Synthesis of methyl 1-(2,3,5-tri-O-acetyl-β-L-ribofuranosyl)-1,2,4- triazole-3-carboxylate from L-ribose: From a laboratory procedure to a manufacturing process

Article abstract of DOI:10.1021/op050051m

A two-step manufacturing process for methyl 1-(2,3,5-tri-O-acetyl-β-L- ribofuranosyl)-1,2,4-triazole-3-carboxylate (1) was developed. In step 1, L-ribose was converted to a β/α mixture of 1,2,3,5-tetra-O-acetyl-L- ribofuranoses (2 and 4). The step contained four chemical transformations and was completed in "one-pot" in approximately 95% yield. The crude step 1 product was reacted with methyl 1,2,4-triazole-3-carboxylate (3) in step 2 to produce 1. The successful utilization of both isomers (2 and 4) in step 2 offered advantages of higher overall yield and a much simplified process by eliminating the isolation of pure 2. The process was successfully scaled up to the pilot plant and subsequently in a manufacturing campaign using commercial production facilities.

Full text of DOI:10.1021/op050051m

Organic Process Research & Development 2005, 9, 583−592
Synthesis of Methyl 1-(2,3,5-Tri-O-acetyl-â-L-ribofuranosyl)-1,2,4-
triazole-3-carboxylate from L-Ribose: From a Laboratory Procedure to a
Manufacturing Process
Pingsheng Zhang,* Zhiming E. Dong, and Thomas P. Cleary
Pharmaceutical Process DeVelopment, Roche Carolina Inc., 6173 East Old Marion Highway,
Florence, South Carolina 29506, U.S.A.
Scheme 1. Kilo-lab synthesis of 1
Abstract:
A two-step manufacturing process for methyl 1-(2,3,5-tri-O-
acetyl-â-
developed. In step 1,
of 1,2,3,5-tetra-O-acetyl-
L
-ribofuranosyl)-1,2,4-triazole-3-carboxylate (1) was
-ribose was converted to a â/r mixture
-ribofuranoses (2 and 4). The step
L
L
contained four chemical transformations and was completed
in “one-pot” in approximately 95% yield. The crude step 1
product was reacted with methyl 1,2,4-triazole-3-carboxylate
(3) in step 2 to produce 1. The successful utilization of both
isomers (2 and 4) in step 2 offered advantages of higher overall
yield and a much simplified process by eliminating the isolation
of pure 2. The process was successfully scaled up to the pilot
plant and subsequently in a manufacturing campaign using
commercial production facilities.
in the literature was tedious, containing 18 extractions and
8 distill-to-dryness operations, and was deemed not suitable
for scale-up. Part of the reason for the process being
cumbersome was the complex nature of the conversion. In
most solvents ribose exists as an equilibrium mixture of five
isomers: one acyclic form, two ribofuranoses, and two
ribopyranoses (Scheme 2, R ) H).3 In early studies, direct
acetylation of ribose under different conditions formed a
mixture of up to five products, with the ribopyranoses being
the major component in most cases.4 Therefore, to achieve
an efficient conversion of ribose to ribofuranoses, a stepwise
approach had to be adopted. It was reported5 that metha-
nolyses of most of the pentoses under acidic conditions
afforded the kinetically favored methyl furanosides first, and
then the furanosides gradually converted to the thermody-
namically favored pyranosides until an equilibrium was
reached (Scheme 2, R ) CH3). If the methanolysis was
conducted under mild conditions and the equilibrium was
stopped before significant amounts of the pyranosides were
formed, the furanosides could be obtained as the major
products. Based on this approach, a strategy to convert
D-ribose to 1,2,3,5-tetra-O-acetyl-â/R-D-ribofuranoses was
developed by Guthrie et al.6 As indicated in Scheme 3,
D-ribose was first converted to methyl ribofuranosides
(methanolysis) in methanol with a catalytic amount of strong
acid. The ribofuranosides were then converted to methyl
2,3,5-tri-O-acetyl-â/R-D-ribofuranosides (acetylation) by
reacting with acetic anhydride under basic conditions,
followed by acetolysis in acetic acid and acetic anhydride
to afford 1,2,3,5-tetra-O-acetyl-â/R-D-ribofuranoses. The
Introduction
Methyl 1-(2,3,5-tri-O-acetyl-â-L-ribofuranosyl)-1,2,4-tria-
zole-3-carboxylate (1) was an intermediate for the synthesis
of Levovirin,1 which was developed as an antiviral agent to
treat hepatitis C. In the early kilo-lab campaigns, 1 was
prepared from 1,2,3,5-tetra-O-acetyl-â-L-ribofuranose (2), as
illustrated in Scheme 1. Commercially available compound
2 was very expensive; therefore the cost of goods became a
serious issue for the commercial success of the drug. To
overcome this problem, we decided to introduce an in-house
production of 2 from considerably less expensive L-ribose.
Only one literature procedure was found for the conver-
sion of L-ribose to compound 2.2 The procedure described
* Corresponding author. Telephone: (843)-629-4241. Fax: (843)-629-4128.
E-mail: Pingsheng.zhang@roche.com.
(3) (a) Lichtenthaler, F. D.; Breunig, J.; Fischer, W. Tetrahedron Lett. 1971,
12, 2825. (b) Bishop, C. T.; Cooper, F. P. Can. J. Chem. 1963, 41, 2743.
(c) Smirnyagin, V.; Bishop, C. T. Can. J. Chem. 1968, 46, 3085.
(4) (a) Furneaux, R. H.; Rendle, P. M.; Sims, I. M. J. Chem. Soc., Perkin Trans.
1 2000, 2011. (b) Brown, G. B.; Davoll, J.; Lowy, B. A. Biochem. Prep.
1955, 4, 70. (c) Zhang, P.; Dong, Z. E. Unpublished results.
(5) (a) Ferrier, R. J.; Hatton, L. R. Carbohydr. Res. 1968, 6, 75. (b) Mowery,
D. F. J. Am. Chem. Soc. 1955, 77, 1667. (c) Bishop, C. T.; Cooper, F. P.
Can. J. Chem. 1961, 40, 224; also see: ref 4b and c.
(1) For biologic properties of Levovirin, see: (a) Fang, C.; Srivastava, P.; Lin,
C. J. Appl. Toxicol. 2003, 23, 453. (b) Lin, C.; Luu, T.; Lourenco, D.; Yeh,
L. T.; Lau, J. Y. N. J. Antimicrob. Chemother. 2003, 51, 93. (c) Watson, J.
Curr. Opin. InVest. Drugs 2002, 3, 680. (d) Tam, R. PCT Int. Appl. WO
01/046212. (e) Zeytinoglu, F. PCT Int. Appl. WO 01/045/642. (f) Lin, C.;
Lau, J. Y. N. J. Pharm. Biomed. Anal. 2002, 30, 239.
(2) (a) Ramasamy, K. S.; Tam, R. C.; Bard, J.; Averett, D. R. J. Med. Chem.
2000, 43, 1019. (b) Ramasamy, K. S.; Hills, L.; Tam, R.; Averett, D. U.S.
Patent 6,130,326, 2000.
(6) Guthrie, R. D.; Smith, S. C. Chem. Ind. (London) 1968, 547.
10.1021/op050051m CCC: $30.25 © 2005 American Chemical Society
Published on Web 08/04/2005
Vol. 9, No. 5, 2005 / Organic Process Research & Development
•
583

Scheme 2. Possible products of direct acetylation of L-ribose and methanolysis under acidic conditions
Scheme 3. Guthrie’s method of converting D-ribose to
tetra-O-acetylribofuranoses
diates so that the number of extractions and distillations could
be significantly reduced. We envisaged that if the acetylation
could be carried out under acidic conditions the overall
conversion could be done in “one-pot”. For instance, the
methanolysis could be conducted in methanol using sulfuric
acid as catalyst. Upon completion of the reaction the mixture
would contain methyl â/R-L-ribofuranosides (5), sulfuric acid,
methanol, and water. Addition of an excess amount of acetic
anhydride to the mixture would lead to the formation of
methyl 2,3,5-tri-O-acetyl-â/R-L-ribofuranosides (6) in a
medium containing sulfuric aid, acetic acid, and unreacted
acetic anhydride, favorable conditions for acetolysis. The
reaction could then be continued until the completion of
acetolysis. An example is shown in Scheme 5, in which
L-ribose was stirred with sulfuric acid in methanol at room
temperature overnight. To this mixture was slowly added
13 equiv of acetic anhydride. The mixture was then stirred
at room temperature for 3 h and 60 °C for 1 h. The reaction
generated a crude product containing 20% 2, 13% 4, 16%
1,2,3,4-tetra-O-acetyl-â-L-ribopyranose (7), and 33% acyclic
peracetate (8). Compound 7 could be formed via several
different pathways; however two conditions had to exist: one
was the opening of the five-member ring (5 and/or 6), and
the other was the presence of a free 5-hydroxy group. The
formation of 8 also required the ring opening of the
ribofuranoses. To minimize the formation of the two byprod-
ucts, it was necessary to carry out the overall conversion
under mild conditions (to prevent the ring opening) and to
acetylate the 5-hydroxy group before conducting acetolysis.
Based on this analysis several changes were made. First, upon
the completion of methanolysis, sodium acetate was added
to neutralize sulfuric acid and stop the acid-catalyzed
equilibria. Second, after neutralization the mixture was
concentrated to partially remove methanol. Acetic anhydride
was then added at ambient temperature, and the mixture was
heated at 100 °C for 2 h to complete the acetylation. Third,
after the mixture was stirred with sulfuric acid at room
temperature for 1 h, it was neutralized with sodium acetate.
Scheme 4. Overall development plan
same strategy was used in the literature synthesis of 2,2 in
which the methanolysis was carried out in methanol and HCl,
the acetylation was conducted in acetic anhydride and
pyridine, and the acetolysis was carried out in a mixture of
acetic anhydride and concentrated sulfuric acid in acetic acid
to afford a mixture of 2 and its R-anomer, 4. Finally, pure 2
was isolated in 57% yield via recrystallization from ethyl
ether.
Our development plan for the project included (1)
developing a more efficient process for the conversion of
L-ribose to 2/4 (Scheme 4, Step 1) starting from the
modification of the literature procedure and (2) exploring
the possibility of converting both 2 and 4 to 1 (Step 2), so
as to avoid isolating pure 2 and to improve the overall yield.
Establishment of the Basic Process for Step 1. The key
to streamlining the literature process was to eliminate the
unnecessary solvent exchanges and the isolation of interme-
584
•
Vol. 9, No. 5, 2005 / Organic Process Research & Development

Scheme 5. A “one-pot” acetylation of L-ribose
Scheme 6. New process for step 1
This reaction mixture was then concentrated and extracted
with dichloromethane to afford a crude product. This time
the crude product contained 66% 2, 24% 4, 1.3% 7, and 2.8%
8. The overall yield was 88%. Although the results were
encouraging, there was a safety concern in the acetylation
step. The reaction of methanol and the hydroxyl groups in 5
with acetic anhydride are strongly exothermic.7 After the
addition of acetic anhydride to the mixture at ambient
temperature, a gradual increase in batch temperature was
observed. The temperature increase accelerated quickly, and
strong external cooling had to be applied to keep the reaction
under control even on small scale batches in the lab. Two
measures were taken to address this problem. One was to
completely remove methanol from the system by solvent-
exchange to acetic acid before the addition of acetic
anhydride. The other was to slowly add acetic anhydride at
high temperature (e.g., 50-80 °C) so that there would be
no acetic anhydride accumulation in the system. After these
modifications a basic framework of the process had been
established, which is illustrated in Scheme 6. This version
of the process contained only one solvent-exchange opera-
tion, one concentration, and one extraction.
Optimization of Step 1. After the establishment of the
basic process for the conversion, the step was further
optimized. The methanolysis was carried out by stirring
L-ribose and methanol at ambient temperature in the presence
a of catalytic amount of concentrated sulfuric acid. The
reaction was an equilibrium between L-ribose/methanol and
5/water. With 0.1-0.2 equiv of sulfuric acid, the reaction
reached its equilibrium within 2 h. Experimental data
indicated that 20 ( 5 °C was the ideal temperature for the
reaction. At higher temperature the formation of methyl
ribopyranosides became significant. At lower temperature
it took a longer time for the reaction to reach equilibrium
while no improvement of conversion and stereoselectivity
was observed. The most effective way to push the equilib-
rium towards the product was to use more methanol. As
demonstrated in the examples in Table 1, when the methanol
was increased from 2 mL/g of L-ribose to 15 mL/g of
L-ribose, the percentage of product in the reaction mixture
Table 1. Concentration effect on the methanolysis
mL of MeOH/
g of L-ribose
GC% (area)
GC% (area)
L-ribose
entry
5
1
2
3
4
2
5
9
91.5-93.4
95.8-97.6
97.2-97.8
98.7
4.5-5.1
1.8-2.5
1.7-1.9
1.4
15
was improved from 92 to 93% to 99%. Ultimately, a
concentration of 5 mL of methanol/g of L-ribose was adopted
as a compromise of conversion and process capacity.
After the methanolysis reached its equilibrium, a base was
added to bring the pH of the mixture to above 4, at which
the equilibrium was halted. Several bases were evaluated for
this purpose. For the neutralization per se any base could be
used, however the choice of base could affect the subsequent
transformations and operations. For instance, when sodium
acetate was used the subsequent acetylation and acetolysis
went normally. The byproduct (Na2SO4), however, has very
low water solubility (4.8 g/100 mL8) which required the use
of large quantities of water during the extraction and resulted
in significant loss of process capacity. Ammonium sulfate
has very good water solubility (71 g/100 mL8). However,
when NH4OAc was tested for the neutralization an intractable
mixture was produced during the acetylation. When pyridine
was tested the acetylation went faster and the acetolysis was
normal. Unfortunately, significant product (2/4) decomposi-
tion was observed during the final concentration of the
reaction mixture. Lithium sulfate has a water solubility of
about 26 g/100 mL,8 so both LiOAc and Li2CO3 were tested
for the neutralization. No adverse effects on the process were
observed in both cases. Lithium carbonate was chosen based
on cost consideration. It is worth noting that the particle size
of Li2CO3 was crucial for the neutralization. Fine powder
must be used. When granular Li2CO3 was tested in some of
lab batches, an intractable mixture was formed after acety-
(7) Reaction calorimetry data measured in our Risk Analysis Lab: ∆Hreaction
-173 kJ/mol, ∆Tad-rise ) 71 °C.
)
(8) Handbook of Chemistry and Physics, 70th ed.; Section B, Physical Constants
of Inorganic Compounds; CRC Press.
Vol. 9, No. 5, 2005 / Organic Process Research & Development
•
585

Scheme 7. Details on the acetylation reaction
Figure 2. Temperature effects on Acetylation.
Figure 1. Kinetics of acetylation at 60 °C.
lation. It was possible that the large particle size caused
inadequate neutralization, which in turn resulted in the
reverse of 5 to L-ribose during the solvent exchange.
During the acetylation compound 5 was converted to 6.
There are three different hydroxyl groups in 5. The acety-
lation should occur in a step-by-step fashion via mono- and
diacetylated intermediates (9, 10, and 11 in Scheme 7).
Monoacetylation, presumably at the primary hydroxyl group,
was fast. For instance, when treated with acetic anhydride
at 50 °C, 5 disappeared completely within 1 h, leading to
the formation of a mixture of 9, 10, 11, and 6. The conversion
of 9-11 to 6 was slower and required a higher temperature
and longer reaction time. Figure 1 illustrates a typical reaction
profile for the acetylation at 60 °C. At the beginning of the
reaction the mixture contained mainly the intermediates
9-11. As the reaction proceeded the intermediates were
gradually converted to the product. Based on this observation
the reaction rate was measured as the consumption of the
total intermediates (the sum of 9, 10, and 11).
Figure 3. Effect of acetic anhydride charge on acetylation at
100 °C.
anhydride were charged, acetylation was completed within
1 h. When 3.8 equiv of acetic anhydride was added, the
reaction was much slower. It took about 4 h for the
intermediates to drop to 0.27%. Reducing the acetic anhy-
dride charge to 3.7 equiv made the reaction even slower. It
took 4 h for the intermediate levels to drop to 0.59%, and
the reaction was completed (total consumption of the
intermediates) in 6 h. When 3.6 equiv of acetic anhydride
were charged, the reaction was not complete after 6 h. As
will be discussed later, to reduce the impurity level during
the acetolysis, it was necessary to tightly control the amount
of acetic anhydride in the reaction mixture. As a compromise,
3.7 equiv of acetic anhydride was charged for the acetylation
process.
Upon the completion of acetylation the reaction mixture
was cooled and treated with sulfuric acid to effect acetolysis.
With the isolation of pure 2 in mind, our initial focus was
how to increase the ratio of 2 to 4 in the reaction. It was
discovered that the ratio was temperature dependent and was
not affected by the â/R ratio of 6. Figure 4 illustrates the
ratio changes of â/R-6 and 2/4 during acetolysis at 0 °C.
The â/R ratio of 6 increased sharply at the beginning of the
reaction and then remained constant. The ratio of 2 to 4 was
around 7 and decreased steadily to about 3 in 6 h. The
following conclusions could be drawn from the observa-
The rate of the acetylation was affected by both reaction
temperature and the amount of acetic anhydride. As indicated
in Figure 2, when 5 equiv of acetic anhydride was charged,
the intermediate level dropped to 0.76% in 8 h and totally
disappeared in 10 h at 60 °C. At 80 °C, the reaction took
less than 4 h. When the temperature was increased to 100
°C, all the intermediates disappeared within 1 h, and no
deterioration of product purity was observed. Based on the
results, 100 °C was adopted for the process.
The effect of acetic anhydride on the acetylation at 100
°C was demonstrated in Figure 3. When 5 equiv of acetic
586
•
Vol. 9, No. 5, 2005 / Organic Process Research & Development

Table 2. Effects of temperature and sulfuric acid on
acetolysis
T,
°C
equiv of reaction GC% GC% GC%
entry
H2SO4
time, h
2
4
2 + 4
1
2
3
4
5
0
0
0
0.97
0.82
0.43
0.82
0.82
0.5
1
6
2
4
67
67
66
71
66
23
23
22
17
13
90
90
88
88
79
-13
-23
These hypotheses were confirmed in our lab, and some
results are listed in Table 2. For instance, the effect of sulfuric
acid on the reaction at 0 °C is demonstrated in entries 1 to
3. With 0.97 equiv of sulfuric acid, the reaction was
completed within 0.5 h and the system reached equilibrium
(ratio of 2 to 4 became constant) within 0.5 h to give about
67% of 2 and 23% of 4. With 0.82 equiv of sulfuric acid,
the reaction completed within 0.5 h but reached equilibrium
in 1 h. When 0.43 equiv of sulfuric acid was used, the
reaction completed in about 5 h and the system reached
equilibrium in about 6 h. The products from all the three
batches had the same composition and purity profiles. Based
on these results, 0.8 equiv of sulfuric acid was adopted for
the process.
Figure 4. â/r ratio changes during acetolysis.
Scheme 8. Proposed mechanism for acetolysis
Reaction temperature affected both reaction rate and â/R
ratio of the products. For instance, in the presence of 0.82
equiv of sulfuric acid, the reaction was completed in 1 h at
0 °C and the 2/4 ratio was 76/23 (entry 2). At -13 °C the
reaction was completed in about 2 h and the 2/4 ratio reached
71/17 (entry 4). At -23 °C the reaction took 4 h and the
2/4 ratio was increased to 66/13 (entry 5). However, at -23
°C higher amounts of impurities were formed which affected
the yield. Clearly, if pure 2 is to be isolated the acetolysis
should be carried out at low temperature (e.g., -13 °C).
However, if both isomers are to be utilized in the subsequent
steps, lower reaction temperature would not offer any benefit.
Using the optimized conditions up to this point the process
produced a crude product that contained about 90% (GC
area%) of 2/4. Two major impurities with close retention
times on GC accounted for most of the other 10%.
Experimental data indicated that (1) the impurity level was
proportional to the amount of acetic anhydride added to the
reaction mixture, as higher levels of the impurities were
observed when more acetic anhydride was used; (2) the
amount of sulfuric acid had no effect on the impurity level;
(3) when pure 2 was subjected to the acetolysis condition
the product was a mixture of 2 and 4, with none of the
impurities observed in the reaction mixture; and (4) when
pure â-6 was subjected to the acetolysis condition the
reaction mixture contained not only 2 and 4 but also the two
impurities. These results clearly indicated that the impurities
were formed from 6 during the acetolysis via a competing
mechanism. As indicated in Scheme 8, the conversion of 6
to 2/4 started with the protonation of the methoxy group,
followed by the leaving of a molecule of methanol to give
intermediates 12 and 13. A competing mechanism would be
the protonation of the oxygen in the five-member ring,
followed by the ring opening to give intermediates 14 and
15 (Scheme 9). Acetylation of the hydroxyl group in the
tions: (1) R-6 was more reactive than â-6. The fast
consumption of R-6 at the beginning of the reaction caused
the increase of the â/R ratio; (2) 2 was formed faster than 4;
(3) there was an equilibrium between 2 and 4, and part of
the newly formed 2 was gradually epimerized to 4. Based
on these observations, a mechanism for the acetolysis
reaction was proposed and is shown in Scheme 8. Multiple
equilibria existed in the reaction mixture, including the
equilibrium between the two anomers of 6, the equilibrium
between 2 and 4, and the equilibrium between 6 and 2/4.
The driving force of the reaction was the consumption of
methanol, which was released from 6 during the reaction,
by acetic anhydride. Both â/R-6 passed through the same
intermediates, 12 and 13, to form 2/4. As a result, the â/R
ratio of 6 should not have any effect on the ratio of 2 to 4.
The ratio of 2 to 4 should be determined by the equilibrium
between the two isomers. Therefore, the major factors that
would most likely affect the 2/4 ratio would be the reaction
temperature and reaction medium. The amount of the sulfuric
acid should affect only the reaction rate, but not the 2/4 ratio.
Vol. 9, No. 5, 2005 / Organic Process Research & Development
•
587

Scheme 9. Possible mechanism for the impurity formation
Table 3. Impurity level changes under different conditionsa
1st portion 2nd portion addition
GC%
area
18 + 19
of Ac2O
(equiv)
of Ac2O
(equiv)
temp
addition
time
entry
(°C)
1
2
3
4
5
6
7
5.1
3.7
3.7
3.7
3.7
3.7
3.7
0
5b
5
5
5
5
NA
5 min
1 h
2 h
3 h
2 h
2 h
6.9
5.7
3.7
3.0
2.9
0.7
0.4
1.4
1.4
1.4
1.4
1.4
1.4
20
25
a All the batches were carried out on a scale of 66.6 mmol of L-ribose. The
first portion of Ac2O was added at 80 °C. The mixture was heated to 100 °C,
held for 4 h, and then cooled to 5 °C. Sulfuric acid (0.82 equiv) was added,
followed by the addition of the second portion of Ac2O at the designated
temperature in a period of time as indicated in the table. b The temperature at
which sulfuric acid was added.
impurities. When 3.7 equiv of the anhydride was added for
the acetylation and then another 1.4 equiv was added in 5
min following the sulfuric acid charge, the impurity level
dropped to 5.7% (entry 2). It was interesting to note that the
addition rate of the second portion of the anhydride affected
the impurity level significantly. When the 1.4 equiv of acetic
anhydride was added slowly in 1 to 3 h, the impurity level
could be reduced to 2.9% (entries 3 to 5). It was also
discovered that the reaction temperature affected the impurity
level. As indicated in entries 5 and 6, the impurity level
decreased to 0.7% from 2.9% when the reaction temperature
was changed from 5 °C to 20 °C. Further increase of the
temperature to 25 °C resulted in lower overall yield due to
higher levels of other impurities. Based on these results, the
second portion of acetic anhydride was added slowly at 20
°C over at least 2 h in the process.
presence of acetic anhydride would lead to the formation of
16 and 17. Addition of acetic acid to 16 and 17, followed
by deprotonation would give compound 18 and 19. To
support this hypothesis, the two impurities were isolated and
identified by spectroscopy as 18 and 19. To further confirm
the structures, the two compounds were independently
synthesized using a literature procedure.9
Upon the completion of the acetolysis the reaction mixture
was transferred to a lithium carbonate solution. The mixture
was then concentrated and extracted. Several solvents
(dichloromethane, ethyl acetate, R,R,R-trifluorotoluene, and
toluene) were evaluated for this purpose, and dichlo-
romethane gave the best results. Two dichloromethane
extractions were adopted in the process, so as to efficiently
extract the product without sacrificing the process capacity.
After optimization of the step 1 process, the step 1 product
(a mixture of 2 and 4) could be obtained from L-ribose in
approximately 95% oVerall yield.
Establishment of the Step 2 Process and Its Optimiza-
tion. Two methods for the fusion of 2 with methyl 1,2,4-
triazole-5-carboxylate (3) (Scheme 1) were used in the early
kilo-lab campaigns. One was the ICN process2 in which the
reaction was carried out at 165-175 °C using bis(p-
nitrophenyl)phosphate as catalyst. The other was a process
developed in our Process Research lab for the production of
Ribavirin10 in which the reaction was conducted at 120-
130 °C in the presence of triflic acid as catalyst and the
product was isolated via crystallization from methanol. In
both processes only pure 2 was used. As mentioned above,
the current step 1 process produced a mixture of 2 and 4 in
After identification of the two impurities, the remaining
challenge was to minimize their formation. According to the
mechanism in Scheme 9, the driving force for the formation
of 18 and 19 was the acetylation of the 4-hydroxyl group
after the ring opening. Apparently, the presence of a high
concentration of acetic anhydride favored this path, which
was consistent with our observations. Lowering the acetic
anhydride concentration during the acetolysis seemed to be
a logical approach. In our “standard” procedure, the total
amount of acetic anhydride charged for both acetylation and
acetolysis was 5.5 equiv and the anhydride was added in
one portion. We found that the charge could be reduced to
5.0 equiv without affecting the reaction. However, the
reduction did not lead to a significant decrease of the impurity
level. To keep the concentration of acetic anhydride low
during the acetolysis, it was decided to add acetic anhydride
in two portions. In the first portion, enough anhydride (3.7
equiv) was added to effect the acetylation. Upon the
completion of the acetylation, sulfuric acid was added,
followed by the addition of the second portion of the
anhydride (1.4 equiv). Studies indicated that the addition
mode had an affect on the impurity level (Table 3). With a
total amount of 5.1 equiv of acetic anhydride used, the
process using a one portion addition generated 6.9% of total
(10) Ribavirin is a drug currently in the market for the treatment of hepatitis C
and is the enantiomer of Levovirin. For the synthesis of Ribavirin, see:
Witkowski, J. T.; Robin, R. K.; Sidwell, R. W.; Simon, L. N. J. Med. Chem.
1972, 15, 1150.
(9) Lichtenthaler, F. D.; Breunig, J.; Fischer, W. Tetrahedron Lett. 1971, 12,
2825.
588
•
Vol. 9, No. 5, 2005 / Organic Process Research & Development

Table 4. Effect of acetic anhydride charge on step 2
yield of
Ac2O
added
(equiv)
isolated
product
(%)
% of hydrolyzed
step 2 product
entry
1
2
3
0
2.66
1.12
0.69
79
81
86
0.9a
2.6a
a Relative to L-ribose.
due to the presence of water in the reaction system. We have
detected seven impurities derived from the hydrolysis of 1:
three “monohydrolyzed” (compounds with one of the three
acetyl groups in 1 removed), three “dihydrolyzed” (com-
pounds with two of the three acetyl groups removed), and
one “trihydrolyzed” product. Due to high solubility of these
impurities in methanol, they stayed in the mother liquor
during the crystallization. In some cases these impurities
accounted for up to 7% of yield losses. Therefore, drying
the dichloromethane solution of Step 1 product became
necessary. Some drying agents, such as sodium sulfate or
magnesium sulfate, could be used to dry the solution. While
adding and filtering the drying powder was not a problem
in the lab, it did raise concerns on the prospect of large-
scale production where handling large amounts of solids
might be problematic and disposal of large quantities of solid
wastes could be costly. Also, 1.5-3.5% of hydrolyzed
products were formed even after the step 1 solution was dried
with sodium sulfate or magnesium sulfate. Therefore, we
decided to evaluate other drying methods. One option was
azeotropic distillation using toluene, but this would add extra
operations to the process. Another option was to use a water-
scavenger to consume the water in situ, such as acetic
anhydride which would react with water to form acetic acid.
The acetic acid and the unreacted anhydride could then be
removed during the concentration step and before adding
triflic acid. One concern for this approach was the possibility
of the reaction of 3 with acetic anhydride. Thus, a mixture
of acetic anhydride and 3 in acetic acid was stirred at 57 °C
for 4 h, 85 °C for 1 h, and 115 °C for 2 h, and no reaction
was observed. After confirming that there was no cross
reaction between acetic anhydride and 3, the effect of acetic
anhydride on step 2 was studied. Some results are listed in
Table 4. In the experiments, a batch of step 1 solution was
divided into three portions. Each portion was mixed with
1.0 equiv of 3, and then two of the batches were treated
with 0.9 equiv and 2.6 equiv of acetic anhydride. All the
batches were carried through step 2. The total percentage of
the hydrolyzed step 2 products and the isolated yields were
compared, and the benefits of adding acetic anhydride were
obvious. For batches with added acetic anhydride (entry 2
and 3), the reactions were faster, the levels of hydrolyzed
step 2 product were lower, and the yield was higher.
Figure 5. Reaction rate of 2 and 4 as the consumption of 3.
a ratio of approximately 3:1. The use of pure 2 in the step
2 reaction would mean the loss of about 25% of the
expensive step 1 product and additional purification opera-
tions for the isolation of pure 2. It would be a significant
advantage in terms of production cost and efficiency if the
mixture of both anomers could be converted to the desired
product.
First, the reaction of pure 2 with 3 was tested using the
Ribavirin process. Thus, a mixture of 2, 1.0 equiv of 3, and
a catalytic amount of triflic acid (1% mol) in methyl acetate
was concentrated and heated at 120-130 °C for ap-
proximately 5 h under vacuum to complete the reaction. The
mixture was then cooled to 70 °C, and methanol was added.
The mixture was stirred until a homogeneous solution was
formed and then slowly cooled to 0 °C to precipitate 1 in
75% yield. To prove the concept that 4 could also be
converted to 1, a pure sample of 4 was isolated from a batch
of Step 1 product. The sample was treated with 3 and triflic
acid under the same condition as above. The reaction
proceeded normally, and product 1 was isolated in 70%
yield! The only difference observed for the two compounds
was that 2 reacted faster than 4, as illustrated in Figure 5.
Having confirmed that both 2 and 4 could be utilized in
Step 2, the crude Step 1 product was tested. When the crude
product was subjected to the same process conditions the
results were not consistent. In the experiments, compound 3
was added to the step 1 dichloromethane extract. The mixture
was concentrated to dryness, and then triflic acid was added
to start the reaction. The main problem was the difficulty in
initiating the reaction. In some cases much higher triflic acid
charges (up to 6 mol %) were needed to complete the
reaction. The use of a high level of triflic acid led to
significant yield loss and colorization of the product. It
seemed that there were some basic materials in the crude
step 1 solution that were consuming triflic acid. To solve
this problem, it was decided to wash the dichloromethane
solution of crude step 1 product with dilute acid (e.g., 5%
sulfuric acid) before starting step 2. Indeed, after the acid
wash the step 2 reaction could be completed within 5 h under
the same conditions in the presence of as low as 1 mol % of
triflic acid.
Ideally, the amount of acetic anhydride charged should
be just enough to consume all the water in the mixture. This
was difficult to achieve due to the fact that the water content
in the Step 1 extract varied from batch to batch. Excess
anhydride could be added; however the unreacted anhydride
In the early stage of the development it was found that
one source of yield loss in step 2 was the hydrolysis of 1
Vol. 9, No. 5, 2005 / Organic Process Research & Development
•
589

Figure 6. Effects of temperature and triflic acid on the reaction
rate.
Figure 7. Yield change and ideal parameter window for step
2.
must be distilled out as completely as possible before
charging triflic acid. Experimental data from additional
studies demonstrated that 37 g (or 0.56 equiv) of acetic
anhydride per batch of 100 g of L-ribose provided the best
results.
0.7 mol % of triflic acid), while the fastest reaction took
only 45 min (at 130 °C and 3.0 mol % of triflic acid).
The effects of the temperature and triflic acid on the yield
are illustrated in Figure 7.11 Higher isolated yield could be
obtained when the reaction was carried out at high temper-
ature and at low dose of triflic acid. If only the isolated yield
is considered when choosing the process parameters, the
reaction should be carried out at around 130 °C using 0.5
mol % triflic acid. However, it was observed during the
studies that the products were significantly darker in color
when the reactions were carried out at temperatures higher
than 125 °C or when a high dose of triflic acid was charged.
Based on Statistical Experimental Design, optimum results
for the process (such as reaction rate, product purity and
color, overall yield, etc.) were achieved under the following
conditions: temperature, 110-120 °C; triflic acid, 0.7-1.5
mol % (identified by the box in Figure 7). Above and to the
right of this area, the purity of the product deteriorated and
the product became darker in color. Below and to the left of
this area, the reaction rate became slow.
In the original Ribavirin process, methanol was used in
the crystallization of the step 2 product. Some concerns were
raised during the evaluation of the process in our lab. For
instance, at the target concentration (440 g crude step product
per liter methanol) the slurry was very thick. This might
cause operational difficulties during the transfer of the batch
from reactor to centrifuge due to pipeline blockage or wall-
cake formation inside the reactor. Increasing the methanol
charge would result in yield loss because of the high
solubility of the product in methanol (45 g/L at 20 °C). Also,
seeding was required to induce the initial precipitation. Thus,
ethanol and 2-propanol (IPA) were evaluated as candidates
to replace methanol in the process. The solubility of 1 in
IPA and ethanol is significantly lower than in methanol (5.5
g/L of IPA and 15 g/L of EtOH). The replacement of
methanol with IPA or ethanol would allow higher solvent
charge, and the slurry would not be as thick. Yield improve-
Due to the fact that the Step 1 product was not isolated,
the total amount of 2 and 4 in the step 1 extract was estimated
via wt % assay of the solution. The amount of 3 charged in
step 2 was then calculated based on the wt % assay. Ideally,
the amount charged should be enough to convert all 2 and 4
to product 1, but not so large an excess as to contaminate
the product with unreacted 3. Our experimental data indicated
that 1.0-1.05 equiv (relative to total amount of 2 and 4
calculated from the wt % assay) of 3 were ideal for the
process. Because the overall yield for step 1 was about 95%,
the amount of 3 added was about 0.95 equiv of L-ribose.
During the Manufacturing campaign the yield on step 1 was
very consistent from batch to batch. Assay analysis of the
step 1 extract was discontinued after the first several batches,
and 0.95 equiv (relative to L-ribose) of 3 was charged in
each batch throughout the campaign with good results.
At this point in development, a basic process for the step
had been established as follows: to a mixture of step 1
extract and 1.05 equiv of 3 was added 0.56 equiv of acetic
anhydride. The mixture was concentrated to remove all the
solvents. To this mixture was added 1.0-1.85 mol % of
triflic acid, and the batch was stirred at 110-130 °C under
20-65 mbar until the reaction was complete. The mixture
was then cooled to around 70 °C. Methanol or ethanol was
added, and the mixture was stirred at the temperature until
a homogeneous solution was formed. The batch was then
cooled to around 45 °C and seeded. The mixture was further
cooled to below 10 °C. The solid was filtered, washed, and
dried to give the step product, 1.
The process was further optimized. The effects of two
major parameters, reaction temperature and the amount of
triflic acid, were studied. The reaction time was affected by
both temperature and the amount of triflic acid. As illustrated
in Figure 6, the reaction rate increased as the reaction
temperature and/or the amount of triflic acid increased. For
instance, the slowest batch took 15 h (at 105 °C and using
(11) The graph was generated through a DOE method. Program used: Version
3.1 of JMP. Method: Response Surface/central composite-orthogonal.
Factors: reaction temperature (105-135 °C) and wt % of triflic acid (0.7-
3.0%).
590
•
Vol. 9, No. 5, 2005 / Organic Process Research & Development

Table 5. Crystallization of 1
at 20 °C while 9.6 g of 95% sulfuric acid were slowly added.
After the addition the mixture was stirred at 20 °C for 3 h.
To this mixture was slowly added lithium carbonate (11.7
g), and the mixture was stirred for 30 min. Methanol (320
g) was distilled out under reduced pressure (bath tempera-
ture: 45 °C, vacuum: 100 mbar). To the mixture was added
acetic acid (360 g), and the distillation was continued until
340 g of distillate were collected (vacuum: 50-100 mbar,
bath temperature: 63 °C, pot temperature should be con-
trolled not to exceed 52 °C). The mixture was sampled and
analyzed for methanol and water content. The total methanol/
water content should be below 3.5% (w/w). The bath
temperature was lowered to 50 °C, and 251.6 g of acetic
anhydride were added. After the addition the mixture was
held for 1 h and then heated to 100 °C and held for 4 h. The
mixture was then cooled to 20 ( 5 °C (pot temperature),
and 52.6 g of 95% sulfuric acid were slowly added. The
addition rate should be controlled so as to ensure that the
pot temperature is 20 ( 5 °C. After the completion of the
addition the mixture was stirred for 30 min at 20 ( 5 °C.
Then, 95.2 g of acetic anhydride were slowly added in 2 h
while maintaining the pot temperature at 20 ( 5 °C. After
the addition the mixture was stirred at 20 ( 5 °C (pot
temperature) for 30 min.
To a 1 L beaker was added 52.1 g of lithium carbonate
and 100 mL of acetic acid. This mixture was stirred and
cooled with an ice-water bath. The above reaction mixture
was drained from the reactor into the beaker with good
stirring. The whole mixture was then transferred back to the
reactor and was concentrated under reduced pressure until
over 419 mL of distillate were collected (vacuum: 60 mbar,
bath temperature: 60 °C, final pot temperature: 57 °C). The
mixture was cooled to 25 ( 5 °C, and to it was added
dichloromethane (150 mL) and water (400 mL). The mixture
was stirred at moderate speed for 30 min. The stirring was
stopped, and the mixture was held for 15 min. The organic
phase was separated, and to the aqueous layer was added
another 150 mL aliquot of dichloromethane. The mixture
was stirred at moderate speed for 15 min and then held for
15 min. The organic phase was separated. Both organic layers
were combined and washed with 160 mL of 4% sulfuric acid.
The pH of the aqueous phase should be below 2 at this point.
The organic phase was separated as a clear light-yellow
solution.
volume, mL/g
step 1 product
wt%
entry
solvent
yield %a
assay
1
2
3
4
5
6
7
MeOH
IPA
IPA
IPA
IPA
IPA
EtOH
2.2
2.9
3.5
3.9
3.9
4.4
2.9
70.9
82.2
82.5
79.9
79.9
79.1
77.1
96.7
84.6
91.3
92.3
92.1
93.2
94.7
a Overall yield from L-ribose.
ment was also possible. Thus, the three alcohols were
compared side by side for the crystallization.
Higher yield was achieved using IPA at a concentration
of up to 4.4 mL/g crude product (compared with 2.2 mL/g
crude product in methanol). However, the purity of the
crystallized product was lower, and the product was much
darker in color. More importantly, it appeared that the
transferability of the crystallization slurry in IPA was not
better than that in MeOH, even though the concentration of
the batch in IPA was only half of that in methanol. Although
more IPA could be used to dilute the slurry without
significantly sacrificing the yield, the approach would
increase the operational volume at this point which was
already the bottleneck volume of the whole process. The
results in ethanol were between those in methanol and IPA.
At the concentration of 2.9 mL/g crude product, ethanol
offered comparable overall yield to that from IPA. The
quality (color and wt % assay) of the products from ethanol
were better than the ones from IPA but was not as good as
the ones from methanol. Some examples are listed in Table
5. At the final concentration, the slurry in ethanol showed
better transferability. As a compromise of product quality
and yield, ethanol was chosen for the process. The experi-
ments seemed to indicate that, in the case of ethanol, better
yield could be achieved at lower hold temperature, so -5
°C was adopted for the process. Because the solubility of 1
in ethanol was low, the yield loss would be small with
increased solvent charge. Therefore, when maximum operat-
ing volume is not an issue in the process, more ethanol should
be used to improve the transferability of the batch.
Four lab demonstration batches were carried out on a scale
of 100 g L-ribose. The overall yield was 72-74%. The major
impurity was the unreacted 3 at 1.2-1.8%. The process was
piloted on 100 kg of L-ribose scale. Six batches were carried
out, with the average yield being 71.0% and average level
of 3 being 1.87%. In a subsequent manufacturing campaign,
16 batches were conducted on the same scale, and the
average yield was 74%. No major operational issues were
encountered in both campaigns.
A 2 L flask was charged with 81 g of 3, the above step
1 solution, and 37 g of acetic anhydride at ambient
temperature. The mixture was distilled at atmospheric
pressure (bath temperature, 90 °C). When the pot temperature
reached 85 °C and the distillation became very slow, vacuum
was applied (up to 30 mbar) and the distillation was
continued for 40 min at 90 °C (bath temperature) and then
for another 40 min at 120 °C (bath temperature, the pot
temperature reached 117 °C). The vacuum was released, and
843 mg of triflic acid was slowly added. After the addition
the vacuum was restored and the mixture was stirred at 115
( 5 °C (pot temperature) for 4 h. Upon the completion of
the reaction the mixture was cooled to 70 °C and to it was
added ethanol (750 mL). When a homogeneous solution was
Experimental Section
Procedure for the Overall Process.12 A dry, clean, 1 L,
four-neck round-bottom flask was charged with 100 g of
L-ribose and 500 mL of methanol. The mixture was stirred
(12) Dong, Z.; Zhang, P. U.S. Pat. Appl. Publ. U.S. 2004034213.
Vol. 9, No. 5, 2005 / Organic Process Research & Development
•
591

formed the mixture was cooled to 50 °C and held until heavy
precipitation formed (seeding might be necessary). The
mixture was then slowly cooled to -5 °C (bath temperature)
in 2 h and held for at least 2 h. The solid was filtered, washed
with 100 mL of cold ethanol to give 250.9 g of step 1 product
as a wet cake. Wt% assay was 75.5%, and overall yield was
74%.
The mother liquor obtained above was extracted with 2
× 100 mL of 3:7 (V/V) mixed solvents of EtOAc/TBME.
The combined organic layers were concentrated to almost
dryness. The residue was subjected to an azeotropic distil-
lation with 20 mL of toluene to remove residual water. The
resulting mixture (13 g) was a colorless oil that contained a
3:1 mixture of 4 and 2. Part of the mixture (12 g) was
subjected to flash column chromatography (140 g silica gel),
eluting with mixed solvents of EtOAc/petroleum ether (9:
31 V/V), to give 4.8 g of 4 (97.1% area purity by GC
analysis) as a colorless oil.
Reaction of 3 with 4. A 250 mL flask was charged with
3 (1.92 g) and a solution of 4 (4.8 g) in methyl acetate (50
mL). The mixture was concentrated at atmospheric pressure
to almost dryness (bath temperature: 90 °C).To this mixture
was added a solution of 22.7 mg of triflic acid in 1 mL of
methyl acetate. The mixture was stirred at 115 ( 5 °C (pot
temperature) under vacuum (30 mbar) for 4 h. Upon the
completion of the reaction the mixture was cooled to 70 °C,
and to it was added ethanol (23 mL). When a homogeneous
solution was formed the mixture was cooled to 50 °C and
held until a heavy precipitate formed. The mixture was then
slowly cooled to -5 °C (bath temperature) in 2 h and held
for 13 h. The solid was filtered, washed with 20 mL of cold
ethanol, and dried under vacuum at 50 °C for 17 h to give
4.1 g (70% yield) of 1 as an off-white solid.
Preparation of Pure 2 and Pure 4. To a 1 L, dry, clean,
round-bottom jacketed flask was added 50.0 g of L-ribose
and 400 g of anhydrous methanol. To this mixture was added
95% sulfuric acid (4.60 g). After the addition the mixture
was stirred at ambient temperature for 3 h. To the vessel
was added lithium carbonate (5.85 g) in one portion, and
the mixture stirred at ambient temperature for 30 min. The
mixture was subject to vacuum distillation (bath tempera-
ture: 30 °C) until 320 g of methanol were collected. The
distillation was stopped, and 103 g of acetic acid were added.
The vacuum distillation was reassumed (at bath temperature
40 °C) until 89 g of distillate was collected. The distillation
was again stopped, and 146 g of acetic acid were added.
Vacuum distillation was reassumed at bath temperature 40
°C and then slowly increased to 50 °C to distill out about
140 g of liquid. To this mixture was added acetic anhydride
(125.8 g). The mixture was heated to 100 ( 5 °C and held
for 5 h. The mixture was then cooled to 20 °C, and to it was
slowly added 26.3 g of 95% sulfuric acid over 30 min, while
maintaining the pot temperature below 25 °C. After the
addition the mixture was stirred for 30 min at 20 ( 5 °C. A
47.6 g amount of acetic anhydride was added slowly over 2
h at 20 ( 5 °C. After addition, the contents were stirred for
1 h. To this mixture was slowly added 26.05 g of lithium
carbonate. After the addition the mixture was stirred for 30
min. The mixture was subject to vacuum distillation at a bath
temperature of 50 °C until about 150 g of liquid were
collected.
Acknowledgment
We wish to thank Dr. Charles Gagliardi and Mr. Dobbert
Locklear for providing thermal stability data; Dr. Yong Jiang,
Ms. Lily Zhang, Ms. Amy O’Neil, Mr. Michael Hall, Ms.
Peggy Chiappetta, and Dr. Stephan Chan for analytical
support; Mr. George Drelich, Mr. Oliver Adair Brewton, and
Mr. Robert Stevenson for piloting the process in the pilot
plant; Dr. Sudhir Panvelker, Mr. Ashishi Tiwari, Mr. Bert
Miranda, and Mr. John Pecca for their contribution in the
Manufacturing campaign; Dr. Jason Yang for coordinating
the project; Dr. Michael Pierce for his support and advice
on the overall project.
To a 3/5 portion of the above residual content was added
water (60 g). The mixture was stirred for 30 min at 50 °C
and then cooled to 20 °C over 1 h and held for at least 30
min. To the slurry was added slowly a mixture of 30 g of
2-propanol and 120 g of water over 1 h. The mixture was
then further cooled to 0-5 °C and aged for at least 2 h. The
solid was filtered, washed with 72 g of water, and dried under
high vacuum at 40 °C for 24 h to afford 38.27 g (60.2%
corrected yield from L-ribose) of pure 2 as a white solid.
Received for review March 29, 2005.
OP050051M
592
•
Vol. 9, No. 5, 2005 / Organic Process Research & Development