Efficient one-pot synthesis of the 2-aminocarbonylpyrrolidin-4-ylthio-containing side chain of the new broad-spectrum carbapenem antibiotic ertapenem.

Article abstract of DOI:10.1021/jo011170c

An efficient synthesis of the 2-aminocarbonylpyrrolidin-4-ylthio containing side chain of ertapenem (MK-0826) is described. Starting material N-(O,O-diisopropyl phosphoryl)-trans-4-hydroxy-L-proline is converted in a one-pot process to (2S)-cis-3-[[(4-mercapto-2-pyrrolidinyl)carbonyl]amino]benzoic acid monohydrochloride in 70-75% overall yield via a series of six reactions. The development of each of these reactions and the isolation of the product is discussed in detail.

Full text of DOI:10.1021/jo011170c

Efficien t On e-P ot Syn th esis of th e
2-Am in oca r bon ylp yr r olid in -4-ylth io-Con ta in in g Sid e Ch a in of th e
New Br oa d -Sp ectr u m Ca r ba p en em An tibiotic Er ta p en em
Karel M. J . Brands,*,† Ronald B. J obson,† Karen M. Conrad,† J . Michael Williams,†
Brenda Pipik,† Mark Cameron,‡ Antony J . Davies,‡ Peter G. Houghton,‡ Michael S. Ashwood,‡
Ian F. Cottrell,‡ Robert A. Reamer,† Derek J . Kennedy,‡ Ulf-H. Dolling,† and Paul J . Reider†
Department of Process Research, Merck Research Laboratories, P.O. Box 2000,
Rahway, New J ersey 07065, and Department of Process Research, Merck, Sharp & Dohme,
Hertford Road, Hoddesdon, Hertfordshire EN11 9BU, U.K.
jos_brands@merck.com
Received December 21, 2001
An efficient synthesis of the 2-aminocarbonylpyrrolidin-4-ylthio containing side chain of ertapenem
(MK-0826) is described. Starting material N-(O,O-diisopropyl phosphoryl)-trans-4-hydroxy-L-proline
is converted in a one-pot process to (2S)-cis-3-[[(4-mercapto-2-pyrrolidinyl)carbonyl]amino]benzoic
acid monohydrochloride in 70-75% overall yield via a series of six reactions. The development of
each of these reactions and the isolation of the product is discussed in detail.
SCHEME 1
In tr od u ction
The discovery and development of new antibiotics has
drawn significant attention in recent years due to the
rising prevalence of multidrug resistant bacteria.1 Er-
tapenem (1) is a new parenteral broad-spectrum carbap-
enem antibiotic2,3 with a unique efficacy profile which is
in late stages of development at Merck.4,5 The compound
is stable against renal dehydropeptidase-I and resistant
to most â-lactamases.6 Once-daily intramuscular or I. V.
dosing is expected to be sufficient for treatment of most
serious upper and lower respiratory tract, urinary tract,
† Merck Research Laboratories.
‡ Merck, Sharp & Dohme.
(1) (a) Chu, D. T. W.; Plattner, J . J .; Katz, L. J . Med. Chem. 1996,
39, 3853-3874. (b) Gold, H. S.; Moellering, R. C. New Engl. J . Med.
1996, 335, 1445-1453. (c) Bonafede, M.; Rice, L. B. J . Lab. Clin. Med.
1997, 130, 558-566. (d) Morris, A.; Kellner, J . D.; Low, D. E. Curr.
Opin. Microbiol 1998, 1, 524-529. (e) Livermore, D. M. Int. Congr.
Symp. Ser. R. Soc. Med. 2001, 247, 61-69.
skin, obstetric and gynecologic infections due to its long
pharmacokinetic half-life.7
(2) For a brief history: Sunagawa, M.; Sasaki, A. Heterocycles 2001,
54, 497-528.
Following conventional carbapenem retrosynthetic
analysis, ertapenem can be assembled from 4-nitroben-
zyl-protected â-methyl carbapenem enolphosphate 28 and
the 2-aminocarbonylpyrrolidin-4-ylthio-containing side
chain 3 (Scheme 1). Many efficient approaches to 2 have
been reported in the literature,9 and this compound is
now commercially available on a large scale.10 In this
paper, we will describe an efficient and practical one-
pot synthesis of 3 that is amenable to large-scale produc-
(3) (a) Kawamoto, I. Drugs Future 1998, 23, 181-189. (b) Coulton,
S.; Hunt, E. In Progress in Medicinal Chemistry; Ellis, G. P., Luscombe
D. K., Eds.; Elsevier Science: New York, 1996; Vol. 33, pp 99-145. (c)
Sader, H. S.; Gales, A. C. Drugs 2001, 61, 553-564.
(4) (a) Sundelof, J . G.; Hajdu, R.; Gill, C. J .; Thompson, R.; Rosen,
H.; Kropp, H. Antimicrob. Agents Chemother. 1997, 1743-1748. (b)
J acoby, G.; Han, P.; Tran, J . Antimicrob. Agents Chemother. 1997,
1830-1831. (c) Livermore, D. M.; Carter, M. W.; Bagel, S.; Wiedemann,
B.; Baquero, F.; Loza, E.; Endtz, H. P.; Van den Braak, N.; Fernandes,
C. J .; Fernandes, L.; Frimodt-Moller, N.; Rasmussen, L. S.; Giama-
rellou, H.; Giamarellos-Bourboulis, E.; J arlier, V.; Nguyen, J .; Nord,
C.-E.; Struelens, M. J .; Nonhoff, C.; Turnidge, J .; Bell, J .; Zbinden, R.;
Pfister, S.; Mixson, L.; Shungu, D. L. Antimicrob. Agents and Chemo-
ther. 2001, 45, 1860-1867. (d) Fuchs, P. C.; Barry, A. L.; Brown, S. D.
Antimicrob. Agents Chemother. 2001, 45, 1915-1918.
(5) (a) Sorbera, L. A.; Rabasseda, X.; Castaner, J .; Leeson, P. A.
Drugs Future 2000, 25, 795-802. (b) Odenholt, I. Expert Opin. Invest.
Drugs 2001, 10, 1157-1166.
(7) Gill, C. J .; J ackson, J . J .; Gerckens, L. S.; Pelak, B. A.; Thompson,
R. K.; Sundelof, J . G.; Kropp, H.; Rosen, H. Antimicrob. Agents
Chemother. 1998, 42, 1996-2001.
(8) Shih, D. H.; Baker, F.; Cama, L.; Christensen, B. G. Heterocycles
1984, 21, 29-40.
(6) Kohler, J .; Dorso, K. L.; Young, K.; Hammond, G. G.; Rosen, H.;
Kropp, H.; Silver, L. L. Antimicrob. Agents Chemother. 1999, 43, 1170-
1176.
(9) Berks, A. H. Tetrahedron 1996, 52, 331-375.
(10) Carbapenem enolphosphate 2 is commercially available from
Takasago, Kaneka, and Nisso companies.
10.1021/jo011170c CCC: $22.00 © 2002 American Chemical Society
Published on Web 06/01/2002
J . Org. Chem. 2002, 67, 4771-4776
4771

Brands et al.
SCHEME 2
SCHEME 3
tion. The coupling of 3 with 2 and the conversion to
ertapenem (1) will be disclosed separately.11
Our strategy for the synthesis of 3 (Scheme 2) is based
on a general and expedient synthesis for protected 2-thia-
5-azabicyclo[2.2.1]heptan-3-ones, which we recently dis-
closed in a preliminary communication.12 Thus, a suitably
protected 4-hydroxyproline derivative 4 was expected to
deliver the key thiolactone 5, which was projected to
undergo aminolysis with m-aminobenzoic acid (MABA).
A team from Sumitomo Pharmaceuticals reported re-
cently on their independent arrival at a similar strat-
egy.13 Thiolactones of the type of 5 have been synthesized
previously.14 However, inefficient multistep approaches
using impractical protecting groups (P ) Ts, Ac) were
used in those studies. We recognized significant advan-
tages in entering a coupling of 2 with a completely
unprotected side chain. Therefore, we carefully selected
the nature of protecting group P in 5 such that depro-
tection of the aminoysis product would afford salt 3
directly. The little-used diisopropyl phosphoryl group
(DIPP)15 met all of our criteria. It is economical, and only
nonvolatile and innocuous cleavage products are gener-
ated under mild acidic conditions. We recently reported
on a novel and practical method for the preparation of
N-(O,O-diisopropylphosphoryl)-trans-4-hydroxy-L-pro-
line (6; DIPP-Hyp).16 The projection that all transforma-
tions leading from 4 to 3 could be performed in the same
pot promised to make our approach particularly practical
and economical.
TABLE 1. Yield of 10 a s a F u n ction of th e Solven t(s)a
entry
solvent
% yield for 10b
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
THFc
<10
<5
36
75
58
42
33
92
86
64
83
45
83
75
53
77
DMEd
EtOAcc,e
MeCNc
DMFd
DMAd
NMPd
d
CH2Cl2
ClCH2CH2Clc,e
cyclohexanoned
fluorobenzenec
c
PhCF3
THF/MeCN 1:1c
MeCN/EtOAc 1:2c
DMF/toluene 1:5c
CH2Cl2/toluene 1:2c
a
According to the procedure detailed in the Experimental
Section, 1.0 equiv of 6 as a 0.15 M solution in solvent(s) at -15 °C
was treated consecutively with 2.1 equiv of DIPEA, 1.05 equiv of
DPPC, 1.1 equiv of pyridine, 1.1 equiv of MsCl and 1.2 equiv of
b
Na2S. Yields of 10 determined by HPLC. c The reaction mixture
became a heavy slurry after DPPC addition and became homoge-
d
neous again after MsCl addition. The reaction mixture was
homogeneous throughout the entire sequence. e The starting mate-
rial is barely soluble in this solvent at a concentration of 0.15 M.
The thiolactone functionality in 5 performs a pivotal
role in our strategy. It not only allows the stereoselective
introduction of a protected mercapto group but also
activates the carbonyl group toward reaction with m-
aminobenzoic acid. Thus, the synthesis of 3 will require
only a single protecting group. Our strategy also guar-
antees high enantiomeric purity of the end product. Any
loss of stereochemical integrity at C-2 or C-4 before the
formation of the thiolactone would result in species that
would not be expected to cyclize and would be readily
separated in a judiciously chosen workup. Only the
unlikely event of a combination of epimerization at C-2
and retention of configuration during the intramolecular
substitution at C-4 would result in the formation of the
enantiomer of 5.
Resu lts a n d Discu ssion
Syn th esis of th e DIP P -P r otected Th iola cton e.
Scheme 3 shows the sequential reactions that were used
to convert DIPP-protected hydroxyproline 6 to thiolactone
10 in a single pot. In the first step, the carboxyl group
was activated as mixed carboxylic phosphinic anhydride
7 via a reaction with diphenylphosphinic chloride (DPPC)
in the presence of diisopropylethylamine (DIPEA). This
intermediate was directly reacted with methanesulfonyl
chloride in the presence of pyridine to produce mixed
anhydride mesylate 8. The latter was then quenched with
aqueous sodium sulfide yielding 9 instantaneously, which
then slowly cyclized to 10. It was found that the overall
yield for this sequence of reactions was critically depend-
ent on the solvent. Table 1 compiles the most important
results of solvent optimization studies. While THF was
the preferred solvent in the preparation of the tert-
butyloxycarbonyl (BOC) and 4-nitrobenzyloxycarbonyl
(pNZ) protected analogues of 5,12 it provided the DIPP-
protected 10 in a disappointingly low yield (Table 1, entry
1). The data of Table 1 clearly show that the best yield
is obtained using dichloromethane as the solvent. In
many solvent systems, mixed anhydride alcohol 7 crys-
tallizes from the reaction mixture as it is formed. Indeed,
this reactive intermediate could be isolated, purified via
(11) Williams, J . M.; Brands, K. M. J .; Skerlj, R.; Dolling, U.-H.
Manuscript in preparation.
(12) Brands, K. M. J .; Marchesini, G.; Williams, J . M.; Dolling, U.-
H.; Reider, P. J . Tetrahedron Lett. 1996, 37, 2919-2922.
(13) Matsumura, H.; Bando, T.; Sunagawa, M. Heterocycles 1995,
41, 147-159.
(14) (a) Yamada, S.; Murakami, Y.; Koga, K. Tetrahedron Lett. 1968,
1501-1506. (b) Verbiscar, A. J .; Witkop, B. J . Org. Chem. 1970, 35,
1924-1927.
(15) Greene, T. W.; Wuts, P. G. M. Protective Groups in Organic
Synthesis, 3rd ed.; Wiley: New York, 1999; p 598.
(16) Brands, K. M. J .; Wiedbrauk, K.; Williams, J . M.; Dolling, U.-
H. Tetrahedron Lett. 1998, 39, 9583-9586.
4772 J . Org. Chem., Vol. 67, No. 14, 2002

Synthesis of a Side Chain of Ertapenam
F IGURE 1.
recrystallization, and characterized. Typically, crystalline
7 went back into solution to form the more soluble 8
during the reaction with methanesulfonyl chloride. How-
ever, there is no correlation between low solubility of 7
and poor overall reaction performance. For entries 1, 3,
4, 9, and 11-16 in Table 1, a slurry was formed
immediately after addition of DPPC. For some of these
experiments, the overall yield was quite acceptable
(>75% for entries 4, 9, 13, 14, and 16). On the other hand,
not all homogeneous reaction mixtures provided 10 in
good yield (cf. entries 2, 5-7, and 10). A relationship
between solvent polarity and reaction performance is not
evident either. Good yields are obtained in systems of
widely varying polarity (cf. entries 4, 8, 9, 11, 13, 14, and
16). Surprisingly, overall yields were significantly lower
in solvents closely related to dichloromethane or proposed
as dichloromethane alternatives17 (entries 9, 11, and 12).
NMR studies provided an explanation for the lower yields
of 10 in particular solvent systems. When 6 was allowed
to react with 1.05 equiv of DPPC in the presence of 1.10
equiv of DIPEA in DMF-d7 at -30 °C, three distinct
proline-containing species were formed in a 73/18/9 ratio.
In situ 13C NMR studies were used to characterize the
major species as 7, the expected mixed anhydride, and
the more abundant minor species as 11, a sym-anhydride
(Figure 1). In the 1H-decoupled 13C spectrum, the nitrogen-
F IGURE 2. Optimization of the amount of sodium sulfide in
the reaction with 8. (Varying amounts of a 1.0 M solution of
sodium sulfide trihydrate in water were rapidly added to a
solution of 8 prepared in the standard manner at -20 °C. The
yield for 10 was determined by HPLC after an identical age
and workup.)
tion of 11 occurred. Mixed anhydride mesylate 8 is more
prone to disproportionation than 7 under the reaction
conditions, which caused some concern for scale-up. It
was demonstrated in a series of control experiments in
dichloromethane that mesylation at higher than optimal
temperatures (-10 °C vs <-15 °C) and/or aging of
solutions of 8 for longer than required to achieve complete
conversion (2 h vs 30 min) resulted in more dispropor-
tionation of 8 into 12 and 13. Consequently, significantly
lower yields for 10 were obtained under these conditions
(80% vs 90-95% under optimal conditions).
It should be noted that several other methods for
activating the carboxyl group in 6 were also examined,
but none of these compared favorably with the mixed
carboxylic phosphinic anhydride method. The best results
were obtained with chloroformates, particularly isobutyl
chloroformate. Combinations with various bases and
solvents and various addition protocols were probed with
this reagent. The optimal yield for 10 (73% according to
HPLC) was obtained when 1.1 equiv of isobutyl chloro-
formate was added to 6 in dichloromethane at -40 °C in
the presence of both DIPEA and pyridine followed by
addition of methanesulfonyl chloride and a quench with
aqueous sodium sulfide as usual (vide infra). Again,
dramatic solvent effects were noted with significantly
lower overall yields in THF or acetonitrile (38 and 25%,
respectively).
bearing methine of 7 was a doublet of doublets (J CP
)
6.4, 4.0 Hz) due to spin-spin coupling to two phosphorus
atoms. In 11, the analogous methine was a doublet
(J CP ) 6.4 Hz) with coupling to only one phosphorus. A
triple-resonance NMR experiment (13C observed with
1
simultaneous H and 31P decoupling) was used to verify
these assignments. The structure of the third species
could not be elucidated. The sym-phosphinic anhydride
12 (which is also always present as a small impurity in
the starting DPPC), was also detected in the sample.
Significant sym-anhydride formation obviously detracts
from the maximum achievable yield for 10 in this process.
A reverse addition protocol (slow addition of a solution
of 6 and DIPEA in DMF to a cold solution of DPPC in
DMF, followed by a typical sodium sulfide quench)
resulted in more sym-anhydride 11 formation and a lower
overall yield for 10 than the normal addition protocol.
The thermal stability of phosphinic carboxylic mixed
anhydrides of N-protected amino acids has been reported
in the literature.18 Solutions of 7 in a variety of solvents
(dichloromethane, DMF, acetonitrile, and THF) were
quite stable according to 31P NMR (no disproportionation
at -25 °C; 10-20% of 11 formed after 16 h at rt).
Addition of DIPEA to the dichloromethane solution at
-25 °C had no effect. However, at rt, significant forma-
A careful definition of the sodium sulfide quench
protocol also proved crucial in obtaining the best yields
for our process. The results of a study determining the
optimum amount of sodium sulfide is given in Figure 2.
The best yield was obtained with 1.2 equiv of sodium
sulfide (the optimal amount of water for the quench was
determined at 0.4 mL water per mL of dichloromethane).
It proved critical to add the aqueous sodium sulfide
solution into the cold solution of 8 in dichloromethane
as fast as possible with vigorous agitation. With a
relatively slow addition (10 min rather than <2 min), the
(17) Ogawa, A.; Curran, D. P. J . Org. Chem. 1997, 62, 450-451.
(18) Ramage, R.; Atrash, B.; Hopton, D.; Parrott, M. J . J . Chem.
Soc., Perkin Trans. 1 1985, 1617-1622.
J . Org. Chem, Vol. 67, No. 14, 2002 4773

Brands et al.
SCHEME 4
SCHEME 5
overall assay yield for 10 was significantly lower (84 vs
90%, in a direct comparison). During the quench, 8 is
instantaneously converted to 9. Completion of the in-
tramolecular displacement of the mesylate in 9 requires
approximately 2 h at 20-25 °C. In the quench, small
amounts of carboxylate 14 are also formed. It was
demonstrated in a control experiment that the ring
closure of 14 to oxylactone 15 is significantly slower than
in the sulfur series, presumably due to a combination of
better nucleophilicity of the thiocarboxylate and higher
ring strain in the oxylactone (Scheme 4).19 An authentic
sample of oxylactone 15 was independently synthesized
from 6 via a Mitsunobu reaction. Any amount of oxylac-
tone 15 carried forward into the aminolysis and depro-
tection reactions (vide infra) eventually produced 16, an
impurity which is poorly rejected in the final crystalliza-
tion of 3. Recognition of the different cyclization rates of
9 and 14 suggested an extraction immediately after
completion of thiolactone formation as a practical way
to control the level of 15 in 10 at less than 0.1%.
Thiolactone 10 can be crystallized after an aqueous
workup. However, using the dichloromethane solution
directly for aminolysis with MABA is preferred since
product losses in the isolation are avoided. Typically, the
overall yield from 6 to 10 is in the 90-95% range under
the fully optimized conditions.
in 2-3 h at rt and yields crude 3 in 90-95% overall yield
for the aminolysis and deprotection reactions (Scheme
5).
Isola tion of 3. With a one-pot process from 6 to 3 in
hand, a practical isolation of the product from the
mixture of acetic acid and hydrochloric acid was needed.
The solubility of the hydrochloride 3 in this system
increases dramatically with increasing amounts of water.
It was indeed established that 3 can be crystallized by
removing most of the water via distillation at reduced
pressure. Although the recovery from this crystallization
was very good (95%), the purity of the solids was not
satisfactory. Moreover, it proved difficult to reduce
residual acetic acid in 3 to the very low levels required
via washing and/or drying. Screening of various other
crystallization solvents led to the identification of 1-pro-
panol and 1-butanol as excellent alternatives. From these
alcohols, 3‚HCl crystallizes with >99 area % purity
according to HPLC analysis. Thus, a crystallization
protocol was developed in which water and acetic acid
were efficiently removed from the crude solution of 3‚
HCl by azeotropic vacuum distillation with 1-butanol.20
Two pseudopolymorphic crystal forms for 3‚HCl were
obtained using this procedure: an unsolvated product
and a thermodynamically less stable (and more soluble)
1:1 solvate with 1-butanol. The control over the desired
crystal form critically depended on seeding in a defined
solvent composition window during the vacuum distilla-
tion. This process was not very robust since there was
also the added complication of formation of significant
amounts of butyl acetate during large scale distillations.
Therefore, a two-stage crystallization was developed.
Crude 3‚HCl was first crystallized from acetic acid via
removal of water, as described above, and filtered. The
resulting crude wet cake was then dissolved in a 9:1
mixture of 1-butanol and water. Slow removal of the
water via an azeotropic distillation in vacuo led to a
reproducible crystallization of the unsolvated product
with a high recovery (85-90% over the two crystalliza-
tions).
Am in olysis a n d Dep r otection . While the reaction
of 10 with aliphatic amines is very fast at rt in most
solvents, the reaction with MABA is relatively slow. This
is probably caused by the combined effects of low nucleo-
philicity of the free aniline and the significant presence
of zwitterionic MABA. Upon warming in several solvents,
the aminolysis of 10 with MABA was not particularly
clean. An important solvent effect was discovered when
the aminolysis was performed in acetic acid. In this
solvent the aminolysis is significantly accelerated at rt,
typically resulting in a complete conversion within 3 h.
Presumably, acid catalysis of the aminolysis via pro-
tonation of the tetrahedral intermediate formed upon
addition of the amine to the carbonyl group more than
offsets the higher degree of protonation of MABA under
these conditions. In practice, the dichloromethane solu-
tion containing 10, prepared as described above, was
added to a solution of 1.0 equiv of MABA in acetic acid
while removing the dichloromethane by distillation. It
proved possible to isolate the aminolysis product 18 via
a workup and crystallization as the corresponding dicy-
clohexylammonium salt. However, once again, the direct
use of the acetic acid solution of 18 for the next step was
preferred. Deprotection of 18 was readily achieved via
addition of a small amount of concentrated hydrochloric
acid to the mixture. This reaction is typically completed
Not surprisingly, 3 forms the corresponding disulfide
19 quite readily (Figure 3). It was discovered that
addition of a small amount of tri-n-butylphosphine (1-5
mol % relative to 3) during the crystallization process
(20) Horsley, L. H. In Azeotropic Data III; Advances in Chemistry
116; Gould, R. F., Ed.; American Chemical Society: Washington, DC,
1973; pp 22-23 and 110.
(19) March, J . Advanced Organic Chemistry, 3rd ed.; J ohn Wiley &
Sons: New York, 1985; pp 307-308.
4774 J . Org. Chem., Vol. 67, No. 14, 2002

Synthesis of a Side Chain of Ertapenam
min). To the resulting solution was added a solution of diphen-
ylphosphinic chloride (98%; 2.10 kg; 8.90 mol) in dichloro-
methane (4 L) at such a rate that the temperature was kept
below -18 °C (20 min). After aging at -20 °C for 40 min, a
solution of pyridine (0.72 kg; 9.10 mol; dried over 4 Å mol-
sieves) in dichloromethane (1 L) was added in one portion,
immediately followed by a solution of methanesulfonyl chloride
(1.07 kg; 9.34 mol) in dichloromethane (5 L) at such a rate
that the temperature was kept below -18 °C (30 min). After
aging at -20 °C for an additional 30 min, a solution of sodium
sulfide trihydrate (1.35 kg; 10.22 mol) in water (19 L) was
added under vigorous agitation as fast as possible (<2 min).
The resulting biphasic mixture (+ 3 °C) was warmed to 25 °C
over 5 min and vigorously agitated for another 2 h. Agitation
was stopped, and the layers were separated. The organic layer
was extracted with 1.0 M hydrochloric acid (20 L), a solution
of sodium bicarbonate (1.65 kg) in water (20 L), and water (15
L). The resulting dichloromethane solution (approximately 55
L) was assayed to contain 2.35 kg of 10 (95% yield). This
solution was added under vacuum into a slurry of 3-aminoben-
zoic acid (98%; 1.15 kg; 7.86 mol) in glacial acetic acid (12.5
L) while distilling the dichloromethane. When all dichloro-
methane had been distilled out, the resulting clear solution
was aged at rt overnight. After conversion to 18 was confirmed
by HPLC, concentrated hydrochloric acid (3.0 L) was added.
The resulting mixture was stirred under N2 for 2 h. After
completion of the deprotection was confirmed by HPLC, the
solution was concentrated to a total volume of approximately
15 L. The solution was flushed in vacuo with additional fresh
glacial acetic acid (30 L), and then toluene (15 L) was added
over 1 h at atmospheric pressure. The slurry was stirred under
N2 overnight, cooled to 0-5 °C, aged for another 1 h, and
filtered. The solids were washed with a mixture of glacial acetic
acid and toluene in a 1:5 ratio (7.5 L), followed by toluene (15
L). The wet cake was dried in vacuo at rt and then redissolved
in a mixture of water (5 L) and 1-butanol (40 L). To the clear
solution was added tri-n-butylphosphine (75 g). The mixture
was concentrated in vacuo to a total volume of approximately
20 L while maintaining the pot temperature below 40 °C. The
solution was flushed with additional 1-butanol (40 L). The
resulting crystal slurry was filtered after aging for 5 h at 0-5
°C. The solids were washed with 1-butanol (5 L) and toluene
(3 × 5 L) and dried in vacuo at 100 °C under a stream of N2.
In this way, 1.88 kg of 3 was obtained (73% overall yield from
6). The solids contained >99 wt % of 3 at >99 area % purity
according to HPLC analysis (Zorbax Eclipse XDB-C18;
150 × 4.6 mm; 5 µm particle size; UV detection at 225 nm;
0.1% perchloric acid/acetonitrile gradient from 90:10 to 50:50
over 50 min; 1.0 mL/min): mp 212 °C; [R]365 ) +41 (c ) 1.0;
F IGURE 3.
assured low levels of 19 in the isolated product (<0.2 area
% by HPLC).21 However, it was also determined that the
addition of tri-n-butylphosphine carried some risk. Add-
ing larger amounts and/or using extended recrystalliza-
tion times at elevated temperatures led to significant
formation of 16, 20, and 21, with the latter predominat-
ing.22 The structures of 16, 20, and 21 were initially
proposed based on LC-MS analysis of the crystallization
mixtures and later corroborated via their independent
synthesis.23
Con clu sion s
An exceedingly efficient and practical synthesis for side
chain 3 of the novel broad-spectrum carbapenem antibi-
otic ertapenem (1) was developed. Compound 3 can be
synthesized in a one-pot operation comprising six chemi-
cal reactions starting from DIPP-hydroxyproline (6). The
product is isolated as the hydrochloride salt with an
overall yield of 70-75%, corresponding to an average
yield of >95% per chemical step. The above-described
process has been implemented on a multikilogram scale
and consistently provides isolated 3‚HCl with >99.5 area
% purity according to HPLC analysis. This chemistry
appears quite general for use in the synthesis of other
2-aminocarbonylpyrrolidin-4-ylthio-containing carbap-
enem side chains.
1
20 mM phosphoric acid); H NMR (400 MHz, D2O) δ 7.96 (s,
1H), 7.76 (d, J ) 7.8 Hz, 1H), 7.61 (d, J ) 7.5 Hz, 1H), 7.45
(m, 1H), 4.69 (br s, 1H), 4.57 (t, J ) 8.4 Hz, 1H), 3.78 (dd, J )
11.7, 6.9 Hz, 1H), 3.68 (m, 1H), 3.32 (dd, J ) 11.7, 7.5 Hz,
1H), 2.96 (m, 1H), 2.13 (m, 1H); 13C NMR (100 MHz, D2O) δ
169.3, 166.8, 136.5, 130.2, 129.3, 126.4, 125.6, 121.6, 59.9, 53.9,
39.1, 34.6. Anal. Calcd for C12H15ClN2O3S: C, 47.60; H, 4.99;
Cl, 11.71; N, 9.25; S, 10.59. Found: C, 47.66; H, 4.93; Cl, 11.76;
N, 9.11; S, 10.57.
Exp er im en ta l Section
All commercially available materials and solvents were used
as received. Reaction temperatures were measured internally,
unless indicated otherwise (rt ) 20-23 °C). Melting points are
not corrected. Optical rotations were measured at 25 °C.
(2S)-cis-3-[[(4-Mer ca p to-2-p yr r olid in yl)ca r bon yl]a m i-
n o]ben zoic Acid Mon oh yd r och lor id e (3). A solution of 6
(2.50 kg; 8.46 mol) in dichloromethane (50 L; dried over 4 Å
molsieves) was cooled to -20 °C, and diisopropylethylamine
(2.35 kg; 18.18 mol; dried over 4 Å molsieves) was added at
such a rate that the temperature was kept below -18 °C (20
N-(O,O-Diisopr opylph osph or yl)-tr a n s-4-h ydr oxy-L-pr o-
lin e (6). Trans-4-Hydroxy-L-proline (200 g; 1.52 mol) was
dissolved in water (500 mL), and the resulting solution was
cooled to 0-5 °C. The pH was adjusted from 5 to 6 to 9 using
a 25% sodium hydroxide solution (18 mL). Diisopropyl phos-
phite (280 g; 1.68 mol) was added in one portion. Sodium
hypochlorite (20 wt %; 640 mL; concentration determined
before use at 2.6 M via iodometric titration) was added over
2.5 h while the pH was maintained at 9.0 (using 25% sodium
hydroxide) and the temperature at 0-5 °C. Excess bleach was
then quenched via addition of sodium bisulfite (30 g; 0.28 mol).
The pH of the resulting solution was adjusted to 2 via slow
addition of concentrated hydrochloric acid (230 mL) while the
temperature was maintained at 0-5 °C. The solution was
(21) (a) Overman, L. E.; Smoot, J .; Overman, J . D. Synthesis 1974,
59-60. (b) Ayers, J . T.; Anderson, S. R. Synth. Commun. 1999, 29,
351-358.
(22) (a) Hoffmann, F. W.; Ess, R. J .; Simmons, T. C.; Hanzel, R. S.
J . Am. Chem. Soc. 1956, 78, 6414. (b) Walling, C.; Rabinowitz, R. J .
Am. Chem. Soc. 1959, 81, 1243-1249. (c) Corey, E. J .; Shimoji, K. J .
Am. Chem. Soc. 1983, 105, 1662-1664.
(23) Details are reported in the Supporting Information.
J . Org. Chem, Vol. 67, No. 14, 2002 4775

Brands et al.
saturated with sodium chloride (170 g) and extracted with
isopropyl acetate (2 × 2 L). The product crystallized from the
combined organic extracts upon partial concentration in vacuo.
The slurry was flushed with fresh isopropyl acetate. The
crystals were filtered, washed, and dried in vacuo at rt to yield
359 g of the product (80% yield): mp 94-95 °C; [R]589 ) -45
(c ) 1.00; EtOH); 1H NMR (400 MHz, CDCl3) δ 7.75 (br, 2H),
4.76 (m, 1H), 4.56 (m, 1H), 4.42 (m, 1H), 4.34 (dd, J ) 12.9,
7.2 Hz, 1H), 3.28 (m, 1H), 3.20 (m, 1H), 2.27 (m, 1H), 2.14 (m,
1H), 1.28 (d, J ) 6.3 Hz, 12H); 31P NMR (162 MHz, CDCl3) δ
5.1; 13C NMR (100 MHz, CDCl3; resolution enhanced) δ 175.7,
71.4 (d, J ) 6.0 Hz), 71.3 (d, J ) 6.0 Hz), 70.5 (d, J ) 8.0 Hz),
59.0 (d, J ) 5.9 Hz), 54.5 (d, J ) 2.9 Hz), 39.1 (d, J ) 8.1 Hz),
23.6 (d, J ) 14.6 Hz), 23.5 (d, J ) 12.3 Hz), 23.5 (d, J ) 14.9
Hz), 23.4 (d, J ) 11.9 Hz). Anal. Calcd for C11H22NO6P: C,
44.75; H, 7.51; N, 4.74; P, 10.50. Found: C, 44.80; H, 7.28; N,
4.57; P, 10.64.
Ack n ow led gm en t. Mrs. Y. Wang and Dr. L. Crock-
er of the Merck Research Laboratories are gratefully
acknowledged for their characterization of the pseudo-
polymorphic crystal forms of 3‚HCl.
Su p p or tin g In for m a tion Ava ila ble: Experimental de-
tails and physical data for compounds 7, 10, 15, 16, and 18-
1
21 as well as H and 13C NMR spectra of 19-21. This material
is available free of charge via the Internet at http://pubs.acs.org.
J O011170C
4776 J . Org. Chem., Vol. 67, No. 14, 2002