Synthesis of the antimalarial API artemether in a flow reactor

Article abstract of DOI:10.1016/j.cattod.2014.04.014

The earlier developed flow protocol for stoichiometric reduction of an important biologically derivedpharmaceutical precursor, artemisinin, to dihydroartemisinin was extended to a sequential reaction toproduce one of the final APIs, artemether. A highly active heterogeneous catalyst was found for theetherification reaction. The use of QuadraSil catalyst allows to eliminate one step of reaction workup. Acomparative Life Cycle Assessment of both reactions has shown advantages of the flow process over theoptimized literature batch protocols. Results of LCA highlight the significance of solvents in pharmaceut-icals manufacture and the advantage of flow technology, enabling small solvent inventories to be used.

Full text of DOI:10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
Catalysis Today xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Synthesis of the antimalarial API artemether in a flow reactor
a
b
c
a
a,∗
Polina Yaseneva , Dorota Plaza , Xiaolei Fan , Konstantin Loponov , Alexei Lapkin
a
Department of Chemical Engineering and Biotechnology, University of Cambridge, Cambridge CB2 3RA, UK
School of Engineering, University of Warwick, Coventry CV4 7AL, UK
School of Chemical Engineering and Analytical Science, The University of Manchester, Manchester M13 9PL, UK
b
c
a r t i c l e i n f o
a b s t r a c t
Article history:
The earlier developed flow protocol for stoichiometric reduction of an important biologically derived
pharmaceutical precursor, artemisinin, to dihydroartemisinin was extended to a sequential reaction to
produce one of the final APIs, artemether. A highly active heterogeneous catalyst was found for the
etherification reaction. The use of QuadraSil catalyst allows to eliminate one step of reaction workup. A
comparative Life Cycle Assessment of both reactions has shown advantages of the flow process over the
optimized literature batch protocols. Results of LCA highlight the significance of solvents in pharmaceut-
icals manufacture and the advantage of flow technology, enabling small solvent inventories to be used.
Received 21 October 2013
Received in revised form 3 April 2014
Accepted 7 April 2014
Available online xxx
Keywords:
Flow chemistry
Artemisinin
©
2014 Elsevier B.V. All rights reserved.
Malaria
Life cycle assessment
Acid catalysis
1
. Introduction
see Fig. 1 [4,5]. This synthetic protocol involves batch stoichiomet-
ric reduction at low temperatures (0–5 C) with a suspension of
◦
Today batch processing dominates manufacturing of pharma-
NaBH , followed by a multi-step work-up procedure to an inter-
4
ceuticals and specialty chemicals. However, the role of continuous
processes, and especially of clean catalytic processes, in these
important areas of chemical manufacturing is rapidly increasing
mediate product dihydroartemisinin, then followed by a batch
catalytic etherification [6]. A one-pot batch protocol for the two
reactions has been published [7] and recently an updated optimized
batch protocol for the two reactions was also reported [8].
[
1]. Key challenges in translating conventional synthetic protocols
into flow processes are typically related to the issues of low solu-
bility of reactants and products, the use of mutually incompatible
solvents in different steps of the multi-step syntheses, the need for
intermediate work-up steps and finally, the lack of standardized
equipment for assembly of flow catalytic processes.
Here we report the first sequential flow synthesis of an impor-
tant medicinal small molecule of a biological origin, in which a
number of the mentioned problems had to be resolved. Small
molecules derived from natural products represent a major source
of new drugs [2]. A well-publicized example is artemisinin, the
WHO-recommended drug for treating uncomplicated Plasmo-
dium falciparum malaria, which is obtained by extraction from
Artemisia annua [3]. The active pharmaceutical ingredients of
the artemisinin-based drugs are ethers, such as artesunate or
artemether, obtained via reduction of artemisinin into dihy-
droartemisinin (DHA) using sodium borohydride in methanol or
ethanol as solvents, followed by etherification to the final APIs,
Recent significant development in the manufacture of
artemisinin-based drugs, is the introduction of a continuous
photochemical process for production of artemisinin from its bio-
logical precursor dihydroartemisinic acid, which could be obtained
either from extraction or biosynthetically [9]. This allows not only
to diversify the supply chain of this important API, but also gives
impetus to the development of a fully continuous production of
the final ethers using the principles of flow chemistry [10].
In our earlier study we reported a flow protocol for the first
reaction shown in Fig. 1, the reduction of artemisinin to DHA [11].
The flow protocol has considerable technical advantages over batch
protocols: a safer operation due to the high heat transfer capacity of
a microreactor, a higher space-time-yield and a higher selectivity to
DHA, compared to best published batch results. However, the flow
process is using a more expensive reducing agent, which is stored
and supplied as a THF solution, and is using THF or Me-THF solvents.
Although Me-THF is being regarded as one of the new generation of
cleaner bio-derived solvents, its life-cycle impact reported by GSK
is quite significant [12]. Thus, we anticipated that the flow proto-
col may have large environmental impact through the increased
life-cycle impact of the overall process due to the more complex
reducing agent and the nature of the solvent used.
∗ Corresponding author. Tel.: +44 0 1223 330141; fax: +44 0 1223 334796.
E-mail address: aal35@cam.ac.uk (A. Lapkin).
http://dx.doi.org/10.1016/j.cattod.2014.04.014
0
920-5861/© 2014 Elsevier B.V. All rights reserved.
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
2
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
Fig. 1. A common route to conversion of artemisinin to artemisinin-based APIs, exemplified by artemether.
The final step in the synthesis, catalytic etherification, can be
performed with inorganic acids, such as HCl [8], or with acidic
resins, such as Amberlyst-15 [6]. However, both catalysts require
downstream neutralization with a base, followed by separation
of the product from the base solution. In order to develop a
fully continuous process it is desirable to eliminate the inter-
mediate work-up steps, as well as to avoid the use of different
solvents.
Recently a sulphonic acid-modified spherical silica material was
introduced as a scavenger, suitable for applications in flow chem-
istry, QuadraSil [13]. The high specific surface area and, more
importantly, the high density of acid sites in comparison with e.g.,
Amberlyst-15, are likely to result in high activity of this material in
acid-catalyzed reactions, if the strength of acid sites is sufficient. Its
spherical morphology is particularly attractive for applications in
flow chemistry due to the low pressure drop, in contrast to unstruc-
tured micro packed beds. Thus, a micro spherical carbon support
with a similar particle size range to that of QuadraSil was earlier
reported in a multiphase mm-scale hydraulic diameter packed bed
reactor [14].
2.2. Batch reduction of artemisinin
Procedures for batch reduction of artemisinin to DHA with
NaBH4 and with superhydride have been described elsewhere
[6,11]. Briefly, artemisinin (200 mg, 0.71 mmol) was suspended in
methanol (10 mL) under moderate stirring and cooled in an ice-
◦
water bath to ca. 4 C. Sodium borohydride (67 mg, 1.77 mmol, 2.5
equivalent) was added in portions to the suspension over a period
of 5 min. The reaction mixture was stirred vigorously under nitro-
gen until TLC showed no reactant left in the reaction mixture (ca.
90 min). Then the reaction mixture was neutralized (pH 5–6) with
50% (v/v) of a mixture of acetic acid/methanol (added in portions
of 50 ␮L each time) and evaporated to dryness under vacuum (at
◦
40 C).
2.3. Work up procedure
Dry residue was extracted with ethyl acetate three times (10 mL
each time). Completeness of extraction was monitored by TLC. The
combined ethyl acetate extracts were dried with Na SO for 6 h, fil-
2
4
The aim of the present study is two-fold: to extend the flow
protocol of conversion of artemisinin to a catalytic etherifica-
tion step, and to perform a cradle-to-gate life cycle assessment
study, which would help to develop understanding of the sources
of environmental impacts in the new process, and thus develop
strategies for their reduction or elimination in further process
optimization.
tered, and evaporated to dryness to give a white flake-like product.
2.4. Batch etherification of dihydroartemisinin
Etherification of a solution of ␣/␤-DHA to artemether with
Amberlyst-15 (0.25 and 1 molar equivalents of active sites/DHA)
in a batch reactor was performed as follows: DHA (0.3138 g;
1
.1035 mmol) was dissolved in 2-Me-THF (45.79 mL) under stirring
2
. Experimental
at ambient temperature. Anhydrous methanol (MeOH/DHA = 12.6
excess) was injected into the flask. Then the calculated amount of
Amberlyst-15 was added to the mixture to start the reaction. Reac-
tion was typically continued for 4 h. Periodic aliquots were taken
for HPLC analysis. The size of the aliquots was sufficiently small
to ignore the influence of the change of volume on the reaction
2.1. Materials and chemicals
To obtain DHA for etherification reaction, artemisinin was
reduced by superhydride solution in a batch reactor according
to a procedure described elsewhere [11]. Artemisinin samples
were kindly donated by Ipca Laboratories Ltd, India and Botani-
cal Extracts EPZ Ltd., Kenya. The obtained DHA was analyzed by
HPLC and dissolved in a mixture of Me-THF (Sigma Aldrich, 99.0%)
and methanol (Analar Normapur, 99.8%, anhydrous) to the concen-
tration of 23.8 mmol L . Then DHA etherification to artemether
was performed in batch and flow reactors with two different types
of acidic catalysts, Amberlyst-15 (Sigma Aldrich, concentration of
◦
kinetics. The same experiment was run at 40 C using an oil bath.
Work-up procedure was as follows: 15 mL of cold water was
added to the reaction mixture. Amberlyst-15 was separated from
the mixture by filtration and washed by 5 mL of anhydrous
methanol. The organic mixture was quenched with 15 mL of 5 wt%
NaHCO3 and stirred for 1 h in an ice bath. 2-Me-THF and methanol
solvents were evaporated. The product was separated from water,
washed several times with water and dried in a Buchner flask.
Etherification of DHA to artemether with QuadraSil-SA (0.25 and
0.41 molar equivalents of active sites/DHA) in a batch reactor was
performed as follows: DHA (0.1452 g; 0.511 mmol) was dissolved in
2-Me-THF (21.18 mL) under stirring at ambient temperature. Anhy-
drous methanol (MeOH/DHA = 12.6 excess) was injected into the
flask. Then the calculated amount of QuadraSil catalyst was added
to the prior prepared mixture and stirred for 4 h. The progress of
−1
−
1
2
−1
acid sites is ∼4.7 mol kg , specific surface area 53 m g ), and
QuadraSil-SA (kindly provided by Johnson Matthey, concentration
−
1
2
−1
of acid sites is ∼1.4 mol kg , specific surface area 700 m g ). Dif-
ferent equivalents of acidic sites to DHA were used and specified
along with the reported results. Sodium bicarbonate (Fisher, 99.7%)
was used in both batch and flow conditions to quench the reaction
catalyzed by Amberlyst-15.
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
Table 1
3
reaction was monitored by HPLC. The same experiment was run at
◦
Impact categories of CML method implemented in this study.
4
0 C using an oil bath.
Results of etherification were quantified in terms of the yield
Category
Name
Unit
of ␤-artemether with respect of the initial concentration of DHA.
In the absence of a standard for ␣-artemether it was not routinely
analysed.
AP
GWP
HTP
FAETP
MAETP
TAETP
POCP
EP
Acidification potential
Climate change
kg SO2-Eq
kg CO -Eq
2
Human toxicity potential
Freshwater aquatic ecotoxicity
Marine aquatic ecotoxicity
Terrestrial ecotoxicity
Photochemical oxidation (summer smog)
Eutrophication potential
Depletion of abiotic resources
Cumulative energy demand
kg 1,4 DCB-Eq
kg 1,4 DCB-Eq
kg 1,4 DCB-Eq
kg 1,4 DCB-Eq
kg ethylene-Eq
kg PO4-Eq
2.5. Flow reduction of artemisinin
Flow reduction of artemisinin to DHA is described in details else-
ADP
CED
kg antimony-Eq
mJ-Eq
where [11]. The same protocol was followed in the current study,
using the same microreactor X3 by Little Things Factory and the
pumps.
Datasets of the production of the materials involved in the pro-
cesses, generated electrical consumptions and transportation were
either extracted from Ecoinvent 2.2 library or modeled in our group.
Pre-chains for NaBH4, superhydride and 2-Me-THF were supplied
by Environmental Clarity, USA. Electricity production mixture in
the UK and standard transport distances were used to model energy
consumption and transportation, respectively [18]. The life cycle
inventory was built for a process scenario of a continuous plant for
converting 20 t per annum of artemisinin to artemether.
2
.6. Flow etherification of dihydroartemisinin
Flow etherification of solutions ␣/␤-DHA to artemether was per-
formed using Omnifit column of 10 mm ID and adjustable catalyst
bed length of 4–12 cm. The column was thermostated in an oil bath
at 40 C. The feed mixtures of DHA in 2-Me-THF and MeOH were
pumped using Kontron 42 and Knauer 100 HPLC pumps. In the case
of Amberlyst-15 catalyst the product mixture was quenched by
◦
5
% aqueous sodium bicarbonate to neutralize the leached acid. In
the case of QuadraSil catalyst the catalyst was diluted with broken
quartz prior to packing.
3. Results and discussion
From our previous study we know that batch reduction of
2.7. HPLC analysis
artemisinin to DHA in methanol with NaBH4 can achieve a 90%
yield in 100 min at 4 C, whereas reduction under flow conditions
◦
One hundred microliter of a liquid from the reaction flask or
from the collected product at the end of the flow reactors was
using 2-Me-THF solvent and superhydride reducing agent results
in a higher yield of 95% and reaction could be run at ambient tem-
perature with a residence time of ca. 20 s [11]. Yield is reported as
combined yield to ␣- and ␤-DHA isomers with the ratio of the two
isomers in solution at the end of reaction 45/55% by NMR [11]. In
the solid state only ␤-DHA isomer is present which, upon dissolu-
tion equilibrates with the ␣-isomer. This interconversion is shown
in Supplementary electronic information, Fig. S1.
taken and quenched with 100 ␮L of 0.5% NaHCO . The solution was
3
diluted with 1 mL of acetonitrile straight before filtering through
a Nylon syringe filter and placing the sample into a HPLC sam-
ple vial rack to reduce the possibility of decomposition of DHA in
acetonitrile.
Shimadzu Prominence HPLC equipped with ELSD and UV/vis
photodiode array detector was used with Thermo Hypersil,
◦
2
50 × 4.6 mm, 5 ␮m, C18 column. ELSD cell temperature was 40 C.
3
.1. Batch etherification of dihydroartemisinin
◦
Column oven temperature was 45 C. Mobile phase composition
was acetonitrile 70%, water 30% v/v and acetic acid 0.1% v/v with a
Etherification of a mixture of ␣/␤-isomers of DHA to artemether
−1
flow rate of 1 mL min . Volume of injection was 20 ␮L with a run
time 17 min. DHA was detected on ELSD and artemether-␤ on the
UV detector using absorbance at 216 nm. DHA and ␤-artemether
were quantified against standards. Due to unavailability of ␣-
artemether standard, it was not routinely quantified. Assuming
that ␣-artemether has a similar response factor as the calibrated ␤-
artemether, in several experiments the yield of ␣-artemether was
estimated to be ca. 16–24% when the yield of ␤-artemether was
between 53 and 74%, with the ratio of ␤/␣ ca. 3. Although a differ-
ent length HPLC column was used to that reported in [8], a similar
order of peaks was observed, which allow us to suggest that the
earlier reported impurity – 9-epi-artemether – was also produced.
Its quantification was not performed; a very crude estimate based
on the same response factor as ␤-artemether suggests <9% yield of
the impurity.
over Amberlyst-15 catalyst was performed at different catalyst
loadings and at different temperatures, see Fig. 2. It was shown that
increasing the molar equivalent of acid with respect to the reactant
results in the increase of ␤-artemether yield: at 170 min of reaction
2.8. Life cycle assessment methodology
To build-up the cradle-to-gate life cycle inventory (LCI) of both
batch and flow production of DHA from artemisinin and artemether
from DHA, and to perform life cycle assessment, Umberto for LCA
5
.6 software was used [15]. The environmental impact scores were
calculated for CML 2001 impact categories, see Table 1 [16], and
energy consumption was evaluated by Cumulative Energy Demand
Fig. 2. Yield of artemether-␤ as a function of reaction time during batch etherifica-
(
CED) method [17].
tion catalyzed by Amberlyst-15. Amount of catalyst and temperatures are shown.
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
4
The rate constants attained with QuadraSil catalyst are con-
siderably higher than those in the case of the more conventional
Amberlyst catalyst, compared for the similar amount of acid sites
in the reaction system. We also highlight that the amount of
Amberlyst required to attain high conversion of DHA is in excess
of stoichiometry of the acid sites to the reactant. It is therefore
ambiguous to call Amberlyst a catalyst in this particular reaction. A
significant leaching of acid was observed with the Amberlyst cata-
lyst (see Supplementary electronic information, Fig. S2). Activity of
the leached acid is not necessarily identical to that of the supported
acid and this may cause the gradual reduction in the reaction rate,
observed in the batch reaction profiles and, hence, the stoichio-
metric excess of solid acid that was required for the reaction. In our
experiments the reactions catalyzed by fresh Amberlyst required
a base quench, without which no product was isolated at the end
of the reactions. This is also related to leaching of acid from the
catalyst. Leaching of acid from Amberlyst was tested with pure
methanol and hence is not related to any upstream impurities.
A potential work-around is to pre-leach any poorly attached acid
groups prior to experiments. This was not tested in this study as
this will lead to the significant reduction in activity: we observe a
clear drop of activity of Amberlyst catalyst over time. This reduc-
tion of activity would make flow process with this catalyst even
less attractive. Thus, current industrial strategy of batch reaction
with Amberlyst followed by a base quench and work up cannot be
effectively directly translated into a flow process.
The reaction with QuadraSil proceeds much cleaner and no acid
leaching was observed (see Supplementary electronic information,
Fig. S2). Stability of acid sites and, most likely, good accessibility of
sites in the mesoporous structure of QuadraSil could be responsible
for this.
In our study fresh Amberlyst catalyst was used in each run and
no attempt to recycle the catalyst was used, mainly due to appar-
ent deactivation during the first run. However, QuadraSil catalyst
was used repeatedly in more than five experiments with no appar-
ent deactivation. A specific long-term flow run and large number
of batch reactions would be required to properly characterize recy-
clability of this catalyst.
Fig. 3. The dependence of artemether-␤ yield on the amount of QuadraSil catalyst
as a function of reaction time in batch etherification at 40 C.
◦
time the samples with 0.25, 0.5 and 1 molar equivalents of active
sites gave 13, 28 and 39% yields, respectively. Increasing of reaction
◦
temperature from ambient to 40 C for the reaction with 1 molar
equivalent of acid sites resulted in the increase of reaction yield
from 39 to 70% at 170 min of reaction time.
The results of etherification of DHA to artemether over QuadraSil
catalyst are shown in Fig. 3. Using a higher ratio of active sites
of QuadraSil to DHA, 0.41 vs 0.25, results in a higher yield of
artemether. Thus, when 0.41 molar equivalents are taken instead
of 0.25, the yield of artemether at 170 min increases from 41 to 57%.
Initial rates were estimated from the batch data at short reac-
tion times and first order rate constants were calculated. MeOH was
used in excess, therefore DHA can be considered as limiting species
and the rate law for conversion of DHA to ␤-artemether can be writ-
ten as Eq. (1), assuming the first-order kinetics with respect to DHA.
0
Applying the boundary conditions, t = 0, C = C
; t = t, C = CDHA, the
DHA
integral form of rate equation can be obtained as Eq. (2).
dCDHA
0
r = −rDHA
=
= kCDHA = kC
(1 − XDHA
)
(1)
(2)
DHA
dt
3.2. Flow etherification reaction
ꢀ
ꢁ
0
ꢂ
ꢃ
C
1
DHA
ln
= ln
= kt
The final process flow scheme is shown schematically in Fig. 5.
CDHA
1 − XDHA
Reaction mixture at the exit of the reduction is quenched with an
acid, which could be used as an additional source of active sites for
the etherification reaction. Therefore, the intermediate purification
step is not critical. However, in the present study the feed into the
second reaction was purified as described in Section 2. The second
reactor is a packed-bed column, operating at ambient or slightly
where k is the first order rate constant (min 1) and XDHA is the
conversion.
Plotting the experimental data obtained from batch exper-
iments against time (shown in Fig. 4), one can see a linear
0
DHA
relationship between ln(C
/CDHA) and t can be established, which
◦
validates the use of the first-order kinetics (Eq. (2)).
elevated temperature (40 C). The second reaction does not require
Fig. 4. First order kinetic plots for QuadraSil and Amberlyst-15 catalysts.
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
5
Fig. 5. Schematic process flow scheme for a tandem conversion of DHA into artemether.
base quench unless free acid is present in the reaction mixture,
originating either as a result of leaching from Amberlyst or in the
case of no work-up after the artemisinin reduction step.
QuadraSil than in the case of Amberlyst. The reason is due to a much
closer to the plug-flow regime in the QuadraSil-packed column.
Nevertheless, such a simple model describes behaviour of the flow
reactor packed with Amberlyst also reasonably well. We later use
this model to estimate the amount of catalyst required for a scaled-
up flow process, in order to calculate LCA impacts for a given rate
of production of artemether.
The flow reactor packed with QuadraSil shows a significantly
better performance, compared with Amberlyst catalyst. A higher
yield is attained at a shorter residence time. This is also clearly seen
by comparing results of the model simulation for both catalysts, see
Supplementary electronic information.
One of the more significant advantages of the QuadraSil catalyst
is its stability. Blank experiments with both catalyst showed that
Amberlyst is continuously leaching acid, see Supplementary infor-
mation. This results in the necessity to quench the reaction with a
weak base in order to obtain the product. In the case of QuadraSil
catalyst the base quench is not necessary.
The packed-bed reactor hydrodynamics was evaluated using a
standard tracer technique. Details of the experiments and exper-
imental data are given in Supplementary electronic information.
Residence time distribution (RTD) analysis confirmed that for the
case of the QuadraSil-packed column the reactor is behaving close
to the ideal plug-flow RTD. In the case of the Amberlyst catalyst the
deviation from the ideal plug flow behaviour is more pronounced.
Therefore, we would expect a larger deviation of the experimen-
tal and simulated results. The latter are required for designing a
large-scale process to enable a life cycle assessment study.
Performance of both catalysts is shown in Fig. 6, along with the
simulated results. First of all we note that a simple plug-flow model
based on the first-order kinetics works a lot better in the case of
Generally, comparison of performance of batch and flow reac-
tions can be done on the basis of space-time-yield (STY) which is
calculated as kg of product produced within 1 h in a reactor of 1 L
volume. This could either involve the work-up stage or not. If the
work-up stage is to be included, the results are likely to be highly
inaccurate for the lab-scale experiments due to significant discrep-
ancies in washing volumes and filtration/drying procedures (times
and volumes).
In the case of the reduction reaction, based on the experimental
results published earlier [11], and using only reaction time, sol-
vent volume and product yield, and excluding work-up, batch and
−
1
−1
flow STYs are 0.011 and 1.60 kg h
L , respectively. The consid-
erable reduction in STY in the flow experiment was attained due to
a massive reduction in reaction time, which was due to a signifi-
cant increase in reaction temperature, as well as the change of the
reducing agent.
In the case of the etherification reaction the reaction times
are comparable between the batch and the flow experiments.
Therefore, we do not expect a massive improvement in the flow
process over a batch one. Based on the reactions performed at
◦
4
0 C with QuadraSil-SA catalyst at 0.41 molar ratio to the reactant,
−
3
−1 −1
h whereas
the batch STY is estimated to be 1.4 × 10 kg L
in the case of a 4 cm length packed bead and liquid flow rate
−
1
−3
−1 −1
h . This
of 0.48 mL min
the estimated STY is 65 × 10 kg L
increase in STY of the flow catalytic process over the batch one is a
manifestation of a well-known effect of flow chemistry: in the case
of the flow process the relative amount of a catalyst to reactant is
orders of magnitude higher than what is typically used in batch.
This results in considerably higher volumetric production rates,
which manifest as larger STY.
3.3. LCA of the combined process
Fig. 6. Comparison of simulated reactor performance (␤-ARM yield) with experi-
mental results using both, Amberlyst-15 catalyst (a) and QuadraSil catalyst (b) at
different flow rates and reactor lengths.
Cradle-to-gate life cycle assessment is performed to reveal the
impact of the significant changes in the materials upon the switch
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
6
Fig. 7. CML impact scores and CED for artemisinin to DHA reaction. Comparison of flow and batch processes and flow process without THF for superhydride.
from the established batch protocols and the new flow synthe-
sis. Due to the change of solvent and reducing agent in the first
reaction, and the change of the catalyst in the second, the envi-
ronmental impacts of the two alternative processes (all batch vs
flow) are expected to be different. To obtain a more detailed infor-
mation about both flow reactions we show separately comparative
assessment of the reduction and of the catalytic etherification.
Reduction in flow with superhydride is compared with reduc-
Looking into its impact in more detail we recognize that this is
largely due to the fact that it is being supplied as a THF solution.
Simulation of an alternative scenario when the reducing agent is
supplied in Me-THF which is then recycled, results in a significant
reduction of CED, see Fig. 8, and also results in an overall lover LCA
impacts of the flow process compared to the batch, see Fig. 7. The
impact of the solvent Me-THF is relatively small, mainly due to its
relatively small inventory in the case of recycle, which is one of the
key features of the flow processes.
tion in batch using NaBH , as the best literature recipe and, what we
4
believe to be, the industrial recipe. Fig. 7 shows CML impacts and
cumulative energy demand (CED) for three alternatives (log scale):
conventional batch reduction of artemisinin to DHA, reduction in
flow using Me-THF solvent and superhydride reducing agent, and
a hypothetical scenario in which THF solvent (reducing agent is
supplied in THF) is removed from the impacts, thus simulating the
scenario when reagent is supplied in Me-THF and recycled along
with the reaction solvent.
In the case of etherification of dihydroartemisinin to artemether
we show comparison of the recently published batch protocol using
HCl as a catalyst [8] with the flow process based on QuadraSil-SA
catalyst. To illustrate the significance of the solvents we performed
analysis of the scaled-up potential flow process without solvent
recycle as well as with solvent recycle. Results are shown in Fig. 9 for
CML impacts and CED and in Fig. 10 for the detailed contributions
towards CED.
It is clear that introduction of the more complex solvent and
the reducing agent result in a significant increase in the environ-
mental impacts and CED if we compare the best batch results with
the reduction of artemisinin in flow. However, we cannot distin-
guish which is the most important factor without running different
process scenarios with the LCA model. Fig. 8 shows the relative con-
tributions of different factors on CED, which allows one to elaborate
the importance of solvent vs energy input vs transport, etc.
From the data in Fig. 8 it is evident that the unoptimized flow
process is dominated by the impact of superhydride reducing agent.
The impacts of the flow process appear to be worse than that
of batch. However, analysis of contributions towards CED, Fig. 10,
shows that the most dominating impact in the case of the flow
process is the solvent. Similar to the analysis performed for the case
of artemisinin reduction step, reduction of solvent inventory by
introducing its recycling within the process allows to significantly
reduce the impacts of the flow process, making it advantageous
over batch technology.
Fig. 8. Individual processes contributions to CED of artemisinin to DHA reaction in
batch and flow conditions.
Fig. 9. CML impact scores and CED for DHA to ARM reaction. Comparison of flow
and batch processes.
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014

G Model
CATTOD-9041; No. of Pages7
ARTICLE IN PRESS
P. Yaseneva et al. / Catalysis Today xxx (2014) xxx–xxx
7
for very useful suggestions with regards to manuscript improve-
ment and methodology.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/j.cattod.
2
014.04.014.
References
[
[
1] T.L. LaPorte, C. Wang, Continuous processes for the production of pharmaceu-
tical intermediates and active pharmaceutical ingredients, Curr. Opin. Drug
Discovery Dev. 10 (2007) 738–745.
2] D.J. Newman, G.M. Cragg, Natural products as sources of new drugs over the
last 25 years, J. Nat. Prod. 70 (2007) 461–477.
Fig. 10. Individual processes contributions to CED of DHA to ARM reaction in batch
and flow conditions.
[3] A.A. Lapkin, P.K. Plucinski, M. Cutler, Comparative assessment of technologies
for extraction of artemisinin, J. Nat. Prod. 69 (2006) 1653–1664.
[
[
[
4] D.L. Klayman, A.J. Lin, N. Acton, J.P. Scovill, J.M. Hoch, W.K. Milhous, et al., Iso-
lation of artemisinin (quinghaosu) from Artemisia annua growing in the United
States, J. Nat. Prod. 47 (1984) 715–717.
5] A. Brossi, B. Venugopalan, L.D. Gerpe, H.J.C. Yeh, J.L. Flippenanderson, P. Buchs,
et al., Arteether, a new antimalarial drug—synthesis and antimalarial proper-
ties, J. Med. Chem. 31 (1988) 645–650.
6] S. Buzzi, A. Presser, M. von Freyhold, Determining A Viable Protocol for
the Derivatisation of Artemisinin Into Dihydroartemisinin, Report for
Medicine for Malaria Ventures, 2007, Available online from www.mmv.org
4
. Conclusions
Derivatization of an important bio-pharmaceutical molecule
artemisinin into its final API artemether was demonstrated as a
sequence of two reactions under flow conditions for the first time.
A highly effective catalyst for etherification of dihydroartemisinin,
QuadraSil, was found. Close to quantitative yield of the product
artemether was obtained. The catalyst shows excellent stability in
the reaction and allows to significantly reduce the complexity of
the product work-up.
The process could potentially be run as a tandem sequence with-
out intermediate work-up of dihydroartemisinin. A detailed LCA
study of both reaction steps shown the importance of solvents in
pharmaceutical processes: replacement of THF by Me-THF, as well
as introduction of solvent recycle lead to significant decreases in
LCA impacts of the flow processes, making them cleaner than the
best literature batch reactions.
(
%
http://www.mmv.org/sites/default/files/uploads/docs/publications/7%20-
204 Derivatisation DHA-AS-Dec14 fincor.pdf).
[7] C. Singh, P. Tiwari, A one-pot conversion of artemisinin to its ether derivatives,
Tetrahedron Lett. 43 (2002) 7235–7237.
[
8] R.W. Stringham, D.S. Teager, Streamlined process for the conversion of
artemisinin to artemether, Org. Process Res. Dev. 16 (2012) 764–768.
[9] D. Kopetzki, F. Levesque, P.H. Seeberger, A continuous-flow process for the
synthesis of artemisinin, Chem. Eur. J. 19 (2013) 5450–5456.
10] A. Lapkin, P.K. Plucinski, Engineering factors for efficient flow processes in
chemical industries, in: S.V. Luis, E. Garcia-Verdugo (Eds.), Chemical Reaction
Processes Under Flow Conditions, RSC, Cambridge, 2010, pp. 1–43.
[
[11] X. Fan, V. Sans, P. Yaseneva, D.D. Plaza, J.M.J. Williams, A. Lapkin, Facile stoichio-
metric reductions in flow: an example of artemisinin, Org. Process Res. Dev. 16
(
2012) 1039–1042.
[
12] R.K. Henderson, C. Jimenez-Gonzalez, D.J.C. Constable, S.R. Alston, G.G.A. Inglis,
G. Fisher, et al., Expanding GSK’s solvent selection guide—embedding sustaina-
bility into solvent selection starting at medicinal chemistry, Green Chem. 13
(
2011) 854–862.
[
13] M.D. Hopkin, I.R. Baxendale, S.V. Ley, An expeditious synthesis of imatinib and
analogues utilising flow chemistry methods, Org. Biomol. Chem. 11 (2013)
Acknowledgements
1822–1839.
[
[
14] X. Fan, P.K. Plucinski, A.A. Lapkin, Liquid phase hydrogenation in a structured
compact reactor, Catal. Today 147S (2009) S313–S318.
15] ifu Institut fur Umweltinformatik, Hamburg, Germany, Umberto Manual Ver-
sion 5, ifeu Institut fur Energie und Umweltforschung, Heidelberg, Germany,
2011.
DP is grateful to University of Warwick for sponsoring her Ph.D.
study. The research leading to these results has received funding
from the European Community’s Seventh Framework Programme
for SYNFLOW under grant agreement no. FP7-NMP-LARGE-3-2009-
[16] J.B. Guinee, Life Cycle Assessment—An Operational Guideline to the ISO
standards—Parts 1–3, Ministry of Housing, Spatial Planning and the Environ-
ment and the Center Of Environmental Science, Leiden University, Netherlands,
2
46461. HPLC instrument was funded by Medicines for Malaria
Ventures (MMV, Switzerland). We gratefully acknowledge Ipca
Laboratories, India and Botanical Extracts EPZ Ltd. Kenya, for sup-
ply of artemisinin and Johnson Matthey for providing QuadraSil-SA
catalyst and its characterization data. We are grateful to Dr John
Suberu for help with the HPLC analytical protocol, and to referees
2
001.
[17] German Association of Engineers-Society for Energy Technology (VDIGET):
Cumulative energy demand - terms, definitions, methods of calculation, VDI
guideline no. 4600:1997. Berlin, Germany: Beuth Verlag GmbH, 1997.
18] Jh.-J. Althaus, Overview and Methodology, in: N.J. Rolf Frischnecht (Ed.), Ecoin-
vent Report 1, Swiss Centre for Life Cycle Inventories, Dubendorf, 2007.
[
Please cite this article in press as: P. Yaseneva, et al., Synthesis of the antimalarial API artemether in a flow reactor, Catal. Today (2014),
http://dx.doi.org/10.1016/j.cattod.2014.04.014