Using NMR solvent water relaxation to investigate metalloenzyme - Ligand binding interactions

Article abstract of DOI:10.1021/jm901537q

This report demonstrates that solvent water relaxation measurements can be used for quantitative screening of ligand binding and for mechanistic investigations of enzymes containing paramagnetic metal centers by using conventional NMR instrumentation at high field. The method was exemplified using prolyl hydroxylase domain containing enzyme 2 (PHD2), a human enzyme involved in hypoxic sensing, with Mn(II) substituting for Fe(II) at the active site. KD values were determined for inhibitors that hinder access of water to the paramagnetic center. This technique is also useful for investigating the mechanism of suitable metalloenzymes, including order of ligand binding and modes of inhibition. 2009 American Chemical Society.

Full text of DOI:10.1021/jm901537q

J. Med. Chem. 2010, 53, 867–875 867
DOI: 10.1021/jm901537q
Using NMR Solvent Water Relaxation to Investigate Metalloenzyme-Ligand Binding Interactions
Ivanhoe K. H. Leung, Emily Flashman, Kar Kheng Yeoh, Christopher J. Schofield, and Timothy D. W. Claridge*
Department of Chemistry, University of Oxford, Chemistry Research Laboratory, 12 Mansfield Road, Oxford OX1 3TA, United Kingdom
Received October 16, 2009
This report demonstrates that solvent water relaxation measurements can be used for quantitative
screening of ligand binding and for mechanistic investigations of enzymes containing paramagnetic
metal centers by using conventional NMR instrumentation at high field. The method was exemplified
using prolyl hydroxylase domain containing enzyme 2 (PHD2), a human enzyme involved in hypoxic
sensing, with Mn(II) substituting for Fe(II) at the active site. KD values were determined for inhibitors
that hinder access of water to the paramagnetic center. This technique is also useful for investigating the
mechanism of suitable metalloenzymes, including order of ligand binding and modes of inhibition.
Introduction
the properties of the paramagnetic metal itself.8,10-12 The
same paramagnetic relaxation enhancement (PRE) effect can
be observed through solvent water proton relaxation rates.
Various groups have made use of this effect to study para-
magnetic metal binding interactions with proteins and other
macromolecules by monitoring changes in the bulk water
relaxation rate.13-17
NMR spectroscopy is widely used for the study of small
molecule-protein interactions; saturation transfer difference
(STDa),1 transferred NOE,2 and chemical shift perturbation
techniques have emerged over recent years to play useful roles
in ligand screening.3-6 Although widely employed, these
techniques may suffer from low intrinsic sensitivity and can
require careful optimization of experimental conditions. For
instance, STD enhancements can be rather weak and these
analyses typically require a large excess of ligand, which may
result in nonspecific binding or cause problems if the ligands
are of limited solubility. The transferred NOE method re-
quires a delicate ligand-protein ratio, and the results can be
difficult to interpret in cases of relatively large ligands such as
for peptide-protein or polysaccharide-protein interactions.
Recently, a novel competition method has been proposed
for screening small molecules binding to paramagnetic me-
talloproteins by employing water as the reporter molecule and
monitoring changes in the bulk water proton longitudinal
relaxation rate (R1).18 In the absence of any ligand, providing
that access is possible, water molecules may chelate to the
paramagnetic metal and undergo exchange with water mole-
cules in the bulk solvent matrix. This leads to a net increase in
the bulk water relaxation rate. The binding of ligands to the
active site will normally hinder or prevent access of water
molecules to the paramagnetic center and so decrease the
observed bulk water relaxation rate.19 Because the exchange
of water will not normally be directly reliant on the exchange
kinetics of the reversibly forming protein-ligand complex (as
in the case for STD-NMR), this technique does not suffer
from the limits imposed by other ligand-based NMR screen-
ing methods.18 For example, while the association of very
strongly binding ligands is often difficult to detect directly
with ligand-observed methods, alterations in water accessi-
bility may still arise under these conditions and remain readily
detectable through water relaxation measurements. Thus,
provided the enzyme under investigation contains a paramag-
netic center at the binding site, this present approach can
overcome the difficulties and limitations often associated with
other ligand based NMR screening techniques.
Chemical shift perturbation using H-15N HSQC requires
1
access to 15N labeled protein and can be difficult to apply in
systems larger than 25 kDa.7 These factors may limit the
applicability of these methods as primary screening tools.
Relaxation enhancement has been used as a tool to study
paramagnetic metalloprotein-ligand interactions since the
early 1960s;8 the technique is based on the fact that the
magnetic moment of an electron is 658 times greater than that
of a proton nuclear spin.9 Therefore, in the presence of a
paramagnetic metalloprotein, the longitudinal and transverse
relaxation rates of nuclei within a binding ligand may be
enhanced, while for nonbinders the relaxation rates will
remain unchanged. The changes in relaxation behavior de-
pend on factors including the applied magnetic field (B0),
molecular correlation times, accessibility of the active site, and
*To whom correspondence should be addressed. Phone: þ44 (0)1865
275 658. Fax: þ44 (0)1865 285 002. E-mail: tim.claridge@chem.ox.ac.
uk.
The effect of paramagnetic relaxation is maximized at low
field, generally between 0.01 to 100 MHz;18 hence the original
application of this solvent-based method for ligand screening
was focused toward the use of specialist low field NMR
instruments or fast field-cycling relaxometers.18,20 While it
was demonstrated that it is possible to conduct screen-
ing experiments at high field (400 MHz) on very small
sample volumes (∼2 μL) using an inhibitor with a KD in the
a Abbreviations: 2OG, 2-oxoglutarate; NOG, N-oxalylglycine; PHD,
prolyl hydroxylase; STD, saturation transfer difference; PRE, para-
magnetic relaxation enhancement; HIF, hypoxia inducible factor; ESI-
MS, electrospray ionization mass spectrometry; ILOEs, interligand
NOEs; CODD, C-terminal oxygen-dependent degradation domain;
NODD, N-terminal oxygen-dependent degradation domain; BNS, bi-
cyclic naphthalenylsulfonyl compound; BIQ, bicyclic isoquinolinyl
compound.
r
2009 American Chemical Society
Published on Web 12/21/2009
pubs.acs.org/jmc

868 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2
Leung et al.
Scheme 1. Outline Reaction Scheme for PHD2 Catalysis; 2OG = 2-Oxoglutarate
nanomolar region, the application to screen weaker inhibitors
has not been demonstrated at high field and may suggest
limited applicability in biochemical research laboratories
employing conventional high-field NMR spectrometers.
By using the catalytic domain of human prolyl hydroxylase
domain 2 (PHD2181-426, or PHD2 hereafter) as a model
system, we have optimized the experimental conditions of this
water relaxation technique at high field and demonstrate its
wider application as a NMR screening technique using con-
ventional instrumentation. Furthermore, we have also de-
monstrated its potential as a method to investigate metallo-
enzyme reaction and inhibition mechanisms by monitoring
the formation of complexes between protein, cofactor and
substrates.
2-fold molar excess of Mn(II) was used to saturate the active
site. However, in cases where the small molecule is a non-
binder or a very weak binder, the presence of free paramag-
netic metal ions in solution can potentially be a source of
artifacts through direct chelation of the free metal and
should be avoided. Therefore, in this situation, a 1:1 molar
ratio of protein to Mn(II) was used so that any contribution
to relaxation rate changes from small molecules chelating to
free metal ions was minimized.
Longitudinal relaxation time constants (T1) were mea-
sured at 500 MHz using the standard inversion recovery
scheme.41 Solvent water T1 constants were measured directly
in a H2O-D2O mixture as opposed to the originally pro-
posed method in which samples are measured in 100%
H2O.18 A H2O-D2O ratio of 12.5%:87.5% was chosen so
that the range of T1 values could be kept roughly between 1.8
to 3.5 s in order to keep the length of inversion recovery
experiments reasonably short. Conventional 3 mm tubes
were used to strike a balance between the ease of titration
and avoiding radiation damping that may arise from the
intense H2O resonance. The small proportion of H2O to D2O
would also assist in the avoidance of radiation damping. The
reduction of filling factor when using a 5 mm probe was
not of concern because of the strong HDO signal being
measured.
Ligand Screening. Figure 1a presents solvent water relaxa-
tion data for three different 2OG titrations from separate
sample preparations by plotting the rate of relaxation (R1)
against the ligand concentration. Analysis in the presence of
succinate, a product of PHD2 catalysis and a relatively weak
inhibitor,42 is plotted for comparison purposes. In this
presentation format, variability between titration curves is
apparent for 2OG and it is not immediately apparent
whether 2OG or succinate is the stronger binder. The
absolute R1 values, such as the initial relaxation rate when
no ligand is added (R1(0)) and the minimum relaxation rate
when a 1:1 complex is formed, are strongly influenced by
variations in the concentrations of paramagnetic ion and
protein. Small variations in Mn(II) and protein concentra-
tions from separate sample preparation stages will therefore
influence the ability to identify differences in binding affi-
nities when comparing across samples.
However, the actual parameter of interest is the relative
change in observed relaxation rate on titration of the ligand.
We therefore propose that in order to normalize these
variations, it is advantageous to plot and compare the
relative relaxation rates R1/R1(0) against the titrated ligand
concentration, as shown in Figure 1b. With this method of
analysis, it is possible to readily distinguish between strong
and weak binders. The curve for a strong binder (2OG37) is
characterized by a steep slope and rapid plateauing at, or
close to, a 1:1 protein-ligand ratio, whereas the curve for a
weak binder (succinate42) is characterized by a gentler slope
which does not plateau until a large ligand excess is reached.
The R1 ratio at which this saturating equilibrium condition is
reached is likely governed by factors including kinetics of the
Results
Methodology. PHD2 is a Fe(II) and 2-oxoglutarate (2OG)
dependent oxygenase that is involved in a key step in human
oxygen sensing as part of the response to limiting oxygen levels
(hypoxia) (for reviews, see refs 21 and 22). Under normal
oxygen levels (normoxia), PHD2 initiates the oxygen depen-
dent degradation of the R-subunit of hypoxia inducible factor
(HIF-1R) by catalyzing the hydroxylation of two prolyl resi-
dues (positions 402 and 564 in human HIF-1R).23,24 The
PHD2-Fe(II) complex forms a ternary complex with the
cofactor 2OG and its protein substrate HIF-R (or the peptide
fragments HIF-1R556-574 (C-terminal oxygen-dependent de-
gradation domain, CODD) and HIF-1R395-413 (N-terminal
oxygen-dependent degradation domain, NODD)) (Scheme 1).
In the case of the peptide substrates being described here, the
hydroxyproline (Hyp) peptides CODDHyp564 or NODDHyp402
are produced, with carbon dioxide and succinate as copro-
ducts.25 PHD2 is a current pharmaceutical target for new
treatments for ischemic disease and anemia.26,27 Following
the finding that PHD activity is inhibited by N-oxalylglycine
(NOG),28,29 a 2OG analogue, several groups have described
the identification of inhibitors of PHD2.30-36 Herein we report
the application of bulk water relaxation measurements to
investigate mechanistic features and inhibition of PHD2.
An absolute requirement for the relaxation-based screen-
ing experiment is the presence of a paramagnetic center
inside the protein that is accessible to bulk water. PHD2 is
unusual among other 2OG oxygenases in that it forms a
stable complex with diamagnetic high spin Fe(II).37
Although there may be small amounts of paramagnetic
Fe(III) species inside the PHD2 active site, quantitative
information is difficult to extract. Previous studies have
indicated that the Fe(II) metal ion in the PHD2 active site
can be substituted by different transition metals.38 Mn(II)
forms a 1:1 complex with apo-PHD238,39 and was selected as
the metal of choice in this study because of its favorable
paramagnetic properties.8,40 By measuring the bulk water
relaxation rates at a fixed Mn(II) concentration and varying
PHD2 concentrations, we confirmed that the binding of
Mn(II) to apo-PHD2 is strong (see Supporting Information,
Figure S1). In the screening experiments described below, a

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2 869
Figure 1. Titrations of 2OG in three separate sample preparations (blue squares, red circles, and red diamonds), and the titration of succinate
(green triangles) into 50 μM PHD2. Plotted against ligand concentration is (a) the directly measured solvent relaxation rate (R1) and (b) the
relative relaxation rate R1/R1(0), in which R1(0) is the initial relaxation rate when no ligand is added. Slight differences in Mn(II) and protein
concentrations from different sample preparations complicate comparisons across samples, but by normalizing the rates as R1/R1(0), binding
differences are made apparent. In this and all subsequent figures, dashed lines are included to aid visualization.
reversibly forming protein-ligand complex, the accessibly of
water molecules, their mode of binding (e.g., directly to metal
or otherwise), and the number of water molecules that are
being displaced. The position of the plateau is also somewhat
sensitive to the intrinsic variability that occurs between
different sample preparations, arising from difficulties in
determining protein concentrations accurately and the slight
variation in metal ion concentration between different stock
solutions. Such variability is apparent for the 2OG titration
data in Figure 1b, and we observe this to be typical behavior
between discrete sample preparations. For clarity therefore,
the normalized rates, R1/R1(0), versus ligand concentration
presentation format is used throughout the remainder of this
report.
should be given when choosing the relative amounts of protein
and ligand for the experiment because there is trade-off bet-
ween protein concentration, total experiment time, and the
sensitivity of the method in differentiating ligands of similar
binding affinity (Figure 2). When using ∼25 μM protein and
100 μM Mn(II), the experimental time is shortened because
there is a higher concentration of free Mn(II) in solution that is
readily accessible to water molecules, thus leading to faster
relaxation rates (R1(0) =0.60 s-1). However, in this case, the
extent of reduction in relaxation rates on ligand addition is
diminished with the titration curve plateauing at around 90%,
reducing the separation between the curves of similar affinity
ligands. In contrast, by using ∼75 μM protein with 100 μM
Mn(II), the variation in rates is enhanced, now plateauing at
around 65%, and the method becomes more sensitive to small
differences in binding affinities. One consequence of this,
however, is that the time required for the experiment is greater
due to the higher concentration of protein bound Mn(II)
The amounts of protein and ligand molecules required for
this relaxation method are comparable to, or less than, other
NMR based screening techniques because it is the intense HDO
signal that is being measured. Nevertheless, consideration

870 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2
Leung et al.
Figure 2. Titration of 2OG into 25 μM PHD2 (blue squares), 50 μM PHD2 (red diamonds), and 75 μM PHD2 (green triangles) with 100 μM
Mn(II). The minimum R1 value depends on the residual relaxation rate of the H2O-D2O mixture, which in turn is dependent upon the amount
of free Mn(II) present in solution.
Figure 3. The titration of 2OG (solid triangles), NOG (solid diamonds), and BNS (solid circles) into 50 μM PHD2 with 100 μM Mn(II). The
data plotted in red and blue are conducted at 500 and 23 MHz, respectively.
species with correspondingly fewer free ions in solution (R1(0)=
0.41 s-1). As comparison, the water relaxation rate observed
measured at 313 K so that the rate of ligand exchange was
increased, enabling a more accurate determination of KD, as
suggested by Bertini et al.18 Such KD measurements become
feasible because only near stoichiometric amounts of ligand
are required.
As anticipated, the minimum relaxation rate is smaller at
low field, indicating greater sensitivity under these condi-
tions; however, the relative change of R1 between the three
ligands, and hence the calculated KD values, are essentially
the same at both low and high field. Qualitatively, results
obtained at high and low fields confirm that the bicyclic
inhibitor BNS is a much stronger binder than 2OG and
NOG. Quantitatively, the KD values obtained for 2OG at
both low and high field are ∼2 μM, while for NOG they are
∼3 μM (see Supporting Information, Figures S2-S5). Note
that these values are slightly higher than the reported values
for the apo-PHD2 solution (50 μM, 12.5% H2O) in the absence
of Mn(II) was 0.11 s-1
.
To investigate the validity and sensitivity of the method at
high field, we then conducted experiments at low field, as in
the work of Bertini et al.,18 and compared the results with our
high field data. Three PHD2-ligand systems were analyzed
for comparison. 2OG is a natural cofactor of PHD2 and has
a KD value of about 1 μM as determined by nondenaturing
ESI-MS,37 while N-oxalylglycine (NOG) is a 2OG mimic
that binds slightly weaker than 2OG.23 A potent bicyclic
naphthalenylsulfonyl inhibitor (BNS) was also tested
(Figure 3).34,36
The low field experiments were conducted in 100% H2O at
23 MHz, and both high and low field experiments were

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2 871
Table 1. Dissociation Constants (KD) Determined by the Solvent
Relaxation Technique at 298 K and IC50 Values34,46 for 2OG, Succinate
and a Range of Inhibitors
The substitution of the glycine moiety of BIQ by bulky
amino acids such as valine also sterically weakens the bind-
ing of BIQ inhibitors to PHD2. In fact, D- and L-Val BIQ
show very similar levels of weak binding as indicated by MS
analyses (Supporting Information, Figure S15). Solvent
relaxation experiments gave KD values of ∼33 μM and
∼35 μM for D- and L-Val BIQ, respectively (see Support-
ing Information, Figures S9-S10). Our results therefore
demonstrate and further support the accuracy and sensitivity
of this technique while using more conventional NMR
instrumentation.
Investigation of PHD2-Mn(II)-2OG-HIF-r Complex
Formation. CODD and NODD are 19 amino-acid peptide
substrates of PHD2 which are fragments of the natural
human HIF-1R; both are hydroxylated by PHD2 at the
trans-C4 position of a prolyl residue.23 In studies on the
binding of larger peptides such as these, it is often difficult to
apply ligand-based NMR techniques; for example, in the
transferred NOE method, it becomes difficult to distinguish
between NOE signals arising from free and bound peptides
because they have the same sign, while in saturation transfer
difference (STD) NMR, peak overlap can make it difficult to
irradiate selectively the protein resonances. Despite the
relatively large size of the NODD and CODD peptide
substrates, the application of the solvent relaxation techni-
que at high field is straightforward because only the HDO
signal is observed.
Previous biophysical and kinetic studies show that PHD2
has a preference for hydroxylating CODD over NODD,47-49
and Km values for CODD and NODD were reported as
37 and 44 μM, respectively.48 Our results suggest CODD is a
significantly stronger binder than NODD, with KD values of
∼14 and ∼85 μM, respectively, derived from solvent relaxa-
tion data (see Supporting Information, Figures S11-S12).
Because neither CODD or NODD directly chelate the metal
center,39 these data demonstrate that it is possible to screen
for compounds that bind in the vicinity of the active site
metal yet need not interact with this directly, provided that
their association displaces water molecules from the active
site metal (or indeed enhances their access) or alters their
exchange with bulk water. These data are also consistent
with the proposed PHD2 mechanism of hydroxylation
which, by analogy with work on other 2OG oxygenases,
involves displacement of water complexed to the active site
iron upon CODD or NODD binding, such that a coordi-
nation position becomes free for an oxygen molecule to
coordinate to the metal prior to the oxidative decarboxyl-
ation step.50-52
We have also investigated whether this technique can be
used for mechanistic studies. CODD and NODD are both
known to form complexes with the PHD2-metal complex
and 2OG. CODD binds to PHD2 in a well-defined cleft,
adopting a loose and extended conformation; 2OG chelates
directly to the metal center in a bidentate fashion.39 A clear
additive effect is observed by the water relaxation technique
when CODD is added to PHD2-2OG or when 2OG is
added to PHD2-CODD (Figure 4). When CODD is first
titrated into PHD2 (red squares, Figure 4a), the curve slowly
plateaus to around 90% at a substrate to protein ratio of 2:1,
indicating formation of a PHD2-CODD complex. Sub-
sequent addition of 2OG (red triangles, Figure 4a) further
decreases the R1/R1(0) ratio to around 70%, indicating the
formation of a PHD2-CODD-2OG complex. An equi-
valent effect is observed when 2OG is titrated into PHD2
for 2OG and NOG,23,37 possibly because in this work we are
analyzing binding to a PHD2-Mn(II) rather than a
PHD2-Fe(II) complex. Thus, both qualitative screening
and the quantitative determination of binding kinetics ap-
pear feasible at high field and with conventional, widely used
NMR instrumentation.
Determination of Inhibitor Binding Affinities. The bicyclic
isoquinolinyl inhibitor (BIQ) is a potent inhibitor of PHD2,
with an IC50 value of 0.07 μM.43-45 Most PHD2/HIF
hydroxylase inhibitors have amino acid side chains that bind
in the 2OG binding pocket, and previous work has shown
that the selectivity of these inhibitors can be modified by
varying the amino acid moiety.33 A series of D/L-amino acid
substituted BIQ analogues have been synthesized; the bind-
ing of some of these to PHD2 is limited by steric con-
straints.46 These were screened for their binding affinities
using the solvent relaxation technique (Table 1). We found
that the introduction of a methyl group (alanine derivative)
altered the selectivity of BIQ binding significantly, with
L-Ala-BIQ binding to PHD2 in a similar fashion to BIQ,
whereas D-Ala-BIQ was a much weaker binder. The relative
weaker binding of D-Ala-BIQ is likely due, at least in part, to
steric hindrance with the side chain of valine-376 in the 2OG
binding pocket.43,46 Solvent relaxation experiments gave KD
values of <1 μM and ∼22 μM for L-Ala-BIQ and D-Ala-BIQ
respectively (see Supporting Information, Figure S8), con-
sistent with the binding data obtained from nondenaturing
ESI-MS (see Supporting Information, Figure S15).

872 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2
Leung et al.
Figure 4. (a) The sequential titration of CODD (solid squares) and 2OG (solid triangles), and (b) simultaneous titration of equal molar CODD
and 2OG (solid diamonds) into 50 μM PHD2 as a function of total ligand concentrations. For (a), different colors are used to represent different
sequences of addition (red: CODD then 2OG; blue: 2OG then CODD).
first (blue triangles, Figure 4a) in which the R1/R1(0) ratio
plateaus at around 80%, and further titration of CODD
(blue squares, Figure 4a) decreases this to around 70%.
The titration of an equimolar mixture of 2OG-CODD
into PHD2 yields a similar net effect although the profile
displays a steeper slope than 2OG or CODD alone due to
the binding of both substrate and cofactor throughout. The
slight differences in the titration end points observed for
the three systems likely arises from the intrinsic variability
that occurs between different sample preparations, as
described for Figure 1b. The observation of these sequen-
tial binding profiles is consistent with previously reported
mechanistic and structural studies of PHD2, in which
PHD2 forms a ternary complex with 2OG and CODD,37
and in which 2OG and CODD are observed to bind to
neighboring sites in the protein active site.39,43 It also
provides insights into the extent of water accessibility upon
the binding of individual substrates. The KDs for 2OG
in the absence and presence of CODD are measured as
0.9 ( 0.1 and 1.0 ( 0.2 μM, respectively. Similar titration
behavior was also observed in ternary systems involv-
ing PHD2-2OG-NODD (see Supporting Information,
Figure S16).
This additive effect also allows the study of different
modes of inhibition. NOG is an unreactive analogue of
2OG, which competes with 2OG for binding to the active
site metal.29,53 As for 2OG, additive effects were observed
when CODD was added to PHD2-NOG, indicating the
formation of a ternary complex (Figure 5a). In contrast,
when CODD (or 2OG) is added to PHD2-BIQ or
PHD2-BNS inhibitor complexes, no additive effects were
observed (Figure 5b). BIQ and BNS chelate directly to the
metal center and therefore compete with 2OG binding, but in
addition they also appear able to block CODD from ap-
proaching the active site metal, consistent with crystallogra-
phy studies39,43 (or alternatively that they serve to completely
displace water from around the metal center such that
subsequent binding effects cannot be detected, although this

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2 873
Figure 5. Binding titration curves showing (a) the titration of CODD (red squares) into a solution of PHD2 þ NOG (red triangles) and (b) the
titration of CODD (red squares) into a solution of PHD2 þ BNS (red triangles). In both, the titration of CODD into PHD2 is also plotted as
reference (blue squares). Additive effects can be observed in (a) indicating the formation of a PHD2-NOG-CODD complex, while in (b) the
BNS inhibitor prevents CODD from approaching the PHD2 active site metal.
seems less likely as they are bidentate metal ligands). Thus,
the additional bulk associated with these ligands also offers a
second mode of inhibition through blocking substrate
binding. Information such as this could prove valuable in
inhibitor optimization.
screening methods by the upper and lower affinity limits
imposed by ligand exchange kinetics.
We have further demonstrated the wider applicability of
this technique to detect multiple binding events such as in the
formation of ternary protein-substrate complexes, which
may have potential applicability in mechanistic studies of
enzyme processes and in the design of enzyme inhibitors.
Other NMR techniques that are able to identify ternary
complex formation, such as interligand NOEs (ILOEs),54,55
lack the relative simplicity and sensitivity of the solvent
relaxation method presented here, requiring very long experi-
ment times and careful optimization of ligand-protein ratios
for their success. The ability to screen the formation of ternary
complexes in a high throughput fashion is therefore poten-
tially powerful in mechanistic studies on enzymes and in drug
design. For instance, Becattini et al. have reported a metho-
dologyfor drug discovery throughthe application of ILOEsin
which a library of small molecules with simple structures was
screened for ternary complex formation.56 Covalent linkage
Discussion and Conclusions
We have herein demonstrated the general applicability of
ligand screening with paramagnetic metalloproteins through
the study of bulk water relaxation, notably using high field
NMR spectrometers. It is important to emphasize that the
technique is relatively simple to carry out both from experi-
mental and data analysis perspectives. The method is highly
sensitivetoligand binding eventsdirectly at, and inthe vicinity
of, the paramagnetic metal center. It enables the differentia-
tion of binding affinities between ligands and, in favorable
cases, allows the value of dissociation constants to be deter-
mined. Furthermore, it is less restricted than many other

874 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2
Leung et al.
of molecules that bind simultaneously may then lead to larger
and potentially more potent inhibitors. Provided the target
enzyme is a paramagnetic metalloprotein, this solvent screen-
ing technique may provide an attractive alternative to ILOEs,
in that it may enable the straightforward identification of
ternary complex formation.
Supporting Information Available: Solvent relaxation mea-
surement for the binding of Mn(II) to PHD2; fitted KD curves
for the binding of small molecule inhibitors to PHD2; ESI-MS
data for the binding of BIQ analogues to PHD2; titration data
for NODD and 2OG; details for the synthesis of BIQ analogues.
This material is available free of charge via the Internet at http://
pubs.acs.org.
We also suggest this technique can be applied in a high-
throughput fashion for the ligand screening of paramagnetic
metalloproteins; by conducting a 2-point assay with and
without potential ligands, a significant change in solvent
relaxation rate may be correlated with ligand binding. The
screening of each potential ligand can be as short as 10 min
and may be performed using robotic sample handling.
References
(1) Mayer, M.; Meyer, B. NMR Characterization of Ligand Binding
by Saturation Transfer Difference NMR Spectroscopy. Angew.
Chem., Int. Ed. 1999, 38, 1784–1788.
(2) Ni, F. Recent Developments in Transferred NOE Methods. Prog.
Nucl. Magn. Reson. Spectrosc. 1994, 26, 517–606.
(3) Fejzo, J.; Lepre, C.; Xie, X. Application of NMR Screening in Drug
Discovery. Curr. Top. Med. Chem. 2003, 3, 81–97.
Experimental Section
(4) Lepre, C. A.; Moore, J. M.; Peng, J. W. Theory and Applications of
NMR-Based Screening in Pharmaceutical Research. Chem. Rev.
2004, 104, 3641–3675.
(5) Zartler, E. R.; Shapiro, M. J. Protein NMR-Based Screening in
Drug Discovery. Curr. Pharm. Des. 2006, 12, 3963–3972.
(6) Ludwig, C.; Guenther, U. L. Ligand Based NMR Methods for
Drug Discovery. Front. Biosci. 2009, 14, 4565–4574.
Materials. All chemicals were from Sigma-Aldrich, with the
exceptions of the CODD (sequence DLDLEMLAPYIPMDD-
DFQL) and NODD (sequence DALTLLAPAAGDTIISLDF)
peptides (Peptide Protein Research Limited), Tris-D11
(Cambridge Isotope Laboratories), and D2O (Apollo Scienti-
fic). BNS was a kind gift from Amgen. BIQ and analogues of
BIQ were synthesized by reported methods.46
(7) Kay, L. E.; Gardner, K. H. Solution NMR Spectroscopy Beyond
25 kDa. Curr. Opin. Struct. Biol. 1997, 7, 722–731.
Expression and Purification of PHD2. Recombinant PHD2
was expressed in Escherichia coli BL21 (DE3) and purified by
cation exchange and size exclusion chromatography as de-
scribed previously.37 Apo-proteins were obtained by incubation
with EDTA after cation exchange chromatography. Purified
PHD2 was buffer exchanged into 50 mM Tris-D11 (100% D2O,
pD 7.1) using Amicon Ultra-4 or Amicon Ultra-15 filter units
(MW cutoff: 10 kDa) (Millipore). At least three rounds of
dilution and concentration were performed at 277 K. PHD2
aliquots were stored at 193 K.
(8) Burton, D. R.; Forsen, S.; Karlstrom, G.; Dwek, R. A. Proton
Relaxation Enhancement (PRE) in Biochemistry: A Critical
Survey. Prog. Nucl. Magn. Reson. Spectrosc. 1979, 13, 1–45.
(9) Mohr, P. J.; Taylor, B. N.; Newell, D. B. CODATA Recommended
Values of the Fundamental Physical Constants: 2006. Rev. Mod.
Phys. 2008, 80, 633–730.
(10) Mildvan, A. S.; Engle, J. L. Nuclear Relaxation Measurements of
Water Protons and Other Ligands. Methods Enzymol. 1972, 26,
654–682.
(11) Mildvan, A. S.; Gupta, R. K. Nuclear Relaxation Measurements of
the Geometry of Enzyme-Bound Substrates and Analogs. Methods
Enzymol. 1978, 49, 322–359.
NMR Experiments. High-field experiments were conducted
at 500 MHz using a Bruker Avance II spectrometer equipped
with a standard 5 mm z-gradient TXI probe. Unless otherwise
stated, all experiments were conducted at 298 K. Conventional
3 mm diameter NMR tubes (Norell or Sigma-Aldrich) contain-
ing 50 μM PHD2 were used in all experiments unless otherwise
indicated. Solutions were buffered using Tris or Tris-D11 (pH
7.5) dissolved in 12.5% H2O and 87.5% D2O. A stock solution
of 800 μM Mn(II) was used throughout and diluted to final
concentrations of 50 or 100 μM. Stock solutions of ligand
molecules were made up freshly to concentrations between 1.6
and 2.5 mM. Samples of 1 μL or 2 μL were titrated into NMR
tubes. Inversion recovery experiments were performed with two
scans with a relaxation delay of at least 5 times T1 between
transients. Pulse tip-angle calibration using the single-pulse
nutation method was undertaken for each sample ensuring
accurate 90° and 180° pulses.57 Curve fitting to allow for the
possibility of radiation damping58,59 was used as a check during
development work and proved that under the conditions em-
ployed there was no contribution from this to the recovery
profiles recorded.
(12) Bertini, I.; Luchinat, C.; Parigi, G. Solution NMR of Paramagnetic
Molecules Applications to Metallobiomolecules and Models; Elsevier
Science: Amsterdam, 2001; pp 75-140.
(13) Eisinger, J.; Shulman, R. G.; Blumberg, W. E. Relaxation En-
hancement by Paramagnetic Ion Binding to Deoxyribonucleic
Acid Solutions. Nature 1961, 192, 963–964.
(14) Mildvan, A. S.; Cohn, M. Magnetic Resonance Studies of the
Interaction of the Manganous Ion with Bovine Serum Albumin.
Biochemistry 1963, 2, 910–919.
ꢀ
(15) Danchim, A.; Gueron, M. Proton Magnetic Relaxation Study of
the Manganese-Transfer-RNA Complex. J. Chem. Phys. 1970, 53,
3599–3609.
(16) Sherry, A. D.; Birnbaum, E. R.; Darnall, D. W. Proton Relaxation
Studies in H2O-D2O Mixtures The Binding of Manganese(II) to
Bovine Serum Albumin. Anal. Biochem. 1973, 52, 415–420.
(17) Budak, H. Water Proton Relaxation Rate Enhancements and
Association Constants for Mn(II) to Serum Proteins Determined
by NMR T1 Measurements. Z. Naturforsch. 2005, 60c, 807–812.
(18) Bertini, I.; Fragai, M.; Luchinat, C; Talluri, E. Water-Based
Ligand Screening for Paramagnetic Metalloproteins. Angew.
Chem., Int. Ed. 2008, 47, 4533–4537.
(19) In some circumstances, it can be envisaged that the binding of
ligands will increase access of water to the paramagnetic center.
(20) Kimmich, R.; Anoardo, E. Field-Cycling NMR Relaxometry.
Prog. Nucl. Magn. Reson. Spectrosc. 2004, 44, 257–320.
(21) Semenza, G. L. Hypoxia-Inducible Factor 1 (HIF-1) Pathway. Sci.
STKE 2007, 407, cm8.
Low-field experiments were conducted using an Oxford In-
struments MQC spectrometer operating at 23 MHz. Conven-
tional 5 mm diameter tubes (Norell) were used throughout. All
experimental details are as for the high field work, except the
solution was buffered in 100% H2O Tris (pH 7.5), and experi-
ments were conducted at 313 K.
(22) Chowdhury, R.; Hardy, A.; Schofield, C. J. The Human Oxygen
Sensing Machinery and its Manipulation. Chem. Soc. Rev. 2008,
37, 1308–1319.
(23) Epstein, A. C. R.; Gleadle, J. M.; McNeill, L. A.; Hewitson, K. S.;
O’Rourke, J.; Mole, D. R.; Mukherji, M.; Metzen, E.; Wilson,
M. I.; Dhanda, A.; Tian, Y. M.; Masson, N.; Hamilton, D. L.;
Jaakkola, P.; Barstead, R.; Hodgkin, J.; Maxwell, P. H.; Pugh,
C. W.; Schofield, C. J.; Ratcliffe, P. J. C. elegans EGL-9 and
Mammalian Homologs Define a Family of Dioxygenases that
Regulate HIF by Prolyl Hydroxylation. Cell 2001, 107, 43–54.
(24) Bruick, R. K.; McKnight, S. L. Oxygen Sensing Gets a Second
Wind. Science 2002, 295, 807–808.
Acknowledgment. We thank Oxford Instruments for
access to the low-field NMR spectrometer and K. Jassat
and Dr. A. Sagidullin (Oxford Instruments) for their advice
and technical assistance withlow-field NMR experiments. We
thank the EPSRC, the BBSRC, European Union, the Well-
come Trust, the Oxford Centre for Advanced Electron Spin
Resonance, and Amgen for funding. We thank Amgen for the
gift of the BNS compound.
(25) Hausinger, R. P. Fe(II)/R-Ketoglutarate-Dependent Hydroxylases
and Related Enzymes. Crit. Rev. Biochem. Mol. Biol. 2004, 39,
21–68.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 2 875
(26) Hewitson, K. S.; Schofield, C. J. The HIF Pathway as a Thera-
peutic Target. Drug Discovery Today 2004, 9, 704–711.
(27) Fraisl, P.; Aragonas, J.; Carmeliet, P. Inhibition of Oxygen Sensors
Schofield, C. J. Structural and Mechanistic Studies on the Inhi-
bition of the Hypoxia-inducible Transcription Factor Hydroxy-
lases by Tricarboxylic Acid Cycle Intermediates. J. Biol. Chem.
2007, 282, 3293–3301.
ꢀ
as a Therapeutic Strategy for Ischaemic and Inflammatory Disease.
Nat. Rev. Drug Discovery 2009, 8, 139–152.
(43) McDonough, M. A.; Li, V.; Flashman, E.; Chowdhury, R.; Mohr,
ꢀ
(28) Jaakkola, P.; Mole, D. R.; Tian, Y. M.; Wilson, M. I.; Gielbert, J.;
Gaskell, S. J.; von Kriegsheim, A.; Hebestreit, H. F.; Mukherji, M.;
Schofield, C. J.; Maxwell, P. H.; Pugh, C. W.; Ratcliffe, P. J.
Targeting of HIF-R to the von Hippel-Lindau Ubiquitylation
Complex by O2-Regulated Prolyl Hydroxylation. Science 2001,
292, 468–472.
(29) Mole, D. R.; Schlemminger, I.; McNeill, L. A.; Hewitson, K. S.;
Pugh, C. W.; Ratcliffe, P. J.; Schofield, C. J. 2-Oxoglutarate
Analogue Inhibitors of HIF Prolyl Hydroxylase. Bioorg. Med.
Chem. Lett. 2003, 13, 2677–2680.
C.; Lienard, B. M. R.; Zondlo, J.; Oldham, N. J.; Clifton, I. J.;
Lewis, J.; McNeill, L. A.; Kurzeja, R. J. M.; Hewitson, K. S.; Yang,
E.; Jordan, S.; Syed, R. S.; Schofield, C. J. Cellular Oxygen Sensing:
Crystal Structure of Hypoxia-Inducible Factor Prolyl Hydroxylase
(PHD2). Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 9814–9819.
(44) Warshakoon, N. C.; Wu, S.; Boyer, A.; Kawamoto, R.; Sheville, J.;
Renock, S.; Xu, K.; Pokross, M.; Evdokimov, A. G.; Walter, R.;
Mekel, M. A Novel Series of Imidazo[1,2-a]pyridine Derivatives as
HIF-1R Prolyl Hydroxylase Inhibitors. Bioorg. Med. Chem. Lett.
2006, 16, 5598–5601.
(30) Ivan, M.; Haberberger, T.; Gervasi, D. C.; Michelson, K. S.;
(45) Guenzler, P.; Volkmar, K.; Stephen, J.; Liu, D. Y.; Todd, W. PCT.
€
Gunzler, V.; Kondo, K.; Yang, H.; Sorokina, I.; Conaway,
Int. Appl. WO 2005007192, 2005.
ꢀ
(46) Stubbs, C. J.; Loenarz, C.; Mecinovic, J.; Yeoh, K. K.; Hindley, N.;
R. C.; Conaway, J. W.; Kaelin, W. G., Jr. Biochemical Purification
and Pharmacological Inhibition of a Mammalian Prolyl Hydro-
xylase Acting on Hypoxia-Inducible Factor. Proc. Natl. Acad. Sci.
U.S.A. 2002, 99, 13459–13464.
ꢀ
Lienard, B. M.; Sobott, F.; Schofield, C. J.; Flashman, E. Applica-
tion of a Proteolysis/Mass Spectrometry Method for Investigating
the Effects of Inhibitors on Hydroxylase Structure. J. Med. Chem.
2009, 52, 2799–2805.
(31) Schlemminger, I.; Mole, D. R.; McNeill, L. A.; Dhanda, A.;
Hewitson, K. S.; Tian, Y. M.; Ratcliffe, P. J.; Pugh, C. W.;
Schofield, C. J. Analogues of Dealanylalahopcin are Inhibitors of
Human HIF Prolyl Hydroxylases. Bioorg. Med. Chem. Lett. 2003,
13, 1451–1454.
(47) Ehrismann, D.; Flashman, E.; Genn, D. N.; Mathioudakis, N.;
Hewitson, K. S.; Ratcliffe, P. J.; Schofield, C. J. Studies on the
Activity of the Hypoxia-Inducible-Factor Hydroxylases Using an
Oxygen Consumption Assay. Biochem. J. 2007, 401, 227.
ꢀ
(32) Hewitson, K. S.; McNeill, L. A.; Schofield, C. J. Modulating the
hypoxia-inducible factor signaling pathway: Applications from
cardiovascular disease to cancer. Curr. Pharm. Des. 2004, 10,
821–833.
(48) Flashman, E.; Bagg, E. A. L.; Chowdhury, R.; Mecinovic, J.;
Loenarz, C.; McDonough, M. A.; Hewitson, K. S.; Schofield,
C. J. Kinetic Rationale for Selectivity Toward N- and C-terminal
Oxygen-Dependent Degradation Domain Substrates Mediated by
a Loop Region of Hypoxia-Inducible Factor Prolyl Hydroxylases.
J. Biol. Chem. 2008, 283, 3808–3815.
€
(33) McDonough, M. A.; McNeill, L. A.; Tilliet, M.; Papamicael, C. A.;
Chen, Q. Y.; Banerji, B.; Hewitson, K. S.; Schofield, C. J. Selective
Inhibition of Factor Inhibiting Hypoxia-Inducible Factor. J. Am.
Chem. Soc. 2005, 127, 7680–7681.
(49) Chan, D. A.; Sutphin, P. D.; Yen, S. E.; Giaccia, A. J. Coordinate
Regulation of the Oxygen-Dependent Degradation Domains of
Hypoxia-Inducible Factor 1R. Mol. Cell. Biol. 2005, 25, 6415–6426.
(50) Zhang, Z.; Ren, J.-S.; Harlos, K.; McKinnon, C. H.; Clifton, I. J.;
Schofield, C. J. Crystal Structure of a Clavaminate Synthase-
Fe(II)-2-Oxoglutarate-Substrate-NO Complex: Evidence for
Metal Centred Rearrangements. FEBS Lett. 2002, 517, 7–12.
(51) Zhang, Z.; Ren, J.; Stammers, D. K.; Baldwin, J. E.; Harlos, K.;
Schofield, C. J. Structural Origins of the Selectivity of the Trifunc-
tional Oxygenase Clavaminic Acid Synthase. Nat. Struct. Biol.
2000, 7, 127–133.
(52) Zhou, J.; Kelly, W. L.; Bachmann, B. O.; Gunsior, M.; Townsend,
C. A.; Solomon, E. I. Spectroscopic Studies of Substrate Interac-
tions with Clavaminate Synthase 2, a Multifunctional Alpha-KG-
Dependent Non-Heme Iron Enzyme: Correlation with Mechan-
isms and Reactivities. J. Am. Chem. Soc. 2001, 123, 7388–7398.
(53) Cunliffe, C. J.; Franklin, T. J.; Hales, N. J.; Hill, G. B. Novel
inhibitors of prolyl 4-hydroxylase. 3. Inhibition by the substrate
analog N-oxaloglycine and its derivatives. J. Med. Chem. 1992, 35,
2652–2658.
(54) London, R. E. Theoretical Analysis of the Inter-Ligand Over-
hauser Effect: A New Approach for Mapping Structural Relation-
ships of Macromolecular Ligands. J. Magn. Reson. 1999, 141, 301–
311.
(55) Li, D.; DeRose, E. F.; London, R. E. The Inter-Ligand Overhauser
Effect: A Powerful New NMR Approach For Mapping Structural
Relationships of Macromolecular Ligands. J. Biomol. NMR 1999,
15, 71–76.
(34) Tegley, C. M.; Viswanadhan, V. N.; Biswas, K.; Frohn, M. J.;
€
Peterkin, T. A. N.; Chang, C.; Burli, R. W.; Dao, J. H.; Veith, H.;
Rogers, N.; Yoder, S. C.; Biddlecome, G.; Tagari, P.; Allen, J. R.;
Hungate, R. W. Discovery of Novel Hydroxy-thiazoles as HIF-R
Prolyl Hydroxylase Inhibitors: SAR, Synthesis, and Modeling
Evaluation. Bioorg. Med. Chem. Lett. 2008, 18, 3925–3928.
ꢀ
(35) Lienard, B. M.; Conejo-Garcıa, A.; Stolze, I.; Loenarz, C.;
´
Oldham, N. J.; Ratcliffe, P. J.; Schofield, C. J. Evaluation of
Aspirin Metabolites as Inhibitors of Hypoxia-Inducible Factor
Hydroxylases. Chem. Commun. 2008, 47, 6393–6395.
(36) Dao, J. H.; Kurzeja, R. J. M.; Morachis, J. M.; Veith, H.; Lewis, J.;
Yu, V.; Tegley, C. M.; Tagari, P. Kinetic Characterization and
Identification of a Novel Inhibitor of Hypoxia-Inducible Factor
Prolyl Hydroxylase 2 Using a Time-Resolved Fluorescence Reso-
nance Energy Transfer-Based Assay Technology. Anal. Biochem.
2009, 384, 213–223.
(37) McNeill, L. A.; Flashman, E.; Buck, M. R. G.; Hewitson, K. S.;
Clifton, I. J.; Jeschke, G.; Claridge, T. D. W.; Ehrismann, D.;
Oldham, N. J.; Schofield, C. J. Hypoxia-Inducible Factor Prolyl
Hydroxylase 2 has a high affinity for Ferrous Iron and 2-Oxyglu-
tarate. Mol. BioSyst. 2005, 1, 321–324.
ꢀ
ꢀ
(38) Mecinovic, J.; Chowdhury, R.; Lienard, B. M. R.; Flashman, E.;
Buck, M. R. G.; Oldham, N. J.; Schofield, C. J. ESI-MS Studies on
Prolyl Hydroxylase Domain 2 Reveal a New Metal Binding Site.
ChemMedChem 2008, 3, 569–572.
ꢀ
(39) Chowdhury, R.; McDonough, M. A.; Mecinovic, J.; Loenarz, C.;
Flashman, E.; Hewitson, K. S.; Domene, C.; Schofield, C. J.
Structural Basis for Binding of Hypoxia-Inducible Factor to the
Oxygen-Sensing Prolyl Hydroxylases. Structure 2009, 17, 981–989.
(40) Niccolai, N.; Tiezzi, E.; Valensin, G. Manganese(II) as Magnetic
Relaxation Probe in the Study of Biomechanisms and of Bio-
macromolecules. Chem. Rev. 1982, 82, 359–384.
(56) Becattini, B.; Pellecchia, M. SAR by ILOEs: An NMR-Based
Approach to Reverse Chemical Genetics. Chem.;Eur. J. 2006,
12, 2658–2662.
(57) Wu, P. S. C.; Otting, G. Rapid Pulse Length Determination in
High-Resolution NMR. J. Magn. Reson. 2005, 176, 115–119.
(58) Mao, X; Guo, J.; Ye, C. Competition Between Radiation Damping
and Transverse Relaxation Effects on NMR Signal Intensities.
Chem. Phys. Lett. 1994, 218, 249–253.
(41) Carr, H. Y.; Purcell, E. M. Effects of Diffusion on Free Precession
in Nuclear Magnetic Resonance Experiments. Phys. Rev. 1954, 94,
630–638.
(59) Mao, X.; Guo, J.; Ye, C. Radiation Damping Effects on Spin-
Lattice Relaxation Time Measurements. Chem. Phys. Lett. 1994,
222, 417–421.
ꢀ
(42) Hewitson, K .S.; Lienard, B. M. R.; McDonough, M. A.; Clifton,
I. J.; Butler, D.; Soares, A. S.; Oldham, N. J.; McNeill, L. A.;