Selectivity in C-alkylation of dianions of protected 6-methyluridine

A regioselective synthesis of 6-ω-alkenyluridines 3, precursors of potent antiviral and antitumor cyclonucleosides 5, is described. While ω-alkenyl halides do not alkylate 6-lithiouridine, compounds 3 were prepared in a regioselective manner by sequential treatment of 6-methyluridine 2 with LTMP or LDA (4 equiv) in THF at −30 °C followed by alkylation with ω-alkenyl bromides.


Introduction
Conformationally restricted C-C bridged cyclonucleosides bearing a linkage between the sugar moiety and the nucleobase, exhibit a broad spectrum of antiviral and antitumor activities [1][2][3][4].Cyclonucleosides are excellent tools for studying the role of the conformational parameters that are critical for the design of new nucleoside drug candidates [4][5][6][7][8].These cyclic compounds are expected to have a beneficial biological impact especially toward enzymatic repair processes [9].
As part of an ongoing program directed by one of us (C.L.) toward the synthesis and development of new cyclonucleosides 5 [5,6], we envisioned that the general transformation outlined in Scheme 1 might afford a facile entry to 5 from dialkenyl precursors 4 by ring-closing metathesis [10][11][12].The strategy relies on the preparation of unknown 6-ω-alkenyluridine key intermediates 3. We report herein that sequential ring lithiation/ methylation of the simple protected uridine 1 leading to 2 fol- lowed by lateral lithiation/alkylation with ω-alkenyl bromides provides a useful regioselective chain-extension procedure and an efficient route to 3.

Results and Discussion
Most methods for the construction of C-substituted nucleosides are based on ring lithiation of nucleoside derivatives followed by their reaction with appropriate electrophiles.Thus, sequential lithiation of 2',3'-O-isopropylideneuridine (6) with LDA in THF (Figure 1) and electrophilic quenching with n-bromobutane was reported to give 6-n-butyl-2',3'-O-isopropylideneuridine (8) in a regiospecific manner (60%) [13].It seems likely that the reaction proceeds via trianion 7 where the 5'-OLi group can easily participate in the stabilization of the 6-lithio intermediate.ω-Alkenyl bromides are known to be poor electrophiles toward organolithiums [14], and indeed, 7 failed to react, in our experiments, with 4-bromo-but-1-ene to give 9.
Literature furnishes little information regarding lateral lithiations in the nucleoside field and the data, scarce as they are, even appear to be inconsistent at first sight.Treatment of 2',3',5'-tri-O-benzoyl-3,6-dimethyluridine (13) with chloroacetone or 2-chloroacetophenone in the presence of LDA (1.2 equiv, THF, −78 °C) afforded 6-(oxiranylmethyl)uridine derivatives 14 exclusively (Figure 3) [29].With 5-chloro-2pentanone, the reaction led to a mixture of 5-and 6-substituted uridine regioisomers 15 and 16 in 47% and 28% yield, respectively.It was suggested that the N-1 sugar moiety in the syn orientation of the nucleoside might affect the access of a very sterically demanding electrophile, such as 5-chloro-2pentanone, to the 6-position.This hypothesis was confirmed by a probe experiment where an even more sterically hindered racemic 3-bromocamphor was used as an electrophile.The corresponding C5-alkylated uridine derivative was obtained as the only recovered product, in low yield (23%), besides the unreacted substrate.
In the past, the regioselectivity of reactions of allyl anions have sometimes been explained using the HSAB theory [34].In the present case, soft electrophiles (ω-alkenyl bromides) are used in the alkylation reaction.However, it is not straightforward to predict the softest center of 18.In the literature, the regioselectivity of lithiation of allyl anions substituted by one nitrogen at the central carbon (C=C(N)-C) has scarcely been studied [38][39][40].Deprotonation of simple enamines or allylamines employing n-BuLi and t-BuLi/t-BuOK produced nitrogen-  substituted allylic anions which undergo protonation, alkylation, trimethylsilylation and reaction with carbonyl compounds and epoxides either exclusively or predominantly at the γ-position [41][42][43].Previous work also showed that cyclic enaminoketones, esters and nitriles were converted into their enolate with n-BuLi and alkylated with a variety of alkylating agents, affording the product of an exclusive γ-alkylation [42][43][44][45].
The results are summarized in Table 1.An excess of LDA (4 equiv) at −70 °C produced a dilithium reagent, which was presumed to be 18, as a yellow solution.The colour faded when allyl bromide (8 equiv/ −70 °C → rt/12 h) was introduced, providing a mixture of regioisomers 3a and 19a which were separated by chromatography (entry 1, 58% and 10% yield, respectively).With LTMP, the delivery of a less acidic conjugated amine (TMP) in the reaction medium could be expected to prevent protonation of lithiated intermediates and thus to allow more efficient trapping by an electrophile [46,47].Indeed, LTMP gave a better yield, but a slight decrease in the regioselectivity was observed (3a/19a 65:20 allowed the assignment of the proton-proton correlations of H7 and the allylic methylene groups.Metalation with s-BuLi/ TMEDA complex was less efficient although the reaction did not lead to degradation products (entry 6).Ring/internal lithiations of uridine derivatives with s-BuLi/TMEDA are usually performed with fully TBDMS-protected ribofuranose nucleosides to allow better regiochemical control and to prevent nucleophilic attack of the base on the sugar moiety [48].

Conclusion
In summary, a straightforward approach to 6-ω-alkenyluridines 3 from readily available protected uridine 1 is proposed.Whereas direct ring alkylation of 6-lithiated uridine 11 with ω-alkenyl bromides failed, our approach relies on lateral lithiation/alkylation of 6-methyluridine 2. The total synthesis and biological properties of cyclonucleosides 5 will be reported separately.
-ene to give 19b, and 3b is formed exclusively.The lateral alkylation of uridine enolate 18 was best accomplished through use of LDA or LTMP as the carbanion generating species, rather than LiHMDS or s-BuLi/TMEDA.The bisallylated products 20 and 21 (Figure4) were obtained in 40% and 18% yield, respectively with LiHMDS at −70 °C and quenched with allyl bromide (entry 5).This result suggests the remetalation of 3a is faster than the destruction of LiHMDS by the excess of allyl bromide.Structure of 20 was confirmed by1H NMR and a two-dimensional COSY experiment, which