Role of Water in Defining the Structure and Properties of B-Form DNA

Abstract

DNA in the cell is rarely naked but normally protein-bound in nucleosomes. Of special interest is the DNA bound to other factors that control its key functions of transcription, replication, and repair. For these several transactions of DNA, the state of hydration plays an important role in its function, and therefore needs to be defined in as much detail as possible. High-resolution crystallography of short B-form duplexes shows that the mixed polar and apolar surface of the major groove binds water molecules over the broad polar floor of the groove in a sequence-dependent varied manner. In contrast, the narrower minor groove, particularly at AT-rich segments, binds water molecules to the polar groups of the bases in a regular double layer reminiscent of the structure of ice. This review is largely devoted to measurements made in solution, principally calorimetric, that are fully consistent with the location of water molecules seen in crystals, thereby emphasizing the substantial difference between the hydration patterns of the two grooves.

1. Crystallographic Studies

High-resolution crystallographic determinations of the structures of short naked duplexes show hydrating waters in the major groove exhibiting preferred but varied sequence-dependent positions on the polar base of the groove [1,2]. Figure 1 shows the water distribution for one particular duplex at a resolution of 1.6 Å. In contrast, for the narrow minor groove of AT-rich sequences, water molecules are regularly arranged in two layers, the lower being H-bonded as in ice to the upper layer and to polar groups on the edges of the bases (Figure 2). It follows that this water is not ordered in the regular way expected of water bound to hydrophobic surfaces. This minor groove water distribution was first noticed crystallographically [3,4], where it was shown that water is fixed by the N3 of A and O2 of T groups of the AT pair. Figure 1 makes it clear that the spine of hydration covers not only AT pairs but also includes an intervening CG pair (also see Refs. [5,6] for further determinations of hydration in crystals of short DNA duplexes). It is worth noting here that NMR studies of DNA in solution [7,8] also revealed the presence of less mobile water in the minor groove, so a clear distinction between the two grooves became apparent. However, these are largely crystal structures, so the disposition of water molecules might differ from that in solution, meaning that measurements are needed in free solution to confirm or contradict the observations made in the solid state.

2. Calorimetric Studies

The stability of naked DNA can be studied thermodynamically using the two well-established techniques of microcalorimetry: DSC and ITC, and the details give an indication of the state of the hydrating water molecule [11]. Such investigations have been performed on duplexes of variable composition and consisting of less than 20 bp with the advantage that they all exhibit a single cooperative transition when thermally denatured. The DSC scan of an all-CG 12 bp duplex in Figure 3 [12,13] demonstrates the presence of a two-state transition centered at 83.6 °C, associated with strand separation, preceded by a gradual non-cooperative accumulation of heat. The most important conclusion from this deconvolution is the clear demonstration of a heat capacity increase upon denaturation—as also seen in protein denaturation—but of a smaller magnitude for DNA (∆Cp = +0.13 kJ/K·mol-bp). This contradicts the prevailing opinion in the literature that there is no change in the heat capacity on DNA melting: the finite value of ∆Cp means that the thermodynamic parameters of DNA vary with temperature. Importantly, changes in the heat capacity of such systems reflect the state of hydrating water molecules. When the hydrophobic surface of DNA (or proteins) becomes hydrated upon denaturation, the newly bound water is weakly held and thus has a greater heat capacity than that of bulk water—so the system exhibits an increase in heat capacity, i.e., a positive contribution to ∆Cp. In contrast, when polar groups become exposed to the solvent, the hydrating waters are tightly bound with a heat capacity less than that of bulk water, resulting in a reduced heat capacity, i.e., is a negative contribution to ∆Cp. The increase in the Accessible Surface Area (ASA), apolar and polar, on DNA melting can be assessed in silico, and it turns out that for DNA, the polar component is about 55% of the total newly exposed surface—in contrast to only ~33% found for globular proteins. The increased negative contribution from exposure of the polar surface explains why, while still positive, ∆Cp for DNA melting is much less than that found for proteins (see Ref. [13] for details).

An important early observation on DNA melting was the finding that the greater the content of GC pairs, the higher the melting temperature, T_m, Ref. [14]. Figure 4 confirms this with DSC scans for three pairs of duplexes differing only in their GC/AT ratios. The observation that the inclusion of AT pairs reduces the melting temperature was immediately explained by assuming that the presence of only two H-bonds in an AT pair was responsible, i.e., the enthalpy of denaturing a GC pair is greater than that of an AT pair. However, the enthalpy of the lower-melting, AT-containing duplexes in Figure 4 is greater than that of the corresponding all-GC duplexes by about 10%.

3. Thermodynamic Characteristics of the Individual Base Pairs

The true explanation for the apparent stabilizing effect of CG pairs involves the role played by water molecules bound in the minor groove to AT pairs. To obtain the enthalpy of forming individual CG and AT pairs requires comparing the enthalpies of melting duplexes of varying compositions and lengths [15]. It was observed that the enthalpy of CG pairs is simply additive, but this is not the case for AT pairs, where it depends to some extent on context, i.e., the nature of the neighboring (flanking) base pairs. If the data for AT pairs are averaged, the values obtained for dissociation of the individual base pairs are as follows in

Table 1. Contributions of CG and AT base pairs to the enthalpy, entropy, Gibbs free energy, and heat capacity increment in DNA dissociation (melting). See Ref. [16].

Base Pair	∆Hcoop (kJ/mol-bp)	∆Scoop (J/K·mol-bp)	∆Gcoop (kJ/mol-bp)	∆Cp (kJ/K·mol-bp)
CG	19.0	36.2	8.2	0.13
AT	28.0	73.5	6.1	0.13

[16]:

A comparison of these values shows that the duplex-stabilizing effect of the CG base pair is greater than that of an AT not because its enthalpic contribution is larger (it is, in fact, less) but because its entropic contribution is substantially smaller than that of AT pairs [15,17]. The question is then as follows: why are the entropic and enthalpic contributions of the AT base pair greater than for CG pairs? The simplest explanation is that dissociation of AT pairs in a duplex is accompanied by the strong loss of ordered waters from the minor groove, and this is responsible for their greater entropy and enthalpy. Hence, the reduced magnitude of the Gibbs free energy, ∆G, for AT pairs.

If the water-hydrating AT pairs in the minor groove are assumed to have the thermodynamic properties of ice (about 1.5 mols of ‘ice’ per AT pair, as suggested by the structure shown in Figure 2), then its subtraction from the observed enthalpy and entropy of an AT pair will yield the intrinsic parameters characterizing the dissociation of an AT pair. The result is an enthalpy of 19 kJ/mol-bp and entropy of 40.5 J/K·mol-bp. So, the enthalpy of dissociating an AT pair is, within experimental error, the same as dissociating a GC pair, meaning that the dissociation/formation of H-bonds is a non-enthalpic process. In contrast, the entropy of dissociating the GC pair is about 4 J/K·mol-bp less than that of forming an AT pair: this extra negative entropy comes from hydrating the polar groups that form the extra H-bond in a GC pair, making the overall entropy less positive.

The overall conclusion is thus that H-bonds are non-enthalpic, and their contribution to duplex stability comes from the entropy gain resulting from dehydration of their polar groups that occurs upon their incorporation into the duplex [16]. In the formalism frequently used to describe the folding of proteins, the formation of base pairs can be viewed as the burying of polar groups in the interior of the macromolecule, i.e., their exclusion from the solvent. This thermodynamic description of the hydrogen bond should not be restricted to those in DNA—equally, it is the case in forming H-bonds in the α—helix [16,18].

With accurate values to hand for the enthalpy and entropy of CG and AT pairs, two factors bear in the question of how the T_m values of primers and probes are best calculated for PCR and other DNA extension reactions. These are: (1) the existence of a heat capacity increase upon DNA melting and (2) enthalpies/entropies of the base pairs that differ considerably from those in the published data tables. When it was thought that the thermodynamic characteristics of the base pairs were independent of temperature, it sufficed to add up enthalpies and entropies for all adjacencies in a duplex, data taken from a single table applicable at all melting temperatures. It is now apparent that account must be taken of the positive increment in ∆Cp on melting and the enthalpies/entropies corrected. An alternative and iterative method of calculating T_m values is proposed in Refs. [13,19], where the matter is explained and discussed in detail.

4. Thermodynamics of Binding Proteins to DNA

A different approach to studying the hydration of DNA is to investigate the thermodynamics of its displacement when protein domains (DNA binding domains, DBDs) bind to defined sites in the major or the minor grooves. Frequently, DNA binding domains insert component elements into both grooves, so such conformationally complex DBDs have been excluded from consideration or have had sub-domain tails removed so that binding is to just one groove, thereby simplifying the analysis. Figure 5 presents data for separate groups of minor groove and major groove binding DBDs. So, what are the general thermodynamic characteristics of DBD binding to DNA? The Gibbs free energies (in red) average about 50 kJ/mol in both grooves, i.e., no obvious preference is shown for a particular groove. However, an important difference is that whereas binding to the major groove (most frequently the insertion of an α-helix) is enthalpy-driven (negative ∆H, in blue), binding to the minor groove is entropy-driven (positive T∆S, in green), particularly at AT-rich regions. Additionally, it can be seen that minor groove binding occurs despite the somewhat positive enthalpy of the process. The enhanced positive entropy factors for minor groove binding cannot be conformationally based: minor groove complexes are not more loose and flexible than major groove complexes. This large positive entropy can only come from displacing ordered water from the minor groove. If the hydrating water molecules in the major groove are more randomly oriented, there is less entropy to be gained from their displacement—but in place of this, DNA binding is driven by the much more extensive van der Walls contacts that give rise to a substantial negative enthalpy. The energetics of DBD binding to DNA provide clear evidence for the presence of ordered water in the minor groove but not in the major groove.

5. Heat Capacity Changes upon DBD Association with DNA

If the thermodynamic characteristics of protein binding to the two grooves of DNA differ in terms of their unequal states of hydration, this should be reflected in the heat capacity changes induced by protein binding to the two grooves. Heat capacity changes are a good proxy for hydration changes since conformational modifications are too insubstantial to have much influence on the heat capacity, but reductions in hydration have a large effect on the heat capacity change that accompanies its release on DBD binding. Dehydration characteristics on DBD binding to the two grooves of DNA have been studied using the same formalism adopted for protein folding/denaturation. An equation was established for the observed magnitude of ∆Cp for protein folding in terms of the loss of accessible surface area (∆ASA in Ų), both apolar and polar in character [21]:

ΔCp^protein = −1.79 × ΔASA^apolar + 0.98 × ΔASA^polar

(1)

in which the ∆Cp coefficients are expressed in J K⁻¹ mol⁻¹ (Ų)⁻¹ and ΔASA are in Ų.

This equation quantifies the reduction in heat capacity resulting from dehydrating apolar surface and the increase in heat capacity upon dehydrating the polar surface of proteins, for which the former effect dominates to give a substantial overall negative value of ∆Cp. When applied to DNA folding (as detailed above), it explains why ∆Cp is still negative—as for proteins—but lesser in magnitude as a result of the greater polar surface buried.

For a particular DBD/DNA interaction (major or minor groove), this equation can be used to calculate the contribution of protein dehydration to the observed (overall) ∆Cp and then, by difference, obtain the contribution from dehydration of the DNA surface. It can be seen from Figure 6 that the contribution to ∆Cp from the protein components (in orange) is fairly similar in the two grooves, but this is not the case for the component from the dehydrating DNA surface. In the major groove, there is a reduction of about −1 J K⁻¹ mol⁻¹ (Ų)⁻¹ in the heat capacity, but this drops to about −0.3 J K⁻¹ mol⁻¹ (Ų)⁻¹ in the minor groove. To deconvolute the DNA contributions into their polar and apolar components, equations of the above type are established with the two coefficients as unknowns. With several such equations for different DBDs, they can be solved simultaneously for each groove. The resulting data [22] show that the apolar coefficients in the two grooves are similar at about −3 J K⁻¹ mol⁻¹ (Ų)⁻¹, i.e., significantly negative and not so different from that derived for proteins. Regarding the dehydration of the polar surface, the two grooves are very different from each other. In the major groove, the polar coefficient is +0.38 J K⁻¹ mol⁻¹ (Ų)⁻¹, which is positive as with proteins, albeit of lower magnitude than the +0.98 J K⁻¹ mol⁻¹ (Ų)⁻¹ given in Equation (1), in reflection of a less polar state in the major groove. The minor groove is dramatically different: the polar coefficient is +2.67 J K⁻¹ mol⁻¹ (Ų)⁻¹, i.e., positive (as expected), but much more positive than one would predict from observation of the minor groove surface. Such a large positive value can only be a consequence of displacing well-ordered water from the polar groups of the minor groove (N₃ of A and O₂ of T bases) [22]. It is worth recalling here that the heat capacity of ice is about half that of bulk liquid water.

The several lines of evidence for ordered water in the minor groove make it clear why the overall ∆Cp for melting DNA, while slightly positive (see Figure 3), is of much lower magnitude than the large increase in Cp for melting proteins. The very substantial and dominant positive contribution to ∆Cp from the hydration of the hydrophobic surface of polypeptides on melting is opposed by only a small negative component from the hydration of polar groups. However, with DNA, the loss of ordered water from the minor groove on melting generates a very substantial negative effect on ∆Cp—a situation not occurring in proteins [13].

6. DNA Bending

Large bends in DNA are most easily generated by protein binding to the minor groove at AT-rich regions, e.g., by HMG boxes or TATA box binding protein (TBP); see Figure 5 (in yellow). Such bending does not appear to demand any extra free energy in that the largest bends (induced by wedge insertion) are found to have the highest affinity: ∆G_nel becomes increasingly negative as the bend angle increases. The characteristic stiffness of the duplex—as seen in its long persistence length of about 45 nm [23], seems to have been eliminated. A reasonable, although partial, explanation for the loss in rigidity is that the stiffness is maintained by the rigid array of water molecules in the minor groove associated with AT pairs. The energetics of DNA bending are discussed in more detail in Refs. [12,24].

7. The Role of Hydration in Enthalpy/Entropy Compensation (EEC)

Modifications to interacting systems frequently lead to compensating alterations in both the enthalpy and entropy of the process, i.e., the Gibbs free energy is barely altered despite large compensating changes occurring to the enthalpy and entropy of the process. This situation is frequently observed when changes are made to ligands that bind to proteins or to DNA. Such enthalpy–entropy compensation (EEC) for a ligand of increased affinity is normally assumed to be a consequence of forming tighter van der Waals contacts to the substrate—contacts that give rise to a more negative enthalpy. However, the additional molecular constraints imposed on both the ligand and the substrate result in a reduction in the conformational entropy, i.e., a more negative ∆S, and this compensates for the more negative enthalpy.

EEC is a widely observed phenomenon, i.e., it appears to be an intrinsic property covering many types of non-covalent systems. In contrast to conformational explanations of EEC, changes in the solvation of a system can also contribute to EEC and frequently dominate the measured enthalpy/entropy components. The contribution to EEC from modified hydration is frequently ignored, and elaborate hypothetical structural explanations offered for the observed compensating enthalpy/entropy changes.

A particularly revealing example of solvation-based enthalpy–entropy compensation is the yeast bZIP DBD from GCN4, binding as a crosslinked homodimer in a scissors grip to target DNA elements of a slightly different sequence, AP-1 and ATF/CREB, Refs. [25,26]. These targets differ in sequence by just one base pair. However, the crystal structures of the complexes with the two DNAs show them to be very similar [27,28]: the two α-helix backbones that contact the DNA major groove overlap each other with an RMSD of only 1.3 Å. The only notable difference between the two structures is the changed interaction of a single conserved Arg sidechain (R243) that binds close to the center of the target DNA, a conformational difference that is quite small and insufficient to explain an enthalpy difference of 52 kJ/mol and a non-electrostatic entropy factor (T∆S) difference of 48 kJ/mol (see Figure 7).

The most reasonable explanation for these very large discrepancies in the entropies and enthalpies of forming the two very similar GCN4 complexes is differences in the number of incorporated ordered water molecules. If we approximate the immobilization/release of water molecules as similar to that of freezing/melting water, it follows that the AP-1 complex has seven or eight more incorporated water molecules than the ATF/CREB complex [25]. It is worth recalling that the binding/release of dynamically constrained water from macromolecular systems is intrinsically compensatory: if such water has an ice-like structure and the temperature is 273 K, there would be no consequent change in the Gibbs energy despite large changes in the enthalpy and entropy. Solvation changes, therefore, represent the principal energetic basis of enthalpy–entropy compensation [26].

8. Conclusions

The solution methods described provide good support for concluding that the distribution of hydrating water molecules evidenced in crystallographic structures is also present in solution. The major groove surface is of mixed polar and apolar nature, leading to a hydration pattern with characteristics similar to those of polypeptide chains and exhibiting a net reduction in heat capacity on its removal by binding proteins. In the much narrower minor groove, particularly at AT-rich sequences, the hydrating waters are bound to the polar groups of the bases in a highly ordered ice-like fashion, providing stiffness to the structure and characterized by a substantial heat capacity increase upon their removal—either as a consequence of binding DBDs or upon stand separation when melting the duplex.