Safe and practical large-scale synthesis of 2-aminoquinoline-6-carboxylic acid benzyl ester

Article abstract of DOI:10.1021/op060044d

An efficient three-step sequence has been developed for the synthesis of 2-aminoquinoline-6-carboxylic acid benzyl ester starting from commercially available 6-quinolinecarboxylic acid. The process features a novel and exceptionally mild conversion of a q

Full text of DOI:10.1021/op060044d

Organic Process Research & Development 2006, 10, 534−538
Safe and Practical Large-Scale Synthesis of 2-Aminoquinoline-6-Carboxylic
Acid Benzyl Ester
Michel Couturier* and Tung Le
Chemical Research & DeVelopment, Pfizer Inc., Eastern Point Road, Groton, Connecticut 06340, U.S.A.
Abstract:
lowing amination protocol leading to the 2-aminoquinoline
5 employs chloroform as a solvent and also produces the
corresponding 2-hydroxyquinoline 6 as a major byproduct.5
Herein, we report our efforts to address these very issues to
develop a safe and scalable process of the title compound 5.
An efficient three-step sequence has been developed for the
synthesis of 2-aminoquinoline-6-carboxylic acid benzyl ester
starting from commercially available 6-quinolinecarboxylic acid.
The process features a novel and exceptionally mild conversion
of a quinoline N-oxide to a 2-aminoquinoline using a triethy-
lamine/ammonium chloride buffered system. The development
of this procedure is especially important since gaseous ammonia,
ammonium hydroxide, and solutions of ammonia in alcohols
all failed to deliver a safe and reliable process.
Results and Discussion
One of our first goals was to identify a derivative of
6-quinolinecarboxylic acid that would impart crystallinity and
avoid the separate preparation of diisopropylcarbodiimide/
tert-butyl alcohol adduct 3. Although the methyl and ethyl
esters have been previously synthesized, the corresponding
2-amino derivatives were poorly soluble and caused dif-
ficulties in the downstream chemistry. The working hypoth-
esis was that the lipophilic nature of the tert-butyl ester
increased the solubility, so we reasoned that a benzyl ester
would behave similarly. Such a derivative would also offer
the choice of either cleavage by hydrogenolysis or saponi-
fication in the later stage of the synthesis. This was especially
important since we needed to commit early to meet the
deadlines before the removal of the protecting group could
be assessed. Hence, the benzyl ester 7 was prepared using
CDI in ethyl acetate and was found to be a crystalline
intermediate, avoiding the chromatographic step (Scheme 2).
The scale-up went as planned except that the acylimidazole
intermediate crystallized out of solution on scale. Although
this event did not occur on laboratory scale, crystallization
was of no consequence to the reaction outcome. The yield
of the benzyl ester 7 was 6.45 kg (94%).
With the benzyl ester 7 in hand, the N-oxidation was then
investigated using the original discovery procedure to secure
material for safety assessment. Results from DSC testing on
the N-oxide 8 showed decomposition with an onset temper-
ature of 168 °C with an energy release of 596 J/g, classifying
it as a high thermal potential. ARC results indicated that the
material has a corrected onset of 137 °C with a time to
maximum rate of greater than a month at the maximum
process temperature. Based on these results, the N-oxide 8
was deemed safe to handle at room temperature. Since the
original process utilized trifluoroacetic anhydride, we inves-
tigated the use of a less corrosive agent and found that
phthalic anhydride,6 which comes as a free-flowing solid,
also gave yields greater than 90% when performed in
dichloromethane or THF. Although the conversion was not
Introduction
In a recent development program of a drug candidate, we
required multi-kilogram quantities of 2-aminoquinoline-6-
carboxylic acid as an ester derivative. A straightforward
approach, which would provide direct entry to the 2-amino-
quinoline core structure, lies in a vicarious amination of
commercially available 6-quinolinecarboxylic acid (1). Al-
though the corresponding Chichibabin reaction has been
reported in earlier literature, the exact regioselectivity of the
aminoquinoline carboxylic acid produced was not estab-
lished.1 Moreover, the reaction conditions using potassium
amide under a sealed vessel raised safety concerns and were
deemed not amenable to large-scale preparation. Alterna-
tively, conversion of quinoline N-oxides to the 2-amino-
quinolines could provide a means to effect the desired
transformation.2 Hence, the original discovery synthesis was
developed based on this approach, and initial assessment of
the procedures employed raised several process related issues.
More specifically, 6-quinolinecarboxylic acid (1) is first
derivatized as the oily tert-butyl ester (2) via the tert-butyl
alcohol adduct 3 of diisopropylcarbodiimide (Scheme 1).3
The latter is also an oily substance that is freshly prepared
and used crude in the process. Of particular concern is the
subsequent oxidation to the N-oxide 4, which has the
potential to be a highly energetic compound. The procedure
also utilizes corrosive trifluoroacetic anhydride.4 The fol-
* To whom correspondence should be addressed. E-mail: michel.a.couturier@
pfizer.com.
(1) Bergstrom, F. W. J. Org. Chem. 1938, 3, 233.
(2) (a) Storz, T.; Marti, R.; Meier, R.; Nury, P.; Roeder, M.; Zhang, K. Org.
Process Res. DeV. 2004, 8, 663. (b) Miura, Y.; Takaku, S.; Fujimura, Y.;
Hamana, M. Heterocycles 1992, 34, 1055. (c) Glennon, R. A.; Slusher, R.
M.; Lyon, R. A.; Titeler, M.; McKenney, J. D. J. Med. Chem. 1986, 29,
2375. (d) Sharma, K. S.; Kumari, S.; Singh, R. P. Synthesis 1981, 316. (e)
van Galen, P. J. M.; Nissen, P.; van Wijngaarden, I.; Ijzerman, A. P.;
Soudijn, W. J. Med. Chem. 1991, 34, 1202. (f) van den Haak, H. J. W.;
Bouw, J. P.; van der Plas, H. C. J. Heterocycl. Chem. 1983, 20, 447.
(3) Mathias, L. J. Synthesis 1979, 561.
(5) Miura, Y.; Takaku, S.; Fujimura, Y.; Hamana, M. Heterocycles 1992, 34,
1059.
(6) Kaczmarerek, L.; Balicki, R.; Nantka-Namirski, P. Chem. Ber. 1992, 125,
(4) Caron, S.; Do, N. M.; Sieser, J. E. Tetrahedron Lett. 2000, 41, 2299.
1965.
534
•
Vol. 10, No. 3, 2006 / Organic Process Research & Development
10.1021/op060044d CCC: $33.50 © 2006 American Chemical Society
Published on Web 04/05/2006

Scheme 1
an issue, the workup and isolation were more problematic
as it produced some emulsions. Since these were manageable
on a 100 g scale, we focused on the scale-up.
complete, we decided to investigate whether this solvent
swap could be accomplished earlier on to avoid the emulsion.
We found that such an operation had no deleterious impact
on the N-oxide 8 and that a clean phase separation was
achieved once the dichloromethane had been replaced with
ethyl acetate. As an added benefit, the material produced in
this process contained no ashes. The overall yield for this
run was 91% (Scheme 3).
Scheme 2
Scheme 3
Complete conversion of the benzyl ester 7 to the N-oxide
8 in the first scale-up was readily achieved in approximately
24 h, but the workup caused major difficulties. More
specifically, once the reaction mixture was quenched in
aqueous sodium thiosulfate/hydrochloric acid and the phases
were separated, subsequent treatment of the organics with
aqueous sodium bicarbonate led to a stable emulsion. In an
effort to achieve separation, the mixture was diluted with
more water and left to settle overnight. At this stage, the
phases were separated, but since some emulsion still
remained at the interface, the aqueous layer was treated with
brine and back-extracted twice with additional methylene
chloride at 35 °C, again allowing it to settle overnight. The
combined organic layers, in which some emulsions remained,
were concentrated and then replaced with ethyl acetate. The
mixture was further concentrated under partial vacuum with
concomitant azeotropic removal of water and then cooled
once the solution was anhydrous with a final volume of 3.5
L/kg (with respect to the starting material). Once 4.3 L/kg
of hexane was added as antisolvent, the solids were isolated
by filtration and dried to yield 4.1 kg of desired N-oxide 8
contaminated with 9.5% of ashes. The lower than expected
yield of 90% and the presence of inorganic material were
due to the emulsions encountered in the process.
It soon became apparent that conversion of the N-oxide
8 to the amine 9 was a major challenge. We first assessed
the original discovery procedure with the exception that
chloroform was replaced with dichloromethane. Hence, the
N-oxide 8 was treated with tosyl chloride, followed by the
addition of ammonium hydroxide. Upon reaction completion,
the organics were concentrated and the residual material was
triturated in ethyl acetate. Unfortunately, our product was
contaminated with the corresponding 2-hydroxyquinoline 10,
which required several triturations to ensure its removal. In
our hands the yield was only 10% (Scheme 4).
We next assessed the effect of various solvents on the
reaction outcome. Since we had originally substituted
chloroform by dichloromethane, we next performed a side-
by-side comparison and found that dichloromethane was a
superior solvent since it did not produce the major dimer
impurity 11 found in the chloroform reaction. While the use
of 1,2-dichloroethane produced similar results to the dichlo-
romethane, there were no significant advantages. Switching
to THF, the reaction cleanly produced yet another new
product characterized as tosylamide. In this solvent system,
the N-oxide was presumably not undergoing sulfonylation
with tosyl chloride. We hypothesized that the initial sulfo-
In a subsequent N-oxidation run, the typical emulsion was
observed during treatment of the dichloromethane layer with
aqueous sodium bicarbonate. Since the dichloromethane is
to be replaced by ethyl acetate after all the extractions are
Vol. 10, No. 3, 2006 / Organic Process Research & Development
•
535

Scheme 4
nylation of the N-oxide 8 by tosyl chloride was in equilibrium
with the desired tosylated N-oxide species. If this were the
case, then switching to a less nucleophillic counterion by
using tosic anhydride should increase the conversion, but
this reaction led to an intractable mixture. We therefore chose
to run all future experiments with tosyl chloride in dichlo-
romethane and focus our efforts on different ammonia
sources.
As mentioned above, the use of ammonium hydroxide as
source of ammonia led to a 1:1 mixture of 2-amino (9) and
2-hydroxyquinoline (10) in approximately 40% yield. Since
the presence of water was the culprit, we performed a
reaction using ammonia gas, which led almost exclusively
to dimer 11 (Scheme 5). In this particular case, the 2-ami-
centration of ammonia led to amidation of the benzylic ester.
Besides the exotherm issue, the rapid addition procedure did
prove reliable in consistently providing a 50-60% yield of
2-aminoquinoline 9.
Since the N-oxide 8 produced in the first run contained
inorganic material, a Celite filtration was implemented prior
to the activation with tosyl chloride. Once this was achieved,
the solution of the tosylated intermediate was cooled to -10
°C to which a precooled solution of ammonia in methanol
at -10 °C was rapidly added. Such a rapid addition was
required to minimize competing nucleophilic addition of the
solvent. In this context, the reaction temperature increased
to 37 °C in a few minutes with minimal reflux in the
condenser. At this stage, 2-aminoquinoline 9 crystallized out
of solution and, upon cooling to -5 °C, could be directly
isolated from the reaction mixture by simple filtration
(Scheme 6). As this material contained some inorganics, the
wet cake was reslurried in water, filtered, and washed with
methanol to yield 216 g (52%) after drying.
Scheme 5
Safety assessment of the foregoing procedure was per-
formed and was deemed unsafe to run on a larger than 25 L
scale due to the large amount of off gassing during the
process. After a series of failed attempts to solve this issue,
a superior alternative was found by the in situ generation of
ammonia from ammonium chloride and triethylamine. This
buffered system now allows a successful inverse addition
of the activated N-oxide to the ammonia source, whereas
such an addition mode into ammonia solutions led to poor
yields (vide supra). This set of reaction conditions is devoid
of any nucleophilic solvent like water and methanol and
provides similar yields to those of the previous method, and
more importantly, the heat of the reaction can now be
controlled by a slow rate of addition of the activated N-oxide
to the suspension of ammonium chloride in TEA/dichlo-
romethane.7 Pilot scale experiments on a 100 g scale provided
the desired compound in 65% yield. With the safety issues
associated with the amination procedure resolved, scale-up
was undertaken where a solution of N-oxide 8 in 4 volumes
of methylene chloride was treated with tosyl chloride, and
this solution was then slowly added to an ammonium chloride
suspension in dichloromethane and triethylamine. An addi-
tion rate of approximately 500 mL/min was required to
maintain a temperature range of 20 to 25 °C. The reaction
mixture was stirred an additional hour, then cooled to -10
°C, and further stirred for 1.5 h and filtered. To remove
triethylammonium chloride from the isolated solids, the
product wet-cake was repulped in water, refiltered, and
washed with methanol cooled at -10 °C. After drying under
vacuum at 35 °C, 1.85 kg (45% yield) of desired 2-amino-
noquinoline product 9 was sufficiently active to react with
the tosylated N-oxide in the ammonia-starved environment
during the gassing procedure.
We then switched to alternative commercially available
ammonia solutions (1 M in methanol and 2-propanol). As
compared to water, we were hoping that the increased steric
hindrance of methanol and 2-propanol would decrease their
competitive addition pathways. This indeed was the case with
methanol, whereas the 2-propanol solution led to an intrac-
table mixture. At this time, we found that the desired
2-aminoquinoline 9 was relatively insoluble in a more
concentrated reaction mixture in dichloromethane/methanol.
This was an especially important finding since the product
could be isolated directly from the reaction mixture by simple
filtration and no longer required multiple reslurries in ethyl
acetate. The success of this reaction, however, was dependent
on the rate of addition of the ammonia solution. More
specifically, the ammonia solution must be added at a fast
rate; otherwise the reaction mixture becomes starved of NH3,
and MeOH then acts as a nucleophile. In fact, the 2-meth-
oxyquinoline 12 was formed almost exclusively when the
addition was performed dropwise. The issue with respect to
a fast addition rate lied in the high amount of energy released
in the reaction. For example, when the ammonia solution in
methanol was added at once, an exotherm of 40 °C was
observed. In an effort to avoid a sudden release of energy,
we reasoned that an inverse addition of the activated species
to the ammonia solution would allow a slow addition rate
since the reaction mixture would not be starved of ammonia.2a
Unfortunately, this mode of addition only provided a 5%
yield of 2-aminoquinoline 9. Presumably, the higher con-
(7) It is noteworthy that the nature of the suspension is ammonium chloride
and not triethylammonium chloride prior to the addition of the activated
N-oxide solution.
536
•
Vol. 10, No. 3, 2006 / Organic Process Research & Development

Scheme 6
Scheme 7
by a brine wash (3 kg of NaCl in 12 L of water). The ethyl
acetate layer was concentrated by vacuum distillation to about
10 L, and hexane (20 L) was added at the rate 4 L/min until
the solids began to precipitate, at which point the rate was
reduced to 2 L/min. The resulting suspension was stirred
for 1 h at 20-25 °C and slowly cooled to 0-3 °C and held
for 4 h before filtering the solid product. The product was
dried in a vacuum oven at 35-40 °C for 24 h to yield 6.45
kg (94.2%) of the title compound; mp 46-48 °C. 1H NMR
(CDCl3, 400 MHz) δ 9.00 (m, 1H), 8.60 (s, 1H), 8.32 (d, J
) 9.0 Hz, 1H), 8.24 (d, J ) 8.0 Hz, 1H), 8.12 (d, J ) 9.0
Hz, 1H), 7.50-7.30 (m, 5H), 5.42 (s, 2H). 13C NMR (100
MHz, CDCl3) δ 166.16, 152.79, 150.33, 137.58, 136.04,
131.37, 130.06, 129.26, 128.91, 128.80, 128.67, 128.57,
128.50, 128.30, 127.62, 122.10, 67.36. HRMS calcd for
C17H14NO2 (M + H+), 264.1025; found, 264.0095.
quinoline 9 were isolated (Scheme 7). It is noteworthy that
the filtrate originating from the reaction mixture also
contained 15 to 20% of the desired product. Since the total
amount of aminoquinoline was enough to support develop-
ment activities, and due to the convenience of isolating the
product directly from the reaction mixture, the matter of
isolating additional 2-aminoquinoline 9 from the filtrate was
not pursued at the time.
1-Oxy-quinoline-6-carboxylic Acid Benzyl Ester (8). To
a mixture of urea hydrogen peroxide (1.67 kg, 17.8 mol),
phthalic anhydride (2.19 kg, 14.6 mol), and methylene
chloride (41.2 L) was added to quinoline-6-carboxylic acid
benzyl ester (7) (2.75 kg, 10.4 mol) under nitrogen. After
stirring for 12 h at 20-25 °C, a solution of sodium thiosulfate
pentahydrate (3.4 kg, 13.7 mol) in water (27.5 L) was added.
After stirring for 30 min, a solution of concentrated HCl (1.2
L) and water (26.3 L) was added. The layers were separated,
and methylene chloride (27.5 L) was added to the aqueous
layer for a second extraction. A solution of sodium bicarbon-
ate (1.4 kg) in water (44 L) was added to the combined
methylene chloride layers. The heterogeneous mixture was
stirred for 15 min, and the methylene chloride was removed
by atmospheric distillation. Upon complete removal of
methylene chloride, ethyl acetate (40 L) was added and the
mixture was stirred for 15 min. The layers were separated,
and the ethyl acetate solution was concentrated by vacuum
distillation to about 10 L. The solution was cooled to room
temperature until crystals form and hexane (12 L) was slowly
added over 20 min. The suspension was cooled to 5-10 °C
and held at that temperature for 4 h before filtering the solids.
The product was dried in a vacuum oven at 35-40 °C for
24 h to yield 2.64 kg of the title compound (90.6%); mp
99-102 °C. 1H NMR (CDCl3, 400 MHz) δ 8.78 (d, J ) 9.0
Hz, 1H), 8.63 (s, 1H), 8.58 (d, J ) 6.0 Hz, 1H), 8.34 (d, J
) 9.0 Hz, 1H), 7.82 (d, J ) 8.5 Hz, 1H), 7.49-7.34 (m,
5H), 5.43 (s, 2H). 13C NMR (100 MHz, CDCl3) δ 165.37,
143.55, 137.40, 135.66, 131.35, 130.68, 130.18, 128.97,
128.84, 128.80, 128.78, 128.72, 128.69, 126.85, 122.15,
120.69, 67.74. HRMS calcd for C17H14NO3 (M + H+),
280.0974; found, 280.0968.
Concluding Remarks
This work demonstrates a safe and practical large-scale
synthesis of 2-aminoquinoline 9, where all the intermediates
are crystalline solids. The process features an exceptionally
mild conversion of a quinoline N-oxide to a 2-aminoquinoline
using a triethylamine/ammonium chloride buffered system.
The development of this procedure was especially important
since gaseous ammonia, ammonium hydroxide, and solutions
of ammonia in alcohols all failed to deliver a safe and reliable
process. The synthesis described herein has successfully
provided multi-kilogram quantities of 9 for use in the
preparation of a drug candidate in a safe and timely manner.
As the manufacturing requirements increase, efforts are
underway to further improve the process for future cam-
paigns.
Experimental Section
Unless otherwise noted, all the operations were performed
in nitrogen purged vessels. All charges and transfers are
performed using an isolated vacuum whenever possible. The
1H and 13C NMR spectra were recorded on a Varian Innova
400 spectrometer. Melting points were obtained from a
Thomas-Hoover Uni-Melt capillary apparatus and are uncor-
rected.
Quinoline-6-carboxylic Acid Benzyl Ester (7). To a
solution of 6-quinolinecarboxylic acid (1) (4.5 kg, 26 mol)
in ethyl acetate (45 L) was added 1,1′-carbonyldiimidazole
(4.64 kg, 28.6 mol). After stirring at 20-25 °C for 4 h,
benzyl alcohol (2.95 kg, 27.3 mol) was added and the
reaction mixture was stirred for 12 h at 20-25 °C. The
reaction was quenched with a solution of concentrated HCl
(4.4 L) in water (22 L). The phases were separated, and the
aqueous layer was extracted with ethyl acetate (13.5 L). The
combined ethyl acetate layers were washed with a solution
of sodium bicarbonate (1.2 kg) in water (13.2 L) followed
2-Amino-quinoline-6-carboxylic Acid Benzyl Ester (9).
To a solution of 1-oxy-quinoline-6-carboxylic acid benzyl
ester (8) (4.1 kg, 14.7 mol) in methylene chloride (20 L)
Vol. 10, No. 3, 2006 / Organic Process Research & Development
•
537

1
was added p-toluenesulfonyl chloride (3.9 Kg, 21 mol) under
nitrogen, and the mixture was stirred for 45 min at 22-25
°C. This solution was added very slowly (400 mL/min) to a
suspension of ammonium chloride (2.4 kg, 44 mol) in
methylene chloride (12 L) and triethylamine (6.8 L, 48 mol)
while keeping the temperature at 25-30 °C. The reaction
mixture was stirred for an additional hour at 22-25 °C before
cooling to -5 °C for 1 h and filtered. The filtered solids
were rinsed with pre-cooled methanol (4.5 L) at -5 °C. This
solid material was then triturated in water (45 L) for 20 min
at 20-25 °C, filtered, and washed with precooled methanol
(9 L) at -5 °C. The product was dried in a vacuum oven at
35-40 °C for 24 h to yield 1.85 kg (45.1%) of the title
compound; mp 199-202 °C. H NMR (CDCl3, 400 MHz)
δ 8.39 (s, 1H), 8.18 (dd, J ) 2.0 Hz, 9.0 Hz, 1H), 7.93 (d,
J ) 9.0 Hz, 1H), 7.62 (d, J ) 9.0 Hz, 1H), 7.46 (d, J ) 7.0
Hz, 2H), 7.42-7.33 (m, 3H), 6.74 (d, J ) 9.0 Hz, 1H), 5.39
(s, 2H), 4.97 (br s, 2H). 13C NMR (100 MHz, CDCl3) δ
166.59, 158.52, 150.61, 139.36, 136.36, 130.98, 130.12,
128.84, 128.46, 126.15, 124.35, 122.84, 112.55, 66.95.
HRMS calcd for C17H15N2O2 (M + H+), 279.1134; found,
279.1120.
Received for review February 24, 2006.
OP060044D
538
•
Vol. 10, No. 3, 2006 / Organic Process Research & Development