Development of a new practical synthesis of a 5-HT2C receptor agonist

Article abstract of DOI:10.1021/op100233f

A new practical synthesis of a 5-HT_2C receptor agonist has been developed and implemented on multikilogram scale. The key step, the selective epoxide opening in the glycidyl tosylate with the aryl Grignard reagent, allowed the incorporation of this commercially available chiral C₃ synthon into the molecule and elaboration of the resulting intermediate into the target aminomethyldihydrobenzofuran without loss of enantiomeric purity.

Full text of DOI:10.1021/op100233f

Organic Process Research & Development 2010, 14, 1438–1447
Development of a New Practical Synthesis of a 5-HT2C Receptor Agonist†
Alexander Gontcharov,*,‡ Chia-Cheng Shaw,§ Qing Yu,§ Sam Tadayon,§ Michel Bernatchez,§ Mark Lankau,§ Michel Cantin,§
John Potoski,⊥ Gulnaz Khafizova,⊥ Gary Stack,| Jonathan Gross,| and Dahui Zhou|
Wyeth Research Chemical DeVelopment and Chemical Sciences, 401 North Middletown Road,
Pearl RiVer, New York 10965, United States, Wyeth Research, Chemical DeVelopment, 1025 Marcel Laurin BouleVard, St.
Laurent, QC H4R 1J6, Canada, and Wyeth Research, Chemical Sciences, 865 Ridge Road,
Monmouth Junction, New Jersey 08852, United States
Abstract:
WAY-255719 (1)2 has been shown to be a potent 5-HT2C
full agonist and demonstrates greater than 100-fold selectivity
over 5-HT2A and 5-HT2B receptors in binding and functional
assays. It has been effective in several animal models predictive
of anti-psychotic activity, with an atypical anti-psychotic profile
and has been selected from a series of analogue compounds
and advanced as a development candidate for treatment of
schizophrenia in Wyeth’s 5-HT2C agonist program.
A new practical synthesis of a 5-HT2C receptor agonist has been
developed and implemented on multikilogram scale. The key step,
the selective epoxide opening in the glycidyl tosylate with the aryl
Grignard reagent, allowed the incorporation of this commercially
available chiral C3 synthon into the molecule and elaboration of
the resulting intermediate into the target aminomethyldihydroben-
zofuran without loss of enantiomeric purity.
Introduction
The 5-HT2C receptor is one of a family of 14 G-protein
coupled receptors which respond to stimulation by the endog-
enous ligand 5-hydroxytryptamine or serotonin.1 Our interest
in 5-HT2C agonism stems from the fact that, in some brain
regions such as the mesolimbic system, dopamine is under the
control of 5-HT2C receptors. 5-HT2C agonists diminish dopamine
receptor firing and dopamine release in these areas without
affecting dopamine levels in other areas, such as the nigrostriatal
track, in which dopamine is important for control of movement.
Reducing dopamine activity in the mesolimbic system has a
beneficial effect on the most florid of the symptoms of
schizophrenia, the so-called positive symptoms. Stimulation of
5-HT2C receptors would also produce an antidepressant effect
similar to selective serotonin reuptake inhibitors which stimulate
serotonin receptors indirectly by increasing synaptic levels of
serotonin. However, because of negative feedback controls, they
take some time to achieve their effect. Direct serotonin agonists
would achieve the same effect immediately. Correspondingly,
5-HT2C agonists are not only active in animal models of
depression, but indicate a faster onset in models which predict
onset.
On a small scale, the lead compound was first prepared as
a racemate (rac-1) via an eight-step synthetic sequence (Scheme
1).2 Most of the SAR analogues of 1 were synthesized in a
similar manner, one variation being the point at which Suzuki
coupling was done: for many analogues it followed the
dihydrobenzofuran ring formation and utilized a reverse
nucleophile-electrophile position of functionalities. In the case
of 1, the alternative approach did not work due to the instability
of o,o-dichlorophenylboronic acid under the conditions of
Suzuki coupling.
To resolve the enantiomers, rac-1 was converted to its Cbz
derivative and subjected to chiral HPLC separation. Subsequent
removal of the Cbz protection gave the enantiomerically pure
1.3
Although the synthesis described above is adequately ef-
ficient for the preparation of the racemic product and, with
minor modifications, could be scaled up to produce a few
Another one of the most widely appreciated functions of
the 5-HT2C receptor is the regulation of appetite and feeding
behavior. Agonists of 5-HT2C decrease appetite and feeding and
thus lead to loss of weight.
(2) (a) Gross, J. L.; Williams, M. J.; Stack, G. P.; Gao, H.; Zhou, D.
Dihydrobenzofuranyl alkanamine derivatives as brain serotonin 2C
receptor agonist and partial agonists and their preparation, pharma-
ceutical compositions and use in the treatment of CNS disorders. Chem.
Abstr. 2008, 149, 471316. U.S. Patent 7,435,837, 2008. (b) Gross,
J. L.; Dunlop, J.; Gao, H.; Grauer, S.; Harrison, B. L.; Huselton, C.;
Kalgaonkar, S.; Lai, M.; Logue, S.; Marquis, K.; Mazandarani, H.;
Stack, G. P.; Sung, A.; Williams, M. J.; Zhang, J.; Zhou, D.;
Rosenzweig-Lipson, S. 2-[Aminomethyl]dihydrobenzofurans as
5-HT2C receptor agonists. Abstracts of Papers of the 238th ACS
National Meeting; American Chemical Society: Washington, DC,
August 2009, MEDI-334.
† Wyeth was acquired by Pfizer on 16 October 2009.
* Corresponding author. E-mail: alex.gontcharov@pfizer.com.
‡ Chemical Development, Pearl River, NY, U.S.A.
§ Chemical Development, Canada.
⊥ Chemical Sciences, Pearl River, NY, U.S.A.
| Chemical Sciences, Monmouth Junction, NJ, U.S.A.
(1) For a review of therapeautic aspects of the 5-HT2C receptor, see:
Rosenzweig-Lipson, S.; Dunlop, J.; Marquis, K. L. Drug News
Perspect. 2007, 20, 565.
(3) It was found later that the enantiomers of the unprotected amine could
also be separated by a normal phase preparative HPLC. The separation
was performed on a 100-g scale and took about 1 week to complete
using a 25 cm × 3 cm column.
1438
•
Vol. 14, No. 6, 2010 / Organic Process Research & Development
10.1021/op100233f  2010 American Chemical Society
Published on Web 11/02/2010

Scheme 1. Medicinal Chemistry approach to preparation of 1
Scheme 2. Retrosynthetic considerations for new
substitutions (if needed), and a number of compounds containing
C3-fragments with epoxides are commercially available in
enantiomerically pure form, particularly glycidol derivatives 3.
The synthetic plan shown in Scheme 3 illustrates the general
strategy for constructing the chiral dihydrobenzofuran moiety.
Epoxide opening with an aryllithium, organocopper, or a
Grignard reagent would create the carbon-carbon bond of the
heterocycle, and such openings are well documented in the
literature.7 The hydroxyl formed as a result of the epoxide
opening would be converted to a leaving group and used to
close the five-membered cycle by a nucleophilic substitution.
The pendant functional group, Y, would need to be a properly
protected amine or a functional group which could be converted
into the amine later so as not to interfere with the ring formation.
Details such as the choice of the protecting groups and reagents,
as well as how and at which point to incorporate the dichlo-
rophenyl substituent, were to be filled in along the way.
Proof of Concept. The key step of the new strategy was
likely to be the opening of the epoxide with the aryl nucleophile.
There are multiple examples of epoxide openings by aryllithium
and Grignard reagents, but the implementation of this chemistry
on scale has been scarce. Reactions between organometallic
reagents and epoxides are catalyzed by the copper(I) salts8 or
Lewis acids (e.g., BF3 etherate).9 Without any catalysts, these
reactions have been reported to be sluggish and are accompanied
by a variety of side products.7a
multikilogram route to 1
hundred grams of material,4 the need for chiral separation at
the end, via either chemical resolution or chromatography,
undoes all of its advantages. This contribution describes
development of a new practical synthesis of 1 and its imple-
mentation for preparation of kilogram quantities of the active
pharmaceutical ingredient (API).
Selection of Synthetic Route. Among a variety of plausible
approaches to the chiral benzodihydrofuran, we selected one
which retrosynthetically puts the molecule together from the
phenolic part and a C3-fragment (Scheme 2). The advantage of
this approach would be the possibility to introduce the chirality
into the target molecule with the chiral C3 fragment, and there
is a considerable selection of compounds of this type com-
mercially available in enantiomerically pure form.5 As the two
furan cycle-forming reactions would be done sequentially, a
proper choice of X1, X2, and Y functionalities as well as the
utilization of protecting groups would be required to ensure the
selectivity of the transformations. An epoxide was considered
a proper choice for the X1, X2 functionalities as it would provide
the necessary regioselectivity (nucleophilic opening usually
occurs at the terminal carbon for steric reasons).6 It may also
provide good chemoselectivity with respect to other nucleophilic
In a model reaction, o-methoxyphenylmagnesium bromide
2a taken in an excess as a commercial THF solution was treated
with benzyl glycidol 3a in the presence of CuI as a catalyst at
-40 °C (Scheme 4). Reaction was complete in 1 h, and no
(7) (a) Silverman, G. S. In Handbook of Grignard Reagents; Silverman,
G. S., Rakita, P. E., Eds.; Chemical Industries Series, Vol. 64; Marcel
Dekker: New York, NY, 1996; pp 322-325. (b) Gaylord, N. G.;
Becker, E. I. Chem. ReV. 1951, 49, 413. (c) Lipshutz, B. H.; Wilhelm,
R. S.; Kozlowski, J. A.; Parker, D. J. Org. Chem. 1984, 49, 3928. (d)
Kurth, M. J.; Abreo, M. A. Tetrahedron 1990, 46, 5085.
(4) The process in Scheme 1 has been successfully scaled up to 70 g for
preparation of 1 and its des-fluoro analogue (unpublished results).
(5) Sheldon, R. A. Chirotechnology: Industrial Synthesis of Optically
ActiVe Compounds; Marcel Dekker: New York, NY, 1993; Chapter
5.
(6) For general reference, see: Lewars, E. G. Oxiranes and Oxirenes. In
ComprehensiVe Heterocyclic Chemistry; Katritzky, A. R., Rees, C. W.,
Eds.; Pergamon Press: Oxford, 1984; Vol. 7, pp 95-129.
(8) (a) Schuda, A. D. C.; Mazzocchi, P. H.; Fritz, G.; Morgan, T. Synthesis
1986, 309. (b) Review: Erdik, E. Tetrahedron 1984, 40, 641.
(9) (a) Eis, M. J.; Wrobel, J. E.; Ganem, B. J. Am. Chem. Soc. 1984, 106,
3693. (b) Alexakis, A.; Vrancken, E.; Mangeney, P. Synlett 1998, 1165.
Vol. 14, No. 6, 2010 / Organic Process Research & Development
•
1439

Scheme 3. General strategy for construction of the chiral dihydrobenzofuran core
Scheme 4. Addition of o-methoxyphenyl Grignard reagent
to glycidol derivatives
Increasing the amount of the catalyst or switching from CuI to
CuCN did not have any effect on the reaction.
The glycidol tosylate opening was successfully scaled up to
10 g of the starting aryl bromide. The mixture of products in
THF was treated with aq NaOH to convert the remaining 4c to
epoxide 5c (Scheme 6). The resulting mixture of 5c and side
product 6c (80:20 area % ratio at 215 nm) was treated with a
mixture of phthalimide and its potassium salt in DMF12 which
cleanly gave the epoxide opening product 7c. The latter was
an amorphous solid, and we were able to separate it from 6c
that we carried along thus far by simple trituration. The yield
of 7c was estimated at 52% based on the amount of the aryl
bromide taken for Grignard formation. Treatment of 7c with
methanesulfonyl chloride in the presence of triethylamine gave
mesylate 8c cleanly as a solid in 95% yield.
starting epoxide remained in the mixture. Alcohol 4a and the
quenched Grignard reagent (anisole) together with some minor
impurities were present in the isolated product mixture.
To allow for more efficient functional group manipulation
in the subsequent synthetic steps, glycidyl tosylate 3b10 was
also subjected to the reaction conditions with o-methoxyphe-
nylmagnesium bromide (1 equiv). The reaction was complete
in 3 h at -40 °C, resulting in the epoxide opening. No tosylate
displacement in the product was observed.
Demethylation of the methoxyaryl in 4a and 4b with BBr3
solution was attempted at this point to study the possibility of
direct cyclization of the resulting hydroxyphenol, but it led to
a complex mixture of products in both cases.
To facilitate deprotection of the phenol in the later interme-
diates, methyl protection in the starting arene was changed to
benzyl protection. Corresponding 2-benzyloxy-5-fluorophenyl-
magnesium bromide 2c (prepared in two steps from 2-bromo-
4-fluorophenol) was reacted with glycidol tosylate (Scheme 5).
The markedly lower reactivity of the new Grignard reagent
toward epoxide opening was apparent from the first experiment,
which could be attributed to both electronic contribution of the
fluoride substituent and the steric influence of the benzyl group;
whereas the reaction with methoxyphenyl Grignard 2a was
complete in 1-3 h, barely a trace of the product 4c formed
with benzyloxyphenyl Grignard 2c under similar conditions.
Raising the reaction temperature to -25 °C and extending the
reaction time to 20 h did drive the epoxide opening to
completion but also caused in situ cyclization of product 4c to
form the terminal epoxide 5c.11 Fortunately, we did not observe
detectable opening of 5c with the starting Grignard reagent 2c.
A problem was encountered when we attempted the cleavage
of the benzyl protection by catalytic hydrogenation. A variety
of conditions was screened, none of which presented anything
appropriate for optimization on a larger scale.13
In parallel with our debenzylation work, we considered
changing the protection on the phenol to a MOM group which
should withstand Grignard reactions and be removable under
mild acidic conditions.14 MOM-protected 2-bromo-4-fluorophe-
nol was prepared using the procedure developed previously at
Wyeth by heating a mixture of the phenol, dimethoxymethane,
DMF, and POCl3 in heptane.15 The Grignard reagent was
prepared by reaction with Mg in THF. Addition to the epoxide,
however, occurred much slower than in previous cases. The
reaction was not complete after 24 h at -23 °C, and the product
mixture was overwhelmed by impurities (over 60% by total
HPLC peak area). After treatment with aq NaOH, it was reacted
with phthalimide-phthalimide K salt mixture. The desired
product failed to crystallize or solidify, giving us no means of
isolation and purification besides flash chromatography. This
route was not pursued any further.
(12) This phthalimide buffer is necessary for the epoxide opening as
phthalimide alone is not nucleophilic enough to open the epoxide,
whereas its presence is crucial in providing acidic catalysis of the
reaction: (a) Lawrence, N. J.; Bushell, S. M. Tetrahedron Lett. 2001,
42, 7671. (b) Williams, T. M.; Crumbie, R.; Mosher, H. S. J. Org.
Chem. 1985, 50, 91.
(10) (a) Klunder, J. M.; Onami, T.; Sharpless, K. B. J. Org. Chem. 1989,
54, 1295. (b) Maruyama, T.; Asada, M.; Shirraishi, T.; Yoshida, H.;
Maruyama, T.; Ohuchida, S.; Nakai, H.; Kondo, K.; Toda, M. Bioorg.
Med. Chem. 2002, 10, 1743.
(13) The list of hydrogenation conditions screened includes Pd/C (50 psi
hydrogen or transfer hydrogenation, no reaction observed), Pd black
(stalled after 10% conversion), Pd(OH)2 on carbon (80% conversion
after high multiple catalyst charges, phthalimide reduction observed).
(14) Greene, T. W.; Wuts, P. G. M. ProtectiVe Groups in Organic Synthesis,
3rd ed.; Wiley & Sons: New York, 1999; pp 257-258.
(11) Theoretically, the enantiomer of epoxide 5c can form by a direct
displacement of the tosylate in 3b without epoxide opening. It has
been reported, however (ref 10), that this does not happen in practice
and 5c forms via the epoxide opening and subsequent closure on the
other side. This was confirmed as we established that no loss of
enantiomeric purity occurred, whereas direct tosylate displacement
would lead to the terminal epoxide of opposite stereo configuration.
(15) Schouten, H. G. Preparation of Alkoxy Methyl Ethers. U.S. Patent
4,500,738, Feb. 19, 1985.
1440
•
Vol. 14, No. 6, 2010 / Organic Process Research & Development

Scheme 5. Addition of o-benzyloxyphenyl Grignard reagent to glycidyl tosylate
Scheme 6. Introduction of the phthalimide fragment and setting up for the dihydrobenzofuran cyclization
Scheme 7. Formation of dihydrobenzofuran ring and removal of phthalimide protection
1
Returning to benzyl-protected phenol derivatives, nonreduc-
tive ether cleavage was attempted and gave encouraging results.
On treatment of 8c with BBr3 in methylene chloride at -78 to
0 °C,16 the benzyl group was cleanly removed, resulting after
aqueous workup, in a mixture of deprotected phenol 9c and
benzyl bromide (Scheme 7). This result suggested that methyl-
protected phenol might work in the reaction sequence as well
as benzyl, additionally easing up steric hindrance around the
aryl-magnesium group during epoxide opening.
Treatment of a methanolic solution of 9c with aq NaOH
led immediately to two new products, one of which was
identified as 10c, the other had the same (M + H)+ in LC/MS
as that of 10c but a different fragmentation pattern. The structure
of the latter side product has not been determined. Screening
of the reagents and conditions for the cyclization allowed
identification of the set in which the reaction occurred cleanly
and no side products formed (Scheme 7) (no conditions were
found in which cyclization occurred without phthalimide ring
opening). Removal of the phthalic amide group was achieved
by heating 10c with hydrazine hydrate in IPA.
which were then analyzed by H NMR and HPLC. Both
experiments showed that two samples contained individual
distinct diastereomers of the amides, confirming the enantio-
meric purity of the amine 1c. HPLC analysis of the diastereomer
mixture established the enantiomeric purity of 1c to be >99%
ee.
This result provided the necessary proof of concept that the
selected synthetic strategy would allow construction of the chiral
dihydrobenzofuran core without loss of the enantiomeric purity
of the starting glycidyl fragment. From here we could proceed
to filling in the missing parts of the synthesis and mapping the
overall synthetic strategy.
One of the remaining pieces to be fitted was incorporation
of the dichloroaryl substituent. It could be added into the
dihydrobenzofuran core 1c by ortho-lithiating at the C-7
position, making the boronic ester and coupling it by Suzuki
chemistry with dichlorobromobenzene. Protection and depro-
tection of the amino group was likely to be required as well,
adding four steps to the linear synthesis. Carrying out the Suzuki
coupling in the beginning of the synthesis and taking into
account available starting materials would add only two steps
to the synthesis and would not require cryogenic lithiation. Thus,
the latter alternative was chosen.
The enantiomeric purity of 1c was confirmed by converting
the product into a pair of the diastereomeric Mosher amides
(16) Fieser, L. F.; Fieser, M. Reagents for Organic Synthesis; Wiley &
Sons: New York, 1967; p 67.
Vol. 14, No. 6, 2010 / Organic Process Research & Development
•
1441

Scheme 8. Synthetic plan for preparation of 1
Scheme 8 illustrates the synthetic plan using methyl protec-
tion of the phenol in place of benzyl this time. The starting
biaryl fragment for construction of the fused bicyclic system
would begin with a Suzuki coupling of commercially available
2-methoxy-5-fluorophenyl boronic acid (11) and o,o-dichloro-
bromobenzene (12), similar to the first step of the Medicinal
Chemistry route. Resulting biaryl 13 was to be selectively
brominated ortho to the methoxy group. Grignard reagent was
to be generated from aryl bromide 14 (methodology would need
to be developed to ensure the aryl chlorides remain intact).
Finally, the chiral dihydrofuran ring would be constructed
following the developed methodology to result in enantiomeri-
cally pure 1.
50 °C afforded 95% conversion in 18 h without noticeable
demethylation. Crystallization of the product after extractive
workup allowed isolation of the product in 80% yield and 98%
purity, leaving most of the unreacted starting material in the
mother liquors. This method was used to prepare first batches
of the product on the laboratory scale.
Further optimization of the reaction conditions in preparation
for the kilolab campaign allowed replacement of toxic dioxane
with acetic acid as the reaction solvent and the use of a catalytic
amount of pTSA as the acid catalyst and achieved a virtually
quantitative conversion. The change also allowed the simplifica-
tion of the isolation: the product was precipitated by addition
of water and isolated by filtration.
Optimization of Reaction Conditions and the Scale-Up
Campaign. The procedure for Suzuki coupling was first
developed by the medicinal chemists and used 5 mol % of
Pd(PPh3)4 as the catalyst, K2CO3 as a base, and dimethoxyethane
as a solvent and afforded the coupled product in 43% yield
after chromatographic purification. A rapid screen of the
catalysts, bases, and reaction conditions allowed the achieve-
ment of complete conversion and 90% isolated yield using the
same catalyst but switching the base to NaOH. Higher reaction
temperature facilitated the conversion and also allowed lowering
the catalyst loading to 2 mol %. For product isolation,
dimethoxyethane was exchanged to heptane, and the insoluble
catalyst residue and byproducts were removed by filtration.
The best reagent for selective bromination of 13 found in
the reagent screen was NBS.17 Bromination with Br2 under a
variety of conditions lacked the selectivity (various amounts
of polybrominated side products formed) and suffered from
demethylation of the methoxy group as a persistent side reaction.
The reaction with NBS required the presence of an acid to effect
good conversion, but the conditions needed optimization to
avoid acid-induced demethylation. Sulfuric acid in dioxane at
In the kilolab campaign, it was possible to run the Suzuki
and bromination steps without isolation of the intermediate solid
material by conducting a solvent exchange from heptane to
acetic acid by vacuum distillation. The total isolated yield of
the brominated biaryl intermediate was 75-80%.
In the next steps of the synthetic sequence, generation of
the Grignard reagent and its reaction with glycidyl tosylate, two
major problems were encountered upon subjecting the biaryl
substrate 2d to the reaction conditions developed for the
monoaryl substrate 2c. First, the Grignard formation with
magnesium metal was not selective, and substantial amounts
of des-chloro- and des-dichloro derivatives were found upon
quenching the Grignard solution with water. The second
problem was the substantially lower reactivity of the Grignard
reagent toward the epoxide than what we observed for 2a or
2c. In fact, only about half of the epoxide reacted with the
Grignard reagent, while the rest was opened by the bromide
anion (when excess of the Grignard reagent was introduced into
the reaction).
To circumvent the first problem, an alternative method for
Grignard generation was employed: the aryl bromide was treated
with i-PrMgCl solution18 at 0 °C to room temperature (Scheme
9). The metal-halogen exchange was complete in 4 h and gave
the desired aryl Grignard solutions with no detectable side
(17) (a) Lambert, F. L.; Ellis, W. D.; Parry, R. J. J. Org. Chem. 1965, 30,
304. (b) Leazer, J. L.; Cvetovich, R.; Tsay, F.-R.; Dolling, U.; Vickery,
T.; Bachert, D. J. Org. Chem. 2003, 68, 3695.
1442
•
Vol. 14, No. 6, 2010 / Organic Process Research & Development

Scheme 9. Preparation of Grignard reagent, epoxide opening, and introduction of the phthalimide group
products. The new approach also solved the problem of the
epoxide opening by the bromide anion as it generated aryl
magnesium chloride instead of bromide which was obtained
in the reaction with magnesium metal.
It took substantial effort to optimize the conditions for the
Grignard addition to the epoxide to identify factors that
determine the outcome of the reaction. The major one was the
reaction temperature which had to be maintained in the -25 to
-20 °C window. Lower temperatures slowed down the reaction
dramatically (at -40 °C it never went past 50% conversion),
whereas higher temperatures led to significant side-product
formation.
The type or amount of the copper catalyst did not play an
important role in the epoxide opening except that without it
the reaction did not proceed at all. Little difference was noticed
among runs with equal loadings of CuCN, Li2CuCl4, and CuI.
Runs with 5 or 10 mol % of CuCN gave very similar results as
well. In laboratory preparation of smaller-scale batches, 7% of
CuCN was used because the commercially supplied material
is stable and nonhygroscopic and has fine particles; thus, it could
be introduced into the Grignard solution as a THF suspension
and could be dispersed easily in the reaction mixture. In the
kilolab runs, CuCN was replaced with CuI to avoid issues
associated with cyanide toxicity. There was no change in the
reaction performance. The copper reagent was used as received
from the vendor, and it was introduced into the Grignard
solution as a solid. However, we had to use a high impeller
speed to keep it from settling on the bottom of the reactor.
The reaction went to completion in 4-5 h at -26 to -22
°C as judged by the disappearance of glycidyl tosylate (a 5 mol
% excess of the Grignard reagent was used; all excess Grignard
byproducts were washed away in the workup). As expected,
the reaction resulted predominantly in hydroxytosylate with a
minor amount of the terminal epoxide (∼10 area % by HPLC).
The mixture could be converted completely into the terminal
epoxide by treatment with aqueous NaOH solution. The epoxide
could be isolated after extractive workup as an oil in a manner
similar to that described for the model substrate. Opening of
the epoxide was done in DMF and required a mixture of
phthalimide and its potassium salt. The reaction was complete
in about 10 h at 75 °C. After aqueous workup, the product was
crystallized by addition of heptane which also purged most of
the impurities carried from the previous steps. The yield was
74% based on glycidyl tosylate.
Alternatively, in a more streamlined procedure used on scale
in the kilolab, conversion to the terminal epoxide was omitted.
Epoxide opening was followed by the aqueous extractive
workup using toluene as an additional solvent to effect the phase
splits. The mixture of the hydroxy tosylate and spontaneously
formed terminal epoxide was isolated as a solution in toluene.
The solvent was changed to DMF using vacuum distillation.
The DMF solution was treated with potassium salt of phthal-
imide. Tosylate displacement in this case could proceed directly
or via the intermediate formation of terminal epoxide 5d induced
by the base. Regardless of the pathway, 2 equiv of the
phthalimide salt were necessary to force the reaction to
completion, the terminal epoxide being the last unreacted species
to disappear. For product isolation, the reaction mixture was
diluted with a mixture of isopropyl acetate (iPAc) and THF
and washed with a weak solution of NaOH in 8% aqueous NaCl
to remove byproduct phthalimide. The composition of the
mixture was carefully optimized to avoid precipitation of the
product, allow the split of the phases, and minimize formation
of emulsions. The temperature of the batch was lowered to 0-4
°C to prevent partial hydrolysis of the phthalimide and loss of
the product into the aqueous washes. The resulting solution was
azeotropically dried, which also removed most of THF (to
e0.1% by GC) and induced the start of product crystallization.
The precipitation was completed by addition of heptane to the
batch. The product was isolated by filtration in 50-60% yield
and 99.5% ee. Major impurities in the isolated product were
des-bromobiaryl byproduct 6d and the unreacted terminal
epoxide 5d (both below 1.5%). Lacking the incorporated amine
functionality (as protected phthalimide), both would be easily
purged when the final amine is converted to the HCl salt.
Sulfonylation of the hydroxyl in 7d (Scheme 10) was carried
out with methanesulfonyl chloride in the presence of triethy-
lamine. On scale, the reaction was carried out in THF which
allowed precipitation of the product by simple addition of water
(18) (a) Wakefield, B. J. Organomagnesium Methods in Organic Synthesis;
Academic Press: New York, 1995; Chapter 3.2.2; (b) Knochel, P.;
Dohle, W.; Gommermann, N.; Kniesel, F. F.; Kopp, F.; Korn, T.;
Sapountzis, I.; Vu, V. A. Angew. Chem., Int. Ed. 2003, 42, 4302.
(19) Kirby, A. J. AdV. Phys. Org. Chem. 1981, 17, 183. We thank a referee
for this suggestion.
(20) The two signals at 7.412 and 7.406 ppm correspond to the two meta-
protons of the dichlorophenyl ring. The small difference in chemical
shift (0.006 ppm) is most likely due to the restricted rotation around
the biaryl bond and the chirality of the molecule. This dissymmetry
of the ortho- and meta-positions of the dichlorophenyl was also
observed in the 13C spectrum of 7d and, to various extents, in the
1
spectra of 8d, 10d, and 1 × HCl, both H and 13C, described in the
following experiments.
Vol. 14, No. 6, 2010 / Organic Process Research & Development
•
1443

Scheme 10. Formation of the dihydrobenzofuran cycle, liberation of the amino group, and preparation of the HCl salt
to the reaction mixture and isolation of the solids by filtration
in 96% yield, 99.3% purity, and without any detectable loss of
enantiopurity.
showed that Form II was more stable in the latter set of solvents
as Form I was found to convert to Form II when suspended in
those solvents at 50 °C for a period of 48 h. Form II also had
slightly lower solubility in water (50 vs 67 mg/mL at 20 °C).
The more stable Form II has been selected as a drug substance.
Experiments designed to work out a process for preparation
of Form II showed that Form I could still precipitate from the
IPA solutions if supersaturation is not carefully controlled during
crystallization. For that reason, we decided not to use anti-
solvents in the process as mixing effects can create high local
concentrations of the antisolvent and induce formation of
Polymorph I. Isopropanol was found to have the most useful
solubility range for the crystallization of the hydrochloride (17
mg/mL at 20 °C and 38 mg/mL at 50 °C) among several
solvents screened, and this range could be greatly expanded
by adding water to the system. Figure 1 shows a dramatic
increase of solubility upon addition of water to the mixture.
However, the solubility difference at high and low temperature
in this solvent system was not large enough to afford good
recovery of the product. Therefore, partial solvent removal
needed to be incorporated into the process.
Under optimized conditions, the MTBE solution of the free
base obtained in the previous operation underwent solvent
exchange to IPA by vacuum distillation. An equimolar amount
of HCl in IPA was added followed by the calculated amount
of water (3 to 4% of the total IPA volume) to ensure complete
dissolution of the salt at 75 °C. The resulting mixture was heated
to 75 °C, cooled to 65-70 °C, and seeded with the hydrochlo-
ride Form II seeds. Once the seed bed was established, the batch
was further cooled to 30-40 °C. The partial vacuum was
applied to the reactor, and a portion of the solvent was distilled
off. Because IPA forms an azeotrope which contains 12 wt %
of water, distillation also resulted in a decrease of the relative
water content in the system, decreasing the solubility of the
salt. Finally, slow cooling of the batch to -10 °C and filtration
of the solids at that temperature afforded 83% yield of the
isolated 1 hydrochloride. The product had 99.5% chemical
As expected, demethylation of 8d occurred readily upon
treatment with boron tribromide at low temperature in meth-
ylene chloride. There was, however, a substantial difference in
behavior of the demethylated product under the reaction
conditions relative to that observed in the model case of 8c. In
the model case, the deprotected product 9c was isolated, and
cyclization was done in a separate step by treatment with
potassium carbonate. Here, the cyclization occurred immediately
upon demethylation. In fact, we attempted to intercept the open-
chain intermediate by quenching the reaction at the half-
conversion point, but only the cyclized product 10d and the
starting methoxyaryl compound 8d could be detected. We
assume the steric repulsion between the dichlorophenyl sub-
stitutent and the phenol led to an enhanced rate of cyclization
relative to the unsubstituted system.19
Further study of reaction conditions allowed replacement of
dichloromethane with toluene. With 1.6 equiv of BBr3, at room
temperature, the reaction was complete in 20 h. For isolation,
methanol was added to the reaction mixture, the volatile
byproducts were removed by distillation, and the product was
isolated by filtration in 79% yield. A slight loss of enantiomeric
purity was observed in the kilolab run: the ee of the isolated
batch was 97.4%, down from 98.6% in the starting material.
The phthalimide protecting group was removed by heating
the substrate with hydrazine hydrate in an ethanol-water
mixture at reflux. Presence of water in the reaction mixture
increased the solubility of phthalylhydrazide and prevented it
from precipitating and forming a thick suspension. The reaction
went to completion in 2 h. The product was extracted into
MTBE, and the solution was washed with dilute aqueous NaOH
to remove phthalylhydrazide (for safety considerations, hydra-
zine levels were monitored in the organic phase in the plant)
and to obtain an MTBE solution of the product as a free base.
Hydrochloride Salt: Polymorphs, Properties, Prepara-
tion. The hydrochloride salt has been selected as the API form
on the basis of its high crystallinity, good solubility in water,
good stability and low hygroscopicity.
A quick and non-exhaustive polymorph screen of the
hydrochloride allowed identification of two forms with very
similar melting points (both had melting endotherms at 134 °C
in DSC scans) but manifestly distinct XRD patterns. Form I
was forming predominantly in solvents in which the hydro-
chloride had low solubility. e.g., MTBE, EtOAc, toluene,
heptane. Form II formed in acetone, water, or IPA where
solubility of the salt was substantial. Stability studies also
Figure 1. Solubility of 1 HCl salt in IPA/water mixture as a
function of water content at 20 °C.
1444
•
Vol. 14, No. 6, 2010 / Organic Process Research & Development

purity and 98.7% ee (a slight increase from 97.4% of the product
in the previous step). The hydrazine level in the isolated solid
was determined at less than 5 ppm, making the material
appropriate for use in the clinical program.
The residual solvent was chased by acetic acid under reduced
pressure (2 × 42.6 kg, distilled down to 30-35 L at 20-36
°C after each charge). NBS (21.5 kg, 121 mol), pTSA (3.23
kg, 17.0 mol), and acetic acid (106 kg) were added to the
concentrate, and the resulting mixture was heated at 55 °C for
22 h. HPLC assay of the reaction mixture showed 99.5%
conversion of the biaryl Suzuki product to the bromide. A
solution of sodium metabisulfite (5.7 kg) in water (20.1 kg)
was added to the reaction mixture, maintaining the batch
temperature between 50 and 60 °C. After ensuring the negative
reaction of the mixture to the starch iodide test, the batch was
cooled to 36 °C, seeded, and held at 30-35 °C for 1 h to form
the crystal bed. The mixture was cooled further to 22 °C over
2 h, and then water (28.2 kg) was added over 35 min. The
resulting slurry was held for 1 h at 20 °C, and then the solid
was filtered and washed with water (3 × 21 kg). The cake was
reslurried with 100 kg of water, filtered again, and washed with
additional amounts of water (3 × 25 kg) to remove residual
acetic acid. The wet product was dried in a vacuum oven at 40
°C until LOD (1.5 h at 50 °C and e10 mmHg) of 0.5% or
lower was reached. The dry solid was tested for levels of
residual acetic acid (0.054% by GC) and water (0.07% by KF
titration) and was acceptable for use in the next step. Yield 24.2
kg (79.3%). For a purified sample, 1H NMR (300 MHz, CDCl3)
δ: 7.42 (m, J ) 8.1 Hz, 2H), 7.39 (dd, J ) 3.0, 7.7 Hz, 1H),
7.30 (dd, J ) 8.1 Hz, 1H), 6.86 (dd, J ) 3.0, 8.0 Hz, 1H), 3.56
(s, 3H). 13C NMR (75 MHz, CDCl3), δ: 158.2 (d, J ) 248
Hz), 151.2 (d, J ) 3 Hz), 135.1, 135.0, 133.0 (d, J ) 9 Hz),
130.0, 128.0, 120.8 (d, J ) 25 Hz), 117.7 (d, J ) 11 Hz), 117.0
(d, J ) 23 Hz), 60.9.
2S-3-[5-Fluoro-3-(2,6-dichlorophenyl)-2-methoxyphenyl]-1-
N-phthalimidopropan-2-ol (7d). A 2 M solution of i-PrMgCl
in THF (14.8 kg, 28.9 mol) was added to a solution of 14 (9.5
kg, 27.1 mol) and THF (33.8 kg), maintaining the batch
temperature at -7 to -2 °C. The reaction mixture was held at
0 °C for 2 h. The completion of transmetalation was established
by HPLC analysis of an aliquot of the reaction mixture
quenched by water for the absence of the starting bromide. Once
the level of the starting material was confirmed to be below
1%, the reactor contents were cooled to -30 to -25 °C, and
CuI (0.36 kg, 1.9 mol) was added to the Grignard solution. The
mixture was stirred at that temperature for 30 min, and then a
solution of glycidyl tosylate (5.90 kg, 25.9 mol) in THF (5.8
kg) was added over 1.5 h, while maintaining the batch
temperature in the range of -30 to -25 °C. The resulting
mixture was stirred at -26 to -22 °C for 3 h and assayed by
HPLC for the amount of unreacted glycidyl tosylate. Reaction
was considered complete when the level fell below 2.0 area
%. The batch was quenched with an aqueous solution of
ammonium chloride (39.6 kg) (solution was prepared from 7.5
kg of NH4Cl and 44.6 kg of water). The batch temperature
during the addition was allowed to rise from -22 to 6 °C. The
temperature was then adjusted to 5-15 °C, and the aqueous
layer was separated and back extracted with toluene (26.6 kg).
The combined organic layers were washed with the aqueous
solution of ammonium chloride (12.5 kg) and water (4 × 14.4
kg, water washes were repeated until neutral pH of the wash)
and filtered through a 2-µm cartridge filter. The solvents were
Conclusions
A new process for synthesis of the 5-HT2C agonist 1 in
optically pure form has been developed and implemented on
scale for preparation of 10 kg of the API in support of the
clinical studies. The key step of the process, the opening of
epoxide with aryl Grignard reagent catalyzed by a copper(I)
salt allowed us to incorporate a commercially available chiral
fragment into the target molecule.
Experimental Section
General. Starting materials, solvents, and reagents were
obtained from commercial sources and were used as received
without purification. HPLC assays were performed on Agilent
1100 chromatographs equipped with PDA detectors. Routine
reaction monitoring was done using a Phenomenex Prodigy
ODS3 4.6 mm × 50 mm column or an equivalent C-18 column
and a standard 8-min gradient program: 10:90 to 90:10 mixture
of acetonitrile-water containing 0.02% of TFA with constant
flow of 1 mL/min. HPLC methods for large-scale reaction
completion assays are given in the Supporting Information.
HPLC determination of the API chiral purity was done using a
Daicel Chiralcel OD-RH, 5 µ, 150 mm × 4.6 mm column,
and an isocratic mixture of 0.05 M solution of KPF6 in water
and acetonitrile in 60:40 ratio with 1 mL/min flow rate (data
collected at 220 nm). RRTs: 0.80 (distomer), 1.00 (eutomer).
NMR spectra were recorded on a Bruker Avance DPX300 or
a Bruker Avance DRX400 NMR spectrometer. Spectra were
referenced by TMS. LC/MS data were obtained on an Agilent
1100 LC system with an Agilent 1100 LC/MS detector. HPLC
conditions: Phenomenex Prodigy ODS3 4.6 mm × 50 mm
column, 90:10 to 10:90 8-min gradient of water/acetonitrile
containing 0.02% HCO2H, flow rate 1 mL/min.
1-Methoxy-2-bromo-4-fluoro-2′,6′-dichlorobiphenyl (14). To
a mixture of 2-bromo-1,3-dichlorobenzene (18.0 kg, 78.8 mol),
5-fluoro-2-methoxyphenylboronic acid (17.6 kg, 92.1% strength,
95.8 mol), Pd(PPh3)4 (1.85 kg, 1.60 mol), and dimethoxyethane
(157 kg) maintained at 70-75 °C was added a solution of
NaOH (15.9 kg, 398 mol) in water (90 kg) over 30 min. The
resulting mixture was heated at reflux (80 °C) for 18 h at which
point the HPLC assay of the reaction mixture showed no
detectable amount of the starting aryl bromide. The batch was
cooled to 20 °C, the layers were separated. The organic layer
was concentrated in vacuum at 20-30 °C to a volume of 30
L. Heptane (123 kg) and water (27.0 kg) were added to the
concentrate, the mixture was filtered through a bag filter to
remove solids, and the phases were separated. The organic
solution was washed with water (2 × 27.0 kg) and then
concentrated under vacuum at 17-21 °C to a final volume of
105 L. Silica gel (Merck, grade 60 mesh 200-400, 3.6 kg)
was added to the concentrated solution, and the slurry was
stirred for 2 h. The silica was filtered off and washed with
heptane (80 kg). The combined filtrate and wash were concen-
trated under reduced pressure at 16-20 °C to a 30-L volume.
Vol. 14, No. 6, 2010 / Organic Process Research & Development
•
1445

distilled off under vacuum (batch temperature 15-23 °C) to
the volume of 18 L. The residual solvents were chased with
toluene (15.4 kg, same conditions as above) and then DMF
(35.6 kg). The final distillation went to the final volume of 43
L and was done at 40-50 °C.
were filtered and washed with a mixture of THF (18 kg) and
water (20 kg), then with water (3 × 55 kg), and finally with
heptane (38 kg). The solid was dried on trays in a vacuum oven
at 70 °C until the LOD (1 h at 85 °C and e10 mmHg) of 0.5%
or less was met. Yield 24.9 kg (96%), strength 97.5%, total
impurities 0.71% (by HPLC), enantiomeric purity 99.0% ee.
Mp > 200 °C dec. 1H NMR (400 MHz, dmso-d6) δ: 7.87 (m,
2H), 7.83 (m, 2H), 7.614 (m, J ) 1.09, 8.13 Hz, 1H), 7.619
(m, J ) 1.09, 8.12 Hz, 1H), 7.50 (m, J ) 8.13, 8.12 Hz, 1H),
7.40 (dd, J ) 3.3, 9.3 Hz, 1H), 7.06 (dd, J ) 3.3, 8.6 Hz, 1H),
5.13 (m, 1H), 3.97 (dd, J ) 9.2, 14.9 Hz, 1H), 3.74 (dd, J )
3.0, 14.9 Hz, 1H), 3.37 (s, 3H), 3.18-3.14 (m, 2H), 2.84 (s,
3H). 13C NMR (100 MHz, dmso-d6) δ: 167.6, 157.6 (d, J )
242 Hz), 152.4 (d, J ) 2 Hz), 134.8, 134.42, 134.37, 134.26,
131.6, 131.5 (d, J ) 10 Hz), 131.4 (d, J ) 8 Hz), 130.8, 128.3,
123.1, 118.7 (d, J ) 22 Hz), 116.7 (d, J ) 24 Hz), 78.5, 60.5,
40.8, 37.6, 33.2.
Potassium salt of phthalimide (7.50 kg, 40.5 mol) was added
to the batch at 20 °C, and then the mixture was heated at 86
°C for 12 h at which point the HPLC analysis of the reaction
mixture showed 2.4 area % of the unreacted terminal epoxide
5c. The mixture was cooled to 10-15 °C, and iPAc (42.0 kg),
THF (10.5 kg), and water (55.0 kg) were added (exotherm was
observed during water addition). The layers were split, the
aqueous layer was back extracted with iPAc (38 kg), and the
organic layers were combined (10.5 kg of THF was added as
a reactor rinse). The resulting solution was washed at 4 °C with
a solution (2 × 38 kg) prepared from NaCl (7.5 kg), NaOH
(0.76 kg), and water (93 kg), then twice with a mixture of water
(40 kg) and THF (10.5 kg). A small amount of brine was added
to the batch during the final washes to aid phase separation.
The resulting solution was distilled under vacuum at 10-20
°C to the volume of 40 L. The product partially crystallized
from solution during distillation. The residual solvents were
chased with iPAc (2 × 37.7 kg), distilling the batch to the same
mark under similar conditions. Heptane (28.8 kg) was added
to the concentrated slurry at 30 °C over 30 min; the slurry was
cooled to 20 °C, and the solids were filtered. The cake was
washed with a 1:1 v/v mixture of iPAc and heptane (2 × 16.2
kg), and the solid was dried on the trays in a vacuum oven at
45 °C until LOD (1 h at 70 °C and e10 mmHg) of 0.5% or
less was obtained. Yield 6.67 kg (51% based on glycidyl
tosylate as the limiting reagent), total impurities 1.13% (by
2R-7-(2,6-Dichlorophenyl)-5-fluoro-2,3-dihydro-2-(N-phthal-
imidomethyl)benzofuran (10d). 8d (24.8 kg, 44.5 mol) was
suspended in toluene (217 kg), and neat BBr3 (12.4 kg, 49.4
mol) was added to the suspension over 33 min at 17-20 °C
(weak exotherm). The initial suspension turned into a red
solution during the addition, and it was stirred in the range of
18-25 °C for 12 h. HPLC assay of a reaction mixture sample
showed 31% of unreacted starting material. After holding further
for 8 h, additional BBr3 (2.9 kg, 11.6 mol) was added, and the
reaction mixture was held longer while reaction progress was
monitored by HPLC. It took 16 h after the last BBr3 addition
for the level of the starting material to drop to 1.7%. At that
point anhydrous methanol (19.8 kg) was added to the batch,
maintaining the temperature below 20 °C (addition was ac-
companied by a strong exotherm; the jacket had to be cooled
to -18 °C to finish addition in 30 min and maintain the
temperature in the desired range). The resulting mixture was
held at 20 °C for 1 h, and then the solvents were distilled under
vacuum to the volume of 105 L, maintaining the batch
temperature at 20-25 °C. MeOH (39.0 kg) was added to the
batch, and the resulting mixture was heated to 70 °C and then
cooled to 0-5 °C over 1 h. The crystallized solid was filtered,
washed with cold (2-8 °C) MeOH (2 × 20 kg), and dried on
the filter at 65 °C with agitation until LOD (85 °C, e 10 mmHg,
1 h) of e0.5% was reached. Yield 15.8 kg (79%), strength
(HPLC) 99.7%, total impurities 0.36%, enantiomeric purity
1
HPLC), enantiomeric purity 99.0% ee. Mp 165-168 °C. H
NMR (300 MHz, CDCl3) δ: 7.86 (m, 2H), 7.72 (m, 2H), 7.412
(m, J ) 1.2, 8.1 Hz, 1H), 7.406 (m, J ) 1.2, 8.1 Hz, 1H), 7.27
(m, J ) 8.1, 8.1 Hz, 1H),20 7.08 (dd, J ) 3.0, 8.8 Hz, 1H),
6.79 (dd, J ) 3.0 Hz, 8.1 Hz, 1H), 4.23 (d5, J ) 3.3, 4.3, 5.7,
7.9, 8.5 Hz, 1H), 3.85 (dd, J ) 3.3, 14.1 Hz, 1H), 3.80 (dd, J
) 8.5, 14.1 Hz, 1H), 3.42 (s, 3H), 2.96 (dd, J ) 4.3, 13.9 Hz,
1H), 2.92 (dd, J ) 7.9, 13.9 Hz, 1H), 2.80 (d, J ) 5.7 Hz, 1H).
13C NMR (CDCl3, 75 MHz), δ: 168.7, 158.4 (d, J ) 244 Hz),
152.2, 135.7, 135.3 (d, J ) 9.0 Hz), 134.1, 132.9 (d, J ) 8.5
Hz), 132.0, 131.8 (d, J ) 9.6 Hz), 129.7, 128.0, 123.4, 118.2
(d, J ) 23 Hz), 116.4 (d, J ) 23 Hz), 70.6, 60.9, 44.1, 36.3.
A total of four batches were run without significant differ-
ences, and the yields were in the 51-58% range (adjusted for
strength).
1
97.8%. Mp 222.5-224.5 °C. H NMR (400 MHz, dmso-d6)
δ: 7.85 (m, 4H), 7.54 (dd, J ) 1.1, 8.1 Hz, 1H), 7.53 (dd, J )
1.1, 8.2 Hz, 1H), 7.41 (dd, J ) 8.1, 8.2 Hz, 1H), 7.19 (dd, J )
2.7, 8.2 Hz, 1H), 6.86 (dd, J ) 2.7, 9.3 Hz, 1H), 5.09 (m, 1H),
3.79 (m, 2H), 3.43 (dd, J ) 9.3, 16.6 Hz, 1H), 3.15 (dd, J )
5.9, 16.6 Hz, 1H). 13C NMR (100 MHz, dmso-d6) δ: 167.8,
156.4 (d, J ) 237 Hz), 152.2, 134.6, 134.5, 134.3, 133.5, 131.5,
130.6, 128.6 (d, J ) 9.4 Hz), 128.1, 123.1, 118.2 (d, J ) 9.5
Hz), 114.9 (d, J ) 25 Hz), 112.9 (d, J ) 25 Hz), 80.0, 41.1,
33.2.
2S-3-[5-Fluoro-3-(2,6-dichlorophenyl)-2-methoxyphenyl]-1-
N-phthalimidopropan-2-yl Methanesulfonate (8d). Methane-
sulfonyl chloride (7.9 kg, 68.9 mol) was added over 40 min to
a stirred mixture of 7d (21.9 kg, 46.2 mol), THF (98 kg), and
Et3N (7.0 kg, 69.3 mol), keeping the batch temperature in the
range of 9-17 °C (strong exotherm). The batch temperature
was adjusted to 20-26 °C, and the mixture was held under
these conditions for 1.5 h. A thick slurry formed in the reactor
during the hold. The batch was analyzed by HPLC for
conversion (0.24 area % of the unreacted starting material was
detected). Water (110 kg) was added to the reactor over 30
min while the batch was maintained at 20-22 °C, and the solids
2R-7-(2,6-Dichlorophenyl)-5-fluoro-2,3-dihydro-2-aminom-
ethylbenzofuran hydrochloride (1 × HCl). Hydrazine hydrate
(5.31 kg, ∼104 mol) was added to a reactor containing a mixture
of 10d (15.3 kg, 34.7 mol), ethanol (36.2 kg), and water (30.6
kg). The resulting mixture was heated at 74-84 °C for 2 h.
1446
•
Vol. 14, No. 6, 2010 / Organic Process Research & Development

Once HPLC assay confirmed completion of the reaction, the
batch was cooled to 45 °C, and water (45.9 kg) and MTBE
(34.0 kg) were added. The mixture was further cooled to 20
°C, the layers were separated, and the aqueous layer was
extracted with MTBE (2 × 17.0 kg). The combined organic
phases were washed with a dilute aqueous solution of NaOH
(2 × 23.1 kg, wash solution was prepared from 0.455 kg of
NaOH and 45.9 kg of water) and then with water (2 × 30.6
kg). The volume of the resulting solution was reduced to 20-24
L by vacuum distillation at 10-20 °C, then the residual solvent
was chased with 2 portions of IPA (53.9 kg each) distilling the
solvent out after each IPA charge to the 20-24 L mark. The
batch was mixed with additional IPA (13.7 kg), and the solution
was filtered through a 10-µm cartridge filter followed by a 0.2-
µm cartridge filter, using 3.0 kg of IPA as a rinse.
All liquid reagents added to the batch after this point were
filtered through a 0.2-µm cartridge filter before the addition. A
solution of HCl in IPA (prepared by dissolution of 3.24 kg of
HCl gas in 18.0 kg of IPA) was added to the solution of the
free base, maintaining the temperature below 25 °C, followed
by 2.05 kg of water. The resulting mixture was heated to 75-78
°C and then, once the complete dissolution of all solids was
visually confirmed, was cooled to 70 °C. The batch was seeded
with Form II of 1 hydrochloride (10 g) at 70 °C, held for 30
min at 70 °C, cooled to 65 °C, and seeded again with the same
amount of seed. The mixture was cooled to 55 °C in 1 h, held
for 1 h, and sampled, and the solids were analyzed by XRPD
to ensure that the correct solid form was crystallizing. The
cooling continued to 30 °C in 1 h; then the vacuum was applied
to the reactor, and the solvent was distilled off until the batch
volume reached 36-40 L (26-36 °C batch temperature). IPA
(42.2 kg) was added to the batch, and the distillation was
repeated to the same mark. The resulting slurry was cooled to
-10 °C over 1 h, held for 1 h, and then filtered at that
temperature. The filter cake was washed with cold (2-8 °C)
IPA and dried first on the filter in the nitrogen flow and then
on trays in vacuum at 55 °C until LOD (65 °C, e10 mmHg,
1 h) of e0.5% was reached. Yield 9.92 kg (83% based on the
amount of 10d), strength 99.5% as hydrochloride salt, chiral
purity 98.6% ee, total impurities 0.21% by HPLC. Mp 134 °C
(by DSC). 1H NMR (400 MHz, dmso-d6) δ: 8.25 (broad s, 3H),
7.57 (apparent d, J ) 8.2 Hz, 2H), 7.45 (dd, J ) 7.8, 8.4 Hz,
1H), 7.24 (dd, J ) 2.6, 8.1 Hz, 1H), 6.90 (dd, J ) 2.6, 9.6 Hz,
1H), 5.05 (d4, J ) 9.2, 7.9, 7.0, 4.5 Hz, 1H), 3.45 (dd, J ) 9.2,
16.6 Hz, 1H), 3.17 (dd, J ) 7.0, 16.6 Hz), 3.10 (dd, J ) 13.4,
4.5 Hz, 1H), 3.04 (dd, J ) 13.4, 7.9 Hz, 1H). 13C NMR (100
MHz, dmso-d6) δ: 156.4 (d, J ) 257 Hz), 151.9, 134.6, 134.3,
133.6, 130.6, 128.9 (d, J ) 11 Hz), 128.4, 128.2, 118.4 (d, J )
9 Hz), 115.1 (d, J ) 25 Hz), 113.0 (d, J ) 25 Hz), 80.0, 42.1,
32.8.
Acknowledgment
We thank colleagues in Discovery Analytical Chemistry
group (Mairead Young) and Analytical Quality Sciences
(Pasquale Crecca) for analytical support and help with chiral
analytical methods. Help from Asaf Alimardanov and Karen
Sutherland with review of the manuscript is greatly appreciated.
Supporting Information Available
1H and 13C NMR spectra of the API and the isolated
intermediates, XPRD patterns of the API polymorphs, HPLC
methods for reaction assays. This material is available free of
charge via the Internet at http://pubs.acs.org.
Received for review August 25, 2010.
OP100233F
Vol. 14, No. 6, 2010 / Organic Process Research & Development
•
1447