Figure 2. Gradient analytical SFC screening affords excellent separation of crude mixture of trans-1 and bicyclic lactone, 3, on
Chromegabond Nitro stationary phase. Conditions: outlet pressure 200 bar, column temperature 35 °C; 1.5 mL/min methanol in
carbon dioxide, gradient of 4% methanol for 4 min, then ramp to 40% methanol @ 2%/min with 3 min hold.
Conventional purification approaches such as silica or
reversed phase chromatography proved fairly ineffective for the
separation of trans-1 and bicyclic lactone, 3; however, screening
of the mixture by supercritical fluid chromatography (SFC)
showed an unusually large separation with the Chomegabond
Nitro stationary phase (Figure 2). A small amount of residual
toluene in the crude reaction mixture appears as a large, early-
eluting peak owing to the comparatively poor UV chromophores
of trans-1 and lactone 3 at the observe wavelength of 215 nm.
Interestingly, lactone 3 is more strongly retained on the
stationary phase than trans-1, opposite what might be expected
based on a simple accounting of polar groups of the two
molecules. This result is somewhat surprising, and may stem
from the fact that the polar ester and alcohol groups in trans-1
are relatively inaccessible, owing to steric encumbrance, with
little possibility for simultaneous interaction of both groups with
the stationary phase. In contrast, the oxygen lone pair electrons
of bicyclic lactone, 3, are considerably more exposed and
available for interaction with the stationary phase, perhaps with
the electron-deficient dinitrobenzene ring of the stationary phase.
Whatever the reason underlying the greater retention of bicyclic
lactone, 3, on this stationary phase, the results clearly indicate
the possibility of facile chromatographic purification of the
desired trans-1.
HPLC. Suitable conditions were identified using an eluent of
5% IPA/heptane, with a step gradient to 60% IPA/heptane,
which improved peak shape and helped to desorb the strongly
retained bicyclic lactone, 3. This HPLC method was employed
in a cGMP preparation of 7.5 kg of trans-1 (Figure 4). In this
campaign, an 11 cm diameter HPLC column filled with 1.5 kg
of the Chromegabond Nitro stationary phase was used to purify
12 kg of crude mixture, affording 7.5 kg of pure trans-1 in
98% purity with 97% recovery. The separation took place over
a period of three work days (∼22 h instrument time) using a
total of 1200 L of solvent, 170 L of which were evaporated to
recover the desired product.
A chromatographic productivity of 5.5 kkd was obtained
during the campaign, with clear indications that further im-
provements could be possible. For example, nearly 20% of each
cycle is spent simply applying sample to the column using an
injector pump with 100 mL/min maximum flow rate - well
suited for general separations, but clearly undersized for this
separation. Simply applying sample at the higher flow rate of
750 mL/min would decrease cycle time by nearly 4 min,
resulting in a productivity of 6.6 kkd. In addition, the step
gradient could almost certainly begin at least 2 min earlier in
the chromatogram, resulting in a further cycle-time reduction
and a corresponding increase in productivity.
Small-scale semipreparative SFC purification of 50 g of the
crude mixture of trans-1 and bicyclic lactone, 3, was carried
out by isocratic elution with 10% EtOH/CO2 on a 2.1 cm i.d.
column, using automated injection and fraction collection, and
conveniently afforded trans-1, with high purity (Figure 3). An
extremely high productivity of 3.5 kkd (kilograms of purified
product per kilogram of stationary phase per day) was obtained
in this SFC purification, with every indication that loading could
be further increased. A chromatographic productivity of 3.5
piqued our interest in the possibility of larger-scale implementa-
tion of this separation, given the difficulty in otherwise accessing
pure trans-1.
Our preferred approach for carrying out separations of
developmental compounds at kilogram scale is to use SFC when
possible, owing to the advantages in speed and solvent savings
often observed relative to the corresponding HPLC approaches.
However, at the time that this study was conducted, we had
not yet built our capabilities for carrying out SFC separations
on larger scale, and such purifications were carried out by
Despite the extraordinary productivity of this method, the
inability to recycle the undesired isomer makes this process
somewhat less than ideal for implementation at the manufactur-
ing scale. Nevertheless, the method did allow rapid and
convenient access to multikilogram quantities of the desired
trans-1, effectively eliminating a significant development chal-
lenge for this project.
Experimental Section
Chemicals. Methanol, 2-propanol, and n-heptane were
obtained from EMD Chemicals (Gibbstown, NJ, U.S.A.).
Compounds 1-3 were obtained from Merck Process Research
(Merck Research Laboratories, Rahway, NJ, U.S.A.).
Chromatographic Stationary Phases. The chiral stationary
phases used for the column screening study were Chiralpak AD,
Chiralpak AS, Chiralcel OD, Chiralcel OJ, and Chiralcel OF
(Chiral Technologies, Inc., West Chester, PA, U.S.A.). The CSP
for the method optimization, loading studies, and final prepara-
tive separation was Chiralpak AD (particle size 20 µm).
622
•
Vol. 13, No. 3, 2009 / Organic Process Research & Development
Welch, Christopher J.
