- Email: [email protected]
Hydrophobicities of the nucleic acid bases: distribution coefficients from water to cyclohexane1
Article No. mb981880
J. Mol. Biol. (1998) 280, 421±430
Hydrophobicities of the Nucleic Acid Bases: Distribution Coefficients from Water to Cyclohexane Phoebe Shih1, Lee G. Pedersen2,3, Paul R. Gibbs1 and Richard Wolfenden1* 1
Department of Biochemistry and Biophysics and 2Department of Chemistry, University of North Carolina, Chapel Hill NC, 27599-7260, USA 3 Laboratory of Quantitative and Computational Biology, National Institute of Environmental Health Sciences, Research Triangle Park, NC, 27709, USA
To establish an experimental scale of hydrophobicities for the nucleic acid bases, comparable with a scale developed earlier for amino acid side-chains, these bases and their parent compounds (purine and pyrimidin-2-one) were converted to n-butylated and tetrahydrofurylated derivatives that are appreciably soluble in cyclohexane, a truly non-polar solvent that dissolves negligible water at saturation. Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, showed no evidence of self-association of the solute in either solvent, and the possibility of speci®c entrainment of water by solutes entering cyclohexane was ruled out by the results of experiments with tritiated water. In both the n-butyl and tetrahydrofuryl series, the bases span a range of 5.3 kcal molÿ1 in their free energies of transfer from water to cyclohexane, and are arranged in the following rank, in order of decreasing hydrophobicity: purine > thymine > adenine > uracil > pyrimidin-2-one > hypoxanthine 5 cytosine 5 guanine. In both series of pyrimidin-2-ones, hydrophobicity decreases with introduction of an amino substituent, but addition of an exocyclic keto group results in a modest enhancement of hydrophobicity; and free energies of transfer are relatively insensitive to the position of N-alkyl substitution. In both series of purines, hydrophobicity decreases with the introduction of exocyclic amino and keto groups, the keto group having the greater effect; and free energies of transfer vary substantially depending on the position of N-alkyl substitution. Several additional compounds were examined to test recent predictions based on SM5.4/A, a quantum mechanical self-consistent-®eld solvation model; and that model was found to yield values in reasonable agreement with the experimental results. # 1998 Academic Press
*Corresponding author
Keywords: nucleic acid bases; hydrophobicity; cyclohexane; partition coef®cients; free energy of transfer
Introduction The major nucleic acid bases participate in many kinds of interactions that involve structural complementarity, including base-pairing during transcription and translation of genetic information; Abbreviations used: THF, tetrahydrofuryl; Bu, butyl; Me, methyl; HPLC, high pressure liquid chromatography; A, adenine; C, cytosine; G, guanine; H, hypoxanthine; P, purine; T, thymine; U, uracil; Z, pyrimidin-2-one; chx, cyclohexane; chf, chloroform; Gtr, free energy of transfer; Kdist, distribution coef®cient; r, correlation coef®cient; FTIR, Fourier transform infrared spectroscopy. 0022±2836/98/280421±10 $30.00/0
enzyme-RNA recognition (particularly in class I aminoacyl-tRNA synthetases; Cusack, 1995); and binding of substrates, coenzymes and effectors by many enzymes. Non-covalent binding interactions of the nucleic acid bases, like those of other biological molecules, involve the stripping away of solvent water from regions of contact between the binding partners. Accordingly, the observed strength of their interactions with other molecules can be considered to include the cost of removing the interacting molecules (at least those parts that make contact with each other) from the solvent to which they were previously exposed (Wolfenden, 1983). In attempts to detect the presence of speci®c attractive or repulsive interactions between the # 1998 Academic Press
422 nucleic acid bases and the sites at which they are bound, it would be desirable to have information about the relative tendencies of the nucleic acid bases to leave water and enter a ``naive'' cavity of low dielectric constant, that does not furnish opportunities for H-bonding or other speci®c attractive interactions. Free energies of solvation of biological molecules also play a signi®cant role in determining the positions of metabolic equilibria in neutral solution. Thus, for example, the difference in free energy of solvation between reactants and products is so large that it fully explains the favorable equilibrium of hydrolysis of ATP to ADP and inorganic phosphate (Williams & Wolfenden, 1985). Equilibria of interconversion of the nucleic acid bases, such as those for the hydrolytic deamination of adenosine (Wolfenden, 1967) and cytidine (Cohen & Wolfenden, 1971), depend in part on the overall strengths of the interactions of reactants and products with watery surroundings, in addition to the relative strengths of covalent bonds that are present in the reactants and products. The hydrophobicities of the nucleic bases are not easy to test experimentally. When equipped with their normal ribose substituents, these bases are so polar that they leave water and enter non-polar solvents at concentrations that are far below the limits of detection by present methods of analysis. However, free energies of solvation have been found, in most cases, to behave as additive functions of the substituent groups that are present in solutes (Leo et al., 1971; Wolfenden et al., 1987). The problem of detection could thus be circumvented, at least in principle, by replacing ribose with less polar substituents, to perform distribution measurements at concentrations suf®cient to allow the bases to be compared with respect to their hydrophobicities. When methyl groups replace substituent ribose, for example, the bases have been shown to leave water and enter chloroform at measurable concentrations (Cullis & Wolfenden, 1981). Unfortunately, chloroform represents an unsatisfactory compromise, because the polarity of chloroform is considerable, and chloroform is known to form H-bonds of signi®cant strength (for a review, see Reichardt, 1988). That dif®culty is compounded by the fact that chloroform absorbs 0.1 M water at saturation, a concentration that is far in excess of the concentrations of N-methylated bases that enter chloroform in distribution experiments. For that reason, ``wet'' chloroform is considerably more polar than pure chloroform, and substantial numbers of water molecules are available in wet chloroform that might form H-bonds with dissolved solutes. Moreover, polar solutes have been shown to ``drag'' extra water molecules into semipolar solvents of moderate polarity in distribution experiments (Tsai et al., 1993). ``Water-dragging'', when that occurs, leads to overestimation of the true hydrophobicity of a solute, and is dif®cult to test experimentally in the case of chloroform, since dissolved water greatly
Hydrophobicities of Nucleic Acid Bases
exceeds the concentration of solutes entering chloroform; i.e. the concentration of ``excess'' water dragged into the chloroform represents a small difference between large numbers, and is consequently dif®cult to determine. For those reasons, it is hardly surprising that in earlier work, some discrepancies were observed between distribution coef®cients observed for methylated derivatives of the bases, between water and 2-butanol or chloroform, resulting in two provisional scales of hydrophobicity. The bases were found to behave somewhat differently in the two solvent systems, in both the range of free energies observed and the ranking of the bases in terms of increasing hydrophobicity (Cullis & Wolfenden, 1981). Similar dif®culties were encountered earlier in experiments involving amino acid side-chains (Radzicka & Wolfenden, 1988), and were ascribed to the considerable H-bonding potential of 2-butanol and chloroform, and the fact that both solvents dissolve substantial concentrations of water at saturation, clouding the interpretation of distribution measurements as a simple index of hydrophobicity. In view of these uncertainties, it seemed desirable to re-examine these questions by obtaining experimental values for the hydrophobicities of the bases, measured with respect to a non-polar condensed phase that contains little water at equilibrium. With a water content at saturation equivalent to that of the vapor phase (Wolfenden & Radzicka, 1994), cyclohexane represents a desirable reference phase in that respect. By replacing substituent ribose with n-butyl and tetrahydrofuryl groups (Figure 1), we have now obtained derivatives of the bases that are capable of entering cyclohexane at measurable concentrations, and to measure the relative distribution coef®cients of nucleic acid bases between neutral aqueous solution and cyclohexane. Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, show no evidence of self-association in either solvent, and experiments with tritiated water eliminate the possibility of signi®cant water-dragging, in this system. In addition to the bases of the genetic material, we
Figure 1. Structures of ribose and the R group (methyl, butyl or tetrahydrofuryl) in nucleic acid base derivatives used in the present study. The site of substitution is N1 for derivatives of pyrimidin-2-one and N9 for derivatives of purine.
423
Hydrophobicities of Nucleic Acid Bases
have examined the behavior of hypoxanthine, and of the parent compounds purine and pyrimidin-2one, in order to evaluate the apparent effects of ring substituents on hydrophobicity. Since the earlier experimental work of Cullis & Wolfenden (1981), molecular dynamics simulation methods have been used to estimate free energies of transfer of the nucleic acid bases or their methylated derivatives from water to the vapor phase (Bash et al., 1987; Cramer & Truhlar, 1992; Mohan et al., 1992; Orozco & Luque, 1993; Elcock & Richards, 1993; Young & Hillier, 1993; Gao, 1994; Miller & Kollman, 1996). Free energies of transfer from water to chloroform have also been estimated by combining those values with computed free energies of transfer from vapor to chloroform (Orozco et al., 1996; Young et al., 1994; Eksterowicz et al., 1997). In addition, Giesen et al. (1997) have predicted partition coef®cients from water to chloroform of several unnatural nucleic acid bases for which experimental data were not available. To test those predictions, we also determined waterto-chloroform partition coef®cients for methylated derivatives of three unnatural bases; 2-amino-9methylpurine, 6-methylamino-9-methylpurine, and 5-bromo-1-methyluracil.
Results Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, showed no evidence of self-association in either solvent; and the possibility of speci®c entrainment of water by solutes entering cyclohexane was ruled out by the results of experiments with tritiated water. Table 1 shows distribution coef®cients of N1-substituted pyrimidines and N9-substituted purine derivatives, between
Table 1. Hydrophobicity of base derivatives Gtr(w!chx) (kcal molÿ1)a Nucleic acid bases Purine Thymine Adenine Uracil Pyrimidin-2-one Hypoxanthine Cytosine Guanine
Butyl derivative 1.05 2.18 2.82 3.65 4.12 5.67 6.03* 6.67*
Tetrahydrofuryl derivative 2.56 4.08 4.25 5.02 5.55 7.14* 7.27* 7.48*
Dipole momentb 3.67 4.56 2.46 4.66 6.49 5.56 7.04 6.75
a Gtr ÿ RT ln K. Distribution coef®cients, K, were determined by the double extraction procedure as described in Materials and Methods. Standard errors on four determinations for each value are 4 0.2 kcal molÿ1. The values with an asterisk (*) have greater experimental uncertainty (4 0.8 kcal molÿ1) because the measured absorbance values are at or near the detection limit. b Computed from the ab initio wave function; units are debyes.
water and cyclohexane at 25 C. The distribution coef®cients of these butylated and tetrahydrofurylated bases were found to vary over a range of four orders of magnitude. Alkylation mixtures were usually found to yield both singly and multiply butylated products, even when only one equivalent of alkylating agent was present, but these products were easily separated by HPLC. Butylation of purine, for example, yields 3-, 7- and 9-n-butylpurines. Water-to-cyclohexane free energies of transfer for this series of butylpurines as well as for N1 and N3-butylated thymine and uracil are included in the ®rst section of Table 2. Several less common derivatives obtained from butylation reaction mixtures included 2-O-butylated cytosine, adenine butylated at the exocyclic amino group, adenine butylated at both N9 and the exocylic amino group, cytosine butylated at both N1 and the exocyclic amino group, and three dibutylated hypoxanthines (1,7; 1,9; 3,7). Their Gtr(w ! chx) values are listed in the second section of Table 2. To explore the effects of the position of alkylation, distribution coef®cients between water and chloroform (Gtr(w ! chx)) were determined for several purines and pyrimidines, and these values are listed in the third section of Table 2. In addition, dipole moments for purine and adenine methylated at N1, N3, N7 and N9, computed from the ab initio wave function, and the experimental and theoretical (Giesen et al., 1997) log K(Cchf/Cw) are included in Table 2.
Discussion Ordering of Gtr(w!chx) for butylated and tetrahydrofurylated bases, compared with earlier findings The present work establishes the hydrophobicities of the nucleic acid bases, determined with respect to a truly non-polar solvent that dissolves negligible water at saturation. In both the n-butyl and tetrahydrofuryl series, the results yield the following ranking in order of decreasing hydrophobicity: purine > thymine > adenine > uracil > pyrimidin-2-one > hypoxanthine 5 cytosine 5 guanine. In Figure 2, the present free energies of transfer for butylated derivatives from water to cyclohexane are used as a basis for plotting the values observed for transfer of tetrahydrofuryl derivatives to water (present data), and the values obtained earlier for transfer of methylated derivatives from water to chloroform (Cullis & Wolfenden, 1981). The similarity of these three scales plotted in Figure 2 indicates that the relative hydrophobicities of nucleic acid bases are little affected by the nature of the substituent, whether that be ribose, methyl, n-butyl or tetrahydrofuryl. Accordingly, these data furnish a self-consistent experimental index of the relative hydrophobicities of the bases themselves.
424
Hydrophobicities of Nucleic Acid Bases Table 2. Free energies of transfer, partition coef®cients, and dipole moments of base derivatives with single or multiple alkyl substitution Base derivatives A. To cyclohexane 9-n-Butylpurine 7n-Butylpurine 3-n-Butylpurine 1-n-Butylpurine 1-n-Butylthymine 3-n-Butylthymine 1-n-Butyluracil 3-n-Butyluracil B. To cyclohexane 2-n-Butoxycytosine 6-n-Butylaminopurine 6-n-Butylamino 9-n-Butylpurine 1-n-Bu- 4-n-butylamino pyrimidin-2-one 1,7-Di-n-Bu-hypoxanthine 1,9-Di-n-Bu-hypoxanthine 3,7-Di-n-Bu-hypoxanthine C. To chloroform 9-Methyladenine* 7-Methyladenine 3-Methyladenine 1-Methyladenine 9-Me-2-aminopurine 6,6-Di-Me-aminopurine 9-Me-6-methylaminopurine 5-Bromo-1-methyluracil 1-Methylthymine* 1-Methyluracil* 3-Methyluracil Thymine 6-Methyluracil Uracil Pyrimindin-2-one 1,3-Dimethyluracil 9-Methylpurine 6-Methylpurine 6-Chloro-purine Purine
Gtr(w!org)a (kcal molÿ1)
logK(Corg/Cw)b Experiment Calculated
1.05 3.55 2.67
ÿ0.77 ÿ2.60 ÿ1.96
2.18 1.84 3.65 3.65
ÿ1.60 ÿ1.35 ÿ2.68 ÿ2.68
1.53 4.08
ÿ1.12 ÿ2.99
ÿ1.36
1.00
ÿ0.95 0.80 1.65 3.00
0.70 ÿ0.59 ÿ1.21 ÿ2.20
1.06
ÿ0.78
2.55 4.12 0.69 0.27 ÿ0.47 0.89 0.61 1.65 1.72 3.09 3.36 4.08 4.22 ÿ0.89 ÿ0.16 2.28 1.86 2.82
ÿ1.82 ÿ3.28 ÿ0.50 ÿ0.20 0.35 ÿ0.66 ÿ0.45 ÿ1.21 ÿ1.26 ÿ2.26 ÿ2.47 ÿ2.99 ÿ3.10 0.65 0.11 ÿ1.67 ÿ1.37 ÿ2.07
Dipole momentc 4.13 6.50 4.72 7.89
ÿ1.6
ÿ1.9
2.71 7.31 4.56 9.16
ÿ0.3 ÿ0.3 ÿ0.3 ÿ1.2
a
Gtransfer ÿ RT ln K. Distribution coef®cients, K, were determined from double extraction procedure as described in Materials and Methods. Cyclohexane was used as the organic layer for butylated and tetrahydrofurylated derivatives and chloroform was used as the organic layer for methylated derivatives. For 1-methyl adenine (pKa 7.2; Brookes & Lawley, 1960), 50 mM Tris-HCl (pH 9) was used as the aqueous phase. Standard errors on four determinations for each value are 4 0.2 kcal molÿ1. Base derivatives with * denote those whose values had been determined previously (Cullis & Wolfenden, 1981). Calculated K values were obtained from Giesen et al. (1997). c Computed from the ab initio wave function; units are debyes. b
Not surprisingly, substantial discrepancies are observed between these values and those determined earlier for the methylated bases using 2-butanol as a reference phase, as to both the ranking of the bases and the span of free energies observed (Cullis & Wolfenden, 1981). That discrepancy can probably be attributed to formation of H-bonds to solvent 2-butanol, a phenomenon encountered earlier in the distribution of amino acid side-chains between water and 1-octanol (Radzicka & Wolfenden, 1988). An additional complicating factor, tending to invalidate alcohols as simple reference phases for the measurement of hydrophobicity, is their tendency to dissolve considerable
concentrations of water (9.4 M in the case of 2-butanol) at saturation (Radzicka & Wolfenden, 1994). Although wet chloroform is moderately polar and dissolves 0.1 M of water at saturation (Leo et al., 1971), this solvent appears to offer a fair compromise as a semi-polar reference solvent, in that a good correlation is observed between values for Gtr(w ! chf) and Gtr(w ! chx) (Figure 2). This correspondence between the two scales is probably fortuitous to some extent, and cannot be expected to apply in every case that might be examined. In that context, it may be worth noting that the 13C chemical shifts of C2, C4, C5, C6 and C8 of 9-nbutylpurine in deuterated chloroform are slightly
425
Hydrophobicities of Nucleic Acid Bases
Figure 2. Gtr(w ! chx) values of butylated base derivatives are plotted against Gtr(w ! chx) values for tetrahydrofurylated derivatives (slope 0.89 0.04, r 0.99) and Gtr(w ! chf) values for methylated derivatives (slope 0.93 0.05, r 0.99).
closer to those in deuterated cyclohexane than to those in deuterated water (unpublished results). Substituent effects Amino and keto functionality on purine and pyrimidin-2-one Gtr(W ! chx) values for butylated purines and pyrimidines are plotted in Figure 3 alongside their structures. Considered as derivatives of the parent compounds purine and pyrimidin-2-one, purines and pyrimidin-2-ones differ in their responses to the introduction of ring substituents. Introduction of an amino or a keto functionality to the purine system decreases hydrophobicity, the keto group having the greater effect. Addition of an amino group to the hypoxanthine ring, already containing a keto group, does not change the hydrophobicity much further. Thus, the effects of substitution fall short of expectations based on any simple additivity relationship. Introduction of an amino substituent to pyrimidin-2-ones decreases hydrophobicity, to an extent similar to that observed in the purines; but addition of a keto group results in a small increase in hydrophobicity. Accordingly, keto substituents exert different effects on water af®nity in the two ring systems. Earlier experimental studies have shown that neighboring atoms in a molecule sometimes interact in ways that appear to reinforce or hinder the polarizing effect of solvating water molecules (Hine &
Figure 3. Structures of purines and pyrimidines are shown alongside the plot of Gtr(w ! chx) for butylated purines and pyrimidines (see Table 1 for experimental details). In nucleic acid bases, the R group is hydrogen and in nucleosides, the R group is ribose. For the base derivatives used in the present study, the R group is a methyl, butyl or tetrahydrofuryl substituent.
Mookerjee, 1975; Wolfenden et al., 1987), as in the present case. Using a fragment-based partitioning model, Giesen et al. (1997) have examined the degree to which adjacent functional groups may interact in a non-linear fashion to in¯uence free energies of distribution. Their analyses indicate that the transferability of an apparent group contribution to the overall partition coef®cient, from one molecule to another, depends strongly on the extent to which the functional group is isolated from other interactions. Halogenated purine and pyrimidine derivatives Bromo substitution at the C5 position of 1-methyluracil increases its tendency to enter the chloroform phase by 0.8 kcal molÿ1, and chloro substitution at C6 of purine enhances its af®nity for the chloroform phase by nearly 1 kcal molÿ1 (Table 2). Although not strictly comparable, these enhancements of apparent hydrophobicity resemble those observed for methyl substitution in
426 both the purine and the pyrimidine systems, as may be seen by comparing Gtr values for 1-methylthymine and 6-methylpurine. Effects of C, N and O-alkylation and multiple alkylation An interesting minor product, obtained by butylation of cytosine, was identi®ed as 2-nbutoxycytosine, by X-ray crystallography, 1H NMR, UV, and FTIR. In terms of Kdist(w ! chx), this compound is more than three orders of magnitude less hydrophobic than is 1-n-butyl cytosine. This effect is similar in magnitude to the effect observed (102.5) when 2-methoxypyridine is compared with N-methyl pyridone, using hexane as the non-polar phase (Cullis & Wolfenden, 1981). In contrast, the hydrophobicity of 6-n-butyladenine is only 8.4-fold lower than that of 9-n-butyladenine. When both N9 and the exocyclic amino group of adenine are butylated, the product actually favors the cyclohexane layer with Kdist 10. Similarly, when both N1 and the exocylic group of cytosine are alkylated, the product also favors the cyclohexane layer, with Kdist 5. Values are given for several dibutylated hypoxanthines in the second section of Table 2. In both purines and pyrimidines, N-methylation is found to produce a much larger enhancement in hydrophobicity than does C-methylation. Thus, N1 and N3 methylation of uracil yielded a more favorable Gtr(w ! chf) by 2.4 kcal molÿ1; whereas methylation at C5 (thymine) and C6 changed Gtr(w ! chf) only by 1 and 0.7 kcal molÿ1, respectively (third section of Table 2). In the purine system, methylation at N9 and C6 resulted in Gtr(w ! chf) of 3 and 0.5 kcal molÿ1, respectively. The difference between the effects of N-methylation and C-methylation seems understandable in terms of the ability of ring N-H groups to form Hbonds to solvent water, and the likelihood that N-methylation results in a major disruption of these interactions. In contrast, the introduction of a methylene increment, by methylation of a carbon atom, has typically been found to enhance hydrophobicity (expressed in terms of Kdist(water ! hydrocarbon) by 0.96 kcal molÿ1, comparable in magnitude with the modest effects observed here (Wolfenden & Lewis, 1976). Effect of the site of alkylation on hydrophobicity Whereas N1 and N3-n-butyluracil have identical partition coef®cients, the hydrophobicity of N7-nbutylpurine differs from that of N9-n-butyl purine by nearly two orders of magnitude. N1, N3 and N9-methyl adenine also varied over a range of more than 300-fold in hydrophobicity. These differences are perhaps not surprising, in view of the likelihood that the different nitrogen atoms, at which substitution occurs, differ from each other in
Hydrophobicities of Nucleic Acid Bases
the strengths and directional preferences of their interactions with solvent water. Relationships between Gtr(w!chx) and dipole moment The dipole moments of purines and pyrimidines, calculated from the ab initio wave function, are included in Table 1 for the underivatized bases, and in Table 2 for methylated bases. For the bases considered as a whole, there is a moderately close relationship between dipole moment and values for Gtr(w ! chx), with r 0.79. Factors other than H-bonding capacity and dipole moment, such as inductive effects and steric crowding, may also contribute to the relative hydrophobicities of the nucleic acid bases and their derivatives. In the pyrimidine series, the low hydrophobicity of cytosine may be related to its high dipole moment, whereas the most hydrophobic of the series, thymine, has the lowest dipole moment. Pyrimidin-2-one and uracil exhibit similar hydrophobicity, perhaps as a result of compensation between H-bonding capacity and dipole moment. In the purine series, the relatively high dipole moment and H-bonding capacity of guanine and hypoxanthine presumably account for their low hydrophobicity. The parent compound purine has a low dipole moment (3.67 debyes) and is least capable of forming H-bonds with solvent water, explaining its high hydrophobicity. Although adenine has an even lower dipole moment (2.46 debyes) than purine, this appears to be more than fully compensated by its ability to form two additional H-bonds to solvent water. The calculated dipole moments of the N-methylated purines and adenine (Table 2) may be helpful in understanding their variations in hydrophobicity. In both series, the H-bonding capacity is equivalent within a given series, rendering the comparison between dipole moment and hydrophobicity relatively straightforward. Plots of Gtr(w ! chx) for butylated purines and of Gtr(w ! chf) for methylated adenine against calculated dipole moments gave r 0.90 and 0.97, respectively. Comparison with theoretical results A recent computational study (Giesen et al., 1997) used the continuum solvation model SM5.4/ A, yielded estimated log K(Cchf/Cw) that were chosen for comparison with our experimental log K(Cchf/Cw) (Table 2) for two reasons. First, this method yielded values for methylated forms of the six common bases (1-Me-T, 1-Me-U, 1-Me-C, 9-MeA, 9-Me-G and 9-Me-H) that were in closer agreement with earlier experimental results (Cullis & Wolfenden, 1981) than were other calculated values of log K(Cchf/Cw) (see Eksterowicz et al. (1997) for a comparison of experimental results with other theoretical results). Second, Giesen et al. (1997) predicted log K(Cchf/Cw) for several methyl-
427
Hydrophobicities of Nucleic Acid Bases
ine. In both systems, halogenation was found to enhance the hydrophobicity of the parent compounds, and the effect was similar to that effect of methylation.
Materials and Methods Chemicals Purine and pyrimidine bases, N1, N3 and N9-methylated adenine; 2-aminopurine, 6,6-dimethylaminopurine, 6-methylaminopurine, 2,6-diaminopurine, 5-bromo-1methyluracil, 6-chloropurine, 6-methylpurine, 3-methyluracil, 6-methyluracil were obtained from either Sigma Chemical Co. or Aldrich Chemical Co. 2-t-Butoxytetrahydrofuran was purchased from Lancaster Chemical Co. Tetrabutylammonium hydroxide and tetramethylammonium hydroxide were obtained from Aldrich Chemical Co. Figure 4. Log K(Cchf/Cw) values, estimated by the contiuum solvation model SM5.4/A (Giesen et al., 1997) are plotted separately against log K(Cchf/Cw) values for the six natural bases determined previously (Cullis & Wolfenden (1981); ®lled circle and broken line, slope 1.30 0.18, r 0.96) and for all nine bases (including the three new derivatives, open circle and ®t by the thick continuous line, slope 1.15 0.18, r 0.93). Experimental log K(Cchf/Cw) values are rounded off to the same signi®cant digit as the calculated values.
Distribution coefficients determination Distribution coef®cients between water (w) and cyclohexane (chx) or chloroform were measured by a doubleextraction procedure (Bone et al., 1983), ®rst extracting the compound from water into cyclohexane at 25 C, separating the phases, then back-extracting from cyclohexane into a fresh aqueous phase. At both stages, the two phases were stirred vigorously for at least one hour at 25 C. Apparent distribution coef®cients were calculated by the following equation: KM!L
ated unnatural nucleic acid bases for which experimental data had not been available, offering an opportunity for a direct experimental test. Experimental and calculated log K(Cchf/Cw) values for 2-amino-9-methylpurine, 6-methylamino-9-methylpurine and 5-bromo-1-methyl-uracil are compared in the third section of Table 2. In Figure 4, calculated log K(Cchf/Cw) values are plotted separately against log K(Cchf/Cw) values for the six natural bases determined previously (broken line) and for all nine bases (thick continuous line), yielding r 0.96, and 0.93, respectively. Inclusion of the three new data points changed the slope of the correlation line slightly, from 1.30 0.18 to 1.15 0.18. However, the near-unit slope of the correlation line indicates that the span and absolute magnitude of the predicted values is approximately correct. Of particular interest is the calculated log K(Cchf/Cw) value of one of the new base derivatives, 5-bromo-1-methyluracil, which was suggested as implying that ``bromine is net hydrophobic in (the C-5) position to about the same extent as a methyl group'' (Giesen et al., 1997). As a bromo group is ordinarily considered to be more electronegative than a methyl group, its tendency to enhance hydrophobicity to the same extent as a methyl group was not, to us, intuitively obvious. That prediction led us to examine the effect of bromo substitution in 5-bromo-1-methyl-uracil, and the effect of chloro substitution in 6-chloropur-
BM LA ÿ LB
where KM ! L is the equilibrium constant for transfer to the less favored solvent, M is the volume of the more favored solvent used in both extractions, B is the concentration of solute in more favored solvent after second or back extraction, L is the volume of the less favored solvent, and A is the concentration of solute in the more favored solvent after ®rst extraction. Whenever possible (solutes with KM ! L > 8 10ÿ4), experiments were performed at initial solute concentrations varying over a range of at least 20-fold (typically 0.3 to 6 mM). Distribution coef®cients were found to be invariant with changing solute concentration, consistent with the view that these values would apply also at in®nite dilution. Furthermore, ultraviolet spectra were taken for solutes that dissolve in signi®cant concentration in cyclohexane, and the lmax of every spectrum examined closely resembles that in water. Therefore, the same tautomeric species are present in both phases. The following lmax values (in nm) were used in the distribution experiments: 9-n-BuA, 261; 6-n-butylaminoP, 269; 6-n-butylamino-9-Bu-P, 270; 1-n-Bu-C, 275; 2-n-butoxy-C, 270; 1-n-butylamino-4n-Bu-Z, 274; 9-n-Bu-G, 253; 9-n-Bu-H, 251; 1,7-di-n-Bu-H, 257; 1,9-di-n-Bu-H, 253; 3,7-di-n-Bu-H, 266; 3-n-Bu-P, 276, 7-n-Bu-P, 266, 9-n-Bu-P, 264; 1-n-Bu-Z, 304; 1-n-BuT, 273; 3-n-Bu-T, 265; 1-n-Bu-U, 268; 3-n-Bu-U, 259; 9-(2THF)-A, 262; 1-(2-THF)-C, 272; 9-(2-THF)-G, 253; 9-(2THF)-H, 250; 9-(2-THF)-P, 263; 1-(2-THF)-Z, 302; 1-(2THF)-T, 269; 1-(2-THF)-U, 264; 1-Me-A, 270; 3-Me-A, 273; 9-Me-A, 262; 6-Cl-P, 265; 6-Me-P, 261; P, 263; 9-MeP, 264; 6-6-dimethylamino-P, 276; 6-methylamino-9-MeP, 268; 2-amino-9-Me-P, 304; 3-Me-U, 258; 6-Me-U, 261; !Me-U, 267; 1,3-di-Me-U, 266; 5-Br-!-Me-U, 283; 1-Me-T, 272; T, 265; Z, 299; U, 259. Distribution coef®cient for each compound was determined based on four or more
428 experiments. Standard errors for each Gtr(w ! chx) value are 4 0.2 kcal molÿ1 for compounds with Gtr(W ! chx) 46 kcal molÿ1 (or KM ! L 5 4 10ÿ5). Gtr(w ! chx) values that are 56 kcal molÿ1 (for compounds with KM ! L 4 4 10ÿ5) are subjected to greater uncertainty (4 0.8 kcal molÿ1) because the measured absorbance values are at or near the detection limit. Water entrainment Experiments with tritiated water showed that solutes entering cyclohexane (9-n-butylpurine, for example, at a concentration of 4 10ÿ2 M in the cyclohexane layer) did not enhance the concentration of water entering cyclohexane above the concentration (9 10ÿ4 M) that enters cyclohexane in the absence of solutes. Thus, water entrainment appeared to be negligible in these experiments. Synthesis of butylated and methylated purine and pyrimidine derivatives Purine and pyrimidine bases were butylated using tetrabutylammonium hydroxide (Meyers & Zeleznick, 1963). A solution of the base (5 mmol) and tetrabutylammonium hydroxide (5 mmol) in water (10 ml) was lyophilized. The hygroscopic residue was heated slowly in a sublimation apparatus under vacuum to a temperature ranging from 160 to 250 C. Heating was continued for three hours at the temperature (160 to 250 C) when a sublimate began to collect on the condenser. The crude mixture collected in the sublimator typically consisted of starting materials and singly and multiply butylated bases. These products were resolved by reverse-phase HPLC, using a preparative Whatman Partisil 10 ODS-2 column (22 mm 500 mm). The starting material was eluted with 100% water, followed by elution of the butylated base bases with a series of 5% steps in the concentration of methanol. Products were identi®ed by their UV and proton nuclear magnetic resonance spectra. The approximate yields and the concentrations of methanol required for elution are as follows: 9-n-Bu-A, HPLC, 35% methanol; yield, 40%; 6-n-butylamino-P, HPLC, 40% methanol; yield, 40%; 6-n-butylamino-9-n-Bu-P, HPLC, 80% methanol; yield, 25%; 1-n-Bu-C, HPLC, 25% methanol; yield, 30%; 2-n-butoxy-C, HPLC, 60% methanol; yield, 35%; 2-Hydroxy-1-n-Bu-4-butylaminopyrimidine (1,4-Di-n-butylcytosine), HPLC, 80% methanol; yield, 20%; 9-n-Bu-G, HPLC, 20% methanol; yield 25%; 9-n-BuH, HPLC, 20% methanol; yield, 30%; 1,7-di-n-Bu-H, HPLC, 75% methanol; yield, 15%; 1,9-di-n-Bu-H, HPLC, 60% methanol; yield, 10%; 3,7-di-n-Bu-H, HPLC, 40% methanol; yield, 10%; 3-n-Bu-P, HPLC, 45% methanol; yield, 15%; 7-n-Bu-P, HPLC, 30% methanol; yield, 65%; 1-n-Bu-T, HPLC, 35% methanol; yield, 50%; 3-n-Bu-T, HPLC, 40% methanol; yield, 35%; 7-n-Bu-P, HPLC, 40% methanol; yield 25%; 9-n-Bu-P, HPLC, 50% methanol; yield, 30%; 1-n-Bu-Z, 1-n-Bu-U, HPLC, 25% methanol; yield, 35%; 3-n-Bu-U, HPLC, 30% methanol; yield, 40%. 2-Aminopurine, 6-methylaminopurine, purine and uracil were methylated using tetramethylammonium hydroxide. The synthetic and puri®cation procedures as described above for the butylation reaction were employed. Products were identi®ed by their UV and proton nuclear magnetic resonance spectra. The approximate yields and the concentrations of methanol required for elution are as follows: 2-amino-9-Me-P, HPLC, 30% methanol; yield, 40%; 6-methylamino-9-Me-P, HPLC,
Hydrophobicities of Nucleic Acid Bases 50% methanol; yield, 40%; 9-Me-P, HPLC, 50% methanol; yield, 20%; 1,3-di-Me-U, HPLC, 75% methanol; yield, 30%. Identification of 2-n-butoxycytosine One product isolated from the butylation of cytosine was shown by 1H NMR to be monobutylated, but its UV spectra in acidic, neutral, and basic conditions did not agree with those reported for N1, N3 or N4-alkylated cytosine. Instead, they were similar to those reported for 2-methoxy-4-amino-pyrimidine (Shugar & Fox, 1952). Since O-alkylation of bases by the present method is unusual, X-ray diffraction and FTIR were used to con®rm its identi®cation as 2-n-butoxycytosine. Crystals of this compound were grown in water at room temperature, and its structure was determined by Dr Peter White at the Single Crystal X-Ray Facility of the Chemistry Department of the University of North Carolina at Chapel Hill. Diffraction data for a monoclinic crystal of 0.2 mm 0.2 mm 0.2 mm dimensions were collected using a Rigaku diffractometer in the omega scan mode at 25 C. In all, 5403 re¯ections were measured, and the RF value for signi®cant re¯ections was 0.121. Structural data were consistent with butylation site at the O2 position. FTIR spectra were obtained from KBr pellets at 25 C at 2 cmÿ1 resolution using a Mattson First Galaxy series 5000 FTIR spectrometer. As expected, this compound lacked a prominent carbonyl absorption near 1625 cmÿ1 that is present in 1-n-butylcytosine and 3-methylcytosine. Synthesis of tetrahydrofuryl purine and pyrimidine derivatives Tetrahydrofuryl derivatives of the bases were prepared using 2-t-butoxytetrahydrofuran as the alkylating agent, as described by Kametani et al. (1977). A mixture of the base (8 mmol) and 2-t-butoxytetrahydrofuran (12 mmol) in dimethylformamide (5 ml) was heated with stirring at 170 C for ®ve hours. After evaporation of the solvent, the product was dissolved in 10 ml of chloroform, washed twice with water, and concentrated to 2 ml before chromatography on a Whatman Partisil 10 ODS-2 preparative HPLC column (22 mm 500 mm). The major products obtained are 1-substituted pyrimidines and 9-substituted purines, although other products were sometimes obtained. The starting material was eluted with 100% water, followed by elution of the tetrahydrofurylated bases with a series of 5% steps in the concentration of methanol. Products were identi®ed by their UV and proton nuclear magnetic resonance spectral data. The approximate yields and concentrations of methanol required for elution are as follows: 9-(2-THF)A, HPLC, 25% methanol; yield, 60%; 1-(2-THF)-C, HPLC, 10% methanol; yield, 55%; 9-(2-THF)-G, HPLC, 10% methanol; yield, 50%; 9-(2-THF)-H, HPLC, 10% methanol; yield, 45%; 9-(2-THF)-P, HPLC, 40% methanol; yield, 45%; 1-(2-THF)-Z, HPLC, 20% methanol; yield, 30%; 1-(2-THF)-T, HPLC, 30% methanol; yield, 35%; 1-(2THF)-U, HPLC, 25% methanol; yield, 55% Calculation of dipole moments Dipole moments for analogs of selected molecules indicated in Table 1 (for underivatized purine and pyrimidines) and Table 2 (for methylated derivatives) were estimated by use of the ab initio quantum mechanical
Hydrophobicities of Nucleic Acid Bases package Gaussian 94 (Frisch et al., 1995) on a Cray model T90 at the North Carolina Supercomputer Center. The initial geometries of the molecules were energy-optimized at the 6-311G (d,p) basis set level, which includes polarization orbitals on all atoms.
Acknowledgments This work was supported by research grants GM18325 to R.W. and HL-06350 to L.G.P. from the National Institutes of Health. P.S. acknowledges the support of a postdoctoral fellowship (PF-4156) from American Cancer Society.
References Bash, P. A., Singh, U. C., Langridge, R. & Kollman, P. A. (1987). Free energy calculations by computer simulation. Science, 236, 564± 568. Bone, R., Cullis, P. & Wolfenden, R. (1983). Solvent effects on equilibria of addition of nucleophiles to acetaldehyde and the hydrophilic character of diols. J. Am. Chem. Soc. 105, 1339± 1343. Brookes, P. & Lawley, P. D. (1960). The methylation of adenosine and adenylic acid. J. Chem. Soc. 539± 545. Cohen, R. M. & Wolfenden, R. (1971). The equilibrium of hydrolytic deamination of cytidine and N4methylcytidine. J. Biol. Chem. 246, 7566 ±7568. Cramer, C. J. & Truhlar, D. G. (1992). Polarization of the nucleic acid bases in aqueous solution. Chem. Phys. Letters, 198, 74 ±80. Cullis, P. M. & Wolfenden, R. (1981). Af®nities of nucleic acid bases for solvent water. Biochemistry, 20, 3024± 3028. Cusack, S. (1995). Eleven down and nine to go. Nature Struct. Biol. 2, 824± 831. Eksterowicz, J. E., Miller, J. L. & Kollman, P. A. (1997). Calculation of chloroform/water partition coef®cients for the N-methylated nucleic acid bases. J. Phys. Chem. ser. B, 101, 10971± 10975. Elcock, A. H. & Richards, W. G. (1993). Relative hydration free energies of nucleic acid bases. J. Am. Chem. Soc. 115, 7930± 7931. Frisch, M. J., Trucks, G. W., Schlegel, H. B., Gill, P. M. W., Johnson, B. G., Robb, M. A., Cheeseman, J. R., Keith, T., Petersson, G. A., Montgomery, J. A., Raghavachari, K., Al-Laham, M. A., Zakrzewski, V. G., Ortiz, J. V., Foresman, J. B., Cioslowski, J., Stefanov, B. B., Nanayakkara, A., Challacombe, M., Peng, C. Y., Ayala, P. Y., Chen, W., Wong, M. W., Andres, J. L., Replogle, E. S., Gomperts, R., Martin, R. L., Fox, D. J., Binkley, J. S., Defrees, D. J., Baker, J., Stewart, J. P., Head-Gordon, M., Gonzalez, C. & Pople, J. A. (1995). Gaussian 95, Gaussian, Inc., Pittsburgh PA. Gao, J. (1994). The hydration and solvent polarization effects of nucleotide bases. Biophys. Chem. 51, 253± 261. Giesen, D. J., Chambers, C. C., Cramer, C. J. & Truhlar, D. G. (1997). What controls partitioning of the nucleic acid bases between chloroform and water?
429 (erratum to be published). J. Phys. Chem. ser. B, 101, 5084± 5088. Hine, J. & Mookerjee, P. K. (1975). The intrinsic hydrophilic character of organic compounds. Correlations in terms of structural contributions. J. Org. Chem. 40, 292± 298. Kametani, T., Kigasawa, K., Hiiragi, M., Wakisaka, K., Kusama, O., Kawasaki, K. & Sugi, H. (1977). Studies on the synthesis of chemotherapeutics. I. Synthesis of 1-(2-tetrahydrofuryl)-5-¯uororuracil. J. Heterocyc. Chem. 14, 473± 475. Leo, A., Hantsch, C. & Elkins, D. (1971). Partition coef®cients and their uses. Chem. Rev. 71, 525± 592. Meyers, T. C. & Zeleznick, L. (1963). Alkylation of the purine nucleus by means of quaternary ammonium compounds. I. Tetraalkylammonium hydroxides. J. Org. Chem. 28, 2087± 2089. Miller, J. L. & Kollman, P. A. (1996). Solvation free energies of the nucleic acid bases. J. Phys. Chem. 100, 8587± 8594. Mohan, V., Davis, M. E., McCammon, J. A. & Pettitt, B. M. (1992). Continuum model calculations of solvation free energies: accurate evaluation of electrostatic contributions. J. Phys. Chem. 96, 6428± 6431. Orozco, M. & Luque, F. J. (1993). Self-consistent reaction ®eld computation of the reactive characteristics of DNA bases in water. Biopolymers, 33, 1851± 1869. Orozco, M., Colominas, C. & Luque, F. J. (1996). Theoretical determination of the solvation free energy in water and chloroform of the nucleic acid bases. Chem. Phys. 209, 19 ± 29. Radzicka, A. & Wolfenden, R. (1988). Comparing the polarities of the amino acids: side-chain distribution coef®cients between the vapor phase, cyclohexane, 1-octanol, and neutral aqueous solution. Biochemistry, 27, 1664± 1670. Reichardt, C. (1988). Solvents and solvent effects. In Organic Chemistry, 2nd edit., pp. 339± 406, VCH Publishers, Weinheim. Shugar, D. & Fox, J. J. (1952). Spectrophotometric studies of nucleic acid derivatives and related compounds as a function of pH. Biochim. Biophys. Acta, 9, 199± 218. Tsai, R.-S., Fan, W., El Tayer, N., Carrupt, P.-A., Testa, B. & Kier, L. B. (1993). Solute-water interactions in the organic phase of a biphasic system. I. Structural in¯uence of organic solutes on the ``water-dragging'' effect. J. Am. Chem. Soc. 115, 9632±9639. Williams, R. & Wolfenden, R. (1985). Solvent water and the biological group-transfer potential of phosphoric and carboxylic anhydrides. J. Am. Chem. Soc. 107, 4345± 4346. Wolfenden, R. (1967). The free energy of hydrolysis of adenosine to inosine and ammonia. J. Biol. Chem. 242, 4711± 4714. Wolfenden, R. (1983). Waterlogged molecules. Science, 222, 1087± 1095. Wolfenden, R. & Lewis, C. A., Jr (1976). A vapor phase analysis of the hydrophobic effect. J. Theoret. Biol. 59, 231± 235. Wolfenden, R. & Radzicka, A. (1994). On the probability of ®nding a water molecule in a nonpolar cavity. Science, 265, 936± 937. Wolfenden, R., Liang, Y.-L., Matthews, M. & Williams, R. (1987). Cooperativity and anticoooperativity in solvation by water: imidazoles, quinones, nitrophenols, nitrophenolate and nitrothiophenolate ions. J. Am. Chem. Soc. 109, 4632±4660.
430
Hydrophobicities of Nucleic Acid Bases
Young, P. E. & Hillier, I. H. (1993). Hydration free energies of nucleic acid bases using an ab initio continuum model. Chem. Phys. Letters, 215, 405± 408. Young, P. E., Hillier, I. H. & Gould, I. R. (1994). Accurate prediction of the solvation of nucleotide base pairs using an ab initio continuum model. J. Chem. Soc. Perkin Trans. 2, 1717±1718.
http://www.hbuk.co.uk/jmb
Edited by I. Tinoco (Received 9 February 1998; received in revised form 14 April 1998; accepted 15 April 1998)
Supplementary material comprising one Table and references is available from JMB Online.
J. Mol. Biol. (1998) 280, 421±430
Hydrophobicities of the Nucleic Acid Bases: Distribution Coefficients from Water to Cyclohexane Phoebe Shih1, Lee G. Pedersen2,3, Paul R. Gibbs1 and Richard Wolfenden1* 1
Department of Biochemistry and Biophysics and 2Department of Chemistry, University of North Carolina, Chapel Hill NC, 27599-7260, USA 3 Laboratory of Quantitative and Computational Biology, National Institute of Environmental Health Sciences, Research Triangle Park, NC, 27709, USA
To establish an experimental scale of hydrophobicities for the nucleic acid bases, comparable with a scale developed earlier for amino acid side-chains, these bases and their parent compounds (purine and pyrimidin-2-one) were converted to n-butylated and tetrahydrofurylated derivatives that are appreciably soluble in cyclohexane, a truly non-polar solvent that dissolves negligible water at saturation. Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, showed no evidence of self-association of the solute in either solvent, and the possibility of speci®c entrainment of water by solutes entering cyclohexane was ruled out by the results of experiments with tritiated water. In both the n-butyl and tetrahydrofuryl series, the bases span a range of 5.3 kcal molÿ1 in their free energies of transfer from water to cyclohexane, and are arranged in the following rank, in order of decreasing hydrophobicity: purine > thymine > adenine > uracil > pyrimidin-2-one > hypoxanthine 5 cytosine 5 guanine. In both series of pyrimidin-2-ones, hydrophobicity decreases with introduction of an amino substituent, but addition of an exocyclic keto group results in a modest enhancement of hydrophobicity; and free energies of transfer are relatively insensitive to the position of N-alkyl substitution. In both series of purines, hydrophobicity decreases with the introduction of exocyclic amino and keto groups, the keto group having the greater effect; and free energies of transfer vary substantially depending on the position of N-alkyl substitution. Several additional compounds were examined to test recent predictions based on SM5.4/A, a quantum mechanical self-consistent-®eld solvation model; and that model was found to yield values in reasonable agreement with the experimental results. # 1998 Academic Press
*Corresponding author
Keywords: nucleic acid bases; hydrophobicity; cyclohexane; partition coef®cients; free energy of transfer
Introduction The major nucleic acid bases participate in many kinds of interactions that involve structural complementarity, including base-pairing during transcription and translation of genetic information; Abbreviations used: THF, tetrahydrofuryl; Bu, butyl; Me, methyl; HPLC, high pressure liquid chromatography; A, adenine; C, cytosine; G, guanine; H, hypoxanthine; P, purine; T, thymine; U, uracil; Z, pyrimidin-2-one; chx, cyclohexane; chf, chloroform; Gtr, free energy of transfer; Kdist, distribution coef®cient; r, correlation coef®cient; FTIR, Fourier transform infrared spectroscopy. 0022±2836/98/280421±10 $30.00/0
enzyme-RNA recognition (particularly in class I aminoacyl-tRNA synthetases; Cusack, 1995); and binding of substrates, coenzymes and effectors by many enzymes. Non-covalent binding interactions of the nucleic acid bases, like those of other biological molecules, involve the stripping away of solvent water from regions of contact between the binding partners. Accordingly, the observed strength of their interactions with other molecules can be considered to include the cost of removing the interacting molecules (at least those parts that make contact with each other) from the solvent to which they were previously exposed (Wolfenden, 1983). In attempts to detect the presence of speci®c attractive or repulsive interactions between the # 1998 Academic Press
422 nucleic acid bases and the sites at which they are bound, it would be desirable to have information about the relative tendencies of the nucleic acid bases to leave water and enter a ``naive'' cavity of low dielectric constant, that does not furnish opportunities for H-bonding or other speci®c attractive interactions. Free energies of solvation of biological molecules also play a signi®cant role in determining the positions of metabolic equilibria in neutral solution. Thus, for example, the difference in free energy of solvation between reactants and products is so large that it fully explains the favorable equilibrium of hydrolysis of ATP to ADP and inorganic phosphate (Williams & Wolfenden, 1985). Equilibria of interconversion of the nucleic acid bases, such as those for the hydrolytic deamination of adenosine (Wolfenden, 1967) and cytidine (Cohen & Wolfenden, 1971), depend in part on the overall strengths of the interactions of reactants and products with watery surroundings, in addition to the relative strengths of covalent bonds that are present in the reactants and products. The hydrophobicities of the nucleic bases are not easy to test experimentally. When equipped with their normal ribose substituents, these bases are so polar that they leave water and enter non-polar solvents at concentrations that are far below the limits of detection by present methods of analysis. However, free energies of solvation have been found, in most cases, to behave as additive functions of the substituent groups that are present in solutes (Leo et al., 1971; Wolfenden et al., 1987). The problem of detection could thus be circumvented, at least in principle, by replacing ribose with less polar substituents, to perform distribution measurements at concentrations suf®cient to allow the bases to be compared with respect to their hydrophobicities. When methyl groups replace substituent ribose, for example, the bases have been shown to leave water and enter chloroform at measurable concentrations (Cullis & Wolfenden, 1981). Unfortunately, chloroform represents an unsatisfactory compromise, because the polarity of chloroform is considerable, and chloroform is known to form H-bonds of signi®cant strength (for a review, see Reichardt, 1988). That dif®culty is compounded by the fact that chloroform absorbs 0.1 M water at saturation, a concentration that is far in excess of the concentrations of N-methylated bases that enter chloroform in distribution experiments. For that reason, ``wet'' chloroform is considerably more polar than pure chloroform, and substantial numbers of water molecules are available in wet chloroform that might form H-bonds with dissolved solutes. Moreover, polar solutes have been shown to ``drag'' extra water molecules into semipolar solvents of moderate polarity in distribution experiments (Tsai et al., 1993). ``Water-dragging'', when that occurs, leads to overestimation of the true hydrophobicity of a solute, and is dif®cult to test experimentally in the case of chloroform, since dissolved water greatly
Hydrophobicities of Nucleic Acid Bases
exceeds the concentration of solutes entering chloroform; i.e. the concentration of ``excess'' water dragged into the chloroform represents a small difference between large numbers, and is consequently dif®cult to determine. For those reasons, it is hardly surprising that in earlier work, some discrepancies were observed between distribution coef®cients observed for methylated derivatives of the bases, between water and 2-butanol or chloroform, resulting in two provisional scales of hydrophobicity. The bases were found to behave somewhat differently in the two solvent systems, in both the range of free energies observed and the ranking of the bases in terms of increasing hydrophobicity (Cullis & Wolfenden, 1981). Similar dif®culties were encountered earlier in experiments involving amino acid side-chains (Radzicka & Wolfenden, 1988), and were ascribed to the considerable H-bonding potential of 2-butanol and chloroform, and the fact that both solvents dissolve substantial concentrations of water at saturation, clouding the interpretation of distribution measurements as a simple index of hydrophobicity. In view of these uncertainties, it seemed desirable to re-examine these questions by obtaining experimental values for the hydrophobicities of the bases, measured with respect to a non-polar condensed phase that contains little water at equilibrium. With a water content at saturation equivalent to that of the vapor phase (Wolfenden & Radzicka, 1994), cyclohexane represents a desirable reference phase in that respect. By replacing substituent ribose with n-butyl and tetrahydrofuryl groups (Figure 1), we have now obtained derivatives of the bases that are capable of entering cyclohexane at measurable concentrations, and to measure the relative distribution coef®cients of nucleic acid bases between neutral aqueous solution and cyclohexane. Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, show no evidence of self-association in either solvent, and experiments with tritiated water eliminate the possibility of signi®cant water-dragging, in this system. In addition to the bases of the genetic material, we
Figure 1. Structures of ribose and the R group (methyl, butyl or tetrahydrofuryl) in nucleic acid base derivatives used in the present study. The site of substitution is N1 for derivatives of pyrimidin-2-one and N9 for derivatives of purine.
423
Hydrophobicities of Nucleic Acid Bases
have examined the behavior of hypoxanthine, and of the parent compounds purine and pyrimidin-2one, in order to evaluate the apparent effects of ring substituents on hydrophobicity. Since the earlier experimental work of Cullis & Wolfenden (1981), molecular dynamics simulation methods have been used to estimate free energies of transfer of the nucleic acid bases or their methylated derivatives from water to the vapor phase (Bash et al., 1987; Cramer & Truhlar, 1992; Mohan et al., 1992; Orozco & Luque, 1993; Elcock & Richards, 1993; Young & Hillier, 1993; Gao, 1994; Miller & Kollman, 1996). Free energies of transfer from water to chloroform have also been estimated by combining those values with computed free energies of transfer from vapor to chloroform (Orozco et al., 1996; Young et al., 1994; Eksterowicz et al., 1997). In addition, Giesen et al. (1997) have predicted partition coef®cients from water to chloroform of several unnatural nucleic acid bases for which experimental data were not available. To test those predictions, we also determined waterto-chloroform partition coef®cients for methylated derivatives of three unnatural bases; 2-amino-9methylpurine, 6-methylamino-9-methylpurine, and 5-bromo-1-methyluracil.
Results Distribution measurements between neutral aqueous solution and cyclohexane, at varying solute concentrations, showed no evidence of self-association in either solvent; and the possibility of speci®c entrainment of water by solutes entering cyclohexane was ruled out by the results of experiments with tritiated water. Table 1 shows distribution coef®cients of N1-substituted pyrimidines and N9-substituted purine derivatives, between
Table 1. Hydrophobicity of base derivatives Gtr(w!chx) (kcal molÿ1)a Nucleic acid bases Purine Thymine Adenine Uracil Pyrimidin-2-one Hypoxanthine Cytosine Guanine
Butyl derivative 1.05 2.18 2.82 3.65 4.12 5.67 6.03* 6.67*
Tetrahydrofuryl derivative 2.56 4.08 4.25 5.02 5.55 7.14* 7.27* 7.48*
Dipole momentb 3.67 4.56 2.46 4.66 6.49 5.56 7.04 6.75
a Gtr ÿ RT ln K. Distribution coef®cients, K, were determined by the double extraction procedure as described in Materials and Methods. Standard errors on four determinations for each value are 4 0.2 kcal molÿ1. The values with an asterisk (*) have greater experimental uncertainty (4 0.8 kcal molÿ1) because the measured absorbance values are at or near the detection limit. b Computed from the ab initio wave function; units are debyes.
water and cyclohexane at 25 C. The distribution coef®cients of these butylated and tetrahydrofurylated bases were found to vary over a range of four orders of magnitude. Alkylation mixtures were usually found to yield both singly and multiply butylated products, even when only one equivalent of alkylating agent was present, but these products were easily separated by HPLC. Butylation of purine, for example, yields 3-, 7- and 9-n-butylpurines. Water-to-cyclohexane free energies of transfer for this series of butylpurines as well as for N1 and N3-butylated thymine and uracil are included in the ®rst section of Table 2. Several less common derivatives obtained from butylation reaction mixtures included 2-O-butylated cytosine, adenine butylated at the exocyclic amino group, adenine butylated at both N9 and the exocylic amino group, cytosine butylated at both N1 and the exocyclic amino group, and three dibutylated hypoxanthines (1,7; 1,9; 3,7). Their Gtr(w ! chx) values are listed in the second section of Table 2. To explore the effects of the position of alkylation, distribution coef®cients between water and chloroform (Gtr(w ! chx)) were determined for several purines and pyrimidines, and these values are listed in the third section of Table 2. In addition, dipole moments for purine and adenine methylated at N1, N3, N7 and N9, computed from the ab initio wave function, and the experimental and theoretical (Giesen et al., 1997) log K(Cchf/Cw) are included in Table 2.
Discussion Ordering of Gtr(w!chx) for butylated and tetrahydrofurylated bases, compared with earlier findings The present work establishes the hydrophobicities of the nucleic acid bases, determined with respect to a truly non-polar solvent that dissolves negligible water at saturation. In both the n-butyl and tetrahydrofuryl series, the results yield the following ranking in order of decreasing hydrophobicity: purine > thymine > adenine > uracil > pyrimidin-2-one > hypoxanthine 5 cytosine 5 guanine. In Figure 2, the present free energies of transfer for butylated derivatives from water to cyclohexane are used as a basis for plotting the values observed for transfer of tetrahydrofuryl derivatives to water (present data), and the values obtained earlier for transfer of methylated derivatives from water to chloroform (Cullis & Wolfenden, 1981). The similarity of these three scales plotted in Figure 2 indicates that the relative hydrophobicities of nucleic acid bases are little affected by the nature of the substituent, whether that be ribose, methyl, n-butyl or tetrahydrofuryl. Accordingly, these data furnish a self-consistent experimental index of the relative hydrophobicities of the bases themselves.
424
Hydrophobicities of Nucleic Acid Bases Table 2. Free energies of transfer, partition coef®cients, and dipole moments of base derivatives with single or multiple alkyl substitution Base derivatives A. To cyclohexane 9-n-Butylpurine 7n-Butylpurine 3-n-Butylpurine 1-n-Butylpurine 1-n-Butylthymine 3-n-Butylthymine 1-n-Butyluracil 3-n-Butyluracil B. To cyclohexane 2-n-Butoxycytosine 6-n-Butylaminopurine 6-n-Butylamino 9-n-Butylpurine 1-n-Bu- 4-n-butylamino pyrimidin-2-one 1,7-Di-n-Bu-hypoxanthine 1,9-Di-n-Bu-hypoxanthine 3,7-Di-n-Bu-hypoxanthine C. To chloroform 9-Methyladenine* 7-Methyladenine 3-Methyladenine 1-Methyladenine 9-Me-2-aminopurine 6,6-Di-Me-aminopurine 9-Me-6-methylaminopurine 5-Bromo-1-methyluracil 1-Methylthymine* 1-Methyluracil* 3-Methyluracil Thymine 6-Methyluracil Uracil Pyrimindin-2-one 1,3-Dimethyluracil 9-Methylpurine 6-Methylpurine 6-Chloro-purine Purine
Gtr(w!org)a (kcal molÿ1)
logK(Corg/Cw)b Experiment Calculated
1.05 3.55 2.67
ÿ0.77 ÿ2.60 ÿ1.96
2.18 1.84 3.65 3.65
ÿ1.60 ÿ1.35 ÿ2.68 ÿ2.68
1.53 4.08
ÿ1.12 ÿ2.99
ÿ1.36
1.00
ÿ0.95 0.80 1.65 3.00
0.70 ÿ0.59 ÿ1.21 ÿ2.20
1.06
ÿ0.78
2.55 4.12 0.69 0.27 ÿ0.47 0.89 0.61 1.65 1.72 3.09 3.36 4.08 4.22 ÿ0.89 ÿ0.16 2.28 1.86 2.82
ÿ1.82 ÿ3.28 ÿ0.50 ÿ0.20 0.35 ÿ0.66 ÿ0.45 ÿ1.21 ÿ1.26 ÿ2.26 ÿ2.47 ÿ2.99 ÿ3.10 0.65 0.11 ÿ1.67 ÿ1.37 ÿ2.07
Dipole momentc 4.13 6.50 4.72 7.89
ÿ1.6
ÿ1.9
2.71 7.31 4.56 9.16
ÿ0.3 ÿ0.3 ÿ0.3 ÿ1.2
a
Gtransfer ÿ RT ln K. Distribution coef®cients, K, were determined from double extraction procedure as described in Materials and Methods. Cyclohexane was used as the organic layer for butylated and tetrahydrofurylated derivatives and chloroform was used as the organic layer for methylated derivatives. For 1-methyl adenine (pKa 7.2; Brookes & Lawley, 1960), 50 mM Tris-HCl (pH 9) was used as the aqueous phase. Standard errors on four determinations for each value are 4 0.2 kcal molÿ1. Base derivatives with * denote those whose values had been determined previously (Cullis & Wolfenden, 1981). Calculated K values were obtained from Giesen et al. (1997). c Computed from the ab initio wave function; units are debyes. b
Not surprisingly, substantial discrepancies are observed between these values and those determined earlier for the methylated bases using 2-butanol as a reference phase, as to both the ranking of the bases and the span of free energies observed (Cullis & Wolfenden, 1981). That discrepancy can probably be attributed to formation of H-bonds to solvent 2-butanol, a phenomenon encountered earlier in the distribution of amino acid side-chains between water and 1-octanol (Radzicka & Wolfenden, 1988). An additional complicating factor, tending to invalidate alcohols as simple reference phases for the measurement of hydrophobicity, is their tendency to dissolve considerable
concentrations of water (9.4 M in the case of 2-butanol) at saturation (Radzicka & Wolfenden, 1994). Although wet chloroform is moderately polar and dissolves 0.1 M of water at saturation (Leo et al., 1971), this solvent appears to offer a fair compromise as a semi-polar reference solvent, in that a good correlation is observed between values for Gtr(w ! chf) and Gtr(w ! chx) (Figure 2). This correspondence between the two scales is probably fortuitous to some extent, and cannot be expected to apply in every case that might be examined. In that context, it may be worth noting that the 13C chemical shifts of C2, C4, C5, C6 and C8 of 9-nbutylpurine in deuterated chloroform are slightly
425
Hydrophobicities of Nucleic Acid Bases
Figure 2. Gtr(w ! chx) values of butylated base derivatives are plotted against Gtr(w ! chx) values for tetrahydrofurylated derivatives (slope 0.89 0.04, r 0.99) and Gtr(w ! chf) values for methylated derivatives (slope 0.93 0.05, r 0.99).
closer to those in deuterated cyclohexane than to those in deuterated water (unpublished results). Substituent effects Amino and keto functionality on purine and pyrimidin-2-one Gtr(W ! chx) values for butylated purines and pyrimidines are plotted in Figure 3 alongside their structures. Considered as derivatives of the parent compounds purine and pyrimidin-2-one, purines and pyrimidin-2-ones differ in their responses to the introduction of ring substituents. Introduction of an amino or a keto functionality to the purine system decreases hydrophobicity, the keto group having the greater effect. Addition of an amino group to the hypoxanthine ring, already containing a keto group, does not change the hydrophobicity much further. Thus, the effects of substitution fall short of expectations based on any simple additivity relationship. Introduction of an amino substituent to pyrimidin-2-ones decreases hydrophobicity, to an extent similar to that observed in the purines; but addition of a keto group results in a small increase in hydrophobicity. Accordingly, keto substituents exert different effects on water af®nity in the two ring systems. Earlier experimental studies have shown that neighboring atoms in a molecule sometimes interact in ways that appear to reinforce or hinder the polarizing effect of solvating water molecules (Hine &
Figure 3. Structures of purines and pyrimidines are shown alongside the plot of Gtr(w ! chx) for butylated purines and pyrimidines (see Table 1 for experimental details). In nucleic acid bases, the R group is hydrogen and in nucleosides, the R group is ribose. For the base derivatives used in the present study, the R group is a methyl, butyl or tetrahydrofuryl substituent.
Mookerjee, 1975; Wolfenden et al., 1987), as in the present case. Using a fragment-based partitioning model, Giesen et al. (1997) have examined the degree to which adjacent functional groups may interact in a non-linear fashion to in¯uence free energies of distribution. Their analyses indicate that the transferability of an apparent group contribution to the overall partition coef®cient, from one molecule to another, depends strongly on the extent to which the functional group is isolated from other interactions. Halogenated purine and pyrimidine derivatives Bromo substitution at the C5 position of 1-methyluracil increases its tendency to enter the chloroform phase by 0.8 kcal molÿ1, and chloro substitution at C6 of purine enhances its af®nity for the chloroform phase by nearly 1 kcal molÿ1 (Table 2). Although not strictly comparable, these enhancements of apparent hydrophobicity resemble those observed for methyl substitution in
426 both the purine and the pyrimidine systems, as may be seen by comparing Gtr values for 1-methylthymine and 6-methylpurine. Effects of C, N and O-alkylation and multiple alkylation An interesting minor product, obtained by butylation of cytosine, was identi®ed as 2-nbutoxycytosine, by X-ray crystallography, 1H NMR, UV, and FTIR. In terms of Kdist(w ! chx), this compound is more than three orders of magnitude less hydrophobic than is 1-n-butyl cytosine. This effect is similar in magnitude to the effect observed (102.5) when 2-methoxypyridine is compared with N-methyl pyridone, using hexane as the non-polar phase (Cullis & Wolfenden, 1981). In contrast, the hydrophobicity of 6-n-butyladenine is only 8.4-fold lower than that of 9-n-butyladenine. When both N9 and the exocyclic amino group of adenine are butylated, the product actually favors the cyclohexane layer with Kdist 10. Similarly, when both N1 and the exocylic group of cytosine are alkylated, the product also favors the cyclohexane layer, with Kdist 5. Values are given for several dibutylated hypoxanthines in the second section of Table 2. In both purines and pyrimidines, N-methylation is found to produce a much larger enhancement in hydrophobicity than does C-methylation. Thus, N1 and N3 methylation of uracil yielded a more favorable Gtr(w ! chf) by 2.4 kcal molÿ1; whereas methylation at C5 (thymine) and C6 changed Gtr(w ! chf) only by 1 and 0.7 kcal molÿ1, respectively (third section of Table 2). In the purine system, methylation at N9 and C6 resulted in Gtr(w ! chf) of 3 and 0.5 kcal molÿ1, respectively. The difference between the effects of N-methylation and C-methylation seems understandable in terms of the ability of ring N-H groups to form Hbonds to solvent water, and the likelihood that N-methylation results in a major disruption of these interactions. In contrast, the introduction of a methylene increment, by methylation of a carbon atom, has typically been found to enhance hydrophobicity (expressed in terms of Kdist(water ! hydrocarbon) by 0.96 kcal molÿ1, comparable in magnitude with the modest effects observed here (Wolfenden & Lewis, 1976). Effect of the site of alkylation on hydrophobicity Whereas N1 and N3-n-butyluracil have identical partition coef®cients, the hydrophobicity of N7-nbutylpurine differs from that of N9-n-butyl purine by nearly two orders of magnitude. N1, N3 and N9-methyl adenine also varied over a range of more than 300-fold in hydrophobicity. These differences are perhaps not surprising, in view of the likelihood that the different nitrogen atoms, at which substitution occurs, differ from each other in
Hydrophobicities of Nucleic Acid Bases
the strengths and directional preferences of their interactions with solvent water. Relationships between Gtr(w!chx) and dipole moment The dipole moments of purines and pyrimidines, calculated from the ab initio wave function, are included in Table 1 for the underivatized bases, and in Table 2 for methylated bases. For the bases considered as a whole, there is a moderately close relationship between dipole moment and values for Gtr(w ! chx), with r 0.79. Factors other than H-bonding capacity and dipole moment, such as inductive effects and steric crowding, may also contribute to the relative hydrophobicities of the nucleic acid bases and their derivatives. In the pyrimidine series, the low hydrophobicity of cytosine may be related to its high dipole moment, whereas the most hydrophobic of the series, thymine, has the lowest dipole moment. Pyrimidin-2-one and uracil exhibit similar hydrophobicity, perhaps as a result of compensation between H-bonding capacity and dipole moment. In the purine series, the relatively high dipole moment and H-bonding capacity of guanine and hypoxanthine presumably account for their low hydrophobicity. The parent compound purine has a low dipole moment (3.67 debyes) and is least capable of forming H-bonds with solvent water, explaining its high hydrophobicity. Although adenine has an even lower dipole moment (2.46 debyes) than purine, this appears to be more than fully compensated by its ability to form two additional H-bonds to solvent water. The calculated dipole moments of the N-methylated purines and adenine (Table 2) may be helpful in understanding their variations in hydrophobicity. In both series, the H-bonding capacity is equivalent within a given series, rendering the comparison between dipole moment and hydrophobicity relatively straightforward. Plots of Gtr(w ! chx) for butylated purines and of Gtr(w ! chf) for methylated adenine against calculated dipole moments gave r 0.90 and 0.97, respectively. Comparison with theoretical results A recent computational study (Giesen et al., 1997) used the continuum solvation model SM5.4/ A, yielded estimated log K(Cchf/Cw) that were chosen for comparison with our experimental log K(Cchf/Cw) (Table 2) for two reasons. First, this method yielded values for methylated forms of the six common bases (1-Me-T, 1-Me-U, 1-Me-C, 9-MeA, 9-Me-G and 9-Me-H) that were in closer agreement with earlier experimental results (Cullis & Wolfenden, 1981) than were other calculated values of log K(Cchf/Cw) (see Eksterowicz et al. (1997) for a comparison of experimental results with other theoretical results). Second, Giesen et al. (1997) predicted log K(Cchf/Cw) for several methyl-
427
Hydrophobicities of Nucleic Acid Bases
ine. In both systems, halogenation was found to enhance the hydrophobicity of the parent compounds, and the effect was similar to that effect of methylation.
Materials and Methods Chemicals Purine and pyrimidine bases, N1, N3 and N9-methylated adenine; 2-aminopurine, 6,6-dimethylaminopurine, 6-methylaminopurine, 2,6-diaminopurine, 5-bromo-1methyluracil, 6-chloropurine, 6-methylpurine, 3-methyluracil, 6-methyluracil were obtained from either Sigma Chemical Co. or Aldrich Chemical Co. 2-t-Butoxytetrahydrofuran was purchased from Lancaster Chemical Co. Tetrabutylammonium hydroxide and tetramethylammonium hydroxide were obtained from Aldrich Chemical Co. Figure 4. Log K(Cchf/Cw) values, estimated by the contiuum solvation model SM5.4/A (Giesen et al., 1997) are plotted separately against log K(Cchf/Cw) values for the six natural bases determined previously (Cullis & Wolfenden (1981); ®lled circle and broken line, slope 1.30 0.18, r 0.96) and for all nine bases (including the three new derivatives, open circle and ®t by the thick continuous line, slope 1.15 0.18, r 0.93). Experimental log K(Cchf/Cw) values are rounded off to the same signi®cant digit as the calculated values.
Distribution coefficients determination Distribution coef®cients between water (w) and cyclohexane (chx) or chloroform were measured by a doubleextraction procedure (Bone et al., 1983), ®rst extracting the compound from water into cyclohexane at 25 C, separating the phases, then back-extracting from cyclohexane into a fresh aqueous phase. At both stages, the two phases were stirred vigorously for at least one hour at 25 C. Apparent distribution coef®cients were calculated by the following equation: KM!L
ated unnatural nucleic acid bases for which experimental data had not been available, offering an opportunity for a direct experimental test. Experimental and calculated log K(Cchf/Cw) values for 2-amino-9-methylpurine, 6-methylamino-9-methylpurine and 5-bromo-1-methyl-uracil are compared in the third section of Table 2. In Figure 4, calculated log K(Cchf/Cw) values are plotted separately against log K(Cchf/Cw) values for the six natural bases determined previously (broken line) and for all nine bases (thick continuous line), yielding r 0.96, and 0.93, respectively. Inclusion of the three new data points changed the slope of the correlation line slightly, from 1.30 0.18 to 1.15 0.18. However, the near-unit slope of the correlation line indicates that the span and absolute magnitude of the predicted values is approximately correct. Of particular interest is the calculated log K(Cchf/Cw) value of one of the new base derivatives, 5-bromo-1-methyluracil, which was suggested as implying that ``bromine is net hydrophobic in (the C-5) position to about the same extent as a methyl group'' (Giesen et al., 1997). As a bromo group is ordinarily considered to be more electronegative than a methyl group, its tendency to enhance hydrophobicity to the same extent as a methyl group was not, to us, intuitively obvious. That prediction led us to examine the effect of bromo substitution in 5-bromo-1-methyl-uracil, and the effect of chloro substitution in 6-chloropur-
BM LA ÿ LB
where KM ! L is the equilibrium constant for transfer to the less favored solvent, M is the volume of the more favored solvent used in both extractions, B is the concentration of solute in more favored solvent after second or back extraction, L is the volume of the less favored solvent, and A is the concentration of solute in the more favored solvent after ®rst extraction. Whenever possible (solutes with KM ! L > 8 10ÿ4), experiments were performed at initial solute concentrations varying over a range of at least 20-fold (typically 0.3 to 6 mM). Distribution coef®cients were found to be invariant with changing solute concentration, consistent with the view that these values would apply also at in®nite dilution. Furthermore, ultraviolet spectra were taken for solutes that dissolve in signi®cant concentration in cyclohexane, and the lmax of every spectrum examined closely resembles that in water. Therefore, the same tautomeric species are present in both phases. The following lmax values (in nm) were used in the distribution experiments: 9-n-BuA, 261; 6-n-butylaminoP, 269; 6-n-butylamino-9-Bu-P, 270; 1-n-Bu-C, 275; 2-n-butoxy-C, 270; 1-n-butylamino-4n-Bu-Z, 274; 9-n-Bu-G, 253; 9-n-Bu-H, 251; 1,7-di-n-Bu-H, 257; 1,9-di-n-Bu-H, 253; 3,7-di-n-Bu-H, 266; 3-n-Bu-P, 276, 7-n-Bu-P, 266, 9-n-Bu-P, 264; 1-n-Bu-Z, 304; 1-n-BuT, 273; 3-n-Bu-T, 265; 1-n-Bu-U, 268; 3-n-Bu-U, 259; 9-(2THF)-A, 262; 1-(2-THF)-C, 272; 9-(2-THF)-G, 253; 9-(2THF)-H, 250; 9-(2-THF)-P, 263; 1-(2-THF)-Z, 302; 1-(2THF)-T, 269; 1-(2-THF)-U, 264; 1-Me-A, 270; 3-Me-A, 273; 9-Me-A, 262; 6-Cl-P, 265; 6-Me-P, 261; P, 263; 9-MeP, 264; 6-6-dimethylamino-P, 276; 6-methylamino-9-MeP, 268; 2-amino-9-Me-P, 304; 3-Me-U, 258; 6-Me-U, 261; !Me-U, 267; 1,3-di-Me-U, 266; 5-Br-!-Me-U, 283; 1-Me-T, 272; T, 265; Z, 299; U, 259. Distribution coef®cient for each compound was determined based on four or more
428 experiments. Standard errors for each Gtr(w ! chx) value are 4 0.2 kcal molÿ1 for compounds with Gtr(W ! chx) 46 kcal molÿ1 (or KM ! L 5 4 10ÿ5). Gtr(w ! chx) values that are 56 kcal molÿ1 (for compounds with KM ! L 4 4 10ÿ5) are subjected to greater uncertainty (4 0.8 kcal molÿ1) because the measured absorbance values are at or near the detection limit. Water entrainment Experiments with tritiated water showed that solutes entering cyclohexane (9-n-butylpurine, for example, at a concentration of 4 10ÿ2 M in the cyclohexane layer) did not enhance the concentration of water entering cyclohexane above the concentration (9 10ÿ4 M) that enters cyclohexane in the absence of solutes. Thus, water entrainment appeared to be negligible in these experiments. Synthesis of butylated and methylated purine and pyrimidine derivatives Purine and pyrimidine bases were butylated using tetrabutylammonium hydroxide (Meyers & Zeleznick, 1963). A solution of the base (5 mmol) and tetrabutylammonium hydroxide (5 mmol) in water (10 ml) was lyophilized. The hygroscopic residue was heated slowly in a sublimation apparatus under vacuum to a temperature ranging from 160 to 250 C. Heating was continued for three hours at the temperature (160 to 250 C) when a sublimate began to collect on the condenser. The crude mixture collected in the sublimator typically consisted of starting materials and singly and multiply butylated bases. These products were resolved by reverse-phase HPLC, using a preparative Whatman Partisil 10 ODS-2 column (22 mm 500 mm). The starting material was eluted with 100% water, followed by elution of the butylated base bases with a series of 5% steps in the concentration of methanol. Products were identi®ed by their UV and proton nuclear magnetic resonance spectra. The approximate yields and the concentrations of methanol required for elution are as follows: 9-n-Bu-A, HPLC, 35% methanol; yield, 40%; 6-n-butylamino-P, HPLC, 40% methanol; yield, 40%; 6-n-butylamino-9-n-Bu-P, HPLC, 80% methanol; yield, 25%; 1-n-Bu-C, HPLC, 25% methanol; yield, 30%; 2-n-butoxy-C, HPLC, 60% methanol; yield, 35%; 2-Hydroxy-1-n-Bu-4-butylaminopyrimidine (1,4-Di-n-butylcytosine), HPLC, 80% methanol; yield, 20%; 9-n-Bu-G, HPLC, 20% methanol; yield 25%; 9-n-BuH, HPLC, 20% methanol; yield, 30%; 1,7-di-n-Bu-H, HPLC, 75% methanol; yield, 15%; 1,9-di-n-Bu-H, HPLC, 60% methanol; yield, 10%; 3,7-di-n-Bu-H, HPLC, 40% methanol; yield, 10%; 3-n-Bu-P, HPLC, 45% methanol; yield, 15%; 7-n-Bu-P, HPLC, 30% methanol; yield, 65%; 1-n-Bu-T, HPLC, 35% methanol; yield, 50%; 3-n-Bu-T, HPLC, 40% methanol; yield, 35%; 7-n-Bu-P, HPLC, 40% methanol; yield 25%; 9-n-Bu-P, HPLC, 50% methanol; yield, 30%; 1-n-Bu-Z, 1-n-Bu-U, HPLC, 25% methanol; yield, 35%; 3-n-Bu-U, HPLC, 30% methanol; yield, 40%. 2-Aminopurine, 6-methylaminopurine, purine and uracil were methylated using tetramethylammonium hydroxide. The synthetic and puri®cation procedures as described above for the butylation reaction were employed. Products were identi®ed by their UV and proton nuclear magnetic resonance spectra. The approximate yields and the concentrations of methanol required for elution are as follows: 2-amino-9-Me-P, HPLC, 30% methanol; yield, 40%; 6-methylamino-9-Me-P, HPLC,
Hydrophobicities of Nucleic Acid Bases 50% methanol; yield, 40%; 9-Me-P, HPLC, 50% methanol; yield, 20%; 1,3-di-Me-U, HPLC, 75% methanol; yield, 30%. Identification of 2-n-butoxycytosine One product isolated from the butylation of cytosine was shown by 1H NMR to be monobutylated, but its UV spectra in acidic, neutral, and basic conditions did not agree with those reported for N1, N3 or N4-alkylated cytosine. Instead, they were similar to those reported for 2-methoxy-4-amino-pyrimidine (Shugar & Fox, 1952). Since O-alkylation of bases by the present method is unusual, X-ray diffraction and FTIR were used to con®rm its identi®cation as 2-n-butoxycytosine. Crystals of this compound were grown in water at room temperature, and its structure was determined by Dr Peter White at the Single Crystal X-Ray Facility of the Chemistry Department of the University of North Carolina at Chapel Hill. Diffraction data for a monoclinic crystal of 0.2 mm 0.2 mm 0.2 mm dimensions were collected using a Rigaku diffractometer in the omega scan mode at 25 C. In all, 5403 re¯ections were measured, and the RF value for signi®cant re¯ections was 0.121. Structural data were consistent with butylation site at the O2 position. FTIR spectra were obtained from KBr pellets at 25 C at 2 cmÿ1 resolution using a Mattson First Galaxy series 5000 FTIR spectrometer. As expected, this compound lacked a prominent carbonyl absorption near 1625 cmÿ1 that is present in 1-n-butylcytosine and 3-methylcytosine. Synthesis of tetrahydrofuryl purine and pyrimidine derivatives Tetrahydrofuryl derivatives of the bases were prepared using 2-t-butoxytetrahydrofuran as the alkylating agent, as described by Kametani et al. (1977). A mixture of the base (8 mmol) and 2-t-butoxytetrahydrofuran (12 mmol) in dimethylformamide (5 ml) was heated with stirring at 170 C for ®ve hours. After evaporation of the solvent, the product was dissolved in 10 ml of chloroform, washed twice with water, and concentrated to 2 ml before chromatography on a Whatman Partisil 10 ODS-2 preparative HPLC column (22 mm 500 mm). The major products obtained are 1-substituted pyrimidines and 9-substituted purines, although other products were sometimes obtained. The starting material was eluted with 100% water, followed by elution of the tetrahydrofurylated bases with a series of 5% steps in the concentration of methanol. Products were identi®ed by their UV and proton nuclear magnetic resonance spectral data. The approximate yields and concentrations of methanol required for elution are as follows: 9-(2-THF)A, HPLC, 25% methanol; yield, 60%; 1-(2-THF)-C, HPLC, 10% methanol; yield, 55%; 9-(2-THF)-G, HPLC, 10% methanol; yield, 50%; 9-(2-THF)-H, HPLC, 10% methanol; yield, 45%; 9-(2-THF)-P, HPLC, 40% methanol; yield, 45%; 1-(2-THF)-Z, HPLC, 20% methanol; yield, 30%; 1-(2-THF)-T, HPLC, 30% methanol; yield, 35%; 1-(2THF)-U, HPLC, 25% methanol; yield, 55% Calculation of dipole moments Dipole moments for analogs of selected molecules indicated in Table 1 (for underivatized purine and pyrimidines) and Table 2 (for methylated derivatives) were estimated by use of the ab initio quantum mechanical
Hydrophobicities of Nucleic Acid Bases package Gaussian 94 (Frisch et al., 1995) on a Cray model T90 at the North Carolina Supercomputer Center. The initial geometries of the molecules were energy-optimized at the 6-311G (d,p) basis set level, which includes polarization orbitals on all atoms.
Acknowledgments This work was supported by research grants GM18325 to R.W. and HL-06350 to L.G.P. from the National Institutes of Health. P.S. acknowledges the support of a postdoctoral fellowship (PF-4156) from American Cancer Society.
References Bash, P. A., Singh, U. C., Langridge, R. & Kollman, P. A. (1987). Free energy calculations by computer simulation. Science, 236, 564± 568. Bone, R., Cullis, P. & Wolfenden, R. (1983). Solvent effects on equilibria of addition of nucleophiles to acetaldehyde and the hydrophilic character of diols. J. Am. Chem. Soc. 105, 1339± 1343. Brookes, P. & Lawley, P. D. (1960). The methylation of adenosine and adenylic acid. J. Chem. Soc. 539± 545. Cohen, R. M. & Wolfenden, R. (1971). The equilibrium of hydrolytic deamination of cytidine and N4methylcytidine. J. Biol. Chem. 246, 7566 ±7568. Cramer, C. J. & Truhlar, D. G. (1992). Polarization of the nucleic acid bases in aqueous solution. Chem. Phys. Letters, 198, 74 ±80. Cullis, P. M. & Wolfenden, R. (1981). Af®nities of nucleic acid bases for solvent water. Biochemistry, 20, 3024± 3028. Cusack, S. (1995). Eleven down and nine to go. Nature Struct. Biol. 2, 824± 831. Eksterowicz, J. E., Miller, J. L. & Kollman, P. A. (1997). Calculation of chloroform/water partition coef®cients for the N-methylated nucleic acid bases. J. Phys. Chem. ser. B, 101, 10971± 10975. Elcock, A. H. & Richards, W. G. (1993). Relative hydration free energies of nucleic acid bases. J. Am. Chem. Soc. 115, 7930± 7931. Frisch, M. J., Trucks, G. W., Schlegel, H. B., Gill, P. M. W., Johnson, B. G., Robb, M. A., Cheeseman, J. R., Keith, T., Petersson, G. A., Montgomery, J. A., Raghavachari, K., Al-Laham, M. A., Zakrzewski, V. G., Ortiz, J. V., Foresman, J. B., Cioslowski, J., Stefanov, B. B., Nanayakkara, A., Challacombe, M., Peng, C. Y., Ayala, P. Y., Chen, W., Wong, M. W., Andres, J. L., Replogle, E. S., Gomperts, R., Martin, R. L., Fox, D. J., Binkley, J. S., Defrees, D. J., Baker, J., Stewart, J. P., Head-Gordon, M., Gonzalez, C. & Pople, J. A. (1995). Gaussian 95, Gaussian, Inc., Pittsburgh PA. Gao, J. (1994). The hydration and solvent polarization effects of nucleotide bases. Biophys. Chem. 51, 253± 261. Giesen, D. J., Chambers, C. C., Cramer, C. J. & Truhlar, D. G. (1997). What controls partitioning of the nucleic acid bases between chloroform and water?
429 (erratum to be published). J. Phys. Chem. ser. B, 101, 5084± 5088. Hine, J. & Mookerjee, P. K. (1975). The intrinsic hydrophilic character of organic compounds. Correlations in terms of structural contributions. J. Org. Chem. 40, 292± 298. Kametani, T., Kigasawa, K., Hiiragi, M., Wakisaka, K., Kusama, O., Kawasaki, K. & Sugi, H. (1977). Studies on the synthesis of chemotherapeutics. I. Synthesis of 1-(2-tetrahydrofuryl)-5-¯uororuracil. J. Heterocyc. Chem. 14, 473± 475. Leo, A., Hantsch, C. & Elkins, D. (1971). Partition coef®cients and their uses. Chem. Rev. 71, 525± 592. Meyers, T. C. & Zeleznick, L. (1963). Alkylation of the purine nucleus by means of quaternary ammonium compounds. I. Tetraalkylammonium hydroxides. J. Org. Chem. 28, 2087± 2089. Miller, J. L. & Kollman, P. A. (1996). Solvation free energies of the nucleic acid bases. J. Phys. Chem. 100, 8587± 8594. Mohan, V., Davis, M. E., McCammon, J. A. & Pettitt, B. M. (1992). Continuum model calculations of solvation free energies: accurate evaluation of electrostatic contributions. J. Phys. Chem. 96, 6428± 6431. Orozco, M. & Luque, F. J. (1993). Self-consistent reaction ®eld computation of the reactive characteristics of DNA bases in water. Biopolymers, 33, 1851± 1869. Orozco, M., Colominas, C. & Luque, F. J. (1996). Theoretical determination of the solvation free energy in water and chloroform of the nucleic acid bases. Chem. Phys. 209, 19 ± 29. Radzicka, A. & Wolfenden, R. (1988). Comparing the polarities of the amino acids: side-chain distribution coef®cients between the vapor phase, cyclohexane, 1-octanol, and neutral aqueous solution. Biochemistry, 27, 1664± 1670. Reichardt, C. (1988). Solvents and solvent effects. In Organic Chemistry, 2nd edit., pp. 339± 406, VCH Publishers, Weinheim. Shugar, D. & Fox, J. J. (1952). Spectrophotometric studies of nucleic acid derivatives and related compounds as a function of pH. Biochim. Biophys. Acta, 9, 199± 218. Tsai, R.-S., Fan, W., El Tayer, N., Carrupt, P.-A., Testa, B. & Kier, L. B. (1993). Solute-water interactions in the organic phase of a biphasic system. I. Structural in¯uence of organic solutes on the ``water-dragging'' effect. J. Am. Chem. Soc. 115, 9632±9639. Williams, R. & Wolfenden, R. (1985). Solvent water and the biological group-transfer potential of phosphoric and carboxylic anhydrides. J. Am. Chem. Soc. 107, 4345± 4346. Wolfenden, R. (1967). The free energy of hydrolysis of adenosine to inosine and ammonia. J. Biol. Chem. 242, 4711± 4714. Wolfenden, R. (1983). Waterlogged molecules. Science, 222, 1087± 1095. Wolfenden, R. & Lewis, C. A., Jr (1976). A vapor phase analysis of the hydrophobic effect. J. Theoret. Biol. 59, 231± 235. Wolfenden, R. & Radzicka, A. (1994). On the probability of ®nding a water molecule in a nonpolar cavity. Science, 265, 936± 937. Wolfenden, R., Liang, Y.-L., Matthews, M. & Williams, R. (1987). Cooperativity and anticoooperativity in solvation by water: imidazoles, quinones, nitrophenols, nitrophenolate and nitrothiophenolate ions. J. Am. Chem. Soc. 109, 4632±4660.
430
Hydrophobicities of Nucleic Acid Bases
Young, P. E. & Hillier, I. H. (1993). Hydration free energies of nucleic acid bases using an ab initio continuum model. Chem. Phys. Letters, 215, 405± 408. Young, P. E., Hillier, I. H. & Gould, I. R. (1994). Accurate prediction of the solvation of nucleotide base pairs using an ab initio continuum model. J. Chem. Soc. Perkin Trans. 2, 1717±1718.
http://www.hbuk.co.uk/jmb
Edited by I. Tinoco (Received 9 February 1998; received in revised form 14 April 1998; accepted 15 April 1998)
Supplementary material comprising one Table and references is available from JMB Online.