Combined nucleobase and backbone modifications enhance DNA duplex stability and preserve biocompatibility

DNA strands containing a triazole linkage flanked on its 3′-side by an aminoethylphenoxazine nucleobase analogue (G-clamp) have been prepared by solid-phase synthesis followed by CuAAC-mediated click oligonucleotide ligation. The stability of the doubly modified DNA duplexes and DNA–RNA hybrids is greatly increased, whereas a single base pair mismatch located at or adjacent to the modifications is strongly destabilising, making triazole G-clamp a potent mismatch/point mutation sensor. A DNA strand containing this unnatural combination was successfully amplified by PCR to produce unmodified copies of the original template, with deoxyguanosine inserted opposite to the G-clamp-triazole nucleotide analogue. This study shows for the first time that a polymerase enzyme can read through a combined backbone/nucleobase modification surprisingly well. These favourable properties suggest new applications for oligonucleotides containing the G-clamp triazole modification in biotechnology, nanotechnology, diagnostics and therapeutics.


Introduction
Templated chemical ligation of alkyne and azide-functionalised oligonucleotides using the CuAAC reaction 1,2 has recently been used to assemble long DNA strands up to 300 bases in length containing a triazole mimic of a DNA phosphodiester linkage. 3he base sequence of DNA strands containing this articial linkage can be copied by DNA polymerases during PCR with high delity, 3 and can also be transcribed by RNA polymerase (Fig. 1). 4 Remarkably, plasmids containing essential 3 and non-essential 5 triazole-containing genes are functional in E. coli despite the apparent dissimilarity of the articial linkage to a normal phosphodiester DNA bridge.Biocompatibility is observed even when triazole linkers are placed just 4 base-pairs apart in opposite strands of DNA. 5 The structural and thermodynamic properties of the biocompatible triazole linkage have been determined by NMR, ultraviolet melting analysis and circular dichroism, revealing a normal B-DNA duplex with reduced resistance to thermal denaturation. 6This relative instability did not detrimentally affect the biocompatibility of the triazole linkage, but increasing the thermodynamic stability of the modied DNA is likely to improve discrimination against mismatched base pairs in more complex biological systems, leading to increased delity and sequence discrimination.The lowering of melting temperature caused by the triazole DNA backbone has also been observed in RNA. 7This suggests that its incorporation into antisense oligonucleotides, 8 miRNA probes, 9 and siRNA constructs 10,11 is likely to lead to reduced potency.This could limit its use in biomedical and diagnostic applications where strategically placed triazole linkages otherwise confer favourable properties on therapeutic a School of Chemistry, University of Southampton, Higheld, Southampton, SO17 1BJ, UK.E-mail: tb2@soton.ac.uk;A.H.El-Sagheer@soton.ac.uk;Fax: +44 (0)2380 592991; Tel: +44 (0)2380 592974 oligonucleotides, protecting them against degradation in vivo. 12ther triazole backbones have been synthesised, [12][13][14] but none give rise to duplexes of the same stability as canonical DNA.
In order to address this we have been investigating approaches to stabilise triazole-containing oligonucleotides, bearing in mind that any modication must also permit readthrough by DNA polymerases.Many chemical adaptations have been developed to increase the thermal stability of nucleic acid duplexes, 15 but with biocompatibility as the major criterion we focused on those that would be expected to produce minimal structural perturbation.
We now report the synthesis and properties of a triazolecontaining DNA analogue that forms more stable duplexes than natural DNA.This is achieved by the incorporation of the aminoethylphenoxazine analogue of 2 0 -deoxycytidine (Gclamp) adjacent to the triazole linkage.This cytosine analogue increases duplex stability by a combination of increased intrahelical base stacking and additional hydrogen bonding to guanine (Fig. 1). 16It has been shown to be effective in a DNA 17,18 and PNA 19 context, and analogues have also been developed to recognise 8-oxoguanine in DNA. 20G-clamp is essentially deoxycytidine with an additional aminoethoxyphenoxy group protruding into the major groove.We reasoned that this structural change might be accommodated by polymerase enzymes without causing mutagenesis, and we have investigated this concept.

Results and discussion
In theory increased duplex stability might be achieved by inserting the G-clamp in the DNA strand at either the 5 0 -or 3 0 -side of the triazole linkage.However, placing the modication at the 5 0 -side would require the synthesis of a novel monomer for use in oligonucleotide synthesis.This could take the form of a resinbound nucleoside consisting of a G-clamp nucleobase, deoxyribose sugar and 3 0 -propargyl group with an additional cleavable attachment point to the solid support; or more likely a 3 0 -propargyl G-clamp nucleoside-5 0 -phosphoramidite monomer for reverse direction oligonucleotide synthesis. 21Both of these are major undertakings, and it is evidently more straightforward to introduce the G-clamp at the 3 0 -side of the triazole linkage.It was apparent to us that this might be achieved using G-clamp phosphoramidite monomer 18 combined with a variation of the methodology we have previously developed to synthesise triazolelinked DNA. 3 This requires two oligonucleotides; one with a 5 0 -azide and the other with a 3 0 -propargyl group, followed by click ligation to join the two oligonucleotides together.
The methodology that is generally used to synthesise 5 0 -azide oligonucleotides 22 involves conversion of the 5 0 -OH group of the terminal deoxyribose sugar to azide whilst the oligonucleotide is attached to the solid support.This transformation (Fig. 2), which is accomplished using methyltriphenoxyphosphonium iodide, provides a general method to substitute the nucleobase on the 3 0 -side of the triazole linkage with unnatural analogues, in this case G-clamp.It has been shown that G-clamp is strongly stabilising at C p C steps in normal DNA, 16 so we decided initially to synthesise and evaluate an oligonucleotide containing the equivalent triazole containing dinucleotide analogue (5-MeC t C C where C C ¼ G-clamp and t ¼ triazole).The 5-methylated analogue of cytosine was used for synthetic convenience. 3 13-mer triazole G-clamp oligonucleotide (ODN-3, Table 1) was synthesised in order to provide a sensitive probe for duplex stability.To accomplish this, the G-clamp phosphoramidite monomer was incorporated at the 5 0 -end of oligonucleotide ODN-1 (Table 1) by standard solid-phase methods and the 5 0 -OH group was converted to azide on the solid support as described above (Fig. 2).This functional group conversion has not been previously carried out on 5 0 -G-clamp oligonucleotides and we were pleased to observe that it proceeded smoothly (Fig. 3A and B).Aer cleavage from the solid support and nucleobase/phosphate deprotection, ODN-1 was obtained in high purity.Ligation of ODN-1 to an excess of 3 0 -propargyl-functionalised ODN-2 6 by the CuAAC reaction 1,2 yielded the triazole-containing 13-mer ODN-3.The reaction proceeded efficiently without the need for a complementary splint (Fig. 3C and D) and excess of 3 0 -propargyl ODN-2 was removed by HPLC purication.Ultraviolet melting studies were performed with ODN-3 hybridised to a series of matched and mismatched complementary DNA and RNA strands (Fig. 4, Table 2).The stability of the duplex formed by ODN-3 and its unmodied DNA complement (ODN-7) was compared to the equivalent unmodied canonical duplex with a central 5-methy cytosine (ODN-4/ODN-7) and to the duplex containing a triazole linkage and a cytosine base in place of G-clamp (ODN-6/ODN-7).ODN-6 was made by clicking ODN-2 with ODN-5 using the method described above.
Incorporation of the G-clamp increased the UV melting temperature (T m ) of the triazole DNA duplex by an impressive 6.5 C compared to the unmodied duplex, and by 12 C compared to the triazole duplex with cytosine in place of the Gclamp (Table 2a).Even greater stabilisation was observed when the triazole G-clamp ODN-3 was hybridised to an RNA complement.In this case the increase in T m was 9.4 C compared to the normal DNA-RNA hybrid, and 12.7 C relative to the triazole duplex with cytosine instead of G-clamp (Table 2b).To dispel concerns that the stabilisation due to the triazole G-clamp is unique to a Me C t C C base stacking step, studies were also carried out to determine the melting temperatures of duplexes containing a T t C C step.The same trends in duplex stability were observed, i.e. triazole G-clamp > canonical DNA > triazole DNA (ESI †).For chemically modied DNA to be useful in a biological context it is essential that it is selective for its chosen target; i.e. duplexes containing mismatched base pairs must be destabilised.To evaluate this, duplexes containing a single methylated or unmethylated C.A or C.C mismatch on either side of the triazole linkage were studied.In the DNA-DNA series the average mismatch destabilisation was 19.9 C for the triazole G-clamp duplex, 18.4 C for the normal non-triazole duplex and 14.8 C for the triazole duplex without G-clamp (Table 2c-e), indicating that triazole G-clamp is the most efficient sensor of DNA mispairing.In the context of DNA-RNA hybrid duplexes the benecial effects of triazole G-clamp were even greater, with an average mismatch destabilisation of 21.9 C compared to 16.7 C for normal DNA and 16.8 C for triazole DNA without the G-clamp (Table 2f-h).Overall these melting studies conrm that the stability of DNA can be greatly improved by incorporation of G-clamp on the 3 0 -side of the articial triazole backbone.Moreover, potent mismatch discrimination can be achieved if the G-clamp nucleobase is directly involved in mispairing, and also when the base directly on the other side of the triazole linkage is mispaired.There is a marked improvement in mismatch discrimination for the combination of triazole and G-clamp compared to normal DNA and DNA containing triazole alone.
Having established that G-clamp is effective in stabilising triazole DNA duplexes, it was important to determine whether DNA polymerases can read through the combination of the triazole and G-clamp to faithfully produce complementary copies.The outcome of these investigations was uncertain as there are no previous reports of replication of DNA strands that contain G-clamp monomers, let alone examples of G-clamp combined with backbone modications.To investigate the polymerase-compatibility of the duplex-stabilising modication we synthesised an 81-mer PCR template (ODN-20) containing triazole G-clamp using a complementary splint (ODN-17).The splint was employed to assist CuAAC-catalysed oligonucleotide   ligation between 3 0 -propargyl ODN-18 and 5 0 -azide ODN-19.The click reaction was efficient and the template was puried and analysed by HPLC and mass spectrometry (Fig. 5A and B).We then carried out experiments to nd out if the doubly-modied linkage can be read through in a linear fashion.This was successful (ESI †).PCR amplication was carried out (Fig. 5C) and the amplicon was puried by agarose gel electrophoresis, inserted into a sequencing vector and analysed by Sanger DNA sequencing (Fig. 5D).The sequencing data from over 20 clones conrmed that the region around the triazole linkage had been read through correctly (ESI †).Overall this series of experiments prove that the G-clamp triazole linkage can be formed efficiently by click ligation, and can be read through faithfully by DNA polymerase enzymes (Fig. 6).This is an important result as it suggests that the triazole G-clamp combination could be used for in vivo applications that involve replication of the modied DNA. 3,5n the context of future therapeutic and diagnostic applications it is important to understand duplex forming properties of oligonucleotides containing multiple modications of triazole G-clamp.It was unclear whether this would confer additional duplex stability or lead to a collapse of the duplex structure due to steric constraints.To this end we used double templated click ligation to synthesise a 13-mer oligonucleotide containing two units of triazole G-clamp and compared the stability of its fully complementary DNA duplex to equivalent unmodied and triazole modied duplexes (ESI †).Very large increases in T m (12.2 C and 22.7 C respectively) were observed conrming the potential of triazole G-clamp as a modication in antisense oligonucleotides.With this result in mind future synthetic strategies will focus on producing the triazole G-clamp dinucleotide phosphoramidite for direct incorporation into oligonucleotides during solid-phase synthesis.

Conclusions
A simple strategy for the synthesis of oligonucleotides containing a G-clamp triazole linkage has been established.The G-clamp is introduced by standard solid-phase oligonucleotide synthesis and the triazole linkage is inserted in a CuAAC reaction between 3 0 -propargyl and 5 0 -azido G-clamp oligonucleotides. 1,2The click reaction works in templated and non-templated modes, although templation is recommended for the synthesis of long oligonucleotides to guarantee an efficient reaction.All the required reagents are available commercially, making this methodology widely accessible.G-clamp triazole-modied DNA has been shown to form duplexes which are more stable than the equivalent unmodied canonical DNA.Interestingly the presence of a single G-clamp nucleobase is sufficient to increase the duplex stability of triazole DNA signicantly beyond that of unmodied DNA, and two additions lead to very large stabilisation.This is likely to be a consequence of increased base stacking and H-bonding, preventing fraying of the base pair adjacent to the triazole linkage. 6Importantly, duplexes containing the triazole G-clamp are greatly destabilised by the presence of mismatched base pairs close to the modication, demonstrating that triazole G-clamp is a powerful discriminatory sensor of Watson-Crick base pairing.The increased duplex stability and mismatch destabilisation suggests that the G-clamp triazole modication could be benecial in antisense, siRNA and miRNA applications.All-triazole DNA backbone analogues [23][24][25][26][27] are being developed for these applications using the click concept, 28 but there are issues related to poor aqueous solubility. 29One solution to this problem is to develop chimeric oligonucleotides containing a combination of normal and triazole backbones.In this context the results presented here, demonstrating the high stability of the G-clamp triazole combination, are very encouraging.This principle of nucleobase compensation to achieve high duplex stability could be applied to other triazole DNA backbone analogues, [12][13][14] but whether or not these would be compatible with accurate read-thorough by DNA polymerases remains to be established.When present in a DNA template, the G-clamp triazole linkage described here can be read through and accurately amplied by PCR, with the G-clamp nucleobase recognised as cytosine.Surprisingly the polymerase enzyme readily accommodates the combined backbone and nucleobase modications.This study opens the way to the wider use of oligonucleotides containing triazole backbones in biotechnology, nanotechnology and various diagnostic and therapeutic applications.

Experimental
Synthesis of 5 0 -azide G-clamp and dC oligonucleotides G-clamp phosphoramidite monomer was obtained from Glen Research.Oligonucleotides were assembled on the 1.0 mmole scale (trityl-off) with 5 0 -dC or 5 0 -G-clamp as described in the ESI.† The protected oligonucleotide attached to the synthesis column was then treated with a 0.5 M solution of methyltriphenoxyphosphonium iodide in DMF (1.0 mL) which was periodically passed through the column via two 1 mL syringes over 15 min at room temperature.The column was then washed several times with dry DMF.To convert the 5 0 -iodo (dC or G-clamp) to 5 0 -azido (dC or G-clamp), sodium azide (15 mg) was suspended in dry DMF (1 mL), heated for 10 min at 70 C then cooled down.The supernatant was taken up into a 1 mL syringe, passed back and forth through the column then le at room temperature overnight (for short oligonucleotides) or for 5 h at 55 C (for long oligonucleotides).The column was then washed with DMF and acetonitrile and dried by passing a stream of argon gas through it.The resultant 5 0 -azide oligonucleotide was cleaved from the solid support and deprotected by exposure to concentrated aqueous ammonia solution for 60 min at room temperature followed by heating in a sealed tube for 5 h at 55 C. Oligonucleotides were puried by HPLC as described in the ESI.†

Synthesis of the G-clamp and normal triazole DNA templates
A solution of Cu I click catalyst was prepared from tris-hydroxypropyltriazole ligand 30 (14.0 mmol in 0.2 M NaCl, 90.0 mL), sodium ascorbate (20.0 mmol in 0.2 M NaCl, 40.0 mL) and CuSO 4 $5H 2 O (2.0 mmol in 0.2 M NaCl, 20.0 mL).This solution was added to the two reacting oligonucleotides (5 0 -azide and 3 0alkyne) and the complementary splint oligonucleotide, (100.0 nmol of each) in 0.2 M NaCl (100 mL).The mixture was kept at room temperature for 2 h.Reagents were then removed by NAP-10 gel-ltration and the ligated triazole DNA product was puried by HPLC as described in the ESI.† For the synthesis of the short (13-mer) triazole containing oligonucleotides, the overall conditions were the same except that a 1.5-fold excess of the alkyne oligonucleotide was used in a non-templated click ligation reaction (i.e.no splint oligonucleotide was added).
Ultraviolet DNA melting studies UV DNA melting was monitored on Cary 4000 Scan UV-Visible Spectrophotometer (Varian) at 3 mM concentration of each oligonucleotide in 10 mM phosphate buffer, 200 mM NaCl at pH 7.0.Spectra were recorded at 260 nm.The samples were initially denatured by heating to 85 C (or 90 C) at 10 C min À1 then cooled to 20 C at 1 C min À1 and heated to 85 C (or 90 C) at 1 C min À1 .Eight successive melting curves were measured and T m values were calculated from their derivatives using Cary Win UV Thermal application Soware.

PCR amplication of the G-clamp triazole template
GoTaq DNA polymerase was used to generate a PCR product from the 81-mer template (ODN-20) which includes a G-clamp triazole linkage.Reagents and conditions: 4 mL of 5Â buffer (Promega green PCR buffer){ was used in a total reaction volume of 20 mL with 5 ng of the DNA template, 0.5 mM of each primer, 0.2 mM dNTP and 0.5 unit of GoTaq polymerase.The reaction mixture was loaded onto a 2% agarose gel in 1Â TBE buffer.PCR cycling conditions: 95 C (initial denaturation) for 2 min then 25 cycles of 95 C (denaturation) for 15 s, 54 C (annealing) for 20 s and 72 C (extension) for 30 s.This was followed by leaving the PCR reaction mixture at 72 C for 5 min.The PCR amplicon was extracted from the gel using a QIAquick Gel Extraction kit.It was then inserted into a vector for automated Sanger DNA sequencing (ESI †).

Fig. 1
Fig. 1 Triazole G-clamp base paired with guanine in complementary DNA.The additional steric bulk of the G-clamp nucleobase relative to cytosine is in red.

Fig. 2
Fig. 2 Conversion of the 5 0 -OH group of a G-clamp oligonucleotide to azide on the solid support.Structure of protected and unprotected G-clamp in inset.

Fig. 5 GFig. 6
Fig. 5 G-clamp triazole DNA template is efficiently chemically ligated and accurately amplified by PCR.(A) Reversed-phase HPLC, (B) ESImass spectrum of 81-mer ODN-20, and (C) 2% agarose gel for the PCR product using ODN-20 as a template.Lane 1; 50bp DNA ladder.Lane 2; control reaction with primers and without the template.Lane 3; PCR reaction using template ODN-20 (5 ng) and primers ODN-21 and ODN-22 (Table 1).(D) Sanger sequencing of cloned PCR template showing the correct sequence (dGCCA) around the site of the original triazole G-clamp.Details of DNA sequencing in ESI.†

Table 1
Oligonucleotides used in this study a a Me

Table 2
Ultraviolet duplex melting studies a a DT m ¼ (T m duplex À T m control duplex).t ¼ central triazole linkage in ODN in column 1 of the table.p ¼ central phosphodiester linkage in ODN in column 1 of the table.C C ¼ G-clamp, Me C ¼ 5-methylcytosine, r ¼ RNA.* Central base pairs ¼ base pairs around triazole site proceeding from