Commercial synthesis of a pyrrolotriazine-fluoroindole intermediate to brivanib alaninate: Process development directed toward impurity control

Article abstract of DOI:10.1021/op400242j

The development of a practical, commercial process for the preparation of 4-fluoro-2-methyl-indol-5-ol and its subsequent coupling with a pyrrolotriazine to form an advanced intermediate of the oncology therapy brivanib alaninate is described. A key aspect is the multikilogram-scale preparation of the fluoroindole intermediate from trifluoronitrobenzene and the subsequent coupling while achieving impurity minimalization. As brivanib alaninate is a high-dose drug, the synthesis of highquality API with low levels of impurities is critical.

Full text of DOI:10.1021/op400242j

Article
pubs.acs.org/OPRD
Commercial Synthesis of a Pyrrolotriazine−Fluoroindole
Intermediate to Brivanib Alaninate: Process Development Directed
toward Impurity Control
Jaan A. Pesti,* Thomas LaPorte, John E. Thornton, Lori Spangler, Frederic Buono, Gerard Crispino,
Frank Gibson, Paul Lobben, and Christos G. Papaioannou
Early Phase Chemical Development, Bristol-Myers Squibb Company, New Brunswick, New Jersey 08923, United States
ABSTRACT: The development of a practical, commercial process for the preparation of 4-fluoro-2-methyl-indol-5-ol and its
subsequent coupling with a pyrrolotriazine to form an advanced intermediate of the oncology therapy brivanib alaninate is
described. A key aspect is the multikilogram-scale preparation of the fluoroindole intermediate from trifluoronitrobenzene and
the subsequent coupling while achieving impurity minimalization. As brivanib alaninate is a high-dose drug, the synthesis of high-
quality API with low levels of impurities is critical.
included the identification of a ‘quality gatekeeper’ intermedi-
ate, a pivotal intermediate that could be prepared pure or easily
purified prior to embarking on the final steps to the API.3 If
successful, the final steps leading to the API would not be
burdened with input-related impurities, a key attribute for a
successful process.
INTRODUCTION
Background and Goals of Process Development.
Brivanib alaninate (1) is an investigational oncology therapy
with the potential for applications against a wide variety of
tumor types and stages of disease progression (Figure 1). The
■
A manufacturing-scale process to make the proposed starting
material pyrrolotriazine core 2 had been previously established
from glycine ethyl ester, ethyl acetoacetate, dimethylformamide
dimethyl acetal and formamide.4 However, the ester moiety of
2 was not at the proper oxidation state for conversion to 1 and
required carbon−carbon bond cleavage, a nontrivial trans-
formation. As for most oxidative cleavages of this sort, the
chemistry required is energetic and potentially hazardous. For
instance, a runaway reaction resulted when an oxidative
cleavage reaction of a derivative of 2 was heated to 60 °C.
Accordingly, a continuous process was developed that over-
came the safety concerns, permitting the metric-ton-scale
manufacture of 3. This molecule was established as the first
regulatory starting material for brivanib alaninate.5 The
preparation of the second proposed regulatory starting material,
fluoroindole 4, existed, but additional process development was
required for large-scale manufacture.
Our team initially engaged in an exploratory program of
process scouting to understand the intrinsic parameters that
impacted the rate, purity, and subsequent workup of the
coupling of 3 and 4 to form the indole−pyrrolotriazine 5. This
work established that 5 was thermally stable, highly crystalline,
and could be facilely purified in high yield on at least 100 g
scale. A particularly attractive feature was the ability of
recrystallization to significantly purge the majority of the
precursors of 3 and 4 to undetectable levels while obtaining
high recoveries of substrate. These qualities supported the
selection of 5 as the aforementioned quality gatekeeper for the
entire process. This permitted the chemical development
program for API to proceed with the understanding that a
Figure 1. Structure of brivanib alaninate (1).
molecule is built upon the versatile medicinal properties of a
pyrrolotriazine core which has been incorporated as well into
several other Bristol-Myers Squibb drug candidates.1 This L-
alanine ester prodrug is an orally available, dual inhibitor of the
vascular endothelial growth factor receptor-2 (VEGFR-2) and
fibroblast growth factor receptor (FGFR) tyrosine kinases.
Early studies indicated likely broad-spectrum antitumor activity
for use in the treatment of hepatocellular carcinoma, colorectal
cancer, and fibroblast growth factor-driven tumors by inhibition
of angiogenesis.2 Clinical and preclinical studies have shown the
important role that the VEGF receptor family of trans-
membrane protein tyrosine kinases have on angiogenesis. A
viable process to make multi-hundred kilogram batches was
required to support clinical investigation of these properties
and establish a commercially viable synthesis.
Our strategy for the process development of brivanib was
focused on removing input-related impurities and avoiding/
minimizing formation of new impurities during the preparation
(Scheme 1). The emphasis on eliminating impurities was
significant since brivanib alaninate is expected to be dosed at
∼800 mg/d, implying any associated impurities could be
ingested at biologically impactful levels. As an aid to achieve
high-purity intermediates, our early process development work
Received: September 1, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1. Convergent synthesis of brivanib alaninate (1)
Scheme 2. Commercial synthesis of fluoroindole 4
process to produce highly pure 5 would be defined, a useful
quality for the immediate precursor of a high-dose drug.
Ultimately, 5 was condensed sequentially with (R)-propylene
oxide (after PIV cleavage) and then a CBZ-protected synthon
and was finally hydrogenated to reveal the drug substance.6
Herein we describe the design and development of our
optimized commercial process, producing the key intermediate
5 in single batches of up to ∼300 kg in 93% yield of ≥99.95%
purity.7,10
from trifluoronitrobenzene while overcoming hazardous
chemistry and the sensitivity of the product and precursors
toward oxidation or basicity. On the basis of the successful
development of this scalable route, 4 became our second
regulatory starting material.
Process Development of Arylpropanone 9. The reaction of
trifluoronitrobenzene 6 with a mixture of potassium tert-
butoxide and ethyl acetoacetate in THF led to ≥99.9%
conversion to the α-ketoester 7.9 The reaction was reprodu-
cible, but variable purity of the product (75−95% purity) was
correlated to the level of KOH in the base, which was remedied
by setting a <5% specification for KOH. The workup was
straightforward, but as the product 7 is an oil, it was more
efficient to concentrate the worked-up product solution and
telescope it into the following decarboxylation step. The
presence of the nitro group required consideration for
explosivity and exothermic events. As such, this and all the
subsequent steps were evaluated by using a combination of
RESULTS AND DISCUSSION
■
Preparation of Fluoroindole 4. An attractive precursor to
fluoroindole 4 is 2,3,4-trifluoronitrobenzene (6, Scheme 2), as
it is inexpensive, available in bulk and properly functionalized to
efficiently form the indole over five steps, concluding with a
modified Reissert indole synthesis.8 Following modification of
an existing route,1b our process has produced metric tons of 4
B
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
DSC, calorimetry, ARC, and other safety-testing protocols to
be assured of the safe operating ranges.11
reducing reagents17 via a Reissert indole cyclization.8 The use
of aqueous sodium dithionite or sodium hydrosulfite as
reductants had the advantages of low cost, safety and mildness
so as not to further impact the sensitive fluoroindole entity. A
trace of HCl was added to all solutions of the process as the
product 4 is not stable at pH 5−8. A further advantage of this
reaction system was the rapid crystallization of the freshly
formed 4 from the reaction medium, isolating the sensitive
product from the remaining reactants.
The reaction proceeded once a methanolic solution of
phenol 10 was added to an aqueous solution of sodium
dithionite. The reaction had to be conducted expeditiously
once the dithionite solution was prepared as it is not stable
under acidic conditions. Upon reduction, the aniline immedi-
ately cyclized and crystallized from the solution. Following
cooling and workup, this process resulted in 85−90% yields of
fluoroindole.18 While requiring administrative controls for
immediate use of the dithionite solution, it is safe, scalable
and high-yielding. The primary challenge now consisted of
addressing the formation and removal of the remaining
impurities, and defining conditions for storage and shipment.
Improper isolation or storage conditions (failure to maintain an
acidic pH or temperature >8 °C) would lead to decomposition
and unacceptable material.
The formation of side products and decomposition products
in this step was extensive at the onset of our investigations. The
missing yield was primarily due to electrophilic aromatic
substitution by SO3 to produce the three arylsulfonate
analogues of 4 (LC−MS analysis). While such polar impurities
were easily washed away with water, the level of sulfonation
varied widely (10−40%), impacting yields. Higher levels of
impurities were correlated to the use of degraded sodium
dithionite; thus, only freshly opened batches of recently
manufactured sodium dithionite were used. The fluoroindole
was also prone to form a variety of impurities such as dimers
and higher oligomers (Figure 2). At pH 5−8, the indole
The subsequent decarboxylation to the propanone 8 was
conducted by heating a H2SO4/acetic acid solution of 7 to 70
°C, achieving a >99.7% conversion. Workup and concentration
resulted in >100% as-is yields of a brown oil and 75−80%
purity in batches up to 450 kg of 6.12 Once again this material
was telescoped into the following step in order to reach a
crystalline intermediate more amenable to purification and
isolation.
The nascent hydroxy group was introduced by methoxide
displacement of the fluorine at the 3-ring position to form
anisole derivative 9.13 The reaction was conducted by refluxing
8 in methanol with 0.8 equiv of K2CO3 (≥99.5% conversion).
None of the regioisomeric aryl-methyl ether was detected due
to the strong directing influence of the nitro group. Screening
the base charge determined that substoichiometric base was
optimum in order to minimize deprotonation of the benzyl
position, which would have deactivated the ring to further
nucleophilic attack.14 The anisole derivative 9 crystallized upon
addition of water, and subsequent MTBE recrystallization along
with charcoal treatment effectively removed any impurities
carried along over the prior two telescoped chemical steps. The
optimized process produced 75−80% yields (≥99.7% purity) in
batches of 8 up to 600 kg.
Synthesis of Nitroaryl Propanone 10. Demethylation of 9
was initially achieved with pyridine hydrochloride1k in a
solvent-free medium by mixing the solids and heating to 160
°C. Following completion, the viscous reaction mixture was
diluted with hydrochloric acid, followed by a tedious
purification/isolation/recrystallization to produce 85−90%
yields of >99.5% purity nitroaryl propanone 10. Complications
for this protocol included inputs that were unstable to the
reaction conditions, handling black viscous oils, formation of
poorly soluble tars, and interfaces that were difficult to detect
during phase separations. The use of carbon treatment
ameliorated some inconveniences, but a superior remedy was
distinctly needed for further scale-up.
Ionic liquids15 are usually used as solvents in that their
atypical physical conditions permit reactivities and selectivities
difficult to achieve in other ways. However, they can also be
used as reagents. Demethylation of the anisole derivative 9
occurs in the chloroaluminate ionic liquid formed from the
complexation of aluminum trichloride and trimethylammonium
chloride in CH2Cl2.15b The resulting reagent, [TMAH]Al2Cl7,
acts as a powerful Lewis acid, coordinating the ethereal oxygen
and strongly activating the methyl group toward attack by
chloride. Subsequent departure of chloromethane produced 10
as a result. The unusual Lewis acidity resulted in a reduction in
the temperature necessary for demethylation from 160 °C in
pyridine hydrochloride to 40 °C under the modified conditions.
At multikilogram scale, trimethylammonium chloride was
added to a slurry of AlCl3 in anhydrous CH2Cl2, followed by
addition of a solution of 9.16 Holding this mixture at 40 °C led
to ≥99.9% methyl cleavage. The chloromethane offgas was
quenched by directing it through a methanolic solution of
aqueous ammonia. The workup progressed by the addition to
water to quench the chloroaluminate species and precipitate the
product as a crystalline solid. High-purity nitroaryl propanone
10 was isolated by centrifugation. Yields from 9 were 85−95%
at >99.4% assay in batches of up to 375 kg.
Figure 2. Oligomers of fluoroindole 4.
polymerized via two distinct pathways. The fluorine could be
displaced by the stabilized phenoxide oxygen to form a
diarylether impurity, or the iminium tautomer would lead to
a C−C-bonded coupling product. The aryl-ether impurity also
further oligomerized to produce impurities up to heptamers19
until the coformed HF increased the acidity of the mixture and
rendered remaining 4 stable. As expected, these sources of
impurities were easily suppressed by maintaining the pH below
4, leading to our protocol of adding dilute hydrochloric acid to
the reaction mixture and all solutions subsequently used in the
isolation to promote compound stability.
Furthermore, the crude fluoroindole typically contained
200−300 ppm of unreacted 10 due to the rapid crystallization
of 4 from the reaction mixture, trapping other components.
Modification of reaction conditions did not further reduce the
Synthesis of Fluoroindole 4. The aryl nitroketone 10 may
be reduced and cyclized to the fluoroindole 4 by a range of
C
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 3. Discovery Chemistry’s preparation of coupled intermediate analogues to 5
levels of residual 10. Thus, a recrystallization was mandated for
the commercial procedure wherein it had been optional for the
earlier protocols. Our recrystallization protocol was based on
dissolution of 4 into cold acidic aqueous methanol followed by
the addition of dilute aqueous HCl to initiate recrystallization.
Filtration of the white crystals which resulted followed by cake
washes with dilute HCl led to 65−75% recoveries of 4.
Preparation of the fluoroindole was piloted successfully up to
282 kg batches, producing 65−75% yields with >98.0% assay
(>99.75 HPLC area).
Figure 3. Sulfur-containing impurities resulting from sodium
metabisulfite addition.
Fluoroindole 4 Stability and Storage Considerations. The
synthetic and purification problems had been solved, but
fluoroindole 4 still required storage at <8 °C to avoid formation
of dimeric impurities. This introduced inconvenience (temper-
ature recorders added to shipments), expense (need for
refrigerated warehouses) and risk (inadvertent warming leading
to compromised product). Of note, our earliest fluoroindole
preparations had not included a recrystallization operation and
were thermally stable in comparison to these latter recrystal-
lized batches. We hypothesized that the addition of the
clarification step which required the filtration of the aqueous
methanolic solution prior to recrystallization, removed a
component that had imparted stability to non-recrystallized
batches. Examination of material removed during the
clarification revealed polar solids that largely consisted of
inorganic sulfur salts. These insoluble salts were derived from
sodium dithionite and most likely retained reducing potential,
possibly protecting the non-recrystallized fluoroindole from
oxidative decomposition pathways that in turn led to the
impurities.
Clearly, some desirable stabilization attribute of fluoroindole
4 was being removed, but rather than return to a process with a
critical variable we could not well control, we screened
antioxidants and identified sodium metabisulfite as a useful
component. If added to the cake washes, the process produced
4 with sulfur levels of 800−7000 ppm (IPC) which provided a
measure of metabisulfite incorporation. The resulting 4
displayed acceptable stability upon accelerated stability testing.
Of less desirability, this modification also produced sulfur-
bridged dimer impurities (Figure 3) as high as 0.3% in
otherwise acceptable batches of fluoroindole.
This process for fluoroindole 4, the second regulatory
starting material defined for brivanib alaninate, has been
satisfactory over several campaigns. Batches of up to 144 kg of
fluoroindole 4 have been made without issue, and over 1 MT of
material has been prepared to date.
Coupling Reaction Process Development for Preparing
the Indole−Pyrrolotriazine 5. Quality Gatekeeper 5 and
Initial Considerations of the Process. We initiated our process
development of the coupling reaction sequence using chemistry
developed by our Discovery colleagues (Scheme 3).1 The
pyrrolotriazine 2 was chlorinated with phosphorous oxychloride
(POCl3) to form the chlorotriazine without isolation after
workup. Subsequent addition of phenol formed a diaryl ether as
a protecting group. The ester was reduced to the aldehyde over
two steps by reduction with lithium aluminum hydride (LAH)
followed by reoxidation by manganese dioxide. Carbon−carbon
bond oxidative cleavage proceeded subsequently by a Baeyer−
Villiger reaction induced by m-chloroperbenzoic acid (m-
CPBA) followed by formate cleavage.
The resulting hydroxyl group was protected as a methyl or
benzyl ether followed by cleavage of the phenyl ether to reveal
the triazine ring hydroxyl. Following another POCl3 chlorina-
tion, addition of fluoroindole 4 furnished the coupled
compound. This formed an analogue to indole−pyrrolotriazine
5 that could be further converted onto brivanib alaninate 1.
While this protocol fully satisfied Discovery Chemistry’s need
for synthetic flexibility to prepare analogues, the nine steps
expended for the functionalization of only one group would be
an issue for further scale-up. While the intermediates
themselves might be viable components for a commercial
process, it was clear a new process was required before
multikilogram preparations could be considered.
We developed methods to reduce the content of these new
impurities to <0.05%, but even such levels led to daughter
impurities three steps later which significantly reduced the
hydrogenation rate to form API. In view of this development,
the metabisulfite treatment was abandoned, the former
recrystallization conditions were reestablished ,and cold storage
for isolated product is now mandated.
Pyrrolotriazine 3 Process Evolution (alternative routes). A
logical change would be incorporation of the indole portion
first in place of the phenol protecting group, permitting
elimination of a protection/deprotection cycle, and then
addressing the C−C cleavage (Scheme 4).
Chlorination followed by addition of fluoroindole 5
produced the expected coupled compound 14 in good yield.1i
D
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 4. Early routes to prepare des-PIV 5
Scheme 5. Conversion of pyrrolotriazine 2 to PIV-protected pyrrolotriazine 3
However, as indoles are prone to oxidation,20 unacceptable
impurity content occurred when the ester was subsequently
oxidatively cleaved. In one approach, the ester was first
exhaustively methylated to form the tertiary alcohol 15. A
subsequent acid-catalyzed oxidative cleavage formed the
substituted pyrrolotriazine 16.5,21 Alternatively, LAH ester
reduction and subsequent tetramethylpiperidine oxide
(TEMPO)-catalyzed oxidation produced the benzylic aldehyde.
E
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 6. Commercial preparation of coupled intermediate 5
Figure 4. FTIR spectroscopy of DIPEA-catalyzed and DIPEA-free chlorinations.
A subsequent Baeyer−Villiger reaction cleaved the carbon−
A better solution for manipulation of energetic intermediates,
carbon bond via m-CPBA. Both reactions were low yielding and
generated numerous impurities.
particularly on multikilogram scale, is the use of flow
chemistry.24 A continuous oxidation by an acid-catalyzed
benzylic hydroperoxide rearrangement was developed which
did not require the prior protection of the triazine ring hydroxyl
group (Scheme 5, route 3).5,25 This proceeded by initial
MeMgBr addition to form the tertiary alcohol, followed by a
carefully defined continuous reaction with hydrogen peroxide
and methanesulfonic acid. The stream was quenched and
worked up to crude 17. Rather than isolate this amorphous and
hydrophilic diol, subsequent pivalate protection was specific for
the pyrrole ring hydroxyl to form crystalline pyrrolotriazine 3.
The PIV protection would have been required in any event as
chlorination with POCl3 on 17 is not regiospecific. This process
from the ester 2 to PIV-protected 3 has been successfully used
by vendors to make several metric tons of pyrrolotriazine 3.
While not a robust process since the continuous oxidation
conditions are narrowly defined, it is now a safe and
reproducible process to permit commercial preparation of a
key starting material.
Clearly, the synthetic plan of shifting the oxidative cleavage
to after the coupling was not viable in view of the indole’s
sensitive nature. The process team explored a wide variety of
alternative routes to synthesize the pyrrolotriazine core in the
correct oxidation state; however, none were superior to the
existing chemistry. Instead, if the strategy was to keep the
coupling step after the establishment of the pyrrole−oxygen
bond, an additional benefit was that convergence would be
established only a few steps from the API. It would fit in well
with our overall strategy of impurity reduction as the product of
a convergent synthesis should be easier to purify.22
The intrinsically hazardous oxidation chemistry would thus
be addressed first at the pyrrolotriazine 3, a superior substrate
with regards to stability as compared to the indole−
pyrrolotriazine 14 (Scheme 5). The ester was converted to
the methyl ketone using the Weinreb amide (route 1),23 but
oxidation under Baeyer−Villiger conditions failed to further
convert to 17. If the aldehyde was formed instead using LAH
reduction and TEMPO oxidation to the aldehyde (route 2),
oxidation with mCPBA did form the diol 17 in 48% overall
yield. This was considered for large-scale work, but route 3 led
to success before this was developed.
Chlorination of Pyrrolotriazenol 3. There is good general
precedent for triazine 3 chlorination with POCl3 to
chloroimidate 18, followed by addition of the fluoroindole 4
to form the indole−pyrrolotriazine 5. (Scheme 6)1m,26,27 We
determined that chlorination proceeded in almost all nonprotic
F
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 7. Hypothesized chlorination mechanism to form chloroimidate 18
solvents examined,28 and that, while a base was not necessary,
the rate and purity was improved by the presence of amine
bases.29 Our optimized final process consisted of charging 2.0
equiv POCl3 to a slurry of 3 in acetonitrile followed by a charge
of 0.9 equiv DIPEA at <40 °C, leading to a clean >99.5%
chlorination conversion after 5−9 h of reflux. The reaction was
quenched into aqueous dipotassium phosphate (K2HPO4).
Isopropyl acetate was included to allow a phase separation from
the aqueous quench layer. A wash with aq K2HPO4 adjusted
the pH to 8−9. Product solutions treated in this manner
possessed 10 days of stability against chloroimidate hydrolysis
back to 3. The solution of 18 obtained from the workup
consisted of a water-saturated solution of isopropyl acetate and
acetonitrile, acceptable for the subsequent coupling reaction.
However, further examination of the chlorination mechanism
revealed significantly more complexity to be understood before
a robust process would be attained.
significant, as once the first chloride has reacted, the remaining
chlorides are kinetically less reactive.31
These observations can best be explained by DIPEA
catalyzing the first step to form interconverting chlorinated
phosphate intermediates followed by a slow step of chloride
(from HCl) displacement of the dichlorophosphonate moiety
to form chloroimidate 18. In comparison, the rate-determining
step for the reaction without base was the initial formation of
the chlorinated phosphate intermediate which then would
subsequently rapidly rearrange to chloroimidate 18. The
transient dichlorophosphonates would not be detected by
FTIR.
While the initial chlorination rates were comparable, the
absence of DIPEA also produced a reaction mixture containing
numerous side products, resulting in a less pure and lower-
yielding product solution. These studies provided clear support
that the inclusion of DIPEA was advantageous, but the
optimum charge remained to be determined.
Optimization of Chlorination Kinetics and Reaction
Parameters. While there are numerous instances of similar
chlorinations,1m,27 the mechanistic understanding of rate and
formation of impurities often differ significantly between
examples and prevent drawing broader conclusions.30 Using
FTIR spectroscopy, different kinetics to form 18 became
apparent, depending on whether DIPEA was present. When
POCl3 was added to a solution of triazine 3 and DIPEA, an
intermediate’s signal immediately appeared and slowly decayed
in intensity as that of chloroimidate 18 grew steadily; whereas
without DIPEA, the rates of 18’s formation and 3’s
disappearance were nearly equal (Figure 4) without detection
of an intermediate. HPLC analysis of the DIPEA-containing
reaction revealed three transient intermediates (tentatively
identified as isomers a, b, and c; Scheme 7) of equivalent mass
attributed to the three possible regioisomeric chlorinated
phosphate intermediates, on the basis of mass spectroscopy
and phosphorous/proton NMR spectroscopy.37,38 Unlike a
similar substrate,30 apparently the possible oligomeric inter-
mediates occasionally seen for other substrates consisting of a
single POCl3 bridging two or more triazine cores are not
The reaction space parameters for DIPEA were determined
via a study comparing equivalents of DIPEA versus reaction
completion and quality for 2.0 equiv of POCl3. This established
that 0.85−0.95 equiv of DIPEA versus triazine 3 produced the
fastest chlorination completion and highest purity of 18. When
more than 1.1 equiv of DIPEA was introduced, the chlorinated
phosphate intermediates only slowly converted to 18. However,
if water was subsequently added, the chlorination rate
increased.
As stated, if no DIPEA is charged, the initial conversion to
the phosphonate ester is slow and rate determining. However,
once the chlorinated phosphate intermediates do form, the
abundance of HCl present due to the absence of base rapidly
converts the dichlorophosphonate onto chloroimidate, and no
intermediate is observed in an overall slow reaction. If 1.0 equiv
or more of DIPEA is present, the first step to chlorinated
intermediates remains rapid. However, the base effectively
complexes all the HCl formed in the first step, and the further
conversion to the chloroimidate 18 becomes rate determining
due to the low concentration of free acid. In this case, the
G
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
phosphonate ester intermediate(s) are detected as long-lived
intermediate(s), but if water is then subsequently added,
hydrolysis of excess POCl3 forms more HCl to drive the second
step. When the optimized level of base (0.9 equiv) is present,
the substochiometric base charge leaves sufficient HCl free32
for the nucleophilic chloride displacement of the dichlor-
ophosphate. The second step for formation of 18 is still rate
determining, but the overall reaction is faster.
Optimization of POCl3 Charge. Stress testing indicated that
as little as 0.7 equiv of POCl3 would lead to high conversions of
chloroimidate, but as the rate is correlated to the POCl3 charge,
reaction completion would have been unreasonably long. In
addition, it was important to avoid undercharging POCl3 below
1.8 equiv as a dimeric impurity, 19A (Figure 5), began to form
advantageous in that a new self-condensation impurity (19B)
formed.35
Understanding of the mechanism and kinetics of the
chlorination of 3 in acetonitrile has allowed the development
of a reaction, workup and storage protocol that met our needs
for a commercial preparation. This procedure has been
reproducible on large scaleour last campaign produced 10
batches totaling ∼1 MT of 18 in solution of 99.5−99.6% HPLC
area %.
Coupling with Fluoroindole To Synthesize Indole−
Pyrrolotriazine 5. Coupling Process Using Acetonitrile and
Ethyl Acetate As Solvents. In comparison to the precise
conditions needed for optimal chlorination, the coupling
reaction of chloroimidate with fluoroindole will proceed well
under a wide variety of reaction conditions. We substituted
DABCO for NaH, which was not safe36 and unnecessarily
strong (pKa = ∼35).37 The base was charged to a mixture of the
fluoroindole 4 and the chloroimidate 18 and heated to 50 °C to
complete the coupling. Dilution of the acetonitrile reaction
mixture with water to isolate the crystalline product produced
90−91% yields and 99.4−99.7% purity.
This coupling procedure remained uneventful for hundreds
of chlorinations until one reaction unexpectedly produced 8
mol % of a new impurity: the overcondensation product 21.
Subsequent investigation established that this impurity had
formed because of an unknown acidic impurity present in the
coupling vessel which catalyzed the formation of 20 (Scheme
8). Upon addition of DABCO, 20 then further reacted to form
the trimeric 21.
As little as 5 mol % of strong acids catalyzed the formation of
20. This sensitivity to acid was worrisome on large scale. The
possibility of forming 21 was eliminated by a simple change in
the process charge order. Instead of adding DABCO last, the
coupling vessel was now initially charged with fluoroindole 4
and DABCO in acetonitrile to form a slurry of the DABCO salt
of 4. Only then was the chloroimidate 18 solution charged. If
any acid was present, the DABCO would neutralize it before 4
was present. This reversal in charge order had no other effect
on the chemistry and has provided a simple remedy against
these impurities.
Along with the change in charging order, both fluoroindole 4
and DABCO were now charged in 10% excess. This would
hasten reaction completion with chloroimidate 18 and
minimize the presence of any 18 remaining in the mature
reaction mixture. The remaining fluoroindole 4 and DABCO
had high water solubility and were removed by the subsequent
water washes, whereas 18 was a problematic impurity. The
initial DABCO/fluoroindole mixture underwent a slurry to
slurry transition to one of indole−pyrrolotriazine 5 and
Figure 5. Self-condensation side products from chlorination.
as a side product.30,33 Up to 4 equiv of POCl3 was used in the
early phases of development. While those chlorinations
proceeded rapidly, our selection of reactor sets was limited
due to the relatively large volume of aqueous K2HPO4 required
during the quench to maintain a homogeneous phase cut. The
charge was optimized at 2.0 equiv which minimized impurities,
produced an acceptable rate and allowed a reasonable quench
volume.
Chlorination Workup and Reaction Modifications. The
subsequent workup for the chlorination required balancing the
hydrolysis of the remaining reagent against the propensity of
chloroimidate 18 to hydrolyze.34 Stress tests had established
that the hydrolysis rate of 18 would increase at pH ≤3 or ≥10.
The hydrolysis of 18 increased the titer of the side product
HCl, which would autocatalyze further hydrolysis. Cold
aqueous K2HPO4 proved to be an advantageous quench
solution, as in case of an accidental overcharge, it would not
result in an overly caustic quench mixture. Three equivalents of
12 wt % K2HPO4 solution hydrolyzed the remaining
phosphoryl chloride quickly to produce a pH 4 mixture, and
a follow-up base wash brought up the pH to 7.5−8.5. Using two
separate K2HPO4 treatments rather than one ameliorated the
otherwise inconveniently high maximum volume (Vmax). Using
a base with more capacity to neutralize acid, i.e. K3PO4, that
required less water to dissolve per base equivalent, was not
Scheme 8. Formation of acid-catalyzed abnormal coupling impurity
H
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 9. Formation of the 22 in the coupling process and conversion to 5
DABCO−HCl over 1−2 h (<0.5% 18). The slurry was diluted
with water and the product collected by filtration. Overall the
process has produced 3 MT of 5 over 19 batches in 93−97%
yields of >99.8% purity. Only three batches required further
aqueous acetone recrystallization as they were above the 0.10%
specification for individual impurities.
Final Workup Modifications To Achieve a Quality Gate
Intermediate. As our process entered phase III, the yields were
excellent, the hazards were controlled, and thus, the focus
remained on optimizing quality with control of impurities to
appropriately low limits. The chloroimidate 18 and its
precursor pyrrolotriazine 3 were the predominant impurities.
Although the chloroimidate was minimized by extending the
coupling reaction, 0.1−0.3% of 18 would routinely became
entrapped within the particles of the heterogeneous mixture
and was effectively isolated from further coupling. This could
be demonstrated by separate analysis of the supernatant and the
solids after 3 h; 90% of the remaining chloroimidate 18 resided
in the solids, and the levels did not decrease during further hold
times or any other modifications of the process.
nearly pure compound that would fulfill our goal of a quality
gate intermediate.
However, all recrystallization systems examined did a poor
job of rejecting 18. We previously had noted the sensitivity of
18 toward hydrolysis back into pyrrolotriazine 3 and
considered if this propensity could be exploited. While the
conversion of one impurity for another seems a poor exchange,
the following step was intolerant of 18, as it would form a
persistent daughter impurity. But the subsequent reaction could
tolerate up to 5% of the pyrrolotriazine 3, which formed an
impurity which was easily purged. Clearly, 3 was preferred.
Once the reaction mixture’s solvents were distilled to reduce
the isopropyl acetate component to <3% to avoid the
appearance of a second phase during the later water addition,
screening various levels of water indicated that only the range
of 2−6% water would produce a solution of the reaction
mixture at reflux. Presumably a level of <2% water did not
completely dissolve all of the salts while >6% water levels
similarly did not completely dissolve the organic components.
Within this range, the now solubilized chlorimidate impurity
completely hydrolyzed back to triazine 3 over 30−90 min at
reflux as it is exposed to water once the particles dissolve. The
90 min reflux limit was important as exceeding this period led
to detectable levels of cleavage of the PIV group. Upon cooling
and seeding at 70 °C, a seed bed was established and
subsequent cooling at 10 °C/h led to controlled recrystalliza-
tion. We were pleased to note that all 8 impurities that required
monitoring were now below the required levels. Recrystalliza-
tion from aqueous acetone is no longer required but if
incorporated, all nonsolvent impurities fall below detection
limits. In summary, the simple incorporation of a controlled
amount of water followed by a recrystallization under carefully
defined conditions completely hydrolyzed all remaining
chloroimidate 18 and rejected the remaining impurities with
no impact to the yield.
Another persistent impurity concern arose from a
fluoroindole precursor. Residual nitroketone 10 would produce
the corresponding 22 once it coupled with chloroimidate 18
(Scheme 9). The coupled nitroketone 22 would not purge from
5, but there was an alternative means to eliminate it. If sodium
dithionite and THF were charged into the mature coupling
mixture, all solids would dissolve at reflux.
As expected, 22 was reduced and converted immediately into
the desired product 5, a satisfactory means to eliminate an
impurity. While subsequent cooling led to recrystallization of
product containing no detectable levels of 22, we felt this
change was too radical at this stage and sought an alternative
resolution that did not require new reagents and solvents.
Instead, we reconsidered the isolation of the indole−
pyrrolotriazine 5 for rejecting these impurities. The current
workup (crystallization of 5 from the slurry of fluoroindole +
DABCO in CH3CN/iPrOAc followed by filtration/washes)
had worked reproducibly to date. Nevertheless, slurry to slurry
transformations can present challenges upon scale-up due to
mass transfer effects which may be dependent on equipment
and operating parameters. We noted the ability of a
recrystallization to remove the nitroketone impurity 22, and
began to explore means to introduce a direct classical
recrystallization of 5. This should negate the propensity of
the previously uncontrolled crystallization to trap chloroimi-
date, 18, and possibly also remove other impurities to deliver a
This new process has been run at up to 284 kg product scale
to produce 5 with reproducibly low impurity levels and
>99.95% purity. Three vendor campaigns have been completed
with yields averaging 93% without any complications while
producing over 1.6 MT of compound. On scale, all batches
were recrystallized as a precaution however the nonrecrystal-
lized material met all the specifications. This revised process
will be used for manufacture of 5 and establishes this
intermediate as a quality gatekeeper for the brivanib
commercial process. The further transformation of 5 into API
is described in ref 6.
I
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
min at 40 °C, then 5 °C/min to 70 °C, then 20 °C/min to 220
°C. Typical Retention Times: isopropyl acetate (RT 7.9
min). acetonitrile (RT 10.5 min).
CAUTION: nitroaromatics are potentially explosive; do not
run these processes without completing appropriate safety
studies and precautions.
CONCLUSIONS
■
A number of process challenges to the preparation of the key
intermediate to brivanib alaninate were overcome over a series
of campaigns. Each campaign established a basis for further
process development that ultimately led to 93% yield and
>99.95% purity of indole−pyrrolotriazine 5. These attributes
enabled the establishment of a gatekeeper intermediate for the
process to API. Key for these accomplishments was an
understanding of the stability of the fluoroindole 4 that enabled
the development of chemistry to prepare this regulatory
starting material, delineating the mechanism of the POCl3
chlorination chemistry that led to a reproducible chlorination
to chloroimidate 18, and finally the development of a
preparation/purification of the indole−pyrrolotriazine 5 that
targeted the efficient destruction of remaining chloroimidate
18.
Ethyl 2-(2,3-Difluoro-6-nitrophenyl)-3-oxobutanoate (7).
Ethyl acetoacetate (739 kg, 5679 mol) was added over 7 h 45
min at 10 °C to a solution of potassium tert-butoxide (637 kg,
5677 mol) in anhyd THF (1625 kg) cooled to 10 °C. The
transfer line was rinsed with THF (80 kg) and 1,2,3-trifluoro-4-
nitrobenzene (6, 447 kg, 2524 mol) was added over 2 h 10 min
at 10−20 °C and the transfer line rinsed again with THF (80
kg). The solution was heated to 22 °C over 20 min and stirred
at this temperature for 2.5 h. After reaction completion was
determined by HPLC (no 6 detected), the pH was adjusted to
7.8 with 5% aq HCl (2050 L) maintaining <30 °C. The bulk
was extracted with ethyl acetate (1370 kg) at 22 °C stirring for
20 min. The phases were allowed to settle for 5 h and
separated. The organic layer was concentrated until no
additional distillate collected under vacuum (P = 700−70
mbar) at 52−58 °C. This is used as is in the next step. An
analytical sample was purified by chromatography of the
concentrated oil on silica gel using heptanes/EtOAc 100/0 to
90/10 and isolating the purest fractions to produce a viscous
EXPERIMENTAL SECTION
■
General. In process methods for 7 and 8 (Steps 1 and 2 for
preparation of 4). Diluent: Acetonitrile. Mobile phase A: 90%
Water, 10% MeCN, 0.05% TFA. Mobile phase B: 90% MeCN,
10% water, 0.05% TFA. Column: Waters SymmetryShield RP8
150 mm × 4.6 mm, 3.5 μm. Column Temperature: 25 °C.
Detector Wavelength: 237 nm. Injection Volume: 10 μL.
Flow Rate: 1.0 mL/min. Program: 0 min 30% B, 30 min 90%
B. Typical Retention Times: 6 (RT 10.71 min, RRT 1.00 R). 7
(RT 17.27 min, RRT 1.61). 8 (RT 8.88 min, RRT 0.83).
In process methods for conversion of 8 to 9 (Step 3 for
preparation of 4). Diluent: Acetonitrile. Mobile phase A: 80%
Water, 20% methanol, 0.05% TFA. Mobile phase B: 80%
Acetonitrile, 20% methanol, 0.05% TFA. Column: Zorbax SB-
Aq, 150 mm × 4.6 mm, 3.5 μm. Column Temperature: 25 °C.
Detector Wavelength: 220 nm. Injection Volume: 10 μL.
Flow Rate: 1.2 mL/min. Program: 0 min 0% B, 25 min 25% B,
40 min 85% B. Typical Retention Times: 8 (RT 17.8 min,
RRT 0.83). 9 (RT 21.4 min, RRT 1.00).
In process method for conversion of 9 to 10 to 4 (Steps 4
and 5 for preparation of 4) and purity method for 4. Diluent:
Acetonitrile/methanol/TFA 800/200/0.5. Mobile phase A:
80% Water, 20% methanol, 0.05% TFA. Mobile phase B: 80%
Acetonitrile, 20% methanol, 0.05% TFA. Column: Zorbax SB-
Aq, 150 mm × 4.6 mm, 3.5 μm. Column Temperature: 25 °C.
Detector Wavelength: 237 nm. Injection Volume: 10 μL.
Flow Rate: 1.2 mL/min. Program: 0 min 0% B, 25 min 25% B,
40 min 85% B. Typical Retention Times: 9 (RT 20.9 min,
RRT 1.71). 10 (RT 14.1 min, RRT 1.15). 4 (RT 12.2 min,
RRT 1.00).
In process method for conversion of 3 to 18 to 5. Diluent:
Acetonitrile. Mobile phase A: 0.05% TFA in water. Mobile
phase B: 0.05% TFA in acetonitrile. Column: Synergi Hydro-
RP 150 mm × 4.6 mm, 4 μm. Column Temperature: 25 °C.
Detector Wavelength: 237 nm. Injection Volume: 10 μL.
Flow Rate: 2 mL/min. Program: 0 min 10% B, 2 min 10% B,
15 min 100% B, 18 min 100% B . Typical Retention Times: 3
(RT 8.1 min). 18 (RT 12.1 min). 5 (RT 12.9 min), 4 (RT 5.8
min).
In process method for isopropyl acetate determination.
Column: Stabilwax-DB, length 30 m, ID 0.32 mm, film 1.00
μm. Injector Temperature: 30 °C. Detector Temperature:
250 C. Injection Volume: 10 μL. Flow Rate: 1.5 mL/min
helium, split flow 50 mL/min. Detector: FID, hydrogen 40
mL/min, air 400 mL/min, make up 25 mL/min. Program: 2
1
yellow oil of 96% HPLC purity: HNMR (CD3CN, 399.78
MHz) 1.08 (t, 3H, J = 6 Hz), 1.86 (s, 3H), 4.00−4.25 (m, 2H),
7.45−7.55 (m, 1H), 7.92−7.98 (m, 1H), 13.13 (s, 1H).
13CNMR (CD3CN, 100.53 MHz) 14.49, 20.24, 62.60, 122.77,
146.98, 148.35, 150.81, 153.29, 155.85, 156.00, 171.81, 176.75.
1-(2,3-Difluoro-6-nitrophenyl)propan-2-one (8). The oil
from the previous step was diluted with glacial HOAc (880 kg)
over 5 min and warmed to 70 °C, and 95−97% H2SO4 (1535
kg) was added at <70 °C over 35 min. The mixture was heated
to 70 °C over 3 h 25 min and stirred at this temperature for 3 h.
After reaction completion was determined by HPLC (0.5% 7
remaining), it was cooled to −5 °C over 7 h 35 min, and water
(3793 kg) was added at <25 °C over 2 h 40 min. The mixture
was stirred at 22 °C for 40 min and allowed to settle for 15 min.
The layers were separated, and the upper aqueous phase was
extracted with ethyl acetate (1873 kg). The combined organic
phases were washed at 22 °C with water (235 kg), ethyl acetate
was added (806 kg), the organic phase was washed with 6%
NaHCO3 solution (3367 kg), stirred for 15 min, and settled for
30 min. After phase separation, it was washed with a 5% brine
(2860 kg) wash by stirring for 10 min and settling for 90 min.
The organic solution was distilled until no further distillate
collects under vacuum (P = 420−55 mbar/34−60 °C) over 9 h
10 min at a jacket temperature of ≤60 °C. The crude product
was isolated as a brown oil (654 kg, 77.6% HPLC purity). This
1
was used as is in the next step. HNMR (DMSO-d6, 399.78
MHz) 2.29 (s, 3H), 4.25 (d, 2H, J = 1.5 Hz), 7.68 (q of d, 1H, J
= 1.5, 9 Hz), 8.05 (q of d, 1H, J = 2, 5 Hz).
1-(2-Fluoro-3-methoxy-6-nitrophenyl)propan-2-one (9).
Potassium carbonate (286 kg, 2069 mol) was added in portions
at 9−25 °C to a solution of 8 (654 kg, 3040 mol) in methanol
(1706 kg). The reaction mixture was heated to reflux over 110
min and stirred for 17 h. After reaction completion
determination by HPLC (0.4% 9), ∼1/3 of the solvent (800
L) was distilled off at 68−70 °C over 7 h 45 min. The solution
was diluted with water, (644 kg) while allowing the internal
temperature to decrease to 39 °C. The solution was seeded to
induce crystallization. The suspension was stirred for 30 min,
J
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
then cooled to 3 °C, and stirred at this temperature for 90 min.
The product was collected by centrifuge over 11 h 45 min, and
the cake was washed with water (372 kg) in two portions. The
wet powder (394 kg) was dried under vacuum at 48−55 °C for
21 h 30 min to yield 375.0 kg (92.8% purity). This was
dissolved in MTBE (1859 kg) by heating to reflux over 4 h 35
min. Charcoal (15 kg) in MTBE (84 kg) was added and stirred
for 40 min at reflux. The slurry was filtered at 56 °C over 2 h 45
min and the line rinsed with hot MTBE (140 kg).
Approximately 2230 L solvent was removed by distillation at
56 °C over 2 h 40 min, cooled to 45 °C, seeded, and held for
30 min. It was finally cooled to 0 °C over 3 h and held for
additional 2 h. The product was collected by centrifuge over 12
h and the wet cake washed three times with cold tert-butyl
methyl ether (total 208 kg). The wet powder was dried under
vacuum at 48−55 °C for 9 h to yield 271 kg of crystals (99.9%
over 60 min, maintaining 20−25 °C. The line was rinsed with
methanol (37 kg). The product began to crystallize during the
addition of 10. The mixture was stirred at 20−25 °C for 4 h,
and reaction completion was determined at <0.5% (none
detected) by HPLC. The reaction mixture was cooled to 0 °C
over 195 min, stirred at this temperature for 5 h, and collected
by centrifuge over 14 h. The wet cake was rinsed with a
solution of 33% aq HCl (13 kg) and water (1070 kg) to yield
220 kg of wet 4, and then dried under vacuum at 44−45 °C for
16 h to yield 164 kg. This was dissolved at 22 °C in a solution
of methanol (411 kg), 33% aq HCl (6 kg) and water (31 kg) by
stirring for 150 min. Cellulose-base filtration aid (7 kg) was
added and the mixture stirred for 30 min. The filtration aid was
filtered off through a pad of additional filter aid over 7.5 h, and
the line was rinsed with degassed methanol (33 kg). A solution
of 33% aq HCl (8 kg) and water (813 kg) was added to the
filtrate over 45 min to precipitate the product. The slurry was
cooled to 0 °C over 70 min, stirred for 5 h, and collected by
centrifuge. The line and the cake were rinsed with a solution of
33% aq HCl (9 kg) and water (650 kg), and the wet powder
was dried under vacuum at 45 °C for 15 h to yield 144.3 kg
(69%, 99.8% assay, >99.95 HPLC area %, no detected GTIs,
total volatiles 0.1%, KF 0.3%). 1HNMR (DMSO-d6, 600 MHz)
2.34 (s, 3H), 6.06 (s, 1H), 6.66 (t, 1H, J = 8.6 Hz), 6.88 (d, 1H,
J = 8.6 Hz), 8.74 (s, 1H), 10.85 (s, 1H). 13CNMR (CDCl3,
125.8 MHz) 13.3, 94.2, 106.0, 111.9, 118.4, 132.3, 135.8, 136.3,
142.6.
1
assay; overall uncorrected yield from 6 = 47.3%). HNMR
(DMSO-d6, 399.78 MHz) 2.27(s, 3H), 3.96(s, 3H), 4.18(d,
2H, J = 4 Hz), 7.20−7.35(m, 1H), 7.95−8.05(m, 1H).
13CNMR (DMSO-d6, 100.53 MHz) 29.60, 56.69, 111.41,
120.17, 122.18, 140.72, 148.04, 150.48, 151.81, 202.95.
1-(2-Fluoro-3-hydroxy-6-nitrophenyl)propan-2-one (10).
Trimethylammonium chloride (466 kg, mol) was added in
portions at ≤25 °C to a previously prepared slurry of aluminum
trichloride (1301 kg, mol) in dry CH2Cl2 (1336 kg) over 150
min. To trap offgasses, the reaction vessel was connected to an
empty vessel, then a scrubber was filled with methanol/25% aq
ammonia/water (1:31.7:1.35 v/v/v) and finally a vessel filled
with dilute sulfuric acid to trap basic vapors. The mixture was
stirred for 1 h at 22 °C. A 32−38 °C solution of 9 (336 kg,
mol) in dry CH2Cl2 (445 kg) was added at 22−33 °C over 35
min, maintaining 3 bar pressure to account for CH3Cl offgas
pressure, followed by a line rinse with CH2Cl2 (87 kg).
(CAUTION: CH3Cl is a hazardous gas!). The mixture was
heated to 42 °C over 65 min and stirred at this temperature for
a total of 5 h 5 min, taking care to maintain P(vessel) at 3 bar
pressure throughout the stirring. HPLC analysis indicated no
starting material, and the mixture was degassed with nitrogen
through the scrubber, cooled to 22 °C over 40 min, and poured
into 2 °C water (3360 kg) at 2−22 °C over 7 h 20 min. The
line was rinsed with CH2Cl2 (87 kg), the mixture was cooled to
0 °C over 80 min and stirred at this temperature for 1 h, and
collected by centrifuge over 20 h 25 min. The wet cake was
washed with water (168 kg). Wet crude 10 was then added to
22 °C water (1680 kg), the slurry was stirred at 22 °C for 1 h
and collected by centrifuge over 10 h 50 min. The wet cake was
washed with water (672 kg), and the cake was checked for the
presence of residual chloride. Additional water washes were
added if necessary. The wet product was dried under vacuum at
50 °C for 13 h 20 min to yield 282 kg of solids (89.4% yield,
4-Chloro-5-methylpyrrolo[1,2-f ][1,2,4]triazin-6-yl pivalate
(18). A slurry of pyrrolotriazine 3 (89.0 kg, 357 mols) in dry
acetonitrile (156 kg, <0.05% water) was treated with
phosphorous oxychloride (66 L, 706 mols) over 8 min at
23−24 °C followed by an acetonitrile (45 kg) rinse.
Diisopropylethylamine (56 L, 321 mol) was added over 20
min at 22−29 °C followed by an acetonitrile (45 kg) rinse, and
the mixture was heated to reflux at 85 °C over 57 min. After
11.5 h, the reaction was sampled, indicating reaction
completion (0.18 HPLC area % 3). The solution was cooled
to 30−35 °C and stored in drums until a similar chlorination
was completed. The reaction mass was transferred into a
quench solution of 12.4 wt % aqueous K2HPO4 solution (2894
kg), isopropyl acetate (743 kg) and acetonitrile (35 kg) over 20
min at 2−10 °C followed by an acetonitrile rinse (39 kg). The
quenched solution was stirred for 1 h at 15−17 °C, allowed to
settle for 1 h and separated. The organic phase was washed with
15 wt % aqueous K2HPO4 solution (398 kg) for 1 h followed
by 1 h of settling before separation to produce a final pH of
8.90 (water layer). The organic layer yielded 985 kg of solution
of 99.5 HPLC A% 18. In addition, 79 kg of acetonitrile rinse
containing more 18 was also collected for the next step. A
second identical chlorination and quench was undertaken that
eventually yielded 960 kg of solution and was combined with
the first solution for the subsequent coupling step. A purified
sample was prepared by concentration of the product solution
to solids, dissolution into 1:1 EtOAc/heptanes, filtration
through a plug of silica gel, concentration, and dissolution
into hot heptanes. Recrystallization was initiated by cooling and
seeding. Following cooling to 5 °C, filtration and pumping, a
light-yellow powder was obtained: DSC mp = 82.4 °C; Anal.
Calcd For C12H14N3O2Cl: C, 53.83; H, 5.27; N, 15.69; Cl,
1
99.4% HPLC assay). HNMR (DMSO-d6, 399.78 MHz) 2.26
(s, 3H), 4.16 (s, 2H), 7.04 (dd, 1H, J = 12,8 Hz), 7.90 (d, 1H, J
= 8 Hz), 11.35−11.50 (m, 1H). 13CNMR (DMSO-d6, 100.53
MHz) 29.72, 115.34, 120.93, 122.49, 139.51, 150.18, 150.89,
151.01, 203.15.
4-Fluoro-2-methyl-1H-indol-5-ol (4). Compound 10 (270
kg, mol) was dissolved in methanol (535 kg) at 27 °C and
cooled to 22 °C. A freshly prepared solution of sodium
hydrosulfite (1216 kg, mol, 83% assay) in water (4050 kg) was
prepared in a separate vessel and at 24 °C (CAUTION: sodium
dithionite has reports of exothermic decomposition when
moist; an aqueous solution was stable up to 170 °C). The
methanolic solution of 10 was added to the aqueous dithionite
1
13.24. Found: C, 53.91; H, 5.20; N, 15.75; Cl, 13.26. HNMR
(DMSO-d6, 399.78 MHz) 1.35 (s, 9H), 2.38 (s, 3H), 8.28 (s,
1H), 8.36 (s, 1H). 13CNMR (DMSO-d6, 100.53 MHz) 9.22,
27.34, 39.41, 107.04, 113.49, 116.61, 139.98, 175.75.
K
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
4-(4-Fluoro-2-methyl-1H-indol-5-yloxy)-5-methylpyrrolo-
[1,2-f][1,2,4]triazin-6-yl pivalate (5). For the purposes of
stoichiometry, a quantitative yield is assumed for the previous
step. The indole 4 (130.0 kg, 787 mol) was dissolved into
acetonitrile (222 kg), stirred 4 h to dissolve, and passed
through a 1 μm filter over 1 h. Additional acetonitrile (30 kg)
was used to wash over the remaining amounts. 1,4-Diaza-
bicyclo[2.2.2]octane (88.0 kg, 785 mols) was charged to the
solution via a rotary feeder over 35 min at 20−29 °C. After 70
min of stirring at 26−27 °C, the solution of 18 prepared above
was charged over 50 min through an inline filter at 26−28 °C
and washed in with the acetonitrile wash that was collected in
the subsequent step. The mixture was stirred for 3 h, at which
point HPLC indicated reaction completion (0.31 HPLC area %
18). The mixture was concentrated by distillation at 200 mbar
at 37−39 °C for 3.5 h until 1290 kg remained in the reactor
(reactor is on a weight cell), acetonitrile (1216 kg) was charged
and an equal weight distilled off, and finally acetonitrile (1222
kg) was charged and 1220 kg distilled off. The isopropyl acetate
content by GC was measured as 2.4 vol % (call point ≤3%) and
KFT as 0.13%. More acetonitrile (278 kg) was added to bring
the volume to the desired concentration. Water (30 kg) was
added and the KFT measured as 2.0% (desired range 2−3%).
The slurry was heated to reflux (80−81 °C) over 43 min to
dissolve the solids and was held 1 h. The solution was cooled to
67 °C over 15 min, forming a thin suspension, and seed crystals
(1.4 kg) in acetonitrile suspension were added. The seed bed
was established over 75 min, and then the slurry was cooled to
24 °C over 3 h 10 min. Water (1780 kg) was added over 150
min at 23−24 °C and held 1 h. The solids were collected by
filtration on a pressure filter over 165 min. A displacement wash
of 297 kg water was done followed by four washes of 667 kg
water each. The cake was dried under vacuum for 3 h to
produce 324 kg of wet (LOD 15.4%) crystals. This was
recrystallized from acetone (1464 kg) by heating over 40 min
to 55 °C, holding 35 min, and cooling to 38 °C over 12 min,
holding 25 min, at which point 1.5 kg of seed crystals were
added. The slurry was stirred 65 min and cooled to 23−25 °C
over 160 min. It was held 152 min, and water (1145 kg) was
added over 155 min at 20−24 °C. The slurry was held 130 min,
and the crystals were collected by centrifuge over 25 h and
washed with water (1176 kg) to produce 298 kg wet crystals.
This was dried at 18−63 °C over 7 h 30 min to produce 264 kg
white powder, (93% yield, 0.05% KFT) after sieving through a
1.2 mm mesh (≥99.95% HPLC A%, assay ≥99.9%, 0.11%
and washed with water (94 L). The solids were dried at 60 °C
over 35 h (KF = 0.04%) to produce 45.0 kg of cream-colored
solids (96.2% recovery, 99.89 HPLC area %, 99.8 wt %).
AUTHOR INFORMATION
Corresponding Author
*E-mail: pesti-office@oprd.acs.org
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We gratefully acknowledge the contributions of Dave Leahy,
Jeff Reinhardt, Michael Wong, Fernando Quiroz, Evan Barlow,
Thomas Kissick, Robert Discordia, Sushil Srivastava, Yueer Shi,
K. Dambalas, Junying Fan, Diya Abdeljabbar, Jollie Godfrey,
Bruce Lotz, Frank Okuniewicz, Andy Stewart, Mary-Ann
Fanning, the personnel and managers of our scale up
laboratories and pilot plants, the collaboration of numerous
researchers of our analytical groups, our vendors, and Simon
Leung and his group for safety related studies. We thank Rod
Parsons, Rob Waltermire, and David Kronenthal for reviewing
this manuscript.
REFERENCES
■
(1) (a) Bhide, R. S.; Cai, Z.-W.; Zhang, Y.-Z.; Qian, L.; Wei, D.;
Barbosa, S.; Lombardo, L. J.; Borzilleri, R. M.; Zheng, X.; Wu, L. I.;
Barrish, J. C.; Kim, S.-H.; Leavitt, K.; Mathur, A.; Leith, L.; Chao, S.;
Qautlet, B.; Mortillo, S.; Jeyaseelan, R., Sr.; Kukral, D.; Hunt, J. T.;
Kamath, A.; Fura, A.; Vyas, V.; Marathe, P.; D’Arienzo, C.; Derbin, G.;
Fargnoli, J. J. Med. Chem. 2006, 49, 2143−2146. (b) Bhide, R. S.; Fan,
J.; Parlanti, L.; Barbosa, S.; Qian, L.; Cai, Z.; Gibson, F. S. PCT Int.
Appl. 2004, WO/2004/009542; CAN 140:146018. (c) Hunt, J. T.;
Mitt, T.; Borzilleri, R.; Gullo-Brown, J.; Fargnoli, J.; Fink, B.; Han, W.
C.; Mortillo, S.; Vite, G.; Wautlet, B.; Wong, T.; Yu, C.; Zheng, P.;
Bhide, R. S. J. Med. Chem. 2004, 47, 4054−4059. (d) Borzilleri, R. M.;
Bhide, R. S.; et al. Bioorg. Med. Chem. Lett. 2005, 1429−1433. (e) Cai,
Z. W.; Zhang, Y.; Qian, L.; Barbosa, S.; Wei, D.; Borzilleri, R. M.;
Zheng, X.; Wu, L.; Fan, J.; Shi, Z.; Wautlet, B. S.; Mortillo, S.;
Jeyaseelan, R.; Kukral, D. W.; Kamath, A.; Marathe, P.; D’Arienzo, C.;
Derbin, G.; Barrish, J. C.; Hunt, J. T.; Lombardo, L. J.; Fargnoli, J.;
Bhide, R. S. J. Med. Chem. 2008, 51, 1976−1980. (f) Borzilleri, R. M.;
Zheng, X.; Qian, L.; Ellis, C.; Cai, Z.-w.; Wautlet, B. S.; Mortillo, S.;
Jeyaseelan, R.; Kukral, D. W.; Fura, A.; Kamath, A.; Vyas, V.; Tokarski,
J. S; Barrish, J. C.; Hunt, J. T.; Lombardo, L. J.; Fargnoli, J.; Bhide, R.
S. J. Med. Chem. 2005, 48, 3991−4008. (g) Padilla, I.; Siu, L. L. Expert
Opin. Investig. Drugs 2011, 20, 577−586. (h) Ayers, M.; Fargnoli, J.;
Lewin, A.; Wu, Q.; Platero, J. S. Cancer Res. 2007, 67, 6899−6906.
(i) Bhide, R.; Cai, Z. W.; Qian, L.; Barbosa, S. PCT WO/2004/
009784, 2004; CAN 140:124558. (j) Qian, C.; Cai, X.; Zhai, H. U.S.
Patent Appl. 2009/0076044, March 19, 2009. (k) Bhide, R. S.; Cai, Z.
W.; Qian, L.; Barbosa, S.; Lombardo, L.; Robl, J. U.S. Patent 7,820,814
B2, October 26, 2010. (l) Hennequin, L. F. A. Patent WO/2002/
12226 A1, Feb 14, 2002. (m) Zheng, B.; Conlon, D. A.; Corbett, R.
1
volatiles. No detectable GTIs.) HNMR (DMSO-d6, 399.78
MHz) 1.40 (s, 9H), 2.43 (s, 3H), 2.48 (s, 3H), 6.32 (s, 1H),
6.96 (dd, 1H, J = 8.3, 8.6 Hz), 7.06 (d, 1H, J = 8.8 Hz), 7.84 (s,
1H), 7.90 (s, 1H), 8.07 (br s, 1H). 13CNMR (DMSO-d6,
100.53 MHz) 8.88, 13.78, 27.35, 39.46, 97.26, 105.65, 106.18
(d, J = 4.0 Hz), 110.16, 111.33, 116.07, 119.03 (d, J = 19.1 Hz),
130.73 (d, J = 11.1 Hz), 136.45 (d, J = 11.1 Hz), 136.64,
138.41, 145.76, 146.46 (d, J = 249.3 Hz), 162.46, 175.91.
Optional Recrystallization for 4-(4-Fluoro-2-methyl-1H-
indol-5-yloxy)-5-methylpyrrolo[1,2-f ][1,2,4]triazin-6-yl piva-
late (5). Crude 5 (46.8 kg, 99.5 HPLC area %) was dissolved
into acetone (267 L) and heated to 54 °C over 40 min and held
for 30 min. This was clarified by passage through a 5 μm in-line
filter to a second reactor over 20 min. A further 14 L of acetone
was used to rinse the reactor. The mixture was heated to 54 °C
to redissolve solids and cooled to 19 °C. Water (187 L) was
added over 66 min, and the slurry was cooled to 5 °C over 56
min. After a 16.5 h hold, the solids were collected by filtration
̆
M.; Chau, M.; Hsieh, D.-M.; Yeboah, A.; Hsieh, D.; Muslehiddinoglu,
̈
J.; Gallagher, W. P.; Simon, J. N.; Burt, J. Org. Process Res. Dev. 2012,
16, 1846−1853.
(2) (a) Saharinen, P.; Eklund, L.; Pulkki, K.; Bono, P.; Alitalo, K.
Trends Mol. Med. 2011, 17, 347−362. (b) Ivy, S. P.; Wick, J. Y.;
Kaufman, B. M. Nat. Rev., Clin. Oncol. 2009, 6, 569−579. (c) Kiselyov,
A.; Balakin, K. V.; Tkachenko, S. E. Expert Opin. Investig. Drugs 2007,
16, 83−107.
(3) Srivastava, S. Quality by Design in Drug Substance Development:
Opportunity or Challenge for Commercialization? Sino Food and
Drug Association Meeting, Beijing, China, June 17, 2010.
(4) Shi, Z.; Kiau, S.; Lobben, P.; Hynes, J.; Wu, H.; Parlanti, L.;
Discordia, R. P.; Doubleday, W.; Leftheris, K.; Dyckman, A. J.;
L
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Wrobleski, S. T.; Dambalas, K.; Tummala, S.; Leung, S. S. W.; Lo, E.
T. Org. Process Res. Dev. 2012, 10, 1618−1625.
Vol. 3, p 77; New York: John Wiley & Sons Inc., 1952. (b) Bird, C.
W.; Cheeseman, G. W. H. Structure of Five-membered Rings with
One Heteroatom. In Comprehensive Heterocyclic Chemistry; Katritzky,
A. R., Rees, C. W., Eds.; Pergamon Press: Oxford, 1984; Vol. 4, pp
246−254.
(21) Hamann, H. J.; Liebscher, J. J. Org. Chem. 2000, 65, 1873−1876.
(22) Convergent reactions typically produce impurities that differ
significantly from the desired product and should be amenable to more
efficient purification. Zhang, T. Y. Chem. Rev. 2006, 106, 2583−2595.
(23) Nahm, S.; Weinreb, S. M. Tetrahedron Lett. 1981, 22, 3815−
3818.
(5) LaPorte, T. L.; Spangler, L.; Hamedi, M.; Lobben, P.; Chan,S. H.;
Muslehiddinoglu, J.; and Wang, S. Y. Org. Process Res. Dev., to be
submitted.
(6) A paper to be submitted to Org. Process Res. Dev. describes the
process development from 5 to the active pharmaceutical ingredient
(API) or brivanib alaninate.
(7) Crispino, G. A.; Hamedi, M.; LaPorte, T. L.; Thornton, J. E.;
̆
Pesti, J. A.; Xu, Z.; Lobben, P. C.; Leahy, D. K.; Muslehiddinoglu, J.;
̈
Lai, C.; Spangler, L. A.: Discordia, R. P.; Gibson, F. S. Process for the
Preparation of [(1R),2S]-2-Aminopropionic acid 2-[4-(4-fluoro-2-
methyl-1H-indol-5-yloxy)-5-methylpyrrolo[2,1-f ][1,2,4]triazin-6-
yloxy]-1-methylethyl ester. U.S. Patent 7,671,199, Mar 2, 2010.
(8) (a) Reissert, A. Ber. 1897, 30, 1030. (b) Baeyer, J. Ber. 1880, 13,
187. (c) Sundberg, R. J. The Chemistry of Indoles; Academic Press: New
York, 1970; pp 176−183.
(24) Pesti, J., Ed. Org. Process Res. Dev. 2012, 16, 840 and manuscripts
following in the same issue: pp 844−1150.
(25) (a) Tetrahedron Lett. 1987 28, 1027−1030. (b) Tetrahedron
Lett. 1993 34, 7667−7668.
(26) A large variety of other chlorinating agents will also form 18, at
least in low yield, such as thionyl chloride, oxalyl chloride,
phosphorous trichloride, and phosphorus pentachloride. Similarly,
phosphorous oxybromide permitted impure preparations of the
analogous bromide. Methanesulfonyl chloride and DIPEA in dichloro-
methane formed the N-mesylate, unreactive when further treated with
4 and DABCO.
(27) (a) Ford, J. G.; Pointon, S. M.; Powell, L.; Siedlecki, P. S. Org.
Process Res. Dev. 2010, 14, 1078−1087. (b) Ford, J. G.; O’Kearney-
McMullan; Pointon, S. M.; Powell, L.; Siedlecki, P. S.; Purdie, M.;
Withnall, J.; Wood, P. O. F. Org. Process Res. Dev. 2010, 14, 1088−
1093. (c) Broxer, S.; Fitzgerald, M. A.; Sfouggatakis, C.; Defreese, J. L.;
Barlow, E.; Power, G. L.; Peddicord, M.; Su, B-N; Tai-Yuen, Y.;
Pathirana, C.; Sherbine, J. P. Org. Process Res. Dev. 2011, 15, 343−352.
(d) Sun, Z.; Wang, H.; Wen, K.; Li, Y.; Fan, E. J. Org. Chem. 2011, 76,
4149−4153. (e) See refs 1m and 33 (f) Thiel, O. R.; Achmatowicz,
M.; Bernard, C.; Wheeler, P.; Savarin, C.; Correll, T. L.; Kasparian, A.;
Allgeier, A.; Bartberger, M. D.; Tan, H.; Larsen, R. D. Org. Process Res.
Dev. 2009, 13, 230−241.
(28) Crispino, G.; Barbosa, S.; Fan, J.; Cai, Z. W. U.S. Patent Appl.
2005/0288289 A1, December 29, 2005. Solvents that permitted
chlorination with POCl3 included toluene, acetonitrile, dimethyl
carbonate, MTBE, ethyl acetate, isopropyl acetate, n-butyl acetate as
well as neat POCl3 itself.
(29) Other productive bases were quinoline, quinoxaline, triethyl-
amine, and most secondary and tertiary amines that were tested.
(30) For an excellent investigation of the chlorination kinetics and
mechanism for 4-quinazolones and the appearance of a similar dimer
see: Arnott, E. A.; Chan, L. C.; Cox, B. G.; Meyrick, B.; Phillips, A.. J.
Org. Chem. 2011, 1653−1661.
(9) Alternative bases to t-BuOK and alternative solvents have been
demonstrated: (a) Parikh, V. D.; Fray, A. H.; Kleinman, E. F. Synthesis
1988, 1567−1569. (b) Arnott, E. A.; Crosby, J.; Evans, M. C.; Ford, J.
G.; Jones, M. F.; Leslie, K. W.; McFarlane, I. M.; Sependa, G. J. Patent
WO/2008/053221 A2; May 8, 2008. The benzyl position acidity of 7
competes with deprotonation of the ethyl acetoacetate by t-BuOK.
Two equivalents of ethyl acetoacetate anion were necessary since one
would be quenched by the product. The side product formed by
nucleophilic displacement of the fluoride by tert-butoxide on the
starting material was minimized by maintaining ≤25 °C during the
addition of ethyl acetoacetate.
(10) All purities are HPLC area % unless otherwise specified.
(11) DSC/calorimetry detected the following exothermic events: 6
kJ/kg at 54 °C; 85 kJ/kg at 124 °C, and 74 kJ/kg at 221 °C. The
various stages of the workup revealed no concerns below 171 °C. Note
that these measurements were specific to our starting materials,
facilities and protocols. Repeating any of this chemistry requires
independent safety examination for prudent chemical practices.
(12) Following H2SO4 addition, exothermic events are detected at 8
kJ/kg at 66 °C, 256 kJ/kg at115 °C, 282 kJ/kg at 295 °C, and after 3 h
of reaction: 279 kJ/kg at 103 °C and 305 kJ/kg at 295 °C, indicating a
reaction temperature of 70 °C permitted a safe operation. The
concentrated oil was stable and displayed a DSC exotherm of 1193 kJ/
kg at 210 °C.
(13) Arnott, E. A.; Crosby, J.; Evans, M.C..; Ford, J. G.; Jones, M. F.;
Leslie, K. W.; Mcfarlane, I. M.; Sependa, G. J. PCT Int. Appl. (2008),
WO/2008/053221 A2 20080508. CAN 148:517536.
(14) Safety examination of the reaction solution established
exotherms of 12 kJ/kg at 108 °C and 250 kJ/kg at 180 °C and no
other points of concern for the rest of the reaction, workup,
recrystallization and drying.
(15) (a) Boovanahalli, S. K.; Kim, D. W.; Chi, D. Y. J. Org. Chem.
2004, 69, 3340−3344. (b) Kemperman, G. J.; Roeters, T. A.;
Hilberink, P. W. Eur. J. Org. Chem. 2003, 9, 1681−1686. (c) Parvulescu,
V. I.; Hardacre, C. Chem. Rev. 2007, 107, 2615−2665.
(16) DSC and safety testing identified exotherms of 208 kJ/kg at 90
°C and 77 kJ/kg at 275 °C, indicating little danger if CH2Cl2 was the
solvent and no other points of concern for the rest of the reaction,
workup, recrystallization and drying. While AlCl3 alone will cleave
aromatic methyl ethers, it requires forcing conditions: Gutman, A.;
Nisnevich, G.; Yudovitch, L. U.S. Patent 7,312,345 B2, December 25,
2007.
(31) Archmatowicz, M. M.; Thiel, O. R.; Colyer, J. T.; Hu, J.; Elipa,
M. V. S.; Tomaskevitch, J.; Tedrwo, J. S.; Larsen, R. D. Org. Process Res.
Dev. 2010, 14, 1490−1500.
(32) The ring nitrogens of the pyrrolotriazine are not sufficiently
basic to complex HCl.
1
(33) The structure of this impurity was established by HNMR and
single crystal X-ray spectroscopy. It became the primary product by
treating triazine 3 with 0.3 equiv of POCl3 and 1.2 equiv of DIPEA.
Similar dimers have been detected by others when reaction conditions
are not optimal: Anderson, N. G.; Ary, T. D.; Berg, J. L.; Bernot, P. J.;
Chan, Y. Y.; Chen, C.-K; Davies, M. L.; DiMarco, J. D.; Dennis, R. D.;
Deshpande, R. P.; Do, H. D.; Droghini, R.; Early, W. A.; Gougoutas, J.
Z.; Grosso, J. A.; Harris, J. C.; Haas, O. W.; Jass, P. A.; Kim, D. H.;
Kodersha, G. A.; Kotnis, A. S.; LaJeunesse, J.; Lust, D. A.; Madding, G.
D.; Modi, S. P.; Moniot, J. L.; Nguyen, A.; Palaniswamy, V.; Phillipson,
D. W.; Simpson, J. H.; Thoraval, D.; Thurston, D. A.; Tse, K.;
Polomski, R. E.; Wedding, D. L.; Winter, W. J. Org. Process Res. Dev.
1997, 1, 300−310 See also ref 2f..
(17) (a) Humphrey, G. R.; Kuethe, J. T. Chem. Rev. 2006, 106,
2875−2911. (b) Hegedus, L. S. Angew. Chem. 1988, 100, 1147−61.
(c) Pindur, U.; Adam, R. J. Heterocycl. Chem. 1988, 25, 1−8.
(18) Safety testing indicated exothermic events of 81 kJ/kg at 164 °C
and 16 kJ/kg at 264 °C and no significant events for all other stages of
the workup, recrystallization and drying.
(34) Archmatowicz et al. has published an excellent analysis and
understanding of the quench of POCl3 in a similar reaction but states
that each reaction is unique and recommends full hazard analysis
before scale-up; see ref 31.
(19) MS would detect components up to MW = 1035.
(20) 2-Methylindoles are known to undergo ring oxidations for many
oxidants: (a) Julian, P. L.; Mueyer, E. W.; Printy, H. C. In The
Chemistry of Indoles; Elderfield, R. C., Ed.; Heterocyclic Compounds,
M
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(35) This structure based on LC−MS data. The analogous impurity
is also known to form when a quinazolinone was chlorinated by
POCl3. See refs 27 (a) and (b).
(36) Handbook of Reactive Chemical Hazards, 3rd ed.; Breherick, L.
Ed.; Butterworths: London, 1975, pp1132−1134.
(37) Buncel, E.; Menon, B. J. Am. Chem. Soc. 1977, 99, 4457−4461.
(38) The assignment for regioisomer (c) is tentative.
N
dx.doi.org/10.1021/op400242j | Org. Process Res. Dev. XXXX, XXX, XXX−XXX