Mesoscale flow chemistry: A plug-flow approach to reaction optimisation

Article abstract of DOI:10.1021/op7000707

In recent years, chemistry in flowing systems has become more prominent as a method of carrying out chemical transformations, ranging in scale from analytical-scale (microchemistry) through to kilogram-scale synthesis (macrochemistry). The advantages are

Full text of DOI:10.1021/op7000707

Organic Process Research & Development 2007, 11, 704−710
Mesoscale Flow Chemistry: A Plug-Flow Approach to Reaction Optimisation
Rob C. Wheeler,* Otman Benali, Martyn Deal, Elizabeth Farrant, Simon J. F. MacDonald, and Brian H. Warrington
GlaxoSmithKline, Gunnels Wood Road, SteVenage SG1 2NI, United Kingdom
Abstract:
medicinal chemistry groups. On this scale, where only a few
milligrams of the target compound are required for screening,
the financial cost benefit from carrying out process optimi-
sation for each chemical transformation is heavily out-
weighed by the need to obtain as much biological information
as quickly as possible. As a result, the synthetic routes
identified are not necessarily the cheapest or most efficient
as the chemistry is often designed to produce multiple
products from common intermediates. However, once a
potential candidate is selected, 50-100 g will be required
for toxicology studies, and this material will generally be
synthesised using the original, nonoptimal route. Once the
candidate passes the toxicological studies, the synthetic route
is usually transferred to Chemical Development in order to
produce 0.5-5 kg material by a Fit For Purpose (FFP) route
for First Time in Human (FTIH) studies. Here, the process
may be modified in order to increase yields, reduce costs,
or remove unattractive steps (e.g., chromatography), although
the tight timeframes involved mean that it is preferable to
make as few modifications to the original route as possible.
Once the compound passes FTIH studies it will progress to
Pre-Clinical Development (PCD), where kilograms of mate-
rial are required for clinical trials. Here, the potential for
lowering the cost of goods and the drive for easier, safer
production are paramount, and the synthesis may be further
modified in order to provide a manufacturing-viable route.
It is clear to see how the chemical route of a drug will
undergo many alterations as it is transferred from high-
throughput chemistry, through medicinal chemistry and
chemical development groups, towards manufacturing. Clearly,
a method of streamlining this process to enable seamless
transfer of the route from one group to another would offer
a potential for large cost- and time savings, resulting in drugs
being available to patients as early as possible. Here, flow
chemistry is an attractive potential solution as it should allow
for rapid early-stage reaction optimisation and direct scale-
up.3 However, it is the area of microchemistry which is
currently attracting the most interest. Numerous chemical
transformations in microreactors have been demonstrated,4
but it is the potential to couple this technology to a flow
assay in order to provide fully automated, iterative lead
generation and optimisation that makes the approach even
more attractive.5
In recent years, chemistry in flowing systems has become more
prominent as a method of carrying out chemical transforma-
tions, ranging in scale from analytical-scale (microchemistry)
through to kilogram-scale synthesis (macrochemistry). The
advantages are readily apparentsincreased control of condi-
tions leading to greater reproducibility, scaleability, and in-
creased safety/reduced losssalthough its acceptance as a viable
synthesis technique has been limited due to its drawbacks,
primarily precipitation, liquid handling, and diffusion of the
reaction within the reactor. Here, we present details of a system
which bridges the gap between micro- and macroreactors and
has enabled fast reaction optimisation (using small amounts of
reagents) and subsequent multigram scale-up using a com-
mercial reactor.
Introduction
The development of new pharmaceutical compounds is a
lengthy, expensive, and dynamic process.1 At early stage
(gene to candidate), the business driver is to identify a
potential drug candidate quickly in order to obtain a strong
intellectual property position as soon as possible. Once a
candidate is identified and the patent is filed, the focus shifts
to getting the drug to market as quickly as possible in order
to maximise the revenue generated (i.e., before the patent
expires or competitor compounds are marketed). In terms
of the chemistry process, the remits for the business units at
each end of this spectrum are clearly very different, and the
chemical route is certain to undergo numerous changes as
the compound progresses from hit generation through to
manufacturing.2 Figure 1 highlights these key processes
involved in development of a drug compound and shows
clearly how the chemical process is developed and/or
optimised within each business unit.
Up until candidate selection, potential drug compounds
are usually synthesised on a milligram scale for in vitro and
in vivo screening, with lead generation generally occurring
via high-throughput methods, and lead optimisation via
* To whom correspondence should be addressed. E-mail: Rob.C.Wheeler@
gsk.com. Telephone: 01438 768631.
(1) (a) Pritchard, J. F.; Jurima-Romet, M.; Reimer, M. L. J.; Mortimer, E.; Rolfe,
B.; Cayen, M. N. Nat. ReV. Drug DiscoVery 2003, 2, 542. (b) MacCoss,
M.; Baillie, T. A. Science 2004, 303, 1810. (c) Rubin, A. E.; Tummala, S.;
Both, D. A.; Wang, C.; Delaney, E. J. Chem. ReV. 2006, 106, 2794. (d)
Herrling, P. L. Progr. Drug Res. 2005, 62, 2.
(2) (a) Butters, M.; Catterick, D.; Craig, A.; Curzons, A.; Dale, D.; Gillmore,
A.; Green, S. P.; Marziano, I.; Sherlock, J.-P.; White, W. Chem. ReV. 2006,
106, 3002. (b) Ragan, J. A.; Murray, J. A.; Castaldi, M. J.; Conrad, A. K.;
Hill, P. D.; Jones, B. P.; Kasthurikrishnan, N.; Li, B.; Makowski, T. W.;
McDermott, R.; Sitter, B. J.; White, T. D.; Young, G. R. ACS Symp. Ser.
2004, 870, 39.
(3) (a) Acke, D. R. J.; Stevens, C. V. Org. Process Res. DeV. 2006, 10, 417.
(b) Wakami, H.; Yoshida, J. Org. Process Res. DeV. 2005, 9, 787.
(4) (a) Salimi-Moosavi, H.; Tang, T.; Harrison, D. J. J. Am. Chem. Soc. 1997,
119, 8716. (b) Mitchell, M. C.; Spikmans, V.; de Mello, A. J. Analyst 2001,
126, 24. (c) Skelton, V.; Greenway, G. M.; Haswell, S. J.; Styring, P.;
Morgan, D. O.; Warrington, B.; Wong, S. Y. F. Analyst 2001, 126, 7. (d)
Garcia-Egido, E.; Spikmans, V.; Wong, S. Y. F.; Warrington, B. H. Lab
Chip 2003, 3, 73.
704
•
Vol. 11, No. 4, 2007 / Organic Process Research & Development
10.1021/op7000707 CCC: $37.00 © 2007 American Chemical Society
Published on Web 06/27/2007

Figure 1. The chemical activities involved in the development of a drug (*FFP ) Fit For Purpose).
Our group is involved in delivering gram-kilogram
quantities of material to support precandidate selection, and
we were interested in utilising flow chemistry as a method
of performing rapid reaction optimisation and scale-up.
Additionally, if we could identify a scalable and efficient
synthetic route at an early stage, we would be able to rapidly
synthesise sufficient material for early toxicology studies on
Figure 2. Commercial flow systems: Syrris AFRICA (left) and
CPC CYTOS (right).
request. However, scaling up a flow process is not necessarily
straightforward, as the physical behaviour of a fluid flowing
through a channel varies with its velocity and channel size.
A more detailed explanation can be found elsewhere,6 but
the basic premise is that, for common organic solvent solu-
tions, laminar flow will occur in small channels/at low flow
rates (e.g., microfluidics) and mixing proceeds purely by
diffusion. Indeed, static mixers are often employed in macro-
scale laminar flow processes to ensure that thorough mixing
occurs.7 In larger channels/at higher flow rates (e.g., macro-
fluidics), fluid behaviour becomes unpredictable and mixing
occurs in a turbulent environment. With this in mind, we
felt that a mesoscale microfluidic system offered the most
flexibility, as it would allow us to identify a scaleable process
using low volumes of reagents. A literature search indicated
that two commercial flow systems were available that seemed
to meet our requirements, the CYTOS8 and AFRICA9 sys-
tems, marketed by CPC and Syrris, respectively (Figure 2).
The CYTOS, designed and marketed by CPC (who also
market a lower volume model, CYTOS-M), is essentially a
stand-alone microreactor which may be connected to any
other peripheral device (e.g., pumps, fractioners, etc.). The
key advantage of the CYTOS is the complex “splitting” of
the reagent inputs into multiple low-volume microchannels,
which provides excellent thermodynamic control for the
mixing of reagents. Additionally, this facilitates high material
throughput via what is essentially a microchemistry process.
Syrris’s AFRICA system, on the other hand, is a complete
flow chemistry system, consisting of pumps and valves to
deliver reagents, a temperature-controlled glass reactor chip,
and a fraction collector. The low volume of the chip means
that the need for a complex mixing approach, such as in the
CYTOS, is not required, although the downside is that the
throughput is reduced.
However, although both of these devices provide an
adequate means of performing flow synthesis, they both
suffer from three major drawbacks:
(5) (a) Dittrich, P. S.; Manz, A. Nat. ReV. 2006, 5, 210. (b) Whitesides, G. M.
Nature 2006, 442, 368. (c) Wong Hawkes, S. Y. F.; Chapela, M. J. V.;
Montembault, M. QSAR Comb. Sci. 2005, 24, 712. (d). Breslauer, D. N.;
Lee, P. J.; Lee, L. P. Mol. BioSyst. 2006, 2, 97.
(6) Draad, A. A.; Kuiken, G. C. D.; Nieuwstadt, F. T. M. J. Fluid Mech. 1998,
377, 267.
(7) Thakur, R. K.; Vial, C.; Nigam, K. D. P.; Nauman, E. B.; Djelveh, G. Chem.
Eng. Res. Des. 2003, 81, 787.
(8) (a). Mauger, C.; Buisine, O.; Caravieilhes, S.; Mignani, G. J. Organomet.
Chem. 2005, 690, 3627. (b) Lui, S.; Fukuyama, T.; Sato, M.; Ryu, I. Org.
Process Res. DeV. 2004, 8, 477.
(1) Reaction Dispersion. In most flowing systems, it is
common practice to first fill the system with an appropriate
solvent before making any injections, so that the system may
equilibrate and be pressurised. However, the problem with
this approach is that, once injected, the reagents/reaction will
start to disperse due to the effect of parabolic flow. This
gives rise to concentration gradients, and potentially a varying
reaction profile across the injected reaction plug. The upshot
(9) (a) Baxendale, I. R,; Deeley, J.; Griffiths-Jones, C. M.; Ley, S. V.; Saaby,
S.; Tranmer, G. K. Chem. Commun. 2006, 2566. (b) Extance, A. R.; Benzies,
D. W. M.; Morrish, J. J. QSAR Comb. Sci. 2006, 25, 484.
Vol. 11, No. 4, 2007 / Organic Process Research & Development
•
705

is that the leading and trailing edges of the reaction should
ideally not be collected, which means that larger volumes
of reagents will be required for reaction optimisation. A
common solution to this issue is to introduce air gaps at the
start and end of the injected solution, although this is by no
means a reliable method as the air may dissolve into the
solution, particularly if heat is being applied.
(2) Liquid Handling. CPC’s CYTOS is operated by
pumping the reagent solutions into the reactor directly
through the supplied pumps. Ideally this approach should
be avoided, as it requires that the pumps exhibit excellent
chemical compatibility over a wide range of reagents and
solvents. A further issue is that the pumps must be cleaned
out thoroughly in between reactions, which increases both
solvent consumption and process time. Syrris avoid this
problem in their AFRICA system by introducing the reagents
into the system via injection loops, a concept commonly used
in HPLC systems. However, the main issue with the
AFRICA system is that it uses dual-syringe pumps to provide
a constant system flowsthis approach clearly limits the flow-
rate range, as the maximum flow achievable is governed by
the size of the syringes and by the rate at which they can
refill without cavitation.
(3) Solid Formation. In all microfluidic devices, the major
issue is precipitation within the channels/capillaries, giving
rise to system blockages. At best, a light blockage may be
relatively easy to clear (e.g., by pumping at higher pressure,
sonication, introduction of a different solvent) and provides
the user with little more than an inconvenience. At worst,
with heavy precipitation, the reactor may become blocked
irreversibly and may have to be discarded. Additionally, the
system must be equipped with pressure detectors and
sufficiently intelligent software to be able to detect when a
blockage occurs, thus enabling unattended operation.
Figure 3. Schematic of the fluidic connections in the PFR
(shows system aspirating reagents).
used a carbon monoxide/ionic liquid flow in the carbony-
lation aryl halides,16 we could find no published material
which demonstrated monophasic synthesis of drug-like
compounds via a plug-flow approach. We realised that this
approach would provide a potential solution to our liquid
handling and dispersion issues and subsequently developed
a plug-flow chemistry system capable of carrying out
chemical reactions on a µL-mL scale, depending on the
quantity of material required. A more detailed description
of the platform can be found elsewhere (article in press),
but in brief, two chemical reagents are loaded into loops
using standard HPLC pumps and valves, and these are then
injected into a PTFE reactor via a simple “T” junction (Figure
3). Liquid handling is performed by a Kawasaki six-axis
robot (Figure 4). The reaction plugs are then tempered, either
via conventional or microwave heating, before being col-
lected automatically (reaction plug detection is triggered via
the difference in the refractive index between the fluorous
and reaction solvents). In this way, optimum reaction
conditions can be determined using low volumes of reagents
(100 µL) and then scaled up directly on the same systemsa
concept we validated on numerous chemical reactions,
including nucleophilic aromatic substitution17 (Scheme 1) and
diazo transfer chemistry18 (Scheme 2). Our perfluorinated
solvent of choice was perfluorodecalin (PFD), as it has low
solubility in most organic solvents (although it is partially
soluble in hexane and diethyl ether), and its high boiling
point (142 °C) allowed us to operate at elevated temperatures.
Even upon strong heating, we observed that organic reaction
plugs appeared to remain intact, and plug flow was still
observed at temperatures in excess of 120 °C. An acceptable
alternative to PFD is perfluoromethyldecalin (PFMD), which
displays physical behaviour similar to that of PFD but has
the added advantage that it has a melting point of -40 °C,
which offers the potential to perform subambient chemistry.
Figure 5 shows a photograph of an organic solvent plug
(orange) surrounded by a PFD in a 0.75 mm i.d. PFA tube.
In practice, the reaction plugs we injected were much larger
than shown (>100 µL), but the image highlights the low
miscibility between the two phases.
Results and Discussion
In seeking to address the issues outlined above, we
became aware of a communication by Song et al.,10 who
demonstrated that a fluorous solvent may be used as a system
solvent in order to enable formation of discrete, aqueous,
microliter-sized droplets in a flowing system. The group have
subsequently demonstrated that the approach is applicable
to numerous applications, including protein crystallisation,11
measurement of reaction kinetics,12 chemical synthesis13 and
high-throughput screening.14 In addition to controlling dis-
persion, this approach enhances mixing within individual
plugs, an effect which has been utilised by Ahmed et al.15
to accelerate biphasic reactions in microreactors.
Although multiphase plug-flow reactions have been
demonstrated by other groups, such as Rahman et al., who
(10) Song, H.; Tice, J. D.; Ismagilov, R. F. Angew. Chem. Int. Ed. 2003, 42,
767.
(11) Zheng, B.; Roach, L. S.; Ismagilov, R. F. J. Am. Chem. Soc. 2003, 125,
11170.
After conducting this proof of concept (PoC) work, we
next aimed to demonstrate optimisation and scale-up of a
(12) Song, H.; Ismagilov, R. F. J. Am. Chem. Soc. 2003, 125, 14613.
(13) Shestopalov, I.; Tice, J. D.; Ismagilov, R. F. Lab Chip 2004, 4, 316.
(14) (a) Zheng, B.; Ismagilov, R. F. Angew. Chem., Int. Ed. 2005, 44, 2520. (b)
Chen, D. L.; Ismagilov, R. F. Curr. Opin. Chem. Biol. 2006, 10, 226.
(15) Rahman, T.; Fukuyama, T.; Kamata, N.; Sato, M.; Ryu, I. Chem. Commun.
2006, 21, 2236.
(16) Ahmed, B.; Barrow, D.; Wirth, T. AdV. Synth. Catal. 2006, 348, 1043.
(17) Gustin, D. J.; Sehon, C. A.; Wei, J.; Cai, H.; Meduna, S. P.; Khatuya, H.;
Sun, S.; Gu, Y.; Jiang, W.; Thurmond, R. L.; Karlsson, L.; Edwards, J. P.
Bioorg. Med. Chem. Lett. 2005, 15, 1687.
(18) Roberts, E.; Sanc¸on, J. P.; Sweeney, J. B. Org. Lett. 2005, 7, 2075.
706
•
Vol. 11, No. 4, 2007 / Organic Process Research & Development

Figure 5. Formation of an organic reaction plug (orange
solution) in a perfluorinated solvent system in 0.75 mm i.d. PFA
tubing.
Scheme 3. Synthetic route to key intermediate C
Table 1. Effect of plug size on the reactivity of the
nucleophilic substitution of 2-nitrofluorobenzene with
tryptamine (Scheme 1)
Figure 4.
run
plug size
% conversion
Automation of the plug-flow reactor concept. Reagent/product
solutions are moved to injection/collection stations by a Ka-
wasaki six-axis robot; reactions are then performed as outlined
in Figure 3.
1
2
3
4
5
0.2
0.5
1.0
2.0
5.0
78
78
76
75
75
Scheme 1. Nucleophilic aromatic substitution
Scheme 2. r-Diazo-â-keto ester synthesis
In carrying out the initial PoC work, we wanted to
determine whether reactions carried out in plug format
observed the same reaction profile regardless of plug volume.
To this end, we studied the nucleophilic aromatic substitution
of 2-nitrofluorobenzene with tryptamine. Reaction plugs of
various volumes were formed and passed through a reactor
for 6.5 min at 80 °C. One drop of each reaction was collected,
diluted with acetonitrile, and analysed by LC/MS (Table 1).
The results show that a consistent reaction profile is observed
for all plugs, and this suggests that reactions may be
optimised on a 200 µL scale and then scaled up to provide
the quantity of material required.
With successful PoC in hand, we next wanted to assess
how effectively we could apply optimal reaction conditions
to different substrates. An internal chemistry project required
2 g each of the R-diazo-â-keto ester, as shown in Scheme 2.
The suggested procedure involved diazo transfer onto the
â-keto ester using trisyl azide (method A, Scheme 2), and
in studying this chemistry for R ) Me, we found that a 15
min reaction at room temperature gave the desired product
in approximately 100% conversion (LC/MS). These condi-
tions were successfully scaled up (via injecting multiple
reaction plugs) to give the diazoester in 85% yield, after
multistep process. A three-step synthetic route (Scheme 3)
was provided to us by a medicinal chemistry group, with 50
g of compound C being required to support precandidate
selection. The chemistry involved oxidation of a secondary
alcohol using NMO/TPAP19 and subsequent trifluorometh-
ylation,20 followed by a hydrogenation step. Here we aimed
to demonstrate that the batch conditions could be optimised
in flow using small quantities of reagents, and that these
conditions could be scaled up to produce the required 50 g
of compound C.
(19) Griffith, W. P.; Ley, S. V.; Whitcombe, G. P.; White, A. D. J. Chem. Soc.,
Chem. Commun. 1987, 1625.
(20) Prakash, G. K. S.; Krishnamurti, R.; Olah, G. A. J. Am. Chem. Soc. 1989,
111, 393.
Vol. 11, No. 4, 2007 / Organic Process Research & Development
•
707

Table 2. Yields for the formation of r-diazo-â-keto esters
% yield
R group
method A
method B
Me
CHF2
CF3
85
34
6
0
82
88
chromatography. However, upon transferring these conditions
to R ) CHF2/CF3 we were unable to obtain the same result
due to the labile nature of the ketone group, and under these
conditions we obtained predominantly ethyldiazoacetate. A
recent article21 suggested that we should be able to convert
ethyldiazoacetate back to the desired product by reacting it
with the appropriate anhydride under basic conditions
(method B, Scheme 2). This approach proved to be more
successful, and excellent yields were observed for the
fluorinated substrates, although no reactivity was observed
with acetic anhydride (summarised in Table 2).
Figure 6. Thales H-Cube.
reaction. Additionally, as the reaction volume for both the
oxidation and trifluoromethylation reactions exceeded the
volume of the reactor, the concept of a discreet reaction plug
became obsolete.
As predicted, the chemistry transferred smoothly, and the
process time for the synthesis of compound B was reduced
to approximately 13 h, including a purification step at each
stage. For the liquid handling of reagents on this scale, we
elected to pump the reagent solutions directly into the system
via HPLC pumpssnot our preferred approach, but one which
allowed us to validate that the chemistry could be success-
fully transferred between systems.
The final step was carried out on the H-cube22 (Figure
6), a system which provides the means to carry out safe and
scaleable hydrogenation reactions. The optimum conditions
were found to be a 0.27 M solution flowing over a 10%
Pd/C cartridge (CatCart 70) at 1.3 mL/min, and we success-
fully synthesised 50 g of compound C in this manner (synthe-
sised via three ∼16 g batches, as the processing time for
each batch was approximately 5 h). In performing the synthe-
sis, we collected the output in fractions to ensure that the
catalyst remained active throughout the course of the run.
Towards the end of the second batch, we observed that the
reactivity began to fall away, and the catalyst had to be re-
placed. The partially reacted material was mixed with the
remaining unreacted solution, and passed back through the
hydrogenator a second time, giving the desired compound
in 100% conversion. For the third and final batch, a new
CatCart was used, and all 16 g was successfully processed
in one run.
Our next project involved scaling up an existing medicinal
chemistry process in order to produce 50 g of compound C
(Scheme 3). The route and conditions were provided to us
from medicinal chemistry and we noted that the first two
steps required slow, controlled mixing of the reagents and
purification via flash column chromatography. The final step
involved hydrogenation, and although the product was
obtained in good yield, the reaction required handling large
volumes of hydrogen gas and removal of the catalyst via
filtrationsa process which is unattractive on a large scale.
We decided to use the batch conditions as a starting point
and transferred the first two steps to our Automated Plug
Flow Synthesis (APFS) system. We observed that each step
could be driven to completion within 15 min via careful
tweaking of the reaction conditions. Purities were also greatly
increased, and the purifications were simplified. (Although
chromatography was still employed, we simply aimed to
remove baseline impurities, and as a result we were able to
massively overload the silica). However, even though we
successfully demonstrated that transfer of the process to a
flow setup would offer us considerable advantages, we
realised that we were limited by the APFS’s throughputs
the projected time to carry out the 50 g synthesis of the first
two steps in this way (excluding purification) would be
approximately 52 h, compared with 40 in batch. A potential
solution to this issue could be to replace the 6-port valves
with 8- or 10-port valves, providing a dual-loop/continuous-
feed system, although the throughput would still be restricted
by the low reactor volume. In order to accomplish the goal
of 50 g synthesis, we turned to CPC’s CYTOS, which is
designed for continual flow applications. Due to the complex
reactor design, which involves splitting the main flow into
multiple, parallel microchannels, we were unable to suc-
cessfully transfer the plug-flow setup to this system as the
injected plugs emerged from the reactor heavily fragmented.
As a result, we elected to attempt the scale-up in the CYTOS
under non-plug-flow conditions, as we felt that any differ-
ences in reactivity at the leading/trailing edges of the reaction
would be minimised by the relatively large scale of the
Overall, the yield for the synthesis of compound C was
increased from 52% in batch to 71% using our optimised
flow approach, and the process time (including purification)
was reduced by approximately 25 h (a further reduction could
have been achieved if hydrogenation could have been
paralleled out to increase the throughput). A summary
comparing the batch and flow processes is given in Table 3.
Conclusion
In summary, we have developed a system capable of
performing chemical reactions on a microliter scale in order
(21) Wang, Y.; Zhu, S. J. Fluorine Chem. 2000, 103, 139.
(22) Jones, R. V.; Godorhazy, L.; Varga, N.; Szalay, D.; Urge, L.; Darvas, F. J.
Comb. Chem. 2006, 8, 110.
708
•
Vol. 11, No. 4, 2007 / Organic Process Research & Development

Table 3. Summary of the processes involved with the
synthesis of alcohol C
been designed to fit on top of a hotplate/stirrer. The reaction
time was then defined by altering the flow rates of the pumps
accordingly (reaction time ) reactor volume/total flow rate).
Generic Flow Setup (CPC Cytos). The system was
configured so that it consisted of a mixer unit (2 mL internal
volume) and 1-3 reactor units (15 mL internal volume each),
and the reagent solutions were pumped through the reactor
using Jasco PU2080 HPLC pumps. The reaction time was
defined by altering the flow rates of the pumps accordingly
(reaction time ) total reactor volume/total flow rate).
N-[2-(1H-Indol-3-yl)ethyl]-2-nitroaniline. APFS system
is used. Solution A ) 2-nitrofluorobenzene (1.41 g, 10
mmol), DMF (15 mL); solution B ) tryptamine (1.60 g, 10
mmol), DMF (15 mL). Reaction plugs were formed (1:1
stoichiometry, v/v) and injected into a reactor, which had
been preheated to 80 °C, at a total flow rate of 0.3 mL/min,
giving a reaction time of 6.5 min. One drop of each reaction
was collected (from approximately the middle of the plug)
into 1.5 mL water for LC/MS analysis.
Diazoester Synthesis. Method A. Ethyl 2-Diazo-3-oxo-
butanoate. APFS system is used. Solution A ) ethylaceto-
acetate (3.2 mL, 25 mmol), acetonitrile (25 mL); solution B
) DABCO (5.5 g, 49 mmol), acetonitrile (35 mL); 11 × 5
mL reaction plugs (1:1 stoichiometry, v/v) were injected
sequentially into a reactor at ambient temperature, with a
total flow rate of 0.45 mL/min, giving a reaction time of 15
min. The combined reaction plugs were quenched into a
stirred flask of water (100 mL), which was subsequently
extracted with ether (3 × 30 mL). The combined organic
extracts were washed with brine (25 mL), dried over
magnesium sulfate, filtered, and concentrated in vacuo to
an oil. Hexane was added and the mixture cooled in an ice
bath to induce precipitation of TrNH2, which was removed
by filtration, and the filtrate was purified by flash column
chromatography (DCM) to afford the diazoester (3.31 g, 85%
yield). 1H NMR (CDCl3) δ 4.28-4.34 (2H, q, J ) 7.2 Hz),
2.48 (3H, s), 1.32-1.36 (3H, t, J ) 7.2 Hz).
process yield
time (h) (%)
step
1
conditions
purification
Batch
slow addition of TPAP evaporate then flash
24
24
69
82
to a DCM solution
of the alcohol/NMO,
keeping temperature
below 25 °C.
chromatography
(14 gsilica/1 g crude)
Stir overnight
2
slow addition of TBAF acidify (2 N HCl) and
to a THF solution of
the ketone/silane at
0 °C; stir overnight
extract with ether,
followed by flash
chromatography
(19 g silica/1 g crude)
3
1
mix reagents and
filter catalyst through
celite and concentrate
solution
6
9
92
86
stir under H2 for 5 h
Flow
15 min reaction,
throughput of
3.14 mL/min
wash with water
and then pour
directly onto a
silica column
(1 g silica to 1 g
crude) to remove
baseline peaks
2
3
5 min reaction,
throughput of
3.4 mL/min
acidify (2 N HCl)
and extract with
ether, followed by
flash chromatography
(2 g silica/1 g crude)
4
84
98
1.5 min reaction time
(approximately),
throughput of
concentrate collected
solution
16
1.3 mL/min
to provide optimal and scaleable reaction conditions. As the
conditions are readily transferable to other flowing systems,
throughput is increased, and processing times are shortened
from traditional batch processes. This system bridges the gap
between micro- and macroflow systems and offers the first
step towards a seamless chemical process in drug discovery.
Method B. Ethyl 2-Diazo-4,4,4-trifluoro-3-oxobutanoate.
APFS system is used. Solution A ) ethyldiazoacetate (2.1
mL, 20 mmol), pyridine (1.8 mL, 22 mmol), DCM (20 mL);
solution B ) trifluoroacetic anhydride (3.5 mL, 30 mmol),
DCM (20 mL); 5 × 5 mL reaction plugs (1:1 stoichiometry,
v/v) were injected sequentially into a reactor at 38 °C, with
a total flow rate of 0.45 mL/min, giving a reaction time of
15 min. The combined reaction plugs were quenched into
water (50 mL), which was subsequently extracted with DCM
(2 × 35 mL). The combined organic extracts were washed
with brine (20 mL), dried over magnesium sulfate, filtered,
Experimental Section
All reagents and solvents were of analytical grade and
were used without further purification. NMR spectra were
recorded on a Bruker Avance Ultrashield 400 using tetra-
methylsilane (TMS) as an internal standard. LC/MS analyses
were carried out on an Agilent series 1100 HPLC coupled
to a Waters Micromass ZQ mass spectrometer. Chromatog-
raphy was performed using an ISCO Combiflash system, or
with IST Isolute Flash Si cartridges.
Generic Flow Setup (APFS). The equipment was set up
as outlined in the schematic in Figure 4. Two reagent
solutions (A and B) were prepared and then aspirated into
loops (20′, 0.03′′ i.d. PFA tubing) on a Vici six-port valve
using Milligat M6 pumps. The reaction plugs were then
formed by using Jasco PU2080 HPLC pumps to drive the
two reagents back out of the loops, where they met at a
PEEK Y-junction (0.75 mm i.d., Anachem). Once formed,
the plug was then flowed through a reactorsa coil of 0.75
mm i.d. PFA tubing (6.75 mL internal volume), which was
incubated by wrapping around a metal cylinder which had
1
and concentrated in vacuo to an oil, the diazoester by H
1
NMR analysis (3.76 g, 88% yield). H NMR (CDCl3) δ
4.34-4.39 (2H, q, J ) 7.2 Hz), 1.34-1.37 (3H, t, J ) 7.2
Hz).
Ethyl 2-diazo-4,4-difluoro-3-oxobutanoate: method B
followed, 82% yield. 1H NMR (CDCl3) δ 6.47-6.74 (1H, t,
J ) 53 Hz), 4.33-4.39 (2H, q, J ) 7.2 Hz), 1.34-1.38 (3H,
t, J ) 7.2 Hz).
Ketone A. CPC CYTOS is used with three reactor units,
giving an internal volume of 47 mL. Solution A ) alcohol
Vol. 11, No. 4, 2007 / Organic Process Research & Development
•
709

(80 g, 294 mmol) + acetonitrile (580 mL); solution B )
TPAP (5.16 g, 14.7 mmol) + NMO (58.5 g, 500 mmol) +
acetonitrile (580 mL). The two solutions were pumped into
the CYTOS at a flow rate of 1.57 mL/min each, giving a
reaction time of 15 min. The output of the reactor was
concentrated in vacuo, diluted with DCM (30 mL), and split
into three batches to simplify the purification. Each batch
was poured onto a silica column (50 g), and the column was
then washed with DCM (180 mL). The eluent was then
concentrated in vacuo to afford the clean ketone A (68.25
g, 86%).
Alcohol B. CPC CYTOS is configured with one reactor
unit, giving an internal volume of 17 mL. Solution A )
ketone (32.5 g, 120 mmol) + TMS‚CF3 (40.4 g, 284 mmol)
+ THF (52 mL); solution B ) TBAF (5.16 g, 14.7 mmol)
+ THF (65 mL). The two solutions were pumped into the
CYTOS at a flow rate of 1.7 mL/min each, giving a reaction
time of 5 min. The temperature was maintained at 25 °C.
The reactor output was directed over stirred, 2 M aqueous
hydrochloric acid (250 mL), which was subsequently ex-
tracted with ether (2 × 100 mL). The combined ethereal
extracts were washed with brine (30 mL), dried over
magnesium sulfate, filtered, and concentrated under reduced
pressure to give the crude title compound. The crude was
then dissolved in DCM (10 mL) and poured onto two silica
columns (50 g); each column was then eluted with DCM
(60 mL fractions). The fractions which contained the product
by TLC were combined and concentrated in vacuo to give
clean alcohol B (34.3 g, 84%).
Alcohol C. The H-Cube was fitted with a CatCart 70 10%
Pd/C cartridge and the system purged with ethanol. The
hydrogen flow was then activated using “full H2” mode at a
temperature of 80 °C. A solution of the alcohol (34.2 g, 100
mmol) in ethanol (340 mL) was then flowed through the
system at 1.3 mL/min, and the system was washed through
with a further 10 mL ethanol before the hydrogen flow was
stopped. The collected material was concentrated in vacuo
to give the hydrogenated compound C (16.61 g, 104% yield).
Acknowledgment
We gratefully acknowledge Rodrigo Zapiain (Warwick
Manufacturing Group) and Alan Stanley (GSK) for providing
elegant and reliable automation, John Holden (Aitken
Scientific) for providing a user-friendly and intuitive software
front-end, and Steven P. Keeling (GSK) for providing the
batch route to compound C.
Received for review March 26, 2007.
OP7000707
710
•
Vol. 11, No. 4, 2007 / Organic Process Research & Development