Process development of selectively benzoylated and fluorinated glycosyl donors

Article abstract of DOI:10.1021/op100053k

Route selection, process development and large-scale preparation of selectively benzoylated and fluorinated d-glucopyranoses, required as glycosyl donors for the synthesis of the SGLT inhibitor SAR7226, are discussed.

Full text of DOI:10.1021/op100053k

Organic Process Research & Development 2010, 14, 623–631
Process Development of Selectively Benzoylated and Fluorinated Glycosyl Donors
Franz J. Weiberth,*,† Harpal S. Gill,† Ying Jiang,† George E. Lee,† Philippe Lienard,‡ Clive Pemberton,† Matthew R. Powers,†
Witold Subotkowski,† Witold Tomasik,§ Benoit J. Vanasse,† and Yong Yu†
Chemical DeVelopment, sanofi-aVentis U.S., 1041 Route 202-206, Bridgewater, New Jersey 08807, U.S.A.
Abstract:
development because it had been used successfully for small
deliveries of 1 and was readily available in multikilogram
quantities (Scheme 2). The objectives for the first pilot-plant
campaigns were to demonstrate a scaleable synthesis of 1 and
deliver drug substance rapidly. The first step in the synthesis
of 6a involved selective tribenzoylation at the 2-, 3-, and
6-positions of 8a using benzoyl chloride in pyridine. Low
reaction temperatures (e0 °C), commonly employed in the
literature,3 were found to be unnecessary. Comparable selectivi-
ties were achieved at 20 °C, with the added benefit of shorter
reaction times. A total of 4.7 equiv of BzCl were necessary to
achieve an optimal distribution of products, typically 85:12:3
normalized HPLC area % (A%) ratio of the desired product,
the tetrabenzoylated byproduct, and underbenzoylated interme-
diates, respectively. The tetrabenzoylated byproduct was more
soluble and easier to remove than the underbenzoylated
intermediates. To mitigate overbenzoylation, the reaction was
continued until the amount remaining of each of two under-
benzoylated intermediates was <3.0 A%. An isolation process
that minimized the carryover of impurities and provided high-
purity product directly from the reaction mixture without a
separate recrystallization was developed. Thus, after water
quench, extraction into toluene, and aqueous washes to remove
benzoic acid and pyridine, the organic phase was partially
concentrated and diluted with n-heptane to facilitate crystal-
lization of 7a. Four batches, providing a total of 147 kg of 7a,
were produced in an average yield of 73% and with >98.8 A%
purity.
Route selection, process development and large-scale preparation
of selectively benzoylated and fluorinated D-glucopyranoses, re-
quired as glycosyl donors for the synthesis of the SGLT inhibitor
SAR7226, are discussed.
Introduction
Sodium-glucose transporters (SGLT) are a family of proteins
that have an important role in transporting glucose across
cellular membranes against a concentration gradient. These
proteins are predominantly expressed in intestinal mucosa
(SGLT1) and kidney tubules (SGLT1 and SGLT2) where they
mediate intestinal absorption and renal reabsorption of glucose.1
SAR7226 (1) emerged as a promising SGLT1/SGLT2 inhibitor,
potentially useful for the treatment of diabetes,2 and multiki-
logram quantities of the drug substance were required to support
preclinical and clinical activities.
Scheme 1 illustrates a retrosynthetic, convergent approach
to 1 that features the coupling of two advanced intermediates,
hydroxypyrazole 2 and glycosyl donor 6 to afford the SAR7226
backbone structure. Intermediate 2 is derived from monoalkyl-
ation of ethyl 4,4,4-trifluoroacetoacetate (3) with 4-methoxy-
benzyl chloride (4) followed by treatment with benzylhydrazine
(5). Glycosyl donor 6 is derived from a suitable galactosugar 8
that is selectively protected, fluorinated with inversion at the
C-4 position, and then manipulated at the anomeric position.
Several synthetic routes and iterations were successfully
demonstrated from kilo scale up to pilot-plant scale as the target
molecule progressed to clinical development and increasing
quantities of drug substance were required. The ultimate goal
was to establish an economical and industrializable synthesis
of SAR7226. This paper summarizes the process development
of glycosyl donor synthons 6.
The next step involved the fluorination of 7a with inversion
of configuration at the C-4 position to afford 9.4 During early
development, diethylaminosulfur trifluoride (DAST) was re-
placed with bis(2-methoxyethyl)aminosulfur trifluoride (BAST),
a more thermally stable fluorinating reagent.5 The fluorination
process likely proceeds through the activated intermediate 12
(Figure 1), analogous to DAST.4 Upon heating, the resulting
activated hydroxyl at C-4 is displaced by fluoride ion with
inversion in a highly selective manner.6 A significant competing
reaction pathway is elimination to give the olefin 13.
Discussion
Development of Glycosyl Donor 6a from 8a. Methyl R-D-
galactose monohydrate (8a) was chosen as the substrate for early
A limited screening of solvents established THF as a suitable
solvent for the fluorination step. The addition of neat BAST
was exothermic and was performed at ambient temperature to
* Author to whom correspondences should be sent. E-mail: franz.weiberth@
sanofi-aventis.com.
† Chemical Development, sanofi-aventis U.S.
‡ Current address: Pharmaceutical Science Development, sanofi-aventis R&D,
Vitry-Sur-Seine, France.
(3) (a) Williams, J. M.; Richardson, A. C. Tetrahedron 1967, 23, 1369.
(b) Reist, E. J.; Spencer, R. R.; Calkins, D. F.; Baker, B. R.; Goodman,
L. J. Org. Chem. 1965, 30, 2312. (c) Rye, C. S.; Withers, S. G. J. Am.
Chem. Soc. 2002, 124, 9756.
§ Chemical Development Support, Analytical Sciences Department, sanofi-
aventis, Bridgewater, New Jersey, U.S.A.
(1) Idris, I.; Donnelly, R. Diabetes, Obes. Metab. 2009, 11, 79.
(2) (a) Frick, W.; Glombik, H.; Kramer, W.; Heuer, H.; Brummerhop,
H.;Plettenburg,O.U.S.Patent2005/0014704A1,2005(WO2004052902;
CAN 141:38810). (b) Frick, W.; Glombik, H.; Kramer, W.; Heuer,
H.; Brummerhop, H.; Plettenburg, O. U.S. Patent 2004/0259819 A1,
2004 (WO2004052903; CAN 141:38811).
(4) Card, P. J. J. Org. Chem. 1983, 48, 393.
(5) Lal, G. S.; Pez, G. P.; Pesaresi, R. J.; Prozonic, F. M.; Cheng, H. J.
Org. Chem. 1999, 64, 7048, and references cited therein.
(6) None of the 4-F epimer of 9 was detected in the product or in the
filtrate by 19F NMR spectroscopy.
10.1021/op100053k  2010 American Chemical Society
Published on Web 05/03/2010
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
623

Scheme 1. Retrosynthesis of SAR7226
Scheme 2. Synthesis of glycosyl donors 6a and 6ba
a Reagents and conditions: a) BzCl, pyr, 20 °C, 73%; b) BAST, THF, 52 °C, 80%; c) AcOH/Ac2O/H2SO4, 50 °C, 58%; d) NH2CH2CH2NH2/AcOH, THF, 20 °C,
quant.; e) BzCl/pyr/Bz2O, NMP, 45%; f) BAST, 2-MeTHF, 52 °C, 77%; g) HBr/AcOH, 55 °C, g 96%.
provide the activated alcohol in a facile manner. The reaction
was then heated at 50 °C to complete the fluorination. Typically,
a 5:1 ratio of fluorinated product 9 to olefin 13 was achieved.
Attempts to substantially improve the selectivity were unsuc-
cessful. Heating rate and reaction temperature marginally
affected the ratio of product to olefin. After a water quench,
extractive workup using toluene, partial concentration of the
organic phase, and crystallization upon addition of n-heptane,
a total of 382 kg of 9 was isolated as a white solid in seven
batches in 80% average yield, and containing e3 A% of olefin
13, the limit for ensuring suitable downstream chemistry.
A key issue for the fluorination process was the safe handling
of neat BAST on large scale in the pilot plant. For the initial
batches (61-mol scale), BAST was purchased in bottles and
first transferred in a fume hood to a Teflon-coated tank prior to
use in the plant. On subsequent larger scales (227-mol scale),
the hazards and the extra processing steps associated with
predispensing the material were eliminated. The neat BAST
was supplied in reusable 100-L pressure vessels constructed of
perfluoroalkoxy copolymer resin (PFA).7,8
to separate the anomers of 10 because both yield 6a upon
deacylation.9 From a process chemistry perspective, however,
the R-anomer was preferred in order to maximize yields because
it was more readily recoverable (the ꢀ-anomer was much more
soluble) and avoided inefficiencies often encountered when
isolating from mixtures. An isolation procedure that removed
the impurity from this complex mixture was developed. After
an extractive workup using toluene and addition of 2-propanol
as an antisolvent, 10 was isolated at 0 °C as a white solid in an
average yield of 59% (222 kg total, five batches) consisting
of a 9:1 or better mixture of R-/ꢀ-anomers and containing only
about 0.6 A% of the open-chain impurity.
The development of a selective anomeric deacylation of 10
was a critical step in implementing a coupling strategy based
on activation of 6a followed by coupling with hydroxypyrazole
2.10 For lab campaigns, selective deacylation was achieved using
hydrazine hydrate as base. Prior to pilot-plant scale, an effort
was made to replace hydrazine hydrate with a less hazardous
reagent (Table 1). Ammonia and morpholine were less chemose-
lective and generated partially debenzoylated impurities. Am-
monium acetate gave poor conversion. Ethylenediamine showed
some potential, although a small amount of debenzoylation was
detected. In comparison, ethylenediamine as its less basic
monoacetate salt11 in THF proved to be a very regioselective
reagent for deacylation, and was employed on scale up. The
workup was designed to allow process telescoping with the
subsequent coupling step to improve overall efficiency and to
Conversion of 9 to 10 was performed using acetic acid, acetic
anhydride, and sulfuric acid as the reagent system.2 Depending
on conditions, varying ratios of three products were obtained:
the desired product 10 as a mixture of R- and ꢀ-anomers and
an impurity tentatively assigned (NMR) as an open-chained
tribenzoylated and triacylated compound. It was not necessary
(7) BAST was manufactured by Air Products and Chemicals, Inc.
(www.airproducts.com) and sold as Deoxo-fluor. The containers were
obtained from Entegris, Inc. (www.entegris.com).
(8) Independent corrosivity studies have not yet been performed. Air
Products and Chemicals, Inc. has a history of successfully manufactur-
ing BAST and performing BAST reactions in Hastelloy equipment.
Accordingly, the BAST reactions and quenches (HF generated) were
performed in the pilot plant in Hastelloy C276 equipment.
(9) In the subsequent deacetylation step, a 4:1 mixture of R/ꢀ-anomers
of 6a is obtained regardless of anomeric composition of starting
10.
(10) Poor conversions (<10 A%) resulted in attempts to directly couple 10
or 11 with 2 in the presence of Lewis acids.
(11) Zhang, J.; Kova´cˇ, P. J. Carbohydr. Chem. 1999, 18, 461.
624
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

route rather than just routine optimization of existing chemistry.
In this context, the major criteria for industrialization were
deemed achieved, or were projected to be achievable after full
optimization, with the exception of CoG of the API.
The material cost to manufacture 6a, based on projected bulk
prices for key materials, was determined to be about 3-fold
higher than the target cost for an economical glycosyl donor.
The three major cost contributors in the synthesis of 6a were
low yield for the conversion of 9 to 10, the cost of the BAST,
and the high cost of R-methyl galactose (8a). Several improve-
ments in the existing synthesis that favorably impacted the cost
of 6a were shown to be feasible, but were not fully developed
in time for the first pilot-plant campaign. For example, a direct
conversion of 9 to 6a was developed and nonaflyl fluoride13
was identified as a cost-effective replacement for BAST.
However, even with these improvements, it was evident that
the cost target would not be achievable and that a synthesis
employing a significantly more economical starting sugar
substrate was required.
Development of Glycosyl Donors from Galactose (8b).
It may seem surprising that the unit cost for bulk quantities of
R-methyl galactose (8a) is high because it is prepared industri-
ally from readily available and inexpensive galactose (8b)14
using conventional Fischer glycosylation reaction conditions,
namely, MeOH as solvent in the presence of an acid catalyst.
However, this glycosylation produces an equilibrium mixture
of ∼3:115 of the desired pyranoside 8a and its ꢀ-anomer together
with R- and ꢀ-furanoside impurities and the open-chain dimethyl
acetal adduct. In-house attempts to substantially improve this
selectivity by screening a variety of conditions and acid catalysts
were unsuccessful.16 Isolating pure 8a from a mixture of
products has proven to be difficult.15 The need for repeat
recrystallizations17 together with low recoveries when crystal-
lizing from mixtures has contributed to the high commercial
price for 8a.
Figure 1. BAST intermediate and elimination byproduct.
avoid a difficult isolation of the anomeric mixture of 6a. Initially,
dichloromethane (DCM) was employed as the extraction
solvent. However, it was difficult to completely remove the THF
in the DCM phase after water washes and concentration.
Residual THF interfered with the downstream chemistry (ace-
timidate coupling with pyrazole 2) by competing with the Lewis
acid catalyst employed in the coupling step. This issue was
overcome by replacing DCM with toluene for extractive
workup. THF was then effectively removed (limit of 0.1
wt %) during partial concentration of the toluene phase. The
resulting toluene solution of 6a (quantitative yield, > 99.5 A%
purity) was used directly in the coupling step.
The four-step synthesis of 6a was shown to be reliable and
reasonably robust, and a total of 202 kg was prepared in the
early pilot-plant campaigns, most of which was elaborated to
prepare 90 kg of SAR7226. The campaign objectives were
achieved: a scaleable synthesis of 1 employing glycosyl donor
6a was developed, enabling accelerated delivery of API to
support preclinical and early clinical studies. However, the long-
term goal was to establish an industrializable synthesis of 1.12
This implies:
• technical scalability and good robustness of the pro-
cesses
• good overall yield and process productivity (space-time-
yield)
• acceptable process safety and industrial hygiene
• tolerable waste generation (environmental impact mini-
mized)
• availability of all materials on commercial scale
• acceptable chemical and physical quality of the drug
substance: impurities are avoided or controlled
• freedom to operate from an intellectual property stand-
point exists or is manageable (e.g., licensing)
• acceptable cost of goods (CoG); API target cost not
exceeded after full process optimization
Attention therefore shifted to developing alternative sub-
strates. It occurred to us that tetrabenzoyl galactose could be a
viable synthon.10 The strategy for synthesizing a suitable
glycosyl donor would be similar to that used for the previous
substrate: blocking of all positions other than the 4-OH, followed
by fluorination and then manipulation at the anomeric position
to generate the desired synthon. Whereas tribenzoylation of 8a,
containing a methoxy group fixed in the R-anomeric position,
is well behaved and reasonably selective, the analogous regi-
oselective 1,2,3,6-tetrabenzoylation (and regioselective tetraa-
Using these criteria, the overall process to synthesize 1 via
glycosyl donor 6a was assessed. Reasonable assumptions were
projected for yield improvements, streamlining workups and
isolations, proper selection and minimization of solvents (e.g.,
minimizing both volume and total number of different solvents
in the overall synthesis and optimizing solvent recoverability),
and anticipated commercial prices of materials. The exercise
was a means to identify potentially insurmountable hurdles to
industrialization that would help define whether significant
efforts were warranted, e.g., the need for an alternative synthesis
(13) Vorbru¨ggen, H. Synthesis 2008, 1165.
(14) Price ranges for bulk quantities: 6-9 €/mol for galactose and 115-
155 €/mol for methyl R-D-galactose monohydrate. The primary
feedstock for these substrates is cow’s milk (certified for human
consumption).
(15) (a) Mowery, D. F., Jr.; Ferrante, G. R. J. Am. Chem. Soc. 1954, 76,
4103. (b) Dale, J. K.; Hudson, C. S. J. Am. Chem. Soc. 1930, 52,
2534. (c) Pater, R. H.; Coelho, R. A.; Mowery, D. F., Jr. J. Org. Chem.
1973, 38, 3272.
(16) Ratios improved slightly to 4:1 at elevated temperatures (120 °C, under
pressure) in attempts to adapt concepts from reports that achieved 12:1
R/ꢀ selectivity using microwave-assisted Fischer glycosylation
conditions: (a) Bornaghi, L. F.; Poulsen, S.-A. Tetrahedron Lett. 2005,
46, 3485. (b) Nu¨chter, M.; Ondruschka, B.; Lautenschla¨ger, W. Synth.
Commun. 2001, 31, 1277.
(12) For a recent discussion on assessing and selecting synthesis routes,
see: Parker, J. S.; Bower, J. F.; Murray, P. M.; Patel, B.; Talavera, P.
Org. Process Res. DeV. 2008, 12, 1060, and references cited
therein. For a review of process chemistry objectives, see: Zhang, T. Y.
Chem. ReV. 2006, 106, 2583.
(17) The ꢀ-anomer is not suitable for use in this instance because it impacts
selectivity in the benzoylation step and would lead to mixtures that
interfere with isolations in downstream chemistry.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
625

Table 1. Selectivity of anomeric deacylation of 10
base (equiv)
equiv
solvent
temp (°C)
time (h)
10 (A%)
6a (A%)
impuritiesa (A%)
NH2NH2 ·H2O
1.1
THF
0
35
0
22
22
22
22
35
5
45
6
0
8
0.3
2
0
23
45
1
100
87
0
NH4OAc
2.4b
THF/MeOH/H2Oc
THF/MeOHd
THF
5
NH3
7.5
1.0
1.2/1.4
1.2/1.4
1.2/1.4
1.3
96
3.7
4
NH2CH2CH2NH2
NH2CH2CH2NH2/AcOH
NH2CH2CH2NH2/AcOH
NH2CH2CH2NH2/AcOH
morpholine
3
94
100
71
52
98
THF
DCM
MTBE
toluene
21
21
21
26
0
6
3
1
a Combined debenzoylated impurities. b Only 10% conversion with 1.2 equiv of NH4OAc and absence of MeOH. c 9:8:1; no reaction in absence of MeOH. d 7:3 ratio.
Scheme 3. Tetrabenzoylation of 8b
cylations in general)18 of 8b is complicated by several factors.
First, 8b is comprised of R/ꢀ-anomers and ring-opened species,
which would generate complex mixtures of products and
provide challenges in isolation and purification. Second, 8b is
a poorly soluble compound, leading to heterogeneous reaction
conditions, at least at the onset of the reaction. As a result,
Figure 2. Benzoylating species.
although there is some differentiation between the five hydroxy
groups in the molecule,3a,19 reported yields and selectivities for
this tetrabenzoylation have been generally poor, and isolations
have often included chromatography.20
pyridinium chloride 16 is the highly reactive species (Figure
2). In contrast, the reactive species in the case of benzoic
anhydride is species 17a or perhaps the benzoyl pyridinium
benzoate 17b in which case the lack of ionic character (17a)
or steric demands lead to improved differentiation of the
hydroxy groups.19,22 However, benzoylations with benzoic
anhydride were sluggish, even at elevated temperatures, and
initially this reagent did not appear to be a viable choice from
the point of view of efficiency and processability on a pilot-
plant time scale.
To address these issues, it was found that an approach using
both benzoyl chloride and benzoic anhydride in tandem had
merit and led to improvements to the regioselective tetraben-
zoylation of 8b. The basis of the tandem approach was to use
an initial deficient amount of benzoyl chloride which would
benzoylate the more active hydroxy groups (e.g., at the 1- and
6-positions) to start the process and solubilize the substrate, then
the less reactive but more selective benzoic anhydride would
benzoylate the less active hydroxy groups and complete the
intended tetrabenzoylation.
Benzoylations of carbohydrates are traditionally conducted
using pyridine as solvent.21 Preliminary experiments showed
that regioselectivity could be improved by using NMP as solvent
with pyridine as base, presumably because of slightly improved
solubilization of 8b in NMP. A mixture of products resulted,
including the desired tetra-Bz 7b, underbenzoylated species,
designated tri-Bz 14, and the overbenzoylated species, penta-
Bz 15 (Scheme 3). Early probes suggested that benzoic
anhydride gave slightly better regioselectivity than benzoyl
chloride. The difference in reactivity of these benzoylating
reagents has been rationalized to be due to the nature of the
reacting species. In the case of benzoyl chloride, benzoyl
(18) In preliminary screens, benzoylation appeared to have more potential
for selectivity than alternative acylations, such as acetylation or
pivaloylation. Also, see: Lindhorst, T. K. Essentials of Carbohydrate
Chemistry and Biochemistry, 2nd ed.; Wiley-VCH: Weinheim, 2003;
p 49.
(19) Generally, anomeric-OH and primary-OH are most reactive, and
equatorial-OH is more reactive than axial-OH. Based on analogy with
benzoylation of methyl R-D-galactopyranoside, 6-OH > 2-OH ≈
3-OH > 4-OH. See: Haines, A. H. Relative Reactivities of Hydroxyl
Groups in Carbohydrates. In AdVances in Carbohydrate Chemistry
and Biochemistry; Tipson, R. S., Horton, D., Eds.; Academic Press:
New York, 1976; Vol. 33, pp 11-109.
(20) Reported yield after chromatography is 38%. (ꢀ-anomer not reported):
(a) Garegg, P. J.; Hultberg, H. Carbohydr. Res. 1982, 110, 261. (b)
Kova´cˇ, P.; Taylor, R. B. Carbohydr. Res. 1987, 167, 153.
(21) Pyridine (or alternative base) in this instance also serves as acyl transfer
reagent (via in situ generated benzoyl pyridinium species) and as HCl
acceptor.
Indeed, after a series of experiments, this proposal demon-
strated real potential. It should be noted that because of the
challenges galactose presented as a substrate, mixtures of the
(22) Selective acylations using acid anhydrides: (a) Horton, D.; Lauterback,
J. H. J. Org. Chem. 1969, 34, 86. (b) Chen, C.-T.; Kuo, J.-H.; Pawar,
V. D.; Munot, Y. S.; Weng, S.-S.; Ku, C.-H.; Liu, C.-Y. J. Org. Chem.
2005, 70, 1188.
626
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Table 2. Selective tetrabenzoylation of 8ba,b
BzCl Bz2O time 14
entry solvent (equiv) (equiv) (h) (A%) (A%) (A%)
previous screenings, final optimizations were performed using
design of experiments (DoE) techniques. The factors investi-
gated in a second-order central composite response surface
design were amount of BzCl (2.9-3.2 equiv), dilution (7-13
v/w parts solvent), and reaction time (14-24 h). The objective
was to maximize 7b while holding 15 to <15 A%. Results are
depicted in the three-dimensional response surface plots for the
desired 7b and 15 byproducts (Figure 3). The parameters that
had the greatest influence on selectivity were dilution and
equivalents of BzCl. The amount of 7b reached highest levels
when using 8-11 parts solvent and about 3.0-3.1 equiv of
BzCl. Levels of 15 increased with increasing amounts of BzCl
and decreasing amounts of NMP. The study predicted that use
of 10 parts of NMP and about 3.05 equiv of BzCl would achieve
the targets for selectivity, conversion, and byproduct distribution.
This prediction was confirmed by experiments (Table 2, entry
11). Finally, in a parallel DoE study, DMA as solvent gave
slightly better selectivity compared to NMP. However, due to
greater industrial hygiene limitations anticipated for DMA, NMP
was selected as the solvent for subsequent optimizations and
scale up.
15
7b
1
2
3
4
5
6
7
8
9
pyridine
3.0
1.5
0
1.5
0
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
3
3
17
5
25
25
19
25
26
19
19
14
7
8
72
40
45
19
6
15
21
42
56
62
63
64
67
71
72
72
75
butanone 5.0
CH3CN
NMP
DMPU
TMU
DMF
DMEU
NMP
2.9
4.6
2.9
2.9
2.9
2.9
2.9
2.9
3.05
3.05
8
2
15
15
6
8
16
10
13
10
9
12
5
10 DMA
11 NMPc
12 DMAc
8
7
7
7
a Conditions: 20 °C (5 °C for entries 1 and 2), 5.5 equiv of pyr (pyr as solvent
for entry 1). b Raw A% HPLC conversions. c Conditions predicted by DoE data.
desired 7b, underbenzoylated species 14,23 and overbenzoylated
15 were obtained even under the best conditions. Apart from
the selectivity of tetrabenzoylation of galactose, one of the major
challenges during process development was isolation of 7b with
purity suitable for downstream chemistry from the complex
mixture of products. Fortunately, under this tandem protocol,
high selectivity at the anomeric center was observed. The ꢀ:R
selectivity improved from about 82:18 to g97:3 as the reaction
proceeded, likely due to in situ thermodynamic anomerization
caused by benzoic acid or pyridine hydrochloride that were
generated during benzoylation.24 This facilitated higher recovery
of 7b. Screening experiments showed that a balance of
selectivity and conversion was required to facilitate isolation.
The reaction was allowed to proceed until HPLC analysis
indicated a product distribution of about 70% 7b, 10% 14, and
12% 15. Allowing the reaction to further benzoylate 14 was
counterproductive because it led to increased overbenzoylation
of 7b, and product tended to oil out during isolation when levels
of 15 exceeded 15 A%.
Further process development efforts focused on the selectiv-
ity, conversion, and a balanced reaction profile by evaluating
several parameters, including the preferred amounts of benzoyl
chloride versus benzoic anhydride, solvent, and base. The
preferred amount of total benzoylating reagent was 4.4 equiv
comprising 2.9 equiv of BzCl and 1.5 equiv of Bz2O. Perfor-
mance in a variety of solvents is summarized in Table 2. DMA
and NMP as solvents gave the highest conversions to 7b with
high selectivity for the ꢀ-anomer. BzCl as sole benzoylation
reagent in NMP gave only 56% conversion to 7b (entry 4). A
separate screening confirmed pyridine as the best base with
respect to tetra/penta selectivity, overall conversion to 7b,
anomeric selectivity, and processability. The main product
obtained with pyridine as solvent was 15.
The optimized, tandem benzoylating reaction sequence was
as follows: benzoyl chloride (3.05 equiv) was added to a
suspension of 8b, NMP and pyridine. After 2 h at 20 °C,
benzoic anhydride25 (1.5 equiv) was added. The reaction was
judged complete after g14 h at 47 °C to give a mixture
containing 5-9 A% of 14, 8-11 A% of 15, and 65-70 A%
of 7b. The process proved to be robust with consistent
performance during scale up.
The challenge that remained was isolating 7b from the rather
complex reaction mixture. This was alleviated to some extent
because the products predominately equilibrated to their ꢀ-ano-
mers under the reaction conditions. Extensive studies were
performed to find a robust isolation process to overcome the
propensity of the product to oil out of solution and to limit the
carryover of 15. Work-up consisted of adding toluene, quench-
ing with water to destroy the excess Bz2O, separating, and
washing. The organic layer was concentrated to about half
volume, and then product was crystallized by dilution with
n-heptane. Seeding was employed to crystallize the first pilot-
plant batch but was found to be unnecessary and was not
employed in subsequent batches. After drying, 7b was obtained
as a white solid in 41-46% yield corrected for g93 wt % purity
(e2% each of 14, R-tetra-Bz anomer, and 15) and with quality
suitable for downstream chemistry without need for recrystal-
lization. This process, starting with inexpensive 8b, represents
significant improvement in terms of yield, selectivity, and ease
of isolation of 7b compared with previously reported proce-
dures. Overall, the scale up of the selective tetrabenzoylation
was successful, and the reaction and isolation processes proved
to be robust: a total of 630 kg of 7b was manufactured in
consistent fashion at two separate facilities.
With the base established as pyridine, the total amount of
benzoylating agent set at ∼4.5 equiv, the target selectivity set
at <15 A% 15, and the solvents narrowed to NMP or DMA by
The next step that needed attention was fluorination of 7b
to give the corresponding 4-fluoroglucopyranose 11. A variety
of fluorinating systems, including BAST, SO2F2, TEA ·HF,
pyridine·HF, TBAF, and inorganic fluoride sources CsF and
(23) Tentative structural assignments. Attempts to isolate tri-Bz species
for characterization were unsuccessful.
(24) In addition, the steric demands of tetrabenzoylation may override
anomeric effects. For acid-catalyzed anomerization, see: (a) Pacsu,
E. Chem. Ber. 1928, 61, 1508. (b) Perrin, C. L.; Armstrong, K. B.
J. Am. Chem. Soc. 1993, 115, 6825. (c) Isbell, H. S.; Frush, H. L. J.
Org. Chem. 1958, 23, 1309. (d) See ref 16a.
(25) Bz2O can also be charged prior to BzCl, where it is an unreactive
bystander until the reaction is heated to 47 °C. Alternatively, Bz2O
can be generated in situ from BzCl and water.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
627

Figure 3. Effect of amounts of BzCl and NMP on conversion to desired product 7b and byproduct 15.
Scheme 4. Fluorination of tetrabenzoylated substrates
organic phase was treated with 1-butanol to crystallize 11 as a
white solid in 74-78% yield and >98.5 A% purity.29
1-BuOH proved to be an ideal antisolvent for the crystal-
lization,30 providing a high level of robustness in the synthesis
because it efficiently solubilized and removed olefin 20 and
other impurities that were generated in varying amounts, while
still affording a relatively high recovery of 11. For example,
relatively impure substrate 7b (85 A%) can be tolerated. The
underbenzoylated impurities that were carried over in 7b were
mostly retained in the 1-BuOH filtrate after fluorination and
isolation, and the overbenzoylated impurities were reduced by
∼5 fold to <1 A%, achieving a purity of >98 A%31 in the
isolated 11. The drawback with using less-pure 7b was the need
to compensate with additional amounts of BAST to complete
the fluorination reaction.
With the fluorinated intermediate in hand, the plan was to
adapt the selective deacylation procedure previously used for
10 to the new substrate 11 to afford glycosyl donor 6a.
However, selective anomeric debenzoylation was much slower
and more challenging compared to the anomeric acetyl group
in the former case. Thus, selective anomeric debenzoylation of
11 was studied using organic and inorganic bases, including
ethylenediamine/AcOH, methylamine, and KOH. The best
conditions (ethylenediamine/AcOH as base, 22 °C, 60 h)
afforded only a 63 A% conversion to 6a, along with 30 A%
unreacted 11 and 7 A% overdebenzoylated impurities, indicating
that debenzoylation at nonanomeric positions was competitive
with anomeric debenzoylation.
KF, were evaluated using three related potential substrates
(Scheme 4), 7b, the triflate 18,26 and the nonaflate 19.27
Analogous to the methyl galactose route, elimination to give
the corresponding olefin 2028 was a competing side reaction
observed in all cases. In some cases, product/olefin ratios were
poor, especially for the activated substrates. The best reagent
was BAST with substrate 7b (2-MeTHF, 50 °C) which provided
a 85:15 ratio of product/olefin. It was desirable to continue
screening fluorinating reagents and conditions in order to replace
BAST. However, to achieve short-term API supplies, we chose
to continue to optimize the BAST fluorination while further
development of alternatives was explored in parallel.
Dioxane as solvent gave the best product/olefin ratio, 87:
13. However, because of the lack of significant superiority over
other ether solvents and because of process safety (peroxide
formation) and industrial hygiene issues, this solvent was not
pursued further. 2-MeTHF was selected as the preferred solvent
for scale up compared to other solvents screened due to the
relatively good product/olefin distribution (85:15) and its
suitability for extractive workup. BAST was added at 20 °C,
and then the temperature was linearly increased to 55 °C over
2 h. The rationale for this heating profile was to provide a
reasonable balance between reaction rate and olefin formation.
The fluorination was complete after 3 h at 55 °C. After
quenching with water and an extractive workup, the final
Because the direct conversion of 11 to 6a was inefficient,
synthesis alternatives were desired. It was found that 11 can be
converted cleanly to either 10 (Ac2O/AcOH/H2SO4, 95% yield)
(29) Fluorination occurred highly selectively with inversion at C-4. None
of the 4-epimer was detected in the product or the filtrate compared
to an authentic sample of the epimer synthesized independently.
(30) 2-PrOH provided a similar performance, but was not employed in any
other step in the synthesis of 1 in the route starting from 8b. In contrast,
1-BuOH was determined to be essential in a subsequent step in
downstream chemistry. Therefore, 1-BuOH was selected as the
preferred solvent because it minimized the total number of solvents
employed in the overall synthesis and enhanced the opportunity for
solvent recovery.
(26) Prepared by treating 7b with Tf2O in DCM in the presence of pyridine.
(27) Prepared by treating 7b with perfluorobutanesulfonyl fluoride in the
presence of DBU.
(28) Olefin 20 was confirmed by comparison to an authentic sample
prepared from 18 by treatment with DBU in DCM and was consistent
1
with reported H NMR data: Blattner, R.; Ferrier, R. J.; Tyler, P. C.
(31) Isolated 11 needed to have a purity g98 A% to ensure acceptable
quality of downstream products.
J. Chem. Soc., Perkin Trans. I 1980, 1535.
628
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

Scheme 5. Indirect conversion of 11 to 6a
suffered from CoG issues, primarily due to high material costs
for the R-methyl galactose (8a) starting material. As a result, a
second glycosyl donor, 6b, was developed starting from the
more economical substrate, galactose (8b). A key aspect in the
synthesis of 6b was the development of a highly selective
tetrabenzoylation of galactose using benzoyl chloride and
benzoic anhydride as reagents in tandem. The new synthesis
reproduced lab-scale expectations, and no major issues
were encountered during scale up. Overall, a total of 157
kg of 6b was prepared, which was subsequently coupled
with 2 to provide 1. The estimated cost to manufacture
6b was significantly less than that for 6a by a factor of 3,
and as a result, CoG was no longer deemed a limitation
in the manufacture of 1.
or 20 (TFA/trifluoroacetic anhydride, cat. TfOH, 97 A% yield),
and that these intermediates could be then converted in high
yield to 6a (Scheme 5). Although these variations showed
synthetic utility, the improvements in yield were offset by the
extra processing steps that were entailed.
Glycosyl donor 6a was shown to be a reliable synthon for
the synthesis of SAR7226 in the first pilot-plant campaigns.
Under standard coupling conditions,32 6a was treated with an
excess of trichloromethylacetonitrile, and the resulting acetimi-
date was coupled with pyrazole 2 in the presence of a Lewis
acid. A major limitation was waste/environmental concerns due
to the formation of trichloroacetamide as a byproduct of the
acetimidate coupling. Therefore, in parallel, efforts were made
to identify a more suitable alternative glycosyl donor.10 In the
meantime, the coupling of the bromo sugar 6b by reaction with
alkoxides had shown feasibility in the synthesis of a back-up
compound, and a decision was made to investigate and develop
6b for the large-scale synthesis of 1.
On lab scale, 6b was prepared by treatment of 11 in DCM
with HBr in AcOH.2 On scale up, AcOH replaced DCM as
solvent. The process was optimized for equivalents of HBr,
volume of AcOH solvent, and temperature. Quantitative con-
version to 6b was achieved using 4 equiv of HBr in AcOH at
55 °C (to improve solubility of the substrate). Work-up consisted
of a water quench followed by partitioning with toluene. The
organic phase was given a series of washes to remove AcOH
and benzoic acid and then was azeotropically dried by partial
concentration under reduced pressure. Two methods of isolation
were developed. After partial concentration, 6b could be isolated
in 87-91% yield and >99 A% purity as a white solid by
crystallization from toluene/n-heptane. For industrial hygiene
and yield/efficiency reasons, after partial concentration, 6b was
isolated as a solution in 2-MeTHF/toluene in g96% yield
(g96.8 A% purity, exclusively R-anomer). This solution was
directly used in the pilot plant in the coupling step as part of a
successful campaign that demonstrated the utility of 6b in the
synthesis of 1. The estimated cost to manufacture 6b was 80%
of the limit target cost for the glycosyl donor needed for the
production of SAR7226, about 3-fold lower compared to that
for 6a on a mole-to-mole basis, primarily because its manu-
facture was more concise and it was prepared from a more
economical sugar substrate.
Experimental Section
Methyl 2,3,6-Tri-O-benzoyl-r-D-galactopyranoside (7a).3a,b
Benzoyl chloride (74.4 kg, 529 mol) was added over 2 h to a
suspension of 8a (24.0 kg, 113 mol) in pyridine (160 kg) at 9
( 3 °C. The mixture was heated at 18 ( 3 °C for 4 h. The
reaction was cooled to 10 °C, and water (190 kg) was added
over 30 min while allowing the temperature to rise to 21 ( 3
°C. The mixture was stirred for g1 h to quench the benzoyl
chloride-pyridine complex. Toluene (190 kg) was added, and
after 15 min, the phases were separated. The organic layer was
washed with 8% aq NaHCO3 (2 × 281 kg), 1 N HCl (2 × 190
kg), and then with water (190 kg) at 25 °C. The toluene phase
(249 kg) was partially concentrated (70-170 Torr, jacket
temperature <90 °C) by collecting 51 kg distillate. After venting
with nitrogen, toluene (33 kg) was added to give a target
concentration of approximately 3.5 v/w parts of toluene based
on the theoretical yield of product. The batch was adjusted to
70 °C and n-heptane (91 kg) was added at 70 ( 3 °C over 34
min. The solution was cooled to 61 °C over 20 min, at which
point crystallization was confirmed. The mixture was held at
about 61 °C for 20 min and then was cooled to 25 °C over 2 h
and held for 1 h. After filtration, washing with a mixture of
n-heptane (31 kg) and toluene (60 kg), and drying (<100 Torr,
45 °C), 7a was obtained as a white solid (40.5-43.5 kg,
71-76% yield, g 98.8 A% by HPLC33). 1H NMR (500 MHz,
CDCl3, δ): 2.85 (br s, 1H, OH), 3.45 (s, 3H, OCH3), 4.35
(apparent t, J ) 6.4 Hz, 1H, H-5), 4.41 (m, 1H, H-4), 4.58-4.68
(m, 2H, H-6), 5.22 (d, J ) 3.3 Hz, 1H, H-1), 5.69-5.73 (m,
2H, H-2 and H-3), 7.33-7.56 (m, 9H, Ar-H), 7.96-8.05 (m, 6H,
Ar-H).
Methyl 2,3,6-Tri-O-benzoyl-4-deoxy-4-fluoro-r-D-glu-
copyranoside (9).2 Bis(2-methoxyethyl)amino sulfur trifluoride
(BAST, 54 kg, 244 mol) was added over 47 min from a PFA-
lined pressure tank to 7a (115 kg, 227 mol) in THF (414 kg)
in a Hastelloy C276 reactor while maintaining 20 ( 5 °C. The
resulting solution was heated and held at 52 °C for 3 h. The
mixture was cooled to 10 °C, and water (460 kg) was added
(initially slowly) over 50 min while maintaining <25 °C.
Toluene (400 kg) was added, and the mixture stirred for 15
min. The organic layer was separated and washed with water
Conclusions
Process development efforts towards two glycosyl donors
applicable to the synthesis of SAR7226 were reported. The first
glycosyl donor, 6a, was prepared in four chemical steps in about
32% overall yield. Although this synthesis proved to be scalable
and was used to prepare 90 kg of SAR7226 (1), the route
(33) Zorbax Eclipse C8, 4.6 mm × 150 mm, 3.5 µ, 220 nm, 35 °C, water/
ACN/TFA, 1 mL/min, gradient program: 40/60/0.1 for 2 min, linear
ramp over 13 min to 10/90/0.1. Relative retention times: tetrabenzoy-
lated adduct of 8a, 0.95; 7a, 1.00.
(32) Schmidt, R. R.; Michel, J.; Roos, M. Liebigs Ann. Chem. 1984, 1343.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
629

(180 kg) and 6% aq NaHCO3 (390 kg), followed by water (2
× 180 kg). The toluene solution was partially concentrated
(70-170 Torr, jacket temperature <90 °C) by collecting 500
kg of distillate. After releasing the vacuum with nitrogen, the
batch was adjusted to 60 °C to ensure dissolution of solids; a
sample was taken to confirm solvent composition (e5 wt %
THF relative toluene). The batch was cooled to 45 °C, and
n-heptane (150 kg) was added over 1 h at 40 ( 5 °C. The batch
was cooled to 20 ( 5 °C over 40 min, held at 20 ( 5 °C for
1 h, and then was filtered, washed with n-heptane (150 kg),
and dried (<100 Torr, 45 °C) to give 9 as a white solid (91.7
kg; 79.4% yield, 97.2 A% by HPLC34). 1H NMR (500 MHz,
CDCl3, δ): 3.47 (s, 3H, OCH3), 4.33 (m, 1H, H-5), 4.60-4.74
(m, 2H, H-6), 4.78 (ddd, J ) 50.0, 10.0, 10.0 Hz, 1H, H-4),
5.17 (apparent t J ) 3.5 Hz, 1H, H-1), 5.21 (dd, J ) 10.1, 3.6
Hz, 1H, H-2), 6.12 (ddd, J ) 15.5, 9.0, 9.0 Hz, 1H, H-3),
7.35-7.60 (m, 9H, Ar-H), 7.98-8.11 (m, 6H, Ar-H); 19F NMR
(282 MHz, CDCl3, δ): -196.52 (dd, J ) 50.9, 15.7 Hz).
1-O-Acetyl-2,3,6-tri-O-benzoyl-4-deoxy-4-fluoro-r-D-glu-
cose (10).2 Acetic acid (74.0 kg, 1233 mol) was charged over
10 min to a suspension of 9 (130.0 kg, 255 mol) in acetic
anhydride (507 kg, 4966 mol) while maintaining 20 ( 3 °C.
Sulfuric acid (96%, 78 kg, 764 mol) was added over 42 min
while maintaining <35 °C. The reaction was gradually heated
to 50 °C and held for 1 h. The mixture was cooled to 44 °C,
diluted with toluene (676 kg), and then cooled to 16 °C. Water
(300 kg) was carefully added over 49 min while maintaining
<28 °C. The mixture was stirred for 1 h at 20 °C. The organic
layer was separated, partitioned with ethyl acetate (210 kg) and
water (330 kg), and then stirred for 30 min at 23 °C. The organic
layer was separated and washed with 6% aq NaHCO3 (690 kg,
then 350 kg) and water (330 kg), then was concentrated
(50-150 Torr, jacket temperature <80 °C) to a volume of about
325 L. The vacuum was released using nitrogen. The batch
was cooled, and 2-propanol (830 kg) was added over 56 min
at 30-40 °C. The suspension was cooled to 20 °C over 30
min, then to 0 °C over 2 h and held for 40 min. After filtering,
washing with 2-propanol (210 kg), and drying (<100 Torr, 45
°C), 10 was obtained as a white product (73-80 kg, 53-58%
yield, g99.2 A% by HPLC35) as a ∼9:1 or better mixture of
R-/ꢀ-anomers.36 1H NMR (300 MHz, CDCl3, δ): 2.20 (s, 3H,
CH3), 4.40-4.45 (m, 1H, H-5), 4.62-4.73 (m, 2H, H-6), 4.85
(ddd, J ) 50.4, 9.6, 9.3 Hz, 1H, H-4), 5.42 (dd, J ) 10.5, 3.6
Hz, 1H, H-2), 6.09-6.19 (m, 1H, H-3), 6.55 (m, 1H, H-1),
7.33-7.59 (m, 9H, Ar-H), 7.90 (d, 2H, J ) 8.4 Hz, Ar-H),
8.00 (d, 2H, J ) 8.4 Hz, Ar-H), 8.07 (d, 2H, J ) 8.4 Hz, Ar-
H); 19F NMR (282 MHz, CDCl3, δ): -196.20 (dd, J ) 50.8,
14.1 Hz).
2,3,6-Tri-O-benzoyl-4-deoxy-4-fluoro-D-glucose (6a).2 Ace-
tic acid (11.5 kg, 191 mol) was added to a solution of 10 (73.1
kg, 136 mol) in THF (482 kg) at about 0 °C. Ethylenediamine
(9.9 kg, 164 mol) was added over 20 min while maintaining
0-6 °C. The mixture was heated to 20 °C over 30 min and
then maintained at about 20 °C for 15 h. The reaction was
cooled to 0 °C and then diluted with toluene (310 kg). Water
(370 kg) was added over 19 min while keeping the temperature
<20 °C. The batch was stirred for 20 min at 20 °C, and then
the phases were allowed to separate. The organic layer was
washed with 1 N hydrochloric acid (180 kg), water (2 × 290
kg), 5% aq NaHCO3 (380 kg), and water (2 × 370 kg) to
achieve pH e 7 for the final water wash. The organic layer
was partially concentrated (50-150 Torr, jacket temperature
<80 °C) by collecting about 325 kg distillate. The batch was
cooled, and toluene (950 kg) was added. Residual water and
THF were removed by collecting 506 kg of distillate. The
resulting solution, containing 0.3 wt % THF (target: e0.1 wt
%), was diluted with 352 kg of toluene and then partially
concentrated by collecting 371 kg of distillate. After venting
and cooling to 20 °C, the resulting solution contained e0.1 wt
% THF, e0.05 wt % water (KF), and 67.4 kg of 6a (theoretical
yield,37 >99.5 A% HPLC38), and was used directly in the
coupling step. R-Anomer (major): 1H NMR, (500 MHz,
DMSO-d6, δ): 4.48-4.68 (m, 3H, H-5, H-6), 5.05 (ddd, J )
50.9, 9.7, 9.5 Hz, 1H, H-4), 5.20 (m, 1H, H-2), 5.49 (apparent
q, J ) 4.2, 3.6 Hz, 1H, H-1), 5.99 (ddd, J ) 14.7, 9.7, 9.5 Hz,
1H, H-3), 7.45-7.71 (m, 9H, Ar-H), 7.54 (m, 1H, OH),
7.84-8.08 (m, 6H, Ar-H); 19F NMR (282 MHz, DMSO-d6,
1
δ): -198.75 (dd, J ) 50.9, 14.7 Hz). ꢀ-Anomer (minor): H
NMR, (400 MHz, DMSO-d6, δ): 4.48-4.68 (m, 3H, H-5, H-6),
5.01 (ddd, J ) 50.9, 9.7, 9.5 Hz, 1H, H-4), 5.19 (m, 1H, H-2),
5.21 (m, 1H, H-1), 5.93 (ddd, J ) 14.7, 9.2, 8.6 Hz, 1H, H-3),
7.45-7.71 (m, 9H, Ar-H), 7.52 (m, 1H, OH), 7.84-8.08 (m,
6H, Ar-H); 19F NMR (376 MHz, DMSO-d6, δ): -199.50 (dd,
J ) 50.9, 14.7 Hz).
1,2,3,6-Tetra-O-benzoyl-ꢀ-D-galactopyranose (7b). Ben-
zoyl chloride (105 kg, 747 mol) was added over a period of 72
min to a suspension of D-galactose (44.0 kg, 244 mol), NMP
(458 kg), and pyridine (107.0 kg, 1352 mol) while maintaining
a temperature of 18-22 °C.39 The reaction has held at about
20 °C for 2 h, and then a solution of benzoic anhydride (83.6
kg, 369 mol) in NMP (37 kg) was added over 31 min at 18-22
°C. The suspension (pyridine ·HCl had precipitated) was held
at 18-22 °C for 1 h, and was then warmed to 47 °C over 0.5 h
to give a pale-yellow solution. After 26 h at 47 °C, the reaction
was judged complete (<9.0 A% tribenzoylated species re-
mained), and the solution was cooled to 22 °C. The mixture
was diluted with toluene (190 kg), and then water (290 kg)
was added over 30 min while maintaining <30 °C. The
batch was warmed to 47 °C and held for 1 h to hydrolyze the
(34) Zorbax Eclipse C8, 4.6 mm × 150 mm, 3.5 µ, 230 nm, 35 °C, water/
ACN/TFA, 1 mL/min, gradient program: 40/60/0.1 for 8 min, linear
ramp over 1 min to 30/70/0.1. Relative retention times: 7a, 0.65; 9,
1.00.
(35) Zorbax Eclipse C8, 4.6 mm × 150 mm, 3.5 µ, 220 nm, 35 °C, water/
ACN/TFA, 1 mL/min, gradient program: linear ramp over 11 min
from 40/60/0.1 to 30/70/0.1. Relative retention times: ꢀ-10, 0.94; 10,
1.00; 9, 1.12.
(36) The reported NMR data correspond to the R-anomer. Important
chemical shifts for the minor ꢀ-anomer are: 1H NMR: δ 2.02 (s, 3H,
CH3), 6.02 (m, 1H, H-1); 19F NMR: -199.26 (dd, J ) 51.9, 14.1
Hz).
(37) HPLC assay data (wt/wt) from lab-scale batches consistently demon-
strated quantitative yield for this process.
(38) Zorbax Eclipse C8, 4.6 × 150 mm, 3.5 µ, 230 nm, 35 °C, water/
ACN/TFA, 1 mL/min, gradient program: linear ramp over 20 min
from 50/50/0.1 to 15/85/0.1. Retention times: 6a anomers, 9.5 and
10.6 min; 10, 13.6 min.
(39) Adding BzCl at lower temperatures did not improve selectivity. Best
selectivity was observed with fast addition (e1 h on pilot-plant
scale).
630
•
Vol. 14, No. 3, 2010 / Organic Process Research & Development

excess benzoic anhydride. After cooling to 20 °C, the phases
were separated. The organic phase was washed with 6% aq
NaHCO3 (2 × 370 kg), diluted with toluene (150 kg) and
washed with water (3 × 350 kg). The toluene phase was
separated and azeotropically dried by collecting about 300 kg
of distillate (70-150 Torr, jacket temperature <80 °C). After
the vacuum was released with nitrogen, the batch was diluted
with toluene (277 kg). The solution was heated to 75 °C, and
n-heptane (97 kg) was added over a period of 25 min at 70-80
°C. The solution was cooled to 50 °C and held for 3 h.
Crystallization commenced within 30 min. The suspension was
cooled to 10 °C over a period of 5 h, held at 10 °C for 3 h,
cooled to -10 °C over g2 h, held at -10 °C for 2 h, and then
was warmed to 20 °C over 30 min and held for 3 h. After
filtering, rinsing with a mixture of toluene (110 kg) and
n-heptane (36 kg), and drying (<100 Torr, 50 °C), 7b was
obtained as a white solid (59.7-65.6 kg corrected for ∼94 A%
HPLC,40 41-45% yield). NMR (500 MHz, DMSO-d6, δ): 4.39
(dd, J ) 5.8, 3.4 Hz, 1H, H-4), 4.53 (m, 2H, H-6), 4.59 (m,
1H, H-5), 5.68 (dd, J ) 10.6, 3.4 Hz, 1H, H-3), 5.91 (dd, J )
10.6, 8.2 Hz, 1H, H-2), 5.95 (br d, 1H, OH), 6.39 (d, J ) 8.2 Hz,
1H, H-1), 7.40-7.70 (m, 12H, Ar-H), 7.85-8.00 (m, 8H, Ar-H).
1,2,3,6-Tetra-O-benzoyl-4-deoxy-4-fluoro-ꢀ-D-glucose (11).
A mixture of 7b (83 kg, 78 kg corrected for 94 A% purity,
131 mol) and 2-MeTHF (190 kg) in a Hastelloy C276 reactor
was heated at 45 °C until dissolution was achieved, and then
the solution was cooled to 20 °C. BAST (37.4 kg, 169 mol)
was added over a period of e60 min while maintaining 15-25
°C. The reaction mixture was heated to 52 °C over 2 h and
then held at 52 °C for 2-3 h. The mixture was cooled to 20
°C and then carefully quenched with water (180 kg) over 45
min while maintaining 15-25 °C. The phases were separated.
The organic layer was washed for 30 min with 6% aq NaHCO3
(390 kg). The aqueous phase was diluted with 5% aq NaCl
(100 kg) prior to separation. The organic phase was diluted with
2-MeTHF (71 kg) and washed with 5% aq NaCl (190 kg). The
resulting milky solution was heated to 57 °C. 1-Butanol (440
kg) was added over 2 h to crystallize the product while
maintaining 57-60 °C. After cooling to 20 °C over 2 h, further
cooling and holding at 5 °C for 1 h, filtering, washing with
1-butanol (2 × 60 kg, 15 °C) and drying (<100 Torr, 50 °C),
58-61 kg of 11 was obtained as a white solid (74-78% yield,
>98.5 A% by HPLC41). 1H NMR (300 MHz, CDCl3, δ): 4.30
(dddd, J ) 9.5, 4.5, 2.5, 2.5 Hz, 1H, H-5), 4.63 (ddd, J ) 12.5,
4.3, 1.0 Hz, 1H, H-6), 4.75 (ddd, J ) 12.5, 2.5, 1.5 Hz, 1H,
H-6′), 4.93 (ddd, J ) 50.1, 9.5, 9.0, 1H, H-4), 5.77 (dd, J )
9.5, 8.3 Hz, 1H, H-2), 5.98 (ddd, J ) 14.1, 9.5, 9.0 Hz, 1H,
H-3), 6.23 (d, J ) 8.3 Hz, 1H, H-1), 7.30-7.60 (m, 12H, Ar-
H), 7.95 (d, J ) 7.8 Hz, 2H, Ar-H), 8.02 (apparent t, J ) 7.8
Hz, 4H, Ar-H), 8.05 (d, J ) 7.8 Hz, 2H, Ar-H); 19F NMR (282
MHz, CDCl3, δ): -199.00 (dd, J ) 50.8, 14.1 Hz).
2,3,6-Tri-O-benzoyl-4-deoxy-4-fluoro-r-D-glucopyrano-
syl Bromide (6b).3c A solution of hydrogen bromide (95 kg of
33 wt % HBr in AcOH, 388 mol, 4.0 equiv) was charged in 22
min to a suspension of 11 (58 kg, 96.9 mol) and AcOH (180
kg) while maintaining 20-30 °C. The mixture was heated to
55 °C over about 1 h and then held at 55 °C for 4 h. The
suspension was cooled, and water (410 kg) was added over 60
min while maintaining <20 °C. Toluene (626 kg) was added,
the temperature was adjusted to 20 °C, and the phases were
separated. The organic layer was washed with water (2 × 300
kg) and 6% aq NaHCO3 (410 kg). The organic layer was
washed with water (2 × 300 kg, to pH e7) and then
concentrated (70-170 Torr, jacket T <80 °C) to a volume of
about 260 L. After venting the reactor with nitrogen and cooling
to 20 °C, the resulting thick slurry was dissolved in 2-MeTHF
(230 kg) to give a solution containing 51.9 kg of 6b (as
determined by wt % HPLC analysis, g 96.8 A% HPLC,42
96.1% yield). 1H NMR (400 MHz, CDCl3, δ): 4.60-4.80 (m,
3H, H-5, H-6), 4.89 (ddd, J ) 50.5, 9.6, 9.0 Hz, 1H, H-4),
5.23 (ddd, J ) 10.1, 4.2, 0.9 Hz, 1H, H-2), 6.21 (ddd, J )
13.5, 10.0, 9.3 Hz, 1H, H-3), 6.76 (br dd, J ) 3.9, 3.1 Hz, 1H,
H-1), 7.39-7.64 (m, 9H, Ar-H), 7.99-8.11 (m, 6H, Ar-H); 19F
NMR (376 MHz, CDCl3, δ): -197.00 (dd, J ) 50.5, 13.5 Hz).
Acknowledgment
We are grateful to the many colleagues within sanofi-aventis
who provided technical, process safety, analytical, sourcing, and
CoG support.
Note Added after ASAP Publication: This paper was
published on the Web on May 3, 2010, with errors in Figure
2. The corrected version was reposted on May 6, 2010.
Received for review February 17, 2010.
OP100053K
(40) Zorbax Eclipse XDB-C8, 4.6 mm × 150 mm, 3.5 µ, 230 nm, 35
°C, water/ACN/TFA, 1 mL/min, gradient program: linear ramp over
5 min from 50/50/0.1 to 45/55/0.1, then linear ramp over 15 min
to 10/90/0.1. Relative retention times: 14, 0.51; 7b, 1.00; 15,
1.32.
(41) Zorbax Eclipse XDB-C8, 4.6 mm × 150 mm, 3.5 µ, 230 nm, 35 °C,
water/ACN/TFA, 1 mL/min, gradient program: 40/60/0.1 for 1 min,
linear ramp over 16 min to 20/80/0.1. Relative retention times: 7b,
0.73; 11, 1.00.
(42) Zorbax Eclipse XDB-C8, 4.6 mm × 150 mm, 3.5 µ, 230 nm, 35 °C,
water/ACN/TFA, 1 mL/min, gradient program: linear ramp over 16
min from 50/50/0.1 to 20/80/0.1. Relative retention times: 6b, 1.00;
11, 1.06.
Vol. 14, No. 3, 2010 / Organic Process Research & Development
•
631