Synthesis of an H3 antagonist via sequential one-pot additions of a magnesium ate complex and an amine to a 1,4-ketoester followed by carbonyl-directed fluoride addition

Article abstract of DOI:10.1021/op300093j

We describe the development of an efficient and scalable process for the preparation of fluorocyclobutane-containing H₃ antagonist, 1. The synthesis was accomplished by the chemoselective addition of a magnesium ate complex and an amine to a 1,4-ketoester in a one-pot sequence, followed by a diastereoselective carbonyl-directed fluorination. The chemoselective addition of the magnesium ate complex to the ketoester benefited from tight stoichiometric control, short addition times, and lower reaction temperatures, and thus was amenable to rapid mixing and excellent heat transfer in a flow reactor.

Full text of DOI:10.1021/op300093j

Article
pubs.acs.org/OPRD
Synthesis of an H3 Antagonist via Sequential One-Pot Additions of a
Magnesium Ate Complex and an Amine to a 1,4-Ketoester followed
by Carbonyl-Directed Fluoride Addition
Joel M. Hawkins,* Pascal Dube, Mark T. Maloney, Lulin Wei, Marcus Ewing, Stephen M. Chesnut,
́
Joshua R. Denette, Brett M. Lillie, and Rajappa Vaidyanathan*,†
Pharmaceutical Sciences, Pfizer Inc., Eastern Point Road, Groton, Connecticut 06340, United States
ABSTRACT: We describe the development of an efficient and scalable process for the preparation of fluorocyclobutane-
containing H3 antagonist, 1. The synthesis was accomplished by the chemoselective addition of a magnesium ate complex and an
amine to a 1,4-ketoester in a one-pot sequence, followed by a diastereoselective carbonyl-directed fluorination. The
chemoselective addition of the magnesium ate complex to the ketoester benefited from tight stoichiometric control, short
addition times, and lower reaction temperatures, and thus was amenable to rapid mixing and excellent heat transfer in a flow
reactor.
isomer (3) were determined. COSMO-RS4 calculations at BP/
INTRODUCTION
■
TZVP/COSMO level of theory in THF using Turbomole5 and
H3 receptors, G protein-coupled receptors which decrease the
COSMOtherm6 software packages revealed that the desired cis
Ca2+ influx into cells, function in a wide variety of tissues as
isomer was more stable than trans isomer 3, but by only 0.245
feedback inhibitors of histamine and other neurotransmitters.1
kcal/mol.7 Furthermore, base-catalyzed equilibration of a
H3 receptor agonists cause sedation by opposing H1-induced
sample of 1 led to an approximately 1:1 mixture of the
wakefulness,2 and H3 receptor antagonists have the potential to
deuterated analogues of 2 and 3 (2-D and 3-D), as observed by
counteract the inhibitory effect of histamine and increase its
presynaptic release. Compound 1 was discovered and
developed within Pfizer as a potent H3 antagonist for multiple
19F NMR spectroscopy (Scheme 1). These results precluded
the use of a thermodynamic stereocontrol strategy, and left the
kinetic control of stereochemistry as the only viable option.
Several approaches to achieve the desired cis disposition of
the carbon substituents on the cyclobutane ring via kinetic
indications related to neurological disorders including excessive
daytime sleepiness and Alzheimer’s disease.3 Herein we report
the identification and development of a scalable route to this
drug candidate.
control were considered (Scheme 2). In principle, this could be
accomplished either during the formation of the cyclobutane
ring, or during the installation of the substituents on the
cyclobutane ring. Control of stereochemistry during the
Scheme 1. Base-catalyzed equilibration of 1
RETROSYNTHETIC ANALYSIS
■
The synthesis of 1 requires control of the relative stereo-
chemistry on the fluorocyclobutane ring, where the amide and
aryl groups are cis to one another. Four general possibilities to
prepare the cyclobutane with desired stereochemistry were
considered: (1) separate and discard the undesired isomer, (2)
separate and epimerize the undesired isomer for recycle, (3)
kinetic stereocontrol, and (4) thermodynamic stereocontrol.
The cost and wastefulness of a late-stage diastereomer
separation and the additional unit operations required for
isomer recycle disfavored the first two options. In order to
determine whether thermodynamic stereocontrol was possible,
the relative energies of free base 2 and the corresponding trans
Received: April 12, 2012
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 2. Analysis of potential methods for kinetic stereocontrol on the cyclobutane ring
formation of the cyclobutane ring by, for instance, a [2 + 2]
addition approach (strategy ‘a’, Scheme 2) seemed unlikely.
Installation of either of the two carbon-bearing substituents on
the cyclobutane ring, i.e. formation of the cyclobutane−arene
C−C bond, or the cyclobutane−carbonyl C−C bond
(strategies ‘b’ and ‘c’, respectively) were also deemed unlikely
to give the necessary kinetic stereocontrol. Stereoselective
kinetic protonation to form the C−H bond (strategy ‘d’) was
considered as a possibility, but calculations predicted that
protonation would occur from the undesired face. Stereo-
selective C−F bond formation (strategy ‘e’) appeared viable,
especially if the carbonyl oxygen could internally ligate to the
carbocation generated by solvolysis of the corresponding
alcohol or activated alcohol as shown in Scheme 2.
Scheme 3. Stereoselective fluorination of alcohol
diastereomers 4 and 5 favoring fluoride isomer 2
The choice of strategy ‘e’ was further supported by two
observations. First, fluorination of either of the alcohol
diastereomers (4 or 5) with bis(2-methoxyethyl)amino sulfur
trifluoride (Deoxo-Fluor) furnished the desired fluorinated
cyclobutane 2 as the predominant diastereomer (Scheme 3).
Second, hydrolysis of free base 2 by extended exposure to
aqueous media gave the alcohol with retention of stereo-
chemistry, i.e., the water approaches the carbocation from the
same face as the fluoride, opposite the amide carbonyl (Scheme
4). This directed us to the bond construction strategy outlined
in Scheme 5, wherein 2 is synthesized via aryl anion addition to
an appropriately substituted cyclobutanone, followed by
fluorination of the resulting alcohol. Further efforts therefore
focused on identifying the optimal carbonyl substituent Y,
metal M, and fluorinating agent X-F, as depicted in Scheme 5.
butanone may react under Lewis acid catalysis.8 The organo-
lithium species (8) was not pursued due to its likely need for
cryogenic conditions.9 Thus, further efforts focused on
organomagnesium reagents, both the conventional Grignard
reagent (10) and other, more reactive, organomagnesium
species such as ate complex 9.10,11
The conventional Grignard reagent (10) could be readily
prepared from arylbromide 14 and magnesium turnings, but it
appeared to lose potency at room temperature over time.
Aliquots quenched with deuterated methanol showed signifi-
cantly decreased ArD/ArH ratios over one week at ambient
CHOICE OF ARYL NUCLEOPHILE
■
A range of aryl organometallic compounds was considered for
the synthesis of 1 (Figure 1). The organo-zinc (11), -aluminum
(12), and -boron (13) compounds were not pursued due to
their anticipated low reactivity, although the strained cyclo-
B
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
only gave partial conversion of aryl bromide 14 to its anion as
judged by quenching aliquots of the reaction with benzalde-
hyde. In contrast, magnesium ate complex 9 could be cleanly
generated by the dose-controlled addition of 1/3 equiv of i-
PrMgCl and 2/3 equiv of n-BuLi at 0 °C (Scheme 6). The
Scheme 4. Solvolysis of fluoride 2 with carbonyl-directed
addition of water
Scheme 6
Scheme 5. Preferred general approach to the stereoselective
synthesis of 2
addition of this reagent to various cyclobutanone coupling
partners was studied in detail (vide infra). All three of the aryl
groups of complex 9 are active nucleophiles, and typically 0.33
equiv of 9 was utilized as described below. Considering that
magnesium ate complexes like 9 are relatively recent additions
to the long history of organomagnesium reagents, the stability
of 9 was examined in order to ensure that it would be a viable,
scalable reagent.
In these experiments, stock solutions of 9 in THF, 2-
MeTHF, and MTBE were stirred under nitrogen at 0, 10, and
20 °C in an Anachem robotic workstation.16 These nine
reactions were run simultaneously, and the solutions were
automatically sampled, quenched with deuterated methanol,
and analyzed by HPLC and mass spectrometry. Profiles of this
data showed that 9 was stable in THF at 0 °C, losing only
about 1% potency in 24 h. The reagent was less stable at higher
temperatures or in the other solvents, losing 3−10% potency
over 24 h, primarily due to the formation of the butylated
impurity 15 (Figure 2) via reaction of 9 with n-butylbromide.
These studies confirmed the stability of 9 as required for it to
be a scalable reagent and favored the use of THF over 2-
MeTHF or MTBE.
temperature under nitrogen in THF, while the stability was
much improved when maintained at 0−5 °C. This suggested
that the Grignard reagent could not be readily stored or
purchased. It would need to be generated in-house, where the
use of magnesium turnings is not preferred.12 Given its
apparent instability at ambient temperature, the conventional
Grignard reagent was not evaluated further.
The formation of Grignard-type species under homogeneous
conditions, thus avoiding magnesium turnings, was pursued
with the successively more reactive reagents i-PrMgCl,13 i-
PrMgCl·LiCl,14 and 1:2 i-PrMgCl/n-BuLi.15 Of these, i-
PrMgCl (2 M in THF) and i-PrMgCl·LiCl (1.3 M in THF)
THE COUPLING STEP
■
It was envisioned that, in its simplest manifestation, the Y group
in the cyclobutanone coupling partner 7 would be an
ethylamino moiety (16, Scheme 7). However, treatment of
magnesium ate complex 9 with the anion of 16 (generated by
Figure 1. Range of potential aryl organometallic nucleophiles.
C
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
was treated with 9. The reaction proceeded at ambient
temperature, remained homogeneous, and provided adduct
19 in situ. The addition of ethylamine and amide coupling
agent T3P to the reaction mixture furnished 5 in 45% overall
yield from 17 (Scheme 8). The main byproduct of the arylation
reaction was des-bromoarene 20, which was formed at ∼25−30
area % according to HPLC. Interestingly, when aliquots of the
arylation reaction were quenched with CH3OD at different
stages of the reaction, the amount of deuterium incorporation
in the des-bromoarene (i.e., the ratio of 20-D:20) decreased
from 0.85:1 at the beginning of the reaction to 0.19:1 at the end
of the reaction. This suggested that some of the des-
bromoarene forms early in the reaction from enolization
(giving 20) of the ketone with a significant amount of 9
remaining (which gives 20-D upon quenching with CH3OD).
Stability tests of the THF solution of magnesium carboxylate
18 showed that homogeneity is maintained for 24 h at 0 °C, yet
precipitation occurred after 15−20 h at 20 °C. This
precipitation at higher rather than lower temperatures
suggested that the precipitation is due to side reactions of the
magnesium carboxylate rather than supersaturation. The
moderate yields, significant des-bromoarene formation, stability
concerns with the magnesium carboxylate solution, and the
need for coupling reagents in the subsequent amidation step led
us to abandon the magnesium carboxylate route in favor of the
keto ester route (Y = OR in 7).
Reaction of 9 with methyl ester 21 proved capricious, leading
to multiple byproducts in addition to the desired adduct, 22.
However, initial small-scale experiments with isopropyl ester 23
at −20 °C suggested that the addition of 9 was cleaner than
with 21, giving an adduct/ArH (24:20) ratio of 5:1 with very
little of the multiple addition products, 25 and 26 (Schemes 9
and 10). Since these initial results with the isopropyl ester were
promising, this reaction was explored in further detail in an
effort to identify the best conditions to effect the trans-
formation. Specifically, the reaction temperature, and the rate
and order of addition were investigated with the goal of
maximizing the product/ArH ratio, minimizing the formation
of multiple addition products (25 and 26), and avoiding the use
of conditions that would require cryogenic tanks for scale-up.
On a small scale, isopropyl ester 23 could be added in just a
few minutes while maintaining the desired reaction temperature
of −20 °C. When scaling up to a 500 mL RC-1 reactor, 23 was
added to a −10 °C solution of 9 over 1 h in order to simulate
Figure 2. Butylated impurity.
Scheme 7
reaction of 16 with i-PrMgCl in THF at −60 °C) did not give
any detectable adduct 4 or its epimer 5, presumably due to the
insolubility of the anion of 16.17 Therefore, further efforts
focused on identifying an appropriate ester or carboxylate
coupling partner.
The ideal Y group in cyclobutanone coupling partner 7
would be aprotic so as not to quench an equivalent of aryl
organometallic 9. Moreover, it would have to be resistant to
aryl anion addition during addition of the aryl anion to the
ketone functionality, and yet be receptive to ethylamine
addition during the amidation step, preferably without the
need for additional coupling or activating reagents. These
seemingly conflicting chemoselectivity requirements for the
addition of organometallic and amine reagents were pursued
with carboxylate and ester groups for C(O)Y in 7 (Scheme 5),
where the ester has the added advantage of not requiring a
carboxyl-activating step during the amidation.
Davies showed that Grignard reagents can be added to the
ketones of ketoacids if the acids are first converted to their
magnesium salts by treatment with Et3N and MgCl2 in THF,
followed by filtration of Et3N·HCl.18 Accordingly, treatment of
acid 17 with MgCl2 and Et3N furnished carboxylate 18, which
Scheme 8
D
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(addition of 23 to 9) implies that the activation energies for the
formation of 25 and 26 are larger than that of the desired
reaction, i.e., lower reaction temperatures result in a significant
reduction in the relative rates of undesired to desired reactions.
Also, the order of addition of the reagents affects the relative
concentration of each species during the addition, which then
affects possible side reactions. In the normal addition mode
(addition of 23 to 9 where intermediates are exposed to excess
9 during the addition), the ester moiety of the alkoxide of 24
can react with 9 giving multiple arene adducts (25 and 26) if
the addition is too slow or too warm. Similarly, in the inverse
addition mode (addition of 9 to 23 where intermediates are
exposed to excess 23 during the addition), the alkoxide of 24
can act as a base and enolize 23, leading to arene 20 if the
addition is too slow or too warm. A third option, simultaneous
addition of 23 and 9, could serve to minimize side reactions by
maintaining a stoichiometric balance between the reacting
species 23 and 9 and was also investigated (vide infra).
The long (1 h) addition time required to absorb the heat of
reaction in a noncryogenic tank seemed problematic with both
addition orders. In a noncryogenic tank, speeding up the
addition would require an even warmer reaction temperature
(to increase ΔT between the reactor and the jacket, and
therefore, the rate of heat transfer), while a lower reaction
temperature would require a longer addition time (to absorb
the heat of reaction with a lower ΔT relative to the jacket).
This quandary required one of two solutions: a cryogenic tank
to lower the reaction temperature so that the desired reaction is
favored, or a reactor configuration that provides stoichiometric
balance of 9 and 23 during the entire reaction along with
efficient heat removal, such as that of a plug flow reactor.
Scheme 9
heat transfer in a typical noncryogenic 1600 L reactor with a
−25 °C cooling jacket.19 These conditions provided a 2:1 ratio
of 24 to 20, but the predominant product of the reaction was
triple addition product 26 (∼30 area %). IR data acquired
during this experiment indicated that ketoester 23 began to
accumulate after one-third of 23 had been added, implying that
all of the organometallic reagent had been consumed by that
point, which is consistent with the formation of the triple
addition product, 26. Reversing the addition order (i.e., adding
a solution of 9 to a solution of 23) at −10 °C decreased the
amount of triple adduct 26 to ∼1−2 area %, but increased the
amount of protonated arene 20 to ∼50 area %.
Some of the possible side reaction pathways leading to 20,
25, and 26 are shown in Scheme 10. The strong temperature
dependence of the selectivity under normal addition conditions
Scheme 10
E
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 3. Kilo Lab flow chemistry rig.
The cryogenic option proved successful: a 40-min normal
addition (addition of 23 to 9) was well tolerated at −37 °C,
providing a 4:1 ratio of 24:20, and no 26 was detected under
these conditions. These conditions could be accommodated in
a cryogenic tank, and provided us with a reasonable option for
future scale-up. Nevertheless, we wanted to avoid the use of
cryogenic conditions so that this reaction could be carried out
in any of our scale-up facilities.
As a convenient, first-pass approximation to the stoichio-
metric balance provided by plug flow, solutions of 9 and 23
were added simultaneously into a batch reactor with
appropriate flow rates to maintain the desired stoichiometry
(a 0.33:1.05 molar ratio of 9 to 23). At a −10 °C reaction
temperature and a 1-h addition time, excellent results were
obtained, with a 24:20 ratio of 4.5:1 and <2% total multiple
addition products. No improvement in selectivity was seen for
simultaneous addition at −30 °C.
mL/min, the residence time would be 6−7 min. The reaction
was quenched by flowing into an agitated tank of methanol.
All experiments were run with 20 mL/min of a solution of 9
in THF (0.17 M) and 10 mL/min of a 1.05 M solution of 23 in
THF, while the reaction temperature was varied from −25 to 0
°C. These experiments indicated that for temperatures as low as
−16 °C, the reaction was complete (as evidenced by the
absence of 20-D when quenched with deuterated methanol)
with the standard flow rates, which correspond to a 6−7 min
residence time. Excellent selectivity was obtained for all
conditions, with the ratio of 24:20 ranging from 3.5−4.5:1
while the level of multiple adducts was 2−3%. At −25 °C, a
small amount of unreacted 9 (∼5%) was seen in the reactor
outlet sample that was quenched with deuterated methanol,
indicating that the reaction was not complete. Overall, these
results indicate that this reaction can be run in flow as warm as
0 °C with acceptable selectivity.
The success of the simultaneous addition protocol led us to
examine the reaction further in a continuous plug flow reactor.
Initial screening was done in a Conjure segmented flow reactor
to demonstrate proof of concept, and the reaction selectivity
was seen to be comparable to the simultaneous addition
conditions (vide supra). This was then successfully scaled up to
a 12-g scale at −5 °C in a Vapourtec reactor system to provide
a 65% in situ yield of 24 (with a ∼4:1 ratio of 24:20).
On the basis of this positive outcome, we scaled up this
reaction in our Kilo Lab flow chemistry rig (Figure 3). In this
configuration, solutions of 9 (0.17 M in THF) and 23 (1.05 M
in THF) were precooled independently before being combined
in a 9-mL static mixer and then flowed through a residence
time coil. The relative flow rates of the two solution streams
were adjusted to attain the desired stoichiometry (0.33
equiv:1.05 equiv of 9:23). The total flow volume for the
system, about 260 mL, was estimated by charging a slug of
colored solution to the organometallic (9) feed tanks and
timing its observed exit from the system. The volume of the
system from the reactor inlet to outlet (static mixer + residence
time coil) was about 190 mL, so that for a total flow rate of 30
AMIDATION AND WORKUP
■
While the reaction conditions were being optimized, con-
current work focused on streamlining the conversion of the
ester (24) to the ethyl amide (5), and its ultimate isolation and
purification. Preliminary experiments involved addition of a
solution of ethylamine in methanol to the reaction mixture at
the end of the Grignard reaction, and heating to 80 °C to
provide the two diastereomers, 5 and 4, in a 7.4:1 ratio. It was
subsequently demonstrated that a simple methanol quench of
the Grignard reaction mixture readily provided the methyl
ester, 22. Treatment of this quenched solution with ethylamine
in methanol furnished a 24:1 ratio of diastereomers (5:4). The
direct amidation of the isopropyl ester in methanol proceeds via
transesterification to methyl ester 22, and may be facilitated by
assistance from the neighboring alkoxide, or via the
intermediacy of a strained lactone (Scheme 11).20
During early development, the procedure to crystallize and
isolate 5 involved a long series of operations with multiple
distillations and extractions. Once the reaction conditions were
F
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
volumetric solvent ratio of 1:1:0.2 methanol/n-heptane/water.
Unfortunately, the addition of water resulted in slow
extractions, and thus we reverted to the simpler methanol/n-
heptane system while suffering only a slight decrease in
selectivity.
Scheme 11
In the optimized isolation procedure, the ethyl amide 5
product solution was distilled to replace methanol with i-
PrOAc, and then a citric acid solution was added to dissolve
salts associated with the Grignard reaction. Since the improved
reaction conditions gave higher selectivity (increased 24:20),
extraction of the methanol solution with n-heptane was not
required prior to distillation. The pH was adjusted to >10 with
aqueous NaOH, and then the phases were separated. The
organic (i-PrOAc) phase was then concentrated and seeded.
Addition of n-heptane followed by cooling to 0 °C led to
crystallization of the product. These streamlined workup and
isolation procedures resulted in a >50% reduction in cycle times
and processing costs for the production of 5.
The crystallization conditions for 5 were optimized to
provide efficient purification while integrating smoothly into
the preceding workup steps. As a result, the early procedure
using the toluene/n-heptane mixture was abandoned in favor of
the i-PrOAc/n-heptane system. When the coupling of 9 with 23
proceeds normally (with a 24:20 ratio of 3−5), ethyl amide 5 is
typically isolated in 98 area % purity with <1% 20, and no
improvement in quality is seen when n-heptane extractions are
performed on the methanol product solution prior to
subsequent workup. For cases when the selectivity in the
Grignard coupling is significantly lower, n-heptane extractions
of the methanol product solution (as described above) may be
needed to decrease the level of 20 prior to crystallization and
isolation.
established for the use of isopropyl ester 24, a concerted effort
was made to streamline the isolation of 5. A comparison of the
early and optimized procedures is shown in Figure 4.
THE FLUORINATION STEP
■
The original scale-up conditions for the fluorination reaction
involved treatment of 5 with 1.5−1.8 equiv of Deoxo-Fluor
(27) at −78 °C in methylene chloride or THF to furnish 2.39
While the ratio of diastereomers (2:3) was ∼8:1, the need to
use, and subsequently quench, an excess of Deoxo-Fluor,
coupled with the need for cryogenic equipment prompted us to
explore alternative conditions. Several deoxofluorination
conditions were evaluated for the conversion of tertiary alcohol
5 to the free-base form of API, 2. These included not only other
fluorinating reagents but also involved attempts to render the
Deoxo-Fluor amenable to scale-up in noncryogenic vessels. In
the initial stages of the screening exercise, it became apparent
that the main byproducts of the fluorination were diastereomer
3 and olefin 28. These two impurities were tracked throughout
the screen, and process development efforts focused on
minimizing, if not eliminating, the formation of these
byproducts (Scheme 12). The three most promising leads
from the initial screen (apart from Deoxo-Fluor) were 2,2-
difluoro-1,3-dimethylimidazolidine (DFI, 29), XtalFluor-E
(30), and N,N,N′,N′-tetramethylfluoroformamidinium hexa-
fluorophosphate (TFFH, 31). A comparison of the best ratios
of the desired product 2 and impurities 3 and 28 formed under
optimized conditions using each of these reagents is presented
in Figure 5.
Figure 4. Comparison of old and improved workup procedures to
isolate 5.
In the original procedure, methanol was removed from the
product solution via distillation, and then the mixture was
quenched with a citric acid quench to dissolve the magnesium
and lithium salts. The mixture was extracted with n-heptane to
remove nonpolar impurities, and the product was extracted into
i-PrOAc at high pH. Next, i-PrOAc was replaced with
acetonitrile, and the mixture was extracted with n-heptane in
order to remove the primary impurity, 20. Finally, solvent
exchange of the product-containing acetonitrile solution to
toluene, followed by the addition of n-heptane, led to
crystallization of 5.
The acetonitrile/n-heptane extraction was especially war-
ranted if selectivity was poor during formation of 24, resulting
in levels of 20 that were too high to purge effectively in the final
crystallization. Further investigation of this extraction revealed
that either methanol or acetonitrile product solutions could be
extracted with n-heptane to selectively remove 20. In addition,
a more selective extraction could be obtained if a third
component, water, is added. Best results were obtained with a
The reaction of 5 with DFI,21 a fuming liquid, furnished 2 in
∼60% yield at −20 °C. However, the reaction mixture quickly
degraded at temperatures near 0 °C. The other issue was that
the urea byproduct could only be partially purged with an
aqueous workup and isolation. These factors, coupled with the
G
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 12
lack of a long-term supplier of DFI prompted us to halt further
development using this reagent.
XtalFluor-E and its analogues are a series of novel reagents
for the deoxofluorination of alcohols and carboxyl groups.22 In
contrast to the Deoxo-Fluor and DFI, these compounds are
crystalline solids. These mild reagents require buffered HF
additives to enable clean reactions with alcohols. In our
investigations, Et3N·2HF23 proved to be the best additive for
our specific alcohol, 5.24 The reaction proceeded smoothly in
THF at 0 °C, providing a mixture of 2 and its diastereomer, 3,
in a 5:1 ratio in ∼60% yield. Attempts to improve this isomeric
ratio by performing the transformation at lower temperatures
failed due to the solidification of the reaction mixture at −15
°C.
TFFH (31) is a nonhygroscopic, crystalline solid used for
peptide synthesis via acylfluoride formation.25 We recently
reported the use of TFFH in deoxofluorination reactions.26
Following the basic reactivity pattern exhibited by the XtalFluor
reagents, our initial focus was to evaluate the impact of HF
additives on the reaction of TFFH with 5. This screen further
Figure 5. A comparison of different deoxofluorinating agents for the
conversion of 5 to 2.
Scheme 13. Optimized process for the synthesis of 1
H
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
reinforced the unique ability of Et3N·2HF to promote the
desired deoxofluorination. Moreover, THF was found to be the
optimal solvent for this transformation and yields up to 60%
were obtained on small scale, with a dr of 5:1 (Figure 5).
Analogous to the DFI reaction, the presence of large quantities
of the urea byproduct proved problematic in this case and led
us to halt our efforts.
Concurrent examination of the Deoxo-Fluor-mediated
deoxofluorination reaction revealed that a lower stoichiometry
of Deoxo-Fluor (<1.5 equiv) could be used when the incoming
5 was of high purity. Moreover, identical impurity profiles and
isomeric ratios (8:1 ratio of 2:3) could be obtained when
performing the reaction at or below −40 °C. Importantly, it
was discovered that addition of a THF solution of 5 to Deoxo-
Fluor allowed us to carry out the reaction at 0 °C, while still
providing similar yields and isomeric ratios. Reaction kinetics
revealed that the reaction of 5 with Deoxo-Fluor was dose
limited and the substrate could be added over 10 min to 3 h
with no loss in conversion or selectivity at a reaction
temperature of 0 °C. When reaction was deemed complete, it
could be held for up to 48 h at 0 °C with no loss in selectivity
or purity. Ultimately, this procedure was streamlined to a point
where the transformation could be carried out with just 1.1
equiv of Deoxo-Fluor at 0 °C.
Upon reaction completion, the reaction mixture was
quenched via subsurface addition into a mixture of aqueous
30% w/w potassium carbonate (9 equiv) and 2-methyltetrahy-
drofuran, at −10−0 °C. Adequate mixing during the quench
was important to prevent localized accumulation of hydro-
fluoric acid, which could degrade the product. Experimental
evidence suggested that a final aqueous pH 10−11 was ideal to
ensure that all of the hydrofluoric acid had been quenched and
converted to potassium fluoride and to ensure that the desired
product resided in the organic layer. The reaction mixture after
the extractive workup was treated with p-toluenesulfonic acid to
form the tosylate salt 1. The overall process for the synthesis of
1 is depicted in Scheme 13.
reaction and in a flow reaction. The isopropyl ester was chosen
to be large enough to protect the ester during the organo-
metallic addition, yet small enough to allow transesterification
with methanol and ultimate amination with ethylamine. The
work up of this step was streamlined to minimize the number
of unit operations and solvent volumes by carefully considering
material and solvent properties and monitoring mass balance
throughout the process. Of the various deoxofluorinating
reagents screened and studied for the second step, Deoxo-
Fluor was found to give high yield, diastereoselectivity and
chemoselectivity in the formation of 1, provided that alcohol 5
was pure, that the solution of 5 was added to the solution of
Deoxo-Fluor, and the reaction was quenched by subsurface
inverse addition with good mixing and pH control. The close
collaboration between synthetic chemists, automation special-
ists, chemical engineers, and analytical chemists greatly
facilitated this process research and development.
EXPERIMENTAL SECTION
■
Preparation of Propan-2-yl 3-[3-fluoro-4-(pyrrolidin-
1-ylmethyl)phenyl]-3-hydroxycyclobutanecarboxylate
24. Batch Process. To THF (150 mL) was added 1-(4-bromo-
2-fluorobenzyl)pyrrolidine 14 (39 g, 150 mmol, 1 equiv), and
the mixture was cooled to −10 °C. Isopropylmagnesium
chloride (2.0 M in THF, 32 mL, 63 mmol, 0.42 equiv) was
added over 20 min at ≤5 °C. n-Butyllithium (2.5 M in hexane,
50 mL, 126 mmol, 0.84 equiv) was added over 30 min at ≤5
°C. After an additional 30 min stir at 0 °C, the reaction was
deemed complete via HPLC (<1% 14 remaining). The reaction
mixture containing the magnesium ate complex 9 was cooled to
−35 °C, and propan-2-yl 3-oxocyclobutanecarboxylate (23, 25
g, 158 mmol, 1.05 equiv) was charged over 40 min at ≤−30 °C.
The reaction mixture was stirred for 1 h at −30 °C, warmed to
20 °C, and then quenched by addition of methanol (40 mL).
The potency of the solution was measured by HPLC, indicating
a yield of 65% of propan-2-yl 3-[3-fluoro-4-(pyrrolidin-1-
ylmethyl)phenyl]-3-hydroxycyclobutanecarboxylate, 24.
Simultaneous Addition Process. The magnesium ate
complex 9 mixture in THF was prepared as above and diluted
with THF to form a 0.5 M solution (Solution A) and held at 0
°C. In a separate vessel, propan-2-yl 3-oxocyclobutanecarbox-
ylate 23 (25 g, 158 mmol, 1.05 equiv wrt 14) was diluted with
THF to form a 1.05 M solution (Solution B). THF (60 mL)
was charged to a separate vessel and cooled to −10 °C.
Solution A and Solution B were added to the vessel containing
THF simultaneously over a period of 1 h using pumps with
flow rates of 5.0 mL/min (Solution A) and 2.5 mL/min
(Solution B). The reaction mixture was stirred at −10 °C for 1
h after the additions were complete, warmed to 20 °C, and then
quenched by the addition of methanol (40 mL). The potency
of the solution was measured by HPLC, indicating a yield of
65% of propan-2-yl 3-[3-fluoro-4-(pyrrolidin-1-ylmethyl)-
phenyl]-3-hydroxycyclobutanecarboxylate, 24.
Continuous Plug Flow Process. The magnesium ate
complex 9 mixture in THF was prepared as above but on a
75 mmol scale of 14 and diluted with THF as needed to form a
0.5 M solution wrt 14 (Solution A) and held at 0 °C. In a
separate vessel, propan-2-yl 3-oxocyclobutanecarboxylate 23
(12.3 g, 79 mmol, 1.05 equiv wrt 14) was diluted with THF to
form a 1.05 M solution (Solution B). The flow reactor was
cooled to −5 °C, and the pumps were started simultaneously to
begin flow of Solution A (1.3 mL/min) and Solution B (0.65
mL/min). On the basis of the reactor volume (10 mL), these
CONCLUSION
■
In summary, a concise process for the manufacture of H3
antagonist 1 was developed in our laboratories. This process
relies on the chemoselective addition of an organometallic
reagent to a cyclobutanone carbonyl in the presence of an ester
group, the addition of an amine to the preserved ester group,
and a diastereoselective deoxofluorination to provide the target
compound. A retrosynthetic analysis which included the
calculation and measurement of the relative stabilities of the
cis and trans isomers of the target molecule and observations of
the stereochemistry of the solvolysis of the fluorocyclobutane
focused our efforts on this strategy. Among the many types of
organomagnesium reagents, a relatively new type of reagent, a
triaryl magnesium ate complex (9), was employed since it can
be prepared in a homogeneous dose-controlled reaction under
noncryogenic conditions and has good reactivity. The stability
of this reagent was rapidly screened vs temperature and solvent
with the aid of an automated workstation, deuterium
quenching, and LC/MS. The addition order, reaction temper-
ature, and addition rate for the addition of organometallic 9 to
ketone 23 was optimized by considering heat removal on scale
and the competing reactions of the starting materials and
intermediates during the addition. The continuous simulta-
neous combination of ketone 23 and organometallic 9 was
preferred and implemented via a simultaneous semibatch
I
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
conditions correspond to a 5-min residence time. The product
solution was flowed into an agitated vessel containing MeOH
(20 mL). The potency of the solution was measured by HPLC,
indicating a yield of 65% of propan-2-yl 3-[3-fluoro-4-
(pyrrolidin-1-ylmethyl)phenyl]-3-hydroxycyclobutanecarboxy-
late, 24. This procedure was scaled successfully to 750 mmol of
the starting material 14.
AUTHOR INFORMATION
Corresponding Author
*rajappa.vaidyanathan@bms.com; joel.m.hawkins@pfizer.com.
■
Present Address
†Process Research and Development, Bristol-Myers Squibb
India Pvt. Ltd., BBRC, Biocon Park, Bommasandra-Jigani Link
Road, Bangalore 560 099, India.
Preparation of N-Ethyl-3-[3-fluoro-4-(pyrrolidin-1-
ylmethyl)phenyl]-3-hydroxycyclobutanecarboxamide 5.
With a solution of propan-2-yl 3-[3-fluoro-4-(pyrrolidin-1-
ylmethyl)phenyl]-3-hydroxycyclobutanecarboxylate, 24 pro-
duced at 150 mmol scale by one of the methods above, a
solvent exchange into MeOH (2 × 265 mL) was performed. To
this solution, ethylamine (2.0 M in MeOH, 300 mL, 600 mmol,
4 equiv) was added and the reaction mixture was stirred at 45
°C for 4 h at which time it was deemed complete by HPLC
(<2% starting material 24). After solvent exchange into i-
PrOAc (3 × 265 mL), a solution of citric acid (72 g, 375 mmol,
2.5 equiv) dissolved in H2O (190 mL) was added. The pH was
adjusted to >10 via slow addition of 50 wt % aqueous NaOH
(60 mL, 1130 mmol, 7.5 equiv) and the phases were separated.
The organic phase was distilled under vacuum to a volume of
about 100 mL, cooled to 20 °C, and seeded. After stirring for 1
h, n-heptane (42 mL) was added over 30 min, and the resultant
suspension was cooled to 0 °C and stirred for 2 h. The
suspension was filtered, and the cake was washed with a 50:50
volume mixture of i-PrOAc/n-heptane (2 × 30 mL). The cake
was dried under vacuum at 50 °C for 12 h to give N-ethyl-3-[3-
fluoro-4-(pyrrolidin-1-ylmethyl)phenyl]-3-hydroxycyclobutane-
carboxamide, 5 as an off-white solid (26.4 g, 55% from 14; 85%
from 24).27
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Kristin Price, Kevin Girard, Dave Damon, Barb
Sitter, and Carlos Mojica for reaction, extraction, and solubility
screening. Jason Mustakis and Yuriy Abramov provided
computational support. Sadia Abid, Andrew Palm, and Joe
Mongillo provided valuable analytical support. Asaad Nematalla
and David Pfisterer modified the flow chemistry rig so we could
successfully execute our process. We also thank Karen
Sutherland for helpful suggestions and guidance during the
course of this work.
REFERENCES
■
(1) (a) Arrang, J.-M.; Garbarg, M.; Schwartz, J.-C. Nature 1983, 302,
832. (b) Leurs, R.; Blandina, P.; Tedford, C.; Timmerman, H. Trends
Pharmacol. Sci. 1998, 19, 177.
(2) Monti, J. M. Life Sci. 1993, 53, 1331.
(3) Wager, T. T.; Pettersen, B. A.; Schmidt, A. W.; Spracklin, D. K.;
Mente, S.; Butler, T. W.; Howard, H., Jr.; Lettiere, D. J.; Rubitski, D.
M.; Wong, D. F.; Nedza, F. M.; Nelson, F. R.; Rollema, H; Raggon, J.
W.; Aubrecht, J; Freeman, J. K.; Marcek, J. M.; Cianfrogna, J.; Cook, K.
W.; James, L. C.; Chatman, L. A.; Iredale, P. A.; Banker, M. J.;
Homiski, M. L.; Munzner, J. B.; Chandrasekaran, R. Y. J. Med. Chem.
2011, 54, 7602.
Preparation of N-Ethyl-3-fluoro-3-[3-fluoro-4-pyrrolidin-1-
ylmethyl)phenyl]cyclobutanecarboxamide 1. Bis(2-
methoxyethyl)amino)sulfur trifluoride (Deoxo-Fluor) (22 mL,
120 mmol, 1.1 equiv) was added to THF (350 mL) that was
precooled to 0 °C. A solution of N-ethyl-3-[3-fluoro-4-
(pyrrolidin-1-ylmethyl)phenyl]-3-hydroxycyclobutanecarboxa-
mide 5 (35 g, 109 mmol, 1.0 equiv) in THF (140 mL) was
added to the Deoxo-Fluor solution over 45 min while
maintaining the temperature at ≤0 °C. The reaction was
deemed complete by HPLC (<1% starting material) after
stirring for 1.5 h at 0 °C, and the reaction mixture was
transferred to an addition funnel. Into a separate vessel were
charged K2CO3, 30% w/v (350 mL, 980 mmol, 8 equiv) and 2-
MeTHF (525 mL); the mixture was cooled to −5 °C. With an
addition funnel, the reaction mixture containing the product
was transferred below the liquid surface into the K2CO3
mixture over a period of about 2 h at ≤0 °C. The phases
were separated, and the organic phase was washed with 20%
aqueous NaCl (175 mL). The organic layer was distilled under
vacuum to remove water and was finally reduced to a volume of
about 200 mL. The solution was filtered to remove salts, and
the solids were washed with 2-MeTHF (40 mL). To the filtrate
(product solution), a solution of p-toluenesulfonic acid
monohydrate (24.3 g, 125 mmol, 1.15 equiv) in 2-MeTHF
(105 mL) was added over 30 min during which time the
product crystallized. The slurry was stirred at 15−20 °C for 2 h.
The product was filtered and washed with 2-MeTHF (70 mL)
and dried under vacuum at 50 °C to provide 1 (42.4 g, 78%
yield) in 94.3 area % purity by HPLC.27
(4) Klampt, A. COSMO-RS: From Quantum Chemistry to Fluid-Phase
Thermodynamics and Drug Design; Elsevier: Amsterdam, 2005.
(5) TURBOMOLE, V6.0; 2009, University of Karlsruhe and
Forschungszentrum Karlsruhe GmbH, 1989−2007; TURBOMOLE
GmbH: Karlsruhe, Germany since 2007.
(6) COSMOtherm, Version C2.1_0110; COSMOLogic GmbH,
Leverkusen, Germany, .
(7) Cyclobutanes are puckered such that cis substituents in 1,3-
disubstituted cyclobutanes are pseudo-diequatorial and thermody-
namically favored. See: Allinger, N. L.; Conia, J. M.; Ripoll, J. −L.;
Tushaus, L. A.; Neumann, C. L. J. Am. Chem. Soc. 1962, 84, 4982.
(8) For lead references on the effect of Lewis acids on the addition of
organometallic reagents to carbonyl compounds, see (a) Hatano, M.;
Suzuki, S.; Ishihara, K. J. Am. Chem. Soc. 2006, 128, 9998. (b) Ipaktschi,
J.; Eckert, T. Chem. Ber 1995, 128, 1171. (c) Zhou, S.; Wu, K. -H.;
Chen, C. -A.; Gau, H. −M. J. Org. Chem. 2009, 74, 3500. (d) Sada, M.;
Matsubara, S. Chem. Lett. 2008, 37, 800. (e) Hatano, M.; Miyamoto,
T.; Ishihara, K. Org. Lett. 2007, 9, 4535. (f) Dosa, P. I.; Fu, G. C. J. Am.
Chem. Soc. 1998, 120, 445.
(9) Eisenbeis, S.; Barilla, M.; Raggon, J.; Stewart, M. Unpublished
results.
(10) For lead references on the addition of Grignard reagents to
cyclobutanones, see (a) Fujiwara, T.; Morita, K.; Takeda, T. Bull.
Chem. Soc. Jpn. 1989, 62, 1524. (b) Casey, B. M.; Eakin, C. A.;
Flowers, R. A. Tetrahedron Lett. 2009, 50, 1264. (c) Hall, H. K., Jr.;
Smith, C. D.; Blanchard, E. P., Jr.; Cherkofsky, S. C.; Sieja, J. B. J. Am.
Chem. Soc. 1971, 93, 121.
(11) The diarylmagnesium reagent was not considered due to the
likely complexity of preparing it on scale. For laboratory methods to
prepare diarylmagnesiums, see Screttas, C. G.; Micha-Screttas, M. J.
Organomet. Chem. 1985, 292, 325.
(12) Note, however, that LiCl can accelerate the formation of
Grignard reagents from magnesium metal. Piller, F. M.; Appukkuttan,
J
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
P.; Gavryushin, A.; Helm, M.; Knochel, P. Angew. Chem., Int. Ed. Engl.
2008, 47, 6802.
(13) (a) Cai, W.; Ripin, D. H. B. Synlett 2002, 273. (b) Turner, R.
M.; Lindell, S. D.; Ley, S. V. J. Org. Chem. 1991, 56, 5739.
(14) (a) Krasovskiy, A.; Knochel, P. Angew. Chem., Int. Ed. Engl.
2004, 43, 3333. (b) Hauk, D.; Lang, S.; Murso, A. Org. Process Res. Dev.
2006, 10, 733. (c) Shi, L.; Chu, Y.; Knochel, P.; Mayr, H. J. Org. Chem.
2009, 74, 2760. (d) Knochel, P.; Krasovkiy, A. U.S. Patent 7,384,580,
2008. (e) Knochel, P.; Krasovskiy, A. U.S. Patent 7,387, 751, 2008.
(15) (a) Gallou, F.; Haenggi, R.; Hirt, H.; Marterer, W.; Schaefer, F.;
Seeger-Weibel, M. Tetrahedron Lett. 2008, 49, 5024. (b) Krasovskiy,
A.; Straub, B. F.; Knochel, P. Angew. Chem., Int. Ed. Engl. 2006, 45,
159. (c) For the application of a trialkyl magnesium ate complex to
prepare a triaryl magnesium ate complex on a large scale via a non-
cryogenic halogen-metal exchange, see: Mase, T.; Houpis, I. N.; Akao,
A.; Dorziotis, I.; Emerson, K.; Hoang, T.; Iida, T.; Itoh, T.; Kamei, K.;
Kato, S.; Kato, Y.; Kawasaki, M.; Lang, F.; Lee, J.; Lynch, J.; Maligres,
P.; Molina, A.; Nemoto, T.; Okada, S.; Reamer, R.; Song, J. Z.;
Tschaen, D.; Wada, T.; Zewge, D.; Volante, R. P.; Reider, P. J.;
Tomimoto, K. J. Org. Chem. 2001, 66, 6775.
(16) For lead references on the effect of solvents on Grignard
reagents, see (a) Ashby, E. C.; Reed, R. J. Org. Chem. 1966, 31, 971.
(b) Sassian, M.; Tuulmets, A. Helv. Chim. Acta 2003, 86, 82.
(17) Species with anionic heteroatoms such as the anion of amide 16
could also suffer from intramolecular or intermolecular anion addition
to the cyclobutanone, protecting the ketone from addition of the aryl
anion.
(18) Davies, A. J.; Scott, J. P.; Bishop, B. C.; Brands, K. M. J.; Brewer,
S. E.; DaSilva, J. O.; Dormer, P. G.; Dolling, U. -H.; Gibb, A. D.;
Hammond, D. C.; Lieberman, D. R.; Palucki, M.; Payack, J. F. J. Org.
Chem. 2007, 72, 4864.
(19) The heat of reaction for the addition of the isopropyl ester was
determined to be 199 kJ or 48 kcal per mol of aryl bromide 14.
(20) Bundesmann, M. W.; Coffey, S. B.; Wright, S. W. Tetrahedron
Lett. 2010, 51, 3879.
(21) Hayashi, H.; Sonoda, H.; Fukumura, K.; Nagata, T. Chem.
Commun. 2002, 1618.
(22) (a) Beaulieu, F.; Beauregard, L.-P.; Courchesne, G.; Couturier,
M.; LaFlamme, F.; L’Heureux, A. Org. Lett. 2009, 11, 5050.
(b) L’Heureux, A.; Beaulieu, F.; Bennett, C.; Bill, D. A.; Clayton, S.;
LaFlamme, F.; Mirmehrabi, M.; Tadayon, S.; Tovell, D.; Couturier, M.
J. Org. Chem. 2010, 75, 3401.
(23) Prepared in situ from Et3N·3HF and 1 equivalent of Et3N.
(24) Et3N·3HF gave low conversion with a dr of 2.5:1 (2:3) whereas
Et3N·HF provided a 3:1 ratio along with 20% of 28. No conversion
was obtained without an additive.
(25) (a) Carpino, L. A.; El-Faham, A. J. Am. Chem. Soc. 1995, 117,
5401. (b) El-Faham, A.; Khattab, S. N. Synlett 2009, 6, 886.
(26) Bellavance, G.; Dube,
(27) See ref 3. for characterization data.
́
P.; Nguyen, B. Synlett 2012, 23, 569.
K
dx.doi.org/10.1021/op300093j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX