Process Development and Scale-Up of a Benzoxazepine-Containing Kinase Inhibitor

Article abstract of DOI:10.1021/acs.oprd.5b00037

The benzoxazepine core is present in several kinase inhibitors, including the mTOR inhibitor 1. The process development for a scalable synthesis of 7-bromobenzoxazepine and the telescoped synthesis of 1 are reported. Compound 1 consists of three chemicall

Full text of DOI:10.1021/acs.oprd.5b00037

Article
pubs.acs.org/OPRD
Process Development and Scale-Up of a Benzoxazepine-Containing
Kinase Inhibitor
Sriram Naganathan,*,†,§ Denise L. Andersen,†,∥ Neil G. Andersen,†,⊥ Stephen Lau,†,# Anders Lohse,‡
and Mads Detlef Sørensen‡
†Exelixis, Inc., 210 East Grand Avenue, South San Francisco, California 94080, United States
‡Niels Clauson-Kaas A/S, Rugmarken 28, Farum DK-3520, Denmark
ABSTRACT: The benzoxazepine core is present in several kinase inhibitors, including the mTOR inhibitor 1. The process
development for a scalable synthesis of 7-bromobenzoxazepine and the telescoped synthesis of 1 are reported. Compound 1
consists of three chemically rich, distinct fragments: the tetrahydrobenzo[f][1,4]oxazepine core, the aminopyridyl fragment, and
the substituted (methylsulfonyl)benzoyl fragment. Routes were developed for the preparation of 3-fluoro-2-methyl-4-
(methylsulfonyl)benzoic acid (17) and tert-butyl 7-bromo-2,3-dihydrobenzo[f ][1,4]oxazepine-4(5H)-carboxylate (2). The
processes for the two compounds were scaled up, and over 15 kg of each starting material was prepared in overall yields of 42%
and 58%, respectively. A telescoped sequence beginning with compound 2 afforded 7.5 kg of the elaborated intermediate 5-
(2,3,4,5-tetrahydrobenzo[f][1,4]oxazepin-2-amine dihydrochloride (6) in 63% yield. Subsequent coupling with benzoic acid 17
gave 7.6 kg of the target compound 1 in 84% yield. The preferred hydrochloride salt was eventually prepared. The overall yield
for the synthesis of inhibitor 1 was 21% over eight isolated synthetic steps, and the final salt was obtained with 99.7% HPLC
purity.
Schmidt rearrangement4 to establish the benzoxazepine
skeleton. Treatment of the chromanone raw material 8 with
sodium azide afforded lactam 9. Regioselective bromination
could be achieved under standard conditions to yield lactam 10.
1. INTRODUCTION
Tyrosine kinases are important enzymes for signal transduction
in cells. Therefore, they are often targets for the treatment of
diseases that are caused by dysregulation of cellular processes,
Reduction of the lactam to the amine was accomplished using
such as cancers. Mammalian target of rapamycin (mTOR) is a
complexed lithium aluminum hydride,5 yielding benzoxazepine
kinase in the phosphatidylinositol-3-kinase (PI3K) family of
enzymes and is implicated in the regulation of cell growth and
11. However, varying levels of debromination were observed
during the reduction,6 and the amount of debrominated
proliferation. Various inhibitors of mTOR have been explored
as possible agents for treatment of various cancers,1 including
product (12, Figure 1) was proportional to the reaction scale:
the benzoxazepine-containing candidate 1.2
on a 500 g scale, up to 40% yield of the byproduct 12 was
obtained.
The step order was shown to be critical. When the
bromination step followed the lactam reduction, the
regioselectivity was severely compromised: the Schmidt
rearrangement on the brominated chromanone gave the lactam
in which the aryl group had migrated, affording the
regiosiomeric lactam. Protection of the amine as the tert-butyl
1.1. Discovery Synthesis of 1. The final product 1
carbamate was achieved in a straightforward manner under the
usual conditions7 using aqueous base and Boc-anhydride to
form 2. While the brominated and debrominated products
could be separated by careful chromatography, such separation
was cumbersome and was not viable for the larger scales
required to advance the clinical programsthe material needs
of the proposed studies translated to requirements of
approximately 12 kg of benzoxazepine 2.
consists of three distinct fragments, which are assembled in four
steps (Scheme 1). First, bromobenzoxazepine 2 is converted to
the corresponding boronic acid 3. Palladium-mediated coupling
with 5-bromo-2-aminopyridine (4) followed by the removal of
the Boc group in aminopyridine 5 yields the penultimate
diamine 6. Final coupling with acid chloride 7 yields the active
pharmaceutical ingredient (API) 1.
While aminopyridine 4 is readily available from several
commercial sources in large quantities, the other two starting
materials, benzoxazepine 2 and acid chloride 7, had to be
synthesized. We report the process development and scale-up
of each of these intermediates, ultimately leading to preparation
of over 7 kg of drug substance 1 to enable clinical studies.
1.2. Discovery Synthesis of Bromobenzoxazepine 2.
The initial synthesis of 2 was conducted beginning with 4-
chromanone (8), as shown in Scheme 2,3 and relied on a
1.3. Discovery Synthesis of (Methylsulfonyl)benzoic
Acid 17. The discovery synthesis of benzoic acid 17 was
carried out in four steps from 2,3-difluorotoluene (13), which is
Special Issue: Sustainable Chemistry
Received: February 3, 2015
© XXXX American Chemical Society
A
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1
Scheme 2
be readily controlled. Residual iron powder inside reaction
vessels may lead to unforeseen complications in future
reactions. Bromodifluorotoluene 14 was purified by vacuum
distillation, which may limit the choice of contract manufac-
turers. Formation of the aryl Grignard reagent required low
temperatures. However, there was another overriding factor
that led us explore alternative reaction conditions: the excess
alkyl Grignard reagent formed the corresponding alkyl
carboxylic acid when treated with carbon dioxide (pentanoic
acid if butylmagnesium chloride was used or isobutyric acid if
isopropylmagnesium chloride was used). It was very difficult to
remove these aliphatic carboxylic acids, and their effect on the
downstream chemistry was unclear. Furthermore, it was
difficult to monitor the reaction of the Grignard reagent with
carbon dioxide.
Figure 1. Structure of debrominated benzoxazepine.
readily available on a large scale (Scheme 3).8 Bromination of
13 was conducted with elemental bromine mediated by iron
powder to give 6-bromo-2,3-difluorotoluene (14). Metal−
bromine exchange using isopropylmagnesium chloride9 fol-
lowed by trapping of the aryl Grignard reagent with dry ice gave
benzoic acid 15. Aromatic nucleophilic substitution with
sodium methylthiolate furnished sulfide 16, and a subsequent
oxidation using Oxone10 gave the desired sulfone 17. While the
overall route itself could be used for scale-up, there were several
issues that precluded the direct scale-up of the initial synthesis.
The iron-powder-mediated bromination had an unpredict-
able induction period, which caused an exotherm that could not
While the SNAr reaction was high-yielding, repeated
additions of aqueous sodium methylthiolate were necessary to
obtain complete conversion to sulfide 16. The oxidation using
Scheme 3
B
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 4
Scheme 5
Oxone was not preferred because the reaction was extremely
volume-intensive and therefore severely limited the throughput.
1.4. Alternative Synthesis of Bromobenzoxazepine 2.
A second approach for the synthesis of intermediate 2 was
explored in which the benzoxazepine ring was formed by a
Mitsunobu reaction (Scheme 4).11 This strategy did not involve
a lactam and thus avoided strong reducing conditions and the
associated dehalogenation. Reductive amination of 5-bromosa-
licylaldehyde (18) with ethanolamine in methanol using
sodium borohydride gave the amine, which was treated with
aqueous base and Boc-anhydride to form diol 19. Subjecting
diol 19 to Mitsunobu conditions gave benzoxazepine 2. While
this approach was used to provide material on an interim basis
in approximately 200 g batches, it was not sustainable for the
longer term because of cost, safety concerns, and complexity of
product isolation.
smaller scales the reaction sequence worked efficiently.
However, we encountered several problems during scale-up.
First, the filtration of imide 22 on a large scale was extremely
slow and took several days, as opposed to the expected few
hours. Second, the yield of intermediate 22 was very low (23%
instead of the expected 70−80%). An investigation of the
various reaction streams revealed that the major product was
acid−amide 23 (Figure 2), presumably formed by hydrolysis of
the imide.13
Figure 2. Structure of the hydrolysis byproduct during imide synthesis.
2. IMIDE-BASED SYNTHESIS OF
BROMOBENZOXAZEPINE 2
Since monitoring of the reaction progress revealed that the
product was exclusively compound 22, the most likely source of
the hydrolysis was the quench. As the aqueous citric acid was
added slowly to prevent excessive heat evolution during the
quench, a sufficiently basic aqueous environment must have
existed, leading to a substantial amount of hydrolysis14 to form
by product 23.
2.1. Process Development of the Borane Reduction. A
suspension of imide 22 in THF was treated with borane·THF
complex for 24−36 h at 50 °C. Borane·THF was added slowly
at the beginning to ensure that hydrogen evolved as a result of
the reaction with the imide could be safely removed.
Calorimetric evaluation showed that the heat evolution (Figure
3) and gas evolution (Figure 4) could be controlled by the rate
of addition. The majority of the heat evolution was observed
after the initial 10% of borane addition, while some heat
evolution was observed until about one-third of the borane·
THF had been added. Care was taken not to heat the reaction
mixture in order to eliminate the risk of formation of
diborane.15
The problem of debromination during the reduction of a
lactam could be mitigated by using borane as the reducing
agent.12 However, an alternative strategy for the construction of
the oxazepine ring was also required since there were several
safety considerations associated with the Schmidt rearrange-
ment step. We chose to form the seven-membered ring by
forming an imide3 and then reducing it to the amine using
borane, thereby avoiding both problems encountered in the
earlier synthesis (Scheme 5). The starting material, 5-
bromosalicylamide (20), is readily available commercially on
a large scale. Treatment of 20 with ethyl bromoacetate in
dimethylacetamide using potassium carbonate as the base
yielded ether 21. Intermediate 21 was not isolated but was
converted directly to the imide using the commercially available
sodium tert-amylate solution in THF. Upon completion of the
cyclization, the reaction mixture was treated with aqueous citric
acid, and the precipitated product, benzoxazepine 22, was
collected by filtration. Imide 22 was reduced using borane·
THF. A large excess of borane was required to effect the
complete reduction of the amine, which resulted in a highly
stable amine−borane complex. The excess borane was first
quenched using methanol, and the amine−borane complex was
broken by refluxing with concentrated hydrochloric acid. On
When the imide had been completely consumed, methanol
was added slowly over a period of 16−24 h to ensure that the
liberated hydrogen was efficiently removed. Calorimetric
evaluation (Figure 5) showed that the hydrogen evolution
was mostly complete after the addition of the first 10% of the
C
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 3. Heat evolution during addition of borane·THF to a THF suspension of imide 22.
Figure 4. Hydrogen evolution during addition of borane·THF to a THF suspension of 22.
Figure 5. Heat evolution during addition of methanol to the borane reduction mixture.
charge of the methanol and was generally controlled by the rate
of addition (Figure 6). Because of the limitation of the air-
handling system’s ability to safely sweep away the hydrogen gas
liberated during the quench of the excess borane employed, the
imide reduction was initially carried out in multiple batches of
approximately 500 g each. The outputs from these batches were
combined and carried forward. At the end of the quench,
concentrated HCl was added, and the reaction mixture was
heated at reflux to break down the amine−borane complex.
The hydrochloride salt of the amine (11·HCl) was isolated by
filtration.
D
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 6. Hydrogen evolution during addition of methanol to the borane reduction mixture.
Scheme 6
Scheme 7
The Boc protection of the salt 11·HCl was conducted in a
straightforward manner using aqueous potassium carbonate and
Boc-anhydride in ethanol. This procedure was utilized to
prepare 2.9 kg of intermediate 2 in one batch and then to
prepare the initial supplies for toxicological studies.
2.2. Scale-Up of the Imide Approach to Bromoben-
zoxazepine 2. A more robust and easily scaled synthetic
approach was required for making larger quantities of
compound 2 (Scheme 6). We investigated alternative ring-
closure approaches in which imide 22 would not be exposed to
aqueous base. We decided to take advantage of the facile
alkaline hydrolysis to produce the acid and isolate it. Thus, the
potassium carbonate-mediated coupling of salicylamide 20 and
ethyl bromoacetate was followed by treatment with aqueous
sodium hydroxide to complete the hydrolysis to give acid 23.
The ring closure mediated by thionyl chloride16 was
modified such that the imide was prepared from salicylamide
20 in two steps. The isolated acid−amide 23 was treated with
thionyl chloride in refluxing toluene to form imide 22 in good
yield. The material could be isolated simply by cooling the
reaction mixture and adding heptane. This two-step procedure
was implemented on an approximately 4 kg scale, and a total of
20 kg of imide 22 was prepared.
The reduction was then scaled up at a contract
manufacturing facility. Several changes were made to the
small-scale procedure to render the process both safe and
efficient, taking advantage of operational and engineering
controls that are available at this facility. First, the borane
treatment was conducted at a lower temperature (35 °C vs 50
°C), and longer reaction times were necessary. Second, a
E
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
reverse quench was implemented as the produced hydrogen gas
was swept directly to an incinerator using a nitrogen sweep. At
the end of the quench, 31% aqueous hydrochloric acid was
added (as opposed to concentrated hydrochloric acid, in order
to avoid the fuming during handling) to break up the amine−
borane complex at 50 °C. The precipitated hydrochloride salt
was isolated by filtration, washed, and dried. The yield starting
from 19.8 kg of imide was 75%.
Minor changes to the Boc-protection step were also made
during scale-up. The reaction was seeded with the product prior
to the addition of Boc-anhydride to facilitate the formation of
larger particles, thereby easing filtration. The yield for this step
was 90%, and 15 kg of intermediate 2 was prepared in this
manner.
Grignard reagent to the ester and oxidative coupling of the
Grignard reagent, respectively. The ester hydrolysis was carried
out with aqueous sodium hydroxide and methanol at 65 °C.
When the ester hydrolysis was complete, the aqueous layer was
acidified. The product was extracted into the organic layer, and
inorganic salts were removed. The product was then extracted
into the aqueous layer using dilute sodium hydroxide, and the
aqueous layer was washed with heptane and acidified to
precipitate the product. Several noteworthy aspects were
observed in the acid−base extractions. When the initial reaction
mixture was treated with dilute acid followed by ester
hydrolysis and acidification (in order to save one operation),
the product quality and yield were significantly lower. The final
heptane wash of the basic aqueous layer was necessary to
ensure that the acid 15 that precipitated upon acidification was
not sticky, in order to facilitate filtration. The Grignard reaction
was scaled up successfully multiple times, starting with 5−8 kg
of bromide 14 per batch, to produce nearly 20 kg of the acid
intermediate 15 in 75−89% yield.
Numerous inorganic and organic bases were screened in a
variety of solvents for the SNAr reaction of acid 15 with sodium
methylthioate to give thioether 16. The best conditions
involved the use of 2.1 equiv of solid sodium methylthioate18
and 2 equiv of DBU in N-methyl-2-pyrrolidone (NMP) at 95
°C. The product was isolated by initial dilution with water,
cooling of the reaction mixture, and acidification. The choice of
acid utilized turned out to be very important (vide infra). Since
the oxidation reaction was run under aqueous conditions, it was
not necessary to fully dry the filtered product, thioether 16. An
assay of the partially dried filter cake was obtained, and the
stoichiometry of the reagents in the oxidation reaction was
adjusted accordingly.
Oxidation of the sulfur in sulfide 16 to give sulfone 17 was
accomplished using hydrogen peroxide in acetic acid, which was
substantially more volume-efficient than the previous Oxone
oxidation. The first oxidation to give the sulfoxide was found to
be highly exothermic, while the second oxidation was observed
to be endothermic and was aided by the exotherm of the first
oxidation. External heat was necessary to complete the
oxidation of the sulfoxide to the sulfone. In small-scale
oxidations we observed significant heat evolution when
hydrogen peroxide was added to a mixture of the starting
material 16 and acetic acid with subsequent heating. If reactions
under these conditions had been scaled up, there would have
been a potential for an unsafe exothermic runaway.
Calorimetric evaluation of the oxidation reaction revealed that
when the mixture of sulfide 16 and acetic acid was first heated
to 60−65 °C and then hydrogen peroxide was added slowly,
the rate of the exotherm was proportional to the rate of
addition of hydrogen peroxide and could be controlled by the
rate of addition. The oxidation was scaled up using this
procedure with 2.6 equiv of hydrogen peroxide. When HPLC
analysis revealed that all of the starting material 16 and the
intermediate sulfoxide were converted to sulfone 17, the excess
oxidant was quenched by the addition of DMSO. The
precipitation of the product was completed by dilution with
water, and sulfone 17 was isolated by filtration. The combined
yield for the conversion of difluorobenzoic acid 15 to sulfone
17 was around 80%.
3. SCALED-UP SYNTHESIS OF BENZOIC ACID 17
Since the discovery route toward acid 17 was short and
straightforward, the same synthetic approach was followed, but
reagents and conditions more suited for large-scale preparation
were employed (Scheme 7). Process development and scale-up
were focused only on the preparation of acid 17 instead of acid
chloride 7 because of poor regioselectivity for amidation and
time constraints.
Bromination of arene 13 was conducted using elemental
bromine but catalyzed by aluminum chloride.17 While the
bromination reaction could be conducted without solvent or
using a variety of solvents, dichloromethane was employed as
the solvent, and addition of bromine as a solution in
dichloromethane at 5−10 °C was preferred, as it provided
the cleanest product and most convenient means of isolation.
When the bromination was conducted without solvent, a
significant yield (nearly 10%) of a dibromo product was
observed. When nonhalogenated solvents were utilized, layer
cuts for aqueous washings during reaction workup were
problematic, presumably because of the high density of
bromoarene 14. Also, the use of higher-boiling solvents
interfered with the distillation of the product, bromoarene 14.
A small amount of a dibromo product was always formed but
could be minimized by controlling the stoichiometry of
bromine. Any remaining starting material 13 and dibromo
products could be removed by distillation. The small amount of
impurities present (up to 3%) did not affect the downstream
chemistry.
A conventional Grignard reaction using elemental magne-
sium was chosen for the synthesis of acid 15. The yield of the
reaction with carbon dioxide was inconsistent, and it could not
be determined when the addition was complete because there
was no convenient way to monitor the progress of the reaction
of the Grignard reagent with carbon dioxide. However, the
product was easily purified using a pH swing, as the sodium salt
of the acid was very water-soluble and the acid itself was very
insoluble.
This property of the acid was exploited by first preparing the
ester using methyl chloroformate as the electrophile and then
saponifying the resulting ester and subsequently isolating the
acid using acid−base extractions. While there was no substantial
difference in the reactivity, 2-methyltetrahydrofuran (2-
MeTHF) was preferred to THF itself, as the layer separations
were considerably more facile and product losses were minimal.
The Grignard reagent was added to a solution of methyl
chloroformate at 0−8 °C. When the chloroformate was added
to the Grignard reagent, significant quantities of the
benzophenone and the biaryl were formed via addition of the
During process development and initial scale-up of the
oxidation reaction, a new impurity was observed in the product,
sulfone 17, at a level of about 1% relative to the parent
compound. This impurity was later-eluting in reversed-phase
F
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 8
Figure 7. Major impurities formed during amidation.
Scheme 9
HPLC, and LC−MS analysis revealed the presence of one
chlorine atom. Examination of its fragmentation pattern and
comparison of this pattern to that obtained for sulfone 17
indicated the presence of a −SO2CH2Cl fragment. The
structure 25 was proposed for the impurity, and it could result
according to the reaction pathway outlined in Scheme 8.
Residual sodium chloride or hydrochloric acid from the use of
hydrochloric acid to neutralize the aqueous base during the
preparation of thioether 16 could be oxidized by hydrogen
peroxide to afford elemental chlorine. Methyl sulfides readily
undergo α-chlorination19 under these conditions to form the
corresponding chloromethyl sulfides (e.g., 24), which can then
be oxidized to the chloromethyl sulfones (e.g., 25) by
additional hydrogen peroxide. This hypothesis was confirmed
by spiking experiments. A sample of sulfide 16, which had been
previously used to produce pure sulfone 17 without
contamination by impurity 25, was spiked with 0.1 equiv of
hydrochloric acid. When this material was subjected to the
typical oxidation conditions, compound 25 was the only major
new product observed, at the expected level. Additional
evidence for this hypothesis included the lack of a chlorine-
containing impurity in the starting material, sulfide 16 (with a
mass corresponding to that of compound 24, a chloromethyl
sulfide), and the absence of impurity 25 upon switching to
concentrated sulfuric acid for the acidification following the
SNAr reaction. While other solutions were possible, such as
additional aqueous washings to ensure removal of hydrochloric
acid or sodium chloride and full drying of sulfide 16, the switch
to sulfuric acid was simpler because it preserved the option to
move ahead with the wet cake.
4. TELESCOPED SYNTHESIS AND SCALE-UP OF API 1
We chose to follow the same route employed during the
discovery synthesis of compound 1. However, several
modifications were necessary to smoothly prepare large
quantities of this material at a variety of contact manufacturers.
The following were areas of emphasis we chose to explore: the
cryogenic conditions required for the metal−halogen exchange
to make the arylboronic acid (or arylboronate) limited the
choice of manufacturers. The conditions for the Suzuki reaction
were harsh and used an expensive catalyst. The choice of
solvents for both the Suzuki reaction and the Boc deprotection
were not preferable for large-scale production. Finally, using the
acid chloride for coupling gave two major impurities, 26 and 27
(Figure 7), which were formed as a consequence of a lack of
regioselectivity.
The synthesis would also be more efficient and productive if
some steps of the sequence could be telescoped, minimizing
losses due to isolation. The ideal opportunity for telescoping
would be the initial three transformations. The Boc-
deprotected diamine 6 would become the first isolated
intermediate and provide a point of control before the final
coupling. When we initiated the process development, it was
not clear whether the free base or a salt would be selected for
clinical development. Thus, we chose to design the process so
G
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
that either could be the final isolated product. The telescoped
sequence leading to diamine 6 is shown in Scheme 9.
except for one major side product, which was tentatively
identified as boroxine 30.22 The boronic acid and the boroxine
remained in the organic layer while the salts were removed in
the acidic aqueous layer. The 3:30 ratio was somewhat
dependent on the temperature at which triisopropyl borate
was added to the magnesiate. At higher temperatures, more
boroxine 30 was produced: the range of the 3:30 ratio was from
about 1.2:1 to 2:1. The combined yield of acid 3 and boroxine
30 was usually greater than 85 area % by HPLC. The Suzuki
reaction was rapid with boronic acid 3 but much slower with
boroxine 30.
Magnesiation using alkyl Grignard reagents in combination
with alkyllithiums is a convenient and efficient alternative to the
use of alkyllithiums alone to effect metal−halogen exchange of
aryl bromides.20 These metal−halogen exchange reactions can
be conducted at significantly higher temperatures, typically −10
to +5 °C, as opposed to −50 °C and below for alkyllithiums
alone. The resulting triaryl magnesiate adds readily to a trialkyl
borate21 to produce the required arylboronate, which can then
be hydrolyzed to give the arylboronic acid. Metal−halogen
exchange was effected by first adding n-butylmagnesium
chloride to a solution of bromide 2 in THF at 0 °C, which
was a stable mixture, followed by the addition of butyllithium to
the mixture. Neither the addition of n-butylmagnesium chloride
nor the addition of butyllithium was particularly exothermic.
However, using dry THF was very important, and a moisture
specification of <0.1% was established to ensure that that
metalation reaction was complete. The progress of metalation
was monitored using HPLC by observing the consumption of
benzoxazepine 2 and the appearance of debrominated
compound 28 (Figure 8) upon aqueous quenching. When
The organic layer was sparged with nitrogen to suppress the
formation of biaryl 31 (Figure 9). Both the solution of boronic
Figure 9. Structure of the biaryl byproduct formed during the Suzuki
reaction.
acid 3 and the aqueous potassium carbonate solution were
sparged with nitrogen, and all of the solids were added under a
nitrogen atmosphere. Several catalyst−ligand combinations
were investigated. High conversion to the product was
observed under most conditions, and thus, the less-expensive,
easily handled catalyst bis(triphenylphosphine)palladium di-
chloride was chosen. There was no observable difference when
the catalyst was preformed or prepared in situ by adding
triphenylphosphine and palladium dichloride to the reaction
mixture. Upon completion of the coupling reaction, the basic
aqueous layer was discarded, and the product was extracted into
dilute aqueous hydrochloric acid.
The workup for the crude purification of the solution of
intermediate 5 had to be carefully choreographed. While the
rejection of major impurities was the primary objective, setting
up the solution for removal of the Boc group was also equally
important. We discovered that higher water content in the
solution of compound 5 was detrimental (it slowed the
deprotection considerably) and that reaction proceeded best in
THF. Furthermore, not all of the impurities that were present
at the end of the coupling reaction were apparent under
standard analytical HPLC conditions.
Figure 8. Debrominated product from aqueous quenching of the aryl
magnesiate.
the metalation was determined to be complete, triisopropyl
borate was added slowly while the temperature was maintained
below −8 °C, and the mixture was stirred overnight at that
temperature. If the reaction between the magnesiate and
triisopropyl borate had not been allowed to proceed to
completion, significant amounts of debrominated compound
28 would have resulted, thus diminishing the yield. Maintaining
a low temperature maximizes the formation of the desired
boronic ester intermediate 29.
The hydrolysis of boronic ester 29 to the corresponding
boronic acid 3 was accomplished using dilute hydrochloric acid
(Scheme 10). While it was possible to conduct the Suzuki
coupling with ester 29, the coupling reaction was significantly
faster with the boronic acid. The hydrolysis was rapid and clean,
Scheme 10
H
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 11
The solution of compound 5 in aqueous acid was treated
with methyl tert-butyl ether (MTBE), causing unknown
impurities to precipitate. Filtration through a pad of Celite
removed this solid and helped simplify the following layer cuts.
The organic layer was discarded, and isopropyl acetate was
added to the aqueous layer. Basicification to pH >9 using 25%
aqueous ammonium hydroxide allowed aminopyridine 5 to be
extracted into the organic layer. Palladium scavenging was
implemented at this point to ensure that the Pd content was at
or below levels determined to be acceptable according to
regulatory guidances.23 A silica-based scavenger, SPM-32 from
PhosphonicS, was chosen for this purpose.24 During the
scavenger treatment, another unknown impurity precipitated
out, but it was removed during the filtration of the scavenger.
We found that adding Celite to the reaction mixture, in
addition to the pad used for filtration, was very helpful in
decreasing the duration of the filtration. Following filtration, the
filter pad was washed with additional isopropyl acetate, and the
solvent was switched to THF by distillation of the isopropyl
acetate under vacuum and addition of THF. The distillation
was repeated, and THF was added to obtain an ∼10% (w/w)
solution of aminopyridine 5 in THF.
It was possible to conduct the extraction and Pd-scavenging
operation in THF by adding THF instead of isopropyl acetate
prior to basicification. However, there were several reasons for
adopting the two-solvent approach. First, the impurity that was
removed during the Pd-scavenging step was carried through to
the next step when THF was used. Second, the water content
in the THF solution of aminopyridine 5 was very high, and
repeated washing with brine was required in order to reduce
the moisture to an acceptable level to allow for smooth
deprotection. Unfortunately, the Boc deprotection reaction did
not proceed well in isopropyl acetate.
are inexpensive and easily handled. Unfortunately, none of the
class-3 solvents25 were suitable for the coupling reaction: while
very high conversion could be achieved for the coupling
reaction itself in most of the solvent systems tested, the
properties of the product were such that there were
insurmountable problems encountered during workup and
isolation.
For initial toxicological studies, acetonitrile/water was chosen
as the solvent system with N-methylmorpholine (NMM) as the
base. Under these conditions, the regioselectivity of the
coupling was very high and the levels of diacylated byproduct
26 were very low. A slight molar excess of the acid (1.2 equiv)
was activated with EDC in a separate reactor and then added to
a solution of amine 6 in acetonitrile/water that had been
neutralized with NMM. The coupling reaction was complete,
according to HPLC analysis, in ca. 90 min at room temperature.
The reaction mixture was washed three times with 6 M
potassium carbonate. Distillative crystallization was employed
to aid in product isolation: acetonitrile was distilled at
atmospheric pressure and continually replaced with MTBE
until the HPLC analysis revealed that the concentration of
product 1 in the supernatant solution was <5 mg/mL.
Achieving this concentration ensured that the product
crystallized in high yield and with high purity. At this stage,
additional MTBE was added to complete the crystallization,
and the mixture was cooled to room temperature. The product
was filtered, washed with MTBE, and then dried under vacuum.
The yield using 0.5−1 kg of 6 was 60−65%, with the
predominant loss of product occurring during the aqueous
carbonate washings. The losses were also more pronounced on
larger scales. The purity by LC area percentage (LCAP) was
between 98.6 and 99.5%.
While this process was adequate for preparation of initial
supplies for toxicological studies, there were several issues that
had to be addressed before embarking on production of drug
substance 1 under cGMP for clinical studies. The loss to
washings had to be minimized, and it was also necessary to
identify a convenient point to institute a polish filtration.
The choice of solvent for the coupling reaction was revisited
for the scale-up, and attention was directed toward a
combination of DMF and dichloromethane, which had also
shown considerable potential. The main advantage of this
solvent system was that the product was very soluble, allowing
for smooth aqueous workup, although slightly higher levels of
impurity 26 were formed. The amount of compound 26 could
easily be controlled by manipulating the stoichiometry of
diamine 6 and acid 17 as well as that of EDC and HOBt. An
excess of diamine 6 was ultimately employed because it limited
the formation of diamide 26 and its availability was no longer
limited. The base was switched to 3 M aqueous sodium
hydroxide, which also helped remove the excess benzoic acid
17. A polish filtration could be also readily incorporated during
The Boc deprotection of aminopyridine intermediate 5 in
THF solution was accomplished by initial heating to ∼35 °C
followed by treatment with concentrated hydrochloric acid.
While the deprotection was ∼99% complete in the first 8 h,
heating was usually continued overnight to ensure complete
deprotection. At the end of the reaction, the mixture was
cooled to room temperature, and the precipitated product was
filtered, washed, and dried. The dihydrochloride salt was
extremely well behaved, nonhygroscopic, and easily filtered and
handled. Typical yields for the preparation of compound 6
using the telescoped sequence ranged from 55 to 64%. On the
largest scale, the yield of dihydrochloride 6 was 63% starting
with 12.3 kg of bromobenzoxazepine 2.
Initial abbreviated polymorph screening of drug substance 1
revealed that there was only one major physical form, which
could be consistently prepared from a variety of solvents. A
series of coupling conditions, reagents, and solvents were
investigated (Scheme 11), and the N-(3-dimethylaminoprop-
yl)-N′-ethylcarbodiimide (EDC)/1-hydroxybenzotriazole
(HOBt) coupling system was chosen because these reagents
I
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 12
a recrystallization, which provided an upgrade in the overall
purity with minimal product loss.
dried under vacuum at 50 °C. Finally, 7.8 kg (95% yield) of 1·
HCl was produced with 99.7% LCAP purity. The final levels of
residual DMF and DCM were below 100 ppm, which are well
below the specified limits of 880 and 600 ppm, respectively.
The final level of residual Pd was below 0.2 ppm, which is also
well below the specified limit of 20 ppm.
Thus, the final process involved the activation of acid 17 in
3:1 dichloromethane/DMF using EDC and HOBt. In a
separate vessel, dihydrochloride 6 was suspended in the same
solvent system, neutralized with 2 equiv of 3 M aqueous
NaOH, and treated with the activated acid at 15 °C. After
overnight reaction, the organic layer was washed twice with
aqueous monobasic potassium phosphate solution. A small
quantity of DMF was added to ensure that the product stayed
in solution, and the organic layer was washed once with 1 M
aqueous NaOH. The product was precipitated from aqueous
DMF using distillative crystallization by distilling out dichloro-
methane and replacing it with water. The product was then
isolated by filtration, washed with water, and partially dried.
The wet cake was dissolved in refluxing THF/water and polish-
filtered while hot. The solution was concentrated to about half
its volume by vacuum distillation and cooled to 10 °C. Seeds
were then added, and the product was crystallized by addition
of MTBE and isolated by filtration. The yield using 6.6 kg of
diamine 6 was 84%, producing 7.6 kg of compound 1 with
98.8% LCAP purity. The purity profile of the material produced
on the larger scale under cGMP was comparable to those of
batches produced on the 0.5−1 kg scale, and no new impurities
were observed.
The HCl salt of compound 1 was selected for clinical studies
involving oral administration. Abbreviated salt and form
screening showed that a stable and consistent form of 1·HCl
was obtained from acetone and acetone/water. When a
suspension of the free base in acetone and concentrated HCl
were combined, the monohydrochloride salt was formed in
good yield (Scheme 12). However, there was an intermediate
stage where the mixture turned sticky and stirring was severely
compromised.
While screening the conditions to optimize the salt
formation, we also investigated the possibility of directly
converting the wet cake of the free base into the desired salt.
Systematic screening of the amount of water present in the
reaction from 0 to 1.5 equiv (w/w) revealed that 0.5 equiv (w/
w) of water was optimal in the salt formation. The physical
properties of the salt formation at lower water contents were
undesirable, while at higher water contents there were
considerably higher losses of product to the filtrate. When 1.5
equiv (w/w) of water was used, the isolated yield was only 71%,
compared with 92−95% using 0.5 equiv (w/w) or less. When
concentrated HCl was added to a heated suspension of free
base 1 in acetone/water, a brief solution resulted, followed by
product crystallization. Unfortunately, the duration of the
solution was variable and too short to enable a polish filtration
at this stage. Therefore, it was decided to utilize filtered
acetone, water, and concentrated HCl for the salt formation
and to conduct the polish filtration during the isolation of the
free base. The crystallization was completed at 20 °C, and the
product was isolated by filtration, washed with acetone, and
5. CONCLUSIONS
We have developed a scalable and efficient process for the
synthesis of 7-bromobenzoxazepine, an important scaffold
present in several kinase inhibitors. By means of this process,
over 15 kg of this intermediate was prepared. Concurrently, a
process for the large-scale preparation of a highly substituted
benzoic acid has also been developed. Finally, we have also
demonstrated a telescoped multikilogram-scale synthesis of a
clinical candidate that incorporates this benzoxazepine core.
Several batches of the drug substance 1 have been prepared to
support the development programs.
6. EXPERIMENTAL SECTION
General Information. All of the reagents and solvents
employed in the synthesis were purchased commercially and
used without further purification. Reactions requiring an inert
atmosphere were conducted under dry nitrogen. NMR spectra
were recorded on a Varian Mercury Plus 400 MHz instrument.
Peaks are reported as observed, without accounting for splitting
due to carbon−fluorine coupling. For 1H NMR spectra,
chemical shifts are reported in parts per million (ppm) relative
to an internal standard of tetramethylsilane; for 13C NMR
spectra, chemical shifts are referenced to the corresponding
deuterated solvent peak. Positive ion FAB high-resolution mass
spectrometry was conducted by Analytical Instrument Group,
Inc. (Raleigh, NC). HPLC analyses were conducted on an
Agilent 1200 series or Waters Empower System instrument
equipped with photodiode array detectors using Waters
SunFire 3.5 μm C18 columns (4.6 mm × 100 mm) and
mobile-phase gradients consisting of 0.05% phosphoric acid in
water and 0.05% phosphoric acid in acetonitrile.
The procedures outlined in this section are those employed
for the reactions run at the largest scale. Analytical character-
izations were obtained on materials that were representative or
purified for that purpose.
2-(4-Bromo-2-carbamoylphenoxy)acetic Acid (23). A
100 L jacketed reactor was charged with 5-bromosalicylamide
(4 kg, 18.5 mol) in dimethylacetamide (DMA) (18.5 L)
followed by potassium carbonate (20.3 mol) and ethyl
bromoacetate (3.24 kg, 19.4 mol). The mixture was heated to
50 °C, and when HPLC revealed that <5% of the
bromosalicylamide remained (ca. 3 h), the reaction mixture
was cooled to room temperature. Then 7.5 L of 20% aqueous
NaOH was added, and the mixture was heated at 50 °C until
no more ester remained (ca. 2 h). The reaction mixture was
cooled to room temperature, and the pH was adjusted to <2
using 6 M aqueous HCl. The precipitated product was
J
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
collected by filtration, washed with water, and dried in a
vacuum oven to yield 4.1 kg of compound 23 as a white solid
(81%).
(15 L) was added slowly while the temperature was maintained
below 25 °C. The organic layer was washed once with 1 M
aqueous HCl (5 L) and 5% aqueous NaHSO3 solution (5 L),
dried over anhydrous MgSO4, and concentrated by atmos-
pheric-pressure distillation. When the internal temperature
reached 50 °C, the product was purified by vacuum distillation
in two portions from a 5 L round-bottom flask. The fraction
boiling between 78 and 83 °C at 20−30 Torr was collected,
yielding two batches of bromide 14 as a colorless liquid (3.13
and 3.46 kg; combined yield 6.59 kg, 85%).
HPLC: 96.7% and 97.2% LCAP, respectively; 2.7% and 2.5%
“regioisomer”; 0.5% and 0.3% dibromo product. 1H NMR (400
MHz, CDCl3): δ 7.27 (m, 1H), 6.91 (dq, 1H), 2.35 (d, 3H).
3,4-Difluoro-2-methylbenzoic Acid (15). A 50 L
jacketed reactor was flushed with nitrogen and charged with
magnesium turnings (852 g, 35 mol) followed by a few crystals
of iodine (0.2 g) and 2-methyltetrahydrofuran (2-MeTHF) (20
L). The temperature of the reactor jacket was set at 45 °C. A
solution of 14 (6.52 kg, 31.5 mol) in 2-MeTHF (7 L) was
prepared, and approximately 0.5−1 L of the bromide solution
was added when the internal temperature reached 35 °C. When
the formation of the Grignard reagent had initiated (at around
40 °C internal temperature), the remainder of the solution of
14 was added at such a rate that the internal temperature was
maintained between 75 and 80 °C (the jacket temperature was
set at 45−50 °C); the addition took ca. 2 h. The brownish
solution was heated at 75−80 °C for an additional 1 h. HPLC
analysis revealed that <1% of the bromide remained. The
Grignard solution was cooled to below 25 °C.
7-Bromobenzo[f ][1,4]oxazepine-3,5(2H,4H)-dione
(22). A suspension of 23 (4.1 kg, 15 mol) in toluene (25 L) was
heated to 90 5 °C, and thionyl chloride (7.12 kg, 52 mol)
was added over ca. 15 min. The reaction mixture was heated at
that temperature until no more starting material remained as
determined by HPLC analysis (ca. 4 h). The reaction mixture
was cooled to 10 °C, and heptane (8.2 L) was added to
precipitate the product. The slurry was aged at that temperature
for 1 h, and the product was collected by filtration, washed with
heptane (8.2 L), and dried in a vacuum oven to yield 2.9 kg of
22 as a white solid (76%).
1H NMR (400 MHz, DMSO-d6): δ 11.54 (s, 1H), 8.07 (d, J
= 2.4 Hz, 1H), 7.80 (dd, J = 8, 2.4 Hz, 1H), 7.17 (d, J = 8 Hz,
1H), 4.81 (s, 2H).
7-Bromo-2,3,4,5-tetrahydrobenzo[f ][1,4]oxazepine
Hydrochloride (11·HCl). A 1000 L Hastelloy C vessel
connected to an incinerator for H2-containing off-gas was
charged with 22 (19.8 kg, 77.33 mol). After a nitrogen purge,
THF (86.4 kg) was added, followed by borane·THF complex
(BH3·THF) (288.4 kg, 328 mol) over ca. 1.5 h from a cylinder
maintaining an inner temperature of 30−32 °C (Caution!
Evolution of hydrogen gas and exotherm). The reaction mixture
was stirred for 4 h at 35 °C, after which an additional portion of
BH3·THF (144.2 kg, 164 mol) was added (Caution! Evolution
of hydrogen gas and exotherm) and the mixture was stirred for
120 h.
A 2500 L glass-lined quench vessel was charged with MeOH
(123.6 kg), and at 45 °C the reaction mixture from above was
added at a rate of 200 kg/h (Caution! Evolution of hydrogen
gas and exotherm). The quench vessel was purged for 1 h with
nitrogen to flush liberated hydrogen to the incinerator.
Aqueous 31% HCl (14.2 kg) was added in portions over 20
min, and the reaction mixture was stirred for 3 h under nitrogen
at 50 °C, cooled to 20 °C over 3 h, and stirred for 8 h. The
precipitated product was filtered, washed with EtOH (33 kg),
and blow-dried overnight to yield 16.5 kg of 11·HCl (75%
yield, 81% potency).
A 100 L jacketed reactor was flushed with nitrogen, charged
with methyl chloroformate (4 kg, 42.3 mol) and 2-MeTHF (7
L), and cooled to below 5 °C. The Grignard reagent solution
was added slowly over ca. 5 h while the internal temperature
was maintained below 8 °C (the jacket temperature was set at
−15 °C). At the end of the addition, the reaction mixture was
warmed to 20 °C and stirred at that temperature for about 1 h.
HPLC showed that <2% of 14 remained and mostly the ester
was present. A cold solution of sodium hydroxide (8 L of 50%
aqueous solution) in water (20 L) and methanol (3 L) was
added over 30 min. The resulting suspension was heated at 60−
65 °C for 8 h. The reaction mixture was analyzed by HPLC to
ensure complete conversion of the ester to the carboxylic acid.
The suspension was then cooled to below 25 °C, and
concentrated HCl (13 L) was added over about 1 h while
the temperature was maintained below 25 °C to obtain a pH of
<2 in the aqueous layer.
The aqueous layer was removed, and water (20 L) was added
to the organic layer, followed by 50% aqueous NaOH solution
(2 L) to obtain a pH of >10 in the aqueous layer. When the
target pH was achieved, heptane (10 L) was added, and the
aqueous layer was washed twice more with heptane (10 L
each). The aqueous layer was then cooled to below 20 °C, and
concentrated HCl (3 L) was added to obtain a pH of <2. The
precipitated product was collected by filtration, washed with
water (2 × 10 L) and heptane (5 L), and dried under vacuum
at 40 °C to obtain 4.82 kg of 15 as a colorless solid (88%).
HPLC assay: 98.1 wt %. 1H NMR (400 MHz, DMSO-d6): δ
13.3 (s, 1H), 7.71 (m, 1H), 7.35 (q, 1H), 2.48 (s, 3H). 13C
NMR (100 MHz, DMSO-d6): δ 167.7, 153.6, 153.5, 151.1,
150.9, 150.3, 150.1, 147.8, 147.7, 129.8, 129.7, 128.7, 128.6,
127.7, 127.6, 115.0, 114.9, 12.6, 12.5. IR: 1681, 1455, 1276,
1012, 771 cm−1. MS (EI) for C8H6F2O2: found 171 ([M −
1H NMR (400 MHz, DMSO-d6): δ 9.77 (br s, 1H), 7.71 (d, J
= 2.4 Hz, 1H), 7.52 (dd, J = 8, 2.4 Hz, 1H), 7.04 (d, J = 8 Hz,
1H), 4.30 (s, 2H), 4.21 (m, 2H), 3.45 (m, 2H).
tert-Butyl 7-Bromo-2,3-dihydrobenzo[f ][1,4]-
oxazepine-4(5H)-carboxylate (2). In a glass-lined 400 L
reactor, 11·HCl (16.5 kg, 62.4 mol) was dissolved in EtOH
(24.5 kg) at 20 °C. Aqueous K2CO3 (16 wt %; 61.8 kg, 71.5
mol) was added, and the mixture was seeded with 25 g seed
crystals of 2. Solid di-tert-butyl dicarbonate (14.9 kg, 68.3 mol)
was added in portions, maintaining a reaction temperature of
<30 °C. The mixture was stirred overnight and filtered. The
product on the filter was washed with 29 kg of a 3:1 mixture of
water and EtOH followed by 30 kg of water and then dried
under a stream of nitrogen to afford compound 2 as a white
solid (15.0 kg, 90%).
6-Bromo-2,3-difluorotoluene (14). A 50 L jacketed
reactor was charged with 2,3-difluorotoluene (13) (4.8 kg,
37.5 mol) and dichloromethane (15 L) and cooled to below 10
°C. Solid aluminum chloride (50 g, 0.375 mol) was added in
one portion. A solution of bromine (6 kg, 37.5 mol) in
dichloromethane (5 L) was added over a period of 3 h while
the temperature was maintained below 10 °C. When HPLC
analysis showed that <5% of 13 remained (1 h at 10 °C), water
K
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
H]−). FAB-MS for C8H6F2O2: found 173.04135 ([M + H]+),
calcd 173.04136.
reactor was charged with THF (54 kg) and N-Boc-7-
bromobenzoxazepine (2) (12.33 kg, 37.6 mol) under an inert
atmosphere and cooled to −10 to −5 °C. A solution of 20%
nBuMgCl in THF (8.77 kg, 15.0 mol) was added over 30 min
while the temperature was maintained below −2 °C. A solution
of 23% n-BuLi in hexane (8.38 kg, 30.1 mol) was added over 60
min, while the temperature was maintained below −2 °C. The
mixture was stirred for 15 min, and HPLC analysis showed that
<5% of 2 remained. The mixture was cooled to −15 to −10 °C,
and triisopropyl borate (9.20 kg, 48.9 mol) was added over 30
min while the temperature was maintained below −8 °C. The
mixture was stirred at −10 to −8 °C for 12−24 h until HPLC
analysis showed that <5% 28 was present. A solution of 2 M
HCl (31 kg) was added over 30 min while the temperature was
maintained below 20 °C. The mixture was stirred for 15−30
min at ca. 20 °C and then allowed to separate, and the aqueous
layer was removed. The THF layer was used as such for the
next step. A sample of 3 for analytical characterization was
prepared by precipitation using hexanes.
3-Fluoro-2-methyl-4-(methylthio)benzoic Acid (16).
An inert 100 L jacketed reactor was charged with 3,4-
difluoro-2-methylbenzoic acid (15) (3.88 kg, 22.5 mol)
followed by NMP (11.6 L, 3 volumes) and DBU (6.9 L, 46.2
mol, 2.05 equiv). The resulting mixture was heated to ca. 50 °C,
and NaSMe (3.3 kg, 47 mol, 2.1 equiv) was added in one
portion to give a thick slurry. The mixture was heated to
approximately 95−100 °C and held at that temperature for 2 h.
The mixture became homogeneous at around 60 °C. Additional
NaSMe was added in two portions (1.4 kg each) to achieve
consumption of the starting material. When HPLC analysis
showed that <5% of 15 remained, the solution was treated with
water (40 L, 10 volumes) over ca. 45 min and cooled to below
25 °C. The reaction mixture was acidified to a pH of <2 with
concentrated sulfuric acid (4.25 L) over ca. 2 h while the
temperature was maintained below 25 °C. The resulting thick
slurry was aged for ca. 1 h, and the product was collected by
filtration and washed with water (2 × 25 L). The filter cake was
partially dried under vacuum in the filter at 50 °C to yield
compound 16 as an off-white solid (8.1 kg, 51.7 wt %; the
purity was established by HPLC analysis against an external
standard, and the material was used as such in the next step). A
sample for analytical characterization could be obtained by
further drying and slurrying in methanol at 50 °C.
1H NMR (400 MHz, CDCl3): δ 8.07−7.03 (m, 3H), 4.87−
4.44 (m, 2H), 4.17−4.06 (m, 2H), 3.86−3.80 (m, 2H), 1.45−
1.41 (m, 9H). 13C NMR (100 MHz, CDCl3): δ 163.7, 155.3,
137.7, 136.9, 131.6, 125.3, 121.4, 80.5, 73.1, 50.7, 49.9, 28.6
(major isomer). IR: 1697, 1603, 1572, 1456, 1409, 1364, 1333,
1304, 1281, 1267, 1240, 1171, 1134, 1111, 1071, 1025, 982,
868 cm−1. FAB-MS for C14H20BNO5: found 293.14334 (M+),
calcd 293.14335.
tert-Butyl 7-(6-Aminopyridin-3-yl)-2,3-dihydrobenzo-
[f ][1,4]oxazepine-4(5H)-carboxylate (5). The THF solu-
tion of 3 from the previous step was subjected to a subsurface
nitrogen sparge for at least 1 h. Potassium carbonate (10.4 kg,
75.3 mol) was dissolved in water (12.3 kg), subjected to a
subsurface nitrogen sparge for at least 1 h, and then added to
the sparged THF solution of the boronic acid. Solid 2-amino-5-
bromopyridine (4) (6.51 kg, 37.6 mol) and bis-
(triphenylphosphine)palladium(II) dichloride (238 g, 3.4
mol) were added under nitrogen. The mixture was heated to
reflux and stirred 12−24 h until HPLC analysis confirmed
>97% conversion.
The mixture was cooled to 25−30 °C, and the aqueous phase
was removed. A solution of 1 M HCl (95 kg) was added over
15 min while the temperature was maintained below 30 °C, to
reach a pH of <2.0. MTBE (23 kg) was added, and the mixture
was stirred for 45 min at 20−25 °C. Celite (3 kg) was added,
and the mixture was filtered over a Celite pad (6 kg) wetted
with MTBE (20 kg). The reactor was rinsed with 1 M HCl (11
kg), and the rinse was used to wash the Celite pad. The
combined filtrate was collected in a clean 250 L reactor, and the
lower aqueous phase containing the product was separated and
washed with MTBE (23 kg).
Isopropyl acetate (54 kg) was added to the aqueous phase,
and the mixture was cooled to 10−15 °C. A 25% aqueous
ammonium hydroxide solution (8.2 kg) was added over 15 min
while the temperature was maintained below 15 °C, to reach a
pH of >9. The mixture was stirred for 15 min at 18−20 °C, and
the lower aqueous phase was removed. 3-Mercaptopropyl ethyl
sulfide silica (336 g) was added, and the mixture was stirred for
12−24 h at 20−25 °C. Celite (3 kg) was added, and the
mixture was filtered through a pad of Celite (6 kg) wetted with
isopropyl acetate (13 kg). The filter pad was washed with
isopropyl acetate (2 × 10 kg), and the volume of the filtrate was
reduced by distillation under vacuum until ∼16 L of residue
remained. THF (100 kg) was added, and the distillation under
1H NMR (400 MHz, DMSO-d6): δ 13.01 (br s, 1H), 7.69 (d,
1H), 7.22 (t, 1H), 2.44 (s, 3H). 13C NMR (100 MHz, DMSO-
d6): δ 168.3, 168.2, 158.4, 156.0, 131.6, 131.4, 128.7, 128.6,
127.3, 127.2, 126.2, 126.0, 123.2, 123.1, 13.8, 12.4, 12.3. IR:
1672, 11412, 1278, 1220, 1178, 923, 755, 766 cm−1. MS (EI)
for C9H9FO2S: found 199.1 ([M − H]−). FAB-MS for
C9H9FO2S: found 200.03065 (M+), calcd 200.03066.
3-Fluoro-2-methyl-4-(methylsulfonyl)benzoic Acid
(17). An inert 100 L jacketed reactor was charged with
compound 16 (8.1 kg, 20.5 mol, 4.1 kg corrected for assay) and
glacial acetic acid (20 L, 5 volumes). The resulting slurry was
heated to 60−65 °C under a nitrogen atmosphere, and 30 wt %
H2O2 solution (5.4 L, 52.9 mol, 2.6 equiv) was added at such a
rate that the temperature remained at 80 5 °C; the addition
required 3−4 h. After 1 h at ca. 95 °C, HPLC analysis showed
that <1% of the intermediate sulfoxide remained. Excess
peroxide was neutralized by the addition of DMSO (1.45 L,
20.5 mol, 1.0 equiv) and heating for 1 h at 95 °C, until a starch/
iodide test for peroxides was negative. The solution was treated
with water (55 L, 12 volumes) over ca. 90 min to furnish a thick
white slurry. The mixture was cooled to 25 °C, aged for ca. 2 h,
and filtered. The filter cake was washed twice with water (18 L
each) and dried under suction for 2 h. The product was further
dried at 50 °C under vacuum to give 4.29 kg of 17 (82% yield
for two steps).
HPLC analysis of the isolated solid: 100% LCAP at λ = 230
1
nm and 102 wt %. H NMR (400 MHz, DMSO-d6): δ 13.74
(br s, 1H), 7.82−7.76 (m, 2H), 3.37 (s, 3H), 2.47 (s, 3H). 13C
NMR (100 MHz, DMSO-d6): δ 167.7, 167.7, 159.1, 156.6,
139.3, 139.3, 131.0, 130.9, 128.2, 128.0, 126.8, 126.4, 126.3,
44.1, 44.1, 12.3, 12.2. IR: 1706, 1681, 1568, 1421, 1406, 1383,
1328, 1303, 1264, 1218, 1189, 1150, 1133, 1045, 976, 967, 925,
848.787. 767, 758 cm−1. MS (EI) for C9H9FO4S: found 231
([M − H]−). FAB-MS for C9H9FO4S: found 233.02827 ([M +
H]+), calcd 233.02827.
tert-Butyl 7-Dihydroxyboryl-2,3-dihydro-5H-benzo[f]-
[1,4]-oxazepine-4(5H)-carboxylate (3). An inert 250 L
L
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
vacuum was continued until ∼18 L of residue remained. THF
(130 kg) was added, and the solution was used as such for the
next step. A sample of 5 for analytical characterization was
prepared following trituration of the crude reaction mixture
with hexanes.
filtrate was collected in a clean reactor, and approximately 80−
90 L of was distilled off at approximately 360 mbar. The
mixture was cooled to 8−12 °C, and seeds of 1 (0.5 wt %) and
MTBE (74 kg) were added. The resulting mixture was stirred
for 19 h, after which the product was isolated by filtration,
washed with MTBE (24 kg), and dried under vacuum at 50 °C
to yield 1 as a white solid (7.61 kg, 84%, 98.8% purity by AN-
HPLC). Compound 1 was observed as a mixture of two
1H NMR (400 MHz, CDCl3): δ 8.29−8.28 (m, 1H), 7.63−
7.08 (m, 4H), 6.58−6.56 (m, 1H), 4.63 (br s, 2H), 4.54−4.46
(d, 2H), 4.07 (t, J = 4.4 Hz, 2H), 3.82 (t, J = 4.2 Hz, 2H), 1.42
(s, 9H). 13C NMR (100 MHz, CDCl3): δ 159.0, 157.8, 155.1,
146.3, 136.5, 133.9, 132.7, 127.6, 126.8, 126.7, 122.1, 108.8,
80.4, 73.2, 50.8, 50.0, 28.7. IR (film): 1674, 1636, 1605, 1486,
1459, 1416, 1386, 1365, 1301, 1282, 1269, 1239, 1219, 1167,
1136, 1110, 1072, 1049, 1026, 1013, 978, 944, 883, 818 cm−1.
MS (EI) for C19H23N3O3: found 342 ([M + H]+). FAB-MS for
C19H23N3O3: found 342.18161 ([M + H]+), calcd 342.18162.
5-(2,3,4,5-Tetrahydrobenzo[f ][1,4]oxazepin-7-yl)-
pyridine-2-amine (6). To the THF solution of 5 from the
previous step was added a solution of 37% HCl (18.5 kg, 188
mol, 5.0 equiv) over 30 min while the temperature was
maintained below 35 °C. The mixture was stirred at 34−36 °C
for 12−24 h until HPLC analysis confirmed >98% conversion
to the product. The resulting suspension was cooled to 24−26
°C and stirred for 1 h. The product was isolated by filtration,
washed with THF (40 kg), and dried under vacuum at 47−50
°C to yield 7.48 kg of 6 (63.3% from 2) as a white solid.
1H NMR (400 MHz, CDCl3): δ 8.28−8.27 (m, 1H), 7.63−
7.60 (m, 1H), 7.31−7.29 (m, 1H), 7.27 (s, 1H), 7.10−7.07 (m,
1H), 6.57−6.55 (m, 1H), 4.47 (br s, 2H), 4.07 (dd, J = 4.4, 3.2
Hz, 2H), 4.02 (s, 2H), 3.25 (dd, J = 4.4, 3.2 Hz, 2H), 1.61 (s,
1H). 13C NMR (100 MHz, CDCl3): δ 159.2, 157.4, 146.2,
136.4, 135.5, 133.6, 127.2, 126.9, 126.1, 121.6, 108.5, 75.4, 53.4,
52.4. IR (film): 1660, 1603, 1487, 1443, 1387, 1320, 1284,
1222, 1171, 1128, 1098, 1055, 1011, 985, 925, 905, 893, 848,
817 cm−1. MS (EI) for C14H15N3O: found 242 ([M + H]+).
FAB-MS for C14H15N3O: found 242.12925 ([M + H]+), calcd
242.12926.
1
rotational isomers in the H and 13C NMR spectra.
1H NMR (400 MHz, DMSO-d6): δ 8.24−8.03 (dd, 1H),
7.79−7.71 (m, 1H), 7.71−7.69 (dd, 0.5H), 7.57−7.57 (d,
0.5H), 7.44−7.40 (m, 1.5H), 7.29−7.19 (dd, 1H), 7.05−7.01
(dd, 1H), 6.64−6.63 (d, 0.5H), 6.54−6.45 (dd, 1H), 6.06 (s,
2H), 4.93−4.31 (m, 2H), 4.31−3.54 (m, 4H), 3.37−3.36 (d,
3H), 2.12−1.77 (d, 3H). 13C NMR (100 MHz, DMSO-d6): δ
167.3, 167.2, 166.6, 166.6, 158.9, 158.9, 158.4, 158.4, 157.4,
157.2, 155.9. 155.8, 145.4, 145.1, 145.1, 144.0, 143.9, 135.0,
134.7, 132.9, 132.8, 129.4, 129.2, 128.2, 128.2, 128.1, 128.0,
127.0, 126.9, 126.8, 125.9, 125.6, 125.4, 123.6, 123.5, 123.3,
123.1, 122.8, 122.0, 122.0, 121.9, 121.9, 121.2, 120.7, 107.8,
107.8, 70.9, 70.8, 51.1, 51.1, 47.4, 46.5, 43.5, 43.5, 43.5, 43.4,
11.0, 10.9, 10.7, 10.6. IR (KBr pellet): 1623, 1487, 1457, 1423,
1385, 1314, 1269, 1226, 1193, 1144, 1133, 1054, 1031, 962,
821, 768 cm−1. MS (EI) C23H22FN3O4S: found 456.2 ([M +
H]+). High-resolution MS (FAB-MS using glycerol as a matrix)
for C23H22FN3O4S: found 456.13943 ([M + H]+), calcd
456.13878.
[7-(6-Aminopyridin-3-yl)-2,3-dihydro-1,4-benzoxaze-
pin-4(5H)-yl][3-fluoro-2-methyl-4-(methylsulfonyl)-
phenyl]methanone Hydrochloride (1·HCl). A clean inert
250 L glass-lined steel reactor was charged with acetone (44
kg), purified water (3.65 kg) and free base 1 (7.30 kg, 16.0
mol). The mixture was heated to 50 °C, and 37% HCl (3.16 kg,
32 mol) was added. The mixture was stirred for 10 min, cooled
to 15−20 °C, and stirred for 1 h. The product was isolated by
filtration, washed with acetone (33 kg), and dried under
vacuum at 50 °C to yield 1·HCl as a white solid (7.81 kg, 95%,
99.7% purity by AN-HPLC).
[7-(6-Aminopyridin-3-yl)-2,3-dihydro-1,4-benzoxaze-
pin-4(5H)-yl][3-fluoro-2-methyl-4-(methylsulfonyl)-
phenyl]methanone (1). An inert 250 L glass-lined reactor
was charged with dichloromethane (105 kg), DMF (19 kg),
HOBt hydrate (3.06 kg,20 mol) and 3-fluoro-2-methyl-4-
(methylsulfonyl)benzoic acid (17) (4.64 kg, 20 mol). The
mixture was cooled to 15 °C and EDC·HCl (3.83 kg, 20 mol)
was added. The resulting mixture was stirred for 1 h.
Another inert 250 L glass-lined reactor was charged with
dichloromethane (105 kg), DMF (19 kg), 5-(2,3,4,5-
tetrahydrobenzo[f ][1,4]oxazepin-7-yl)pyridin-2-amine bis-
(hydrochloride) (6·HCl) (6.60 kg, 21 mol), and 3 M NaOH
solution (14.2 kg, 42 mol). The mixture was cooled to 15 °C,
and the activated benzoic acid from the first reactor was added
over 45 min. The resulting mixture was stirred for 18 h.
Stirring was stopped and the phases were allowed to separate.
The aqueous layer was discarded, and the lower organic layer
was washed with 1 M monobasic sodium phosphate solution (2
× 50 kg). DMF (19 kg) was added, and the organic phase was
washed with 1 M NaOH (50 kg). Dichloromethane (80−90 L)
was removed by distillation, and water (53 kg) was added. The
distillation was continued to remove another 30−50 L of
dichloromethane, resulting in a suspension, which was cooled
to 15−20 °C, filtered, and washed with water (25 kg).
The wet cake was returned to the reactor, and THF (128 kg)
and water (11.4 kg) were added. The mixture was heated to
reflux, and the solution was polish-filtered while hot. The
Analyses: OVI: DMF < 100 ppm, DMC < 100 ppm, acetone
= 3081 ppm, MTBE < 100 ppm, iPAc < 100 ppm, THF < 100
ppm. Heavy metals: Pd ≤ 0.2 ppm, others < 20 ppm (USP
1
⟨231⟩). H NMR (400 MHz, DMSO-d6), equimolar amounts
of two rotamers: δ 8.20−8.40 (br s, 2H), 8.33 (s, 0.5H), 8.31
(d, J = 2.8 Hz, 0.5H), 8.15 (d, J = 2.0 Hz, 0.5H), 7.96 (dd, J =
9.7, 2.0 Hz, 0.5H), 7.70−7.78 (m, 1.5H), 7.55−7.57 (m, 0.5H),
7.51−7.55 (m, 0.5H), 7.28 (d, J = 8.6 Hz, 0.5H), 7.17 (d, J =
3.1 Hz, 0.5H), 7.15 (d, J = 5.1 Hz, 0.5H), 7.05−7.11 (m, 1.5H),
6.83 (d, J = 2.7 Hz, 0.5H), 4.86−4.99 (m, 1H), 4.29−4.56 (m,
1H), 4.10−4.27 (m, 2H), 3.93−4.04 (m, 0.5H), 3.45−3.65 (m,
1.5H), 3.37 (s, 1.5 H), 3.35 (s, 1.5H), 2.12 (d, J = 2.0 Hz,
1.5H), 1.76 (d, J = 2.0 Hz, 1.5H). 13C NMR (100 MHz,
DMSO-d6), equimolar amounts of two rotamers: δ 168.1,
167.5, 159.4, 159.2, 159.1, 156.6, 153.9, 153.8, 144.6, 142.9,
142.3, 133.0, 132.7, 130.0, 129.9, 129.7, 129.5, 129.1, 129.0,
128.9, 128.8, 128.5, 127.7, 127.6, 127.5, 127.1, 126.9, 124.4,
124.3, 124.1, 122.7, 122.1, 121.6, 114.4, 71.2, 51.7, 51.3, 47.9,
46.9, 44.3, 44.2, 11.7, 11.4.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: sriram.naganathan@dermira.com.
M
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(11) Mitsunobu, O. Synthesis 1981, 1. Hughes, D. L. Org. React. 1992,
42, 335. For a recent review, see: But, T. Y. S.; Toy, P. H. Chem.
Asian J. 2007, 2, 1340.
(12) Lane, C. F. Chem. Rev. 1976, 76, 773. Brown, H. C.;
Krishnamurthy, S. Aldrichimica Acta 1979, 12, 3. Ho, T.-L.; Fieser,
M.; Smith, J.; Fieser, L. Borane-Tetrahydrofuran. In Fieser and Fieser’s
Reagents for Organic Synthesis; Wiley: New York, 2006. Young, S. D.;
Wiggins, M. J.; Huff, J. R. J. Org. Chem. 1988, 53, 1114.
Present Addresses
§S.N.: Dermira, Inc., 275 Middlefield Road, Suite 150, Menlo
Park, CA 94025.
∥D.A.: Threshold Pharmaceuticals, 170 Harbor Way, Suite 300,
South San Francisco, CA 94080.
⊥N.G.A.: Astex Pharmaceuticals, Inc., 4140 Dublin Boulevard,
Suite 200, Dublin, CA 94568.
#S.L.: Gilead Sciences, 333 Lakeside Drive, Foster City, CA
94404.
(13) Structural assignment of the regiochemistry of the acid−amide
1
is tentative and based on preliminary spectral analyses (MS and H
NMR). It is possible that the acid and amide groups are opposite of
what is shown. No attempt was made to rigorously assign the
regiochemistry.
Notes
The authors declare no competing financial interest.
(14) Edward, J. T.; Terry, K. A. J. Chem. Soc. 1957, 3527.
(15) Atkins, W. J.; Burkhardt, E. R.; Matos, K. Org. Process Res. Dev.
2006, 10, 1292. Atkins, W. J. Borane-THF BASF Safety and Handling
Presentation; BASF Corporation: Evans City, PA, 2009.
(16) Luk’yanov, B. S.; Maksimenko, A. A.; Borodkin, G. S.;
Maksimenko, A. A. Chem. Heterocycl. Compd. 2003, 39, 1539.
(17) De la Mare, P. B. D. Electrophilic Halogenation; Cambridge
University Press: Cambridge, U.K., 1976.
(18) While both solid NaSMe and NaSMe in aqueous solution are
commercially available, use of the aqueous solution is very inefficient,
and it is very difficult to completely consume 15.
(19) Douglass, I. B.; Martin, F. T. J. Org. Chem. 1950, 15, 795.
Douglass, I. B. J. Am. Chem. Soc. 1952, 74, 5770. Senning, A.;
Lawesson, S.-O. Acta Chem. Scand. 1962, 16, 117.
ACKNOWLEDGMENTS
■
We gratefully acknowledge Otute Akiti and Sean Cusack for
evaluating the safety risks associated with the borane reduction
and oxidation steps. We thank Nathan Guz, Matthew Pfeiffer,
and Norma Tom for their assistance with the borane reduction;
James Robinson, Tri Tran, and Sharique Zuberi for their
assistance with the imide synthesis; James Tam, Jim Zhan, and
Shigang Zhang for help with analytical development and
support; Rahul Manek and Shital Rane for their contributions
to salt and form screening and obtaining solid-state data; Bill
Atkins (BASF Corporation, Evans City, PA) for helpful
suggestions and guidance regarding safe handling of borane·
(20) Lau, S. Y. W.; Hughes, G.; O’Shea, P. D.; Davies, I. W. Org. Lett.
2007, 9, 2239 and references therein.
(21) Ito, S.; Kubo, T.; Morita, N.; Matsui, Y.; Watanabe, T.; Ohta, A.;
Fujimori, K.; Murafuji, T.; Sugihara, Y.; Tajiri, A. Tetrahedron Lett.
2004, 45, 2891.
THF; and Jens Wittenberg and Martin Volkert (BASF
̈
Corporation, Ludwigshafen, Germany) for their efforts in the
scale-up of the borane reduction. We also thank Jo Ann Wilson
and Norma Tom for their help in reviewing the manuscript.
(22) Brown, H. C. Boranes in Organic Chemistry; Cornell University
Press: Ithaca, NY, 1972. While the structure of boroxine 30 was not
rigorously established, support for the structural assignment was
obtained by subjecting a sample of boronic acid 3 to dehydrating
conditions. HPLC analysis revealed that 3 was cleanly converted to 30.
(23) Guideline on the specification limits for residues of metals
catalysts or metal reagents, EMEA/CHMP/SWP/4446/2000. ICH
Q3D Guideline for Elemental Impurities (ICH, Dec 16, 2014).
(24) Galaffu, N.; Man, S. P.; Wilkes, R. D.; Wilson, J. R. H. Org.
Process Res. Dev. 2007, 11, 406. It was not immediately apparent at the
time of this work that a Pd-scavenging operation was compulsory, and
the process as designed may have been able to reduce the Pd content
to acceptable levels. However, because of constraints of time and
project need, a scavenging operation was introduced in order to be
certain.
REFERENCES
■
(1) Fruman, D. A.; Rommel, C. Nat. Rev. Drug Discovery 2014, 13,
140 and references therein. Nuss, J. M.; Tsuhako, A. L.; Anand, N. K.
Annu. Rep. Med. Chem. 2009, 44, 339 and references therein.
(2) Anand, N.; et al. Benzoxazepines as Inhibitors of PI3K/mTOR
and Methods of their Use and Manufacture. U.S. Patent 8,648,066,
Feb 11, 2014. Aay, N.; et al. Benzoxazepines as Inhibitors of PI3K/
mTOR and Methods of their Use and Manufacture. U.S. Patent
8,637,499, Jan 28, 2014.
(3) Kamei, K.; Maeda, N.; Nomura, K.; Shibata, M.; Katsuragi-Ogino,
R.; Koyoma, M.; Nakajima, M.; Inoue, T.; Ohno, T.; Tatsuoka, T.
Bioorg. Med. Chem. 2006, 14, 1978. Chupak, L. S.; et al. Preparation of
N-Substituted Bicyclic Oxazolidinones as Antibacterial Agents. WO
2004069244 A1, Aug 19, 2004.
(25) ICH Q3C (R5) Impurities: Guideline for Residual Solvents (ICH,
Feb 4, 2011).
(4) Schmidt, K. F. Ber. Dtsch. Chem. Ges. 1924, 57B, 704. Wolff, H.
Org. React. 1946, 3, 307.
(5) Ashby, E. C.; Schwartz, R. D. Inorg. Chem. 1972, 11, 919. Ashby,
E. C.; Lin, J. J.; Goel, A. B. J. Org. Chem. 1978, 43, 1557.
(6) Kwiecien, H.; Baumann, E. Pol. J. Chem. 2000, 74, 141.
́
(7) Green, T. W.; Wuts, P. G. M. Protective Groups in Organic
Synthesis; Wiley-Interscience: New York, 1999; pp 518−525, 736−739.
(8) Takeuchi, C. S.; Kim, B. G.; Blazey, C. M.; Ma, S.; Johnson, H.
W. B.; Anand, N. K.; Arcalas, A.; Baik, T. G.; Buhr, C. A.; Cannoy, J.;
Epshteyn, S.; Joshi, A.; Lara, K.; Lee, M. S.; Wang, L.; Leahy, J. W.;
Nuss, J. M.; Aay, N.; Aoyama, R.; Foster, P.; Lee, J.; Lehoux, I.;
Munagala, N.; Plonowski, A.; Rajan, S.; Wollfrey, J.; Yamaguchi, K.;
Lamb, P.; Miller, N. J. Med. Chem. 2013, 56, 2218. Anand, N.; et al.
Benzoxazepine Derivatives as mTOR Kinase Inhibitors and Their
Preparation and Use for the Treatment of Cancer. WO 2010118208,
Oct 14, 2010. Schmitt, M.; et al. Preparation of 3-Cyclopropyl-4-(3-
amino-2-methylbenzoyl)pyrazole Derivatives as Herbicides. WO
2006061074, June 15, 2006.
(9) For example, see: Knochel, P.; Dohle, W.; Gommermann, N.;
Kneisel, F. F.; Kopp, F.; Korn, T.; Sapountzis, I.; Vu, V. A. Angew.
Chem., Int. Ed. 2003, 42, 4302.
(10) Trost, B. M.; Curran, D. P. Tetrahedron Lett. 1981, 22, 1287.
N
DOI: 10.1021/acs.oprd.5b00037
Org. Process Res. Dev. XXXX, XXX, XXX−XXX