Synthesis and structural characterization of piperazino-modified DNA that favours hybridization towards DNA over RNA.

We report the synthesis of two C4'-modified DNA analogues and characterize their structural impact on dsDNA duplexes. The 4'-C-piperazinomethyl modification stabilizes dsDNA by up to 5°C per incorporation. Extension of the modification with a butanoyl-linked pyrene increases the dsDNA stabilization to a maximum of 9°C per incorporation. Using fluorescence, ultraviolet and nuclear magnetic resonance (NMR) spectroscopy, we show that the stabilization is achieved by pyrene intercalation in the dsDNA duplex. The pyrene moiety is not restricted to one intercalation site but rather switches between multiple sites in intermediate exchange on the NMR timescale, resulting in broad lines in NMR spectra. We identified two intercalation sites with NOE data showing that the pyrene prefers to intercalate one base pair away from the modified nucleotide with its linker curled up in the minor groove. Both modifications are tolerated in DNA:RNA hybrids but leave their melting temperatures virtually unaffected. Fluorescence data indicate that the pyrene moiety is residing outside the helix. The available data suggest that the DNA discrimination is due to (i) the positive charge of the piperazino ring having a greater impact in the narrow and deep minor groove of a B-type dsDNA duplex than in the wide and shallow minor groove of an A-type DNA:RNA hybrid and (ii) the B-type dsDNA duplex allowing the pyrene to intercalate and bury its apolar surface.

INTRODUCTION

Watson–Crick base pair recognition between nucleic acids forms the basis for a multitude of biological functions in nature. For instance, storage, transfer and expression of genetic information in living systems are governed by the ability of nucleic acids to self-assemble in a perfectly reliable manner using Watson–Crick base pairing. One goal in the field of nanobiotechnology is to design and build functional nano-scale structures. The simple and reliable Watson–Crick dsDNA framework is a superb scaffold for nano-scale assemblies as components can be self-assembled using base pairing and the DNA helix has a regular and predictable topology (1). Indeed in the field of DNA nanotechnology, impressive two-dimensional (2D) and 3D structures have been built (2–4).

To achieve the goal of functional DNA nano-structures of the size of naturally occurring enzymes, ≈10-nm dimensions, high-density and precise decoration and functionalization of the DNA scaffold is required. Thus, obtaining control at the Ångstrøm scale is essential to accomplish control of nano-scale assemblies (5,6). One way of doing so is to engineer functionalized nucleotide analogues that can be built into nano-structures organized on a DNA scaffold (7). For use in functional DNA nano-assemblies, two minimum requirements to a nucleotide analogue are that it should retain the Watson–Crick base pairing integrity and that it should be fairly straightforward to functionalize it with a variety of chemically diverse moieties.

4′-C-(N-Methylpiperazino)methyl modified DNA (Y, Figure 1) is an attractive molecule for use as a building block in the nano-assembly approach. First, it is modified in the sugar moiety and thus is likely to retain the Watson–Crick pairing ability; second, the amino functionalities may be protonated at neutral conditions; and, third, the distal nitrogen atom in the piperazino moiety is easily derivatised. 4′-C-(N-Methylpiperazino)methyl–DNA stabilizes duplex formation with target DNA strands (8) and, in addition, the introduction of a pyrene unit via a carbonyl linkage (Z, Figure 1) is tolerated and gives a distinct DNA selectivity (9). Thus, in general, 4′-C-piperazinomethyl–DNA appears as an appealing building block for functionalization of dsDNA with additional moieties facing the minor groove and it has potential use in Ångstrøm scale engineered nano-assemblies.

Figure 1.

The four nucleotides examined in this study with the nomenclature of protons used in NMR studies indicated.

In the context of therapeutic applications, 4′-C-piperazinomethyl–DNA holds some interesting potential as well. First, the piperazinomethyl modification is tolerated without loss of thermostability when hybridizing with complementary RNA (8,9). Second, the positively charged piperazino moiety shields the negatively charged sugar–phosphate backbone, thus potentially improving cell membrane penetration. Third, the piperazino moiety allows introduction of functionalities internally in oligonucleotides thus giving the possibility of altering and optimizing the pharmacokinetic and pharmacodynamic properties of anti-sense oligonucleotides.

As a proof of principle for the further functionalization of 4′-C-(N-methylpiperazino)methyl–DNA, we have introduced a pyrene moiety connected to the piperazino group with a butanoyl linker. We chose to introduce a pyrene moiety because of its fluorescence properties and to investigate the effect of having a fluorophore and a positively charged moiety juxtaposed with a flexible linker in the dsDNA duplex scaffold (10,11).

In this article, we extend our previous studies by reporting in detail the synthesis of the two derivatives, X and W, of 4′-C-(N-methylpiperazino)methylthymidine (Figure 1) and characterize their behaviour in the context of duplex formation at the atomic level. This we achieve using ultraviolet (UV) spectrophotometry, circular dichroism (CD), fluorescence, nuclear magnetic resonance (NMR) spectroscopy and molecular modelling.

MATERIALS AND METHODS

Synthesis

The synthetic routes to the monomers of X, Y, Z and W, are outlined in Schemes 1 and 2. We have reported the synthesis of phosphoramidites 4, 8 and 11a in preliminary form without experimental details and detailed characterization data (8,9) and full experimental details are now included in the Supplementary Data. Modified oligonucleotides were synthesized on 1 µmol scale by the phosphoramidite approach using standard conditions on an automated DNA synthesizer. The coupling yields were >90% for modified phosphoramidites and ≈99% for DNA phosphoramidites. The composition of all oligonucleotides was verified by matrix-assisted laser desorption/ionization-mass spectrometry (MALDI-MS) analysis and the purity by capillary gel electrophoresis. Two 12-mer oligonucleotides were synthesized for detailed NMR analysis (Scheme 3). The 4′-C-piperazinomethyl-modified strand (pip) was modified on nucleotide positions X2, X8 and X10 and the 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl-modified strand (pyrenepip) contained a single modified nucleotide at position W8. Unmodified DNA strands were purchased from DNA Technology, Århus, Denmark and RNA strands were purchased from Dharmacon RNA Technologies, Boulder, Colorado.

Scheme 1.

(i) (a) Tf2O, pyridine, (b) N-methylpiperazine THF, 40%, (ii) TBAF, THF, 76%, (iii) (iPr2N)P(Cl)O(CH2)2CN, DIPEA, CH2Cl2, 63%, (iv) (a) Tf2O, anh. pyridine, CH2Cl2, r.t., (b) piperazine, anhydrous THF, 50°C, 54%; (v) FmocCl, anh. pyridine, r.t., 98%; (vi) Et3N·3HF, pyridine hydrochloride, THF, r.t., 70%; (vii) (iPr2N)P(Cl)O(CH2)2CN, DIPEA, CH2Cl2, r.t., 64%. T = thymin-1-yl; DMT = 4,4′-dimethoxytrityl.

Scheme 2.

(i) 1-Pyrenebutyric acid, CDI, anhydrous CH2Cl2, r.t., 75%; (ii) 1-pyrenecarboxylic acid, CDI, anhydrous CH2Cl2, r.t., 74%, (iii) TBAF, THF, r.t., 82% (for 9a) and 80% (for 9b); (vi)) (iPr2N)P(Cl)O(CH2)2CN, DIPEA, CH2Cl2, r.t., 67% (for 10a) and 64% (for 10b) T = thymin-1-yl, DMT = dimethoxytrityl.

Scheme 3.

The numbering schemes used for the pip:DNA and pyrenepip:DNA duplexes. X denotes 4′-C-piperazinomethyl-T residues and W denotes a 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl-T residue.

UV, CD and fluorescence emission spectra

For thermal denaturation experiments, samples with duplex concentrations of ∼1 μM were prepared with 10 mM sodium phosphate buffer (pH 7.0), 0.1 mM EDTA and varying concentrations of sodium chloride (see Tables 1–3 for details). To assure proper hybridization between strands, the samples were annealed from 70°C. The temperature-dependent absorbance at 260 nm was measured in a 10-mm cuvette on a PerkinElmer Lambda 20 spectrophotometer equipped with a thermo regulated Peltier element (from 15 to 80°C with a temperature rise of 1°C/min). The melting temperatures (Tm values) were determined as the local maxima of the first derivative of the absorbance versus temperature curves. Furthermore, the UV absorbance trace of the pyrenepip:DNA duplex from 300 to 370 nm was measured from 10°C to 80°C with 10°C intervals.

Solutions of concentration of ∼25 μM of dsDNA, pip:DNA, pyrenepip:DNA, DNA:RNA, pip:RNA and pyrenepip:RNA duplexes were prepared with 10 mM sodium phosphate buffer (pH 7), 1 mM EDTA and 85 mM sodium chloride for CD spectra which were measured in a 1-mm cuvette on a Jasco J-600A spectropolarimeter from 200 to 360 nm at 25°C.

Fluorescence emission spectra of the single-stranded pyrenepip and the pyrenepip:DNA and pyrenepip:RNA duplexes were recorded from 350 to 600 nm at 25°C with an excitation wavelength of 350 nm in a 10-mm quartz cuvette on a Perkin–Elmer LS 55 luminescence spectrometer.

NMR spectroscopy

dsDNA, pip:DNA and pyrenepip:DNA duplex NMR samples were prepared by mixing equimolar amounts of the two single strands in 40 mM NaCl and 10 mM phosphate buffer (pH 7) to duplex concentrations between 0.5 and 2 mM in 500 µl D2O or 10:90 D2O/H2O. NMR spectra were recorded on Varian Inova spectrometers operating at either 500 or 800 MHz. TOCSY, DQF-COSY, NOESY, WATERGATE-NOESY, jr-NOESY, 1H-13C HSQC and 1H-31P COSY spectra were recorded. For the pyrenepip:DNA duplex, 2D spectra were also recorded with 10% v/v acetonitrile added to the sample. Full experimental details are included in the Supplementary Data (Tables S3 and S8). The 1H, 13C and 31P resonances were assigned using standard methods (Tables S1, S2, S6 and S7). All spectra were processed with NMRPipe (12), 1D spectra were analysed using NMRDraw and 2D spectra were analysed using Sparky (13).

Chemical shift changes induced by pH in the pip:DNA duplex were measured in 20 selective 200 ms NOESY spectra recorded in the pH interval 3.5–11. The pH value in the sample was adjusted using NaOD and DCl solutions, measured (before and after NMR experiments) using a Hamilton spintrode and corrected for the deuterium effect from pH meter reading (pH = pD – 0.4) (14). The pKa values of the amino groups in the piperazinomethyl modifications were determined by fitting the piperazinomethyl protons’ chemical shift changes as a function of the pH value to the Henderson–Hasselbalch equation (15). Separate fittings were performed for the low and high pH regions, respectively.

Molecular dynamics simulations

The two modified thymidine nucleotides were built using the xleap module of AMBER, the piperazino modified nucleotide with the distal nitrogen protonated and the pyrenepiperazino-modified nucleotide with the proximate nitrogen protonated. The atomic charges of the two modified nucleotides where calculated using the RESP methodology (16).

The NMR-guided model of the pip:DNA duplex

A dsDNA helix was build with a standard B-type helical structure in the nucgen module of AMBER. The helix was initially relaxed using simulated annealing and energy minimization with X-PLOR 3.851 (17), applying the force field of Cornell et al. (18) with 1H chemical shift, sugar pucker, backbone, glycosidic angle and minor groove width restraints. The modifications with their proper charges were then added to the dsDNA helix in the xleap module of AMBER to form the starting model of the pip:DNA helix. During subsequent refinement, only the piperazinomethyl modifications were allowed to move freely while applying broad distance restraints with a force constant of 50 kcal mol−1 Å−2 between the modification and the duplex. The distance restraints were derived from NOE volumes in a 100-ms NOESY spectrum (Table S5). Two pseudo atoms were introduced in the calculation for the Hb and Hc protons in the piperazine ring, respectively. Ten structures with initial random atom velocities where subjected to a simulated annealing protocol with the temperature at 600 K for 4 ps and then cooled to 300 K over 24 ps with a time step of 1 fs. In the final step, structures were energy minimized.

Molecular dynamics simulation of the pyrenepip:DNA duplex

Two different models of the pyrenepip:DNA were built with the pyrene moiety intercalated in the G6:C19 – A7:T18 and C5:G20 – G6:C19 base pair steps, respectively. These initial models were subjected to free molecular dynamics (MD) simulations using the generalised Born solvent model, with a salt concentration of 0.2 M, as implemented in the AMBER package (19,20). Initially, the two models were equilibrated for 100 ps by raising the temperature from 0 to 300 K with weak positional restraints on duplex atoms. The free MD at 300 K was carried on for 2.0 ns with a temperature coupling of 2.0 ps and no cut-off for non-bonded interactions.

RESULTS AND DISCUSSION

The free hydroxyl group in the known nucleoside 1 (21) (Scheme 1) was converted into the very labile triflate moiety by treatment with trifluoromethanesulfonic anhydride (Tf2O) in pyridine (8,9). The crude triflate was treated with N-methylpiperazine to introduce the piperazino moiety. This reaction proceeded very slowly and resulted only in a moderate 40% yield of nucleoside 2. The tert-butyldimethylsilyl (TBDMS) group of nucleoside 2 was subsequently removed by standard conditions using tetrabutylammonium fluoride (TBAF) in THF yielding nucleoside 3 in 76% yield. Nucleoside 3 was converted into phosphoramidite 4 in a yield of 63% using 2-cyanoethyl N,N-diisopropylphosphoramidochloridite. The choice of introducing N-methylpiperazine instead of unsubstituted piperazine was made since we assumed that a side reaction leading to dimers was a potential risk if unsubstituted piperazine was used. However, the very slow substitution reaction indicated that this side reaction would be insignificant if a large excess of piperazine was used. The introduction of unsubstituted piperazine on nucleoside 1 was therefore tried using the same conditions as with N-methylpiperazine and resulted in nucleoside 5 in 54% yield. The secondary amino group was subsequently protected with an Fmoc group by treatment with FmocCl in anhydrous pyridine. The Fmoc protecting group is compatible with the use of this monomer for the so-called post-oligonucleotide conjugation, i.e. derivatization of the distal amino group of the piperazine moiety subsequent to completion of oligonucleotide (ON) synthesis. The TBDMS group was removed by treatment with triethylammonium trishydrofluoride and pyridinium chloride in THF yielding nucleoside 7 in 70%. Nucleoside 7 was converted into phosphoramidite 8 in a yield of 64% by the same procedure as previously described. The phosphoramidites 4 and 8 were used for incorporation of monomer X and Y, respectively, into ONs using automated DNA synthesis on polystyrene-based solid supports.

For the pyrene modified building blocks, two different linkers between the pyrene moieties and the piperazino unit were chosen, a butanoyl and a carbonyl linker. Introduction of the pyren-1-ylcarbonyl and pyren-1-ylbutanoyl moieties was done on the solid support subsequent to completion of oligonucleotide synthesis by a modified version of the ‘on resin conjugation’ method published by Mokhir et. al. [(22); Figure 2]. Phosporamidite 4 (Scheme 1) was thus incorporated into ONs that were transformed into ON17-ON20. This was accomplished by selective removal of the Fmoc protecting group using 20% piperidine in DMF, followed by coupling with the relevant carboxylic acid in the presence of HBTU and DIPEA. The ONs were then cleaved off the solid support and deprotected using saturated aqueous ammonia, and purified by reversed-phase high-performance liquid chromatography (HPLC). Finally, the O5′-DMT group was cleaved off and the product ON was precipitated from ethanol using standard methods (Figure 2).

Figure 2.

The post-ON synthesis approach.

Unfortunately, the ‘on resin conjugation’ approach described above was not applicable for incorporation of more than two pyrene modifications. Therefore, the phopshoramidites of the modified monomers (11a and 11b, Scheme 2) were synthesized. Nucleoside 5 was treated with 1-pyrenecarboxylic acid or 1-pyrenebutyric acid in the presence of 1,1′-carbonyldiimidazole (CDI) to afford nucleosides 9a (74%) and 9b (75%), respectively. These amides were desilylated using TBAF in THF yielding nucleosides 10a (82%) and 10b (80%), respectively, which were phosphitylated to furnish the desired phopshoramidites 11a (67%) and 11b (64%), respectively. Phosphoramidites 11a and 11b were used to synthesize ON21, ON22, ON24, ON25 and ON27-31.

Thermal denaturation using UV spectrophotometry

In general, the incorporation of X and Y is tolerated well in duplexes (Table 1). Towards complementary DNA, an increase in melting temperatures between ∼1 and 4°C per modification is observed whilst towards complementary RNA, little effect is seen (ΔTm ≈ −0.5–1.0°C). Upon lowering of the salt concentration, the stabilizing effect of X and Y towards DNA hybridization was equal or slightly increased. This is consistent with a positive charge placed at the piperazino moiety, which would lower the electrostatic repulsion between the DNA strands a priori.

Table 1.

Melting temperatures for - and Y-modified duplexes

		DNA	RNA	DNA	RNA
		40 mM Na+		110 mM Na+
ON1	5′-d(GTG-ATA-TGC)	23	21	29	28
ON2	5′-d(GTG-AYA-TGC)b	26 (+3.0)	21 (0.0)	33 (+4.0)	28 (0.0)
ON3	5′-d(GTG-AXA-TGC)c	24 (+1.0)	20 (−1.0)	31 (+2.0)	28 (0.0)
ON4	5′-d(GCA-XAX-CAC)c	29 (+3.0)	26 (+2.5)	34 (+2.5)	26 (−1.0)
ON5	5′-d(GYG-AYA-YGC)b	31 (+2.7)	22 (+0.3)	35 (+2.0)	29 (+0.3)
ON6	5′-d(GXG-AXA-XGC)c	31 (+2.7)	20 (−0.3)	34 (+1.7)	26 (−0.6)
		10 mM Na+		110 mM Na+
ON7	5′-d(GTG-TTT-TGC)	16	17	32	32
ON8	5′-d(GTG-YTY-TGC)	22 (+3.0)	17 (0.0)	35 (+1.5)	30 (−1.0)
ON9	5′-d(GTG-XTX-TGC)	22 (+3.0)	17 (0.0)	36 (+2.0)	29 (1.5)
ON10	5′-d(CTA-TCT-GTC-GTT-CTC-TGT)	38	46	57	59
ON11	5′-d(CTA-XCT-GTC-GTT-CTC-TGT)	40 (+2.0)	45 (−1.0)	58 (+1.0)	59 (0.0)
ON12	5′-d(CTA-XCX-GTC-GTT-CTC-TGT)	42 (+2.0)	45 (−0.5)	58 (+1.0)	59 (0.0)
ON13	5′-d(CTA-TCT-GTC-GTT-CTC-XGT)	40 (+2.0)	44 (−2.0)	58 (+1.0)	59 (0.0)
ON14	5′-d(CTA-TCT-GTC-GTT-CXC-XGT)	42 (+2.0)	44 (−1.0)	59 (+1.0)	58 (−0.5)
ON15	5′-d(CTA-XCX-GTC-GTT-CXC-XGT)	44 (+1.5)	44 (−0.5)	60 (+0.7)	57 (−0.5)

aShown are melting temperatures (Tm values) towards DNA and RNA complements and in brackets the changes in melting temperatures (ΔTm) per modification calculated relative to the unmodified duplexes;X denotes 4′-C-piperazinomethyl-T residues and Y denotes 4′-C-(N-methylpiperazino)methyl-T residues.

bData taken from ref. (8).

cData taken from ref. (9).

When the pyrene moiety is introduced (Z and W), the thermostability towards DNA increases and concomitantly, the stability towards RNA decreases (Table 2). Hence, a substantial preference for DNA over RNA is observed (6–16°C). In general, there appears to be little difference between the two different building blocks except in the case of ON17 which displays an exceptional low melting temperature towards its RNA complement. Multiple insertions of Z and W lead to a decrease in the raise of melting temperature. Notably, the melting temperature actually drops upon the incorporation of the third Z modification in a 9-mer duplex (compare ON21 and ON19 in Table 2).

Table 2.

Melting temperatures for Z- and W-modified duplexes

		110 mM Na+
		DNA	RNA
ON16	5′-d(GTG-ATA-TGC)	29	28
ON17	5′-d(GTG-AZA-TGC)b	36 (+7.0)	20 (−8.0)
ON18	5′-d(GTG-AWA-TGC)	37 (+8.0)	28 (0.0)
ON19	5′-d(GCA-ZAZ-CAC)b	37 (+4.0)	23 (−2.5)
ON20	5′-d(GCA-WAW-CAC)	39 (+5.0)	20 (−4.0)
ON21	5′-d(GZG-AZA-ZGC)b	31 (+0.7)	<10
ON22	5′-d(GWG-AWA-WGC)	40 (+3.7)	<10
ON23	5′-d(GTG-TTT-TGC)	32	32
ON24	5′-d(GTG-TZT-TGC)b	41 (+9.0)	28 (−4.0)
ON25	5′-d(GTG-TWT-TGC)	41 (+9.0)	30 (−2.0)
ON26	5′-d(AAA-TGA-TGG-CTG-C)	48	44
ON27	5′-d(AAA-TGA-ZGG-CTG-C)	56 (+8.0)	45 (+1.0)
ON28	5′-d(AAA-TGA-WGG-CTG-C)	55 (+7.0)	43 (−1.0)
ON29	5′-d(AAA-WGA-TGG-CWG-C)	56 (+4.0)	36 (−4.0)
ON30	5′-d(AAA-ZGA-ZGG-CZG-C)	59 (+3.7)	31 (−4.3)
ON31	5′-d(AAA-WGA-WGG-CWG-C)	57 (+3.0)	29 (−5.0)

aShown are melting temperatures (Tm values) towards DNA and RNA complements and in brackets the changes in melting temperatures (ΔTm) per modification calculated relative to the unmodified duplexes; Z denotes 4′-C-(N-pyren-1-ylcarbonyl)piperazinomethyl-T residues and W denotes 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl-T residues.

bData taken from ref. (9).

The 3′-flanking nucleobase may play a role in the stabilization by pyrene moieties as observed for 2′-O-pyrene modified ONs (23). We tested this sequence dependence with ON24 and ON25. As observed in Table 2 (ON24 and ON25 versus ON17 and ON18), it is of little importance whether a purine or a pyrimidine is 3′-juxtaposed to the pyrene-carrying nucleotide. This indicates that the stabilizing behaviour by Z and W is not related to the type of adjacent nucleobase.

NMR studies of 4′-C-piperazinomethyl-modified dsDNA

pip:DNA duplex structure

Chemical shifts are excellent markers of the local structure in molecules and in a dsDNA duplex they depend strongly on its precise helical structure. We observe a very good agreement between the 1H chemical shift values calculated for a standard B-type helix and both the dsDNA and pipDNA duplexes at 25°C (Table S1) (24). Especially, we would like to emphasize the good agreement of the H2′ chemical shifts, as in a B-type helix, the H2′ protons are positioned close to the nucleobase in the 3′ direction but exactly in a node where they do not sense the ring current effect. A small change in this position would therefore lead to a change of H2′ chemical shift (24). In addition, all other NMR data support the B-type duplex geometry of the pip:DNA duplex, i.e. characteristic inter- and intra-strand NOE contacts (Figure S1), S-type puckered deoxyriboses as shown by H1′↔H2′ and H1′↔ H2′′ COSY cross peaks (25) (Figure S2) and 31P chemical shifts within the narrow 0.7 p.p.m. interval that is characteristic for a standard B-type helix (26). Thus, it is apparent that the three 4′-C-piperazinomethyl modifications are accommodated without altering the structure of the helix itself from that a B-type duplex.

The 4′-C-piperazinomethyl modifications

The melting temperatures for the dsDNA and pip:DNA duplexes show that at low salt concentration the introduction of three 4′-C-piperazinomethyl modifications results in an increase in Tm of 7°C (Table 3). When raising the salt concentration, the increase in Tm is reduced to 5°C. The salt-dependent effect indicates that the piperazinomethyl groups are positively charged at neutral pH and able to reduce the electrostatic repulsion between the two DNA strands.

Table 3.

Melting temperatures for duplexes studied by NMR spectroscopy

	15 mM Na+		50 mM Na+		100 mM Na+
	DNA	RNA	DNA	RNA	DNA	RNA
DNA	26	24	38	37	43	40
pip	33 (+2.3)	25 (+0.3)	43 (+1.7)	36 (−0.3)	48 (+1.7)	40 (0.0)
pyrenepip					52 (+9.0)	40 (0.0)

aShown are melting temperatures (Tm values) towards DNA and RNA complements and in brackets the changes in melting temperatures (ΔTm) per modification calculated relative to the unmodified duplexes.

To determine which nitrogen atoms of the piperazinomethyl modifications are protonated at physiological pH and their pKa values, we followed pH-dependent 1H chemical shifts changes of the protons within the modifications (Figure 3). From the extent of chemical shift changes observed in different positions, it appears that the distal nitrogen atom is preferably protonated at neutral pH, whereas the proximal one is not. The chemical shifts show two distinct jumps and correspondingly, high and low pKa values were determined for each of the three modifications (Tables 4 and S4). As compared to the free N-methylpiperazine molecule, the first protonation is facilitated in the pip:DNA duplex, while surprisingly the second protonation is not. The interaction of the modifications with the duplex and the different degree of interaction with the solvent as compared to a free N-methylpiperazine molecule is the origin of the difference.

Figure 3.

pH titration curves for the X10 piperazinomethyl moiety in the pip:DNA duplex, Ha (red), Hb (green) and Hc (blue). The curves fitted to the data are shown in grey and the titration curve for N-methylpiperazine is also shown in grey.

Table 4.

pKa values for the three 4′-C-piperazinomethyl-T (X) residues in the pip:DNA duplex determined by Hill plot analyses of proton chemical shift variations

Residue	pKa
X2	3.6 ± 0.1
	Nd
X8	4.0 ± 0.1
	10.7 ± 0.1
X10	4.0 ± 0.1
	10.7 ± 0.1
N-methylpiperazine	4.7 ± 0.0
	9.0 ± 0.0

For each of the three modifications, NOE contacts from the protons in the piperazine ring to the H1′ of its own and the previous nucleotides’ sugar ring is observed. X8 and X10 both have an adenosine neighbour in the 5′-direction (A7 and A9, respectively) and NOEs from the protons in the piperazine rings to the H2 protons of these adenosines are also detected (Figure S3). In addition, two cross-strand NOEs are observed: X8Hc ↔ C19H1′ and X10Hc ↔ A17H2. To summarize, all DNA protons from which NOEs to the piperazino modifications are observed are positioned in the minor groove and in the 5′-direction of the C4′ substitution sites.

The pip:DNA model

We created a model of the pip:DNA duplex starting from a canonical B-type dsDNA duplex using distances derived from NOEs between DNA and piperazinomethyl protons to derive the positions of the modifications. This model is sufficiently accurate in terms of elucidating the interactions between the modifications and the remainder of the pip:DNA duplex.

In a B-type helical structure, the C4′-H4′ bond is positioned at the edge of the minor groove oriented in an almost perpendicular angle towards the backbone of the complementary strand. The centrelines through the two nitrogen atoms in the piperazine rings of X8 and X10 have the same perpendicular direction, but are simultaneously projected towards the bottom of the minor groove, reaching it between the first and second base pair in the 5′-direction of the modified strand. Thus, the piperazinomethyl modifications are buried in the minor groove and they have a suitable size to span and efficiently fill out the bottom part of the groove and hence make numerous van der Waals contacts (Figure 4).

Figure 4.

A representative model structure of the pip:DNA duplex viewed into the minor groove. Nucleobase atoms are coloured light blue, sugar–phosphate backbone atoms dark blue and atoms of the piperazinomethyl groups red.

In contrast to X8 and X10, the piperazinomethyl modification of X2 points into the solvent in the 5′-direction, where it is allowed to span a larger conformational space and makes contacts with the deoxyribose sugar of C1. Interaction with the minor groove in the 5′-direction, as observed for X8 and X10, is not possible as there is only one base pair in its 5′-direction. This suggests that a modification in the first or second position from the 5′-end of a DNA strand would stabilize the duplex less than modifications at other positions, which indeed is the trend observed for other 4′-C modifications (27). The difference in pKa values between X2 and X8 and X10 (Table 4) confirms that the piperazinomethyl moiety of X2 is located in a different environment.

Although using fairly wide NOE distance restraints in the calculation of the pip:DNA model, all models converged into identical piperazinomethyl positions in the minor groove. This rigidity of the piperazinomethyl moieties is most likely not a picture of the true dynamics but rather a representation of the best fit of the average position of the modifications biased by the force field used. A very rigid conformation would have given rise to two, four and four different chemical shift values for the protons in positions a, b and c, respectively, in the piperazinomethyl groups. However, only one, two and two different chemical shifts are observed which suggests that the piperazine ring is undergoing fast ring flip and rotation around the NN-axis in the minor groove.

The piperazinomethyl modification induces a higher increment in the melting temperature per modification (∼2–3°C at low salt concentrations) as compared to unbranched 4′-C-aminoalkyl modifications (∼0.5°C) (27). The pKa values we have determined show that this effect is not due to double protonation of the piperazine ring at neutral pH. The distal nitrogen atom of the piperazine ring is protonated at neutral pH but a salt bond to the negatively charged phosphate group in the backbone appears unlikely judging from the model structure and the absence of any conspicuous 31P chemical shift differences. A salt bond to the modified nucleotide’s own phosphate group (in the 5′-direction) is structurally possible but again the absence of large 31P chemical shift differences do not indicate this, and furthermore the inter- and cross-strand NOEs to the bottom of the minor groove would not agree with this scenario.

The relatively high increase in the melting temperature at even higher salt concentrations also suggest that the stabilization observed for the 4′-C-piperazinomethyl modification is not due to a better effective neutralisation of the negatively charged phosphate backbones. Instead, the piperazinomethyl modification’s higher degree of rigidity, its larger size and its perfect filling of the bottom of the minor groove, hence improving its ability to form van der Waals and electrostatic interactions to both strands of the duplex, seems to favour the hybridization towards DNA as compared to the smaller and presumably much more dynamic 4′-C-aminoalkyl modifications. The connection between the larger degree of interaction with the minor groove and the increase in stability can explain why a slightly higher stabilization of dsDNA is achieved with the moderately larger 4′-C-(N-methylpiperazino)methyl modification (8).

NMR studies of 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl-modified dsDNA

pyrenepip:DNA duplex structure

Molecules that bind nucleic acid duplexes can be classified as either intercalators or groove binders depending on their mode of binding. Although pyrene is an aromatic and planar molecule, the usual characteristics of an intercalator, pyrenes chemically tethered to DNA duplexes are both found intercalating (23,28,29) and binding in the grooves (30,31) depending on the type of helix, the attachment site, the linker to the duplex and the sequence of the oligonucleotide (10,11).

Conformational dynamics that on the NMR scale is too fast to give rise to peaks from each single conformation and yet not fast enough to show only the average line, occupies an intermediate dynamical regime leading to broad resonance lines. This is the case for the protons in the central part of the pyrenepip:DNA duplex at and near the binding site of the pyrenepiperazino modification. The severe line broadening hampered analysis of spectra at 25°C and, consequently, we increased the temperature to 47°C where the dynamics are faster but the formation of the duplex remains intact. Yet, many lines are still severely broadened at this temperature. For some NMR experiments, we added 10% v/v acetonitrile, commonly used for solvation of aromatic compounds like pyrene, to the NMR sample to increase the correlation time. This resulted in somewhat sharper lines for some resonances.

The 1H chemical shift differences detected at 47°C for the non-exchangeable protons of the pyrenepip:DNA duplex deviates at most 0.4 p.p.m. from the dsDNA duplex with the largest changes (>0.1 p.p.m.) observed for the four central nucleotides, C5 to W8, in the modified strand and for T16 to C19 in the unmodified strand (Figure S5). Except for T16, these nucleotides are all positioned in the 5′-direction of the pyrenepiperazino modification. The protons for which considerable chemical shift changes are observed are spread between both the minor and major grooves in a B-type helix. The imino and amino protons belonging to nucleobases of the four central base pairs in the pyrenepip:DNA duplex were not observed in the NMR spectra due to either conformational line broadening or fast exchange. A single resonance, which was impossible to assign, was observed at 10.9 p.p.m. This is 1.7 p.p.m. upfield of the nearest imino proton signal from the dsDNA duplex. Intercalating agents generally cause an upfield shift of the adjacent imino protons owing to the ring current effect from the intercalator (32) and the unassigned resonance at 10.9 p.p.m. could be an imino proton shifted by the ring current of an intercalating pyrene moiety. Indeed, a shift of 1.6 p.p.m. was observed for imino protons neighbouring the intercalators in an intercalating nucleic acid modified duplex (29). At the two ends of the duplex, the inter-strand amino to imino proton NOEs and their similar chemical shift values as compared to the dsDNA duplex showed the formation of a normal Watson–Crick base-pairing pattern.

To make room for an intercalator in a given base pair step, the duplex has to unwind to increase the distance between the stacked base pairs. Such an extension will lead to a weakening or even break of the intra-nucleotide NOE connectivities normally observed. In the 300-ms NOESY spectrum, it was possible to follow the inter-nucleotide H6/H8 ↔ H1′, H6/H8 ↔ H2′ and H6/H8 ↔ H2′′ connectivities in the modified DNA strand from C1 to C5 and again from A7 to C12 (Figure 5). Except for a very weak G6H8↔ C5H2′′ cross-peak, the NOE connectivity is thus broken in the C5-G6 and G6-A7 steps. However, the intra-nucleotide NOEs observed for these nucleotides are also very weak and the G6H8 ↔ G6H1′, A7H8 ↔ H1′ and A7H8 ↔ A7H2′′ cross-peaks are not observed at all. In the unmodified DNA strand, the NOE connectivity can be followed all the way through the strand, albeit the inter-nucleotide NOEs between T18 and C19 are weakened, and between C19 and G20, only the G20H8 ↔ C19H2′′ cross-peak is observed. Thus, the NOE connectivities appear to be broken or weakened in the C5:G20 to G6:C19 and G6:C19 to A7:T18 base pair steps. Nevertheless, it is not possible to tell with complete certainty if the NOE contacts are missing because of an unwinding and lengthening of the duplex caused by intercalation or because of the line broadening observed for this part of the duplex.

Figure 5.

The aromatic–H1′ region of the 300-ms NOESY spectrum of the pyrenepip:DNA duplex. The aromatic–H1′ NOE connectivity pathways are shown for the modified strand (red) and for the DNA strand (blue). The aromatic resonance positions are shown. Missing or very weak sequential NOE connectivities are marked with asterisks.

For nucleic acids, the chemical shifts of the 31P atoms in the sugar–phosphate backbone depends on the two backbone torsion angles α (O3′-P-O5′-C5′) and ζ (C3′-O3′-P-O5′) (26). For a regular B-type helix, the 31P chemical shifts are found within a narrow interval, since the α and ζ torsion angles are in gauche(−), gauche(−) conformations. Unwinding of the duplex in connection with intercalation changes these angles to a gauche(−), trans conformation in the intercalation site to lengthen the backbone which gives a ∼1.0–1.5 p.p.m. downfield shift for the 31P chemical shift (26). The 31P chemical shifts span a rather large interval of 1.7 p.p.m. in the pyrenepip:DNA duplex. Noteworthy chemical shift differences between the pyrenepip:DNA and dsDNA duplexes are observed for the six phosphorus atoms joining the four central base pairs steps show a modest downfield shift of 0.3 and 0.4 p.p.m., respectively, G6P and T18P in the G6:C19 to A7:T18 step display a higher but asymmetric downfield shift of 1.3 and 0.7 p.p.m., respectively, while A7P and A17P in the A7:T18 to W8:A17 base pair step differ with an upfield shift of 0.2 p.p.m. and a down field shift of 0.3 p.p.m., respectively.

The nearly identical CD spectra of the pyrenepip:DNA and dsDNA duplexes in the 220–310-nm range suggest that the pyrenepip modification leaves the bulk part of the overall helical geometry unperturbed (Figure 6). Both CD spectra show the characteristic features of a B-type helical environment with positive and negative bands of almost equal intensity at ∼275 nm and ∼250 nm, respectively, and the zero crossings near 260 nm (33).

Figure 6.

(A) Temperature-dependent UV absorbance curves for the pyrenepip:DNA duplex. (B) Fluorescence emission spectra of the single-stranded pyrenepip strand and the pyrenepip:DNA and pyrenepip:RNA duplexes. (C) CD spectra of the pyrenepip:DNA and DNA:DNA duplexes.

The 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl modification

The incorporation of a single 4′-C-(N-pyren-1-ylbutanoyl)piperazinomethyl modification into a DNA duplex gives an increase in melting temperature of 9°C . This is comparable to other pyrene-modified dsDNA duplexes where the pyrene is intercalating (9,23,34). As the pyrenepip:DNA duplex melts into two single strands, the pyrene UV absorbance band in the 300–360-nm region shifts to a shorter wavelength (Figure 6). Stacking with nucleobases shifts the absorption band of pyrene towards longer wavelengths (31,35), thus, our UV melting profile suggests intercalation of the pyrene moiety.

The fluorescence emission spectrum of a molecule depends on the fluorophore itself and its environment. For pyrene, the fluorescence emission decreases upon intercalation because of the stacking interactions with the nucleobases (35). When the pyrenepip–DNA strand is hybridized with the complementary DNA strand, we indeed observe a reduced fluorescence emission (Figure 6), which is an indication of at least partial stacking of the pyrene moiety.

As for the DNA protons in the central part of duplex, the NMR resonances emanating from the pyrenepiperazino modification are very broad and consequently not all of them could be assigned unambiguously. When compared to free pyrene, the 1H chemical shifts from the pyrene moiety in the modification are subject to a substantial upfield shift (0.7–1.4 p.p.m.), which is characteristic for an intercalating pyrene unit as it experiences the ring current effect of the adjacent nucleobases. The large line widths made analysis of the NOESY spectra extremely difficult as only a few NOE contacts between the modification and the DNA were detectable and then again only by inspection of individual 1D slices of the 2D spectra. For the pyrene moiety NOEs were observed for H6/7/8 ↔ G6H1′, H6/7/8 ↔ G6H2′/H2′′, H2 ↔ C19 H1′ and H3 ↔ C19H6. The latter NOE contact is most interesting as H6 of C19 is located in the major groove of the DNA duplex. Cross-peaks from the pyrene moiety to protons in both the minor (H1′, H2′ and H2′′) and major (H6) grooves is very strong evidence for intercalation of the pyrene where it penetrates through the DNA duplex. From the linker part of the modification, we exclusively observe NOEs to protons in the minor groove: Hb/c ↔ W8H1′ and Hb/c ↔ A7H1′, Hd ↔ T18H1′ and Hf ↔ C19H1′. These NOEs unambiguously positions the piperazino ring and the butanoyl linker in the minor groove pointing in its 5′-directions, as also observed for the piperazinomethyl-modified DNA duplex.

The pyrenepip:DNA model

Taken together, the data presented above do not give a clear-cut picture of the structural implications of the pyrenepiperazino modification. However, the amount of circumstantial evidence pointing towards an intercalative mode of pyrene binding is substantial, albeit the intercalation site in the duplex cannot be assigned to a single base pair step. There are breaks/weakenings of NOE connectivities in both the C5:G20 to G6:C19 and G6:C19 to A7:T18 steps and upfield 31P chemical shifts are observed in both these base pair steps. The broad resonances observed for the site near the modification show that intermediate conformational exchange is taking place. This exchange might potentially be a shift between the pyrene intercalating in the two base pair steps mentioned above, which are the second and third base pair step in the 5′-direction of the modified nucleotide. Rapid on/off rates between multiple intercalation sites and therefore the observation of only the time-averaged properties have previously been shown to result in only a weakening and not a complete break of the NOE connectivities (36).

To examine if both intercalation sites are acceptable in a dsDNA duplex, we constructed two model structures with the pyrene moiety intercalated in either of the two possible base pair steps. Both models conserved the overall B-type helical structure, as experimentally indicated by the lack of chemical shift changes for the non-central base pairs and the CD curves, and the pyrene moiety is nicely accommodated in both base pair steps (Figures 7 and 8). The larger degree of 31P chemical shift changes in the G6:C19 to A7:T18 site indicates that this is the preferred site for intercalation. This fact is supported by the superior agreement with experimental NOE contacts for this model structure (Figure 8). To reach the intercalation site, the linker part of the modification makes contacts with the bottom of the minor groove in both models. Depending on which intercalation site to reach, the linker is either extended (C5:G20 to G6:C19) or more compact (G6:C19 to A7:T18), thus the linker is both adequately long and flexible to promote intercalation in both possible sites. Even though the possible surface contact between the intercalator and the nucleobases in a given base pair step influences the sequence specificity of an intercalator, it is usually the non-intercalating groups attached to it that determines where it intercalates through interactions in either the major or minor groove. Inspection of the two model structures reveals that the linker does not appear to make hydrogen bonds or other strong interactions to the minor groove. Thus, there is no apparent reason for the discrimination of the two intercalation sites and it is somehow a combination of the stacking interaction of the pyrene and the van der Waals interactions of the linker in the minor groove that governs the intercalation preference. Only fairly weak interactions are present in both model structures which supports that multiple intercalation sites are populated and in exchange with each other.

Figure 7.

Model structures of the pyrenepip:DNA duplex viewed into the minor groove. (A) The structure with the pyrene moiety intercalated in the C5:G20 to G6:C19 base pair step and (B) in the G6:C19 to A7:T18 step. The colouring scheme is as in Figure 5.

Figure 8.

Close-ups of the intercalation sites with interatomic distances indicated between protons in the modification and DNA for which NOEs are observed. (A) The pyrene is intercalating in the C5:G20 to G6:C19 base pair step and (B) in the G6:C19 to A7:T18 step.

Although not tested with a model, there are also some weak indications of intercalation in the A7:T18 to W8:A17 base pair step. Even though the intensity of the NOE connectivities in this step are not significantly affected, the 31P chemical shift and the chemical shift difference pattern in the unmodified strands suggest the possibility of intercalation in this site. Such an arrangement also seems possible for the linker in the pyrenepip:DNA duplex.

Structural basis for DNA/RNA discrimination

Both of the modifications discussed in this paper possess the ability to effectively discriminate between binding to complementary DNA and RNA. As gleaned from Tables 1–3, both modifications yield substantial increases in thermostability when hybridized with complementary DNA, which is in contrast to the virtually unaltered melting temperatures when hybridized with complementary RNA. Modifications on C4′ favour the C2′-endo sugar conformation which potentially prearranges the nucleotide better for a B-type dsDNA duplex than for an A-like DNA:RNA hybrid (37). Two further factors are important. First, in a B-type dsDNA duplex with sugars adopting C2′-endo puckers, the C4′–H4′ bonds are pointing towards the opposite minor groove wall, whereas in a DNA:RNA hybrid with O4′-endo deoxyribose sugar puckers, this bond is pointing from the edge of the groove into the solvent. Second, in a B-type dsDNA duplex, the minor groove is narrow (∼5.2 Å) and deep, while in a DNA:RNA hybrid adopting an intermediate, A-like geometry it is wider (∼8–9 Å) and more shallow. Both of these factors fit nicely with the DNA/RNA discriminatory properties of the piperazinomethyl-modified DNA. The wider minor groove in a DNA:RNA hybrid leads to less favourable electrostatic interactions between the charged piperazinomethyl group and the groove and also reduces van der Waals interactions. In addition, the piperazinomethyl moiety is turned away from the groove and into the solvent, thus decreasing interaction between the duplex and the piperazinomethyl moiety dramatically.

In general, nucleic acid analogues with positively charged C4′-modifications tend to increase the binding stability towards DNA, the optimum result being ∼1.5°C per modification for an aminoethyl modification, and slightly decrease the binding stability towards RNA (27,38). Modelling of 4′-C-aminoalkyl modifications show that the modifications are most likely to interact with their own DNA strand, possibly by interaction with the phosphate group located in the 3′-direction (27). In common for the charged C4′- modifications, they consist of a flexible linker capped with a positively charged moiety. 4′-C-piperazinomethyl–DNA differs as the positively charged moiety is a bulky and rigid moiety that spans the minor groove and interacts with both strands of the DNA duplex rather than interacting solely with its own strand.

All data on the pyrenepip:DNA duplex, including the fluorescence spectrum where the emission is lower in a duplex context than for the free single-strand, are consistent with intercalation of the pyrene moiety. With a shorter carbonoyl linker instead of the butanoyl linker, a similar increase in the thermostability (7°C versus 9°C) was observed for hybridization with complementary DNA and the fluorescence emission spectra similarly indicated intercalation of the pyrene (9). By contrast, the length of the linker has a very different effect on the hybridization with RNA. This can be explained from the fluorescence spectra which indicate that the carbonoyl- and butanoyl-linked pyrenes interact in diverse manners with DNA:RNA hybrids. With the shorter linker, the pyrene moiety is completely excluded from the duplex in a DNA:RNA hybrid (9). In this case, the thermostability is decreased by 8°C for one modification, probably because of the lack of hydrophobic interactions between the pyrene unit and the interior of the duplex. The change in thermostability is likely to reflect the penalty paid for placing the large aromatic pyrene unit in the shallow minor groove. On the other hand, the fluorescence emission spectra of the present modification with the longer butanoyl linker show that the pyrene unit is interacting with the base pairs in a DNA:RNA hybrid, which leads to a net null effect on the thermostability of the modified DNA:RNA hybrid. The difference in stability between the pyrenepip:DNA and pyrenepip:RNA duplexes probably stems from both the poorer accommodation of the piperazino group and the linker in the hybrid minor groove as discussed above and less efficient intercalation of the pyrene chromophore in the A-like geometry of the pyrenepip:RNA hybrid. Intercalating chromophores are less well accommodated in an A-like helix structure, as the neighbouring nucleotides require a larger change in their glycosidic angles to fit the intercalating geometry, and consequently, a considerable distortion is introduced in the adjacent base pairs (39).

CONCLUSION

We have presented the synthesis of two 4′-C-modified DNA analogues, both carrying a positively charged piperazino moiety located in the minor groove of duplexes and one of them functionalised with a butanoyl-linked pyrene. Both modifications adhere to Watson–Crick base pairing and bind more strongly to complementary DNA than RNA. The reason for this discrimination of strand type is that the cationic and bulky piperazino moiety with limited flexibility fits snugly in the deep, narrow and electrostatic predominantly negative minor groove in B-type duplexes. In addition, the pyrene intercalates in a B-type dsDNA duplex and thereby adds further stability to the duplex. In conclusion, we have shown that 4′-C-piperazino modifications are a promising scaffold for dsDNA functionalities embedded in the minor groove and that such modifications efficiently discriminates between DNA and RNA binding.

SUPPLEMENTARY DATA

Supplementary data are available at NAR online.

FUNDING

The Danish National Research Foundation. Funding for open access charges: Nucleic Acid Center, Department of Physics and Chemistry, University of Southern Denmark; Danish Centre for Scientific Computing.

Conflict of interest statement. None declared.