Probing the Limits of Selectivity in a Recognition-Mediated Reaction Network Embedded within a Dynamic Covalent Library

Article abstract of DOI:10.1002/chem.201503740

Two recognition-mediated reaction processes operating through reactive binary complexes drive resolution of a 24-component dynamic covalent library, assembled from individual aldehydes and nucleophiles. The effectiveness of the library resolution and sele

Full text of DOI:10.1002/chem.201503740

DOI: 10.1002/chem.201503740
Full Paper
&
Combinatorial Chemistry
Probing the Limits of Selectivity in a Recognition-Mediated
Reaction Network Embedded within a Dynamic Covalent Library
[a]
[a, b]
[a]
Tamara Kosikova, Harry Mackenzie,
and Douglas Philp*
Abstract: Two recognition-mediated reaction processes op-
erating through reactive binary complexes drive resolution
of a 24-component dynamic covalent library, assembled
from individual aldehydes and nucleophiles. The effective-
ness of the library resolution and selective amplification of
one recognition-enabled species over another is limited by
the difference in the rates of the recognition-mediated react-
ive processes and the strength of the recognition processes
employed in the dynamic system.
Introduction
components under thermodynamic control. When a DCL is in-
structed by the addition of an external stimulus, the equilib-
rium distribution is altered and the library components sta-
bilised by the interaction with the stimulus are amplified. How-
ever, the purely thermodynamic selection limits the magnitude
of amplification that can be achieved within the DCL. In order
to attain increased selectivity within DCLs, we have to employ
methods that couple DCLs to irreversible processes and allow
the target to be removed from a pool of exchanging compo-
nents through kinetic selection. In the past, several successful
examples, demonstrating target amplification using both
[
1]
Complexity in the world around us does not simply develop
as a function of the increasing number of components within
a given system. Instead, it emerges within systems that possess
the ability to dynamically self-organise and in which feedback
between components is present. The capacity of individual
components within a system to interact allows these individual
elements, with established properties, to come together and
form constructs of higher order, giving rise to emergent behav-
[2]
iour and properties. Simultaneously, the interconnectedness
of all components enables complex systems to constantly
adapt and evolve in response to environmental stimuli. Our in-
ability to control and predict the behaviour of complex sys-
tems is an essential feature that allows us to differentiate them
from systems that are merely complicated, such as machines
or engines, where components interact in a controlled and
[
6]
[7]
chemical reactions and physical processes, have been re-
[
8]
ported. Recently, we have reported the use of a single, recog-
nition-mediated irreversible 1,3-dipolar cycloaddition reaction
as the means of selecting and amplifying a specific target from
within a dynamic library composed of interconverting nitrones
and imines. Specifically, we explored two different kinetic
modes for achieving DCL resolution (Figure 1), driven by:
[3]
predictable manner. Only by exploring the role of connectiv-
ity and the fundamental relationship between the components
of interconnected networks can we begin to develop an
understanding of how complex systems may have played
a role in the transition from simple compounds to the first
self-sustaining system.
[
8b]
1) formation of a reactive binary complex (AB, Figure 1), or
[
8a]
2) template-mediated self-replication
(SR, Figure 1). In this
work, we examine the factors that govern the selectivity in
a DCL, assembled from individual aldehydes and nucleophiles,
incorporating a reaction network involving not one, but two
recognition-mediated processes operating through a reactive
binary complex. With specific focus on this mode of selection
within the DCL, we demonstrate how changing the strength of
recognition processes within the system influences the effect-
iveness of resolution within the library.
The study of a complex chemical network begins with its
design. One method that has been shown to be particularly ef-
fective in generating complex interconnected systems is dy-
[4]
namic covalent chemistry (DCC). The general combinatorial
approach and reversible bond formation allow facile construc-
[5]
tion of dynamic covalent libraries (DCLs) of interconverting
Results and Discussion
[
a] T. Kosikova, H. Mackenzie, Prof. D. Philp
School of Chemistry and EaStCHEM, University of St. Andrews
North Haugh, St. Andrews, KY16 9ST (UK)
E-mail: d.philp@st-andrews.ac.uk
Our dynamic covalent library (Figure 2) is based on a 4”4
matrix of aldehydes, A to D, and nucleophiles, W to Z, com-
prised of three amines and one hydroxylamine, which afford
an exchange pool of 12 imines and 4 nitrones. The total of 24
interconverting components represents a dynamic system that
[
b] H. Mackenzie
UCB, 216 Bath Road, Slough, Berks, SL1 3WE (UK)
19
can be analysed readily by F NMR spectroscopy—the pres-
Supporting information and ORCID(s) from the author(s) for this article are
available on the WWW under http://dx.doi.org/10.1002/chem.201503740.
ence of either an aryl-F or trifluoromethyl (ÀCF ) tag on each
3
Chem. Eur. J. 2016, 22, 1831 – 1839
1831
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
The exchange pool
Initially, we needed to confirm that our library forms a dynamic
exchange pool. Determination of the equilibrium distribution
of our DCL in the absence of any recognition-mediated
reactions will enable us to quantify the magnitude of perturba-
tion of the DCL composition upon addition of instructing
maleimide. In a typical experiment, an equimolar solution of al-
dehydes and nucleophiles ([A] to [D] and [W] to [Z]=20 mm)
was prepared in CD Cl2 saturated with para-toluenesulfonic
2
acid monohydrate (PTSA). The solution was incubated at 273 K
and the product distribution was determined by quantitative
1
9
1
Figure 1. Resolution of a dynamic covalent library (DCL) depends on the
composition of the library and the mode of resolution employed. AB de-
notes a resolution mode driven by reaction of library members in a recogni-
tion-mediated reaction operating through a binary reactive complex. SR de-
notes a resolution mode driven by reaction of library members in a recogni-
tion-mediated autocatalytic reaction controlled by a self-complementary
template.
F{ H} NMR spectroscopy (282.4 MHz) after five days (for de-
tails, see the Supporting Information). However, as a result of
insensitivity of the ÀCF group to changes in its surrounding
3
electronic environment, we were unable to resolve individual
resonances for the Y-imines bearing the trifluoromethyl tag.
The distribution of condensation products in the library ex-
change pool is presented in Figure 4. The equilibrium position
for the formation of nitrones from hydroxylamine Z lies far to-
wards the product side, with almost complete conversion
(>99%) to the corresponding nitrones. The two recognition-
enabled nitrones, AZ and BZ, are present at a higher concen-
tration than non-recognition nitrones CZ and DZ. Interestingly,
the more electron rich pyridine ring present in aldehyde A
makes it less reactive towards nucleophiles than aldehyde B,
resulting in a slightly higher concentration of BZ (6.72 mm)
than AZ (6.39 mm).
nucleophilic component facilitating the identification of indi-
vidual components in the library.
The library is designed to contain components equipped
with two distinct features: a reactive site and a recognition
element. Formation of a reactive nitrone is provided by reac-
tion of aldehydes with hydroxylamine Z. The library also con-
tains two aldehydes bearing a site capable of recognising
a carboxylic acid: A (4,6-dimethyl amidopyridine) and B (6-
methyl amido-pyridine). Overall, the library exchange pool con-
[8]
tains only two components, AZ and BZ, equipped with both
a reactive and a recognition element, required for a recogni-
tion-mediated irreversible 1,3-dipolar cycloaddition reaction
Kinetic selection in the absence of the DCL
Prior to instructing the DCL through addition of maleimide M,
we needed to rule out inherently different reactivities of the
two nitrones towards maleimides as a potential factor con-
tributing to the selectivity for the formation of any one cyclo-
adduct. In this respect, we constructed a series of kinetic ex-
periments (Figure 5) that would establish the reactivity of both
AZ and BZ nitrones in the absence of any dynamic compo-
nents. Firstly, the individual reactions of each nitrone with a rec-
ognition-disabled maleimide Mdis were examined. The carb-
oxylic acid on Mdis is protected as the methyl ester, rendering
it incapable of recognition-mediated interactions and by exam-
ining its reaction with the two recognition nitrones, we can es-
tablish the rate constants and diastereoselectivity associated
with the bimolecular pathways.
[
8b]
with a maleimide (M). The reactions proceed by formation
of the [AZ·M] or [BZ·M] binary complexes (Figure 3) afforded
by the recognition of the carboxylic acid bearing maleimide by
the amidopyridine recognition site, thus facilitating selective
formation of cis cycloadducts.
In order for selective amplification of recognition-enabled
species from within a DCL to take place, the irreversible cyclo-
addition must accomplish the formation of cis-AZ and cis-BZ
fuelled only by the building blocks distributed amongst the
pool of simple components. Whilst both AZ and BZ are cap-
able of molecular recognition with a carboxylic acid, the
strength of their association constants are expected to be dif-
ferent as a result of the changes in the electronic environment
of the pyridine moieties. The pyridine unit on AZ contains an
additional methyl group, increasing the electron density on
the pyridine ring, and therefore, the strength of the association
with the maleimide ÀCOOH. We estimated the association con-
stants for the formation of [AZ·M] and [BZ·M] binary com-
plexes, based on a Ka measured for a structurally similar
Secondly, analysis of the reaction of AZ and BZ with a recog-
nition-enabled maleimide M allows us to determine the rate
and selectivity enhancement gained through recognition in
a purely kinetic scenario. Thirdly, probing the reactivity of both
nitrones under competition conditions, in the presence of
either recognition-disabled maleimide Mdis or recognition-en-
abled maleimide M provides us with a direct measurement of
selectivity afforded through kinetic selection only. These three
isolated kinetic experiments will, ultimately, allow us to evalu-
ate the extent of contribution of dynamic selection to the
overall selectivity in the recognition-network. We employed
three experiments in order to probe the selectivity within
À1
phenylacetic acid, to be 315 and 160m , respectively (see the
Supporting Information). The two remaining nitrones, CZ and
DZ, which do not contain a recognition element, can also
react with maleimide M, however, only through the slow and
unselective bimolecular reaction, affording both the trans and
cis diastereoisomeric products, typically in a 3:1 ratio.
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1832
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 2. Dynamic covalent library composed of four aldehydes, A to D and four nucleophiles, W to Z. Corresponding exchange pool components and cyclo-
adducts, formed upon reaction with recognition-enabled maleimide M and recognition-disabled Mdis, containing aldehyde A are in black, aldehyde B are in
grey, aldehyde C and aldehyde D are white.
a DCL. Analysis of the exchange pool distribution in the ab-
sence of any maleimide was described in Figure 4. Two add-
itional DCL experiments were designed to complement the
competition kinetics performed with Mdis and M maleimides in
the absence of DCL, and are presented in a later section.
In each kinetic experiment, a sample containing the
nitrone(s) and the maleimide at 20 mm was prepared in CD Cl2
2
saturated with PTSA and the product evolution was monitored
1
19
1
at regular intervals by H NMR (500.1 MHz) or F{ H} NMR
(470.3 MHz) spectroscopy at 273 K (for details, see the Support-
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1833
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 3. The cis selective 1,3-dipolar cycloaddition between nitrones AZ
and BZ with maleimide M proceeds through the recognition-mediated for-
mation of [AZ·M] and [B·M] reactive binary complexes.
Figure 5. Flow chart outlining a series of experiments designed to probe the
origins of selectivity in a recognition-mediated network embedded in a dy-
namic covalent library. Both the DCL and the irreversible reactions have to
be examined in isolation and in a competition scenario in the presence of
recognition-disabled maleimide Mdis and recognition-enabled maleimide M
(grey box).
maleimide M clearly demonstrated the effect of enabling mo-
lecular recognition, revealing conversion to cycloadducts of
>
85% after 15 h and more than 40:1 selectivity for the cis dia-
stereoisomer. Concentration–time data for the four isolated ki-
netic experiments were fitted to the appropriate kinetic model
(
see the Supporting Information), constructed to encompass
all the interactions in each system as well as the estimated
association constants, thus allowing us to extract kinetic pa-
rameters for the individual reactions of M and Mdis with nitro-
nes AZ or BZ (Table 1).
Employing the fitting procedure (see the Supporting Infor-
mation), we were able to determine the bimolecular and rec-
ognition-mediated rate constants, as well as the respective ef-
fective molarities (EMs) associated with AZ and BZ nitrones.
Nevertheless, when these extracted parameters were em-
ployed in a kinetic model simulating the competition reaction
between AZ and BZ with maleimide M, we observed a dis-
agreement between the estimated K values used in the fitting
a
of individual kinetics and the K values required for achieving
a
the best fit between our model and the competition kinetic
data. Such difference suggests that system-level interactions,
arising from the transfer of the individual recognition-mediated
processes to an environment enforcing competition for
a shared building block, exerts an unforeseen effect on the ap-
Figure 4. a) Distribution of the DCL exchange pool after five days at 273 K,
as determined by b) F{ H} NMR spectroscopy (282.4 MHz), relative to 1-
19
1
parent K values for the two recognition processes. Namely,
a
the fitted association constants, specific to the aldehydes and
maleimide employed in our system under competition condi-
fluoro-4-nitrobenzne as the internal standard ([A] to [D] and [W] to
[
Z]=20 mm in CD
drate). The resonances for Y-imines could not be resolved as a result of the
insensitivity of the ÀCF tag to changes in its electronic environment.
2 2
Cl saturated with para-toluenesulfonic acid monohy-
À1
tions, were determined as 220 and 170m for [AZ·M] and
3
[
BZ·M], respectively (ratio of 1.3:1).
The difference in relative efficiencies of the two recognition
ing Information). Kinetic experiments with recognition-disabled
maleimide Mdis confirmed that, in the absence of recognition,
the cycloaddition reactions are slow and unselective, resulting
in less than 20% conversion to cycloadducts after 15 h, with
cis/trans diastereoisomeric ratio of approximately 1:2.3 in
each experiment. The outcome of analogous reactions with
processes manifested itself during the competition for a limited
supply of maleimide M (Figure 6a). Both recognition-enabled
nitrones, AZ and BZ, exhibited similar reaction efficiency when
instructed with M, reaching 52 and 44% conversion to recog-
nition-enabled products cis-AZ and cis-BZ (ratio of 1.2:1) after
15 h, respectively (Figure 6b).
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1834
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Table 1. Kinetic and association constants determined for the reaction of AZ and BZ with M and Mdis by kinetic simulation and fitting of experimental
data using the SimFit program.
trans-AZ
cis-AZ
trans-BZ
cis-BZ
À1 À1
À4
À5
À4
À4
À5
À4
bimolecular rate constant [m
s
]
1.24”10
–
–
315
5.57”10
1.24”10
2.23
1.30”10
–
–
160
5.93”10
1.18”10
1.99
À1
recognition-mediated rate constant [s
]
effective molarity [m]
association constant K
À1
a
[m
]
315
160
Figure 6. a) The 1,3 dipolar cycloaddition between nitrone AZ and BZ and recognition-enabled maleimide M gives rise to sets of two diastereoisomeric prod-
ucts: i) trans-AZ and cis-AZ, and ii) trans-BZ and cis-BZ. b) Concentration versus time plot for this reaction ([AZ]=[BZ]=[M]=20 mm, CD Cl saturated with
2
2
19
1
para-toluenesulfonic acid monohydrate, 08C) as determined by F{ H} NMR (470.3 MHz) spectroscopy. The data points for the formation of trans-AZ closely
mirror the data for trans-BZ, making it difficult to observe the trans-AZ data points on the plot.
Fitting of the concentration versus time data revealed that
the enhancement in the recognition-mediated reaction relative
to bimolecular rate constant (effective molarity, EM) is largely
similar for both reactions, with an EM ratio of 1.1:1 for AZ/BZ.
Therefore, the uneven population of cis-AZ and cis-BZ is pre-
dominantly influenced by the different affinities of AZ and BZ
for M, with smaller contribution exerted also by the individual
variations in reactivity. We envisaged that competition for
a shared building block might present a different outcome
when the reaction network is embedded in a dynamic library.
More specifically, we hoped that the absence of preformed
nitrones would allow the added maleimide to selectively pull
material from the exchange pool and form the product with
a higher association constant preferentially, amplifying the dif-
ference between the two cis products in the processes.
and it was clear that simply incorporating an irreversible reac-
tion into a DCL did not afford selectivity for the recognition
species.
We next examined the effect of recognition-enabled
maleimide M on the library distribution and product selectivity.
An equimolar solution containing all reaction components ([A]
to [D] and [W] to [Z]=[M]=20 mm) was prepared in CD Cl
2
2
19
1
saturated with PTSA and the solution examined by F{ H} NMR
spectroscopy (282.4 MHz) after five days at 273 K. Addition of
the recognition-enabled maleimide M had a profound effect
on the distribution of the product pool (Figure 7b). Recogni-
tion-enabled cycloadducts cis-AZ and cis-BZ are present in
a ratio of 1.1:1, at a significantly higher concentration than any
other product, with 45 and 41% conversion respectively.
The higher K value for the recognition processes in cis-AZ
a
manifested itself in the marginally higher proportion of this
product than that of cis-BZ. The effect of introducing recogni-
tion was also immediately noticeable within the exchange
pool. The library components equipped with the recognition
site are significantly depleted in response to the decreasing
amounts of AZ and BZ nitrones. The release of the W and X
exchange pool components from their A- and B-imines results
in increased concentrations of the C- and D-imines, incorporat-
ing these components. Figure 7c shows the magnitude of the
recognition effect on the DCL distribution, highlighting the dif-
ference between the concentrations of exchange pool in the
recognition-enabled scenario relative to exchange in the ab-
sence of any maleimide. The increased concentration of non-
recognition imines illustrates how the connectivity of a network
can allow system-wide effects to emerge.
Recognition-mediated resolution
In order to ascertain that simply performing the cycloaddition
reaction in the presence of a DCL does not generate any sig-
nificant changes in composition, the DCL was first addressed
with a recognition-disabled maleimide M . Under recognition-
disabled conditions, all four nitrones can only react through
dis
a slow, bimolecular reaction. A library sample ([A] to [D] and
[
W] to [Z]=[M ]=20 mm, CD Cl saturated with PTSA) was
dis
1
2
2
19
analysed by F{ H} NMR spectroscopy (282.4 MHz) after five
days at 273 K (Figure 7a). The exchange pool distribution re-
mained largely unaffected by the slow and unselective cyclo-
addition reactions, with only 40% overall conversion to all 8
cycloadducts. The four nitrones exhibited similar reactivities,
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1835
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 7. The DCL was instructed with: a) recognition-disabled maleimide Mdis, and b) recognition-enabled maleimide M. Distribution of the libraries was
19
1
examined by F{ H} NMR spectroscopy (282.4 MHz) after five days at 273 K, relative to 1-fluoro-4-nitrobenzne as the internal standard ([A] to [D] and [W] to
Z] and [M] or [Mdis]=20 mm in CD Cl saturated with para-toluenesulfonic acid monohydrate). Resonances for Y-imines (in all experiments) and for cis-CZ:
[
2
2
b) could not be unambiguously assigned and their concentrations were not determined. c) The change in exchange pool concentration in a DCL instructed
with M relative to the uninstructed exchange pool experiment after five days.
Origins of limited selectivity
components were removed selectively from the mixture of
interconnected components despite the fact that they were
present as minor components. Employing dynamic conditions
together with recognition-mediated irreversible reaction re-
vealed that the small difference in the strength of association
constants cannot be amplified beyond the limit imposed by
kinetic selection.
Selectivity in a system with two simultaneous and irreversible
kinetically driven processes is inherently limited by the reaction
conditions and the kinetic parameters governing these pro-
cesses. The isolated kinetic experiment, where AZ and BZ
nitrones competed for a limited amount of recognition
maleimide M, revealed that the more strongly binding amido-
pyridine unit allows AZ to associate and react to a greater
extent than BZ nitrone. The final ratio in this experiment was
approximately 1.2:1 in favour of cis-AZ. We hoped that the
ratio would be amplified upon transferring the competing rec-
ognition-mediated reaction network to a dynamic environ-
ment. However, embedding the network within a DCL showed
a negligible effect on the formation of these two products and
selectivity remained essentially unchanged (Figure 8).
In order to probe the origins of the limited selectivity in our
dynamic recognition-mediated network, we constructed a kin-
etic model incorporating all of the reactions leading to the for-
mation of exchange pool products (for details, see the Sup-
porting Information). Rather than individually measuring rate
constants for each reaction, we decided to identify trends in
the reactivity of each component from the distribution of the
exchange pool in isolation. Simulation of the exchange pool in
the absence of any maleimide (for details, see the Supporting
Information) showed a good agreement with the experimental
data (Figure 4) presented already. Utilising our simulation, we
were able to vary the rates of exchange reactions with respect
to the rate of cycloaddition, setting them to be much higher,
very similar or significantly lower than the rate of cycloaddition
reaction. Comparison of the simulated DCL distribution to the
experimentally determined results revealed that our system be-
haves akin to a fast exchange simulation (for details, see the
Supporting Information). The rates of exchange reactions have
a specific effect on the selectivity for recognition products, as
well as on the ratio of cis-AZ/cis-BZ within the library. Any im-
balance in the concentrations of AZ and BZ generated through
the irreversible reactions is erased by exchange with other li-
brary members containing the recognition and reactive mater-
ials (A, B and Z), thus allowing the recognition-mediated reac-
tion to proceed efficiently. In our library, the starting compo-
nents are present at 20 mm, a concentration considerably
While the dynamic selection did not provide the desired in-
crease in selectivity, the recognition events exerted a dramatic
effect on the product distribution (Figure 7b vs. a) in the li-
brary and facilitated selection of species equipped with recog-
nition elements effectively. AZ, BZ, A, B and Z exchange pool
above the K for both competing association processes. The
Figure 8. Concentration of cis-AZ (black) and cis-BZ (grey) formed under kin-
etic conditions after 15 h (cis-AZ/cis-BZ=1.2:1) and within the environment
of DCL after five days (cis-AZ/cis-BZ=1.1:1). The area of the circle denoted
by the dotted line represents 100% conversion of maleimide M, whereas
the relative areas of the pie charts are scaled to represent the actual conver-
sion to recognition-enabled products under kinetic selection (96%) and kin-
etic and dynamic selection (85%).
d
fast exchange reactions allow a significant concentration of
both nitrones to build up, enabling both recognition processes
to operate efficiently. In this respect, the dynamic scenario be-
comes similar to a simple kinetic experiment. We envisaged
that operating the system at concentrations closer to or below
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1836
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 9. Outcome of simulations examining the effect of initial concentration on: a) ratio of cis-AZ/cis-BZ, b,c) conversion to all cycloadducts and percentage
of the recognition-enabled products in the product pool, in reaction network under kinetic selection and within a dynamic environment after five days, re-
spectively. Four sets of conditions with AZ/BZ processes at different K
ing to the fitted parameters determined for our experimental system, Conditions II to IV simulated either a higher ratio of EM (II), K
a
and EM ratios were examined: Condition I employed the K
a
and EM ratio correspond-
(III) or both EM and K
values (IV). Simulations were performed using the SimFit software package. In each condition, grey rectangles represent the concentration range that af-
a
a
[9]
fords the best selectivity for overall formation of cycloadducts and the highest percentage of recognition-enabled species in the cycloadduct pool (N.B.: in all
cases, the x-axis is represented in logarithmic scale).
the K for the formation of both binary complexes, [AZ·M] and
ing initial conditions and reaction parameters on the ratio of
cis-AZ/cis-BZ formed, the overall conversion to cycloadducts
and percentage of recognition-enabled products (cis-AZ and
cis-BZ) present in the product pool. Condition I employed the
Ka and EM ratio corresponding to the fitted parameters deter-
mined for our experimental system. Conditions II to IV simulat-
ed either a higher ratio of EM (II), K (III) or both EM and K
values (IV) for the AZ/BZ recognition processes. These simula-
tions allowed us to elucidate the effect of varying kinetic par-
ameters at different concentrations on the ratio of cis-AZ/cis-
BZ (Figure 9a) and conversion to cycloadducts after a specific
time (5 days) in a kinetic scenario (Figure 9b) and in the envir-
onment of a DCL (Figure 9c).
d
[
BZ·M], would enable the initial imbalance towards one recog-
nition product to be propagated. In this respect, we set out to
investigate what conditions promote selectivity in a reaction
network embedded in a DCL and whether the same conditions
have a comparable influence on the network under simple kin-
etic selection. In order to do so, we examined the effect of
varying three variables: 1) initial concentration, 2) ratio of asso-
ciation constants for the formation of [AZ·M] and [BZ·M] and
a
a
3
) ratio of effective molarities for reactions of AZ and BZ
nitrones. These three variables were examined under four dif-
ferent simulation conditions, I to IV (Figure 9).
Each simulation examined a different ratio of K values and
a
EM ratios at ten different initial concentrations, ranging from
Condition I enabled us to determine what influence different
initial concentration would have on our experimental system.
0
.05 to 200 mm. We focused on examining the effect of chang-
1837
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
At 20 mm initial concentration (Condition I), we found the
simulated ratio of cis-AZ/cis-BZ within a DCL to be very similar
to the experimentally determined values for dynamic selection,
confirming it to be an accurate model of our dynamic library
system. The simulated ratio of cis-AZ/cis-BZ is lower in the DCL
selection than in the kinetic selection, and continues to fall
with decreasing concentration of starting materials (for details,
see the Supporting Information). This observation, while
counterintuitive, can be attributed to the recognition nitrone
AZ being present at lower concentrations than the nitrone BZ
in the library. As mentioned previously, the higher electron
density around the pyridine ring in BZ nitrone makes the alde-
hyde component B less reactive towards hydroxylamine Z. This
effect becomes more pronounced at lower initial concentra-
tions (for details, see the Supporting Information), effectively
eroding the advantage available to AZ nitrone as a result of its
achieved when the selection occurs from the DCL. However,
the optimum concentration ranges for selectivity (grey zones,
Figure 9) are significantly narrower when selection occurs from
the DCL as opposed to purely kinetically. The narrowing of the
optimum concentration window is proportional to the number
of compounds present within the DCL—the larger the library,
the narrower the window will be.
Ultimately, while the simulations here allow us to determine
the effect of different recognition and reaction parameters on
the selectivity generated in this system by two reactive binary
complexes within a DCL, nonlinear reaction processes, such as
autocatalysis, are an obvious extension and implementing
these processes within DCLs may, in principle, generate more
dramatic selectivity for a single product, this expectation
awaits experimental testing.
higher K value. A similar effect on the cis-AZ/cis-BZ ratio was
not observed in conditions II to IV, likely because the increas-
a
Conclusions
ing K and EM ratios compensate for the lower reactivity. In-
a
A simple recognition event has a profound effect on the distri-
bution of a dynamic library and can be used to amplify chem-
ical species from within a number of interconverting compo-
nents. Nevertheless, the rapid exchange reactions within the
dynamic library and the experimental conditions employed
enable both recognition-mediated reaction processes to oper-
ate efficiently, thereby allowing kinetic selection to prevail.
Through simulations and kinetic fitting, we were able to de-
velop an understanding of the intricate interconnected net-
work and formulate a set of rules that govern the selectivity
within the dynamic system and how it compares to selectivity
in a purely kinetic scenario. Finally, we have demonstrated that
creasing either the K or EM ratio (Condition II and III) revealed
a
that it is possible for a lower concentration to have the desired
effect on the selectivity within a system, if the K or EM ratios
a
for the two competing processes are larger than those ex-
pressed by AZ and BZ in our experimental system. In fact,
analysis of Condition IV revealed that a concurrent increase in
both K and EM ratio affords an even more significant increase
a
in selectivity for one recognition cycloadduct in the system.
Nevertheless, as observed experimentally for nitrone AZ, we
know from experience that increasing the strength of recogni-
tion on element A can potentially also affect the reactivity of
this component towards nucleophiles and its ability to ex-
change within the library. However, as we have no means of
predicting the extent to which the aldehyde reactivity might
be affected, we cannot account for it in the simulation and the
rates of exchange reactions are kept constant throughout the
simulations. We further examined the influence of simulated
initial conditions on the overall conversion to cycloadducts
and the percentage of recognition-enabled species in the
product pool. The simulations showed that there is an opti-
mum window of concentration for obtaining the best selectiv-
ity for recognition products within each set of conditions. Fur-
thermore, we found that selectivity for products formed
through recognition-mediated reactions is affected more
strongly by decreasing concentration within a DCL as com-
pared to under kinetic selection. A drop in selectivity for recog-
nition species is also observed at higher concentrations as
a result of increased contribution of the bimolecular pathway
working at concentrations below the K values of the recogni-
d
tion processes can positively influence the difference in se-
lectivity between two amplified species. However, in order to
induce significant selectivity in a recognition-mediated system,
the difference in the association constants and the effective
molarities of the employed recognition and reactive elements
must be much greater than in our current experimental
system. The results presented here suggest that the successful
resolution of DCLs, where multiple recognition-mediated path-
ways are open to the system, may be impractical. Therefore,
resolution of DCLs using irreversible recognition-mediated ap-
proaches may have to rely on the interplay of multiple kinetic
selection modes, for example, binary reactive complexes and
autocatalytic templates, in order to be effective. This strategy
is currently under investigation in our laboratory.
to the overall product pool. However, this effect is more no- Acknowledgements
ticeable in conditions employing the rather low experimental
EM ratio.
These kinetic simulations demonstrate that a clear set of
This work was supported by the award of a Postgraduate Stu-
dentship from EPSRC (EP/K503162/1) to T.K. The research data
supporting this publication can be accessed at http://
dx.doi.org/10.17630/2a427a7e-d025-492e-9301-377e18459060.
rules exists that describes library performance. A substantial in-
crease in selectivity for cis-AZ over cis-BZ can only be achieved
under Condition IV. In this situation, both the ratio of associ-
ation constants and the ratio of effective molarities for the two
recognition-mediated processes within the reaction network
embedded are 10. In these circumstances, better selectivity is
Keywords: combinatorial chemistry · cycloaddition reaction ·
kinetics · molecular recognition · NMR spectroscopy
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1838
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
[
[
1] a) J.-M. Lehn, Angew. Chem. Int. Ed. 2015, 54, 3276; b) J.-M. Lehn, Angew.
Chem. Int. Ed. 2013, 52, 2836; Angew. Chem. 2013, 125, 2906; c) R. J.
Sarma, J. R. Nitschke, Angew. Chem. Int. Ed. 2008, 47, 377; Angew. Chem.
R. Rodriguez, J. K. M. Sanders, S. Balasubramanian, Angew. Chem. Int. Ed.
2008, 47, 2677; Angew. Chem. 2008, 120, 2717; i) M. Chung, C. M. Heb-
ling, J. W. Jorgenson, K. Severin, S. J. Lee, M. R. Gagne, J. Am. Chem. Soc.
2008, 130, 11819; j) P. C. Gareiss, K. Sobczak, B. R. Mcnaughton, P. B.
Palde, C. A. Thornton, B. L. Miller, J. Am. Chem. Soc. 2008, 130, 16254;
k) O. Sijbren, S. Kay, Top. Curr. Chem. 2007, 277, 267; l) R. M. Bennes, D.
Philp, Org. Lett. 2006, 8, 3651; m) B. Brisig, J. K. M. Sanders, S. Otto,
Angew. Chem. Int. Ed. 2003, 42, 1270; Angew. Chem. 2003, 115, 1308;
n) O. Ramstrçm, J.-M. Lehn, Nat. Rev. Drug Discovery 2002, 1, 26;
o) R. L. E. Furlan, G. R. L. Cousins, J. K. M. Sanders, Chem. Commun. 2000,
1761.
2
008, 120, 383; d) V. Campbell, J. Nitschke, Synlett 2008, 20, 3077; e) T.
Vicsek, Nature 2002, 418, 131; f) S. H. Strogatz, Nature 2001, 410, 268;
g) G. M. Whitesides, R. F. Ismagilov, Science 1999, 284, 89; h) I. R. Epstein,
K. Showalter, J. Phys. Chem. 1996, 100, 13132.
2] a) A. G. Salles, S. Zarra, R. M. Turner, J. R. Nitschke, J. Am. Chem. Soc.
2
013, 135, 19143; b) N. Giuseppone, Acc. Chem. Res. 2012, 45, 2178; c) A.
Vidonne, D. Philp, Eur. J. Org. Chem. 2009, 593; d) P. T. Corbett, J. K. M.
Sanders, S. Otto, Angew. Chem. Int. Ed. 2007, 46, 8858; Angew. Chem.
2
007, 119, 9014; e) R. Eelkema, M. M. Pollard, J. Vicario, N. Katsonis, B. S.
Ramon, C. W. M. Bastiaansen, D. J. Broer, B. L. Feringa, Nature 2006, 440,
63; f) G. Ashkenasy, R. Jagasia, M. Yadav, M. R. Ghadiri, Proc. Natl. Acad.
[6] a) Y. Zhang, O. Ramstrçm, Chem. Eur. J. 2014, 20, 3288; b) L. Hu, O. Ram-
strçm, Chem. Commun. 2014, 50, 3792; c) Y. Zhang, L. Hu, O. Ramstrçm,
Chem. Commun. 2013, 49, 1805; d) M. Sakulsombat, Y. Zhang, O. Ram-
strçm, Top. Curr. Chem. 2011, 322, 55; e) V. del Amo, D. Philp, Chem. Eur.
J. 2010, 16, 13304; f) P. Vongvilai, O. Ramstrçm, J. Am. Chem. Soc. 2009,
131, 14419; g) M. Angelin, P. Vongvilai, A. Fischer, O. Ramstrçm, Chem.
Commun. 2008, 46, 768; h) P. Vongvilai, R. Larsson, O. Ramstrçm, Adv.
Synth. Catal. 2008, 350, 448; i) P. Vongvilai, M. Angelin, R. Larsson, O.
Ramstrçm, Angew. Chem. Int. Ed. 2007, 46, 948; Angew. Chem. 2007, 119,
966; j) J. D. Cheeseman, A. D. Corbett, R. Shu, J. Croteau, J. L. Gleason,
R. J. Kazlauskas, J. Am. Chem. Soc. 2002, 124, 5692.
1
Sci. USA 2004, 101, 10872.
[
[
3] a) K. Ruiz-Mirazo, C. Briones, A. de La Escosura, Chem. Rev. 2014, 114, 285;
b) J. J. Peyralans, S. Otto, Curr. Opin. Chem. Biol. 2009, 13, 705; c) J. R.
Nitschke, Nature 2009, 462, 736; d) R. F. Ludlow, S. Otto, Chem. Soc. Rev.
2
4
008, 37, 101; e) J. Stankiewicz, L. H. Eckardt, Angew. Chem. Int. Ed. 2006,
5, 342; Angew. Chem. 2006, 118, 350; f) M. Kindermann, I. Stahl, M. Rei-
mold, W. M. Pankau, G. von Kiedrowski, Angew. Chem. Int. Ed. 2005, 44,
750; Angew. Chem. 2005, 117, 6908.
6
4] a) A. Herrmann, Chem. Soc. Rev. 2014, 43, 1899; b) Y. Jin, C. Yu, R. J.
Denman, W. Zhang, Chem. Soc. Rev. 2013, 42, 6634; c) E. Moulin, G.
Cormos, N. Giuseppone, Chem. Soc. Rev. 2012, 41, 1031; d) M. E. Belo-
wich, J. F. Stoddart, Chem. Soc. Rev. 2012, 41, 2003; e) P. T. Corbett, J. Lec-
laire, L. Vial, K. R. West, J.-L. Wietor, J. K. M. Sanders, S. Otto, Chem. Rev.
[7] a) J. M. A. Carnall, C. A. Waudby, A. M. Belenguer, M. C. A. Stuart, J. J.-P.
Peyralans, S. Otto, Science 2010, 327, 1502; b) M. Barboiu, F. Dumitru, Y.-
M. Legrand, E. Petit, A. van der Lee, Chem. Commun. 2009, 2192; c) M.
Angelin, A. Fischer, O. Ramstrçm, J. Org. Chem. 2008, 73, 3593; d) W. L.
Noorduin, T. Izumi, A. Millemaggi, M. Leeman, H. Meekes, W. J. P.
Van Enckevort, R. M. Kellogg, B. Kaptein, E. Vlieg, D. G. Blackmond, J. Am.
Chem. Soc. 2008, 130, 1158.
[8] a) J. W. Sadownik, D. Philp, Angew. Chem. Int. Ed. 2008, 47, 9965; Angew.
Chem. 2008, 120, 10113; b) J. W. Sadownik, D. Philp, Org. Biomol. Chem.
2015, 13, 10392.
2
006, 106, 3652; f) S. J. Rowan, S. J. Cantrill, G. R. L. Cousins, J. K. M. Sand-
ers, F. J. Stoddart, Angew. Chem. Int. Ed. 2002, 41, 898; Angew. Chem.
002, 114, 938.
2
[
5] a) J. Li, P. Nowak, S. Otto, J. Am. Chem. Soc. 2013, 135, 9222; b) S.
Hamieh, V. Saggiomo, P. Nowak, E. Mattia, R. F. Ludlow, S. Otto, Angew.
Chem. Int. Ed. 2013, 52, 12368; Angew. Chem. 2013, 125, 12594; c) R. A. R.
Hunt, S. Otto, Chem. Commun. 2011, 47, 847; d) P. Besenius, P. A. G. Cor-
mack, R. F. Ludlow, S. Otto, D. C. Sherrington, Org. Biomol. Chem. 2010, 8,
[9] D. Sievers, G. von Kiedrowski, Chem. Eur. J. 1998, 4, 629.
2414; e) S. M. Turega, C. Lorenz, J. W. Sadownik, D. Philp, Chem. Commun.
2
008, 4076; f) S. Xu, N. Giuseppone, J. Am. Chem. Soc. 2008, 130, 1826;
Received: September 17, 2015
Revised: November 30, 2015
g) P. Besenius, P. A. G. Cormack, J. Liu, S. Otto, J. K. M. Sanders, D. C. Sher-
rington, Chem. Eur. J. 2008, 14, 9006; h) A. Bugaut, K. Jantos, J.-L. Wietor,
Published online on December 22, 2015
Chem. Eur. J. 2016, 22, 1831 – 1839
www.chemeurj.org
1839
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim