Synthesis of a NO-releasing prodrug of rofecoxib

Article abstract of DOI:10.1021/jo051712g

A newly developed synthesis of a NO-releasing prodrug of rofecoxib is described. The highly productive process consists of five chemical steps and produces prodrug 1 in an overall 64% yield from commercially available 3-phenyl-2-propyn-1-ol (4). The synth

Full text of DOI:10.1021/jo051712g

Synthesis of a NO-Releasing Prodrug of Rofecoxib
F. Conrad Engelhardt,* Yao-Jun Shi, Cameron J. Cowden, David A. Conlon, Brenda Pipik,
George Zhou, James M. McNamara, and Ulf-H. Dolling
Department of Process Research, Merck & Company, Inc., P.O. Box 2000,
Rahway, New Jersey 07065-0900
conrad_engelhardt@merck.com
ReceiVed August 12, 2005
A newly developed synthesis of a NO-releasing prodrug of rofecoxib is described. The highly productive
process consists of five chemical steps and produces prodrug 1 in an overall 64% yield from commercially
available 3-phenyl-2-propyn-1-ol (4). The synthesis is highlighted by the carbometalation reaction of
propargyl alcohol 4 to generate the tetrasubstituted olefin core, sulfone acid 2. Additionally, two alternate
end-game strategies to prepare NO-COXIB 1 from this intermediate were explored and developed: (1)
a convergent synthesis where a bromonitrate side chain is introduced in one step and (2) a two-step
sequence that first installs the requisite six-carbon ester side chain followed by chemoselective nitration.
Introduction
Americans take low-dose aspirin daily to reduce their risk of a
cardiovascular event and, because of the potential of increased
GI damage, are unable to benefit from the advancements offered
by COX-2 selective drugs. Awareness of these protective effects
and the emerging incompatibility of a low-dose aspirin regimen
with COX-2 selective inhibitors led to the development of a
new class of drugs, the COX inhibiting nitric oxide donators
(CINODs). CINODs are a new class of antiinflammatory and
analgesic drugs that may minimize gastrointestinal toxicity
compared with standard nonsteroidal antiinflammatory drugs
(NSAIDs) by virtue of this nitric oxide donation. This drug class
is characterized by a COX inhibiting drug or prodrug linked to
a nitric oxide moiety. In animal studies, CINODs lead to marked
reductions in gastroduodenal and small bowel injury. Data have
suggested that CINODs have the potential to be clinically
beneficial with an improved gastrointestinal safety profile that
will open up the possibility of codosing with patients who take
low-dose aspirin.
Nonselective nonsteroidal antiinflammatory drugs (NSAIDs)
are widely used in the treatment of pain and arthritis, but these
benefits are counterbalanced by gastrointestinal (GI) toxicity
which is a common side effect of dual cyclooxygenase (COX-1
and COX-2) activity. This prompted the search for safer
therapies, including the development of COX-2 selective
inhibitors such as rofecoxib (VIOXX). Rofecoxib demonstrated
high selectivity for the inhibition of the induced COX-2 isoform
while sparing inhibition of the “housekeeping” COX-1, resulting
in superior GI tolerability.
Although COX-2 inhibitors do not interfere with the anti-
platelet activity of aspirin, they significantly increase the
gastrointestinal injury caused by aspirin.1 Currently, 26 million
(1) Fiorucci, S.; Santucci, L.; Wallace, J. L.; Sardina, M.; Romano, M.;
Soldado, P.; Morelli, A. Proc. Natl. Acad. Sci. U.S.A. 2003, 88, 10937-
10941.
10.1021/jo051712g CCC: $33.50 © 2006 American Chemical Society
Published on Web 12/17/2005
480
J. Org. Chem. 2006, 71, 480-491

Synthesis of a NO-Releasing Prodrug of Rofecoxib
FIGURE 1. Rofecoxib starting material.
SCHEME 1. Prodrug for Rofecoxib (VIOXX) and Nitric Oxide
SCHEME 2. Retrosynthetic Analysis
The novel CINOD (1) shown in Scheme 1 is a prodrug for
both rofecoxib (VIOXX) and nitric oxide. Under physiological
conditions, the acetate will be removed via esterase activity and
subsequent ring closure will release the COX-2 selective
inhibitor, rofecoxib. Additionally, the released side chain is
expected to generate 1,6-hexanediol and nitric oxide (Scheme
1).
the use of a magnesium-mediated carbometalation reaction to
prepare (Z)-2,3-biaryl-4-acetoxybut-2-enoic acid2 to gain access
to NO-containing COX-2 inhibitors such as 1. In this paper,
we report the successful implementation of this approach and
discuss alternate end-game strategies to completing the synthesis
of NO-containing 1 from the sulfide acid 3.
The novel COX-2 selective, nitric oxide compound 1 was in
development for the treatment of pain, while hopefully being
gastric sparing. As part of our program toward the synthesis of
prodrug 1, a convergent strategy was envisioned where a nitrate-
containing C6 side chain could be directly coupled to the (Z)-
4-alkoxybut-2-enoic acid core 2 (Scheme 2). In turn, methyl
sulfone 2 could be accessed by simple oxidation of the sulfide
acid 3 precursor. The key step to the synthesis, however, is the
one-pot construction of the tetrasubstituted alkene core of 1
using a Grignard carbometalation reaction between aryl Grignard
5 and propargyl alcohol 4.
The regio- and stereocontrolled preparation of tetrasubstituted
alkenes remains a difficult synthetic challenge. Recently, work
by Fallis and co-workers highlighted the versatility and power
of magnesium-mediated carbometalation reactions to access
well-defined olefin systems. Inspired by this work, we examined
Results and Discussion
(Z)-2,3-Biaryl-4-hydroxybutenoic acid 7 represents a good
entry point to the new class of NO-COXIB inhibitors (Figure
1).3 Rofecoxib (6) appeared to be the ideal starting material, as
it already contained the requisite biaryl-Z-olefin along with the
appropriate carboxylic acid and allyl alcohol oxidation states
(Figure 1). Unfortunately, any attempts to prepare ring-opened
products directly using rofecoxib were unsuccessful. Although
ring opening was difficult, reduction of rofecoxib was possible
yielding the pseudosymmetric diol 8. Unfortunately, this
molecule still required a desymmetrization reaction, a protecting
group manipulation, and oxidation to arrive at the desired (Z)-
(2) The synthesis of a series of (Z)-2,3-biaryl-4-acetoxybut-2-enoic acids
was reported in International Patent Publication WO 96/13483 (1996).
(3) Internal communication with Merck Medicinal Chemistry.
J. Org. Chem, Vol. 71, No. 2, 2006 481

Engelhardt et al.
results and systematically investigated the carbometalation
reaction sequence in its three parts: (1) the carbometalation,
(2) the carboxylic acid installation, and finally (3) the acetate
trap.
1. Carbometalation. Several concerns emerged regarding the
carbometalation step: (1) the preparation of Grignard reagent
5 from the less reactive aryl chloride, (2) the excess aryl
Grignard reagent required, and (3) the use of a mixed-solvent
system. Fallis and co-workers routinely report the use of chloro-
based aryl and vinyl Grignard reagents in the carbometalation
reaction. They note that the bromo-derived Grignard reagents
consistently give lower yields than the chloro counterparts.6 This
was an issue as aryl bromides are preferable for the preparation
of Grignard reagents because they are cheaper and more reactive
than the corresponding aryl chlorides. It is also generally easier
to prepare a Grignard reagent from the bromide precursor. We
investigated whether there was a significant advantage in using
a chloride-based Grignard reagent in this reaction. Our experi-
ments7 confirmed the observations by Fallis indicating that the
Grignard reagents derived from aryl or vinyl chlorides are
superior reagents in the carbometalation reaction. The chloride-
derived Grignard reagent reacted more rapidly than the corre-
sponding bromo Grignard to give the desired vinyl Grignard
intermediate 9, and the reaction was generally cleaner with fewer
byproducts observed. One possible rationale for this difference
is the potential presence of an unfavorable Schlenk equilibrium
that may exist in the bromo Grignard case.8
Aware of the advantages of employing an aryl chloride
Grignard, we turned to the preparation of 4-thiomethylphenyl-
magnesium chloride 5 from 4-chlorothioanisole 14 (Scheme 4).
In the initial procedure, 4-chlorothioanisole 14 was added to a
suspension of magnesium (∼20 mesh, 1.25 equiv) and 1,2-
dibromoethane (2.5 mol %) in refluxing THF (Scheme 4). These
conditions always resulted in generation of 4-thiomethylphen-
ylmagnesium chloride 5, but the reaction was unpredictable;
the problem was controlling the initiation of the reaction, a
critical factor for large scale. Typically, 20-60% of the
4-chlorothioanisole 14 had to be charged to observe any reaction.
Although there was never a propensity toward a runaway
reaction, we sought to control the onset of the Grignard
formation as a safety priority. It was discovered that the addition
of 1-2 mol % of the product Grignard 5 to the system was the
FIGURE 2. Proposal to trap the carboxylate/alkoxide intermediate.
2,3-biaryl-4-hydroxybutenoic acid 7. Through these challenges,
it became evident that rofecoxib was not the ideal starting
material; instead, we focused our attention on other methodolo-
gies to prepare hydroxybut-2-enoic acid 7.
As previously mentioned, Fallis has described a method for
preparing 2,3-disubstituted butenolides,4 including rofecoxib, via
Grignard-mediated carbometalation chemistry with propargyl
alcohols (Figure 2). Monosubstituted butenolides and saturated
γ-lactones have also been prepared in a similar fashion using a
hydromagnesiation reaction.5 Presumably, preparation of these
butenolides proceeds through a (Z)-4-alkoxybut-2-ene carbox-
ylate intermediate prior to ring closure (Figure 2). Trapping and
isolation of these tetrasubstituted alkene intermediates have not
previously been described in the context of carbometalation
chemistry. We report that these reactive intermediates can be
intercepted using acetic anhydride to generate a (Z)-2,3-biaryl-
4-acetoxybut-2-enoic acid 3, the core of prodrug 1. The four-
component coupling strategy employed herein involves the
addition of an aryl Grignard reagent 5 to 3-phenyl-2-propyne-
1-ol 4 followed by a carbon dioxide quench and an acetate trap
(Scheme 2). Naturally, challenges arose in the implementation
of such a complex strategy. Each challenge will be addressed
with particular emphasis on the installation of the carboxylic
acid moiety and the subsequent preparation of the acetate-
protected allylic alcohol intermediate.
We initially examined the use of 4-thiomethylphenylmagne-
sium chloride 5 and 3-phenyl-2-propyn-1-ol 4 to prepare vinyl
Grignard intermediate 9 (Scheme 3). Further treatment of this
magnesium chelate 9 with excess carbon dioxide followed by
acetic anhydride generated (Z)-2,3-biaryl-4-acetoxybut-2-enoic
acid 3, albeit in low yield (6%), with a majority of the reaction
mixture being comprised of the following side-product impuri-
ties (Scheme 3). Among the byproducts was thioanisole 10
which is generated from the initial deprotonation of 3-phenyl-
2-propyn-1-ol with aryl Grignard 5. Alkenes 12a,b can be
explained by incomplete reaction of vinyl Grignard 9 with
carbon dioxide followed by a protic workup or acetate quench.
Thioanisic acid 11 results from the reaction of excess aryl
Grignard 5 with carbon dioxide. Finally, lactone 13 results from
rapid ring closure of the carboxylate/alkoxide (Figure 2) after
inefficient trapping with acetic anhydride. Despite low yields
for the desired product 3, we were encouraged by these initial
(6) Fallis, A. G.; Wong, T.; Tjepkema, M. W.; Audrain, H.; Wilson, P.
D. Tetrahedron Lett. 1996, 37, 755-758.
(7) A rate study in a mixed solvent system (cylcohexane/THF 1:3)
confirmed that the 4-thiomethylphenylmagnesium chloride 5 reacted more
quickly and cleanly than the bromo derivative both at room temperature
(23 °C) and at reflux in THF (65 °C). The reaction employing the chloro-
based Grignard reagent also went to completion within 4 h, whereas the
bromo-based reagent stalled at 40-50% conversion and could not be driven
further towards completion.
(8) For a discussion of Schlenk equilibrium of Grignard reagents, see:
(a) Dessy, R. E.; Green, S. E. I.; Salinger, R. M. Tetrahedron Lett. 1964,
21, 1369-1374. (b) Smith, M. B.; Becker, W. E. Tetrahedron 1966, 22,
3027-3036. (c) Smith, M. B.; Becker, W. E. Tetrahedron 1967, 23, 4215-
4227.
(4) Fallis, A. G.; Wilson, P. D.; Forgione, P. Tetrahedron Lett. 2000,
The preparation of a 2 M solution of 4-thiomethylphenylmagnesium bromide
in THF resulted in a heterogeneous solution with significant precipitates
observed. Alternatively, the 4-thiomethylphenylmagnesium chloride reagent,
also prepared as a 2 M solution in THF, appeared homogeneous.
41, 17-20.
(5) Takayori, I.; Okamoto, S.; Sato, F. Tetrahedron Lett. 1990, 44, 6399-
6402.
482 J. Org. Chem., Vol. 71, No. 2, 2006

Synthesis of a NO-Releasing Prodrug of Rofecoxib
SCHEME 3. (Z)-2,3-Biaryl-4-acetoxybut-2-enoic Acid via Carbometalation and Acetate Trapping
SCHEME 4. Preparation of 4-Thiomethylphenylmagnesium
Chloride
we explored the use of a single solvent (THF) in these reactions
and found that only 3-4 h is necessary to affect full conversion
to vinyl Grignard intermediate 9 under these conditions.
To summarize, the new developments to the carbometalation
step include: (1) demonstration of the superiority of the chloro-
based Grignard reagent, (2) implementation of a sacrificial base,
MeMgCl, for the initial deprotonation, (3) reduction in the
charge of 4-thiomethylphenylmagnesium chloride (1.1 vs 3.2
equiv), and (4) simplification by employing THF as the sole
solvent in the reaction. When all of these improvements were
combined, the carbometalation step provided intermediate 9
(>95% conversion) in just 4 h at 65 °C in THF (Scheme 7).15
2. Carboxylic Acid Installation and Acetate Trap. With
access to vinyl Grignard intermediate 9, we looked to the next
best, often providing controlled initiation using only a 5-10%
initial charge of aryl chloride 14.9 Later, accumulated kinetic
data alleviated any remaining safety concerns over the prepara-
tion of this aryl Grignard and confirmed that the reaction could
be controlled.
The use of excess (3.2 equiv) expensive chloro Grignard
reagent employed in the carbometalation step was a concern,
especially as 1 equiv simply serves as a “sacrificial base”
deprotonating propargyl alcohol 4. A significant improvement
was made by substituting a sacrificial alkyl10 Grignard reagent,
MeMgCl, for the deprotonation of the alkynol 4.11 The rate and
reactivity profile for the subsequent Grignard carbometalation
reaction after employing methylmagnesium chloride (1.00 equiv)
as a sacrificial base matched what was seen previously with no
new byproducts observed. This modification allowed us to
reduce the charge of our more valuable 4-thiomethylphenyl-
magnesium chloride 5 from the previous 3.2 equiv routinely
reported in the literature to only 1.1 equiv and resulted in a
much cleaner reaction.12
(11) Initially, i-PrMgCl was used as the sacrificial base, but an excess
charge of this reagent ultimately led to the generation of a mixed anhydride
and the difficult to reject isobutyl ester impurity upon reaction with dianion
15 (see below). This problem was mitigated by substituting methylmag-
nesium chloride as the sacrificial Grignard reagent. In this case, excess
MeMgCl generated acetate upon reaction with carbon dioxide and subse-
quent reaction with acetic anhydride would simply regenerate acetic
anhydride, which would still allow formation of the desired (Z)-4-alkoxybut-
2-enoic acid 3 (see below).
After investigating the nature of the Grignard reagent, the
effects of other variables in the carbometalation reaction were
examined. Fallis routinely reports the use of a mixed-solvent
system requiring a nonpolar solvent such as cyclohexane or
toluene used in conjunction with THF (Scheme 3).13 In
accordance with the literature procedure, initial studies of the
carbometalation reaction were run in a two-solvent system
(cyclohexane/THF). However, it has been shown that in limited
cases THF can work as the sole solvent.14 Following this lead,
(12) The amount of thioanisole 10 and thioanisic acid 11 byproducts
has been reduced significantly by employing an alkylmagnesium chloride
reagent (MeMgCl) as a sacrificial base. The only byproduct now is methane
gas.
(13) Previous investigations have suggested that solvent plays an
important role in these carbometalation reactions. Nonpolar solvents such
as cyclohexane were essential for the direct condensation of the magnesium
chelate intermediate with aldehydes. Toluene/THF mixtures (1:1) have been
shown to provide significant improvements in some reactions.
(14) Forgione, P.; Fallis, A. G. Tetrahedron Lett. 2000, 41, 11-15.
(9) This method has the advantage of being indigenous to the process
and introduces no new chemical entities to the system. Tilstam, U.;
Weinmann, H. Org. Process Res. DeV. 2002, 6, 906-910.
(10) It has been shown that in the absence of copper iodide (CuI) alkyl
Grignard reagents do not readily participate in the carbometalation reaction.
See the following: (a) Tessier, P. E.; Penwell, A. J.; Souza, F. E. S.; Fallis,
A. G. Org. Lett. 2003, 5, 2989-2992. (b) Douboudin, J. G.; Jousseaume,
B.; Saux, A. J. Organomet. Chem. 1979, 168, 1-11.
J. Org. Chem, Vol. 71, No. 2, 2006 483

Engelhardt et al.
SCHEME 5. Unoptimized Acetate Trapping of an Alkoxide Intermediate
steps of the reaction sequence: (1) installation of the carboxylic
acid moiety and (2) trapping of the alkoxide intermediate
(Scheme 2). Using the optimized conditions described above,
we generated vinyl Grignard intermediate 9 and then treated it
with carbon dioxide16 to afford the (Z)-4-hydroxybut-2-enoic
acid17 in high yield. The final step in the one-pot reaction
sequence, the interception of the proposed reactive intermediate
carboxylate/alkoxide dianion 1518 via trapping with an acety-
lating reagent, proved the most challenging to understand. A
variety of acetylating agents including AcCl, Ac2O, AcCN,19
and AcBr were investigated, and the best results were achieved
using Ac2O. Adding excess acetic anhydride (2.0 equiv) to
carboxylate/alkoxide 15 at 30 °C afforded the desired acid 3 in
∼50% yield along with ∼25% of butenolide 13 (Scheme 5).
Initial efforts20 to improve the yield of 3 while minimizing 13
were unsuccessful. Over the course of these early studies, the
overall yield of the reaction (product 3 plus lactone 13) was
consistently high (70-90%), independent of any alterations in
acylation conditions, and the ratio of product 3 to lactone 13
was typically 2:1.21 Lactone 13 most likely results from
inefficient trapping of dianion intermediate 15; however, we
were concerned that it may form competitively with 3 under
the acetylation reaction conditions. Reaction monitoring using
ReactIR indicated that under the reaction conditions no buteno-
lide22 13 was formed prior to or during the acetic anhydride
quench (Figure 2), and only upon protic workup was 13
observed (25% yield).23
In addition to monitoring lactone 13, the ReactIR revealed
that a significant amount of “free” carbon dioxide (2330-2350
cm-1) remained dissolved in the THF (ca. 0.25 M at 25 °C)
after all of vinyl Grignard intermediate 9 was consumed. Further
observation of the carbon dioxide signal showed that over time
the concentration of free CO2 in solution slowly declined (Figure
2) but could not be accounted for in the headspace of the
reaction, indicating that the gas was reacting in solution. We
were also able to show that increasing the age after introduction
of carbon dioxide had a detrimental effect on the overall yield
of the reaction and reduced the ratio of product 3 to lactone 13
after trapping with acetic anhydride.13 Taking all of this
information into account, we proposed the following: excess
free CO2 in the reaction system may be slowly reacting with
the alkoxide anion of 15 to form a carbonate species 16, in
essence protecting the allylic alkoxide from subsequent acety-
lation. Subsequent protic workup after Ac2O introduction
releases carbon dioxide, and the resulting (Z)-4-hydroxybut-2-
enoic acid cyclizes rapidly to provide butenolide 13 (Chart 1).24
Variation of the temperature, concentration, and solvent led
to only a slight change in the ratio of product to lactone, so
attention was turned instead to the introduction of additives to
the reaction system, primarily with the intent of removing the
carbon dioxide from the system in an effort to prevent formation
of carbonate 16. It was the addition of alkoxide bases (0.5 equiv)
such as NaOEt or KOtBu that gave encouraging results, as the
percent of lactone 13 in the reaction product mixture was
reduced from the typical 30-35% to 15 and 11%, respectively.
Again, in situ ReactIR experiments were conducted, this time,
to understand the critical role potassium tert-butoxide plays in
(15) As assayed by quenching into methanol and quantifying trisubstituted
alkene 12a.
(16) The CO2 quench is best performed by bubbling excess CO2 through
the reaction mixture while maintaining a slightly elevated temperature range
(30-40 °C) to afford a reasonable reaction rate. Although lower temper-
atures facilitate carbon dioxide solubility in THF, incorporation of the carbon
dioxide moiety was significantly reduced.
(17) (Z)-4-hydroxybut-2-enoic acid 7 cannot be isolated, and it cyclizes
upon protic workup to give butenolide 13. The presence of butenolide 13
serves as a viable assay for the presence of (Z)-4-hydroxybut-2-enoic acid
in this step.
(18) ReactIR confirmed that butenolide 13 (CdO: 1758 cm-1) is not
present under the reaction conditions and is generated only upon protic
workup. For reference, the carbonyl stretch for the opened carboxylate/
alkoxide 15 is much broader (1650-1500 cm-1).
(19) (a) Cain, B. R. J. Org. Chem. 1976, 41, 2029-2031. (b) Holy, A.;
Soucek, M. Tetrahedron Lett. 1971, 2, 185-188.
(20) The reaction at the acetate quench stage was evaluated in different
solvent systems (THF, MeCN, toluene, and DMF). The effect of temperature
on the reaction was also evaluated (0 °C, 22 °C, and 100 °C). Finally,
additives such as bases (nBuLi, KHMDS, CaH2, and Hunig’s Base) and
Lewis Acid salts (CaCl2, LiCl, LiBr, KBr, KCl, MgBr2, NaI, and ZnCl2)
were added to try to alter the ratio of desired acid 3 to lactone 13.
Encouragingly, variability in the ratio of 3/13 was observed under all of
these various reaction conditions; unfortunately, the best ratio observed was
2:1, and thus, 30-35% of the product mixture was being lost as the lactone
13 byproduct.
(22) Butenolide 13 slowly grew (1758 cm-1) but amounted to only a
0.7% yield over the course of 14 h in the presence of carbon dioxide.
(23) The final amount of butenolide 13 observed at the end of the reaction
increased in direct proportion to the length of the carbon dioxide age prior
to the acetylation step. Longer carbon dioxide exposures (i.e., >3 h) had a
detrimental effect on the ratio of acid 3 to lactone 13.
(24) Gas evolution has been observed upon the addition of acetic
anhydride on larger scales supporting the notion of the carbonate protecting
group. Longer carbon dioxide exposures prior to the acetate trapping step
resulted in an increase in the observed butenolide 13, presumably due to a
larger concentration of carbonate 16. Also, free carbon dioxide is observed
after acetic anhydride is added to the system.
(21) The ratio of desired acid 3/butenolide 13 was 2:1. Efforts to increase
this ratio were unsuccessful.
484 J. Org. Chem., Vol. 71, No. 2, 2006

Synthesis of a NO-Releasing Prodrug of Rofecoxib
CHART 1. IR Showing Carbonate Formation
SCHEME 6. Increased Product to Lactone Ratio in the Absence of Carbon Dioxide
TABLE 1. Carbon Dioxide Effect on Product to Lactone Ratio
suppressing the formation of butenolide 13.25 It is believed that
potassium tert-butoxide acts as a carbon dioxide scavenger
sequestering CO2 as a tert-butyl carbonate species.26 An
experiment was designed to access both the carboxylate 17 and
the alkoxide 15 in a controlled manner from the sulfide acid 3
while monitoring the intermediate species with the ReactIR
(Scheme 6). Adding isopropylmagnesium chloride (1 equiv) to
the sulfide acid 3 cleanly afforded carboxylate 17, and subse-
quent addition of 2 equiv of methylmagnesium chloride gener-
ated the dianion alkoxide 15 with an end of reaction observed
after 45 min (the reaction could be monitored by observing the
(3/13)
entry
1
conditions
ratio 3/13
yield (%)
63
(a) CO2 (30 min, room temperature)
(b) Ac2O (excess, room temperature)
(a) CO2 (30 min, room temperature)
(b) KOtBu (0.50 equiv)
(c) Ac2O (excess, room temperature)
(a) CO2 (30 min, room temperature)
(b) KOtBu (0.50 equiv)
2.5:1
2
3
14:1
86
65
(25) Spectral information garnered from the ReactIR experiment de-
scribed in Scheme 6. Each compound has a unique infrared CdO stretching
frequency that can be readily distinguished from other reactive intermediates.
Note that the carbonyl of both the acid and the acetate protecting group
can be used to identify each intermediate.
3.5:1
(c) CO2 (30 min, room temperature)
(d) Ac2O (excess, room temperature)
decay of the acetate signal at 1740 cm-1). Finally, quenching
with an excess of acetic anhydride cleanly afforded carboxylate
17 with no IR eVidence of lactone 13. Protic workup and HPLC
analysis confirmed a ratio of >39:1 (3/13) for this reaction in
85% yield. This experiment verified that acetate/acid 3 could
be obtained in high yield by intercepting dianion intermediate
15 using an acetate trap.
The experiments in Table 1 were designed to generate the
same intermediate, alkoxide 15, and to probe the effect residual
carbon dioxide has on the ratio of product 3/lactone 13. Entry
1 showed that exposing alkoxide 15 to carbon dioxide prior to
the acetate quench significantly impacted the yield (3 + 13) of
the reaction (63%) and reduced the ratio of product to lactone
(2.5:1). Entry 2, on the other hand, went a step further and
attempted to negate the detrimental effect of carbon dioxide
exposure by “scrubbing” the reaction solution with potassium
(26) All other attempts to remove the carbon dioxide from the reaction
medium, including purging, sparging, and thermal techniques, have failed
to give favorable results.
J. Org. Chem, Vol. 71, No. 2, 2006 485

Engelhardt et al.
SCHEME 7. Optimized Carbometalation Step
CHART 2. IR Response of CO2 at 2350 cm-1
isolated as a pure28 crystalline solid by an in situ preparation of
the corresponding magnesium salt 18 (Scheme 7). This was
accomplished in practice by first buffering the crude reaction
mixture with KOH29 after the acetic anhydride quench and then
washing with an aqueous magnesium chloride solution.30 Upon
separating the layers and adding an antisolvent (i.e., MTBE,
EtOAc, and IPAc) to the THF layer, a solid crystallized. Pursuit
of this isolation procedure led to the optimized protocol where
water31 was added and isopropyl acetate (2 volumes) was used
as the antisolvent for crystallization. The highly productive
carbometalation reaction fully optimized now allowed for the
isolation of >99% pure magnesium salt as a hydrate in high
yield from readily available starting materials. This one-pot
sequence of reactions constructs the entire carbon backbone of
our targeted NO-COXIB, leaving only a sulfide to sulfone
oxidation and nitrate side chain incorporation (Scheme 2).
The entire optimized process for the carbometalation proce-
dure includes the following: (1) a one-pot, single solvent (THF)
process; (2) MeMgCl as a sacrificial base; (3) 4-thiomethyl-
phenylmagnesium chloride (1.10 equiv); (4) Ac2O as the
acetylating agent; (5) KOtBu (0.5 equiv) as an additive to “scrub
out” excess CO2; (6) a pH adjust using KOH; and (7) formation
and isolation as the hydrated magnesium salt 18. The optimum
conditions routinely provided 18 in 85% isolated yield and 99%
purity (Scheme 7).
tert-butoxide. During the experiment, the carbon dioxide peak
is observable at 2339 cm-1 and corresponds to free gas in the
THF solution. The addition of potassium tert-butoxide resulted
in an immediate and sharp decrease in the concentration of CO2
(see Chart 2). At this point, acetic anhydride was added and
the reaction was quenched. The overall yield of the reaction
under these conditions rebounded to 86%, and the ratio of
product 3 to lactone 13 increased (14:1). The observed ratio
and shown decrease in free CO2 in solution due to the addition
of potassium tert-butoxide support the conjecture that this
additive acts as a carbon dioxide scavenger. This effectively
“deprotects” alkoxide 15 allowing for efficient acetate quenching
to afford the desired trapped intermediate 17. The last experi-
ment in the table, entry 3, examines whether the detrimental
effect of carbon dioxide returns when CO2 is reintroduced after
scrubbing the reaction with KOtBu. A significant amount of
butenolide 13 was generated under these conditions again with
a lower yield (65%). This implies that excess carbon dioxide is
detrimental to the success of the reaction, and adding the
appropriate amount27 of potassium tert-butoxide is critical. Upon
the introduction of the mildly acidic acetic anhydride, the tert-
butoxide sequestered CO2 is rendered unstable and is released
into solution (see Chart 2), further confirming that there is a
reservoir of carbon dioxide covalently bound as a carbonate
species.
3. Salt Break and Oxidation. Although direct oxidation of
magnesium salt 18 was possible, the oxidation step was plagued
with low yields and erratic results due to the presence of
magnesium32 (2.5 wt %) and chloride.33 A salt break step was
introduced to convert magnesium salt 18 to the free acid 3 and
to remove these inorganic impurities. The salt break was easily
achieved by dissolving the magnesium salt 18 in DMF and
adding to it a 2 M aqueous solution of acetic acid at 40 °C
(Scheme 8). Free acid 3 prepared by this procedure contained
<100 ppm of metals and chloride and was typically >99 LCAP.
(28) Typical purity was >98.5% by HPLC analysis at 210 nm.
(29) Mother liquor losses during the isolation of the magnesium salt had
been variable and high (6-25%). Work revealed that the liquor losses were
closely tied to the pH of the crystallization solution. Potassium hydroxide
(aqueous, 30-45 wt %) was added to the crude reaction to buffer the pH
of the solution in the range of 6.4-7.2, as acetic acid was generated after
acetate trapping with Ac2O. This resulted in reproducible mother liquor
losses of 5-7% (∼3 mg/mL). One equivalent of KOH proved optimum as
larger charges resulted in a pH >7.4 which is basic enough to generate
insoluble Mg(OH)2 (upon introduction of the MgCl2 solution), causing thick
emulsions. KOH was selected as the base because of its low cost, and having
potassium as the counterion introduces no new metal species into the reaction
system that could interfere with our isolation of the desired magnesium
salt product.
(30) The magnesium chloride (1.4 M) solution is prepared using MgCl2‚
6H2O. (Caution! Do not use MgCl2 to prepare an aqueous solution of
MgCl2, as introduction of water is potentially explosiVe because of the
extremely high exothermic heat of solVation for this salt.) The MgCl2
solution serves mainly as an aqueous wash for the THF layer, and the high
molarity of the solution allows facile separation between the aqueous and
THF (organic) layers.
Having achieved high solution assay yields for acid 3, we
addressed the next issue of how to isolate the sulfide acid in
the presence of numerous byproducts including thioanisic acid
and butenolide 13 (3-6%) without resorting to chromatography.
We were fortunate to discover that acid/acetate 3 could be
(27) The amount of KOtBu added to the carbometalation reaction is
optimized on the basis of the volume of THF in the reaction and the
solubility of CO2 in THF at ambient temperature (ca. 0.25 M). Calculations
based on the volume of THF and the solubility of carbon dioxide provided
the insight to the appropriate amount of KOtBu (0.5 equiv) that needs to
be added.
(31) Water (3-5% volume) was critical to the success of the crystal-
lization, as the isolated magnesium salt is a highly hydrated species requiring
8-10 water molecules per mole of salt.
486 J. Org. Chem., Vol. 71, No. 2, 2006

Synthesis of a NO-Releasing Prodrug of Rofecoxib
SCHEME 8. Salt Break, Oxidation, and Initial End-Game
Many methods were examined for the oxidation of the sulfide
acid 3 to the corresponding methyl sulfone 2, including catalytic
tungsten/H2O2, mCPBA, Oxone, MMPP, sodium perborate, and
H2O2/acetic acid. The main problems with these methods
included formation of rofecoxib 6, isomerization of the double
bond, oxidative cleavage of the double bond, and/or difficulty
in isolating the sulfone acid 2 from the byproducts of the
oxidation method. Hydrogen peroxide in acetic acid provided
the most suitable oxidation conditions due to cost effectiveness,
the ability to control the rate of reaction, and ease of product
isolation. Hydrogen peroxide in acetic acid is regularly used as
an oxidation system for the preparation of sulfones.34 For large-
scale reactions, we routinely avoid this system because of the
possible generation of peracetic acid, a dangerous and potentially
explosive compound. Using a ReactIR probe, we examined the
reaction, which indicated that peracetic acid was not being
produced under the reaction conditions (hydrogen peroxide in
acetic acid at 40 °C). This process has now been demonstrated
on a 50 kg scale to provide methyl sulfone 2 in >95% yield
from sulfide acid 3 (Scheme 8).
To complete the synthesis, incorporation of the side chain
remained. The initially reported procedure35 coupled sulfone acid
2 with excess dibromohexane 22 to afford alkyl bromide 23.
Then, the final product was obtained by treating the primary
bromide with silver nitrate36 to afford NO-COXIB (1) (Scheme
8). A large excess of alkylating reagent and stoichiometric silver
salt was used in an effort to minimize the formation of dimeric
compound 24. Unfortunately, despite the use of a 20-fold excess
of 1,6-dibromohexane, impurity 24 was still observed (∼4%)
which necessitated extensive purification to remove this impurity
from the final product. Two alternate end-game approaches were
examined to avoid these complications and remove any exposure
of the final compound to stoichiometric amounts of silver (vide
infra).
End-Game Strategy. With the core of the molecule in hand,
two alternate strategies for the final coupling of the side chain
were explored. The first is a convergent approach (route A)
where nitrate side chain 26 was prepared from 1,6-bromohexanol
25 and then coupled to sulfone acid 2 (Figure 3). This route,
although initially attractive, suffered from the generation of the
same difficult to remove impurities. The second strategy (route
B) involves direct coupling of 1,6-bromohexanol 25 with acid
2 followed by nitration to yield prodrug 1. Both coupling
strategies will be discussed in detail below.
(32) Under the harsh reaction conditions, 30% aqueous H2O2 (5 equiv)
in AcOH at 40 °C, epoxidation of the Z olefin is possible in the presence
of a metal catalyst such as magnesium. Hydrolysis of the formed oxirane
followed by further oxidation can then lead to olefin cleavage products.
Convergent Strategy. The convergent synthesis of 1, via
route A, hinged upon the successful preparation of nitrate side
chain 26 from 1,6-bromohexanol 25 and the subsequent
coupling. To fully exploit the advantages of the convergent
approach, a safe and economical synthesis of 6-bromohexyl
nitrate 26 was needed. Additionally, as this compound is an
oil, no purification of the product could be easily accomplished
before coupling to give NO-COXIB (1), and hence, we needed
to produce material as cleanly as possible. We initially used a
H2SO4/HNO3 nitration of 6-bromohexanol 26 under neat reac-
tion conditions.37 Although successful, obvious safety concerns
emerged in response to this procedure. Alternatively, we turned
to the use of an in situ prepared nitrating reagent, acetyl nitrate
(35) Internal communication with Merck Medicinal Chemistry.
(36) Reaction of an alkyl halide with silver nitrate to give a nitrate
ester: Boschan, R.; Merrow, R. T.; van Dolah, R. W. Chem. ReV. 1955,
55, 485-510.
(37) Olah, G. A.; Malhotra, R.; Narang, S. C. Nitration: Methods and
Mechanisms; VCH: New York, 1989 .
(33) Residual chloride (>200 ppm) in the sulfide acid salt led to the
formation of chlorosulfone.
(34) Oxidation of aryl sulfides to aryl sulfones using hydrogen peroxide
in acetic acid: Russell, G. A.; Pecoraro, J. M. J. Org. Chem. 1979, 44,
3990-3991.
J. Org. Chem, Vol. 71, No. 2, 2006 487

Engelhardt et al.
FIGURE 3. End-game strategy.
SCHEME 9. Preparation of 6-Bromohexyl Nitrate
SCHEME 10. Convergent Coupling
28 (AcONO2),38 formed from nitric acid (90%) and acetic
anhydride39 (Scheme 9). With an excess of acetyl nitrate 28 in
the nitration step, the reaction cleanly afforded 6-bromohexyl
nitrate 26 (Scheme 9), ready for use in the coupling step.
The coupling of sulfone acid 2 with prepared 6-bromohexyl
nitrate 26 (2 equiv) was carried out using potassium carbonate
(1.0 equiv) in DMF (10 volumes) at 20-22 °C. The reaction
provided nitrate 1 after 2-3 h in 93% yield. Unfortunately,
product 1 was also accompanied by two impurities, alkyl
bromide 23 (∼2%) and dimer 24 (∼2%), both of which were
not effectively rejected during a recrystallization of 1 (Scheme
10). Of major concern was the toxicity risk associated with
having any bromide impurity 23 present in the final drug
substance 1 as it is a potential active alkylating agent. Therefore,
in practice, a silver nitrate treatment was implemented to convert
bromide 23 into product 1 and silica chromatography was then
used to remove the dimer 24.
Although the coupling step initially appeared straightforward
and high yielding (93%), the two impurities produced, bromide
23 and dimer 24, cast a shadow on this process. Excellent
chemoselectivity was observed in the coupling reaction, as the
selectivity for bromide displacement by the carboxylate anion
vs nitrate displacement was 124:1 (Scheme 11). Unfortunately,
it is the combination of several unfavorable side reactions that
leads to significant amounts of these two key impurities, bromide
23 and dimer 24 (vide infra).40
The origin of impurities 23 and 24 is relevant to any new
synthesis, and their formation was shown to stem from multiple
sources. Bromide impurity 23 emerges via two potential
(38) The generation of nitrating reagent AcONO2 using acetic anhydride
and nitric acid and its use in the nitration of an alcohol: Black; Babers
Organic Syntheses, 1939; Vol. 19, pp 64-66.
(39) We have attempted to replace dichloromethane (DCM) with a more
environmentally benign solvent, but we observed either overnitration or
recovered starting material. So far, solvents tested include THF, DME, IPAc,
chlorobenzene, MTBE, Ac2O, CH3CO2H, and MeCN.
(40) We have not been able to reject either of these impurities by
recrystallization without considerable yield losses of the desired final
product.
488 J. Org. Chem., Vol. 71, No. 2, 2006

Synthesis of a NO-Releasing Prodrug of Rofecoxib
SCHEME 11. Reaction Pathways in the Coupling of Sulfone Acid and 6-Bromohexyl Nitrate
pathways; see Scheme 11. During the alkylation step, the nitrate
group (-ONO2) of 26 is susceptible to displacement by the
carboxylate of sulfone acid 2 (Scheme 11), thus affording
bromide 23. Additionally, the nitrate group is also susceptible
to displacement by a bromide ion (KBr) generated in situ from
the coupling; thus, 1 can be converted directly to bromide 23.
Similarly, the formation of the dimer impurity 24 can be
rationalized (Scheme 11). The carboxylate of sulfone acid 2
can slowly displace the nitrate group of 1 during the alkylation
reaction leading directly to dimer 24. The primary bromide of
any formed compound 23 can also be displaced by an unreacted
carboxylate of sulfone acid 2, leading to additional dimer 24.
Alternative End-Game. The convergent route to NO-
COXIB (1) incorporated the coupling of sulfone acid 2 with
6-bromohexyl nitrate 26 (Scheme 10); unfortunately, this
reaction was complicated by side reactions (see Scheme 11)
leading to two difficult to remove impurities, bromide 23 and
dimer 24. We sought a more controlled route to NO-COXIB
(1) that would minimize formation of these impurities. We
anticipated that coupling of sulfone acid 2 with 6-bromohexanol
25 would afford alcohol 27, and without the presence of an
electrophilic nitrate leaving group, bromide 23 and dimer 24
should not be generated. Two concerns emerged with this
synthetic approach: (1) the potential instability of 6-bromo-
hexanol under the basic coupling conditions, and (2) the
chemoselectivity of the critical nitration of the primary alcohol
in the presence of various sensitive functional groups including
two aromatic rings. In practice, 1.10 equiv of 6-bromohexanol
25 coupled with sulfone acid 2 at 40 °C yielded alcohol 27 in
4-5 h (∼92% isolated yield) with no bromide impurity 23 and
only small amounts of dimer 24 observed (<0.3% by HPLC
analysis), resulting from the presence of 1,6-dibromohexane 22
as an impurity in commercial 6-bromohexanol 25.
The nitration of alcohol 27 assumes greater importance in
the newly improved synthesis as it becomes the final step. We
initially found that competitive nitration of the phenyl group
(compound 30 in Scheme 12) was problematic, but this was
mitigated if the nitrating mixture (HNO3/Ac2O) was added to
the substrate and if the temperature was controlled (-10 to -15
°C). In this case, the nitration leading to over-nitrated compound
30 was limited to <0.5% (by HPLC analysis) and this level
was reduced to <0.1% upon crystallization to give 1.41
As acetyl nitrate is documented42 to be explosive at temper-
atures above 50 °C, we were concerned about its long-term use
in our process. We thought that increasing the carbon content
of the reagent would lead to a safer system and prevent the
formation of the toxic tetranitromethane. Various anhydrides
were examined (propionic, isobutyric, n-butyric, valeric, and
hexanoic), and n-butyric anhydride was deemed optimal because
(41) Acetylation to give 29 was observed when the alcohol 27 was added
rapidly to the nitrating mixture (HNO3/Ac2O), and as long as the addition
time was extended to over 30 min, acetylation was completely suppressed.
(42) Butler, A. R.; Hendry, J. B. J. Chem. Soc. B: Phys. Org. 1971,
102-105.
J. Org. Chem, Vol. 71, No. 2, 2006 489

Engelhardt et al.
SCHEME 12. Coupling and Nitration: New and Improved Route
Then, 250 mL of potassium tert-butoxide (1 M in THF, 0.50 equiv)
was charged to the homogeneous reaction mixture. The batch was
stirred at 32 °C for 1 h. Finally, the batch temperature was adjusted
to 23 °C and 94.5 mL of acetic anhydride (2 equiv) was added
while maintaining a temperature of 30-45 °C. The batch was stirred
for 1.5 h and then cooled to 10 °C. Then, the reaction was quenched
with 93.5 mL of 30 wt % potassium hydroxide (28.1 g KOH, 1.0
equiv) and diluted with a deionized water flush (300 mL). The batch
was aged at 40 °C for 7 h.
Next, 150 mL of an aqueous 2 M magnesium chloride solution
was added and the batch was stirred for 15 min at 40 °C with strong
agitation. Agitation was then ceased, and the two layers were
allowed to settle and were separated. To the organic layer was added
100 mL of deionized water, and finally, 2.3 L of isopropyl acetate
was added slowly over 2 h at 20 °C to complete the crystallization
of the batch.
The batch was then cooled to 0 °C, and the slurry was filtered
and washed with 500 mL of deionized water and 500 mL of cold
(0 °C) isopropyl acetate. Filtration yielded 160.3 g of the desired
crystalline hydrated magnesium salt product 18 (73.8% yield).
Because of the various hydrate forms of the magnesium salt 18,
this compound was recrystallized as the free acid 3 in the subsequent
step prior to full characterization (see below).
Salt Break. Carboxylic Acid (3). A solution of the magnesium
salt 18 (2.63 kg corrected, 7.31 mol) in DMF (8 L) was slowly
added to aqueous acetic acid (26 L, 2 M, 56 mol) at 35-40 °C.
The precipitated free acid 3 was isolated by filtration, and the wet
cake was washed with 20% aqueous DMF (6.6 L) and then twice
with water (6.6 L). The product was dried at 40 °C under vacuum
to yield 2.4 kg of the desired acid 3 as a tan crystalline solid (96%
yield, mp ) 149-150 °C). 1H NMR (400 MHz, CDCl3): δ (ppm)
10.87 (bs, 1H), 7.18-7.15 (m, 3H), 7.12-7.08 (m, 2H), 7.04-
6.99 (m, 4H), 5.26 (s, 2H), 2.42 (s, 3H), 1.96 (s, 3H). 13C NMR
(100 MHz, CDCl3): δ (ppm) 173.5, 171.2, 143.6, 138.8, 136.1,
134.4, 133.9, 130.0, 129.8, 128.3, 127.9, 125.7, 65.3, 20.8, 15.4.
FTIR: (cm-1) 3176, 3044, 2927, 1736, 1667, 1598, 1491, 1445,
1369, 1231, 1175, 1092, 1037, 815, 718, 706. HRMS calcd for
C19H18O4S (M - H) m/z 341.0848, found 341.0846.
of its ease of workup, its reasonable rejection of any butyroyl
ester formed, and the anticipated safety improvement (Scheme
12). With this modified procedure, the yield of isolated crude
NO-COXIB (1) is generally 92-95% with excellent purity
(>98%) as well.
Conclusions
Traditionally, tetrasubstituted alkenes with defined stereo-
chemistries represent a challenging synthetic motif in organic
chemistry. In this paper, we have demonstrated a highly efficient
four-step one-pot carbometalation reaction that constructs a
tetrasubstituted alkene with remarkable stereocontrol. The
Grignard carbometalation reaction transforms a simple propargyl
alcohol 4 into (Z)-2,3-biaryl-4-acetoxybut-2-enoic acid 3, the
core of prodrug 1. Previously, this type of transformation has
been used to prepare butenolides,6 but we have shown that the
intermediate (Z)-4-alkoxybut-2-ene carboxylate intermediate
(Figure 2) can be intercepted with an appropriate acetate trap
(Ac2O). The optimization of this four-step process required an
intricate understanding of the reaction intermediates which was
obtained by using ReactIR. The sequestering of excess carbon
dioxide by introducing an alkoxide base (potassium tert-
butoxide) was instrumental in the success of the subsequent
acetate trapping reaction to generate the opened product 3.
Finally, the synthesis of the NO-COXIB drug was completed
via two different end-game strategies designed to control
impurity and byproduct levels.
Experimental Section
Grignard Carbometalation. Magnesium Salt (18). A flask is
charged with 300 mL of THF, and the vessel is flushed with
nitrogen. This was followed by the addition of 66.1 g of 3-phenyl-
2-propyn-1-ol (4). The batch was then cooled to approximately 5
°C, and 175 mL of methylmagnesium chloride (3 M, 1.05 equiv)
was added slowly over 30 min, achieving a final batch temperature
between 25 and 30 °C. Then, 268 mL of 4-thioanisolemagnesium
chloride was added (2.05 M in THF, 1.10 equiv), and the batch
was heated to reflux (65-70 °C). The batch was stirred at this
temperature for 3 h and then was cooled to approximately 18 °C.
Carbon dioxide (dry) was then charged subsurface, and the reaction
was aged at 30 to 35 °C for 2 h.
Protonated Intermediate (12a). This intermediate 12a was
recrystallized from hexanes (mp ) 88-90 °C). 1H NMR (400 MHz,
CDCl3): δ (ppm) 7.23-7.04 (m, 9H), 6.70 (s, 1H), 4.46 (d, J )
1.5, 2H), 2.50 (s, 3H), 1.68 (s, -OH, 1H). 13C NMR (100 MHz,
CDCl3): δ (ppm) 141.0, 138.1, 136.6, 135.3, 129.5, 129.4, 128.2,
127.1, 127.0, 126.9, 68.7, 15.8. FTIR: (cm-1) 3224, 3079, 3044,
3017, 2913, 2858, 1598, 1494, 1445, 1397, 1348, 1058, 815, 753,
7
The vessel pressure was then vented, and a series of purges were
completed to remove residual carbon dioxide in the headspace.
698, 670. HRMS calcd for C16H16OS (M + Li) m/z 263.1082,
found 263.1087.
490 J. Org. Chem., Vol. 71, No. 2, 2006

Synthesis of a NO-Releasing Prodrug of Rofecoxib
Lactone (13) Byproduct. This impurity, 13, was recrystallized
from MTBE/THF (2:1) for characterization (mp ) 144-146 °C).
1H NMR (400 MHz, CDCl3): δ (ppm) 7.45-7.37 (m, 5H), 7.24
(d, J ) 8.7, 2H), 7.17 (d, J ) 8.6, 2H), 5.17 (s, 2H), 2.49 (s, 3H).
13C NMR (100 MHz, CDCl3): δ (ppm) 173.7, 155.6, 143.0, 130.6,
129.5, 129.0, 128.9, 127.9, 127.0, 125.9, 125.6, 70.5, 15.1. FTIR:
(cm-1). HRMS calcd for C17H14O2S (M + H) m/z 283.0793, found
283.0789.
The reaction mixture was cooled to ∼20 °C, and IPAc (26 L)
was introduced. Then, ice-cold water (19 L) was added slowly to
maintain the temperature at <30 °C. The mixture was stirred for
0.5 h and settled to give two clear layers. The aqueous layer was
separated and back-extracted with IPAc (19 L). The combined
organic layer was washed with water (2 × 19 L). The organic layer
was concentrated in vacuo to ∼13 L and flushed with 13 L of IPAc.
The resulting solution’s concentration was adjusted to 170-180
mg/mL (∼15 L, KF <200 µg/mL). To this solution was added 5.5
L of n-heptane at 20-24 °C, followed by addition of ∼25 g of the
seed (∼1 wt % based on 95% yield), and the solution was stirred
for 1-2 h to provide a good seed bed (supernatant ∼50 mg/mL) at
18-20 °C. The remaining n-heptane (16.5 L) was introduced over
1-2 h and then aged for an additional 8 h. The slurry was cooled
to -5 to 0 °C to reduce the supernatant concentration to <1.5 mg/
mL. It was then filtered, and the cake was washed with cold
premixed IPAc/n-heptane (1:4, 8 L) and air-dried at 22 °C under
nitrogen for 12 h. The isolated crystalline solid (2.58 kg, 95 wt %)
was obtained in 90% yield (mp ) 59-62 °C). 1H NMR (400 MHz,
CDCl3): δ (ppm) 7.70 (d, J ) 8.5, 2H), 7.29 (d, J ) 8.5, 2H),
7.12-7.07 (m, 3H), 7.01-6.97 (m, 2H), 5.11 (s, 2H), 4.20 (t, J )
6.6, 2H), 3.55 (t, J ) 6.5, 2H), 2.96 (s, 3H), 1.98 (bs, -OH, 1H),
1.92 (s, 3H), 1.64 (quintet, J ) 6.8, 2H), 1.49 (quintet, J ) 6.8,
2H), 1.30 (m, 4H). 13C NMR (100 MHz, CDCl3): δ (ppm) 170.5,
168.0, 143.7, 139.4, 139.0, 137.6, 135.1, 130.3, 129.4, 128.3, 128.1,
127.1, 65.7, 64.7, 62.6, 44.4, 32.5, 28.4, 25.7, 25.3, 20.7. FTIR:
(cm-1) 3543, 3058, 2934, 2864, 1743, 1722, 1598, 1445, 1376,
1314, 1238, 1141, 1092, 1078, 961, 850, 753, 732, 704. HRMS
calcd for C25H30O7S (M - H) m/z 473.1634, found 473.1636.
Nitration: Synthesis of NO-COXIB (1). Nitric acid (HNO3)
(344.6 mL, 7.33 mol) was added over 20 min to a cooled solution
of n-butyric anhydride (1.375 kg, 8.69 mol) in dichloromethane
(10 L) with the internal temperature remaining below 5 °C. After
stirring for 2 h at 0 °C, the solution, to be cooled to -15 °C, and
a solution of alcohol 27 (2.20 kg, 4.64 mol) in dichloromethane
(7.3 L) was added over 30 min, maintaining the temperature below
-10 °C. The reaction was stirred at -15 °C for 30 min. The reaction
was quenched by addition of K3PO4 solution (8 L of 2 M aqueous
solution). Then, toluene (10 L) was added and the layers were
separated. The organic layer was washed with aqueous urea (20 L
of 0.5%) and then solvent was switched to toluene (24 L final
volume). The resulting toluene solution was filtered through a plug
of silica gel (3.72 kg) with toluene/EtOAc elution (50 L of 90:10
mixture). Concentration of the eluent to 20 L volume was followed
by addition of heptane (2 L) at 35 °C to obtain a seed bed. Further
addition of heptane (17 L) was made, and filtration gave crude 1.
This material was recrystallized from toluene/heptane to give pure
1 as a white crystalline solid (2.05 kg, 90% yield, mp ) 79-80
°C). 1H NMR (400 MHz, CDCl3): δ (ppm) 7.72 (d, J ) 8.4, 2H),
7.30 (d, J ) 8.4, 2H), 7.15-7.09 (m, 3H), 7.04-6.98 (m, 2H),
5.14 (s, 2H), 4.38 (t, J ) 6.6, 2H), 4.22 (t, J ) 6.5, 2H), 2.98 (s,
3H), 1.94 (s, 3H), 1.66 (quintet, J ) 6.7, 4H), 1.40-1.27 (m, 4H).
13C NMR (100 MHz, CDCl3): δ (ppm) 170.5, 168.0, 143.7, 139.6,
139.4, 137.5, 135.2, 130.4, 129.5, 128.3, 128.2, 127.2, 73.2, 65.5,
64.7, 44.4, 28.3, 26.7, 25.5, 25.3, 20.8. FTIR (cm-1): 2941, 2865,
1743, 1715, 1625, 1383, 1314, 1279, 1224, 1203, 1155, 1106, 1037,
961, 871, 705. HRMS calcd for C25H29NO9S (M - H) m/z
518.1485, found 518.1486.
Sulfide Oxidation. Sulfone (2). A mixture of acid 3 (2.56 kg,
7.30 mol) in acetic acid (24 L) was heated to 40 °C, and hydrogen
peroxide (3.26 L, 36.5 mol) was added over 15 min. After 6-8 h,
the reaction was diluted with water (48 L). The mixture was seeded,
and the temperature was held at 40 °C for 1 h and then allowed to
cool slowly to room temperature over 2 h. The batch was then
cooled further to -10 °C and held at this temperature for 1 h. The
product was isolated by filtration, washed with 7 L of water, and
dried under vacuum to afford sulfone acid 2 (2.46 kg, 89.8%) as a
1
white crystalline solid (mp ) 153-154 °C). H NMR (400 MHz,
CDCl3): δ (ppm) 9.73 (bs, COOH, 1H), 7.75 (d, J ) 8.6, 2H),
7.30 (d, J ) 8.7, 2H), 7.20-7.14 (m, 3H), 7.08-7.05 (m, 2H),
5.28 (s, 2H), 3.00 (s, 3H), 1.95 (s, 3H). 13C NMR (100 MHz,
CDCl3): δ (ppm) 172.2, 171.0, 143.5, 142.3, 139.7, 136.2, 134.9,
130.3, 129.8, 128.5 (2 peaks), 127.2, 64.9, 44.5, 20.8. FTIR: (cm-1
)
3190, 3079, 3024, 2927, 1736, 1598, 1494, 1445, 1307, 1231, 1155,
1099, 1085, 968, 781, 767, 663. HRMS calcd for C19H18O6S
(M - H) m/z 373.0746, found 373.0747.
End-Game Strategy 1: Convergent Approach. A. Bromoni-
trate Preparation (26). Nitric acid (90% w/w) (1.45 kg, 20.7 mol)
was added over 1 h to a solution of acetic anhydride (2.53 kg, 24.8
mol) in dichloromethane (20 L) maintained at -10 °C. This mixture
was then stirred at 0 °C for 1 h before a solution of 6-bromohexanol
25 (2.50 kg, 13.8 mol) in dichloromethane (20 L) was added over
1.5 h, maintaining the temperature below 0 °C. The reaction was
stirred for 30 min, quenched into K2HPO4 solution (10 L of 1 M),
and stirred for 14 h before the layers were separated. The organic
layers were washed with urea solution (5 L of 10% w/w solution),
water (20 L), and brine (10 L of saturated aqueous). The organic
solution was then concentrated to afford the crude 6-bromohexyl
nitrate 26 (3.19 kg, 100 wt %, quant) as a colorless oil.
B. Coupling of Nitrate to Give NO-COXIB (1). To a 100 L
flask was charged 20 L of DMF, solid sulfone acid 2 (2.72 kg,
6.97 mol), liquid bromonitrate 26 (3.99 kg, 17.2 mol), and 4.4 L
of DMF to give a clear solution at room temperature. To this
resulting solution was added powder K2CO3 (0.98 kg, 7.09 mol)
in one portion at 20 °C, followed by 2.0 L of DMF for rinsing,
and the mixture was then stirred at 20-22 °C for 2-3 h.
Next, ethyl acetate (30 L) was introduced and then cold water
(30 L) was added slowly to maintain the temperature at <30 °C.
The mixture was stirred for 0.5 h and settled to give two clear
layers. The aqueous layer was separated and back-extracted with
EtOAc (25 L). A combined organic layer was washed with water
(2 × 20 L) and then saturated brine solution (26 L). The organic
layer was concentrated in vacuo to ∼20 L, followed by addition of
∼20 L of n-heptane at 18-22 °C, and stirred for 1-2 h to provide
a white slurry of the product. The remaining n-heptane (5 L) was
introduced over 1 h to afford a thick slurry. The slurry was cooled
to 0-5 °C to reduce the supernatant concentration to <1.5 mg/
mL. It was then filtered, and the cake was washed with cold
premixed EtOAc/n-heptane (1:3, 12 L) and air-dried at 23 °C under
nitrogen for 12 h. The isolated white crystalline solid (3.56 kg)
was obtained in 93.6% yield.
Acknowledgment. Thanks to Roy Helmy and Peng Wang,
our analytical team. Thanks to Liam Dunne and James Corry
for the observation that potassium hydroxide buffers the pH to
allow high yields from carbometalation reactions.
End-Game Strategy 2: Linear Approach. A. Preparation of
Alcohol (27). To a 50 L glass vessel equipped with an overhead
stirrer, a thermocouple, and a nitrogen inlet was charged 9 L of
DMF, liquid bromohexanol 25, solid sulfone acid 2, and 2 L of
DMF for rinsing to give a clear solution at room temperature. To
this resulting solution was added powder K2CO3 in one portion at
20-22 °C, followed by 2 L of DMF for rinsing. Then, the solution
was stirred at 20-22 °C for 10 min and then heated to 40-45 °C
for 3-5 h.
Supporting Information Available: Spectroscopic character-
ization (1H, 13C, and FT-IR) for compounds 1-3, 12a, 13 (1H and
13C only), and 27. This material is available free of charge via the
Internet at http://pubs.acs.org.
JO051712G
J. Org. Chem, Vol. 71, No. 2, 2006 491