Single-molecule kinetics and footprinting of DNA bis-intercalation: the paradigmatic case of Thiocoraline

DNA bis-intercalators are widely used in molecular biology with applications ranging from DNA imaging to anticancer pharmacology. Two fundamental aspects of these ligands are the lifetime of the bis-intercalated complexes and their sequence selectivity. Here, we perform single-molecule optical tweezers experiments with the peptide Thiocoraline showing, for the first time, that bis-intercalation is driven by a very slow off-rate that steeply decreases with applied force. This feature reveals the existence of a long-lived (minutes) mono-intercalated intermediate that contributes to the extremely long lifetime of the complex (hours). We further exploit this particularly slow kinetics to determine the thermodynamics of binding and persistence length of bis-intercalated DNA for a given fraction of bound ligand, a measurement inaccessible in previous studies of faster intercalating agents. We also develop a novel single-molecule footprinting technique based on DNA unzipping and determine the preferred binding sites of Thiocoraline with one base-pair resolution. This fast and radiolabelling-free footprinting technique provides direct access to the binding sites of small ligands to nucleic acids without the need of cleavage agents. Overall, our results provide new insights into the binding pathway of bis-intercalators and the reported selectivity might be of relevance for this and other anticancer drugs interfering with DNA replication and transcription in carcinogenic cell lines.


S1. Force-jump experiments
We performed force-jump experiments to recover the equilibrium FEC of Thiocoraline-DNA and to measure the kinetic rates of binding. The experiments were performed in two ways: (i) bidirectional stepwise force-jumps, and (ii) unidirectional force-jumps from low force. In the first type of experiments ( Figure S1a), the force was increased stepwise every ∼5 min up to a maximum preset force. Then the protocol was reversed and the force stepwise decreased down to the initial value. By doing so, we obtained a forward and backward relaxation curve for each force. The relaxations were well described by a single-exponential function, from which we obtained the forward and backward asymptotic extensions ( Figure S1b). The equilibrium extension at each force x In these experiments, the force was increased from 2 pN to a preset value, and the relaxation of the molecular extension was followed until reaching saturation (∼ 10 min). Consecutive force-jumps between 2 pN and the different preset forces were performed in this way, and fitted to a single-exponential function ( Figure S1d). The x [T hio] eq (F ) values obtained with both methods were fully compatible. From this alternative measurement, we also obtained the total kinetic rate k tot (F ) at each force ( Figure 3, main text). All force jumps were performed with a force-feedback operating at 1 kHz in which force was changed in less than 0.1 s and maintained constant to the preset value. Relaxation of the molecular extension (green) for two forces from panel (a). Data is fitted to a single exponential function (purple) from which the equilibrium extension x [Thio] eq (F ) at each force is determined. (c) Typical unidirectional force-jump experiment. The force (red) is increased from 2 pN to a preset value, and the relaxation of the molecular extension (green) is measured until reaching saturation. Then the force is decreased again to 2 pN and the measurement repeated for a set of different forces. (d) Relaxation of the molecular extension (green) at two forces from panel (c). Data is fitted to a single exponential function (purple), obtaining x [Thio] eq (F ) and the total kinetic rate k tot at each force. Experiments correspond to 100 nM Thiocoraline. .

S2. DNA stretching experiments in the overstretching region
The DNA pulling curves presented in the main text are performed up to a maximum stretching force of 40 pN. This force is below the overstretching transition (F OS ∼63 pN at 100 mM NaCl), and allows us to obtain metastable FECs at the highest pulling speed of 1.5 µm/s for all Thiocoraline concentrations ([Thio]≤1 µM). However, if the maximum pulling force is increased to 70 pN, FECs show longer molecular extensions and larger hysteresis as compared to 40 pN, due to the greater intercalation achieved at the highest forces ( Figure   S2a). In fact, at 100 nM (concentration used for the experiments shown in the main text), hysteresis is observed also at the fastest pulling speed of 1.5 µm/s ( Figure S2b), and we cannot pull fast enough to reach the metastable regime at this higher force. As previously observed for other intercalators [S1 , S2 ], a tilted plateau related to the overstretching transition is observed. If experiments are performed at lower bis-intercalator concentrations, the equilibration time of the binding reaction τ eq increases and the metastable FECs can be obtained again at the fastest pulling speed ( Figure S2d-f, red). As expected, when we decrease ligand concentration the molecular extension becomes shorter (due to the lower fraction of ligand bound at low forces), and the slope of the force plateau decreases approaching the overstretching curve of naked dsDNA. These findings are in agreement with a mechanism in which the bis-intercalator binds to B-DNA, elongating and unwinding the double helix, and increasing the energy required to force-melt or fully unwind the bis-intercalated DNA (converting B-DNA into S-DNA). At concentrations below 10 nM, fast pulling experiments show a molecular extension slightly longer (<5%) than naked dsDNA and an overstretching plateau that is slightly tilted, indicative of a few bound intercalators ( Figure S2e, f). Remarkably, at this low concentrations slower pulling rates show negative hysteresis along the overstretching region (the force is higher in the releasing as compared to the stretching part of the pulling cycle, Figure S2d-f). These results suggest that Thiocoraline is able to bind either force-melted regions or S-DNA, stabilizing the DNA double helix and therefore reducing the molecular extension at forces above the overstretching transition. In these experiments, once the force is reduced below the overstretching transition, the molecular extension remains longer than during stretching, confirming that Thiocoraline has bis-intercalated DNA at forces above the overstretching plateau. This is particularly visible in the 3 nM experiment ( Figure S2e). In this condition, the stretch part of the cycle for the fast (dark red) and slow (dark blue) pulling rate are very similar below the overstretching region. However, during releasing at the slow pulling rate (light blue) the apparent molecular extension is ∼700 nm longer than that observed at a fast pulling rate (light red), suggesting that most of the binding has happened at F ≥ F OS . Recent evidence that force-melted DNA and S-DNA coexist at physiological conditions, and their similar extension, makes difficult to unambiguously attribute binding to any of both [S3 , S4 ].

S5. Three-state kinetic model
We propose a model in which the binding reaction has an on-pathway intermediate state in which only one of the two intercalating moieties is bound to DNA (Figure 6a, main text).
The model considers two different rates for binding: one rate corresponding to binding one chromophore when the ligand is unbound (α on ) that leads to the mono-intercalated intermediate state. This rate is expected to be concentration-dependent. Then there is a second binding rate (α on ) that corresponds to intercalate the second moiety (once Thiocoraline is in the mono-intercalated intermediate state) in order to reach the fully bis-intercalated state.
This kinetic rate is expected to be independent of ligand concentration. Finally, we consider that there is a third kinetic rate (α off ) that corresponds to the rate at which any of the intercalating moieties de-intercalates. For simplicity reasons, we considered a minimal model in which this unbinding rate is the same both to remove an intercalating moiety from the fully bound bis-intercalated state, or the mono-intercalated intermediate state.
The kinetics of the model can be described by means of the following master equations: dP ++ dt = α on (P +− + P −+ ) − 2α off P ++ (S1a) dP +− dt = P −− α on + P ++ α off − (α off + α on )P +− (S1b) where P ++ , P +−/−+ and P −− correspond respectively to the probabilities of a site to be: in the fully bound state (bis-intercalated), in the intermediate state (mono-intercalated), or non-occupied (naked DNA). The occupancy of the intermediate state can be grouped by defining P 1 = P +− + P −+ :

S5.1. Determining the microscopic rates α on and α off from wash-off experiments
For wash-off experiments it can be assumed that α on =0 (as no free-ligand is present in the buffer), and the set of equations can be analytically solved by finding the eigenvalues and eigenvectors of the reduced system: The eigenvalues of this system are determined as: The system therefore shows two characteristic timescales (a faster one, λ − and a slower one λ + ) that can be directly related to the two off-rates observed in the wash-off experiments (k off,fast , k off,slow ), finding that: From the experimental values k off,fast and k off,slow at different forces (Figure 5b, main text) we can determine the force-dependence of the intrinsic rates α on (F ) and α off (F ) (Figure 6b, main text). Using these rates, the master equations can be solved for any set of initial conditions (P ++ (0), P 1 (0) and P −− (0)), finding solutions that will be of the form: where A fast and A slow are the amplitudes of the fast and slow relaxation modes. In our experiments, we have access to the change in molecular extension as bis-intercalators bind or unbind. From the model depicted in Figure 6a (main text), the change in molecular extension x(t) can be related to the probability of the bis-intercalated and intermediate states as: where we have assumed that the intermediate state contributes to a fractional extension, f , of that corresponding to a fully bound bis-intercalator (as we will later see f ∼ 0.4 maximizes the agreement between model and experiments). Consequently, it is useful to define the normalized fractional elongation of the molecule χ(t), that goes from 0 (naked DNA) to 1 (maximal elongation due to full bis-intercalation): For the wash-off experiments, the kinetics of the system (Eq. S2) can be analytically solved using the α on (F ) and α off (F ) values shown in Figure 6b (main text). By assuming, that the molecule is initially in the fully bis-intercalated state (P ++ (0) = 1, P 1 (0) = P −− (0) = 0), we can determine the expected amplitude of the slow and fast decay rates (A off,fast and A off,slow ). The ratio between these amplitudes (A off,slow /(A off,slow +A off,fast )), can be compared to the experimental values obtained, finding compatible results ( Figure 6c and Figure S7). The agreement between theory and experiment is maximized for an intermediate state that has an extension 0.4 times that of a fully bound bis-intercalator ( Figure S7).
This value is close to half that of a fully bound bis-intercalator, supporting the hypothesis that the intermediate state is mono-intercalated. To convert the molecular extension measured in the experiments to the fractional elongation χ(t), we assumed a maximum binding density (n = 4) of bis-intercalated ligands, with each bis-intercalated ligand elongating DNA by ∆l 0,Thio = 0.68 nm: where N bp is the number of base pairs of the DNA molecule, n is the binding density (=4) and φ F,WLC a WLC correction to the molecular extension due to the force applied to the molecule. For this correction we used the fact that the elastic properties of intercalated DNA are similar to those of naked DNA, and therefore N bp ∆l 0,Thio φ F,WLC ∼ 2x DNA (F ) obtaining the simplified expression.
For the force-jump experiments, as far as α on (F ) ≥ α on (F ), the experiments are well described. In these experiments, we assumed that the molecule starts in equilibrated conditions at 2 pN (dP ++ /dt = dP 1 /dt = dP −− /dt = 0) with rates α off (F = 2 pN), α on (F = 2 pN) and α on (F = 2 pN). The experiment starts with a force-jump in which these kinetics rates are changed, and therefore the master equations are solved by considering the previous initial conditions and the new kinetic rates at each force (15-55pN). The values of α off (F ) and α on (F ) are determined using an extrapolation of the wash-off experiments (blue and red regions in Figure 6b, main text). In this situation α on (F), is the rate-limiting step of the reaction, explaining the fact that force-jump experiments show little variation when the concentration is increased from 100 nM to 1 µM .
Finally, we assessed the binding kinetics for experiments in which the initial conditions correspond to naked DNA (P −− = 1), with the determined kinetic rates. In this case, the binding kinetics are well described considering an α on that increases with force with a similar slope as for α on (Figure 6b, main text). These values are found to be compatible with the Figure S9. Consecutive unzipping curves in the presence of Thiocoraline. In this figure several DNA unzipping curves are shown superimposed using a different color for each curve. Each unzipping curve contains a few binding events (typically between 0 and 2). By overlapping a large number of unzipping curves it can be seen that the ligand binds to many positions along the DNA sequence. Statistics on the preferred binding sites of the ligand are obtained by collecting a large number of unzipping curves (N > 100) where the different binding sites become populated. Figure S10. Sequence-selectivity of Thiocoraline clamping positions. Normalized probability of observing a binding event in which Thiocoraline clamps the central dinucleotide step XX (of a tetra nucleotide motif NXXN) if different reading frames are considered. Binding at the tetranucleotide position were blockage is observed (reading frame 0, red) is compared to alternative distances from the peak position (black, blue, purple, cyan). Sequence-selectivity is only observed for the reading frame 0, whereas other reading frames show probabilities compatible with a random distribution (dashed gray). Only for the reading frames ±1 (black and blue) a lower peak at CG is also observed, indicative of the error associated with the method.