Development of an efficient Pd-catalyzed coupling process for axitinib

Article abstract of DOI:10.1021/op400088k

The manufacturing process of axitinib (1) involves two Pd-catalyzed coupling reactions, a Migita coupling and a Heck reaction. Optimization of both of these pivotal bond-formation steps is discussed as well as the approach to control impurities in axitinib. Essential to the control strategy was the optimization of the Heck reaction to minimize formation of impurities, in addition to the development of an efficient isolation of crude axitinib to purge impurities.

Full text of DOI:10.1021/op400088k

Article
pubs.acs.org/OPRD
Development of an Efficient Pd-Catalyzed Coupling Process for
Axitinib
Brian P. Chekal,‡ Steven M. Guinness,‡ Brett M. Lillie,‡ Robert W. McLaughlin,‡ Charles W. Palmer,†
Ronald J. Post,‡ Janice E. Sieser,‡ Robert A. Singer,*,‡ Gregory W. Sluggett,† Rajappa Vaidyanathan,‡
and Gregory J. Withbroe‡
‡Chemical Research and Development, Pfizer Worldwide Research and Development, Groton Laboratories, Eastern Point Road,
Groton, Connecticut 06340, United States
†Analytical Research and Development, Pfizer Worldwide Research and Development, Groton Laboratories, Eastern Point Road,
Groton, Connecticut 06340, United States
ABSTRACT: The manufacturing process of axitinib (1) involves two Pd-catalyzed coupling reactions, a Migita coupling and a
Heck reaction. Optimization of both of these pivotal bond-formation steps is discussed as well as the approach to control
impurities in axitinib. Essential to the control strategy was the optimization of the Heck reaction to minimize formation of
impurities, in addition to the development of an efficient isolation of crude axitinib to purge impurities.
an (THP) to provide intermediate 5. The first C−C bond-
forming step involves a Heck reaction between 5 and 2-
vinylpyridine (6) to afford 7. To set the stage for the final bond
formation, the nitro group is reduced, and the amine undergoes
a Sandmeyer reaction7 to provide 6-iodoindazole 9. After
coupling 9 with thiol 10 using Pd(dppf)Cl2 as catalyst,5 THP-
protected axitinib (11) is obtained. Deprotection of 11 with p-
TsOH in methanol affords crude axitinib (1a) as an ethyl
acetate solvated form. Crude axitinib is converted to the desired
solid form through dissolution in methanol and acetic acid
followed by treatment with activated carbon or Florisil and
displacement of solvent with xylenes to crystallize axitinib’s
anhydrous form 4 (1b).
One disadvantage of this route is that the potentially
genotoxic intermediates 7 and 8 may require stringent control
in the drug substance. In addition, the Sandmeyer reaction is
relatively low yielding and not an operation that is desirable to
conduct toward the end of a manufacturing route due to the
energetic nature and potential toxicity of the intermediates
involved.
INTRODUCTION
■
Vascular endothelial growth factor (VEGF) antagonists have
been recognized as an important class of pharmaceutical agents
for development due to their efficiency in controlling growth
and proliferation of cancer cells.1 Axitinib (1) is a VEGF
inhibitor in Pfizer’s oncology development program currently
undergoing late-stage clinical trials for the treatment of solid
tumors, including metastatic renal cell carcinoma (RCC).2,3
During the review of axitinib by the Food and Drug
Administration (FDA), the oncology drug advisory committee
unanimously endorsed approval of this drug and cited the
different toxicity profile as beneficial in providing more options
to patients. On January 27, 2012, axitinib was approved with
the trade name INLYTA for treatment of patients in the United
States with advanced renal cell carcinoma after failure of one
prior systemic therapy.
When considering a manufacturing process for axitinib
(Figure 1), the vinylpyridine moiety could be installed through
A number of important elements were learned regarding the
Heck reaction for this key bond formation which controls the
trans-olefin geometry. The THP protecting group on 5 blocks
the indazole nitrogen from forming a Michael adduct with 6.
The Heck reaction involving 5 is facilitated by the electron-
deficient nitro group present, which otherwise is challenged at
undergoing insertion due to the electron-rich character at the 3-
position of indazole. The main disadvantage of carrying out the
Heck reaction early in the process is that the olefin moiety is
potentially prone to degradation. This being the case, it is a
more prudent strategy to carry out the Heck reaction toward
Figure 1. Structure of axitinib (1) and bond-forming strategy.
a Heck reaction,4 and the diaryl thioether could be constructed
through a Migita5 cross-coupling or SNAr strategy. To enable
manufacture of axitinib for early clinical trials, a scalable route
was identified (Scheme 1) using these approaches.6 The
sequence begins with iodination of 6-nitroindazole (2) and a
subsequent protection of 3-iodoindazole 3 with tetrahydropyr-
Special Issue: Transition Metal-Mediated Carbon-Heteroatom Cou-
pling Reactions
Received: April 5, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 1. First-generation synthesis of axitinib
Scheme 2. Second-generation synthesis of axitinib
the end of the route rather than rely on processing and
engineering controls to mitigate potential degradation products
through a lengthy synthetic route.
prompted a revision of the route. A second-generation route
was developed which reversed the order of the key bond-
formation steps to allow for a shorter process (Scheme 2)8
while also addressing the Pd control strategy.9
The final recrystallization and polymorph control possess
undesirable unit operations involving displacement of acetic
acid and methanol with xylenes as well as treatment with
activated charcoal or another processing agent (such as Florisil)
for Pd removal. For this reason an essential goal of
development was focused on designing a milder approach for
accessing the desired solid form, while controlling levels of Pd
without processing agents.
The first-generation synthesis for axitinib (Scheme 1) was
sufficient for preparing early clinical supplies and had identified
key bond-formation steps. The use of the Sandmeyer reaction
(conversion of 8 to 9) late in the process was highly
undesirable from a manufacturing perspective, and this
In this approach, 6-iodoindazole (12) is prepared by
commercial suppliers from 6-aminoindazole through a
Sandmeyer reaction.7 The Migita coupling of 12 with 10 is
accomplished using Xantphos as a supporting ligand for Pd.10
After formation of 13 is complete, 14 is formed in situ with
addition of iodine and base. A Heck reaction between 14 and 6
at 110 °C is carried out using 1,8-bis(dimethylamino)-
naphthalene (Proton-Sponge) as an additive to suppress the
indazole nitrogen undergoing a Michael addition with 6. After a
workup, crude axitinib (1a) is further purified to control Pd and
organic impurities. Crude axitinib is first recrystallized from
NMP and methanol in the presence of 1,2-diaminopropane
B
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(1,2-DAP) and 1,2-bis(diphenylphosphino)ethane (DIPHOS)
for Pd removal.9 Following the recrystallization, reslurries from
THF and methanol control organic impurities. As in the prior
route, a final recrystallization from acetic acid, methanol, and
xylenes is necessary to obtain form 4 of axitinib (1b).
Development of this second-generation process was undertaken
to ensure control of process impurities and to consistently
deliver high-quality drug substance for commercial manufactur-
ing.
found to form 15 readily in solution if oxygen is not excluded
by purging with inert gas. In the absence of 12, mercapto dimer
15 is likely to undergo oxidative addition with Pd to afford a
stable complex 16 as proposed below in Scheme 3.12
It was found that this problem of inhibition by 15 is
overcome by addition of zinc metal13 or a related reducing
agent.14 Under ideal manufacturing conditions with efficient
inertion of the headspace, the use of zinc has been unnecessary,
since it is predominantly dissolved oxygen that promotes
formation of 15, potentially leading to catalyst inhibition.
Additional optimization of the coupling reaction includes the
use of a milder base, sodium bicarbonate, which allows for
lower reaction temperatures (50 °C) as shown in Scheme 4.
With use of a weaker base, the amount of thiolate
(deprotonated 10) is minimized. To further minimize potential
inhibition of the Pd catalyst, 10 is slowly added to the reaction
mixture after all other components have been charged.
While steps 1−2 are highly efficient, it was recognized that
some batches of 6-iodoindazole (12) appeared to inhibit the Pd
catalyst for the Migita coupling. The problem was traced to low
levels of sulfur being present in 12 as a result of the means of its
preparation. The root cause of this was found to be the quench
of the Sandmeyer reaction used to convert 6-aminoindazole to
12, and that by avoiding sodium thiosulfate during the quench,
sulfur is no longer formed in the process.15
The conditions for iodination at step 2 remained unchanged
from those originally developed. During development, it was
observed that excess base and higher reaction temperatures led
to the formation of sulfoxide impurity 18 (Figure 2). In
addition, premixing of the iodine and base gave poor
conversion of the iodination reaction. To achieve complete
conversion while minimizing sulfoxide formation, 2 equiv of
iodine is charged followed by 2 equiv of aqueous potassium
hydroxide. The reaction mixture is transferred into aqueous,
methanolic ascorbic acid to quench the excess iodine during
crystallization of crude 14. In carrying out the telescoped
process, crude 14 is isolated as a yellow-to-brown colored
crystalline material in about 80% yield.
The impurities tracked across steps 1−2 are shown in Figure
2. Compound 17 arises from unreacted 12 in step 1 undergoing
iodination in step 2. The step 1 product, 13, is a common
impurity found in 14 due to incomplete conversion in step 2.
As mentioned earlier, 18 arises from competitive oxidation of
14 under the reaction conditions of step 2. The Xantphos used
in step 1 gives rise to a family of impurities. One set of
impurities is derived from oxidation, which provides mono- and
bis-oxides of Xantphos. Another set of impurities results from
exchange of the aryl groups of Xantphos with indazole during
the coupling reaction, as proposed in Scheme 5.16 The initial
product of the exchange reaction is a derivative of Xantphos
with an indazole incorporated into the phosphine ligand
(compound 23) as well as a byproduct resulting from coupling
RESULTS AND DISCUSSION
■
Overall, the second-generation route to axitinib (Scheme 2) is
extremely efficient with a well-conceived sequence of steps;
therefore, this route was deemed appropriate for commercial
development. The first two telescoped steps required minimal
optimization and provided 14 as a reliable intermediate for
early control of impurities in the route. The Heck reaction was
recognized as a step that required substantial development due
to the need to control various process impurities resulting from
the relatively high reaction temperatures used. To limit
formation of a homocoupled dimer of 14, a large amount of
LiBr was originally included in this step,11 but it would be ideal
to avoid or limit the use of this additive. The subsequent
reslurries of 1a, while effective in purity control, could be
unreliable due to the heterogeneous nature of the process.
Therefore, better control of the formation and purge of
impurities within the Heck reaction became a primary goal, in
addition to control of residual Pd and solid form to enable
robust production of axitinib.
Development of the Migita Coupling and Iodination.
The initial work on the commercial manufacturing route for
step 1 demonstrated that a Pd catalyst derived from Xantphos
provided the optimal reactivity in the Migita coupling. A
primary concern focused on thiol 10 acting as a Pd catalyst
inhibitor. This point was best exemplified by combining the Pd
catalyst and 10 in the absence of 12, which upon subsequent
addition of 12 and base led to no reaction. This is most likely
due to the Pd catalyst coordinating irreversibly with at least 2
equivalents of 10. Related to this problem is inhibition caused
by the mercapto dimer, 15, an impurity formed through
oxidation (shown in Scheme 3). Samples of 10 have been
Scheme 3. Proposed inhibition of Pd catalyst by 15
undergoing oxidative addition
Scheme 4. Optimized process for steps 1, 2, and 2R
C
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 2. Process impurities formed in steps 1−2.
Scheme 5. Proposed mechanism for exchange of aryl groups on Xantphos in step 1 coupling based on observed side products
of a phenyl group with 10 (compound 24). More than one
phenyl group can undergo exchange on Xantphos, leading to a
complex mixture of byproducts.
catalysis.17 To further facilitate oxidative insertion of Pd,
electron-withdrawing protecting groups for the indazole
nitrogen were examined. From this study, the acyl protecting
group emerged as the best option since it could be introduced
in situ with acetic anhydride and Hunig’s base. With inclusion of
the acyl protecting group, the undesired side reaction of the
indazole moiety undergoing Michael addition with 2-vinyl-
pyridine (6) is suppressed, while also lowering the temperature
of the Heck reaction from 110 °C down to 90 °C.
Equally important as the acyl protecting group is the choice
of catalyst. A catalyst derived from Pd(OAc)2 and Xantphos
was found to enable highly efficient coupling while minimizing
formation of saturated impurity 26 (Figure 3) which is
challenging to purge due to its structural similarity to 1. In
the absence of Xantphos, 4 mol % of Pd(OAc)2 typically
enables only 10−20% conversion to 30 over 24 h at 90 °C,
while the optimal conditions (4 mol % of Pd and Xantphos)
achieve complete conversion during this time. The relative
stoichiometry of the Pd and Xantphos not only influences the
rate of the reaction, but also determines the extent to which 26
is formed. Excess Xantphos inhibits the Pd catalyst, most likely
through blocking coordination of the desired substrates with a
second equivalent of Xantphos on the Pd catalyst. In contrast,
using an excess of Pd relative to Xantphos leads to increased
levels of 26, since unligated Pd promotes formation of 26. For
example, using a 2:1 molar ratio of Pd to Xantphos results in
levels of about 1% of 26 in solution, while using a 1:1
stoichiometry limits levels to only 0.2%. For these reasons, the
process is typically operated near a 1:1 stoichiometry of
Xantphos and Pd.
While the reaction temperature and the composition of the
catalyst significantly impact the levels of side products, another
key element in the process is the stoichiometry of 2-
vinylpyridine, 6. Using 4.5−8 equivalents of 6 was found to
lower the levels of a number of impurities (Figure 3), including
those arising from dehalogenation (13), rearrangement of the
acyl group (27), and inadvertent coupling with thiol (28). The
trends in Table 2 show that the level of 26 is generally
unaffected by the amount of 6; however, the formation of these
other tracked impurities are well controlled with excess 6. In
the case of 28, the reversibility of C−S bond formation is at
work.18 During the Heck reaction, the Pd catalyst is capable of
breaking the C−S bond that was formed in the first step of the
Control of Impurities through Recrystallization at
Step 2R. Originally, reslurries of 14 were used to control
impurities in the process at step 2, but it became apparent that a
recrystallization of 14 would be necessary to control the
Xantphos-related impurities in addition to the other process-
related impurities. To efficiently control impurities and ensure
reliable performance of the downstream Heck reaction, a
recrystallization at step 2R was developed. Dissolution of crude
14 is initially achieved in a blend of NMP and 1,2-DAP. To this
solution is added methanol followed by water as the
antisolvents. The 1,2-DAP is included to aid in color removal
and to help solubilize Pd during isolation. In general, the
recrystallization controls all of the organic impurities well as
shown in Table 1, with the most prevalent impurities being 13
Table 1. Ranges of impurities in commercial scale batches of
14
observed range in
crude 14 (%)
observed range in
recrystallized 14 (%)
impurity
13
17
18
0.52−2.4
0.13−1.4
≤0.1
0.19−0.79
0.08−0.32
≤0.05
total
impurities
1.7−5.0
0.45−1.2
and 17, typically, each at a level below 1%, while all other
impurities are at levels below 0.2%. The recrystallization step
2R yields 14 in about a 85−90% recovery and ≥98% overall
purity. This recrystallization ensures that none of the impurities
introduced early in the process impact the purity profile of the
drug substance.
Development of the Heck Reaction. The further
development of the Heck reaction was essential to streamline
the overall process. The use of Proton-Sponge (Scheme 2) was
an intriguing way of minimizing formation of 25 (Figure 3);
however, its stoichiometric use was not practical and challenged
the control strategy with the difficulty of its purge due to its
highly crystalline nature. The oxidative addition of 14 onto Pd
is challenged by the relatively electron-rich indazole system, as
proven by the need for a 3-iodo moiety to achieve efficient
D
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 3. Typical impurities generated during the Heck reaction.
Table 2. Levels of impurities monitored in step 5 (deprotection) reaction mixture on lab scale which arise at step 4 (Heck
reaction) based on the amount of 2-vinylpyridine (6)
6 (equiv)
% 26 (saturated impurity)
% 13 (dehalogenated impurity)
% 27 (rearranged impurity)
% 28 (C−S coupling)
% total impurities
1.5
3.0
6.0
8.0
0.19
0.33
0.29
0.41
5.48
3.72
2.40
2.10
14.83
7.20
3.27
2.98
2.72
1.06
0.31
0.28
41
30
23
20
Scheme 6. Optimized conditions for the Heck reaction, steps 3−5
process between 10 and 12. In doing so, the liberated 10 is able
to undergo coupling with acylated 14 (intermediate 29) to
afford 28 after deprotection. While some reports describe using
tertiary amines as solvents to alleviate the C−S bond
scrambling,18b,c in this case it is sufficient to increase the
stoichiometry of 6.
14, Hunig’s base, and acetic anhydride. Once the acylation to
afford intermediate 29 is complete, 6 (6 equiv) is charged, and
the reaction is heated for 24 h at 90 °C. After completing the
Heck reaction at step 4, acylated product 30 is deprotected in
situ (step 5) with 1,2-diaminopropane (1,2-DAP) in a solution
of THF, followed by crystallization with water to yield 1c as a
THF solvated form in about 75% for the three steps in ≥99%
chemical purity (Table 3).
Prior work has demonstrated the effectiveness of 1,2-DAP in
combination with chelating phosphine ligands for maintaining
Pd in solution while crystallizing 1.9 For the optimized process,
1,2-DAP was chosen not only for solvating Pd but also for
enabling deprotection of the acyl group prior to crystallization
The structure of the amine base has a profound effect on the
efficiency of the catalyst in the Heck reaction. When
substituting tributylamine for N,N-diisopropylethylamine (Hu-
nig’s base) in step 4, the conversion is only about 75% after 24
h, over which time the reaction is normally >98% complete.
Even more striking, is that using N-ethylmorpholine or 4-
dimethylaminopyridine enables only 34% conversion after 24 h.
Using tertiary amine bases with structures similar to N,N-
diisopropylethylamine provide optimal results, which is
consistent with prior reports.19 For example, N,N-dicyclohex-
ylmethylamine and N,N-diisopropylmethylamine both have
roughly the same level of reactivity as N,N-diisopropylethyl-
amine, enabling reaction completion within 24 h at 90 °C.
To streamline operations, the protection (step 3), Heck
reaction (step 4), and deprotection (step 5) are executed in a
telescoped manner (Scheme 6). Typically, the Pd(OAc)2 and
Xantphos are charged to NMP at the start of step 3 to enable
formation of the catalyst, which is followed by the addition of
Table 3. Levels of impurities in commercial-scale batches of
isolated crude axitinib (1c)
impurity
observed range (%)
13
25
26
28
0.21−0.30
≤0.05
0.13−0.17
0.11−0.13
0.51−0.82
≤20−42 ppm
total impurities
residual Pd
E
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Scheme 7. Conversion of crude axitinib to form 4 via an isopropanol solvate
of crude axitinib. While the 1,2-DAP and Xantphos in the
process were found to be effective in maintaining the Pd in
solution during the crystallization of 1c, it was discovered that
the stoichiometry of 6 impacts the effectiveness of the Pd
removal. Remarkably, when using 6 equiv of 6 in combination
with 4 equiv of 1,2-DAP, 1c is consistently isolated with Pd
levels below 50 ppm. Using less of 6 (such as 3 equiv) or less of
1,2-DAP (such as only 2−3 equiv) results in Pd levels that are
typically an order of magnitude higher. On production scale,
the Pd levels have ranged from 42 ppm to below 20 ppm.
Control of Solid Form. While the purity and Pd level are
well controlled during the isolation of 1c (crude axitinib) at
step 5, additional processing controls are required for solid
form. 1c is isolated as a THF solvated form and requires
conversion to an anhydrous form suitable for formulation.
Earlier in development, anhydrous form 4 (1b) was
manufactured for use in the clinic. Prior to the polymorph
conversion, a recrystallization from NMP, 1,2-DAP, and
methanol was implemented to ensure that Pd was sufficiently
purged (Scheme 2). After the recrystallization, axitinib was
dissolved in a combination of acetic acid and methanol followed
by distillation from xylenes to convert axitinib from an acetic
acid solvate to 1b. Unfortunately, the conditions required for
distillation of acetic acid also result in acylation at the 1-position
of the indazole ring of axitinib to provide intermediate 30 as an
impurity (structure shown in Scheme 6). During studies of
solvated forms, it was found that an isopropanol solvated form
(1d) will readily undergo conversion to 1b by heating in
heptane. Consequently, this more preferable solid form
conversion was incorporated into the process at step 5R
following the recrystallization from NMP, 1,2-DAP, and
methanol. The recrystallized product at step 5R is reslurried
in isopropanol and dried to remove excess isopropanol. The
solids are subsequently reslurried in heptane at 90 °C for
conversion to 1b (step 6 in Scheme 7).
The implementation of this improved process facilitated
access and discovery of other anhydrous forms. Soon after this
process was implemented, forms 6, 25, and 41 were prepared
and characterized. All of these forms were found to be more
stable than form 4. In all, there are five anhydrous solid forms
known for axitinib, forms 1, 4, 6, 25, and 41.20 Form 41 (1e) is
the most stable anhydrous form and therefore was selected for
commercial development. To access 1e, the same recrystalliza-
tion was implemented prior to the polymorph conversion. After
recrystallization from NMP, 1,2-DAP, and methanol, the
isopropanol solvate (1d) was accessed and converted to 1e
by a reslurry in refluxing ethanol with seed crystals. Ethanol was
found to be an ideal solvent since the anhydrous forms are
thermodynamically favored as compared to the ethanol solvate
near room temperature. Most other solvents convert axitinib to
a solvated form, for which over 70 solvates have been
characterized to date.20a,21
Using this process, recrystallization of crude axitinib with
subsequent reslurries in isopropanol and ethanol afford 1e in
80% yield and ≥99.9% purity with Pd levels below 10 ppm.
Typically, all impurities are below 0.10% due to the efficient
control offered by the recrystallization and polymorph
conversion. An advantage of isolating such a highly crystalline
form (mp 225 °C) is that impurities are more readily purged
from the process, including Pd. Spiking studies demonstrated
that processing 1c through step 6 with an ingoing Pd level of up
to 300 ppm consistently yields 1e with Pd levels below 10 ppm.
While the process to prepare 1e as a slurry in ethanol
performed suitably for development purposes, it was highly
desirable to design a homogeneous process that would be more
robust and would not be as reliant on stirring rate or vessel
configuration. A process involving the seeding of a saturated
solution of axitinib was developed and is shown in Scheme 8.
Scheme 8. Integrating the solid form conversion with the
final recrystallization at step 6
For this process, the polymorph conversion is integrated with
the recrystallization step by replacing methanol with ethanol as
the antisolvent. After adding a solution of axitinib in NMP/
THF to hot ethanol, seed crystals of 1e are introduced to the
super-saturated solution to enable formation of the desired
polymorph in a highly robust manner. A relatively slow cooling
rate is used to avoid nucleation of any solvated forms. The
overall solvent composition is vital to produce the desired form,
as too much THF or NMP will also favor solvated forms.
By directly preparing the desired solid form (1e) in the
recrystallization, there is no longer any need for isolations of
intermediate solvated forms after that of 1c at step 5. With this
streamlined process for step 6, 1e is isolated in 80−85% yield
based on ingoing 1c and is obtained in ≥99.9% purity (Table
4).
CONCLUSIONS
■
The second-generation route (Scheme 2) was deemed to be an
optimal manufacturing route due to the short length and late
introduction of the potentially unstable olefin. This route was
further developed and optimized to about 50% overall yield for
the process (Scheme 9). Essential to the efficient assembly in
the route is the implementation of two Pd-catalyzed couplings.
F
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
column (2.7 μm; 100 mm × 3.0 mm) (Advanced Materials
Technology, Wilmington, DE), running an acetonitrile/0.05%
trifluoroacetic acid gradient was used for assessment of steps 1
and 2. Mobile phase solution A was 0.05% trifluoroacetic acid
in water (v/v), and mobile phase solution B was 0.05%
trifluoroacetic acid in acetonitrile (v/v). Gradient elution was
performed as shown in Table 5 below with a run time of 20 min
Table 4. Maximum levels of impurities in commercial scale
batches of 1e
impurity
max. % in manufacturing batches
13
25
26
28
≤0.05
≤0.05
≤0.05
≤0.05
≤0.05%
7 ppm
total impurities
residual Pd
Table 5. Gradient program for steps 1−2
time (min)
mobile phase A (%)
mobile phase B (%)
0
15
20
90
50
10
10
50
90
For the Migita coupling at step 1 it is important to use a mild
base with efficient inertion of the headspace to minimize
inhibition of the Pd catalyst. In the Heck reaction, the electron-
rich iodoindazole 14 suffers from inefficient oxidative addition
which is overcome in part with in situ acylation. The acyl group
not only facilitates reactivity with the Pd catalyst but also blocks
formation of 25 as a side product. The choice of Xantphos in
the Heck reaction limits the formation of impurity 26, which is
difficult to purge. Ultimately, Xantphos contributes to the
control of Pd in a synergistic manner with 1,2-diaminopropane
and 6 during isolation of 1c at step 5.
Control of solid form is critical for axitinib as there are five
known anhydrous forms. As the synthesis was optimized and
the drug substance was routinely prepared in high purity, these
solid forms were identified and characterized. From these
studies, form 41 (1e) emerged as the most stable form which
was later nominated for commercial development. As a benefit,
the relatively high crystallinity of 1e facilitates control of
impurities, including the purge of Pd. By incorporating the
polymorph conversion into the recrystallization at step 6, the
conversion of 1c to 1e ensures consistent purge of organic
impurities as well as Pd.
and a re-equilibration time of approximately 5 min. The flow
rate was 0.7 mL/min and the column temperature was 50 °C.
The injection volume was 5 μL, and UV detection was
performed at 220 nm. Sample and standard solutions were
prepared in water/acetonitrile/trifluoroacetic acid (500:500:0.5,
v/v/v) at a nominal concentration of 0.1 mg/mL.
For steps 4 and 5 and axitinib purity and assay, the HPLC
column was a Symmetry C18 150 mm × 4.6 mm, 5 μm
(Waters Corp., Milford, MA). Mobile phase solution A was
composed of water/acetonitrile/trifluoroacetic acid
(900:100:0.5, v/v/v), and mobile phase solution B was
water/acetonitrile/trifluoroacetic acid (200:800:0.5, v/v/v).
Gradient elution was performed as shown in Table 6 below
Table 6. Gradient program for steps 4 and 5 and axitinib
assay/purity
time (min)
mobile phase A (%)
mobile phase B (%)
0
1
100
100
68.5
0
0
0
EXPERIMENTAL SECTION
18
40
41
31.5
100
100
■
General. All reactions were carried out in dry reaction
vessels under an atmosphere of dry nitrogen unless otherwise
specified. All reagents and solvents were used as received
without further purification unless otherwise specified. For
most steps, reactions were monitored, and purity was assessed
by HPLC, using an Agilent 1100 series instrument. A Halo C18
0
with a run time of 41 min and a 4-min re-equilibration time.
The flow rate was 1.0 mL/min, and the column temperature
was 30 °C. The injection volume was 10 μL, and UV detection
Scheme 9. Optimized manufacturing route for axitinib
G
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
was performed at 220 nm. Axitinib sample and standard
solutions were prepared in water/acetonitrile/acetic acid
(700:300:30, v/v/v) at a nominal concentration of 0.2 mg/mL.
Crude 2-(3-Iodo-1H-indazol-6-ylthio)-N-methylbenza-
mide (14) (Steps 1−2). To a nitrogen-purged reactor vessel at
ambient conditions is charged NMP (82.4 kg) followed by 12
(40 kg, 164 mol, 1.0 equiv), Xantphos (0.96 kg, 0.010 equiv),
tris(dibenzylideneacetone)dipalladium [Pd2(dba)3] (0.76 kg,
0.005 equiv), and sodium bicarbonate (15.16 kg, 1.1 equiv).
The reactor is purged, and the mixture is held at a target of 25
°C and stirred for about 30 min. A solution of 10 (28.78 kg,
1.05 equiv) in NMP (61.8 kg) is added over about 1 h. The
batch mixture is heated to a target of 50 °C. The reaction is
held for at least 2 h and then sampled for step 1 reaction
completion. The mixture is cooled to a target of 25 °C after
passing the step 1 reaction completion (typically not more than
5% of 12 remaining). A solution of iodine (87.5 kg, 2.0 equiv)
in NMP (47.4 kg) is added at a rate that allows maintaining the
temperature near a target of 25 °C. A solution of potassium
hydroxide (43.0 kg, 2.0 equiv) in water (40 L) is added over at
least 30 min with a target of 25 °C. The reaction mixture is held
for at least 5 h, and then sampled for step 2 reaction
completion. After passing the reaction completion (typically
not more than 7% of 13 remaining), the reaction mixture is
quenched into a solution of ascorbic acid (25.36 kg, 1.75 equiv)
in water (200 L) over about 1 h at a target of 45 °C. Methanol
(120 L) is added to the slurry. After stirring the suspension for
about 3 h at 45 °C, the slurry is filtered warm. The filtercake is
washed with water (120 L) followed by methanol (120 L), and
then dried to give crude 14 as a brown crystalline solid (62 kg,
92% yield uncorrected by assay).
Recrystallization of 2-(3-Iodo-1H-indazol-6-ylthio)-N-
methylbenzamide (14) (Step 2R). To a purged reactor
vessel is added crude 14 (62 kg, 151.5 mol) followed by NMP
(207 kg) and 1,2-diaminopropane (1.2 kg) at ambient
conditions. This solution is heated to about 60 °C and stirred
to ensure dissolution of the solids. Methanol (338 L) is charged
while maintaining temperature at a target of 55 °C. Water (335
L) is then added while maintaining a target of 55 °C and
precipitating the product from solution. The resulting slurry is
held at temperature, then cooled to approximately 20 °C and
granulated for about 3 h. The slurry is filtered. The filter cake is
reslurried with water (200 L), reslurried with methanol (135
L), and blown dry with nitrogen. The solids were dried and
isolated to give recrystallized 14 as a pale-yellow crystalline
solid (45.3 kg, 73% yield uncorrected by assay) in 99.5% purity.
1H NMR (DMSO-d6, 300 MHz) δ (ppm) 13.53 (s, 1H), 8.35
(q, J = 4.7 Hz, 1H), 7.56 (s, 1H), 7.51−7.40 (m, 2H), 7.36−
7.23 (m, 3H), 7.13 (dd, J = 8.5, 1.3 Hz, 1H), 7.06−7.01 (m,
1H), 2.76 (d, J = 4.7 Hz, 3H). 13C NMR (DMSO-d6, 75 MHz)
δ (ppm) 168.2, 141.3, 137.6, 135.6, 134.4, 130.7, 128.2, 126.7,
126.7, 125.9, 122.0, 114.7, 94.1, 26.5; mp 219−221 °C.
reaction is complete (typically not more than 2% of 29
remaining), the reaction mixture is cooled to a target of 50 °C.
The reaction mixture is diluted with THF (121 kg) and
passed through a polishing filter. The batch is further diluted
with THF (40 kg) and 1,2-diaminopropane (33 kg, 4 equiv) at
a target of 50 °C. After stirring the solution at a target of 50 °C
for at least 30 min, water (496 L) is slowly charged while
maintaining the temperature near a target of 50 °C. The
mixture is stirred at a target of 50 °C for about 12 h and then
cooled to a target of 15 °C. The product slurry is filtered. The
cake is washed with water (136 L). The crude material is
reslurried warm in THF (121 kg) with 1,2-diaminopropane
(2.2 kg, 0.25 equiv) for ∼4 h and then cooled to ∼15 °C for
filtration. The filtered solids are washed with THF (121 L),
dried, and isolated to afford crude axitinib 1c (THF solvate),
34.0 kg (80% yield uncorrected by assay) in 99.3% purity.
Residual Pd content was determined to be not more than 20
ppm by atomic absorption spectroscopy. Mp 216−217 °C.
Recrystallization and Polymorph Conversion to Form
41 of (E)-N-Methyl-2-(3-(2-(pyridin-2-yl)vinyl)-1H-inda-
zol-6-ylthio)benzamide (1e) (Step 6). To a purged reactor
vessel is added 1c (34.0 kg, 88.0 mol) followed by NMP (87.6
kg) and THF (30.3 kg). The reactor is purged, and the mixture
is heated to 68 °C and stirred to allow for dissolution of the
ingoing material. In a separate vessel, ethanol (631 kg) is added
(process can use either absolute ethanol or methanol denatured
ethanol), and the reactor is heated to about 73 °C and stirred.
Once axitinib has completely dissolved in NMP/THF in the
first vessel, this solution is filtered to be speck-free and is
transferred into the second vessel containing the ethanol. To
this hot supersaturated solution is added an amount of form 41
axitinib seed crystals (1.70 kg, 0.05 equiv). The heating is
increased to a reflux, and the mixture is held for about 7 h. The
batch is slowly cooled to 20 °C at a rate of about 0.07 °C/min.
After holding at the final isolation temperature, the slurry is
filtered and washed with ethanol (102 L) to displace the
mother liquor and remove residual NMP. After drying, 1e is
isolated as a white crystalline solid (28.0 kg, 82% yield
uncorrected by assay) in 99.9% purity. The Pd level was
determined to be of not more than 6 ppm by atomic absorption
1
spectroscopy. H NMR (DMSO-d6, 300 MHz) δ (ppm) 13.35
(1H, s), 8.61 (1H, d, J = 3.8 Hz), 8.39 (1H, q, J = 4.4 Hz), 8.21
(1H, d, J = 8.8 Hz), 7.96 (1H, d, J = 16.4 Hz), 7.85−7.76 (1H,
m), 7.66 (1H, d, J = 7.8 Hz), 7.61 (1H, s), 7.58 (1H, d, J = 16.5
Hz), 7.50 (1H, dd, J = 5.7 Hz), 7.36−7.23 (3H, m), 7.19 (1H,
dd, J = 8.4 Hz, 1.2 Hz), 7.05 (1H, dd, J = 7.5, 1.5 Hz), 2.78
(3H, d, J = 4.5 Hz). 13C NMR (DMSO-d6, 75 MHz) δ (ppm)
168.2, 155.2, 149.8, 142.4, 142.2, 137.3, 136.0, 132.9, 130.6,
129.5, 128.1, 126.5, 25.9, 124.1, 123.0, 122.9, 122.1, 120.6,
115.1, 26.5. Mp 224−225 °C.
AUTHOR INFORMATION
■
Crude (E)-N-Methyl-2-(3-(2-(pyridin-2-yl)vinyl)-1H-in-
dazol-6-ylthio)benzamide (1c) (Steps 3−5). Pd(OAc)2
(1.00 kg. 0.04 equiv), Xantphos (2.56 kg, 0.04 equiv), and
NMP (107 kg) are charged to a nitrogen-purged reaction vessel
to form the reaction catalyst. After this, Hunig’s base (43 kg, 3
equiv), and 14 (45.3 kg, 110.6 mol, 1 equiv) are added under
ambient conditions. The reaction is heated to a target of 50 °C,
and acetic anhydride (23 kg, 2 equiv) is added to effect
acylation. Once acylation is complete (not more than 2% of 14
remaining), 2-vinylpyridine (70 kg, 6 equiv) is added. The
heating is further increased to a target of 90 °C. Once the Heck
Corresponding Author
*E-mail: robert.a.singer@pfizer.com.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Paul Glynn for overseeing the scale-up of this
technology in the Pilot Plant. We thank John O’Connell and
Philomena Enright for oversight of the scale-up in commercial
manufacture. We are grateful for the work of the Structure
H
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
of the coupling reaction. Alternatives to zinc were also found,
including ascorbic acid and poly(methylhydrosiloxane).
(15) Xiang, Y.; Caron, P.-Y.; Lillie, B. M.; Vaidyanathan, R. Org.
Process Res. Dev. 2008, 12, 116−119.
(16) Exchange of aryl groups on Xantphos and related
triarylphosphines has been observed, see: (a) Yin, J.; Zhao, M. M.;
Huffman, M. A.; McNamara, J. M. Org. Lett. 2002, 4, 3481−3484.
(b) Yin, J.; Buchwald, S. L. Org. Lett. 2000, 2, 1101−1104.
(c) Hamann, B. C.; Hartwig, J. F. J. Am. Chem. Soc. 1998, 120,
3694−3703. (d) Segelstein, B. E.; Butler, T. W.; Chenard, B. L. J. Org.
Chem. 1995, 60, 12−13.
Elucidation Group who identified many of the impurities
encountered during development. We thank Dave Damon,
Barbara Sitter, Carrie Wager, and John Lucas for carrying out
screening of both metal-catalyzed coupling reactions to verify
the optimal catalysts were selected for development. We
acknowledge Gregg Tavares, Yanqiao Xiang, and Matthew
Frierson for their assistance in the analytical research and
development. We thank John Barry and Jerome McCormick for
providing the analytical data from the axitinib commercial
manufacturing campaigns.
(17) The 3-bromoindazole derivative of 14 fails to react with the Pd
catalyst and remains unchanged under the reaction conditions.
(18) (a) Yamamoto, T.; Sekine, Y. Inorg. Chim. Acta 1984, 83, 47−
53. (b) Murata, M.; Buchwald, S. L. Tetrahedron 2004, 60, 7397−7403.
(c) Kreis, M.; Braese, S. Adv. Synth. Catal. 2004, 347, 313−319.
(19) (a) Littke, A. F.; Fu, G. C. J. Am. Chem. Soc. 2001, 123, 6989−
7000. (b) Hills, I. D.; Fu, G. C. J. Am. Chem. Soc. 2004, 126, 13178−
13179.
(20) (a) Campeta, A. M.; Chekal, B. P.; Abramov, Y. A.; Meenan, P.
A.; Henson, M. J.; Shi, B.; Singer, R. A.; Horspool, K. R. J. Pharm. Sci.
2010, 99, 3874−3886. (b) Chekal, B. P.; Campeta, A. M.; Abramov, Y.
A.; Feeder, N.; Glynn, P. P.; McLaughlin, R. W.; Meenan, P. A.; Singer,
R. A. Org. Process Res. Dev. 2009, 13, 1327−1337.
(21) Samas, B.; Seadeek, C.; Campeta, A. M.; Chekal, B. P. J. Pharm.
Sci. 2011, 100, 186−194.
REFERENCES
■
(1) Ferrara, N. Oncologist 2004, 9 (Suppl. 1), 2−10.
(2) (a) Hu-Lowe, D. D.; Zou, H. Y.; Grazzini, M. L.; Hallin, M. E.;
Wickman, G. R.; Amundson, K.; Chen, J. H.; Rewolinski, D. A.;
Yamazaki, S.; Wu, E. Y.; McTigue, M. A.; Murray, B. W.; Kania, R. S.;
O’Connor, P.; Shalinsky, D. R.; Bender, S. L. Clin. Cancer Res. 2008,
14, 7272−7283. (b) Kania, R. S.; Bender, S. L.; Borchardt, A. J.;
Braganza, J. F.; Cripps, S. J.; Hua, Y.; Johnson, M. D.; Johnson, T. O.,
Jr.; Luu, H. T.; Palmer, C. L.; Reich, S. H.; Tempczyk-Russell, A. M.;
Teng, M.; Thomas, C.; Varney, M. D.; Wallace, M. B. PCT Int. Appl.
WO/2001/002369, 2001.
(3) Rini, B. I.; Escudier, B.; Tomczak, P.; Kaprin, A.; Szczylik, C.;
Hutson, T. E.; Michaelson, M. D.; Gorbunova, V. A.; Gore, M. E.;
Rusakov, I. G.; Negrier, S.; Ou, Y.-C.; Castellano, D.; Lim, H. Y.;
Uemura, H.; Tarazi, J.; Cella, D.; Chen, C.; Rosbrook, B.; Kim, S.;
Motzer, R. J. Lancet 2011, 378, 1931.
(4) (a) Beletskaya, I. P.; Cheprakov, A. V. Chem. Rev. 2000, 100,
3009−3066. (b) Farina, V. Adv. Synth. Catal. 2004, 346, 1553−1582.
(5) (a) Migita, T.; Shimizu, T.; Asami, Y.; Shiobara, J.; Kato, Y.;
Kosugi, M. Bull. Chem. Soc. Jpn. 1980, 53, 1385−1389. (b) Kondo, T.;
Mitsudo, T.-a. Chem. Rev. 2000, 100, 3205−3220. (c) Hartwig, J. F.
Acc. Chem. Res. 2008, 41, 1534−1544.
(6) Babu, S.; Dagnino, R., Jr.; Ouellette, M. A.; Shi, B.; Tian, Q.;
Zook, S. E. PCT Int. Appl. WO/2006/048745, 2006.
(7) Boulouard, M.; Schumann-Bard, P.; Butt-Gueulle, S.; Lohou, E.;
Stiebing, S.; Collot, V.; Rault, S. Bioorg. Med. Chem. Lett. 2007, 17,
3177−3180.
(8) Ewanicki, B. L.; Flahive, E. J.; Kasparian, A. J.; Mitchell, M. B.;
Perry, M. D.; O’Neill-Slawecki, S. A.; Sach, N. W.; Saenz, J. E.; Shi, B.;
Stankovic, N. S.; Srirangam, J. K.; Tian, Q.; Yu, S. U.S. Pat. Appl. , U.S.
2006094881 A1 20060504, 2006.
(9) (a) Flahive, E. J.; Ewanicki, B. L.; Sach, N. W.; O’Neill-Slawecki,
S. A.; Stankovic, N. S.; Yu, S.; Guinness, S. M.; Dunn, J. Org. Process
Res. Dev. 2008, 12, 637−645. (b) Flahive, E. J.; Ewanicki, B. L.; Yu, S.;
Higginson, P. D.; Sach, N. W.; Morao, I. QSAR Comb. Sci. 2007, 26,
679−685.
(10) Itoh, T.; Mase, T. Org. Lett. 2004, 6, 4587−4590.
(11) LiCl and other lithium salts have been used to influence the
product distribution in the Heck reaction, presumably through
affecting the Pd during migratory insertion while also impacting the
extent of Pd maintained in solution. See (a) Okano, K.; Tokuyama, H.;
Fukuyama, T. J. Am. Chem. Soc. 2006, 128, 7136. (b) Schmidt, A. F.;
Halaiqa, A. A.; Smirnov, V. V. Synlett 2006, 2861.
(12) (a) Zanella, R.; Ros, R.; Grazian, M. Inorg. Chem. 1973, 12,
2736−2738. (b) Lam, C, T.; Senoff, C. V. Can. J. Chem. 1973, 51,
3790−3794. (c) Canich, J. A. M.; Cotton, F. A.; Dunbar, K. R.;
Falvello, L. R. Inorg. Chem. 1988, 27, 804−811. (d) Louie, J.; Hartwig,
J. F. J. Am. Chem. Soc. 1995, 117, 11598−11599.
(13) Zinc has been shown to enable couplings of aryl disulfide
species previously. For nickel couplings, see: Taniguchi, N. J. Org.
Chem. 2004, 69, 6904−6906. For Pd couplings using dppf as the
supporting ligand, see: Fukuzawa, S.-i.; Tanihara, D.; Kikuchi, S. Synlett
2006, 2145−2147.
(14) When charging 10 to the Pd catalyst in the absence of 12
initially, this leads to complete inhibition of the catalyst; however,
charging catalytic zinc reactivates the catalyst, enabling full conversion
I
dx.doi.org/10.1021/op400088k | Org. Process Res. Dev. XXXX, XXX, XXX−XXX