Reaction development and mechanistic study of a ruthenium catalyzed intramolecular asymmetric reductive amination en route to the dual orexin inhibitor suvorexant (MK-4305)

Article abstract of DOI:10.1021/ja202358f

The first example of an intramolecular asymmetric reductive amination of a dialkyl ketone with an aliphatic amine has been developed for the synthesis of Suvorexant (MK-4305), a potent dual Orexin antagonist under development for the treatment of sleep di

Full text of DOI:10.1021/ja202358f

ARTICLE
pubs.acs.org/JACS
Reaction Development and Mechanistic Study of a Ruthenium Catalyzed
Intramolecular Asymmetric Reductive Amination en Route to the
Dual Orexin Inhibitor Suvorexant (MK-4305)
,
†
‡
†
‡
†
†
Neil A. Strotman,* Carl A. Baxter, Karel M. J. Brands, Ed Cleator, Shane W. Krska, Robert A. Reamer,
†
§
Debra J. Wallace, and Timothy J. Wright
†
Department of Process Chemistry, Merck Research Laboratories, Rahway, New Jersey 07065, United States
‡
Department of Process Chemistry, Merck Sharp and Dohme Research Laboratories, Hertford Road, Hoddesdon, Hertfordshire,
EN11 9BU, U.K.
§
Department of Chemical Engineering, Merck and Company, Rahway, New Jersey 07065, United States
S Supporting Information
b
ABSTRACT: The first example of an intramolecular asym-
metric reductive amination of a dialkyl ketone with an aliphatic
amine has been developed for the synthesis of Suvorexant (MK-
4
305), a potent dual Orexin antagonist under development for
the treatment of sleep disorders. This challenging transforma-
tion is mediated by a novel Ru-based transfer hydrogenation
catalyst that provides the desired diazepane ring in 97% yield
and 94.5% ee. Mechanistic studies have revealed that CO2,
produced as a necessary byproduct of this transfer hydrogena-
tion reaction, has pronounced effects on the efficiency of the Ru
catalyst, the form of the amine product, and the kinetics of the transformation. A simple kinetic model explains how product
inhibition by CO leads to overall first-order kinetics, but yields an apparent zero-order dependence on initial substrate
2
concentration. The deleterious effects of CO on reaction rates and product isolation can be overcome by purging CO from
2
2
the system. Moreover, the rate of ketone hydrogenation can be greatly accelerated by purging of CO or trapping with nucleophilic
2
secondary amines.
’
INTRODUCTION
Merck & Co. recently disclosed the evaluation of the novel
dual Orexin antagonist 1 (Suvorexant, MK-4305) as a potential
1
treatment for insomnia in phase III clinical trials (Figure 1). The
2
first multikilogram synthesis of 1 was recently reported. How-
ever, preparation of the chiral diazepane ring moiety was
achieved through a racemic reductive amination and classical
resolution sequence to afford a 27% overall yield from 2-MSA to
Figure 1. Dual Orexin antagonist MK-4305 (1).
(
R)-3 (Scheme 1). It was projected that performing the reductive
amination in an asymmetric fashion would remove a step from
the synthesis and greatly improve the overall yield and produc-
tivity of the synthesis.
of the species involved in the catalytic cycle and uncovered the
profound effect of CO on the equilibrium levels of the active
catalyst and rate of reaction.
2
The racemic process alone faces a series of challenges: the
requirement for preferential reduction of imine over acyclic
ketone, a challenging 7-membered ring formation, and the need
for intra- versus intermolecular reaction. In addition to these
concerns, an asymmetric process presents the added difficulty of
’ DEVELOPMENT OF ASYMMETRIC REDUCTIVE
AMINATION
A variety of different approaches were considered for the
transformation of 2-MSA (MSA: methanesulfonic acid) to (R)-3
providing chiral induction at an imine showing little steric
3
differentiation (Me vs CH CH NR ) (Scheme 2). Despite
2
2
2
(via 4) including direct hydrogenation (H ) and transfer
2
these challenges, we report the first asymmetric reductive
amination of a dialkyl ketone with an alkyl amine. In addition,
mechanistic studies have provided insight toward the structures
Received: March 15, 2011
Published: May 02, 2011
r 2011 American Chemical Society
8362
dx.doi.org/10.1021/ja202358f
|J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
Scheme 1. Racemic and Asymmetric Reductive Amination Routes
Scheme 2. Asymmetric Reductive Amination of 2-MSA
hydrogenation (HCO H or 2-propanol (IPA)), which are both
enantioselectivity could be highly influenced through changes to the
2
13
known to give faster reduction of imines than ketones
ligand architecture. An evaluation of 30 ligand analogues based on
both DPEN and trans-1,2-diaminocyclohexane (DACH) scaffolds
(Figure 2) showed some promising product enantioselectivities
(Table 1, entries 1ꢀ7). Although the parent TsDPEN ligand gave
essentially racemic product (entry 1), the product enantioselectivity
was improved significantly by using more sterically bulky ligands
(entries 3 and 4), with the 2,4,6-triisopropylphenylsulfonyl-DPEN
ligand (TIPPS-DPEN, 9) proving to be the most effective. A
notably higher ee was obtained in MeCN than in IPA, THF, 5:2
4
(
Scheme 2). An extensive screen of direct hydrogenation
conditions involving Rh, Ir, Ru, and Pd precatalysts with ∼200
chiral phosphine ligands resulted in up to 72% ee for (R)-3
5
,6
(
Scheme 2). Other approaches involving a chiral phosphoric
3 7
acid catalyst or a chiral borohydride provided the product in
significantly lower enantioselectivities (15 and 40% ee, respec-
tively; Scheme 2).
Given these moderate enantioselectivities, we turned our
8
attention to transfer hydrogenation conditions using variants
formic acid/Et N, or in a variety of other solvents.
3
of Noyori’s (S,S)-RuCl(p-cymene)(TsDPEN) (5) (TsDPEN =
N-tosyl-1,2-diphenylethylenediamine) catalyst, which have proven
Both the enantioselectivity and yield were improved dramatically
14
when no effort was made to preform imine 4 (entry 8). Changing
9
4
6
useful for reduction of both ketones and imines. While there are a
few successful examples of the use of this catalyst for asymmetric
the η arene portion of the catalyst to benzene or hexamethylben-
zene led to decreased selectivity (entries 9 and 10). Moving to
the isoelectronic pentamethylcyclopentadienyl (Cp*) rhodium or
iridium analogues (entries 11 and 12) failed to produce any
10
reductive amination of alkyl-aryl ketones (acetophenones), only
very poor selectivities have been achieved for dialkyl ketones or even
11,12
15
alkyl-aryl ketones in an intramolecular sense.
Studies on transfer
improvement. Reducing the temperature from 35 to 0 °C for
hydrogenation of ketones to produce alcohols involving modifica-
the reaction in MeCN not only afforded the expected increase
tions to this TsDPEN ligand suggested that the reactivity and
in enantioselectivity, but also led to a large increase in yield
8
363
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
Figure 2. Select transfer hydrogenation ligands employed in this study.
Table 1. Optimization of Asymmetric Reductive Amination Using Transfer Hydrogenation Catalysts
entry
ligand
M
arene
solvent
MeCN
temp (°C)
HCO
2
H (equiv)
Et
3
N (equiv)
other base (equiv)
ee (%)
assay yield (%)
a
1
6
7
8
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
9
Ru
Ru
Ru
Ru
Ru
Ru
Ru
Ru
Ru
Ru
Rh
Ir
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
benzene
35
35
35
35
35
35
35
35
35
35
35
35
35
0
5
5
5
5
5
5
2
2
NaO
NaO
NaO
NaO
NaO
NaO
NaO
2
2
2
2
2
2
2
CH
CH
CH
CH
CH
CH
CH
19
13
46
52
39
42
21
70
30
9
45
31
21
19
31
28
39
54
33
42
11
38
53
72
59
a
2
MeCN
MeCN
a
3
2
a
4
MeCN
2
a
5
THF
2
a
6
IPA
2
a
7
none
100
5
40
4
8
9
MeCN
MeCN
5
4
10
11
12
13
14
15
16
17
18
19
20
Me
6
C
6
MeCN
5
4
Cp*
Cp*
MeCN
5
4
45
37
71
80
80
84
86
90
94
94.5
MeCN
5
4
Ru
Ru
Ru
Ru
Ru
Ru
Ru
Ru
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
p-cymene
DCM
5
4
MeCN
5
4
DCM
0
5
4
b
MeCN
0
5
0
K
2
K
3
K
2
K
2
K
2
CO
3
(4)
(4)
(4)
(4)
(4)
82
c
MeCN
0
5
0
PO
CO
CO
CO
4
54
MeCN
0
3.5
2.5
2.5
2.5
3
3
3
88
d
d
DCM
ꢀ5
0
2.5
2.5
>96
MeCN/PhMe
CH (2 equiv) and MgSO
>97
a
b
c
Aged substrate at 40 °C with NaO
yield at 72% conversion. Combined substrate with 4 equiv K
2
4
for 2 h prior to start of reaction to form the imine. 82% yield at 93% conversion. 54%
d
2
CO
3
in solvent and 0.1 vol of water to carry out deprotonation, and filtered before use.
(
entry 14 vs 8). Although a similar increase in enantioselectivity was
achieved for this process using only 3 mol % catalyst (entry 18). It
observed in CH Cl upon lowering the temperature, the yield
was eventually found that the beneficial role of inorganic base was to
2
2
16
increase was minimal (entries 13 and 15). When examining the
effect of other bases on this process, it was discovered that while
inorganic bases led to slower reaction rates, employing K CO or
precipitate the mesylate counterion as KOMs. By first deprotonat-
ing 2-MSA with K CO and then filtering, this reaction performed
2
3
exceptionally well in DCM, providing the product in 94% ee and
nearly quantitative yield (entry 19). When these asymmetric
reductive amination conditions were implemented on a larger scale,
3 was formed in 94% ee and >97% yield (Scheme 3). The amine
could be isolated in a single operation as its HCl salt 3-HCl in 87%
2
3
K PO did lead to an increase in ee (entries 16 and 17).
3
4
Further optimization of this system revealed that a mixture of
K CO and Et N provided both high product selectivities and high
reaction rates. By simultaneously optimizing the concentrations of
substrate, HCO H, Et N, and K CO , 90% ee and 88% yield was
2
3
3
17
overall yield and affording >99.5% ee. Alternatively, CH Cl could
2
3
2
3
2
2
8
364
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
Scheme 3. Asymmetric Reductive Amination
Figure 3. Possible forms of the free-base of 2-MSA.
Scheme 4. Equilibrium between Ketone and Imines form of
2
Free Base
be replaced with a 1:1 mixture of MeCN/toluene, giving the same
18
high selectivity and yield (entry 20). The level of enantioselectivity
obtained in this reaction is remarkable considering that this is the
only example of asymmetric reductive amination of a dialkylketone
with an aliphatic amine.
At this point, the asymmetric reductive amination reaction was
examined in more detail to determine if any further improvements in
productivity (higher rate, lower catalyst loading, lower solvent volume,
etc.) could be made. Until now, the form in which the free-base of 2-
MSA existed had not been studied and the relative contributions of
ketone, imine, and aminal were not known (Figure 3). While RPLC-
MS showed exclusively the mass of the ring-opened ketone form (2)
Figure 4. [4] vs time at various initial concentrations of 4.
Since even 1 equiv of water proved immiscible in CD Cl , the
2
2
equilibrium was studied in a 1:1 mixture of CD CN/toluene-d . The
3
8
equilibrium lies almost completely toward the imine isomer when no
water is added (Scheme 4). Even with the addition of 5 equiv of water,
(or aminal), it was hypothesized that imine 4, which should be the
actual species undergoing transfer hydrogenation, was likely the major
species under the nearly anhydrous reaction conditions. More
importantly, if the imine were not the major form, the transfer
hydrogenation could be greatly accelerated if conditions were altered
to favor formation of this species.
the equilibrium still greatly favors the imine isomer with a K of 2.7. It
eq
is interesting to note that, as in CD Cl , the corresponding aminal
2
2
species was not observed. Addition of triethylamine or formic acid did
not affect the position of the equilibrium.
1
H NMR spectroscopy in CD Cl showed the presence of
2
2
primarily 4, with only 5ꢀ10% of 2 and no signals for the correspond-
’
KINETICS OF REDUCTIVE AMINATION
ing aminal. Although the NMR signals for compounds 4 and 2
13
showed similar coupling patterns, C NMR spectroscopy clearly
The kinetic behavior of the reaction network was studied as
showed the ketone carbonyl in the minor species (2) (δ 207.3 ppm).
part of an effort to reduce the catalyst loading for the asymmetric
8
365
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
imagine how 3 could inhibit this reaction since other secondary
amines had proven essentially inert. Another possibility was
reversibility of the transfer hydrogenation reaction. The fact that
the enantioselectivity of 3 did not erode upon extended reaction
times, and that subjecting enantiopure (R)-3 to the reaction
conditions did not produce any of the undesired enantiomer
(
(S)-3) did not support this. Another possibility was a depen-
dence of the rate on the concentration of formic acid, but a series
of experiments showed that this was not the case. The only
remaining hypothesis was that the CO being produced through-
2
out the reaction was causing the inhibition, although it is
generally believed that CO extrusion by transfer hydrogena-
2
9b
tion catalysts is irreversible. To test this hypothesis, the
transfer hydrogenation reaction was conducted under standard
conditions with addition of CO : a >10-fold decrease in rate
2
was observed. This observation of CO inhibition prompted
2
us to fully explore the origin of this previously unreported
phenomenon.
’
STUDY OF CATALYTIC SPECIES
t 0
Figure 5. Ln([4] /[4] ) vs time at various initial concentrations of 4.
Transfer hydrogenation using Noyori’s RuCl(p-cymene)-
(
TsDPEN) catalyst is believed to proceed through a pathway
wherein an 18-electron Ruꢀhydride undergoes either concerted
Table 2. Observed Rate Constants (kobs) and Initial Rates As
a Function of Initial Substrate Concentration at 0 °C
23
proton and hydride addition to a ketone or hydride transfer to
24,25
an imine or iminium species
to produce an alcohol or amine,
ꢀ
1
1ꢀ
[
4]
0
(M)
k
obs = (ꢀd[4]/dt)/[4]
0
(s
)
0
initial rate = kobs[4] (M s )
respectively, and generate an unsaturated 16-electron Ru species
4,9,22
ꢀ
ꢀ
ꢀ
5
ꢀ6
ꢀ6
ꢀ6
(Scheme 5).
With formic acid as the terminal hydrogen donor,
0
.1250
.0625
.03125
3.56 ( 0.26 ꢁ 10
7.31 ( 0.31 ꢁ 10
15.6 ( 0.43 ꢁ 10
4.45 ꢁ 10
4.57 ꢁ 10
4.88 ꢁ 10
5
5
the Ruꢀhydride is regenerated, producing CO as a byproduct.
2
0
0
1
A study was undertaken to investigate by H NMR spectros-
copy how rapidly the precursor 10 reacts with triethylamine and
formic acid to irreversibly generate the active Ruꢀhydride
catalyst. Within 2 min at ꢀ5 °C, precursor 10 was completely
consumed, indicating very rapid activation. However, two sur-
prising observations emerged from this study.
reductive amination step. An initial experiment under the
standard conditions (0.125 M of 4 in 1:1 MeCN/toluene)
provided a curve that appeared to be overall first-order
(
(
Figure 4) and yielded a linear plot of ln[4] versus time
First, despite the disappearance of 10, Ruꢀhydride was not
1
9ꢀ21
Figure 5).
Surprisingly, when the same experiment was
the only, or even the major species formed (Scheme 6). It was
performed with [4] at 0.0625 M, the observed rate constant
0
(
kobs = (ꢀd[4]/dt)/[4]), was twice that seen at the higher
Scheme 5. Mechanism of Transfer Hydrogenation Using
concentration rather than being identical, as expected (Table 2).
When the concentration of 4 was decreased further to 0.03125
M, the kobs again doubled, indicative of a zero-order dependence
on 4. This raised the question of how a reaction that has no
RuCl(p-cymene) (ArSO DPEN) Catalysts
2
apparent order in substrate (rate is independent of [4] ) could at
0
the same time appear to follow overall first-order kinetics.
Elegant kinetic experiments performed by Wills and co-work-
ers using tethered derivatives of 5 have demonstrated that
transfer hydrogenation may initially be zero-order in substrate,
when substrate concentration is high and catalyst regeneration is
rate-limiting, but become first-order in substrate as the reaction
22
progresses and substrate reduction becomes rate-limiting.
Depending on the exact catalyst derivative that Wills employed,
kinetics varied widely between exhibiting complete first-order
behavior and nearly complete zero-order behavior. Our kinetic
curves are dominated by first-order kinetics, presumably due to a
first-order dependence on [4], which is emphasized in the log
plots. However, a first-order dependence on [4] should render
log plots with the same slope regardless of the initial concentra-
tion of 4, contrary to what we observed.
Scheme 6. Conversion of RuꢀChloride to RuꢀHydride via
RuꢀFormate
The observed reaction kinetics are strongly suggestive of
product inhibition, where the reaction rate decreases as the
product concentration increases. However, it was difficult to
8
366
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
Figure 6. Diastereomers of RuH(p-cymene)(ArSO -DPEN).
2
shown that a 95:5 mixture of a species identified as the
Ruꢀformate complex and the anticipated Ruꢀhydride complex
was formed. The assignment of the Ruꢀformate species was
made based on the integration of a high frequency formate
Figure 7. Equilibrium between Ruꢀformate and Ruꢀhydride.
ꢀ
þ
proton (separated from a resonance for HCO2 HNEt3 ) and
the TIPPS-DPEN and p-cymene ligands. A thorough search of
the literature revealed that this type of species had only
been reported once before by Ikariya, who had monitored the
rate of decomposition of Ruꢀformate(p-cymene)(TsDPEN) to
Ru. While both the TIPPS and Ts substituted Ruꢀhydrides favor
the (R)-configuration at Ru, it is interesting that such a high level
of the (S)-isomer is present, especially with the TIPPS substi-
tuent. Since the rate of isomerization of RuH(p-cymene)-
(TIPPS-DPEN) is so high even at ꢀ5 °C, we were unable to
determine if both diastereomers are competent catalysts for
transfer hydrogenation, or if both give similar product enantios-
electivities or reaction rates.
2
6
Ruꢀhydride. Longer age times (>3 h) of the mixture at ꢀ5 °C
did not lead to additional conversion from the Ruꢀformate to
the Ruꢀhydride species. This was surprising since conversion of
Ruꢀformate to Ruꢀhydride is a necessary step in the transfer
hydrogenation of 4, but was occurring at a significantly slower
rate than the catalytic reaction proceeds. The only explanation
that we could see for this behavior was the existence of an
equilibrium between the Ruꢀformate and Ruꢀhydride species,
The potential equilibrium between Ruꢀformate and Ruꢀ
hydride species was studied with the RuH(p-cymene)(TsDPEN)
catalysts ratherthantheTIPPSanalogue becauseitprovidedNMR
spectra that were much clearer and easier to interpret. RuH(p-
cymene)(TsDPEN) was generated from the Ruꢀchloride com-
27
10a
which will be discussed below.
plex using Noyori’s method with KOH in IPA and a solution of
The second surprising observation was that the RuH(p-
cymene)(TIPPS-DPEN) was not a single species as anticipated,
but instead existed as a 3:1 ratio of two species showing hydride
resonances at ꢀ5.63 and ꢀ5.62 ppm, respectively. After 30 min
at ꢀ5 °C, this ratio had completely switched to 1:2, where it
remained. These signals result from the two diastereomers of the
catalyst, due to chirality at Ru (Figure 6). A NOESY experiment
showed an NOE from the Ru-H at ꢀ5.63 ppm to the CHNH2
methine proton, which can only be achieved in the isomer
bearing an (S)-configuration at Ru. Likewise, an NOE was seen
the resulting Ruꢀhydride in CD Cl was exposed to CO .
2
2
2
Resonances were immediately seen for a Ruꢀformate complex,
including a diagnostic high frequency resonance for O CH, and
2
complete disappearance of the Ruꢀhydride resonances. This
result provided definitive evidence for the reversibility of
Ruꢀformate conversion to Ruꢀhydride since CO was the only
2
30
possible source of formate.
Similar to experiments with RuCl(p-cymene)(TIPPS-DPEN), a
0.1 M solution of RuCl(p-cymene)(TsDPEN) was treated with
formic acid and triethylamine, and a ∼95:5 mixture of Ruꢀformate
to Ruꢀhydride was observed. However, as the concentration of the
Ruꢀchloride solution was lowered to concentrations which are
more similar to those used in the actual asymmetric reductive
amination reaction (0.00375 M; equivalent to [Ru] at 3 mol %
catalyst), the proportion of Ruꢀhydride increased greatly
(Figure 7). This shift toward the right in the equilibrium can be
from the Ru-H at ꢀ5.62 to the CHNSO methine proton,
2
consistent with a structure having an (R)-configuration at Ru.
Noyori reported, based on X-ray crystallographic analysis, that
the major diastereomer in RuH(p-cymene)(TsDPEN) contains
the (R)-configuration at Ru (<1% minor diastereomer by NMR
2
8
spectroscopy). More recently Wills demonstrated that tethered
variants of 5 can exist as mixtures of diastereomers and that this
ratio of diastereomers may change over the course of a reaction.
explained by a concomitant decrease in [CO ] as the total
concentration of Ru is decreased. These equilibrium ratios provide
2
29
Preparation of RuH(p-cymene)(TsDPEN) by treatment of
RuCl(p-cymene)(TsDPEN) with formic acid and triethylamine
was also examined by H NMR spectroscopy. Initially, a 65:1
a rough estimate of the equilibrium constant of ∼0.005 M, although
there is significant error in this estimate since the CO that is
2
1
generated partitions between the solvent and the headspace of the
isomer ratio was observed, consistent with Noyori’s observations.
As expected, the major isomer (Ruꢀhydride resonance at ꢀ5.83
ppm) showed an NOE from the hydride proton to the CHNSO2
proton, consistent with the (R)-configuration at Ru. Interest-
ingly, after 2 h at room temperature, the ratio had changed to
NMR tube. When the 75:25 mixture of Ruꢀhydride/Ruꢀformate
was treated with CO , all NMR resonances associated with
2
the Ruꢀhydride disappeared to give the Ruꢀformate complex as
31
the only Ru species, providing further evidence for an equilibrium
process.
∼
8:1 due a growth of the resonance corresponding to the minor
These dramatic effects of CO on the proportion of Ruꢀ
2
isomer at ꢀ6.13 ppm. This isomerization between the two
diastereomers may occur through the 16-electron unsaturated
intermediate, which is generated after each hydride transfer
event, and which can add hydrogen from formic acid to either
catalyst face generating both diastereomers. Diastereomer inter-
conversion could also be achieved through partial dissociation of
hydride present in the reaction mixture help to explain the kinetic
observations involving CO inhibition. However, the possibility that
2
CO could interact with the substrate by forming a carbamate with
2
2, lowering the concentration of 4 and hence leading to a slower
reaction rate, was also considered. No more than a trace of this
carbamate species (<2%) was observed when a solution of 2/4 in
CD Cl was charged with CO . The product of the transfer
the ArSO -DPEN ligand and coordination to the opposite face of
2
2
2
2
8
367
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
Scheme 7. Equilibrium between 3 and 11
hydrogenation (3), on the other hand, readily formed carbamate 11
when CO was added, with the ratio of 11:3 increasing with higher
2
CO pressure and lower temperature (Scheme 7). An HMBC
2
experiment conducted on 11 demonstrated a correlation between
the carbamate carbon (δ 160.9 ppm) and the three protons geminal
to the carbamate nitrogen, providing strong evidence for this
structural assignment, and differentiating it from a protonated form
of 3.
’
FURTHER KINETIC STUDIES
vF si gt ui mr ee .8. Concentrations of 4, 3, 11, and Ruꢀhydride ([RuH] ꢁ 50)
The kinetics of the transfer hydrogenation reaction were
further investigated by NMR spectroscopy. When the transfer
hydrogenation reaction was carried out at 0 °C, 3, 4, 11, and the
Scheme 8. Catalytic Cycle for Transfer Hydrogenation Using
Formic Acid
32
Ruꢀhydride could be monitored (Figure 8).
The
[
Ruꢀhydride] reached a maximum at ∼1 h, and then slowly
decreased to ∼40% of this maximum value over the course of the
reaction. Throughout the reaction, both the [3] and the [11]
increased, maintaining a ratio of ∼1.2:1 favoring carbamate 11. It
is interesting to note that the reaction of 3 with CO to form 11
2
reduces the amount of CO in the system, attenuating the
2
deleterious effect of CO on the transfer hydrogenation reaction.
2
Disappearance of 4 followed approximate first-order kinetics
ꢀ
5
ꢀ1
with kobs = 8.81 ( 0.16 ꢁ 10
s , showing a 2.5-fold increase
33
from earlier experiments run at ꢀ5 °C.
Starting from the differential rate expression (eq 1) derived from
the catalytic cycle (Scheme 8), and employing a steady-state approx-
imation for the Ruꢀhydride species (eq 2), eq 3 can be derived. Since
[
Ruꢀhydride] had been measured throughout the reaction, eq 1
ꢀ ꢀ1
2 2
1
could be solved for k . A value of k = 0.22 ( 0.02 M
s
was
34
obtained. Rate constant k can be obtained from eq 3, which, at high
1
[
4] (early in the reaction), may be simplified to ꢀd[4]/dt = k [Ru] ,
ꢀ ꢀ1 ꢀ1
1
1
0
1
giving k = 0.0025 s . Finally, k = 0.18 M
from the equilibrium expression K = k /k which relates
Ruꢀhydride and CO with Ruꢀformate.
s
was obtained
ꢀ1
eq
1
ꢀ1
35
2
However, eq 3 still contains [CO ], and therefore, an analy-
2
tical solution expressed only in terms of [4] cannot be offered. An
accurate representation of [CO ] in terms of [4] must include its
2
generation as the reaction progresses, the equilibrium with amine
3
to give carbamate 14, and partitioning between the solution
Equation 5 clearly explains the dichotomous behavior of this
kinetic system, wherein overall rates are unaffected by changing
36
and gas phases in the reaction vessel. However, a simple
approximation is [CO ] = [4] ꢀ [4], where an equimolar
[4] , but a given reaction exhibits a first-order decay of 4. When
2
0
0
amount of CO is generated as imine 4 is consumed. Substituting
[4] was changed from 0.125 to 0.0625 M to 0.03125 M
2
0
this expression into eq 3 yields eq 4. Since the term k 1[4] is
(Table 2), both [4] and [4] changed proportionally, resulting
ꢀ
0
0
much larger than (k ꢀ k )[4] þ k (due to similar values of
in no net effect on reaction rate (ꢀd[4]/dt), and corresponding
2
ꢀ1
ꢀ1
1
rate constants k and k and the small magnitude of k (vide
supra)), this expression further simplifies to eq 5. Substituting the
to a 2- and 4-fold increase in k = (ꢀd[4]/dt)/[4]. In contrast,
2
1
obs
throughout the course of a reaction, the denominator in eq 5 (or
eq 3) remains constant, while [4] in the numerator follows a first-
order decay, leading to the observed first-order kinetics. A
physical interpretation of this kinetic data is that even though
the reaction of Ruꢀhydride with imine 4 is fast compared to the
conversion of Ruꢀformate to Ruꢀhydride, the generation of
rate constants that we obtained into eq 5 yields a composite of
these rate constants (k = k k [Ru] /(k [4] )). This pro-
obs
ꢀ1
1 2
0
ꢀ1
0
ꢀ
5
duces kobs = 9.89 ꢁ 10
s
, which can be compared with k
=
obs
ꢀ
5
ꢀ1
8
.81 ꢁ 10
s
obtained from a curve fit of [4] versus time
Figure 8), suggesting that the highly simplified eq 5 represents a
useful model for this reaction.
(
3
7,38
CO as a byproduct inhibits the reaction as it progresses, pushing
2
8
368
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
the equilibrium away from Ruꢀhydride, and yielding a decrease
in rate and an apparent first-order dependence on 4.
Scheme 9. Transfer Hydrogenation of Acetophenone
ꢀ
d½4ꢂ
¼
k2½RuHꢂ½4ꢂ
ð1Þ
dt
ꢀ
d½RuHꢂ
dt
¼
¼
0
k ½RuO CHꢂ ꢀ k ½RuHꢂ½CO ꢂ ꢀ k ½RuHꢂ½4ꢂ
1
2
ꢀ1
2
2
ð2Þ
ꢀ
d½4ꢂ
k1k2½Ruꢂ ½4ꢂ
0
¼
ð3Þ
dt
kꢀ1½CO2ꢂ þ k2½4ꢂ þ k1
ꢀ
d½4ꢂ
k1k2½Ruꢂ ½4ꢂ
0
ꢃ
ð4Þ
ð5Þ
dt
kꢀ1½4ꢂ þ ðk2 ꢀ kꢀ1Þ½4ꢂ þ k1
0
ꢀ
d½4ꢂ k1k2½Ruꢂ ½4ꢂ
0
ꢃ
dt
kꢀ1½4ꢂ0
’
REMOVAL OF CO FROM REACTION MIXTURE
2
The above mechanistic investigations suggested that purging
of CO during an asymmetric reductive amination in acetoni-
2
Figure 9. Conversion of acetophenone vs time as a function of amine.
trile/toluene on larger scale would accelerate the reaction rate.
Side-by-side experiments were conducted in Multimax vessels
equipped with ReactIR probes to directly compare the effects of
nitrogen purging with no purging. Even though an IR probe
could not effectively distinguish between compounds 4 and 3, it
’
SUMMARY
The first asymmetric reductive amination of a dialkyl ketone
with an alkylamine was developed using a new variant of
Noyori’s (S,S)-RuCl(p-cymene)(ArSO DPEN) catalyst. This
ꢀ1
clearly showed a strong band for CO at 2339 cm . The vessel
2
2
undergoing the nitrogen sweep showed <5% of the [CO ] found
process proceeds with high yield (>97%) and high enantios-
electivity (>94%) to form a diazepane ring in the key step of the
synthesis of a novel dual Orexin inhibitor and has been run on
>100 kg scale. Despite the requirements of low temperature,
high pH, high dilution, and bulky catalyst needed to provide
high yield and ee, a reduction of the catalyst loading to 2 mol %
proved feasible. In the course of this work, the profound impact
of CO2 on the equilibrium between Ruꢀhydride and
Ruꢀformate catalyst forms and the corresponding negative
2
in the sealed control vessel. This decrease in [CO ] was
2
accompanied by a ∼60% increase in the reaction rate, allowing
the loading of catalyst 10 to be decreased to 2 mol %. Addition-
ally, isolated yields were always greater in purged reactions (by up
to 20%) since there was no appreciable formation of carbamates
and subsequent loss into basic aqueous layers during work-up.
To investigate whether rate acceleration by CO removal is a
2
general phenomenon, the transfer hydrogenation of acetophe-
none with this same catalyst system was examined. The effects of
purging should be even more pronounced in a ketone reduction
reaction, since there are no amines produced to absorb some of
the CO . An alternative to CO purging could be accomplished
effects that CO had on reaction rate were studied. The
2
potential of formation of carbamates of the product with CO2
was noted as this can reduce the overall isolated yield. Incor-
porating a purge of CO into the process led to an increase in
the rate of reaction and allowed for product isolation with a
nearly quantitative yield. Finally, significant rate increases in the
transfer hydrogenation of acetophenone (either via purging
of the reaction mixture, or trapping with secondary amines)
were realized, suggesting that this may be a generally useful
technique.
2
2
2
by trapping with a secondary amine, similar to the carbamate
formation in the reductive amination step. Thus, the effects of
purging and trapping of CO on the transfer hydrogenation rates
2
of acetophenone (Scheme 9) were examined when using 10 mol
%
Ru under conditions that were otherwise identical to those
used for the reductive amination.
Addition of piperidine and, particularly, pyrrolidine proved
effective at trapping CO , resulting in rate enhancements of
2
’
EXPERIMENTAL SECTION
∼
2.5- and 3.7-fold, respectively, versus reactions using triethyla-
39,40
mine (Figure 9).
Purging CO proved even more effective,
Asymmetric Reductive Amination in CH Cl and Upgrade
2
2
2
resulting in a remarkable 9.5-fold increase in rate over a sealed
reaction using triethylamine. In all cases, the transfer hydrogena-
tion reaction of acetophenone gave 98.3ꢀ98.9% ee.
via HCl Salt (R)-3. Catalyst Preparation. Dichloromethane (68 mL),
ligand (S,S)-9 (300 mg), [RuCl (p-cymene)] (191 mg), and triethy-
lamine (1.29 g) were combined to give an orange slurry which was
2
2
8
369
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
heated to 40 °C for 3 h leading to an orange/red solution which was
cooled to room temperature.
Fujii, A.; Hashiguchi, S.; Ikariya, T.; Noyori, R. J. Am. Chem. Soc. 1996,
118, 4916–4917.
(5) For examples of asymmetric imine hydrogenation or reduc-
In a second vessel, dichloromethane (200 mL) was charged and
degassed via nitrogen pressure purges. Bis MSA salt 2-MSA (10 g) and
anhydrous potassium carbonate powder (325 mesh; 11.6 g) were
added followed by water (1.16 mL). The vessel contents were then
cooled to ꢀ10 to ꢀ5 °C. Triethylamine (5.34 g) was charged to the
mixture followed by the preformed catalyst solution. The solution was
degassed by nitrogen purge and then formic acid (1.28 g) was added.
This was followed by a second formic acid charge of 2.60 g after 30 min.
The reaction temperature was maintained at ꢀ10 to ꢀ5 °C and the
mixture aged for 16 h.
tive amination using H , see: (a) Blaser, H.-U.; Buser, H. P.; Jalett,
2
H. P.; Pugin, B.; Spindler, F. Synlett 1999, 867. (b) Kadyrov, R.;
Reirmeier, T. H. Angew. Chem., Int. Ed 2003, 42, 5472–5474. (c)
Kadyrov, R.; Riermeier, T. H.; Dingerdissen, W.; Tararov, V.; B €o rner,
A. J. Org. Chem. 2003, 68, 4067–4070. (d) Chi, Y.; Zhou, Y.-G.;
Zhang, X. J. Org. Chem. 2003, 4120–4122. (e) B €o rner, A. Synlett
2
005, 203–211.
6) Noyori’s preformed Ru catalysts of the form RuCl (diamine)-
(
2
(
bis-phosphine) exhibited low yields (<30%) and ee’s (<20%).
7) Yamada, K.; Takeda, M.; Wakuma, I. Tetrahedron Lett.
981, 3869.
8) Attempts to carry out transfer hydrogenation using [RuCl (p-
(
The reaction was quenched by addition of NaOH solution (123 g,
1
1
.0 M, aq), the layers were allowed to separate, and the lower organic
(
2
layer was collected. The organic layer was then concentrated, then
toluene (68 mL) and DMAc (34 mL) were added followed by
anhydrous HCl (2.26 g). The salt was cooled to 0 °C over 3 h, and
the solids were isolated by filtration The solids were dried at 40 °C to
afford 5.48 g of amine HCl salt (R)-3-HCl, 87% yield, >99.9% purity
2
cymene)] and amino alcohol ligands in IPA with KOH resulted in none
of the desired reductive amination product and only some ketone
reduction.
(9) RuCl(p-cymene)(Ts-DPEN) was developed for transfer hydro-
genation of ketones using IPA as the hydrogen donor. It was later shown
that IPA could be replaced by formic acid to give an irreversible reaction
1
and 99.7% ee. 3 (free base): H NMR (500 MHz, CD Cl ) δ 7.22 (d,
2
2
through generation of CO . (a) Hashiguchi, S.; Fujii, A.; Takehara, J.;
J = 2.1 Hz, 1H), 7.146 (d, J = 8.2 Hz, 1H), 6.93 (dd, J = 8.4, 2.6 Hz, 1H),
2
Ikariya, T.; Noyori, R. J. Am. Chem. Soc. 1995, 117, 7562–7563. (b) Fujii,
A.; Hashiguchi, S.; Uematsu, N.; Ikariya, T.; Noyori, R. J. Am. Chem. Soc.
3
3
.90 (ddd, J = 14.5, 6.1, 4.1 Hz, 1H), 3.80 (dt, J = 13.7, 3.3 Hz, 1H),
.68ꢀ3.56 (m, 2H), 3.20 (dt, J = 14.2, 3.5 Hz, 1H), 2.92 (ddd, J = 14.2
1
996, 118, 2521–2522.
Hz, 11.0, 3.5 Hz, 1H), 2.77 (dqd, J = 9.9, 6.4, 3.3 Hz, 1H), 1.94 (ddt, J =
1
3
(10) (a) Noyori, R.; Hashiguchi, S. Acc. Chem. Res. 1997, 30, 97–102.
1
4.3, 6.0, 3.2 Hz, 1H), 1.59ꢀ1.46 (m, 2H), 1.10 (d, J = 6.4 Hz, 3H);
NMR (126 MHz, CD Cl ) δ 164.10, 148.30, 146.06, 129.39, 119.97,
16.09, 109.55, 55.76, 51.68, 48.13, 46.59, 38.40, 23.58; IR (cm ):
C
(b) Palmer, , M. J.; Wills, M. Tetrahedron Asym. 1999, 10, 2045–2061.
2
2
(11) Wills found that RuCl(p-cymene)(TsDPEN) gave nearly ra-
ꢀ
1
1
1
2
cemic products in similar intramolecular reductive amination reactions
to give 6 or 7-membered rings, and gave very low yields of 7-membered
ring products. Williams, G. D.; Pike, R. A.; Wade, C. E.; Wills, M. Org.
Lett. 2003, 5, 4227–4230.
þ
þ
249, 1571, 1639; LCMS: [M þ H] calcd for C H ClN O þ H ,
1
3
16
3
66.1; found 265.9, 267.9 (30%).
’
ASSOCIATED CONTENT
Supporting Information. Experimental procedures, char-
S
b
acterization data, and copies of NMR spectra. This material is
available free of charge via the Internet at http://pubs.acs.org.
(
12) Wills has shown that certain monocyclic imines can be hydro-
’
AUTHOR INFORMATION
genated with up to 60% ee using N-alkylated variants of 5. Martins,
J. E. D.; Redonde, M. A. C.; Wills, M. Tetrahedron: Asymmetry 2010,
Corresponding Author
neil_strotman@merck.com
2
1, 2258–2264.
13) Mohar, B.; Valleix, A.; Desmurs, J.-R.; Felemez, M.; Wagner, A.;
Mioskowski, C. Chem. Commun. 2001, 2572–2573.
14) We initially attempted to preform imine 4 by treatment with
(
(
’
ACKNOWLEDGMENT
sodium formate and MgSO
respectively.
4
to deprotonate 2-MSA and remove water,
We would like to thank Mike Ashwood, Brian Bishop, John
(15) Xiao and co-workers reported highly enantioselective imine
Edwards, David Lieberman, and Faye Sheen (Department of
Process Research, Hoddesdon, U.K.), and Paul Fernandez, Tony
Moses, Marguerite Mohan, Doris Glykys and Robert Scogna
hydrogenation and reductive amination of acetophenone derivatives
using a Cp*IrX complex with 9 using a chiral phosphoric acid additive.
Additionally, they reported reductive amination of unbranched dialkyl
ketones using a Cp*IrX Me C SO -DPEN complex and a chiral
(Department of Chemical Process Development and Commer-
5
6
2
cialization, Rahway, NJ) for experimental input.
phosphoric acid that proceeded with high ee. This system was not
applicable to aliphatic amines and the addition of chiral phosphoric
acids to our process led to no improvement in rate or ee. (a) Li, C.;
Wang, C.; Villa-Marcos, B.; Xiao, J. J. Am. Chem. Soc. 2008,
130, 14450–14451. (b) Li, C.; Villa-Marcos, B.; Xiao, J. J. Am. Chem.
Soc. 2009, 131, 6967–6969.
’
REFERENCES
(1) Cox, C. D.; J. Med. Chem. 2010, 53, 5320–5332.
(2) Baxter, C. A.; Cleator, D.; Brands, K. M. J.; Edwards, J. S.;
Reamer, R. A.; Sheen, F. J.; Stewart, G. W.; Strotman, N. A.; Wallace,
D. J. Org. Process. Res. Dev. 2011, 15, 367–375.
3
(16) In experiments with only Et N as base, we found that when
mesylate ion is present in solution as ammonium salts, the reaction
proceeds quickly but with lower enatioselectivity. When mesylate is
removed from solution by precipitation as the potassium or cesium salt
(3) The one exception to the ubiquitous low selectivity for reduction
of unbranched dialkylimines was from the work of MacMillan et al.
which demonstrated asymmetric reductive amination of 2-butanone
with 4-methoxyaniline using a chiral phosphoric acid derivative. Storer,
R. I.; Carrera, D. E.; Ni, Y.; MacMillan, D. W. C. J. Am. Chem. Soc. 2006,
1
(as determined by H NMR), the reaction is slower, but proceeds with
high ee.
(17) During formation and manipulation of the DBT (dibenzo-
yltartrate) salt of 3, up to 20% of an impurity had formed that was
determined to be isomer a, which could be formed through bicyclic
5.6-ring species b. While weak acids like pivalic acid readily catalyzed this
1
28, 84–86.
4) It has been shown that transfer hydrogenation proceeds >1000
times faster for imines than for structurally similar ketones: Uematsu, N.;
(
8
370
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371

Journal of the American Chemical Society
ARTICLE
process, stronger acids led to minimal isomerization (see Supporting
Information for effect of acid pK ; Table S1).
a
was estimated using data between 200 and 600 min, since in this region
the value of [Ruꢀhydride] was not changing drastically, and there was a
relatively small error in the rate measurement compared to late in the
reaction.
(
35) The equilibrium constant was determined by measuring
1 13
[
Ruꢀhydride] and [Ruꢀformate] by H NMR and [CO
2
] by
C
NMR (relative to the CD CN/toluene-d solvent mixture) in a sealed
3
8
NMR tube at 0 °C using the following relationship: K = k /k
=
eq
1
ꢀ1
[
Ruꢀhydride][CO
2
]/[Ruꢀformate] = 0.014 M.
(
36) [CO ] = [4]
2 g
2
0
ꢀ [4] ꢀ [11] ꢀ [CO = [3]
2
]
g
t
ꢀ [CO
2 g
] , where
[
CO ] = moles of CO in gas phase divided by volume of reaction
(
18) A tethered variant of 5, [(s,s)-teth-TsDPEN-RuCl], originally
2
mixture. This does not take into account the minor impact of the
Ruꢀhydride/Ruꢀformate equilibrium on [CO ].
37) See Supporting Information (Figure S4) for overlaid plots of
experimental and calculated [4] vs time ([4] calculated using kobs
[Ru] /(kꢀ1[4] )).
38) A major source of error in this rate law is the use of the steady-
state approximation for [Ruꢀhydride]. It is clear from Figure 8 that
developed by Wills, provided low conversion and 63% ee under
otherwise identical conditions. Hayes, A. M.; Morris, D. J.; Clarkson,
G. J.; Wills, M. J. Am. Chem. Soc. 2005, 127, 7318–7319.
2
(
=
k
1
k
2
0
0
(
[
Ruꢀhydride] changes considerably over the course of the reaction, as
the initially formed Ruꢀformate is converted to Ruꢀhydride to estab-
(19) [HCO
2 0 0
H] = [TEA] = 0.3125 M in all reactions.
lish equilibrium and then after reaching a maximum, Ruꢀhydride begins
(20) Since the product (3) of asymmetric reductive amination
to decrease as CO is produced and the equilibrium is shifted toward
2
generally precipitated late in the reaction as the formate salt, kinetics
were generally studied by setting up a series of identical reactions for one
run and quenching them at various times. This was the only way that we
could accurately determine the ee of the product as a function of
conversion/time.
Ruꢀformate.
(39) Relative rates assuming first-order kinetics in [acetophenone].
2
(40) Amine 3 appears to be a highly effective trapping agent for CO .
When 1 equiv of acetophenone was added to the standard asymmetric
reductive amination reaction of 4, the acetophenone was hydrogenated
4 times faster than acetophenone in a similar reaction without 4 (29% vs
8% conversion). The high nucleophilicity of this diazepane ring in 3 is
consistent with the relative order of nucleophilicity of secondary amines:
piperidine < pyrrolidine < piperazine < homopiperidine. Brotzel, F.;
Chu, Y. C.; Mayr, H. J. Org. Chem. 2007, 72, 3679–3688.
ꢀkt
(
21) Rates determined by fitting [4] = [4]
0
e
.
(22) Cheung, F. K.; Lin, C.; Minissi, F.; Criville, A. L.; Graham,
M. A.; Fox, D. J.; Wills, M. Org. Lett. 2007, 9, 4659–4662.
23) Casey and Johnson provided evidence for the concerted nature
(
of the proton and hydride transfer to ketones through deuterium kinetic
isotope effect studies. Casey, C. P.; Johnson, J. B. J. Org. Chem. 2003,
68, 1998–2001.
(24) It was shown by B €a ckvall that the presence of acid was critical
for activity toward imine hydrogenation with this catalyst system,
suggesting that an iminium species, rather than an imine is actually
being reduced Åberg, J. B.; Samec, J. S. M.; B €a ckvall, J.-E. Chem.
Commun. 2006, 2771–2773.
(25) Both Wills and Pihko have suggested an ionic transition state
involving attack of the Ruꢀhydride moiety on a protonated imine or
iminium ion. (a) Martins, J. E. D.; Clarkson, C. J.; Wills, M. Org. Lett.
2
009, 11, 847–850. (b) Evanno, L.; Ormala, J.; Pihko, P. M. Chem.—Eur.
J. 2009, 15, 12963–12967.
(
26) Koike, T.; Ikariya, T. Adv. Synth. Catal. 2004, 346, 37–41.
(27) For examples of equilibria between metal hydrides and CO2
with formates see the following, and references therein: Creutz, C.;
Chou, M. H. J. Am. Chem. Soc. 2009, 131, 2794–2795.
(
28) Haack, K.-J.; Hashiguchi, S.; Fujii, A.; Ikariya, T.; Noyori, R.
Angew. Chem., Int. Ed. Engl. 1997, 36, 285–288.
29) Cheung, F. K.; Clarke, A. J.; Clarkson, G. J.; Fox, D. J.; Graham,
M. A.; Lin, C.; Crivill ꢀe , A. L.; Wills, M. Dalton Trans. 2010,
9, 1395–1402.
30) Since the Ruꢀhydride is a saturated 18-electron species, its
reaction with CO most likely does not involve CO coordination to Ru
since this would require either partial ligand dissociation or an η to η
ring slip, both high energy processes. A more likely explanation is that
CO hydrogen bonds to the NH of the catalyst, adds hydride to
(
3
(
2
2
6
4
2
2
generate formate, and since formate is an insufficiently strong base to
deprotonate the Ru-coordinated NH moiety (unlike alkoxide, from
2
ketone reduction), the formate then coordinates to Ru.
(
31) See Supporting Information (Figure S1).
(
32) No precipitation of 3 as the formate salt was observed in an
NMR tube.
ꢀkt
(
33) Rate determined by fitting [4] = [4]
0
e
(See Supporting
Information, Figure S2 for curve fit).
(
34) See Supporting Information (Figure S3) for a plot of k
2
vs time,
where k = (ꢀd[4]/dt)/([RuH]*[4]) = kobs/[RuH]. This value for k2
2
8
371
dx.doi.org/10.1021/ja202358f |J. Am. Chem. Soc. 2011, 133, 8362–8371