Innovatives Supercomputing in Deutschland
inSiDE • Vol. 11 No. 1 • Spring 2013
current edition
about inSiDE
index  index prev  prev next  next

Dynamics and interaction of membrane-spanning protein helices in health and disease

Proteins and Lipid Membranes

Proteins perform countless functions thus enabling living cells to maintain their metabolism, to grow, to communicate with other cells etc. About 25% of all proteins in any organism are embedded within lipid membranes. Membranes define and structure individual cells and are built from a great variety of different lipid molecules. The task of membrane proteins is to mediate regulated fusion of cells, allow for solute exchange with the cell exterior, regulate release of cellular products, mediate lipid translocation from one side of the membrane to the other one, cleave other membrane proteins etc. Most membrane proteins are anchored by helical transmembrane domains (TMDs, Fig. 1). Long regarded as dull membrane anchors, TMDs are increasingly recognized to guide assembly of subunits to larger functional complexes and to be dynamic [1]. Molecular Dynamics (MD) simulations done at single-atom resolution are well-suited to investigate molecular properties of protein helices, surrounding lipids and solvent by computing time series of structures. From these ensembles of structures conformational and dynamical properties are evaluated and related to the underlying molecular interaction forces. For the production runs (typically 100,000 atoms for > 150 ns) we use the description of atomic interactions given by the CHARMM forcefield [2,3] and the MD algorithm as implemented in NAMD [4]. Each simulation project compares a large set of TMDs with different amino acid compositions and sequences in isotropic solvent as well as in membranes of different lipid composition. Validation of the simulations by experimental measurements connects models of exquisite structural detail to physical reality.

One production run uses up to 120 cores at the fat node island of SuperMUC and consumes approximately 100,000 hours of CPU time (three months wall-clock time). In contrast to the LRZ Ultra-Violett Linux cluster and the HLRB II system, the short queue times at the SuperMUC HPC system allowed us to run the simulations in a relative short time. Due to the large number of cores per machine and the high amount of latent memory, the fat nodes are quite perfect for NAMD with its high number of I/O operations and CPU communications, highly depending on low interconnection latencies. First try runs with the Sandy Bridge architecture of SuperMUC’s thin nodes showed very similar performance which will allow us to extend our simulations to larger membrane patches and a higher number of embedded transmembrane helices which require more than 1,000 cores.

TMD Helix-Helix Interactions can Support Protein/Protein Interactions

Most membrane proteins form multi-subunit complexes and an hitherto unknown number of them do so as a result of specific TMD helix-helix interactions. MD simulations can provide models of the interfaces and the physical forces that govern the interaction. In one example, we have modeled the sequence-specific interaction of the TMD of sulphhydryl oxidase, a membrane protein that aids folding of other proteins. Experimental work had identified amino acids that are responsible for mutual recognition of these helices and computational work resulted in a homodimeric structure that matches the experimental results (Fig. 2).

Structural Dynamics Supports Protein Function

While the 3-dimensional structures of tens of thousands of different water-soluble proteins have been determined within the last decades, it is becoming increasingly clear that the functional cycles of many proteins require conformational transitions between different structural states, i.e., structural dynamics. Likewisely, membrane proteins are increasingly recognized as dynamic entities. TMDs are particularly important for the function often being located at the structural interface between ligand-sensing extracellular and intracellular signaling domains. TM helices can change their orientations relative to each other on µs time scales; in addition, the helices themselves are intrinsically dynamic structures in a sense that they exhibit local dynamics (like transiently hydrogen-bond opening and side-chain rotations) as well as global motions (like bending, twisting and elongation) within ns (Fig. 3). TMD interaction and dynamics also appear to affect their ability to serve as substrates for intramembrane proteases. Specifically, both factors have been suggested to define the specificity by which the TMD of amyloid precursor protein is cleaved to fragments eliciting Alzheimer’s disease.

Figure 1: A pictorial view on a membrane protein anchored with a helical transmembrane domain (red) in a lipid patch (grey, only shown partially).

Transmembrane Helix Dynamics Depends on Amino Acid Sequence

While there is ample evidence for the functional significance of the slow (µs) relative movements of TMDs, only little is known about the role of fast (ns) helix dynamics. MD simulations provide rich molecular detail at the ns time scale and are thus employed by us to study local TM helix dynamics. In a systematic analysis, we have designed a large set of low-complexity model TMDs de novo and modeled their dynamics in isotropic solvent. Interestingly, the results show that the structural fluctuations along the helix strongly depend on their sequences (Fig. 4) Thus, helix dynamics is more specific than just Brownian motion; rather, its sequence-dependence shows that it is a structural property that can be optimized by successive mutations in evolution and thus support the functional diversification of membrane proteins. Our results show [5] that the backbone dynamics i) increases at the termini of the helices and ii) depends on the primary structure. The dependence on sequence is particularly important since the helix-destabilizing amino acid types found here have previously been found to be overrepresented in the TMDs of membrane fusogenic proteins and isolated TMDs are known to induce membrane fusion in vitro. Further, this work yielded mechanistic insights into the molecular determinants of backbone dynamics. Specifically, we found that the nature of the amino acid side chains affects the flexibility of the helix. In essence, those side chains form a molecular ‘scaffold’ around the central invariant backbone of a helix and the stability of that scaffold is determined by side-chain size, structure and dynamics which determine the mutal interactions.

These MD calculations closely fit experimental work (CD spectroscopy, hydrogen/deuterium-exchange kinetics). While the overall time scale for isotope exchange is determined by the slow chemical exchange (time scale of hours), the variations along the sequence depend on local dynamics which is reliably sampled within the 100 ns MD trajectories. Therefore the results from MD could be used to calculate and predict experimental data recorded on a much longer time scale with excellent accuracy (Fig. 5).

Figure 2: Transmembrane helix-helix interactions stabilize the dimer of sulphhydryl oxidase. The residues of one of the two helices responsible for mutual recognition are shown as blue spheres. Transparent shapes around the atoms represent their van der Waals radii. The second helix is represented by its molecular surface (light grey). The C-terminus of the helices is at the top.

Helix Dynamics in Lipid Membranes

Another question relates to the potential impact of TMDs on the structure of the membrane and vice versa. Simulating our model TMDs in lipid bilayer patches indicate that the dynamics of membrane-embedded helices are strongly reduced by the membrane. This is expected due to the absence of destabilizing water from the membrane interior. Further, the results nicely demonstrate specific molecular bonding between protein and lipids such that local protein/lipid networks are generated. Further, insertion of the TMDs has profound consequences for the structure of the membrane. Membranes are highly fluid as their constituent lipids are quite flexible themselves. Indeed, the long-range order, i.e., the way lipids are organized, changes drastically around a TMD (Fig. 6). Changes in membrane order underly complex biological processes, such as membrane fusion and lipid exchange. This work is thus expected to shed light on the mechanisms by which membrane proteins catalyze these functions.

Figure 3: Patch of a plasma membrane and magnifications illustrating backbone dynamics and interactions of monomeric and dimeric transmembrane helices.

Helix Dynamics in Alzheimer’s Disease

Alzheimer’s disease affects currently ~40 million people on earth and is expected to increase dramatically as a result of increasing life-spans. Although the etiology of the disease is still controversial, it is clear that cleavage of the Amyloid precursor protein (APP) TMD by γ-secretase is an essential step leading to it. It is believed that this TMD forms a dimer which is cleaved at various sites. Successive cleavage leads to the build-up of Aβ peptides in the brain that ultimately kill nerve cells. The toxicity of these peptides depends on their size and size depends on the site of cleavage. Thus, understanding how individual cleavage events are influenced by the structure and dynamics of the APP TMD helix will provide a rationale for understanding a key step in this disease. Simulations of the APP TMD revealed a number of surprising results [6]. For one thing, that part of the helix that engages in forming the dimer is much more dynamic than the actual cleavage domain. Possibly, a highly flexible dimerization domain facilitates movement of the helix within γ-secretase or the final release of cleavage products. The nature of cleavage products, i.e., their length, may be determined by the relative helix dynamics at the site where cleavage begins. Indeed, we find that a number of parameters describing local helix unfolding point to enhanced dynamics, viz. cleavage, at the site leading to the less toxic product (Aβ 40) compared to the site prompting the slightly longer and more toxic Aβ 42. Comparing monomeric and dimeric helices reveals another interesting feature. While dimerization rigidifies the dimerization region, as expected, it destabilizes the cleavage domain. A more detailed analysis of the trajectories suggests that this destabilization is due to increased solvation within the dimer. Thus, dimerization could actually promote initial cleavage. Contrary to expectation, the dynamics of the natural substrate, the APP TMD, is similar to that of several non-substrates. However, we find that the APP TMD dynamics is unleashed by certain amino acid exchanges. Some of these exchanges are also known to change the cleavage pattern. Taken together, it appears as if local changes in backbone dynamics can have a pronounced impact on the efficiency of the cleavage reactions. There are several hereditary forms of Alzheimer’s disease that are caused by individual amino acid exchanges within the APP TMD and that are characterized by profoundly changed ratios of the Aβ peptides. Elucidating whether such exchanges exert their effects via changing helix dynamics will be a rewarding practical application of this work.

Figure 5: Site-specific helix dynamics for TMD model peptides with increasing content of helix-destabilizing residues (for sequences see Fig. 4) (a) Local deuterium-hydrogen exchange rates calculated from MD. Exchange competence is determined by the fraction of open intrahelical hydrogen-bonds and the presence of water molecules in close contact to the backbone amide groups. Both properties can be reliably retrieved from MD, as is demonstrated by the close correspondence between experimental and calculated exchange kinetics in (b). Experimental values are shown in dark grey, MD-derived values are shown in colors.
Figure 6: The local lipid composition around a transmembrane helix differs drastically from the bulk values and depends on the orientation of the TMD (N-terminal, C-terminal) as well as on the lipid type (DOPC, DOPS, DOPE). Interaction modes are exemplified by lipids complexed with the charged residues (lysine, grey and blue spheres) flanking the hydrophobic core (grey trace) of oligo-leucine L16: (A) hydrogen-bonding to the lipid phosphate group (orange), (B) hydrogen-bonding to lipid carbonyl groups (red), (C) indirect interaction of lipids in the second shell.


[1] Langosch, D., and Arkin, I.T. (2009). Interaction and Conformational Dynamics of Membrane-Spanning Protein Helices. Protein Sci. 18, 1343-1358.

[2] MacKerell, Jr., A. D., Bashford, D., Bellott, M., Dunbrack Jr., R.L., Evanseck, J.D., Field, M.J., Fischer, S., Gao, J., Guo, H., Ha, S., Joseph- McCarthy, D., Kuchnir, L., Kuczera, K., Lau, F.T.K., Mattos, C., Michnick, S., Ngo, T., Nguyen, D.T., Prodhom, B., Reiher, III, W.E., Roux, B., Schlenkrich, M., Smith, J.C., Stote, R., Straub, J., Watanabe, M., Wiorkiewicz-Kuczera, J., Yin, D., and Karplus, M. (1998) All-atom Empirical Potential for Molecular Modeling and Dynamics Studies of Proteins. J. Phys.Chem. B 102, 3586-3616.

[ 3] Feller, S., and MacKerell, Jr., A.D. (2000) An Improved Empirical Potential Energy Function for Molecular Simulations of Phospholipids, J. Phys. Chem. B 104, 7510-7515.

[4] Phillips, J.C., Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E., Villa, E., Chipot, C., Skeel, R.D., Kalé, L., and Schulten, K. (2005). Scalable Molecular Dynamics with NAMD. J. Comp. Chem., 1781-1802.

[5] Quint, S., Widmaier, S., Minde, D., Langosch, D., and Scharnagl, C. (2010). Residue-Specific Side-Chain Packing Determines Backbone Dynamics of Transmembrane Model Helices. Biophys. J. 99, 2541-2549.

[6] Pester, O., Barret, P., Hornburg, D., Hornburg, P., Pröbstle, R., Widmaier, S., Kutzner, C., Dürrbaum, M., Kapurniotu, A., Sanders, C.R., Scharnagl, C., and Langosch, D. (2013). The Backbone Dynamics of the Amyloid Precursor Protein Transmembrane Helix Provides a Rationale for the Sequential Cleavage Mechanism of γ-Secretase. J. Am. Chem. Soc. 135, 1317-1329.

• Dieter Langosch (1)
• Matthias Mörch (1)
• Christina Scharnagl (2)
(1) Technische Universität München, Lehrstuhl für Chemie der Biopolymere
(2) Technische Universität München, Fakultät für Physik

top  top