The MemProtMD database: a resource for membrane-embedded protein structures and their lipid interactions

Abstract

Integral membrane proteins fulfil important roles in many crucial biological processes, including cell signalling, molecular transport and bioenergetic processes. Advancements in experimental techniques are revealing high resolution structures for an increasing number of membrane proteins. Yet, these structures are rarely resolved in complex with membrane lipids. In 2015, the MemProtMD pipeline was developed to allow the automated lipid bilayer assembly around new membrane protein structures, released from the Protein Data Bank (PDB). To make these data available to the scientific community, a web database (http://memprotmd.bioch.ox.ac.uk) has been developed. Simulations and the results of subsequent analysis can be viewed using a web browser, including interactive 3D visualizations of the assembled bilayer and 2D visualizations of lipid contact data and membrane protein topology. In addition, ensemble analyses are performed to detail conserved lipid interaction information across proteins, families and for the entire database of 3506 PDB entries. Proteins may be searched using keywords, PDB or Uniprot identifier, or browsed using classification systems, such as Pfam, Gene Ontology annotation, mpstruc or the Transporter Classification Database. All files required to run further molecular simulations of proteins in the database are provided.

Figure 1.

The number of unique PDB, UniProt and Pfam accessions represented in the MemProtMD database over time. A selection of landmark structures are indicated by red lines, denoting their PDB release date and include the first Porin (44), K⁺ channel (45), aquaporin (46), solute transporter (47), Sec translocon (48), adrenergic receptor (49), glutamate receptor (50), TRP channel (51) and voltage-gated Ca²⁺ channel (52). The total number of PDB structures, UniProt proteins and Pfam families are shown.

Figure 2.

Overview of the MemProtMD pipeline and database. Membrane protein structures deposited in the PDB are first converted to MARTINI representation. Lipids and solvent are then added with random orientations and a 1 μs self-assembly simulation is performed. The simulation is converted back to atomistic representation using CG2AT and several analyses are performed. Analysis results, as well as metadata derived from the original PDB structure and metadata downloaded from a range of databases are deposited into the MemProtMD database, which is then used to perform multiple sequence alignments, automatically classify membrane proteins, perform text searches and perform ensemble analyses.

Figure 3.

Online analysis and visualization of a simulation of the A_2A adenosine receptor. (A) Pre-rendered images available to view and download showing (from left to right) protein with membrane surface shaded to show thinning (red) and thickening (blue), protein with CG lipid head-group beads (yellow: glycerol, red: phosphate, blue: choline) and protein surface shaded according to contacts with lipid head-groups. (B) Chain-by-chain overview of the protein, showing violin plot of protein density along membrane normal, classification of chain membrane interactions (‘Membrane spanning’), topology summary (‘7 membrane spanning α helices’) and residue count. (C) Online 3D protein view showing protein in sphere representation coloured by membrane contact type (yellow: acyl tails, red: lipid head-groups, blue: solvent). The selected chain (Chain A) is shown in full colour, whilst other parts of the protein are desaturated. (D) Topology viewer showing tilt and curvature of TM helices. Non-continuous loops are shown as dashed lines. The membrane thickness and degree of protein-induced deformation is shown. (E) Sequence viewer showing several metrics along the amino acid sequence: (from top to bottom) contacts with lipids flipping from one leaflet to another; contacts with water and ions; contacts with lipids, broken down by chemical group; stacked plot of contacts with acyl tails, head-groups and solvent; position of the protein and local membrane thickness; protein secondary structure and amino acid sequence.

Figure 4.

Online sequence view for 25 simulations of the UniProt AA2AR_HUMAN entry (Uniprot id: P29274) contacts with lipid head-groups (red) and lipid acyl tails (yellow) are shown arranged along the protein amino acid sequence, with each row representing a single simulated structure. A consensus secondary structure is shown at the bottom of each row for each PDB file, shaded darker where secondary structure is more highly conserved within the structures. TM domains are shown by a box around the secondary structure, shaded according to mean depth within the membrane, from red (shallow) to yellow (deep), where a certain residue was not present in a structure the cell is shaded grey.