Process development for a novel pleuromutilin-derived antibiotic

Article abstract of DOI:10.1021/op900104g

A scalable synthesis of a novel pleuromutilin-based antibiotic is reported. The synthesis features the scale-up of an interesting skeletal rearrangement of the pleuromutilin core and isolation of a highly purified product despite starting with relatively

Full text of DOI:10.1021/op900104g

Organic Process Research & Development 2009, 13, 729–738
Process Development for A Novel Pleuromutilin-Derived Antibiotic
Yemane Andemichael, Jun Chen, Jacalyn S. Clawson, Wenning Dai, Ann Diederich, Susan V. Downing, Alan J. Freyer,
Peng Liu, Lynette M. Oh,* Daniel B. Patience, Sonja Sharpe, Joseph Sisko,* Julie Tsui, Frederick G. Vogt, Jun Wang,
Lori Wernersbach, Edward C. Webb, James Wertman, and Leon Zhou
Chemical DeVelopment, GlaxoSmithKline, 709 Swedeland Road, UW2820, PO Box 1539,
King of Prussia, PennsylVania 19406-0939, U.S.A
Abstract:
A scalable synthesis of a novel pleuromutilin-based antibiotic is
reported. The synthesis features the scale-up of an interesting
skeletal rearrangement of the pleuromutilin core and isolation of
a highly purified product despite starting with relatively impure
pleuromutilin. The use of Design of Experiment (DoE) and
Principal Component Analysis (PCA) tools to achieve these goals
Figure 1. Compound 1.
is also discussed. Furthermore, the novel coupling of a carbamate
and N-acyl-imidazole to produce the imidodicarbonate portion of
suitable methods to construct the relatively scarce unsym-
the target molecule is described.
Our search for alternative synthetic routes focused on finding
metrical imidodicarbonate functionality found in 1 without using
highly toxic or overly sensitive reagents.3 Although several
possibilities were quickly identified, we focused on that which
is described in Scheme 2 because of its brevity, lower cost,
Introduction
Antibiotic resistance is a growing concern in the
and anticipated desirable manufacturing characteristics. This
medical community. The commitment by GSK towards
paper describes our efforts to bring these concepts to fruition.
alleviating this problem is exemplified by the research
At the outset of our work, several strategic decisions needed
to be addressed in order to define the actual isolated intermedi-
ates. It was known from the initial scale-up of the chemistry in
Scheme 1 that intermediates 3 and 4 were difficult to crystallize
and isolate. The natural reflex in this circumstance was to avoid
isolation altogether, that is, to convert 2 to 9 without isolation
of 3 or 4 (Scheme 2). After only minimal effort it became clear
that from a chemistry perspective this approach was quite
feasible. However, there were two overriding considerations that
forced us to back away from this tactic. First was the recognition
that pleuromutilin (2) is a fermentation product, and batches of
2 that arrived in our laboratories for process development
contained no less that 35 different impurities, the structures of
which were mostly unknown. While there was an expectation
that the fermentation, crystallization, and isolation processes
would ultimately be optimized so as to reduce the variability
and amounts of the impurities found in 2, we reasoned that the
crystallization and isolation of additional intermediates from
Scheme 2 would help to secure the necessary and consistent
purity of 1A. Second, feedback from our manufacturing partners
within GSK indicated that the benefits derived from avoiding
the isolation of sequential intermediates would be offset by the
increased reactor residence time and an overall decrease in
‘process velocity’.
efforts that led to the discovery of compound 1, a novel
pleuromutilin-derived antibiotic. The antibacterial activity
of pleuromutilin and its derivatives has been known for
more than 50 years,1 but clinical development of such
agents has lagged. In 2007, GSK received marketing
authorization for Altabax, itself an agent derived from
pleuromutilin, for the topical treatment of impetigo. More
recently, compound 1 also showed promising antibacterial
properties worthy of further investigation, and this paper
describes the chemical process development that enabled
its production on larger scale.
The synthetic plan shown below in Scheme 1 was
initially employed to make compound 1 (as the succinic
acid salt, 1A) and analogues, and for that purpose it was
well-suited.2 However, for the production of large quanti-
ties of 1, several shortcomings of the route needed to be
overcome. Notable among these issues was the desire to
avoid phosgene (and its derivatives) to alleviate worker
safety concerns, the need to avoid AgOCN due to cost
and processing issues, and decreasing the number of
isolated intermediates in order to reduce overall costs.
Furthermore, the physical properties of a number of the
isolated intermediates in this route were suboptimal and
rendered large-scale production particularly cumbersome.
* Authors for correspondence. E-mail: lynette.m.oh@gsk.com. Telephone:
610-270-5492. Fax: 610-270-4829. E-mail: joe.sisko@gsk.com. Telephone: 610-
270-4360. Fax: 610-270-4829.
(3) For literature methods of preparing imidodicarbonates, see: Itoh, T.;
Watanabe, M.; Fukuyama, T. Synlett 2002, 8, 1323. Hajra, S.;
Bhowmick, M.; Maji, B.; Sinha, D. J. Org. Chem. 2007, 72, 4872.
Singh, O. V.; Han, H. Org. Lett. 2007, 9, 4801. Hageman, H. Angew.
Chem., Int. Ed. Engl. 1977, 16, 743. Gorbatenko, V. I. Tetrahedron
1993, 49, 3227.
(1) Kavanaugh, F.; Hervey, A.; Robbins, W. J. Proc. Natl. Acad. Sci.
U.S.A. 1951, 37, 570.
(2) Brook, G.; Burgess, W.; Colthurst, D.; Hinks, J. D.; Hunt, E.; Pearson,
M. J.; Shea, B.; Takle, A. K.; Wilson, J. M.; Woodnutt, G. Bioorg.
Med. Chem. 2001, 9, 1221.
10.1021/op900104g CCC: $40.75  2009 American Chemical Society
Published on Web 07/02/2009
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
729

Scheme 1. Original synthesis of compound 1Aa
a Reagents and conditions: (a) (CH3O)3CH, MeOH, H2SO4, 30 °C; (b) KOH, MeOH, H2O, 65 °C; (c) THF, pyridine, triphosgene, H2O, CH2Cl2; (d) CH2Cl2, cat.
pyridine, AgOCN; (e) EtOAc, HCl, NH4OH to pH ) 5; (f) CH2Cl2, aq K2CO3, 1-PrOH, succinic acid.
Scheme 2. Improved route to compound 1Aa
a Reagents and conditions: (a) (CH3O)3CH, MeOH, H2SO4, 72-77%; (b) MeOH, aq KOH, heptane, tert-butyl methyl ether, chlorosulfonyl isocyanate, Et3N, H2O,
toluene, 85%; (c) N-methyl pyrrolidinone, NaO-tert-pentoxide, aq citric acid; (d) toluene, HCl, aq NH4OH, 2-PrOH, EtOAc, 83% for 2 steps; (e) 2-PrOH, H2O,
succinic acid, 92%.
After re-examining our plan it was felt that 4 would
be the best intermediate to isolate, since any method for
converting 3 to 4 would require aqueous conditions, but
the subsequent conversion of 4 to 9 would almost certainly
require anhydrous conditions. Hence, isolation and drying
of 4 prior to conversion to 9 seemed logical. In practice,
the conversion of 2 to 4 without isolation of 3 was actually
quite simple to accomplish with good overall yields.
However, despite significant efforts, we simply could not
develop a reliable process for the crystallization of 4. As
such we refocused our efforts on the isolation of 3 fol-
lowed by the conversion of 3 to 9 without isolation of 4.
730
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

Figure 2. PCA score plot of IR spectra of batches of pleuromutilin 2.
Scheme 3. Convergence of 2 and impurity 12 to
As noted above, production of pleuromutilin (2) is ac-
complished by a complex fermentation process, followed by
extraction with an organic solvent and subsequent crystallization.
However, a practical limitation of this process is that batches
of 2 contained varying amounts of approximately 35 individual
impurities, the structures of which were mostly unknown and
difficult to accurately quantify due to their low concentrations
and poor UV absorbances. The complex impurity profile of 2
made it difficult to set a conventional specification for quality
control. Principal component analysis (PCA) is a powerful
technique4 of identifying patterns in data and expressing the
data in such a way as to highlight their similarities and their
differences, i.e., impurity profiles. We decided to conduct a
study of the variability in the purity of 2 using multivariate
analysis of the diamond attenuated total reflectance (DATR)/
IR spectra from various batches. DATR/IR is a rapid, less time-
consuming technique than our HPLC method, enabling dozens
of analyses to be collected in a short amount of time. The IR
spectra were collected on three replicates from each batch, and
principal component analysis was performed on these spectra.5
The PCA score plot of lab and plant batches of 2 is shown in
Figure 2.
Each data point in the score plot represents one sample
spectrum. These data points are grouped in the plot based on
their relative similarities and differences. The proximity of the
data points situated in the plot indicates similarity of batches
of 2 in terms of overall purity and impurity content. A number
of distinct clusters can be discerned. The horizontal axis (the
first principal component) in the plot corresponds primarily to
the content of 2, with batches to the right (quadrants 1 and 2)
having higher level of purity. The second principal component,
as represented by the vertical axis, is believed to reflect the
intermediate 4
variation of some of the impurities residing in the batches. Lab
and plant batches from the typical fermentation process (higher
purity) are in two separate clusters; both of which are found in
quadrants 1 and 2. Lower-purity batches, such as those produced
from “atypical” fermentation processes (where the feedstock
was altered), or batches containing high levels of impurity 12
(Scheme 3) are clustered to the left in quadrants 3 and 4. Thus,
it was encouraging to us that the DATR/IR data used in
combination with PCA techniques were correlating to HPLC
purity profiles. Despite the high variability and complexity of
the purity of 2, the technique grouped batches of 2 based on
their overall purity and impurity profile, making it easier for
visualization. More importantly, it allowed us to begin identify-
ing which batches were acceptable for use, thus helping us
toward our goal of setting specifications. One of those studies
is described here. As mentioned previously, impurity 12
(Scheme 3) is a prominent impurity which varied in amounts
ranging from 1 to 25%. In order to understand how the variable
amounts of 12 might impact on the stage 1 process, an online
DATR/IR method was developed to monitor the kinetics of
this chemistry.6 IR data from two reactions which utilized
batches of 2 situated in different quadrants (second vs fourth)
in the PCA score plot (Figure 2) were evaluated to gain
(4) For a leading reference on PCA see: Eriksson, L.; Johansson, E.;
Kettaneh-Wold, N.; Wold, S. Multi- and MegaVariate Data Analysis;
Umetrics Academy: Umea, Sweden, 2001.
(5) Umetrics SIMCA-P, version 11multivariate data analysis software was
used in this study. The DATR-IR data were collected using a Smiths
Detection (Danbury, CT) FT-IR spectrometer equipped with a diamond
attenuated total reflection (DATR) sampling interface. The IR spectral
data in the range from 800 to 1800 cm-1 were normalized by regressing
it against the average spectrum, and the resultant (filtered) data were
subjected to the PCA analysis.
(6) The in situ infrared data were collected using Mettler-Toledo Au-
toChem (Columbia, MD) MonARC Process IR with Neotecha
recirculation sampling loop. .
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
731

Table 1. Factor set points and ranges for stage 1
initial ranges
set-points (half-factorial study) (RSM study)
ranges
factor
MeOH
H2SO4
trimethyl orthoformate
(TMOF)
4.2 vol
0.3 vol
1.6 vol
2.0-6.0 vol
0.1-0.5 vol
0.6-2.0 vol
4.0-8.0 vol
0.05-1.0 vol
0.4-1.6 vol
temperature
30 ° C
20-40 ° C
0-40 ° C
Scheme 4. Formation of impurity 14 from 3
Figure 3. Stage 1 online IR profiles.
understanding of the extent of the impact. These two reactions
were conducted using batches of 2 that contained approximately
25% and 6% of acetate 12, respectively. The reaction profiles
are overlaid in Figure 3. The superimposable profiles for both
the consumption of starting materials and the formation of
product indicate these two reactions proceeded at virtually the
same reaction rate from the beginning to the end. Therefore,
the presence of 12 at levels as high as 25% in batches having
different impurity profiles does not have much impact on the
outcome of the reaction.
Concurrently, it was shown that 12 was converted to 13 in
stage 1, and we reasoned that the stage 2 hydrolysis would
render this problem moot as 3 and 13 should both be converted
to 4. In contrast, little was known about the structures of the
remaining impurities which greatly complicated the data analysis
during our studies to optimize the conversion of 2 to 3 (Vide
infra).
The strategic choice for the stage 1 conversion of 2 to 3
was made to ‘protect’ the alcohol at C-11 while allowing for
selective manipulation at the latent hydroxyl group on C-14.
Although this reversible process had previously been reported,7
the practical challenges brought on by the variable purity profile
of 2 made it difficult to optimize this reaction and ensure its
robustness. The mechanism for the formation of 3 is believed
to involve a number of discrete steps including epimerization
and corresponding ring flip at the cyclopentane junction of 2
followed by oxonium ion formation at C-3 under conditions
which promote ketalization. Finally, a stereospecific hydride
transfer from C-11 to C-3 produces 3. Ensuring that these steps
occur while controlling competitive product degradation under
the reaction conditions made this a challenging process to
optimize.
ranges. Additional runs at wider dilutions and temperature
settings established the ranges for the subsequent response
surface modeling (RSM) study.8 We elected to generate a
response surface model using a central composite design, the
main goals of which were to maximize conversion to 3 during
extended hold periods (up to 24 h) and to obtain a reaction
mixture with an impurity profile that would perform well in
the isolation procedure (Vide infra). Principally, we were most
concerned with maximizing the consumption of the starting
material 2 and minimizing the formation of the major impurity,
identified as alkene 14 (Scheme 4), since both had proved
difficult to purge.
To maximize production of 3, the most important factors
were temperature and an interaction between H2SO4 volumes
and temperature. Lower temperatures and lower acid amounts
should produce more of the product (3) (Figure 4). Note also
that, at lower acid amounts (e.g., 0.25 vol), the gentler slope
suggests less variation in the amount of 3 produced across the
temperature range and therefore would likely be a more robust
process in the event of temperature excursions on scale up.
Similarly, reactions conditions with (a) higher temperatures
and less acid or (b) lower temperatures and more acid would
minimize the amount of starting material 2 remaining (Figure
5).
The RSM also indicated that the interaction between H2SO4
volumes and temperature (Figure 6) was most influential on
Reaction Optimization
We began by conducting sequential design of experiment
(DoE) studies of the variables and factors which we believed
would influence the yield and purity profile of the reaction
(Table 1).
The initial half factorial designs indicated that there was
significant curvature in the linear responses that merited further
investigation. Moreover, the analysis of the reaction space also
suggested that the reaction optimum may be outside of the initial
Figure 4. Effect of interaction of H2SO4 and temperature on
product (3) levels.
(7) Berner, H.; Schulz, G.; Schneider, H. Tetrahedron 1980, 36, 1807.
732
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

Figure 7. Operating range for H2SO4 charge and temperature.
Figure 5. Effect of interaction of H2SO4 and temperature on
pleuromutilin (2) levels.
35 °C with no loss in purity.9 We were gratified to confirm
these predictions with three additional plant runs at 40 kg scale
that produced 72-77% isolated yields with nearly impurity
profiles identical to those above. The slightly lower-yielding
reaction was a direct result of using a batch of 2 that was 3-4%
less pure than those used in the other runs.
Crystallization Development
Since the solvent for stage 1 was methanol, crystallization
development began by investigating water as an antisolvent.
The original route (Scheme 1) utilized water as an antisolvent
to precipitate 3, but due to limited development work at that
time, the process had isolation issues on the initial scale-up to
15 kg. Therefore, redevelopment of the crystallization began
with solubility studies of purified 3 in varying ratios of methanol
and water at 25 °C. This investigation revealed a new crystalline
form of 3 that was confirmed by microscopy and thermal
analysis10 (Figure 8). Ripening experiments with this sample
as seed material yielded a pure sample of the new form. Along
with its higher melting point, this ripening suggested that the
new form was the more stable of the two forms. The habit of
the new crystal form (bladed) was also more desirable than that
of the original form (acicular) because the needle-like (or
acicular) crystals gave poorly mixing slurries and difficult
filtrations. Accordingly, ongoing development had the added
focus of maintaining form control to isolate the newly discov-
ered form.
The solubility was measured gravimetrically at three ratios
of methanol and water and four temperatures. This enabled a
seeded cooling crystallization in which water is added to the
methanol solution to saturate the solution at the desired seeding
temperature prior to adding seeds. Scaling the process to a 1-L
vessel revealed two issues-slight oiling prior to seeding and
sudden nucleation of the undesired form during the cooling
profile. Analysis of the oil phase showed that it was primarily
Figure 6. Effect of interaction of H2SO4 and temperature on
alkene 14 levels.
the level of alkene 14 generated. At lower temperatures, the
level of the alkene impurity 14 remained relatively low and
constant over the range of acid volumes, while at higher
temperatures the level of impurity 14 increased more signifi-
cantly (Figure 6).
Recalling that all these optimization studies were conducted
based on a 24 h reaction time and that data from our
crystallization work (Vide infra) indicated that approximately
4% of the alkene 14 and 3% of pleuromutilin 2 could be purged,
an operating range was defined to maximize yield and achieve
our purity targets (Figure 7).
The major conclusion from the central composite study is
that, in order to maximize product formation while minimizing
the amounts of 2 and 14 to levels that could be comfortably
purged by our crystallization protocol, less acid (0.25 volumes),
lower temperature (25 °C), and 7 volumes of MeOH with 1.3
volumes of TMOF should be employed. Importantly, our initial
plant runs at 40 kg scale confirmed these results. Two successive
runs at 25 °C required 24 h to reach completion and gave
isolated yields of 79% and 75%, both with >97% purity by
HPLC. With the goal of improving our process velocity,
additional studies also indicated that reaction times could be
significantly shortened to 6-7 h by conducting the reaction at
(8) Design-Expert, version 7.0.3; Stat-Ease, Inc.: Minneapolis, Minneso-
tawas used to analyze the data. The study type was a response surface
modeling based on central composite design (30 experiments). Overlay
plot (operating range) was generated using graphical optimization
technique.
(9) Presented by James Wertman at the Dynochem User Group Meeting
2007, May 15-16, 2007, Philadelphia PA.
(10) This data was collected on a TA Instruments 2920 DSC, with a 10
°C/min temperature ramp rate.
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
733

Figure 8. Microscopy and thermal analysis of the new form of 3.
Scheme 5. Base-mediated formation of impurity 15
acetate impurity 13. We quickly found that for reactions with
>5% of 13, the addition of 5-10% of methyl isobutyl ketone
limited the impact of 13 on the crystallization by preventing
this oiling and minimized the amount of 13 found in the isolated
product.
With regards to nucleation of the undesired form, several
methods were tested to improve the form control and suppress
nucleation, including increased seed loading, improving the
chemical and form purity of the seeds, and changing from a
dry addition to a slurry addition of seeds. However, the most
significant improvement was through the modification of the
cooling profile. Although the seeding temperature was not
changed, a longer hold time after seeding was implemented to
ensure complete desupersaturation. It was also determined that
a cooling rate of <0.10 °C/min after this hold time was required
to prevent nucleation of the undesired form. In the event of an
unsuccessful form control, it was also determined that a ripening
of the solution could convert the product completely to the
desired form. This was a crucial contingency, considering that
the undesired form yielded a compressible product cake with a
cake resistance 2 orders of magnitude larger than the desired
form, which led to excessively long filtration times. Careful
attention to the temperature control and cooling rate on scale-
up prevented the need to employ the ripening protocol for any
of the batches, and only small amounts of the undesired form
could be observed in in-process samples. The key signal of
success of the crystallizations was the smooth isolations by
centrifugation (which can exacerbate poor filtrations through
excessive cake compression). No operation issues were ob-
served, and the filtration time was as predicted for the desired
form. Also, in line with expectations was an average adjusted
yield of 84% and total impurity levels <3%, despite impurity
levels ranging from 5 to 17% in the input material.
to provide anhydrous intermediate 4, problems with emulsions
in the plant forced us to develop a more efficient transition to
the carbamoylation step. In the revised procedure, alcohol 3
was heated in aqueous KOH in MeOH to induce hydrolysis
followed by an extraction of 4 into heptane and an azeotropic
distillation to remove any residual water. Slow addition of
chlorosulfonyl isocyanate (CSI) to the heptane solution of 4
generated intermediate 16 which was hydrolyzed to 9 by
treatment with water and triethylamine in high yields and
consistently high purity after isolation by crystallization (Scheme
6). Once again our plant results closely matched the lab results.
Four consecutive runs at 24 kg scale provided 9 in isolated
yields of 77-87% with purities all exceeding 99% by HPLC.
Completion of the synthesis required only the coupling
of 9 and 1011 to produce 7 followed by acid-mediated
conversion to 11 (Scheme 7). The choice of solvent for
these transformations was deemed critical due to the poor
solubility of 10 in most solvents and the desire to use a
solvent during the conversion of 7 to 11 that was inert to
strong acids like HCl. The coupling reaction was initially
run by combining 9 and sodium tert-pentoxide (2.5 equiv)
in 10 volumes of THF at 0-5 °C for 5-10 min, followed
by addition of portions of 10 as a solid. Although this
gave acceptable yields in our initial plant campaign, there
were several aspects of this chemistry that we wanted to
explore. First, we wondered if this transformation could
be accomplished with a milder base that did not require
strictly anhydrous conditions. Next, we wanted to find a
means of adding the base as a solution or suspension
because addition of sodium tert-pentoxide as a solid
clogged the hopper and lines in the plant. Finally, we
looked to find a solvent that was capable of making stable
Having secured a relatively robust and efficient manner of
obtaining highly pure product 3, the remainder of the synthesis
became considerably less complex. As noted above, obtaining
anhydrous 4 was deemed imperative since we expected conver-
sion to 5 would utilize water-sensitive reagents. Towards this
goal, we examined the hydrolysis of 3 to 4 under phase transfer
conditions in toluene. In fact, hydrolysis of 3 (and to some extent
13) could be achieved with Bu4NCl (3%) and aqueous KOH
in high yield, while formation of the unexpected impurity 15
was kept to <1% (Scheme 5).
(11) Compound 10 was readily prepared in high yield from commercially
available trans-4-aminocyclohexanol by sequential treatment with
di-tert-butyl dicarbonate (Boc2O) and N,N′-carbonyldiimidazole
(CDI).
While the extractive workup of 4 in toluene followed by
azeotropic removal of water permitted an isolation-free process
734
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

Scheme 6. Conversion of alcohol 4 to carbamate 9
Scheme 7. Formation of intermediate 11
solutions or suspensions of the base, providing high yields
of the product and also leading to a simple workup.
An initial base screen confirmed that bulky alkoxide
bases such as sodium tert-pentoxide provided the best
conversions, while weaker inorganic bases (K2CO3,
K3PO4) gave no conversion at all. Furthermore, NMP was
found to be a superior reaction solvent that also allowed
for a much simpler process. Dissolving sodium tert-
pentoxide in 2 volumes of NMP allowed for a controlled
addition of base to a homogeneous mixture of 9 and 10
that had been dissolved in 3 volumes of NMP and cooled
to 5 °C. After 30 min, the product (as its sodium salt)
could be precipitated by adding the reaction mixture to a
large excess of water. Because the sodium salt of 7 readily
deliquesced upon storage for a few days, we added citric
acid to the water quench to produce a buffered system at
pH 4 that smoothly precipitated 7 as a fine white powder
that could be easily stored. Two runs at 40 kg scale
produced excellent results with very high purity and
isolated yields of 7 estimated to be >90% that were used
directly in the next step without drying.
We had envisioned toluene as the ideal choice for the
conversion of 7 to 11 because of its chemical stability
towards conc HCl and because it was expected to provide
a clean phase separation from the aqueous phase after
reaction completion. Indeed, we found that simply mixing
7 in toluene (4 volumes) followed by a water wash to
remove any residual NMP from the previous step, then
adding 1.5 volumes of conc. HCl at 15 °C for ∼4 h
smoothly produced 11 with high conversions (Scheme 8).
The Boc group was typically cleaved in the first 30-60
min, at which point the primary amine intermediate 17
became fully soluble in the aqueous layer. Stirring for a
Scheme 8. Formation of 11 via intermediate 17
few additional hours effected the skeletal rearrangement
to produce 11 as the HCl salt. Separating the toluene phase
efficiently removed any nonpolar impurities that formed
during the reaction while retaining the product in the
aqueous acid layer. Isolation of the free base of 11
involved slow addition of the aqueous layer to a mixture
of ammonium hydroxide and EtOAc to which 2-PrOH had
been added. After removing the bottom aqueous layer, a
solvent exchange into 2-PrOH followed by a seeded,
cooling crystallization produced crystalline 11 in high
yields and purity. It is interesting to note that, while
2-PrOH was the crystallization solvent due to the modest
solubility of 11 in this solvent, its presence in the EtOAc
extraction was essential for efficient extraction from water.
Once again, the plant runs at 65 kg scale gave highly pure
product and yields of >80% of 11 over 2 steps.
The final step in the sequence involved formation of
the succinic acid salt (1A). Initial salt screening of
compound 1 revealed that crystalline succinic acid salts
could be made both with a 1:1 and 0.5:1 (succinic acid:
1) stoichiometry.12,13 From the outset it was presumed that
the 1:1 stoichiometry would be easier to control, and so
this became our target. In addition, the high aqueous
solubility of both salts (>10 mg/mL) suggested that
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
735

Experimental Section
(1R,2R,4S,6R,7R,9R)-4-Ethenyl-2,4,7,14-tetramethyl-9-
(methyloxy)-3-oxotricyclo[5.4.3.01,8]tetradec-6-yl Hydroxy-
acetate (3). Compound 2 (40 kg, 1 equiv) was dissolved
in methanol (221.6 kg, 7 vol) and trimethyl orthoformate
(50.4 kg, 1.3 vol) at 20 °C. Concentrated sulphuric acid
(18.4 kg, 0.25 vol) was added over 15 min at <30 °C.
The mixture was heated to 35 °C and stirred for 7 h.
Triethylamine (29.0 kg, 1 vol) was added, and the mixture
was heated to 55 °C. Water (175.2 L, 4.4 vol) and methyl
isobutyl ketone (3.2 kg, 0.1 vol) were added while
maintaining the temperature at 50-60 °C. The solution
was cooled to 45 °C, and seed crystals (0.4 kg) were
added. The solution was cooled to 40 °C and held there
for 1.5 h before cooling to 30 °C over 1.5 h and holding
for an additional 2.5 h. The solution was further cooled
to 15 °C and held for 12 h then cooled to 0 °C for 1 h.
The product was isolated by centrifuge, washed twice with
water (120 L, 3 vol each), and dried under vacuum at 25
°C for 15 h to give 31.3 kg (75% yield) of compound 3
as a white powder.
Figure 9. Salt ratios of 1A as a function of succinic acid input.
physical properties of the API, such as particle size, would
not likely become critical quality attributes and that
manufacturing attributes would dominate our selection
criteria.
Solubility data from a solvent screen in combination
with the results from a preliminary polymorph screen
identified 1-PrOH as the initial solvent of choice. How-
ever, high residual 1-PrOH levels (9-12%) were found
in our first scale-up batches of API after drying for 24-30
h. The choice of solvent was later re-evaluated, and a
2-PrOH/H2O system (9:1) was shown to give the product
that was the same form by XRPD and contained less
residual solvent (0.5-2%) after drying for 24 h.
A study was conducted to optimize the required
succinic acid charge to ensure the reproducible formation
of the 1:1 salt, 1A. The results showed that in 9:1 2-PrOH
/H2O, at least 1.6 equiv of succinic acid was needed to
prepare 1A (Figure 9). Because of its high water solubility,
the process for making 1A involved initial dissolution of
1 in a 4:1 ratio of 2-PrOH/water (2.5 volumes) with 1.6
equiv of succinic acid. This mixture was heated until
complete dissolution followed by a final clarifying filtra-
tion of the product. Subsequent additions of 2-PrOH as
the antisolvent and slow cooling led to a high-yielding
and controlled process for production of 1A. The scale-
up of this process to 2 kg proceeded uneventfully to give
the final API in 92% isolated yield with >99% purity.
1H NMR δ (ppm) (CDCl3, 700 MHz): 6.65 (1H, dd, J )
17.5 Hz, 10.5 Hz), 5.89 (1H, d, J ) 10.5 Hz), 5.32 (1H, d, J )
10.5 Hz), 5.04 (1H, d, J ) 17.5 Hz), 4.11 (2H, dd, J ) 27.3
Hz, 17.5 Hz), 3.47 (1H, m), 3.23 (3H, s), 2.92 (1H, q, J ) 6.3
Hz), 2.50 (1H, m), 2.01 (2H, m), 1.73 (1H, d, J ) 11.2 Hz),
1.58 (2H, m), 1.46 (1H, m), 1.32 (1H, m), 1.25 (1H, m), 1.24
(3H, s), 1.20 (3H, s), 1.16 (1H, m), 1.09 (1H, m), 1.01 (3H, d,
J ) 6.41 Hz), 0.80 (3H, d, J ) 7.0 Hz). 13C NMR δ (ppm)
(CDCl3, 176 MHz): 215.5, 172.9, 140.5, 118.8, 83.4, 73.9, 64.5,
61.7, 57.2, 54.3, 47.9, 45.5, 45.3, 44.7, 43.6, 40.7, 31.0, 29.8,
29.0, 25.9, 20.6, 16.8 16.1.
(1R,2R,4S,6R,7R,9R)-4-Ethenyl-2,4,7,14-tetramethyl-9-
(methyloxy)-3-oxotricyclo[5.4.3.01,8]tetradec-6-yl Carbamate
(9). Methanol (43.6 kg) and 3 (21.8 kg, 1.0 equiv) were
warmed to 40-45 °C until dissolution is observed. A
solution of KOH (6.3 kg, 2.0 equiv) in water (10.9 L)
was added to the methanol mixture, and the contents were
heated to 60-63 °C for 2 h. The mixture was cooled to
40 °C, and heptane (89.4 kg) and water (65.4 L) were
added. The mixture was stirred at 40-50 °C for 15 min,
and the aqueous layer was removed. The heptane layer
was washed with water (50 L) at 40 - 45 °C. Heptane
was removed by vacuum distillation to a final volume of
∼40 L. tert-Butyl methyl ether (113 kg) was added, and
the resulting solution was cooled to ∼5 °C. Chlorosulfonyl
isocyanate (10.2 kg, 1.3 equiv) was added over 20 min at
∼15 °C. After 75 min, an additional portion of chloro-
sulfonyl isocyanate (1.6 kg, 0.2 equiv) was added to drive
the reaction to completion. After stirring for an additional
30 min, water (54.5 L) was added slowly, at <30 °C.
Triethylamine (5.7 kg, 2 equiv) was added at <35 °C
during the addition, warmed to 45 °C, and stir-
red for 1 h and then at 20 °C for 12 h. The reaction was
warmed to 35 °C, and the aqueous layer was removed.
The organic layer was washed with water (43.6 L), and
the organic layer was removed. The solution was con-
centrated under vacuum to a final volume of ∼44 L, and
Conclusion
A scalable synthesis of the novel pleuromutilin-based
antibiotic 1A was developed. The synthesis features the scale-
up of an interesting skeletal rearrangement of the pleuromutilin
core and isolation of a highly purified product despite starting
with relatively impure pleuromutilin. Furthermore, the novel
coupling of a carbamate 9 and N-acylimidazole 10 to produce
the imidodicarbonate portion of 1 is described and optimized.
The route has been scaled up in plant to produce high-quality
API.
(12) A salt with a ratio of 2:1 (succinic acid:1) could also be formed via
the liquid-assisted drop grinding technique, but as this technique is
limited to lab-scale preparations, it was not considered a viable option
for scale-up. For details of the liquid-assisted drop grinding technique,
see: Friscic, T.; Trask, A. V.; Jones, W.; Motherwell, W. D. Angew.
Chem., Int. Ed. 2006, 45, 7546.
(13) Later studies showed that succinic acid salts of 1 could be formed via
crystallization with a continuum of stoichiometries ranging from 0.5:1
(SA:1) to 1.4:1. A detailed analytical study of the various salts was
conducted and reported. See: Clawson, J. S.; Vogt, F. G.; Brum, J.;
Sisko, J.; Patience, D. B.; Dai, W.; Sharpe, S.; Jones, A. D.; Pham,
T. N.; Johnson, M. N.; Copley, R. C. P. Cryst. Growth Des. 2008, 8,
4120. Details of our efforts to controllably manufacture salts with
various stoichiometries will be discussed in future communications.
736
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development

an additional portion of heptane (29.8 kg) was added. The
solution was again concentrated to ∼44 L and heated to
60-65 °C, and heptane (29.8 kg) and toluene (9.4 kg)
were added. The contents were stirred for 30 min at 60-65
°C and then cooled over 2 h to 0-5 °C. The reactor
contents were held at 0-5 °C for 1 h, and the solids were
isolated by centrifuge. The vessel and cake were rinsed
with heptane (44.9 kg). The solids were dried under
vacuum at ∼30 °C for 12 h to yield 17.9 kg (85% yield)
of compound 9 as a white solid.
(11). A reactor was charged with toluene (262 L, 4 vol,
based on wt/wt assay of compound 7 (256 kg crude, 65.5
kg at 100%, 1 equiv)). The mixture was heated to 60-65
°C and held for 30-40 min. Stirring was stopped to allow
phase separation, and the bottom aqueous portion was
discarded. To the top organic layer was added water (196.8
L), and the mixture was stirred at 70-75 °C for 20-30
min. Stirring was stopped to allow phase separation, and
the bottom aqueous portion was discarded. The contents
in the reactor were cooled to 10-15 °C, and conc HCl
(116.1 kg) was added over 30-40 min at <30 °C. Once
the HCl addition was complete, the reaction temperature
was raised to 20-25 °C, and the reaction was stirred 2 h.
The agitation was stopped, and the lower aqueous layer
containing the product was separated. To the top organic
portion was added conc HCl (19.7 kg, 0.27 vol), and the
mixture was stirred for ∼10 min and settled. The bottom,
aqueous layer was combined with the original aqueous
acid layer, and the organic layer was discarded. The
combined aqueous solutions were added over 1.5 h into a
second vessel containing a mixture of ammonium hydrox-
ide (30% aqueous, 137.8 kg), water (65.5 L), 2-PrOH (65.5
L), and ethyl acetate (295.9 kg) that has been cooled to
∼15 °C. The temperature in the vessel was maintained at
<30 °C during the addition, and the final pH was >9 after
complete addition. After stirring the mixture at 20-25
°C for 0.5 h, the bottom aqueous layer was separated and
discarded. To the organic layer was added water (196.8
L), and the mixture was stirred at 20-25 °C for 0.5 h.
The aqueous layer was discarded, and the organic portion
was concentrated to a final volume of ∼200 L. 2-PrOH
(257.8 kg) was added, and the mixture was concentrated
under vacuum to a final volume of ∼330 L. The mixture
was cooled to 0-5 °C over 2 h and held at this temperature
for 1 h. The mixture was filtered via centrifuge and the
wet cake rinsed with cold 2-PrOH (103 kg). The solid
product was dried under vacuum (∼60 °C) to a final LOD
of <20% to give 55.1 kg (83% yield) of product 11.
1H NMR δ (ppm) (DMSO-d6, 400 MHz): 6.22 (1H,
dd, J ) 17.7, 11.5 Hz), 5.46 (1H, d, J ) 8.1 Hz), 5.10
(1H, dd, J ) 17.5 Hz, 1.7 Hz), 5.05 (1H, dd, J ) 11.0
Hz, 1.7 Hz), 4.48 (1H, m), 4.26 (6H, br), 3.77 (2H, m),
3.42 (1H, d, J ) 5.5 Hz), 2.57 (1H, m), 2.35 (1H, m),
2.18 (1H, m), 2.08 (3H, m), 1.88 (2H, m), 1.76 (2H, m),
1.64 (2H, m), 1.46 (2H, m), 1.37 (3H, s), 1.28 (6H, m),
1.13 (2H, m), 1.06 (3H, s), 1.03 (12H, d, J ) 6.1 Hz),
0.81 (3H, d, J ) 6.8 Hz), 0.63 (3H, d, J ) 6.6 Hz). 13C
NMR δ (ppm) (DMSO-d6, 100 MHz): 217.1, 150.8, 150.5,
140.9, 115.2, 73.4, 72.7, 69.9, 62.0, 57.4, 49.0, 45.0, 44.0,
43.8, 41.6, 36.3, 34.0, 33.2, 30.1, 29.8, 28.2, 26.6, 25.5,
24.5, 16.1, 14.7, 11.5.
1H NMR δ (ppm) (DMSO-d6, 400 MHz): 6.72 (1H, dd, J
) 17.5, 10.6 Hz), 6.47 (2H, br), 5.50 (1H, d, J ) 10.0 Hz),
5.24 (1H, d, J ) 10.6 Hz), 4.93 (1H, d, J ) 17.5 Hz), 3.36
(1H, m), 3.12 (3H, s), 2.86 (1H, q, J ) 6.4 Hz), 2.31 (1H, dd,
J ) 15.3, 10.0 Hz), 2.04 (1H, m), 1.93 (1H, m), 1.82 (1H, m),
1.64 (1H, d, J ) 11.5 Hz), 1.53 (1H, d, J ) 15.5 Hz), 1.40
(2H, m), 1.24 (1H, m), 1.15 (3H, s), 1.12 (1H, m), 1.10 (3H,
s), 1.07 (1H, m), 0.90 (3H, d, J ) 6.4 Hz), 0.81 (3H, d, J )
7.0 Hz). 13C NMR δ (ppm) (DMSO-d6, 100 MHz): 214.5,
156.0, 140.9, 117.6, 82.4, 69.8, 63.2, 56.3, 53.3, 47.0, 44.8,
44.2, 42.8, 39.6, 30.1, 28.9, 28.3, 25.4, 20.3, 16.2, 15.4.
trans-4-({[(1,1-Dimethylethyl)oxy]carbonyl}amino)cyclo-
hexyl (1R,2R,4S,6R,7R,9R)-4-Ethenyl-2,4,7,14-tetramethyl-
9-(methyloxy)-3-oxotricyclo[5.4.3.01,8]tetradec-6-yl Imidodi-
carbonate (7). N-Methylpyrrolidinone (82.4 kg), 9 (40.0
kg, 1.0 equiv) and 10 (40.0 kg, 1.2 equiv) were slurried
together for 30 min and then cooled to 5 °C. A solution
of sodium tert-pentoxide (29.2 kg, 2.5 equiv) in NMP
(123.6 kg) was added to the slurry containing 9 and 10
over 30 min. The reactor containing the pentoxide base
was rinsed with NMP (41.2 kg), and the rinse was added
to the reactor vessel. After 30 min, a solution of citric
acid (0.76 kg) and water (7.2 L) was added to the reaction,
and the resulting mixture was warmed to ∼25 °C and
stirred for 15 min until dissolution was observed. The
mixture was added over 20 min to a second vessel
containing a solution of citric acid (69.4 kg) and water
(650 L). The resulting slurry was filtered via centrifuge
in six separate portions and washed with water (at least
133 L per portion), and the product (compound 7, 258.5
kg ‘wet’) was stored and used without further drying in
the next step. Yield was assumed to be 100% for the
purposes of calculating charges in stage 4.
1H NMR δ (ppm) (CDCl3, 400 MHz): δ 6.70 (1H, dd, J )
16.0, 8.0 Hz), 5.74 (1H, d, J ) 8.0 Hz), 5.28 (1H, d, J ) 12.0
Hz), 4.99 (1H, d, J ) 16.0 Hz), 4.70 (1H, m), 4.45 (1H, d, J )
6.6 Hz), 3.45 (2H, m), 3.23 (3H, s), 2.88 (1H, m), 2.48 (1H,
m), 2.21 (1H, m), 2.03 (6H, m), 1.70 (3H, m), 1.57 (6H, m),
1.46 (9H, s), 1.25 (3H, s), 1.20 (3H, s), 1.17 (1H, m), 1.01
(3H, d, J ) 4.0 Hz), 0.87 (3H, d, J ) 4.0 Hz). 13C NMR δ
(ppm) (CDCl3, 100 MHz): 215.1, 155.2, 150.2, 150.1, 140.2,
118.4, 83.0, 79.3, 74.4, 74.0, 64.0, 56.7, 53.7, 48.4, 47.5, 45.2,
44.8, 44.0, 43.2, 40.2, 30.6, 29.9, 29.4, 28.5, 28.3, 25.6, 20.3,
17.6, 16.5, 15.7.
trans-4-Aminocyclohexyl (1S,2R,3S,4S,6R,7R,8R)-4-Ethe-
nyl-3-hydroxy-2,4,7,14-tetramethyl-9-oxotricyclo[5.4.3.01,8]-
tetradec-6-yl Imidodicarbonate Succinic Acid Salt (1A). A
reactor was charged with 11 (2.2 kg of the bis-(2-propanol)
solvate, 1.8 kg of 1 based on wt/wt assay), 2-PrOH (4.4
L), and water (1.1 L). Succinic acid (658 g, 1.6 equiv)
was charged, and the contents of the reactor were stirred
trans-4-Aminocyclohexyl (1S,2R,3S,4S,6R,7R,8R)-4-Eth-
enyl-3-hydroxy-2,4,7,14-tetramethyl-9-oxotricyclo[5.4.3.01,8]-
tetradec-6-yl Imidodicarbonate Bis-(2-propanol) Solvate
Vol. 13, No. 4, 2009 / Organic Process Research & Development
•
737

and heated to 70-75 °C until a clear solution was
obtained. The solution was filtered through a 1 µm in-
line filter into a crystallization reactor. The original reactor
was rinsed with 2-PrOH (5.5 L) that was then sent through
the filter and into the crystallization reactor. The contents
of the crystallization reactor were heated to 70-75 °C to
ensure complete dissolution and then cooled to ∼65 °C
over 15 min and seeded with 18 g (1.0 wt %) of compound
1A seeds that were suspended in 2-PrOH (64 mL). The
slurry was held at ∼65 °C for 30-60 min and then cooled
to 20 °C over 1 h. Additional 2-PrOH (11.0 L) was then
added to the slurry over 30 min, and the mixture was
cooled to 0-5 °C over 1 h. The product was isolated by
filtration, rinsed with 2-PrOH (7.0 L), and dried under
vacuum at 40-60 °C for ∼24 h. The product weighed
2.03 kg (92% yield).
1.48 (1H, m), 1.38 (1H, m), 1.36 (5H, s), 1.24 (5H, m),
1.05 (3H, s), 1.00 (1H, m), 0.80 (3H, d, J ) 7.0 Hz),
0.61 (3H, d, J ) 7.0 Hz). 13C NMR δ (ppm) (DMSO-d6,
176 MHz): 217.1, 175.5, 150.7, 150.5, 140.9, 115.2, 72.7,
72.3, 70.0, 57.3, 47.9, 44.9, 43.9, 43.8, 41.6, 36.3, 34.0,
32.2, 30.1, 28.8, 28.3, 28.2, 26.6, 24.5, 16.1, 14.7,
11.5.
Acknowledgment
We thank Thomas Wrzosek and James Ward for their
assistance in analyzing and interpreting the results of the DoE
studies.
Supporting Information Available
Copies of the NMR spectra of all intermediates and final
product as well as the raw data for the DoE study. This material
is available free of charge via the Internet at http://pubs.
acs.org.
1H NMR δ (ppm) (DMSO-d6, 700 MHz): 9.20 (5H,
b), 6.20 (1H, dd, J ) 17.7, 11.2 Hz), 5.45 (1H, d, J ) 8.3
Hz), 5.09 (1H, dd, J ) 17.5, 1.8 Hz), 5.04 (1H, dd, J )
11.2, 1.8 Hz), 4.48 (1H, m), 3.41 (1H, d, J ) 5.8 Hz),
2.98 (1H, m), 2.36 (1H, s), 2.24 (4H, s), 2.17 (1H, m),
2.07 (3H, m), 1.95 (4H, m), 1.65 (1H, m), 1.61 (1H, m),
Received for review April 23, 2009.
OP900104G
738
•
Vol. 13, No. 4, 2009 / Organic Process Research & Development