High-Throughput Assay for Enantiomeric Excess Determination in 1,2- and 1,3-Diols and Direct Asymmetric Reaction Screening

Article abstract of DOI:10.1002/chem.201701923

A simple and efficient method for determination of the yield, enantiomeric/diasteriomeric excess (ee/de), and absolute configuration of crude chiral diols without the need of work-up and product isolation in a high throughput setting is described. This ap

Full text of DOI:10.1002/chem.201701923

A Journal of
Accepted Article
Title: High-Throughput Assay for Enantiomeric Excess Determination
in 1,2- and 1,3-Diols and Direct Asymmetric Reaction Screening
Authors: Elena G. Shcherbakova, Valentina Brega, Vincent M. Lynch,
Tony D. James, and Pavel Anzenbacher
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201701923
Link to VoR: http://dx.doi.org/10.1002/chem.201701923
Supported by

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
High-Throughput Assay for Enantiomeric Excess Determination
in 1,2- and 1,3-Diols and Direct Asymmetric Reaction Screening
Elena G. Shcherbakova[a], Valentina Brega[a], Vincent M. Lynch[b], Tony D. James[c], and Pavel
Anzenbacher, Jr. *[a]
Abstract: A simple and efficient method for determination of the yield,
enantiomeric/diasteriomeric excess (ee/de), and absolute
Currently, the most common methods used for the determination
of enantiomeric purity involve and
1H- 19F-NMR
configuration of crude chiral diols without the need of work-up and
product isolation in a high throughput setting is described. This
approach utilizes a self-assembled iminoboronate ester formed as a
product by dynamic covalent self-assembly of a chiral diol with an
enantiopure fluorescent amine such as tryptophan methyl ester or
tryptophanol and 2-formylphenylboronic acid. The resulting
diastereomeric boronates display different photophysical properties
and allow for fluorescence-based ee determination of molecules
containing a 1,2- or 1,3-diol moiety. This method has been utilized for
the screening of ee in a number of chiral diols including atorvastatin,
a statin used for the treatment of hypercholesterolemia. Noyori
asymmetric hydrogenation of benzil was performed in a highly parallel
fashion with errors <1% ee confirming the feasibility of the systematic
examination of crude products from the parallel asymmetric synthesis
in real time and in a high-throughput screening (HTS) fashion.
spectroscopy,[13],[14],[15],[16] circular dichroism (CD),[17],[18],[19],[20]
chiral-phase high performance liquid chromatography (HPLC),
HPLC coupled with circular dichroism (HPLC-CD), or chiral gas
chromatography (GC).[21] Such techniques are, however, better
suited for serial analyses while the need for the simple parallel
platforms remains largely unmet. For this reason, the HTS
methods allowing for the use of simple commercial instrument
platforms in conjunction with multi-well formats are sought.[18]
We have recently reported the first fluorescence-based assay for
the rapid parallel determination of the ee value of amines and
amine derivatives utilizing the dynamic covalent self-assembly
approach.[22],[23] Our new approach for the determination of the ee
of chiral diols (Figure 1 bottom) is based on the dynamic covalent
self-assembly[24],[25],[26],[27] of an enantiomerically pure fluorescent
amine reporter with 2-formylphenylboronic acid (FPBA) and a
molecule containing a 1,2- or 1,3-diol moiety (Figure 1 top).
Introduction
Chiral diols represent essential structural motifs found in
pharmaceuticals,[1] biologically active natural products,[2] synthetic
precursors and intermediates,[3] or chiral catalysts/auxiliaries
used in
a
number of asymmetric transformations.[4],[5],[6]
Asymmetric dihydroxylation of alkenes,[7],[8] asymmetric
hydrogenation of ketones,[9],[10] and dicarbonyl reduction by single
enzyme[11] are among the most advantageous synthetic
procedures for the preparation of chiral diols. Recent advances in
the development of methods for the preparation of 1,2- and 1,3-
diols necessitate the development of high-throughput screening
techniques for the fast determination of enantiomeric excess (ee),
which is essential for the optimization of chiral catalysts and
libraries of auxiliaries for asymmetric transformations where the
ability to analyse a large number of samples in parallel fashion is
highly desirable.[12]
[a]
E. G. Shcherbakova, Dr. V. Brega, Prof., Dr. P. Anzenbacher, Jr
Department of Chemistry
Figure 1. Concept of dynamic covalent self-assembly for enantiomeric excess
determination in chiral diols and asymmetric reaction characterization. Top: 1,2-
or 1,3-Diols self-assemble with an enantiomerically pure fluorescent amine
reporter and 2-formylphenylboronic acid. Bottom: The diol analytes/substrates
and fluorescent amines used in this study.
Bowling Green State University
Bowling Green, OH, USA 43403
E-mail: pavel@bgsu.edu
Dr. V. M. Lynch
Department of Chemistry
University of Texas at Austin
Austin, TX 78712, USA
Prof., Dr. T. D. James
[b]
[c]
Department of Chemistry, University of Bath
Claverton Down, Bath BA2 7AY, UK
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
Results and Discussion
The successful formation of the iminoboronate ester/oxazolidine
Table 1. Affinity constants Ka (× 106 M−1) a corresponding to L-tryptophan
methyl ester (L-TrpOMe) and L-tryptophanol (L-TrpOH) based complexes
complexes
was
confirmed
by
matrix-assisted
laser
with selected chiral diols[a]
.
desorption/ionization (MALDI) mass spectrometry (see the
Supplemental Information), single-crystal X-ray diffraction
analysis, and fluorescence spectroscopy. As expected, the
reaction between L-tryptophanol (1,2-amino alcohol), FPBA, and
a diol yields an oxazolidine boronate ester.[26] Figure 2 shows the
X-ray crystal structures of the two diastereomeric oxazolidine
Ka (× 106 M-1)
L-TrpOMe
L-TrpOH
0.64
4.33
4.72
3.15
4.07
4.50
7.71
5.15
2.31
11.0
8.40
3.37
0.40
2.60
3.34
(S)-α-Glycerol chlorohydrin
(R)-α-Glycerol chlorohydrin
(S)-(+)-Phenylethylene glycol
(R)-(−)-Phenylethylene glycol
(+)-Dimethyl L-tartrate
1.65
0.47
0.47
1.39
1.04
0.96
1.10
0.36
0.20
0.32
1.83
0.45
0.49
0.38
2.34
(‒)-Dimethyl D-tartrate
(+)-Diethyl L-tartrate
(‒)-Diethyl D-tartrate
(1S,2S,3R,5S)-(+)-Pinanediol
(1R,2R,3S,5R)-(‒)-Pinanediol
(S,S)-(‒)-Hydrobenzoin
meso-Hydrobenzoin
(R,R)-(+)-Hydrobenzoin
(3S,5S)-Atorvastatin
(3R,5R)-Atorvastatin
[a] All titrations were performed in acetonitrile/water (5% v/v). The Kas were
calculated based on the change in fluorescence intensity upon the addition
of each chiral diol. The Kas were calculated using non-linear least-square
fitting, errors of the fitting were < 20%.
boronate esters self-assembled from L-tryptophanol, FPBA and
(R,R)- or meso-hydrobenzoin in a 1:1:1 stoichiometry.
Figure 2. Crystal structures of oxazolidine boronate esters self-assembled from
FPBA, L-tryptophanol, and (R,R)-HBZ (left), and meso-HBZ (right).
Figure
3 shows the fluorometric titration profiles of the
The choice of tryptophan derivatives was motivated by the
potential for bright fluorescence. The variation of the
spectroscopic properties originate from different geometrical
arrangement around the boron center to induce an intramolecular
collisional quenching. Due to a strong enantiomer-induced
fluorescence signal (amplification or quenching), fluorometric
titrations enable a quantitative estimation of chiral enrichment in
a diverse set of diols (Figure 1 bottom). As chiral fluorescent
reporter moieties, we selected enantiomerically pure L-tryptophan
methyl ester and L-tryptophanol (Table 1).
enantiomeric pair of diethyl tartrate upon the formation of the
oxazolidine boronate ester with L-tryptophanol. The formation of
diasteriomeric complexes exhibit high association constants with
both enantiomers of diethyl tartrate (Figure 3, and Table 1), but
(+)-diethyl-L-tartrate displays ca. 20-fold stronger fluorescence
amplification. The graph insets show the nonlinear least-square
fit of the data using a 1:1 binding model. The respective
association constants were (Ka, M−1) Ka(L) = 7.7 × 106 M−1 and Ka(D)
= 5.2 × 106 M−1 for (+)-diethyl-L-tartrate and (-)-diethyl-D-tartrate,
respectively.
Figure 3. Fluorescence titrations of L-tryptophanol - FPBA (1:1, 40 μM) with diethyl-tartrate enantiomers (0-100 μM) in MeCN. λex = 280 nm. Left: (+)-diethyl-L-
tartrate; Center: (–)-diethyl-D-tartrate; Right: Isotherms show the difference in signal amplification for the two diethyl tartrate esters (λem = 330 nm)
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
As another example of the utility of the present method we show
the ee analysis of the statin-type blockbuster drug atorvastatin
used for the treatment of hypercholesterolemia[28],[29] and the
prevention of cardiovascular diseases.[30],[31]
Atorvastatin exists in four stereoisomers, but commercialized is
only the enantiopure (3R,5R)-7-[2-(4-fluorophenyl)-3-phenyl-4-
(phenylcarbamoyl)-5-propan-2-ylpyrrol-1-yl]-3,5-
dihydroxyheptanoic acid form.[11],[32] Figure 4 bottom shows that
addition of (3S,5S)-atorvastatin, the inactive enantiomer,[33],[34] to
the FPBA/L-tryptophanol mixture induces an intensity
enhancement of L-tryptophanol florescence. Conversely, the
fluorescence of the L-tryptophanol is quenched upon the addition
of the biologically active (3R,5R)-atorvastatin (Figure 4 top).
such as supervised method linear discriminant analysis (LDA)
were used.[35],[36],[37] LDA reduces the dimensionality of the data
used to identify the samples and separate the data set into groups
of clusters based on similarities in the response data, and to help
identify possible patterns and functional relationship between the
clusters. The leave-one-out cross-validation protocol confirmed
the ability of the tryptophan methyl ester/tryptophanol based
fluorescent complexes to differentiate between 7 enantiomeric
pairs of chiral diols with 100% correct classification for 320 data
points (16 by 20 repetitions).
Figure 5 illustrates the graphical output of LDA where we see a
tight grouping of the data points within the individual clusters and
a clear segregation of the clusters, which corresponds to large
differences in the fluorescence of the various diols and their
enantiomers within the large response space defined by the first
three canonical factors (F1-F3) of LDA.
The qualitative analysis above confirms that the covalent self-
assemblies can be used to recognize a variety of enantiomeric
diols using their fluorescence and should allow for the quantitative
determination of their ee.
An important feature observed in the qualitative analysis (Figure
5 left) of the chiral diols is the distance between the enantiomers
that reflects the difference in the fluorescence behavior of the
diastereomeric assemblies of the diols. In fact, the larger the
distance, the lager the difference between the behavior of the
enantiomers. Particularly interesting is the example of
hydrobenzoin, for which we tested the S,S-, R,R and meso-form.
The clusters corresponding to the three forms are well separated
suggesting that a quantitative analysis will be possible.
Next, the loading test of the sensing system was performed using
five different diols at various levels of ee (0-100%) (Figure 5 right).
In the simultaneous high-throughput experiment the 47 data-
points with 24 repetitions in each cluster (1128 individual
measurements) were dispensed and recorded in 1536 well plate
within minutes. The LDA shows 100% correct classification of all
clusters and clear ee-dependent trends for all the five diols.
Interestingly, the S,S-diethyl- and S,S-dimethyl tartrate show a
high similarity at ee < 70% (at ee > 70% S,S-diethyl- and S,S-
dimethyl tartrate show significantly different response). This is not
unreasonable considering the high structural similarity of these
compounds.
Figure 4. Fluorescence titrations of L-tryptophanol - FPBA (1:1, 40 μM) with
atorvastatin enantiomers. Top: stepwise addition of (3R,5R)-atorvastatin (0-100
μM) in MeCN. λex = 280 nm. Bottom: addition of (3S,5S)-atorvastatin (0-80 μM)
in MeCN. λex = 280 nm.
Following in this vein we also performed a quantitative analysis of
the ee with mixtures of atorvastatin enantiomers. First, we used
LDA to confirm a discriminatory potential of the FPBA-amine
assemblies to distinguish various ee values (demonstrated by
100% classification and superior separation of the data clusters).
The regression analysis using the support vector machine
(SVM)[38] algorithm enables deconvolution of the multidimensional
data to the actual ee values of (3R,5R)-atorvastatin and leave-
one-out cross-validation approach validates the predictive
potential of the developed model (see Supplemental Information).
The plot of the predicted versus actual ee value for (3R,5R)-
atorvastatin shows that the developed SVM model allowed
simultaneous correct identification of two validation samples. The
low prediction error (<2% ee) confirms the ability of the self-
assembled system to correctly signal the levels of enantiomeric
excess in complex molecules.
Further tests with various enantiomeric diols show that different
diols also either quench or amplify the fluorescence albeit to
different degrees. The complementarity in association affinities of
the complexes, together with the analyte-specific signal
amplification/quenching output, indicates that the combination of
enantiopure tryptophan methyl ester/tryptophanol based
complexes will be suited for recognition and identification using
fluorescence-based HTS methods.
First, a qualitative analysis was performed to confirm the ability of
the assemblies to differentiate between structurally similar
enantiomeric pairs of chiral diols. The fluorescence intensities
were recorded using a conventional microplate reader in 384-well
plate (16 data clusters/20 repetitions). The response patterns
associated with the assemblies were obtained in the form of
multivariate data sets. Hence, pattern recognition techniques
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
Figure 5. Graphical 3D outputs of LDA. Left: the results of qualitative LDA of various chiral diols. LDA output shows a perfect separation of the enantiomeric pairs
and an excellent recognition capability with 100% correct classification of all 16 data-points corresponding to the individual analytes. The response space is defined
by the first three factors (F1-F3) of LDA comprising 99.9% of total variance. Right: he result of the semi-quantitative LDA for simultaneous determination of ee of
five diols. Clear ee-dependent trends in individual responses to each diol were identified. The response space is defined by the first three factors (F1-F3) of LDA
comprising 95.2% of total variance.
The SVM confirms that the of L-tryptophanol based oxazolidine
iminoboronate esters yield a highly sensitive and linear response
that can be used to quantify the concentration, ee and absolute
configuration of atorvastatin in a sample.
However, high throughput settings may require simple analysis
method without using sophisticated statistical methods. Simplest
method of obtaining absolute configuration and ee of reaction
mixtures from fluorescence intensity (FI) is by utilizing a standard
curve method. For our purpose, a standard curve is defined as a
graph of FI plotted on the Y axis, and various ee of standards
along the X axis. The resulting graph of FI vs ee obtained by linear
regression of points referring to FI readings from a number of
mixtures of known ee used for calibration is linear (Figure 6). The
absolute configuration and ee of controls (validation samples) and
unknowns (for example asymmetric reaction samples) can be
determined by interpolation of their fluorescence reading on the
graph.
Table 2. The results of Atorvastatin ee determination using linear regression.
Actual
Abs. config.
Fluorimetric Assay
Abs. config.
ee[a] (%)
R,R 1.40
Entry
1
2
3
4
5
6
7
8
ee(%)
3
R,R
R,R
R,R
R,R
R,R
R,R
R,R
R,R
S,S
S,S
S,S
S,S
S,S
S,S
10
20
30
40
60
85
95
25
45
60
65
70
95
R,R
R,R
R,R
R,R
R,R
R,R
R,R
S,S
S,S
S,S
S,S
S,S
S,S
13.2
23.0
31.5
42.0
58.7
84.6
95.8
25.8
44.9
57.1
65.2
72.6
96.3
9
10
11
12
13
14
[a] Average value from two independent trials with 20 repetition in each trial.
Figure 6. Standard curve obtained from L-tryptophanol-FPBA assembly (1:1,
40 μM) with atorvastatin enantiomers at various ee (black circle •). 14 samples
of unknown atorvastatin enantiomer excess (blue and red circles) were
simultaneously correctly analysed.
Here, the linearity of the fluorescence response to atorvastatin ee
suggests the estimation of the ee using standard curve will yield
an excellent linear correlation. The standard samples were
prepared using mixtures of the known ee of the atorvastatin
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
across a whole range of ee from pure (3R,5R)-atorvastatin to pure
(3S,5S)-atorvastatin. Figure 6 shows the results of quantitative
linear regression (R2=0.999) between fluorescence intensity and
ee of atorvastatin. This linear model was used to determine
simultaneously 14 samples of unknown ee with excellent
precision (Table 2).
and prediction respectively, with the root-mean-square errors
(RMSEs) of calibration (C), cross-validation (CV), and prediction
(P) below 1% ee.
Table 3. The results of benzil reduction.
1H NMR[a]
Abs. config.
HT Fluorimetric assay
Abs. config.
R,R
Asymmetric reaction screening.
Entry
2
2A
3
3A
4
4A
Ru-cat
S,S
S,S
R,R
R,R
ee(%)
97.5
≥99
99
≥99
4
ee(%)[b]
95.4
R,R
R,R
S,S
S,S
R,R
R,R
In general, there is a need to screen libraries of catalysts for their
ability to yield products with high enantioselectivity. This requires
time-efficient yet effective protocols for the systematic
examination of crude products of asymmetric reactions in real
time and in a HTS fashion. To illustrate the real-life utility of the
present method, we decided to demonstrate the possibility of
evaluating reaction mixtures of stereoselective reactions, both the
yield and ee in small amounts (<1 mg) in crude form (without
product crystallization) in high-throughput fashion.
R,R
99.5
S,S
98.3
S,S
99.7
S,S : R,R
S,S : R,R
R,R
3.75
3.7
R,R
2.87
[a] Determined by 1H NMR from multiple peaks integration. [b] Errors for
determination of ee < 1%.
As a benchmark for testing our method, we used the Noyori
asymmetric transfer hydrogenation of benzils catalyzed by
Next, the validated model was used to determine the ee in the six
entries of unknown ee obtained from the Noyori reaction mixtures
(Table 3). As expected, samples 2 and 2A (were prepared using
RuCl[(S,S)-Tsdpen](p-cymene) catalysts), as well as samples 3
and 3A (prepared using RuCl[(R,R)-Tsdpen](pcymene)) were
predicted as pure (R,R)-hydrobenzoin and (S,S)-hydrobenzoin
respectively, and their ee were determined to be 95.4% ee for the
crude product (entry 2), 99.5% ee for the recrystallized part (entry
2A), and 98.3% ee for the crude mixture (entry 3), 99.7% ee for
the recrystallized sample (entry 3A), which is in perfect agreement
with the 1H NMR[15] results and literature values.[9]
RuCl[(R,R)-Tsdpen](p-cymene)
or
RuCl[(S,S)-
Tsdpen](pcymene) catalysts with a formic acid/triethylamine
mixture, which affords S,S- and R,R-hydrobenzoin almost
quantitatively with excellent diastereomeric and enantiomeric
purity (Fig. 7).[9], [10]
Unknowns 4 and 4A were prepared using a 1:1 mixture of
RuCl[(S,S)-Tsdpen](p-cymene) catalysts and RuCl[(R,R)-
Tsdpen](p-cymene) catalysts, respectively, and gave a small
excess of (R,R)-hydrobenzoin. The corresponding ee were
predicted as 3.7% ee for the crude product (entry 4), and 2.8% ee
for the recrystallized part of reaction mixture (entry 4A). To
perform the ee assignment using the present method in 20
measurements per entry takes < 1 min. Each measurement
requires only nanograms of the benzil substrate. Perhaps most
importantly, the comparison of the results obtained for crude and
crystallized samples shows that our method is not sensitive to
traces of catalysts or reaction additives and may be applied to the
determination of ee in crude product mixtures.
We have also tested sensitivity of the assay to the presence of
meso-hydrobenzoin. It was found, that the method is not sensitive
up to 5% of meso-form (See SI). Above this value, a new
calibration set comprising meso-hydrobenzoin must be developed,
and consequently, method could be used for simultaneous
determination of ee and de.
Figure 7. Noyori asymmetric hydrogenation of benzil with ruthenium catalyst
yields hydrobenzoin in a high diastereomeric and enantiomeric purity.
Upon completion of the reaction, the products were divided into
two portions, crude (after solvent evaporation; entry 2, 3, and 4),
and a second portion was recrystallized from methanol (samples
2A; 3A; 4A). As a negative control, an incomplete reaction run
without the presence of competent catalyst (entry #1 as a mixture
of benzil, benzoin and inactive Ru-catalyst, not shown in Table 3)
was also performed. For each entry, the absolute configuration
1
and ee was determined using H NMR[15] and compared to the
results of the fluorimetric assay (Table 3).
As expected, the linearity of the fluorescence response to ee was
also observed for the hydrobenzoin in a standard calibration
procedure. This was first confirmed in the titration experiment
(Figure 8, Left) as well as in a high throughput assay (Figure 8,
Bottom). Both experiments provide almost identical dependence
(Table 4) suggesting the validity of the high-throughput and
standard curve approach.
Towards that end, we generated a calibration dataset using pure
enantiomers of hydrobenzoin. The resulting multivariate data set
was analyzed by SVM and 25 data points from the calibration set
were used to develop/calibrate the linear regression model. The
model was then validated using four independent samples (not
part of the calibration dataset) and their ee was determined. The
corresponding correlation graph (See SI) between the predicted
ee and the actual ee is characterized by a regression (R2) of
0.999842, 0.999818 and 0.999823 for calibration, cross-validation
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
Figure 8. Standard curves generated from hydrobenzoin samples with various enantiomeric composition (ee range from 100% (R,R)-hydrobenzoin to 100% (S,S)-
hydrobenzoin). Left: fluorescence titration profile of L-tryptophanol - FPBA (1:1, 40 μM) with hydrobenzoin standards. Inset: Standard curve obtained for FI vs known
ee used for calibration. Right: standard curve for FI readings from a number of mixtures of known ee used for calibration in HT fluorescence assay. Six unknown
asymmetric reaction samples (four red circles, two blue circles), their absolute configuration and ee were simultaneously correctly determined.
Conclusions
Table 4. The results of benzil reduction from standard curve linear fitting.
FI Titration [a]
In summary, we have demonstrated that the dynamic covalent
self-assembly of an enantiomerically pure fluorescent amine
reporter with 2-formylphenylboronic acid (FPBA) and a molecule
Entry
2
Ru-cat
S,S
Abs. config.
ee(%)[a] [c]
96.5
97.9
96.9
99.9
5.8
R,R
R,R
S,S
S,S
R,R
R,R
2A
3
S,S
R,R
containing
a 1,2- or 1,3-diol moiety form diastereomeric
3A
4
4A
R,R
S,S : R,R
S,S : R,R
complexes that display different fluorescence properties, an
output signal measured using standard fluorescence plate reader
in a high-throughput fashion. The chiral diol-induced changes in
fluorescence are highly sensitive to the interplay between
components of assembly, steric arrangement, and diol-receptor
affinity. More importantly, these spectroscopic signatures are
quantitatively related to the amount of analyte and absolute
configuration. The ability to determine ee was tested on a set of 6
diol enantiomeric pairs including meso-hydrobenzoin, and anti-
hypercholesterolemia drug atorvastatin. The qualitative study
using the pattern recognition procedure (LDA) allowed for
identification of all enantiomers tested. Likewise, the quantitative
analysis of ee of atorvastatin confirmed the ability to quantify drug
purity with an error in ee of <2% using a simple standard curve
linear regression protocol. The final proof for the utility of the
method was a high throughput determination of ee in reaction
mixtures and products of Noyori asymmetric reduction of benzil to
hydrobenzoin. Here, the tested (6 samples in 20 repetitions) and
independently validated system was able to determine absolute
configuration and ee, which was compared to the ee data
obtained from NMR measurements. The error of the
determination was < 1% ee. Finally, the artificial neural network
(ANN) was employed for a simultaneous determination of ee and
hydrobenzoin concentration. This experiment, was performed in
a high-throughput fashion. Each reaction requires ca 10-20
ng/well of the benzil substrate and product mixtures can be
analyzed in the same well. The present method enables
measurement of ee in crude and crystallized samples, confirming
that our method is not sensitive to traces of catalysts or reaction
9.2
[a] Average value from two independent readings. [b] Average value from two
independent trials with 20 repetition in each trial. [c] Errors for determination
of ee < 2%
Because the tryptophanol based self-assembly depends both on
the hydrobenzoin concentration and the change in ee value of the
respective enantiomers, both can be calculated using artificial
neural network (ANN).[39],[40] ANN utilizes a more complex and
expanded calibration set that covers different total concentration
values together with various ee in parallel with calibration set
prepared using different unknown concentrations (Noyori
asymmetric reaction outcomes). ANN was employed to calculate
the total hydrobenzoin concentration and ee of the hydrobenzoin
in Noyori asymmetric hydrogenation reaction mixtures (See
Supplemental Information). The resulting data (Table S21 of SI)
show that the known ee and concentrations are in perfect
agreement with the data calculated by the artificial neural network
(ANN) and that ANN using the fluorescence data obtained enable
simultaneous determination of ee and hydrobenzoin
concentration. This is, in part, due to the fact that benzil, as most
of the aromatic ketones is not fluorescent. Thus, all the
fluorescence in the reaction mixture originates from hydrobenzoin
stereoisomers. This in turn means that the present method allows
determination of the reaction yield even in reactions where the
starting material is not fully converted to products.
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
immediately read after incubation with a BMG PHERAstar Plus multi-mode
microplate reader.
additives. Together, the data presented here indicate that
enantiopure tryptophan methyl ester/tryptophanol based
assemblies are well suited for instant and accurate determination
of the absolute configuration of chiral diols. These features are
utilized in a simple, fast but also inexpensive way for monitoring
the ee and total yield of diol enantiomers using a microscale HTS
platform.
Data handling.
The response patterns associated with the sensor array are acquired in a
form of multivariate data set. Raw data were subjected to the Student’s t-
test to exclude 4 data points out of 24 repetitions in order to examine the
difference (standard deviation) between data within the set of repetitions.
The coefficient of variability among the data within the class of 20
repetitions did not exceed 4% and obtained data were then analyzed using
linear discriminant analysis (LDA) and artificial neural network (ANN)
without any further pretreatment. Support vector machine (SVM) coupled
with principal component analysis (PCA) and partial least squares (PLS)
methods were applied for quantitative regression analyses of multivariate
data sets. The training set was used to calibrate the model producing the
root mean square error of calibration (RMSEC). The model was validated
using 5-fold cross-validation producing the root mean square error of
cross-validation (RMSECV). The validity and predictive ability of the
developed model was then tested using independent data sets (validation
samples). The predictive accuracy of a model was evaluated by the value
of the root mean square error of prediction (RMSEP). The linear least
squares method was used for finding the coefficients of polynomial
equations that are a best linear fit to a set of X,Y data corresponding to ee
and FI of the standards and unknown samples. A polynomial equation
expresses the dependent variable Y (FI) as a first-order polynomial in the
independent variable X (ee) generate a straight line (Y=a+bX, where a is
the intercept and b is the slope), those coefficients (a, b) were used to
predict values of X for each measured Y. In all these cases, Y is a linear
function of the parameters a and b.
Experimental Section
Materials and methods
Starting materials, all reagents, and organic solvents were obtained from
commercial suppliers and used without further purification.
Fluorescence.
Steady-state fluorescence emission and excitation measurements were
recorded between 290 nm and 550 nm. Solution of L-tryptophanol and L-
tryptophan methyl ester complexes were excited at 280 nm in
spectrophotometric grade acetonitrile. The emission from probes was
scanned in
1 nm step with appropriate excitation and emission
monochromators band pass settings with dwell time 0.35 sec. under
ambient conditions. Titration isotherms were constructed from changes in
the fluorescence maximum at 330-335 nm.
Noyori asymmetric Ru catalyzed transfer hydrogenation of benzil.
Compounds 1, 2, 2A, 3, 3A, 4, 4A were prepared from the set of parallel
asymmetric reductions of benzil catalyzed by RuCl[(R,R)-Tsdpen](p-
cymene) or RuCl[(S,S)-Tsdpen](p-cymene) catalysts at 40°C with
Accession Numbers
The accession number for the meso-hydrobenzoin, oxazolidine
boronate ester complex reported in this paper is CCDC 1543343,
and for the (R,R)-hydrobenzoin, oxazolidine boronate ester
complex reported in this paper is CCDC 1543342.
substrate/catalyst molar ratio (S/C) of 1500 in
a DMF solution of
HCOOH/N(C2H5)3. Crude reaction products were divided in two parts, one
portion of each sample was used without any further purification (sample
1; 2; 3; 4), and the second portion was recrystallized from methanol at -
40°C (samples 2A; 3A; 4A).
Determination of association constants (Ka, M−1) by fluorescence
titrations.
Acknowledgements
P.A. acknowledges support from NSF (CHE-0750303 and DMR-
1006761)
Then each reaction mixture has to be diluted 100 times with the same
solvent to a final concentration of 40 μM and a final volume of 2 mL directly
in a square quartz cuvette with 10mm path length. Titration isotherms
(normalized fluorescence intensity vs concentration of diols) show the
nonlinear least-squares best fits of the data to 1:1 binding model and yield
association constants (Ka, M−1).
Keywords: Self-assembly • Fluorescence • Enantiomeric
excess • Diols • Asymmetric Catalysis
[1]
[2]
[3]
R. N. Patel, Coord. Chem. Rev. 2008, 252, 659–701.
T. Hudlicky, A. J. Thorpe, Chem. Commun. 1996, 1993.
H. C. Kolb, M. S. VanNieuwenhze, K. B. Sharpless, Chem. Rev. 1994,
94, 2483–2547.
High-throughput array experiments design.
The array experiments were performed in 384 well density MATRICAL
MP100-1-PS Microplates, and 1536 well density chemically resistant
microplates MatriPlates MCR111-1-1, black and total working volume 120
μl and 5 μl respectively. The fluids were dispensed using a robotic high-
precision 16-channel liquid handling system Nanodrop Express (BioNex
Solution Inc.) with fixed-tip head 2x8 configuration. Each experiment was
performed in 24 repetitions. For the control experiments an equal amount
of acetonitrile was added instead of the chiral diol solutions. The printed
plate was then incubated (3 min) and centrifuged to establish the
[4]
J. Seyden-Penne, Chiral Auxiliaries and Ligands in Asymmetric
Synthesis, Wiley, New York, 1995.
T. P. Yoon, Science 2003, 299, 1691–1693.
Q.-L. Zhou, Ed. , Privileged Chiral Ligands and Catalysts, Wiley-VCH,
Weinheim, Germany, 2011.
K. B. Sharpless, K. Akashi, J. Am. Chem. Soc. 1976, 98, 1986–1987.
M. A. Andersson, R. Epple, V. V. Fokin, K. B. Sharpless, Angew.
Chem. Int. Ed. 2002, 41, 472–475.
[5]
[6]
[7]
[8]
[9]
K. Murata, K. Okano, M. Miyagi, H. Iwane, R. Noyori, T. Ikariya, Org.
Lett. 1999, 1, 1119–1121.
Org. Synth. 2005, 82, 10.
[10]
[11]
equilibrium
in
the
formation
of
diastereoisomeric
Y. Chen, C. Chen, X. Wu, Chem. Soc. Rev. 2012, 41, 1742.
iminoboronate/oxazolidine-boronate ester complexes. Plates were
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
[12]
D. Leung, S. O. Kang, E. V. Anslyn, Chem Soc Rev 2012, 41, 448–
479.
[27]
A. Wilson, G. Gasparini, S. Matile, Chem Soc Rev 2014, 43, 1948–
1962.
[13]
[14]
D. Parker, Chem. Rev. 1991, 91, 1441–1457.
A. M. Kelly, Y. Pérez-Fuertes, S. Arimori, S. D. Bull, T. D. James, Org.
Lett. 2006, 8, 1971–1974.
[28]
[29]
B. W. McCrindle, L. Ose, A. D. Marais, J. Pediatr. 2003, 143, 74–80.
P. Jones, S. Kafonek, I. Laurora, D. Hunninghake, Am. J. Cardiol.
1998, 81, 582–587.
[15]
[16]
[17]
[18]
[19]
A. M. Kelly, Y. Pérez-Fuertes, J. S. Fossey, S. L. Yeste, S. D. Bull, T.
D. James, Nat. Protoc. 2008, 3, 215–219.
S. L. Yeste, M. E. Powell, S. D. Bull, T. D. James, J. Org. Chem.
2009, 74, 427–430.
K. W. Bentley, Y. G. Nam, J. M. Murphy, C. Wolf, J. Am. Chem. Soc.
2013, 135, 18052–18055.
H. H. Jo, C.-Y. Lin, E. V. Anslyn, Acc. Chem. Res. 2014, 47, 2212–
2221.
[30]
P. S. Sever, B. Dahlöf, N. R. Poulter, H. Wedel, G. Beevers, M.
Caulfield, R. Collins, S. E. Kjeldsen, A. Kristinsson, G. T. McInnes, et
al., The Lancet 2003, 361, 1149–1158.
M. R. Law, BMJ 2003, 326, 1423–0.
M. Korhonova, A. Doricakova, Z. Dvorak, PLOS ONE 2015, 10,
e0137720.
B. R. Krause, R. S. Newton, Atherosclerosis 1995, 117, 237–244.
T. A. Kocarek, Drug Metab. Dispos. 2002, 30, 1400–1405.
M. Otto, Chemometrics: Statistics and Computer Application in
Analytical Chemistry., Wiley-VCH, Weinheim; Cambridge, 1998.
R. G. Brereton, Applied Chemometrics for Scientists, John Wiley &
Sons, Chichester, Englandꢀ; Hoboken, NJ, 2007.
P. Anzenbacher, Jr., P. Lubal, P. Buček, M. A. Palacios, M. E.
Kozelkova, Chem. Soc. Rev. 2010, 39, 3954.
N. Cristianini, J. Shawe-Taylor, An Introduction to Support Vector
Machines: And Other Kernel-Based Learning Methods, Cambridge
University Press, Cambridgeꢀ; New York, 2000.
J. A. Burns, G. M. Whitesides, Chem. Rev. 1993, 93, 2583–2601.
S. H. Shabbir, C. J. Regan, E. V. Anslyn, Proc. Natl. Acad. Sci. 2009,
106, 10487–10492.
[31]
[32]
[33]
[34]
[35]
E. A. Dragu, J.-V. Naubron, A. Hanganu, A. C. Razus, S. Nica,
Chirality 2015, 27, 826–834.
[20]
[21]
[22]
K. W. Bentley, D. Proano, C. Wolf, Nat. Commun. 2016, 7, 12539.
W. H. Pirkle, T. C. Pochapsky, Chem. Rev. 1989, 89, 347–362.
E. G. Shcherbakova, T. Minami, V. Brega, T. D. James, P.
Anzenbacher, Angew. Chem. Int. Ed. 2015, 54, 7130–7133.
E. G. Shcherbakova, V. Brega, T. Minami, S. Sheykhi, T. D. James, P.
Anzenbacher, Chem. - Eur. J. 2016, 22, 10074–10080.
M. Hutin, G. Bernardinelli, J. R. Nitschke, Chem. - Eur. J. 2008, 14,
4585–4593.
Y. Pérez-Fuertes, A. M. Kelly, A. L. Johnson, S. Arimori, S. D. Bull, T.
D. James, Org. Lett. 2006, 8, 609–612.
E. Galbraith, A. M. Kelly, J. S. Fossey, G. Kociok-Köhn, M. G.
Davidson, S. D. Bull, T. D. James, New J Chem 2009, 33, 181–185.
[36]
[37]
[38]
[23]
[24]
[25]
[26]
[39]
[40]
This article is protected by copyright. All rights reserved.

10.1002/chem.201701923
Chemistry - A European Journal
FULL PAPER
Entry for the Table of Contents
FULL PAPER
Self-assembled fluorescent probes:
incorporates chiral diols, allowing for
simultaneous determination of ee, de, and
product yields in a complicated matrix with
only 10-20 ng/well of the substrate.
Fluorescence is directly proportional to
the absolute configuration and
Elena G. Shcherbakova, Valentina
Brega, Vincent M. Lynch, Tony D.
Jame], and Pavel Anzenbacher, Jr.*
Page No. – Page No.
concentration of the diols. The
High-Throughput Assay for
applicability of this simple approach,
demonstrated on the products of Noyori
asymmetric hydrogenation of benzil and
lipid-lowering drug atorvastatin in a highly
parallel fashion with errors <1% ee
confirming the feasibility of assaying
crude products in real time, without any
work up, as required for high-throughput
Enantiomeric Excess Determination
in 1,2- and 1,3-Diols and Direct
Asymmetric Reaction Screening
screening
.
This article is protected by copyright. All rights reserved.