Love-Hate ligands for high resolution analysis of strain in ultra-stable protein/small molecule interaction

Article abstract of DOI:10.1016/j.bmc.2014.07.029

The pathway of ligand dissociation and how binding sites respond to force are not well understood for any macromolecule. Force effects on biological receptors have been studied through simulation or force spectroscopy, but not by high resolution structural experiments. To investigate this challenge, we took advantage of the extreme stability of the streptavidin-biotin interaction, a paradigm for understanding non-covalent binding as well as a ubiquitous research tool. We synthesized a series of biotin-conjugates having an unchanged strong-binding biotin moiety, along with pincer-like arms designed to clash with the protein surface: 'Love-Hate ligands'. The Love-Hate ligands contained various 2,6-di-ortho aryl groups, installed using Suzuki coupling as the last synthetic step, making the steric repulsion highly modular. We determined binding affinity, as well as solving 1.1-1.6 ? resolution crystal structures of streptavidin bound to Love-Hate ligands. Striking distortion of streptavidin's binding contacts was found for these complexes. Hydrogen bonds to biotin's ureido and thiophene rings were preserved for all the ligands, but biotin's valeryl tail was distorted from the classic conformation. Streptavidin's L3/4 loop, normally forming multiple energetically-important hydrogen bonds to biotin, was forced away by clashes with Love-Hate ligands, but Ser45 from L3/4 could adapt to hydrogen-bond to a different part of the ligand. This approach of preparing conflicted ligands represents a direct way to visualize strained biological interactions and test protein plasticity.

Full text of DOI:10.1016/j.bmc.2014.07.029

Bioorganic & Medicinal Chemistry 22 (2014) 5476–5486
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Love–Hate ligands for high resolution analysis of strain
in ultra-stable protein/small molecule interaction
Michael Fairhead , Di Shen b, , Louis K. M. Chan , Ed D. Lowe , Timothy J. Donohoe , Mark Howarth a,
a,
b
a
b,
⇑
⇑
a
Department of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, UK
Department of Chemistry, University of Oxford, Chemistry Research Laboratory, 12 Mansfield Road, Oxford OX1 3TA, UK
b
a r t i c l e i n f o
a b s t r a c t
Article history:
The pathway of ligand dissociation and how binding sites respond to force are not well understood for
any macromolecule. Force effects on biological receptors have been studied through simulation or force
spectroscopy, but not by high resolution structural experiments. To investigate this challenge, we took
advantage of the extreme stability of the streptavidin–biotin interaction, a paradigm for understanding
non-covalent binding as well as a ubiquitous research tool. We synthesized a series of biotin-conjugates
having an unchanged strong-binding biotin moiety, along with pincer-like arms designed to clash with
the protein surface: ‘Love–Hate ligands’. The Love–Hate ligands contained various 2,6-di-ortho aryl
groups, installed using Suzuki coupling as the last synthetic step, making the steric repulsion highly
modular. We determined binding affinity, as well as solving 1.1–1.6 Å resolution crystal structures of
streptavidin bound to Love–Hate ligands. Striking distortion of streptavidin’s binding contacts was found
for these complexes. Hydrogen bonds to biotin’s ureido and thiophene rings were preserved for all the
ligands, but biotin’s valeryl tail was distorted from the classic conformation. Streptavidin’s L3/4 loop,
normally forming multiple energetically-important hydrogen bonds to biotin, was forced away by clashes
with Love–Hate ligands, but Ser45 from L3/4 could adapt to hydrogen-bond to a different part of the
ligand. This approach of preparing conflicted ligands represents a direct way to visualize strained
biological interactions and test protein plasticity.
Received 12 May 2014
Revised 16 July 2014
Accepted 17 July 2014
Available online 24 July 2014
Keywords:
Protein crystallization
Biotin
Avidin
Mechanobiology
Chemical biology
Ó 2014 Elsevier Ltd. All rights reserved.
1
. Introduction
Many effects in biology depend on force, including sensing of
difficult to reconcile these values with specific molecular features.
Simulation has been important in bridging this gap, suggesting that
there are new molecular interactions made as ligands are extracted
1
6,7
faithful chromosome segregation, restraining cancer cell prolifer-
from the binding site by an external force. However, there has
2
3,4
ation, and adaptation to exercise. Some of the forces involved in
these events have been measured and key components for trans-
ducing and responding to force determined.3 However, there are
still major questions at the molecular level as to how force changes
structure and reactivity. Force spectroscopy, particularly Atomic
Force Microscopy (AFM), can identify how the energy barriers to
ligand dissociation from biological receptors vary according to
the pulling rate, as well as providing a distance from the ground
been limited experimental verification of such analyses. Also, sim-
ulation has often used simplifications to binding partners, as well
as time-scales of unbinding that are orders of magnitude faster
6
–8
than occur in the cell.
This time-scale difference can distort
the energy landscape, leading to the disappearance of energy bar-
5
,9,10
riers which would be significant at lower pulling rates.
X-ray crystallography is the most widely used method for high
resolution analysis of ligand binding interactions, but it is not pos-
sible to apply force directly to every ligand in the X-ray beam. Here
through the synthesis of a series of chemical probes, we aimed to
use crystallography to gain molecular insight into how force
affected ligand interaction. We made use of the streptavidin–biotin
system, a pinnacle of molecular recognition. Streptavidin is a pro-
5
state to the transition state for unbinding. Nevertheless it is very
Abbreviations: D, Dead, non-binding streptavidin mutant; EC50, half-maximal
effective concentration; L3/4, loop connecting b-strands 3 and 4; LH, Love–Hate
ligand; rmsd, root mean square deviation; SA, wild-type core streptavidin; SAe, core
streptavidin with C-terminal hexaglutamate tag.
ꢁ14
8
tein with an affinity for the vitamin biotin of 4 ꢀ 10
M. As well
⇑
Corresponding authors. Tel.: +44 (0)1865 275649; fax: +44 (0)1865 285002
as being a favored model system, streptavidin is one of the most
widely applied linkages or bridges in bioorganic chemistry and
biotechnology. There is extensive study of this system from
(
T.J.D.); tel.: +44 (0)1865 613233; fax: +44 (0)1865 613201 (M.H.).
E-mail addresses: timothy.donohoe@chem.ox.ac.uk (T.J. Donohoe), mark.howarth@
1
1
bioch.ox.ac.uk (M. Howarth).
These authors contributed equally.
http://dx.doi.org/10.1016/j.bmc.2014.07.029
0
968-0896/Ó 2014 Elsevier Ltd. All rights reserved.

M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
5477
site-directed or random mutagenesis, AFM, and simulation; also
groups pointing back towards the protein, designed to introduce
a broad van der Waals clash adjacent to the binding site)
(Fig. 1A). The biotin core must be unmodified, in order to avoid
interrupting the fundamental interactions with streptavidin
streptavidin’s symmetry and rigidity make it favorable for high-
resolution crystallography.6
,9,11–14
Streptavidin’s affinity for biotin
relates to a hydrophobic cage around biotin, extensive van der
Waals interactions, and network of cooperative hydrogen
1
3,14
a
(hydrogen bonding network shown in Figure 1B).
We avoided
1
4
bonds. Hydrogen bonds to biotin include those from the loop
between b-strands 3 and 4 (L3/4) of streptavidin. In the absence
of biotin, L3/4 is commonly disordered and/or shows high temper-
ature factors; L3/4 closes over the valeryl tail in all structures
where biotin analogs or biotin conjugates are bound.13–15
having a large group within one or two atoms from the biotin side-
chain carbonyl, based on prior insight into such biotin conjugates
1
7,18
designed to be resistant to biotinidase cleavage.
We also
required ligands to have sufficient aqueous solubility for soaking
into crystals to achieve high occupancy. We sought to avoid heter-
ogeneity in ligand position, to help the ligand be resolved in the
crystal structure and so that ligand strain would be at the same
point on the protein in all of the unit cells. Ligands were designed
to have ready synthetic accessibility, as well as to have a highly
modular synthesis, through varying the di-ortho aryl groups to
tune streptavidin repulsion (Fig. 1C).
We synthesized a tailored series of biotin conjugates, Love–Hate
(
LH) ligands, designed to leave the biotin core unmodified, so that
specific interactions with the receptor could be retained, but to
introduce a repulsive element through steric clash at positions dis-
tant to the binding site (Fig. 1). Using the series of LH ligands, as
well as exploiting our orientation-controlled divalent streptavidin
1
6
tetramers, we gained insight into strain within ligand complexes
through X-ray crystallography and equilibrium binding analysis.
Initially, we designed the ligand LH1, which contains an ethyl-
ene diamine spacer between biotin and a sterically hindered
2
,6-diarylbenzene moiety (Fig. 1C). LH1 was assembled from com-
mercially available benzoic acid (1) (Scheme 1A) that underwent
C–H diiodination using Pd(OAc) and I with PhI(OAc) as oxidant
2
2
. Results
2
2
2
1
9
to give (2) in 74% yield. 2,6-Diiodobenzoic acid (2) was then
coupled with N-Boc ethylenediamine (3) under amide coupling
conditions using EDCꢂHCl and HOBt, to give amide (4) in 56% yield.
.1. Design and synthetic route to the first Love–Hate ligand
We required ligands with the appropriate balance of attraction
2
Next, Suzuki cross coupling with PhB(OH) was then successfully
(
through an unmodified biotin core) versus repulsion (through
Figure 1. Principle and chemical structures of Love–Hate ligands. (A) Cartoon of strategy for ligand design. Protein in blue. B represents the high affinity ligand and dotted
lines the stabilizing polar interactions. (B) Hydrogen-bonding contacts made by streptavidin to biotin, shown as dotted lines, with the Loop between strands 3 and 4 (L3/4) in
blue. (C) Love–Hate (LH) ligand structures 1–4.

5
478
M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
performed on the hindered diiodide (4).20 The crude material was
directly treated with trifluoroacetic acid for the deprotection of the
Boc group, which subsequently allowed final amide coupling onto
biotin NHS ester, to afford LH1 in 16% yield over 3 steps. LH1
identity and purity were validated by mass spectrometry
and was reduced in situ to the 2,6-dibromohydrazine (6) using
2
1
SnCl
2
.
With hydrazine (6) in hand, we found that amide coupling
with biotin under EDCꢂHCl and HOBt in DMF provided the optimal
conditions, since hydrazide (7) directly precipitated out of solution
upon completion of the reaction. Finally, Suzuki coupling, without
extensive optimization, afforded the bulky LH2 ligand in 18% yield.
LH3 and LH4 were prepared by the same route using the corre-
sponding aryl boronic acids. The structure and purity of LH2, LH3
and LH4 were validated by mass spectrometry (Section 4.1) and
NMR (Supplementary methods). Given the limited aqueous solu-
bility of LH1, LH2 and LH3 (<2 mM in 10% DMSO), when further
increasing the size of the adjunct rings in designing LH4, to favor
solubility we used a substituent with multiple heteroatoms, sulfo-
nylmorpholinyl. LH4 retained similar solubility in 10% DMSO to the
other LH ligands.
(
Section 4.1) and NMR (Supplementary methods).
2
.2. Crystal structure of streptavidin bound to LH1
Streptavidin forms a dimer of dimers as its quaternary struc-
1
3,14
ture, with each subunit a b-barrel of 8 antiparallel b-strands.
LH1 gave a 1.1 Å structure in complex with streptavidin, with both
binding sites in the asymmetric unit containing well defined ligand
(Table 1, Fig. 2A and B and Supplementary Fig. S1). Despite the
di-ortho phenyl groups, L3/4 of streptavidin was shut and all
hydrogen bond contacts to biotin were preserved, with minimal
difference to complexes with biotin itself,14 and an extra potential
hydrogen bond from LH1 to Ser112 (Fig. 2C and D).
2.4. Crystal structures of streptavidin bound to second series of
Love–Hate ligands
Since the ethylene diamine spacer of LH1 allowed flexibility in
positioning of the repulsive moieties, we designed ligands LH2–4,
which have a linker three atoms shorter between biotin and the
sterically hindered group (Fig. 1C).
Crystallization trials were set up with the second series of LH
ligands. Despite several attempts we could not obtain a well-
resolved crystal structure of LH2. With wild-type streptavidin we
observed low ligand occupancy for LH2, complicated by twinning.
Therefore we used our trans-divalent streptavidin, where only the
2
.3. Synthetic route to second series of Love–Hate ligands
1
and 3 subunits on opposite sides of the tetramer could bind bio-
1
6
The synthesis of the second series of Love–Hate ligands, LH2–4,
tin ligands (Supplementary Fig. S2),
so avoiding any clash
utilized a similar cross-coupling strategy, but with diversification
of the sterically hindered groups using cross coupling at the latest
stage possible (Scheme 1B). Commercially-available 2,6-dibromo-
between adjacent ligands. The other two subunits were ‘Dead’,
through three mutations to key biotin binding residues.22 Such
mixed tetramers can be readily generated by refolding from inclu-
sion bodies a combination of Alive and Dead subunits. As a result of
2
aniline (5) was treated with NaNO to generate the diazonium salt
Scheme 1. Chemical synthesis route of Love–Hate ligand 1 (A) and Love–Hate ligands 2–4 (B).

M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
5479
Table 1
Data collection, refinement statistics, and structure validation for the Love–Hate ligand/Streptavidin structures
Data collection
SA-LH1 (PDB 4CPE)
SA-LH3 (PDB 4CPF)
trans-Divalent SA-LH4 (PDB 4CPH)
A86D SA-LH4 (PDB 4CPI)
Space group
I222
I222
C121
C121
Asymmetric unit
Unit cell size (Å)
Unit cell angles (°)
Completeness (%)
Multiplicity
Dimer
Dimer
Tetramer
Tetramer
46.59 ꢀ 93.93 ꢀ 104.32
90 ꢀ 90 ꢀ 90
93.1 (55.0)
5.8 (3.2)
47.08 ꢀ 94.12 ꢀ 104.79
90 ꢀ 90 ꢀ 90
93.0 (65.1)
6.1 (2.4)
90.74 ꢀ 47.49 ꢀ 104.88
90 ꢀ 101.14 ꢀ 90
96.6 (95.5)
81.26 ꢀ 81.25 ꢀ 90.69
90 ꢀ 103.84 ꢀ 90
99.6 (99.6)
*
3.3 (3.2)
3.3 (3.3)
Refinement
Resolution range (Å)
Total reflections
Unique reflections
Mean I/sigma(I)
R-merge
1.06–20.37
557,763 (17,822)
96,250 (5,646)
18.14 (2.46)
0.0472 (0.4556)
0.132/0.154
2,165
1.14–32.75
477,826 (13,372)
79,019 (5,486)
18.81 (2.83)
0.0524 (0.339)
0.138/0.165
2,224
1.64–41.90
172,843 (16,293)
52,123 (5,108)
8.35 (1.93)
0.0788 (0.593)
0.200/0.232
4,069
1.54–39.45
280,431(28,198)
84,162 (8,419)
15.61 (2.34)
0.0512 (0.622)
0.157/0.182
4,355
Rwork/Rfree
No. of atoms
Protein
1,861
1,955
3,709
3,846
Heteroatoms
Water
B factors (mean of all atoms, Å )
Wilson Plot
78
226
43
226
106
254
198
311
2
9.73
15.00
11.68
18.60
18.96
28.60
17.55
25.10
Mean B value
Rmsd
Bond length (Å)
Bond angle (Å)
LH ligand occupancy
LH ligand density correlation
0.011
1.42
0.83/0.81
0.91/0.95
0.010
1.39
0.82
0.96
0.009
1.23
0.83/0.86
0.83/0.82
0.008
1.17
0.92/0.91
0.95/0.95
Structure factors
MolProbity clashscore
Poor rotamers (%)
Ramachandran outliers (%)
Ramachandran favored (%)
MolProbity score
2.17 (97th percentile)
1.08
0
2.33 (96th percentile)
0.51
0
3.28 (98th percentile)
1.35
0
0.9 (99th percentile)
1.28
0
97.55
1.11 (95 percentile)
96.48
1.24 (90 percentile)
97.73
1.28 (98 percentile)
98.19
th
th
th
th
0.86 (100 percentile)
Residues with bad bonds (%)
Residues with bad angles (%)
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
Statistics for the highest resolution shell are shown in parenthesis.
*
Completeness for the highest resolution shells of SA-LH1 and SA-LH3 were lower than expected due to some spots diffracting beyond the boundaries of the detector. The
highest resolution shells with high completeness for SA-LH1 are 1.16–1.18 Å (98%) and 1.26–1.28 Å (99%) for SA-LH3.
the Glu
6
tag on the Alive subunits, the mixed tetramers can subse-
bonds to three residues normally hydrogen-bonding to biotin
(Ser45, Asn49 and Ser88) (Fig. 3C and D).
quently be separated using anion exchange chromatography into
forms with a defined number and orientation of functional biotin
binding sites. Unfortunately, despite dealing with the high levels
of twinning, we were still unable to resolve clear density for the
LH2 ligands.
However, we were able to obtain a 1.1 Å resolution crystal
structure for LH3 bound to wild-type streptavidin. In one subunit,
LH3 was clearly resolved in all its regions (Fig. 3A and B and
Supplementary Fig. S3). The ligand had strikingly distorted the
binding site, displacing L3/4 of streptavidin and preventing
Asn49 and Ser88 forming polar contacts with biotin (Fig. 3C and
D). The conformation of the rings had led the carbonyl of the
methoxyester to bend back within hydrogen bonding distance of
the cis-NH of biotin. Normally the cis-NH of biotin hydrogen-bonds
to Ser45 (Fig. 1B), but the displaced Ser45 may be able to form an
alternative hydrogen bond to the carbonyl of this same methoxyes-
ter (Fig. 3C and D). The guanidino of Arg84 was well positioned for
Examining the water molecules resolved within 4 Å of the
bound ligand, there was no clear change for SA/LH1 as compared
to SA/biotin. However in the case of LH3, the bending over of
L3,4 led to two water molecules being resolved at the streptavi-
din/ligand interface, forming a variety of polar contacts (2023,
2053, CHAIN B). trans-Divalent SA/LH4 had only one extra water
molecule resolved at the streptavidin/ligand interface (2010,
CHAIN C), although this structure was solved at lower resolution.
1
6
2.5. Binding strength of Love–Hate ligand series
We characterized the dissociation constant of the ligand series
for streptavidin. The exceptional affinity and stability of the bio-
tin/streptavidin interaction mean that many standard assays are
unsuitable, including surface plasmon resonance and isothermal
titration calorimetry. Also the amount of ligand available from
the multi-step syntheses prohibited certain approaches. We
a putative cation–
p
interaction23 with the central phenyl ring of
LH3 (Supplementary Fig. S3E).
d
obtained relative K values using an equilibrium competition
LH4, with the largest headgroup on each phenyl ring, sulfonyl-
morpholinyl, did not yield crystals with wild-type streptavidin.
However, LH4 was able to form crystals with trans-divalent strep-
tavidin at 1.6 Å resolution. The density of the biotin core of LH4
was well resolved (Fig. 3A and B). Part of the central phenyl ring
and one of the pendant phenyl rings of LH4 were seen clearly
but the rest of LH4 had substantial disorder (Fig. 3B and
Supplementary Fig. S4). LH4 also yielded major changes to the
binding site: L3/4 was displaced and LH4 failed to form hydrogen
assay, based on the amount of biotin-4-fluorescein dissociation at
increasing ligand concentration. The fluorescence of biotin-4-fluo-
rescein is quenched ꢃ90% upon binding to streptavidin, while fluo-
2
4
rescence is regained upon release. To avoid any interaction
between bulky ligands binding to adjacent streptavidin binding
sites, ligand binding was tested on a monovalent streptavidin
(SAe1D3) with three subunits which do not bind biotin and one
1
6,22
subunit that binds biotin with wild-type affinity.
Biotin itself
of
was used in the assay as reference ligand, with
a
K
d

5
480
M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
Figure 2. Crystal structure of streptavidin bound to LH1. (A) Crystal structure of adjacent binding sites in a streptavidin tetramer bound to LH1. Streptavidin is shown in green
with a surface representation. LH1 is shown as its van der Waals surface, with carbons in cyan. (B) 2mFo-DFc electron density of LH1 (from chain B) contoured at 1 (blue)
and cropped to within 1.5 Å of ligand atoms, overlaid with the ligand structure in stick format (green). (C) Overlay of biotin and the streptavidin L3/4 loop (green, from PDB
RY2) with the equivalent part from the LH1/streptavidin structure (pink). Ligands are shown in stick format and protein in ribbon format. (D) Schematic of hydrogen bond
r
3
network from streptavidin to LH1, with parental interactions to biotin in black, interactions lost for LH1 compared to biotin in red, and putative new interactions for LH1 in
green.
ꢁ
14
8
wild-type streptavidin.26 LH4 binding was greatly impaired com-
pared to biocytin for A86D streptavidin, with a 150-fold decrease
in binding (Table 2).
4
ꢀ 10
M, as defined by Green, 1990. Desthiobiotin was used as
a weaker ligand, also of known affinity, ꢃ100-fold weaker than
biotin as reported by Hirsch and coworkers.25 We confirmed that
the assay had reached equilibrium by incubating for a further
2
days with no change in results. Despite the bulky substituents,
2.6. A86D streptavidin crystal structure with LH4
LH1 showed minimal difference in equilibrium binding to biotin,
with a 1.1-fold reduction in affinity (Table 2 and Supplementary
Fig. S5A). LH2–4 showed reduced affinity compared to biotin
To understand an even more strained interaction (as judging
from the measured EC50 value, Table 2) we solved the structure
of the LH4 ligand bound to A86D streptavidin at 1.5 Å resolution
(Table 1). The structure of the complex showed clear density for
two of the four copies of LH4 present in the asymmetric unit
(Fig. 4A and B and Supplementary Fig. S6). The pendant arms had
a very different orientation for LH4/A86D streptavidin compared
to LH4/trans-divalent streptavidin. The valeryl tail for LH4/A86D
projected well away from the classic orientation for biotin (3.6 Å
between the carbonyl carbons) (Fig. 4C). L3/4 was again displaced
into an open conformation, with loss of Asn49 and Ser88 hydrogen
bonds. However, novel putative hydrogen bonds were formed:
from Asn85 to one of the sulfonyl groups, and from Asp86 to one
of the hydrazide NH groups. As with the streptavidin/LH3 struc-
ture, Ser45 did not form the classic hydrogen bond to the cis-NH
of the ureido ring, but in a new orientation Ser45 could form a
putative hydrogen bond to the carbonyl of the valeryl tail. One of
the sulfonyl oxygens was close enough to form an intra-ligand
hydrogen bond to the cis-NH of biotin (Fig. 4C and D). Similar to
the trans-divalent SA/LH4 structure, A86D SA/LH4 had one water
molecule resolved at the streptavidin/ligand interface (2022,
CHAIN D) that was not observed in SA/biotin structures.1
(Table 2) but the effect was smaller than would be expected based
on the major changes in the binding interface from the crystal
structures. Also, the relative affinity was counter-intuitive, with
LH4 showing stronger streptavidin binding than its smaller relative
LH3 (Table 2).
Desiring to span a wider range of affinities, we explored the rea-
sons for the reduced effect on affinity of such a sterically hindered
biotin ligand. The degrees of freedom in LH2–4 still allow avoid-
ance of much of the clash with streptavidin. Therefore we created
an A86D streptavidin mutant to restrict this freedom. Steric clash
of Asp86 was designed to direct a different orientation of the cen-
tral phenyl ring of the Love–Hate ligand, so favoring the desired
repulsive interactions from van der Waals clash of the diphenyl
branches with the surface of streptavidin.
There is potential electrostatic repulsion between neighboring
carboxylates of biotin and Asp86 within A86D streptavidin. There-
fore biocytin (biotinyl-L-lysine) was used as a reference ligand for
A86D streptavidin, giving a similar binding curve as biotin showed
with wild-type streptavidin (Supplementary Fig. S5B). Biocytin and
biotin were previously found to have equivalent affinity for
3,27

M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
5481
Figure 3. Crystal structures of streptavidin bound to LH3 or LH4. (A) Crystal structure of adjacent binding sites in a streptavidin tetramer bound to LH3 (left) or LH4 (right).
Streptavidin is shown in green with a surface representation. LH3 or LH4 is shown as its van der Waals surface, with carbons in cyan. (B) 2mFo-DFc electron density of LH3
(
left, from chain B) or LH4 (right, from chain C) contoured at 1
r (blue) and cropped to within 1.5 Å of ligand atoms, overlaid with the ligand structure in stick format (green).
(
C) Overlay of biotin and the streptavidin L3/4 loop (green, from PDB 3RY2) with the equivalent part from the LH3 (left) or LH4 (right) streptavidin complex (pink). Ligands are
shown in stick format and protein in ribbon format. (D) Schematic of hydrogen bond network from streptavidin to LH3 (left) or LH4 (right), with parental interactions to
biotin in black, interactions lost for LH compared to biotin in red, and putative new interactions for LH in green.

5
482
M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
Table 2
Competition assay for Love–Hate ligand binding affinity
Ligand
Biotin
Desthiobiotin
Lipoic acid LH1
LH2
LH3
LH4
Biocytin A86D LH4 A86D
EC50, nM (95% CI)
28 (20–39) 5,410 (4,580–6,390) n/a
31 (17–56) 251 (157–401) 178 (124–257) 138 (87.7–216) 19 (17–22)
2,840 (2,220–3,640)
0.97
150
2
R
0.98
0.99
190
n/a
n/a
0.94
1.1
0.96
9
0.98
6.4
0.96
4.9
0.99
1
Fold-decrease in affinity 1a
b
Biotin-4-fluorescein was incubated with streptavidin and the indicated concentration of each ligand for 48 h at 37 °C and fluorescence was determined. The concentration of
2
ligand for 50% dissociation of biotin-4-fluorescein (EC50) was determined, along with the 95% confidence intervals (CI) for the EC50 and the correlation coefficient (R ) for the
fit. Lipoic acid served as a control, with no dissociation of biotin-4-fluorescein induced (n/a, not applicable). For wild-type streptavidin binding sites, the EC50 for the ligand
was divided by the EC50 for biotin, to give a relative decrease in binding affinity. For A86D streptavidin binding sites, the EC50 for the ligand was divided by the EC50 for
biocytin. The fluorescence curves are shown in Fig. S5. The one-site competitive binding fit was calculated using GraphPad Prism, giving an EC50 with 95% confidence intervals
2
(
CI) as well as a R Pearson correlation coefficient value.
a
Defined as 1 for all experiments with wild-type streptavidin.
Defined as 1 for the experiments with streptavidin A86D.
b
2
.7. Comparison of Love–Hate/streptavidin complex structures
LH3–4, including Ser45 interacting with different parts of the
ligands. Given the surprising finding that LH3 and LH4 bind well
to streptavidin even without many classic interactions of the biotin
binding site, these ligands may serve as a foundation to create
streptavidin mutants that bind well to such biotin conjugates, for
clinical imaging and pre-targeted cancer radiotherapy and to
overcome blocking by endogenous biotin in the bloodstream.35 It
is always possible that a crystal structure captures one out of
various structural isoforms present in solution; interpretable
occupancy of only half of the accessible binding sites in some of
our structures suggests that other binding modes of LH ligands
may be possible and will require future investigation.
Overlaying the Love–Hate ligand complex structures with the
streptavidin/biotin structure, the organization of the ureido and
thiophene rings of biotin in each structure was unchanged within
error to biotin itself (Fig. 4E). Focusing on the L3/4 loop shows
how the loop was shut with LH1, just as for biotin (Fig. 2C) but
pushed into different open conformations in the presence of LH3
(
(
Fig. 3C), LH4 (Fig. 3C) or LH4 bound to the A86D mutant
Fig. 4C). The valeryl tail was unchanged compared to biotin for
LH1 but gradually more distorted from biotin in LH3, LH4 and
LH4/A86D (Fig. 4E).
Other structures have been solved for avidin or streptavidin in
3
6–40
complex with reduced affinity ligands
but not with the
3
. Conclusion
sequential extension of the ligand series employed here for gradual
introduction of repulsion. The crystal structure of the D128A strep-
tavidin mutant bound to biotin showed that a water had filled the
extra space occupied by the Asp128 head-group and the biotin was
Here we have generated a modular series of ligands to intro-
duce repulsion into an ultra-stable protein/small molecule interac-
tion and analyzed these ligand/protein complexes at high
resolution by X-ray crystallography. The four crystal structures
show the plasticity of this outstanding ligand interaction. The
hydrogen bonds surrounding the ureido and thiophene rings of
biotin survived for all Love–Hate ligands but biotin’s valeryl tail
could change orientation and the L3/4 loop, previously considered
essential for high affinity binding,18,28,29 could be displaced by the
Love–Hate ligand arms. A distortion of L3/4 has recently been sug-
gested for avidin bound to a ferrocene-biotin conjugate.30 In one
subunit, L3/4 is closed. In the other observed subunit, L3/4 is disor-
4
1
somewhat displaced from the wild-type structure. With our
ligands we found no replacement of contacts to the ureido groups
by water molecules. This result is consistent with evidence from
simulation that the cooperativity of the five hydrogen bonds to
streptavidin from biotin’s ureido group is central to the exceptional
thermodynamic and mechanical stability of this interaction.42 Rel-
evant crystallographic insights to strained complexes also come
from colicins trapped in a complex with non-cognate anti-toxins
4
3
via a disulfide. In addition, quantum mechanical calculations
have shown that proteins do not always bind the lowest energy
3
0
dered but the electron density for the ferrocene-biotin is weak, so
it is unclear whether L3/4 has been displaced by the ligand.
There has been substantial interest in making biotin conjugates
resistant to biotinidase, an enzyme in blood which cleaves the
amide bond to release biotin. Biotinidase interferes with clinical
applications using streptavidin or avidin, such as pre-targeted
4
4
conformer of a small molecule ligand.
It is a synthetic challenge to create ligands with the right bal-
ance of repulsion, rigidity and solubility. Covalent tethering of bio-
tin onto streptavidin in such a way as to strain biotin binding
might be possible. However, the large free-energy of biotin binding
might instead trap a locally unfolded conformation of streptavidin
at the tethering site, or promote inter-tetramer interactions and so
aggregation. In future studies, it will be worth investigating even
further repulsion in the streptavidin-biotin system and exploring
the Love–Hate ligand approach in other biological systems where
3
1
radiotherapy in cancer. For designed biotinidase-resistant conju-
gates, small changes adjacent to the biotin in a variety of cyanoco-
balamin conjugates (e.g., addition of a methyl group to the
biotinamide nitrogen) caused large decreases in streptavidin bind-
1
7,18
ing affinity.
In contrast, the large arms on LH3 and LH4 caused
3
,5
force changes function.
The modular nature of the synthetic
gross disruption of the biotin binding site, with displacement of
L3/4 and loss of multiple hydrogen bonding contacts to the biotin
core, but with only small change in affinity. Point mutation of res-
idues on L3/4 contacting biotin previously caused great loss of bio-
tin binding affinity: S45A streptavidin has ꢃ900-fold reduced
route that was developed here should aid in the preparation of a
wide variety of synthetic analogs to address these fundamental
questions.
2
8
affinity for biotin. Also, based on structures of the low off-rate
4. Experimental
3
2
mutant traptavidin, we previously identified Ser45 as a key
determinant of biotin dissociation.29 The effect of Asn49 or Ser88
4.1. Ligand synthesis
mutation on biotin K
d
has not been fully evaluated but circularly-
permuted streptavidin lacking L3/4 had biotin affinity reduced
4.1.1. General synthetic methods
more than a million-fold.3
3,34
This discrepancy could be rational-
ized if LH2-LH4 formed extra hydrophobic contacts or hydrogen
bonds. Potential new interactions (i.e., not present to biotin itself),
All solvents and reagents were used as commercially supplied,
without extra precautions unless otherwise stated. All chemical
reagents were from Sigma–Aldrich unless stated. Anhydrous
chloroform and acetonitrile were dried by filtration through an
including hydrogen bond or cation–p interactions, were found for

M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
5483
Figure 4. Crystal structure of A86D streptavidin bound to LH4. (A) Structure of adjacent binding sites in A86D streptavidin bound to LH4. A86D streptavidin is shown in green
with a surface representation. LH4 is shown as its van der Waals surface, with carbons in cyan. (B) 2mFo-DFc electron density of LH4/A86D streptavidin (from chain D)
contoured at 1
r (blue) and cropped to within 1.5 Å of ligand atoms, overlaid with the ligand structure in stick format (green). (C) Overlay of biotin, the streptavidin L3/4 loop
and residue 86 (green, from PDB 3RY2) with the equivalent part from the LH4/A86D streptavidin structure (pink). Ligands are shown in stick format. (D) Schematic of
hydrogen bond network from A86D streptavidin to LH4, with parental interactions to biotin in black, interactions lost for LH4 compared to biotin in red, and putative new
interactions for LH4 in green. (E) Comparison of all LH conformations. Overlay of conformation bound to wild-type streptavidin for biotin (in green, from PDB 3RY2) compared
to LH1, LH3 and LH4. Biotin bound to wild-type streptavidin is also overlaid with LH4 bound to A86D streptavidin. Ligands are shown in stick format.
activated alumina purification column. Petrol refers to petroleum
ether in the boiling range 30–40 °C. Flash column chromatography
Tensor 27 and analyzed by OPUS software. Melting points (MP)
were obtained using a Leica VMTG heated-stage microscope and
are uncorrected.
(
FCC) was performed using Merck Kieselgel 60 silica (40–63
H nuclear magnetic resonance (NMR) spectra were recorded on
l
m).
1
1
3
a Bruker AV400 (400 MHz) or Bruker AVII500 (500 MHz).
C
4.1.2. 2,6-Diiodobenzoic acid (2)
NMR spectra were recorded on a Bruker AV400 (101 MHz) or
AVII500 (126 MHz) as stated. Chemical shifts (d, in ppm) are
reported relative to residual solvent peaks. Coupling constants
Benzoic acid (1) (500 mg, 4.09 mmol), Pd(OAc)
0.20 mmol, Alfa Aesar), iodobenzene diacetate (PIDA) (2.00 g,
6.14 mmol) and I (1.56 g, 6.14 mmol) were dissolved in DMF
2
(45.9 mg,
2
1
are quoted to the nearest 0.1 Hz for H NMR and to the nearest
(20 mL) under argon in a sealed tube. The reaction mixture was
stirred at 100 °C for 48 h, then cooled to rt, diluted with ethyl ace-
tate (50 mL) and washed with aqueous HCl (1 M, 50 mL ꢀ 4). The
1
3
1
Hz for C NMR. Mass spectra under the conditions of electro-
spray ionization (ESI) were recorded on a Fisons Platform II ana-
lyzed using a Time of Flight system. Fourier transform infrared
spectra (FTIR) were recorded as evaporated thin films on a Bruker
2 4
organic phase was washed with brine, dried over Na SO and
concentrated in vacuo. The residue was recrystallized from ethyl

5
484
M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
acetate and petrol (EtOAc/petrol, 10:1) to give the product as a
4.1.6. 2,6-Dibromophenylhydrazine (6)
light yellow solid (1.13 g, 74%). Melting point = 179–180 °C (Lit:
Glassware was washed with diethyl ether before use. A solution
of NaNO (276 mg, 4.00 mmol) in H O (1.7 mL) was added drop-
2 2
wise to a white suspension of 2,6-dibromoaniline (5) (500 mg,
2.00 mmol) in aqueous HCl (37%, 5 mL) while cooling with an
ice-water bath. The reaction mixture was stirred at 0 °C for 1 h to
4
5
1
1
2
1
84–187 °C ); H NMR (500 MHz, CDCl
3
) d 7.84 (d, J = 7.9 Hz,
H), 6.82 (t, J = 7.9 Hz, 1H); C NMR (126 MHz, CDCl ) d 172.5,
44.4, 138.7, 132.1, 90.7. Spectroscopic data are consistent with
13
3
19
those previously reported.
give a yellow suspension. Then a solution of SnCl
2
(760 mg,
4
.1.3. tert-Butyl (2-aminoethyl)carbamate (3)
A solution of di-tert-butyl dicarbonate (2.10 mL, 9.20 mmol) in
4.00 mmol) in aqueous HCl (37%, 1.2 mL) was added dropwise at
0 °C. The reaction was stirred for 30 min before being warmed to
rt and further stirred for 2 h to give a thick yellow suspension.
The reaction was cooled to 0 °C, and solid sodium hydroxide was
added until the pH was 9. The resulting yellow suspension was
chloroform (45 mL) was added via a syringe pump over 3 h to a
solution of ethylenediamine (3.10 mL, 45.8 mmol) in chloroform
(
50 mL) at 0 °C. The reaction was stirred for 24 h at rt to give a
white suspension which was then washed with saturated Na CO
100 mL ꢀ 3). The organic layer was dried over Na SO and concen-
trated in vacuo to yield the title compound, 3, as a colorless oil
2
3
2 4
extracted with diethyl ether, dried over Na SO , concentrated in
(
2
4
vacuo, and purified via column chromatography (petrol/ether,
9:1) to afford the product, 6, as a yellow solid (354 mg, 66%). 6
was highly reactive towards acetone, ethyl acetate and acetoni-
trile; decomposition was also observed in methanol. Thus, careful
1
(1.32 g, 90%); H NMR (400 MHz, CDCl
(m, 2H), 2.72 (m, 2H), 1.37 (s, 9H), 1.26 (br. s, 1H); C NMR
(101 MHz, CDCl ) d 156.2, 79.1, 43.4, 41.8, 28.4. Spectroscopic data
3
) d 5.20 (br. s, 1H), 3.10
13
3
handling was crucial to obtain pure material. Melting point = 79–
46
max (thin film)/cm 3228, 2416, 1427, 755, 716, 658; 1
ꢁ1
are consistent with those previously reported.
82 °C;
m
H
NMR (400 MHz, MeOD) d 7.60 (d, J = 8.1 Hz, 2H), 6.92 (t, J = 8.0 Hz,
1
3
4
.1.4. tert-Butyl (2-(2,6-diiodobenzamido)ethyl)carbamate (4)
tert-Butyl (2-aminoethyl)carbamate (3, 400 mg, 2.50 mmol),
1H), 4.91 (br s, 3H); C NMR (101 MHz, MeOD) d 145.1, 132.7,
+
6 6 2 2
126.1, 117.6; HRMS (ESI ) calculated for [C H Br N
+H]+ 264.8970,
2
,6-diiodobenzoic acid (2, 778 mg, 2.08 mmol), N-(3-dimethylami-
found 264.8981 (
D
ꢁ3.2 ppm).
0
nopropyl)-N -ethylcarbodiimide hydrochloride (EDCꢂHCl, 399 mg,
2
2
.08 mmol, Fluorochem) and 1-hydroxybenzotriazole (HOBt,
81 mg, 2.08 mmol, Alfa Aesar) were dissolved in MeCN (25 mL)
0
4
.1.7. N -(2,6-Dibromophenyl)-biotinyl hydrazide (7)
2
,6-Dibromophenylhydrazine (6, 302 mg, 1.14 mmol), D-biotin
in a sealed tube and the reaction was stirred at 60 °C for 12 h.
The crude mixture was filtered through a short pad of silica gel,
eluted with ethyl acetate, then recrystallized with dichlorometh-
ane to yield the product, 4, as a white solid (600 mg, 56%). Melting
(
185 mg, 0.757 mmol, Fluorochem), N-(3-dimethylaminopropyl)-
0
N -ethylcarbodiimide hydrochloride (218 mg, 1.14 mmol, Fluoro-
chem), and HOBt (154 mg, 1.14 mmol, Alfa Aesar) were dissolved
in DMF (15 mL). The yellow solution was stirred at rt for 18 h to
give a white suspension which was then filtered to give the prod-
uct, 7, as a white solid (287 mg, 74%). Melting point = 168–170 °C;
ꢁ1
point = 74–75 °C;
160, 2029, 766, 689; H NMR (500 MHz, CDCl
J = 7.9 Hz, 2H), 6.70 (br. s, 1H), 6.68 (t, J = 7.9 Hz, 1H), 5.20 (br. s,
mmax (thin film)/cm 3351, 3256, 2976, 2931,
1
2
3
) d 7.73 (d,
H), 3.55 (m, 2H), 3.42 (m, 2H), 1.39 (s, 9H); 1 C NMR (126 MHz,
CDCl ) d 170.4, 156.6, 147.0, 138.8, 131.6, 92.2, 79.6, 40.7, 39.8,
2
0
ꢁ1
3
[a
]
D
+21.6 (c 0.25, MeOH);
m
max (thin film)/cm 3269, 2929,
1
1
2
360, 2341, 1696, 1440; H NMR (500 MHz, MeOD) d 7.53 (d,
3
+
J = 8.2 Hz, 2H), 6.81 (t, J = 8.0 Hz, 1H), 4.50 (m, 1H), 4.29 (dd,
J = 7.9, 4.4 Hz, 1H), 3.20 (ddd, J = 8.7, 6.0, 4.6 Hz, 1H), 2.94 (dd,
J = 12.9, 5.0 Hz, 1H), 2.72 (d, J = 12.6 Hz, 1H), 2.25 (t, J = 7.2 Hz,
2
8.4; High Resolution Mass Spectrometry (HRMS) (ESI ) calculated
+
18 2 2 3
for [C14H I N O +Na] 538.9299, found 538.9278 (D 3.9 ppm).
1
3
2
H), 1.81–1.53 (m, 4H), 1.50–1.38 (m, 2H); C NMR (126 MHz,
0
0
00
0
4
.1.5. Biotinamidoethyl-[1,1 :3 ,1 -terphenyl]-2 -carboxamide
LH1)
tert-Butyl (2-(2,6-diiodobenzamido)ethyl)carbamate (4, 50.0 mg,
.0969 mmol), phenyl boronic acid (35.4 mg, 0.290 mmol),
Pd(PPh (11.2 mg, 0.00969 mmol, Strem Chemicals), aqueous
Na CO (2 M, 1 mL, Acros) and 1,2-dimethoxyethane (4 mL) were
MeOD) d 175.0, 166.1, 144.4, 134.2, 125.6, 115.6, 63.3, 61.6, 57.0,
(
+
4
1.1, 34.2, 29.8, 29.4, 26.4; HRMS (ESI ) calculated for [C16H20Br
2
+
4 2
N O S+Na] 512.9566, found 512.9553 (D 1.2 ppm).
0
3 4
)
4
.1.8. General procedure for the Suzuki coupling
N -(2,6-Dibromophenyl)-biotinyl hydrazide (7, 50.0 mg, 0.102 mmol,
2
3
0
placed in a sealed tube, and the mixture was degassed with argon.
The reaction mixture was then heated at 100 °C for 2.5 h, allowed
to cool to rt, filtered through a short pad of silica gel, eluted with
diethyl ether, and concentrated in vacuo. The crude material was
redissolved in dichloromethane (0.3 mL) and TFA (74
was added. The yellow solution was stirred at rt for 5 h, concen-
trated in vacuo and then redissolved in DMF (0.3 mL) before -bio-
tin N-hydroxysuccinimide ester (33.1 mg, 0.0969 mmol), and
triethylamine (176 L, 1.26 mmol) were added. The mixture was
further stirred at rt for 48 h and purified via column chromatogra-
1
equiv), the appropriate boronic acid (0.306 mmol, 3 equiv),
Pd(OAc)
clohexylphosphino-2 ,4 ,6 -triisopropylbiphenyl (XPhos) (3.6 mg,
.00765 mmol, 0.075 equiv), aqueous Na CO (2 M, 0.32 mL, Acros),
2
(1.1 mg, 0.00508 mmol, 0.05 equiv., Alfa Aesar), 2-dicy-
0
0
0
0
2
3
lL, 1 mmol)
and 1,2-dimethoxyethane (0.64 mL) were placed in a sealed tube,
and degassed with argon. The suspension was then heated at 80 °C
for 48 h, and the crude mixture was directly purified by column
D
chromatography (CH
2
Cl
2
/MeOH, 95:5).
l
0
0
0
00
0
phy (CH
2
Cl
2
/MeOH, 9:1) to afford the product, LH1, as a white solid
4.1.9. N -([1,1 :3 ,1 -Terphenyl]-2 -yl)-biotinyl hydrazide (LH2)
Benzeneboronic acid (37.2 mg) was subjected to the general
procedure for Suzuki coupling with 7 to give the title compound
(LH2, 13.1 mg, 26%) as a yellow solid. Melting point = 111–
20
(
8.2 mg, 16%). Melting point = 134–136 °C; [
a
]
D
+27.2 (c 0.25,
ꢁ1
CHCl
3
);
m
max (thin film)/cm
3313, 2925, 1697, 1647, 1541,
1
1
457, 761, 702; H NMR (500 MHz, MeOD) d 7.56 (m, 1H), 7.48
2
0
ꢁ1
(
m, 2H), 7.44–7.39 (m, 5H), 6.91 (br. s, 1H), 4.50 (dd, J = 7.7,
112 °C; [
3057, 2929, 2860, 1701, 1455, 1420, 758, 703; 1H NMR
(500 MHz, CDCl ) d 7.56 (m, 4H), 7.47 (t, J = 7.7 Hz, 4H), 7.37 (m,
a
]
D
3
+75.6 (c 0.25, CHCl ); mmax (thin film)/cm 3254,
4
2
.6 Hz, 1H), 4.30 (dd, J = 7.9, 4.4 Hz, 1H), 3.20 (m, 1H), 2.98 (m,
H), 2.93 (dd, J = 12.8, 4.9 Hz, 1H), 2.84 (m, 2H), 2.72 (d,
3
1
3
J = 12.6 Hz, 1H), 2.07 (t, J = 7.6 Hz, 2H), 1.75–1.39 (m, 6H);
NMR (126 MHz, MeOD) d 176.0, 172.4, 166.1, 142.0, 141.5, 136.9,
C
2H), 7.16 (m, 2H), 7.08 (m, 1H), 6.44 (br. s, 1H), 6.27 (br. s, 1H),
6.01 (br. s, 1H), 5.30 (br. s, 1H), 4.43 (m, 1H), 4.19 (m, 1H), 3.03
1
3
30.2, 130.2, 130.0, 129.3, 128.7, 63.4, 61.6, 57.0, 41.1, 39.9, 39.7,
(m, 1H), 2.83 (dd, J = 12.8, 4.9 Hz, 1H) 2.63 (d, J = 12.6 Hz, 1H),
1.62–1.08 (m, 8H); C NMR (126 MHz, CDCl ) d 171.3, 163.6,
3
142.8, 140.4, 131.2, 130.4, 129.0, 128.8, 127.3, 122.1, 61.7, 60.1,
+
13
6.8, 29.8, 29.5, 26.6; HRMS (ESI ) calculated for [C31
H
34
N
4
O
3
+
S+Na] 565.2244, found 565.2223(
D
3.7 ppm).

M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
5485
+
5
5.3, 40.4, 33.2, 28.2, 28.0, 24.5; HRMS (ESI ) calculated for
amounts, 0.5–3 lL, of crystallization buffer (75% ammonium sul-
+
[
C
H
28 30
N
4
O
2
S+Na] 509.1982, found 509.1974 (
D
1.5 ppm).
fate, 25% 1 M sodium acetate pH 4.5) in a sitting drop 2 well Swis-
sci plate (Hampton Research). Crystals were typically obtained
after a few hours at rt and reached a maximum size after 1–2 days.
Crystals of A86D streptavidin with LH4 were obtained by mixing
equal amounts of 1 mM protein with 2 mM LH4 in PBS with 30%
2-methyl-2,4-pentanediol (MPD). This stock was used to set up sit-
ting drops in a 2 well Swissci plate with the best diffracting crystals
(2 weeks at rt) obtained with condition A9 of the Morpheus screen:
0.1 M bicine/Trizma base pH 8.5, 10% w/v polyethylene glycol
20,000, 20% v/v polyethylene glycol monomethyl ether 550,
30 mM magnesium chloride and 30 mM calcium chloride.50
Prior to data collection, crystals were mounted and immersed
into liquid nitrogen, using 6 M sodium formate as a cryoprotectant
(LH1) or 30% MPD with 2 mM ligand (LH3 and LH4) for crystals
obtained using ammonium sulfate. Crystals of A86D streptavidin
bound to LH4 were frozen directly without further cryoprotectant.
All crystallographic data were collected at the Diamond Light
Source, Didcot, UK. Data for SA/LH1, trans-divalent SA/LH4 and
A86D SA/LH4 were collected at 100 K at the IO2 beamline using
a Pilatus 6 M detector and 0.98 Å wavelength beam. Data for
SA/LH3 were collected at 100 K at the IO4 beamline using a Pilatus
6 M-f detector and 0.98 Å wavelength beam.
0
0
0
00
4
4
.1.10. Dimethyl 2 -(biotinyl)hydrazinyl)-[1,1 :3 ,1 -terphenyl]-
0
0
,4 -dicarboxylate (LH3)
-Methoxycarbonylphenyl boronic acid (54.0 mg) was sub-
jected to the general procedure of Suzuki coupling with 7 to give
4
the title compound (LH3, 11.6 mg, 19%) as a colorless solid. Melting
2
0
ꢁ1
point = 93–95 °C; [
299, 2949, 2363, 2342, 1706, 1400, 1278; H NMR (500 MHz,
CDCl ) d 8.13 (d, J = 8.5 Hz, 4H), 7.65 (d, J = 8.5 Hz, 4H), 7.17 (m,
a
]
D
3
+28.8 (c 0.25, CHCl ); mmax (thin film)/cm
1
3
3
2
5
(
2
H), 7.12 (m, 1H), 6.48 (d, J = 4.1 Hz, 1H), 6.20 (d, J = 4.4 Hz, 1H),
.83 (br. s, 1H), 5.01 (br. s, 1H), 4.45 (dd, J = 7.6, 5.0 Hz, 1H), 4.21
m, 1H), 3.95 (s, 6H), 3.02 (m, 1H), 2.88 (dd, J = 12.9, 5.0 Hz, 1H),
.84 (d, J = 12.9 Hz, 1H), 1.67–1.12 (m, 8H); 13C NMR (126 MHz,
CDCl ) d 171.6, 166.9, 163.3, 145.2, 142.7, 130.8, 130.4, 130.1,
29.1, 129.0, 122.4, 61.7, 60.0, 55.3, 52.3, 40.4, 33.0, 28.2, 28.0,
3
1
2
+
+
4.6; HRMS (ESI ) calculated for [C32
ꢁ1.5 ppm).
34
H N
4
O
6
S+Na] 625.2091,
found 625.2101 (
D
0
00
0
0
00
4
2
.1.11. N -(4,4 -Bis(morpholinosulfonyl)-[1,1 :3 ,1 -terphenyl]-
-yl)-biotinyl hydrazide (LH4)
0
4
-(4-Morpholinylsulfonyl)phenylboronic acid (83.0 mg) was
subjected to the general procedure for Suzuki coupling with 7 to
give the title compound (LH4, 32.7 mg, 41%) as a yellow solid.
4.2.3. Structure solution and refinement
20
Melting point = 162–165 °C; [
film)/cm
a
]
D
+22.4 (c 0.25, CHCl
3326, 2921, 2859, 2247, 1698, 1165, 728; H NMR
500 MHz, CDCl ) d 7.83 (d, J = 8.5 Hz, 4H), 7.75 (d, J = 8.5 Hz, 4H),
.19–7.10 (m, 3H), 7.02 (br. s, 1H), 6.10 (br. s, 1H), 5.99 (br. s,
3
);
m
max (thin
Data were auto-indexed and integrated using the xia2 program
ꢁ1
1
51
upon collection. Structures were phased by molecular replace-
2
7
(
3
ment from PDB 3RY1 (apo-streptavidin) and further refined
5
2
7
using Phenix, through a Python graphical user interface. Models
were altered to fit better the electron density using Coot.53
Throughout refinement, all data were included from lowest to
the highest resolution and anisotropic temperature factors were
refined. Models were continually evaluated using MolProbity.54
Crystallographic Information File (CIF) restraints for LH ligands
were calculated using the GRADE server (Global Phasing Ltd):
http://grade.globalphasing.org/cgi-bin/grade/server.cgi. Refine-
ment statistics and other structure factors for the four structures
are shown in Table 1. Structures were visualized and images for
figures prepared using PyMOL (Schrödinger, LLC). Electron density
maps were visualized and images prepared (contoured at 1 rmsd)
1
H), 5.13 (br. s, 1H), 4.43 (dd, J = 7.6, 5.0 Hz, 1H), 4.21 (dd, J = 7.6,
4
.7 Hz, 1H), 3.77 (m, 8H), 3.07 (m, 8H), 3.04 (m, 1H), 2.84 (dd,
1
3
J = 12.9, 5.0 Hz, 1H), 2.64 (d, J = 12.6, 1H), 1.75–1.06 (m, 8H);
C
NMR (126 MHz, CDCl
3
) d 172.1, 163.5, 145.5, 142.7, 133.8, 131.2,
1
2
8
29.8, 129.2, 128.4, 122.3, 66.1, 61.7, 60.0, 55.5, 46.0, 40.4, 33.1,
8.2, 28.0, 24.6; HRMS (ESI ) calculated for [C36H N O S +Na]
44 6 8 3
+
+
07.2275, found 807.2301 (
D
ꢁ3.3 ppm).
4
4
.2. Biological assays
.2.1. Constructs, protein expression and purification
pET21 containing core streptavidin (SA), core streptavidin
4
7
55
with the CCP4 mg program. Hydrogen bonds were determined
tag (SAe),16 A86D streptavidin and Dead
48
in Coot, with a 3.5 Å cut-off. 2mFo-DFc images, mFo-DFc images,
53
with a C-terminal Glu
streptavidin (D) (the N23A, S27D, S45A mutant of core streptavi-
din showing negligible biotin binding) have been previously
6
2
2
and omit maps of the ligand in each structure are presented in
Supplementary Figures S1, S3, S4 and S6. The electron density for
residues 46–48 in the A86D streptavidin/LH4 structure was not
clearly defined and these residues were modeled using standard
stereochemical restraints in the subunit where LH4 ligand was
clearly observed. In the A86D subunit where LH4 was only partially
observed, 46–48 was not modeled. The coordinates and structure
factors for the crystal structures been deposited in the Protein Data
Bank with accession codes 4CPE for wild-type streptavidin with
LH1, 4CPF for wild-type streptavidin with LH3, 4CPH for trans-
divalent streptavidin with LH4, and 4CPI for A86D streptavidin
with LH4. The best resolved ligand for each structure is bound to
chain B (4CPE), chain B (4CPF), chain C (4CPH) and chain D (4CPI).
described. pET21 containing A86D core streptavidin with
a
C-terminal Glu tag (A86D-SAe) was cloned by QuikChange from
6
48
SAe, using primers previously described. The A86D-SAe construct
was verified by sequencing; GenBank Accession Number KJ401123.
All streptavidin subunits were expressed in Escherichia coli and
then refolded from inclusion bodies by dilution into PBS as previ-
4
9
ously described. After refolding, the protein was first purified
16
on iminobiotin-sepharose as described. Monovalent streptavidin
SAe1D3), A86D monovalent streptavidin (A86D-SAe1D3), and 1,3
trans-divalent streptavidin (SAe2D2) were purified by ion-
(
16
exchange chromatography. All proteins were dialyzed into PBS
and stored at 5–10 mg/mL. Protein concentrations were calculated
from A280 via ProtParam. The concentrations of all the streptavidin
forms refer to the concentration of monomer.
4.2.4. Biotin-4-fluorescein equilibrium assay
Monovalent streptavidin or A86D monovalent streptavidin
(
55 nM) was incubated for 1 h at rt with 11 nM biotin-4-fluores-
4
.2.2. Crystallization and data collection
cein (Life Technologies) in 90 L total volume. Then 10 L of ligand
l
l
Crystals of the streptavidin/LH ligand complexes were obtained
by the sitting drop crystallization method. Ligands were dissolved
at 100 mM in DMSO. Briefly, 1 mM protein was incubated with
at varying concentrations in DMSO was added and incubated for
48 h at 37 °C in closed PCR tubes. Samples were allowed to cool
to rt for 1 h before being transferred to a black 96 well plate. The
fluorescence increase of biotin-4-fluorescein upon unbinding from
streptavidin was monitored from kex = 485 nm and kem = 520 nm
using a PHERAstar FS plate reader (BMG Labtech). The percentage
2
mM ligand for 1 h at rt in PBS with 10% DMSO. Samples were
then centrifuged for 3 min at 17,000 g to remove any precipitate.
L of the protein ligand solution was then mixed with varying
2
4
1
l

5
486
M. Fairhead et al. / Bioorg. Med. Chem. 22 (2014) 5476–5486
dissociation was calculated as (signal with ligand ꢁ signal without
ligand)/(signal without quenching ꢁ signal without ligand) ꢀ 100.
The signal without quenching was taken as the biotin-4-fluores-
cein fluorescence in the absence of streptavidin. Experiments were
22. Howarth, M.; Chinnapen, D. J.; Gerrow, K.; Dorrestein, P. C.; Grandy, M. R.;
Kelleher, N. L.; El Husseini, A.; Ting, A. Y. Nat. Methods 2006, 3, 267.
2
2
3. Gallivan, J. P.; Dougherty, D. A. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 9459.
4. Kada, G.; Falk, H.; Gruber, H. J. Biochim. Biophys. Acta 1999, 1427, 33.
25. Hirsch, J. D.; Eslamizar, L.; Filanoski, B. J.; Malekzadeh, N.; Haugland, R. P.;
Beechem, J. M.; Haugland, R. P. Anal. Biochem. 2002, 308, 343.
repeated 3 times. Biotin, biocytin, desthiobiotin and (±)-a-lipoic
2
2
6. Mock, D. M.; Nyalala, J. O.; Raguseo, R. M. J. Nutr. 2001, 131, 2208.
7. Le Trong, I.; Wang, Z.; Hyre, D. E.; Lybrand, T. P.; Stayton, P. S.; Stenkamp, R. E.
Acta Crystallogr. D Biol. Crystallogr. 2011, 67, 813.
acid were from Sigma-Aldrich and stocks prepared at 100 mM in
DMSO. The one-site competitive binding fit was calculated using
GraphPad Prism, giving an EC50 with 95% confidence intervals as
28. Hyre, D. E.; Le Trong, I.; Freitag, S.; Stenkamp, R. E.; Stayton, P. S. Protein Sci.
2000, 9, 878.
2
well as a R Pearson correlation coefficient value.
2
3
9. Chivers, C. E.; Koner, A. L.; Lowe, E. D.; Howarth, M. Biochem. J. 2011, 435, 55.
0. Plazuk, D.; Zakrzewski, J.; Salmain, M.; Blauz, A.; Rychlik, B.; Strzelczyk, P.;
Bujacz, A.; Bujacz, G. Organometallics 2013, 32, 5774.
Acknowledgments
31. Wilbur, D. S.; Hamlin, D. K.; Chyan, M. K.; Kegley, B. B.; Pathare, P. M. Bioconjug.
Chem. 2001, 12, 616.
M.F. and M.H. were funded by the Biotechnology and Biological
Sciences Research Council (BBSRC). E.D.L. was funded by the Uni-
versity of Oxford Department of Biochemistry. L.K.M.C. was funded
by Oxford Synthesis Connections (OxSynC) from the University of
Oxford and the Engineering and Physical Sciences Research Council
32. Chivers, C. E.; Crozat, E.; Chu, C.; Moy, V. T.; Sherratt, D. J.; Howarth, M. Nat.
Methods 2010, 7, 391.
3. Chu, V.; Freitag, S.; Le Trong, I.; Stenkamp, R. E.; Stayton, P. S. Protein Sci. 1998,
3
7, 848.
34. Le Trong, I.; Chu, V.; Xing, Y.; Lybrand, T. P.; Stayton, P. S.; Stenkamp, R. E. Acta
Crystallogr. D Biol. Crystallogr. 2013, 69, 968.
3
5. Park, S. I.; Shenoi, J.; Frayo, S. M.; Hamlin, D. K.; Lin, Y.; Wilbur, D. S.; Stayton, P.
S.; Orgun, N.; Hylarides, M.; Buchegger, F.; Kenoyer, A. L.; Axtman, A.; Gopal, A.
K.; Green, D. J.; Pagel, J. M.; Press, O. W. Clin. Cancer Res. 2011, 17, 7373.
6. Dixon, R. W.; Radmer, R. J.; Kuhn, B.; Kollman, P. A.; Yang, J.; Raposo, C.; Wilcox,
C. S.; Klumb, L. A.; Stayton, P. S.; Behnke, C.; Le Trong, I.; Stenkamp, R. J. Org.
Chem. 1827, 2002, 67.
(
EPSRC). We thank Claire Chivers for cloning of A86D streptavidin.
3
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmc.2014.07.029.
37. Le Trong, I.; Aubert, D. G.; Thomas, N. R.; Stenkamp, R. E. Acta Crystallogr. D Biol.
Crystallogr. 2006, 62, 576.
38. Terai, T.; Maki, E.; Sugiyama, S.; Takahashi, Y.; Matsumura, H.; Mori, Y.;
Nagano, T. Chem. Biol. 2011, 18, 1261.
39. Weber, P. C.; Wendoloski, J. J.; Pantoliano, M. W.; Salemme, F. R. J. Am. Chem.
Soc. 1992, 114, 3197.
References and notes
1
2
.
.
Rago, F.; Cheeseman, I. M. J. Cell Biol. 2013, 200, 557.
Carey, S. P.; D’Alfonso, T. M.; Shin, S. J.; Reinhart-King, C. A. Crit. Rev. Oncol.
Hematol. 2012, 83, 170.
Vogel, V.; Sheetz, M. P. Curr. Opin. Cell Biol. 2009, 21, 38.
Hoefer, I. E.; den Adel, B.; Daemen, M. J. Cardiovasc. Res. 2013, 99, 276.
Evans, E. A.; Calderwood, D. A. Science 2007, 316, 1148.
Grubmuller, H.; Heymann, B.; Tavan, P. Science 1996, 271, 997.
Dror, R. O.; Dirks, R. M.; Grossman, J. P.; Xu, H.; Shaw, D. E. Annu. Rev. Biophys.
40. Pazy, Y.; Kulik, T.; Bayer, E. A.; Wilchek, M.; Livnah, O. J. Biol. Chem. 2002, 277,
30892.
41. Freitag, S.; Chu, V.; Penzotti, J. E.; Klumb, L. A.; To, R.; Hyre, D.; Le Trong, I.;
Lybrand, T. P.; Stenkamp, R. E.; Stayton, P. S. Proc. Natl. Acad. Sci. U.S.A. 1999, 96,
8384.
42. DeChancie, J.; Houk, K. N. J. Am. Chem. Soc. 2007, 129, 5419.
43. Meenan, N. A.; Sharma, A.; Fleishman, S. J.; Macdonald, C. J.; Morel, B.; Boetzel,
R.; Moore, G. R.; Baker, D.; Kleanthous, C. Proc. Natl. Acad. Sci. U.S.A. 2010, 107,
10080.
44. Liebeschuetz, J.; Hennemann, J.; Olsson, T.; Groom, C. R. J. Comput. Aided Mol.
Des. 2012, 26, 169.
3
4
5
6
7
.
.
.
.
.
2
012, 41, 429.
8
9
.
.
Green, N. M. Methods Enzymol. 1990, 184, 51.
Rico, F.; Moy, V. T. J. Mol. Recognit. 2007, 20, 495.
1
0. Izrailev, S.; Stepaniants, S.; Balsera, M.; Oono, Y.; Schulten, K. Biophys. J. 1997,
2, 1568.
1. Laitinen, O. H.; Hytonen, V. P.; Nordlund, H. R.; Kulomaa, M. S. Cell. Mol. Life Sci.
006, 63, 2992.
45. Kulhanek, J.; Pytela, O. Collect. Czech. Chem. Commun. 1997, 62, 913.
46. Aubineau, T.; Cossy, J. Chem. Commun. 2013, 49, 3303.
47. Sano, T.; Pandori, M. W.; Chen, X.; Smith, C. L.; Cantor, C. R. J. Biol. Chem. 1995,
270, 28204.
7
1
2
1
1
2. Sano, T.; Vajda, S.; Cantor, C. R. J. Chromatogr. B Biomed. Sci. Appl. 1998, 715, 85.
3. Weber, P. C.; Ohlendorf, D. H.; Wendoloski, J. J.; Salemme, F. R. Science 1989,
48. Crozat, E.; Meglio, A.; Allemand, J. F.; Chivers, C. E.; Howarth, M.; Venien-Bryan,
C.; Grainge, I.; Sherratt, D. J. EMBO J. 2010, 29, 1423.
49. Howarth, M.; Ting, A. Y. Nat. Protoc. 2008, 3, 534.
50. Gorrec, F. J. Appl. Crystallogr. 2009, 42, 1035.
243, 85.
1
1
4. Stayton, P. S.; Freitag, S.; Klumb, L. A.; Chilkoti, A.; Chu, V.; Penzotti, J. E.; To, R.;
Hyre, D.; Le Trong, I.; Lybrand, T. P.; Stenkamp, R. E. Biomol. Eng. 1999, 16, 39.
5. Zimbron, J. M.; Heinisch, T.; Schmid, M.; Hamels, D.; Nogueira, E. S.; Schirmer,
T.; Ward, T. R. J. Am. Chem. Soc. 2013, 135, 5384.
6. Fairhead, M.; Krndija, D.; Lowe, E. D.; Howarth, M. J. Mol. Biol. 2014, 426, 199.
7. Wilbur, D. S.; Chyan, M. K.; Pathare, P. M.; Hamlin, D. K.; Frownfelter, M. B.;
Kegley, B. B. Bioconjug. Chem. 2000, 11, 569.
8. Wilbur, D. S.; Hamlin, D. K.; Chyan, M. K. Bioconjug. Chem. 2006, 17, 1514.
9. Mei, T. S.; Giri, R.; Maugel, N.; Yu, J. Q. Angew. Chem., Int. Ed. 2008, 47, 5215.
0. Suau, R.; Rico, R.; Najera, F.; Ortiz-Lopez, F. J.; Lopez-Romero, J. M.; Moreno-
Manas, M.; Roglans, A. Tetrahedron 2004, 60, 5725.
51. Winter, G. J. Appl. Crystallogr. 2010, 43, 186.
52. Terwilliger, T. C.; Grosse-Kunstleve, R. W.; Afonine, P. V.; Moriarty, N. W.;
Zwart, P. H.; Hung, L. W.; Read, R. J.; Adams, P. D. Acta Crystallogr. D Biol.
Crystallogr. 2008, 64, 61.
53. Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Acta Crystallogr. D Biol.
Crystallogr. 2010, 66, 486.
54. Davis, I. W.; Murray, L. W.; Richardson, J. S.; Richardson, D. C. Nucleic Acids Res.
2004, 32, W615–W619.
55. McNicholas, S.; Potterton, E.; Wilson, K. S.; Noble, M. E. Acta Crystallogr. D Biol.
Crystallogr. 2011, 67, 386.
1
1
1
1
2
2
1. Gu, W.; Wang, S. F. Eur. J. Med. Chem. 2010, 45, 4692.