Multi-kiloscale enantioselective synthesis of a vitronectin receptor antagonist

Article abstract of DOI:10.1021/op0499021

The development of a novel, cost-effective synthesis of the vitronectin receptor antagonist SB-273005 became necessary as the compound proceeded to Phase 1. A practical synthesis of the compound presented challenges to the process chemist. Chief among the challenges was developing an enantioselective route to the compound. Second was either developing a scalable Mitsunobu coupling of the side chain to the main body or finding alternate chemistry. In this paper we will describe the chemistry we developed which allowed us to make over a hundred kilograms of SB-273005 by a process that we believe is suitable for even larger scale manufacturing.

Full text of DOI:10.1021/op0499021

Organic Process Research & Development 2004, 8, 738−743
Multi-Kiloscale Enantioselective Synthesis of a Vitronectin Receptor
Antagonist
Michael D. Wallace,* Michael A. McGuire, Marvin S. Yu,† Lynn Goldfinger, Li Liu, Wenning Dai, and Susan Shilcrat
GlaxoSmithKline, Synthetic Chemistry Department, 709 Swedeland Road, P.O. Box 1539,
King of Prussia, PennsylVania 19406
Abstract:
The development of a novel, cost-effective synthesis of the
vitronectin receptor antagonist SB-273005 became necessary
as the compound proceeded to Phase 1. A practical synthesis
of the compound presented challenges to the process chemist.
Chief among the challenges was developing an enantioselective
route to the compound. Second was either developing a scalable
Mitsunobu coupling of the side chain to the main body or
finding alternate chemistry. In this paper we will describe the
chemistry we developed which allowed us to make over a
hundred kilograms of SB-273005 by a process that we believe
is suitable for even larger scale manufacturing.
Figure 1. Active benzazepine compounds.
The integrin family of transmembrane glycoproteins that
acts as cell adhesion receptors and signal transducers include
the victronectin receptor, Rvâ3.1 The vitronectin receptor,
Rvâ3, is known to assist a wide variety of biological
processes, and as a consequence of this broad activity, it
was anticipated that suitably designed antagonists would be
useful in the treatment of inflammation,2 cardiovascular
disorders,3 cancer,4 and osteoporosis.5 Two potent second
generation Rvâ3 receptor antagonists intended for the
treatment of osteoporosis have been reported6 (Figure 1).
Compound 1 (SB-273005), a 2,4,9-trisubstituted-2-benz-
azepine-3-one, was discovered by our colleagues in the
Medicinal Chemistry group at SmithKline Beecham, and
their original synthesis was disclosed with the synthesis of
compound 2.6 Simultaneous with the Medicinal Chemistry
work, new routes were being investigated by members of
the Process Chemistry team, and as SB-273005 approached
the clinic, a more practical route was deemed a necessity.
The topic of this report is to describe the final synthetic route
to SB-273005 which is notable for its novelty, brevity, ease
to scale, and cost-effectiveness.
Synthetic methods for 2-benzazepin-3-ones are not ex-
tensive, but there are examples and probably the most
common method for construction is the lactamization of
appropriate ortho substituted benzylamines.7 Indeed, that was
the selected method described above by the Medicinal
Chemistry group and was blueprinted as the foundation of
our plan as well. More recently, that same lactam bond
connection strategy was employed by Van Rompaey in the
preparation of 2,4-disubstituted 2-benzazepin-3-ones8 (eq 1,
Scheme 1). A different method to generate the C-N bond
was employed by Busacca, whereby a dehydrative cyclization
between an amide and ketone yielded an acylimine which
was successfully hydrogenated to generate the desired
benzazepinone structure9 (eq 2, Scheme 1). In a method that
did not construct the benzazepinone ring system through
C-N bond formation, researchers at Merck modified the
Pictet-Spengler reaction (usually reserved for isoquinoline
synthesis) and cyclized onto an acylimine10 (eq 3, Scheme
1). Other researchers as well have used subtle variations on
Pictet-Spengler chemistry to afford the desired benzazepi-
none skeleton.11,12 The Schmidt, and Beckman, rearrange-
ments of â-tetralones may also be used to prepare 2-benza-
zepin-3-ones.13
* To whom correspondence should be addressed. E-mail: michael.d.wallace@
gsk.com.
† Current address: Fluorous Technologies, Inc., 970 William Pitt Way,
Pittsburgh, PA, 15238.
(1) Samanen, J.; Jonak, Z.; Rieman, D.; Yue, T.-L. Curr. Pharm. Des. 1997,
3, 545.
(2) Storgard, C. M.; Stupak, D. G.; Jonczyk, A.; Goodman, S. L.; Fox, R. I.;
Cheresh, D. A. J. Clin. InVest. 1999, 103, 47.
(3) Matsuno, H.; Stassen, J. M.; Vermlyn, J.; Deckmyn, H. Circulation 1994,
90, 2203.
(4) Carron, C. P.; Meyer, D. M.; Pegg, J. A.; Engleman, V. W.; Nickols, M.
A.; Settle, S. L.; Westlin, W. F.; Ruminski, P. G.; Nickols, G. A. Cancer
Res. 1998, 58, 1930.
(5) Engleman, V. W.; Nickols, G. A.; Ross, F. P.; Horton, M. A.; Griggs, D.
W.; Settle, S. L.; Ruminski, P. G.; Teitelbaum, S. L. J. Clin. InVest. 1997,
99, 2284.
(6) Miller, W. H.; Alberts, D. P.; Bhatnager, P. K.; Bondinell, W. E.; Callahan,
J. F.; Calvo, R. R.; Cousins, R. D.; Erhard, K. F.; Heerding, D. A.; Keenan,
R. M.; Kwon, C.; Manley, P. J.; Newlander, K. A.; Ross, S. T.; Samanen,
J. M.; Uzinskas, I. N.; Venslavsky, J. W.; Yuan, C. C.-K.; Haltiwanger, R.
C.; Gowen, M.; Hwang, S.-M.; James, I. E.; Lark, M. W.; Rieman, D. J.;
Stroup, G. B.; Azzarano, L. M.; Salyers, K. L.; Smith, B. R.; Ward, K. W.;
Johanson, K. O.; Huffman, W. F. J. Med. Chem. 2000, 43, 22.
(7) Croisier, P.; Rodriguez, L. Denmark Patent # DE2733869A1, 1978.
(8) Van Rompaey, K.; Van den Eynda, I.; DeKimpe, N.; Tourwe, D.
Tetrahedron 2003, 59, 4421.
(9) Busacca, C. A.; Johnson, R. E. Tetrahedron Lett. 1992, 33(2), 165.
(10) de Laszlo, S. E.; Bush, B. L.; Doyle, J. J.; Greenlee, W. J.; Hangauer, D.
G.; Halgren, T. A.; Lynch, R. J.; Schorn, T. W.; Siegl, P. K. S. J. Med.
Chem. 1992, 35, 833.
(11) Wittekind, R. R.; Lazarus, S. J. Heterocycl. Chem. 1971, 8, 495.
(12) Kamochi, Y. Daiichi Yakka Daigaku Kenkyu Nenpo 1988, 19, 19.
738
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development
10.1021/op0499021 CCC: $27.50 © 2004 American Chemical Society
Published on Web 08/17/2004

Scheme 1. Alternate Methods of Benzazepinone Synthesis
Scheme 3. Synthesis of SB-273005a
Scheme 2. Retrosynthetic Analysis of SB-273005
a (a) Br2, CH2Cl2, reflux, 65%. (b) (1) MeOH, HCl, rt; (2) itaconic acid, Et3N,
Pd(OAc)2, P(o-tolyl)3, Bu4NBr, CH3CN, 80%. (c) DCA, [RuCl2(R-BINAP)]2-
Et3N, 60 psi of H2, 60 °C, MeOH/H2O, 84%. (d) H2SO4, MeOH, reflux, 86%.
(e) (1) trifluoroethylamine-HCl, ZnCl2, CH3CN, reflux; (2) NaBH(OAc)3, DMA;
(3) TFA, toluene, reflux, 72%. (f) (1) PPh3, DIAD, 6-methylamino-2-py-
ridineethanol (compound 3), TBME; (2) LiOH H2O, THF/H2O, 50 °C, 66%.
3 derivatized as a suitable leaving group (vide infra).
Benzazepinone 4 would be synthesized from the chiral diester
aldehyde 5 after reductive amination with commercially
available trifluoroethylamine. Diester aldehyde 5 was figured
to arise from diesterification after asymmetric hydrogenation
of unsaturated diacid 6, which was reasoned as a Heck
product from coupling to itaconic acid after bromination of
the readily available 3-hydroxy benzaldehyde 7.
The bromination of inexpensive 3-hydroxybenzaldehyde
7 on a 100 kg scale using Br2 in CH2Cl2 to afford the known
bromoaldehyde 815 (Scheme 3) proceeded in reproducible
isolated yields of 63-68%. As expected the main problem
with this chemistry was formation of the regioisomeric
4-bromo-3-hydroxy benzaldehyde. Fortunately, the desired
product crystallized from the reaction mixture without
contamination from the regioisomer. Additionally, this simple
procedure increased the throughput and minimized the waste.
In an effort to address the modest yield, investigations after
the plant campaign revealed an improved 75-78% yield
when the starting substrate was treated with bromine in
glacial acetic acid. Recent articles using both sets of
conditions report yields consistent with ours.16,17
As mentioned above, our synthetic design of SB-273005
hinged upon the construction of an asymmetric synthesis of
an appropriate ortho substituted benzylamine intermediate
(which would come about via reductive amination) but
additionally relied upon Mitsunobu coupling (if not some
other classical means) of the newly prepared chiral 2,4,9-
trisubstituted benzazepinone to a 2,6-disubstituted pyridyl
alcohol (Scheme 2). Retrosynthetically, SB-273005 was
targeted as coming from a routine phenol alkylation of
benzazepinone 4 with the previously reported disubstituted
pyridine 3.14 The alkylation would be brought about via
Mitsunobu in the worst case or more likely with compound
(13) Smalley, R. K. Azepines. In ComprehensiVe Heterocyclic Chemistry, 1st
ed.; Katrizky, A. R., Ed.; Pergamon Press: Elmsford, NY, 1984; Vol. 7,
pp 529-532.
(14) Compound 3 was synthesized in 47% yield over seven steps (including
two steps for purification). Beginning with commercially available 2-amino
picoline, the amine was protected with (Boc)2O, followed by N-methylation.
The C6 methyl was homologated with diethyl carbonate, followed by LiBH4
reduction of the ester. The amine was deprotected, and purification was
carried out by conversion of the crude product to the aqueous soluble formate
salt, followed by liberation back to the free base after the impurities were
extracted. See ref 6 for details.
(15) Danckwortt, P. Ber. Dtsch. Chem. Ges. 1910, 42, 463.
(16) Kaiser, F.; Schwink, L.; Velder, J.; Schamlz, H. G. J. Org. Chem. 2002, 67
(26), 9248.
(17) Leo, P. M.; Morin, C.; Philouze, C. Org. Lett. 2002, 4, 2711.
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
739

Scheme 4. Catalyst Screen
Figure 2. Intramolecular aldol side reaction.
The Heck reaction between bromoaldehyde 8 and itaconic
acid was carried out by first converting the aldehyde into
the dimethylacetal. Simple stirring in methanol with catalytic
HCl effected this transformation, and the reaction progress
was easily monitored by 1H NMR (the transformation could
also be monitored by ReactIR). Treatment of the dimethy-
lacetal with itaconic acid, excess Et3N, tetrabutylammonium
bromide (TBAB), and 1.0% Pd(OAc)2 in a 1:3 molar ratio
with respect to tri-o-tolylphosphine generated the desired
product in 80% yield. In general, the E diastereoisomer was
1
the only product observed by H NMR. The workup and
isolation procedure was tedious and wasteful (mutilple
extractions, mutiple solvents, etc.), but it allowed for
purification of the product by extraction into aqueous base
and residual palladium levels were always less than the heavy
metal specification.18 Additionally, the disposal concerns and
decreased throughput were justified by the novelty of the
chemical process. Other researchers have coupled itaconate
esters to bromoanisoles,19,20 but our report is the first to
address direct coupling of itaconic acid to such a structurally
complex aryl bromide.
When the Heck chemistry was carried out directly on
bromoaldehyde 8, intramolecular aldol condensation chem-
istry of the product occurred as a side reaction (Figure 2).
The intramolecular aldol chemistry was presumably base
catalyzed given the excess triethylamine and the lower pKa
of the methylene protons of the Heck product. Given the
ease with which the acetal of bromoaldehyde 8 was formed
and reversed (the aldehyde was easily regenerated upon
exposure to aqueous acid), this simple additional transforma-
tion permitted us to sidestep the above aldol problem.
With unsaturated diacid 6 in hand, the asymmetric
hydrogenation was next examined. Putting this chemistry into
practice was more straightforward than we had anticipated.
For example, it was known that certain 2(E)-alkylidene
itaconic acid derivatives had been successfully hydrogenated
in the presence of chiral rhodium,21 and ruthenium,22 sources.
In addition, the application of Ru2Cl4[(+)-BINAP]2(Et3N)
on 2-acylaminoacrylic acids,23 and itaconic acid itself,24 has
produced high yields of the hydrogenated products in good
enantiomeric excesses. Therefore, several commercially
available catalysts, including the ruthenium based example
above, were screened for their ability to hydrogenate an
intermediate diacid (or its corresponding diester) from one
of our earlier syntheses (Scheme 4).
Entries 5 and 6 provided the most promising results;
however, the results from the rhodium DUPHOS catalyst
(entry 6) were not always reproducible. Additionally, the
ruthenium catalyst was less expensive,25 more easily handled
(air stable, easier to prepare, etc.), and routinely provided
reproducible results. Therefore, it was selected for further
development, and when it was applied to unsaturated diacid
6 in methanol (i.e., the dimethylacetal of compound 6) in
the presence of H2 at 60 psi and 60 °C, consistent yields
and ee’s were generated (80-85% chemical yield and 90-
95% ee). Although the results were encouraging, the
hydrogenated diacid would not easily crystallize, and thus a
brief crystallization study was undertaken.
Various dicarboxylate salts were examined (Na, K, Et3N,
etc.), but recrystallization of the dicylohexylamine (DCA)
salt proved to be the most consistent method. Recrystalli-
zation of the DCA salt improved the chemical purity from
95% to >99% and the ee from 94% to >99%. From these
results, it was envisaged that DCA could be directly added
to the hydrogenation reaction mixture. Fortunately this turned
out to be the case, and the crude DCA salt isolated from
precipitation of the reaction mixture was generally >98%
chemically pure and 98% enantiomerically pure. The general
conditions that were developed for this transformation were
2.1 equiv of DCA, 0.2 mol % catalyst, in aqueous methanol
at 60 psi of H2 and 60 °C. A simple solvent swap into
acetonitrile effected product precipitation in 84% chemical
yield with purity and ee as described above. The catalyst
was never recycled. The product was assigned an absolute
(18) Interestingly, the palladium level was of no consequence as it was observed
that Heck product contaminated with 1000 ppm of residual palladium
underwent asymmetric hydrogenation with ee comparable to the standard
case. This result was somewhat surprising to us because we thought that
bound palladium might competitively catalyze racemic hydrogenation.
(19) Buchwald, S.; Gurtler C. Chem.sEur. J. 1999, 5(11), 3107.
(20) Devi, A. R.; Rajaram, S. Synth. Comm. 1999, 29(4), 591.
(21) Talley, J. J. U.S. Patent 4,939,288, 1990.
(22) Genet, J. P.; Ratovelomanana-Vidal, V.; Mallart, S.; Pfister, X.; Bischoff,
L.; Cano De Andrade, M. C.; Darses, S.; Galopin, C.; Laffitte, J. A.
Tetrahedron: Asymmetry 1994, 5(4), 675.
(23) Ikariya, T., J., Ishii, Y.; Kawano, H.; Arai, T.; Saburi, M.; Yoshikawa, S.;
Akutagawa, S. J. Chem. Soc., Chem. Commun. 1985, 922. This paper also
describes catalyst preparation.
(24) Kawano, H.; Ishii, Y.; Ikariya, T.; Saburi, M.; Yoshikawa, S.; Uchida, Y.;
Kumobayashi, H. Tetrahedron Lett. 1987, 28(17), 1905.
(25) The cost of Ru2Cl4[(+)-BINAP]2(Et3N) was considerably cheaper due to
the fact that it was prepared in house (see ref 23 above for preparation).
740
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

configuration of (S) on the basis of X-ray analysis. In general,
material of this quality was then elaborated to API without
deterioration of chirality.
The diesterification of diacid acetal 9 to afford diester
aldehyde 5 was carried out uneventfully. Refluxing concen-
trated H2SO4 in MeOH served the purpose. The byproduct
DCA sulfate salts were easily removed by filtration after a
solvent swap into EtOAc, but the filtrate had to be tediously
washed with H2O, 10% H2SO4, and NaHCO3 before another
solvent swap and isolation in CH3CN. Typical yields were
85-90%, and the purity was sufficient for direct exposure
to the subsequent set of conditions.
Figure 3. Elimination problem.
statistical arrays of experiments28 were designed using all
of the above listed leaving groups and counterions (solvent
changes, and concentrations, were also examined). Unfor-
tunately, alkylation was never the major product under any
set of conditions, and many times vinyl pyridine 16 was the
exclusive product.29
The conversion of diester aldehyde 5 to the desired
benzazepinone 4 was carried out by a reductive amination-
cyclization sequence. First, the Shiff base was formed with
trifluoroethylamine-HCl, and catalytic ZnCl2, in refluxing
CH3CN with azeotropic removal of water. The Schiff base
intermediate was then reduced to the 2°-amine with NaBH-
(OAc)3 in DMA. Typically, the Schiff base chemistry was
Ironically, the chemistry that we had least desired to
optimize turned out to be our last hope. The Mitsunobu
chemistry was unique in that it was the only method that
afforded the desired product in a reasonable yield (80%).
This fact necessitated its development. In general, the most
reproducible results were achieved using PPh3 and DIAD
as activating agents and TBME as reaction solvent. Typically,
the product was contaminated by 10% of starting material
benzazepinone 4 and 10% vinyl pyridine 16 in addition to
the triphenylphosphine oxide and DIAD hydrazide. Starting
material was easily removed with a 1 N NaOH wash.
Removal of triphenylphosphine oxide was not quantitative,
but approximately 75-80% of it could be removed by
precipitation from TBME. The DIAD hydrazide, the vinyl
pyridine, and the residual 20% phosphine oxide were not
easily removed. Initially, large scale chromatography was
considered, but most eluent mixtures did not provide efficient
resolutions.
Fortunately, we were able to take advantage of the fact
that the subsequent step was a saponification. After the
phosphine oxide was filtered off, and the organic layer was
washed with 1 N NaOH, the crude mixture in TBME was
diluted with an aqueous solution of LiOH‚H2O and the
bilayered mixture was refluxed. The desired product was
sufficiently aqueous soluble as a carboxylate salt and the
organic side products were easily removed by extraction into
the organic layer. SB-273005 then precipitated from the
aqueous solution after dilution with methanol and acidifica-
tion with HCl. The crude product was directly recrystallized
from MeOH to afford 99% pure SB-273005 in 66% yield
over two steps. This streamlined process permitted an
otherwise undesirable reaction to be scaled and dramatically
reduced any cost that would have been associated with a
bulk chromatography.
1
monitored by H NMR. The desired chemistry was always
complicated by incomplete consumption of the aldehyde, and
after reduction the desired amine product was usually
contaminated with 10-15% of the benzyl alcohol product
even after workup.26 After NaHCO3 quench, extraction into
EtOAc, and solvent swap into toluene, the crude amine was
directly cyclized.
Cyclization of amine diester was conducted with TFA in
refluxing toluene. Lactam formation was generally complete
within 3-4 h and was unaccompanied by competitive lactone
formation from the benzyl alcohol contaminant in the starting
material. In addition, as reported earlier,27 none of the
regioisomeric eight-membered ring lactam product was
observed. Presumably, the transannular and Pitzer strains are
reduced in the transition state leading to the seven-membered
ring. After workup (aqueous H2SO4, aqueous NaHCO3, and
H2O), the 2,4,9-trisubstituted-2-benzazepin-3-one compound
4 was precipitated from toluene in 72% yield, with >97%
chemical purity and >98% ee.
Finally, benzazepinone 4 was coupled to the aforemen-
tioned 2-methylamino-6-ethanol pyridine 3 via Mitsunobu
chemistry. In the planning stages, the Mitsunobu was thought
to be the last method by which we would choose to construct
the desired product. Mitsunobu chemistry is environmentally
unfriendly, atom uneconomical, and complicated to purify
due to the phosphine oxide and hydrazide byproducts.
Therefore, the simple Williamson etherification (i.e., phe-
noxide alkylation with an appropriate leaving group at the
ethyl terminus of the 2,6-disubstituted pyridyl alcohol) was
considered first.
The overall cost to produce SB-273005 on scale was
$2100.00/kg which we believe was a very competitive price
given the originality of the overall synthesis. The synthesis
included a novel Heck reaction, a new asymmetric hydro-
What was perceived to be a routine phenoxide alkylation
was terminally complicated by variable amounts of elimina-
tion to vinyl pyridine 16 (Figure 3). Simple modification of
the terminal hydroxy of disubstituted pyridine 3 as a tosylate
or mesylate (using potassium as the phenoxide counterion)
did not increase the amount of alkylation. At that point,
(28) The equipment used for the numerous experiments was the Anachem SK233
coupled to an Agilent 1100 HPLC.
(29) The chronic failure of this O-alkylation chemistry was thought to be inherent
in the electrophile. A preferred anti conformation of the leaving group and
pyridine ring would force the nucleophile in an SN2 type mechanism syn
periplanar to the pyridine ring thereby generating unfavorable steric
interactions.
(26) Unfortunately, ReactIR could not detect the iminium ion, but presumably
the rate of consumption of the iminum ion was faster than the aldehyde
reduction.
(27) See ref 7.
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
741

genation, and a unique application of the Mitsunobu etheri-
fication. The yield for the brief seven-step synthesis was 15%
(76% average yield for each step). Finally, the synthetic route
was scaled to 100 kg quantities and could be easily scaled
to manufacture much greater amounts.
vessel was maintained at 60 psi of pressure and the contents
were heated at 60 °C. The mixture was stirred for 30 h before
1H NMR confirmed the consumption of starting material.
The contents were cooled to ambient temperature and filtered.
Approximately half of the filtrate was removed by distilla-
tion. To the residual mixture was added 400 L of CH3CN,
and the mixture was distilled to approximately half volume.
The dilution/distillation process was repeated twice. The
residual mixture was cooled to 20-25 °C, and it was held
at that temperature for 6 h. The solid was then collected,
rinsed with cold CH3CN, and dried in vacuo (40° C, 20 in.
Hg) to afford 111 kg of 9, 84% yield, 98% pure by HPLC:
Hypersil HyPurity Elite C18, 150 mm × 4.6 mm i.d., 5 µ;
conditions, isocratic 90% A/10% B [A ) 100% H2O/0.1%
TFA, B ) 100% CH3CN/0.1% TFA], run time ) 12 min,
flow rate ) 2.0 mL/min, λ ) 223 nm. >98.0% ee by chiral
HPLC: Chiralcel OD-H, 250 mm × 4.6 mm i.d., 5 µ;
conditions, isocratic 93% hexane/7.0% IPA/0.1% TFA, run
time ) 30 min, flow rate ) 1.0 mL/min, sample concentra-
tion ) 1.0 mg/mL, λ ) 223 nm, desired product (S) ) 25
Experimental Section
Reagents from commercial sources were used as is.
Tetramethylsilane was used as the internal standard for NMR
work.
2-[(2-Formyl-4-hydroxyphenyl)methylidene]succinic
Acid (6). 6-Bromo-3-hydroxybenzaldehyde (62 kg, 310 mol,
1.0 equiv) was dissolved in 253 L of MeOH. The resulting
solution was filtered. To the solution was added concentrated
HCl (1.5 kg, 41.6 mol, 0.13 equiv). The mixture was stirred
at room temperature for 2 h. After confirmation by 1H NMR
that the dimethyl acetal was completely formed, Et3N (159
kg, 1571 mol, 5.0 equiv) was added, followed by 615 L of
CH3CN. The reaction mixture was purged with N2. To the
mixture was added itaconic acid (41 kg, 315 mol, 1.0 equiv),
Pd(OAc) (0.70 kg, 3.1 mol, 0.01 equiv), P(o-tolyl) (2.9
2
3
1
min. H NMR (400 MHz, d6-DMSO): δ 1.06-1.88 (over-
kg, 9.3 mol, 0.03 equiv), and Bu4NBr (10 kg, 31 mol, 0.1
equiv). The resulting reaction mixture was heated at reflux
for 10 h at which point the reaction was deemed complete
by HPLC. After cooling to room temperature, about half of
the reaction solvent was removed by distillation. Aqueous
KOH solution (235 L at 15% w/w) was added at room
temperature. The aqueous solution was washed with 250 L
of TBME. The aqueous solution was acidified to pH 1 using
282 L of 3 N HCl solution. The acidic aqueous solution was
extracted with 4 × 250 L of TBME. The combined TBME
extracts were filtered. About half of the reaction solvent was
removed by distillation. To the mixture was added 250 L of
acetonitrile. The distillation/dilution process was then re-
peated 3 times. The heterogeneous solution was cooled to
-10 °C for 2 h, and the resulting precipitate was collected.
The cream-colored solid product was dried under vacuum
(50 °C, 20 in. Hg) to afford 61.3 kg of 6, 79% yield, 98%
pure by HPLC: Ultrasphere ODS 5 µ, 150 mm × 4.6 mm
i.d.; conditions, isocratic 20% A/80% B[ A ) 90% H2O/
10% CH3CN/0.1% TFA, B ) 90% CH3CN/10% H2O/0.1%
TFA], run time ) 30 min, flow rate ) 1.0 mL/min, λ )
lapping multiplets, 40 H), 2.51 (m, 4H), 2.13 (dd, 1 H), 2.22
(dd, 1 H), 2.43 (dd, 1 H), 3.03 (dd, 1 H), 2.50 (m, 1 H),
3.22-3.23 (2 s, 6 H), 5.40 (s, 1 H), 6.63 (dd, 1 H), 6.87 (d,
1 H), 6.95 (d, 1 H), 9.20 (bs, 1 H). 13C NMR (75 MHz,
d6-DMSO) 177.6, 175.0, 155.6, 137.5, 131.6, 129.1, 115.4,
113.7, 101.4, 53.5, 53.5, 52.7, 44.7, 38.0, 33.1, 31.6, 25.8,
24.7 ppm. MS (Ion Mode: ESI) m/z ) 275 [M + Na]+.
Dimethyl (2S)-2-[(2-Formyl-4-hydroxyphenyl)methyl]-
butanedioate (5). (S)-2-Carboxyl-4-[(2-formyldimethyl-
acetal-4-hydroxyphenyl)butyric acid, bis(dicycyclohexyl-
amine) salt 9 (99 kg, 150 mol, 1.0 equiv) was dissolved in
620 L of MeOH. To the solution was added concentrated
H2SO4 (49 kg, 500 mol, 3.3 equiv). The mixture was heated
at reflux for 19 h at which point the reaction was deemed
complete by HPLC. The mixture was cooled to 20-25 °C.
Approximately 75% of the MeOH was removed by distil-
lation. To the residual mixture was added 485 L of EtOAc.
Approximately 70% of the volume was removed by distil-
lation. The dilution/distillation process was repeated 3 times.
The solvent was adjusted to approximately 500 L. The
mixture was cooled to 0-5 °C, and it was held at that
temperature for 1 h. The solid was removed by filtration.
The filtrate was extracted with 2 × 240 L of water, 270 L
of 10% aqueous H2SO4, and 250 L of 8% aqueous NaHCO3.
Approximately half of the organic layer was removed by
distillation. To the residual mixture was added 240 L of CH3-
CN, and the volume was reduced by half. The dilution/
distillation process was repeated. The volume was adjusted
to approximately 500 L with CH3CN. The CH3CN solution
was assayed and found to contain 36 kg of 5, 86% yield,
>96% pure by HPLC: Symmetry C18, 4.6 mm i.d. × 250
mm, 5 µ; conditions, isocratic 80% A/10% B [A ) 90:10:
0.1 H2O/CH3CN/TFA, B ) 90:10:0.1 CH3CN/H2O/TFA],
run time ) 15 min, flow rate ) 1.0 mL/min, λ ) 220 nm.
1H NMR (400 MHz, d6-DMSO): δ 2.48, 2.60 (dd, 2 H),
2.94 (m, 1 H), 3.13 (m, 1 H), 3.48 (s, 3 H), 3.54 (s, 3 H),
6.98 (dd, 1 H), 7.09 (d, 1 H), 7.21 (d, 1 H), 9.75 (bs, 1 H),
1
220 nm. H NMR (400 MHz, d6-DMSO): δ 3.17 (s, 2 H),
7.10 (dd, 1 H), 7.22 (d, 1 H), 7.30 (d, 1 H), 8.11 (s, 1 H),
10.01 (s, 1 H), 10.23 (s, 1 H), 12.51 (bs, 2 H). 13C NMR
(75 MHz, d6-DMSO) δ 192.8, 172.6, 168.6, 158.6, 138.9,
135.5, 131.4, 128.6, 128.0, 121.4, 117.8, 34.0 ppm. MS (Ion
Mode: ESI) m/z ) 251 [M + H]+, 273 [M + Na]+.
(S)-2-Carboxyl-4-[(2-formyldimethylacetal-4-hy-
droxyphenyl)]butyric Acid, Bis(dicyclohexylamine) Salt
(9). 2-[(2-Formyl-4-hydroxyphenyl)methylidene]succinic acid
6 (50 kg, 200 mol, 1.0 equiv) was dissolved in 455 L of
refluxing MeOH. After 4 h at reflux, the solution was cooled
to ambient temperature and filtered. After the filtrate had
been transferred to an appropriate hydrogenation vessel, DCA
(77 kg, 425 mol, 2.1 equiv) was added, followed by 50 L of
water. After purging and venting with nitrogen, [RuCl2(R-
BINAP)] -TEA (0.125 kg, 0.25 wt %) was added. After
2
purging and venting with hydrogen 3 times, the reaction
742
•
Vol. 8, No. 5, 2004 / Organic Process Research & Development

10.10 (s, 1 H). 13C NMR (75 MHz, d6-DMSO) δ 193.3,
174.4, 172.1, 157.0, 135.2, 133.5, 131.3, 121.5, 117.8, 57.9,
51.9, 43.5, 35.5, 33.0 ppm. MS (Ion Mode: ESI) m/z ) 303
[M + Na]+.
fluoroethyl)-2,3,4,5-tetrahydro-1H-benzo[c]azepin-4-yl]acetic
acid methyl ester (50 kg, 151 mol, 1.0 equiv) was added
500 L of TBME. To the heterogeneous mixture was added
6-(methylamino)-2-pyridineethanol (25.3 kg, 166 mol, 1.1
equiv) that was dissolved in 250 L of TBME. To the mixture
was added PPh3 (43.5 kg, 166 mol, 1.1 equiv). The
heterogeneous mixture was cooled to 0-5 °C. To the cooled
mixture was added DIAD (33.5 kg, 166 mol, 1.1 equiv). The
mixture was stirred at 20-25 °C for 2-3 h at which point
the reaction was deemed complete by HPLC. The mixture
was concentrated to approximately one-third of its original
volume. The mixture was cooled to 0-5 °C and stirred at
that temperature for 30 min. The solids were removed by
filtration. The filtrate was extracted with 100 L of 3 N NaOH
and 100 L of brine. To the organic layer was added LiOH‚
H2O (13.9 kg, 332 mol, 2.2 equiv) that was dissolved in 125
L of water. The mixture was heated at 50-55 °C for 2-3 h
at which point the reaction was deemed complete by HPLC.
The mixture was cooled to 20-25 °C, and it was diluted
with 375 L of water. The layers were separated, and the
aqueous layer was washed with 250 L of TBME. To the
aqueous layer was added 125 L of MeOH. The aqueous
methanolic mixture was acidified with 10-12 L of concen-
trated HCl until the pH ) 6.0-7.0. The mixture was seeded
with authentic product. The mixture was further acidified
with 10-12 L of concentrated HCl until pH ) 5.0-5.5. The
mixture was cooled to 0-5 °C and stirred at that temperature
for 30 min. The crude precipitate was collected by filtration.
The crude product was dissolved in 450 L of hot MeOH,
seeded with pure product, and then allowed to stand at 20-
25 °C for 16 h. The recrystallized product was collected by
filtration. The product was then oven dried in vacuo (40 °C,
20 in. Hg) to afford 44.9 kg of 1, 66% yield, 98% pure by
HPLC: YMC Basic, 4.6 mm i.d. × 150 mm, 5 µ; conditions,
gradient 80% H2O/CH3CN to 20% H2O/CH3CN over 10 min,
run time ) 15 min, flow rate ) 2.0 mL/min, λ ) 233 nm.
>98% ee by chiral HPLC: Chiralcel OJ 250 mm × 4.6 mm
i.d., 10 µ; isocratic 70% hexane/30% H2O/0.1% TFA/0.1%
diethylamine, run time ) 20 min, flow rate ) 1.0 mL/min,
λ ) 234 nm. 1H NMR (400 MHz, d6-DMSO): δ 2.37, 2.66
(dd, 2 H), 2.66, 2.98 (dd, 2 H), 3.74 (m, 1 H), 4.19 (m, 2
H), 4.13, 5.28 (d, 2 H), 6.78 (m, 1 H), 6.79 (m, 1 H), 7.00
(d, 1 H), 2.93 (t, 2 H), 4.25 (t, 2 H), 2.73 (d, 3 H), 6.32 (q,
1 H), 6.25 (dd, 1 H), 6.41 (dd, 1 H), 7.28 (dd, 1 H), 11.90
(bs, 1 H). 13C NMR (75 MHz, d6-DMSO) δ 174.7, 173.1,
159.3, 156.2, 155.9, 137.0, 135.1, 131.3, 128.0, 124.7, 114.9,
113.9, 110.6, 105.1, 66.9, 52.0, 46.7, 37.3, 36.4, 36.1, 33.7,
27.9 ppm. MS (Ion Mode: ESI) m/z ) 452 [M + 1]+.
[(S)-9-Hydroxy-3-oxo-2-(2,2,2-trifluoro-ethyl)-2,3,4,5-
tetrahydro-1H-benzo[c]azepin-4-yl]acetic Acid Methyl
Ester (4). To (S)-2-(2-formyl-4-hydroxybenzyl)succinic acid
dimethyl ester 5 (56 kg, 200 mol, 1.0 equiv) in 562 L of
CH3CN was charged trifluoroethylamine-HCl (32.5 kg, 240
mol, 1.2 equiv) and ZnCl2 (13.5 kg, 100 mol, 0.5 equiv).
The mixture was heated at reflux, while water and hydro-
chloric acid were azeotropically removed. After 2-3 h the
water content was found to be <1.0%. The mixture was
cooled to 20-25 °C. To the imine solution was added NaBH-
(OAc)3 (92.5 kg, 436 mol, 2.2 equiv) that was dissolved in
244 L of DMA. After 30 min the reaction was deemed
complete by HPLC. The mixture was quenched with 160 L
of 8.0% aqueous NaHCO3. The mixture was then further
diluted with 1120 L of 8% aqueous NaHCO3 and 372 L of
TBME. After the layers were separated, approximately half
the volume of the organic layer was removed by distillation.
The residual mixture was diluted with 265 L of toluene, and
it was distilled until approximately half volume. The dilution/
distillation process was repeated. The residual mixture was
diluted with toluene to approximately 400 L. To the mixture
was added TFA (15.4 kg, 135 mol, 0.7 equiv). The mixture
was heated at reflux for 4 h at which point the reaction was
deemed complete by HPLC. The mixture was cooled to 20-
25 °C and then distilled to approximately half volume. To
the mixture was added 270 L of EtOAc. The mixture was
washed with 270 L of 10% aqueous H2SO4, 270 L of 5%
aqueous NaHCO3, and 270 L of water. The organic layer
was distilled to approximately one-third volume. To the
residual mixture was added 230 L of toluene. The mixture
was distilled to approximately half volume. The dilution/
distillation process was repeated twice. The mixture was
cooled to 0-5 °C and held at that temperature for 30 min.
The precipitate was collected, washed with toluene, and oven
dried in vacuo (40 °C, 20 in. Hg) to afford 47.6 kg of 4,
72% yield, 99% pure by HPLC: Ultrasphere ODS 150 mm
× 4.6 mm i.d., 5 µ; conditions, isocratic 80% A/20% B [A
) 90% H2O/10% CH3CN/0.1% TFA, B ) 90% CH3CN/
10% H2O/0.1% TFA], run time ) 20 min, flow rate ) 1.0
mL/min, λ ) 240 nm, 98% ee by chiral HPLC: Chiralpak
AD 250 mm × 4.6 mm i.d., 10 µ; conditions, isocratic 85%
hexane/IPA, run time ) 20 min, flow rate ) 1.0 mL/min, λ
1
) 230 nm. H NMR (400 MHz, d6-DMSO): δ 2.48, 2.71
(dd, 2 H), 2.64, 2.93 (dd, 2 H), 3.77 (m, 1 H), 3.58 (s, 3 H),
4.14, 4.30 (dq, 2 H), 4.07 (d, 1 H), 5.26 (d, 1 H), 6.57 (d, 1
H), 6.63 (dd, 1 H), 6.90 (d, 1 H), 8.29 (bs, 1 H). 13C NMR
(75 MHz, d6-DMSO) δ 174.6, 172.0, 154.9, 134.8, 131.2,
125.9, 124.7, 115.5, 114.7, 52.1, 51.2, 46.7, 36.2, 36.0, 33.7
ppm. MS (Ion Mode: ESI) m/z ) 354 [M + Na]+.
Acknowledgment
The authors thank the Analytical Sciences groups at
SmithKline Beecham and GlaxoSmithKline.
S-(-)-9-[2-[6-(Methylamino)pyridin-2-yl]-1-ethoxy]-3-
oxo-2-(2,2,2-trifluoroethyl)-1,2,4,5-tetrahydro-2-benzaepine-
4-acetic Acid (1). To [(S)-9-hydroxy-3-oxo-2-(2,2,2-tri-
Received for review May 14, 2004.
OP0499021
Vol. 8, No. 5, 2004 / Organic Process Research & Development
•
743