Cyclic mismatch binding ligand CMBL4 binds to the 5'-T-3'/5'-GG-3' site by inducing the flipping out of thymine base.

A newly designed cyclic bis-naphthyridine carbamate dimer CMBL4: with a limited conformational flexibility was synthesized and characterized. Absorption spectra revealed that two naphthyridines in CMBL4: were stacked on each other in aqueous solutions. The most efficient binding of CMBL4: to DNA was observed for the sequence 5'-T-3'/5'-GG-3' (T/GG) with the formation of a 1:1 complex, which is one of possible structural elements involved in the higher order structures of (TGG)n repeat DNA triggering the genome microdeletion. Surface plasmon resonance assay also showed the binding of CMBL4: with TGG repeat DNA. Potassium permanganate oxidation studies of CMBL4: -bound duplex containing the T/GG site showed that the CMBL4: -binding accelerated the oxidation of thymine at that site, which suggests the flipping out of the thymine base from a π-stack. Preferential binding was observed for CMBL4: compared with its acyclic variants, which suggests the marked significance of the macrocyclic structure for the recognition of the T/GG site.

INTRODUCTION

DNA encodes the genetic information necessary for life (1). DNA structures and sequences are allied intrinsically with many vital biological processes (2,3). Errors in polymerase reactions, chemical degradations of nucleotide bases, and several physical factors (UV or ionizing radiations) are responsible for the formation of imperfect DNA structures with damaged nucleotide bases, abasic sites, bulge bases and mismatched base pairs (4–9). Furthermore, the secondary structures of trinucleotide repeat DNA causing neurological diseases involve a number of mismatched base pairs and are known to be responsible for repeat expansion (10–15). Alteration of gene expression due to changes in DNA sequences can be lethal if they produce a nonfunctional variant of an essential protein. Sequence-specific recognition of these damaged DNA structures by small molecules is a promising tool and is an active research field in molecular biology providing a basis for novel chemotherapeutic agents (16–27).

So far our group has succeeded in synthesizing a series of compounds named mismatch-binding ligands (MBLs) containing two heterocycles, typically 1,8-naphthyridine moieties linked by a linker either at (a) C-2 or (b) C-7 positions (Figure 1A). These molecules were selectively bound to the characteristic DNA sequences involving bulge or mismatched base pairs (G-G, G-A, A-A or C-C) through the formation of selective hydrogen bonding with nitrogenous bases and stacking of the naphthyridine units with the neighboring base pairs (23,28–39).

Figure 1.

(A) Connecting schemes for the MBLs with two 1,8-naphthyridine heterocycles at (a) C-2 and/or (b) C-7 position. (B) Structure of CMBL4. (C) Possible parallel-stacked arrangement of two naphthyridine in CMBL4. Arrows indicated the direction of hydrogen bonding. (D) Hydrogen bonding pattern of N-acyl-2-amino-1,8-naphthyridine unit with guanine. (E) Schematic representation of the proposed binding mode of CMBL4 with two nucleotide guanine (GG) bulges inducing the flipped out of T.

Linkers connecting two heterocycles play an important role in offering a conformation that is suitable for the simultaneous binding of two heterocycles to the target nucleotide bases (36). Due to the increased conformational freedom of two heterocycles connected by a long linker, the necessary conformational changes for the binding should be accompanied by the penalty of entropy loss. To minimize the extra energy required for the conformational change of MBLs for the complex formation with DNA, we decided to design cyclic MBLs (CMBLs), where two heterocycles are connected by linkers at both (a) C-2 and (b) C-7 positions, to provide conformational restrictions to the dynamic motion of the two heterocycles.

The class of macrocyclic compounds where a suitable bridging unit and spacer group connects nonadjacent positions of aromatic units is known as cyclophanes (40,41). Several cyclophane derivatives have been effectively utilized for host–guest complexation and undergo selective interactions with nucleotides, amino acids, nucleic acids and proteins (21,41–54), among which the water-soluble functional cyclophanes are especially interesting due to their better biomimetic scope. Cyclobisintercalands (CBIs) are the most extensively studied water-soluble macrocycles belonging to the class cyclophane; they comprise homo- or heterodimeric planar aromatic units (such as acridine, naphthalene, phenanthridine, phenazine, tetracationic porphyrin, anthracene, imidazolium, bipyridine and/or biphenyl) (42–54). Two interactive aromatic units of CBIs stack between two separate base pairs for the recognition of nucleic bases and mispaired DNA sites, such as mismatches, bulges, and an abasic site.

In the course of our studies on hydrogen-bonding-mediated sequence-specific recognition of nucleic acids by synthetic ligands and to obtain deeper insights into the interaction of the bis-naphthyridine compound with nucleic acids, we herein reported the synthesis, photophysical properties and DNA-binding activity of CMBL4. We found that CMBL4 was selectively bound to the sequence 5′-T-3′/5′-GG-3′ (T/GG) with 1:1 stoichiometry. Potassium permanganate oxidation showed that thymine (T) in the segment was flipped out from the π-stack. Since the T/GG site is an important structural element of (TGG) repeat causing microdeletion (55–58), CMBL4 could be a useful probe for the studies on (TGG) repeat and related disorders. We investigated the binding of CMBL4 with TGG repeat DNA by using a surface plasmon resonance (SPR) assay, which clearly showed the binding of CMBL4 to TGG repeat DNA.

MATERIALS AND METHODS

Synthesis and conformation analysis of CMBL4

CMBL4 was synthesized from tert-butoxycarbonyl (Boc) protected methylcarbamoylnaphthyridine dimer (MCND). Detail synthesis (Supplementary Figure S4) and characterization of CMBL4 are given in the Supplementary data. We also prepared control compound naphthyridine carbamate dimer (NCD) (30) and MCND (36) and 2-(N-methoxycarbonyl)amino-7-aminomethyl-1,8-naphthyridine (monomer) (30) using previously reported procedure to evaluate the effect on interaction with nucleic acid upon macrocyclization. Conformational analysis of CMBL4 was carried out using Amber* force field with mixed torsional/low mode sampling method (Maestro version 10.0.013, Schrödinger Inc.). Further energy optimization of the lowest energy conformation of CMBL4 obtained from Maestro was carried out using Gaussian 5.0.9. by DFT calculation (wb97xd/6-311+(G)).

Measurements of absorption and fluorescence spectra

UV and fluorescence spectra were measured in 10 mM sodium cacodylate buffer (pH 7.0) and 100 mM NaCl using BECMAN COULTER DU800 UV/Visible spectrophotometer and JASCO FP-8500 spectrofluorometer respectively. Fluorescence spectra were measured at absorption maxima.

Measurements of thermal denaturation profiles of duplexes

Thermal denaturation profiles of the duplexes 5′-d(GTC CAG X GCA ACG)-3′)/5′-d(CGT TGC YZ CTG GAC)-3′ and 5′-d(GTC CAG WX GCA ACG)-3′)/5′-d(CGT TGC YZ CTG GAC)-3′ with general formula d(5′-X-3′/5′-YZ-3′) and d(5′-WX-3′/5′-YZ-3′) respectively (5 μM each strand) were measured in sodium cacodylate buffer (10 mM, pH 7) containing sodium chloride (100 mM) using SHIMADZU UV-2700 spectrometer equipped with SHIMADZU TMSPC-8 temperature controller using a 1 cm path length cell. The absorbance of the samples was monitored at 260 nm from 2°C to 95°C with heating rate of 1°C /min. Tm values were calculated by using the median or differential method.

Circular dichroism (CD) measurements

Circular dichroism (CD) measurements of the duplex containing 5′-T-3′/5′-GG-3′ (d) were carried out on a J-725 CD spectropolarimeter (JASCO) using a 1.0 cm path length cell. CD titration spectra of DNAs (7.5 μM) were measured while titrating with ligand (0, 4, 6, 8, 10, 12, 14, 16 and 18 μM) at ambient temperature in 10 mM sodium cacodylate (pH 7.0) containing 100 mM sodium chloride.

Cold-spray ionization time-of-flight mass (CSI-TOF-MS) measurements of ligand–DNA complex

Samples were prepared by mixing d DNA duplex (20 μM) and ligand (40, 60 or 80 μM) in 50% MeOH in water containing 0.1 M ammonium acetate. Mass spectra were obtained using BRUKER micrOTOF II mass spectrometer. Spray temperature was cooled at −10°C during the injection with a sample flow rate of 0.18 ml/h. Nitrogen gas was used as a desolvation gas as well as a nebulizer.

Matrix-assisted laser desorption/ionization time-of-flight mass (MALDI-TOF-MS) measurements

MALDI-TOF-MS spectra of DNA oligomers were measured using BRUKER ultraflex III. DNA oligomers were mixed with HPA (3-Hydroxy picolinic acid) (0.5 μM; 10 mg/ml)/DAC (Diammonium citrate) (0.5 μM; 1 mg/ml) matrix, which were applied to anchor chip followed by ionization using 60–70% laser irradiation.

Surface plasmon resonance (SPR) assay

Surface plasmon resonance (SPR) single cycle kinetics assay was performed using Biacore T200 platform (GE Healthcare, Life Science). Immobilization of the oligomers on the Series S sensor chip SA surface was carried out using avidin-biotin coupling in HEPES–NaCl running buffer (HBS-N) (0.01 M, HEPES, 0.15 M NaCl, pH 7.4). 5′-Biotinylated oligonucleotides (10 μM aqueous stock) were diluted to 0.2 μM in 10 mM HEPES–500 mM NaCl, was injected to reach the response of around 500 RU. Blank immobilization was performed in the flow cell 1 to permit reference subtraction. Ligand solution (1 mM aqueous) was diluted using HBSEP+ buffer (0.01 M HEPES, 0.15 M NaCl, 3.0 mM EDTA, pH 7.4, 0.005% (v/v) Surfactant P20). Sensorgrams were obtained in the concentration range of 0.25–4.0 μM. All sensorgrams were corrected by reference subtraction of blank flow cell response and buffer injection response.

Reactions and analysis methods used for the Investigation of (T/GG) binding site

Potassium permanganate (KMnO4) was prepared as a 2 mM stock solution. KMnO4 oxidation reaction of d DNA duplexes (5 μM each) were carried out with KMnO4 (0.1 mM) in sodium cacodylate buffer (10 mM, pH 7) containing sodium chloride (100 mM) in presence and absence of ligand at 0°C for a duration of 180 min. HPLC profiles were recorded with injection of 50 μl of reaction mixture after every 30 min. Isolated oxidized product was treated with piperidine at 90°C for 30 min. Phosphorylated termini of the cleaved DNAs were removed by the alkaline phosphatase treatment at 37°C for 30 min. HPLC profiles were recorded using CHEMCOBOND 5-ODS-H column and acetonitrile/0.1M triethylammonium acetate (TEAA) as mobile phase. Adenosine was added as an internal standard (adenosine is eluted at retention time of 5.9 min).

RESULTS

Molecular design, synthesis and characterization of CMBL4

CMBL4 consisted of two 2-amino-1,8-naphthyridine heterocycles connected by two linkers at both (a) C-2 and (b) C-7 positions to provide conformational restrictions to the dynamic motion of two heterocycles (Figure 1A and B). The heterocycle had a hydrogen-bonding surface fully complementary with that of guanine (Figure 1D). Both linkers contained a secondary amino group, which functions as a positively charged site upon protonation at neutral pH to gain water solubility and ensure electrostatic attraction with the negatively charged phosphate groups of DNA. Arrangement of two heterocycles in CMBL4 could be either parallel or antiparallel in the hydrogen-bonding direction, and stacked or unstacked conformation (Figure 1C and Supplementary Figure S1). In parallel arrangement, with either stacked or unstacked conformations, hydrogen-bonding surfaces of two N-acyl-2-amino-1,8-naphthyridine units in CMBL4 point in the same direction, which provides the possibility to bind to two consecutive guanines (GG) if these Gs were involved in the structure with a dynamic motion and were therefore kinetically accessible to CMBL4 (Figure 1E). Conformational analysis of CMBL4 using an Amber* force field with mixed torsional/low mode sampling method (Maestro version 10.0.013; Schrödinger Inc.) showed that the most stable conformation and subsequent nine conformations possessed parallel-stacked structures (Supplementary Figure S2; Tables S1 and S2). Energy minimization of the lowest energy conformation of CMBL4 by DFT calculation (wb97xd/6-311+(G)) showed a parallel-stacked structure (Supplementary Figure S3), which suggests the possibility of binding to two consecutive guanines.

CMBL4 was synthesized from tert-butoxycarbonyl Boc-protected MCND (36) in the following three steps: deprotection of the carbomethoxy group, amide coupling and finally deprotection of the Boc group. Details of the synthesis (Supplementary Figure S4) and the characterization of products are given in the Supplementary Data.

Conformational studies of CMBL4 by absorption and fluorescence spectra

NCD and MCND contained two 1,8-naphthyridine units linked in an acyclic framework at either C-2 or C-7 positions, respectively, and monomer, which contained a 1,8-naphthyridine scaffold common to all three compounds (CMBL4, NCD and MCND) (Figure 2A). The absorption spectra of CMBL4 in aqueous buffer solution (pH 7) containing 100 mM sodium chloride was compared with those of parent molecules NCD, MCND, and monomer (Figure 2B). Intense absorption band at 318 nm with a shoulder peak at 332 nm was observed for CMBL4, whereas dual absorption peaks at 318 and 332 nm with almost the same intensity were observed for NCD and monomer. MCND showed similar absorption spectra with that of CMBL4, but with a different ratio of absorption at 320 and 332 nm. These differences in the absorption spectra between CMBL4 and NCD suggested that the ground electronic state of CMBL4 would be significantly perturbed from that of NCD by the macrocyclization, whereas the ground electronic state of MCND became similar to that of CMBL4, which shows that the three-atom linker connecting at the C-7 position has a larger effect on the conformation.

Figure 2.

(A) Structures of NCD, MCND, and a monomer unit. (B) Absorption and (C) fluorescence spectra excited at absorption maxima (318 nm for CMBL4 and NCD, 320 nm for MCND and 332 nm for monomer) of CMBL4 (blue), NCD (red), MCND (green) and monomer (black) (10 μM) in 10 mM sodium cacodylate buffer (pH 7.0) and 100 mM NaCl. Fluorescence of monomer is reduced by a factor 10 in the intensity for clarity. (D) Absorption spectra of CMBL4 with variable concentrations (10, 25, 50 and 100 μM). Vertical arrow indicates the increased ligand concentration. The inset shows the linear correlation of absorption at 318 nm against CMBL4 concentration with correlation factor of >0.99. (E) Absorption spectra of CMBL4 with increasing percentage of methanol (0, 10, 25, 50, 79 and 100% (dotted line)). Vertical arrow indicates the increased percentage of methanol. (F) Tm increase of the duplex d (5 μM) in the presence of CMBL4, NCD, MCND and monomer (10 μM) in 10 mM sodium cacodylate buffer (pH 7.0) and 100 mM NaCl.

Fluorescence spectra of these compounds reflected the differences in the absorption spectra (Figure 2C). While monomer emitted broad and intense fluorescence that peaked at 364 nm, CMBL4 showed emissions not only at 364 nm, but also at 438 nm. This dual emission was almost similar to that of MCND with a higher I438/I364 ratio of 0.49 compared to 0.36 for MCND. The fluorescence spectrum of NCD was observed neither at 364 nm nor 438 nm, but at 390 nm. Since the absorption spectra of NCD was similar to that of monomer, the broad and structureless emission of NCD that peaked at 390 nm is likely due to excimer fluorescence. On the other hand, the emission at 438 nm observed for MCND and CMBL4 is not due to the excimer fluorescence, but is due to the characteristic ground electronic state of the molecule, which is likely due to the intramolecular stacking of two naphthyridines in an aqueous buffer solution. The absorption at 318 nm of CMBL4 did not shift regardless to the concentration and fulfil the Lambert–Beer law (Figure 2D and inset), whereas the absorption spectra markedly changed upon increasing the solution hydrophobicity from 0% to 100% methanol. The absorption spectrum of CMBL4 in 100% methanol resembled that of NCD in 100% aqueous buffer solution with dual absorption bands at 320 and 332 nm (Figure 2E). In fact, the fluorescence spectrum of CMBL4 in methanol resembled that of NCD in aqueous buffer with a broad peak at 400 nm (Supplementary Figure S5B). In addition, 1H NMR spectra of CMBL4 showed a significant up-field shift of signals in D2O compared with those in CD3OD (Figure 3). These spectroscopic analyses indicated that naphthyridine moieties in CMBL4 are in intramolecularly stacked conformations in aqueous solutions.

Figure 3.

Expanded 1H NMR spectra (aromatic region) of CMBL4 in (A) D2O and (B) CD3OD.

Sequence dependency of CMBL4 binding evaluated by thermal melting temperature (Tm) analyses

As CMBL4 consists of two structural units of N-acyl-2-amino-1,8-naphthyridine with three hydrogen-bonding groups fully complementary to that of G (Figure 1D), we anticipated the possibility of CMBL4 binding to two consecutive Gs, if these Gs were kinetically accessible for CMBL4. We chose a single thymine opposite two consecutive Gs to form the T/GG site. The local T/GG structure should be dynamic enough for CMBL4 to creep into the site. Importantly, the T/GG site is one of the possible structural elements involved in the higher order structures of (TGG) repeat DNA. The long TGG repeat has been revealed to trigger the recurrent microdeletion from the long arm of paternal chromosome 14 (14q32.2) (55–58). The ligand binding to the (TGG) repeat may have potential to suppress the microdeletion.

The two consecutive Gs and T opposite GG were proven essential for CMBL4 binding as we describe below (see also Table 2 and Supplementary Figure S7). The T/GG site was practically prepared by hybridizing the T strand (d: 5′-d(GTC CAG T GCA ACG)-3′) and the GG strand (d: 5′-d(CGT TGC GG CTG GAC)-3′) (cf. Figure 1E). The effects of CMBL4 on the binding to the T/GG site were first investigated by measuring the melting temperature (Tm) of the d duplex (5 μM). An increase of Tm (ΔTm) by 9.0 (±0.5)°C was observed in the presence of 10 μM CMBL4 (Table 1, Figure 2F, and Supplementary Figure S6) and ΔTm increased in a concentration-dependent manner to 31.8°C with 100 μM CMBL4. NCD showed a much weaker effect on ΔTm than CMBL4 with an increase to 4.9 and 10.6°C with 10 and 100 μM, respectively. MCND and monomer showed negligible Tm increases, which further confirm the importance of the cyclic structure for the binding to the T/GG site (Table 1, Figure 2F, and Supplementary Figure S6).

Table 2.

ΔT values of the different bulge/mismatch duplexes d(5′-X-3′/5′-YZ-3′) with CMBL4a

Entry	X/YZ	Tm (–)b	Tm (+)c	ΔTmd
	X/GG
1	T/GG	43.8 (0.7)	52.8 (0.8)	9.0 (0.5)
2	C/GG	56.6 (0.3)	56.7 (0.3)	0.1 (0.1)
3	A/GG	46.5 (0.2)	48.4 (0.2)	1.9 (0.1)
4	G/GG	45.0 (0.4)	48.2 (0.2)	3.2 (0.2)
	T or C/YY
5*	T/AA	49.6 (0.2)	50.9 (0.5)	1.3 (0.4)
6*	T/TT	37.5 (0.3)	40.1 (0.1)	2.6 (0.3)
7**	T/CC	-	-	-
8*	T/II	41.1 (0.1)	43.1 (0.1)	2.0 (0.1)
9*	C/AA	43.8 (0.2)	45.6 (0.1)	1.8 (0.2)
10	C/TT	41.4 (0.1)	44.0 (0.1)	2.6 (0.3)
11	C/CC	40.0 (0.1)	43.2 (0.5)	3.2 (0.6)
	T/YG or GZ
12*	T/AG	45.3 (0.2)	46.0 (0.4)	0.7 (0.3)
13*	T/TG	40.1 (0.1)	41.0 (0.3)	0.9 (0.1)
14*	T/CG	39.0 (0.2)	39.5 (0.1)	0.5 (0.1)
15*	T/GA	47.2 (0.3)	48.3 (0.2)	1.1 (0.2)
16*	T/GT	40.0 (0.1)	42.1(0.5)	2.1 (0.5)
17*	T/GC	39.0 (0.1)	41.5 (0.5)	2.5 (0.5)

aThermal denaturation profile of the duplex 5′-d(GTC CAG X GCA ACG)-3′/5′-d(CGT TGC YZ CTG GAC)-3′ (5 μM) was measured in sodium cacodylate buffer (10 mM, pH 7) containing sodium chloride (100 mM). The temperature was increased at a rate of 1°C/min. Tm values (°C) were calculated by median method. All measurements were made three times and standard deviations are shown in parentheses.

bTm values of the oligomers.

cTm values of oligomers in the presence of ligands (10 μM).

dΔTm was calculated as the difference between Tm(+) and Tm(–).

*Tm values (°C) were calculated using differential method.

**Unable to determine the Tm value due to the biphasic nature of the Tm curve.

Table 1.

Comparison of ΔTm [°C] values of the duplex d (5′-T-3′/5′-GG-3′)a

Ligand	Tm (–)b	Tm (+)c	ΔTmd
CMBL4	43.8 (0.7)	52.8 (0.8)	9.0 (0.5)
NCD	43.8 (0.7)	48.7 (0.4)	4.9 (0.4)
MCND	43.8 (0.7)	44.3 (0.1)	0.5 (0.2)
monomer	43.8 (0.7)	44.9 (0.1)	1.1 (0.1)

aThermal denaturation profile of the duplex d (5 μM) was measured in sodium cacodylate buffer (10 mM, pH 7) containing sodium chloride (100 mM). The temperature was increased at a rate of 1°C/min. Tm values (°C) were calculated by median method. All measurements were made three times and standard deviations are shown in parentheses.

bTm values of the duplex d.

cTm values of duplex d in the presence of ligands (10 μM).

dΔTm was calculated as the difference between Tm (+) and Tm (–).

The sequence-dependent binding of CMBL4 to the T/GG site was confirmed by measuring the Tm of 17 kinds of duplexes with general formula 5′-d(GTC CAG X GCA ACG)-3′/5′-d(CGT TGC YZ CTG GAC)-3′ (X/YZ) (Table 2 and Supplementary Figures S7 and S8). Among those duplexes, the T/GG site containing duplex showed the most prominent ΔTm.

The first set of measurements (entries 1–4) was for G-bulge flanking G-X base pairs (X/GG). The highest ΔTm of 9.0°C was observed for the T/GG site containing a G-bulge flanking G-T mismatch (entry 1), whereas a negligible increase of Tm was obtained for C/GG containing a G-bulge flanking the thermodynamically stable G-C base pair (entry 2), which shows that the G-T mismatch is important for the CMBL4 binding. Modest ΔTm values for A/GG (entry 3) and G/GG (entry 4) containing G-bulge flanking G-A and G-G mismatches, respectively, further highlighted the significant role of the G-T mismatch flanking G-bulge for the binding of CMBL4. The second set of measurements is for GG control having YY (AA, TT, CC or II) opposite T (entries 5–8) and C (entries 9–11). Modest ΔTm values (1.3-3.2°C) were observed for all entries. The third set of measurements consisted of T/YG or T/GZ (entries 12–17) showed negligible or small ΔTm for all entries.

To verify the possibility of unidentified sequences for CMBL4 binding, we investigated the sequences consisting of T and G (Table 3) and found only small ΔTm values for all entries. In addition, neither single G-G mismatch (entry 9) nor fully matched (FM) sequence CC/GG (entry 1) were the sites of CMBL4 binding. The result of TT/GG was particularly significant (entry 2). Although the TT/GG site contained the T/GG site, CMBL4 binding to this site was not observed. The results shown in Table 3 suggest that the combination of two consecutive GG opposite the single T was necessary for the binding of CMBL4 and stabilization of the duplex.

Table 3.

ΔT values of the different bulge/mismatch duplexes d(5′-WX-3′/5′-YZ-3′) with CMBL4a

Entry	(WX/YZ)	Tm (–)b	Tm (+)c	ΔTmd
1	CC/GG	68.0 (0.5)	66.6 (0.1)	–1.4 (0.7)
2	TT/GG	52.5 (0.1)	53.1 (0.1)	0.6 (0.1)
3	TG/TG	48.4 (0.1)	48.2 (0.1)	–0.2 (0.4)
4	TG/G	46.0 (0.4)	46.9 (0.6)	0.9 (0.4)
5	G/TG	48.5 (0.2)	48.9 (0.1)	0.4 (0.4)
6	GT/GT	46.5 (0.2)	47.8 (0.8)	1.3 (0.7)
7	GT/G	46.4 (0.2)	47.9 (0.1)	1.5 (0.3)
8	G/GT	46.3 (0.3)	47.6 (0.3)	1.3 (0.3)
9	G/G	54.2 (0.4)	53.3 (0.1)	–0.9 (0.4)

aThermal denaturation profile of the duplex 5′-d(GTC CAG WX GCA ACG)-3′/5′-d(CGT TGC YZ CTG GAC)-3′ (5 μM) was measured in sodium cacodylate buffer (10 mM, pH 7) containing sodium chloride (100 mM). The temperature was increased at a rate of 1°C/min. Tm values (°C) were calculated by median method. All measurements were made three times and standard deviations are shown in parentheses.

bTm values of the oligomers.

cTm values of oligomers in the presence of ligands (10 μM).

dΔTm was calculated as the difference between Tm(+) and Tm(–).

Complex formation between CMBL4 and T/GG site

The formation of the CMBL4-bound d duplex was confirmed by CSI-TOF-MS. In the presence of 40 μM CMBL4 (2 molar equivalent to d duplex) (Figure 4 and Supplementary Figure S9A and B) ions corresponding to the 1:1 complex between CMBL4 and d duplex were observed as a predominant complex ([d]5– (m/z: found 1752.592; calcd. 1752.591) and [d]6– (m/z: found 1460.320; calcd. 1460.325) (Figure 4). Formation of 1:2 [d-2CMBL4]5– (m/z: found 1855.91; calcd. 1855.90) complex was detectable with increased concentrations of CMBL4 (60 and 80 μM) (Supplementary Figure 9C and D), but its intensity was much smaller than that of the 1:1 complex, which was still the dominant species. When the ligand concentration was increased, a concomitant increase of the ion peak corresponding to the unbound ligand was observed with the almost disappearance of the ion peak corresponding to d. These data clearly indicated that CMBL4 bound to d duplex preferentially with 1:1 stoichiometry.

Figure 4.

CSI-TOF-MS of duplex d (20 μM each) in 50% aqueous methanol and 100 mM ammonium acetate in the presence of CMBL4 (40 μM). The sample solution was cooled to –10°C during the injection with a flow rate of 0.18 ml/h.

In the presence of NCD, formation of a much less intense peak corresponding to the ions [d]5– (m/z: found: 1750.001; calcd: 1749.991) (Supplementary Figure S9E) indicated the reduced binding affinity of acyclic compound NCD to d duplex compared to the cyclic variant CMBL4 (Supplementary Figure S9B). Another acyclic compound, MCND, did not produce any ion peaks corresponding to the d complex formation (Supplementary Figure S9F). These CSI-TOF-MS data clearly indicate the necessity of the cyclic structure for the efficient formation of the complex with duplex d.

The CD spectra of the duplex d were measured in the presence and absence of CMBL4 to see if any structural changes could be induced upon ligand binding (Figure 5A). An induced CD band was observed at 353 nm. The molar ellipticity of the positive peak at 353 nm increased with decreasing molar ellipticity at 280 nm as the ligand concentration increased (0, 4, 6, 8, 10, 12, 14, 16 and 18 μM). The association constant (Ka) of CMBL4 for the duplex d was calculated to be 4.3 × 105 M–1 (Kd = 2.3 μM) by curve fitting to the 1:1 binding isotherm (Figure 5B).

Figure 5.

(A) CD titration experiments for the binding of CMBL4 to the duplex d (7.5 μM each) were measured in sodium cacodylate buffer (10 mM, pH 7.0) and NaCl (100 mM) at 25°C in the absence and presence of CMBL4 (0, 4, 6, 8, 10, 12, 14, 16 and 18 μM). Bold lines represent the CD curves in the absence and presence of highest concentrations of CMBL4 (0 and 18 μM, blue and red lines, respectively). (B) Induced CD signals at 353 nm were plotted against the concentration of CMBL4. The solid line represents the best fits of the data to a 1:1 binding isotherm.

Chemical probing of the binding site of CMBL4

Upon ligand binding, the T/GG site possibly induced the flipping out of the T base. We investigated the reactivity of the T base in the T/GG site of the d duplex toward potassium permanganate oxidation by reverse-phase HPLC (Figure 6, Supplementary Figure S10) (32,48,59,60). Oxidation of the C5-C6 double bond of the T base with potassium permanganate forming 5,6-dihydroxy-5,6-dihydrothymine (thymine glycol, Tg) would be more feasible for the T base in the extrahelical position compared to that in the intrahelical position of the duplex (61). Treating the d duplex (5 μM) with potassium permanganate (100 μM) in the presence of CMBL4 (10 μM) at 0°C for 180 min in buffer solution (pH 7.0) resulted in the formation of one major product at a retention time of 14.6 min with concomitant decrease of the d strand (Figure 6D). The resultant product was isolated by HPLC, analyzed by MALDI-TOF MS, and identified as the Tg-containing d (d(Tg)) (m/z: found 3994.67, calcd. 3994.62 [M+H]+) (Supplementary Figure S13A). The isolated d(Tg) was then treated with piperidine at 90°C for 30 min to give new products (Figure 6E), which suggests strand cleavage of the d(Tg) at the site of oxidation. MALDI-TOF MS analysis of resultant products indicated that these were the phosphorylated oligomers 5′-d(GTC CAG)-PO3H-3′ (GTCCAGp) (m/z: found 1873.03, calcd. 1873.20 [M+H]+) and 5′-HO3P-d(GCA ACG)-3′ (pGCAACG) (m/z: found 1882.12, calcd. 1882.21 [M+H]+) (Supplementary Figure S13B and C). Phosphorylated termini of the two identified oligomers were hydrolyzed by treatment with alkaline phosphatase to produce 5′-d(GTC CAG)-3′ and 5′-d(GCA ACG)-3′. The identities of these two oligomers were firmly confirmed by the same retention times with the authentic oligomer by co-injection in reverse-phase HPLC (Supplementary Figure S11).

Figure 6.

(A) Schematic representation of oxidation of duplex d by potassium permanganate (KMnO4) in the presence of CMBL4 followed by piperidine and alkaline phosphatase (AP) treatment. HPLC profiles for the reaction of the duplex d (5 μM) with KMnO4 (100 μM) at 0°C (B) just after addition of KMnO4, (C) after 180 min of KMnO4 addition in the absence of ligand and (D) after 180 min of KMnO4 addition in the presence of CMBL4 (10 μM). (E) HPLC profiles after piperidine heating of isolated d(Tg) at 90°C for 30 min. The peak marked with asterisk could not be identified. (F) HPLC profiles for the reaction of the fully match (FM) duplex d(5′-CC-3′/5′-GG-3′) (5 μM) with KMnO4 (100 μM) at 0°C in the presence of CMBL4 (40 μM). Adenosine was added as an internal standard (adenosine peak with retention time 5.9 min was not shown for clarity. Full range HPLC profiles were submitted in supplementary material).

Neither the potassium permanganate oxidation of FM duplex containing the 5′-CC-3′/5′-GG-3′ (CC/GG) in the presence of CMBL4 (Figure 6F, Supplementary Figure S12) nor of the d duplex in the absence of CMBL4 (Figure 6C) under the same conditions showed the formation of any prominent new products, which indicated that the oxidation of T in the T/GG site of the d duplex was induced by the CMBL4 binding to the GG site. These results supported the proposed mode of binding of CMBL4 to the T/GG site in the duplex involving the flipping out of the T from the base stacking, which makes it more susceptible towards oxidation. The potassium permanganate oxidation of the duplex d in the presence of NCD (Supplementary Figure S10b) showed a much reduced rate of formation of d(Tg). These observations were fully consistent with the lower ΔTm value (4.9°C, cf. Table 1) obtained for NCD than that obtained for CMBL4 (9.0°C).

DISCUSSION

We have designed bis-naphthyridine cyclophane CMBL4 based on the CMBL concept. The significance of two consecutive Gs opposite the single T for the CMBL4 binding was confirmed by the sequence-specific binding studies using thermal melting temperature (Tm) analysis. The Tm of duplex d (T/GG) in the presence of CMBL4 and its acyclic variants (NCD and MCND) suggested the importance of a cyclized structure. CSI-TOF-MS of the CMBL4-bound d duplex confirmed the formation of a 1:1 d complex. The formation of a 1:2 d complex was detectable with much smaller intensity compared with the 1:1 complex. Even after increasing the concentration of CMBL4, we observed a 1:1 complex as the predominant stoichiometry. In the presence of NCD, formation of a less-intense peak corresponding to the 1:1 d complex was observed. However, the ion peak corresponding to the formation of a d complex was not detected. These data clearly support the improved binding affinity of CMBL4 compared with its acyclic variants. Flipping out of the T from the π-stack at the T/GG site upon CMBL4 binding was confirmed by the increasing rate of potassium permanganate oxidation of T in the presence of CMBL4. A much reduced rate of formation of the oxidized product d(Tg) in the presence of acyclic variants (NCD and MCND) also strongly support the increasing affinity for the binding of CMBL4 compared with its acyclic variants. The dissociation constant of CMBL4 with T/GG was calculated to be 2.3 μM. The binding affinity is higher than that of previously reported G-bulge binding molecule (Kd = 29 μM) (28), but is lower than that of the G-G mismatch-binding molecule (29). The T/GG site is one of the possible structural elements involved in the higher order structures of (TGG) repeat DNA. The long TGG repeat has been revealed to trigger recurrent microdeletion in the long arm of paternal chromosome 14 (14q32.2) (49–52). The ligand binding to (TGG) repeat may have potential to suppress the microdeletion.

Several groups have been developing small molecules binding to the characteristic sequences of DNA/RNA by applying different approaches. Zimmerman et al. focused on the rational design of small molecules comprised of melamine-acridine conjugates to target CTG and CUG repeats causing myotonic dystrophy type 1 (MD1) (24). Miller et al. focused on the development of small molecules for CUG repeats by dynamic combinatorial chemistry, which is based on the techniques of molecular design and a novel combinatorial method of small-molecule evolution (25). In contrast, Disney et al. developed small molecules based on high-throughput screening from chemical library to target selective genome sequences such as CGG repeat RNA (26,27). Our group has also reported MBLs that can stabilize the secondary structures of trinucleotide repeats. Among these MBLs, NCD binds with different trinucleotide repeat sequences with decreased efficiency along with the most strongly binding CGG repeat, probably because of the conformational freedom of the longer linker length (33). With the much shorter three-atom linkers, MCND showed improved binding to the d(GAA) repeats (36).

We have evaluated the binding of CMBL4 to TGG repeat DNA and UGG repeat RNA by using SPR assay. SPR single-cycle kinetics analysis using d(TGG)8 and r(UGG)8 immobilized sensor chip showed a concentration-dependent response for CMBL4 binding (Supplementary Figure S14a) with both repeats. Although the quantitative binding analysis is difficult for the repeat sequence because of the ambiguity of the binding stoichiometry and heterogeneity of the multiple binding sites, the SPR data clearly showed the binding of CMBL4 to both repeats. The SPR response for the r(UGG)8 sequence was smaller than for the repeat DNA, which suggests that CMBL4 bound to UGG repeat RNA with lower binding affinity than TGG repeat DNA. The binding mode of CMBL4 to the RNA repeats is not necessarily the same as for the DNA repeats, which is currently unknown. These data indicated that the CMBL4 binding to T/GG site could be extended to the d(TGG) repeat DNA.

CONCLUSION

In conclusion, a new molecule CMBL4 designed based on a CMBL concept was synthesized and confirmed to bind to the T/GG site in a sequence-specific manner. The cyclophane-type structure of CMBL4 provided a characteristic electronic structure of stacked naphthyridines in aqueous solutions. The binding of CMBL4 to the T/GG site was much more efficient than that of its acyclic variants, and the binding of CMBL4 to the T/GG site induced the T to flip out from the π-stack. The 1:1 complex between CMBL4 and the T/GG site was shown by CSI-TOF-MS. These studies revealed the potential of CMBLs for the recognition of noncanonical DNA structures, which was not efficient using acyclic MBL variants. SPR single-cycle kinetics analyses suggested the binding of CMBL4 with TGG repeats, which indicated that the CMBL4 binding to T/GG site could be extended to the TGG repeat DNA. A molecular modeling simulation of CMBLs suggests that the orientation of two naphthyridines could be modulated by their linker length and structures. These advanced molecular designs may expand the possibilities of CMBLs binding to other DNA or possibly RNA structures of biological significance to explore more complex biological systems.