Development of a robust and sustainable process for nucleoside formation

Article abstract of DOI:10.1021/op300335d

A practical and robust process for the synthesis of an Isatoribine pro-drug was demonstrated. The process relies on a streamlined glycosylation carried out in xylene and an effective regioselective enzymatic hydrolysis that can be run in a semicontinuous way. Analysis of the process mass intensity established the high impact from an environmental standpoint of our process improvement.

Full text of DOI:10.1021/op300335d

Article
pubs.acs.org/OPRD
Development of a Robust and Sustainable Process for Nucleoside
Formation
Fabrice Gallou,* Manuela Seeger-Weibel, and Pierre Chassagne
Novartis Pharma AG, Chemical and Analytical Development, CH-4002 Basel, Switzerland
ABSTRACT: A practical and robust process for the synthesis of an Isatoribine pro-drug was demonstrated. The process relies
on a streamlined glycosylation carried out in xylene and an effective regioselective enzymatic hydrolysis that can be run in a
semicontinuous way. Analysis of the process mass intensity established the high impact from an environmental standpoint of our
process improvement.
n the course of one of our development projects, we became
interested in an oral pro-drug (1) of Isatoribine (2), a
nucleoside analogue potentially useful for the treatment of
patients with chronic hepatitis C and potentially other viral
infections (see Figure 1). Delivered intravenously, it is a specific
While the synthesis of the base moiety (6) was quite
expeditive, the final glycosylation and selective ester hydrolysis
presented numerous challenges. The hygroscopicity and poor
solubility of the base led, for example, to a tedious, low yielding,
and unreliable process in the glycosylation step. As for the final
selective deprotection step, very low throughput, poor
productivity, and yield, as well as high cost due to a high
catalyst loading, were observed. All these aspects contributed to
a very low overall performance of the initial process and a
resulting high drug substance cost, both from the raw material
end (large amount of solvent, high loading of enzyme required)
and the processing cost. In addition, the API and all
intermediates also suffered from poor physical properties that
rendered all purifications by crystallization difficult. Our efforts
therefore focused on these last two transformations, with the
goal to rapidly come up with a practical solution for an
improved and scalable process.
I
GLYCOSYLATION
Figure 1. Targeted API (1).
■
We started our investigations on the glycosylation step, as it
appeared pivotal to the overall synthetic approach. We were
indeed hoping we could gain not only a better understanding of
the handling of the base, but also use this know-how for its
purification in the previous step. Earlier investigations had
showed the Vorbrueggen protocol or variants thereof were
Toll-like receptor agonist that induces the production of
cytokines such as α interferon. The parent drug, Isatorabine
(2), is obtained in vivo after full hydrolysis of the esters and
1,2
oxidation of the base moiety.
The original synthesis of 1 relied on the stepwise preparation
of the base (5) followed by glycosylation and selective
hydrolysis steps to the desired active pharmaceutical ingredient
3
−7
optimal for the formation of the desired bond.
Preliminary
requirement was a complete persilylation of the base, typically
reported in chlorinated solvents, or acetonitrile. In our case, a
combination of the very strong silylating agent, BSA and
acetonitrile seemed necessary due to the overall low reactivity
of our substrate. This led however to numerous side-reactions
such as the decomposition via the Ritter manifold in the
glycosylation step, or hydrolysis of the sensitive acetate
protecting groups. Initial monitoring of this persilylation step
via in situ IR revealed that the double silylation occurred quite
readily and could be easily monitored using this technique. The
impact of residual water was rapidly demonstrated to be critical
(
API) (see Scheme 1). Key glycosylation, by coupling of 5-
Aminothiazolo[4.5-d]pyrimidine-2,7-(3H,6H)-dione (6) with
,2,3,5-tetra-O-acetyl-β-D-ribofuranose (7) using N,N-bis-
trimethylsilyl)acetamide (BSA) in the presence of trimethyl-
1
(
silyl trifluoromethanesulfonate (TMSOTf) in acetonitrile
provided the product after direct isolation by precipitation.
Further purification by recrystallization from acetonitrile gave
5
-amino-3-(2′,3′,5′-tri-O-acetyl-β-D-ribofuranosyl)-3H-thiazolo-
[
4,5-d]pyrimidin-2,7-dione (8) in 70% yield. It was then
selectively hydrolyzed with Novozyme 435 (Candida antarctica
lipase) in a mixture of acetone and sodium phosphate buffer at
pH 7 to give the desired API, 5-amino-3-(2′,3′-di-O-acetyl-β-D-
ribofuranosyl)-3H-thiazolo[4,5-d]pyrimidin-2-one (1). The
product was further purified by crystallization from a mixture
of ethanol and heptane to give 87% of a white powder.
(
the presence of 0.5% water could severely impede the
silylation). A solvent screening was conducted first. BSA
Received: November 18, 2012
Published: February 20, 2013
©
2013 American Chemical Society
390
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
Scheme 1. First Generation Process
could not be used as a solvent as it would lead not only to
extensive decomposition of the product of the glycosylation,
but also to an increase of the overall process cost. Acetonitrile
allowed for sufficient solubility but raised other challenges
mixture was heterogeneous at the beginning of the reaction,
xylene offered the opportunity to reduce the water content
from the hygroscopic base prior to the persilylation thanks to a
very efficient azeotropic drying. In addition, at the end of the
persilylation, the excess BSA and its resulting byproducts could
be removed by distillation to minimize the decomposition of
the product to overhydrolysis products A, diols B, C and triol D
(Figure 2) and bis-sugar byproducts which could not be
isolated and characterized. An interesting feature of the process
was the visual transition from an heterogeneous mixture to a
clear solution once the persilylation had completed. The
distillation besides allowed for maintaining the overall
throughput attractive as the volume would remain constant
despite the successive addition of the various components. The
additional benefit was the opportunity to recycle xylene and
silyl acetamide, which was achieved by treatment with
hexamethyldisilane.
(
byproducts resulting from Ritter manifold). Dichloromethane
and dicholoroethane were also known solvents for such
transformation but were clear undesirable options for environ-
mental reasons. We therefore turned our attention to aromatic
solvents. Toluene has been reported to allow efficient
8
glycosylation in certain cases but did not allow solubilization
of our specific base, resulting in almost no reactivity. We
therefore switched to the more polar aromatic and higher
9
boiling point solvents xylene, cymene and anisole. To our
delight, we observed that xylene already displayed sufficient
desirable solubilization properties and reactivity to allow for the
persilylation in a few minutes at 60 °C (entries 1−5, Table 1).
a
Table 1. Solvent Effect in the Glycosylation Step
In the absence of silyl acetamide byproducts, a smaller
amount of decomposition was observed (no more than 5−8%
A+B+C+D observed) (see table 2, entries 2, 3, 4). However,
the presence of other uncharacterized regioisomer and bis-sugar
byproducts was still observed. Trimethylsilyl triflate had
appeared as the optimal acid catalyst from the various Lewis
b
c
d
entry
solvent
assay yield
isolated yield
purity
1
2
3
4
5
CH CN
80%
3
toluene
xylene
45%
ca. 85%
ca. 85%
ca. 85%
67%
>99%
>99%
>99%
cymene
anisole
50−55%
48−50%
acids evaluated (SnCl , ZnCl , TiCl , BF .OEt , AlCl ). The
4
2
4
3
2
3
a
peracetylated sugar starting material also displayed limited
stability under the reaction conditions (ca. 50% decomposition
after 24 h at 80 °C), as well as the product itself (ca. 20%
decomposition after 24 h at 80 °C). These observations
prompted us to evaluate the influence of the sugar addition
Reaction conditions: BSA (5 equiv), TMSOTf (1.1 equiv),
tetraacetylated D-ribose (1.1 eq, added in 1 h as a solid), 40 °C, 16
b
h. Assay yield determined by HPLC by comparison with reference
c
material. Isolated yield after crystallization from EtOAc/Solvent of
d
reaction. Purity determined by HPLC at 210 nm.
(
Table 3). Addition of the sugar in solution, as a slurry or as a
We therefore focused on this option as other high boiling
point solvents would eventually be problematic. Although the
solid had no impact on the outcome of the reaction (entries 1,
2, and 3). The solid addition was therefore preferred in order to
Figure 2. Main byproducts from glycolysation.
3
91
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
a
Table 2. Influence of the Temperature and BSA and Its Byproducts’ Effect on the Glycosylation Step
b
c
d
entry
solvent
distillation after the persilylation
temp (°C)
assay yield
isolated yield
purity
1
2
xylene
xylene
yes
no
20
60
51% (optimal after 10−12 h)
67%
>15% A + B + C + D
3
4
xylene
xylene
yes
yes
60
80
85%
65%
65%
>99%
5
−8% A + B + C + D
<0.6% A + B + C + D
85% (optimal after 2−3 h)
>99%
5
−8% A + B + C + D
<0.6% A + B + C + D
a
b
Reaction conditions: BSA (5 equiv), TMSOTf (1.1 equiv), tetraacetylated D-ribose (1.1 equiv, added in 1 h as a solid), 4−5 h. Assay yield
determined by HPLC by comparison with reference material. Isolated yield after crystallization from EtOAc/xylene of reaction. Purity determined
by HPLC at 210 nm.
c
d
a
Table 3. Influence of the Sugar Addition
Overall, the process appeared quite robust, as very similar
yields and purity were observed from our multigram laboratory
experiments. The following options were eventually selected as
the most efficient and with the least amount of side-products
isolated
yield
b
c
d
entry
1
sugar addition
solid in 1 h
assay yield
85%
purity
>99%
65%
65%
65%
57%
65%
65%
(
see Figure 2): the portionwise addition of the peracetylated
5
−8% A + B +
<0.6% A + B +
C + D
C + D
sugar (1.5 equiv) in 2−3 h, a slightly longer addition time
preferred in our production facilities, and reaction over 12 h at
2
3
4
5
6
solution in 1 h
slurry in 1 h
85%
−8% A + B +
C + D
85%
−8% A + B +
C + D
75−80%
>99%
5
<0.6% A + B +
C + D
6
0 °C (Scheme 2), resulting in a ca. 85% chemical yield. Our
solvent screening had shown that a combination of aromatic
solvent and an ester could result in crystalline material of very
high purity. Therefore, we just had to proceed with an aqueous
workup followed by distillation to a ca. 12−18% mixture in
xylene (determined by GC in the laboratory and by weighing of
the distillate in our plant), resuspension in isopropyl acetate,
demonstrated to be the optimal ester for purification and
throughput reasons, and slow cooling to 0 °C, resulting in
highly crystalline material that could be isolated by simple
filtration and drying to result in an overall 65% yield of product
as a single isomer, in a purity >99%, with the hydrolysis
products as the main byproducts. The critical content of xylene
for the crystallization was at the time determined by the weight
of the distillate, hence a rather large range to allow for a robust
process. This process was repeated with various grades of base
and turned once again extremely robust, in terms of both
chemical yield and isolated purity, which was one of our major
concerns to be able to produce on time. We had besides
demonstrated a more than 4-fold improvement in our
throughput, based on the development of an heterogeneous
process, and the recycling of the various components of the
reaction, which can be further illustrated by our overall
performance in process mass intensity, the currently accepted
>99%
5
<0.6% A + B +
C + D
solid in single
portion
>99%
5
−8% A + B +
<0.6% A + B +
C + D
C + D
solid in 10 portions 85%
over 1 h
>99%
5
−8% A + B +
<0.6% A + B +
C + D
C + D
solid in 10 portions 85%
over 4 h
>99%
5
−8% A + B +
<0.6% A + B +
C + D
C + D
a
Reaction conditions: xylene, BSA (5 equiv), TMSOTf (1.1 equiv),
tetraacetylated D-ribose (1.1 equiv), 60 °C, 4−5 h at the end of the
sugar addition. Assay yield determined by HPLC by comparison with
reference material. Isolated yield after crystallization from EtOAc/
xylene of reaction. Purity determined by HPLC at 210 nm.
b
c
d
minimize the overall volume of the reaction. The rate of the
sugar addition was also found to be critical. Decomposition of
the sugar was found to take place as demonstrated by the
formation of bis-sugar side-products and required controlled
addition of the sugar to minimize formation of these side-
products (entries 4, 5, 6).
10
metrics to express the greenness of a process.
Scheme 2. Final Glycosylation Process
3
92
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
Scheme 3. Original Hydrolysis Step
This practical process to glycosylate a sugar with a
nucleobase was later demonstrated to be general, operationally
simple, highly practical, and robust. It is of particular interest for
highly hygroscopic and poorly soluble nucleobases and is based
on the use of a high boiling point, nonpolar solvent, which
allows for the azeotropic drying of the nucleobase and the
removal of silylating agent byproducts after the initial
persilylation step.
HYDROLYSIS
■
The final step consisted in the regioselective hydrolysis of the
acetate of the primary alcohol (see Scheme 3). Enzymes have
long been demonstrated to be the best option in cases where
such selectivity is expected from a transformation and in the
presence of sensitive functional groups. In our case, Candida
antarctica lipase had been identified as the lipase of choice from
an earlier screening. Its polymer immobilized version Novozym
Figure 3. Solvent screening for the enzymatic hydrolysis.
was required. This resulted in an immediate improvement in
the volume efficiency by ca. 30%. In practice, this meant that
the same given reactor could allow us to produce 30% extra
product within the same time frame.
1
1
435 was the most attractive option from a long-term process
standpoint, as we could envision enzyme recycling. Immobi-
lized lipases are generally used to perform biotransformations
of most interesting industrial applications that quite often take
place in nonaqueous media.
The process was originally conducted in acetone or tert-
butanol, under phosphate buffer conditions to allow for the
hydrolysis and avoid further hydrolysis of the byproducts B, C,
and D (Figure 2). Under these conditions, completion was
observed after up to 2 days, and substantial amounts of mono-
and diacetates were observed, as a consequence of the long
reaction time. Furthermore, an overall very low volume
efficiency was required (>50 L total organic solvent/kg product
recovered), which rendered the process all the more unsuitable
for large scale, along with a high catalyst loading.
We continued our optimization of the volume efficiency and
demonstrated that the enzyme was actually very robust in this
solvent system. No deactivation was observed even with
relatively high concentration of starting material (see Table 4),
and a robust 85% isolated yield was obtained after
crystallization from MTBE/2-MeTHF.
a
Table 4. Optimization of the Concentration
conc (mol/L)
0.1
0.12
0.135
0.15
b
isolated yield
ca. 85% ca. 85% ca. 85% ca. 85%
c
HPLC purity
94.3%
94.6%
95.3%
94.5%
reaction volume efficiency (overall
volume/kg product (1)
ca. 26
ca. 21
ca. 17
ca. 13
a
Reaction conditions: Novozym 435, 2-MeTHF, pH 6.3−6.5, 25−27
We first screened a variety of mono- and biphasic solvent
systems. Solvent selection is known to be a critical factor for
b
c
°
C, 16 h. Isolation by crystallization from TBME/2-MeTHF. Purity
determined by HPLC at 210 nm.
1
2
biocatalysis, as organic solvents are known to influence the
1
3−15
enzyme activity and selectivity.
Solvents are indeed
reported to impact an enzyme conformation by interacting
with the hydration layer essential for catalysis and by altering
hydrophobic or H-bonding in the core of the protein as well as
Ultimately, an improvement of the throughput by a factor >2
was achieved for this batch process. The purity obtained at this
stage was sufficient to meet the higher specifications expected at
the selected salt form of the API.
16
protein solvation sites.
Our screening confirmed that tert-butanol and acetone were
some of the optimal solvents from a reactivity standpoint.
However, alternatives were identified such as isopropanol, 2-
butanol, and 2-MeTHF (see Figure 3).
The latter especially attracted our attention, as solubility
studies had revealed the product to be more than 50 times
more soluble in 2-MeTHF than in TBME, the extraction
solvent used in the first generation process. Together, this
meant a dramatic reduction of the solvent requirements for the
workup in particular, as no additional water immiscible solvent
We then turned our attention to the high catalyst loading
required. The long-term goal was to accelerate the enzymatic
process in such a way that could minimize the reaction time
and avoid further hydrolysis of the diacetate product to the
monohydrolyzed and fully hydrolyzed byproducts. Such a
continuous or semicontinuous process could allow for recycling
of the immobilized enzyme. In order to accelerate the process,
we screened a series of additives, such as various Lewis acids
17
18
(FeCl , for example ), the inorganic salts NaCl and LiCl ), or
3
19
other additives (PEG ). An interesting but moderate effect
3
93
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
with LiCl and PEG was observed (5−10% acceleration) so that
the added benefit did not appear significant enough to justify
such a change (see Figure 4).
Figure 5. Enzyme activity over 15 cycles. The reaction was conducted
as described in the Experimental Section. The percentages indicated
are HPLC area at 210 nm. The overhydrolysis side-products and the
desired product have the same extinction coefficient at that
wavelength.
Figure 4. Additive effect.
Fortunately, the high stability of the enzyme provided an
even simpler handle for acceleration, while maintaining the
desired selectivity. A ca. 20% reaction acceleration was indeed
observed when increasing the temperature to 35 °C but
unfortunately only in acetone, thus requiring a second
immiscible solvent for the isolation, most probably 2-MeTHF
as previously identified.
overhydrolysis be observed (yield down to 88.6% after the 10th
cycle and up to 2% overhydrolysis byproducts), presumably due
to a much longer contact time of substrate with the enzyme.
Nevertheless, we had demonstrated that the enzyme could be
20
recycled very efficiently to result in an optimal process.
With these considerations, a system was designed using a
column packed with Novozym 435 and swollen with acetone.
At that stage, a feed of the starting material in acetone at a
concentration of 1.6 mol/L and 35 °C would be circulated at
various flow rates in a loop mode until full conversion was
observed, as indicated by the plateau observed during the in situ
pH measurement. The pH was constantly adjusted to 6.3−6.5
with a phosphate buffer. It was rapidly discovered that the
reaction could be >98% complete within a few hours with this
simple setup, with ca. 1.3% overhydrolysis byproduct. This was
achieved at this stage with a relatively high enzyme loading of 3
equivalent weight ratio of enzyme to substrate and a 1.6 mL/L
flow rate (see Table 5 for reaction monitoring). By comparison,
GREEN CHEMISTRY METRICS
■
To illustrate our process improvements, we wanted to measure
some of the well-accepted green metrics such as the process
mass intensity. The process mass intensity (PMI) is a well
accepted metric of greenness currently recommended by the
ACS Green Chemical Pharmaceutical Roundtable. It is defined
as the quantity of raw material input in kilograms per the
quantity of bulk API in kilograms. For the latter, several PMIs
were calculated on solvents, water, and reagents (see Table
21
6
).
Table 6. Process Mass Intensity for the Old and New
Processes on the Glycosylation and Hydrolysis Steps
a
Table 5. Head-to-Head Comparison of Batch/Semibatch
Processes
old
new process in
new process w/
recycling
a
hydrolysis step
process w/o recycling
step PMI substrate,
reagents, solvents
37.9
25.0
16.0
batch process
85%
semibatch process
isolated yield
purity
90%
step PMI solvents
step PMI aqueous
step PMI
32.6
37.2
75.1
19.6
9.6
12.3
9.6
94.3% to 95.3%
95.4% to 98.2%
1% total B, C, D
overhydrolysis
3%−5% total B, C, D
34.5
25.5
hydrolysis step
new process in
2-MeTHF
new process in 2-
MeTHF w/ recycling
a batch mode hydrolysis would result in the same conversion
within 16 h and would display 3−5% overhydrolysis, with the
same catalyst loading.
step PMI substrate,
reagents, solvents
280.2
23.5
11.9
step PMI solvents
step PMI aqueous
step PMI
277.6
51.8
21.5
43.6
67.1
11.2
43.6
55.5
Encouraged by these preliminary results, we then studied the
impact of the contact time. It was then demonstrated that the
fastest flow rate and residence time would minimize over-
hydrolysis to the di- and monoacetate byproducts B, C, and D
from 1.3% to <1.0% when going from 1.6 mL/min to 6.4 mL/
min. The next obvious development was to reuse the enzyme.
We were here delighted to find out that, after up to 15 uses, no
significant loss of activity was observed with yields of the
desired monohydrolyzed product oscillating between 95.4%
and 98.2%, and less than 1.6% overhydrolysis in all cases
332.0
For the glycosylation step, a 2- and even 3-fold improvement
of the overall mass intensity was achieved without or with
recycling, respectively, due in part to a better choice of organic
solvent to promote the reaction and to the superior physical
properties of our selected solvents that did not require an
additional water-immiscible solvent for the extraction. Besides,
the product of the glycosylation, 7, could be isolated and
purified by crystallization without extensive operations and
(
Figure 5). Only when the flow was interrupted and the
column stored overnight could a drop in reactivity and slight
3
94
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
water washings in the workup, hence giving a ca. 4-fold
improvement in the mass intensity on water.
1H, J = 12.0, 3.6 Hz), 4.36−4.31 (m, 1H), 4.23 (dd, 1H, J =
1
3
12.0, 5.6 Hz), 2.13 (s, 3H), 2.12 (s, 3H), 2.05 (s, 3H) ppm; C
For the enzymatic hydrolysis step, a very interesting >6-fold
improvement in mass intensification was observed in batch
using 2-MeTHF as the organic solvent. For the semicontinuous
process, the benefits were not as obvious, as it could only
perform comparably in acetone, which subsequently required a
water-immiscible organic solvent.
NMR (100 MHz, CDCl ): δ 170.8, 169.7, 169.0, 160.9, 155.3,
3
1
50.5, 104.3, 85.6, 78.5, 72.0, 70.0, 62.7, 20.7, 20.6, 20.5 ppm.
+
LR-MS (EI): m/z: calcd for C H N O SH [M + H ]: 427.1,
16
18
4
8
found: 427.1. HPLC purity: 99.2% with major byproduct A
observed in 0.19%; HPLC assay: 99.4%; titration: 98.8%; water
content: 0.2%.
In conclusion, a pratical and robust process for the synthesis
of an Isatoribine pro-drug was demonstrated. The process relies
on a streamlined glycosylation process in xylene and an
effective and regioselective enzymatic hydrolysis with a much
improved environmental impact. The catalytic activity of the
immobilized lipase was demonstrated to be very robust, as the
enzyme displayed an excellent behavior as catalyst in both
aspects of high levels of activity, selectivity, and excellent
operational stability. This process was further developed in a
semicontinuous mode and showed to proceed in an even more
efficient manner from a throughput standpoint.
(
2R,3R,4R,5R)-2-(Hydroxymethyl)-5-(5-amino-2-
oxothiazolo[4,5-d]pyrimidin-3(2H)-yl)tetrahydrofuran-
,4-diyl Diacetate (1). Batch Process. Peracetylated product
8; ca. 1.0 kg) was charged to an inert reactor. 2-MeTHF (2.2
3
(
L) was added at room temperature, and the suspension was
stirred at room temperature. Phosphate buffer (15.8 L) was
added at room temperature. (The phosphate buffer was freshly
prepared according to the following recipe: 450 g of
disodiumhydrogenophosphate was dissolved in 32 L of water,
and ca. 67 mL of phosphoric acid (ca. 85%, Fluka 79617) was
added to adjust the pH to 6.95−7.05. The density was
measured to be 1.00027.) The pH was adjusted to 6.50 with
EXPERIMENTAL SECTION
■(
2R,3R,4R,5R)-2-(Acetoxymethyl)-5-(5-amino-2-
oxothiazolo[4,5-d]pyrimidin-3(2H)-yl)tetrahydrofuran-
0
.25 mol/L phosphoric acid. Novozym 435 (370 g) was added,
and the suspension was stirred at IT 25−27 °C at ca. 150 rpm
until completion. The pH was adjusted to 6.3−6.5 with a 0.5
mol/L disodium hydrogenophosphate solution about every 4 h.
At completion, the reaction mixture was filtered through a
Nutsche filter to remove the enzyme, and the cake was rinsed
with 2-methyltetrahydrofuran (1 L). 2-Methyltetrahydrofuran
was added to the filtrate, and the resulting biphasic mixture was
stirred at IT 35−40 °C for 15 min. Stirring was stopped, and
the phases were separated. The aqueous phase was extracted a
second time with 2-methyltetrahydrofuran (3 L) at IT 35−40
°C, and the organic extracts were combined and washed with
deionized water (9 L) at IT 20−25 °C. To the organic phase
was added activated charcoal (8.8 g), and the resulting
suspension was stirred at IT 20−25 °C for 2 h and filtered
through a pad of Celite. The cake was rinsed with 2-
methyltetrahydrofuran (3 L). The resulting yellowish solution
was concentrated at IT 40−45 °C (IT < 50 °C) and ca. 180
mbar to about one-fourth of the volume and crystallized from
3
,4-diyl Diacetate (8). A solution of the 5-aminothiazolo[4,5-
d]pyrimidin-2(3H)-one (6) (0.85 kg, 5.1 mol) in xylene (12 L)
was distilled at 70−80 °C (100−150 mbar), removing ca. 1 L of
solvent. The solution was cooled to 60 °C, and BSA (2.62 kg,
1
2.9 mol) was added dropwise and stirred for 1 h. The reaction
mixture was distilled again, removing ca. 1 L of solvent, excess
BSA, and byproducts. The reaction was cooled down to 60 °C,
and TMSOTf (1.2 kg, 5.4 mol) was added dropwise, followed
by the tetraacetylated D-ribose (7) (1.7 kg, 5.4 mol), which was
added continuously over the course of 90 min. The reaction
mixture was stirred at 65 °C for 16 h. At completion, the
temperature was reduced to 35−40 °C and isopropylacetate (6
L) was added. The resulting mixture was stirred at that
temperature for 10 min, and 8% sodium bicarbonate solution
was added at a rate such that the temperature remained below
4
1
5−50 °C. The resulting mixture was stirred at 35−40 °C for
5 min. Stirring was stopped, and the phases were separated.
The desired product is in the organic phase. The aqueous
extract was extracted one more time. 20% sodium chloride
solution was added, and the resulting mixture was stirred for 15
min at 20−25 °C. Stirring was stopped, and the phases were
separated. The desired product is in the organic phase.
Isopropylacetate was charged, and the resulting suspension
was refluxed for 0.5 h (internal temperature 85−90 °C, jacket
temperature 95 °C) to collect the solid on the side of the
reactor. When a clear solution was obtained, with no more solid
on the vessel, the temperature was reduced to 50−55 °C
TBME/2-MeTHF to result in 0.78 kg of desired product in
1
9
6
6.1% purity. H NMR (400 MHz, CDCl ): δ 8.22 (s, 1H),
3
.28 (d, 1H, J = 3.2 Hz), 6.08 (dd, 1H, J = 5.6, 2.8 Hz), 5.72
(
dd, 1H, J = 7.2, 6.0 Hz), 5.50 (bs, 2H), 4.32 (bs, 1H), 3.98
(dd, 1H, J = 12.0, 2.8 Hz), 3.85 (bs, 1H), 2.19 (s, 3H), 2.09 (s,
3H) ppm. LRMS (EI): m/z: calcd for C H N O SH [M +
14
16
4
7
+
H ]: 385.3, found: 385.3.
Semicontinuous Process. A 1 cm diameter Nutsche filter
was charged with ca. 10 g of Candida antarctica lipase Novozym
(
jacket temperature 55 °C) and some seed product was added.
The seeds remained suspended, and the resulting mixture was
stirred at that temperature for 0.5 h until a fine suspension was
observed. The temperature was decreased to 20−25 °C in 2 h
and to 0−5 °C in 1 h. The suspension was held at that
temperature for 2 h, and the solid was filtered through a glass
sintered funnel (diameter 14 cm) under vacuum. The wet cake
was rinsed with cold isopropylacetate (ca. twice the cake
volume). The solid was dried for 16 h in a vacuum oven at 50
4
35. A solution of adduct (ca. 10 g dissolved in 9 mL of acetone
and 16 mL of pH 7.0 phosphate buffer) was passed through the
filter at ca. 1.6 mL/min (ca. 0.2 bar pressure) until completion.
The pH of the filtered mixture was continuously maintained
between 6.3 and 6.5 with a Na HPO solution. The reaction
was complete after ca. 2 h. The process was repeated 10 times.
The phases were then separated, and the aqueous phase was
extracted one time with ca. 200 mL of 2-methyltetrahydrofuran.
The combined organic phases were washed once with water
and concentrated under reduced pressure to give the crude
product in >90% yield and with <1% overhydrolysis byproduct.
2
4
°
C and 50 mbar to provide 1.4 kg of a white solid (single
1
1
isomer observed by H NMR). H NMR (400 MHz, CDCl ): δ
3
8.16 (s, 1H), 6.15 (d, 1H, J = 3.2 Hz), 6.06 (dd, 1H, J = 5.6, 2.8
Hz), 6.01 (dd, 1H, J = 7.2, 6.0 Hz), 5.29 (bs, 2H), 4.51 (dd,
3
95
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396

Organic Process Research & Development
Article
AUTHOR INFORMATION
Corresponding Author
Tel.: +41 61 324 9911. Fax: +41 61 324 8536. E-mail address:
fabrice.gallou@novartis.com.
■
*
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The authors would like to thank Bernard Riss, Caspar Vogel,
Christine Richardson, Pascal Beney, Sadja Elmont, and Andreas
Knell for their participation on the project.
REFERENCES
■
(
1) Xiang, A. X.; Webber, S. E.; Kerr, B. M.; Rueden, E. J.; Lennox, J.
R.; Haley, G. J.; Wang, T.; Ng, J. S.; Herbert, M. R.; Clark, D. L.; Banh,
V. N.; Li, W.; Fletcher, S. P.; Steffy, K. R.; Bartkowski, D. M.;
Kirkovsky, L. I.; Bauman, L. A.; Averett, D. R. Nucleosides, Nucleotides
Nucleic Acids 2007, 26, 635.
(
7
(
1
2) Fletcher, S.; Steffy, K.; Averett, D. Curr. Opin. Invest. Drugs 2006,
, 702.
3) Vorbrueggen, H.; Krolikiewicz, K.; Bennua, B. Chem. Ber. 1981,
14, 1234.
(
4) Vorbrueggen, H.; Hoefle, G. Chem. Ber. 1981, 114, 1256.
5) Vorbrueggen, H.; Bennua, B. Chem. Ber. 1981, 114, 1279.
6) Vorbrueggen, H.; Ruh-Pohlenz, C. Org. React. 2000, 55, 1.
7) Vorbrueggen, H. Acc. Chem. Res. 1995, 28, 509.
8) Li, N.-S.; Piccirilli, J. A. J. Org. Chem. 2006, 71, 4018.
9) Our internal industrial grade xylene was used which consisted of a
(
(
(
(
(
mixture of all three isomers, ortho, meta, and para, in an average ratio
of <0.5/5/1, respectively.
(
10) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.;
Manley, J. B. Org. Process Res. Dev. 2011, 15, 912.
11) Candida antarctica lipase B (CAL-B, Novozym435 , 7300 PLU/
@
(
g). For typical uses of this lipase, see, for example: (a) Orrenius, C.;
Norin, T.; Hult, K.; Carrea, G. Tetrahedron: Asymmetry 1995, 6, 3023.
(
b) Orrenius, C.; Ohrner, N.; Rotticci, D.; Mattson, A.; Hult, K.;
Norin, T. Tetrahedron: Asymmetry 1995, 6, 1217. (c) Uppenberg, J.;
Ohrner, N.; Norin, M.; Hult, K.; Kleywegt, G. J.; Patkar, S.; Waagen,
V.; Anthomen, T.; Alwyn Jones, T. Biochemistry 1995, 34, 16832.
(
12) Kinoshita, M.; Ohno, A. Tetrahedron 1996, 52, 5397.
(
13) Burke, P. A.; Griffin, R. G.; Klibanov, A. M. Biotechnol. Bioeng.
1
(
1
(
993, 42, 87.
14) Margolin, A. L.; Tai, D. F.; Klibanov, A. M. J. Am. Chem. Soc.
987, 109, 7885.
15) Sakurai, T.; Margolin, A. L.; Russell, A. J.; Klibanov, A. M. J. Am.
Chem. Soc. 1988, 110, 7236.
(
(
16) Dordick, J. S. Biotechnol. Prog. 1992, 8, 259.
17) Takahashi, K.; Tamaura, Y.; Kodera, Y.; Mihama, T.; Saito, Y.;
Inada, Y. Biochem. Biophys. Res. Commun. 1987, 142, 291.
18) Talukder, M. M. R.; Beatrice, K. L. M.; Song, O. P.; Puah, S.;
Wu, J. C.; Won, C. J.; Chow, Y. Energ Fuels 2008, 22, 141.
(
(
(
(
19) Secundo, F.; Carrea, G. J. Mol. Catal. B: Enzym. 2002, 19, 93.
20) Gallou, F.; Beney, P. WO 2008-US68197 20080625.
21) Anastas, P. T.; Warner, J. C. Green Chemistry: Theory and
Practice; Oxford University Press: New York, 1998; p 30.
3
96
dx.doi.org/10.1021/op300335d | Org. Process Res. Dev. 2013, 17, 390−396