Biophysical characterization of intrinsically disordered peptides through molecular dynamics simulations and solution nuclear magnetic resonance spectroscopy
Loading...
Authors
Rosenman, David Jacob
Issue Date
2015-08
Type
Electronic thesis
Thesis
Thesis
Language
ENG
Keywords
Biology
Alternative Title
Abstract
Here, the ensembles of Aβ were simulated using all atom models and explicit water. Further, we employed enhanced sampling through replica exchange molecular dynamics (REMD) using longer timescales than had been employed in the past. Indeed, we observe that these extensive timescales are needed to reach convergence for the emerging β structure of the Aβ monomer simulations. Further, the Aβ variants were each simulated using multiple force fields, including the state of the art OPLS-AA/TIP3P and AMBER99sb-ILDN/TIP4P-Ew combinations. We have also collected various experimental data, primarily at the atomistic level using nuclear magnetic resonance (NMR) spectroscopy. These data, including chemical shifts, J-couplings, solvent exchange rates, and pressure coefficients (measures of the sensitivity of chemical shifts to pressure), are not only observables that can be directly compared to values back-calculated from simulation to validate these runs, but also, collectively provide their own description of the secondary structure biases in the Aβ ensemble that can be compared to that of simulation.
The final chapter summarizes our findings and proposes new research directions to expand and improve our characterization of IDPs. Ongoing REMD simulations of Aβ peptides, including those run with CHARMM22*/TIP3SP parameters are also described. Simulation of Aβ40 using this CHARMM combination reduces the compaction of the ensemble, while preserving similar tertiary structure biases as the other force field combinations explored in this thesis and showing good agreement to NMR data.
We have further attempted to assess the robustness of our simulation techniques through their application in modeling the conformational landscapes for subpeptides of SOD1 and α-synuclein, involved in the pathogenesis of amyotrophic lateral sclerosis and Parkinson’s disease, respectively. The modeled fragments were of comparable length to Aβ, and some of these are known to be intrinsically disordered. Förster resonance energy transfer (FRET) and small-angle X-ray scattering (SAXS) data were used to inform the ensemble shape of the SOD1 fragments, while NMR measurements were provided for the α-synuclein subpeptide. Unfortunately, comparison to these experimental data yielded poor agreement for these systems. The simulations of SOD1 peptides suggest that our parameters are overly predisposed to sample collapsed states. Meanwhile, simulations between different force fields yielded very different descriptions of the α-synuclein fragment, and neither representation produced observables that agreed well with experimental values. These descriptions inform important weaknesses in our simulation parameters, but also suggest that the concurrences in the characterizations made through simulation and experiment for Aβ are non-trivial.
While these trends were common to our approaches to studying Aβ, examination at finer levels revealed key inconsistencies. For example, the two force fields used here model the local structure and electrostatics of the A21-A30 central region in very different ways. Unsurprisingly, the FAD mutations we studied which lie in this region also had very different effects on the structural ensembles observed between the different force fields. Unfortunately, the available experimental data, in general, remain too coarse to inform which model, if any, is correct.
Representing 20-30% of the human proteome, intrinsically disordered peptides (IDPs) are highly dynamic sequences that do not assume a single, defined native state under physiological conditions. These peptides are implicated as causative agents in various human neurodegenerative diseases, primarily due to their susceptibility to misfold and aggregate. Further, the ambiguous and diverse free energy landscapes for these systems challenge the validity and applicability of our computational modeling techniques, and preclude the application of many of the experimental methods we currently and proficiently use to characterize well-folded proteins. This thesis defines the ensemble structural biases and individual conformations of the monomer state of various disease-relevant IDPs through the union of simulations and experiments, in order to better understand their role in aggregation and disease pathology. Through their application to these challenging systems, this work also aims to identify previously unknown weaknesses and strengths of these biophysical techniques in order to improve their application to other systems.
The majority of this thesis investigates the amyloid β (Aβ) monomer, a prototypic IDP whose aggregates are implicated in the pathogenesis of Alzheimer’s disease (AD), a neurodegenerative disease that affects 36 million people worldwide. Relatively subtle modifications to the peptide, including a two residue change in peptide length (Aβ40 vs Aβ42) and point mutations in the Aβ sequence that are linked to familial AD (FAD), result in profound changes to the aggregation and toxicity of the peptide. We hypothesize these changes are fundamentally linked to the intrinsic disorder of monomer state; these disordered ensembles may be easily perturbed by small changes in physicochemistry, allowing them to sample new conformers that can seed different aggregates.
Here, simulations and experiment were used to describe different variants of Aβ and ascertain the nature and extent of these changes at the monomeric level. Numerous computational characterizations of Aβ monomers exist in the literature, but little consensus exists among these studies. We suspect that simulations of IDPs are particularly sensitive to the conditions of the run; as such, our approach to model Aβ emphasizes explicit representation, broad sampling, and parameter generalizability.
The union between experiment and simulation with multiple force fields reveals a consistent model of Aβ supported by multiple methods. This model describes wild type Aβ40 with antiparallel β-hairpin between L17-A21 and A30-L34, while residues A21-A30 forms an intervening loop region that rarely interacts with the majority of the protein. Meanwhile, Aβ42 contributes new β-hairpin motifs involving V40-I41, with a new turn involving residue G37. These structural trends for the Aβ monomer are of particular interest because they increase the solvent exposure of hydrophobic side chains and because they correlate to intrapeptide models for oligomers and fibrils as determined by solid state NMR. This suggests that these conformations may represent the seeds of aggregation for these higher order forms.
The final chapter summarizes our findings and proposes new research directions to expand and improve our characterization of IDPs. Ongoing REMD simulations of Aβ peptides, including those run with CHARMM22*/TIP3SP parameters are also described. Simulation of Aβ40 using this CHARMM combination reduces the compaction of the ensemble, while preserving similar tertiary structure biases as the other force field combinations explored in this thesis and showing good agreement to NMR data.
We have further attempted to assess the robustness of our simulation techniques through their application in modeling the conformational landscapes for subpeptides of SOD1 and α-synuclein, involved in the pathogenesis of amyotrophic lateral sclerosis and Parkinson’s disease, respectively. The modeled fragments were of comparable length to Aβ, and some of these are known to be intrinsically disordered. Förster resonance energy transfer (FRET) and small-angle X-ray scattering (SAXS) data were used to inform the ensemble shape of the SOD1 fragments, while NMR measurements were provided for the α-synuclein subpeptide. Unfortunately, comparison to these experimental data yielded poor agreement for these systems. The simulations of SOD1 peptides suggest that our parameters are overly predisposed to sample collapsed states. Meanwhile, simulations between different force fields yielded very different descriptions of the α-synuclein fragment, and neither representation produced observables that agreed well with experimental values. These descriptions inform important weaknesses in our simulation parameters, but also suggest that the concurrences in the characterizations made through simulation and experiment for Aβ are non-trivial.
While these trends were common to our approaches to studying Aβ, examination at finer levels revealed key inconsistencies. For example, the two force fields used here model the local structure and electrostatics of the A21-A30 central region in very different ways. Unsurprisingly, the FAD mutations we studied which lie in this region also had very different effects on the structural ensembles observed between the different force fields. Unfortunately, the available experimental data, in general, remain too coarse to inform which model, if any, is correct.
Representing 20-30% of the human proteome, intrinsically disordered peptides (IDPs) are highly dynamic sequences that do not assume a single, defined native state under physiological conditions. These peptides are implicated as causative agents in various human neurodegenerative diseases, primarily due to their susceptibility to misfold and aggregate. Further, the ambiguous and diverse free energy landscapes for these systems challenge the validity and applicability of our computational modeling techniques, and preclude the application of many of the experimental methods we currently and proficiently use to characterize well-folded proteins. This thesis defines the ensemble structural biases and individual conformations of the monomer state of various disease-relevant IDPs through the union of simulations and experiments, in order to better understand their role in aggregation and disease pathology. Through their application to these challenging systems, this work also aims to identify previously unknown weaknesses and strengths of these biophysical techniques in order to improve their application to other systems.
The majority of this thesis investigates the amyloid β (Aβ) monomer, a prototypic IDP whose aggregates are implicated in the pathogenesis of Alzheimer’s disease (AD), a neurodegenerative disease that affects 36 million people worldwide. Relatively subtle modifications to the peptide, including a two residue change in peptide length (Aβ40 vs Aβ42) and point mutations in the Aβ sequence that are linked to familial AD (FAD), result in profound changes to the aggregation and toxicity of the peptide. We hypothesize these changes are fundamentally linked to the intrinsic disorder of monomer state; these disordered ensembles may be easily perturbed by small changes in physicochemistry, allowing them to sample new conformers that can seed different aggregates.
Here, simulations and experiment were used to describe different variants of Aβ and ascertain the nature and extent of these changes at the monomeric level. Numerous computational characterizations of Aβ monomers exist in the literature, but little consensus exists among these studies. We suspect that simulations of IDPs are particularly sensitive to the conditions of the run; as such, our approach to model Aβ emphasizes explicit representation, broad sampling, and parameter generalizability.
The union between experiment and simulation with multiple force fields reveals a consistent model of Aβ supported by multiple methods. This model describes wild type Aβ40 with antiparallel β-hairpin between L17-A21 and A30-L34, while residues A21-A30 forms an intervening loop region that rarely interacts with the majority of the protein. Meanwhile, Aβ42 contributes new β-hairpin motifs involving V40-I41, with a new turn involving residue G37. These structural trends for the Aβ monomer are of particular interest because they increase the solvent exposure of hydrophobic side chains and because they correlate to intrapeptide models for oligomers and fibrils as determined by solid state NMR. This suggests that these conformations may represent the seeds of aggregation for these higher order forms.
Description
August 2015
School of Science
School of Science
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY