A Photoresponsive Stiff‐Stilbene Ligand Fuels the Reversible Unfolding of G‐Quadruplex DNA

Abstract The polymorphic nature of G‐quadruplex (G4) DNA structures points to a range of potential applications in nanodevices and an opportunity to control G4 in biological settings. Light is an attractive means for the regulation of oligonucleotide structure as it can be delivered with high spatiotemporal precision. However, surprisingly little attention has been devoted towards the development of ligands for G4 that allow photoregulation of G4 folding. We report a novel G4‐binding chemotype derived from stiff‐stilbene. Surprisingly however, whilst the ligand induces high stabilization in the potassium form of human telomeric DNA, it causes the unfolding of the same G4 sequence in sodium buffer. This effect can be reversed on demand by irradiation with 400 nm light through deactivation of the ligand by photo‐oxidation. By fuelling the system with the photolabile ligand, the conformation of G4 DNA was switched five times.


.1 FRET melting assays
Fluorescence resonance energy transfer (FRET) melting assays were performed according to the procedure reported by De Cian and co-workers [1] on Roche LightCycler 480 qPCR instrument. In these assays, the oligonucleotides of interest were obtained labelled at the 5' and 3' ends with FAM (a fluorescence donor) and TAMRA (a fluorescence quencher) respectively.
In the folded state, proximity of the donor and quencher mean that FAM fluorescence is not observed since energy is transferred non-radiatively to TAMRA by FRET. As the temperature is raised and the secondary structure denatures, the fluorophores move further apart and hence the fluorescence signal increases. From the resulting curve, the characteristic melting temperature (T1/2) is defined as that at which the normalised fluorescence signal equals 0.5. The change in melting temperature (ΔTm) induced by a small molecule ligand compared to that of the oligonucleotide in the absence of ligand provides an indication of the ligand's ability to stabilise the G4 structure. The assay is shown in schematic form below: S3 All oligonucleotides used were purchased from Eurogentec (Belgium), purified by HPLC and delivered dry. Oligonucleotide concentrations were determined by UV-absorbance using a NanoDrop 2000 Spectrophotometer from Thermo Scientific. The oligonucleotides used were: µM, 2 µM, 5 µM or 10 µM. Each sample was tested in duplicate on the same plate, and each experiment was tested in triplicate to assess the reproducibility of all results. Appropriate control experiments were also carried out for each sample set. Data processing was carried out using Origin 9, with ΔT1/2 used to represent ΔTm.

Circular dichroism titrations
Circular Dichroism (CD) titrations were recorded using a Jasco J-810 spectrometer fitted with a Peltier temperature controller. Measurements were taken in a quartz cuvette with a path length of 5 mm, at 20 °C, at a 100 nm / min scanning speed at 1 nm intervals, with a 1 nm bandwidth.
The CD spectra were recorded between 450 and 200 nm, and baseline corrected for the buffer used. The oligonucleotide sequence used was: telo23 (human telomeric G-quadruplex): 5'-

Circular dichroism kinetic studies
Kinetic studies of the conformational switch were undertaken by monitoring the time course of the evolution of ellipticity at 273 nm at 20 °C following the addition of ligand. (E)-1 (10 eq.) was added to the bottom of an empty cuvette (5 mm path length) in 4 µL DMSO. Following stabilisation of the background CD signal, telo23 (1 eq., 1 mL of 4.22 µM solution in 100 mM sodium phosphate buffer, pH 7.4) was injected rapidly to facilitate mixing and the change in CD monitored for 2500 sec. The addition point of ligand was visible by a spike in HT voltage and t = 0 defined as the moment this returned to a stable value ( Figure S7b). Three independent repeats were conducted and the change in ellipticity at 273 nm plotted as a function of time.
Fitting to a single exponential function (Δθ = A*exp(-t/τ) + c) was carried out using Prism 7 to determine of the characteristic folding time (τ).

UV-visible spectroscopy
UV spectra were recorded on a Thermo Scientific BIOMATE 3S UV-vis Visible Spectrophotometer at ambient temperature. Measurements were taken in a quartz cuvette with a path length of 10 mm using in slow scanning mode at 0.5 nm intervals. The UV-visible spectra were recorded between 450 nm and 200 nm and baseline corrected for the buffer used. suppression. DOSY experiments were conducted using the PGSE-based DOneshot sequence. [2] The final NMR samples contained 600 µL of 175 µM telo23 DNA. Samples were annealed before use by heating for 2 minutes at 90°C and then placed immediately into ice. Aliquots of (E)-1 (10 mM in DMSO-d6) were added to yield 2 equiv. portions and mixed thoroughly. NMR spectra were recorded immediately after the addition of ligand. Data were processed using MestReNova software (version 11.0.2).

Photoirradiation experiments
The 800 nm fundamental output of a Ti:Sapphire ultrafast amplifier (Libra, Coherent) was used to generate 397 nm via second harmonic generation for photoirradiation experiments. The spectral profile of the 397 nm light used is shown in Figure S11. The measured beam spot size (diameter) at the sample region is 0.4 cm and the energy used to irradiate solutions was 30 mW with a power density of 1.91 W cm -2 . Experiments involving (E)-1 in the absence of telo23 were conducted using 1.5 mL solution in a 3 mL quartz cuvette with a path length of 10 mm.
The concentration of the ligand was 10 µM in sodium phosphate buffer (100 mM, pH 7.4). The photoreaction was followed using UV-visible spectroscopy using the procedure detailed above (Section 1.4). Experiments involving telo23 were conducted using 1 mL solution in a 2 mL quartz cuvette with a path length of 5 mm. The concentration of DNA was 4.22 µM in sodium phosphate buffer (100 mm, pH 7.4). The effect of photoirradiation of G4 topology was S6 monitored by circular dichroism spectroscopy using the procedure detailed above (Section 1.2).
For experiments involving the repeated addition of (E)-1, 10 eq. of ligand were added by aliquot (82 µL) of 600 µM (E)-1 solution in water. The system was allowed to equilibrate for 20 min at 20 °C following addition of ligand before the CD spectrum was recorded.
Photoirradiation of NMR samples was conducted by transferring the sample a 2 mL quartz cuvette with a path length of 5 mm.       Crude photoproduct Ketone 7 S16 Figure S15: CD spectra of telo23 G4 before and after 400 nm irradiation in absence of (E)-1 Figure S16: Eventual photofatigue of the (E)-1/telo23 conformational switch showing spectra initial spectrum (green), spectra following addition 10 eq. portions of (E)-1 (black; switches 1, 3, 5, and 7) and spectra following photoirradiation (red; switches 2, 4, 6 and 8). Though a topological switch can be clearly observed over the 8 cycles, little recovery of the negative band at 265 nm and over 50% reduction in the maximum at 295 nm is observed following switch 5.

Supplementary tables
We performed a docking calculation of (E)-1 to telo22 G4 DNA (PDB code: 143D) [3] in order to predict all the available high-affinity binding modes. The ligand molecule was optimised using PerkinElmer Chem3D software using the MM2 forcefield. Docking calculations were performed using AutoDock Vina [4] with the DNA structure kept fixed in its original crystal conformation throughout the docking procedures. The highest affinity binding pose was then submitted to long microsecond molecular dynamics simulations in order to predict the stability of the binding pose.

Standard (unbiased) molecular dynamics simulations
All the simulations were performed using the Grgmacs-5.0 software package. [5] The recently introduced parm-BSC1 force field was used for the DNA parameterization. For the ligand, the General Amber Force Field (GAFF) were used to generate parameters. [6,7] The charges were calculated using the restrained electrostatic potential (RESP) fitting procedure. [8] The RESP fit was performed onto a grid of electrostatic potential points calculated at the HF/6-31G(d) level as recommended by the force-field designers and recent literature. [7,8] To start the simulations, each docked (E)-1/telo22 complex was solvated in a cubic box with the dimension of 74 x 74 x 74 Å 3 along with 12982 TIP3P explicit water molecules. [9] An extra 19 Na+ ions were added to neutralize the system. [10] The (E)-1/telo22 complexes were minimized prior to the equilibration and production run as follows: the minimization of the solute hydrogen atoms on the DNA and the ligand was followed by the minimization of the counterions and the water molecules in the box. In the next step, the DNA backbone along with the all the heavy atoms on the ligand were kept frozen, and the solvent molecules with counterions were allowed to move during a 50 ps MD run, to relax the density of the whole system. In the next step the nucleobases were relaxed in several minimization runs with decreasing force constants applied to the DNA backbone atoms, however, a few phosphate atoms were kept restrained with a force constant of 2.39 kcal.mol-1 . Å -2 . After the full relaxation, the system was slowly heated to the room temperature to 300K using V-rescale thermostat with a coupling constant of 0.5 ps employing an NVT (constant-temperature, constant-volume) ensemble. [11,12] As the system reached the temperature of interest, the equilibration simulation was performed for 10000 ps (10 ns) using an NPT ensemble with S18 Berendsen thermostat and Berendsen barostat, and 0.5 ps was used again as the coupling constant for both temperature and pressure, respectively. [13] Finally, the production run was set for 1000000000 ps (1μs) using Nose-Hoover thermostat [14,15] and Parrinello-Rahman barostat [16] with the same coupling constant as previously taken in the equilibration simulation in the NPT ensemble. All the simulations were carried out under the periodic boundary conditions (PBC). The particle-mesh Ewald (PME) method was used to calculate the electrostatic interactions with in a cut-off of 10 Å. [17] The same cut-off was used for Lennard-Jones (LJ) interactions. All simulations were performed with a 1.0 fs time step.

Well-tempered metadynamics
We performed a well-tempered metadynamics (WTMetaD) simulation of (E)-1 binding to the telo22 G4 DNA. [18] The WTMetaD also helps to understand the binding/unbinding mechanism of (E)-1. The simulation started with a well-equilibrated structure generated from the docking pose-1 (vide infra). As discussed above, the docking poses had been run for 1 μs in MD simulations; however, the preliminary structure for the WTMetaD simulation was taken after 20 ns of MD simulation, which was found to be a stable ligand binding conformation.
The same systems and MD settings as described previously were used for the WTMetaD simulation. The plumed 2.3 plugin was used to carry out the simulation with the Gromacs-5.0.7 code. [19] The bias potential was calculated according to the WTMetaD scheme as follows: where the deposition rate, ω, and deposition stride, τG, of the Gaussian hills were set to 0.358 kcal.mol-1·ps-1 (1.5 kJ.mol-1·ps-1) and 1.0 ps, respectively. The bias factor (T + ΔT)/T was set to 15, and the final free energy surface (FES) was calculated as follows: where the V(s,t) is the bias potential added to the Collective Variables (CV) used and the T represents the simulation temperature. ∆T is the difference between the temperature of the CV and the simulation temperature. The bias potential is grown as the sum of the Gaussian hills deposited along the chosen CV space and finally the sampling of particular CV space can be controlled with the tuning of the ∆T parameter.

S19
To describe the different ligand binding conformations during the metadynamics simulation, we used two Collective Variables (CV) and they are: (i) the distance (d) between the centre of mass (COM) of the middle G-tetrads and the heavy atoms of (E)-1 and (ii) a torsion angle is used. The torsion is defined between two points in the ligand to two points in the G-tetrad. In the ligand, the two points are considered as the COM of the two indane residues.

MD analysis results
We carried out 1 μs long MD simulation of the lowest-energy docking pose. Figure

Metadynamics results
The sampling of the ligand binding/unbinding mechanism was studied using the WTMetaD simulation. The simulation started with a well-equilibrated structure from the docking pose discussed above. As the simulation progressed, several binding/rebinding events were observed. Figure S18 depicts the separation distance between the centre of mass (COM) of the bases containing the central G-tetrad of telo22 i.e., the COM of G3, G9, G15, and G21 to the COM of the heavy atoms of the ligand (E)-1. Carefully looking at the distance separation and molecular visualisation, it is apparent that the ligand unbinds from DNA within the first ~15 ns and rapidly starts sampling the unbound state, i.e. the solvated state. Before the unbinding, the ligand visits the major groove and slides on top of the DNA interacting with A7 and A19 bases with stacking interactions (pose B in Figure S19). While sampling in this particular state, the ligand tail is partially interacting with thymine bases such as T5, T18 and T18 which are present on top of the DNA and always fraying toward solvent (pose C in Figure S19). As the time progresses, the separation distance increases and the ligand lose interactions with bases A7 and A19. At the final stage of unbinding, the ligand is interacting with T17 and T18 bases and leaves toward solvent (pose D in Figure S19). After the complete dissociation, the ligand mostly samples the unbound state and partially interacts with all the solvent exposed thymine bases until approximately 90 ns, where it enters to the major groove. During this process, all three G-tetrads are partially broken with rapid conformational change in the DNA which helps the ligand to partially intercalate (pose E in Figure S19). The 1st intercalation takes place at ~95 ns, making stacking interactions with G9 base which remain after ~10 ns of rigorous sampling. In the later stage of the simulation, the ligand returns towards G9 base and finally a full intercalation takes place at ~150 ns (pose F in Figure S19). At this stage of the simulation, the ligand mainly interacts with the guanine bases in the G-tetrads such as G3, G4, G9, G21 and G20, and also with A7 in a sandwich manner. After sampling for a further ~35 ns, the ligand leaves the DNA and again starts sampling the solvent region. In the meantime, several binding/unbinding processes is occurred (see Figure S18, the sampling region between ~350−525ns) and the DNA backbone RMSD is already reached to ~1.0 nm which corresponds towards the denaturation of the telo22 G4 structure. At ~525 ns, the ligand again enters to the binding pocket and intercalates with G21, G8, G15, G2 and G14 (see pose G in Figure S19).
Finally, at ~582 ns, the telo22 G4 begins to unfold, with the backbone RMSD reaching ~1.5nm within the next 100 ns. These findings (see main manuscript) lead us to conclude that ligand (E)-1 is able to unfold the telo22 G4 DNA upon binding thorough rapid major groove binding and intercalation, reinforcing our experimental observations. S21 Figure S18: The separation distance between telo22 and (E)-1 in the WTMetaD simulation.
The highlighted sections in the curve represents all the possible events on the unfolding of telo22 DNA and each successive events is denoted with the corresponding colour codes. Figure S19: The schematic representation of the unbinding of (E)-1 and the unfolding of telo22 DNA sampled in the WTMetaD simulation.

Synthetic procedures and compound characterisation General experimental
Chemicals were purchased and used without further purification. Dry solvents were obtained by distillation using standard procedures, or by passage through a column of anhydrous alumina using equipment from Anhydrous Engineering (University of Bristol) based on the Grubbs' design. [20] Reactions requiring anhydrous conditions were performed under N2; glassware and needles were either flame dried immediately prior to use, or placed in an oven (150 °C) for at least 2 h and allowed to cool in a desiccator or under reduced pressure. Liquid reagents, solutions or solvents were added via syringe through rubber septa; solid reagents were added via Schlenk type adapters. Reactions were monitored by TLC on Kieselgel 60F254 (Merck), with UV light (254 nm) detection and by staining with basic potassium permanganate solution. Flash column chromatography was performed according to Still and co-workers, [21] using Proton NMR was consistent with literature data. [22] (Z)-3: Proton and carbon NMR were consistent with literature data. [22] (E)-6,6'-di(pyridin-4-yl)-2,2',3,3'-tetrahydro-1,1'-biindenylidene, (E)-4 Bromide (