5864
Biochemistry 2009, 48, 5864–5873 DOI: 10.1021/bi900493n
Membrane Attachment Facilitates Ligand Access to the Active Site in Monoamine Oxidase A† Rossen Apostolov,‡ Yasushige Yonezawa,‡ Daron M. Standley,‡ Gota Kikugawa,§ Yu Takano,‡ and Haruki Nakamura*,‡ ‡ §
Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan, and Institute of Fluid Science, Tohoku University, 2-1-1 Katahira-Aobaku, Sendai, Miyagi 980-8577, Japan Received March 23, 2009; Revised Manuscript Received May 15, 2009
ABSTRACT: Monoamine oxidase membrane enzymes are responsible for the catalytic breakdown of extra- and
intracellular neurotransmitters and are targets for the development of central nervous system drugs. We analyzed the dynamics of rat MAOA by performing multiple independent molecular dynamics simulations of membrane-bound and membrane-free forms to clarify the relationship between the mechanics of the enzyme and its function, with particular emphasis on the significance of membrane attachment. Principal component analysis of the simulation trajectories as well as correlations in the fluctuations of the residues pointed to the existence of three domains that define the global dynamics of the protein. Interdomain anticorrelated movements in the membrane-bound system facilitated the relaxation of interactions between residues surrounding the substrate cavity and induced conformational changes which expanded the active site cavity and opened putative pathways for substrate uptake and product release. Such events were less pronounced in the membrane-free system due to differences in the nature of the dominant modes of motion. The presence of the lipid environment is suggested to assist in decoupling the interdomain motions, consistent with the observed reduction in enzyme activity under membrane-free conditions. Our results are also in accordance with mutational analysis which shows that modifications of interdomain hinge residues decrease the activity of rat MAOA in solution.
Monoamine oxidase (MAO)1 enzymes catalyze the oxidative deamination of various biogenic monoamines (1, 2). They break down neurotransmitters such as serotonin and dopamine in the synaptic cleft and thus regulate the function of the central nervous system. Enzyme malfunction leads to various psychological disorders such as aggressive behavior, criminality, social phobias, depression, and substance abuse, which has made these proteins one of the major targets for antidepressant drugs (3, 4). MAO inhibitors have also been successfully used for suppressing the production of neurotoxins like MPP+ (5, 6), an agent directly linked to the development of Parkinson’s disease. Moreover, they have found application in the treatment of oxidative stress conditions (7, 8) because one of the products of the amine oxidation is hydrogen peroxide, a very powerful source of † We are grateful to the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan for a Grant-in Aid for Scientific Research on Priority Area “Structures of Biological Macromolecular Assemblies” (513-18054013 and 513-20051013). *To whom correspondence should be addressed. E-mail: harukin@ protein.osaka-u.ac.jp. Telephone: +81 (0)6 6879 4311. Fax: +81 (6) 6879 8636. 1 Abbreviations: MAO, monoamine oxidase; MD, molecular dynamics; PCA, principal component analysis; ENM, elastic network model; POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; FAD, flavine adenine dinucleotide; rmsd, root-mean-square deviation; TM, transmembrane.
pubs.acs.org/Biochemistry
hydroxyl radicals. Clinical trials have given hope for their possible use as agents against Alzheimer’s disease (9), Huntington’s disease (10), and amyotrophic lateral sclerosis (11). The spectrum of serious diseases that can possibly be treated by controlling MAO function has stimulated extensive research, including mutagenesis studies (2) and docking simulations (12-15) in the hope of improving our understanding of their structure-function relationship and kinetics. MAOs comprise two isoenzymes, MAOA and MAOB, which exhibit different substrate sensitivities due to differences in the size, shape, and chemical environment of the active site pockets (16). MAOA is also the first membrane protein discovered to form a single transmembrane helix, clearly seen in the crystal structure of rat MAOA (17). Experiments have suggested that membrane incorporation is important for protein function since, under membrane-free conditions, rat MAOA (18) and human MAOA (19) activities drop approximately 4 times, while truncation of the transmembrane helix of human MAOB decreases its activity by a factor of 10 (20). Many of the MAO structures are resolved at very high resolution and provide detailed information about the active site, which is needed for successful design of new inhibitors. Unfortunately, due to difficulties in obtaining noncoagulating apo forms, all of the deposited structures contain various bound inhibitors. The substrate binding regions are found deep in the interior of the extracellular domain unit, and
Published on Web 05/20/2009
r 2009 American Chemical Society
Article there is no easily identifiable route for substrate access to the active site. The lack of inhibitor-free crystal structures makes it difficult to infer the conformational changes that define the function of the enzyme. These changes may include collective displacements of domains to facilitate access to the catalytic region for substrate attack. It is possible to identify such domains and the direction of their displacements using several computational methods. In this work, we analyzed the dynamics of rat MAOA using molecular dynamics (MD) simulations (21) of the full-length protein embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer and of the membrane-unbound system by truncating the extracellular unit in solvent, with principal component analysis (PCA) of the trajectories. Rat MAOA was chosen because, at the time, the crystal structure [Protein Data Bank (PDB) entry 1o5w] was the only member of the MAO family that had a resolved transmembrane helix (TMH). The length of the TMH is essential in investigating how the protein dynamics is affected by membrane insertion. Moreover, the CR rmsd between 1o5w and a more recently determined structure (PDB entry 2z5x) is less than 0.7 A˚. Thus, although the resolution of 1o5w (3.2 A˚) is not as good as that of 2z5x (2.2 A˚), we are sure that our statistical analyses based on time averages of structural ensembles derived from MD simulations are sufficient. The results from the PCA analysis, along with elastic network model (ENM) analysis (22, 23), suggested the existence of three dynamical domains within the extracellular unit. The dynamics of these domains differed significantly, depending on whether the protein was embedded in the membrane, suggesting that membrane anchoring might have a direct role in enhancing enzyme function. METHODS The initial structure of the rat MAOA monomer was obtained from the Protein Data Bank [entry 1o5w (17)]. The crystal structure does not include the first nine N-terminal and last six C-terminal residues; therefore, they were excluded from the model. Instead, acetyl and N-methylamine residues were used to cap the termini. Missing hydrogen atoms were added using the tplgene module of the myPresto simulation suite (24). The flavine adenine dinucleotide (FAD) cofactor was added to the simulation system after structure optimization and electrostatic potential (ESP) charge calculation using Gaussian (25) with the 6-31G* basis set. No inhibitor or substrate was included in the model. Rat MAOA forms a homodimer in the crystal structure. However, other homologues, such as human MAOA, are found to crystallize and retain catalytic activity in vitro as monomers (16), while bovine MAOB is shown to exist and function as a tetramer or larger oligomeric complex (26). To the best of our knowledge, there is no evidence showing that dimerization is required for rat MAOA to function. Therefore, we performed all simulations on the monomer. Atomic interactions were modeled according to the CHARMm22 force field (27). Polar hydrogen bonds, including water molecules, were treated as rigid bodies which allowed for a 2 fs time step. Rigid body treatment was based on integration of the Euler equations of motion of a rigid body, as described in ref (28). Electrostatic interactions were calculated using the smooth particle-mesh Ewald method (29) with a grid size of 1 A˚ per cell. A cutoff distance of 12 A˚ was used for the Ewald real space and van der Waals truncation. The Ewald
Biochemistry, Vol. 48, No. 25, 2009
5865
convergence parameter R was set to 0.35 A˚-1. All calculations were conducted with the cosgene MD engine of myPresto (24) on three hardware platforms: SGI Altix 4000 server, Intel Xeon 3.2 GHz cluster, and MDGRAPE-3 special purpose accelerator board (30). The membrane-bound system was modeled by embedding the monomer in a pre-equilibrated POPC lipid bilayer of 163 chains. Two opposing POPC chains, one from each leaflet of the bilayer, were removed, and the TM helix of MAOA was inserted into the cavity. The whole system was then solvated in 17125 water molecules. In addition, 76 sodium and 78 chloride ions were added to neutralize the total charge and bring the tonicity of the solvent to physiological levels of 0.9%. The complete system consisted of 81647 atoms. The size of the rectangular cell was 72 A˚ 94 A˚ 124 A˚. The energy of the system was locally optimized using 3000 steps of conjugated gradient minimization. This was followed by a 100 ps equilibration of the membrane chains and the solution under NPT (constant pressure and constant temperature) conditions, while keeping the positions of the protein and the cofactor atoms fixed. The constraints were then slowly relaxed for an additional 100 ps. Production runs were conducted without any constraints using an NPT ensemble. Temperature and pressure were controlled by coupling the system to a Nose-Anderson barostat and thermostat with coupling times of 2 and 0.1 ps, respectively, for achieving constant conditions of 1 bar and 310 K. A snapshot 2 ns from the trajectory, when the root-mean-square deviation (rmsd) of the CR atom fluctuations had reached a plateau, was used as a base for the construction of the membrane-unbound system. Residues from Leu501 to Ile521 that form the transmembrane (TM) helix were removed and substituted with an N-methylamine cap. Lipids were also removed and replaced with water. The new system had dimensions of 72 A˚ 92 A˚ 96 A˚ with 63605 atoms. Energy minimization and equilibration were performed in the same way as in the membrane-bound system. We performed three independent, 20 ns long production runs for each of the systems for a total of 120 ns of trajectory data. For each run, we used a different random distribution of initial velocities. Trajectory snapshots were taken every 2 ps for subsequent analysis. Conformational changes of the protein during the six simulation trajectories were monitored by calculating the rmsd of the CR atoms after aligning them to the initial minimized structure using the standard method for mass-weighted least-squares fitting. The degree of collective motions between different groups of atoms was assessed from the cross-correlation matrix with elements ri -Æ! ri æÞ 3 ð! rj -Æ! rj æÞæ= Corrij ¼ Æð! qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! ! ! ! Æð ri -Æ ri æÞ2 æÆð rj -Æ rj æÞ2 æ Each element represents the correlation of the movements of atoms i and j. In the case of fully linearly correlated movements, Corrij equals 1, while in the case of anticorrelated motions it is -1 and in the case of noncorrelated or perpendicular motions 0. The simulation trajectories were analyzed for dominant collective displacements using PCA of the fluctuations of CR atoms (31, 32). The average structure of each trajectory was projected along chosen high-eigenvalue modes. The two extreme projections were then used for dynamical domain analysis using DynDom (33). Structural domains in the static crystal structure were assigned using Protein Domain Parser (PDP) (34). The search for putative
5866
Biochemistry, Vol. 48, No. 25, 2009
Apostolov et al.
routes leading to the active site was done using Caver (35), which provides the direction, size, and identity of atoms defining the tunnel walls. Each structure was searched for multiple tunnels sorted by the size of the narrowest part of the tunnel. Water, lipid, and ions were removed from the trajectory snapshots, and the search was performed for six tunnels starting at 4 A˚ from the riboflavin reactive site of FAD. The ENM method was implemented in the myPresto program suite following the algorithm in the ElNemo (22) server. We used only the CR atoms of each residue of the crystal structure and let them interact with other residues within a 9 A˚ cutoff via a spring potential (23). They were assigned the same mass, 12.01 amu. Figures of the MAOA structure were prepared using VMD (36). Sequence alignment was done using clustalw (37) and seaview (38). Graphs were prepared using gnuplot (39), and the manuscript was prepared with Lyx (40) and Jabref (41). RESULTS AND DISCUSSION Overview of MAOA Topology. Most of the mass of MAOA is concentrated into a compact extracellular structure, which is attached at its C-terminus to the membrane through a single 21-residue transmembrane R-helix. The FAD cofactor and the active site are buried deep in the interior of the protein. The structure can be functionally divided into two domains, the FAD-binding domain (domain F, residues 10-50, 227-287, and 427-473) and the substrate binding domain (17). Analysis of MD simulation data as well as analysis of the crystal structure using PDP suggests that the substrate binding domain is actually composed of two subdomains. We will refer to these as membrane binding domain M (residues 112-208, 288-300, and 408426) and substrate binding domain S (residues 51-111, 209-226, 301-407, and 474-486). As indicated by the residue ranges given above, the domains consist of noncontinuous segments, implying a tight coupling between them. The domains share a large amount of surface area, and their separation is not obvious from the static structure. Domain M is thought to be partially embedded in the membrane (17), as confirmed by MD simulations of the highly homologous MAOB (42). In Figure 1a, a schematic view of the major structural elements is shown. A snapshot at 5 ns from the membrane-bound system trajectory is shown in Figure 1b. Disordered outer loop L, transmembrane helix TM, and the active site pocket are also shown with the corresponding crystal structure (Figures 1c,d). Residues lining the active site cavity were defined as those for which at least one atom could be found within 6 A˚ of the inhibitor in the crystal structure. These residues form two walls that are part of domains M and S. The first wall comprises residues entirely from substrate binding domain S: 97, 108, 305, 323-325, 335-337, 350, and 352. The second wall is constructed by residues mainly from membrane binding domain M: 66-69, 197, 207-210, 215, 406, and 444. Residues 110, 111, and 180-182 are positioned between the two walls, on opposite sides of the active site cavity. Dynamics and Stability of the Membrane-Bound and Membrane-Unbound Enzymes. After
