Kilogram synthesis of a selective serotonin reuptake inhibitor

Article abstract of DOI:10.1021/op700125z

Process development of a selective serotonin reuptake inhibitor (1) is described. The synthesis features Nishiyama catalystmediated asymmetric cyclopropanation of vinyl indole 2 with ethyldiazoacetate to install the trans-disubstituted cyclopropane. The active pharmaceutical ingredient (1) was prepared in 13 chemical steps with 9 isolations and proceeded in an overall yield of 34%. ? 2008 American Chemical Society.

Full text of DOI:10.1021/op700125z

Organic Process Research & Development 2008, 12, 168–177
Kilogram Synthesis of a Selective Serotonin Reuptake Inhibitor
Robert Anthes, Osagie Bello, Serge Benoit, Chien-Kuang Chen,* Elisabeth Corbett, Richard M Corbett, Albert J. DelMonte,
Stephane Gingras, Robert Livingston, Justin Sausker, and Maxime Soumeillant
Process Research and DeVelopment, Bristol-Myers Squibb Company, One Squibb DriVe, P.O. Box 191,
New Brunswick, New Jersey 08903-0191, U.S.A.
Abstract:
Process development of a selective serotonin reuptake inhibitor
(1) is described. The synthesis features Nishiyama catalyst-
mediated asymmetric cyclopropanation of vinyl indole 2 with
ethyldiazoacetate to install the trans-disubstituted cyclopropane.
The active pharmaceutical ingredient (1) was prepared in 13
chemical steps with 9 isolations and proceeded in an overall yield
of 34%.
Figure 1. Asymmetric cyclopropanation approach to 1a and
1b.
restricted side chain.3 Recently, a neuroscience drug candidate
Introduction
The neurotransmitter 5-hydroxytryptamine (serotonin) plays
(1a, Figure 1) fitting this description entered development and
required material for toxicological and first-in-human studies.
This paper details the development of an expedient and efficient
chemical route that showcased a catalytic asymmetric cyclo-
propanation reaction and the production of kilogram quantities
of the maleate salt 1a, as well as the corresponding HCl salt
1b.
many key physiological roles,1 and not surprisingly, selective
serotonin reuptake inhibitors (SSRIs) are an important class of
drugs that have widespread application.2 One approach to the
optimization of the activity of an SSRI is the derivatization of
an indole core through incorporation of a conformationally
* To whom correspondence should be addressed. Telephone: 732-227-6863.
Fax: 732-227-3932. E-mail:chien.chen@bms.com.
(1) (a) Nagatomo, T.; Rashid, M.; Muntasir, H. A.; Komiyama, T.
Pharmacol. Ther. 2004, 104, 59. (b) Middlemiss, D. N.; Price, G. W.;
Watson, J. M. Curr. Opin. Pharmacol. 2002, 2, 18.
Results and Discussions
The original synthesis3h,4 of 1 (Scheme 1) by our Discovery
colleagues provided the first 40 g required for early toxicology
studies. This route involved cyclopropanation of a chiral
enamide through the use of an expensive chiral auxiliary,
employed hazardous reagents including diazomethane, sodium
hydride, and lithium aluminum hydride, whose use at scale is
preferred to be limited, and required chromatographic purifica-
tion of several intermediates.
(2) (a) Currently available drugs in this class include Citalopram,
Escitalopram,Fluoxetine,Fluvoxamine,Paroxetine,andSertraline.Malinin,
A.; Oshrine, B.; Serebruany, V. Med. Hypothesis 2004, 63, 103. (b)
Chen, J.; Mabjeesh, H.; Greenstein, A. Urology 2003, 61, 197. (c)
Dimmock, P. W.; Wyatt, K. M.; Jones, P. W.; Shaughn O’Brien, P. M.
Lancet 2000, 356, 1131. (d) Moore, K. Int. J. Gynecol. Obstet. 2004,
86 (1), S53. (e) Pogarell, O.; Poepperl, G.; Mulert, C.; Hamann, C.;
Sadowsky, N.; Riedel, M.; Moeller, H.-J.; Hegerl, U.; Tatsch, K Eur.
Neuropsychopharmacol. 2005, 15, 521–524.
Scheme 1. Discovery synthesis of 1a
(3) (a) King, H.; DMeng, Z.; Denhart, D.; Mattson, R.; Kimura, R.;
Dedong, W.; Gao, Q.; Macor, J. E Org. Lett. 2005, 7, 3437. (b) Taylor,
E. W.; Nikam, S.; Weck, B.; Martin, A.; Nelson, D. Life Sci. 1987,
41, 1961. (c) Vangveravong, S.; Kanthasamy, A.; Lucaites, V. L.;
Nelson, D. L.; Nichols, D. E. J. Med. Chem. 1998, 41, 4995. (d)
Ezquerra, J.; Pedregal, C.; Lamas, C.; Pastor, A.; Alvarez, P.; Vaquero,
J. J. Tetrahedron Lett. 1996, 37, 683. (e) Macor, J. E.; Blank, D. H.;
Post, R. J. Tetrahedron Lett. 1994, 35, 45. (f) Macor, J. E.; Blank,
D. H.; Ryan, K.; Post, R. J. Synthesis 1997, 443. (g) Schmitz, W. D.;
Denhart, D. J.; Brenner, A. B.; Ditta, J. L.; Mattson, R. J.; Mattson,
G. K.; Molski, T. F.; Macor, J. E. Bioorg. Med. Chem. Lett. 2005, 15,
1619. (h) Mattson, R. J.; Catt, J. D.; Denhart, D. J.; Deskus, J. A.;
Ditta, J. L.; Higgins, M. A.; Marcin, L. R.; Sloan, C. P.; Beno, B. R.;
Gao, Q.; Cunningham, M. A.; Mattson, G. K.; Molski, T. F.; Taber,
M. T.; Lodge, N. J. J. Med. Chem. 2005, 48, 6023.
(4) (a) Mattson, R. J.; Denhart, D. J.; Deskus, J. A.; Ditta, J. L.; Marcin,
L. R.; Epperson, J. R.; Catt, J. D.; King, D.; Higgins, M. A. (Bristol-
Myers Squibb) Cyclopropylindole derivatives as selective serotonin
reuptake inhibitors, U.S. patent US6777437 B2, Aug 17, 2004. (b)
Mattson, R. J. (Bristol-Myers Squibb) Cyclopropylindole derivatives
as selective serotonin reuptake inhibitors, U.S. patent US6822100 B2,
Nov 23, 2004. (c) Mattson, R. J.; Denhart, D. J.; Deskus, J. A.; Ditta,
J. L.; Marcin, L. R.; Epperson, J. R.; Catt, J. D.; King, D.; Higgins,
M. A. (Bristol-Myers Squibb) Cyclopropylindole derivatives as
selective serotonin reuptake inhibitors, PCT Int. Appl. WO 0279152,
2002.
a Reagents and conditions: (a) POCl3, DMF (75%); (b) TsCl, Et3N (65%);
(c) NaH, diethyl 2-(N-(1R)-(+)-2,10-camphorsultam)-2-oxoethylphosphonate
(58%); (d) CH2N2 (62%); (e) LAH, THF (81%); (f) (COCl)2, DMSO; (g) HNMe2,
NaBH(OAc)3; (h) NaOH, EtOH (77%, 3 steps).
168
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development
10.1021/op700125z CCC: $40.75
2008 American Chemical Society
Published on Web 01/17/2008

Scheme 2. Synthesis of vinyl indole 2
Drawing on prior Process group experience,5 an alternative
route to 1a involving catalytic asymmetric cyclopropanation of
vinyl indole 2 with ethyl diazoacetate was considered (Figure
1). The feasibility of this approach was demonstrated by our
Discovery colleagues.6 In this paper we present details of the
development and plant-scale implementation of this route, which
enabled production of kilogram quantities of 1a.
with both 1 M methylmagnesium bromide and chloride in THF.
Methylmagnesium iodide gave low conversion and resulted in
poor product quality. Reactions in dimethoxyethane gave less
than 5% conversion. Optimized conditions for the production
of alcohol 6 entailed the addition of methylmagnesium chloride
in THF (1.1 equiv) to a solution of 5 at 15–30 °C. No extractive
workup was needed as addition of dilute hydrochloric acid to
the reaction medium (total volume 25 mL/g) gave rise to a
coarse solid that converted into fine crystalline material in 1 h
with stirring. More extensive aging of the slurry was of further
benefit, as this led to lower levels of residual magnesium in
isolated product 6. On kilogram scale (6.4–6.5 kg) the yield
ranged between 91% and 95% (95.5–97.6 HPLC area % purity).
Indolyl alcohol 6 underwent dehydration to vinyl indole 2
in the presence of catalytic p-toluenesulfonic acid upon warming
in toluene.9 A competing side reaction, however, led to
Preparation of the Vinyl Indole Cyclopropanation Sub-
strate. In developing a scalable route to 1a, the Discovery
synthesis of vinyl indole 26 was modified to four direct-drop
processes (Scheme 2).7 Initial development efforts toward a
scalable preparation of 3-formyl-5-cyanoindole (4b) showed that
acetonitrile was the most favorable solvent from the standpoint
of operability (facile stirring) and acceptable product quality.
Addition of 3 to a mixture of commercially prepared Vilsmeier
reagent and acetonitrile (7 mL/g)8 was exothermic (65.27 kJ/
kg solution), but the heat evolution could be controlled by
adjusting the addition rate. Upon reaction completion, water
(16–17 mL/g) was added, and warming to 70 °C for 1 h
facilitated hydrolysis of the iminium salt 4a. During cooling,
aldehyde 4b crystallized spontaneously and was isolated in high
yield with only minimal loss (4%) to the mother liquor. Upon
scale-up, this procedure gave excellent yield, reproducibility,
and purity (5.0 kg, 95% yield, 99.2–99.7 HPLC area % purity).
After a thorough solvent screen, acetonitrile was chosen for
tosylation of indole 4b. Addition of diisopropylethylamine (i-
Pr2NEt) to an acetonitrile slurry of 4b and p-TsCl led to a mild,
exothermic reaction (temperature increase from 23 to 30 °C)
and dissolution of solids. As the reaction proceeded, 5 precipi-
tated as a crystalline solid. Addition of water (5 mL/g, final
concentration 17 mL/g) led to increased recovery to 97–98%
yield (>99 HPLC area % purity) on laboratory scale. This
procedure successfully produced N-tosyl indole 5 on ap-
proximately 3.5 kg scale in 96–99% yield (99.3–99.5 HPLC
area % purity).
1
significant levels of a dimeric impurity. H NMR analysis
suggested the impurity to be 7 (Figure 2), the product of reaction
between vinyl indole and a putative intermediate carbocation.
In looking for an alternate dehydration process, we found that
heating the trifluoroacetyl derivative of 6 with excess i-Pr2NEt
also gave facile conversion to vinyl indole 2. Although this
reaction proceeded well in a variety of solvents, a new dimeric
impurity was observed. 1H NMR evidence suggested the new
dimer to be one or more stereoisomeric Diels–Alder adducts
(8, Figure 2). Selection of suitable workup conditions (vide
infra) minimized formation of these side products.
Figure 2. Impurities from the dehydration of vinyl alcohol 6.
Commercially available methyl Grignard reagents were
evaluated for their ability to effect the transformation of
aldehyde 5 to indolyl alcohol 6. High conversions were observed
As would be expected, the rate of dimer formation increased
with increasing product concentration. Timely analytical as-
sessment and the ability to effect rapid temperature changes
on scale became important. To minimize dimer formation,
kilogram-scale preparation of vinyl indole 2 entailed controlled
addition of trifluoroacetic anhydride (TFAA) to a solution of 6
and i-Pr2NEt in DMF (5 L/kg) while maintaining the batch
temperature below 25 °C. React-IR monitoring experiments
(5) Similar disconnection was employed during multi-kilogram scale
preparation of a melatonin antagonist. Simpson, J. H.; Godfrey, J.;
Fox, R.; Kotnis, A.; Kacsur, D. J.; Hamm, J.; Totleben, M.; Rosso,
V.; Mueller, R.; Delaney, E.; Deshpande, R. P. Tetrahedron Asymmetry
2003, 14, 3569.
(6) Marcin, L. R.; Denhart, D. J.; Mattson, R. J. Org. Lett. 2005, 7, 2651.
(7) The Drug Discovery conversion of aldehyde 5 to vinyl indole 2
involved Wittig olefination. Although the Wittig chemistry was
effective, two-step methyl Grignard addition followed by dehydration
was found to be operationally simpler.
(9) Sundberg, R. J.; Amat, M.; Fernando, A. M. J. Org. Chem. 1987, 52
(8) mL/g refers to milliliters of solvent per grams of substrate.
(14), 3151.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
169

performed prior to scale-up demonstrated reaction of TFAA
with alcohol 6 was nearly instantaneous. The solution subse-
quently was heated to 96–99 °C to effect elimination. Upon
completion,10 water was added to quickly lower the batch
temperature and thus suppressed the competing dimerization
process. Incorporation of ethanol during cooling enabled a
controlled crystallization. Further cooling to 0 °C allowed for
isolation of vinyl indole 2 in good yield containing less than
2% of 8. Glass plant implementation of this process was
demonstrated on approximately 7.7 kg scale (85% yield; 97.8
HPLC area % purity).
gave a rich purple solution similar to that seen when using
isolated Nishiyama catalyst. Addition of a 2 M solution of
freshly prepared EDA in toluene13 to the catalyst/substrate
solution at 55–65 °C gave predominantly trans-cyclopropane
9a (dr ) 10:1, er ) 93:7). Although the reaction could be
performed in a variety of solvents, with the notable exception
of acetonitrile,14 THF was selected because of favorable
azeotropic properties. Although higher temperatures led to faster
reaction rates with minimal effect on the enantioselectivity,
concerns over the high-temperature stability of EDA15 prompted
us to set the operating temperature at 55–65 °C.
Cyclopropanation and Hydrolysis. Prior successful large-
scale cyclopropanation of a styrene derivative5 using Nishiya-
ma’s catalyst11 and ethyldiazoacetate (EDA) suggested this
system also would be suitable for cyclopropanation of 2
(Scheme 3).12 Good enantio- and diasteroselectivities for
During scale-up, a reaction sample was analyzed soon after
the EDA addition had begun to ensure the in situ generated
catalyst was active. This procedure was followed to help prevent
a build-up of EDA in the reactor in the event the catalyst
preparation failed. With the assurance that EDA was being
consumed, the remaining portion was then added over 8 h.
Cyclopropyl ester 9a was produced as a 10:1 mixture of trans/
cis-cyclopropane isomers, in agreement with development runs.
Chiral HPLC analysis of the crude mixture indicated an
enantiomeric excess (ee) of 86% for the major trans-isomer
9a.
Scheme 3. Asymmetric cyclopropanation and
chemoselective ester hydrolysis
While the cyclopropanation proceeded well and with good
selectivity, the resulting cyclopropylester 9a proved difficult to
crystallize without chromatographic removal of the catalyst. All
attempts to isolate 9a from the reaction mixture gave poor yields
and highly colored material. Owing to the difficulty in procuring
clean ester 9a and on the basis of our prior experience with
related chemistry,5,11d,13f we sought to develop a process wherein
cyclopropanation followed by ester hydrolysis would give acid
10a directly. A consequence of this work was the discovery
that hydrolysis of the desired trans-ester under basic conditions
was faster than hydrolysis of the undesired minor cis-ester.
However, this process was complicated by competing hydration
of the cyano group, as well as detosylation. Phase transfer
cyclopropanation of 2 with the Nishiyama catalyst (enantiomeric
ratio (er) ) 93:7, diastereomeric ratio (dr) ) 9.2:1)6 and our
group’s prior experience in preparing and using commercially
available (R)-pybox ligand on kilogram scale,5 coupled with
the need to advance the active pharmaceutical ingredient (API)
quickly to enable toxicity studies, encouraged us to focus
development work on this catalytic system in lieu of initiating
a catalyst screen.While initial development work was performed
with isolated Nishiyama catalyst, we also investigated in situ
catalyst generation. This approach allowed us to circumvent
the use of ethylene gas required for isolation of the stable
catalyst and also improved cycle times. In addition, the in situ
method avoided loss of catalyst to mother liquors. As such,
combination of (R)-Pybox and dichloro(p-cymene)ruthenium(II)
dimer in THF followed by addition of 2 and heating to 55 °C
(13) (a) Kotnis, A. S.; Simpson, J. H.; Deshpande, R. P.; Kacsur, D. J.;
Hamm, J.; Kodersha, G.; Merkl,W.; Domina, D.; Wang, S. S. Y.
Manuscript in preparation. (b) Clark, J. D.; Shah, A. S.; Peterson, J. C.;
Patelis, L.; Kersten, R. J.; Heemskerk, A. H. Thermochim. Acta 2002,
386, 73. (c) Clark, J. D.; Shah, A. S.; Peterson, J. C.; Patelis, L.;
Kersten, R. J.; Heemskerk, A. H.; Grogan, M.; Camden, S. Thermo-
chim. Acta 2002, 386, 65. (d) Armstrong, R. K. J. Org. Chem. 1996,
31, 618. (e) Searle, N. E. Organic Syntheses; Wiley: New York, 1963;
Collect. Vol. IV, p 424. (f) DelMonte, A. J.; Dowdy, E. D.; Watson,
D. J. Development of Transition Metal-Mediated Cyclopropanation
Reactions. In Topics in Organometallic Chemistry; Larsen, R. D., Ed.;
Springer: Berlin, New York, 2004; Vol. 6, Organometallics in Process
Chemistry; pp 97-122.
(10) A 4 min HPLC method was developed to minimize exposure of the
product to the reaction conditions.
(11) (a) Nishiyama, H.; Itoh, Y.; Matsumoto, H.; Park, S.-B. J. Am. Chem.
Soc. 1994, 116, 2223. (b) Park, S.-B.; Murata, K.; Matsumoto, H.;
Nishiyama, H. Tetrahedron Asymmetry 1995, 6, 2487. (c) Nishiyama,
H.; Itoh, Y.; Sugawara, Y.; Matsumoto, H.; Aoki, K.; Itoh, K. Bull.
Chem. Soc. Jpn. 1995, 68, 1247. (d) Totleben, M. J.; Prasad, J. S.;
Simpson, J. H.; Chan, S. H.; Vanyo, D. J.; Kuehner, D. E.; Deshpande,
R.; Kodersha, G. A. J. Org. Chem. 2001, 66, 1057.
(14) Results from the solvent screen at 30 mL/g with the slow addition
(16 h) of 2.2 M EDA intoluene are as follows: in THF (50–60 °C),
86–89% ee with 9.7–10.8:1 trans:cis ratio with 91–99% conversion;
in ethyl acetate (50 °C), 87.1% ee with 9.7:1 trans:cis ratio with 98.6%
conversion; in toluene (50 °C), 88.2% ee with 11.4:1 trans:cis ratio
and 95.9% conversion; in acetone (reflux), 82.6% ee with 10.7:1 trans:
cis ratio and 87.8% conversion; in CH2Cl2 (reflux), 92.4% ee with
13.1:1 trans:cis ratio and 56.5% conversion; in MTBE (50 °C), 87.6%
ee with 12.9:1 trans:cis ratio and >99% conversion; in methylisobu-
tylketone (50 °C), 86.3% ee with 14.7:1 trans:cis ratio and 82.7%
conversion; in butylacetate (50 °C), 85.6% ee with 11.6:1 trans:cis
ratio and 98.7% conversion; in triflurotouluene (50 °C), 86.3% ee
(trans:cis ratio ND) and 67.9% conversion; in acetonitrile no conver-
sion was achieved, most likely due to complexation of acetonitrile
with ruthenium (See: Cadierno, V.; Gamasa, M. P.; Gimeno, J.;
Inglesias, L.; Garcia-Granda, S. Inorg. Chem. 1999, 38, 2874.
(15) Clark, J. D.; Shah, A. S.; Peterson, J. C. Thermochim. Acta 2002,
177, 392–393.
(12) (a) Subsequent to the work described here, significant advances in
the field of cyclopropanation chemistry were reported by many research
groups. For representative examples, see: Lebel, H.; Marcoux, J.-F.;
Molinaro, C.; Charette, A. B Chem. ReV. 2003, 103, 977. (b) Liao,
W.-W.; Li, K.; Tang, Y J. Am. Chem. Soc. 2003, 125, 13030. (c) Long,
J.; Yuan, Y.; Shi, Y. J. Am. Chem. Soc. 2003, 125, 13632. (d)
Aggarwal, V. K.; Fang, G. Y.; Meek, G. Org. Lett. 2003, 5, 4417. (e)
Risatti, C. A.; Taylor, R. E. Angew. Chem., Int. Ed. 2004, 43, 6671.
(f) Moreau, B.; Charette, A. J. Am. Chem. Soc. 2005, 127, 18014. (g)
Johansson, M. J.; Gorin, D. J.; Staben, S. T.; Toste, F. D. J. Am. Chem.
Soc. 2005, 127, 18002. (h) Bykowski, B.; Wu, K.-H.; Doyle, M. P.
J. Am. Chem. Soc. 2006, 128, 16038. (i) Thompson, J. L.; Davies,
H. M. L. J. Am. Chem. Soc. 2007, 129, 6090. (j) Xie, H.; Zu, L.; Li,
H.; Wang, J.; Wang, W. J. Am. Chem. Soc. 2007, 129, 10886.
170
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

catalysis only accentuated the rates of these side reactions.
Typical yields of 10a using these protocols ranged from 30%
to 50%.
% purity). Filtration of an acetonitrile solution of 10a through
a 10 µm filter prior to addition of (R)-(+)-R-methylbenzylamine
proved beneficial by lowering residual ruthenium from the
cyclopropanation to <310 ppm (initial value >3100 ppm). The
resolution was performed in two batches (6.4 and 9.4 kg),
providing salt 10b in 56–64% yield, 97.8–98.6 HPLC area %
purity, and 95–96% de.
Alternatively, hydrolysis under acidic conditions at high
temperature gave a much cleaner reaction profile with no
detosylation and minimal hydration of the cyano group.
Significantly, the trans-isomer still hydrolyzed 3–4 times faster
than the cis-ester. This selectivity, as well as the difference in
solubility between the cis-ester and trans-acid ultimately led to
an efficient process for separating the undesired cis-ester from
the product. After screening various acids, triflic acid was
identified as the most efficient for catalyzing the hydrolysis.
Crystallization development work led to the identification of
aqueous acetonitrile for isolation of 10a. Additionally, toluene
was found to inhibit the hydrolysis and lower the recovery of
acid 10a. Thus, upon completion of the cyclopropanation, water
(3 mL/g) was added, and most of the toluene and THF were
removed by azeotropic distillation. The displaced volume was
replaced with acetonitrile. Upon completion of the partial solvent
exchange, hydrolysis was then conducted with triflic acid (2.5
equiv) in acetonitrile/water (3–5:1 v/v). In order to drive the
hydrolysis to completion, the byproduct ethyl alcohol was
removed continuously by distillation. Acetonitrile and water
were added throughout the distillation to maintain a constant
volume. This protocol successfully drove the hydrolysis reaction
to 97–98% completion. Acid 10a was crystallized by the
introduction of additional water and was isolated by simple
filtration.
Reduction of Acid 10 to Alcohol 11. Direct reduction of
the R-methylbenzylamine salt 10b with borane-THF complex
(BH3 ·THF) led to unacceptable levels of nitrile reduction
products. Reduction of the free acid 10a, on the other hand,
was highly selective when carried out at sufficiently low
temperature (below 10 °C). Solutions of free acid 10a could
be produced by suspending salt 10b in tert-butyl methyl ether
and THF, washing with sulfuric acid, and azeotropic drying.
The liberated acid solution was then filtered through a 5 µm
filter to remove insoluble ruthenium residues prior to reduction.
Reaction of solutions of 10a with BH3 ·THF (1.15 equiv, 3.45
hydride equiv versus 3.00 theoretical requirement) typically gave
high conversion with less than 2% competing nitrile reduction.
The amine byproduct stemming from nitrile reduction was
purged to a great extent through precipitation of the alcohol 11
from solvent systems containing dilute hydrochloric acid.
Ultimately the alcohol was obtained in 96% yield with greater
than 98 HPLC area % purity.
Calorimetry studies showed two distinct exothermic events
during the reaction sequence (Figure 3).17 The first, associated
with acid deprotonation, was easily controlled by modulation
of the addition rate of the reagent. Nearly all of the hydrogen
offgassing was observed during this initial phase. The second
exotherm, associated with reduction of the carboxylate, showed
an induction period and peak heat flow extending beyond the
completion of the reagent addition. This had significant safety
ramifications and required careful consideration before initiating
large scale processing.
Scale-up of the cyclopropanation/hydrolysis was conducted
in two separate batches (7.6 and 7.8 kg), providing acid 10a in
crude yields of 93% and 100% (91.0 and 83.8 HPLC area %
purity, 85.4% and 87.6% ee, respectively). The ratio of trans-
acid 10a to cis-ester 9b increased from 10:1 in the reaction
mass to 40–150:1 in isolated acid 10a.
Partial Resolution of Acid 10a. The intermediacy of acid
10a provided an opportunity to enhance the enantiopurity
through selective crystallization of a diastereomeric salt (Scheme
4). Using the bottom-up strategy and parallel crystallization
Scheme 4. Augmentation of chiral purity via resolution of
10a
techniques,16 (R)-(+)-R-methylbenzylamine was identified as
the preferred base, as it gave salt 10b in high analytical purity
(>98 HPLC area % purity) and diastereomeric purity (95–96%
de). Various solvent systems were evaluated, with acetonitrile
offering the best combination of operability (efficient stirring),
yield (82–84%), analytical purity (>98 HPLC area % purity),
and de (95–96%). Isopropyl alcohol yielded 10b having good
de (96%) but poor operability, whereas ethyl and n-butyl
acetates produced 10b of lower de (94%), and THF/heptane
gave good de (97%) but lower overall purity (96 HPLC area
Figure 3. Strong secondary exotherm attributable to carboxy-
late reduction. * Qr ) Heat of reaction.
Efforts to conduct the reaction at elevated temperature and
push the reduction kinetics into an addition-controlled regime
were unsuccessful as a result of decreased selectivity for
carboxylate versus nitrile reduction. The low-temperature
(17) For a detailed analysis of a similar system see: Org. Process Res.
DeV. 2004, 6, 1072–1075.
(16) Chen, C.-K.; Singh, A. A. Org. Process Res. DeV. 2001, 5, 508.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
171

procedure adopted on scale involved a combination of modula-
tion of the borane addition rate and active cooling to maintain
the batch temperature below 10 °C throughout the reduction
sequence. The modulated borane addition permitted careful
reaction temperature monitoring and the opportunity to im-
mediately halt the dosing in the event an unexpected temperature
increase occurred. Additional safety measures implemented in
the glass plant included the low-temperature transport and
storage of BH3 ·THF solution,18as well as oxygen monitoring
in the reactor headspace. The procedure was conducted on two
batches (5.4 and 6.4 kg) and provided alcohol 11 (Scheme 5)
in 90–96% yield, 98.3 HPLC area % purity, and 96% ee. The
ruthenium levels for both batches were below 85 ppm.
water/CH2Cl2, or water/methyl acetate. Subsequent addition of
dimethylamine rapidly afforded the corresponding iminium
species, and reduction (1.3 equiv of sodium triacetoxyborohy-
dride) typically was complete within 30 min.
Sodium hypochlorite lots examined during laboratory de-
velopment varied widely in potency. Due to this variability,
the titre of the material to be used during the campaign was
determined immediately prior to use.20
The primary impurity formed in the oxidation/reductive
amination sequence was carboxylic acid 10a, which could be
minimized by operating under oxygen-free conditions.19b If
precautions to eliminate oxygen were not taken, the amount of
acid 10a formed was as high as 10%. Upon scale-up, the
amount of 10a formed was less than 1%. A slow addition rate
of the bleach also was crucial, as more rapid addition led to
the competing formation of unidentifed side-products.
In the reductive amination, development work indicated that
dichloromethane or methyl acetate performed well. Methyl
acetate ultimately was selected as dichloromethane led to
alkylation of the amine. Methanol initially was selected as a
suitable solvent for crystallization of the free-base 14a; however,
the crystalline solid obtained was a solvate, which melted readily
upon drying under vacuum, even at 20–25 °C. To overcome
this problem, a study to identify a suitable salt of 14a was
conducted. This work revealed that HCl salt 14b could be
crystallized from 2-propanol in high recovery. Following this
lead, solvent exchange to 2-propanol after workup of the
reductive amination, followed by addition of a solution of HCl
in 2-propanol (formed by adding acetyl chloride to 2-propanol)
produced crystalline 14b in high yield (88%) and purity (99.8
HPLC area % purity). This process also was effective in further
reducing the ruthenium content of the batch from 80 ppm in
11 to 30–40 ppm in 14b. Such ruthenium levels, however, were
not acceptable for the final intermediate. To address this issue,
we tested various ZetaCarbon pads21 and Smopex resins.22
ZetaR53SP was found to be the most effective for lowering
residual ruthenium levels (80 to 7 ppm), while also minimizing
product adsorption to the filter media (3%).
Scheme 5. Borane reduction of carboxylic acid 10
Transformation of Alcohol 11 to Amine 14. Early work
dedicated to conversion of alcohol 11 to final intermediate 14
entailed transformation of the hydroxyl group in 11 to a leaving
group, followed by substitution by dimethylamine (Scheme 6).
Analogs possessing a variety of leaving groups including
mesylate, tosylate, trifluoroacetate, and phosphate (12a–12d)
were screened. Of these, only the mesylate could be prepared
reproducibly, although it was prone to decomposition. In
contrast, the chloride 12e could be prepared cleanly using
Vilsmeier reagent in DMF. Chloride displacement by dimethy-
lamine, however, was slow. Exposure to >12 equiv of dimeth-
ylamine at 60 °C for 40 h gave incomplete conversion. Addition
of sodium iodide led to complete conversion in 20–24 h at 60
°C; however, amine 14 proved difficult to isolate under these
reaction conditions. As a result, alternatives to the displacement
chemistry were investigated.
Scheme 6. Conversion of alcohol 11 to amine 14
The process for converting alcohol 11 to hydrochloride salt
14b was demonstrated on 3.5 and 4.5 kg scale (75 and 88%
yield; 99.7–99.8 HPLC area % purity). The Zeta filtration
method also scaled up effectively, as the final ruthenium content
was <10 ppm (initial values 30 and 83 ppm).
Detosylation/Maleate Salt Formation. The procedure used
by the Discovery team3h,4 for conversion of final intermediate
14a to 1a involved detosylation with sodium hydroxide, an
extractive workup and wash sequence, and maleate salt forma-
tion. Development team goals for this process included reducing
the number of operations and enhancing the enantiomeric excess
(95.6% ee in the final intermediate 14).
Work on detosylation of final intermediate 14b and conver-
sion to the maleate salt 1a began with investigation of
nucleophile/solvent combinations. Using parallel screening
Free base 14a had been prepared previously by our Dis-
covery colleagues3h,4 via Swern oxidation of 11 followed by
reductive amination. As an alternative to Swern conditions,
oxidation with TEMPO19a was explored. Optimal conditions
consisted of combining 11 with sodium bicarbonate (1.1 equiv),
TEMPO (0.025 equiv), and NaOCl (1.2 equiv) in water/THF,
(20) Jeffery, G. H.; Bassett, J.; Mendham, J.; Denney, R. C. Vogel’s
Textbook of QuantitatiVe Chemical Analysis, 5th ed.; Prentice Hall:
Singapore, 1989; pp 396–397..
(18) Chem. Eng. News 2002, 80, 7.
(19) (a) Anelli, P. L.; Biffi, F. M.; Quici, S. J. Org. Chem. 1987, 52, 2559.
(b) de Nooy, A. E. J.; Besemer, A. C.; van Bekkum, H. Tetrahedron
1995, 51, 8023.
(21) http://www.cuno.com/labstore/dfs_zetacarbon.shtml.
(22) http://www.jmcatalysts.com/pct/marketshome.asp?marketid)69&id)
250.
172
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

techniques, compound 14b was subjected to a wide range of
nucleophiles (metal alkoxides, silanolates, carbonates, tetrabu-
tylammonium fluoride, and others), in an array of protic,
ethereal, and polar aprotic solvents. Notably, conditions utilizing
THF as solvent or cosolvent often led to clean detosylation.
The screen identified metal alkoxides and silanolates as effective
reagents for rapid detosylation (1–2 h, 40–60 °C) in several
solvents including alcohols, THF, and combinations thereof.
Investigations later focused on alcohol-free systems, in order
to eliminate the possibility of forming the toxic alkyl tosylate
byproduct.
This work revealed that treatment of 14b with 10 N NaOH
in THF in various volume ratios led to complete detosylation
in 4 h at 65 °C. Final conditions called for use of 50 wt %
NaOH (2.5 mL/g 14b) in THF (3.0 mL/g 14b) at 65–68 °C for
3–4 h. Detosylated free base 1 could be separated efficiently
from the byproduct sodium tosylate via phase split, after dilution
of the reaction mass with water, THF, and MTBE (volume ratio
7.5:1:3.3). Typical product loss to the aqueous portion was
<6%, and residual sodium tosylate in the organic phase was
<2%.
Higher yields of 1a were obtained through crystallization
from a dry organic phase having Karl Fischer (KF) values of
<0.35% (w/w). KF values taken of the organic phase im-
mediately after phase separation ranged from 3.3% to 3.7% (w/
w). To achieve the desired KF value, azeotropic removal of
water was effected by using dry 1:1 THF/MTBE mixtures and
concentrating to approximately 2.5 mL/g of input (14b) until
the KF was <0.8% (w/w). Upon meeting the KF criterion, the
concentrated batch was charged with THF (KF after charge
<0.35%) to a volume corresponding to 7 mL/g of input (14b).
Conveniently, as the drying process progressed, any remaining
sodium tosylate precipitated from the mixture and could be
removed efficiently by filtration. Addition of a solution of maleic
acid (1.1 equiv) in THF afforded a slurry of 1a which filtered
well. Two THF/MTBE (1:1 v/v) rinses provided maleate salt
1a in >99.5 HPLC area % purity and >98% ee.
Two kilogram-scale reactions (1.8 and 1.7 kg 14b) were
conducted in a 40 L Hastelloy vessel, compatible with hot 50
wt % NaOH. Upon reaction completion, the thick reaction mass
was diluted with water (9 L/kg), THF (1.1 L/kg), and MTBE
(2.9 L/kg) and transferred to a borosilicate 50 L vessel. Phase
separation, distillative drying, filtration through 10 and 0.5 µm
polish filters in series, and maleic acid addition/crystallization
proceeded similarly in the first kilogram batch as it did in the
laboratory with four significant exceptions. First, the final
MTBE concentration during crystallization was higher than in
typical laboratory runs (12% versus 6% by GC analysis).
Additionally, aspects relating to product quality including
particle size and enantiopurity were different from those of
development runs. The particle size in the first batch was larger
than observed in laboratory runs (D90 ) 264 µm, as opposed
to 31 µm), and the ee was lower than expected (97% versus
>98% in laboratory runs). Finally, during the first batch, rapid,
uncontrolled crystallization occurred 2 h subsequent to comple-
tion of the maleic acid addition, whereas in laboratory runs
crystallization consistently occurred during the maleic acid
addition. This run generated 1.24 kg of 1b (83.3% yield) with
99.1 HPLC area % purity.
The deviations in the first kilogram batch from the laboratory
runs were addressed during a second batch through modification
of the crystallization protocol. To better control the final MTBE
content, the distillation end points were modified to depend on
internal batch volume instead of distillate volume. Additionally,
a seeding protocol was incorporated to control the onset of
crystallization and to improve particle size and ee. After
implementing these modifications, the final MTBE content of
the slurry was 5.7% and the isolated product had a D90 value
of 41 µm. This second batch produced 1.1 kg of 1b (77% yield)
with 99.5 HPLC area % purity and 98% ee.
HCl Salt (1b) Development. Subsequent to the delivery
described above, an alternate form of API 1 possessing different
pharmacokinetic properties was desired. After extensive screen-
ing, the HCl salt (1b) was identified. Paramount to the efficient
crystallization of the HCl salt was the method of its preparation.
HCl introduced as an aqueous solution produced a hygroscopic
foam. Alternatively, addition of anhydrous HCl in MTBE
(prepared by reaction of oxalyl chloride with 1 equiv of water
or acetyl chloride and 1 equiv of methanol) to an MTBE
solution of free base generated a fluffy white solid. Upon
prolonged stirring this amorphous suspension transformed into
crystalline material. As crystalline material had now been
isolated, all future experiments could be seeded, avoiding the
slow transition from amorphous to crystalline material. HCl
prepared from acetyl chloride and ethanol performed well under
these conditions and was used for all subsequent experiments.
We anticipated performing the detosylation of 14b as it was
previously conducted and modifying the workup and crystal-
lization to prepare the HCl salt. Crystallization development
indicated MTBE was a suitable solvent, providing good product
recovery and augmentation of the enantiomeric excess. Further
examination indicated higher ee values could be obtained by
incorporation of ethanol in the crystallization system. Use of
3–8 mL ethanol per gram of 14b was critical, as smaller charges
resulted in gummy material and larger charges resulted in
unacceptable product losses to the mother liquor. To provide
additional control over the crystallization, seeds were added to
the batch prior to adding the HCl.
Two batches (1.8 kg each) of 14b were converted success-
fully to HCl salt 1b using this protocol. The phase split was
conducted as before, but the distillation procedure was changed
in that only MTBE (as opposed to 1:1 THF/MTBE) was used
to replenish the volume distilled. Polish filtration to remove
residual sodium tosylate provided a clear MTBE solution. To
this was added a slurry of seeds (1b, 6–8 g) in MTBE, followed
by a solution of HCl in ethanol, prepared by the addition of
acetyl chloride to ethanol at <10 °C. To ensure a controlled
addition of HCl to the free base, a metering pump calibrated at
<50 mL/min was used for the addition of approximately 8 L
of HCl solution. Crystallization occurred gradually after about
20% of the HCl was charged. After overnight stirring, salt 1b
was collected by filtration under nitrogen. The two batches
provided 1b in 88–89% yield, with 99.9 HPLC area % purity
and 98.5–98.6% ee.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
173

Conclusions
In summary, we have described an efficient (9 isolations),
(2 × 9.3 L). The material was dried in vacuo at 55–60 °C until
constant weight to give 6.46 kg (98.5% yield) of tosylate 5.
Mp: 244.6–246.0 °C. 1H NMR (d6-DMSO, 400 MHz) δ 10.08
(s, 1H), 9.06 (s, 1H), 8.48 (d, J ) 1.0 Hz, 1H), 8.15 (d, J ) 8.1
Hz, 1H), 8.05 (d, J ) 8.6 Hz, 2H), 7.86 (dd, J ) 8.6 and 1.5
Hz, 1H), 7.47 (d, J ) 8.1 Hz, 2H), 2.35 (s, 3H); 13C NMR
(d6-DMSO, 100 MHz) δ 199.09 (C), 186.52 (CH), 147.03 (C),
139.86 (CH), 135.99 (C), 132.85 (C), 130.75 (CH), 129.36
(CH), 127.41 (CH), 125.86 (CH), 120.66 (CH), 118.59 (C),
114.52 (CH), 107.81 (C), 21.09 (CH3). IR 3121.3, 2229.7,
1677.6, 1382.7, 1172.7 cm-1. Anal. Calcd for C17H12N2O3S:
C, 62.95; H, 3.72; N, 8.63; S, 9.88. Found: C, 62.94; H, 3.69;
N, 8.64; S, 9.69.
Preparation of 3-(1-Hydroxy-vinyl)-1-(toluene-4-sulfo-
nyl)-1H-indole-5-carbonitrile (6). To a stirred suspension of
aldehyde 5 (6.35 kg, 19.6 mol) in THF (22.3 L) at 0–5 °C was
added 3 M methylmagnesium chloride in THF (7.05 L, 21.2
mol) over 50 min while maintaining the temperature <15 °C.
After a THF rinse (3.1 L) the solution was stirred at 0–5 °C
until completion of the reaction by HPLC. The reaction was
quenched by addition of 0.3 N aqueous HCl (78.2 L) over 45
min while maintaining the temperature <10 °C. The resulting
slurry was stirred for 60 min at 15–25 °C and the product was
collected by filtration. The cake was washed with water (2 ×
19.8 L) and dried in vacuo at 55–60 °C until constant weight
to give 6.35 kg (95.2% yield) of alcohol 6. Mp: 155.6 °C. 1H
NMR (d6-DMSO, 400 MHz) δ 8.23 (d, J ) 1.5 Hz, 1H), 8.08
(d, J ) 8.6 Hz, 1H), 7.91 (d, J ) 8.1 Hz, 2H), 7.75 (d, J ) 1.0
Hz, 1H), 7.72 (dd, J ) 8.6 and 1.5 Hz, 1H), 7.40 (d, J ) 8.6
Hz, 2H), 5.33 (d, J ) 5.1 Hz, 1H), 4.97 (qt, 1H), 2.32 (s, 3H),
1.44 (d, J ) 6.6 Hz, 3H); 13C NMR (d6-DMSO, 100 MHz) δ
172.6 (C), 158.1 (C), 146.0 (C), 136.6 (C), 133.8 (C), 130.4
(CH), 128.9 (C), 128.2 (C), 127.6 (CH), 126.8 (CH), 126.1
(CH), 124.0 (CH), 119.1 (C), 114.2 (CH), 105.6 (C), 81.5 (C),
61.5 (CH), 23.8 (CH3), 21.0 (CH3). Anal. Calcd for
C18H16N2O3S: C, 63.51; H, 4.73; N, 8.23; S, 9.42. Found: C,
63.45; H, 4.75; N, 8.19; S, 9.27.
Preparation of 1-(Toluene-4-sulfonyl)-3-vinyl-1H-indole-
5-carbonitrile (2). To a stirred solution of alcohol 6 (9.55 kg,
28.1 mol) and DIPEA (10.9 kg, 84.2 mol) in N,N-dimethyl-
formamide (DMF, 46.1 L) at 0–5 °C was added trifluoroacetic
anhydride (TFAA, 7.07 kg, 33.7 mol) over 30 min while
maintaining the temperature <25 °C. A DMF rinse (1.6 L) was
added, and the solution was heated to 96–98 °C until <1 area
% (AP) of 6 remained by HPLC. Upon completion of the
reaction, water (23.9 L) was added over 10 min, followed by
ethyl alcohol (EtOH, 32.4 L) over 8 min, and water (23.9 L)
over 4 min while maintaining the temperature >56 °C. The
solution was stirred at 50–55 °C for 20 min and cooled to 0–5
°C for 1 h. The product was collected by filtration. The cake
was washed with a 1:4 mixture of EtOH/water (2 × 28.7 L)
and dried in vacuo at 55–60 °C until constant weight to give
7.69 kg (85.0% yield) of vinyl indole 2. Mp: 133.2–133.6 °C.
1H NMR (d6-DMSO, 400 MHz) δ 8.43 (d, J ) 1.0 Hz, 1H),
8.22 (s, 1H), 8.11 (d, J ) 8.6 Hz, 1H), 7.92 (dm, J ) 8.6 Hz,
2H), 7.76 (dd, J ) 8.6 and 1.5 Hz, 1H), 7.40 (dm, J ) 8.6 Hz,
2H), 6.86 (dd, J ) 17.2 and 11.2 Hz, 1H), 6.00 (d, J ) 17.2
Hz, 1H), 5.38 (d, J ) 11.6 Hz, 1H), 2.31 (s, 3H); 13C NMR
13-step synthesis of 1. All key intermediates were crystalline,
and no special processing equipment was needed, ensuring
maximum manufacturing site flexibility. In situ preparation of
the Nishiyama catalyst for the cyclopropanation eliminated the
use of ethylene gas, thereby greatly simplifying the process and
improving safety. A simple chemical resolution with the
inexpensive chiral amine (R)-(+)-R-methylbenzylamine was
developed to enhance the enantiomeric excess of intermediates.
Compound 14b was accessible via a straightforward process
from cyclopropylmethyl alcohol 11. Finally, a simple process
for detosylation of 14b and API crystallization was found,
providing 1b in 34% overall yield from 5-cyanoindole (3). The
overall process was demonstrated on multikilogram scale.
Experimental Section
All reactions were performed under a nitrogen atmosphere.
All reagents purchased from vendors were used as received
unless otherwise indicated. Chiral analysis was obtained on a
Shimadzu HPLC with a Chiracel AD or AD-H column unless
stated otherwise. Proton and carbon NMR data were collected
on a Bruker AC-300 at 300 MHz for proton and 75 MHz for
carbon or on a Bruker AVANCE 400 at 400 MHz for proton
and 100 MHz for carbon. Melting points were obtained with a
Mettler-Toledo FP 62 melting point instrument by measurement
of the change of luminous intensity during the melting process.
Preparation of 3-Formyl-1H-indole-5-carbonitrile (4b).
To a stirred suspension of (chloromethylene)dimethylammo-
nium chloride (Vilsmeier reagent, 5.31 kg, 41.5 mol) in
acetonitrile (CH3CN, 19.7 L) at 15–25 °C was added a solution
of 1H-indole-5-carbonitrile 3 (4.90 kg, 34.5 mol) in CH3CN
(14.8 L) over 25 min while maintaining the temperature <45
°C. The resulting suspension was adjusted to 25 °C. Upon
completion of the reaction, water (43.5 L) was added. The
reaction mixture was heated at 70 °C for 1 h and then cooled
to 0–5 °C. After stirring at 0–5 °C for 1 h, the solids were
collected by filtration, and the cake was washed with water (2
× 14 L). The material was dried in vacuo at 55–60 °C until
constant weight to yield 5.58 kg (95.1% yield) of indole 4b.
Mp: 275.0–275.9 °C. 1H NMR (d6-DMSO, 400 MHz) δ 12.56
(s, 1H), 9.99 (s, 1H), 8.47 (d, J ) 3.0 Hz, 1H), 8.44 (s, 1H),
7.69 (d, J ) 8.6 Hz, 1H), 7.62 (d, J ) 8.1 and 1.5 Hz, 1H); 13C
NMR (d6-DMSO, 100 MHz) δ 185.34 (CH), 140.20 (CH),
138.78 (C), 126.36 (CH), 125.66 (CH), 123.94 (C), 119.87 (C),
117.98 (C), 113.92 (CH), 104.37 (C). Raman 2222.2, 1650.9,
1416.8 cm-1. Anal. Calcd for C10H6N2O: C, 70.58; H, 3.55;
N, 16.46. Found: C, 70.35; H, 3.59; N, 16.57.
Preparation of 3-Formyl-1-(toluene-4-sulfonyl)-1H-in-
dole-5-carbonitrile (5). To a stirred suspension of 4b (3.44
kg, 20.2 mol) and p-toluenesulfonyl chloride (4.16 kg, 21.8 mol)
in CH3CN (52.9 L) at 20–25 °C was added N,N-diisopropyl-
ethylamine (DIPEA, 4.04 L, 3.00 kg, 23.2 mol) over 21 min
while maintaining the temperature at 20–25 °C. Upon comple-
tion of the reaction, water (17.2 L) was added to the resulting
slurry, and the suspension was stirred for 60 min. The product
was collected by filtration. The cake was washed sequentially
with a 1:1 (v/v) mixture of CH3CN/water (18.5 L) and water
174
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

(d6-DMSO, 100 MHz) δ 146.2 (C), 136.3 (C), 133.5 (C), 130.5
(2CH), 128.4 (C), 128.2 (CH), 126.9 (CH), 126.5 (CH), 126.2
(CH), 125.7 (CH), 120.2 (CH), 119.0 (C), 116.9 (CH2), 114.3
(CH), 106.5 (C), 21.0 (CH3). Raman 3069.2, 3017.8, 2920.2,
2224.4, 1640.7, 1541.3, 771.6, 123.0 cm-1. Anal. Calcd for
C18H14N2O2S: C, 67.07; H, 4.37; N, 8.69; S, 9.94. Found: C,
66.84; H, 4.39; N, 8.63; S, 9.90.
The filtrate was heated to 65 °C and (R)-(+)-R-methylbenzyl-
amine (3.21 kg, 26.5 mmol, 1.09 equiv) was added over 18
min. The reaction mixture was cooled to 20 °C over 3 h and
was then held at 0–5 °C for 1 h. The resulting solid collected
by filtration and washed with CH3CN (2 × 18.6 L). After drying
at 45 °C in vacuo, 6.84 kg of salt 10b was obtained (64.0%
yield, 97.8 HPLC AP purity, 95.5% ee, Ru ) 310 ppm). Mp:
176.4–177.2 °C. 1H NMR (CDCl3, 300 MHz) δ 8.04 (dm, J )
8.4 Hz, 1H), 7.85 (s, 1H), 7.74 (dm, J ) 8.4 Hz, 2H), 7.56 (dd,
J ) 7.7, 1.5 Hz, 1H), 7.29 (m, 9H), 5.28 (s (br), 3H), 4.23 (q,
J ) 6.8 Hz, 1H), 2.35 (s, 3H), 2.34–2.26 (m, 1H), 1.75–1.69
(m, 1H), 1.52 (d, J ) 6.8 Hz, 3H), 1.48–1.43 (m, 1H), 1.17–1.11
(m, 1H); 13C NMR (CDCl3, 75 MHz) δ 179.80, 146.11, 140.75,
137.09, 135.12, 131.44, 130.56, 129.23, 128.65, 128.16, 127.22,
126.49, 125.08, 124.19, 123.66, 119.72, 114.79, 106.96, 51.31,
24.77, 22.61, 22.02, 15.45. Anal. Calcd for C28H27N3O4S: C,
67.04; H, 5.43; N, 8.38; S, 6.39. Found: C, 67.32; H, 5.58; N,
8.09; S, 6.51.
Preparation of Ethyldiazoacetate (EDA) in Toluene. To
a stirred solution of sodium tetraborate decahydrate (1.07 kg,
2.81 mol) in water (21.5 L) at 25 °C was added sodium nitrite
(5.58 kg, 80.9 mol) followed by glycine ethyl ester hydrochlo-
ride salt (10.8 kg, 77.0 mol). Upon complete dissolution, toluene
(24.9 L) was added, and the resulting biphasic mixture was
cooled to 0 °C. A 2% (w/w) phosphoric acid solution in water
was added over 30 min while maintaining the temperature <20
°C until the pH was between 3.7 and 4.5 (addition of 60.3 kg
resulted in a pH of 3.9). The organic layer was washed
successively with water (10.8 L) and 8% (w/w) of aqueous
sodium bicarbonate (2 × 21.9 L). The combined aqueous
washes were neutralized with a 20 wt % solution of phosphoric
acid in water. The organic layer was assayed for EDA content
by HPLC, NMR and GC and was held overnight at 10 °C
before being transferred to a pressure bomb. The reactor was
rinsed with toluene (17.3 L), which was transferred to the
pressure bomb as well. Residual EDA in the 200-L reactor was
neutralized by addition of a 50 wt % solution of acetic acid in
water (86 kg).
Preparation of (1S,2S)-2-[5-Cyano-1-(toluene-4-sulfonyl)-
1H-indol-3-yl]-cyclo-propanecarboxylicAcid,(R)-(+)-r-Meth-
ylbenzylamine Salt (10b). To a stirred solution of dichloro(p-
cymene)ruthenium(II) dimer (0.37 kg, 0.61 mol) and Pybox
(4R) ligand (0.382 kg, 1.27 mol) in THF (37.6 L) at 20–25 °C
was added a solution of vinyl indole 2(7.79 kg, 24.2 mol) in
THF (36.1 L). The resulting mixture was heated to 55–56 °C,
and 2 M solution of EDA in toluene (25 kg) was added at 3
L/h. Upon completion of the addition, HPLC indicated the
amount of unreacted 2 was <1 AP. The reaction mixture was
cooled to 20–25 °C, and water (23.4 L) was charged. The batch
was concentrated to 30–35 L and cooled to 50 °C, and CH3CN
(58.5 L) was added. The resulting solution was cooled to 10
°C, and CF3SO3H (triflic acid, 9.25 kg, 61.6 mol) was added
over 10 min while maintaining the temperature e25 °C during
the addition. The mixture was concentrated until the batch
temperature reached 82 °C. A volume of CH3CN/water (ratio
greater than 4:1) equivalent to the distillate was added, and three
additional distillation/addition cycles were conducted. The
reaction mixture was cooled to 72 °C, and 17.4 L of water was
added while maintaining the temperature >65 °C. The mixture
was cooled to 20–25 °C, and water (20 L) was added to the
resulting slurry. The mixture was stirred for 1 h at 20–25 °C,
and the solid was collected by filtration. The cake was washed
twice with water (19.8 L,15.2 L), and the solid was dried at 50
°C in vacuo to yield 9.37 kg of acid 10a (100% crude yield,
83.8% HPLC AP purity, 87.2% ee, Ru ) 3100 ppm).
Preparation of 3-((1S,2S)-2-Hydroxymethyl-cyclopropyl)-
1-(toluene-4-sulfonyl)-1H-indole-5-carbonitrile (11). A mix-
ture of salt 10b (6.42 kg, 12.8 mol), THF (26.7 kg), methyl
tert-butyl ether (53.4 kg), and 10% aqueous sulfuric acid
solution (27.9 kg) was agitated until dissolution. The phases
were separated, and the organic phase washed sequentially with
10% aqueous sulfuric acid (27.9 kg), 15% brine (29.8 kg), and
26% brine (32.6 kg). The organic phase was passed through a
5 µm polish filter, and 36.3 kg of solvent was distilled at
atmospheric pressure. THF (20.4 kg) was charged, and an
additional 24.3 kg of solvent was distilled (final batch temper-
ature 66 °C, KF 0.20%). After holding for 16 h at 20–25 °C,
the solution was cooled to 0 °C and 1 M BH3 ·THF in THF
(12.9 kg, 14.7 mol, 1.15 equiv) was charged over 90 min,
modulating the addition rate as necessary to maintain the batch
temperature below 10 °C. The O2 content of the headspace was
continuously monitored and maintained below 0.4% throughout
the reduction sequence. After heat evolution had ceased, the
mixture was warmed to 22 °C over 30 min, and additional 1
M BH3 ·THF (1.2 kg, 1.4 mol, 0.11 eq) was added over 10
min. HPLC analysis indicated <1.5% starting material remain-
ing, and water (2.6 kg) and CH3CN (6.0 kg) were added. Dilute
HCl (prepared from 88.3 kg water and 3.2 kg conc HCl) was
charged over 45 min, and the resultant slurry was agitated for
90 min. The product 11 was collected by filtration, washed with
water (2 × 7.5 kg), and dried in vacuo until LOD <1.0% to
yield 4.5 kg of alcohol 11 as an off-white powder (12.3 mol,
1
96% yield, 98.3 HPLC AP purity). Mp: 139.0–140.5 °C. H
NMR (d6-DMSO, 400.13 MHz) δ 8.29–8.28 (m, 1H), 8.05 (d,
J ) 8.59 Hz, 1H), 7.90–7.86 (dm, J ) 8.45 Hz, 2H), 7.73 (dd,
J ) 8.59, 1.76 Hz, 1H), 7.65 (d, J ) 1.76 Hz, 1H), 7.38 (dm,
J ) 8.45 Hz, 2H), 4.70 (t, J ) 5.81 Hz, 1H), 3.59–3.52 (m,
1H), 3.31–3.22 (m, 1H), 2.31 (s, 3H), 1.85–1.79 (m, 1H),
1.23–1.14 (m, 1H), 1.07–1.01 (m, 1H), 0.85–0.79 (m, 1H); 13C
NMR (d6-DMSO, 100.62 MHz) δ 145.90, 136.11, 133.77,
131.15, 130.37, 127.88, 126.80, 125.23, 125.06, 123.49, 119.24,
114.26, 105.77, 63.81, 24.61, 21.04, 10.95, 10.74. Anal. Calcd
for C20H18N2O3S: C, 65.55; H, 4.95; N, 7.64; S, 8.75. Found:
C, 65.62; H, 5.10; N, 7.71; S, 8.39.
A slurry of crude 10a (9.29 kg, 24.4 mol) in anhydrous
CH3CN (115 L) was heated at 75 °C until dissolution occurred.
The solution was cooled to 35 °C and was passed through a 5
µm cartridge filter. The filter was rinsed with CH3CN (7.2 L).
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
175

Preparation of 3-((1S,2S)-2-Dimethylaminomethyl-cyclo-
propyl)-1-(toluene-4-sulfonyl)-1H-indole-5-carbonitrile, Hy-
drochloride, Hemihydrate (14b). A stirred mixture of alcohol
11 (3.38 kg, 9.22 mol), TEMPO (36 g, 0.23 mol), sodium
bicarbonate (856 g, 10.2 mol), methyl acetate (33.7 L), and
water (8.83 L) was cooled to 0–5 °C. The reactor was flushed
5 times with nitrogen, and an 11 wt % solution of sodium
hypochlorite (NaOCl, 7.19 kg, 10.6 mol) was charged over 140
min while maintaining the internal temperature at 0–5 °C. HPLC
analysis indicated the AP of 11 was <1% relative to the
aldehyde intermediate 13. The reaction was quenched by
addition of ethyl alcohol (0.99 kg, 21 mol), and was warmed
to 20–25 °C over 30 min. The organic layer was passed through
a 0.5 µm cartridge filter, and the filtrate was cooled to 0–5 °C.
A 2 M solution of dimethylamine in THF (7.85 kg, 18.5 mol)
was charged to the batch, followed by a rinse of methyl acetate
(0.4 L). The reaction mixture was stirred for 10–15 min, and
the vessel was purged 5 times with nitrogen. Sodium tri-
acetoxyborohydride (2.53 kg, 11.9 mol) was added over 30 min,
leading to a 7–10 °C temperature increase. After the addition,
the reaction mixture was warmed to 20–25 °C for 15 min. Water
(17 L) was added, followed by stirring for 15 min. The aqueous
layer was separated and back extracted with methyl acetate (6.75
L). The combined organic layers were washed sequentially with
sodium bicarbonate solution (1.55 kg sodium bicarbonate in
17.0 L of water) and sodium chloride solution (4.43 kg sodium
chloride in 14.5 L of water). The organic phase was passed
sequentially through a Zeta R53SP filter (six 8′′ diameter pads)
at a rate of 2 L/min, and a 0.45 um polish filter. The organic
phase was then concentrated to 17 L. 2-Propanol (IPA, 41 L)
was charged, and the solution was concentrated to 27.2 L. The
resulting mixture was warmed to 72 °C and additional IPA (3.4
L) was added. Concurrently, a solution of HCl/IPA was
prepared by the addition of 1.16 kg (14.8 mol) of acetyl chloride
over 50 min to IPA (6.8 L) while maintaining the temperature
<18 °C. The resulting HCl/IPA solution was charged over 10
min to the batch while maintaining the temperature above 65–70
°C. The mixture was cooled to 62–65 °C over 25 min, during
which time crystallization started. The thickening slurry was
held for 30 min at 62 °C with vigorous agitation, cooled to 20
°C over 30 min, and held at 20 °C for 1 h. The crystals were
collected by filtration and washed with IPA (6.8 L) followed
by MTBE (6.8 L). The solid was dried at 50 °C in vacuo until
LOD was <1%, affording 3.50 kg (88.1%, 99.8 HPLC AP
purity, Ru ) 8 ppm) of salt 14b as an off-white solid. Mp:
Preparation of 3-((1S,2S)-2-Dimethylaminomethyl-cyclo-
propyl)-1H-indole-5-carbonitrile, Maleic Acid Salt (1a). A
mixture of salt 14b (1.7 kg, 4.0 mol), THF (4.6 kg), and 50 wt
% aqueous NaOH (6.6 kg) was heated at 68 °C in a 40 L
Hastelloy vessel for 4 h with vigorous stirring. The batch was
cooled to 30–35 °C, and water (5.0 L) was added while
maintaining the temperature <35 °C, followed by addition of
THF (2.0 L) and MTBE (3.4 L). The mixture was transferred
to a 50 L borosilicate reactor. Additional water (10.3 L) and
MTBE (3.4 L) were added, and the mixture was stirred at 20
°C for 15 min. The phases were separated, and the organic phase
was concentrated to 4.2 L. The reactor was charged with a 1:1
(v/v) mixture of THF/MTBE (7.7 L). The mixture was
concentrated again, and the KF of the batch was measured to
be 0.32 wt %. To the batch was added THF (7.6 L), and the
mixture was held at 25 °C. The solution was then passed
sequentially through 10 µm and 0.45 µm polish filters. The filter
train was rinsed with THF (1.5 L), and seeds (1a, 10 g in THF
(0.11 L)) were added to the resulting batch. A solution of maleic
acid (0.51 kg, 4.39 mol) in 1.93 L of THF was added through
a 10 µm Cuno filter. The resulting slurry was stirred at 20–25
°C for 15 h, and the solid was collected by filtration. The cake
was washed sequentially with THF (2.48 L) and a 1:1 (v/v)
mixture of THF/MTBE (2.4 L). The product was dried at 35–40
°C in vacuo until the LOD was <1% to yield 1.08 kg of API
1a (76.6% yield, 99.5 HPLC AP purity and 98.3% ee). Mp:
167 °C. 1H NMR (d6-DMSO, 300 MHz) δ 1.01 (ddd, J ) 8.8,
4.8, 4.8 Hz, 1H), 1.14 (ddd, J ) 8.1, 5.3, 5.3 Hz, 1H), 1.30 (m,
1H), 2.09 (ddd, J ) 8.3, 5.3, 5.3 Hz, 1H), 2.83 (s, 6H), 3.13
(d, J ) 13.1, 7.5 Hz), 3.27 (d, J ) 13.1, 7.1 Hz, 1H), 6.04 (s,
2H), 7.32 (d, J ) 2.3 Hz, 1H), 7.41 (dd, J ) 8.5, 1.5 Hz, 1H),
7.50 (d, 8.4 Hz, 1H), 8.18 (s, 1H), 9.50 (s broad, 1 H), 11.49
(s, 1H); 13C NMR (CD3OD, 75 MHz) δ 171.29, 140.29, 137.11,
129.18, 125.81, 125.65, 125.16, 122.34, 117.92, 114.04, 102.87,
63.12, 43.65, 17.10, 14.92, 14.05. Anal. Calcd for C19H21N3O4:
C, 64.21; H, 5.96; N, 11.82. Found: C, 63.86; H, 5.69; N, 11.45.
Preparation of 3-((1S,2S)-2-Dimethylaminomethyl-cyclo-
propyl)-1H-indole-5-carbonitrile, HCl Salt (1b). The deto-
sylation described above was applied to 1.7 kg (4.0 mol) of
14b with no modification of the protocol through the reaction
and phase separation phases. After phase separation, the batch
was concentrated under atmospheric pressure until the internal
volume was 4.0 L. MTBE (7.55 kg) was added, and distillation
was resumed until the internal volume was 4.0 L. The KF of
the concentrated batch was measured to be 0.47%. THF (0.84
kg) and MTBE (3.13 kg) were added to the batch, and the
mixture was cooled to 20–25 °C (KF 0.26%). The batch was
passed through 10 µm and 0.5 µm filters in series. The filter
train was rinsed with MTBE (0.95 kg). To the stirred batch
were added seeds (7 g, 1b) as a slurry in MTBE (1.68 kg),
followed by EtOH (1.88 kg). A solution of HCl in EtOH
(prepared by addition of acetyl chloride (0.31 kg, 3.95 mol) to
EtOH (2.71 kg)) was added over 1 h to the batch. The resulting
slurry was held for 18 h at 20–25 °C, and the solids were
collected by filtration. The cake was washed with 95:5 (v/v)
MTBE/EtOH (2.5 L) and MTBE (2 × 2.4 L). After drying at
50 °C in vacuo (LOD <1%), 0.964 kg of API 1b was isolated
(88.7% yield, 99.8 HPLC AP purity, 98.7% ee). Mp: 176.5–177.5
1
212.0–213.3 °C. H NMR (d6-DMSO, 400 MHz) δ 10.60 (s,
1H), 8.40–8.38 (m, 1H), 8.06 (d, J ) 8.59 Hz, 1H), 7.89 (dm,
J ) 8.34 Hz, 2H), 7.79 (s (br), 1H), 7.74 (dd, J ) 8.59, 1.51
Hz, 1H), 7.39 (dm, J ) 8.34 Hz, 2H), 3.26–3.18 (m, 1H),
3.11–3.03 (m, 1H), 2.75 (s, 6H), 2.31 (s, 3H), 2.20–2.14 (m,
1H), 1.49–1.40 (m, 1H), 1.34–1.27 (m, 1H), 1.13–1.07 (m, 1H);
13C NMR (d6-DMSO, 100 MHz) δ 145.98, 136.10, 133.71,
130.73, 130.41, 128.02, 126.83, 125.35, 124.06, 123.59, 119.20,
114.29, 105.91, 59.53, 41.45, 21.06, 16.47, 12.86, 12.02. Anal.
Calcd for C22H24ClN3O2S.1/2 H2O: C, 60.19; H, 5.74; Cl, 8.08;
N, 9.57; S, 7.30. Found: C, 59.77; H, 5.52; Cl, 8.56; N, 9.70;
S, 7.26.
176
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

°C. 1H NMR (d6-DMSO, 400.13 MHz) δ 11.61 (s, 1H), 10.87
(s broad, 1H), 8.25 (s, 1H), 7.49 (d, J ) 8.59 Hz, 1H), 7.41
(dd, J ) 8.59, 1.51 Hz, 1H), 7.32 (d, J ) 1.51 Hz, 1H),
3.24–3.07 (m, 2H), 2.82–2.68 (m, 6H), 2.20–2.13 (m, 1H),
1.37–1.27 (m, 1H), 1.17–1.10 (m, 1H), 1.06–1.00 (m, 1H); 13C
NMR (d6-DMSO, 125.77 MHz) δ 137.88, 127.08, 124.30,
123.84, 123.76, 120.83, 116.22, 112.70, 100.37, 59.73, 41.52,
41.34, 16.02, 12.83, 12.39. Anal. Calcd for C15H18ClN3: C,
65.32; H, 6.57; N, 15.23; Cl, 12.85. Found: C, 65.03; H, 6.62;
N, 15.23; Cl, 12.92.
scale-up runs. In addition, we thank Ms. Zerene Clarin and Dr.
Olav Lyngberg for preparative chromatography work, Dr. Jeff
Bien for API crystallization work, and the safety group for
hazard assessment studies. Simon Leung, Ehrlic Lo, Frank
Okuniewicz, Srinivas Tummala and Steve Wang performed heat
flow analysis on the carboxylic acid reduction. We also thank
Drs. David Kronenthal, James Sherbine, Richard Mueller,
Robert Discordia, Jaan Pesti, Melanie Miller, and Edward
Delaney for helpful discussions during the preparation of the
manuscript.
Acknowledgment
The authors wish to thank the BMS glass plant operations
staff and Analytical R & D staff for their valuable support during
Received for review June 5, 2007.
OP700125Z
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
177