Practical synthesis of the new carbapenem antibiotic ertapenem sodium

Article abstract of DOI:10.1021/jo0501442

A practical synthesis for the large-scale production of the new carbapenem antibiotic, [4R,5S,6S]-3-[[(3S,5S)-5-[[(3-Carboxyphenyl)amino]carbonyl]-3- pyrrolidinyl]thio]-6-[(1R)-1-hydroxyethyl]-4-methyl-7-oxo-1-azabicyclo[3.2.0] hept-2-ene-2-carboxylic aci

Full text of DOI:10.1021/jo0501442

Practical Synthesis of the New Carbapenem Antibiotic Ertapenem
Sodium
J. Michael Williams,*,† Karel M. J. Brands,† Renato T. Skerlj,† Ronald B. Jobson,†
George Marchesini,† Karen M. Conrad,† Brenda Pipik,† Kimberly A. Savary,† Fuh-Rong Tsay,†
Peter G. Houghton,† D. Richard Sidler,† Ulf-H. Dolling,† Lisa M. DiMichele,† and
Thomas J. Novak‡
Departments of Process Research and Analytical Research, Merck Research Laboratories,
Post Office Box 2000, Rahway, New Jersey 07065
mike_williams@merck.com
Received January 24, 2005
A practical synthesis for the large-scale production of the new carbapenem antibiotic, [4R,5S,6S]-
3-[[(3S,5S)-5-[[(3-Carboxyphenyl)amino]carbonyl]-3-pyrrolidinyl]thio]-6-[(1R)-1-hydroxyethyl]-4-
methyl-7-oxo-1-azabicyclo[3.2.0]hept-2-ene-2-carboxylic acid monosodium salt (ertapenem sodium,
1), has been developed. The synthesis features the novel use of 1,1,3,3-tetramethylguanidine as
base for the low-temperature reaction of a thiol, derived from trans-4-hydroxy-L-proline, with the
carbapenem nucleus activated as the enol phosphate. Hydrogenolysis of a p-nitrobenzyl ester is
effected using a palladium on carbon catalyst to give an overall yield for the two steps of 90%. The
use of bicarbonate in the hydrogenolysis was key in providing protection of the pyrrolidine amine
as the sodium carbamate improving both the performance of the reaction and the stability of the
product. This discovery made processing at manufacturing scale possible. Experimental evidence
for the formation of the sodium carbamate is provided. A remarkably expedient process for the
simultaneous purification and concentration of the aqueous product stream relies on ion-pairing
extraction for the removal of the water-soluble 1,1,3,3-tetramethylguanidine. Crystallization then
affords 59-64% overall yield of the monosodium salt form of the product.
Introduction
produced new antibiotics with improved properties for
the treatment of human infections.3
The rising emergence of bacteria resistant to existing
antibacterial therapy is responsible for the prevailing
interest in new agents for the fight against these threats
to human health.1 The carbapenem class of antibiotics
has become widely used in treating infections caused by
a broad range of pathogens.2 The development of new
carbapenems is an important part of our effort to address
the problem of multi-drug-resistant bacteria and has
Imipenem, the first carbapenem antibiotic approved for
clinical use, has a remarkable spectrum of activity and
remains on the last line of defense in treatment of many
of the most serious infections.4 As a consequence of poor
solution stability toward hydrolysis and susceptibility to
enzymatic inactivation, however, imipenem must be
administered four times daily along with cilastatin, an
inhibitor of the enzyme, human renal dehydropeptidase-1
(DHP-1), which is responsible for inactivation of the
antibiotic.5
† Merck Process Research.
‡ Merck Analytical Research.
(3) Sunagawa, M.; Sasaki, A. Heterocycles 2001, 54, 497-528.
(4) (a) Kahan, J. S.; Kahan, F. M.; Goegelman, R.; Currie, S. A.;
Jackson, M.; Stapley, E. O.; Miller, T. W.; Miller, A. K.; Hendlin, D.;
Mochales, S.; Hernandez, S.; Woodruff, H. B.; Birnbaum, J. J. Antibiot.
(Tokyo) 1979, 32, 1-12. (b) Kropp, H.; Sundelof, J. G.; Kahan, J. S.;
Kahan, F. M.; Birnbaum, J. Antimicrob. Agents Chemother. 1980, 17,
993-1000.
(1) (a) Goldhaber, M. Science 1994, 266, 1462. (b) Verhoef, J. Adv.
Exp. Med. Biol. 2003, 531, 301-13.
(2) (a) Coulton, S.; Hunt, E. In Progress in Medicinal Chemistry;
Ellis, G. P., Luscombe, D. K., Eds.; Elsevier Science: 1996; Vol. 33, p
99-145. (b) Hashizume, T.; Morishima, H. Drugs Future 2000, 25,
833-841.
10.1021/jo0501442 CCC: $30.25 © 2005 American Chemical Society
Published on Web 08/13/2005
J. Org. Chem. 2005, 70, 7479-7487
7479

Williams et al.
antibiotic therapy, these new agents challenge existing
methods of synthesis. In the most convergent approach
to these compounds, construction of a carbon-sulfur bond
is made more difficult by the increased complexity of the
reaction partners and the increased steric demands on
bringing them together. The introduction of charged and
nucleophilic functionality, while important in giving the
antibiotic the desired properties, has made purification
and isolation of the product more difficult. Herein, we
describe our strategy and discoveries leading to a practi-
cal synthesis for the new antibiotic ertapenem sodium 1
allowing production on multi-kilogram scale.11
Results and Discussion
Ertapenem is an amino dicarboxylic acid and as such
can exist in four forms in solution. The zwitterionic
monocarboxylate and the dicarboxylate salts are the
predominant forms in the pH range where the best
solution stability is observed.12 The monosodium salt is
the predominant form in the range from pH 4 to 7; the
disodium salt becomes the predominant form above pH
7. Ertapenem was found to crystallize as the monosodium
salt from a concentrated aqueous solution at pH 5.5 on
addition of a specific combination of alcohol solvents.
Attempts to crystallize other salts were not successful.
The monosodium salt, therefore, became the target for
isolation of the product.
Since isolation of the product would require achieving
a concentration in excess of 100 mg/mL, we examined
the solution stability at that concentration and found that
the rate of degradation in aqueous solution at 2 °C is
about 1%/h across the pH range which affords the best
solution stability (pH 5-8). Although hydrolysis contrib-
utes to the degradation of ertapenem, the formation of
dimeric degradates accounts for most of the loss in this
pH range at high concentration.13
As with Meropenem, attack of the pyrrolidine amine
on the â-lactam is responsible for formation of dimeric
degradates.14 In the case of ertapenem, other degradates
resulting from bimolecular reactions appear to arise from
attack by the two carboxylates on the â-lactam, leading
initially to anhydrides, which can then undergo intramo-
lecular acyl transfer to give the more stable products that
are observed.15 The degradates have been characterized
by NMR analysis of samples isolated by preparative
HPLC.12
FIGURE 1.
Since imipenem was introduced in the 1980s, there has
been a substantial effort to identify carbapenem antibiot-
ics with improved pharmacokinetics while retaining a
broad spectrum of activity. The introduction of a â-methyl
substituent to the carbapenem nucleus proved to be a
key advance in achieving this goal.6 Meropenem, discov-
ered at Sumitomo Pharmaceuticals, was the first car-
bapenem antibiotic to be developed for clinical use as
single agent therapy.7 Multiple daily dosing, however,
was still required.
Ertapenem sodium8 is a new broad-spectrum carbap-
enem antibiotic that has demonstrated efficacy against
the growing number of cephalosporin-resistant bacteria.9
In clinical trials, the antibiotic showed an improved
pharmacokinetic profile in comparison with other car-
bapenems allowing single agent therapy and once-daily
dosing.10 Ertapenem sodium is the active ingredient in
INVANZ which was recently approved in the United
States as an IV or IM treatment for moderate to severe
upper and lower respiratory tract, urinary tract, skin,
obstetric, and gynecologic infections. While the â-methyl
carbapenems clearly represent a significant advance in
(5) Norrby, S. R.; Alestig, K.; Bjornegard, B.; Burman, L. A.; Ferber,
F.; Huber, J. L.; Jones, K. H.; Kahan, F. M.; Kahan, J. S.; Kropp, H.;
Meisinger, M. A.; Sundelof, J. G. Antimicrob. Agents Chemother. 1983,
23, 300-7.
(6) Shih, D. H.; Baker, F.; Cama, L.; Christensen, B. G. Heterocycles
1984, 21, 29-40.
(7) (a) Sunagawa, M.; Matsumura, H.; Inoue, T.; Fukasawa, M.;
Kato, M. J. Antibiot. (Tokyo) 1990, 43, 519-32. (b) Sunagawa, M.;
Matsumura, H.; Inoue, T.; Fukasawa, M.; Kato, M. J. Antibiot. (Tokyo)
1991, 44, 459-62. (c) Fukasawa, M.; Sumita, Y.; Harabe, E. T.; Tanio,
T.; Nouda, H.; Kohzuki, T.; Okuda, T.; Matsumura, H.; Sunagawa, M.
Antimicrob. Agents Chemother. 1992, 36, 1577-9.
(8) Ertapenem was discovered at Zeneca Pharmaceuticals (now
AstraZeneca) and was licenced by Merck. Betts, M. J.; Davies, G. M.;
Swain, M. L. Antibiotic Compounds. U.S. Patent 5,478,820, 1995.
(9) (a) Jones, R. N. J. Chemother. 2001, 13, 363-76. (b) Hilliard, N.
J.; Johnson, C. N.; Armstrong, S. H.; Quarles, S.; Waites, K. B. Int. J.
Antimicrob. Agents 2002, 20, 136-40.
Recognizing the limitations that product stability
would impose on purification and isolation of the product,
(11) For a recent review covering synthesis of carbapenems, see:
Singh, G. S. Mini-Rev. Med. Chem. 2004, 4, 69-92. A scaleable
procedure for preparation of the carbapenem doripenem, avoiding
chromatographic purification, was recently reported: Nishino, Y.;
Kobayashi, M.; Shinno, T.; Izumi, K.; Yonezawa, H.; Masui, Y.;
Takahira, M. Org. Process Res. Dev. 2003, 7, 846-850.
(12) Crocker, L. S.; Wang, Y.; McCauley, J. A. Org. Process Res. Dev.
2001, 5, 77-79.
(13) Sajonz, P.; Natishan, T. K.; Wu, Y.; Williams, J. M.; Pipik, B.;
DiMichele, L.; Novak, T.; Pitzenberger, S.; Dubost, D.; Almarsson, O¨ .
J. Liq. Chrom. Relat. Technol. 2001, 24, 2999-3015.
(14) Takeuchi, Y.; Sunagawa, M.; Isobe, Y.; Hamazume, Y.; Noguchi,
T. Chem. Pharm. Bull. 1995, 43, 689.
(10) Gill, C. J.; Jackson, 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. (b) Majumdar, A. K.; Musson,
D. G.; Birk, K. L.; Kitchen, C. J.; Holland, S.; McCrea, J.; Mistry, G.;
Hesney, M.; Xi, L.; Li, S. X.; Haesen, R.; Blum, R. A.; Lins, R. L.;
Greenberg, H.; Waldman, S.; Deutsch, P.; Rogers, J. D. Antimicrob.
Agents Chemother. 2002, 46, 3506-11.
(15) An anhydride has been proposed as an intermediate in the
bimolecular degradation of imipenem: (a) Smith, G. B.; Schoenewaldt,
E. F. J. Pharm. Sci. 1981, 70, 272-6. (b) Ratcliffe, R. W.; Wildonger,
K. J.; DiMichele, L.; Douglas, A. W.; Hajdu, R.; Goegelman, R. T.;
Springer, J. P.; Hirshfield, J. J. Org. Chem. 1989, 54, 653-660. (c)
Smith, G. B.; Dezeny, G. C.; Douglas, A. W. J. Pharm. Sci. 1990, 79,
732-40.
7480 J. Org. Chem., Vol. 70, No. 19, 2005

New Carbapenem Antibiotic Ertapenem Sodium
SCHEME 1
we set out to address this problem in design of the
synthesis. Since the removal of protecting groups would
likely detract from the yield overall and create byproducts
that would have to be removed in purification of the
product, we intended to minimize the use of protecting
groups. At the outset, we realized the importance of each
reaction consistently performing near perfection, not only
to achieve the highest possible yield, but also to limit the
need for operations to remove undesired side-products.
The processing time associated with these operations was
expected to be very costly in terms of degradation and,
thus, yield. Herein we describe how this strategy led to
an exceedingly practical synthesis which forms the basis
for the manufacture of this important compound.
1st Generation Synthesis. Carbapenems related to
imipenem have generally been constructed through reac-
tion of a thiol with an activated form of the bicyclic
carbapenem nucleus.16 Since appearing in the first
reported synthesis of a â-methyl carbapenem,7 the enol
phosphate 2 has seen extensive use in the preparation
of a large number of carbapenems for testing as antibac-
terial agents3,4 and is now commercially available in bulk
quantities.17 For the coupling reaction, functionality in
the thiol has generally been protected to avoid competing
reactions and to improve the organic solubility and
stability of the product after coupling. The thiols required
for the synthesis of ertapenem were derived through a
highly efficient process starting with trans-4-hydroxyl-
L-proline as previously reported.18 This process allowed
preparation of thiol substrates for the coupling reaction
arrayed both with and without protecting groups at the
carboxylic acid and amine functions. Initially, we chose
to protect the amine as a p-nitrobenzyl carbamate al-
lowing deprotection concurrent with hydrogenolysis of the
p-nitrobenzyl ester. Considering the difference in reactiv-
ity between carboxylate and thiolate toward soft electro-
philes, we expected that we would be able to define
conditions for the coupling reaction favoring the desired
reaction so that protection of the carboxylic acid would
not be required.19
compromised selectivity for the desired reaction. Al-
though some improvement in rate could be achieved by
using more base, the reduction in reaction time was
modest and consistently achieving complete conversion
remained a problem.
The first solution to this problem came from an
experiment with a very different goal. After completion
of the reaction, we found that three extractions were
required to remove the amine salts and solvent (DMF)
prior to hydrogenolysis of the protecting groups. It was
necessary to perform extractions at this stage since the
product of the hydrogenolysis would be water soluble and
removal of DMF from the aqueous solution would not be
practical. We were interested in avoiding these extrac-
tions, and it occurred to us that it might be possible to
crystallize the product from the reaction mixture as a
carboxylate salt if a suitable amine were used as base in
the coupling. Filtration then would afford the product free
of the coupling reaction solvent and all but one equiv of
amine. Dicyclohexylamine often gives crystalline car-
boxylate salts. In the reaction using dicyclohexylamine
as base, crystalline solids were observed, but these solids
proved to be the diphenyl phosphate salt of dicyclohexy-
lamine and not the salt of the carbapenem product.
Although the experiment did not produce the desired
result, we observed that the coupling reaction was much
faster, achieving complete conversion in only 4 h. Diiso-
propylamine is considerably less expensive and lower in
molecular weight compared with dicyclohexylamine and
proved to be an excellent base for this reaction. We did
not uncover any evidence for reaction with the â-lactam.
It was clearly important to consistently achieve high
conversion in the coupling reaction to avoid carrying
unreacted thiol into the hydrogenation where it would
undoubtedly poison the catalyst.20 Following standard
conditions reported in the carbapenem literature, reac-
tions of 3 with 2 using a moderate excess (2.1 equiv) of a
tertiary amine base such as diisopropylethylamine or
triethylamine in DMF at -30 °C proved to be slow,
requiring 24-48 h to approach complete conversion
(Scheme 1). Increasing the temperature significantly
The rate increase observed in using dicyclohexylamine
or diisopropylamine suggests that the secondary amines
afford a substantially higher proportion of thiolate in the
rate-limiting equilibrium between thiol and thiolate
compared with the tertiary amine bases. The dissociation
constants reported for diethylamine and triethylamine
in DMF are 10.4 and 9.25, respectively.21 The dissociation
constant for thiol 3 in DMF was estimated to be 11.2.22
(16) An alternate, less convergent approach has been reported:
Prashad, A. S.; Vlahos, N.; Fabio, P.; Feigelson, G. B. Tetrahedron Lett.
1998, 39, 7035-7038.
(17) Interest in the synthesis of 1-â-methyl carbapenems generated
a variety of solutions to this problem: Berks, A. H. Tetrahedron 1996,
52. The carbapenem nucleus has become commercially available with
the carboxylic acid protected as the p-nitrobenzyl ester from Nisso,
Kaneka, and Takasago.
(18) (a) Brands, K. M. J.; Marchesini, G.; Williams, J. M.; Dolling,
U.-H.; Reider, P. J. Tetrahedron Lett. 1996, 37, 2919-2922. (b) Brands,
K. M.; Jobson, R. B.; Conrad, K. M.; Williams, J. M.; Pipik, B.;
Cameron, M.; Davies, A. J.; Houghton, P. G.; Ashwood, M. S.; Cottrell,
I. F.; Reamer, R. A.; Kennedy, D. J.; Dolling, U. H.; Reider, P. J. J.
Org. Chem. 2002, 67, 4771-6.
(19) Pearson, R. G. J. Am. Chem. Soc. 1963, 85, 3533-3539.
(20) Rylander, P. N. Catalytic Hydrogenation in Organic Synthesis;
Academic Press: New York, 1979.
(21) Demange-Guerin, G. Talanta 1970, 17, 1075-1084. The dis-
sociation constant for diisopropylamine in DMF has not been reported.
(22) The dissociation constants for the carboxylic acid and thiol
functions were determined as a function of DMF concentration in water
by titration. Extrapolation to 100% DMF gave dissociation constants
of 8.2 and 11.2, respectively. McCauley, J. A. Unpublished results (May
5, 1994). For validation of this approach, see: Sistovaris, N.; Hamachi,
Y.; Kuriki, T. Fresenius J. Anal. Chem. 1991, 340 345-349.
J. Org. Chem, Vol. 70, No. 19, 2005 7481

Williams et al.
SCHEME 2
The use of secondary amines for formation of the C-S
bond in the synthesis of carbapenems is unprecedented.
After complete conversion had been achieved, the
reaction was quenched into a mixture of dilute phosphoric
acid and ethyl acetate. DMF and amine salts were
removed at this stage through multiple extractions. The
ethyl acetate solution was then combined with Pd on
carbon catalyst in water for hydrogenolysis of both
p-nitrobenzyl ester and p-nitrobenzyl carbamate. The pH
of this biphasic reaction was controlled between 7.1 and
7.2 during the reaction by the addition of NaOH.
Following hydrogenolysis, two unanticipated problems
were encountered. The filtration to remove the catalyst
was very slow and the level of Pd in the filtrate was
unacceptably high. Both problems were overcome by
adjusting the pH to <6 prior to filtration. The Pd level
in the resulting filtrate was dramatically reduced (from
>1000 to <2 ppm); the filtrate was no longer black but
light yellow. At the lower pH, however, the product was
less stable. Adjusting the pH to 7-8 improved solution
stability, but the salts introduced in the two pH adjust-
ments added to the load of impurities that had to be
removed from the aqueous solution to allow isolation of
the product. The time required for these operations
contributed to the increasing level of degradates. In lab-
scale experiments, the assay yield of ertapenem was 75
to 80%.23 Yields were somewhat lower with more degra-
dates formed on larger scale where each operation
requires more time.
Following the catalyst filtration, the product was
purified using a hydrophobic resin column. The best
purification was achieved by first extracting residual
ethyl acetate using dichloromethane. Ethyl acetate af-
fected retention on the polystyrene-based column; whereas,
residual dichloromethane did not. Since the product has
charged functionality at the pH of the solution (pH 7-8)
loaded onto the column, it is not readily adsorbed on the
solid resin stationary phase and the presence of ethyl
acetate adversely affects retention on the column. The
column was eluted using dilute acetone (2%) in aqueous
sodium bicarbonate (0.05 M) to give a solution of ertap-
enem with reduced levels of degradates, reaction byprod-
ucts, and salts in 80-85% recovery. The solution, how-
ever, was quite dilute (5-10 g/L), and significant
concentration was required before the product could be
crystallized. Nanofiltration, with diafiltration used to
reduce the level of inorganic salts, allowed concentration
of the aqueous solution at below 10 °C.24 Initially, this
concentration step was conducted at about pH 5.5 to
allow more complete removal of salts. Later, we found
that the pH of the rich cut could be adjusted to <6 using
HCl and then back to pH 7 using NaOH to improve the
stability of ertapenem during nanofiltration. These op-
erations converted sodium bicarbonate, used in elution
of the hydrophobic resin column, to sodium chloride
which was more efficiently removed by nanofiltration.
Nonetheless, the time required for these operations,
particularly as the concentration increased, resulted in
degradation of the product. When the desired concentra-
tion range was reached, the pH of the solution was
adjusted to 5.5 using acetic acid and the monosodium salt
was then crystallized from solution by adding methanol
and 1-propanol. A costly recrystallization was typically
required to produce pure product.
2nd Generation Synthesis. The productivity of the
1st generation synthesis was clearly limited by the
hydrophobic resin column purification. The subsequent
concentration (about 20×) that was needed in order to
crystallize the product was inefficient and resulted in
product degradation. The resin column was primarily
introduced to remove byproducts of the hydrogenolysis
reaction. We expected therefore that the potential for
eliminating the resin column from the isolation process
would be improved if the coupling reaction could be
performed without a protecting group on the pyrrolidine
amine. Furthermore, the pH adjustment, which had been
introduced to control the level of solubilized palladium,
increased the degradate and inorganic salt load thus
limiting options for improving efficiency in purification
and isolation of the product. Recovery in the recrystal-
lization of 1, which was often required to remove degra-
dates from the isolated product, was only about 80%
providing a clear incentive for devising a synthesis where
recrystallization would not be required.
In our initial effort to make use of the hydrochloride
salt 5, we identified serious challenges (Scheme 2). The
product of the coupling reaction was now an amino acid,
and it was no longer possible to use simple extractions
to remove DMF and amine salts after the coupling
without substantial losses of product. Therefore, the
solvent and salts from the coupling reaction would have
to be removed after the hydrogenolysis. N-Ethylpyrroli-
dinone (NEP) was chosen as solvent to allow extraction
from an aqueous solution containing ertapenem following
hydrogenolysis. The list of bases suitable for use in the
coupling and affording salts that could be removed from
an aqueous solution was quite limited. We expected that
sodium hydroxide would be a sufficiently strong base to
give complete deprotonation of the thiol for the coupling
(23) For consistency, yields reported throughout are based on 2.
(24) Antonucci, V.; Yen, D.; Kelly, J.; Crocker, L.; Dienemann, E.;
Miller, R.; Almarrson, O¨ . J. Pharm. Sci. 2002, 91 (4), 923-932.
7482 J. Org. Chem., Vol. 70, No. 19, 2005

New Carbapenem Antibiotic Ertapenem Sodium
SCHEME 3
FIGURE 2.
and that sodium salts could be removed from the aqueous
product stream by nanofiltration after the hydrogenoly-
sis.
In proposing to use sodium hydroxide in this reaction,
however, it was obvious that we would need to change
the way the coupling reaction was carried out. The base
could clearly not be added to a cold mixture of enol
phosphate and thiol. Thus, it became necessary to form
the thiolate separately to avoid hydrolysis of the carba-
penem; addition of more than 3 equiv of a solution of
sodium hydroxide in water to a solution of thiol 5 in NEP
after rigorous degassing was found to ensure complete
conversion in the subsequent coupling reaction. Any
excess NaOH, however, led to hydrolysis of the carba-
penem. We discovered that isopropyl acetate could be
added to the thiolate solution to consume excess NaOH;
the resulting sodium acetate and 2-propanol formed were
innocuous in the coupling. When an NEP solution of the
carbapenem enol phosphate 2 was added to the cold NEP
solution of the thiolate, complete conversion could be
achieved at low temperature. Good selectivity was ob-
served as long as the temperature was maintained below
-40 °C. The water introduced with the sodium hydroxide
to ensure solubility of the sodium salts, however, caused
the reaction mixture to become very viscous below -45
°C. After the reaction was complete, the mixture was
quenched into aqueous buffer, and the resulting solution
was used directly in the hydrogenation without extractive
purification.
In preliminary experiments, the overall yield after
hydrogenolysis was significantly lower than had been
achieved with the 1st generation synthesis (45%, cf. 75-
80%) even though the coupling reaction was performing
well. Dimer degradates and one side-product were ob-
served at significantly higher levels. It was at this stage
that an important and timely discovery was made. It was
found that the formation of dimer degradates arising
from reaction of the pyrrolidine amine with the carba-
penem was suppressed in solutions containing sodium
bicarbonate. We proposed that this observation could be
explained by the formation of the sodium carbamate 8
(Figure 2), which was subsequently confirmed by NMR.25
The formation of the carbamate affords protection of the
pyrrolidine amine inhibiting attack on the carbapenem.26
The addition of 3 equiv of sodium bicarbonate prior to
hydrogenolysis suppressed the formation of dimer deg-
radates as well as the main side-product of the reaction
so that yields became comparable to those achieved in
the 1st generation synthesis where the pyrrolidine amine
had been protected as the p-nitrobenzyl carbamate.27 The
bicarbonate also serves to maintain the pH between 8
and 7 making it unnecessary to add sodium hydroxide
during the hydrogenolysis.
Conclusive evidence for formation of the carbamate 8
1
was obtained in a H-13C heteronuclear multiple-bond
correlation (HMBC) 2-D NMR experiment.28 Three-bond
correlations were observed between the carbamate car-
bonyl carbon and protons on the proximal pyrrolidine
carbons (2′ and 5′).
The main side-product of the hydrogenolysis was later
identified. Mass by LCMS (M + 1 ) 581) suggested that
the product was a p-aminobenzyl derivative of ertap-
enem. The fragmentation pattern, however, indicated
that this compound was not simply the intermediate
resulting from reduction of 7 (i.e. 11, Scheme 3). Struc-
ture 9 is consistent with the fragmentation pattern in
the mass spectrum and was subsequently confirmed by
(27) For background on the kinetics of carbamate formation see: (a)
Caplow, M. J. Am. Chem. Soc. 1968, 90, 6795-6803. (b) Ewing, S. P.;
Lockshon, D.; Jencks, W. P. J. Am. Chem. Soc. 1980, 102, 3072-3084.
For characterization of carbamates by NMR, see: (c) Archer, R. A.;
Kitchell, B. S. J. Org. Chem. 1966, 31, 3409-3410. (d) Lemieux, R.
U.; Barton, M. A. Can. J. Chem. 1971, 49, 767-776. (e) Morrow, J. S.;
Keim, P.; Gurd, F. R. N. J. Biol. Chem. 1974, 249, 7484-7494. (f) Pratt,
R. F.; Dryjanski, M.; Wun, E. S.; Marathias, V. M. J. Am. Chem. Soc.
1996, 118, 8207. Methods for the preparation of alkyl carbamates from
amines using carbon dioxide have been reported: (g) McGhee, W.;
Riley, D.; Christ, K.; Pan, Y.; Parnas, B. J. Org. Chem. 1995, 60, 2820-
2830. (h) Dean, D. C.; Wallace, M. A.; Marks, T. M.; Melillo, D. G.
Tetrahedron Lett. 1997, 38, 919-922. (i) Salvatore, R. N.; Shin, S. I.;
Nagel, A. S.; Jung, K. W. J. Org. Chem. 2001, 66, 1035-1037.
(28) Bax, A.; Summers, M. F. J. Am. Chem. Soc. 1986, 108, 2093-
2094.
(25) The carbamate form of meropenem was also characterized.
Almarsson, O¨ .; Kaufman, M. J.; Stong, J. D.; Wu, Y.; Mayr, S. M.;
Petrich, M. A.; Williams, J. M. J. Pharm. Sci. 1998, 87, 663-666.
(26) The use of carbon dioxide in protection of an amine is very
limited. Formation of a carbamate in the hydrogenation of a nitro group
to a primary amine was used to suppress hydrolysis of the nitro olefin
starting material catalyzed by the amine product. Semikolenov, V. A.;
Simokova, I. L.; Golovin, A. V.; Burova, O. A.; Simirnova, N. M. In
Heterogeneous Catalysis and Fine Chemicals; Blaser, H. U., et al., Eds.;
Stud. Surf. Sci., Catal. 1997, 108, 225-262. The influence of bicarbon-
ate and carbon dioxide on reactions containing amines is often not
recognized or considered in interpreting results.
J. Org. Chem, Vol. 70, No. 19, 2005 7483

Williams et al.
SCHEME 4
the reaction of carbon dioxide at the secondary hydroxyl
to give 14. The resulting anhydride 15 would then
undergo rearrangement to afford 13.33 This degradate is
typically observed at about 0.5-1% by area relative to 1
in the hydrogenolysis reaction mixture. The rate of
formation appears to be primarily dependent on the
concentration of carbon dioxide as would be expected from
the proposed mechanism.
A Practical Synthesis. We had made significant
advances in the 2nd generation synthesis, and discoveries
from this effort would develop into key elements for a
practical synthesis. The coupling reaction, however, was
clearly quite complex and not suitable for consistent,
reliable production. There was still a potential strategic
advantage in avoiding protection of the pyrrolidine
amine. We had shown that protection was not required
for the coupling reaction, but selectivity was very much
dependent on temperature. Reducing the temperature
further was not an option as long as sodium hydroxide
was used as base. It was clear that a nonnucleophilic,
organic-soluble base would offer significant advantages.
1,1,3,3-Tetramethylguanidine (TMG) proved to be an
excellent alternative allowing short reaction times (less
than 2 h) at temperatures as low as -60 °C. When TMG
was used as base in the coupling reaction, the yield
through hydrogenolysis improved from 75-80% to about
90%.
NMR analysis of an isolated sample.29 This side-product
most likely results from alkylation of the pyrrolidine
amine during hydrogenolysis. Reduction of the nitro
group gives the hydroxylamine 10 which is observed as
an intermediate in the reaction mixture.30 Further reduc-
tion of 10 would produce 11 which has not been observed.
Fragmentation of 11 as depicted in Scheme 4 would
produce 12 which would react readily with available
nucleophiles.31 Reaction at the pyrrolidine amine would
give 9. Formation of the carbamate 8 makes the pyrro-
lidine unavailable and thereby inhibits the formation of
9.
Amines are known to affect the performance of pal-
ladium-catalyzed hydrogenation reactions.32 Often hy-
drogenation reactions of compounds containing the amine
function or producing an amine-containing product are
performed under acidic conditions so that the amine is
protonated limiting interaction with the catalyst. Al-
though we do not have direct evidence of a beneficial
effect on catalyst performance in this case due to the
complexity of the reaction, it seems reasonable to believe
that formation of a carbamate would also suppress
interaction of an amine with the catalyst and thereby
improve catalyst performance.
With the coupling and hydrogenolysis reactions per-
forming well and the impurity load consequently reduced,
the goal of eliminating the resin purification was begin-
ning to appear attainable, but the challenges in avoiding
the pH adjustment to reduce solubilized palladium levels
and in removing TMG from the aqueous solution after
the hydrogenolysis remained.
Control of the level of palladium in the filtrate after
hydrogenolysis continued to require pH adjustment prior
to filtration. This procedure added processing time and
introduced inorganic salts that had to be removed in
order to crystallize the product. More importantly, the
rate of degradation in the reaction mixture increased
substantially below pH 7 compounding the impurity load
on purification. In trying to understand the source of
solubilized palladium, a practical solution was realized.
We suspected that colloidal Pd(0) was responsible for the
black color of the filtrate, and this Pd(0) might arise from
the reduction of Pd(II) that had been leached from the
catalyst support by some ligand in the reaction mixture.
If the Pd were reduced on the carbon support to Pd(0)
prior to exposure to constituents of the reaction mixture
capable of ligating Pd(II), the level of solubilized Pd might
be reduced. Some improvement was observed in using
commercially available prereduced catalysts, but the low
levels required were realized only when the catalyst was
reduced in the reactor and not exposed to air before
introducing the substrate. When the catalyst was pre-
reduced in this way, the reaction mixture could be filtered
at pH 7-8 to give a yellow solution nearly free of
palladium. Indeed, we subsequently showed that ertap-
The use of bicarbonate in the process was beneficial
overall, but there was one adverse consequence that we
came to realize. One degradate, identified as the oxazi-
none 13 (Scheme 4), was the direct result of reaction of
the carbapenem with carbon dioxide. A sample was
prepared using high-pressure carbon dioxide and the
structure confirmed by NMR and LCMS analysis. We
propose that the â-lactam ring is opened as a result of
(29) Fragments of mass 370 and 338 amu, corresponding to scission
of the C-S bonds, were consistent with 9. Consecutive losses of CO2
(535 and 491 amu) were also observed suggesting the presence of two
carboxylic acids.
(30) By LCMS M + 1 ) 597 with fragments of mass 265 and 233
amu, corresponding to cission of the C-S bonds.
(31) (a) Filar, L. J.; Winstein, S. Tetrahedron Lett. 1960, 1 (46),
9-16. (b) Dome´, M.; Wakselman, M. Bull. Soc. Chim. Fr. 1975, 576-
582. (c) Le Corre, G.; Guibe-Jampel, E.; Wakselman, M. Tetrahedron
1978, 34, 3105-3112.
(32) Czech, B. P.; Bartsch, R. A. J. Org. Chem. 1984, 49, 4076-
4078 and references therein.
(33) (a) Johnson, D. A.; Hardcastle, G. A., Jr. J. Am. Chem. Soc.
1961, 83, 3534-3535. The carbamate intermediate was subsequently
characterized by NMR: See ref 27c,e.
7484 J. Org. Chem., Vol. 70, No. 19, 2005

New Carbapenem Antibiotic Ertapenem Sodium
SCHEME 5
enem solubilizes palladium from an unreduced Pd-on-
carbon catalyst.34
Ion exchange chromatography held some promise as a
means for removing TMG. We found, however, that the
loading would limit productivity and the high salt
concentration required to elute the column would make
extended diafiltration necessary.
A lead in solving the TMG removal problem came from
an observation made in experiments intended to define
an efficient means for extracting NEP from the aqueous
solution. In screening a series of solvents, we found that
1 equiv of TMG (out of 3.3 charged to the coupling
reaction) was readily extracted from the aqueous solu-
tion. The 1 equiv of diphenyl phosphate (DPP) present
as a normal side product of the coupling reaction was
extracted simultaneously. Additional extractions did not
reduce the level of TMG in the aqueous layer. This
observation suggested that the TMG was forming an
organic soluble ion pair with DPP.35 We found that other
ion-pairing agents (lipophilic sulfonates and carboxylates)
were not as efficient as DPP in partitioning the TMG into
the organic layer and some created serious emulsion
problems. Although commercial DPP could be added to
improve extraction efficiency, we recognized that in
production sufficient DPP would be produced in 3-4
batches to efficiently remove TMG for a single batch if
the DPP were recycled into the process from the waste
streams. An extraction protocol was developed that uses
a solution of DPP in isoamyl alcohol to reduce the TMG
level from about 3.5 to 0.2 equiv (relative to 1) while
concentrating the aqueous solution about 4x as needed
for a direct crystallization of the product with minimal
losses to the organic extract. This extraction also removes
byproducts of the hydrogenolysis. Thus, this single
operation eliminates both the inefficient resin column
purification and nanofiltration concentration operations
required in the earlier syntheses. The extraction is best
performed as a continuous operation using a centrifugal
extractor.36
experiments were performed under conditions expected
to favor specific reactions and the resulting side products
were characterized by LCMS.
Addition of the carbapenem enol phosphate 2 or the
preformed thiolate of 5 (Scheme 5) to a reaction mixture
after conversion was complete resulted in the formation
of products corresponding to adducts of each with the
product of the coupling reaction (M + 1 ) 1205 and 877,
respectively). The latter product was correlated with a
product after hydrogenolysis that was also observed when
ertapenem 1 was reacted with 5 in water at pH 7.5.
Warming a typical coupling reaction mixture after com-
plete conversion had been achieved resulted in the
increase of a product (M + 1 ) 1221) believed to come
from reaction of two molecules of the desired product 16.
When base was added at higher temperature (-25 rather
than -60 °C), a second product believed to be the result
of a reaction between the carbapenem enol phosphate 2
and thiol 5 (M + 1 ) 861) is observed at higher than
typical levels. Clearly, the ability to maintain the low
temperature during the reaction is important in achiev-
ing high selectivity for the desired product and in
minimizing further reaction of the product. It is also
important to use a sufficiently strong base to give a fast
reaction and to add the base as quickly as possible to
minimize reactions between starting materials and prod-
uct.
With the basis for a practical synthesis established,
we examined the details of each reaction to understand
the controlling features and ensure optimal performance.
The success of the coupling reaction derives from achiev-
ing near perfect chemoselectivity for the desired reaction.
The enol phosphate contains two electrophilic sites, and
the thiol contains three nucelophilic sites. The product
of each of the possible combinations would retain the
potential to react further. Under less than optimal
conditions, many of the possible products of these un-
desired reactions are indeed observed. Although none of
the side products have been unambiguously identified,
The amount of the disulfide 1818b (Figure 3) corre-
(34) The level of palladium in solution for a slurry of 10% Pd on
carbon in water containing ertapenem and 2.5 equiv of sodium
bicarbonate (pH 7.1) was more than 100× higher than the control (Pd
undetected < 0.3 mg/L).
(35) For a review covering the factors influencing partitioning for
charged species see: Leo, A.; Hansch, C.; Elkins, D. Chem. Rev. 1971,
71, 525-616.
FIGURE 3.
(36) (a) Jacobsen, F. M.; Beyer, G. H. AIChE J. 1956, 2, 283-289.
(b) Todd, D. B.; De Cicco, R. W.; Daiser, H. R. Chem. Eng. Prog. 1965,
61 (5), 74-77. (c) Podbielniak, W. J.; Kaiser, H. R.; Ziegenhorn, G. J.
In Chemical Engineering Progress Symposium Series: The History of
Penicillin Production; AIChE: New York, 1970; Vol. 66 (100), p 45-
50. (d) Todd, D. B. Chem. Eng. 1972, July 24, 152-158. (e) King, M.
L.; Foreman, A. L.; Orella, C.; Pines, S. H. Chem. Eng. Prog. 1985, 81
(May), 36-39.
sponding to thiol 5 which was formed in the coupling
reaction was variable even when the solvent was rigor-
ously degassed prior to use. We found a report that
N-methylpyrrolidinone (NMP) forms hydroperoxides upon
exposure to air and NEP would be expected to do the
J. Org. Chem, Vol. 70, No. 19, 2005 7485

Williams et al.
TABLE 1. Bicarbonate and Hydrogenolysis
Performance
synthesis resulted in a practical and highly productive
synthesis for manufacturing an unstable carbapenem.
NaHCO3 NaOH pH > 7c 1 yieldd
10
9
dimersf
entrya (equiv)b
(equiv)
(%)
area %e area %e area %e
Experimental Section
1
2
3
4
5
0
1.0g
0.8
0.3
0
77
85
90
91
90
0.6
0.0
0.0
0.2
0.1
5.9
3.7
1.4
0.9
1.2
4.4
1.6
1.0
0.8
0.1
1.0
2.0
3.0g
4.0
All commercially available materials were used as received.
Compound 5 was prepared as described previously.18b
(4R,5S,6S,8R,2′S,4′S)-3-[[2-[[(3-Carboxyphenyl)amino]-
carbonyl]pyrrolidin-4-yl]thio]-4-methyl-6-(1-hydroxy-
ethyl)-7-oxo-1-azabicylclo[3.2.0]hept-2-en-2-carboxylic
Acid Sodium Salt (Ertapenem Sodium 1). N-Ethylpyrro-
lidinone (NEP, 1.33 L) containing 0.5% (w/v) water was
charged into a 3 L flask equipped with an air-driven overhead
stirrer, thermocouple, and a nitrogen/vacuum inlet. The tem-
perature was adjusted to about 0 °C and [4R,5S,6S,8R)-3-
[(diphenoxyphosphinyl)oxy]-6-(1-hydroxyethyl)-4-methyl-7-oxo-
1-azabicyclo[3.2.0]hept-2-en-2-carboxylic acid (4-nitro-phenyl)
methyl ester (2, 170.0 g, 0.286 mol) was charged. After inertion
(vacuum/nitrogen), 0.7 mL of tri-n-butylphosphine followed by
(2S-cis)-3-[[(4-mercapto-2-pyrrolidinyl)carbonyl]amino]ben-
zoic acid monohydrochloride (5, 85.8 g, 0.280 mol) was charged.
The flask was inerted (vacuum/nitrogen), and the solution was
cooled to -55 to -60 °C. 1,1,3,3-Tetramethylguanidine (TMG,
110 g, 0.936 mol) was charged with vigorous agitation main-
taining the internal temperature below -50 °C. The reaction
mixture was maintained at -50 to -55 °C for 1 h and warmed
to -40 °C. The reaction was complete as determined by HPLC
analysis (YMCbasic; 250 × 4.6 mm; 5 µm particle size; UV
detection at 254 nm, 0.05% phosphoric acid/acetonitrile 85:15
for 5 min then gradient to 20:80 over 15 min; 1.2 mL/min;
relative retention times: 5, 0.37; 18, 0.38; diphenyl phosphate,
0.96; 16, 1.00; 2, 1.50).
A 2 gal autoclave equipped with a pH probe was charged
with a slurry of 5% Pd on C catalyst (127 g, 48.3% dry wt, 29
mmol Pd) in 1.8 L of water. The vessel was purged with
nitrogen and placed under hydrogen (70 psi) for 1 h, then
vented and placed under nitrogen. Sodium hydroxide (5.0 N,
171 mL, 0.86 mol) was charged by pump maintaining the pH
below 9 with carbon dioxide. The temperature was adjusted
to 5 °C.
The coupling reaction mixture was charged into the auto-
clave maintaining the temperature below 15 °C and introduc-
ing carbon dioxide as needed to maintain the pH below 9. The
reaction flask was rinsed with 113 mL of NEP, and the pH
was adjusted to about 8 using carbon dioxide. The temperature
was adjusted to 5 °C and the vessel was placed under hydrogen
pressure (70 psi). The temperature was increased to 20 °C over
30 min. After 2 h, the pH was about 7, and the hydrogen was
vented and the vessel was purged with nitrogen. The reaction
mixture was cooled to 5 °C and activated carbon (19 g dry
weight) was charged. The mixture was filtered using a
stainless steel pressure filter, and the catalyst cake was
washed with 0.5 L of water.
0
a The reactions were performed in a 1 L autoclave with the
exception of entry 4 which was run in a 2 gal reactor. b NaHCO3
was generated from NaOH and carbon dioxide. Equiv are based
on NaOH charged. c NaOH added during the hydrogenolysis to
maintain the pH above 7. d Yield based on 2. e HPLC area %
relative to 1. f Degradates resulting from reaction of the pyrrolidine
amine and â-lactam. g The pH was adjusted with NaOH to 6.5
prior to charging the coupling reaction mixture into the autoclave.
same.37 The amount of disufide observed in the reaction
could be reduced by storing the NEP under nitrogen, but
some increase in comparison with what was present in
the starting material was still observed. The addition of
a small amount (1 mol %) of tri-n-butylphosphine prior
to charging the thiol suppressed formation of the disulfide
without adversely impacting performance of the hydro-
genolysis.38
The use of bicarbonate in the hydrogenolysis reaction
was key in achieving high yields with the unprotected
thiol. The effect of bicarbonate on the hydrogenolysis was
clearly demonstrated in a series of reactions using 0, 1,
2, 3, and 4 equiv of bicarbonate (Table 1). Sodium
hydroxide was added during the reactions as needed to
maintain the pH above 7. The yield improved with the
increasing charge of bicarbonate from 77% without
bicarbonate to 91% with 3 equiv. The increase in yield
was a result of suppressed dimer formation (4.4 to 0.8
area %) and a reduction in 9 from about 5.9 to 1.2 area
%. Although the level of dimers was further reduced by
using 4 equiv of bicarbonate, the yield was not improved.
Conclusions
Ertapenem presented a significant challenge in devel-
oping a practical synthesis for manufacturing. Our
strategy in minimizing the use of protecting groups
created opportunities that led to an efficient and highly
productive synthesis. By using TMG as base, the coupling
reaction is completed at a rate and temperature that
allows high selectivity for the desired product. The use
of bicarbonate in the synthesis affords protection of a
pyrrolidine amine as the carbamate allowing improve-
ments in productivity and performance in the hydro-
genolysis of a p-nitrobenzyl ester. The solution stability
of the final product was also improved reducing the
amount of degradation observed so that the extended
time required in manufacturing could be tolerated.
Overall, minimizing the burden on purification made it
possible to use a novel ion-pairing extraction for purifica-
tion and concentration prior to isolation of the product
by crystallization. The innovation in developing this
The combined filtrate and wash (3.8 L containing 128 g of
product, 90% yield) was extracted with 9.5 L of isoamyl alcohol
containing diphenylphosphoric acid (0.950 mol, 238 g), 50%
NaOH (563 mmol, 45 g), and water (120 g) using a counter-
current centrifugal extractor.39 The resulting aqueous solution
was extracted with 2.5 L of isoamyl alcohol to give an aqueous
solution of ertapenem (140 g/L).
The pH was adjusted to 5.5 with acetic acid. The product
was crystallized by adding volumes of methanol and 1-pro-
panol equal to the batch volume at -10 °C. The solid was
isolated by filtration and washed with 0.8 L of a mixture of
2-propanol and water (85:15 v/v), then with 1.3 L of methyl
acetate containing 2% (w/v) water at 5 °C. The solid was dried
with a nitrogen stream under vacuum to a final water content
of about 17%. The overall yield on multikilogram scale was
(37) (a) Drago, R. S.; Riley, R. J. Am. Chem. Soc. 1990, 112, 215.
(b) Patton, D. E.; Drago, R. S. J. Chem. Soc., Perkins Trans. 1 1993,
1611.
(38) Larger amounts of tri-n-butylphosphine poison the palladium
catalyst affecting performance of the hydrogenolysis.
(39) On lab scale, two CINC V-2 single stage centrifugal separators
(Costner Industries Texas, LP, formerly CINC.; http://www.cit-ind-
.com/) were used in series.
7486 J. Org. Chem., Vol. 70, No. 19, 2005

New Carbapenem Antibiotic Ertapenem Sodium
59-64% based on 2. On lab scale, yields were somewhat lower
as a result of physical losses in the extraction step: HPLC
analysis (YMCbasic; 250 × 4.6 mm; 5 µm particle size; UV
detection at 225 nm, 0.05% phosphoric acid/acetonitrile 90:10
for 1 min then gradient to 60:40 over 19 min; 1.0 mL/min)
relative retention times: TMG, 0.30; NEP, 0.68; hydrolysis
product, 0.76; 13, 0.84; ertapenem 1, 1.00; 9, 1.06; 10, 1.11;
dimer degradates, broad at 1.36; diphenyl phosphate, 1.53; 16,
1.94. UV (nm, H2O) 294; FT-IR (Nujol mull, cm-1) 3650-3600,
1751, 1695, 1559, 1459, 1377, 771; 1H NMR (500.13 MHz,
dioxane as reference at δ ) 3.75) δ 7.86 (m, 1H), 7.71 (m, 1H),
7.65 (m, 1H), 7.47 (t, 1H, J ) 8.0 Hz), 4.62 (t, 1H, J ) 8.3 Hz),
4.21 (om, 1H), 4.18 (dd, 1H, J ) 9.5, 2.4 Hz), 4.07 (m, 1H),
3.82 (dd, 1H, J ) 12.3, 6.8 Hz), 3.47 (dd, 1H, J ) 12.3, 5.6
Hz), 3.42 (dd, 1H, J ) 6.0, 2.4 Hz), 3.31 (m, 1H), 3.02 (m, 1H),
2.20 (m, 1H), 1.27 (d, 3H, J ) 6.4 Hz), 1.17 (d, 3H, J ) 6.8
Hz); 13C NMR (100.61 MHz) δ 177.3, 175.3, 168.4, 167.7, 138.4,
138.1, 137.0, 134.5, 130.0, 127.0, 124.9, 122.5, 65.9, 60.9, 59.5,
56.7, 53.2, 43.5, 41.4, 35.5, 20.9, 16.7; FTICR/MS calculated
for [C22H25N3O7S + H]+: 476.1486; found: 476.1498.
Acknowledgment. We would like to acknowledge
C. A. Bazaral, Jr., A. Houck, and A. Newell for perform-
ing hydrogenolysis reactions; J. Corte, J. Kukura, A.
Wieczorek, A. Mahajan, E. Fisher, and C. Orella for
their contribution in developing and performing the
centrifugal extraction; D. Dubost and O¨ . Almarrsson for
key observations on solution stability; J.A. McCauley
for determining dissociation constants for 3; T. Natis-
han, P. Sajonz, Y. Wu, Y. Wang, and T. Wang for
analytical support.
Supporting Information Available: Experimental de-
tails for the preparation or isolation of a reference sample and
characterization data supporting the proposed structures for
8, 9, and 13 are provided. This material is available free of
charge via the Internet at http:// pubs.acs.org.
JO0501442
J. Org. Chem, Vol. 70, No. 19, 2005 7487