Combination of a customized robotic system with a TLC scanner for high-throughput reaction screening

Article abstract of DOI:10.1021/op0341972

Combination of a new and highly efficient robotic system for high-throughput reaction screening with a thin-layer chromatography (TLC) scanner is described. The system consists of a parallel synthesis robot capable of performing up to several hundred reactions in parallel and a second robot for analytical sample workup, dilution, and TLC-spotting. The automatically prepared TLC plates are analyzed with a modern TLC scanner which gives rapidly semi-quantitative analytical results on the outcome of the single reactions. Several automated chemistry examples clearly demonstrate the high efficiency of this combined system and its superiority compared to more classical approaches, e.g., HPLC analyses.

Full text of DOI:10.1021/op0341972

Organic Process Research & Development 2004, 8, 482−487
Combination of a Customized Robotic System with a TLC Scanner for
High-Throughput Reaction Screening
Christian Dinter,* Hilmar Weinmann,* Claudia Merten, Armin Schu¨tz, Thorsten Blume, Michael Sander,
Michael Harre, and Harribert Neh
Schering AG, Process Research, Automated Process Optimization, D-13342 Berlin, Germany
Abstract:
An additional problem is the rapid analytical characteriza-
Combination of a new and highly efficient robotic system for
high-throughput reaction screening with a thin-layer chroma-
tography (TLC) scanner is described. The system consists of a
parallel synthesis robot capable of performing up to several
hundred reactions in parallel and a second robot for analytical
sample workup, dilution, and TLC-spotting. The automatically
prepared TLC plates are analyzed with a modern TLC scanner
which gives rapidly semi-quantitative analytical results on the
outcome of the single reactions. Several automated chemistry
examples clearly demonstrate the high efficiency of this com-
bined system and its superiority compared to more classical
approaches, e.g., HPLC analyses.
tion of a high number of reactions, e.g. in catalytic reaction
screening, where a lot of different parameters such as catalyst,
ligands, additives, solvents, temperatures, etc. have to be
tested. By using traditional analytical testing methods, such
as, e.g., HPLC, relatively long overall analyses times have
to be accepted even if short columns and gradient methods
are used. Therefore, this creates a new bottleneck in the
overall process of rapid reaction screening. To overcome this
problem several approaches using color or fluorescence
assays for reaction high-throughput screening were published
recently.4-6 However, these assay systems usually give more
or less qualitative hints on the reaction outcome and are
therefore not always sufficient for reaction screening in an
industrial chemical process R & D. To fill this gap we
developed a new, rapid and reliable method for determination
of reaction outcomes based on automated spotting of
analytical samples on thin-layer chromatography (TLC)
plates in combination with a modern TLC scanner.
Introduction
The way in which chemical process research and devel-
opment is performed has been changing dramatically over
the past decade. This is due to the high pressure on chemical
development departments arising from a steadily increasing
output of new candidates from drug discovery units as well
as increased competition on the market which makes the
rapid development of innovative drugs crucial for economic
success. To cope with these challenges many pharmaceutical
companies heavily invested in robotic systems to accelerate
process R & D. Most of these early systems were focusing
on reaction optimization,1-3 and a lot of successful examples
of automated reaction optimization have been published in
the past few years.
However, in the early stages of process research when a
new drug candidate has just been handed over from the
Medicinal Chemistry department, a lot of experiments have
to be performed to find the best starting points for a further
and more detailed optimization. Route scouting for improved
alternative synthetic approaches requires a high number of
reactions to be performed rapidly. A drawback in this early
development phase is that usually only very limited quantities
of starting materials and core intermediates are available.
This means that the screening experiments have to be
performed using very small volumes; however, the results
have to be reliable and reproducible also on a larger scale.
Results and Discussion
After the successful establishment of a first automated
system for process optimization (Bohdan Process Develop-
ment Workstation and Sample Preparation Workstation)7
within our Automated Process Optimization group we set
out to expand the concept of laboratory automation to small-
scale reaction screening, route scouting, and polymorphism
studies.
On the basis of our specifications we made the decision
to establish for these purposes a custom-tailored system of
Zinsser Analytic.8 The whole robotics system consists of
three parts (see Figures 1 and 2).
The first robot (CALLI) is responsible for the preparation
of the individual reaction vials and the dispensing and
weighing of solid starting materials. It is capable of dosing
solids into reaction vials of different sizes (1-25 mL) with
two variable solid-dispensing pipets “VARIX” (see Figure
3). Six different reagents and starting materials can be kept
under inert atmosphere in powder storage containers in an
argon box. The reaction vials are also kept in an argon box.
Solids can be dosed in amounts between 1 mg and 2 g and
* To whom correspondence should be addressed. E-mail: Christian.Dinter@
Schering.de and Hilmar.Weinmann@Schering.de.
(4) Stambuli, J. P.; Hartwig, J. F. Current Opinion in Chemical Biology 2003,
7, 420-426.
(5) Hartwig, J. F.; Pawlas, J.; Nakao, Y.; Kawatsura, M. J. Am. Chem. Soc.
2002, 124, 3669-3679.
(1) Harre, M.; Tilstam, U.; Weinmann, H. Org. Process Res. DeV. 1999, 3,
304-318.
(2) Owen, M.; DeWitt, S. Laboratory Automation in Chemical Development.
Process Chemistry in the Pharmaceutical Industry; Gadamasetti, K. G., Ed.;
Marcel Dekker Inc.: New York, 1999.
(6) Hartwig, J. F.; Kawatsura, M. Organometallics 2001, 20, 1960-1964.
(7) Weinmann, H.; Schulz, C.; Wessa, T.; Harre, M.; Tilstam, U.; Neh, H. Org.
Process Res. DeV. 2001, 5, 335-339.
(3) Okamoto H.; Deuchi, K. Lab. Rob. Autom. 2000, 12, 2-12.
(8) See http://www.zinsser-analytic.com.
482
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development
10.1021/op0341972 CCC: $27.50 © 2004 American Chemical Society
Published on Web 04/22/2004

Figure 1. Overview of the high-throughput robotic system.
Figure 2. CALLI, LISSY, and SOPHAS.
accuracy of approximately (4%. Especially the bigger errors
were obtained by handling very small amounts of solvents
(about 30 µL) such as dichloromethane. Two slurry cannulas
are available as gripper tools. As a new tool on SOPHAS
an additional ceramic needle was integrated for handling
corrosive reagents such as POCl3, SOCl2, or SO2Cl2. Looking
for an easy evaluation reaction for the suitability of the
ceramic needle we tested the formation of benzoic acid
chloride using a solution of SOCl2 in ethyl acetate as well
as neat SOCl2. In both cases the benzoic acid (dissolved in
ethyl acetate) was transformed quantitatively into the acid
chloride using 2 equiv of SOCl2 without any damage to the
ceramic needle. A quick analytical TLC showed no starting
material left.
Solid dispensing can be achieved on the SOPHAS robot
in a way similar to that for the CALLI robot by using solid-
dispensing pipets. Independent addition of two reactive gases
to the reaction vials is possible. Agitation is done by shaking
the reactor blocks which is also possible under complete inert
atmosphere (nitrogen or argon). Reaction products can be
isolated on a filtration unit by applying either vacuum
filtration or positive pressure on the individual filters either
on a 24-filter cartridge rack or a 96-well filter plate. Solvents
can be changed during longer synthesis sequences by
evaporation on a custom-tailored evaporation unit (see Figure
4 ). Finally analytical samples can be automatically trans-
ferred to the third robot LISSY via a shuttle system.
The third robot LISSY is responsible for sample workup
(quenching and dilution) and reformatting into various
analytical sample vials for HPLC or HPLC-MS systems.
This robot takes over the sample racks from the shuttle to
SOPHAS and can store them in a stacker for three to five
sample racks. On top of this, LISSY is also capable of
spotting 1-µL samples automatically in a parallel fashion on
TLC plates (see Figure 5). Direct spraying with detection
Figure 3. Solid dispensing with CALLI on the balance.
the exact weight that was transferred to the individual vial
is determined by weighing on an integrated balance. De-
pending on the physical properties of the solid reactants, the
accuracy is about (5%. Solvents can be added automatically
by using two cannulas. The amount of solvent is modified
to the actual amount of dispensed solid by software calcula-
tion.
On the synthesis robot SOPHAS it is possible to handle
up to seven reaction blocks with reaction volumes varying
between 1 mL (96-well format) and 20 mL (8-well format).
Reactions can be performed in a temperature range between
-50 °C (limited by the cooling unit) and +140 °C. Extensive
temperature studies revealed an interaction of the block
format and the real internal solution temperature. For each
block and vial format we measured external and internal
temperature and detected an overall deviation of about 3-5
°C. Within one reaction block the temperature distribution
is quite homogeneous, varying around (1 °C. For rapid
dispensing of solvents and liquid reagents four parallel
stainless steel three-channel-cannulas are used. Depending
on the type of solvent, our liquid handling parameters were
developed to add the desired solvents and solutions with an
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
483

Figure 4. Evaporation unit and reaction-block handling on SOPHAS.
Scheme 1. Fast Optimization of Aryl Ether Synthesis 3
and separation. With the use of the experimental procedures
from the Medicinal Chemistry protocol, the optimization of
TLC condition succeeded very fast. After obtaining the best
TLC scanning conditions the reaction was screened on the
SOPHAS robot. In a single screening run we varied four
different parameters: temperature, base, solvent, and addi-
tive.
Figure 5. Automated sample spotting on TLC plates with
LISSY.
Two reaction blocks (each with 24 glass reaction vials of
4-mL volume) were filled with 1.0 equiv of phenol 1 (about
300 mg) and 1.1 equiv of bromoethylester 2. After adding
different solvents (acetone, toluene, acetonitrile, DMF,
butanone, ethanol), bases (Cs2CO3, K2CO3, KOtBu, NaOEt
all added as solids), and additives (different iodides to
promote in situ Finkelstein reaction) one reactor block was
shaken at ambient temperature and another one at 55 °C.
Our reaction array was a selection of 48, taking solubilty of
substrate and reagents into account. Aliquots of the reaction
vials were taken after 2, 6, and 16 h, quenched, diluted, and
delivered to our automated TLC-spotting system LISSY via
the shuttle system. Preparation of the TLC plates with the
optimal mobile-phase system and subsequent analysis of the
plates in the TLC-scanning apparatus showed directly
qualitatively and semi-quantitatively the reaction outcomes.
Figure 7 shows two different reaction conditions of the
screening. It is fairly easy to notice that condition B is
superior to condition A. Best conditions for this transforma-
tion in this first screening were 1.0 equiv of phenol, 1.1 equiv
of K2CO3, 0.1 equiv of NaI in acetone. Relative area
integration of the sample TLC revealed a conversion of more
than 87%. The experimental results were transferred back
to the laboratory and were reproduced sucessfully on a 10-g
scale (85% isolated yield) using classical mechanical stirring
for optimal agitation. The whole reaction screening and
analysis sequence (48 reactions, 154 TLC lanes) took only
2 days plus 1 day for reaction preparation and planning
processes.
agents is possible and can be used to easily detect certain
reaction products by color assays.
Thin-layer chromatography as an analytical method for
high-throughput determination of reaction outcome is cur-
rently a hot topic in industrial and academic synthesis
laboratories. After we had performed the studies described
in this article, two other papers appeared which reported
about the use of high performance thin-layer chromatography
for the assessment of the quality of combinatorial libraries9
and for the detection of unexpected reactions in organome-
tallic combinatorial catalysis.10 For our initial studies we used
the commercially available CAMAG TLC-SCANNER 3
which allowed measurements via adsorption and fluorescence
in a wavelength area from 190 to 800 nm (see Figure 6).11
The scanner was integrated in the sample preparation
workflow of the TLC spotting robot LISSY.
Our first investigation was focused on the synthesis of
an alkoxy-benzimidazole derivative 3 (Scheme 1) which was
used as an early intermediate in the synthesis of a potential
development candidate.12 First, the analytical TLC assay was
developed by screening wavelength and mobile solvent phase
(also mixed-solvent systems) for optimal product detection
(9) Salo, P. K.; Pertovaara, A. M.; Salo, V.-M. A.; Salomies, H. E. M.;
Kostiainen, R. K. J. Comb. Chem. 2003, 5, 223-232.
(10) Lavastre, O.; Touzani, R.; Garbacia, S. AdV. Synth. Catal. 2003, 345, 974-
977.
(11) See http://www.camag.com.
(12) For related compounds, see: Ezquerra, J.; Lamas, C. Tetrahedron 1997,
37, 12755-12764.
484
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Figure 6. CAMAG TLC-Scanner 3 and a typical HPTLC plate (20 spots per 10 cm × 10 cm developed plate, spotted by LISSY).
Figure 7. Time-conversion plots of two representative reaction conditions for the aryl ether synthesis 3.
Scheme 2. Reaction Screening for Optimization of
∆15-Estrone Methyl Ether-15-ene 6
6 in low yield. We set up a broad reaction screening by
varying solvents (DMSO, dichloromethane), temperatures
(room temperature, 40 and 70 °C), type and equivalents of
N-oxide (MPO, NMO, and TEMPO), type of base (seven
organic bases, three different buffered waters, pH 6.5-7.5,
inorganic salts such as NaHCO3, NaCO3), and amount of
additional base. The amount of IBX was dispensed with the
CALLI robot, and the reproducibility was very good (nearly
all samples within a range of (2 mg). More than 100 single
experiments in five robotic runs were performed on SOPHAS,
taking just a fraction of the full array into account. Due to
the number of experiments and analytical samples (four
samples per reaction) we used the SOPHAS robotic system
combined with the TLC scanner for fast reaction analysis.
Additionally, we measured all samples again by traditional
HPLC analysis and compared the obained results of both
analytical methods.15 Figure 8 shows the two independent
analytics (TLC and HPLC) for the same reaction conditions.
The correlation is very good.
Another detailed analysis of the TLC-scanning methodol-
ogy was performed on the oxidation of estrone methyl ether
4 to ∆15-estrone methyl ether 6 (Scheme 2). In this case
emphasis was put on the comparison of the analytical TLC
results with classical HPLC analysis. The classical prepara-
tion of enone 6 involved four chemical steps including
ketalization, bromination, dehydrobromination, and deketali-
zation chemistry. Due to this lengthy synthesis we were
attracted by a recent report from K. C. Nicolaou et al. who
had developed a mild and selective method for this type of
transformation.13
Good conditions for this reaction were found: 1.0 equiv
of silyl enol ether 5, 1.5 equiv of IBX, and 1.5 equiv of
MPO 1.0 equiv of Hu¨nigs base in a solvent mixture of
methylene chloride and DMSO at room temperature gave
about 80% conversion. Methylene chloride proved to be
necessary for solubility of the starting material, thus increas-
Oxidation of silyl enol ether 5 by using IBX‚MPO14 in a
fast proof-of-concept experiment yielded the desired enone
(15) For correct monitoring of the reaction products by TLC due to different
scanning wavelengths between TLC and HPLC the absolute area values
had to be corrected. The silyl enol ether 5 hydrolyzes on silica plates, thus
yielding the ketone 4. For easier comparison of TLC and HPLC results we
took the sum of these two compounds, due to the very similar responding
factors in HPLC.
(13) Nicolaou, K. C.; Gray, D. L. F.; Montagnon, T.; Harrison, S. T. Angew.
Chem. 2002, 114, 1038-1042; Angew. Chem., Int. Ed. 2002, 41, 996-
1000.
(14) IBX ) o-iodoxybenzoic acid; MPO ) p-methoxy pyridine N-oxide
(hydrate); A 1:1 complex is formed in situ.
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
485

Figure 8. Correlation of TLC and HPLC chromatograms of the same reaction sample.15
Table 1. Some representative reaction conversions of the
alkylation (8 of 48) after 20 h
ing yield and decreasing reaction temperature. Overall, the
TLC scanner results show the same trends of each reaction
as does the corresponding HPLC analysis. The major
advantage of TLC is the enormous reduction of time for the
analytical data collection. In this study we analyzed more
than 420 reaction samples via TLC and HPLC. TLC was
about 15-20-fold faster than HPLC (TLC: 8 h versus
HPLC: 140 h) giving the same qualitative and semi-
quantitative information. Additionally, the consumption of
pure reagent grade solvents for the analytics was dramatically
reduced, resulting in cost savings and a lower amount of
waste solvent.
NaI
T
conversion
[%]
reaction
solvent
DMF
base
additive [°C]
1
2
3
4
5
6
7
8
Cs2CO3
Cs2CO3
K2CO3
K2CO3
-
+
-
+
-
+
-
+
20
20
55
55
55
55
55
55
51
61
75
83
77
84
73
87
DMF
DMF
DMF
acetonitrile Cs2CO3
acetonitrile Cs2CO3
acetone
acetone
K2CO3
K2CO3
Currently, the implementation of an automated enzyme
screening methodology and the integration of catalyst librar-
ies for transition metal catalysis are ongoing.16
acetonitrile, acetone, butanone, ethanol) and 1.1 equiv of
bromethylester (neat) were added, using the four stainless
steel needles. The reaction block was shaken with a speed
of 650 rpm at room temperature or at 55 °C. Samples were
taken out of the reaction mixture after 2, 8, and 20 h and
quenched with ethyl acetate (containing 1% of AcOH).
Finally, the samples were diluted to a concentration of 1 mg/
mL with the LISSY robot, and 2 µL were spotted on glas
TLC plates (Merck HPTLC plates 10 cm × 10 cm, silica
gel 60 F254). The development of the plates was performed
with a mixture of hexane/ethanol (80:20 vol/vol), and the
plates were scanned with the TLC Scanner at a wavelength
of 450 nm (tungsten lamp) (see Table 1).
General Procedure for Oxidation of Estrone Methyl
Ether 4. Each vial of a 24-vial reflux reaction block was
filled with 1.5 equiv of solid IBX with the CALLI robot.
Then, the addition of the 1.5 equiv of N-oxide (MPO, NMO,
and TEMPO) and 1.0 equiv of silyl enol ether 5 was
performed manually. One milliliter of DMSO (or DMSO/
dichloromethane as 5:2 mixture) was added with the four
stainless steel needles an the reaction was shaken at 75 °C
or room temperature. Reaction samples were taken after 1,
4, 10, and 20 h, hydrolyzed with saturated NaHCO3 solution,
and extracted with dichloromethane. A sample of the organic
layer was diluted with acetonitrile to a concentration of 0.5
mg/mL for HPLC (column: Kromasil C8, 5 µm; mobile
phase: acetonitrile/water, gradient: 60-100% acetonitrile
in 15 min; flow rate: 1 mL/min; UV detector 220 nm;
retention time of 5 ) 12.70 min, of 4 ) 3.23 min) and
analytical TLC. One microliter of the samples was spotted
on glass TLC plates (Merck HPTLC plates 10 cm × 10 cm,
silica gel 60 F254) with the LISSY robot. The development
of the plates was performed with a mixture of hexane/ethanol
Conclusions
A new and highly efficient robotic system for high-
throughput reaction screening was established. Automated
TLC preparation in combination with a modern TLC scanner
was developed as a powerful tool for high-throughput semi-
quantitative determination of reaction outcome. Several
automated chemistry examples clearly demonstrate the high
efficiency of this combined system and its superiority
compared to more classical approaches, e.g., HPLC analyses.
The new high-throughput TLC method is thus far only
limited by the accuracy of the spotting of small sample
amounts to the TLC plate and by the detection limit of the
scanner. Further miniaturization holds a great potential for
even higher sample throughput with this analytical method.
Experimental Section
Solvents and reagents were used as received from
commercial and internal suppliers. HPLC was performed on
a DIONEX system (P580 pump, Gina 50 Autosampler, and
HP1100 UV detector). All analytical TLC plates were
scanned on a CAMAG TLC Scanner 3.
General Procedure for Alkylation of Alkoxy-benzim-
idazole Derivative 3. Each vial of a 24-vial reflux reaction
block was filled with 1.0 equiv of starting material and 1.1
equiv of solid base (Cs2CO3, K2CO3, KOtBu, NaOEt) on the
SOPHAS robot. Additionally, 0.1 equiv of NaI was added
to reaction vials 13-24. Then, 700 µL of solvent (DMF,
(16) A recent publication on automated enzyme-screening methodology: Yazbeck,
D. R.; Tao, J.; Martinez, C. A.; Kline, B. J.; Hu, S. AdV. Synth. Catal.
2003, 345, 524-532.
486
•
Vol. 8, No. 3, 2004 / Organic Process Research & Development

Table 2. Some represenative reaction conversions (8 of 120) of the IBX oxidation after 20 hours
base additive
N-oxide
[1.5 equiv]
[1.5 equiv]
T
[°C]
conversion
HPLC [%]
conversion
TLC [%]
reaction
solvent
DMSO
DMSO
DMSO
DMSO
DMSO
DMSO/CH2Cl2
DMSO/ CH2Cl2
DMSO/ CH2Cl2
(more than 10 screened)
1
2
3
4
5
6
7
8
-
-
-
-
-
75
75
75
75
75
20
20
20
42
73
50
54
81
69
71
82
43
72
49
54
80
69
70
81
MPO
NMO
TEMPO
MPO
MPO
MPO
MPO
NEt3
NEt3
Hunigs base
Hunigs base
(80/20 vol/vol), and the plates were scanned with the TLC
scanner at a wavelength of 200 nm (deuterium lamp) (see
Table 2). See also reference 15.
very thankful to E. Adam and J. Krueger for excellent
technical assistance.
Received for review December 22, 2003.
OP0341972
Acknowledgment
We thank T. Wessa, K. Lovis, S. Sokolowsky (all
Schering AG) for very stimulating discussions. We are also
Vol. 8, No. 3, 2004 / Organic Process Research & Development
•
487