Efficient Synthesis of α-Branched Purine-Based Acyclic Nucleosides: Scopes and Limitations of the Method

An efficient route to acylated acyclic nucleosides containing a branched hemiaminal ether moiety is reported via three-component alkylation of N-heterocycle (purine nucleobase) with acetal (cyclic or acyclic, variously branched) and anhydride (preferentially acetic anhydride). The procedure employs cheap and easily available acetals, acetic anhydride, and trimethylsilyl trifluoromethanesulfonate (TMSOTf). The multi-component reaction is carried out in acetonitrile at room temperature for 15 min and provides moderate to high yields (up to 88%) of diverse acyclonucleosides branched at the aliphatic side chain. The procedure exhibits a broad substrate scope of N-heterocycles and acetals, and, in the case of purine derivatives, also excellent regioselectivity, giving almost exclusively N-9 isomers.


Introduction
Structurally modified nucleosides and nucleotides belong to the most studied class of biologically active compounds, especially for their potential antiviral and anticancer properties [1][2][3][4]. Among them, so called acyclic nucleoside and nucleotide analogues have played an important role in the combat with various viral infections [3, 5,6]. Acyclic nucleosides (acyclonucleosides), in general, are defined as heterocyclic compounds (usually nucleobase or modified nucleobase) containing one or more functional groups (usually hydroxyls) on the aliphatic side chain. The discovery of acyclovir (Zovirax) [7][8][9] and (S)-DHPA [10], the active ingredient of Duviragel (marketed in former Czechoslovakia), as anti-herpetic drugs (Figure 1), sparked the real interest in synthesis and biological evaluation of various types of acyclonucleosides.

Introduction
Structurally modified nucleosides and nucleotides belong to the most studied class of biologically active compounds, especially for their potential antiviral and anticancer properties [1][2][3][4]. Among them, so called acyclic nucleoside and nucleotide analogues have played an important role in the combat with various viral infections [3, 5,6]. Acyclic nucleosides (acyclonucleosides), in general, are defined as heterocyclic compounds (usually nucleobase or modified nucleobase) containing one or more functional groups (usually hydroxyls) on the aliphatic side chain. The discovery of acyclovir (Zovirax) [7][8][9] and (S)-DHPA [10], the active ingredient of Duviragel (marketed in former Czechoslovakia), as anti-herpetic drugs (Figure 1), sparked the real interest in synthesis and biological evaluation of various types of acyclonucleosides.

Results and Discussion
Treatment of 6-chloropurine (1a) with 2-bromomethyl-1,3-dioxolane (2a) and acetic anhydride (3a) in a presence of Lewis acid was chosen as a model reaction (Table 1). Since we observed formation of alkylated products for the first time in acetonitrile (MeCN), it was used as a suitable solvent for initial optimization of the reaction conditions. At first, the influence of various Lewis acids and of temperature on the reaction course was studied. Thus, an equimolar mixture of starting compounds 1a, 2a, and 3a with corresponding Lewis acid was stirred in MeCN for 15 min either at room temperature or at 70 • C (analogous treatment at 0 • C gave generally much lower yields, data not shown). The results are summarized in Table 1. The best conversion and regioselectivity (only N-9 isomer 4 formed, no N-7 isomer 5 observed) was achieved with the use of SnCl 4 and trimethylsilyl trifluoromethanesulfonate (TMSOTf), up to 83% isolated yields, both at room temperature and at 70 • C. With other Lewis acids used, the conversion was much lower (below 40%) and/or regioselectivity was worse, although N-9 isomer 4 always remained the major product. Increased temperature led to significantly increased yields of 4, compared to room temperature, when FeCl 3 , AlCl 3 , or TiCl 4 were employed as Lewis acids, but the yields were still lower compared to those with SnCl 4 or TMSOTf. Furthermore, prolongation of the reaction time had no positive effect on the conversion (data not shown).
Next, the influence of the solvent was carefully evaluated. The above standard reaction procedure (general procedure A) was carried out in a wide range of solvents (Table 2). In some solvents (DMSO, DMF, NMP, THF, cyclohexane, methanol, pyridine), no reaction was observed, while in other solvents only traces of product (toluene, acetic acid) or low yields of desired product (acetone, DCM) were obtained. High conversion (between 67% and 54%) was observed in nitromethane, EtOAc, and dioxane, but regioselectivity was generally worse and small amounts of undesired N-7 isomer 5 (4-6%) were observed or isolated. The best solvent for the glycosylation proved to be originally used MeCN with an 83% yield of compound 4 (Tables 1 and 2). isomer 4 formed, no N-7 isomer 5 observed) was achieved with the use of SnCl4 and trimethylsilyl trifluoromethanesulfonate (TMSOTf), up to 83% isolated yields, both at room temperature and at 70 °C. With other Lewis acids used, the conversion was much lower (below 40%) and/or regioselectivity was worse, although N-9 isomer 4 always remained the major product. Increased temperature led to significantly increased yields of 4, compared to room temperature, when FeCl3, AlCl3, or TiCl4 were employed as Lewis acids, but the yields were still lower compared to those with SnCl4 or TMSOTf. Furthermore, prolongation of the reaction time had no positive effect on the conversion (data not shown).  Based on the above results (Table 1), the following reaction procedure (general procedure A) was selected for further optimization studies: equimolar (2.0 mmol) mixture of 6-chloropurine (1a), 2-bromomethyl-1,3-dioxolane (2a), acetic anhydride (3a) and TMSOTf in 10 mL of solvent at room temperature for 15 min.
Next, the influence of the solvent was carefully evaluated. The above standard reaction procedure (general procedure A) was carried out in a wide range of solvents (Table 2). In some solvents (DMSO, DMF, NMP, THF, cyclohexane, methanol, pyridine), no reaction was observed, while in other solvents only traces of product (toluene, acetic acid) or low yields of desired product (acetone, DCM) were obtained. High conversion (between 67% and 54%) was observed in nitromethane, EtOAc, and dioxane, but regioselectivity was generally worse and small amounts of undesired N-7 isomer 5 (4-6%) were observed or isolated. The best solvent for the glycosylation proved to be originally used MeCN with an 83% yield of compound 4 (Table 1 and 2). The above reaction conditions (general procedure A) developed for the model reaction (Tables  1 and 2) were slightly optimized into general method B (Scheme 2), where 1.5 equivalent of TMSOTf (instead of 1.0 equivalent) was used, giving comparable or slightly better yield of 4 (85% vs. 83%). It should be noted, that further increase of amount of TMSOTf during the reaction course did not lead to further improvements. With this methodology (general procedure B) in hands, we decided to explore the substrate scope and limitations of this alkylation procedure with regard to Nheterocycles, acetals, as well as acid anhydrides and acyl chlorides.
First, we carried out the reaction with variously substituted purines (including aza/deazapurines, where purine numbering system was used) and with a wide range of cyclic acetals (Scheme 2). Most starting materials were commercially available, some acetals, however, had to be prepared (see the experimental section). It was found, that most of the modified purines were good substrates and afforded the expected products in good to moderate yields, especially with certain type of acetals. For example, when 2-bromomethyl-1,3-dioxolane (2a) was used as the reaction partner, all purine analogues used gave the corresponding products in 54-85% yields, with the exception of unprotected guanine (1h) when only undesired N-7 isomer 23 was obtained in a 15% yield. Moreover, a variety of cyclic acetals, both 5-membered 1,3-dioxolanes (2a, 2b, 2g, 2k, and 2m) and 6-membered 1,3-dioxanes (2c and 2i), afforded the desired products (yields in the range of 11-85%). Nevertheless, when the acetal used contained an aldehyde group (2f), a free hydroxyl (2j) or an azido group (2n), no reaction was observed. Furthermore, the reaction did not take place when 1,3-dioxolanes disubstituted at the acetal carbon atom (compounds 2h and 2l) were used as the reaction partners.

Solvent
Yield The above reaction conditions (general procedure A) developed for the model reaction (Tables 1  and 2) were slightly optimized into general method B (Scheme 2), where 1.5 equivalent of TMSOTf (instead of 1.0 equivalent) was used, giving comparable or slightly better yield of 4 (85% vs. 83%). It should be noted, that further increase of amount of TMSOTf during the reaction course did not lead to further improvements. With this methodology (general procedure B) in hands, we decided to explore the substrate scope and limitations of this alkylation procedure with regard to N-heterocycles, acetals, as well as acid anhydrides and acyl chlorides. In general, desired regioselectivity of the synthetic procedure was great, affording almost exclusively N-9 isomers, with two exceptions: treatment of 2-amino-6-chloro-7-deaza-8-azapurine (1g) with 2-bromomethyl-1,3-dioxolane (2a) gave N-8 isomer 22 (78%) and already mentioned  First, we carried out the reaction with variously substituted purines (including aza/deazapurines, where purine numbering system was used) and with a wide range of cyclic acetals (Scheme 2). Most starting materials were commercially available, some acetals, however, had to be prepared (see the experimental section). It was found, that most of the modified purines were good substrates and afforded the expected products in good to moderate yields, especially with certain type of acetals. For example, when 2-bromomethyl-1,3-dioxolane (2a) was used as the reaction partner, all purine analogues used gave the corresponding products in 54-85% yields, with the exception of unprotected guanine (1h) when only undesired N-7 isomer 23 was obtained in a 15% yield. Moreover, a variety of cyclic acetals, both 5-membered 1,3-dioxolanes (2a, 2b, 2g, 2k, and 2m) and 6-membered 1,3-dioxanes (2c and 2i), afforded the desired products (yields in the range of 11-85%). Nevertheless, when the acetal used contained an aldehyde group (2f), a free hydroxyl (2j) or an azido group (2n), no reaction was observed. Furthermore, the reaction did not take place when 1,3-dioxolanes disubstituted at the acetal carbon atom (compounds 2h and 2l) were used as the reaction partners.
When 2-amino-6-chloropurine (1d) was treated with equimolar amount of acetic anhydride (general procedure B), desired product 18 was observed in traces only (UPLC-MS Acquity Waters, USA, H-Class Core System) and the bis-acetylated product 17 was isolated as the major product (49% yield). Interestingly, this was the only case when the exocyclic amino group underwent acetylation during the reaction and up to now, we have no good explanation for this observation. In order to increase the yield, 2 equivalents of acetic anhydride were used next (under otherwise identical conditions), and compound 17 was obtained in a 59% yield. On the other hand, compound 17 was prepared in a 72% yield, when the reaction started directly from N-2 acetylated 2-amino-6-chloropurine 1e instead of 1d.
Quite a surprise for us was the reaction with carbamate 2k. Carbamates are, in general, relatively unstable under acidic conditions. However, treatment of carbamate 2k with N-2 acetylated 2-amino-6-chloropurine 1e gave desired product 20 in a 64% yield.
Next, the developed alkoxyalkylation methodology (general procedure B) was carried out with selected purines and with various acyclic acetals (Scheme 3). Depending on the starting materials, the reactions afforded products in low to high yields (23-82%), with the exception of acetals containing either unprotected amino (compound 2p) or hydroxyl (compound 2q) group, where no reaction was observed.
Next, the developed alkoxyalkylation methodology (general procedure B) was carried out with selected purines and with various acyclic acetals (Scheme 3). Depending on the starting materials, the reactions afforded products in low to high yields (23-82%), with the exception of acetals containing either unprotected amino (compound 2p) or hydroxyl ( compound 2q) group, where no reaction was observed.
Scheme 3. Preparation of acyclonucleosides by multi-component reaction-the scope of purines and acyclic acetals a,b . a General procedure B. b Isolated yields. c nr-no reaction.
As shown before, the presence of double bond in acetal 2m (Scheme 2) was tolerated under the reaction conditions. We were subsequently interested in using acetal with a triple bond. Thus, treatment of purines 1b and 1l (Scheme 3) with 3,3-diethoxy-1-propyne (2r), diethyl acetal of propynal, afforded desired products 34 (36%) and 36 (31%), respectively, in acceptable yields. In order to extend the substrate scope of the studied synthetic procedure, reactions of various N-heterocycles with selected cyclic and acyclic acetals were carried out next (Scheme 4). The reactions afforded acceptable to high yields (21-88%) of target compounds, no reaction was observed only in the case when 1H-indole (1s) and acetal 2o were used as reaction partners. However, when the hydroxyl group of acetal 2q (Scheme 3) was acetylated, as in acetal 2t (Scheme 4), desired alkylated product 37 was isolated in a 68% yield. The best yield (88% of 38 as inseparable mixture of diastereoisomers) was obtained during the reaction of 3-deazapurine derivative 1n with racemic acetal 2u, which contains both nitrile and ester groups, demonstrating that these functional groups were well-tolerated.
Furthermore, the reaction of compound 1q demonstrated that nitro group and even unprotected carboxylic acid are tolerated during the procedure and product 41 was isolated in a 40% yield. In case of thiazolium trifluoromethanesulfonate 40, which is highly soluble in water, the extraction had to be replaced by the purification on C18-silica gel column. Nevertheless, it can be concluded that described procedure can be applied to a large variety of N-heterocyclic compounds.
The last goal of the study was to test whether different anhydrides of carboxylic acids or even acyl chlorides could be exploited for such methodology (Scheme 5). For this purpose, our model reaction of 6-chloropurine (1a) with 2-bromomethyl-1,3-dioxolane (2a) under standard reaction conditions (general procedure B) was chosen, where acetic hydride (3a) was replaced by other anhydrides or acyl chlorides. acetal 2u, which contains both nitrile and ester groups, demonstrating that these functional groups were well-tolerated.
The regioselectivity was usually great, as demonstrated by formation of products 39 (1,2,4triazole derivative), 41 (pyrazole derivative) and 45 (benzotriazole derivative). The only exception was the reaction of tetrazole 1r with acetal 2m, where regioisomers 42 (26%) and 43 (21%) were isolated in almost equimolar ratio. Furthermore, the reaction of compound 1q demonstrated that nitro group and even unprotected carboxylic acid are tolerated during the procedure and product 41 was isolated in a 40% yield. In case of thiazolium trifluoromethanesulfonate 40, which is highly soluble in water, the extraction had to be replaced by the purification on C18-silica gel column. Nevertheless, it can be concluded that described procedure can be applied to a large variety of N-heterocyclic compounds.
The last goal of the study was to test whether different anhydrides of carboxylic acids or even acyl chlorides could be exploited for such methodology (Scheme 5). For this purpose, our model reaction of 6-chloropurine (1a) with 2-bromomethyl-1,3-dioxolane (2a) under standard reaction conditions (general procedure B) was chosen, where acetic hydride (3a) was replaced by other anhydrides or acyl chlorides. When pivalic anhydride (3b) or benzoic anhydride (3c) were used, the corresponding acylated acyclonucleosides 46 (51%) and 47 (39%), respectively, were obtained in satisfactory yields. The yields were, however, lower compared to the reaction with acetic anhydride (85% of 4). This methodology could be exploited in the synthesis where pivaloyl (Piv) or benzoyl (Bz) protected hydroxyl intermediate would be beneficial for subsequent reaction steps. Moreover, these protecting groups could be easily removed in situ. The reaction, though, did not proceed with p-toluenesulfonic anhydride (3d, no product observed; the expected product 48 would be probably too reactive/unstable), nor with stearic anhydride (3f, traces of product 50 observed in UPLC-MS, impossible to isolate due to its high lipophilicity).
When anhydrides were replaced with the corresponding acyl chlorides, the glycosylation did not proceed well. The reaction of 1a and 2a with acetyl chloride (3g) or benzoyl chloride (3i) did not afford any isolable product (only traces of 4 and 47 were detected by UPLC-MS), while analogous procedure with anhydrides afforded target compounds 4 and 47 in 85% and 39% yields (Scheme 5). On the other hand, when decanoyl chloride (3h) was employed, the reaction afforded the expected acylated product 51 in a 35% yield. In certain cases it may be desirable to prepare lipophilic compounds (e.g., in order to be cell penetrable) and also their isolation can be conveniently done with simple extraction into an organic solvent. When pivalic anhydride (3b) or benzoic anhydride (3c) were used, the corresponding acylated acyclonucleosides 46 (51%) and 47 (39%), respectively, were obtained in satisfactory yields. The yields were, however, lower compared to the reaction with acetic anhydride (85% of 4). This methodology could be exploited in the synthesis where pivaloyl (Piv) or benzoyl (Bz) protected hydroxyl intermediate would be beneficial for subsequent reaction steps. Moreover, these protecting groups could be easily removed in situ. The reaction, though, did not proceed with p-toluenesulfonic anhydride (3d, no product observed; the expected product 48 would be probably too reactive/unstable), nor with stearic anhydride (3f, traces of product 50 observed in UPLC-MS, impossible to isolate due to its high lipophilicity).
When anhydrides were replaced with the corresponding acyl chlorides, the glycosylation did not proceed well. The reaction of 1a and 2a with acetyl chloride (3g) or benzoyl chloride (3i) did not afford any isolable product (only traces of 4 and 47 were detected by UPLC-MS), while analogous procedure with anhydrides afforded target compounds 4 and 47 in 85% and 39% yields (Scheme 5). In order to reveal the plausible reaction mechanism, the reaction (general procedure B with 1a, 2a, and 3a) was run in deuterated MeCN directly in the NMR tube. Intermediate 52 (Scheme 6) was observed immediately after the reaction set up, together with a small amount of ethylene glycol diacetate (53) formed as a by-product (its structure was compared with commercially available sample). In order to confirm the structure of compound 52, it was prepared independently in a 91% yield by treatment of 2-bromomethyl-1,3-dioxolane (2a) and acetic anhydride (3a) in the presence of sulfuric acid (a modified procedure of Rosowsky et al. [28] for the preparation of 2-acetoxyethyl acetoxymethyl ether). Compound 52 was fully characterized and its NMR spectra were identical with those of the intermediate observed during the NMR experiment. Thus, the first reaction step of the synthesis of acyclic nucleosides is acetolysis of the corresponding acetal, in this case of compound 2a. We also performed reaction of compound 52 with 6-chloropurine (1a) and TMSOTf (1.5 equivalent) in MeCN (analogy to general procedure B) and product 4 (Scheme 6) was isolated in a 30% yield. Thus, the multi-component one-pot approach affords much better yields of the target compound 4 (85% versus 30%) and eliminates the reaction step needed for preparation of the alkylating agent (e.g., compound 52). synthesis of acyclic nucleosides is acetolysis of the corresponding acetal, in this case of compound 2a.
We also performed reaction of compound 52 with 6-chloropurine (1a) and TMSOTf (1.5 equivalent) in MeCN (analogy to general procedure B) and product 4 (Scheme 6) was isolated in a 30% yield. Thus, the multi-component one-pot approach affords much better yields of the target compound 4 (85% versus 30%) and eliminates the reaction step needed for preparation of the alkylating agent (e.g., compound 52). Scheme 6. Study of the reaction course and synthesis of intermediate 52.

Materials and Methods
General Methods. Unless otherwise stated, solvents were evaporated at 40 °C/2 kPa and prepared compounds were dried at 30 °C at 2 kPa. Reaction flasks were heated in aluminum heating blocks. Tetrahydrofuran, dioxane, and acetonitrile were dried by activated neutral alumina (Drysphere). Dimethylformamide was dried by activated molecular sieves (3 Å). Other dry solvents were purchased from commercial suppliers. Analytical TLC was performed on silica gel pre-coated aluminum plates with the fluorescent indicator Merck 60 F254 (Sigma-Aldrich, Prague, Czech Republic). Flash column chromatography was carried out by Teledyne ISCO CombiFlash Rf200 with a dual absorbance detector (Teledyne ISCO, Lincoln, NE, USA). HRMS spectra (ESI + or EI + ) were recorded on LTQ Orbitrap XL spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with the ESI or EI ionization method. NMR spectra were recorded on Bruker (Rheinstetten, Germany) Avance III (400 MHz or 500 MHz) spectrometers and referenced to the residual solvent signal or a specified additive (see Supplementary Materials). The structure determination for all compounds and the assignment of NMR signals was done with the help of a combination of 1D proton and carbon experiments with 2D H,H-COSY, H,C-HSQC, and H,C-HMBC experiments. Standard experiments from the Bruker pulse-sequence library were used. Purity of compounds was measured on Waters Scheme 6. Study of the reaction course and synthesis of intermediate 52.

Materials and Methods
General Methods. Unless otherwise stated, solvents were evaporated at 40 • C/2 kPa and prepared compounds were dried at 30 • C at 2 kPa. Reaction flasks were heated in aluminum heating blocks. Tetrahydrofuran, dioxane, and acetonitrile were dried by activated neutral alumina (Drysphere). Dimethylformamide was dried by activated molecular sieves (3 Å). Other dry solvents were purchased from commercial suppliers. Analytical TLC was performed on silica gel pre-coated aluminum plates with the fluorescent indicator Merck 60 F254 (Sigma-Aldrich, Prague, Czech Republic). Flash column chromatography was carried out by Teledyne ISCO CombiFlash Rf200 with a dual absorbance detector (Teledyne ISCO, Lincoln, NE, USA). HRMS spectra (ESI + or EI + ) were recorded on LTQ Orbitrap XL spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) with the ESI or EI ionization method. NMR spectra were recorded on Bruker (Rheinstetten, Germany) Avance III (400 MHz or 500 MHz) spectrometers and referenced to the residual solvent signal or a specified additive (see Supplementary Materials). The structure determination for all compounds and the assignment of NMR signals was done with the help of a combination of 1D proton and carbon experiments with 2D H,H-COSY, H,C-HSQC, and H,C-HMBC experiments. Standard experiments from the Bruker pulse-sequence library were used. Purity of compounds was measured on Waters UPLC-MS system (Santa Clara, CA, USA) that consisted of Waters UPLC H-Class Core System (column Waters ACQUITY UPLC BEH C18 1.7 mm, 2.1 × 100 mm), Waters ACQUITY UPLC PDA detector and Mass spectrometer Waters SQD2. The universal LC method was used (eluent H 2 O/MeCN, gradient 0-100%, run length 7 min) and MS method (ESI + and/or ESI − , cone voltage = 30 V, mass detector range 100-1000 Da). Purity of the final compounds was >95%, unless otherwise stated.
Preparation of starting material. Starting compounds and reagents were purchased from commercial suppliers (Sigma-Aldrich, Prague, Czech Republic; Fluorochem, Hadfield, UK; Fisher Scientific, Pittsburgh, PA, USA; Carbosynth, Compton, UK; TCI Europe, Zwijndrecht, Belgium) and used without further purification. In several cases, we had to prepare the starting material that was not commercially available. Compounds 2d [29], 2i [30], and 2m [31] were prepared according to the published procedures.

Conclusions
In summary, we have developed an efficient strategy for the preparation of acylated purine-based acyclonucleosides branched at the hemiaminal ether carbon atom. The three-component reaction takes advantage of cheap and easily available starting material, namely acetals, acetic anhydride and trimethylsilyl trifluoromethanesulfonate (TMSOTf), and of simple procedure (one-pot reaction, room temperature, short reaction time). The general procedure is based on equimolar reaction mixture of N-heterocycle, acetal, acetic anhydride, and TMSOTf (1.5 equivalent is preferable) in MeCN and offers the target acyclonucleosides in moderate to high yields (up to 88%). The substrate scope is relatively broad, both as for the N-heterocycle and as for the acetal (cyclic or acyclic) and many functional groups are tolerated (e.g., double and triple bonds, halogen, tosyl, cyano, keto, and ester groups). It was shown that starting 1,3-dioxolanes can be branched either at C-2 or at C-4/C-5 positions; however, no reaction was observed with acetals bearing azido group and free hydroxyl or amino groups. The N-heterocycles usually react with a great regioselectivity and also the unprotected exocyclic amino group is mostly well-tolerated. Although other acid anhydrides (e.g., pivalic anhydride) can be alternatively used, the best yields were obtained with acetic anhydride.
This efficient, simple and fast alkoxyalkylation methodology offers an efficient approach to diverse protected (acylated) acyclic nucleosides which can serve as convenient intermediates for subsequent synthesis of various target compounds (e.g., acyclonucleosides and acyclic nucleoside phosphonates). The methodology is currently being exploited in our laboratory in order to prepare variously branched acyclonucleosides and acyclic nucleoside phosphonates, which are further studied as potential inhibitors of enzymes of purine metabolism (e.g., adenylate cyclases, purine nucleoside phosphorylases, and 6-oxopurine phosphoribosyltransferase) with the aim to develop potential therapeutic agents.