Process development and scale-up of AG035029

Article abstract of DOI:10.1021/op0601621

A practical process for the synthesis of PPARγ agonist AG035029 was developed which involved six steps along the longest linear path. The process development utilized automated technology and computational chemistry extensively to accelerate the speed of

Full text of DOI:10.1021/op0601621

Organic Process Research & Development 2007, 11, 30−38
Process Development and Scale-up of AG035029
James Saenz, Mark Mitchell, Sami Bahmanyar, Nebojsa Stankovic, Michael Perry, Bridgette Craig-Woods, Billie Kline,
Shu Yu,* and Kim Albizati
Chemical Research & DeVelopment, Pfizer Global R&D-La Jolla 10578 Science Center DriVe,
San Diego, California 92121, U.S.A.
Abstract:
A practical process for the synthesis of PPARγ agonist
AG035029 was developed which involved six steps along the
longest linear path. The process development utilized automated
technology and computational chemistry extensively to acceler-
ate the speed of the project in the areas of catalyst screening,
Figure 1. AG035029.
reaction optimization, mechanistic studies, and polymorph
control.
tography, most of the transformations proceeded in accept-
able yield and several intermediates are crystalline when the
purity is high. Importantly, it was revealed early in the
development that the penultimate intermediate (10) is a
workable solid that could be purified by recrystallization,
which provided a good control point for the purity of the
AG035029. However, two major issues still remained. The
Jacobsen indazole synthesis2 involved intermediate formation
of a hazardous diazonium salt, and the synthesis is long and
inefficient.
Introduction
AG035029 is a potent PPARγ agonist developed at Pfizer
for potential treatment of type II diabetes. In the U.S. 17
million people have type 2 diabetes mellitus (T2D), and the
worldwide number is estimated to be over 100 million. The
cost related to T2D in the U.S. is estimated at over $120
billion annually. Approximately 60% of diabetic patients do
not reach the treatment goal of HbA1c < 7% using the
currently available therapeutics as mono- or polypharmacy.
Glitazones (TZDs) are commercially successful PPARγ
agonists; however, there is significant room for improvement
in their safety/efficacy profile.1 AG035029 (Figure 1)is a
non-TZD PPARγ agonist with weak cross activity for
PPARR. Based on animal models, it is expected that
AG035029 might differentiate itself from the currently
marketed TZDs with respect to potency and possibly lipid
control. In order to support toxicity studies and clinical
studies, the material demands of the project escalated
exponentially in a very short period of time such that 10 kg
of drug substance were needed to proceed to Phase 1 study.
In addition to classical process chemistry techniques, the use
of a variety of automated technologies coupled with input
from computational chemistry experiments allowed develop-
ment of a synthetic route to this deceptively simple com-
pound in a rather short period of time. We present the
combined results of these studies in this manuscript to
demonstrate the advantages of a multidisciplinary approach
to chemical process development.
An alternative indazole formation method reported in the
literature3 was evaluated for suitability. Unfortunately, due
to the difference in substitution patterns, the high efficiency
was not achieved on our substrate. Jacobson indazole
synthesis appeared to be superior in terms of efficiency. A
calorimetric study of the Jacobsen indazole synthesis revealed
that no significant exotherm or off-gassing occurred in the
process. This was in agreement with the proposed mecha-
nism2 of the reaction in which the diazonium salt does not
accumulate; it is consumed as it is produced. After careful
study and consideration, it was felt that the risks involved
in the Jacobsen indazole synthesis could be managed.
Two further issues were factored into our initial approach
to the project. Soon after the project was initiated, a vendor
was identified to provide compound 4 in large quantities (50
kg), although at substantial cost. Because of the low
efficiency of a linear route, the high price of 4 mandated
the introduction of this piece at a later stage in the synthesis.
Further, differential scanning calorimetry data suggested that
phenyl oxazole containing intermediates in the linear route
employed by Discovery tend to be high-energy compounds;
early introduction of the phenyl oxazole group should be
avoided for production safety considerations. These two key
factors necessitated the study of a more convergent approach
to AG035029.
Results and Discussion
Evaluation of the Medicinal Chemistry Synthesis. The
synthetic route employed by the medicinal chemistry team
(Scheme 1) was evaluated to gain insights and to identify
potential hurdles to scale-up. Although the route is linear
and all the intermediates and API were isolated by chroma-
Convergent Route to AG-035029. A more convergent
route was easy to envisage (as shown in Scheme 2) and was
quickly reduced to practice. The synthesis started from
compound 11, a readily commercially available material. The
* To whom correspondence should be addressed. E-mail: shu.yu@pfizer.com.
(1) Schoonjans, K.; Auwerx, J. The Lancet 2000, 355, 1008.
(2) Troendlin, F.; Werner, R.; Ruechardt, C. Chem. Ber. 1978, 111, 367-78.
(3) Caron, S.; Vazquez, E. Synthesis 1999, (4), 588-592.
30
•
Vol. 11, No. 1, 2007 / Organic Process Research & Development
10.1021/op0601621 CCC: $37.00 © 2007 American Chemical Society
Published on Web 12/20/2006

Scheme 1. Modified medicinal chemistry route to AG035029
Scheme 2. Synthesis of AG035029 via a convergent route
nitro group was reduced to give the aniline (12) by catalytic
hydrogenation using sulfur poisoned Pd/C as catalyst, which,
after Jacobsen indazole synthesis, afforded 13. Conjugate
addition of the indazole at N1 to ethyl acrylate followed by
debenzylation yielded phenol 15. The phenol was then treated
with the tosylate 5 to form the ether bond in 10. Hydrolysis
of 10 yielded AG035029 in an overall yield comparable to
that from the linear route, without any optimization work
having been performed to this point. Since this route is
convergent, it has an advantage over the linear route in terms
of efficiency and cost. It was decided to pursue optimization
of the convergent route.
During the pilot run, some weak links in the synthesis
were identified. For instance, the reduction of the nitro group
in 11 was not totally selective, with some debenzylation
being observed. Also, extensive double addition occurred
during the Michael reaction, as shown in Scheme 3 (13 to
16). Further, the byproduct 16 was very difficult to remove,
even with chromatography. To address these and other
process-related issues, each step in the convergent route was
Vol. 11, No. 1, 2007 / Organic Process Research & Development
•
31

Scheme 3. Alkylation of indazole 13
Figure 2. IR indicates involvement of an intermediate.
studied individually. Screening of different catalysts on a
High-Pressure ChemScan automated reactor system promptly
identified 3% Pt/C as the catalyst of choice. In the presence
of this catalyst, no debenzylation was observed while the
reduction of the nitro group was complete in 18 h. Jacobson
indazole synthesis was found to require no more than 2 equiv
of iso-amyl nitrite and afforded intermediate 13 consistently
in greater than 80% overall yield from 11. This reaction was
followed closely by HPLC to ensure each intermediate in
the process was completely formed before the next reagent
was added.
The bottleneck of the convergent route appeared to be
the introduction of the N1- side chain. Initially, compound
14 was prepared from 13 via Michael addition to ethyl
acrylate. Unfortunately, the desired product was always
obtained in an admixture with 16, apparently from double
Michael addition (Scheme 3). This byproduct could be
present at levels approaching 25% and was more pronounced
when the reaction was scaled-up. In addition, the byproduct
was extremely hard to remove through crystallization or
chromatography and could not be purged during down stream
chemistry. Use of lower temperatures (e.g., -15 °C) appeared
to reduce the level of 16 for small batch sizes (1 g), but
upon scale-up this effect was diminished. To overcome the
problem, an alkylation was envisaged in lieu of the Michael
reaction (Scheme 3). Alkylation of indazole 13 with ethyl
3-bromopropionate gratifyingly did not furnish the undesired
double Michael product 16; however a mixture of isomers
14 and 17 was now obtained.
Furthermore this was a capricious process; sometimes the
desired 1-alkylated product 14 was obtained exclusively,
whilst on other occasions up to 20% of the undesired
2-isomer 17 was obtained despite apparently identical
reaction conditions. The reaction was clearly not fully
understood and hence was out of control. Experience has
shown that such processes are prime candidates for study
with ReactIR, which immediately revealed the participation
32
•
Vol. 11, No. 1, 2007 / Organic Process Research & Development

involved its formation. The following mechanism was
proposed, proceeding through the intermediacy of a Vils-
meier-like species (at too low a concentration to observe
spectroscopically) which could react to form the observed
intermediate I (Figure 5). The simulated profile is compared
to the scaled IR data in Figure 2.
In a subsequent experiment the reagents were charged at
room temperature and the mixture was slowly heated to 90
°C. ReactIR revealed that the indazole started to get
consumed at approximately 30 °C and was completely
consumed by 40 °C, furnishing the intermediate species.
Decay of the intermediate to product was apparent once the
reactor temperature exceeded 60 °C (Figure 6).
Figure 3. Product derived through first-order decay of an
intermediate species.
of an intermediate species (absorbing at 1549 cm-1), ruling
out a simple SN2-like mechanism (Figure 2).
The IR data also illustrated that product was formed in-
line with the rate of decay of the intermediate and not the
rate of decay of the starting indazole. Furthermore, this was
a first-order process as deemed from the linearity of the ln-
(intensity) versus time plot (Figure 3).
Figure 5. Proposed mechanism.
The reaction was also studied using neat toluene as solvent
in place of 80:20 toluene/DMF, and under these conditions
ReactIR did not indicate involvement of an intermediate
species. Rather the reaction proceeded with slight induction
at 80 °C, which had completely disappeared by 90 °C
furnishing nth order kinetics, typical for an SN2 type process
(Figure 4). Most importantly the selectivity was always poor
in the absence of DMF. These results appear to suggest that
different mechanisms operate in pure toluene and 80:20
toluene/DMF.
Figure 6. Temperature ramp experiment.
This data suggested that the intermediate was responsible
for the selective alkylation and that DMF was intimately
The key to success for the process was clear; the reaction
mixture needed an isothermal hold at 40 °C to allow for
Figure 4. Reaction profile in absence of DMF.
Vol. 11, No. 1, 2007 / Organic Process Research & Development
•
33

Figure 7. SNi mechanism.
Figure 9. Isomer conversion.
Figure 8. Computational measurements support SNi.
hourly intervals. The isomerization progress is plotted in
Figure 9.
complete formation of the intermediate species, followed by
heating to 90 °C to allow for collapse of the intermediate to
product. The process was now highly reproducible in the
laboratory and was successfully up-scaled to >20 kg at the
Pfizer pilot facility. Based upon these kinetic measurements,
an SNi mechanism is proposed (Figure 7) and is supported
by computational calculations (see figures in ref 4),4 con-
sistent with a similar process proposed by Hamel.5
Computational measurements also allowed an alternate
mechanism proceeding through a five-membered transition
state to be ruled out (Figure 8).
Earlier in the study it was observed that even for cases
when product was obtained with poor selectivity, isomer-
ization to the thermodynamic 1-alkylated product 14 could
be achieved with prolonged heating. The isomerization was
studied in an 80:20 toluene/DMF mixture at three different
temperatures over a 16 h period, analyzed by HPLC at
The half-life for the interconversion was 12.5, 3.7, and
3.4 h at 65, 75, and 85 °C, respectively. Presumably
isomerization occurs via a dissociative mechanism to ethyl
acrylate; however a fast cage-like recombination to product
is envisaged since the double-Michael byproduct 16 was not
observed. In this respect the isomerization is completely
analogous to the proposed SNi mechanism in Figure 7. The
difference lies in the readily reversible formation of the
intermediate III, allowing the thermodynamic equilibration
to occur at much lower temperatures. In this latter case,
heating is required solely to overcome the energy barrier to
the rate-limiting (and entropically favored) Grobe-like frag-
mentation of 17. Once dissociated, rapid recombination to
furnish desired N1-alylated product (14) is inevitable.6 Under
the optimized conditions, the alkylation yielded the desired
14 cleanly. After the workup, the intermediate (14) was
34
•
Vol. 11, No. 1, 2007 / Organic Process Research & Development

Scheme 4. Improved convergent route
present in MTBE solution that could be used in the next
step without further purification.
into DCM, speck-free filtration, followed by solvent ex-
change into iso-propyl alcohol always gave highly crystalline
material. The API is freely soluble in IPA at reflux and
crystallizes out of the solution during cooling. This added
crystallization also upgraded the API to >99% purity, with
no qualified impurities of >0.3% observed, and no unquali-
fied impurities of >0.1% were present. The material that
was produced this way affords form A API as a freely
moving powder. As a bonus, the particle size of the API
had a distribution of D50 ) 10 micron and D90 ) 36 micron,
which served the formulation of Phase I drug product and
Phase II drug product well.
This process had been employed to produce three lots of
API, 1.5 kg of non-GMP, 1.5 kg of GMP, and 12.6 kg of
GMP, without incidents. In the third campaign, the GMP
starting materials (13 and 5) were manufactured outside
Pfizer using the process provided by Pfizer (vide supra). The
downstream chemistry was carried out in-house, which is
summarized in Scheme 4. In this campaign, an overall yield
of 66% from 13 was achieved. The material passed the spec
without incident.
Polymorphs. A second form (form B) was found during
the polymorph screening. Although form B has a lower
melting point (164 °C; form A has mp of 173 °C), solubility
studies indicated that the new form does have lower
solubility, indicating it is the more stable form.7a This
enantiotropic system was studied in considerable detail, and
a transition temperature of 95 °C was determined.7b Efforts
were made to convert form A to form B in a controlled
manner, our experiments indicated that spontaneous nucle-
ation always generates form A, and form B could be
produced only when induced. Studies of the meta-stable zone
for form B in DME/heptane using HEL Automate and
Lasentec FBRM established a cooling/seeding regiment that
consistently gave form B. The robustness of the process has
been demonstrated on a 500 g scale. Planning of conversion
on a 10 kg scale is underway and will be reported in due
course.
Formation of 10 via Tandem Telescoping. The next step
was the debenzylation of compound 14 to set the stage for
the ether formation. Screening of a variety of commercial
catalysts revealed that Degussa’s Pd(OH)2/C was the best
choice when MTBE is the solvent. It rapidly catalyzed the
debenzylation, yet the over reduction of the indazole ring
system was limited. Debenzylation by transfer hydrogenation
(Pd/C, HCOONH4, ethanol, rt) afforded the desired phenol
15 cleanly, with only a trace of over-reduction. The drawback
of transfer hydrogenation was the operational inconvenience:
a solvent exchange was required to make the ethanolic
solution and the residue salt in the ethanolic solution after
the reaction sublimized during the solvent exchange in the
next step, a manageable inconvenience yet extremely an-
noying. When the pros and cons of the two options were
weighed, it was felt the use of molecular hydrogen was a
better choice, especially when considering the small amount
of the over-reduction byproduct could be easily removed in
the next step. Thus, the MTBE solution of compound 15
was converted to an acetonitrile solution by solvent exchange,
which was allowed to react with tosylate 5 in the presence
of CssCO3 at 60 °C to give the penultimate intermediate (10).
The product was easily isolated by addition of water as an
antisolvent to precipitate it out of the solution. Normally the
product was >97% pure that could be used directly in the
next step. Reslurring the intermediate in i-PrOH at room
temperature could upgrade any lot to >97% in greater than
90% recovery without fail. The overall yield for the three
steps from intermediate 13 was >70%.
End Game. Hydrolysis of the penultimate intermediate
(10) in THF afforded the crude API in the anionic form.
Partitioning the crude reaction mixture in water and MTBE
allowed most of the impurities to go to the organic layer
and upgrade the API in the aqueous layer to >98% pure.
Unfortunately, acidification of the aqueous layer often times
afforded the API as a thick oil instead of a solid. On the
other hand, the sequence of extracting the API as a free acid
Vol. 11, No. 1, 2007 / Organic Process Research & Development
•
35

Conclusion
An efficient and safe process for the preparation of drug
scaled-up in the kilo lab and pilot plant on a multikilogram
scale without incident.
candidate AG035029 was developed which provided many
significant advantages over the original process used by
Discovery chemists. The process development employed new
technologies extensively, which made it possible for us to
develop a scalable process expeditiously. The process was
Experimental Section
General. All melting points are uncorrected. 1H and 13
C
NMR spectra were recorded on a BRUKER-SPECTROSPIN
300 UltraShield instrument operating at 300 and 75.5 MHz,
respectively, using CDCl3 or d6-DMSO as the solvent.
Electron ionization low-resolution spectra were determined
at an ionizing voltage of 70 eV. Electron spray low-resolution
spectra were determined at an ionizing voltage of 70 eV.
Column chromatography was carried out on Universal silica
gel (32-63 µm) using the indicated solvents as eluents.
HRMS data were determined on a Micromass Q-TOF-2 mass
spectrometer. All the starting materials and solvents were
used as received without further purification.
(4) Note: Three mechanistic possibilities were proposed to describe the forma-
tion of the key intermediate observed in the kinetic studies. Density
functional theory calculations (B3LYP/6-31G(d)) were carried out on simpler
model systems to explore these mechanisms. The first two mechanisms
proposed a rearrangement of the substituent from N2 to N1 of I via a five-
membered ring transition state as shown in Figure 10. The computed acti-
vation energy barrier (Eact) for this rearrangement was extremely high (∼68
kcal/mol). This can be attributed to the overwhelming strain involved in
the five-membered ring transition state to achieve the N2 to N1 rearrange-
ment. Transition structure geometry and activation energy barrier for the
proposed N2 to N1 rearrangement via five-membered ring transition state.
3-Benzyloxy-2-methyl-phenylamine (12). A 2.5 L Parr
flask was charged with 3% Pt/C (62% water, 3.30 g, 3 wt
%) and then charged with a solution of 11 (110.0 g, 0.453
mol) in ethanol (1.6 L). The reaction was pressurized with
hydrogen gas at 50 psi. After 30 min, the pressure dropped
to 5 psi and was refreshed to 50 psi. After 1 h (from the
commence of the reaction), the pressure dropped to 3 psi
and was refreshed to 50 psi. After 3 h (from the commence
of the reaction), the pressure dropped to 40 psi and was
refreshed to 50 psi. The reaction was monitored by HPLC
for the disappearance of 11 and the formation of 12. At 18
h, the reaction was complete. The reaction mixture was
filtered through a pad of Celite 521. The pad was rinsed
with EtOH (200 mL). The combined filtrates were concen-
trated leaving an amber oil (94.82 g, 98%). NOTE: Oil
crystallizes upon sitting. The material showed satisfactory
purity and was used directly in the next step without further
1
purification. Mp 42-44 °C. H NMR (300 MHz, DMSO):
δ 1.95 (s, 3H), 4.81 (s, 2H), 5.00 (s, 2H), 6.26 (d, J ) 8.1
Hz, 1H), 6.28 (d, J ) 8.1 Hz, 1H), 6.80 (dd, J ) 8.1, 8.1
Hz, 1H), 7.25-7.47 (m, 5H). 13C NMR (75.5 MHz,
DMSO): δ 9.3, 69.2, 100.5, 108.0, 108.8, 126.0, 127.2,
127.5, 128.4, 137.9, 147.6, 156.7. Anal. Calcd for C14H15-
NO: C, 78.84; H, 7.09; N, 6.57. Found: C, 78.81; H, 7.15;
N, 6.60.
4-Benzyloxy-1H-indazole (13). A 5 L three-neck flask,
equipped with a mechanical stirrer, temperature probe, and
condenser was charged with 3-benzyl-2-methyl-phenylamine
(80.0 g, 0.375 mol) and 3.2 L of toluene, followed by KOAc
(40.5 g, 0.412 mol) and Ac2O (141.6 mL, 1.50 mol). After
stirring for 1 h, the mixture became very thick. The reaction
was monitored by HPLC for the disappearance of starting
material; after 2.5 h, all the starting material was consumed.
To the thick suspension was charged isoamylnitrite (100.8
The third mechanism proposed the disproportionation of II (see Figure 7)
via a Grobe type fragmentation as shown followed by a fast cage
recombination to form the intermediate. The computed activation energy
barrier (Eact) for this fragmentation is 3.9 kcal/mol in ꢀ ) 37.2, the dielectric
constant for DMF (3.4 kcal/mol in the gas phase, ꢀ ) 1). This implies that
this proposed mechanism is indeed feasible and consistent with the observed
experimental data. Transition structure geometry and activation energy
barrier for the proposed Grobe fragmentation in the rate-limiting step of
the reaction.
(5) Hamel, P.; Girard, M. J. Org. Chem. 2000, 65, 3123-3125.
(6) (a) Katritzky, A. R.; Perumal, S.; Fan, W. Q. J. Chem. Soc., Perkin Trans.
2
1990, (12), 2059-62. (b) Katritzky, A. R.; Yannokopoulou, K.;
Kuzmierkiewicz, W.; Aurrecoechea, J. M.; Palenik, G. J.; Koziol, A. E.;
Szczesniak, M. J. Chem. Soc., Perkin Trans. 1 1987, (12), 2673-9. (c)
Katritzky, A. R.; Kuzmierkiewicz, W.; Perumal, S. HelV. Chim. Acta. 1991,
74, 1936-40. (d) Yamazaki, T.; Baum, G.; Shechter, H. Tetrahedrom Lett.
1974, 15, 4421-4424.
(7) (a) Karpinski, P. H. Chem. Eng. Technol. 2006, 29, 233-238. (b) Kline, B.
J.; Saenz, J.; Stankovic, N.; Mitchell, M. B. Org. Process Res. DeV. 2006,
10 (2), 203-211.
36
•
Vol. 11, No. 1, 2007 / Organic Process Research & Development

mL, 0.750 mol). The resulting mixture was heated to 80 °C
and stirred for 18 h, at which time HPLC of an aliquot
indicated the reaction was complete. Heating was stopped
allowing the reaction mixture to cool to 40 °C, and then 4.15
M NaOH (904 mL, 3.75 mol) was charged followed by TBA-
OH (1 M in MeOH, 37.5 mL, 37.5 mmol). The resulting
mixture was stirred slowly (150 rpm) for 4 h, at which time
HPLC of an aliquot suggested the hydrolysis was complete.
The mixture was allowed to cool to room temperature, and
then the layers were separated. The organic layer was washed
with water (3 × 1600 mL). The resulting organic layer was
concentrated to a wet solid. The residue was dissolved into
DCM (800 mL) and then charged with heptane (800 mL).
The resulting mixture was concentrated to half the volume
(≈ 800 mL) at 40 °C under a vacuum (450 mbar). The
resulting mixture was allowed to cool to 20 °C and then
was filtered. The filtered solids were washed with heptane
(400 mL) and dried to afford 67.44 g (80% yield) of the
title compound as a light yellow solid. Mp 118-123 °C. 1H
NMR (300 MHz, DMSO): δ 5.27 (s, 2H), 6.62 (d, J ) 7.5
Hz, 1H), 7.10 (d, J ) 8.7 Hz, 1H), 7.23 (dd, J ) 8.1, 8.1
Hz, 1H), 7.29-7.46 (m, 3H), 7.48-7.55 (m, 2H), 8.06 (s,
1H), 13.05 (s, 1H). 13C NMR (75.5 MHz, DMSO) δ 69.2,
100.8, 102.9, 114.8, 127.1, 127.5, 127.8, 128.4, 130.9, 137.0,
141.7, 151.9. Anal. Calcd for C14H12N2O: C, 74.98; H, 5.39;
N, 12.49. Found: C, 75.02; H, 5.30; N, 12.41.
(8.5 g, 10 wt %%) under argon and then charged with an
MTBE (1500 mL) solution of 14. The reaction was pres-
surized with hydrogen gas to 47 psi. The pressure was
refreshed after 10 min to 47 psi. The reaction was shaken
for 12 h total. The reaction was monitored by HPLC for the
disappearance of 14 and the formation of 15. After agitation
overnight, all of 14 was consumed. The reaction mixture was
filtered through a pad of Celite 521. The pad was rinsed
with MTBE (170 mL). The combined filtrates were carried
on to the next step. An analytical sample was prepared by
concentrating the solution to afford an oil, which solidified
on standing. Slurring the solid in MTBE/heptane gave an
off-white solid after filtration and drying. Mp 103-105 °C.
1H NMR (300 MHz, DMSO): δ 1.08 (t, J ) 7.0 Hz, 3H),
2.88 (t, J ) 6.6 Hz, 2H), 3.99 (q, J ) 7.1 Hz, 2H), 4.55 (t,
J ) 6.4 Hz, 2H), 6.42 (d, J ) 7.5 Hz, 1H), 7.02 (d, J ) 8.3
Hz, 1H), 7.16 (dd, J ) 8.3, 7.3 Hz, 1H), 8.03 (s, 1H), 10.07
(s, 1H). 13C NMR (75.5 MHz, DMSO): δ 13.8, 33.9, 43.9,
60.0, 100.2, 103.6, 115.1, 127.4, 130.7, 141.4, 151.0, 170.7.
Anal. Calcd for C12H14N2O3: C, 61.53; H, 6.02; N, 11.96.
Found: C, 61.54; H, 5.94; N, 11.91.
Toluene-4-sulfonic Acid 2-(5-Methyl-2-phenyl-oxazol-
4-yl)-ethyl Ester (5). A 3 L four-neck flask equipped with
a mechanical stirrer, temperature probe, and nitrogen bubbler
was charged 1.0 L of anhydrous DCM, followed by 2-(5-
methyl-2-phenyl-oxazol-4-yl)-ethanol (4, 101.15 g, 0.50
mol). The flask was then immersed in a bucket of ice-water,
after stirring for half an hour, the internal temperature
dropped to 3 °C. The flask was then charged with TsO2O
(179.5 g, 0.55 mol) followed by (i-Pr)2NEt (131 mL, 0.75
mol); the temp surged from 3 °C to 6 °C in this process.
The reaction mixture was allowed to stir at ambient tem-
perature for 18 h, at which time HPLC of an aliquot indicated
the reaction was complete. The reaction mixture was poured
into a 4 L separation funnel, and 1 L of water was added.
After partitioning, the aqueous phase was washed once with
1 L of DCM. The combined organic layers were washed
three times with 1 L of water. The organic phase was
concentrated on rotary evaporator at 26 °C bath temperature.
The material crystallized at the end of the concentration. The
solid was further dried in a high vacuum for 8 h to afford
170.5 g of the title compound. Mp 93-95 °C. 1H NMR (300
MHz, DMSO): δ 2.11 (s, 3H), 2.25 (s, 3H), 2.76 (t, J ) 6.0
Hz, 2H), 4.24 (t, J ) 5.9 Hz, 2H), 7.25 (d, J ) 7.9 Hz, 2H),
7.43-7.53 (m, 3H), 7.62 (d, J ) 8.3 Hz, 2H), 7.74-7.83
(m, 2H). 13C NMR (75.5 MHz, DMSO): δ 9.6, 20.7, 25.0,
69.2, 125.4, 127.0, 127.4, 129.0, 129.8, 130.0, 131.3, 132.0,
144.7, 145.2, 158.3. Anal. Calcd for C19H19NO4S: C, 63.85;
H, 5.36; N, 3.92. Found: C, 63.93; H, 5.45; N, 3.93.
3-{4-[2-(5-Methyl-2-phenyl-oxazol-4-yl)-ethoxy]-inda-
zol-1-yl}-propionic Acid Ethyl Ester (10). A 3 L four-neck
flask, equipped with a mechanical stirrer, temperature probe,
and condenser was charged with 1.50 L of a homogeneous
solution of 15 (∼105 g, 0.449 mol based on 13) in MTBE
in portions. The flask was marked for the volumes of 0.60,
1.1, and 1.5 L, respectively. The total volume was reduced
to 0.60 L by distilling the MTBE out at atmospheric pressure,
under the protection of nitrogen. Heating was then stopped,
3-(4-Benzyloxy-indazol-1-yl)-propionic Acid Ethyl Es-
ter (14). A 1 L three-neck flask, equipped with a mechanical
stirrer, temperature probe, chart recorder, and condenser was
charged with 13 (95.0 g, 0.424 mol), DMF (475 mL), cesium
carbonate (275.5 g, 0.847 mol), and ethyl bromopropionate
(55.0 mL, 0.424 mol) in that order. The reaction was heated
to 40 °C while vigorous stirring was maintained. (NOTE:
The reaction appears to be autocatalytic and may continue
to heat past 40 °C. In this case, the reaction continued to
heat to 59 °C before it came back down to 40 °C in 1 h.)
The reaction was monitored by HPLC for the disappearance
of 13, and after 3 h all the starting material was consumed.
The reaction was then heated to 90 °C and maintained at
this temperature for 1.25 h. After cooling to 40 °C, the
reaction vessel was charged with MTBE (950 mL) followed
by water (950 mL) while stirring to dissolve all the salts.
The layers were then separated, and the organic layer was
set aside. The aqueous layer was then washed with MTBE
(2 × 500 mL). The organic layers were combined and
concentrated to about half the original volume, leaving 1 L
of solution. The production should satisfactory purity and
1
was used in the next step without further purification. H
NMR (300 MHz, DMSO): δ 1.06 (t, J ) 7.2 Hz, 3H), 2.89
(t, J ) 6.6 Hz, 2H), 3.97 (q, J ) 7.2 Hz, 2H), 4.58 (t, J )
6.6 Hz, 2H), 5.26 (s, 2H), 6.65 (d, J ) 7.2 Hz, 1H), 7.17-
7.44 (m, 5H), 7.46-7.54 (m, 2H), 8.05 (s, 1H). 13C NMR
(75.5 MHz, DMSO): δ 13.9, 34.0, 44.0, 60.1, 69.3, 101.1,
102.6, 115.4, 127.4, 127.6, 127.9, 128.5, 130.5, 136.9, 141.2,
151.9, 170.7. Anal. Calcd for C19H20N2O3: C, 70.35; H, 6.21;
N, 8.64. Found: C, 70.29; H, 6.29; N, 8.66.
3-(4-Hydroxy-indazol-1-yl)-propionic Acid Ethyl Ester
(15). A 2.5 L Parr flask was charged with 20% Pd(OH)2
Vol. 11, No. 1, 2007 / Organic Process Research & Development
•
37

and the contents in the flask were allowed to cool until
boiling ceased. The flask was charged with the balance 0.70
L of the MTBE solution of 15. The total volume was reduced
to 0.60 L by distilling the MTBE out at atmospheric pressure,
under the protection of nitrogen. Heating was once again
stopped, and the contents in the flask were allowed to cool
until boiling ceased. The flask was charged with 1.5 L of
acetonitrile. The total volume was reduced to 0.60 L by
distilling the solvent out at atmospheric pressure, under the
protection of nitrogen. Heating was stopped, and the contents
in the flask were allowed to cool to room temperature. A
sample was taken and checked with NMR. No sign of the
presence of MTBE was observed.
The flask was then charged with acetonitrile until a total
volume of 1.1 L was reached, followed by cesium carbonate
(219.3 g) and 5 in that order. The reaction mixture was heated
to 65 °C and stirred for a total of 18 h. The reaction was
monitored with HPLC (TFASH) for the disappearance of
15; after 18 h, the reaction was complete. The reaction was
cooled to room temperature and then charged with water
(1.25 L). The resulting mixture was stirred for 6 h at room
temperature, followed by storing at 4 °C for 18 h, at which
time a thick precipitate formed. The mixture was filtered.
The cake was washed with 300 mL of water and sucked
dry. The cake was transferred to a glass dish and dried at 40
°C/28 in. vacuum for 18 h. The cake weighed 162.0 g. Mp
96-98 °C. 1H NMR (300 MHz, DMSO): δ 1.06 (t, J ) 6.9
Hz, 3H), 2.88 (t, J ) 6.6 Hz, 2H), 3.02 (t, J ) 6.0 Hz, 2H),
3.97 (q, J ) 7.2 Hz, 2H), 4.37 (t, J ) 6.6 Hz, 2H), 4.57 (t,
J ) 6.6 Hz, 2H), 6.60 (d, J ) 7.2 Hz, 1H), 7.19 (d, J ) 8.4
Hz, 1H), 7.27 (dd, J ) 8.1, 7.5 Hz, 1H), 7.43-7.53 (m, 3H),
7.87-7.93 (m, 2H), 7.95 (s, 1H). 13C NMR (75.5 MHz,
DMSO): δ 9.9, 13.9, 25.6, 34.0, 44.0, 60.1, 66.5, 100.6,
102.5, 115.2, 125.4, 127.1, 127.4, 129.1, 130.1, 130.3, 132.7,
141.1, 145.3, 152.0, 158.4, 170.7. Anal. Calcd for C24H25N3O4:
C, 68.72; H, 6.01; N, 10.02. Found: C, 68.69; H, 5.97; N,
9.95.
A 5 L three-neck flask, equipped with a mechanical stirrer,
temperature probe, and condenser, was charged with the
crude AG035029, MTBE (1.6 L), and EtOH (200 mL), in
that order. The reaction mixture was heated to reflux and
stirred for a total of 18 h, cooled to room temperature, and
then stirred for 18 h; a thick precipitate formed at this point.
The mixture was filtered. The cake was washed with 200
mL of MTBE and sucked dry. The cake was transferred to
a glass dish and dried at 40 °C/ 28 in. vacuum for 18 h. The
cake weighed 72.0 g (83%). Mp 173 °C. 1H NMR (300 MHz,
DMSO): δ 2.39 (s, 3H), 2.81 (t, J ) 6.8 Hz, 2H), 3.01 (t,
J ) 6.4 Hz, 2H), 4.35 (t, J ) 6.4 Hz, 2H), 4.52 (t, J ) 6.4
Hz, 2H), 6.59 (d, J ) 7.2 Hz, 1H), 7.16-7.30 (m, 2H), 7.42-
7.52 (m, 3H), 7.86-7.93 (m, 2H), 7.94 (s, 1H). 13C NMR
(75.5 MHz, DMSO): δ 9.9, 25.6, 33.9, 44.1, 66.6, 100.6,
102.6, 115.2, 125.4, 127.1, 127.4, 129.1, 130.1, 130.2, 132.7,
141.1, 145.3, 152.0, 158.4, 172.2. Anal. Calcd for C22H21N3O4:
C, 67.51; H, 5.41; N, 10.74. Found: C, 67.38; H, 5.26; N,
10.65.
AG035029 Form B. A 3 L four-neck flask, equipped with
a mechanical stirrer and temperature probe, was charged with
form A AG035029 (178 g, 0.454 mol), followed by 1.8 L
of DME. The mixture was heated to reflux and maintained
for 1 h. The mixture became a homogeneous solution. The
mixture was cooled to 60 °C, and 1.45 g of form B
AG035029 seed were added immediately. The resulting
mixture was allowed to stir for 18 h at this temperature. XRD
of an aliquot indicated the material was form B. Heating
stopped, and the mixture was allowed to cool to room
temperature and stir for 24 h. The mixture was filtered,
washed with 180 mL of DME, and sucked dry. The cake
was dried in the oven at 40 °C/28 in. vacuum for 6 h to
1
afford 143 g of form B AG035029. Mp 164 °C. H NMR
(300 MHz, DMSO): δ 2.39 (s, 3H), 2.81 (t, J ) 6.8 Hz,
2H), 3.01 (t, J ) 6.4 Hz, 2H), 4.35 (t, J ) 6.4 Hz, 2H), 4.52
(t, J ) 6.4 Hz, 2H), 6.59 (d, J ) 7.2 Hz, 1H), 7.16-7.30
(m, 2H), 7.42-7.52 (m, 3H), 7.86-7.93 (m, 2H), 7.94 (s,
1H). 13C NMR (75.5 MHz, DMSO): δ 9.9, 25.6, 33.9, 44.1,
66.6, 100.6, 102.6, 115.2, 125.4, 127.1, 127.4, 129.1, 130.1,
130.2, 132.7, 141.1, 145.3, 152.0, 158.4, 172.2. Anal. Calcd
for C22H21N3O4: C, 67.51; H, 5.41; N, 10.74. Found: C,
67.38; H, 5.26; N, 10.65.
AG035029 Form A. A 5 L three-neck flask, equipped
with a mechanical stirrer and temperature probe, was charged
with compound 10 (93.0 g, 0.221 mol), THF, and a solution
of 37.2 g (0.664 mol) of KOH in 0.35 L of water in that
order. The progression of the reaction was monitored by
HPLC for the disappearance of 10; after 3 h, the reaction
was complete. The reaction was charged with water (1.0 L)
followed by MTBE (1.0 L). The resulting mixture was stirred
followed by separation of layers. The organic layer was
discarded. The aqueous layer was treated with 180.9 g (1.33
mol) of KHSO4, and 1.0 L of DCM was added; the mixture
was thoroughly agitated followed by separation of layers.
The aqueous layer was discarded. The organic layer was then
concentrated to dryness.
Acknowledgment
We thank Profs. E. J. Corey and Larry Overman as well
as Drs. Trevor Laird, Wei-Guo Su, Jay Srirangam, Ben Borer,
Sally Gut, and Keith McCarthy for helpful discussions.
Received for review August 3, 2006.
OP0601621
38
•
Vol. 11, No. 1, 2007 / Organic Process Research & Development