«DNA Bending and Structural Waters in Major and Minor Grooves of A-tracts. Monte Carlo Computer Simulations * ‡ § Alexander V. Teplukhin, Valery ...»
October 9, 1996
DNA Bending and "Structural" Waters in Major and Minor Grooves of
A-tracts. Monte Carlo Computer Simulations
* ‡ §
Alexander V. Teplukhin, Valery I. Poltev and Victor B. Zhurkin
Institute of Mathematical Problems of Biology, Russian Academy of Sciences,
Pushchino, Moscow Region, 142292, Russia
Institute of Theoretical and Experimental Biophysics, Russian Academy of Sciences, Pushchino,
Moscow Region, 142292, Russia
§ Laboratory of Mathematical Biology, NCI, NIH, Bg. 12B, Rm. B116, MSC 5677, 12 South Dr., Bethesda, MD 20892-5677 E-mail: email@example.com
DNA, as a linear molecule, is quite susceptible to the environment, each nucleotide being substantially exposed to solvent. At the same time, the internal interactions stabilizing any one particular conformation of DNA, are relatively weak, compared to globular proteins. As a consequence, DNA conformations are aptly responsive to external effects, and the local structure of DNA can vary over a wide range. In turn, the structural variability of DNA, such as the sequencespecific bending and curvature, is important for a number of basic processes, including protein binding and gene regulation (for reviews see Trifonov, 1986; Crothers et al., 1990; Hagerman, 1990, 1992; Harrington and Winicow, 1994; Travers, 1995; Olson and Zhurkin, 1996).
Bent DNA is characterized by periodically modulated sizes of the major and the minor grooves (Drew and Travers, 1985; Burkhoff and Tullius, 1987). The atom arrangements in the grooves differ substantially for different sequences; thus, the possibility exists that some of sequences can be solvated with more structural waters than others. This raises the important question of the correlation between the sequence-dependent structures and energetics of DNA duplex on the one hand, and the features of its hydration shell on the other hand. In particular, it is of interest to examine the role of solvation effects in stabilizing the bent conformations of the double helix.
To address this question, we have studied structural water patterns formed in the DNA grooves in a sequence-specific fashion, and compared them to bulk waters. We have chosen Monte Carlo (MC) simulations as a tool to reveal the most favorable modes of solvation of DNA, and have analyzed duplexes containing the so-called A-tracts (contiguous runs of several adenines in one strand). The reason for doing so is given below.
Most of the curved sequences studied so far, starting historically with kinetoplast DNA, contain Atracts periodically alternating with "mixed" GC-rich spacers (Marini et al., 1982; Diekmann, 1986;
Koo et al., 1986; Hagerman, 1988; Beutel and Gold, 1992). Strong DNA curvature is likely to be a consequence of the difference in conformation and flexibility between the A-tracts and the GC-rich spacers. Among other factors, solvation effects could be related to this difference. Currently there is no consensus in the literature regarding explanation of this phenomenon at the atomic level. Here we mention only briefly the well known discrepancies between the "A-tract" or "junction" model (Crothers et al., 1990; Haran et al., 1994), the "non-A-tract" model (Calladine et al., 1988; Goodsell et al., 1994; Bruckner et al., 1994) and various "wedge" models, both static and flexible (Bolshoy et al., 1991; De Santis et al., 1990; Zhurkin et al., 1991; Olson et al., 1993). All these models, however, have one feature in common, which appears to be consistent with the whole multitude of 3 experimental data, namely, that the A-tracts in solution are characterized by a more negative roll of base-pairs and compressed minor groove, compared to the "mixed" GC-rich sequences.
At the macroscopic level of hundreds of base pairs, the DNA curvature is easily detectable by retardation on polyacrylamide gel. This retardation crucially depends on the environment: type of cations, activity of water and temperature. Divalent cations increase DNA retardation on gels (Marini et al., 1984; Shlyakhtenko et al., 1990; Nickol and Rau, 1992; Brukner et al., 1994). By contrast, addition of alcohol (EtOH and MPD) and increase in temperature gradually diminish the curvature of the kinetoplast DNA (Marini et al., 1984; Sprous et al., 1995). Notice, however, that when the "spacer" sequences located between the A-tracts are AT-rich (T5, TATAT or TCTCT), the temperature dependence is more complicated, and the maximum curvature is observed at 25°-35°C (Diekmann, 1987; Shlyakhtenko et al., 1990, 1992). All these data, taken together, reveal the role of solvent in moderating macroscopic DNA bending and curvature.
At the local level, the NMR data indicate that the conformation and dynamics of the A-tracts are different at low temperature (5°-15°C) and at room temperature or higher (Leroy et al., 1988;
Nadeau and Crothers, 1989; Katahira et al., 1990). Traditional interpretation attributes this difference to the changes in the base-pair inclination (Marini et al., 1984), in particular, to a negative roll at low temperature (Lipanov and Chuprina, 1987; Nadeau and Crothers, 1989). This interpretation is consistent with the temperature dependence of the NMR and gel electrophoresis measurements mentioned above. By analogy with the X-ray fiber structure of poly(dA):poly(dT) also having negative roll of bases and a narrow minor groove, the low temperature conformation of the A-tract in solution is denoted B' form (Arnott et al., 1983; Alexeev et al., 1987; Chuprina, 1987).
Several detailed hypotheses have been developed to elucidate the unusual properties of An:Tn tracts in solution, but none of them accounts for all available data (Diekmann et al., 1992). One of them, the "hydration spine" hypothesis, considers a regular positioning of the "structural" waters in the minor groove as the main factor stabilizing the B'-like conformation of A-tracts in solution (Chuprina, 1985, 1987). This model, based on the crystal structure of the B-DNA dodecamer (Drew and Dickerson (1981), successfully predicted the larger number of bound waters in An:Tn tracts compared to the alternating A-T sequences (Buckin et al., 1989). It remains unclear, however, why the regular hydration spine in the narrow minor groove should be energetically preferable compared to water strings in the relatively wide minor groove of the "standard" B-form (Prive et al., 1987).
Furthermore, this concept based exclusively on the formation of the hydration spine in the minor groove, fails to explain why the DNA curvature depends upon the nature of the exocyclic groups in the purine 6- and 7-positions in the major groove (Diekmann et al., 1987; Seela et al., 1989). The hydration of the minor groove alone cannot account for the noticeable difference in the electrophoretic nobilities of the DNAs containing A4:T4 and (AATT)2 tracts (Koo et al., 1986;
Hagerman, 1988). In addition, the amount of water released upon binding of netropsin to poly(dA):poly(dT) is substantially larger than would be expected if the frozen-in water melted only from the minor groove (Marky and Kupke, 1989). The optical, spectroscopic and calorimetric studies of the B' → B "premelting" transition in the An:Tn fragments also point to melting and release of "bound" water into the bulk solvent (Herrera and Chaires, 1989; Chan et al., 1990). So, a thorough examination of DNA hydration, depending on DNA conformation, especially in the major groove, is necessary. One more reason to pay attention to the major groove is that the third strand of the triplex is located there (Felsenfeld et al., 1957). Finally, many proteins bind in the major groove, and the "structural" waters appear to be an important component of the protein-DNA interface (Otwinowski et al., 1988; Shakked et al., 1994).
A number of computer simulations of DNA hydration have been performed starting with pioneering works by Clementi and Corongiu (reviewed in Clementi (1983)), and up to recent times (McConnell et al., 1994; Cheatham and Kollman, 1996; Yang and Pettitt, 1996). These studies use both MC simulation techniques, (Clementi, 1983; Poltev et al, 1988; Subramanian and Beveridge, 1989; Eisenhaber et al., 1990; Teplukhin et al., 1991; Vovelle and Goodfellow, 1993 and references therein) and molecular dynamics, MD (Seibel et al., 1985; Swaminathan et al., 1991; Chuprina et al., 1991; Miaskiewicz et al., 1993; Fritsch et al., 1993; McConnell et al., 1994). Hydration free energies were also evaluated for DNA, based on calculation of the solvent accessible surface areas (Raghunathan et al., 1990, and references therein). In addition, the potential-of-mean-force approach was developed recently for computing the hydrophilic hydration of DNA (Hummer and Soumpasis, 1994; 1995).
The main emphasis of these studies was on the hydration shell in the minor groove, in particular on the hydration spine. E.g., our earlier results (Poltev et al., 1988; Teplukhin et al., 1991) indicate
Results of computations for the major groove are less explicit, in accord with the X-ray and NMR data indicating a more disordered state of water shell in the major groove (Prive et al., 1991;
Kubinec and Wemmer, 1992; Liepinsh et al., 1992; Fawthrop et al., 1993). The role of specific hydration patterns in the major groove in stabilizing a particular conformation of DNA in solution, to the best of our knowledge has never been discussed. Among very few detailed results published so far, we mention the "regions of high water density... between consecutive adenines and at the center of base-pair steps" (Hummer and Soumpasis, 1994). Similarly, when analyzing the B' form, we found earlier that the water molecules bridging N7 and H(N6) atoms of the two adjacent adenines often make a third hydrogen bond with an N7 atom of the adenine whose H(N6) takes part in this bridging (see Figure 1 in Poltev et al., 1988). These water molecules can also be treated in a spine-like manner, by analogy with the minor groove. This possibility was not systematically analyzed, however. The present study was undertaken to fill this gap.
Selecting the method for simulating the DNA hydration, we took into account the following considerations. In all Monte Carlo (MC) studies published so far, hydration of DNA has been simulated either for the crystal and fiber structures, or for the energy minimized DNA conformations. This approach is disadvantageous, however, since the DNA flexibility is ignored.
On the other hand, the MD simulations of flexible oligonucleotide duplexes were performed by several groups of authors (Seibel et al., 1985; Zielinski et al, 1988; Miaskiewicz et al., 1993; Fritsch et al., 1993), but recently Beveridge and his colleagues concluded that "MD on DNA must be carried out for at least an order of magnitude longer than previously expected, and perhaps even longer" (McConnel et al., 1994). Although remarkable progress has been achieved in this area during the last years, and now the overall behavior of the double helix is predicted quite realistically, the details of the DNA structure are not entirely consistent with the experimental data.
For example, the calculated equilibrium twisting angle for B-DNA is by several degrees less than the solution measurements (Cheatham and Kollman, 1996), so the MD-simulated B-form has certain features usually inherent in A-form. Moreover, the longest nanosecond-scale MD simulations published recently (Cheatham and Kollman, 1996; Yang and Pettitt, 1996) reveal essential force field dependence of the resulting DNA structure (B-like and A-like structures respectively).