Small-Molecule-Based Self-Assembled Ligands for G-Quadruplex DNA Surface Recognition.

Most drugs are small molecules because of their attractive pharmacokinetics, manageable development and manufacturing, and effective binding into the concave crevices of bio-macromolecules. Despite these features, they often fall short when it comes to effectively recognizing the surfaces of bio-macromolecules. One way to overcome the challenge of biomolecular surface recognition is to develop small molecules that become self-assembled ligands (SALs) prior to binding. Herein, we report SALs made from 8-aryl-2'-deoxyguanosine derivatives forming precise hydrophilic supramolecular G-quadruplexes (SGQs) with excellent size, shape, and charge complementarity to G-quadruplex DNA (QDNA). We show that only those compounds forming SGQs act as SALs, which in turn differentially stabilize QDNAs from selected oncogene promoters and the human telomeric regions. Fluorescence resonance energy-transfer melting assays are consistent with spectroscopic, calorimetric, and light scattering studies, showing the formation of a "sandwichlike" complex QDNA·SGQ·QDNA. These results open the door for the advent of SALs that recognize QDNAs and potentially the surfaces of other bio-macromolecules such as proteins.

Introduction

Historically, the development of small-molecule ligands has been the major focus of pharmacology and chemical biology because they are ideally suited to bind to concave crevices such as receptor pockets and the active sites of enzymes. A small size generally also improves the pharmacokinetic properties of drugs and cell permeability and enables standard pharmaceutical manufacturing practices. When it comes to recognizing bio-macromolecular surfaces, however, small molecules often fall short of their desired affinity and specificity unless the macromolecule has a so-called “hot spot” on its surface.[1,2] In general, small molecules simply lack the affinity and selectivity toward the mostly flat (convex) surfaces of bio-macromolecules such as proteins.[3] Targeting a small concave cavity with a small molecule is feasible because in the drug design process, the contact surface area can be readily maximized. Once inside a pocket, a small molecule is shielded from the constant jostling by solvent molecules, to a much greater extent than what would be apparent on a molecular surface.

Successful strategies for bio-macromolecular surface recognition include the use of large structures such as antibodies, nucleic acids such as aptamers,[5,6] or nanoparticles.[7−10] The larger size of these nanostructures enables a larger binding surface area and multivalent interactions. However, is it possible to cover large portions of bio-macromolecular surfaces using small molecules as building blocks for SALs?

What is an inherent SAL? It is the one whose supramolecular structure is dictated by the information encoded in its subunits, preferably one which is precise. Every supramolecular assembly has a critical (or lower) self-assembly concentration, which is defined as the concentration at which half of the subunits are in the assembled state.[11,12] It is, thus, valuable to develop inherent SALs whose formation occurs in a highly cooperative process, to maximize their selectivity, affinity, and other critical attributes. The more cooperative the self-assembly process, the sharper the transition between monomer and assembly with a drastic nonlinear change in a relatively narrow concentration range. Cooperative self-assembly favors the formation of specific supramolecular structures, and it can arise from allosteric effects and/or from mutually supporting molecular surfaces.[13,14]

The strategy we present here to develop inherent SALs relies on small molecules that “grow” before reaching their target (Figure A). There is some precedent for “conditional” SALs (Figure B) where the simultaneous binding of small molecules as aggregates (e.g., dimers and trimers) happens only in the presence of targets such as the active sites of proteins[15] or the groove of double-stranded DNA (dsDNA).[16] However, such systems generally do not form precise supramolecular assemblies prior to binding their target,[17] resulting in poor selectivity if the interactions between the subunits are weak or ambiguous. An ideal (inherent) SAL should be stable enough prior to binding to ensure good affinity and selectivity, especially when targeting a biomolecular surface. Affinity and selectivity could then be fine-tuned as in traditional drug development strategies [e.g., structure–activity relationship (SAR)]. Herein, we report the development of inherent SALs for QDNA formed via the self-assembly of 8ArG derivatives into SGQs (Figure C,D). We also provide the structural characterization of the supramolecular complex formed between a QDNA and an SGQ acting as an SAL, which provides critical insights into the specific noncovalent interactions driving the complex formation.

Figure 1

Cartoon depiction of inherent vs conditional self-assembled ligands (SALs) and structural depictions of a supramolecular G-quadruplex (SGQ) made from 8-aryl-2′-deoxyguanosines (8ArGs) acting out as an inherent SAL for a G-quadruplex DNA (QDNA). (A) Inherent SALs self-assemble prior to binding, thus increasing the selectivity for a desired macromolecular target. (B) Conditional (or templated) SALs self-assemble only in the presence of a target, which acts as a template. (C) Kekulé structures of compound 1 and its corresponding tetrad 14.[4] (D) Schematic depiction of the formation of a supramolecular complex (116·QDNA) between hexadecameric SGQ 116 and the telomeric unimolecular QDNA structure formed by d[AG3(T2AG3)3] (pdb 1KF1). (E) Comparison between a single vs four molecules of derivative 1 binding to QDNA.

Guanine-rich DNA sequences able to form QDNA structures are abundant on the telomeres and the promoter regions of some oncogenes.[18−22] They are attractive pharmacological targets because of their putative role in the regulation of cellular processes such as replication and cell survival. The development of ligands that bind QDNA with high selectivity and affinity remains a challenge that must be overcome to make QDNAs a viable target for pharmacological intervention.[18,23−26] Effective QDNA ligands must differentiate a desired target QDNA from dsDNAs and from alternative QDNA structures adopting different topologies (Figure ).[27−32] The search for viable QDNA ligands ranges from a wide variety of small molecules to more innovative designs such as conditional SALs where a QDNA templates the organization of ligands on top of the G-tetrads exposed to the solvent.[27−37] In the strategy described above, the size of a single molecule of a G-derivative such as 1 is not large enough to cover significant portions of the surface of a QDNA (Figure E). By contrast, upon self-assembly, the resulting SGQ having its analogous exposed flat G-tetrads enables the formation of more effective complexes (Figure D) between QDNA structures and the inherent SALs reported in here.

Figure 4

Stabilization of oligodeoxynucleotides (ODNs) (expressed as ΔT1/2) in the presence of 8ArGs (64 μM) in sodium (A,C) or potassium buffers (B). FRET melting assays for 0.5 μM ODN in 10 mM LCB with (B) 10 mM KCl/90 mM LiCl (hTelo), 1 mM KCl/99 mM LiCl (c-MYC and c-KIT2 oncogenes), and (A,C) 100 mM NaCl. In some cases, the ΔT1/2 values for 2 or the error bars on the graphs are too small to notice. The molecular models that show the main topologies adopted by hTelo, c-MYC, and c-KIT2 in K+-buffer were prepared from the pdb codes 1KF1,[49]1XAV,[50] and 2KYP, respectively. The structure of hTelo in Na+ buffer was prepared from the pdb code 143D.

Results and Discussion

Self-Assembly Studies of Hydrophilic 8ArG Derivatives

In aqueous media, successful self-assembly of precise (discrete and well-defined) supramolecules held together by noncovalent interactions is very challenging. We have reported that in H2O/D2O (9:1) with at least 5 mM KCl, compound 1 forms the hexadecameric SGQ 116 in contrast to 2, which does not self-assemble under these conditions (i.e., at or below low millimolar concentrations) because of the lack of the meta-carbonyl phenyl group attached to the C8 of the guanine moiety.[4,38] Compound 3 is a new derivative of 1, synthesized to evaluate the impact of longer ester chains to their self-assembly into SGQs and the subsequent binding to QDNAs.

The propensity of compounds 1–4 (prepared as previously reported[4,38,39] and as described in the Supporting Information) to self-assemble into SGQs in buffer solutions similar to those used in the fluorescence resonance energy-transfer (FRET) melting assays (e.g., 10 mM LCB) was investigated using a combination of 1H NMR (Figures and S18–S21), dynamic light scattering (DLS) (Figure A), and differential scanning calorimetry (DSC) (Figure B). Under potassium buffer conditions (LCB with 100 mM KCl and 4 mM 8ArGs), the 1H NMR signatures as well as the relative hydrodynamic diameter (Dh) and thermodynamic stability determined by DLS and DSC confirm that only 1 and 3 self-assemble into SGQs 116 (Dh = 4.7 nm) and 316 (Dh = 5.0 nm), whereas 2 (Dh = 1.2 nm) and 4 (Dh = 1.9) remain as monomers.[40] An additional test to compare the propensity of 8ArGs to self-assemble into SGQs is to replace the promoting cation from potassium to sodium in the buffer. The latter is less efficient at promoting the formation of SGQs,[41] resulting in a shift in the equilibrium away from the hexadecameric SGQs 116 and 316. 1H NMR spectra reveal that in sodium buffer (LCB with 100 mM NaCl and 4 mM 8ArGs), both 2 and 4 remain as monomers, whereas 1 and 3 self-assemble into SGQs, albeit with a larger proportion of octameric SGQs, specifically, 89% for 18 and 37% for 38, which relative to potassium buffer represents an increase of 21 and 9%, respectively. Given that both 2 and 4 remain as monomers in the concentration range tested, they were suitable controls in the QDNA-binding studies.

Figure 2

8ArG self-assembly. (Top) Kekulé structures of G-derivatives 1–4 and molecular surfaces for SGQs 116 and 316 overlaid on the corresponding polytube representations. Red and blue regions in the molecular surfaces indicate the locations of the methylenes and the carbonyl group highlighted in the Kekulé structures, respectively. (Bottom) 1H NMR spectra of 8ArGs (5 mM) were measured in 10 mM lithium cacodylate buffer pH 7.2 (LCB) with 100 mM KCl and 100 mM NaCl. The expansions show the regions of the N1H and the sugar where red highlights the signals of hexadecamer species[38] and light blue highlights those of octamer species. Compounds 2 and 4 show no evidence of self-assembly at the concentrations shown. The meaning of the subscripts in the peak labels is as follows: W = Watson–Crick edge; o = outer tetrads; and i = inner tetrads.

Figure 3

(A) Hydrodynamic diameter (Dh in nm) measured by DLS and (B) DSC endotherms in 10 mM LCB with 100 mM KCl. Samples were 4 mM 1–4, and the main species in solution were SGQs for 1 and 3 and monomers (Mon.) for 2 and 4.

Because SGQs have a lower self-assembly concentration in the micromolar range where the FRET assays were performed, it is critical to determine if they were present in significant amounts or if the equilibrium was shifted mostly to the monomeric subunits. The thiazole orange (TO) fluorescence enhancement assay is an attractive way to gain further insights into this issue, considering the sensitivity limitations of NMR in that concentration range. TO fluorescence enhancement assay, which is widely used to track the binding of ligands to QDNA,[42,43] is nonfluorescent when free in solution but becomes highly fluorescent upon complexation with QDNA because of the restricted rotations that limit nonradiative energy dissipation. In a similar fashion, we hypothesized that TO could report on the presence of SGQs, given some of their shared features with QDNAs. Fluorescence measurements of TO with and without 2 (120 μM) show no enhancement, which contrasts with a fourfold enhancement in the presence of equal amounts of 3 (Figure S43). We hypothesize that the smaller magnitude of the enhancement, relative to that reported for polyanionic QDNAs, is, in part, because of the Coulombic repulsion expected for the cationic TO binding to the polycationic SGQ. Nonetheless, this enhancement is consistent with the presence of SGQs, which offer a large solvent-exposed area suitable for binding.

Can 8ArG SGQs Selectively Stabilize QDNA Structures?

The samples of 1–4 (5 mM in LCB with 100 mM NaCl) used for NMR analysis were diluted to 64 μM and used in a FRET melting assay with the QDNA sequence derived from the human telomeric sequence (hTelo) (Figure A). The rationale for using Na+ instead of K+ was to test the hypothesis that only the more robust SGQs would stabilize the hTelo QDNA structure. The experimental data validate this hypothesis because the stabilization of the QDNAs follows the same trend as the propensity of compounds 1–4 to self-assemble into SGQs. Specifically, compound 3, which forms the most stable SGQ, stabilized the hTelo QDNA with ΔT1/2 of ∼9 °C, whereas compound 1, which forms a less stable SGQ, resulted in a smaller stabilization (ΔT1/2 = 1 °C) of the same QDNA. By contrast, compounds 2 and 4, which under these conditions do not self-assemble into SGQs, destabilized the hTelo QDNA instead. This destabilization might result from the alternative interactions with the loops and/or grooves or even direct competition between the G-derivatives and the guanine moieties in the QDNAs. The importance of an SGQ acting as an inherent SAL toward the hTelo QDNA is underscored by the following: (a) because 1 and 4 are isomers, the lack of stabilization by the latter demonstrates the importance of the stereochemical arrangement of the carbonyl group (meta vs para) and the corresponding SGQ formation (Figure A) and (b) in 1 and 3, both of which have a meta-carbonyl phenyl moiety, the degree of stabilization is directly proportional to the inherent stability of the corresponding SGQs in NaCl buffer (Figures and 4).

In addition to hTelo, we evaluated if the SGQs formed by 1 or 3 could interact and stabilize the QDNA structures derived from the promoter regions of the oncogenes c-MYC (FmycT) and c-KIT2 (Fkit2T)[44,45] in various buffers, with compound 2 serving as a negative control (i.e., no SGQ formation) for FRET melting assays. These sequences enable the evaluation of the binding of SGQs as a function of their structural and topological features (Figure B,C). FRET melting profiles in LCB, with NaCl or KCl, demonstrate that 3 and, to a lesser extent, 1 differentially stabilize QDNA structures in a dose-dependent manner (Figures S25–S27), in contrast with compound 2, which shows no significant stabilization. The highest stabilization by 1 and 3 (64 μM), respectively, was observed for the structure of F21T in K+ buffer (ΔT1/2 ≈ 12 and 15 °C) and for FmycT in Na+ buffer (ΔT1/2 ≈ 10 and 13 °C). However, only 3 significantly stabilized the QDNA structures for both hTelo (in Na+ buffer) and c-KIT2 (in both K+ and Na+ buffer), which highlights a selectivity for different QDNA topologies. For FmycT in Na+ buffer, the slightly larger stabilization induced by 3 over 1 (ΔT1/2 ≈ 13 and 10 °C, respectively) and the greater steepness of the melting curves suggest a more cooperative dissociation of the resulting complexes formed by the putative SGQs made by 1 or 3 with these QDNAs.[46] The corresponding measurements for 1 in Li+ buffer, by contrast, showed no significant stabilization of either of the QDNAs[47,48] or the SGQs.[41]

To determine if the observed behavior resulted primarily from simple electrostatic interactions, we compared 3 and 2 as representative G-derivatives with, respectively, the highest and lowest propensity to self-assemble into SGQs. Neither 3 nor 2 imparted significant stabilizations of the dsDNA (Figures S39 and S43). This indicates that neither electrostatic complementarity alone nor additional nonspecific interactions drive the complexation of 3 to QDNA, but instead is its higher propensity to form SGQs that act as SALs.

We hypothesize that the observed selectivity (Figure B,C) results from specific interactions between the ester side chains in the SGQs formed by 1 and 3 with the loops of the various QDNAs tested. For example, under potassium buffer conditions (Figure B), 1 leads to greater discrimination than 3 despite the greater stabilization induced by the latter. The greater flexibility enabled by the longer side chains in compound 3 can adjust to the various topologies adopted by the QDNAs, which in turn results in a lower degree of discrimination.

Although the FRET melting assays provide compelling support for the SGQs acting as SALs for QDNAs, to further support this hypothesis, we performed a series of NMR experiments using a simplified tetramolecular QDNA formed by the hTelo sequence 5′-TTAGGG-3′.[51] The rationale for using a simplified QDNA structure is as follows: (a) a lower complexity of the spectra facilitates their interpretation; (b) higher overall symmetry enhances the signal-to-noise ratio; (c) because the selected QDNA dimerizes (via its 3′ end), disruption of the dimer (QDNA2) provides a convenient way to track the formation of the complex as has been reported for small-molecule ligands;[51−53] and (d) intercalation of the SGQ into QDNA2 to form a QDNA·SGQ·QDNA complex would likely preserve an overall high symmetry, which also simplifies the spectral interpretation.

The spectra for the N1H “imino” groups in QDNA2 (4.7 mM ssDNA concentration, 50 mM potassium buffer pH 7.05, 300 mM KCl, and 90:10 of H2O/D2O) show three sets of signals corresponding to the consecutive imino hydrogens from G6, G5, and G4 at 10.5, 10.9, and 11.3 ppm, respectively (Figure Ai).[52] Addition of a single equivalent of 1 leads to the disruption of the 1H NMR spectrum of QDNA2 (Figure Aii), which after adding a second equivalent reveals the signals corresponding to SGQ 116 (Figure Aiii).[38] After equilibrating for 48 h, the spectrum shows a new species with three sets of signals from QDNAs (G4, G5, and G6) and two from the SGQ 116, intercalated in between (Figures Aiv and 6A). These results provided the first indication of the formation of a “sandwichlike” complex of general formula QDNA·116·QDNA, which is further supported by the DSC, DLS, and 2D NMR measurements discussed below.

Figure 5

NMR, DSC, and DLS experiments are consistent with the formation of a sandwich complex. (A) 1H NMR (500 MHz) spectra for the titration of 1 into a solution of the dimer of tetramolecular QDNAs formed by 5′-d[T2AG3]-3′ (4.7 mM in single-stranded DNA (ssDNA), 50 mM potassium buffer pH 7.05, 300 mM KCl, 90:10 of H2O/D2O). (i) 0 equiv, (ii) 1 equiv at t0, (iii) 2 equiv at t0, and (iv) 2 equiv at t48h. (B) Expansions of the 1H–1H nuclear Overhauser enhancement spectroscopy (NOESY) spectrum of a sample with 2 equiv of 1 at t48h (4.58 mM in ssDNA, 50 mM potassium buffer pH 7.02, 300 mM KCl, 90:10 of H2O/D2O). The nuclear Overhauser effect (NOE) correlations between G6/H2β′ and 1o correspond to the expanded region shown in Figure A. (C) DSC endotherms for 116 (3.3 mM in 1), QDNA2 (1.4 mM in ssDNA), and the QDNA·116·QDNA complex (1.4 mM in ssDNA and 3.2 mM in 1). (D) DLS measured the hydrodynamic diameter (Dh) for 116, QDNA, and their complex. Conditions: 116 (5 mM), QDNA dimer (4.5 mM ssDNA), and the supramolecular complex (4.5 mM ssDNA) in 78 mM potassium phosphate buffer pH 7.0, 467 mM KCl, 90:10 of H2O/D2O. The meaning of the subscripts in the peak labels in part A are: M = monomer; g = groove edge; WC = Watson–Crick edge; o = outer tetrads; and i = inner tetrads.

Figure 6

Molecular models derived from the NMR studies. (A) Depiction for the formation of the QDNA·116·QDNA complex (red/blue) from its constituents 116 (red) and QDNA2 ((d[T2AG3]4)2, light blue). The box on the right highlights the key NOE contacts highlighted in Figure B. (B) Molecular model of the interface between the subunits in the QDNA dimer. Top and side views of the molecular surfaces (i,ii) and polytube representation (iii,iv) of the tetrads at the 3′–3′ interface in QDNA2. (C) Molecular model of the interface between the SGQ (116) and one of the two QDNA subunits in the QDNA·116·QDNA complex. Top and side views (v,vi) of the molecular surface and polytube representation (vii,viii) of the tetrads at the interface between G6 (blue) and 1o (red) in the QDNA·116·QDNA complex. All models were prepared with Maestro 9.3 and minimized by MacroModel using the OPLS2005 force field. For the meaning of the subscripts, see the caption of Figure .

DSC measurements (Figure C) are consistent with the formation of a QDNA·116·QDNA complex (1.4 mM in ssDNA and 3.2 mM in 1) of greater thermodynamic stability than its individual constituents QDNA2 (1.4 mM in ssDNA) and 116 (3.3 mM in 1). The DSC thermogram reveals two main transitions, suggesting a stepwise dissociation of the QDNA·116·QDNA complex into two putative intermediate species of general formula QDNA·1 (where n = 1–3). The half-height of the peak (T1/2), corresponding to the QDNA·116·QDNA complex, is narrower than those of both QDNA2 and 116, a strong indication of a more cooperative dissociation of the former.[54] The enhanced stability and cooperativity are likely the result of conformational adaptability and the large number of attractive noncovalent interactions between the components in the QDNA·116·QDNA complex.

DLS experiments indicate the QDNA·116·QDNA complex has a hydrodynamic diameter of 5.8 nm (Figure D), which agrees with that (5.6 nm) obtained from the molecular model (Figures and S45). The corresponding Dh values for the individual components, QDNA2 (3.6 nm) and SGQ 116 (4.8 nm), are also in good agreement with those of the molecular models of 3.6 and 5.0 nm, respectively. The diameter of the QDNA·116·QDNA complex is smaller than the result of a simple addition of its individual constituents because of the interpenetration of the corresponding hydrodynamic spheres driven by attractive noncovalent interactions. This is also consistent with the molecular model (Figures and S45), which shows a compact structure with significant interpenetration of each individual QDNA component within the hydrodynamic shell of the SGQ 116.

The 2D NOESY spectrum of a sample containing 2 equiv of 1 (t48h; Figure Aiv) has one set of broad signals with NOE correlations, revealing interactions between QDNA2 and the SGQ 116 (Figures B and S44). Specifically, there are multiple NOE correlations, indicating interactions between the G6 tetrad in the QDNA and the subunits in the outer tetrads (1o) of the SGQ 116 such as (i) 1o/N1H with G6/N1H; (ii) 1o/N2HWC (WC = Watson–Crick edge) with G6/H8; and (iii) G6/H2β′ (Figure B) with 1o/N2HWC and multiple hydrogens of the aromatic group at the C8 of the guanines in the outer tetrads of 116 such as 1o/H13, 1o/H15, 1o/H14, and 1o/H11. The NOE correlations are consistent with a system where the outer tetrads (1o) of the SGQ 116 interacts with a QDNA having all anti glycosidic bonds (at least for G4–G6) and the individual strands adopting a right-handed helix.[51]

Combining our NMR data with those for the reported NMR structure of QDNA2,[51−53] we constructed a model for QDNA2 and the QDNA·116·QDNA complex using Maestro 9.3 and minimized it by MacroModel with the OPLS2005 force field (Figures and S45). The QDNA dimer has a 3′–3′ interface (tail-to-tail) (Figures , S45, and S46) with opposite polarity between the strands with the sugars of the two G6 tetrads at the interface pointing in opposite directions. For both, QDNA2 and the QDNA·116·QDNA complex, within each QDNA subunit, the strands have the same polarity, the tetrads are in a head-to-tail arrangement, and the sugars are directed in the same direction (Figure S46). In the QDNA·116·QDNA complex, the tail of the outer tetrad of the SGQ, 14(o), interacts with the tail of the G6 tetrad. Thus, the interaction at the interface between the QDNA and the SGQ leads to a partial overlap between the bases with opposite polarities (Figures C, S45, and S46). In 14(o), the proximity of the ribose O6 to the N2Hg (g = groove edge) in G6 allows the formation of intertetrad H-bonds that stabilize the QDNA·116·QDNA complex. The complex is also stabilized by the attractive Coulombic interactions between the polycationic 116 and the polyanionic QDNA2, which results in a net charge of +1 for the complex. The molecular model of QDNA·116·QDNA shows that many of the cationic ammonium groups at the end of the ester chains of 1 accommodate within the grooves of the QDNA subunits in close contact with the phosphate backbone.

Conclusions

The results presented here showcase the possibility of using 8ArGs to develop SALs for QDNA structures. This strategy is akin to fragment-based drug discovery methods, but one in which the fragments are connected using noncovalent interactions. Some of the attractive features of using SGQs as SALs include their precise supramolecular structure and their capacity to recognize relatively flat molecular surfaces. These findings are also amenable to SAR studies to optimize the binding for specific targets. Furthermore, because the recognition of protein surfaces shares similar challenges, it is conceivable that the presented strategy could be adapted to these purposes, and efforts along these lines are already showing promise.[55]

Experimental Procedures

General Methods

Unless otherwise specified, all reagents and solvents were obtained from commercial sources and used without further purification. Column chromatography was performed using silica gel 60, 0.04–0.063 mm (from Sorbent Technologies). For thin-layer chromatography (TLC) analysis, silica gel 60 F254 glass-backed plates from Sorbent Technologies were used. Visualization of the TLC plates was effected with UV light. All compounds were characterized with 1H, 13C, and 2D NMR techniques such as 1H–1H correlation spectroscopy and/or 1H–13C heteronuclear multiple quantum coherence. All chemical shifts are reported in parts per million relative to the residual undeuterated solvent as an internal reference. The following abbreviations are used to explain the multiplicities: s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet; and b, broad.

General NMR Methods

NMR experiments were recorded on (1) a Bruker Avance DRX-400 spectrometer equipped with a 5 mm PABBO BB-1 H/D Z-GRD probe with a nominal frequency of 400.15 MHz for proton; (2) a Bruker AVANCE AV-500 spectrometer equipped with a 5 mm PABBO probe with nominal frequencies of 500.13 MHz for proton and 125.76 MHz for carbon; or (3) a Bruker AVANCE III Ascend Aeon 700 MHz spectrometer equipped with a 5 mm cryoprobe with nominal frequencies of 700.26 MHz for proton and 176.08 MHz for carbon. The 1H NMR water suppression experiments were done in H2O/D2O (9:1), and sodium 3-(trimethylsilyl)propionate-2,2,3,3-d4 was used as the internal standard to calibrate the spectra at 0 ppm. For each sample, the more suitable water suppression experiments were used, either (1) standard presaturation pulse sequence with the excitation frequency set above the water peak around 4.7 ppm or (2) standard excitation sculpting technique.

DLS Experiments

Measurements of the hydrodynamic diameter (Dh) of the species in solution were done at 25 °C using the globular protein parameter. Prior to measurements, samples were centrifuged in a Daigger Mini-Centrifuge model Sprout, and then 0.30 μL of the supernatant was transferred to a quartz ultra micro cell. For the analysis of compounds 1–4 (4 mM) in 10 mM LCB with 100 mM KCl, a Zetasizer Nano ZS model ZEN3600 from Malvern with a 4 mW laser of 632.8 nm wavelength at a backscatter angle of 173° was used. The instrument software was Malvern Zetasizer version 7.10. For the analysis of QDNA2, SGQ 116, and the QDNA·116·QDNA system in 78 mM potassium phosphate buffer pH 7.0 with 467 mM KCl, a Brookhaven Instruments BI-90 Plus particle size analyzer with a diode laser at a scattering angle of 90°, a wavelength of 657 nm, and a laser power of 40% was used.

DLS measurements were done for the samples of QDNA2 dimer (4.5 mM ssDNA), 116 (5 mM 1), and the supramolecular complex (4.5 mM ssDNA) in 78 mM potassium phosphate buffer pH 7.0 with 467 mM KCl and 90:10 of H2O/D2O. The sample with the supramolecular complex was also used for its NMR characterization.

DSC Experiments

The experiments were performed on a VP-DSC microcalorimeter from MicroCal, and Origin (v.7) was used for data processing. Measurements were made using a heating rate of 1 °C/min. All samples were centrifuged in a Daigger Mini-Centrifuge model Sprout and degassed prior to their transference to the DSC instrument. The thermodynamic stability of the species formed by compounds 1–4 (4 mM 1) was evaluated in 10 mM LCB with 100 mM KCl. For the structural analysis of SGQ 116 as a SAL for QDNA2, DSC measurements were done for QDNA2 (1.4 mM ssDNA), SGQ 116 (3.3 mM 1), and the QDNA·116·QDNA complex (1.4 mM ssDNA and 3.2 mM 1) in 78 mM potassium phosphate buffer pH 7.0 with 467 mM KCl and 90:10 of H2O/D2O.

Measurement of TO Fluorescence Spectra

The samples evaluated have 10.7 μM TO with or without 120 μM 8ArG, in 10 mM LCB with 10 mM KCl and 90 mM LiCl. We chose the concentration of 120 μM 8ArG because it is also the concentration of 8ArGs in the stock solution used for the dose–response FRET melting assays. Therefore, the samples were prepared as described below within the subsection titled “Buffered Solutions of the ODNs and of 8ArGs”. The only modification to this procedure is that at the end of the sample preparation, the TO was added from a stock solution of TO in dimethyl sulfoxide (2 mM). All fluorescence measurements were made using the spectrofluorophotometer RF-5301PC from Shimadzu (λex = 501 nm, emission range = 518–720 nm, at ∼25 °C).

FRET Melting Assays

Oligonucleotide Stock Solutions in Deionized Water

All oligonucleotides were purchased from Sigma-Aldrich as reversed-phase high-performance liquid chromatography (RP-HPLC)-purified. The lyophilized ODNs were dispersed in deionized water, and the absorbance of a dilute sample at 260 nm (after 5–10 min at 90 °C) was measured. The experimental concentrations of the stock solutions were calculated using the average value of the absorbance, a cell path length value of 10 mm, and the ODN molar extinction coefficients reported by the supplier, ε (M–1 cm–1): 215 000 (F21T), 236 800 (FmycT), 199 100 (Fkit2T), and 258 900 (FduplexT). The DNA sequences used on the FRET melting assays were modified at the 5′ and 3′ end with 6FAM (F) and TAMRA (T), respectively. F21T: FAM-G3TTAG3TTAG3TTAG3-TAMRA, FmycT: FAM-TTGAG3TG3TAG3TG3TAA-TAMRA, and Fkit2T: FAM-G3CG3CGCGAG3AG4-TAMRA.

Buffered Solutions of the ODNs and of 8ArGs

All solutions for 8ArGs (∼120 μM) and the ODNs (∼4 μM) were in 10 mM LCB with KCl/LiCl (10 mM/90 mM or 1 mM/99 mM), 100 mM NaCl, or 100 mM LiCl. The solutions were prepared by mixing the DNA or 8ArGs, from a stock solution in deionized water with aliquots from stock solutions of LCB (100 mM), KCl (100 mM), NaCl (1 M), or LiCl (1 M) and additional deionized water. K+-containing solution in LCB with 10 mM KCl/90 mM LiCl was prepared for F21T, FduplexT, and selected 8ArGs. For FmycT, Fkit2T, and also selected 8ArGs, K+-containing solutions in LCB were prepared with 1 mM KCl/99 mM LiCl which is the suggested KCl concentration for oncogene sequences.[56] The samples of the ODNs were subsequently submitted to the protocol described below to promote the DNA folding into the corresponding QDNA or dsDNA structures. The samples of 8ArGs were stored in the refrigerator for at least 2 h before their use in the FRET melting assays.

DNA Folding into QDNA Structures

To induce the folding of the oligonucleotides into unimolecular QDNAs, the K+-, Na+-, and Li+-containing solutions of F21T, FmycT, and Fkit2T were heated at 90 °C for 5–10 min, kept on ice for ∼7 h, and then stored in the refrigerator (4 °C) for at least 12 h.

Protocol for FRET Melting Assays[57,58]

The final K+-, Na+-, and Li+-containing samples were prepared on a 105 μL scale by diluting the ODNs (F21T, FmycT, or Fkit2T) up to 0.5 μM and adding different concentrations of 8ArGs (0.5–64 μM) from the corresponding buffered solution.

FRET assays were performed as a high-throughput screen in a 96-well plate using a total sample volume of 25 μL, using a Stratagene MX3005P real-time PCR (qPCR) instrument. For the melting profiles after excitation at 492 nm, the emission of FAM at 520 nm was monitored from 25 to 95 °C. The samples were left stabilizing for 1 min at 25 °C, and the temperature was increased by 1 °C for 70 cycles, with an average end-point measurement of the temperature (average from three measurements). Analysis of the data was done using Excel and OriginPro8 to normalize the emission of FAM between 0 and 1. T1/2 is defined as the temperature for which the normalized emission is 0.5, and the values are the mean of 2–4 experiments.

NMR Studies for the Formation of a Complex between 1 and a Dimeric QDNA Structure

General Experimental Details

All NMR analyses were performed using Shigemi NMR tubes. The nanopure water and buffer used for the preparation of all samples were filtered through Fisherbrand 0.45 μm nylon syringe filters. The DNA sequence 5′-d(TTAGGG)-3′ was purchased from Sigma-Aldrich as RP-HPLC-purified, and concentration values were calculated from the nanomoles reported on the technical data provided by Aldrich.

Formation of the Dimeric QDNA Species

To prepare the DNA stock solution, the DNA was dispersed in concentrated potassium buffer with KCl (pH ≈ 7), nanopure water, and D2O. The resulting solution was kept at 10–11 mM of the DNA strand to promote the formation and dimerization of the tetramolecular QDNAs. This process was monitored by 1H NMR analysis.

Preparation of Samples for the Titration, DLS, and DSC Experiments

Once it is confirmed by NMR that the main species in the DNA stock solution is the dimer of tetramolecular QDNA structures, we proceed to dilute the sample to reach 4.5–5.0 mM DNA and divided the sample into two fractions with equal amounts of the QDNA dimer. One fraction is for the direct analysis of the QDNA dimer species by DLS and DSC experiments. The other fraction of the QDNA dimer is used for the titration experiment.

Titration of 1 into a Sample of the QDNA Dimer

In a typical titration experiment, before the addition of aliquots of compound 1, the sample that originally only has the QDNA dimer species is freeze-dried and suspended in H2O/D2O (9:1). The process of freeze drying, resuspension, increasing the amount of compound 1, and NMR analysis were repeated for subsequent data points in the titration experiment. Once the supramolecular complex had fully formed (Figure Aiv), further NMR, DLS, and DSC experiments were performed.

Molecular Modeling

General Methods

The molecular models were constructed using Maestro v. 9.5 and minimized with MacroModel v. 8.0.315 (both form Schrodinger, LLC, New York, 2007) using the force field OPLS2005 and water as the continuum solvent. The details for the construction and structural analysis of SGQ 116 were previously reported by our research group[4,38] and used as the starting point to create the molecular model of SGQ 316, 38.

Molecular Models for the Structures of the QDNA Dimer and the Supramolecular Complex

The “initial structure” for the molecular model was prepared by removing the thymines and adding the corresponding OHs to the first structure in the pdb 1NP9.[52,53] The constructed molecular structure of QDNA2 was created by using two entries of the initial structure and applying the necessary rotations and alignments to minimize repulsive interactions. Construction of the model for the supramolecular complex QDNA·116·QDNA was made by combining the constructed molecular model of the QDNA2 and the model of SGQ 116 which was developed by our research group,[4] followed by the necessary rotations and alignments. Each molecular model was minimized with MacroModel, the force field OPLS2005, water as the continuum solvent, and constrains applied to all guanines and metals.