Exploring the relation between amplification and binding in dynamic combinatorial libraries of macrocyclic synthetic receptors in water

Article abstract of DOI:10.1002/chem.200701413

Herein we describe an extensive study of the response of a set of closely related dynamic combinatorial libraries (DCLs) of macrocyclic receptors to the introduction of a focused range of guest molecules. We have determined the amplification of two sets of diastereomeric receptors induced by a series of neutral and cationic guests, including biologically relevant compounds such as acetylcholine and morphine. The host-guest binding affinities were investigated using isothermal titration calorimetry. The resulting dataset enabled a detailed analysis of the relationship between the amplification of selected receptors and host-guest Gibbs binding energies, giving insight into the factors affecting the design, simulation and interpretation of DCL experiments. In particular, two questions were addressed: Is amplification by a given guest selective for the best receptor? And does the best guest induce the largest amplification of a given receptor? Our experimental results and computer simulations showed that the relative levels of amplification of hosts by a guest are well-correlated with their relative affinities, and simulations have confirmed previous observations that amplification can be selective for the best receptor when only modest amounts of guest are used. In contrast, the correlation between guest binding and the extent of amplification of a given receptor across a wide range of guests tends to be poorer, because every guest has its own unique set of affinities for competing receptors in the DCL. This implies that the results of screening a DCL for selective receptors by comparing the response of the mixture to two different guests should be interpreted with caution. DCLs are complex mixtures in which all compounds are connected through a set of equilibria. Obtaining quantitative information about all host-guest binding constants from such systems will require the explicit and simultaneous consideration of all of the main equilibria within a DCL.

Full text of DOI:10.1002/chem.200701413

FULL PAPER
DOI: 10.1002/chem.200701413
Exploring the Relation between Amplification and Binding in Dynamic
Combinatorial Libraries ofMacrocyclic Synthetic Receptors in Water
Peter T. Corbett, Jeremy K. M. Sanders, and Sijbren Otto*[a]
Abstract: Herein we describe an exten-
sive study of the response of a set of
closely related dynamic combinatorial
libraries (DCLs) of macrocyclic recep-
tors to the introduction of a focused
range of guest molecules. We have de-
termined the amplification of two sets
of diastereomeric receptors induced by
a series of neutral and cationic guests,
including biologically relevant com-
pounds such as acetylcholine and mor-
phine. The host–guest binding affinities
were investigated using isothermal ti-
tration calorimetry. The resulting data-
set enabled a detailed analysis of the
relationship between the amplification
of selected receptors and host–guest
Gibbs binding energies, giving insight
into the factors affecting the design,
simulation and interpretation of DCL
experiments. In particular, two ques-
tions were addressed: Is amplification
by a given guest selective for the best
receptor? And does the best guest
induce the largest amplification of a
given receptor? Our experimental re-
sults and computer simulations showed
that the relative levels of amplification
of hosts by a guest are well-correlated
with their relative affinities, and simu-
lations have confirmed previous obser-
vations that amplification can be selec-
tive for the best receptor when only
modest amounts of guest are used. In
contrast, the correlation between guest
binding and the extent of amplification
of a given receptor across a wide range
of guests tends to be poorer, because
every guest has its own unique set of
affinities for competing receptors in
the DCL. This implies that the results
of screening a DCL for selective recep-
tors by comparing the response of the
mixture to two different guests should
be interpreted with caution. DCLs are
complex mixtures in which all com-
pounds are connected through a set of
equilibria. Obtaining quantitative infor-
mation about all host–guest binding
constants from such systems will re-
quire the explicit and simultaneous
consideration of all of the main equili-
bria within a DCL.
Keywords: disulfides
·
dynamic
combinatorial chemistry · macrocy-
cles · molecular recognition · water
Introduction
not always as successful as one would wish, prompting the
exploration of alternative approaches to synthetic receptors
that are less dependent on a detailed understanding.[2] One
such alternative is provided by dynamic combinatorial
chemistry.[3] In this method a set of relatively simple build-
ing blocks are designed, prepared and combined. A reversi-
ble reaction is used to link the building blocks together to
produce an equilibrium mixture of potential receptors (a dy-
namic combinatorial library or DCL). The binding energy
between host and guest can then be harnessed to drive the
synthesis of (ideally) the best receptor by shifting the equi-
librium of receptors towards those with the highest affinities
(Figure 1).
The design of successful synthetic receptors requires being
able to balance i) the strength of the non-covalent interac-
tions formed between the host and the guest; ii) the cost or
benefit of desolvating the host and the guest; iii) the entrop-
ic penalty associated with placing the host and guest in con-
formations suitable for binding; and iv) the benefit or cost
of any non-covalent intra-host[1] or intra-guest interactions
made or broken during binding. With our current imperfect
understanding of all these interactions this balancing act is
[a] Dr. P. T. Corbett, Prof. J. K. M. Sanders, Dr. S. Otto
Department of Chemistry, University of Cambridge
Lensfield Road, Cambridge CB21EW(UK)
Fax : (+44)1223-336017
Dynamic combinatorial chemistry has spread into many
different areas, using an increasing number of reversible re-
actions to generate not only hosts but also guests for a wide
range of targets[3e,4] as well as sensors,[5] and catalysts.[6] In
some cases, spectacular levels of amplification of receptors
with exceptionally high affinity[4a] or remarkable structure[7]
E-mail: so230@cam.ac.uk
Supporting information for this article is available on the WWW
under http://www.chemeurj.org/ or from the author.
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
2153

Figure 1. Bifunctionalized building blocks combine to form a dynamic combinatorial library of macrocycles that are potential receptors. The amounts of
the individual macrocycles depend on their thermodynamic stabilities as represented by the depth of the corresponding well in the free energy landscape.
Addition of a guest or template that selectively binds to one of the macrocycles induces a shift in the equilibrium towards this species at the expense of
all other library members.
have been observed. A fundamental understanding of the
complex and sometimes counterintuitive behavior of DCLs
in response to the introduction of the template has started
to emerge as a result of theoretical investigations by Severin
and co-workers[8] and ourselves.[9]
a DCL made from such building blocks to the original guest
should then lead to re-formation of the corresponding re-
ceptor, while different guests may lead to new receptors. We
chose reversible disulfide chemistry for constructing the
[11]
DCLs. We and others[12] have demonstrated that disulfide
Key to most applications of DCLs is the presumption that
library members that bind strongly to a template get ampli-
fied efficiently. This paper describes a detailed study of the
binding and amplification of two pairs of diastereomeric re-
ceptors from a DCL by several different but structurally re-
lated guests, including biologically and pharmaceutically rel-
evant compounds such as acetylcholine and morphine alka-
loids. It represents the most extensive dynamic combinatori-
al dataset to date, allowing us, for the first time, to probe
the relationship between binding strength to a particular
host in a DCL and the guest-induced amplification of this
host across a wide range of guests. From the data obtained
in this study of selected hosts in DCLs of moderate com-
plexity, we can see where the results of DCL experiments
can be correctly analyzed and interpreted using relatively
simple techniques, and identify situations where more so-
phisticated methods are required.
DCLs can be generated by mixing thiol building blocks in
aqueous solution at close to neutral pH. Oxygen from the
air is sufficient to irreversibly oxidize the thiols into the de-
sired disulfides, producing water as the only by-product.
While oxidation is taking place, the mixture contains disul-
fides and thiols, allowing for equilibration through nucleo-
philic attack of thiolate anions on the disulfides, displacing a
new thiolate anion in the process. Thus, exchange is catalytic
in thiolate anion; it can be switched off by allowing oxida-
tion to go to completion or by addition of acid to protonate
the thiolate.
In particular, two key questions will be addressed: Do the
best guests induce the strongest amplifications? Are the
best hosts the most amplified? Some results of these studies
have previously been communicated.[6a,9c,10] The paper is
structured as follows: We first describe the design and syn-
thesis of the building blocks, followed by the analysis of the
DCLs, their response to guest molecules, and the characteri-
sation of selected host molecules. We then describe the ther-
modynamics of host–guest binding. Finally, we come to the
essence of the investigation where we explore the relation
between binding and amplification.
Results and Discussion
Design and synthesis ofbuilding blocks : Building blocks
were designed to contain structural elements of previously
known successful synthetic receptors for specific guests.
These receptor fragments were equipped with suitable rever-
sible covalent attachment points. We reasoned that exposing
Scheme 1. Design of the dithiol building blocks was inspired by the Koga/
Breslow (1)[13] and Dougherty (2)[14] cyclophane receptors.
2154
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
We selected two previously
described water soluble hosts
1[13] and 2,[14] which were known
to bind strongly to anionic and
cationic guests, respectively.
The hosts were built up from
subunits that would allow the
introduction of two thiol groups
with minimal effect on the over-
all structure of the receptor
(Scheme 1). Both receptors
were known to be tolerant to
minor structural variations such
as changes to the length of the
alkyl spacer in 1[13b] and the
nature of the phenyl spacer in
2.[14e,n] We started our experi-
ments on the Koga/Breslow re-
ceptor 1, and successfully syn-
thesized the required dithiol
building block 3. Unfortunately,
when exposed to the conditions
required for disulfide oxidation
and exchange (pH 7–9), this
building block decomposed.[15]
We were more successful with
the building blocks 4 and 5 that
were inspired by the Dougherty
cyclophane 2.[6a,9c,10]
Scheme 2. Building block synthesis. i) NaBH4, RT, aq Na2CO3; ii) Me2N-C(S)-Cl, DABCO, 08C!RT, DMF;
iii) 3 h at 2408C, Ph2O; iv) dimethyl acetylenedicarboxylate, 1.25 h at 1908C, Ph2O; v) 1.75m KOH in diethyle-
neglycol, 30 min at 1058C.
Table 1. Experimentally observed binding constants and free energies, enthalpies and entropies of binding of
templates T1–T13 to hosts 12a and 13a and amplification factors (AFs) for hosts 12a and b and 13a and b in-
duced by the same templates (5 mm) in libraries A and B.[a]
Guest
Host K[b]
[mꢀ1
DGo[b]
[kJmolꢀ1
DHo[b]
[kJmolꢀ1
TDSo[b]
[kJmolꢀ1
]
AF(A)[a] AF(B)[a]
G
]
E
]
G
]
N
12a
12b
13a
13b
2.1”103 (0.3) ꢀ19.0 (0.3)
ꢀ15.1 (7.6) 4.2 (2.1)
ꢀ22.4 ꢀ3.6
5
3.2
4.5
3.4
5.7
1.6
2.10”103
ꢀ18.9
Dithiol building block 4 was
synthesized as a racemic mix-
ture as outlined in Scheme 2.
The key step is a Diels–Alder
reaction between the protected
precursors 8 and dimethylacety-
lenedicarboxylate which pro-
ceeded in 84% yield. The thiol
groups were introduced starting
from aromatic alcohols through
12a
12b
13a
13b
4.5”104 (0.5) ꢀ26.5 (0.3)
2.0”105 (0.0) ꢀ30.2 (0.1)
ꢀ23.8 (0.8) 2.8 (1.1)
ꢀ38.3 (1.7) ꢀ8.1 (1.7)
4.2
2.6
1.8
1.4
16.2
2.6
12a
12b
13a
13b
1.30”105[c,d]
6.40”105[c]
ꢀ29.1[c,d]
ꢀ33.1[c]
ꢀ23.7[c,d]
ꢀ40.6[c]
5.4[c,d]
ꢀ7.5[c]
a
Newman–Kwart rearrange-
12a
12b
13a
13b
3.6”105 (0.3) ꢀ31.7 (0.2)
3.3”105 (0.0) ꢀ31.5 (0.0)
ꢀ19.0 (0.9) 12.7 (1.1)
ꢀ30.8 (1.1) ꢀ1.0 (0.7)
10.6
3.2
6.2
2.2
10.9
3.2
ment on the O-thiocarbamate
intermediate 7, following meth-
ods described by Field and En-
gelhardt.[16] A similar strategy
was used to synthesize building
block 5. Both compounds can
easily be obtained in 10–30 g
scale.
12a
12b
13a
13b
8.5”102 (1.0) ꢀ16.7 (0.4)
6.5”103 (0.5) ꢀ21.7 (0.2)
0.6
0.5
0.7
0.7
5.5
1.2
12a
12b
13a
13b
5.6”104 (0.2) ꢀ27.1 (0.1)
3.2”104 (0.3) ꢀ25.7 (0.2)
ꢀ18.5 (0.4) 8.7 (0.5)
ꢀ30.6 (1.5) ꢀ4.9 (1.7)
11.2
4
9.1
3.9
14.9
1.3
Guest-induced amplification of
hosts from DCLs of macrocy-
clic disulfides: In previous stud-
ies using disulfide DCLs made
from dithiol building blocks in-
cluding 4 and 5 we have ob-
served that introducing tem-
plates T2–T4 and T13 (Table 1)
12a
12b
13a
13b
7.1
5.4
3
2.8
9.7
5.3
9.9”104 (1.0) ꢀ28.5 (0.3)
ꢀ12.6 (0.9) 16.0 (1.2)
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2155

S. Otto et al.
and investigate the stereoselec-
tivity of their amplification for
a wide range of guests. We suc-
ceeded in separating all isomer-
ic receptors from each other
and from the remaining library
members using an isocratic
HPLC method as shown in
Figure 2.
The assignments of the dia-
stereomers of hosts 12 and 13
were conducted by 1H NMR
spectroscopy. In case of host 12
the analysis was straightfor-
ward: the diastereomeric prod-
ucts have different symmetries
and give rise to different num-
bers of signals in the 1H NMR
spectrum. For the diastereomer
in which all three subunits have
Table 1. (Continued)
Guest
Host K[b]
[mꢀ1
DGo[b]
[kJmolꢀ1
DHo[b]
[kJmolꢀ1
TDSo[b]
[kJmolꢀ1
]
AF(A)[a] AF(B)[a]
G
]
U
]
R
]
U
12a
12b
13a
13b
3.9”103 (0.9) ꢀ20.4 (0.6)
6.5”103 (2.0) ꢀ21.7 (0.6)
ꢀ31.6 (0.7) ꢀ11.1 (1.2) 2.1
2
1.6
1.9
7.1
2.3
12a
12b
13a
13b
7.9”104 (0.2) ꢀ28.0 (0.1)
4.9”104 (0.0) ꢀ26.8 (0.0)
ꢀ18.8 (0.6) 9.2 (0.5)
ꢀ26.2 (0.2) 0.6 (0.2)
9.2
4.7
10.8
6.7
22
3.6
12a
12b
13a
13b
2.4”105 (0.0) ꢀ30.7 (0.1)
4.0”104 (0.0) ꢀ26.3 (0.0)
ꢀ19.5 (0.3) 11.2 (0.4)
ꢀ25.4 (1.8) 0.9 (1.7)
11.4
5.5
12.1
6.9
9.7
1.1
12a
12b
13a
13b
1.1”106 (0.2) ꢀ34.5 (0.6)
9.3”104 (1.0) ꢀ28.3 (0.4)
ꢀ19.5 (0.6) 15.0 (1.1)
ꢀ26.5 (1.3) 1.9 (1.6)
15.5
3.8
20.8
6.1
6.9
0.4
12a
12b
13a
13b
2.7”104 (0.2) ꢀ25.3 (0.2)
ꢀ33.6 (3.6) ꢀ8.3 (3.4)
5.7
2.8
1.3
6.6
3.4
the same chirality (12b)
a
simple NMR spectrum very
similar to that of the building
block is expected, whereas the
trimer in which the subunits
have different chiralities (12a;
RR,RR,SS or SS,RR,RR) all
three subunits are inequivalent.
Isolation of the predominant
stereoisomer by preparative
HPLC and subsequent analysis
by NMR showed three, rather
than one set of signals per mac-
rocycle, proving it to be 12a.
1.4
12a
12b
13a
13b
5.4”105 (0.4) ꢀ32.7 (0.2)
ꢀ46.4 (0.7) ꢀ13.7 (0.9) 11.2
18.1
13.9
3.7
7.2
2.6”104 (0.6) ꢀ-25.1 (0.6) ꢀ42.9 (4.0) ꢀ17.8 (4.5)
0.8
[a] The DCLs in set A contained 5 mm 4, and the libraries in set B contained 3.33 mm 4 and 1.67 mm 5. Errors
are shown in parentheses and are based on the reproducibility of the data over 2 or 3 separate experiments.
[b] Determined using isothermal titration calorimetry using 10 mm borate buffer pH 9.0 at 298 K. [c] Data
taken from ref. [6a]. [d] Data for a mixture of 12a and 12b.
For host 13 a similar analysis
is inconclusive as both diaste-
reomers will give two sets of
induced dramatic changes in the library compositions, lead-
ing to the amplification of hosts 12 and 13.[6a,10] The struc-
ture of the selected dynamic combinatorial receptors differs
from the expected disulfide analogue of the Dougherty re-
ceptor 2. This difference may be due to the more flexible
nature of the -CH2-O- units in 2 as compared to the -S-S-
units in the disulfide analogues. Flexibility is important as
host 2 is reported to bind guest T2 in a partially collapsed
conformation[14] and it is likely that the more rigid disulfide
analogue of 2 is unable to adopt a similar collapsed confor-
mation. Alternatively, or additionally, the increased length
and different bond angles of the disulfide linkages may
cause a difference in binding.
Since building block 4 was used as a racemic mixture
hosts 12 and 13 were obtained as mixtures of stereoisomers.
Later we reported the unexpected observation that template
T5 amplifies receptor 14 containing four units of building
block 4.[17] Remarkably, a strong preference for one out of
four possible diastereomeric products was observed. This
observation prompted us to return to receptors 12 and 13
signals. However, since one of the diastereomers (13b;
RR,SS) is a meso compound while the other (13a) is pro-
duced as a racemic mixture of RR,RR and SS,SS, addition of
a suitable chiral shift reagent should give separate signals
for the two enantiomers of the racemate without producing
additional signals for the meso isomer. We have isolated a
mixture of both diastereomers of 13 by preparative HPLC
1
and analyzed this material by H NMR in D2O. Addition of
homochiral ammonium salt 15 induced marked changes in
the NMR spectrum as shown in Figure 3. The signals due to
the bridgehead protons (5.5–5.1 ppm) are most revealing, as
only two peaks are expected per diastereomer. Thus, four
signals are expected in the absence of the chiral guest (cf.
Figure 3a), and six in the presence of the guest. Although
some of the signals in Figure 3 overlap, this is indeed what is
observed. Integration of the signals indicates that 13a is the
predominant diastereomer.
Having established the stereochemistry of the diastereo-
mers of hosts 12 and 13 we set out to study the selectivity in
the amplification of these four compounds using a large
2156
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
guests for receptor 2. Following
this precedent we selected
guests T1–T11 (Table 1). We
also included a number of bio-
logically relevant templates:
acetylcholine (T8) and two
morphine alkaloids: morphine
itself (T12), and the N-methy-
lated morphine derivative T13.
Two sets of 14 DCLs were
made: one biased towards the
formation of 12 (set A) and one
biased towards the formation of
13 (set B). The libraries in set
A contained 5 mm 4, and those
in set B contained 3.33 mm 4
and 1.67 mm 5. The libraries
were prepared by dispersing the
building block(s) in water and
adjusting the pH to 8.0. In each
set, 13 DCLs contained a tem-
plate (T1 to T13, respectively;
at 5 mm concentration), while
no template was added to the
14th DCL. The libraries were
allowed to oxidize in capped
HPLC vials and stirred at 258C
for three weeks. Precipitation
occurred in the libraries to
which T3 had been added, and
these libraries were therefore
not analyzed further. All of the
other libraries remained clear
Figure 2. Part of the HPLC analysis (Waters Symmetry reversed phase
C18 column 25.0 cm”4.6 mm, 5 mm particle size, 50:32:18:0.1 H2O/
MeCN/IPA/TFA) of a DCL made from 4 (3.33 mm), 5 (1.67 mm) and
5 mm quinuclidinium iodide (T9) showing separation of the diastereo-
mers of hosts 12 and 13. Additional minor library members can be ob-
served at 18.7 min and elsewhere in the trace.
Figure 3. Part of the 1H NMR spectra of a mixture of stereoisomers of
host 13 (in 49 mm pD 8.9 sodium borate buffer) in the presence of in-
creasing amounts of homochiral ammonium ion 15, showing the signals
of the racemic mixture of RR,RR and SS,SS 13a split in two while the
signals of the meso stereoisomer 13b are shifted but otherwise unaffect-
ed. a) [15]=0 mm; b) 0.13 mm; c) 0.63 mm; d) 1.18 mm.
series of different but structurally related templates. The
choice of the templates was partially motivated by the ex-
tensive studies by the Dougherty laboratory,[14] which had
demonstrated that quaternary amines and imines were good
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2157

S. Otto et al.
Table 2. Experimentally observed selectivity in the amplification of hosts
12a vs 12b; 13a vs 13b and 13a vs 12a by various templates (5 mm) in
DCLs of set B made from 4 (3.33 mm) and 5 (1.67 mm).
solutions. The library compositions were analyzed by HPLC
and the amplification factors for receptors 12a and b and
13a and b were computed. Amplification factors are defined
as the concentration of a library member in a templated li-
brary, divided by its concentration in the corresponding un-
templated library, and can be calculated by dividing the
areas of the HPLC peaks corresponding to these compounds
in the templated libraries and the untemplated libraries. The
results are summarized in the last two columns of Table 1.
These four receptors represent the major amplified spe-
cies in most of the libraries. The tetramer 14 was the major
host in libraries templated by T5,[17] and a significant minor
constituent in libraries of set A templated by T9,[9a] and to a
lesser extent T10 and a few others. It was not possible to
identify 14 in the untemplated libraries of set B. Other hosts
were visibly amplified as minor constituents of the libraries,
accounting for a few percent of the total peak area in vari-
ous experiments. However, as the aim of the present study
is to compare binding and amplification across a wide range
of templates, we here only report results on the two most
widely amplified hosts 12 and 13.
Template AF
A
(12b) AF
U
(13b) AF
N
N
T1
T2
T4
T5
T6
T7
T8
T9
T10
T11
T12
T13
1.3
1.3
2.8
1.1
2.3
1.1
1.0
1.6
1.7
3.4
2.0
1.3
3.6
6.2
3.4
4.6
11
1.8
3.1
6.1
8.7
16
1.3
9.0
1.8
7.8
1.6
3.2
3.6
2.0
0.75
0.33
0.20
0.20
0.90
4.5
Table 2 shows the selectivity of the amplification of hosts
12 and 13 induced by the templates in the libraries in set B.
Modest diastereoselectivities were observed: In almost all
cases, 12a is more strongly amplified than 12b, and 13a is
more strongly amplified than 13b suggesting that 12a and
13a are better hosts than their respective diastereomers, ir-
respective of the guest. In the untemplated libraries, the
ratios of 12a to 12b and 13a to 13b are 1.3:1 and 0.4:1, re-
spectively. In general, the preference for the amplification
of 13a over b is more pronounced than the preference for
12a over b. This difference may be due to hosts 13 present-
ing a smaller, more elongated cavity, leading to a tighter fit,
where steric interactions between the hosts and guests are
likely to be more important. In hosts 12, the cavities are
larger and more spherical, leading to a looser fit. Amongst
the templates that were analyzed the bulky and rigid T11
was the most diastereoselective, particularly in the case of
13. However, in all cases the diastereoselectivities were sig-
nificantly less than the more than 30-fold selectivity we pre-
viously observed for host 14.[17] We also analyzed the selec-
tivity in the amplification of receptor 12a versus 13a and
found that bulky guests preferred the larger host 12a.
Figure 4. Ratios of the diastereomers of 12 in experimental DCLs of set
A (5 mm 4), vs those of set B (3.33 mm 4 and 1.67 mm 5) for various tem-
plates (5 mm).
1.6 mm T13. We deliberately used a sub-stoichiometric quan-
tity of the template. If amplification would be enantioselec-
tive, the preferred enantiomer of the building block would
be recruited by the template to give one enantiomer of the
trimer. No template would remain to amplify the remaining
building block, which should therefore form mostly macro-
1
cycles other than trimer. A H NMR spectrum of the result-
In addition, we have studied the selectivity in the amplifi-
cation of receptor 12a vs 12b in the DCLs of set A (cf.
Table 1). As the ratio of the amplification factors for a pair
of diastereomers should be proportional to the ratio of their
binding constants (see the Supporting Information), this
ratio should remain the same across different DCLs. Indeed,
we observe a good correlation between the diastereoselec-
tivities in the amplification of 12 in libraries A and B
(Figure 4). This can interpreted as evidence that the DCLs
have all reached equilibrium and also provides an indication
of the accuracy of the experimental data.
ing mixture was taken with T13 acting as a chiral shift re-
agent, giving separate peaks for the enantiomers of 12a and
b in approximately 1:1 ratio, indicating that the amplifica-
tion of 12 was not significantly enantioselective.
Host–guest binding: The binding of templates T1–T13 to the
hosts 12a and 13a was studied using isothermal titration mi-
crocalorimetry (ITC) in 10 mm borate buffer pH 9.0.[18] The
results are shown in Table 1 from which a number of trends
are apparent. Firstly, the unquaternised templates (T1 and
T12) both have binding constants more than an order of
magnitude lower than those of their quaternised counter-
parts (T2 and T13). Similar observations have been reported
by Dougherty et al.[14f,j] and are evidence for the involve-
In an attempt to probe whether enantioselective amplifi-
cation was occurring in the presence of homochiral template
T13, we equilibrated 10 mm of rac-4 in the presence of
2158
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
ment of cation–p interactions in host–guest binding. Second,
large guests tend to prefer the larger host 12a, while small
templates tend to prefer the smaller host 13a. Third, tem-
plates with rather elongated aromatic structures such as T2
and T3 are bound particularly strongly by 13a.
Binding by 12 and 13 can be expected to be driven by two
major contributors: cation–p and hydrophobic interactions,
each with their own enthalpic and entropic components.
Whereas cation–p interactions are most likely enthalpy
driven, hydrophobic interactions have a more complex ther-
modynamic signature which can vary from entropy driven
(small curved solutes) to enthalpy driven (planar solutes or
cavities).[19]
For most of the templates measured, the enthalpies of
binding to 13a are around ꢀ10 kJmolꢀ1 stronger than the
corresponding enthalpies of binding to 12a (Figure 5). The
more favorable enthalpy contribution is almost completely
offset by an entropic penalty of around 10 kJmolꢀ1. This
pattern is consistent with 13a presenting a tighter, more
elongated cavity than 12a, with greater conformational con-
straints on binding but also with the prospect of forming a
more intimate host–guest contact. The fact that the flatter,
aromatic guests tend to have more enthalpy-driven binding,
in particular to 13a, supports this notion.
presumption by investigating the correlation between the
guest induced amplifications for a selected host and its bind-
ing to these guests.
Are the best hosts the most amplified? It is well established
that a breakdown of the correlation between binding and
amplification can occur when two or more library members
compete for a scarce building block.[8,9a,c] Library members
that use relatively small numbers of that building block per
host have a competitive advantage over those members that
contain larger numbers of the particular building block.
Thus, low oligomers tend to have an advantage over high
oligomers, and hetero oligomers over homo oligomers. An
intuitive way of rationalizing this behavior is illustrated by
the examples in Figure 6. Given a fixed amount of building
block a DCL can produce twice the number of cyclic dimers
than it can produce cyclic tetramers. Thus, for every four
building blocks the system can gain twice the guest–dimer
binding energy against only one guest–tetramer binding
energy, provided there is sufficient template available.
Hence, the system may prefer forming dimers even in cases
where the tetramer is actually the strongest binder. Similar
arguments apply in a mixed building block library, when
guest binding affinity is associated with one of the building
blocks. For example, if only the white building blocks in
Figure 6 bind the guest, the system tends to form heterodim-
ers even though the homodimer may be the strongest
binder. These effects will only occur in the presence of
excess template. When the template concentration is re-
duced to substoichiometric levels the system will revert to
selectively producing the best receptor.
Figure 5. Correlation between experimentally determined enthalpy and
&
entropy of binding of guests T1–T13 to hosts 12a (”) and 13a ( ). The
lines connect data points corresponding to the same guest.
Figure 6. In the presence of excess guest (T) there is an inherent prefer-
ence for the formation of small oligomers over large oligomers and for
the formation of hetero oligomers over homo oligomers because, with a
fixed amount of building blocks, the system is able to produce more of
the small or hetero oligomers than of the large or homo oligomers.
Correlation between guest binding and amplification: The
original naïve conception of dynamic combinatorial chemis-
try was that the best hosts in a DCL would be most ampli-
fied upon addition of a guest. Conversely, when exposing a
given DCL to a variety of guests, one would expect that the
guest that elicits the strongest response from the library is
the one that is most strongly bound. Recent investigations
by Severin[8] and ourselves[9a,c] have demonstrated that the
first of these two presumptions is not always correct and de-
pends on experimental conditions in a predictable way.
Below we will present further evidence that supports this
notion. With the present dataset we are also, for the first
time, able to experimentally test the validity of the second
The mixed building block DCLs of set B contain homo-
trimers 12 in competition with heterotrimers 13. In a previ-
ous study on the amplification induced by guest T9 at differ-
ent concentrations ranging from 0.5 to 10 mm we have
shown that at high concentrations of this guest the DCL
“misbehaves” and heterotrimer 13a is somewhat more am-
plified than homotrimer 12a despite the latter having a
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2159

S. Otto et al.
somewhat higher affinity for the guest.[9c] Upon reducing the
template concentration to a sub-stoichiometric level the
better receptor 12a is the most amplified.
The present dataset allows a more elaborate investigation
of the competition between 12a and 13a encompassing
guests T1–T13 and comparing binding affinities with am-
plification factors. In an ideal case of two isomeric receptors
We have conducted computer simulations of these libra-
ries using our DCLSim software[9a,b] with the aim of investi-
gating how the values of the slope and intercept depend on
experimental conditions. Simplified versions of the DCLs
were simulated, where only the two species under consider-
ation (12a and 13a) were allowed to bind to the template,
using the experimentally observed binding constants from
Table 1. No other experimental values were incorporated in
the model, thus excluding any other host–guest interactions
which undoubtedly occur. This methodology has previously
been shown to reproduce the main trends in experimental
DCLs, and allows us to rapidly explore the key aspects of
their behavior.[9] Figure 7b shows the results of the simula-
tions (full details are given in the Supporting Information).
Comparing these data with that in Figure 7a shows that,
while the individual simulated data points deviate from their
experimental counterparts, the trends in the simulated data
closely match those of the experimental data. This agree-
ment is all the more remarkable for the fact that the model
does not include other major binding constants, such as
those to the tetramer 15 and to the minor diastereomers
12b and 13b, and serves to further validate the robustness
of simplified DCL simulations.
R1 and R2 a plot of log(AFR1/AFR2) vs. log(KR1/KR2) for the
G
various guests will be a straight line passing through the
origin (when binding constants are equal, amplification fac-
tors should be equal) and having a slope of 1 (the ratio of
amplification factors equals the ratio of binding con-
stants).[20] Our experimental DCLs, in which we used build-
ing block and template concentrations of 5 mm, show a clear
deviation from this ideal behavior (Figure 7a). Data points
falling in the white quadrants of the plot represent DCLs
where the best binder of the two receptors has the higher
amplification factor. In the grey quadrants, the DCL selects
the “wrong” library member. This undesired behavior
occurs only for a relatively narrow range of binding events,
where K13a < K12a < 3.2”K13a
.
We have performed similar simulations for different guest
concentrations and found that, as expected, at lower guest
concentrations the bias towards hetero-oligomers disappears
and the intercept of plots of the type of Figure 7 approach
zero (Figure 8). Also the correlation between amplification
factor and binding constant (as reflected in the value of the
correlation coefficient R2) improves upon reducing the
guest concentration, in line with previous observations.[8,9a,c]
However, the guest concentration appears to have little
effect on the slope of the lines (i.e., the sensitivity of the
DCL). The main experimental parameter that determines
the slope turned out to be the ratio of the building blocks 4
and 5 in the DCL.
Figure 7. Correlation between ratios of amplification factors and binding
constants of 12a and 13a for a range of guests in (a) experimental and
(b) simulated DCLs of set B (3.33 mm 4; 1.67 mm 5 and 5 mm template).
The shaded parts of the graph signify “misbehaving” DCLs where the
most amplified of the two receptors is not the strongest binder.
Inspection of Figure 7a reveals that we still obtain an es-
sentially linear relationship between log(AF12a/AF13a) versus
log(K12a/K13a) but the y intercept (ꢀ0.37) is clearly not zero
and the gradient (0.74) is significantly less than 1. The non-
zero intercept reflects a systematic bias towards the amplifi-
cation of the heterotrimer 13a over the homotrimer 12a by
a factor 2.3 (i.e. 100.37), in line with the inherent preference
for the amplification of hetero oligomers over homo oligo-
mers. The reduced slope of the line indicates that the sensi-
tivity of the DCL (i.e., the extent to which a difference in
guest binding between 12a and 13a is reflected in a differ-
ence in amplification) is suboptimal. However, the correla-
tion between relative amplification and relative binding is
good (R2 =0.97), suggesting that if the binding constants for
a few host/guest combinations are known, it should be possi-
ble to deduce the relative affinities for other guests with rea-
sonable accuracy.
~
&
Figure 8. Dependence of the slopes ( ), intercepts ( ) and correlation co-
efficients (”) of the plots of log(AF12a/AF13a) vs log(K12a/K13a) on a) the
concentration of the guests at a fixed 2:1 ratio of 4 and 5, and b) the 4:5
ratio at a fixed guest concentration of 5 mm in simulated DCLs.
In our experimental mixed building block DCLs we used
building blocks 4 and 5 in a 2:1 ratio and also the simula-
tions shown in Figure 8a were performed using this library
composition. Varying the ratio of 4 to 5 has profound effects
2160
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
on the slope, intercept and correlation coefficient of the re-
lationship between log(AF12a/AF13a) and log(K12a/K13a). In-
creasing the fraction of 5 brings the slope close to unity,
while the correlation between binding and amplification
reaches a maximum close to [4]/[5] ratio of 7:3 and the inter-
cepts is nearly zero at a [4]/[5] ratio of 8:2. Thus, the opti-
mum [4]/[5] ratio for a close to zero y intercept is not the
optimum for R2, which in turn is different from the opti-
mum ratio that gives a slope close to unity. These results
suggest that there are no simple guidelines for choosing
building block ratios in mixed libraries when competing re-
ceptors are present. Thus, when screening mixed building
block libraries it may be beneficial to set up and analyze
several experiments at different building block ratios.
Do the best guests induce the strongest amplifications? This
question is key in assessing the potential of DCLs for the
development of selective receptors that specifically favor a
certain guest (X) over another (Y). Ideally one should be
able to identify selective receptors by comparing the prod-
uct distributions in a DCL exposed to X with that exposed
to Y. Inspection of the data in Table 1 reveals that a large
difference in amplification factors between two guests cor-
rectly predicts a large difference in binding constants in
some cases, but not in others. For example in library B guest
T1 amplifies host 12a 6.4 times better than guest T5, while
the difference in binding constant is only 2.5-fold. In con-
trast, in the same library guest T4 amplifies host 12a 8.9
times better than guest T1, while the difference in binding
constant is more than two orders of magnitude larger (a
factor of 424). We have analyzed the selectivity in the am-
plification for all possible guest combinations for hosts 12a
and 13a in libraries A and B. The results are shown in
Figure 9 and demonstrate that, while a correlation between
selectivity in binding and amplification is evident, considera-
ble scatter exists. Nevertheless, the instances where large se-
lectivities in amplification are observed are invariably asso-
ciated with highly selective receptors, although the reverse
may not necessarily be true.
In an attempt to understand the origin of the scatter ob-
served in Figure 9 we have analyzed the relationship be-
tween the guest-binding strength and the amplification
factor of a given receptor in more detail. In the absence of
complicating factors such as competition from other library
members, this relationship is expected to be sigmoidal: For
weakly binding guests the host will be essentially unampli-
fied, and so small changes in the binding energy will have
little effect, producing a flat region of the curve. For strong-
ly binding guests, the amplification of the host will be limit-
ed by the availability of the building blocks, again producing
a flat region. Only at intermediate binding energies will the
amplification factor vary significantly with the guest binding
energy.
Figure 9. Correlation between relative binding constants and relative am-
plification factors of receptors a) 12a and b) 13a comparing two guests X
and Y corresponding to any combination of T1–T13 in the experimental
&
DCLs of class A (&) and B ( ).
hosts 12a and 13a in the DCLs of set B. Amplification of
host 12a in the simple DCLs of set A, which contain only
one building block, appears to correlate fairly well with the
host–guest binding energy. Amplification of the same host
in the more complex DCLs of set B shows a somewhat in-
creased amount of scatter, while the amplification of host
13a in the same libraries is hardly correlated at all to the
guest binding energy. Thus, under the conditions of our ex-
periments, the correlation between host–guest binding
energy and amplification factor is different for different li-
braries and different for different receptors.
Although some scatter is expected through experimental
error, there is another more plausible cause for the variabili-
ty of the correlation: competition for the guest by other li-
brary members. Figure 11 shows the results of a simulation
of a simple model DCL made from a single building block,
consisting of one non-binding dimer, two trimers and one
non-binding tetramer (full details are given in the Support-
ing Information). The solid line shows how the amplification
of one of the trimers varies as a function of the affinity to-
wards a variety of guests, when the other trimer essentially
does not bind. As expected, a sigmoidal relationship is ob-
served. The dashed lines show the results of giving the com-
peting trimer a progressively higher but constant binding
energy. Thus, competition from a single library member
causes the sigmoidal curves to shift further left as the com-
Figure 10a shows the experimental results for the amplifi-
cation of 12a induced by guests T1–T13 (where available)
as a function of the host–guest binding constant in DCLs of
set A. Figure 10b and c show the corresponding data for
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2161

S. Otto et al.
single species, with a single amplification factor, and weight-
ed average binding constants did not substantially improve
the correlations.
Reducing the template concentration can be expected to
improve the correlations of the type shown in Figures 9 and
10 to some extent by removing the competition from rela-
tively weakly binding small and/or hetero oligomers. How-
ever, using lower guest concentrations is unlikely to lead to
perfect correlations as the disruptive effect of truly competi-
tive binders in the DCL will be unaffected by reducing the
template concentration. These expectations were confirmed
by performing simulations of the DCLs at a reduced substoi-
chiometric template concentration of 1.5 mm. A comparable
drop in template concentration has been shown to be suffi-
cient to cause T9 to go from preferentially amplifying its
weaker binder 13a to preferentially amplifying the stronger
binder 12a.[9c] The simulations were based on a minimal
model system in which only a homo- and a heterotrimer had
affinities for the guests, using the free energies of binding in
Table 1 (see Supporting Information for details), which
varied only from the model used to generate Figure 7b in
the concentration of the template. The results for the DCLs
of set B (Figure 12) show only marginally better correlations
between selectivities of amplification and binding, as com-
pared to the experimental data in Figure 9, which was ob-
tained using a template concentration of 5 mm.
Figure 10. Experimentally observed amplification factors (AFs) as a func-
tion of the host–guest Gibbs binding energy of a) host 12a in DCLs of
set A (5 mm 4), b) host 12a and c) host 13a in DCLs of set B (3.33 mm 4
and 1.67 mm 5).
petitive binding becomes stronger. These shifts are caused
by a competitor which, rather unrealistically, binds all guests
with the same affinity. In experimental DCLs, the binding
strength of any competing species is likely to vary from one
guest to the next, resulting in considerable scatter.
Figure 11. Amplification of a trimeric receptor in simple simulated DCLs,
in the presence of competitors with fixed free energies of binding. Full
details of the simulations are provided in the Supporting Information.
Figure 12. Correlation between relative binding constants and simulated
relative amplification factors of receptors a) 12a and b) 13a comparing
two guests X and Y corresponding to any combination of T1–T13 in
DCLs of class B using a template concentration of 1.5 mm.
In our experimental systems, there are many library mem-
bers that may compete with 12a and 13a: the diastereomers
12b and 13b, the homotetramer 14 and its diastereomers,
and in fact any other library member. Interestingly, the
minor diastereomers 12b and 13b do not appear to be a
major source of scatter. Treating the pairs of isomers as a
2162
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
Conclusion
Experimental Section
Materials and methods: 1H NMR spectra were recorded on Bruker
DRX-500, DRX-400 or DPX-250 instruments or on a DRX instrument
fitted with a cryoprobe. 13C NMR spectra were recorded on Bruker
DPX-400 (100 MHz) or DRX-500 (125 MHz) instruments. Chemical
shifts are quoted in parts per million with reference to solvent signals.
All chemicals were purchased from Acros, Aldrich, Avocado or Fluka in
reagent grade quality or better and used without further purification. All
solvents used in synthesis were distilled prior to use and dry solvents
were freshly distilled from CaH2 under argon. Ultrapure water was ob-
tained from a Millipore water purification system, and when used for
HPLC, passed through a 0.45 mm Millipore filter. HPLC grade MeCN,
MeOH and 2-PrOH (Fisher, Romil) were passed through a 0.45 mm Mil-
lipore filter and used without further purification.
We have made dynamic combinatorial libraries (DCLs)
from dithiols 4 and 5 in water and analyzed the amplifica-
tion of diastereomeric receptors 12a and b and 13a and b in
response to the introduction of guests T1–T13. Nearly all
guests induced larger amplifications of receptor 12a as com-
pared to 12b. Similarly receptor 13a was preferred over its
diastereomer 13b, essentially irrespective of the guest. In
contrast, the selectivity in amplification between 12a and
13a depended on the nature of the guest, with larger spheri-
cal guests preferring 12a, while smaller more elongated
guests preferred the smaller receptor 13a. We have investi-
gated the thermodynamics of binding of guests T1-T13 to
receptors 12a and 13a. Affinities span three orders of mag-
nitude, ranging from 8.5”102 mꢀ1 for binding of T5 to host
12a to an impressive 5.4”105 mꢀ1 for binding of T13 to 12a.
By comparing binding and amplification data two key ques-
tions were addressed: Is amplification by a given guest se-
lective for the best receptor? And does the best guest
induce the largest amplification of a given receptor? We
performed computer simulations to further explore the be-
havior of the DCLs under different “experimental” condi-
tions. The results of the experimental and theoretical inves-
tigations show a good—albeit biased—correlation between
relative amplification and relative affinity for different hosts
for the same guests. Furthermore, the theoretical investiga-
tions were in line with other studies[8,9a,c] that showed that
this bias could be alleviated, leading to the reliably selective
amplification of the best binder, by using a modest amount
of guest. However, the correlation between host–guest bind-
ing and the amplification factor for a given host across a
wide range of guests (i.e., the selectivity of the host for the
guests) was less satisfactory, largely as a result of every
guest having its own unique pattern of affinities for the vari-
ous competing receptors in the mixture. Thus, a plot of the
ratio of the amplification factors induced by guests X and Y
versus the ratio of the underlying host–guest binding con-
stants showed considerable scatter. Nevertheless the highest
guest selectivities were characterized by the biggest differ-
ence in amplification factor.
Analytical HPLC was carried out on Hewlett Packard 1050 and 1100 sys-
tems coupled to UV analyzers and the data were processed using HP
Chemstation software. Separations were performed on reversed phase
Waters Symmetry C18 columns (25.0 cm”4.6 mm, 5 mm particle size) for
analytical HPLC. Except where otherwise stated, the chromatography
was carried out at 458C using Jones Chromatography or Anachem
column ovens and using UV detection at 320 nm.
LC-MS was conducted on a Hewlett Packard 1050 system, coupled to a
diode array detector and a Micromass Platform LC quadrupole mass an-
alyser, controlled using a combination of MassLynx and OpenLynx soft-
ware. Separations were achieved using reversed phase columns in the
same manner as the analytical HPLC analyses. Mass analysis was con-
ducted in the negative ion mode.
Preparative HPLC was conducted on
a Gilson preparative HPLC
system, equipped with a Gilson 305 pump, two Gilson 306 pumps, a
Gilson 806 manometric module, a Gilson 811C dynamic mixer, a Gilson
model 231 sample injector, a Gilson FC204 fraction collector, and a
Gilson 401 dilutor. UV detection was performed originally using a Gilson
115 UV detector, and later using a Shimadzu SPD-6 A UV detector. The
data were originally collected using a HP3395 integrator, and later a
computer running Gilson Unipoint software. Separation were performed
using a Nucleodur C18 preparative column (25.0 cm”2.1 cm, 100 Š,
5 mm), with a Nucleodur C18 guard column (5.0 cm”2.1 cm, 100 Š,
5 mm). Except where otherwise stated, a Grant water bath was used to
heat the column, and a length of tubing between the pump and the
column, to 458C. This arrangement ensured that the mobile phase was at
the correct temperature before it entered the column.
Microcalorimetry: Isothermal calorimetry measurements were conducted
by using a MCS-ITC calorimeter from MicroCal, LLC, Northampton,
MA, USA. A single 3 mL aliquot and 29 aliquots of 10 mL were titrated
into the calorimetric cell every 3 minutes. The data were analyzed using
the customized ITC module of the Origin 5.0 software package and a
least squares fitting procedure to fit the data to the appropriate binding
model. All measurements were carried out at 298 K.
In conclusion, assessing the absolute binding affinity of a
certain guest for a certain host from the extent to which the
guest induces the amplification of this host is not necessarily
reliable in a quantitative sense, but is in most cases still
qualitatively useful. Most importantly, when strong amplifi-
cations or strongly selective amplifications are observed the
selected receptors are indeed strong or highly selective bind-
ers, even though the reverse is not necessarily true. To fur-
ther extend the information that can be obtained from
DCLs, we are currently developing methodology that allows
also quantitative information on host–guest binding to be
extracted from the amplification factors by analyzing the be-
havior of the entire DCL, instead of focusing on selected
amplification events.[21]
Guests T1, T5–T8, T10 and T12 are commercially available. Guests
T2,[14a] T3,[6a] T4,[6a] T9,[22] T11,[22] and 15[23] were prepared as reported
previously.
N-Methylmorphinium iodide (T13): The compound was prepared by re-
acting morphine (0.50 g, 1.75 mmol) with methyl iodide (10 mL,
160 mmol) in acetonitrile (100 mL) for 72 h at room temperature fol-
lowed by recrystallization from acetonitrile. 1H NMR (500 MHz, D2O):
d=1.98 (d, 1H), 2.47 (dt, 1H), 2.83 (dd, 1H), 3.22 (s, 3H), 3.23–3.34 (m,
3H), 3.30 (s, 3H), 3.37 (t, 1H), 3.44 (d, 1H), 4.02 (t, 1H), 4.27 (m, 1H),
4.98 (dd, 1H), 5.30 (dt, 1H), 5.65 (dm, 1H), 6.56 (d, 1H), 6.65 ppm (d,
1H); 13C NMR (125 MHz, D2O): d=23.49, 29.30, 33.32, 41.34, 50.21,
54.15, 56.08, 65.41, 69.89, 90.03, 117.90, 120.20, 122.50, 125.55, 129.01,
133.13, 138.44, 145.31 ppm; elemental analysis calcd (%) for C 50.60, H
5.19, N 3.28; found: C 50.39, H 5.19, N 3.27.
Anthracene-2,6-diol (6):[24] Sodiumborohydride (9.6 g, 0.25 mol) was dis-
solved in a 1m solution (240 mL) of sodium carbonate in water. 2,6-Dihy-
droxyanthraquinone (4.8 g, 20 mmol) was added in portions while stirring
at room temperature. After the initial gas evolution had stopped the so-
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2163

S. Otto et al.
lution was gently refluxed for 15 min. The reaction mixture was trans-
ferred to a large beaker and acidified using 250 mL 3m HCl (caution: ex-
tensive foaming). The precipitated product was filtered and taken up in
acetone and filtered over a short pad of celite. The solvent was evaporat-
ed and the compound dried under vacuum overnight, yielding 3.99 g
(19 mmol, 95%) anthracene-2,6-diol. The product was used without fur-
ther purification in the next step. 1H NMR ([D6]DMSO, 400 MHz): d=
7.06 (dd, 2H), 7.12 (d, 2H), 7.80 (d, 2H), 8.11 (s, 2H), 9.63 ppm (s, 2H).
(CD3OD, 400 MHz): d=5.53 (s, 2H), 6.94 (dd, 2H), 7.24 (d, 2H),
7.33 ppm (d, 2H); 13C NMR (CD3OD, 100 MHz): d=53.33, 125.23,
125.77, 126.83, 129.83, 142.52, 146.59, 148.99, 168.61 ppm; LC-MS: m/z:
calcd: 379.0075; found: 379.0060 [M+Na+].
3,5-Bis(dimethylthiocarbamoyloxy)benzoic acid methyl ester (10):
Methyl 3,5-dihydroxybenzoate (5.0 g, 29.7 mmol) was dissolved in anhy-
drous DMF (20 mL) under a nitrogen atmosphere. The solution was
cooled to 08C and DABCO (13.3 g, 119 mmol) was added in portions. To
the resulting suspension a solution of N,N-dimethylthiocarbamoyl chlo-
ride (14.7 g, 119 mmol) in DMF (20 mL) was added dropwise at 0–58C.
Where the reaction mixture “solidified” more DMF was added to enable
efficient stirring. The suspension was allowed to warm to room tempera-
ture and stirred for 24 h. The reaction mixture was poured into water
(200 mL), filtered and the residue was washed with ethanol to give 9.22 g
(26.9 mmol, 90%) a white crystalline powder. The product was pure
enough for most purposes, but can be recrystallized from ethanol (m.p.
135–1368C). 1H NMR (CDCl3, 400 MHz): d=3.34 (s, 3H), 3.44 (s, 3H),
3.88 (s, 3H), 7.05 (t, 1H), 7.63 ppm (d, 2H); 13C NMR (CDCl3,
100 MHz): d=39.84, 44.33, 53.38, 122.66, 123.86, 134.25, 154.92, 166.41,
187.96 ppm; LC-MS: m/z: calcd: 365.0606; found: 365.0605 [M+Na+].
Dimethylthiocarbamic acid O-(6-dimethylthiocarbamoyloxyanthracen-2-
yl) ester (7): Crude anthracene-2,6-diol (6; 9.0 g, 42.8 mmol) was dis-
solved in 80 mL anhydrous DMF under a nitrogen atmosphere. The solu-
tion was cooled to 08C and DABCO (28.8 g, 257 mmol) was added in
portions. To the resulting suspension a solution of N,N-dimethylthiocar-
bamoyl chloride (suspect carcinogen!) (31.8 g, 257 mmol) in anhydrous
DMF (40 mL) was added dropwise at 0–58C. In cases where the suspen-
sion appeared to “solidify”, more DMF was added. The suspension was
allowed to warm to room temperature in a melting ice bath overnight
and stirred for a total of 24 h. The reaction was monitored by TLC (silica
gel, 1% methanol in chloroform; Rf starting material: 0.6; Rf product:
0.2). The reaction mixture was filtered and the residue was washed thor-
oughly with water and finally with a small amount of ethanol to give
9.65 g (25.1 mmol, 59%) of a yellow-brown powder. The product was
pure enough to be used in the next step, but can be recrystallized from
chloroform/acetone (m.p. 282–2848C). 1H NMR (CDCl3, 400 MHz): d=
3.41 (s, 6H), 3.49 (s, 6H), 7.26 (dd, 2H), 7.61 (d, 2H), 7.97 (d, 2H),
8.37 ppm (s, 2H); 13C NMR (CDCl3, 100 MHz): d=41.41, 45.88, 121.42,
125.90, 128.86, 131.63, 132.78, 134.03, 153.69, 190.31 ppm; LC-MS: m/z:
calcd: 407.0864; found: 407.0859 [M+Na+].
3,5-Bis(dimethylcarbamoylsulfanyl)benzoic acid methyl ester (11): O-Thi-
ocarbamate 10 (8.0 g, 23.3 mmol) was suspended in diphenyl ether
(80 mL) and heated under a nitrogen atmosphere on a sand bath to 230–
2408C for 3 h. The reaction was monitored by TLC (silica gel, CHCl3/
CH3CN 9:1, Rf starting material: 0.4, Rf product: 0.15). After the reaction
mixture was allowed to cool to 30–408C it was poured into 160 mL
hexane and slowly allowed to cool to 48C. The product (7.37 g,
21.5 mmol, 92%) was obtained after filtration and extensive washing
with warm hexane as light beige crystals (m.p. 140–141.58C). 1H NMR
(CDCl3, 400 MHz): d=3.05 (brd, 6H), 3.88 (s, 3H), 7.82 (t, 1H),
8.16 ppm (d, 2H); 13C NMR (CDCl3, 100 MHz): d=36.42, 51.81, 129.64,
130.81, 136.65, 145.78, 165.07, 165.27 ppm; LC-MS: m/z: calcd: 365.0606;
found: 365.0616 [M+Na+].
Dimethylthiocarbamic acid S-(6-dimethylcarbamoylsulfanylanthracen-2-
yl) ester (8): Crude O-thiocarbamate 7 (8.0 g, 20.8 mmol) was suspended
in diphenyl ether (100 mL) under a nitrogen atmosphere and heated to
2308C for 3 h. The reaction mixture was allowed to slowly cool to room
temperature. The product was isolated by filtration (gentle heating to
melt the diphenyl ether may be necessary) and washed extensively with
hexane to afford 7.0 g (18.2 mmol, 88%) of beige crystals. The product
was pure enough to be used in the next step, but can be recrystallized
3,5-Dimercaptobenzoic acid (5): Under a nitrogen atmosphere S-thiocar-
bamate 11 (7.0 g, 20.4 mmol) was suspended in a 1.75m solution (70 mL)
of KOH in diethyleneglycol that had been purged with argon for 2 h. The
solution was heated at 1058C for 30 min. After the solution had cooled
down to room temperature 500 mL of water (purged) was added fol-
lowed by rapid addition of 10% HCl (55 mL). The precipitate was fil-
tered (or centrifuged in cases where the precipitate is fine), washed ex-
tensively with water and dried under vacuum overnight to give 3.50 g
(18.8 mmol, 92%) of a white powder. The product can be recrystallized
from degassed ethanol/water (2 g acid in a refluxing mixture of ethanol
(28 mL) and water (48 mL) under an inert atmosphere (m.p. 221–2238C).
1H NMR (CD3OD, 400 MHz): d=7.40 (s, 1H), 7.64 ppm (s, 2H);
13C NMR (CD3OD, 100 MHz): d=126.97, 132.69, 133.06, 134.87,
168.13 ppm; LC-MS: m/z: calcd: 185.9809; found: 185.9808 [M+].
1
from chloroform/acetone (m.p. 277–2798C). H NMR (CDCl3, 400 MHz):
d=3.06 (brs, 6H), 3.14 (brs, 6H), 7.50 (dd, 2H), 7.98 (d, 2H), 7.97 (d,
2H), 8.37 ppm (s, 2H); 13C NMR (CDCl3, 125 MHz): d=36.69, 126.23,
126.27, 128.48, 130.98, 131.34, 131.83, 135.11, 166.58 ppm; LC-MS: m/z:
calcd: 407.0854; found: 407.0859 [M+Na+].
Dimethyl
noanthracene-11,12-dicarboxylate (9): A mixture of the S-thiocarbamate
(7.0 g, 18.2 mmol) and dimethyl acetylenedicarboxylate (6.58 g,
2,6-bis(dimethylcarbamoylsulfanyl)-9,10-dihydro-9,10-ethe-
8
46.3 mmol) in diphenyl ether (40 mL) under a nitrogen atmosphere was
heated for 75 min at 1908C. The reaction was monitored by TLC (silica
gel, 10% acetonitrile in chloroform; Rf starting material: 0.35; Rf prod-
uct: 0.15). The reaction mixture was allowed to cool to room temperature
and hexane was added and the solvents were decanted to give an oily
product that solidified upon standing. The product was purified by chro-
matography (silica gel, gradient elution 5 to 10% acetonitrile in chloro-
form) yielding 8.08 g (15.3 mmol, 84%) of a yellow solid. Note: use of
ethanol-stabilized chloroform gives poor separation. An analytical
sample was obtained by recrystallization from acetone/petroleum ether
(60–80) (m.p. 195–1968C). 1H NMR (CDCl3, 400 MHz): d=3.06 (brs,
6H), 3.14 (brs, 6H), 7.50 (dd, 2H), 7.98 (d, 2H), 7.97 (d, 2H), 8.37 ppm
(s, 2H); 13C NMR (CDCl3, 100 MHz): d=38.48, 53.50, 53.97, 77.61,
125.99, 127.23, 132.32, 134.65, 145.57, 146.10, 148.10, 167.05, 168.37 ppm;
LC-MS: m/z: calcd: 549.1130; found: 549.1129 [M+Na+].
Dynamic combinatorial libraries: The relevant building blocks were dis-
solved in water, with sufficient 2.5m aqueous NaOH solution to fully de-
protonate the thiols and carboxylic acids on the building blocks, using
sonication where necessary. The pH was then adjusted to 8 using 1m
aqueous HCl and 2.5m aqueous NaOH solutions. Aliquots of 500 mL of
building block solutions were added to 2 mL HPLC vials, and mixed with
500 mL of a template solution and a magnetic stir bar. The vials were
then capped, and stirred for at least three weeks at 298 K.
Receptor 12a was isolated by preparative HPLC from DCLs prepared
using 10 mm 4 and 10 mm T13 or T11. Aliquots of 500 mL were chromato-
graphed (20 mLminꢀ1 50:50:0.1 H2O/MeCN/TFA, ambient temperature).
For each run, fractions were collected at 0.25 minute intervals, typically
from 14–17.75 minutes. Examination of the chromatograms of this region
revealed two overlapping peaks. Fractions containing pure 12a were
combined and the solvent was removed in vacuo. The material obtained
from 20 injections was re-constituted in 5 mL 1:1 MeCN/H2O, and re-
chromatographed (20 mLminꢀ1 50:50:0.1 H2O/MeCN/TFA, ambient tem-
perature) in one injection. The collected fractions corresponding to the
main peak were dried in vacuo, redissolved in 50 mm borate buffer using
the minimum number of 1.5 mL aliquots to achieve complete dissolution
(typically 1 mgmLꢀ1), and separated into 1.5 mL centrifuge tubes. Ali-
2,6-Dimercapto-9,10-dihydro-9,10-ethenoanthracene-11,12-dicarboxylic
acid (4): Under
a nitrogen atmosphere S-thiocarbamate 9 (2.0 g,
3.8 mmol) was suspended in a 1.75m solution (70 mL) of KOH in diethy-
leneglycol that had been purged with argon for 2 h. The solution was
heated at 1058C for 30 min. After the solution had cooled down to room
temperature 500 mL of water (purged) was added followed by rapid ad-
dition of 10% HCl (55 mL). The very fine precipitate was centrifuged,
washed extensively with water and dried under vacuum overnight to give
1.15 g (18.8 mmol, 85%) of a beige powder (m.p. decomp). 1H NMR
2164
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166

Dynamic Combinatorial Libraries
FULL PAPER
quots of 100 mL 2m HCl was added to each tube, the contents of the
tubes were vortexed, and the resultant suspensions were centrifuged for
15 minutes at 13,000 rpm. The pellets were washed twice by addition of
500 mL 40 mm HCl, resuspension by vortexing, and recentrifugation for
15 minutes at 13000 rpm. The final pellets were dried in vacuo overnight.
This procedure typically gave 4 mg of 12a per 10 mL of DCL solution.
1H NMR (500 MHz, CD3OD, 300 K): d=7.57 (s, 2H), 7.47 (s, 2H), 7.38
(s, 2H), 7.26 (d, 2H, J=7.6 Hz), 7.20 (m, 4H), 7.01 (d, 2H, J=7.9 Hz),
6.89 (m, 4H), 5.82 (s, 2H), 5.81 (s, 2H), 5.72 ppm (s, 2H); elemental
analysis calcd (%) for 12a + 6H2O: C 55.37, H: 3.61, N 0.00; found: C
55.37, H 3.46, N 0.00.
Chem. Soc. 2006, 128, 8459–8467; i) A. Valade, D. Urban, J. M.
Beau, ChemBioChem 2006, 7, 1023–1027; j) S. A. Poulsen, J. Am.
Soc. Mass Spectrom. 2006, 17, 1074–1080; k) S. Tsujita, M. Tanada,
T. Kataoka, S. Sasaki, Bioorg. Med. Chem. Lett. 2007, 17, 68–72;
l) A. Valade, D. Urban, J. M. Beau, J. Comb. Chem. 2007, 9, 1–4;
m) P. Vongvilai, M. Agnelin, R. Larsson, O. Ramstrçm, Angew.
Chem. 2007, 119, 966–968; Angew. Chem. Int. Ed. 2007, 46, 948–
950; n) F. Bulos, S. L. Roberts, R. L. E. Furlan, J. K. M. Sanders,
Chem. Commun. 2007, 3092–3093; o) J. Y. Liu, K. R. West, C. R.
Bondy, J. K. M. Sanders, Org. Biomol. Chem. 2007, 5, 778–786.
[5] a) A. Buryak, K. Severin, Angew. Chem. 2005, 117, 8149–8152;
Angew. Chem. Int. Ed. 2005, 44, 7935–7938; b) A. Buryak, K. Sever-
in, J. Comb. Chem. 2006, 8, 540–543.
[6] a) B. Brisig, J. K. M. Sanders, S. Otto, Angew. Chem. 2003, 115,
1308–1311; Angew. Chem. Int. Ed. 2003, 42, 1270–1273; b) L. Vial,
J. K. M. Sanders, S. Otto, New J. Chem. 2005, 29, 1001–1003.
[7] R. T. S. Lam, A. Belenguer, S. L. Roberts, C. Naumann, T. Jarrosson,
S. Otto, J. K. M. Sanders, Science 2005, 308, 667–669.
[8] a) Z. Grote, R. Scopelliti, K. Severin, Angew. Chem. 2003, 115,
3951–3955; Angew. Chem. Int. Ed. 2003, 42, 3821–3825; b) K. Se-
verin, Chem. Eur. J. 2004, 10, 2565–2580; c) I. Saur, K. Severin,
Chem. Commun. 2005, 1471–1473.
[9] a) P. T. Corbett, S. Otto, J. K. M. Sanders, Chem. Eur. J. 2004, 10,
3139–3143; b) P. T. Corbett, S. Otto, J. K. M. Sanders, Org. Lett.
2004, 6, 1825–1827; c) P. T. Corbett, J. K. M. Sanders, S. Otto, J. Am.
Chem. Soc. 2005, 127, 9390–9392; d) P. T. Corbett, J. K. M. Sanders,
S. Otto, Angew. Chem. 2007, 119, 9014–9017; Angew. Chem. Int. Ed.
2007, 46, 8858–8861.
[10] S. Otto, R. L. E. Furlan, J. K. M. Sanders, Science 2002, 297, 590–
593.
[11] S. Otto, R. L. E. Furlan, J. K. M. Sanders, J. Am. Chem. Soc. 2000,
122, 12063–12064.
Receptor 13a was isolated by preparative HPLC from DCLs prepared
using 6.66 mm 4, 3.33 mm 5 and 10 mm T2. The DCLs were diluted 1:1
with MeCN + 0.4% TFA, and 5 mL aliquots were chromatographed
(20 mLminꢀ1 55:30:15:0.1 H2O/MeCN/IPA/TFA, retention time typically
15.5–20 min). The collected fractions from four such injections were com-
bined, dried in vacuo and redissolved in 10 mL 1:1 MeCN/H2O. Aliquots
of 500 mL of this solution were chromatographed again to separate 13a
from 13b (20 mLminꢀ1 50:50:0.1 H2O/MeCN/TFA). For each run, frac-
tions were collected at 0.25 minute intervals, from 12.5–20.0 minutes. Ex-
amination of the chromatograms of this region revealed two overlapping
peaks. Fractions containing pure 13a were combined and the solvent was
removed in vacuo. The material obtained from 20 injections was re-con-
stituted in 5 mL 1:1 MeCN/H2O, and rechromatographed a third time
(20 mLminꢀ1 50:50:0.1 H2O/MeCN/TFA) in a single injection. The col-
lected fractions corresponding to the main peak were dried in vacuo, and
reprecipitated as described for 12a. This procedure typically gave 1 mg of
13a per 10 mL of DCL solution. 1H NMR (500 MHz, CD3OD): d=5.62
(s, 2H), 5.67 (s, 2H), 7.11 (d, 2H), 7.19 (d, 2H), 7.24 (m, 4H), 7.56 (s,
2H), 7.61 (s, 2H), 7.91 (s, 2H), 8.07 ppm (s, 1H); elemental analysis
calcd (%) for 13a + 4H2O: C 53.51, H 3.34; found: C 53.27, H 3.39; LC-
MS: m/z: calcd for: 914.9591; found: 914.9602 [M+Na+].
[12] O. Ramstrçm, J.-M. Lehn, ChemBioChem 2000, 1, 41–48.
[13] a) K. Odashima, A. Itai, Y. Iitaka, K. Koga, J. Am. Chem. Soc. 1980,
102, 2504–2505; b) K. Odashima, T. Soga, K. Koga, Tetrahedron
Lett. 1981, 22, 5311–5314; c) K. Odashima, A. Itai, Y. Iitaka, K.
Koga, J. Org. Chem. 1985, 50, 4478–4484; d) J. Winkler, E. Coutou-
liargyropoulou, R. Leppkes, R. Breslow, J. Am. Chem. Soc. 1983,
105, 7198–7199.
The mixture of13a and 13b for the experiments described in Figure 3
was obtained using only the first chromatographic step described above.
Acknowledgements
[14] a) M. A. Petti, T. J. Shepodd, D. A. Dougherty, Tetrahedron Lett.
1986, 27, 807–810; b) T. J. Shepodd, M. A. Petti, D. A. Dougherty, J.
Am. Chem. Soc. 1986, 108, 6085–6087; c) T. J. Shepodd, M. A. Petti,
D. A. Dougherty, J. Am. Chem. Soc. 1988, 110, 1983–1985; d) D. A.
Stauffer, D. A. Dougherty, Tetrahedron Lett. 1988, 29, 6039–6042;
e) M. A. Petti, T. J. Shepodd, R. E. Barrans, D. A. Dougherty, J. Am.
Chem. Soc. 1988, 110, 6825–6840; f) D. A. Dougherty, D. A. Stauff-
er, Science 1990, 250, 1558–1560; g) D. A. Stauffer, R. E. Barrans,
D. A. Dougherty, J. Org. Chem. 1990, 55, 2762–2767; h) D. A.
Stauffer, R. E. Barrans, D. A. Dougherty, Angew. Chem. 1990, 102,
953–956; Angew. Chem. Int. Ed. Engl. 1990, 29, 915–918; i) A.
McCurdy, L. Jimenez, D. A. Stauffer, D. A. Dougherty, J. Am.
Chem. Soc. 1992, 114, 10314–10321; j) P. C. Kearney, L. S. Mizoue,
R. A. Kumpf, J. E. Forman, A. McCurdy, D. A. Dougherty, J. Am.
Chem. Soc. 1993, 115, 9907–9919; k) J. E. Forman, R. E. Marsh,
W. P. Schaefer, D. A. Dougherty, Acta Crystallogr. Sect. A 1993, 49,
892–896; l) J. E. Forman, R. E. Barrans, D. A. Dougherty, J. Am.
Chem. Soc. 1995, 117, 9213–9228; m) S. M. Ngola, D. A. Dougherty,
J. Org. Chem. 1996, 61, 4355–4360; n) S. M. Ngola, P. C. Kearney, S.
Mecozzi, K. Russell, D. A. Dougherty, J. Am. Chem. Soc. 1999, 121,
1192–1201.
[15] We believe that upon deprotonating of the thiol rapid nucleophilic
attack takes place on a methyl group on the quaternary nitrogen
center, resulting in the irreversible methylation of the thiol.
[16] L. Field, P. R. Engelhardt, J. Org. Chem. 1970, 35, 3647–3655.
[17] P. T. Corbett, L. H. Tong, J. K. M. Sanders, S. Otto, J. Am. Chem.
Soc. 2005, 127, 8902–8903.
[18] Due to problems with overlapping and tailing peaks in the prepara-
tive HPLC, it was not possible to obtain sufficient quantities of the
minor diastereomers 12b and 13b to perform the corresponding
We are grateful to Ana Belenguer for valuable assistance with the HPLC
analyses and isolation and to the Royal Society and EPSRC for financial
support.
[1] a) S. Otto, Dalton Trans. 2006, 2861–2864; b) Z. Rodriguez-Docam-
po, S. I. Pascu, S. Kubik, S. Otto, J. Am. Chem. Soc. 2006, 128,
11206–11210.
[2] J. K. M. Sanders, Chem. Eur. J. 1998, 4, 1378–1383.
[3] For reviews, see: a) C. Karan, B. L. Miller, Drug Discovery Today
2000, 5, 67–75; b) J.-M. Lehn, A. V. Eliseev, Science 2001, 291,
2331–2332; c) O. Ramstrçm, T. Bunyapaiboonsri, S. Lohmann, J.-M.
Lehn, Biochim. Biophys. Acta 2002, 1572, 178–186; d) B. de Bruin,
P. Hauwert, J. N. H. Reek, Angew. Chem. 2006, 118, 2726–2729;
Angew. Chem. Int. Ed. 2006, 45, 2660–2663; e) P. T. Corbett, J. Lec-
laire, L. Vial, K. R. West, J.-L. Wietor, J. K. M. Sanders, S. Otto,
Chem. Rev. 2006, 106, 3652–3711.
[4] For recent references, see: a) L. Vial, R. F. Ludlow, J. Leclaire, R.
Perez-Fernandez, S. Otto, J. Am. Chem. Soc. 2006, 128, 10253–
10257; b) S. M. Voshell, S. J. Lee, M. R. Gagne, J. Am. Chem. Soc.
2006, 128, 12422–12423; c) M. Bru, I. Alfonso, M. I. Burguete, S. V.
Luis, Angew. Chem. 2006, 118, 6301–6305; Angew. Chem. Int. Ed.
2006, 45, 6155–6159; d) B. R. McNaughton, B. L. Miller, Org. Lett.
2006, 8, 1803–1806; e) L. Milanesi, C. A. Hunter, S. E. Sedelnikova,
J. P. Waltho, Chem. Eur. J. 2006, 12, 1081–1087; f) B. Danieli, A.
Giardini, G. Lesma, D. Passarella, B. Peretto, A. Sacchetti, A. Silva-
ni, G. Pratesi, F. Zunino, J. Org. Chem. 2006, 71, 2848–2853; g) S. A.
Poulsen, L. F. Bornaghi, Bioorg. Med. Chem. 2006, 14, 3275–3284;
h) B. L. Shi, R. Stevenson, D. J. Campopiano, M. F. Greaney, J. Am.
Chem. Eur. J. 2008, 14, 2153 – 2166
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
2165

S. Otto et al.
analyses for these compounds. Furthermore, we were not able to
record the binding of T7 to 12a by ITC, as the binding appeared to
be almost enthalpy-neutral, leading to an extremely weak signal.
[19] S. Otto, J. B. F. N. Engberts, Org. Biomol. Chem. 2003, 1, 2809–2820.
[20] See Supporting Information for the derivation of this relationship.
[21] R. F. Ludlow, J. Liu, H. Li, S. L. Roberts, J. K. M. Sanders, S. Otto,
Angew. Chem. 2007, 119, 5864–5866; Angew. Chem. Int. Ed. 2007,
46, 5762–5764.
[22] Y. Kondo, R. Uematsu, Y. Nakamura, S. Kusabayashi, J. Chem. Soc.
Faraday Trans. 1 1988, 84, 111–116.
[23] J. Lacour, L. Vial, C. Herse, Org. Lett. 2002, 4, 1351–1354.
[24] P. Boldt, Chem. Ber. 1967, 100, 1270–1280.
Received: September 6, 2007
Published online: December 14, 2007
2166
www.chemeurj.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2008, 14, 2153 – 2166