16046
J. Phys. Chem. B 2009, 113, 16046–16058
Noncovalent Interactions Involving Histidine: The Effect of Charge on π-π Stacking and T-Shaped Interactions with the DNA Nucleobases Cassandra D. M. Churchill and Stacey D. Wetmore* Department of Chemistry and Biochemistry, UniVersity of Lethbridge, 4401 UniVersity DriVe, Lethbridge, Alberta, T1K 3M4 Canada ReceiVed: August 14, 2009; ReVised Manuscript ReceiVed: October 7, 2009
Detailed (gas-phase) MP2/6-31G*(0.25) potential energy surface scans and CCSD(T) energy calculations at the complete basis set (CBS) limit were used to analyze the (face-to-face) stacking and (edge-to-face) T-shaped interactions between histidine (modeled as imidazole) and the DNA nucleobases. For the first time, a variety of relative monomer arrangements between both neutral and protonated histidine and the natural nucleobases were considered to determine the effects of charge on the optimum dimer geometry and binding strength. Our results reveal that protonation of histidine changes the preferred relative orientations of the monomers and propose that these geometric differences may be combined with experimental crystal structures to assess the protonation state of histidine in different environments. It is also found that protonation affects the nucleobase binding preference, as well as the magnitude of the stacking and T-shaped interactions. Indeed, the maximum possible stacking and T-shaped interactions involving the neutral histidine range between approximately 20 and 45 kJ mol-1, while this range increases to 40-105 kJ mol-1 upon protonation, which represents an up to 330% enhancement. Although an increase in the interaction energies upon protonation of histidine is expected, the present work provides a measure of the magnitude of this enhancement in the gas phase and reveals that the amplification is almost entirely due to larger electrostatic contributions. The relative strengthening of different classifications of dimers upon protonation leads to stronger T-shaped interactions than stacking energies for protonated histidine, while the stacking and T-shaped interactions involving neutral histidine are of comparable magnitude. Thus, there is a significant difference in the nature of the πcation-π interactions involving protonated histidine and the π-π interactions involving neutral histidine. The calculated strengths of the interactions studied in the present work suggest that both neutral and cationic histidine contacts will provide significant stabilization to DNA-protein complexes. Although solvation effects will decrease the magnitude of the reported interactions, our results are applicable to a variety of low-polarity, biologically-relevant environments such as nonpolar enzyme active sites. Therefore, our calculations suggest that these interactions may also be important for many biological processes. The proposed significance of these interactions is supported by the large number of histidine-nucleobase contacts that appear in experimental crystal structures. The highly accurate (MP2/6-31G*(0.25)) preferred structures and (CCSD(T)/CBS) binding strengths reported in the present work can be used as benchmarks to analyze the performance of existing, or to develop new, molecular mechanics force fields for use in large-scale molecular dynamics (MD) studies of DNA-protein complexes. Introduction As we gain a greater understanding of the chemistry of biological processes, it is clear that a variety of interactions occur between the molecular components of life. It is also becoming apparent that these interactions are responsible for essential biological processes. Although strong forces, such as ionic and covalent bonds, have long been accepted to govern the chemistry of biological systems, weaker noncovalent interactions have more recently been identified to play important roles in the assembly, structure, and stability of biomolecules.1,2 For example, hydrogen bonding and stacking between the nucleobases are responsible for the formation and stability of DNA double helices.3,4 Furthermore, noncovalent contacts can preside over the interactions between biological macromolecules.5 For example, interactions between DNA and protein components contribute to the strength and specificity associated with DNA-protein binding.5 * Corresponding author. E-mail:
[email protected].
Essential noncovalent interactions between biomolecules include hydrogen-bonding, hydrophobic, van der Waals, charge transfer, X-H · · · π, and π-π interactions.1,2 Among this list, associations between two π systems are perhaps the least understood. Although conjugated molecules may interact in a variety of geometries, the structures generally fall into two broad categories: stacked (face-to-face), where molecular π clouds are parallel, and T-shaped (edge-to-face), where molecular π clouds are perpendicular. While both stacking and T-shaped interactions involving DNA and/or protein components are commonly found in nature,6 relatively little is known about the strength of these contacts compared to other noncovalent interactions. Despite difficulties studying individual π-π interactions experimentally, computational chemistry has provided a wealth of useful information about contacts between conjugated biological molecules. For example, calculations have been used to characterize stacking interactions between the DNA nucleobases (Figure 1),3,7-22 as well as interactions between the natural nucleobases and other ring systems designed for biochemical
10.1021/jp907887y CCC: $40.75 2009 American Chemical Society Published on Web 11/11/2009
Noncovalent Interactions Involving Histidine
Figure 1. Structure of the natural DNA nucleobases (adenine (A), cytosine (C), guanine (G), and thymine (T)), as well as neutral (HIS) and protonated (HIS+) histidine. Truncated computational models used in the present work are generated by replacing all atoms highlighted in blue with a single hydrogen atom.
J. Phys. Chem. B, Vol. 113, No. 49, 2009 16047 interactions between the π-containing components of DNA and proteins. First, its unique pKa allows histidine to be neutral or protonated under biological conditions.55 Second, a large number of histidine-nucleobase contacts have been identified in experimental crystal structures.6,45 Due to the importance of protonated histidine in nature, the interactions between protonated histidine (modeled as the imidazolium cation) and other aromatic amino acids have been investigated.45,54 However, studies on the interactions between protonated histidine and DNA components are limited. Although Rooman et al.45 compare the strength of stacking interactions between protonated or neutral histidine and adenine, only select geometries were considered. Therefore, a full understanding of the interactions between histidine and adenine, as well as the other DNA nucleobases, is still missing. The present work systematically examines both the stacking and T-shaped interactions in heterodimers between all natural DNA nucleobases and protonated histidine (Figure 1). The geometry and strength of these interactions will be directly compared with data for the corresponding neutral histidine heterodimers52 to unveil the effects of cationic charge on these noncovalent interactions. In addition to revealing vital information about interactions between central biomolecules, this work will provide insight on the fundamental differences and similarities between π-π and πcation-π interactions. Computational Details
or medical applications.23-32 Calculations have also been used to gain a greater appreciation of the strength of X-H · · · π interactions between biomolecules by studying T-shaped heterodimers composed of small molecules and the nucleobases.33 Similarly, π-π interactions between the four aromatic amino acids (phenylalanine, tyrosine, tryptophan, histidine),34-38 as well as their interactions with small molecules,33,39-43 have been studied. More recently, π-π stacking and T-shaped interactions between conjugated DNA and protein building blocks were analyzed.44-53 All of these studies have proven that both stacking and T-shaped interactions between biomolecules are stronger than initially anticipated and therefore must be taken into account when developing a complete appreciation of the structure and/or function of biological macromolecules.4 The computational studies mentioned above shed light on the preferred geometries and magnitude of binding between two neutral, conjugated biomolecules. However, relatively little research has focused on understanding the effect of charge on stacking and T-shaped interactions, and these effects cannot be easily predicted. In our group, the interactions between (neutral) natural or (cationic) damaged DNA nucleobases and the (neutral) aromatic amino acids were investigated in attempts to reveal how DNA repair enzymes selectively identify alkylated nucleobases over their neutral counterparts.46,48,49 This work showed that the cationic charge on the nucleobase significantly affects both the geometry and strength of DNA-protein stacking and T-shaped interactions and thereby suggests that differences between neutral and cationic π-π interactions can play an integral role in the function of DNA repair enzymes.46,48,49 Indeed, a recent publication emphasizes that a distinction must be made between π-π and πcation-π interactions, which are rarely studied despite being distinct in nature and significantly stronger than the (neutral) π-π contacts.54 Although the effect of charge on DNA-protein binding due to nucleobase mutations is of great interest, charge may also be present in the amino acid. Among the amino acids, histidine provides an excellent case study for the effect of charge on
All monomers were optimized in a planar (Cs) geometry at the MP2/6-31G(d) level of theory. The DNA nucleosides were modeled by replacing the deoxyribose sugar moiety (highlighted in blue in Figure 1) with a hydrogen atom. Similarly, histidine was modeled as imidazole, where the protein backbone in the biomolecule is replaced with a hydrogen atom at the β carbon (highlighted in blue in Figure 1). We note that these truncations are commonly used in the literature.44-52 Furthermore, previous work in our group supports their use since the magnitude of the calculated π-π component of the interaction energy between DNA nucleobases and (conjugated) amino acids is unchanged by the inclusion of the biological backbone.53 The counterpoisecorrected, gas-phase potential energy surfaces for all dimers were scanned using MP2 single-point calculations with the 6-31G*(0.25) basis set, which replaces the standard d-exponent for second row atoms (0.80) with 0.25.9,12,14 The MP2/631G*(0.25) combination has previously been shown to yield both stacking and T-shaped interactions between the natural nucleobases or (cationic) 3-methyladenine and aromatic amino acids comparable to CCSD(T) results estimated at the complete basis set (CBS) limit.49,52 For example, the MP2/6-31G*(0.25) and CCSD(T)/CBS adenine-histidine stacking interactions are within 0.6 kJ mol-1, while the corresponding T-shaped interactions are within 1 kJ mol-1.52 Similarly, MP2/6-31G*(0.25) recovers 107% of the protonated histidine-adenine stacking interaction and 98-104% of the corresponding T-shaped interactions.56 Since the 6-31G*(0.25) basis set was specially designed to produce accurate stacking interactions9,12,14 and relies on a cancellation of higher-order correlation effects and basis set incompleteness errors, the good agreement between MP2/ 6-31G*(0.25) and CCSD(T)/CBS T-shaped interactions is somewhat surprising. Nevertheless, the most cost-effective method was implemented in the present work due to the large number of data points considered on the potential energy surfaces of these dimers (see below) and its success for the benchmark test cases mentioned above. The MP2/6-31G*(0.25) potential energy surface scans were conducted by systematically
16048
J. Phys. Chem. B, Vol. 113, No. 49, 2009
Churchill and Wetmore
Figure 2. Definition of the geometric variables considered and increments used in (a) stacking potential energy surface scans and (b) T-shaped potential energy surface scans (angle of “edge” rotation (θ), vertical separation (R1), angle of rotation (R), and horizontal displacement (R2)). See Figure 3 for θ edges considered.
changing the relative monomer orientations as performed in our previous publications46-49,52,53 and outlined in the following subsections. Stacking Interactions. To generate stacked (face-to-face) histidine-nucleobase heterodimers, the monomers were aligned according to their centers of mass. Initial relative orientations of the monomers were defined such that the model glycosidic bond (N1-H in the pyrimdines and N9-H in the purines) and the model histidine backbone were aligned in parallel. Two relative orientations of the monomer molecular planes were considered in neutral histidine dimers, where the first is defined by stacking histidine and the nucleobases in the orientation shown in Figure 1, while the second, indicated with a prime (′) symbol, is obtained by mirror flipping histidine in Figure 1 prior to stacking with the nucleobases. Due to symmetry, only one relative monomer orientation was considered in protonated histidine heterodimers. From these initial structures, three variables were altered: the vertical separation (R1), angle of rotation (R), and the horizontal displacement (R2), where the definition of each variable is provided in Figure 2a. First, R1 was varied in 0.1 Å increments and then held constant at the preferred value for the remaining calculations. Next, R was rotated in 30° increments in a right-hand sense about the axis that passes through the centers of mass and is perpendicular to the molecular planes. Using the optimal R1 and R values, R2 was considered, where one monomer was shifted across the face of the other monomer in 0.5 Å increments across a grid. T-Shaped Interactions. T-shaped (edge-to-face) interactions between nucleobase edges and the histidine face, as well as histidine edges and the nucleobase faces, were considered. The edges (θ, Figure 2b) were chosen such that either an atom (indicated by a number in our nomenclature) or a bond (indicated by a letter in our nomenclature) in one monomer is directed toward the π system of the second monomer, where all edges considered are defined in Figure 3. Initial structures for histidine-edge heterodimers were generated by aligning the centers of mass of histidine and the nucleobase. Initial structures for the nucleobase edge heterodimers were generated by aligning the nucleobase edge atom under consideration and the center of mass of histidine or the center of the nucleobase bond under consideration and the center of mass of histidine. For each T-shaped heterodimer, the potential energy surface was sequentially scanned as a function of R1, R, and R2 as discussed for the stacked structures. It should be noted that there are slight differences in the definition of these variables for T-shaped interactions due to differences in the starting structure for the scans. For example, in nucleobase-edge dimers, R1 is the
Figure 3. Nucleobase and histidine edges (θ, Figure 2) considered in T-shaped heterodimers, where in our nomenclature a number indicates an atom is directed toward the monomer π face and a letter indicates a bridged structure involving more than one atom directed toward the π face.
distance between the amino acid center of mass and a nucleobase atom (numbered edges) or the midpoint of the line connecting two nucleobase atoms (lettered edges). For heterodimers involving protonated histidine edges, adjustments to the scanning procedure were made such that R and R2 scans were performed in succession until convergence in these variables, as well as the binding energy, was reached. Due to the resulting large number of points considered on the potential energy surface of protonated histidine-edge dimers, as well as the molecular symmetry of protonated histidine, a smaller subset of edges were considered for protonated histidine. Additional details and justification for changes in the scanning methodology for protonated histidine T-shaped dimers will be provided in the Results and Discussion section. For any given dimer pair, our detailed potential energy surface scans involved up to 147 calculations for stacked monomers, 3617 for T-shaped interactions involving a nucleobase edge, and 1925 for T-shaped interactions involving a histidine edge. All calculated energies are available in the Supporting Information. However, due to the large data set generated in our work, only dimer orientations that lead to the strongest interaction energies are reported in the tables and discussed in the text. All calculations were performed with Gaussian 03.57 Results and Discussion The present study compares the π-π interactions between neutral or protonated (cationic) histidine (modeled as imidazole) and the natural DNA nucleobases. A variety of different relative
Noncovalent Interactions Involving Histidine
J. Phys. Chem. B, Vol. 113, No. 49, 2009 16049
TABLE 1: Strongest MP2/6-31G*(0.25) Stacking Interactions between Neutral or Protonated Histidine and the DNA Nucleobases as Determined by Varying R1 (Å), r (°), and R2 ((X-displacement, Y-displacement), Å)a,b dimer
R1
∆E
R
∆E
R2
∆E
c
A:HIS C:HISc G:HISc T:HIS′c
3.3 3.5 3.3 3.4
-18.1 -7.9 -23.0 -18.8
180 150 300 120
-28.3 -24.2 -32.4 -25.5
(0, -1.0) (-0.5, -1.5) (1.5, 1.0) (-1.0, 0)
-29.7 -26.9 -35.3 -26.8
A:HIS+ C:HIS+ G:HIS+ T:HIS+
3.2 3.3 3.2 3.4
-54.0 -44.2 -54.6 -23.0
270 330 30 270
-58.4 -44.8 -57.0 -27.0
(0, 0) (2.0, -1.0) (-1.5, 2.0) (1.0, 0.5)
-58.4 -59.9 -69.0 -35.1
a See Figure 1 for structures of amino acids and nucleobases and Figure 2 for definitions of R1, R, and R2. b The initial relative orientation of the monomers was obtained by directly overlaying the structures in Figure 1. The prime (′) indicates that histidine was mirror flipped (with respect to Figure 1) prior to stacking with the nucleobase. c Ref 52.
monomer orientations will be considered including stacked and T-shaped structures involving either a nucleobase or a histidine edge. An emphasis will be placed on understanding the effect of charge on the preferred geometries and the magnitude of the binding energies in the histidine heterodimers. Our analysis will be conducted using data from detailed potential energy surface scans as outlined in the Computational Details. Since a large number of data points were considered on the potential energy surface for each dimer, we focus on the major conclusions and general trends and use select examples to illustrate our most important findings. The entire data set provided in the Supporting Information provides a more complete understanding of any given dimer. The results for each type of interaction will be discussed separately below. Stacking Interactions. As mentioned in the Computational Details, stacking interactions between the DNA nucleobases and neutral or protonated histidine were calculated as a function of three variables (R1, R, and R2, Figure 2a). Therefore, the effects of each geometric variable on the magnitude of the stacking interaction will be analyzed in the present section, where the maximum binding energies for the most favorable stacking orientation of each heterodimer with respect to each variable are summarized in Table 1. Only the monomer orientations leading to the strongest (most negative) stacking interaction are reported in the table and discussed in the text. The stacking energy sharply decreases in magnitude as a function of vertical separation (R1) at short distances due to increased repulsion and decreases in magnitude more slowly at longer distances due to decreased dispersion-attraction forces (see Supporting Information). However, when relatively small changes to the preferred R1 value that leads to the strongest binding energy are considered, the interaction energies are found to be weakly dependent upon the vertical separation. This is not surprising since the potential energy surfaces associated with stacked complexes are well-known to be shallow with respect to this variable. The preferred R1 value in nucleobase-histidine heterodimers tends to be smaller for the purines than the pyrimidines due to increased dispersion, where smaller vertical separations have been previously reported for purine heterodimers compared with pyrimidine heterodimers.31,58 However, the differences are relatively small (
