The solvent density in the minor groove is depicted for presentation here as the linear superposition of 16 individual configurations (snapshots) collected at equally spaced intervals along the production segment of the Monte Carlo realization. Water molecules found within the first-shell radii of the N-9, C-4, N-3, C-2, and N-1 atoms of adenine, of the N-3, 02, and N-1 of thymine, of the N-9, C-4, N-3, C-2, and N2 of guanine, and of the 02 and N-1 of cytosine were assigned as the first coordination shell of the minor groove. A second shell comprised of all water molecules forming a primary coordination to first-shell waters was defined to further elaborate the minor-groove hydration. Proximity analysis turned up -50 waters of hydration in the minor groove of the dodecamer, 12 primary and 38 secondary with respect to the bases.
The calculated solvent density in the minor groove of the duplex of our model d(CGCGAATTCGCG) dodecamer is shown in Fig. 2. The small circles outside the nucleic acid are the positions of the oxygen atoms of the solvent water molecules, with filled points indicating the first-shell coordination running along the floor of the minor groove and the open points indicating the position of second-shell groove waters (some of these second-shell groove waters may reside in a first-shell coordination with atoms of the sugarphosphate backbone). An individual point simply indicates a water molecule is present in one or another of the solvent configurations contributing to the statistical state of the system. The important information conveyed by this figure is the clustering of points, indicating the concentration of solvent density in that region. The calculated minor-groove solvent density for the model dodecamer (Fig. 2) shows considerable localization and provides theoretical evidence for the existence of discrete hydration sites in the minor groove and for the existence of a well-defined ordered water structure in this region corresponding to Dickerson's observed spine of hydration.
The nature of the calculated spine is further revealed by a display of waters in the first shell of the minor groove and the nucleotide bases of the dodecamer (Fig. 3). The first-shell coordination consists of water molecules hydrogen bonded to the polar atoms along the floor of the minor groove, and more or less situated in the plane of the nucleotide base pairs. By contrast, the first-shell waters in the crystallographic spine of hydration were found to bridge the 0 and N atoms of successive base pairs. Some discrepancies in the calculated and observed results are expected due to the difference between the canonical B-form and the crystallographic dodecamer structure as well as the approximations inherent in the intermolecular force field. The preferred calculated hydration site for an A-T base pair turns out to be the adenine N-3 acceptor site. In the C-G region the guanine N2 donor site is clearly favored over. the cytosine O2 acceptor site. The localization of solvent density is a bit more pronounced in the C G than in the APT region, reflecting the more demanding geometrical restrictions of a donor compared with an acceptor site on the solute (28). The calculated spine of hydration can be traced out by following the hydrogen-bonded water network up the minor groove. The calculated spine, as in the crystal structure, is not just a first-shell entity but involves for the most part an alternation of first-shell and second-shell water molecules of the minor groove. This can be more clearly seen in Figs. 4 and 5, where snapshots of the hydration of the minor groove are shown in molecular detail with the nucleic acid "trellis" removed. Note that once the second shell is involved the spine is not necessarily a single path, but an incipient water network in which the motif of tetrahedral coordination characteristic of liquid water coordination is quite discernible. This point was noted in the discussion of the crystal structure as well (ref. 7, but see also ref. 8), even though some of the participating waters were crystallographically disordered and thus not observed. The water network in the C-G region (Fig. 5) shows some interesting water polygons of order 4 and 5.
An important result from our calculation is the extent to which the C G as well as the APT region can support an ordered water structure in the minor groove. We find the COG region, with the floor of the minor groove lined with guanine N2 donor groups and cytosine O2 acceptor sites, to be just as capable of supporting a spine of hydration as is an APT region. The calculated spine of hydration in the minor groove of our model dodecamer extends completely from one end of the structure to the other, with the interface between the C-G and A-T regions dealt with easily by the geometric flexibility of water-water hydrogen bonding interactions. The penetration of water into the DNA is of course greater for ART tracts. Thus our calculations support the general idea of the spine of hydration, but indicate that the spine is not necessarily specific to A-T-rich regions as originally suspected. An ordered water network in the minor groove of DNA supported by C G base pairs provides an attractive alternative explanation of the positive binding entropies for netropsin complexed with poly[d(G-C)]poly- [(C-G)] observed by Marky and Breslauer (15).* The role of the Dickerson spine of hydration in the preferential stabilization of B-DNA specific for A-T-rich sequences also needs to be reconsidered in light of this result.
A cautionary note on the relationship of the spine of hydration to themodynamic stability of the DNA duplex has been aired by our group (29). We pointed out that ordered water in the minor groove may be in an energetically favorable state but is entropically unfavorable [enthalpy-entropy compensation (30)] and that more disordered water in the major groove, quite capable of favorable energetics, could well be as important to the stability of B-DNA as the minor-groove spine of hydration. At this point we must emphasize that the spine of hydration as observed crystallographically and as calculated as described herein, is a structural entity for which the corresponding energetics are not unequivocally established.
In conclusion, we note with considerable interest the extent of localization of the calculated hydration sites in the A-T and G-C regions found in the simulation and also the difference in penetration of the groove waters into DNA for A-T and C-G base pairs. This raises another potentially interesting thermodynamic issue, the extent to which ordered water structure in DNA is sequence dependent. A particular sequence that supports an unusually ordered water structure could be particularly favorable for binding substrate due to the extra entropy increase on desolvation, providing an additional entropic drive to the binding free energy (27). We believe that the present calculations and results indicate that further theoretical and experimental investigation of this idea would be worthwhile.
Very helpful discussions with Dr. Helen Berman of Fox Chase Cancer Research Institute and Prof. Kenneth J. Breslauer of Rutgers University are gratefully acknowledged. D.L.B. particularly acknowledges the hospitality and scientific discussions and interactions of the 1986 CECAM Workshop on Nucleic Acid Hydration (D. E. Westhof, organizer) at the University of Paris, Orsay and the "morte subite" group. This research was supported by grants from the National Institutes of Health (GM-37909), National Science Foundation (CHE-8696117), the Office of Naval Research, as well as the generosity of Merck Sharpe and Dohme Research Laboratories and the National Bureau of Standards. Computer facilities were provided by the Pittsburgh Supercomputer Center and Wesleyan University.
Footnotes
*A referee cautions that positive binding entropies are also observed for intercalators that presumably do not seriously influence the minor groove bound water.
Answers to all your Biology Questions
