Development of a scalable synthesis of a GPR40 receptor agonist

Article abstract of DOI:10.1021/op1003055

Early process development and salt selection for AMG 837, a novel GPR40 receptor agonist, is described. The synthetic route to AMG 837 involved the convergent synthesis and coupling of two key fragments, (S)-3-(4-hydroxyphenyl) hex-4-ynoic acid (1) and 3-

Full text of DOI:10.1021/op1003055

ARTICLE
pubs.acs.org/OPRD
Development of a Scalable Synthesis of a GPR40 Receptor Agonist
Shawn D. Walker,* Christopher J. Borths, Evan DiVirgilio, Liang Huang, Pingli Liu, Henry Morrison,
Kiyoshi Sugi, Masahide Tanaka, Jacqueline C. S. Woo, and Margaret M. Faul
Chemical Process Research and Development, Amgen Inc., Thousand Oaks, California 91320, United States
S Supporting Information
b
ABSTRACT: Early process development and salt selection for AMG 837, a novel GPR40 receptor agonist, is described. The
synthetic route to AMG 837 involved the convergent synthesis and coupling of two key fragments, (S)-3-(4-hydroxyphenyl)hex-4-ynoic acid
(1) and 3-(bromomethyl)-40-(trifluoromethyl)biphenyl (2). The chiral β-alkynyl acid 1 was prepared in 35% overall yield via classical
resolution of the corresponding racemic acid (()-1. An efficient and scalable synthesis of (()-1 was achieved via a telescoped sequence of
reactions including the conjugate alkynylation of an in situ protected Meldrum’s acid derived acceptor prepared from 3. The biaryl bromide 2
was prepared in 86% yield via a 2-step SuzukiꢀMiyaura couplingꢀbromination sequence. Chemoselective phenol alkylation mediated by
tetrabutylphosphonium hydroxide allowed direct coupling of 1and 2to afford AMG 837. Due to the poor physiochemical stability of the free
acid form of the drug substance, a sodium salt form was selected for early development, and a more stable, crystalline hemicalcium salt
dihydrate form was subsequently developed. Overall, the original 12-step synthesis of AMG 837 was replaced by a robust 9-step route
affording the target in 25% yield.
’ INTRODUCTION
Knoevenagel condensation of 4-hydroxybenzaldehyde (4) with
Meldrum’s acid (5) in water provided benzylidene malonate 3.3
Protection of the phenol group of 3 as the corresponding THP
ether4 6, followed by addition of propynylmagnesium bromide,
afforded the conjugate addition product 7. Thermolysis of 7 in
diethyl ketone/water resulted in decarboxylation with concomi-
tant cleavage of the THP group to provide the racemic acid
(()-1. Resolution of (()-1 was accomplished by chiral chro-
matography or via diastereomeric salt formation with (1S,2R)-
1-amino-2-indanol followed by salt break to furnish 1.
Type 2 diabetes is a growing health concern that afflicts over
150 million people worldwide. Chronic hyperglycemia associated
with diabetes can result in damage and dysfunction of various organs,
including the eyes, kidneys, nerves, and heart. Activation of the
G-protein coupled receptor GPR40 amplifies glucose-stimulated
insulin secretion and lowers plasma glucose concentrations in multi-
ple animal models of insulin resistance and obesity. Therefore,
GPR40 agonists are being investigated as potential new therapeutic
agents for the treatment of type 2 diabetes.1 Herein we report the
development of an efficient synthesis of AMG 837, a novel agonist of
GPR40 that was selected for clinical development.
As shown in Scheme 1, AMG 837 possesses a chiral β-alkynyl acid
fragment connected via an ether linkage to a trifluoromethyl-sub-
stituted biaryl. Thus, a convergent synthetic route would comprise the
synthesis and coupling of two key fragments, (S)-3-(4-hydroxyphe-
nyl)hex-4-ynoic acid (1) and 3-(bromomethyl)-40-(trifluoromethyl)-
biphenyl (2). Due to the paucity of scalable synthetic methods avail-
able for enantioselective alkynylation at the outset of this program,2
efforts focused on the development of a racemic synthesis coupled
with resolution to prepare the chiral intermediate 1. Key to the success
of this approach was the development of a practical synthesis of the
Meldrum’s acid derived acceptor 3. Synthesis of the biaryl 2 via
SuzukiꢀMiyaura cross-coupling was evaluated using various combi-
nations of arylboronic acidꢀaryl halide coupling partners to minimize
unit operations and cost-of-goods. The preparation and coupling of
these intermediates posed a number of challenges for large-scale
synthesis. This paper describes how these challenges were addressed
during the development of a robust process for preparing a pharma-
ceutically acceptable salt form of AMG 837.
In order to prepare sufficient quantities of 1 to supply the
program, a number of limitations needed to be addressed. Although
the initial Knoevenagel condensation reaction was reversible, it was
driven forward in this case by the low solubility (<0.1 mg/mL) of the
product 3in water. In practice, a hot (50ꢀ60 °C) slurry of Meldrum’s
acid (5) in water was added to a solution of aldehyde 4 in water, and
the resulting mixture was aged at 75 °C. This hetereogeneous (slurry-
to-slurry) process produced a sticky suspension that filtered slowly
and required recrystallization of 3 from acetone/water to increase the
purity to >97%. Significantly, as the scale of the Knoevenagel conden-
sation increased, the isolated yield of 3 decreased from 70% to 80%
yield on 1ꢀ3 kg scale to 40ꢀ50% yield on 75ꢀ80 kg scale.
The THP protection step suffered from variable yields (<50ꢀ
90%) and incomplete reaction, often stalling at 80ꢀ95% con-
version. Charging additional catalyst (PPTS) or 3,4-dihydro-2H-
pyran (DHP) did not improve conversion, and concomitant
formation of black polymeric material was observed.5 In addition,
complete solvent switch from dichloromethane to acetone was
required to isolate reasonable yields of solid 6, necessitating a
lengthy distillation. This process afforded 6 in relatively low
purity (90ꢀ96%) and was not considered robust for scale-up due
to the instability of the THP group.
’ RESULTS AND DISCUSSION
Original Synthesis of Hexynoic Acid 1. The original 6-step
synthesis of hexynoic acid 1 is summarized in Scheme 2.1b
Received: November 14, 2010
Published: March 27, 2011
r
2011 American Chemical Society
570
dx.doi.org/10.1021/op1003055 Org. Process Res. Dev. 2011, 15, 570–580
|

Organic Process Research & Development
ARTICLE
Scheme 1. Key Intermediates for the Synthesis of AMG 837
Scheme 2. Original Synthesis of 1
The conjugate addition of propynylmagnesium bromide to
alkene 6 afforded 7 in 80ꢀ89% yield, depending on the purity of
the starting material. However, during the extended reaction
time (∼3 days) required for hydrolysisꢀdecarboxylation of 7,
several unidentified impurities were generated. The racemic acid
(()-1 was isolated from this process in 67ꢀ74% yield and
85ꢀ95% purity. Recrystallization or purification via formation of
a bis-diisopropylamine salt was used to provide (()-1 in >95%
purity.6 Resolution via diastereomeric salt formation with 1.0
equiv of (1S,2R)-1-amino-2-indanol in isopropyl alcohol (IPA)
afforded 1 in ∼93% ee (28% yield) after two crystallizations and a
salt break. Further upgrade of the optical purity to our target of
>97% ee was accomplished with chiral chromatography. The
overall yield of 1 was <10%, and a more efficient and controlled
process was required.
Table 1. Reaction Conditions for the Knoevenagel
Condensation
entry
solvent
conversion (%)
note
1
2
3
water
organic solvents
water/toluene (10/1)
85
<10
>97
fine solids
MTBE, IPAc,THF, toluene, MeCN
spherical beads (agglomeration)
Optimized Synthesis of Hexynoic Acid 1. To investigate the
reduction in yield as the scale of the Knoevenagel condensation
increased, stability studies were conducted on both Meldrum’s
acid (5) and the condensation product 3. It was determined that
Meldrum’s acid decomposed rapidly in water above ∼35 °C,
with a half-life of ∼2 h at 50 °C.7 Thus, the longer times required
to heat up and transfer aqueous slurries of Meldrum’s acid on
larger scale resulted in partial hydrolysis and contributed to the
decreased yields of 3. Although 3 was stable in dry organic
solvents including acetone, it readily hydrolyzed in miscible
solvent/water mixtures.8 Consequently, during recrystallization
of 3 from acetone/water, a retro-Knoevenagel reaction pathway
was further eroding isolated yields. Studies directed at identifying
milder condensation conditions revealed that the conversion was
incomplete in water at 30ꢀ35 °C (Table 1, entry 1) and that the
reaction did not proceed in dry organic solvents (entry 2).9
However, addition of 10% of an immiscible organic solvent to
water, including isopropyl acetate, toluene, tert-butyl methyl
ether and 2-butanone, improved the reaction conversions to
>97%. Among the solvent systems tested, toluene/water (entry
3) was unique in its ability to promote agglomeration of the
product 3 into relatively uniform spherical beads (Figure 1).10
These beads were easily filtered and dried quickly to provide 3 in
571
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
Figure 1. Spherical beads formed during agglomeration of 3 (size compared with standard gage rods).
salt-free solution of 9 was treated directly with propynylmagne-
sium bromide (1.1 equiv) to furnish the adduct 10.17 Aqueous 2
N NaOH (4 equiv) was then added to the crude solution of 10,
and the mixture was heated at 50 °C for 1ꢀ2 h to complete
hydrolysis of the acetate group.18 After pH adjustment, separa-
tion of the aqueous phase and solvent swap from THF to toluene,
11 was isolated via filtration as a white solid in 87% overall yield
for the 3-step sequence. The purity of 11 by HPLC was 96 wt %,
containing 3 wt % of the bis-acid 12 (see Scheme 4).19
Decarboxylation of 11 in 9:1 DMF/water (3 vol) at 100 °C for
1ꢀ2 h afforded a clean solution (>98% HPLC purity) of racemic
acid (()-1 (Scheme 4). Formation of an intermediate diacid 12
was observed during the course of the reaction, which is
consistent with a stepwise hydrolysisꢀdecarboxylation pathway.
The pure product (()-1 could be obtained via crystallization
from IPAc/heptane, or the solution of (()-1 conveniently
telescoped into the subsequent resolution step.
Figure 2. Representative waterfall plot (starting material and product
absorptions at 2 min intervals) for in situ reaction monitoring via Raman
spectroscopy.
Classical resolution of (()-1 was studied with >20 commer-
cially available chiral amines and confirmed the selection of
(1S,2R)-1-amino-2-indanol as the optimum agent. It was found
that using only 0.55 equiv of the amine in n-propanol (5 vol) gave
chiral salt 13 with high diastereomeric excess (94ꢀ97% de).
Slurrying the isolated salt in additional hot n-propanol (2.5 vol)
improved the purity to >98% de with ∼28ꢀ34% overall yield
(out of a maximum of 50%).
99% yield (>99 wt % purity) with <0.03 wt % residual water by
Karl Fischer titration (KF).11 Due to the high quality of this
material, recrystallization was not required.
In situ Raman spectroscopy was also used to monitor the
course of the reaction, since the starting aldehyde and product 3
showed distinct absorptions at 1587 and 1560 cmꢀ1, respectively.12
Under the optimized conditions the reaction was complete in <1 h
at 33 °C (Figure 2); however, the batch was typically aged for 2ꢀ3 h
to promote agglomeration before filtering.
Attention then turned to developing a through process to the
racemic acid (()-1. To circumvent difficulties associated with
the labile THP protecting group, alternative protection strategies
were investigated. In situ protection of the phenol group of 3 as
the corresponding phenoxide salt13 8 was evaluated, but the
subsequent addition of propynylmagnesium bromide typically
stalled at <70% conversion.14
Additional studies revealed that the yield for the resolution
could be increased to 39ꢀ41% (98% de) by performing the
initial salt formation in acetonitrile (10 vol) and the hot slurry in
acetonitrile containing 6ꢀ8 wt % water (5 vol total). Salt break
with aqueous acid, extraction with ethyl acetate and crystal-
lization (or solvent-swap to THF for the next step) afforded 1 in
35% overall yield. The optimized synthesis of hexynoic acid 1 is
summarized in Scheme 5.
Original Synthesis of Bromide 2. The original 3-step synthe-
sis of bromide 2 is shown in Scheme 6. SuzukiꢀMiyaura coupling
(SMC) of 3-bromobenzoic acid (14) and 4-trifluoromethylphe-
nylboronic acid (15) was carried out according to the method of
Tiffin and co-workers20 using catalytic palladium on carbon to
provide biaryl 16. Borane reduction of 16 afforded the benzylic
alcohol 17. Treatment of 17 with thionyl bromide in dichlor-
omethane produced the bromide 2, which was recrystallized
from hexane or heptane. The overall yield of 2 was 72ꢀ82%, and
early development work focused on modifications to make the route
Further experimentation revealed that acetate 9 met our
requirements as an inexpensive group that could be rapidly
deployed and cleaved (Scheme 3). Treatment of 3 with acetyl
chloride (1.1 equiv) in the presence of N-methylmorpholine (1.2
equiv) in THF (5 vol15) smoothly furnished 9. The insoluble
N-methylmorpholine hydrochloride produced in the reaction
was conveniently removed via in-line filtration,16 and the resulting
572
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
Scheme 3. Telescoped Process to 11
Scheme 4. Decarboxylation Pathway to (()-1
Scheme 5. Optimized Synthesis of 1
more amenable to scale-up. The key issues identified were (i) during
reagent charges for the SMC reaction, addition of aqueous Na2CO3
to a slurry of 14 and 15 resulted in vigorous CO2 off-gassing and
foaming of the batch; (ii) the bulk density of 16 was very low (∼0.1
g/mL), which made handling and transfer of the powder difficult; (iii)
during reduction of 16 to 17, a delayed exotherm was observed after
collected solids were dissolved in hot acetone (50 °C, 9 vol) and
polish filtered, and water (12 vol) was added. The mixture was
then cooled to 10 °C to crystallize 16. Using this process, a 19.7
kg batch of acid 16 was prepared in 91% yield and >99% purity
with improved bulk density (0.2 g/mL).
The reduction of 16 to 17 was originally carried out by slowly
adding 1.2 equiv of BH3 THF (1.0 M in THF) to a solution of 16 in
the addition of BH3 THF was complete, which was considered
3
3
THF at 0 °C, followed by gradual warming of the mixture to 30 °C
over several hours.22 A development run performed with 40 g of 16
demonstrated that when borane addition was complete, the initial
exotherm subsided, but only 14% conversion to product had
occurred. As the mixture was subsequently warmed, a second
exotherm initiated at ∼4 °C that rapidly heated the reactor contents
to 9 °C despite jacket cooling. This result was consistent with the
accumulation and exothermic sequential reduction of an acylox-
yborane intermediate.23 In order to achieve conditions in which
accumulation of intermediates was not significant, the reduction was
potentially hazardous for scale-up; and (iv) the bromination reaction
was performed in dichloromethane, and emulsions were frequently
observed during aqueous workup.
Optimized Synthesis of Bromide 2. The catalyst loading in
the SMC reaction was successfully lowered from 1.1 wt % to 0.13
wt % palladium21 with complete conversion in 3 h at 75 °C. The
issue with reaction foaming was addressed by changing the order
of addition of reagents. By charging an aqueous solution of
Na2CO3 to the reactor first followed by addition of the other
reagents (14 and 15), the foaming was greatly suppressed and a
stirrable slurry was obtained. After reaction completion, the
mixture was filtered through a pad of Celite to remove the
catalyst, the pH of the filtrate was adjusted to <2 with aqueous
HCl, and the product 16 was directly isolated by filtration. The
performed at higher temperature. Thus, addition of BH3 THF over
3
1ꢀ2 h to a solution of 16 in THF at 45 °C smoothly afforded
alcohol 17. At this temperature the reaction was feed-controlled and
reaction conversion was proportional to the fraction of BH3 THF
3
573
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
Scheme 6. Original Synthesis of 2
Scheme 7. Optimized Synthesis of 2
Scheme 8. Improved Synthesis of 2
dosed. A modification of this process was also performed on 20 kg
scale using borane generated in situ from BF3 THF and NaBH4
3
(eq 1).
4BF3 THF þ 3NaBH4 f 4BH3 THF þ 3NaBF4 ð1Þ
3
3
The controlled addition of borane was replicated by slowly
adding BF3 THF to a mixture of 16, NaBH4, and B(OMe)3 in
3
THF.24 Lab-scale development runs indicated that the reaction
under these conditions was again feed-controlled and that
reaction conversion (by HPLC) was proportional to the fraction
isopropyl acetate to heptane to crystallize the product, biaryl alcohol
17 was isolated in 95% yield (containing 6 ppm residual Pd).
Bromination with thionyl bromide in toluene completed the 2-step
sequence to 2.
of BF3 THF added. During large scale production, the progress of
3
the reaction during the BF3 THF addition was confirmed by
3
Synthesis of AMG 837. The Medicinal Chemistry route
began with Fischer-esterification of 1 in methanol to afford 20,
followed by base-mediated phenol alkylation and ester hydrolysis
to provide the free-acid form of AMG 837 (Scheme 9). Both
intermediate esters 20 and 21 were isolated as oils after filtration
through silica gel and concentration of the filtrate. The original
conditions for alkylation of 20 employed cesium carbonate in acetone
at room temperature, and the solid paste remaining at the end of the
reaction was removed by filtration through Celite. The filtration was
slow and was anticipated to be even more difficult on larger scale. We
found potassium carbonate (1.5 equiv) to be a suitable alternative
base for the alkylation of 20 (1.0 equiv) with 2 (1.1 equiv), and the
reaction proceeded smoothly in 2 vol of refluxing acetone to provide
21.28 The resulting heterogeneous mixture was easily filtered to
remove residual base, and then solvent switched to ethanol for the
next step. Since the ester 21 was rather unstable,29 solutions of the
product were not held at this stage but instead processed forward.
Aqueous sodium hydroxide (2.0 equiv) was added to the ethanolic
solution of 21 to hydrolyze the ester as well as scavenge excess alkylat-
ing agent 2.30 The resulting mixture was neutralized with aqueous
HCl, concentrated and extracted with isopropyl acetate.31 The iso-
propyl acetate phase was washed with water and dried via azeotropic
distillation to afford a solution of AMG 837 in 92ꢀ96% assay yield.
Although this 3-step process was successfully employed in multiple
batches to prepare ∼20 kg AMG 837 overall, a chemoselective
phenol alkylation process was subsequently developed that elimi-
nated the need for the ester protection and hydrolysis steps. Extensive
base and solvent screening revealed that treatment of 1with 2.0 equiv
constant generation of heat (internal reactor temperature mon-
itoring) to avoid the accumulation of starting material or intermedi-
ates. During the process, the reactor was vented to a scrubber
containing 5% methanolic sodium methoxide to decompose any
borane in the vent-gas. After aqueous workup, extraction with
toluene and azeotropic drying via distillation, a solution of alcohol
17 (>99% purity, containing <200 ppm water) in toluene was
obtained and used directly into the next step.
Bromination of 17 with SOBr2 was slower in toluene than in
dichloromethane solvent, and only 94% conversion was observed
after 3 h at 40 °C. However, in the presence of 0.1 equiv of Et3N,
the reaction proceeded with >99% conversion to 2 via feed-
controlled addition of SOBr2 over 2 h to a warm (40 °C) solution
of 17.25 After sequential washes with water and aqueous NaH-
CO3, the organic phase was concentrated and solvent-swapped
to heptane. The heptane phase was decolorized by filtration
through a thin pad of silica gel, concentrated to 2.5 vol and cooled
to crystallize 2. Using this 3-step process, 21 kg of bromide 2
(99.9% purity, <4 ppm residual Pd) was prepared in 82% overall
yield (Scheme 7).
Improved Synthesis of Bromide 2. Subsequent research
focused on shortening the synthesis of 2. In order to eliminate the
reduction step, direct preparation of 17 from the less expensive raw
materials 18 and 19 was investigated (Scheme 8).26 A screen of
SuzukiꢀMiyaura coupling conditions revealed that the reaction
proceeded smoothly using a Pd/PCy3 catalyst with potassium
phosphate in 2:1 THF/water at 65 °C.27 After extraction with
isopropyl acetate, charcoal treatment, and azeotropic distillation from
574
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
Scheme 9. Three-Step Synthesis of AMG 837
Scheme 10. Chemoselective Coupling to AMG 837
of 40% aqueous tetrabutylphosphonium hydroxide (n-Bu4POH) in
THF, followed by addition of 1.0 equiv bromide 2, afforded AMG
837 in 89% assay yield (Scheme 10). Under these conditions, the
alkylation was highly selective (>100:1) for the desired ether
formation over competing esterification.32 A homogeneous, single-
phase mixture was maintained throughout the process and upon
reaction completion the batch was concentrated and pH adjusted to
liberate the free acid. Although a crystalline form (mp 83 °C) of
AMG 837 free acid was identified, the molecule’s fatty-acid-like
structure made crystallization difficult, and the material exhibited
poor stability during storage.33 Consequently, formation of a phar-
maceutically acceptable salt was pursued in order to improve the
stability of the drug substance and provide a better control point for
purification.
A lysine salt of AMG 837 was initially prepared by the
Medicinal Chemistry team to enable screening and preliminary
toxicology studies, but the severe hygroscopicity of this form
precluded further development. After an intensive salt-screening
effort, a sodium salt of AMG 837 was selected as a suitable form
for early clinical development.34 Although the sodium salt was poorly
crystalline, it was significantly more stable than the free acid, could be
reproducibly prepared in high yield and purity, and could rapidly
supply the clinical program. In practice, after workup of the alkylation
step, a solution of AMG 837 in isopropyl acetate was solvent-switched
to acetonitrile and treated with aqueous NaOH. The resulting sodium
salt of AMG 837 was isolated by filtration as a crystalline acetonitrile
solvate.35 The crystallization process was effective in rejecting process
impurities (<90% pure acid could be converted to >99.5% purity
sodium salt), but vacuum oven drying the product to attain low
residual acetonitrile levels proved difficult on kilo-scale.36 Use of
higher drying temperatures (>70 °C) to accelerate desolvation was
unsuccessful and resulted in the formation of diene impurities 22 and
23 (1ꢀ2% each) from alkyne isomerization.37
through deionized water) through the cake to remove acetoni-
trile. The filter cake was stirred periodically, and the moist
nitrogen sweeps were alternated with dry nitrogen to control
the water content of the solid and prevent it from becoming
sticky and compacted.38 Using this process, the sodium salt of
AMG 837 was isolated in 88ꢀ95% yield with high chemical
purity (>99% LCAP,39 >96 wt %, 1ꢀ2% water) and contained
<400 ppm residual acetonitrile. Although the sodium salt was
suitable for this phase of development, a more crystalline, stable
and non-hygroscopic form was preferred for long-term develop-
ment. Additional final form screening identified a highly crystalline
hemicalcium salt dihydrate that possessed these quality attributes.34
However, physical characterization of the calcium salt by differential
scanning calorimetry (DSC) showed a broad endotherm centered at
85 °C due to dehydration leading to an amorphous form. The
amorphous material was less chemically stable than the crystalline
dihydrate, so understanding and control of dehydration vs amor-
phous content was considered key to the development of a robust
process.40 Attempts at direct conversion of the free acid to the calcium
salt using Ca(OH)2 or Ca(OAc)2 in a variety of solvents produced
amorphous powders. However, it was noted that in contrast to the
high solubility of the sodium salt in water (>100 mg/mL at pH 8), the
calcium salt was poorly soluble (<0.005 mg/mL), and this difference
was exploited in designing a salt-exchange crystallization process.
Thus, slow addition of an aqueous solution of calcium chloride to
a solution of the sodium salt of AMG 837 in 3:2 methanol/water
(7 vol) at 25 °C, followed by aging and filtration, provided the
hydrated hemicalcium salt wet cake (Scheme 11).41 Conveniently,
the initially isolated sodium salt MeCN-solvate could be used directly
in this salt exchange process without the need for a prior drying/
desolvation step.
In contrast, it was noted that analytical samples left open to air
for some time prior to testing showed consistently low residual
acetonitrile levels. This finding was consistent with dynamic
vapor sorption (DVS) data indicating sample weight loss at
35ꢀ55% relative humidity that corresponded to loss of acetoni-
trile from the sample and a form change (confirmed by XRPD) to
the desired unsolvated form. Integrating this process knowledge,
a mild “wet” drying approach was successfully scaled up (4ꢀ6 kg
batches) via isolation of the salt by filtration and then drying on
the filter by passing humidified nitrogen (dry nitrogen bubbled
Studies were then conducted to evaluate the impact of drying
temperature on the amorphous content of the calcium salt, as
determined by quantitative Raman spectroscopy.42 As summarized
in Table 2, the desired crystalline dihydrate form was maintained by
drying the wet cake at 25ꢀ40 °C (entries 1 and 2), whereas partial
conversion to amorphous form was observed at 60 °C (entry 3) and
complete loss of crystallinity occurred at 85 °C(entry4). Duringthe
process, bulk solvent and water were shown to be readily removed at
e40 °C to reproducibly afford crystalline AMG 837 hemicalcium
dihydrate in 89ꢀ94% overall yield (Scheme 11) with excellent
575
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
Scheme 11. Final Salt Formation
Gradient: mobile phase A, 40% MeOH, 60% water contain-
ing 0.1% perchloric acid; mobile phase B, 40% MeOH, 60%
acetonitrile
Table 2. Drying Studies on Hemicalcium Salt Dihydrate Wet
Cakea
entry
drying temp (°C)
crystallinity (%)b
water content (wt %)c
time (min)
0
% B
10
10
50
65
95
95
10
10
1
2
3
4
25
40
60
85
g97
g97
89
4.3
4.0
3.7
5.0
e3
<1.0
15.0
35.0
45.1
50.0
50.1
55.0
a 64 h drying time in vacuum oven with a nitrogen sweep. b Percentage
target polymorph determined by quantitative Raman spectroscopy; see
ref 42. c Determined by Karl Fischer titration; theoretical value for
dihydrate = 3.9 wt % water.
chemical purity (>99% LCAP, >95 wt %, KF: 3.9ꢀ4.5 wt % water)
containing <1500 ppm residual methanol.
5-(4-Hydroxy-benzylidene)-2,2-dimethyl-[1,3]-dioxane-4,
6-dione (3). 4-Hydroxybenzaldehyde (1.50 kg, 12.3 mol), Mel-
drum’s acid (2.10 kg, 14.6 mol), water (18.0 L) and toluene (2.1
L) were stirred at 23 °C for 60 min, and the internal temperature
was increased to 33 °C over 40 min and then aged for 60 min. A
reaction aliquot was filtered, and HPLC analysis of the super-
natant indicated complete reaction (<0.5 mg/mL 4-hydroxyben-
zaldehyde vs completion target of <2 mg/mL). The yellow
suspension was aged at 33 °C for an additional 2 h to promote
agglomeration of the product into spherical beads. The beads
(diameter ∼0.25 mm) were filtered and washed with water (2 ꢁ
6 L). The wet cake was dried at 50 °C under vacuum to constant
weight to provide 3 as yellow solid beads (3.02 kg, 99%, >98
’ CONCLUSIONS
Early process development, synthesis and salt selection for the
GPR40 agonist AMG 837 has been described. Improvements to
the original 12-step Medicinal Chemistry synthesis led to a
process-enabled route that was satisfactory to supply preclinical
and early clinical studies. Additional process research afforded an
improved route (9 steps, 25% overall yield) and identified a
stable, crystalline hemicalcium salt dihydrate form of the drug
substance that was suitable for long-term development.
’ EXPERIMENTAL SECTION
1
LCAP,38 >98 wt %, <0.03 wt % water). H NMR (300 MHz,
Mass spectra analyses and high-resolution mass spec analyses
were measured via fast-atom bombardment (FAB) or electro-
spray ionization (ESI). Optical rotations were recorded in cells
with 1 dm path length. Enantiomeric excess was determined by
chiral HPLC analysis. Assay yields were calculated by quantita-
tive HPLC analysis against a purified reference standard. Isolated
yields were corrected for product potency (wt % purity by
quantitative HPLC). HPLC in-process-test method: YMC Pro
C18 s-3 column (3.0 mm ꢁ 150 mm), 35 °C, flow rate 1.0 mL/
min, UV detection at 220 nm. Gradient: mobile phase A, 40%
MeOH, 60% water containing 0.1% perchloric acid; mobile
phase B, 40% MeOH, 60% acetonitrile
CDCl3) δ10.95 (s, 1H), 8.26 (s, 1H), 8.18 (d, J=8.80Hz, 2H), 6.91
(d, J = 8.80 Hz, 2H), 1.73 (s, 6H); 13C NMR (100 MHz, CDCl3) δ
26.83, 26.84, 103.90, 109.87, 115.80, 115.81, 123.03, 137.89, 137.90,
156.99, 163.35, 163.63. Anal. Calcd for C13H12O5: C, 62.90; H, 4.87.
Found: C, 63.10; H, 4.84. IR: 3262, 1745, 1696, 1571, 1274, 1194,
1038 cmꢀ1; mp 202ꢀ203 °C.
5-(1-(4-Hydroxyphenyl)but-2-ynyl)-2,2-dimethyl-1,3-di-
oxane-4,6-dione (11). To a cold (5 °C), stirred solution of 3 (1.00
kg, 4.03 mol) and N-methylmorpholine (489 g, 4.83 mol) in
anhydrous THF (5.0 L) was added acetyl chloride (348 g, 4.43 mol)
at a rate that did not allow the internal batch temperature to rise above
25 °C. After the addition, the mixture was aged at 25 °C for 30 min
(HPLC analysis showed >99% conversion to 9). The mixture was in-
line filtered (to remove N-methylmorpholine hydrochloride salt),
rinsing with anhydrous THF (2 ꢁ 1.5 L). The filtrate was cooled to
15 °C, and a solution of 1-propynylmagnesium bromide (8.9 L, 0.5 M
solution in THF, 4.45 mol) was added at a rate that did not allow the
internal batch temperature to rise above 30 °C. After the addition was
complete, the mixture was stirred at 25 °C for an additional 30 min
(HPLC analysis showed >99% conversion to 10). Aqueous 2 M
NaOH (8.0 L) was added at a rate that did not allow the internal
batch temperature to rise above 30 °C. The mixture was then heated
to 50 °C and held for 30 min (HPLC analysis showed >99%
time (min)
% B
25
50
65
95
95
25
0
10
28
35
36
37
HPLC purity method: YMC Pro C18 s-3 column (3.0 mm ꢁ
150 mm), 35 °C, flow rate 0.6 mL/min, UV detection at 220 nm.
576
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
conversion to 11). The mixture was cooled to 0 °C, and aqueous 5 M
HCl (∼4.5 L) was added at a rate that did not allow the internal batch
temperature to rise above 30 °C (final pH, 2ꢀ3). The layers were
separated, and the organic phase was concentrated (45 °C) and
solvent-swapped to toluene by distillation (∼18 L distillate collected)
until GC assay indicated <5 wt % THF in the reaction pot. The
resulting mixture in toluene (∼9L) was cooledto23°Cover2hand
stirred for an additional 2 h. The batch was filtered, and the filter cake
was washed with toluene (2 ꢁ 2 L). The filter cake was dried under
vacuum at 50 °C to afford 11 as a white solid (1.01 kg, 87% yield, 96.5
were dissolved (15ꢀ20 min.), the agitation was stopped, and the
layers were separated (pH aqueous phase, 1ꢀ2). The organic
phase was washed with water (2 ꢁ 1.0 L), dried over MgSO4,
filtered, and concentrated. The product can be solvent-swapped into
THF for the next step or further dried (vacuum oven, 30 °C) to
afford 1 as a white solid (900 g, 99% yield, >99 LCAP). 1H NMR
(300 MHz, DMSO-d6) δ 1.77 (d, J = 2.48 Hz, 3H), 2.53ꢀ2.60 (m,
2H), 3.79ꢀ3.95 (m, 1H), 6.63ꢀ6.75 (m, 2H), 7.14 (d, J = 8.48 Hz,
2H), 9.31(s, 1H), 12.21 (s, 1H); 13C NMR (100 MHz, DMSO-d6)
δ 3.21, 32.68, 42.96, 77.90, 80.83, 115.08, 115.09, 128.15, 128.16,
131.35, 156.12, 171.85. HRMS (ESI) calcd for (M þ H): 205.08592,
1
LCAP, 96.2 wt %, 3.2 wt % bis-acid 12). H NMR (400 MHz,
found 205.08609. IR: 3232, 1693, 1598, 1512, 1216, 1157 cmꢀ1
;
CDCl3) δ 7.40 (m, 2H); 6.78 (m, 2H); 4.85 (s(b), 1H); 3.82 (s,
1H); 1.88 (d, 3H, J = 2.5 Hz); 1.72 (s, 3H); 1.63 (s, 3H). For bis-acid
12: 1H NMR (300 MHz, CDCl3) δ 1.73 (d, J = 2.34, Hz, 3H), 3.50
(d, J = 10.82 Hz, 1H), 4.02 (dd, J = 10.82, 2.48 Hz, 1H), 6.67 (d, J =
8.62, Hz, 2H), 7.12 (d, J = 8.62 Hz, 2H), 9.32 (s, 1H), 12.77 (s, 2H).
(3S)-3-(4-Hydroxyphenyl)-hex-4-ynoic acid (1S,2R)-1-
Amino-2-indanol Salt (13). A solution of 11 (50.0 g, 0.173
mol) in DMF (137 mL) and water (13.7 mL) was heated to
100 °C over 30 min and then aged for an additional 60 min
(CAUTION: carbon dioxide evolution). HPLC analysis of a
reaction aliquot indicated >99% conversion. The mixture was
cooled to room temperature, and tert-butyl methyl ether (250 mL)
was added. The organic phase was washed with 50% aqueous brine
(500 mL). The aqueous phase was separated and extracted with tert-
butyl methyl ether (250 mL). The combined organic phases were
washed with 50% aqueous brine (250 mL). The organic phase was
distilled, and anhydrous acetonitrile was added until the water
content in the acetonitrile phase was <1 wt % and GC analysis
showed <1 wt % residual DMF. The resulting solution of (()-1 in
acetonitrile (∼205 mL) was added to a solution of (1S,2R)-1-amino-
2-indanol (14.3 g, 0.095 mol, 0.55 equiv) in anhydrous acetonitrile
(245 mL) at 65ꢀ70 °C over 1 h, and the mixture was aged for 3 h at
70 °C. The resulting suspension was cooled to room temperature
over 2 h, and aged for 1 h. The mixture was cooled to 0ꢀ5 °C, aged
for 1 h, and filtered, and the collected solid was washed with cold
(0ꢀ5 °C) acetonitrile (2 ꢁ 40 mL). The wet cake (68% de 13) was
charged into a 500-mL round-bottom flask. Acetonitrile (225 mL)
and water (20 mL) were charged. The mixture was heated to 70 °C
over 0.5 h and held for an additional 3 h at 70 °C. The mixture was
cooled to room temperature over 2 h and aged for 1 h. The mixture
was cooled to 0ꢀ5 °C, aged for 1 h, and filtered, and the collected
solid was washed with cold (0ꢀ5 °C) acetonitrile (2 ꢁ 60 mL). The
wet cake was dried in a vacuum oven at 50 °C for 3 days to afford 13
as a white solid (24.5 g, 40% yield, >99 LCAP, 97.8% de). Chiral
HPLC analysis: ChiralPak AD-H, 40 °C, mobile phase hexane/
isopropyl alcohol/TFA (v:v:v) 850:150:2, flow rate 1.0 mL/min, UV
detection at 226 nm, tR (min) major: 9.5, minor: 6.7; 1H NMR (300
MHz, DMSO-d6) δ 1.76 (d, J = 2.54 Hz, 3H), 2.32ꢀ2.48 (m, 2H),
2.83 (dd, J = 16.04, 2.93 Hz, 1H), 3.02 (dd, J = 16.14, 5.77 Hz, 1H),
3.84 ꢀ 3.96 (m, 1H), 4.24 (d, J = 5.28 Hz, 1H), 4.36ꢀ4.42 (m, 1H),
6.67 (d, J = 8.61 Hz, 2H), 7.12 (d, J = 8.61 Hz, 2H), 7.17ꢀ7.28 (m,
4H), 7.33ꢀ7.48 (m, 2H); 13C NMR (100 MHz, DMSO-d6) δ 3.30,
33.32, 45.43, 57.67, 71.47, 77.08, 82.05, 114.96, 114.97, 124.78,
124.79, 126.35, 127.87, 128.12, 128.13, 128.14, 132.44, 141.20,
141.28, 155.91, 174.03. Anal. Calcd for C21H23NO4: C, 71.37; H,
6.56; N, 3.96. Found: C, 71.48; H, 6.56; N, 3.95. IR: 2881, 1613,
1510, 1409, 1237, 1093 cmꢀ1; mp 202ꢀ203 °C.
[R]D = þ10.09 (c 0.11; CHCl3); mp 127 °C.
[3-(4-Trifluoromethylphenyl)phenyl]methanol (17). To a
solution of 3-(hydroxymethyl)phenylboronic acid 18 (46.3 g,
305 mmol, 1.1 equiv) in THF (400 mL) and water (200 mL)
were added K3PO4 (117.6 g, 555 mmol, 2.0 equiv), 1-chloro-
4-(trifluoromethyl)benzene 19 (50.0 g, 277 mmol, 1.0 equiv),
Pd2(dba)3 (2.54 g, 2.7 mmol, 0.01 equiv) and PCy3 (1.9 g, 6.6
mmol, 0.024 equiv). The mixture was degassed and refilled with
N2 four times and then heated at 65 °C for 20 h. The mixture was
cooled to room temperature and filtered through a pad of Celite,
washing with THF. The THF was removed via distillation at
reduced pressure, and IPAc (600 mL) was added. The phases
were separated, and the organic phase was washed with 10%
aqueous Na2CO3 (300 mL) and water (2 ꢁ 300 mL). Charcoal
(10 g) was added to the organic phase, and the mixture was
heated at 70 °C for 0.5 h. The mixture was filtered through a pad
of Celite, washing with IPAc. The IPAc solution was concen-
trated at reduced pressure and solvent-swapped to heptane
by distillation (80ꢀ85 °C). The resulting heptane solution
(∼500 mL, containing <1 wt % IPAc) was cooled to 57 °C,
seeded (1 wt %), and cooled to room temperature over 2 h, and
the resulting slurry aged for 2 h. The mixture was cooled to
0ꢀ5 °C, aged for 0.5 h, and filtered. The filter cake was washed
with cold (0ꢀ5 °C) heptane (2 ꢁ 100 mL) and dried in a
vacuum oven at 50 °C overnight to afford 17 as a white solid
(66.5 g, 95% yield, 98.3 wt %). 1H NMR (400 MHz, DMSO-d6)
δ 7.88 (2H, d), 7.81 (2H, d), 7.68 (1H, s), 7.59 (1H, d), 7.47
(1H, t), 7.39 (1H, d), 5.28 (1H, t), 4.60 (2H, d); 13C NMR (100
MHz, DMSO-d6) δ 144.3, 143.5, 138.3, 128.9, 127.4, 126.5,
125.7, 125.3, 125.0, 62.7. IR (neat): 3334, 1322, 1165, 1122,
1111, 1071, 1016, 843, 737 cmꢀ1. Anal. Calcd for C14H11F3O: C,
66.66; H, 4.40. Found: C, 66.81; H, 4.41; mp 64ꢀ66 °C.
[3-(4-Trifluoromethylphenyl)phenyl]methanol (17) by in
Situ Generated Borane Reduction of 16. To a mixture of
sodium borohydride (3.15 kg, 83.3 mol) in tetrahydrofuran (35.0
kg) was added a solution of 16 (19.7 kg, 74.0 mol) in tetrahy-
drofuran (56.0 kg) over 2 h so as to maintain the internal
temperature of the batch below 25 °C. Trimethyl borate (2.31
kg, 22.2 mol) was charged, and the mixture was heated to 44 °C.
Then 45% BF3 THF (13.2 kg, 87.6 mol) was slowly added to the
3
reactor at a rate that did not allow the batch temperature to
increase above 50 °C (addition of BF3 THF occurred over 2 h
3
with the reaction temperature between 41 and 49 °C during the
addition, and constant heat generation was monitored to avoid
accumulation of unreacted borane). The mixture was aged at
41ꢀ43 °C for 2 h, and the reaction was judged to be complete
(<0.5% starting material detected vs target: <1.0% by HPLC
analysis). The reaction was cooled to 10 °C, excess borane was
quenched by the addition of acetone (8.59 kg), and water (47.2 kg)
was added. After removal of the tetrahydrofuran by distillation under
(3S)-3-(4-Hydroxyphenyl)-hex-4-ynoic Acid (1). To a stir-
red suspension of 13 (1.57 kg, 4.44 mol) in ethyl acetate (6.0 L)
was added a solution of NaHSO4 H2O (714 g, 5.17 mol) in
3
water (9.0 L). The mixture was stirred vigorously until all solids
577
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
atmospheric pressure, toluene (51.1 kg) and water (9.9 kg) were
added, the phases were allowed to separate, and the water layer was
discarded. The upper organic layer was successively washed with a
solution of NaHCO3 (1.42 kg) in water (23.6 kg) and water (23.6
kg). The organic layer was concentrated under reduced pressure to
remove about half of the toluene (26.9 kg distillate collected), and
fresh toluene (21.3 kg) was added (water concentration of the
resultant solution was 55 ppm). The mixture was filtered (washing
with 8.5 kg toluene), and the filtrate was concentrated via distillation
to afford a clear solution of 17 (18.5 kg assay, >99 LCAP, 99% yield)
in toluene (55.4 kg) that could be used directly in the next step.
3-(Bromomethyl)-40-(trifluoromethyl)biphenyl (2). To a
solution of 17 (18.5 kg, 73.3 mol) in toluene (55.4 kg) was
charged triethylamine (0.74 kg, 7.31 mol), and the resulting solution
was heated to 35 °C. SOBr2 (16.9 kg, 81.3 mol) was slowly added at a
rate that did not allow the internal temperature of the batch to
increase above 40 °C (addition of SOBr2 occurred over 2.3 h with the
reaction temperature between 35 and 40 °C during the addition).
The mixture was aged for 0.5 h at 40 °C, and HPLC analysis
indicated that the reaction was complete (<0.1% starting material
detected vs target: <1.0% starting material). Excess SOBr2 was
quenched by addition of water (34.3 kg) and stirring for 45 min.
(Note: addition of water is exothermic. A 7 °C temperature increase
was observed during the quench). The phases were then allowed to
separate, and the lower (aqueous) layer was discarded (aqueous layer
pH < 1). The upper organic layer was washed twice with a solution of
NaHCO3 (2.05 kg) in water (34.3 kg) (aqueous layer pH after twice
washing was 7.0). The organic layer was washed with water (37.3
kg), and after phase separation the upper organic layer was con-
centrated via distillation (53.6 kg distillate collected). Heptane (31.9
kg) was added, and the mixture was heated to 50 °C. The heptane
solution was filtered through silica gel (2.9 kg), and the silica gel was
washed with heptane (25.5 kg) preheated at 50 °C. The combined
filtrate and washes were concentrated (54.4 kg distillate collected),
additional heptane (25.5 kg) was added, and the mixture was heated
to 50 °C. The resulting heptane solution was cooled to 40 °C, seeded
with 2 (0.1 wt %, 0.019 kg), and cooled to 0 °C over 5 h for
crystallization. The mixture was further aged at 0 °C for 4 h and then
filtered. The filter cake was washed with cold (0 °C) heptane (8.29
kg) and then dried at 30 °C for 8 h, to afford 2 as a white solid (20.9
concentrated to dry azeotropically. The resulting solution of AMG
837 (9.50 g assay, 89% yield) could be further concentrated to
dryness or solvent-swapped to acetonitrile for the next step.
AMG 837 Hemicalcium Salt Dihydrate. To a solution of
AMG 837 free acid (113.1 g, 0.258 mol, 97.6 LCAP) in aceto-
nitrile (250 mL) was added aqueous 5 N NaOH (51.6 mL, 0.258
mol, 1.0 equiv), and the clear yellow solution was stirred at 20 °C
for 30 min. Additional acetonitrile antisolvent (1.60 L) was
added over 2.5 h, and the resulting slurry was aged at 20 °C
for 15 h. The mixture was filtered, and the filter cake was washed
with acetonitrile (250 mL). The cake was dried on the filter
under a nitrogen stream for 20 min and then further dried in a
vacuum oven at 20 °C for 16 h to afford AMG 837 sodium salt
(MeCN solvate) as a white solid (122.4 g, 99.1 LCAP, 90.1 wt %,
containing 7.3 wt % acetonitrile, 95% corrected yield). An
analytical sample was further dried for characterization and
1
displayed: H NMR (400 MHz, DMSO-d6) δ 7.90 (d, J = 8.2
Hz, 2H), 7.82 (d, J = 8.2 Hz, 2H), 7.81 (bs, 1H), 7.69 (dt, J = 6.3,
1.9 Hz, 1H), 7.53 (t, J = 6.3 Hz, 1H), 7.52 (d, J = 6.3 Hz, 2H),
7.26 (d, J = 8.6 Hz, 2H), 6.93 (d, J = 8.6 Hz, 2H), 5.15 (s, 2H),
4.01 (m, 1H), 2.33 (dd, J = 14.7, 6.9 Hz, 1H), 2.18 (dd, J = 14.7,
7.3 Hz, 1H), 1.74 (d, J = 2.3 Hz, 3H); 13C NMR (100 MHz,
DMSO-d6) δ 174.2, 156.6, 143.9, 138.7, 138.3, 136.1, 129.3,
128.3, 127.9 (q, 2JCꢀF = 31.7 Hz), 127.6, 127.5, 127.0 (q, 1JCꢀF
=
271.9 Hz), 126.5, 126.2, 125.8 (q, 3JCꢀF = 3.8 Hz), 114.4, 83.3,
76.4, 69.0, 47.5, 33.9, 3.3. HRMS (m/z): [MH]þ calcd for
C26H20F3NaO3, 461.1335 found, 461.1326. IR (KBr): 3590,
3064, 1640ꢀ1500, 1413, 1328, 1124, 1112, 1071, 789 cmꢀ1. The
solvated AMG 837 sodium salt (120.5 g, 90.1 wt %, 0.235 mol,
1.0 equiv) obtained as described above was dissolved in methanol
(578 mL) and water (289 mL) to afford a clear solution. A
solution of calcium chloride dihydrate (17.3 g, 0.118 mol, 0.5
equiv) in water (35 mL) was added dropwise over 50 min at
20 °C, and the resulting slurry was aged for 16 h. The mixture was
filtered, and the filter cake was washed with 2:1 methanol/water
(2 ꢁ 150 mL). The cake was dried on the filter under a nitrogen
stream for 2 h and then further dried in a vacuum oven at 40 °C
with a nitrogen sweep for 36 h to afford AMG 837 hemicalcium
salt dihydrate as a white solid (121.0 g, 99.2 LCAP, 99.4 wt %,
99.6% ee, 4.6 wt % water, 1600 ppm methanol, 94% overall
yield from the free acid). Chiral HPLC analysis: ChiralPak
AD-RH (150 mm ꢁ 4.6 mm), 40 °C, mobile phase: acetoni-
trile/methanol/water/TFA (v:v:v:v) 193:193:115:1, flow rate:
1.0 mL/min., UV detection at 254 nm, tR (min) major: 9.0,
minor: 5.4.; Ion Chromatography (IC): Dionex IonPac CS12A
(250 ꢁ 4 mm), 35 °C, mobile phase: 30 mM MSA in water, flow
rate: 1.0 mL/min. Calcd: 4.2% Ca, found 4.1% Ca. 1H NMR (400
MHz, DMSO-d6) δ 7.87 (d, J = 8.2 Hz, 2H), 7.79 (d, J = 8.2 Hz,
2H), 7.77 (bs, 1H), 7.67 (bd, J = 6.8 Hz), 7.49 (m, 2H), 7.28 (d,
J = 8.6 Hz, 2H), 6.92 (d, J = 8.6 Hz, 2H), 5.11 (s, 2H), 4.05 (m,
1H), 2.44 (dd, J = 15.2, 7.0 Hz, 1H), 2.28 (dd, J = 15.2, 7.2 Hz,
1H), 1.72 (d, J = 2.2 Hz, 3H); 13C NMR (100 MHz, DMSO-d6)
δ 177.1, 156.7, 143.9, 138.7, 138.2, 135.5, 129.2, 128.4, 127.9 (q,
1
kg, 99.9 LCAP, 90% yield). H NMR (400 MHz, CDCl3) δ
7.76ꢀ7.68 (4H, m); 7.63 (1H, br s); 7.56ꢀ7.52 (1H, m);
7.50ꢀ7.43 (2H, m); 4.58 (2H, s); 13C NMR (100 MHz, CDCl3)
δ ppm 144.0, 140.4, 138.6, 129.8, 129.5, 128.7, 127.9, 127.4, 127.3,
125.8, 122.9, 33.1. IR (neat): 1327, 1265, 1166, 1122, 1112, 1070,
843, 735, 699 cmꢀ1. Anal. Calcd. for C14H10BrF3: C, 53.36; H, 3.20.
Found: C, 53.37; H, 3.22; mp 62ꢀ64 °C.
(S)-3-[4-(40-Trifluoromethyl-biphenyl-3ylmethoxy)-phenyl]-
hex-4-ynoic Acid (AMG 837). To a stirred solution of 1 (5.00 g,
24.5 mmol) in THF (35 mL) was added aqueous tetrabutylpho-
sphonium hydroxide (34.5 g, 2.04 equiv, 40 wt % aqueous solution),
and the resulting clear solution was cooled to ꢀ5 °C. A solution of 2
(7.71 g, 1.0 equiv) in THF (15 mL) was added over ∼30 min so as
to maintain the internal temperature of the batch between ꢀ5 and
0 °C. The mixture was aged for an additional 30 min and then
warmed to room temperature and aged for 20 h. The mixture was
concentrated at reduced pressure (T < 40 °C) to remove THF, and
the resulting aqueous solution was washed sequentially with hexane
(20 mL) and 1:1 MTBE/hexane (30 mL). The aqueous phase was
treated with 1 N aqueous HCl (33 mL) to adjust to pH to 2ꢀ3 and
extracted with EtOAc (100 mL). The layers were separated, and
the EtOAc phase was washed with water (50 mL) and then
2JCꢀF = 32.1 Hz), 127.6, 127.5, 126.5, 126.2, 125.7 (q, 3JCꢀF
=
4.3 Hz), 124.3 (q, 1JCꢀF = 271.4 Hz), 114.4, 82.7, 76.9, 69.0, 46.4,
33.1, 3.3. IR (KBr): 3545, 1562, 1509, 1436, 1327, 1241, 1126,
1016, 790 cmꢀ1
.
’ ASSOCIATED CONTENT
S
Supporting Information. This material is available free
b
of charge via the Internet at http://pubs.acs.org.
578
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
’ AUTHOR INFORMATION
discussion on agglomeration effects, see: Pietsch, W. Agglomeration
in Industry; Wiley-VCH: Weinheim, 2005, Vols. 1 and 2. For a
recent review of hydrophobic effects and reactions in/on
water, see: Butler, R. N.; Coyne, A. G. Chem. Rev. 2010,
110, 6302.
Corresponding Author
*Tel: 805-313-5152. Fax: 805-375-4532. E-mail: walkers@
amgen.com.
(11) Since the next step employed a Grignard reagent, a target of
<0.1% water was set.
(12) Raman spectroscopy is particularly useful here since it is
suitable for analysis of heterogeneous mixtures and is insensitive to
aqueous solvents.
(13) Bases screened to generate phenoxide 8 included: LiHMDS,
LiH, KHMDS, K2CO3, NaHMDS, NaH, and Na2CO3.
(14) This was attributed to the low solubility of the phenoxide salts
in THF. Addition of polar co-solvents (e.g., TMEDA, NMP, DMPU)
did not improve conversions. Similar results were obtained using a TMS-
ether protecting group.
’ ACKNOWLEDGMENT
The authors thank Judy Ostovic, Tiffany Correll, J. Preston,
Kyung Gahm, Yong Xie, Jan Fang and Nina Cauchon from
Analytical R&D; Jonathan Houze, Michele Akerman, Jiwen Liu,
J. Luo, Mike Schmitt, Rajiv Sharma and Julio Medina from
Medicinal Chemistry; Seth Huggins, Aashish Maheshwari, Ken
McRae, Jerad Manley, Joseph Tomaskevitch, Justin Tvetan and
Greg Sukay from Process Engineering Development; Maria Silva
Elipe, Mike Ronk and Randy Jensen for assistance with NMR
structure determinations; and Janan Jona, M. Bhupathy, Rob
Larsen, Mike Martinelli, Paul Reider and Jerry Murry for support
and encouragement.
(15) 1.0 vol = 1.0 L/kg.
(16) Solubility of N-methylmorpholine hydrochloride in THF at
room temperature is <2 mg/mL as determined via turbidity
measurements.
(17) 1,4-Addition occurred with >95% chemoselectivity with
this addition mode. However, when a solution of acetate 9
was added to a solution of propynylmagnesuim bromide, HPLC
analysis indicated a mixture containing 54% 10, 20% 11 and
16% 9.
(18) Although the saponification was typically complete in 1 h, stress
tests indicated that the mixture could be aged for 24 h without loss of
yield or purity. Apparently, formation of an enolate anion suppresses
ester hydrolysis of the malonate moiety.
’ REFERENCES
(1) (a) Medina, J. C.; Houze, J. B. Annu. Rep. Med. Chem. 2008,
43, 75. (b) Akerman, M.; Houze, J.; Lin, D. C. H.; Liu, J.; Luo, J.; Medina,
J. C.; Qiu, W.; Reagan, J. D.; Sharma, R.; Shuttleworth, S. J.; Sun, Y.;
Zhang, J.; Zhu, L. Chem. Abstr. 2005, 143, 326090; Patent WO/2005/
086661 A2, 2005.
(2) As a step to addressing this synthetic challenge we recently
developed an asymmetric conjugate alkynylation of Meldrum’s acid
derived acceptors providing practical access to chiral β-alkynyl acids:
(a) Cui, S.; Walker, S. D.; Woo, J. C. S.; Borths, C. J.; Mukherjee, H.;
Chen, M. J.; Faul, M. M. J. Am. Chem. Soc. 2010, 132, 436. For a
discussion of enantioselective approaches to AMG 837, see: (b) Woo, J.
C. S.; Cui, S.; Walker, S. D.; Faul, M. M. Tetrahedron 2010, 66, 4730.
(c) Yazaki, R.; Kumagai, N.; Shibasaki, M. Org. Lett. 2011, 13, 952. For
Cu-catalyzed alkynylation of Meldrum's acid derived acceptors with aryl
acetylenes see: (d) Kn€opfel, T. F.; Carreira, E. M. J. Am. Chem. Soc. 2003,
125, 6054. For Rh-catalyzed alkynylation of Meldrum's acid derived
acceptors with TMS-acetylenes see: (e) Fillion, E.; Zorzitto, A. K. J. Am.
Chem. Soc. 2009, 131, 14608. For recent reviews of asymmetric
conjugate alkynylation, see: (f) Fujimori, S.; Kn€opfel, T. F.; Zarotti, P.;
Ichikawa, T.; Boyall, D.; Carreira, E. M. Bull. Chem. Soc. Jpn. 2007,
80, 1635. (g) Trost, B. M.; Weiss, A. H. Adv. Synth. Catal. 2009, 351, 963.
(3) Bigi, F; Carloni, S.; Farrari, L.; Maggi, A.; Satori, G. Tetrahedron
2001, 42, 5203.
(19) The level of bis-acid in the product was dependent on the
heating time during distillative solvent switch from THF to toluene.
Small levels of bis-acid 12 are not a concern since 12 is also decarboxy-
lated to (()-1 in the next step.
(20) Dyer, U. C.; Shapland, P. D.; Tiffin, P. D. Tetrahedron Lett.
2001, 42, 1765.
(21) A 5 wt % charge of NEChemcat E-type Pd/C (50% wet
containing 5% Pd, corresponding to 0.125 wt% Pd overall)
(22) BH3-THF solutions were stored refrigerated and used shortly
after being received. The titer of individual lots was confirmed by use-
test prior to large-scale use. CAUTION: BH3-THF solutions are known
to decompose over time, particularly at room temperature and above.
Thus, lots shipped or stored for extended periods at room temperature
may have low borane titers and can accumulate pressure. See safety
highlights in(a) Atkins, W. J., Jr.; Burkhardt, E. R.; Matos, K. Org. Process
Res. Dev. 2006, 10, 1292. (b) Potyen, M.; Josyula, K. V. B.; Schuck, M.;
Lu, S.; Gao, P.; Hewitt, C. Org. Process Res. Dev. 2007, 11, 210. (c)
Guercio, G.; Manzo, A. M.; Goodyear, M.; Bacchi, S.; Curti, S.; Provera,
S. Org. Process Res. Dev. 2009, 13, 489.
(4) Miyashita, N.; Yoshikoshi, A.; Grieco, P. A. J. Org. Chem. 1977,
42, 3772.
(23) Lobben, P. C.; Leung, S.S-W.; Tummala, S. Org. Process Res. Dev
2004, 8, 1072. Initial exotherm results from deprotonation of acid 16 to
form an acyloxyborane intermediate that accumulates at low tempera-
ture. The second exotherm during warming can be explained by the
sequential reduction of this acyloxyborane intermediate to a trialkoxy-
boroxine which librates 17 upon aqueous workup.
(5) This presumably results from polymerization of DHP;
see: (a) Burrows, R. C. Polym. Prepr. 1965, 6, 600. (b) Kuethe, J. T.;
Tellers, D. M.; Weissman, S. A.; Yasuda, N. Org. Process Res. Dev. 2009,
13, 471.
(6) The efficiency of the classical resolution step was highly depen-
dent on the purity of the starting racemic acid. Consequently, a target
purity of >95% was established.
(7) HPLC analysis indicated that <5% of Meldrum’s acid hydrolyzed
in water at 32°C after 2 h. In contrast, 48% hydrolyzed at 50 °C and
>99% hydrolyzed at 79 °C over the same time period. Meldrum’s acid is
known to decompose to acetone and malonic acid in water: Kaczvinsky,
J. R., Jr.; Read, S. A. J. Chromatogr. 1992, 575, 177.
(8) HPLC analysis indicated that 11% of 3 hydrolyzed in 1:1
acetone/water at room temperature after 1 h and 85% hydrolyzed
after 16 h.
(9) Reaction conversion in THF could be increased to 75% by
adding 10 mol% DMAP to the mixture.
(10) Product 3 readily agglomerated into beads in water that
contained 10ꢀ20% vol of toluene. Agglomeration was not efficient
in water that contained more than 20% or less than 5% toluene. For a
(24) In the absence of B(OMe)3, thick gels were obtained which
were difficult to stir. We postulate that the in situ borane reduction may
produce a polymeric product such as A that is responsible for the gel.
Thus, the role of B(OMe)3 may be to cleave the polymeric chain to
smaller fragments by disproportionation. Although B(OMe)3 has been
reported to accelerate borane reductions of carboxylic acids (e.g., see:
Lane, C. F.; Myatt, H. L.; Daniels, J.; Hopps, H. B J. Org. Chem. 1974,
579
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580

Organic Process Research & Development
ARTICLE
39, 3052), in this case the reduction rates were similar in the presence or
absence of B(OMe)3.
(35) Crystallization studies using focused beam reflectance measure-
ments (FBRM, Lastentec) indicated significant particle attrition under
high sheer agitation resulting in increased fines and poor filtration rates.
To ensure good filtration on scale, the agitation should be reduced to
maintain just-suspended solids after complete addition of NaOH and
acetonitrile antisolvent.
(36) Upon oven drying, initial acetonitrile levels in the wet cakes
(typically 20ꢀ30 wt %) rapidly decreased but remained steady at ∼8 wt
%, consistent with the theoretical monosolvate value.
(37) Analytical samples of the conjugated diene degradation pro-
ducts were isolated by HPLC and their structures determined by 1D and
2D NMR analyses and HRMS. Related thermal and base-mediated
prototropic rearrangements of acetylenic compounds to conjugated
dienes have been reported: Jones, E. R. H.; Mansfield, E. H.; Whiting,
M. C. J. Chem. Soc. 1954, 3208. For a review see:Iwai, I. Mechanisms of
Molecular Migrations; Interscience: New York, 1969; Vol. 2.
(38) The sodium salt is hygroscopic above 55% relative humidity
(RH) with >30 wt% gain up to 95% RH by vapor sorption analysis. Static
drying times on an Aurora filter varied from 2 days to >1 week depending
on the batch and filter size. Bench-top modeling of unit operations
(50ꢀ500 g scale pilots), indicated drying cycle times could be reduced
using a rotary (<50 h) or fluid bed dryer (<5 h).
(25) The reactor was vented to a scrubber containing 5 N aqueous
NaOH to neutralize HBr and SO2 off-gases.
(26) Relative Aldrich catalogue prices are illustrative. Aryl halides:
14, $1.7/g (25 g bottle); 19, $0.13/g (100 g bottle). Boronic acids: 15,
$30/g (5 g bottle); 18, $14/g (10 g bottle).
(27) For a lead reference on Pd/PCy3-catalyzed SMC reactions, see:
Kudo, N.; Perseghini, M.; Fu, G. C. Angew. Chem., Int. Ed. 2006,
45, 1282.
(28) Alkylation reactions evaluated in THF, EtOAc or MTBE
solvents resulted in slower reaction rates and incomplete conversions.
The rate of the alkylation reaction in acetone (2ꢀ20 h for >99%
conversion) was dependent on the particle size of the K2CO3 employed.
Granular base was typically employed but faster reactions occurred with
finer powder.
(29) The main degradation pathway is via oxidative cleavage of the
alkyne moiety. A pure sample of the largest degradant was obtained by
flash chromatography, and the structure proposed on the basis of 1D
(1H, 13C) and 2D (HMBC) NMR experiments (CDCl3) is a mixture of
keto-ester (91%) and its enol tautomer (∼9%) and is consistent with the
measured LCMS Mþ (428) value. For example, when a crude sample of
neat 21 was allowed to stand open to air at room temperature for 17 h,
5.1% keto-ester was generated (HPLC assay).
(39) LCAP = liquid chromatography area percent.
(40) For a report describing online humidity analysis during factory
drying of a hydrate, see: Cypes, S. H.; Wenslow, R. M.; Thomas, S. M.;
Chen, A. M.; Dorwart, J. G.; Corte, J. R.; Kaba, M. Org. Process Res. Dev.
2004, 8, 576.
(41) Typically, the hemicalcium salt wet cake contained ∼14 wt%
water (by KF titration) before drying.
(42) A multivariate curve resolution (MCR) method based on
second derivative Raman measurements of the pure crystalline and
amorphous forms was used. For full details, see: Xie, Y.; Tao, W.;
Morrison, H.; Chiu, R.; Jona, J.; Fang, J.; Cauchon, N. Int. J. Pharm.
2008, 362, 29.
(30) The bromide 2 also has high solubility in many organic solvents
including the subsequent crystallization solvents and was not detected in
the API by HPLC assay.
(31) At least half of the initial ethanol charge was distilled to prevent
formation of an emulsion during the isopropyl acetate extraction.
(32) Details and scope of this method have been reported elsewhere:
Liu, P.; Huang, L.; Faul, M. M. Tetrahedron Lett. 2007, 48, 7380.
(33) Oxidative degradation (presumably via the intermediacy of a
keto-acid) led to accumulation of a methyl ketone as the largest single
impurity in aged samples of AMG 837. For example, when solid AMG
837 free acid was aged at 44 °C overnight (open air), ∼1% methyl
ketone was detected by HPLC analysis. A solution of AMG 837 in
acetonitrile (5 vol) at room temperature for 1 week accumulated 4%
methyl ketone.
(34) For a detailed discussion of salt screening for AMG 837, see:
Morrison, H.; Jona, J.; Walker, S. D.; Woo, J. C. S.; Li, L.; Fang, J. Org.
Process Res. Dev. 2011, 15, 104.
580
dx.doi.org/10.1021/op1003055 |Org. Process Res. Dev. 2011, 15, 570–580