Avidin and streptavidin ligands based on the glycoluril bicyclic system

Article abstract of DOI:10.1039/b605081f

Glycoluril derivatives with a carboxylic acid side chain have been synthesized and shown to bind to both avidin and streptavidin. Introduction of a valerate side chain in glycoluril led to an increased binding to both proteins only when the valerate group was bound to a N atom and with the proper stereochemistry [(+)-enantiomer]. On the other hand, introduction of the valerate side chain either on the bridgehead carbon or on the N atom with the opposite stereochemistry [(-)-enantiomer] led to a decrease in binding constant compared with unsubstituted glycoluril. Direct spectrophotometric competitive titration of each protein with a racemic ligand allowed measurement of the enantioselectivity of the ligand-protein complexation, together with the binding constant of the two enantiomers. In the case of the N-substituted glycoluril, the extension of the side chain by one methylene group, from valerate to caproate, led to an increase in the binding constant to both proteins. Docking studies using AutoDock 3.05 have been performed in order to predict the binding modes of these ligands to streptavidin. The effect of the stereochemistry and the position of the side chain on the binding constant to streptavidin is discussed in view of the predicted binding modes. The Royal Society of Chemistry 2006.

Full text of DOI:10.1039/b605081f

View Article Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/obc | Organic & Biomolecular Chemistry
Avidin and streptavidin ligands based on the glycoluril bicyclic system†
Pedro Hidalgo-Fern a´ ndez, Eva Ayet, Ivan Canal and Joan-Antoni Farrera*
Received 7th April 2006, Accepted 26th June 2006
First published as an Advance Article on the web 14th July 2006
DOI: 10.1039/b605081f
Glycoluril derivatives with a carboxylic acid side chain have been synthesized and shown to bind to
both avidin and streptavidin. Introduction of a valerate side chain in glycoluril led to an increased
binding to both proteins only when the valerate group was bound to a N atom and with the proper
stereochemistry [(+)-enantiomer]. On the other hand, introduction of the valerate side chain either on
the bridgehead carbon or on the N atom with the opposite stereochemistry [(−)-enantiomer] led to a
decrease in binding constant compared with unsubstituted glycoluril. Direct spectrophotometric
competitive titration of each protein with a racemic ligand allowed measurement of the
enantioselectivity of the ligand–protein complexation, together with the binding constant of the two
enantiomers. In the case of the N-substituted glycoluril, the extension of the side chain by one
methylene group, from valerate to caproate, led to an increase in the binding constant to both proteins.
Docking studies using AutoDock 3.05 have been performed in order to predict the binding modes of
these ligands to streptavidin. The effect of the stereochemistry and the position of the side chain on the
binding constant to streptavidin is discussed in view of the predicted binding modes.
Introduction
Avidin (Av) and streptavidin (Sav) are tetrameric proteins, with one
1
binding site per subunit, widely used in biochemical applications.
The basis of this prevalent use is the high association constant of
1
5
−1
the complexes that biotin (1) forms with both Av (K
a
≈ 10 M )
1
3
−1
and Sav (K
a
≈ 2.5 × 10 M ), which makes the binding process
2
considered to be almost irreversible. This strong binding between
biotin (1) and Av or Sav has promoted the use of these systems in
3–10
the construction of different supramolecular structures.
Binding studies of biotin analogues to Av and Sav have
suggested a major contribution from the ureido group, with a lower
contribution from the valerate side chain and the thiolane ring.
Av and Sav can also bind a large number of molecules with lower
17
11–15
affinity than for biotin. One of these compounds is glycoluril (2a),
which was shown to bind Sav with an association constant of 4 ×
0 M . X-Ray diffraction studies of the complex Sav–glycoluril
X-Ray diffraction studies of both Av-1 and Sav-1 complexes
showed a similar network of hydrogen-bonding interactions
between the ureido group of biotin (1) and five protein residues,
whereas the carboxylate group of biotin formed hydrogen-bonding
interactions with five protein residues in the case of Av and
with only two protein residues in the case of Sav. This different
hydrogen-bonding network of the carboxylate, together with
differences in the hydrophobic interactions to aromatic residues, is
consistent with the stronger binding of biotin (1) to Av compared
5
−1 18
1
showed hydrogen-bonding interactions between one of the ureido
groups of glycoluril (2a) and the same five protein residues involved
in the binding of biotin (1). The oxygen atom of the second ureido
group of glycoluril also contributes to the binding with a hydrogen-
bonding interaction to the same protein residue that binds to the
sulfur atom of biotin. Considering the similar hydrogen-bonding
pattern in both Sav–biotin and Sav–glycoluril complexes, it was
concluded that the differences in binding constant between both
complexes predominantly reflected the missing interactions due to
the absence of the valerate side chain in glycoluril (2a).
16
with Sav.
The objective of the present work was to quantify the contri-
◦
bution to the DG of the binding of a carboxylic acid side chain
that is covalently linked to either a bridgehead carbon atom or
to a nitrogen atom of glycoluril. For that reason we synthesized
Departament de Quimica Organica, Facultat de Qu ´ı mica, Universitat de
Barcelona, Mart ´ı i Franqu e´ s 1, 08028-Barcelona, Spain. E-mail: jfarrera@
ub.edu
5
-(3,7-dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-1-yl)pentanoic acid
†
Electronic supplementary information (ESI) available: binding models
(
2b), 5-(3,7-dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-2-yl)pentanoic
for competitive titrations with ligands, and binding isotherms for competi-
tive titrations of (S)Av–HABA complexes with glycoluril-type ligands. See
DOI: 10.1039/b605081f.
19
acid (2c), and 6-(3,7-dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-2-
yl)hexanoic acid (2d).
This journal is © The Royal Society of Chemistry 2006
Org. Biomol. Chem., 2006, 4, 3147–3154 | 3147

View Article Online
◦
Scheme 1 Reagents and conditions: (i) SOCl
2
; (ii) CH
2
N
2
, hexane–Et
2 2 2 2
O (1 : 1), rt,1 h; (iii) dimethyldioxirane, acetone, 0 C; (iv) H NCONH , H O–EtOH
◦
◦
(
2 : 1), pH = 3, 4 C, 3 d; (v) pH = 1, 4 C, 5 d; (vi) KOH–H
2
O, rt, 24 h; (vii) pH = 1–2.
Results and discussion
respectively. Unlike compound 2b, which is achiral, compounds 2c
and 2d are chiral. Since the two enantiomers of a racemic ligand
are expected to have different binding constants to a given protein,
a small amount of racemic 2c was resolved into its enantiomers by
Compound 2b was obtained from 5-ethoxycarbonylpentanoic acid
by reacting the corresponding acyl chloride with diazomethane to
give the diazoketone 3 in 75% yield (Scheme 1). The diazoketone
23
RP-HPLC using a chiral stationary phase.
3
was subsequently oxidized with dimethyldioxirane in acetone to
In order to prove that these ligands bind to Av and Sav we used
yield glyoxal 4, together with the corresponding hydrate.
ꢀ
the dye 2-(4 -hydroxyphenylazo)benzoic acid (HABA), which has
Attempts to condense compound 4 to urea in standard
conditions for the synthesis of glycolurils (toluene or benzene,
with a catalytic amount of TFA and azeotropic removal of
water) didn’t afford glycoluril 5 due to competing side reactions
involving glyoxal 4 (Table 1). These side reactions of glyoxal
been shown to bind to both Av and Sav at the same binding site as
biotin. This dye, in its unbound form, has an absorption maximum
at 348 nm, which shifts to 500 nm when bound to Av or Sav.
Addition of any of these ligands to an aqueous solution of HABA
and either Av or Sav resulted in a decrease in the absorption
at 500 nm and the corresponding increase in the absorption at
24
4
could be minimized by reacting urea and glyoxal 4 at low
temperature. However, in spite of water being formed in the
condensation of urea and glyoxals, this type of reaction can
proceed in aqueous or hydroalcoholic media with moderate to
3
48 nm, proving that all of them bind to the same binding sites as
HABA and biotin. Since the binding constants of HABA to Av
and Sav were already known (K
5
3
−1
a
= 1.7 × 10 and 7.3 × 10 M
20–22
good yields.
Depending on the acidity of the aqueous medium
1,13
respectively), the binding constants of the new ligands to Av and
Sav should also be measurable in spectrophotometric competition
experiments with HABA. For every ligand, the plot of absorbance
at 500 nm (A500) vs. the ratio [ligand]/[subunits] was fitted to a 1 : 1
binding model assuming that each subunit of the protein behaves
the condensation of urea and glyoxal affords either glycoluril,
when pH = 1–2, or the intermediate 4,5-dihydroxyimidazolidin-
2
-one, when pH = 3–4. The best result, 28% yield, was obtained
2
when urea and glyoxal 4 were allowed to react in EtOH–H
O
◦
(
1 : 2) at 4 C in two stages: first at pH = 3 in order to
25,26
as an independent binding site, without cooperative effects.
form the corresponding 4,5-dihydroxyimidazolidin-2-one, and
then at pH = 1 in order to form glycoluril 5 (Table 1). Finally,
saponification of the ester group of 5 afforded ligand 2b in almost
quantitative yield.
The reasonably good fits obtained (shown in Fig. 1–3 and in the
ESI†) between the experimental and calculated curves support our
assumption of independent behaviour between the binding sites.
In order to compare binding constants of different compounds it
is preferable that all of them are measured with the same technique
and under the same experimental conditions. For that reason we
measured the binding constant of glycoluril (2a) to Sav by the
spectrophotometric competitive titration method, yielding a value
which was circa one order of magnitude lower than the value
19
Compounds 2c and 2d were obtained from x-aminoacids
which, by reaction with potassium cyanate in aqueous medium, af-
forded the corresponding x-ureidocarboxylic acids. Condensation
of these x-ureidocarboxylic acids to 4,5-dihydroxyimidazolidin-
21
2
-one afforded the ligands 2c and 2d in 45 and 50% yield
Table 1 Optimization of the synthesis of 5
◦
Entry
Solvent
Catalyst
Temperature/ C
Time
Yield (%)
1
2
3
4
Benzene
TFA
Reflux
16 h
16 h
24 h
3 d
ca. 1
ca. 1
5
H
H
H
2
2
2
O–EtOH (2 : 1)
O–EtOH (2 : 1)
O–EtOH (2 : 1)
HCl (pH = 1)
HCl (pH = 1)
(i) HCl (pH = 3)
50
25
4
28
(
ii) HCl (pH = 1)
4
5 d
3
148 | Org. Biomol. Chem., 2006, 4, 3147–3154
This journal is © The Royal Society of Chemistry 2006

View Article Online
a
Table 2 Binding constants of glycoluril and its derivates to avidin and streptavidin
Avidin
Streptavidin
−
1
DG/kJ mol−1
−1
−1
Ligand
K
a
/M
S
K
a
/M
DG/kJ mol
S
6
3
7
7
5
5
7
5
4
2
2
a
b
2.2 (± 0.6) × 10
8.5 (± 1.0) × 10
2.6 (± 1.0) × 10
−36.2 ± 0.7
−22.4 ± 0.3
−42.3 ± 1.0
4.7 (± 0.5) × 10
76 ± 10
−26.5 ± 0.3
−10.7 ± 0.3
−35.7 ± 1.2
−36.8 ± 0.9
−23.1 ± 0.5
−21.9 ± 0.7
−38.0 ± 1.7
−24.1 ± 1.2
6
(
(
(
+)-2c
−)-2c
± )-2d
217
103
1.8 (± 0.8) × 10
2.8 (± 1.0) × 10
1.1 (± 0.2) × 10
7.0 (± 2.0) × 10
4.5 (± 2.0) × 10
1.7 (± 1.0) × 10
164
265
b
b
6 b
b
3
.5 (± 2.0) × 10
−43.0 ± 1.6
4
1.2 (± 0.3) × 10
−29.0 ± 0.6
b
b
b
b
3 b
6 b
4 b
b
b
b
2
.6 (± 2.0) × 10
4.1 (± 2.0) × 10
.0 (± 2.0) × 10
−30.9 ± 2.5
b
−43.4 ± 1.3
b
4
−32.0 ± 1.4
a
b
Determined by spectrophotometric competition with HABA, in 0.1 M phosphate buffer at pH = 7.0. Obtained from the competitive tritation of the
protein–HABA complex with the racemic ligand.
obtained by surface plasmon resonance using other experimental
18
conditions (Table 2). On the other hand, owing to the lack of a
carboxylic acid side chain in glycoluril (2a), it was expected that
a similar binding constant to both Av and Sav would be found.
However, the binding constant to Av resulted in being almost
two orders of magnitude higher than to Sav, as also happens in
1,14,24,27
the complexes of biotin and other ligands.
In the case of
glycoluril this difference in the binding constants to both proteins
can only be ascribed to differences in the hydrophobic and van
der Waals interactions of the bicyclic system, and to differences
in the hydrogen-bonding interactions involving the ureido groups.
Analogously, in the case of biotin, apart from the well established
difference of interactions of the valerate side chain towards Av
16
and Sav, the interactions involving the bicyclic system of biotin
could also make an important contribution to the difference of its
binding constant to both proteins.
The glycoluril derivatives with a valerate side chain bound to a
bridgehead carbon atom (2b) or to a nitrogen atom [(+)-2c and
(
−)-2c] also have a higher binding affinity to Av than to Sav (Fig. 1
and Table 2). However, the difference in the DG of binding (DDG)
for a given ligand to Av and Sav is about two times smaller in the N-
substituted glycolurils (+)-2c and (−)-2c than in the C-substituted
glycoluril 2b (6.6, 5.9 and 11.7 kJ mol− respectively).
1
Fig. 2 Absorption change at 500 nm in the titrations of the complex
Sav–HABA with ligands 2c in phosphate buffer 0.1 M at pH = 7.0. A:
Sav 11.8 lM (tetramer) and HABA 59.7 lM with ligand (−)-2c. B: Sav
1
1.9 lM (tetramer) and HABA 59.7 lM with ligand (+)-2c. C: Sav 12.1 lM
Fig. 1 Absorption change at 500 nm in the titration of Av 9.2 lM
(tetramer) and HABA 67.6 lM with ligand (± )-2c. The solid line represents
the fit of the data to the (1 : 1) binding model, and to the model considering
different binding constants for each enantiomer in titration C.
(
tetramer) and HABA 34.7 lM with ligand 2b in phosphate buffer 0.1 M,
pH = 7.0. The solid line represents the fit of the data to the (1 : 1) binding
model.
influence on the free energy of binding of the resulting ligand to
Av and Sav. The binding is weakest when the valerate side chain is
bound to a glycoluril bridgehead carbon (2b). On the other side,
the strongest binding is obtained when the valerate side chain is
bound to a glycoluril nitrogen atom with the proper configuration
in the stereogenic centers [(+)-2c]. The decrease in the free energy
For the chiral compounds 2c the (+)-enantiomer was found to
bind to both Av and Sav about 13 kJ mol stronger than the
−
1
(
−)-enantiomer (Fig. 2 and Table 2), which is an indication of
the better fit of the (+)-enantiomer inside the binding pocket of
both proteins. The position and stereochemistry of the bicyclic
system where the valerate side chain is bound has an important
−
1
of binding on going from 2b to (+)-2c is of 19.9 kJ mol for
This journal is © The Royal Society of Chemistry 2006
Org. Biomol. Chem., 2006, 4, 3147–3154 | 3149

View Article Online
−
1
Av and of 25.0 kJ mol for Sav. Smaller decreases in the free
consuming, stage of racemic resolution. Finally, titration with the
enantiomerically pure optimal ligands would allow confirmation
of the previously found binding constants as well as to their
assignment to the corresponding enantiomer of the ligand.
energy of binding are obtained on going from 2b to (−)-2c, being
−
1
−1
6
.6 kJ mol for Av and of 12.4 kJ mol for Sav. It is clear from
these results that introduction of a valerate side chain in glycoluril
will lead to an increase in the binding constant to Av and Sav only
when the side chain is bound to a nitrogen atom and with the
proper stereochemistry [(+)-enantiomer]. This increase in binding
afinitty could be due to new hydrogen-bonding interactions of the
carboxylate group together with van der Waals interactions of the
We have applied this methodology to the competitive spec-
trophotometric determination of binding constants, although its
general character makes it also applicable to other methods,
either direct or competitive. In order to check this methodology
we first used racemic 2c, because it had already been resolved
and the binding constants of both enantiomers to Av and Sav
had been independently measured. Competitive titration of the
complex Sav–HABA with racemic 2c afforded an experimental
curve (Fig. 2C) that was fitted to a 1 : 1 binding model. But this
time the binding model included two different binding constants –
one for each enantiomer of the ligand. Non-linear fitting of the
calculated absorbance to the experimental absorbance at 500 nm
25,28
valerate side chain.
In other cases, either with the valerate side
chain bound to the bridgehead carbon or to the nitrogen atom
with the wrong stereochemistry [(−)-enantiomer], there will be a
decrease in binding constant compared to unsubstituted glycoluril
2a).
Although the side chain in 2b and 2c – valerate – is the same as
in biotin (1), the constitution and stereochemistry at the linkage
point between the bicyclic system and the side chain is different
for these compounds. Owing to these differences, the valerate
side chain might not be the most appropriate one for substituted
glycolurils in order to make hydrogen-bonding interactions to
specific residues of Av and Sav. In the case of biotin it has been
found that shortening or extending the carboxylic acid side chain
by a single methylene group leads to a markedly different network
of hydrogen-bonding interactions with Av and Sav. For that
reason we were interested in finding out whether an extension of
the carboxylic acid side chain of N-substituted glycolurils would
lead to an increase in the binding constant to Av or to Sav. Ligand
(
6
3
−1
led to the binding constants of 2.8 × 10 and 7.0 × 10 M for
the two enantiomers, which agree reasonably well with the values
obtained in the titrations with the pure enantiomers (Table 2).
In order to further check this method with another protein, the
complex Av–HABA was titrated with racemic 2c. Non-linear
fitting of the experimental curve (see ESI†) to the same 1 : 1 binding
7
5
−1
model led to the binding constants of 3.5 × 10 and 2.6 × 10 M
16
for the two enantiomers (Table 2). The reasonably good agreement
between the binding constants obtained by titration with each
enantiomerically pure ligand and by a single titration with the
corresponding racemate, with two different proteins, supports the
validity of this methodology.
2
d, which has five methylene groups in the side chain, was obtained
as a racemate. We were interested to know, without having to
ressolve the racemic 2d, whether any one of its enantiomers would
bind to Av or to Sav stronger than the enantiomers of 2c, and also
the enantioselectivity of the binding to each protein.
When this methodology was applied to the competitive titration
7
−1
of Av–HABA with racemic 2d, a Kmajor of 4.1 × 10 M and a
5
−1
K
minor of 4.0 × 10 M were obtained. In the most likely case
where the Kmajor corresponded to the 2d-enantiomer with the
same stereochemistry as (+)-2c, the extension of the side chain
by one methylene group would lead to an increase in the binding
constant of both enantiomers to Av. The increase in stability
(DDG) of the Av complexes would be higher for the weak-binding
There are many examples in the literature on the simultaneous
determination of two binding constants in a single titration of
an enantiomerically pure host with a racemic guest, or vice versa
an enantiomerically pure guest with a racemic host, when the
two diastereomeric complexes show separate signals. This is the
−
1
enantiomer (3.0 kJ mol ) than for the strong-binding enantiomer
1
29
−1
case in H NMR spectroscopy of some host–guest complexes
and affinity capillary electrophoresis of some diastereomeric ion-
pairs.
(1.1 kJ mol ).
In an analogous competitive titration of Sav–HABA with
3
0,31
1
6
−1
H NMR titration of a racemic host with a racemic
racemic 2d (Fig. 3), a Kmajor of 4.5 × 10 M and a Kminor of 1.7 ×
4 −1
guest leading to an averaged chemical shift can also give the two
binding constants of the diastereomeric complexes, but only when
the ratio between both binding constants can be obtained from
10 M were obtained. Assuming that the Kmajor corresponded to
the 2d-enantiomer with the same stereochemistry than (+)-2c, the
32
a separate experiment. However, it is very common to have
a racemic ligand that binds to an enantiomerically pure chiral
receptor (e.g. a protein) leading to a single averaged response for
the two diastereomeric complexes. In this case it is also possible to
simultaneously obtain the binding constants of both enantiomeric
ligands by means of a single titration of the enantiomerically pure
33
receptor with the racemic ligand. Obviously this methodology
does not tell us to which enantiomer of the ligand each binding
constant corresponds. But it has the great advantage of giving
us the binding constant of both enantiomeric ligands, as well
as the enantioselectivity, in a single titration of the receptor
with the racemic ligand, without having to separate the two
enantiomers. The application of this methodology should allow
the fast screening of racemic ligands. Only those racemic ligands
showing the appropriate binding constants (or activities) and
enantioselectivity would proceed to the next, time- and resources-
Fig. 3 Absorption change at 500 nm in the titration of Sav 10.6 lM
(
0
tetramer) and HABA 52.7 lM with ligand (± )-2d in phosphate buffer
.1 M, pH = 7.0. The solid line represents the fit of the data to the
(1 : 1) binding model considering different binding constants for each
enantiomer.
3
150 | Org. Biomol. Chem., 2006, 4, 3147–3154
This journal is © The Royal Society of Chemistry 2006

View Article Online
extension of the side chain by one methylene group would also
lead to an increase in the binding constant of both enantiomers to
Sav. But now the increase in stability (DDG) of the Sav complexes
might not be long enough to form hydrogen-bonding interactions
with Asn 49 and Ser 88. On the other hand, the most populated
cluster of docked 2b, which did not contain the lowest energy
conformation, had binding modes in which the bicyclic system of
2b was slightly tilted in relation to that of biotin (Fig. 4B). The
lower stability of this binding mode might come from less effective
hydrogen bonding and van der Waals interactions.
When (1R,5S)-2c and (1R,5S)-2d, the ligands with a stereo-
chemistry more similar to that of biotin (1), were docked to Sav,
the predicted binding mode showed for both ligands an orientation
within the binding site similar to that of crystallographic biotin
(1) (Fig. 4C). For these ligands the best scored conformation
corresponded to the most populated cluster.
In the case of (1S,5R)-2c, maintaining the valerate side chain
in a position similar to that of bound biotin would imply that
its bicyclic system could not be superimposable to that of its
enantiomer, or to that of bound biotin. Instead, the docking
results predicted that the bicyclic system of (1S,5R)-2c would be
almost superimposable to that of its enantiomer (Fig. 4D), while
the valerate side chain would lie in a different position to that of
its enantiomer, or to that of biotin. In this binding mode, appart
from potentially less favourable van der Waals interactions, one
hydrogen-bonding interaction might be lost because of the valerate
side chain substitution on the N atom close to Ser 45.
−
1
would be higher for the strong-binding enantiomer (2.3 kJ mol )
−
1
than for the weak-binding enantiomer (1.0 kJ mol ).
Apart from this increase in the binding constant to both Av
and Sav, the extension of the side chain by one methylene group
leads also to changes in the enantioselectivity S (defined as the
ratio of the binding constants of both enantiomers to a given
receptor). While in the case of Av, the extension of the side chain
by one methylene group leads to decrease in the enantioselectivity,
in the case of Sav it has the opposite effect (Table 2). This different
behaviour is consistent with the markedly different effect of the
extension of the side chain of biotin on the network of hydrogen-
16
bonding interactions to Av, compared to Sav.
We were interested in predicting the binding mode of ligands 2a–
d to Sav, in order to explain the major effects of the stereochemistry
and the position of the side chain on the binding constant.
Docking programs have been used to predict the binding site
and bound conformation of flexible ligands to macromolecules
of known structure, which are maintained rigid. Among these
34–36
programs, AutoDock
(Lamarckian genetic algorithm) has
been successfully used in reproducing the experimental binding
37
conformation of a wide range of ligand–protein complexes.
In this sense, AutoDock has been shown to reproduce the
The results obtained by docking studies could explain the major
differences in binding affinities, although it is difficult to extract
conclusions about the predicted absolute binding free energies and
hydrogen-bonding distances because the structure of the protein
was maintained rigid. More detailed studies based on molecular
dynamic methods are in progress.
experimentally observed binding conformation of biotin (1) to
34
˚
Sav with a rmsd of 0.66 A. Since the binding of biotin (1) to Sav
12,38
leads to an ordering of a surface protein loop,
we assumed that
the protein conformation in the Sav-2 complexes would be more
similar to that in the biotin complex (Sav-1) than to that in the
uncomplexed Sav. Consequently, we proceeded to dock ligands
2
a–d to the complexed Sav conformation.
In the first place, biotin (1) and glycoluril (2a) were docked to
Conclusion
Sav in order to test the reliability of the docking formalism and of
the assumed protein conformation. The predicted binding mode
of biotin (1) and glycoluril (2a) reproduced their crystallographic
In this paper we describe the synthesis of glycoluril derivatives with
a carboxylic acid side chain, showing that they bind to both avidin
and streptavidin at the same binding site as biotin. Glycoluril
derivatives with a valerate side chain bound either to a bridgehead
carbon atom or to a nitrogen atom have a higher binding affinity
to Av than to Sav. Moreover, the position and stereochemistry
of the bicyclic system where the valerate side chain is bound has
an important influence on the binding affinity of the resulting
ligand. Introduction of a valerate side chain in glycoluril led to an
increased binding to both proteins only when the valerate group
was bound to a N atom and with the proper stereochemistry [(+)-
enantiomer]. On the other hand, introduction of the valerate side
18,38
˚
with rmsd values of 0.37 and 1.29 A
complex coordinates
respectively. When 2b was docked to Sav, the best scored binding
◦
mode had the bicyclic system rotated approximately 107 about the
bridge bond, compared with the bicyclic systems of bound biotin
(
1) and bound glycoluril (2a). In this binding mode (Fig. 4A) only
one of the urea groups in 2b was almost superimposable to that
in biotin (1) or in glycoluril (2a), with the possibility of forming
similar hydrogen-bonding interactions. Besides, the valerate side
chain, although following a similar direction to that in biotin, in 2a
Fig. 4 Superimposition of binding modes of ligands 2b–d with Sav as found by AutoDock. A: lowest energy binding mode of 2b (yellow) and
crystallographic biotin (green); B: most populated cluster of 2b (yellow) and crystallographic biotin (green); C: lowest energy binding mode of (1R,5S)-2d
(
red) and crystallographic biotin (green); D: lowest energy binding modes of (1R,5S)-2c (blue) and (1S,5R)-2c (orange).
This journal is © The Royal Society of Chemistry 2006
Org. Biomol. Chem., 2006, 4, 3147–3154 | 3151

View Article Online
chain either on the bridgehead carbon or on the N atom with
the opposite stereochemistry [(−)-enantiomer] led to a decrease in
binding constant compared to the unsubstituted glycoluril. Direct
spectrophotometric competitive titration of each protein with a
racemic ligand allowed measurement of the enantioselectivity
of the ligand–protein complexation, together with the binding
constant of the two enantiomers. In the case of the N-substituted
glycoluril, the extension of the side chain by one methylene group,
from valerate to caproate, led to an increase in the binding constant
to both proteins. The effect of the stereochemistry and the position
of the side chain on the binding constant to streptavidin could be
explained in view of the binding modes predicted from docking
studies using AutoDock 3.05.
6-(3,7-Dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-2-yl)hexanoic acid
(2d)
6
-Ureidohexanoic acid (1.74 g, 10 mmol) was dissolved in 60 ml of
◦
water at 90 C. The pH of the solution was adjusted to 1 with conc.
HCl and 4,5-dihydroxyimidazolidin-2-one (903 mg, 10.5 mmol)
was added. The resulting mixture was heated to reflux for 1 h and
was then cooled to 4 C and the precipitate collected after 24 h and
recrystallized from water. Yield: 50%; mp: 179–183 C. H NMR
◦
◦
1
(
(
200 MHz, DMSO-d
m, 2H, –CH–CH–); 3.19–3.05 (m, 1H, N–CHH); 2.97–2.83 (m,
–COOH); 1.44 (m, 4H,
6
): 7.40 (s, 1H, NH); 7.23 (s, 2H, NH); 5.20
1
CH
1
H, N–CHH); 2.17 (t, J = 7 Hz, 2H, CH
2
−
1
2
); 1.18 (m, 2H, CH
2
). IR (KBr): mmax/cm : 3290, 1740, 1708,
+
682, 1509. ESI-MS: 279.10 ([M + Na] ). Elemental analysis (%)
calc. for C
found: 44.98, H 6.51, N 20.69, O 27.82.
9
H
14
N
4
O
4
·0.6H
2
O: C 44.97, H 6.49, N 20.98, O 27.56;
Experimental
Ethyl-7-diazo-6-oxoheptanoate (3)
General
Ethyl 5-(chloroformyl)pentanoate (1.203 g, 6.25 mmol) was dis-
solved in 30 ml of anhydrous hexane under a nitrogen atmosphere
1
H NMR spectra were recorded on a Varian Gemini (200 MHz) or
Varian Mercury spectrometer (400 MHz) in the indicated solvent.
Chemical shifts were reported in parts per million (ppm) relative
to tetramethylsilane (0.0 ppm), CDCl
and added dropwise to a solution of diazomethane in Et
2
O
(
80 ml, 15.1 mmol). The reaction mixture was stirred for 1 h
3
(7.26 ppm), or DMSO-
(2.50 ppm) as an internal standard. C NMR spectra were
at room temperature and then the excess diazomethane and the
solvent were removed under reduced pressure to yield 1.199 g
of a yellow oil. The crude diazoketone was purified by column
chromatography (gradient from hexane–AcOEt 3 : 1 to 1 : 1) to
1
3
d
6
recorded on a Varian Gemini (50 MHz) or Varian Mercury
spectrometer (100 MHz). Chemical shifts were given in parts1
per million (ppm) relative to DMSO-d
standard. Infrared spectra were recorded on a Perkin-Elmer 681
instrument. UV-Vis spectra were recorded on a Varian CARY
5
6
(39.5 ppm) as an internal
−1
yield 0.912 g of a yellow oil (75%). IR (film) mmax/cm : 3094, 2981,
1
2
940, 2871, 2104, 1731, 1641. H NMR (200 MHz, CDCl
3
): 5.26 (s
–O–CO–), 2.32 (m, 4H,
), 1.66 (m, 4H, CH –CH ).
broad, 1H, –CH=N
CH –COOEt and CH
.25 (t, 3H, CH –CH –O).
), 4.13 (q, 2H, –CH
–CO–CH=N
2
2
00 Scan spectrometer. Mass spectra were recorded on a Waters
2
2
2
2
2
Micromass ZQ spectrometer for ESI spectra. Analytical thin layer
chromatography was performed on Merck Silica Gel 60 F254
1
3
2
plates and visualized with UV light and developed by exposure to
Ethyl-5-(3,7-dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-1-
yl)pentanoate (5)
ꢀ
Cl
2
(gas) prior to soaking in a solution of 4,4 -methylenebis(N,N-
dimethylaniline). Reverse phase HPLC was performed on a C18
Kromasil silica-based column. Melting points were determined in
a variable temperature optical microscope and are uncorrected.
Optical rotation was measured on a Perkin Elmer 241 MC
polarimeter.
Ethyl 7-diazo-6-oxoheptanoate (42 mg, 0.21 mmol) was dissolved
in 5 ml of acetone, and cooled with an ice bath. Dimethyldioxirane
solution (1.5 equiv.) was added and evolution of nitrogen was
observed. The reaction mixture was allowed to warm to room
temperature and react for 30 min. The solvent was removed under
reduced pressure and the desired glyoxal 4 was obtained as an oil,
pure enough for the next reaction. The glyoxal 4 was dissolved
Materials
in a H O–EtOH mixture (2 : 1) and cooled on an ice bath. Urea
2
Avidin (affinity purified, Sigma), streptavidin (affinity purified,
Sigma), 5-ethoxycarbonylpentanoic acid (Aldrich) and glycoluril
(61.4 mg, 1.02 mmol) was added and the pH of the solution was
adjusted to 3 with 0.05 M HCl. The reaction mixture was allowed
21
◦
(
6
2
Acros) were used as received. 4,5-Dihydroxyimidazolidin-2-one,
to react for 3 d at 4 C. Then the pH of the solution was adjusted
39
40
◦
-ureidohexanoic acid, dimethyldioxirane and 5-(3,7-dioxo-
to 1 with 0.1 M HCl, and the reaction was kept at 4 C for a
19
,4,6,8-tetraazabicyclo[3.3.0]oct-2-yl)pentanoic acid (2c) were
further 5 d. The solvents were removed under reduced pressure
and the solid obtained was loaded (as a solid dispersion in silica)
into a chromatography column and eluted with an AcOEt–MeOH
gradient. The product was then further purified by RP-HPLC
prepared acording to literature procedures.
(
(
40% MeOH in H
2
O isocratic) to yield 16 mg of a white solid
(
+)- and (−)-5-(3,7-Dioxo-2,4,6,8-tetraazabicyclo-
◦
+
28%); mp: 255–258 C. ESI-MS: 293.1 ([M + Na] ). IR (KBr):
[3.3.0]oct-2-yl)pentanoic acid (2c)
−
1
1
m
max/cm : 3225, 2941, 1730, 1681, 1507. H NMR (200 MHz,
): 7.16 (s, 2H, NH); 7.06 (s, 2H, NH); 4.85 (s, 1H, NH–
CH–NH); 4.03 (q, J = 7.2 Hz, 2 H, –CH –O–CO), 2.26 (t, J =
7 Hz, 2H, –CH –COOEt); 1.62 (m, 2H, –CH –CH –COOEt); 1.48
(t, J = 7 Hz, 2H, –C–CH –CH ); 1.26 (m, 2H, C–CH –CH –CH );
1.16 (t, J = 7.2 Hz, 3H, O–CH –CH ).
Racemic resolution of 5-(3,7-dioxo-2,4,6,8-tetraazabicyclo-
3.3.0]oct-2-yl)pentanoic acid at milligram scale was achieved by
DMSO-d
6
[
2
ꢀ
R
HPLC using as chiral stationary phase a Chirobiotic T (ASTEC
2
2
2
23
22
Inc, USA) column. (+)-2c: [a]D = +50.5 (c = 0.12, MeOH);
−)-2c: [a]D = −48.5 (c = 0.09, MeOH).
2
2
2
2
2
2
2
(
2
3
3
152 | Org. Biomol. Chem., 2006, 4, 3147–3154
This journal is © The Royal Society of Chemistry 2006

View Article Online
5
-(3,7-Dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-1-yl)pentanoic acid
(b) Binding model for a competitive titration with a racemic
ligand (or a mixture of two ligands). This binding model also
assumes that the four identical subunits of each protein (Av or Sav)
(
2b)
Ethyl-5-(3,7-dioxo-2,4,6,8-tetraazabicyclo[3.3.0]oct-1-yl)penta-
noate (24 mg, 0.088 mmol) was dissolved in 8 ml of an aqueous
solution of KOH (0.012 M, 0.096 mmol). The reaction mixture
was stirred overnight at room temperature. Then, 0.3 ml of a KOH
solution (0.098 M, 0.029 mmol) were added and the reaction was
stirred for a further 6 h. The reaction mixture was then acidified
to pH 2 with 0.1 M HCl and evaporated in vacuo. The solid
residue was redissolved in the minimum quantity of water (about
behave independently, so that the equations of the 1 : 1 binding
25,26
model
and L
S) (Scheme 3). It was assumed that there was no interaction
between the two ligands L and L , and that they would bind
can also be used. But now there are two ligands (L
1
2
) that compete with the indicator (I) for the protein subunits
(
1
2
the same protein with different binding constants (K
L
S
and K
L S
2
1
respectively, see ESI†).
8
ml) and loaded onto a column packed with MCI Gel CHP20P
resin, washed with 50 ml of water and then the product eluted
with water–MeOH (2 : 1, v/v), yielding 22 mg of a white solid
◦
+
(
(
(
1
2
1
1
quantitative); mp: 243–245 C. ESI-MS: 265.0 [M + Na ]. IR
Scheme 3
−
1
1
KBr): mmax/cm : 3220, 2948, 1745, 1683, 1558, 1508. H NMR
400 MHz, DMSO-d ): 7.16 (s, 2H, NH); 7.06 (s, 2H, NH); 4.85 (s,
–COOH); 1.62 (m,
6
H, NH–CH–NH); 2.18 (t, J = 7 Hz, 2H, CH
H, CH –CH
–COOH); 1.48 (t, J = 7 Hz, 2H, –C–CH
.26 (m, 2H, –C–CH –CH –CH ). C (100 MHz, DMSO-d
2
In order to simultaneously measure the binding constant of
both enantiomeric ligands to a given protein by means of a
single titration with the racemic ligand, the general procedure
for spectrophotometric competitive titrations was also followed.
Least-squares fitting of the calculated absorbance (Acalc) [eqn.
2
2
2
–CH
2
);
):
1
3
2
2
2
6
75.8, 161.9, 77.1, 69.3, 39.4, 35.1, 25.9, 23.6. Elemental analysis
·0.5H O: C 43.03, H 6.02, N 22.30, O
8.66; found: 43.54, H 5.64, N 22.05, O 28.77.
(%) calc. for C
9
H
14
N
4
O
4
2
2
(
1)] to the experimental absorbance (Aexp) afforded in a single
experiment the binding constant of each enantiomeric ligand.
This method can also be applied to non-racemic mixtures of two
ligands, as long as their molar ratio is known. The reasonably good
fit between the experimental and calculated curves supports our
assumption of independent behaviour between the binding sites.
General procedure for the spectrophotometric competitive titrations
Spectrophotometric competitive titrations were performed on a
1
.5 mL cell of 1 cm pathlength. Aliquots of a 0.1–3.0 mM solution
of the ligand (either single or racemic) in phosphate buffer (0.1 M,
pH = 7.0) were added to a 5–15 lM solution of the protein and
Docking studies
3
0–70 lM of HABA in phosphate buffer (0.1 M, pH = 7.0). UV-
Vis absorption spectra were recorded from 700 to 200 nm, 5 min
after each addition, in order to monitor the changes in absorbance
at 500 and 348 nm (due to the displacement of the protein-bound
dye by the ligand).
34–36
AutoDock 3.05
was used in the docking studies. The X-ray
structure of the dimeric streptavidin–biotin complex was retrieved
41
38
from the Protein Data Bank (entry code: 1SWE ), water
molecules were removed and the resulting pdb file was converted
to the tetrameric form by symmetry operations using the Swiss-
(
a) Binding model for a competitive titration with an enan-
42
PdbViewer. Polar hydrogens were added and Kollman charges
were assigned to the macromolecule using the AutoDockTools
tiomerically pure ligand (or a single ligand). This binding model
assumes that the four identical subunits of each protein (Av
or Sav) behave independently, without cooperative effects. This
assumption allows us to consider the competition of the indicator
43
program. One of the four biotin molecules was removed from its
binding site and affinity grids of 80 × 80 × 80 points, separated
˚
0.375 A and centered in the empty binding site, were constructed
(
I) and the ligand (L) for the tetrameric protein as equivalent to a
using the Autogrid program. Charges and active torsions were
added to the ligands using the AutoDockTools program and non-
polar hydrogens were merged with the carbons. The Lamarckian
genetic algorithm (LGA) was used in order to perform the
conformational search. For each ligand, 50 independent docking
runs of 100 individuals and 10 000 generations were performed
using the default values for the AutoDock parameters. Results
were clustered and ranked according native AutoDock scoring
function.
competition of I and L for the monomeric protein subunits (S).
Since Av and Sav have only one binding site per subunit, the
25,26
equations of the 1 : 1 binding model
and ESI†).
can be used (see Scheme 2
Scheme 2
Acknowledgements
Least-squares fitting of the calculated absorbance (Acalc) [eqn.
1)] to the experimental absorbance (Aexp) led to the optimal value
for the binding constant between S and L. The reasonably good
fit between the experimental and calculated curves supports our
assumption of independent behaviour between the binding sites.
(
This research was supported by Generalitat de Catalunya
(2005SGR000653 and 2002BEAI400180) and by Ministerio de
Ciencia y Tecnologia (BQU2002-02424) of Spain. Both P. H.-
F. and E. A. thank Generalitat de Catalunya for predoctoral
fellowships, and I. C. thanks Universitat de Barcelona for a
predoctoral fellowship. We thank Professor Roeland J. M. Nolte,
Acalc = eI(500)[I] + eS·I(500)[S·I]
(1)
This journal is © The Royal Society of Chemistry 2006
Org. Biomol. Chem., 2006, 4, 3147–3154 | 3153

View Article Online
Dr Maria Llu ¨ı sa P e´ rez, Dr Jaume Farr a` s and Dr Oriol Pineda for
helpful discussions.
19 A. N. Kravchenko, E. Y. Maksareva, P. A. Belyakov, A. S. Sigachev,
K. Y. Chegaev, K. A. Lyssenko, O.V. Lebedev and N. N. Makhova,
Russ. Chem. Bull., 2003, 52, 192–197.
2
0 A. Wu, J. C. Fettinger and L. Isaacs, Tetrahedron, 2002, 58, 9769–9777.
References
21 E. Grillon, R. Gallo, M. Pierrot, J. Boileau and E. Wimmer, Tetrahedron
Lett., 1988, 29, 1015–1016.
22 S. Gautam, R. Ketcham and J. Nematollahi, Synth. Commun., 1979, 9,
863–870.
23 K. Steinbach and S. Berger, Chromatographia, 1997, 46, 92–94.
24 N. M. Green, Biochem. J., 1965, 94, C23.
25 H.-J. Schneider and A. K. Yatsimirsky, Principles and Methods in
Supramolecular Chemistry, John Wiley & Sons, Chichester, 2000.
26 K. A. Connors, Binding constants. The measurement of molecular
complex stability, John Wiley & Sons, New York, 1987.
27 J.-A. Farrera, P. Hidalgo-Fernandez, J. M. Hannink, J. Huskens, A. E.
Rowan, N. A. J. M. Sommerdijk and R. J. M. Nolte, Org. Biomol.
Chem., 2005, 3, 2393–2395.
28 D. H. Williams and M. S. Westwell, Chem. Soc. Rev., 1998, 27, 57–64.
29 G. Tarkanyi, J. Chromatogr., A, 2002, 961, 257–276.
30 P. Bartak, P. Bednar, L. Kubacek, M. Lammerhofer, W. Lindner and
Z. Stransky, Anal. Chim. Acta, 2004, 506, 105–113.
31 J. P. Liu, H. Coffey, D. J. Detlefsen, Y. Li and M. S. Lee, J. Chromatogr.,
A, 1997, 763, 261–269.
1
2
3
N. M. Green, Methods Enzymol., 1990, 184, 51–67.
N. M. Green, Adv. Protein Chem., 1975, 29, 85–133.
M. Ahlers, W. Mueller, A. Reichert, H. Ringsdorf and J. Venzmer,
Angew. Chem., 1990, 102, 1310–1327; M. Ahlers, W. Mueller, A.
Reichert, H. Ringsdorf and J. Venzmer, Angew. Chem., Int. Ed. Engl.,
1
990, 29(11), 1269–1285.
L. Haeussling, B. Michel, H. Ringsdorf and H. Rohrer, Angew. Chem.,
991, 103, 568–572; L. Haeussling, B. Michel, H. Ringsdorf and H.
4
5
1
Rohrer, Angew. Chem., Int. Ed. Engl., 1991, 30(5), 569–572.
H. Ebato, J. N. Herron, W. Mueller, Y. Okahata, H. Ringsdorf and P.
Suci, Angew. Chem., 1992, 104, 1064–1066; H. Ebato, J. N. Herron, W.
Mueller, Y. Okahata, H. Ringsdorf and P. Suci, Angew. Chem., Int. Ed.
Engl., 1992, 31(8), 1087–1090.
6
7
W. Mueller, H. Ringsdorf, E. Rump, G. Wildburg, X. Zhang, L.
Angermaier, W. Knoll, M. Liley and J. Spinke, Science (Washington,
D. C.), 1993, 262, 1706–1708.
K. Fujita, S. Kimura, Y. Imanishi, E. Rump, J. van Esch and H.
Ringsdorf, J. Am. Chem. Soc., 1994, 116, 5479–5480.
32 A. Dalla Cort, J. I. M. Murua, C. Pasquini, M. Pons and L. Schiaffino,
Chem.–Eur. J., 2004, 10, 3301–3307.
33 E. Buergisser, R. J. Lefkowitz and A. DeLean, Mol. Pharmacol., 1981,
19, 509–512.
34 G. M. Morris, D. S. Goodsell, R. S. Halliday, R. Huey, W. E. Hart,
R. K. Belew and A. J. Olson, J. Comput. Chem., 1998, 19, 1639–
1662.
8
9
H. Ringsdorf and J. Simon, Nature, 1994, 371, 284.
P. Ringler, W. Muller, H. Ringsdorf and A. Brisson, Chem.–Eur. J.,
1
997, 3, 620–625.
1
0 J. M. Hannink, J. J. L. M. Cornelissen, J. A. Farrera, P. Foubert, F. C. De
Schryver, N. A. J. M. Sommerdijk and R. J. M. Nolte, Angew. Chem.,
Int. Ed., 2001, 40, 4732–4734.
1
1
1
1
1
1
1
1
1 O. Livnah, E. A. Bayer, M. Wilchek and J. L. Sussman, Proc. Natl.
Acad. Sci. U. S. A., 1993, 90, 5076–5080.
2 P. C. Weber, D. H. Ohlendorf, J. J. Wendoloski and F. R. Salemme,
Science (Washington, D. C.), 1989, 243, 85–88.
3 P. C. Weber, J. J. Wendoloski, M. W. Pantoliano and F. R. Salemme,
J. Am. Chem. Soc., 1992, 114, 3197–3200.
35 G. M. Morris, D. S. Goodsell, R. Huey and A. J. Olson, J. Comput.
Aided Mol. Des., 1996, 10, 293–304.
36 D. S. Goodsell and A. J. Olson, Proteins: Struct., Funct., Genet., 1990,
8, 195–202.
37 B. D. Bursulaya, M. Totrov, R. Abagyan and C. L. Brooks, 3rd,
J. Comput. Aided Mol. Des., 2003, 17, 755–763.
38 S. Freitag, I. Le Trong, L. Klumb, P. S. Stayton and R. E. Stenkamp,
Protein Sci., 1997, 6, 1157–1166.
39 G. R. Stark, Biochemistry, 1965, 4, 1030–1036.
40 W. Adam, J. Bialas and L. Hadjiarapoglou, Chem. Ber., 1991, 124,
2377.
41 H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H.
Weissig, I. N. Shindyalov and P. E. Bourne, Nucleic Acids Res., 2000,
28, 235–242.
4 P. C. Weber, M. W. Pantoliano, D. M. Simons and F. R. Salemme,
J. Am. Chem. Soc., 1994, 116, 2717–2724.
5 L. Pugliese, A. Coda, M. Malcovati and M. Bolognesi, J. Mol. Biol.,
1
993, 231, 698–710.
6 Y. Pazy, T. Kulik, E. A. Bayer, M. Wilchek and O. Livnah, J. Biol.
Chem., 2002, 277, 30892–30900.
7 A. E. Rowan, J. A. A. W. Elemans and R. J. M. Nolte, Acc. Chem. Res.,
1
999, 32, 995–1006.
8 B. A. Katz, B. Liu and R. Cass, J. Am. Chem. Soc., 1996, 118, 7914–
920.
42 N. Guex and M. C. Peitsch, Electrophoresis, 1997, 18, 2714–2723.
43 M. F. Sanner, J. Mol. Graphics Modell., 1999, 17, 57–61.
7
3
154 | Org. Biomol. Chem., 2006, 4, 3147–3154
This journal is © The Royal Society of Chemistry 2006