Enantioselective synthesis of a key intermediate in a new process for orlistat using asymmetric hydrogenation and a grignard reagent promoted lactone cyclization

Article abstract of DOI:10.1021/op060208q

A new enantioselective synthesis of Orlistat suitable for large-scale preparation is described. Therein, the first isolated key intermediate (R)-3-hexyl-5,6-dihydro-4-hydroxy-6-undecyl-2H-pyran-2-one (12) is prepared via (a) the asymmetric hydrogenation of methyl 3-oxotetradecanoate to (S)-3-hydroxytetrade-canoate (9); (b) the acylation of 9 with 2-bromooctanoyl halide (bromide/chloride) to (R)-3-[(2-bromo-1-oxooctyl)oxy]-tetradecanoic acid methyl ester (11) and finally (c) the tert-butyl magnesium chloride promoted cyclization of 11 to the single enantiomer 12. The single enantiomer intermediate 12, previously published as a mixture of enantiomers 2, has been carried on through several steps to Orlistat (1) without any process changes.

Full text of DOI:10.1021/op060208q

Organic Process Research & Development 2007, 11, 524−533
Enantioselective Synthesis of a Key Intermediate in a New Process for Orlistat
Using Asymmetric Hydrogenation and a Grignard Reagent Promoted Lactone
Cyclization
Mark A. Schwindt,*,† Michael P. Fleming,† Yeun-Kwei Han,† Lewis M. Hodges,† David A. Johnston,† Roger P. Micheli,†
Chris R. Roberts,† Roger Snyder,† Robert J. Topping,† Kurt Pu¨ntener,‡ and Michelangelo Scalone‡
Roche Colorado Corporation, Boulder Technology Center, 2075 North 55th Street, Boulder, Colorado 80301, U.S.A, and
F. Hoffmann-La Roche Ltd., Pharmaceuticals DiVision, Synthesis & Process Research, Grenzacherstrasse 124,
CH-4070 Basel, Switzerland
Abstract:
A new enantioselective synthesis of Orlistat suitable for large-
scale preparation is described. Therein, the first isolated key
intermediate (R)-3-hexyl-5,6-dihydro-4-hydroxy-6-undecyl-2H-
pyran-2-one (12) is prepared via (a) the asymmetric hydrogena-
tion of methyl 3-oxotetradecanoate to (S)-3-hydroxytetrade-
canoate (9); (b) the acylation of 9 with 2-bromooctanoyl halide
(bromide/chloride) to (R)-3-[(2-bromo-1-oxooctyl)oxy]-tetrade-
canoic acid methyl ester (11) and finally (c) the tert-butyl
magnesium chloride promoted cyclization of 11 to the single
enantiomer 12. The single enantiomer intermediate 12, previ-
ously published as a mixture of enantiomers 2, has been carried
on through several steps to Orlistat (1) without any process
changes.
Figure 1. Orlistat (1).
Scheme 1. First-generation, optical resolution based
synthesis of Orlistat (1)
Introduction
Orlistat (1), the active pharmaceutical ingredient of
Xenical, is a novel antiobesity agent that selectively inhibits
gastrointestinal lipases, reducing the absorption of dietary
fat by about 30% (Figure 1).1 Also known as (-)-tetrahy-
drolipstatin or N-formyl-(S)-leucine (S)-1-[(2S,3S)-3-hexyl-
4-oxo-2-oxetanyl]methyldodecyl ester, 1 is the hydrogenated
derivative of lipstatin, a naturally occurring lipase inhibitor
isolated from Streptomyces toxytricini.2 (-)-Tetrahydrolip-
statin has been the research target of numerous synthetic
groups.3
A diagram of the first-generation Orlistat synthesis based
on the optical resolution of the racemic key intermediate 4
is shown in Scheme 1.3u The isolated intermediates are the
racemic dihydropyranone 2, racemic all-cis δ-lactone 3, the
enantiomerically pure 1:1 PEA-salt 4, and the (S,S,R)-
oxetanone 5.
The required relative configuration of the three stereo-
centers in 3 was established during the diastereoselective,
heterogeneous hydrogenation of 2, which afforded selectively
3 as a 1:1 mixture of the two all-cis enantiomers. Subse-
quently, the required absolute configuration for each of these
centers was selected by the optical resolution of 4. Finally,
the fourth chiral center of Orlistat was introduced during the
Mitsunobu coupling reaction of 5 with N-formyl-L-leucine.
The objective to elaborate a new, enantioselective, second-
generation synthesis of 1 was to develop and demonstrate a
more efficient process with an increased throughput. The
strategy therefore was to retain the final reaction sequence
4 to 1 but to avoid the first-generation’s late-stage resolution
* To
whom
correspondence
should
be
addressed.
E-mail:
mark.schwindt@roche.com.
† Roche Colorado Corporation.
‡ F. Hoffman-La Roch Ltd.
(1) (a) Hauptman, J. B. 8th International Congress on Obesity, Paris, France,
August 1998. (b) Prous, J.; Mealy, N.; Castaner, J. Drugs of the Future
1994, 19, 1003.
(2) (a) Weibel, E. K.; Hadvary, P.; Hochuli, E.; Kupfer, E.; Lengsfeld, H. J.
Antibiot. 1987, 40, 1081. (b) Hochuli, E.; Kupfer, E.; Mauer, R.; Meister,
W.; Mercadal, Y.; Schmidt, K. J. Antibiot. 1987, 40, 1086.
524
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development
10.1021/op060208q CCC: $37.00 © 2007 American Chemical Society
Published on Web 01/09/2007

Scheme 2. Second-generation, new enantioselective
synthesis of Orlistat (1)
Figure 2. New retrosynthetic strategy: (S)-hydroxyester 9, (R)-
dihydropyrone 12 and (3S,4S,6R)-lactone 13.
of 4 by the use of enantiomerically pure intermediate 12
(Figure 2) rather than its racemate analogue 2.
A few of the envisioned advantages of this second-
generation strategy were the following: (a) the only asym-
metric center to be created by asymmetric synthesis would
involve a well-precedented asymmetric reduction of a
â-ketoester;3e,4,5 (b) the throughput from 12 to 5 would be
greatly increased due to avoidance of the undesired enanti-
omer of 2; (c) the process development from 12 to 1 was to
be expedited by utilization of much of the first-generation
process technology; and (d) the critical last step and
crystallization were to be left unchanged.
Several synthetic routes6 to 12 were evaluated preliminar-
ily. Scheme 2 illustrates the final optimized magnesium-
mediated route.7
The magnesium-mediated route was a basic extension of
an earlier approach reported by Ramig and co-workers from
Roche Nutley, who described the acylation of 9 to the diester
11 and the subsequent zinc-mediated Reformatsky cyclization
of 11 to 12 in about a 60% yield.3f,8
In their publication, Ramig and co-workers did not
mention the use of magnesium for the cyclization step but
(3) (a) Barbier, P.; Schneider, F. HelV. Chim. Acta 1987, 70, 196. (b) Pons, J.
M.; Kocienski, P. Tetrahedron Lett. 1989, 30, 1833. (c) Fleming, I.;
Lawrence, N. J. Tetrahedron Lett. 1990, 31, 3645. (d) Chadha, N. K.; Batcho,
A. D.; Tang, P. C.; Courtney, L. F.; Cook, C. M.; Wovkulich, P. M.;
Uskokovic, M. R. J. Org. Chem. 1991, 56, 4714. (e) Case-Green, S. C.;
Davies, S. G.; Hedgecock, C. J. R. Synlett 1991, 11, 781. (f) Landi, J. J.,
Jr.; Garofalo, L. M.; Ramig, K. Tetrahedron Lett. 1993, 34, 277. (g)
Hanessian, S.; Tehim, A.; Chen, P. J. Org. Chem. 1993, 58, 7768. (h)
Pommier, A.; Pons, J.-M. Synthesis 1994, (Spec. Issue), 1294. (i) Giese,
B.; Roth, M. J. Braz. Chem. Soc. 1996, 7, 243. (j) Paterson, I.; Doughty,
V. A. Tetrahedron Lett. 1999, 40, 393. (k) Wedler, C.; Costisella, B.; Schick,
H. J. Org. Chem. 1999, 64, 5301. (l) Dirat, O.; Kouklovsky, C.; Langlois,
Y. Org. Lett. 1999, 1, 753. (m) Ghosh, A. K.; Liu, C. Chem. Commun.
1999, 17, 1743. (n) Parsons, P. J.; Cowell, J. K. Synlett 2000, 1, 107. (o)
Ghosh, A. K.; Fidanze, S. Org. Lett. 2000, 2, 2405. (p) Honda, T.; Endo,
K.; Ono, S. Chem. Pharm. Bull. 2000, 48, 1545. (q) Sato, M.; Nakashima,
H.; Hanada, K.; Hayashi, M.; Honzumi, M.; Taniguchi, T.; Ogasawara, K.
Tetrahedron Lett. 2001, 42, 2833. (r) Bodkin, J. A.; Humphries, E. J.;
McLeod, M. D. Tetrahedron Lett. 2003, 44, 2869. (s) Thadani, A. N.; Batey,
R. A. Tetrahedron Lett. 2003, 44, 8051. (t) Polkowska, J.; Lukaszewicz,
E.; Kiegiel, J.; Jurczak, J. Tetrahedron Lett. 2004, 45, 3873. (u) One process
uses lauroyl aldehyde, hexylketoester, and N-formyl-L-leucine as key raw
materials (Karpf, M.; Zutter, U. U.S. Patent 5245056A, 1993, Eur. Pat. Appl.
EP 443449 A2, 1991).
(4) Heiser, B.; Broger, E.; Crameri, Y. Tetrahedron: Asymmetry 1991, 2, 51.
(5) A preliminary report on this topic has been given in: (a) Schmid R.; Scalone
M. ComprehensiVe Asymmetric Catalysis III; Jacobsen, E. N., Pfalz, A.,
Yamamoto, H., Eds.; 1999; p 1446. (b) Schmid, R.; Broger, E. A.;
Cereghetti, M.; Crameri, Y.; Foricher, J.; Lalonde, M.; Mu¨ller, R. K.;
Scalone, M.; Schoettel, G.; Zutter, U. Pure Appl. Chem. 1996, 68, 131.
(6) Harrington, P. J.; Hodges, L. M.; Pu¨ntener, K.; Scalone, M. Eur. Pat. Appl.
EP 1127886 A1, 2001; U.S. Patent 6552204 B1, 2000.
(7) (a) Fleming, M. P.; Han, Y.-K.; Hodges, L. M.; Johnston, D. A.; Micheli,
R. P.; Pu¨ntener, K.; Roberts, C. R.; Scalone, M.; Schwindt, M. A.; Topping,
R. J. U.S. Patent 6858749 B2, 2005; U.S. Patent 6743927 B2, 2004; U.S.
Patent 6818789 B2, 2004. (b) A preliminary account of this work has been
accepted for publication: Birk, R.; Karpf M.; Pu¨ntener K.; Scalone M.;
Schwindt M.; Zutter U. Chimia 2006, 60, 561, Special Issue: Process
Chemistry.
mentioned two papers which described magnesium-metal
based cyclizations to γ-lactones in moderate yields.9,10
A
number of researchers have found ways to utilize Grignard
reagents in the preparation of ketones from carboxylic acid
derivatives by various methods.11 The large-scale use of
magnesium rather than zinc is advantageous due to fewer
disposal issues with magnesium salt wastes. Also, the use
of magnesium would allow utilization of the higher-boiling
and less-flammable tetrahydrofuran rather than diethyl ether.
However, the use of magnesium under Ramig’s conditions
failed to afford 12 in an acceptable yield. The following
discussion describes the development of a novel enantiose-
lective synthesis of (R)-3-hexyl-5,6-dihydro-4-hydroxy-6-
undecyl-2H-pyran-2-one 12 as a key intermediate in a new
synthesis of Orlistat from the conception of the idea to the
demonstration of a potential second generation large-scale
process.
Results and Discussion
Synthesis of (R)-Dihydropyrone 12. The first reaction
sequence in the new, enantioselective synthesis of 1 consists
of three distinct chemical reactions: (a) the asymmetric
(9) Kelly, T. R.; Chandrakumar, N. S.; Cutting, J. D.; Goehring, R. R.; Weibel,
F. R. Tetrahedron Lett. 1985, 26, 2173.
(10) Li, W.-R.; Ewing, W. R.; Harris, B. D.; Joullie, M. M. J. Am. Chem. Soc.
1990, 112, 7659.
(11) O’Neill, B. T. Nucleophilic Addition to Carboxylic Acid Derivatives. In
ComprehensiVe Organic Synthesis; Trost, B. M., Fleming, I., Eds.; 1991; p
398.
(8) (a) Ramig, K. M.; Landi, J. J., Jr. U.S. Patent 5274143, 1993. (b) Ramig,
K. M.; Landi, J. J., Jr. U.S. Patent 5420305, 1995.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
525

Scheme 3. Structure of various chiral diphosphine ligands
hydrogenation of 6 to (R)-hydroxyester 9; (b) the acylation
of 9 to the (R)-bromodiester 11; and (c) the cyclization of
11 to (R)-dihydropyranone 12 with only 12 being isolated
as a solid intermediate.
a. Asymmetric Hydrogenation of 6: The substrate for
the asymmetric hydrogenation, the â-ketoester 6, became
commercially available during the course of our investiga-
tions but was first prepared in-house on a multi-100 kg scale
by condensation of lauroyl chloride with the magnesium salt
of methyl acetoacetate in up to 87% yield.7a,12 Originally
the starting material (S)-hydroxyester 9 was prepared by a
Raney-Nickel/tartaric acid catalyzed asymmetric hydrogena-
tion of 6. This heterogeneous hydrogenation process was
successfully used in the first in-house production of 9 on a
multi-100 kg scale (84-91% ee).5 However, a single use of
the catalyst was economically not feasible, and a reuse
proved to be on a large scale not practicable because after
each run a laborious remodification with tartaric acid was
required to maintain its activity and selectivity. Furthermore,
one crystallization step was necessary to enhance the ee of
9 to >99%. Therefore this process was not further developed.
The promising results obtained with the ruthenium catalyzed
asymmetric hydrogenation of 6 using (S)-BINAP3e or a (S)-
BIPHEMP4 based catalysts (97% ee) resulted in a focus on
this approach with the goal to elaborate a technically feasible,
highly enantioselective (>99% ee) hydrogenation process
which affords enantiomerically pure 9 and avoids an ee-
enhancing crystallization step.
Scheme 4. Synthesis of the hydrogenation catalyst
The pre-catalysts 8a-m for asymmetric hydrogenation
of 6 were prepared by reaction of the chiral diphosphines
(structures cf. Scheme 3) with commercially available [RuCl2-
(COD)]n (exemplified with 8b, Scheme 4).4,13 As expected,
the diacetoxy complex 8b was catalytically inactive (Table
1, entry 1), but addition of hydrogen chloride generated a
catalytically active species.5,14 Although only 2 molar equiv
of hydrogen chloride are formally required to exchange the
two coordinated acetates in 8b and form the proposed
dichloro catalyst (Scheme 4),15 the addition of up to a 20-
fold excess was extremely advantageous leading to highly
increased reaction rates and higher enantioselectivities. The
exact role of the excess of chloride remains unknown; we
assume it drives the pre-catalyst to the active catalyst, which
we suggest to write as [RuCln((R)-MeOBIPHEP)], not
excluding the presence of dimeric species.16 Suitable chloride
sources were both aqueous and anhydrous hydrogen chloride.
Interestingly, lithium chloride led to a very active catalyst,
whereas other additives such as HBr, Bu4NI, HBF4, or
p-toluenesulfonic acid were much less effective.
Table 1. Influence of additive and additive/Ru molar ratioa
conversion (%)
equiv of
additive
9,
% ee
entry
additive
1 h
4 h
1
2
3
4
5
6
7
8
<1
4
<1
13
35
99.9
>99.9
>99.9
99.9
78
n.d.
36c
26
96
99
99
97
HClb
2
4
HClb
12
72
90
90
87
31
1
HClb
10
20
20
20
20
20
20
20
HClb
HClaq
LiCl
HBraq
HBF4
p-TsOH
Bu4NI
46
9
10
11
15
96
<1
33c
62
50
<1
n.d.
a 8b/additive, S/C 50 000, 70 bar of H2, 80 °C, MeOH (30 wt %, argon
distilled); 15 g scale. b Anhydrous methanolic HCl (prepared by addition of AcCl
to MeOH). c (S)-9 was formed.
(12) (a) Viscontini, M.; Merckling, N. HelV. Chim. Acta 1952, 282, 2280. (b)
Sotoguchi, T.; Yuasa, Y.; Tachikawa, A.; Harada, S. U.S. Patent 5945559,
1999.
(13) Chan, A. S. C.; Laneman, S. Inorg. Chim. Acta 1994, 223, 165.
(14) King, S. A.; Thompson, A. S.; King, A. O.; Verhoeven, T. R. J. Org. Chem.
1992, 57, 6689.
(15) (a) Noyori, R.; Ohkuma, T.; Kitamura, M.; Takaya, H.; Sayo, N.;
Kumobayashi, H.; Akutagawa, S. J. Am. Chem. Soc. 1987, 109, 5856. (b)
Noyori, R. Acta Chem. Scand. 1996, 50, 380.
(16) (a) King, S. A.; DiMichele, L. Chem. Industries 1995, 62, 157. (b) Ohkuma
T.; Noyori, R. CompehensiVe Asymmetric Catalysis I; Jacobsen, E. N., Pfalz,
A., Yamamoto, H., Eds.; 1999; p 207.
The influence of a variety of chiral diphosphines on the
hydrogenation rate and enantioselectivity in the catalysts of
type [Ru(OAc)2(diphosphine)]/20 HClaq was investigated
(Table 2). Very high enantioselectivities of >99% were
achieved only by applying atropisomeric diphosphines such
as (R)-BIPHEMP 7a, (R)-MeOBIPHEP 7b, (R)-3,5-t-Bu-
MeOBIPHEP 7c, and (R)-2-Furyl-MeOBIPHEP 7h.17 The
most active catalyst (98% conversion within 1 h at S/C
526
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Table 2. Influence of diphosphine liganda
conversion (%)
4 h
Scheme 5. Synthesis of (R)-bromodiester 11a
9,
% ee
entry
diphosphine
1 h
16 h
1
2
3
4
7a
7b
7c
7d
7e
7f
7g
7h
7i
7j
7k
7l
7m
7b
45
26
66
98
32
1
<1
1
8
7
2
9
<1
56
>99.9
94
>99.9
>99.9
99.7
2
<1
5
45
34
10
13
<1
99.9.
99
99
99
94
>99.9
5
>99.9
4
1
20
99.9
91
60
22
<1
>99.9
96
6
n.d.
n.d.
99
97
64
72
39
n.d.
96
7b
8
i.e., S/C 10 000, had to be employed to achieve complete
conversion. In contrast, at a hydrogen pressure of 70 bar the
hydrogenation rate increased, but the enantioselectivity
slightly decreased. Surprisingly, a substrate concentration of
40-60 wt % was necessary for both a high reaction rate
and a high enantioselectivity. Even at a concentration of 96
wt %, the hydrogenation reached 99% conversion within 16
h (S/C 50 000) and yielded 9 with 97% ee. On a large scale,
the hydrogenation was carried out preferentially at concen-
trations between 40 and 60 wt %. As expected, a further
increase of the reaction temperature brought about an increase
of the reaction rate but also a decrease of the enantioselec-
tivity. Finally, a process temperature of 80 °C was selected.
Under the best conditions, the hydrogenation of 200 g of
6 (2-L autoclave) ran to completion within 4 h, affording 9
with >99% purity and >99% ee. A total of greater than 2
tons of 9 in up to multi-100 kg batches have been produced
accordingly. All batches met set specifications (>99.9%
conversion, >99% ee).
9
10b
11
12
13c
14d
a [Ru(OAc)2(diphosphine)]/20 HClaq, S/C 50 000, 70 bar H2, 60 °C, MeOH
(30 wt %, argon distilled); 15 g scale. b S/C 5000. c [RuCl(p-cy-mene)(7m)]Cl/
18 HClaq used as catalyst (S/C 5000). d [RuCl(p-cy-mene)(7b)]Cl/18 HClaq used
as catalyst (S/C 50 000).
50 000) contained the very sterically demanding (R)-3,5-t-
Pe-MeOBIPHEP 7d as the chiral diphosphine ligand. How-
ever, with this catalyst the ee of 9 dropped to 94%. Due to
the large-scale accessibility of the parent diphosphine (R)-
MeOBIPHEP, 8b was selected as the pre-catalyst for the
asymmetric hydrogenation of 6. The intermediate 8b proved
to be remarkably air stable. Storage of 8b on the bench at
room temperature for 6 months caused neither decrease in
catalytic activity nor enantioselectivity at S/C 50 000 com-
pared with freshly prepared 8b or a sample of 8b, which
had been stored at room temperature under an argon
atmosphere for 6 months. The 31P NMR spectra of air-
exposed 8b showed the formation of a small amount of the
bis-oxide of 7b (bis-oxide of 7b/8b ∼1:10). Evidently, this
impurity had no influence on the formation of the catalyti-
cally active species. In contrast, the catalyst formed in
solution by treatment of 8b with HCl in methanol was highly
air sensitive: stirring this catalyst solution for 10 min in air
caused a total loss of activity. Therefore, total exclusion of
air during both the catalyst formation step and the hydro-
genation was crucial for high catalyst performance. Not-
withstanding, complete conversion and 99% ee were achieved
with an S/C ratio as high as 250 000 (40 bar H2, 22 h), no
matter if high-purity18 or technical-grade gases19 and solvents
were used, thus indicating the commercial potential and the
robustness of the asymmetric hydrogenation step. Finally,
in order to achieve a high throughput on a large scale, an
S/C ratio of 50 000 at 40 bar hydrogen pressure, which led
to complete conversion within ca. 4 h, was selected for scale-
up. The hydrogenation also proceeded selectively at a lower
hydrogen pressure (e.g., 5 bar), but a 5-fold amount of 8b,
b. Acylation of 9: The reaction of (R)-hydroxyester 9
with 2-bromooctanoyl halide 10 in toluene under modified
Schotten-Baumann20 conditions provided a 95% yield of
(R)-bromodiester 11a, which was a new, nonracemic inter-
mediate for the subsequent synthesis of Orlistat 1.
Initial laboratory preparations of 11a involved adding
pyridine in a 10 mol % excess to a hexane solution of 9 and
10 at 0 °C. This process was modified successfully to run
in toluene with the addition of 10 to 9 in the presence of
pyridine. However, use of pyridine on a large scale was
undesirable due to odor concerns. Early attempts to use
Schotten-Baumann conditions in which aqueous K2CO3 was
added last to a toluene solution of reactants, or to use solid
K2CO3 with a hexane solution of reactants, were unsuccess-
ful. Pyridine was successfully replaced with 4-dimethylami-
nopyridine (DMAP) in a 2-mol % excess. Attempts to reduce
the amount of DMAP using additional NaHCO3 or KHCO3
led to the finalized conditions employing 5-mol % DMAP
and 1.5 equiv of KHCO3.
The 2-bromooctanoyl bromide 10 was initially prepared
by conversion of octanoic acid to octanoyl chloride using
thionyl chloride, followed by bromination. Different bromi-
nation conditions produced varying amounts of 10a and 10b
ranging from 60/40 to 98/2 (Scheme 5). The influence of
the ratio 10a/10b was investigated in the next reaction of
the process: 10a/10b containing as much as 18 wt % of
10b has been used successfully, with less than 1% 11b being
formed. The level of 11b present in the subsequent,
(17) (a) Schmid, R.; Cereghetti, M.; Heiser, B.; Scho¨nholzer, P.; Hansen, H.-J.
HelV. Chim. Acta 1988, 71, 897. (b) Cereghetti, M.; Foricher, J.; Heiser,
B.; Schmid, R. Eur. Pat. Appl. EP 398132, 1990. (c) Foricher, J.; Schmid,
R. World Pat. Appl. WO 9315090, 1993. (d) Foricher, J.; Schmid, R. Eur.
Pat. Appl. EP 0926152, 1998. (e) Foricher, J.; Heiser, B.; Schmid, R. Eur.
Pat. Appl. EP 0530335, 1992.
(18) Nitrogen and hydrogen ex Carbagas 60 (g99.99990% purity).
(19) Nitrogen and hydrogen ex Carbagas 25 (99.5% purity).
(20) (a) Schotten, C. Ber. 1884, 17, 2544. (b) Baumann, E. Ber. 1886, 19, 3218.
(c) For a review see: Sonntag, N. O. V. Chem. ReV. 1953, 52, 237.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
527

Scheme 6. Intermediates and major products of the
magnesium-metal reaction to 12
of these derivatives brought about an improved selectivity
to 12. Several solvents were investigated as potential
alternatives to tetrahydrofuran. The use of toluene, diiso-
propyl ether, dimethylformamide, or 1,4-dioxane resulted in
no formation of 12. In ethers such as methyl tert-butyl ether,
di-n-butyl ether, 1-methyltetrahydrofuran, 1,2-dimethoxy-
ethane, and diethyl ether, 12 was formed, but yields were
lower than those in tetrahydrofuran. All of these screening
investigations resulted in lowering the yield of 12.
c.2. Activation of 11a by Magnesium Metal. The second
phase of exploration investigated more closely the magnesium-
metal parameters such as particle size,21 entraining reagents,22
and reagent additives.23 A too small magnesium particle size
(both powders and flakes) resulted in the formation of thick,
unmanageable slurries, and yields were not improved. In
contrast, coarse magnesium chips (approximately 1/8 to1/4 inch
turnings) were most convenient for large scale production
and allowed the reuse of any remaining magnesium metal
as a heel left in the reactor. Magnesium-metal reactions
typically require a surface activator or entrainment reagent
to maintain optimum reaction activity. Other entraining
reagents were investigated in an attempt to avoid the use of
the toxic 1,2-dibromoethane. The use of trace amounts of
cyclopentadiene, iodine, zinc chloride, or iron (III) bromide
all appeared to increase the rate of Grignard formation as
magnesium-mediated ring closure to 12 was a concern since
the presence of 11b could significantly decrease the reaction
rate and lead to a much reduced yield.
The commercially available 2-bromooctanoyl bromide
10a contained variable amounts of 10b typically ranging
from 2-4%. The commercial 10a was used on a large-scale
producing ca. 2 tons of 11a. The overall yield for the four
batch campaign was 100.9% with an assay of 95.0%
illustrating the excellent yields and robustness required for
a large-scale process.
1
confirmed in the H NMR analysis indicated by the loss of
the bromine after quenching the reaction mixture. However,
in no instance was a significant increase in the isolated yield
of 12 observed. Evidently, the key rate-limiting reaction of
this step was slow ring closure.
Various additives were investigated in an attempt to either
increase the cyclization rate or to minimize protonation of
the intermediate magnesium enolate 14. The reaction of 1,2-
dibromoethane with magnesium not only furnishes a clean
metallic surface on which 11a can react but also results in
the formation of ethylene and magnesium bromide. Since
experiments in which all the 1,2-dibromoethane was added
prior to the addition of 11a, which is a less-effective surface-
cleaning mode of addition, gave yields of approximately
48%, the presence of magnesium bromide was suspected to
be enhancing the desired cyclization. Experiments using
either magnesium chloride or magnesium iodide in the
absence of 1,2-dibromoethane produced yields comparable
to experiments performed in the presence of 1,2-dibromo-
ethane alone. However, the addition of numerous other Lewis
acids produced no significant yield increases beyond 50%.
Some of the studied Lewis acids were zinc bromide, zinc
chloride, magnesium chloride, magnesium bromide, mag-
nesium iodide, copper(I) bromide, and diethylaluminum
chloride. Addition of chlorotrimethylsilane or acetic anhy-
dride as a prequench to react preferentially with the methanol
c. Cyclization of 11: The initial approach was to carry
out the ring closure by treatment of a solution of (R)-
bromodiester 11a in tetrahydrofuran with magnesium metal
as shown simplistically in Scheme 6. Maintaining a clean
magnesium surface by the addition of 1,2-dibromoethane was
required to consistently achieve yields in the 40-50% range.
The major side product was the (R)-diester 15, which could
arise from protonation of the magnesium enolate 14 by either
12 or the methanol generated as a cyclization byproduct
(Scheme 6). The R-diester could also be generated by
equilibrium between the two possible enolates of 15 or from
an intermolecular equilibrium between the various esters and
their enolates. Although their yields were modest, these early
experiments clearly indicated the potential of a second-
generation process. The next stage of development work was
to expand efforts to increase the yield from the initial 40-
50% range.
c.1. Screening Studies. The first priority was to look at
the main aspects of the reaction, such as (a) the best metal
reagent; (b) the appropriate functional group(s) on 11 to
facilitate ring closure; and (c) potentially large solvent effects.
A quick screening of zinc, manganese, lithium, and magne-
sium showed that magnesium had the best potential. In an
attempt to increase the cyclization rate and enhance the yield
of 12, potential alternatives to the methyl ester function of
11a such as the corresponding morpholino amide, the N,O-
dimethylhydroxylamide, the nitrile, the acid chloride, and
the mixed pivaloyl anhydride were tested. However, none
(21) (a) Walborsky, H. M.; Topolski, M. J. Am. Chem. Soc. 1992, 114, 3455.
(b) Baker, K. V.; Brown, J. M.; Hughes, J.; Skarnulis, A. J.; Sexton, A. J.
Org. Chem. 1991, 56, 698. (c) Rieke, R. D.; Li, P. T.-J.; Burns, T. P.; Uhm,
S. T. J. Org. Chem. 1981, 46, 4323. (d) Hill, C. L.; Vander, S.; Aande, J.
B.; Whitesides, G. M. J. Org. Chem. 1980, 45, 1020.
(22) Garst, J. F.; Ungvary, F.; Batlaw, R.; Lawrence, K. E. J. Am. Chem. Soc.
1991, 113, 6697.
(23) (a) Guindon, Y.; Rancourt, J. J. Org. Chem. 1998, 63, 6554. (b) Cardillo,
G.; Gentelucci, L.; Tolomelli, A.; Tomasini, C. Tetrahedron 1999, 55, 6231.
528
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

reaction side product did not afford any yield benefits. The
same held true for the addition of amine bases such as
diisopropylethylamine and triethylamine. Apparently, these
hindered bases did not effectively prevent a labile proton
from causing side products. None of these additives aided
in the ring-closure reaction or in stopping the side-reaction
protonation to 15.
still not well understood.25 Recent reviews still indicate some
controversy on whether this pathway may be a radical-type
reaction (single electron transfer) or quenched by an available
labile proton (two electron transfer).26 The presence of a
competitive Grignard reagent, which could also react with
the acidic protons in 11a, a transient intermediate, 12, or
the methanol byproduct, might possibly decrease production
of 15. A study of several simple Grignard reagents (methyl,
isopropyl, sec-butyl, and tert-butyl) did result in reduction
of 15 as a side product and with optimized conditions
improved yields of 12 up to the 60-75% range. Since the
reaction of 11a occurred rapidly with magnesium metal even
at low temperatures, an experiment was performed without
the magnesium metal. Interestingly, the reaction produced
the same yields when using 2 to 3 equiv of the Grignard
reagent. This development removed the complexities of
maintaining a clean metal surface with the toxic 1,2-
dibromoethane. Analyses of periodic, quenched-reaction
samples suggested that although all of the 11a was being
immediately consumed through rapid metal-halogen ex-
change, the cyclization product was being formed at a much
slower rate. Although the trans-metalation from tert-butyl
magnesium chloride to 14 was rapid even at -78 °C,
acceptable cyclization rates of the intermediate enolate 14
still required temperatures of approximately 60 °C. The tert-
butyl Grignard reagent gave superior product yields over
longer times than other comparable Grignard reagents due
to less decomposition once the reaction approached comple-
tion. The proposed mechanism for the reaction of tert-butyl
magnesium chloride with 11a is shown in Scheme 7.
The apparent lower propensity of tert-butyl magnesium
chloride to form side products may be due to its steric bulk,
which decreases its nucleophilicity and its propensity to react
with the tert-butyl bromide byproduct. Since solutions of tert-
butyl magnesium chloride in tetrahydrofuran are com-
mercially available or can readily be prepared from tert-butyl
chloride and magnesium-metal, this Grignard reagent was
chosen for further development. This new variation does not
fit into either of the traditional zinc Reformatsky27 or
magnesium Barbier28 reactions. The utilization of this
different halogen-magnesium exchange reaction seems to
be part of a growing field.29
c.3. Fine-Tuning of the Magnesium-Metal Process
Parameters. Thus far, investigations had failed to increase
the yield of 12. By timed quench studies, the formation of
the Grignard 14 from 11 was observed to be faster than the
ring closure to 12. Decreasing the production of 15 should
result in an improved yield of 12. A closer study of the
reaction parameters, such as temperature, stoichiometry, order
of addition and workup procedures, was initiated. A study
of various reaction temperatures indicated that cyclization
rates were optimal at 60 °C. Since a reasonable alternative
to 1,2-dibromoethane was not found, a study was done to
minimize the 1,2-dibromoethane amounts required for Grig-
nard initiation and formation of the magnesium enolate 14.
It was found that once a magnesium heel had been formed,
whether it was initiated by iodine or 1,2-dibromoethane, the
reaction proceeded reasonably well. 1,2-Dibromoethane (1
equiv) added with the 11a solution gave the best results with
isolated 12 yields of about 50%. A 0.5 equiv amount of 1,2-
dibromoethane resulted in an approximately 5-10% lower
yield. No 1,2-dibromoethane gave 10-20% lower yield. It
was observed that 11b reacted at a much slower rate in the
Grignard reaction than 11a. The use of less than 1 molar
equiv of 1,2-dibromoethane and 15-20 mol % 11b (relative
to 11a) resulted in the yield decreasing from 45-50% to
approximately 35%. However, the yield in the presence of
15-20% 11b could be increased to 45-50% by using 1
molar equiv of 1,2-dibromoethane. Although the mechanistic
reason for this yield dependency on the 1,2-dibromoethane
charge is unclear, the additional 1,2-dibromoethane might
be causing the conversion of 11b to 11a and, therefore, a
10b limit of <2% in 10a was set to have a consistent
parameter. Although these parameter changes did not in-
crease yields, close observations gave a hint as to how to
increase the yield above 50%.
c.4. Activation of 11a by Grignard Reagents. The key
hint on improving 12 yields was that somehow the magne-
sium-metal was affecting the reaction. The 11a substrate was
complicating the situation due to having other enolizable
carbonyls that could complex with the magnesium metal and
having easily detached protons that were available for
intramolecular reduction of the 14 intermediate to 15.
Fortunately, no double addition side products were observed.
Magnesium-metal insertion reactions into carbon-halogen
bonds were first reported by Victor Grignard around 1900.24
The reduction of a carbon-magnesium-bromide bond to a
carbon-hydrogen bond and the double addition of a Grig-
nard reagent to an ester are common problems with
magnesium-metal reactions. Although the history of Grignard
reactions is long, the mechanism of Grignard reactions is
The commercial solutions of ca. 1 M tert-butyl magne-
sium chloride showed a very low propensity to form a
precipitate upon storage at 20-25 °C. Since the proposed
cyclization mechanism suggested that 3 molar equiv of
Grignard reagent were required, a series of experiments was
performed by the slow addition of 11a to 2.5-5.0 equiv of
tert-butyl magnesium chloride at 60 °C. With the slow
addition no exothermic reaction was observed, even on scale-
(25) Handbook of Grignard Reagents; Silverman, G. S., Rakita, P. E., Eds.;
Marcel Dekker, Inc.; New York, 1996.
(26) (a) Root, K. S.; Hill, C. L.; Lawrence, L. M.; Whitesides, G. M. J. Am.
Chem. Soc. 1989, 111, 5405. (b) Garst, J. F.; Soriaga, M. P. Coord. Chem.
ReV. 2004, 248, 623.
(27) Reformatsky, S. Ber. 1887, 20, 1210.
(28) Barbier, P. C. R. Acad. Sci. 1899, 128, 110.
(29) (a) Ren, H.; Knochel, P. Chem. Commun. 2006, 726. (b) Abarbri, M.;
Thibonnet, J.; Berillon, L.; Dehmel, F.; Rottlaender, M.; Knochel, P. J. Org.
Chem. 2000, 65, 4618.
(24) Grignard, V. Compt. Rend. 1900, 130, 1322.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
529

Scheme 7. Simplified reaction pathway for the
halogen-magnesium exchange reaction to 12
Table 4. Quench study
wt % yield of 12 in solution after
entry acid equiv pH
0 h
3 h
22 h
48 h
1
2
3
4
HCl
3.2
3.1
2.1
1.5
0
1.3
7
88
86
87
89
89
90
90
89
88
91
87
75
86
91
citric
citric
citric
8
67
produced a residue, which was then slurried in hexane. The
product was isolated after filtering and drying.
The workup procedure required both an extraction solvent
system and a crystallization solvent system. The presence
of magnesium salts in the reaction mixture necessitated an
aqueous-acid workup and, consequently, displacement of
tetrahydrofuran with a relatively water-immiscible extraction
solvent. A crystallization-solvent system which did not
involve distillation of the product to dryness was also needed.
Ideally, a single-solvent system could satisfy all requirements.
A number of potential extraction solvents were studied,
including ethyl acetate, isopropyl acetate, methylene chloride,
methyl tert-butyl ether, and toluene. All of these solvents
worked fairly well with the main differences being the
amount of solvent required to completely dissolve 12. Since
methyl tert-butyl ether and heptane were utilized in the first-
generation process, these two solvents were ultimately
selected as the extraction and crystallization solvents,
respectively.
Table 3. Influence of t-BuMgCl/11a molar ratio
t-BuMgCl/11a
isolated
yield
(%)
entry
(mol/mol)
Since the instability of 12 to either strongly acidic or
strongly basic conditions was known, citric acid, which is a
relatively weak acid, was investigated as an alternative to
hydrochloric acid. In a laboratory stability study, a methyl
tert-butyl ether/tetrahydrofuran slurry of a crude reaction
mixture was transferred into various quench mixtures as
shown in Table 4. When 3.1-3.2 equiv of acid were utilized,
the 48 h samples indicated that approximately a 5% extra
yield loss was suffered in the hydrochloric acid experiment
relative to citric acid. As expected, it was found that
quenching with only 1.5 equiv of citric acid resulted in
significant degradation of 12 within 22 h, while using 2.1
equiv of citric acid resulted in stable assays of 12 out to at
least 22 h. The final use of 3 equiv of citric acid (relative to
11a) represented an attempt to build robustness into the
process by keeping the pH slightly on the acidic side. A
second citric acid wash (0.5 equiv) was found to significantly
decrease the amount of magnesium salts remaining in the
organic phase.
1
2
3
4
5
6
2.50
2.75
3.00
3.25
3.50
5.00
66.4
74.0
78.4
74.5
72.9
72.6
up. The results (Table 3) show that the yield was fairly
constant when a range of 2.75-5.0 equiv of Grignard reagent
was used.
The final major yield-enhancing breakthrough was the
discovery that simultaneous addition of tert-butyl magnesium
chloride and a solution of 11a in tetrahydrofuran to a small
amount of tetrahydrofuran at 60 °C produced yields of
approximately 85%. It was unclear why this mode of addition
was so beneficial, but apparently, this addition mode further
decreased undesired side reactions due to excess Grignard
reagent. The reverse addition of tert-butyl magnesium
chloride to 11a minimized excess Grignard reagent but
resulted in a yield decrease of 5-10%.
Since the hydrogenation of 12 to 13 in the first-generation
process had been shown to be highly sensitive to salts
entrained in 12, numerous water washes were required to
decrease these salts to an acceptable level. The citric acid
workup typically required five water washes to decrease the
conductivity of the last aqueous phase to <200 µ siemens,
which was a level that had been shown to ensure acceptable
hydrogenation rates. As the water washing progressed, the
pH of the aqueous washes gradually increased from ap-
proximately 2.5 to approximately 3.5.
c.5. Workup of the Cyclization Mixture. Concurrent
with alterations in the cyclization step itself, the subsequent
workup procedure underwent numerous modifications through-
out both the research and development phases. The initial
workup of the Grignard process involved decanting the
product-containing tetrahydrofuran solution away from the
magnesium-metal heel, distilling off the tetrahydrofuran,
dissolving the residue in ethyl acetate, and washing the
organic phase with hydrochloric acid followed by water.
Finally, removal under a vacuum of all the ethyl acetate
530
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

Several experiments, which utilized various batches of
both tert-butyl Grignard reagent and 11a, were performed
under the final conditions. Data from these experiments
confirmed that the process was capable of producing 80-
88% yields of 12 having 99-100% purity. A large-scale
production campaign produced ca. 760 kg of 12. The overall
average large-scale yield from 11a was 83.7% with a purity
of >99.5%.
resulting 12 was enantiomerically pure and not a racemate
(as in 2). This improvement removed the late stage resolution
required at 4 in the original synthesis and allowed doubling
the throughput in the later steps. In summary, the goal set at
the beginning of the work, i.e., to create a significantly more
efficient process by obviating the late-stage resolution present
in the first-generation process, was achieved using inexpen-
sive commercially available starting materials.
Synthesis of (3S,4S,6R)-δ-Lactone 13. In the second
reaction sequence, no major changes from the first-generation
process were needed to diastereoselectively reduce enantio-
merically pure 12 to enantiomerically pure 13.3u A large-
scale production campaign using the Raney Nickel catalyzed
heterogeneous hydrogenation of 12 afforded ca. 360 kg of
13 in 86% yield with an average purity >98%. No other
diastereomer was observed above 1.5%.
Synthesis of PEA-Salt 4. The subsequent third reaction
sequence was also carried out in analogy to the first-
generation process.3u Hydroxy-δ-lactone 13 was first THP-
protected, followed by saponification of the δ-lactone,
selective benzylation of the δ-hydroxy group, and deprotec-
tion of the THP-ether. After treatment with (S)-phenylethy-
lamine, the enantiomerically pure PEA-salt 4 was isolated
in double the yield of 4 to that produced by the first-
generation optical resolution based process. Isolation of 4
via crystallization remained necessary to meet the previously
set purity specifications and to ensure a substrate quality
suitable for further transformation into 1. A large-scale
production of the PEA-salt 4 produced over 250 kg in 88%
yield.
Final Steps to Orlistat. There were virtually no differ-
ences in the last reaction sequence between the first- and
second-generation Orlistat processes for the conversion of
4 to 1.3u Benzenesulfonyl chloride promoted â-lactonization
of 4 and subsequent debenzylation afforded the (S,S,R)-
oxetanone 5 in 85% yield (Step 4). In the last step (Step 5),
the N-formyl-L-leucine building block was introduced using
Mitsunobu conditions to yield Orlistat 1 on a multi-10 kg
scale in 80% yield. The second-generation process produced
1 that was equivalent and met all first-generation specifica-
tions.
Experimental Section
General. All reagents and solvents, if not otherwise noted,
were obtained from commercial suppliers and used without
further purification. All gases used in the large-scale produc-
tion runs were of technical grade (99.5% purity). The
diphosphines and ruthenium complexes were either pur-
chased from Strem (7e,g,j-m) or prepared according to or
in analogy to, respectively, the literature (7a-d,f,h,17i,30 8a-
m).4,13 Melting points were determined on a Bu¨chi 510 and
are uncorrected. 1H NMR spectra (at 250 MHz on a Bruker
AC 250E spectrometer or at 300 MHz on a General Electric
QE300 spectrometer) were recorded by using TMS as an
internal standard. Chemical shifts are given in ppm (δ), and
J values, in hertz. ESI-MS spectra were measured on a
Perkin-Elmer Sciex API III. Elemental analyses were carried
out on a Carlo Erba 1160 elemental analyzer.
For the determination of conversion and chemical purity
of 9 (both enantiomers) by GC (column: DB-1, injector:
280 °C, detector: 330 °C, oven: 50-320 °C/5 °C per min,
carrier gas: He (70 kPa)); a sample of the methanolic
reaction mixture was evaporated to dryness, 2 mg of the
residue were dissolved in 500 µL of pyridine and 500 µL of
BSTFA/1% TMCS, and the mixture was stirred for 1 h at
room temperature. GC retention times: 22.8 min, 8 (both
enantiomers); 23.5 and 23.7 min 6 (i.e., cis and trans silylenol
ethers of 6).
For enantiomeric purity determination of 9 by GC
(column: Optima-240, injector: 280 °C, detector: 300 °C,
oven: 200-290 °C/3 °C per min, carrier gas: He (80 kPa));
a 2 mg sample of the evaporated residue from above was
dissolved in 1 mL of dichloromethane, and then 10 mg of
(S)-Trolox-methyl ether, 15 mg of 1,3-dicyclohexylcarbo-
diimide, and 0.1 g of 4-dimethylaminopyridine were added
to the solution. The resulting mixture was stirred for 30 min
at 60 °C and filtered (Acrodisc PFTE 0.45 µm). GC retention
times: 26.8 min, (S)-enantiomer of 9; 27.6 min, 9.
Methyl (R)-3-Hydroxytetradecanoate (9): A 200-mL
stainless steel addition device which was connected to a 2-L
autoclave was charged in the air with 1 N aqueous hydrogen
chloride (0.62 mL, 0.62 mmol) and methanol (100 mL). The
addition device was sealed and purged with nitrogen.
Afterwards, the autoclave was charged in the air with 8b
(27.9 mg, 31.2 × 10-6 mol, S/C 5 × 104), 6 (400 g, 1.56
mol), and methanol (658 mL). The autoclave was sealed and
purged with nitrogen. The bottom valve of the addition
device was opened, and the aqueous-methanolic solution of
hydrogen chloride was introduced into the autoclave. The
Conclusion
The comparison of first- and second-generation Orlistat
processes shows significant improvements for the new
synthetic route. The overall yield of the enantioselective,
second-generation process for 1 from 6 amounts to 41-48%,
which compares very favorably with the 18-22% achieved
with the first-generation, optical resolution based process of
4. Particularly, the herein described hydrogenation process
proved to be highly robust, such that the large-scale
production of â-hydroxyester 9 with very high ee (>99%)
and purity (>99%) by asymmetric hydrogenation of â-ke-
toester 6 was straightforward and practical. The acylation
of 9 and subsequent cyclization of 11 to the single enantiomer
12 proved to be a robust and high yielding process also. The
new enantioselective synthesis utilized the same chemistry
for the later steps as the original process. The key aspect of
the improved enantioselective synthesis was that now the
(30) Laue, C.; Schro¨der, G.; Arlt, D.; Grosser, R. Eur. Pat. Appl. EP 0643065,
1994.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
531

reaction mixture was stirred for 30 min at 20 °C. Nitrogen
was then replaced by hydrogen (30 bar), and the mixture
was heated to 80 °C. After 25 min at a reaction temperature
of 80 °C, the pressure had reached 31 bar. The hydrogen
pressure was set to a constant 40 bar. After 4 h, the conver-
sion was complete (>99.9%, 99.0% ee), and the autoclave
was cooled to 20 °C. The hydrogen pressure was released,
and the autoclave was purged with nitrogen. The slightly
yellow solution was transferred to a 2-L Erlenmeyer flask,
and the autoclave was washed with methanol (300 mL). The
methanolic solutions were combined. If desired, the methanol
solution of 9 can be utilized directly in the next step.
Rotatory evaporation of the combined methanolic solution
to dryness at 50 °C/30 mbar gave crude 9 (397.1 g) as a
white, waxy solid. Crude 9 (10.9 g) was dissolved in hexane
(52 mL) at 30 °C and then cooled in an ice bath. At 15 °C,
a few seed crystals of pure 9 were added to promote
crystallization. The resulting suspension was stirred for 4 h
at 0-5 °C and then stored overnight at -15 °C. The white
precipitate was filtered off, washed with ice-cold hexane (20
mL), and dried under a vacuum (0.1 bar) for 6 h at room
temperature to yield pure 9 (10.2 g, 92% yield) as white
crystals with 99.7% purity (GC %AN) and 99.7% ee. Mp
40 °C. MS (EI): 259 (M + H+). 1H NMR (250 MHz, CDCl3)
δ 4.10-3.90 (m, 1H), 3.71 (s, 3H), 2.84 (d, 1H, J ) 2.5
Hz), 2.52 (dd, 1H, J ) 10.3, 1.7 Hz), 2.41 (dd, 1H, J )
10.3, 5.5 Hz), 1.6-1.2 (m, 20H), 0.88 (t, 3H, J ) 4.2 Hz).
Anal. calcd for C15H30O3: C, 69.72; H, 11.70. Found: C,
69.57; H, 11.79. Ru content (X-ray fluorescence): 6 ppm.
Alternatively, a 2000-L autoclave was charged with a
methanolic solution of 6 and the pre-catalyst 8b (S/C 50 000).
After the autoclave was purged with technical grade nitrogen,
1 N HCl (20 equiv relative to 8b) in methanol was added
through a 3-L autoclave-addition device previously purged
with technical grade nitrogen to form the air-sensitive
catalyst. Nitrogen was replaced by technical grade hydrogen
(40 bar), and the temperature was set to 40 °C. After 4-16
h the conversion was complete, and the methanolic solution
of 9 was used directly in the next step.
(R)-3-[(2-Bromo-1-oxo-octyl)oxy]-tetradecanoic Acid
Methyl Ester (11a). (R)-hydroxyester 9 (351.1 g, 547.6
mmol, 40.3 wt % solution in methanol) was added to a 1-L,
three-neck, jacketed, round-bottom flask, which was equipped
with a distillation condenser, an addition funnel, a nitrogen
inlet, a thermocouple, and a mechanical stirrer. Most of the
methanol was removed by vacuum distillation. Toluene (360
mL) was added to the residue, and solvents were again
removed by vacuum distillation. 4-Dimethylaminopyridine
(3.3 g, 27 mmol), KHCO3 (78.75 g, 786.5 mmol), water (40
mL), and toluene (25 mL) were added to the reaction mixture.
The mixture was stirred and cooled to 11 °C. 2-Bromooc-
tanoyl bromide (173.4 g, 602 mmol, 96.4 wt % 2-bromooc-
tanoyl bromide, 2.5 wt % 2-bromooctanoyl chloride) was
added to the stirred, reaction mixture at approximately 100
mL/h over 1.5 h. The mixture was stirred for 2 h at 10 °C
and then sampled for reaction completion (0.07 wt % 9).
Water (108 mL) was added, and the mixture was stirred for
15 min at 10 °C. The mixture was allowed to settle, and the
phases were separated. The organic phase was washed with
water (2 × 108 mL, 10 °C). Solvents and residual water
were removed by vacuum distillation and gave 11a as an
orange oil (261.8 g, 103% yield, typically 4-5 wt % toluene).
1H NMR (300 MHz, CDCl3) δ 5.28 (dt, 1H, J ) 6.4 Hz),
4.17 (dd, 1H, J ) 7.5 Hz, J ) 7.5 Hz), 3.67 (s, 3H), 2.60
(m, 2H), 2.00 (m, 2H), 1.65 (m, 2H), 1.25 (m, 26H), 0.88
(m, 6H); 13C NMR (75 MHz, CDCl3) δ 170.60, 170.56,
169.26, 169.15, 72.15, 72.10, 51.91, 51.83, 46.18, 46.12,
46.05, 38.92, 38.87, 35.12, 34.96, 34.92, 34.17, 33.95, 32.01,
31.62, 31.45, 29.96, 29.72, 29.60, 29.54, 29.45, 29.36, 29.14,
28.82, 28.64, 27.29, 25.08, 25.00, 22.79, 22.60, 14.20, 14.09.
(R)-3-Hexyl-5,6-dihydro-4-hydroxy-6-undecyl-2H-py-
ran-2-one (12). Tetrahydrofuran (60 mL) was added to a
2-L, three-neck, jacketed, round-bottom Morton flask, which
was equipped with a distillation condenser, two syringe
pumps, a nitrogen inlet, a thermocouple, and a mechanical
stirrer. The mixture was heated to 60 °C. (R)-Bromo diester
11a (89.7 g, 95.0 wt %, 184 mmol) was placed into one of
the syringe pumps. tert-Butyl magnesium chloride/tetrahy-
drofuran (685 mL, 1.02 M total base, 0.94 M active base,
645 mmol) was placed into the other syringe pump. Both
solutions were added simultaneously over a 5-h addition
period (addition rates of 17.2 mL/h and 137 mL/h, respec-
tively). Once the additions were complete, the reaction
mixture was held at 60 °C for 2 h, cooled to 20 °C over 3
h, and held at 20 °C for 12.5 h. The reaction mixture was
sampled for reaction completion (glc analysis). The reaction
mixture was concentrated by vacuum distillation to ca. 400
mL (ca. 50% of the original volume). tert-Butyl methyl ether
(275 mL) was added to the mixture to produce a fluid, cream-
colored suspension. Tetrahydrofuran (17.5 mL) was added
to the suspension. Another 2-L, three-neck, jacketed, round-
bottom Morton flask, which was equipped with a distillation
condenser, a nitrogen inlet, a thermocouple, and a mechanical
stirrer, was used to prepare a solution of citric acid (43.32
g) and water (142 mL). The cream-colored suspension was
slowly pumped into the stirred, aqueous citric acid solution
over 60 min via Teflon tubing. The suspension flask was
rinsed through the tubing into the quench vessel with 17.5
mL of tetrahydrofuran. The resulting reaction mixture was
heated to 30 °C, and 75 mL of tert-butyl methyl ether were
added. Stirring was continued for ca. 2 h. After the mixture
was allowed to settle, the lower aqueous phase was drained
from the upper product-containing organic layer. A solution
of citric acid (6.75 g) and water (111 mL) was added to the
remaining organic layer. The mixture was stirred and allowed
to settle, and the aqueous layer was separated from the
product-containing organic phase. The remaining organic
layer was washed with water (5 × 88 mL). The mixture was
concentrated by vacuum distillation to ca. 400 mL. n-Heptane
(460 mL) was added over ca. 1 h resulting in a very thin
suspension. The mixture was concentrated again by vacuum
distillation to ca. 400 mL, and n-heptane (184 mL) was added
slowly. The suspension was stirred for 2 h at 25 °C, cooled
to 5 °C over 1 h, and then held at 5 °C for 13.5 h. The
slurry was filtered, rinsed with cold n-heptane (184 mL),
and dried. The product 12 was isolated as a white solid (56.12
532
•
Vol. 11, No. 3, 2007 / Organic Process Research & Development

1
g, 86.5% yield, 99.99% purity (GC wt/wt)). H NMR (300
W. Walther, S. Brogly, and C. Tournoux for GC analyses;
D. Spiess and M. Ha¨ss for skillful technical assistance; and
Drs. M. Karpf, U. Zutter, R. Schmid, and R. Birk for helpful
discussions. We wish to thank our colleagues from Roche
Colorado Corporation for the analytical support, especially
J. Lay, I. Nuiry, and Dr. L. Bryant. We wish to thank our
colleagues from Roche Colorado Corporation for their
contributions on other parts of the project, Dr. G. Semones,
Dr. P. Harrington, Dr. G. Withers, Dr. M. Ronn, C. Evanson,
Dr. R. Nalitham, Dr. J. Brown, Dr. H. Moorlag, H.
Hafezzadeh, Dr. R. Dauer, Dr. H. Khatri, Dr. B. DeHoff,
Dr. M. Kopach, Dr. B. Hughes, Dr. G. Rowe, Dr. F. Todd,
Dr. T. Foderaro, and B. Smallwood. We wish to thank our
colleagues from Roche Ireland Limited for running the final
API step on the pilot-plant scale for us.
MHz, CDCl3) δ (keto form) 4.68 (m, 1H), 3.41 (t, 1H, J )
5.6 Hz), 2.72 (dd, 1H, J ) 3.0 Hz, J ) 15.9 Hz), 2.42 (dd,
1H, J ) 11.5 Hz, J ) 18.8 Hz), 2.00-1.49 (m, 6H), 1.49-
1.20 (m, 24H), 0.88 (m, 6H); (enol form) 8.07 (s, 1H), 4.32
(m, 1H), 2.51 (m, 2H), 2.27 (m, 2H), 1.75 (m, 1H), 1.61
(m, 1H), 1.52-1.15 (m, 26H), 0.88 (m, 6H).
Typically, the magnesium route generated a 1:1 mixture
of the enol and keto forms. The enol form slowly equilibrated
to the keto form in CDCl3, which contained a trace of
hydrogen chloride. The keto-form NMR spectrum was
identical to that found in the literature.31 The ratio of the
keto/enol forms was solvent dependent. The NMR data for
the enol-form was obtained using a malonate synthetic route.6
Acknowledgment
We wish to thank our colleagues from F. Hoffmann-La
Roche Ltd Basel for the analytical support, especially Dr.
Received for review October 9, 2006.
OP060208Q
(31) Fehr, M. J.; Consiglio, G.; Scalone, M.; Schmid, R. J. Org. Chem. 1999,
64, 5768.
Vol. 11, No. 3, 2007 / Organic Process Research & Development
•
533