

David E. Shaw E-mail:

[email protected]

Phone:

(212) 478-0260

Fax:

(212) 845-1286

ACS Paragon Plus Environment

1

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 35

Abstract Chaperonins—ubiquitous facilitators of protein folding—sequester misfolded proteins within an internal cavity, thus preventing protein aggregation during the process of refolding. GroEL, a tetradecameric bacterial chaperonin, is one of the most studied chaperonins, but the role of the internal cavity in the refolding process is still unclear. It has been suggested that, rather than simply isolating proteins while they refold, the GroEL cavity actively promotes protein folding. A detailed characterization of the folding dynamics and thermodynamics of protein substrates encapsulated within the cavity, however, has been difficult to obtain by experimental means, due to the system’s complexity and the many steps in the folding cycle. Here, we examine the influence of the GroEL cavity on protein folding based on the results of unbiased, atomistic molecular dynamics simulations. We first verified that the computational setup, which uses a recently developed state-of-the-art force field that more accurately reproduces the aggregation propensity of unfolded states, could recapitulate the essential structural dynamics of GroEL. In these simulations, the GroEL tetradecamer was highly dynamic, transitioning among states corresponding to most of the structures that have been observed experimentally. We then simulated a small, unfolded protein both in the GroEL cavity and in bulk solution, and compared the protein’s folding process within these two environments. Inside the GroEL cavity, the unfolded protein interacted strongly with the disordered residues in GroEL’s C-terminal tails. These interactions stabilized the protein’s unfolded states relative to its compact states and increased the roughness of its folding free-energy surface, resulting in slower folding compared to the rate in solution. For larger proteins, which are more typical GroEL substrates, we speculate that these interactions may allow substrates to more quickly escape kinetic traps associated with compact, misfolded states, thereby actively promoting folding.


2

Page 3 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Chaperonins—essential proteins1,2 common to all living organisms—are key components of the cellular machinery that facilitates the proper folding of complex, multidomain proteins and helps maintain protein homeostasis.3,4 They are generally composed of 14–18 subunits arranged in two rings.5 Each ring forms a cavity in which an unfolded or misfolded substrate is sequestered, allowing it to fold or refold, while avoiding aggregation with other proteins.6,7 ATP hydrolysis provides the energy required for opening and closing the cavity,8–11 and for capturing and releasing the substrate8,12,13 during the many steps of the catalytic cycle.

The Escherichia coli GroEL is one of the most studied chaperonins, as it is normally found in the cytosol14 and can be produced in large amounts and easily purified.15,16 GroEL is an oligomer composed of 14 identical subunits arranged in two symmetric rings,17 with each subunit composed of 546 amino acids. Its tertiary structure consists of an apical and an equatorial domain, separated by an intermediate region. Protein substrates initially bind to hydrophobic patches on the apical domain that line the interior of the central cavity.18 ATP binding to the equatorial domain induces a large allosteric conformational change in the apical domain that enables the binding of the cochaperonin GroES and the locking of the substrate within the central cavity. This conformational change is reversed upon ATP hydrolysis, and GroES and the protein substrate are then released.19 The slow rate of hydrolysis (~0.1 s−1) allows sufficient time for the substrate to fold in the isolation of the GroEL cavity, preventing aggregation.20

There is some debate about the role played by the GroEL cavity, if any, in actively assisting protein folding and refolding. A number of studies employing a range of substrates and experimental techniques have been performed to assess the influence of the GroEL cavity on protein folding. In some studies, GroEL appears to have only limited influence on the folding mechanism,21,22 supporting the idea that the GroEL cavity acts as a passive anti-aggregation device.23 Other studies of folding in the cavity have found some enhancement in folding rates


3


Page 4 of 35

and folding efficiency compared to folding in solution.24–28 Folding rates and efficiency were found to decrease upon the mutation of multiple residues within the cavity wall, leading to the hypothesis that the GroEL cavity actively promotes folding.24,25,28,29 The interpretation of some of the data is not straightforward,23 however, since substrates in solution can be subject to varying degrees of aggregation, and since GroEL mutations can affect not just the interactions with the substrate but also a number of steps in the catalytic cycle, including ATP hydrolysis rates.30 The direct observation of protein folding inside the GroEL cavity with high spatial and temporal resolution would help shed light on the role of GroEL in the folding process. With a few notable exceptions,26,31 however, it has proven difficult to obtain a detailed structural characterization of substrates inside the GroEL cavity using experimental methods.

To examine the influence of the GroEL cavity on protein folding, we performed atomistic molecular dynamics (MD) simulations of the folding of a small protein, HP35, inside the GroEL cavity, and compared the folding rates and thermodynamics with those of the same protein folding in solution. Although GroEL substrates are typically larger and have more complex folding free-energy landscapes than does HP35, the faster folding rate of this smaller protein makes it possible to study its folding inside the GroEL cavity in fully atomistic detail with a physics-based force field. This computational setup enables a detailed description of the highly heterogeneous cavity environment—containing ordered and disordered regions, amino acids with diverse properties, ions, water, and ATP—any aspect of which could potentially influence the folding rate and mechanism of protein substrates.

To minimize potential sources of bias, we first performed simulations to verify that the computational setup—including a recently developed physics-based force field that more accurately describes the aggregation propensity of unfolded states32—was able to recapitulate the key features of the structural dynamics of the GroEL tetradecamer. In these simulations, we found that GroEL was highly dynamic, transitioning on the tens-of-microseconds timescale among most of the structures that have been observed experimentally. The GroEL


4

Page 5 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


conformational ensemble observed in simulation was also found to be consistent with smallangle X-ray scattering (SAXS) data.

Having performed this validation, we then simulated an unfolded protein inside the GroEL cavity and in bulk solution. When folding in the GroEL cavity, we observed that the unfolded substrate strongly and almost exclusively interacted with GroEL’s disordered C-terminal tails, which stabilized the substrate’s more expanded states at the expense of the most compact states and increased the roughness of (i.e., slowed down diffusion on) its folding free-energy surface. The combination of these two effects resulted in this small substrate folding over an order of magnitude more slowly than it did in solution. We speculate that the interior of the GroEL cavity has evolved to provide a mildly destabilizing environment, which may help larger proteins escape from the compact, misfolded kinetic traps that are characteristic of their more complex folding free-energy landscapes, thus actively promoting the folding of these proteins.

Results

Conformational dynamics of the GroEL tetradecamer

As a first step of our investigation, we validated our computational approach by performing simulations of the GroEL tetradecamer on the hundred-microsecond timescale and comparing the results to SAXS, cryo-EM, and X-ray structural data. A number of X-ray diffraction17,33–41 and cryo-EM11,42–45 studies have provided atomic-resolution structures of the different stages of GroEL’s catalytic cycle, shedding light on the essential ingredients of the reaction mechanism. Computational studies have complemented these experiments by investigating the mechanism of interconversion between different states,46–48 and how some of the experimentally observed structural features may be relevant for substrate binding, folding, and unfolding.29,49–51 These


5


Page 6 of 35

studies have suggested that isolated GroEL monomers might be highly flexible.52–54 Due to steric and electrostatic constraints, conformational transitions within a ring are expected to be strongly cooperative and possibly concerted.46,47 The size and complexity of the system, however, have to date made it difficult to computationally characterize the conformational dynamics of the GroEL tetradecamer with atomic resolution on the timescales at which biologically important, large-scale conformational transitions might occur. In addition to providing validation for our computational approach, our simulations thus also further our understanding of the conformational dynamics of the GroEL tetradecamer.

The X-ray structure of the GroEL-ATP14 complex (PDB entry 1SX3)36 was used to generate starting structures of apo GroEL and GroEL-ATP7. Excess ATP molecules and Mg2+ ions were removed, and the 22 missing C-terminal amino acids, including the Gly-Gly-Met repeats, were added using Maestro.55 The systems were solvated in a 200 × 200 × 200 Å water box containing 50 mM NaCl. The final system was composed of ~1 million particles. Using Anton,56 a specialpurpose machine capable of running very long MD simulations, trajectories 470 µs and 250 µs in length were obtained for apo GroEL and GroEL-ATP7, respectively, using the a99SB-disp force field32 (see Methods for further details).

The conformational ensemble observed in the apo GroEL simulation was compared to SAXS data.57 To provide a baseline for the analysis, we calculated SAXS scattering intensities for the 35 tetradecamer experimental structures present in the PDB using FoXS58,59 (Figure 1A). As observed previously,57 no single experimental structure recapitulated well the SAXS scattering intensities between 0 and 0.2 Å−1 (best χ2 of 9.2). The agreement improves if a linear combination of the 35 experimental structures with optimized coefficients is used to fit the data (χ2 = 4.6, Figure 1A). Interestingly, the optimization of the coefficients of the linear combination results in two very different structures, accounting for 99% of the weight: the cryo-EM structure of apo GroEL (PDB entry 1GRF,45 89%) and the cryo-EM structure of the Rd-open:Rd5 state (PDB entry 4AB3,11 10%), suggesting some degree of heterogeneity in the apo-GroEL


6

Page 7 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


conformational ensemble. The simulation ensemble provides a similarly reasonable model for the SAXS data (Figure 1B): A χ2 of 4.4 is obtained using the average of 35 structures taken at fixed intervals from the apo-GroEL simulation, and the deviation can be reduced to a χ2 of 1.6 upon optimization of the coefficients of the linear combination.

Visual inspection of the simulation trajectories revealed large conformational transitions in some of the monomers on the tens- to hundreds-of-microseconds timescale. A Cα-RMSD analysis showed that, while most of the monomers remained close to their original conformation, reversible transitions occasionally occurred to conformations more than 20 Å Cα RMSD from the starting structure (Figure 2). These large-scale motions generally involved the reorientation of the apical domain with respect to the equatorial domain. Conformational changes within the domains were much smaller, on the order of a few Å RMSD (Figure S1). Conformational transitions were not tightly concerted, and as a result GroEL’s 7-fold symmetry was broken in these simulations. Non-symmetric states have also been observed in X-ray35 and cryo-EM11,42 experiments, but their prevalence is difficult to assess, as symmetry is often imposed in the analysis of the experimental data.

Some degree of cooperativity was observed, however: When one subunit underwent a large conformational change, for example, a number of interactions with the adjacent subunits were broken, and as a consequence the adjacent subunits were more prone to experience similar conformational changes. Possibly as a result, most of the large conformational transitions in GroEL were localized on one of the two rings, whereas the other ring experienced a lesser degree of large-scale motions (Figure 2E, F). The amount of conformational flexibility and asymmetry was higher in GroEL-ATP7—where ATP binds to only one of the rings—a finding consistent with experimental observations.11

To characterize in detail the conformational space explored by GroEL, we performed a kineticclustering analysis of the simulations.60–62 In this analysis, the simulations of the 14 apo GroEL


7


Page 8 of 35

monomers were treated as multiple independent simulations of the same system. This approximation substantially increases the amount of available statistics, allowing for a robust partitioning of the simulation into a small number of states, but implicitly assumes that the motions of monomers are uncorrelated. A simulation several orders of magnitude longer would be necessary to collect enough statistics to go beyond this approximation (e.g., by treating nearest-neighbor correlation explicitly); we expect that this may be necessary to obtain highly accurate transition rates, but it would not change the structural characterization of the states. In the GroEL-ATP7 simulation, the two rings had markedly different dynamics (Figure 2C, D), and were analyzed separately; only the model obtained from the ATP-bound ring, which had the largest fluctuations, is described in the main text. The kinetic-clustering analysis yielded 5–7 clusters, and transitions occurred between these clusters on the 10–100 μs timescale (Figure 3A, B).

Although in most clusters the equatorial subunit was essentially rigid, substantial flexibility was observed in the apical subunit (Figure 3A, B). Despite this considerable flexibility, most of the average cluster structures closely match (RMSD < 5 Å) one or more of the structures determined by X-ray diffraction17,33–41 or cryo-EM11 (Table 1), and are representative of the variety of conformations observed in the different steps of the catalytic cycle. For apo GroEL, the only exception is a conformation observed in ~13% of the simulation time; it is unclear if this conformation was the result of residual force field inaccuracy, or if this highly flexible, lowpopulation structure has simply not been captured in any experiment to date. For several of the average cluster structures, the Cα RMSD from the cryo-EM structures is surprisingly small ( 90%, or as the set of conformations in which the Cα RMSD from the experimentally determined structure (PDB ID 2F4K) was 90%) of interactions involved the highly conserved Gly-Gly-Met repeats in the C-terminal tails (Figure 7), suggesting that these residues play a prominent role. The tails are both hydrophobic and highly disordered, two characteristics that may increase their affinity for an unfolded substrate. To assess the relative importance of these contributions, we performed additional control simulations in which we either mutated the C-terminal methionine residues to serine (GroEL-GGS) or entirely removed the last 18 Cterminal residues (GroEL-ΔC18). In the GroEL-GGS simulation, interactions with the Cterminal tail were greatly reduced (Figure 7), but still accounted for ~50% of the contacts, suggesting that both the hydrophobic character and the disordered nature of this region play a role in determining its affinity for disordered states. In both the GroEL-GGS and GroEL-ΔC18 mutants, compact unfolded states had roughly the same stability as in aqueous solution (Figure 5), and the timescale of the decay of the autocorrelation function for the SASA and Rg in the unfolded states had values intermediate


12

Page 13 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


between those of water and wild-type GroEL (Figure 6). The folding rate of HP35 in these two mutants (~200 µs in GroEL-GGS and ~130 µs in GroEL-ΔC18) also appeared to be intermediate between the value calculated in water solution (27 ± 6 µs) and that estimated in wild-type GroEL (~500 µs). These results indicate that the hydrophobic character of the C-terminal tails is critical for the destabilization of compact states, but the observation that completely removing the Cterminal tails still resulted in a ~5-fold reduction in the HP35 folding rate suggests that there may be other components of the GroEL cavity, not identified in this study, that also play a nonnegligible role in modulating folding kinetics.

Previous experimental studies have indicated that the C-terminal tail is important for preventing the escape of protein substrates27,64 and for enhancing protein folding rates, possibly through nonspecific interactions with misfolded intermediates.24,27 In our simulations, the C-terminal tails were found to interact strongly with and stabilize expanded states at the expense of compact states, including the folded state, and to slow down diffusion in conformational space such that folding inside the GroEL cavity was slower than in bulk solution for HP35. This result is consistent with previous experiments demonstrating that misfolded proteins tend to interact with these residues,24,27,75 but it runs counter to the expectation, based on experimental observations and theoretical considerations, that the folding rate inside the GroEL cavity should be faster than or equal to the rate in aqueous solution. Previous simulation76 and experimental studies22 have suggested that the overall effect of GroEL on folding rates may depend on the substrate and the type of interaction it has with the cavity. We speculate that the result obtained here may be somewhat specific to small, fast-folding protein domains characterized by a smooth energy landscape with no long-lived folding intermediates or kinetic traps. In the case of more typical GroEL substrates, for which escaping long-lived kinetic traps may be rate-limiting for folding, interaction with the C-terminal tails may destabilize misfolded states, thus increasing the rate of escape from kinetic traps and decreasing overall folding times.


13


Page 14 of 35

Methods

MD simulation setup and parameters

The X-ray structure of the GroEL-ATP14 complex (PDB entry 1SX3)36 was used to generate the starting structures of apo GroEL and GroEL-ATP7. The X-ray structure of the ATP-bound GroEL2-GroES2 complex (PDB entry 3WVL)33 was used as a starting structure to model the GroEL cavity in the HP35 folding simulations. The excess ATP molecules and Mg2+ ions in apo GroEL, GroEL-ATP7, and one of the GroEL-GroES subunits in the GroEL2-GroES2 complex were removed. The 22 missing C-terminal amino acids, including the Gly-Gly-Met repeats, were added using Maestro.55

The apo GroEL and GroEL-ATP7 systems were solvated in a 200 × 200 × 200 Å water box containing 50 mM NaCl, and the final systems were composed of ~1 million particles. For the simulations with HP35, one HP35 protein was placed in the center of the GroEL cavity, either in a folded or unfolded conformation (see below); the system was then solvated in a 160 × 160 × 160 Å water box containing 150 mM KCl. The final systems were composed of ~470,000 particles. Simulations of 150 µs and 250 µs were performed on an Anton machine56 for apo GroEL and GroEL-ATP7 using the a99SB-disp force field.32 13 simulations were performed with HP35 encapsulated within the GroEL cavity. Six simulations were started with HP35 in the unfolded state, and three simulations with HP35 in the folded state. In two of the simulations performed starting from the unfolded state of HP35, the C-terminal methionine residues were mutated to serine, and in four simulations the last 18 residues of the C-terminal tails (residues 531–548) were eliminated. Each simulation was run for a length of time ranging from 100 to 250 µs.


14

Page 15 of 35 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Prior to performing MD simulations on Anton, each system was subject to energy minimization and equilibrated at 300 K and 1 atm using Desmond,77 with harmonic restraints on the protein Cα atoms. Anton simulations were performed in the NPT ensemble using the Multigrator framework78 with a Nosé-Hoover thermostat79,80 (τrel 1.67 ps, update frequency 40 steps) and a MTK barostat (τrel 16.7 ps, update frequency 400 steps).81 The mass of all hydrogen atoms was set to 4 a.m.u. and the time step to 4.0 fs.61 As a control for this computational setup, a preliminary 32-μs run of apo GroEL was performed using normal hydrogen masses and a 2.5-fs time step. The conformational ensemble observed in this initial simulation, although far more limited, was consistent with that observed in the 470-μs simulation performed with heavy hydrogen masses and a 4-fs time step.

Analysis

SAXS scattering intensities were calculated with the FOXS software,59 allowing for offset optimization. To verify the importance of modeling the hydration sphere, we repeated the calculation either without optimization of the hydration shell parameters or using the SASMOL code.82 The differences observed were negligible, possibly because of the large size of the protein and the small scattering angles considered.

The kinetically metastable states were identified using a kinetic-clustering algorithm60,61 that minimizes the deviation between the autocorrelation functions of 105 Cα-Cα distances calculated for the kinetic model and those measured in simulation.

The ionic concentration inside the GroEL cavity was calculated by counting the number of ions and water molecules within a cylinder of 50-Å radius and 55-Å height centered in each cavity. We verified that the results do not depend strongly upon the choice of different shapes or sizes to define the cavity volume.


15


Page 16 of 35

For the purpose of the analysis of native contacts in the folding of HP35 and of contacts between HP35 and the GroEL cavity, a contact between two residues was defined as a Cα-Cα distance

Atomic-Level Description of Protein Folding inside the GroEL Cavity

Recommend Documents