Efficient Preparation of a Key Intermediate in the Synthesis of Roxifiban by Enzymatic Dynamic Kinetic Resolution on Large Scale

Article abstract of DOI:10.1021/op0300239

Additional information is presented for the transformation of a kinetic resolution into a dynamic kinetic resolution of the isobutyl ester 5b to form the acid 2 in high yield and ee via the intermediacy of a thioester 6c, (Pesti, J. A.; Yin, J.; Zhang, L.-H; Anzalone, L. J. Am. Chem. Soc. 2001, 123, 11075-11076.). The development of optimized reaction conditions for the preparation of 6c, its dynamic kinetic resolution to 2, and the scale-up of both reactions into a pilot plant are described.

Full text of DOI:10.1021/op0300239

Organic Process Research & Development 2004, 8, 22−27
Efficient Preparation of a Key Intermediate in the Synthesis of Roxifiban by
Enzymatic Dynamic Kinetic Resolution on Large Scale¶
Jaan A. Pesti,* Jinguao Yin,† Lin-hua Zhang,‡ Luigi Anzalone,| Robert E. Waltermire, Philip Ma, Edward Gorko,
Pat N. Confalone,∇ Joseph Fortunak, Charlotte Silverman,O John Blackwell,# J. C. Chung,§ Michael D. Hrytsak,
Mary Cooke, Lakisha Powell, and Charles Ray
Process Research and DeVelopment, Bristol-Myers Squibb Pharmaceutical Research Institute, PO Box 4000,
Princeton, New Jersey 08543-4000, U.S.A.
Abstract:
and dampen the typical peak/valley pharmacokinetic profile
seen for other therapies. This property, among others,
encouraged us to develop a safe, efficient, and inexpensive
large-scale process to prepare roxifiban to advance it into
the clinic.2,3 Herein we describe the development of a
practical process for the preparation of the key intermediate
2, including the transition from a kinetic to a dynamic kinetic
resolution and the scale-up of this chemistry in the pilot
plant.4
Additional information is presented for the transformation of
a kinetic resolution into a dynamic kinetic resolution of the
isobutyl ester 5b to form the acid 2 in high yield and ee via the
intermediacy of a thioester 6c, (Pesti, J. A.; Yin, J.; Zhang, L.-
H; Anzalone, L. J. Am. Chem. Soc. 2001, 123, 11075-11076.).
The development of optimized reaction conditions for the
preparation of 6c, its dynamic kinetic resolution to 2, and the
scale-up of both reactions into a pilot plant are described.
Recent reports have identified roxifiban (1) as a potent
and selective antagonist of the platelet glycoprotein IIb/IIIa
receptor.1 This isoxazoline derivative holds strong promise
for the prevention and treatment of a wide variety of
thrombotic diseases resulting from undesired platelet adhe-
sion such as stroke, acute myocardial infarction, transient
ischemic attack, and unstable angina. The efficacy of existing
anti-platelet therapies is well established. However, none
possesses all of the key attributes desired by clinicians,
namely selectivity for the platelet, freedom from adverse
effects, activity versus all agonists, rapid onset of action,
and oral activity. The ability of roxifiban to bind to both
activated and resting platelets is unlike other known IIb/IIIa
antagonists. This reservoir of 1 residing on unactivated
platelets is expected to result in increased anti-platelet activity
Results and Discussion
Background. The chiral acid 2 has been previously
resolved by chiral preparative HPLC, by fractional crystal-
lization of its cinchonidine salts,5 or via an enzyme-mediated
kinetic resolution of 5a, itself prepared by a highly efficient
[3 + 2] cycloaddition (Scheme 1).1 As the n-butyl butenoate
4a was not readily available, we found a better alternative
to be the isobutyl ester 4b which is inexpensive and available
in large quantities. This reaction regiospecifically produced
5b and was smoothly scaled up in the pilot plant.2,3
Fortuitously, the conditions previously established for
enzymatic resolution of the n-butyl ester 5a were also
effective for the i-butyl ester 5b.2,3,5 The R-enantiomer was
selectively hydrolyzed, isolated, and purified after separation
from the remaining S-ester by simple filtration, eventually
on large scale in our pilot plant. The unreacted S-isomer was
subsequently racemized and returned for another cycle of
¶ This work was conducted while the authors were part of the DuPont
Pharmaceutical Company and its predecessor, the DuPont Merck Pharmaceutical
Company.
* To whom correspondence should be addressed. Telephone: (609) 252-3703.
Fax: (609) 252-7825. E-mail: jaan.pesti@bms.com.
† Current address: Schering-Plough Research Institute, Chemical Process
Research & Development, 1011 Morris Ave., U-13-2000 Union, NJ 07083.
‡ Current address: Gilead Sciences, Inc., 333 Lakeside Drive, Foster City,
CA 94404.
| Current address: Johnson & Johnson, Pharmaceutical Research & Develop-
ment, L.L.C., Welsh and McKean Roads, Spring House, PA 19477.
3 Current address: EI duPont de Nemours and Company, Dupont Agricultural
Products, Stine/Haskell Laboratories, Newark, DE 19714.
Current address: Abbott Labs, Global Process R&D, 1401 Sheridan Road,
North Chicago, IL 60064-6291.
O Current address: Analytical Development, AstraZeneca Pharmaceuticals,
NLW2, 1800 Concord Pike, Wilmington, DE 19850-5437.
# Current address: Rhodia Pharma Solutions, 383 Phoenixville Pike, Malvern,
PA 19355.
§ Current address: Boehringer Ingelheim Pharmaceuticals, Inc., 900 Ridgebury
Road, Ridgefield, CT 06877.
(2) Zhang, L.-h; Chung, J. C.; Costello, T. D.; Valvis, I.; Ma, P.; Kauffman,
G. S.; Ward, R. J. Org. Chem. 1997, 62, 2466-2470.
(1) (a) Xue, C.-B.; Mousa, S. A. Drugs Future 1998, 23, 707-711. (b) Xue,
C.-B.; Wityak, J.; Sielecki, T. M.; Pinto, D. J.; Batt, D. G.; Cain, G. A.;
Sworin, M.; Rockwell, A. L.; Rodercik, J. J.; Wang, S.; Orwat, M. J.; Frietze,
W. E.; Bostrom, L. L.; Liu, J.; Higley, C. A.; Rankin, F. W.; Tobin, A. E.;
Emmett, G.; Lalka, G. K.; Sze, J. Y.; Di Meo, S. V.; Mousa, S. A.; Thoolen,
M. J.; Racanelli, A. L.; Hausner, E. A.; Reilly, T. M.; DeGrado, W. F.;
Wexler, R. R.; Olson, R. E. J. Med. Chem. 1997, 40, 2064-2084.
(3) Zhang, L.-h; Anzalone, L.; Ma, P.; Kauffman, G. S.; Storace, L.; Ward, R.
Tetrahedron Lett. 1996, 37, 4455-4458.
(4) Parts of this work have been previously published: Pesti, J. A.; Yin, J.;
Zhang, L.-h; Anzalone, L. J. Am. Chem. Soc. 2001, 123, 11075-11076.
(5) Wityak, J.; Sielecki, T. M.; Pinto, D. J.; Emmett, G.; Sze, J. Y.; Liu, J.;
Tobin, E.; Wang, S.; Jiang, B.; Ma, P.; Mousa, S. A.; Wexler, R. A.; Olson,
R. E. J. Med. Chem. 1997, 40, 50-60.
22
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development
10.1021/op0300239 CCC: $27.50 © 2004 American Chemical Society
Published on Web 12/24/2003

Scheme 1. Current method for resolution of 22
Scheme 2. Retro-Michael/Michael racemization mechanism of (S)-6 in the presence of base4
resolution. While efficient and very amenable to further scale-
up, this protocol of enantiomer separation, racemization,
purification, and enzymatic resolution recycle was inconve-
nient and lengthy. We therefore decided to streamline the
process by searching for a dynamic kinetic resolution.
While dynamic kinetic resolutions are not trivial to
develop,6 such a change would offer other advantages beyond
eliminating the need to recycle (S)-5b. For most kinetic
resolutions, the enantiomeric excess of the product (eep)
begins to deteriorate as the conversion approaches 50%. This
occurs since the ratio of the less-reactive to the more-reactive
enantiomer will continue to increase until the advantage of
the enzyme’s discrimination between the enantiomers is
overcome. At that point reaction products derived from the
less-reactive isomer begin to predominate, and the eep
deteriorates. Establishment of a dynamic kinetic resolution
should maintain a high eep at all times7 and avoid the
degradation in eep experienced by dynamic kinetic resolutions
at J50% conversions.
Development of the Thioester Resolution to 2. All of
our efforts to achieve the dynamic kinetic resolution of 5b
by manipulation of reaction conditions (higher pH) or the
substrate (appending electron-withdrawing groups) in the
presence of lipases were unsuccessful; 55% eep was not
exceeded in any instance. Instead we decided to consider
the use of thioesters, suggested by the publication of the first
examples of the dynamic kinetic resolution of thioesters.8
These reactions proceed by the exploitation of their ability
to lower the pKa of the adjacent protons, sometimes suf-
ficiently so as to permit facile racemization. While these
examples also required the presence of an R-thiophenyl
substituent to sufficiently lower the pKa,9 such a correspond-
ing change within our chemistry would have been unreason-
able for a viable process. Instead, we have previously
established that the racemization of 5b can occur under
relatively mild basic conditions.2,10 This is due to a facile
retro-Michael/Michael reaction equilibrium between the
isoxazoline ring and derived acrylate.4 We anticipated that
the combination of this ring-cleavage equilibrium and the
conversion of the ester to a thioester 6 would enable
racemization to occur under conditions mild enough to also
support enzymatic resolution (Scheme 2). This premise
proved to be true.
A useful conversion of 5b to 2 via thioesters would require
several prerequisites: the discovery of an efficient thioester/
enzyme couple, the establishment of thioester racemization
conditions, optimization of the best combination of thioester/
enzyme/racemization conditions, and ultimately an excellent
preparation of the selected thioester to overcome the inherent
disadvantage of an extra synthetic step between 5b and 2.
Each of these components was studied individually to allow
an orderly approach to developing an efficient enzymatic
dynamic kinetic resolution.
A variety of thioesters were quickly prepared by the
conversion of racemic 2 to the acid chloride, followed by
reaction with the copper salts of the requisite thiols.11,12
(Table 1). We screened these thioesters against 21 enzymes
(see Experimental Section for further details) and looked for
combinations that produced both high conversion and eep.13
(6) (a) Ward, R. S. Tetrahedron: Asymmetry 1995, 6, 1475-1490. (b) Noyori,
R.; Tokunaga, M.; Kitamure, M. Bull. Chem. Soc. Jpn. 1995, 68, 36-56.
(c) Faber, K. Chem. Eur. J. 2001, 7, 5004-5010. (d) Caddick, S.; Jenkins,
K. Chem. Soc. ReV. 1996, 25, 447-456. (e) Gihani, M. T. E.; Williams, J.
M. J. Curr. Opin. Chem. Biol. 1999, 3, 11-15.
(9) The authors later considerably broadened the scope of the new reaction:
Um, P.-J.; Drueckhammer, D. G. J. Am. Chem. Soc. 1998, 120, 5605-
5610.
(7) (a) Chen, C. S.; Fujimoto, Y.; Girdaukas, G.; Sih, C. J. Am. Chem. Soc.
1982, 104, 7294-7299. (b) Fulling, G.; Sih, C. J. Am. Chem. Soc. 1987,
109, 2845-2846.
(8) Tan, D. S.; Gunter, M. M.; Drueckhammer, D. G. J. Am. Chem. Soc. 1995,
117, 9093-9094.
(10) See ref 4/Supporting Information for discussion of the mechanism of this
racemization.
(11) Reissig, H.-U.; Scherer, B. Tetrahedron Lett. 1980, 21, 4259-4262.
(12) Adams, R.; Reifschneider, W.; Ferretti, A. Organic Syntheses; Wiley &
Sons: New York, 1973; Collect. Vol. 5, pp 107-110.
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
23

Table 1. Thioesters prepared for initial screeninga
Table 2. Racemization conditions for thioester 6ca
entry
thioester
compound
yield (%)
entry
amine (equiv)
reaction period % racemizedb (%)
1
2
3
4
5
6
7
8
phenyl
ethyl
6a
6b
6c
6d
6e
6f
92
73
86
>99
71b
94
1
2
3
4
5
6
7
8
9
trioctylamine (0.2)
trioctylamine (0.2)c
trioctylamine (2.0)
triethylamine (2.0)
triethylamine (2.0)c
trimethylamine (0.2)
trimethylamine (1.0)
trimethylamine (2.0)
trimethylamine (2.0)c
145 h
8 h
161 h
141 h
100 min
50 h
146 h
91 h
24 h
9
49
15
47
52
7
50
50
49
50
0
n-propyl
n-butyl
i-propyl
s-butyl
tert-butyl
i-butyl
6g
6h
89
91
10 trimethylamine (3.0)
83 h
97 h
a Reported as crude yields for 5-10-g scale experiments. b Prepared via the
silylated mercaptan/AlCl3 method.
11 noned
a Reaction conditions: 1.0-1.5 g 6c, 0.2 N NaH2PO4, pH ) 9.2, 40 °C, and
Triton X-100. b A value of 50% represents one-half-life or 50% ee. c 10% v/v
toluene added as well. d PS-30 lipase present as well.
While the mixture was basic (pH ≈ 9-10), we were not
seeking to establish a racemization at this point but rather
to identify a specific thioester/enzyme pair that would
produce 2 in high eep. These preliminary screening experi-
ments and several individual 1-2-g experiments produced
two useful observations. Amano PS-30 lipase, the same
enzyme that was optimal for the esters, was the only one
that displayed acceptable enantioselectivity and reaction rate,
at least for some thioesters. This was not unexpected as PS-
30 had previously resolved two esters related to 6. As regards
the choice of thioester, only those with linear alkyl chains
(ethyl, n-propyl, and n-butyl) were efficiently (1-2 days)
hydrolyzed with acceptable enantioselectivity (>85% eep).14
Remarkably, branching on the thioalkyl moiety (i.e., s-butyl,
i-propyl, etc.) led to a significantly poorer conversion and
eep. This is in contrast to the example of the i-butyl ester 5b
and indicated that the enzyme’s active site is more restrictive
of thioesters than oxoesters. Having identified three likely
thioesters and an inexpensive enzyme, the focus shifted to
the search for racemization conditions.
To quantitate the racemization rates, the chiral n-propyl
thioester 6c was prepared in the previously described manner
but from resolved acid 2. A heterogeneous mixture of this
thioester, PS-30 lipase, and Triton X-100 in pH 9.25
phosphate buffer held at 40 °C for 4 days led to no
racemization. More basic reaction conditions would have
adversely affected the lipase;15 therefore, we considered the
addition of amines and/or toluene, which can significantly
increase thioester racemization rates while retaining the
activity of the enzyme.8,16 The results of these experiments
are listed in Table 2.
1. The combination of toluene (10% v/v; 1.5 mL/g of
thioester) with any amine dramatically increased the race-
mization rate.17
2. Trioctylamine was ineffectual unless toluene was also
present.
3. The addition of g2 equiv of trimethylamine produced
an acceptable, but not optimal, racemization rate without the
necessity of toluene.
Of further significance was that in no instance was ester
hydrolysis detected. This increased confidence that nonen-
zymatic hydrolysis would not degrade the eep once dynamic
kinetic resolution conditions were established.
There was now present sufficient information from these
individual studies to consider the combination of the best
examples. A dynamic kinetic resolution should result by
matching the established choices of enzyme (Amano PS-
30), class of thioesters (C2-C4 linear chains), and racemi-
zation conditions (toluene, amine, phosphate buffer, surfac-
tant). Key experiments are listed in Table 3.
An enantioselective dynamic kinetic resolution had oc-
curred although the conversion was incomplete when toluene
was present (entries 4 and 8). Despite the increase in the
rate of racemization induced by toluene, the enzyme appeared
to be significantly inhibited or inactivated by the solvent
under these reaction conditions. Fortunately, our earlier
investigations had established that trimethylamine alone
would induce a sufficient but suboptimal racemization rate.
Modifying the reaction mixture by introducing two equiva-
lents of trimethylamine without the presence of toluene
produced a >99% conversion to (R)-2 with 97.6% eep after
22.1 h (entry 3). The “enantiomeric ratio” E of 74 determined
for this reaction under nonracemizing conditions predicted
a maximum eep of 97.3%,4,18 in satisfyingly close agreement
to that found above. The ethyl and n-butyl analogues (6b
and 6d, respectively) hydrolyzed nearly as stereoselectively
(entries 6 and 7) but 6c produced the best overall process in
regards to workup, yield, purity, and safety. We believe these
are the first examples wherein enzymatic resolution is
coupled with a retro-Michael/ Michael equilibrium, a unique
means to achieve a dynamic kinetic resolution.4
The following observations redirected our subsequent
research:
(13) On the basis of our previous success with Amano lipase PS30 for the
resolution of isoxazolines, newly prepared thioesters were typically tried
first with PS30 in 1-2 g experiments rather than within the formal screen.
Reaction for all experiments were followed by chiral LC which conveniently
indicated both the extent of hydrolysis and the eep of the 2 produced. The
majority of these screening experiments produced low conversion with poor
enantioselectivity and therefore only the better results noted for each thioester
are listed in the Experimental Section.
(14) We did not test the shortest “linear” alkyl thioester, methyl thioester, due
to the problems of stench, the boiling point of methanethiol, and the poor
resolution of the oxoester analogue.
(15) Amano Enzyme U.S.A. Co., Ltd. Company enzyme catalog and technical
information, 2001.
(16) (a) Chang, C.-S.; Tsai, S.-W.; Kuo, J. Biotechnol. Bioeng. 1999, 64, 120-
126. (b) Chang, C.-S.; Tsai, S.-W. Biochem. Eng. J. 1999, 3, 239-242. (c)
Lin, C.-N.; Tsai, S.-W. Biotechnol. Bioeng. 2000, 69, 31-38.
(17) The amine was still necessary. When the only additive was toluene and the
pH was adjusted to ∼9.5 with aqueous NaOH, no racemization was detected.
(18) Stecher, H.; Faber, K. Synthesis 1997, 1-16.
24
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

Table 3. Enzymatic dynamic kinetic resolution optimization experimentsa
entry
thioester
amine (equiv)
reaction period (h)
conversion (%)
eep (R)
1
2
3
4
5
6
7
8
9
6c
6c
6c
6c
6c
6b
6d
6c
6c
6c
trimethylamine (0.2)
trimethylamine (1.0)
trimethylamine (2.0)
trimethylamine (2.0)b
trimethylamine (3.0)
trimethylamine (2.0)
trimethylamine (2.0)
trioctylamine (0.2)b
triethylamine (1.1)
none
65
48
22
70
20
16
48
168
66
7
61
>99
>99
94
>99
>99
>99
87
96.5
95.6
97.6
96.8
97.3
96.3
97.3
97.2
96.2
96.5
>99
23
10
a Reaction conditions: 1-2 g 6c, 40 °C, phosphate buffer, Triton X-100, and Amano PS30 charged at 0.1 the weight of ester. The pH was maintained at 9.3-9.5
by 5 N NaOH added via a pH stat device. b Reaction medium also contained 10% v/v toluene.
Scheme 3. Direct conversion of 5b to thioesters19
of salt to the third wash since the low acid content led to
emulsions otherwise. These results were encouraging and
permitted further scale-up.
This chemistry was adapted to the pilot plant with the
precaution of using a bleach and sodium hydroxide-filled
vessel as a thiol scrubber, followed by the combustion of
any remaining volatiles in the thermal oxidizer. While we
had planned to use n-hexyllithium to deprotonate the thiol,
Only one serious hurdle remained. A good preparation
a sudden shortfall of the bulk supply forced our reliance upon
of the thioester was required to make commercial use of this
new dynamic kinetic resolution. The procedure used to
prepare thioesters during the screening process presented a
hazardous and lengthy path for a viable process. We have
n-butyllithium instead. As butane was now produced as the
conjugate acid of the deprotonation of propanethiol, a
concern was now the presence of the low-boiling (bp ) -0.5
°C) hydrocarbon in solution. This would complicate attaining
previously reported an efficient means for converting 5b
toluene reflux (111 °C) later and was also a safety concern.
directly to a variety of thioesters19 by the aluminum chloride-
The butane was stripped from the solution of the lithium
mediated reaction of silylated mercaptans with esters (Scheme
salt by sparging with nitrogen at 30 °C for 6 h. The
volatilized hydrocarbon was directed to the thermal oxidizer,
and its heat of combustion provided an unconventional means
to measure the endpoint of the sparge. After adding back 20
3).20 While the preparation of the thioester in this manner
formally adds an extra step, the synthesis of 5b had been
well-established in the pilot plant, and it would be attractive
to retain this intermediate in our new route. The combination
of a high-yield process to convert 5b directly to 6c and the
subsequent thioester dynamic kinetic resolution was a
reasonable alternative to the longer ester kinetic resolution/
kg (∼6% of total weight) of heptane to make up for the
solvent loss to the sparge, the solution of lithium silylsulfide
was ready to be reacted with 5a.
To allow safe handling of the reactive aluminum chloride,
unreacted ester recovery/racemization/recycle route and
our engineers designed a solids addition device which
would allow the luxury of the additional thioester preparation
attached to a reactor port and allowed charging of a
step.
preweighed quantity of aluminum chloride into an inerted
chamber. When appropriate, internal fins were actuated and
the solids dropped into the reaction mixture. This worked
very well with no significant solids adhesion within the
Our main objective during the process optimization phase
for preparing 5b was to discover a substitute for aluminum
chloride so as to eliminate the need to handle a fuming and
hazardous solid. We considered several likely Lewis acids
device. Subsequent chemistry proceeded exactly as at smaller
that are liquids (boron tribromide and boron trifluoride
scale and we prepared 45 kg of thioester 6c (98.6 wt %) in
91% yield in our initial batch.
complexed with various ethers) without success. The butyl
ether complex of boron trifluoride produced the best results
Scale-up of the enzymatic step was even more facile. Lab-
(6% solution yield after 6 days), which was obviously not
scale work (up to 183 g of 5b) defined the basic parameters
of the work, as described above, and was multiplied 10-fold
competitive with our existing conditions. Instead, our
engineers developed a safe means to handle the water-
when introduced into the kilolab (1.860 kg). The results were
reactive aluminum chloride (described below), and the
excellent (1.309 kg of 2, 99.6 wt %, g99.5% ee, 88.7%
yield). Only two points were important for further scale-up.
The time needed to attain reaction completion (34 h; >99
LC area %) was not comparable to that of laboratory
chemistry was maintained unchanged.
We explored any large-scale effects within this step by
increasing the charge of 5b 10-fold to 2 kg in our kilolab.
This experiment proceeded nearly flawlessly in 93% crystal-
lized yield (99.0 wt %). Our only change was the addition
particle size, vessel size, and other physical factors rendered
experiments (20 h). Perhaps differences in stirring efficiency,
mixing less efficient in the kilolab and resulted in a slower
conversion. Second, foaming occurred in the receiver during
(19) Pesti, J. A.; Yin, J.; Chung, J. C. Synth. Commun. 1999, 2, 3811-3820.
(20) Mukaiyama, T.; Takeda, T.; Atsumi, K. Chemistry Lett. 1974, 187-188.
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
25

the filtration of the reaction mass. Foaming could be
controlled by various means (greater receiver volume,
reduced vacuum, pressure applied to the top of the cake in
place of vacuum below the cake, etc.). The lack of adverse
observations now permitted an attempt at our final objec-
tive: further scale-up into the pilot plant.
an extra step to prepare the thioester and provides additional
possibilities for the preparation of roxifiban.
Experimental Section
General.21 Large-scale reactions were conducted in 200-
and 300-gal glass-lined reactors and were vented through a
scrubber containing 50 kg of 5-10% bleach, 65 kg of 30%
sodium hydroxide, 8 kg of 2-propanol, and 25 kg of water.
The gas stream was further directed to a thermal oxidizer.
[3-(4-Cyanophenyl)-4,5-dihydroisoxazol-5-yl]thioace-
tic Acid, S-Isobutyl Ester (6c). To a solution of THF (189
kg) and 1-propanethiol (24.5 kg, 322 mol) was charged a
solution of 2.5 M n-butyllithium in hexane (90.6 L, 226.5
mol) over 130 min at <15 °C. Chlorotrimethylsilane (37.5
kg, 345 mol) was charged over 62 min at 0-15 °C. Nitrogen
was sparged through the solution as the reaction temperature
was raised to 30 °C for 6 h to remove dissolved butane.
Heptane (20 kg) was charged to make up for the volume
loss that occurred during the stripping operation, and the
solution was cooled to 20 °C. The slurry was filtered through
a Dacron bag, and cartridge filters in series into a solution
of isobutyl 2-[3-(4-cyanophenyl)-4,5-dihydro-5-isoxazolyl]-
acetate (5b) (50.0 kg, 96.5 wt %, 168.5 mol) followed by a
mixture of THF (22 kg) and heptane (17 kg). After 10 min,
the solution was cooled to e10 °C. Aluminum chloride (30.0
kg, 225 mol) was charged through a solids charging adapter
over 35 min at e20 °C, followed by heptane (5 kg), and the
slurry was heated to 65 °C over 75 min. After an hour at 65
°C, LC indicated an 6c:5b area ratio of 97.03:0.20. After
110 min, the reaction was cooled to e20 °C. Water (213 L)
and toluene (114 kg) were charged over 1 h at <30 °C. The
phases were separated, and the organic layer was washed
with water (2 × 100 L). A final wash was mixed with sodium
chloride (14.1 kg) to aid the separation. The organic phase
was reduced in volume by vacuum distillation at 40-45 °C
and 50-75 mmHg vacuum. After 308 kg had been collected,
heptane (256 kg) was charged at 35-40 °C over 68 min to
crystallize 6c. The slurry was cooled to 0-5 °C over 3 h
and held 9 h. The crystals were filtered, and the cake was
washed with 0-5 °C heptane (2 × 51 kg). The crystals were
dried at 50 mmHg and 45 °C for 40 h to yield 44.9 kg (98.6
wt %, 91% yield). This material was identical to that
previously reported.19
The thioester (44.9 kg) was converted into 2, again with
careful attention to capturing the volatile thiol produced
during the hydrolysis. As regards design, the plant arrange-
ment differed from the kilolab preparations only in that pH
was not automatically controlled. A mixture of sodium
phosphate, trimethylamine, surfactant, 6c, and Amano PS-
30 lipase were reacted until HPLC analysis indicated
sufficient conversion (96%). This was conducted in a manner
similar to that of the kinetic resolution process to produce
28.4 kg of 2 (>99.9 wt %) in 80.4% overall yield. While
the eep was only 94.4% at the end of the reaction, the acid
precipitation of crude 2 raised it to 98.6%, and subsequent
recrystallization purified it to >99.9%. The multikilogram
preparation ran as planned, and no changes were required.
This high yield and well-functioning process leading to a
pure product fulfilled our goal of transforming the kinetic
resolution of 5b into a dynamic kinetic resolution and
successfully demonstrating this chemistry in the pilot plant.
While the thioester process to 2 is certainly more
interesting chemistry when compared to the kinetic resolu-
tion, it can be difficult to decide between these two
established routes. It is not unreasonable to argue that the
need to prepare 6c and handle malodorous reagents out-
weighs the benefits of the subsequent dynamic kinetic
resolution. The preference expressed by potential contractors
between the kinetic and dynamic resolutions should be noted.
Their judgment was that the higher conversion and overall
yield of the new process outweighed the disadvantages.
Obviously, the selection between the two will be driven by
the experience and equipment available to the chemist, and
either route may be superior under different circumstances.
In any event, the dynamic kinetic resolution does supply a
new option for the preparation of the key roxifiban inter-
mediate 2.
Conclusions
The existing kinetic resolution of the isobutyl ester 5b to
form the acid 2 has been transformed into a dynamic kinetic
resolution via the intermediacy of the thioester 6c. Optimized
reaction conditions for this dynamic kinetic resolution were
determined by a series of screening trials examining the
interactions of enyzmes and thioesters, racemization condi-
tions of the thioester 6c, and finally by dynamic kinetic
resolution experiments derived from the previous two
screens. The preparation of 6c was improved by synthesizing
it directly from 5b with silylated propanethiol and aluminum
chloride and by developing means to conduct this chemistry
on large scale safely. The subsequent dynamic kinetic
resolution scaled up to the pilot plant facilely and proceeded
at large scale in a well-behaved manner. This new chemistry
avoids the inconvenient recycling of the unreacted enanti-
omer required by the kinetic resolution by instead introducing
(R)-2-[3-(4-Cyanophenyl)-4,5-dihydroisoxazol-5-yl]ace-
tic Acid (2). A solution of water (653 L), sodium phosphate
monobasic monohydrate (67.5 kg), and aqueous 24% tri-
methylamine solution (73.5 kg, 0.297 kmol) was treated with
sodium hydroxide solution (33%, 24.4 kg, 0.201 kmol) at
<30 °C to raise the pH to 9.09. Triton X-100 (3.7 kg), 6c
(44.9 kg, 0.156 kmol) and lipase Amano PS-30 (4.5 kg) were
charged sequentially, and the resulting mixture was heated
to 40 °C over 1 h. The mixture was stirred at 40-43 °C,
and the pH of the mixture was adjusted every 2-6 h to 9.2-
9.5 by adding 33% NaOH solution (∼3 kg each time; total
of 21 kg charged). After 41 h, HPLC indicated an area ratio
of 2:6c of 96.1:3.7. The reaction mixture was cooled to
(21) HPLC methods are described in ref 4 and the corresponding Supporting
Information, and in ref 19.
26
•
Vol. 8, No. 1, 2004 / Organic Process Research & Development

10 °C, Celite 545 (24.1 kg) was charged and the mixture
was filtered through more filter aid slurried in water. A total
of 180 L of water was used to wash through the remaining
product solution. The filtrate was cooled to 10 °C, isopropyl
acetate (20 kg) was added, and the mixture was acidified to
pH 3.09 by the addition of 85% phosphoric acid (55.1 kg)
over 1.75 h at <10 °C. The mixture was cooled to 5 °C,
and the solids were collected by filtration. The cake was
washed with 5 °C water (3 × 45 L) and then slurried with
absolute ethanol (188 kg) at 0-5 °C. The wet solids were
then recrystallized from ethanol (523 kg) by heating to reflux
over ∼1 h to dissolve all the solids. The mixture was then
cooled to 5 °C over 4.5 h and held at 0-5 °C for 2 h. The
resulting slurry was filtered, washed with ethanol (76 kg),
and dried at 50 °C at 50 mmHg for 24 h to yield 28.4 kg
(>99.9 wt %, >99.9 ee %, 80.4% yield) of 2. This material
was identical to that previously reported.2,5
Enzyme Screening Experiments.13 The typical experi-
ment consisted of a mixture of 20-25 mg of thioester, 10-
20 mg of lipase, 3-4 drops of Triton X-100, and 2-3 mL
of phosphate buffer, pH 8, maintained at 40-45 °C for 18-
24 h while either shaken or stirred by a Teflon stir bar in a
capped vial. The following Amano enzymes were screened
against the thioesters listed in Table 1: lipase AP12
(Aspergillus niger), lipase AY30 (Candida rugosa), lipase
G (Penicillilum camembertii), lipase GC20 (Geotricum
candidum), lipase L10 (C. lipolytica), lipase AK (Pseudomo-
nas fluorescens), lipase FAP (Rhizopus oryzae), lipase N (R.
niVeus), lipase PS30 and PS (Pseudomonas cepacia), lipase
AY (C. rugosa), lipase D (R. delemar), lipase MAP10
(Mucor sp.), lipase CES (P. burkholderia), and lipase R10
(P. roqueforti). These lipases are classified as triacylglycerol
acylhydrolases, (EC 3.1.1.3). Screened as well were the
Amano proteases S (Bacillus stearothermophilus)) and N (B.
subtilis), both classified as EC 3.4.24.28. The following
Sigma enzymes were screened: acylase I (Aspergillus spp.,
EC 3.5.1.14), lipase II PPL (porcine pancreas, EC 3.1.1.3),
protease XXIII (A. oryzae), and PLE (porcine liver esterase,
C 3.1.1.1). The better results for each thioester follow: 6a:
no specificity >40% eep noted for any enzyme. 6b: Amano
PS30 (∼99% eep after workup at ∼92% conversion). 6c:
Amano PS30 (∼97% eep after workup at ∼85% conversion),
Amano CES (74% eep at 79% conversion). 6d: Amano PS30
(87% eep at ∼60% conversion; Amano AK (95% eep at
∼47% conversion). 6e: Amano PS30 (31% eep at ∼12%
conversion). 6f: Amano PS30 (68% eep at ∼3% conversion).
6g: Amano R10 (59% eep at ∼15% conversion). 6h: Amano
AY (60% eep at ∼1.5% conversion).
Acknowledgment
We thank the following colleagues for their contributions
in the development of this chemistry: Robert Wethman,
Dennis Potter, Bill Boucher, Walt Conway, Joseph Renai,
Dawn Blackburn, Elton Watson, Fred Maccherone, and Jeff
Koskey. We are grateful for insightful suggestions in the
preparation of the manuscript from Robert Discordia, David
Kronenthal, Edward Delaney, Robert DiCosimo, and Am-
barish Singh.
Received for review May 16, 2003.
OP0300239
Vol. 8, No. 1, 2004 / Organic Process Research & Development
•
27