Entropic Enhancement of Protein-DNA Affinity by Oxygen-to-Sulfur Substitution in DNA Phosphate

Abstract

Dithioation of DNA phosphate is known to enhance binding affinities, at least for some proteins. We mechanistically characterized this phenomenon for the Antennapedia homeodomain-DNA complex by integrated use of fluorescence, isothermal titration calorimetry, NMR spectroscopy, and x-ray crystallography. By fluorescence and isothermal titration calorimetry, we found that this affinity enhancement is entropy driven. By NMR, we investigated the ionic hydrogen bonds and internal motions of lysine side-chain NH₃⁺ groups involved in ion pairs with DNA. By x-ray crystallography, we compared the structures of the complexes with and without dithioation of the phosphate. Our NMR and x-ray data show that the lysine side chain in contact with the DNA phosphate becomes more dynamic upon dithioation. Our thermodynamic, structural, and dynamic investigations collectively suggest that the affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate is largely due to an entropic gain arising from mobilization of the intermolecular ion pair at the protein-DNA interface.

Introduction

Chemically modified oligonucleotides have drawn considerable interest as potential therapeutic reagents (1–3). One type of such modifications is the thioation of phosphate groups, which replace one or two nonbridging oxygens with sulfur atoms. It is known that mono- or dithioation of oligonucleotides increases resistance to nucleases and improves cell penetration properties (4–6). It was recently discovered that phosphorothioate is also naturally present in some bacterial genomes (7,8). Thioation of the phosphate retains the overall charge and similar tetrahedral covalent geometry of DNA phosphate. Interestingly, compared to unmodified DNA, thioated DNA often exhibits stronger binding affinity, at least for some proteins and in some positions (9–13). Due to these properties, short DNA duplexes containing phosphoromonothioate or phosphorodithioate groups can effectively serve as decoy molecules that inhibit particular transcription factors involved in pathogenesis (14–16).

From a physicochemical point of view, however, the protein-DNA affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate may appear counterintuitive, especially given the following two facts. First, sulfur atoms in organic compounds tend to serve as relatively poor hydrogen bond acceptors compared with oxygen atoms. For example, the boiling point of methanethiol (CH₃SH) is lower than that of methanol (CH₃OH) by 59°C; and the boiling point of 2-mercaptoethanol (HO-CH₂-CH₂-SH) is lower than that of ethylene glycol (HO-CH₂-CH₂-OH) by 39°C (17). Second, the electronegativity of the sulfur atom is weaker than that of the oxygen atom (2.58 vs. 3.44 by Pauling scale) (17). Despite these characteristics of sulfur atoms, how can the oxygen-to-sulfur substitution in DNA phosphate enhance protein-DNA association?

Recently, we gained important insight into this question. In our previous NMR studies of the HoxD9 homeodomain-DNA complexes (9), we found that the mobility of the Lys side-chain NH₃⁺ group is enhanced upon the oxygen-to-sulfur substitution of the DNA phosphate group, which forms an intermolecular ion pair at the molecular interface. The entropic impact of this mobilization on the binding free energy was estimated from the changes in NMR order parameters and bond-rotation correlation times. The data suggested that the mobilization of the intermolecular ion pair can at least partially account for affinity enhancement by the oxygen-to-sulfur substitution of DNA phosphate. However, this was indecisive because neither structural details around the ion pairs nor thermodynamic (i.e., enthalpic and entropic) data on binding was available for the HoxD9-DNA complexes.

In the current work, we resolve this issue and further examine the role of ion-pair dynamics in affinity enhancement by the oxygen-to-sulfur substitution of DNA phosphate. For this purpose, using fluorescence spectroscopy, isothermal titration calorimetry (ITC), NMR spectroscopy, and x-ray crystallography, we characterize the sequence-specific interactions of the Antennapedia (Antp) homeodomain with unmodified and dithioated DNA. The Antp homeodomain is practically more useful than the HoxD9 homeodomain due to higher solubility and stability of the free state under physiological conditions. Furthermore, in previous studies by other research groups, the Antp homeodomain has been extensively characterized by biochemical methods (18,19) as well as by biophysical methods such as ITC (20), NMR (21–23), and x-ray crystallography (24). Using this well-suited system, we investigate how the oxygen-to-sulfur substitution in a DNA phosphate group influences protein-DNA association in terms of thermodynamics, internal motions, and structure.

Materials and Methods

Protein preparation

A synthetic gene encoding the 60 amino-acid residues of the fruit fly Antp homeodomain (RKRGRQTYTRYQTLELEKEFHFNRYLTRRRRIEIAHALSLTERQIKIWFQNRRMKWKKEN) with C39S mutation (24) was subcloned into the NdeI/ HindIII sites of the pET-49b vector (EMD Millipore, Billerica, MA). Escherichia coli strain BL21(DE3) was transformed with this plasmid and cultured at 37°C in 4 L of M9 minimal media containing kanamycin (30 μg/ml) as well as ammonium chloride and glucose as the sole nitrogen and carbon sources, respectively. At OD₆₀₀ ≈ 0.8, protein expression was induced by adding isopropyl β-D-thiogalactopyranoside (0.4 mM) to the media, and the E. coli culture was continued at 18°C for an additional 16 h. The cells were harvested and disrupted by sonication in a buffer containing 20 mM Tris•HCl (pH 7.5), 1 mM EDTA, 500 mM NaCl, 2 mM DTT, 5% glycerol and a protease inhibitor cocktail (Roche, South San Francisco, CA). The supernatant of the lysate was loaded onto a SP-FF column (GE Healthcare) equilibrated with 50 mM Phosphate buffer (pH 7.5) and 500 mM NaCl, and the protein was eluted with a gradient of 500–1500 mM NaCl. Fractions containing the Antp homeodomain were pooled and concentrated to ∼10 ml and then loaded onto a S100 size-exclusion column (GE Healthcare, Pittsburgh, PA) equilibrated with a buffer of 50 mM Tris•HCl (pH 7.5), 1 mM EDTA, and 400 mM NaCl. The protein solution was loaded onto a Resource-S cation-exchange column (GE Healthcare) equilibrated with a buffer of 50 mM Tris•HCl (pH 7.5), 1 mM EDTA, and 600 mM NaCl, and then eluted with a gradient of 600–1400 mM NaCl. Purity of the protein was confirmed to be >95%. The protein was quantified using ultraviolet absorbance at 280 nm together with an extinction coefficient of 15,470 M⁻¹ cm⁻¹.

DNA preparation

The sequences of DNA strands used in this work are shown in the figures together with the data. The DNA strand containing a phosphorodithioate group was synthesized on an ABI Expedite 8909 DNA synthesizer with standard dA/dC/dG/dT-phosphoramidites and dC-thiophosphoramidite (AM Biotechnologies (Houston, TX)/Glen Research (Sterling, VA)), and purified via anion-exchange chromatography using a Mono-Q column as described (25). All the other DNA strands were purchased from Integrated DNA Technologies (Coralville, IA) and purified by anion-exchange chromatography. For preparation of double-stranded DNA, complementary strands were annealed and minor single-stranded DNA excess due to the uncertainty in measuring single strand concentrations was removed by anion-exchange chromatography.

Fluorescence-based affinity measurements

The affinities of the Antp homeodomain for the unmodified and dithioated target DNA duplexes were determined using two different methods. One is based on fluorescence anisotropy as a function of protein concentration (0.1–500 nM). Fluorescence arising from tetramethylrhodamine (TAMRA) attached to the 3′–terminus of DNA (3.3 nM) was measured using an ISS PC-1 spectrofluorometer (Champaign, IL). Excitation and emission wavelengths used were 533 and 580 nm, respectively. The titration experiments were performed at 25°C using a buffer of 10 mM sodium phosphate (pH 5.8) and 150 mM NaCl. The dissociation constant K_d was calculated from the anisotropy data via nonlinear least squares fitting with

$$A_{obs}=A_{free}+\left(A_{bound}-A_{free}\right)\left(P+D+K_d-\sqrt{{\left(P+D+K_d\right)}^2-4PD}\right)/\left(2D\right),$$ (1)

where A_obs is the observed anisotropy; A_bound and A_free are those of protein-bound DNA and free DNA; and P and D are total concentrations of the protein and the probe DNA, respectively. For each DNA, the affinity measurements were repeated three times.

We also measured K_d values using a fluorescence-based competition assay. In this assay, solutions of 10 nM 3′-TAMRA-labeled target DNA, 50 nM Antp homeodomain, and unlabeled competitor DNA (0.5–6400 nM) were made and fluorescence anisotropy was measured as a function of the competitor concentration. We used the following equation for analysis of the competition assay data:

$$r=\frac{C{K}_{d,p}+K_{d,p}{K}_{d,c}-P{K}_{d,p}+2P{K}_{d,c}-K_{d,p}\sqrt{{\left(C+K_{d,c}-P\right)}^2+4P{K}_{d,c}}}{2\left\{C{K}_{d,p}-\left(K_{d,p}-K_{d,c}\right)\left(K_{d,p}+P\right)\right\}},$$ (2)

$$A_{obs}=\left(1-r\right)A_{free}+r{A}_{bound},$$ (3)

where K_d,c and K_d,p are the dissociation constants for the competitor and probe DNA duplexes, respectively; and C is the concentration of competitor DNA. When a value of either K_d,c or K_d,p is known, the other dissociation constant can be determined from the competition assay data via nonlinear least squares fitting. Eq. 2 assumes D << P. When P << C is also satisfied (e.g., for the data for the nonspecific competitor DNA), Eq. 2 becomes equivalent to a popular form based on binding polynomial (26):

$$r=\frac{P/K_{d,p}}{1+C/K_{d,c}+P/K_{d,p}}.$$ (4)

The dissociation constant K_d was calculated from the fluorescence anisotropy data via nonlinear least squares fitting using MATLAB (The MathWorks, Natick, MA).

ITC measurements

Using a MicroCal VP-ITC microcalorimeter (Northampton, MA), the ITC experiments were carried out at 25°C for three different DNA duplexes. Two of these DNA duplexes include the Antp target sequence and one of them is dithioated at the Lys-57 interaction site. The third DNA duplex is nonspecific 15-bp DNA. Each solution of DNA or protein was extensively dialyzed to a buffer of 10 mM sodium phosphate (pH 5.8) and 150 mM NaCl. Each titration experiment consists of one 5 μl and twenty 12 μl injections of Antp homeodomain into the cell, which initially contained a 1.41 ml solution of DNA. Concentrations of the molecular components are given in the figure legends. The interval between the injections was 4 min. For both unmodified and dithioated DNA, the ITC experiment was repeated three times. To account for the heat of protein dilution, a control experiment was also performed using titration of the protein into the buffer under the identical experimental conditions. Heat of the control titration was subtracted from the original titration data. To avoid influence of heat from nonspecific association, data with relatively low concentrations of the Antp homeodomain with DNA in excess were used to measure the binding enthalpies for the specific complex formation. The ITC data were analyzed with the Origin 7.0 software. The affinity data from the fluorescence experiment together with the ITC data were used to determine binding entropy as described previously (20).

NMR spectroscopy

All NMR experiments for the Antp homeodomain-DNA complexes were performed using Bruker Avance III spectrometers (Fällanden, Switzerland) operated at a ¹H frequency of 600, 750, or 800 MHz. A 280 μl solution of 0.8 mM Antp homeodomain (¹⁵N- or ¹³C/¹⁵N-labeled) and 1.2 mM DNA in a buffer of 10 mM sodium phosphate (pH 5.8) and 20 mM NaCl was sealed into the inner tube (outer diameter 4.1 mm) of a 5-mm coaxial NMR tube (Shigemi, Allison Park, PA). D₂O for NMR lock was separately sealed in the outer layer of the coaxial tube to avoid deuteration of NH₃⁺ groups (i.e., NDH₂⁺ and ND₂H⁺ species) (27). Backbone ¹H/¹³C/¹⁵N resonances were assigned using 2D HSQC spectra and 3D HNCO, HN(CA)CO, HNCA, HN(CO)CA, CBCA(CO)NH, HNCACB, and HBHA(CO)NH spectra (28). Side-chain ¹H/¹³C resonances were assigned using 3D H(CCO)NH, C(CO)NH, HCCH-TOCSY, and HCCH-COSY spectra (28). Lys side-chain NH₃^{+ 1}H/¹⁵N resonances were assigned using Lys-selective 2D HISQC (29), (H2C)N(CC)H-TOCSY (30), and H2(C)N (31) spectra and 3D H3NCECD (29), HDHE(CDCE)NH3 (29), and H3NCG (30) spectra. Scalar coupling between lysine side-chain ¹⁵N and DNA phosphate/phosphorodithioate ³¹P nuclei across hydrogen bonds (^h3J_NP) was analyzed by 2D H3(N)P and spin-echo ^h3J_NP-modulation difference constant-time HISQC experiments with a Bruker cryogenic QCI ¹H/¹³C/¹⁵N/³¹P probe at the ¹H-frequency of 600 MHz as described previously (9). Backbone ¹⁵N longitudinal and transverse relaxation rates (denoted by R₁ and R₂, respectively) were measured at 25°C, from which the molecular rotational correlation time and anisotropy of rotational diffusion of the Antp homeodomain-DNA complexes were determined. For Lys side-chain NH₃⁺ groups, ¹⁵N R₁ and R₂ relaxation rates, and heteronuclear NOE were measured at 25°C as described previously (32). For the NH₃⁺ groups, the generalized order parameters S²_axis and the reorientation correlation time τ_i for the symmetry axis, and the bond-rotation correlation time τ_f were determined using Mathematica as described previously (32,33).

X-ray crystallography

The Antp homeodomain-DNA complexes with and without dithioation for DNA phosphate at the Lys-57 interaction sites were crystallized using the conditions described by Fraenkel and Pabo (24). Crystals were grown at 17°C over several days using the sitting-drop vapor diffusion method. For the unmodified DNA complex, x-ray diffraction data were collected on a Rigaku FR-E++DW with an R-AXIS-IV++ image plate detector using Cu radiation from a single crystal. Two hundred 1/2 degree width frames were collected. The diffraction images were processed and scaled using HKL3000, using an I/sigma of 1.0 resolution cutoff criteria. The data for the dithioated DNA complex were collected at the advanced photon source beamline 19ID at a wavelength of 0.97921 Å using a Quantum 315 detector. To collect complete data without overloads two data sets were collected at different resolutions, exposure times, frame widths (1° and 1/2°), and beam attenuation, from the same crystal. Images were processed and scaled using HKL2000, using the CC^∗>0.5 criteria for resolution cutoff. Refinement was performed using Phenix (34), with TLSMD (35) determined TLS parameters, weight optimization, and DNA restraints. The crystallographic phase was determined using the molecular replacement method. Model building and validation was performed in Coot (36). The atomic coordinates of the crystal structures of the unmodified and dithioated complexes have been deposited to Protein Data Bank (PDB accession codes 4XID and 4XIC, respectively).

Results

We compared the two DNA complexes of the Antp homeodomain with 15-bp DNA duplexes with identical base sequences: one with no chemical modification, the other with phosphorodithioate at the Lys-57 interaction site (Fig. 1a). We conducted the same set of biophysical experiments for these two complexes to study the impacts of the oxygen-to-sulfur substitution on protein-DNA association.

Figure 1.

Enhancement of binding affinity for the Antp homeodomain by dithioation of the DNA phosphate at the Lys-57 interaction site. (a) Dithioation of DNA phosphate. (b) Binding isotherm as measured by the fluorescence anisotropy-based titration experiment. Fluorescence anisotropy for TAMRA-labeled 15-bp duplexes (red, unmodified DNA; blue, dithioated DNA) was measured as a function of the concentration of the Antp homeodomain. The probe DNA concentration was 3.3 nM in this experiment. Dissociation constants determined from the three replicates together with Eq. 1 were K_d = 11 ± 1 nM for the unmodified DNA and K_d = 2.9 ± 0.2 nM for the dithioated DNA (c) Competition assay data used to indirectly determine K_d constants. The total concentrations of the probe DNA and the protein were 10 and 50 nM, respectively, in this experiment. The measurement was replicated three times. The fitting calculations using Eqs. 2 and 3 together with the K_d constant for the competitor gave K_d = 15.3 ± 0.8 nM for the unmodified DNA and K_d = 4.0 ± 0.5 nM for the dithioated DNA. The solid lines represent best-fit curves. To see this figure in color, go online.

Affinity enhancement by dithioation of DNA phosphate

Using two different fluorescence-based assays with TAMRA-labeled DNA as a fluorescent probe, we examined the influence of the dithioation on target DNA association of the Antp homeodomain at 150 mM NaCl. By protein titration assays, in which TAMRA fluorescence anisotropy is monitored as a function of the protein concentration, we directly measured the dissociation constants K_d for the complexes with 15-bp DNA duplexes (Fig. 1b). From the protein titration data, we determined values of K_d of the Antp homeodomain to be 11 ± 1 nM for the unmodified DNA and 2.9 ± 0.2 nM for the dithioated DNA.

Because the affinities were close to the limit of the measurable range in the protein titration assay and the probe concentration used was comparable to K_d values, we also analyzed affinities by competition assays that allow for K_d determination for high-affinity systems. These assays involve competitor DNA (15-bp) that contains the same 6-bp target sequence but differs from the probe DNA in the other parts. In this competition assay, the difference in sequence between the probe and competitor DNA is important because otherwise, transfer of the dithioated strand to competitor DNA can occur. We used 50 nM protein and 10 nM probe DNA and varying concentrations of the competitor DNA. The fluorescence anisotropy changes as the unlabeled competitor increasingly outcompetes the probe DNA (Fig. 1c). From these data together with the affinity of the competitor DNA, we determined the K_d values for the unmodified and dithioated complexes to be 15.3 ± 0.8 and 4.0 ± 0.5 nM, respectively. Thus, both data sets indicate that the oxygen-to-sulfur substitution in the DNA phosphate enhances the binding affinity for the Antp homeodomain by a factor of ∼4. From the K_d data, the change in binding free energy ΔΔG upon the oxygen-to-sulfur substitution in the DNA phosphate was calculated to be −0.8 ± 0.1 kcal/mol for the Antp-DNA complexes.

Thermodynamic impact of phosphorodithioate on association

For the thermodynamic characterization, we adopted the approach of Dragan et al. (20) that combines the fluorescence and ITC methods. In this approach, the binding enthalpy ΔH is obtained from the ITC data and the binding free energy ΔG is obtained from the fluorescence anisotropy-based titration data. Because ITC allows for direct observation of the heat from association, ΔH can be measured directly even if the dissociation constant K_d is too small to determine by ITC. For this approach to be valid, however, the fluorescent probe (i.e., TAMRA) should not perturb the binding properties of DNA. We confirmed this by a competition assay using TAMRA-labeled and unlabeled DNA with the identical sequence: the obtained K_d of the unlabeled DNA was virtually the same within experimental uncertainties (Fig. 2a).

Figure 2.

Competition assay data used to assess unlabeled specific DNA (a) and nonspecific DNA (b) duplexes. The total concentrations of the probe DNA and the Antp homeodomain were 10 and 50 nM, respectively. The fitting calculations using Eqs. 2 and 3 together with the K_d constant for the probe gave K_d = 10 ± 2 nM for the specific DNA and K_d = (7.9 ± 0.8) × 10³ nM for the nonspecific DNA. To see this figure in color, go online.

For the 15-bp DNA duplexes with and without the dithioation at the Lys-57 interaction site, we conducted the ITC experiments in which the Antp homeodomain was injected into the DNA solutions to the final molar ratio of 5.3 (Fig. 3a). When the molar ratio is larger than 1, we observed nonmonotonic change of the heat effect over a wide range of the molar ratio, suggesting the presence of weak association events. Judging from the fluorescence-based K_d data, however, it is very unlikely that this effect is due to the specific association with the target site. These ITC data clearly suggest the presence of the multiple-binding sites (37,38), most likely due to nonspecific association of DNA with additional protein molecules (38). In fact, the crystal structures of the Antp homeodomain-DNA complexes suggest that two additional protein molecules can nonspecifically interact with DNA regions that are not covered by the protein bound to the target site. Using the fluorescence-based competition assay, we measured the apparent affinity for a nonspecific 15-bp DNA duplex that does not contain the core recognition sequence for Antp (Fig. 2b). The apparent K_d constant of this nonspecific DNA was (7.9 ± 0.8) × 10³ nM, which is comparable to the sample concentrations used in the ITC experiment. With this affinity, influence of the nonspecific association on the fluorescence data for the specific DNA duplexes (Fig. 1) is virtually negligible because the sample concentrations are lower than the apparent K_d constant for nonspecific complexes. We also conducted the ITC experiment for the same 15-bp nonspecific DNA duplex (Fig. 3b). The ITC data suggest that nonspecific association of the first and second protein molecules generate opposite heat effects. These data suggest that the abnormal ITC profiles at molar ratios higher than 1 (Fig. 3a) are due to nonspecific association.

Figure 3.

Thermodynamic impact of dithioation of DNA phosphate at the Lys-57 interaction site. (a) ITC data for the specific DNA duplexes with 21 injections of the Antp homeodomain, leading to the final molar ratio of 5.3. The DNA concentration was 9 μM. Because the sample concentrations were comparable to the K_d for nonspecific association and the DNA duplexes are long enough to nonspecifically interact with additional protein molecules, heat from the nonspecific interactions was also observed after the target site was saturated. (b) ITC data for the nonspecific DNA duplex with 21 injections of the Antp homeodomain, leading to the final molar ratio of 6.6. The DNA concentration was 8 μM DNA. The heat effect of the first and second nonspecific associations were opposite in sign. (c) ITC data for the specific DNA duplexes with 21 injections of the Antp homeodomain leading to the final molar ratio of 1.8. The DNA concentration was 17 μM. Green dotted lines correspond to the heat effect arising from association of the Antp homeodomain with the target site on the DNA duplexes. (d) Changes in binding free energy (ΔΔG), enthalpy (ΔΔH), and entropic terms (−TΔΔS) upon the oxygen-to-sulfur substitution of the DNA phosphate. The increase in binding entropy ΔΔS was determined to be 2.5 ± 0.6 cal K⁻¹ mol⁻¹. See also Table 1. To see this figure in color, go online.

When the molar ratio is significantly <1, the contribution of the nonspecific association to the heat effect is virtually negligible, because the affinity for the target site is stronger than that for nonspecific DNA by a factor >500. To analyze the binding enthalpy for the specific association, we conducted additional ITC experiments for the specific DNA duplexes using a larger number of injections in a narrower range of molar ratio (up to 1.8) (Fig. 3c). In this second set of ITC data, the heat effects at molar ratio <0.9 were virtually constant. Using these data at low molar ratio, we obtained the enthalpies of the specific association. Interestingly, the binding enthalpies were virtually identical for the unmodified and dithioated DNA. Table 1 summarizes the obtained binding free energy, enthalpy, and entropy for each DNA. Differences in the binding free energy (ΔΔG) and its enthalpic (ΔΔH) and entropic (−TΔΔS) terms clearly indicate that the affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate is entropy driven (Fig. 3d).

Table 1.

Thermodynamic parameters on protein-DNA association measured for the Antp homeodomain-DNA complexes at 25°C

Complex	ΔG (kcal/mol)a	ΔH (kcal/mol)b	ΔS (cal K−1mol−1)c
Unmodified DNAd	−10.89 ± 0.08	−8.64 ± 0.06	7.5 ± 0.4
Dithioated DNAe	−11.66 ± 0.08	−8.68 ± 0.07	10.0 ± 0.4

Each uncertainty represents the standard error of the mean.

^aMeasured from the fluorescence anisotropy-based titration data.

^bMeasured with the ITC data (Fig. 3c).

^cCalculated by ΔG = ΔH – TΔS.

^dComplex with unmodified DNA.

^eComplex with DNA containing phosphorodithioate at the Lys-57 interaction site.

NMR of interfacial Lys NH₃⁺ groups

To investigate the impact of the oxygen-to-sulfur substitution in the DNA phosphate on the intermolecular ion pairs with the protein side chain, we extensively characterized the Lys side-chain NH₃⁺ groups of the Antp homeodomain-DNA complexes by NMR. Fig. 4a shows the Lys NH₃⁺-selective HISQC spectra (29) recorded for the complexes with unmodified or dithioated 15-bp DNA at 15°C and pH 5.8. Four of six Lys side-chain NH₃⁺ groups of the protein clearly showed the ¹H-¹⁵N crosspeaks. By Lys-specific (27,29–31) and general (39) triple resonance experiments, these NH₃⁺ signals were assigned to Lys-46, Lys-55, Lys-57, and Lys-58. The NH₃⁺ groups of Lys-46, Lys-55, and Lys-57 form ion pair with DNA phosphate (24). The NH₃⁺ groups of Lys-2 and Lys-18 were not observed, presumably due to their rapid hydrogen exchange.

Figure 4.

NMR evidence for the hydrogen bonds between the Lys NH₃⁺ and DNA phosphate/phosphorodithioate groups. (a) Overlaid HISQC spectra recorded for the Lys NH₃⁺ groups in the unmodified (red) and dithioated (blue) DNA complexes of the Antp homeodomain. (b) The H3(N)P correlation spectrum recorded for the dithioated DNA complex. This spectrum shows ¹H-³¹P crosspeaks arising from coherence transfers via hydrogen bond scalar coupling ^h3J_NP between Lys ¹⁵N and DNA ³¹P nuclei. The signals represent the direct evidence for ionic hydrogen bonds with DNA phosphate (³¹P ∼−2 ppm) or phosphorodithioate (∼108 ppm). To see this figure in color, go online.

NMR evidence for ionic hydrogen bonds

A remarkable feature of NMR investigations on Lys NH₃⁺ groups is that quantitative measurements of relatively small (<1 Hz) scalar couplings between ¹⁵N and other nuclei is feasible because of extremely slow intrinsic ¹⁵N transverse relaxation of NH₃⁺ groups (27,32,40). For interfacial Lys NH₃⁺ groups forming a contact ion pair (CIP) with DNA, hydrogen-scalar coupling ^h3J_NP between protein side-chain NH₃^{+ 15}N and DNA ¹³P nuclei across a hydrogen bond could be detectable (9,27). In fact, the H3(N)P spectrum (9) recorded for the Antp homeodomain – dithioated DNA complex (Fig. 4b) clearly shows ¹H-³¹P crosspeaks that arise from heteronuclear ¹⁵N-³¹P scalar couplings across a hydrogen bond between Lys NH₃⁺ and DNA phosphate / phosphorodithioate groups. Of importance, this spectrum shows a crosspeak from the Lys-57 NH₃⁺ group with ³¹P resonance at 107.7 ppm, a typical ³¹P chemical shift for phosphorodithioate, which is substantially different from that for DNA phosphate (∼−2 ppm) (41,42). By the spin-echo ^h3J_NP-modulation constant-time HISQC experiment (9), the ^h3J_NP constant was measured to be 0.23 Hz for the Lys-57-phosphorodithioate ion pair. The ^h3J_NP constants for the Lys-46 and Lys-55, which form a CIP with DNA phosphate, were measured to be 0.62 and 0.41, respectively. Lys-58 did not exhibit ^h3J_NP coupling, which is reasonable because this residue does not form a CIP in the crystal structures. The ^h3J_NP data represent direct evidence for the hydrogen bonds between the Lys side-chain NH₃⁺ and DNA phosphate/phosphorodithioate groups in the complexes.

Mobility of interfacial Lys NH₃⁺ groups

To investigate the impact of the oxygen-to-sulfur substitution on the dynamics of ionic interactions between protein side chain and DNA phosphate, we conducted ¹⁵N relaxation analysis for the Lys side-chain NH₃⁺ groups of the two Antp-homeodomain DNA complexes at 25°C. Using 800- and 600-MHz NMR spectrometers, we measured heteronuclear ¹H-¹⁵N NOE and ¹⁵N R₁ and R₂ relaxation rates for the Lys NH₃⁺ groups of these complexes. The measured values of these parameters are given in Table 2. Upon dithioation of the DNA phosphate, only the Lys-57 NH₃⁺ group exhibited significant changes in ¹⁵N relaxation parameters (Fig. 5, a and b; see also Table 2). These most likely reflect a change in mobility rather than in covalent geometry, because NH₃⁺ groups are known to retain almost ideal tetrahedral geometry even in an ion pair with an acidic group (43). From the ¹⁵N relaxation data, we determined the order parameters S²_axis and reorientation correlation times τ_i for symmetry axes of NH₃⁺ groups and Cε-Nζ bond-rotation correlation times τ_f (Table 2). As we observed previously for the HoxD9 homeodomain, the interfacial Lys NH₃⁺ groups (Lys-46, Lys-55, and Lys-58) exhibited high mobility with S²_axis < 0.6 despite the ionic interaction with DNA. Among the four NH₃⁺ groups analyzed, only the Lys-57 NH₃⁺ group exhibited statistically different S²_axis values for the two complexes (Fig. 5c), which is reasonable because this is the only Lys residue close to the oxygen-to-sulfur substitution site. The S²_axis value for the complex with dithioated DNA was significantly smaller than that for the complex with unmodified DNA (0.34 vs. 0.54). These results indicate that the intermolecular ion pair of Lys-57 gets mobilized upon the oxygen-to-sulfur substitution in DNA phosphate, although our ^h3J_NP data clearly indicate the presence of a hydrogen bond between the NH₃⁺ and phosphorodithioate groups.

Table 2.

¹⁵N relaxation and dynamics parameters measured for the Lys side-chain NH₃⁺ groups of the Antp homeodomain-DNA complexes at 25°C and pH 5.8

	Lys-46 NH3+	Lys-55 NH3+	Lys-57 NH3+	Lys-58 NH3+
Complex with Unmodified DNA
800 MHz
15N R1 (s−1)	1.01 ± 0.02	0.49 ± 0.03	0.84 ± 0.01	0.28 ± 0.01
15N R2,ini (s−1)a	2.62 ± 0.06	1.60 ± 0.25	2.53 ± 0.04	1.05 ± 0.04
1H-15N NOE	−2.34 ± 0.08	−2.40 ± 0.23	−2.67 ± 0.03	−2.47 ± 0.06
600MHz
15N R1 (s−1)	1.17 ± 0.03	0.58 ± 0.05	0.98 ± 0.01	0.30 ± 0.01
1H-15N NOE	−2.93 ± 0.10	−2.90 ± 0.46	−2.84 ± 0.20	−3.09 ± 0.20
Dynamicsb
S2axis	0.49 ± 0.01	0.34 ± 0.07	0.54 ± 0.03	0.24 ± 0.01
τf (ps)	265 ± 11	60 ± 84	77 ± 53	5.4 ± 0.6
τi (ps)	0.0 ± 0.1	326 ± 170	550 ± 272	217 ± 14
Complex with Dithioated DNA
800 MHz
15N R1 (s−1)	1.02 ± 0.01	0.50 ± 0.01	0.33 ± 0.01	0.31 ± 0.01
15N R2,ini (s−1)a	2.74 ± 0.07	1.83 ± 0.17	1.39 ± 0.02	1.10 ± 0.03
1H-15N NOE	−2.49 ± 0.05	−2.75 ± 0.09	−2.35 ± 0.02	−2.60 ± 0.03
600MHz
15N R1 (s−1)	1.17 ± 0.02	0.53 ± 0.02	0.35 ± 0.01	0.32 ± 0.01
1H-15N NOE	−2.88 ± 0.07	−3.00 ± 0.12	−2.63 ± 0.02	−2.93 ± 0.03
Dynamicsb
S2axis	0.51 ± 0.01	0.43 ± 0.05	0.34 ± 0.01	0.26 ± 0.01
τf (ps)	243 ± 9	27 ± 12	8.4 ± 0.2	7.7 ± 0.3
τi (ps)	0.0 ± 0.01	212 ± 70	248 ± 6	211 ± 9

Signals from the Lys-2 and Lys-18 NH₃⁺ groups were not observed due to rapid hydrogen exchange with water.

^aThe initial rate for intrinsically biexponential ¹⁵N transverse relaxation of NH₃⁺ (32).

^bSymbols are defined in Fig. 5c. The molecular rotational correlation time and anisotropy were determined to be 9.9 ns and 1.3, respectively, from backbone ¹⁵N relaxation data.

Figure 5.

Mobilization of the Lys-57 NH₃⁺ group by dithioation of the interacting DNA phosphate group. (a) Lys NH₃^{+ 15}N R₁ relaxation (¹H frequency, 600 MHz) for the unmodified (red) and dithioated (blue) DNA complexes. (b) Lys NH₃^{+ 15}N R₂ relaxation (¹H frequency, 800 MHz) for the unmodified (red) and dithioated (blue) DNA complexes. (c) Changes in order parameters S²_axis for the symmetry axis of Lys NH₃⁺ groups upon dithioation of the DNA phosphate. To see this figure in color, go online.

Crystal structures of the complexes

For structural investigation, we crystallized the Antp homeodomain-DNA complexes with and without the dithioation for that phosphate. DNA duplexes with 5′-overhangs were used for the crystallization (Fig. 6). X-ray diffraction data for both complexes were collected to 2.7 Å. The crystals of these two complexes gave the same space group P4₃2₁2 and virtually the same cell dimensions. Although we obtained the crystals under the conditions described by Fraenkel and Pabo, the space group of our crystals was different from the previous one, P222₁ (24). This might be due to a slight difference in the N-termini of the protein constructs: Ours contains a Met residue (from the initial codon) before the first residue of the 60-amino-acid Antp homeodomain, whereas the corresponding N-terminal addition is Met-Glu in the construct of Fraenkel and Pabo (24). The asymmetric unit of our crystals contained two complexes. Refinement produced structures with good geometry and no Ramachandran outliers with free-R values of 28.7% for the unmodified DNA complex and 27.0% for the dithioated DNA complex. Table 3 summarizes the crystallographic data and statistics. Fig. 6a shows a superposition of our crystal structures of the dithioated (PDB ID: 4XIC) and unmodified (PDB ID 4XID) DNA complexes of the Antp homeodomain. The overall backbone structures of these complexes were virtually the same with root mean-square deviation (RMSD) being only 0.45 Å for the backbone atoms. DNA groove widths/depths calculated with the CURVES+ program (44) were virtually the same for the unmodified and dithioated DNA complexes (Fig. 6b). These results indicated that dithioation of a single phosphate does not impact the overall structure of the protein-DNA complex.

Figure 6.

Structural impact of dithioation of the DNA phosphate at Antp Lys-57 interaction site. (a) Superposition of the crystal structures of the Antp homeodomain-DNA complexes with (blue) and without (red) dithioation. The phosphate and phosphorodithioate groups at the modification site are indicated by an arrow. (b) Major and minor groove widths of the DNA double helices of the complexes. Solid and dotted lines show the data for structures 1 and 2, respectively, which are two independent structures in the asymmetry unit (see the main text). The values were calculated with the CURVES+ program (44). Horizontal arrows indicate values for the canonical B-form. (c and d) Electron density maps together with the structures of the Lys-57 side chain-DNA phosphate (c)/phosphorodithioate (d) ion pairs. Two independent structures in the asymmetric unit are shown for each complex. (e) DNA backbone torsion angles relevant to the phosphorodithioate and phosphate groups at the Lys-57 interaction site. To see this figure in color, go online.

Table 3.

Crystallographic data collection and refinement statistics

	Unmodified DNA Complex (PDB ID: 4XID)	Dithioated DNA Complex (PDB ID: 4XIC)
Crystallographic Data Collection
X-ray Source	Rigaku FRE++	APS BL-19ID
Wavelength (Å)	1.5418	0.9792
Space Group	P43212	P43212
Unit cell parameters (Å,°)	α =96.44, b =96.44, c =90.14, α = β = γ = 90	α =96.54, b =96.54, c =89.55, α = β = γ = 90
Sample temperature (K)	85	90
Resolution range (Å)	37.6–2.7	37.4–2.7
Total reflections	22043	18507
Nonanomalous reflections	12006	10913
Completeness (%)	99.0	99.8
Multiplicity	7.8	8.9
Rmerge	0.054	0.127
Rpim	0.024	0.043
Refinement
Rwork (%)	23.3	21.6
Rfree (%)	28.7	27.2
Bond RMSD from ideal values (Å)	0.005	0.006
Angle RMSD from ideal values (°)	0.746	0.823
Ramachandran plot: favored (%)	98.2	100
Allowed (%)	1.8	0
Outliers (%)	0	0
No. non-H atoms: protein	1055	1074
DNA	1222	1218
Water	15	14

Lys-57-DNA interactions in the crystal structures

We compared the structural details of the interactions between Lys-57 side chain and DNA phosphate/phosphorodithioate for the crystal structures of the Antp homeodomain-DNA complexes. Fortunately, because of the presence of two independent molecules per asymmetric unit, our crystallographic data provide some insight into the structural dynamics relevant to the oxygen-to-sulfur substitution. The two structures of the unmodified DNA complex show similar ion-pairing interactions between the Lys-57 side chain and the DNA phosphate involving an ionic hydrogen bond (i.e., CIP) (Fig. 6c). This observation is consistent with the crystal structure of Fraenkel and Pabo (24) as well as with our NMR hydrogen bond scalar coupling ^h3J_NP data. In contrast, the two structures of the dithioated DNA complex show significantly different interactions between the DNA phosphorodithioate and the Lys-57 side chain (Fig. 6d). In one of the structures (structure 1), the NH₃⁺ group is in contact with phosphorodithioate, forming a CIP (N…S distance, 3.0 Å). In the other structure (structure 2), the Lys-57 side chain exhibited a solvent-separated ion pair with phosphorodithioate (N…S distance, 5.5 Å). These results suggest that the Lys-57 side chain in the dithioated DNA complex is more dynamic than in the unmodified DNA complex. The phosphorodithioate itself also appears to be more dynamic than the corresponding phosphate. The DNA backbone torsion angles ε, ζ, α, and β for the phosphorodithioate differ by 10–20° between the two structures of the dithioated DNA complex, whereas the corresponding differences are <2° for the unmodified DNA complex (Fig. 6e). Although these crystallographic data show two distinct modes, our NMR spectra show a single crosspeak from the Lys-57 NH₃⁺ group. ¹⁵N relaxation dispersion data for the NH₃⁺ groups also show no evidence that Lys-57 undergoes fast exchange between multiple states on a μs-ms timescale. The internal motions indicated by low order parameters from the ¹⁵N relaxation data are in a ps-ns timescale. The discrete states observed in the crystal structures might be due to the difference in temperature (25°C in the NMR relaxation experiments versus ∼−185°C in the x-ray experiments), water activity, or crystal packing force. Nonetheless, it is important to note that the NMR and crystallographic data are consistent in that the Lys side chain becomes more dynamic upon the oxygen-to-sulfur substitution in DNA phosphate.

Discussion

Mobilization of intermolecular ion pair by dithioation of DNA phosphate

Our NMR and crystallographic data collectively indicate that the oxygen-to-sulfur substitution in DNA phosphate can mobilize the intermolecular ion pair. How could this mobilization occur? We speculate that the mobilization could be related to a relatively flat energy surface of the H•••S hydrogen bond and a larger effective radius of sulfur (1.84 Å for sulfur vs. 1.40 Å for oxygen) (17). Recent theoretical quantum chemical studies have shown that, compared to H•••O hydrogen bonds, the enthalpy for H•••S hydrogen bonds is slightly smaller and that its energy surface is flatter (45,46). Due to the flatter energy surface for sulfur, a slight deviation from ideal hydrogen bond geometry causes only a marginal increase in enthalpy and may allow for a wider spatial distribution of a Lys side-chain NH₃⁺ group, especially with the relatively large ionic radius of sulfur. This effect may make the intermolecular ion pair more dynamic. Weaker interactions of phosphorodithioate with water molecules (47) could also mobilize the intermolecular ion pair.

Entropic gain due to mobilization of the intermolecular ion pair

By using the experimental order parameters S²_axis together with a particular motional model, we could roughly estimate the entropic gain due to mobilization of the Lys-57 NH₃⁺ group upon the oxygen-to-sulfur substitution in DNA phosphate. Assuming the diffusion in a cone model (48), which gives entropic difference equal to $k_{\mathrm{B}}ln\left[\left\{3-{\left(1+8S_{axis,a}\right)}^{1/2}\right\}/\left\{3-{\left(1+8S_{axis,b}\right)}^{1/2}\right\}\right]$ (k_B, the Boltzmann constant), the increase in entropy for the symmetry axis of the Lys-57 NH₃⁺ group by the oxygen-to-sulfur substitution is estimated to be 1.0 cal K⁻¹ mol⁻¹. Additionally, difference in rotational entropy (S_rot) of an NH₃⁺ group can contribute to the entropic gain. As considered previously for CH₃ groups (49), the rotational entropy (S_rot) of an NH₃⁺ group is indirectly related to the bond-rotation kinetics because the probability distribution function for the bond torsion angle depends on the energy barrier for rotation. Provided that the Eyring equation is applicable to NH₃⁺ rotation (33), experimental τ_f data along with the analytical expression (49) of S_rot suggest that NH₃⁺ rotational entropy could increase by ∼0.3 cal K⁻¹ mol⁻¹ upon the oxygen-to-sulfur substitution in the interacting DNA phosphate group. The overall entropic gain arising from mobilization of the NH₃⁺ group (i.e., reorientational + rotational) is thus estimated to be ∼1.3 cal K⁻¹ mol⁻¹. Relatively high mobility of the phosphorodithioate group, which is implicated by our crystallographic data (see Fig. 6e), can further increase the entropic gain for the ion pair. Although the mobility of the intermolecular ion pairs may depend on ionic strength, a previous NMR study showed only weak ionic strength dependence for the Arg side-chain dynamics in peptide-RNA complex (50). If the oxygen-to-sulfur substitution in DNA phosphate increased the ion-pair mobility to a similar degree at 20 and 150 mM NaCl, this entropic effect would be comparable to the observed increase in binding entropy (ΔΔS = 2.5 ± 0.6 cal K⁻¹ mol⁻¹) upon the dithioation.

Consideration on other entropic effects

Our thermodynamic data clearly indicate that the affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate is entropy driven. As described previously, mobilization of the intermolecular ion pair seems to make a significant contribution to this entropic gain. Here, we consider other entropic effects that can contribute to the affinity enhancement.

The polyelectrolyte effect, which arises from release of condensed counterions upon DNA-association, could contribute to the entropic difference (51). The entropic term of this effect is given by $\Delta S_{PE}=-z\psi Rln\left[M^{+}\right]$, where z is the number of DNA phosphates that interact with the protein; R is the gas constant; [M⁺] is the concentration of cations; and ψ is the number of the released cations per phosphate (ψ ≤ 1), which is given as a function of the charge and axial phosphate distance along DNA (51,52). Because the parameter z is identical for the unmodified and dithioated DNA complexes and only a single phosphate group is substituted to phosphorodithioate in the dithioated DNA, the entropic change in the polyelectrolyte effect upon the oxygen-to-sulfur substitution (ΔΔS_PE) is given by $-\left({\psi }_S-{\psi }_O\right)Rln\left[M^{+}\right]$, in which ψ_Oand ψ_S are for phosphate and phosphorodithioate, respectively. Because ψ_O = 0.88 for unmodified DNA (52) and ψ_S ≤ 1, the upper limit of ΔΔS_PE at 0.15 M NaCl is calculated to be 0.4 cal K⁻¹ mol⁻¹. Actual ΔΔS_PE could even be negative, because the weaker charge density of phosphorodithioate could cause ψ_S < ψ_O. Thus, it is very unlikely that the polyelectrolyte effect contributes to the entropic enhancement of the affinity by the oxygen-to-sulfur substitution.

The hydrophobic effect could also contribute to the entropic enhancement of the affinity. Difference in the hydrophobic effect between the unmodified and dithioated DNA complexes may arise from a different degree of solvent-exposure around the modification site. The entropic term in cal K⁻¹ mol⁻¹ units for this effect can be roughly estimated by using an empirical equation of Spolar and Record (51), $\Delta S_{HE}=0.32\Delta A_{np}ln\left(T/386\right)$, where ΔA_np, the total change in accessible surface areas (in Ų) of nonpolar atoms due to binding; and T, temperature (in K). Because Lys-57 of the dithioated complex is more exposed to solvent (i.e., smaller ΔA_np) in structure 1 as shown in Fig. 6, this entropic effect should be smaller for the dithioated complex, and therefore cannot account for our entropic data. Strictly speaking, however, more accurate assessment of the hydrophobic effect requires quantitative information on desolvation entropy for both phosphate and phosphorodithioate groups, though qualitatively, the lower charge density of the phosphorodithioate group should weaken its interactions with water (53).

Conclusions

This work provides mechanistic insight into affinity enhancement by dithioation of DNA phosphate. Our ITC and fluorescence data clearly indicate that the affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate is entropy driven. Our NMR and x-ray data show that upon the oxygen-to-sulfur substitution, the Lys side chain in contact with the DNA phosphate becomes more dynamic, yielding a significant gain in binding entropy (ΔΔS). This entropic effect seems to make a major contribution to the affinity enhancement by dithioation of DNA phosphate. Recently, based on the crystal structure of the dithioated RNA in the free state (not in the complex), Egli and coworkers (47) speculated that phosphorodithioate may enhance binding of RNA to the Ago2 protein via an increased hydrophobic effect. However, our data suggest that mobilization of intermolecular ion pairs play a more important role in the affinity enhancement by the oxygen-to-sulfur substitution in DNA phosphate, at least for the Antp and HoxD9 homeodomains. Our current study sheds light on the role of ion-pair dynamics in protein-DNA association. Enhancement of the ion-pair dynamics may be an effective strategy to improve design of high-affinity oligonucleotides that efficiently inhibit transcription factors for therapeutic purposes.

Author Contributions

J.I. designed research; L.Z., D.N., M.A.W., and J.I. performed research; K.M.A. and D.G.G. provided dithioated DNA strands; L.Z. and J.I. wrote the article.

Editor: Jeff Peng.