Synthesis of the NK1 receptor antagonist GW597599. Part 3: Development of a scalable route to a key chirally pure arylpiperazine urea, a happy end

Article abstract of DOI:10.1021/op9002032

GW597599 1 is a novel NK-1 antagonist currently under investigation for the treatment of central nervous system disorders and emesis. The initial chemical development synthetic route, derived from the one used by medicinal chemistry, involved several haza

Full text of DOI:10.1021/op9002032

Organic Process Research & Development 2009, 13, 1100–1110
Synthesis of the NK1 Receptor Antagonist GW597599. Part 3: Development of a
Scalable Route to a Key Chirally Pure Arylpiperazine Urea, A Happy End
Giuseppe Guercio,*,† Sergio Bacchi,*,† Alcide Perboni,† Corinne Leroi,§ Francesco Tinazzi,† Ilaria Bientinesi,† Marie Hourdin,⊥
Michael Goodyear,† Stefano Curti,† Stefano Provera,‡ and Zadeo Cimarosti†
Chemical DeVelopment and Analytical Chemistry, GlaxoSmithKline Medicines Research Centre, Via Fleming 4,
37135 Verona, Italy, Sanofi-AVentis R&D Centre, 195 route d’Espagne BP 13669, 31036 Toulouse Cedex 1, France, and
Pharmasynthese, Process Industrialization - 57 rue GraVetel, 76320 Saint Pierre le`s Elbeuf, France
Abstract:
2. Initial Synthetic Route
The synthetic route used for the synthesis of the first few
grams of GW597599 is shown in Scheme 1.
2.1. Experimental Details. The key stage to access the final
compound was the coupling between the racemic ketopiperazine
7 and the racemic benzylamine 12 to give the hindered
quaternary urea 8 via the synthesis of the carbamoyl chloride
intermediate 13.
GW597599 1 is a novel NK-1 antagonist currently under investiga-
tion for the treatment of central nervous system disorders and
emesis. The initial chemical development synthetic route, derived
from the one used by medicinal chemistry, involved several
hazardous reagents, gave low yields and produced high levels of
waste. Through a targeted process of research and development,
application of novel techniques and extensive route scouting, a new
synthetic route for GW597599 was developed. This paper reports
the optimisation work of the third and last stage in the chemical
synthesis of GW597599 and the development of a pilot-plant-
suitable process for the manufacturing of optically pure arylpi-
perazine derivative 1. In particular, the process eliminated the use
of triphosgene in the synthesis of an intermediate carbamoyl
chloride, substantially enhancing safety, overall yield, and
throughput.
To prepare this reactive acyl chloride, the ketopiperazine was
treated with the highly toxic triphosgene,2 in a mixture of
dichloromethane and triethylamine followed by addition of
benzylamine 12 plus an auxiliary base.
1. Introduction
The addition of the benzylamine was followed by a solvent
swap between dichloromethane and acetonitrile, allowing
shortening of the reaction time by performing the coupling at
the acetonitrile refluxing temperature. The base was needed to
capture the hydrochloric acid evolved in the step. However, to
avoid the risk of losing the triethylamine during the solvent
exchange, the higher-boiling N,N-diisopropylethylamine was
also added. The amide group in the piperazine ring was then
reduced by treatment with a solution of borane in THF at reflux.
The resulting crude product was purified by chromatography
to isolate a single diastereoisomer of 1 which underwent a
further classical resolution procedure to obtain the optically pure
GW597599. The final compound was then crystallized as its
pharmaceutically active acetate salt.
2.2. Initial Evaluation of a Possible Enantiospecific Ap-
proach. It was evident that the possibility to access the
enantiopure ketopiperazine 73 and benzylamine 12 would have
produced a single diastereoisomer 14. However, the attempt to
reduce the amide group in the piperazine ring by treatment with
a solution of borane in THF at reflux led to the partial
epimerization at the chiral centre of this ring (Scheme 2). The
proton R to the ureidic moiety, being benzylic in nature, is in
fact very acidic.
Substance P (SP) is a member of the neurokinin family that
exerts its pleiotropic role by preferentially binding to the
neurokinin receptor classified as NK1.1 Substance P and the
NK1-receptors that mediate its activity are present in the brain
stem centres that elicit the emetic reflex. Nausea and vomiting
have been reported as the most distressing side effects of
chemotherapy, and the disruptive effects of these symptoms
on patients’ daily lives have been well documented. In light of
the need for continued routine use of emetogenic chemotherapy,
effective prevention of chemotherapy-induced nausea and
vomiting (CINV) is a central goal for physicians administering
cancer chemotherapy. GW597599 1 is a novel NK-1 antagonist
currently being developed to relieve these symptoms. This and
other NK-1 antagonists have also been investigated for the
treatment of CNS disorders such as depression. In this paper
we describe the successful efforts to define a synthetic route
for large-scale manufacturing.
* Authors to whom correspondence may be sent. E-mail: giuseppe.2.guercio@
gsk.com; sergio.k.bacchi@gsk.com.
† Chemical Development, GlaxoSmithKline Medicines Research Centre.
‡ Analytical Chemistry, GlaxoSmithKline Medicines Research Centre.
§ Sanofi-Aventis R&D Centre.
⊥ Pharmasynthese.
(1) Di Fabio, R.; Griffante, C.; Alvaro, G.; Pentassuglia, G.; Pizzi, D. A.;
Donati, D.; Rossi, T.; Guercio, G.; Mattioli, M.; Cimarosti, Z.;
Marchioro, C.; Provera, S.; Zonzini, L.; Montanari, D.; Melotto, S.;
Gerrard, P. A.; Trist, D. G.; Ratti, E.; Corsi, M. J. Med. Chem. 2009,
52, 3238–3247.
(2) Rosiak, A.; Hoenke, C.; Christoffers, J. Eur. J. Org. Chem. 2007, 26,
4376–4382.
(3) Guercio, G.; Bacchi, S.; Goodyear, M.; Carangio, A.; Tinazzi, F.; Curti,
S. Org. Process Res. DeV. 2008, 12, 1188–1194.
1100
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development
10.1021/op9002032 CCC: $40.75  2009 American Chemical Society
Published on Web 10/02/2009

Scheme 1. Initial synthetic route
Scheme 2. Enantiospecific version of the initial synthetic route
3. Development of a Convergent Process
Scheme 3. New potential route
3.1. Synthetic Strategy. An alternative process to avoid this
partial epimerization was envisaged. The disconnection of 1
into the two chiral fragments 16 and 18,4 shown in Scheme 3,
provided the basis of a more efficient route. This encompassed
the need for the last-stage resolution and addressed the issue
of the partial racemisation in the final chemical step caused by
borane.
3.2. Development of a Selective Arylpiperazine Nitrogen
Monoprotection. The synthesis of the carbamoyl chloride of
the piperazine dihydrochloride 18 instead of the keto-derivative
14 required the use of a protecting group to discriminate
between the two nitrogen atoms (Scheme 4). Three different
protecting groups were tried: 4-nitrobenzylchloroformate (4-
(4) Guercio, G.; Manzo, A. M.; Goodyear, M.; Bacchi, S.; Curti, S.;
Provera, S. Org. Process Res. DeV. 2009, 489–493.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1101

Figure 1. Optical microscopy (polarized light 125×) of the two different batches of piperazine dihydrochloride 18.
Scheme 4. Selective nitrogen monoprotection
To understand the lack of control in selectivity, the batch of
starting piperazine dihydrochloride 18 with very large crystals
was suspended in ethyl acetate at 25 °C, treated with triethyl-
amine and Boc2O portionwise. To be noted is that a clear
solution was not observed at this point because both the
piperazine and Et3N·HCl are insoluble solids in ethyl acetate.
Thus, the acid exchange should occur only on the surface of
the crystals, driving a partial conversion to the monoprotected
compound 19. Due to its higher solubility, the compound 19
converted into the double-protected compound 21 by reaction
with the Boc2O still present. Several alternatives were tested
trying to improve the process: increasing the amount of ethyl
acetate to solubilise the piperazine 18; changing the solvent and
increasing the temperature. More than 100 volumes of ethyl
acetate were required to get complete solubilisation. Since the
undesirable dichloromethane was the only solvent better than
ethyl acetate in solubilising piperazine 18, it was decided to
operate on the temperature. On stirring 18 in the presence of
triethylamine for 2 h at 40 °C, cooling back to 20 °C and adding
Boc2O, the expected ratio ca. 95:5 mono versus diprotected was
obtained. The stirring period at 40 °C seemed to be fundamental
for breaking down the bigger crystals achieving a better
dissolution rate, increasing the availability of the substrate and
improving the acid exchange between 18 and the triethylamine.
So, without the need of an extra particle size reduction step,
the double protection was limited.
3.3. Screening of Coupling Agents. Using 19 as the starting
material for the downstream chemistry, we investigated possible
coupling agents (Scheme 5). In particular carbonyl-di-imida-
zole,8 bis(p-nitrophenyl) carbonate,9 triphosgene,10 and phos-
gene11 were tested.
On reacting 19 with carbonyl-di-imidazole the corresponding
acylimidazole was obtained in quantitative yields, but this did
not react directly with 16 if the imidazole moiety was not
preliminary quaternised with iodomethane. The reaction proved
to be extremely difficult to interpret by HPLC and afforded 20
in only 35% a/a. A similar behaviour was observed with bis(p-
nitrophenyl) carbonate; the intermediate carbonate was obtained
with an almost quantitative yield, but the subsequent reaction
Table 1. Comparison between different nitrogen protecting
groups
cost for 1
protecting
groupa
area %
area %
area %
mmol of
monoprotected diprotected starting 18 protecting agentb
4-NBCF
CbzCl
Boc2O
93.0
77.4
94.0
3.2
8.6
5.0
3.8
14.0
1.0
1.15
0.06
0.39
a Note: the protecting agents were used equimolar to the starting material in the
same dilution conditions. b In euros. Raw material prices were referred to Aldrich
catalogue 2009-2010.
NBCF),5 benzylchloroformate chloride (CbzCl)6 and di-tert-
butyl-dicarbonate (Boc2O).7 Results for each of those are shown
in Table 1.
The balance between selectivity, cost, and ease of removal
pointed us toward the Boc2O.
Surprisingly, when the optimised conditions were applied
to two different batches of piperazine 18, in one case we
observed the expected ratio of 94:6 of 19:21, while in the other
one a 75:25 ratio was found, with no evident explanation for
the decrease in selectivity. Cooling the reaction mixture led to
an even worse ratio of about 50:50, with a large amount of
unreacted starting piperazine 18 recovered after aqueous
workup. A microscopy study highlighted a difference between
the two batches that was considered of use in the rationalisation
process. The batch that reacted as expected had a smaller particle
size distribution with respect to the batch that had a higher
amount of diprotected derivative 21, which showed the presence
of very large crystals (Figure 1).
(5) Shields, J. E.; Carpenter, F. H. J. Am. Chem. Soc. 1961, 83, 3066–
3070.
(6) Atwell, G. J.; Denny, W. A. Synthesis 1984, 12, 1032–1033.
(7) Houghten, R. A.; Beckman, A.; Ostresh, J. M. Int. J. Pept. Protein
Res. 1986, 27 (6), 653–658.
(9) Izdebski, J.; Pawlak, D. Synthesis 1989, 6, 423–425.
(10) Jacobsen, E. J.; Stelzer, L. S.; Belonga, K. L.; Carter, D. B.; Im, W. B.;
Sethy, V. H.; Tang, A. H.; VonVoigtlander, P. F.; Petke, J. D. J. Med.
Chem. 1996, 39 (19), 3820–3836.
(11) Zeidan, T. A.; Wang, Q.; Fiebig, T.; Lewis, F. D. J. Am. Chem. Soc.
2007, 129 (32), 9848–9849.
(8) Lam, P. Y. S.; Ru, Y.; Jadhav, P. K.; Aldrich, P. E.; DeLucca, G. V.;
Eyermann, C. J.; Chang, C.-H.; Emmett, G.; Holler, E. R.; et al. J. Med.
Chem. 1996, 39 (18), 3514–3525.
1102
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 5. Coupling
the benzylamine 16) we carried out a design of experiment
(DoE): 30 experiments; central composite design; factors:
equivalents of triphosgene from 0.35 to 0.45, volumes of the
solution of protected piperazine 19 from 2 to 10, equivalents
of benzylamine 16 from 1.2 to 1.5, and volumes of the solution
of benzylamine 16 from 5 to 15. In the end we expected to
achieve also a better yield.
By interpreting the results of the DoE study we were able
to understand the impact of the equivalents of triphosgene on
the self-condensation of benzylamine into the bis-urea 27, and
we were able to identify the best reaction conditions to improve
both yield and quality (0.38 equiv of triphosgene, 1.4 equiv of
benzylamine 16, amount of solvent not a critical factor (Figure
2)). Having identified the best conditions to minimize the
impurity 27, we further investigated the fate of the other
impurities (Scheme 6). For example, the incomplete Boc
protection of piperazine 18, as made explicit in section 3.2, led
to the compound 25, due to the reaction of both piperazine
nitrogens with triphosgene followed by the coupling with two
benzylamine molecules. Starting piperazine 18 was found in
the final isolated product as a result of deprotection of the di-
Boc piperazine 21 still present at the end of the coupling. Large
amounts of benzylamine gave higher conversion to the desired
urea, but the excess of the reagent led to the formation of two
other impurities, 25 and 26. The triethylamine adduct 26 was
also observed with a long reaction time at high temperature.
Interestingly enough, in the same conditions we also identified
an impurity, 28, possibly generated by the reaction of benzyl-
amine with triphosgene followed by the reaction with Et3N and
subsequent loss of ethylene, the same as for the formation of
26. A low amount of triphosgene, potentially due either to
incomplete addition of reagent or a too-high level of quenching
water, led to another bis-urea 24, Via the coupling of the
carbamoyl chloride 22 with the mono-Boc piperazine 19.
Following deprotection, 24 led to 29. A short reaction time
meant incomplete conversion of the carbamoyl chloride 22 that
reacted with the desired urea 1, giving a further undesired bis-
urea 23 (Scheme 6).
with 16 proceeded with a negligible conversion. The first
trial with triphosgene gave an uncorrected yield of 78% molar
versus a 49% achieved with a solution of phosgene in toluene.
Even after screening a combination of solvents (acetonitrile,
dichloromethane, toluene, and ethyl acetate) and bases (Et3N,
iPr2NEt), the best result was as high as an uncorrected yield of
58%. Those promising results with triphosgene convinced us
to drop the less safe phosgene coupling agent.
3.4. Triphosgene Coupling Optimisation. Once triphos-
gene was selected as the coupling agent, all the subsequent
experiments aimed to maximise both yield and purity of the
final product.
The starting conditions of the process were: triphosgene (0.45
equiv) was dissolved in ethyl acetate (12 vol), and the solution
was cooled to 0 °C. A solution of N-Boc protected piperazine
19 (1 equiv) and triethylamine (1.5 equiv) were added followed
by a stirring period of 30 min at 0 °C. The resulting suspension
was heated to reflux and treated with a solution of benzylamine
16 (1.5 equiv as free base) and triethylamine (1.2 equiv) in ethyl
acetate (15 vol). The reaction mixture was kept at reflux for
6 h before being cooled to 25 °C. Among the several observed
impurities, bis-benzylamine derivative 27 resulted in being the
predominant one.
With the aim of reducing the total volumes of solvent,
reducing the number of equivalents of triphosgene, reducing
the amount of the expensive benzylamine 16, and producing a
lower formation of the main byproduct 27 (coming again from
Having understood the nature and the root cause of the
impurities, we devised the reaction conditions to minimize them
all. The bis-benzylamine derivative 27 was already kept under
Figure 2. Effect of equivalents of benzylamine 16 and triphosgene on conversion to desired product 20 and main impurity 27.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1103

Scheme 6. Process and impurities in the triphosgene process
Scheme 7. New potential alternative to triphosgene
control by applying the conditions suggested by the DoE study.
In addition, similarly to 27, not having basic nitrogens, 25 and
26 also could not precipitate during the crystallisation step.
By ensuring the completeness of the formation of the
carbamoyl chloride we prevented the formation of both 24 and
its derivative 29. On the other hand, the complete consumption
of the carbamoyl chloride buffered the formation of 23. The
impurity 28 resulted in being only a minor one, controlled Via
the crystallisation conditions.
3.5. Deprotection. Trifluoroacetic acid in tert-butylmethyl
ether was initially used to remove the Boc protection. However
tert-butylmethyl ether is a potential human carcinogen,12 and it
was nicely replaced with ethyl acetate, a solvent already in use
in the previous steps.
for the deprotection, leading to a biphasing system. A higher
yield of 70-75% was obtained when the intermediate salt was
first washed with base and then the resulting solution treated
with a stoichiometric amount of methanesulphonic acid.
3.6. Development of an Alternative to Triphosgene. As
the production scale increased, the safety issues concerning the
use of triphosgene13 prompted us to reinvestigate alternative
reagents, even if the preliminary work on other coupling agents
was not too promising. However, we investigated the little
known use of the carbon dioxide/thionyl chloride mixture14 for
the synthesis of the carbamoyl chloride. Between the two
partners of the coupling reaction, piperazine 19 and the
benzylamine 16, the latter gave a higher conversion for the
transformation described in Scheme 7.
The GW597599 initial selected salt was the acetate, but it
proved to be chemically unstable at temperatures above 60 °C;
thus, after a salt selection screening, the mesylate was chosen
as the preferred version due to an improved stability and better
physical properties profile such as a higher melting point, a
higher crystallinity, and a single isolated polymorph. A further
advantage was the capability of methanesulphonic acid to
remove the Boc group while forming the desired salt. However,
a direct salt formation from the reaction gave only a moderate
26% yield, probably due to the too large excess of acid required
The possible mechanism (Scheme 8) described in the
literature14 for the carbamoyl formation passes through an
unstable intermediate triethyl ammonium salt (31, Scheme 8),
(12) Toxicology and Industrial Health; CRC Press: Boca Raton, FL, 1995;
Vol. 2, No. 2, pp 119-149.
(13) Urben, P., Ed. Bretherick’s Handbook of ReactiVe Chemical Hazards,
7th ed.; Academic Press: Burlington, MA, 2008; Vol. 1, pp 429-
430.
(14) McGhee, W. D.; Pan, Y.; Talley, J. J. Tetrahedron Lett. 1994, 35,
839.
1104
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 8. Initial carbamoyl chloride alternative process
whereas the role of pyridine remained unclear, probably acting
as a trap for the sulphur dioxide formed during the reaction.
It should be said that the reaction suffered from variability
of yield without a clear rationale, maybe due to the instability
of the salt 31 involved in the equilibrium. The reaction was
improved by trapping the intermediate salt 31 in order to push
the equilibrium towards completion. The use of trimethylsilyl
chloride15 was found to be the right choice, leading to a higher
conversion of the starting material and a less variable profile.
The resulting hypothetical trimethylsilyl species 32 forced the
equilibrium between salt 31 and starting material 16 (Scheme
9), even when running the reaction at room temperature.
This three-phase reaction required intense studies to become
scalable; studies focused not only on the optimization of the
equivalents of reagents but also on engineering issues.
Figure 3. Carbon dioxide uptake versus inlet pressure.
Even on small scale the rate of carbon dioxide flow appeared
to be essential and required specific investigations. Reactions
were performed in a closed system, namely a hydrogenation
apparatus (Endeavour16), to check the amount of gas consumed
during the reaction and to evaluate the effect of an overpressure.
Additional tests using an infrared probe (Figure 3) aimed to
assess the amount of carbon dioxide absorbed in the solvent.
All these experiments highlighted the fact that the absorption
of carbon dioxide during the equilibrium between the free
secondary amine 16 and the carbamoyl acid derivative 31
(Scheme 9) was low, leading to a 16:30 ratio of only 19:1. A
relevant absorption of carbon dioxide occurred after the addition
of trimethylsilyl chloride which drove the equilibrium to the
carbamoyl silane intermediate (32). It was also demonstrated
that the carbon dioxide flow was needed all along the reaction,
even after the addition of the thionyl chloride, due to its
capability to remove other gases from the headspace. In fact
GC-MS spectra observed very low levels of sulphur dioxide
and trace quantities of hydrogen chloride in the reactor
headspace, supporting the hypothesis that the carbon dioxide
flow is necessary to remove sulphur dioxide, driving the
equilibrium toward completion. An additional proof is shown
in Figure 3; when the reaction was performed under overpres-
sure of carbon dioxide in the closed system (Endeavour), it was
faster even before the addition of TMSCl, possibly due to the
higher solubility of carbon dioxide in the solvent, but slower
in the second part due to the impossibility of the carbon dioxide
acting as scrubber for the sulphur dioxide (Figure 3, red line).
Vice versa the carbon dioxide flow (Figure 3, green line)
provided a slow reaction before the addition of TMSCl, but it
turned out to be much faster upon addition of TMSCl.
Figure 4. Nine bases to screen were selected using a PCA
viewer.
While keeping the supply of carbon dioxide constant, a first-
reagents screening was performed to simplify the procedure,
with the aim of replacing pyridine with a safer base. A principal
component analysis (PCA)17 was employed to ensure diversity
in reagents for the screening (Figure 4). The investigation
showed that each of the two bases (triethylamine and pyridine)
played a different role and that pyridine was indeed the most
suitable base for the second part of the reaction.
Further DoE investigations elucidated the impact of the
thionyl chloride concentration and the temperature on the overall
yield. The DoE study (D-optimal design, 20 experiments, seven
factors: from 1 to 2 equivalents of thionyl chloride, temperature
T1 from 0 to 25 °C before TMSCl addition, temperature T2 from
0 to 25 °C after TMSCl addition, from 5 to 10 volumes of
ethyl acetate, from 1.5 to 2.5 equivalents of triethylamine, from
1 to 2 equivalents of TMSCl, from 1.5 to 2.5 equivalents of
pyridine) proved to be quite complex, due to several constraints
required to ensure that regular chemistry took place: such as
constant carbon dioxide flow, the total quantity of bases greater
than the quantity of chloride reagents (triethylamine + pyridine
> TMSCl + SOCl2), and the two different temperatures to be
(17) Each class of compounds (for example solvents and reagents) can be
described by relevant chemical descriptors (e.g., dielectric constant,
boiling point, logP, etc.) that generate vectors in a multidimensional
space. The PCA is a statistical approach that allows reducing the
descriptors to three principal components; in this way, solvents that
are expected to behave similarly are close in space. For a more
statistical discussion: Eriksson, L., Johansson, E., Kettaneh-Wold, N.,
Wold, S. Multi- and MegaVariate Data Analysis, Principles and
Applications; Umetrics Academy: Umeå, Sweden, 2001; pp 43-48.
(15) Knausz, D.; Kolos, Z.; Meszticzky, A.; Csakvari, B. Kemiai Kozle-
menyek 1992, 74 (1-2), 147–159.
(16) Endeavor, Argonaut, is a system made of eight reactor vessels with
independent temperature and pressure controls. It works on a maximum
scale of ∼5 mL, and it permits feeding a reactant in the closed vials,
monitoring the pressure trend versus time.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1105

Scheme 9. Final alternative triphosgene process
Figure 5. Results from robustness assessment for the formation of the carbamoyl chloride (all equivalents are referred to benzylamine
starting material).
1106
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

Scheme 10. Stage 3, final manufacturing process
applied before and after the trimethylsilyl chloride addition (T1
< T2, Scheme 9 and Figure 5). The overall setup of experiments
was very challenging, taking into account a three-phase,
multistep reaction, a carbon dioxide flow rate to be kept
constant, the evaporation of the solvent, the formation of volatile
salts (triethylamine hydrochloride), the high exothermicity of
reagents addition, the corrosion problems due to traces of
hydrochloric acid gas evolution. In the end, it was possible to
extract some key information from the study (Figure 5). In
particular, the use of a large excess of thionyl chloride led to a
faster reaction (graph 1, Figure 5). It is important to note that
the amount of this reagent was not pushed to the suggested
limit, because a high excess was leading to a very dark reaction
mixture and to further problems in the workup procedure. Two
other parameters were found important and interacting, such
as the temperature after the trimethylsilyl chloride addition and
the amount of solvent (graph 2, Figure 5). Those two parameters
can be strongly related to the concentration of carbon dioxide
in the reaction media. The temperature was maintained and
carefully controlled at 25 °C, while volumes of solvent were
increased up to 10. The temperature before the trimethylsilyl
chloride addition was found to be less important, so that the
global procedure was simplified by using a single temperature
all the way through. The interactions between trimethylsilyl
chloride and the two bases, triethylamine and pyridine, were
also relevant (graphs 3 and 4, Figure 5).
or with triethylamine. Both of them were controlled Via the
crystallisation conditions as mentioned before.
Finally, the identification of the best monoprotection conditions
as described in section 3.2 minimised the presence of the impurities
18, 21, and 25.
4. Conclusions
In this contribution we describe the extensive process definition
made in the Chemical Development department to identify a safe,19
robust, and scalable process for the third and last stage of the
synthesis of GW597599, 1 (Scheme 10).
In the first large-scale manufacturing campaign of the three-
stage process3,4 (Scheme 11), the overall yield was 49% molar;
employed were two solvents, five reactors, and a silica treatment
to remove some impurities; for each kilogram of output about 190
kg of waste was generated. The final large-scale manufacturing
campaign, after all the improvements described in the current paper
and in the previous two publications,3,4 took advantage of the
convergent route and the new coupling reaction; on 100 kg of input,
using two solvents and only three reactors and no silica treatment,
a total of 58 kg of waste for each kilogram of output was produced
(Figure 6 and Table 2). The yield moved up to about 80-85%
molar, resulting in a final drug substance with an excellent purity
greater than 98% w/w.
The main points coming out of the DoE robustness assessment
are summarized in Figure 5.
In summary, based on the DoE analysis, the key critical steps
in this process are the initial saturation with carbon dioxide and
the reaction with trimethylsilyl chloride. It is of fundamental
importance to ensure the reaction is complete before the addition
of thionyl chloride.
Having in hand the best reaction conditions for the carbamoyl
chloride synthesis18 of the benzylamine 16, we managed to
avoid the formation of the impurities 23, 24, 28, and 29, all
based on the piperazine fragment as per Scheme 6. Of course,
we still had to deal with the main impurities, compounds 26
and 27, produced respectively by the reaction of the benzyl-
amine carbamoyl chloride with the unreacted benzylamine 16
Figure 6. Metrics of the evolution of the process.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1107

Scheme 11. Overall final manufacturing process
Table 2. Metrics of the evolution of the process
initial process to
support early
toxicological studies
1st large-scale
preparation to support
early clinical development
2nd large-scale
preparation to support
late clinical development
final route
to support
possible launch
overall yield
% mol
mass intensity
(kg materials/1 kg output)
relative cost of
1 to final route
2
6.4
360
2.4
12
45
109
1
2830
39
327
2.3
Experimental Section
(bt, J ) 12.0 Hz, 1H); 2.97 (dt, J ) 11.4, 2.0 Hz, 1H); 3.71
(dd, J ) 10.4, 2.4 Hz, 1); 3.86 (m, 2H); 6.97 (m, 2H); 7.54
(dd, J ) 8.3, 6.3 Hz, 1H).
Generic acidic HPLC method used: column type Luna C18;
mobile phase A: 0.05% TFA/water and B: 0.05% TFA/
acetonitrile; gradient: 0 min 100% A to 8 min 95% B; flow 1
mL/min; column temperature 40 °C; detector UV DAD @ 220
nm.
MS (ES+) m/z 295.4 (MH+); generic acidic HPLC, retention
time 3.7 min; purity ∼ 95% a/a.
Bis(1,1-dimethylethyl) (2S)-2-(4-fluoro-2-methylphenyl)-
1,4-piperazinedicarboxylate 21: 1H NMR (400 MHz, DMSO-
d6) δ 1.23 (s, 9H); 1.30 (bs, 9H); 2.32 (s, 3H); 3.27 (m, 1H);
3.5-3.7 (m, 4H); 3.89 (dt, J ) 13.7, 4.5 Hz, 1H); 5.03 (t, J )
5.3 Hz, 1H); 6.95 (td, J ) 8.7, 2.7 Hz, 1H); 7.01 (dd, J ) 10.1,
2.7 Hz, 1H); 7.19 (dd, J ) 8.2, 6.2 Hz, 1H).
1,1-Dimethylethyl (3S)-3-(4-fluoro-2-methylphenyl)-1-
piperazinecarboxylate 19. Piperazine dihydrochloride 18 (600
g, 2.2 mol) was suspended in ethyl acetate (8400 mL), then
triethylamine (732 mL, 5.25 mol) was added. The mixture was
stirred at 40 °C for 2 h and then cooled to 20 °C, and a solution
of di-tert-butyl dicarbonate (Boc2O, 480 g, 2.2 mol) in ethyl
acetate (2400 mL) was added over 1.5 h, maintaining a very
fast stirring rate. The suspension was stirred at 20 °C for 15
min. Water (1800 mL) was added, and the aqueous layer was
separated. After two further washes with water (2 × 1800 mL),
the organic phase was concentrated under vacuum to 1500 mL,
diluted with ethyl acetate (3600 mL), and concentrated again
to 1500 mL. The solution was finally cooled to 20 °C and used
as such in the next step.
MS (ES+) m/z 395.5 (MH+); generic acidic HPLC, retention
time 6.4 min; purity ∼ 90% a/a.
1,1-Dimethylethyl (3S)-4-{[{(1R)-1-[3,5-bis(trifluorometh-
yl)phenyl]ethyl}(methyl)amino]carbonyl}-3-(4-fluoro-2-meth-
ylphenyl)-1-piperazinecarboxylate 20 (Initial Method). Triph-
osgene (156 g, 0.52 mol) was dissolved in ethyl acetate
(4800 mL) under nitrogen. The solution was cooled to 0
°C. A solution of 19 (equivalent to 400 g of 19, 1.36 mol)
and triethylamine (285 mL, 2 mol) was added over 30
min. The mixture was stirred at 0 °C for 30 min, then the
suspension was heated to reflux, and a solution of 16 (free
base, 1.9 mol) and triethylamine (226 mL, 1.62 mol) in
ethyl acetate (5900 mL) was added over 15 min. The
reaction mixture was stirred at reflux for 6 h then cooled
1H NMR (400 MHz, DMSO-d6) δ 1.42 (s, 9H); 2.35 (s,
3H); 2.4-2.6 (bm, 2H); 2.69 (td, J ) 11.7, 2.5 Hz, 1H); 2.82
(18) Bientinesi, I.; Cimarosti, Z.; Guercio, G.; Leroi, C.; Perboni, A. Patent
WO 2007048642, 2007.
(19) See Supporting Information.
1108
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development

to 25 °C. An aqueous solution of sodium hydroxide 10%
(3200 mL) was added. The basic aqueous phase was
discharged, and HCl 4% (4000 mL) was added. After
discharging the acid aqueous layer, the organic mixture
was washed three times with NaCl 13% w/w (2000 mL).
The solution was evaporated to dryness to obtain the
desired product (920 g, purity about 85%) to be used as
such in the synthesis of compound 1.
N,N′-Bis{(1R)-1-[3,5-bis(trifluoromethyl)phenyl]ethyl}-
N,N′-dimethylurea 27: 1H NMR (400 MHz, DMSO-d6) δ 1.58
(d, J ) 7.0 Hz, 6H); 2.67 (s, 6H); 5.05 (q, J ) 7.0 Hz, 2H);
7.92 (bs, 4H); 7.97 (bs, 2H).
MS (ES+) m/z 569.4 (MH+); generic acidic HPLC, retention
time 7.7 min; purity > 90% a/a.
(2S)-N,N-Diethyl-2-(4-fluoro-2-methylphenyl)-1-pipera-
1
zinecarboxamide 28: H NMR (600 MHz, HPLC-NMR,
1H NMR (600 MHz, CDCl3) δ 1.48 (s, 9H); 1.54 (d, J )
7.1 Hz, 3H); 2.43 (s, 3H); 2.78 (s, 3H); 3.04-3.15 (m, 2H);
3.26-3.37 (m, 2H); 3.80-4.03 (m, 2H); 4.58 (bs, 1H); 5.59
(q, J ) 7.2 Hz, 1H); 6.81 (t, J ) 8.1 Hz, 1H); 6.88 (d, J ) 9.9
Hz, 1H); 7.20 (dd, J ) 8.1, 5.9 Hz, 1H); 7.54 (bs, 2H); 7.75
(bs, 1H).
MS (ES+) m/z 592.6 (MH+); generic acidic HPLC, retention
time 7.4 min; purity ∼ 90% a/a.
D2O/ACN) δ 0.92 (t, J ) 7.3 Hz, 6H); 2.31 (s, 3H); 2.99 (m,
2H); 3.30 (m, 5H); 6.88 (m, 2H); 7.29 (m, 1H).
MS (ES+) m/z 294.38 (MH+); HPLC Zorbax Eclipse XDB
C8; mobile phase A: water and B: acetonitrile; gradient: 0 min
55% A to 18 min 20% A. Flow 1 mL/min; detector UV DAD
@215 nm. Retention time 2.00 min. Purity ∼ 90% a/a.
(2S,2′S)-1,1′-(Oxomethanediyl)bis[2-(4-fluoro-2-methylphe-
nyl)piperazine] 29: 1H NMR (600 MHz, HPLC-NMR, D2O/
ACN) δ 2.20 (s, 6H); 3.08-3.13 (m, 2H); 3.22-3.27 (m, 2H);
3.31-3.37 (m, 2H); 3.38-3.44 (m, 2H); 3.64-3.70 (m, 2H);
3.80-3.86 (m, 2H); 4.73 (dd, J ) 11.2, 3.5 Hz, 2H); 6.83-6.88
(m, 2H); 6.90 (d, J ) 9.9 Hz, 2H); 7.06-7.14 (m, 2H).
MS (ES+) m/z 415.5 (MH+); generic acidic HPLC, retention
time 2.9 min; purity > 90% a/a.
(2S)-N-{(1R)-1-[3,5-Bis(trifluoromethyl)phenyl]ethyl}-2-
(4-fluoro-2-methylphenyl)-4-{[(2S)-2-(4-fluoro-2-methylphe-
nyl)-1-piperazinyl]carbonyl}-N-methyl-1-piperazinecarbox-
amide 23: 1H NMR (600 MHz, DMSO-d6) δ 1.51 (d, J ) 7.1
Hz, 3H); 2.33 (s, 6H); 2.52 (dd, J ) 12.9, 8.8 Hz, 1H); 2.59 (s,
3H); 2.67-2.76 (m, 2H); 2.76-2.87 (m, 2H); 2.94-3.03 (m,
1H); 3.06-3.12 (m, 1H); 3.29 (dd, J ) 13.2, 9.6 Hz, 1H);
3.33-3.40 (m, 2H); 3.53-3.63 (m, 2H); 4.27 (dd, J ) 8.8, 3.3
Hz, 1H); 4.49 (dd, J ) 8.9, 3.4 Hz, 1H); 5.20 (q, J ) 6.9 Hz,
1H); 6.66 (t, J ) 8.1 Hz, 1H); 6.86 (td, J ) 8.4, 2.6 Hz, 1H);
6.93-6.97 (m, 2H); 7.17 (dd, J ) 8.8, 6.0 Hz, 1H); 7.43 (dd,
J ) 8.8, 6.3 Hz, 1H); 7.71 (bs, 2H); 7.96 (bs, 1H).
MS (ES+) m/z 712.7 (MH+); generic acidic HPLC, retention
time 5.7 min; purity ∼ 90% a/a.
{(1R)-1-[3,5-Bis(trifluoromethyl)phenyl]ethyl}methylcar-
bamic Chloride 30. Benzylamine malate 16 (1140 g, 2.81 mol)
was suspended at 20 °C in ethyl acetate (3400 mL). Na2CO3
15% (3400 mL) was added, and the mixture was stirred until
dissolution. The phases were separated, and the organic layer
was washed with NaCl 20% w/w (3400 mL). The solution was
concentrated to 2100 mL, ethyl acetate (3000 mL) was added,
and the solution was concentrated again to 2100 mL. Further
ethyl acetate (3000 mL) was added. A cycle of vacuum and
CO2 was applied to the vessel, then a CO2 flow was maintained
for the whole process. Et3N (516 mL, 3.7 mol) was added, and
the reaction mixture was stirred at 20 °C for 30 min. Trimeth-
ylsilyl chloride (540 mL, 4.25 mol) was added over 30 min
(exothermic step), and the reaction mixture was stirred for a
further 30 min at 20 °C. Pyridine (462 mL, 5.7 mol) was added
followed by a 20 min addition of thionyl chloride (306 mL,
4.2 mol). The reaction mixture was stirred at 20 °C for 6 h
under a CO2 atmosphere. The reaction mixture was washed
twice with 28% aqueous racemic malic acid (2 × 680 mL).
The organic layer was washed with water (3000 mL) and then
with Na2CO3 15% w/w (3000 mL). Ethyl acetate (3600 mL)
was added, and the solution was concentrated to 1800 mL and
used as such in the next step.
Bis(1,1-dimethylethyl) (3S,3′S)-4,4′-(oxomethanediyl)bis[3-
(4-fluoro-2-methylphenyl)-1-piperazinecarboxylate] 24: 1H
NMR (400 MHz, DMSO-d6 at 70 °C) δ 1.47 (2s, 18H); 2.31
(s, 3H); 3.5-3.9 (bm, 6H); 5.03 (t, 1H); 6.9-7.0 (m, 2H); 7.18
(m, 1H).
MS (ES+) m/z 615.7 (MH+); generic acidic HPLC, retention
time 7.2 min; purity ∼ 90% a/a.
(2S)-N,N′-Bis{(1R)-1-[3,5-bis(trifluoromethyl)phenyl]ethyl}-
2-(4-fluoro-2-methylphenyl)-N,N′-dimethyl-1,4-piperazinedi-
carboxamide 25: 1H NMR (600 MHz, DMSO-d6) δ 1.50 (d,
J ) 6.9 Hz, 3H); 1.55 (d, J ) 7.1 Hz, 3H); 2.31 (s, 3H); 2.62
(s, 3H); 2.67 (s, 3H); 2.93 (dd, J ) 13.2, 10.7 Hz, 1H);
3.03-3.09 (m, 1H); 3.25-3.31 (m, 1H); 3.42 (dd, J ) 13.0,
3.2 Hz, 1H); 3.44-3.48 (m, 1H); 3.53-3.59 (m, 1H); 4.57 (dd,
J ) 10.4, 3.8 Hz, 1H); 5.18 (q, J ) 7.1 Hz, 1H); 5.25 (q, J )
6.9 Hz, 1H); 6.81 (dt, J ) 8.4, 2.5 Hz, 1H); 6.95 (dd, J ) 10.0,
2.6 Hz, 1H); 7.27 (dd, J ) 8.5, 6.0 Hz, 1H); 7.72 (bs, 2H);
7.97 (bs, 2H); 7.98 (bs, 1H); 8.00 (bs, 1H).
1H NMR (600 MHz, CDCl3) δ 1.67-1.74 (m, 3H), 2.89
(bs, 3H), 5.73-5.83 (m, 1H), 7.75 (bs, 2H), 7.87 (bs, 1H).
MS (ES+) m/z 334.66 (MH+); HPLC column type X-Terra
RP18; mobile phase A ammonium hydrogen carbonate 5 mM
pH ) 10/acetonitrile 90/10% v/v and B ammonium hydrogen
carbonate 5 mM pH ) 10/acetonitrile 10/90% v/v; gradient: 0
min 58% B to 6 min 63% B then to 7 min 100% B. flow 1
mL/min; column temperature 40 °C; detector UV DAD @210
nm. Retention time 6.2 min. Purity > 90% a/a.
1,1-Dimethylethyl (3S)-4-{[{(1R)-1-[3,5-bis(trifluorometh-
yl)phenyl]ethyl}(methyl)amino]carbonyl}-3-(4-fluoro-2-meth-
ylphenyl)-1-piperazinecarboxylate 20 (Final Method). A
solution of 30 was added at 25 °C to a mixture of solution 19
MS (ES+) m/z 789.7 (MH+); generic acidic HPLC, retention
time 7.9 min; purity > 99% a/a.
N-{(1R)-1-[3,5-Bis(trifluoromethyl)phenyl]ethyl}-N′,N′-di-
1
ethyl-N-methylurea 26: H NMR (600 MHz, DMSO-d6) δ
1.03 (t, J ) 7.1 Hz, 6H); 1.54 (d, J ) 6.9 Hz, 3H); 2.54 (s,
3H); 3.11 (q, J ) 7.0 Hz, 4H); 5.10 (q, J ) 6.8 Hz, 1H); 7.92
(bs, 2H); 7.94 (bs, 1H).
MS (ES+) m/z 371.34 (MH+); generic acidic HPLC, reten-
tion time 6.5 min; purity ∼ 90% a/a.
Vol. 13, No. 6, 2009 / Organic Process Research & Development
•
1109

and Et3N (408 mL, 2.93 mol). The mixture was heated at reflux
for at least 19 h then cooled to 25 °C. Diethylamine (66 mL,
0.64 mol) was added, and the solution was stirred for 1 h to
quench the unreacted 30. Aqueous NaCl 20% w/w (1200 mL)
was added, the phases were separated, and the solution was
used as such in the next step.
Hz, 1H); 7.29 (dd, J ) 8.5, 6.0 Hz, 1H); 7.71 (bs, 2H); 8.02
(bs, 1H); 8.71 (bs, 1H); 9.02 (bs, 1H).
ES+: m/z 492 [MH - CH3SO3H]+, 341, 221; ES-: m/z 586
[M - H]-; 95 [CH3SO3]-.
13C NMR (150 MHz, DMSO-d6) δ 16.37, 18.81, 30.54,
39.79, 42.41, 45.70, 46.58, 52.41, 53.42, 112.48, 116.55, 121.02,
123.19 (d), 127.19, 127.44 (d), 130.34 (d), 134.00, 138.56,
144.79, 160.89, 163.2.
1H NMR (600 MHz, CDCl3) δ 1.48 (s, 9H), 1.54 (d, J )
7.1 Hz, 3H), 2.43 (s, 3H), 2.78 (s, 3H), 3.04-3.15 (m, 2H),
3.26-3.37 (m, 2H), 3.80-4.03 (m, 2H), 4.58 (bs, 1H), 5.59
(q, J ) 7.2 Hz, 1H), 6.81 (t, J ) 8.1 Hz, 1H), 6.88 (d, J )
9.9 Hz, 1H), 7.20 (dd, J ) 8.1, 5.9 Hz, 1H), 7.54 (bs, 2H),
7.75 (bs, 1H).
IR (Nujol mull, cm-1): 1653 (str. CdO), 1600 (str. CdC
aromatic) (cm-1).
HPLC column type Betabasic C18; mobile phase A: buffer
ammonium hydrogen carbonate 5 mM pH ) 10/methanol 40/
60% v/v and B buffer ammonium hydrogen carbonate 5 mM
pH ) 10/methanol 10/90% v/v; gradient: 0 min 100% A to 20
min 100% B. flow 1 mL/min; column temperature 40 °C;
detector UV DAD @210 nm. Retention times 1: 13 min, purity
>98%.
HPLC column type Chiralpack AD; mobile phase n-hexane/
ethanol 86/14% v/v + 0.2% v/v purified water; flow 1 mL/
min; column temperature 25 °C; detector UV DAD @210 nm.
Retention time 1: 4.56 min and opposite enantiomer 4.15 min,
other diastereomers 5.20 and 14.2 min, respectively.
MS (ES+) m/z 592.6 (MH+); generic acidic HPLC, retention
time 7.4 min; purity ∼ 90% a/a.
(2S)-N-{(1R)-1-[3,5-Bis(trifluoromethyl)phenyl]ethyl}-2-
(4-fluoro-2-methylphenyl)-N-methyl-1-piperazinecarboxam-
ide Methanesulphonate 1. To the organic phase of 20,
MeSO3H (720 mL, 11.1 mol) was added over 10 min, and the
mixture was heated at 40 °C and stirred for 1 h. Ethyl acetate
(3000 mL) was added followed by NH3 (13%, 2400 mL). The
phases were separated, and the organic layer was washed with
water (2 × 2400 mL). The reaction mixture was concentrated
to 1800 mL, ethyl acetate (6000 mL) was added, and the
mixture was concentrated finally to 3600 mL. Methanesulphonic
acid (144 mL, 2.2 mol) was added dropwise at 25 °C and the
solution seeded with 1 (3 g). Isooctane (7200 mL) was added
over 90 min. The suspension was stirred for at least 6-8 h and
filtered, and the solid was washed three times with an ethyl
acetate/isooctane 1:2 mixture (3 × 1800 mL). The resulting
solid was dried under vacuum at 40 °C for at least 3 h, obtaining
1100 g of 1 (98% w/w purity, theoretical yield ) 85% from
18).
Acknowledgment
We thank Lucilla Turco for the spectroscopic characterisation
of the compounds and for valuable discussions. We also thank
Giulio Camurri and Luca Martini for the HPLC method
development, Francois Ricard for computational modelling
studies and Gilles Laval for his support defining the process.
Supporting Information Available
Process safety evaluation of the process and chiral HPLC
of 1. This information is available free of charge via the Internet
at http://pubs.acs.org.
1H NMR (600 MHz, DMSO-d6) δ 1.48 (d, J ) 6.9 Hz, 3H);
2.31 (s, 3H); 2.39 (s, 3H); 2.74 (s, 3H); 2.95 (t, J ) 12.2 Hz,
1H); 3.00-3.06 (m, 1H); 3.27 (dd, J ) 12.5, 2.3 Hz, 1H);
3.28-3.35 (m, 1H); 3.38-3.43 (m, 1H); 3.47 (dt, J ) 13.3,
3.1 Hz, 1H); 4.49 (dd, J ) 11.8, 3.3 Hz, 1H); 5.35 (q, J ) 6.5
Hz, 1H); 6.84 (td, J ) 8.4, 2.6 Hz, 1H); 7.01 (dd, J ) 10.2, 2.7
Received for review August 3, 2009.
OP9002032
1110
•
Vol. 13, No. 6, 2009 / Organic Process Research & Development