Electrostatic fields near the active site of human aldose reductase: 2. New inhibitors and complications caused by hydrogen bonds

Article abstract of DOI:10.1021/bi200930f

Vibrational Stark effect spectroscopy was used to measure electrostatic fields in the hydrophobic region of the active site of human aldose reductase (hALR2). A new nitrile-containing inhibitor was designed and synthesized, and the X-ray structure of its complex, along with cofactor NADP⁺, with wild-type hALR2 was determined at 1.3 A resolution. The nitrile is found to be in the proximity of T113, consistent with a hydrogen bond interaction. Two vibrational absorption peaks were observed at room temperature in the nitrile region when the inhibitor binds to wild-type hALR2, indicating that the nitrile probe experiences two different microenvironments, and these could be empirically separated into a hydrogen-bonded and non-hydrogen-bonded population by comparison with the T113A mutant, in which a hydrogen bond to the nitrile is not present. Classical molecular dynamics simulations based on the structure predict a double-peak distribution in protein electric fields projected along the nitrile probe. The interpretation of these two peaks as a hydrogen bond formation-dissociation process between the probe nitrile group and a nearby amino acid side chain is used to explain the observation of two IR bands, and the simulations were used to investigate the molecular details of this conformational change. Hydrogen bonding complicates the simplest analysis of vibrational frequency shifts as being due solely to electrostatic interactions through the vibrational Stark effect, and the consequences of this complication are discussed.

Full text of DOI:10.1021/bi200930f

Article
pubs.acs.org/biochemistry
Electrostatic Fields near the Active Site of Human Aldose Reductase:
2. New Inhibitors and Complications Caused by Hydrogen Bonds
Lin Xu,† Aina E. Cohen,‡ and Steven G. Boxer*,†
†Department of Chemistry, Stanford University, Stanford, California 94305-5080, United States
‡Stanford Synchrotron Radiation Laboratory, 2575 Sand Hill Road, Menlo Park, California 94025, United States
S
* Supporting Information
ABSTRACT: Vibrational Stark effect spectroscopy was used to measure electrostatic fields
in the hydrophobic region of the active site of human aldose reductase (hALR2). A new
nitrile-containing inhibitor was designed and synthesized, and the X-ray structure of its
complex, along with cofactor NADP+, with wild-type hALR2 was determined at 1.3 Å
resolution. The nitrile is found to be in the proximity of T113, consistent with a hydrogen
bond interaction. Two vibrational absorption peaks were observed at room temperature in
the nitrile region when the inhibitor binds to wild-type hALR2, indicating that the nitrile
probe experiences two different microenvironments, and these could be empirically
separated into a hydrogen-bonded and non-hydrogen-bonded population by comparison
with the T113A mutant, in which a hydrogen bond to the nitrile is not present. Classical
molecular dynamics simulations based on the structure predict a double-peak distribution in
protein electric fields projected along the nitrile probe. The interpretation of these two peaks
as a hydrogen bond formation−dissociation process between the probe nitrile group and a
nearby amino acid side chain is used to explain the observation of two IR bands, and the simulations were used to investigate the
molecular details of this conformational change. Hydrogen bonding complicates the simplest analysis of vibrational frequency
shifts as being due solely to electrostatic interactions through the vibrational Stark effect, and the consequences of this
complication are discussed.
he organization of charged, polar, and polarizable residues
The quantitative basis of this approach derives from the
Tin the tertiary structure of a protein creates large internal relationship between changes in protein electric fields,
electrostatic fields, which are expected to affect many aspects of
(MV/cm) and observed molecular vibrational
protein function.1−3 Vibrational Stark effect (VSE) spectrosco-
frequency shifts, Δvobs (cm−1):
̅
py has been developed in recent years as a strategy to measure
the magnitude and direction of protein electrostatic fields.4−9
(1)
This method has been applied in several systems, including
where |Δ | (cm−1/(MV/cm) is the Stark tuning rate, a measure
of the sensitivity of the transition frequency to an electric field,
h is Planck’s constant, and c is the speed of light. We have
shown that nitriles have a desirable combination of extinction
coefficient in a relatively uncluttered region of the IR spectrum
and a relatively large Stark tuning rate.21 Nitriles can be
incorporated on unnatural or modified amino acids or
inhibitors.5,6,22 The dot product in eq 1 can be reduced to a
measure of the projection of the protein electric field along the
myoglobin,4 human aldose reductase (hALR2),5,6 Pseudomonas
putida ketosteroid isomerase (KSI),7,8 and ribonuclease S
(RNase S).9 hALR2 is a particularly attractive target for such
studies as a large number of very high-resolution structures are
available.10−12 hALR2 is a 36 kDa aldo-keto reductase that
catalyzes the conversion of aldehydes derived from carbohy-
drate metabolism to corresponding alcohols using NADPH as
the reducing reagent.13,14 Because its involvement in the polyol
pathway is believed to be responsible for many long-term
diabetic complications, researchers have been working for
decades on developing tight-binding inhibitors of hALR2 for
potential pharmacological treatment of diabetes.10−20 Among a
variety of hALR2 inhibitors, there is a class of carboxylate-type
inhibitors developed at the Institute for Diabetes Discovery
(IDD), whose molecular scaffold allows for a variety of
chemical modifications while still maintaining tight binding
affinity.19 This provides us the starting point for designing
novel IDD-type inhibitors with Stark probes such as nitriles5,6
to study the electrostatic environment in hALR2.
−CN bond vector, Δ
, because Δ
has been shown
CN
to be parallel to the transition dipole moment, which in turn is
parallel to the −CN bond.23 The magnitude of the Stark tuning
rate, |Δ CN|, is obtained by measuring the vibrational Stark
spectrum in an external electric field, either for closely related
compounds in organic glasses or in the protein,9 and the
Received: June 17, 2011
Revised: August 5, 2011
Published: August 22, 2011
© 2011 American Chemical Society
8311
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
direction of the −CN bond is obtained by X-ray crystallog-
the binding affinity of hALR2 by 200-fold.19,24 This
modification also makes the orientation of the probe more
unambiguous in the active site by preventing phenyl ring
rotation. As described in the following, the structure of the
tertiary complex of wild-type hALR2-2-NADP+ was determined
by X-ray crystallography at 1.3 Å resolution. This structure
shows that the nitrile probe is buried deep in the hydrophobic
region of the binding pocket and well protected from water. It
also shows that the probe forms a short contact with the side
chain of residue T113. Interestingly, we found two IR peaks in
the nitrile region when 2 bound to WT hALR2, and IR spectra
of two hALR2 mutants helped in the assignment of their origin.
Classical molecular dynamics (MD) simulations were per-
formed to provide a molecular-level understanding of observed
IR features. The complications due to hydrogen bonds for
using VSE spectroscopy to study protein electrostatic fields are
discussed.
raphy. Thus, observed frequency shifts (ΔvCobNs) can be
̅
translated into values for the projection of changes in the
protein electric field along the −CN bond axis, Δ
, and
this can be compared with simulations that are widely used to
estimate such fields.
In part 1 of this series, a nitrile probe was delivered to the
hydrophobic specificity pocket of hALR2 through the binding
of an IDD-type inhibitor, 1, (5-chloro-2-{[(4-cyanobenzyl)-
amino]carbonyl}phenoxy)acetic acid, that we prepared (Figure
MATERIALS AND METHODS
■
Inhibitor Synthesis. The nitrile-containing inhibitor 2, (5-
chloro-2-{[(4-cyano-3-nitrobenzyl)amino]carbonyl}phenoxy)-
acetic acid, was synthesized according to the procedure
described in Scheme 1. All starting materials were purchased
from Aldrich and used without further purification. All
reactions were conducted under nitrogen protection. Products
from each step were purified by either flash column
chromatography (FCC) using silica gel (Aldrich, 230−400
mesh, 60 Å) or high-performance liquid chromatography
(HPLC) (Shimadzu). Proton nuclear magnetic resonance
spectra (1H NMR) were recorded on a 300 MHz Varian
Gemini instrument and mass spectra on a ZQ single-
quadrupole electron spray mass spectrometry (MS) instrument.
4-Cyano-3-nitrobenzyl Bromide. N-Bromosuccinimide
(NBS, 2.4 g) and azobisisobutyronitrile (AIBN, 180 mg)
were added to a 16 mL solution of 2 g of 4-methyl-2-
nitrobenzonitrile in anhydrous carbon tetrachloride (CCl4).
The reaction mixture was heated at 80 °C overnight. The
solution was filtered to remove solids and put under vacuum to
remove excess CCl4. The residue was dissolved in ethyl acetate
(EtOAc) and applied to FCC equilibrated in hexane. The
product was eluted with 60% (v/v) EtOAc in hexane. The
product was identified by the appearance of new peaks on
NMR spectra: 4-methyl-2-nitrobenzonitrile (8.3 ppm, s, 1 H;
8.1 ppm, d, 1 H; 7.8 ppm, d, 1 H) versus 4-cyano-3-nitrobenzyl
bromide (8.5 ppm, s, 1 H; 8.2 ppm, d, 1 H; 8.1 ppm, d, 1 H; 4.9
ppm, s, 2 H).
4-Cyano-3-nitrobenzyl Azide. The bromide product was
dissolved in 60 mL of anhydrous dimethylformamide (DMF);
500 mg of sodium azide was added in small portions over 10
min. The reaction mixture was stirred at room temperature for
24 h. Sodium bromide was removed by filtration through
Celite; the filtrate was extracted with EtOAc and water, and the
organic layer was dried over MgSO4. The solids of MgSO4 were
removed by filtration, and excess solvent was removed under
vacuum. The product was identified by the appearance of new
peaks on the NMR spectrum of 4-cyano-3-nitrobenzyl azide
(8.4 ppm, s, 1 H; 8.2 ppm, d, 1 H; 8.0 ppm, d, 1 H; 4.8 ppm, s,
2 H).
Figure 1. Structural model of X-ray data for the hALR2−2−NADP+
complex. The cofactor NADP+ is colored red, inhibitor 2 yellow, and
the nitrile probe green. The structure of inhibitor 16 is shown at the
top left.
1).6 Changes in electrostatic fields due to mutations in the
vicinity of the nitrile based on a modeled structure were
quantified using nitrile stretching frequency shifts and
compared with calculation results. According to the modeled
structure of the hALR2−1−NADP+ complex, a hydrogen bond
donor from an amino acid side chain in the vicinity of the nitrile
probe might complicate the interpretation of observed
frequency shifts as perturbations in electric fields in the sense
described by eq 1, where Δ
is meant to describe the
electric field change due to a change in the organized
environment, not specific, local chemical interactions such as
hydrogen bonds, which may have an electrostatic component,
but may also depend on overlap of the wave functions on the
donor and acceptor, a contribution that is not included in eq
1.8,9
In this work, we have modified the molecular structure of 1
by introducing the nitro functionality at the meta position of
the amine moiety [inhibitor 2, (5-chloro-2-{[(4-cyano-3-
nitrobenzyl)amino]carbonyl}phenoxy)acetic acid (Figure 1)].
This molecule was initially designed for the purpose of
increasing inhibitor binding affinity because it has been
shown that adding a nitro group in the meta position increases
4-Cyano-3-nitrobenzylamine. The azide product was dis-
solved in a 12 mL mixture of tetrahydrofuran (THF) and water
(10:1, v/v). Triphenylphosphine (1.8 g) was added in small
portions over 10 min, and the reaction mixture was stirred at
room temperature overnight. Excess solvent was removed
8312
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Scheme 1. Synthetic Pathway of 2, (5-Chloro-2-{[(4-cyano-3-nitrobenzyl)amino]carbonyl}phenoxy)acetic Acid
under vacuum, and the residue was dissolved in dichloro-
methane and extracted with a 0.2 M HCl aqueous solution. The
aqueous phase was combined and cooled to 0 °C; 1.0 M NaOH
was added until the pH of the solution reached 10, and the
product was extracted with dichloromethane, washed with
saturated NaCl (aqueous), and dried over MgSO4. The solids
were removed by filtration. Excess solvent was removed under
vacuum. The product was identified by NMR (8.4 ppm, s, 1 H;
8.1 ppm, d, 1 H; 7.9 ppm, d, 1 H; 3.9 ppm, s, 2 H; 2.1 ppm, s, 2
H) and MS [m/z 177.9 (ES+), 178.2 g/mol (calculated)].
4-Chloro-N-(4-cyano-3-nitrobenzyl)-2-hydroxylbenza-
mide. The amine product was dissolved in a 6 mL solution of
1.0 g of 4-chloro-2-hydroxybenzoic acid and 100 mg of
hydroxybenzotriazole (HOBt) in anhydrous THF. A 2 mL
solution of 1.1 g of N,N′-dicyclohexylcarbodiimide (DCC) in
THF was added dropwise over several minutes, and the
reaction mixture was stirred at room temperature overnight.
Solids were removed by filtration. The residue was subjected to
reverse phase HPLC (30 to 60% acetonitrile in water, 5 mL/
min, 30 min), and excess solvent was removed by
lyophilization.
Ethyl (5-Chloro-2-{[(4-cyano-3-nitrobenzyl)amino]-
carbonyl}phenoxy)acetate. The amide product was dissolved
in an 8 mL solution of 300 mg of K2CO3 in anhydrous acetone,
and 200 μL of ethyl bromoacetate was added. The solution was
heated at 56 °C overnight; 1.0 M HCl was added until the pH
of the solution reached 1, and the mixture was extracted with
EtOAc. The organic phase was combined, washed with
saturated NaCl, and dried over MgSO4. The solids of MgSO4
were removed by filtration. Excess solvent was removed under
vacuum. The residue was subjected to reverse phase HPLC (30
to 60% acetonitrile in water, 5 mL/min, 30 min). Excess solvent
was removed by lyophilization.
was removed by lyophilization. The final product was a white
powder and was characterized by NMR (9.3 ppm, t, 1 H; 8.4
ppm, s, 1 H; 8.1 ppm, d, 1 H; 7.9 ppm, d, 1 H; 7.8 ppm, d, 1 H;
7.3 ppm, s, 1 H; 7.2 ppm, d, 1 H; 4.9 ppm, s, 2 H; 4.7 ppm, d, 2
H) and MS [m/z 390.0 (ES+), 389.7 g/mol (calculated)].
Protein Expression and Purification. Expression and
purification were conducted as described previously.6 The gene
for WT hALR2 was obtained from A. Podjarny in the pET-15b
expression vector (Novagen). Mutations were performed using
the Quikchange mutagenesis kits (Stratagene) with polymerase
chain reaction primers purchased from Elim Biopharmaceut-
icals, Inc. All plasmids were transformed into Escherichia coli
strain BL21(DE3) (Novagen). Cells were incubated at 37 °C
for 3 h. Expression was then induced by addition of IPTG (1
mM), followed by growth for an additional 4 h. The cells were
then pelleted, flash-frozen, and stored at −20 °C. To purify the
protein, the cell lysate was resuspended and sonicated. After the
cell debris had been removed, the lysate was applied to a Ni-
NTA agarose column (Qiagen), and His-tagged protein was
eluted using an imidazole gradient. The His tag was cut off
using thrombin from bovine plasma (Aldrich). The cleaved
product was loaded onto a 5 mL Hitrap Q HP anion exchange
column (GE Healthcare) and purified by fast protein liquid
chromatography (FPLC) with gradient elution (0 to 120 mM
NaCl, 5 mL/min, 12 min). The purity of the final cleaved
product was confirmed by sodium dodecyl sulfate−polyacryla-
mide gel electrophoresis and mass spectrometry.
Kinetic Measurements of hALR2. Kinetic experiments
were performed using DL-glyceraldehyde as a substrate and
NADPH as a cofactor in 20 mM Hepes buffer (pH 7) at room
temperature as previously described.6 Reaction rates were
measured by following the decrease in NADPH adsorption at
340 nm over 5 min. The data were fit to the Michaelis−Menten
equation to determine kcat. Inhibition curves were determined
at a saturating substrate concentration. The slope of the
decrease in NADPH absorption was divided by blank and fit to
eq 2 to yield the apparent inhibition constant
(5-Chloro-2-{[(4-cyano-3-nitrobenzyl)amino]carbonyl}-
phenoxy)acetic Acid. The product was dissolved in 7.5 mL of
ethanol and treated with 1.2 mL of 2.0 M NaOH. The solution
was stirred at room temperature for 4 h; 1.0 M HCl was added
until the pH of the solution reached 7. Most solvent was
removed under vacuum. The concentrated solution of product
is adjusted to pH 1, and a yellow precipitate appeared, which
was dissolved and extracted with EtOAc. The organic phase was
combined and washed with saturated NaCl and dried over
MgSO4. The solids of MgSO4 were removed by filtration.
Excess solvent was removed under vacuum. The residue was
further subjected to reverse phase HPLC (40 to 50%
acetonitrile in water, 10 mL/min, 60 min). Excess solvent
(2)
where E is the enzyme concentration, I is the inhibitor
concentration, and KI is the apparent inhibition constant.25,26
All data were averaged from two parallel experiments.
X-ray Crystallography. The hanging drop method was
used to obtain crystals (EasyXtal Tool 24 culture plates,
8313
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Table 1. Measured Kinetic Constants, Nitrile Vibrational Frequencies, and Full Widths at Half-Maximum (fwhm) of Inhibitor 2
in the Specificity Pocket of WT and Mutant hALR2 at Room Temperature
sample
kcat (s−1
)
KI (nM)
frequency (cm−1
)
fwhm (cm−1
)
buffer
T113A
WT
−
−
2242.2 ± 0.1
2237.8 ± 0.1
10.6 ± 0.1
5.8 ± 0.1
0.20 ± 0.01
0.33 ± 0.01
0.19 ± 0.01
30 ± 1
107 ± 7
32 ± 1
2238.2 ± 0.1
2238.8 ± 0.1
2252.3 ± 0.1
2247.4 ± 0.3
11.4 ± 0.4
6.4 ± 0.1
9.5 ± 0.2
8.2 ± 1.1
T113S
Qiagen). WT hALR2 was cocrystallized with the oxidized form
of coenzyme NADP+ (Sigma) and inhibitor 2 at room
temperature (1:2:2 protein:inhibitor:coenzyme ratio). Hanging
drops were made by the mixing mother liquor [0.06 M citric
acid, 0.04 M Bis-tris propane, and 12% (w/v) PEG 3350 (pH
4.2)] with a holoenzyme solution (10 mg/mL, 20 mM Hepes,
pH 7.0). Crystals were observed after equilibration for 3 days.
Cryofreezing was conducted via quick transfer into a
stabilization solution (25% PEG 3350), then into a cryo-
protecting solution (55% PEG 3350), and finally into liquid
nitrogen. Experiments were conducted at X-ray beamline 9-2 at
the Stanford Synchrotron Radiation Light source (SSRL). Data
sets were processed with HKL2000 and PHENIX. Data
collection and refinement statistics are provided as Supporting
Information (Table S1). The Protein Data Bank entry is 3T42.
IR Measurements of hALR2. IR spectra were recorded
with a Bruker Vertex 70 FT-IR spectrometer at both room
temperature and low temperatures. Samples were illuminated
through a notch filter that transmitted 2000−2500 cm−1 light,
and spectra were recorded with a liquid nitrogen-cooled indium
antimonide detector at 1 cm−1 resolution. For room-temper-
ature measurements, the sample cell was made of sapphire
windows separated by 75 and 100 μm Teflon spacers. Solutions
of holoenzyme (1:1:2 protein:2:NADP+ ratio) were equili-
brated for 1 h at 4 °C, flushed with an excess of 20 mM Hepes
buffer (pH 7), and concentrated to a final concentration of 2
mM. Variable-temperature IR spectra were recorded using a
temperature-controlled water-jacketed cell from Lake Shore
Cryotronics, Inc. Ethylene glycol was added to the water bath
to push temperatures below 0 °C. The samples were prepared
by dissolving lyophilized powder of a 4 mM holoenzyme
solution in an equal volume of a 40% (w/w) trehalose/water
mixture. The use of trehalose ensures a fluid state below 0 °C,
and this process also forms an optical-quality glass at cryogenic
temperatures. Equilibration for 5 min at each temperature was
allowed before data were collected. To obtain a spectrum at 77
K, the sample cell was made of sapphire windows separated by
30 and 60 μm Teflon spacers. The cell was filled with a solution
of 4 mM WT hALR2 holoenzyme or a mixture of 4 mM WT
hALR2 and 4 mM T113A holoenzyme in 40% trehalose buffer
and immediately immersed in liquid nitrogen in a custom
cryostat.6 All spectra were averaged on three parallel measure-
ments and fit using one or two Gaussian functions with Matlab.
Stark Spectrum of Inhibitor 2. The sample cell was made
of sapphire windows coated with 4.5 nm of Ni separated by a
pair of 30 μm spacers and attached by copper wires to a high-
voltage DC power supply (Trek Instruments Inc.).6 The cell
was filled with a solution of 50 mM inhibitor 2 in 2-
methyltetrahydrofuran (MeTHF) and immediately immersed
in liquid nitrogen in a custom liquid nitrogen cryostat. Applied
fields of 0.7−1.6 MV/cm were synchronized with a home-built
control unit; 128 scans with the field applied were alternated
with 128 scans with no applied field, and the field-on-minus-
field-off difference spectrum was obtained. The spectra were fit
as described previously.5,6,21 The Stark tuning rate of the nitrile
probe on 2 reported is the average of three separate
experiments.
Molecular Dynamics Simulations. The crystallographic
structure of the hALR2−2−NADP+ complex was used in MD
simulations. Amino acid mutations were created using Pymol.27
The calculation strategy used here follows closely the approach
developed in our lab for the ribonuclease S system containing
nitrile probes.9 Molecular dynamics simulations were con-
ducted in Gromacs 3.3.1 using the AMBER-'99 force field.28
Inhibitor 2 and the cofactor were parametrized using the
Antechamber and Leap package with GAFF atom type and
AM1BCC charges.29 More details about the parametrization of
2 can be found in the Supporting Information (Figure S1 and
Table S3). The protein was solvated with explicit SPC water
molecules.30 Nonbonded cutoffs of 10 Å were used.
Simulations were conducted with periodic boundary con-
ditions, and the calculation of long-range electrostatics was
treated with the particle mesh Ewald model. Simulations were
equilibrated (20 ps energy minimization, followed by 20 ps of
heavy-atom position restrained refinement) and run using the
Nose-Hoover thermostat and the Parinello-Raman barostat.
Replica exchange molecular dynamics (REMD, 4 ns) were
performed with 24 replicas from 298 to 339.45 K, ordered in
temperature with a spacing of 1.75 K per replica. Electric fields
at the midpoint of the nitrile bond were calculated at every 2 fs
step for all MD trajectories. Atomic positions were saved at
every 2 ps step only for the 298 K trajectories. Therefore, in the
4 ns 298 K trajectories, 2 × 106 electric field data were recorded
and 2 × 103 structures were saved. Structures were taken every
60 ps from the 298 K trajectories of the REMD simulation over
the last 3 ns, generating 50 unique starting structures to run 50
independent parallel 1 ns simulations (50 × 1 ns). In the 50 ×
1 ns simulation, 25 × 106 electric field data were recorded;
structures were not saved because of limited hard drive space.
Electric field histograms were plotted using the 50 × 1 ns
simulation data set (the vertical axis represents the percentage
of electric field counts within a bin of 1 MV/cm, based on all 25
× 106 electric field data) and fit using one or two Gaussian
functions. Correlation plots between electric field data and
geometrical parameters were based on 4 ns 298 K trajectories
from the REMD simulations, with 2 × 103 data points, each
representing one electric field value and the corresponding
structural data at that time point.
RESULTS AND DISCUSSION
■
Catalysis and Structure. All proteins studied in this work
displayed normal catalytic activity and strong binding affinity
toward inhibitor 2 (Table 1, columns 2 and 3). For example,
WT hALR2 showed a kcat of 0.33 s−1, close to the result
reported previously.6,14 2 bound to WT hALR2 with an
apparent inhibition constant (KI) of 107 nM, compared to a
value of 2 μM for binding of inhibitor 1 to WT hALR2;6 the
binding affinity has been increased by approximately 20-fold.
8314
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
basis of the short contact observed in the crystallographic
structure with the hydroxyl group of T113, it seems likely that
the higher-energy, more intense band is due to the hydrogen-
bonded form. To test this idea, two single-point mutations were
examined, also shown in Figure 3B. T113A, in which the side
chain hydroxyl group is removed, displays a single symmetric
peak whose frequency is close to the lower-frequency peak of
WT. T113S, which also has a side chain hydroxyl group,
displays a major peak close to mutant T113A, with a small but
reproducible shoulder on the high-energy side. These results
are consistent with a well-documented result for simple nitriles
in hydrogen-bonding and non-hydrogen-bonding solvents: the
nitrile frequency typically shifts to higher wavenumbers upon
accepting hydrogen bonds from hydroxylic solvent mole-
cules.32−45 Qualitatively, Purcell and Drago32 have argued that
when nitriles form a complex with Lewis acids (including a
hydrogen bond), the electron lone pair is alleviated from the
nitrogen nonbonding orbital that overlaps with nitrile bonding
orbitals, resulting in an increase in the nitrile force constant and
a blue shift in the nitrile stretch frequency.
To improve our understanding of the origin of the two peaks
of WT hALR2, we obtained IR spectra as a function of
temperature (Figure 3C and Table S2 of the Supporting
Information). The high-frequency (hydrogen-bonded) peak
shifts to higher frequencies and narrows with a decrease in
temperature. The blue shift of the hydrogen-bonded peak could
be explained by the strengthening of hydrogen bonds at low
temperatures,9 with an accompanying decrease in bandwidth
due to an increasingly more restrained system. The low-
frequency (non-hydrogen-bonded) peak becomes broader and
more asymmetric as the temperature is lowered, possibly
suggesting that there may be more than one non-hydrogen-
bonded configuration (the center frequency of the non-
hydrogen-bonded peak first shifts slightly to the blue and
then to the red). The area ratio of the high-frequency peak to
the low-frequency peak increases as the temperature is lowered
(the total integrated area increases by ∼1.3-fold), and at 77 K,
essentially only a single high-frequency peak is observed. This
suggests that the hydrogen-bonded form is more favorable at
low temperatures. We also measured the IR spectrum for
mutant T113A at the same sample concentration at 77 K: a
single peak, which is 17 cm−1 lower in frequency than the WT
peak, is observed. This suggests that in the WT spectrum, the
high-frequency peak at 77 K should be attributed to the
hydrogen-bonded form. The ratio of the area of the nitrile peak
of WT to that of T113A at 77 K is 1.7 ± 0.1. It is reasonable to
assume that most of the population is in the hydrogen-bonded
form at 77 K for 2 bound to WT hALR2, so we can obtain an
approximate value of the ratio of the integrated peak area for
hydrogen-bonded and non-hydrogen-bonded nitrile species at
77 K. The result is consistent with previous reports of simple
model compounds: the integrated intensities for the hydrogen-
bonded forms are higher than those of the non-hydrogen-
bonded forms. For example, the −CN stretch for CH3CN has
twice the integrated intensity in H2O that it does in the pure
solvent,33 and for HCN, the CN stretching intensity increases
by a factor of 1.4 upon formation of a complex with HF.34
Assuming that the ratio of intensities does not change
drastically with temperature, the relative populations of the
hydrogen-bonded and non-hydrogen-bonded forms of 2 in WT
hALR2 at room temperature can be estimated from IR
absorption. The ratio of the area of the two peaks at room
temperature is 1.9 ± 0.3 [the value used here was obtained in
Figure 2. Electron density map for the inhibitor binding pocket (2Fobs
− Fcalc contoured at 1σ in blue mesh), identifying relevant residues
involved in inhibitor binding.
The fact that 2 binds strongly to WT and mutant hALR2
ensures the feasibility of using this inhibitor to carry the probe
into the protein active site for IR experiments and for X-ray
crystallography. [We attempted to cocrystallize WT hALR2
with inhibitor 1 (Figure 1). Only crystals of the hALR2−
NADP+ complex were obtained, likely reflecting the much
poorer binding of 1 than 2 at the active site. Likewise, soaking
native crystals of the hALR2−NADP+ complex in the presence
of 1 failed.]
The hALR2 holoenzyme folds as an eight-stranded α/β-
barrel structure (Figure 1). The active site is a well-resolved
region located at the C-terminal loop (Figure 2), and 2 binds
similarly to previously studied hALR2−IDD inhibitor−NADP+
complexes.11,12 The carboxylate moiety of 2 is tightly anchored
via hydrogen bond interactions with Y48, H110, and W111, as
well as electrostatic interaction with cofactor NADP+. There is
also hydrophobic π-stacking interaction between the amine
moiety of 2 and the W111 side chain. The nitro functionality
forms a hydrogen bond with the L300 backbone amide, which
accounts for the enhanced binding affinity of 2, consistent with
earlier work on nitro-substituted IDD inhibitors.24 The nitrile
probe is buried in the hydrophobic interior and well protected
from close-range interaction with solvent molecules. A short
contact exists between the nitrile group and the T113 side
chain (N···O distance of 2.84 Å), which is a common binding
mode when an inhibitor opens the specificity pocket of hALR2
(e.g., similar effects are seen for IDD594, zopolrestat and
IDD38811,18,31). For residue T113, no sign of multiple-side
chain occupancy was observed in the electron density map.
IR Spectroscopy. The vibrational Stark spectrum of
inhibitor 2 is shown in Figure 3A. The magnitude of the
Stark tuning rate of the nitrile probe on 2 is found to be 0.53
cm−1/(MV/cm), which is somewhat smaller than what has
been observed previously for nitrile groups on phenyl rings
(e.g., inhibitor 1 has a Stark tuning rate of 0.77 cm−1/(MV/
cm).5,6 This is likely due to the electron withdrawing property
of the adjacent nitro group. Detailed analysis of the Stark
spectrum can be found in the Supporting Information.
FTIR spectra of 2 bound to WT and mutant hALR2 were
recorded to obtain information about electrostatic fields and/or
hydrogen bond interactions in the vicinity of the nitrile probe.
As seen in Figure 3B and Table 1 (columns 4 and 5), when 2 is
bound to WT hALR2, two peaks separated by ∼14 cm−1 were
observed in the nitrile region at room temperature. On the
8315
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Figure 3. (A) Vibrational Stark spectrum of inhibitor 2 dissolved in MeTHF at 77 K. The black curve is the difference spectrum (field on minus field
off) scaled to sample an absorbance of 1 and an applied field strength of 1.0 MV/cm. The purple, blue, and green curves represent contributions
from the zeroth, first, and second derivatives, respectively, of the vibrational absorption spectrum of 2. The red curve is the overall fit, and the
second-derivative contribution gives a |Δ CN| of 0.53 cm−1/(MV/cm). More detailed analysis of the Stark spectrum can be found in the Supporting
Information. (B) Vibrational absorption spectra in the nitrile region of 2 in the active sites of WT and two mutants of hALR2 at room temperature.
(C) Temperature dependence of IR absorption spectra of 2 bound to WT hALR2 in 40% trehalose buffer and compared with the low-temperature
absorbance of 2 bound to the T113A mutant. (D) van’t Hoff plot derived from variable-temperature IR spectra.
20 mM Hepes buffer; the use of trehalose changed the area
ratio at 25 °C slightly (see Table S2 of the Supporting
Information)], from which we estimate a ratio of the two
species of 1.1, which can be used later for comparisons with
room-temperature MD simulations. We can calculate the
equilibrium constant, Keq, of the equilibrium
MD Simulations and Electrostatics of the Wild Type.
To further investigate the molecular origins of these spectra, we
have performed classical MD simulations. Classical MD applies
Newton’s laws of motion to a group of interacting atoms using
a variety of force fields. In hydrogen bond interactions, both
electrostatic interactions (dominant at long range) and
exchange repulsion (short range) are important components.
At the outset, we attempt to employ a very simple model to
calculate IR absorption spectral shifts using the Stark tuning
rate as the bridge between calculated changes in electric fields
and IR shifts: we extract information about electric fields from
the MD output and from this obtain information about the
structures, dynamics, and populations that relate to room-
temperature IR results. In the particular case of hydrogen bonds
to nitriles, quantum effects are generally considered to be more
significant than electrostatics.32−45 As mentioned earlier,
Purcell and Drago explained the observed blue shifts of nitrile
frequency upon formation of hydrogen bonds by loss of an
electron lone pair from the nonbonding orbital of the nitrile
triple bond on the basis of molecular orbital calculations.32
Reimers and Hall noted that when there is strong and
directional solvent−solute interaction, such as a hydrogen bond
or proton transfer, this interaction will be of primary
importance, and the nitrile frequency becomes less sensitive
to its electrostatic environments.35 Therefore, we note at the
at various temperatures. Fitting the temperature dependence of
Keq to the van’t Hoff equation
(3)
where R is the molar gas constant, ΔHΘ is the standard
enthalpy change, and ΔSΘ is the standard entropy change, we
can extract thermodynamic parameters for the reaction shown
above. A van’t Hoff plot is shown in Figure 3D, giving a ΔHΘ
of −9.4 kJ mol−1 and a ΔSΘ of −34 J mol−1 K−1. These results
will be discussed later. In addition, the equilibrium constant at
77 K can be calculated using ΔHΘ and ΔSΘ to be ∼4 × 104,
consistent with the fact that at cryogenic temperature only a
single high-frequency peak is observed.
8316
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Table 2. Calculated Peak Positions and fwhm Values for the
Electric Field Distribution along the Nitrile Probe in WT
and Mutant hALR2. Measured and Predicted Changes in
Electric Fields (Non-hydrogen-bonded Peaks vs Mutant
a
T113A) and fwhm Values (MV/cm)
a
MD
experiment
position
Δ
fwhm
Δ
fwhm
WT
−15.6
−41.9
−16.1
−7.4
0.5
−
−
23.6
22.1
15.2
20.8
24.5
0.8
21.5
−
10.9
12.1
−
−
−
1.9
T113A
T113S
8.7
−41.5
−
−
Figure 4. Calculated electric field histogram projected on the nitrile
probe in WT and mutant hALR2 bound to 2 extracted from MD
simulations and using eq 1 (cf. Figure 3B).
a
The measured Δ
and fwhm are calculated using the
observed spectra and the Stark tuning rate of 2 in MeTHF (|Δ CN| =
0.53 cm−1/(MV/cm).
outset that modeling the results as entirely electrostatic in
origin will incorrectly assign the spectral shifts of the hydrogen-
bonded and non-hydrogen-bonded populations. However, this
anticipated discrepancy does not imply that the force field used
in the simulations is wrong, because the energies of a nitrile in
MD are not determined solely by electrostatics. Especially, the
Lennard-Jones parameters in our simulations were calibrated by
Corcelli et al. against a DFT calculation of the interaction
between acetonitrile and water in a linear hydrogen bonding
geometry for a more accurate description of the hydrogen bond
interaction between the lone pair of the nitrogen atom and
water.44 The validity of the structural and population
information then largely depends on the transferability of
these parameters into a chemically different system (different
donor, acceptor, and hydrogen bond configurations), which can
be tested by comparing simulated spectral features with
experimental results. In our case, the most direct linkage
between them is the population information. That is, if the
populations that emerge from our analysis are similar to those
observed experimentally, this would suggest that the Lennard-
Jones parameters are valid so that the energetics of a nitrile falls
into a reasonable regime.
Classical MD simulations on the nanosecond time scale were
performed, and electric fields at every MD step were recorded
giving a histogram of protein electric fields projected along the
nitrile probe, as shown in Figure 4 and Table 2 (columns 2 and
4). MD predicts a double-peak field distribution for WT
hALR2, consistent with the two peaks observed in the room-
temperature IR spectrum (Figure 3B). For T113A, MD
predicts a single peak, again consistent with the single peak
observed for T113A. Although MD will not identify the two
peaks in the WT spectrum, we can deduce their identity
through the comparison with T113A: the single peak of T113A
is close to the more positive field peak of WT hALR2;
therefore, this peak is the non-hydrogen-bonded peak, while
the more negative field peak is the hydrogen-bonded peak.
More quantitatively, the MD simulations predict that in WT
hALR2, the population ratio of hydrogen-bonded to non-
hydrogen-bonded forms is 1.5, which is quite close to the
experimental value estimated above to be 1.1. As for T113S,
MD predicts a double-peak field distribution, and the area ratio
of the more positive peak is larger than that of WT hALR2.
With the basic confirmation that the simulations provide
reasonable agreement for the populations and fields, we can
exploit the MD trajectories to further investigate the micro-
scopic interactions between individual atoms that lead to the
field distributions in a given protein, something that is not
possible in the experiments. Specifically, we extracted structural
information from the MD output and used this to establish
correlations between computed electric fields and geometric
parameters. Figure 5A shows the dependence of the electric
field projected on the inhibitor −CN bond on the dihedral
angle representing T113 side chain rotation for WT hALR2.
The relatively constant angle values indicate that the T113 side
chain conformation does not change much during course of the
simulation. As seen in Figure 5B, however, the two-distance
correlation (the distance between the T113 hydroxyl hydrogen
atom and the nitrogen atom of the probe and the distance
between the hydrogen atom and the oxygen atom of the T113
carbonyl group) displays two-state behavior. To explain this, we
propose a model of a hydrogen bond motif involving one donor
and two acceptors. This is based on the observation of a short
contact between the nitrile probe and T113 hydroxyl group (as
seen in the low-temperature X-ray structure), and also on the
fact that threonine [serine also (see below)] is frequently found
to form an intraresidue hydrogen bond.11,46 Dynamically, the
hydroxyl group switches between hydrogen bonds to the nitrile
probe and the amide carbonyl group of T113. Panels C and D
of Figure 5 show the correlation between the computed electric
fields and the angle and distance parameters of the hydrogen
bond between the T113 hydroxyl group and the nitrile probe.
In the field−angle correlation, the two-state behavior is seen
again. All of the electric field data are sorted in Figure 5E by
using geometrical criteria for hydrogen bond formation [angle
of ≤60° and distance of ≤3.5 Å (see the legend and panel F of
Figure 5, which shows snapshots)]. This results in two
distinguished peaks. The more negative field peak corresponds
to hydrogen-bonded nitrile species and the other to non-
hydrogen-bonded species. This is consistent with the
observation that the single peak of T113A on the electric
field histogram is close to the more positive field peak of WT
hALR2. As expected, the electrostatic model we employed
predicts a red shift in the frequency of the nitrile probe upon
hydrogen bond formation, failing to capture the blue shift
observed in IR experiments due to the dominating quantum
effects in nitrile hydrogen bond formation. Finally, the
calculated lifetimes for the two species are both on the
picosecond time scale (9.7 ps for the hydrogen-bonded form
8317
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Figure 5. MD trajectory analysis for WT hALR2. (A) Correlation of the calculated electric field projected on the inhibitor −CN bond with the
dihedral angle formed by the C, CA, CB, and OG1 atoms of T113. Each data point represents one electric field value and the corresponding
structural data at that time point. (B) Correlation of two distances in the MD simulation: the distance between T113 HG1 and INH N2 and the
distance between T113 HG1 and T113 O (see the inset of panel A for labeling). (C) Correlation of the calculated electric field projected on the
inhibitor −CN bond and the angle formed by the HG1 and OG1 atoms of T113 and the N2 atom of INH. (D) Correlation of calculated electric
field projected on the inhibitor −CN bond and distance between T113 OG1 and INH N2. (E) Calculated electric fields projected on the inhibitor
−CN bond sorted according to hydrogen bond geometrical criteria. The red solid curve represents the whole data set (2001 data points from 4 ns
REMD); the black dashed curve represents configurations in which the inhibitor nitrile is hydrogen-bonded while the carbonyl group of T113 is not
hydrogen-bonded (61%, angle of ≤60°, distance of ≤3.5 Å), and the blue dashed curve represents configurations in which the nitrile probe is not
hydrogen-bonded while the T113 carbonyl group is hydrogen-bonded (39%). (F) Detailed representation of the T113−OH···2−CN species (left)
and T113−OH···T113−CO species (right).
and 6.2 ps for the non-hydrogen-bonded form).47 In summary,
by analyzing the MD trajectories for WT hALR2, we propose
that the source of the two peaks in the protein electric field
distribution projected on the probe −CN bond is the
dynamical making and breaking of a hydrogen bond between
the nitrile probe and T113 side chain hydroxyl group, which
viewed from the nitrile IR band gives rise to the two peaks seen
in the IR spectrum, and further that the T113 hydroxyl group is
always hydrogen-bonded, either to the inhibitor nitrile group or
to its amide carbonyl group.
Following this analysis, the thermodynamic parameters
obtained earlier actually correspond, in the context of the
entire protein−inhibitor complex, to the equilibrium between
two types of hydrogen bonds, a hydrogen bond between T113
and its amide carbonyl group and one between T113 and the
inhibitor nitrile:
From other studies, hydrogen bond energies are typically
8318
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
Figure 6. MD trajectory analysis of hALR2 T113S. (A) Correlation of electric fields with the dihedral angle formed by the C, CA, CB, and OG
atoms of S113. (B) Electric fields sorted according to serine side chain conformation. The blue curve represents the full data set (2001 data points).
The green curve (72%) represents the data set for state 1. The pink curve (28%) represents the data set for state 2. (C) Electric fields in state 1
sorted according to the hydrogen bond formation criteria. The green solid curve represents the entire data set (1436 data points); the red dashed
curve represents subconfigurations in which the inhibitor nitrile is hydrogen-bonded while the imidazole nitrogen atom of H306 is not hydrogen-
bonded (23%, angle of ≤60°, distance of ≤3.5 Å), and the blue dashed curve represents subconfigurations in which the nitrile probe is not hydrogen-
bonded while the imidazole nitrogen atom of H306 is hydrogen-bonded (77%). (D) S113−OH···2−CN subspecies in state 1 (left) and S113−
OH···H306−imidazole subspecies (right). (E) Electric fields in state 2 sorted according to the hydrogen bond formation criteria. The peak solid
curve represents the full data set (565 data points). The red dashed curve represents subconfigurations in which the inhibitor nitrile is hydrogen-
bonded while the carbonyl group of S113 is not hydrogen-bonded (63%, angle of ≤60°, distance of ≤3.5 Å). The blue dashed curve represents
subconfigurations in which the nitrile probe is not hydrogen-bonded while the S113 carbonyl group is hydrogen-bonded (37%). (F) S113−OH···2−
CN subspecies in state 2 (left) and S113−OH···S113−CO subspecies (right).
approximately 10−30 kJ mol−1.48−53 The enthalpy change
obtained from the van’t Hoff plot in Figure 3D for the
difference between a hydrogen bond with the nitrile as the
acceptor and one with the amide carbonyl as the acceptor
(ΔHΘ = −9.4 kJ mol−1) is reasonable in terms of the order of
magnitude. The negative sign indicates that the nitrile nitrogen
atom is a better hydrogen bond acceptor than the amide
oxygen atom in this case, although previous studies, e.g., ab
initio calculations50 and Abraham’s empirical hydrogen bond
basicity scale,54 suggested the contrary. This discrepancy is
likely due to an idiosyncratic behavior of this restrained system.
The negative entropy change (ΔSΘ = −34 J mol−1 K−1) implies
that the configuration on the left in Figure 5F is less disordered.
It can be interpreted as forming a stronger hydrogen bond
limits the movement of molecules, leading to a decrease in the
system entropy.
MD Simulations and Electrostatics of T113S. For
T113S, a similar approach was used, and the results suggest
both similarities and differences between this mutant and WT
hALR2. We first examine the dependence of electric fields on
8319
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
the S113 side chain rotation. As seen in Figure 6A, the data set
is composed of two populations, denoted state 1 and state 2.
Compared with threonine in WT hALR2, in which the methyl
group largely blocks side chain rotation (Figure 5A), the serine
side chain is more flexible in T113S, occupying two possible
conformations, as illustrated in panels D (state 1) and F (state
2) of Figure 6. By sorting electric field data according to the
dihedral angle parameter (Figure 6B), we find that the two
populations contribute differently to the electric field
distribution. The prediction of multiple side chain conforma-
tions for S113 is consistent with a recent paper on
crystallographic studies of hALR2 mutants bound with IDD
inhibitors with halo substituents on phenyl para sites.31 The
two side chain conformations of S113 in our system predicted
by MD simulations resemble the two side chain conformations
of S113 detected in structures of T113S−IDD388 or −IDD594
complexes, and the predicted percentages of two configurations
as depicted in panels D and F of Figure 6 are close to the
measured fractional occupancies of the S113 side chain in
crystal structures of T113S−IDD388 or −IDD594 complexes.
(In our system, the predicted percentages of state 1 and state 2
are 72 and 28%, respectively. In the crystal structure of the
T113S−IDD388 complex, the percentages of the two
corresponding configurations are 72 and 28%, respectively,
and in the T113S−IDD594 complex, they are 68% and 32%,
respectively.31) In state 1, there is a hydrogen bond motif
similar to that in WT hALR2. In this case, the imidazole
nitrogen atom in H306 acts as one hydrogen bond acceptor
while the other acceptor is the nitrogen atom on the probe
nitrile, and the hydrogen bond donated by the S113 side chain
hydroxyl group switches between them dynamically. If we use
the same hydrogen bond criteria (angle of ≤60° and distance of
≤3.5 Å) to sort electric field data for state 1 (Figure 6C), the
peak at a more negative electric field corresponds to the S113−
OH···2−CN subspecies; the other corresponds to the S113−
OH···H306−imidazole subspecies. Snapshots for the two
subspecies in state 1 are given in Figure 6D. In the case of
state 2, there is also a similar hydrogen bond motif, in which the
donor is the S113 side chain hydroxyl group, one acceptor is
the nitrogen atom on the probe nitrile, and the other acceptor
is the amide carbonyl oxygen atom of S113. Electric field data
are sorted using the same criteria (angle of ≤60° and distance
of ≤3.5 Å), and two separated peaks are seen again (Figure 6E).
Snapshots for the S113−OH···2−CN species and S113−
OH···S113−CO species in state 2 are given in Figure 6F,
which has a configuration similar to that of WT hALR2 (cf.
Figure 5F). Overall, this could explain the IR spectrum of 2
bound to T113S, a major peak with a high-frequency tail,
attributed to hydrogen bonding, but more complicated than
that of WT: unlike WT hALR2, in T113S, the serine residue
displays two possible side chain conformations, each inter-
preted as a hydrogen bond making and breaking process, and
state 1, the dominant configuration, supports a higher ratio of
non-hydrogen-bonded nitrile species.
expected according to eq 1. As a result, this purely electrostatic
model predicts a frequency shift in the opposite direction to
what is observed when a hydrogen bond is formed. For the
same reason, when there are hydrogen bond interactions
between the nitrile probe and the surrounding environment,
the idea of using the nitrile as a Stark probe to measure
electrostatic fields in proteins solely on the basis of eq 1 is
questionable: environment-induced frequency shifts may no
longer respond linearly to electric fields but are more
dominated by the quantum mechanical nature of such
interactions, and these effects may depend upon details of
the geometry of a particular hydrogen bond.35,45,55 We recently
presented a method to separate effects of hydrogen bonds on
IR frequency shifts from the environmental electrostatic
contribution in the case of −SCN probes by correlating the
13C NMR chemical shift and the IR frequency.8,56 This
approach is relatively simple to apply to −SCN probes as 13C
can be directly introduced during labeling, but it would be more
difficult to apply to a complex inhibitor such as 2 because of the
synthetic effort, though we are currently developing a synthetic
strategy that may allow facile preparation of 13C-labeled
versions of aromatic nitrile-containing inhibitors. One interest-
ing application would be using 13C-labeled inhibitor 1 to
acquire tandem IR and NMR measurements in the systems
studied in part 1 of this series, where an X-ray structure could
not be obtained, so that we can detect whether the nitrile probe
accepts hydrogen bonds from nearby amino acid residues in the
absence of crystallographic structures. If a comparison is made
within a series of mutants, pH changes or inhibitor changes,
and their added electrostatic contribution is the goal, then a
demonstration that a hydrogen bond is retained or lost by the
NMR−IR correlation is very useful for vetting the data. It
would also be interesting to measure |Δ CN| for both hydrogen-
bonded and non-hydrogen-bonded populations to determine
whether the Stark tuning rate is affected by a hydrogen bond. It
has been shown in one case that hydrogen-bonded nitrile
probes have a Stark tuning rate similar to those of non-
hydrogen-bonded ones.56 We attempted to measure the
vibrational Stark effect of inhibitor 2 in situ in WT hALR2,
but the relatively low extinction coefficient and limited
solubility of hALR2 in buffered solution precluded this
measurement.
While the interpretation of frequency shifts of electric fields
with calculated shifts may be compromised when the nitrile is
hydrogen-bonded, in cases where the nitrile is not hydrogen-
bonded, we can compare changes in electric fields (Δ
)
measured by IR given the Stark tuning rate with predictions
made by MD (Table 2, columns 3 and 5). It can be seen that
the calculation strategy performs well for WT hALR2 and
T113S in terms of the direction, but not magnitude. Treating
the line width as being due to inhomogeneous broadening from
the distribution of local electric fields interacting with Δ CN, we
found the simulation reproduces the observed trend for the
three non-hydrogen-bonded peaks (Table 2, columns 4 and 6);
i.e., T113A shows the narrowest fwhm, while WT hALR2
shows the broadest fwhm. Thus, the basic computational
strategy is qualitatively successful, but not quantitatively, likely
because of errors in the force field and/or insufficient sampling
in the MD simulations. There are limited examples in the
literature demonstrating the improvement of force fields in the
pursuit of precise assignment of structural features based on
Concluding Comments about Complications Caused
by Hydrogen Bonds. As discussed earlier, many experiments
reveal that the nitrile stretch frequency shifts to higher
frequencies upon accepting a hydrogen bond. Using a purely
electrostatic model, when the −CN bond interacts with a
hydroxyl group in a configuration as in Figure 5F, the difference
dipole moment of the nitrile bond Δ
(pointing from the
nitrogen atom to carbon atom) is aligned along the electric field
exerted by the hydroxyl dipole, and a red shift in energy is
CN
8320
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
(4) Park, E. S., Andrews, S. S., Hu, R. B., and Boxer, S. G. (1999)
Vibrational Stark Spectroscopy in Proteins: A Probe and Calibration
for Electrostatic Fields. J. Phys. Chem. B 103, 9813−9817.
(5) Suydam, I. T., Snow, C. D., Pande, V. S., and Boxer, S. G. (2006)
Electric Fields at the Active Site of an Enzyme: Direct Comparison of
Experiment with Theory. Science 313, 200−204.
(6) Webb, L. J., and Boxer, S. G. (2008) Electrostatic Fields Near the
Active Site of Human Aldose Reductase: 1. New Inhibitors and
Vibrational Stark Effect Measurements. Biochemistry 47, 1588−1598.
(7) Sigala, P. A., Fafarman, A. T., Bogard, P. E., Boxer, S. G., and
Herschlag, D. (2007) Do Ligand Binding and Solvent Exclusion Alter
the Electrostatic Character within the Oxyanion Hole of an Enzymatic
Active Site. J. Am. Chem. Soc. 129, 12104−12105.
electric field calculations, but several important examples have
been presented using the system of carbon monoxide
photodissociation in myoglobin.57−60 A genuine QM/MM
simulation might be able to better capture the properties of the
hydrogen-bonded populations. This system, which contains a
hydrogen-bonded and non-hydrogen-bonded nitrile in the
same macromolecular system, can serve as a test system for
establishing benchmarks in advanced simulation methodology,
and the independent predictions of the time scale for making
and breaking hydrogen bonds can, in principle, be tested using
two-dimensional IR techniques.61 Such experiments are
currently in progress.
(8) Fafarman, A. T., Sigala, P. A., Herschlag, D., and Boxer, S. G.
(2010) Decomposition of Vibrational Shifts of Nitriles into Electro-
static and Hydrogen-Bonding Effects. J. Am. Chem. Soc. 132, 12811−
12813.
(9) Fafarman, A. T., and Boxer, S. G. (2010) Nitrile Bonds as
Infrared Probes of Electrostatics in Ribonuclease S. J. Phys. Chem. B
114, 13536−13544.
(10) Podjarny, A., Cachau, R. E., Schneider, T., Van Zandt, M. C.,
and Joachimiak, A. (2004) Subatomic and Atomic Crystallographic
Studies of Aldose Reductase: Implications for Inhibitor Binding. Cell.
Mol. Life Sci. 61, 763−773.
(11) Howard, E. I., Sanishvili, R., Cachau, R. E., Mitschler, A.,
Chevrier, B., Barth, P., Lamour, V., Van Zandt, M. C., Sibley, E. O.,
Bon, C., Moras, D., Schneider, T. R., Joachimiak, A., and Podjarny, A.
(2004) Ultrahigh Resolution Drug Design. I: Details of Interactions in
Human Aldose Reductase-Inhibitor Complex at 0.66 Å. Proteins 55,
792−804.
(12) Ruiz, F., Hazemann, I., Darmanin, C., Mitschler, A., Van Zandt,
M., Joachimiak, A., El-Kabbani, O., and Podjarny, A. (2011) The
Crystallographic Structure of Aldose Reductase IDD393 Complex
Confirms Leu300 as a Specificity Determinant, manuscript to be
submitted.
(13) Bohren, K. M., Page, J. L., Shankar, R., Henry, S. P., and
Gabbay, K. H. (1991) Expression of Human Aldose and Aldehyde
Reductases. J. Biol. Chem. 266, 24031−24037.
(14) Bohren, K.M., Grimshaw, C. E., Lai, C. J., Harrison, D. H.,
Ringe, D., Petsko, G. A., and Gabbay, K. H. (1994) Tyrosine-48 Is the
Proton Donor and Histidine-110 Directs Substrate Stereochemical
Selectivity in the Reduction Reaction of Human Aldose Reductase:
Enzyme Kinetics and Crystal Structure of the Y48H Mutant Enzyme.
Biochemistry 33, 2021−2032.
(15) Kador, P. F. (1988) The Role of Aldose Reductase in the
Development of Diabetic Complications. Med. Res. Rev. 8, 325−352.
(16) Srivastava, S. K., Ramana, K. V., and Bhatnagar, A. (2005) Role
of Aldose Reductase and Oxidative Damage in Diabetes and the
Consequent Potential for Therapeutic Options. Endocr. Rev. 26, 380−
392.
(17) Kaul, C. L., and Ramarao, P. (2001) The Role of Aldose
Reductase Inhibitors in Diabetic Complications: Recent Trends.
Methods Find. Exp. Clin. Pharmacol. 23, 465−475.
(18) Urzhumtsev, A., Tete-Favier, F., Mitschler, A., Barbanton, J.,
Barth, P., Urzhumtseva, L., Biellmann, J.-F., Podjarny, A., and Moras,
D. (1997) A ‘specificity’ pocket inferred from the crystal structures of
the complexes of aldose reductase with the pharmaceutically important
inhibitors tolrestat and sorbinil. Structure 5, 601−612.
(19) Van Zandt, M. C., Sibley, E. O., McCann, E. E., Combs, K. J.,
Flam, B., Sawicki, D. R., Sabetta, A., Carrington, A., Sredy, J., Howard,
E., Mitschler, A., and Podjarny, A. (2004) Design and synthesis of
highly potent and selective (2-arylcarbamoyl-phenoxy)-acetic acid
inhibitors of aldose reductase for treatment of chronic diabetic
complications. Bioorg. Med. Chem. 12, 5661−5675.
(20) Steuber, H., Zentgraf, M., Podjarny, A., Heine, A., and Klebe, G.
(2006) High-resolution Crystal Structure of Aldose Reductase
Complexed with the Novel Sulfonyl-pyridazinone Inhibitor Exhibiting
an Alternative Active Site Anchoring Group. J. Mol. Biol. 356, 45−56.
ASSOCIATED CONTENT
■
S
* Supporting Information
(1) Analysis of Stark spectroscopy data, (2) crystallographic
data, (3) temperature-dependent IR data, and (4) para-
merization of inhibitor 2. This material is available free of
charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: sboxer@stanford.edu. Phone: (650) 723-4482. Fax:
(650) 723-4817.
Funding
This work was supported in part by a grant from the National
Institutes of Health (GM27738) to S.G.B.
ACKNOWLEDGMENTS
■
We thank Drs. Aaron Fafarman and Lauren Webb for useful
discussions about all aspects of this work. We thank Prof.
Michael Van Zandt for useful discussions about inhibitor
design. We also thank Dr. Sushant Malhotra for helpful
discussions about the inhibitor synthesis pathway and Dr. Jorge
Zuniga from the Brunger lab at Stanford for assisting us in
setting up robot screening trays for obtaining the initial
crystallization conditions. Portions of this research were
conducted at the Stanford Synchrotron Radiation Light source,
a national user facility operated by Stanford University on
behalf of the U.S. Department of Energy, Office of Basic Energy
Sciences. The SSRL Structural Molecular Biology Program is
supported by the Department of Energy, Office of Biological
and Environmental Research, and by the National Institutes of
Health, National Center for Research Resources, Biomedical
Technology Program, and the National Institute of General
Medical Sciences.
ABBREVIATIONS
■
hALR2, human aldose reductase; VSE, vibrational Stark effect;
IR, infrared; IDD, Institute for Diabetes Discovery; FCC, flash
column chromatography; DMF, dimethylformamide; MeTHF,
2-methyltetrahydrofuran; MD, molecular dynamics.
REFERENCES
■
(1) Honig, B., and Nicholls, A. (1995) Classical Electrostatics in
Biology and Chemistry. Science 268, 1144−1149.
(2) Warshel, A., and Papazyan, A. (1998) Electrostatic effects in
macromolecules: fundamental concepts and practical modeling. Curr.
Opin. Struct. Biol. 8, 211−217.
(3) Simonson, T. (2001) Macromolecular electrostatics: Continuum
models and their growing pains. Curr. Opin. Struct. Biol. 11, 243−252.
8321
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322

Biochemistry
Article
(21) Andrews, S. S., and Boxer, S. G. (2000) Vibrational Stark Effects
of Nitriles I. Methods and Experimental Results. J. Phys. Chem. A 104,
11853−11863.
(22) Fafarman, A. T., Webb, L. J., Chuang, J. I., and Boxer, S. G.
(2006) Site-Specific Conversion of Cysteine Thiols into Thiocyanate
Creates an IR Probe for Electric Fields in Proteins. J. Am. Chem. Soc.
128, 13356−13357.
(23) Andrews, S. S., and Boxer, S. G. (2002) Vibrational Stark Effects
of Nitriles II. Physical Origins of Stark Effects from Experiment and
Perturbation Models. J. Phys. Chem. A 106, 469−477.
(24) Steuber, H., Heine, A., and Klebe, G. (2007) Structural and
Thermodynamic Study on Aldose Reductase: Nitro-substituted
Inhibitors with Strong Enthalpic Binding Contribution. J. Mol. Biol.
368, 618−638.
(25) Lamour, V., Barth, P., Rogniaux, H., Poterszman, A., Howard,
E., Mitschler, A., Van Dorsselaer, A., Podjarny, A., and Motas, D.
(1999) Production of crystals of human aldose reductase with very
high resolution diffraction. Acta Cystallogr. 55, 721−723.
(26) Kurono, M., Fujiwara, I., and Yoshida, K. (2001) Stereospecific
interaction of a novel spirosuccinimide type aldose reductase inhibitor,
AS-3201, with aldose reductase. Biochemistry 40, 8216−8226.
(27) DeLano, W. L. http://www.pymol.org.
(28) Sorin, E. J., and Pande, V. S. (2005) Exploring the Helix-Coil
Transition via All-Atom Equilibrium Ensemble Simulations. Biophys. J.
88, 2472−2493.
(29) http://ambermd.org/antechamber/antechamber.html.
(30) Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F., and
Hermans, J. (1981) Interaction models for water in relation to protein
hydration. In Intermolecular Forces (Pullman, B., Ed.) p 331, Kluwer
Academic Publishers, Dordrecht, The Netherlands.
(31) Koch, C., Heine, A., and Klebe, G. (2011) Tracing the Detail:
How Mutations Affect Binding Modes and Thermodynamics
Signatures of Closely Related Aldose Reductase Inhibitors. J. Mol.
Biol. 406, 700−712.
(32) Purcell, K. F., and Drago, R. S. (1966) Studies of the Bonding in
Acetonitrile Adducts. J. Am. Chem. Soc. 88, 919−924.
(33) Bertie, J. E., and Lan, Z. (1997) Liquid Water-Acetonitrile
Mixtures at 25 °C: The Hydrogen-Bonded Structure Studied through
Infrared Absolute Integrated Absorption Intensities. J. Phys. Chem. B
101, 4111−4119.
(34) Curtiss, L. A., and Pople, J. A. (1973) Molecular Orbital
Calculation of Some Vibrational Properties of Ilse Complex between
HCN and HF. J. Mol. Spectrosc. 48, 413−426.
(35) Reimers, J.R., and Hall, L. E. (1999) The Solvation of
Acetonitrile. J. Am. Chem. Soc. 121, 3730−3744.
(36) Eaton, G., Pena-Nunez, A. S., Symons, M. C. R., Ferrario, M.,
and McDonald, I. R. (1988) Spectroscopic and Molecular Dynamics
Studies of Solvation of Cyanomethane and Cyanide Ions. Faraday
Discuss. Chem. Soc. 85, 237−253.
(37) Nyquist, R. A. (1990) Solvent-Induced Nitrile Frequency Shifts:
Acetonitrile and Benzonitrile. Appl. Spectrosc. 44, 1405−1407.
(38) Fawcett, W. R., Liu, G., and Kessler, T. E. (1993) Solvent-
Induced Frequency Shifts in the Infrared Spectrum of Acetonitrile in
Organic Solvents. J. Phys. Chem. 97, 9293−9298.
(43) Aschaffenburg, D. J., and Moog, R. S. (2009) Probing Hydrogen
Bonding Environments: Solvatochromic Effects on the CN Vibration
of Benzonitrile. J. Phys. Chem. B 113, 12736−12743.
(44) Lindquist, B. A., Haws, R. T., and Corcelli, S. A. (2008)
Optimized Quantum Mechanics/Molecular Mechanics Strategies for
Nitrile Vibrational Probes: Acetonitrile and para-Tolunitrile in Water
and Tetrahydrofuran. J. Phys. Chem. B 112, 13991−14001.
(45) Chaban, G. M. (2004) Anharmonic Vibrational Spectrospcopy
of Nitriles and Their Complexes with Water. J. Phys. Chem. A 108,
4551−4556.
(46) Eswar, N., and Ramakrishnan, C. (2000) Deterministic features
of side-chain main-chain hydrogen bonds in globular protein. Protein
Eng. 13, 227−238.
(47) Kim, Y. S., and Hochstrasser, R. M. (2005) Chemical exchange
2D IR of hydrogen-bond making and breaking. Proc. Natl. Acad. Sci.
U.S.A. 102, 11185−11190.
(48) Wofford, B. A., Eliades, M. E., Lieb, S. G., and Bevan, J. W.
(1987) Determination of dissociation energies and thermal functions
of hydrogen-bond formation using high resolution FTIR spectroscopy.
J. Chem. Phys. 87, 5674−5680.
(49) Feyereisen, M. W., Feller, D., and Dixon, D. A. (1996)
Hydrogen Bond Energy of the Water Dimmer. J. Phys. Chem. 100,
2993−2997.
(50) Rablen, R. P., Lockman, W. J., and Jorgensen, L. W. (1998) Ab
Initio Study of Hydrogen-Bonded Complexes of Small Organic
Molecules with Water. J. Phys. Chem. A 102, 3782−3797.
(51) Lantz, M. A., Jarvis, S. P., Tokumoto, H., Martynski, T., Kusumi,
T., Nakamura, C., and Miyake, J. (1999) Streching the α-helix: A direct
measure of the hydrogen-bond energy of a single-peptide molecule.
Chem. Phys. Lett. 315, 61−68.
(52) Wang, Z., Wu, C., Lei, H., and Duan, Y. (2007) Accurate an
Initio Study on the Hydrogen-Bond Pairs in Protein Secondary
Structures. J. Chem. Theory Comput. 3, 1527−1537.
(53) Zhang, Y., and Wang, C. (2008) Estimation on the
Intramolecular 10-Membered Ring N−H···OC Hydrogen-Bonding
Energies in Glycine and Alanine Peptides. J. Comput. Chem. 30, 1251−
1260.
(54) Abraham, M. H., Grellier, P. L., Prior, D. V., Morris, J. J., and
Taylor, P. J. (1990) Hydrogen Bonding. Part 10. A Scale of Solute
Hydrogen-Bond Basicity using log K Values for Complexation in
Tetrachloromethane. J. Chem. Soc., Perkin Trans. 2 4, 521−529.
(55) Choi, J., Oh, K., Lee, H., Lee, C., and Cho, M. (2008) Nitrile
and thiocyanate IR probes: Quantum chemistry calculation studies and
multivariate least-square fitting analysis. J. Chem. Phys. 128, 134506.
(56) Fafarman, A. T., Sigala, P. A., Schwans, J. P., Fenn, T. D.,
Herschlag, D., and Boxer, S. G. Quantitative, Directional Measurement
of Electric Field Heterogeneity Within an Enzyme Active Site,
manuscript submitted.
(57) Meller, J., and Elber, R. (1998) Computer Simulations of
Carbon Monoxide Photodissociation in Myoglobin: Structural
Interpretation of the B States. Biophys. J. 74, 789−802.
(58) Nutt, D. R., and Meuwly, M. (2003) Theoretical Investigation
of Infrared Spectra and Pocket Dynamics of Photodissociated
Carbonmonoxy Myoglobin. Biophys. J. 85, 3612−3623.
(59) Anselmi, M., Aschi, M., Nola, A. D., and Amadei, A. (2007)
Theoretical Characterization of Carbon Monoxide Vibrational
Spectrum in Sperm Whale Myoglobin Distal Pocket. Biophys. J. 92,
3442−3447.
(39) Jamroz, D., Stangret, J., and Lindgren, J. (1993) An Infrared
Spectroscopic Study of the Preferential Solvation in Water-Acetonitrile
Mixtures. J. Am. Chem. Soc. 115, 6165−6168.
(40) Getahun, Z., Huang, C., Wang, T., De Leon, B., DeGrado, W.
F., and Gai, F. (2003) Using Nitrile-Derivatized Amino Acids as
Infrared Probes of Local Environment. J. Am. Chem. Soc. 125, 405−
411.
(41) Huang, C., Wang, T., and Gai, F. (2003) Temperature
dependence of the CN stretching vibration of a nitrile-derivatized
phenylalanine in water. Chem. Phys. Lett. 371, 731−738.
(42) Maienschein-Cline, M. G., and Londergan, C. H. (2007) The
CN Stretching Band of Aliphatic Thiocyanate is Sensitive to Solvent
Dynamics and Specific Solvation. J. Phys. Chem. A 111, 10020−10025.
(60) Plattner, N., and Meuwly, M. (2008) The Role of Higher CO-
Multipole Moments in Understanding the Dynamics of Photo-
dissociated Carbonmonoxide in Myoglobin. Biophys. J. 94, 2505−2515.
(61) Zheng, J., and Fayer, M. D. (2007) Hydrogen Bond Lifetimes
and Energetics for Solute-Solvent Complexes Studied with 2D-IR
Vibrational Echo Spectroscopy. J. Am. Chem. Soc. 129, 4328−4335.
8322
dx.doi.org/10.1021/bi200930f|Biochemistry 2011, 50, 8311−8322