ARTICLE pubs.acs.org/JPCB
Permeation of Nucleosides through Lipid Bilayers Chenyu Wei*,†,§ and Andrew Pohorille*,‡,§ †
NASA Ames Research Center, Mail Stop 229-1, Moffett Field, California 94035, United States NASA Ames Research Center, Mail Stop 239-4, Moffett Field, California 94035, United States § Department of Pharmaceutical Chemistry, University of California San Francisco, San Francisco, California, United States ‡
ABSTRACT: Elucidating mechanisms that facilitate primordial synthesis of information polymers is central to understanding the origins of life. One such mechanism might have been the recently discovered diastereoselectivity of membranes favoring uptake of ribose (Sacerdote, M. G.; Szostak, J. W. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 6004), which might have promoted its preferential incorporation into nucleic acids. To determine whether the same mechanism was available if nucleosides rather than sugars were supplied to ancestral cells, we carry out molecular dynamics simulations of their permeation through a lipid bilayer. We find that the free energy barriers to permeation of ribo-adenosine and arabino-adenosine through the 1-palmitoyl2-oleoyl-sn-glycero-3-phosphatidylcholine membrane are quite similar, equal to 10.0 and 10.4 kcal/mol, respectively. The corresponding permeability coefficients are also similar, equal to 9.1 10-7 and 5.3 10-7 cm/s. The 10-fold increase in permeability of membranes to ribose over its diastereomers is not preserved for nucleosides because in contrast to free aldopentoses they exist in the furanose rather than pyranose form. This change eliminates the possibility of forming a network of favorable, intramolecular interactions between exocyclic, hydroxyl groups that stabilizes ribose, but not its diastereomers, inside membranes. Thus, uptake of nutrients provided selective advantage to primordial RNA only if the species absorbed through cell walls were sugars rather than nucleosides.
’ INTRODUCTION Genetic information for all organisms on earth is encoded in DNA or RNA, which are chain polymers capable of forming double helices. There is a consensus that primordial cells used only RNA for this purpose.1 The discovery that RNA molecules are also capable of catalyzing chemical reactions2,3 gave raise to the RNA World hypothesis for early life.4 This hypothesis posits that contemporary life, based on mutual coupling between nucleic acids and proteins, was preceded by a biochemistry in which RNA molecules played the dual role of self-replicating information carriers and enzymes. Even though the RNA World has become a canonical concept to explain the origins and early evolution of life,5-9 a number of important questions and unresolved issues persist. One such question is the following: why was ribose selected for its central role in the earliest polymers, even though analogs of RNA containing other sugars, capable of forming stable double helices and therefore suitable for storing and transferring genetic information, have been synthesized in the laboratory10-12 and could have existed in the prebiotic environment? A closely related question is why was synthesis of RNA from activated ribonucleotides not blocked by their diastereomers, which only differ in the position of exocyclic substituents in the furanose ring? Typically, incorporation of these diastereomers into nucleic acids is facile but terminates elongation of nascent chains or causes single-strand breaks.13-15 r 2011 American Chemical Society
If not repaired, these defects form the molecular basis for anticancer and antiviral activity of a number of arabinonucleosides.16,17 In the origin of life context, this implies that polymerization of primordial RNA took place in an environment in which the concentration of ribonucleotides or their activated derivatives was much higher than the concentration of the corresponding diastereomers. Recently, Sacerdote and Szostak18 made an unexpected finding that ribose permeates membranes markedly faster than its diastereomers: arabinose, xylose, and lyxose. This unusual phenomenon has been observed for bilayers made of both phospholipids, such as 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC), and much simpler amphiphiles, such as myristoleic and oleic acids. Membranes made of fatty acids and their mixtures with other simple amphiphilic molecules, such as long chain alcohols, are believed to resemble cell walls of the earliest ancestors of cells.19-22 On this basis Sacerdote and Szostak hypothesized that in the absence of contemporary, sophisticated mechanisms for achieving selectivity in synthesis and transport of nutrients, incorporation of ribose into information polymers might have been kinetically favored because it could have been supplied more quickly to primordial cells. Then, it would have been more readily available for synthesis of Received: December 21, 2010 Revised: February 16, 2011 Published: March 15, 2011 3681
dx.doi.org/10.1021/jp112104r | J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B
ARTICLE
Figure 1. Ball-and-stick diagram for nucleoside molecule. (a) Ribo-adenosine (rA); (b) arabino-adenosine (aA); (c) xylo-adenosine (xA).
nucleotides and their subsequent polymerization.18 However, the reason for the observed differences in permeation of ribose and its diastereomers remained unclear, especially since they could not be accounted for by conventional models for predicting rates of permeation.18,23 To explain the results of Sacerdote and Szostak, we carried out detailed computer simulations of how sugar molecules of interest move from water to the membrane.23 Not only did these simulations reproduce the measured rates of permeation but they also provided an explanation for the observed phenomenon. In both water and membrane, ribose and its diastereomers exist preferentially in the pyranose form. In aqueous solution, the exocyclic hydroxyl groups attached to the six-membered ring interact favorably with water molecules. Inside the membrane, however, water molecules are absent and, instead, the same groups of ribose form a network of favorable, hydrogen-bonding interactions with themselves. In other diastereomers, these groups are positioned differently and a similar network is not possible. As a result, the free energy barrier to the transfer of ribose across membranes is lower than the barrier to the transfer of arabinose and xylose. Consequently, ribose permeates membranes faster than its diastereomers. The hypothesis relies on an assumption that sugars were supplied to protocells and the subsequent synthesis of the building blocks for RNA proceeded inside the cell. To preserve the excess of ribose in the interior of the cell, synthetic processes must have occurred on time scales comparable to the rates of ribose permeation. A different scenario is also possible in which nucleosides or their activated derivatives were synthesized in the environment and they, rather than sugars, permeated membranes. The viability of this mechanism was demonstrated in laboratory experiments in which template-directed synthesis of RNA inside vesicles was driven by exogenous delivery of activated monomers.24-27 To examine the consequences of the latter scenario for possible preferential incorporation of ribose into primordial nucleic acids, we extend our earlier simulations to nucleosides containing the sugars of interest. In nucleosides, these sugars no longer exist in the six-membered, pyranose form, but instead are locked in the five-membered, furanose ring geometry. This changes the geometry of intramolecular interactions between the exocyclic groups, making them less likely to form a hydrogen-bonded network. As a consequence of these changes, preferential permeation of ribose over its diastereomers might not be preserved for the corresponding nucleosides and nucleotides. If this is the case, it leads to concrete constraints on the hypothesis put forward by Sacerdote and Szostak: the hypothesis can be valid only if the species transported across primordial cell walls were sugars rather than nucleosides or their activated derivatives. Oppositely, if preferential permeation of ribose extends to RNA monomers the hypothesis applies,
independently on which of these species were transported across the protocellular membrane. Our study consists of two stages. First, we calculate the partition coefficients of three nucleosides, ribo-adenosine (rA), arabino-adenosine (aA), and xylo-adenosine (xA), between water and decane, which serves as a model nonpolar phase. These three compounds are shown in Figure 1. According to the Overton rule,28 permeation rates of different solutes correlate with their partition coefficients between water and a nonpolar solvent similar to the interior of the membrane. Since measured partition coefficients are available for several analogs of the solutes of interest, not only does this stage provide a quick assessment of permeabilities but also serves as a test of the accuracy of our approach. In the second stage, we carry out simulations in which rA and aA are transferred across the POPC membrane, calculate the corresponding free energy changes and estimate permeability coefficients of the membrane to both species. The paper closes with conclusions about mechanisms of preferential transport across membranes and the validity of the selective permeation hypothesis.
’ METHODS Molecular Dynamics Simulations. Adenine nucleosides were simulated in two systems. One consisted of a lamella formed by 106 decane molecules directly in contact with a lamella containing 988 water molecules. The dimensions of the simulation box were 29 29 79 Å, with the z direction perpendicular to the interface. The average thickness of the decane and water lamella was 45 and 34 Å, respectively. The second system consisted of a bilayer built of 142 POPC molecules in contact with a lamella containing 5903 water molecules, enclosed in a 69 69 72 Å simulation cell. All-atom MD simulations were carried out using the package NAMD.29 The CHARMM force field30 was used to describe the nucleosides, decane, and the POPC membrane. Water molecules were represented as the TIP3P model. A time step of 1 fs was used to integrate the equations of motion. The temperature was kept at 298 and 303 K for the water-decane and water-POPC systems, respectively, using the Langevin friction force scheme with the damping coefficient at 5 ps-1. The pressure of 1 atm was maintained along the z-direction, whereas the x, y dimensions of the simulation cell were kept fixed. For the POPC bilayer, it ensured that the surface area per lipid headgroup was 68.3 Å2, equal to the value measured in recent X-ray scattering experiments.31 Periodic boundary conditions were applied in the three spatial directions. Long-ranged electrostatic interactions were calculated using Particle Mesh Ewald scheme with the grid 32 32 128 for the water-decane system, and 72 72 72 3682
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B for the water-POPC system. The cutoff for van der Waals (VDW) interactions was set to 13.5 Å. Free Energy Calculations. Adaptive Biasing Force (ABF) Simulations. The free energy profile, A(z), for transferring a nucleoside across the water-decane or water-POPC interface was calculated as a function of the order parameter z, defined as the z-component of the distance between the center of mass of the solute and the center of mass of either the decane lamella or the POPC bilayer. To do so, the ABF method32,33 was used. This method relies on integrating forces along the chosen coordinate in unconstrained MD simulations, during which the positiondependent average force is subtracted from the instantaneous force in an adaptive manner. This ensures uniform sampling of the order parameter independently of the shape of A(z). For the water-decane systems, free energy calculations were divided into four “windows” along z, each 6 Å wide. Harmonic restraints at the edges of each window kept the solute in the prescribed region. ABF simulations 10 ns long were carried out in the window located in water whereas simulations in the windows located in decane or at the interface were 20-30 ns long. For the water-POPC system, three ABF windows were used to cover the range of -25 Å < z < -3 Å, which extended from bulk water to the center of the POPC membrane (z = 0 Å). ABF simulations 80-100 ns long were performed in each window. To improve statistics in the membrane region, which is characterized by slow dynamics, additional ABF simulations 50-80 ns long were carried out in two windows spanning the range -10 Å < z < 6 Å and -7 Å < z < -3 Å. For the membrane center region of -3 Å < z < 3 Å, ABF simulations 80 ns long, in which updating the average force was suppressed for the last 20 ns, were performed to estimate A(z) for two different orientations of the nucleosides in the membrane (see below). The ABF method was also used to calculate the free energy profile, A(χ) around the glycosidic angle χ (C4-N9-C10 -O40 , see Figure 1a for atom labeling) in the adenine nucleosides. For this purpose trajectories 12 ns long were obtained for the solutes in water, decane, and at the center of the POPC membrane. Free Energy Perturbation (FEP) Method. FEP method was used to calculate the free energy differences between rA and aA molecule in water, decane, and at the center of the POPC membrane. The calculations were done using the dual-topology paradigm with a soft-core potential scheme,34,35 as implemented in the NAMD code. A strategy in which transformation of van der Waals interactions was delayed to avoid possible strong, spurious electrostatic repulsion was applied with the shift coefficient set to 5 and electrostatic interactions turned on at λ = 0.5. Details of this approach are described in the Supporting Information to a recent review of the FEP method.36 The Bennett Acceptance Ratio method37 was used in all calculations. The transformation of rA to aA in the aqueous and decane phases were carried out in 10 windows of equal size. In each window, the free energy difference was estimated from a 4 ns MD trajectory using the procedures, consistency checks and controls as recently recommended.36 For the transformation at the center of the POPC membrane, two sets of FEP calculations were performed, starting from two independent initial configurations. In one set, the transformation from rA to aA was carried out in 10 evenly spaced windows. The other set consisted of 16 windows. The first and the last two windows were twice as large as the remaining windows. Correspondingly, the trajectories used to estimate the change in free energy in these windows were twice as long (8 ns vs 4 ns).
ARTICLE
FEP calculations were also performed for transforming ribofuranose to arabinofuranose in both water and decane. The free energy difference was estimated from 1 ns trajectories obtained for each of 10 evenly spaced windows. The upper limit of the standard error for estimates of free energy differences obtained from the ABF method was calculated using a previously derived formula,38 as applied in our previous study on solute permeation through membranes.23 Standard error for individual windows in FEP was calculated as described by Bennett37 using the formula given in a recent review.36 The number of statistically independent configurations for the forward and backward transformations was estimated by correcting the total numbers of MD steps for the correlation length.39,40 Standard error for the whole transformation was calculated assuming that standard errors in individual windows were uncorrelated.
’ RESULTS AND DISCUSSIONS Partition Coefficients of Nucleosides between Water and Decane. According to the Overton rule,28 permeation rates of
solutes across membrane correlate with their hydrophobicity, which can be measured as their partition coefficients between water and a nonpolar solvent. Previously, we calculated the water/decane partition coefficients for aldopentoses and demonstrated that the Overton rule correctly predicts the higher permeability of membranes to ribose than its diastereomers.23 Here, we extend these calculations to nucleosides to determine whether the rule remains valid for these substantially larger solutes and to test the accuracy of the CHARMM force field for nucleosides.30 The partition coefficient is defined as P = e-ΔA/kBT, where ΔA is the free energy difference between the solute dissolved in water and the nonpolar solvent. We calculated this quantity across the water-decane system not only for rA, aA, and xA, which are the molecules of interest in this study, but also for adenine, 20 -deoxyadenosine (20 -dA) and 20 ,30 -dideoxyadenosine (20 ,30 ddA). The last three solutes were included because in contrast to rA, aA, and xA their partition coefficients were measured experimentally.41 The results are summarized in Table 1. Shown in Figure 2 are the calculated free energy profiles, A(z), for transferring different solutes from bulk water to decane. The difference between the end points of these profiles yields ΔA. For adenine, ΔA is equal to 9.4 ( 0.3 kcal/mol, which corresponds to a partition coefficient of 1.3 10-7, in close agreement with the experimental value of 3.7 10-7.41 For 20 -dA and 20 ,30 -ddA, ΔA is equal to 11.7 ( 0.2 and 9.0 ( 0.2 kcal/mol, respectively. Qualitatively, ΔA for 20 ,30 -ddA is expected to be lower because replacing one of the exocyclic hydroxyl groups of 20 -dA with a hydrogen atom in 20 ,30 -ddA decreases the solubility in water, which in turn reduces ΔA. The corresponding partition coefficients, equal to 2.6 10-9and 2.5 10-7 for 20 -dA and 20 ,30 ddA, respectively, are approximately an order of magnitude lower than the measured values of 8 10-8 and 3.6 10-6.41 This discrepancy can be markedly reduced by using an alternative charge model of the sugar ring in which the partial charge on oxygen atom, O40 , is reduced from -0.5, as used in CHARMM 27,30 to -0.4, as used in the earlier version of CHARMM.42 To ensure electroneutrality, the partial charges on the two carbon atoms (C10 and C40 ; see Figure 1a for atom labeling) bonded to the oxygen atom were reduced from 0.16 to 0.11. These modifications reduce ΔA for 20 -dA and 20 ,30 -ddA by -0.9 and 3683
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B
ARTICLE
Table 1. Free Energy Differences and Corresponding Partition Coefficients between Water and Decane Phase for Several Nucleosidesa ΔA (alternative charge model42)
change to the alternative charge model
(kcal/mol)
(kcal/mol)
(kcal/mol)
P
exp -7
3.7 10-7 8 10-8 3.6 10-6
adenine 20 -dA
9.4 ( 0.3 11.7 ( 0.2
9.4 10.8
-0.9
1.3 10 1.2 10-8
20 ,30 -ddA
9.0 ( 0.2
8.2
-0.8
0.9 10-6 -9
rA
12.6 ( 0.2
11.9
-0.7
1.9 10
aA
12.5 ( 0.2
11.8
-0.7
2.2 10-9
xA (30 -endo)
12.7 ( 0.2
12.0
-0.7
1.6 10-9
xA (20 -endo)
11.2 ( 0.2
10.5
-0.7
2.0 10-8
11.2
-0.7
b
xA a
ΔA (CHARMM2730)
11.9
-0.8 kcal/mol, respectively. This in turn increases the corresponding partition coefficients to 1.2 10-8 and 0.9 10-6, putting them in good agreement with the experimental values. Even more importantly, the calculated ratio of the partition coefficients for 20 ,30 -ddA and 20 -dA using either set of charges is approximately equal to 100, which is quite close to the ratio of 200 obtained experimentally.41 In conjuction with the data presented further in this paper, this indicates that the charge model influences only ΔA for individual species but not their relative values with respect to each other. These relative values are of interest here because they determine how the calculations should be interpreted. During the simulations we monitored the conformational equilibria of the nucleosides in both water and decane. In water, the sugar ring of rA was found to exist predominantly in the 20 -endo state (87%) whereas aA occupied mostly the 30 -endo conformation (84%). Very similar preferences, 63 and 76%, respectively, were determined from NMR measurements.43 These preferences are slightly enhanced in decane. Good agreement with experiment was also found for syn/anti equilibrium around the glycosidic bond. Simulations of rA in aqueous solution yielded equal populations of the anti and syn states in comparison with 75% anti conformation estimated on the basis of NMR experiments.43 In decane, the anti population increases to 85%. A similar trend was observed for aA; the population of the anti state increased from 64 to 98% during transfer from the aqueous solution to decane. In general, nucleosides appear to be more flexible in water than in nonpolar solvents, which can be attributed to stronger solute-solvent interactions. They effectively compete with intramolecular interactions in the nucleosides, preventing the solute from becoming locked in a single conformational state. Even though the calculated and experimentally determined conformational equilibria are in good agreement, on two occasions we observed differences that extend beyond the uncertainties in both methods. In the current and previous44 simulations, the dihedral angle γ around the C40 -C50 exocyclic bond (see Figure 1 for notation) is predominantly in the gaucheþ state whereas NMR data45 for 20 ,30 -ddA in water yield a 40 and 10% occupancy of the trans and gauche- states besides a majority population (50%) in the gaucheþ state. This discrepancy appears to be an artifact arising from the fact that the CHARMM force field was designed primarily to reproduce properties of nucleic acid polymers, which all have γ fixed at gaucheþ. Fortunately, underestimating the trans state is not expected to affect the properties investigated here because the hydroxyl group attached to C40 is too far from other exocyclic hydroxyl groups to be involved in intramolecular interactions.
0
6.7 10-9
Experimental data are from Xiang and Anderson (1994). Calculated using the experimental equilibrium between 3 -endo and 20 -endo of 72-28%.46 b
Figure 2. Free energy profiles for the transfer of several nucleosides from bulk water to decane. Black, red, green, blue, yellow, and brown curve is for rA, aA, xA, 20 -dA, 20 ,30 -ddA, and adenine, respectively.
Another discrepancy between calculations and experiments is in sugar pucker of xA. In our simulations, the dominant conformation is 20 -endo whereas, according to experimental estimates, xA exists mainly (72%) in the 30 -endo state.46 To remove this discrepancy, simulations of the transfer from water to decane were carried out with xA constrained to the 30 -endo and 20 -endo conformations. Then, the free energy of transfer was estimated taking Boltzmann averages of the free energies of transfer in these two conformations using experimentally determined populations in water of 72 and 28%.46 Alternatively, one might attempt to modify torsional potentials in the sugar ring to reproduce experimental results. However, considering the paucity of structural data for xylo nucleosides, this procedure would be largely ad hoc. In general, despite minor disagreements between simulations and experiments, we conclude that the CHARMM force field is appropriate for the study carried out here. For rA, aA, and xA, which are of main interest here, ΔA is nearly identical, respectively, equal to 12.5 ( 0.2, 12.6 ( 0.2, and 11.9 ( 0.2 kcal/mol. Modifications of the partial charges in the furanose ring, as described above, uniformly lower ΔA by 0.7 kcal/mol. The corresponding partition coefficients for the three nucleosides are similar and are equal to 1.9 10-9 for rA, 2.2 10-9 for aA, and a somewhat higher value of 6.7 10-9 for xA. FEP calculations in which rA in water or in decane is transformed to aA confirm these results. The corresponding free energy difference, ΔΔA, estimated from these calculations is equal to 0.4 kcal/ mol, substantially less than 1.1 kcal/mol obtained from similar 3684
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B
ARTICLE
Figure 4. Free energy profile for the transfer of ribo-adenosine (black curve) and arabino-adenosine (red curve) from water to POPC membrane center.
Figure 3. (a) Snapshot from ribo-adenosine (rA) in water-POPC membrane MD simulation. (b) Schematic plot of rA in the orientation with its base in the up direction along z, compared to the sugar (labeled as “up”), and in the opposite orientation (labeled as “down”). (c) Free energy profile of arabino-adenosine in the membrane interial. Dotted curves, with molecule in the up or down orientation (as defined in (b); solid black curve, symmetrized free energy.
transformations of sugars alone.23 These results are consistent with the fact that log P for rA and aA measured in the wateroctanol system are also very close (-1.05 for rA47 and -1.11 for aA48). Combining the results obtained here with those from our previous study23 leads to a conclusion that the Overton rule predicts different membrane permeabilities to ribose and its diastereomers but similar permeabilities to rA and its diastereomers. The latter prediction was directly tested in simulations described below. Free Energy Calculation of Nucleoside Molecule Across Membrane. We calculated the free energy profile, A(z), across the water-POPC bilayer interface for two nucleosides, rA and aA. Special interest in aA was motivated by the fact that this nucleoside is the only diastereomer of rA that has the 30 hydroxyl group in the same position. This group links with the next subunit in RNA. As a consequence, arabino derivatives are the most efficient competitors of ribonucleotides because they can incorporate into the growing polymer without substantial structural strain.
The calculations require special care due to slow conformational and orientational dynamics of the nucleosides. Specifically, anti/syn interconversion becomes quite slow inside the membrane. Its time scale appears to be of the order of 100 ns, compared to just a few nanoseconds in water. This means that satisfactory equilibration of the nucleosides around the glycosidic angle in the membrane would require simulations on the microsecond time scale. To avoid such extended simulations we carried out two sets of separate calculations inside the bilayer in which the nucleosides were either in the anti or the syn conformation. This yielded, respectively, free energy profiles, Aanti(z) and Asyn(z). A similar procedure was not necessary for sugar pucker, as the interconversion between 20 -endo and 30 endo conformations is markedly faster (∼1 ns in water and ∼10 ns in the membrane). Inside the membrane, the nucleosides prefer to adopt an elongated orientation along the z-direction with the hydrophilic sugar moiety remaining closer to the interface with water than the nucleobase. This is illustrated in Figure 3a, which is a snapshot from MD simulations. Because of the symmetry of the bilayer, this implies that the nucleosides reorient as they cross the center of the bilayer. Such reorientation was occasionally observed in unrestricted simulations, but its time scale was too slow to obtain equilibrated free energy profiles. Thus, in analogy to the approach taken for the anti and syn conformations, we carried out separate simulations for two different orientations of the nucleosides near the center of the bilayer (-3 Å < z < 3 Å). To describe molecular orientation, a parameter θ was defined as the angle between the vectorBr =Br (N7) þBr (N9) - 2 Br (C8) (see Figure 3b), and the z-axis. As illustrated in Figure 3b, two orientation ranges were defined: (1) for 0° e θ e 90° the nucleoside is orientated with its nucleobase directed up, toward the outgoing interface, compared to the sugar; (2) for 90° e θ e 180° the nucleobase is oriented down, toward the incoming interface. As seen in Figure 3c, the up orientation is favored on the incoming side of the membrane whereas the down orientation is preferred on the outgoing side. As expected, the two profiles are approximately symmetric with respect to the center of the bilayer and increase rapidly once the center of mass of the nucleoside crosses the midplane of the membrane. From the two profiles, it follows that at a fixed z one orientation is clearly preferred, unless the nucleoside is located very close to the center of the bilayer. For example, at z = ( 1.5 Å the ratio between the two orientations, estimated from the free energy difference between the two profiles, is as high as 30. 3685
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B
ARTICLE
Figure 5. Population of the dihedral angles of the hydroxyl group OH2, OH3, and OH4 of (a) ribo-adenosine; (b) arabino-adenosine in the waterPOPC system. Black curve, bulk water region (-25 Å < z < -17 Å); red curve, interface region (-17 Å < z < -10 Å); green curve, interface region (-10 Å < z < -3 Å); blue curve, membrane center region (-3 Å < z < -3 Å).
The calculations described above can be combined to obtain A(z) for the full range of z. In the region of -25 Å < z < -10 Å, which is away from the membrane and extends into the bulk water, equilibration between anti and syn conformations is sufficiently rapid to allow for estimating A(z) from ABF calculations without imposing any restraints on the glycosidic bond. In the region of -10 Å < z < -3 Å inside the membrane, A(z) is evaluated by weighting separate free energy profiles for the anti and syn states according to the Boltzmann distribution ( " #) -Asyn ðzÞ -Aanti ðzÞ AðzÞ ¼ - kB T ln Fanti exp þ Fsyn exp kB T kB T ð1Þ where A(z) is calculated relative to the free energy in bulk water, kB is the Boltzmann constant, T is temperature. and Fanti and Fsyn are the populations of the anti and syn state, respectively, in aqueous solution. In the center of the bilayer (-3 Å < z < 3 Å), two orientational states were also included in a similar fashion. The profiles for both states (0° e θ e 90° and 90° e θ e 180°) were symmetrized and the populations in these states were taken as equal by symmetry at z = 0. The full free energy profiles for both rA and aA are shown in Figure 4. They are similar with the free energy barriers to the transfer of rA and aA across the POPC membrane estimated at 10.0 ( 0.1 and 10.4 ( 0.1 kcal/mol, respectively. The small difference in the free energy barrier equal to only 0.4 ( 0.1 kcal/mol was confirmed in FEP calculations in which rA was alchemically transformed to aA either in bulk water or at the center of the POPC bilayer. In water, the free energy difference ΔAwater(rA f aA) is equal to -0.11 ( 0.07 kcal/ mol whereas in the membrane the corresponding difference ΔAPOPC (rA f aA) has been found to be 0.3 ( 0.1 or 0.4 ( 0.1 kcal/mol in two independent calculations. This yields ΔΔA = ΔAPOPC - ΔAwater= 0.4 or 0.5 ( 0.1 kcal/mol, consistent with the results obtained using ABF. In the FEP calculations, in the membrane only the anti state was considered because it provides the dominant contribution to ΔΔA due to strong preference for this conformation, while both anti and syn states were considered in the bulk water phase. For comparison, ΔΔA between transferring arabinose and ribose from water to the center of the POPC bilayer is markedly higher, equal to 2.3 kcal/mol.23
Permeability of the Nucleoside Molecules. Using the diffusion model in the steady state approximation, the permeability Pm of the membrane to the nucleosides depends on the free energy profiles A(z) and the position-dependent diffusion coefficient, Dz(z), along z as follows49,50 Z z2 AðzÞ=kB T 1 e dz ð2Þ ¼ Pm Dz ðzÞ z1
In bulk aqueous solution, Dz(z) was calculated from the Einstein relation applied to equilibrium simulations of aA in a box of pure water. In the membrane and in the interfacial region, Dz(z) was obtained from the autocorrelation function of random force acting on the center of mass of the solute at different, fixed positions along the z direction.51 The diffusion coefficient, equal to 1.0 10-5 cm2/s in bulk water, calculated in the NVE ensemble to avoid potential nonphysical effects of the Langevin thermostat, decreases to 0.4 10-6 cm2/s at the water-POPC interface and then increases slightly to 0.5 10-6 cm2/s at the center of the POPC membrane. Similar changes of the diffusion coefficient with the solute position in the membrane were observed previously.23,50,52,53 Dz(z) for rA is assumed to be similar to that for aA. The permeability coefficient estimated from eq 2 using the calculated A(z) and Dz(z) is equal to 9.0 10-7cm/s for rA and 5.3 10-7cm/s for aA. Thus, the permeation of rA is only slightly favored over that of aA (with the ratio of 1.7). A similar result is predicted from the partition coefficients measured across the water-octanol interface47,48 and calculated across the waterdecane system, both of which are nearly the same for rA and aA. The preference for ribose permeating membranes at a rate 10 times faster than that of its diastereomers,18 such as arabinose, xylose, and lyxose, is shown not to be preserved for the ribose-based nucleosides. This can be explained by considering how intramolecular interactions differ in these two cases. For sugars, which exist primarily in the pyranose form, MD simulations23 revealed that in ribose a strong hydrogen bondlike network between four exocyclic hydroxyl groups forms in nonpolar environments. A similar network does not exist in other aldopentoses, such as arabinose or xylose, because the hydroxyl substituents are not in the same positions. In aqueous solution the network is not stable because water molecules successfully compete for hydrogen-bonding interactions with the OH groups of the sugars. As a result, ribose is stabilized relative to its 3686
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B diastereomers in nonpolar environments but not in water. This lowers the free energy barriers to its permeation through membranes and increases the corresponding permeability coefficient. In contrast, the sugar ring in nucleosides is locked in the furanose form. In the five-membered ring the O-H...O angle between consecutive exocyclic groups is less conducive to favorable interactions than the angle in the pyranose form. In addition, two of the hydroxyl groups are replaced either by the base attached to C10 or by the exocyclic group attached to C40 . The stable hydrogen bondlike network seen in the pyranose ring of ribose is thus disrupted, and favorable intramolecular interactions inside the membrane are markedly reduced. Shown in Figure 5 are the populations of the dihedral angles formed by the exocyclic hydroxyl groups with the sugar ring of rA and aA. As can be seen in this figure, rotation of the OH groups is quite facile in aqueous solution. Therefore, all three staggered rotamers (trans, gaucheþ, and gauche-) are occupied in water. In the membrane interior, the population of the gaucherotamer of rA is markedly reduced, but both trans and gaucheþ states still remain significantly populated. A similar distribution of the OH dihedral angles was found for aA. This is in contrast to free ribose in which all four OH groups are locked in a specific conformation that facilitates formation of hydrogen bonds between consecutive exocyclic groups (see Figure 5 in ref 23). This confirms the observation that the difference in conformational preferences between ribose and its diastereomers in the membrane interior largely disappears for nucleosides. To show that the absence of favorable intramolecular interactions in ribose-based nucleosides, compared to their diastereomers, is a general feature of the furanose sugar ring, we carried out FEP calculations in which ribofuranose was transformed to arabinofuranose. The difference in free energy between these two sugars is quite small in both water and decane and is equal to 0.12 ( 0.1 and -0.06 ( 0.1 kcal/mol, respectively. For comparison, the same free energy changes for the sugars in the pyranose form in hexadecane and water are considerably larger, equal to 1.06 and -0.88 kcal/mol.23
’ SUMMARY AND CONCLUSIONS We studied unassisted transport of nucleosides, ribo-adenosine and arabino-adenosine, across POPC membrane using atomic-level MD simulations. Analysis of the orientation-dependent free energy landscape inside the membrane indicates that the most likely permeation pathway taken by the nucleosides involves flipping near the center of the membrane. This allows for avoiding high free energy barrier associated with being in an unfavorable orientation with respect to the nearest interface. Such a mechanism, which could also be involved in permeation of other large molecules, such as drugs possessing asymmetric charge distributions and non-negligible molecular dipoles, has to be accounted for in estimating permeation rates. The free energy barriers for crossing the POPC membrane and the corresponding permeability coefficients are quite similar for rA and aA. These molecules also exhibit similar partition coefficients measured across the water/octanol interface or calculated in the water/decane system, which indicates that the Overton rule applies to nucleosides, as it does to sugars.23 Even though the permeation coefficient of POPC membranes to xA was not calculated, larger partition coefficient of this nucleside across the water-decane interface compared to the corresponding coefficients for rA and aA indicates that rA does not permeate membranes faster than xA. The substantial reduction of the 10fold preference for transport of ribose over arabinose or
ARTICLE
xylose18,23 is attributed to the disruption of the stable, hydrogen bondlike network formed in nonpolar solvents by exocyclic groups of ribose, but not its diasteromers. The disruption occurs because the ribose ring in nucleosides is no longer in the pyranose form, but instead adopts the furanose form. Thus, the kinetic mechanism of selective permeation is not available to nucleosides. This provides a firm constraint that relates uptake of nutrients to the synthesis of the earliest polymers at the origins of life. Ribonucleotides could have had selective advantage over their diastereomers only if sugars permeated walls of primitive cells, and subsequent synthesis and polymerization of nucleotides took place inside cells. If, however, nucleosides or nucleotides were synthesized exogenously and then permeated cell membrane the same selectivity mechanism could not operate.
’ AUTHOR INFORMATION Corresponding Author
*E-mail: (C.W.)
[email protected]; (A.P.)
[email protected].
’ ACKNOWLEDGMENT This work was supported by the NASA Exobiology Program. NASA Advanced Supercomputing (NAS) Division provided computational resources needed to carry out this study. ’ REFERENCES (1) Joyce, G. F. New Biol. 1991, 3, 399. (2) Kruger, K; Grabowski, P. J.; Zaug, A. J.; Sands, J.; Gottschling, D. E.; Cech, T. R. Cell 1981, 31, 47. (3) Guerrier-Takada, C.; Gardiner, K.; Marsh, T.; Pace, N.; Altman, S. Cell 1983, 35, 849. (4) Gilbert, W. Nature 1986, 319, 618. (5) The RNA World; Gestland, R. F., Cech, T. R., Atkins, J. F., Eds.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 1999. (6) Orgel, L. E. Crit. Rev. Biochem. Mol. Biol. 2004, 39, 99. (7) Ellington, A. D.; Chen, X.; Robertson, M.; Syrett, A. Int. J. Biochem. Cell Biol. 2009, 41, 254. (8) Chen, X.; Li, N.; Ellington, A. D. Chem. Biodiversity 2007, 4, 633. (9) Robertson, M. P.; Joyce, G. F. The Origins of the RNA World. Cold Spring Harb. Perspect. Biol. 2001, 10.1101/cshperspect.a003608. (10) Deamer, D. W.; Pashley, R. M. Origins Life Evol. Biospheres 1989, 19, 21. (11) Mautner, M.; Leonard, D.; Deamer, D. Planet. Space Sci. 1995, 43, 139. (12) Dworkin, J.; Deamer, D.; Sandford, S.; Allamandola, L. Proc. Nat. Acad. Soc. U.S.A. 2001, 98, 815. (13) Sampath, D.; Rao, V. A.; Plunkett, W. Oncogene 2003, 22, 9063. (14) Hayakawa, Y.; Kawai, R.; Otsuki, K.; Kataoka, M.; Matsuda, A. Bioorg. Med. Chem. Lett. 1998, 8, 2559. (15) Azuma, A.; Huang, P.; Matsuda, A.; Plunkett, W. Mol. Pharmacol. 2001, 59, 725. (16) Grant, S. Adv. Cancer Res. 1988, 72, 197. (17) Nucleoside Analogs in Cancer Therapy; Cheson, B. D., Keating, M. J., Plunkett, W., Eds.; Marcel Dekker, Inc.: New York, 1997. (18) Sacerdote, M. G.; Szostak, J. W. Proc. Nat. Acad. Sci. U.S.A. 2005, 102, 6004. (19) Hargreaves, W. R.; Deamer, D. W. Biochemistry. 1978, 17, 3759. (20) Walde, P.; Wick, R.; Fresta, M.; Mangone, A.; Luisi, P. L. J. Am. Chem. Soc. 1994, 116, 11649. (21) Hanczyc, M. M.; Fujikawa, S. M.; Szostak, J. W. Science 2003, 302, 618. (22) Pohorille, A.; Deamer, D. W. Res. Microbiol. 2009, 160, 449. (23) Wei, C.; Pohorille, A. J. Am. Chem. Soc. 2009, 29, 10237. 3687
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
The Journal of Physical Chemistry B
ARTICLE
(24) Chakrabarti, A. C.; Breaker, R. R.; Joyce, G. F.; Deamer, D. W. J. Mol. Evol. 1994, 39, 555. (25) Oberholzer, T.; Albrizio, M.; Luisi, P. L. Chem Biol. 1995, 2, 677. (26) Monnard, P. A.; Luptak, A.; Deamer, D. W. Philos. Trans. R. Soc. London, Ser. B 2007, 362, 1741. (27) Mansy, S. S.; Schrum, J. P.; Krishnamurthy, M.; Tobe, S.; Treco, D. A.; Szostak, J. W. Nature 2008, 454, 122. (28) Overton, E. Vierteljahrsschr. Naturforsch. Ges. Zurich 1899, 44, 88. (29) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kale, L.; Schulten, K. J. Comput. Chem. 2005, 26, 1781. (30) MacKerell, A. D., Jr.; et al. J. Phys. Chem. B 1998, 102, 3586. (31) Kucerka, N.; Tristram-Nagle, S.; Nagle, J. F. J. Membr. Biol. 2005, 208, 193. (32) Darve, E.; Pohorille, A. J. Chem. Phys. 2001, 115, 9169. (33) Darve, E.; Rodríguez-Gomez, D.; Pohorille, A. J. Chem. Phys. 2008, 128, 144120. (34) Zacharias, M.; Straatsma, T. P.; McCammon, J. A. J. Chem. Phys. 1994, 100, 9025. (35) Beutler, T. C.; Mark, A. E.; van Schaik, R. C.; Gerber, P. R.; van Gunsteren, W. F. Chem. Phys. Lett. 1994, 222, 529. (36) Pohorille, A.; Jarzynski, C.; Chipot, C. Phys. Chem. B 2010, 114, 10235. (37) Bennett, C. H. J. Comput. Phys. 1976, 22, 245. (38) Rodriguez-Gomez, D.; Darve, E.; Pohorille, A. J. Chem. Phys. 2004, 120, 3563. (39) Straatsma, T. P.; Berendsen, H. J. C.; Stam, A. J. Mol. Phys. 1986, 57, 89. (40) Flyvbjerg, H.; Petersen, H. G. J. Chem. Phys. 1989, 91, 461. (41) Xiang, T. X.; Anderson, B. D. J. Membr. Biol. 1994, 140, 111. (42) MacKerell, A. D., Jr.; Wiorkiewicz-Kuczera, J.; Karplus, M. J. Am. Chem. Soc. 1995, 117, 11946. (43) Stolarski, R.; Dudycz, L; Shugar, D. Eur. J. Biochem. 1980, 108, 111. (44) Foloppe, N.; Nilsson, L. J. Phys. Chem. B 2005, 109, 9119. (45) Plavec, J.; Koole, L. H.; Chattopadhyyaya, J. J. Biochem. Biophys. Method 1992, 25, 253. (46) Akhrem, A. A.; Mikhailopulo, I. A.; Abramov, A. F. Org. Magn. Reson. 1979, 12, 247. (47) Ford, H., Jr.; Merski, C. L.; Kelley, J. A. J. Liq. Chromatogr. 1991, 14, 3365. (48) Hansch, C.; Leo, A. Log P Database; Pomona College Medicinal Chemistry Project: Claremont, CA, 1987. (49) Diamond, J. M.; Szabo, G.; Katz, Y. J. Membr. Biol. 1974, 17, 148. (50) Marrink, S. J.; Berendsen, H. J. C. J. Phys. Chem. 1994, 98, 4155. (51) Kubo, R. Rep. Prog. Phys. 1966, 29, 255. (52) Marrink, S. J.; Sok, R. M.; Berendsen, H. J. C. J. Chem. Phys. 1996, 104, 9090. (53) Xiang, T. X.; Anderson, B. D. Adv. Drug Delivery Rev. 2006, 58, 135.
3688
dx.doi.org/10.1021/jp112104r |J. Phys. Chem. B 2011, 115, 3681–3688
