Full text of DOI:10.1021/op200174k

ARTICLE
pubs.acs.org/OPRD
Use of an Iridium-Catalyzed Redox-Neutral Alcohol-Amine Coupling
on Kilogram Scale for the Synthesis of a GlyT1 Inhibitor
Martin A. Berliner,*,† Stꢀephane P. A. Dubant,‡ Teresa Makowski,† Karl Ng,† Barbara Sitter,† Carrie Wager,†
and Yinsheng Zhang†
†Chemical Research and Development, Pharmaceutical Sciences, Pfizer Global Research and Development, Groton Laboratories,
558 Eastern Point Rd, Groton, Connecticut 06340, United States
‡Chemical Research and Development, Pharmaceutical Sciences, Pfizer Global Research and Development, Sandwich Laboratories,
Ramsgate Road, Sandwich, Kent CT13 9NJ, United Kingdom
S Supporting Information
b
ABSTRACT: A recent development for the efficient and environmentally friendly synthesis of aliphatic amines is the transition-
metal-catalyzed redox-neutral coupling of an alcohol and an amine, generally referred to as a “borrowing hydrogen” reaction. In this
work, we describe the first kilogram-scale application of this technology in the synthesis of PF-03463275, a GlyT1 inhibitor
developed for the treatment of schizophrenia. Using (Cp*IrCl2)2 the reaction has been optimized to achieve catalyst loadings lower
than 0.05 mol % iridium (S/C g 2000) while retaining reasonable reaction times (<24 h). Water and a tertiary amine are essential
for high catalytic activity, resulting in dramatically increased reaction rates compared to existing literature protocols. Methods
for iridium removal are also described.
’ INTRODUCTION
oxidation of 4 provides aldehyde 5, which is utilized without
purification in a reductive amination with 3-fluoro-4-chloroben-
zylamine (6) to generate secondary amine 7. Installation of the
amide in 9 is accomplished by coupling with N-methylimidazole-
4-carboxylic acid (8) under typical conditions. After column
chromatography, the Boc group in 9 is removed with hydrogen
chloride to generate pyrrolidine 10. In the final step, the N-methyl
group on the pyrrolidine is introduced using a reductive amina-
tion with formaldehyde. Following cleanup by silica gel chroma-
tography, 1 is rendered crystalline by a reslurry in ethyl acetate.
Interestingly, all bond-forming steps in the synthesis of 1 from 2,
6, and 8 generate new carbonÀnitrogen bonds, and no carbonÀ
carbon bonds are formed.5
The primary drivers for the development of a new route for the
synthesis of 1 focused on operational issues. Most intermediates
(3, 4, 5, 7, and 9) are oils, and their salts, while solids, proved to
be hygroscopic and unsuitable for handling and storage. We also
sought to avoid formaldehyde use in the final step while eliminating
the need for protecting groups. We were mindful of the potential
difficulties engendered by the presence of unprotected amines in
the API and precursors. In practice, this latter concern proved
to be the major driver for the selection of chemistry during
process development, as many common transformations uti-
lize exogenous amine bases that must be removed during work-
up. Given this constraint, a step-reordered route to 1 was
designed to address many of the existing concerns in the
discovery synthesis while incorporating our desired changes
(Scheme 2).6 This modified route is shorter and employs
formaldehyde earlier in the process in a telescoped debenzyla-
tion/reductive amination reaction to eliminate the need for
The construction of carbonÀnitrogen bonds is central to the
preparation of bioactive organic molecules. Significant chemical
literature is associated with heterocycle synthesis and amine
acylation, and more recent and substantial efforts have resulted
in development of efficient methods for the transition-metal-
catalyzed construction of arylÀnitrogen bonds.1 In contrast, the
transition-metal-catalyzed synthesis of aliphatic CÀN bonds has
received comparatively less attention.2 These reactions typically
proceed via a series of discrete steps involving sequential de-
hydrogenation of the alcohol (to form a carbonyl compound)
followed by imine formation and subsequent reduction (Figure 1).
As a result of the action of the catalyst as a hydride shuttle be-
tween starting material and product, these processes are typically
described in the literature as “borrowing hydrogen” or “hydrogen
autotransfer” reactions to reflect the redox-neutral nature of the
overall process.
From a process development perspective, application of this
chemistry offers several potential advantages over existing tech-
niques (e.g., direct alkylation, reductive amination). In addition
to a significant improvement in atom economy (water is the only
byproduct of the reaction), the redox-neutral process features
operational simplicity and reduced environmental impact through
solvent and waste stream reduction. In this report, we describe
the optimization and kilogram-scale application of an iridium-
catalyzed redox-neutral coupling of an alcohol and a substituted
benzylamine as applied to the synthesis of PF-03463275 (1), a
glycine transporter type 1 (GlyT1) inhibitor that has potential as
a therapy for schizophrenia.3
The seven-step medicinal chemistry synthesis of PF-
03463275 is shown in Scheme 1.3a The synthesis starts from
exo-N-benzyl-3-azabicyclo[3.1.0]hexane-6-methanol (2),4 which
is converted to the N-Boc alcohol 4 via amino alcohol 3. Swern
Received: June 28, 2011
Published: August 13, 2011
r
2011 American Chemical Society
1052
dx.doi.org/10.1021/op200174k Org. Process Res. Dev. 2011, 15, 1052–1062
|

Organic Process Research & Development
ARTICLE
formaldehyde quantification in the API. A new penultimate
intermediate, secondary amine 12, was envisioned to function
as an amine donor for acylation of carboxylic acid 8. This route
also introduces the pivotal aliphatic carbonÀnitrogen bond
construction between alcohol 11 and amine 6. However, it is
important to note that when we committed to this route we
had not identified the exact method we wanted to employ for
this coupling step.
’ INITIAL PROCESS DEVELOPMENT
Synthesis of Alcohol 11. Debenzylation of 2 is the first step of
the medicinal chemistry synthesis and scaled easily without sig-
nificant modification. The product, amino alcohol 3, participates
in a facile reductive amination with formaldehyde under a hy-
drogen atmosphere with Pd(OH)2/C as catalyst.7 We determined
that the two transformations could be conducted sequentially but
not simultaneously in the same vessel, as the presence of formalde-
hyde suppresses the N-debenzylation reaction.8 In the optimized
process (Scheme 3), after debenzylation is complete (16 h,
25 °C, 50 psig H2) a slight excess of aqueous formaldehyde
solution (37% w/w) is introduced. Hydrogen is reintroduced to
the reactor, and uptake ceases after 4À8 h, at which time 11 is the
only observed product. Aqueous dimethylamine is then added
and hydrogenation continued. This final step serves to consume
any unreacted formaldehyde9 and results in formation of tri-
methylamine, which is easily removed by distillation during work-
up. Water, methanol, and other volatiles are removed by azeo-
tropic distillation with toluene, and alcohol 11 is isolated as a
viscous oil by concentration under reduced pressure, typically
in yields of approximately 90% for the two steps. While not
needing purification before the next step, vacuum distillation of
11 does result in improved color and shelf life. This sequence was
Figure 1. Redox-neutral (borrowing hydrogen, hydrogen autotransfer)
aminations.
Scheme 1. Medicinal Chemistry Synthesis of PF-03463275
Scheme 2. Planned Process Route
1053
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
Scheme 3. Final Process Conditions, Synthesis of 11
Scheme 4. Two-Step Synthesis of 12
Figure 2. Impurities from the oxidationÀreduction process.
Scheme 5. Preparation of Labeled 6 and First Synthesis of
[M + 5]-12
’ DEVELOPMENT OF IR-CATALYZED REDOX-NEUTRAL
PROCESS
successfully executed in our kilo lab several times (up to 3 kg
scale) and on larger scale (10 kg) by an external vendor.
Implementation of a Two-Step Alcohol Amination. For the
first campaign, we chose to implement a two-step oxidation/
reductive amination sequence with aldehyde 13 (Scheme 4) as a
nonisolated intermediate due to concerns about its stability and
water solubility. Activated DMSO oxidations10 proved superior
and of the several variants available, the CoreyÀKim,11 ParikhÀ
Doering,12 and Swern13 oxidations were most robust for gram-
scale reactions.14 Of these, Swern and ParikhÀDoering processes
provided the best reaction impurity profiles in the subsequent
step, and on the basis of operational simplicity the ParikhÀ
Doering conditions were chosen for scale-up. Once the oxidation
was complete, methanol and amine 6 were added directly to the
reaction mixture. Imine formation was observed to be rapid by
1H NMR assay and in situ IR reaction monitoring.15 Imine 14
was then reduced to the amine by portionwise addition of solid
sodium borohydride.16 After a lengthy work-up,17 diamine 12
was crystallized as its nonhygroscopic bis(hydrochloride) salt by
addition of excess anhydrous HCl in 2-propanol. These crystal-
lization conditions purge most of the process impurities formed
in this step (Figure 2), which are primarily the bis-alkylation
byproduct 15 and N-methylated amines 16 and 17. The latter
two impurities are formed by reduction of methyleneiminium
species derived from reaction of DMSO-derived Pummerer
degradants18 with 6 and 12. The overall yield of 12 in this two-
step sequence is 30À45%. A new process was required, and
counterintuitively, we uncovered the answer while preparing a
small-scale batch of isotopically labeled 1.
Initial Application: Synthesis of Stable Isotope-Labeled 12.
Because of the presence of a single chlorine atom in 1, the
synthesis of a M + 5 isotopically labeled isomer was required to
use as an internal standard for quantitative bioanalytical studies
using mass spectrometry. Benzylamine 6 was identified as the
preferred site for labeling, and [M + 4]-6 (19) was readily pre-
pared from 1-iodo-3-chloro-4-fluorobenzonitrile via cyanation
with in situ prepared [13C/15N]CuCN19 followed by reduction
with nickel boride20 in d1-methanol (Scheme 5). Condensation
of 19 with alcohol 1 using the conditions described above
provided very poor yields of 20 due to the complexity of the
process when operating on milligram scale, even when 20 was
isolated by column chromatography after work-up.
In 2003, Fujita and Yamaguchi21 reported that (Cp*IrCl2)2
catalyzes the N-alkylation of anilines and primary benzylic amines
with primary alcohols at modest temperatures (90À100 °C) with
good selectivity for the formation of secondary amines. Direct
application of these conditions to the coupling of alcohol 11 and
amine 19 proved immediately successful (Scheme 6). Moreover,
the condensation was competent under microwave conditions,
proceeding to completion in minutes with no significant differ-
ence in yield. By pretreating both starting materials with D2O and
MeOD to fully exchange all N- and O-bound hydrogens for
deuterium and taking the resulting D2O-wet mixture into the
reaction, we discovered that an additional deuterium can be
introduced in the product. The deuteration occurs selectively on
the imine carbon and provides the requisite [M + 5] isomer 20 in
1054
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
Scheme 6. Preparation of 20 Using Cp*Ir Chemistry
conducted in the absence of water stalled and produced a yellow
inorganic precipitate. The use of 2À3% water on a volume basis
prevented this precipitation event and facilitated reactions pro-
ceeding to completion, typically in 4À5 h at 100 °C. Additionally,
excess 6 in the product mixture proved challenging to purge
during crystallization of 12. Attempts to identify an alternative
salt with better crystallization and purification properties failed,
so amine 6 was deliberately undercharged (0.9 equiv per equivalent
of alcohol 11) in order to facilitate product isolation and avoid
chromatographic purification of the API.25
We also looked at purging iridium from the reaction product.
Scavenging resins were ineffective at reducing iridium levels in
product-containing solutions.26 Using our existing crystallization
process, typical iridium content in isolated 12 bis(hydrochloride)
salt was approximately 1000 ppm. This level proved acceptable as
a 5- to 10-fold reduction of iridium levels was observed through
the remainder of the synthesis.
Scheme 7. Possible Mechanism for Deuterium
Incorporation
Unfortunately, this process proved only partially successful
when executed on scale (1.5 kg). We observed that the water
added to this reaction efficiently distilled as a toluene azeotrope
over the first hour of processing and was retained in the reactor
vapor trap.27 This unintentional dehydration (the vapor trap
essentially functioned as a DeanÀStark trap) led to the reaction
stalling at 70% conversion, and once stalled, return of the water to
the reactor did not rescue the process. After work-up, a mixture of
HCl salts of 12 and 6 were crystallized from 2-propanol in a 2:1
molar ratio, with an overall yield of 55% calculated on a mass
basis. Ineffective washing of the filter cake due to cake cracking
resulted in 6000 ppm Ir in isolated 12, which was higher than
expected on the basis of pilot experiments. Finally, the presence
of 6 necessitated a chromatographic cleanup of API manufac-
tured from this batch of 12 to meet clinical release specifications.
In preparation for a reload campaign, a series of experiments
were conducted to address the issues encountered during the first
scale-up of the redox-neutral chemistry. The effect of reactor
configuration was addressed by laboratory pilots that verified that
semicontinuous return of water to the reaction mixture did not
adversely impact the reaction rate and reduced the likelihood for
stalled reactions. A more directed effort was made to develop an
improved and more reliable iridium purge process by optimizing
the crystallization and isolation of 12 as its bis(hydrochoride)
salt. The most important factor affecting iridium content in 12
proved to be the protocol and volume of wash solvent of the
filter cake. A slight modification of our standard filtration pro-
tocol prevented cake cracking,28 and adding more wash solvent
(2-propanol) until the filtrate ran clear and colorless resulted in a
10-fold decrease in iridium levels in 12. As a side benefit, the
increased volume of cake wash also improved the purging of 6
such that it was routinely reduced to levels below 1% in
development pilots. Even with these processing changes, the
reaction on scale proceeded to only 85% conversion (measured
as a ratio of desired product 12 to amine 6) before stalling. After
work-up, crystallization according to the newly optimized con-
ditions provided a 55% yield of 12 bis(hydrochloride) as a white
crystalline solid in >98% purity, with <1% benzylamine 6 and 360
ppm iridium.
good yield and high isomeric purity. We propose that an equilibrium
process that facilitates rapid H-to-D exchange on the metal is
responsible for formation of a deuteroidirium(III) species (i.e., C),
which serves as the reductant (Scheme 7).22 The source of the
deuterium has not been conclusively proven, but likely possibi-
lities include D2O and/or a N-deutero-trialkylammonium spe-
cies. Notably, in the absence of added D2O, the reaction between
11 and 19 generates [M + 4]-20 and little of the desired [M + 5]
product. Finally, in the synthesis of 20 we observe no higher-
mass isomers, indicating that for this substrate under these
reaction conditions the alcohol dehydrogenation and amine reduc-
tion are essentially irreversible.
Process Optimization for First Kilogram-Scale Campaigns.
The need to resupply nonlabeled 1 led us to implement the Fujita
and Yamaguchi chemistry for the synthesis of 12. Our initial work
involved a series of directed screens to evaluate alternative catalysts
and identify critical process parameters. In addition to (Cp*IrCl2)2,
we looked at (Ru(p-cymene)Cl2)2/DPPF23 and Ru(COD)Cl2/
XantPhos24 with three different inorganic bases (NaHCO3,
K2CO3, and Na2CO3). Reactions using Ru-based catalysts stalled
with large amounts of imine present, while Ir-catalyzed reactions
proceeded to completion. Iridium-catalyzed reactions did not
proceed to any significant extent with less than 3% iridium
present and were more reliable when 5% Ir was employed. The
importance of water in the reaction was unexpected, as reactions
Robustness Optimization. For the next API campaign, our
primary goal was to improve the robustness of the Ir-catalyzed
coupling reaction. Small-scale use tests of 6 and 11 in threaded
vials went to completion but on 10 g scale in a jacketed reactor
under nitrogen the reaction stalled at 60% conversion. Additional
experiments in open vessels revealed the consistent nonoptimal
1055
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
Table 1. Summary of Ir-Catalyzed Reactions (5% Catalyst) in
Sealed Reaction Vesselsa
Table 2. Product Distribution from Catalyst Optimization
Experimentsa
composition (molar ratio)c
solvent reaction time
entry Ir (mol %)b (vol)
(h)
12
6
15
1
2
3
4
5
6
7
8
5
1.0
1.0
1.0
0.2
0.2
0.2
0.2
0.2
<5
<5
5
89
86
85
86
86
85
84
2
5
7
6
5
6
7
9
9
1
0.5
0.25
0.1
0.05
0.033
0
7
10
17
17
24
24
9
10
9
ratiob
8
100
entry scale (g)
reactor type
GL18
6 (equiv) 12
6
15 yield (%)c
a All reactions were conducted on 1 g scale in capped glass test tubes
(5, 1, 0.5 mol % Ir) or 10 g scale in a 100 mL 316SS Parr Reactor using
1 N aqueous NaHCO3 as base at 110 °C for the indicated time.
1
2
3
4
1
10
0.9
0.9
1
89
81
88
88
2
0
5
5
9
18
7
65
60
74
78
Parr
b Monomeric iridium. Measured by H NMR analysis of a sample of
c
1
10
Parr
the reaction mixure.
1400
8 L Parr (2 runs)
1
7
a All reactions were conducted with 0.025 equiv (Cp*IrCl2)2, 0.05 equiv
K2CO3, 5 vol toluene in a sealed reactor (GL18 threaded capped test
tube or 100 mL stirred 316SS Parr reactor) for 4À5 h at 100 °C.
b Reaction monitoring was via 1H NMR analysis of samples; ratios are
reported as molar ratios (see Supporting Information). All reactions
proceeded to completion (no detected 11). c Yield is of pure isolated 12
bis(hydrochloride) salt after work-up.
Table 3. Product Distribution from Temperature Optimiza-
tion Experiments
composition (molar ratio)
entry temp (°C) reaction time (h)
12
6
15
1
2
3
4
80
100
110
130
72
24
85
86
86
86
5
6
5
4
8
8
performance of the reaction, with rate variability and catalyst
inactivation observed even when obvious water was present in
the mixture. We note that in reviewing the available literature on
this reaction, including a recently published review by Fujita and
co-workers,29 it was apparent that method development was
conducted in sealed vessels without any investigation of the role
of water.
A defining experiment in this series serves to illustrate the
dramatic improvement in reaction performance observed in sealed
reactors. The previous optimal conditions for this process utilized
0.9 equiv amine, which was deliberately undercharged in order to
compensate for catalyst deactivation and facilitate isolation of 12
(Table 1, entry 1). When this reaction was run for the first time at
10 g scale in a Parr reactor, 6 and 8 were completely consumed,
forming a mixture of 12 and the tertiary amine 15 (ratio 81:18;
Table 1, entry 2) from which 12 bis(hydrochloride) was isolated
in 60% yield. Improved yields were realized by adjusting reagent
stoichiometry to 1:1 (entry 3). An increase in the amount of HCl
used to crystallize 12 improved crystallization robustness and
reduced the amount of Ir in the isolated salt by ca. 20%. On scale
(1.4 kg over 2 runs in a 8 L Hastelloy pressure reactor; entry 4)
this reaction went to completion (no detected alcohol 11) and
provided 12 in 78% yield, a dramatic improvement over the
previous optimized conditions.
17
10
9
<16
2. Catalyst Loading. Catalyst loading was optimized in a series
of experiments summarized in Table 2. On gram scale in capped
glass test tubes, reactions using 0.5 mol % Ir or greater were
complete in 5 h or less (entries 1À3). On this scale, lower catalyst
loadings (0.25 mol % Ir or lower) resulted in inconsistent results
due to catalyst inactivation. We ascribe this to the very small
amounts of catalyst utilized in these experiments (3 mg or less),
so screening was continued at 10 g scale. Again, the reaction
proved competent as the catalyst charge was reduced a further
order of magnitude (entries 4À7) to less than 0.05 mol % Ir
(S/C > 2000). Minor and insignificant changes in overall reaction
mixture composition accompany catalyst reduction (see Table 2).
As a control experiment, a mixture of alcohol 11 and amine 6
were combined with an aqueous solution of NaHCO3 and held in
the reactor for 24 h at 110 °C (entry 8) to verify that no reaction
occurred. Most notably, no degradation was observed, and upon
adding 0.025 mol % (Cp*IrCl2)2 (0.05 mol % Ir) and reheating,
the reaction proceeded to completion in 17 h (entry 6).
3. Reaction Temperature. Changing the reaction temperature
resulted in substantial differences in the rate of the reaction, but
no significant differences in reaction composition were observed
across the tested temperature range (Table 3).32
Efficiency Optimization. At the completion of this campaign,
we investigated several additional factors that were expected to
influence the reaction outcome.30
1. Base and Water. A brief survey of base (K2CO3 or
NaHCO3) and amount of added water (none to 5 equiv)
revealed no significant difference in reaction outcome across all
combinations, so the use of a saturated aqueous sodium bicar-
bonate solution (approximately 1 N concentration at 25 °C) was
standardized for all future experiments.31
4. Reaction Solvent. The performance of nonparticipating
solvents in the reaction was evaluated under robust conditions
(0.25 mol % Ir catalyst, 1 vol solvent). Toluene provided the best
result (Table 4, entry 1), with tert-butanol (entry 2), a mixture of
tert-butanol and toluene (entry 3), and isopropyl acetate (entry 4)
also giving acceptable results (Table 3). Notably, the use of water
1056
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
Table 4. Effect of Solvent on Reaction Performancea
Table 6. Low-Catalyst Scale-Up Runs
Ir used 12
6
15 11
yield
Ir
composition (molar ratio)
entry 11 (kg) (mol %) (%) (%) (%) (%) (12 2HCl) (ppm)
3
entry
solvent
toluene
12
6
15
11
1
2
2.4
2.4
0.0625 80.6 8.4 9.3 1.7 4.81 kg (76%) 9.2
0.0750 83.6 8.1 8.0 0.3 4.54 kg (69%) 6.9
1
86
81
83
84
76
75
75
64
61
41
22
16
6
9
12
11
8
2
t-BuOH
7
6
3
t-BuOH/tol (1:1)
i-PrOAc
Scheme 8. Synthesis of 1
4
8
5
DMAc
10
2
4
10
11
6
DMSO
12
13
10
34
7
water (no base)
water
13
16
5
8
10
9
CH3CN
neat
10
11
12
29
32
38
29
46
46
EtOAc
2-MeTHF
a All reactions were conducted on 1 g scale using 1 vol solvent,
0.25 mol % Ir catalyst, 0.5 mol % aqueous NaHCO3 base at 110 °C
for 20À24 h.
Table 5. Iridium in Isolated Samples of 12 2HCla
3
Ir in reaction
(mol %)
Ir in isolated 12 2HCl
Ir reduction
(fold)
3
entry
(ppm)
above (0.05 mol % Ir, aqueous NaHCO3 solution, 0.2 L/kg
toluene, 110 °C). Due to decreased solvent use, the reaction was
conducted on larger scale (2.4 kg 11) in a single run in a 8 L
Hastelloy pressure reactor. Reaction performance was slower on
scale; after 18 h the reaction was only 70% complete. Facility
scheduling constraints necessitated that we add an additional
0.0125 mol % iridium to drive the reaction to completion, which
was reached after 40 h (<2% 11 remaining). The rate discrepancy
between pilots and the scale-up was subsequently determined to
be due to an initial temperature overshoot in the laboratory
reactor that resulted in an accelerated initial reaction rate and
faster overall reaction performance. For the second campaign, a
slight initial increase in catalyst loading and reaction temperature
(to 115 °C) avoided the necessity of adding additional catalyst,
affording complete conversion (<0.5% 11 remaining) in 36 h.
1
2
3
4
5
5
300 (300, 360)a
140
68
0.25
0.1
37
16
8 (9, 7)a
62
0.05
0.033
62
4
80
a Numbers in parentheses represent valued obtained on batches run on
scale (>1 kg 11). Iridium quantitation by ICP/MS.
alone as solvent was inferior, although catalytic activity is
improved in the absence of base (entries 7, 8).33 Similar results
were obtained using polar aprotic solvents (entries 5, 6). Poor
catalyst activity was observed in ethyl acetate (entry 11) and
2-methyltetrahydrofuran (entry 12) and in the absence of solvent
(entry 10). With the exception of the reaction in ethyl acetate, in
which appreciable ester hydrolysis occurred, all reactions were
clean (i.e., no degradation products were apparent), highlighting
the wide range of functional group compatibility exhibited by this
catalytic system.
5. Work-Up and Isolation Modifications. By reducing catalyst
charge to less than 0.1 mol % Ir, the amount of carbonate base
added to the reaction was reduced to very low levels and did not
need to be removed via aqueous extraction prior to work-up.
Consequently, the work-up of this reaction was eliminated, and
the crude reaction mixture directly exchanged into 2-propanol
via azeotropic displacement of toluene at reduced pressure prior
to salt formation. Filter cake wash volumes were maintained as
they serve a dual role of reducing iridium and purging of 6 in the
product to below 1%. Iridium reduction ranged between 60- and
140-fold during crystallization (Table 5) and generated 12
bis(hydrochloride) salt that did not require additional iridium
mitigation in subsequent operations.
’ FINAL ACYLATION AND IRIDIUM CONTROL
Two variations of the API synthesis have been implemented
on scale. We initially conducted a two-step process involving
acylation of 12 with acid chloride 22, which was prepared from 8
under typical conditions (Scheme 8). For the acylation, dichlor-
omethane is a preferred solvent,34 and the reaction is rapid using
excess dimethylethylamine (DMEA) as base.35 This base was
chosen strictly on the basis of its low boiling point (38 °C) rather
than on any intrinsic difference in reactivity between DMEA
and more commonly used tertiary amine bases. As a result of
the basicity of the API, the work-up of this reaction required
full neutralization with aqueous base36 prior to a water wash
to prevent extraction of salts of 1 into the aqueous phase. If
necessary, an iridium scavenging operation was conducted at this
stage. Solvent displacement with ethyl acetate facilitated removal
of CH2Cl2 and DMEA, and after concentration of the mixture to
∼3 L/kg, 1 crystallized upon cooling as white needles. Heptane
was added as antisolvent prior to isolation to increase yield.
In some campaigns we observed small amounts of off-colored
Performance of Optimized Conditions on Scale. After these
efforts were complete, two additional campaigns were conducted
(Table 6). The first employed the optimized conditions described
1057
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
Table 7. Removal of Iridium from 1 Using Scavenging
Agentsa
Table 8. Effect of Added Tertiary Amine Base on Model
Reaction Performancea
Ir amount removal of Ir
main product
recovery (%)
entry
resin name
(ppm)
(%)
1
2
3
4
5
6
7
8
Ecosorb C908
86
65
82
86
78
87
79
87
83
83
93
91
90
92
93
95
91
90
Ecosorb C941
Norit CA1 (B)
Darco KBB (B)
Norit CA1
104
60
entry
additive (amount)
conversion (%)
99
1
none
20
CUNO SASS R55
Darco KB-B 278106
Biotage AC 2604-1
64
2
1 N NaHCO3 (0.25 mol %)
1 N NaOH (1 mol %)
20
83
3
0
81
4
30% w/w Bu4NOH in water (1 mol %)
N-methylmorpholine (1 equiv)
N-methylpiperidine (1 equiv)
N-methylpyrrolidine (1 equiv)
N-methylpyrrolidine (5 mol %)
1,4-dimethylpiperizine (5 mol %)
TMEDA (5 mol %)
0
a The batch of 1 used for this study had ∼475 ppm Ir. Iridium
5
92
quantitation by ICP/MS.
6
94
7
94
waxy solids in the filter cake; these proved to be a complexmixture
containing 1 and several other API-related components.37
A
8
95
repeat of the reaction work-up proved sufficient to purge these
materials and generate clinical grade API.
9
95
10
11
12
94
Acid chloride 22 demonstrates nonideal storage and handling
qualities, and its synthesis, while straightforward, employs DMF
with oxalyl chloride, which can form N,N-dimethylcarbamoyl
chloride (DMCC), a suspected human carcinogen.38 While not a
significant issue in this process due to the aqueous base work-up
after the acylation, implementation of a direct coupling protocol
between 12 and acid 8 promised to improve the overall efficiency
of the process and reduce overall solvent use. We elected to
implement a T3P-mediated process primarily because of the
water solubility of its degradants (propanephosphonic acid salts)
upon quenching and work-up with aqueous base. This reaction
also performed best in dichloromethane, and work-up and
product isolation were conducted as before, obtaining similar
overall yields and product quality.
Methods for Iridium Control. Prior to development of the
low-catalyst amine coupling reaction, we needed a way to control
iridium at the API stage to meet product quality guidelines. A
screen of several dozen scavenging agents and carbons quickly
revealed that few proved effective at iridium sequestration with-
out incurring significant loss of material (Table 7).39 For all
experiments, 100 wt % of scavenger was employed in organic
solvent for 14À24 h at ambient temperature. Darco KBB (entry 4)
was selected for use in subsequent scale-up efforts on the basis of
recovery efficiency and low cost. In practice, we observed typical
iridium reductions of ∼85% (for example, 360 ppm Ir in 12 was
reduced to 60 ppm Ir in crystallized API).
DABCO (5 mol %)
N-methylpyrrolidine (1 equiv)b
93
87 (65)c
a Except for entry 12, all reactions were conducted in sealed test tubes on
1 g (alcohol) scale at 115 °C for 16 h. Reactions with 1 equiv additive
were run without solvent; others employed 1 vol toluene. Conversion is
measured as the ratio of reaction product(s) to starting amine. b 10 g
scale with 0.05 mol % (Cp*IrCl2)2 at 130 °C for 40 h. The reaction
stalled at 87% conversion (TON = 870). See the Supporting Informa-
tion for analytical details. c Isolated yield of 23 monohydrochloride salt;
see the Experimental Section for the procedure.
N-ethylpyrrolidine were utilized due to formation of significant
amounts of N-ethylated byproducts from competitive amine
cross-coupling processes.40,41 In general, bases that do not read-
ily form enamines upon dehydrogenation appear to be preferred.
Inorganic carbonate bases seem to be unnecessary but not dele-
terious, but hydroxides kill catalytic activity (entries 2 and 3).42
A larger-scale run (entry 12) with stoichiometric N-methylpyr-
rolidine as solvent proceeded to partial conversion (catalyst
TON = 870, 65% isolated yield), indicating that further optimi-
zation of this reaction is required to achieve optimal (>2000
TONs) catalyst performance. These reactions appear to perform
better above the azeotropic boiling point of the solvent/water
mixture, and therefore the chemistry is better suited for sealed
vessels or for flow equipment where the temperature can be
significantly elevated above the boiling point. Water may also
play a role in facilitating catalyst turnover but we have not yet
quantified its influence. Finally, the distribution of products
(starting materials, the desired product and overalkylated products)
is dependent on starting material steric demands and stoichiom-
etry, is influenced by the reaction solvent and additives, and does
not appear to be overly influenced by reaction temperature.
Because of this, each substrate would need to be optimized in-
dividually.
’ DISCUSSION AND CONCLUSIONS
We have discovered that several factors are essential for ensuring
robust catalyst performance. The most important is that use of a
tertiary amine base appears to be necessary to realize improved
turnover. This effect was briefly explored on a model system by
evaluating the reaction between phenethylamine and 3-phenyl-1-
propanol, which has been reported to be quite sluggish under
standard conditions in the absence of any additives.33a Significant
improvement in overall reaction performance was obtained when
any one of several common tertiary amine bases were added in
substoichiometric (5 mol % as a solution in 1 vol toluene) or
stoichiometric (100 mol %; reaction solvent) quantities (Table 8).
Unsatisfactory results were obtained when triethylamine and
In conclusion, the optimized conditions for the iridum-
catalyzed coupling reaction between alcohol 11 and amine 6 as
described and implemented above represent a significant im-
provement over existing literature examples, in which a catalyst
loading of 1 mol % or greater is utilized. Our initial experience
optimizing and scaling this reaction was decidedly mixed due to
poor reproducibility and high catalyst loading. However, we were
1058
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
able to successfully address these issues in subsequent campaigns.
In doing so, we have identified critical process parameters that
allowed us to achieve a 20- to 100-fold decrease in catalyst loading
over the existing best-in-class procedures for the coupling of
similar compounds.43 While the role of water and the amine base
in this reaction have not been fully quantified, we have demon-
strated that their presence is critical to achieving efficient and
robust catalyst performance on both small and large scale. The
finding that the intermediate acidic iridium hydride species can
be utilized for isotopic enrichment of substrates through H-D
exchange with D2O was unexpected. The significant operational
and environmental advantages of this process over traditional
oxidation-reductive amination sequences should serve to encou-
rage chemists faced with similar bond-forming challenges to in-
vestigate this technology. Finally, experiments to further eluci-
date the reaction mechanism and develop generalized conditions
for high turnover solution-phase reactions in batch- and flow-
mode are in progress and will be reported in due course.
period the reactor contents were cooled and sampled; the re-
action was 70% complete. An additional charge of catalyst (1.42 g,
1.8 mmol, 0.0000625 equiv) and aqueous sodium bicarbonate
solution (3.5 mL) were added to the mixture, which was purged
and heated to 110 °C for 16 h. The contents were cooled and
sampled; the reaction was complete (<2% alcohol 11 remaining).44
The mixture was then transferred to a 100 L glass-lined reactor,
diluted with 7.2 L 2-propanol, and distilled under reduced
pressure (0.2 bar). When the pot volume reached ∼7 L, 2-propanol
(7.2 L each) was added, and distillation was continued, again
reducing the volume to ∼7 L. After a second dilution with
2-propanol (7.2 L), the mixture was concentrated to 7 L, cooled
to ambient temperature, and transferred slowly into a reactor
containing a preheated (50 °C) solution of 5 N HCl in
2-propanol (9.31 kg, 2.75 equiv). Once the freebase had been
added, the resulting slurry was cooled to ambient temperature at
10 °C/h and granulated for 4 h. The product was isolated by
filtration, and care was taken to stop the filtration immediately
before the cake was exposed to nitrogen. The filtercake was then
washed with 2-propanol (72 L), and the resulting product dried
to constant mass in a vacuum oven (50 °C, 0.05 bar). A total of
4.81 kg (14.1 mol, 76% yield) of 12 was isolated as a free-flowing,
nonhygroscopic white crystalline powder. Mp 232 °C. The
spectroscopic properties of the salt are nonideal, and therefore
a sample was freebased for characterization. 1H NMR (400 MHz,
CDCl3, freebase): δ 7.34 (dd, J = 2.1, JHF = 7.1 Hz, 1 H), 7.13
(ddd, J = 2.2, 8.3 Hz, JHF = 4.8 Hz, 1 H), 7.01 (app t, J = 8.8 Hz,
JHF = 8.8 Hz, 1 H), 3.71 (s, 2 H), 2.95 (d, J = 8.8 Hz, 2 H), 2.40 (d,
J = 6.8 Hz, 2 H), 2.25 (s, 3 H), 2.28 À 2.22 (m, 2H), 1.30 (tt, J =
3.3, 6.8 Hz, 1 H), 1.18 - 1.13 (m, 2 H). 13C NMR (100 MHz,
CDCl3, freebase): δ 157.0 (d, JCF = 247 Hz), 137.7 (d, JCF = 3.7
Hz), 130.0, 127.5 (d, JCF = 7.2 Hz), 120.7 (d, JCF = 17.5 Hz),
’ EXPERIMENTAL SECTION
((1R,5S,6r)-3-Methyl-3-azabicyclo[3.1.0]hexan-6-yl)methanol
(11). An inerted 30 L Hastelloy pressure reactor was charged with
Pd(OH)2 (20% Pd content, 50% wet, 275 g), methanol (20 L,
7.2 L/kg) and ((1R,5S,6r)-3-benzyl-3-azabicyclo[3.1.0]hexan-6-
yl)methanol (2, 2.75 kg, 13.5 mol, 1 equiv). After three purge/
vent cycles, the mixture was warmed to 30 °C, and the reactor
was pressurized with hydrogen gas (50 psig). When hydrogen
uptake ceased (approx 16 h), the reactor was purged, and a
solution of formaldehyde in water (37% w/w, 1.32 kg, 16.2 mol,
1.2 equiv) was charged in one portion. The reactor was purged
and repressurized with hydrogen gas at 50 psig, and the mixture
was held at 30 °C for 8 h. Upon cessation of hydrogen uptake
(8 h), the reactor was purged, and dimethylamine in water (40%
w/w, 0.61 kg, 5.4 mol, 0.4 equiv) was charged in one portion.
Hydrogenation was continued at 30 °C and 50 psig for another
4À8 h. The resulting reaction mixture was filtered to remove
catalyst, and the solvent was displaced from methanol into
toluene via a constant-volume distillation under reduced pres-
sure, removing methanol, water, dimethylamine, and trimethy-
lamine. The mixture was filtered to remove residual colloidal
palladium residues, and after additional concentration, the resulting
light yellow toluene solution of 11 was removed from the reactor.
A gravimetric assay of the solution indicated a total of 1.55 kg of
116.3 (d, JCF = 20.4 Hz), 57.1, 52.27, 51.5, 41.4, 22.3, 20.2. 19
F
NMR (376 MHz, CDCl3, freebase): δ À119 (m, 1F). IR (thin
film, cmÀ1): 2886, 2776, 1499, 1449, 1406, 1346, 1247, 1059,
846, 818. HRMS (ES+): calcd for C14H19ClFN2 (M + H)
268.1221, found 269.1221.
1-Methyl-1H-imidazole-4-carbonyl Chloride Hydrochlor-
ide (22). A 75 L glass-lined reactor was charged with 1-methyl-1H-
imidazole-4-carboxylic acid (1.2 kg, 9.52 mol, 1 equiv), dichlor-
omethane (12 L, 10 volumes), and DMF (12 mL, 1 vol/vol%).
The resulting slurry was cooled to 5 °C, and oxalyl choride
(1.33 kg, 1.10 equiv) was added over 45 min while maintaining
the temperature between 0 and 10 °C. The reaction mixture was
held for 2 h at 5 °C, warmed to 25 °C, and held for an additional
8 h. Heptane (12 L, 10 volumes) was added, and the slurry
granulated 15 min. The product was isolated by filtration,
washed with heptane (12 L, 10 volumes), and dried under a
nitrogen stream for at least 6 h. A total of 1.62 kg (94% yield) of
acid chloride HCl salt 22 was isolated as a fluffy off-white
hygroscopic solid and was stored tightly sealed and protected
from moisture to prevent degradation. 1H NMR (DMSO-d6):
δ 14.6 (br s, 1H), 9.16 (s, 1H), 8.30 (s, 1H), 3.85 (s, 3H). 13C
NMR (DMSO-d6): δ 159.0, 138.5, 127.6, 125.1, 35.8.
PF-03463275 (1), Acid Chloride Route, with Carbon Treat-
ment To Reduce Iridium. A 100 L glass-lined reactor was
charged with 12 bis(hydrochloride) salt (2.44 kg, 7.14 mol, 1 equiv,
360 ppm Ir), acid chloride 22 (1.62 kg, 8.57 mol, 1.2 equiv), and
dichloromethane (24.4 L, 10 volumes). The resulting slurry was
cooled to 5 °C, and N,N-dimethylethylamine (2.19 kg, 3.25 L,
30 mol, 4.2 equiv) was added at a rate to maintain the internal
temperature below 10 °C. The resulting thin suspension was
1
11 was present (12.2 mol, 90% yield). H NMR (400 MHz,
CDCl3): δ 3.34 (d, J = 7.2 Hz, 2 H) 3.04 (br. s., 1 H) 2.97 (d, J =
9.0 Hz, 2 H) 2.26 (m, 2 H) 2.23 (s, 3 H) 1.42 (tt, J = 7.2, 3.4 Hz, 1
H), 1.25 (m, 2H). 13C NMR (100 MHz, CDCl3): δ 64.7, 57.0
(2C), 41.5, 22.5, 21.5 (2C). IR (thin film, cmÀ1): 3316 (br),
2889, 2785, 1450, 1348, 1222, 1146, 1033, 991, 843. HRMS (ES+):
calcd for C7H14NO (M + H) 128.1075, found 128.1074.
N-(3-Chloro-4-fluorobenzyl)-1-((1R,5S,6s)-3-methyl-3-az-
abicyclo[3.1.0]hexan-6-yl)methanamine Bis(hydrochloride)
(12). A 8 L Hastelloy pressure reactor was charged with alcohol
11 (2.4 kg, 18.6 mol, 1 equiv), 3-chloro-4-fluorobenzylamine
(2.97 kg, 18.6 mol, 1 equiv), pentamethylcyclopentadienyliri-
dium(III) chloride dimer (3.70 g, 4.6 mmol, 0.00025 equiv), a
1 N aqueous sodium bicarbonate solution (9.2 mL, 9.2 mmol,
0.0005 equiv), and toluene (470 mL, 0.2 L/kg). The reactor was
purged with nitrogen and then sealed and heated to an internal
temperature of 110 °C and held for 24 h. The maximum internal
pressure during the reaction was 25 psig. At the end of this time
1059
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
warmed to 25 °C and held for 30 min, at which time the acylation
was complete (<1% 12 remaining). A solution of sodium carbonate
(3.18 kg, 30 mol, 4.2 equiv) in water (29 L) was added in one
portion, and the layers were mixed for at least 30 min, allowed to
settle, and split. The lower organic phase was returned to the
reactor and washed with water (11 L). The layers were again split.
The lower organic phase was returned to the reactor, a slurry of
Darco KBB (1.22 kg, 50 wt %) in dichloromethane (5 L) was
added to the solution of product, and the slurry was mixed for 4 h.
This mixture was filtered through a polishing filter into a speck-
free reactor, and the filter was washed with dichloromethane
(10 L). This product-rich solution was then displaced into ethyl
acetate by azeotropic distillation at atmospheric pressure, reach-
ing a final volume of 8 L. At this point the solution was cooled to
40À42 °C and held to crystallize 1. The slurry was then cooled to
30 °C, and heptane (20 L) was added, and further cooled to
15 °C. After granulating for 3 h, the product was isolated by
filtration, washed with 25 L of a 4:1 heptane/ethyl acetate mixture,
and dried under vacuum. A total of 2.14 kg of 1 was isolated (79%
filtrate was extracted with brine and water (twice), dried with
MgSO4, and concentrated to give 1.83 g (74%) of a white solid.
1H NMR (400 MHz, CDCl3): δ 7.81 (m, 1H), 7.52 (m, 1H),
7.30 (m, 1H). 13C NMR (400 MHz, CDCl3): δ 160.8 (d, 1C),
134.7 (s, 1C), 132.5 (d, 1C), 122.8 (d, 1C), 117.9 (d, 1C), 116.8
(s, 1C, 13CN), 109.5 (d,1C). LC-MS (ES+): m/z 158 (M + H)
1-(3-Chloro-4-fluorophenyl)(13C,2H2)methan(15N)amine (19).
Nitrile 18 (0.85 g, 5.4 mmol, 1 equiv) was mixed with NiCl2
(1.2 g, 9.2 mmol, 1.7 equiv) in MeOD (35 mL). NaBD4 (1.55 g,
37 mmol) was added portionwise over 30 min, and then the
reaction mixture was allowed to warm to ambient temperature
and held for 1.5 h. The mixture was then diluted with methanol
(10 mL) and filtered through Celite. The filtrate was partitioned
between ethyl acetate (30 mL) and water, and the organic phase
was then filtered again through Celite to give a clear, colorless
solution. This material was dried with MgSO4 and concentrated,
and the residue was purified by flash chromatography on silica gel
(EtOAc to 6% MeOH/EtOAc) to provide 0.5 g (57%) of 19 as a
clear, colorless oil. LC-MS (ES+) m/z 164.2, 146.1 (M À NH2).
1H NMR (400 MHz, CDCl3): δ 7.32 (m, 1H), 7.15(m, 1H),
7.01(m, 1H). 13C NMR (400 MHz, CDCl3): δ 158.5 (d, 1C),
140.2 (s, 1C), 128.7 (s, 1C), 127.4 (s, 1C), 120.8 (s, 1C), 116.5
(d, 1C), 45.8 (d, 1C, 13CD2NH2).
1
yield). Mp 125 °C. H NMR (600 MHz, 373 K, DMSO-d6):
δ 7.59 (d, J = 2 Hz, 1H), 7.54 (bs, 1H), 7.44 (d, J = 6.9 Hz, 1H),
7.27 (m, 1H), 7.26 (d, J = 2 Hz, 1H), 4.98 (br s, 2H), 3.69 (s, 3H),
3.57 (br s, 2H), 2.74 (d, J = 8.7 Hz, 2H), 2.19 (m, 2H), 2.17 (s,
3H), 1.31 (tt, J = 3, 7 Hz, 1H), 1.26 (m, 2H). 13C NMR (150
MHz, 373 K, DMSO-d6): δ 164.8, 157.6 (d, JCF = 246 Hz),
138.7, 138.6, 138.4 (d, JCF = 4 Hz), 130.7, 129.2 (d, JCF = 7 Hz),
127.5, 120.6 (d, JCF = 18 Hz), 117.7 (d, JCF = 21 Hz), 57.8 (2C),
50.5, 50.0, 41.8, 34.3, 23.5 (2C). Anal. Calcd for C19H22ClFN4O:
C, 60.55; H, 5.88; Cl, 9.41; F, 5.04; N, 14.87. Found: C, 60.42; H,
5.97; N, 14.79. HRMS (ES+): calcd for C19H23ClFN4O (M + H)
377.1544, found 377.1539.
1-(3-Chloro-4-fluorophenyl)-N-{[(1R,5S,6s)-3-methyl-3-aza-
bicyclo[3.1.0]hex-6-yl](2H1)methyl}(13C,2H2)methan(15N)amine
(21). Alcohol 11 (160 mg, 1.26 mmol, 1.03 equiv) was mixed
with 19 (200 mg, 1.22 mmol, 1 equiv) in MeOD (5 mL) and
D2O (1 mL), and the resulting solution was stirred for 15 h. The
solvent was evaporated, and the residue was redissolved and
concentrated from MeOH (5 mL) and D2O (1 mL) three more
times to fully deuterate the amine and alcohol. The residue was
dissolved in d8-toluene (2 mL), and K2CO3 (20 mg, 12 mol %)
and (Cp*IrCl2)2 (30 mg, 0.05 mol %) were added. The mixture
was heated at 120 °C bath temperature for 18 h. At this time the
mixture was diluted with dichloromethane (10 mL) and washed
with water (5 mL) three times and concentrated. The residue was
then dissolved in dichloromethane and extracted three times
with 1 N HCl (5 mL each time), and the combined aqueous
phases were basified with NaOH to pH 10 and extracted three
times with dichloromethane (10 mL each time). The resulting
organic phase was dried with MgSO4 and concentrated to
provide 230 mg of crude product (69% yield) that was used
PF-03463275 (1), Carboxylic Acid Route with T3P and No
Carbon Treatment. A 200 L glass-lined reactor was charged with
12 bis(hydrochloride) salt (4.54 kg, 13.3 mol, 1 equiv), 1-methyl-
1H-imidazole-4-carboxylic acid (1.76 kg, 13.9 mol, 1.05 equiv),
N,N-dimethylethylamine (4.13 kg, 6.1 L, 56.5 mol, 4.25 equiv),
and dichloromethane (23 L, 5 volumes). The resulting slurry was
cooled to 10 °C, and T3P (50% solution in ethyl acetate, 10.6 kg,
9.9 L, 16.6 mol, 1.25 equiv) was added at a rate to maintain the
internal temperature below 10 °C. The resulting thin suspension
was held for 1 h at 10 °C and then warmed to 25 °C and held for
8 h, at which time the reaction was complete (<1% 12 remaining).
A solution of sodium carbonate (5.91 kg, 56 mol, 4.2 equiv) in
water (51 L) was added in one portion, and the layers were mixed
for 60 min, allowed to settle, and split. The upper organic phase
was washed twice with water (23 L each wash). The layers invert
for the water washes. After the last water wash, the lower organic
phase was filtered through a polishing filter into a speck-free 100 L
reactor. This product-rich solution was then displaced into ethyl
acetate by azeotropic distillation at atmospheric pressure, reach-
ing a final volume of 18 L. At this point the solution was cooled to
40À42 °C and held to crystallize 1. The slurry was then cooled to
20 °C, and heptane (36 L) was added. After granulating for 3 h,
the product was isolated by filtration, washed with 36 L of a 3:1
heptane/ethyl acetate mixture, and dried under vacuum. A total
of 4.34 kg of 1 was isolated (87% yield).
1
without further purification. H NMR (400 MHz, CDCl3):
δ 7.38 (dd, 1H), 7.27 (m, 1H), 7.03 (m, 1H), 2.70 (m, 3H),
2.25 (m, 2H), 2.20 (s, 3H), 1.30À1.20 (m, 3H). 13C NMR (100
MHz, CDCl3): δ 157.1 (d, 1C), 137.4 (d, 1C), 129.2 (s, 1C),
127.9 (d, 1C), 119.2 (d, 1C), 116.6 (d, 1C), 56.2 (s, 2C), 49.3 (m,
1C, 13CD2N), 47.8 (m, 1C), 40.7 (d, 1C), 21.7(s, 2C), 19.3 (d,
1C). LC-MS (ES+) 274.2 (M + H, M + 5 isomer).
3-Phenyl-N-(1-phenylethyl)propan-1-amine Hydrochlor-
ide (23). 3-Phenylpropanol (10 mL, 1 equiv; 73.5 mmol), phene-
thylamine (9.45 mL, 1 equiv). and N-methylpyrrolidine (7.7 mL,
1 equiv) were combined in a 75 mL heavy-walled glass pressure
vessel with a magnetic stirbar and Teflon bushing closure equipped
with a perfluororo O-ring45 (ChemGlass cat. no. GC-1880-30).
Pentamethylcyclopentadienyliridium(III) chloride dimer (30 mg,
0.0005 equiv) was added in one portion. The headspace of the
mixture was purged, and the vessel was sealed and heated in a
130 °C oil bath for 40 h, at which time the reaction had
proceeded to 87% conversion. The volatiles were removed by
coevaporation with 2-propanol, and the resulting yellow oil dried
under reduced pressure. This material was dissolved in 25 mL
3-Chloro-4-fluoro(cyano-13C,15N)benzonitrile (18). 3-Chloro-
4-fluoro-1-iodobenzene (3.09 g, 11.7 mmol) was combined with
(13C,15N)-copper(I) cyanide (10.9 g, 10.9 mmol) in DMF. The
mixture was heated to 110 °C for 4 h and then cooled to ambient
temperature. Ethyl acetate (30 mL) and water (30 mL) were
added, and the mixture was filtered through Celite. The resulting
1060
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
2-propanol, and hydrogen chloride (5 N in 2-propanol, 22 mL,
1.5 equiv) was added in one portion. Heptane (100 mL) was then
added slowly, at which time the mixture became cloudy and the
product crystallized from solution. An additional 50 mL of heptane
was added to facilitate stirring, and the resulting white slurry was
granulated for 18 h. At this time the white solids were isolated
by filtration and were washed with 20 mL of a 1:4 mixture of
2-propanol/heptane and dried to provide 13.2 g (47.8 mmol,
(6) A rapid development timeline dictated that we commit to raw
materials ordering before route selection. Consequently, 2, 6, and 8 were
selected to allow for maximum flexibility in choosing a manufacturing
route for the initial batch of 1.
(7) (a) Rylander, P. N. Hydrogenation Methods; Academic Press Inc.:
San Diego, CA, 1985. (b) For a representative procedure, see: Aurelio,
L.; Box, J. S.; Brownlee, R. T. C.; Hughes, A. B.; Sleebs, M. M. J. Org.
Chem. 2003, 68, 2652–2667.
(8) For an example of a reaction where debenzylation proceeds in
the presence of formaldehyde, see: Kende, A. S.; Liu, K.; Brands, K. M. J.
J. Am. Chem. Soc. 1995, 117, 10597–10598. It appears that increased
catalyst loading is necessary to retain catalytic activity.
(9) If not removed, small amounts of unreacted formaldehyde (as
polymer) retained in crude isolates of 11 resulted in formation of
N-methyl derivatives of amine 6 and 12 in the next step (see Figure 2).
(10) (a) Mancuso, A.; Swern, D. Synthesis 1981, 165–185. (b)
Tidwell, T. Synthesis 1990, 857–870.
1
65% yield) of the title compound. Mp 142 °C. H NMR (400
MHz, CDCl3): δ 10.20 (br. s., 1 H) 9.82 (br. s., 1 H) 7.61À7.54
(m, 2 H) 7.45À7.33 (m, 3 H) 7.21À7.08 (m, 3 H) 7.04 (m, 2 H)
4.15 (dqd, J = 9.7, 6.9, 2.9 Hz, 1 H) 2.70À2.58 (m, 2 H)
2.58À2.41 (m, 2 H) 2.27À2.13 (m, 2 H) 1.83 (d, J = 6.9 Hz,
3 H). 13C NMR (100 MHz, CDCl3): δ 139.9, 135.9, 129.3,
129.2, 128.4, 128.2, 127.8, 126.1, 59.0, 45.1, 32.6, 27.2, 20.4.
(11) Kim, K. S.; Cho, I. H.; Yoo, B. K.; Song, Y. H.; Hahn, C. S.
J. Chem. Soc., Chem. Commun. 1984, 762–763.
’ ASSOCIATED CONTENT
(12) Parikh, J. R.; Doering, W. V. E. J. Am. Chem. Soc. 1967, 89,
5505–5507.
(13) Mancuso, A. J.; Huang, S.; Swern, D. J. Org. Chem. 1978, 43,
S
Supporting Information. Synthesis and experimental
b
procedures for the preparation of 2, a full list of resins examined
for Ir scavenging for 1, ReactIR monitoring data for the reductive
amination synthesis of 12, an NMR in-process analysis for the
reaction between 6 and 11, and 1H and 13C spectra of 11, 12, 22,
23, and 1. This material is available free of charge via the Internet
at http://pubs.acs.org.
2480–2482.
(14) Due to lack of a chromophore, the initial reaction completion
and product quality assays were typically conducted using GC and
NMR, with the best data obtained by taking NMR spectra of aliquots of
the reaction mixture.
(15) Imine 14 is not stable to HPLC analysis or TLC, unlike the
imine product by condensation of N-Boc-protected aldehyde 5 and
amine 6. See the Supporting Information for analytical and reaction
monitoring details for this sequence.
(16) Granular NaBH4 was employed to reduce contact reactivity
with the methanol-rich solution of imine during addition.
(17) First, aqueous sodium hydroxide was added to quench un-
reacted borohydride and break the amine-boron complex, which other-
wise persisted after an aqueous work-up and complicated product
isolation. Removal of methanol and of triethylamine and pyridine via
azeotropic distillation with water gave a biphasic mixture that was
extracted with ethyl acetate and then exchanged to 2-propanol.
(18) Activated DMSO reagents are known to react with base at
elevated temperatures to provide methylthiomethyl electrophiles via
Pummerer rearrangement. See, for example: van der Linden, J. J. M.;
Hilberink, P. W.; Kronenburg, C. M. P.; Kemperman, G. J. Org. Process
Res. Dev. 2008, 12, 911–920. This process was demonstrated by running
this reaction using DMSO-d6 as solvent, which resulted in formation of
d2-methylated (M + 16) isomers of 16 and 17.
’ AUTHOR INFORMATION
Corresponding Author
*Phone: +1-860-715-2505. E-mail: martin.a.berliner@pfizer.com.
’ ACKNOWLEDGMENT
The authors acknowledge the efforts of the staff of the Groton
Kilo Lab who executed this chemistry on scale. We thank Angel
Diaz, Robert Williams, Zhaohui Lei, Bob Rafka, Bao Nguyen,
Melissa Rockwell, Zhongli Zhang, Todd Zelescki, and Darlene
Hall for project support. We also thank Pfizer colleagues Rob
Singer and Joel Hawkins for helpful suggestions and Josh Dunetz
and Dan Bowles for proofreading the manuscript.
’ REFERENCES
(19) For an example preparation of a labeled benzonitrile, see:
Matloubi, H.; Shafiee, A.; Saemian, N.; Shirvani, G.; Daha, F. J.
J. Labelled Compd. Radiopharm. 2004, 47, 31–36.
(20) Khurana, J. M.; Kukreja, G Synth. Commun. 2002, 32, 1265–1269.
(21) Fujita, K.; Li, Z.; Ozeki, N.; Yamaguchi, R. Tetrahedron Lett.
2003, 44, 2687–2690.
(1) Leading reviews: (a) Surry, D.; Buchwald, S. Angew. Chem., Int.
Ed 2008, 47, 6338–6361. (b) Magano, J.; Dunetz, J. R. Chem. Rev. 2011,
111, 2177–2250.
(2) For a review, see: Guillena, G.; Ramꢀon, D. J.; Yus, M. Chem. Rev.
2010, 110, 1611–1641.
(22) Krische and co-workers have postulated that deprotonation of
an iridium(III) hydride by a carbonate base resulting in formation of an
iridium(I) species is a key step in the enantioselective coupling between
alcohols and allyl nucleophiles. See: Han, S. B.; Kim, I. S.; Krische, M. J.
Chem. Commun. 2009, 7278 and references therein.
(3) (a) Lebel, L.; Tunucci, D.; Valentine, J.; Lowe, J. A., III; Hou, X.;
Schmidt, C.; David Tingley, F., III; McHardy, S.; Kalman, M.; DeNinno,
S.; Sanner, M.; Ward, K. Bioorg. Med. Chem. Lett. 2009, 19, 2974–2976.
(b) For a review, see: Wolkenberg, S.; Sur, C. Curr. Top. Med. Chem.
2010, 10, 170–186.
(23) Hamid, M. H. S. A.; Williams, J. M. J. Chem. Commun. 2007, 725.
(24) PersonalcommunicationfromPfizer colleague Dr. Robert A. Singer.
(25) Amide 21 is a byproduct formed in the final step from acylation
of 6 with 8. We could not separate 21 from 1 by nonchromatographic
methods.
(4) Brighty, K.; Castaldi, M. Synlett 1996, 1097–1099. We utilized a
different route to prepare 2 to facilitate safe execution on scale. See the
Supporting Information.
(5) A recent report (Roughley, S. D.; Jordan, A. M. J. Med. Chem.
2011, 54, 3451–3479) analyzing the types of reactions utilized in current
medicinal chemistry publications has found that ca. 65% of all transfor-
mations are bond-forming processes, and of those, over 80% are CÀN
bond-forming reactions, with CÀC bond-forming reactions a distant
second at 15%. In the direct experience of this author this trend toward
late-stage fragment assembly through carbonÀheteroatom coupling
reactions has become much more prevalent over the past decade.
1061
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062

Organic Process Research & Development
ARTICLE
(26) Resins currently marketed for iridium removal postdate this
work but were tested on 1 and did not provide useful reductions in
iridium content. See the Supporting Information for full details.
(27) The fixed reactors in our kilolab are equipped with descending
condensers. U-Shaped vapor traps are installed on the distillate return
piping to prevent reflux vapors from reaching the condenser.
(28) For this compound, we observed cracking and channeling
immediately after the last rinse solvent had been pushed through the
filtercake. Cracking was prevented by halting the filtration and adding
additional rinse solvent when the liquor level was approx 0.5À1 cm
above the cake surface. Interestingly, not all batches of crystallized 12
bis(HCl) salt demonstrated this tendency.
(29) Fujita, K.; Enoki, Y.; Yamaguchi, R. Tetrahedron 2008, 64,
1943–1954.
(30) Timing constraints necessitated that we focus on improve-
ments to the existing Ir-catalyzed chemistry. Plans to reinvestigate the
effectiveness of other Ir- and Ru-derived catalysts were abandoned as a
result of the discontinuation of this project.
(31) After the completion of this optimization effort, we determined
that inorganic carbonate bases or water proved to be unnecessary to
achieve high turnover numbers (TONs). Stronger aqueous bases
(NaOH, R4NOH) led to catalyst inactivation.
(32) A preliminary experiment under recirculating flow conditions
(150 °C loop temperature, 10 min residence time, 30 h reaction time)
generated a similar product distribution.
(33) Williams and co-workers have described the use of (Cp*IrI2)2
in amination reactions in water in the absence of exogenous base.
(a) Saidi, O.; Blacker, A. J.; Lamb, G. W.; Marsden, S. P.; Taylor, J. E.;
Williams, J. M. J. Org. Process Res. Dev 2010, 14, 1046–1049. (b) Saidi,
O.; Blacker, A. J.; Farah, M. M.; Marsden, S. P.; Williams, J. M. J. Chem.
Commun. 2010, 46, 1541–1543.
(34) Poor solubility of the mono-HCl salt of 12 results in signifi-
cantly reduced reaction rates and lower terminal conversions in solvents
other than CH2Cl2, even when excess base was employed.
(35) Only partial conversion of 12 to 1 was observed under
SchottenÀBaumann conditions due to competitive and extremely rapid
hydrolysis of the water-soluble 22 to 8.
(36) Sodium carbonate was employed to minimize amide hydrolysis.
(37) This mixture contains 1 HCl and ionic components resulting
3
from reaction of 1 with 22 and subsequent degradation.
(38) (a) Stare, M.; Laniewski, K.; Westermark, A.; Sj€ogren, M.; Tian,
W. Org. Process Res. Dev. 2009, 13, 857–862. (b) Levin, D. Org. Process
Res. Dev. 1997, 1, 182.
(39) The full results of this screen are described in the Supporting
Information.
(40) This process (amine cross-coupling catalyzed by Cp*Ir com-
plexes) has been documented by Williams and co-workers. Saidi, O.;
Blacker, A. J.; Farah, M. M.; Marsden, S. P.; Williams, J. M. J. Angew.
Chem., Int. Ed. 2009, 48, 7375–7378.
(41) Amine alkylation is also observed using more hindered bases. For
example, reactions using diisopropylethylamine as base (0.05 or 1 equiv) ge-
nerate the desired product and a variety of cross-coupled products resulting
from amine exchange, including mixed N-ethyl and N-isopropyl amines.
Higher temperatures appear to accelerate the amine exchange process.
(42) Balcells and co-workers have conducted a theoretical mechan-
istic study of this reaction using carbonate as the base. See: Balcells, D.;
Nova, A.; Clot, E.; Gnanamgari, D.; Crabtree, R. H.; Eisenstein, O.
Organometallics 2008, 27, 2529–2535.
(43) Fujita and Yamaguchi have recently described the (Cp*IrCl2)2-
catalyzed amination of sulfonamides with alcohols, which also proceeds
with a catalyst loading of 0.05 mol % Ir. The increased acidity of the
sulfonamide is responsible for this catalytic activity and a strong base
(KOt-Bu) has been identified to be necessary to facilitate turnover. See:
Zhu, M.; Fujita, K.; Yamaguchi, R. Org. Lett. 2010, 12, 1336–1339.
(44) See the Supporting Information for the NMR in-process
analysis of this reaction.
(45) Viton O-rings that are shipped with this equipment have poor
chemical compatibility with N-methylpyrrolidine.
1062
dx.doi.org/10.1021/op200174k |Org. Process Res. Dev. 2011, 15, 1052–1062