https://orcid.org/0000-0001-7500-985X
Abstract
A bisbenzamidine DNA binding agent can be easily dimerized by alkylation of its terminal amidine groups to afford an extended minor groove binder with over 20-fold enhanced DNA affinity toward extended A/T-rich sites. Nuclear magnetic resonance experiments in combination with molecular dynamics simulation studies provide structural insight into the insertion of this compound in the DNA minor groove, and antimicrobial assays demonstrate that the increased affinity translates into higher antileishmanial activity.
Introduction
Small molecules that target DNA can modulate gene expression, repair and cell replication, and have thus found application as anticancer, antimicrobial and antiparasitic agents (1–3). However, despite their potential, the clinical application of these DNA binding agents has been frustrated by their relatively low affinity, their poor sequence selectivity and their preference for short DNA sites. In this context, the development of agents that can target longer DNA sequences remains a key goal that could lead to more effective treatments with fewer adverse reactions, but remains a challenging goal (4–6). This is particularly important for biologically relevant sequences, such as poly[d(A/T)] repeats found in the mitochondrial DNA network of some parasites.
Cationic bisbenzamidines preferentially bind A/T-rich sequences over those containing G/C pairs, but with low affinity and poor selectivity toward particular combinations of A/T pairs (7). Despite their limited sequence discrimination, the interaction of some bisbenzamidines with short A/T-rich sites has led to their therapeutic application as antileishmanial agents (2,8–10). In this context, and inspired by the work of Brent L. Iverson, who demonstrated the potential of cooperative interactions to enhance the recognition properties of DNA intercalators (11–13), we envisioned that bisbenzamidine dimers could show cooperativity and high-affinity DNA binding, while simultaneously recognizing longer A/T-rich sequences. Here, we show that dimeric bisbenzamidines obtained in a single synthetic step from the corresponding monomers display excellent binding affinity toward composite d(A/T) sites containing two AAATTT monomer binding sequences (AAATTT X AAATTT). Moreover, we have found that the increased DNA affinity and selectivity for extended sequences result in higher antiparasitic activity of these dimeric agents.
Materials and methods
Synthesis of 1 and 2
Reagents were purchased from commercial sources and used without further purification. Acetonitrile (HPLC grade) and trifluoroacetic acid (TFA) were purchased from Fisher Scientific. Dimethyl sulfoxide (DMSO) was purchased from Sigma–Aldrich and dried over molecular sieves. Analytical ultra-high-performance liquid chromatography–mass spectrometry (UHPLC–MS) was performed with an Agilent 1260 Infinity II series LC/MS using an SB C18 (1.8 μm, 2.1 mm × 50 mm) analytical column from Phenomenex. Standard conditions for analytical UHPLC consisted of a linear gradient from 5% to 95% of solvent B in 12 min at a flow rate of 0.35 ml/min [A: water–0.1% TFA; B: acetonitrile–0.1% TFA]. Compounds were detected by ultraviolet (UV) absorption at 222, 270 and 330 nm. Electrospray ionization mass spectrometry (ESI/MS) was performed with an Agilent 6120 Quadrupole LC/MS model in positive scan mode using direct injection of the purified compound solution into the MS detector. Peptide purification was performed by preparative reverse-phase high-performance liquid chromatography (RP-HPLC) with an Agilent 1260 Infinity II series LC with an Agilent 1260 Binary Pump, using a Sunfire Prep C18 OBD (5 μm, 19 mm × 150 mm) reverse-phase column from Waters. Standard conditions for preparative RP-HPLC consisted of an isocratic regime during the first 2 min, followed by linear gradients (adjusted for each compound) of solvent B during 30 min (A: water–0.1% TFA; B: acetonitrile–0.1% TFA).
Bisbenzamidine (1). Isophthalaldehyde (515 mg, 3.73 mmol) and sodium cyanoborohydride (345 mg, 5.22 mmol) were added to a round-bottom flask and dissolved in MeCN/H2O (1:1) under nitrogen. 4-Aminobenzamidine dihydrochloride (1.58 g, 7.40 mmol) dissolved in MeCN/H2O (1:1) was added, and the mixture was stirred for 1 h at room temperature. The solution was diluted with MeCN/H2O (1:1), filtered and purified by preparative RP-HPLC. The combined fractions from the HPLC purification were freeze-dried, yielding a white solid that was identified as the TFA salt of the desired product, 1 (1.27 g, 57%). 1H-NMR (300 MHz, DMSO-d6, δ): 4.37 (s, 4H), 6.69 (d, J = 9 Hz, 4H), 7.18–7.41 (m, 6H), 7.61 (dd, 4H), 8.75 (s, 8H). 13C-NMR: 45.8, 111.5, 112.7, 125.8, 126.1, 128.5, 129.7, 139.3, 153.5, 164.1. UHPLC–MS(ESI): [M + 2H]2+ calculated for C22H26N6 = 373.49, found 373.24 (Supplementary Figure S1); MP: 233–247°C.
Bisbenzamidine dimer (2). Bisbenzamidine 1 (150 mg, 0.25 mmol) was dissolved in dried DMSO (1.0 ml) in a dried Schlenk flask under a nitrogen atmosphere. NaH (40 mg, 1.0 mmol, 4 equiv., 60% dispersion in mineral oil) was added and the mixture was stirred at room temperature for 15 min, followed by dropwise addition of a solution of 1,4-bis(bromomethyl)benzene (66 mg, 0.25 mmol, 1.0 equiv.) in dried DMSO (1 ml). The resulting mixture was stirred at room temperature for 24 h. The resulting solution was diluted with 30 ml of Milli-Q water, filtered and purified by RP-HPLC. The combined fractions from the purification step were collected and freeze-dried, yielding an orange pale solid identified as the TFA salt of the desired product, 2 (84.7 mg, 26%). 1H-NMR (300 MHz, DMSO-d6, δ): 4.36 (s, 8H), 4.62, (d, J = 6 Hz, 4H), 6.69 (dd, J = 9, 2 Hz, 8H), 7.19–7.42 (m, 15H), 7.54 (s, 2H), 7.56–7.66 (m, 6H), 8.67 (s, 2H), 8.77 (d, J = 7 Hz, 8H), 9.07 (s, 2H), 9.76 (s, 2H). 13C-NMR: 44.7, 45.8, 111.5, 112.7, 113.8, 125.8, 126.0, 127.6, 128.5, 129.7, 135.4, 139.3, 153.1, 158.5, 162.3, 164.3. UHPLC–MS(ESI): [M + H]+ calculated for C52H54N12 = 847.09, found 847.40; calculated for [M + 2H]2+ = 424.56, found 424.30; calculated for [M + 3H]3+ = 284.70, found 283.36; calculated for [M + 4H]4+ = 212.77, found 212.8 (Supplementary Figure S2); MP: 225–242°C.
SPR studies
The Biacore optical biosensor system by Cytiva, Global Life Science Solutions USA LLC was used to conduct surface plasmon resonance (SPR) measurements. The system was equipped with four channels, and streptavidin-derivatized (SA) CM4 sensor chips were used for the experiments. The SA chips were prepared and the biotinylated DNAs, A3T3 (5′-biotin-GGC AAATTT CAG TTTTT CTG AAATTT GCC-3′) and A3T3cA3T3 (5′-biotin-GGC AAATTT C AAATTT CAG TTTTT CTG AAATTT G AAATTT GCC-3′), were immobilized on chip surfaces, specifically on cells 3 and 4, respectively.
The ligand solutions were prepared using degassed and filtered 50 mM Tris–HCl buffer at pH 7.4 with 100 mM NaCl concentrations and 0.05% (v/v) surfactant P20. A series of ligand concentrations, ranging from 2 nM to 1 μM, were then injected over the DNA-immobilized sensor chip at a flow rate of 100 μl/min for 180 s, followed by buffer flow for ligand dissociation (600−1800 s). After each sample run, the sensor chip surface was regenerated by injecting an acidic 10 mM glycine solution (pH 2.5) for 30 s, followed by several buffer injections to establish a stable baseline for subsequent cycles.
Data analysis was performed using a previously described method (14), which involved subtracting the reference response from the blank cell (cell 1) from the response in each flow cell containing DNA to give a signal (RUobs, response units) directly related to the amount of bound ligand. The expected maximum response (RUmax) per bound ligand in the steady-state region was determined from the molecular weight of the DNA, the ligand molecular weight, and the refractive index gradient ratio of the ligand and DNA. KaleidaGraph 4.0 software was used to plot RUobs versus free ligand concentration (Cfree). r = K1 × Cfree/1 + K1 × Cfree. The equilibrium binding constants (K1) were determined using a one-site binding model, where r = (RUobs/RUmax) represented the moles of bound compound/mole of DNA hairpin duplex, and K1 was the macroscopic binding constant. RUmax in the equation was used as a fitting parameter to evaluate the stoichiometry of the ligand–DNA complex, and that value was compared with the predicted maximal response per bound ligand. Kinetic analysis was achieved by globally fitting the ligand-binding sensorgrams using a standard 1:1 kinetic model with incorporated mass transport-limited binding parameters, as described previously (15,16).
Promastigote viability assay
Leishmania major (MHOM/JL/80/Friedlin) strain, genetically modified by the integration of the mCherry gene into the 18S ribosomal RNA locus, was routinely cultured at 27°C in M-199 medium (Gibco) supplemented with 0.04 M HEPES (pH 7.4), 0.01 mM adenine, 0.01 mg/ml hemin, 0.2 μg/ml biopterin, 0.0001% (w/v) biotin, 10% (v/v) heat-inactivated fetal bovine serum, and an antibiotic cocktail comprising 100 U/ml penicillin and 100 μg/ml streptomycin.
Compounds 1 and 2 were pre-dispensed into black, clear-bottom 96-well plates (Greiner Bio-One). For IC50 value estimations, eight-point one in three dilution curves were generated with a top concentration of 100 μM. Promastigotes were collected in the exponential growth phase and seeded into the microplates containing the compounds, at a density of 100 000 cells per well. Amphotericin B (AMB) at 10 μM and 0.33% (v/v) DMSO were used as maximum effect and no effect controls, respectively. After an incubation period of 72 h, the fluorescence, emitted by viable promastigotes, was detected using a Spark plate reader (Tecan) with excitation at 575 nm and emission at 620 nm.
Computational studies
DNA and ligand preparation
cgA3T3cA3T3cg DNA sequence was built using the nab module of AMBER Tools 21 as right-handed β-DNA (17). The DNA protein coordinates for cgcgA2T2cgcg DNA sequence were taken from PDB ID 1PRP (18). The ligand geometry was minimized using a restricted Hartree–Fock method and a 6-31G(d) basis set, as implemented in the ab initio program Gaussian 09. Partial charges were derived by quantum mechanical calculations by using Gaussian 09, as implemented in the R.E.D. Server (version 3.0) (19,20), according to the RESP model (21). The missing bonded and nonbonded interaction parameters were assigned, by analogy or through interpolation, from those already present in the AMBER database (ff14SB) (22).
Docking
These studies were performed with AutoDock Vina (23) and AutoDock 4.2.6 (24).
Minimization of the DNA/ligand complexes
The cgcgA2T2cgcg/1 and cgA3T3cA3T3cg/2 DNA complexes obtained by docking were immersed in a truncated octahedron of TIP3P water molecules (∼4800 and ∼12 800, respectively) and neutralized by addition of sodium ions. Molecular mechanics parameters from bsc1 (for DNA) (25) and the ff14SB (for ligand) force fields were assigned to the DNA structures with the LEaP module of AMBER Tools 21. The system was minimized in four stages: (i) minimization of the ligands (1000 steps, first half with the steepest descent and the other half with the conjugate gradient); (ii) minimization of the solvent and ions (5000 steps, first half with the steepest descent and the other half with the conjugate gradient); (iii) minimization of the purine and pyrimidine bases, water and ions (5000 steps, first half with the steepest descent and the other half with the conjugate gradient); and (iv) final minimization of the whole system (5000 steps, first half with the steepest descent and the other half with the conjugate gradient). A positional restraint force of 50 kcal/(mol Å2) was applied to these unminimized atoms during the first three stages (i–iii).
Simulations of the DNA/ligand complexes
Molecular dynamics (MD) simulations were performed using the pmemd.cuda_SPFP module in the AMBER 20 suite of programs (26–28). Periodic boundary conditions were applied, and electrostatic interactions were treated using the smooth particle mesh Ewald method (29), with a grid spacing of 1 Å. The cutoff distance for the unbound interactions was 9 Å. The SHAKE algorithm was applied to all bonds containing hydrogen atoms using a tolerance of 10–5 Å (30) and an integration step of 2.0 fs (31). The minimized system was then heated at 300 K and 1 atm by increasing the temperature from 0 to 300 K over 100 ps and by maintaining the system at 300 K for another 100 ps. A positional restraint force of 50 kcal/(mol Å2) was applied to all P, OP1, OP2, O3′, C3′ and O5′ atoms during the heating stage. Finally, the system was equilibrated at constant volume (200 ps with positional restraints of 5 kcal/(mol Å2) to P, OP1, OP2, O3′, C3′ and O5′ atoms) and constant pressure (another 100 ps with positional restraints of 5 kcal/(mol Å2) to P, OP1, OP2, O3′, C3′ and O5′ atoms). The positional restraints were gradually reduced from 5 to 1 kcal/(mol Å2) (5 steps, 100 ps each), and the resulting systems were allowed to equilibrate (100 ps) without further restraints. Unrestrained MD simulations were carried out for 100 ns. System coordinates were collected every 10 ps for further analysis. The molecular graphics program PyMOL and UCSF Chimera were used for visualization and depiction of the DNA/ligand complex structures. The cpptraj module in AMBER 16 was used to analyze the trajectories and to calculate the root-mean-square deviation (RMSD) of the DNA during the simulation (32). For DNA measurements, P, OP1, OP2, O3′, C3′ and O5′ atoms were considered, while for the ligands all heavy atoms were used.
Minimization of ligands 1 and 2
They were performed following the same protocol as for DNA/ligand complexes, except in step 3 in which a positional restraint force of 50 kcal/(mol Å2) was applied to the carbon atoms in the central aromatic ring (connector).
Simulations of ligands 1 and 2 in water
Heating and equilibration steps were carried out as for DNA/ligand complexes, with positional restraint force of 50 kcal/(mol Å2) applied to the carbon atoms of the central aromatic ring. Unrestrained MD simulations were carried out for 20 ns and system coordinates were collected every 20 ps for further analysis.
Binding free energy calculations
The binding free energy was calculated by the MM/PBSA approach (33–37), as implemented in AMBER Tools 21. ante-MMPBSA.py module was used to create topology files for the complex, DNA and ligands, and binding free energies were calculated with the MMPBSA.py module. A single trajectory approach was used to calculate binding free energies considering only the last 80 ns of the 100-ns MD trajectories. Both Poisson–Boltzmann and generalized Born implicit solvation models were employed, providing similar results.
Results and discussion
Our work started by analyzing the DNA-bound structure of the previously described bisbenzamidine 1, which can be synthesized in good yield in a single reductive amination step from commercially available 4-aminobenzamidine (33). This monomeric benzamidine shows good affinity for A/T-rich DNA sites with dissociation constants in the low μM range, and particularly for the sequence AAATTT with an approximate dissociation constant (KD) of 0.7 μM (38), and therefore appeared to us as an ideal starting point for the development of dimeric bisbenzamidines.
Binding of 1 to the DNA by NMR and in silico studies
Nuclear magnetic resonance (NMR) experiments were carried out to monitor the interactions between 1 and the Dickerson dodecamer, a well-known DNA construct that features a central AATT site d(CGCGAATTCGCG) (39). NMR assignments for the dodecamer and its 1:1 complex with 1 were carried out using TOCSY and NOESY experiments in H2O and D2O. The analysis of the NMR data showed that both the ligand and the DNA display significant chemical shift perturbations, indicating a molecular recognition event (Figure 1). For the dodecamer, major changes were observed at the A6–C9 region, especially for T7 H4′ and T8 H4′, as well as for the anomeric H1′ protons of A6, T7, T8 and C9. Fittingly, all these hydrogens are oriented toward the minor groove, strongly suggesting that the ligand is targeting that region. Moreover, chemical shift perturbations of the corresponding H5′ protons of T7, T8 and C9 were also observed. These hydrogen atoms are also oriented toward the minor groove and were readily identified since they are now strongly shielded (δ 3.5–3.8 ppm) instead of being overlapped in the crowed region at δ ∼ 4.0–4.3 ppm. Moreover, several strong and medium-sized nucleotide–ligand intermolecular NOEs could also be identified in a non-ambiguous manner (Supplementary Table S1).
Left: Chemical shift perturbations of the Dickerson dodecamer. Right: Putative 3D model of the complex between the Dickerson dodecamer and ligand 1. The water molecules in the vicinity of the minor groove are shown as semitransparent sticks and the nucleotides that are more strongly affected by the interaction with 1 are highlighted. The 1PRP crystal structure was used as a reference structure and the ligand obtained by MD studies using the experimental NOE data. Snapshot taken after 100 ns of simulation. Structure made with UCSF Chimera 1.16 (18).
This experimental information was used to generate a putative 3D model of the DNA/1 complex, using the available crystal structure of the Dickerson dodecamer in complex with pentamidine (PDB ID 1PRP) (40) and manually docking the ligand following the observed intermolecular NOEs. The resulting binary DNA/1 complex was further studied by MD simulations using the molecular mechanics force field ff14SB of AMBER 21. One hundred nanosecond simulations were performed. These types of in silico studies have proven to be a powerful tool for studying the strength of the interactions of a ligand bound to a biomacromolecule along with the overall arrangement of the complex.
The generated 3D structure for the cgcgA2T2cgcg/1 DNA complex (Figure 1) proved to be stable and no significant structural changes were observed during the whole simulation. Thus, the analysis of the RMSD of the whole DNA backbone as well as compound 1 showed relatively low RMSD values (2.1 and 2.1 Å, respectively; Supplementary Figure S3). The proposed 3D structure of the complex provides a very satisfactory agreement with the observed NMR data. In fact, H5 and H6 at the phenyl moieties of the ligand display short distances with the ribose H1′, H4′ and H5′ protons of A6, T7, T8 and C9, located at the minor groove, in full agreement with the NOE observations. Moreover, the three aromatic rings of the ligand are placed above H4′ and/or H5′ of the ribose moieties of T7, T8 and C9 at both strands establishing CH–π interactions. The ligand would be anchored to the groove through diverse water-mediated hydrogen-bonding interactions with the phosphate group of the nucleotides T7, T8 and C9, and the bases of G4, A5, A6 and T8 (Supplementary Figure S4). These interactions fittingly explain the observed chemical shift shielding for those ribose protons upon complex formation (Figure 1).
This structure suggested that a dimeric bisbenzamidine DNA binder could be assembled by a simple head-to-tail dimerization strategy connecting the amidinium moieties of two bisbenzamidines bound to adjacent A/T-rich sites.
Synthesis of the bisbenzamidine dimer 2
Given the known nucleophilicity of amidines in SN2 reactions (41), we decided to synthesize linear dimers connected through their terminal amidine groups by reacting bisbenzamidine 1 with various bis(bromomethyl)benzene linking reactants. However, the dimerization reaction proved to be more difficult than initially expected and we could not isolate any relevant product when 1 was treated with 1,3-bis(bromomethyl)benzene. Fortunately, when the reaction was carried out with the isomeric 1,4-bis(bromomethyl)benzene, we could isolate the desired dimer 2 in a 26% yield using NaH as base in DMSO (Figure 2 and Supplementary Data). Additionally, we isolated a small proportion of the bisbenzamidine trimer in a 15% yield after preparative RP-HPLC purification and observed traces of the tetrameric product as well, which unfortunately could not be isolated in sufficient quantity and purity for a complete characterization. Although the overall efficiency of the synthesis is reduced by the time-consuming RP-HPLC purification step, the good chromatographic separation of the reaction components should allow large-scale purification using reverse-phase flash chromatography.
Synthesis of bisbenzamidine dimer 2.
Spectroscopic characterization of the DNA binding of 2
Having at hand the desired dimer 2, we began to explore its DNA binding properties by circular dichroism (CD) spectroscopy (42–44). Thus, incubation of a 5 μM solution of a short oligonucleotide containing a composite site with two adjacent AAATTT benzamidine binding sequences (A3T3cA3T3, 5′-… AAATTT c AAATTT …-3′) with 2 resulted in the appearance of an intense positive induced CD band at ∼330 nm, consistent with the insertion of the tetraamidine 2 into the minor groove of the DNA, as shown in Figure 3A (42–44). A control experiment with an oligonucleotide, G3C2, lacking the A/T-rich binding sites showed a much weaker induced CD band upon addition of 2, further confirming that such CD band must arise from the insertion of the bisbenzamidine units in the DNA minor groove (Figure 3B).
CD spectra of 2 with an A/T-rich oligonucleotide and a G/C-rich control. (A) 5 μM solution of A3T3cA3T3 (dashed line), and the same solution in the presence of 5 equiv. of 2 (solid line); UV spectrum of 2 is shown on top for reference. (B) Same as in panel (A) but with the control G/C-rich oligo G3C2. Experiments were run in 20 mM Tris–HCl buffer (pH 7.5) and 100 mM NaCl. A3T3cA3T3 hairpin DNA: 5′-GGC AAATTT c AAATTT CAG TTTTT CTG AAATTT G AAATTT GCC-3′ (A/T-rich binding site in upper case, T5 loop in italics); G3C2 hairpin DNA: 5′-GGCAGGCCCAGC TTTTT GCTGGGCCTGCC-3′ (T5 loop in italics).
DNA binding studies by SPR
Initial experiments to characterize the DNA binding of 1 and 2 using fluorescence titrations showed that 2 had very slow kinetics and advised a different approach for a precise quantitative determination of their DNA dissociation constants (Supplementary Figure S10). Thus, Biacore SPR methods were utilized to evaluate quantitatively binding of the monomer diamidine 1 and the dimer tetraamidine 2 to hairpin DNA with A/T-rich sequences of 5′-AAATTT-3′ (15). For this purpose, a 5′-biotin-tagged DNA with a single site was immobilized in one cell of a streptavidin-coated CM4 sensor chip, while a similar DNA with two sites was immobilized on a second sensor cell (15,45,46).
The binding results of the two compounds were found to be significantly different. The monomer displayed association and dissociation rates that were faster than can be measured by SPR chip technology (Figure 4). Therefore, the sensorgrams for the monomer appeared as square waves due to the rapid on–off kinetics. The steady-state single-site affinity binding function was used to calculate the dissociation constant (KD) of the monomer 1, resulting in an affinity of 4900 nM (Figure 4) for the single-site sequence. On the other hand, the dimer 2 binding to the two-site sequence exhibited slow association kinetics [ka = (7.7 ± 2) × 105 M−1 s−1] and much slower dissociation kinetics [kd = (2.7 ± 3) × 10−3 s−1], resulting in a very remarkable much stronger binding than the monomer. The dissociation constant for the dimer was determined to be 3.4 ± 2 nM because of the very low dissociation rate constant (Figure 4). The monomer binding constant to the two-site sequence is 4800 nM, which is within experimental error of the single-site binding constant. On the other hand, the dimer binding to the single-site sequence is 181 nM, which is much weaker than the value for the two-site sequence. These values are as expected since the monomer can bind to both sites in the two-site sequence, while the dimer can only partially bind to the single-site sequence.
(A) Representative SPR sensorgrams and affinity analysis of 1 and 2 binding to the DNA oligos A3T3 and A3T3cA3T3. Insets in all cases show 1:1 steady-state affinity fitting (A) A3T3/1 complex at 0.05, 0.07, 0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 2.0, 3.0 and 5.0 μM 1, (B) A3T3cA3T3/1 complex at same concentrations of 1, (C) A3T3/2 complex at 0.02, 0.03, 0.05, 0.07, 0.1, 0.2, 0.3, 0.5 and 0.7 μM of the dimeric bisbenzamidine 2 and (D) A3T3cA3T3/2 complex. Concentrations of 2, from bottom to top, are 20, 30, 50, 70 and 100 nM. The gray lines are the kinetic fitting of the results with a single-site function. The experiments were conducted in Tris–HCl buffer (50 mM Tris–HCl, 100 mM NaCl, 1 mM EDTA, pH 7.4) at 25°C.
It is worth noting that the core unit of the monomer features six single bonds, and rotation about those bonds gives a large array of conformations in solution (Supplementary Figure S7A); the diamidine can rotate all six bonds very rapidly toward reaching the required conformation for minor groove insertion. In contrast, the dimer tetraamidine 2 has the same six bonds in each binding unit for a total of 12, and four additional single bonds in the linker (Supplementary Figure S7B). Given the large number of possible molecular conformations that 2 can adopt, it is not surprising that the association of the dimer is much slower than that of the monomer. The on rate in SPR for the second order interaction of 2 with A3T3cA3T3 DNA at 30 nM, a concentration significantly above the KD, for example, has an apparent half-life of 43 s. However, once bound, the dimer forms many favorable interactions with the AAATTT sites in the minor groove, and the dissociation rate is quite low, resulting in strong binding. The plasticity differences of both ligands can be easily visualized on studying their intrinsic motion in water by MD simulations (Supplementary Figure S5). For example, for ligand 2, displacements of the position of its terminal amidine groups of up to ∼27 Å were identified, while of only ∼9 Å for compound 1. While each monomeric unit in the dimer can be quite dynamic, with a single unit moving in and out of the groove quite rapidly, the rate for both to simultaneously move out of the groove for complete complex dissociation is quite low. The half-life for first-order dissociation reaction is >250 s, indicating strong binding. It is evident that linking two relatively weak binding units has resulted in a compound with much stronger binding.
Computational modeling of the dimer 2
To get a deeper insight into the insertion of 2 in the DNA minor groove, we performed in silico studies, taking as reference the plausible three-dimensional structure of the monomer 1 bound to the Dickerson dodecamer derived from the NMR data. The cgA3T3cA3T3cg DNA sequence was built using the nab module of AMBER Tools 21 as right-handed B-DNA. The most plausible cgA3T3cA3T3cg/2 DNA complex obtained by docking was subjected to 100 ns of MD simulation to provide a more realistic picture of the ligand conformation upon binding. To this end, the model in a truncated octahedron of TIP3P water molecules and neutralized by addition of sodium ions using the molecular mechanics force field ff14SB of AMBER 21 was then subjected to 100 ns of dynamic simulation. The CGCGAATTCGCG/1 DNA complex was also studied.
Fitting the experimental observations, dimer 2 was found in our simulations to insert snugly into the minor groove and nearly symmetrically with respect to the central G/C base pair allowing each of the two bisbenzamidine monomers to interact with an AAATTT site (Figure 5A and Supplementary Figure S8). Both the complex and the ligand proved to be very stable during the 100-ns MD run because no significant changes were identified during the whole simulation, showing RMSD average values of 2.5 and 2.2 Å for DNA backbone and ligand, respectively (Supplementary Figure S6). Despite the great flexibility of the dimer 2 mentioned above, this DNA sequence captures the optimal helical arrangement of the ligand thanks to two direct hydrogen-bonding interactions with the bases of nucleotides T14 and C9, which were not identified for monomer 1 (Figure 5B). These direct contacts observed in our model proved to be reasonably stable during the simulation as they were observed during >60% of the simulation (Figure 5C). Among both contacts, the interaction with T14 showed to be stronger with an average distance between the involved atoms of 2.8 Å. The strength of the interaction was further supported by comparison of the binding free energies of ligands 1 and 2 to CGCGAATTCGCG and CGAAAATTTCAAATTTCG DNA sequences, respectively, which were calculated using the MM/PBSA method in explicit water (generalized Born) as implemented in AMBER (Table 1). This method estimates the free energy difference between two states, i.e. bound and unbound states of two solvated molecules, and can be directly correlated to dissociation constants as ΔGbinding = −2.303RT log K. The affinity of 2 for A/T-rich sequences proved to be stronger (50 kcal/mol) than that of the parent ligand, in agreement with the experimental data, which might be caused by its ability to establish direct contacts with the nucleotide bases.
![(A) Proposed three-dimensional model of the cgA3T3cA3T3cg/2 DNA complex obtained by MD simulation studies. Compound 2 is shown in spheres. Snapshot taken after 60 ns of the MD simulation. (B) Detailed view of the main hydrogen-bonding interactions (dashed lines) of 2 with cgA3T3cA3T3cg. The proposed direct contacts of the ligand with the nucleotide bases (C9, T14 and T15) are also highlighted. These strong contacts were not observed for the CGCGAATTCGCG/1 DNA complex. (C) Variation of the relative distances between the ligand and DNA atoms involved in direct hydrogen-bonding interactions during the whole simulation: d1 [O2 (T14 chain B) and H52 (2) atoms], d2 [O2 (T14 chain B) and H02 (2) atoms] and d3 [O2 (C9) and H55 (2) atoms].](https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/narmolmed/1/4/10.1093_narmme_ugae022/3/m_ugae022fig5.jpeg?Expires=1747898894&Signature=XiXHuTjTYI1Vk9hfUw52t6y0IP38WhkUpG7k2Znq7j0znQxVzkqjxEEXgb1MJyeKNg~4EuB6NDYHdx-mtv67FMZN30A-XKv9~m4f9tByVSFExsXIbMYwA4RkT8ZEfyieUhb8X6idCAySxkBglZ5btq~Zb1Vds~fNiv5AIdKr6fBF9LK1qkf5O1XVAp1aKmhP32UpuSvA1JOCZOKNCDmo~ntu9oe~-kWzRvruMDUUEUNDog58UZSkywn1Dz12TO9lL4~xazEZG~tdNocey-ePG7epcSsWgJkRXxtbcbTw76OneGljaCgNiuGA9mQceo3FjBJfEi~yjX8GEFm4As1Bcg__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
(A) Proposed three-dimensional model of the cgA3T3cA3T3cg/2 DNA complex obtained by MD simulation studies. Compound 2 is shown in spheres. Snapshot taken after 60 ns of the MD simulation. (B) Detailed view of the main hydrogen-bonding interactions (dashed lines) of 2 with cgA3T3cA3T3cg. The proposed direct contacts of the ligand with the nucleotide bases (C9, T14 and T15) are also highlighted. These strong contacts were not observed for the CGCGAATTCGCG/1 DNA complex. (C) Variation of the relative distances between the ligand and DNA atoms involved in direct hydrogen-bonding interactions during the whole simulation: d1 [O2 (T14 chain B) and H52 (2) atoms], d2 [O2 (T14 chain B) and H02 (2) atoms] and d3 [O2 (C9) and H55 (2) atoms].
Calculated binding free energies using MM/PBSA
| Ligand | Complex | ΔG (kcal/mol) | ΔΔG |
|---|---|---|---|
| 1 | CGCGAATTCGCG/1 | −42.2 ± 0.2a | +50 |
| 2 | CGAAAATTTCAAATTTCG/2 | −92.0 ± 0.2a | 0 |
| Ligand | Complex | ΔG (kcal/mol) | ΔΔG |
|---|---|---|---|
| 1 | CGCGAATTCGCG/1 | −42.2 ± 0.2a | +50 |
| 2 | CGAAAATTTCAAATTTCG/2 | −92.0 ± 0.2a | 0 |
Only the last 80 ns of the whole MD simulation were considered for the calculations.
aStandard error of mean.
Calculated binding free energies using MM/PBSA
| Ligand | Complex | ΔG (kcal/mol) | ΔΔG |
|---|---|---|---|
| 1 | CGCGAATTCGCG/1 | −42.2 ± 0.2a | +50 |
| 2 | CGAAAATTTCAAATTTCG/2 | −92.0 ± 0.2a | 0 |
| Ligand | Complex | ΔG (kcal/mol) | ΔΔG |
|---|---|---|---|
| 1 | CGCGAATTCGCG/1 | −42.2 ± 0.2a | +50 |
| 2 | CGAAAATTTCAAATTTCG/2 | −92.0 ± 0.2a | 0 |
Only the last 80 ns of the whole MD simulation were considered for the calculations.
aStandard error of mean.
Improved antiparasitic activity of the dimer 2
Following the in vitro characterization, and given that pentamidine, another bisbenzamidine, is currently used as a second-line drug for the treatment of leishmaniasis, we decided to evaluate the antileishmanial activity of our bisbenzamidine monomer and dimer (9,47,48). The antiparasitic effect of both compounds was assessed on L. major (MHOM/JL/80/Friedlin) strain, genetically modified to constitutively produce the mCherry fluorescent protein (mCh) when parasites are viable. mCh+L. major promastigotes were routinely cultured at 27°C (see ESI for experimental details) and the IC50 value was determined by seeding the promastigotes at a density of 100 000 cells per well, into 96-well plates containing 3-fold dilutions of compounds 1 or 2. AMB at 10 μM and 0.33% (v/v) DMSO were used as positive and negative controls, respectively. After an incubation period of 72 h, we measured the fluorescence at 620 nm, emitted by viable promastigotes. Interestingly, although both compounds were active against the Leishmania extracellular form, the dimer (IC50= 2.5 ± 0.1 μM) exhibited an IC50 10-fold lower than the monomer (IC50= 32.9 ± 3.6 μM), thus demonstrating that the oligomerization promotes a significant increase on the antileishmanial activity, as shown in Figure 6. This observation is consistent with the highest DNA binding affinity showed by the dimer 2, in comparison with the monomer, since one of the proposed mechanisms of action of the bisbenzamidines is that their interaction with DNA disrupts the activity of replication-associated proteins, thereby leading to the inhibition of parasite growth (48). For comparison, pentamidine, which is used for the treatment for cutaneous leishmaniasis, was slightly less potent than the dimer 2, showing an IC50 of 6.6 ± 0.5 μM, and the percentage of promastigote viability after the treatment with AMB at 10 μM, which was taken as reference for maximum effect, was 4.6 ± 0.2%, while dimer 2 showed a value of 6.4 ± 1.5% at similar concentration. Altogether, these data reinforce the potential of the dimerization strategy for improving the potency of antileishmanial agents.
Dose–response curves of the bisbenzamidine monomer 1 (white circles) and dimer 2 (black circles) against mCh+ L. major promastigotes. The dose–response data for pentamidine are also shown as reference (white diamonds). Promastigotes were incubated at 27°C for 72 h with different concentrations of each compound and fluorescence at 620 nm, emitted by viable promastigotes, was measured in a microplate reader.
In summary, monomeric bisbenzamidines can be readily dimerized through an alkylation reaction with the bis-electrophile 1,4-bis(bromotethyl)benzene; the resulting dimeric agent exhibits high affinity for tandem A/T-rich DNA sequences because of the cooperative interaction of the two monomers with their respective sites. Importantly, such increased interaction derives from a much lower dissociation kinetics, a strategy that could be exploited for the design of other DNA binding agents. Detailed structural studies and molecular modeling demonstrate that these new agents maintain their ability to form tight contacts with the edge of the minor groove. Importantly, we have also found that the increased interaction with A/T-rich sites translates into antileishmanial activities for the dimer that are at least one order of magnitude higher than those of the monomer, as measured by their IC50 values.
Data availability
The data underlying this article are available in the article and in its online supplementary material. These include detailed synthetic procedures, NMR spectra for compounds 1 and 2, and additional figures for the MD simulation studies (Supplementary Figures S3–S6). The three-dimensional structures of the DNA complexes CGCGAATTCGCG/1 (PDB1) and CGAAAATTTCAAATTTCG/2 (PDB2) are also provided. We have submitted the two structures in the papers to ModelArchive: ‘Bisbenzamidine bound to the Dickerson dodecamer’ (https://www.modelarchive.org/doi/10.5452/ma-d86z7) and ‘Bisbenzamidine dimer bound to a composite DNA sequence featuring two A3T3 bisbenzaminide binding sites’ (https://www.modelarchive.org/doi/10.5452/ma-p6lcv).
Supplementary data
Supplementary Data are available at NARMME Online.
Acknowledgements
We thank Renata Calo-Lapido for preliminary experiments. Molecular graphics images were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by National Institutes of Health P41 RR001081) and the molecular graphics program PyMOL. All authors are grateful to the Centro de Supercomputación de Galicia (CESGA) for use of the Finis Terrae computer. JJB also acknowledge CIBERES, an initiative of Instituto de Salud Carlos III (ISCIII, Madrid, Spain).
Funding
Agencia Estatal de Investigación [RTI2018-099877-B-I00, PID2022-136963OB-I00, PID2021-127702NB-I00]; Xunta de Galicia [ED431C 2021/29, ED431G 2023/03]; European Regional Development Fund; CICECO [IF/00894/2015, UIDB/50011/2020, UIDP/50011/2020, LA/P/0006/2020]; EDCTP [RIA2020S-3301]; ISCIII [PI21CIII/00005]; CIBERINFEC [CB21/13/00018].
Conflict of interest statement. None declared.
Notes
Present address: Mateo I. Sánchez, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK.
References
Author notes
The first three authors should be regarded as Joint First Authors.
Comments