Growing ecosystem of deep learning methods for modeling protein-protein interactions.

Numerous cellular functions rely on protein-protein interactions. Efforts to comprehensively characterize them remain challenged however by the diversity of molecular recognition mechanisms employed within the proteome. Deep learning has emerged as a promising approach for tackling this problem by exploiting both experimental data and basic biophysical knowledge about protein interactions. Here, we review the growing ecosystem of deep learning methods for modeling protein interactions, highlighting the diversity of these biophysically informed models and their respective trade-offs. We discuss recent successes in using representation learning to capture complex features pertinent to predicting protein interactions and interaction sites, geometric deep learning to reason over protein structures and predict complex structures, and generative modeling to design de novo protein assemblies. We also outline some of the outstanding challenges and promising new directions. Opportunities abound to discover novel interactions, elucidate their physical mechanisms, and engineer binders to modulate their functions using deep learning and, ultimately, unravel how protein interactions orchestrate complex cellular behaviors.

Ceramide-1-phosphate transfer protein enhances lipid transport by disrupting hydrophobic lipid-membrane contacts.

Cellular distributions of the sphingolipid ceramide-1-phosphate (C1P) impact essential biological processes. C1P levels are spatiotemporally regulated by ceramide-1-phosphate transfer protein (CPTP), which efficiently shuttles C1P between organelle membranes. Yet, how CPTP rapidly extracts and inserts C1P into a membrane remains unknown. Here, we devise a multiscale simulation approach to elucidate biophysical details of CPTP-mediated C1P transport. We find that CPTP binds a membrane poised to extract and insert C1P and that membrane binding promotes conformational changes in CPTP that facilitate C1P uptake and release. By significantly disrupting a lipid's local hydrophobic environment in the membrane, CPTP lowers the activation free energy barrier for passive C1P desorption and enhances C1P extraction from the membrane. Upon uptake of C1P, further conformational changes may aid membrane unbinding in a manner reminiscent of the electrostatic switching mechanism used by other lipid transfer proteins. Insertion of C1P into an acceptor membrane, eased by a decrease in membrane order by CPTP, restarts the transfer cycle. Most notably, we provide molecular evidence for CPTP's ability to catalyze C1P extraction by breaking hydrophobic C1P-membrane contacts with compensatory hydrophobic lipid-protein contacts. Our work, thus, provides biophysical insights into how CPTP efficiently traffics C1P between membranes to maintain sphingolipid homeostasis and, additionally, presents a simulation method aptly suited for uncovering the catalytic mechanisms of other lipid transfer proteins.

Remembering the Work of Phillip L. Geissler: A Coda to His Scientific Trajectory.

Phillip L. Geissler made important contributions to the statistical mechanics of biological polymers, heterogeneous materials, and chemical dynamics in aqueous environments. He devised analytical and computational methods that revealed the underlying organization of complex systems at the frontiers of biology, chemistry, and materials science. In this retrospective we celebrate his work at these frontiers.

Static Disorder has Dynamic Impact on Energy Transport in Biomimetic Light-Harvesting Complexes.

Despite extensive studies, many questions remain about what structural and energetic factors give rise to the remarkable energy transport efficiency of photosynthetic light-harvesting protein complexes, owing largely to the inability to synthetically control such factors in these natural systems. Herein, we demonstrate energy transfer within a biomimetic light-harvesting complex consisting of identical chromophores attached in a circular array to a protein scaffold derived from the tobacco mosaic virus coat protein. We confirm the capability of energy transport by observing ultrafast depolarization in transient absorption anisotropy measurements and a redshift in time-resolved emission spectra in these complexes. Modeling the system with kinetic Monte Carlo simulations recapitulates the observed anisotropy decays, suggesting an inter-site hopping rate as high as 1.6 ps-1. With these simulations, we identify static disorder in orientation, site energy, and degree of coupling as key remaining factors to control to achieve long-range energy transfer in these systems. We thereby establish this system as a highly promising, bottom-up model for studying long-range energy transfer in light-harvesting protein complexes.

Membrane hydrophobicity determines the activation free energy of passive lipid transport.

The collective behavior of lipids with diverse chemical and physical features determines a membrane's thermodynamic properties. Yet, the influence of lipid physicochemical properties on lipid dynamics, in particular interbilayer transport, remains underexplored. Here, we systematically investigate how the activation free energy of passive lipid transport depends on lipid chemistry and membrane phase. Through all-atom molecular dynamics simulations of 11 chemically distinct glycerophospholipids, we determine how lipid acyl chain length, unsaturation, and headgroup influence the free energy barriers for two elementary steps of lipid transport: lipid desorption, which is rate limiting, and lipid insertion into a membrane. Consistent with previous experimental measurements, we find that lipids with longer, saturated acyl chains have increased activation free energies compared to lipids with shorter, unsaturated chains. Lipids with different headgroups exhibit a range of activation free energies; however, no clear trend based solely on chemical structure can be identified, mirroring difficulties in the interpretation of previous experimental results. Compared to liquid-crystalline phase membranes, gel phase membranes exhibit substantially increased free energy barriers. Overall, we find that the activation free energy depends on a lipid's local hydrophobic environment in a membrane and that the free energy barrier for lipid insertion depends on a membrane's interfacial hydrophobicity. Both of these properties can be altered through changes in lipid acyl chain length, lipid headgroup, and membrane phase. Thus, the rate of lipid transport can be tuned through subtle changes in local membrane composition and order, suggesting an unappreciated role for nanoscale membrane domains in regulating cellular lipid dynamics.

Breakage of Hydrophobic Contacts Limits the Rate of Passive Lipid Exchange between Membranes.

The maintenance of heterogeneous lipid compositions among cellular membranes is key to biological function. Yet, even the simplest process that could be responsible for maintaining proper lipid distributions, passive lipid exchange of individual molecules between membranes, has eluded a detailed understanding, due in part to inconsistencies between experimental findings and molecular simulations. We resolve these discrepancies by discovering the reaction coordinate for passive lipid exchange, which enables a complete biophysical characterization of the rate-limiting step for lipid exchange. Our approach to identify the reaction coordinate capitalizes on our ability to harvest over 1000 unbiased trajectories of lipid insertion, an elementary step of passive lipid transport, using all-atom and coarse-grained molecular dynamics simulations. We find that the reaction coordinate measures the formation and breakage of hydrophobic contacts between the membrane and exchanging lipid. Consistent with experiments, free energy profiles as a function of our reaction coordinate exhibit a substantial barrier for insertion. In contrast, lipid insertion was predicted to be a barrier-less process by previous computational studies, which incorrectly presumed the reaction coordinate to be the displacement of the exchanging lipid from the membrane. Utilizing our newfound knowledge of the reaction coordinate, we formulate an expression for the lipid exchange rate to enable a quantitative comparison with experiments. Overall, our results indicate that the breakage of hydrophobic contacts is rate limiting for passive lipid exchange and provide a foundation to understand the catalytic function of lipid transfer proteins.

Exploiting Chromophore-Protein Interactions through Linker Engineering To Tune Photoinduced Dynamics in a Biomimetic Light-Harvesting Platform.

Creating artificial systems that mimic and surpass those found in nature is one of the great challenges of modern science. In the context of photosynthetic light harvesting, the difficulty lies in attaining utmost control over the energetics, positions and relative orientations of chromophores in densely packed arrays to transfer electronic excitation energy to desired locations with high efficiency. Toward achieving this goal, we use a highly versatile biomimetic protein scaffold from the tobacco mosaic virus coat protein on which chromophores can be attached at precise locations via linkers of differing lengths and rigidities. We show that minor linker modifications, including switching chiral configurations and alkyl chain shortening, lead to significant lengthening of the ultrafast excited state dynamics of the system as the linkers are shortened and rigidified. Molecular dynamics simulations provide molecular-level detail over how the chromophore attachment orientations, positions, and distances from the protein surface lead to the observed trends in system dynamics. In particular, we find that short and rigid linkers are able to sandwich water molecules between chromophore and protein, leading to chromophore-water-protein supracomplexes with intricately coupled dynamics that are highly dependent on their local protein environment. In addition, cyclohexyl-based linkers are identified as ideal candidates to retain rotational correlations over several nanoseconds and thus lock relative chromophore orientations throughout the lifetime of an exciton. Combining linker engineering with judicious placement of chromophores on the hydrated protein scaffold to exploit different chromophore-bath couplings provides a clear and effective path to producing highly controllable artificial light-harvesting systems that can increasingly mimic their natural counterparts, thus aiding to elucidate natural photosynthetic mechanisms.

Predictions for α-Helical Glycopeptide Design from Structural Bioinformatics Analysis.

Glycosylation not only impacts the functions of glycoproteins but can also improve glycoprotein stability and folding efficiency-characteristics that are desirable for protein engineering and therapeutic design. To further elucidate the effects of N-glycosylation on protein structure and to provide principles useful for the rational design of α-helical glycopeptides, we investigate stabilizing protein-sugar interactions in α-helical glycosylation sites using an integrated structural bioinformatics analysis and molecular dynamics simulation approach. We identify two glycan conformations with an Asn χ1 of 180° or 300° that are amenable to α-helical structure in natural α-helical glycosylation sites in the Protein Data Bank. A combination of sterics and favorable intraglycopeptide enthalpy explains the existence of only these two conformations. Furthermore, we catalog all known protein-sugar interactions that utilize these conformational modes. The most common interactions involve either a Glu residue at the -4 position interacting with the GlcNAc whose Asn has χ1 = 300° or a Glu residue at the +4 position interacting with the GlcNAc whose Asn has χ1 = 180°. Via metadynamics simulations of model α-helical glycopeptides with each of these two interactions, we find that both interactions are stabilizing as a result of favorable electrostatic intraglycopeptide interactions. Thus, we suggest that incorporating a Glu at either the -4 or +4 position relative to an N-linked glycan may be a useful strategy for engineering stable α-helical glycoproteins.

Computational methods to design cyclic peptides.

Cyclic peptides (CPs) are promising modulators of protein-protein interactions (PPIs), but their application remains challenging. It is currently difficult to predict the structures and bioavailability of CPs. The ability to design CPs using computer modeling would greatly facilitate the development of CPs as potent PPI modulators for fundamental studies and as potential therapeutics. Herein, we describe computational methods to generate CP libraries for virtual screening, as well as current efforts to accurately predict the conformations adopted by CPs. These advances are making it possible to envision robust computational design of active CPs. However, unique properties of CPs pose significant challenges associated with sampling CP conformational space and accurately describing CP energetics. These major obstacles to structure prediction likely must be solved before robust design of active CPs can be reliably achieved.

Insights into How Cyclic Peptides Switch Conformations.

Cyclic peptides have recently emerged as promising modulators of protein-protein interactions. However, it is currently highly difficult to predict the structures of cyclic peptides owing to their rugged conformational free energy landscape, which prevents sampling of all thermodynamically relevant conformations. In this article, we first investigate how a relatively flexible cyclic hexapeptide switches conformations. It is found that, although the circular geometry of small cyclic peptides of size 6-8 may require rare, coherent dihedral changes to sample a new conformation, the changes are rather local, involving simultaneous changes of ϕi and ψi or ψi and ϕi+1. The understanding of how these cyclic peptides switch conformations enables the use of metadynamics simulations with reaction coordinates specifically targeting such coupled two-dihedral changes to effectively sample cyclic peptide conformational space.