Probing the Molecular Recognition of a DNA·RNA Hybrid Duplex

DNA·RNA hybrid duplexes are the nucleic acid structures least explored as targets in drug discovery. There is a dearth of model ligands displaying convincing structure or sequence selectivity; the secondary structures of the hybrid nucleic acid hosts are variable and complex.^[1,2] Ligands selective for hybrid duplexes have potential therapeutic applications as telomerase and RNaseH inhibitors.

Assays for investigating small molecule interactions with DNA·RNA hybrid duplexes have only recently been developed and described.^[3] Herein their application to the discovery of a small molecule that specifically recognizes poly(dA) ·poly(rU) is reported. An unexpected binding mode was uncovered, establishing a principle that extrapolations from pure DNA or RNA binding are invalid when considering DNA·RNA hybrid duplexes.

Competition dialysis (Fig. 1) showed that the non-classical intercalator 4,6-bis-[4'-[[2"-(dimethylamino)ethyl]mercapto]phenyl]pyrimidine dihydrobromide 1 bound preferentially to quadruplex and triplex DNA structures and to the hybrid duplex poly(dA)·poly(rU). In duplex binding, strong preferences for both backbone and base orientation were evident: association with the poly(dA)·poly(rU) hybrid duplex had ~20-fold preference over the equivalent RNA, 3-fold over DNA, and 7-fold over the alternative poly(rA)·poly(dT) hybrid.

Figure 1.

Competition dialysis data for compound 1, all data are the mean±SD of 3 separate determinations.

Relative binding preferences for duplex structures were confirmed in a melting of mixtures assay in which UV melting profiles were obtained for a mixture of all four A·T(U) duplexes in the absence and presence of low concentrations of ligand, Fig. 2.^[3,4] Significant shifting of the melting transition of the poly(dA)·poly(rU) hybrid was accompanied by a lesser shift in the DNA duplex melting curve at higher ligand loading.

Figure 2.

Melting of mixtures data for the four A·T(U) duplexes (each 10 µM(bp)) in the presence of increasing concentrations of ligand 1.

CD experiments (Fig. 3a, c) further elucidated the binding interaction. As ligand 1 was titrated into the hybrid solution, the intensity of nucleic acid ellipticity (244 and 275 nm) decreased but the envelope of the spectrum retained its starting shape: indicative of reduced chirality in the oligonucleotide (unwinding) due to intercalation, without alteration of the global conformation. Negative induced CD bands observed for the ligand (320 and 364.5 nm) were also consistent with intercalation.^[5] Intercalation was confirmed by a preliminary analytical ultracentrifugation (AUC) experiment in which the sedimentation coefficient decreased upon ligand loading in a manner consistent with unwinding and lengthening of the duplex (Fig S3). Plots of induced CD intensity vs. ligand:base pair ratio (Fig. 3a, inset) had an inflection at a ratio of 1:10. The 10 bp per ligand binding stoichiometry was confirmed by a Job plot (Fig. 3b).

Figure 3.

a) CD titration of ligand 1 into poly(dA)·poly(rU) duplex; inset, induced CD as a function of the [1]/[poly(dA)·poly(rU)]_bp ratio. b) Job plot for mixtures of compound 1 with poly(dA)·poly(rU) at a constant 100 µM. c) CD titration of ligand 1 into poly(dA)· [poly(dT)]₂ triplex, the spectrum of the free nucleic acid is shown as a heavy line; inset, induced CD as a function of the [1]/[poly(dA)·{poly(dT)}₂]_bt ratio.

CD titration spectra using poly(dA)·[poly(dT)]₂ triplex DNA, Fig. 3c, reflected those obtained for the hybrid duplex, showing unwinding of the triplex and negative induced CD in the ligand (300, 360.8, 328 nm). The inflection in the induced CD plots (Fig. 3c, inset) occurred at a mixing ratio of 0.15, i.e. ~7 bt per ligand.

Isothermal titration calorimetry (ITC) probed thermodynamic details of the binding interaction. The thermogram for ligand dilution indicated disaggregation of the ligand; data analysis using an isodesmic self-aggregation model yielded K_agg = 4.1±0.7×10² M⁻¹ and ΔH_agg = −7.3±0.6 kcal mol⁻¹ (Fig. S4, S5). The thermogram for ligand–hybrid interaction indicated (at least) two types of binding site (Fig. 4). ITC data were analyzed using a model describing two consecutive binding events and a model describing two independent binding events; both models incorporated simultaneous ligand self aggregation (Scheme S1). Strong parameter correlation meant that the two types of site could not be quantified independently in a statistically-meaningful way. Nevertheless, analysis of parameter correlation plots concluded that the two binding events are a major binding site of 10 bp in combination with a minor binding event involving ligand association with (potentiallyfraying) ends and gaps (supporting information). The overall binding stoichiometry is consistent with other experimental data for this system and simulated data based on binding parameters from ITC reproduced UV titrations well (Fig. S11).

Figure 4.

Integrated heat effects for titration of ligand 1 into poly(dA)·poly(rU) fitted using a model involving a major binding site in combination with a minor binding event (see supporting information).

In conclusion, biarylpyrimidine 1 shows a strong preference for binding the poly(dA)·poly(rU) hybrid over all other duplex nucleic acid structures and as a DNA·RNA hybrid binder, inhibited RNaseH^[3] with IC₅₀ = 29.5 ± 0.2 µM (Fig. 5).

Figure 5.

Inhibition of the RNaseH digestion of poly(dA)·poly(rU) by compound 1; IC₅₀ = 29.5 ± 0.2 µM.

All experimental data indicated that each ligand occupied a 10 bp site; CD and AUC data unambiguously showed nucleic acid intercalation. The non-classical intercalator has flexibility of torsion angles between the three aromatic rings. This flexibility may contribute to the binding preference for the A-like hybrid by allowing the ligand to accommodate closely to propeller-twisted base pairs.^[6] Ligand 1 also showed binding preferences for triplex and quadruplex nucleic acids. Moreover, triplex poly(dA)· [poly(dT)]₂ also exists in an A-like helix. It seems, therefore, that the binding preferences discovered are selection for the global A-like structure. In contrast to major groove-binding aminoglycosides,^[7] the binding event did not drive a change in the global nucleic acid conformation. Even though the two ligand classes share correlating preferences for triplex structures^[8] there was also no evidence that ligand 1 induced hybrid duplex→triplex transitions.

Ligand 1 did not bind to the other A-form structure in Fig. 1, the RNA duplex poly(rA)·poly(rU), an observation in contrast to the aminoglycosides, where binding to equivalent RNA duplexes was strong.^[7] An explanation may lie in the grooves and the dimensions of the ligand. The ligand possesses two cationic side chains and it may be postulated that one lies in the minor groove. Although the 2'-OH and associated solvation waters of a single RNA strand offer potential hydrogen bonding sites to the terminal ammonium functionality, these groups occlude the minor groove of the pure RNA duplex. Such steric constraints would also account for the reduced affinity of bulkier analogs of ligand 1 for the hybrid duplex (Fig. S12). Indeed the hybrid minor groove appears inaccessible to all but the smallest groups: pentamidine, berenil, Hoechst 33258 and DAPI all failed to stabilize either of the hybrid duplexes in the melting of mixtures assay.^[3]

Binding rules for DNA·RNA duplexes are ill-defined and this study shows that naïve extrapolation from duplex binding models is inappropriate. The structural basis for the ~10 bp major binding site size is not yet clear but is equivalent to ~1 ligand per helical turn. The intercalator ethidium has been shown to exclude 3 or 7 bp when bound to the poly(dA)·poly(rU) hybrid.^[9,10] and precedent for long-range transmission of binding effects over tens to hundreds of bp exists in the daunorubicin-driven Z→B conversion of [poly(dGdC)]₂. [11]

DNA·RNA hybrids and DNA triplexes both adopt structures intermediate between classic B-form DNA and A-form RNA. The correlation between binding data for ligand 1 and its analogs to these two structures is striking (Figs. 3 and S13) and demonstrates that the search for structure and sequence selective ligands for DNA·RNA hybrid duplexes should start with re-evaluation of DNA triplex-targeting compounds.

Experimental Section

Compound 1 was synthesized as previously described.^[12,13] Oligonucleotides for CD and calorimetry were obtained from Midland Certified Reagent Co, Midland, TX (poly(dA), lot 111997, 126–1200 nucleotides; poly(rU) lot 102197, 500–2600 nucleotides). All other oligonucleotides were purchased from Sigma, Milwaukee, WI, USA or Poole, UK.

CD spectra were recorded on a Jasco J-810 spectropolarimeter with Peltier. UV melting experiments were performed on a Cary 400Bio spectrophotometer equipped with Peltier. Other spectra were recorded on a Jasco V-550 spectrophotometer and Tecan Safire₂ microplate reader. ITC was performed on a Microcal VP-ITC calorimeter.

Competition dialysis,^[3,14–17] melting of mixtures^[3,4] and RNaseH assays^[3,4] followed published protocols.

Job plots: the protocol followed that set out by Jenkins^[18] but detected by UV. Mixtures of 1 and poly(dA)·poly(rU) to a summed concentration of 100 µM in BPES pH 6.00 were set up in duplicate on a 96 well microtitre plate. Reference solutions had buffer in place of nucleic acid. Spectra were recorded from 230–450 nm. Absorbance data for the drug maximum at 356 nm were extracted and the free drug absorbance subtracted from that of the drug-nucleic acid complex before plotting A₃₅₆ vs. mole fraction of ligand.

CD Titrations: nucleic acid solutions in BPES pH 6.00 (1 ml) were prepared to have A₂₅₇ ~0.8 and placed in a 1 cm pathlength quartz cuvette equipped with stirrer. Concentrated ligand solutions were prepared in the same buffer to minimize volume changes during titration. Aliquots were added until precipitation of the ligand-nucleic acid complex was evident. Raw data were corrected for dilution before plotting and analysis.

ITC: all experiments were performed in BPES pH 6.00. Oligonucleotides were dissolved in buffer, dialyzed against buffer (3 changes of solution) and the final dialysate used as solvent for all reactions and dilution of reagents. In a typical experiment, a solution of poly(dA)·poly(rU) (0.5 mM(bp)) was loaded into the calorimeter cell and ligand 1 (0.25 mM) loaded into the syringe; aliquots (10 µl) were added at intervals of 360 s. To determine the heats of dilution, a solution of 1 (5 mM) was titrated (15 µl aliquots, 360 s) into buffer. Other control titrations of buffer into nucleic acid and buffer into buffer were also performed. Data were treated in Origin (Microcal, Inc) to generate integrated heat effects per injection (Δh). These were analyzed using IC ITC following published procedures.^[19,20] Ligand self-aggregation parameters were determined from ligand dilution experiments and held constant in subsequent analyses of binding data.