An integrated chemo-enzymatic route for preparation of β-thymidine, a key intermediate in the preparation of antiretrovirals

Article abstract of DOI:10.1021/op100208x

A chemo-enzymatic method for production of β-thymidine, an intermediate in the synthesis of antiretrovirals, is described. Guanosine and thymine were converted by means of enzymatic transglycosylation to yield 5-methyluridine (5-MU), which was reproducibly synthesised at a 10-20-L scale in 85% yield at a final product concentration of ～80 g?L^-1. A downstream processing (DSP) protocol was designed to remove reaction components interfering with the subsequent synthetic step. The crystallised 5-MU produced in the biocatalytic reaction was found to behave similarly to commercially available 5-MU, and the integration of the initial biocatalytic and subsequent three-step chemical process to β-thymidine was successfully demonstrated at bench scale.

Full text of DOI:10.1021/op100208x

Organic Process Research & Development 2011, 15, 258–265
An Integrated Chemo-enzymatic Route for Preparation of ꢀ-Thymidine, a Key
Intermediate in the Preparation of Antiretrovirals
Gregory E. R. Gordon,† Moira L. Bode,*,† Daniel F. Visser,† M. Jerry Lepuru,† Jacob G. Zeevaart,† Nasheen Ragubeer,†
Molala Ratsaka,† David R. Walwyn,‡ and Dean Brady†
CSIR Biosciences, Ardeer Road, Modderfontein, South Africa 1645, and ARVIR Technologies (Pty) Ltd. Postnet Suite 300,
PriVate Bag X30500, Houghton, South Africa 2041
Abstract:
Fermentation processes have been used for ꢀ-thymidine
production, but generally low thymidine concentrations are
achieved (0.5-7 g·L-1),5 resulting in additional processing costs
for product isolation. The low concentration of ꢀ-thymidine
produced may be attributed to tight metabolic control that is
difficult to overcome.6
A chemo-enzymatic method for production of ꢀ-thymidine, an
intermediate in the synthesis of antiretrovirals, is described.
Guanosine and thymine were converted by means of enzymatic
transglycosylation to yield 5-methyluridine (5-MU), which was
reproducibly synthesised at a 10-20-L scale in 85% yield at a
final product concentration of ∼80 g ·L-1. A downstream process-
ing (DSP) protocol was designed to remove reaction components
interfering with the subsequent synthetic step. The crystallised
5-MU produced in the biocatalytic reaction was found to behave
similarly to commercially available 5-MU, and the integration of
the initial biocatalytic and subsequent three-step chemical process
to ꢀ-thymidine was successfully demonstrated at bench scale.
Methods of nucleoside synthesis involving use of biocata-
lysts4 have also been developed. Mitsui Chemicals produce
deoxynucleosides by chemical conversion of D-glucose to
deoxyribose phosphate, followed by the enzyme-catalyzed
addition of the required base.3 Chemo-enzymatic routes allow
for the increased volumetric productivity of chemical methods
to be combined with the selectivity of biological methods.
Hori et al.7 published extensive work in which they applied
enzymes in the transglycosylation of inosine with thymine to
yield 5-methyluridine (5-MU), a compound that can be used
as an intermediate in the synthesis of ꢀ-thymidine.8 Unfortu-
nately, the reported yields and productivities were low. How-
ever, Ishii et al.9 (1989) showed that by using guanosine as the
ribose donor and whole cells of Erwinia carotoVora as a
biocatalyst, it was possible to produce 5-MU in a 74% yield at
increased starting substrate concentrations of 300 mM guanosine
[8.5 percentage mass guanosine per mass of reaction mixture
(% m ·m-1)], albeit over a 24-48 h period. More recently we
have demonstrated guanosine conversions of greater than 95%
and improved 5-MU yields of 85% with a volumetric produc-
tivity of up to 10 g ·L-1 ·h-1.10 From a commercial supply
perspective, guanosine has the benefit of being available cheaply
and in large quantities as it can be derived from disodium
Introduction
ꢀ-Thymidine is required in multitonne quantities as a
precursor to the anti-AIDS drugs stavudine (d4T)1 and zidovu-
dine (AZT).2 There are four broad approaches to nucleoside
production: extraction from natural sources, chemical synthesis,
fermentation, and biocatalysis.
ꢀ-Thymidine may be obtained from natural sources through
hydrolysis of DNA. Companies such as Yamasa Shoyu Co.
(Tokyo, Japan) and Reliable Biopharmaceuticals Corp. (St.
Louis, U.S.A.) have digested hundreds of tonnes of salmon milt
each year to isolate tonne quantities of individual nucleosides,
including ꢀ-thymidine. However, the use of natural resources
is inefficient as typically 100 tonnes of salmon only yields a
total of approximately 55 kg of the above deoxynucleosides in
roughly equal amounts.3
Traditional chemical methodologies for nucleoside produc-
tion suffer from numerous disadvantages, such as the use of
toxic metal reagents, like silver and tin, which are used to
activate substrates.4 In addition, often both the R and ꢀ anomers
are generated, and the desired product needs to be separated
from the mixture.
(5) (a) Lee, H. C.; Ahn, J. M.; Lee, S. N.; Kim, J. H. Biotechnol. Lett.
2004, 26, 265–268. (b) Anderson, D. M.; Liu, L.; Podkovyrov, S.;
Wang, B. WO/2001/02580 A1, 2001.
(6) Lee, H. C.; Kim, J. H.; Kim, J. S.; Jang, W.; Kim, S. Y. Appl. EnViron.
Microbiol. 2009, 75, 2423–2432.
(7) (a) Hori, N.; Watanabe, M.; Yamazaki, Y.; Mikami, Y. Agric. Biol.
Chem. 1989, 53, 197–202. (b) Hori, N.; Watanabe, M.; Yamazaki,
Y.; Mikami, Y. Agric. Biol. Chem 1989, 53, 2205–2210. (c) Hori, N.;
Watanabe, M.; Sunagawa, K.; Uehara, K.; Mikami, Y. J. Biotechnol.
1991, 17, 121–131. (d) Hori, N.; Watanabe, M.; Mikami, Y. Biocatal.
Biotrans. 1991, 4, 297–304. (e) Hori, N.; Uehara, K.; Mikami, Y.
Agric. Biol. Chem. 1991, 55, 1071–1074.
* Author to whom correspondence should be addressed. E-mail: mbode@
csir.co.za.
† CSIR Biosciences.
‡ ARVIR Technologies (Pty) Ltd.
(1) Becouarn, S.; Czernecki, S.; Valery, J.-M. Nucleosides Nucleotides
1995, 14, 1227–1232.
(8) Mansuri, M. M.; Starret, J. E., Jr.; Wos, J. A. J. Org. Chem. 1989,
54, 4780–4785.
(2) (a) Glinski, R. P.; Khan, M. S.; Kalamas, R. L.; Sporn, M. B. J. Org.
Chem. 1973, 38, 4299–4305. (b) Czernecki, S.; Valery, J.-M. Synthesis
1991, 239–240. (c) Kulig-Adamiak, A.; Bełdowicz, M.; Kamin´ski,
J.; Lewicka, L.; Ka¸kol, B.; Wis´niewska, A.; Kazimierczuk, Z. Przemysl
Chemiczny 2004, 83, 553–558.
(9) Ishii, M.; Shirae, H.; Yokozeki, K. Agric. Biol. Chem. 1989, 53, 3209–
3218.
(10) (a) Visser, D. F.; Bode, M. L.; Gordon, G. E. R.; Rashamuse, K. J.;
Hennessy, F.; Louw, M. E.; Brady, D. PCT/IB2008/054810, 2008.
(b) Visser, D. F.; Rashamuse, K. J.; Hennessy, F.; Gordon, G. E. R.;
van Zyl, P. J.; Mathiba, K.; Bode, M. L.; Brady, D. Biocatal. Biotrans.
2010, 28, 245–253. (c) Gordon, G. E. R.; Visser, D. F.; Bode, M. L.;
Raseroka, N.; Brady, D. Submitted to J. Biotechnol. 2010.
(3) Nagler, P.; Grayson, I. Chim. Oggi 2003, 21, 17–21.
(4) Wilson, L. J.; Hager, M. W.; El-Kattan, Y. A.; Liotta, D. C. Synthesis
1995, 12, 1465–1479.
258
•
Vol. 15, No. 1, 2011 / Organic Process Research & Development
10.1021/op100208x  2011 American Chemical Society
Published on Web 11/02/2010

Scheme 1. Chemo-enzymatic process for the preparation of ꢀ-thymidine
guanosine-5′-monophosphate which is manufactured in Japan
by yeast fermentation as a flavourant.11
and this then couples to thymine in a synthesis step to yield
5-MU. The reaction is catalysed by the enzymes purine
nucleoside phosphorylase (PNP) and uridine phosphorylase
(UP) for the phosphorolysis and synthesis steps, respectively.
This is a coupled two-step reaction performed in one pot
(Scheme 1). At laboratory scale the highest yield of 5-MU
attained was 85%, in a 1-L glass baffled reactor at a productivity
of 10 g ·L-1 ·h-1.10c The substrate thymine (4.7% m ·m-1) is
used in slight excess to the cosubstrate guanosine (9.3% m ·m-1)
in a mole ratio of 1.15:1, and the two enzymes required (PNP
and UP) were added at 2000 units each. Even at the reaction
temperature of 60 °C, this is a slurry reaction due to low
substrate solubility, and effective mixing is thus important.
However, during the small laboratory-scale optimisation of the
biocatalytic reaction, a crucial factor influencing the success of
the reaction was avoidance of excessive agitation.10c
On scale-up to 10 L, a glass-lined 16-L reactor with anchor
stirrer configuration (CR-16, Bu¨chi, Switzerland) was selected
to investigate whether this met the requirements. Reaction
reproducibility and robustness were evaluated by performing
three runs under identical conditions. The results obtained (Table
1) show that the biocatalytic reaction was both scalable and
reproducible with respect to guanosine conversion and 5-MU
yield.
Previously, we have defined the process-operating window
for the biocatalytic process using 1-L scale reactions.10c This
contribution describes the development of a kilogram-scale
chemo-enzymatic process to produce ꢀ-thymidine. This process
comprises a biocatalytic step to synthesise 5-methyluridine,
followed by chemical transformations to yield ꢀ-thymidine.
However, the integration of biocatalytic reactions into chemical
synthesis is a critical aspect in technology adoption.12 Hence,
the aim of the current study was to successfully scale the
biocatalytic reaction to 10-20 L and integrate this step with
the chemical process for the production of ꢀ-thymidine as
depicted in Scheme 1.
Results and Discussion
Biocatalytic Synthesis of 5-MU. The biocatalytic synthesis
of 5-MU was carried out by transglycosylation of guanosine
and thymine. In this reaction a ribose moiety is in effect
transferred from guanosine, a purine nucleoside, to thymine, a
pyrimidine nucleoside, to yield 5-MU and guanine. Ribose-1-
phosphate is generated in situ by phosphorolysis of guanosine,
(11) Demain, A. L.; Sanchez, S. In Fermentation microbiology and
biotechnology; El-Mansi, M., Bryce, C. F. A.; Eds.; CRC Press; Boca
Raton, FL, 2007; Chapter 4, p 117.
(12) Wohlgemuth, R. New Biotechnol. 2009, 25, 204–213.
Vol. 15, No. 1, 2011 / Organic Process Research & Development
•
259

Table 1. Biocatalytic reaction reproducibility
Subsequent removal of the acetyl protecting groups in the
presence of sodium methoxide gives the desired ꢀ-thymidine.
These reactions were all initially evaluated at small scale,
starting with commercially available 5-methyluridine and then
repeated with our biocatalytically produced 5-MU, before being
scaled to bench scale.
Investigation of Factors Influencing DABT Yield. A key
to the success of this process was integration of the biocatalytic
reaction and the subsequent chemical transformations. The first
step was to identify any factors negatively influencing the yield
of DABT from 5-MU using commercially available 5-methy-
luridine. This has direct bearing on the purity requirements of
the material entering the first chemical step and was used to
define the DSP for the biocatalytic reaction.
guanosine
conversion
(%)
mole
balance
(%)
normalised
5-MU yield
(%)a
5-MU
reaction
rxn 1
rxn 2
yield (%)
92.9
94.3
96.9
94.7
2.03
2.14
92.2
91.4
81.1
88.2
6.21
7.04
109
108
95.2
104
7.61
7.32
84.6
84.9
85.1
84.9
0.30
0.36
rxn 3
average (%)
SD
RSDb
a 5-MU yield normalised to
a mole balance of 100%. The reaction was
performed at 60 °C over 26 h. Reaction conditions: 9.3% m ·m-1 guanosine, and
4.7% m·m-1 thymine (a molar ratio of 1 to 1.15), 1600 U of each enzyme per
litre. b RSD: relative standard deviation, (SD/X) × 100.
The general process followed from 5-MU to DABT was
first described by Mansuri et al.8 This reaction proceeds Via an
anhydro intermediate and can result in the formation of the
byproducts thymine, 2′,3′,5′-tri-O-acetylmethyluridine (TAMU),
and 3′-O-acetyl-2′,5′-dibromothymidine (ADBT), as shown in
Scheme 2.13 One of the possible factors leading to byproduct
formation is the presence of water or alcohol which interferes
with acetylation at the 5′-position, allowing bromination to occur
at this position instead. On the basis of the proposed reaction
mechanism,13 in situ water formation is shown to take place,
and Mansuri et al.8 used an excess of acetyl bromide to 5-MU
over and above that required by reaction stoichiometry in order
to compensate for this. The formation of thymine due to acidic
cleavage of the nucleoside while undergoing acetylation has
also been reported.14
In order to define the specification of the 5-MU feed required
for reasonable DABT yields we experimented with the DABT
synthesis reaction conditions. Reaction parameters such as
reaction solvent and substrate concentration were altered, and
the effects of typical contaminants following biocatalytic
reactions, such as water and phosphate salts, were also evalu-
ated. The results of this investigation were used to define the
DSP requirements of the 5-MU produced in the biocatalytic
step (see below).
Figure 1. Guanosine conversion (2), 5-MU yield (]), and
productivity (9) on (a) a 10-L scale in a CR-16 reactor
(triplicate reactions) with anchor stirrer and (b) on a 20-L scale
using a CR-26 with a retreat blade impeller (single reaction).
Table 2 shows the yield of DABT obtained according to
the different conditions tested. The base case small-scale
reaction (experiment 1) used for these investigations, based on
the method of Mansuri et al.,8 was performed in acetonitrile at
2.5% m ·m-1 5-MU under reflux conditions using an excess of
acetyl bromide to 5-MU (mole ratio of >4.5 to 1). The solvent
and acetyl bromide were distilled from the mixture after
reaction, followed by partitioning between dichloromethane and
water. The workup was slightly modified compared to the
original procedure in that a saturated brine wash, instead of a
water wash, was applied to minimise transfer of organics into
the aqueous layer. A 92% yield was obtained using this
procedure. This yield is lower than that reported by Mansuri et
al.,8 but higher than that obtained by Shiragami et al.13 (48%),
who commented on their failure to reproduce the original yield
reported. Huang and Chu15 modified the method by introducing
acetic acid and HBr and were able to achieve a 63% yield.
However, the reaction completion (the point at which no
further conversion was observed) was delayed and took 10 h,
compared with 7 h at 1-L scale. As a result the 5-MU
productivity decreased to 7 g ·L-1 ·h-1 (Figure 1a). Lab-scale
studies (1 L) conducted using an anchor stirrer with and without
baffles showed that proper mixing under low-shear conditions
was crucial for optimum reaction performance (results not
shown). It was therefore anticipated that the slightly lower
productivity could be compensated for by reactor engineering
and agitation optimisation. Subsequent replacement of the
anchor stirrer with a retreat blade impeller in a 20-L reaction
(CR-26 Bu¨chi, Switzerland) returned the productivity to ap-
proximately 10 g ·L-1 ·h-1 (Figure 1b).
Chemical Synthesis of ꢀ-Thymidine from 5-MU. Synthesis
of DABT from 5-MU. The first step in the conversion of 5-MU
to ꢀ-thymidine is reaction with acetyl bromide to produce 3′,5′-
di-O-acetyl-2′-bromothymidine (DABT, Scheme 1). Conversion
of DABT into 3′,5′-di-O-acetylthymidine (DAT) is achieved
by hydrogenolysis catalysed by sponge nickel in methanol.
(13) Shiragami, H.; Ineyama, T.; Uchida, Y.; Izawa, K. Nucleosides
Nucleotides 1996, 15, 47–58.
(14) Bera´nek, J.; Hrebabecky, H. Nucleic Acids Res. 1976, 3, 1387–1400.
(15) Huang, H.; Chu, C. K. Synth. Commun. 1990, 20, 1039–1046.
260
•
Vol. 15, No. 1, 2011 / Organic Process Research & Development

Scheme 2. Preparation of DABT from 5-MU
Table 2. Factors influencing DABT yield
necessary for the 5-MU produced in the biocatalytic reaction.
The results show the anticipated decrease in DABT yield to
73%.
DABT
yield (%)
experiment
parameter altered
Phosphate is used in catalytic quantities in the biocatalytic
reaction, where it also functions as a buffer. A synthetic mixture
of 5-MU and phosphate buffer salts (roughly representing a
typical biocatalytic reaction process stream) was prepared and
subjected to the standard reaction conditions (experiment 5).
The purpose of this experiment was to determine if the presence
of phosphate salts had any detrimental effect on reaction
performance with respect to DABT yield, and if separation of
5-MU from salts was required. The results showed that the
presence of phosphate salts had an adverse effect on the reaction
performance, as the amount of DABT produced was lower
(52%) and the formation of a major byproduct was observed.
This reaction is particularly dilute and hence has a very low
volumetric productivity. Therefore, the concentration of 5-MU
in the reaction was increased from 2.5% m ·m-1 to 7.5% m ·m-1
in order to increase reactor productivity. Experiment 6 clearly
shows that the increased concentration had a deleterious effect
on DABT yield, with only 60% product being obtained.
Finally, 5-methyluridine produced in the biocatalytic reaction
was used directly without purification in order to test the effect
on DABT yield. Results from Table 2 suggested that the
presence of water had a deleterious effect on DABT yield, and
thus the material was thoroughly dried prior to use. HPLC
analysis showed the purity of 5-MU to be 80% m ·m-1, with
the rest of the mass being attributable to inorganic salts, thymine,
guanosine, and guanine. During addition of acetyl bromide to
the reaction, the mixture turned very dark and remained turbid
throughout the reaction. This was in contrast to reactions on
pure 5-MU, where all solids dissolved during addition of acetyl
bromide and the reaction mixture turned beige. Other than this,
by HPLC monitoring, the course of the reaction was identical
to that of the pure material. However, subsequent separation
of the organic and aqueous layers during reaction workup
proved difficult as salts precipitated at the solvent interface. On
a small scale this problem could be ameliorated to some extent
1
2
3
base case reaction
quench and neutralisation
reaction solvent
(ethyl acetate)
water addition
phosphate addition
substrate conc. increased
to 7.5% m ·m-1
92
84
59
4
5
6
73
52
60
Previous reports had indicated that loss of DABT by
cleavage to thymine was possible during lengthy exposure to
acidic conditions.14 Thus, in experiment 2, in order to minimise
exposure to acidic conditions, the reaction mixture was neutral-
ized using solid sodium bicarbonate after initial quenching of
excess acetyl bromide by the slow addition of methanol. The
reaction mixture was then filtered to remove solids, and solvent
was removed under vacuum. The crude organic residue was
dissolved in dichloromethane (DCM) and washed with water
to remove inorganic salts. The quench procedure led to a
reduced yield of 84%. As no thymine was detected by HPLC,
either with or without neutralisation, this step was not deemed
necessary.
The use of a water-miscible solvent, such as acetonitrile, as
reaction solvent increased the number of handling steps required
for product isolation, and thus replacement with ethyl acetate
was tested (experiment 3). Unfortunately, this resulted in
increased byproduct formation, with an unidentified product and
DABT being present in roughly equal quantities.
As discussed above, it has been reported13 that the presence
of water or an alcohol has an adverse effect on DABT
formation, by preventing acetylation of the 5′-position and
resulting in ADBT formation (see Scheme 2). Experiment 4
was therefore conducted in the presence of excess water (8.3%
m·m-1 water on 5-MU) in order to investigate its influence on
byproduct formation. This also has a bearing on the dryness
Vol. 15, No. 1, 2011 / Organic Process Research & Development
•
261

Figure 3. Solubility of 5-MU (9) or ([) thymine in isobutanol.
Figure 2. Solubility of 5-MU (9), guanosine (2), thymine ([),
and guanine (0) in phosphate buffer as a function of temperature.
of reaction components in phosphate buffer versus isobutanol
(Figures 2 and 3). Isobutanol has low water miscibility, while
also providing the advantage of relative ease of drying compared
to water. Hence, residual 5-MU could be recovered from the
crystallisation liquor by removal of water, followed by recrys-
tallisation from isobutanol in order to eliminate inorganic salts.
The final DSP, as depicted in Figure 4, gave 84% recovery
of 5-MU at a purity of >90% m ·m-1. Losses of 5-MU observed
during the guanine recovery could potentially be avoided
through further washing of the guanine filter cake, where 10%
of the 5-MU is lost. This would increase the overall 5-MU
recovery to 94%. It is anticipated that further optimisation of
the water crystallisation step and mother liquor recycling to
recover 5-MU would decrease the proportion of material
requiring isobutanol purification.
Integration of Biocatalytic Step into the Multistep Bench-
Scale Synthesis of ꢀ-Thymidine. There is a growing applica-
tion of biocatalysis in the synthesis of pharmaceutical com-
pounds and intermediates.16 Crucial to the success of these
processes is successful integration with the chemical steps.
5-MU ex. biocatalytic reaction and purified according to the
method shown in Figure 4 was used to evaluate the DABT
yield at small scale on biocatalytically produced material. The
results obtained showed the 5-MU to behave identically to
commercially available material. The chemical steps were
subsequently scaled up, and the bench-scale preparation of
ꢀ-thymidine from 5-methyluridine produced by biocatalysis was
successfully demonstrated. 5-MU ranging in purity from 87 to
95% m ·m-1 purity was converted into ꢀ-thymidine in good
yield. In general, yields obtained were similar to those obtained
at laboratory scale, indicating no significant problems with scale-
up (Table 3). The only significant variation occurred in the
preparation of DABT, where we achieved a DABT yield of
92% in small-scale reactions, and only 78% at bench scale. It
is considered likely that this reduction in yield at bench scale
was the result of loss of material to the aqueous layer during
DSP, mediated by water-miscible acetonitrile still present in
the product residue. Thymidine recovery was not optimal and
recoveries of 70-75% were obtained after collection of a single
crop of crystals, with no recycling of the mother liquor.
However, based on solubility data of thymidine in methanol,
thymidine recovery can be significantly improved to 90% with
recycling of the mother liquor. The overall yield of ꢀ-thymidine
from guanosine was 69% for laboratory scale and 56% for
by the use of a 50% brine solution, rather than saturated brine,
but on scale-up this separation would present difficulties.
Therefore, on balance, it was recommended that during DSP
the 5-methyluridine be purified of salts before use in the first
chemical step. On the basis of results from Table 2 and the
above result, conditions chosen for scale-up were a 2.5% m ·m-1
concentration of 5-MU in acetonitrile, with no quenching or
neutralisation of the reaction mixture. In addition, it was
determined that the 5-MU entering the reaction should be
thoroughly dried and be free of salts.
Down-Stream Processing of 5-MU. Integration of the
5-MU synthesis reaction with later synthetic steps required
removal of interfering reaction components after the biocatalytic
step. Since the presence of water impacted negatively on DABT
yield (Table 2), the 5-MU prepared in the biocatalytic reaction
for use in the subsequent ꢀ-thymidine preparation was required
to be free of moisture, and removal of phosphate salts was also
considered essential.
The first DSP step involved the removal of guanine. As a
result of the low guanine solubility at any temperature (Figure
2), this component could be removed by filtration. Hence, on
reaction completion the reaction mixture was heated to 90 °C
to maximise solubility of all other reaction components and was
filtered hot to recover guanine as a byproduct.
Although any guanosine present would undergo acetylation
during the preparation of DABT, the biocatalytic conversion
of guanosine was ∼95%, and guanosine solubility was low;
hence, a negligible amount of guanosine was measurable in the
process stream after filtration. It was anticipated that guanosine
levels less than 0.1% m ·m-1 would not have an adverse effect
on the ꢀ-thymidine process.
Finally, removal of thymine was desirable to enhance 5-MU
purity. Removal of unreacted thymine could be achieved by
cold filtration at 4 °C, as 5-MU remains in solution at this
temperature for short periods. Residual thymine at low levels
(0.2% m ·m-1), could be effectively removed during the brine
wash step of the DABT DSP. Thereafter, recovery of 5-MU
was possible by maintaining the reaction mixture at low
temperature to allow crystallisation from this aqueous solution.
However, a significant proportion of the 5-MU was not
recovered in this step and remained in solution. A number of
different solvents were tested for their suitability to separate
the residual 5-MU from inorganic salts. Isobutanol was identi-
fied as the most promising on the basis of the relative solubilities
(16) Pollard, D. J.; Woodley, J. M. Trends Biotechnol. 2007, 25, 66–73.
262
•
Vol. 15, No. 1, 2011 / Organic Process Research & Development

Figure 4. Isolation and recovery of 5-MU and guanine.
Table 3. Yields obtained in the preparation of ꢀ-thymidine
from 5-MU
effective overall synthesis of ꢀ-thymidine to be demonstrated
at bench scale.
lab-scale yield
(%) ex. biocat
bench-scale yield
(%) ex. biocat
Conclusions
reactions
Integration of biocatalytic 5-MU production at bench scale
with preparation of an advanced pharmaceutical intermediate
such as ꢀ-thymidine has been successfully demonstrated for
the first time. Guanosine conversions of 95% and 5-MU reaction
yields of 85% were achieved at an average 5-MU productivity
of 10 g ·L-1 ·h-1. 5-MU from the biocatalytic reaction was
isolated and purified using a combination of filtration, crystal-
lisation, evaporation, and isobutanol extraction and was recov-
ered in 84% yield at a purity of 90% m ·m-1. The material was
demonstrated to be of a suitable quality for the subsequent
preparation of ꢀ-thymidine without adversely affecting the
chemical transformations.
5-MU
90
92
90
92
85
78
DABT
DAT
ꢀ-thymidine
85
100
bench scale, not taking into account recoveries of 5-MU and
thymidine, as these were not optimised. Working on a recovery
of 90% for each of these products, the overall yield becomes
56% for laboratory scale and 45% for bench scale. This
compares favorably to the yield of 28% obtained for a chemical
method from tetra-O-acetylxylofuranose at small scale.17
The use of a biocatalytic step ensured formation of only the
ꢀ-anomer of 5-MU, thereby maximising the overall yield and
avoiding a complicated isolation step. This enabled a cost-
Experimental Section
General Methods. Guanosine and guanine standards were
purchased from Sigma-Aldrich (Germany), while thymine and
(17) Alla, V. R. R.; Mukund, K. G.; Sista, V. S. L. U.S. Patent 5,596,087,
1997.
Vol. 15, No. 1, 2011 / Organic Process Research & Development
•
263

5-methyluridine standards were purchased from NSTU Chemi-
cals (China). These compounds were quantitatively analysed
by HPLC, using a Waters Alliance model 2609 instrument with
a Synergi 4 µm Max-RP 150 mm × 4.6 mm column and a
UV detector at 260 nm. Samples were prepared by dissolving
the required amount of sample in sodium hydroxide (10 M, 1
mL) and then made up to the required volume using deionised
water so as to ensure the sample concentration was within the
linear region of the calibration curve. The injection volume was
10 µL, and elution was achieved with ammonium acetate, (25
mM, pH 4.0) at a flow rate 1 mL ·min-1 and at ambient
temperature.
Analytical standards of DABT8 and DAT17 were prepared
and used for quantitative analysis and for determining product
yields. ꢀ-Thymidine was purchased from NSTU Chemicals
(China) for use as a standard. HPLC analysis of these
compounds was carried out on a Hewlett-Packard series 1100
instrument using a 5 µm Waters Spherisorb ODS2 4.6 mm
×250 mm column under a gradient elution program at a flow
rate of 1 mL ·min-1. The UV detector was set to 254 nm.
Gradient elution was performed, with ammonium acetate (50
mM) decreasing from 90 to 20% as methanol increased from
10 to 80%. Retention times were 5-MU 5 min, thymidine 6.5
min, DAT 8 min, and DABT 10 min. NMR spectra were
recorded using a 400 MHz Varian INOVA instrument.
Small-Scale Synthesis. 3′,5′-Diacetyl-2′-bromothymidine
(DABT). To 5-methyluridine (10.0 g, 39 mmol), in a three-
necked round-bottomed flask fitted with a thermometer, reflux
condenser, and dropping funnel, was added acetonitrile (389
g). This slurry was heated to reflux, and 5.7 equiv of acetyl
bromide (27.32 g, 222 mmol) was added dropwise over 30 min.
During the course of addition, the solids dissolved to leave a
yellow solution. The reaction was heated for an additional 30
min, and then solvent was removed under reduced pressure
through two NaOH traps to remove HBr. The residue was taken
up in dichloromethane, and this was washed twice with saturated
brine. The organic layer was dried over MgSO4 (2 g), and
filtered, and the solvent was removed to yield a brown residue
(16.9 g) containing DABT (14.6 g, 92% yield, as adjudged by
HPLC).
1 h. After this time additional MeOH (119 g) and Amberlite
resin (IR-120, 30 g, Rohm and Haas, U.S.A.) were added and
stirred at room temperature for 30 min. The beads were filtered
off and washed with MeOH (3 × 40 g). Solvent was removed
in Vacuo from the reaction mixture and combined methanol
washings to yield a white solid containing thymidine (5.1 g,
92% by HPLC). Thymidine was purified by a single recrys-
tallisation from MeOH, with a recovery of 75%.
Bench-Scale Biocatalytic 5-MU Synthesis. Substrate Prepa-
ration. Guanosine, 9.3% m ·m-1 (1040 g, 3.67 mol), and 50
mM sodium phosphate buffer solution (9 600 g, pH 7.5) were
charged to a 16-L reactor (CR-16 Bu¨chi, Switzerland) while
stirring at 100 rpm (anchor stirrer). To this was added 4.8%
m·m-1 thymine (535 g, 4.24 mol) as a prewetted paste with
phosphate buffer. The reactor was then heated to 60 °C, while
stirring, and the reaction mixture was sampled (in triplicate,
1-3 g) prior to enzyme addition for determination of initial
concentrations. A portion of phosphate buffer solution (300 g,
50 mM, pH 7.5) was retained and used to dissolve and
quantitatively transfer the enzymes to the reactor. The reaction
mixture was maintained at 60 °C, while stirring at 100 rpm
and then sampled at intervals.
Enzyme Preparation. Lyophilized powders of enzyme,
purine nucleoside phosphorylase (PNP) ex Bacillus halodurans
overexpressed in Escherichia coli and uridine phosphorylase
(UP) ex E. coli overexpressed in E. coli, were produced in-
house according to our procedures as described by Visser et
al.10a,18 and stored at 4 °C. Approximately 16 kU each of PNP
(3.114 g, at a specific activity 5.14 U ·mg-1) and UP (3.758 g,
at a specific activity 4.27 U ·mg-1) were then dissolved in 100
mL of phosphate buffer prior to addition to the reaction mixture.
Isolation and Recovery of 5-MU (DSP). Isolation and
RecoVery of Guanine (Byproduct). Immediately after biocata-
lytic reaction (at 60 °C) the mixture was heated to 80-90 °C
and then filtered hot to remove guanine. The hot reaction
mixture was centrifuged at 80 °C, using a basket centrifuge
(10-µm mesh filter cloth, 1000 rpm, 30 min). The filter cake
was washed or slurried in hot deionised water (90 °C, 2-5 L)
and dried at 1500 rpm for 20 min. The filter-cake was initially
air-dried overnight in a fume-cupboard and then dried further
in a vacuum oven at 55 °C. From this process 464 g of guanine,
with a purity of 90% m ·m-1, was obtained.
5-MU Crystallisation. The filtrate (ex hot filtration) was
allowed to cool down from 80 to 30 °C over 4 h, followed by
cooling down from 30 to 4 °C over 2 h. The suspension was
then cold filtered to remove thymine. The reaction mixture was
then maintained at 4 °C for 24-48 h. During the cooling
process, the solution was stirred at 200-300 rpm using an
anchor stirrer. The 5-MU crystallised and was recovered by
conducting a cold filtration at 4 °C. The filter-cake was placed
in a fume cupboard, air-dried overnight, and then dried further
in a vacuum oven at 55 °C until dry. This crystallization yielded
486 g of 5-MU (a recovery of 57%), at an average purity of
between 88% m ·m-1 and 95% m ·m-1.
3′,5′-Diacetylthymidine (DAT). Without further purification,
the 16.9-g residue obtained above containing DABT (14.6 g,
36 mmol) was dissolved in MeOH (63.4 g) and transferred to
a Parr reactor. Sodium bicarbonate (4.6 g, 1.5 equiv) and Ni
catalyst A-5000 (4.5 g, ∼15% m ·m-1 catalyst loading, assume
50% wet) were added. The reaction was warmed to 35 °C, and
the H2 pressure was set at 9 bar. After 30 min the temperature
was raised to 40 °C for the remainder of the reaction. After 2 h
the reaction mixture was filtered through Celite 535 to remove
catalyst. Solvent was removed in Vacuo, and the residue was
taken up in ethyl acetate (90 mL). Precipitated salts were
removed from the ethyl acetate solution by filtration, after which
solvent was removed in Vacuo to yield a brown residue (12.7
g) containing DAT (10.6 g, 90% yield, by HPLC).
5-MU Crystallisation from Isobutanol. Residual 5-MU after
crystallisation could be recovered from the mother liquor. Water
ꢀ-Thymidine. Without further purification, DAT prepared as
described above (9.3 g residue containing 7.4 g DAT, 22.7
mmol) was dissolved in MeOH (29.2 g), and NaOMe (1.4 g,
1.1 equiv) was added. The reaction was warmed at 40 °C for
(18) Visser, D. F.; Hennessy, F.; Rashamuse, K.; Louw, M.; Brady, D.
Extremophiles 2010, 14, 185–192.
264
•
Vol. 15, No. 1, 2011 / Organic Process Research & Development

was removed using a Bu¨chi evaporator by gradually heating at
60-80 °C for 8-10 h. The crude 5-MU solid recovered was
dried under vacuum. The crude 5-MU was dissolved in
isobutanol (3 g 5-MU per 100 g of isobutanol) by heating the
slurry to reflux (110 °C) for 2-3 h in a 15-L glass reactor (GR-
15, Bu¨chi, Switzerland). The slurry mixture was then filtered
hot to remove the inorganic salts and insoluble organics such
as thymine. The filtrate containing 5-MU was then allowed to
crystallise from isobutanol, and 5-MU was recovered by
carrying out a cold filtration at 4 °C. The material was dried
further in a vacuum oven at 55 °C to remove residual solvent.
The material was then resuspended in water and recrystallised.
During this crystallization 230 g of 5-MU was recovered, which
corresponds to 27% of the total 5-MU produced, at an average
purity of between 88% m ·m-1 and 95% m ·m-1.
starting material remained. The reaction mixture was removed
from the Parr reactor under positive pressure and filtered through
two layers of Whatman No. 1 filter paper to retain the catalyst.
Solvent was removed by distillation at 60 °C under reduced
pressure in a 20-L rotary evaporator under rotation. To this
residue was added ethyl acetate (7.7 kg), and once the residue
had dissolved, MgSO4 (120 g) was added for drying. The salts
were removed by filtration through Whatman no. 1 filter paper,
and a small amount of ethyl acetate was used to rinse the
precipitated material. Solvent was removed on a 20-L rotary
evaporator under reduced pressure to give a brown residue (908
g) containing DAT (628 g, 85% yield by HPLC).
ꢀ-Thymidine. Brown residue (839 g) containing DAT (562
g, 1.72 mol) obtained from hydrogenolysis was dissolved in
MeOH (2.3 kg) in a 10-L rotary evaporator flask, and to this
was added sodium methoxide (100 g, 1 equiv). The reaction
was warmed while rotating the 20-L rotary evaporator at 44
°C for 1 h at atmospheric pressure, after which time HPLC
analysis showed the reaction to be complete, with 100%
conversion. Methanol (3.2 kg) was added, followed by 1 kg of
Amberlite IR-120 resin (Rohm and Haas, U.S.A.) to adjust the
alkaline reaction solution, and the resultant pH was found to
be 5. The beads were removed by filtration through Whatman
no. 1 filter paper and washed three times with methanol (total
of 2.3 kg). The solvent fractions were combined, and the solvent
was removed under vacuum on a 20-L rotary evaporator to
leave a residue of mass 778 g, containing 417 g thymidine
(100% yield) by HPLC. To this residue was added MeOH (3224
g), and the mixture was transferred to a glass reactor (GR-15)
and heated for 2 h. The solution was cooled while stirring, and
the slurry was transferred to a vessel fitted with an overhead
stirrer, stirred at 4 °C overnight, and then filtered and washed
with ice-cold MeOH. The recovered solid was dried under
vacuum to give a white powdery material (295.6 g, 71%
recovery). Analysis showed the material to have a purity of
99% m ·m-1 by HPLC and a melting point of 185-186 °C.
Proton NMR spectra and melting point are in agreement with
previously reported data for ꢀ-thymidine.19,20
Bench-Scale Synthesis. 3′,5′-Diacetyl-2′-bromothymidine
(DABT). Acetonitrile (13.9 kg) was transferred into a 26-L glass-
lined reactor (CR-26, Bu¨chi, Switzerland), and stirring com-
menced. 5-MU (424 g at 90% purity, 1.48 mol) in solid form,
produced in the biocatalytic reaction described above, was added
to the reactor, and the reaction was heated to 75 °C. A peristaltic
pump, primed with acetonitrile (0.5 kg), was used for acetyl
bromide transfer. Acetyl bromide (1.042 kg, 8.48 mol) was
added over a period of 0.5 h. After addition, the pump lines
were rinsed with acetonitrile (0.5 kg). The reaction was sampled
30 min after addition of acetyl bromide, and HPLC showed
the reaction to be complete with no residual 5-MU detectable.
Solvent was removed by distillation to leave a brown residue
inside the reactor. Dichloromethane (5.8 kg) was added,
followed by a 50% brine solution (5.87 kg). The two layers
were stirred for 10 min at 100 rpm and then transferred into a
separating vessel. The layers were allowed to stand for 45 min
before separation. The organic layer was transferred back to
the reactor and an additional portion of 50% brine was added
(5.87 kg). This was stirred for 10 min at 100 rpm and then
transferred into a separating vessel, where separation was carried
out after 15 min. Dichloromethane (983 g) was used to rinse
the reactor, and this was added to the combined brine layers
and allowed to stand overnight before separation. All organic
layers were combined, and the solvent was removed in Vacuo
to leave a beige-coloured foam (575 g) containing DABT (468
g, 78% yield by HPLC). Multiple batches of DABT were
combined to provide material for the subsequent reaction.
3′,5′-Diacetylthymidine (DAT). To DABT prepared as de-
scribed above (1167 g of residue containing 914 g DABT, 2.26
mol) was added methanol (3.78 kg), and the residue was
dissolved by mixing the material gently at 40 °C. Material was
transferred to an 8-L Parr reactor. Sodium bicarbonate (270 g,
1.4 equiv) and Ni catalyst A-5000 (260 g, ∼15% m ·m-1)
(Johnson Matthey Catalysts, TN, U.S.A.) were added to the
reaction. The reaction was warmed to 35 °C while stirring at
1000 rpm. When the reaction reached temperature, the H2
pressure inside the Parr reactor was set to 9 bar (131 psi). The
reaction was sampled hourly and stopped after 2 h, when no
Acknowledgment
We thank Prof. B. Zeelie (NMMU), Dr J. Dudaz and S. F.
Marais (CSIR) for help and guidance, Mr. K. Mathiba and Mrs.
N. Raseroka for conducting HPLC analysis. We gratefully
acknowledge the financial support, and resources from CSIR,
LIFElab and Arvir Technologies in funding this project, without
which this work would not have been possible.
Received for review July 30, 2010.
OP100208X
(19) Spectral database for organic compounds. SDBS No. 1178 (http://
riodb01.ibase.aist.go.jp/sdbs/ (National Institute of Advanced Industrial
Science and Technology, 07/10/2010).
(20) Reese, C. B.; Sanghvi, S. J. Chem. Soc., Chem. Commun 1983, 87,
7–879.
Vol. 15, No. 1, 2011 / Organic Process Research & Development
•
265