Communications
structure (PDB entry 2X52) as well as the dimensions and
orientational flexibility of the spin labels (Figures S4 and S5),
reveals that the ligand bridges pairs of adjacent primary
binding sites (B1C2/C2B1 or B2C1/C1B2).
If we decrease the lectin concentration to a WGA dimer/
ligand molar ratio of 1:4, only a fraction (< 40%) of the
divalent ligand is expected to bind to WGA. Mobility
measurements (Figure S2, top) also suggest only partial
binding. Partial binding is also reflected in the distance
distribution between the two spin labels within 12 (Figure 1,
top) with the low-distance peak featuring a broader shoulder
at the lower distance end and a shift of its maximum to a
slightly smaller distance. This suggests a superposition of the
distance distributions for bound and for unbound ligands,
respectively. Corresponding simulations suggest about 40%
bound ligands. In addition, the measurements of the shorter
distances using cw spectra at low temperatures (Figure S1,
top) indicate a reduced—compared to in the absence of WGA
(Figure S1, bottom)—but still significant presence of short
distances (r< 1.5 nm) allocated to unbound ligands. While the
relative binding fraction of the divalent ligand is decreased at
the reduced protein concentration, the number of bound
ligands per protein is increased, that is, larger than 1
(Figure 2c). Consequently, additional distances appear in
the distribution (Figure 1, top) corresponding to interligand
spin interactions between different molecules of 12 bound to
the very same WGA dimer. While a quantitative analysis for
multiple spin interactions is difficult,[12] a prominent peak at
r= 4.3 nm and a broad feature for r< 4 nm can be identified.
These distances correspond very well to the interligand
distances between the primary binding sites predicted from
the crystal structure (cf. Figure S4, distances between oxygens
in the 6-position of GlcNAc residues: 4.2, 2.6, and 3.5 nm).
The interligand distances between the secondary binding sites
D2A1 and D1A2 (up to 5.3 nm) are not found.
For the divalent ligands 21 and 22 with the short linker we
find in the absence of WGA a situation similar to that for 11
and 12, respectively. Rotational diffusion of 21 is fast (tC =
120 ps, Figure S6, bottom), distances shorter than 1.5 nm are
present (Figure S7, top), and the distance distribution
obtained from DEER analysis of 22 features a single broad
peak with a maximum at 1.8 nm (Figure 3, bottom). This
maximum occurs at the same distance as for 12. This can be
explained by the gauche effect of the oxygen atoms in the
linker region of 12 leading to a folded conformation that is
well known for oligo(ethylene glycol) chains.[13] Upon addi-
tion of WGA, the results for the shorter ligand 22 completely
from those obtained with 12. Even in the presence of an
eightfold molar excess of the WGA dimer, the main
component of the cw spectrum of 21 at room temperature
remains the fast motion regime as in the absence of WGA.
Only a small fraction features a reduced mobility (Figure S6,
top). This finding suggests a low binding affinity which is in
accordance with the IC50 value of 2. In the distance
distribution obtained with 22 under these conditions
(Figure 3, top), the same short distances as those in absence
of WGA are found (including distances below 1.5 nm; cf.
Figure S7). Distances of r= 2.3 nm corresponding to the
stretched, site-bridging ligand conformation do not appear
Figure 3. Distance distributions for 22 in the absence (bottom) and
presence of WGA (molar ratio WGA dimer/ligand 22 8:1, top). Small
peaks at r=3.1 and 3.9 nm (bottom) are not significant (see the
Supporting Information).
significantly. In contrast, distances around 3 nm are present
that are attributed to interligand distances between spin-
labeled ligands bound to C1B2 and C2B1. From these findings
we conclude that there is a small fraction of monovalently
bound divalent ligands which preferably bind to C1B2 and
C2B1 (Figure 2d). The absence of the typical prominent peak
at r= 2.3 nm observed for 12 shows that under these
conditions adjacent binding sites, for example C1B2 and
B2C1, are not occupied simultaneously. Since the crystal
structure does not suggest steric hindrance of simultaneous
binding to adjacent binding sites, the most plausible expla-
nation for this observation is that the binding affinity for
C1B2 and C2B1 is higher than for the other two primary
binding sites.
In a solution of monovalent ligand 31 and WGA (each
33 mm), only weak binding is detected by EPR spectroscopy.
The main contribution to the dipolar evolution in the DEER
experiment originates from the background signal of a
homogeneous three-dimensional spin distribution. However,
there is a small contribution from interacting spins of singly
labeled ligands 31 bound to the very same WGA dimer. The
peak at 2.3 nm which is characteristic for binding to adjacent
binding sites is missing (Figure 4, bottom) as in the case of 22.
Only longer distances allocated to distances between the
binding sites C1B2 and C2B1 (around 3 nm, cf. Figure 3, top)
and between B1C2 and B2C1 (around 4 nm) occur. Not until
the WGA concentration is decreased to a protein/ligand
molar ratio of 1:7, resulting in more ligands bound per
protein, does the peak attributed to adjacent binding sites
appear (Figure 4, top). The increase of the relative intensity of
the peak allocated to the occupied binding sites B1C2 and
B2C1 (approximately 4 nm) with decreasing protein/ligand
ratio, again, indicates the lower binding affinity of these sites
compared to C1B2 and C2B1.
In summary, our results show a detailed picture of the
molecular mechanism of the binding of mono- and divalent
ligands to WGA. Applying a combination of state-of-the-art
EPR techniques, we obtained, for the first time, structural
evidence for multivalent protein–ligand interactions in solu-
tion. The chelating binding of the divalent ligand 12, which has
a linker long enough to bridge adjacent binding sites, is
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2011, 50, 8428 –8431
Braun, Patrick
