Development of a scalable palladium-catalyzed α-arylation process for the synthesis of a CGRP antagonist

Article abstract of DOI:10.1016/j.tet.2013.04.057

The Pd-catalyzed α-arylation of cycloheptapyridyl ketone is a key complexity-building step in the synthesis of BMS-846372, a CGRP antagonist. A first-generation process utilized Pd(OAc)₂/P^tBu ₃·HBF₄ catalyst system with a strong base NaO ^tBu. Although this process was demonstrated on multi-kilo scale, the harsh conditions led to non-selective metal catalyzed processes, which generated several operational, quality, and throughput issues. By acquiring detailed knowledge around several important process parameters, we were able to design an efficient and scalable second-generation α-arylation process using a Pd(OAc)₂/RuPhos catalyst system with the weaker base, K ₃PO₄ in tert-amyl alcohol. This new weak base process was high yielding, efficient, and superior in several respects compared to the strong base process. The strategy behind the reaction and isolation development and the process considerations important to scaling a catalytic reaction from laboratory to manufacturing scale will be discussed.

Full text of DOI:10.1016/j.tet.2013.04.057

Accepted Manuscript
Development of a Scalable Palladium-catalyzed α-Arylation Process for the Synthesis
of a CGRP Antagonist
Lopa V. Desai, Michael Hay, David Leahy, Carolyn Wei, Dayne Fanfair, Thorsten
Rosner, Yi Hsiao
PII:
S0040-4020(13)00595-4
10.1016/j.tet.2013.04.057
TET 24257
DOI:
Reference:
To appear in: Tetrahedron
Received Date: 2 March 2013
Revised Date: 12 April 2013
Accepted Date: 15 April 2013
Please cite this article as: Desai LV, Hay M, Leahy D, Wei C, Fanfair D, Rosner T, Hsiao Y,
Development of a Scalable Palladium-catalyzed α-Arylation Process for the Synthesis of a CGRP
Antagonist, Tetrahedron (2013), doi: 10.1016/j.tet.2013.04.057.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Development of a Scalable Palladium-catalyzed α-Arylation Process
for the Synthesis of a CGRP Antagonist
Lopa V. Desai*, Michael Hay*, David K. Leahy, Carolyn Wei, Dayne Fanfair,
Thorsten Rosner, Yi Hsiao
Corresponding authors’ email addresses: lopa.desai@bms.com; michael.hay@bms.com
Corresponding authors’ phone numbers: 732-227-5443; 732-227-6267
Corresponding authors’ fax numbers: 732-227-3002
Chemical Development, Bristol-Myers Squibb, 1 Squibb Dr., New Brunswick, NJ
08903, USA
Graphical Abstract:
Strong Base Conditions
4 mol% Pd loading
Alumina (8Kg/Kg) Column
60-65% Isolated Yield
First Generation
F
O
O
F
N
N
TIPSO
TIPSO
7:1 dr in-process
12:1 dr isolated
Weak Basic Conditions
2 mol% Pd loading
75-80% Isolated Yield
Second Generation
1

ACCEPTED MANUSCRIPT
Development of a Scalable Palladium-catalyzed α-Arylation Process
for the Synthesis of a CGRP Antagonist
Lopa V. Desai*, Michael Hay*, David Leahy, Carolyn Wei, Dayne Fanfair,
Thorsten Rosner, Yi Hsiao
Chemical Development, Bristol-Myers Squibb, 1 Squibb Dr., New Brunswick, NJ
08903, USA
Corresponding authors’ email addresses: lopa.desai@bms.com; michael.hay@bms.com
Corresponding authors’ phone numbers: 732-227-5443; 732-227-6267
Corresponding authors’ fax numbers: 732-227-3002
Key Words: Catalysis, α-Arylation, Palladium, heteroaromatic ketone
1. Abstract
The Pd-catalyzed α-arylation of cycloheptapyridyl ketone is a key complexity-
building step in the synthesis of BMS-846372, a CGRP antagonist. A first generation
process utilized Pd(OAc)2/PtBu3.HBF4 catalyst system with a strong base NaOtBu.
Although this process was demonstrated on multi-kilo scale, the harsh conditions led to
non-selective metal catalyzed processes which generated several operational, quality and
throughput issues. By acquiring detailed knowledge around several important process
parameters, we were able to design an efficient and scalable second generation α-
arylation process using a Pd(OAc)2/RuPhos catalyst system with the weaker base, K3PO4
in t-amyl alcohol. This new weak base process was high yielding, efficient and superior
in several respects compared to the strong base process. The strategy behind the reaction
and isolation development and the process considerations important to scaling a catalytic
reaction from laboratory to manufacturing scale will be discussed.
2. Introduction
2

ACCEPTED MANUSCRIPT
Secretion of Calcitonin Gene Related Peptide (CGRP), a 37 amino acid
neuropeptide, is known to increase during a migraine attack, leading to abnormal vascular
activity of the intracranial arteries.1 BMS-846372 is a potent CGRP antagonist with the
potential to be effective in relieving migraines.2 We envisioned synthesizing BMS-
846372 from intermediates 1 and 2 using a palladium-catalyzed α-arylation reaction
followed by reduction of the carbonyl group.3 The α-arylation reaction is an attractive
transformation for the rapid construction of C-C bonds alpha to a carbonyl moiety,4
Miura,5 Buchwald,6 Hartwig7 and others8 have shown that this transformation is effective
for the coupling of various aryl halides with esters, amides, aliphatic ketones, aldehydes
and more recently heteroaromatic ketones.9 Both electron deficient and rich aryl halides
have been successfully employed, rendering the α-arylation reaction versatile and
synthetically valuable8a. However, to the best of our knowledge, this transformation has
not been applied on a manufacturing scale, even though this disconnection is strategically
constructive. Herein, we present a diasteroeselective and scalable synthesis of advanced
intermediate 3 through a palladium-catalyzed α-arylation reaction. We will provide a
detailed discussion on the process considerations important to scaling a catalytic reaction
from laboratory to manufacturing scale.
Figure 1: Retrosynthetic Strategy to Access BMS-846372
F
O
F
O
Br
F
O
F
+
NH
F
N
N
N
N
F
O
TIPSO
N
TIPSO
O
BMS-846372
(3)
(1)
(2)
3. Results and Discussion
2.1 First Generation Development: Strong Base α-Arylation
3

ACCEPTED MANUSCRIPT
Our initial reaction development began by screening ligands, bases, and catalysts
in various solvents at 120 °C to effect the α-arylation reaction of 1 with aryl bromide 2.
Notably, the reaction afforded the product as a 6:1 thermodynamic ratio of diastereomers,
whereby the major trans-diastereomer 3 was the desired product (Scheme 1). Due to tight
timelines, we were not able to perform an extensive screen. From the few solvents that
we screened, toluene was the most effective for the desired transformation. After
screening various Pd(II) and Pd(0) pre-cursors, Pd(OAc)2 and Pd2(dba)3 afforded the
product in reasonable yield. As shown in Table 1, Pd(OAc)2 was found to be the metal
precursor of choice with all the ligands (e.g. entries 17 and 18), with Cs2CO3, K3PO4, and
NaOtBu as effective bases. Interestingly, other carbonate, phosphate and tert-butoxide
bases (Na2CO3, K2CO3, Li3PO4, Na3PO4, LiOtBu and KOtBu) afforded <5% of the
desired product suggesting a strong correlation between ligand and the base. A greater
yield of product 3 was obtained when ligands S-Phos, MePhos and PtBu3·HBF4 were
used with NaOtBu as a base (Table 1, entries 9, 12 and 17).10 From these results, we
chose to focus our initial optimization and development work on the
PtBu3·HBF4/NaOtBu conditions due to the ready availability of the phosphine ligand.
Table 1: Representative Screening Resultsa
Entry Catalyst
Ligand
S-Phos
Base
Na3PO4
Conversion (%)b
1
2
Pd(OAc)2
Pd2dba3
1%
16%
38%
50%
5%
DPEPhos
CxPOMeCy
CxPOMeCy
XantPhos
XantPhos
Binap
KOtBu
Cs2CO3
K3PO4
3
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd2dba3
4
5
K3PO4
6
Cs2CO3
Cs2CO3
Cs2CO3
NaOtBu
K3PO4
14%
19%
27%
40%
6%
7
8
S-Phos
9
S-Phos
10
MePhos
4

ACCEPTED MANUSCRIPT
11
12
13
14
15
16
17
18
Pd2dba3
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd(OAc)2
Pd2dba3
MePhos
MePhos
NaOtBu
NaOtBu
K2CO3
45%
50%
1%
PtBu3.HBF4
DtBPF
K3PO4
2%
DtBPF
NaOtBu
K3PO4
44%
45%
50%
40%
PtBu3.HBF4
PtBu3.HBF4
PtBu3.HBF4
NaOtBu
NaOtBu
a Conditions: 1 (1 equiv), 2 (1.1 equiv), Base (1.3 equiv), Pd precursor (0.05 equiv Pd), monodentate ligand
(0.1 equiv)/ bidentate ligand (0.05 equiv), base (2 equiv), toluene (10 ml/g 1), 120 °C
b Conversion based on HPLC area percent of product and starting material
Scheme 1: Pd-catalyzed a-Arylation of 1 with Arylbromide 2
O
F
F
O
O
Br
F
F
F
F
N
N
N
TIPSO
TIPSO
TIPSO
(1)
(2)
(3)
(4)
Optimization of this lead result focused on examining the effect of concentration
and temperature on the reaction kinetics and impurity profile. We observed a large
discrepancy between the overall conversion and the solution yield of the product (Figure
2) resulting from the formation of several impurities (Figure 3), some of which were
observed to be highly colored. After conducting several control experiments and
understanding the origin and fate of some of the impurities we realized that majority of
the impurities, including the highly colored ones, resulted from undesired base-induced
decomposition of the starting material and the product. We reasoned that lowering the
reaction temperature would slow down the rate of this decomposition. Indeed, lowering
the reaction temperature to 95 °C slowed down the rate of decomposition, which
addressed the discrepancy between conversion and yield, but did not entirely mitigate it.
Our approach to mitigate this discrepancy began with careful consideration of each
5

ACCEPTED MANUSCRIPT
reaction parameter. Under these conditions, the reaction is sensitive to a) ligand particle
size, wherein finely powdered ligand improved the yield of 3, b) moisture and c) oxygen
content. Due to the instability of the product to the highly basic conditions, the base
charge had to be within 1.3–1.5 equivalents and the reaction had to be cooled to 25 °C
shortly after reaching completion in order to maximize the solution yield.
Figure 2: Reaction Profile of the
Temperatures
α-Arylation Reaction at Different Concentrations and
Conditions: 1 (1 equiv), 2 (1.1 equiv), NaOtBu (1.3 equiv), Pd(OAc)2 (0.04 equiv), PtBu3.HBF4 (0.04
equiv), solvent (7 or 3 mL/g of 1), 120 or 95 °C
The major impurities observed for the PtBu3·HBF4/NaOtBu system were 5 and 6
(Figure 3).11 We believe 6 is formed via the mechanism illustrated in Figure 4. Complex
6

ACCEPTED MANUSCRIPT
11, formed from the reaction of enolate 10 with oxidative addition complex 9, could
undergo cyclopalladation with the aromatic C-H bond forming complex 13.12 Carbon-
carbon bond forming reductive elimination followed by nucleophilic displacement of the
fluoride by the enolate affords 6. Based on the mechanism, we hypothesized that the
ligand to metal ratio should have an effect on the formation of 6. Excess ligand would
occupy a coordination site on palladium in complex 11/12 and disfavor the agostic
interaction with the aromatic C-H bond. We examined ligand:metal ratios from 1:1 to 3:1
and observed that higher ligand content reduced the level of 6 formed during the course
of the reaction, although at the expense of reaction rate (Table 2).13
Figure 3: Impurities Formed During the
α
-Arylation Reaction
F
F
F
F
O
O
O
O
F
O
OH
N
N
N
N
TIPSO
OTIPS
(8)
TIPSO
TIPSO
(5)
(6)
(7)
Figure 4: Proposed Mechanism for the Formation of Impurity 6
7

ACCEPTED MANUSCRIPT
Ar
Ar
PdII
PdII
H
O
O
Br
[Pd0]
[Pd]II Br
F
F
O
N
N
F
F
TIPSO
(12)
TIPSO
Ar-Br
(2)
(9)
(11)
N
TIPSO
(10)
F
F
Ar PdII
O
F
O
O
N
N
N
TIPSO
TIPSO
TIPSO
(6)
(14)
(13)
Table 2: Effect of L:M ratio on Reaction Performance and Formation of Impurity 6a
Entry
L:M
0.9:1
1:1
% Conversion
HPLC Area % of 6
1
2
3
96
91
95
2.1
1.6
0.3
2:1
a Conditions: 1 (1 equiv), 2 (1.1 equiv), NaOtBu (1.3 equiv), Pd(OAc)2 (0.04 equiv), PtBu3.HBF4, Toluene
(4 ml/g 1), 95 °C, 4 h
Other reaction impurities, including 5, 7 and 8 are related to the reaction of
starting material 1 or product 3 with NaOtBu. Based on control experiments, we found
that subjection of desired product 3 to NaOtBu in toluene at 95 °C in the presence of
atmospheric oxygen led to the formation of impurity 7, which could undergo base
promoted reaction to afford 8, presumably via a pinacolone based rearrangement. In the
absence of oxygen, heating 3 in the presence of NaOtBu led to the formation of
benzofuran 5 (Scheme 3).
Scheme 3: Un-catalyzed Reaction of Product Forming Impurities 5, 7 and 8
8

ACCEPTED MANUSCRIPT
F
O
Base, ∆
N
TIPSO
F
O
(5)
F
F
N
F
F
O
TIPSO
(3)
O
O
OH
Base, O2
Base
∆
N
N
TIPSO
OTIPS
(8)
(7)
Based on this work, we employed the following optimized conditions: 4 mol%
Pd(OAc)2, 4 mol% PtBu3·HBF4, 1.3 equiv NaOtBu, 1.2 equiv ArBr (2) in toluene at 95
°C for 4 hours. These conditions afforded 3 in 82% solution yield as a 6:1 mixture of
diastereomers (trans:cis). After performing the reaction on scale, we next needed to
remove the metal, starting material and product related impurities. At the onset of our
isolation development work, we observed that these impurities greatly impacted
crystallization performance and thus several methods for removing them were explored.
We initially examined carbon treatments, filtration through derivatized and underivatized
silica gels, and resins. These treatments afforded streams that contained impurities 6, 5,
and other unidentified impurities which a) increased the propensity for oiling during the
crystallization, b) adversely affect the performance of downstream chemistry and c)
resulted in isolated material that was colored and of poor quality. We found that washing
the reaction stream with saturated aqueous NH4Cl followed by filtration through
activated alumina, constructed from 8 kg of alumina per kg of 1, using 20 L/kg 1 of
toluene, successfully removed these unidentified colored impurities and provided
adequate crystallization performance i.e. affording solutions with 97% potency and 98%
HPLC area purity.
9

ACCEPTED MANUSCRIPT
Upon achieving acceptable quality of the post reaction stream, a controlled seeded
crystallization protocol was developed using IPA/water (1:1 v/v). This procedure resulted
in isolation of the product in 65-67% yield (98 wt% potency) with mother liquor losses of
~ 12 M%. Notably, the diastereomeric ratio of the isolated product increased from 6:1 to
10:1 trans:cis. We demonstrated the crystallization on 62 kg, 85 kg and 95 kg scale
providing 3 in 61% average yield and 97.4% average HPLC area % purity (Table 3). The
decrease in the yield of 3 is due to the level of 6, batch 2, Table 3. It was difficult to
control impurity 6 at a specific level or explain the variability in its levels between
batches. Inconsistency in the performance of the alumina column resulted in poor quality
and purity of 3, batch 3, Table 3.
Table 3: Pilot Plant Batches with Strong Base α
-Arylation Processa
Batch
1 (kg)
62
3 (kg)
49
Yield (%)
HPLC Area % Purity
1
2
3
65
56
63
98.5
98.6
95.0
85
57
95
72
a
Conditions: 1 (1 equiv), 2 (1.1 equiv), NaOtBu (1.3 equiv), Pd(OAc)2 (0.04 equiv), PtBu3.HBF4 (0.04
equiv), toluene (4 ml/g 1), 95 °C, 4 h
2.2 Second Generation Development: Weak Base α-Arylation
Although the PtBu3·HBF4/NaOtBu process was implemented successfully on
multi-kilo scale, we identified several issues that needed to be addressed in order to
achieve a commercially viable α-arylation process. These issues included a) the
instability of the product and starting material to the reaction conditions; b) the variability
in impurity levels; c) the tedious and inconsistent alumina filtration; and d) the necessity
of several different solvents which resulted in inefficient distillations.
The root of all concerns discussed above was the use of a strong base NaOtBu.
Hence, we refocused our efforts toward using weaker bases. We postulated that the use of
weakly basic conditions would result in better stability of 1 and 3, resulting in improved
impurity profile, reduced level of colored impurities and thus avoid the need for an
alumina filtration. We were also determined to find a catalyst/ligand system that would
10

ACCEPTED MANUSCRIPT
minimize the formation of impurity 6. Our optimization efforts included a broader
number of solvents compared to our earlier work and utilized the preformed Pd(0)
catalyst, Pd(PtBu3)2 to reduce issues of active catalyst generation. Table 4 illustrates a
subset of the bases and solvents included in the screen. It is important to note that polar
aprotic solvents such as DMF, NMP or DMSO were detrimental to the reaction. Notably,
the best combinations of conversion and yield were observed when using t-amyl alcohol
in combination with several bases. The optimal base/solvent combination for the reaction
was found to be K3PO4 in t-amyl alcohol.
Table 4: Second Generation Base/Solvent Optimization for α
-Arylation Reactiona
Entry
Base
Solvent
Toluene
Conversion (%)b
Product (%)
1
2
K2CO3
K2CO3
K2CO3
Cs2CO3
Cs2CO3
Cs2CO3
Na3PO4
K3PO4
K3PO4
K3PO4
0
0
0
0
MeTHF
3
t-Amyl OH
Toluene
14
11
8
12
9
4
5
MeTHF
7
6
t-Amyl OH
t-Amyl OH
Toluene
7
6
7
0
0
8
13
22
72
11
18
58
9
MeTHF
10
t-Amyl OH
a
Conditions: 1 (1 equiv), 2 (1.1 equiv), base (3 equiv), Pd(PtBu3)2 (0.04 equiv), solvent (10 ml/g 1), 80 °C.
b Conversion based on HPLC Area Percent of Product and Starting material
With optimal base and solvent conditions in hand we focused on improving the
catalyst for this transformation. The initial optimization employed [(allyl)PdCl]2 in
combination with several ligands. Representative results are shown in Table 5. We
observed that bidendate ligands (Xantphos, DPE-Phos) generally were not effective,
providing approximately 30% conversion. When using catalysts derived from
monodentate ligands, we observed that simple trialkylphosphines possessing a large cone
angle (PtBu3) promoted the desired arylation reaction; whereas smaller trialkyl phosphine
11

ACCEPTED MANUSCRIPT
derived catalysts (PCy3) led to minimal conversion. Based on these findings, we
systematically studied the effect of a variety of bulky, biaryl phosphine ligands on the
reaction. Biaryl phosphines containing cyclohexyl groups on the phosphine center were
found to give significantly higher yields than biaryl phosphines containing tert-butyl or
aryl groups. Furthermore, the presence of a coordinating functional group (OMe, OiPr)
on either ring of the biaryl system led to improved conversion and yield (Table 6).
Table 5: Ligand Screening Resultsa
Entry
Ligand
Conversion (%)b Cone Product (%)
Angle
1
2
3
4
5
6
P(tBu)2Me·HBF4
PCy3·HBF4
PtBu3·HBF4
P(Ad)2nBu
DPE Phos
0
0
16114
17015
1828
ND
0
0
45
7
34
2
34
N/A
30
XantPhos
30
N/A
24
a Conditions: 1 (1 equiv), 2 (1.1 equiv), base (3 equiv), [(Allyl)PdCl]2 (0.0125 equiv), Ligand (0.05 equiv),
t-amyl alcohol (10 ml/g 1), 90 °C.
b Conversion based on HPLC Area Percent of product and starting material
Table 6:Effect of Substituted Biaryl Phosphines on Reaction Conversiona
Entry
Ligand
R
R'
R''
Conversion Product
Product/
Conversion
0.76
(%)b
(%)
34
5
1
2
PtBu3·HBF4 N/A
John Phos
N/A
H
N/A
H
45
H
25
0.20
12

ACCEPTED MANUSCRIPT
3
4
5
6
7
8
MePhos
X-Phos
Me
iPr
H
iPr
H
H
H
28
66
72
76
96
98
12
52
54
65
90
93
0.42
0.79
0.75
0.86
0.94
0.95
Dave Phos NMe2
H
S Phos
RuPhos
OMe OMe
H
OiPr
iPr
OiPr
iPr
H
Brett Phos
OMe
a
Conditions: 1 (1 equiv), 2 (1.1 equiv), base (3 equiv), [(Allyl)PdCl]2 (0.0125 equiv), Ligand (0.05 equiv),
t-amyl alcohol (10 mL/g 1), 80 °C.
b Conversion based on HPLC Area Percent of product and starting material
Based on the performance of X-Phos, RuPhos and S-Phos compared to other
biaryl phosphines, we drew several conclusions with respect to ligand structure and
reaction performance. First, 2,6-disubstitution on the bottom aromatic ring leads to better
in process purity and improved conversion. Second, the presence of a coordinating
substituent on the upper aromatic ring led to significant improvements in the reaction
impurity profile and conversion. For example, superior reactivity was observed with
BrettPhos.
After screening various Pd(II) and Pd(0) precursors, we selected the following
experimental conditions: Pd(OAc)2, BrettPhos, and K3PO4 in t-amyl alcohol.
Unfortunately, we could not obtain sufficient amounts of Brett Phos to satisfy our
campaign needs, while adequate quantities of RuPhos were available within our project
timelines. Using RuPhos in place of BrettPhos resulted in modest reduction of solution
yields (90% vs 95%) and required higher catalyst loading (2.5 mol% vs. 1.5 mol%) to
achieve similar reaction kinetics (Figure 5). After surveying a broad range of reaction
parameters using the Pd(OAc)2/RuPhos conditions, we found that several variables
directly impacted the rate and impurity profile of the reaction (Table 7). Specifically,
increasing the base stoichiometry (entry 1 vs. entry 2), minimizing base particle size
(entry 1 vs entry 6), minimizing reaction water content (entry 4 vs entry 5), and
increasing temperature (entry 2 vs 3) accelerated the reaction rate while maintaining a
consistent impurity profile. Notably, prolonged reaction times led to inferior impurity
profile (entry 4). Based on these studies we obtained the following optimal conditions:
13

ACCEPTED MANUSCRIPT
2.5 mol% Pd(OAc)2, 5.0 mol% RuPhos, 4.0 equiv K3PO4 in t-amyl alcohol (4 L/kg 1) at
95 °C. Under these reaction conditions, product 3 was obtained in 90% solution yield.
Figure 5: Reaction Profile of
Pd(OAc)2/RuPhos
α-Arylation Reaction Using Pd(OAc)2/BrettPhos and
Table 7: Optimization Studies Using Pd(OAc)2/RuPhos/K3PO4 in t-Amyl Alcohola
Entry
Base
Base
Particle
Size (in)
0.025
Temperature Water Content Time (h)b Product
Equiv
(°C)
(equiv)
(%)
1
2
3
4
5
4
6
4
4
4
85
85
0
0
16
12
9
94
93
92
89
93
0.025
0.025
105
85
0.4
0.4
0
0.075
34
19
0.075
85
a Reaction carried out using 1 (1 equiv), 2 (1.2 equiv), Pd(OAc)2 (0.025 equiv), RuPhos (0.05 equiv), t-amyl
alcohol (4 mL/g 1)
b Reaction time represents time required to reach 98% conversion
A key concern throughout this development work was the heterogenous nature of
the reaction. Employing an insoluble base, such as K3PO4, in large scale reactions using
over head stirring can often be problematic, leading to variable reaction rates and
14

ACCEPTED MANUSCRIPT
potential reaction stalling. In these systems, the only base available to the reactants is on
the surface of the base particle. Therefore, understanding base particle size and
maximization of available base surface area is extremely important for achieving
consistent reaction performance. Our initial analysis indicated that minimization of base
particle size increased the base surface area and increased the reaction rate. A
complicating factor in this analysis was the water content of the reaction mixture, which
caused the solid base to agglomerate, increasing base particle size, and slowing the
reaction rate. Carrying out the reaction using K3PO4 of various particle sizes and water
content (KF, Karl Fischer) showed that in order to obtain full conversion in less than 15
hours, we needed to control the base particle size to less than 400 µm and base water
content to <3.0 wt% (Figure 6). Additional studies comparing base particle size to
agitation rate indicated that controlling base particle size to less than 400 µm is optimal to
achieve maximum base suspension under normal agitation conditions.
Figure 6: Effect of Potassium Phosphate Particle Size and KF on Reaction Rate
15

ACCEPTED MANUSCRIPT
Using these reaction conditions, we observed impurities 5, 7 and 8 (Figure 4), but
notably, impurity 6 was not formed. Impurities 7 (α-hydroxylation) and 8
(rearrangement) can be controlled to <2 % by controlling headspace oxygen content to
below 0.5 % during the reaction. Impurity 5 is controlled to less than 2 % by using fewer
equivalents of base and maintaining the reaction temperature below 100 °C. In addition
to these impurities, we also observed the ligand-related impurities RuPhos and RuPhos
oxide. Notably, we were also able to detect the oxidative addition complex 9 (Figure 7)
by HPLC and tentatively characterized 9 by HPLC-MS, providing us with a reliable
method to confirm active catalyst formation. We were also able to confirm complex 9 by
an independent synthesis of [(Ruphos)PdAr(Br)] from Pd2dba3 and Ruphos.
Figure 7: Oxidative Addition Complex 9 Formed in the
α
-Arylation Reaction
F
F
Cy
Cy
P
Pd
OiPr
iPrO
(9)
NB: Cationic complex 9 has been characterized by LCMS and MS/MS.
Our next challenge was to develop an isolation protocol that provided high quality
product directly from the crude reaction mixture, eliminating the need for time
consuming distillations and solvent exchanges. Our optimized conditions to remove
K3PO4 employed the addition of 1 M HCl to the reaction mixture to dissolve the residual
base and acidify the stream, resulting in rapid and clean phase splits. Due to the high
solubility of the product in a wide range of organic solvents such as heptane, the only
viable antisolvent for the crystallization was water. Unfortunately, the immiscibility of t-
amyl alcohol and water required the use of a water miscible co-solvent in this
crystallization. After screening several co-solvents in different ratios we found we found
1:2:2.5 of t-amyl alcohol:IPA:water (v/v/v) was the optimal ratio for the ternary solvent
crystallization.
16

ACCEPTED MANUSCRIPT
Removal of the oxidative addition complex 9, which persisted during the
crystallization, was challenging and led to higher palladium levels in the isolated
material. There was a strong correlation between pH of the mother liquor and the
palladium content in the isolated material (Table 8). As a result, we reasoned that we
might be able to change the nature of complex 9 by manipulating the pH, for example, by
protonating the ligand in an acidic media. Indeed, as we lowered the pH of the stream to
<0.7, we obtained lower levels of palladium in the isolated cake (entries 1 and 2, Table
8). At lower pH, however, we started to observe desilylation of the product. Although this
procedure provided a path forward to generate material of the desired color and quality,
we felt it would be difficult to consistently achieve a mother liquor pH of <0.7 while
maintaining product stability.
Table 8: Effect of pH on Product Quality and Appearance
Entry
pH
% loss % loss
Isolated Potency
HPLC Area %
Purity
Pd
(ppm)
98
ML
15%
13%
11%
11%
wash
3%
yield %
76%
%
1
2
3
4
0.43
0.69
1.3
100.0%
99.0%
96.6%
95.7%
100.0%
4%
78%
100.0%
440
2%
80%
100.0%
760
1.64
2%
82%
99.8%
1200
In order to develop a more robust isolation process, we developed a strategy to
degrade and remove complex 9. We found that treatment of the HCl washed reaction
stream with aqueous cysteine (0.5 wt%) at 50 °C led to complete decomposition of
complex 9 over the course of 4 hours, resulting in the formation of a proposed “Pd-
cysteine” complex 10 and RuPhos (Figure 8). The decomposition of complex 9 could be
clearly monitored by HPLC due to its UV absorbance. After removal of the lean aqueous
phase, 3 could be crystallized, using the aforementioned t-amyl alcohol/IPA/water, as an
off-white solid. Unfortunately, the isolated material was contaminated with residual
RuPhos, which crystallizes readily under these conditions. Our strategy to remediate
RuPhos from isolated 3 was based upon the basicity of this ligand, which has a pKa of
17

ACCEPTED MANUSCRIPT
7.1, and the pH effect observed earlier (Table 8). Based on this data, we determined that
acidification of the crystallization medium to a pH of ≤4 should lead to protonation of the
ligand rendering it soluble in the partially aqueous mother liquor and product desilylation
was not an issue at this pH range. Therefore, we modified the crystallization conditions to
employ a 1:2:2.5 volume ratio of t-amyl alcohol:IPA: aq. H3PO4 (0.25 M), resulting in
complete removal of RuPhos from isolated 3. The optimized crystallization conditions
were employed on 700 g scale to provide 3 in 73 M% isolated yield in high quality (99.8
HPLC Area % purity, 97 wt% potency).
Figure 8: Decomposition of Oxidative Addition Complex 9 with Cysteine
F
F
pKa=7.1
Cy
Cy
Cy
P
P
Pd
OiPr
Cy
iPrO
Aq. L-Cysteine
Br
"Pd/Cysteine" +
t-Amyl Alcohol
50 C
iPrO
OiPr
(10)
RuPhos
(9)
Table 9: Comparison of the First and Second Generation α-Arylation Processes
Process
Reaction
Yield %
80%
Isolated
Yield %
61%
HPLC Area %
Purity
# Unit
Operations
NaOtBu/PtBu3.HBF4
97.4%
7
4
K3PO4/RuPhos
90%
73%
99.8%
4. Conclusion
Two processes for the α-arylation reaction between cycloheptyl pyridyl ketone 1
and difluorobromo benzene 2 were described. The first generation process utilized
Pd(OAc)2/PtBu3.HBF4 catalyst system with a strong base NaOtBu. Although this process
was demonstrated on multi-kilo scale it suffered from several operational, quality and
throughput issues. As a result a second generation process was developed using a
Pd(OAc)2/RuPhos catalyst system with the weaker base K3PO4 in t-amyl alcohol. This
18

ACCEPTED MANUSCRIPT
new weak base process was superior in several respects compared to the strong base
process. As presented in Table 9, the new conditions afford higher yield, improved
quality of product and reduced the number of unit operations. To the best of our
knowledge, this is the first example of the use of Pd(OAc)2/RuPhos/K3PO4 in t-amyl
alcohol for an α-arylation reaction on scale. In order to scale the palladium catalyzed α-
arylation reaction we addressed several important process considerations: a) catalyst
generation and stability; b) base particle size; c) stability of starting material and product
under the reaction conditions; d) moisture and oxygen content; e) understanding impurity
origin and fate and f) removal of palladium species and impurities. By acquiring detailed
knowledge around each of the process considerations, we were able to design an efficient
and scalable α-arylation process.
5. Acknowledgements
The authors wish to acknowledge David Kronenthal, Susanne Kiau, Martin Eastgate,
Francisco Gonzalez Bobes and Rajendra Deshpande for reviewing the paper and helpful
discussions. The authors also want to acknowledge Charles Pathirana, Michael
Peddicord, Wei Ding, Jonathan Marshall, Frank Rinaldi, and Merrill Davies for impurity
characterization and analytical support. Additionally Victor Rosso, Jose Tabora, Xinhua
Qian, Renee Randazzo, Vidya Iyer, and Alan Braem are acknowledged for experimental
design and campaign support.
6. Experimental
a.
Instrumentation
NMR spectra were obtained on a Bruker 400 (399.96 MHz for 1H; 100.57 MHz for 13C)
spectrometer. 1H NMR chemical shifts are reported in parts per million (ppm) relative to
TMS, with the residual solvent peak used as an internal reference. Multiplicities are
reported as follows: singlet (s), doublet (d), doublet of doublets (dd), triplet of doublets
(td), triplet (t), multiplet (m), and broad resonance (br).
b. Procedure and Characterization
(6S,9R)-6-(2,3-difluorophenyl)-9-((triisopropylsilyl)oxy)-6,7,8,9-tetrahydro-
Hcyclohepta[b]pyridin-5-one (3)
19

ACCEPTED MANUSCRIPT
First Generation Strong Base Process:
A mixture of toluene (14.9 L), 2 (3.50 kg, 18.1 mol), 1 (4.92 kg, 14.8 mol), palladium
acetate (0.132 kg, 0.588 mol), tri-tertbutylphosphonium tetrafluoroborate (0.174 kg,
0.600 mol), and sodium tert-butoxide (1.88 kg, 19.6 mol) was heated to 95 °C for 4 h.
The resulting dark mixture was cooled to 25 °C and diluted with ethyl acetate (44.5 L),
washed with saturated ammonium chloride solution (2x24.7 L), phase split and
concentrated the organic to ~ 25 L under vacuum. The dark solution was absorbed onto a
packed bed of alumina (19.8 kg) and eluted with toluene (203 L). The fractions
containing 3 were combined and solvent swapped to IPA and concentration to 3 L/Kg of
3. The IPA solution brought to 45 °C and Water (1.5 L/Kg) was added to the solution
over 10 min, and the mixture seeded with 5 wt% 3. The mixture was cooled to 40 °C over
2.5 h and Water ( 1.5 L/Kg) was added to the solution over 50 min. The resulting slurry
was cooled to 25 °C over 1 h and aged for an additional 1 h followed by filtration. The
cake was washed with a solution of IPA/water (1:1, 9.17 L) and dried under vacuum (at
atmospheric pressure, 50 °C) for 24 h to afford 3 (4.15 kg, 63 %, 12:1 dr). [α] 20 D
+166.7 (c 3.12, MeOH); Mp = 93.5-94.2 °C; 1H NMR (400 MHz, DMSO d-6) δ 8.66 (dd,
J = 4.8, 1.8 Hz, 1H), 7.91 (dd, J = 7.7, 1.7 Hz, 1H), 7.53 (dd, J = 7.7, 5.0 Hz, 1H), 7.36-
7.29 (m, 1H), 7.20-7.14 (m, 1H), 7.04-7.01 (m, 1H), 5.22(dd, J = 4.0, 2.0 Hz, 1H), 4.65
(dd, J = 12.3, 2.2 Hz, 1H), 2.29-2.21 (m, 2H), 2.10-2.02 (m, 1H), 1.99-1.93 (m, 1H),
13
1.04-0.95 (m, 3H), 0.92 (d, J = 7.1 Hz, 9H), 0.86 (d, J = 7.1 Hz, 9H); C NMR (125
MHz, DMSO d-6) δ 204.4, 160.4, 159.4, 150.9, 149.8 (dd, J = 245.2, 13.2 Hz), 147.6
(dd, J = 244.4, 12.4 Hz), 137.1, 132.8, 132.7 (d, J = 12.4 Hz), 125.0, 123.8, 115.8 (d, J =
16.8), 76.2, 49.0, 32.6, 26.7, 17.6, 11.4; 19F NMR (376 MHz, DMSO d-6) δ -139.1 (ddd,
J = 21.8, 10.3, 5.7 Hz), -142.4 (d, J = 20.7 Hz); IR (KBr) 3444 (br), 2943, 2866, 1689,
1488, 1264, 1083 cm-1; HRMS Calcd for C25H34F2NO2Si: 446.2321; HRMS found [M +
H]+: 446.2317.
Second Generation Weak Base Process:
A 10 L reactor equipped with an overhead stirrer was charged with 1 (0.63 kg, 1.89 mol),
t-amyl alcohol (2.5 L), potassium phosphate tribasic (1.63 kg, 7.56 mol), palladium
acetate (10.43 g, 0.046 mol), RuPhos (43.36 g, 0.092 mol), and 2 (0.43 kg, 2.23 mol)
20

ACCEPTED MANUSCRIPT
under a nitrogen atmosphere. The resulting suspension was heated to 95 °C until < 2 %
of 1 remained as judged by HPLC analysis. The mixture was cooled to 15 °C followed
by slow addition of aq. HCl (5.7 L, 1 M) to the mixture while maintaining the batch
temperature below 30 °C. The resulting biphasic mixture was agitated for 15 min at 25
°C . Agitation was stopped and the lower aqueous layer was removed. The rich organic
phase was diluted with t-amyl alcohol (1.4 L) and treated with an aqueous solution of L-
cysteine (5.50 L, 5 wt %). The resulting biphasic mixture was heated to 50 °C for 2 h
followed by cooling to 25 °C. Agitation was stopped and the lower aqueous layer was
removed and the rich organic phase was polish filtered thru a 1 υm line. The rich organic
filtrate was concentrated to 3 L/kg 1 via vacuum distillation (75 torr, 50 °C batch
temperature) followed by dilution with isopropanol (0.7 L). The distillation was
continued with constant addition of isopropanol until the ratio of isopropanol:t-amyl
alcohol >75:25 % v/v as judged by GC analysis. The batch was returned to atmospheric
pressure and heated to 40–45 °C. Aq. H3PO4 (0.7 L) was slowly added to the mixture
while maintaining the batch temperature >40 °C followed by the addition of seeds of 3 (7
g, 1 wt %). The resulting slurry was aged at 40 °C for 1 h. Aq. H3PO4 (1.1 L) was
slowly added to the batch while maintaining the batch temperature > 40 °C. The slurry
was held for 1 h at 40 °C, followed by cooling to 0 °C over 6 h. The slurry was isolated
by filtration. The wet cake was washed with 47/35/18 %v/v aq. H3PO4/isopropanol/t-
amyl alcohol (1x2.1 L), 50/50 %v/v isopropanol/water and dried (50 °C/100 torr) to
provide 3 (0.60 kg, 73 %) as an off-white solid.
7. References and Notes
1 a) Goadsby, P. J. Drugs 2005, 65, 2557. (b) Bell, I. M.; Bednar, R. A.; Fay, J. F.;
Gallicchio, S. N.; Hochman, J. H.; McMasters, D. R.; Miller-Stein, C.; Moore, E. L.;
Mosser, S. D.; Pudvah, N. T.; Quigley, A. G.; Salvatore, C. A.; Stump, C. A.; Theberge,
C. R.;Wong, B. K.; Zartman, C. B.; Zhang, X.-F.; Kane, S. A.; Graham, S. L.; Vacca, J.
P.; Williams, T. M. Bioorg. Med. Chem. Lett. 2006, 16, 6165. (c) Goadsby, P. J.; Lipton,
R. B.; Ferrari, M. D. N. Eng. J. Med. 2002, 346, 257. (d) Olesen, J.; Diener, H. C.;
21

ACCEPTED MANUSCRIPT
Husstedt, I.W.; Goadsby, P. J.; Hall, D.; Meier, U.; Pollentier, S.; Lesko, L. M. N. Eng. J.
Med. 2004, 350, 1104
2 Luo, G.; Chen, L.; Conway, C. M.; Denton, R.; Keavy, D.; Gulianello, M.; Huang, Y.;
Kostich, W.; Lentz, K. A.; Mercer, S. E.; Schartman, R.; Signor, L.; Browning, M.;
Macor, J. E.; Dubowchik, G. M. ACS Med. Chem. Lett. 2012, 3, 337 and references
therein.
3 Leahy, D.K.; Fan, Y.; Desai, L.V.; Chan, C.; Zhu, J.; Luo, G.; Chen, L.; Hanson, R.L.;
Sugiyama, M.; Rosner, T.; Cuniere, N.; Guo, Z.; Hsiao, Y.; Gao, Q, Org. Lett., 2012, 14,
4938.
4 Culkin, D. A.; Hartwig, J. F. Acc. Chem. Res. 2003, 36, 234.
5 a) Satoh, T.; Kawamura, Y.; Miura, M.; Nomura, M. Angew. Chem., Int. Ed. Engl.
1997, 36, 1740.; b) Miura, M.; Nomura, M. Top. Curr. Chem. 2002, 219, 211. c) Y.;
Satoh, T.; Miura, M.; Nomura, M Tetrahedron Lett. 2002, 43, 101.; d) Miura, M.; Satoh,
T. Palladium in Organic Synthesis; Springer: Berlin/ Heidelberg, 2005; Vol. 14. e) Terao,
Y.; Fukuoka, Y.; Satoh, T.; Miura, M.; Nomura, M Tetrahedron Lett. 2002, 43, 101.
6 a) Palucki, M.; Buchwald, S. L. J. Am. Chem. Soc. 1997, 119, 11108.; b) Moradi, W.
A.; Buchwald, S. L. J. Am. Chem. Soc. 2001, 123, 7996.; c) Martín, R.; Buchwald, S. L.
Angew. Chem., Int. Ed. 2007, 46, 7236
7 a) Hamann, B. C.; Hartwig, J. F. J. Am. Chem. Soc. 1997, 119, 12382.; b) Hama, T.;
Liu, X.; Culkin, D.; Hartwig, J. F. J. Am. Chem. Soc. 2003, 125, 11176.; c) Liu, X.;
Hartwig, J. F. J. Am. Chem. Soc. 2004, 126, 5182.; d) Vo, G. D.; Hartwig, J. F. Angew.
Chem., Int. Ed. 2008, 47, 2127
8 a) Bellina, F.; Rossi, R. Chem. Rev. 2010, 110, 1082; b) Burtoloso, A. C. B. Synlett
2009, 320; c) Lloyd-Jones, G. C. Angew. Chem. Int. Ed. 2002, 41, 953
9 Desai, L.V.; Ren, D.T.; Rosner, T, Org. Lett., 2010, 12, 1032
10 Although these conditions were attractive due to the unprecedented application of
CxPOMeCy ligand in an α-arylation reaction, we ran into several impurity and palladium
removal issues during isolation upon scale-up
11 All impurities were characterized and their structures predicted based on chemical
formulas and fragmentation pattern obtained by MS/MS and LCMS.
22

ACCEPTED MANUSCRIPT
12 Lyons, T. W.; Sanford, M. S., Chem. Rev., 2010, 110, 1147 and references therein.
13 Hartwig has proposed a mono-ligated complex to be the catalytic active species for
arylation of esters with arylbromides when using PtBu3 ligand with NaOtBu. Hama, T.;
Hartwig, J. F.; Org. Lett. 2008, 10, 1545
14 Rahman, M. M.; Liu, H. Y.; Prock, A.; Giering, W. P. Organometallics 1987, 6, 650
15 Tolman, C. A. Chem. Rev. 1977, 77, 313
23