Process development and scale up of a glycine antagonist

Article abstract of DOI:10.1021/op9001824

A synthetic route amenable to large-scale synthesis of the glycine antagonist (2R,4E)-7-chloro-4-(2-oxo-1-phenyl-pyyrrolidin-3-ylidene)- 1,2,3,4-tetrahydroquinoline-2-carboxylic acid, (2R,3R,4R,5S)-6-(methylamino) hexane-1,2,3,4,5-penta-ol 12 is presented

Full text of DOI:10.1021/op9001824

Organic Process Research & Development 2009, 13, 1130–1140
Process Development and Scale Up of a Glycine Antagonist
Adam Banks,† Gary F. Breen,† Darren Caine,‡ John S. Carey,*,† Christopher Drake,‡ Michael A. Forth,† Asa Gladwin,‡
Simone Guelfi,§ Jerome F. Hayes,† Paolo Maragni,§ David O. Morgan,† Paul Oxley,† Alcide Perboni,§ Matthew E. Popkin,†
Fiona Rawlinson,† and Guillaume Roux†
Chemical DeVelopment, GlaxoSmithKline Pharmaceuticals, Old Powder Mills, Leigh, nr. Tonbridge, Kent TN11 9AN, United
Kingdom, Chemical DeVelopment, GlaxoSmithKline Pharmaceuticals, Gunnels Wood Road, SteVenage, Hertfordshire SG1 2NY,
United Kingdom, and Chemical DeVelopment, GlaxoSmithKline Pharmaceuticals, Via Fleming 4, 37135 Verona, Italy
Abstract:
Results
The discovery route to compound 12 is shown (Scheme 1).2
A synthetic route amenable to large-scale synthesis of the glycine
antagonist (2R,4E)-7-chloro-4-(2-oxo-1-phenyl-pyyrrolidin-3-ylidene)-
1,2,3,4-tetrahydroquinoline-2-carboxylic acid, (2R,3R,4R,5S)-6-
(methylamino)hexane-1,2,3,4,5-penta-ol 12 is presented. The route
consists of four stages of chemistry. Stage 1 starts from 5-chloro-
2-iodoaniline hydrochloride and is a three-step telescoped stage
consisting of an imine formation with ethyl glyoxalate, Mannich
reaction using vinyloxytrimethylsilane, and subsequent Wittig
reaction with (2-oxo-1-phenyl-3-pyrrolidinyl)triphenylphospho-
nium bromide. The stage 1 product (4E)-2[(5-chloro-2-iodophe-
nyl)amino]-4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)butanoic acid
ethyl ester 17 is subjected to an enzyme-catalysed kinetic resolution
to prepare the single (2R)-enantiomer 19 as the ethyl ester. Stage
3 is the intramolecular Heck reaction to yield (2R,4E)-7-chloro-
4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)-1,2,3,4-tetrahydroquinoline-
2-carboxylic acid ethyl ester 31. The final stage is ester saponifi-
cation and meglumine salt formation to afford the drug candidate
molecule 12. In total, more than 300 kg of target 12 was produced
with a purity >99.9%. Aspects of route selection as well as
elements of process understanding and control are discussed.
The route starts with tert-butyl lactate 1 which is acylated with
acryloyl chloride to give 2. Oxidative cleavage of the alkene
gave the glyoxalate ester derivative 3. The imine 5 was formed
by condensation with 5-chloro-2-iodoaniline 4.3,4 A highly
diasteroselective Lewis acid-promoted addition of allyltributyltin
gave compound 6 as single diastereoisomer. A second oxidative
alkene cleavage gave aldehyde 8. An alternative sequence was
to treat the imine 5 with vinyloxytrimethylsilane 7 in the
presence of a catalytic quantity of Lewis acid to give directly
the aldehyde 8, but this time as an 85:15 mixture of diastere-
oisomers at the newly created chiral centre. Aldehyde 8
underwent a highly E-selective Wittig reaction upon reaction
with (2-oxo-1-phenyl-3-pyrrolidinyl)triphenylphosphonium bro-
mide 95,6 to give alkene 10. Intramolecular Heck cyclisation of
alkene 10 gave the tetrahydroquinoline 11 with the exocyclic
E double bond as the major product. The final step of the
synthesis was hydrolysis of the chiral auxiliary to give the target
carboxylic acid 12. The overall sequence from tert-butyl lactate
1 to tetrahydroquinoline 12 using vinyloxytrimethylsilane 7 as
the nucleophile was seven chemical steps and an overall yield
of 26%. Initially the acid was isolated as the sodium salt.
However, this salt did not possess the required pharmaceutical
development attributes as it was hygroscopic and lacked
sufficient photostability. During the development process the
counterion was switched to the N-methyl-D-glucamine salt (also
known as meglumine), which overcame these shortcomings.
Although the discovery routes had been used by the
discovery scientists to prepare tens of grams of tetrahydro-
quinoline 12, the initial drug substance requirements for
chemical development scientists was several kilograms, so an
assessment of the suitability of the route to provide these inflated
Introduction
The chiral tetrahydroquinoline carboxylic acid 12 is an orally
bioavailable glycine antagonist and has been identified as a
potential drug candidate molecule for the treatment of nicotine
craving.1 Gram quantities of the compound were initially
prepared by the discovery scientists, whereupon the scientists
within Chemical Development were charged with preparing
kilogram quantities of the compound to support drug safety
assessment, pharmaceutical development, and clinical trials
activities.
In this contribution we describe the following: (1) Our
assessment of the discovery synthesis. (2) Some route evaluation
work and our rationale around route selection. (3) The route of
synthesis used to deliver early development supplies. (4) The
route of synthesis used to deliver >300 kg of the drug candidate
molecule along with aspects of process understanding and
process control.
(2) (a) Di Fabio, R.; Alvaro, G.; Bertani, B.; Donati, D.; Giacobbe, S.;
Marchioro, C.; Palma, C.; Lynn, S. M. J. Org. Chem. 2002, 67, 7319–
7328. (b) Di Fabio, R.; Alvaro, G.; Bertani, B.; Donati, D.; Pizzi,
D. M.; Gentile, G.; Pentassuglia, G.; Giacobbe, S.; Spada, S.; Ratti,
E.; Corsi, M.; Quartaroli, M.; Barnaby, R. J.; Vitulli, G. Bioorg. Med.
Chem. Lett. 2007, 17, 1176–1180.
(3) (a) Di Fabio, R.; Alvaro, G.; Bertani, B.; Giacobbe, S. Can. J. Chem.
2000, 78, 809–815. (b) Van de Lande, L. M. F. Recl. TraV. Chim.
Pay Bas. 1932, 51, 98–113.
* Author to whom correspondence may be sent. E-mail: john.s.carey@
gsk.com.
(4) Large scale quantities of 5-chloro-2-iodoaniline hydrochloride were
obtained from third party suppliers and supplied as the hydrochloride
salt.
† GlaxoSmithKline Pharmaceuticals, Tonbridge, U.K.
‡ GlaxoSmithKline Pharmaceuticals, Stevenage, U.K.
§ GlaxoSmithKline Pharmaceuticals, Verona, Italy.
(1) (a) Di Fabio, R. WO 99/64411. (b) Orlandi, A. WO 01/42238. (c)
Chiamulera, C.; Reggiani, A.; Trist, D. G.; Teneggi, V. WO 2005/
053693.
(5) (a) Dolan, S. C.; Perboni, A.; Maragni, P. WO 98/39341. (b) Ikuta,
H.; Shirota, H.; Kobayashi, S.; Yamagishi, Y.; Yamada, K.; Yamatsu,
I.; Katayama, K. J. Med. Chem. 1987, 30, 1995–1998.
(6) Large scale quantities of (2-oxo-1-phenyl-3-pyrrolidinyl)triphenylphos-
phonium bromide were obtained from third party suppliers.
1130
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development
10.1021/op9001824 CCC: $40.75  2009 American Chemical Society
Published on Web 09/23/2009

Scheme 1a
a Reagents and conditions: i) Acryloyl chloride, NEt3, DMAP, CH2Cl2, 88%. ii) OsO4, NaIO4, THF, H2O, 100%. iii) 4, MgSO4, Tol, ∆, 100%. iv) allyltributyltin,
SnCl4, CH2Cl2, -78 °C, 77%. v) O3, PPh3, CH2Cl2, -78 °C, 55%. vi) 7, Yb(OTf)3, CH2Cl2 -20 °C, 55%. vii) 9, DBU, acetonitrile, -20 °C, 70%. viii) Pd(PPh3)4,
NEt3, Tol, ∆, 85%. ix) LiOH, THF, H2O then HCl, 90%.
Scheme 2a
a Reagents and conditions: i) 14, LiCl, DBU, acetonitrile, 50%, 15:16 ) 5:1.
amounts was undertaken.7 We had serious concerns about the
use of the osmium and tin reagents, as they would present
significant problems with respect to both worker and patient
safety as well as having a serious negative environmental
impact. The overall cost of goods for this route was high; one
of the main contributors was the chiral auxiliary (R)-tert-butyl
lactate 1. This reagent was introduced at the beginning of the
synthesis and could not be effectively recycled as it was not
cleaved until the end of the synthesis. There were few solid
crystalline intermediates in this route. Isolation of solid inter-
mediates by crystallization helps to define yields but more
importantly expels impurities, which is important from a quality
by design perspective. From a throughput perspective the route
is relatively long, seven steps, and many of the intermediates
required purification by chromatography, which is both a time-
and solvent-consuming unit operation.
As an alternative to the Heck reaction to form the tetrasub-
stituted alkene, we briefly explored some olefination strategies
such as the use of Wittig reagents, e.g. 9 or the corresponding
phosphonate esters, e.g. 14 (Scheme 2).8 The best result obtained
for the olefination of ketone 13 was a 50% yield of a 5:1 mixture
of (E)-:(Z)-alkene isomers 15 and 16, respectively. This was
achieved using phosphonate ester 14 under the conditions
described by Masamune and Roush.9 As these preliminary
results were not competitive with the use of the Heck reaction
for this key transformation, we did not pursue it further and
decided to focus upon routes that would incorporate the Heck
reaction.
The discovery route to tetrahydroquinoline 12 relied upon a
chiral auxiliary to control the addition of a nucleophile to an
imine. In the case of the addition of allyltributyl stannane the
levels of diastereoselectivity were excellent; the price to be paid,
however, was the use of toxic tin reagents at very low
When considering the synthesis of tetrahydroquinoline 12
there are two key synthetic features, namely the C-2 chiral centre
which has the (R) configuration and the tetra-substituted alkene
that has the (E) configuration.
(8) Bacchi, S.; Almi, M.; Guelfi, S.; Perboni, A. ‘Synthesis of 4-substituted-
2-carboxy tetrahydroquinolines Via intermolecular olefination reaction.’
Poster presented at IASOC 2002 congress. September 2002, Ischia,
Italy.
(7) Butters, M.; Catterick, D.; Craig, A.; Curzons, A.; Dale, D.; Gillmore,
A.; Green, S. P.; Marziano, I.; Sherlock, J.-P.; White, W. Chem. ReV.
2006, 106, 3002–3027.
(9) Blanchette, M. A.; Choy, W.; Davis, J. T.; Essenfield, A. P.;
Masamune, S.; Roush, W. R.; Sakai, T. Tet. Lett. 1984, 25, 2183–
2186.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1131

Scheme 3
to a larger scale, a revised imine-forming process was developed
whereby aniline 4 and ethyl glyoxalate were heated to reflux
under Dean and Stark conditions for 1 h. This established an
equilibrium mix of 20 (64%): 22 (8%): 23 (24%): 24 (2%): 4
(2%), whereupon a much lower charge of magnesium sulphate
(10 wt %) was added to promote the conversion to imine 20
and a final equilibrium mix of 20 (92%): 23 (3.5%): 24 (4%)
was produced. The scaling up of heterogeneous reactions is
often fraught with difficulties, and in the case of this imine-
forming reaction it was believed that, as the scale increased,
the magnesium sulphate became aggregated and adhered to the
walls of the reactor, and thus was unable to participate in the
process. In the revised process, aggregation of the magnesium
sulphate was prevented by first removing the water, using a
Dean and Stark trap, as it was produced. Many other Lewis
acids were screened for their ability to promote imine formation.
Of these, magnesium bromide and zinc iodide were particularly
effective; however, these particular Lewis acids were not
compatible with telescoping into the subsequent Mannich
reaction as aldehyde 21 proved to be somewhat unstable in their
presence.
The second step in the telescoped sequence was the Lewis
acid-promoted Mannich addition of vinyloxytrimethylsilane 7
to imine 20 to afford aldehyde 21. This reaction was not without
difficulties as the reaction scale was increased. Initially ytterbium
triflate was used to promote the transformation; however, it was
found that the isolated yield was very sensitive to the reaction
time. If the reaction time was much longer than 30 min, then
a very low yield was obtained. This situation was felt to be
unsatisfactory for scale up. The reason for this was the formation
of the reactive hydroxy-tetrahydroisoquinoline 25 Via a
Friedel-Crafts reaction (Figure 2). A brief survey of Lewis
acids identified trimethylsilyl triflate as a potential alternative,
as the isolated yield seemed consistent regardless of reaction
time. However, when this procedure was repeated on a large
scale, some variability was observed. Ten batches were run on
a 100 kg plus scale, and the yield (over the three steps shown
in Scheme 3) varied from 47% to 62% with an average of 56%.
A more in-depth investigation using process analytical tech-
niques (PAT) was undertaken. Monitoring the reaction with
either an in situ ReactIR probe or in a tube in an NMR machine
indicated that no reaction occurred when the trimethylsilyl
triflate was added to a premixed solution of imine 20 and vinyl
silane 7. Aldehyde 21 was only produced when the water,
initially assumed only to quench the reaction, was added. This
issue had been masked as the samples for in-process control
had been quenched prior to analysis. These observations came
as surprise but at least indicated the source of the variability in
the process, as the reaction was not actually occurring until
the quench was added. The requirement for a proton source to
promote a Mannich reaction has precedent in work by Snapper
and Hoveyda10 which indicated that 1.0 equiv of isopropanol
was required to be present. The possibility that triflic acid was
the active promoter for the reaction was discounted as PAT
monitoring of reactions containing triflic acid again showed no
reaction had occurred until the water quench had been added.
temperatures. The alternative was the addition of the vinyloxy-
trimethylsilane 7, but this was a less selective reaction. In either
case the isolated yield from imine 5 to aldehyde 8 was on the
order of 50%. We reasoned we might be able to achieve similar
yields based upon a resolution strategy and thereby use
commercially available ethyl glyoxalate instead of oxalate 3.
This would reduce the number of stages and remove a
potentially tricky oxidation reaction from the sequence. In
addition we would look to recycle the unwanted isomer from
the resolution as it was felt that the chiral centre at C-2 bearing
an ester or acid function could be labile. Having made the
decision about the overall route strategy (Scheme 3), there were
still issues and options to be resolved as well as the short- and
long-term drug substance supplies to deliver. Our first goal was
a route to the racemic alkene 17, the first intermediate on our
proposed route.
The 5-chloro-2-iodoaniline 4 was obtained from the suppliers
as the hydrochloride salt, and this was liberated as the free base
by washing a toluene solution with aqueous potassium carbon-
ate. To form imine 20, aniline 4 and ethyl glyoxalate were
heated to reflux under Dean and Stark conditions in the presence
of magnesium sulphate (Scheme 4). On a small laboratory scale,
an equilibrium mixture was established within 1 h, this mix
contained 93% of the desired imine 20 along with small
amounts of the following byproduct N,O-hemiaminal 22 (1%),
aminal 23 (3%) and N,O-ethyl acetal 24 (2%), (Figure 1). It
was assumed that the ethyl glyoxalate solution was the source
of the ethanol incorporated into N,O-ethyl acetal 24. When the
reactions were scaled to a much larger scale, the final position
of the equilibrium shifted, on a 112-kg scale the equilibrium
position of 20 (78%): 22 (4%): 23 (6%): 24 (11%): 4 (1%)
was established after 5 h. Although this equilibrium position
was very different to that obtained on a small scale, it was not
believed that this would have a detrimental affect on the overall
efficiency of the process as it could be thought that compounds
22, 23, and 24 are all precursors to imine 20 under the
subsequent Lewis acid conditions. After scaling the reactions
(10) Josephsohn, N. S.; Snapper, M. L.; Hoveyda, A. H. J. Am. Chem.
Soc. 2004, 126, 3734–3735.
1132
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 4a
a Reagents and conditions: i) Tol, aq K2CO3 then MgSO4, EtO2CCHO, ∆. ii) 7, TMSOTf, -15 °C. iii) 9, DBU, acetonitrile, 56% (over 3 steps).
As previously mentioned, this three-step telescoped stage
of chemistry was run repeatedly on a >100 kg scale, and the
average yield was 56%. In total, approximately 1000 kg of
alkene 17 was produced by this process.
With a concise synthesis of the carbon framework of the
molecule in place, albeit in racemic form, the end-game of the
Figure 1
synthesis had to be put in place. Here there were two choices
(Scheme 3): Approach 1, perform the intramolecular Heck
reaction then undergo some sort of resolution process; Approach
2, perform the resolution and then form the tetrahydroquinoline
by the intramolecular Heck reaction. Our initial scale up
campaign to afford approximately 10 kg of drug substance used
Approach 1. This 10 kg of drug substance was used to fund
key repeat dose safety assessment studies and would define the
quality standard that subsequent material would have to meet.
Approach 2 was used to prepare >300 kg of drug candidate for
clinical studies.
Figure 2
It would therefore appear that the reaction is actually promoted
by triflic acid and a proton source, i.e., water, in the case
described here. To gain greater control over the process it would
be necessary to add the proton source, e.g., water, isopropanol,
or aqueous acetic acid, in a controlled fashion.11
Approach 1, the Heck reaction was performed on alkene 17
using 2 mol % of PdCl2 and 5 mol % triphenylphosphine, these
conditions were preferable to the use of Pd(PPh3)4 as both the
charge of palladium and the reaction time were both reduced
(Scheme 5). The tetrahydroquinoline 18 was isolated by
crystallisation in 72% yield. The isolated product did on
occasion contain ca. 2-3% of the trans-endo isomeric byprod-
uct 26 (Figure 3) and high levels of residual palladium, ca. 2000
ppm. The presence of either of these impurities did not affect
the quality of the drug substance produced as they were
effectively cleared in the subsequent chemistry. A resolution
of the two enantiomers of 18 was achieved by way of an
enzyme-catalysed hydrolysis of the ester functionality. A full
screen of hydrolytic enzymes was undertaken, and from it a
hit was achieved using a crude preparation from Aspergillus
niger lipase. Further investigation showed that it was ferulic
acid esterase that was the active component in the preparation
and so aqueous preparations of ferulic acid esterase were
obtained and used in the development of the process. The ferulic
acid esterase hydrolysis of racemic ester 18 was highly selective
for the hydrolysis of the (R)-enantiomer, with a selectivity factor
E > 100. At the end of the reaction the desired (R)-enantiomer
12 could be isolated from the mixture as the meglumine salt in
37% yield, and the unreacted (S)-enantiomer of ester 25 could
also be recovered in 55% yield. The intermediate grade drug
substance was recrystallised from acetone and water to expel
to low levels the oxidized quinoline impurities 27 and 28 (Figure
Although our stated aim was a route to racemic alkene 17
Via aldehyde 21, reports of asymmetric Mannich reactions have
been published in the literature.10,12 Our attempts to apply the
reported catalysts/conditions to our system resulted in low
conversions and low enantiomer excesses being obtained.13
The Wittig reaction between aldehyde 21 and the phospho-
rane formed from phosphonium salt 9 and 1,8-diazabicycl-
[5.4.0]undec-7-ene proved to be straightforward. The product
17 was formed in solution as a 95:5 ratio of (E):(Z) double
bond isomers. The product was then isolated from solution by
way of a seeded crystallisation process from isopropanol. The
product 17 was isolated as a white solid, with a purity as
determined by HPLC of >99.3% PAR (peak area ratio), with
the levels of the unwanted (Z)-isomer <0.3% PAR.
(11) Due to the decreased requirement for the drug candidate molecule we
were unable to complete the work to gain full understanding and
control of this process.
(12) (a) Kobayashi, S.; Hamada, T.; Manabe, K. J. Am. Chem. Soc. 2002,
124, 5640–5641. (b) Kobayashi, S.; Matsubara, R.; Nakamura, Y.;
Kitagawa, H.; Sugiura, M. J. Am. Chem. Soc. 2003, 125, 2507–2515.
(c) Cordova, A.; Barbas, C. F. Tet. Lett. 2003, 44, 1923–1926. (d)
Nakamura, Y.; Matsubara, R.; Kiyohara, H.; Kobayashi, S. Org. Lett.
2003, 5, 2481–2484.
(13) Subsequent to our investigations there have been a considerable number
more publications on the asymmetric Mannich reaction including
examples using acetaldehyde. (a) Yang, J. W.; Chandler, C.; Stadler,
M.; Kampen, D.; List, B. Nature 2008, 452, 453–454. (b) Hayashi,
Y.; Okano, T.; Itoh, T.; Urushima, T.; Ishikawa, H.; Uchimaru, T.
Angew. Chem., Int. Ed. 2008, 47, 9053–9058. (c) Kano, T.; Yamagu-
chi, Y.; Maruoka, K. Angew. Chem., Int. Ed. 2009, 48, 1838–1840.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1133

Scheme 5a
a Reagents and conditions: i) PdCl2, PPh3, NEt3, Tol, ∆, 72%. ii) Ferulic acid esterase, 0.1 M sodium citrate buffer, DMSO, 38 °C then aq meglumine, acetone, 12
) 37%, 25 ) 55%. iii) Acetone, H2O, 90%. iv) DBU, acetone, ∆, 88%.
18 was not always possible. (5) There was considerable concern
over the robustness of the ferulic acid esterase process. Although
the reaction was highly selective for the hydrolysis of the (R)-
enantiomer, it was never possible to get the reaction to proceed
past 40% conversion. Many different factors that could possibly
affect the reaction were investigated, including temperature,
enzyme loading, concentration, solvent composition, alternative
Figure 3
buffers, and reaction pH, but none seemed to increase the level
of conversion. It was the physical nature of the reaction that
led to some clues as to why complete conversion was not
observed. To get the reaction to proceed it was necessary to
dissolve the substrate 18 in warm dimethyl sulphoxide and then
to add this to a solution of the enzyme in citrate buffer.
Immediately upon mixing, a precipitate formed. Further inves-
tigation by microscopy suggested that this was an amorphous
coprecipitate of substrate and enzyme, and if the precipitate was
filtered off and placed in a warm oven, then the reaction still
proceeded. If the enzyme was challenged with crystalline
substrate, then no reaction occurred. Amorphous 18 was
prepared but was unstable with respect to undergoing a solid-
state transition and crystallising. Subjecting amorphous 18 to
the dimethyl sulphoxide/citrate buffer conditions again promoted
rapid crystallisation. Our rationale for the incomplete conversion
is that an amorphous coprecipitate of the enzyme and substrate
is initially formed. The amorphous ester is a substrate for the
enzyme, but at the same time partial crystallisation is occurring
which means the reaction cannot go to completion. Since what
was causing the amorphous coprecipitation was not understood
or controllable, it was felt that the reaction would not be robust
and not the basis of a longer-term manufacturing strategy
capable of producing hundreds or thousands of kilograms.
The alternative synthetic approach [Approach 2 (Scheme 3)]
to perform a resolution and then carry out the intramolecular
Heck reaction was investigated. Again, a resolution of the two
enantiomers of 17 was achieved by way of an enzyme-catalysed
hydrolysis of the ester functionality. A full screen of hydrolytic
enzymes was undertaken, and from it several hits were
identified. Two enzymes were identified that showed modest
selectivity for the hydrolysis of the (R)-enantiomer of 17. These
were porcine pancreatic lipase (PPL) and trypsin derived from
bovine pancreas. Additionally, two enzymes were identified that
showed very high selectivity (E > 500) for the hydrolysis of
Figure 4
4). By means of the route shown in Scheme 5 14 kg of the
target drug substance 12 (as meglumine salt) was prepared. The
overall quality of the batches produced was excellent, area of
main peak by HPLC >99.5%, all individual impurities by HPLC
<0.15%, chiral purity by HPLC >99.5%, and residual palladium
<10 ppm. Additionally it was shown that the unreacted (S)-
enantiomer 26 could be recycled to the racemic compound 18
upon heating with DBU in acetone in 88% yield.
Although 14 kg of drug substance had been produced by
Approach 1 as outlined in Scheme 4, it was felt there were
several significant shortcomings in the process that would mean
it would have been difficult to produce hundreds of kilograms
Via this route. (1) As a general rule if a classical resolution
strategy is going to be employed, then it should be performed
as early as possible in the synthetic sequence.14 (2) The large
workup volumes encountered in the enzymatic resolution stage
severely restricted throughput. (3) Solutions of ferulic acid
esterase were not freely available. (4) Although the unwanted
enantiomer 25 could be recycled to give the racemic ester 18,
a buildup of impurities during this process was observed. These
impurities included the trans-endo isomer 26 (which in inde-
pendent experiments was shown not to be a substrate for ferulic
acid esterase) as well as levels of the quinolines 27 and 28.
Producing drug substance of high quality from the recycled ester
(14) Zang, T. Chem. ReV 2006, 106, 2583–2595.
1134
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 6a
the (S)-enantiomer of 17. These were lipases from Mucor miehei
and Candida cylindrecea. It was decided to try and pursue the
use of M. miehei using the polymer-supported reagent lipozyme
to develop a process. This decision was based upon the
following points: (1) Both the PPL and the trypsin are derived
from animal sources. To reduce the risk of the drug substance
being contaminated with transmissible spongiform encaphalo-
pathies (TSE) and therefore comply with the leading regulatory
authorities it was decided not to use these reagents. (2) The
enzymes PPL and trypsin only showed modest selectivity of
the hydrolysis of the (R)-enantiomer, so a further up-grade of
chiral purity would be required. (3) The enzymes M. miehei
and C. cylindrecea hydrolysed the unwanted (S)-enantiomer,
so this would mean introducing a further synthetic step to cleave
the ester at the end of the synthesis. This would pose a risk
that the synthetic sequence would be longer than the one shown
in Scheme 5. However, it was felt that the final-stage ester
hydrolysis should be a clean transformation, and therefore it
should be possible to incorporate the meglumine salt formation
into a final stage of chemistry and produce good-quality drug
substance if the quality of the product arising from the
intramolecular Heck reaction was controlled. This would
eliminate the need for a final discrete recrystallisation step to
upgrade product quality and therefore save a stage of chemistry.
In developing a resolution procedure for racemic ester 17
using lipozyme we looked to build upon our experience of
lipase-catalysed kinetic resolutions using supported enzymes15
(Scheme 6). Using 88% w/w tert-butanol16 as solvent, the
kinetic resolution could be driven to completion in ca. 18 h if
30% by weight of lipozyme was used. The progress of the
reaction was determined by measuring the chiral purity of the
reacting ester; once the ratio of enantiomers (R):(S) was >98.0:
<2.0, then the reaction was deemed complete. The chiral purity
of acid 29 produced during the reaction was always >99.5%.
A charge of 30% by weight of lipozyme would have been
uneconomical, so efforts were made to reduce the charge. As
had been previously shown15 control of pH was an effective
tool, with the optimum pH for this transformation being within
the range pH 6.1-6.3. As the pH of the reaction mixture
decreased during the reaction, a dilute ammonia solution was
added to maintain a constant pH. By controlling the pH in this
way we could achieve a reaction time of <24 h using a 7% by
weight charge of lipozyme. Our attempts to try and recycle the
enzyme resin proved fruitless, as we showed in control
experiments that the enzyme was being denatured during the
process by the ethanol that was liberated. Further experimenta-
tion showed that increasing the level of tert-butanol present in
the solvent facilitated a further reduction in lipozyme charge,
without extending the reaction time. With a tert-butanol level
of 95% w/w, the charge of lipozyme could be reduced to 3%
by weight. During the scale up to 139-kg scale, we opted for
the convenience of using 88% w/w tert-butanol.
a Reagents and conditions: i) Lipozyme, t-BuOH, NH3, pH 6.2, 40 °C, 45%.
ii) AcCl, EtOH, 50 °C, 44% over 2 steps. iii) NaOEt, EtOH, 0 °C, 89%. iv)
PdCl2, PPh3, NEt3, Tol, ∆, 74%. v) NaOH, aq THF, 0 °C then aq meglumine,
acetone, 92%.
Once the reaction was complete, the enzyme resin was
removed by filtration. Controlled addition of 0.5 equiv of
aqueous sodium bicarbonate solution caused the (R)-ester 19
to crystallise from solution as a white solid and the (S)-
enantiomer of acid 29 to remain in solution as the sodium salt.
The (R)-ester 19 was consistently produced in a 45% yield, and
in total, 600 kg was prepared. During the crystallisation process
the chiral purity of ester 19 increased to 99.7%, and the purity
as determined by HPLC was >99.5% PAR.
To increase the overall throughput of the sequence and to
improve the economics it would be necessary to recycle the
sodium salt acid 29 back to racemic ester 17. The filtrate from
the enzyme-catalysed kinetic resolution was concentrated by
distillation to remove the tert-butanol, the sodium salt was
neutralised, and the parent acid was then extracted into ethyl
acetate.17 The solvent was exchanged for ethanol, and a solution
of hydrogen chloride in ethanol was added to promote the
esterification to the (S)-ester 30, a reaction which took ca. 20 h
to complete at 50 °C. Ester 30 could be isolated as a white
crystalline solid by seeding and then adding water. The (S)-
ester 30 was produced in a 44% yield over two steps from
racemic ester 17, and in total, 420 kg was prepared.
Racemisation of (S)-ester 30 into racemic ester 17 could be
achieved using sodium ethoxide in ethanol. Under the basic
reaction conditions the product was prone to isomerisation to
(15) Atkins, R. J.; Banks, A.; Bellingham, R. K.; Breen, G. F.; Carey, J. S.;
Etridge, S. K.; Hayes, J. F.; Hussain, N.; Morgan, D. O.; Oxley, P.;
Passey, S. C.; Walsgrove, T. C.; Wells, A. S. Org. Process Res. DeV.
2003, 7, 663–675.
(17) Attempts to isolate the parent acid 29 as crystalline solid were
unsuccessful. Acidification of solutions of the sodium salt lead to
amorphous material being produced, although this material filtered
well the material lost much of its structure during drying. A crystalline
potassium salt of 29 was isolated from ethyl acetate.
(16) 88% w/w tert-butanol contains by weight 88% of tert-butanol and
12% of water. This grade of solvent represents the azeotropic boiling
mixture at atmospheric pressure, is commercially available in bulk
and remains a liquid down to-10 °C.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1135

Figure 5
Figure 6
Table 1. Effect of solvent and ligand on intramolecular
form compound 32 (Figure 5). This side reaction could be
minimised to <2% by reducing the reaction temperature to 0
°C and reducing the charge of sodium ethoxide to 0.25 equiv.
Again the reaction progress was determined by using chiral
HPLC, the end point of which was determined as the ratio of
(S)-ester 30:(R)-ester 19 < 51.0:>49.0 and was achieved within
24 h. Ester 17 could be isolated as a white crystalline solid by
neutralisation with AcOH and then addition of water. This
recycled racemic ester 17 was produced in an 89% yield, and
in total, 380 kg was prepared. The purity of recycled ester 17
as determined by HPLC was equivalent to the de noVo ester.
Importantly, when the recycled ester 17 was subjected to the
lipozyme-promoted kinetic resolution, the reaction proceeded
in a fashion analogous to the de noVo material, and the quality
of the (R)-ester 19 produced was again of equivalent quality
and performance.
A modified one-pot procedure for the combined esterification
and racemisation reactions was demonstrated on a small,
laboratory scale. This used Amberlyst A-15 resin as catalyst
for the esterification. The isolated yield of 74% was comparable
with the two-stage procedure, but given the sensitivity of the
racemisation protocol it was decided that a greater understanding
of the reaction parameters, especially around the charge of
sodium ethoxide, would be needed ahead of scale up.
The efficiency with which the resolution and racemisation
protocols worked vindicated our faith in such an approach. The
overall yield of (R)-ester 19 based upon recycling the unwanted
enantiomer just the once increases from 45% to 63%; if this
protocol were to go into continual manufacturing, then the
overall efficiency would be 74% as the unwanted enantiomer
would be continually recycled.
The second step in Approach 2 was to carry out the
intramolecular Heck reaction on a single enantiomer of sub-
strate. We had some reservation as to whether the pre-existing
reaction conditions utilising excess triethylamine in refluxing
toluene would be amenable to substrate 19 as both the substrate
19 and product 31 had been shown to racemise at elevated
temperature in the presence of base. Fortunately, no racemisation
was observed under the standard conditions used for racemic
substrate 17. As well as forming the desired product 31 the
reaction also gave the identified major byproducts: the trans-
endo isomer 26, the des-halo compound 33 (Figure 6), and a
small number of unidentified minor byproducts that were
formed at levels <0.5% PAR. As was previously noted,2 the
product distribution obtained from this reaction depends upon
the reaction conditions used. A considerable number of ligands
(phosphines, palladocycles, carbenes), palladium sources, sto-
Heck reaction
ratio
of 31:26:33
PPh3 (equiv)
solvent
yield (%)
0
toluene
DMF
60
70
17:78:3
10:82:8
83:11:6
85:6:8
0
1
toluene
toluene
toluene
DMF
97
2.0
2.5
2.5
4
>99
>99
97
>99
88:7:5
31:61:8
88:7:5
toluene
a Peak area ratios determined by HPLC at 245 nm.
ichiometries, solvents, and bases were screened.18 Table 1 shows
just a few selected results that show the requirement for 2.5
equiv of triphenylphosphine with respect to palladium. It was
observed that the product distribution remained constant
throughout the course of the reaction, thus discounting an
equilibration of products once the reaction was complete. The
screening experiments showed that similar results could be
obtained using other high-boiling aromatic hydrocarbon solvents
such as xylenes or mesitylene and that bases such as tributy-
lamine could be used in place of triethylamine. The product
distribution observed is consistent with the reaction proceeding
Via the mechanism shown in Scheme 7.2 The origin of the
reaction selectivity is not apparent, but it could be concluded
that the reaction conditions used help prolong the lifetime of
intermediate 34 to enable rotation to occur and then reductive
elimination to occur Via Ha rather than elimination of Hb to
form the trans-endo isomer. The level of palladium chloride
used was 2 mol % and that of the triphenylphosphine 5 mol
%; these materials are relatively cheap, and their contribution
to the overall cost of goods was so low that there was no desire
to try and reduce the levels. As discussed previously, the quality
of the Heck product 31 would directly affect the quality of the
drug candidate molecule 12. The two key quality-critical
attributes we were concerned with were the levels of residual
palladium and the level of trans-endo isomer 26. The procedure
outlined in Scheme 5 was used to fund the key repeat dose
safety assessment studies and therefore, in conjunction with
the ICH guidelines, set the bench mark for quality for future
material. That route was tolerant of levels of palladium and
trans-endo isomer 26 much greater than those in Scheme 6.
Many palladium scavengers have been reported in the litera-
(18) (a) Amatore, C.; Jutand, A. Acc. Chem. Res. 2000, 33, 314–321. (b)
Beletskaya, I. P.; Cheprakov, A. V. Chem. ReV. 2000, 100, 3009–
3066. (c) Whitcombe, N. J.; Hii, K. K.; Gibson, S. E. Tet. 2001, 57,
7449–7476. (d) Littke, A. F.; Fu, G. C. Angew. Chem., Int. Ed. 2002,
41, 4176–4211. (e) Prashad, M. Top. Organomet. Chem. 2004, 6, 181–
203. (f) Farina, V. AdV. Synth. Catal. 2004, 346, 1553–1582.
1136
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 7
Scheme 8a
a Reagents and conditions: i) Tol, aq K2CO3 then MgSO4, EtO2CCHO, ∆. ii) 7, TMSOTf, -15 °C. iii) 9, DBU, acetonitrile, 60% (over 3 steps). iv) Lipozyme,
t-BuOH, NH3, pH 6.2, 40 °C, 40%. v) PdCl2(Ph3)2, aq K2CO3, 1,4-dioxane, ∆.
ture;19 it was found that the addition of 6 mol % of trimercap-
totriazine20 at the end of the reaction and then continuing to
heat at reflux reduced the levels of residual palladium in ester
31 from ca. 2000 ppm to <15 ppm. The excess trimercaptot-
riazine and the palladium-trimercaptotriazine complex could
be removed by filtration along with triethylammonium iodide.
After an aqueous workup the ester 31 could be isolated from a
toluene solution by the controlled addition of isooctane. On an
87 kg input scale, ester 31 was consistently isolated in 74%
yield, and in total, 245 kg were produced. The quality of the
intermediate was excellent, no des-halo 33 was detected, and
the levels of the trans-endo isomer 26 were in the range
0.16-0.21% PAR, the target specification for this impurity
being <0.25% PAR. No racemisation of the chiral centre was
observed during the reaction, and the chiral purity of the isolated
product was 100%.
There are aspects of quality by design that could be
incorporated into this stage of chemistry. One such example
would be the use of the commercially available preformed
bis(triphenylphosphine)palladium(II) dichloride catalyst. Given
the sensitivity of the reaction outcome to the ratio of palladium
to ligand, it would increase the consistency of the reaction in
terms of both profile and duration if only a single reagent needed
to be charged and the ratio of palladium to ligand to be fixed.
It should be noted that the ratios listed in Table 1 are not
normalised for the relative response of each component, so any
fine-tuning of the reaction profile would need quantitative
analysis to maximise the amount of desired product 31 being
formed. Since impurities are going to be produced, then the
impurity that is easiest to expel (in this case the des-halo
compound 33) should be formed in preference to the trans-
endo compound 26 which is more difficult to remove.
An analogous route using the bromo-substituted Heck
precursor was also investigated briefly, Scheme 8. The Heck
reaction precursor 37 was prepared from 2-bromo-5-chloro-
aniline 353b as a single enantiomer by applying the methods
used in Schemes 4 and 6. The conditions used for ester 19 were
not suitable for this substrate. Using triphenylphosphine as
ligand a screen of solvents and bases showed that a product
distribution of desired product 31:trans-endo isomer 26:des-
(19) (a) Galaffu, N.; Man, S. P.; Wilkes, R. D.; Wilson, J. R. H. Org.
Process Res. DeV. 2007, 11, 406–413. (b) Pink, C. J.; Wong, H.-T.;
Ferreira, F. C.; Livingston, A. G. Org. Process Res. DeV. 2008, 12,
589–595. (c) Flahive, E. J.; Ewanicki, B. L.; Sach, N. W.; O’Neill-
Slawecki, S. A.; Stankovic, N. S.; Yu, S.; Guinness, S. M.; Dunn, J.
Org. Process Res. DeV. 2008, 12, 637–645, references therein.
(20) 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.;
Humora, M. J.; Anderson, N. G. Org. Process Res. DeV. 1997, 1,
311–314.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1137

halo compound 33 of 86:9:0.5 could be achieved using
potassium carbonate as base and dioxane as solvent. Again no
racemisation at C-2 was observed.
routes of synthesis. The development routes used a resolution
strategy, whereas the discovery route used a chiral auxiliary
approach; by recycling the unwanted enantiomer the throughput
and economics could be increased and potentially problematic
oxidation steps removed. By performing the resolution strategy
earlier in the sequence, improved throughput was achieved, and
a more readily available supported enzyme could be used. Using
PAT afforded greater process understanding of the Mannich
reaction to form aldehyde 21, but due to resourcing constraints
a complete solution to the problem could not be obtained. The
quality attributes of the target compound 12 could be achieved
by control of the quality of the penultimate intermediate 31,
controlling the reaction temperature during the saponification,
and by controlling the final crystallisation.
The final stage in the synthetic sequence required saponifica-
tion of the ethyl ester 31 to the carboxylic acid 12, formation
of the meglumine salt, and then a crystallisation that controlled
not only the chemical purity but also the solid-state attributes
as well. The quality attributes of ester 31 had already been
controlled, that is chiral purity, related impurities by HPLC,
and residual Pd. It was felt that only the chiral purity could be
eroded during the final processing step. To refine and optimize
the reaction conditions a statistical design of experiment (DoE)
approach was used; the factors studied were reaction temper-
ature (range 0-20 °C), amount of sodium hydroxide, volumes
of tetrahydrofuran and water, and the rate at which the aqueous
sodium hydroxide was added. The most significant factor
affecting product quality was the reaction temperature, with the
low temperature being optimum. The other factors were then
modified to ensure that the reaction was complete within 4 h
and that a homogeneous reaction solution was obtained at the
end point. Under the optimised reaction conditions only 0.3%
racemisation was observed. No upgrade of chiral purity was
observed during the crystallisation of form 2. Once the reaction
was complete, the excess sodium hydroxide was neutralized
with hydrochloric acid, and the parent free acid 12 was extracted
and solvent exchanged into acetone. The meglumine salt was
formed by addition of aqueous N-methyl-D-glucamine, the
solution composition was adjusted to make the solution
metastable with respect to the salt, and then a seed of form 2
was added to induce crystallisation which was followed by
addition of further acetone antisolvent and temperature lowering
to increase recovery. Two solid-state forms of salt 12 were
known. Form 1 was a metastable channel hydrate, while form
2 was a highly crystalline, nonsolvated, nonhydrated form; it
was form 2 that was required. On a 50 kg input scale, salt 12
was consistently isolated in 92% yield; in total, >300 kg were
produced. The quality of the final drug candidate molecule was
excellent, with the main peak >99.9% PAR; no related impuri-
ties were detected above 0.1% PAR, the chiral purity was
99.7%, the level of residual Pd <10 ppm, and the crystalline
form was always form 2.
Experimental Section
(4E)-2[(5-Chloro-2-iodophenyl)amino]-4-(2-oxo-1-phenyl-
pyrrolidin-3-ylidene)butanoic Acid Ethyl Ester (17). 5-Chloro-
2-iodoaniline hydrochloride (112 kg, 386 mol) was suspended
in toluene (927 kg) and treated with a solution of potassium
carbonate (56 kg, 405 mol) in water (560 L) at 20-25 °C for
30 min. The phases were separated, and the organic phase was
washed with water (560 L). Anhydrous magnesium sulphate
(84.0 kg, 699 mol) was added and the solution dried by heating
at reflux with a Dean and Stark trap. A 50% w/w solution of
ethyl glyoxalate in toluene (119 kg, 582 mol) was added over
30 min, maintaining reflux. The reaction mixture was main-
tained at reflux for 5 h and water (8.8 L) collected in the Dean
and Stark trap. The reaction mixture was cooled to -15 °C,
and vinyloxytrimethylsilane (67 kg, 578 mol) was added over
10 min. Trimethylsilyl triflate (4.3 kg, 19.3 mol) was added
over 5 min maintaining the temperature at -15 to -19 °C.
After stirring for 30 min water (560 L) was added, allowing
the temperature to rise to 20-25 °C. The phases were separated,
and the organic phase was washed with water (560 L). The
organic phase was filtered to remove residual inorganic solids
and then concentrated by distillation under reduced pressure,
maintaining the internal temperature below 60 °C to leave a
solution of aldehyde 21 with a residual volume of 780 L.
Phosphonium salt 7 (195 kg, 388 mol) was suspended in
acetonitrile (386 kg) and 1,8-diazabicyclo[5.4.0]undec-7-ene (63
kg, 414 mol) added; the mixture was stirred at 18-22 °C for
1 h and 40 min to form a solution of the phosphorane. The
toluene solution of aldehyde 21 (780 L) was added to the
phosphorane solution over 50 min. The mixture was stirred for
45 min, and then water (560 L) was added. After stirring for
15 min the phases were separated, and the organic phase was
concentrated by distillation under reduced pressure to leave a
residual volume of 370 L. Isopropanol (704 kg) was added and
the temperature adjusted to 28 °C. Seed crystals of alkene 17
(125 g) were added, and the solution was aged at 25-30 °C
until crystallisation was established. The slurry was cooled to
22 °C, aged for 2 h, then cooled and aged at 0 °C for 1 h. The
product was isolated by filtration, and the filtercake was washed
with isopropanol (354 kg) and then dried under vacuum at
45-50 °C to afford alkene 17 as a white solid (113 kg, 56%
yield over three steps).
Following Approach 2 as outlined in Scheme 3 and as
exemplified in Scheme 6, >300 kg of the chiral tetrahydroqui-
nolone 12 was produced. This approach overcame many of the
shortcomings outlined with Approach 1. The resolution was
performed earlier in the synthetic sequence, and the unwanted
enantiomer was recycled to produce material of quality similar
to that of the de noVo material; this improved throughput and
reduced cost. By controlling the quality of the penultimate
intermediate 31, the quality of the final compound obtained met
all specifications without the need for a recrystallisation, and
this kept the number of stages of chemistry to a minimum.
Conclusions
In conclusion, in this paper we have described our Chemical
Development effort to support a glycine antagonist to be used
for treatment of nicotine craving. The development chemists
delivered >300 kg of drug candidate molecule using two related
Mp ) 83-85 °C; 1H NMR (400 MHz, CDCl3) δ 7.72 (m,
2 H), 7.56 (m, 1 H), 7.39 (m, 2 H), 7.16 (tt, J ) 7.3, 1.1 Hz,
1138
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

1 H), 6.58 (tt, J ) 7.7, 2.8 Hz, 1 H), 6.51-6.49 (m, 2 H), 4.87
(d, J ) 8.0 Hz, 1 H), 4.29-4.24 (m, 3 H), 3.88 (t, J ) 6.9 Hz,
2 H), 2.85-2.73 (m, 4 H), 1.30 (t, J ) 7.1 Hz, 3 H); 13C NMR
(100 MHz, CDCl3) δ 171.6, 166.5, 146.5, 139.9, 139.6, 136.2,
135.6, 128.9, 127.1, 124.7, 119.7, 119.6, 111.1, 83.05, 62.0,
55.8, 45.3, 32.1, 21.7, 14.3; IR (neat cm-1): 3361, 1730, 1690,
1667. LRMS (CI + ve) m/z 525 (M+ + H).
was washed with acetone/water (400 mL; 1:1 v/v). The
filtercake was dried to leave ester 18 as a yellow solid (175 g,
88%).
(2R,4E)-2[(5-Chloro-2-iodophenyl)amino]-4-(2-oxo-1-phen-
yl-pyrrolidin-3-ylidene)butanoic Acid Ethyl Ester (19) -
Scheme 6. Racemic ester 17 (139 kg, 265 mol) and lipozyme
RM IM-supported enzyme (9.8 kg,) were slurried in 88% w/w
tert-butanol (1239 kg) at 39-41 °C, and the pH was maintained
at pH 6.1-6.3 by the addition of a solution of 1.5 M ammonia
in 88% w/w tert-butanol (typically 73.5 L). Once the reaction
was complete (as measured by chiral HPLC: (R)-ester 19:(S)-
ester 30 > 98.0:2.0, time 19 h), the resin was removed by
filtration and washed with 88% w/w tert-butanol (113 kg). To
the combined filtrates at 40 °C was added sodium bicarbonate
(427 kg; 2.6% w/w; aq; 132 mol). The mixture was stirred at
40 °C for 1 h during which period ester 19 crystallized from
solution. The slurry was cooled to 5 °C over 2 h 30 min and
then was aged at 3-4 °C for 1 h. The product was isolated in
a centrifuge and was washed with a mixture 88% w/w tert-
butanol (113 kg) and water (139 L) and then dried under
vacuum at 40-45 °C to afford ester 19 as a white solid (63 kg,
45% yield).
The mother liquors and cake wash, containing acid 29 as
the sodium salt, were concentrated by distillation under reduced
pressure to leave a residual volume ) 485 L. The concentrate
contained approximately 62 kg of acid 29.
HPLC in process control method: Chiracel OD column 250
mm × 4.6 mm, 5 µm. Eluent: 80% n-heptane, 20% EtOH +
0.1% TFA. Isocratic for 20 min. Flow rate 1.0 mL/min. Detector
at 255 nm. Injection volume 20 uL. Typical retention times:
(R)-enantiomer of acid, 7.4 min; (S)-acid 29, 8.0 min; (S)-ester
30, 9.4 min; (R)-ester 19, 10.0 min.
(2S,4E)-2[(5-Chloro-2-iodophenyl)amino]-4-(2-oxo-1-phen-
yl-pyrrolidin-3-ylidene)butanoic Acid Ethyl Ester (30) -
Scheme 6. To a reactor were charged a solution of sodium salt
of acid 29 (1100 kg; ca. 13% w/w; aq; 283 mol) and ethyl
acetate (1130 kg). Hydrochloric acid (115 kg; 2 M; aq) was
added, and pH of aqueous layer ) 6.0-6.5. The layers were
separated, and then the organic layer was concentrated by
distillation at atmospheric pressure to leave a residual volume
) 265 L. Ethyl acetate (890 kg) was added and then the solution
concentrated by distillation at atmospheric pressure to leave a
residual volume ) 320 L. Ethanol (500 kg) was added to give
a solution of acid 29 and the temperature adjusted to 45-50
°C. To a separate reactor was charged ethanol (110 kg), and
acetyl chloride (15.4 kg, 196 mol) was added over 40 min,
maintaining temperature <35 °C. The solution of hydrogen
chloride in ethanol was added to the solution of acid 29 in
ethanol and then continued stirring at 45-50 °C until the
reaction was complete (as measured by HPLC: ester 30:acid
29 > 95.0:5.0, time 20 h). The solution was cooled to 17-22
°C, and seed crystals of ester 30 (25 g) were added. The slurry
was cooled to 8-10 °C, and then water (315 L) was added.
The slurry was further cooled to 0-5 °C and aged for 1 h. The
product was isolated by filtration, and the filtercake was washed
with ethanol/water (265 kg; 3:1 v/v) and then dried under
vacuum at 45-50 °C to afford ester 30 as a white solid (141
kg, 44% yield over two steps).
(2R,4E)-7-Chloro-4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)-
1,2,3,4-tetrahydroquinoline-2-carboxylic Acid, (2R,3R,4R,5S)-
6-Methylamino-hexane-1,2,3,4,5-penta-ol (12) - Scheme 5.
Ester 18 (10.0 kg, 25.22 mol) was dissolved in dimethyl
sulphoxide (55 L) at 50 °C and then cooled to 40 °C. The ester
solution was added to a solution of ferulic acid esterase (4 L;
4% w/w; aq) in sodium citrate buffer (125 L; 0.1 M; aq) and
dimethyl sulphoxide (20 L) at 35 °C, and the resulting slurry
was stirred at 38 °C for 20 h. Butanone (200 L) was added, the
solution was cooled to 20-25 °C, and the phases were
separated. To the aqueous layer was added brine (50 L; 20%
w/w; aq), and then the mixture was extracted with butanone
(100 L). The butanone extracts were combined and then washed
with brine (160 L; 6% w/w; aq), brine (200 L; 6% w/w; aq),
and then brine (80 L; 20% w/w; aq). Further butanone (100 L)
was added and the solution concentrated by distillation at
atmospheric pressure to leave a residual volume ) 40 L.
Acetone (270 L) was added followed by a solution of N-methyl-
D-glucamine (2.46 kg, 12.61 mol) in water (12 L) at 20 °C.
The resulting slurry was aged at 20 °C for 1 h. The product
was isolated by filtration, and the filtercake was washed with
acetone (100 L) and then dried under vacuum at 40 °C to afford
intermediate grade salt 12 as a yellow solid (5.27 kg, 37%).
The mother liquors and wash containing ester 25 were
concentrated by distillation at atmospheric pressure to leave a
residual volume ) 100 L. Water (185 L) was added at 20 °C
and the resulting slurry aged at 20 °C for 1 h. The product was
isolated by filtration, and the filtercake was washed with water
(50 L) and then dried under vacuum at 50 °C to afford ester 25
as a yellow solid (4.50 kg, 45%).
The intermediate grade salt 12 (7.90 kg, 14.0 mol) was
suspended in water (24 L) and warmed to 55 °C to achieve
complete dissolution and then filtered into a clean reactor, and
the lines were rinsed through with water (8 L). The solution
was cooled to and maintained at 40 °C, and acetone (95 L)
was added Via a filter. Seed crystals of product 12 (form 2,
16 g) were added. Acetone (190 L) was added Via a filter, and
then the resulting slurry was cooled to 20 °C and aged for 1 h.
The slurry was cooled to 2 °C and aged for 1 h. The product
was isolated by filtration, and the filtercake was washed with
acetone (2 × 40 L) and then dried under vacuum at 50 °C to
afford salt 12 as a yellow solid (7.11 kg, 90%).
(2R,S,4E)-7-Chloro-4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)-
1,2,3,4-tetrahydroquinoline-2-carboxylic Acid Ethyl Ester
(18) - Scheme 5. Ester 25 (200 g, 504 mmol) was suspended
in acetone (1.4 L), and then 1,8-diazabicycl[5.4.0]undec-7-ene
(18.8 mL, 126 mmol) was added. The mixture was heated at
reflux for 4 h and then cooled to 40 °C, and acetone (1.0 L)
was added. The mixture was then cooled to 20-25 °C and
hydrochloric acid (1.0 L; 0.1 M; aq) added over 20 min. The
resulting slurry was aged and then filtered, and the filtercake
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1139

(2R,S,4E)-2[(5-Chloro-2-iodophenyl)amino]-4-(2-oxo-1-
phenyl-pyrrolidin-3-ylidene)butanoic Acid Ethyl Ester (17)
- Scheme 6. Ester 30 (140 kg, 267 mol) was suspended in
ethanol (540 kg) at -2 to +2 °C. Sodium ethoxide in ethanol
(21.6 kg; 21% w/w; 67 mol) was added, maintaining the
temperature within the range -2 to +2 °C and then continued
stirring at this temperature until the reaction was complete (as
measured by chiral HPLC: (S)-ester 30:(R)-ester 19 < 51.0: 49.0,
time 24 h). Acetic acid (4.0 kg, 67 mol) was added and the
slurry stirred for 30 min. Water (140 L) was added and the
slurry aged at -2 to +2 °C for 30 min. The product was isolated
by filtration, the filtercake was washed with ethanol/water (160
kg; 78:22 w/w) and then dried under vacuum at 45-50 °C to
afford ester 17 as a white solid (125 kg, 89% yield).
mol) was added, whilst maintaining the temperature at 1-2 °C.
The mixture was stirred at -1 to +1 °C until the reaction was
complete (as measured by HPLC: ester 31:acid 12 < 1.0:99.0,
time 2 h 45 min). Hydrochloric acid (78 L; 2.5 M; aq; 195
mol) was added, maintaining the temperature at 2 °C. Dichlo-
romethane (519 kg) was added, the solution was warmed to
20-25 °C and the phases separated. The aqueous phase was
extracted with dichloromethane (259 kg), then the combined
organic extracts were washed with demineralised water (195
L) before being filtered into a clean reactor and concentrated
by distillation under reduced pressure to leave a residual volume
) 146 L. Acetone (308 kg) was added Via a filter and the solu-
tion was concentrated by distillation under reduced pressure to
leave a residual volume ) 146 L. The temperature of the
concentrate was adjusted to 42 °C, and a prewarmed solution
of N-methyl-D-glucamine (23.9 kg, 123 mol) in demineralised
water (122 L) was added Via a filter. Acetone (78 kg) was added
Via a filter, the temperature was adjusted to 41 °C, and seed
crystals of product 12 (form 2, 245 g) were added. Acetone
(577 kg) was added Via a filter and the slurry stirred at 40 °C
for 1 h. The slurry was cooled to 4 °C and aged for 1 h. The
product was isolated by filtration, and the filtercake was washed
with acetone (2 × 77 kg) and then dried under vacuum at
35-40 °C to afford salt 12 as a yellow solid (63.9 kg, 92%
yield).
Mp ) 186 °C; Anal. Calcd for C20H17ClN2O3 ·C7H17NO5:
C, 57.50; H, 6.08; N, 7.45; Cl, 6.29. Found: C, 57.5; H, 5.8; N,
7.3; Cl, 6.1. 1H NMR (400 MHz, D2O) δ 7.58-7.44 (m, 4 H),
7.31 (t, J ) 6.9 Hz, 1 H), 7.23 (d, J ) 8.8 Hz, 1 H), 6.78 (d,
J ) 2.0 Hz, 1 H), 6.68 (dd, J ) 8.4, 1.9 Hz, 1 H), 4.10 (m, 1
H), 3.92 (t, J ) 5.8 Hz, 1 H), 3.89-3.72 (m, 5 H), 3.72-3.61
(m, 3 H), 3.31 (dd, J ) 13.9, 4.5 Hz, 1 H), 3.23 (dd, J ) 13.0,
3.6 Hz, 1 H), 3.19 (dd, J ) 13.0, 9.3 Hz, 1 H), 3.08 (m, 2 H),
2.77 (s, 3 H); 13C NMR (100 MHz, D2O) δ 181.5, 172.2, 148.0,
140.4, 140.2, 136.6, 131.0, 130.6, 127.6, 126.3, 123.7, 119.9,
117.5, 115.6, 72.3, 72.1, 72.0, 69.5, 64.1, 57.4, 52.5, 48.5, 34.4,
29.6, 27.9; IR (neat cm-1) 3392, 1737, 1218; HRMS calcd for
C20H17ClN2O3 + H+ 369.1005, found 369.0995;
(2R,4E)-7-Chloro-4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)-
1,2,3,4-tetrahydroquinoline-2-carboxylic Acid Ethyl Ester
(31) - Scheme 6. To a reactor were charged ester 19 (87 kg,
166 mol), triphenylphosphine (2.177 kg, 8.3 mol), palladium(II)
chloride (0.589 kg, 3.32 mol) and toluene (1060 kg). Triethy-
lamine (22.3 kg, 220 mol) was added, and the mixture was
heated at reflux until the reaction was complete (as measured
by HPLC: alkene 17:ester 19 < 1.0:99.0, time 3 h 45 min).
Trimercaptotriazine (1.74 kg, 9.81 mol) was added and mixture
heated at reflux for a further 1 h. The solution was cooled to
45 °C, and the solids were removed by filtration. The filtercake
was washed with toluene (152 kg). With the temperature
maintained at 40-45 °C, the combined filtrates were washed
twice with water (2 × 696 L) before being concentrated by
distillation under reduced pressure to leave a residual volume
) 680 L. The concentrate temperature was adjusted to 45 °C,
and isooctane (602 kg) was added over 1 h 40 min whilst
maintaining the temperature at 41-44 °C. The resulting slurry
was cooled and aged at 18 °C for 2 h. The product was isolated
by filtration, the filtercake was washed with heptane/isooctane
(2 × 177 L, 50:50 v/v) and then dried under vacuum at 45-50
°C to afford ester 31 as a yellow solid (48.7 kg, 74% yield).
1
Mp ) 160-162 °C; H NMR (400 MHz, CDCl3) δ 7.68
(m, 2 H), 7.39 (m, 2 H), 7.17 (tt, J ) 7.3, 1.0 Hz, 1 H),
7.14-7.12 (m, 1 H), 6.65-6.63 (m, 2 H), 4.79 (s, 1 H),
4.22-4.03 (m, 4 H), 3.85 (ddd, J ) 9.1, 7.8, 6.0 Hz, 1 H),
3.77 (ddd, J ) 9.1, 7.9, 4.9 Hz, 1 H), 3.48-3.42 (m, 1 H),
3.17-3.12 (m, 2 H), 1.17 (t, J ) 7.1 Hz, 3 H); 13C NMR (100
MHz, CDCl3) δ 172.4, 168.3, 145.7, 139.8, 137.6, 135.7, 129.1,
128.9, 125.2, 124.6, 119.8, 118.3, 116.8, 114.4, 61.6, 53.9, 45.6,
28.0, 25.2, 14.1; IR (neat cm-1): 3369, 1711, 1664; LRMS (CI
+ ve) m/z 397 (M+ + H).
Acknowledgment
We thank the following Chemical Development team
members for their contributions to the project; Mark Armitage,
Paul Blackler, Stephen Feely, Pat Harris, Jeegna Patel, Anne
Pierce, Iain Pitfield, Christian Schorn.
Supporting Information Available
1H and 13C NMR spectra for compounds 26, 27, and 28.
This material is available free of charge via the Internet at
http://pubs.acs.org.
(2R,4E)-7-Chloro-4-(2-oxo-1-phenyl-pyrrolidin-3-ylidene)-
1,2,3,4-tetrahydroquinoline-2-carboxylic Acid, (2R,3R,4R,5S)-
6-Methylamino-hexane-1,2,3,4,5-penta-ol (12) - Scheme 6.
Ester 31 (48.7 kg, 123 mol) was dissolved in tetrahydrofuran
(300 kg) and demineralised water (161 L) and was cooled to 0
°C. A solution of sodium hydroxide (78 L; 10% w/w; aq; 195
Received for review July 16, 2009.
OP9001824
1140
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development