Molecular Orbital Study of the Interaction between MgATP and the Myosin Motor Domain Using the PM6 Method

Hiroshi KAGAWA, Hiroto KIKUCHI, Qi GAO and Toshikiho OGIHARA


1 Introduction

In several previous studies, our research group applied quantum-chemical calculations to large-scale biomolecules to understand biological reactions at a fundamental level. In this paper, we provide background information about these studies and present new results.
Kagawa et al. studied the structure of adenosine triphosphate (ATP) by using the PM3 method in the MOPAC computer program [1 - 3]. Furthermore, Kagawa and Mori studied the initial process of ATP hydrolysis in the myosin motor domain using the PM3 algorithm in MOPAC97 [4, 5]. Analyzing the electronic states of the model structure by means of a molecular orbital correlation diagram, they considered that (A) H2O(1181) corresponds to lytic water and (B) the highest occupied molecular orbitals (HOMOs) indicate that nonbridging O atoms of g-phosphate constitute the reaction site of ATP hydrolysis in the model structure. Based on these considerations, they proposed hypothetical mechanisms for the initial step of the hydrolysis reaction. The initial step of hydrolysis involves dissociation of lytic water to H and OH ions by electron flow into the antibonding unoccupied molecular orbital (MO) of O-H.
The entire myosin motor domain is too large to be used for MO calculations, even if semi-empirical MO methods are used; therefore, Kagawa et al. extracted only a small part of the domain for MO calculations. They also studied the effect of the amino acid charge and reported the relation between the inorganic phosphate charge and the phosphate structure [6]. Subsequently, Kagawa reviewed quantum studies on myosin-based ATP hydrolysis [7].
In this study, we recalculate the previous model using the PM6 method of MOPAC2009 [8] instead of PM3, and we compare the results of previous studies with those of the present study. The models under investigation are constructed using 1MMD data [9] obtained from the Protein Data Bank (PDB). Then, to discuss the mechanisms governing subsequent steps in the hydrolysis reaction, we investigate a new model developed using 1VOM data [10]. The 1MMD structure is thought to be identical to the structure formed in the initial step of ATP hydrolysis, while the 1VOM structure is thought to be identical to the structure formed at a slightly later step after the initial hydrolysis step.
The electronic state and three-dimensional (3D) structure of the new model complex are calculated using PM6, along with optimization of the g-phosphate and the atoms surrounding the g-phosphate. The results indicate that the proposed structure may be responsible for initiating the dissociation of g-phosphate from ATP.

2 Materials and Methods

The structures used in this study are constructed from the following X-ray coordinate files in the PDB: 1MMD [9] and 1VOM [10]. 1MMD, which was used in previous studies by Kagawa and Mori [4, 5], is the 3D structure of the myosin motor domain of Dictyostelium discoideum myosin II with an ATP analog, ADPBeF3. 1VOM, which is used in combination with 1MMD in the present study, has a structure almost identical to that of 1MMD but uses another analog, ADPVO4. ADPBeF3 is obtained from ATP by replacing g-phosphate with BeF3. VO4 in ADPVO4 is considered to originate from the reaction between VO3 and lytic water. These structures consist of ca. 6000 atoms, including the O atoms of water. Even semi-empirical methods cannot perform MO calculations for the entire system within a realistic time; therefore, we extracted only small parts of the system.
Because of the limited number of atoms available in WinMOPAC in which the MOPAC97 program is implemented, Kagawa and Mori [4, 5] selected only MgADPBeFx (x = 3 in the PDB 1MMD data), nine amino acid residues near the phosphates of ATP, and seven O atoms of water molecules near the ATP analog [H2O(1001), H2O(1002), H2O(1033), H2O(1175), H2O(1181), H2O(1343), and H2O(1344)]. Selected amino acid sequences are a sequence of a part of the P-loop (phosphate binding loop) [11] composed of six amino acid residues, Ser181-Gly182-Ala183-Gly184-Lys185-Thr186, and a sequence of three amino acid residues, Ser236-Ser237-Arg238, located near the triphosphate portion of ADPBeFx.
Although in the present there is no such small limit on the number of atoms in WinMOPAC and MOPAC2009, we use the same structure from 1MMD and a similar structure from 1VOM for comparison. The 1VOM structure consists of the same nine amino acids, along with one O atom of VO4 (998:O4) and six O atoms of water molecules [H2O(3), H2O(53), H2O(102), H2O(695), H2O(696), and H2O(697)] [10]. To restructure MgADP-BeF3 for modeling MgATP, Be was replaced with P and F3 was replaced with O3. Similarly, to restructure MgADPVO4 for modeling MgATPH2O, V was replaced with P. Furthermore, in both cases, H atoms missing in the X-ray structures were added to the model structure, OH was added to each C terminal of the amino acid chains, and two H atoms were added to each N terminal. The total charge of MgATP was set to -2, which is a reasonable value for a physiological environment [12], and that of the remaining portion of the structure was set to 0 or +2. Only H and O atoms added to the N terminals were optimized using the PM3 method in MOPAC97 or the PM6 method in MOPAC2009. The PM3 method can be used for calculations of structures that contain Mg, whereas the PM6 method can be used for calculations of structures that contain either Mg or V. Kagawa and Mori analyzed the interactions between the two parts-MgATP-2w and the cage structure-of a complex called the model complex [4, 5]. The "2w" of MgATP-2w indicates two water molecules which coordinate tightly with the Mg ion. The cage is composed of nine amino acid residues and the other five water molecules. In this work, we considered only the model complexes as a whole and did not consider the components of the complexes.
The MO calculations of the PM3 method were performed using MOPAC97 of WinMOPAC ver. 2, developed by Fujitsu Co. ( and also using MOPAC2009 developed by Stewart Computational Chemistry (; those of the PM6 method were performed using MOPAC2009. Graphic images of MOs were also visualized by means of a molecular modeling program, Winmostar ver. 3.80k developed by Tencube Labs. ( with an open-source Java viewer for chemical structures in 3D, Jmol ( Input data for MOPAC97 and MOPAC2009 were obtained using two molecular modeling programs: (1) FREE WHEEL ver. 0.57Q (not supported anymore), developed by Butch Software Studio, and (2) Winmostar ver. 3.80k.

3 Results and Discussion

As mentioned above, we considered only the model complexes as a whole in this study. MO calculations were performed under the following conditions:
  1. The PDB data used is either 1MMD or 1VOM.
  2. The model complex with original atoms, namely, BeF3 or VO4, or the model complex with replaced atoms, namely, PO3 or PO4.
  3. The total amino acid charge is set to either 0 by itself (not in solution) or to its usual value of +2 in a neutral aqueous solution; in either case, the total charge of MgATP is set to its usual value of -2 in a neutral aqueous solution.
  4. The MO calculation method is PM3 or PM6.
The results are summarized in Figure 1 and shown in Table 1.

Table 1. Location of HOMO in various cases.
MethodTotal charge of amino acids1MMD1VOM
PM30R, W (a)RR, W (c)-
PM60R (b)RR (d)R
+2AAA (e)A
R: Reaction sites except lytic water, namely, Lys185, nonbridging O of g-phosphate, and nonbridging O of b-phosphate (HOMO distribution on these atoms is different in different cases); W: lytic water; A: adenine of ATP; -: no data because PM3 has no parameter corresponding to V. Pictures of HOMO lobes with indexes (a)-(e) are shown in Figure 1.

Figure 1. HOMO (MO 287) of the MgATP-S1Dc model complex constructed using 1MMD and 1VOM. (a): 1MMD, PM3; (b): 1MMD, PM6; (c): 1VOM, PM3; (d): 1VOM, PM6, when the total amino acid charge is 0. (e): 1VOM, PM6, when the total amino acid charge is +2. The colors of the lobes (red and blue meshes) denote the opposite phase signs for MO. The color of each ball denotes the atomic species: gray, carbon; blue, nitrogen; red, oxygen; orange, phosphorus; green, magnesium; white, hydrogen. The sticks denote chemical bonds. The numerical letter 4 denotes the fourth oxygen atom of VO4 (PO4) given in the PDB datum. See (a)-(d) in Table 1. The lobes in other cases of "A" in Table 1 are similar to (e) in appearance.

First, we consider the cases wherein the total amino acid charge is 0. In both the MgATP-S1Dc model complex constructed from 1MMD and PM3 (studied by Kagawa and Mori [4, 5]) and that constructed from 1VOM and PM3, HOMO appears on the reaction site where lytic water, nonbridging O atoms, and Lys185 are present; however, the lobes on the lytic water molecule and O atoms of Pg are small in the case of 1VOM and PM3. When PM6 is used instead of PM3, there is no lobe on the lytic water molecule and the lobes of O atoms of Pg become smaller. The difference between the calculation results of PM3 and PM6 could be due to the slight difference between the parameters of PM3 and PM6. The same result (i.e., absence of lobe on lytic water molecule) is obtained for the MgADPBeF3-S1Dc and MgADPVO4-S1Dc model complexes.
Next, we consider the cases wherein the total amino acid charge is +2. In all cases, HOMO appears on adenine of ATP.
Further studies may encounter two significant problems. First, HOMO appears at entirely different locations depending on the amino acid charge. Second, HOMO appears at the lytic water molecule when PM3 is used (although lobes at the lytic water are too small to see in Figure 1 (c)) but not when PM6 is used (Table 1). We anticipate that these problems will be solved in the near future.
In a previous study [5], the mixing of states resulted in charge transfer to the unoccupied MO of H2O(1181) in 1MMD. This charge transfer is predicted to weaken the O40-H121 bond of the water molecule. We consider that this weakening of the O-H bond constitutes the initial stage of hydrolysis.
In this context, we attempted to elucidate the subsequent steps that occur after the O-H bond is weakened. To determine the step that occurs after the binding of H and OH in the water molecule to g-phosphate, we used 1VOM for the case wherein the H and OH ions of lytic water are bound to g-phosphate. Positional optimization of the g-phosphate and its surrounding atoms revealed an increase in the distance between Pg and the bridging O bound to Pb; this increase is considered to initiate the dissociation of g-phosphate from ATP. The distance between Pg and bridging O atom increased from 1.7939 A to 2.0616 A (Figure 2). As shown in Figure 2 (b), the bond between Pg and bridging O disappears. As a next step, we need to study the dissociation mechanism in greater detail.

Figure 2. Molecular structure of the MgATP-S1Dc model complex constructed from 1VOM. The color of each ball denotes the atomic species as the same as in Figure 1. The gray sticks denote chemical bonds and coordinations, whereas coordinations are not displayed in Figure 1. The numerical letters 1 - 4 denote the number of four oxygen atoms of VO4 (PO4) given in the PDB datum, and the alphabet letter O denotes the bridging O atom bound to Pb. (a): Molecular structure determined by positional optimization of hydrogen atoms. (b): Result of the further positional optimization of g-phosphate and its surrounding atoms. The distance between Pg and the bridging O atom bound to Pb increased from 1.7939 A to 2.0616 A, and the bond between them disappeared in a picture prepared using Winmostar.

ATP is supported by H bonds from surrounding amino acids; therefore, in computational studies, it is difficult to alter the structure of ATP. Furthermore, ATP alone cannot undergo hydrolysis. We suppose it is necessary for myosin to participate in the hydrolysis reaction as an enzyme. It is probably difficult to reproduce the hydrolysis reaction using only nine amino acid residues. Efforts are underway to carry out partial optimization of both the present complex as well as larger complexes with more amino acid residues. In the future, it may be necessary to use QM/MM-MD methods.

Mori and Kagawa first met as part-time teachers conducting computer-training classes at Azabu University. Kagawa later studied under Mori as a computational researcher, and both researchers maintained close personal contact. Kagawa is therefore deeply indebted to Mori and wishes to continue this study in his memory. The authors pray for the repose of his soul.


[ 1] J. J. P. Stewart, J. Comput. Chem., 10, 209-220 (1989).
[ 2] J. J. P. Stewart, J. Comput. Chem., 10, 221-264 (1989).
[ 3] H. Kagawa, Y. Nagai, K. Mori, Bull. Lib. Arts & Sci. Nippon Med. Sch., 22, 79-93 (1997), (in Japanese).
[ 4] H. Kagawa, K. Mori, Bull. Lib. Arts & Sci. Nippon Med. Sch., 26, 1-6 (1999).
[ 5] H. Kagawa, K. Mori, J. Phys. Chem. B, 103, 7346-7352 (1999).
[ 6] H. Kagawa, Y. Tamura, S. Kawauchi, K. Mori, K. Suzuki, Bull. Lib. Arts & Sci. Nippon Med. Sch., 31, 1-10 (2001).
[ 7] Hiroshi Kagawa, J. Nippon Med. Sch., 74, 4-10 (2007).
[ 8]
[ 9] A. J. Fisher, C. A. Smith, J. B. Thoden, R. Smith, K. Sutoh, H. M. Holden, I. Rayment, Biochemistry, 34, 8960-8972 (1995).
[10] C. A. Smith, I. Rayment, Biochemistry, 35, 5404-5417 (1996).
[11] C. A. Smith, I. Rayment, Biophys. J., 70, 1590-1602 (1996).
[12] W. Saenger, Principles of Nucleic Acid Structure, Springer-Verlag, New York (1987), Chapter 5.