our simulation; free ion 1 (red), near Aspl 13 (blue), near Asp107 (black) and free ion 2 (green). The thin lines show the location of Na+ in the control experiment ...
Bioscience Reports, Vol. 18, No. 1, 1998
Computational Observation of an Ion Permeation Through a Channel Protein Atushi Suenaga,1 Yuto Komeiji,2 Masami Uebayasi,3 Toshiyuki Meguro,1 Minoru Saito,4 and Ichiro Yamato1,5 Received February 9, 1998; accepted February 17, 1998 The ion permeation process, driven by a membrane potential through an outer membrane protein, OmpF porin of Escherichia coli, was simulated by molecular dynamics. A Na+ ion, initially placed in the solvent region at the outer side of the porin channel, moved along the electric field passing through the porin channel in a 1.3 nsec simulation; the permeation rate was consistent with the experimentally estimated channel activity (108- 109/sec). In this simulation, it was indicated that the ion permeation through the porin channel proceeds by a "push-out" mechanism, and that Aspll3 is an important residue for the channel activity. KEY WORDS: Channel protein; molecular dynamics; ion permeation process; channel activity. ABBREVIATIONS: MD, molecular dynamics; RMS, root mean square.
INTRODUCTION
Gram-negative bacteria have two membraneous structures at their surfaces, cytoplasmic and outer membranes [1,2]. Outer membranes contain phospholipids, lipopolysaccharides, and several outer membrane proteins [2-4]. The major components of the outer membrane proteins in E. coli are OmpF, OmpC, and OmpA proteins. OmpF and OmpC proteins are porins, which have permeation pores for hydrophilic molecules with a molecular weight less than 600 Da. The crystal structure of OmpF porin was determined [5], as the first three dimensional structure of membrane proteins that function as transporters. OmpF porin is a homo-trimer of the OmpF polypeptide with a molecular weight of 36 kDa (Plate 1A). The porin 1
Department of Biological Science and Technology, Science University of Tokyo, 2641 Yamazaki, Noda-shi, Chiba 278-8510, Japan. 2 Supermolecular Science Division, Electrotechnical Laboratory, 1-1-4 Umezono, Tsukuba-shi, Ibaraki 305-0045, Japan. 3 Speciflc Metabolism Laboratory, National Institute of Bioscience and Human Technology, 1-1-3 Higashi, Tsukuba-shi, Ibaraki 305-0046, Japan. 4 Parallel Application Laboratory, Real World Computing Partnership, 1-6-1 Takezono, Tsukuba-shi, Ibaraki 305-0032, Japan. 5 To whom correspondence should be addressed.
39
40
Suenaga, Komeiji, Uebayasi, Meguro, Saito, and Yamato
monomer has a large 16-stranded antiparallel j3-barrel structure enclosing the transmembrane pore (Plate 1B). The monomer functions as an independent channel. The pore is constricted around half-way through the membrane by a long loop (L3; loop 3) inside the barrel. The constriction zone is surrounded by clusters of positively and negatively charged amino acid residues (Plate 1C). L3 has several glutamic and aspartic amino acids; so, L3 is postulated as an electrical gate of the channel [6-8]. Without an applied membrane potential, the channel of the porin is open and the current increases proportionally to the applied voltage according to Ohm's law [9]. The ion conductance is very high and the rate of ion movement was estimated to be 108~9 ions/sec [9]. The conductance gradually decreases above a certain voltage (threshold voltage for OmpF porin is about 150mV [4]), which suggests that the channel becomes closed [9]. Owing to these functional properties, the ion permeation process through porin is taken as a representative of the channel activity and is studied using biochemical and electrophysiological methods. OmpF porin is a good target to study the molecular mechanism of channel activity by molecular dynamics (MD) simulation with present-day computer power, since the permeation rate is very fast. Previous MD studies of porin [10-12] mainly discussed the dynamics of L3 in vacuo to understand the role of L3 in channel closure but did not report the dynamics of ions and water molecules. In this report, we first provide information about ion behavior bound or released from the amino acids at the constriction zone of the porin. We also describe the entire permeation process of the ion through the porin, which reveals several characteristic features of the channel activity difficult to obtain by experimental methods. MATERIALS AND METHODS Generation of Initial Structure
In our simulation, the X-ray structure of OmpF porin monomer [5] was placed in a box filled with TIP3P [13] water molecules (Plate 1B). To mimic the situation that the porin resides in a biological membrane, water molecules were removed from the part which corresponds to the position of the lipid bilayer. The center of porin was 37 A along the Z-axis. Water molecules within the range of 37 ± 12 A along the Z-axis were removed, since the length of the hydrophobic part in the lipid bilayer is considered about 25 A [14]. Thus, the space corresponding to the lipid bilayer part was left vacant and the other parts, including inside of the porin, were filled with water molecules. (The box size was 62 x 64 x 88 A.) To keep the lipid bilayer part vacant during simulation, the positions of the two layers of water molecules, next to the bilayer part, were fixed and other water molecules were left free. All the charged residues of the porin were neutralized by placing counter ions, except for those making up ion pairs (Glu62-Arg82 and Argl32-Glu347; Plate 1C). In summary, 41 Na+ and 25 C1- were placed near charged amino acids. An additional free sodium ion (free ion 1) was placed in the solvent region outside the pore (at 55 A along Z-axis) as a free ion to investigate the movement. Then at 0.8 nsec during the simulation, we placed another free Na+ (free ion 2) at a similar position to the initial free ion 1 near the inlet of the pore.
Plate 1. The crystal structure of OmpF porin. (A) Porin trimer viewed from the outside of the cell is shown using the coordinates obtained from X-ray crystallography [5]. Only the main chain of the porin trimer is shown. (B) The initial structure for the molecular dynamics simulation. Only main chain backbone (green) is shown, and water molecules (red) are indicated in the CPK model. Upper side corresponds to the outside of cell. Z-axis represents the direction perpendicular to the membrane. The electric field (inside negative) is in the opposite direction of the arrow of Z-axis. (C) Constriction zone of OmpF porin. Shown is the constriction zone ranging from 30 to 45 A along the Z-axis viewed from the direction perpendicular to the membrane. L3 represents the long loop (yellow) which constricts the pore approximately half-way along the barrel. Side chains of the charged residues at the constriction zone are shown; those of negatively charged residues are shown in blue and those of positively charged residues are drawn in red. (D) Root mean square fluctuations of amino acid residues during the 800-1300 psec simulation. (E) Time course of the locations of ions along the Z-axis. The thick lines show the location of Na+ in our simulation; free ion 1 (red), near Aspl 13 (blue), near Asp107 (black) and free ion 2 (green). The thin lines show the location of Na+ in the control experiment where no free Na+ was placed in the simulation; near Aspl13 (blue) and near Aspl07 (black).
42
Suenaga, Komeiji, Uebayasi, Meguro, Saito, and Yamato
Application of Membrane Potential Voltage of 500 mV was applied across the pore to mimic the membrane potential, where the additional force exerted on a charge (qi) during simulation was calculated as where Fi is the force due to the voltage acting on atom i, qt is the charge of atom i, and E is the electric field. We used an electric field of 500 mV/100 A, beyond the threshold value, to enhance the ion permeation rate (Plate 1B). Condition of Molecular Dynamics Simulation Amber 3.0 Revision A [15] was used throughout this study. An Amber86 force field [16] was employed for the simulation. All covalent bonds were constrained to their equilibrium values using the SHAKE algorithm [17] and a time step of 1 fsec was used. Electrostatic interaction was calculated up to the range of 20 A from each atom on residue base, whereas the calculation of van der Waals interaction was cut off by 9 A [18]. Watanabe et al. [12] showed that constrained MD was required to keep the porin structure close to the experimental structure during simulation. Therefore, the positions of amino acid residues at the lipid bilayer side of the porin were fixed and those at inner side of the porin channel were left free. Thus, the amino acid residues kept frozen show no fluctuation, whereas those allowed to move show certain fluctuation values above zero (Plate 1D). RESULTS AND DISCUSSION Channel Activity The simulation with the applied voltage covered 1300 psec at 300 K. Main chain root mean square (RMS) displacement of the simulated structure from the X-ray structure was very low (-1.5 A; not shown). RMS fluctuations of L3 (residue 101-134), which was important for ion permeation [6-8,19], showed that L3 was relatively flexible (Plate 1D); this has been demonstrated by other MD simulations [10-12] and we thought that our simulation condition was satisfactory to reveal the essential features of dynamics during the ion permeation process. Counter ions placed near charged residues of the porin remained close to their initial locations, except two sodium ions placed near Asp107 and Aspll3. Those two ions migrated along the electric field generated by the applied voltage. Plate 1E shows the location of ions along the Z-axis during the simulation. The counter ions placed near Asp107 and Aspl13 (in Loop 3) moved off along the electric field to the outside of the channel (Plate 1E). The first free Na+ (free ion 1) migrated electrophoretically starting at the position of the pore inlet into the constriction zone. It did not interact with charged amino acids until it encountered Aspl13. It pushed out the counter Na+ ion initially placed near Aspl13 (Plate 1E; 760 psec). The second free Na+ (free ion 2) then migrated to Aspl13 in a similar manner pushing out the free ion 1 which was bound to Aspl13 (Plate 1E; 1160 psec).
MD of Porin
43
As a whole, during this 1.16 nsec simulation one Na+ ion passed across the porin channel. The ion conductance through the single channel has been determined as 0.82 nS at 1 M NaCl with an applied voltage of 120mV [20]. Thus, the calculated rate (0.614 ions/1 nsec at 0.2 M NaCl with a membrane potential of 500 mV) is in good accord to the ion permeation rate (1 ion/1.16 nsec) estimated from our simulation. (In our simulation, the ion concentration was 2 Na+ in the upper solvent part above the constriction zone which corresponded to 0.2 M Na+.) Our simulation is the first demonstration of the voltage dependent channel activity of the porin. Without an applied voltage, ions, including counter ions placed near charged amino acids, stayed close to the initial positions [21]. Furthermore, without added free Na+, the counter Na+ ion placed near Aspl13 did not come off the amino acid even with an applied voltage, but rather migrated uphill along the applied electric field to equilibrium position; in contrast, the Na+ placed near Asp 107 migrated off the amino acid in all cases, suggesting that Asp107 was not stably neutralized under the applied voltage. Other counter ions (Na+ or Cl~) slightly migrated, but their ion pairings with neighboring charged amino acids were maintained. If we put a free Cl~ in our system, it will move in the opposite direction of Na+ according to the electric field. Although the analysis of such movement of anions may give additional information, we followed only the permeation process of a free Na+ in this simulation, since our simulation should be enough to provide essential features of the channel activity and the inclusion of free anions in our system will make the analysis more complicated.
Ion Permeation Process
Figure 1 shows schematic snapshots at the constriction zone during the simulation. Counter ions located near AsplO? and Aspll3 began to separate from Aspl07 and Aspll3 at 200psec (Plate IE; Fig. 1, 250psec). The free ion 1 (Na 1) formed ion pairing with Aspll3 at 750psec (Fig. 1, 750psec). Free ion 2 (Na2) moved into the pore slowly in a similar manner to free ion 1 and pushed out free ion 1, which was bound to Aspll3 at 1160psec (Fig. 1, 1160psec). Then, free ion 2 formed the ion pairing with Aspl 13. This process is an ion-exchange reaction. These results indicate that the ion permeation of porin proceeds by a "push-out" mechanism. Finally, free ion 1 permeated completely from the outer to the periplasmic side of the porin via AspllS (Fig. 1, 1300psec). As a whole, our simulation showed that Aspl 13 plays an important role for the channel activity of OmpF porin, especially for ion selectivity of the porin, and that other charged amino acids may not be so important for the channel activity. In a mutagenesis experiment, the mutant porin (Aspl13—>Gly) showed different cation selectivity [19] from the wild-type porin, although the ion conductance did not change much. We think that the result of our simulation is consistent with the observed characteristics of the channel activity. Further studies using free energy calculations should be useful for understanding the ion selectivity of the porin.
44
Suenaga, Komeiji, Uebayasi, Megnro, Saito, and Yamato
Fig. 1. Model of the ion permeation process in porin. Na+ indicates the counter ion placed near Aspl 13. Na 1 indicates free ion 1, and Na 2 indicates free ion 2. W indicates the water molecule which mediates the ion pairing between the ion and the amino acid.
Ion Releasing/Binding Process
Figure 2 shows the coordination numbers of water molecules around the free ion 1 and the Na+'s placed near Asp 107 and Aspll3 as a function of time. During the time period of 0~170psec, the coordination number around the Na+ near Asp 107 and Aspl13 was about 4. The experimental observations and other simulations indicated that the coordination number around a free Na+ was 5-6 [21,22]. Therefore, the low coordination numbers around the Na+ near Asp 107 and Aspll3 were due to the binding to respective Asp's. The coordination number of the Na+ at Aspll3 increased to 5-6 during 170~600psec. During the time period of 170-500 psec, the distance between Aspl13 and the Na+ was a little longer (4-5; Fig. 1,250 psec; first intermediate of ion releasing process) compared with the initial distance (3-4 A). After 500 psec, the distance (7-8 A) between the ion and Aspll3 was constant for a while until 600 psec (second intermediate state; Fig. 1, 550 psec). The releasing of the counter Na+ from Asp 107 proceeded similarly but with a much shorter period (Plate IE, 200-400psec). In the first intermediate state of these ion releasing processes, the Na+ formed ion pairing with the amino acid mediated with one water molecule. In the second intermediate state, another water molecule was further inserted and the Na+ formed ion pairing with amino acid residues mediated with two water molecules. The free ion 1 formed ion pairing with Aspll3 at 750psec (Plate IE; Fig. 1, 750psec). After 700psec, the distance between the free ion 1 and Aspll3 became closer, then the free ion 1 formed ion pairing with Aspl 13 mediated with one water molecule (750 psec). This ion pairing was the transient intermediate until 760 psec.
45
MD of Porin
Then the free ion rapidly formed direct ion pairing with Aspll3-O52 (Fig. 1, 760psec). This ion pairing was stable until free ion 2 came close to Aspll3. The coordination number around free ion 1 was consistent with these states of binding (Fig. 2). When free ion 2 came close to Aspll3, free ion 1 moved from the position near Aspll3-O52 to Aspll3-O 5 l and free ion 2 formed ion pairing with Asp113O$2 mediated with one water molecule (Fig. 1, 1155psec). This intermediate state was transient (1155-1160 psec) and free ion 2 rapidly formed direct ion pairing with Aspl \3-Og2. These results indicate that the binding process of the ion to the amino acid involves a transient intermediate state, forming ion pairing mediated with one water molecule. After a while (50 psec), free ion 1 began to separate from Asp 113. This releasing process of free ion 1 (1210-1300 psec) was similar to that of the counter ions placed near AsplOV and Aspl 13. In addition, the coordination number around the free Na+ showed that only two of the 5-6 water molecules hydrated to the Na+ should be removed to cross the constriction zone. The porin channel activity is very high, indicating that the free energy stabilization at the bound ion state to a specific site is not so high, consistent with this change of the coordination number. Other channels (voltage dependent Na+-channel, acetylcholin receptor etc.) have lower channel activities than OmpF porin. In such channels, the binding energy of ions to a specific site may be very high and the number of water molecules that should be removed may be more than 2. Hypothesis on the Open/Close Mechanism of the Porin Channel
Based on our simulation study, we discuss several possibilities of the open/close mechanism of the channel. Classically, we think that the opening and closing of a
Fig. 2. Plots of the coordination number around counter ions (near Asp 107 and Aspl 13) and a free ion versus time during the simulation. The continuous thick line shows the coordination number around the counter ion near AsplOV; the broken line shows the coordination number around the counter ion near Aspl 13; the continuous thin line shows the coordination number around the free ion 1.
46
Suenaga, Komeiji, Uebayasi, Meguro, Saito, and Yamato
channel takes place by a conformational change of the protein as a primary choice of the mechanism. Here we hypothesize another possible mechanism. Around the constriction zone, there are ten charged amino acids. At first, we tried an MD simulation without any counter ions near these ten charged amino acids. The free Na+ migrated into the channel part, bound to Aspl21, and it stayed there for 500 psec simulation. This is consistent with the result of the present report that counter ions other than those near Asp 107 and Aspll3 stayed close and stable to the initial positions during 1.3 nsec MD. And the estimate of pKa of this Aspl21 was 9.63 [23], suggesting that this Aspl21 should be neutral. Next, we tried an MD simulation with counter ions near all these ten charged amino acids. Counter ions placed near Arg82 and Argl32 (Cl~) rearranged even uphill along the electric field during the initial 200 psec to their equilibrium positions. This rearrangement caused the filling up of the pore, blocking the ion permeation pathway. We continued the simulation for a further 300 psec, but the free Na+ seemed interrupted to enter the constriction zone by these rearranged counter ions. Therefore, we thought that the simulation of the ion permeation process was sensitive to the initial configurational structure of counter ions. In this reported simulation, these counter ions (Cl~ at Arg82 and Argl32, Na+ at Glu62 and Glu347) were omitted, assuming that these charged amino acids formed ion pairings with each other (Glu62-Arg82 and Argl32-Glu347). These ion pairings were stable during 1.3 nsec. From these observations, we speculate that certain counter ion configurations correspond to the open state of the channel and the other configurations correspond to the closed state (that is, the pore was just filled with counter ions and blocked). We further speculate that the channel open/close transition resulted from the process that the membrane potential (above the threshold value, 150mV) creates stable binding sites for such counter ions that fill up and block the pore. In our simulation, we applied 500 mV as the membrane potential far beyond the threshold value for OmpF porin. Nevertheless, we think that the channel state we used was that of the open configuration, and thus, that we could follow the ion permeation process. This possibility of an open/close mechanism is difficult to verify experimentally. But it is an-attractive alternative to the conformational change model. This possibility can be confirmed by performing long time MD simulations starting from various initial configurations of counter ion arrangements (the number of the initial configurations to be tested with and without counter ions for the ten charged amino acids counts 210~1000 initial structures). Furthermore, the possibility can also be tested by examining the effect of site-directed mutations of these amino acids (Glu62, Glu347, Arg82, and Argl32) on the open/close behavior of the channel activity. The mutations of Arg82 or Argl32 brought about different threshold values of the membrane potential [19], which is not inconsistent with our proposal. Previously, we carried out short time MD simulations with and without counter ions near these ten charged amino acids without a membrane potential [21]. Without counter ions, ion pairings of Glu62-Arg82 and Argl32-Glu347 were stable and there was an open pore at the constriction zone. However, with counter ions, ion pairings of Glu62Arg82 and Argl32-Glu347 were broken and the pore seemed closed, filled with counter ions, as described above. Apparently, this observation is inconsistent with the experimental result that the pore is open without an applied voltage. At present,
MD of Porin
47
we think that we need a much longer simulation to follow the opening from the closed state of the pore without an applied voltage. CONCLUSION From the results and discussion of our simulation, we conclude: (1) L3 loop, especially Asp 113, is important for the channel activity (especially for ion selectivity), which agrees with the experimental result [19]; (2) other charged amino acids at the constriction zone such as Asp 107, Glull7, Lysl6, Arg42, and LysSO play no or a small role in the channel properties (such as ion selectivity, permeation rate, voltage dependence, and so on); (3) the ion pairing amino acids at the constriction zone, Glu62, Arg82, Argl32, and Glu347 may be essential to determine the characteristics of the channel activity (especially of the voltage dependent open/close). As discussed in the text, this study can be expanded in three directions: (1) free energy calculations using appropriate structures in the trajectory to investigate the ion selectivity (specificity) determinants; (2) quantum mechanical MD to investigate the precise mechanism of the ion binding/releasing process at the constriction zone; (3) MD starting from various configurations of ion placement (counter ions) with long time simulations to confirm the open/close mechanism as the ion stacking mechanism rather than the conformational change. Finally, it will be a very useful enterprise to perform a long time MD of porin trimer in lipids (phospholipids and lipopolysaccharides) surrounded with water with an accurate calculation of the electrostatic interaction to follow steady state permeation of Na+ and Cl~. ACKNOWLEDGMENTS We are grateful to Dr. J. P. Rosenbusch for his encouragement. We thank RIPS at the Agency of Industrial Science and Technology for providing time on the Cray C916 supercomputer. This work was partly supported by the RING Program of the Agency of the Industrial Science and Technology (to Y.K. and M.U.). REFERENCES 1. Cronan, J. E., Gennis, R. B., and Maloy, S. R. (1987) in: Escherichia coli and Salmonella typhimurium (F. C. Neidhardt et al., eds.), American Society for Microbiology, Washington, DC, pp. 24-30. 2. Nikaido, H. and Vaara, M. (1987) in: Escherichia coli and Salmonea typhimurium (F. C. Neidhardt et al., eds.), American Society for Microbiology, Washington, DC, pp. 7-22. 3. Hancock, R. E. W. (1987) J. Bacterial. 169:929-933. 4. Rosenbusch, J. P. (1990) Experientia 46:167-173. 5. Cowan, S. W. et al. (1992) Nature 358:727-733. 6. Benson, S. A., Occi, J. L. L., and Sampson, B. A. (1988) /. Mot. Biol. 203:961-970. 7. Misra, R. and Benson, S. A. (1988) J. Bacterial. 170:3611-3617. 8. Lou, K.-L. et al. (1996) J. Biol. Chem. 271:20669-20675. 9. Schindler, H. and Rosenbusch, J. P. (1981) Proc. Natl. Acad. Sci. USA 78:2302-2306. 10. Bjorksten, J., Soares, C. M., Nilsson, O., and Tapia, O. (1994) Prat. Engng 74:487-493. 11. Soares, C. M., Bjorksten, J., and Tapia, O. (1995) Prat. Engng. 8:5-12. 12. Watanabe, M., Rosenbusch, J. P., Schirmer, T., and Karplus, M. (1997) Biophys. J. 72:2094-2102. 13. Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W., and Klein, M. L. (1983) / Chem. Phys. 79:926-935.
48
Suenaga, Komeiji, Uebayasi, Meguro, Saito, and Yamato
14. Gennis, R. B. (1989) in: Biomembranes (C. R. Cantor, ed.), Springer-Verlag, New York, pp. 36-84. 15. Seibel, G., Singh, U. C., Weiner, P. K., Caldwell, J., and Kollman, P. A. (1989) AMBER 3.0 Revision A, University of California, San Francisco, CA. 16. Weiner, S. J., Kollman, P. A., Nguyen, D., and Case, D. A. (1986) /. Comput. Chem. 23: 230-252. 17. Ryckaert, J., Ciccotti, G., and Berendsen, H. J. C. (1977) J. Comput. Phys. 23:327-341. 18. Komeiji, Y., Uebayasi, M., Someya, J., and Yamato, I. (1991) Prot. Engng 4:871-875. 19. Saint, N., Lou, K.-L., Widmer, C., Luckey, M., Schirmer, T., and Rosenbusch, J. P. (1996) /. Biol. Chem. 271:20676-20680. 20. Buehler, L. K., Kusumoto, S., Zhang, H., and Rosenbusch, J. P. (1991) J. Biol. Chem. 266:2444624450. 21. Suenaga, A., Yamato, I., and Saito, M. (1997) in: Progress in Cell Research (M. Sokabe, A. Auerbach, and F. Sigworth, eds.), Towards molecular biophysics of ion channels, Elsevier Sci. Pub., Amsterdam. Vol. 6, pp. 217-232. 22. Moore, W. J. (1962) in: Physical chemistry, 3rd ed., Prentice-Hall, NJ, USA. 23. Karshikoff, A., Spassov, V., Cowan, S. W., Ladenstein, R., and Schirmer, T. (1994) J. Mol. Biol. 240:372-384.