Research Method

The potential of each structural biology methodology at a general level, within our network, will be now developed to demonstrate that

- each approach is essential for a any activity in SBMP,
- the network gathers all the expertise required to pursue SBMP at a high,
- real complementarities do exists between the various methodologies.

Specific achievements and potential of X-ray crystallography

A large majority of the 40000 structures deposited in the PDB were solved by X-ray crystallography. This demonstrates the power of the method for getting structures at nearly atomic resolution, providing three-dimensional well ordered crystals are obtained. Although the number of membrane proteins In the PDB is much lower than the soluble ones, with a few exceptions, most of them were obtained by Xray crystallography (1). These structures provide a very powerful basis for rational drug design or for raising functional mechanisms that will be ascertained by complementary approaches as described In this application. High resolution membrane protein structures also highlight specific protein-lipid interactions that are of interest for understanding how proteins are stabilized in the native membrane (2). However, membrane proteins are more difficult to crystallize and subsequently crystals are more fragile to handle. Crystal are obtained either by the standard vapor diffusion method starting from proteins solubilized in detergent, or from lipidic phases such as cubic phases (3). Both methods are now automated and utilize nanovolumes for each trial. Several robots for nanodrop crystallization are available (Grenoble, Oeiras) and Grenoble will purchase a robot for cubic phases in 2008. In the last 10 years, the use of synchrotron radiation for protein crystallography has largely developed and severalm dedicated beam-lines are easily accessible at various synchrotron sources (i.e. European beam-lines or French beam-line at ESRF in Grenoble). In particular microfocussed beams make it now possible to collect complete data sets on crystals that have only a few microns in the 3 dimensions, as first illustrated by bacteriorhodopsin (4). Working with very small crystals also facilitates the freezing procedure, which is necessary to prevent high radiation damage to the crystals. Indeed, membrane protein crystals, probably because they contain not only water but also amphiphiles, are more difficult to cryo-cool than others. In order to check intrinsic crystal quality, we will also screen directly the diffraction from the crystallization plates. This method developed at the IBS (5) allows to screen automatically (using a robot developed and located at the French beam-line BM30A at ESRF) the different wells of a crystallization plate to identify protein crystals (versus salt crystals) and eventually collect data from crystals that were not taken out from their growing medium. Because of the limitem number of unique membrane protein structures known, most of the structures have to be solved by mutiple isomorphous replacement (MIR) or multiple anomalous dispersion (MAD), or related methods (i.e. SAD). When proteins are overexpressed in E. coli, they can be produce with seleno-methionines and MAD is the method of choice. For others, heavy atoms will be obtained either by co-crystallization or by soaking. Very powerful programs are now available for the calculation of phases, the model building and refinement and are done in a similar way as for soluble proteins. However, because membrane protein crystals are generally of lower quality as for soluble proteins (lower resolution, anisotropy, twinning,…), solving their structures is not yet an automatic procedure. Therefore, it is important to train young researchers in protein crystallography and to maintain a high level expertise In this field.

Figure 1

Figure 1: Schematic representation of several steps involved in X-ray Crystallography, namely crystallization, X-ray data collection and structure determination.


  1. White, Protein Sci., 13, 1948-1949 (2004);,
  2. Palsdottir and Hunte, BBA, 1666, 2-18 (2004).
  3. Landau and Rosenbusch, PNAS, 93, 14532-14535 (1996).
  4. Pebay-Peyroula et al., Science, 277, 1676-1681, (1997).
  5. Jacquamet et al., Structure, 12, 1219-1225 (2004).

Specific achievements and potential of electron crystallography

Electron crystallography is currently the only method to solve the structure of a membrane protein embedded in a lipid bilayer. To this end a membrane protein is overexpressed, purified and crystallized by reconstitution in presence of lipids (1). Since the membrane protein is integrated In the bilayer, the possibility exists that a fragile membrane protein may be crystallized into 2D crystals, but cannot be assembled into 3D crystals. This and the fact that the most native structure of that membrane protein can be assessed drives a few groups to push the technology further. Yoshi Fujiyoshi has pushed the technology for acquiring the information from 2D crystals by imaging and by diffraction. If large well ordered crystals are available, data can be collected In Kyoto that allow the structure of the pertinent membrane protein to be solved to better than 2 Å resolution (2). A new 300 kV FEG transmission electron microscope will be available in Basel in 2008, offering similar capacities as the Kyoto instruments.

Figure 2

Figure 2: 2D crystallization (2DX) of membrane proteins and electron crystallography. A) Dialysis driven removal of the detergent is the most frequently used 2DX method (1). While the detergent is eliminated, lipids and membrane proteins assemble into an ordered aggregate. Other ways to bring the detergent below the critical micelle concentration and thus promoting ordered aggregation of solubilized lipids and proteins concern dilution (3), adsorption by BioBeads (4) or by cyclodetrin (5). B) High quality 2D crystals exhibit uniform thickness, a size of several μm, and characteristic shape. Scale bar represents 1 μm. C) Electron diffraction reveals the crystal quality. In favourable cases, the highest reflections are at (2 Å)-1, as marked in the inset, which shows the outermost reflections. D) Image processing is key to extract the information from high-resolution images and electron diffraction patterns for mergeing it into a 3D potential map.

2D crystallization is achieved by removing or diluting the detergent that keeps both the membrane protein and the lipid in solution (3). Upon reaching the critical micellar concentration (cmc), aggregation of membrane protein and lipid occurs as result of the hydrophobic effect. Cyclodextrin is an efficient agent for detergent adsorption, and has recently been used for 2D crystallization (5). These ring-shaped molecules exposing an inner hydrophobic surface bind detergent molecules AT stoichiometric ratios. Detergent removal is controlled by addition of cyclodextrin to the ternary mixture containing detergent, lipid and membrane protein. Based on this, a 2DX robot was developed that operates on a 96 well plate format and which is programmed to achieve the optimum detergent removal rate or assembly kinetics. Initial experiments with several aquaporins and with aerolysin have demonstrated superior crystal growth than with conventional methods. The software is critical for efficiently extracting the information buried in images and diffraction patterns, and to merge these experimental data into a 3D potential map. We are currently implementing the BASIC ideas of the MRC electron crystallography package in a modern software that is easier to use and runs on all platforms. Thus, processing of large stacks of images or diffraction patterns can be automated, thereby allowing data to be extracted from noisy images at an enhanced rate (6).


  1. Braun, T. & Engel, A. In Encyclopedia of Life Sciences., pp. A000304,2005.
  2. Gonen, T. et al. Nature 438, 633-8, 2005.
  3. Remigy, H. et al. FEBS Lett 555, 160-9, 2003.
  4. Rigaud J.L. , et al. J Struct Biol 118, 226-235, 1997.
  5. Signorell, G. A. et al. J Struct Biol 157, 321-8, 2007.
  6. Philippsen, et al. J Struct Biol 157, 28-37, 2007.

Specific achievements and potential of AFM and SMFS

AFM and SMFS provide novel bionanotechnological approaches to characterize the structurefunction relationship of MPs [1,2]. In all measurements the MPs remain embedded in the membrane bilayer and exposed to buffer solution at ambient temperatures [2]. It is not required to label, stain or to fix the protein membranes, which are just investigated as they are, structurally intact and functionally fully active. Both approaches high-resolution AFM imaging and SMFS of MPs hale been established by the scientific groups of A. Engel and D. Muller. High-resolution AFM imaging demonstrated its capability to observe single MPs at sub-nanometer resolution. The superb signalto- noise ratio of the AFM allows single protruding peptide loops connecting alpha-helices or betastrands to be visualized on single proteins embedded in the bilayer. Hence, crystalline order is not required for such analyses, although tight packing of the MP is advantageous. In many applications AFM has been applied to visualize oligomeric states and structural assemblies of functional MPs [3]. Simultaneously when being imaged at high-resolution, AFM enables to detect and map conformational fluctuations of delicate structures and to determine local flexibilities of the protein surface [2]. Since AFM allows investigating MPs in their physiological relevant environment it became possible to directly follow functional related conformational changes of native MPs AT molecular resolution. Examples encompass gap junctions from the rat liver cell [4], OmpF porin from E. coli [2], and a bacterial surface layer [2].

Figure 3

Figure 3: Observing the oligomeric state, supramolecular assembly and function of native membrane proteins by AFM. (A) Proton-driven rotors from spinach chloroplast FoF1-ATP synthase [3]. (B) Sodium-driven rotors from Ilyobacter tartaricus FoF1-ATP synthase. (C) High-light-adapted native photosynthetic membrane from Rsp. Photometricum[3]. (D) Pore complexes of perfringolysin O (PFO), a prototype of the large family of pore-forming cholesterol-dependent cytolysins (CDCs). Image courtesy of Z. Shao (Virginia). (E) The oligomeric state of bovine rhodopsin in native disc membranes [3]. (F) Structrual organsiation of the light-harvesting complex I photosynthetic core complex of Rsp. Rubrum [3]. (H) Extracellular surface of gap junction hemichannels from rat liver Wells recorded at pH 7.6 [6]. In presence of aminosulfonate compounds the hemichannels open their channel entrance with increasing pH from the closed (pH 6.0) to open (pH 7.6).

Applied to MPs SMFS enables to detect their unfolding and folding pathways, their structural stability, and to probe their energy landscape and refolding kinetics [5]. SMFS detects molecular interactions stabilizing structurally stable segments such as transmembrane α-helices, polypeptide loops or fragments hereof. The strength of these molecular interactions stabilizing the MP structure depend on environmental variations such as the oligomeric assembly of MPs, temperature changes, point mutations, electrolyte and pH variations [6]. Recently, both groups for the first time followed the ligand binding of a single Na+-ion to individual antiporters. The strength and location of the molecular interaction caused by the ligand binding could be mapped onto the protein structure at a precision of ±2 amino acids [8]. It became also possible to observe inhibitor binding and deactivating the same antiporter [7]. Additionally, the SMFS spectra reveals the “fingerprint” of the functional state of a protein. This functional fingerprint can be used to identify whether a single antiporter is in the activated, deactivated or inhibited state [5]. Similar experiments characterizing functionally important interactions of bovine rhodopsin have been established by both groups as well [8]. These molecular interactions stabilize the three dimensional structure of the GPCR and hold highly conserved residues at the place needed for proper functioning [5,8]. The relevance of Zinc stabilizing these structures could be demonstrated and the importance of the highly conserved disulfide bond to stabilize the molecular interactions traced [9]. Such insights deliver detailed insights into how molecular interactions stabilize secondary structure elements of MPs and how they switch their functional state.

In this network project the groups of A. Engel and D. Muller will further develop AFM and SMFS to characterize the structure-function relationship of native MPs and protein membranes. In particular high-resolution AFM will be developed further to observe the native oligomeric state and supramolecular assembly of proteins in reconstituted and native cell membranes. Applying novel ultrafast and –sensitive AFM imaging techniques will improve the capability to observe functional related conformational changes and dynamic assemblies of MPs. SMFS will be further developed to detect and to understand the molecular interactions that drive the supramolecular assembly and switch the functional state of MPs. Network members will be teached to apply the approaches and technology developed to image their MPs and cell membranes at high-resolution. Basic teaching will focus applying AFM and SMFS on well established and characterized samples such as bacteriorhodopsin of the native purple membrane, bovine rhodopsin being embedded in the native disc membrane of eye, and OmpF from E.coli. Furthermore, the A. Engel and H. Vogel labs will express, purify, reconstitute and provide MP samples, which have not been structurally characterized so far. In the lab of H. Vogel native MPs and native protein membranes will be functionally characterized. A. Engels lab will further focus onto the 2D crystallization of these examples for their structural assessment by EM crystallography. Advanced teaching of the students will include adaptation and development of the application of AFM and SMFS to answer key questions on new hitherto not characterized MPs and protein membranes.


  1. Drake et al., Science 243, 1586-8 (1989).
  2. Engel, A et al., Nature Structural Biology 7, 715-8, (2000).
  3. Seelert, H et al., Nature 405, 418-9, (2000); Fotiadis, D. et al., Nature 421, 127-8 (2003); Fotiadis, D. et al., J Biol Chem 279, 2063-8 (2004); Scheuring S. et al., Science 309, 484-7 (2005).
  4. Müller D.J. et al., EMBO J 21, 3598-607 (2002); Yu, J. et al., J Biol Chem 282, 88905-904 (2007).
  5. Kedrov, A. et al., Annu Rev Biophys Biomol Struct 36, 233-60 (2007).
  6. Janovjak H., et al., EMBO J 22, 5220-9, (2003).
  7. Kedrov A., et al., et al., EMBO reports 6, 668-74 (2005).
  8. Fotidais, D. et al., Curr Opin Struct Biol 16, 252-9 (2006).
  9. Park, P., et al., J Biol Chem 282, 11377-85 (2007).

Specific achievements and potential of NMR

Nuclear Magnetic Resonance (NMR) provides a spectroscopic method to study molecular structure and dynamics in solution and/or non-crystalline environment (solid-state NMR). Thus far, protein structures determined by NMR have mostly been obtained using aqueous solutions1 or, in the context of membranes, using protein preparations in a detergent or micelle environment. In the latter case, 3D structures for several β-barrel MPs and a pentameric α-helical protein could be determined (for a recent review see Ref.2). In addition, solid-state NMR methods become increasingly useful to not only study lipid bilayer itself3 but also to determine molecular structures and dynamics of membrane embedded proteins (see, e.g., Ref. 4 and references therein). In general, solid-state NMR studies can be conducted using microcrystalline or frozen samples and they can also be extended to lipid-bilayer preparations5. Comparison between detergent and lipid bilayer preparations can, for example, be used to delineate the influence of the surrounding lipid matrix upon protein structure, as recently demonstrated6 for a receptor-transducer complex of sensory rhodopsin from Natronomonas pharaonis (Figure).

Figure 4

Figure 4: Cartoon representation of solubilized and reconstituted samples of the sensory rhodopsin receptor (dark blue) – transducer (light blue) complex. Both proteins are first isolated independently in micelles (orange), then mixed together to form a 1:1 complex. Such preparations could be used for solutionstate NMR studies. After reconstitution into purple membrane lipids (pink), solid-state NMR studies can be performed. Figure adapted from Ref. 6

NMR hence offers complementary spectroscopic means to study structure, ligand binding, complex formation or the influence of the surrounding lipid matrix on the atomic level. Often, such studies can be conducted in a functional environment. Because of the intrinsic short-range order of NMR interactions such approaches are highly complementary to microscopy methods. Crystallinity is not a requirement making NMR studies also highly useful In combination with diffraction techniques.

Progress in NMR of MPs is intimately related to methodological and instrumental advancements and heavily relies on molecular biology tools that provide increasing possibilities to express, solubilize and isotope-labeled MPs for NMR studies. An initial training network as outlined here would enable students and postdocs do learn and apply these methods in European laboratories that are world-wide leaders in the field of NMR of MPs. In contrast to the current situation where most groups are experts in a limited number of research aspects, this strategy would create researchers that are not only familiar with a whole arsenal of state-of-the-art NMR methods but it would allow them to combine NMR with a variety of other biophysical approaches ranging from microscopy to molecular dynamics simulations to study MPs with unprecedented detail and comprehensiveness.


  1. K. Wüthrich, NMR of proteins and nucleic acids, Wiley Interscience, New York, 1986.
  2. L. K. Tamm and B. Liang, Progress in Nuclear Magnetic Resonance Spectroscopy, 2006, 48, 201-210.
  3. J. Seelig, Quarterly Reviews of Biophysics, 1977, 10, 353-418.
  4. S. J. Opella and F. M. Marassi, Chemical Reviews, 2004, 104, 3587-3606.
  5. M. Baldus, Current Opinion in Structural Biology, 2006, 16, 618-623.
  6. J. P. Klare, E. Bordignon, M. Doebber, J. Fitter, J. Kriegsmann, I. Chizhov, H. J. Steinhoff and M. Engelhard, Journal of Molecular Biology, 2006, 356, 1207-1221.

Specific achievements and potential of theory and modelling

Molecular modelling approach is nowadays routinely used for investigation of many aspects of MP research like protein structure prediction, stability and dynamics, and finally modelling of protein-ligand and protein-protein complex structures [1]. Simulations can provide dynamical view of proteins and can be used to build models of other conformational states (like open and closed conformations of channels or substates during receptor activation) and explore different modes of binding in protein-ligand and protein-protein complexes. Mostly unexplored protein-lipid interface (because of lack of experimental structures) can affect the topology, stability, assembly, traffic and enzymatic activity of membrane proteins [2]. Increasing computer capacity allows unbiased simulations of lipid and membrane-active peptides. With an increasing number of high resolution structures of MPs, which also enables homology modelling of more structures, a wide range of MPs can now be simulated over time spans that capture essential biological processes [3]. TransMPs comprise 20-30% of the genome but, because of experimental difficulties, they represent less than 1% of known structures. Lack of MP structures makes computational prediction an important method for obtaining novel structures. Recent advances In computational techniques hale been combined with experimental data to constrain modelling 3-D structures of proteins [4] and also their dimers and higher oligomers [5].

Figure 5

Figure 5: Oligomer of rhodopsin. Ellipses mark rhodopsin dimers. (a) AFM image of the membrane of native dinks from rod outer segments. (b) Theoretical model (1N3M) of rhodopsin oligomer (after removal of cytoplasmic loops of rhodopsin) simulated in the membrane. Small balls denote positions of phosphorus atoms in phospholipids, red  etanolamine, green – serine headgroups.

Rhodopsin proved to be useful not only as a template for building other G protein-coupled receptors (GPCRs) but also their function. Currently it is believed that most GPCRs exist and act as dimers. Based on AFM images of rhodopsin oligomers in native rod cell disc membranes it was possible to design a model of such oligomer (1N3M in Protein Data Bank) [6] which was later confirmed experimentally by mutagenesis and crosslinking experiments not only for rhodopsin but also for dopamine D2 receptor. We also investigated a functional role of oligomeric state of rhodopsin and built structures of rhodopsin complexes with its G protein [7], rhodopsin kinase [1] and recently with arrestin [8]. Simulation methods helped us to investigate mechanisms that are responsible for recognition of activated rhodopsin by G protein and then, after phosphorylation, by arrestin. Using special techniques like steered molecular dynamics we were able to trace movements of the whole domains of proteins. Subsequent simulations including explicite membrane will help to understand many aspects of membrane proteins like stability, mechanism of signal passing and also facilitate drug design.


  1. Filipek S et al. in Structural Genomics on Membrane Proteins (Lundstrom KH, Ed.) pp 331-48, Taylor and Francis, (2006).
  2. Schneiter R et al., Appl. Microbiol. Biotechnol., 73, 1224-32, (2007).
  3. Ash WL et al., BBA-Biomembr., 1666, 158-89, (2004).
  4. Fleishman SJ et al., Curr. Opin. Struct. Biol., 16, 496-504, (2006).
  5. Reggio PH et al., AAPS J., 8, E322-36, (2006).
  6. Fotiadis D et al., Nature, 421, 127-8, (2003).
  7. Filipek S et al., Photochem. Photobiol. Sci., 3, 628-38, (2004).
  8. Modzelewska A et al., Cell Biochem. Biophys., 46, 1-15, (2006).