Design of an integrated process of chromatography, crystallization and racemization for the resolution of 2′,6′-pipecoloxylidide (PPX)

Article abstract of DOI:10.1021/op200268h

An integrated process for the chiral separation of the industrially relevant substance 2′,6′-pipecoloxylidide (PPX), an intermediate in the manufacture of a number of anesthetics, was developed. By combining three different techniques, chromatography, crystallization, and racemization, high productivity was achieved. All unit operations were executed using a common solvent system, full recycling, and a minimum of solvent exchanges or removals. The target molecule was obtained with an enantiopurity of >99.5 wt %.

Full text of DOI:10.1021/op200268h

Article
pubs.acs.org/OPRD
Design of an Integrated Process of Chromatography, Crystallization
and Racemization for the Resolution of 2′,6′-Pipecoloxylidide (PPX)
Jan von Langermann,† Malte Kaspereit,‡ Mozaffar Shakeri,§ Heike Lorenz,† Martin Hedberg,⊥
Matthew J. Jones,⊥ Kerstin Larson,⊥ Bjorn Herschend,⊥ Robert Arnell,⊥ Erik Temmel,†
̈
Jan-Erling Backvall,§ Achim Kienle,∥,¶ and Andreas Seidel-Morgenstern*,†,□
̈
†Max Planck Institute for Dynamics of Complex Technical Systems, Physical-Chemical Foundations of Process Engineering,
Sandtorstraße 1, 39106 Magdeburg, Germany
‡Friedrich-Alexander University Erlangen-Nuremberg, Department for Separation Science and Technology, Egerlandstraße 3,
91058 Erlangen, Germany
§Department of Organic Chemistry, Arrhenius Laboratory, Stockholm University, 106 91 Stockholm, Sweden
⊥AstraZeneca AB, Pharmaceutical Development, Medicines Development, 151 85 Sodertalje, Sweden
̈
̈
∥Max Planck Institute for Dynamics of Complex Technical Systems, Process Synthesis and Process Dynamics, Sandtorstraße 1,
39106 Magdeburg, Germany
¶
Otto-von-Guericke University Magdeburg, Chair of Automation/Modeling and □Chair of Chemical and Process Engineering,
Universitatsplatz 2, 39106 Magdeburg, Germany
̈
ABSTRACT: An integrated process for the chiral separation of the industrially relevant substance 2′,6′-pipecoloxylidide (PPX),
an intermediate in the manufacture of a number of anesthetics, was developed. By combining three different techniques,
chromatography, crystallization, and racemization, high productivity was achieved. All unit operations were executed using a
common solvent system, full recycling, and a minimum of solvent exchanges or removals. The target molecule was obtained with
an enantiopurity of >99.5 wt %.
bupivacaine, and the wider safety margin of (S)-3 allows the use
of higher concentrations and doses compared to that for
bupivacaine, resulting in a lower risk of systemic toxicity and
ensuring better surgical anesthesia.2
1. INTRODUCTION
1.1. Potential Improvements in the Manufacture of
(S)-2,6-Pipecoloxylidide. 2′,6′-Pipecoloxylidide 1 (PPX) and
its derivatives 2−4 (Figure 1) were first prepared and tested as
On an industrial scale, racemic 1 is prepared either from
racemic pipecolinic acid or via hydrogenation of 2′,6′-
picolinoxylidide.3 The (S)-enantiomer is then obtained via
diastereomeric salt resolution using O,O-dibenzoyl-L-tartaric
acid leaving the (R)-enantiomer as waste.4 A further option is
chiral pool synthesis from L-lysine, which has been reported to
give (S)-1 in a reasonable, 38% overall yield from CBz-mono-
protected L-lysine.3a The latter route could perhaps become
competitive after further optimization with an approach using
resolution, but the apparently much more expensive5 mono-
CBz-protected analogue of L-lysine has been reported to be
available through a two-step procedure from L-lysine at no
more than 50−55% overall yield.6 This results in an overall
yield of the route from L-lysine to (S)-1 of around 19% over six
synthetic steps.
Furthermore, there are a number of different methods for
making enantiomerically pure (S)-pipecolinic acid published in
the literature,7 and this is certainly a useful starting material for
making compound (S)-1.8 However, the cost of (S)-pipecolinic
acid5 is unfavorable when compared to that of the available
Figure 1. Structures of 2′,6′-pipecoloxylidide 1 and its derivatives
mepivacaine 2, ropivacaine 3, and bupivacaine 4.
local anesthetics in 19571 and 1 was found to be a useful inter-
mediate for the synthesis of the commercial drugs mepivacaine
2, ropivacaine 3, and bupivacaine 4. For example, bupivacaine is
commonly used in its racemic form as an anesthetic during
childbirth due to its long-lasting effect. Unfortunately, there are
known toxicities associated with this drug. However, the pure
(S)-enantiomer of 3 has a lower systemic toxicity than racemic
Special Issue: INTENANT
Received: September 30, 2011
Published: January 20, 2012
© 2012 American Chemical Society
343
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
alternatives. Consequently, the shorter and more high-
yielding3c,d resolution-based routes are currently more attractive
from an economic perspective. The current AstraZeneca internal
process design utilizes a diastereomeric salt resolution without
recovery of the distomer. It is therefore of interest to investigate
whether the process can be improved by integration of unit
operations by coupling chromatography, enantioselective
crystallization, and racemization for the recovery of the distomer
in accordance with the INTENANT-concept.9 The outcome of
this study is described in this article. The knowledge gained
from using 1 as a case study within INTENANT is also of general
value for future products within the AstraZeneca development
portfolio.
1.2. Coupling of Chromatography, Crystallization,
and Racemization. Given the need to produce enantiopure
(S)-PPX, all available separation strategies were compared with
the aim of designing an efficient, fully integrated separation
process. The main goal was to identify an improved process
employing a combination of different unit operations while
optimizing yield, productivity, and enantiopurity. Preliminary
work within the INTENANT project revealed early on that a
combination of chromatography, crystallization and racemiza-
tion is a suitable approach. Further details are given in the
paper by Kaspereit et al. in this issue of Organic Process Research
& Development.10
discussed below (see section 3.3) in addition to good compati-
bility with the other unit operations (see sections 2 and 4).
Furthermore, the boiling point of DBE is significantly higher
than that of any of the other components in the eluent system,
ensuring good separability both within and outwith the
integrated process.
Within this process, preparative chromatography employing
an eluent system containing DBE, ethanol, and diethylamine
(DEA) provides an enantioenriched solution, which is
concentrated via distillation. The concentration step also serves
to remove ethanol and DEA from the solvent. The concentrated
solution is transferred to the crystallization process where the
solid, enantiopure product is obtained. The residual mother
liquor from the crystallization is recycled into the chromato-
graphic separation since it contains an excess of the desired
(S)-PPX. Simultaneously the (R)-enantiomer-rich stream from
the initial chromatographic separation is racemized and
eventually recycled into the separation cycle. Racemization of
PPX is achieved by a homogeneous catalytic reaction employing
the Shvo catalyst.27b Since only simple concentration adjust-
ments are required between the different operations, process
control is straightforward. Fortunately, azeotropes are absent in
the eluent system (dibutylether/ethanol/diethylamine), and
purification of the spent solvent mixtures is therefore not critical.
This is primarily of relevance for the (re)adjustment of the
eluent system for the next chromatographic separation cycle.
The main advantage of the process is its simple concept
requiring only standard laboratory equipment, thus keeping costs
low. All solvents as well as the undesired (R)-PPX are almost
fully recycled, resulting in high atom efficiency. The inves-
tigations of the individual unit operations (chromatography,
crystallization and racemization) within the common solvent
system are described and discussed in the following sections
(2 − 4). The results are evaluated and summarized with respect
to their usability for an integrated resolution process for PPX.
Nonetheless, the combination of these different unit opera-
tions for the separation of PPX is challenging, since the
individual operations require operating conditions that are not
easily coupled. One critical issue identified early in the develop-
ment is the need for a common solvent system capable of
linking the individual unit operations (Figure 2) while reducing
2. CHROMATOGRAPHY
2.1. Choice of Chromatographic System. The choice of
the stationary phase and the mobile phase is decisive for the
performance of a chromatographic separation. Generally, the
aim is to have high resolution, sufficient solubility, and low
retention times. A significant number of chiral stationary phases
(CSPs) and solvents were screened by pulse experiments on a
250 mm × 4.6 mm column packed with preparative CSPs with
20 μm particle size, and ‘Chiralpak IC’ was identified as the
most suitable CSP for PPX.
With respect to the process concept in Figure 2, the mobile
phase selected should also be compatible with the individual unit
operations. A 85/15/0.1 (v/v/v) DBE/ethanol/diethylamine
mixture was chosen as chromatographic solvent (separation
factor α = 1.42). Ethanol and diethylamine volumes were
selected to optimize retention times while ensuring sufficient
separation efficiency. Beyond 85% of DBE the separation
efficiency increases only slightly, but retention times rise
significantly. The high boiling point of DBE together with the
absence of azeotropes ensures efficient removal of ethanol and
DEA in the subsequent solvent evaporation.
Figure 2. Integrated resolution process for PPX. Initially the racemic
feed is enriched by chromatography. The resulting (S)-PPX-rich
stream is then subjected to crystallization after which the mother
liquor is returned to the chromatographic enrichment. The (R)-PPX-
rich stream is racemized and then returned to the chromatographic
separation. The primary function of the intermediate distillations is to
adjust the composition of the solutions.
the need for solvent swaps or intermediate separations. This
also included the consideration of secondary factors: preferably
low toxicity of all substances involved, low cost and easy
workup in the downstream processes. A manageable, low
process complexity is also favored to ensure the practicability
under industrial manufacturing conditions. Based upon these
considerations, dibutylether (DBE) was chosen as the most
suitable solvent, as it meets all the requirements mentioned
above. The selection criteria included a suitable solubility of
PPX for the final enantioselective crystallization, which is
Despite the high separation factor, full baseline separation
was unfortunately never observed. This is due to a peculiar
adsorption behavior, which will be discussed in more detail
below. However, in view of the combined process this is less
relevant, since chromatography is required to perform only a
partial separation and enrichment. In fact, the combination of
344
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
chromatography and crystallization11−13 required by the scheme
in Figure 2 is particularly useful in cases where chromatographic
performance is limited.
2.2. Column Model. The chromatographic separation step
was designed on the basis of a mathematical column model.
Here the transport-equilibrium model is applied, which consists
of the following component mass balances
Concerning the spatial discretization of the eqs 1, the stage
number Nax = 1250 corresponding to the Dax predicted from
the correlation by Chung and Wen16 leads to unacceptably high
sim
ax
computation times. Therefore a lower value of N = 500 was
chosen for these simulations. The corresponding coefficients
sim
k
were recalculated from eq 2. This speeds up calculations by
f,i
a factor of 4 without observable loss of predictive quality.
Competitive adsorption isotherms q*i(c) were determined by
̅
2
∂q
∂t
∂c
∂t
∂c
∂z
∂ c
1 − ε
ε
i
i
i
i
the inverse method, see for example ref 18. In this method,
which was applied simultaneously to 12 different experimental
chromatograms, the differences between simulated and experi-
mental elution profiles are minimized using an optimizer (here:
fminsearch in Matlab) that adjusts the isotherm parameters.
As mentioned earlier, initial experiments indicated a partially
collaborative adsorption. Thus, several different isotherm
models were investigated. A sufficiently accurate description
was achieved using the following expressions:
+
+ u
= D
ax
2
(1a)
∂z
∂q
∂t
i
= k [q* (c ) − q ], i = (1, 2)
f,i
i
̅
i
(1b)
In these equations, ci and qi are the concentrations of com-
ponent i in the fluid and the loadings on the CSP, respectively.
The competitive adsorption equilibrium is denoted by q*i(c); ε
̅
is the bed porosity, u = Q/(εA) is the interstitial fluid velocity,
Q the volumetric flow rate, and Dax the axial dispersion
coefficient. Equation 1b describes mass transfer between the
two phases by a solid film driving-force approach.
I
b
I
s
i
q = γ·c + q
i
i
I
1 + ∑ b c
j j
j
II
2
The model equations were implemented in Matlab and solved
by an explicit backward−forward finite difference scheme that
approximates the axial dispersion term in eq 1a numerically (see
refs 14 and 15 for details).
2.3. Parameter Determination. Before applying the
above column model the unknown parameters in eqs 1a and 1b
have to be determined experimentally. Table 1 lists the parameter
b c + 2b c + b c c
i
i
q,i i
m R S
2
II
s
+ q
,
II
1 + ∑ b c + ∑ b c +b c c
k
k
q,n n m R S
k
n
i = (R, S); j, k, n = R...S
(3)
This isotherm model implies a simultaneous linear (γ),
Langmuir-type (I) and competitive−collaborative (II) adsorp-
tion behaviors. The number of free parameters was reduced
to 8 by requiring that the initial slopes correspond to the
Table 1. Determination of model parameters for eqs 1a and
a
1b
measured values in Table 1, i.e. q*i(c→0) = Ki. The parameters
̅
are valid in the investigated concentration range up to 70 g/L
parameter
value
determination procedure
(rac.), specifically: γ = 0.6331, qsI = 64.7 g/L, bSI = 0.0316 L/g,
Q [mL/min] 2
chosen as compromise between efficiency and
throughput
bRI = 0, qsII = 46.8 g/L, bSII = 0.1035 L/g, bRII = 0.2056 L/g, bq,S
=
0.0056 L2/g2, bq,R = 0.0258 L2/g2, and bm = 0.0188 L2/g2.
ε
0.706
7.52
retention time t0 = εV/Q of nonadsorbing DBE
The model was validated by simulating independent experi-
ments. Figure 3 shows two decisive examples where asymmetric
Ki
from retention times tR,i of small racemate
injections, using
10.25
160
tR,i = t0[1+(1−ε)/ε Ki]
Napp,i
from small racemate injections, using Ni ≈ 5.54tR2,i
/
w520, with w50 as peak
185
width at half its height
Dax[m2/s]
kf,i[1/s]
sim
2.84 × 10−7 Chung and Wen correlation,16 value corresponds
to Nax = uL/(2 Dax)=1250
1.043
0.686
from eq 2 for Dax = 2.84 × 10−7 m2/s
D
[m2/s] 7.10 × 10−7 corresponds to Nax = 500 (chosen for simulation)
ax
sim
k
f,i
[1/s]
1.485
0.895
from eq 2 for Nax = 500
a
The index i = (R, S) denotes the two enantiomers, where S is weaker
adsorbing.
Figure 3. Validation of the model (lines) against independent experi-
ments (symbols) with asymmetric injection compositions. (Left)
Injection of a 1:3 (S:R) mixture. (Right) Injection of a 3:1 (S:R) mixture.
Injection volume 2 mL, total injection concentration 20 g/L.
values obtained from simple injections of small pulses at dilute
conditions at 25 °C.
Standard methods are used for determination of the para-
meters given in Table 1. Merely the determination of the mass
transfer coefficients, kf,i, requires further explanation. These
were calculated from the measured apparent stage numbers,
Napp,i, the axial dispersion coefficient, Dax, and the retention
factors, k′i, using the equation reported by Guiochon et al.17
mixtures were injected. The agreement between experiment
and model prediction is very good.
2.4. Steady-State Recycling Chromatography. Chro-
matographic separation was performed by a specific operating
mode known as steady-state recycling (SSR).19,20 In this process
a large injection is applied, which results in only partially
resolved elution profiles. The sufficiently purified leading and
trailing edges of the profile are collected as product fractions.
A constant amount of fresh feed is added to the remaining
2
⎛
⎜
⎝
⎞
⎟
2D
u
k′
i
L
u
ax
H (u) =
=
+ 2
i
N
1 + k′ k′ k
⎠
app,i
i
i f ,i
(2)
345
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
Figure 4. Schematic setup of SSR chromatography (left) with elution profiles and cut times in the steady state (right). Here, recycle and fresh feed
are mixed (mixed recycle mode, MR-SSR).19 For this mode shortcut design methods are available.21,22 Alternatively, the mixed fraction can be
recycled directly to the column (closed-loop mode, CL-SSR).20
unresolved part and then reinjected. After several cycles, a
periodic steady state is attained. The principle is illustrated in
Figure 4.
Under preparative conditions a properly designed SSR process
can achieve significantly higher productivity, higher product
concentrations, and lower eluent consumption than batch
chromatography at the same purity and yield.21
The main design parameters for SSR chromatography are the
four cut times (cf. Figure 4) that define the product fractions and
the injection volume. Two shortcut design methods are
available21,22 that allow a simple determination of the cut times
for any purity requirements. Here the method proposed by
Kaspereit and Sainio21 is applied, requiring only a simple
evaluation of single chromatograms, and a rapid design study was
performed by taking the required “input chromatograms” from
simulations. It was found that, due to the specific adsorption
behavior of PPX, performance improves particularly well when
decreasing the purity requirement from the pure enantiomer to
lower values. Due to this behavior the injection concentration
should also be optimized. A rough estimation indicated that a
purity of 95% for both fractions and a fresh feed concentration of
25 g/L (racemic mixture) represent useful conditions for the
coupled process. Figure 5 shows the performance of the SSR
process as a function of the injection volume. It can be seen that
the SSR process (small symbols) clearly outperforms conven-
tional batch chromatography (large symbols). The optimum
injection volume for SSR is approximately 2.5 mL.
The above procedure was applied to design several experi-
mental SSR runs for different purity requirements and different
start-up conditions. In all cases very good agreement between
the experimental results and design specifications was achieved.
As an example, Figure 6 shows the results for the optimum
injection volume of 2.5 mL and the desired purity of 95%. In
this example, the transient start-up period of the process was
essentially eliminated by using the steady-state injection
concentrations predicted by the model for the first two
injections (see refs 21 and 22 for details on this approach).
Steady-state is readily obtained after about three cycles. Elution
profiles, collected fraction concentrations, and purities (sym-
bols) are very close to the simulation results.
Figure 5. Optimized performance of the MR-SSR process as function
of the injection volume for a fresh feed concentration of 25 g/L
(rac.). Open symbols - productivity (product amount per time and
volume of stationary phase). Filled symbols - solvent requirement
(solvent amount per product amount). Large symbols at the left
mark the optimized batch chromatography process that achieves the
same yield (i.e., so-called touching band conditions, no waste
fraction).
Note that a further performance improvement is possible by
choosing the first cut time closer to the first product peak. In
the current case this would result in a temporal overlap of
recycling and injection, which is not accounted for in the
shortcut design method. If the resulting slight purity deviations
cannot be accepted, this can be dealt with by using the model
to fine-tune the cut times, or by temporarily storing the recycle
fraction and reinjecting it in the next-but-one cycle (as was
assumed for Figure 6). Since the purpose of the experiments
performed was to assess process feasibility, such overlap was
prevented by applying safety margins between the injections.
In summary, a highly efficient chromatographic system was
found, which was executed under reduced purity requirements.
A full separation of PPX was neither possible nor necessary,
since the enantioenriched solution was transferred to a
crystallization process yielding the enantiopure product. See
Kaspereit et al. within this issue of Organic Process Research &
Development for further details.10
346
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
Figure 6. Results of an SSR experiment with 16 cycles. Sampled elution profiles of the final cycle (left), concentrations in collected fractions
(middle), and fraction purities (right) in comparison to model predictions. Symbols - experimental results. Lines - simulation results. Filled symbols/
thick lines mark the (S)-enantiomer (the first fraction); open symbols/thin lines mark the (R)-enantiomer (second fraction). Experimental
conditions: fresh feed 25 g/L (rac.); injection volume 2.52 mL; flow rate 2 mL/min; cut times 3.45, 6.29, 7.35, 11.0 min. Injections performed every
7.55 min (corresponds to the cycle time of 11.0 − 3.45 min).
Table 2. Melting temperatures and heats of fusion of the
pure enantiomer and the racemic compound of PPX, both in
their neutral forms
3. CRYSTALLIZATION
Previous crystallization process development and understand-
ing was limited to the hydrochloride salt of PPX, which was
obtained directly from synthesis and used for further processing.
However, the protonated form of PPX is not suitable for
use in the proposed integrated process, and therefore the final
crystallization was redesigned for the neutral form. The
characterization of the crystallization behavior of racemic as
well as enantiopure PPX required for the design process
revealed new opportunities for process solutions for manufac-
ture of the single enantiomer.
melting temperatures -
heat of fusion -
Tf / K
ΔHf / kJ·mol‑1
own work
literature25
own work
literature25
enantiomer
racemate
404.84
387.85
403.15
385.15
25.84
24.12
24.19
23.13
3.1. Physical/Chemical Properties of PPX: Solid Phase
Behavior and Melt Phase Diagram. The solid phase
behavior of the neutral form of PPX defines the overall design
of the integrated process, since the requirements of the crystalli-
zation process define the purity required from the preceding
chromatographic enrichment.23 Here, X-ray powder diffraction
(XRPD) was used to characterize the phase behavior of PPX.
As shown in Figure 7, the powder patterns of the racemate
Figure 8. Binary melt phase diagram of PPX, neutral form; liquidus
curves calculated using the Schroder-van Laar and Prigogine-Defay
̈
equations are indicated by dotted and dashed lines. The eutectic
composition lies at x(S)-PPX = 0.67 (only one-half of the phase
diagram is shown), Te - eutectic temperature, Tf - temperature on the
liquidus line.
similar. The values determined here compare favorably to those
reported by Nemak et al.25
As no decomposition or solid phase transitions were
observed during the measurements on the pure enantiomer
and the racemic compound, the binary phase diagram was
determined. The system was found to behave almost ideally,
the experimentally determined liquidus and solidus lines match
the calculated Schroder-van Laar and Prigogine-Defay lines23
Figure 7. XRPD-patterns of PPX; racemate (black) and enantiomer
(red), indicating a racemic compound-forming behavior.
̈
(Figure 8), and the eutectic composition was found at a mole
fraction of x = 0.67. The eutectic composition is a critical
factor in the integrated process design, as it represents the
minimum enrichment required from the chromatographic step
and the lower composition limit of the system for crystalliza-
tion of the pure enantiomer. This specific position of the
eutectic is favorable for a combination of enantioenrichment via
chromatography and enantioselective crystallization, since only
a moderate enrichment is required.13,26
and the enantiomer are clearly distinct, indicating a racemic
compound-forming system.
The melting temperatures, the corresponding heats of fusion
(Table 2), and the melt phase diagram (see Figure 8) support
the interpretation of the diffraction patterns. The enantiomer
melts at a higher temperature than the racemate, approximately
405 and 388 K respectively, while the heats of fusion are very
347
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
Figure 9. XRPD-screening for solvent-mediated polymorphism and solvate formation for racemic PPX. No polymorphs were identified and only two
solvates (with dichloromethane and methanol at the top of the graph) were found. The results were verified by solution NMR (data not shown). No
solvates were found for the pure PPX-enantiomer.
As seen in the phase diagram (Figure 8) and also verified by
a Tammann plot (data not shown), the PPX-system shows
complete immiscibility in the solid state.
3.2. Solvate and Polymorph Screening. A polymorph
and solvate screen was included early in the process develop-
ment as part of the solvent screen and in order to identify
potential problems with undesired crystal modifications.23 The
results from the screen are also of value to the chromatography
and racemization process development as they ensure that the
most suitable solvent for all operations can be identified. Here,
dibutylether was identified as the most suitable solvent for all
operations.
Only two solvates were found, a dichloromethane solvate
and a methanol solvate. Consequently, these solvents were not
considered in the further investigations. XRPD-patterns for the
solid phases obtained in the screening process are shown in
Figure 9.
While the solution−crystallization-based polymorph screen
was negative, one polymorph of the racemic compound was
identified via melt crystallization. This polymorph has a distinct
XRPD pattern (Figure 10a) and is clearly metastable as it
transforms back to the stable modification with a half-life of
∼300 min (Figure 10a,b). Full thermal characterization is
precluded by an increased transition rate at elevated temper-
atures. As this polymorph is a polymorph of the racemic com-
Figure 10. (a) Transition of the metastable polymorph of the racemic
compound to the stable modification as followed by XRPD; (b) tracking
pound and the target species here is the pure enantiomer, its
existence is of no relevance to the process.
of the intensity evolution for the peak at 10.2° (marked with an arrow in
3.3. Enantioselective Crystallization. The ternary sol-
(a)), revealed a half-life of ∼300 min for the metastable modification.
The measurements were executed at ambient temperature.
ubility phase diagram of PPX was determined in the selected
process solvent, dibutylether. Solubility data were measured in
the temperature range from 5 to 55 °C, and the solubilities of
both the pure enantiomer and the racemic compound were
shown to increase with temperature (Figure 11).
out at the gram-scale at MPI-Magdeburg, later repeated at a
larger scale employing several 100 g of PPX at AstraZeneca,
Sodertalje (results not shown here). The small scale experiment
̈
̈
While the eutectic composition is independent of temper-
ature, a subtle shift of the eutectic composition determined from
the ternary phase diagram measurements to x = 0.7 compared to
the composition of the eutectic from the binary phase diagram
(x = 0.67) was observed. It is likely that this difference is simply
due to experimental error and for the purpose of crystallization
process development the solution value is used. Solid phase
analysis confirmed that only the stable PPX modifications, as
shown in Figure 7, were present.
started with an enantioenriched solution of 93.4% (S)-PPX in
the eluent system dibutylether/ethanol/diethylamine (DBE/
EtOH/DEA), which was obtained from SSR-chromatography
as described in section 2.4. The dilute, unsaturated solution from
chromatography was concentrated by partial solvent evaporation,
which removed all remaining ethanol and diethylamine and a
fraction of the dibutylether. The concentrated solution was
finally charged to the crystallization vessel and (S)-PPX was
obtained with an enantiopurity of ≥99.5%, which fulfills the
minimum requirement of 99.0%. The large-scale-validation at
Initial validation of the suggested separation route of SSR-
chromatography and enantioselective crystallization was carried
348
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
catalyst (Figure 12) was identified (ref.27a−d and below).
(S)-PPX was used for the initial studies, since it was more
easily available than its (R)-form.
Figure 12. Reaction scheme for the PPX-racemization. Use of alcohols
such as ethanol or isopropanol results in an undesired side reaction,
which necessitates the removal of ethanol from the chromatography
eluent prior to racemization.
Figure 11. Ternary phase diagram of PPX in dibutylether. The
solubility increases with temperature, and the eutectic composition is
independent of temperature. Due to the advantageous position of the
eutectic composition and the sufficiently strong temperature depend-
ence of the solubility, enantioselective crystallization of enantiopure
PPX is feasible by means of a simple cooling crystallization.
Initial racemization experiments included isopropanol in
the solvent in order to enhance the solubility of PPX, as the
solubility in dibutylether is limited (section 3.3). The results
of these studies are given in Table 3. Substoichiometric
Table 3. Racemization of PPX in the presence of
a
isopropanol
AstraZeneca confirmed the results from the small-scale experi-
ments, underlining the general applicability of the integrated
process. The mother liquor, depleted but still enriched with
respect to (S)-PPX (≥70%), was recycled into the chromato-
graphic separation after adjustment of the solvent to the specifi-
cation of the eluent system (as shown in the general process
scheme, Figure 2). During the entire crystallization process, no
formation of byproduct, which may interfere with subsequent
chromatography or racemization cycles, was observed.
In summary, the crystallization process facilitates the produc-
tion of enantiopure PPX by crystallization from dibutylether. A
sufficient enantiopurity of PPX was achieved and the remaining
mother liquor was recycled to chromatography, ensuring an
optimal separation process. Racemization of the undesired
(R)-enantiomer is essential to raise the overall-yield of the PPX-
separation process as without this operation the combination
of chromatography and crystallization do not provide sufficient
process improvement.
entry
isopropanol (equiv.)
time (h)
ee (PPX)
byproduct
1
2
3
4
0
1
1
1
1
2
<1
2.4
1
−
0
0.5
b
1
0
c
10
22
1.1
4.5
9.4
a
Reaction conditions: Shvo’s catalyst (2 mol %), 0.5 mmol (S)-PPX,
b
isopropanol (x equiv), dibutylether (2 mL), temp. 140 °C. The
presence of 1 equiv of isopropanol in the racemization mixture
prolonged crystallization of PPX to over 12 h. The presence of 10
c
equiv of isopropanol in the racemization mixture does not allow
crystallization of PPX due to enhanced solubility.
(0.5 equiv) and stoichiometric addition of isopropanol to the
reaction mixture did not affect the racemization rate (Table 3,
compare entries 2 and 3 with entry 1) and no byproducts were
observed. The stoichiometric addition of isopropanol to the
racemization mixture resulted in an extended crystallization time
of PPX from one hour to one night at room temperature. When
(S)-PPX was racemized in the presence of 10 equivalents of
isopropanol, the racemization rate decreased (Table 1, entry 4)
and byproduct formation according to Figure 12, was observed.
Consequently the use of isopropanol was avoided during
PPX-racemization since the corresponding impurities will
eventually accumulate throughout the process. Accumulation
within the recycled material is likely to have a detrimental effect
upon both chromatography and crystallization.
A major advantage of the proposed separation process is the
full recycling of (R)-PPX and all other chemicals used, including
the repeated use of the racemization catalyst. Considering its
cost, recycling of the catalyst is almost mandatory for an
economically advantageous process. Furthermore, the racemi-
zation of highly concentrated PPX solutions is desirable at the
manufacturing scale. Therefore the effect of catalyst recycling
and high PPX concentrations was investigated and the results
are summarized in Table 4.
4. RACEMIZATION
Resolution processes starting from racemic mixtures are limited
to a maximum yield of 50%, namely the eutomer fraction of
the racemic mixture, while the distomer is merely waste. A
racemization of the distomer is therefore highly desirable in
order to recover the eutomer from the distomer stream and to
increase the overall process yield and efficiency.
Within INTENANT, investigation of racemization techniques
included several studies aimed at developing novel methodo-
logies to racemise important structural classes such as amino
acids and their derivatives, amines and alcohols. In case of PPX,
prior state-of-the-art was to use stoichiometric amounts of
sodium ethoxide29 for racemization. Very long reaction times
(>5 d) were required and several side reactions resulted in
undesired impurities. Other studies demonstrated the successful
use of metal catalysts27 and heating in polar protic solvents.28
In the course of the INTENANT work a novel homogeneous
catalytic method for the racemization of (R)-PPX using Shvo’s
349
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
three) is not possible for a racemic compound-forming system
when starting from a racemic mixture and requires prior enrich-
ment to a composition above the eutectic. Clearly, options two
and three lend themselves to coupling within an integrated
process, where the gain in productivity and reduced solvent
consumption of a chromatographic process with lower product
purity requirements may outweigh the cost of an additional
crystallization operation. For PPX, asymmetric synthesis (option
four) is not viable for the simple reason that no effective process
is known. As a consequence options 2 and 3 as well as a racemi-
zation reaction were developed as coupled processes, which
successfully improved the overall process efficiency. Even though
neither of the processes, taken in isolation, operates in an optimal
manner, the combination of the three operations allows their
disadvantages to be minimized while optimizing the entire
process with respect to productivity, efficiency and yield. The
proposed process was finally validated in lab- and large scale
experiments.
On the basis of the encouraging results of the SSR chromato-
graphy experiments (section 2.4), the process was initially
combined with a crystallization step. A longer production run
was conducted in order to obtain a sufficient amount of a
partially enriched solution that was then subjected to an enantio-
selective crystallization. The same operating parameters as those
reported in Figure 6 were applied in an experiment with 126
cycles (16 h). This resulted in 775 mL of the first product
fraction with a purity of 93.4% (design value 95%) with respect
to the eutomer. The overall productivity is about 270 g d−1 L−1,
which is lower than the predicted 730 g d−1 L−1 (Figure 5).
The reduced productivity is due to the safety margins applied
between the injections, as discussed above. From the solution
collected, a total of 470 mg of solid (S)-enantiomer was crystalli-
zed in enantiomerically pure form, which corresponds to a yield
of 54%. It must be noted that a rather conservative crystallization
procedure was applied at this initial stage in order to keep a
sufficient distance to the eutectic composition at 70% in solution.
At this point racemization of the undesired enantiomer was
not considered as the process described above was still under
development. The full integrated process was validated at
Table 4. Racemization of high concentrations of PPX and
repeated use of mother liquor
a
b
entry
1
batch no.
time (h)
ee (PPX)
yield
1
2
3
4
1
1
1
1
0
15
40
70
86
c
2
91
92
c
3
c
4
∼100
a
Reaction conditions: Shvo catalyst (1 mol %), (S)-PPX 1 mmol,
b
Bu2O (2 mL), temp. 140 °C. Yield = (mass of solids isolated after
reaction)/ (initial mass of substrate − mass of substrate taken for
c
sampling). Volume of mother liquor removed by syringe after
reaction was 1.6 mL. Thus, the crystals from the reaction were washed
with a further 400 μL of Bu2O and were then transferred to the new
racemization solution.
Racemization reactions employing higher concentrations
(∼10%, entry 1) and lower concentrations (∼5%, entry 2) of
(S)-PPX were completed in 1 h. With continuous recycling of
the catalyst the catalytic efficiency decreases, while the
enantiomeric excess obtained increases simultaneously from
0% (full racemization) to 15%, 40%, and 70%, respectively.
This effect can be explained by a loss of Shvo’s catalyst in the
mother liquor by 10% based on sample volume in addition to
deactivation of the catalyst during the racemization reactions. In
solution, the catalyst is known to be sensitive to oxygen, the
presence of which results in deactivation. It is worth mentioning
that the reaction mixture gradually turned red from the first to
the fourth batch of racemization. The loss of catalytic activity
was significantly reduced at the large scale, possibly as a result of
the reduced relative surface area of the solvent exposed to the
atmosphere, thus reducing the effect of oxygen.
5. INTEGRATED PROCESS
Combining chromatography, crystallization and racemization
into an integrated process provides for significant improvements
compared to a single purification process. Table 5 comparing
the original process with the individual unit operations and the
combined process, highlights these improvements.
AstraZeneca, Sodertalje, on a larger scale employing several
̈
̈
Starting with option one, the classical diastereomeric salt
resolution as ‘state of the art’ was originally used to obtain the
enantiopure product. The very easy setup and simple process
control is tarnished by the limited maximum yield of 50% and
high waste production. However, from an industrial point of
view, this process was preferred for many years due to low cost.
Option two, chromatographic separation, is an interesting
alternative, but full separation is precluded due to high cost
and the large eluent volumes required. Costs, productivity,
and solvent consumption can be reduced significantly if the
purity requirements from chromatography can be relaxed (see
ref 10), while the purity requirements for the product can
then no longer be met. Enantioselective crystallization (option
100 g of PPX. These experiments confirmed the validity of the
general approach and its applicability to industrial processes.
6. CONCLUSION
A novel, integrated, chiral resolution process combining
chromatography, crystallization, and racemization was inves-
tigated for 2′,6′-pipecoloxilidide (PPX). A significant improve-
ment in process efficiency was obtained by the systematic
coupling of these unit operations. Preparative chromatography
was used to obtain an enantioenriched solution from the initial,
racemic mixture of PPX. Despite the fact that the ‘chromato-
graphic separation’ employed provides unsatisfactory resolution
Table 5. Comparison of PPX-resolution strategies
feasibility for
separation of PPX
option
type
advantage(s)
disadvantage(s)
1
classical resolution-state of the simple process design
art-
waste disposal, max. 50%
yield
yes
2
3
4
chromatographic separation
purity freely adjustable
cost (for bulk production)
enantioenrichment necessary
no process available
complexity
no
no
no
yes
enantioselective crystallization simple process design
asymmetric synthesis
low waste generation
5 (combination of 2 + 3 +
racemization)
integrated process
high yield and productivity, waste
reduction
350
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
when assessed in isolation, the reduced purity require-
ments lead to improved productivity, reduced eluent volumes,
and an output stream that is sufficiently enriched for
enantioselective crystallization. An enantioselective crystallization
is then used to isolate the pure (S)-PPX with a purity of ≥99.5%,
while the remaining, enriched mother liquor is recycled into the
chromatographic separation. The (R)-PPX-rich chromatography
stream is racemized using a homogeneous catalyst, which yields
a racemic mixture of PPX that is also recycled into the
chromatography enrichment, together with fresh feed. The basis
for the entire process is, on the one hand, the existence of a
common solvent system that can be used in all three unit
operations and, on the other hand, the operating conditions for
the individual unit operations that minimize their disadvantages
while improving the overall process.
liquid and solid phases were separated by filtration and analyzed
using HPLC, XRPD, and NMR.
Racemization Conditions. A flame-dried 20-mL reaction
vessel was charged with (S)-PPX and the catalyst; the vessel
was closed, evacuated, and flushed with argon three times. The
solvent and any additive were subsequently added via syringe.
The mixture was stirred at the indicated temperature for a
given period of time. After completion of the reaction time,
the mixture was cooled to 0 °C for 30−40 min and the PPX
crystallized was isolated by filtration. The loss of optical purity
was determined by HPLC analysis (AD column, i-hexane/
ethanol/triethylamine; 85:15:0.1, tR(S) = 10.33, tR(R) = 12.65 min).
Racemic and enantiopure compounds were used as references.
The multiple use of Shvo catalyst was investigated using a
similar procedure, with the exception that racemized PPX was
removed after crystallization and fresh (S)-PPX was charged
prior the next racemization run.
7. EXPERIMENTAL SECTION
(rac)-PPX and (S)-PPX were provided by AstraZeneca both
in their neutral forms and as hydrochloride salts. All other
chemicals were obtained in analytical purity and were used
without any further purification. For racemization studies
dibutylether was dried over molecular sieves prior to use.
Chromatographic Separation. The eluent system con-
sisted of a mixture of DBE/EtOH/DEA. 85/15/0.1 (v/v/v).
A Chiralpak IC column from Chiral Technologies (West
Chester/PA, U.S.A.) was used throughout the whole study.
Temperature control of the column (at 25 °C) was provided by
a Lauda Ecoline RE306 (Lauda, Germany) thermostat. A
modular experimental setup was applied consisting of K-1001
HPLC pumps, UV detector Smartline 2500, and K-6 and K-17
valves for switching (injection, fractionation, and recycling) (all
Knauer GmbH, Germany). The system was controlled via an
RS232 interface and used a custom Labview program (Labview
7.0, National Instruments, U.S.A.). Injections in the SSR
experiments were performed through an HPLC pump using a
K-6 valve as selection valve.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: seidel-morgenstern@mpi-magdeburg.mpg.de. Tele-
phone: +49-391-6110-401. Fax: +49-391-6110-521.
ACKNOWLEDGMENTS
■
This work was funded by the European Union (EU-project
“INTENANT, Integrated Synthesis and Purification of
Enantiomers”; NMP2-SL-2008-214129).
REFERENCES
■
(1) Ekenstam, B.; Egner
11, 1183−1190.
(2) See, for instance, the AstraZeneca product monograph for
ropivacaine hydrochloride 2004 and references therein.
́
, B.; Pettersson, G. Acta Chem. Scand. 1957,
(3) (a) Adger, B.; Dyer, U.; Hutton, G.; Woods, M. Tetrahedron. Lett.
1996, 37, 6399−6402. (b) Rinderknecht, H. Helv. Chim. Acta 1959, 42,
1324. The most cost-efficient and high-yielding routes to racemic 1
judged by overall yield and raw material costs are found in (c) Tullar, B.
F.; Bolen, C. H. GB 1166802, 1969 and 1968. (d) Soni R. R.; Koftis, T.;
Georgopoulou, I; Karagiannidou, E. PCT Int. Appl. WO/2009/089842,
2009.
HPLC analysis of collected samples was performed on an
Agilent HP 1100 using a Chiralcel OD column (150 mm ×
4.6 mm, 5 μm) with n-hexane/isopropanol/triflouroacetate
(85/15/0,1 v/v) as solvent and a UV-detector.
Solid-Phase Investigations and Solubility. Differential
scanning calorimetry (DSC) experiments were performed on a
DSC131 from Setaram (France). The start temperature was
30 °C, the final temperature was 130 °C, and a heating rate of
2 K/min was applied. In order to produce controlled mixtures
for phase diagram determination, weighed amounts of racemate
and/or enantiomer were mixed and dissolved in acetone, and
after evaporation of the solvent the solid was ground to a fine
powder. Ten to fifteen milligrams of the respective mixtures
were used for DSC experiments.
Solid samples were analyzed throughout by X-ray powder
diffraction to identify the solid phases present. An X’Pert Pro
diffractometer (PANalytical GmbH, Kassel, Germany) with Cu
Kα1 radiation was used. Measurements covered a 2θ range from
3° to 40° with a step size of 0.0167°. Investigations of solvates
included the use of NMR to identify possible solvates and
exclude the possibility of solvent mediated solvatomorphism.
Solubility measurements were performed by applying a
classical isothermal method. A known composition of PPX in
dibutylether was prepared (leaving an excess of the solid phase),
and the vessels were placed into a temperature-controlled
bath and magnetically stirred at a constant temperature (within
0.01 K) until equilibrium was attained. Subsequently, the
(4) Tullar, B. F. J. Med. Chem. 1971, 14, 891−892. This procedure
gave a rather low yield for the diastereomeric salt resolution of 26%
(50% maximum). However, internal optimization at former Astra
raised the yield to a robust level of 45% by changing solvent and
process conditions (not shown).
(5) Commercial availability and pricing data were investigated using
the MDL ISIS 2.5 acd commercial catalogue database, available within
AstraZeneca, to obtain both lab-scale and bulk-scale prices. Bulk costs
were arbitrarily set to 10% of the Aldrich or Acros catalogue prices for
each raw material, based upon experience with other compounds. The
relative prices for L-lysine/racemic pipecolinic acid/picolinic acid
appear to be 1:9:1 if prices from the same manufacturer are used for all
three chemicals. With respect to absolute numbers, picolinic acid
appears to cost at most 50 USD/kg as assessed from the database. Z-
protected L-lysine and enantiopure (S)-pipecolinic are both expensive
and cost of the order of 500 USD/kg in bulk and according to the
database. These cost/price levels may, of course, change depending on
demand and other factors.
(6) The most high-yielding procedure was found in Bence, A. K.;
Crooks, P. A. Synth. Commun. 2002, 32, 2075. For an earlier reference
where a lower yield was reported, see Baldwin, J. E.; Killin, S. J.;
Adlington, R. M.; Spiegel, U. Tetrahedron 1988, 44, 2633.
(7) A Scifinder search gave 59 reactions describing the synthesis of
(S)-pipecolinic acid, out of which a few describe methods involving
resolution of racemic material.
351
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352

Organic Process Research & Development
Article
(8) Reported yields for the amide-coupling to give optically pure 1
are low when starting from enantiopure (S)-pipecolinic acid:
(a) Tsuchida, T.; Kaneko, K.; Matsumoto, N. European patent EP
1433782; 2005. (b) Pilli, R. A.; Santos, L. S. Brasilian patent BR
2003002250, 2005 (patent written in Portuguese; short abstract in
English is available). (c) Rangisetty, J. B.; Pullagurla, M. R.; Muthiah,
R. J. J., Jobdevairakkam, C. N., U.S. Patent Appl. 2006/0276654, 2006.
(9) Information about INTENANT is available from the homepage
www.intenant.eu.
(10) Kaspereit, M. et al. Org. Process Res. Dev. 2012, 10.1021/
op200285v, INTENANT special section this issue.
(11) Kaspereit, M.; Gedicke, K.; Zahn, V.; Mahoney, A. W.; Seidel-
Morgenstern, A. J. Chromatogr. A 2005, 1092, 43−54.
(12) Amanullah, M.; Mazzotti, M. J. Chromatogr. A 2006, 1107,
36−45.
(13) Gedicke, K.; Kaspereit, M.; Beckmann, W.; Budde, U.; Lorenz, H.;
Seidel-Morgenstern, A. Chem. Eng. Res. Des. 2007, 85, 928−936.
(14) Antos, D.; Seidel-Morgenstern, A. Chem. Eng. Sci. 2001, 56,
6667−6682.
(15) Antos, D. Gradient Techniques in Preparative Chromatography −
Modelling and Experimental Realization. PhD Thesis, Oficyna
Wydawnicza Politechniki Rzszowskiej, Rzeszow, Poland, 2003.
(16) Chung, S.; Wen, C. AIChE J. 1968, 14, 857−866.
(17) Guiochon, G.; Shirazi, D. G.; Felinger, A.; Katti, A. M.
Fundamentals of Preparative and Nonlinear Chromatography; Academic
Press: Waltham, U.S.A., 2006.
(18) Felinger, A.; Cavazzini, A.; Guiochon, G. J. Chromatogr. A. 2003,
986, 207−225.
(19) Bailly, M.; Tondeur, D. Chem. Eng. Sci. 1982, 37, 1199−1212.
(20) Grill, C. M.; Miller, L. J. Chromatogr. A 1998, 827, 359−371.
(21) Kaspereit, M.; Sainio, T. Chem. Eng. Sci. 2011, 66, 5428−5438.
(22) Sainio, T.; Kaspereit, M. Sep. Purif. Technol. 2009, 66, 9−18.
(23) Jacques, J.; Collet, A.; Wilen, S. H. Enantiomers, Racemates and
Resolutions; Krieger Publishing Company: Malabar, USA, 1994.
(24) Collins, A. N.; Sheldrake, G. N.; Crosby, J. Chirality in Industry:
The Commercial Manufacture and Applications of Optically Active
Compounds; Wiley & Sons: Chichester, UK, 1992.
́
́
(25) Nemak, K.; Acs, M.; Kozma, D.; Fogassy, E. J. Therm. Anal.
1997, 48, 691−696.
(26) Lorenz, H.; Sheehan, P.; Seidel-Morgenstern, A. J. Chromatogr.
A 2001, 908, 201−214.
(27) Some recent examples of racemization of amines using metal
́
catalysis: (a) Pam
̈
̀
ies, O.; Ell, O. A.; Samec, J. S. M.; Hermanns, N.;
, L. K.;
Zhao, D.; Sortais, J.-B.; Paetzold, J.; Hoben, C.; Backvall, J.-E. Chem. Eur.
Backvall., J.-E. Tetrahedron Lett. 2002, 43, 4699−4702. (b) Thalen
́
̈
J. 2009, 15, 3403−3410. (c) Paetzold, J.; Backvall, J. E. J. Am. Chem. Soc.
̈
́
2005, 127, 17620−17621. (d) Thalen, L. K.; Hermanns, N.; Backvall.,
̈
J.-E. Tetrahedron Lett. 2010, 51, 6802−6802. (e) Parvulescu, A. N.;
Jacobs, P. A.; De Vos, D. E. Chem.-Eur. J. 2007, 13, 2034−2043. (f) Kim,
M.-J.; Kim, W.-H.; Han, K.; Choi, Y.K.; Park. J. Org. Lett. 2007, 9, 1157−
̌ ́
1159. (g) Jerphagnon, T.; Gayet, A. J. A.; Berthiol, F.; Ritleng, V.; Mrsic,
N.; Meetsma, A.; Pfeffer, M.; Minnaard, A.J.; Feringa, B. L.; de Vries, J. G.
Chem.-Eur. J. 2009, 15, 12780−12790. (h) A useful review article on
racemization methodologies in general including metal-mediated
racemization: Ebbers, E. J.; Ariaans, G. J. A.; Houbiers, J. P. M.;
Bruggink, A.; Zwanenburg, B. Tetrahedron 1997, 53, 9417−9476.
(28) Dyer, U. C.; Langston, M.; Woods, M. PCT WO 001996022281
A1; 1996.
(29) Internal experiments by Peter Jaksch at AstraZeneca studying
racemization of (R)-1 using a stoichiometric ethoxide base at room
temperature in EtOH, resulted in clean but slow racemization which
was complete in 5 days. Increasing the temperature, of course, resulted
in a much faster racemization but also caused significant byproduct
formation.
352
dx.doi.org/10.1021/op200268h | Org. ProcessRes. Dev. 2012, 16, 343−352