Evolution of the process for the preparation of a selective erbb vegf receptor inhibitor

Article abstract of DOI:10.1055/s-0032-1317540

An efficient synthetic route to the potent and selective ErbB VEGF receptor inhibitor, BMS-690514 (1) is described. Strategic modifications in both approach and procedure addressed several issues, which led to a safe, efficient, and economical process for the preparation of multi-kilogram quantities of 1. The convergent route involves alkylation of a suitably protected (3R,4R)-4-aminopiperidin-3-ol with the triethyl(alkyl)ammonium salt of a functionalized pyrrolotriazine 3a followed by deprotection to provide 1 as the crystalline free base. Georg Thieme Verlag Stuttgart - New York.

Full text of DOI:10.1055/s-0032-1317540

CLUSTER
▌305
cEluvsteor lution of the Process for the Preparation of a Selective ErbB VEGF
Receptor Inhibitor
Synthesis of a Selective ErbB VEGF Receptor Inhibitor
Boguslaw Mudryk,* Amit Joshi,* Adrian Ortiz, Ian S. Young, James R. Sawyer, Bin Zheng, Masano Sugiyama,
Zhongping Shi, Jale Müslehiddinoğlu,1 R. Michael Corbett,2 David R. Kronenthal, David A. Conlon
Chemical Development, Bristol-Myers Squibb Company, One Squibb Drive, New Brunswick, NJ 08903, USA
Fax +1(732)2273001; E-mail: boguslaw.mudryk@bms.com; E-mail: amit.joshi@bms.com
Received: 01.10.2012; Accepted: 17.10.2012
C–N bond
Abstract: An efficient synthetic route to the potent and selective
ErbB VEGF receptor inhibitor, BMS-690514 (1) is described. Stra-
tegic modifications in both approach and procedure addressed sev-
eral issues, which led to a safe, efficient, and economical process for
the preparation of multi-kilogram quantities of 1. The convergent
route involves alkylation of a suitably protected (3R,4R)-4-amino-
piperidin-3-ol with the triethyl(alkyl)ammonium salt of a function-
alized pyrrolotriazine 3a followed by deprotection to provide 1 as
the crystalline free base.
HO
formation
H2N
HN
N
OMe
N
N
N
BMS-690514 (1)
Key words: antitumor agents, protected piperidines, protecting
groups, heterocycles, Schiff bases
fragment coupling,
deprotection
R1
R2
N
HN
N
OMe
HO
The human epidermal growth factor receptor (HER or
EGFR) and the vascular endothelial growth factor recep-
tor (VEGFR) signaling pathways are implicated in pro-
cesses governing tumor growth and proliferation. The
development of potent and selective inhibitors of HER
and VEGFR may provide additional clinical benefit in the
treatment of non-small-cell lung cancer (NSCLC), meta-
static breast cancer, and other solid malignancies. As part
of a drug discovery program at Bristol-Myers Squibb, the
functionalized pyrrolotriazine BMS-690514 (1) was iden-
tified as an orally active, selective, and potent dual inhib-
itor of both HER and VEGFR.3,4
OH
+
N
N
N
H
2
3a
Scheme 1 Retrosynthetic analysis of BMS-690514 (1)
To support non-clinical and clinical studies, multi-kilo-
gram quantities of both pyrrolotriazine 3a and piperidine
2 were required. A scalable and efficient route to function-
alized pyrrolotriazine 3a was developed and is summa-
rized retrosynthetically in Scheme 2. This work has been
described in detail in a previous report.5
Herein, we report the process development and demon-
stration on multi-kilogram scale of a synthetic route to 1
that supports preclinical and clinical studies. In addition,
we describe our laboratory development efforts towards
an optimized end game, resulting in a process that should
be viable on a commercial scale.
OH
EtO2C
amination,
reduction
HN
OMe
HO
N
N
N
The retrosynthetic analysis of 1 is outlined in Scheme 1.
Strategic C–N bond scission simplifies the target mole-
cule into its constituent parts: enantiomerically pure pi-
peridine 2 and pyrrolotriazine 3a. In the forward sense,
these fragments could be united by activation of the pri-
mary alcohol of 3a, and alkylation with variably protected
piperidine 2. To ensure the success of this key step, we an-
ticipated that a judicious selection of protecting groups
would be required for the C-4 amino group present in 2
(see below). Finally, removal of the protecting groups
would provide the target structure 1 as the crystalline free
base.
N
N
N
formamide
condensation
3a
Br
CO2Et
CO2Et
NH2⋅MSA
C-alkylation,
CO2Et
CO2Et
cyclodehydration
+
H
N
N
N
NaO
CBz
Scheme 2 Retrosynthetic analysis of pyrrolotriazine core 3a
In 1998, Langlois and Calvez reported the synthesis of
(3S,4S)-l-benzyl-4-N-benzylamino-3-hydroxypiperidine
[(+)-4a] through ring expansion of (2S,3S)-l-benzyl-3-N-
SYNLETT 2013, 24, 0305–0312
Advanced online publication: 23.11.2012
DOI: 10.1055/s-0032-1317540; Art ID: ST-2012-Y0838-C
© Georg Thieme Verlag Stuttgart · New York
0
9
3
6
-
5
2
1
4
1
4
3
7
-
2
0
9
6

306
B. Mudryk et al.
CLUSTER
benzylamino-2-hydroxymethylpyrrolidine,6 which was attempts to prepare and directly couple the mesylate of 3a
derived from (S)-pyroglutamic acid.7 Our original route resulted in an unsatisfactory impurity profile. Our col-
(see Scheme 3) to the enantiomer [(–)-4a] utilized a simi- leagues in Discovery Chemistry had previously shown
lar synthetic sequence with sequential Boc protection and that the corresponding bromide reacted with triethylamine
selective debenzylation as the last two steps to arrive at in- to produce the triethyl(alkyl)ammonium bromide, which
termediate 2a (R1 = Boc, R2 = H).4h This sequence re- could then be used to N-alkylate multiple piperidine de-
quired some procedural modifications to enable the safe, rivatives in good yields.4h After surveying other reagents
efficient, and reproducible preparation of piperidine 4a on for activation of the alcohol, we found that the tetraalkyl-
scale (see the Supporting Information for more details). ammonium derivative 3b was formed most efficiently by
Ultimately, the seven-step process proceeded in 22% displacement of the mesylate by triethylamine. Derivative
overall yield, and was successfully scaled to 70 kg without 3b was then coupled efficiently with 2a to produce 5 in
major issues. The final two transformations (4a → 2a; excellent yield and purity. Interestingly, replacing trieth-
Boc protection, followed by selective debenzylation) ylamine with diisopropylethylamine did not lead to for-
were telescoped to provide high quality 2a after crystalli- mation of the desired tetralkylammonium salt,
zation. With the orthogonally protected piperidine 2a and presumably for steric reasons.
pyrrolotriazine 3a in hand, we turned our attention to the
One of the major impurities (7) from this step originated
remaining transformations of the process, as shown in
from an undesired Friedel–Crafts alkylation pathway (see
Scheme 4.
Scheme 5). This impurity became a significant concern
To this end, the fragment coupling of 3a with the piperi- due to the poor downstream purging of subsequently
dine 2a required activation of the primary alcohol. Initial formed derivatives. Upon identifying the triethyl(al-
kyl)ammonium salt as the preferred coupling partner, ad-
ditional parameters (solvent, temperature, and
stoichiometry) were studied to determine their impact on
CO2H
steps
N
O
NH
O
the level of the Friedel–Crafts impurity 7. The key process
parameter for control of the level of this impurity was
found to be solvent, wherein N-methyl-2-pyrrolidinone
(NMP) produced the lowest level of the impurity and was
therefore selected for further development.8
O
Ph
(R)-pyroglutamic
acid
conjugate addition,
reduction
Bn
NR1
OH
Bn
H
N
The exothermic nature of both the mesylation and subse-
quent quaternization in NMP required these stages of the
process to be performed at low temperature (–20 to 0 °C),
which also helped control levels of impurity 7. However,
once 3b was formed, coupling with 2a required 8–10 h at
55 °C for the reaction to reach completion. Efforts to ac-
celerate the coupling by increasing the temperature result-
ed in increased levels of 7. We reasoned that the use of a
ring expansion
N
OH
N
R2
Bn
4a: R1 = H, R2 = Bn
2a: R1 = Boc, R2 = H
Boc2O, 4-DMAP
then H2, Pd(OH)2
Scheme 3 Summary of the original synthetic route to fragment 2a
HO
Boc
1
N
HN
N
OMe
R
HN
N
OMe
N
fragment coupling
Bn
N
N
Bn
Boc
N
N
N
OH
5
3a, R = OH
N
H
3b, R = NEt3
2a
2
alcohol activation
Boc removal
HO
HO
H
3
H2N
N
HN
OMe
N
HN
OMe
N
debenzylation
Bn
N
N
N
N
N
N
6
1: BMS-690514
Scheme 4 Final steps to target structure 1
Synlett 2013, 24, 305–312
© Georg Thieme Verlag Stuttgart · New York

CLUSTER
Synthesis of a Selective ErbB VEGF Receptor Inhibitor
307
HO
Boc
N
HN
OMe
R
HN
N
OMe
R
HN
N
OMe
N
Bn
N
N
N
N
N
N
Friedel–Crafts
alkylation
N
3
piperidine
coupling
H
N
H
N
HN
OMe
N
N
N
N
N
N
N
OMe
OMe
N
N
7
R = OH or Et3N
Scheme 5 Origin of the Friedel–Crafts impurity 7
stronger base might accelerate the reaction. Indeed, the
use of 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU; 1–2
equiv) resulted in faster reactions (3 h at 55 °C); however,
the yield was reduced by almost 5% due to the formation
of a new impurity, 8, arising from a competitive displace-
ment process as illustrated in Scheme 6, therefore no ad-
ditional studies were conducted with stronger bases.9
OMs
HN
OMe
Et3N
MsCl (1.3 equiv)
Et3N (3.9 equiv)
NMP
N
3a
–20 to 0 °C
N
N
HO
MeO
Bn
Boc
Boc
N
N
OH
NMP
MeCN
55 °C
HN
N
OMe
N
Bn
N
N
N
N
N
H
N
N
OMe
Et3N
N
H
2a
3b
N
80%
N
5
N
N
OMe
Et3N
Scheme 7 Antepenultimate step: alkylation of piperidine 2a and pyr-
rolotriazine 3a
N
N
N
creased impurity levels.10 We also found that the presence
of trace levels of oxygen during the coupling reaction led
to highly colored reaction streams that resulted in an iso-
lated product with a pink color. Recrystallization did not
remove the colored impurities, which persisted in the
downstream chemistry. We subsequently observed that
the colored impurities were removed in the API (active
pharmaceutical ingredient) debenzylation step, which
used Pd(OH)2/C for hydrogenolysis, suggesting that use
of activated carbon could result in color removal. An op-
tional carbon treatment (Darco G60) was defined to re-
move color from the reaction stream during the formation
of 6 (see below) if necessary.11
8
Scheme 6 Origin of the pseudodimeric impurity 8
To ensure that our conversion and impurity specifications
were achieved, we settled on the following optimized stoi-
chiometry for this step: 3.9 equiv of triethylamine, 1.3
equiv of methanesulfonyl chloride and 1.1 equiv of 2a.
Combination of this stoichiometry and temperature range
allowed consistent control of 7 to levels below 0.4 LCAP
(liquid chromatography area percent). Addition of water
to the process stream after reaction completion allowed
direct precipitation of 5 and isolation by filtration. From a
processing perspective, it was found that including an
equivolume amount of acetonitrile as a cosolvent in the
coupling step improved the crystallization of 5, with min-
imal impact on the level of 7. The finalized process, as
shown in Scheme 7, was successfully scaled up to ca. 100
kg scale.
Additional effort was invested in making the crystalliza-
tion of 5 more robust. Although the original crystalliza-
tion of 5 (addition of water to the reaction stream at 25 °C)
was simple to perform, it was inconsistent with regards to
several key parameters. In particular, the time required for
the crystallization to initiate varied, and in some instances
the product oiled out of solution. Addition of 1 wt% of 5
seed at 40 °C prior to water addition, and slow cooling of
the slurry to 25 °C over 2 h, resulted in a more consistent
crystallization. In addition, the new procedure offered the
advantage of an additional 40% purging of the Friedel–
As the project advanced and additional material was re-
quired to support clinical trials, we reinvestigated the cou-
pling process with the aim of improving it further. We
observed that loss of acetonitrile and triethylamine during
the coupling could lead to the reaction stalling, and in-
© Georg Thieme Verlag Stuttgart · New York
Synlett 2013, 24, 305–312

308
B. Mudryk et al.
CLUSTER
Crafts pseudodimer impurity 7. Two batches (ca. 30 kg) the isolated solid, which was further reduced to acceptable
were executed in our pilot plant, with an average yield of levels in the API crystallization.
76%, to afford the isolated product with a purity of 99.4
LCAP and 100% ee, and an acceptable level of 7 (0.3
LCAP).
Three low-level impurities were observed in isolated 6.
Compound 9 (Figure 1) resulted from deprotection of the
Friedel–Crafts impurity 7, which was carried into the re-
The penultimate step (Scheme 8) consists of acid-mediat- action with input 5. Only modest reduction (ca. 40%) of
ed Boc deprotection of 5 to provide 6 as a salt. Carbon di- this impurity was observed during the isolation of 6, fur-
oxide and isobutylene are byproducts of this reaction. For ther confirming the necessity of controlling its formation
safety reasons, the acid was added slowly at 50 °C to con- in the preparation of 5. The tert-butyl ether impurity 10
trol the rate of CO2 off-gassing. Most of the isobutylene originated from reaction of the tert-butyl cation, formed
remains in solution and is trapped by water to form t- during Boc removal, with the product.13 Although 10 did
BuOH. The penultimate salt is then free based by slow ad- not purge appreciably, its level was low enough (typically
dition of 1 M NaOH, which affords crystalline 6.
less than 0.2 LCAP) to not pose quality issues down-
stream. The final impurity of note in isolated 6 was the
starting material 5.
HO
Boc
N
HO
H
HN
N
OMe
N
Bn
N
N
HN
N
OMe
N
Bn
N
N
5
N
1. MSA (3.5 equiv)
MeCN–H2O
55 °C
H
N
H2N
OMe
N
m-anisidine
(GTI, byproduct)
2. 1 M NaOH (4.5 equiv)
20–30 °C
N
N
88%
OMe
9
HO
H
O
H
N
HN
OMe
N
Bn
N
HN
N
OMe
N
Bn
N
N
N
N
N
6
10
HO
Scheme 8 Penultimate step: Boc deprotection of intermediate 5
Boc
N
HN
N
OMe
N
Bn
In the first process iteration, aqueous hydrochloric acid in
isopropanol was used for the deprotection. Although Boc
removal appeared simple from a chemistry perspective, it
was complicated by the fact that m-anisidine was liberated
as a byproduct of acidic hydrolysis of 5 and 6 (Scheme 8).
m-Anisidine has been identified as a genotoxic impurity
(GTI), therefore the level of this compound in the final
isolated API needed to be controlled to below 5 ppm.12
The observed level of m-anisidine in the reaction mixture
was 700–1600 ppm, with up to 250 ppm present in isolat-
ed 6. Based on the ability of the API crystallization to
purge m-anisidine, a specification of <250 ppm of m-an-
isidine in isolated 6 was set.
N
N
5
Figure 1 Key impurities for the penultimate step
Due to concerns with respect to m-anisidine formation,
the reaction was closely monitored and not allowed to
proceed past 98% conversion. As a result, some residual 5
(<2 LCAP) remained in the isolated product, which was
acceptable since 5 was readily purged in the API step. We
also monitored the process for acetamide (a potential car-
cinogen) due to the use of aqueous acid in acetonitrile at
elevated temperatures. Analysis of the reaction stream re-
vealed that acetamide was present at low levels (ca. 35
ppm), but was purged to undetectable levels in the isolated
solid. This process was run at ca. 50 kg scale (two batches)
to produce 6 in 94% yield with an average purity of 98.5
LCAP.
As an alternative to HCl, we found that methanesulfonic
acid in acetonitrile produced lower levels of m-anisidine
in-process (200–400 ppm vs. 700–1600 ppm with
HCl/IPA) and had the added advantage of slightly higher
isolated yields (95 vs. 92%). The modified process consis-
tently generated 6 with less than 100 ppm m-anisidine in
Synlett 2013, 24, 305–312
© Georg Thieme Verlag Stuttgart · New York

CLUSTER
Synthesis of a Selective ErbB VEGF Receptor Inhibitor
309
The final step (see Scheme 9) involved removal of the 50% during the crystallization and posed the greatest lia-
benzyl group by hydrogenolysis [20% Pd(OH)2/C, 30 psig bility. To ensure that acceptable levels of 11 were ob-
H2, isobutanol/toluene, sodium carbonate,14 50 °C]. Upon served in the API, based on our purging studies, we set a
completion of the reaction and filtration of the catalyst, specification of <100 ppm for formaldehyde in isobuta-
the crude stream was washed with water and solvent ex- nol. The N-isobutyl impurity 13, however, purged effi-
changed to toluene, which facilitated the direct crystalli- ciently in the crystallization and was not a concern. The
zation of 1. The key to success of the hydrogenolysis was N-ethyl impurity 12 was derived from reaction of the
the judicious choice of solvent. Factors considered includ- product with residual acetonitrile carried over from the
ed: (1) solubility of the input and product; (2) product loss previous step16 and purged to >75% during the crystalliza-
during the work-up; (3) degree of palladium leaching; (4) tion. To control the formation of 12, we ensured that the
reaction kinetics; (5) impurity profile, and (6) robustness level of acetonitrile present in 6 was less than 0.7%.17 Iso-
of the crystallization to ensure the correct API polymorph. lation of 1 consisted of a water wash to remove the sodium
Alcohol solvents or tetrahydrofuran (THF) were required carbonate, solvent swap to toluene by constant volume
to solubilize 6, however, these solvents led to palladium distillation (to less than 1% isobutanol), seeding (3 wt%,
leaching and unacceptable palladium levels in the final at 85–89 °C), and then slow cooling to 20 °C. This proto-
product. The addition of toluene as a cosolvent reduced col reduced the levels of all the impurities (11–13) gener-
the amount of palladium leaching and, importantly, led to ated in the hydrogenolysis to acceptable levels. The
the correct polymorph of the API during crystallization. A crystallization also effectively purged m-anisidine from as
mixture of isobutanol/toluene (1:4 ratio) offered the ad- high as 800 ppm to below 5 ppm.18 Whereas the crystalli-
vantage of near quantitative partitioning of the product zation offered minimal reduction in the level of the deben-
into the organic phase during the aqueous wash. In this zylated version of the Friedel–Crafts-derived impurity 9,
solvent system, Pearlman’s catalyst [20% Pd(OH)2/C] optimization of the coupling step (3a → 5) allowed it to
was found to be optimal, based on reaction rate and impu- be controlled to acceptable levels upstream. This process
rity profile.
yielded 1 (89% yield, potency of 99.3 wt%) on ca. 30 kg
scale as a white solid, and consistently produced the cor-
rect crystal form of the API.
During the hydrogenolysis, three impurities of concern
were generated (11–13; Scheme 9B) in addition to deriv-
atives of the three impurities present in compound 6 (9, Despite the success and robustness of our initial deprotec-
10, and 5) that also underwent debenzylation in this step.15 tion endgame sequence, opportunities for improvement
The N-alkyl impurities, 11 and 13, were formed by reduc- were identified in the following areas: (1) Boc deprotec-
tive amination of formaldehyde and isobutyraldehyde, re- tion (penultimate step) resulted in the liberation of the
spectively, which were contaminants present in genotoxic impurity m-anisidine, and (2) benzyl group hy-
isobutanol. Additionally, isobutyraldehyde could form by drogenolyis (API step) produced three potentially diffi-
dehydrogenation (oxidation) of isobutanol mediated by cult to purge N-alkyl impurities. Due to these liabilities it
the Pd catalyst. The N-methyl impurity 11 purged only became evident that an alternative endgame strategy
A
HO
HO
reaction conditions
H2, Pd(OH)2, i-BuOH–toluene (20:80)
0.2% aq Na2CO3, 50 °C, 30 psig
BnHN
H2N
HN
N
OMe
N
HN
N
OMe
N
isolation
N
N
- water wash to remove Na2CO3
- solvent swap into toluene, <1% i-BuOH
- seeding with controlled cooling
N
N
1: BMS-690514
6
89%
B
HO
HO
HO
H
H
N
H
N
N
HN
N
OMe
N
HN
OMe
HN
N
OMe
N
N
N
N
N
N
N
N
N
11: N-Me impurity
12: N-Et impurity
13: N-i-Bu impurity
<0.4
<0.6
<1.5
crude LCAP:
<0.25
<0.15
<0.15
isolated LCAP:
Scheme 9 (A) Hydrogenolysis of intermediate 6; (B) key impurities (11–13) derived from hydrogenolysis of 6
© Georg Thieme Verlag Stuttgart · New York
Synlett 2013, 24, 305–312

310
B. Mudryk et al.
CLUSTER
would need to be identified. In addition, whereas the pro- Upon initial evaluation of this alternative protection strat-
cess utilized to generate piperidine 2a performed well on egy, there were concerns that the perceived instability of
a multi-kilogram scale (see Scheme 3), it was laborious Schiff bases might render this approach impractical. A
and operationally intensive because each transformation survey of the literature provided a recent example that uti-
required multi-operation steps, decreased throughput and lized methyl isobutyl ketone (reaction solvent) to mask a
resulted in higher manufacturing costs (longer plant time). primary amine in situ as the Schiff base. This allowed se-
lective alkylation of a secondary amine present within the
All of the above considerations resulted in the develop-
molecule.20 Further precedence was garnered from the
ment of a second-generation, more efficient synthesis of
demonstration that (1R,2R)-2-aminocyclohexanol pro-
duces stable Schiff bases with benzaldehyde, in contrast
the target structure 1. We first sought to develop an alter-
native preparation of piperidine 4a. This study has been
to the (1R,2S)-2-aminocyclohexanol, which yields the ox-
recently reported19 and is summarized retrosynthetically
azolidine (Scheme 12).21 With this knowledge providing
in Scheme 10. This alternative chemistry resulted in the
a degree of confidence, the first task at hand was selection
development of an expedient and simple racemic synthe-
of the appropriate aldehyde component. We imposed
sis of 4a from pyridine and benzyl chloride followed by a
three selection criteria: (1) the aldehyde must be inexpen-
classical resolution.
sive, (2) the resultant Schiff base intermediates must be
isolable crystalline solids, and (3) these isolated solids
Bn
Bn
NH
NH
must possess sufficient stability to enable isolation and
prolonged storage. After screening a series of aldehydes,
we found that p-anisaldehyde satisfied all these require-
ments.
classical
resolution
OH
OH
OAc
CO2H
Ph
N
N
Bn
Bn
(+)-4a + (R)-O-acetylmandelic acid
(rac)-4a
OH
OH
PhCHO
PhCHO
process highlights
5 steps
NH2
N
Ph
Ph
regioselective
epoxide opening
2 isolations
27% overall yield
99.8% ee
(1R,2R)-2-
aminocyclohexanol
only product
O
OH
O
steps
+
Ph
Cl
N
H
N
NH2
N
Bn
(1R,2S)-2-
only product
aminocyclohexanol
Scheme 10 Second generation synthesis of 4a
Scheme 12 Precedence for stable Schiff base vs. oxazolidine forma-
tion
Secondly, it was quickly recognized that these end game
liabilities / impurities were a consequence of protecting
group choice, and therefore could conceivably be avoided
by simply adjusting the protecting group strategy without
changing the overall synthetic strategy. As an alternative,
a Schiff base strategy was envisaged (Scheme 11) to re-
place the initial Boc/Bn strategy (Scheme 7).
This alternative sequence was successfully demonstrated
on a 300 g scale. Both benzyl groups present in 4a were
removed under hydrogenolysis conditions (H2; 30 psig,
Pd/C). The resulting crude ethanolic stream of 14 was
then treated with p-anisaldehyde and the ethanol was ex-
1. MSA, aq MeCN
2. H2, Pd(OH)2, 30 psig
Boc/Bn
PG strategy
HO
HO
PG1
H2N
N
PG2
HN
N
OMe
HN
N
OMe
N
N
N
N
N
N
Schiff base
PG strategy
1: BMS-690514
1. aq IPA, oxalic acid
2. aq KOH
Scheme 11 Comparison of protecting group strategies
Synlett 2013, 24, 305–312
© Georg Thieme Verlag Stuttgart · New York

CLUSTER
Synthesis of a Selective ErbB VEGF Receptor Inhibitor
311
the oxalate salt (97% yield). It is important to point out
that a simple and mild acid hydrolysis has replaced two
problematic steps, Boc removal and hydrogenolysis. Fi-
nally, conversion of the API into the desired polymorph
required removal of oxalic acid (aq KOH) followed by
crystallization from toluene (91% from 15) as previously
described. This process provided high quality API
(Scheme 14) of the desired crystal form.22
NH
NH2
OH
OH
Pd(OH)2, H2
N
94 %
N
H
14
4a
OMe
A robust synthesis of 1, which is a potential therapeutic
agent for the treatment of lung and other cancers, has been
described. The original route was used to prepare >100 kg
of 1, but suffered shortcomings that would hinder larger
future deliveries. The use of a Schiff base for primary ni-
trogen protection during coupling of the two key frag-
ments (3+2), alleviated many of the issues encountered
with the previously used Boc/benzyl protecting groups
(m-anisidine liberation during Boc removal and N-alkyl
impurities generated during hydrogenolysis). This im-
proved end game strategy was successfully demonstrated
and produced 232 g of API of similar quality to the prod-
uct obtained through the previous route. In addition to the
technical advantages garnered by use of the Schiff base
strategy, the yield was also improved for this new process
(67 vs. 62%).
HO
OHC
N
NH
MeO
95 %
2b
Scheme 13 Preparation of Schiff base 2b
changed with toluene. During this process, the reaction
was azeotropically driven to completion by removal of
water, with concurrent crystallization of the desired prod-
uct from the reaction mixture. Schiff base intermediate 2b
was isolated in 95% yield over these two steps (Scheme
13).
Subsequent coupling was performed as previously de-
scribed (Scheme 14), however, in this case, NMP was
used as the sole solvent rather than as a cosolvent with
acetonitrile. Omission of acetonitrile allowed for a direct
crystallization of the product through the addition of water
upon reaction completion. This process resulted in a 76%
yield of the isolated Schiff base 15 as its monohydrate.
The level of the equivalent Schiff base pseudodimer was
similar to the level of 7 produced under the original cou-
pling conditions. Hydrolytic deprotection of the Schiff
base was conducted under mild conditions (4% aq IPA,
1.5 equiv oxalic acid) and resulted in the isolation of 1 as
Acknowledgment
We would like to thank Drs. Robert Waltermire and Jean Tom for
careful review of this manuscript. We also acknowledge the analy-
tical support provided by Drs. Su Pan, Vera Leshchinskaya, Lydia
Breckenridge, Charles Pathirana and Mr. Jonathan Karten. We
thank our colleagues in Chemical Development Operations for their
contributions to the development of these processes. We would also
like to thank Dr. Yeung Chan, Dr. Daniel Hsieh, Dr. Simon Leung,
Shawn Springfield, Agnes Yeboah, and Melissa Chau for their con-
tributions to this work.
HO
HO
MeO
⋅H2O
N
N
NH
N
HN
OMe
HO
HN
OMe
MeO
2b
N
MsCl, Et3N, NMP
–20 to 0 °C
N
N
N
N
55 °C
N
76%
15
3a
4% aq IPA
oxalic acid
97%
HO
HO
H2N
H2N
O
N
N
HN
N
OMe
HN
OMe
aq KOH
i-BuOH–toluene
HO
N
OH
N
N
N
N
94%
O
1
1 oxalate salt
Scheme 14 Alternative end game strategy
© Georg Thieme Verlag Stuttgart · New York
Synlett 2013, 24, 305–312

312
B. Mudryk et al.
CLUSTER
Cossy, J.; Dumas, C.; Michel, P.; Pardo, G. M. Tetrahedron
Lett. 1995, 36, 549. (c) See also: Cossy, J.; Dumas, C.;
Pardo, G. M. Synlett 1997, 905.
Supporting Information for this article is available online at
http://www.thieme-connect.com/ejournals/toc/synlett.SnoIufproig
m
iotSrat
ungIifoop
r
t
(7) Thottathil, J. K.; Moniot, J. L.; Mueller, R. H.; Wong, M. K.
Y.; Kissick, T. P. J. Org. Chem. 1986, 51, 3140.
References and Notes
(8) Use of NMP produced approximately 1 LCAP of the
Friedel–Crafts-derived impurity. As a comparison, CH2Cl2,
which was the worst performer, produced ca. 12 LCAP of
the impurity.
(9) The structure of 8 was proposed based on LCMS data.
Specifically, the C–N connectivity cannot be confirmed.
(10) Loss of acetonitrile and triethylamine, which form a low-
boiling azeotrope (b.p. 55 °C), was attributed to the nitrogen
sweep used to ensure efficient inertion and prevent
formation of colored impurities. This issue was resolved
during development runs by using a nitrogen blanket, and
was not observed in the pilot plant batches.
(11) Carbon treatment after formation of 5 resulted in ca. 5% Boc
deprotection even at r.t. and, as a result, the carbon treatment
should be implemented during processing of 6 if required.
(12) For a discussion on genotoxic impurities, see: Snodin, G. J.
Regul. Toxicol. Pharmacol. 2006, 45, 79.
(1) New address: Celgene Corp., Basking Ridge, NJ, USA.
(2) New address: Dupont, Newark, DE, USA.
(3) For Bristol-Myers Squibb patents covering 1, see: (a) Wei,
C.; Norris, D. J. Crystalline forms of (3R,4R)-4-amino-1-({4-
[(3-methoxyphenyl)amino]pyrrolo[2,1-f][1,2,4]triazin-5-
yl}methyl)piperidine-3-ol; US Patent 0191375, 2007.
(b) Fink, B. E.; Gavai, A. V.; Vite, G. D.; Chen, P.;
Mastalerz, H.; Norris, D. J.; Tokarshi, J. S.; Zhao, Y.; Han,
W.-C. Preparation of pyrrolo[2,1-f][1,2,4]triazine
derivatives as HER1, HER2, HER4 kinase inhibitors and
antiproliferative agents; WO Patent 066176, 2005. (c) Fink,
B. E.; Gavai, A. V.; Vite, G. D.; Chen, P.; Mastalerz, H.;
Norris, D. J.; Tokarski, J. S.; Zhao, Y.; Han, W.-C.
Pyrrolotriazine Compounds as Kinase Inhibitors; US
Patent 7,141,571 B2, 2006.
(4) For earlier discovery work in the HER and VEGFR program,
see: (a) Borzilleri, R. M.; Cai, Z.; Ellis, C.; Fargnoli, J.; Fura,
A.; Gerhardt, B.; Goyal, B.; Hunt, J. T.; Mortillo, S.; Qian,
L.; Tokarski, J.; Vyas, V.; Barri Wautlet, B.; Zhenga, X.;
Bhidea, R. S. Bioorg. Med. Chem. Lett. 2005, 15, 1429.
(b) Fink, B. E.; Vite, G. D.; Mastalerz, H.; Kadow, J. H.;
Kim, S.-H.; Leavitt, K. J.; Du, K.; Crews, D.; Mitt, T.;
Wong, T. W.; Hunt, J. T.; Vyas, D. M.; Tokarski, J. S.
Bioorg. Med. Chem. Lett. 2005, 17, 4774. (c) Mastalerz, H.;
Chang, M.; Chen, P.; Dextraze, P.; Fink, B. E.; Gavai, A.;
Goyal, B.; Han, W.-C.; Johnson, W.; Langley, D.; Lee, F.
Y.; Marathe, P.; Mathur, A.; Oppenheimer, S.; Ruediger, E.;
Tarrant, J.; Tokarski, J. S.; Vite, G. D.; Vyas, D. M.; Wong,
H.; Wong, T. W.; Zhang, H.; Zhang, G. Bioorg. Med. Chem.
Lett. 2007, 17, 2036. (d) Mastalerz, H.; Chang, M.; Gavai,
A.; Johnson, W.; Langley, D.; Lee, F. Y.; Marathe, P.;
Mathur, A.; Oppenheimer, S.; Tarrant, J.; Tokarski, J. S.;
Vite, G. D.; Vyas, D. M.; Wong, H.; Wong, T. W.; Zhang,
H.; Zhang, G. Bioorg. Med. Chem. Lett. 2007, 17, 2828.
(e) Mastalerz, H.; Chang, M.; Chen, P.; Fink, B. E.; Gavai,
A.; Han, W.-C.; Johnson, W.; Langley, D.; Lee, F. Y.;
Leavitt, K.; Marathe, P.; Norris, D.; Oppenheimer, S.;
Sleczka, B.; Tarrant, J.; Tokarski, J. S.; Vite, G. D.; Vyas, D.
M.; Wong, H.; Wong, T. W.; Zhang, H.; Zhang, G. Bioorg.
Med. Chem. Lett. 2007, 17, 4947. (f) Cai, Z.-W.; Wei, D.;
Borzilleri, R. M.; Qian, L.; Kamath, A.; Mortillo, S.;
Wautlet, B.; Henley, B. J.; Jeyaseelan, R.; Tokarski, J.; Hunt,
J. T.; Bhide, R. S.; Fargnoli, J.; Lombardo, L. J. Bioorg.
Med. Chem. Lett. 2008, 18, 1354. (g) Gavai, A. V.; Fink, B.
E.; Fairfax, D. J.; Martin, G. S.; Rossiter, L. M.; Holst, C. L.;
Kim, S.-H.; Leavitt, K. J.; Mastalerz, H.; Han, W.-C.; Norris,
D.; Goyal, B.; Swaminathan, S.; Patel, B.; Mathur, A.; Vyas,
D. M.; Tokarski, J. S.; Yu, C.; Oppenheimer, S.; Zhang, H.;
Marathe, P.; Fargnoli, J.; Lee, F. Y.; Wong, T. W.; Vite, G.
D. J. Med. Chem. 2009, 52, 6527. (h) Fink, B. E.; Norris, D.;
Mastalerz, H.; Chen, P.; Goyal, B.; Zhao, Y.; Kim, S.-H.;
Vite, G. D.; Lee, F. Y.; Zhang, H.; Oppenheimer, S.;
Tokarski, J. S.; Wong, T. W.; Gavai, A. V. Bioorg. Med.
Chem. Lett. 2011, 21, 781.
(13) The linkage of the tert-butyl group to oxygen was confirmed
by HMBC NMR experiments.
(14) The aqueous carbonate was used in the first iteration of the
process to neutralize hydrogen bromide, a byproduct derived
from des-bromination of the brominated penultimate
derivative at the pyrrole fragment; see ref. 4. Although the
carbonate was not needed, it was retained in the revised
process to accommodate regulatory requirements and keep
the key process parameters unchanged.
(15) In general, impurities (organic, residual solvents, or
inorganic) do not add any therapeutic value, may present
toxicity concerns and are therefore controlled to very low
levels in the active pharmaceutical ingredient.
(16) Nacario, R.; Kotakonda, S.; Fouchard, D. M. D.;
Tillekeratne, L. M. V.; Hudson, R. A. Org. Lett. 2005, 7,
471.
(17) From a procedural perspective, it was noted that this
impurity continued to form after the reaction reached its
endpoint. As a result, we utilized an internal Raman probe,
rather than HPLC after a set amount of time. This allowed us
to stop the reaction when it was complete, minimizing
additional conversion of 1 into 12.
(18) It should be noted that we set a more stringent requirement
of 250 ppm m-anisidine in isolated 6 (see above) to ensure
that safe levels of this impurity would be present in the drug
substance.
(19) (a) Young, I. S.; Ortiz, A.; Sawyer, J. R.; Conlon, D. A.;
Buono, F. G.; Leung, S. W.; Burt, J. L.; Sortore, E. Org.
Process Res. Dev. 2012, 16, 1558. (b) Ortiz, A.; Young, I. S.;
Sawyer, J. R.; Hsiao, Y.; Singh, A.; Sugiyama, M.; Corbett,
R. M.; Chau, M.; Shi, Z.; Conlon, D. A. Org. Biomol. Chem.
2012, 10, 5253. (c) First generation: 7 steps, 3 isolations, 26
operations, 22% overall yield. Second generation: 5 steps, 2
isolations, 10 operations, 29% overall yield resulting in 60%
reduction in cost. Number of steps refers to number of
intermediates that could be isolated.
(20) Laduron, F.; Tamborowski, V.; Moens, L.; Horváth, A.; De
Smaele, D.; Leurs, S. Org. Process Res. Dev. 2005, 9, 102.
(21) Gil-Av, E. J. Am. Chem. Soc. 1959, 81, 1602.
(22) First generation: three steps, 62% yield, three N-alkyl
impurities. Second generation: two steps, 67% yield, no N-
alkyl impurities. Number of steps refers to number of
intermediates that could be isolated.
(5) Zheng, B.; Conlon, D. A.; Corbett, R. M.; Chau, M.; Hsieh ,
D.-M.; Yeboah, A.; Hsieh, D.; Müslehiddinoğlu, J.;
Gallagher, W. P.; Simon, J.-N.; Burt, J. Org. Process Res.
Dev. 2012, 16, 1846.
(6) (a) Langlois, N.; Calvez, O. Synth. Commun. 1998, 28, 4471.
(b) For other examples of this type of ring expansion see:
Synlett 2013, 24, 305–312
© Georg Thieme Verlag Stuttgart · New York

Copyright of Synlett is the property of Georg Thieme Verlag Stuttgart and its content may not be copied or
emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.