Larger laboratory scale synthesis of 5-methyluridine and formal synthesis of its L -enantiomer

A larger laboratory scale synthesis (>60 g per run) of 5-methyluridine is presented. The critical intermediate 1,2-O -isopropylidene- α -D-ribofuranose was prepared from very cheap D -glucose via D -allose. Its L -enantiomer was obtained from L -arabinose via L -glucose

Introduction 5-Methyluridine 1 is a component of the ribonucleic acids from which it can be isolated, however demand for it far exceeds a supply in this way.5-Methyluridine is a starting compound for the synthesis of 3´-azido-2´,3´dideoxythymidine or AZT (Zidovudine, Retrovir) 3 (via 2´-deoxy compound 2) and also to obtain 2´,3´-dideoxy-2´,3´-didehydrothymidine or d4T (Zerit) 4, both used as potent inhibitors of the reverse transcriptase, a critical enzyme necessary for multiplication of the HIV virus responsible for the AIDS epidemic. 1The sugar moieties present in 1-4 belong to the D series.L-Thymidine 6 is very active inhibitor of the reverse transcriptase of the hepatitis B virus (HBV). 2 Compound 6 is marketed under the name Telbivudine (Sebvio, Tyzeka) and is the enantiomer of 2. As such it can be prepared from the L 5-methyluridine 5 via deoxygenation at its 2´-position. 3onsidering the demand for both enantiomers 1 and 5, uniform access to both of them is an attractive synthetic goal.The stereochemical characteristics of 1,2;5,6-di-O-isopropylidene-α-D-glucofuranose 7 permit its transformation into 1,2-O-isopropylidene-α-D-ribofuranose 12 4 and further to 1-O-acetyl-2,3,5-tri-O-benzoylβ-D-ribofuranose 15 and finally to 5-methyluridine 1.It is obvious that the same sequence of reactions performed on L-glucose will furnish L 5-methyluridine 5. Compound 15 can be directly prepared form D-ribose using HCl-MeOH (→ 13), 5,6 H2SO4-MeOH 7 or AcCl-MeOH 8 followed by benzoylation (→ 14) and acetolysis (→ 15), but since D-glucose is abundant and is one of the cheapest chiral compounds available, we used it to obtain D-ribose and finally 1 at a larger laboratory scale (> 60 g per run) in excellent cumulative yields of ca.90% for the sequence 15→ 17→ 1.We also tried to obtain 1,2-O-isopropylidene-α-L-ribofuranose 32 using Lglucose which was obtained from L-arabinose via the nitromethane one-carbon extension (Fischer-Sowden reaction 9,10 which is a carbohydrate version of the Henry reaction, [11][12][13][14][15] ) followed by a Nef reaction (sodium aci salt of nitroalditols, H2SO4). 10The other versions of this process include ozonization, 16,17 and peroxomolybdate-H2O2; 17,18 non-carbohydrate examples include among others "basic silica gel", 19 NaClO2 20 and H2O2-K2CO3; 21 the Nef reaction has been recently reviewed. 22,23This met with a limited success due to difficulties in separation of the necessary L-glucose 27 from the L-mannose 28 by chromatography (see below).A much better result was obtained using reasonably priced L-xylose which was converted to the L-ribose 32 in batches of ca. 15 g per run.Since 32 can be elaborated to produce L 5-methyluridine 5 by the same set of conditions as those applied to reach 5-methyluridine 1 from the D-ribose 12, we can claim that a formal synthesis of L 5 was also realized.

Results and Discussion
A transformation of D-glucose into D-ribose via inversion of the configuration at the C3 position followed by fission of the C5-C6 bond (dehomologation) is known, and we followed the published procedures with some additional modifications.Alternative methods to produce D-ribose have been reviewed. 24The critical inversion of configuration at the C3 position of D-glucose to get D-allose 9 requires the transient ulose 8. Reduction of 8 proceeds with practically complete stereoselectivity (→ 9) using NaBH4 but application of the LiAlH4 gives mixtures of 7 and 9. 25 Some care must be exercised during the oxidation step (7→ 8).The cheapest and operationally easiest oxidant is DMSO-Ac2O mixture, 26 even though foul smelling divalent sulfur compounds are formed.The other oxidants include DMSO-P2O5, 27 pyridinium dichromate-Ac2O, 28 pyridinium chlorochromate, 29 and RuCl3-NaIO4, 30 (RuO2-NaIO4 31 was reported to also form Baeyer-Villiger overoxidation products) and finally Dess-Martin periodinane. 32DMSO-Ac2O is not a very active system and its application may result in incomplete conversion of 7 into 8, and consequently, the next reduction step may furnish a mixture of the allose 9 and unreacted glucose 7, which are difficult to separate.In fact, unreacted 7 was reported to be present even after 24 h of reaction time. 26The same kind of problem has been observed before in different context. 33The very active CrO 3 -Py-Ac 2 O mixture 34 provided complete oxidation in less than 1 hour at room temperature for 25-39 g batches of 7. A weak point of this method is formation of insoluble tars composed of the reduced chromium compounds complexed with pyridine.A very interesting one-pot oxidation-reduction procedure was published which consists of the addition of the NaBH4 directly to the DMSO-oxalyl chloride oxidation mixture, 35 although this was performed on small scale.The dehomologation step (9→ 10→ 11→ 12), i.e. a transformation of the D-allose (a hexose) to D-ribose (a pentose) can be performed using orthoperiodic acid H5IO6 followed by NaBH4 reduction. 36,37The orthoperiodic acid is strong enough (pKa ca 3.3 38 ) to promote a hydrolysis of the more reactive C5-C6 acetonide in 9 to liberate a diol 10 which was subsequently cleaved by the same reagent to furnish the aldehyde 11 which in turn was subjected to NaBH4 reduction to yield the ribo compound 12.This procedure is very attractive and in fact it was successfully used on a small scale (1 g) in good yields (e.g.29→ 32, see below).When applied to 9 at a 27 g scale however, it furnished the 1,2-O-isopropylidene-α-D-ribofuranose 12 in yield as low as 25%.The reason for this is unclear, but probably can be traced to incomplete removal of the iodic acid HIO3 formed during oxidation of the C5-C6 diol.HIO3 crystallized in the reaction mixture and must have been filtered off.In the case of incomplete crystallization and removal, NaBH4 reduced it to the transient hypoiodous acid 39 which presumably degraded the aldehyde 11 oxidatively.To avoid this it was much better to perform separately the hydrolysis (9→ 10) using 0.8% H2SO4 in MeOH-H2O mixture 40 followed by NaIO4 oxidation (→ 11) and final NaBH 4 reduction to get the 1,2-O-isopropylidene-α-D-ribofuranose 12 in much better yield (9→ 12, 76%).Removal of the acetonide (70% aq.AcOH, 80 °C) followed by Fischer type glycosylation (MeOH, cat.H2SO4 or HCl) furnished kinetically controlled furanosides 13 which, upon conventional benzoylation (BzCl, Py) followed by acetolysis (Ac2O, AcOH, cat.H2SO4), furnished the necessary 1-O-acetyl-2,3,5-tri-O-benzoyl-β-Dribofuranose 15.The best cumulative yields for this sequence was 45% based on 12.There are some variations in the published conditions and the yields of 15 obtained by acetolysis. 5,6,41Cimpoia et al. 41 reported that the best procedure was with a decreased amount of H2SO4 in relation to AcOH and Ac2O and low temperature to avoid formation of the open-chain acetal acetates.Transformations of 12 to 15 were performed without isolation of the intermediates.The compound 15 must be thoroughly dried since traces of i-PrOH used for its crystallization react with SnCl4 during the next coupling step to liberate HCl which may compromise the yield of 17.The compound 15 was finally coupled with the trimethysilylated thymine (thymine, HMDS, cat.(NH4)2SO4, bp., 3-4 h) under the influence of SnCl4 in CH3CN 42 to get 2,3,5-tri-O-benzoyl-β-D-ribofuranosyl thymine 17 in nearly quantitative yield.Alternative versions of this coupling include different activating groups at the anomeric position, like chloride, 43 methylcarbonate, 44 1,2-epoxide, 45 N-phenyl trifluoracetimidate, 46 or S-tolyl group, 47 different catalysts like TMSOTf, 48 BiBr3, 49 or Ph3PAu + -N(SO2CF3)2 , 50 different pattern of protection of the ribofuranosyl moiety like 1,2,3,5-tetra-O-acetate 49 or 1,2-di-O-acetyl-3,5-di-O-benzoate, 51 different activation of the thymine moiety via bis(tributylstannylation) 52 rather than bis(trimethylsilylation), and finally a solvent-free ball milling procedure. 53Compound 17 was finally deprotected using Zemplén conditions (MeOH, cat.NaOMe).Methyl benzoate (an oil) formed during this reaction interferes with the crystallization of 1 and for this reason was removed by partition between CHCl3 and water (1 remained in the water phase).The 5-methyluridine 1 formed in this way was isolated in ca 90% yield for two steps (coupling and deprotection).It is interesting to note that the commercial 1,2,3,5-tetra-O-acetyl-β-D-ribofuranose 16 under the same coupling conditions furnished 2,3,5-tri-O-acetyl-β-D-ribofuranosyl thymine 18 in lower yield (69.5%) even though a clear spot-to-spot reaction also took place.This probably can be attributed to inferior stabilization of the reactive cation 20 in comparison to the benzylic cation 19.Scheme 1 summarizes this part of the project.Scheme 1. Synthesis of 5-methyluridine 1 starting from D-glucose via D-allose.
L-Ribose, 54,55 a necessary substrate to obtain the L 5-methyluridine 5 is a known compound that can be obtained from L-arabinose via epimerization at the C2 position catalyzed by molybdic acid (Bílik reaction), 56,57 starting from D-fructose, 58 D-galactose, 59 D-lyxose, 60 or D-ribose via transposition of the C1-C5 position. 61,62he latter transformation is possible due to the enantiotopic relationship between both -CH2OH groups in Dribitol (and also in any alditol which has a plane of symmetry like galactitol or allitol.In fact D-galactose was transformed into L-galactose 63 by the same kind of transposition).However, having accomplished a synthesis of 1 using D-glucose, we wanted to apply L-glucose for the same purpose to obtain the L enantiomer 5. L-Glucose can be prepared from D-glucose by the published procedure. 64It is also commercially available but due to the prohibitively high price we tried to obtain it from reasonably priced L-arabinose via a one carbon atom extension.This approach is shown in the Scheme 2. The Henry reaction (aldehyde/ketone, nitroalkane, base) [9][10][11][12][13][14][15] performed on L-arabinose and CH3NO2 and NaOMe 9,10 furnished a mixture of crystalline 1-deoxy-1nitro-L-glucitol 21 and 1-deoxy-1-nitro-L-mannitol 22 in approximately equal proportion.This addition to diastereotopic sides of the carbonyl group should follow the Felkin-Anh model 65 as shown in 23 and 24 (Figure 1) where the C3-C5 fragment is treated as a large group and the C2-OH as a medium one.The privileged attack on the Re side should furnish the R configuration at the C2 atom, i.e. the L-gluco diastereoisomer 21 should predominate.It is not a case, though.A reason for this is unclear, but one can consider extensive hydrogen bonding between a solvent (MeOH) and the C=O group.This may override the steric and stereoelectronic factors which control the Felkin-Anh transition state.Also, a fully formed carbonyl group might not even have been present.Rather, reactive forms might have been the species with the hemiacetal rings being partially opened with retained α or β configuration.This may additionally influence the steric outcome.Theoretical calculations show that the transition states during the Henry reactions are such, that the negatively charged NO2 moiety is far away from the carbonyl oxygen 66 atom as in 23a and 24a which apparently have the same energies.Consequently, both epimeric products 21 and 22 are formed without much stereoselection.This is additionally influenced by reversibility of the addition in basic medium.Irrespective of the mechanism, there is no preference for 21 over 22.The 21/22 mixture was described as being separable by tedious fractional crystallization 9,10 and the resulting 1-nitro-L-glucitol and 1-nitro-Lmannitol were subjected separately to the Nef reaction to give L-glucose and L-mannose, respectively.However, since we needed the 2,3;5,6-di-O-isopropylidene-L-mannofuranose for another project we performed the Nef reaction (a.21/22, NaOH; b.H2SO4) 10 without fractional crystallization.A mixture of Lglucose 25 and L-mannose 26 thus obtained was subjected to isopropylidenation (acetone, H2SO2, ultrasound) 67 to yield a mixture of the di-O-isopropylidenated compounds 27 and 28 which were separated by vacuum-dry chromatography 68,69 at this stage.Eluted first was the L-mannose 28 followed by more polar Lglucose 27.The NMR characteristics of both 27 and 28 are identical to those of their D-enantiomers.Due to small difference of the Rf values of 27 and 28, their separation was successful at small scale only (ca 2 g of mixture per run).The 1,2;5,6-di-O-isopropylidene-α-L-glucofuranose 27 was then subjected to the oxidation/reduction (a.CrO3-Py-Ac2O; 34 b.NaBH4) sequence to invert the configuration at the C3 position to produce the L-allose 29 which in turn was subjected to dehomologation (a.H5IO6, b.NaBH4 36,37 ) at a 1 g scale which furnished the 1,2-O-isopropylidene-α-L-ribofuranose 32 in a cumulative 69% yield (29→ 32).As mentioned above for the D-enantiomer this process did not function well at elevated scale.Considering the overall length of this process, difficulties during the separation of 27 and 28, and the low overall yield of the 1,2-O-isopropylidene-α-L-ribofuranose 32, a more efficient route was devised using L-xylose.This commercially available pentose was transformed in one pot 70 [a.acetone, H2SO4; b. partial neutralization with Na2CO3, H2O] to the 1,2-O-isopropylidene-α-L-xylose.Selective protection of the primary OH group via silylation (t-BuMe2SiCl, imidazole; t-BuPh2Si- 71 and Tr- 72 were published for the same purpose) was nearly quantitative (34→ 35).The compound 35 was subjected to inversion of configuration at the C3 position via oxidation-reduction (a.CrO3-Py-Ac2O; 34 b.NaBH4) followed by desilylation to furnish the 1,2-Oisopropylidene-α-L-ribofuranose 32 with a total selectivity and good yield.Compound 32 is the L enantiomer of the intermediate D 12, and can be further elaborated to obtain the L 5-methyluridine 5.

Conclusions
In conclusion, a larger laboratory scale route to obtain 5-methyluridine 1 is described using D-glucose as a precursor of the pivotal 1,2-O-isopropylidene-α-D-ribofuranose 12.Its enantiomer L 32 was obtained starting from either L-arabinose via L-glucose and L-allose, or (much better) from L-xylose.

Experimental Section
General.EtOAc was dried by azeotropic removal of water; ca.20% of a forerun was rejected and the rest was distilled.Acetone and CH2Cl2 were dried by shaking with P2O5 during 20 min, rapid filtration and distillation.DMF was dried by azeotropic removal of water using benzene or toluene (ca.20% of the volume of DMF).Pyridine was dried by storage over KOH.MeOH was dried by Mg/I2 method.The 1 H and 13 C spectra were obtained on the Varian 300 MHz spectrometer unless otherwise stated.Exact mass measurements were obtained on the Waters Xevo G2-XS QTof spectrometer.Optical rotations were measured on the Jasco P-2000 241 automatic polarimeter at ca 26 °C.Moisture-sensitive reactions were performed using protecting atmosphere of argon dried by passage through "blue silica gel".Evaporations of the solvents were performed at ca 40 °C.MgSO4 was used to dry the extracts.Column chromatography was performed using silica gel 70-230 mesh from the Fluka.TLC chromatography was performed on the 0.2 mm silica gel aluminum plates (Fluka) and the spots were reveled using 10% H 2 SO 4 in MeOH and heating at ca. 110 ° C.

1,2;5,6-Di-O-isopropylidene-α-D-allofuranose 9.
A round-bottom flask equipped with a reflux condenser and magnetic stirring bar was charged with CH2Cl2 700 (mL) and CrO3 (60 g, 600 mmol).The flask was immersed in ice-water, and pyridine (97 mL, 1200 mmol) was added portion wise during 10 min.The cooling bath was removed.After stirring for 1 h at rt, the dark brown mixture was cooled again with ice-water bath and 7 (39.3g, 151 mmol) was added portionwise.After each addition of 7 a small volume of Ac2O was added.The total volume of Ac2O was 75 mL, 61.3 g, 600 mmol.After these intermittent additions which required 15 min the cooling bath was removed and oxidation continued for 25 min counting from the end of additions.TLC showed complete conversion of 7 Rf 0.49 into more polar ulose 8 Rf 0.33 (hexane -EtOAc 1:1).Most of CH2Cl2 was evaporated below 40 °C and 1:1 mixture of toluene -EtOAc, (500 mL) was added.This resulted in precipitation of insoluble black tar.The supernatant was decanted and the solid residue was washed twice with the same solvent.The combined solutions were passed through a short silica gel column prepared in toluene -EtOAc 1:2 using over pressure.The column was eluted with the same solvent system and product-containing fractions were evaporated.Xylenes (100 mL) were added and evaporation was continued to expel all residual pyridine.The oil obtained was dissolved in 96% EtOH (300 mL), cooled in ice-water bath and NaBH4 (5 g, 132 mmol) was added portionwise while maintaining magnetic stirring.The cooling bath was removed and stirring was continued overnight.TLC showed allose 9 Rf 0.36 slightly less polar than the ulose 8 (hexane -EtOAc 1:1).Acetone (10 mL) was added to destroy the excess of NaBH4 and most of the volatiles were evaporated.The residue was taken up in CH2Cl2 and washed with water.The organic phase was dried (MgSO4) and evaporated to yield solid crude 9 (31.1 g, 79% for two steps).Mp 72-75 °C (hexane -EtOH), [ 8 mmol) was dissolved in dry EtOAc (150 mL) and stirred with H5IO6 (99% pure, 28.7 g, 124.6 mmol) added in three portions.A white precipitate soon appeared.TLC showed the aldehyde 11 Rf 0.29 (hexane -EtOAc 1:2); the substrate 9 has Rf 0.36 (hexane -EtOAc 1:1).After 2 h filtration was performed using a sintered glass and the precipitate was washed with EtOAc.The volatiles were removed by evaporation.Some solid material appeared.EtOAc (50 mL) was added and filtration/evaporation was repeated.The residual oil was briefly dried on an oil pump, dissolved in 96% EtOH (100 mL), cooled in icewater bath and treated with NaBH4 (5 g, 132 mmol).The cooling bath was removed.Rf of the product 12 was 0.46 in CH2Cl2 -MeOH 20:1.5.After 3 h most of EtOH was evaporated and the residue was dissolved in CH2Cl2 and this solution was washed with water.The water phase was back extracted twice with CH2Cl2.The combined organic phases were dried, filtered and evaporated.Purification by chromatography CH2Cl2 -MeOH 20 :1.4 furnished 12 (4.9g, 25%).B. Via stepwise hydrolysis and NaIO4/NaBH4 treatment.To a cold (ice bath) solution of 9 (35 g, 134.6 mmol) in MeOH (300 mL) was added cold 0.8% H2SO4 (250 mL), and the mixture was left overnight at rt. TLC showed that 9 (Rf 0.36 in hexane -EtOAc 1:1) reacted to form 10 Rf ~0 in the same solvent system, or Rf 0.21 in CH2Cl2 -MeOH 9:1.Amberlite IRA 410 (OH -) was added to neutralize the acid and was removed by filtration and washed with MeOH.The volatiles were removed by evaporation to yield crude 10 as a syrup.A small amount of this material was purified by chromatography using CH2Cl2 -MeOH 9:1 to get the crystalline material, mp 130-134 °C (EtOAc -EtOH); lit. 32mp.131-133 °C (Et2O -MeOH) The bulk of the crude product was dissolved in 96% EtOH (250 mL) and treated with a suspension of NaIO4 (30 g, 140 mmol), in H2O (100 mL) with magnetic stirring.A white precipitate started to deposit immediately.After 4 h TLC showed a conversion of 10 into the aldehyde 11 Rf 0.29 in hexane -EtOAc 1:2.The whole mixture was filtered on a sintered glass and the solid material was washed with EtOH.NaBH 4 (6 g, 158.7 mmol) was added to the cold (ice bath) filtrate and the mixture was stirred magnetically during 5 h at rt.The work-up and purification as described above furnished 12 (16.9g, 76% for three steps).12: mp 85 °C (hexane -EtOAc), [α]D 26 + 69 (c 2.2, EtOH); lit. 73 The conversion of 12 into 15 was performed without isolation of the intermediates.The acetonide 12 (25 g, 132 mmol) in 70% AcOH (150 mL) was maintained at 80 °C during 3 h, whereupon the volatiles were removed by evaporation below 40 °C.Coevaporation with xylenes and drying on an oil pump furnished glassy material, which was dissolved in dry MeOH (400 mL) and the mixture was cooled down in ice bath and magnetically stirred.H2SO4 (97%, 2 mL) was added slowly and the mixture was left for 18 h in a refrigerator.Saturated aq Ba(OH)2 was added to neutrality and precipitated BaSO4 was removed by filtration through Celite.The volatiles were removed on an evaporator and on an oil pump.To the residue was added pyridine(300 mL), followed by BzCl (54 mL, 64.6 g, 460 mmol), added dropwise under magnetic stirring and with cooling in an ice bath and under a blanket of argon.After an overnight reaction, TLC (CHCl3 -MeOH 9:1) showed two products having Rf 0.68 and 0.80, presumably both anomeric compounds 14.Water (5 mL) was added to hydrolyze the excess of BzCl and 2 h later an extraction was performed (CH2Cl2 -ice -5N HCl).The organic phase was washed with aq Na2CO3, water (2 x), dried and evaporated.To the residue (58 g) was added AcOH (40 mL) and Ac2O (90 mL).The solution was chilled in an ice -salt bath and conc.H2SO4 (14 mL) was added dropwise under a blanket of argon and with manual swirling.The flask was closed with a rubber septum and left at ca. -5 °C for 10 h.TLC showed the compound 15 Rf 0.60 (hexane -EtOAc, 2:1) together with less and more polar byproducts.CH2Cl2 was added and the solution was transferred to a separatory funnel charged with ice and water and extraction was performed.The organic layer was washed with water (3 x), dried and evaporated.Addition of i-PrOH resulted in spontaneous crystallization.Filtration, washing with cold i-PrOH and prolonged drying on an oil pump furnished 15 (29.8 g, 45% cumulative yield), which was the best yield obtained.mp 130-133 °C (i-PrOH), [α]D 26 +43.1 (c 2, CHCl3); lit. 5 131-132 °C, [α]D +45.1 (c 1.32 CHCl3). 1 H (CDCl3): 8.01-7.32(15H), 6.44 (1H, s), 5.91 (1H, dd, J 5.0 Hz and 6.4 Hz), 5.79 (1H, d, J 4.8 Hz), 4.82-4.75(2H, unresolved), 4.52 (1H, dd, J 5.7 Hz and 13.9 Hz), 2.00 (3H, s). 13 C (CDCl3): 169.2, 166.1, 165.5, 165.2, 133.8, 133.7, 133.4,130.0, 129.9, 128.7, 128.6, 98.6, 80.2, 75.2, 71.6, 63.9, 21.0.

Figure 1 .
Figure 1.Stereochemistry of formation of nitroalcohols 21 and 22 according to the Felkin-Anh model (A) and according to the ref.66 (B).

Figure 1 .
Figure 1.Stereochemistry of formation of nitroalcohols 21 and 22 according to the Felkin-Anh model (A) and according to the ref.66 (B).