Debottlenecking the Synthesis Route of Asenapine

Article abstract of DOI:10.1021/op700240c

The discovery synthesis of asenapine that was used for the manufacture of drug substance batches up to 10 kg contained two chemical steps that were major bottlenecks for scale-up. One of these steps involved a magnesium/methanol reduction of an enamide moiety that was severely hampered by safety and efficiency problems. The other step was a laborious chromatography and isomerization cycle that was marked by a poor yield and extremely low throughput The safety issues of the magnesium/methanol reduction could be solved by adding portions of magnesium to a solution of the enamide. In addition, an alternative process for the conversion of the mixture of cis- and trans-lactam into the desired trans-isomer was developed, circumventing the chromatographic separation.

Full text of DOI:10.1021/op700240c

Organic Process Research & Development 2008, 12, 196–201
Debottlenecking the Synthesis Route of Asenapine
Marco van der Linden, Theo Roeters, Ramon Harting, Edwin Stokkingreef, Arjan Sollewijn Gelpke, and Gerjan Kemperman*
Department of Process Chemistry, Organon NV, P.O. Box 20, 5340 NH Oss, The Netherlands
Abstract:
synthesis was scaled up to about 10 kg. Scaling up this process
to commercial production was deemed impossible for at least
part of the synthesis route (see the box in Scheme 1). The
synthesis of unsaturated lactam 4 from 5-chloro-2-phenoxy-
phenylacetic acid 1 was considered to be well scalable, although
optimization of the yields and reaction volumes was needed.
The conversion of trans-lactam 5 into asenapine maleate
proceeded well, but the need to decrease the reaction volumes
and to improve the yield of the final recrystallization was
recognized. However, the true bottleneck for scale-up of this
synthesis was found to be the conversion of unsaturated lactam
4 into trans-lactam 5. This report discusses the development
and scale-up of alternative processes for this conversion.
The discovery synthesis of asenapine that was used for the
manufacture of drug substance batches up to 10 kg contained two
chemical steps that were major bottlenecks for scale-up. One of
these steps involved a magnesium/methanol reduction of an
enamide moiety that was severely hampered by safety and
efficiency problems. The other step was a laborious chromatog-
raphy and isomerization cycle that was marked by a poor yield
and extremely low throughput. The safety issues of the magnesium/
methanol reduction could be solved by adding portions of
magnesium to a solution of the enamide. In addition, an alternative
process for the conversion of the mixture of cis- and trans-lactam
into the desired trans-isomer was developed, circumventing the
chromatographic separation.
2. Results and Discussion
Identification of Bottlenecks in the Conversion of 4 into
5. The conversion of unsaturated lactam 4 into trans-lactam 5
comprises two steps. The first step involves magnesium/
methanol reduction of the double bond, which gives rise to the
formation of the desired trans-lactam 5 and its cis-isomer 6 in
an unfavorable ratio of approximately 1:4. In addition, a
significant amount of side products are formed because the
combined amount of 5 and 6 is only 60-80% of the reaction
product.
1. Introduction
Asenapine is a novel psychopharmacologic agent being
developed for the treatment of schizophrenia and bipolar
disorder. The pharmacologic profile, kinetics, metabolism, and
safety and efficacy studies in human volunteers and schizo-
phrenic patients of asenapine have been comprehensively
reviewed.1 Asenapine has a unique in vitro receptor affinity
profile, with the highest affinity in its class for an array of
serotonin (5-HT2A, 5-HT2B, 5-HT2C, 5-HT5A, 5-HT6, 5-HT7),
dopamine (D1, D3, D4), R-adrenergic receptors (R1A, R2A, R2B,
R2C), and histamine (H1, H2) receptors but with minimal affinity
for muscarinic receptors.2 In a recently published phase III
clinical trial, asenapine was efficacious and well-tolerated in
the treatment of patients with acute exacerbation of schizo-
phrenia.1
The original synthesis of asenapine has been published.3,4
This synthesis, which is outlined in Scheme 1, starts from
5-chloro-2-phenoxyphenylacetic acid 1, which can be prepared
via several different routes, such as the route described by Harris
et al.5 The synthesis of asenapine as shown in Scheme 1 has
been used for the supply of active pharmaceutical ingredient
(API) required for preclinical and clinical development. To
prepare adequate amounts of API, the original laboratory
The process involves addition of unsaturated lactam 4 in
toluene solution to a suspension of magnesium in a mixture
of toluene and methanol. The magnesium was activated by using
the carcinogenic dibromoethane before addition of the unsatur-
ated lactam. During this process, large amounts of hydrogen
gas are formed because of the inevitable exothermic side
reaction of magnesium with methanol. An additional drawback
of such a process is that all of the potential energy of the
magnesium, in contact with methanol, is present in the reactor,
resulting in a large accumulation of heat. Because there is no
control over the rate in which the heterogeneous reaction
between magnesium and methanol takes place, the maximum
scale at which one can safely operate such a process is limited
and is determined by the cooling and venting capacity of the
reactor. Figure 1 shows the result of a reaction calorimetry
experiment of this reaction. Clearly, the reaction is not dose-
controlled. Because of the potential danger of the accumulated
heat, the maximum scale was estimated to be 10-15 kg. In
addition, the reaction of starting material 4, which is dissolved
in a mixture of methanol and toluene, with the reducing mixture
of magnesium and methanol is highly exothermic. This restricts
the rate of addition of the solution of starting material 4 to the
magnesium suspension as well. The longer the addition time,
the more magnesium is lost in reaction with methanol, resulting
in the formation of hydrogen gas. As a consequence, an excess
of magnesium (i.e., 11 equiv) is required to achieve complete
* To whom correspondence should be addressed. Ph: +31-(0)-412669375.
Fax: +31-(0)-412663513. E-mail: gerjan.kemperman@organon.com.
(1) Potkin, S. G.; Cohen, M.; Panagides, J. J. Clin. Psychiatry. 2007, 68,
1492–1500.
(2) Shahid, M.; Walker, G. B.; Zorn, S. H.; Wong, E. H. F. J. Psychop-
harmacol. 2007, in press.
(3) Vader, J.; Kaspersen, F.; Sperling, E.; Schlachter, I.; Terpstra, A.;
Hilberink, P.; Wagenaars, G. J. Labelled Compd. Radiopharm. 1994,
34, 845–869.
(4) van der Burg, W. J. Tetracyclic derivatives and pharmaceutical
compositions of matter. U.S. patent 4,145,434, 1979.
(5) Harris, T. W.; Smith, H. E.; Mobley, P. L.; Manier, D. H.; Sulser, F.
J. Med. Chem. 1982, 25, 855–858.
196
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development
10.1021/op700240c CCC: $40.75
2008 American Chemical Society
Published on Web 02/12/2008

Scheme 1. Synthesis of asenapine
conversion of starting material 4. Finally, the poor control causes
an extremely vigorous reaction, leading to a significant amount
of side products.
The second issue is the conversion of a 1:4 mixture of the
desired trans-lactam 5 and the unwanted cis-lactam 6 into pure
trans-lactam 5. This step involves an isomerization and chro-
matography loop. First, partial isomerization of the unwanted
cis-isomer 6 into the trans-isomer 5 by using DBN leads to a
trans:cis ratio of 1:3, which is the thermodynamic equilibrium.
The trans-lactam 5 and cis-lactam 6 are then separated by
chromatography over silica gel. This provides one fraction with
the desired trans-lactam 5 and a fraction of the unwanted isomer
cis-lactam 6. The cis-isomer 6 is isomerized again using DBN,
after which the chromatographic separation is repeated. After
this cycle has been repeated three times, the three fractions of
trans-lactam 5 are collected and crystallized, furnishing one
batch of trans-lactam 5. The overall yield of this procedure is
32%. Clearly the isomerization/chromatography loop is ex-
tremely elaborate and low yielding, which results in high cost
and low throughput.
In summary, the magnesium/methanol reduction of lactam
4 is hampered by safety problems and by poor product
selectivity. The conversion of the resulting mixture of 5 and 6
into the desired trans-lactam 5 suffers from low throughput and
low yield and accordingly limited capacity and high cost.
Development of an Alternative Process for the Magne-
sium/Methanol Reduction. Initial attempts to develop an
alternative process for magnesium/methanol reduction involved
catalytic hydrogenation. A variety of palladium, platinum,
rhodium, ruthenium, and iridium catalysts were tested using
different ligands and solvents at pressures varying between 1
and 5 bar. Under most conditions no reaction was observed.
With several palladium catalysts, only dechlorination took place.
It was also investigated whether zinc or lithium could be
used to reduce unsaturated lactam 4. In addition, the use of
Figure 1. Output of a reaction calorimetry experiment on the
Mg/MeOH reduction. The solid line presents the heat flow, and
the dotted line depicts the addition of unsaturated lactam 4.
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
197

reaction with methanol to form hydrogen gas. Consequently,
much less magnesium was required for a complete conversion
of the starting material (i.e., only 3 equiv). This was beneficial
to the safety of the process because only 2 mol instead of 10
mol of hydrogen gas was being formed over a controlled period
of time. Another advantage was found in a better control of
the temperature, resulting in a less vigorous reaction and
formation of significantly fewer side products. The reversed
addition process led to a product containing more than 93% of
trans-lactam 5 and cis-lactam 6 together, compared with
60-80% with the original process.
Development of an Alternative Process for the Isomer-
ization and Chromatography Loop. A complicating factor
in the preparation of trans-lactam from the cis/trans mixture
was that the thermodynamic equilibrium between the cis- and
trans-lactam was in favor of the unwanted cis compound 6. It
was found that the cis-isomer could be selectively crystallized
from toluene, leading to a mother liquor that consisted of 5
and 6 in a more favorable 1:1 ratio. Nevertheless, the batchwise
chromatography process was not amenable to scale-up because
of the low throughput. Because the aim of the chromatography
step was the separation of two components, application of
simulating moving bed chromatography (SMB) was a potential
alternative. Separation of cis- and trans-lactam by SMB
appeared to be potentially scalable against acceptable cost.
However, prepurification would be required; with SMB, it is
important to start with a relatively pure mixture of two
components. Therefore, alternative solutions were sought.
A solubility study of cis- and trans-lactam in various solvents
pointed out that the cis-lactam was 10 times less soluble in
toluene, whereas the trans-lactam was 2 times less soluble in
methanol. This finding suggested that a preferred crystallization
process might be possible. In the laboratory, preferred crystal-
lization worked reasonably well with pure mixtures of cis- and
trans-lactam, but the yield of the crystallizations was too low
when the reaction mixtures contained other impurities.
For both chromatography and preferred crystallization, the
yield was affected heavily by the unfavorable cis/trans ratio at
thermodynamic equilibrium and by the need to perform a
prepurification on the crude cis/trans mixture. Therefore, we
attempted to convert the cis- and trans-lactam into derivatives
that have a more favorable trans/cis ratio at thermodynamic
equilibrium. Conceptually, the idea was to open the five-
membered lactam ring of compounds 5 and 6, whereby the
corresponding amino acid derivatives 9 and 10 would be formed
as shown in Scheme 2. It is plausible that the trans-amino acid
9 was more thermodynamically stable than the cis-isomer 10,
such that after equilibration the preferred trans-isomer 9 should
predominate. When the ring closure of trans-amino acid 9 is
accomplished, trans-lactam 5 was selectively provided.
It was found that ring-opening of trans-lactam 5 and
cis-lactam 6 could be achieved by treating a reaction
mixture with potassium hydroxide in refluxing ethanol
(Scheme 3). Under these conditions, the boiling temper-
ature of the solvent exceeded 100 °C, resulting in a fast
reaction. The ring-opening reaction was completed in
about 4 h. The reaction mixture comprised trans-amino
acid 9 and cis-amino acid 10 in a ratio of approximately
Figure 2. Output of a reaction calorimetry experiment on the
reversed addition magnesium/methanol reduction. The solid line
represents heat flow; the dotted line shows the addition of the
magnesium portions.
magnesium in combination with less acidic alcohols was
explored with the intention of suppressing hydrogen gas
formation. It appeared that the reducing power of zinc was
insufficient; no reaction was observed. However, a Birch
reduction, using lithium in ammonia, was found to be too strong
because a complex mixture of products was obtained using this
procedure. Lactam 4 also could not be reduced with magnesium
in ethanol or propanol.
Since it had been found that only magnesium in combination
with methanol was able to reduce the double bond of lactam 4,
it was investigated how this reaction could be transformed into
a safer and more efficient procedure. To this end, the reaction
was investigated by dosing the magnesium in portions to a
solution of unsaturated lactam 4 instead of dosing the lactam
to the magnesium. In principle, portionwise addition of mag-
nesium should allow for better heat control because the amount
of energy introduced in the reactor is directly related to the
portion size. It would be challenging to see whether the reaction
would proceed with portionwise addition, given that the
activation of a successive portion of magnesium could not be
controlled.
The reversed addition process evolved into a procedure in
which part of the magnesium turnings (i.e., approximately 0.6
equiv) was first suspended in toluene. The magnesium was
activated with iodine instead of dibromoethane, which was
followed by addition of a solution of the unsaturated lactam 4
in a toluene/methanol mixture. Subsequently, the required
amount of magnesium was added in several portions. In this
way, the reaction rate and the reaction temperature were
perfectly controlled because the maximum amount of heat that
can be generated was related to the size of the portions of
magnesium. This should overcome the problem of accumulation
encountered in the original procedure. Figure 2 illustrates a
reaction calorimetry experiment of the reversed addition
magnesium/methanol reduction of lactam 4. It is evident from
Figure 2 that the process was dose controlled and thus, in
principle, scalable.
Another advantage of the reversed addition process was that
the magnesium is only reacting with methanol in the presence
of starting material. As a result, little magnesium was lost in a
198
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

Scheme 2. Concept of ring-opening and -closing toward trans-lactam 5
Scheme 3. Developed processes for the conversion of unsaturated lactam 4 into trans-lactam 5
10:1, suggesting an isomerization concomitant to the ring-
opening process. By using deuterium labeling, it was
demonstrated that the isomerization of the amino acid
isomers proceeds via deprotonation of the R-hydrogen. It
was found that the trans-amino acid could be crystallized
as a zwitterion, but the hydrochloride salt could be
crystallized in a higher yield. After crystallization, the
trans-amino acid contained less than 1% of the cis-isomer.
The reaction could also be conducted in n-butanol or
n-propanol. The higher boiling point of n-butanol led to
a shorter reaction time for the ring opening, but the workup
was found to be more difficult as a result of the polarity
of n-butanol and its partial immiscibility with water. The
relatively high solubility of the amino acid in n-butanol,
even at high or low pH, caused loss of product during the
acid/base extractions.
Next, it was investigated how to effect ring closure of
the trans-amino acid to obtain the desired trans-lactam
5. It was found that ring closure could be achieved in
several ways. First, the hydrochloride salt of the trans-
amino acid (compound 9a in Scheme 3) could be
converted into the acid chloride derivative by treatment
with thionyl chloride. When the acid chloride was reacted
with a base such as triethylamine, the trans-lactam was
obtained in a yield of 75%. A second method involves
heating a suspension of the hydrochloride salt 9a in
toluene in the presence of silica gel for 24 h. In this way
the trans-lactam could be obtained in yields varying from
80% to 86%. The reaction could also be accomplished
without the use of silica gel albeit at a higher temperature.
Thus, heating 9a in xylene for 24 h led to trans-lactam in
similar yields. A third option was to use the zwitterion 9,
which upon heating in toluene, showed a fast and clean
ring-closing reaction, providing the trans-lactam in a yield
of 90%. Although the third option is preferable for scale-
up, because of the neutral and relatively mild conditions,
the isolation of the zwitterion 9 led overall to lower yields.
We therefore attempted to find a basic additive that could
be used to generate the zwitterion 9 in situ from the amino
acid hydrochloride salt 9a. This would have the advantage
of carrying out the reaction in a stainless-steel reactor
instead of a glass-lined reactor, avoiding the risk of
electrical discharge. A series of salts was investigated (e.g.,
sodium hydrogen carbonate, sodium carbonate, sodium
hydroxide, ammonium hydroxide, and sodium acetate),
and all worked nearly equally well. Sodium acetate was
selected for scale-up because it was found to result in the
fastest reaction. Also NaOH and NH4OH have been
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
199

Table 1. Batch analysis data of trans-lactam 5
a/a % by HPLC
RRT 0.71
batch
5
6
RRT 0.67
RRT 0.73
RRT 0.80
others (total)
Discovery Route to trans-Lactam 5
Lot 1
Lot 2
Lot 3
Lot 4
Lot 5
Lot 6
95.67
97.86
98.51
98.86
97.59
98.64
<0.05
<0.05
0.13
0.24
0.16
0.42
0.21
0.63
0.37
1.65
0.66
0.20
0.09
0.67
0.24
<0.05
<0.05
0.05
1.99
0.92
0.49
0.20
0.80
0.38
<0.05
0.27
0.09
0.15
0.13
0.11
0.15
0.22
0.10
0.12
<0.05
0.06
Developed Route to trans-Lactam 5
Lot 7
99.82
99.80
99.90
99.87
99.86
99.88
0.08
0.11
0.06
0.09
0.08
0.09
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
0.10
0.09
<0.05
<0.05
0.06
Lot 8
Lot 9
Lot 10
Lot 11
Lot 12
<0.05
1
successfully scaled up in the plant and were used to
produce 20-kg batches of trans-lactam 5.
in parts per million (ppm). H NMR chemical shifts were
referenced to TMS as internal standard. Mass spectra were recorded
on a PE SCIEX API 165. The synthesis of 11-chloro-2,3-dihydro-
2-methyl-1H-dibenz[2,3;6,7]oxepino[4,5-c]pyrrol-1-one (4) was
performed according to the findings of Vader et al.3
The new processes for the conversion of unsaturated lactam
4 into trans-lactam 5 have been outlined in Scheme 3.6 The
processes are highly telescoped. The reaction mixture from the
magnesium reduction is directly used in the ring-opening
process. After filtration of the hydrochloride salt 9a, the solvent-
wet crystals are immediately transferred back into the reactor
for the ring-closing reaction. The overall yield of the conversion
is 65%, which is significantly higher than the 32% obtained
via the original procedure. In addition, the processes were found
to be well scalable and are currently operated at batch sizes of
120 and 250 kg, respectively.
Table 1 shows comparisons of the impurity profile of a
number of batches of trans-lactam 5 from the original chro-
matography process with that of batches from the developed
processes. It appears that the quality of the trans-lactam made
via the new processes is significantly better than that of the
original processes. Although the trans-lactam batches made via
the original route had a variable impurity profile, with some
impurities being present at levels up to 2%, the trans-lactam
from the developed process has a consistent impurity profile
characterized by a purity >99.8%. This high level of control
of the purity of trans-lactam 5 minimized the risk for introduc-
tion of new impurities into asenapine via the new manufacturing
processes.
Reduction of 11-Chloro-2,3-dihydro-2-methyl-1H-dibenz-
[2,3;6,7]oxepino[4,5-c]pyrrol-1-one (4) via the Reversed
Addition Method. Under a nitrogen atmosphere magnesium
turnings (2.2 kg, 90.6 mol) were added to a solution of iodine
(8.5 kg, 22.0 mol) in toluene (100 L). This was followed by
addition of a solution of 11-chloro-2,3-dihydro-2-methyl-1H-
dibenz[2,3;6,7]oxepino[4,5-c] pyrrol-1-one (4) (45 kg, 151.2
mol) in a mixture of methanol (225 L) and toluene (225 L)
over a period of approximately 30 min at a temperature below
40 °C. Subsequently, over a period of 5 h, 7 portions of
magnesium (1.3 kg, 53.5 mol) were added. The reaction mixture
was diluted with toluene (900 L) and neutralized with a mixture
of 18% hydrochloric acid (220 L) and water (540 L). The layers
were separated, and the water layer was extracted twice with
toluene (600 L). The combined toluene layers were washed with
water (450 L) and evaporated to dryness. This gave a mixture
of trans-11-chloro-2,3,3a,12b-tetrahydro-2-methyl-1H-dibenz[2,3:
6,7]oxepino[4,5-c]pyrrol-1-one (5) and cis-11-chloro-2,3,3a,12b-
tetrahydro-2-methyl-1H-dibenz[2,3:6,7]oxepino[4,5-c]pyrrol-1-
one (6) (45.3 kg, ∼100%) in a ratio of 1:4 as determined by
1H NMR and HPLC.
In conclusion, safe and scalable processes have been
developed for the manufacture of trans-lactam 5, which is the
final intermediate in the synthesis of asenapine maleate. The
developed processes led to an improvement of the yield for
the conversion of 4 into 5 from 32% to 65%. In addition, the
developed processes have resulted in a significant improvement
of the impurity profile of compound 5.
Preparation of trans-11-Chloro-2,3,3a,12b-tetrahydro-2-
methyl-1H-dibenz[2,3:6,7]oxepino[4,5-c]pyrrol-1-one (5). Po-
tassium hydroxide was added (144 kg, 2571 mol) to a mixture
of cis-11-chloro-2,3,3a,12b-tetrahydro-2-methyl-1H-dibenz[2,3:
6,7]oxepino[4,5-c]pyrrol-1-one (6) and trans-11-chloro-2,3,3a,12b-
tetrahydro-2-methyl-1H-dibenz[2,3:6,7]oxepino[4,5-c]pyrrol-1-
one (5) (45.3 kg,150.5 mol cis:trans ratio 4:1) dissolved in
ethanol (675 L). The reaction mixture was concentrated by
distillation of part of the ethanol (300 L) until a temperature of
106 °C was reached. The mixture was maintained at reflux
temperature for 4 h. The reaction mixture was cooled and diluted
with ethanol (180 L), water (450 L), and toluene (225 L). The
pH of the solution was adjusted to 1 by using 18% hydrochloric
acid (∼510 L) at a temperature below 55 °C. The toluene layer
was separated. The aqueous phase was partly evaporated to
Experimental Section
General Information. Nuclear magnetic resonance spectra
were recorded on a Bruker DPX 400. Chemical shifts were reported
(6) Kemperman, G. J.; van der Linden, J. J. M.; Reeder, M. Intermediate
compounds for the preparation of trans-5-chloro-2-methyl-2,3,3a,12b-
tetrahydro-1H-dibenz[2,3;6,7]oxepino[4,5-c]pyrrole. Patent WO2006/
106136, 2006.
200
•
Vol. 12, No. 2, 2008 / Organic Process Research & Development

remove the ethanol and subsequently cooled to ambient tem-
perature to start crystallization of the product. The crystals of
trans-8-chloro-10,11-dihydro-11-[(methylamino)methyl]-dibenz-
[b,f]oxepin-10-carboxylic acid hydrochloride (9a) were filtered
and washed once with water (25 L). The wet crystals were
directly transferred back in the reactor containing a suspension
of sodium acetate (3.2 kg, 39 mol) in toluene (350 L). The
reaction mixture was heated and distilled until a temperature
of 92 °C was reached while maintaining the volume by adding
toluene. After the reaction mixture was cooled to 70 °C, a
suspension of sodium acetate (9.5 kg, 116 mol) in toluene (50
L) was added. The reaction mixture was heated to reflux for
3 h and then cooled to 70 °C, followed by filtration to remove
salts. Evaporation of the toluene followed by crystallization of
the product from methanaol furnished 28.8 kg (64% overall)
of trans-11-chloro-2,3,3a,12b-tetrahydro-2-methyl-1H-dibenz[2,3:
6,7]oxepino[4,5-c]pyrrol-1-one (5): mp 150.6 °C; 1H NMR data
(measured at 399.87 MHz in CDCl3 relative to TMS) δ (ppm)
3.03 (s, 3H, N-CH3), 3.53–3.65 (m, 3H, CHꢀ + CH), 3.82 (m,
1H, CH), 4.03 (m, 1H, CHR), 7.01–7.28 (m, 6H, ArH), 7.87
(dd, 1H, ArH).
Acknowledgment
Drs. Leo Letendre, Michael Reeder, and Tim Stuk (Pfizer
Inc.) are acknowledged for the pleasant collaboration and for
their contribution to the process development activities for
asenapine.
Supporting Information Available
Additional synthetic methods. This information is available
free of charge via the Internet at http://pubs.acs.org.
Received for review October 26, 2007.
OP700240C
Vol. 12, No. 2, 2008 / Organic Process Research & Development
•
201