Non-Catalyzed Click Reactions of ADIBO Derivatives with 5-Methyluridine Azides and Conformational Study of the Resulting Triazoles

Copper-free click reactions between a dibenzoazocine derivative and azides derived from 5-methyluridine were investigated. The non-catalyzed reaction yielded both regioisomers in an approximately equivalent ratio. The NMR spectra of each regioisomer revealed conformational isomery. The ratio of isomers was dependent on the type of regioisomer and the type of solvent. The synthesis of various analogs, a detailed NMR study and computational modeling provided evidence that the isomery was dependent on the interaction of the azocine and pyrimidine parts.


Introduction
Copper-free click reactions based on the strain-promoted alkyne-azide cycloaddition reaction (SPAAC) were discovered by Wittig and Krebs in 1961 [1]. During examination of the properties of cyclooctyne, they observed its rapid reaction with phenyl azide to yield a single triazole product [1,2]. In 2004, Bertozzi and co-workers first used the SPAAC with biotinylated cyclooctyne as a bioorthogonal reaction to modify biomolecules and living cells [3]. Various derivatives of cyclooctyne have been developed to improve the kinetics of cycloaddition (Fig 1) [4,5].
Incorporation of an electron-withdrawing fluorine in cyclooctyne leads to a significant increase in the reaction rate [4,5]. The use of dibenzocyclooctyne results in further acceleration due to the additional ring strain caused by the phenyl rings [4,5]. The introduction of nitrogen into cyclooctyne further improves the reaction rate [4,5] and facilitates binding of the necessary appendix for labeling or reaction with other substrates. In 2010, an aza-dibenzocyclooctyne motif (Fig 1), a combination of DIBO and DIMAC, was developed and used for the PEGylation of enzymes [6]. The use of ADIBO derivatives for a wide range of biological applications is exemplified by the work of Kjems and co-workers, who used an ADIBO moiety to ligate DNA to macromolecules to produce DNA conjugates with polymers, proteins and other large biomolecules [7] Pfeifer and co-workers then prepared ADIBO-activated glass slides for the immobilization of diagnostic peptides [8]. ADIBO derivatives have also been used to label membrane bilayers [9], 5´-capped RNA [10], antibodies [11] and proteins [12] The modification of nanoparticles with ADIBO for biological purposes is another research area [13][14][15]. ADIBO derivatives often serve as F-18 probes [16][17][18][19][20] or 64 Cu radiolabeled probes [21][22][23] for PET imaging. Surprisingly, although ADIBO derivatives are widely used for copper-free click reactions, the structures of the final triazoles have been fully described for only three simple compounds in a single article [24] In that study, the reaction was performed between polymethoxy azocine and a few simple azides: 5-azidopentanoic acid, benzyl azide and 4-azidophenyl isothiocyanate. The reaction proceeded with slight regioselectivity, and no unexpected behavior was observed in the NMR spectra.
In our study, we investigated copper-free click reactions of ADIBO derivative 7 with azides derived from 5-methyluridine prepared using our newly developed procedures. The use of simple nucleosides led to the formation of triazoles, and the structures of the triazoles were characterized. Full characterization of the final products is crucial for describing bioorthogonal reactions. Copper-free click reactions with nucleobases using dibenzoazocine derivatives have been described in only two articles [25,26]. In the first, the structures of the final triazoles synthesized on azides derived from purine-based acyclovir and ganciclovir were not determined by standard analytical technique [26]. Wnuk and co-workers then described the reaction of 5-azidouridines and 8-azidopurines with dibenzoazocine to afford "a mixture of several inseparable regioisomers" identified by HPLC, although the reaction could afford in principle only two regioisomers [25].
The structure of triazoles derived from ADIBO and oligonucleotides or nucleosides has not been studied to date, although the vicinity of the bulky dibenzoazocine group to the nucleic base can play a significant role in the conformation of the DNA duplex, RNA strain assembly or protein tertiary structure when used for biomolecule labeling.
Here, we report the first results of a conformational study of triazoles formed directly on nucleosides at the 5´position. This triazole formation could be used for the non-catalyzed 5´end chemical labeling of oligonucleotides to bind DNA/RNA probes to other molecules or surfaces to enable target delivery or immobilization. The results were also verified for a derivative of 5-azidomethylene uridine to reveal potential difficulties with the labeling of oligonucleotides via base derivatization. To avoid negative or false-positive effects of the immobilized/labeled nucleic acid in a biological assay, an effect of biomolecule modifications to its structure should be elucidated.

Preparation of 5´-azides
Our synthesis of 5´-azidoderivative 5 was based on a two-step approach starting from 5-methyluridine (Scheme 1). The total yield of this reaction reached approximately 60%, which is more efficient than the previously described synthesis [27] beginning with protection of 5-methyluridine by acetone. The newly developed synthetic protocol was also successfully tested in the synthesis of protected 5´-azidoderivative 6 starting from protected 5-methyluridine 2, prepared in a very good yield using the described procedure (Fig 2) [28].

Copper-free click reactions
First, we studied the copper-free click reactions of azides 5 and 6 by treatment of dibenzoazocine derivative 7 (Fig 3), which was prepared as described previously [29]. This compound was highly reactive, and all reactions in methanol were nearly instantaneous. Both products were produced as a mixture of two regioisomers in a 1:1 ratio. The conversion of this reaction depended on the amount of azocine 7. Complete conversion of the azides required at least 1.4 equivalents of 7.
Isomers 8a,b and 9a,b were successfully separated by semi-preparative HPLC, yielding products in >99% purity. All isomers were subjected to detailed NMR study. at 150.75 ppm (Fig 6) likely hindered the rotation of the pyrimidine ring relative to the other parts of the molecule. Identical results were obtained in d 6 -DMSO.
The 1 H spectra of compounds 9a and 9b in d 6 -DMSO were measured at various temperatures (25°C, 50°C, 100°C and cooling back) to determine whether the number of isomers was affected by temperature. The spectral pattern was essentially unaffected until +50°C. Signal coalescence was finally observed at +100°C (Fig 7).
Identical results were obtained for standard 19 F spectra of 9a and 9b measured in DMSO, with signal coalescence at 100°C (Fig 8).
To characterize the relationship between the number of isomers and the type of solvent, we extended the number of tested solvents to include acetone, D 2 O, MeOD, d 6 -DMSO and d 7 -DMFA for derivatives 8a and 9a, which were selected as representative model compounds. Changing the solvent not only shifted the signals but also affected their ratio (Fig 9).
The number of isomers of compounds 8a and 9a remained constant; only the ratio was affected. The dependence of the isomeric ratio on solvent is presented in Table 1.
To determine whether the presence of isomers was caused by the nucleoside part of molecule or by s-cis, s-trans isomery of the amide groups on the azocine moiety, we prepared triazoles 10-12 substituted on nitrogen by only hydrogen, the sterically bulkier coumarin, and by 2´,3´,5´tribenzoyl-5-methyluridine, which mimics the bulky surrounding in a nucleotide ( Fig 10).
For triazoles 10a and 11a, we observed only one set of signals, with no isomery. Surprisingly, the 1 H spectra of triazole 11b contained at least three sets of signals (Fig 11), similar to the 13 C NMR spectra in which more than one set of signals was detected. These results confirm that the type of substituent on the triazole strongly influences the number of isomers in NMR spectra. Moreover, the presence of isomery in 11b and the lack of isomery in derivative 11a indicate that the position of the aliphatic part of azocine relative to the triazole substituent is crucial for the number of isomers formed. These results also demonstrate that the presence of amide bonds in the azocine part of the molecule does not affect the isomery observed in the NMR spectra.
Regioisomers 12a and 12b were inseparable under several HPLC conditions; the retention times of the isomers were nearly identical on semi-preparative C 18 columns. The 1 H NMR spectrum of the mixture of 12a,b in CDCl 3 revealed the presence of additional isomers, similar to compounds 8a,b and 9a,b. Thus, all possible isomers can also be expected when labeling oligonucleotides via nucleobase derivatization.
To clearly identify the origin of the isomery, triazoles 8a,b, 9a,b and 11a,b were subjected to computational study.

Computational study
According to the NMR study described above, we assumed that the observed conformations of compounds 8a,b and 9a,b were the result of a combination of rotation about two bonds between the triazole and ribose rings. The rotations of these bonds were studied as the changes of two dihedral angles involving backbone atoms N 15 N 14 C 13 C 12 and N 14 C 13 C 12 C 11 (for numbering see Fig 12).
Although the conformational changes are dependent on the solvent, we simplified the quantum calculation to a vacuum to assess the ability of the compounds to form stable conformers whose distributions further depend on the solvents. The theoretical model used in this study was the B3LYP method with a 6-31G(d,p) basis set in Gaussian 09 [32]. The optimized geometry was determined for all individual structures. Using the optimized structures, the potential energy surface (PES) was scanned with 10-degree increments of rotation, up to a total of 360 degrees for every dihedral angle.  To determine the configurations with energies at the local minima, the dependencies of the energies on both dihedral angles were examined. The conformation with the lowest energy was selected as the zero point on the energy scale for each structure. The potential local energy minima were subsequently determined, and the fractional populations from the Maxwell-Boltzmann distribution were estimated according to the following standard equation at 300 K: where N i is the number of molecules in the configuration with the energy E i of a total number N of all molecules at temperature T and κ is the Boltzmann constant. The sum in the denominator is over all our configurations with particular energies E j . The resulting total populations (the sum of the populations of the conformers with energy lower than or equal to the corresponding energy) in the energy are presented in Fig 13. These dependencies revealed that the most frequent conformers have potential energies lower than 30 kJ/mol.
Compound 8a formed 20 local minima with energy lower than 30 kJ/mol, and one conformer significantly predominated with a population close to 14% (see Table 2). Compound 8b formed 16 local minima with energy < 30 kJ/mol, and 12 and 15 local minima satisfied these criteria for derivatives 9a and 9b, respectively. The local minima representing conformers with populations greater than 0.5% are summarized in Table 2 (a list of all minima with energy lower than 30 kJ/mol is presented in the S1-S4 Tables).
Analyses of compounds 8a,b and 9a,b with respect to changes in both dihedral angles revealed that the position of the aliphatic chain in structures (a) (intended 8a/9a) and (b) (intended 8b/9b) differed. For the (a) structures, both the left and right positions of the aliphatic chain (with respect to the triazole ring-see Fig 14) were observed (Fig 14A and 14B), but the positions on the right site were characteristic for only three local minima, 18, 19 and 20, with higher energies (26.03-26.79 kJ/mol) and a low conformers population (below 0.01%) for derivative 8a; two local minima-8 (19.94 kJ/mol) and 11 (23.59 kJ/mol)-with populations less than 0.01% were observed for derivative 9a. However, the (b) structures were characterized by the right position only (Fig 14C).
In addition, the interactions between the 5-methyluridine and the aliphatic chain as well as the distinct intermolecular hydrogen bonds were observed. The most frequent conformer at  the first local minima of derivative 8a (see Table 2) maintained the aliphatic chain in direct interaction with the pyrimidine ring. This interaction was enabled by a hydrogen bond between the carbonyl of trifluoroacetyl group 37 and imide hydrogen 3 of the pyrimidine ring (see . In the conformer representing the second most populated local minimum, the trifluoroacetyl carbonyl group of the aliphatic chain interacted with the ribose OH hydrogen (see . Similar to conformation 2, the other local minima of derivative 8a summarized in Table 2 exhibited interactions of the trifluoroacetyl group with the ribose OH hydrogens (see S79 Fig).
Local minimum 1, in which the uracil carbonyl group directly interacts with the ribose hydroxyl group, predominated for derivative 8b (Fig 16). In local minimum 4, interaction of both ribose OH hydrogens with the trifluoroacetyl carbonyl group was observed (see S80 Fig).
Structures at other local minima (2, 3, 5 and 6) were again fixed by hydrogen bonds between the ribose OH hydrogen and uracil carbonyl group (see S80 Fig).
In derivative 9a, in the conformer of the first local minimum, the aliphatic chain was located close to the pyrimidine ring; however, no hydrogen bond was observed between them (Fig 17-1). The conformer with local minimum 2 ( Table 2) was characterized by the location of the pyrimidine part of the molecule distant from the aliphatic chain (Fig 17-2). Moreover, in the conformer with local minimum 3 (Table 2), an interaction between the pyrimidine imide group 3 and trifluoroacetyl carbonyl group 37 was detected (Fig 17-3).  The derivative 9b was characterized by two predominant local minima with a total population of approximately 28% (Table 2). These two local minima, which differed only slightly in their combination of dihedral angles, exhibited a close position of the aliphatic chain and pyrimidine ring, although no hydrogen bond was observed between them (Fig 18-1

and 18-2).
To elucidate the conformational changes and relationship among individual local minima, we determined the pathways of the conformational changes with appropriate energy. The pathways included all local minima of the appropriate derivatives with intrinsic energy lower than or equivalent to 30 kJ/mol (and a few intermediate states with higher energies).
In derivative 8a, the transition from local minimum 1 to 2 was connected with energy of 13.76 kJ/mol, whereas other changes required relatively high energy. Although the transition between minima 2 and 3 required low energy, any transition to another local minimum involved overcoming a relatively high energy barrier. Thus, the transition to local minimum 4 and 5 was relatively complicated (see Fig 19).
Identical diagrams were determined for structures 8b, 9a and 9b (see S80, S82 and S83 Figs). Similar to 8a, the transitions among individual local minima were energetically demanding.
This conformational analysis using quantum mechanical investigations revealed that derivatives 8a, 8b, 9a and 9b can form more local minima on the PES in which different interactions between the aliphatic chain (bearing a fluorine atom), uracil and ribose were observed, consistent with the differentiation of signals in the 1 H and 19 F NMR spectra.

Conclusions
In summary, an alternative approach to the synthesis of 5-and 5´-azido derivatives of thymidine riboside was developed. These derivatives were successfully converted to 5-and 5´-1, 2,3-triazol-1-yl derivatives via copper-free click reactions using dibenzoazocine derivative 7.
The NMR spectra of all formed triazoles revealed the presence of conformational isomers., The isomery does not originate from the potentially expected s-cis/s-trans isomery of the amide groups in the aliphatic chain bound to the azocine moiety, what was confirmed by NMR analysis. 1 H-15 N HMQC and 1 H-15 N HMBC NMR spectra predicted the formation of conformational isomers due to hindered rotation of the pyrimidine and azocine parts of the molecule Non-Catalyzed Click Reactions of ADIBO Derivatives and 19 F- 19 F EXSY experiments proved exchange between individual conformers. The computational study of the studied compounds revealed that their isomery is caused either by different positions of the aliphatic chain relative to the nucleoside part of the molecule or by the formation of intramolecular hydrogen bonds. The similar isomery was studied subsequently also for derivatives, in which the nucleoside was replaced by hydrogen (10a) or coumarine scaffold (11a/11b). For the substrates 10a and 11a no isomery was observed, while for derivative 11b at least three conformers were detected.
The formation of triazoles on a nucleic acid sequence can distort the nucleobase, resulting in deformation of the nucleic acid structure, which can have a negative effect on imaging processes. However, this distortion could potentially be exploited in various biological applications that are based on the deformation of nucleic acids.

Experimental Section
LC/MS analyses were performed using UHPLC/MS on a UHPLC chromatograph with a PDA detector and a single quadrupole mass spectrometer; a C18 column was used at 30°C and a flow rate of 600 μl/min. The mobile phase consisted of (A) 0.01 M ammonium acetate in water and (B) acetonitrile, with B linearly programmed to change from 10% to 80% over the course of 2.5 min and then to maintain this concentration for 1.5 min. The column was re-equilibrated at 10% B for 1 min. APCI ionization was operated with a discharge current of 5 μA, vaporizer temperature of 350°C and a capillary temperature of 200°C.
Purification was performed by semi-preparative HPLC on a reverse-phase C18 column, 20 x 100 mm, with 5-μm particles. The mobile phase consisted of acetonitrile and a 10 mM aqueous ammonium acetate gradient over 6 min.
NMR spectra were measured in DMSO-d6 or CDCl 3 on 500 MHz and 400 MHz spectrometers. Chemical shifts (δ are reported in parts per million (ppm), and coupling constants (J) are reported in Hertz (Hz). Acetate salts exhibited a singlet at 1.7-1.9 ppm in 1 H NMR spectra and two resonances at 173 and 23 ppm in 13 C spectra.
HRMS analysis was performed using a high-resolution mass spectrometer operating in positive full-scan mode (120 000 FWMH) in the range of 200-900 m/z. The settings for electrospray ionization were as follows: oven temperature of 300°C, sheath gas of 8 arb. units and source voltage of 1.5 kV. The acquired data were internally calibrated with diisooctyl phthalate as a contaminant in methanol (m/z 391.2843). Samples were diluted to a final concentration of 20 μmol/l with 0.1% formic acid in water and methanol (50:50, v/v). The samples were injected by direct infusion into the mass spectrometer.
and 5-hydroxymethylene-uracil [33] were prepared as described in the literature.
(TIF) S4 Table. The fractional and the percentage population for the local minima of the structure 9b. (TIF)