The synthesis of a 5HT2C receptor agonist

Article abstract of DOI:10.1021/op700121y

This report describes the large-scale synthesis of 1 that features a Fischer indole strategy to form an advanced intermediate followed by reduction to the indoline to construct the tetracyclic core of the molecule. Resolution using dibenzoyl-D-tartaric acid affords access to a single enantiomer, from which a Suzuki coupling builds in the biaryl functionality. Deprotection followed by salt formation furnishes the desired target molecule.

Full text of DOI:10.1021/op700121y

Organic Process Research & Development 2007, 11, 985–995
The Synthesis of a 5HT2C Receptor Agonist
Lindsay A. Hobson,* William A. Nugent,§ Stephen R. Anderson,‡ Subodh S. Deshmukh,† James J. Haley III, Pingli Liu,¶
Nicholas A. Magnus,# Patrick Sheeran, James P. Sherbine, Benjamin R. P. Stone, and Jiang Zhu|
Chemical Process Research and DeVelopment Department, Bristol Myers Squibb Company, One Squibb DriVe,
New Brunswick, New Jersey 08903, U.S.A.
Abstract:
C, and D). The Drug Discovery chemistry team had used this
approach for their preparation of analogues in this series, thereby
having the ability to vary the aryl group. With this approach
the aryl portion could be introduced by a Suzuki coupling
reaction of the required boronic acid and a halogenated
tetracycle. Scheme 1 outlines the retrosynthetic analysis of 1,
which shows the initial disconnection of the aryl linker to give
the tetracyclic core 2 and the required boronic acid 3. The
tetracyclic indoline 2 can in turn be obtained from reduction of
the corresponding indole 4, which can be obtained from a
Fischer indolization of 4-piperidone monohydrate hydrochloride
with hydrazine 5. Hydrazine 5 can be obtained from bicyclic
amine 6 via formation of the N-nitroso compound followed by
reduction.
The known bicyclic amine 63 can be obtained in two ways,
as shown in Scheme 2, either by a Beckmann rearrangement
of the chromanone oxime 7 or by alkylation/ring opening/
cyclization of 2-benzoxazolinone 8.
The key difference between the Drug Discovery chemistry
route and the process route was the method used to introduce
the chirality. The Discovery Chemistry route utilized a chiral
chromatographic separation of the racemic BOC-protected
indoline, 19, prior to the Suzuki coupling reaction. During the
development of the process route the possibility of a resolution
process was investigated, which would provide a more cost-
efficient process.
This report describes the large-scale synthesis of 1 that features a
Fischer indole strategy to form an advanced intermediate followed
by reduction to the indoline to construct the tetracyclic core of
the molecule. Resolution using dibenzoyl-D-tartaric acid affords
access to a single enantiomer, from which a Suzuki coupling builds
in the biaryl functionality. Deprotection followed by salt formation
furnishes the desired target molecule.
Introduction
The 5HT2C receptor is believed to be one of two receptors,1
along with 5HT1B, responsible for the major satiety enhancing
effect of 5HT on the central nervous system. Target molecule
1 was selected as a promising drug candidate for the treatment
of obesity.2 To this end, the production of kilogram quantities
of 1 was required to support drug development (i.e., toxicology,
formulation development, and pharmacologic assessment). To
meet these requirements the identification and execution of a
scalable synthetic sequence was essential and is outlined below.
Within the Chemical Process R & D group both routes to 6
were investigated. The route that had been utilized by the
Discovery chemists employed the alkylation of 2-benzoxazoli-
none with 1-bromo-3-chloropropane using sodium hydride as
base in a mixture of THF and DMPU.4 This resulted in a 68%
yield of the alkylated benzoxazolinone 9, Scheme 3. The major
impurity in the reaction was shown to be the alkylated dimer
10. A small amount of the bromoalkylated product, 11, forms,
which should also be converted to the desired product 6.
Treatment of 9 with potassium hydroxide in 2-methoxyethanol
Results and Discussion
The synthetic approach to 1 utilizes as the key step a Fisher
indole cyclization to assemble the tetracyclic core (rings A, B,
* To whom correspondence should be addressed. E-mail: lindsay.hobson@
bms.com. The research described was carried out at the former Dupont
Pharmaceuticals Co.
§ Current address: Vertex Pharmaceuticals, Inc., Cambridge, MA, U.S.A.
‡ Current address: Chemical and Pharmaceutical Development, Wyeth
Research, Pearl River, NY 10965, U.S.A.
† Current address: Chemical Research and Development, Pfizer Inc., Groton,
CT 06340, U.S.A.
(3) (a) Fevig, J. M.; Mitchell, I. S.; Lee, T.; Chen, W.; Cacciola, J.
Substituted pyrroloquinolines and pyridoquinolines as serotonin ago-
nists and antagonists. PCT Int. Appl. WO2002059124, 2002 (Bristol-
Myers Squibb Company, USA). (b) Sebok, P.; Levai, A.; Timar, T.
Heterocycl. Commun. 1998, 4, 547. (c) Meshcheryakova, L. M.;
Zagorevskii, V. A.; Orlova, E. K. Khim. Geterotsiklicheskikh Soedin.
1980, 6, 853. (d) Orlova, E. K.; Sharkova, N. M.; Meshcheryakova,
L. M.; Zagorevskiim, V. A.; Kucherova, N. F. Khim. Geterotsikli-
cheskikh Soedin. 1975, 9, 1262. (e) Sam, J.; Valentine, J. L.; Richmond,
C.W. J. Pharm. Sci. 1968, 57, 1763. (f) Strupczewski, J. T.; Bordeau,
K. J.; Chiang, Y.; Glamkowski, E. J.; Conway, P. G.; Corbett, R.;
Hartman, H. B.; Szewczak, M. R.; Wilmot, C. A.; Helsley, G. C.
J. Med. Chem. 1995, 38, 1119.
¶ Current address: Incyte, Wilmington, DE, U.S.A.
# Current address: Eli Lilly and Company, Chemical Product Research and
Development Division, Indianapolis, IN 46285, U.S.A.
Current address: Federal Bureau of Investigation, 2500 East T.C. Jester,
Houston, TX 77008, U.S.A.
| Current address: STS Chemical Services, Chemical Development, Roche
Palo Alto LLC, Palo Alto, CA, U.S.A.
(1) (a) Bickerdike, M. J.; Vickers, S. P.; Dourish, C. T. Diabetes, Obes.
Metab. 1999, 1, 207. (b) Halford, J. C. G.; Blundell, J. E. Ann. Med.
2000, 32, 222. (c) Kennett, G. A. Idrugs 1998, 1, 456.
(2) Robichaud, A. J.; Lee, T.; Deng, W.; Mitchell, I. S.; Haydar, S.; Chen,
W.; McClung, C. D.; Calvello, E. J. B.; Zawrotny, D. M. Preparation
of substituted heterocycle fused gamma-carbolines. PCT Int. Appl.
WO2000077010, 2000 (Du Pont Pharmaceuticals Company, USA).
(4) Smolinski, S.; Szneler, E. Cf. reference to alkylate 2-aminophenol with
1,3-dibromopropane. Zesz. Nauk. Uniw. Jagiellonskiego, Prace Chem.
1980, 25, 19.
10.1021/op700121y CCC: $37.00
Published on Web 10/18/2007
2007 American Chemical Society
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
985

Scheme 1. Retrosynthetic analysis
in 2-methoxyethanol over a period of 48 h in 55% yield.5 In
order to optimize the yield of this reaction, an investigation using
both potassium hydroxide and potassium tert-butoxide in a
number of solvents was undertaken. This revealed that the
combination of THF and potassium tert-butoxide gave better
reaction impurity profiles compared to potassium hydroxide.
Upon addition of potassium tert-butoxide to a solution of 9 in
THF at 0 to 10 °C, HPLC analysis indicated the formation of
a new compound, which was shown to be compound 12, where
potassium tert-butoxide had opened up the oxazolinone ring to
provide a BOC-protected amine. Upon heating to 55 to 60 °C,
12 was readily converted to cyclic ether 13. Treatment of 13
with either methanesulfonic acid or 5 M hydrochloric acid
removed the BOC group and afforded either the methane-
sulfonic acid salt or the hydrochloride salt of 6 in 70–80% yield,
Scheme 4.
It is possible that the two steps from 2-benzoxazolinone to
compound 6 could be conducted sequentially without isolation
of intermediate 9; however this was not explored. An alternative
approach to 6 was examined in parallel to this work. This
involved the treatment of 4-chromanone with hydroxylamine
to form the known oxime 14.3b,6 Treatment of oxime 14 with
a hydride source such as di-isobutylaluminium hydride would
effect the Beckmann rearrangement to give the required 1,5-
benzodiazepine 6, Scheme 5.3d,7,8
Scheme 2. Retrosynthetic analysis of the 1,5-benzodiazepine
at reflux effected the ring opening of the oxazolinone followed
by displacement of the chloride and decarboxylation to generate
6 in 70% yield.
Due to the low yields obtained, the hazards associated with
the use of sodium hydride on scale, and the toxicity of
2-methoxyethanol, a study was performed to investigate alter-
nate bases, solvents, and temperatures in conjunction with a
reduction of reagent stoichiometry. The objective was to
optimize an alternate base/solvent combination that would
minimize the formation of the dimer impurity and improve the
yield, while resulting in a more scalable synthesis. Table 1
summarizes the different reaction conditions investigated to
optimize the reaction along with the results.
The course of the reaction is significantly impacted by
solvent, reagent choice, and stoichiometry. The choice of base
had a dramatic effect on the formation of dimer. Changing the
base from sodium hydride to potassium tert-butoxide dropped
the amount of dimer from 12.8 area % to <0.03 area % (HPLC)
and also increased the yield to 92%. Investigating additional
potassium bases showed that potassium hydroxide led to
increased formation of the dimer, whereas potassium carbo-
nate was comparable to potassium tert-butoxide in minimizing
the dimer formation. The choice of solvent and temperature
were also factors that affected the amount of dimer. Although
there were many acceptable alternatives, the conditions chosen
for scale-up were potassium carbonate in dimethylacetamide
(DMAC) at 15 to 25 °C, with 2 equiv of the alkylating reagent.
The subsequent step involved opening of the oxazolinone
ring in the presence of base with closure on the alkyl chain to
give the 1,5-benzodiazepine 6. The literature indicates that this
transformation can be carried out using potassium hydroxide
4-Chromanone was treated with a slight excess of both
hydroxylamine hydrochloride and triethylamine in methanol at
70 °C for 3 h. The process was investigated with and without
triethylamine. It was found that the yields for the two processes
were similar (92–95%); however the reaction time was slower
when triethylamine was not present, 5 h versus 2 h. Upon
completion of the reaction, water was added and the oxime
directly crystallized from the reaction solution. The product was
dried at 60 °C/25 mmHg in order to produce material with a
(5) Sam, J.; Valentine, J. L.; Aboul-Enein, M. N. J. Pharm. Sci. 1971,
60, 1370.
(6) Pratap, R.; Gupta, R. C.; Anand, N. Indian J. Chem. Sect. B 1981, 20,
1063.
(7) Nagarajan, K.; Kulkarni, C. L.; Venkateswarlu, A. Indian J. Chem.
1974, 12, 247.
(8) (a) Evans, D.; Lockhart, I. M. J. Chem. Soc. 1965, 4806. (b) Huckle,
D.; Lockhart, I. M.; Wright, M. J. Chem. Soc. 1965, 1137. (c)
Grunewald, G. L.; Dahanukar, V. H.; Ching, P.; Criscione, K. R.
J. Med. Chem. 1996, 39, 3539.
986
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

Scheme 3. Alkylation of 2-benzoxazolinone
Table 1. Effects of solvent/base/temperature/stoichiometry
solvent
temp, °C
time, h
base (equiv)
Br(CH2)3Cl equiv
area % 9
area % 10
area % 11
yield %
THF/ DMPU
THF/ DMPU
THF/ NMP
NMP
65
7
4
1.5
5
19
5
NaH (1.15)
tBuOK (1.05)
tBuOK (1.05)
K2CO3 (2)
1.1
3
65
12.8
ND
ND
ND
3.7
22
51
92
77
85
94
75
75
89
86
92
80
92
ND
ND
5
62
3
95.6
95
92.8
87.3
92
98.5
96.2
95
20
2
NMP
25
K2CO3 (1.5)
KOH (1.0)
1.5
3.1
4
3.5
2.5
ND
ND
1
IPA
IPA
NMP
DMAC
DMAC
82
6.7
82
3.5
1.15
3
KOH (1.0)
7.7
0–10
0–10
15–25
KOtBu (1.15)
KOtBu (1.15)
K2CO3 (1.5)
2
1.46
2.8
2
6
2
3.1
2.2
Scheme 4. Formation of 1,5-benzodiazepine 6 from
benzoxazolinone 9
Oxime 14 was then treated with 3 equiv of di-isobutylalu-
minium hydride in toluene at low temperature (-5 to -15 °C).
Following the addition, the reaction was allowed to warm to
15 °C, and after a 1 h hold time the reaction was found to be
complete by HPLC analysis. The reaction was quenched by
the addition of a solution of Rochelle’s salt. The quench was
addition controlled to limit the release rate of the hydrogen and
isobutane gases generated.9 Following the workup of the
reaction mixture, the product was treated with HCl in IPA to
form the hydrochloride salt, which was isolated in 90% yield
with a purity of 98.5%.
The preferred method to prepare 6b was determined to be
from 4-chromanone due to the higher yield over the two steps.
With 1,5-benzodiazepine 6b in hand, the next step required
N-nitrosation3d,10 to give the N-nitroso compound 15, which,
following reduction, would produce hydrazine 5 (Scheme 6)
in readiness for the Fischer indolization step.
Treatment of a solution of 6b in hydrochloric acid with a
solution of sodium nitrite in water generated the N-nitroso
compound 15, which was taken directly into the reduction
without isolation. Reduction of 15 to the hydrazine 5 was
achieved by heating with sodium dithionite in a mixture of
ethanol and sodium hydroxide for 2 h.11 A small amount of
the N–N cleavage product, 6, was observed, but this was
minimized by shortened reaction times. Hydrazine 5 was not
isolated but taken directly into the Fischer indole reaction.
water content of less than 0.5%. This level of water was shown
to be acceptable for use in the next reaction using di-
isobutylaluminum hydride. The isolated yield of oxime 14 was
high, 94% (25 kg scale), with a purity of 98.7 wt %. However,
it was found that the crystalline product showed instability under
typical storage conditions (closed polyethylene bag, inside a
fiber drum at room temperature). The product reverted back to
4-chromanone, as identified by HPLC. This occurred after
storage for ∼6 h, with the HPLC purity of oxime 14 changing
from 100 area % to 83 area %. The cause of the autoreversion
has not been clearly established, but laboratory stability studies
showed that the presence of residual acid induced reversion.
Additional studies showed that the compound was stable at 22
°C overnight, 100 °C under vacuum, and as a slurry in water
(process water from the plant) at 100 °C for 3 h. Prior to the
next step, the 4-chromanone was removed from the product
by slurrying the material in methanol/water followed by filtration
to isolate the product, which was then stored under vacuum in
a tray drier to ensure that no further degradation took place.
Further studies demonstrated that reagents such as nitric acid,
hydrogen peroxide, and bleach also converted the product back
to starting material.
(9) Perrault, W. R.; Shephard, K. P.; LaPean, L. A.; Krook, M. A.;
Dobrowolski, P. J.; Lyster, M. A.; McMillan, M. W.; Knoechel, D. J.;
Evenson, G. N.; Watt, W.; Pearlman, B. A. Org. Process Res. DeV.
1997, 1, 106.
(10) (a) Toscano, L.; Grisanti, G.; Fioriello, G.; Seghetti, E. For similar
N-Nitrosations see. J. Med. Chem. 1976, 208. (b) Inoue, N. Bull. Chem.
Soc. Jpn. 1968, 41, 2078.
(11) (a) For nitroso reductions see: Conant, J. B.; Corson, B. B. Organic
Syntheses; . Wiley: New York, Collect. Vol. II, 1943; p 33. (b) Ishii,
H.; Murakami, Y.; Ishikawa, T. Chem. Pharm. Bull. 1990, 38, 597.
(c) Overberger, C. G.; Herin, L. P. J. Org. Chem. 1962, 27, 417.
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
987

Scheme 5. Formation of 1,5-benzodiazepine 6 from 4-chromanone
Scheme 6. Formation of hydrazine 5 from
1,5-benzodiazepine 6
Scheme 8. Reduction of indole 4
Scheme 7. Formation of indole 4 from hydrazine 5
applicable to this substrate was an asymmetric hydrogenation
of an indole using Rh(acac)(cod) with a chiral ligand to produce
monosubstituted indolines in high ee’s.15 The catalytic hydro-
genation of compound 4 proved challenging, since the indole
double bond is both tetrasubstituted and stabilized by the
heteroaromaticity of the indole ring. Conditions that were
sufficiently vigorous to overcome these factors also promoted
competitive hydrogenation of the phenyl ring to afford 17,
Scheme 8. To minimize this problem, we focused on reductions
performed under acidic conditions. Protonation of the indole
to form an indolinium ion was expected to suppress the
aromaticity of the indole nucleus and thus activate the double
bond toward hydrogenation.
Adams catalyst (platinum oxide) gave higher conversion and
selectivity for 16 as compared with 5% ruthenium on carbon
or 5% platinum on carbon under a variety of conditions. Formic
acid has been advocated as a solvent for the hydrogenation of
indoles.16 However, the use of formic or acetic acid alone or in
combination with methanesulfonic acid for the hydrogenation
of 4 resulted in low conversions. Higher conversions (50–100%)
were achieved using 1:1 ethanol/aqueous HBF4 as solvent,17
but selectivity was limited and in one case resulted in complete
conversion to 17 (as a mixture of diastereomers). The transfer
hydrogenation of indoles to indolines is known.18 However,
attempted transfer hydrogenations of 4 under the reported
conditions afforded only recovered starting material.
Treatment of 5 with 4-piperidone monohydrate hydrochloride
in IPA furnished indole 4, Scheme 7.7
Due to the product’s low solubility in IPA, isolation was
achieved via a precipitative crystallization, which also entrained
ammonium chloride in the product. To remove the ammonium
chloride, a slurry of the material in water was performed and
produced a fine particulate solid that was comprised of
agglomerated needles with a particle size distribution in the 5–30
µm range and thus filtered very poorly. To improve the physical
properties of the isolated material and the performance of the
filtration, a controlled crystallization using 2-propanol/water
(1:9) was developed, which resulted in material of larger particle
size (50–100 µm plates) and thus faster filtration. The indole
was isolated following the improved recrystallization
conditions in an overall yield of 73% with a typical purity
of >99.7 wt %.
Reduction to the indoline from indole 4 was now required.
The literature describes many methods for the reduction of
indoles to indolines, most of which include the use of reagents
such as borohydrides/cyanoborohydrides with a strong acid,12
borane,13 or catalytic hydrogenations.14 One example potentially
Our best results for catalytic hydrogenation were obtained
using platinum oxide catalyst in 90:10 dichloromethane/
methanesulfonic acid (10% catalyst, 30 psi H2, 25 °C, 24 h).
These conditions afforded 100% conversion of 4, but selectivity
was only 7/3 in favor of 16. In parallel, achiral reductions were
also being investigated. Due to delivery requirements and the
lack of success with chiral reductions, we turned our attention
(12) (a) Robinson, B. Chem. ReV. 1969, 69, 785. (b) Gribble, G. W.;
Nutaitis, C. F. Org. Prep. Proc. Int. 1985, 17, 317, and references
therein. (c) Srikrishna, A.; Reddy, T. J.; Viswajanani, R. Tetrahedron
1996, 52, 1631. (d) Kumar, Y.; Florvall, L. Synth. Commun. 1983,
13, 489. (e) Berger, J.; Davidson, F.; Langford, G. E. J. Med. Chem.
1977, 20, 600. (f) Kiguwa, Y. Chem. Pharm. Bull. 1978, 26, 108. (g)
Wakamatsu, T.; Inaki, H.; Ogawa, A.; Watanabe, M.; Ban, Y.
Heterocycles 1980, 14, 1441. (h) Gribble, G. W.; Nutaitis, C. F.; Leese,
R. M. Heterocycles 1984, 22, 379. (i) Hutchins, R. O.; Natale, N. R.
Org. Prep. Proc. Int. 1979, 11, 201.
(13) (a) Maryanoff, B. E.; McComsey, D. F.; Nortey, S. D. J. Org. Chem.
1981, 46, 355, and references therein. (b) Biswas, K. M.; Dhara, R.;
Roy, S.; Mallik, H. Tetrahedron 1984, 40, 4351. (c) Monti, S. A.;
Schmidt, R. R., III., Tetrahedron 1971, 27, 3331. (d) Elliot, A. J.;
Guzik, H. Tetrahedron Lett. 1982, 19, 1983.
(15) Kuwano, R.; Sato, K.; Kurokawa, T.; Karube, D.; Ito, Y. J. Am. Chem.
Soc. 2000, 122, 7614. For asymmetric hydrogenation of a BOC-
protected 2,3-disubstituted indole, see: Kuwano, R.; Kashiwabara, M.
Org. Lett. 2006, 8, 265.
(16) Kikugawa, Y.; Kashimura, M. Synthesis 1982, 785.
(17) Magee, M. P.; Norton, J. R. The use of tetrafluoroboric acid was based
on a report regarding imine hydrogenation. J. Am. Chem. Soc. 2001,
123, 1778.
(14) (a) Jokela, R.; Lounasmaa, M. Tetrahedron 1989, 45, 303. (b) Collot,
V.; Schmitt, M.; Marwah, P.; Bourguignon, J.-J. Heterocycles 1999,
51, 2823. (c) Coulton, S.; Gilchrist, T. L.; Graham, K. Tetrahedron
1997, 53, 791.
(18) Khai, B. T.; Arcelli, A. Tetrahedron Lett. 1985, 26, 3365.
988
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

Scheme 9. Formation of indoline 16
cooled to -5 to -10 °C and treated with a solution of bromine
in dichloromethane. The reaction was complete 30 min after
the addition, at which point n-heptane was added to crystallize
the product as the hydrobromide salt. On a >9.5 kg scale
product 20 was obtained in >88% yield over two steps with a
purity of >97 wt %.
With aryl bromide 20 in hand, the Suzuki coupling reaction22
was investigated. The literature describes a number of potential
conditions that can be used for Suzuki couplings. The initial
conditions evaluated were those implemented in the Discovery
Chemistry route to the compound which utilized barium
hydroxide as the base and tetrakis(triphenylphosphine) pal-
ladium as the catalyst in aqueous dimethoxyethane. The required
boronic acid could be synthesized from 4-bromo-3-trifluorom-
ethylphenol in two steps as shown in Scheme 12: protection of
the phenol group with 2-bromopropane followed by formation
of the boronic acid with n-butyllithium and triisopropylborate
followed by acidification. The boronic acid used in the scale-
up work was purchased from a vendor.
The reaction of 20 with boronic acid 22 was complete in
3 h when using a 2.5 mol % catalyst loading as shown in
Scheme 13. Workup of the reaction to isolate compound 23
proved capricious. Compound 23 could not be isolated as a
solid and also contained high levels of palladium (2500-3500
ppm). It was found that treatment of a solution of 23 with
Picachem carbon23 directly after the Suzuki coupling did not
sufficiently remove the residual palladium to an acceptable level.
Washing the organic phase with a solution of tris(hydroxym-
ethyl)aminomethane24 was likewise inefficient in removing the
residual palladium. Ultimately, the residual palladium was
removed by treatment of the organic phase with a 20% sodium
carbonate solution of trithiocyanuric acid.25
Upon treatment of the organic phase with trithiocyanuric
acid/sodium carbonate and cooling to <5 °C a solid palladium-
containing precipitate formed, which was easily removed via
filtration. The aqueous phase was then discarded and the organic
phase was treated with Picachem carbon 80PN to further reduce
palladium levels to less than 100 ppm. The organic phase was
then solvent exchanged to isopropyl acetate. The isopropyl
acetate solution of 23 was heated to 87 to 89 °C, and
methanesulfonic acid was added over a period of 30 min to
effect the BOC group cleavage. After holding for an additional
1 h at 87 to 89 °C, the reaction was complete and the solution
was cooled to <30 °C. Following aqueous workup, the organic
phase was treated with diphenylacetic acid to generate the
diphenylacetic acid salt of 1. On a 1 kg scale the yield was
65–70% with 30 ppm of palladium in the diphenylacetic acid
salt and with a purity of 99.4 wt %.
to the stoichiometric reduction of 4 using triethylsilane in
trifluoroacetic acid, Scheme 9.19
Treatment of indole 4 with 3.5 equiv of triethylsilane in 5
volumes of trifluoroacetic acid at 45 °C for 24 h provided
indoline 16. Upon completion of the reduction (95% conver-
sion), water was added and the organic phase that contained
side products was discarded. The aqueous layer was then
washed with n-heptane to further remove the siloxane byprod-
ucts before the acidic phase was basified. The aqueous phase
was cooled to 5 °C, treated with sodium hydroxide until the
pH had been adjusted to 12 or higher, and then extracted with
isopropyl acetate followed by a solvent exchange to methanol
to provide a crude solution of indoline 16.
During a crystallization screen of indoline 16 with chiral
acids it was found that dibenzoyl-D-tartaric acid gave the desired
enantiomer in high ee in 20% yield. During the course of the
development work it was found that a higher yield could be
obtained if indoline 16 was initially treated with dibenzoyl-L-
tartaric acid. This removed the undesired enantiomer from the
mixture, thus enriching the mother liquors in the desired
enantiomer to a ratio of 67:33, Scheme 10. After treating the
mother liquors with sodium hydroxide and dissolving the free
base in dichloromethane, a solvent exchange to methanol was
carried out. The methanol solution containing the enriched
desired enantiomer of the indoline was then treated with
dibenzoyl-D-tartaric acid to afford the product, 18. The resolu-
tion was demonstrated on pilot plant scale to produce 25 kg of
18 in 25–30% yield (99.1 wt %, >97% enantiomeric excess).20
With the desired enantiomer of the indoline in hand the next
step required protection of the piperidine nitrogen followed by
bromination of the aromatic ring to provide the precursor for
the Suzuki coupling reaction. To this end a solution of indoline
18 in dichloromethane was treated with an excess of aqueous
sodium hydroxide, followed by di-tert-butyldicarbonate to
directly afford the BOC-protected indoline 19, Scheme 11.
During the development of the bromination reaction it was
found that molecular bromine was preferred over N-bromo-
succinimide, affording 20 as its hydrobromide salt, which could
be crystallized directly from the reaction mixture. In contrast,
N-bromosuccinimide gave a difficult-to-crystallize gum, even
after an aqueous workup. However, in using molecular bromine
for this reaction, the temperature needed to be maintained below
0 °C to minimize competitive BOC cleavage to compound 21.21
The organic phase from the BOC-protection containing 19 was
The Suzuki reaction proceeded well on scale; however the
filtration of the large amount of solid palladium-containing
precipitate from the trithiocyanuric acid treatment proved slow
(22) Miyaura, N.; Suzuki, A. Chem. ReV. 1995, 95, 2457. Suzuki, A. J.
Organmet. Chem. 1999, 147, and references therein.
(19) Ketcha, D. M.; Lieurance, B. A. Tetrahedron Lett. 1989, 30, 6833.
(20) The absolute stereochemistry of the free base of compound 18 had
previously been determined by single-crystal X-ray in the Discovery
Chemistry group, and 18 was compared to authentic samples using
HPLC retention times.
(21) Compound 21 was not isolated and could be carried through to the
Suzuki reaction without impacting the step.
(23) Picachem carbon was identified from adsorbent screening studies and
was found to be the best in terms of loading and removal. For further
information on Picachem carbon see www.picacarbon.com/en/.
(24) Private communication with J. Sherbine.
(25) Rosso, V. W.; Lust, D. A.; Bernot, P. J.; Grosso, J. A.; Modi, S. P.;
Rusowicz, A.; Sedergran, T. C.; Simpson, J. H.; Srivastava, S. K.;
Anderson, N. G. Org. Process Res. DeV. 1997, 1, 311.
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
989

Scheme 10. Resolution of indoline 16
Scheme 11. Formation of bromoindoline 20
Scheme 12. Formation of boronic acid 22
were slightly higher (<50 ppm vs <30 ppm) than those from
the two-treatment protocol, but the simplified process was more
amenable to scale-up. The palladium level could be reduced
further during the final step, which involved the conversion to
the selected final pharmaceutical salt form. This improved
procedure was run on a 10 kg scale to give the diphenylacetic
acid salt 1a in 55% yield, 98.1 area % with <50 ppm of
palladium with 1 to 2% of the product being left in the liquors
during the crystallization.
and troublesome due to the large amount of solid generated
and the need to filter through Celite. To avoid this type of
filtration in the future, the reaction was further optimized to
reduce the amount of palladium used. During the development
work on the Suzuki reaction it was found that a solvent mixture
of isopropyl acetate and water worked as well as aqueous DME.
Implementing this change eliminated a solvent exchange.
Alternate bases were also investigated, and potassium carbonate
and cesium carbonate were found to be comparable to barium
hydroxide in terms of reaction rate and impurity profile. The
catalyst loading could be lowered to 1 mol % without impacting
the reaction conversion, but the reaction rate was slower, as
would be expected. Modifying the Suzuki reaction conditions
to potassium carbonate as the base and isopropyl acetate as the
solvent and 2 mol % of catalyst allowed a streamlined procedure
for the deprotection reaction followed by salt formation.
Employment of alumina or triphenylphosphine to remove
residual palladium showed no significant improvements from
the trithiocyanuric acid/carbon treatments. One promising
alternative was to treat both the organic phase following the
Suzuki reaction and also the organic phase after the deprotection
with carbon. Further investigation showed that one treatment
of the organic phase after the deprotection reaction with
Picachem carbon 80PN (10% wt/wt) at 50 °C for 2 h provided
an efficient removal. Following formation of the diphenylacetic
acid salt residual levels of palladium were <50 ppm. The levels
The final step in the preparation was a salt exchange from
the diphenylacetic acid salt, 1a, to a pharmaceutically acceptable
salt form for delivery of the drug. The citrate salt was chosen
as the initial form for development of the API. Aqueous sodium
hydroxide was used to extract the free base into isopropyl
acetate. The solvent was then exchanged to an ethanol/n-heptane
(70:30) mixture, and citric acid was added to form the citrate
salt, 1b, Scheme 14.
The citrate salt was isolated in 92% yield, 100.7 wt % on a
900 g scale, and was shown to have 100% chiral purity by SFC
and <14 ppm of residual palladium.
Conclusion
An efficient route has been developed and demonstrated on
multikilogram scale for the synthesis of 1, a potential drug
candidate for the treatment of obesity. A chemical resolution
was utilized on scale to provide the desired enantiomer of the
indoline, which could be efficiently progressed via a Suzuki
coupling to provide the API.
Experimental Section
NMR spectra were obtained at 25 °C in CDCl3 except as
indicated, and field strengths for the various nuclei were as
follows: 1H (400.1 MHz), 13C (100.6 MHz), 19F (376.5 MHz).
Coupling constants (J) are given in hertz.
990
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

Scheme 13. Formation of the API
Scheme 14. Formation of the desired API salt form
(E)-Chroman-4-one oxime (14). 4-Chromanone (25 kg, 169
mol) was charged to a 100 gallon glass-lined reactor followed
by hydroxylamine hydrochloride (12.9 kg, 185 mol) and
methanol (37.6 kg). Agitation was set to 100 rpm and the jacket
temperature to 20 °C. Triethylamine (19.1 kg, 189 mol) was
charged to the reactor over 15 min, adjusting the addition rate
in order to maintain the batch temperature at <30 °C. The
agitation was increased to 150 rpm and the batch set to heat to
80 °C in jacket mode over a period of 30 min. The batch began
to reflux at ∼70 °C and was held at reflux with a jacket
temperature of 80 °C for 2 h or until HPLC indicated >99.5%
conversion. The batch was then cooled to 50 °C, and water
(250 kg) was added above surface over a 60 min period. The
batch was then cooled to 15 °C over 90 min. The batch was
held at 15 °C for 30 min with the agitator set at 60 rpm. A
sample of the slurry was taken and the mother liquor submitted
for analysis to ensure <0.5 wt % of compound 14 remained.
The batch was then filtered on a Nutsche filter using a 1 µm
Dacron cloth. The cake was washed with 98 kg of purified water
in three portions. The wet cake was then dried in a tray drier at
60 °C under full vacuum for 12 h or until the KF level was
<0.3 wt % to provide 25.7 kg of compound 14 in 93% yield
(100 area %, 100.3 wt %, <0.1% MeOH). Metal results were
Fe 1.6 ppm, Cd 1 ppm, Ba 0.8 ppm, all other metals e0.5 ppm.
2,3,4,5-Tetrahydrobenzo[b][1,4]oxazepine hydrochloride
(6b). Compound 14 (25.6 kg, 157 mol) was charged to a 50
gallon glass-lined reactor followed by anhydrous tetrahydrofuran
(53.8 kg), which was charged through the dip leg using nitrogen
pressure. The agitation was set to 90 rpm and the batch
temperature to 25 °C. A sample of the solution was taken for
KF analysis to ensure that the KF was <500 ppm. To a 200
gallon glass-lined reactor venting to a flame arrestor was added
DIBAl-H (50 wt % solution in toluene, 185.8 kg, 653 mol)
using ∼10 psig nitrogen pressure on the DIBAl-H cylinder. On
completion of the addition the cylinder block valve was closed
and the DIBAl-H cylinder was slowly vented to 5 psig. Toluene
(2 kg) was charged to a stainless steel pressure cylinder. The
nitrogen feed port on the DIBAl-H cylinder was connected to
the toluene cylinder and the toluene cylinder pressurized to
15–20 psig with nitrogen. The feed line was flushed by
transferring the toluene from the pressure cylinder into the
reactor followed by blowing the line with nitrogen. The reactor
containing the DIBAl-H was cooled to ensure the contents were
<2 °C, the reactor was vented to the flame arrestor, and the
nitrogen purge was 0.5 scfm. The solution of compound 14 in
THF was transferred using nitrogen pressure to the reactor
containing the DIBAl-H such that the addition was above
surface and that the reaction temperature was <10 °C over 2 h
40 min. On completion of the addition, the batch was warmed
to 25 °C over 50 min. THF (4 kg) was used to rinse the 50
gallon reactor and then transferred using nitrogen pressure to
the reaction vessel. The nitrogen purge was reduced to 0.1 scfm,
and the batch was aged at 25 °C for 2 h or until HPLC indicated
>99.5% conversion. A 2-propanol cylinder was connected to
a nitrogen line and pressurized to ∼5 psig nitrogen pressure.
2-Propanol (10.5 kg) was then charged to the batch using
nitrogen pressure over a period of 30 min. The addition rate
was controlled using a needle valve in order to maintain the
temperature of the batch at <30 °C and control the evolution
of hydrogen. To a 300 gallon glass-lined reactor was added
water (203 L) followed by potassium sodium tartrate (Rochelle’s
salt) (200 kg, 709 mol). Sodium hydroxide (50 wt %, 39.5 kg)
was then added to the solution of Rochelle’s salt using deadhead
vacuum, and the charging line was rinsed with water (2 L).
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
991

The batch temperature of the quench solution was checked to
be <30 °C, and the reactor was aligned to vent to the thermal
oxidizer. At this point the reaction solution was transferred to
the quench tank using nitrogen pressure and a needle valve to
control the addition rate and to maintain the reactor temperature
between 25 and 30 °C. The addition was carried out above
surface and took 2 h 50 min. During the addition ∼22 000 L
of isobutane gas was expected to be generated. After the
addition was complete, toluene (5 kg) was charged to the 100
gallon reaction tank through the dip leg using deadhead vacuum
and then transferred to the quench tank using a slight nitrogen
pressure. Dry nitrogen purges were then carried out through
the dip leg and headspace of the quench tank at 1 scfm each.
The temperature was set to 30 °C in batch mode, and the agi-
tation was increased to 125 rpm. During the heat-up the
isobutane gas dissolved in the toluene was released. The
nitrogen sparge through the dip leg was continued for 75 min
until the isobutane gas evolution had stopped. At this point the
nitrogen purge through the dip leg was stopped and the
headspace nitrogen purge was reduced to 0.1 scfm. The re-
maining isobutane dissolved in the toluene phase was then
vacuum stripped from the reaction mass by applying ∼600
mmHg vacuum and gradually increasing the vacuum to 300
mmHg in 25 mmHg steps, each time allowing for the oxidizer
gas flow rate to stabilize. The batch was held at 300 mmHg
for 30 min until the isobutane evolution had ceased. The reactor
vents were aligned with the process scrubber, the agitation was
stopped, and the phases were allowed to separate. The aqueous
phase was discharged into drums. The organic phase was
discharged into drums, and a toluene rinse (10 kg) was charged
to the reactor and combined with the organic phase. The
combined organic phase was sampled for KF (criterion <1.0
wt %). The organic layer was recharged to the 300 gallon glass-
lined reactor through a 0.5 µm cartridge filter using deadhead
vacuum of ∼300 mmHg. The agitation was set to 75 rpm, and
HCl in 2-propanol (3.5 kg, 5.5 N) was charged through the dip
leg using a Teflon-coated wilden pump. The HCl was charged
through the dip leg to avoid losing HCl to the vapor space
during the addition. The batch was then heated to 70 °C in batch
mode. When the batch reached 70 °C, a further charge of HCl
in 2-propanol (32 kg) was added through the dip leg over a
period of 60 min. On completion of the addition, the batch was
held at 70 °C for 30 min, after which time the batch was cooled
to 20 °C over 3 h and the agitation was set to 50 rpm. When
the batch temperature reached <25 °C, a sample was taken to
ensure that the mother liquor contained <1.0 wt % of compound
6b. The slurry was filtered on a Nutsche filter containing a 1
µm Dacron filter cloth. Toluene (50 kg) was charged to the
reactor using deadhead vacuum and agitated at 100 rpm for 5
min. The toluene was then discharged to wash the cake in two
portions. The wet cake was dried in a tray drier at 50 °C and
200 mmHg vacuum until a constant weight was obtained. The
product was isolated as a white solid (26.4 kg) in 91% yield
with 100 area % and 79.1 wt % (based on freebase).
°C to form a homogeneous solution. To a 100 gallon glass-
lined reactor was charged compound 6b (24 kg, 129 mol)
followed by water (108 L). The solution was cooled to 10 °C,
and hydrochloric acid (37 wt %, 14.8 kg) was then charged to
the reactor while maintaining the temperature at <20 °C. The
solution of sodium nitrite was charged to the 100 gallon reactor
containing compound 6b using nitrogen pressure and controlling
the addition such that the temperature was maintained between
10 and 20 °C. Water (2 kg) was used to rinse the 30 gallon
reactor and transferred to the reaction vessel. The batch was
warmed to 22 °C and held at 22–25 °C for 60 min or until
HPLC showed >99.0% conversion to compound 15. The batch
was cooled to 0–10 °C, and ethanol (80.3 kg) was charged using
deadhead vacuum. Sodium hydroxide (50 wt %, 76.1 kg) was
then charged over a 30 min period using deadhead vacuum (500
mmHg to ensure minimal loss of HCl) while keeping the
temperature <20 °C. The pH was checked to be >7, and the
batch was cooled to 0–8 °C with 40 rpm agitation. Sodium
hydrosulfite (61.9 kg, 355 mol) was charged to the reactor
through the manway followed by inertion of the reactor. The
agitation rate was increased to 100 rpm, and the batch was
heated to reflux for 2 h or until the reaction completion was
>99.0 % conversion by HPLC to compound 5. The reaction
was cooled to 20 °C, and the resultant slurry was filtered on a
Nutsche filter. The product-rich filtrate was recirculated through
the reactor until solids were no longer present in the reactor.
The cake was washed with ethanol (37.9 kg), and the wash
was discharged to the reactor containing the filtrate. The
combined filtrate and wash were then distilled under vacuum
(50–150 mmHg) while keeping the batch temperature <40 °C
until the liquid volume remaining in the reactor was ∼200 L.
A second cake wash with ethanol (37.9 kg) was then transferred
and the distillation continued until ∼200 L remained and the
amount of ethanol remaining was <5 vol %. The batch was
cooled to 20–25 °C, and MTBE (59.7 kg) was charged using
deadhead vacuum (∼600 mmHg). The biphasic mixture was
agitated for 20 min and the phases were separated (30 min).
The lower aqueous phase was extracted a second time with
MTBE (48 kg). The combined organic phase was then
recharged to a 100 gallon reactor and distilled under vacuum
(∼50 mmHg) at 35–40 °C until ∼60 L remained in the reactor.
2-Propanol (95.7 kg) was charged using deadhead vacuum, and
the distillation continued until ∼150 L remained in the reactor,
when a further 95.7 kg of 2-propanol was charged. A sample
was taken to ensure that MTBE was <5 vol % and water was
<5 wt %. Additional 2-propanol (25.2 kg) was added in order
to obtain a 12 wt % IPA solution, and the batch was then cooled
to 5–8 °C. 4-Piperidone monohydrate hydrochloride (21.2 kg,
138 mol) was charged to the reactor, and the batch was heated
to 40 °C. When the batch temperature was >35 °C, a solution
of HCl in 2-propanol (5.5 N, 27.8 kg) was added through the
dip leg using deadhead vacuum. The batch was then heated at
reflux for 2 h or until HPLC conversion indicated >98.5%.
The batch was then cooled to <20 °C and filtered on a Nutsche
filter using a 1 µm Dacron filter cloth. 2-Propanol (67.9 kg)
was added to the reactor and discharged to wash the cake in
three portions. The wet cake was deliquored using a nitrogen
purge and by applying vacuum (∼100 mmHg) for 13 h. Water
6,7,9,10,11,12-Hexahydro-5H-[1,4]oxazepino[2,3,4-hi]py-
rido[4,3-b]indole hydrochloride (4). To a 30 gallon glass-lined
reactor was charged sodium nitrite (9.8 kg, 142 mol) followed
by water (21 L). The mixture was agitated at 80 rpm and 20
992
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

(175 L) was charged to a 100 gallon glass-lined reactor followed
by the wet cake (97.5 kg). The slurry was heated to 105 °C
over a 90 min period. A clear solution was visible at 83 °C,
and the batch was held at 90 °C for 20 min. The solution was
cooled in batch mode from 90 to 80 °C over 30 min, 80 to 60
°C over 40 min, and then 60 to 5 °C over 60 min. Following
a 60 min hold at 5 °C, the slurry was filtered on a Nutsche
filter containing a 1 µm Dacron filter cloth and washed with
precooled water (3–5 °C) (60 L) in three portions. The wet cake
was dried at 70 °C and 50 mmHg vacuum until constant weight
to provide compound 4 (24.6 kg) in 72% yield (100 wt %)
over the three steps.
4: 1H NMR (400 MHz, DMSO-d6) δ 2.29 (m, 2H), 3.05 (t,
J ) 6.1 Hz, 2H), 3.47 (t, J ) 6.4 Hz, 2H), 4.03 (t, J ) 5.6 Hz,
2H), 4.26 (m, 4H), 6.69 (d, J ) 7.9 Hz, 1H), 6.93 (t, J ) 7.8
Hz, 1H), 7.09 (d, J ) 7.8 Hz, 1H) and 9.71 (s, 1H); 13C NMR
(100 MHz, DMSO-d6) δ 19.44, 30.39, 40.65, 46.29, 71.14,
103.33, 110.30, 110.68, 119.90, 126.58, 128.33, 132.46, 144.75.
HRMS: calcd for C14H16N2O 229.1341 [M + 1]; found
229.1340 [M + 1].
(8aS,12aR)-6,7,9,10,12,12a-Hexahydro-5H-[1,4]oxazepi-
no[2,3,4-hi]pyrido[4,3-b]indole hemidibenzoyl-D-tartrate (18).
Compound 4 (15 kg, 56.7 mol) was charged to a 100 gallon
glass-lined reactor. Trifluoroacetic acid (111 kg, 974 mol) was
added to the reactor using deadhead vacuum followed by
triethylsilane (23.1 kg, 199 mol). The agitation was set to 150
rpm to ensure adequate mixing. The resultant solution was
heated to 45 °C and held for at least 24 h until the reaction
reached >94% conversion. The reaction was cooled to <5 °C,
and water (56 L) was added. The batch temperature was
adjusted to 20 °C, and the phases were agitated for 15 min and
then allowed to settle. The product-rich aqueous phase was
separated, then recharged to the reactor. n-Heptane (38 kg) was
added to the reactor, and the biphasic mixture was agitated for
15 min and then allowed to settle. The aqueous phase was
separated and washed a second time with n-heptane (38 kg) to
remove the siloxane byproducts. The aqueous phase was
separated, recharged to the reactor, and cooled to 5 °C. A
solution of NaOH was prepared in a separate 30 gallon glass-
lined reactor using water (80 L) and adding 50 wt % NaOH
(86 kg) while keeping the temperature <30 °C. The prepared
solution of NaOH was then charged to the product-rich aqueous
phase such that the temperature did not exceed 30 °C (∼45
min). Water (53 L) was used to rinse the NaOH vessel and
transferred to the product-containing vessel. The batch temper-
ature was adjusted to 20 °C, and a sample was taken for pH
(pH > 12). Isopropyl acetate (65 kg) was added to the aqueous
phase and the mixture agitated for at least 15 min. The phases
were separated, and the aqueous phase was extracted with IPAc
(2 × 65 kg). The combined organic phase was recharged to
the 100 gallon glass-lined reactor and washed with a 20% brine
solution (45.5 kg). The organic phase was then distilled under
vacuum (100 mmHg), gradually increasing the jacket temper-
ature to 50–55 °C. The distillation was continued until 30–40
L remained in the reactor. Methanol (108 kg) was added to the
reactor, and the distillation was continued at 200 mmHg and a
jacket temperature of 50–60 °C. The distillation was continued
until ∼40 L remained in the reactor. A second charge of
methanol (107 kg) was added and the distillation continued until
the level was ∼60 L. Methanol (100 kg) was charged to the
reactor, a sample was taken for residual isopropyl acetate
(criterion e5%), and the batch was cooled to 20 °C. The
solution was transferred through a 0.5 µm cartridge filter to
remove any salts into a 100 gallon glass-lined reactor, and
methanol (39.9 kg) was added to make a total solution weight
of 178.3 kg. The batch was then heated to 50–55 °C, and water
(52 L) was added while maintaining the batch temperature at
50–55 °C. To a separate 30 gallon glass-lined reactor was
charged dibenzoyl-L-tartaric acid (3.95 kg, 11 mol) followed
by methanol (16.5 kg) and water (5 L). The mixture was agitated
until a solution formed. The dibenzoyl-L-tartaric acid solution
was then charged to the solution of indoline. The solution was
seeded with 65 g of the dibenzoyl-L-tartaric acid salt of the
indoline and the slurry held at 50 ( 3 °C for 1 h. The batch
was then cooled to 35 °C over 2 h. When the batch reached 40
°C, a sample of the slurry was filtered and the cake was
submitted for analysis (criterion: g98% undesired enantiomer).
The batch was further cooled to 25 °C over 1 h. The batch was
filtered on a Nutsche filter and the cake washed with methanol
(17 kg). The combined mother liquor and cake wash were
recharged to a 100 gallon glass-lined reactor and distilled under
vacuum (150 mmHg), heating to 50–55 °C until 30–40 L
remained in the vessel. Methylene chloride (120 kg) was
charged to the reactor followed by a solution of sodium
hydroxide (4.7 kg of 50% NaOH in 61 L of water), maintaining
the temperature at <30 °C. The pH of the reaction mixture was
checked to be >10, and after agitating the biphasic mixture for
15 min the phases were allowed to separate. The organic layer
was separated and washed with a 10% brine solution (66 kg).
After agitating for 15 min the phases were allowed to separate.
The organic phase was then distilled at atmospheric pressure
until there was 30–40 L remaining, at which point methanol
(124 kg) was charged and the distillation continued at 300
mmHg and a jacket temperature of 50 °C. The distillation was
continued until ∼60 L remained. Methanol (92 kg) was then
charged to the reactor, and a sample was taken for residual
DCM (criterion e5%). The batch was then heated to 50 ( 3
°C. Water (42 L) was charged to the reactor while maintaining
the batch temperature at 50 ( 3 °C. In a separate 30 gallon
glass-lined reactor a solution of dibenzoyl-D-tartaric acid (3.95
kg, 11 mol) in methanol (13 kg) and water (4 L) was prepared
and then discharged to a drum. The solution of dibenzoyl-L-
tartaric acid (14 kg) was added to the reactor containing the
product-rich solution and then seeded with compound 18 (64.5
g, 0.15 mol). The remaining dibenzoyl-L-tartaric acid solution
was then charged to the slurry, and the slurry was held at 50 (
3 °C for 1 h. The batch was then cooled to 25 °C over 3 h. On
reaching 25 °C a sample was taken, the slurry was filtered, and
the cake and mother liquor were submitted for analysis
(criterion: cake g98% desired enantiomer; filtrate 30–35%
desired enantiomer present). The slurry was filtered on a
Nutsche filter and washed with methanol (17 kg). The wet cake
was dried in a tray drier at 70 °C under vacuum to provide the
product as a white solid (5.8 kg) in 27% yield. The solid had
a wt % of 99.3%, 100 area %, and the enantiomeric ratio was
98.9:1.1 desired/undesired. 18: DSC 196.89 °C; 1H NMR (400
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
993

MHz, DMSO-d6) δ 1.85–2.03 (m, 4H), 2.18 (t, J ) 12 Hz,
1H), 2.29 (m, 1H), 2.75 (td, J ) 12 and 4 Hz, 1H), 2.98–3.08
(m, 3H), 3.25–3.31 (m, 2H), 3.61 (t, J ) 12 Hz, 1H), 4.30–4.34
(m, 1H), 5.60 (s, 1H), 6.60–6.73 (m, 3H), 7.47 (t, J ) 8 Hz,
2H), 7.60 (t, J ) 8 Hz, 1H) and 7.97 (2H, d, J ) 8 Hz); 13C
NMR (100 MHz, DMSO-d6) δ 21.57, 30.72, 37.74, 38.82,
44.88, 49.97, 62.75, 71.94, 75.39, 118.12, 119.91, 120.67,
128.46, 128.86, 129.49, 129.55, 130.94, 133.29, 133.75, 142.49,
146.42, 165.53, and 170.19; HRMS calcd for C14H19N2O
231.1497 [M + 1]; found 231.1488 [M + 1].
(m, 2H), 3.77 (m, 1H), 4.37 (dt, J ) 12 and 4 Hz, 1H), 6.85
(1H, s) and. 6.92 (1H, s); 13C NMR (100 MHz, CDCl3) δ 25.76,
30.69, 41.55, 41.67, 49.26, 49.67, 64.76, 72.08, 111.00, 120.48,
122.14, 137.10, 142.04 and 147.02; HRMS calcd for
C14H18N2OBr 309.0602 [M + 1]; found 309.0602 [M + 1].
(8aS,12aR)-2-[4-Isopropoxy-2-(trifluoromethyl)phenyl]-
6,7,8a,9,10,11,12,12a-octahydro-5H-[1,4]oxazepino[2,3,4-
hi]pyrido[4,3-b]indole diphenylacetate (1a). To a 50 gallon
glass-lined reactor was charged water (17 L) followed by
potassium carbonate (7.2 kg, 52 mol), compound 20 (8.5 kg,
17.35 mol), and 4-isopropoxy-2-trifluoromethylphenylboronic
acid, 22 (5.8 kg, 23.4 mol). Isopropyl acetate (59.3 kg) was
charged, and agitation was set to 180 rpm. The nitrogen purge
was set to sparge through the dip tube. The reactor was degassed
by pressurization to 1000 mmHg, venting to 760 mmHg and
applying vacuum (300 mmHg). This was repeated three times.
Tetrakis(triphenylphosphine)palladium (0) (0.44 kg, 0.38 mol)
was slurried in isopropyl acetate (2 kg) and transferred to the
reactor using deadhead vacuum. The contents of the batch were
heated to reflux (∼85 °C) and held for 3 h or until HPLC
analysis indicated >99.5% conversion. The reaction mass was
allowed to cool to 20 °C. A Nutsche filter was prepared with
a 3–5 µm Dacron filter bag and Celite 545 (15 kg), and water
was added to wet the Celite. The contents of the reactor were
then filtered through the Celite, and the filtrate was collected.
Isopropyl acetate (73.9 kg) was added to the reactor and then
discharged to the Nutsche filter to wash the cake. The filtrate
was charged back into a 50 gallon glass-lined reactor, and the
phases were allowed to settle for 15 min. The bottom aqueous
layer was discharged. The organic layer was then distilled under
vacuum (100 mmHg), keeping the batch temperature <50 °C
until ∼51 L remained in the vessel and the water content was
<1% by KF. The solution was then heated to reflux (∼85 °C).
Methanesulfonic acid (3.3 kg, 34.3 mol) was charged to a
Teflon-lined charging cylinder and then added to the reactor
using nitrogen pressure. The addition rate was controlled by
an in-line needle valve and took ∼30 min. The reaction began
to off-gas CO2 and isobutene ca. two-thirds of the way through
the addition (max. 380 L of each). The reaction mixture was
kept at ∼85 °C for a further 1 h or until HPLC analysis indicated
no starting material remaining. The reaction was cooled to 20
°C, and a solution of potassium carbonate (7.2 kg, 52 mol) in
water (43 L) was added over a 30 min period; the addition rate
was controlled to maintain the pressure in the reactor below
900 mmHg. The reaction mixture was stirred for 15 min, and
the pH of the aqueous phase was checked to be >7. The reactor
contents were filtered through water wet Celite 545 (8 kg) and
washed with isopropyl acetate (73.9 kg). The filtrate was
recharged to the vessel, and the phases were allowed to settle
over a 15 min period. The bottom aqueous layer was then
discharged. About 100 mmHg was then applied to the reactor,
and the organic phase was distilled, keeping the batch temper-
ature <50 °C until ∼85 L remained in the reactor. Picachem
carbon 80PN (1 kg, 10 wt %) was charged to a 100 gallon
glass-lined reactor. The organic phase was transferred to the
100 gallon reactor followed by a rinse with isopropyl acetate
(5 kg). The slurry was heated to 50 °C, held for 2 h, and then
cooled to 20–30 °C. The slurry was filtered through Celite 545
(8aS,12aR)-tert-Butyl-2-bromo-6,7,9,10,12,12a-hexahydro-
5H-[1,4]oxazepino[2,3,4-hi]pyrido[4,3-b]indole-11(8aH)-car-
boxylate (20). To a 100 gallon glass-lined reactor was charged
compound 18 (9 kg, 21.97 mol) followed by methylene chloride
(77.7 kg) and water (61 L). Sodium hydroxide (5.3 kg, 132.5
mol, 50 wt %) was charged to the reactor via pump while
maintaining the temperature below 30 °C. The pH of the
aqueous phase was checked to be pH > 7. Di-tert-butyldicar-
bonate (5.2 kg, 23.82 mol) was added followed by a rinse of
the charging line with methylene chloride (1.5 kg). The biphasic
mixture was stirred for 60 min at 25 °C. HPLC analysis of the
reaction mixture showed the absence of starting material. The
two phases were separated, and the organic phase was recharged
to a second vessel (50 gallon glass-lined reactor). A sample of
the organic phase was taken for KF (criterion <0.5 wt %). The
organic phase was cooled to -5 °C in batch mode. A bromine
solution (3.68 kg, 23.02 mol) in methylene chloride (1.7 kg)
was charged to a Teflon-lined stainless steel charging cylinder.
The bromine solution was transferred to the reactor containing
the organic phase using ∼5 psig of nitrogen pressure, subsurface
over a period of ∼30 min, keeping the temperature at -5 ( 3
°C. Following the charge the charging line was rinsed with
methylene chloride (1 kg) and transferred to the reactor. On
completion of the addition, HPLC analysis indicated the absence
of starting material. n-Heptane (58.4 kg) was charged to the
reactor using a pump over a 45 min period, keeping the batch
temperature between -5 and 0 °C. During the addition the
product began to crystallize. The reaction mass was held
between -5 and 0 °C for 15 min. The product was collected
by filtration on a 30 in. jacketed Nutsche filter using a 1 µm
Dacron filter cloth. n-Heptane (17.5 kg) was charged to the
reactor and cooled to 0 °C and used to wash the cake. The
cake was dried in a tray drier at 50 °C/50 mmHg for 18.5 h to
give the product, 20, as a light brown solid, 9.8 kg (88%
corrected yield, 97 wt %, 95.2 area %), with 4.8 area % of
compound 21 (4.1 wt %).
1
20: DSC 117.85 °C; H NMR (400 MHz, CDCl3) δ 1.44
(s, 9H), 1.88–1.90 (m, 2H), 2.02–2.03 (m, 2H), 2.52–2.62 (m,
1H), 3.21–3.40 (m, 4H), 3.42–3.82 (m, 4H), 4.35–4.45 (m, 1H),
6.89 (d, J ) 1.8 Hz, 1H), and 6.92 (d, J ) 1.8 Hz, 1H); 13C
NMR (100 MHz, CDCl3) δ 24.49, 28.41, 30.84, 39.13, 41.22,
44.77, 49.97, 64.36, 72.15, 79.62, 110.99, 121.02, 122.56,
135.60, 142.18, 146.78, and 154.76; HRMS calcd for
C19H26N2O3Br 409.1127 [M + 1]; found 409.1146 [M + 1].
1
21: DSC 104.21 °C; H NMR (400 MHz, CDCl3) δ
1.76–1.82 (m, 1H), 1.90–1.96 (dq, J ) 16 and 4 Hz, 1H), 2.06
(m, 2H), 2.45–2.55 (m, 2H), 2.85 (m, 2H), 2.99 (dd, J ) 12
and 8 Hz, 1H), 3.08–3.14 (dt, J ) 12 and 4 Hz, 1H), 3.31–3.38
994
•
Vol. 11, No. 6, 2007 / Organic Process Research & Development

(10 kg) and washed with isopropyl acetate (73.9 kg). A sample
was removed to determine Pd content for information only. The
filtrate was charged back into the 50 gallon glass-lined reactor
through a 0.2 µm cartridge filter using deadhead vacuum. About
100 mmHg vacuum was applied to the reactor, and the solution
was distilled, keeping the batch temperature <50 °C until ∼53
L remained in the reactor. A sample was taken for wt % water
and concentration of the compound (criterion: <1 wt % and
11–12 wt %, respectively). If the 11–12 wt % is not met,
isopropyl acetate can be added or further distillation performed.
n-Heptane (20.5 kg) was charged to the reactor, and the batch
was heated to 60–65 °C. Diphenylacetic acid (3.7 kg, 17.4 mol)
was charged to a 30 gallon glass-lined reactor followed by
isopropyl acetate (13.1 kg). The batch was heated to 55–60 °C
to obtain a homogeneous solution. The diphenylacetic acid
solution (∼4 L) was charged to the rich organic phase at 60
°C. The reaction mixture was seeded with a seed slurry (150 g
in 0.7 kg of n-heptane). Once the batch was seeded, the
remainder of the diphenylacetic acid solution was transferred
to the slurry over a period of 30 min. The 30 gallon reactor
was rinsed with isopropyl acetate (2 kg) and transferred to the
slurry. The reaction mass was allowed to cool to 0 °C over a
period of 2 h and held at 0 to 5 °C for 1 h. A sample was taken
for mother liquor concentration and was found to be 2.1 wt %.
The product was filtered on a 1 µm Dacron filter bag and
washed with a cooled mixture (0 to 5 °C) of n-heptane/IPAc
(14.4 kg:7.8 kg). The product was dried at 50 °C/50 mmHg
for 21 h to give a white solid, 6.1 kg (55%, 98.1 area%).
1a: DSC 162.80 °C; 19F NMR (376 MHz, MeOH-d4) δ
-58.83; 1H NMR (400 MHz, MeOH-d4) δ 1.36 (d, J ) 8 Hz,
6H) 2.00–2.15 (m, 3H), 2.30 (m, 1H), 2.47–2.58 (m, 2H), 3.12
(td, J ) 12 and 4 Hz, 1H), 3.20–3.26 (m, 2H), 3.31–3.51 (m,
3H), 3.73–3.79 (m, 1H), 4.40–4.50 (m, 1H), 4.69 (septet, J )
8 Hz, 1H), 4.94 (s, 1H), 6.74 (s, 1H), 6.77 (s, 1H) and 7.10–7.33
(m, 13H); 13C NMR (100 MHz, MeOH-d4) 19.25, 19.74, 29.08,
36.31, 37.48, 43.38, 48.45, 59.64, 61.18, 68.62, 70.37, 111.75,
111.81, 116.44, 117.18, 119.54, 121.19, 123.91, 124.35, 126.20,
127.12, 127.46, 130.61, 131.19, 131.82, 139.92, 140.30, 144.39,
155.39, 177.42; HRMS calcd for C24H28N2O2F3 433.2103 [M
+ 1]; found 433.2106 [M + 1].
sodium hydroxide (66.6 g, 1.67 mol) in water (3.6 L) was
charged to the slurry, and the biphasic mixture was stirred at
20 °C for 10 min. The two phases were separated, and the
organic phase was recharged to the vessel. To this was added
a second charge of sodium hydroxide (66.6 g, 1.67 mol) in water
(3.6 L). The biphasic mixture was stirred for 10 min, and the
two phases were separated. The organic phase was washed with
water (3.6 L) and then evaporated to dryness on the rotary
evaporator. The residue was dissolved in ethanol (2.7 L) and
water (25 mL). Citric acid (266 g, 1.39 mol) was added, and
the mixture was heated to 50 °C. n-Heptane (2.7 L) was
charged, maintaining the temperature at ∼50 °C. The solution
was cooled to 35 °C and seeded with ∼0.5 g of seeds. Cooling
was continued to 25 °C. n-Heptane (3.6 L) was added over a
period of 1 h. The slurry was filtered and washed with
n-heptane/EtOH (70:30) (1 L). The product was dried on the
filter under a nitrogen blanket for 24 h. The product was
obtained as a white solid, 798.6 g (92%, 100 wt %, 100% chiral
purity, residual Pd 14 ppm, 99.55 area % at 265 nm).
1
1b: 19F NMR (376 MHz, MeOH-d4) δ -58.76; H NMR
(400 MHz, MeOH-d4) δ 1.36 (d, J ) 8 Hz, 6H), 2.00–2.40
(m, 4H), 2.50–2.64 (m, 2H), 2.80 (dd J ) 12 and 16 Hz, 4H),
3.20 (m, 1H), 3.36–3.45 (m, 3H), 3.50–3.60 (m, 2H), 3.73–3.79
(m, 1H), 4.40–4.50 (m, 1H), 4.69 (septet, J ) 8 Hz, 1H), 6.74
(s, 1H), 6.82 (s, 1H), 7.13 (dd, J ) 4 and 8 Hz, 1H), and
7.09–7.23 (m, 2H); 13C NMR (MeOH-d4, 100 MHz) δ 22.59,
22.95, 32.43, 39.54, 40.90, 45.27, 46.68, 51.82, 64.54, 71.91,
73.73, 74.89, 115.06, 115.12, 119.73, 120.59, 122.93, 124.52,
127.24, 130.78, 133.98, 134.46, 135.19, 143.31, 147.73, 158.71,
175.54, and 180.04.
Acknowledgment
We thank our Drug Discovery colleagues previously at
Dupont Pharmaceuticals for useful discussions. We are also
thankful to Dr. Pat N. Confalone, Dr. Rodney Parsons, and Dr.
Robert Waltermire for helpful discussions. Special thanks to
Barbara Lord, John Castoro, Heidi Heck, Yun K. Ye, and
Thomas Neiss for analytical support during the development
of this work.
(8aS,12aR)-2-[4-Isopropoxy-2-(trifluoromethyl)phenyl]-
6,7,8a,9,10,11,12,12a-octahydro-5H-[1,4]oxazepino[2,3,4-
hi]pyrido[4,3-b]indole citrate (1b). The diphenylacetate salt
(895.0 g, 1.4 mol) was slurried in IPAc (3.6 L). A solution of
Received for review June 1, 2007.
OP700121Y
Vol. 11, No. 6, 2007 / Organic Process Research & Development
•
995