Process development and scale-up of the potential thiazolidinedione antidiabetic candidate PNU-91325

Article abstract of DOI:10.1021/op025549s

An efficient six-step synthesis has been developed for the preparation of the thiazolidinedione analogue PNU-91325 (3) from the commercially available olefin 12. This process involves a novel epoxide ring opening with a deactivated phenol under phase-transfer conditions. Significant improvements were made in the oxidation of a secondary alcohol to the ketone and the 1,4-reduction of an enone from a previous process. Overall, this route allows for the preparation of PNU-91325 in 25% yield.

Full text of DOI:10.1021/op025549s

Organic Process Research & Development 2002, 6, 721−728
Process Development and Scale-Up of the Potential Thiazolidinedione
Antidiabetic Candidate PNU-91325
Donald E. Carpenter, Rick J. Imbordino, Mark T. Maloney, Jeffery A. Moeslein, Michael R. Reeder,* and Allen Scott
Early Process Research and DeVelopment, Pharmacia Corporation, Kalamazoo, Michigan 49001, U.S.A.
Abstract:
An efficient six-step synthesis has been developed for the
preparation of the thiazolidinedione analogue PNU-91325 (3)
from the commercially available olefin 12. This process involves
a novel epoxide ring opening with a deactivated phenol under
phase-transfer conditions. Significant improvements were made
in the oxidation of a secondary alcohol to the ketone and the
1,4-reduction of an enone from a previous process. Overall, this
route allows for the preparation of PNU-91325 in 25% yield.
Figure 2.
Introduction
Pioglitazone analogue 5-{4-[2-(5-ethylpyridin-2-yl)-2-
oxoethoxy]benzyl}-1,3-thiazolidine-2,4-dione (3) was an
Pharmacia wished to conduct studies that might serve to
evaluate the biological properties of 3; thus, bulk quantities
antidiabetic thiazolidinedione candidate that was being
were needed for this study. The former Upjohn Company
evaluated for possible clinical development for the treatment
has had experience with the synthesis of various pioglitazone
of noninsulin-dependent diabetes mellitus (NIDDM).1 Ketone
analogues, and thus we had ample synthetic methodology
3, although not identified as a metabolite of pioglitazone (7),
to draw from as well as a significant amount of alcohol 11
has been added to a list of putative metabolites on the basis
which serves as a convenient starting material.
of analogy to alcohol 1 and known ketone metabolite 4
(Figure 1).
Results and Discussion
Our synthetic strategy for the synthesis of 3 (Scheme 1)
began with dehydration of 11 with KOH to afford olefin
12.6 Epoxidation of 12 followed by subsequent ring opening
of 13 with 4-hydroxybenzaldehyde provided aldehyde 14.
Knoevenagel condensation with 2,4-thiazolidinedione af-
forded olefin 15 which was reduced with NaBH4 to provide
diastereomeric alcohols 2. Oxidation of the carbinol to the
(4) (a) Clark, D. A.; Goldstein, S. W.; Volkmann, R. A.; Eggler, J. F.; Holland,
G. F.; Hulin, B.; Stevenson, R. W.; Kreutler, D. K.; Gibbs, E. M.; Krupp,
M. N.; Merrigan, P.; Kelbaugh, P. L.; Andrews, E. G.; Tickner, D. L.;
Suleske, R. T.; Lamphere, C. H.; Rajeckas, F. J.; Kappeler, W. H.;
Figure 1.
McDermott, R. E.; Hutson, N. J.; Johnson, M. F. J. Med. Chem. 1991, 34,
319. (b) Hulin, B.; Clark, D. A.; Goldstein, S. W.; McDermott, R. E.;
Dambek, P. J.; Kappeler, W. H.; Lamphere, C. H.; Lewis, D. M.; Rizzi, J.
P. J. Med. Chem. 1992, 35, 1853.
The structure of 3 is similar to that of other thiazolidino-
nediones (Figure 2) that are or have been in clinical
development such as pioglitazone (7)2 (Takeda Chemical
(5) Cantello. B. C. C.; Cawthorne, M. A.; Cottam, G. P.; Duff, P. T.; Haigh,
Industries, Inc./Upjohn), troglitazone (8)3 (Sankyo and Parke-
D.; Hindley, R. M., Lister, C. A.; Smith, S. A.; Thurlby, P. L. J. Med.
Chem. 1994, 37, 3977.
(6) For our initial work, we contracted with a vendor to convert a large lot of
Davis), englitazone (9)4 (Pfizer), and BRL-49653 (10)5
(SmithKline Beecham).
alcohol 11 to olefin 12. Additional 12 was subsequently purchased from
Koei Chemical Company, Limited.
* Author for correspondence: E-mail: michael.r.reeder@pharmacia.com.
(1) Tanis, S. P.; Parker, T. T.; Colca, J. R.; Fisher, R. M.; Kletzein, R. F. J.
Med. Chem. 1996, 39, 5053 and references cited therein.
(7) Thurkauf, A.; Mattson, M. V.; Richardson, S.; Miradeghi, S.; Ornstein, P.
L.; Harrison, E. A., Jr.; Rice, I. C. C.; Jacobson, A. E. Monn, J. A. J. Med.
Chem. 1992, 8, 1323.
(2) Sohda, T.; Momose, Y.; Meguro, K.; Kawamasu, Y.; Sugiyama, Y.; Ikeda,
H. Arzeim-Forsch./Drug Res. 1990, 40, 37.
(3) Yoshioka, T.; Fujita, T.; Kanai, T.; Aizawa, Y.; Kurumada, T.; Hasegawa,
K.; Horikoshi, H. J. Med. Chem. 1989, 32, 421.
(8) The onset temperature of decomposition (56 °C) for the epoxide was
determined from ARC and RC1 experiments and is representative of the
neat oil. Further safety analysis for the PTC chemistry showed no
exothermic decomposition of the epoxide at 80 °C while in solution.
10.1021/op025549s CCC: $22.00 © 2002 American Chemical Society
Published on Web 08/16/2002
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
721

ketone afforded the target pioglitazone analogue 3.
by GLC) without any significant impurities (Scheme 2).
Optimization of this epoxidation of olefin 12 included a
temperature-controlled addition of a tert-butyl alcohol solu-
tion of acetic acid (1 equiv) and 12 to a stirring aqueous
slurry of NBS cooled to 0 °C. After 1 h, the resultant
bromohydrin 17 was treated with an excess of 2 N NaOH
and stirred at 0 °C until reaction was complete (Scheme 3).
The crude epoxide product was extracted into MTBE and
used without purification in step two.
Scheme 1
Scheme 3
Olefin 12 as a neat oil has been observed to polymerize
upon standing even when stored cold with inhibitors, and it
is recommended that it be used as quickly as possible. Little-
to-no decomposition or polymerization has been observed
for 13 stored in solution (MTBE) for up to 4 months;
however, due to its low (56 °C) onset of thermal decomposi-
tion,8 it is recommended that 13 be stored cold in solution
and used as soon as possible after preparation.
Ring Opening of Epoxide 13: Nonmetal-Catalyzed.
Ring opening of epoxide 13 remains problematic in this
preparation. A significant amount of development has gone
into this reaction trying to control the regiochemical ring
opening of 13. As seen in Scheme 4, there are two positions
that can undergo nucleophilic ring opening. Most common
is for the nucleophile to attack at the unhindered C-2 terminal
position affording the desired ring-opened product 14.
Alternatively, attack at more hindered C-1 is considerably
more likely to occur in 13 because of the resonance
stabilization that can exist. Given our target, we wished to
achieve a regioselective C-2 ring-opening of epoxide 13.
Three classes of reactions were tested to effect this ring
opening: (1) nonmetal assisted, (2) metal-assisted, and (3)
phase-transfer conditions (PTC).
Epoxidation of Olefin 12. Our initial epoxidation condi-
tions for 12 focused on work reported by Thurkauf et al.7
Thurkauf’s procedure called for treating 2-vinylpyridine with
N-bromosuccinimide (NBS) in a solution of dioxane, water,
and acetic acid. Subsequent treatment of the reaction mixture
with Na2CO3 afforded the desired epoxide in 66% yield.
Employing these conditions on 12, epoxide 13 was obtained
in ca. 66% (Scheme 2). While we were able to reproduce
Thurkauf’s yield, many other byproducts were formed during
this reaction, and purification of 13 required column chro-
matography. In addition to the problematic chemistry,
carcinogenic dioxane as a solvent was unattractive from a
scale-up perspective. Substituting THF for dioxane in
Tarkuaf’s procedure did increase the yield to around 80%
with the formation of one (major) unidentified impurity.
Several modifications such as temperature, use of different
bases, the rate and order of addition of reagents, and the
solvent/water ratios failed to suppress this impurity.
Scheme 4
Scheme 2
A preliminary literature search for epoxide ring opening
with phenols revealed two recent reports in which sodium
phenoxides opened the ring of 2-pyridyloxirane predomi-
nantly at C-2 at 60-90 °C in DMF in 50-68% yield.9 Since
a procedure was available for preparation of the potassium
salt of 4-hydroxybenzaldehyde from our previous scale-up
of other pioglitazones, we chose to substitute the potassium
salt into our reaction. Using a 2:1 ratio of potassium salt to
epoxide at 60-65 °C, we obtained a mixture of 14 and 18
in 65% yield. These products were purified by chromatog-
However, the use of tert-butyl alcohol/water (1:5) as the
solvent afforded the epoxide 13 in 90% yield (>92% purity
722
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

raphy on silica gel, providing a 43% yield of 14 as pale
yellow crystals and a 5% yield of 18 as an oil. Variation of
the molar ratios, dipolar aprotic solvent, temperature, time,
and the use of 4-hydroxybenzaldehyde as proton donor, failed
to produce a significant improvement in the outcome. A
considerable amount of insoluble brown tar was produced
in all cases, making workup of these reactions difficult.
Ring Opening of Epoxide 13: Metal-Assisted Ring
Opening. In an effort to suppress the undesired ring opening
at C-1, we explored using a Lewis acid to affect the
regiochemistry. Evidence in the literature suggested that ring
opening of 2-pyridyl epoxides with water, halides, phenols,
amines, and alcohols catalyzed by Mg(II), Cu(II), Ni(II), and
Zn(II) metal ions is highly regioselective for C-2 attack.10
This regiochemical control can be rationalized by chelation
of the bidentate metal to the oxygen lone pair on the oxirane
as well as to the pyridine nitrogen via 19 favoring attack at
C-2 (Scheme 5).
Singh and Sekar recently reported that catalytic Cu(OTf)2
catalyzes the ring opening of epoxides with a variety of
amines to afford good yields of the resulting amino alco-
hols.11 Initial results employing Singh’s chemistry on 12
seemed promising. Epoxide 13 was treated with 4-hydroxy-
benzaldehyde (2 equiv) and Cu(OTf)2 (10 mol %) in THF
at room temperature (Scheme 6). HPLC analysis of the crude
reaction mixture indicated a high selectivity for C-2 attack,
and consumption of 13 was complete within 24 h. However,
isolation of the crude product was problematic as the reaction
produced a thick insoluble black tar isolating 14 in an
unacceptable 20-50% yield.
Scheme 6
Scheme 5
In an attempt to improve the isolated yield, several
reaction parameters were investigated. The concentration was
found to have a dramatic effect on the rate of reaction.
Control experiments found that the optimal reaction rate was
determined to be at a reaction concentration of 2.6 M based
on phenol. Reactions run at lower concentration often did
not go to completion coupled with compromised attack at
the C-1 position. Decreasing the amount of catalyst to as
low as 1 mol % did not reduce the amount of tar formation
and required elevated temperatures to drive the reaction to
completion. The stoichiometry of the phenol was investi-
gated, and it was found that employing 2 equiv seemed to
give the best yield and rate of reaction. Decreasing this
amount to 1.5 equiv or lower led to incomplete reaction and
increased impurity formation. It did appear that THF was
the solvent of choice, however, mixtures of THF and MTBE
were found to work as well. Cu(OTf)2 was found to be
insoluble in MTBE. Tar formation for reactions carried out
at room temperature were found to be comparable to those
reactions carried out at 45 °C, while heating the reactions
above 45 °C often led to increased amounts of C-1 attack
and other impurities. We did not investigate the origin for
tar formation; however, a control reaction in which epoxide
13 was stirred with Cu(OTf)2 (10 mol %) in THF at room
temperature, resulted in decomposition of 13 with ensuing
tar formation. Likewise, it is conceivable that epoxide 13
could potentially react with alcohol 14 to give a polymeric
Our first attempt at this chelation-controlled opening of
13 was done with the magnesium salt of 4-hydroxybenzal-
dehyde. This compound was easily prepared from com-
mercially available magnesium methoxide and 16 in methan-
ol at room temperature. The solvent was removed in vacuo,
and the solid was dried under high vacuum with heat.
Epoxide 13 was then treated with this magnesium salt (0.5
equiv) and 4-hydroxybenzaldehyde (1 equiv) as proton donor.
After 2 days at 70 °C in DMF a 62% crude yield of product
was obtained, consisting of 14 (80%) and 18 (1%). While
the use of divalent magnesium uniformly reduced the amount
of undesired isomer to < 2%, the use of monovalent
potassium or sodium salts of 4-hydroxybenzaldehyde resulted
in 15-20% undesired product. However, problematic with
both sets of reaction conditions was the formation of a dark,
insoluble tar that made workup difficult. Due to the low onset
of thermal decomposition for 13 we hoped to avoid this tar
formation by running this reaction at temperatures below 50
°C; however, these failed to suppress tar formation. While
it was found that the small amount of wrong isomer 18 that
was formed in these metal-catalyzed reactions could be
removed in the next step by the usual crystallization from
methanol (vide infra), more of a concern was developing an
effective method to isolate the product from this tar.
12
material. Other Lewis acids such as Mg(OTf)2 or CoCl2
gave consistent results similar to that for Cu(OTf)2, whereas
reactions carried out with Ti(O-iPr)4 stalled at 50% remaining
13.
From the experimental evidence obtained thus far, em-
ploying divalent metal catalysts appeared to favor attack at
(9) (a) Maoleon, D.; Pujol, M. D.; Rosell, G. J. Med. Chem. 1988, 31, 2122.
(b) Hanzlik, R. P.; Edelman, M.; Michaely, W. J.; Scott, G. J. J. Am. Chem.
Soc. 1976, 98, 1952.
(10) (a) Hanzlik, R. P.; Hamburg, A. J. Am. Chem. Soc. 1978, 100, 1745. (b)
Hanzlik, R. P.; Michaely, W. J. J. Chem. Soc., Chem. Comm. 1975, 113.
(11) Singh, V. K.; Sekar, G. J. Org. Chem. 1999, 64, 287.
(12) Iqbal, J.; Pandey, A. Tetrahedron Lett. 1990, 31, 575.
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
723

the C-2 position; however, the ensuing tar formation coupled
with problematic workup and isolation of crude product made
this approach unattractive from a scale-up perspective.
Further development on this approach was abandoned at this
time.
of PEG 4000 and heated to 80 °C (eq 3).8
Ring Opening of Epoxide 13: Phase-Transfer Condi-
tions (PTC). We were hopeful that we could achieve a
regioselective ring opening of 13 and suppress tar formation
using phase-transfer conditions. There are few examples in
the literature for the preparation of phenolic ethers using
PTC13 and we were concerned at the outset for this chemistry
due to the poor nucleophilicity of 16. Initially, we screened
three readily available PT catalysts we had in-house, A175
(methyl-tribuylammonium chloride), A336 (tricaprylyl-
methylammonium chloride) and BTBAB (benzyl(tributyl)-
ammonium bromide). To test this reaction, 1 N NaOH (0.5
equiv), the PT catalyst, and epoxide 13 were added to a
solution of 4-hydroxybenzaldehyde (1.5 equiv) in MTBE/
H2O (1:1), and the reaction was vigorously stirred for 12 h
(eq 1).
After stirring for 24 h, the reaction was cooled to room
temperature, and HPLC analysis indicated no 13. To our
pleasure, we found that little to no tar resulted from these
conditions and that workup proved much easier, providing
a crude dark brown/black solid in 50-70% mass recovery.
Unfortunately, the authors’ high regiochemical claim could
not be applied to this reaction, and between 10 and 20 area
% of 18 resulted. Products 14 and 18 could not be easily
separated at this stage; however, these regioisomers could
be separated during the purification of the Knoevenagel
products (vide infra). Although these conditions resulted in
up to as much as 20% of 18, the lack of tar formation made
this attractive for scale-up.
A direct comparison of the ring opening of 13 under PTC
or with the magnesium salt of 16 revealed that both methods
gave comparable yields. The PTC process proved operation-
ally superior with respect to lack of tar formation, whereas
the magnesium process resulted in significant tar formation
that required a tedious and difficult workup and was thus
abandoned at this time. Because epoxide 13 would be used
as a solution in MTBE in this reaction, we examined if any
problems would arise by having this solvent in our phase-
transfer chemistry. We were pleased to find that introducing
13 dissolved in MTBE (1 kg epoxide/L MTBE) caused no
adverse effects on the reaction, yield, tar formation, or
regiochemical outcome.
In our first scale-up run of step 3, we allowed this reaction
to proceed until less than 1% of 13 remained. While we were
able to reach this limit, impurities generated during this
prolonged reaction (28 h at 80 °C) resulted in a 67% yield
of 14 and 18. Knowing that we could upgrade in the next
step, we decided to raise the level of 13 that remained to
3%, thus decreasing the reaction time to about 17 h. While
use tests in the laboratory showed that running this reaction
until 3% of 13 remained caused a decrease in the yield of
up to 10%, the amount of impurities generated were
significantly less.
Unfortunately, only recovered 13 was observed for all PT
catalysts tested. We felt that adding a Lewis acid to this
reaction might overcome the lack of reactivity; we thus hoped
that we could introduce inexpensive CuSO4 to aid in both
reactivity and regioselectivity. Repeating these reactions with
CuSO4 (100 mol %) we observed complete consumption of
13 in the HPLC assay; however disappointingly, a significant
amount of tar was formed, and a low isolated yield of 14
resulted (eq 2). Decreasing the amount of CuSO4 decreased
the amount of tar, but also decreased the yield and eroded
the regioselectivity.
Knoevenagel Condensation Following previous pio-
glitazone methodology,2,15 only minor modifications were
made to include removal of half the solvent (methanol) prior
to crystallization and the addition of water washes to re-
move residual pyrrolidine from the product. Although we
did not study this, we were concerned that small amounts
of pyrrolidine could interfere with the cobalt-catalyzed
reduction of 15 to 2. Additional water washes of the product
reduced the pyrrolidine level to below 3% which was an
acceptable level giving a good yield in the methanol
trituration.16
Disappointed in our preliminary PTC results, we subse-
quently performed a literature search which revealed a report
by Bhattacharyya and Maiti who report the ring opening of
oxiranes with sulfides and sulfones catalyzed by poly-
(ethylene glycol) 4000 (PEG 4000).14 An important aspect
of this paper was the authors’ claim that these conditions
were highly regiospecific, giving ring opening from the less
hindered site in the absence of a Lewis acid. Therefore, to
test this claim, epoxide 13 was added to the preformed
sodium phenoxide in toluene/H2O with a catalytic amount
Treating a mixture 14 and 18 (4:1) with 2,4-thiazo-
lidinedione and pyrrolidine in CH3OH (35-45 °C) after
(13) McKillop, A.; Fiaud, J.-C.; Hug, R. P. Tetrahedron 1974, 30, 1379 and
references cited therein.
(14) Bhattacharyya, P.; Maiti, A. K. Tetrahedron 1994, 50, 10483.
(15) Sohda, T.; Ikeda, H.; Megura, K. Chem. Pharm. Bull. 1995, 43, 2168.
(16) As determined by 1H 400 MHz NMR.
724
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

workup gave a mixture of the olefin products 15 and 21 (97:
3) (Scheme 7). The crude product mixture was purified by
triturating the solids at reflux in methanol (20 mL/g) which
reduced the level of 21 to below 0.5% with a ∼90%
recovery. Solids purified in this manner were isolated as a
solvate and contained about 7% methanol.
described in the earlier syntheses was avoided, and simple
extraction into EtOAc was effective. A near quantitative yield
for this reduction has been observed, providing 2 in greater
than 95 area% purity.
Although 2 could be isolated and purified, this added an
additional purification step to our synthesis, and it was
desirable, given the excellent yield and high purity of 2, to
not isolate this material but to use it crude in the subsequent
oxidation. This idea presented a few challenges to include
extraction of 2 into CH2Cl2 in which 2 was not significantly
soluble, as well as determining if having THF (from the
reduction) in the subsequent oxidation reaction would be
problematic. The former was easily tested, and it was found
that after reduction of 15, extraction of crude 2 into CH2Cl2
proved quite facile (probably due to the THF present);
however, 2 was found to precipitate from this solution if
left standing at 20-25 °C. The next question that needed
attention was the impending THF that was present in the
CH2Cl2 from the reduction step. Control reactions conducted
employing up to as much as 20 vol % THF in the oxidation
found that this did not compromise reactivity or product
quality. Because repeated distillation of the THF from CH2-
Cl2 would prove cumbersome upon scale-up, it was decided
to add the appropriate amount of DMSO which was to be
used for the oxidation and subsequently distill the THF and
CH2Cl2 under vacuum, leaving the product as a solution in
DMSO. This method proved successful at removing all or
most of the THF from the DMSO, and oxidation of 2 using
this solution has been carried out without incident (vide
infra).
Scheme 7
Reduction of Olefin 15. At this point in the synthesis
we hoped to apply the chemistry developed in-house for
previous pioglitazone syntheses for the reduction 15 to 2.17
The initial reduction of 15 using these conditions proved-
problematic in that the reaction would stall after 80%
conversion (eq 4).
Oxidation of 2. Prior unoptimized efforts for the oxidation
of 2 to 3 in our group on small kilogram scale proved
problematic with respect to isolation of the final product.19
Purification of 3 required silica gel chromatography followed
by several upgrade recystallizations giving 3 in 40-45%
yield. While many oxidation conditions were tested, only
conditions reported by Taber were successful.1,20 Following
a modified Taber procedure, oxidation of 2 was carried out
by treatment of a slurry of P2O5 (2.5 equiv) in CH2Cl2 cooled
to 0 °C with a solution of 2 dissolved in DMSO (8 equiv)/
CH2Cl2 followed by the addition excess Et3N (eq 5).
Attempts to restart this reduction by adding either more
NaBH4 or CoCl2/DMG (dimethylglyoxime) catalyst did not
consume the remaining 15. In addition to this reduction
stalling, we saw the need to develop a less cumbersome
workup.
We quickly identified conditions to prevent this reduction
from stalling by adjusting the reaction stoichiometry.17,18
Optimal conditions for reduction were found to be: to a
stirring solution of 13 (1 equiv), CoCl2 (0.0006 equiv), and
DMG (0.029 equiv) in H2O, THF, and 1 N NaOH (1.6:1:
1.1) was added dropwise a solution of NaBH4 (1.44 equiv)
dissolved in 0.2 N NaOH, maintaining the internal temper-
ature between 15 and 20 °C. The observed blue/purple
reaction mixture gives a visual indication that reduction is
taking place, and consequently, when the reaction becomes
yellow, no reduction is occurring. The ideal pH range for
this reduction was determined to be between 9 and 11. We
found that if the reaction pH becomes greater than 11, it
can be restarted by lowering the pH with acetic acid. Once
the reduction was complete, excess acetone was added to
consume any remaining NaBH4, and the crude product 2 was
extracted into EtOAc. The complex workup procedure
While oxidation of 2 to 3 occurred without incident, 10-
25% of an impurity was formed under these reaction
conditions. Although not identified prior to our efforts, the
known Swern oxidation pathway would expect to yield 23
as a potential byproduct (Scheme 8).21
Scheme 8
(17) Huber, J.; Hewitt, B. Pharmacia Corp., Early Process Research
Development, Kalamazoo, MI. Unpublished results.
&
(18) Leutenegger, U.; Madin, A.; Pflatz, A. Angew Chem., Int. Ed. Engl. 1989,
28, 60.
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
725

Disappointingly, the use of the more bulky base diisopropyl-
ethylamine (DIPEA) did not suppress it’s formation; thus,
this impurity was isolated and characterized and found to
be consistent with ketone 24 and not the initial hypothesis
of 23 (Figure 3).
observed. It was found that if water (pH ) 7) was added to
the crude reaction mixture, the amount of 24 observed after
the workup was significantly higher than it was prior to
workup. Alternatively, if the crude oxidation mixture was
quenched into water (pH ) 7), the amount of 24 before and
after workup remained the same. Current conditions now call
for this oxidation to be quenched into water (pH ) 7).
Unfortunately, after extraction with methylene chloride, the
resultant product did not crystallize, presumably due to
residual DMSO present after the extractions. Subsequent
aqueous washes of the combined CH2Cl2 layers were not
successful at removing this DMSO. Alternatively, the crude
oil could be dissolved in EtOAc and washed with water (pH
) 7) which after removal of the solvent, provided 3 as an
off-white to light-yellow/orange solid. Recrystallization of
this material from absolute ethanol (25 mL/g) provided 3 as
a colorless solid.
Figure 3.
Formation of 24 can be rationalized by the fact that as
oxidation of 2 to 3 slows, 3 could react with any activated
DMSO complex providing 24. To test this hypothesis, 3 was
submitted to the oxidation conditions and after stirring at
room temperature for 24 h provided 24 in quantitative yield.
After reexamination of the current oxidation conditions using
2.5 equiv of P2O5 and 8 equiv of DMSO, it seemed
reasonable that if the amount of activated DMSO complex
was decreased, less alkylation on the imide would result.
Therefore, oxidation of 2 to 3 was carried out using 2.5 equiv
of P2O5 and 5 equiv of DMSO stirring at 0 °C, and the
reaction progress was monitored by HPLC analysis. Nor-
mally, 2 is dissolved in DMSO prior to addition; however,
by employing less DMSO in this reaction we encountered
solubility problems and had to add 2 as a DMSO slurry.
After stirring for 50 h, oxidation stalled at 75% conversion,
however, only 7% of 24 was detected by HPLC. In an effort
to understand the relative rate of oxidation, we carried out
an experiment in which we removed reaction samples over
a 50-h period and recorded the relative amounts of 3 and
24. We were surprised to find that almost all of the oxidation
was complete within 45 min, and that as reaction time
continued, both 2 and 3 decreased. This analysis was key to
understanding that to obtain the highest yield for this
oxidation, we needed to allow 2 to react with an excess of
activated DMSO complex at low temperature for about 1 h
and then quench the reaction. To test this hypothesis,
oxidation of 2 now employed 5 equiv P2O5 and 10 equiv of
DMSO in CH2Cl2 at -30 °C followed by the addition of 3
equiv of DIPEA; HPLC analysis showed that 15% of 2
remained after 45 min and that very little (approximately
4%) of 24 was produced. Alternatively, running this reaction
at 0 °C gave complete consumption of 2 and between 3 and
7% of 24 which could be removed in the recrystallization
step (eq 6).
Conclusions
Herein we have described a six-step synthesis of the
antidiabetic candidate 3 in an overall 25% yield from olefin
12. The regiochemical ring opening of epoxide 13 with
4-hydroxybenzaldehyde proved difficult; thus, we chose a
nonselective phase-transfer reaction on the basis of ease of
workup, giving up to as much as 20% of the undesired ring-
opened product. This mixture was carried on crude into the
Knoevenagel condensation at which time the undesired
product could be removed. The reduction conditions in step
5 were greatly improved from past analogue syntheses with
a simplified extractive workup. Controlling the stoichiometry,
temperature, and reaction time proved vital to obtaining a
high yield in the final oxidation reaction.
Experimental Section
1
General Procedures. H and 13C spectra were obtained
using a Bruker Avance 400 in dilute d6-DMSO or CDCl3
solution. Chemical shifts are reported as δ (ppm) values from
Me4Si as an internal standard.
5-Ethyl-2-oxiran-2-ylpyridine (13). To a dry, nitrogen-
purged 4000-L glass-lined reactor was added NBS (133 kg,
747.4 mol, 1.22 equiv) and water (815 L), and the resultant
slurry was cooled to 0 °C. To a separate dry nitrogen-purged
4000-L glass-lined reactor (add reactor) was added tert-butyl
alcohol (122 kg) followed by ethyl vinyl pryidine 12 (81.6
kg, 612.6 mol, 1 equiv) and this homogeneous dark solution
was cooled to 0 °C at which time glacial acetic acid (40.5
kg, 674.4 mol, 1.1 equiv) was added, resulting in a 4 °C
exotherm. This tert-butyl alcohol/12/HOAc solution was then
added to the NBS slurry at rate to maintain the internal
temperature below 10 °C. Once this addition was complete,
the reaction was allowed to stir at ca. 10 °C. After 1 h, the
dark-yellow reaction slurry was assayed by TLC and GLC,
indicating that no 12 remained. The reaction was then
quenched with excess 2 N NaOH (1200 L) at a rate to
maintain the internal temperature below 25 °C. During this
addition the reaction color changed from a bright yellow to
a very dark brown. After the addition of the 2 N NaOH, the
reaction stirred for 30-60 min and was judged complete by
GLC analysis. The crude epoxide was then extraced into
During our development of this oxidation, we noticed that
the quench method seemed to affect the amount of 24
(19) Conway, B.; Veley, M. Pharmacia Corp., Early Process Research &
Development, Kalamazoo, MI. Unpublished results.
(20) Taber, D. F.; Amedio, J. C., Jr.; Jung, K.-Y. J. Org. Chem. 1987, 52, 5621.
(21) Swern, D.; Omura, K. Tetrahedron 1978, 34, 1651.
726
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development

MTBE (305 kg, 412 L) followed by the addition of DW
water (496 L) and sodium chloride (158 kg). The aqueous
layer was separated and subsequently extracted twice with
MTBE (147 kg, 193 L), and the combined organic layers
were then distilled under vacuum to an approximate 1:1
volume of product to MTBE and then transferred into drums.
The crude isolated yield for this reaction was found to be
95% (87.1 kg of epoxide 13), and the purity for this material
was 92.7 area % (GLC). This epoxide was used without
further purification, and these drums where stored cold until
4000-L reactor that contained crude 14 (87 kg, 320 mol, 1
equiv) and was stirred at ca. 25 °C until all the solids were
dissolved at which time pyrrolidine (22.9 kg, 320 mol, 1
equiv) was added. This reaction mixture was then heated to
ca. 38 °C to 42 °C for at least 1 h and then assayed. Once
complete, the contents were cooled to ca. 25 °C, and the pH
was adjusted to less than 7 with acetic acid (21.1 kg). The
reaction was then distilled to about 100 L and then cooled
to ca. 20 °C to effect crystallization of the desired product.
These solids were collected by filtration and washed twice
with cold (-5 °C) methanol (110 L) followed by washing
with 20 °C water (3 × 160 L) and then blown dry with heated
single-pass nitrogen (35-40 °C), affording 70.4 kg (59%)
of 15 as a dark solid. Recrystallization of crude 15: To a
dry, nitrogen-purged 4000-L glass lined reactor was added
crude PNU-276559 (15) (70.4 kg, 190 mol) followed by
methanol (1050 L). The resultant slurry was then heated to
reflux for at least 60 min. Note: Not all solids are dissolVed
after this time. The reaction was then cooled to ca. 25 °C
over at least 30 min at which time the reaction was further
cooled to ca. 0 °C over a 30-min period and stirred for at
least 30 min at this temperature. The resulting solids were
collected by filtration and washed with cold (0 °C) methanol
(3 × 115 L). The solids were dried with single-pass nitrogen
heated to ca. 35-40 °C until the residual methanol content
was between 6 and 8 wt %, affording 66.8 kg (95%) of 15
as a slightly colored solid. 1H NMR (d6-DMSO, 400 MHz)
δ 8.40 (s, 1H), 7.44 (s, 1H), 7.66 (m, 1H), 7.52 (m, 2H),
7.11 (d, 2H), 5.83 (bs, 1H), 4.98 (bs, 1H), 4.39 (m, 1H),
4.19 (m, 1H), 2.61 (q, J ) 7.6 Hz, 2H), 1.19 (t, J ) 7.6 Hz,
3H); 13C NMR (d6-DMSO, 100 MHz) δ 168.29, 167.80,
160.74, 158.70, 148.32, 138.00, 136.23, 132.39, 132.14,
135.85, 120.97, 120.64, 115.81, 72.83, 72.18, 25.37, 15.70.
5-{4-[2-(5-Ethylpyridin-2-yl)-2-hydroxyethoxy]benzyl}-
1,3-thiazolidine-2,4-dione (2). To a clean dry 5-L round-
bottom flask was added alcohol 15 (100 g, 0.29 mol), CoCl2
(0.045 g, 0.0002 mol), DMG (0.91 g, 0.0078 mol) followed
by water (295 mL), THF (175 mL), and 1 N NaOH (185
mL). To this resulting solution was added at room temper-
ature, NaBH4 (15.7 g, 0.42 mol) dissolved in 0.2 N NaOH
(240 mL) over 30 min, resulting in a deep-purple solution.
The reaction was adjusted to pH ≈ 10 with acetic acid and
monitored by HPLC until complete (less than 3% 15
remaining). Once judged complete, the reaction was quenched
by slow addition of acetone (150 mL). After stirring for 15
min, the reaction was was adjusted to pH ≈ 3 with 6 N HCl
and then to pH ) 7 with 2 N NaOH. The product was
extracted into CH2Cl2 (3 × 500 mL) and then concentrated
to dryness to afford 103 g (>100%) of alcohol 2 as a solid.
This material could be used without purification. Spectral
data were consistent with literature values.1
1
needed for step 2. H NMR (CDCl3, 400 MHz) δ 8.41 (s,
1H), 7.36 (m, 1H), 7.17 (m, 1H), 3.61 (s, 1H), 2.99 (m, 1H),
2.95 (m, 1H), 2.80 (q, J ) 7.6 Hz, 2H), 1.90 (t, J ) 7.6 Hz,
3H); 13C NMR (CDCl3, 100 MHz) δ 154.31, 148.96, 138.67,
135.94, 119.28, 52.59, 50.01, 28.33, 25.67, 15.15.
4-[2-(5-Ethylpyridin-2-yl)-2-hydroxyethoxy]benzalde-
hyde (14). To a dry, nitrogen-purged 4000-L glass-lined
reactor was added 4-hydroxybenzaldehyde (108.8 kg, 890.
mol, 1.5 equiv), poly(ethylene glycol) 4000 (PEG4000) (14.6
kg) and 1 N NaOH (843 L), resulting in a dark homogeneous
solution. To this mixture was added toluene (896 L) followed
by the crude epoxide 13/MTBE solution (87.1 kg, 583.8 mol,
1 equiv), and the contents were heated to ca. 78 °C for
approximately 13-17 h at which time the reaction was
assayed by HPLC. Once complete (<3% starting 13), the
reaction was cooled to room temperature, and the layers were
allowed to separate. The aqueous layer was separated and
extracted with toluene (3 × 405 L), and the combined organic
layers were washed twice with 1 N NaOH (594 L) and once
with DW water (405 L). The organic layer was distilled under
vacuum to about 300 L at which time the crude reaction
was transferred into drums to determine yield. Note: It has
been obserVed that concentration of this material to low
Volume can cause the product to precipitate from solution.
The crude yield for this reaction was determined to be 57.3%
(92.3 kg product 14) whose HPLC assay indicated a 18.5/
81.5 product ration of 18 to 14. The crude product was then
charged back into the reactor, the drums were rinsed with
toluene, and the mixture was vacuum-distilled to dryness,
1
affording a black solid. 14: H NMR (CDCl3, 400 MHz) δ
9.86 (s, 1H), 8.41 (s, 1H), 7.78 (m, 2 H), 7.57 (m, 2 H),
7.40 (m, 2 H), 7.02 (m, 2 H); 5.14 (t, J ) 5.6 Hz, 1 H), 4.67
(s, 1 H), 4.27 (m, 2 H), 2.65 (q, J ) 5.2 Hz, 2 H), 1.23 (t,
J ) 5.2 Hz, 3 H); 13C NMR (CDCl3) δ 190.69, 163.6, 155.63,
148.09, 138.74, 136.28, 131.86, 130.11, 120.76, 114.87,
1
72.58, 71.20, 25.71, 15.20. (18) H NMR (CDCl3) δ 9.75
(s, 1 H), 8.36 (s, 1 H), 7.67 (m, 2 H), 7.41 (m, 1 H), 7.20
(m, 1 H), 6.94 (m, 2 H), 5.41 (t, J ) 5.6 Hz, 1 H), 4.00 (d,
J ) 5.6 Hz, 2 H), 2.56 (q, J ) 7.6 Hz, 2 H), 1.16 (t, J ) 7.6
Hz, 3 H); 13C NMR (CDCl3) δ 189.66, 161.77, 153.60,
148.07, 138.02, 135.55, 130.92, 129.35, 119.74, 114.94,
80.30, 64.64, 24.76, 14.06.
(5Z)-5-{4-[2-(5-Ethylpyridin-2-yl)-2-hydroxyethoxy]-
benzylidene}-1,3-thiazolidine-2,4-dione (15). To a dry,
nitrogen-purged 4000-L glass-lined reactor (add reactor) was
added thiazolidine-2,4-dione (41.4 kg, 353 mol, 1.1 equiv)
followed by methanol (1150 L) and then stirred at ca. 25 °C
until dissolved. This solution was then transferred into a
5-{4-[2-(5-Ethylpyridin-2-yl)-2-oxoethoxy]benzyl}-1,3-
thiazolidine-2,4-dione (3). To a dry 250-mL round-bottom
flask was added P2O5 (9.52 g, 67.1 mmol) followed by CH2-
Cl2 (40 mL) giving a stirrable slurry. Once the reaction
cooled to 0 °C, 2 (5.0 g, 13.42 mmol) dissolved in DMSO
Vol. 6, No. 5, 2002 / Organic Process Research & Development
•
727

(9.5 mL, 10 equiv) was added dropwise, maintaining an
internal temperature of about 0 °C, causing the contents to
become orange. After stirring for 15 min, DIPEA (7 mL,
40.26 mmol) was added dropwise, maintaining an internal
temperature of about 0 °C, causing the contents to become
homogeneous. After stirring for 45 min, the reaction mixture
was poured into water (pH ) 7). The product was extracted
into EtOAc (3 × 50 mL), the combined organic layers were
washed with water (pH ) 7) (3 × 50 mL), and the solvent
was removed in vacuo. The resultant solid was then
recrystallized from ethanol (25 mL/g) to afford 3.72 g (75%)
of 3 as a white solid. Spectral data were consistent with
literature values.1
Acknowledgment
We thank Dr. Richard Kurtz, Dr. Krister Lundmark and
Dr. Kerstin Nordstrom for their helpful discussions. We also
thank Deidre DeKam for her work on the oxidation of 2 to
3. The hard work of our pilot plant, process safety, and
documentation staff is duly recognized.
Received for review May 13, 2002.
OP025549S
728
•
Vol. 6, No. 5, 2002 / Organic Process Research & Development