Unravelling mechanistic features of organocatalysis with in situ modifications at the secondary sphere

Article abstract of DOI:10.1038/s41557-019-0258-1

Secondary-sphere interactions serve a fundamental role in controlling the reactivity and selectivity of organometallic and enzymatic catalysts. However, there is a dearth of studies that explicitly incorporate secondary-sphere modifiers into organocatalytic systems. In this work, we introduce an approach for the in situ systematic modification of organocatalysts in their secondary sphere through dynamic covalent binding under the reaction conditions. As a proof-of-concept, we applied boronic acids as secondary-sphere modifiers of N-heterocyclic carbenes that contained a hydroxy handle. The bound system formed in the reaction mixture catalysed the enantioselective benzoin condensations of a challenging substrate class that contains electron-withdrawing groups. Linear regression coupled with data visualization served to pinpoint the divergent origins of enantioselectivity for different substrates and decision tree algorithms served to formulate selection criteria for the appropriate secondary-sphere modifiers. The combination of this highly modular catalytic approach with machine-learning techniques provided mechanistic insights and guided the streamlined optimization process of a gram-scale reaction at low organocatalyst loading.

Full text of DOI:10.1038/s41557-019-0258-1

Articles
https://doi.org/10.1038/s41557-019-0258-1
Unravelling mechanistic features of organocatalysis
with insitu modifications at the secondary sphere
Vasudevan Dhayalanꢀ ꢀ1,2, Santosh C. Gadekarꢀ ꢀ1,2, Zayed Alassadꢀ ꢀ1,2 and Anat Miloꢀ ꢀ1*
Secondary-sphere interactions serve a fundamental role in controlling the reactivity and selectivity of organometallic and
enzymatic catalysts. However, there is a dearth of studies that explicitly incorporate secondary-sphere modifiers into
organocatalytic systems. In this work, we introduce an approach for the insitu systematic modification of organocatalysts in
their secondary sphere through dynamic covalent binding under the reaction conditions. As a proof-of-concept, we applied
boronic acids as secondary-sphere modifiers of N-heterocyclic carbenes that contained a hydroxy handle. The bound system
formed in the reaction mixture catalysed the enantioselective benzoin condensations of a challenging substrate class that
contains electron-withdrawing groups. Linear regression coupled with data visualization served to pinpoint the divergent
origins of enantioselectivity for different substrates and decision tree algorithms served to formulate selection criteria for
the appropriate secondary-sphere modifiers. The combination of this highly modular catalytic approach with machine-learn-
ing techniques provided mechanistic insights and guided the streamlined optimization process of a gram-scale reaction at
low organocatalyst loading.
raditionally, optimizing the reactivity and selectivity of cata- mode of the modifier was required to be orthogonal to catalytic
lytic reactions is accomplished by modifying the catalyst activity to avoid catalyst inhibition. Likewise, the modifier was to be
T
structure prior to performing the reaction. Here we pro- located in close proximity to the active site and possess the capacity
pose a method to fine-tune the catalyst structure under reaction for dynamic and non-covalent interactions that could guide reac-
conditions by promoting the binding of a modifier to its second- tivity and selectivity. To achieve this goal, we selected to harness
ary sphere. The secondary or outer coordination sphere was first the propensity of aryl boronic acids (BAs) to form boronic ester
described more than a century ago by Alfred Werner in the con- bonds with hydroxyl groups (Fig. 1a). The practical benefit of this
text of organometallic complexes as groups that are not directly binding mode stems from the availability and structural diversity of
bonded to a metal, but are rather coordinated to the ligands1–3. This BAs and the ease of identifying and preparing organocatalysts with
definition has since been expanded to any moiety in the molecu- pendant hydroxyl groups. Boronic ester bonds can be described
lar microenvironment of coordination compounds that influences as dynamic covalent bonds, which are considered to have a dual
the orientation and electronic properties of their ligands by intro- nature, because—depending on the conditions (solvent, tempera-
ducing non-covalent interactions, such as hydrogen bonding, elec- ture, pKa, additives)—they can undergo rapid exchange, similar to
trostatic forces and hydrophobic effects4–6. Although it is now well a non-covalent bond, or exhibit stability, similar to a covalent bond.
established that such interactions have a fundamental effect on the Whereas the application of dynamic covalent chemistry has flour-
reactivity and selectivity in enzymatic catalysis7–9, metal-mediated ished in the fields of supramolecular and materials chemistry, it
processes and transition metal chemistry4–6,10–14, the systematic has been limited in organocatalysis15–27. We hypothesized that BAs
study of secondary-sphere interactions in organocatalysis remains would bind to the catalyst under reaction conditions and affect its
uncommon. The secondary sphere comprises moieties that are not reactivity and selectivity by promoting dynamic covalent and non-
an integral part of an active site, yet are located in proximity to it covalent π and hydrogen bonding interactions with the substrates
and are involved in its mechanism of action through dynamic or at the different steps of the catalytic cycle. Concomitantly, the
non-covalent interactions. These interactions control the geometry application of mathematical modelling, coupled with mechanistic
and electronic properties of the reaction intermediates and transi- experiments, would uncover the guiding principles by which BAs
tion state(s). In this work, we establish a straightforward and sys- influence catalysis.
tematic strategy to modify the secondary sphere of organocatalysts
As a proof-of-concept, we sought to identify a known organo-
by combining a highly modular catalytic system with mathematical catalytic system in which a hydroxyl group was located in proximity
modelling techniques to facilitate the discovery and prediction of to the active site and was explicitly invoked to explain the reactiv-
secondary-sphere design principles. This methodology is predi- ity and selectivity. Such a system was reported by Zeitler, Connon
cated on modifying the structure of an organocatalyst in situ under and co-workers in the context of the benzoin condensation (Cat5F
reaction conditions and providing a tunable handle for the nuanced (Fig. 1b))28,29. The benzoin condensation has been studied for over
recognition of substrates by their geometry and electronic proper- a century and involves the coupling of two aldehydes to afford a
ties, and thus streamline the optimization, mechanistic study and benzoin product through the inversion of polarity (umpolung) of
discovery of organocatalytic systems.
one of the two aldehydes (Fig. 1c shows the commonly accepted
By promoting the in situ binding of a modifier to the catalytic catalytic cycle). This reaction is influenced by non-covalent inter-
system under reaction conditions, we intended to curtail the need actions30–32 and, indeed, the organocatalytic system described by
for prior synthetic steps to fine-tune catalytic scaffolds. The binding Zeitler and Connon was considered the most efficient in terms of
1Department of Chemistry, Ben-Gurion University of the Negev, Beer Sheva, Israel. 2These authors contributed equally: Vasudevan Dhayalan,
Santosh C. Gadekar, Zayed Alassad *e-mail: anatmilo@bgu.ac.il
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
a
d
Cat5F (6 mol%)
BA (7.2 mol%)
HO
O
O
Cl
OH
HO
HO
Cat
O
THF, 20 °C, 3Å MS
Rb2CO3 (4 mol%)
+
Cl
Cl
B
Cat
B
HO
100
80
60
40
20
0
24 h
6 h
b
HO
O
O
Cat5F (4 mol%)
THF, 0–18 °C
Rb2CO3 (4 mol%)
R
R
R
–
BF4
N
+
F
N
N
F
Ph
Ph
OH
Ar
F
F
F
Cat5F
c
+
e
–
N
BF4
N
N
Cat5F (6 mol%)
4-CF3Ph-BA (7.2 mol%)
HO
O
O
N
R
THF, 20 °C, 3Å MS
Rb2CO3 (4 mol%)
R
R
R
Base
100
80
60
40
20
0
HO
O
With BA
Without BA
O
Ar
δ+
*
N
N
R
Ar
R
Ar
+
+
N
N
N
N
H
–
H
N
N
HO
O
–
O
R
Ar
N
N
δ–
O
N
OH
δ+
R
Substrate R Group
Fig. 1 | the influence of secondary-sphere modulation of NhCs on the benzoin reaction. a, Conceptual framework: stabilizing boronic ester bonds in situ
under catalytically relevant conditions to tune the reactivity and selectivity at the secondary sphere. b, The optimal organocatalyst for the benzoin reaction
to date developed by Zeitler, Connon and co-workers contains a hydroxyl group invoked in non-covalent interactions and is limited in its scope of substrates
with electron-withdrawing groups28. c, Commonly accepted catalytic cycle for the benzoin condensation. d, Comparison of the enantioselectivity obtained in
the benzoin condensation using several BA modifiers and 3-chlorobenzaldehyde as the substrate at 6ꢀhours and at 24ꢀhours. Error bars represent the range of
two measurements. 9-Anth, 9, anthracene. e, Comparison of the enantioselectivity obtained in the benzoin condensation with different electron-withdrawing
substituents in the presence and absence of 4-CF3Ph-BA. Dashed horizontal lines represent the average enantioselectivity across each set: 90% e.e. (e.r. 95:5)
with 4-CF3Ph-BA and 60% e.e. (e.r. 80:20) in its absence. Reactions were run for 6 hours. Error bars represent the s.d. between at least two measurements.
catalyst loading and selectivity to date33. Nonetheless, the paucity Reactions that were conducted for 24 h with different BAs led to a
of examples of organocatalytic enantioselective benzoin condensa- range of enantioselectivity values up to 97:3 (Fig. 1d). To our sur-
tions of aldehydes with electron-withdrawing substituents provided prise, the full set of BAs studied led to e.r. values above 95:5 for 6 h
an ideal opportunity to examine the concept of leveraging second- reactions. Consequently, we set out to study a scope of electron-
ary-sphere interactions by the in situ introduction of BAs.
withdrawing substrates in the presence of the best performing
BA, 4-(trifluoromethyl)phenylboronic acid (4-CF3Ph-BA). The
e.r. obtained for the substrates in the presence of 4-CF3Ph-BA was
results and discussion
Catalytic experiments. Reactions were performed with a model between 88:12 and 98.5:1.5 and was consistently improved and
substrate, 3-chlorobenzaldehyde, reported by Zeitler, Connon more reproducible compared to the same reaction in the absence
and co-workers to afford an enantiomeric ratio (e.r.) of 84:16. of BA (Fig. 1e).
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
a
b
2.5
2.0
1.5
1.0
4-F
4-Cl
3-F
4-I
µL
4-Br
4-COOMe
3-Br
3-Cl
4-CF3
µB
4-CN
3,5-Cl
3-I
R2 = 0.92
y = 0.92x + 0.15
3-CF3
3,5-Br
B5
3-COOMe
Average-Q23Fold = 0.83
L
3-NO2
Validation n = 500
c
∆∆G⧧
2.5
µL
1.0
1.5
Measured ∆∆G⧧ (kcal mol–1
∆∆G⧧ = 4.87–0.60B5–0.11L-0.23 µL
2.0
2.5
B5
)
2.0
1.5
1.0
0.5
0
L
With acetic
acid
With Cat5H
∆∆G⧧norm = 1.85–0.50B5–0.10L-0.29 µL
4-F (98.25:1.75)
3-F (98:2)
3-COOMe (94:6)
4-CN (93.5:6.5)
3-NO2 (88:12)
d
–
–
–
BF4
F
BF4
BF4
HO
O
Cat (6 mol%)
4-CF3Ph-BA (7.2 mol%)
O
N
N
N
N
N
+
+
+
N
N
N
N
F
R
Ph
Ph
Ph
Ph
Ph
Ph
THF, 20 °C, 3Å MS
Rb2CO3 (4 mol%)
Acetic acid (0 or 10 mol%)
R
R
OH
F
F
OH
Cat5H
OH
F
Cat5F
CatMes
Fig. 2 | Multivariate model of the enantioselectivity for 6ꢁhour reactions across a set of electron-withdrawing aldehydes. a, Parameters that were
identified as predictive for the enantioselectivity in this reaction. b, Measured versus predicted enantioselectivity represented as ∆∆G‡ꢀ=ꢀ−RTln([R]/[S])
given by a model that included three molecular descriptors (the normalized parameters were obtained by subtracting the mean and dividing by the
s.d.). Goodness-of-fit is represented by R2 and the average Q2 of a threefold (Q32fold) validation performed 500 times on randomized sets. c, Graphical
representation of the relative contribution of each parameter to the predicted enantioselectivity (grey diamonds) that highlights an inverse correlation of
all three parameters with the resulting enantioselectivity and the putative differences in the origin of enantioselectivity for different substrates. Light grey
diamonds represent the enantioselectivity on the addition of either acetic acid (blue arrow) or Cat5H (orange arrow). d. Conditions and different catalysts
tested in the reaction.
Substrate trend analysis for 6hour reactions. The obtained results we were in the position to probe experimentally the underlying fac-
did not follow a simple electronic or steric trend (Fig. 1e). Therefore, tors that regulate the enantioselectivity each parameter signifies.
we sought to identify a multivariate model to shed light on the The juxtaposition of parameter contributions and ∆∆G‡ values
origin of enantioselectivity in the presence of 4-CF3Ph-BA. The highlights the inverse correlation between each of these parameters
substrate was the only reaction component that changed between and the resultant enantioselectivity. Namely, the e.r. was diminished
experiments and, therefore, molecular descriptors that pertained to for substrates that were longer, wider and had a strong dipole on
the geometric, steric and electronic features of the substrates were the L axis.
collected. Hammett values34, Sterimol parameters35, vibration fre-
To interrogate the role of each parameter in predicting enantiose-
quencies and intensities36, and NBO charges37 were evaluated. As lectivity, we focused on two substrates that afford similar e.r. values,
the models obtained with these known parameters were not satis- but for which different parameters predominantly predict the selec-
factory in terms of their goodness-of-fit (Supplementary Section tivity, 3-carbomethoxybenzaldehyde and 4-cyanobenzaldehyde
5 gives details and several examples), we decided to add the com- (Fig. 2c). As μL is aligned with the aldehyde moiety it was speculated
ponents of the dipole moment using a fixed Cartesian coordinate that it could represent the directional electron-withdrawing capac-
system across all substrates (Fig. 2a). It was hypothesized that this ity of the substituents on the benzoin product and, therefore, could
new parameter system would be advantageous to describe the direc- be related to base-catalysed racemization. To probe this assump-
tional electrostatic and electronic effects. Indeed, with the introduc- tion further, correlations between μL and the other parameters
tion of the dipole moment components, a highly predictive model were sought and, indeed, a strong correlation emerged between the
was identified (Fig. 2b). The model included the component of the dipole moment on the length axis and the charges on the aldehyde
dipole moment on the L (maximal length) axis (μL) and two ste- moiety (Supplementary Fig. 17 gives details). To mitigate the postu-
ric Sterimol parameters, the maximal length of the aldehyde ring lated effect of the base, catalytic reactions with each of the substrates
with its substituents (L) and the maximal width perpendicular to in the presence of 10mol% of acetic acid were performed. It was
the length (B5 (Fig. 2a)). To better understand the meaning of these anticipated that if the dipole moment is related to base-catalysed
parameters, a graphical representation of each parameter’s contri- racemization, the addition of acetic acid would have a greater influ-
bution to the predicted enantioselectivity was produced by applying ence on 4-cyanobenzaldehyde because μL is the main parameter
the model to the respective parameters for several of the substrates that predicts its enantioselectivity (Fig. 2c, blue). Indeed, follow-
(Fig. 2c; Supplementary Table 7 gives further details). By doing so ing our prediction, with the addition of acetic acid, the ∆∆G‡ value
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
a
b
Cat5F (6 mol%)
BA (7.2 mol%)
HO
∠
HO
O
tor
O
B
B
CF3
CH3
Predictor importance
0.02
HO
HO
<38.7
≥38.7
2-CF3-Ph 0.81*
R
THF, 20 °C, 3Å MS
Rb2CO3 (4 mol%)
µB
R
100
R
HO
9-Anth
2,6-MeO-Ph
<–0.8
3-CF3
≥0.8
100
80
60
40
20
0
0.016
0.012
0.008
0.004
0
No boronic acid
1.03*
<1.9
µL
≥1.9
80
60
40
20
0
∠
tor
4-CF3-Ph 1.90*
3,5-CF3-Ph
<8.6
≥8.6
3-MeO-Ph
1.68*
2-MeO-Ph 1.48*
3-Me-Ph
3-F-Ph
2-Me-Ph
2-F-Ph
2-Ph-Ph
O
O
4-MeO-Ph
4-Me-Ph
4-F-Ph
3,5-MeO-Ph
3,5-F-Ph
C6F5
Measured versus
F3C
H3C
predicted ∆∆G⧧
R2 = 0.93
Ph
Average L3-Fold = 0.15
*Predicted ∆∆G⧧
0
18
36
54
72
0
18
36
54
72
Time (h)
Time (h)
c
d
HO
O
O
O
O
O
Conditions Cl
Cl
OH
+
+
2.0
Cl
Cl
Cl
Cl
2,6-OMe-Ph
9-Anth
Benzoin
Aldehyde
Benzoic acid
Benzil
2.0
∠
Loss of ∆∆G‡
tor
1.0
1.5
1.0
0.5
0.0
0.8
0.6
0.4
0.2
0.0
1.5
1.0
0.5
0.0
3,5-CF3-Ph
C6F5
R2=0.92
2-Ph-Ph
2-OMe-Ph
y = 0.92x + 0.04
Av-Q210-Fold = 0.88
Validation n = 500
4-CF3-Ph
νC–B
0.0
0.5
Measured ∆∆G ⧧ loss (kcal mol–1
tor Torsion angle
1.0
1.5
2.0
)
∠
∆qC–B
νC–B
Charge difference C–B
Frequency of vertical C–B stretch
)
)
se
se
at
%
HF
%
a
a
l
∆∆G‡norm = 0.58 + 0.16 ∆qC–B + 0.16 νC–B + 0.31 ∠tor
C
6
l
T
B
B
o
o
+
t
m
m
a
4
C
(
(
Fig. 3 | Disentangling the effect of racemization. a, Model for 24ꢀhour reactions with 3-chlorobenzaldehyde as the substrate and different BAs: decision
tree (left (see text)) and estimation of each predictor’s importance based on the mean standard error (MSE) decrease at each node (right). Bold black
numbers are the predicted ∆∆G‡(in kcal mol–1). Goodness-of-fit is represented by R2 and the average L3fold (predictive loss) value of a threefold validation
∠
performed 500 times on randomized sets. tor, torsion angle. b, Erosion of enantioselectivity over time with and without BAs for electron-donating and
withdrawing substrates 3-methylbenzaldehyde and 3-trifluoromethylbenzaldehyde. Error bars represent the range of two measurements. c, Racemization
experiments starting from enantioenriched 3-chlorobenzoin: the left axis represents product distribution ratios and the right axis represents the loss of
enantioenrichment in ∆∆G‡. Error bars represent the s.d. between at least two measurements. d. Linear regression model and structural features for loss
of enantioenrichment in racemization experiments of enantioenriched 3-chlorobenzoin with different BAs. νC–B, charge difference C–B; qC–B, frequency of
vertical C–B stretch.
for 3-carbomethoxybenzaldehyde remained 1.6kcalmol–1, whereas Cat5H has an increased pKa value compared to Cat5F, and thus one
for 4-cyanobenzaldehyde it increased from 1.55 to 2.02kcalmol–1 could consider this to be the origin of its increased enantioselec-
(Fig. 2c, blue arrow). We suspected that the maximal width, B5, may tivity28,39,40. However, if this were the case, CatMes, which is esti-
be related to electrostatic effects given that such effects were found mated to have a greater pKa value39,40, would be expected to increase
computationally to have a strong influence on enantioselectivity in enantioselectivity further, but in fact it significantly decreases it
N-heterocyclic carbene (NHC) catalysis38. Based on the assumption (Supplementary Table 13). This outcome fits well with electrostatic
that the substituents on the catalyst aryl ring would have a decisive interactions, which would presumably be disrupted due to the steric
influence on the electrostatic interactions, the reaction of each of hindrance and torsional constraints introduced by a mesityl ring.
these two substrates was performed using two different catalysts Overall, these results highlight the unique capacity of multivariate
(Cat5H and CatMes (Supplementary Section 6.2)). It was anticipated modelling and data visualization approaches to distinguish the ori-
that if the maximal width is representative of the electrostatic inter- gins of enantioselectivity for different substrates even when their
actions with the catalyst, we would see a greater effect on 3-car- e.r. value is similar. As demonstrated, such insights can facilitate the
bomethoxybenzaldehyde because B5 is the main parameter that tailored optimization of reaction outcomes.
predicts its enantioselectivity (Fig. 2c, orange). In accord with our
predictions, for 3-carbomethoxybenzaldehyde the ∆∆G‡ increased BA trend analysis for 24h reactions. Due to the erosion in enan-
to 2.02kcalmol–1 (Fig. 2c, orange arrow) in the presence of Cat5H
,
tioselectivity during 24h reactions and the observation that adding
whereas for 4-cyanobenzaldehyde it did not change significantly. acetic acid improved the enantioselectivity for certain substrates, it
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
b
c
a
0
–0.2
–0.4
–0.6
–0.8
–1
H
–
BF4
F
H
H
0.15
0.10
0.05
0.00
N
N
+
H
N
Ha
F
F
F
O
Hb
F
Hc
HO
HO
B
CF3
y = –0.50x 2 + 0.49x + 0.01
R 2 = 0.99
Ka = 27 M–1
Hc
H
H
∆
G = 2 kcal mol–1
0.0
0.5
1.0
H
N
N
F
+
F
–1.2
Mole fraction catalyst
H
N
Ha
F
0
5
10
15
OH
BA:catalyst ratio
F
O
B
F
Hc
d
Hc
–
BF4
CF3
Hc
HO
BA
Ha
Hb
Cat
20 °C
15 °C
10 °C
5 °C
0 °C
–5 °C
–10 °C
–15 °C
–20 °C
–30 °C
8.0
7.8
7.6
7.4
7.2
7.0
6.8
6.6
ppm
Fig. 4 | Binding studies of BA to NhC in thF. a, Proposed dynamic covalent binding. b, Job’s plot reveals a 1:1 binding stoichiometry. c, NMR binding studies
using a global fit of three proton peak shifts (Supplementary Section 3.2 gives details). d, NMR at lower temperatures reveals slower exchange rates and
stabilization of the boronic ester.
seemed likely that the 24h reaction results represent two pro- by following the reaction over time for 3-methyl- and 3-trifluo-
cesses: benzoin product formation and subsequent racemization. romethylbenzaldehyde (Fig. 3b). In both cases, in the absence
As the outcome was a result of these two competing processes, we of BA the initial enantioselectivity decreased over time and the
were unable to identify a model to describe the 24h reactions by reproducibility of the results was limited41. Adding 4-MePh-BA or
linear regression. We turned to decision tree analysis and were 4-CF3Ph-BA led to an improved stability of enantioselectivity over
able to identify a model that highlighted the BA structural fea- time and the latter shows a superior performance over longer peri-
tures related to the determination of selectivity at 24h reactions ods of time, which is especially notable in the presence of electron-
(Fig. 3a). Regression trees resemble flow charts, in which each withdrawing substrates.
node (diamond) represents an attribute, each branch represents the
splitting criteria of this attribute (on the left the values below the Racemization analysis. Although the decision tree model serves
split and on the right those equal to or above it) and each leaf, or to classify BAs based on their structural features and is predictive
terminal node (green circle), contains a class of BAs (green) that of the observed enantioselectivity at 24h, it is limited in terms of
lead to a predicted value, in our case the obtained ∆∆G‡ (black). delineating the racemization mechanism. To probe the mechanistic
The top node in a tree is the root node, which in this case is the underpinnings of the racemization process, several control experi-
dihedral or torsion angle. BAs that bear a relatively high torsion ments were designed in which enantioenriched 3-chlorobenzoin was
angle between the BA moiety and the aromatic residue lead to the subjected to reaction conditions while omitting certain components.
lowest enantiomeric excess at 24h, predicted at 0.81kcalmol–1. The The erosion of enantiomeric excess and the product distribution
component of the dipole moment in the direction of the width were analysed by HPLC (Fig. 3c and Supplementary Section 6.6).
(μB) serves to distinguish 3-CF3Ph-BA, which has a relatively large From these results, it appears that BA suppresses both the base-
electron-withdrawing group on the B axis and is predicted to give and NHC-catalysed racemization processes as well as the oxidation
∆∆G‡ =1.03kcalmol–1. The component of the dipole moment in process to benzil. The reversibility of the benzoin condensation has
the direction of the length (μL) classifies the BAs with the highest previously been described in the literature42–44, and the accretion of
enantioselectivity, predicted at 1.9kcalmol–1, which contains the aldehyde in the presence of BA suggests that it may be involved in
largest electron-withdrawing groups on the L axis, 4-CF3 and 3,5- the retro-benzoin process. Overall, these results suggest that BAs are
CF3. Finally, an additional limitation on the torsion angle separates probably involved in both the selectivity- and turnover-determining
the remaining BAs with longer 2-substituent groups, 2-MeO and steps of the benzoin reaction. This assertion was supported by the
2-Ph, from the rest of the set.
exceedingly different reaction progress profiles that emerged by fol-
We sought to probe whether the racemization process over lowing the change in product yield over time in the presence and
long reactions was unique to electron-withdrawing substrates absence of BA (Supplementary Table 25).
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
a
HO
D
HO
D
HO
D
O
O
O
O
O
Cat5F (6 mol%)
BA (7.2 mol%)
Cl
O
Cl
O
Cl
H
H
+
+
+
+
THF, 20 °C, 3Å MS
Rb2CO3 (4 mol%)
24 h
Cl
Cl
Cl
Cl
Cl
Cl
Cl
e.r.
e.r.
e.r.
–
D/H
D/H
D/H
Equiv
1
Equiv
1
Product ratio
–
–
97.5:2.5
96:4
97.5:2.5
97:3
0
4-CF3Ph-BA
4-MePh-BA
no BA
96:4
0.47
0.48
0.53
0.43
0.1
0.1
0
–
96:4
0.42
0
0
53:47
27:73
0.47 50.5:49.5 23:77
b
c
Ar
+
–
BF4
N
Ar
+
N
HO
D
Cat5F (6 mol%)
HO
D
O
O
Cl
N
N
Rb2CO3 (4 mol%)
N
F
F
Ar =
F
H
–
F
F
THF, 20 °C,
3Å MS, 24 h
N
R
O
Cl
Cl
Cl
Cl
O
Base
N
R
H
Ar
e.r.
e.r.
+
N
D/H
90:10
D/H
HO
H
O
HO
D
O
E
Ar
N
Stable
adduct
97:3
With Cat5F
70:30
25:75
80:20
O
O
+
N
N
N
B
Without Cat5F
90:10
R
H
O
Ph
Ph
R
E
A
Stable
adduct
Ar
Ar
d
HO
O
+
O
Cat5F (0.1 mol%)
N
N
R = Me
R = CF3
752.13
N
N
BA (0.15 mol%)
Cl
H
–
B
D
698.15
D
–
N
N
HO
R
O
THF, 20 °C, 3Å MS
Cl
Cl
O
Rb2CO3 (0.067 mol%)
20 h
R
50 mmol
7 g
e.r.
97.5:2.5
97.5:2.5
97:3
% yield
27
Ar
4-CF3Ph-BA
4-MePh-BA
N
N
O
O
699.15
61
N
OH
700.16
D
H
753.13
754.13
2,4-diMePh-BA
C
88
R
Breslow
intermediate
697.14
751.12
701.16
755.13
m/z
m/z
Fig. 5 | Cross-over experiments between the benzoin product and an additional aldehyde. a, Enantioenriched and deuterated 3-chlorobenzoin and
2-chlorobenzaldehyde under reaction conditions with and without BAs. The enantioenrichment and deuterium incorporation were determined for
the remaining 3-chlorobenzoin and for the cross-benzoin product. Complete erosion of enantioenrichment and substantial deuterium scrambling
reveal the full reversibility of the reaction in the absence of BA. b, Proposed catalytic cycle and the formation of a stable acyl anion adduct with
2-chlorobenzaldehyde, which is presumed to be further stabilized by BA. Mass spectrometry (matrix-assisted laser desorption/ionization–time of flight) is
consistent with the formation of a stabilized 1:1:1 adduct of NHC:BA:2-chlorobenzaldehyde. One plausible stable adduct that fits with the observed m/z is
proposed. c. Racemization and deuterium-loss experiments (24ꢀhours) in the presence of base with and without the NHC catalyst reveal that the erosion
of enantioenrichment and the loss of the deuterium label under reaction conditions are enhanced by the presence of the NHC catalyst. d. Streamlined
optimization of a gram-scale reaction at a 0.1ꢀmol% NHC loading with different BAs.
To probe the influence of changes to the structure of BA on the from 24h reactions and racemizations are not correlated and only
racemization process we ran 24h racemization reactions with dif- the racemization results are somewhat correlated to the pKa of the
ferent BAs (Supplementary Section 6.5). We were able to identify a BAs (Supplementary Figs. 21–23). Taken together with the more
highly predictive model that reflected the structural features leading pronounced importance of the torsion angle in predicting 24h reac-
to racemization (Fig. 3d). Akin to the model for 24h reactions, the tions, we hypothesize that beyond decelerating base catalysed race-
torsion angle was key and BAs with minimal torsion angles led to mization, BAs are involved in inhibiting the retro-benzoin reaction
a decreased racemization. In addition, several features that related by binding to NHC or one of its reactive intermediates.
to the BA moiety were implicated in describing the observed race-
mization. Enhanced racemization was observed using BAs with Binding studies. The binding between BA and Cat5F was studied by
higher C–B vibration frequencies and larger differences in charge NMR (Supplementary Section 3 gives experimental details). A Job
between the boron and the adjacent carbon on the aromatic moiety. plot was obtained in THF-d8 that suggests that BA and Cat5F afford
Increased charge differences and higher vibration frequencies entail a 1:1 complex (Fig. 4a,b)45. A 1:1 binding constant of 27M−1 was
a relative negative charge on carbon, which decreases the electro- calculated based on fast exchange equilibria on the NMR timescale,
philicity of boron, thus decreasing its affinity to sequester the base. which corresponds to 2kcalmol–1 (Fig. 4c)46. Based on these results,
However, a less-electrophilic boron would also be less likely to bind the weakly bound Cat5F–BA complex could be hydrogen bonded
to a hydroxy group on the NHC catalyst or any hydroxy-containing or could involve the release of a water molecule to form a dynamic
intermediate on or off the catalytic cycle. This option would incur a boronic ester bond (Fig. 4b). To study these possibilities, variable
higher steric penalty, and thus could be reflected by the decrease in temperature 1H NMR measurements were performed in THF-d8. At
selectivity observed at increased torsion angles. Note that the results room temperature, both the proton on the stereogenic centre, Ha
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
(orange, Fig. 4d), and the hydroxyl peak of the Cat5F, Hb (green, the mechanism and mode of BA binding throughout the catalytic
Fig. 4d), shifted upfield by 0.1ppm on the introduction of BA, which cycle, and the precise origin of reactivity and selectivity.
rules out hydrogen bonding as we would expect the BA hydroxylic
proton (blue, Fig. 4d) to be shifted as well. BAs have been shown Gram-scale reactions. Based on the insight gained through the
to form dynamic-covalent bonds with hydroxy groups47, therefore, mechanistic studies of this system, we wished to exploit the stabi-
a fast exchange on the NMR scale was hypothesized between a lization afforded by BAs to perform a gram-scale reaction at a low
bound boronic ester and the free Cat5F and BA. To test this pro- catalyst loading (Fig. 5d). The reaction was performed at a 50mmol
posal, low-temperature NMR measurements were performed. With scale (7g) with a catalyst loading of 0.1mol%, with 0.15mol% BA
the decrease in temperature, the Cat5F hydroxyl peak, Hb (green), and 0.067mol% base. Such mild conditions and low catalyst load-
broadened and decreased in intensity, the BA hydroxyl peak, OH ing are unprecedented for the benzoin condensation with an NHC
(blue), shifted downfield, as did the protons at the ortho positions, catalyst. We started by applying 4-CF3Ph-BA and, although the e.r.
Hc (red) (Fig. 4d). These results fit well with the stabilization of the was 97.5:2.5, the yield even after 72h was only 34% (Supplementary
proposed boronic ester at lower temperatures. Note that under the Section 6.14). It was noted that under these benign conditions, race-
reaction conditions base and molecular sieves were added, which mization over time was negligible. We turned to additional BAs that
are both known to promote the formation of boronic esters47–49
.
led to the least loss of enantioselectivity in the 24h reactions, and of
these we opted for electron-donating BAs with the assumption that
Cross-over experiments with isotopic substitution. The non- they may afford higher turnover frequencies based on their increased
homogeneous reaction conditions are limiting for kinetic stud- yields (Supplementary Table 15). This left us with the BAs on the lower
ies, so cross-over reactions of deuterated 3-chlorobenzoin with left node of the regression tree predicted at ∆∆G‡ =1.68kcalmol–1.
2-chlorobenzaldehyde were performed to test the influence of We selected 4-MePh-BA and, indeed, the e.r. at 20h was also 97.5:2.5,
BAs on the different steps in the proposed catalytic cycle (Fig. 5a). but the yield increased to 60%. Finally, we hypothesized that an addi-
2-chlorobenzaldehyde was selected because it was found to form tional 2-Me group would (1) not influence the torsion of the BA and
stable and less-reactive acyl anion adducts50, and therefore it was (2) not significantly increase the dipole moment in the direction of L
hypothesized that this experiment would produce predominantly or B5, and thus should not erode enantioselectivity. Yet, we expected
one cross-over product (Fig. 5a). Indeed, in the absence of BA, the an increase in the yield due to the addition of an electron-donating
reaction led to an equivalent amount of 3-chlorobenzoin and one group. Therefore, the reaction was run with 2,4-diMePh-BA, which
cross-over benzoin isomer, both with negligible enantioenrichment. led to an e.r. of 97:3 and a yield of 88% after 20h.
The 3-chlorobenzoin deuterium incorporation loss was significant
Based on the premise that dynamic covalent binding with BAs
and a slight deuterium uptake was observed for the cross-over could serve to modify the secondary sphere of organocatalysts, we
product (Supplementary Table 33 gives details). Based on our con- selected a system in which such interactions could be envisioned
trol experiments (Fig. 3c), it was postulated that the racemization near a catalytic site. As a proof-of-concept, we were able to regu-
and deuterium loss could arise from both the reversibility of the late the reproducibility and improve the selectivity of the benzoin
benzoin reaction and a base-catalysed process (Fig. 5b). To distin- condensation for a challenging substrate class that contained
guish between these two options, additional racemization control electron-withdrawing groups. Furthermore, the development of a
experiments were performed in which deuterated 3-chlorobenzoin machine-learning-based decision tree for the BA selection stream-
was subjected to reaction conditions in the absence of BA with and lined the optimization of a gram-scale benzoin reaction with the
without the NHC catalyst (Fig. 5c and Supplementary Table 36). lowest catalyst and base loading to date using an NHC organocata-
The enantioenrichment and deuterium loss was much more sig- lyst. We propose that the approach developed in this work could be
nificant in the presence of the NHC catalyst. This result supports applied to systematically modify and regulate the secondary sphere
the hypothesis that a major amount of deuterium loss stems from of numerous organocatalytic systems to give an increased stability
the complete reversibility of the benzoin reaction to release a pro- and reproducibility, an industrially relevant catalyst loading and a
tonated aldehyde, which can then undergo a nucleophilic attack tunable handle for controlling both reactivity and selectivity.
by another Breslow intermediate (Fig. 5b). To our surprise, in the
presence of BA no cross-over, deuterium loss or enantioenrichment
loss were observed (Fig. 5a). In combination with the racemization
experiments, these results shed light on the role of BA in mitigat-
ing racemization. We hypothesize that BA increases the energetic
barrier for the release of the protonated aldehyde in the presence of
NHC, probably by binding to the reactive intermediates and transi-
tion states under the reaction conditions. As the 2-chlorobenzal-
dehyde forms stable acyl anion adducts and the BA is presumed to
increase the barrier for the release of the aldehyde, it was expected
that an extremely stable acyl anion adduct could be formed in the
presence of BA. Accordingly, mass spectrometry analysis (matrix-
assisted laser desorption/ionization–time of flight) was consistent
with the formation of a stable 1:1:1 NHC:BA:2-chlorobenzaldehyde
adduct in the presence of two different BAs (Fig. 5b). These results
strongly suggest that BAs are not only involved in regulating the
selectivity and mitigating the base and NHC-induced racemization,
but also take part in the turnover determining step(s) of this reac-
tion by binding to NHC and its reactive intermediates. In addition,
experiments with a catalyst that does not contain a hydroxy group
were performed and support our hypothesis that this is not merely
cooperative catalysis, but rather a modification at the secondary
sphere (Supplementary Section 6.11). We are currently conducting
Methods
General experimental procedure for 6 and 24h benzoin reactions in the
presence of BAs. Enantioselective benzoin reactions were run in 8ml vials with
screw-cap septa and stirred at 500revolutions per minute (r.p.m.). A dry, argon-
fushed 8ml vial equipped with a magnetic stir bar was charged with Rb2CO3
(0.045mmol) and a 3Å molecular sieve (MS) (50mg) and dried at 200°C in the
oven overnight. Te vial was then heated with a heat gun under an argon fow for
5min and cooled to room temperature. Te NHC catalyst (0.066mmol) and aryl
BA (0.079mmol) were added and fushed with argon followed by the addition
of 1ml of anhydrous THF. Te reagents were allowed to stir for 10min at 20°C
under an argon atmosphere and then arylaldehyde (1.1mmol) was added. Te
reaction mixture was stirred at 20°C for 6 or 24h. Te reaction was quenched
with distilled acetic acid (30µl, 0.48mmol) and stirred for 10min. Te solvent
was removed under a nitrogen fow. To calculate the NMR yields, a stock solution
of 1,3,5-triisopropylbenzene (1mmol) in CHCl3 (10ml) was prepared and each
reaction mixture was dissolved in 1ml of the stock solution. An aliquot of the
dissolved reaction mixture (200µl) was used for HPLC enantioenrichment analysis
by fltering through silica gel (500mg) and then washing with CHCl3 (5×2ml).
Afer evaporating the solvent in vacuo, the crude product was dissolved in
isopropyl alcohol and the sample was ready for HPLC analysis.
General experimental procedure for 24h racemization reactions of
enantioenriched 3-chlorobenzoin with different BAs. A dry, argon-flushed 8ml
vial equipped with a magnetic stir bar was charged with Rb2CO3 (0.045mmol)
and a 3Å MS (50mg) and dried at 200°C in the oven overnight. The vial was
then heated with a heat gun under an argon flow for 5min and cooled to room
density functional theory and kinetic studies to further investigate temperature. Cat5F (0.066mmol) and aryl BA (0.079mmol) were added and flushed
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
with argon. The reactions were performed in batches of ten and enantioenriched
3-chlorobenzoin was weighed out to prepare 1.7g in 11ml of anhydrous THF,
which should amount to 0.55mmol in each 1ml. For each batch of reactions, the
effective amount (mmol) of 3-chlorobenzoin was determined either by HPLC or
NMR using 1,3,5-triisopropylbenzene as the internal standard. 3-chlorobenzoin
(1ml) was added into the reaction vail and the mixture was stirred (500rpm)
at 20°C for 24h. The reaction was quenched with distilled acetic acid (30µL,
0.48mmol) and stirred for 10minutes. The solvent was removed under nitrogen
flow. To calculate NMR yields a stock solution of 1,3,5-triisopropylbenzene
(1mmol) in CHCl3 (10mL) was prepared and each reaction mixture was dissolved
in 1mL of the stock solution. An aliquot of the dissolved reaction mixture (200µL)
was used for HPLC enantioenrichment analysis by filtering through silica gel
(500mg) then washing with CHCl3 (5×2mL). After evaporating the solvent, the
crude product was dissolved in isopropyl alcohol and the sample was ready for
HPLC analysis.
11. Ward, T. R. et al. Exploiting the second coordination sphere: proteins as host
for enantioselective catalysis. Chimia 57, 586–588 (2003).
12. Uraguchi, D., Ueki, Y. & Ooi, T. Chiral organic ion pair catalysts
assembled through a hydrogen-bonding network. Science 326,
120–123 (2009).
13. Meeuwissen, J. & Reek, J. N. H. Supramolecular catalysis beyond
enzyme mimics. Nat. Chem. 2, 615–621 (2010).
14. Leenders, S. H. A. M., Gramage-Doria, R., de Bruin, B. & Reek, J. N. H.
Transition metal catalysis in confned spaces. Chem. Soc. Rev. 44,
433–448 (2014).
15. Rowan, S. J., Cantrill, S. J. & Cousins, G. R. L. Dynamic covalent chemistry.
Angew. Chem. Int. Ed. 41, 898–952 (2002).
16. Bapat, A. P., Roy, D., Ray, J. G., Savin, D. A. & Sumerlin, B. S. Dynamic-
covalent macromolecular stars with boronic ester linkages.
J. Am. Chem. Soc. 133, 19832–19838 (2011).
17. Bull, S. D. et al. Exploiting the reversible covalent bonding of boronic acids:
recognition, sensing, and assembly. Acc. Chem. Res. 46, 312–326 (2013).
18. Wilson, A., Gasparini, G. & Matile, S. Functional systems with orthogonal
dynamic covalent bonds. Chem. Soc. Rev. 43, 1948–1962 (2014).
19. Schaufelberger, F. & Ramström, O. Dynamic covalent organocatalysts
discovered from catalytic systems through rapid deconvolution screening.
Chem. Eur. J. 21, 12735–12740 (2015).
General experimental procedure for crossover experiments. A dry, argon-
flushed 8mL vial equipped with a magnetic stir bar was charged with Rb2CO3
(0.045mmol) and 3Å MS (50mg) and dried at 200°C in the oven overnight. The
vial was then heated with a heat gun under an argon flow for 5min and cooled to
room temperature. Cat5F (0.066mmol) and aryl BA (0.099mmol) were added and
flushed with argon. Enantioenriched 3-chlorobenzoin-d (120mg, 0.42mmol, in
600µl of anhydrous THF) and 2-chlorobenzaldehyde (47µ, 0.42mmol) were added
into the reaction vail and the mixture was stirred (500r.p.m.) at 20°C for 24h. The
reaction was quenched with distilled acetic acid (30µl, 0.48mmol). An aliquot of
the reaction mixture (50µl) was diluted with 950µl of acetonitrile, filtered through a
0.45µm microfilter and used for HPLC product distribution analysis. The rest of the
reaction mixture was purified by silica gel column chromatography (EtOAc:hexane
20:80) to determine the enantioenrichment and deuterium incorporation.
20. Lascano, S. et al. Te third orthogonal dynamic covalent bond. Chem. Sci. 7,
4720–4724 (2016).
21. Seifert, H. M., Ramirez Trejo, K. & Anslyn, E. V. Four simultaneously
dynamic covalent reactions. Experimental proof of orthogonality.
J. Am. Chem. Soc 138, 10916–10924 (2016).
22. Zhou, Y., Li, L., Ye, H., Zhang, L. & You, L. Quantitative reactivity scales
for dynamic covalent and systems chemistry. J. Am. Chem. Soc. 138,
381–389 (2016).
General experimental procedure for large-scale reactions. A dry, argon-flushed
25ml round-bottom flask equipped with a magnetic stir bar was charged with
Rb2CO3 (0.033mmol) and a 3Å MS (30mg). The flask was then heated with a
heat gun under argon flow for 10min and cooled to room temperature. Cat5F
(0.050mmol) and aryl BA (0.075mmol) were added, followed by the addition
of 3-chlorobenzaldehyde (50mmol), 15ml of anhydrous THF and water (30µl).
The reaction mixture was stirred at 20°C for 20 or 40h. The reaction was
quenched with distilled acetic acid (4.5ml) and purified by column
chromatography (EtOAc:hexane 20:80).
23. Akgun, B. & Hall, D. G. Fast and tight boronate formation for click
bioorthogonal conjugation. Angew. Chem. Int. Ed. 55, 3909–3913 (2016).
24. Moulin, E., Cormos, G. & Giuseppone, N. Dynamic combinatorial chemistry
as a tool for the design of functional materials and devices. Chem. Soc. Rev. 41,
1031–1049 (2012).
25. Teichert, J. F., Mazunin, D. & Bode, J. W. Chemical sensing of polyols
with shapeshifing boronic acids as a self-contained sensor array.
J. Am. Chem. Soc. 135, 11314–11321 (2013).
26. Wong, C.-H. & Zimmerman, S. C. Orthogonality in organic, polymer,
and supramolecular chemistry: from Merrifeld to click chemistry.
Chem. Commun. 49, 1679–1695 (2013).
27. Wiskur, S. L. & Anslyn, E. V. Using a synthetic receptor to create an
optical-sensing ensemble for a class of analytes:a colorimetric assay for the
aging of scotch. J. Am. Chem. Soc. 123, 10109–10110 (2001).
28. Baragwanath, L., Rose, C. A., Zeitler, K. & Connon, S. J. Highly
enantioselective benzoin condensation reactions involving a bifunctional
protic pentafuorophenyl-substituted triazolium precatalyst. J. Org. Chem. 74,
9214–9217 (2009).
Data availability
All data generated or analysed during this study are available in this published
article and its Supplementary Information files, or from the corresponding author
upon request. Experimental procedures, results, characterization data, spreadsheets
of parameters used in the models and MATLAB scripts used for model
identification are accessible online as Supplementary Information.
Received: 22 June 2018; Accepted: 22 March 2019;
Published: xx xx xxxx
29. O’Toole, S. E. & Connon, S. J. Te enantioselective benzoin condensation
promoted by chiral triazolium precatalysts: stereochemical control via
hydrogen bonding. Org. Biomol. Chem. 7, 3584–3593 (2009).
30. Langdon, S. M., Legault, C. Y. & Gravel, M. Origin of chemoselectivity in
N-heterocyclic carbene catalyzed cross-benzoin reactions: DFT and
experimental insights. J. Org. Chem. 80, 3597–3610 (2015).
31. Maji, R. & Wheeler, S. E. in Aromatic interactions: Frontiers in Knowledge
and Application (eds Darren W Johnson, D. W. & Hof, F.) 18–38 (Te Royal
Society of Chemistry, Cambridge, 2017).
references
1. Warren, J. J., Lancaster, K. M., Richards, J. H. & Gray, H. B. Inner- and
outer-sphere metal coordination in blue copper proteins. J. Inorg. Biochem. 115,
119–126 (2012).
2. Werner, A. Zur kenntnis des asymmetrischen kobaltatoms.
Ber. Dtschn Chem. Ges. 45, 121–130 (1912).
3. Werner, A. Über die raumisomeren kobaltverbindungen.
Justus Liebig’s Ann. Chem. 386, 1–272 (1912).
32. Paul, M., Breugst, M., Neudorf, J. M., Sunoj, R. B. & Berkessel, A. Keto-enol
thermodynamics of Breslow intermediates. J. Am. Chem. Soc. 138,
5044–5051 (2016).
4. Cook, S. A. & Borovik, A. S. Molecular designs for controlling the local
environments around metal ions. Acc. Chem. Res. 48, 2407–2414 (2015).
5. Reedijk, J. Coordination chemistry beyond Werner: interplay between
hydrogen bonding and coordination. Chem. Soc. Rev. 42, 1776–1783 (2013).
6. Shook, R. L. & Borovik, A. S. Role of the secondary coordination sphere in
metal-mediated dioxygen activation. Inorg. Chem. 49, 3646–3660 (2010).
7. Baschieri, A., Bernardi, L., Ricci, A., Suresh, S. & Adamo, M. F. Catalytic
asymmetric conjugate addition of nitroalkanes to 4-nitro-5-styrylisoxazoles.
Angew. Chem. Int. Ed. 48, 9342–9345 (2009).
33. Flanigan, D. M., Romanov-Michailidis, F., White, N. A. & Rovis, T.
Organocatalytic reactions enabled by N-heterocyclic carbenes.
Chem. Rev. 115, 9307–9387 (2015).
34. Hansch, C., Leo, A. & Taf, R. W. A survey of Hammett substituent
constants and resonance and feld parameters. Chem. Rev. 91,
165–195 (1991).
35. Verloop, A., Hoogenstraaten, W. & Tipker, J. in Drug Design Vol 7
(ed. Ariens, E. J.) 165–207 (Academic, 1976).
36. Milo, A., Bess, E. N. & Sigman, M. S. Interrogating selectivity in catalysis
using molecular vibrations. Nature 507, 210–214 (2014).
37. Glendening, E. D., Landis, C. R. & Weinhold, F. Natural bond orbital
methods. WIREs Comput. Mol. Sci. 2, 1–42 (2011).
8. Kawai, H., Kusuda, A., Nakamura, S., Shiro, M. & Shibata, N. Catalytic
enantioselective trifuoromethylation of azomethine imines with
trimethyl(trifuoromethyl)silane. Angew. Chem. Int. Ed. 48, 6324–6327 (2009).
9. Nicolaou, K. C., Liu, G., Beabout, K., McCurry, M. D. & Shamoo, Y.
Asymmetric alkylation of anthrones, enantioselective total synthesis
of (−)- and (+)-viridicatumtoxins B and analogues thereof: absolute
confguration and potent antibacterial agents. J. Am. Chem. Soc. 139,
3736–3746 (2017).
38. Maji, R. & Wheeler, S. E. Importance of electrostatic efects in the
stereoselectivity of NHC-catalyzed kinetic resolutions. J. Am. Chem. Soc. 139,
12441–12449 (2017).
39. Massey, R. S., Collett, C. J., Lindsay, A. G., Smith, A. D. & O’Donoghue, A. C.
Proton transfer reactions of triazol-3-ylidenes: kinetic acidities and
carbon acid pKa values for twenty triazolium salts in aqueous solution.
J. Am. Chem. Soc. 134, 20421–20432 (2012).
10. Sahu, S. et al. Secondary coordination sphere infuence on the reactivity of
nonheme iron(ii) complexes: an experimental and DFT approach. J. Am.
Chem. Soc. 135, 10590–10593 (2013).
NAtUre CheMiStry | www.nature.com/naturechemistry

Nature Chemistry
Articles
50. Collett, C. J. et al. Rate and equilibrium constants for the addition of
40. Niu, Y. et al. Experimental and computational gas phase acidities of conjugate
acids of triazolylidene carbenes: rationalizing subtle electronic efects.
J. Am. Chem. Soc. 139, 14917–14930 (2017).
41. Companyó, X. & Burés, J. Distribution of catalytic species as an indicator
to overcome reproducibility problems. J. Am. Chem. Soc. 139,
8432–8435 (2017).
42. Enders, D. & Henseler, A. A direct intermolecular cross-benzoin type
reaction: N-heterocyclic carbene-catalyzed coupling of aromatic
aldehydes with trifuoromethyl ketones. Adv. Synth. Catal. 351,
1749–1752 (2009).
43. Enders, D. N-heterocyclic carbene catalysed asymmetric cross-benzoin
reactions of heteroaromatic aldehydes with trifuoromethyl ketones.
Chem. Commun. 46, 6282–6284 (2010).
44. Enders, D., Niemeier, O. & Henseler, A. Organocatalysis by N-heterocyclic
carbenes. Chem. Rev. 107, 5606–5655 (2007).
45. Renny, J. S., Tomasevich, L. L., Tallmadge, E. H. & Collum, D. B. Method of
continuous variations: applications of Job plots to the study of molecular
associations in organometallic chemistry. Angew. Chem. Int. Ed. 52,
11998–12013 (2013).
46. Tordarson, P. Determining association constants from titration experiments
in supramolecular chemistry. Chem. Soc. Rev. 40, 1305–1323 (2011).
47. Neel, A. J., Milo, A., Sigman, M. S. & Toste, F. D. Enantiodivergent
fuorination of allylic alcohols: data set design reveals structural interplay
between achiral directing group and chiral anion. J. Am. Chem. Soc. 138,
3863–3875 (2016).
N-heterocyclic carbenes into benzaldehydes: a remarkable 2-substituent efect.
Angew. Chem. Int. Ed. 127, 6991–6996 (2015).
Acknowledgements
This research was supported by the Israel Science Foundation (Grant no. 1193/17).
We thank M. Sigman, D. Toste, A. Neel and D. Pappo for fruitful discussions.
D.V. acknowledges the PBC for a postdoctoral fellowship. S.C.G. acknowledges the
Kreitman Graduate School for a postdoctoral fellowship. Z.A. acknowledges the
Kreitman Graduate School for the chemo-tech scholarship. Mass spectra measurements
were performed with the help of M. Shema-Mizrachi and M. M. Karpasas.
Author contributions
All the authors designed and performed the experiments and analysed the data. The
Supplementary Information was compiled by S.C.G. and D.V., the product distribution
by HPLC was performed by Z.A. and mathematical modelling was performed by A.M.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41557-019-0258-1.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to A.M.
48. Hall, D. G. Boronic Acids: Preparation and Applications in Organic Synthesis,
Medicine and Materials 1–133 (Wiley, Weinheim, 2011).
49. Martínez-Aguirre, M. A. & Yatsimirsky, A. K. Brønsted versus Lewis
acid type anion recognition by arylboronic acids. J. Org. Chem. 80,
4985–4993 (2015).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2019
NAtUre CheMiStry | www.nature.com/naturechemistry