The synthesis of GV143253A: A case study for the use of analytical and statistical tools to elucidate the reaction mechanism and to Optimize the process

Article abstract of DOI:10.1021/op100097c

GV143253A 1 is a broad-spectrum injectable I-lactam belonging to the class of trinem antibiotics. This article describes the work which enabled a detailed process understanding via several analytical techniques and the subsequent optimization of a key int

Full text of DOI:10.1021/op100097c

Organic Process Research & Development 2010, 14, 840–848
The Synthesis of GV143253A: A Case Study for the Use of Analytical and Statistical
Tools to Elucidate the Reaction Mechanism and to Optimize the Process
Giuseppe Guercio,*,† Alcide Perboni,† Francesco Tinazzi,† Luca Rovatti,‡ and Stefano Provera‡
Chemical DeVelopment and Analytical Chemistry, GlaxoSmithKline Medicines Research Centre, Via Fleming 4,
37135, Verona, Italy
Abstract:
niques). Our group developed an alternative route4 which has
been used for the synthesis of about 100 g of 1 (Scheme 1).
However, all synthetic steps suffered from low yields and
nonreproducible reaction conditions. In particular, stage 2 was
the most troublesome with a yield of ∼33%. In this step
ꢀ-enoneazetidinone 4 was converted, using a suitable oxalyl
chloride 5, into the oxalimide 6 (step 2a) that reacted with the
diethoxymethylphosphine without isolation to afford the un-
stable and again nonisolable phosphorane intermediate 7 that
was finally cyclized (step 2b) at higher temperature to give the
desired compound 8. To reach an acceptable cost of goods for
1, the required target yield for this stage was ∼60%. In order
to achieve this, we initiated a thorough mechanistic investigation.
GV143253A 1 is a broad-spectrum injectable ꢀ-lactam belonging
to the class of trinem antibiotics. This article describes the work
which enabled a detailed process understanding via several
analytical techniques and the subsequent optimization of a key
intermediate. By means of a combined application of 31P NMR
spectroscopy and MS spectrometry, the main impurities
have been identified and the reaction mechanisms clarified.
Moreover, a design of experiments (DoE) approach was
applied which substantially improved the overall yield.
1. Introduction
In recent years, bacteria have enhanced their mechanism of
resistance to the most common antibacterial drugs. In particular,
Methicillin-resistant Staphylococcus aureus (MRSA) represents
one of the most challenging and threatening diseases globally.1,2
Among the various compounds screened, exomethylenyl
trinem 13 has emerged as a very potent anti-Gram positive agent
with a remarkable activity against MRSA and also against some
important Gram negative pathogens.
In this contribution we describe the success of efforts made
with the aim of understanding the reaction mechanisms of two
critical steps to improve the synthetic route for large-scale
manufacturing. In particular, reaction monitoring was performed
in real time using both 31P NMR spectroscopy and MS
spectrometry. The output of these studies was combined with
a statistical tool in order to achieve a process improvement
ensuring an appropriate scalability for the production of clinical
batches.
3. Design of Experiment (DoE) Study on Step 2a
The synthesis of the oxalimide5 6 was originally conducted
using 2.0 equiv of FmOCOCOCl 5 and 2.2 equiv of N,N-
diisopropylethylamine (DIPEA) with a typical yield of 45-50%.
The conditions were optimized using a statistical approach
coupled to a parallel equipment6 (DoE, see Figure 1) to come
up with an increased yield of 90-95% when changing the
stoichiometry slightly (1.5 equiv of 5 and 2.0 equiv of DIPEA).
4. Analytical Techniques for the Elucidation of Step 2b
The literature-suggested mechanism for step 2b, the phos-
phorane formation 7 and its cyclization to 8, is depicted in
Scheme 2.7 The first equivalent of diethoxymethylphosphine
deoxygenates 6 to form a phosphine oxide and the highly
reactive carbene 10 which is readily trapped by a second
equivalent of the phosphine to form the phosphorane 7. This
unstable intermediate spontaneously cyclizes at higher temper-
ature, leading to the desired 8 plus a second equivalent of
phosphinoxide as a classical Wittig reaction.
The optimization of this synthetic step in the same way as
was done for the oxalimide formation failed. All the studied
parameters, e.g., temperature, equivalents of phosphine, solvents,
etc., did not give the desired outcome, the solution yield being
always lower than 50% and the isolated yield lower than 45%.
2. Synthesis
The original synthesis of 1 was a linear process starting from
the commercially available 4-acetoxyazetidinone. This synthetic
approach was not ideal for further scale-up, as all of the
intermediates possessed physical properties that were unsuitable
for purification (e.g. by using standard crystallisation tech-
* To whom correspondence should be addressed. E-mail: giuseppe.2.guercio@
gsk.com.
(4) Maragni, P.; Mattioli, M.; Pachera, R.; Perboni, A.; Tamburini, B.
Org. Process Res. DeV. 2002, 6, 597.
† Chemical Development.
‡ Analytical Chemistry.
(5) Plantan, I.; Selic, L.; Mesar, T.; Stefanic, A. P.; Oblak, M.; Prezelj,
A.; Hesse, L.; Andrejasic, M.; Vilar, M.; Turk, D.; Kocijan, A.; Prevec,
T.; Vilfan, G.; Kocjan, D.; Copar, A.; Urleb, U.; Solmajer, T. J. Med.
Chem. 2007, 50 (17), 4113.
(1) Klevens, R. M.; Morrison, M. A.; Nadle, J.; Petit, S.; Gershrnan, K.;
Ray, S.; Harrison, L. H.; Lynfield, R.; Dumyati, G.; Townes, J. M.;
Craig, A. S.; Zell, E. R.; Fosheim, G. E.; Mcdougal, L. K.; Carey,
R. B.; Fridkin, S. K. JAMA 2007, 298 (15), 1763.
(2) Klein, E.; Smith, D. L.; Laxminarayan, R. Emerging Infect. Dis. 2007,
13 (12), 1840.
(6) The software used for the statistical analysis was Design Expert DX
5. A central composite design (10 experiments with two central points)
was performed considering FmOCOCOCl and DIPEA equivalents.
The parallel equipment used was SK233 - Anachem.
(7) Battistini, C.; Scarafile, C.; Foglio, M.; Franceschi, G. Tetrahedron
Lett. 1984, 25 (22), 2395, and references therein.
(3) Rossi, T.; Andreotti, D.; Tedesco, G.; Tarsi, L.; Ratti, E.; Feriani, A.;
Pizzi, D. A.; Gaviraghi, G.; Biondi, S.; Finizia, G. PCT Int. Appl.
1998, 9821210.
840
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development
10.1021/op100097c  2010 American Chemical Society
Published on Web 06/08/2010

Scheme 1. Initial overall process
Moreover, no impurities were identified using a UV detector
coupled to an isocratic HPLC system.
4.1. MS Data. The reactions were performed both in toluene
and in chloroform (the literature-suggested solvents) and
monitored by MS spectrometry using ad hoc equipment and
the procedure as described in the Experimental Section.
All the MS signals are associated with the species [M
+ H]+ due to the positive electrospray ionization mode
used. Regarding the reaction performed in toluene: (a) the
signal at m/z 665 is associated with the protonated starting
oxalimide 6; (b) the signal at m/z 785 is associated with
the desired phosphorane 7, formed quickly in the first
minutes; (c) the MS signal at m/z 893 was tentatively
attributed to an impurity resulting from the addition of a
second equivalent of phosphine to the phosphorane
intermediate and subsequent loss of ethylene; (d) three
other signals were also observed at m/z values of 801,
773, and 727, respectively. The signal at m/z 727 was
attributed to the loss of ethanol from the species m/z 773
and the latter to the loss of ethylene from the species m/z
801 (Figure 2a).
A different approach to this step was needed; in particular,
the understanding of the reaction mechanism was required via
the identification of the side products by means of different
analytical techniques. The probe of a MS spectrometer was
connected directly to a reactor to check the “in process”
molecular weight of the products of the reactions. Moreover,
some reactions were performed in a NMR tube checking the
31P signals during the reaction in order to understand the nature
and the relative amounts of the obtained intermediates contain-
ing phosphorous and their conversion to the final compound
(the 1H and 13C signals were too complicated to be of use due
to the presence of conformers, side products, and diastereoi-
somers in mixture).
The MS spectra obtained from the reaction performed in
chloroform (Figure 2b) contained the same ion signals observed
using toluene even if at different levels plus a number of new
minor peaks. These observations suggested that the reactions
are solvent dependent and chloroform does not seem to be the
best solvent for the first part of the reaction.
At the end of the intermediate formation phase (about
60 min) the temperature of the reaction vessel was raised
to 70 °C, and the cyclization to the desired compound 8
(m/z 633) started. Figure 3 gives the MS spectra obtained
at the end of the reaction, and the following observations
can be highlighted: (a) m/z 633 is a major peak in both
spectra as well as m/z 785 and m/z 893; (b) the spectra
Figure 1. Oxalimide optimization via DoE.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
841

Scheme 2. Step 2b literature-suggested mechanism
obtained in toluene contained also a major signal at m/z
847, tentatively attributed to the loss of ethanol from m/z
893; (c) other minor peaks were present at different levels
in both spectra, among them, it is worth underlining the
signal m/z 757 (loss of ethylene from the phosphorane 7)
and the signal m/z 769 associated with an adduct formed
into an MS ion source [8 + (EtO)2PMe + H]+.
A kinetic reaction profile was finally obtained by reporting
the relative intensities of the main MS signals as a function of
the reaction time. The initial fast decrease of the signal
associated with the starting oxalimide 6 was accompanied with
a fast increase of the signal associated with the phosphorane 7.
Then, when the reaction mixture was heated, the cyclic 8 started
forming with the simultaneous disappearance of the phospho-
Figure 2. MS spectra obtained at the end of the phosphorane formation performed in toluene (a) and in chloroform (b).
842
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

Figure 3. MS analysis of cyclization performed in toluene (a) and in chloroform (b).
Figure 4. Kinetic profile in toluene.
rane 7, the initial formation of the impurity 893 and its decrease,
while 847 increased (Figure 4).
Figure 5 shows the phosphorane 7 synthesis both in toluene
and in chloroform. From left to right we could identify the
unreacted phosphine (175 ppm), the phosphorane intermediate
7 (3-4 peaks due to different species in equilibrium in the
region 74-75 ppm), a hydride (phosphine side product always
present in the starting material, 28 ppm), the phosphinoxide
(side product of the reaction, 27 ppm), and an impurity present
in larger amounts in chloroform than in toluene. Its chemical
4.2. 31P NMR Investigation. NMR analyses were required
in order to obtain structural information regarding these side
products as well as quantitative data. The reactions were
performed at different temperatures directly in the NMR tubes,
using deuterated toluene and chloroform as reaction solvents,
respectively.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
843

Figure 5. NMR analysis of oxalimide formation performed in toluene and chloroform.
Figure 6. NMR analysis of cyclization performed in toluene and chloroform.
shift (-29 ppm) seems to suggest a penta-coordinated phos-
phorous tetraalkoxy-alkyl.
compound bearing another phosphorous in the same
molecule, the second one being a pentavalent two-alkoxy-
two-alkyl-hydroxyl. This species resulted in a larger
amount in toluene with respect to that in chloroform.
Moreover, the excess of integral at 76 ppm, phosphorane-
like phosphorous, was balanced by the integration of the
resonance at 51 ppm, suggesting the presence of a further
molecule with two phosphorous atoms, the second one
being coordinated to two-alkyl-alkoxy-oxygen double
During the cyclization (Figure 6), apart from degrada-
tion peaks of the phosphine, the phosphorane 7 peak is
decreasing due to its cyclization, but a signal remains at
75 ppm. This resonance could be still in agreement with
phosphorus linked to a dialkoxy-alkyl-carbon double bond.
Interestingly, this signal showed the same integration with
the signal at -12/-13 ppm, suggesting the presence of a
844
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

Scheme 3. Suggested structures and amount of the impurities
bond. Finally, the initially observed peak at -29 ppm was
found again in relatively larger amounts in chloroform
than in toluene.
exocyclic double bond9 of the phosphorane intermediate. It can
further lose ethanol under the MS conditions, giving the 15 (m/z
847).
Finally, the peak m/z 757 could represent 16, formed as a
consequence of the loss of ethylene from the phosphorane
intermediate 7, only seen in toluene, due to the instability of
the phosphorane intermediate at high temperature.
4.3. Involved Intermediates and Proposed Mechanism.
Having identified the mass of the major impurities and obtained
some structural information by 31P NMR chemical shifts as well
as their relative amounts, it is possible to suggest the structures
of all the species involved in the reaction (Scheme 3).
The peak m/z 801 can be associated with 11, which would
correspond to the phosphorus detected at -29 ppm in the NMR
spectra, produced in a lower amount in toluene than in
chloroform. 11 can lose ethylene, giving 12 (the m/z 773) and
ethanol giving 138 (the m/z 727). 13 was isolated by column
A summary of MS and 31P NMR data related to each of the
species here cited is reported in Table 1. It is worth noting that
some extra peaks were detected in 31P NMR spectra. An
investigation on the degradation of the starting phosphinoxide
was conducted in both chloroform (at 55 °C) and toluene (at
70 °C), leading to the identification of some major derivatives
1
responsible for these extra peaks by means of standard H
1
chromatography and fully characterized by H/13C/31P NMR
1
1
decoupled-31P and H coupled-31P NMR, and 2D H-31P
spectroscopy, confirming its structure and, indirectly, its parents.
The identification of these structures supports the hypothesis
of an intramolecular cyclization mechanism of the carbene
intermediate 10 with the phosphinoxide.
The peak m/z 893 is in agreement with 14, characterized by
the two phosphorus set of signals at 76 and 51 ppm in 31P NMR
spectra collected in chloroform. This species exists, according
to NMR results, both in the open keto- and in the cyclic eno-
tautomer 17, in larger amount in toluene than in chloroform,
and comes from a subsequent addition of phosphine to the
correlation experiments.
On the basis of this data, the following reaction mechanism
is postulated (Scheme 4): from oxalimide 6, after the addition
of phosphine, carbene intermediate 10 and one equivalent of
phosphinoxide is formed. Carbene 10 can react with the second
equivalent of phosphine, giving the desired phosphorane 7, or
react with the phosphinoxide, leading to impurity 11. At higher
temperatures phosphorane 7 cyclizes to the final trinem ester 8
but can also add another equivalent of phosphine, giving
impurity 14 which can lose ethylene to give 15. Phosphorane
7 itself can lose ethylene, giving impurity 16. Moreover,
(9) Castelijns, M. M. C. F.; Schipper, P.; Van Aken, D.; Buck, H. M. J.
Org. Chem. 1981, 46 (1), 47.
(8) Labaudiniere, L.; Burgada, R. Tetrahedron 1986, 42 (13), 3521.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
845

Table 1. Summary of MS and 31P NMR data for the step 3 impurities
product
MS (m/z)
31P NMR δ (ppm, CDCl3)
31P NMR δ (ppm, toluene-d8)
(EtO)2PMe excess
HP()O)(Me)(OEt)e
(EtO)2P()O)Mee
(EtO)2P(OMe)e
n.a.
175.9a, 176.2b
175.3a, 175.3b
27.9a, 26.5b
27.2a, 26.1b
n.d.a, 136.9b
n.d.a, 135.7b
n.d.a,b
n.a.
31.3a, 30.7b
153
28.6a, 28.3b
n.a.
n.d.a, 137.9b
(HO)2P(OMe)e
n.a.
n.d.a, 136.1b
e
HP()O)(OEt)2
n.a.
n.d.a, 4.8b
phosphorane 7
785
75 (various peaks)a, 75.4 (br.)b
76 (various peaks)a, 75.4b
-29.1a, -29.5b
n.d.d
11
12
13
14
15
16
17
801
-28.5a, -28.3b
773
n.d.d
727
n.d.a,b,d, 43.6c
n.d.a, 44.3b
893.40440f
847.36527f
757
n.d.a, 75.4, 74.4/52-48 (various peaks)b
n.d.a,b
n.d.d
n.d.d
n.d.d
n.d.d
893.40440f
n.d.a, 75.4, 74.4/-10, -16 (br.)b
n.d.a, 75.4, 74.7/-11.1, -13.7b
a Phosphorane 7 formation (spectra collected at 25 °C). b Cyclization reaction (spectra collected at 55 °C in CDCl3, 70 °C in toluene-d8). c Isolated material (spectra
collected in CDCl3 at 25 °C). d Not detected or unclear in the expected chemical shift regions. e Degradation products of the starting phosphinoxide identified in a blank
reaction in both CDCl3 (at 55 °C) and toluene-d8 (at 70 °C). f Accurate mass measurements carried out using a VG Autospec mass spectrometer, FAB positive ionization
mode, and PEG 1000 internal standard. The differences between measured and calculated accurate mass values are e5 ppm.
Scheme 4. Reaction mechanism
impurity 11 can lose ethylene and ethanol, giving 12 and 13,
respectively. Thus, it appeared that we had a clash between the
reaction conditions necessary to prevent the formation of
the impurities. In particular, the impurities of the first part of
the reaction were due to a competitive reaction of carbene
intermediate 10 with the phosphinoxide and the phosphine, and
thus, the impurities seemed to be preventable by using an excess
of phosphine. However, the impurities of the second part were
due to the excess of phosphine and were also solvent dependent:
toluene seems to work better in the first part of the reaction
and worse in the second, while chloroform works worse in the
first part of the reaction and better in the second.
reactions to 11, 12, and 13 and then concentrate the reaction
mixture to eliminate toluene and the unreacted phosphine;
perform the second part of the reaction in chloroform to reduce
the side reactions to 14, 15, and 16, emphasizing the desired
pathway.
At this point different DoE studies10 were performed to
optimize the new conditions; in the synthesis of the phosphorane
7 we focused our effort on the optimization of the equivalents
of phosphine to be used, volumes of solvent, and temperature.
As expected, increasing the equivalents of phosphine resulted
in a higher yield in the phosphorane 7 (Figure 7).
5. Optimization of Step 2b
(10) The software used for the statistical analysis was Design Expert DX
5. A central composite design (20 experiments with 6 central points)
was performed considering phosphine equivalents, concentration, and
temperature. As output, the ratio between the HPLC areas of the
phosphorane and of naphthalene was considered as the internal
standard. The parallel equipment used was SK233 - Anachem.
Having identified the structures of the principal impurities
and the cause of their formation, it was possible to devise a
new powerful procedure: perform the first part of the reaction
in toluene using a large excess of phosphine to reduce the side
846
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development

probe of a Quattro II (Micromass - Waters) mass spectrometer
through a fused silica capillary. A positive nitrogen pressure
(10 psi) applied to the vessel generated a 2.5 µL/min continuous
flow of reaction mixture which entered the ESI probe. An
additional makeup solvent (95:5 methanol water at 75 µL/min)
was added to the electrospray ionisation (ESI) probe allowing
an increase in the ionisation efficiency in positive mode.
7.2. NMR Spectroscopy. 31P NMR spectra were collected
using a Varian VXR 300 NMR spectrometer operating at 121.4
MHz. The small-scale reactions were followed directly in 5 mm
NMR tubes at controlled operating temperatures in accordance
with the effective actual reaction conditions. Both deuterated
chloroform (CDCl3) and toluene (toluene-d8) were used as
solvents.
7.3. SyntheticDetails.9H-Fluoren-9-ylmethyl(1S,5E,8aS,8bR)-
1-(1-{[(1,1-dimethylethyl)(dimethyl)silyl] oxy}ethyl)-2-oxo-5-(4-
pyridinylmethylidene)-1,2,5,6,7,8,8a,8b-octahydroazeto[2,1-
a]isoindole-4-carboxylate (8). ꢀ-Enoneazetidinone oxamate 4
(10 kg, 1 equiv) was suspended in dichloromethane (80 L) at
20 °C, and then aqueous NaHCO3 4% (80 L) was added. The
mixture was stirred, and the aqueous layer was separated and
then counterextracted with dichloromethane (25 L). The two
organic phases were mixed together and then were concentrated
to 50 L. The solution was cooled to 0 °C, DIPEA (6.78 L, 2
equiv) was added, and then a solution of FmOCOCOCl 7 (8.52
kg, 1.5 equiv) in dichloromethane (5 L) was added in 30 min.
Toluene (110 L) was added 15 min after the end of the addition;
then the solution was washed with aqueous NaCl 10% (80 L),
and the organic solution of 6 was then dried via azeotropic
distillation. To this solution was added (EtO)2PMe (50% in iso-
octane, 18 L, 3 equiv), and this mixture stirred at 30 °C for 50
min and was then concentrated under vacuum to about 30 L.
Cyclohexane (300 L) was added and the mixture heated at 90
°C for 70 min and then cooled to 20 °C. The solution was
filtered through an alumina pad (20 kg), the pad was washed
Figure 7. Phosphorane 7 optimization via DoE.
Regarding the cyclization, solvent screening was performed
to identify a better solvent and substitute for the undesired
chloroform.
At the end of the study (Figure 8) the safer cyclohexane/
toluene mixture was selected. Applying the above-mentioned
conditions, stage 2 global yield reached the target objective
(∼60%) with a net decrease in the observed impurities.
6. Summary
In this report we described how a combination of NMR and
MS analyses made possible not only identifying the structure
of the main impurities but also quantifying their amounts. On
the basis of these data it was possible to understand the reaction
mechanism and to design a better process that matched our
target yield.
The final process was successfully scaled up on 10 kg scale
and allowed progression to the subsequent clinical studies.
7. Experimental Section
7.1. MS Spectrometry. The reactions were conducted in a
reaction vessel directly connected to the electrospray triaxial
Figure 8. Screening of solvents for the cyclization to 8.
Vol. 14, No. 4, 2010 / Organic Process Research & Development
•
847

with toluene (2 × 30 L), and the organics were concentrated
to 20 L. Di-isopropyl ether (82 L) was added in 30 min and
the suspension cooled to 0 °C for 1 h. The suspension was
filtered, and the solid obtained was washed with cold di-
isopropyl ether (16.5 L) and dried at 40 °C under vacuum,
yielding 8 (7.5 kg, 60%).
1H NMR (500 MHz, CDCl3): δ 0.12 (s, 3H); 0.13 (s, 3H);
0.93 (s, 9H); 1.27 (d, J ) 6.3 Hz, 3H); 1.60-1.50 (m, 2H);
2.10-1.95 (m, 2H); 2.22 (m, 1H); 3.02 (m, 1H); 3.12 (m, 1H);
3.37 (dd, J ) 5.9, 3.0 Hz, 1H); 4.26 (m, 1H); 4.30 (m, 1H);
4.40 (m, 2H); 4.50 (m, 1H); 6.44 (m, 1H); 7.07 (d, J ) 6.6 Hz,
2H); 7.28 (m, 2H); 7.39 (m, 2H); 7.63 (d, J ) 7.0 Hz, 1H);
7.72 (d, J ) 7.0 Hz, 1H); 7.75 (m, 2H); 7.82 (d, J ) 6.6 Hz,
2H).
Acknowledgment
We thank Damiano Papini for the HPLC analytical support,
Carla Marchioro for the valuable NMR discussion, Tino Rossi,
Gilberto Vigelli, and Pieter Westerduin for their useful
suggestions.
Received for review April 12, 2010.
OP100097C
848
•
Vol. 14, No. 4, 2010 / Organic Process Research & Development