Understanding Ligand Interaction with Different Structures of G

Jul 20, 2012 - Sachin Dev Verma, Nibedita Pal, Moirangthem Kiran Singh, Him Shweta, Mohammad Firoz Khan, and Sobhan Sen*. Spectroscopy Laboratory ...
0 downloads 0 Views 478KB Size
Download PDF
Article pubs.acs.org/ac

Understanding Ligand Interaction with Different Structures of G‑Quadruplex DNA: Evidence of Kinetically Controlled Ligand Binding and Binding-Mode Assisted Quadruplex Structure Alteration Sachin Dev Verma, Nibedita Pal, Moirangthem Kiran Singh, Him Shweta, Mohammad Firoz Khan, and Sobhan Sen* Spectroscopy Laboratory, School of Physical Sciences, Jawaharlal Nehru University (JNU), New Delhi 110067, India S Supporting Information *

ABSTRACT: The study of ligand interaction with G-quadruplex DNA is an active research area, because many ligands are shown to bind G-quadruplex structures, showing anticancer effects. Here, we show, for the first time, how fluorescence correlation spectroscopy (FCS) can be used to study binding kinetics of ligands with G-quadruplex DNA at the single molecule level. As an example, we study interaction of a benzo-phenoxazine ligand (Cresyl Violet, CV) with antiparallel and (3 + 1) hybrid G-quadruplex structures formed by human telomeric sequence. By using simple modifications in FCS setup, we describe how one can extract the reaction kinetics from diffusion-coupled correlation curves. It is found that the ligand (CV) binds stronger, by an order of magnitude, to a (3 + 1) hybrid structure, compared to an antiparallel one. Ensemble-averaged time-resolved fluorescence experiments are also carried out to obtain the binding equilibrium constants (K) of ligand-quadruplex interactions in bulk solution for the first time, which are found to match very well with FCS results. Global analysis of FCS data provides association (k+) and dissociation (k−) rates of the ligand in the two structures. Results indicate that stronger ligand binding to the (3 + 1) hybrid structure is controlled by the dissociation rate, rather than the association rate of ligand in the quadruplexes. Circular dichroism (CD) and induced-CD spectra show that the ligand not only binds at different conformations in the quadruplexes, but also induces antiparallel structure to form a mixed-type hybrid structure in Na+ solution. However, in K+ solution, the ligand stabilizes the (3 + 1) hybrid structure. Molecular docking studies predict the possible differences in binding sites of the ligand inside two quadruplexes, which strongly support the experimental observations. Results suggest that different binding modes of the ligand to the quadruplex structures actually assist the alteration of structures differently.

H

critical length of telomere. At this stage the cell enters a state of senescence and subsequently the cell dies.11−13 However, in cancer cells telomere gets extended during replication due to the activation of telomerase that leads to abnormal growth of cells.13 G-quadruplex structures are found to have inhibitory effects on the catalytic activity of telomerase.14 The existence of G-quadruplex in vivo15 has provided extra importance to these structures, because they can act as promising anticancer targets for ligands to block telomerase activity.3,4,13−16 To date, five different structures have been reported for human telomeric sequence. In presence of Na+ ions, a basket-

uman telomeres protect chromosome from nuclease degradation, recombination, and end-to-end fusion.1−4 In human cells, telomeric DNA contains a ∼2−10 kb duplex region of repetitive guanine-rich (TTAGGG)n sequence, terminated by 3′-end single-strand overhang of ∼200 nucleobases.1−4 Under physiological ionic conditions, the 3′end overhang can form four-stranded guanine quartet structures, called G-quadruplex (Figure 1).5−10 These structures are stabilized by monovalent cations such as Na+ and K+ and/or by small molecules, forming multiple G·G·G·G stacks through Hoogsteen-type hydrogen bonds among the guanines.5−10 The structure and stability of human telomeres are related to genetic stability,11 cell aging,12 and cancer.13 During cell division, the length of telomere gets shortened which leads to a © 2012 American Chemical Society

Received: June 11, 2012 Accepted: July 20, 2012 Published: July 20, 2012 7218

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

structure. Molecular docking studies on the ligand-quadruplex complexes predict the possible differences in binding sites of CV inside the structures, which strongly support experimental observations. Our results suggest that, based on binding modes, the ligand can alter the G-quadruplex structures differently, in the presence of different ions. Although the extensive use of CVs as neural stains in histological studies has been reported,24 this paper is probably the first one showing that CV can possibly serve as a potential ligand to target G-quadruplex forming DNA. In fact, a very recent report by Balasubramanian and co-workers showed that, out of 173 molecules, only 2 molecules having a benzophenoxazine core-group, like CV has, can actually downregulate the c-KIT expression in human gastric carcinoma cells.20 Several other acridine, anthraquinone, quindoline, porphyrin, carbocyanine, berberine, perylene, isoquinolinebased molecules are found to bind and stabilize G-quadruplex structures.3,4,13,17−23 Many of them have shown promising anticancer effect.3,17−20,23 Several techniques, such as UV−vis absorption and fluorescence,25−27 circular dichroism (CD),28−30 X-ray,31 nuclear magnetic resonance (NMR),32 isothermal titration calorimetry (ITC),33 surface plasmon resonance (SPR),20,26 etc., have been used to study ligand−quadruplex interactions. However, FCS stands out to be unique, because it has the capability of recording molecular interactions at the single molecular level.34,35 FCS is used extensively to study molecular diffusion and size, chemical kinetics, and molecular interactions in vitro and in vivo.34−46 Previously, FCS has been used to study the competition of duplex−quadruplex interconversion,43 and pH-dependent structure variation of C-rich i-motifs.44 However, to the best of our knowledge, FCS has never been used in the context of ligand−quadruplex interaction. This paper will show how FCS can be used efficiently to study ligand−quadruplex interactions at the single-molecule level to obtain the reaction kinetic parameters, using a laser beam that underfills the back-aperture of objective and a large signal-collection pinhole. We also show how ensemble-averaged time-resolved fluorescence decay and anisotropy measurements can be performed to compliment the FCS results. CD and molecular modeling studies are also carried out to obtain comprehensive picture of the interactions. Results provide important clues about the ligand binding characteristic to Gquadruplex structures, that stronger binding of a ligand to a preformed structure is favored possibly when the ligand further induces similar topology within that structure.

Figure 1. Schematic structures of human telomeric G-quadruplex DNA: (A) basket-type antiparallel structure in Na+, (B) (3 + 1) hybrid structure in K+ solution. (C) Molecular structure of Cresyl Violet.

type antiparallel structure (Figure 1A) is found for sequence d[AGGG(TTAGGG)3] in solution.5 However, in the presence of K+ (ions more abundant in intracellular environment), at least three structures(3 + 1) Form-1 hybrid (Figure 1B), (3 + 1) Form-2 hybrid and Form-3have been found in solution.6−9 However, the K+-containing crystal shows only a propeller-type all-parallel G-quadruplex structure.10 The concept that stabilization of G-quadruplex structures can be achieved through targeting the telomeric 3′-end overhang with small molecules is one propitious route for anticancer drug development.3,4,13,17,18 G-quadruplexes are three-dimensional (3D) structures that have diverse loop and G-tract orientations.5−10 Targeting and stabilizing these G-quadruplexes with small molecules indeed require structure-based (drug) molecule design. Many quadruplex-stabilizing ligands have been reported.3,4,17−23 Some of them are publicized as hopeful anticancer agents.18−21 Nonetheless, the effective drug design and their efficiency to bind/stabilize G-quadruplex structures rely on some central questions: • How does the ligand interact with different G-quadruplex structures? • Are these interactions kinetically controlled? • How does the ligand induce quadruplex structures in the presence of ions? • Which ligand structures have potential for use as anticancer drugs? However, the most important question is this: How does one study these interactions, especially at their single molecule level? This paper tackles these questions by studying a ligand’s (Cresyl Violet, CV; Figure 1C) interaction with two different human telomeric G-quadruplex structuresbasket-type antiparallel and (3 + 1) hybrid structuresat the (near) singlemolecule level using fluorescence correlation spectroscopy (FCS) for the first time. FCS results reveal at least an order-ofmagnitude difference in the binding affinities of the ligand in two structures. Ensemble-averaged time-resolved fluorescence measurements are also performed, and the results support the FCS data. Analysis of the FCS data with the kinetic model finds stronger binding of CV to (3 + 1) hybrid structure is controlled by the dissociation rate (k−) of CV, rather than its association rate (k+). Circular dichroism (CD) and induced-CD spectra also show that CV not only binds in different conformation in the quadruplexes, but also induces the antiparallel structure toward mixed-type hybrid topology in Na+ solution. However, in K+ solution CV further stabilizes the (3 + 1) hybrid

■

RESULTS AND DISCUSSION At first, the different structures in presence of the two cations are confirmed in CD. Spectra confirm basket-type antiparallel structure in Na+ and (3 + 1) hybrid structure in K+ solutions (see Figure 5A, presented later in this work, and the text below). (See the Supporting Information for details regarding the materials and methods.) Steady-State Fluorescence Data. Figure 2 shows the fluorescence emission and excitation spectra of free CV and when bound to G-quadruplex structures. The excitation spectrum (Figure 2, inset) of free CV shows a peak at ∼590 nm. However, when CV binds to the quadruplex formed in either Na+ or K+, the excitation spectra shift to the red side by ∼25 nm, resulting in peaks located at ∼615 nm. This red shift indicates that the complexes are formed in the ground-state of CV. The emission yield of CV decreases drastically and the 7219

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

G (τ ) =

⟨δF(t )δF(t + τ )⟩ ⟨F(t )⟩2

(1)

where δF(t) is fluorescence fluctuation at time t, δF(t + τ) is that after a delay τ, and ⟨F(t)⟩ is the average fluorescence intensity. If fluctuations arise from only translation diffusion of dye into and out of a 3D Gaussian volume with radial (r) and axial length (l), then GD(τ) can be modeled with the function34−46 −1 ⎛ r ⎞ 2 ⎛ τ ⎞⎤ 1 ⎛ τ⎞ ⎡ ⎜1 + ⎟ ⎢ 1 + ⎜ ⎟ ⎜ ⎟ ⎥ ⎝ l ⎠ ⎝ τD ⎠⎥⎦ ⟨N ⟩ ⎝ τD ⎠ ⎢⎣

−1/2

G D(τ ) =

Figure 2. Relative fluorescence emission spectra of CV in buffer (black line), and bound to antiparallel (red line) and (3 + 1) hybrid Gquadruplex (blue line). Inset shows normalized excitation spectra (CV in buffer (black dots), antiparallel (red dots), and (3 + 1) hybrid Gquadruplex (blue dots)). The fluorescence intensity decreases ∼12fold in antiparallel and ∼33-fold in (3 + 1) hybrid G-quadruplex, relative to CV in buffer. Arrow in the inset indicates the direction of shift in the excitation spectra of CV upon binding to the Gquadruplexes.

(2)

where ⟨N⟩ is the average number of fluorescent species inside the volume and τD is the translational diffusion time. The parameter τD can be expressed in terms of diffusion coefficient (D) as τD =

r2 4D

(3)

FCS is ideal for measuring reaction kinetics if association and dissociation of a fluorescent ligand to a host change the fluorescent state (high-to-low or low-to-high) of the ligand, and if the time of reaction falls within the time window of correlation.41,43,44 This happens in the present system. Free CV has high fluorescence. But, upon binding, fluorescence gets quenched (see Figure 2). This leads to fluorescence fluctuations during CV’s dwell time inside the observation volume, which get correlated to provide as coupled reaction term (τR) in the diffusion (τD) of bound CV. This interaction can be treated as a single-step bimolecular reaction, where one CV is interacting with one quadruplex, having association and dissociation rates of k+ and k−, respectively. The reaction can then be written as

spectra shift to the red side upon binding to either of the Gquadruplex structures. The decrease in fluorescence intensity arises mainly because of two reasons: (i) shifting of the excitation spectra to the red side, because of ground-state complex formation, which reduces the excitation probability of bound CV at the excitation wavelength; and (ii) electron transfer from guanines to CV, which promotes nonradiative pathways in CV. The relative change in fluorescence intensities of CV among two structures indicates different efficiencies of electron transfer. Guanine is highly rich electron donor, and CV can act as an effective electron acceptor.47 The efficiency of electron transfer from a donor to an acceptor strongly depends on the relative distance and orientation between donor and acceptor. The relative change in fluorescence intensity of bound CV implies that it binds at different positions (and possibly at different orientations) inside two structures, such that the efficiencies of electron-transfer change in the two structures. Two types of binding modes are possible for a molecule like CV inside G-quadruplex: stacking with G-tetrad and groove binding. The larger quenching in the (3 + 1) hybrid quadruplex suggests a possible π-type stacking of CV with G-tetrad, which can promote efficient electron transfer. However, relatively less quenching in antiparallel structure indicates lesser efficiency of electron transfer, which suggests a possible groove binding of the ligand inside this structure. Molecular docking studies actually support these ideas (see below). Note that, because the reaction time (τR) of CV with quadruplex is found to be much slower (on the scale of tens of microseconds) than the excitedstate lifetime of CV (on the scale of nanoseconds) and tripletstate conversion (microsecond), the complexation equilibrium can be considered only in the ground-state of CV. Thus, interferences from excited-state processes of CV in the measured reaction kinetics can be neglected. FCS Data. FCS correlates fluorescence fluctuations that originate from changes in concentration of fluorescent species due to translational motion into and out of a tiny observation volume (on the order of femtoliters), and/or from the changes in chemical properties of the fluorophore due to reaction and/ or complex formation. In FCS, the autocorrelation function G(τ) is given by34−46

k+

CV + quadruplex ⇄ CV − quadruplex k−

(4)

The equilibrium binding constant (K) can be defined as K=

k+ k−

(5)

The reaction is considered to be pseudo-first-order in nature with relaxation time (τR) as45,46 τR = (k+[quadruplex] + k −)−1

(6)

To model the correlation curves, obtained from titration of CV with G-quadruplexes, we adopted the general solution of GR(τ) as in eq 7, with the condition that τR is much faster than diffusion times of free (τA) or bound (τB) CV. G R (τ ) =

−1/2 −1 ⎛ ⎛ r ⎞ 2 ⎛ τ ⎞⎤ τ ⎞ ⎡ 1 ⎜1 + ⎟ ⎢1 + ⎜ ⎟ ⎜ ⎟⎥ ⎝ l ⎠ ⎝ τD̅ ⎠⎥⎦ NA + NB ⎝ τD̅ ⎠ ⎢⎣

×(1 + AR e−τ / τR )

(7)

where NA and NB are, respectively, the number of free and bound dyes inside observation volume. AR is the amplitude of the reaction term. The concentration-dependent variables τD̅ and τR can be expressed as45,46 7220

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry τD̅ ([quadruplex]) = with

Article

methods). Under this condition, the average number (NA + NB) of CV molecules was calculated to be ∼5−10. Global Analysis of FCS Data. Any attempt to extract reliable values of chemical kinetic parameters demands a simple reaction model that can describe the experimental data. The kinetic parameters are extracted by analyzing the correlation curves with a target global fit using eq 7. This equation incorporates τD̅ and τR expressions from eqs 8 and 9, where parameters are strongly correlated. Hence, the fit to individual correlation curve was highly influenced by the statistical parameter correlations. Because the association and dissociation of CV with the quadruplex take place within the dwell time of CV inside the observation volume; only the average diffusion of the molecules is observed with a mean diffusion time, τD̅ . This τD̅ contains diffusion times of free CV (τA) and bound CV (τB) through eq 8. τA (455 μs) was easily determined from the correlation curve of CV in buffer. However, only the approximate τB values were obtained from the fit. The concentration-dependent τD̅ and τR parameters are calculated using parameter values of τA (fixed), τB, K, and k− (see the Supporting Information). During the fit, the parameters were shared among all correlation curves. The geometry factor (l/r) was known from calibration of the FCS setup (see the Supporting Information) and was kept fixed during analysis. AR was kept as a free parameter to follow its variation with the Gquadruplex concentration. As expected, AR shows an initial rise and then saturation with increasing quadruplex concentration (see the Supporting Information). Global fit was started with initial (fixed) guess values of the parameters, and then released one-by-one in consecutive runs. This procedure was used several times to ensure significant improvement in residuals and parameter values. Keeping the K-value fully free raised uncertainties in τB and k− values. Fixing a suitable range for K reduced this uncertainty. A constrained range of 1 × 105−2 × 106 was used for K in the fits. This range was fixed because Kvalues obtained from time-resolved fluorescence data fall in this range (see below). The values of fitted parameters are listed in Table 1.

τA(1 + K[quadruplex]) 1 + (τA /τB)K[quadruplex]

DB τ = A τB DA

(8)

and τR ([quadruplex]) = [k −(1 + K[quadruplex])]−1

(9)

A set of concentration-dependent correlation curves for a particular G-quadruplex is analyzed with a target global fit using eq 7, incorporating τD̅ and τR expressions from eqs 8 and 9. Concentration-dependent AR values are obtained as fitting parameters. Binding equilibrium (K) and dissociation rate (k−) constants are obtained directly from the fit. Correlation of free CV in buffer, fitted with eq 2, yields τD = 455 ± 2 μs. Upon the addition of preformed G-quadruplex, the autocorrelation curves get monotonically faster near the characteristic reaction time (τR) (see Figure 3). This is a

Table 1. Reaction Kinetics Parameters from FCS Global Fit parameter

antiparallelb

−1

(0.13 ± 0.005) × 10 (17 ± 0.26) × 103 (2.21 ± 0.1) × 109 455 (fixed) 630 (±15)

K (M ) k− (s−1) k+ (M−1 s−1)a τA (μs) τB (μs)

Figure 3. Normalized fluorescence correlation curves of free CV (1 nM) and at different concentrations (0.3−10 000 nM) of Gquadruplexes, along with their global fits. Residuals are plotted in lower panels. FCS data for CV: (A) antiparallel G-quadruplex and (B) (3 + 1) hybrid G-quadruplex. The change in correlation curves near the reaction time is indicated by the arrows. Inset in each panel shows the reaction and diffusion regions when they are matched at 100 μs. For the sake of clarity, only 9 curves out of 23 curves are plotted.

(3 + 1) hybridb 6

(1.7 ± 0.02) × 106 (2.4 ± 0.03) × 103 (4.08 ± 0.1) × 109 455 (fixed) 650 (±10)

Values calculated using eq 5. bError is calculated from weighted fits using the standard deviation of data points.

a

The binding equilibrium constants (K) obtained for CV in two G-quadruplex structures differ by more than an order of magnitude: K = 1.7 × 106 M−1 and 0.13 × 106 M−1 in (3 + 1) hybrid and antiparallel structures, respectively. This indicates that CV has a much higher binding preference to (3 + 1) hybrid structure over the antiparallel one. Molecules such as CV, having a planar benzo-phenoxazine core-group, possess delocalized π-electrons, which can promote stable π-stacking with the G-tetrad. Molecular docking study does confirm such end-stacking/intercalation of CV inside this structure (see below). On the other hand, diagonal and edgewise loops present near the ends of antiparallel structure may actually introduce some steric hindrance to form π-stacked CV

typical feature of reaction-coupled diffusive correlation curves.37,45,46 Figure 3A shows autocorrelation curves of CV titrated with antiparallel quadruplex in Na+, and Figure 3B shows that for (3 + 1) hybrid structure in K+ solution. It is seen that τR typically ranges from ∼30 μs to 100 μs (see inset in Figure 3 and the Supporting Information), which is an order of magnitude faster than the diffusion time of either free CV (455 μs) or bound CV (∼630−650 μs). This condition is only achieved by using a laser beam (diameter of ∼1 mm) that under-fills the objective back-aperture and large collection pinhole (100 μm) (see the Supporting Information for the 7221

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

Figure 4. Fluorescence decays (with IRF and fits) of CV in the presence of different concentrations of (A) antiparallel and (B) (3 + 1) hybrid Gquadruplex. (C) Concentration-dependent fraction (ab) of the lifetime of bound CV in the presence of antiparallel (red) and (3 + 1) hybrid Gquadruplex (blue). Anisotropy decays (with global fits) of CV in the presence of different concentrations of (D) antiparallel and (E) (3 + 1) hybrid G-quadruplex. (F) Concentration-dependent fraction of anisotropy decay time of bound CV in antiparallel (red) and (3 + 1) hybrid (blue) quadruplex. Only 10 decays and 8 anisotropy data points (out of 17 data points) are plotted for the sake of clarity. Arrows indicate the direction of change with concentration.

smaller than pure diffusion-controlled association rate of two species calculated by the Smoluchowski equation (kdiff ≈ 1010 M−1 s−1).46 Time-Resolved Fluorescence Data. To confirm the values of binding equilibrium constants obtained from FCS data, we measured ensemble-averaged time-resolved fluorescence decays of CV in buffer and in the presence of different concentrations of G-quadruplex in Na+ and K+ solutions. It should be noted that, although time-resolved fluorescence has been used earlier to monitor ligand−quadruplex interactions,25,26 this method has never been used to calculate the binding constants of ligand−quadruplex interactions. Figures 4A and 4B plot the fluorescence decays of CV in the presence of different concentrations of antiparallel and (3 + 1) hybrid quadruplex structures, respectively. It is readily seen that the rate-of-change and spread-of-change of the decays are different in two systems, suggesting a substantial difference in CV interaction with the two structures. The decay of free CV could be fitted with a single exponential function (τf = 2.6 ns). However, with increasing quadruplex concentration, the decays develop multiexponential features with faster average lifetimes. This decrease in lifetime confirms the effect of electron transfer from guanines to CV. A set of (normalized) titrated decays were analyzed (deconvoluting instrument response function) in a global target fit. Two exponential time constants were needed to fit the decays (see eq 10). Two time componentsone for free CV (τf) and another for bound CV (τb)were linked in all decays, along with their relative contributions (af and ab) in the global fit.

complex. This may otherwise facilitate groove-binding of CV, as is seen in molecular docking studies. This gives rise to lower binding affinity of CV to this structure. SPR results of Balasubramanian and co-workers show similar binding phenomena for other substituted benzo-phenoxazine molecules with c-KIT sequences.20 They saw two benzo-phenoxazine molecules bind to c-kit2 by an order of magnitude stronger than c-kit1, albeit bind less strongly to human telomeric sequence, compared to c-kit2.20 Here, we show that CV can bind very strongly (K = 1.7 × 106 M−1) to the (3 + 1) hybrid G-quadruplex, which is comparable to the binding affinity (Kd ≈ 1 μM) of substituted benzo-phenoxazine molecules to ckit2.20 The evidence of kinetically controlled CV binding to different structures of human telomeric G-quadruplex DNA can be realized by following the association and dissociation rates of CV in the two structures. Global analysis of FCS data provide a dissociation rate (k−) of 2.4 × 103 s−1 for CV in the (3 + 1) hybrid structure, compared to 17 × 103 s−1 in the antiparallel structure (see Table 1). The 7-fold lower dissociation rate in the (3 + 1) hybrid structure, compared to the antiparallel structure, is consistent with the fact that endstacking/intercalation of CV in the (3 + 1) hybrid structure promotes stronger complex formation. Hence, CV has a lower propensity to dissociate out of this structure. However, the probable (loose) groove-binding of CV inside antiparallel structure promotes a weaker complex formation where easy dissociation of CV is possible. The calculated association rates (k+) show only a mere 1.85-fold increase in the (3 + 1) hybrid structure, compared to the antiparallel structure (see Table 1). This clearly suggests that, although association plays some role initially to form the complex, the lower dissociation rate actually promotes a stronger and long-lived CV-quadruplex complex in the case of (3 + 1) hybrid structure. Nonetheless, the present study reports some of the highest association rates (Table 1) for a molecule like CV to G-quadruplex DNA, albeit

⎛ t ⎞ ⎛ t⎞ I N (t ) = a f exp⎜ − ⎟ + ab exp⎜ − ⎟ ⎝ τf ⎠ ⎝ τb ⎠

(10)

During the global fit, the τf value was kept fixed at 2.6 ns (lifetime for free CV) and τb was kept as a free global (linked) parameter that yielded a value of 290 ps. A monotonic increase 7222

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

(normalized) arf and arb can again be expressed as in eq 11, through which K can be obtained. Global fitting of the entire dataset with eq 12, using two (linked) time constants and their contributions as free global parameters, provides a value of τrb = 3.8 ns (for fixed τrf = 250 ps). Figure 4F shows the variation of arb with quadruplex concentration in two cases. Variation shows similar features as those shown in Figure 4C. Fitting with eq 11 finds K = 0.11 × 106 M−1 and 1.6 × 106 M−1 in the antiparallel and (3 + 1) hybrid structures, respectively (see Table 2). The values are similar to the FCS and lifetime data (see Tables 1 and 2). At this point, it is important to note that, although the binding constants of a chemical reaction can be obtained from various bulk spectroscopic studies, it is difficult to measure the accurate values of association (k+) and dissociation (k−) rates and characteristic reaction time (τR) from ensemble-averaged techniques. Our study shows that FCS is unique for this purpose, which directly provides the kinetic parameters of the reaction. The present data clearly demonstrate how the ligand interacts with G-quadruplexes at their molecular length scale to have specific binding affinities to different G-quadruplex structures. CD and Induced-CD Data. The peak positions and directions of CD spectrum depend on the chirality of a (macro)molecule, predicting its topology.9,28−30 Figure 5A

in ab and a corresponding decrease in af were observed with increasing quadruplex concentration (see the Supporting Information). This suggests more and more complexes are formed with increasing concentration. Taking into account the fact that samples are equilibrium mixtures of free and bound CV, the independent (normalized) fractions of af and ab can be expressed in terms of the binding equilibrium constant (K) as af =

1 1 + (Fb/Ff )K[quadruplex]

ab =

(Fb/Ff )K[quadruplex] 1 + (Fb/Ff )K[quadruplex]

(11)

where Ff and Fb are the fluorescence intensities of free and bound CV, respectively. Either of the af and ab expressions could be used to extract K by fitting their concentrationdependent variation. However, analyzing the variation of ab is much more reliable, because the increase in its value directly follows the fraction of complex formation. We fitted the (normalized) variation of ab, using its expression in eq 11, where the ratio Fb/Ff (at 625 nm) was obtained from the steady-state spectra in Figure 2 (excitation at 470 nm in TCSPC setup, instead of 532 nm, gave similar Fb/Ff values for the samples). Figure 4C plots the variation of ab for both antiparallel and (3 + 1) hybrid quadruplex. Plot clearly shows the difference in binding affinities. The K-values obtained are 0.1 × 106 M−1 and 1.56 × 106 M−1 in the case of antiparallel and (3 + 1) hybrid structures, respectively (see Table 2). These values are in good agreement with those obtained from FCS (also see Table 1). Table 2. Kinetics Parameters Obtained from Time-Resolved Fluorescence and Anisotropy Measurements parameter K (M−1) (time-resolved decays) K (M−1) (anisotropy decays) (Fb/Ff)a a b

antiparallelb

(3 + 1) hybridb

(0.1 ± 0.01) × 106

(1.56 ± 0.12) × 106

(0.11 ± 0.008) × 106 0.082

(1.6 ± 0.1) × 106 0.03

Values obtained from steady-state fluorescence data (Figure 2). Errors are obtained directly from fits.

Another good way to monitor CV−quadruplex complex formation is through measurement of fluorescence (rotational) anisotropy decays. Free CV in solution can rotate very fast. Thus, anisotropy decay of free CV gives a single rotational time constant of 250 ps. However, when CV binds to G-quadruplex, the entire (bulky) complex rotates much slower. Figures 4D and 4E show the anisotropy decays of CV at various concentrations of antiparallel and (3 + 1) hybrid G-quadruplex DNA, respectively. The decays become slower with increasing concentration. A closer look reveals that the rates of slowing of decays are drastically different in the two cases. The decays could be nicely fitted with the sum of two exponential (rotational) time constants: one for free CV (τrf ) and one for bound CV (τrb), as shown in eq 12. ⎡ ⎛ t ⎞⎤ ⎛ t ⎞ r(t ) = r0⎢a fr exp⎜ − r ⎟ + abr exp⎜ − r ⎟⎥ ⎢⎣ ⎝ τf ⎠ ⎝ τb ⎠⎥⎦

arf

Figure 5. (A) Circular dichroism (CD) spectra of G-quadruplex formed in the presence of Na+ (red) and K+ (blue) ions. Black curve shows CV-induced quadruplex formation in the absence of ions. Green dotted curve shows DNA (annealed) without ions and CV. (B) Induced-CD spectrum of CV in the absence of ion (black). (C) CD spectra (annealed) of G-quadruplexes in the presence of only K+ (blue dash) and when CV is added (blue line −1:1 of CV-quadruplex). Plot also includes CD spectra in the presence of only Na+ (red dash) and when CV is added (red line −1:1 of CV-quadruplex). Arrows show CV-induced change in the spectra. (D) Induced CD of CV in the presence of K+ (blue solid) and Na+ (red solid).

shows the CD spectra of quadruplex formed in Na+ and K+ solutions. In Na+ solution, the spectrum shows two positive peaks, at ∼250 and ∼295 nm, and a large negative peak at ∼265 nm, indicating a basket-type antiparallel structure (red curve).29,30 For the structure formed in K+ solution, CD shows a large positive peak at ∼290 nm, indicating contribution from antiparallel conformation. It also shows a positive shoulder at ∼275 nm, a small hump at ∼255 nm, and a

(12)

arb

where and are the respective contributions from the free and bound ligand. Similar variations in these contributions with concentration are observed as in the case of lifetime data. The 7223

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

negative peak at ∼238 nm, indicating contributions from parallel topology. Overall, this spectrum predicts the (3 + 1) hybrid structure of G-quadruplex.9,29,30 One important finding in this study is that CV binds and induces the human telomeric sequence to form quadruplex structures both in the absence and the presence of ions. Figure 5A plots the CD spectrum of CV-induced topology in the absence of ions. The spectrum contains a small negative peak at ∼235 nm and a positive peak at ∼255 nm, along with two positive shoulders at ∼270 and ∼295 nm. The small negative peak at ∼235 nm and the positive shoulder at ∼270 nm indicate the weak contribution of the parallel conformation. However, the positive shoulder at ∼295 nm indicates the contribution from the antiparallel topology, albeit smaller than the pure antiparallel structure formed in Na+ solution. Overall, the spectrum indicates a weakly formed mixed-type hybrid topology. To further confirm the binding and inducing effect of CV, we monitored the induced CD of CV. Figure 5B shows the full CD spectra from 220 nm to 700 nm. CV has (S1 ← S0) absorption near 600 nm and does not have chirality. Thus, no CD can be observed for free CV. However, it showed pronounced induced CD with positive and negative peaks near 600 nm, confirming CV binding and forming Gquadruplex structure in the absence of ions. Figure 5C shows the effect of ligand binding in preformed quadruplex structures in the presence of Na+ and K+. In Na+ solution, the ligand induces antiparallel topology where the characteristic negative peak at ∼265 nm disappears and a positive plateau near this wavelength appears. The 295 nm peak decreases, suggesting a lowering of antiparallel topology. Thus, CV induces the antiparallel structure to form a mixed-type hybrid structure in the presence of Na+ ion, similar to that observed in the absence of ions. On the other hand, in the K+ solution, the ligand induces the positive peak at 290 nm to shift to 285 nm, along with pronounced positive shoulders at ∼275 and ∼260 nm. These changes indicate that CV further stabilizes the (3 + 1) hybrid structure. A similar effect was observed for bromine-substituted guanines (BrG) in telomeric DNA. These guanines are shown to stabilize G-quartets more, compared to pure guanines.9 The effects seen above suggest that CV can stabilize the (3 + 1) hybrid structure through π−π type endstacking/intercalation near G-tetrads. Induced-CD is measured to obtain the binding conformations of the ligand in the two quadruplex structures. Figure 5D shows the full CD spectra. This plot clearly finds different conformations of ligand binding in two structures. In the (3 + 1) hybrid structure, the induced CD shows an all-positive spectrum, possibly indicating a planar π-stacked CV near G-tetrads. However, in the antiparallel structure, the induced-CD shows a combination of positive and negative peaks, similar to that observed in the absence of ions, which probably indicate groove/loop binding of CV in this structure. These findings are fascinating, because of the fact that K+ ions are abundant in an intracellular environment. Thus, the formation of parallel and/or hybrid-parallel G-quadruplex structures is preferred in a cellular environment, which can then be stabilized by ligands such as CV. However, in the antiparallel structure, CV induces the formation of a mixed-type hybrid topology. Because of possible groove-binding and competition with Na+ ions, CV cannot stabilize this newly formed structure to the same extent as in the (3 + 1) hybrid structure. These results suggest that alterations of preformed quadruplex structures depend on how the ligand binds to a specific structure in the presence of ions: End-stacking or

intercalation of the ligand stabilizes the quadruplex structure through π−π interaction, whereas groove/loop binding of the ligand alters the topology by inducing G-tracts and side-loop orientations in the structure. Molecular Docking Studies. To obtain the binding sites of ligand in G-quadruplexes, we performed a docking study of CV with antiparallel (PDB entry 143D)5 and (3 + 1) hybrid (PDB entry 2GKU)6 structures in AUTODOCK.48 Docking results reveal two possible binding sites for the ligand in the (3 + 1) hybrid quadruplex: One end-stacked CV with the first guanine in first G-tetrad near the 5′-end, and another intercalated between two G-tetrads near the 3′-end (see Figure 6A). Both binding modes are found to be almost equally

Figure 6. Cartoons showing possible binding sites of CV inside (A) the (3 + 1) hybrid (PDB entry 2GKU) and (B) the antiparallel (PDB entry 143D) G-quadruplex structures obtained in the docking study. Images on the left-hand side show side views of CV stacking and groove-binding; images on the right-hand side show the models from other angles with a van der Waals surface.

probable and energetically similar. Either of these binding modes can form a stable CV-quadruplex complex in the (3 + 1) hybrid structure. These binding modes also explain the lower dissociation rate of CV in this structure. As soon as CV associates with the (3 + 1) hybrid structure, because of the strong π-interaction with G-tetrads, CV refrains from going out of the structure, giving lower dissociation rates. In the case of the antiparallel structure, two possible binding sites are obtained: one shows that CV is end-stacked with the third G-tetrad near the 5′-end, and another where CV is bound to the largest groove formed by the first and second G-tracts (see Figure 6B). Steady-state and time-resolved fluorescence show that the fluorescence quenching and the decrease in lifetime of CV is higher in the (3 + 1) hybrid structure than in the antiparallel structure. Tight binding of CV through π-type endstacking/intercalation can actually promote efficient electron transfer, which quenches the fluorescence more and make the lifetime faster in the (3 + 1) hybrid structure, compared to the antiparallel structure. However, the relatively less quenching and higher lifetime of CV in the antiparallel structure suggest a probable (loose) groove-binding than a π-stacking. The docking study actually supports this by showing that groovebinding is energetically favored over π-stacking in the antiparallel structure. Thus, molecular docking data provided 7224

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

the first-hand information about possible binding sites of CV inside two G-quadruplex structures, which strongly support the experimental results.

Notes

The authors declare no competing financial interest.

■

■

ACKNOWLEDGMENTS This work is supported by Department of Information Technology (No. 12(4)/2007-PDD), and Department of Science and Technology (No. SR/FTP/PS-16/2007), Govt. of India. We thank Prof. Sudipta Maiti for many helpful technical discussions on FCS setup, as well as for providing data collection software. We thank Dr. Pritam Mukhopadhyay for allowing us to use the spectrophotometer and fluorimeter, and Dr. Aseem Mishra for providing a few components for FCS setup. We also thank Dr. Pratik Sen and Dr. Pradipta Bandhopadhyay for helpful discussion on molecular docking studies. TCSPC and CD data were collected at Advanced Instrumentation Research Facility (AIRF), JNU. S.D.V. and H.S. thank the University Grants Commission (UGC), and N.P., M.K.S., and M.F.K. thank the Council Of Scientific and Industrial Research (CSIR) for fellowships.

CONCLUSION This paper has made a detailed and quantitative study of a ligand’s interaction with two different structures of human telomeric G-quadruplex DNA. It shows how fluorescence correlation spectroscopy (FCS) can be employed to monitor the kinetic steps of a ligand’s (Cresyl Violet, CV) interaction with antiparallel and (3 + 1) hybrid G-quadruplex structures at the (near) single-molecule level. We show how, by incorporating underfilled objective back-aperture and large confocal pinhole in a FCS setup, one can achieve the condition where the reaction time (τR) is made separable from diffusion time (τD). Under this condition, by correlating fluorescence fluctuations and analyzing a full set of titrated correlation curves with global fitting, we were able to measure all the important kinetic parametersK, k+, k−, and τRof the ligand−quadruplex interaction. Some of these parameters are difficult to obtain from ensemble-averaged experiments, which FCS can provide directly. Results reveal that CV binds stronger, by an order of magnitude, to the (3 + 1) hybrid structure, compared to the antiparallel structure. The difference in binding affinities is found to be kinetically controlled by the dissociation rate rather than the association rate of the ligand in the G-quadruplex structures. Kinetic data imply that, as soon as the complex is formed due to association, the 7-times-lower dissociation rate of CV in the case of the (3 + 1) hybrid structure (compared to the antiparallel structure) promotes a stronger and long-lived complex. Measurements of ensembleaveraged time-resolved fluorescence and anisotropy decays provided the binding constants of CV in quadruplexes in bulk solution that fully match with FCS results. Through molecular docking, we saw that the ligand can bind through π-stacking and/or intercalation in the (3 + 1) hybrid structure. However, an energetically favored groove-binding was found in the antiparallel structure. CD and induced-CD spectra confirm the difference in binding modes of CV inside two structures. CD spectroscopy also shows that CV not only binds to the Gquadruplexes, but also alters the structures differently in the presence of different ions. These alterations are assisted by the binding modes of the ligand. End-stacking or intercalation of ligand can stabilize the quadruplex structure through π−π interactions, whereas the loop/groove binding of ligand can actually alter the topology by inducing G-tract and side-loop orientations. We believe that the methods described here and the kinetic information obtained in this study can be extremely useful to characterize small molecules for designing better ligands to target G-quadruplex DNA at their single-molecule level.

■

■

(1) Cech, T. R. Cell 2004, 116, 273−279. (2) Verdun, R. E.; Karlseder, J. Nature 2007, 447, 924−931. (3) Ou, T.; Lu, Y.; Tan, J.; Huang, Z.; Wong, K.; Gu, L. ChemMedChem 2008, 3, 690−713. (4) Neidle, S. Curr. Opin. Struct. Biol. 2009, 19, 239−250. (5) Wang, Y.; Patel, D. J. Structure 1993, 1, 263−282. (6) Luu, K. N.; Phan, A. T.; Kuryavyi, V.; Lacroix, L.; Patel, D. J. J. Am. Chem. Soc. 2006, 128, 9963−9970. (7) Phan, A. T.; Luu, K. N.; Patel, D. J. Nucleic Acids Res. 2006, 34, 5715−5719. (8) Phan, A. T.; Kuryavyi, V.; Luu, K. N.; Patel, D. J. Nucleic Acids Res. 2007, 35, 6517−6525. (9) Lim, K. W.; Amrane, S.; Bouaziz, S.; Xu, W.; Mu, Y.; Patel, D. J.; Luu, K. N.; Phan, A. T. J. Am. Chem. Soc. 2009, 131, 4301−4309. (10) Parkinson, G. N.; Lee, M. P. H.; Neidle, S. Nature 2002, 417, 876−880. (11) Hackett, J. A.; Feldser, D. M.; Greider, C. W. Cell 2001, 106, 275−286. (12) Bodnar, A. G.; Ouellette, M.; Frolkis, M.; Holt, S. E.; Chiu, C. P.; Morin, G. B.; Harley, C. B.; Shay, J. W.; Lichtsteiner, S.; Wright, W. E. Science 1998, 279, 349−352. (13) Neidle, S.; Parkinson, G. Nat. Rev. Drug Discovery 2002, 1, 383− 393. (14) Zahler, A. M.; Williamson, J. R.; Cech, T. R.; Prescott, D. M. Nature 1991, 350, 718−720. (15) Lipps, H. J.; Rhodes, D. Trends Cell Biol. 2009, 19, 414−422. (16) Oganesian, L.; Bryan, T. M. BioEssays 2007, 29, 155−165. (17) Monchaud, D.; Teulade-Fichou, M. P. Org. Biomol. Chem. 2008, 6, 627−636. (18) Burger, A. M.; Dai, F.; Schultes, C. M.; Reszka, A. P.; Moore, M. J.; Double, J. A.; Neidle, S. Cancer Res. 2005, 65, 1489−1496. (19) Drygin, D.; Siddiqui-Jain, A.; O’Brien, S.; Schwaebe, M.; Lin, A.; Bliesath, J.; Ho, C. B.; Proffitt, C.; Trent, K.; Whitten, J. P.; Lim, J. K. C.; Hoff, D. V.; Anderes, K.; Rice, W. G. Cancer Res. 2009, 69, 7653− 7661. (20) McLuckie, K. I. E.; Waller, Z. A. E.; Sanders, D. A.; Alves, D.; Rodriguez, R.; Dash, J.; McKenzie, G. J.; Venkitaraman, A. R.; Balasubramanian, S. J. Am. Chem. Soc. 2011, 133, 2658−2663. (21) Sun, D.; Liu, Y.; Liu, D.; Zhang, R.; Yang, X.; Liu, J. Chem. Eur. J. 2012, 18, 4285−4295. (22) Chang, C. C.; Chu, J. F.; Kao, F. J.; Chiu, Y. C.; Lou, P. J.; Chen, H. C.; Chang, T. C. Anal. Chem. 2006, 78, 2810−2815. (23) Jain, A. K.; Paul, A.; Maji, B.; Muniyappa, K.; Bhattacharya, S. J. Med. Chem. 2012, 55, 2981−2993.

ASSOCIATED CONTENT

S Supporting Information *

Materials, methods, and supporting data. This material is available free of charge via the Internet at http://pubs.acs.org.

■

REFERENCES

AUTHOR INFORMATION

Corresponding Author

*Tel.: +91-11-26738803. Fax: +91-11-26717537. E-mail: sens@ mail.jnu.ac.in. 7225

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226

Analytical Chemistry

Article

(24) Yu, D. K.; Yoo, K. Y.; Shin, B. N.; Kim, I. H.; Park, J. H.; Lee, C. H.; Choi, J. H.; Cho, Y. J.; Kang, I. J.; Kim, Y. M.; Won, M. H. Brain Res. 2012, 1437, 50−57. (25) Bhasikuttan, A. C.; Mohanty, J.; Pal, H. Angew. Chem., Int. Ed. 2007, 46, 9305−9307. (26) Maiti, S.; Chaudhury, N. K.; Chowdhury, S. Biochem. Biophys. Res. Commun. 2003, 310, 505−512. (27) Zhang, Z.; Sharon, E.; Freeman, R.; Liu, X.; Willner, I. Anal. Chem. 2012, 84, 4789−4797. (28) Paramasivan, S.; Bolton, P. H. Nucleic Acids Res. 2008, 36, e106. (29) Paramasivan, S.; Rujan, I.; Bolton, P. H. Methods 2007, 43, 324− 331. (30) Hudson, J. S.; Brooks, S. C.; Graves, D. E. Biochemistry 2009, 48, 4440−4447. (31) Campbell, N. H.; Parkinson, G. N.; Reszka, A. P.; Neidle, S. J. Am. Chem. Soc. 2008, 130, 6722−6724. (32) Phan, A. T.; Kuryavyi, V.; Gaw, H. Y.; Patel, D. J. Nat. Chem. Biol. 2005, 1, 167−173. (33) Haq, I.; Trent, J. O.; Chowdhry, B. Z.; Jenkins, T. C. J. Am. Chem. Soc. 1999, 121, 1768−1779. (34) Magde, D.; Elson, E.; Webb, W. W. Phys. Rev. Lett. 1972, 29, 705−708. (35) Rigler, R.; Mets, Ü .; Widengren, J.; Kask, P. Eur. Biophys. J. 1993, 22, 169−175. (36) Maiti, S.; Haupts, U.; Webb, W. W. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 11753−11757. (37) Haustein, E.; Schwille, P. Annu. Rev. Biophys. Biomol. Struct. 2007, 36, 151−169. (38) Shi, X.; Foo, Y. H.; Sudhaharan, T.; Chong, S.-W.; Korzh, V.; Ahmed, S.; Wohland, T. Biophys. J. 2009, 97, 678−686. (39) Pal, N.; Verma, S. D.; Singh, M. K.; Sen, S. Anal. Chem. 2011, 83, 7736−7744. (40) Nag, S.; Bandyopadhyay, A.; Maiti, S. J. Phys. Chem. A 2009, 113, 5269−5272. (41) Ishii, K.; Tahara, T. J. Phys. Chem. B 2010, 114, 12383−12391. (42) Sasmal, D. K.; Mondal, T.; Mojumdar, S. S.; Choudhury, A.; Banerjee, R.; Bhattacharyya, K. J. Phys. Chem. B 2011, 115, 13075− 13083. (43) Kumar, N.; Sahoo, B.; Varun, K. A. S.; Maiti, S.; Maiti, S. Nucleic Acid Res. 2008, 36, 4433−4442. (44) Choi, J.; Kim, S.; Tachikawa, T.; Fujitsuka, M.; Majima, T. J. Am. Chem. Soc. 2011, 133, 16146−16153. (45) Al-Soufi, W.; Reija, B.; Felekyan, S.; Seidel, C. A M.; Novo, M. ChemPhysChem 2008, 9, 1819−1827. (46) Al-Soufi., W.; Reija, B.; Novo, M.; Felekyan, S.; Kuehnemuth, R.; Seidel, C. A. M. J. Am. Chem. Soc. 2005, 127, 8775−8784. (47) Kerller, D. I.; Kamat, P. V. J. Phys. Chem. 1991, 95, 4406−4410. (48) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.; Belew, R. K.; Olson, A. J. J. Comput. Chem. 1998, 19, 1639− 1662.

7226

dx.doi.org/10.1021/ac3015998 | Anal. Chem. 2012, 84, 7218−7226