Continuous flow hydrogenation of a pharmaceutical intermediate, [4-(3,4-dichlorophenyl)-3,4-dihydro-2H-naphthalenyidene]-methylamine, in supercritical carbon dioxide

Article abstract of DOI:10.1002/adsc.200700395

The hydrogenation of complex pharmaceutical intermediate, rac-sertraline imine has been optimized as a continuous flow process utilizing a palladium/calcium carbonate (Pd/CaCO₃) catalyst and hydrogen in supercritical carbon dioxide (scCO_2<

Full text of DOI:10.1002/adsc.200700395

COMMUNICATIONS
DOI: 10.1002/adsc.200700395
Continuous Flow Hydrogenation ofa Pharmaceutical Intermedi-
ate, [4-(3,4-Dichlorophenyl)-3,4-dihydro-2H-naphthalenyidene]-
methylamine, in Supercritical Carbon Dioxide
Peter Clark,a Martyn Poliakoff,a,* and Andy Wellsb,*
a
Chemistry Department, University of Nottingham, University Park, Nottingham, NG7 2RD, U.K.
Fax : (+44)-115-951-3058 E-mail: martyn.poliakoff@nottingham.ac.uk
AstraZeneca, Process R&D, Charnwood, Bakewell Road, Loughborough, LE11 5RH, U.K.
b
E-mail: andrew.wells@astrazeneca.com
Received: June 26, 2007; Revised: August 10, 2007; Published online: November 21, 2007
Supporting information for this article is available on the WWW under http://asc.wiley-vch.de/home/.
Abstract: The hydrogenation of complex pharma-
ceutical intermediate, rac-sertraline imine has been
optimized as a continuous flow process utilizing a
palladium/calcium carbonate (Pd/CaCO3) catalyst
and hydrogen in supercritical carbon dioxide
(scCO2). Superior levels of selectivity were obtained
in the flow system possibly attributable to the heat
transfer properties of scCO2 which help to remove
excess heat from the catalyst surface.
The application of scCO2 as a solvent for the syn-
thesis of pharmaceuticals has not yet been fully realiz-
ed. This is because synthetic research in scCO2 has fo-
cused largely on model substrates and compounds rel-
event to the bulk and fine chemicals industry.[11–13]
This article explores the use of scCO2 as a solvent for
the hydrogenation of a complex pharmaceutical inter-
mediate,
[4-(3,4-dichlorophenyl)-3,4-dihydro-2H-
naphthalenyidene]methylamine, also known as rac-
sertraline imine 1.
Keywords: carbon dioxide; diastereoselectivity; hy-
drogenation; sertraline; supercritical fluids
(4S)-Sertraline imine 1a is an intermediate in the
synthesis of cis-(1S,4S)-sertraline hydrochloride 2,
marketed under the name, Zoloftꢀ (Scheme 1), a
multi-billion dollar drug developed originally by
Pfizer for the treatment of depression and other anxi-
ety related disorders.[14,15] Today Zoloftꢀ is manufac-
The excessive use of organic solvents in synthesis, pu- tured in relatively few steps by first reacting 1-naph-
rification and cleaning is the largest single contributor thol, 3, and 1,2-dichlorobenzene, 4, with aluminium
to overall chemical waste in the pharmaceutical indus- chloride to provide the racemic intermediate ketone.
try. Therefore, a new technology that minimizes the Simulated moving bed (SMB) chromatography is then
use of organic solvents will help make pharmaceutical used to separate efficiently the (4S)- and (4R)-enan-
and fine chemical synthesis more sustainable. Super- tiomers. Enantiomerically pure (4S)-ketone is then re-
critical fluids (SCFs) and, in particular, supercritical acted with methylamine to afford the imine 1a. Dia-
CO2 (scCO2), have emerged as alternative green sol- stereoselective hydrogenation puts in the second
vents which can help the pharmaceutical industry chiral centre to afford (4S)-sertraline before conver-
attain this goal.[1,2] For instance, SCF chromatogra- sion to the hydrochloride salt 2 (Scheme 1).
phy[3,4] and SCF extraction[5,6] have proved to be pow-
We chose to study the hydrogenation of imine 1a in
erful tools for the separation of active pharmaceutical scCO2 for a number of reasons. It has both regio- and
ingredients (APIs) from complex mixtures. scCO2 has chemoselectivity challenges, and a large body of infor-
also more recently been applied to anti-solvent pre- mation exists relating to this hydrogenation utilizing
cipitation of drug particles from solution.[7–10] scCO2 is conventional molecular solvents in batch processes.
highly suitable for pharmaceutical processing because Separation of dechloro by-products can be problemat-
it is non-toxic and has relatively mild critical parame- ic and a key goal is to see if continuous hydrogenation
ters (Tc =318C, Pc =74 bar) so that the advantages of in scCO2 can control aryl dechlorination. We used
near-critical and scCO2 can be achieved, whilst oper- rac-imine 1 as the substrate, rather than the optically
ating at relatively low temperatures where degrada- pure (4S)-enantiomer 1a. Therefore, upon hydrogena-
tion of sensitive molecules is unlikely.
tion, formation of all four stereoisomers, 5a–d, is pos-
Adv. Synth. Catal. 2007, 349, 2655 – 2659
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2655

Peter Clark et al.
COMMUNICATIONS
Scheme 1. Current Pfizer synthesis to make Zoloftꢀ. (a) AlCl3; (b) simulated moving bed chromatography (SMB); (c)
MeNH2, EtOH; (d) Pd/CaCO3, H2, EtOH; (e) HCl, EtOH.[14]
sible, the goal being to maximize formation of the re- times are often required to achieve high levels of con-
quired cis-isomers, 5a and 5b (Scheme 2).
version. scCO2 is particularly attractive as a solvent
In the original Pfizer process, reduction of rac- for hydrogenation reactions since H2 and scCO2 are
imine 1, was performed using NaBH4 to yield the cis- completely miscible. The absence of interphase mass
and trans-diastereoisomers in equal quantities. Pfizer transport effects means that hydrogenation can be be
then switched to a process using a Pd/C catalyst and performed efficently in a continuous flow system.[16]
H2 to enhance diastereoselectivity and provide a cis: High throughput and increased safety due to a small-
trans ratio of 70:30. Despite improved diastereoselec- er reactor volume make continuous flow technology
tivity, the presence of dechlorinated by-products led an attractive alternative to batch processing. A sche-
to downstream purification issues. In their current matic of our flow apparatus is shown in Figure 1.
process, Pfizer use a Pd/CaCO3 catalyst which pro-
vides excellent diastereoselectivity (cis:trans ratio of
95:5) and <1.0% dechlorinated by-products.[15]
The limited solubility of H2 in conventional liquid
phase hydrogenation reactions can be rate limiting.
Therefore, large presures of H2 and long reaction
Figure 1. The continuous flow reactor is built from commer-
cially available components. The substrate is mixed with
high pressure CO2 and H2 before it is pumped over a fixed-
bed reactor containing heterogeneous catalyst. System pres-
sure is controlled via a back pressure regulator (BPR). De-
pessurization at the end of system affords separation of the
solvent CO2 (gas) from the product (liquid). A more de-
tailed description can be found elsewhere.[17]
Scheme 2. Diastereoselective hydrogenation of imine
should be performed to maximize formation of cis-amines
5a and 5b.
1
2656
asc.wiley-vch.de
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2007, 349, 2655 – 2659

Continuous Flow Hydrogenation of a Pharmaceutical Intermediate
COMMUNICATIONS
Table 1. Initial catalyst screening for the continuous hydro-
genation of imine 1 in scCO2.
Catalyst[a]
T
Conv.
cis:trans
Chemoselectivity
[%][b]
[8C] [%][b]
ratio[c]
2a–2d
6–10
Pt/C
Pd/C
Pd/
80
80
80
98
98
98
56:44
87:13
95:5
100
94
96
0
6
4
CaCO3
[a]
Reaction conditions: system pressure=175 bar; mass of
5% metal catalyst=0.5 g; H2:substrate ratio=3:1; con-
centration of
1 in THF=0.2M; flow rates: CO2 =
1.0 mLminÀ1, organic=0.4 mLminÀ1.
[b]
[c]
Calculated by GLC using normalized peak areas
Calculated by HPLC using peak areas.
The initial challenge was to find a catalyst that
would exhibit high levels of activity in a flow system
while delivering high levels of diastereo- and chemo-
selectivity. Since imine 1 is a solid under ambient con-
ditions, a concentrated THF solution of imine 1 was
pumped into the reactor. EtOH can also be used as
co-solvent; however, the solubility of imine 1 is signif-
icantly lower than in THF.
Figure 2. By-products 6–9 formed during the hydrogenation
of imine 1 over the Pd/C and Pd/CaCO3 catalysts were iden-
tified by GC-MS. By-product 10 was identified using NMR
and comparison with an authentic sample.
Three different heterogeneous catalysts were
screened. The first, Pt/C, had been shown in model
studies to be selective towards imine hydrogenation Pd/CaCO3 catalyst, imine 1 underwent 99% conver-
in the presence of aryl chlorides.[18] However, it was sion with 80% selectivity to amine 10. The other
found that while dechlorination did not occur, diaste- products were amines 5a–d.
reoselectivity was poor (entry 1, Table 1) and varia-
tion in temperature and other perturbations did not H2-starved conditions, it is proposed that 10 is formed
lead to any significant improvement. via a Pd-catalyzed disproportionation of imine 1 to
Given the high yield of naphthylamine 10 under
Changing from Pt/C to a Pd/C catalyst gave a sig- naphthylamine 10 and amines 5a–d or via an oxida-
nificant improvement in diastereoselectivity but a tive dehydrogenation mechanism (Scheme 3). Howev-
small amount of dechlorination was observed er, we cannot rule out the conversion of imine 1 to 10
(entry 2). Even without optimization, the third cata- via the undetected oxidized intermediate 11.
lyst Pd/CaCO3 gave a cis:trans ratio of 95:5, the same
As further evidence of the oxidative mechanism
as that reported for the current Pfizer process with a leading to 10, imine 1 was reacted with 3-dichloro-5,6-
similar catalyst. Chemoselectivity was also high with dicyano-p-benzoquinone, DDQ, yielding a product
only 4% by-products detected (entry 3).
identical by NMR and GLC to 10 isolated in the hy-
Amines 6–8 are formed through dechlorination of drogenation experiments.[19]
one or both of the aryl chloro groups, while by-prod-
uct 9 is formed through further hydrogenolysis of the
-NHMe group and increases in concentration at tem-
peratures>808C (Figure 2). Note that we did not ob-
serve formation of a stable carbamate of the secon-
dary amine product. On occasions where solids were
found in the product stream, these were identified as
HCl salts, the HCl being produced via dechlorination.
As far as we are aware, the oxidized by-product
naphthylamine 10 has not previously been reported as
a by-product in any Zoloftꢀ literature. Amine 10 rep-
resents ca. 50% of all by-products formed in the ex-
periments in Table 1. However, at >808C with
H2:substrate ratios of <3:1, 10 was detected in larger
Scheme 3. Amine 10 can be formed from imine 1 over a Pd
quantities; for instance, with no H2 at 1208C, over the catalyst on active sites that are devoid of adsorbed H2.
Adv. Synth. Catal. 2007, 349, 2655 – 2659
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
2657

Peter Clark et al.
COMMUNICATIONS
the nitro group to occur. To establish whether this
happens in our system, a series of reactions was per-
formed in THF + C3H8 (Figure 3). scC3H8 (Tc =
978C, Pc =43 bar) was chosen because it has previous-
ly been used as a solvent for hydrogenation reac-
tions.[22]
Reactions performed in C3H8 exhibit selectivity at
least as high as in scCO2, suggesting that CO poison-
ing is unlikely to be the cause of the enhanced che-
moselectivity. However, SCFs are known to exhibit
greater diffusivity and lower viscosity than organic
solvents which makes them excellent at removing
heat.[23] Therefore, the enhanced chemoselectivity
might be attributable to the ability of scC3H8 or
scCO2 to prevent hot-spot formation on the catalyst
bed, which would otherwise lead to by-products.[24] It
should be noted, however, that C3H8 is less suitable
Figure 3. Comparison of overall chemoselectivity towards
chlorinated products 5a–d, c, and dechlorinated products
6–9, g, in the presence and absence of SCFs. Traces are
~
^
labelled: & performed in THF only;
scCO2 +THF *;
scC3H8 + THF **. Reaction conditions: H2 P=1 bar; 0.05 g than scCO2 for industrial applications since it is
of 5% Pd/CaCO3 cat.; 4.5 mL of 0.05M solution in THF; re-
action time=40 min; * CO2 P=175 bar; ** C3H8 P=
175 bar.
highly flammable.
Optimization of hydrogenation of 1 in the flow
system using scCO2 showed that by working at 408C
over Pd/CaCO3, in the presence of a large excess of
At >1008C, Pd is known to catalyze the oxidative H2, it was possible to suppress almost completely all
aromatization of dehydro-1-napthylamines.[20] Forma- dechlorination and dehydrogenation side reactions,
tion of 10 may occur, even with excess H2, as a result achieving a chemoselectivity of >99%. Diastereose-
of compaction of the Pd catalyst inside the flow reac- lectivity was also excellent at a cis:trans ratio of 97:3,
tor. This makes it difficult for H2 to be evenly distrib- which is comparible to the current Pfizer process
uted and could lead to hydrogen-deficient active sites (Table 2).
in the catalyst bed, thereby promoting disproportiona-
In summary, we report the successful hydrogenation
tion and oxidative dehydrogenation of 1 rather than of a complex pharmaceutical intermediate in scCO2.
hydrogenation. This unexpected discovery of 10 high- The hydrogenation of rac-sertraline imine 1 can be
lights possible problems associated with compactable performed in continuous flow with higher levels of
hydrogenation catalysts.
diastereoselectivity and comparible levels of chemose-
It is important to identify any specific advantages lectivity to the current process. Under optimized con-
of using scCO2 compared to conventional solvents. dtions this yields cis-sertraline (5a and 5b) in 97%
Therefore a series of hydrogenation reactions was yield (including mass balance from flow apparatus).
conducted in THF and also THF + scCO2 mixtures While diastereoselectivty appears to be affected pri-
inside a batch autoclave at 25–1208C. Temperature marily by the choice of metal and catalyst support,
had little effect on diastereoselectivity; the cis:trans chemoselectivty is affected significantly by the pres-
ratio remaining at between 92:8 and 93:7 in both sol-
vent systems. However, there was a dramatic differ-
ence in the temperature dependence of chemoselec-
tivity (Figure 3). Reactions in THF only undergo a
change in chemoselectivity at a much lower tempera-
Table 2. Comparison between the scCO2 flow system and
the Pfizer batch system for the hydrogenation of rac-sertra-
line imine 1.
System
P
T
Conv.
cis:trans
Chemoselectivity
[%][b]
ture than those in the presence of scCO2. For in-
stance, at 808C, selectivity towards total dechlorinated
by-products is reduced from 78% in THF to only
20% when scCO2 is present.
In 2005, Ikariya et al. reported the effects of scCO2
on the chemoselective hydrogenation of chloro-
nitroaromatics over a Pt/C catalyst where lower levels
of dechlorinated by-products were found for reactions
performed with scCO2 compared with those with-
out.[21] It was suggested that CO, formed via the Pt-
catalyzed reverse water-gas-shift reaction, poisons se-
lectively the catalytic sites responsible for promoting
dechlorination, while still allowing hydrogenation of
[bar] [8C] [%][b]
ratio[c]
5a–5d
6–10
[a]
scCO2
175 40
99
5
97: 3
99
<1
Batch[15]
1
2
99
95: 5
99 <1[d]
[a]
Reaction conditions: system pressure=175 bar; mass of
5% Pd/CaCO3 catalyst=0.5 g; H2:substrate ratio=10:1,
concentration of 1 in THF=0.2M; flow rates: CO2 =
1.0 mLminÀ1, organic=0.4 mLminÀ1.
[b]
[c]
[d]
Calculated by GLC using normalized peak areas.
Calculated by HPLC using peak areas.
The presence of 10 is not reported in any Zoloftꢀ litera-
ture.
2658
asc.wiley-vch.de
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2007, 349, 2655 – 2659

Continuous Flow Hydrogenation of a Pharmaceutical Intermediate
COMMUNICATIONS
[3] W. Huang, X. Zhang, H. Liu, J. H. Shen, H. L. Jiang,
Tetrahedron Lett. 2005, 46, 5965.
[4] F. Fusaro, M. Hanchen, M. Mazzotti, G. Muhrer, B.
Subramaniam, Ind. Engin. Chem. Res. 2005, 44, 1502.
[5] E. Reverchon, I. De Marco, J. Supercrit. Fluids 2006,
38, 146.
[6] M. Molnar, E. Szekely, B. Simandi, S. Keszei, J. Lovasz,
E. Fogassy, J. Supercrit. Fluids 2006, 37, 384.
[7] P. Pathak, M. J. Meziam, T. Desai, Y. P. Sun, J. Super-
crit. Fluids 2006, 37, 279.
ence of scCO2. It is suggested that the excellent heat
transfer properties of the scCO2 help to maintain ex-
ceptional levels of chemoselectivity, even at elevated
temperature. In the future it is hoped that scCO2 flow
systems will be successfully applied to other reactions
and substrates of interest to the pharmaceutical indus-
try. It will also be interesting in the future to investi-
gate the economic drivers to change from a batch pro-
cess in conventional solvents to flow reaction in a
scCO2 by comparing the investment and operation
costs of the two processes.
[8] P. Ginty, M. J. Whitaker, K. M. Shakesheff, S. M.
Howdle, Mater. Today 2005, 8, 42 .
[9] M. Charoenchaitrakool, F. Dehghani, N. R. Foster, Int.
J. Pharmaceut 2002, 239, 103.
[10] E. Reverchon, R. Adami, J. Supercrit. Fluids 2006, 37,
Experimental Section
1.
[11] X. L. Xie, C. L. Eckert, Ind. Eng. Chem. Res. 2004, 43,
7907.
[12] E. J. Beckman, J. Supercrit. Fluids 2004, 28, 121.
[13] M. Solinas, A. Pfaltz, P. G. Cozzi, W. Leitner, J. Am.
Chem. Soc. 2004, 126.
[14] G. J. Quallich, Chirality 2005, 17, S120.
[15] G. P. Taber, D. M. Pfisterer, J. C. Colberg, Org. Process
Res. Dev. 2004, 8, 385.
All catalysts used in this study were supplied by Johnson
Matthey and used as supplied. The starting material, rac-
imine 1 was prepared using literature procedures with a
purity of >98%. Analysis was performed by NMR (Jeol
270 MHz). For all hydrogenation studies, conversion and
chemoselectivity were calculated using GLC (Shimadzu
2010) fitted with an FID detector. Relative response factors
were used for all accurate calculations. In all cases, an achi-
ral RTX-5 column (0.52mm ID, 0.25 mm film thickness) was
used and He as a carrier gas. A Waters 600 E HPLC ma-
chine was used to calculate accurately the cis:trans ratio. It
was fitted with a UV-VIS detector set to 154 nm. A chiral
HPLC column (Diacel Chirasil OD-H column, 30 cm) was
used with a mobile phase of 2% 2-PrOH in n-hexane with
0.2% diethylamine. The flow rate was set to 1.0 mLminÀ1
and the temperature isothermal at 258C. By-product iden-
tification was performed using a Thermo-Finnegan Polaris
Q ion trap GC-MS fitted with an RTX-5 MS column at-
tached via a heated transfer line to the ion trap mass spec-
trometer (CI mode) and He carrier gas. The ion-trap was
used in the CI mode. All batch reactions were conducted
inside a stainless steel autoclave (Mk type 1) sealed to a
torque of 220 Nm equipped with a magnetic stirrer bar.
[16] P. Licence, J. Ke, M. Sokolova, S. K. Ross, M. Poliakoff,
Green Chem. 2003, 5, 99.
[17] P. Stephenson, P. Licence, S. K. Ross, M. Poliakoff,
Green Chem. 2004, 6, 521.
[18] Model studies were performed in the continuous flow
apparatus using N-[1-(4-chlorophenyl)ethylidene]me-
thanamine (below) to assess the chemoselectivity of
imine hydrogenation vs. dechlorination. Results are not
reported here.
[19] Experimental procedure and data can be found in the
Supporting Information.
Acknowledgements
[20] M. Chakrabarty, S. Sarkar, Y. Harigaya, Synthesis 2003,
2292.
[21] T. I. Ikariya, M. Tada, Y. Iwasawa, Chem. Commun.
2005, 924.
We thank EPSRC, the CRYSTAL Faraday Partnership, As-
traZeneca and Thomas Swan & Co. Ltd. for support, Kem-
protec Ltd. for chemicals, and M. Guyler, P. Fields and R.
Wilson for technical support.
[22] S. van den Hark, M. Härrçd, Appl. Catal. A: Gen.
2001, 210, 207.
[23] M. A. McHugh, V. J. Krukonis, Supercritical Fluid Ex-
traction: Principles and Practise, Vol. X, 2nd edn., But-
terworth-Heinmann, Boston, MA, 1994.
[24] C. A. Mantelis, F. Lavanchy, T. Meyer, J. Supercrit.
Fluids 2007, 40, 376.
References
[1] M. Perrut, Stp Pharma Sciences 2003, 13, 83.
[2] B. Subramaniam, R. A. Rajewski, K. Snavely, J. Phar-
maceut. Sci. 1997, 86, 885.
Adv. Synth. Catal. 2007, 349, 2655 – 2659
ꢁ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
2659