Determination of an Ensemble of Structures Representing the Denatured State of the Bovine Acyl-Coenzyme A Binding Protein

Kresten Lindorff‐Larsen, Sigríður Kristjánsdóttir, Kaare Teilum, Wolfgang Fieber, Christopher M. Dobson, Flemming M. Poulsen, Michele VendruscoloJournal of the American Chemical Society 2004 · DOI

TL;DR

This paper introduces a novel computational method, Monte Carlo Replica Sampling (MCRES), to determine the structural ensembles of denatured proteins. By combining experimental NMR data with this method, the researchers reveal that the denatured state of bovine acyl-coenzyme A binding protein (ACBP) is highly heterogeneous and exhibits persistent, long-range interactions, suggesting that sequence-dependent properties play a crucial role in shaping even unfolded protein structures.

What Problem Does This Paper Address?

Understanding the structure of denatured proteins is crucial for deciphering protein folding mechanisms, stability, and even the formation of pathological aggregates like amyloid fibrils. However, denatured proteins exist as a dynamic, heterogeneous ensemble of conformations, not a single static structure. Traditional structure determination methods, which rely on interpreting experimental data as representing a single molecular structure, are ill-suited for this task. Experimental data from techniques like Nuclear Magnetic Resonance (NMR) spectroscopy, particularly Paramagnetic Relaxation Enhancement (PRE), provide distance information that is inherently averaged over this ensemble. The challenge, therefore, is to develop computational approaches that can effectively handle this ensemble averaging and extract meaningful structural information from such averaged experimental data.

Methods

The core of this paper lies in the development and application of the Monte Carlo Replica Sampling (MCRES) method. This approach extends the concept of "multiconformer refinement" by simulating multiple non-interacting copies (replicas) of the protein in parallel. Experimental restraints, derived from PRE NMR data, are then compared to the ensemble averages of distances calculated across these replicas. As the simulation progresses, each replica can explore different conformational regions, contributing to an ensemble that is compatible with the averaged experimental data.

Key aspects of the method include:

  • Ca Model: For computational efficiency, the protein is represented by a simplified Ca-only model.
  • Monte Carlo Sampling: A crank-shaft move is used to explore conformational space.
  • Ensemble Averaging: Experimental distance restraints are compared to the average distances calculated across all simulated replicas.
  • MCRES: This specific implementation allows for efficient sampling of highly heterogeneous ensembles.
  • Cross-validation: The quality of the determined ensembles was assessed by withholding subsets of experimental restraints and evaluating how well the simulations could predict them.

The researchers also employed standard techniques for protein expression, purification, and NMR data acquisition, including site-directed spin-labeling and HSQC experiments. For all-atom model generation and energy minimization, they used MAXSPROUT and CHARMM with the EEF1 potential.

Datasets

The primary dataset used in this study consists of Paramagnetic Relaxation Enhancement (PRE) data obtained from site-directed spin-labeling experiments on the bovine acyl-coenzyme A binding protein (ACBP). These experiments were performed on five different cysteine mutants of ACBP at three different concentrations of guanidine hydrochloride (GuHCl): 1.6 M, 1.9 M, and 3.0 M. The PRE data provided interresidue distance restraints, which were then used as input for the MCRES simulations.

Additionally, the researchers used:

  • 15N-1H HSQC NMR spectra to monitor protein states and analyze peak intensities.
  • Hydrodynamic radius (Rh) measurements obtained via NMR diffusion experiments to validate the overall size of the simulated ensembles.
  • High-temperature molecular dynamics (MD) simulations of ACBP to generate a reference set of conformations for comparison and to establish an empirical relationship between Rg and Rh.
  • A random-coil model (simulated with excluded volume but no specific restraints) to serve as a reference for identifying residual structure.

Key Insights and Ideas

  • Denatured states are not random coils: Contrary to the simplistic view of denatured proteins as completely disordered, this study provides strong evidence for persistent, non-random residual structure.
  • Long-range interactions persist: The most striking finding is the identification of significant long-range interactions between two specific regions of the protein (residues ~15-35 and ~60-86), even in the denatured state.
  • Sequence dictates structure, even when unfolded: These persistent interactions are attributed to the inherent sequence-dependent properties of the protein, specifically the patterns of helical propensity and hydrophobicity.
  • MCRES is a powerful tool for ensemble determination: The development and successful application of MCRES demonstrate its capability to efficiently sample and characterize highly heterogeneous conformational ensembles, overcoming a major hurdle in studying denatured proteins.
  • Native-like structural propensities in the denatured state: The identified residual structure in the denatured state appears to be related to the native structure, suggesting that these propensities are fundamental to the protein's sequence.

Key Results

  • The MCRES method, with an optimized number of 20 replicas, successfully sampled denatured state ensembles that were consistent with experimental PRE data and hydrodynamic radius measurements.
  • The calculated hydrodynamic radius of the simulated ensembles converged to a value between 25 and 26 Å across all denaturant concentrations, closely matching the experimentally determined value of 25.1 ± 0.3 Å in 3.0 M GuHCl.
  • Cross-validation analysis showed that using multiple replicas significantly improved the accuracy of predicting experimental distance restraints, with the largest gains observed when moving from one to two replicas, and continued improvement up to approximately 20 replicas.
  • Detailed analysis of the denatured state ensembles revealed persistent long-range interactions between residues in the N-terminal helical region (A2, ~15-35) and the C-terminal helical region (A4, ~60-86).
  • The extent of residual structure was found to decrease with increasing denaturant concentration, suggesting these interactions are more prevalent under less denaturing conditions.
  • The identified residual structure was found to be remarkably similar across different denaturant concentrations, and also comparable to structures observed in the acid-denatured state, indicating these features are not artifacts of the denaturant itself.

Limitations and Future Work

  • Ca Model Simplification: The use of a Ca-only model, while computationally efficient, omits detailed atomic interactions and side-chain specificities. However, the authors argue that the experimental restraints guide the sampling to biologically relevant conformations.
  • Generalizability of Nrep: The optimal number of replicas (Nrep = 20) was determined for ACBP. The authors acknowledge that this value may not generalize to all systems and suggest that methods for determining optimal Nrep should be applied on a case-by-case basis.
  • Denaturant Perturbation: While denaturants are necessary to stabilize the denatured state, they can also perturb its structure. The authors note this inherent issue but highlight the advantage of varying denaturant concentration to study these effects.

Future work suggested by the authors includes:

  • Extending the MCRES approach to other types of NMR data, such as NOEs, chemical shifts, and residual dipolar couplings, to characterize a wider range of non-native protein states.
  • Applying this methodology to proteins with intrinsically disordered regions, even under native conditions.
  • Further investigating the role of sequence-dependent propensities in shaping the structure of unfolded proteins.

Summary updated

Abstract

The denatured state of a protein contains important information about the determinants of the folding process. By combining site-directed spin-labeling NMR experiments and restrained computer simulations, we have determined ensembles of conformations that represent the denatured state of the bovine acyl-coenzyme A binding protein (ACBP) at three different concentrations of guanidine hydrochloride. As the experimentally determined distance information corresponds to weighted averages over a broad ensemble of structures, we applied the experimental restraints to a system of noninteracting replicas of the protein by using a Monte Carlo sampling scheme. This procedure permits us to sample ensembles of conformations that are compatible with the experimental data and thus to obtain information regarding the distribution of structures in the denatured state. Our results show that the denatured state of ACBP is highly heterogeneous. The high sensitivity of the computational method that we present, however, enabled us to identify long-range interactions between two regions, located near the N- and C-termini, that include both native and non-native elements. The preferential formation of these contacts suggests that the sequence-dependent patterns of helical propensity and hydrophobicity are important determinants of the structure in the denatured state of ACBP.
BibTeX
@article{lindorfflarsen2004determination,
  author = {Kresten Lindorff‐Larsen and Sigríður Kristjánsdóttir and Kaare Teilum and Wolfgang Fieber and Christopher M. Dobson and Flemming M. Poulsen and Michele Vendruscolo},
  title = {{Determination of an Ensemble of Structures Representing the Denatured State of the Bovine Acyl-Coenzyme A Binding Protein}},
  journal = {Journal of the American Chemical Society},
  year = {2004},
  doi = {10.1021/ja039250g},
}
Figure