Extending the σ-Hole Motif for Sequence Specific Recognition of the DNA Minor Groove

Abstract

The majority of current drugs against diseases, such as cancer, can bind to one or more sites in a protein and inhibit its activity. There are, however, well known limits on the number of druggable proteins and complementing current drugs with compounds that could selectively target DNA or RNA would greatly enhance therapeutic progress and options. We are focusing on the design of sequence-specific DNA minor groove binders that, for example, target the promoter sites of transcription factors involved in a disease. We have started with AT specific minor groove binders that are known to enter human cells and have entered clinical trials. To broaden the sequence-specific recognition of these compounds, we have identified several modules that have H-bond acceptors that strongly and specifically recognize G•C base pairs. A lead module is a thiophene-N-alkyl-benzimidazole σ-hole based system with terminal phenyl-amidines where the optimum compounds have excellent affinity and selectivity for a G•C base pair in the minor groove. Efforts are now focused on optimizing this module. We have previously optimized the alkyl group. In the work described here we are evaluating modifications to the compound aromatic system with the goal of improving GC selectivity and affinity as with the N-alkyl modifications. The lead compounds from these studies retain the thiophene-N-alkyl-BI module but have halogen substituents adjacent to an amidine group on the terminal phenyl-amidine. Other improved compounds in this set have modified amidines and conversion of the amidine to an imidazoline, for example, resulted in a strong binding compound with good specificity.

Graphical Abstract

Introduction

There are intense research efforts underway to design new synthetic compounds that have excellent selectivity for targeting RNA or DNA sequences and/or structures^1–11. Such compounds could have a major impact on the treatment of important for diseases that are the result of aberrant behavior of transcription factors. Because nucleic acids are at an early stage in the disease development process, targeting them has many advantages, especially since many transcription factors are classified as “undruggable”¹². Relative to drugs that bind to protein sites, however, drugs that are specific for DNA are relatively rare and except for aminoglycosides, are even more rare with RNA. Small molecule binding sites in proteins are relatively well-understood but the number of druggable proteins is only a very small part of the genome¹³. With this limit on protein-drug binding sites, the interest in therapeutic agents that are specific for RNA or DNA has rapidly increased.

Duplex DNA sequences and sequence-dependent local microstructures offer useful and specific targets for binding small molecules^3,5,9,10. Major non-duplex structural forms of nucleic acids, such as quadruplexes, also offer the possibility of selective binding interactions and a targeted therapeutic effect^14,15. With RNA an enormous variety of different structural forms are known and being discovered, but the selection of specific target regions and structures that will have a therapeutic effect and finding compounds specific for the structure remains a major challenge^1,2,4,6,13.

Our development efforts have focused on novel compounds to bind with high selectivity properties in the DNA minor groove. Results so far in several laboratories have found the strongest binding, optimum recognition of DNA and most specific synthetic compounds bind in the DNA minor groove. Compounds that have served as models for AT sequence-specific minor groove binding over an extended time period include pentamidine, Hoechst 33258, netropsin, DAPI, furamidine, and analogs of all of these compounds. Such compounds are classical minor groove agents, they are all highly AT sequence-specific binders, and they have been used in therapeutic applications or for biotechnology purposes such as cell staining^16–26. The compounds generally have H-bond donor groups, such as benzimidazoles, amides, and amidines/guanidiniums, that interact with the A-N3 and T=O groups on the edges of A·T base pairs (bps) at the floor of the minor groove. The G-NH₂ that projects into the groove interferes with the binding of these compounds both sterically and electronically. To recognize a G·C bp then it is necessary to incorporate H-bond accepter groups into designed, synthetic minor groove binders^27,28. The GC recognizing compounds should have a shape that allows space for the G-NH₂ H-bond, follows the curvature of the minor groove and can also recognize AT bps in the target sequence. Needless to say, the design and preparation of such compounds is a challenge and to date has only rarely been accomplished^9,10,27–36.

To initiate a project to systematically develop compounds that could specifically recognize G·C bps, a focused library of approximately 200 heterocyclic, cationic minor-groove binders with H-bond acceptor groups was evaluated. This analysis produced several initial compounds that could bind to a single G·C bp in an AT context. Three new compounds that bound strongly and specifically to a single G·C bp flanked by A·T bps (Figure S1 in Supporting Information) were selected for the initial study. The accepter groups in these compounds are azabenzimidazole, pyridine and a thiophene-N-methylbenzimidazole (N-MeBI) σ-hole motif which is preorganized into a conformation to match the curvature of the minor groove⁷. We have recently explored changing the methyl substituent on the thiophene-N-MeBI unit in a successful project to enhance the binding selectivity of that module⁹. The idea behind these substituent changes was to better recognize microstructural variations in the minor groove in addition to sequence recognition^9,37,38. The results with the modified compounds were quite successful, with a modest loss in affinity to single G•C sequences, but with a major gain in binding selectivity over closely related sequences. The sequences that were tested, for example, were a single G•C sequence with an AAAGTTT binding site, and for selectivity a pure AT-binding sequence with an AAATTT site and a GC containing the binding site, AAAGCTTT (Figure 1B). These binding sites were incorporated in hairpin DNA duplexes for binding analysis.

Figure 1.

A) Chemical structure of the modified compounds from DB2457 used in this study. B) The DNA sequences used in this study; DNA sequences used for SPR studies were labeled with 5’-biotin.

In this report, we have built on the results with the modified thiophene-N-MeBI module, which illustrated that modified structures could have significantly improved recognition properties. We have addressed three specific, important questions in this work: what is the effect on binding affinity and selectivity of (i) modifying the terminal amidine group, (ii) incorporating different halogen substituents at different positions in the aromatic systems, and (iii) changing the basic structure of the thiophene-N-MeBI aromatic system? Modifications to include different halogens are based on successful, previous results with a -Cl substituent⁹. The amidine structure modifications as well as replacing the N-alkyl-BI-thiophene group by other chemically and stereochemically similar structures (Figure 1) are exploratory in order to establish the limits on the thiophene-N-alkyl-BI chemical space. The results with these new compounds illustrate new methods of design in the thiophene-σ-hole system with excellent specificity for recognition of G•C bps can be obtained even without the thiophene or alkyl-BI groups.

Experimental

Materials

In the DNA thermal melting (T_m) and circular dichroism (CD) experiments, the hairpin oligomer sequences were used (Figure 1B). In SPR experiments, 5′-biotin-labeled hairpin DNA oligomers were used. All DNA oligomers were obtained from Integrated DNA Technologies, Inc. (IDT, Coralville, IA) with reverse-phase HPLC purification and mass spectrometry characterization.

The buffer used in T_m and CD experiments was 50 mM Tris-HCl, 100 mM NaCl, 1 mM EDTA, pH 7.4 (TNE 100). The biosensor-surface plasmon resonance (SPR) experiments were performed in filtered, degassed TNE 100 with 0.05% (v/v) surfactant P20.

UV-vis Thermal Melting (T_m)

DNA thermal melting experiments were performed on a Cary 300 Bio UV-vis spectrophotometer (Varian). The concentration of each hairpin DNA sequence was 3 μM in TNE 100 using 1 cm quartz cuvettes. The solutions of DNA and ligands were tested with a ratio of 2:1 [ligand] / [DNA]. All samples were increased to 95 °C and cooled down to 25 °C slowly before each experiment. The spectrophotometer was set at 260 nm with a 0.5 °C/min increase beginning at 25 °C, which is below the DNA melting temperature and ending above it at 95 °C. The absorbance of the buffer was subtracted, and a graph of normalized absorbance versus temperature was created using KaleidaGraph 4.0 software. The ΔT_m values were calculated using a combination of the derivative function and estimation from the normalized graphs.

Biosensor-Surface Plasmon Resonance (SPR)

SPR measurements were performed with a four-channel Biacore T200 optical biosensor system (GE Healthcare, Inc., Piscataway, NJ). A streptavidin-derivatized (SA) CM5 sensor chip was prepared for use by conditioning with a series of 180 s injections of 1 M NaCl in 50 mM NaOH (activation buffer) followed by extensive washing with HBS buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.05% P20, pH 7.4). Biotinylated-DNA samples (AAATTT: 5’-biotin-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-biotin-CCAAAGTTTGCTCTCAAACTTTGG-3’; AAAGCTTT: 5’-biotin-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’) of 25–30 nM were prepared in HBS buffer and immobilized on the flow cell surface by noncovalent capture. Flow cell 1 was left blank as a reference, while flow cells 2–4 were immobilized separately by manual injection of biotinylated-DNA stock solutions (flow rate of 1 μL/min) until the desired amount of DNA response units (RU) was obtained (150–250 RU). Ligand solutions were prepared with degassed and filtered TNE 100 with 0.05% (v/v) surfactant P20 by serial dilutions from a concentrated stock solution. Typically, a series of different ligand concentrations (2 nM to 500 nM) were injected over the DNA 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 cycle, the sensor chip surface was regenerated with a 10 mM glycine solution (pH 2.5) for 30 s followed by multiple buffer injections to yield a stable baseline for the following cycles. The reference response from the blank cell was subtracted from the response in each flow cell containing DNA to give a signal (RU, response units) that is directly proportional to the amount of bound compound. The predicted maximum response per bound compound in the steady-state region (RU_max) was determined from the DNA molecular weight, the amount of DNA on the flow cell, the compound molecular weight, and the refractive index gradient ratio of the compound and DNA. RU was plotted as a function of free ligand concentration (C_free), and the equilibrium binding constants were determined with a one-site binding model (K₂ = 0).

$$r=\left(K_1·C_{\text{free}}+2K_1·K_2·{\text{C}}_{\text{free}}{}^2\right)/\left(1+K_1·C_{\text{free}}+K_1·K_2·C_{\text{free}}{}^2\right)$$ (1)

where r represents the moles of bound compound per mole of DNA hairpin duplex, K₁ and K₂ are macroscopic binding constants (for a single-site or 1:1 model K₂ = 0), and C_free is the free compound concentration in equilibrium with the complex. RUmax in the equation was used as a fitting parameter, and the obtained value was compared to the predicted maximal response per bound ligand to evaluate the stoichiometry. Kinetic analysis was performed by globally fitting the binding results for the entire concentration series using a standard 1:1 kinetic model with integrated mass transport-limited binding parameters as described previously^39,40.

Circular Dichroism (CD)

Circular dichroism experiments were performed on a Jasco J-1500 CD spectrometer in 1 cm quartz cuvette at 25 °C. A buffer scan as a baseline was collected first in the same cuvette and subtracted from the scan of the following samples. The hairpin DNA sequence AAATTT: 5’-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-CCAAAGTTTGCTCTCAAACTTTGG-3’ (5 μM), Figure 1B, in TNE 100 was added to the cuvette prior to the titration experiments, and then the compound was added to the DNA solution and incubated for 10 min to achieve equilibrium binding for the DNA-ligand complex formation. For each titration point, four spectra were averaged from 500 to 230 nm wavelength with a scan speed of 50 nm/min, with a response time of 1 s. Baseline-subtracted graphs were created using the KaleidaGraph 4.0 software.

Competition Electrospray Ionization Mass Spectrometry (ESI-MS)

Electrospray Ionization Mass Spectrometry (ESI-MS) analyses were performed on a Waters Q-TOF micro Mass Spectrometer (Waters Corporate, Milford, MA) equipped with an electrospray ionization source (ESI) in a negative ion mode. DNA sequences AAATTT (5’-CCAAATTTGCCTCTCGAAATTTGG-3’), AAAGTTT (5’-GCCAAAGTTTGCCTCTGCAAACTTTGGC-3’) and AAAGCTTT (5’-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’), for ESI-MS experiments were purified by dialyzing it in 50 mM ammonium acetate buffer (pH 6.7) at 4 °C with 3x buffer exchange. Test samples were prepared in 50 mM ammonium acetate with 10% v/v methanol at pH 6.7 and introduced into the ion source through a direct infusion at 5 μl/min flow rate. The competitive experiments were done by mixing a ligand and DNAs with different sequences at different ratios. The instrument parameters were typically as follows: capillary voltage of 2800 V, sample cone voltage of 30 V, extraction cone voltage of 1.0 V, desolvation temperature of 70 °C, and source temperature of 100 °C. Nitrogen was used as nebulizing and drying gas. A multiply charged spectra were acquired through a full scan analysis at mass range from 300–2500 Da and then deconvoluted to the spectra presented. MassLynx 4.1 software was used for data acquisition and deconvolution.

Fluorescence Spectroscopy Binding Determinations

Fluorescence spectra were recorded on a Cary Eclipse Spectrophotometer, with excitation and emission slit width as determined depending on the concentrations of ligands. The free compound solutions at different concentrations were prepared in an appropriate buffer in TNE 100, and DNA sequence (AAATTT: 5’-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-CCAAAGTTTGCTCTCAAACTTTGG-3’; AAAGCTTT: 5’-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’) aliquots were added from a concentrated stock. All titration spectra were collected after allowing an incubation time of 10 min. DB2802 was excited at 342 nm and DB2803 was excited at 336 nm based on molecular absorbance from UV-vis spectroscopy. Emission spectra of these compounds were monitored from 200 nm wavelength range. All the fluorescence titrations were performed at 25 °C. Then the Fluorescence titration spectrums and fitting plots were made in Kaleidagraph 4.0 software to determine the K_D value.

Molecular Dynamics (MD) Simulations.

Structure optimization of DB2789 was performed by using DFT/B3LYP theory with the 6–31+G* basis set in Gaussian 09 (Gaussian, Inc., 2009, Wallingford, CT) with Gauss-view 5.09⁴¹. Partial charges were derived using the RESP fitting method (Restrained Electrostatic potential)^42,43. AMBER 16 (Assisted Model Building with Energy Refinement) software suite was used to perform molecular dynamics (MD) simulations⁴⁴. Canonical B-form ds[5’-CGAAAGTTTCG-3’][5’-CGAAACTTTCG-3’] DNA was built in Nucleic Acid Builder (NAB) tool in AMBER. AMBER preparation and force field parameter files required conducting molecular dynamics simulations for the DB2789 molecule by using ANTECHAMBER⁴⁵. Specific atom types assigned for DB2789 molecule were adapted from the ff99 force field. Most of the force field parameters for DB2789 molecule were derived from the existing set of bonds, angles and dihedrals for similar atom types in parm99 and GAFF force fields⁴⁶. Some dihedral angle parameters were obtained from previously reported parametrized data^47,48. The molecular structure with specific atom types used for the DB2789 molecule is shown in Figure S3. Parameters of DB2789 in frcmod file are listed at the Table S1.

The AutoDock Vina program was used to dock the DB2692 in the minor groove of DNA to obtain the initial structure for the DB2789-DNA complex⁴⁹. MD simulations were performed in explicit solvation conditions where the DNA-DB2789 complex was placed in a truncated octahedron box filled with TIP3P water using xleap program in AMBER. Sodium ions were used to neutralize the system. A 10 Å cutoff was applied on all van der Waals interactions. The MD simulation was carried out using the Sander module with SHAKE algorithm applied to constrain all bonds. Initially, the system was relaxed with 500 steps of steepest-descent energy minimization. The temperature of the system was then increased from 0 K to 310 K over 10 ps under constant-volume conditions. In the final step, the production run on the system was subsequently performed for 500 ns under NPT (constant-pressure) conditions.

Results

Chemistry

Scheme 1 outlines the synthesis of the final diamidines 6a-h. The nitro derivatives 2a-e were obtained by reaction of the bromo derivatives 1a-c with different boronic acid esters under the Suzuki coupling conditions using palladium tetrakis triphenylphosphine as a catalyst in dioxane and potassium carbonate solution. These nitro compounds were reduced to the corresponding amines 3a-e using tin dichloride dihydrate in ethanol. The diamines were converted to the corresponding benzimidazole dinitriles 5a-g on reaction with different aldehydes with the aid of sodium metabisulphite in DMSO. The intermediate dinitriles were converted to the final diamidines through reaction with either lithium bis(trimethylsilyl)amide in THF followed by deprotection of the silylated intermediate using HCl gas in ethanol, or applying Pinner methodology where the dinitriles are suspended and stirred in ethanolic HCl to form the imidate ester hydrochloride that was allowed to react with different amines to afford the final diamidines 6a-h. The synthesis of the diamidine 11 is described in Scheme S1 (supplemental information). 4-Ethynylbenzonitrile 7 was allowed to react with the bromo derivative 8 under Sonogoshira coupling conditions using palladium (II)bis(triphenylphosphine) dichloride and copper iodide as catalysts in a mixture of THF and triethylamine to afford the aldehyde 9. This intermediate aldehyde was converted to the benzimidazole 10 and then to the final diamidine 11, as described in Scheme 1. Compound 18 was prepared as described in Scheme S2. The intermediate bromo derivative 14 was prepared by reaction of the iodo derivative 12 and the tin compound 13 utilizing Stille cross-coupling conditions using palladium tetrakis triphenylphosphine as a catalyst in dioxane. The bromo derivative 14 was allowed to couple with cyanophenylboronic acid to afford the mono nitrile intermediate 15 by applying Suzuki conditions. This nitrile was brominated using NBS in DMF to furnish the intermediate bromo compound 16. This bromo derivative was converted to the dinitrile intermediate 17 using Suzuki coupling conditions. The final diamidine 18 was prepared after reaction lithium bis(trimethylsilyl)amide with the dinitrile as described before. Scheme S3 describes the synthesis of the final diamidine 25, starting from the bromoisatin 19. 2-Acetylthiophene 20 and 19 were heated in ethanolic potassium hydroxide to furnish the quinoline carboxylic acid derivative 21. The carboxylic acid 21 and copper cyanide in NMP heated at reflux led to cyanation and decarboxylation, furnishing the cyanoquinoline derivative 22. Compound 22 was subjected to bromination using NBS, followed by Suzuki coupling to produce the dinitrile 24. This dinitrile was converted to the final amidine 25 using the lithium bis(trimethylsilyl)amide as described before. The synthesis of the diamidines 29a,b is described in Scheme S4. Suzuki coupling of the bromo derivative 26 with cyanophenylboronic acid followed by condensation with different amines to produce the dinitriles 28a,b. Pinner conditions were applied to this dinitrile to produce the final compounds 29a,b.

Scheme 1. Reagents and conditions:

a) 4-cyanophenylboronic acid derivatives, Pd(PPh₃)₄, Na₂CO₃/H₂O, dioxane, reflux; b) SnCl₂/EtOH, reflux; c) Na₂S₂O₅/DMSO, 130 °C; d) i- LiN(TMS)₂/THF, ii- HCl/EtOH.OR i- HCl/EtOH. ii- appropriate diamine.

DNA Thermal Melting: Screening for Relative Binding Affinity

Changes in thermal melting temperature (T_m) of DNA provide a rapid way for initial screening of ligands for binding affinity with different DNA sequences (Table 1). Three related DNA sequences that we have successfully used in comparative studies⁹ were selected for testing (Figure 1B). In order to determine the binding selectivity of our compounds, we chose two quite similar sequences, AAATTT and AAAGCTTT, as the control sequences for the AAAGTTT target site.

Table 1.

Thermal Melting Studies (ΔT_m, ℃) of All Test Compounds (Figure 1A) with Pure AT and Mixed DNA Sequences.^a

ΔTm (℃)	AAATTT 70 ℃	AAAGTTT 66 ℃	AAAGCTTT 67 ℃
DB2457b	6	14	5
DB2708b	4	14	5
DB2798 (6a)	3	6	2
DB2759b	3	12	3
DB2801 (6b)	3	10	3
DB2789 (6c)	3	12	4
DB2740b	5	17	7
DB2788 (6d)	5	14	6
DB2647 (6e)	2	4	1
DB2650 (6f)	5	3	2
DB2786 (6g)	4	15	5
DB2787 (6h)	3	11	2
DB2770 (11)	2	7	4
DB2655 (18)	9	13	6
DB2654 (25)	6	10	2
DB2802 (29a)	9	8	3
DB2803 (29b)	3	7	2

^a.ΔT_m = T_m (the complex) - T_m (the free DNA). 3 μM DNA sequences were studied in TNE100 with the ratio of 2:1 [ligand]/[DNA]. An average of two independent experiments with a reproducibility of 0.5 ℃. Full DNA sequences: AAATTT: 5’-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-CCAAAGTTTGCTCTCAAACTTTGG-3’; AAAGCTTT: 5’-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’.

^b.Reported compounds.

The parent compound, DB2457, with an N-methyl substituent was previously reported⁷ and has preferential binding to the single G·C bp sequence and weaker binding to the pure AT and two G·C bps sequences (Table 1). A compound with N-isopropyl substituents, DB2708, has similar binding to AAAGTTT as DB2457, and significantly weaker binding to the pure AT sequence for improved selectivity as desired⁹. DB2759 with a -Cl substituted on the phenyl ring keeps strong binding to AAAGTTT while binding with the two control sequences drops to an undetectable amount under the standard conditions.

Based on the successful results with modified N-substituents in improving the binding selectivity, focused modifications of the thiophene compounds in other parts of the structure is attractive to search for improved affinity and selectivity. First, considering the improved selectivity of the -Cl derivative, DB2759, -F, -Br halogen modifications were made as well as changing the position of the -Cl substituent to the opposite phenyl ring (Table 1). Both -Cl and the -Br compounds bind with the AAAGTTT sequence strongly and with good selectivity compared to AAATTT and AAAGCTTT. Surprisingly, the -F substituent binds much weaker to AAAGTTT compared to other halogen compounds. We also tested two compounds with modifications on the amidine group, DB2786, and DB2787. The isopropyl substituent obviously improves the binding to AAAGTTT, while the imidazoline compound has similar binding to all three sequences as DB2708⁹. In addition to N-isopropyl substituents, we also tested an N-Ph substituted compound, DB2740, and found that it has strong binding and good selectivity for the single G•C sequence. DB2788 with a -Cl substituent was also prepared based on the success of the substituent in previous studies. In this compound, however, the -Cl modification did not significantly increase the binding selectivity for the N-Ph substituted compound.

In order to explore the effects on binding from variations of the basic aromatic structure, we also introduced a C≡C bond linkage into the structure (DB2770) to change the connection between phenyl rings. The acetylene linkage, however, reduces the binding by a large amount. Changing the N-RBI into a pyridine group produces DB2655, which binds strongly to AAAGTTT but with lower selectivity than the BI compounds. Changing the N-RBI into a quinoline (DB2654) decreases the single G·C bp sequence binding affinity but maintains a high level of selectivity. In addition to the N-RBI module, the thiophene was modified to a thiazole ring (DB2647 and DB2650) and, surprisingly, the simple change to a thiazole causes the compounds to entirely lose the function of the thiophene σ–hole system with the consequence of very poor binding. In addition to the thiazole, the thiophene was replaced with benzothiophene, DB2802 and DB2803. For both methyl and isopropyl BI substituted compounds, the binding affinity to the target AAAGTTT sequence decreases with a decrease in binding selectivity. The N-Me derivative, DB2802, binds to the pure AT test sequence in a similar manner to the single G·C bp sequence.

Biosensor-SPR: Methods for Quantitative Binding

A biacore sensorchip (CM5) functionalized with streptavidin was used to immobilize 5’-biotin labeled hairpin duplex DNA sequences (Figure 1B) in flow cells 2–4 and flow cell 1 was left as a blank, for background subtraction. With different compounds in the flow solutions, we were able to determine comparative binding constants for all the derivatives (Table 2). Sensorgrams were obtained and are shown for representative compounds with the different DNAs (Figure 2). With the AAATTT and AAAGCTTT sequences, most compounds showed significantly weaker binding than to AAAGTTT, with faster on and off rates. No kinetics and limited K_D values could be accurately determined for the pure AT and the two G•C bps DNA sequences with most compounds. However, with the single G•C bp containing AAAGTTT sequence, excellent sensorgrams were obtained at low concentrations for most derivatives and representative SPR sensorgrams are shown in Figure 2. Kinetics fits with a one-site model were excellent and allowed a determination of K_D values. Kinetics fits were generally required to determine K_D values since many sensorgrams did not reach a steady-state level, especially at the lower concentrations.

Table 2.

Summary of Binding Affinity (K_D, nM) for the Interaction of All Test Compounds with 5’-Biotin-labeled DNA Sequences using Biosensor-SPR Method.^a

KD (nM)	AAATTT	AAAGTTT	AAAGCTTT
DB2457b	222	4	192
DB2708b	1020	4	223
DB2798 (6a)	NB	632	NB
DB2759b	NB	14	NB
DB2801 (6b)	NB	14	NB
DB2789 (6c)	NB	11	NB
DB2740b	NB	3	107
DB2788 (6d)	NB	10	957
DB2647 (6e)	NB	NB	NB
DB2650 (6f)	NB	NB	NB
DB2786 (6g)	713	1	55
DB2787 (6h)	NB	11	1890
DB2770 (11)	NB	354	NB
DB2655 (18)	108	12	315
DB2654 (25)	NB	62	NB

^a.All the results in this table were investigated in Tris-HCl buffer (50 mM Tris-HCl, 100 mM NaCl, 1 mM EDTA, 0.05% P20, pH 7.4) at a 100 μL min^–1 flow rate. NB means no detectible binding. The listed binding affinities are an average of two independent experiments carried out with two different sensor chips and the values are reproducible within 10% experimental errors. Full DNA sequences: AAATTT: 5’-biotin-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-biotin-CCAAAGTTTGCTCTCAAACTTTGG-3’; AAAGCTTT: 5’-biotin-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’.

^b.Reported compounds.

Figure 2.

Representative SPR sensorgrams for (A–C), DB2801, (D–F), DB2787 in the presence of AAATTT, AAAGTTT, and AAAGCTTT hairpin DNAs. In (A, C, D, and F), the concentrations of sensorgrams are 2–100 nM of each compound from bottom to top. In (B), the concentrations of DB2801 from bottom to top are 15, 30, 50, 70, and 100 nM; In (E), the concentrations of DB2787 from bottom to top are 15, 30, 50, 90, 100 nM. In (B and E), the solid grey lines are best-fit values for the global kinetic fitting of the results with a single site function. Full DNA sequences: AAATTT: 5’-biotin-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-biotin-CCAAAGTTTGCTCTCAAACTTTGG-3’; AAAGCTTT: 5’-biotin-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’.

While the parent compounds, DB2457 and DB2708, have strong, selective binding to the AAAGTTT sequence, it is necessary to increase the selectivity of new compounds to a higher level for biological applications. As previously reported, DB2759 with -Cl on the aromatic system binds to AAAGTTT with a 14 nM K_D and is very selective with no detectable binding to AAATTT and AAAGCTTT⁹. To find the optimum halogen substituent, different halogen substituents were incorporated and tested in the aromatic systems: DB2798 (-F), DB2801 (-Br) and DB2789 with -Cl on the other phenyl of the aromatic system. The different modification positions of -Cl do not make a major difference in binding affinity when DB2789 and DB2759 are compared. The -Br modification maintains similar binding affinity and specificity to AAAGTTT as with the -Cl derivatives (Figure 2). It is interesting that the smaller, electron-withdrawing atom, F, causes the binding affinity to become 40 times weaker than the -Cl modified compound (K_D = 632 nM vs 14 nM). The potential intramolecular hydrogen bonding between the adjacent -F and -H of the amidine group limits the rotation of amidine on the phenyl ring to adopt an appropriate angle for binding to DNA minor groove and significantly reduces the binding affinity of the amidine group with the adjacent T=O of DNA.

As previously reported, DB2740 with an N-Ph substituent binds strongly to AAAGTTT but with limited selectivity versus AAAGCTTT⁹. The above results suggested that a -Cl substituent would improve the selectivity of the N-phenyl compound. To evaluate this point, DB2788 was synthesized and it binds to AAAGTTT slightly weaker than DB2740, but with 10-fold better binding selectivity to AAAGCTTT. This result shows that a -Cl on the aromatic systems, adjacent to an amidine group, can increase the AAAGTTT binding selectivity of different thiophene-N-RBI type compounds.

To evaluate the effects of changes in the chemical space of the compounds, two groups were also introduced into the amidines of DB2708. DB2786 with isopropyl amidine groups represents the strongest binding (K_D = 1 nM) to AAAGTTT in this series. Unfortunately, DB2786 loses binding selectivity with a 55 nM K_D for AAAGCTTT. DB2787 with amidine replaced by imidazoline binds to AAAGTTT similar to the -Cl compound but with slightly worse selectivity (Figure 2). In order to confirm the importance of the thiophene-N-RBI in the core system of the original compounds. Five compounds with basic structure variations of the thiophene σ-hole system were prepared. With the introduction of an alkyne between thiophene and phenyl ring of DB2708, the binding to AAAGTTT dramatically decreases. The replacement of N-RBI with pyridine, DB2655 gives a strong binding compound but with poor selectivity. The replacement of N-RBI-phenyl with quinoline, DB2654, gives a compound with good selectivity but weaker binding affinity. Surprisingly, the thiazole-N-RBI substituents destroyed the binding ability for all three DNA sequences. This is a surprising and puzzling result given the success of thiazole in other minor groove binders⁵⁰. These results indicate the thiophene, or related derivative such as benzothiophene, is necessary but can be combined with the other kinds of N containing heterocyclic groups for a 1, 4 N···S interaction to produce a σ-hole system.

Circular dichroism (CD): Determination of DNA Binding Mode

CD titration experiments are an effective and convenient method of evaluating the binding mode and the saturation limit for compounds binding with DNA sequences. CD spectra monitor the asymmetric environment of the binding of the compound to DNA and therefore, can be used to obtain information on the binding mode^51,52. No CD signals have been observed at 300–450 nm wavelength for the free compounds. However, in Figure 3 (left, middle), on the addition of the compounds into DNA, substantial positive induced CD signals (ICD) are seen in the absorption region between 300 and 450 nm. These positive ICD signals in this region indicate a minor groove binding mode by these ligands, as expected from their structures. When the Ligand/DNA ratio is titrated to 1:1, the saturated spectra indicate 1:1 binding between ligand and AAAGTTT. However, in Figure 3 (right), when we change sequence from AAAGTTT to AAATTT, the ICD is low and indicates weak binding between DB2788 and the pure AT sequence. In summary, as can be seen from Figure 3, DB2786 and DB2788 give CD spectra typical of complexes in the minor groove of the AAGTTT sequences with a 1:1 stoichiometry, in agreement with SPR results and modeling.

Figure 3.

Circular dichroism spectra for the titration of representative compounds, DB2786, DB2788, with a 5 μM AAAGTTT or AAATTT sequence in TNE100 at 25℃. Arrows indicate the changes. Full DNA sequences: AAATTT: 5’-CCAAATTTGCCTCTGCAAATTTGG-3’; AAAGTTT: 5’-CCAAAGTTTGCTCTCAAACTTTGG-3’

Fluorescence Titrations: Binding affinity

Fluorescence spectroscopy is an effective method to quantitatively measure the interactions between biomolecules and small molecules, such as a small organic molecule binding with DNA. If a change in the fluorescence intensity accompanies the binding of two species, this can be used to monitor the binding interaction and determine the stoichiometry of binding using equilibrium titration methods^53,54.

In order to study the binding specificity of DB2802 and DB2803 with benzothiophene, which have non-optimum properties for SPR studies, fluorescence titration experiments were conducted. The fluorescence titration spectra and binding affinity fitting plots are shown in Figure S2 (Supporting Information). The fluorescence intensity of DB2802 increases on titration with AAATTT. For AAAGTTT and AAAGCTTT, the fluorescence intensities both decrease in the titration. From DNA thermal melting results, the binding affinity between DB2802 with AAATTT and AAAGTTT are similar and stronger than AAAGCTTT. DB2803 with an N-isopropyl group binds to AAAGTTT strongly compared to the pure AT and AAAGCTTT sequences. The results agree with our previous analysis of the N-isopropyl benzimidazole that showed strong binding to AAAGTTT with good selectivity. The above results indicate that the isopropyl group on benzimidazole matches the microstructure of AAAGTTT better than methyl group and increases the binding specificity for both the thiophene and benzothiophene groups.

Competition electrospray ionization mass spectrometry (ESI–MS): A Direct Determination of Binding Stoichiometry and Binding Specificity with Relative Binding Affinity

Mass spectrometry is an excellent method for the evaluation of relative binding affinity and specificity as well as stoichiometry. It provides a good test of experimental results obtained from other methods, such as SPR, where macromolecule-ligand stoichiometry is obtained by fitting the signal at different concentrations^8,55. For this series of compounds, we chose two representative thiophene-BI substituents, DB2787, and the parent compound DB2708, for the competition ESI-MS experiments. DB2787 binds to sequence AAAGTTT with 11 nM K_D value and good selectivity. DB2708 binds to AAAGTTT with 4 nM K_D value but the binding selectivity is less than DB2787. In Figure 4A, three free DNA peaks are shown for AAATTT (m/z = 7302), AAAGCTTT (m/z = 7921), and AAAGTTT (m/z = 8540). On addition of DB2787, the peak for AAAGTTT (m/z = 8540) decreases with the simultaneous appearance of a new peak at m/z = 9070 that is characteristic of a 1:1 AAAGTTT–DB2787 complex (Figure 4B). There is no appearance of any complex peak with the other DNA sequences. As shown in Figure 4C, the same three DNA sequences are used to test DB2708. On the addition of DB2708, the peak of AAAGTTT almost disappeared with a new peak at m/z = 9019 that is characteristic of a 1:1 AAAGTTT–DB2708 complex (Figure 4D). However, another small new peak at m/z = 8540 appeared and is consistent with a 1:1 AAAGCTTT-DB2708 complex. Since the binding affinity between the DNA sequence AAAGCTTT and DB2708 is 223 nM K_D (AAAGTTT/AAAGCTTT binding ratio is 56), with the increasing titration of DB2708 (up to 4:1 ratio), the binding between DB2708 and AAAGCTTT can be observed. Moreover, DB2787 with excellent selectivity (AAAGTTT/AAAGCTTT binding ratio is 172) only shows the AAAGTTT complex under these conditions. When excess ligands were added in the DNA mixture solution, only 1:1 Ligand/DNA complex peaks were observed. These results confirm the 1:1 stoichiometry between ligands and DNA, which agrees with CD results.

Figure 4.

ESI-MS negative mode spectra of the competition binding of sequences AAATTT (5’-CCAAATTTGCCTCTCGAAATTTGG-3’), AAAGTTT (5’-GCCAAAGTTTGCCTCTGCAAACTTTGGC-3’) and AAAGCTTT (5’-CCAAAGCTTTGCTCTCAAAGCTTTGG-3’) (10 μM each); with 40 mM DB2787 or DB2708 in buffer (50 mM ammonium acetate with 10% methanol (v/v), pH 6.8). (A and C) The ESI-MS spectra of free DNA mixture. (B and D) The ESI-MS spectra of DNA mixture with compounds. The ESI-MS results shown here are deconvoluted spectra and molecular weights are shown with each peak.

Molecular Structure: Molecular Dynamics

To help better understand the structural basis of molecular recognition of DNA sequences with a single G·C bp in an AT context, molecular dynamics (MD) simulations for the -Cl substituted N-isopropyl compound, DB2789, with the [5’-CGAAAGTTTCG-3’][5’-CGAAACTTTCG–-3’] duplex sequence were conducted. The 500 ns MD simulation has been performed by using the Amber 16 software package in the presence of 0.15 M NaCl⁴⁴. Force constants for DB2789 were determined as described previously and in the Methods Section and added to the force field for the simulations^46,47. Three optimum H-bonds in the complex have been observed. Both amidine groups form -N–H to T=O H-bonds (Figure 5A) that are an average of 2.9 Å in length. The amidines also form frequent highly dynamic H-bonds to terminal water molecules that move in and out of the minor groove. These water molecules frequently also form H-bonds with A·T bps at the floor of the minor groove and help link the compound to the specific binding site in the groove and stabilize the complex. The third strong H-bond is from the central G-NH that projects into the minor groove to the unsubstituted imidazole N in the BI group of DB2789, Figure 5A, to account for much of the binding selectivity of DB2789. Additional selectivity in binding is provided by the -CH group of the six-member ring of BI that points into the minor groove. This -CH forms a dynamic close interaction with the -N3 of the dG base of the central G·C bp. Additional direct stabilizing interactions are formed by other -CH groups that point to the floor of the minor groove from the two phenyls of DB2789. These -CH groups form significant dynamic interactions with A-N3 and T=O groups on the bases at the floor of the minor groove. Although DB2789 is optimally oriented along the minor groove with appropriate twist to match the minor groove curvature, the bulky -Cl group at the -ortho position of the bottom amidine group restricted the degree of freedom of the rotation of the -amidine group (Figure 5B, D). On the other hand, the top amidine group shows dynamic amidine group rotations due to the constraint reduction (Figure 5C, E).

Figure 5.

Molecular Dynamics (MD) model of DB2789 bound to an AAAGTTT site: (A) A wire-ball and stick model viewed into the minor groove of the AAAGTTT binding site with bound DB2789. The DNA bases are represented in tan-white-red-blue-yellow(C–H–O–N–P) color scheme and DB2789 is magenta-white-blue-yellow-green (C–H–N–S-Cl) color scheme. The important interactions between different sections of the DB2789-DNA complex are illustrated in yellow dashed lines. The terminal amidine groups form strong hydrogen bonds with the carbonyl groups of dT=O. The imidazole-N makes a strong hydrogen bond interaction with the exocyclic NH of dG. The BI–C–H that points to the floor of the minor groove forms a strong interaction with the N3 of dG in the minor groove. (B) and (C) represent the atoms of two amidine groups that were selected for the dihedral angle plots in (D) and (E).

Discussion

We are engaged in a project that first time is converting strictly AT specific heterocyclic diamidine DNA minor groove binders into compounds that can bind to mixed AT and GC sequences of DNA¹¹. This involves the creation of a set of modules that can recognize a single G•C bp and combining the modules or truncated versions of them to recognize a range of complex sequences, such as promoter sites of transcription factors. The ability to target such sites could have a major impact on understanding of transcription factor function as well as therapeutic development. An example is the PU.1 transcription factor that is involved in the development of a number of cancers such as AML. There are no drugs effective against PU.1, and it has been classified as “undruggable”. In cases like PU.1 it seems more effective to target the promoter sequence rather than the transcription factor for the treatment of AML⁵⁶.

The first step in this project involved the design and synthesis of a broad array of heterocyclic diamidines that could recognize a single G•C bp in an AT context^7,9,57 (Figure S1). This effort produced several compounds that were selective for the PU. 1 promotor target sequence. One of our lead compounds was based on a thiophene-N-methyl-benzimidazole σ-hole system that is preorganized to bind to the DNA minor groove (Figure 1).

Our goal in the research described in this report is to probe the limits on the thiophene-N-MeBI module to determine if there are improved structures for strong and specific binding to G•C bps. As noted in the Introduction, three specific, important questions are addressed: what is the effect on binding affinity and selectivity of (i) modifying the amidine groups, (ii) incorporating different halogen substituents in the aromatic systems, and (iii) modifying the basic structure of the thiophene-N-MeBI aromatic system? In the halogen series, our initial compound had a -Cl substituent substituted adjacent to the amidine on the phenyl connected to the N-alkyl-BI group. To determine if this is the optimum halogen, -Br and -F were substituted in the same position. In quantitative SPR studies, the -Br group performed identically to the -Cl with good binding and excellent selectivity for the single G•C bp sequence. This substituent diversity may be helpful in targeting different cell types. Somewhat surprisingly, the -F substituent gave very poor binding, which we attribute to its locking the adjacent amidine into an unfavorable orientation for H-bonding to DNA through H-bonding to an adjacent amidine -NH. Putting the -Cl on the phenyl at the other end of the compound gave both stronger binding and excellent specificity for the single G•C bp sequence. The benzimidazole was also modified with an N-phenyl substituent, which gave quite strong binding but poor selectivity for the two G•C bps DNA sequence. A -Cl was added to determine if its enhanced selectivity would be available in this system. The N-phenyl-Cl compound has somewhat weaker binding, as with other -Cl derivatives, but with improved selectivity, although not as much improvement as desired.

Modifying the amidines with isopropyl groups gave the strongest binding of any compound in the series to the single G•C bp sequence, but unfortunately, the binding to the two G•C bps DNA was also enhanced. Converting the amidine to an imidazoline slightly reduced the binding to the single G•C bp sequence but the selectivity for the single G•C bp sequence over 150. A major question was whether the thiophene-N-alkyl-BI module was required or whether other σ-hole structures with H-bond acceptors would recognize the single G•C bp sequence. The first compounds in these experiments were what we thought was a conservative substitution of thiazole for thiophene to make two isomeric compounds, DB2740 and DB2788. It was a surprise that neither of these compounds gave any detectible binding to any of the three test DNAs, especially, since the triazole group has performed well in other minor groove binders⁵⁰. To maintain the σ-hole the N-alkyl-BI group was replaced with a pyridine, DB2655, and quinoline, DB2654. The pyridine had satisfactory binding to the single G•C bp sequence but poor selectivity for the two G•C bps sequence and especially for the pure AT sequence. The quinoline had significantly reduced binding to the single G•C bp sequence but no detectible binding to the AT or GC test sequences.

In conclusion, there are three lead compounds from this design, synthesis and biophysical studies of GC recognition agents. All three compounds have halogen substituents adjacent to one amidine. Either amidine seems to give excellent results for the single G•C bp sequence. All three compounds have K_D values in the quite strong 10 – 15 nM range for the single G•C bp sequence with no detectible binding to the AT and GC test sequences. The twist of the amidine caused by the halogen reduces binding to all sequences, but the reduction with AT and GC sequences is much larger. We have previously noted the microstructural variations in the minor groove among the three sequences and this difference obviously works to significantly enhance the selectivity of compounds with the extra twist on the amidines or size of the N-BI substituent⁹. The next compound to consider and perhaps test in cells is the imidazoline, DB2787. It also has good binding to the single G•C bp sequence, excellent selectivity over the AT sequence, which is the natural binding sequence for these type heterocycle diamidines and related compounds. Relative to the halogen derivatives, the imidazoline has slightly reduced selectivity for the two G•C bps sequence. This project has now shifted toward an emphasis on recognizing two G•C bps in the same sequences as will be necessary for strong binding with selectivity to other binding sites on the PU.1 promoter as well as recognizing promoters for other transition factors such as HOX A9⁵⁸.