Development of a scalable process for CI-1020, a novel endothelin antagonist

Article abstract of DOI:10.1021/op010200a

The process development of a route for preparing CI-1020 on pilot-plant scale is described in 55% overall yield. Hydrocyanation conditions are described which use acetone cyanohydrin catalyzed by tetramethylammonium hydroxide and which provide the desired

Full text of DOI:10.1021/op010200a

Organic Process Research & Development 2001, 5, 226−233
Development of a Scalable Process for CI-1020, A Novel Endothelin Antagonist1
James E. Ellis,* Edward M. Davis, Gary J. Dozeman, Edward A. Lenoir, Daniel T. Belmont, and Phillip L. Brower
Pfizer Global Research and DeVelopment, Holland Laboratories, Pfizer Inc., 188 Howard AVenue,
Holland, Michigan 49424, U.S.A.
Abstract:
the testing program. This article describes the development
of a scalable synthesis of CI-1020.
Original Synthesis. Our Discovery group using the
synthetic route shown in Scheme 1 developed a very useful
synthesis of CI-1020.
The process development of a route for preparing CI-1020 on
pilot-plant scale is described in 55% overall yield. Hydrocya-
nation conditions are described which use acetone cyanohydrin
catalyzed by tetramethylammonium hydroxide and which
provide the desired ketonitrile intermediate (4) in 85% yield
with excellent quality. The penultimate intermediate (7), a
hydroxybutenolide, is prepared in a two-step process using an
aldol condensation followed by acid-catalyzed ring closure to
give product in 86.8% yield. The active pharmaceutical ingredi-
ent (API) (8) is prepared by ring-opening of the hydroxy-
butenolide with sodium carbonate to provide the sodium salt.
The use of ReactIR to monitor the API reaction is described.
ReactIR was required to determine an endpoint for the reaction.
The use of chromatographic analysis to determine the endpoint
was not possible. The API and the penultimate hydroxybuteno-
lide are not separable by chromatographic methods.
This synthetic route was based on earlier work for
5
preparation of hydroxybutenolides. The first step of the
synthesis was the preparation of the chalcone. A standard
Claisen-Schmidt reaction was run in which p-methoxyac-
etophenone (1) was reacted with piperonal (2) in ethanol with
a catalytic amount of base to give the desired chalcone (3)
6
in 95% yield. The chalcone was hydrocyanated using the
Lapworth procedure of sodium cyanide with acetic acid to
7
give the desired ketonitrile (4) in 86% yield. The ketoester
(5) was prepared in 95% yield by alcoholysis of 4 using
p-toluenesulfonic acid monohydrate in refluxing dioxane/
8
methanol. The penultimate hydroxybutenolide intermediate
(7) was prepared in a two-step sequence. The first step was
an aldol reaction between 5 and 3,4,5-trimethoxybenzalde-
hyde (6) catalyzed with sodium methoxide to give a mixture
of double bond isomers of CI-1020 (8). The reaction mixture
was then treated with excess acetic acid at reflux providing
the ring-closed hydroxybutenolide structure (7) in 59% yield.
The last step of the synthesis was reaction of 7 with aqueous
sodium hydroxide solution followed by lyophilization to
provide the desired sodium salt CI-1020 (8) in 88% yield.
Development of a Revised Synthesis. A very efficient
synthesis for preparing CI-1020 on small scale had been
developed by our Discovery group. However, several
potential problems in the synthesis were identified as the
process was reviewed from a scale-up perspective. In solving
these problems a revised synthesis was developed, which is
shown in Scheme 2.
Synthesis of the Ketonitrile (4). As noted above, the
synthesis of the chalcone (3) was quite straightforward giving
the desired product in nearly quantitative yield. However,
the original Claisen-Schmidt reaction conditions gave a
product slurry that was quite thick and difficult to transfer
from the reaction flask. Because the solubility of 3 in ethanol
was very low, the reaction could be diluted 3- to 5-fold and
still obtain the chalcone in 98% yield on pilot scale.
Introduction
In 1985 initial reports were published concerning isolation
of peptidic compounds with very potent vasoconstrictor
activity from endothelial tissues. These compounds were
designated endothelins and were found to be peptides of 21
amino acids. Since then this class of compounds has been
identified in tissues throughout the body. Elevated levels of
endothelins have been found to be associated with several
disease states. They have possible effects on ischemic stroke,
congestive heart failure, pulmonary hypertension, and a
number of other cardiovascular conditions.
A series of novel hydroxybutenolide compounds with
activity as endothelin antagonists was discovered recently
in our Discovery endothelin research program. These
compounds are non-peptides and have selective activity at
nanomolar concentrations with endothelin A receptors.
Kilogram-scale quantities of CI-1020 (8) were needed for
2
3
,4
*
Author for correspondence. Telephone: 616 392-2375, ext. 2387. Fax: 616
3
92-8916. E-mail: James.Ellis3@Pfizer.com.
(
1) Based in part on presentation given at 8th Midwest Pharmaceutical Process
Chemistry Consortium Symposium, West Lafayette, Indiana, October 1999.
2) For a review: Ellis, R. L. N. Engl. J. Med. 1995, 333, 356.
3) Doherty, A. M.; Patt, W. C.; Edmunds, J. J.; Berryman, K. A.; Reisdorph,
B. R.; Plummer, M. S.; Shahripour, A.; Lee, C.; Cheng, X.; Walker, D. M.;
Haleen, S. J.; Keiser, J. A.; Flynn, M. A.; Welch, K. M.; Hallak, H.; Taylor,
D. G.; Reynolds, E. E. J. Med. Chem. 1995, 38, 1259.
(
(
(5) Allen, C. F. H.; Frame, G. F.Can. J. Res. 1932, 6, 605.
(6) Kohler, E. P.; Chadwell, H. M. Organic Syntheses, 2nd ed. ; Wiley: New
York, 1941; Collect. Vol. I, p 78.
(7) Ishii, H.; Chen, I.; Ishikawa, T. J. Chem. Soc., Perkin Trans. 1 1987, 671;
Davey, W.; Tivey, D. J. J. Chem. Soc. 1958, 1230; Lapworth, A.; Wechsler,
E. J. Chem. Soc. 1910, 97, 38.
(8) For a review: Sandler, S. R.; Karo, W. Organic Functional Group
Preparations, 2nd ed.; Academic Press: New York, 1986; Vol. 3, pp 314-
350.
(
4) Reynolds, E. E.; Kaiser, J. A.; Haleen, S. J.; Walker, D. M.; Olszewski, B.;
Shroeder, R. L.; Taylor, D. G.; Hwang, O.; Welch, K. M.; Flynn, M. A.;
Thompson, D. M.; Edmunds, J. J.; Berryman, K. A.; Lee, C.; Reisdorph,
B. R.; Cheng, X. M.; Patt, W. C.; Doherty, A. M. J. Pharmacol. Exp. Ther.
1
995, 273, 1410.
2
26
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development
10.1021/op010200a CCC: $20.00 © 2001 American Chemical Society and The Royal Society of Chemistry
Published on Web 04/19/2001

Scheme 1. Original synthesis of CI-1020
Scheme 2. Revised process of CI-1020
The initial preparation of the ketonitrile (4) used the
Lapworth procedure of potassium cyanide and acetic acid
to hydrocyanate the chalcone. The hydrocyanation reaction
was run in ethoxyethanol because it was an excellent solvent
for chalcones. The reaction proceeded to completion in
approximately 1 h at reflux using 2 mol of potassium cyanide
and 1.5 mol of acetic acid per mole of substrate.
Two major problems hindered the scale-up of this process.
The first significant problem with the Lapworth method from
a scale-up perspective was the generation of large amounts
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
227

of hydrogen cyanide from partial neutralization of the excess
potassium cyanide. This process would have required special
equipment to prevent release of hydrogen cyanide.
The second problem was that the reaction mixture became
highly colored as the reaction proceeded. Since the color
carried through the remainder of the synthesis to the bulk
pharmaceutical compound, it was necessary to purify the
ketonitrile by recrystallization and carbon treatment to
remove the colored impurities. Attempts to directly isolate
Since the slow reaction rate using sodium carbonate
catalyst was also related in part to the low solubility of the
base in the reaction solvent, several other bases were
evaluated, which might have better solubility in the reaction
system. Catalytic triethylamine did not give complete reac-
tion even after 40 h of reaction time. Aqueous potassium
hydroxide gave no improvement in rate and also gave an
oily product. Catalytic aqueous potassium cyanide did not
significantly improve the reaction rate. Aqueous potassium
carbonate did improve the reaction rate but still required 20-
24 h for complete reaction.
At this point, 25% tetrabutylammonium hydroxide in
water was tried as a soluble form of hydroxide. This base
could also function as a phase transfer catalyst to assist in
the delivery of the cyanide in the Michael reaction. The
reaction proceeded very well, reaching completion after about
4 h. However, the presence of water in the reaction system
appeared to adversely affect the reaction by reducing the
solubility of the chalcone in acetone. The next base evaluated
was 25% tetramethylammonium hydroxide in water, which
also gave complete reaction in about 4 h. Further studies
used this reagent, since it has nearly twice the molar
concentration compared to the tetrabutylammonium salt.
Thus, the reaction could be run with the same stoichiometry
and only half the amount of added water.
As the reaction conditions were optimized, it was found
that both the amount of cyanohydrin and base could be
reduced substantially. However, as the amounts of cyano-
hydrin and base were reduced, the reaction times increased.
A maximum desirable reaction time of 10 h was chosen,
which was obtained using 1.25 mol of cyanohydrin per mole
of substrate with 5 mol % of tetramethylammonium hydrox-
ide. The reduction in the amount of cyanohydrin in the
reaction was a very significant improvement from a develop-
ment perspective since it facilitated the waste treatment by
reducing the amount of excess cyanide to be destroyed. In
addition, the increase in concentration gave an increase in
throughput.
4
that was not colored were unsuccessful. Consequently, we
began to look for a suitable alternative method to overcome
these problems.
There are a number of potentially useful alternative
methods available in the literature for hydrocyanation of R,â-
enones. Some of the methods, such as using hydrogen
cyanide or trialkylaluminum cyanides, raise significant safety
issues on scale.9 The method which appeared to be the
most practical for scale-up was the base-catalyzed Nazarov
hydrocyanation using acetone cyanohydrin as the cyanide
source.11
,10
The process evaluation was begun using a sodium
carbonate-catalyzed hydrocyanation with acetone cyanohy-
drin. In the literature, this reaction is typically run in alcoholic
solvents. Therefore, several different solvents were evalu-
ated: methanol, ethanol, ethoxyethanol, and also acetone.
In all of the solvents, except acetone, the reaction mixture
darkened considerably at reflux. Since minimization of color
in the isolated product was desired, acetone was chosen for
further studies.
Initially, a reaction stoichiometry of 2 mol of acetone
cyanohydrin and 0.25 mol of a 10% aqueous solution of base
per mole of 3 was used, which Betts and Davey had
described as optimal.1 Under these conditions, the sodium
carbonate-catalyzed hydrocyanation with acetone as solvent
proceeded smoothly to the desired product with good purity
and excellent color. However, the reaction rate was slow with
the sodium carbonate-catalyzed method, requiring more than
1b
40 h to complete the reaction on scale. This problem was
partly due to the reaction in acetone starting out as a slurry,
since the chalcone and sodium carbonate have low solubility
in acetone. On pilot scale, the solid sodium carbonate
presented additional problems. It settled into the vortex under
the reactor agitator, making it less available for reaction and
acted to block the bottom valve of the reactor.
An additional benefit in using acetone as the reaction
solvent is that it simplified the waste treatment. The use of
acetone avoided mixed organic solvents in the waste stream
since acetone is released in the reaction of acetone cyano-
hydrin. More importantly, the acetone could be separated
from the water phase easily by atmospheric distillation, since
it does not azeotrope with water. This allowed the treatment
of the water phase with sodium hypochlorite to destroy the
The slow reaction rate was also a function of the electron-
donating characteristics of the substituents on the chalcone,
which are known to slow the reaction.1 The electron-
donating substituents decrease the electrophilicity of the
chalcone double bond, resulting in a decreased rate of
Michael addition by the cyanide anion.
2
13
residual cyanide. This distillation separation was used
successfully on pilot scale using a reactor connected to an
atmospheric scrubber containing sodium hypochlorite to trap
any HCN vapors. The acetone distillate contained less than
1
0 ppm of hydrogen cyanide. The residual cyanide in the
(
9) For a review: Nagata, W.; Yoshioka, M. In Organic Reactions; Dauben,
W. O., Ed.; John Wiley and Sons Inc.: New York, 1977; Vol. 25, p 255.
water was treated with sodium hypochlorite without any
problems. By avoiding alcoholic solvents, the possibility of
the aqueous layer containing residual alcohol was eliminated,
which could have been dangerous since alcohols can form
(
(
10) Liotta, C. L.; Dabdoub, A. M.; Zalkow, L. H. Tetrahedron Lett. 1977, 1117.
11) (a) Nazarov, I. N.; Zav’yalov, J. Gen. Chem. USSR (Engl. Transl.) 1954,
24, 475; Chem. Abstr. 1955, 49, 6139f. (b) Betts, B. E.; Davey, W. J. Chem.
Soc. 1958, 4193. (c) Acetone cyanohydrin is a commercially available raw
material. An estimated 500,000 metric tons were used in the U.S. in 1989:
Encyclopedia of Chemical Technology; Howe-Grant, M., Ed., Wiley: New
York, 1993; Vol. 7, p 829.
14
explosive alkyl hypochlorites.
(
12) Ellis, J. E.; Davis, E. M.; Brower, P. L. Org. Process Res. DeV. 1997, 1,
(13) Jenks, W. R. In Encyclopedia of Chemical Technology; Grayson, M., Ed.;
2
50.
Wiley: New York, 1979; Vol. 7, p 320.
228
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

The optimized hydrocyanation process resulted in a
process that ran quite consistently on scale-up. The process
was combined with the chalcone process. The hydrocyanation
conditions on pilot scale were 1.25 equiv of acetone cyano-
hydrin with 5 mol % of tetramethylammonium hydroxide.
The reaction was complete in 9 h at reflux and provided
excellent quality product in 85-95% yield for the two steps.
Synthesis of the Ketoester (5). The alcoholysis of the
ketonitrile (4) to the ketoester (5) using the Pinner reaction/
hydrolysis process ran quite well on small scale but had some
problems on scale-up. Under the initial reaction conditions,
the ketonitrile was refluxed with p-toluenesulfonic acid
monohydrate in a 1:1 mixture of dioxane and methanol for
Scheme 3. Hydroxybutenolide formation
16 h. The isolated crude product was reslurried with water
to remove the by-product ammonium p-toluenesulfonate that
had cocrystallized with the product, giving a 59% yield of a
blue/green solid. The color of this product carried through
to the active pharmaceutical ingredient (API).
Several changes were made to improve the reaction. Since
methanol was necessary for the Pinner reaction, the first
process change was to eliminate dioxane as one of the
reaction solvents, leaving only methanol. The use of dioxane
was undesirable on-scale since it is a suspected carcinogenic
agent. Since Pinner reactions are most commonly run using
hydrogen chloride gas, the use of hydrogen chloride was
of 9 and 10 into conjugation between the two carbonyl
groups, with rapid concomitant cyclization to the stable
hydroxybutenolide.
8
evaluated for the reaction. Hydrogen chloride was easier to
use than p-toluenesulfonic acid, and the by-product am-
monium chloride was more soluble in methanol and less
likely to coprecipitate with the product than ammonium
p-toluenesulfonate. The reaction was run at reflux for 16 h,
and then water was added to hydrolyze the acetimidate
intermediate. The changes worked quite well, giving better
isolated yields and purity. However, the isolated product still
was quite colored, varying from blue to green. We believe
this color may result from trace amounts of highly colored
quinone methide impurities arising from partial cleavage of
The original process used freshly prepared sodium meth-
oxide in methanol for the aldol reaction. The reaction mixture
was heated at reflux for 16 h and then acidified with acetic
acid. The acidified reaction mixture was heated at reflux for
an additional 10 h. The mixture was cooled to crystallize
the hydroxybutenolide in 70-75% yield.
In the initial optimization of this process, sodium hy-
droxide, potassium hydroxide, sodium ethoxide, and sodium
methoxide were evaluated as basic catalysts for the aldol
reaction. When the reaction was run in water with sodium
hydroxide, the main product observed was the carboxylic
acid resulting from hydrolysis of the ketoester. The use of
potassium hydroxide in ethanol gave a low yield of the
desired product in low purity. The best base/solvent com-
bination for the aldol reaction in terms of both yield and
purity was sodium ethoxide, freshly prepared with sodium
metal, in ethanol. Commercial sodium ethoxide gave a highly
colored product in poor yield. Since the use of sodium metal
was undesirable on scale-up, the use of commercial sodium
methoxide in alcohols was evaluated and found to be superior
to commercial sodium ethoxide. Either dry sodium methylate
or sodium methoxide in methanol were found to work equally
well in this reaction. Dry sodium methylate was chosen for
use on scale-up to minimize the amount of methanol in the
reaction.
the ether functionalities in the substrate or product under the
Pinner reaction conditions.15
At this point, the process was changed to use aqueous
hydrochloric acid in methanol to further simplify the process.
The ketonitrile (4) was refluxed in aqueous methanolic HCl
for 8-18 h and then cooled to crystallize a very lightly
colored product. The small amount of color now present did
not carry through to the API. The optimized process ran very
well on pilot scale, providing ketoester (5) in 91-95% yield
with 98% purity by HPLC.
Synthesis of the Hydroxybutenolide (7). The next step
in the synthetic route was preparation of the penultimate
hydroxybutenolide (7). This process has two distinct reaction
phases as shown in Scheme 3 below. The first reaction was
an aldol coupling of 3,4,5-trimethoxy benzaldehyde to 5 in
The solvents for the aldol reaction that were studied
included water, methanol, ethanol, and 2-propanol (IPA). As
noted above, water with sodium hydroxide caused hydrolysis
of the ketoester. The hydroxybutenolide (7) was more soluble
in methanol compared to higher alcohols, thus potentially
producing lower yields. The range of solubility of 7 in the
alcohols was methanol > ethanol > IPA with roughly 4 g
5
the presence of strong base. The second reaction was
neutralization of the mixture of isomeric aldol adducts with
acetic acid and isomerization of the isomeric double bonds
(14) Brethereck, L. Handbook of ReactiVe Chemical Hazards, 2nd ed.; Butter-
worths: London, 1979; p 805.
(15) Filar, L. J.; Winstein, S. Tetrahedron Lett. 1960, 25, 9.
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
229

of 7 per liter in methanol to roughly 2.4 g of 7 per liter in
IPA. On the basis of the solubility data, initially IPA was
thought to be the ideal solvent for this reaction. However,
this was not the case because the sodium salts of the isomers
Scheme 4. API formation
9
and 10 formed during the aldol reaction were so insoluble
that they formed an unstirrable reaction mixture. Addition
of acid did not make the mixture stirrable. Ethanol was
chosen as the reaction solvent because 7 had moderate
solubility in that solvent. The minimum amount of ethanol
for the reaction was that which would completely dissolve
the sodium methylate. The practical reaction concentration
was found to be about 1.25 M ketoester (5) in ethanol.
Initially, it was presumed the aldol reaction required reflux
temperatures of about 70 °C. However, during the optimiza-
tion study it was found that when the reaction was run at
lower temperatures it produced far fewer by-products, thus
providing higher yield and purity. At 30 °C the reaction was
complete after about 3 h, the same time it takes at reflux.
The in-process HPLC purity improved from about 80% to
almost 90% by area normalization.
The optimized process was successfully piloted, providing
an improved yield of hydroxybutenolide to 80-85% com-
pared to the 59% obtained with the original process. The
improved process also provided HPLC purities of 90-95%
by area normalization.
Synthesis of the API (8). The final step of the synthesis
of CI-1020 was preparation of the sodium salt, which was
water-soluble and could be used in IV formulation of the
API. The hydroxy lactone ring of 7 was opened by reaction
with a base to give the API (8) as shown in Scheme 4. The
main impurities of concern in this reaction were the geo-
metric isomers 9 and 10 shown in Scheme 4, which resulted
from movement of the double bond out of conjugation with
the ester carbonyl.
The order of addition was also critical to minimizing
formation of by-products. The sodium alkoxide must be
preformed or predissolved in the solvent. Reactions in which
we tried to form the alkoxide in the presence of the other
reagents gave little or no yield. The best reaction conditions
were to add the alkoxide solution to a vessel containing both
5
and 3,4,5-trimethoxybenzaldehyde (6). When base was
added to a solution of only ketoester without 3,4,5-tri-
methoxybenzaldehyde, a large impurity peak appeared in the
in-process assay, particularly at temperatures >40 °C.
The neutralization was accomplished with acetic acid. No
other acids were studied for the neutralization. However,
from experiments directed at converting CI-1020 (8) back
to hydroxybutenolide, strong acids such as hydrochloric acid
did not allow isomerization of 9 and 10 to hydroxybutenolide
to occur. The components of the reaction mixture precipitated
immediately upon adding the strong acids. In the case of
acetic acid, the acidified reaction mixture remained in
solution, allowing the isomerization to 7. No advantage was
found from cooling the reaction during addition of the acetic
acid. Temperatures below 65 °C did not produce isomer-
ization to any great extent, and thus the scaled-up runs were
isomerized at reflux. Studies on the amount of acetic acid
required indicated a practical minimum of about 50% excess
to do the isomerization and to neutralize the excess meth-
oxide. When less than 50% excess of acetic acid was used,
the isomerization either was very slow or did not go to
completion. On-scale, a 60% excess of acetic acid was used,
which gave an isomerization completion time of 10 to 11 h
based on the in-process assay.
The choice of base was the first parameter in the
development of this step that was studied. A base was needed
which was strong enough to form the sodium salt but weak
enough to minimize olefin isomerization. The bases tried
included sodium hydroxide, sodium methoxide, sodium
carbonate, and sodium bicarbonate. Sodium hydroxide had
been used in the Discovery preparation of 8. However, when
sodium hydroxide was used for larger-scale runs, significant
amounts of the olefin isomers were obtained in the API
prepared. The first larger-scale experiments of the API
process used sodium methoxide. Careful analysis determined
that this material was contaminated with as much as 40% of
the isomers. The major breakthrough on this reaction was
the use of sodium carbonate, as the base that had the balance
of properties needed. The sodium carbonate was used in
equimolar quantities because it was converted in the reaction
to sodium bicarbonate, which precipitated since it was nearly
insoluble in methanol. The sodium bicarbonate precipitate
was filtered off to give a clear solution of the API sodium
salt. In subsequent reactions, 1.2 equiv of sodium carbonate
was used to achieve a reasonable reaction rate with complete
conversion to product.
The choice of reaction solvent was determined primarily
by two major factors. First, a solvent was needed in which
the product was reasonably soluble because a pre-crystal-
lization filtration was necessary to remove the insoluble
carbonates and other particulates at the end of the reaction
to ensure the product was suitable for use in IV formulations.
Second, the product needed to be isolated by crystallization.
In the original work, lyophilization had been used to isolate
Upon completion of the isomerization, the reaction was
cooled to 0 °C over several hours, aged at that temperature,
and then filtered. The crude cake was washed with cold
ethanol followed by water. In the initial work on this step
large amounts of sodium acetate were found to coprecipitate
with the product. The sodium acetate was easily removed
using a water reslurry since 7 is very insoluble in water.
230
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

the product and was undesirable for large-scale work. The
crystallization from a range of solvents including water,
methanol, ethanol, isopropyl alcohol (IPA), methyl tert-butyl
ether (MTBE), ethyl acetate, and toluene was evaluated. Of
these solvents, only ethanol showed any promise for use in
a single-solvent system. Either the API was so soluble in a
solvent that it would not crystallize, as with water and
methanol, or it was very insoluble, making a pre-crystal-
lization filtration impossible. Several mixed-solvent systems
were evaluated in which the primary solvent, that the product
was very soluble in, could be removed using an azeotropic
distillation, leaving the product as a slurry in a solvent in
which it is essentially insoluble, for example, water/MTBE,
water/ethyl acetate, methanol/ toluene, and methanol/ethyl
acetate. In all of these systems, the product failed to
crystallize. Instead it usually formed a gum during the
distillation. Solvent-exchange of IPA for methanol was also
studied. This showed the most promise of any system. Finally
it was found that if the reaction was run in a minimum
amount of methanol, the product could be precipitated simply
by adding IPA, eliminating the need for a distillation.
The next critical part was determining the minimum
volume of methanol the reaction could be run in. Early in
our scale-up work it was discovered that using larger volumes
of methanol and then distilling off the excess solvent prior
to crystallization with IPA could be problematic. In several
runs the product crystallized during the methanol distillation.
The resulting product appeared to be a methanol solvate that
was very difficult to dry. The solvated material required
several days of drying using vacuum oven temperatures
greater than 110 °C to obtain product with less than 0.5%
methanol. Although a slight color is introduced with these
temperatures, no degradation was detected.
steadily during the reaction with the rate of formation of 10
approximately twice that of 9. In the case of 10, concentra-
tions up to 0.5% (area normalization) were acceptable in-
process since the concentration was reduced 5- to 10-fold
when the API is isolated. However, the concentration of 9
was of concern since it was usually present in the isolated
product in the same percentage as the crude reaction. This
problem was easily monitored by HPLC. If the reaction was
stopped prior to completion because of high levels of 9, the
main effect was decreased yield with the unreacted 7 filtered
off with the sodium bicarbonate by-product.
Since an HPLC method was not found that was able to
resolve penultimate intermediate from the final product, other
methods for determining the hydroxybutenolide content of
the API were evaluated. Methods using IR were evaluated
because of significant differences in the IRs of 7 and 8. A
quantitative IR method was developed. In this method an
analyte peak characteristic for hydroxybutenolide was identi-
fied which initially appeared to have few interferences from
other compounds in the reaction mixture. This method was
useful for isolated API but was not useful as an in-process
assay because isolation of the API from each sample was
required before running the assay. In addition, once pure
samples of the double-bond isomers, 9 and 10 (the other
major impurities in the API reaction), were isolated they were
found to have an IR band in the same region as the
hydroxybutenolide analyte peak. Thus, the utility of the
quantitative IR for assay of hydroxybutenolide content was
diminished.
However, the ReactIR was found to be an effective
method for monitoring reaction progress. In this method,
sampling at intervals during the reaction allowed monitoring
of the formation of 8 in the reaction mixture. The other main
The premature crystallization could be avoided by dis-
solving the product formed in a minimum volume of
methanol. A solution of 8 in methanol of about 0.55 molarity
was used for the pilot-scale lots. The product could then be
crystallized by the addition of 2.5 to 3 vol of IPA to the
methanol reaction solution.
Initially, the API reaction was run at reflux but gave quite
inconsistent results at that temperature. As the reaction was
studied more thoroughly, higher reaction temperatures were
found to favor formation of the impurities 9 and 10. The
API reaction must be run at temperatures less than 30 °C to
minimize these impurities.
The most significant problem in the development of the
API process was the determination of reaction completion.
This was very difficult to this point since no chromatographic
method capable of separating hydroxybutenolide from the
sodium salt was found. Hydroxybutenolide and the sodium
salt readily interconverted in solution in a pH-dependent
equilibrium. The two compounds gave a single peak in the
chromatograms for all HPLC systems studied to date.
Without a good method for completion of reaction, alterna-
tive methods for determining the reaction rate had to be
evaluated. However, HPLC was useful for monitoring the
levels of 9 and 10, which form during the sodium salt
formation reaction. The levels of the two impurities grew
components of the reaction mixture, 7, Na
2 3
CO , and NaH-
CO , have low solubility in methanol and consequently do
3
not contribute much to the absorbance spectra obtained. The
end point of the reaction was determined by the plateauing
of the absorbance of several peaks in the IR spectrograms.
-
1
The IR band at 1598 cm was the primary analytical band
monitored since it is associated with the carbonyl stretch of
the forming carboxylate. At least one other peak was
monitored to avoid being fooled by early plateauing in a
single peak. This method was initially studied on lab-scale
experiments. The reactions were run directly in the ReactIR.
Several parameters were studied in these experiments:
powdered versus granular sodium carbonate, presence of
water, reaction temperature, and reaction time. The type of
sodium carbonate had a slight impact on the reaction. The
powdered carbonate, which was used in the laboratory, ran
slightly faster than the granular carbonate that was typically
used in the pilot plant. The powdered form was used in the
subsequent pilot-plant work. The presence of a small amount
of water was found to have no effect on the reaction. Initially,
it had been thought that water might help to solubilize the
sodium carbonate and thus facilitate the reaction, but this
was not the case.
As expected, the reaction proceeded faster at temperatures
up to 50 °C. However, the rate of formation of the isomeric
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
231

Figure 1. ReactIR study.
impurities was also found to be much higher. Consequently,
the optimization experiments were run at 20-25 °C. The
laboratory experiments run at 20-25 °C were generally done
in 42-44 h. Using these data, a lot in the pilot plant was
run using a 44 h reaction time for the batch. Analysis of the
isolated product determined that it contained higher amount
of the isomers than had been seen previously. This suggested
that the reaction was complete well before the 44 h stop time
since the rate of isomerization increases as the amount of
hydroxybutenolide decreases. In the next pilot-plant run, in-
process samples of the reaction were taken at regular
intervals, and the samples were run on the ReactIR to gather
data to determine completion of reaction. In addition, the
samples were assayed by HPLC to determine the isomer
content. The collected data is shown in Figure 1. As the data
were being collected, a plateau developed for both analyte
peaks at about 28 h, suggesting that the reaction might be
done. However, after about 0.5 h the peaks started to increase
in size again briefly and then plateaued again for several
hours. It is not clear what caused the increase in peak size
at 28.5 h, but since the reaction is heterogeneous, the
remaining hydroxybutenolide may have gone into solution
and reacted at that point. The reaction was finally considered
to be complete at about 30 h, a significantly shorter reaction
time than was used in the previous run. The material obtained
was the best quality product produced. The shorter reaction
times on scale-up were attributed to much better mixing of
the heterogeneous reaction mixture in pilot-plant reactors
compared to the lab-scale reactors.
that step. The yields for all of the steps of the synthetic route
were improved to greater than 85% on-scale.
Several issues remain to be resolved on this synthesis of
CI-1020. The most important of these involve the API.
Further work would need to be done to develop an analysis
that would determine the hydroxybutenolide content in API.
This work is necessary only if an IV formulation remains
the formulation of choice. If an oral formulation were to be
developed, that formulation would probably use the hy-
droxybutenolide. This would obviate the need for an assay
to separate the hydroxybutenolide from the current API. More
work also needs to be done on control of the crystallization
of the API to avoid formation of the methanol solvate that
has been obtained in some small-scale lots.
Unfortunately, work on this project has been terminated.
Toxicology studies on CI-1020 demonstrated that it caused
arteriopathy in dogs and monkeys. No safe dose of CI-1020
was found which would have allowed clinical studies to
proceed.
Experimental Section
All of the reagents used in this study were purchased from
the Aldrich Chemical Co. Solvents were purchased from
commercial bulk sources. All of the reactions were run under
1
nitrogen atmosphere. The H spectra were taken on a Varian
200 MHz spectrometer; chemical shifts are expressed in ppm
from tetramethylsilane as an internal standard. The melting
points were determined on a Buchi melting point apparatus
and are uncorrected. HPLC for the in-process assay and final
product assay were performed on a Perkex 5 µ C-18 125 ×
The process optimization done to date provided a process
for CI-1020, which ran successfully on pilot scale. However,
4
5
.6 mm column with a mobile phase of 45% acetonitrile and
5% 0.1 M sodium acetate buffer adjusted to pH 6 with the
®
more work using ReactIR needs to be done to have a fully
detector at 254 nm. The analyses by HPLC are expressed
by % area normalization for individual components.
optimized manufacturing process for CI-1020.
2
-(Benzo[1,3]dioxo-5-yl)-4-(4-methoxyphenyl)-4-oxobu-
Summary
tyronitrile (4). 3-(Benzo[1,3]dioxo-5-yl)-1-(4-methoxyphe-
nyl)-prop-2-en-1-one (9.3 kg, 32.9 mol) was slurried into
32 L of acetone at room temperature. Acetone cyanohydrin
(3.5 kg, 41.1 mol) and a 25% aqueous solution of tetra-
methylammonium hydroxide (0.75 L, 2.08 mol) were added
to the slurry. The reaction was heated at reflux for 10 h.
The mixture was cooled, and 18 L of water was added to
Overall, the original route was significantly improved into
a scalable process that was run successfully on pilot-plant
scale. The synthesis used only common solvents for all
processing stepssethanol, methanol, acetone, and IPA. The
cyanide waste generated in hydrocyanation step was mini-
mized, contained, and easily destroyed in the workup for
232
•
Vol. 5, No. 3, 2001 / Organic Process Research & Development

crystallize the product. The resulting slurry was then cooled
to 0-10 °C for at least 3 h. The product was collected by
centrifugation and dried in vacuo at 40-50 °C. The desired
10 h at 70 °C, the reaction was allowed to cool to 0-10 °C
over 3 h and then held at that temperature range for 8 h.
The resulting slurry was collected by centrifugation and
washed with cold ethanol and then with water. The filter
cake was immediately reslurried at 20 °C in water (30 L)
for 0.5 h, filtered, and washed with water and then with cold
ethanol. The white solid cake was dried in vacuo at 50-55
product (8.6 kg, 85% theory) was obtained with a purity of
1
9
9% (area normalization) by HPLC. mp 93.9-94.9 °C; H
NMR (200 MHz, d
1H, dd, CHCH
ArCHCH ), 5.95(2H, s, OCH
.5(4H, m, ArH), 7.9(2H, d, ArH).
-(Benzo[1,3]dioxol-5-yl)-4-(4-methoxyphenyl)-4-oxo-
butyric Acid Methyl Ester (5). 2-(Benzo[1,3]dioxo-5-yl)-
-(4-methoxyphenyl)-4-oxobutyronitrile (8.5 kg, 27.5 mol)
6
-DMSO) 3.4(1H, dd, CHCH
HCO), 3.9(3H, s, OMe), 4.5(1H, t,
O), 7.3(1H, d, ArH), 7.35-
a
HCO), 3.6-
(
b
2
2
°
C for 18-20 h to give 3-(benzo[1,3]dioxol-5-yl)- 5-hy-
7
droxy-5-(4-methoxyphenyl)-4-[(3,4,5-trimethoxyphenyl)-
methyl]-2(5h)-furanone (11.2 kg, 86.8%) with a purity of
2
9
7.6% (area normalization) by HPLC. mp 182.4-183.5 °C;
4
1
H NMR (200 MHz, d
3
5
5
-pyridine) 3.6(6H, s, OCH
), 3.8(3H, s, OCH ), 4.1(2H, s, ArCH
.95(2H, s, OCH
3
),
and 37% hydrochloric acid (7.2 kg, 72 mol) were slurried
into methanol (43 L) at room temperature. The reaction
mixture was heated to reflux for 18 h and then cooled to 0
.7(3H, s, OCH
3
3
2
),
2
O), 6.3(2H, s, ArH), 6.95-7.25(5H, m,
ArH) 7.9(2H, d, ArH), 10.6(1H, broad, OH).
Z)-Sodium 2-(benzo[1,3]dioxol-5-yl)-4-(4-methoxyphe-
nyl)-4-oxo-3-(3,4,5-trimethoxyphenyl)-but-2-enoate (8). 7
7.0 kg, 13.8 mol) and sodium carbonate powder (1.7 kg,
6.0 mol) were slurried into methanol (25 L). This slurry
(
5 °C and held at that temperature range for 12 h. The
(
product was collected by centrifugation and washed with
methanol. The wet product was dried in vacuo at 50 °C to
give 2-(benzo[1,3]dioxol-5-yl)-4-(4-methoxyphenyl)-4-ox-
obutyric acid methyl ester (8.6 kg, 91.5%) as an off-white
(
1
was stirred for 31 h at 18-22 °C. The reaction mixture was
filtered to remove the sodium bicarbonate by-product and
the filter rinsed with methanol (10 L). IPA (120 L) was added
to crystallize the product at 18-22 °C. The slurry was cooled
at 0 ( 5 °C for 12 h. The slurry was collected by centri-
fugation and washed with IPA. The wet product cake was
dried in vacuo at 40-50 °C for 26 h to give 5.9 kg (83.6%)
of CI-1020 as a white solid with a purity of 99.92% (area
solid with a purity of 97.9% (area normalization) by HPLC.
1
H NMR (200 MHz, CDCl
.7(3H, s, CO CH ), 3.9(3H, s, OCH
HCO), 4.1(1H, m, ArCHCH ), 5.95(2H, s, OCH
.9(5H, d, ArH), 7.9(2H, d, ArH).
Preparation of 3-(Benzo[1,3]dioxol-5-yl)-5-hydroxy-5-
4-methoxyphenyl)-4-[(3,4,5-trimethoxyphenyl)methyl]-
(5h)-furanone (7). Sodium methylate (2.3 kg, 42.6 mol)
3
) 3.2(1H, dd, CHCH
), 3.9(1H, dd, CHCH
O), 6.7-
a
HCO),
3
2
3
3
b
-
2
2
6
(
2
1
normalization) by HPLC. mp 258.3-259.3 °C; H NMR (200
and ethanol SD3A (16 kg) were charged to a reactor, allowed
to warm to near reflux until the methylate had dissolved,
and then cooled to 35 °C. To a separate reactor was added
MHz, d
.56(6H, s, OCH
6.2(2H, s, ArH), 6.9-7.0(5H, m, ArH) 7.85(2H, d, ArH).
Anal. Calcd for C28 Na: C, 63.64; H, 4.77. Found: C,
5
-pyridine) 3.38(2H, s, ArCH
2
), 3.52(3H, s, OCH
3
),
3
3
), 3.76(3H, s, OCH
3
), 6.0(2H, s, OCH O),
2
2-(benzo[1,3]dioxol-5-yl)-4-(4-methoxyphenyl)-4-oxobutyr-
ic acid methyl ester (8.6 kg, 25.1 mol) and 3,4,5-trimethoxy
benzaldehyde (4.9 kg, 25 mol). The sodium methylate/
ethanol solution was then transferred to a second reactor.
The reaction was held at 27-32 °C for 3 h, first by cooling,
later by heating. After 3 h, glacial acetic acid (4.2 kg, 69.9
mol) was added and the reaction heated to 68-72 °C. After
25 9
H O
63.65; H, 4.60. Na assay Calcd: 4.35 Found: 4.15.
Received for review January 3, 2001.
OP010200A
Vol. 5, No. 3, 2001 / Organic Process Research & Development
•
233