Catalytic Asymmetric Reduction of a 3,4-Dihydroisoquinoline for the Large-Scale Production of Almorexant: Hydrogenation or Transfer Hydrogenation?

Article abstract of DOI:10.1021/op400268f

Several methods are presented for the enantioselective synthesis of the tetrahydroisoquinoline core of almorexant (ACT-078573A), a dual orexin receptor antagonist. Initial clinical supplies were secured by the Noyori Ru-catalyzed asymmetric transfer hydrogenation (Ru-Noyori ATH) of the dihydroisoquinoline precursor. Both the yield and enantioselectivity eroded upon scale-up. A broad screening exercise identified TaniaPhos as ligand for the iridium-catalyzed asymmetric hydrogenation with a dedicated catalyst pretreatment protocol, culminating in the manufacture of more than 6 t of the acetate salt of the tetrahydroisoquinoline. The major cost contributor was TaniaPhos. By switching the dihydroisoquinoline substrate of the Ru-Noyori ATH to its methanesulfonate salt, the ATH was later successfully reduced to practice, delivering several hundreds of kilograms of the tetrahydroisoquinoline, thereby reducing the catalyst cost contribution significantly. The two methods are compared with regard to green and efficiency metrics.

Full text of DOI:10.1021/op400268f

Article
pubs.acs.org/OPRD
Catalytic Asymmetric Reduction of a 3,4-Dihydroisoquinoline for the
Large-Scale Production of Almorexant: Hydrogenation or Transfer
Hydrogenation?
Gerard K. M. Verzijl,† Andre H. M. de Vries,† Johannes G. de Vries,† Peter Kapitan,‡ Thomas Dax,‡
́
Matthias Helms,‡ Zarghun Nazir,‡ Wolfgang Skranc,‡ Christoph Imboden,∥,⊥ Juergen Stichler,∥
Richard A. Ward,§ Stefan Abele,*,∥ and Laurent Lefort*,†
†DSM Innovative Synthesis BV, P.O. Box 18, 6160 MD Geleen, The Netherlands
‡DSM Fine Chemicals Austria, St. Peter Strasse 25, 4021 Linz, Austria
§GlaxoSmithKline, Medicines Research Centre, Gunnels Wood Road, Stevenage SG1 2NY, United Kingdom
∥Actelion Pharmaceuticals Ltd., Gewerbestrasse 16, 4123 Allschwil, Switzerland
S
* Supporting Information
ABSTRACT: Several methods are presented for the enantioselective synthesis of the tetrahydroisoquinoline core of almorexant
(ACT-078573A), a dual orexin receptor antagonist. Initial clinical supplies were secured by the Noyori Ru-catalyzed asymmetric
transfer hydrogenation (Ru-Noyori ATH) of the dihydroisoquinoline precursor. Both the yield and enantioselectivity eroded
upon scale-up. A broad screening exercise identified TaniaPhos as ligand for the iridium-catalyzed asymmetric hydrogenation
with a dedicated catalyst pretreatment protocol, culminating in the manufacture of more than 6 t of the acetate salt of the
tetrahydroisoquinoline. The major cost contributor was TaniaPhos. By switching the dihydroisoquinoline substrate of the Ru-
Noyori ATH to its methanesulfonate salt, the ATH was later successfully reduced to practice, delivering several hundreds of
kilograms of the tetrahydroisoquinoline, thereby reducing the catalyst cost contribution significantly. The two methods are
compared with regard to green and efficiency metrics.
transfer hydrogenation with a chiral N-sulfonated diamine−
Ru(II)−η6-arene catalyst has been the method of choice.4 In
most cases, the reducing agent is a 5:2 HCOOH/Et3N mixture
(TEAF), leading to the formation of the desired product with
high enantiomeric excess (ee). The initial trials with the Noyori
catalyst performed at Actelion Pharmaceuticals Ltd. were
promising. Using a standard Ru(cymene)(Ts-DPEN) catalyst,
the desired product (Scheme 2) was obtained in high yield
INTRODUCTION
In 2005, Actelion Pharmaceuticals disclosed a 1,2,3,4-
tetrahydroisoquinoline derivative (almorexant, 1) (Scheme 1)
■
Scheme 1. Almorexant and Its Key Intermediates
Scheme 2. Asymmetric Transfer Hydrogenation of 3
as an efficient antagonist for orexin receptors. Orexins are
hypothalamic peptides that play an important role in the sleep−
wake cycle of mammals. At the time, almorexant was the first
compound active on both the orexin OX(1) and OX(2)
receptors. Because of its novel mechanism of action, it appeared
as a very promising drug for the treatment of sleeping
disorders.1 The preparation of almorexant requires the
synthesis of an enantiopure key intermediate, 2, that could be
obtained by asymmetric reduction of the corresponding 3,4-
dihydroisoquinoline 3.
Historical Background of the Preparation of Enantio-
pure 2. A large number of methods for the enantioselective
reduction of this class of cyclic imines have been reported in the
literature.2 Since the seminal paper of Noyori and co-workers,3
(95%) with good ee’s (81−95%) at substrate to catalyst (S:C)
ratio of 500 on a 250−300 g scale. The formation of the HCl
salt of the product in methanol allowed enantiomeric
enrichment up to 99% ee. Unfortunately, these good results
were not reproducible on a larger scale. During a 30 kg
campaign performed at a toll manufacturer, a severe drop in the
yield (57−60%) and ee (76−80%) was observed in addition to
Received: September 26, 2013
Published: November 22, 2013
© 2013 American Chemical Society
1531
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
the presence of a byproduct, the N-formylated tetrahydroiso-
quinoline. Such a process with varying results was judged unfit
for longer-term production. Alternative routes relying on non-
enantioselective hydrogenation and classical resolution were
designed. Upon screening of a set of chiral acids, D-tetranilic
acid5 was identified as a good resolving agent, leading to 41%
yield of the desired diastereoisomer with 98% ee. The
undesired enantiomer could be recycled via oxidation to the
imine with low-cost bleach.6 Although this protocol delivered
the desired intermediate with a high level of purity, an
asymmetric synthesis route was more desirable in order to
increase the overall yield and limit the numerous solvent
switches due in part to the recycling steps.
Hence, a route relying on catalytic asymmetric hydro-
genation (AH) of the cyclic imine was investigated. Hydro-
genation of 3,4-dihydroisoquinoline with molecular hydrogen
has also been reported with different metals such as Ti, Ru, Rh,
and Ir.7 Catalyst screening was performed by a commercial
screening company using 11 ligands from four classes of
bidentate phosphines along with various metal precursors and
solvents. Modest results were obtained with Ir (maximum 56%
ee).
There have been several reports by DSM on catalytic
asymmetric hydrogenation processes using the proprietary
MonoPhos library (codeveloped with the University of
Groningen)8 or other readily available catalytic systems and/
or ligands.9 In this paper, we present the initial search for a
cost-efficient asymmetric hydrogenation catalyst, resulting in
the selection of a very effective iridium−TaniaPhos-based
catalyst for the enantioselective conversion to 2 that was used
on a ton scale. However, the high cost contribution of this latter
catalyst triggered the search for a more competitive alternative.
We chose to revisit the route based on asymmetric transfer
hydrogenation (ATH) that was originally developed by
Actelion Pharmaceuticals Ltd. A second process for the
multikilogram production of intermediate 2 based on a Noyori
Ru catalyst was then developed. The pros and cons of the two
catalytic technologies−asymmetric hydrogenation and asym-
metric transfer hydrogenation−for this particular compound
will be presented as well as a justification for our final catalyst
choice.
1. Solvent Screening with a Phosphoramidite Ligand-
Based Hydrogenation Catalyst. The DSM screening
procedure for an asymmetric hydrogenation catalyst first
involves the identification of the best reaction conditions for
a given class of catalysts. A representative member of the
catalyst class is then chosen and tested in different solvents at
different H2 pressures and temperatures in the presence or
absence of additives. The experiments are carried out in a
medium-throughput parallel Endeavor reactor where up to
eight hydrogenations can be run in parallel.10 On the basis of
earlier results, we anticipated that Ir would be more efficient
than Rh for the catalytic asymmetric hydrogenation of imine 3,
using as the ligand a phosphoramidite with substituents at the
3- and 3′-positions (L2 in Figure 1) instead of the standard
ligand MonoPhos (L1).11 In Figure 1, we show the results of
the solvent screening.
As is often the case in asymmetric hydrogenations, the
solvent played an important role. Nonprotic solvents such as
dichloromethane (DCM), ethyl acetate (EtOAc), methyl tert-
butyl ether (MTBE), and methyl ethyl ketone (MEK) gave the
best results in terms of ee together with acceptable conversions.
The highest conversions were obtained with acetic acid as the
solvent or as an additive in DCM, albeit with low ee’s. Acids
can indeed scavenge the amine product as its salt and prevent
catalyst inhibition.12 (S)-MonoPhos was also tested under the
same conditions in DCM and gave 67% conversion with only
17% ee, confirming the superiority of bulkier phosphoramidites
for the desired transformation.
2. Reaction Condition Screening Using Four Ligands
and Robotic Equipment. With the aid of robotic equipment
and a high-throughput A96 parallel reactor,13 we explored the
influence of Ir precursors and additives in 96 reactions in
parallel. Since the discovery of the Metolachlor imine
hydrogenation catalyst,14 iodine is commonly used to improve
the performance of Ir catalysts in imine hydrogenation.15
Phthalimide has also been reported as an efficient additive.16
Two Ir precursors, Ir(COD)2BF4 and [Ir(COD)Cl]2 (cationic
and neutral, respectively) were used together with two bulky
phosphoramidite ligands (R)-L2 and (S,S,S)-L3 (the latter
being sterically hindered on the amine side), the axially chiral
ligand L4, and a ferrocene-based ligand from the Solvias
collection, L5. The resulting catalysts were tested in four
solvents (EtOAc, MIBK, DCM, and IPA) in the absence or
presence of I2 or phthalimide as an additive, thus generating a
total of 96 different catalytic mixtures, all tested in parallel
within 24 h (Scheme 4).
The color-coded table in Scheme 4 shows that most of the
ee’s obtained are lower than 50%. Regardless of the ligand, the
precursor [Ir(COD)Cl]2 (Ir 2) appeared to be much better
than Ir(COD)2BF4 (Ir 1). In the case of the two
phosphoramidite ligands, the best results were obtained in
EtOAc (62−66% ee), while the use of IPA resulted in very low
ee’s. Overall, iodine had a negative effect on the performance of
the phosphoramidite/Ir catalysts, while the presence of
phthalimide did not have much influence. From the non-
phosphoramidite reactions only one specific combination
stands out: the ferrocene-based ligand L5 in the presence of
I2 with EtOAc as the solvent and [Ir(COD)Cl]2 as the metal
source gave 76% ee.
RESULTS AND DISCUSSION
■
The cyclic imine 3 was prepared in three steps from
commercially available trans-4-(trifluoromethyl)cinnamic acid
(Scheme 3).1 The ring formation was accomplished via the
Bischler−Napieralski reaction of the corresponding amide, with
the product being isolated as its mesylate salt, 3·CH3SO3H.
Depending on the reduction method used, the substrate was
either used as such or first converted to the free base.
Scheme 3. Synthetic Route towards Imine 3
3. Further Investigation of the MonoPhos Lead.
Having set the reaction conditions for the Ir/phosphoramidite
catalyst ([Ir(COD)Cl]2 as the Ir source, no additive, EtOAc as
the solvent), we varied the phosphoramidite ligands in order to
1532
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
Figure 1. Solvent screening with Ir/3,3′di-Me-MonoPhos (L2) for the hydrogenation of 3 to 2. Conditions: [Ir(COD)Cl]2 (0.005 mmol), L2 (0.02
mmol), 3 as the free base (0.2 mmol), solvent (5 mL), and H2 (25 bar) at 50 °C for 16 h.
a
Scheme 4. Layout of A96 Reactor Screening and the Obtained ee’s
a
Conditions: [Ir(COD)Cl]2 (0.00055 mmol), ligand (1 or 2 equiv/Ir), 3 as the free base (0.03 mmol), solvent (1.5 mL), and H2 (25 bar) at 50 °C
for 16 h. Almost all of the conversions were above 90% and are not mentioned here for clarity.
increase the enantioselectivity. A large number of ligands were
tested (Scheme 5), and the importance of the substitution at
the 3- and 3′-positions of the binaphthol backbone was
confirmed. In the absence of these substituents, the ee’s were
quite low. The same was true when very bulky groups such as
tert-butyl, neopentyl, or mesityl were present at the 3- and 3′-
positions. Phenyl substituents gave the best ee’s (79%). On the
other hand, no clear trends were found with the substitution
pattern on the amino group. Taddol- or catechol-based
phosphoramidites were also active but not particularly
enantioselective.
enrichment would occur upon crystallization of the product. Up
until now, hydrogenation had been performed at low S:C ratios
(i.e., between 25−200 mol:mol). However, decreasing the
catalyst amount by more than 1 order of magnitude was still
required to make the transformation cost-effective for large-
scale production. Unfortunately, a strong decrease in activity
and enantiomeric excess was observed at S:C ratios above 200.
At S:C = 300, incomplete conversion and an enantiomeric
excess of 44% was obtained with di-Ph-MonoPhos (L6) instead
of 79% at S:C = 25. Similarly, upon use of 3,3′-di-MePipPhos
(L7), the performance of the catalyst dropped from full
conversion and 69% ee at S:C = 200 to 67% conversion and
48% ee at S:C = 500. Several explanations17 are possible for this
behavior, the most obvious one being the presence of an
Although the best enantiomeric excesses obtained with
phosphoramidites remained modest, we decided to further
optimize the catalytic system, anticipating that enantiomeric
1533
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
Scheme 5. Enantiomeric Excesses (ee’s) for Selected Examples of Phosphoramidite Ligands Tested in the Ir-Catalyzed
a
Hydrogenation of 3
a
Conditions: [Ir(COD)Cl]2 (0.002 mmol), ligand (0.004 mmol), 3 as the free base (0.1−0.2 mmol, S:C = 25−50), EtOAc (5 mL), and H2 (25 bar)
at 50 °C for 2 h. Conversions are not shown but were >20% for all examples.
a
Scheme 6. Results of the Screening of 28 Ligands from the Solvias Collection
a
Conditions: [Ir(COD)Cl]2 (0.00163 mmol), ligand (0.00326 mmol) (structures are shown in the Supporting Information), 3 as the free base
(0.0326 mmol, S:C = 10), EtOAc (5 mL), and H2 (25 bar) at 50 °C for 18 h. Conversion was >90% for all examples except for M002-1 (73%) (see
the Supporting Information).
Scheme 7. Asymmetric Hydrogenation of 3 with Ir/TaniaPhos
impurity in the substrate that acts as a poison for the catalyst.
Another explanation is substrate and/or product inhibition.
Increasing the S:C ratio favors the binding of the substrate and/
or the product to the metal center. Eventually the chiral ligand
is displaced, leading to the formation of inactive or non-
enantioselective Ir species. Because phosphoramidites are
monodentate ligands, their displacement by N-containing
compounds may occur more readily than for bidentate ligands.
4. Further Improvement of the Ferrocene Lead. In
contrast to phosphoramidites, the ferrocene ligand/Ir catalyst
exhibited improved performance in the presence of iodine.
Using the high-throughput screening platform, we tested
another 28 ferrocenyl disphosphine ligands available in house
with [Ir(COD)Cl]2 as the precursor in EtOAc as a solvent with
and without iodine as an activator (Scheme 6).
To our delight, one single ligand out of the whole series,
TaniaPhos T002-1 (R = R′ = cyclohexyl) resulted in full
conversion and the highest ee ever obtained in our hands
(94%).18 Although this was not the case for all of the ligands
tested, the addition of I2 was crucial to render the Ir/TaniaPhos
combination highly enantioselective. The second-best results
were obtained using JosiPhos-type ligands with or without
iodine. Initially, the results were variable until we realized that
this was caused by incomplete coordination of the ligand to the
Ir precursor. When we performed the first catalyst preparation
in the A96 reactor, we allowed for a long incubation time of the
1534
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
Scheme 8. Three-Step Process from 3·CH3SO3H to 2·HOAc As Performed in the Laboratory Phase (Lab) and in Production
a
(Prod.)
a
The concentrations of 3·CH3SO3H and 3 were 6.6 and 23% w/w in the lab and production, respectively. The concentration of the intermediate
solution containing 3 was adjusted to 5 and 18% w/w in the lab and production, respectively.
Ir precursor with the chiral ligand (around 1 h) prior to the
addition of iodine since 96 reaction mixtures had to be
prepared in parallel. In our subsequent trials where we tested
only the successful TaniaPhos, the time for formation of the
precatalyst was shortened to 10−20 min. Surprisingly, the
complexation was not complete under these conditions, and
upon addition of I2, active but non-enantioselective Ir species
were formed, leading to a lower ee of the product. Hence, once
the Ir precursor and the ligand were allowed to form the
complex in DCM (at rt for 4 h, or at 40 °C for 30 min) prior to
the addition of I2, the results obtained in the A96 reaction (2.5
mL of solvent) were reproducible on a larger scale (70 mL of
solvent) (Scheme 7).19
5. Scale-Up of Asymmetric Hydrogenation. For the
pilot plant campaign (six batches of 45 kg) and the validation
campaign (10 batches of 750 kg), a lab improvement program
was initiated with the following targets. First, the number of
solvents used in the process for the production of the desired
HOAc salt of 2 should be reduced (see Scheme 8). In the lab
experiments, four solvents (EtOAc, DCM, toluene, and
heptane) had been used. Second, we wanted to reduce the
amount of catalyst because of the high costs of Ir and
TaniaPhos (target: S:C > 1500).
MeOH improved the activity of the catalyst up to S:C = 3000
without loss of ee and activity. The exact role of MeOH is not
fully understood. It is assumed that it can coordinate to the Ir
center and either favors its activation by I2 and/or stabilizes the
active species. Unfortunately, the catalytic system appeared to
be very poorly soluble in MeOH. Therefore, the MeOH
preparation could not be used for large-scale production.
However, preparing the catalyst in a 1:1 DCM:MeOH mixture
appeared to be a good compromise, as it combined the benefits
of MeOH (albeit less than for pure MeOH) with good
solubility. With this preparation method, an S:C ratio of 2000
was reached during the validation campaign. The hydro-
genation in toluene also appeared to be much more tolerant to
high substrate concentration. In EtOAc, increasing the
concentration of free imine above 5.1% w/w led to a decrease
in ee, while in toluene a concentration as high as 18% w/w
could be used.
Finally, the final crystallization of the acetate salt of 2 could
be efficiently done in toluene at 0 °C, thereby avoiding the use
of heptane. Gratifyingly, enantiomeric enrichment from 91% ee
to 99.8% ee occurred during the crystallization. Residual DCM
and MeOH from the catalyst preparation had to be removed in
order to maintain a high yield.
A first improvement was quickly made when we observed
that the hydrogenation ran smoothly in toluene as the solvent
(using the same conditions as in EtOAc). The next step was to
verify whether efficient freebasing could be done in toluene.
The imine salt 3·CH3SO3H is less soluble in toluene than in
EtOAc and more water had to be used. However, since a more
concentrated solution of NaOH is compatible with toluene, the
overall volume of liquids could be reduced, allowing a larger
batch size (up to 23% w/w of free imine). More importantly,
the solvent switch for the final crystallization of 2·HOAc could
be avoided since the hydrogenation was performed in toluene
instead of EtOAc.
The next logical step was to test the preparation of the
catalyst in toluene. Although the Ir/TaniaPhos complex is very
soluble in toluene, a black precipitate formed immediately upon
addition of the iodine. This solid was tested in the
hydrogenation with toluene as the solvent, giving low
conversion and ee. As already mentioned, the performance of
the catalyst appeared to be dependent on its mode of
preparation, and further investigations led to a significant
reduction of the amount of catalyst. During a study performed
by Solvias, it was discovered that preparing the catalyst in
The proven acceptable range (PAR) study revealed the
importance of several other parameters. The iodine to Ir ratio
was ideally kept above 1.8. A lower amount led to lower
conversion and ee. The TaniaPhos ligand to Ir ratio also had to
be carefully controlled between 0.5 and 2.0. When the
hydrogenation was performed at 0 °C, the reaction rate was
about 4 times slower than at rt. On the other hand, at 35 °C,
the enantiomeric excess dropped to 87%. Concerning the
pressure, the reaction was effective at 3 bar. After conversion to
the free base, the imine could be stored for a few days as a
toluene solution. However, there was formation of some low-
level byproducts that acted as poison for the catalyst, leading to
a slight decrease in conversion of the hydrogenation at a similar
S:C ratio (from 99% to 95% conversion after 4 days of storage).
After 1 week of storage, the conversion was entirely inhibited,
and no hydrogenated product was observed. The main solvent
in the process is toluene. During the last step, the free amine 2
was precipitated as its acetate salt from the toluene solution
upon addition of glacial acetic acid. Traces of HOAc have a very
detrimental effect on the catalyst performance (low ee and
conversion). For that reason, the recycling process of toluene
involved an extraction with aqueous NaOH and water prior to
1535
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
Table 1. Details of the 10 Hydrogenation Batches with In-Process Control (IPC) by HPLC at Different Reaction Times and
Chiral Purities of the Batches
batch
1
2
3
4
5
6
7
8
9
10
S:C
1200
1500
2000
2500
2000
2000
2000
2000
2000
2000
HPLC area of 2 (%)
IPC 1 h
<0.1
−
−
0.6
−
−
20.4
1.2
0.8
−
36.8
6.0
25.0
2.0
0.7
−
7.1
1.0
0.7
7.0
0.7
0.5
−
17.1
1.7
−
1.4
1.0
0.7
−
2.0
1.0
0.5
−
IPC 2 h
IPC 3 h
2.0
IPC 4 h
−
−
2.0
1.1
chiral purity (% ee)
96.4
96.3
96.2
95.9
96.2
96.2
96.5
96.4
96.5
96.4
Scheme 9. Asymmetric Transfer Hydrogenation for the Production of 2
a
Table 2. Screening of Reaction Conditions in the Asymmetric Transfer Hydrogenation of 3 (or 3·CH3SO3H)
entry
conditions
T (°C)
time (h)
conv. (%)
ee (%)
N-formyl (area %)
1
2
3
4
5
6
7
8
3, DCM, TEAF = 5:2 (5.8 equiv), S:C = 1000/1
23
23
23
23
23
28
28
31
42
22
40
65
48
90
20
24
100
100
100
73
85
91
86
87
89
80
89
90
13.3
2.1
3·MsOH, DCM, TEAF = 5:2 (5.8 equiv), S:C = 1000/1
3·MsOH, AcOEt, TEAF = 5:2 (5.8 equiv), S:C = 1000/1
3·MsOH, IPA, TEAF = 5:2 (5.8 equiv), S:C = 1000/1
3·MsOH, EtOH, TEAF = 5:2 (5.8 equiv), S:C = 1000/1
3·MsOH, DCM, TEAF = 5:2 (3 equiv), S:C = 2000/1, reflux
3·MsOH, DCM, TEAF = 1:1 (5.8 equiv), S:C = 1500/1, reflux
3·MsOH, DCM, TEAF = 1:1 (3 equiv), S:C = 2000/1, reflux
6.3
2.4
55
1.2
100
99
15.6
4.2
100
2.0
a
The catalyst used for all experiments was (p-cymene)Ru(Cl)2(Ts-DPEN), see Scheme 9, reaction performed on 9 g scale, reflux, 28−32 °C, p = 550
mbar. In the column “conditions”, the parameter that is varied from one experiment to the other is underlined.
its reuse in hydrogenation. We found that up to 0.1% v/v of
water could be tolerated in the hydrogenation process.
However, on a slightly larger scale (9 g), the reaction became
slower (full conversion after 40 h; Table 2, entry 1) and slightly
less enantioselective (85% ee), but more worryingly, a side
product that had been overlooked so far was detected by HPLC
(13 area %) and identified as the N-formylated analogue of 2
(Scheme 9).
The validation campaign produced 6 t of the acetate salt of 2
in a total of 10 batches (see Table 1). The asymmetric
hydrogenation was performed in a large-scale reactor loaded
with 750 kg of the free amine under 6 bar H2. Gradually, the
S:C ratio was increased starting at 1200 for the first batch up to
2500 for the fourth batch. However, at this high S:C ratio, the
reaction did not reach completion even after 4 h. Therefore, the
remaining batches were run at S:C = 2000. The ee after the
asymmetric hydrogenation step was routinely ∼96% across the
campaign. After crystallization, all 10 batches met the
specification. Starting from 7.6 mT of imine salt, 6 mT of the
amine salt was produced (86% yield).
Although reasonably high S:C ratios were reached during the
validation campaign, the cost of the catalyst, and more
specifically of TaniaPhos, still represented a significant part of
the total production cost. Therefore, it was decided to continue
the search for a cheaper alternative. We turned towards
asymmetric transfer hydrogenation with the goal of demon-
strating the scalability of this transformation at high S:C ratios.
6. Screening for a Transfer Hydrogenation Catalyst.
Confirming the results obtained earlier with a standard Noyori
catalyst was rather straightforward (Scheme 9). The first
experiment at S:C = 1000 immediately gave the desired product
with full conversion and high ee (90%) after 18 h. The reducing
agent used was a mixture of formic acid and triethylamine in
the usual 5:2 azeotropic ratio as disclosed by Noyori.3
Reasoning that the mesylate salts of 2 and 3 would be less
prone to react with formic acid, we decided to introduce a
stoichiometric amount of CH3SO3H simply by undertaking the
transfer hydrogenation reaction directly from the imine salt, 3·
CH3SO3H (i.e., without prior freebasing). The unknown was
whether the imine salt would remain sufficiently active towards
transfer hydrogenation and how much the presence of an
additional acid would affect the enantioselectivity.20 To our
satisfaction, the reaction worked very well (Table 2, entry 2).
Full conversion was retained within 22 h with only 2.1 area %
for the N-formyl byproduct and an enantioselectivity of 91%.
Since DCM is not the most preferred solvent for large-scale
production, we performed a solvent screen (EtOAc, IPA,
EtOH; entries 3−5), but none of those solvents outperformed
DCM. In protic solvents, the formation of N-formylated 2 was
minimal, but unfortunately, the reaction was not complete even
after an extended time. The next step was to lower the amount
of catalyst. At S:C = 2000, the reaction went to completion but
required 3 days under reflux (entry 6), leading to a large
amount of N-formylated byproduct. It became apparent that
the key to limiting the formation of the byproduct was to
accelerate the reaction. It is known that the activity of the
1536
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
a
Table 3. Comparison of the Two Catalytic Routes (AH and ATH) towards the Key Chiral Intermediate 2
entry
parameter
substrate
AH
ATH
justifications
no need for freebasing in ATH
1
2
3
3
<
<
=
3·CH3SO3H
Ru-Noyori
catalyst
Ir/TaniaPhos
Ru cheaper than Ir, ATH ligand much cheaper
rated equally
operating conditions
S:C = 1500, 16−18 °C,
S:C = 1500, 28−32 °C,
6 bar H2
550 mbar
4
5
reactor technology
H2 infrastructure and
pressure reactor
<
>
standard reactor
ATH allows for more flexibility in multipurpose assets
reaction time
batch time
cycle time
10 h
97 h
18 h
32 h
115 h
44 h
long reaction time with Ru cat slows down throughput by a
factor of 2
6
relative concentration
(kg of product/m3)
1.0
<
1.4
ATH can be run at higher concentrations
7
8
solvents
toluene, heptane
>
>
<
>
=
DCM, toluene
ATH requires one solvent switch and utilizes DCM
solution ee
PMI24
96%
89%
9
22 (15 without H2O)
27 (21 without H2O)
10
11
relative variable cost/kg
relative fixed cost/kg
1.0
1.0
0.65
0.98
economically attractive ATH catalyst system
ATH is slower but run at higher concentration and less
equipment needed; effects are balanced
a
Entries 6, 10, and 11 have been normalized for the asymmetric hydrogenation (AH) process.
catalyst in transfer hydrogenation can vary significantly with the
pH (i.e., with the HCOOH:Et3N ratio).21 Using 1:1 TEAF
(entry 7), we indeed observed that the reaction occurred much
faster and, as expected, with a lower amount of side product.
Finally, by using a smaller excess of TEAF (entry 8), we could
perform the reaction at S:C = 2000 with an acceptable level of
byproduct that could be removed during workup.
The cleavage of the formyl group of the side product by base
or acid was not successful. Treatment with diluted bases or
acids showed no reaction, while strong acids or bases led to
decomposition. Extraction of 2 into an aqueous phase as its
acetate salt while leaving the N-formyl side product in the
organic phase appeared to be a good strategy. The N-
formylated 2 was depleted to ca. 1 area %. However, even
better was the direct crystallization of 2·HOAc after a solvent
switch to 4:1 toluene/heptane. In this case, the amount of N-
formyl side product was 0.2−0.6 area %.
7. Scale-Up of the Transfer Hydrogenation. This
process was used to produce two batches of 18 and 12 kg.
Both batches were run in a large-scale Hastelloy reactor at S:C
= 1500. The reaction was maintained under reflux between 28
and 32 °C at 550 mbar for 8 h. This allowed the efficient
removal of CO2 and therefore avoided its conversion to CO,
which would act as a poison for the catalyst.22 The two batches
performed equally well (89.7% ee and 89.4% ee, respectively).
After a solvent switch, the crystallization of the acetate salt of 2
was performed in toluene. Starting from 18 kg of 3·CH3SO3H,
14.5 kg of 2·HOAc was produced with a yield of 87%. Another
larger pilot campaign was run with 5 batches resulting in several
hundreds of kilograms of 2.23
8. Comparison of the Hydrogenation and Transfer
Hydrogenation Processes. Although our study of the
asymmetric transfer hydrogenation route was still ongoing, it
was sufficiently advanced to allow for a fair comparison of the
two routes. Table 3 summarizes the main features of both
technologies, highlighting their specific advantages. Both
processes have pros and cons. In the transfer hydrogenation
process, the imine salt 3·CH3SO3H can be used as a substrate,
thereby avoiding the freebasing step (entry 1). Although the
process based on transfer hydrogenation is slower than the one
based on hydrogenation (entry 5), the fixed costs of the two
processes are identical because higher concentrations can be
used with ATH. From an ecological and environmental point of
view, the hydrogenation technology is slightly better, as
indicated by the calculated process mass intensity (PMI)24
(entry 9). From an economic standpoint, the Noyori process is
more favorable because of the overall lower costs of goods,
mainly driven by the catalyst price (entries 2 and 10). For both
processes, the reaction conditions are mild and well within an
acceptable range for production (entry 3), although ATH offers
more flexibility as it can be carried out in any standard reactor
(entry 4).
CONCLUSION
■
Via catalyst screening on a lab scale, we discovered two efficient
catalytic systems for the enantioselective synthesis of the
intermediate 2·HOAc towards almorexant, one relying on
asymmetric hydrogenation using an Ir complex with a
ferrocene-based ligand and the other one relying on asymmetric
transfer hydrogenation using a Ru catalyst with a diamine
ligand. Both catalysts were studied further and appeared to be
suited for large-scale manufacturing. In case the production had
continued,23 a final choice would have been guided mainly by
the cost of goods for the two routes, considering the availability
of the production units and the delivery time of the raw
materials.
EXPERIMENTAL SECTION
■
(S)-6,7-Dimethoxy-1-(4-(trifluoromethyl)phenethyl)-1,2,3,4-
tetrahydroisoquinoline Acetate (2·HOAc).
Hydrogenation on a Production Scale. Freebasing. A
large-scale reactor was charged with toluene (2600 kg) and
purified water (1200 kg). 3·CH3SO3H (750 kg, 1.63 kmol) was
added, followed by 50% NaOH (160 kg). The mixture was
stirred for 1 h. After phase separation, the lower phase was
removed, and the organic phase was washed with purified water
(3 × 1000 kg). Water present in the organic phase was
removed by azeotropic distillation. The solution was transferred
into the 6.3 m3 hydrogenation reactor.
Precatalyst Preparation. A reactor was charged with MeOH
(3.2 L) followed by I2 (0.413 kg, 1.63 mol). A second reactor
was charged with [Ir(COD)Cl]2 (0.257 kg, 0.38 mol),
TaniaPhos (0.58 kg, 0.815 mol), DCM (22 L), and MeOH
(32 L). The mixture was stirred for 3 h at rt, after which time
1537
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
Present Address
the I2 solution was added. The mixture was stirred for 30 min
and then transferred to the hydrogenation reactor.
⊥C.I.: Newron Suisse S.A., Birsigstrasse 4, CH-4054 Basel,
Switzerland.
Hydrogenation. The reactor was pressurized with H2 (6 bar
min.) without stirring, and the pressure was released. The
reactor was again pressurized to 6 bar H2, and the stirrer was
started. The temperature was maintained between 16 and 18
°C. When the starting material was depleted to less than 2.0%
a/a (HPLC analysis), the reaction was stopped.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank our colleague Math Boesten from the analytical
group for his invaluable support in HPLC method develop-
ment. S.A. is indebted to Dr. Thomas Weller for constant
support and fruitful input. C.I., J.S., and S.A. thank Dr. Erhard
Bappert and Dr. Felix Spindler (Solvias AG) for their work on
the optimization of the AH conditions.25
Crystallization. A fraction of the reaction solvent was
removed by distillation under low pressure. Additional toluene
was added to adjust the concentration of 2 to between 7.8 and
9% (w/w). Acetic acid (98 kg, 1.63 kmol) was slowly dosed
while maintaining the temperature at 20 °C. The mixture was
stirred for 30 min. After filtration and drying, the final product
was packed into 40 kg drums. Overall yield of 2·HOAc: 86%,
REFERENCES
■
1
>96% ee. H and 13C NMR data corresponded to those in the
(1) (a) Weller, T.; Koberstein, R.; Aissaoui, H.; Clozel, M.; Fischli W.
WO 2005/118548, 2005. (b) In 2008, GSK and Actelion
Pharmaceuticals Ltd. entered into an exclusive worldwide collabo-
ration for Actelion’s almorexant, an orexin receptor antagonist in phase
III development with first-in-class potential as a treatment for primary
insomnia. (c) Part of this work was preliminarily disclosed: Abele, S.;
Dax, T. In Almorexant: Evolution of the Manufacturing Process.
Organic Process Research and Development Conference, Barcelona,
September 13−15, 2010.
literature.1a
Transfer Hydrogenation on a Production Scale. A
large-scale reactor was charged with 3·CH3SO3H (18 kg, 39.17
mol), DCM (59.9 kg), and formic acid (5.4 kg, 117.3 mol). The
solution was cooled to 0−10 °C, and Et3N (11.88 kg, 117.4
mol) was slowly added while maintaining the temperature
below 15 °C. The solution was heated to 20−25 °C, stirred for
1 h, and further heated to 30−32 °C. The batch was brought to
reflux at this temperature by applying vacuum (∼550 mbar)
and kept refluxing for 10 min before the catalyst, (chloro)-
{[(1R,2R)-(−)-2-amino-1,2-diphenylethyl](4-toluenesulfonyl)-
amido}(p-cymene)ruthenium(II), was added [17 g, 26.7 mmol
dissolved in degassed DCM (0.5 L)]. The reaction was
continued until the amount of starting material was less than
2.0% a/a (89.7% ee). The reactor was vented, and additional
DCM was added to compensate for the loss occurring during
the reaction. The batch was cooled to 20−25 °C and diluted
with DCM (29.9 kg). The resulting solution was washed with
water (28.1 kg), saturated aqueous NaHCO3 (43.7 kg), and
water (15.8 kg). The organic phase was concentrated to 27 L
by distilling off the solvent at 40−50 °C. Toluene (48.2 kg) was
added, and the batch was concentrated again by distillation
(final volume 27 L). Another portion of toluene (108 kg) was
added. Sampling for IPC showed DCM < 0.5% w/w,
concentration of 2 = 9.5−11.5% w/w, ee >85%. At 20−25
°C, HOAc (2.23 kg, 37.14 mol) was added. The resulting
suspension was stirred for 1 h at 20 °C and for 1 h at 0 °C. The
product was isolated by centrifugation and rinsed with toluene
(19.8 kg). The damp product was deliquored under a stream of
N2 and dried under reduced pressure at 40 °C until the toluene
content was below 1.8% w/w. Overall yield of 2·HOAc: 87%,
(2) (a) Chrzanowska, M.; Rozwadowska, M. D. Chem. Rev. 2004,
104, 3341−3370. (b) Xie, J.-H.; Zhu, S.-F.; Zhou, Q.-L. Chem. Rev.
2011, 111, 1713−1760. (c) Nugent, T. C.; El-Shazly, M. Adv. Synth.
Catal. 2010, 352, 753−819.
(3) Uematsu, N.; Fujii, A.; Hashiguchi, S.; Ikariya, T.; Noyori, R. J.
Am. Chem. Soc. 1996, 118, 4916−4917.
(4) (a) Mao, J.; Baker, D. C. Org. Lett. 1999, 1, 841−843. (b) Vedejs,
E.; Trapencieris, P.; Suna, E. J. Org. Chem. 1999, 64, 6724−6729.
(c) Tietze, L. F.; Zhou, Y.; Topken, E. Eur. J. Org. Chem. 2000, 2247−
̈
2252. (d) Tietze, L. F.; Rackelmann, N.; Sekar, G. Angew. Chem., Int.
Ed. 2003, 42, 4254−4257. (e) Tietze, L. F.; Rackelmann, N.; Muller, I.
̈
Chem.Eur. J. 2004, 10, 2722−2731. (f) Mujahidin, D.; Doye, S. Eur.
J. Org. Chem. 2005, 2689−2693. (g) Roszkowski, P.; Maurin, J. K.;
Czarnocki, Z. Tetrahedron: Asymmetry 2006, 17, 1415−1419. (h) Li,
L.; Wu, J.; Wang, F.; Liao, J.; Zhang, H.; Lian, C.; Zhu, J.; Deng, J.
Green Chem. 2007, 9, 23−25. (i) Canivet, J.; Suss-Fink, G. Green Chem.
̈
2007, 9, 391−397. (j) Werner, F.; Blank, N.; Opatz, T. Eur. J. Org.
Chem. 2007, 3911−3915. (k) Czarnocki, S. J.; Wojtasiewicz, K.;
́
́
Jozwiak, A. P.; Maurin, J. K.; Czarnocki, Z.; Drabowicz, J. Tetrahedron
2008, 64, 3176−3182. (l) Martins, J. E. D.; Clarkson, G. J.; Wills, M.
Org. Lett. 2009, 11, 847−850. (m) Martins, J. E. D.; Redondo, M. A.;
Wills, M. Tetrahedron: Asymmetry 2010, 21, 2258−2264.
(5) (2S,3S)-2,3-Dihydroxy-N-o-tolylsuccinic acid.
(6) (a) Erbeck, S.; Koberstein, R.; Soi, A.; Zistler, A. WO2010/
064212A, 2010. (b) Later, a small-scale protocol for a similar recycling
with trichloroisocyanuric acid as an oxidation reagent was reported.
See: Bolchi, C.; Pallavicini, M.; Fumagalli, L.; Straniero, V.; Valoti, E.
Org. Process Res. Dev. 2013, 17, 432−437.
1
99.7% ee. H and 13C NMR data corresponded to those in the
literature.1a
(7) (a) Willoughby, C. A.; Buchwald, S. L. J. Am. Chem. Soc. 1994,
116, 11703−11714. (b) Morimoto, T.; Suzuki, N.; Achiwa, K.
Tetrahedron: Asymmetry 1998, 9, 183−187. (c) Guiu, E.; Claver, C.;
ASSOCIATED CONTENT
■
́
Benet-Buchholz, J.; Castillon, S. Tetrahedron: Asymmetry 2004, 15,
S
* Supporting Information
3365−3373. (d) Jackson, M.; Lennon, I. C. Tetrahedron Lett. 2007, 48,
1831−1834. (e) Li, C.; Xiao, J. J. Am. Chem. Soc. 2008, 130, 13208−
13209. (f) Chang, M.; Li, W.; Zhang, X. Angew. Chem., Int. Ed. 2011,
50, 10679−10681. (g) Ruzic, M.; Pecavar, A.; Prudic, D.; Kralj, D.;
Scriban, C.; Zanotti-Gerosa, A. Org. Process Res. Dev. 2012, 16, 1293−
1300. (h) Berhal, F.; Wu, Z.; Zhang, Z.; Ayad, T.; Ratovelomanana-
Vidal, V. Org. Lett. 2012, 14, 3308−3311. (i) Xie, J.-H.; Yan, P.-C.;
Zhang, Q.-Q.; Yuan, K.-X.; Zhou, Q.-L. ACS Catal. 2012, 2, 561−564.
(8) For reviews of the use of MonoPhos ligands in asymmetric
hydrogenation, see: (a) van den Berg, M.; Feringa, B. L.; Minnaard, A.
J. Enantioselective Alkene Hydrogenation: Monodentate Ligands. In
Analytical methods and the procedure for the catalyst
screening. This material is available free of charge via the
Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Authors
Laurent.lefort@dsm.com
stefan.abele@actelion.com
1538
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539

Organic Process Research & Development
Article
The Handbook of Homogeneous Hydrogenation; de Vries, J. G., Elsevier,
C. J., Eds.; Wiley-VCH: Weinheim, Germany, 2007; Vol. 2. p 995.
(b) de Vries, J. G. Asymmetric Olefin Hydrogenation Using
Monodentate BINOL- and Bisphenol-Based Ligands: Phosphonites,
Phosphites, and Phosphoramidites. In Handbook of Chiral Chemicals;
Ager, D. J., Ed.; CRC Press: Boca Raton, FL, 2005; p 269. For a
recent applied example, see: (c) Lefort, L.; Boogers, J. A. F.; Kuilman,
T.; Vijn, R. J.; Janssen, J.; Straatman, H.; de Vries, J. G.; de Vries, A. H.
M. Org. Process Res. Dev. 2010, 14, 568−573.
(9) For example, see: Busscher, G. F.; Lefort, L.; Cremers, J. G. O.;
Mottinelli, M.; Wiertz, R. W.; de Lange, B.; Okamura, Y.; Yusa, Y.;
Matsumura, K.; Shimizu, H.; de Vries, J. G.; de Vries, A. H. M.
Tetrahedron: Asymmetry 2010, 21, 1709−1714.
(10) http://www.biotage.com/DynPage.aspx?id=23682 (accessed
Aug 20, 2013).
(11) We have indeed observed that 3,3′-substitution on the
binaphthol backbone is very important in the case of Ir catalysts. A
possible explanation is that it prevents the rapid deactivation of these
catalyst by dimerisation. See: Giacomina, F.; Meetsma, A.; Panella, L.;
Lefort, L.; de Vries, A. H. M.; de Vries, J. G. Angew. Chem., Int. Ed.
2007, 46, 1497−1500.
(12) The beneficial effect of acids or acylating agents as product
scavengers has been demonstrated in imine hydrogenation. For
example, see: Hou, G.; Tao, R.; Sun, Y.; Zhang, X.; Gosselin, F. J. Am.
Chem. Soc. 2010, 132, 2124−2125.
(13) Robotic equipment: Zinsser Lissy liquid handling robot placed
inside a glovebox, see http://www.zinsserna.com/lissy_liquidhandling.
htm. Parallel hydrogenation autoclave: Premex A96 reactor, see
http://www.premex-reactorag.ch/index.php?page=464 (accessed Aug
20, 2013).
(14) Blaser, H. U.; Buser, H. P.; Coers, K.; Hanreich, R.; Jalett, H. P.;
Jelsch, E.; Pugin, B.; Schneider, H. D.; Spindler, F.; Wegmann, A.
Chimia 1999, 53, 275−280.
(15) The role of I2 in Ir-catalyzed hydrogenation was first identified
in 1990. See: Chan, Y. N. C.; Osborn, J. A. J. Am. Chem. Soc. 1990,
112, 9400−9401.
(16) (a) Morimoto, T.; Achiwa, K. Tetrahedron: Asymmetry 1995, 6,
2661−2664. (b) Morimoto, T.; Suzuki, N.; Achiwa, K. Heterocycles
1996, 43, 2557−2560. (c) Zhu, G.; Zhang, X. Tetrahedron: Asymmetry
1998, 9, 2415−2418.
(17) Heller, D.; de Vries, A. H. M.; de Vries, J. G. Catalyst Inhibition
and Deactivation in Homogeneous Hydrogenation. In The Handbook
of Homogeneous Hydrogenation; de Vries, J. G., Elsevier, C. J., Eds.;
Wiley-VCH: Weinheim, Germany, 2007; Vol. 3, p 1483.
(18) During the screening, TaniaPhos T002-1 was used, but it led to
the undesired enantiomer. After the initial discovery, the opposite
hand of the ligand (T002-2) was used.
(19) De Vries, A. H. M.; Domin, D.; Helms, M.; Imboden, C.;
Kobersteinn, R.; Nazir, Z.; Skranc, W.; Stanek, M.; Tschebull, W.;
Verzijl, G. K. M. WO 2009/083903, 2009.
(20) Protonation of the imine was suspected to be mandatory for the
reaction. See: Blackmond, D. G.; Ropic, M.; Stefinovic, M. Org. Process
Res. Dev. 2006, 10, 457−463.
(21) Wu, X.; Li, X.; King, F.; Xiao, J. Angew. Chem., Int. Ed. 2005, 44,
3407−3411.
(22) Blacker, A. J. Enantioselective Transfer Hydrogenation. In The
Handbook of Homogeneous Hydrogenation; de Vries, J. G., Elsevier, C. J.,
Eds.; Wiley-VCH: Weinheim, Germany, 2007; Vol. 3, p 1215.
(23) The development of almorexant was abandoned in 2011.
(24) (a) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.;
Manley, J. B. Org. Process Res. Dev. 2011, 15, 912−917. (b) Abele, S.;
Inauen, R.; Spielvogel, D.; Moessner, C. J. Org. Chem. 2012, 77, 4765−
4773.
(25) To be published elsewhere.
1539
dx.doi.org/10.1021/op400268f | Org. Process Res. Dev. 2013, 17, 1531−1539