Spermine-DNA interactions: A theoretical study

Transcription

1 Proc. Nad. Acad. Sci. USA Vol. 83, pp , August 1986 Biophysics Spermine-DNA interactions: A theoretical study (polyamines/dna conformation/energy minimiation/molecular mechanics/molecular modeling) BURT G. FEUERSTEIN*, NAGARAJAN PATTABIRAMANt, AND LAURENCE J. MARTON*t *Brain Tumor Research Center of the Department of Neurological Surgery, and the Departments of tpharmaceutical Chemistry and *Laboratory Medicine, Schools of Medicine and Pharmacy, University of California, San Francisco, CA Communicated by Rudi Schmid, February 24, 1986 ABSTRACT Models for the interaction of spermine and DNA were studied by performing conformational energy calculations on spermine and molecular mechanics calculations on major and minor groove complexes of spermine and oligomers of DNA. Docked into the major groove of B-DNA, spermine stabilies the complex by aimig interactions between proton acceptors on the oligomer and proton donors on spermine. This is achieved by bending the major groove of DNA over spermine and altering oligomer sugar puckering and interstrand phosphate distances. By comparison, Liquori's minor groove model appears to be less stable than the major groove model. This evidence favors a preferential binding of spermine to certain sites in DNA, which provides a powerful force for the modification of DNA conformation. Polyamines are aliphatic polycationic compounds that are found in all cells and have a significant role in the regulation of normal and malignant cell proliferation (1). Polyamine biosynthesis is highly regulated, and intracellular levels of polyamines can change rapidly by orders of magnitude when cell growth is stimulated (1). It has been proposed that polyamines affect growth by interacting with DNA. Evidence for these interactions obtained from experiments conducted in cell-free systems includes the ability of polyamines to precipitate DNA (2) and to raise the melting temperature of natural DNAs (3). Other results obtained by electron microscopy and micrococcal nuclease digestion of DNA show that polyamines can produce an organied condensed DNA structure (4). On the cellular level, depletion of intracellular polyamines by treatment with inhibitors of polyamine biosynthesis produces effects consistent with models in which polyamine interactions alter DNA conformation (5). It has been shown recently that polyamines can cause a B to Z conformational transition in both poly[d(g-m5c)] and poly[d(g-c)] (6, 7). In these studies, significant and specific changes in DNA structure can take place at low polyamine/ base-pair ratios. Differences in the ability of two spermidine analogs with the same charge but different structures to cause the B to Z transition suggest that a specific interaction causes the transition (8). Even though the interactions between DNA and polyamines have been described in detail, the molecular basis for an interaction has not been well-characteried. A major controversy in the field involves the question of whether or not the interaction is specific. In theoretical studies, polyamines have been treated as point concentrations of positive charge that interact with a DNA molecule considered to be a linear concentration of negative charge (9); in this counterion condensation model, chemical structure is ignored and specificity in interaction is not considered. Experimental results consistent The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C solely to indicate this fact with this model have been published (9-11). In other models, structures of both DNA and polyamines are taken into account. In one of these models, the tetracationic polyamine spermine bridges the minor groove ofdna to give a complex in which the two amine groups on one end of spermine interact with the phosphates on one DNA strand, and the two amine groups on the other end interact with phosphates on the opposite strand (12). Although this type of model allows preferential binding to the minor groove, base-specific interactions are ignored. Other models include crystal structures of DNA complexed with spermine. Drew and Dickerson (13) have described a dodecamer B-DNA crystal in which spermine asymmetrically bridges the major groove, and Quigley has described a Z-DNA crystal that contains two spermine molecules (14). Interactions of spermine with specific atoms have not been well-defined in either crystal structure. Because of this controversy, we have begun a detailed theoreticalinvestigation ofthe molecularnature ofpolyamine- DNA interactions. We have calculated the conformational energies for spermine and used energetically favorable conformations to construct possible models for DNA-spermine interactions. Because DNA conformation changes when spermine is introduced, we minimied the energy of our model complexes by allowing the conformation of both spermine and oligomer to change. METHODS We calculated conformational energies for spermine by varying all the possible single bonds using van der Waals and electrostatic potential functions through conformations staggered at angles of 6, 18, and 3. After favorable conformations had been determined, spermine was docked to the B-DNA oligomers d(g-c)5-d(g-c)5 and d(a-t)5-d(a-t)5 by using the MIDAS computer graphics program. Model oligomers and their complexes with spermine were constructed from the coordinates of Arnott and Hukins (15) and were displayed and manipulated on an Evans and Sutherland Picture System 2. Molecular mechanics calculations were performed by using the Assisted Model Building and Energy Refinement (AMBER) program. 11 Potential energy was evaluated as a function of bond length, bond angle, torsion angle, van der Waals and electrostatic potentials, and hydrogen bonds. Appropriate partial charges and constants were taken from Singh and Koilman (16) and Weiner et al. (17). Structures were refined until the root-mean-square gradient was <.1 kcal/mol (1 cal = 4.18 J). A distance-dependent dielectric constant was used to evaluate electrostatic interac- To whom reprint requests should be addressed. IJarvis, L., Huang, C., Ferrin, T. & Langridge, R. (1985) UCSF MIDAS User's Guide (Univ. of California, San Francisco). IIWeiner, P. K., Singh, U. C., Kollman, P. A., Caldwell, J. & Case, D. A. (1986) Program AMBER UCSF (Univ. of California, San Francisco), Unix version 2..

2 Biophysics: Feuerstein et al. tion energies. The charge on spermine was calculated by using the Gaussian 8-UCSF program (16). RESULTS AND DISCUSSION Our calculations show that in spermine conformations with the lowest energies, a fixed distance is maintained through the butyl group between the secondary amines with the aminopropyl end groups remaining relatively flexible. Using the calculated length for the diaminobutyl group, we searched for interactions between amine proton donors in spermine and proton acceptors in oligomer and found that the N-7 positions of alternating purine/pyrimidine sequences were probable binding sites. This selection is supported by the data of Drew and Dickerson (13) that shows binding of spermine to alternating purine/pyrimidine sequences across the major groove of B-DNA. Therefore, we modeled possible conformations by choosing N-7 binding sites for spermine in alternating purine/pyrimidine sequences, and we considered five models for spermine-dna complexes. Two models were of oligomers alone and two were of oligomer-spermine complexes. As a model that represents the spermine-dna complexes proposed by Tsuboi (3) and Liquori et al. (12), we considered d(a-t)5-d(a-t)5 complexed with spermine bridging the minor groove. Spermine-DNA Model. The total energy and the intramolecular and interaction energies of each energy-minimied model are listed in Table 1. Because these energies were calculated in vacuo, they cannot give a complete description of behavior in solution. A more complete characteriation of the real interactions in solution should consider the effects of ions, water, and entropy. Qualitatively, however, it is evident that the interaction of spermine with DNA lowers the total energy considerably, an expected result for the interaction of the positively charged spermine with the linear distribution of negative charges on DNA. Our detailed examination of these energy-minimied structures is discussed below. Interaction of Spermine with d(g-c)5d(g-c)5. The interaction energies of spermine with the oligomers are listed in the third row of Table 1. The large negative values indicate that significant stability results when spermine is complexed with them; for d(g-c)5d(g-c)5, the energy of stabiliation is kcal/mol. A view down the helical axis of the spermine-d(g-c)5 d(g- C)5 complex is shown in Fig. 1. Dashed lines represent hydrogen bonds between spermine and oligomer. There are two interactions at each amine group of spermine. The primary amines have formed two hydrogen bonds with phosphate oxygens. The secondary amines not only have formed the expected hydrogen bonds with the N-7 positions of guanine, but they also have pulled phosphate oxygens into positions favorable for the formation of hydrogen bonds. Introduction of spermine into the major groove of DNA thus provides a nidus for the stabiliing interactions we find in this model. In Fig. 2, the interaction energies of spermine with sugar/phosphates is plotted against the base sequence of the Table 1. Calculated energies in kcal/mol for oligomers alone and for spermine-oligomer complexes Model A B C D E Total energy Energy of DNA Spermine-DNA interaction energy Energy of spermine A, d(g-c)5*d(g-c)5; B, spermine-d(g-c)5d(g-c)5; C, d(a-t)5s d(a-t)5; D, spermine-d(a-t)5 d(a-t)5; E, Liquori spermine- d(a-t)5sd(a-t)5- Proc. Natl. Acad. Sci. USA 83 (1986) 5949 FIG. 1. View down the helical axis of the spermine-d(g-c)5.d(g- C)5 complex. The probable hydrogen bonds formed in the complex are shown as dashed lines. Each amine interacts with two proton acceptors. oligomer. The lower panel represents one strand of alternating purine and pyrimidine residues in the figure, running 5' to 3', and the upper panel represents the complimentary 3' to 5' strand. The interactions of spermine with the d(g-c) oligomer, shown as dashed lines, occur in two patterns. In the lower panel, two sugar/phosphates interact with spermine. The interaction at the 3-purine position is much stronger than the interactions at the neighboring 4-pyrimidine position. In the complimentary strand, however, three relatively equal energy interactions occur between spermine and three neighboring sugar/phosphate groups. These two patterns can be seen in Fig. 2 for the energy-minimied conformation of the spermine-d(g-c)5*d(g-c)5 complex. On the right side of the complex, three phosphate oxygens interact with spermine. Two of the oxygens, OA and OB, are bonded to a single phosphorus atom directly below spermine, while the other oxygen OA is on the same backbone but is bonded to a different phosphate atom. In this pattern, one phosphate provides twice the number of hydrogen bonds as the other, and therefore provides more stabiliation energy. The second pattern of spermine interaction is shown on the left side of Fig. 1. Three different phosphate oxygens (OAs) from three different residues interact with spermine. This pattern produces three approximately equal energies of stabiliation from three different phosphates as shown for 14-pyrimidine, 15-purine, and 16-pyrimidine in Fig. 2. It is interesting that two different patterns of phosphate a:.. o 2-5 -J :y 3'End 2PYR 19PUR 18PYR 17PUR 16PYR 15PUR 14PYR 13PUR 12PYR 11 PUR 5'End O -5-1L t% % I 5'End 1 PUR 2 PYR 3 PUR 4 PYR 5 PUR 6 PYR 7PUR 8PYR 9PUR 1OPYR 3'End NUCLEOTIDE SEQUENCE FIG. 2. Energy of interaction of spermine with sugar/phosphates. o, d(a-t)5 d(a-t)5; m, d(g-c)5-d(g-c)5.

3 595 Biophysics: Feuerstein et al. oxygen-spermine interactions are found within these two complimentary strands. Both local charge density and structure depend on the pattern of interaction; use of three residues instead of two per ligand allows, by exclusion, less total specific ligand binding per oligomer. We believe that these two patterns of interaction belie other, possibly multiple, configurations for binding of spermine to DNA phosphate oxygens that may be important for specificity of interaction. Fig. 3 is an energy plot for the interactions of spermine with bases in the major groove models. The single energy minimum for the d(g-c) oligomer in the upper and lower plots suggests that the interaction of spermine with bases is very specific and involves only one base in each strand. This interaction and its implications will be discussed below when the case of d(a-t)5sd(a-t)5 is considered. DNA Conformation in the Spermine-d(G-C)s5d(G-C)s Complex. The large negative electrostatic interaction between spermine and the d(g-c) oligomer allows it to change conformation to better accommodate spermine. The intramolecular energies for d(g-c)5d(g-c)5 alone and complexed with spermine are given in the second row of Table 1. Binding of spermine increases the energy of the oligomer from kcal/mol to kcal/mol. Thus, spermine binding lowers the total energy of the complex despite its significant destabiliing effects on oligomer conformation. These findings raise the question of how spermine binding into the major groove destabilies the oligomer. The spermine-d(g-c) oligomer complex before energy minimiation is shown in Fig. 4a and the energy-minimied complex is shown in Fig. 4b. The most dramatic aspect of the energy-minimied complex is the 25 bend in the oligomer produced by the interaction, which can be seen in Fig. 4b by imagining the average helical axis at the bottom of the oligomer molecule and comparing it with the axis at the top. We have performed similar calculations using counterions to neutralie the DNA phosphate backbone; this produces an almost identical bend in DNA. The result is reminiscent of a bend apparently stabilied by spermine in the crystal structure of trnaphe (18). The bend could be either a change of direction over several residues or a kink that occurs at only one position. It is possible to distinguish between these two possibilities from a plot of the stacking energy versus the position of the base pairs (Fig. 5). The control stacking energies for d(g-c)5-d(g- C)5 are represented by the open squares, which show the variance and symmetry expected for a naked piece of DNA. The same model complexed with spermine, represented by the solid squares, has an asymmetric increase in the stacking energies between base pairs 3-4, 6-7, and 7-8. The increases j < 4 3' -2 --o~~ - Vb~~v 5' End 1 PUR 2 PYR ~~F'-- 3PUR 4PYR 5PUR 6PYR 7PUR 8PYR 9PUR 1OPYR 3End NUCLEOTIDE SEQUENCE FIG. 3. Energy of interaction of spermine with bases. o, d(a- T)5-d(A-T)5; o, d(g-c)5 d(g-c)5. - Proc. Natl. Acad. Sci. USA 83 (1986) a b FIG. 4. Views into the major groove of d(g-c)5-d(g-c)5 with spermine in place before (a) and after (b) energy minimiation. Note the decrease in distance across the major groove (n) and the increase across the minor groove (o) after energy minimiation was performed. Squares and circles represent the same points on the helix, and are included for comparison between a and b. of no more than 1 kcal/mol of base pair can be partially attributed to decreases in intrastrand and interstrand phosphate distances discussed below. These data show that the bend does not significantly disrupt base stacking. Fig. 4b confirms that the bend occurs over several base pairs with no evidence of a kink that either occurs over one base pair or disrupts stacking. We note that other groups using similar minimiation techniques have obtained conflicting results regarding a bend in DNA produced by thymine dimers (19, 2). They found a bend only when DNA conformation was altered before energy minimiation. We have not altered B-DNA conformation before energy minimiation and the uncomplexed oligomer has no bend after minimiation. Therefore, the simple docking of spermine has produced the bend. Even though base stacking is not greatly disrupted when spermine is complexed into the major groove, other changes in the structure of the oligomer have occurred. First, the bend in the d(g-c) oligomer has been produced by folding the major groove over spermine, which allows the maximum interaction to occur between the oligomer and spermine. This effect can be seen by comparing the short distance across the major groove in Fig. 4b (solid squares) with the much longer distance across the major groove in Fig. 4a (solid squares). The change in distance across the major groove brings the interstrand phosphates closer to each other, which increases their energy of repulsion. Second, the minor groove has widened opposite the major groove, as shown in Fig. 4b (solid

4 -j ( -i c) > -2 a: wi Biophysics: Feuerstein et al. I I % I II I, % I % I, %d I I " % "Cr 1% " b' BASE PAIR NUMBER 9 1 FIG. 5. Base stacking energies of oligomers that have been minimied with and without complex formation with spermine. o, d(a-t)5-d(a-t)5; *, spermine-d(a-t)5 d(a-t)5 complex; o, d(g- C)5-d(G-C)5; *, spermine-d(g-c)qd(g-c)5 complex. Proc. Natl. Acad. Sci. USA 83 (1986) 5951 circles), compared to the much narrower minor groove in the unminimied oligomer in Fig. 4a (solid circles). Thus, the increased width of the minor groove is a consequence of the bend: as the oligomer pivots on its helical axis to enclose spermine in the major groove, the minor groove opens on the opposite side of the helix. The orientation of phosphates also changes when d(g- C)5d(G-C)5 is complexed with spermine. As described above, a change in the conformation of the oligomer allows a network of interactions to form between proton donors in spermine and proton acceptors in DNA. To position the phosphate oxygens advantageously, the distance between the appropriate neighboring phosphorus atoms has been shortened from =7 A to 5.5 A. This has been accomplished by altering sugar puckering, which shortens the distance between interstrand phosphates. This shortening contributes to the bend, which allows the appropriate phosphate groups to interact with the amines in spermine. The electrostatic potentials for minimum energy conformations of d(g-c)5d(g-c)5 are shown as color-coded surfaces in Fig. 6. Blue is the most positive potential, green and yellow are intermediate potentials, and red is the most negative potential. The structure on the right is the energyminimied uncomplexed oligomer and that on the left is the energy-minimied oligomer complexed with spermine in the major groove. (For clarity, spermine is not shown in the complex.) As noted above, sies of the major and minor grooves are structurally different in these polymers. The fit of spermine into the major groove is not only structural, however. It is evident that the major groove of the polymer becomes intensely negative after complexing with spermine. This effect increases both the stabiliation of the complex and the specificity of binding. Thus, both electrostatic and structural specificity are evident in the major groove model. The major changes in the conformation of the energyminimied major groove spermine-oligomer complex are the bend and its electrostatic consequences and changes in distances between intrastrand phosphates caused by sugar puckering. The prediction of a bend is compatible with studies that show compaction of DNA after treatment with polyamines (4). Moreover, the changes in sugar puckering and phosphate distances are of particular interest because of their similarity to well-characteried features of A- and Z- DNA (ref. 21, pp , 22-24, ). In both A- and Z-DNA, the sie of the grooves is altered, but in Z-DNA the helical sense is altered as well. Because polyamines cause a transition from the B to both Z and A states, these energy minimiation results suggest a basis for either the B to A or B to Z transitions (6, 7, 22). Structure of the Major Groove Spermine-d(A-T)5 d(a-t)s Complex. The interactions of spermine within the major groove of d(a-t)5-d(a-t)5 are qualitatively similar to its interactions within the major groove of d(g-c)q5d(g-c)5. The interaction energy of spermine with d(a-t)5-d(a-t)5 listed in the third row of Table 1 is kcal/mol, a value similar to the kcal/mol obtained for the spermine-d(g- C)5d(G-C)5 complex. Spermine complexed into the major groove of d(a-t)5-d(a-t)5 forms a network of interactions with proton acceptors on the oligomer, does not greatly disrupt base stacking (Fig. 5), and limits interactions to a specific region of the oligomer (Fig. 2 as explained above, FIG. 6. Electrostatic potential surface in kcal/mol for the d(g-c) oligomer complexed (Left) and not complexed (Right) with spermine. The most negative surface is coded red (x < -6), yellow (-4 > x < -6) and green (2 > x > -4) are intermediate, and blue (x > 2) is the least negative. Note the increased negativity in the major groove of the complexed polymer (Left). For clarity, spermine is not shown.

5 5952 Biophysics: Feuerstein et al. and Fig. 3, as explained more fully below). Major changes in the conformation of the oligomer are the folding of the major groove over spermine, the increase in the width of the minor groove, and the shortening of the distances between intrastrand phosphates caused by an alteration in sugar puckering. In addition, electrostatic complimentarity is preserved in the spermine-oligomer complex, but the absolute values of its electrostatic potential are somewhat reduced. Plots of the interaction energies for the complex between spermine and d(a-t)5 d(a-t)5 are shown in Figs. 2 and 3. The interactions with phosphates and sugars shown in Fig. 3 have been explained above. Plots of the energy of interaction of spermine with bases in each nucleotide versus the nucleotide sequence are shown in Fig. 3. Spermine binds to bases of residues 5 and 15 for both d(g-c)5d(g-c)5 (dashed lines) and d(a-t)5-d(a-t)5 (solid lines). Binding of spermine to bases in d(g-c)5d(g-c)5 occurs in two residues only, but binding for d(a-t)5-d(a-t)5 is spread over neighboring residues as well. As shown in Fig. 1, a hydrogen bond forms between the secondary amine groups of spermine and the N-7 positions of purine bases in d(g-c)5-d(g-c). While this interaction forms the basis for spermine binding to d(a-t)5-d(a-t)5, a nonbonded interaction also appears between the 5-methyl groups of thymine and the methylene groups of spermine (not shown). This difference in binding interactions forms a basis for possible specificity of spermine binding to DNA. Spermine-d(A-T),5d(A-T)5 Minor Groove Model. We have also considered the minor groove model proposed by Tsuboi (3) and Liquori et al. (12). In their model, spermine bridges the minor groove and interacts with the phosphate backbone on either side of the groove. We chose to study spermine complexed with the d(a-t) oligomer because it has a higher concentration of negative charge in the minor groove than the d(g-c) oligomer, which should provide a more favorable interaction with the positively charged spermine. The last two columns in Table 1 compare the energies for the major and minor groove models for the d(a-t) oligomer. The total energy of the major groove model is >7 kcal/mol more stable and the interaction of spermine with the oligomer is 12 kcal/mol more favorable in the major than in the minor groove model. This result is supported by the finding of a cavity created in the minor groove beneath the spermine four-carbon bridge (not shown). Because these calculations were performed in vacuo, results cannot yield precise quantitative information for the complex in solution. However, these qualitatively large energy differences and our structural results support a preference for spermine binding to the major groove in the case of d(a-t). In the case of d(g-c), electrostatic effects make the minor groove model even less feasible. Spermine Conformation in the Minimied Models. The intramolecular energy for spermine complexed to the d(g-c) and d(a-t) oligomers for the major groove model and to the d(a-t) oligomer for the minor groove model are listed in the bottom row of Table 1. The lowest energy was found for the d(a-t) minor groove model and the highest energy was for the d(g-c) major groove model. Thus, spermine in the d(g-c) oligomer has the least favorable conformation of the three considered. Spermine alone was found to have a total energy of 63.9 kcal/mol. In each minimied model, the three central torsion angles for spermine, which control the distance between the secondary amines, are all near 18 (data not shown). This finding is consistent with our calculation that in the most stable configuration of spermine the central diaminobutane is all trans. Proc. Natl. Acad Sci. USA 83 (1986) The results of the theoretical calculations presented here show that spermine docked into the major groove spontaneously produces a bend in B-DNA. Our evidence supports the major groove over the minor groove model on both structural and energetic grounds. Moreover, these models provide a hypothesis for explaining the significant modifications in DNA structure known to occur in solution. We are continuing our molecular mechanics studies by considering other base sequences and binding positions for spermine and other conformations of DNA. We are also beginning studies of the molecular dynamics of the system and are conducting experiments of spermine-dna interactions by using physical techniques in synthetic oligomers to test aspects of the model we propose. We thank Peter Kollman and U. C. Singh for helpful discussions, Robert Langridge for the use of the MIDAS program and the facilities of the Computer Graphics Lab at the University of California at San Francisco, and Neil Buckley for editing and preparing the manuscript. This work was supported in part by National Institutes of Health Program Project Grant CA-13525, National Institutes of Health Grant RR 181 (to the Computer Graphics Lab), the National Institutes of Health National Cooperative Drug Discovery Group Grant CA-3766, and the Andres Soriano Cancer Research Fund. 1. Tabor, C. W. & Tabor, H. (1984) Annu. Rev. Biochem. 53, Heby,. & Agrell, A. (1971) Hoppe-Seyler's Z. Physiol. Chem. 352, Tsuboi, M. (1964) Bull. Chem. Soc. Jpn. 37, Gosule, L. C. & Schellman, J. A. (1976) Nature (London) 239, Marton, L. J. (1985) West. J. Med. 142, Behe, M. & Felsenfeld, G. (1981) Proc. Natl. Acad. Sci. USA 78, Chen, H. H., Behe, M. J. & Rau, D. C. (1984) Nucleic Acids Res. 12, Thomas, T. J., Bloomfield, V. A. & Canallakis, M. (1985) Biopolymers 24, Bloomfield, V. A. & Wilson, R. W. (1981) in Polyamines in Biology and Medicine, eds. Morris, D. R. & Marton, L. J. (Dekker, New York), pp Braunlin, W. H., Strick, T. J. & Record, M. T. (1982) Biopolymers 21, Wemmer, D. E., Srivenugopal, K. S., Reid, B. R. & Morris, D. R. (1985) J. Mol. Biol. 185, Liquori, A. M., Constantino, L., Cresceni, V., Elia, B., Giglio, E., Puliti, R., Desantix, S. M. & Vitigliani, V. (1967) J. Mol. Biol. 24, Drew, H. R. & Dickerson, R. E. (1981) J. Mol. Biol. 151, Quigley, G. J. (1982) in Molecular Structure and Biological Activity, eds. Griffin, J. F. & Duax, W. L. (Elsevier-North Holland, Amsterdam), pp Arnott, S. & Hukins, D. W. L. (1972) Biochem. Biophys. Res. Commun. 47, Singh, U. C. & Kollman, P. A. (1984) J. Comput. Chem. 5, Weiner, S. J., Kollman, P. A., Case, D., Singh, U. C., Ghio, C., Alagona, G. & Weiner, P. K. (1984) J. Am. Chem. Soc. 16, Quigley, G. J., Teeter, M. M. & Rich, A. (1978) Proc. Natl. Acad. Sci. USA 75, Pearlman, D. A., Holbrook, S. R., Pirkle, D. H. & Kim, S. H. (1985) Science 227, Rao, S. N., Keepers, J. W. & Kollman, P. (1984) Nucleic Acids Res. 12, Saenger, W. (1984) Principles of Nucleic Acid Structure (Springer, New York). 22. Minyat, E. E., Ivanov, V. I., Krityn, A. M., Minchenkova, L. E. & Schyolinka, A. K. (1979) J. Mol. Biol. 128,