Search

Machine learning has played transformative roles in numerous chemical and biophysical problems such as protein folding where large amount of data exists. Nonetheless, many important problems remain challenging for data-driven machine learning approaches due to the limitation of data scarcity. One approach to overcome data scarcity is to incorporate physical principles such as through molecular modeling and simulation. Here, we focus on the big potassium (BK) channels that play important roles in cardiovascular and neural systems. Many mutants of BK channel are associated with various neurological and cardiovascular diseases, but the molecular effects are unknown. The voltage gating properties of BK channels have been characterized for 473 site-specific mutations experimentally over the last three decades; yet, these functional data by themselves remain far too sparse to derive a predictive model of BK channel voltage gating. Using physics-based modeling, we quantify the energetic effects of all single mutations on both open and closed states of the channel. Together with dynamic properties derived from atomistic simulations, these physical descriptors allow the training of random forest models that could reproduce unseen experimentally measured shifts in gating voltage, ∆V1/2, with a RMSE ~ 32 mV and correlation coefficient of R ~ 0.7. Importantly, the model appears capable of uncovering nontrivial physical principles underlying the gating of the channel, including a central role of hydrophobic gating. The model was further evaluated using four novel mutations of L235 and V236 on the S5 helix, mutations of which are predicted to have opposing effects on V1/2 and suggest a key role of S5 in mediating voltage sensor-pore coupling. The measured ∆V1/2 agree quantitatively with prediction for all four mutations, with a high correlation of R = 0.92 and RMSE = 18 mV. Therefore, the model can capture nontrivial voltage gating properties in regions where few mutations are known. The success of predictive modeling of BK voltage gating demonstrates the potential of combining physics and statistical learning for overcoming data scarcity in nontrivial protein function prediction.

To help cells cope with protein misfolding and aggregation, Hsp70 molecular chaperones selectively bind a variety of sequences (“selective promiscuity”). Statistical analyses from substrate-derived peptide arrays reveal that DnaK, the E. coli Hsp70, binds to sequences containing three to five branched hydrophobic residues, although otherwise the specific amino acids can vary considerably. Several high-resolution structures of the substrate -binding domain (SBD) of DnaK bound to peptides reveal a highly conserved configuration of the bound substrate and further suggest that the substrate-binding cleft consists of five largely independent sites for interaction with five consecutive substrate residues. Importantly, both substrate backbone orientations (N- to C- and C- to N-) allow essentially the same backbone hydrogen-bonding and side-chain interactions with the chaperone. In order to rationalize these observations, we performed atomistic molecular dynamics simulations to sample the interactions of all 20 amino acid side chains in each of the five sites of the chaperone in the context of the conserved substrate backbone configurations. The resulting interaction energetics provide the basis set for deriving a predictive model that we call Paladin (Physics-based model of DnaK-Substrate Binding). Trained using available peptide array data, Paladin can distinguish binders and nonbinders of DnaK with accuracy comparable to existing predictors and further predicts the detailed configuration of the bound sequence. Tested using existing DnaK-peptide structures, Paladin correctly predicted the binding register in 10 out of 13 substrate sequences that bind in the N- to C- orientation, and the binding orientation in 16 out of 22 sequences. The physical basis of the Paladin model provides insight into the origins of how Hsp70s bind substrates with a balance of selectivity and promiscuity. The approach described here can be extended to other Hsp70s where extensive peptide array data is not available. Author summary Molecular chaperones are proteins that help prevent misfolding and aggregation of their substrates. This is a complex task, as the cell is very crowded, and the chaperone must efficiently bind to misfolded regions of the client protein while avoiding well-folded proteins. An additional confounding detail is that proteins can enter the binding cleft of some Hsp70s in two orientations, a fact often unaccounted for by existing sequence-based models. Here, we developed a model to describe how client proteins bind to DnaK, the E. coli Hsp70, using physics-based molecular dynamics simulations to quantify the interactions between a variety of peptide substrates and key sites on DnaK. The resulting model, which we call Paladin, provides a physical basis to understand how DnaK binds to specific peptides. Given a sequence, Paladin can predict the precise residues that bind at a specific site on DnaK and can further explain challenging features like the binding orientation, which are typically not predicted by sequence-only models. The Paladin model could be used to design ‘super-binder’ therapeutic peptides to inhibit chaperones like DnaK in E. coli.

Intrinsically disordered proteins (IDPs) are frequently associated with human diseases such as cancers, and about one-fourth of disease-associated missense mutations have been mapped into predicted disordered regions. Understanding how these mutations affect the structure-function relationship of IDPs is a formidable task that requires detailed characterization of the disordered conformational ensembles. Implicit solvent coupled with enhanced sampling has been proposed to provide a balance between accuracy and efficiency necessary for systematic and comparative assessments of the effects of mutations as well as post-translational modifications on IDP structure and interaction. Here, we utilize a recently developed replica exchange with guided annealing enhanced sampling technique to calculate well-converged atomistic conformational ensembles of the intrinsically disordered transactivation domain (TAD) of tumor suppressor p53 and several cancer-associated mutants in implicit solvent. The simulations are critically assessed by quantitative comparisons with several types of experimental data that provide structural information on both secondary and tertiary levels. The results show that the calculated ensembles reproduce local structural features of wild-type p53-TAD and the effects of K24N mutation quantitatively. On the tertiary level, the simulated ensembles are overly compact, even though they appear to recapitulate the overall features of transient long-range contacts qualitatively. A key finding is that, while p53-TAD and its cancer mutants sample a similar set of conformational states, cancer mutants could introduce both local and long-range structural modulations to potentially perturb the balance of p53 binding to various regulatory proteins and further alter how this balance is regulated by multisite phosphorylation of p53-TAD. The current study clearly demonstrates the promise of atomistic simulations for detailed characterization of IDP conformations, and at the same time reveals important limitations in the current implicit solvent protein force field that must be sufficiently addressed for reliable description of long-range structural features of the disordered ensembles.

Achieving facile specific recognition is essential for intrinsically disordered proteins (IDPs) that are involved in cellular signaling and regulation. Consideration of the physical time scales of protein folding and diffusion-limited protein-protein encounter has suggested that the frequent requirement of protein folding for specific IDP recognition could lead to kinetic bottlenecks. How IDPs overcome such potential kinetic bottlenecks to viably function in signaling and regulation in general is poorly understood. Our recent computational and experimental study of cell-cycle regulator p27 (Ganguly et al., J. Mol. Biol. (2012)) demonstrated that long-range electrostatic forces exerted on enriched charges of IDPs could accelerate protein-protein encounter via "electrostatic steering" and at the same time promote "folding-competent" encounter topologies to enhance the efficiency of IDP folding upon encounter. Here, we further investigated the coupled binding and folding mechanisms and the roles of electrostatic forces in the formation of three IDP complexes with more complex folded topologies. The surface electrostatic potentials of these complexes lack prominent features like those observed for the p27/Cdk2/cyclin A complex to directly suggest the ability of electrostatic forces to facilitate folding upon encounter. Nonetheless, similar electrostatically accelerated encounter and folding mechanisms were consistently predicted for all three complexes using topology-based coarse-grained simulations. Together with our previous analysis of charge distributions in known IDP complexes, our results support a prevalent role of electrostatic interactions in promoting efficient coupled binding and folding for facile specific recognition. These results also suggest that there is likely a co-evolution of IDP folded topology, charge characteristics, and coupled binding and folding mechanisms, driven at least partially by the need to achieve fast association kinetics for cellular signaling and regulation.

Peptide-based packaging systems show great potential as safer drug delivery systems. They overcome problems associated with lipid-based or viral delivery systems, vis-a-vis stability, specificity, inflammation, antigenicity, and tune-ability. Here, we describe a set of 15 & 23-residue branched, amphiphilic peptides that mimic phosphoglycerides in molecular architecture. These peptides undergo supramolecular self-assembly and form solvent-filled, bilayer delimited spheres with 50-200 nm diameters as confirmed by TEM, STEM and DLS. Whereas weak hydrophobic forces drive and sustain lipid bilayer assemblies, these all-peptide structures are stabilized potentially by both hydrophobic interactions and hydrogen bonds and remain intact at low micromolar concentrations and higher temperatures. A linear peptide lacking the branch point showed no self-assembly properties. We have observed that these peptide vesicles can trap fluorescent dye molecules within their interior and are taken up by N/N 1003A rabbit lens epithelial cells grown in culture. These assemblies are thus potential drug delivery systems that can overcome some of the key limitations of the current packaging systems.

To understand the interplay of residual structures and conformational fluctuations in the interaction of intrinsically disordered proteins (IDPs), we first combined implicit solvent and replica exchange sampling to calculate atomistic disordered ensembles of the nuclear co-activator binding domain (NCBD) of transcription coactivator CBP and the activation domain of the p160 steroid receptor coactivator ACTR. The calculated ensembles are in quantitative agreement with NMR-derived residue helicity and recapitulate the experimental observation that, while free ACTR largely lacks residual secondary structures, free NCBD is a molten globule with a helical content similar to that in the folded complex. Detailed conformational analysis reveals that free NCBD has an inherent ability to substantially sample all the helix configurations that have been previously observed either unbound or in complexes. Intriguingly, further high-temperature unbinding and unfolding simulations in implicit and explicit solvents emphasize the importance of conformational fluctuations in synergistic folding of NCBD with ACTR. A balance between preformed elements and conformational fluctuations appears necessary to allow NCBD to interact with different targets and fold into alternative conformations. Together with previous topology-based modeling and existing experimental data, the current simulations strongly support an "extended conformational selection" synergistic folding mechanism that involves a key intermediate state stabilized by interaction between the C-terminal helices of NCBD and ACTR. In addition, the atomistic simulations reveal the role of long-range as well as short-range electrostatic interactions in cooperating with readily fluctuating residual structures, which might enhance the encounter rate and promote efficient folding upon encounter for facile binding and folding interactions of IDPs. Thus, the current study not only provides a consistent mechanistic understanding of the NCBD/ACTR interaction, but also helps establish a multi-scale molecular modeling framework for understanding the structure, interaction, and regulation of IDPs in general.

Notch receptors are core components of the Notch signaling pathway and play a central role in cell fate decisions during development as well as tissue homeostasis. Upon ligand binding, Notch is sequentially cleaved at the S2 site by an ADAM protease and at the S3 site by the γ-secretase complex. Recent X-ray structures of the negative regulatory region (NRR) of the Notch receptor reveal an auto-inhibited fold where three protective Lin12/Notch repeats (LNR) of the NRR shield the S2 cleavage site housed in the heterodimerization (HD) domain. One of the models explaining how ligand binding drives the NRR conformation from a protease-resistant state to a protease-sensitive one invokes a mechanical force exerted on the NRR upon ligand endocytosis. Here, we combined physics-based atomistic simulations and topology-based coarse-grained modeling to investigate the intrinsic and force-induced folding and unfolding mechanisms of the human Notch1 NRR. The simulations support that external force applied to the termini of the NRR disengages the LNR modules from the heterodimerization (HD) domain in a well-defined, largely sequential manner. Importantly, the mechanical force can further drive local unfolding of the HD domain in a functionally relevant fashion that would provide full proteolytic access to the S2 site prior to heterodimer disassociation. We further analyzed local structural features, intrinsic folding free energy surfaces, and correlated motions of the HD domain. The results are consistent with a model in which the HD domain possesses inherent mechanosensing characteristics that could be utilized during Notch activation. This potential role of the HD domain in ligand-dependent Notch activation may have implications for understanding normal and aberrant Notch signaling.