Three-Dimensional Interaction Homology: Deconstructing Residue-Residue and Residue-Lipid Interactions in Membrane Proteins.

A method is described to deconstruct the network of hydropathic interactions within and between a protein's sidechain and its environment into residue-based three-dimensional maps. These maps encode favorable and unfavorable hydrophobic and polar interactions, in terms of spatial positions for optimal interactions, relative interaction strength, as well as character. In addition, these maps are backbone angle-dependent. After map calculation and clustering, a finite number of unique residue sidechain interaction maps exist for each backbone conformation, with the number related to the residue's size and interaction complexity. Structures for soluble proteins (~749,000 residues) and membrane proteins (~387,000 residues) were analyzed, with the latter group being subdivided into three subsets related to the residue's position in the membrane protein: soluble domain, core-facing transmembrane domain, and lipid-facing transmembrane domain. This work suggests that maps representing residue types and their backbone conformation can be reassembled to optimize the medium-to-high resolution details of a protein structure. In particular, the information encoded in maps constructed from the lipid-facing transmembrane residues appears to paint a clear picture of the protein-lipid interactions that are difficult to obtain experimentally.

3D interaction homology: The hydrophobic residues alanine, isoleucine, leucine, proline and valine play different structural roles in soluble and membrane proteins.

The aliphatic hydrophobic amino acid residues-alanine, isoleucine, leucine, proline and valine-are among the most common found in proteins. Their structural role in proteins is seemingly obvious: engage in hydrophobic interactions to stabilize secondary, and to a lesser extent, tertiary and quaternary structure. However, favorable hydrophobic interactions involving the sidechains of these residue types are generally less significant than the unfavorable set arising from interactions with polar atoms. Importantly, the constellation of interactions between residue sidechains and their environments can be recorded as three-dimensional maps that, in turn, can be clustered. The clustered average map sets compose a library of interaction profiles encoding interaction strengths, interaction types and the optimal 3D position for the interacting partners. This library is backbone angle-dependent and suggests solvent and lipid accessibility for each unique interaction profile. In this work, in addition to analysis of soluble proteins, a large set of membrane proteins that contained optimized artificial lipids were evaluated by parsing the structures into three distinct components: soluble extramembrane domain, lipid facing transmembrane domain, core transmembrane domain. The aliphatic residues were extracted from each of these sets and passed through our calculation protocol. Notable observations include: the roles of aliphatic residues in soluble proteins and in the membrane protein's soluble domains are nearly identical, although the latter are slightly more solvent accessible; by comparing maps calculated with sidechain-lipid interactions to maps ignoring those interactions, the potential extent of residue-lipid and residue-interactions can be assessed and likely exploited in structure prediction and modeling; amongst these residue types, the levels of lipid engagement show isoleucine as the most engaged, while the other residues are largely interacting with neighboring helical residues.

Systematized analysis of secondary structure dependence of key structural features of residues in soluble and membrane-bound proteins.

Knowledge of three-dimensional protein structure is integral to most modern drug discovery efforts. Recent advancements have highlighted new techniques for 3D protein structure determination and, where structural data cannot be collected experimentally, prediction of protein structure. We have undertaken a major effort to use existing protein structures to collect, characterize, and catalogue the inter-atomic interactions that define and compose 3D structure by mapping hydropathic interaction environments as maps in 3D space. This work has been performed on a residue-by-residue basis, where we have seen evidence for relationships between environment character, residue solvent-accessible surface areas and their secondary structures. In this graphical review, we apply principles from our earlier studies and expand the scope to all common amino acid residue types in both soluble and membrane proteins. Key to this analysis is parsing the Ramachandran plot to an 8-by-8 chessboard to define secondary structure bins. Our analysis yielded a number of quantitative discoveries: 1) increased fraction of hydrophobic residues (alanine, isoleucine, leucine, phenylalanine and valine) in membrane proteins compared to their fractions in soluble proteins; 2) less burial coupled with significant increases in favorable hydrophobic interactions for hydrophobic residues in membrane proteins compared to soluble proteins; and 3) higher burial and more favorable polar interactions for polar residues now preferring the interior of membrane proteins. These observations and the supporting data should provide benchmarks for current studies of protein residues in different environments and may be able to guide future protein structure prediction efforts.

3D Interaction Homology: Computational Titration of Aspartic Acid, Glutamic Acid and Histidine Can Create pH-Tunable Hydropathic Environment Maps.

Aspartic acid, glutamic acid and histidine are ionizable residues occupying various protein environments and perform many different functions in structures. Their roles are tied to their acid/base equilibria, solvent exposure, and backbone conformations. We propose that the number of unique environments for ASP, GLU and HIS is quite limited. We generated maps of these residue's environments using a hydropathic scoring function to record the type and magnitude of interactions for each residue in a 2703-protein structural dataset. These maps are backbone-dependent and suggest the existence of new structural motifs for each residue type. Additionally, we developed an algorithm for tuning these maps to any pH, a potentially useful element for protein design and structure building. Here, we elucidate the complex interplay between secondary structure, relative solvent accessibility, and residue ionization states: the degree of protonation for ionizable residues increases with solvent accessibility, which in turn is notably dependent on backbone structure.

3D interaction homology: Hydropathic interaction environments of serine and cysteine are strikingly different and their roles adapt in membrane proteins.

Atomic-resolution protein structural models are prerequisites for many downstream activities like structure-function studies or structure-based drug discovery. Unfortunately, this data is often unavailable for some of the most interesting and therapeutically important proteins. Thus, computational tools for building native-like structural models from less-than-ideal experimental data are needed. To this end, interaction homology exploits the character, strength and loci of the sets of interactions that define a structure. Each residue type has its own limited set of backbone angle-dependent interaction motifs, as defined by their environments. In this work, we characterize the interactions of serine, cysteine and S-bridged cysteine in terms of 3D hydropathic environment maps. As a result, we explore several intriguing questions. Are the environments different between the isosteric serine and cysteine residues? Do some environments promote the formation of cystine S-S bonds? With the increasing availability of structural data for water-insoluble membrane proteins, are there environmental differences for these residues between soluble and membrane proteins? The environments surrounding serine and cysteine residues are dramatically different: serine residues are about 50% solvent exposed, while cysteines are only 10% exposed; the latter are more involved in hydrophobic interactions although there are backbone angle-dependent differences. Our analysis suggests that one driving force for -S-S- bond formation is a rather substantial increase in burial and hydrophobic interactions in cystines. Serine and cysteine become less and more, respectively, solvent-exposed in membrane proteins. 3D hydropathic environment maps are an evolving structure analysis tool showing promise as elements in a new protein structure prediction paradigm.

3d interaction homology: The structurally known rotamers of tyrosine derive from a surprisingly limited set of information-rich hydropathic interaction environments described by maps.

Sidechain rotamer libraries are obtained through exhaustive statistical analysis of existing crystallographic structures of proteins and have been applied in multiple aspects of structural biology, for example, crystallography of relatively low-resolution structures, in homology model building and in biomolecular NMR. Little is known, however, about the driving forces that lead to the preference or suitability of one rotamer over another. Construction of 3D hydropathic interaction maps for nearly 30,000 tyrosines reveals the environment around each, in terms of hydrophobic (π-π stacking, etc.) and polar (hydrogen bonding, etc.) interactions. After partitioning the tyrosines into backbone-dependent (ϕ, ψ) bins, a map similarity metric based on the correlation coefficient was applied to each map-map pair to build matrices suitable for clustering with k-means. The first bin (-200° ≤ ϕ < -155°; -205° ≤ ψ < -160°), representing 631 tyrosines, reduced to 14 unique hydropathic environments, with most diversity arising from favorable hydrophobic interactions with many different residue partner types. Polar interactions for tyrosine include surprisingly ubiquitous hydrogen bonding with the phenolic OH and a handful of unique environments surrounding the tyrosine backbone. The memberships of all but one of the 14 environments are dominated (>50%) by a single χ(1)/χ(2) rotamer. The last environment has weak or no interactions with the tyrosine ring and its χ(1)/χ(2) rotamer is indeterminate, which is consistent with it being composed of mostly surface residues. Each tyrosine residue attempts to fulfill its hydropathic valence and thus, structural water molecules are seen in a variety of roles throughout protein structure.

Unintended consequences? Water molecules at biological and crystallographic protein-protein interfaces.

The importance of protein-protein interactions (PPIs) is becoming increasingly appreciated, as these interactions lie at the core of virtually every biological process. Small molecule modulators that target PPIs are under exploration as new therapies. One of the greatest obstacles faced in crystallographically determining the 3D structures of proteins is coaxing the proteins to form "artificial" PPIs that lead to uniform crystals suitable for X-ray diffraction. This work compares interactions formed naturally, i.e., "biological", with those artificially formed under crystallization conditions or "non-biological". In particular, a detailed analysis of water molecules at the interfaces of high-resolution (≤2.30 Å) X-ray crystal structures of protein-protein complexes, where 140 are biological protein-protein complex structures and 112 include non-biological protein-protein interfaces, was carried out using modeling tools based on the HINT forcefield. Surprisingly few and relatively subtle differences were observed between the two types of interfaces: (i) non-biological interfaces are more polar than biological interfaces, yet there is better organized hydrogen bonding at the latter; (ii) biological associations rely more on water-mediated interactions with backbone atoms compared to non-biological associations; (iii) aromatic/planar residues play a larger role in biological associations with respect to water, and (iv) Lys has a particularly large role at non-biological interfaces. A support vector machines (SVMs) classifier using descriptors from this study was devised that was able to correctly classify 84% of the two interface types.