Conformational Behavior of Flavin Adenine Dinucleotide: Conserved

Nov 12, 2014 - Metabolic enzymes utilize the cofactor flavin adenine dinucleotide (FAD) to catalyze essential biochemical reactions. Because these enz...
2 downloads 0 Views 5MB Size
Download PDF
Article pubs.acs.org/JPCB

Conformational Behavior of Flavin Adenine Dinucleotide: Conserved Stereochemistry in Bound and Free States Gopi Kuppuraj,*,† Dennis Kruise,†,∥ and Kei Yura†,‡,§ †

Center for Informational Biology, Ochanomizu University, 2-1-1 Otsuka, Bunkyo, Tokyo 112-8610, Japan Graduate School of Humanities and Sciences, Ochanomizu University, 2-1-1 Otsuka, Bunkyo, Tokyo 112-8610, Japan § National Institute of Genetics, Yata 1111, Mishima, Shizuoka 411-8540, Japan ∥ Institute for Life Science and Technology, Hanzehogeschool Groningen, 9747 AS/9700 RM Groningen, The Netherlands ‡

S Supporting Information *

ABSTRACT: Metabolic enzymes utilize the cofactor flavin adenine dinucleotide (FAD) to catalyze essential biochemical reactions. Because these enzymes have been implicated in disease pathways, it will be necessary to target them via FADbased structural analogues that can either activate/inhibit the enzymatic activity. To achieve this, it is important to explore the conformational space of FAD in the enzyme-bound and free states. Herein, we analyze X-ray crystallographic data of the enzyme-bound FAD conformations and sample conformations of the molecule in explicit water by molecular dynamics (MD) simulations. Enzyme-bound FAD conformations segregate into five distinct groups based on dihedral angle principal component analysis (PCA). A notable feature in the bound FADs is that the adenine base and isoalloxazine ring are oppositely oriented relative to the pyrophosphate axis characterized by near trans hypothetical dihedral angle “δV” values. Not surprisingly, MD simulations in water show final compact but not perfectly stacked ring structures in FAD. Simulation data did not reveal noticeable changes in overall conformational dynamics of the dinucleotide in reduced and oxidized forms and in the presence and/or absence of ions. During unfolding−folding dynamics, the riboflavin moiety is more flexible than the adenosine monophosphate group in the molecule. Conversely, the isoalloxazine ring is more stable than the variable adenine base. The pyrophosphate group depicts an unusually highly organized fluctuation illustrated by its dihedral angle distribution. Conformations sampled from enzymes and MD are quantified. The extent to which the protein shifts the distribution from the unbound state is discussed in terms of prevalent FAD shapes and dihedral angle population.

■

exposure by noncovalently binding to FAD.4 The mechanism occurs by electron transfer in which the isoalloxazine ring is activated by light energy and acts as an electron donor to repair the damaged DNA.5−9 Other enzymes that bind FAD, for example, succinate dehydrogenases, riboflavin synthases, and acyl CoA dehydrogenases, are involved in critical metabolic processes making them active drug targets.10−20 If the validity of these FAD-based metabolic targets or other disease areas is to be fully investigated, one of the approaches to target them is through designing FAD structure-based pharmacophores. Hence, it is important to elucidate various distinct con-

INTRODUCTION

The nucleotide coenzyme flavin adenine dinucleotide (FAD) catalyzes biological reactions involving electron transfer. FAD consists of a riboflavin group (a derivative of vitamin B2) connected to the terminal phosphate moiety of adenosine diphosphate (Figure 1). The riboflavin group itself is comprised of a linear alcohol chain, adonitol, and a flavin ring system called isoalloxazine. The catalytic function of the FAD is concentrated in this isoalloxazine ring. In catalytic reactions involving NAD(P)H as substrates, a hydride ion is transferred from or to the isoalloxazine ring, resulting in either reduction or oxidation of the substrate molecule in the enzyme’s active site.1 The presence of an isoalloxazine ring in the structure also allows FAD to have photochemical properties.2,3 For example, the enzyme photolyase repairs damaged DNA caused by UV © XXXX American Chemical Society

Received: July 29, 2014 Revised: October 25, 2014

A

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 1. Schematic diagram of FAD “backbone” structure. Hydrogen atoms, oxidation states (FAD2−/FADH2), stereochemistry, and valences are omitted for clarity. Adenosine monophosphate moiety (AMP) in red, riboflavin moiety (RFN) in blue. Dihedral angles about four chemical bonds (highlighted in bold) are indicated at the center bond pair and are defined in Supporting Information S1. Atoms of the virtual dihedral angle, δV, are colored in green.

field.40 Their rationale for using a NAD molecule instead of ATP is attributed to the structural similarity of NAD with FAD, where in the former the two aromatic bases are connected by two sugar groups linked through diphosphates. These two simulation studies show that FAD in extended conformation at the start folds to a closed conformation in which the isoalloxazine and adenine rings form stacking interactions. Most studies have focused on the comparative analysis of protein/enzyme-bound cofactors, namely, ADP, ATP, NAD(P)(H), FAD, and CoA molecules, all sharing the common adenine, ribose sugar, and diphosphate regions (see above). Some works are successful in comparing the conformation of the bound ligands and identifying common motifs involved, if any, in cofactor recognition and binding.1,41−48 The successful simulation works of van den Berg et al. and Radoszkowicz et al. are aimed at finding correlations with experimental steady-state and time-resolved fluorescence properties of FAD.39,40 In this work, we combine and compare X-ray crystallography records in the Protein Data Bank (PDB) to provide “static” features of FAD conformations found in proteins with the molecular dynamics simulations of FAD in water to analyze a range of flexible conformations. The properties of enzyme-bound conformations found in crystal structures are reported on the basis of dihedral angle principal component analysis between adenine and isoalloxazine rings, and the molecular dynamics simulations of free FAD in water are extended to cover 200 ns. In addition, the effects of the reduction of the FAD to FADH2 and role of counterions on the conformations have not been studied before. Therefore, we have analyzed conformations of FAD in both oxidation states, namely, FAD and FADH2. In particular, we have analyzed whether the reduction in aromaticity of the isoalloxazine ring system affects the overall conformational properties of the cofactor. We show that the relative population shift in conformational parameters of FAD in these two media is facilitated through a specific group of localized structural fluctuations in the coenzyme that give rise to folding−unfolding dynamics.

formations of FAD in bound and unbound states. These data provide vital knowledge on how FAD conformations are conserved within/across host enzymes, and by extension, the shape of the binding pocket imposed favored geometries and the inherent property of the FAD itself. A variety of studies have analyzed and compared conformations of coenzymes.21−25 Moodie and Thornton, for example, examined X-ray crystal structures of nucleotides and concluded that the conformations are extended in enzymebound structures and closed in unbound structures.23 In another study, the conformational diversity of ATP, NAD, and FAD coenzymes is analyzed across and within the protein families.26 The results show that most superfamilies bind ligands in a similar conformation. Enzyme-bound cofactors, particularly NAD and NADP, are studied by their dihedral angle distributions and also in relating conformations to enzyme function (EC) based on binding pockets/folds (utilizing either CATH/SCOP).26−28 Indeed, such studies have led to the creation of a CoFactor database, which provides online access to information extracted from the literature on 27 cofactors.29−31 In a recent comparative work, Stegemann and Klebe utilize self-organizing neural network analysis on ADP, ATP, NAD(P), FAD, and CoA molecule dihedral angles and binding cavities to study ligand conformation across host proteins.32 Previous analysis on FAD conformations to the bound proteins categorizes them as either extended or “butterfly”.1 The conformation of the photolyase-bound FAD differs from other redox-enzyme-bound FAD conformations studied from 32 PDB structures.33−37 The former shows FAD in an extended, I-shaped structure with the distance between the adenine and isoalloxazine ring systems in the range 13−16 Å. In the protein dodecin, FAD binds in a distinct closed structure as a result of stacking interactions between adenine and isoalloxazine moieties.38 Molecular dynamics (MD) simulations of FAD in the gas phase and in aqueous solution indicate that, irrespective of its oxidation state, the coenzyme is found to be flexible in both media; the distances between the adenine and isoalloxazine rings range from 4 to 16 Å with no preferred interring distances.11 More accurate MD simulations of FAD in explicit water for 8 ns have been carried out by van den Berg et al. using GROMOS 96 force field combining parameters from the flavin mononucleotide (FMN) and the ATP ligands.39 On the other hand, Radoszkowicz and co-workers utilize functional groups of FMN and NAD molecules to build the FAD force

■

MATERIALS AND METHODS PDB Structure Analysis. To analyze the structural diversity of the FAD, the Ligand Expo database,49 which contains crystal structures of enzyme-bound FAD deposited in the Protein Data Bank (PDB),50 was searched for redundant enzyme-bound FAD structures. For structures in multiple domains/chains, only one conformer was extracted. We did not differentiate the B

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

technique.56 Cutoff distances of van der Waals and electrostatic interactions were set to 10 Å. Coordinates of solutes were retrieved at every 2 ps during the entire production run. The solvent accessible surface area (SASA) was computed with a probe radius of 1.4 Å. The Gibbs free energy (ΔG) of the FAD unfolding−folding motions in water was estimated using the Boltzmann algorithm available in the g_sham program distributed with GROMACS. They were shown as energy landscape plots of two principal components (PC1 and PC2). Interaction energies of five dominant FAD configurations with the water molecules were computed by “freezing” the molecule in its distinct configuration as found in its corresponding group and by averaging energies over 100 ns dynamics runs. Unless otherwise indicated, the MD conformational properties were derived from FADH2 (system 2) because the reduced form of FAD is least studied previously.

oxidized (FAD) and reduced (FADH2) forms of FAD in crystal structures, as they differ only by hydrogen atoms. A resolution cutoff was not applied. However, structures with missing atoms were not included in the data set. A total of 927 redundant FAD entries were generated using this criterion. They were then treated with the GeoPCA method51 that utilizes dihedral angle based principal component analysis to project conformations in 2D space. Dihedral angles were computed using the Bio3d v2.0 package in R.52 MD Simulations in Water. Simulations were performed using the GROMACS 4.5.5 program53 to sample FAD2− and FADH2 conformations in water. The starting structure in the extended conformation was taken from the PDB entry 1bgn. The force field of FAD2− was obtained from the GlycoBioChem PRODRG2 server54 by uploading the heavy atom coordinates of 1bgn. For FADH2, explicit hydrogens were added to N1 and N5 atoms of the isoalloxazine. To assess the effects of ions and oxidation−reduction states of the molecule on the conformational dynamics, three simulation systems, namely, FAD2−, FADH2, and FADH2 without ions were set up (see the Introduction). The systems and their components were detailed in Table 1. Structures were solvated by placing

■

RESULTS FAD Conformations in Enzyme Structural Space. Principal component analysis of the enzyme-bound FAD’s 13 dihedral angles resulted in at least five distinct groups (Figure 2). The conformational differences between these groups are reflected by their separations in the projection. In addition, several small groups are noted, wherein the nucleotide conformations are quite diverse. Conformational groups correspond to a mixture of redox enzymes catalyzing various substrates. The enzymes are p-cresol methylhydroxylase, glutathione reductase, pyruvate oxidase, ferredoxin reductase, and photolyase. Because of this assortment of functions, the structure−functional correlation is least annotated. The five groups and their corresponding enzymatic reactions (characterized by their EC numbers) are listed in Table 2. The conformations of FAD bound to the enzymes of a similar functional class (similar EC numbers) are highly similar (small RMSD separation). There are also a few outliers and structures that do not form any groups (filled gray diamonds). To determine if these distinct FAD conformations interact differently in water, the interaction energies of these five dominant configurations with water from the crystal structures

Table 1. Simulation Systems and Their Components solute no. of waters ions

system 1

system 2

system 3

FAD2− 4573 2Na+

FADH2 4439 2Na+ 2Cl−

FADH2 4441 none

them in a pre-equilibrated cubic box of SPC water molecules that extended to 13 Å from the edge of the solutes. They were then minimized with steepest descent for 200 steps and equilibrated under NVT and NPT ensembles at 300 K and 1 atm for 500 ps in each phase with position restraints on the solutes. 200 ns unrestrained production dynamics were then performed on each system for 1+e8 steps with a time step of 2 fs. The LINCS algorithm was applied to constrain all bonds.55 Electrostatic interactions were updated using the Ewald

Figure 2. Enzyme-bound FAD conformations from dihedral angle data composed of the first two principal components (PC1 and PC2). XY scales are in arbitrary units. The variance of the PCs is indicated in parentheses. Each shape represents a FAD conformer. Different shapes are used to distinguish each group. Structures that do not form groups and outliers are represented as filled gray diamonds. C

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

of FAD are computed. The electrostatic and van der Waals interaction energies of the groups (in kJ/mol) are −915.5 and −135.8 [1], −930.5 and −138.2 [2], −913.5 and −142.0 [3], −908.5 and −136.3 [4], and −916.8 and −137.9 [5], respectively. The electrostatic contributions to the energies in [2] and [4] are distinct, but in other groups, their values are in the same range. All groups have similar van der Waals interaction energies, except [3]. To detect if there are differences in amino acid propensities of the five distinct groups of enzyme-bound FAD configurations, residue propensities around the adenine and isoalloxazine ring systems are computed at 4.0 Å from these functional groups. The propensities are presented as log-odds plots (Figures 3 and 4), which show notable relationships on ring recognition at the residue level. The adenine ring of all groups prefers to be surrounded by Val and Ile. Group 3 favors Gly and Metthe latter is disfavored by other groups. In addition to other residues, adenines of group 4, strangely, are surrounded by charged Arg and His, and group 5 leans toward Asn. The residue propensities surrounding isoalloxazines are, rather unsurprisingly, aromatic side chain residues, except Phe. With minor exceptions, both rings strongly disfavor charged residues: Asp, Glu, Lys, and polar uncharged Gln. Description of Overall Shape of Enzyme-Bound FAD Structures. To measure the overall shape of FAD bound to various enzymes, the distances between terminal “backbone” atoms (RN9A−N10, Figure 1) leading to the ring systems are calculated. The distance distribution (Figure 5) shows that FAD favors an overall extended shape populating values RN9A−N10 > 11.0 Å with a majority taking values around 15.0 and 16.0 Å. Nevertheless, some FAD structures exhibit RN9A−N10 < 8.0 Å favoring a compact structure. A comparison of FAD conformations across this distribution is shown by computing pairwise rmsd’s taken from representative structures at 16.0, 14.0, 12.0, 10.0, 6.0, and 4.0 Å, respectively (Table 3). The conformational diversity of FAD bound to various enzymes is evident from the rmsd values in Table 3. The rmsd values between the extended conformations (1zp4 and 1tdk) and closed FAD conformations (3k87 and 1iqu) are clearly high (>3.0 Å). The conformation of FAD bound to the oxidoreductase (3k87), and in dodecin (PDB entry 2ccb), is unusual in that it is highly folded. The adenine and isoalloxazine rings in this structure are separated by only 3.8 Å (RN9A−N10), whereas most other enzyme-bound FAD conformations are separated by at least 10.0 Å or greater. A likely reason for the atypical conformations of closed FADbound enzymes is because of their function. Another way to elucidate the shape of the coenzyme is to determine the orientation of the adenine and isoalloxazine rings relative to the pyrophosphate group. The virtual dihedral angle “δV” is defined to describe this orientation of adenine and nicotinamide rings relative to the pyrophosphate axis.57 Consequently, if the two rings are on the same side of the P−P axis, then δV takes value close to 0°, and if they are on opposite sides relative to the P−P axis, δV is 180°. Figure 6 illustrates the distribution of the δV in FAD defined by N9A− PA−P−N10 atoms (Figure 1). The enzyme-bound FADs surprisingly exhibit a strong preference for near trans orientation (∼180°), where the adenine and isoalloxazine rings are oppositely oriented relative to the pyrophosphate moiety. Conformational Analysis of FAD in Water. The trajectory of the RN9A−N10 distances for the 200 ns simulation

[4] [5]

[3]

[1]

[2]

EC numbers

1.1.1.158, 1.1.1.204, 1.1.3.7, 1.1.3.10, 1.1.3.13, 1.1.3.22, 1.1.99.5, 1.2.3.3, 1.2.3.31, 1.2.99.2, 1.3.1.2, 1.3.1.34, 1.3.3.1, 1.3.3.6, 1.3.5.1, 1.3.99.1, 1.3.99.20, 1.4.3.2, 1.4.3.3, 1.4.3.4, 1.5.3.1, 1.5.3.10, 1.5.3.11, 1.5.3.14, 1.5.3.15, 1.5.5.1, 1.5.99.8, 1.5.99.12, 1.6.2.4, 1.6.4.2, 1.6.4.5, 1.6.4.8, 1.6.6.1, 1.6.99.2, 1.7.3.1, 1.7.99.5, 1.8.1.4, 1.8.1.7, 1.8.1.9, 1.8.99.2, 1.10.99.2, 1.11.1.1, 1.14.13.2, 1.14.13.7, 1.16.1.1, 1.17.1.4, 1.17.3.2, 1.17.99.1, 1.18.1.2, 1.21.3.3, 2.1.1.74, 2.1.1.148, 2.2.1.6,, 2.3.1.12, 2.3.1.61, 2.5.1.26, 4.1.1.47, 4.1.3.18, 4.1.99.3, 5.4.99.9 1.1.1.158, 1.1.1.204, 1.1.3.10, 1.1.3.22, 1.1.3.38, 1.1.3.41, 1.1.3.6, 1.1.99.5, 1.13.12.9, 1.14.13.2, 1.18.1.2, 1.2.3.3, 1.2.99.2, 1.3.1.2, 1.3.3.1, 1.3.3.4, 1.3.5.1, 1.3.99.1, 1.3.99.2, 1.3.99.20, 1.3.99.3, 1.3.99.7, 1.4.1.13, 1.4.3.2, 1.4.3.3, 1.4.3.4, 1.5.1.20, 1.5.3.1, 1.5.3.11, 1.5.3.6, 1.5.99.8, 1.6.4.2, 1.6.4.8, 1.6.99.2, 1.7.3.1, 1.7.99.5, 1.8.1.4, 1.8.1.5, 1.8.1.7, 1.8.1.9, 1.8.99.2, 2.1.1.148, 2.2.1.6, 4.1.2.10, 5.4.99.9 1.1.1.158, 1.1.1.204, 1.1.3.10, 1.1.3.13, 1.1.3.21, 1.1.3.22, 1.1.3.4, 1.1.3.41, 1.1.3.6, 1.1.3.7, 1.11.1.1, 1.14.13.10, 1.14.13.2, 1.14.13.23, 1.14.13.3, 1.14.13.39, 1.16.1.1, 1.17.1.4, 1.17.3.2, 1.17.99.1, 1.18.1.2, 1.18.1.3, 1.2.2.2, 1.2.3.3, 1.2.99.2, 1.21.3.3, 1.3.1.2, 1.3.3.1, 1.3.3.6, 1.3.5.1, 1.3.99.1, 1.3.99.10, 1.3.99.17, 1.3.99.2, 1.3.99.3, 1.3.99.7, 1.4.3.11, 1.4.3.16, 1.4.3.2, 1.4.3.3, 1.4.3.4, 1.5.1.20, 1.5.3.1, 1.5.3.11, 1.5.3.6, 1.5.99.12, 1.5.99.8, 1.6.2.2, 1.6.2.4, 1.6.4.2, 1.6.4.5, 1.6.4.8, 1.6.5.2, 1.6.6.1, 1.6.99.2, 1.6.99.7, 1.7.3.1, 1.8.1.12, 1.8.1.4, 1.8.1.5, 1.8.1.7, 1.8.1.9, 2.1.1.148, 2.1.1.74, 2.2.1.6, 2.3.1.168, 2.7.13.3, 2.7.7.2, 4.1.3.18, 4.1.99.3 1.1.3.13, 1.1.3.38, 1.14.13.2, 1.14.13.39, 1.18.1.2, 1.2.3.3, 1.3.99.1, 1.3.99.17, 1.3.99.3, 1.4.3.4, 1.5.1.20, 1.5.99.12, 1.5.99.8, 1.6.4.5, 1.8.1.4, 1.8.1.9, 1.8.99.2, 2.2.1.6, 4.1.99.3 1.1.1.204, 1.1.3.10, 1.11.1.1, 1.17.99.1, 1.18.1.2, 1.2.99.2, 1.3.5.1, 1.3.99.1, 1.4.3.2, 1.4.3.3, 1.5.3.1, 1.5.3.11, 1.5.99.8, 1.6.4.2, 1.8.1.7, 4.1.99.3

group

Table 2. Five Major FAD Conformation Groups (in Figure 2) and the Corresponding EC Numbers of Their Bound Enzymes

The Journal of Physical Chemistry B

D

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 3. Log-odds plot of amino acid residue propensities around the adenine ring in the five groups of enzyme-bound flavin adenine dinucleotides. Positive and negative Y-values indicate preference and “dislike”, respectively.

Figure 4. Log-odds plot representing amino acid residue propensities surrounding the isoalloxazine ring in the five groups of enzyme-bound flavin adenine dinucleotides. Positive and negative Y-values indicate preference and “dislike”, respectively.

riboflavin portion (RFN) measured by RN10−P distance shows similar folding efforts from the initial RN10−P = 8.8 Å. The time scales also show that the conformational fluctuations in the RFN region are far greater than the AMP moiety. This is not unexpected because the former consists of a linear flexible hydrocarbon chain. In the final folded conformation, AMP and RFN portions settle at RN9A−PA ∼ 6.7 Å and RN10−P ∼ 5.6 Å, respectively, indicating that they have adopted compact conformations on their own (RN9A−PA > RN10−P) but in a relatively stretched configuration when compared to RN9A−N10 distances. To summarize, the predominant form of FADH2 in water is a folded conformation characterized by an average RN9A−N10 ∼ 4.5 Å. The final folded conformation contains adenine and isoalloxazine rings that are not perfectly stacked unlike in the folded crystal structure of FAD in the enzyme dodecin or oxidoreductase (3k87).

obtained from system 2 is shown in Figure 7. The starting open conformation continues only for the first 1000 ps of the simulation, followed by partial folding events around 6 ns, and becomes completely folded at 13 ns (RN9A−N10 ∼ 4.0 Å). The compact structure persists predominantly until 60 ns. This is followed by multiple unfolding attempts at time scales between 60 and 100 ns. Around 100−140 ns of the simulation, a series of folded FADH2 conformations are generated. Between 140 and 150 ns, FADH2 attempts to form extended structures, followed by another series of closed conformations between 155 and 190 ns. A short series of unfolding structures are generated before finally settling to a folded conformation at 200 ns. During the simulation run, it is noticed that trajectory of the adenosine monophosphate (AMP) region defined by the RN9A−PA distance starts at 7.0 Å but attempted to form compact conformations on its own (Figure 8). In a similar vein, the E

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 7. Trajectory of RN9A−N10 distance fluctuations in system 2 during the 200 ns simulation run.

Figure 5. Distribution of enzyme-bound FAD conformations as a function of RN9A−N10 distance (Å).

Table 3. Pairwise rmsd at Atomic Positions (in Å) for Enzyme-Bound FAD Conformers Representative of Different RN10−N9A Distances PDB (RN9A−N10, Å)

3k87

1iqu

1gvh

2g37

1zp4

1tdk

3k87 (3.9) 1iqu (6.2) 1gvh (10.0) 2g37 (12.1) 1zp4 (14.2) 1tdk (16.0)

0 4.3 3.6 5.5 6.3 7.1

4.3 0 4.3 4.3 3.4 3.4

3.6 4.3 0 3.2 3.9 4.5

5.5 4.3 3.2 0 1.6 2.2

6.3 3.4 3.9 1.6 0 2.3

7.1 3.4 4.5 2.2 2.3 0

Figure 8. Trajectories of RN9A−PA and RN10−P distances in system 2 during the 200 ns simulation run.

of systems 2 and 3 of FADH2 are displayed in Figure 9. The distance oscillations show that there are some changes in the extent of folding−unfolding patterns over time. However, overall the trajectories are similar and on average FADH2 without ions stays in a closed conformation ending at RN9A−N10 = 4.5 Å. To elucidate the conformational dynamics and differences between two oxidation states of FAD, FAD2−, and FADH2, the trajectories of RN9A−N10 distances between the two states (systems 1 and 2) are plotted (Figure 10). The plot indicates that, although the time histories of the distances show variations, in particular FADH2 (black) has more unfolding events than its oxidized form (red); by and large, the FAD mostly stays in a closed conformation in both states including

Figure 6. Distribution of the virtual dihedral angle, δV.

To investigate the effect of ions on the conformations and unfolding−folding dynamics, time histories of RN9A−N10 values F

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 9. Trajectory of RN9A−N10 distance fluctuations in systems 2 (black) and 3 (blue) during the 200 ns production dynamics.

Figure 11. Comparative distributions of RN9A−N10 distances in the enzyme-bound and water environments.

projected on the eigenvectors of its covariance matrix (see the Materials and Methods section). Figure 12 shows the

Figure 10. Time events of RN9A−N10 distances in FADH2 (black) and FAD2− (red). Only the last 40 ns of the production runs are shown.

the final structure. These results indicate that FADH2 prefers to unstack easily compared to its oxidized form, indicating a loss of aromaticity on the isoalloxazine ring in the former appears to have reduced the stacking effects between the two moieties. To gain a better description about the shape of the FAD in two environments, the distributions of RN9A−N10 distances of sampled conformations from systems 1 and 2 (bold line) and enzyme-bound states (dashed line) are plotted (Figure 11). Examination of the line plots shows the relative stability of the shape of the molecule in both environments, namely, folded in water and extended in the enzymes, and the degree to which the enzyme enforces a population shift from unbound FAD conformations. The free-energy landscape (FEL) can provide a quantitative description of FAD folding−unfolding motions. Thus, the probability of dinucleotide conformations occupying the folded and unfolded states is computed from the 200 ns trajectory and

Figure 12. Converged free energy landscape of FADH2 dynamics in water as a function of PC1 and PC2. The XY axes are in arbitrary scales. The global free energy minimum of the final folded conformation is shown in dark blue. Representative conformations from other “basins” are also shown.

converged free energy landscape for FADH2 dynamics in water. The deepest free-energy minimum of FADH2 (dark blue) corresponds to the compact conformation characterized by the intramolecular complex formation between adenine and isoalloxazine rings, as shown by the 200 ns time events. Notably, there are also other populated folded states (various shades of light blue), with higher free energy where the ring systems are semiclosed. The ensemble of conformations found in this basin correlates with the intermediate folded structures occurring during the 200 ns trajectory (Figure 7). Similar FEL G

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 13. Distributions of neighboring bonded dihedral angle pairs. Scatter plots demonstrating dihedral angle conservation and relative population shift of FAD in crystal structures (blue) from water (black). Only select dihedrals are shown.

plots for systems 1 and 3 are presented in Supporting Information S2. Analysis of Dihedral Angle Space in Free and Enzyme-Bound FAD Conformations. To ascertain the conserved geometry of FAD conformations, the distributions of neighboring bonded dihedral angle pairs (Figure 13) in both environments are computed. Comparison of the distributions of N-glycosidic and ribityl bonds leading to the ring systems are also shown. Select distributions are displayed as scatter maps. Other dihedral distributions are shown in Supporting Information S3. Angles along the sugar and anhydride linkage leading to the phosphates, namely, βA and αA, take values around 160−180°/ −180° and 60−90°, respectively, in proteins. The ζA and D8 dihedral angles of the diphosphate groups show an unusual symmetrical distribution in water (see also Supporting Information S4 for a clear picture); they are populated around 90° upon binding to enzymes. The adonitol bonds represented by dihedrals, D2−D7, show strong conservation for the trans orientation in bound enzymes (∼180°). The adenine base about the N-glycosidic bond (χA) shifts exclusively to the anti region in the enzyme-bound state. The ribityl bond (D1) connecting the isoalloxazine ring concentrated at 90° during dynamics spreads to −90° in proteins. The dihedral angle orienting the sugar moiety is mostly concentrated in the gauche region (−60° ∼ γA ∼ 60°) in both free and bound molecules. Finally, the virtual dihedral angle, δV, that is near trans in proteins orients close to 0° in water, indicating the two ring systems lie on the same side relative to the P−P axis. This relative distribution change is shown in Figure 14. Dihedral Angle Dynamics of FADH2 in Water. To establish whether dihedral angles change in a concomitant manner with respect to overall conformational fluctuations resulting in folding−unfolding dynamics, trajectories of several dihedral angles from system 2 are plotted (in Figure 15 and Supporting Information S5). Dihedral angle changes are observed for the χA, D3, and D4 values, which correlated with

Figure 14. δV angle distributions derived from FAD in water and enzymes. Notice the change in the orientation of the ring systems relative to the diphosphate axis.

the major changes in the distance trajectories shown in Figure 7. The major transitions between the initial extended form and the oscillations occurring between 4 and 5, 60 and 100, 140 and 160, and around 190 ns resulting in the unfolded FAD conformations are mostly correlated with transitions in χA and D2−D5 angles. No significant dihedral fluctuations are noticed at other time scales, indicating that relatively stable compact conformations are produced. The other dihedral angles are also found to be flexible during folding−unfolding arrangements. However, significant changes are unnoticeable. Among the dihedral transitions of ring systems (χA and D1), adenine base H

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

Figure 15. Time evolution of select dihedral angles obtained from FADH2 in water.

fluctuates significantly during dynamics, but the isoalloxazine transitions are not very substantial. This indicates that, during unfolding, the adenine ring adopts different orientations to relieve from the stacking interactions, while the isoalloxazine relatively remains in more or less identical orientations. Time events of the hypothetical dihedral angle, δV, suggest that, during folding−unfolding transitions, the values range from 0 to 180°, indicating closer and farther orientation of ring systems.

■

leads to a large population shift in conformational parameters favoring specific geometries in the protein structural space. Past works have attempted to classify the binding conformation of FAD in the protein context. Our intention is not to regurgitate these studies even though the data set we have created is more exhaustive than any previous efforts.23,26,32 One of the previous works on 91 redundant entries showed two groups, with one dominant group and a second smaller group.32 Our study on an extensive redundant data set, on the contrary, reveal that conformations of FAD bound to various enzymes segregate into five distinct groups. Although this result is different, FAD conformations predominantly favor an open state when bound to proteins, which is consistent with previous works.1,23,26 While extended conformations maximize residue interactions and stability, it is surprising that FAD adopts only a restricted set of distinct conformations (five groups) upon

CONCLUDING DISCUSSION

Our results demonstrate the conformational features of flavin adenine dinucleotide (FAD) bound in the protein space and in solution. Through GeoPCA, molecular dynamics, and dihedral angle analysis, we show that the FAD−enzyme binding event I

dx.doi.org/10.1021/jp507629n | J. Phys. Chem. B XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry B

Article

stabilization in the closed conformation characterized by small RN9A−N10 distances, it appears that RFN and AMP groups hold enough structural flexibility to adopt comparatively extended configurations and be relieved of intramolecular contacts with close functional groups in solution. No appreciable differences in final conformations are noticed between oxidized and reduced FAD and between FADH2 with and without counterions; in particular, the stacking interactions seem to have lessened in FADH2 due to the loss of aromaticity in the isoalloxazine ring. These two results are open to interpretation because the stacking or pyrophosphate-cation properties described by the available PRODRG2 force field can have limitations.65 However, replication of the simulation data, consistent with previous literature, suggests that the last compact stacking conformation of FAD in both oxidation states and with/without ions noticed in this study is also feasible. Nevertheless, they warrant additional theoretical and experimental scrutiny. Scatter plots of select dihedral angle distributions in the enzyme-bound state and in water reveal that the protein imposes the observed population change and they are not native features of the FAD. Quantification of FAD shapes according to prevailing structures (dihedral scatter plots, RN9A−N10, and δV) noticed in protein and in water shows how stable these specific structures are in respective media and the degree of conformational transition in changing the environment. Thus, the protein-binding domain/fold forces FAD to an extended conformational state that is least populated in solution.1,26,32,66,67 The exceptions to these are the crystal structures of FAD bound to the enzymes dodecin and chlorophenol monooxygenase (PDB entry 3k87), where FAD is characterized by a perfectly stacked folded conformation (see the Introduction and Results sections). The favored folded structures in water seem to be a delicate interplay between reduction in FAD surface area and hydration of the aromatic ring systems (see Supporting Information S6). Previous studies have shown that adding alcohol/urea to FAD in water renders unfolding of the molecule.40,61,62 Therefore, it appears that solvent plays a big role in extending the dinucleotide conformation. Indeed, our preliminary results on propensities of residues around the ring systems in enzymebound FAD structures point out that mostly hydrophobic interactions dominate (see the Results section). This was further confirmed in a very recent MD study employed to understand the nature of the interaction between FAD and denaturant urea.68 Results describe that urea exchanges as many water molecules around adenine and isoalloxazine rings of FAD, resulting in unfolding the dinucleotide.

binding to different classes of enzymes, despite possessing 13 dihedral angles about rotational bonds. Previous comparative studies on adenine and phosphate moieties of enzyme-bound FAD concluded that cofactor recognition is achieved not through the molecule as a whole but by partial recognition of its local structure.1,26,32,58,59 In particular, previous groups have reported a strong preference of the anti conformation of the χA dihedral angle about the Nglycosidic bond and relevance of diphosphate binding to the Rossmann fold or its variants.23,26,32 Our dihedral angle data from crystal structures, intuitively, are in accord with these observations. Our analysis also adds the following structural features characteristic of bound-FAD conformations to the existing literature: (i) the ζA and D8 dihedral angles about the pyrophosphates are conserved at syn and gauche regions, and (ii) the D1 angle about the isoalloxazine ring is populated mostly at the syn conformation. Through the virtual dihedral angle δV, we establish that the adenine and isoalloxazine rings adopt a near trans orientation relative to the pyrophosphate axis upon binding to the enzymes, which was not reported elsewhere (to the best of our knowledge). It is, however, relevant to note that the enzyme-bound FAD dihedral angles are extracted from varying resolution protein X-ray crystal data as opposed to only coenzyme crystal structures. Furthermore, crystallographic conditions in the refinement of most protein structures can significantly affect the conformation of the final structure. Consequently, the dihedral angle data include noise, such as unrealistic temperature factors. Nonetheless, a reasonable approximation of the dihedral angle distributions can be deduced because of our large data set of independently solved enzyme-bound FAD crystal structures available in the PDB. Several experiments and MD simulations show that a folded FAD conformation exists in solution.1,39,40,60−64 Fluorescence titration experiments and anisotropic measurements propose a representation in which the two aromatic rings stack at a distance of centers of mass