Unraveling the physical chemistry and the mixed binding modes of complex DNA ligands by single molecule stretching experiments

Abstract

In this work we present a complete methodology to unravel the physical chemistry and the mixed binding modes of complex DNA ligands. Single molecule stretching experiments were performed with complexes formed between a DNA binding drug that exhibits multiple mixed binding modes (Berenil) and the biopolymer. From these experiments we determine the changes of the two basic mechanical properties, the contour and persistence lengths, as a function of the drug concentration in the sample. Combining a modeling analysis for the two mechanical properties, we were able to extract the physicochemical parameters of the interaction and to determine the effective binding mechanisms. In particular, we have shown that in this case the binding modes can be modulated by changing the ionic strength of the surrounding buffer: for high ionic strengths (150 mM), Berenil behaves as a typical minor groove ligand in its interaction with λ-DNA; while, for low ionic strengths (10 mM), the drug also partially intercalates into the double-helix. The methodology developed in the present analysis can be promptly applied to other complex DNA ligands, therefore allowing one to investigate and decouple different binding mechanisms.

---

1 Introduction

DNA can interact with small molecules such as drugs and other types of ligands in many different ways. Classic intercalation, for instance, is characterized by the insertion of a flat aromatic molecule between two adjacent DNA base pairs. Such process induces strong structural perturbations on the double-helix structure, increasing the DNA contour length and unwinding the double-helix.^1–6 Groove binding, on the other hand, is usually characterized by a combination of electrostatic, van der Waals and hydrogen bonds between the ligand molecules and the DNA minor or major groove floor.^4,7 This kind of interaction can also induce structural perturbations on the double-helix, such as bendings and twists, depending on the structure and on the net charge of the ligand. Finally, ligands such as some platinum-based compounds^8–11 and drugs from the family of furocoumarins^12–14 can also form covalent bonds with the DNA base-pairs.

There is also a considerable number of ligands that can interact with DNA by multiple mixed binding modes, for example, combining intercalation and groove binding. In some cases the ligand molecule has distinct portions which bind to DNA by different modes.^15–22 In other cases there is only a single binding mode for the entire ligand molecule, which can be changed upon determined conditions such as the type of surrounding buffer and the DNA base-pair sequence.^23,24 Some studies have suggested that there is a correlation between ligands that exhibit multiple mixed binding modes and those that exhibit anticancer activity by inhibiting the activity of topoisomerases.^25,26 Therefore, there is a considerable interest in identifying and characterizing such types of ligands and their interactions with the DNA molecule.

Berenil [1,3-bis(4′-amidinophenyl) triazene] (Fig. 1), also known as Diminazen or Azidin, is a compound used to treat some animal diseases as an antitrypanosomal or antiprotozoal agent. In the past it was classified as a typical minor groove binder,^27,28 but some studies have pointed that it can also intercalate at certain DNA sequences,^29,30 although the intrinsic details about the coupling of these binding modes remain unknown.^29,31 In addition, all the studies concerning the DNA–Berenil interaction were performed using ensemble-averaging techniques such as microcalorimetry, viscometry, and various optical spectroscopies.^28–32 To the best of our knowledge, a single molecule study on the DNA–Berenil interaction is lacking. Such kind of study is important because it gives insights on the mechanical behavior of individual DNA–ligand complexes, from where the relevant physicochemical properties of the interaction can be extracted.³³ In addition, such kind of study can be used to identify and decouple different binding modes of a ligand.^33–35

Fig. 1 Chemical structure of the Berenil molecule.

In order to get new insights on the understanding of DNA interactions with complex ligands that exhibit multiple mixed binding modes, here we have used optical tweezers to perform single molecule stretching experiments with λ-DNA–Berenil complexes at various ligand concentrations, in order to determine the effects of Berenil binding on the mechanical properties of the DNA molecule (contour and persistence lengths). Combining a modeling analysis for these two properties, we were able to: (a) extract the physicochemical parameters of the interaction (binding constants, cooperativity degree, etc.); (b) determine the effective binding mechanisms; and (c) determine the role of the buffer ionic strength on the interaction. Such methodology can be promptly applied to other complex DNA ligands, therefore allowing one to investigate and decouple different binding mechanisms.

2 Materials and methods

The samples in this work consist of λ-DNA molecules (New England Biolabs Cat. N3011) end-labeled with biotin in a Phosphate Buffered Saline (PBS) solution. Berenil was purchased from Sigma-Aldrich (Cat. D7770) and used without further purification. We have carried all the experiments in two different PBS solutions with [Na] = 150 mM and [Na] = 10 mM in order to investigate the role of the ionic strength on the Berenil interaction with λ-DNA. The sample chamber consists of an o-ring glued in the surface of a glass coverslip, where the working solution is deposited. One end of the DNA molecules is attached to the coverslip surface, which is coated with streptavidin,³⁶ while the other end is attached to a streptavidin-coated polystyrene bead with a diameter of 3 μm (Bangs Labs Cat. CP01N).

Our optical tweezers consist of a 1064 nm ytterbium-doped fiber laser (IPG Photonics) operating in the TEM₀₀ mode, mounted on a Nikon Ti–S inverted microscope with a 100× NA 1.4 objective. The tweezers is previously calibrated by using the Stokes force calibration procedure before the experiments. We use a very soft optical tweezers, with a trap stiffness κ = (2.8 ± 0.2) pN μm⁻¹, in order to perform high resolution measurements in the low-force regime (<5 pN). Once calibrated, the tweezers is used to trap the polystyrene bead attached to a DNA molecule. By moving the microscope stage using a piezoelectric actuator, we stretch the DNA while monitoring the changes of the bead position in the tweezers' potential well, using videomicroscopy. In order to guarantee that the DNA–Berenil interaction will not be disturbed by using the tweezers, we limit the maximum stretching forces to ∼2 pN, working therefore within the entropic regime. The Marko–Siggia worm-like chain (WLC)³⁷ expression for the entropic force is then used to fit the experimental data and to extract the basic mechanical parameters: the persistence length A and the contour length L.

Before adding the drug, the bare DNA molecule chosen to perform the experiments is carefully tested by performing at least 6 repeated stretching measurements in order to obtain the average values and the error bars (standard error of the mean) of the mechanical properties. The results obtained are A₀ = (48.3 ± 1.1) nm and L₀ = (16.8 ± 0.4) μm in PBS 150 mM; and A₀ = (45.9 ± 3.0) nm and L₀ = (16.4 ± 0.3) μm in PBS 10 mM. The Berenil concentration in the sample chamber was then changed by using micropipettes, maintaining the same DNA molecule trapped by the tweezers. We wait ∼30 minutes for drug equilibration with DNA. We have found that this time interval is sufficient for the complexes to achieve the chemical equilibrium, verifying that the mechanical properties stabilizes at their equilibrium values after such time. Thus, we determine the average mechanical properties and the error bars of this particular λ-DNA–Berenil complex at the chosen Berenil concentration, by performing 6 repeated stretching experiments. Then, the series of stretching experiments were repeated scanning various Berenil concentrations. Finally, in order to evaluate the variability of the results over different DNA molecules, we use different samples to repeat the entire experiment (scanning all the chosen concentrations) for 4 different DNA molecules. The results and error bars reported below in Fig. 4 and 5 for the mechanical properties are averages over these 4 different DNA molecules. All experiments were carried out at room temperature (23 °C). More details about the experimental methods and procedures can be found in ref. 23.

3 Results and discussion

In Fig. 2 we show some representative force–extension curves obtained for three different Berenil concentrations in the PBS 150 mM buffer. The fittings to the WLC model are also shown (solid lines). Fig. 3 shows the similar result obtained in the PBS 10 mM buffer. Berenil affects both the contour and persistence lengths of the DNA molecule upon binding, which can be observed by analyzing the shape of the force–extension curves.

Fig. 2 Representative force–extension curves of the DNA–Berenil complexes obtained for three different Berenil concentrations in the PBS 150 mM buffer. The fittings to the WLC model are also shown (solid lines).

Fig. 3 Representative force–extension curves of the DNA–Berenil complexes obtained for three different Berenil concentrations in the PBS 10 mM buffer. The fittings to the WLC model are also shown (solid lines).

In Fig. 4 we show the behavior of the average contour length L of the λ-DNA–Berenil complexes (normalized by the bare λ-DNA value L₀) as a function of the Berenil concentration in the sample C_T, obtained for the two ionic strengths used. Observe that the contour length presents a simple monotonic increase, saturating at L/L₀ ∼ 1.14 for C_T > 1 μM in the PBS 150 mM buffer; and at L/L₀ ∼ 1.35, also for C_T > 1 μM, in the PBS 10 mM buffer. Such behavior suggests that intercalative binding possibly plays a role in this system, since a nearly similar behavior of the contour length is usually verified for intercalating ligands.^12,38,39 Nevertheless, such behavior can also be related to possible changes on the DNA solenoidal tertiary structure caused by groove binding of the ligand, as suggested by Reinert.³¹

Fig. 4 Contour length L of the λ-DNA–Berenil complexes (normalized by the bare λ-DNA value L₀) as a function of the Berenil concentration in the sample C_T, obtained both in PBS 150 mM (red circles) and PBS 10 mM (blue squares).

In Fig. 5 we show the behavior of the average persistence length A for the same λ-DNA–Berenil complexes as a function of C_T, obtained in both buffers. Observe that this mechanical parameter exhibits a non-monotonic behavior, increasing from the bare DNA result until reaching a certain maximum value, and then decreasing until ∼41 nm for C_T > 3.5 μM. The main difference here is that the maximum value reached by the persistence length is much higher (∼90 nm) in PBS 10 mM than in PBS 150 mM (∼56 nm). In any case, this non-monotonic behavior strongly suggests that intercalation, if present, certainly is not the unique binding mechanism. In fact, the typical change induced by intercalators on the DNA persistence length is considerably different.^23,40

Fig. 5 Persistence length A of the same λ-DNA–Berenil complexes as a function of C_T, obtained in both buffers. Observe that this mechanical parameter exhibits a non-monotonic behavior, increasing from the bare DNA result until reaching a certain maximum value, and then decreasing until ∼41 nm for C_T > 3.5 μM.

Before we proceed with the quantitative analysis to determine the physicochemical parameters of the λ-DNA–Berenil interaction, some general considerations can be drawn from the data of Fig. 4 and 5. Firstly, observe that Berenil binds strongly to the λ-DNA molecule, since saturation occurs at very low ligand concentrations. Secondly, as mentioned above, two different binding mechanisms, or a mixed binding mechanism, should play a role here in order to explain the non-monotonic behavior of the persistence length. Alternatively, as found by Reinert for calf thymus DNA (ctDNA),³¹ another possibility is the existence of only one single binding mode (groove binding) which causes conformational changes on the double-helix structure (bendings and unwinding) that are related to the behavior of the mechanical parameters. Using viscometry, Reinert has found an increase both in the contour and persistence lengths as Berenil binds for low drug concentrations, in agreement with our results. Higher drug concentrations could not be tested in these experiments due to technical reasons.³¹

If the first scenario is true (two different binding modes), the contour length data suggest that one of these mechanisms should be intercalation. Nevertheless, the non-monotonic behavior of the persistence length strongly suggests that minor groove binding plays an important role here.^15,33,40 The following analysis will show that the pure mechanical data of Fig. 4 and 5 are sufficient to answer if these modes are mixed or independent, i.e., how they are coupled/mixed. There are some compounds such as the chemotherapeutic drug actinomycin D that have two distinct portions, one that intercalates and other that binds to the DNA minor groove.^15,17–22 Thus, in this case a single ligand molecule exhibits the two binding modes that simultaneously occur, i.e., the modes are completely coupled. Other ligands such as the anthracyclines (doxorubicin, daunomycin) intercalate, but have a strong tendency to form self-aggregates composed of a few molecules. Since only one molecule of these aggregates can intercalate, the others remain outside the double-helix interacting with the phosphate backbone.^23,24 This is another example of two binding mechanisms that cannot be dissociated easily, unless it is possible to modify the buffer composition to hinder the ligand self-association. There are, however, some types of ligands that exhibit independent binding modes that are, in principle, dissociable, such as Hoechst 33258,³⁵ β-cyclodextrin⁴¹ and psoralen.¹² In the first two cases, a particular binding mode can be strongly favored by controlling the ligand concentration in the sample.^35,41 In the last case, the binding modes can be controlled by modulating the sample illumination with ultraviolet light.^12,13,42

In the second scenario (a single binding mode), the changes measured both in the persistence and contour lengths arise from the conformational changes such as bendings and twists induced in the double-helix structure due to Berenil binding at the minor groove, without intercalation.

In order to advance in the above discussion, we use here a recently developed quenched-disorder statistical model to extract the physical chemistry of the λ-DNA–Berenil interaction from the persistence length data.^33,43 This model has been used in recent works to investigate DNA interactions with various ligands, and it has been proven extremely useful and efficient to determine the physicochemical parameters of the interaction and to decouple multiple binding modes.^15,33,35,43 Basically, the approach consists in fitting the persistence length data with the equation

where A₀ is the bare DNA persistence length, A₁ is the local persistence length due to the binding of a single ligand molecule (or a single bound cluster of molecules if they form aggregates), A₂ is the local persistence length if two ligand molecules (or two bound clusters) became nearest neighbors, r is the bound site fraction (fraction of bound DNA base-pairs) and r_max is the maximum value of r.^33,43 Observe that the boundary values for the bound ligand fraction are r = 0 (no bound ligand, i.e., bare DNA) and r = r_max (DNA saturated with bound ligand).

The bound site fraction r can be expressed by a convenient binding isotherm involving the relevant physicochemical parameters of the interaction.³³ Here we have tested some different binding isotherms plugged into eqn (1) in order to get insights on the mechanisms of Berenil binding to λ-DNA. The technical details of the fitting procedure were described elsewhere.³³ The first isotherm tested was the Hill binding isotherm, which can be written as

where K is the equilibrium binding association constant, n is the Hill exponent (a cooperativity parameter which is a lower bound for the number of cooperating ligand molecules involved in the reaction), and C_f is the free ligand concentration in solution, which is related to the total ligand concentration C_T by C_f = C_T − C_b, where C_b is the bound ligand concentration. Note that r = C_b/C_bp, where C_bp is the DNA base-pair concentration in solution.

Observe that such isotherm has only one single association constant. Thus, it can be applied to DNA–ligand systems which have only a single binding mode or two or more coupled modes that simultaneously occur, in this last case giving the effective parameters of the interaction.¹⁵

A relatively easy way to use the Hill model to decouple two binding modes is expressing the effective binding isotherm as a sum of two Hill processes,³⁵ i.e.,

Another possibility is using the sequential binding isotherm, which is adequate if the two binding modes do not occur simultaneously. It reads

In Fig. 6 we show the experimental data and the three fittings mentioned above for the PBS 150 mM data. The fitting performed with the sequential binding isotherm (green dashed line) do not agree well with the experimental results, returning some unmeaning physicochemical parameters. As we will show below, this result is related to the fact Berenil binding at high ionic strengths exhibits a considerably high positive cooperativity, which cannot be accounted for by using the sequential binding isotherm. Such positive cooperativity induces a nearly simultaneous binding of a few Berenil molecules in each binding event.³³ Thus, a sequential binding model cannot be the scenario at least for high ionic strengths. On the other hand, observe that the fittings performed with the single Hill (blue solid line) and double Hill (red dashed line) isotherms are indistinguishable, both explaining with accuracy the experimental data. In fact, the physicochemical parameters returned by these two fittings are identical; and when using the double Hill isotherm we found n₁ ∼ n₂ and K₁ ∼ K₂. This result strongly indicates that there is only a single effective binding mechanism for the λ-DNA–Berenil interaction at high ionic strengths. Thus, if intercalation occurs here, it should be strongly coupled/mixed to the groove binding mode. Berenil intercalation was previously reported by Pilch et al. for other types of DNA, although they have found that the existence of such mode strongly depends on the specific DNA sequence.²⁹

Fig. 6 Persistence length data (black circles) obtained in PBS 150 mM and the fittings to our quenched disorder statistical model⁴³ using three different binding isotherms: sequential binding (green dashed line), single Hill (blue solid line) and double Hill (red dashed line). The sequential binding isotherm cannot explain the experimental data, while the fittings to a single or a double Hill process are indistinguishable, indicating that there is only a single effective binding mechanism.

From the fitting with the single Hill process, we find the physicochemical parameters K = (7.9 ± 0.7) × 10⁵ M⁻¹, n = 7 ± 2.5, r_max = 0.4 ± 0.1, A₁ = (75 ± 7) nm and A₂ = (41 ± 1) nm. The result found for the equilibrium association constant K is compatible to results reported in the literature for various different DNA sequences^28–30 and confirms the strong binding of Berenil to the λ-DNA. In addition, the result found for the Hill exponent suggests that Berenil binding occurs here with a high positive cooperativity, with ∼7 drug molecules binding simultaneously to the double-helix at each binding event.³³ Cooperativity was also reported in previous works which have investigated the binding of Berenil to other DNA sequences.²⁸ The value of r_max indicates that approximately ∼40% of the DNA base-pairs are occupied by Berenil at saturation, a result compatible with other investigations.^28,29 Finally, the results found for the local persistence lengths A₁ and A₂ reflect the fact that Berenil binding firstly increases the bending rigidity of the double-helix at low concentrations and then decreases it at higher concentrations. Such behavior occurs for many ligands which exhibit complex binding to DNA, and can arise from the competition of two distinct binding modes^15,35 as well as from conformational changes on the double-helix induced by the ligand at a certain concentration.^41,43

In Fig. 7 we show the similar fittings obtained for the PBS 10 mM data. The double Hill isotherm is not shown here because it is again identical to the single Hill isotherm, with n₁ ∼ n₂ and K₁ ∼ K₂. Observe that, for low ionic strengths, both the single Hill (blue solid line) and sequential binding (green dashed line) isotherms explain well the experimental data. Nevertheless, when using the sequential binding isotherm, the fitting returns K₁ ∼ K₂, indicating again that there is only a single effective binding mode here. In the present case, such binding mode should be non-cooperative since it can be explained with a sequential binding isotherm, which assumes that individual molecules bind nearly sequentially to the DNA.

Fig. 7 Persistence length data (black circles) obtained in PBS 10 mM and the fittings to our quenched disorder statistical model⁴³ using two different binding isotherms: sequential binding (green dashed line) and single Hill (blue solid line). The fitting with the sequential binding isotherm returns K₁ ∼ K₂, indicating again that there is only a single binding mode here. The double Hill isotherm is not shown here because it is again identical to the single Hill isotherm.

From the fitting with the sequential binding isotherm, we found K₁ = K₂ = (1.3 ± 0.5) × 10⁶ M⁻¹. From the fitting with the single Hill process, we found K = (1.7 ± 0.4) × 10⁶ M⁻¹, n = 1.6 ± 0.6, and also r_max = 0.4 ± 0.1. These results indicate that the binding affinity increases for lower ionic strengths (the equilibrium constant is approximately twice the value found in the PBS 150 mM buffer), but the cooperativity between Berenil molecules strongly decreases, becoming very close to a complete non-cooperative system (Hill exponent ∼ 1 (ref. 15 and 33)). In non-cooperative binding reactions the ligand molecules bind individually, not influencing the binding of the subsequent ones. This result is strongly correlated to the net charge of the Berenil molecules (+2). At high ionic strengths, the repulsive electrostatic interaction between these molecules is strongly screened, making cooperativity possible. Nevertheless, at low ionic strengths, such repulsive interaction can difficult the subsequent binding of Berenil molecules close to previously bound sites of the double-helix. Thus, we expect a strong decrease of the Hill exponent as the ionic strength of the buffer is lowered, as measured here.

In summary, from the results of the fitting analysis performed in both buffers, it is clear that there is only a global effective binding mode in the interaction between Berenil and λ-DNA, and not two independent modes. What is not clear yet is if this effective binding mode involves partial intercalation of Berenil molecules into the double-helix. In fact, such information could not be deduced from the analysis performed here with the persistence length data. As mentioned before, we have two possibilities: a mixed binding mode involving partial intercalation and groove binding, or alternatively, a groove binding mechanism which induces conformational changes on the double-helix structure, probably unwinding, that increases the DNA contour length. The first scenario was proposed by Pilch et al., who have studied Berenil interaction with some particular DNA sequences at low ionic strengths.²⁹ The second scenario was proposed by Reinert, who has studied Berenil interaction with ctDNA at a high ionic strength.³¹ In order to clarify which is the case here for the Berenil interaction with λ-DNA, we return to the contour length data shown in Fig. 4. If one supposes that the increase verified for this parameter is due only to intercalative binding, it may satisfy the relation^33,38,44

with γ = δ/Δ, where Δ = 0.34 nm is the mean distance between two consecutive base pairs in the bare DNA molecule at B-form and δ is the increase of this length caused by a single intercalated molecule.⁴⁴

For typical intercalators one has δ ∼ 0.34 nm and γ ∼ 1.^4,6,44 Thus, determining the values of δ and γ is a simple way to quantify the role of intercalation in a DNA–ligand binding reaction. From Fig. 4, we have Θ_max = 0.14 for the PBS 150 mM data and Θ_max = 0.35 for the PBS 10 mM data. Since we have determined r_max ∼ 0.4 for both buffers data from the persistence lengths fittings, it is straightforward to obtain γ_{PBS 150} ∼ 0.14/0.4 ∼ 0.35 and γ_{PBS 10} ∼ 0.35/0.4 ∼ 0.88. Thus, δ_{PBS 150} ∼ 0.35 × (0.34 nm) ∼ 0.12 nm and δ_{PBS 10} ∼ 0.88 × (0.34 nm) ∼ 0.3 nm. One should observe that in PBS 10 mM, the value found for γ (and consequently for δ) is very close to the result expected for intercalators. Nevertheless, in PBS 150 mM, the values of γ and δ are much smaller, but compatible to the result found by Reinert for the interaction of Berenil with ctDNA at high ionic strengths.³¹

Thus, based on the analysis performed with our experimental data, we propose the following scenario for the interaction between Berenil and λ-DNA: the ionic strength of the buffer solution is the factor that determines if the partial intercalation of Berenil occurs. By increasing the ionic strength of the buffer, one decreases the affinity of Berenil to partially intercalate into the double-helix of λ-DNA. In any case, such intercalation, when exists, is always mixed/coupled with minor groove binding, which is the main binding mechanism of the λ-DNA–Berenil interaction. Both intercalation and groove binding contribute here to increase the DNA contour length, but such increase is much stronger when partial intercalation is present at low ionic strengths, due to the intrinsic nature of this type of interaction. The fact that the existence of the intercalative binding mode depends on the ionic strength is strongly correlated to the net charge of the Berenil compound (+2). Indeed, our results show that electrostatic driven interactions play a very important role in this DNA–ligand system. This is the reason why the variation of the ionic strength affects significantly the maximum values obtained for the mechanical properties (contour and persistence lengths), the physicochemical parameters (equilibrium constant and Hill exponent), and the whole binding mechanism.

4 Conclusion

By using single molecule stretching and a new methodology of data analysis, we were able to show that one can determine the binding mechanisms and the physicochemical parameters of the DNA interaction with the complex ligand Berenil. In particular, we show that the interaction can be modulated by changing the ionic strength of the surrounding buffer. For high ionic strengths (150 mM), Berenil behaves as a typical minor groove DNA ligand. For low ionic strengths (10 mM), otherwise, the drug also partially intercalates into the double-helix. Besides such conclusions, our analysis promptly allows one to extract the physicochemical parameters (equilibrium constants, cooperativity degree, bound ligand fraction at saturation, etc.) of the interaction for each ionic strength studied. Our methodology can be readily applied to other DNA ligands, allowing one to investigate and decouple different binding mechanisms for complex ligands which present multiple mixed binding modes to DNA.

Acknowledgements

This work was supported by the Brazilian agencies: Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

References

1. L. D. Williams, M. Egli, Q. Gao and A. Richa, Struct. Funct., Proc. Conversation Discip. Biomol. Stereodyn., 7th, 1992, 1.
2. M. F. Brana, M. Cacho, A. Gradillas, B. de Pascual-Teresa and A. Ramos, Curr. Pharm. Des., 2001, 7, 1745–1780.
3. M. K. Goftar, N. M. Kor and Z. M. Kor, Int. J. Adv. Biol. Biomed. Res., 2014, 2, 811–822.
4. A. Sischka, K. Tönsing, R. Eckel, S. D. Wilking, N. Sewald, R. Rios and D. Anselmetti, Biophys. J., 2005, 88, 404–411.
5. J. B. Chaires, N. Dattagupta and D. M. Crothers, Biochemistry, 1982, 21, 3933–3940.
6. H. Fritzsche, H. Triebel, J. B. Chaires, N. Dattagupta and D. M. Crothers, Biochemistry, 1982, 21, 3940–3946.
7. B. H. Geierstanger and D. E. Wemmer, Annu. Rev. Biophys. Biomol. Struct., 1995, 24, 463–493.
8. F. A. P. Crisafuli, E. C. Cesconetto, E. B. Ramos and M. S. Rocha, Integr. Biol., 2012, 2012, 568–574.
9. K. Stehlikova and H. Kostrhunova, Nucleic Acids Res., 2002, 30, 2894–2898.
10. X.-M. Hou, X.-H. Zhang, K.-J. Wei, C. Ji, S.-X. Dou, W.-C. Wang, M. Li and P.-Y. Wang, Nucleic Acids Res., 2009, 37, 14001410.
11. N.-K. Lee, J.-S. Park, A. Johner, S. Obukhov, J.-Y. Hyon, K. J. Lee and S.-C. Hong, Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 2009, 79, 041921.
12. M. S. Rocha, A. D. Lúcio, S. S. Alexandre, R. W. Nunes and O. N. Mesquita, Appl. Phys. Lett., 2009, 95, 253703.
13. D. W. Ussery, R. W. Hoepfner and R. R. Sinden, Methods Enzymol., 1992, 212, 242–262.
14. G. D. Cimino, H. B. Gamper, S. T. Isaacs and J. E. Hearst, Annu. Rev. Biochem., 1985, 54, 1151–1193.
15. E. C. Cesconetto, F. S. A. Junior, F. A. P. Crisafuli, O. N. Mesquita, E. B. Ramos and M. S. Rocha, Phys. Chem. Chem. Phys., 2013, 15, 11070–11077.
16. T. Paramanathan, I. Vladescu, M. J. McCauley, I. Rouzina and M. C. Williams, Nucleic Acids Res., 2012, 40, 49254932.
17. D. J. Patel, S. A. Kozlowski, J. A. Rice, C. Broka and K. Itakura, Proc. Natl. Acad. Sci. U. S. A., 1981, 78, 7281–7284.
18. W. Müller and D. M. Crothers, J. Mol. Biol., 1968, 35, 251–290.
19. H. M. Sobell, S. C. Jain, T. D. Sakore and C. E. Nordman, Nature, New Biol., 1971, 231, 200–205.
20. R. Eckel, R. Ros, A. Ros, S. D. Wilking, N. Sewald and D. Anselmetti, Biophys. J., 2003, 85, 1968–1973.
21. T. Berge, N. S. Jenkins, R. B. Hopkirk, M. J. Waring, J. M. Edwardson and R. M. Henderson, Nucleic Acids Res., 2002, 30, 2980–2986.
22. P. L. Hamilton and D. P. Arya, Nat. Prod. Rep., 2012, 29, 134–143.
23. E. F. Silva, R. F. Bazoni, E. B. Ramos and M. S. Rocha, Biopolymers, 2016 DOI:10.1002/bip.22998.
24. C. Pérez-Arnaiz, N. Busto, J. M. Leal and B. García, J. Phys. Chem. B, 2014, 118, 1288–1295.
25. A. Y. Chen and L. F. Liu, Annu. Rev. Pharmacol. Toxicol., 1994, 34, 191–218.
26. D. S. Pilch, C. Yu, D. Makhey, E. J. LaVoie, A. R. Srinivasan, W. K. Olson, R. R. Sauers, K. J. Breslauer, N. E. Geacintov and L. F. Liu, Biochemistry, 1997, 36, 12542–12553.
27. L. H. Pearl, J. V. Skelfy, B. D. Hudson and S. Neidle, Nucleic Acids Res., 1987, 15, 3469–3478.
28. H.-U. Schmitz and W. Hubner, Biophys. Chem., 1993, 48, 61–74.
29. D. S. Pilch, M. A. Kirolos, X. Liu, G. E. Plum and K. J. Breslauer, Biochemistry, 1995, 34, 9962–9976.
30. F. Barceló, M. Ortiz-Lombardía and J. Portugal, Biochim. Biophys. Acta, 2001, 1519, 175–184.
31. K. E. Reinert, Nucleic Acids Res., 1983, 11, 3411–3430.
32. C. Caramelo-Nunes, T. Tente, P. Almeida, J. C. Marcos and C. T. Tomaz, Anal. Biochem., 2011, 412, 153–158.
33. M. S. Rocha, Integr. Biol., 2015, 7, 967–986.
34. M. J. McCauley and M. C. Williams, Biopolymers, 2009, 91, 265–282.
35. E. F. Silva, E. B. Ramos and M. S. Rocha, J. Phys. Chem. B, 2013, 117, 7292–7296.
36. I. Amitani, B. Liu, C. C. Dombrowski, R. J. Baskin and S. C. Kowalczykowski, Methods Enzymol., 2010, 472, 261–291.
37. J. F. Marko and E. D. Siggia, Macromolecules, 1995, 28, 8759–8770.
38. M. S. Rocha, M. C. Ferreira and O. N. Mesquita, J. Chem. Phys., 2007, 127, 105108.
39. J. E. Coury, L. McFail-Isom, L. D. Williams and L. A. Bottomley, Proc. Natl. Acad. Sci. U. S. A., 1996, 93, 12283–12286.
40. R. F. Bazoni, C. H. M. Lima, E. B. Ramos and M. S. Rocha, Soft Matter, 2015, 11, 4306–4314.
41. P. S. Alves, O. N. Mesquita and M. S. Rocha, J. Phys. Chem. Lett., 2015, 6, 3549–3554.
42. M. S. Rocha, N. B. Viana and O. N. Mesquita, J. Chem. Phys., 2004, 121, 9679–9683.
43. L. Siman, I. S. S. Carrasco, J. K. L. da Silva, M. C. Oliveira, M. S. Rocha and O. N. Mesquita, Phys. Rev. Lett., 2012, 109, 248103.
44. M. S. Rocha, Phys. Biol., 2009, 6, 036013.

---