The activating oxydianion binding domain for enzyme-catalyzed proton transfer, hydride transfer, and decarboxylation: Specificity and enzyme architecture

Article abstract of DOI:10.1021/ja5123842

The kinetic parameters for activation of yeast triosephosphate isomerase (ScTIM), yeast orotidine monophosphate decarboxylase (ScOMPDC), and human liver glycerol 3-phosphate dehydrogenase (hlGPDH) for catalysis of reactions of their respective phosphodianion truncated substrates are reported for the following oxydianions: HPO₃^2-, FPO₃^2-, S₂O₃^2-, SO₄^2- and HOPO₃^2-. Oxydianions bind weakly to these unliganded enzymes and tightly to the transition state complex (E?S?), with intrinsic oxydianion Gibbs binding free energies that range from -8.4 kcal/mol for activation of hlGPDH-catalyzed reduction of glycolaldehyde by FPO₃^2- to -3.0 kcal/mol for activation of ScOMPDC-catalyzed decarboxylation of 1-β-D-erythrofuranosyl)orotic acid by HOPO₃^2-. Small differences in the specificity of the different oxydianion binding domains are observed. We propose that the large -8.4 kcal/mol and small -3.8 kcal/mol intrinsic oxydianion binding energy for activation of hlGPDH by FPO₃^2- and S₂O₃^2-, respectively, compared with activation of ScTIM and ScOMPDC reflect stabilizing and destabilizing interactions between the oxydianion -F and -S with the cationic side chain of R269 for hlGPDH. These results are consistent with a cryptic function for the similarly structured oxydianion binding domains of ScTIM, ScOMPDC and hlGPDH. Each enzyme utilizes the interactions with tetrahedral inorganic oxydianions to drive a conformational change that locks the substrate in a caged Michaelis complex that provides optimal stabilization of the different enzymatic transition states. The observation of dianion activation by stabilization of active caged Michaelis complexes may be generalized to the many other enzymes that utilize substrate binding energy to drive changes in enzyme conformation, which induce tight substrate fits. (Table Presented).

Full text of DOI:10.1021/ja5123842

Article
pubs.acs.org/JACS
The Activating Oxydianion Binding Domain for Enzyme-Catalyzed
Proton Transfer, Hydride Transfer, and Decarboxylation: Specificity
and Enzyme Architecture
Archie C. Reyes, Xiang Zhai, Kelsey T. Morgan, Christopher J. Reinhardt, Tina L. Amyes,
and John P. Richard*
Department of Chemistry, University at Buffalo, SUNY, Buffalo, New York 14260-3000, United States
ABSTRACT: The kinetic parameters for activation of yeast
triosephosphate isomerase (ScTIM), yeast orotidine mono-
phosphate decarboxylase (ScOMPDC), and human liver
glycerol 3-phosphate dehydrogenase (hlGPDH) for catalysis
of reactions of their respective phosphodianion truncated
2−
substrates are reported for the following oxydianions: HPO3
,
FPO32−, S2O32−, SO4 and HOPO32−. Oxydianions bind
2−
weakly to these unliganded enzymes and tightly to the
transition state complex (E·S‡), with intrinsic oxydianion
Gibbs binding free energies that range from −8.4 kcal/mol for activation of hlGPDH-catalyzed reduction of glycolaldehyde by
2−
FPO3 to −3.0 kcal/mol for activation of ScOMPDC-catalyzed decarboxylation of 1-β-D-erythrofuranosyl)orotic acid by
HOPO32−. Small differences in the specificity of the different oxydianion binding domains are observed. We propose that the
2−
large −8.4 kcal/mol and small −3.8 kcal/mol intrinsic oxydianion binding energy for activation of hlGPDH by FPO3 and
S2O32−, respectively, compared with activation of ScTIM and ScOMPDC reflect stabilizing and destabilizing interactions between
the oxydianion −F and −S with the cationic side chain of R269 for hlGPDH. These results are consistent with a cryptic function
for the similarly structured oxydianion binding domains of ScTIM, ScOMPDC and hlGPDH. Each enzyme utilizes the
interactions with tetrahedral inorganic oxydianions to drive a conformational change that locks the substrate in a caged Michaelis
complex that provides optimal stabilization of the different enzymatic transition states. The observation of dianion activation by
stabilization of active caged Michaelis complexes may be generalized to the many other enzymes that utilize substrate binding
energy to drive changes in enzyme conformation, which induce tight substrate fits.
Loop-phosphodianion binding interactions anchor substrates
to enzymes and are expressed at the Michaelis complex.
However, the 11−13 kcal/mol intrinsic Gibbs phosphodianion
binding free energy, determined from the ratio of the second-
order rate constant kcat/Km (Figure 1) for the enzyme-catalyzed
reaction of the whole substrate and (kcat/Km)E for reaction of the
truncated substrate alone, exceeds the total substrate binding
energy for the reactions catalyzed by TIM, OMPDC and GPDH,
so that a significant fraction of these dianion binding interactions
are expressed specifically at the transition state for the catalyzed
reaction.10,12,14,16 We separated the anchoring and activating
interactions of the phosphodianion by cutting the covalent
connection to the reacting substrate10,12,14,16 and characterizing
activation of enzyme catalysis of a phosphodianion truncated
substrate by phosphite dianion. The dianion binding energy
recovered as activation by the phosphite dianion piece (6−8
kcal/mol) is determined from the ratio of the third-order rate
constant kcat/KHPiKd (Figure 1) for the enzyme-catalyzed
reaction of the pieces and the second order rate constant (kcat/
Km)E for reaction of the truncated substrate alone.10,12,14,16,26
INTRODUCTION
■
Studies on the origin of enzymatic rate enhancements often focus
on understanding the intermolecular electrostatic, hydrogen
bonding and hydrophobic interactions responsible for stabiliza-
tion of enzymatic transition states.1 These interactions may be
characterized for a particular enzyme in mutagenesis studies, with
the aim of determining whether the sum of stabilizing protein−
ligand interactions is sufficiently large to account for the observed
transition state stabilization.2−4 We are interested in the more
general problem of defining paradigms for enzyme architecture,
which favor large enzymatic rate accelerations across broad
spectra of enzyme-catalyzed reactions.4−9
Enzyme-catalyzed decarboxylation by orotidine monophos-
phate decarboxylase (OMPDC),10 proton transfer by triose-
phosphate isomerase,11−13 and by OMPDC,14,15 hydride transfer
by α-glycerol phosphate dehydrogenase (GPDH),16 and
phosphoryl transfer by α-phosphoglucomutase17 proceed
through strikingly different transition states, yet each transition
state is stabilized by 11−13 kcal/mol by interactions with the
nonreacting phosphodianion of substrate. These enzymes
interact with substrate through a gripper loop, and we have
proposed that the wide distribution of these loops,18−22 and of
phosphate cups,23 reflects their common function in providing a
large intrinsic phosphodianion binding Gibbs free energy.6,24,25
Received: December 4, 2014
Published: January 2, 2015
© 2015 American Chemical Society
1372
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
rationalize enzyme specificity through ligand driven protein
conformational changes that induce tight fits for the substrate at
the enzyme active site.22,28−36 Its power is derived from the
generality of a mechanism that provides specificity in the binding
of any nonreactive substrate fragment to the enzyme−transition
state complex, provided the binding energy is utilized to stabilize
an active enzyme−substrate cage that shows specificity in
binding to the reaction transition state complex.6,24
We are interested in determining and comparing structure−
function relationships for the dianion binding sites of yeast TIM
(ScTIM), yeast OMPDC (ScOMPDC) and human liver GPDH
(hlGPDH). ScOMPDC-catalyzed decarboxylation of (1-β-D-
erythrofuranosyl)orotic acid (EO) to form (1-β-D-
erythrofuranosyl)uracil (EU) is activated by a series of five
tetrahedral oxydianions, which bind weakly to unliganded
OMPDC and tightly to the enzyme−transition state complex,
with the following intrinsic oxydianion binding energies (kcal/
mol): SO32−, −8.3; HPO32−, −7.7; S2O32−, −4.6; SO42−, −4.5;
HOPO32−, −3.0; HOAsO32−, no activation detected.7 These
results highlight the cryptic function for the phosphodianion-
binding site of OMPDC, which provides Gibbs binding free
energy to stabilize the Michaelis complex, while serving the more
significant role of a dianion activation site. In this paper we
examine the specificity of the dianion activation sites at ScTIM
and hlGPDH, respectively, for activation of proton transfer and
hydride transfer by a series of structurally homologous
oxydianions. These new data allow a comparison of the kinetic
parameters for activation of ScTIM, hlGPDH and ScOMPDC by
2−
Figure 1. Reactions of whole substrates RCH2OPO3 (kcat/Km) and
the pieces RH + HPO32− (kcat/KHPiKd) catalyzed by OMPDC, TIM and
GPDH. A total 11−13 kcal/mol Gibbs phosphodianion binding free
energy for RCH2OPO32− and 6−8 kcal/mol phosphite dianion binding
energy for R−H is utilized in the stabilization of the transition states for
the respective enzyme-catalyzed decarboxylation, proton and hydride
transfer reactions.10,12,16,26 The 4−6 kcal/mol difference between the
phosphodianion and phosphite dianion binding energy is the entropic
advantage that arises from the covalent connection between the
substrate pieces.4,5,27
2
2−
five oxydianions: HPO32−, FPO3 , S2O32−, SO42− and HOPO3
.
Our results reveal a striking similarity in the specificity of these
enzymes for dianion activation, which is consistent with a similar
architecture for the three dianion binding sites.
The transition states for the enzyme-catalyzed reactions of the
whole substrate and the substrate pieces were examined by
comparing the effects of site-directed mutations at TIM and at
OMPDC on the kinetic parameters kcat/Km (M−1 s−1) and kcat/
EXPERIMENTAL SECTION
■
Materials. Water was from a Milli-Q Academic purification system.
Bovine serum albumin, fraction V (BSA) was purchased from Roche.
DEAE-Sepharose and Sephacryl S-200 were purchased from GE
Healthcare. DEAE-Sephadex A25, Dowex 50WX4−200R, NADH
(disodium salt), NAD+ (free acid), glycolaldehyde dimer, triethylamine
(≥99.5%), triethanolamine hydrochloride and D,L-dithiothreitol
(DTT) were purchased from Sigma-Aldrich. Protease inhibitor tablets
(Complete) were purchased from Roche. Ammonium sulfate (enzyme
grade), sodium bicarbonate, sodium hydroxide (1.0 N) and hydro-
chloric acid (1.0 N) were purchased from Fisher. Sodium phosphite
(dibasic, pentahydrate) was purchased from Fluka. The water content
was reduced to Na2HPO3·0.4H2O as previously described.12 Glyco-
laldehyde labeled with 13C at carbon-1 ([1-13C]-GA, 99% enrichment of
13C, 0.09 M in water) was purchased from Omicron Biochemicals.
Deuterium oxide (99% D) and deuterium chloride (35% w/w, 99.9% D)
were purchased from Cambridge Isotope Laboratories. Imidazole was
recrystallized from benzene. Deuterium oxide (99.9% D) and deuterium
chloride (35% w/w, 99.9% D) were from Cambridge Isotope
Laboratories. Orotidine 5′-monophosphate (OMP) was synthesized
enzymatically from 5-phosphoribosyl-1-pyrophosphate and orotic
acid,37 and 1-(β-D-erythrofuranosyl)orotic acid (EO) was available
from an earlier study.10 All other chemicals were reagent grade or better
and were used without further purification.
K
HPiKd (M−2 s−1) for the respective bimolecular and termolecular
enzymatic reactions (Figure 1). In both cases a slope of 1.0 was
observed for a logarithmic plot of kcat/Km (M−1 s−1) against kcat/
K
HPiKd (M−2 s−1) for reactions catalyzed by the respective
wildtype enzyme and by a broad range of mutant enzymes. This
shows that the mutations result in the same change in the
activation barriers for the enzyme-catalyzed reactions of whole
substrate and the substrate in pieces.4,5 The results are consistent
with reactions that proceed through transition states of
essentially the same structure and that interact in the same
manner with the protein catalyst, but with an entropic advantage
to reaction of the whole substrate. They allow partitioning of the
total 11−13 kcal/mol Gibbs dianion binding free energy into the
6−8 kcal/mol binding energy expressed at the transition state for
the catalyzed of reaction phosphite dianion, and the additional
4−6 kcal/mol entropic (anchoring) effect from connecting the
substrate pieces (Figure 1).4,5
Our model for enzyme activation by oxydianions follows from
the function of the oxydianion to drive closure of a flexible
gripper loop over the enzyme-bound substrate.6,11,25 The ligand-
driven conformational change converts enzymes from an inactive
open form to an active closed form, which traps the substrate in a
catalytically active protein cage. The dianion acts by providing
the electrostatic glue required to lock enzymes in their closed
forms. However, the dianion is a spectator, in the sense that its
binding has little or no effect on the chemistry at the catalytic
site.6,24,25 This proposal folds back onto an earlier suggestion to
A stock solution of unlabeled glycolaldehyde dimer (200 mM
monomer) used for studies on GPDH was prepared by dissolving the
dimer in water and storing the solution for at least 3 days at room
temperature to allow for breakdown to the monomer.12 A solution of
labeled [1-13C]-glycolaldehyde (1 mL of a 90 mM solution in H2O) used
for studies on ScTIM was reduced to a volume of ca. 100 μL by rotary
evaporation, 5 mL of D2O was added, and the volume was again reduced
to ca. 100 μL by rotary evaporation. This procedure was repeated twice
more and ca. 900 μL of D2O was added to the final solution to give a
1373
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
volume of ca. 1 mL. The stock solution of [1-13C]-GA in D2O was stored
at room temperature to minimize the content of glycolaldehyde dimer,12
OMPDC was monitored by following the decrease in absorbance at 279
nm.10,15 The assay solution (1 mL) contained OMP (≈0.05 mM) at pH
7.1 (30 mM MOPS), 25 °C and I = 0.10 maintained with NaCl.
GPDH was assayed by monitoring the oxidation of NADH by
DHAP.45 The standard assay mixture contained 100 mM triethanol-
amine (pH 7.5), 1 mM DHAP, 0.20 mM NADH, 0.1 mg/mL BSA, 50
μM DTT, and ca. 0.67 nM GPDH at I = 0.12 (NaCl). The kinetic
parameters kcat and Km for the GPDH-catalyzed reaction of DHAP at
saturating 0.20 mM NADH was determined from the fit of kinetic data
to the Michaelis−Menten equation. The value of Ki for competitive
inhibition of GPDH-catalyzed reduction of DHAP by HPO32− at pH 7.5
and I = 0.12 (NaCl) was determined by examining the effect of
1
and the concentration of [1-13C]-GA was determined by H NMR
spectroscopy.11 The barium salt of D-glyceraldehyde 3-phosphate
diethyl acetal was prepared by a literature procedure.38 The solutions
in D2O used for studies of the ScTIM-catalyzed reactions of [1-13C]-GA
were prepared as described in earlier work.11
Enzymes. The C155S mutant of OMPDC from Saccharomyces
cerevisiae (ScOMPDC) was prepared according to a literature procedure
and stored at −80 °C.39,40 The enzyme was defrosted and dialyzed at 4
°C against 10 mM MOPS (50% free base) at pH 7.1 containing 100 mM
NaCl, unless noted otherwise. Wildtype TIM from Saccharomyces
cerevisiae (ScTIM) was expressed from the TIM-deficient tpiA− λDE3
lysogenic strain of Escherichia coli, purified by a literature procedure, and
stored at −80 °C in 25 mM Tris-HCl at pH 8.0 and I = 0.1 (NaCl)
containing 20% glycerol.5
increasing concentrations of DHAP (0.01−0.50 mM) on the initial
2−
velocity for reaction in the presence of 15 mM and 30 mM HPO3
.
Enzyme-Catalyzed Reactions of Phosphodianion Truncated
Substrates. OMPDC. The decarboxylation of EO (0.1 mM) catalyzed
by ScOMPDC [14 μM] in the presence of 2−40 mM FPO32− and 5 mM
MOPS at pH 7.0, 25 °C and I = 0.14 (NaCl) was monitored at 283
nm.10,15 Reactions (1 mL total volume) were initiated by the addition of
50 μL of ScOMPDC in 100 mM MOPS (pH 7.0) and were monitored
for up to 10 h. These reactions obeyed excellent first-order kinetics with
stable end points at 10 reaction halftimes. Values of kobs (s−1) were
obtained from the fits of the absorbance vs time trace to a single
exponential decay. The apparent second-order rate constants (kcat/
Km)obs (M−1 s−1) for ScOMPDC-catalyzed decarboxylation of EO were
calculated using the relationship (kcat/Km)obs = kobs/[E].
The plasmid pDNR-dual donor vector containing the gene for wild-
type human liver GPDH (hlGPDH) gene insert was purchased from the
Harvard plasmid repository. The insert gene was subcloned into a
bacterial expression vector pET-15b from Novagen and used for
transformations of cells from E. coli strain Bl 21(DE3). These cells were
then grown overnight in 200−300 mL of LB medium that contained 100
μg/mL ampicillin at 37 °C. This culture was diluted into 4 L of LB
medium (100 μg/mL ampicillin), and grown at 37 °C to OD600 = 0.6, at
which point 0.6 mM isopropyl-1-thio-D-galactoside was added to the
culture to induce protein expression. After 12 h of overexpression, the
cells were harvested and stored in 20 mL of 25 mM Tris-HCl buffer that
contains 0.5 M NaCl at pH 7.9.
The cell pellets were suspended in 25 mM Tris-HCl at pH 7.9 in the
presence of protease inhibitors (Complete) and lysed using a French
press. The lysate was diluted to 40 mL with the same buffer and
centrifuged at 13 000 rpm for 60 min. Solid ammonium sulfate was
added to the lysate supernatant to give a 30% saturated solution, which
was then centrifuged at 14 000 rpm for 20 min. The supernatant was
mixed with solid ammonium sulfate to give a 40% saturated solution.
After 20 min the mixture was centrifuged at 14 000 rpm for 20 min and
the solid was dissolved in 25 mL of 25 mM Tris-HCl buffer at pH 7.9.
The protein solution was dialyzed overnight against 25 mM Tris buffer
pH 7.9 at 4 °C. The dialysate was loaded onto a DEAE-Sepharose ion-
exchange column previously equilibrated against 25 mM Tris-HCl at pH
7.9. The column was eluted with 800 mL of a linear 0−150 mM gradient
of NaCl in the same buffer. The protein concentration was determined
from the UV absorbance at 280 nm, using the extinction coefficient of
18 450 M−1cm−1 calculated for a subunit molecular weight of 37 500 Da
using the ProtParam tool available on the Expasy server.41,42 The
fractions that contained GPDH were pooled, concentrated, and further
purified over a Sephacryl S-200 column, equilibrated with 25 mM Tris
pH 7.9 that contains 0.15 M NaCl; and, eluting with the same buffer
solution. The GPDH obtained from this column was judged to be
homogeneous by gel electrophoresis. Fractions with OD280 > 1 were
pooled, concentrated and stored at −80 °C in 20% glycerol, 25 mM Tris-
HCl buffer at pH 7.9, and 100 mM NaCl. The final yield of GPDH from
a 2 L bacterial culture was typically 100 mg.
ScTIM. The reactions of [1-13C]-GA in D2O catalyzed by wildtype
ScTIM were monitored by 1H NMR spectroscopy, as described
previously.11 The frozen enzyme was thawed and exhaustively dialyzed
against 30 mM imidazole buffer (20% free base) in D2O at pD 7.0 and I =
0.024 (NaCl). The TIM-catalyzed reactions of [1-13C]-GA in the
presence of dianion activators were initiated by the addition of enzyme
to give reaction mixtures (850 μL in D2O) containing 20 mM [1-13C]-
GA, 20 mM imidazole (20% free base), up to 30 mM dianion in D2O at
pD 7.0 and I = 0.1 (NaCl). ScTIM was added to give the following range
of final concentrations for the different dianion activators: 20−40 μM
2
ScTIM, HPO32−; 30−100 μM ScTIM, FPO3 ; 80−240 μM ScTIM,
2−
SSO32−; 20−80 μM ScTIM, SO42−; 30−130 μM ScTIM, HPO4 In
each case, 750 μL of the reaction mixture was transferred to an NMR
tube, the 1H NMR spectrum was recorded immediately and spectra were
then recorded at regular intervals. The remaining reaction mixture was
incubated at 25 °C and was used to conduct periodic assays of the
activity of TIM. No significant loss in TIM activity was observed during
any of these reactions. These reactions were generally followed for ca.
2−
30−40% reaction: 3−14 h, HPO32−; 5−12 h, FPO32−; 18−72 h, SO4
;
and 15−80 h, HPO42−. The reactions in the presence of SSO32−were
followed for ca. 40−60% reaction over 5−11 h. Once sufficient data had
been obtained, the protein was removed by ultrafiltration and the
solution pD was determined. There was no significant change in pD
observed during any of these reactions.
These reactions were monitored by 1H NMR analyses, as described in
previous work.11 Observed first-order rate constants for the
disappearance of [1-13C]-GA, kobs (s−1) were determined from the
slopes of linear semilogarithmic plots of reaction progress against time
covering the first 30−60% of the reaction (eq 1), where fS is the fraction
of [1-13C]-GA remaining at time t. The apparent second-order rate
constants, (kcat/Km)obs (M−1 s−1), were calculated using eq 2, where fhyd
= 0.94 is the fraction of [1-13C]-GA present in the unreactive hydrate
form, and [E] is the concentration of ScTIM.12
1H NMR Analyses. 1H NMR spectra at 500 MHz were recorded in
D2O at 25 °C using a Varian Unity Inova 500 spectrometer that was
shimmed to give a line width of ≤0.5 Hz for the most downfield peak of
the double triplet due to the C-1 proton of the hydrate of [1-13C]-GA.11
Spectra (16−64 transients) were obtained using a sweep width of 6000
Hz, a pulse angle of 90° and an acquisition time of 6 s. A total relaxation
delay of 120 s (>8T1) between transients was used to ensure that
accurate integrals were obtained for the protons of interest.43,44
Baselines were subjected to a first-order drift correction before
determination of integrated peak areas. Chemical shifts are reported
relative to that for HOD at 4.67 ppm.
Enzyme Assays. The TIM-catalyzed isomerization of GAP was
monitored by coupling the formation of the DHAP to the oxidation of
NADH catalyzed by GPDH.11,45 The standard assay mixture (1.0 mL)
contained 100 mM triethanolamine buffer (pH 7.5, I = 0.1), 0.1 mg/mL
BSA, 0.2 mM NADH, 2 mM GAP, 1 unit of GPDH and 0.01−0.05 nM
ScTIM at I = 0.10 (NaCl). The decarboxylation of OMP catalyzed by
ln fs = −kobs
t
(1)
kobs
(1 − fhyd )[E]
(kcat/Km)obs
=
(2)
hlGPDH. Stock solutions of hlGPDH (7 or 14 mg/mL) were dialyzed
exhaustively against 20 mM triethanolamine at 4 °C. Assay mixtures for
the hlGPDH-catalyzed reduction of GA in the absence of phosphite
contained 10 mM TEA (pH 7.5), 20−60 mM GA, 0.20 mM NADH and
0.05 mM hlGPDH at I = 0.12 (NaCl). Assay mixtures for the hlGPDH-
1374
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
catalyzed reduction of GA in the presence of dianions contained 10 mM
TEA (pH 7.5), 2−60 mM GA, 0.20 mM NADH, a measured
concentration of the dianion at I = 0.12 (NaCl) and the following
enzyme concentrations: 0.4 μM hlGPDH, HPO32−; 0.2 μM hlGPDH,
FPO32−; 4 μM hlGPDH, SO42−; 4 μM hlGPDH, HOPO32−; 8 μM
hlGPDH, S2O32−. The initial velocity of enzyme-catalyzed reduction of
GA by NADH was determined by monitoring the decrease in
absorbance at 340 nm over a period of 20 min for the dianion activated
hydride transfer reactions, and a period of 120 min for the unactivated
hydride transfer reactions. The kinetic parameters were determined
from the nonlinear least-squares fits of kinetic data to the appropriate
kinetic Scheme, using the fitting software provided by the Prism
graphing program (GraphPad).
Scheme 2
described in earlier work.8,11,47 Apparent second-order rate
constants (kcat/Km)obs (M−1 s−1) for the ScTIM-catalyzed
reactions were determined as described in the Experimental
Section. Figure 3 shows the dependence of (kcat/Km)obs for
RESULTS
■
The decarboxylation of EO (0.10 mM) to form EU catalyzed by
OMPDC in the presence of fluorophosphate dianion activator
(Scheme 1) was monitored spectrophotometrically at 283
Scheme 1
Figure 3. Dependence of (kcat/Km)obs for ScTIM-catalyzed turnover of
the free carbonyl form of [1-13C]-GA in D2O on the concentration of
added dianion for reactions at pD 7.0, 25 °C and I = 0.1, NaCl. The solid
●
,
lines show the fits of the data to eq 3 derived for Scheme 2: (A)
2−
2−
2−
2−
2−
■
; , HOPO3 .
▼
●
▲
; , SO4
HPO3
;
, FPO3 . (B) , S2O3
ScTIM-catalyzed turnover of the carbonyl form of [1-13C]-GA in
D2O, on the concentration of dianion activator for reactions at
pD 7.0, 25 °C and I = 0.1, NaCl. These data were fit to eq 3
derived for Scheme 2, using (kcat/Km)E = 0.062 M−1 s−1
determined for the unactivated reaction, to give the values of
KX and (kcat/Km)E·X reported in Table 1. The yields of [2-13C]-
GA (22%), [2-13C, 2-2H]-GA (54%) and [1-13C, 2-2H]-GA
(25%) obtained from reactions activated by HPO32− are similar
to the yields determined in earlier work for the reactions of
[1-13C]-GA catalyzed by TIM from chicken muscle (cTIM) and
Trypanosoma brucei (TbbTIM).11,48 The small differences (not
reported) in the yields of products from ScTIM-catalyzed
reactions of [1-13C]-GA activated by different dianions show that
dianions influence the partitioning of the enediolate reaction
intermediate. These product yields will be reported in a separate
publication.
hlGPDH-Catalyzed Reduction of GA. GPDH from rabbit
muscle follows an ordered mechanism, with NADH binding first
followed by DHAP.49 Initial velocities, vi, of the reduction of
DHAP by NADH catalyzed by hlGPDH at pH 7.5 (10 mM
TEA), 25 °C and I = 0.12 (NaCl) were determined by
monitoring the decrease in absorbance at 340 nm. The essentially
identical Michaelis−Menten plots (not shown) of vi/[E] (s−1)
against [DHAP] (carbonyl form) observed when [NADH] is
held constant at 0.10 mM or 0.20 mM require Km ≪ 0.10 mM for
NADH. The fit of data to the Michaelis−Menten equation gave
(Km)obs = 0.052 mM for DHAP (carbonyl form, fcar = 0.45),50
when GPDH is saturated by NADH, kcat = 240 s−1 for turnover of
the ternary E·NADH·DHAP complex and kcat/Km = 4.6 × 106
M−1 s−1. These kinetic parameters are similar to kcat = 130 s−1, Km
= 0.13 mM and kcat/Km = 1.0 × 106 M−1 s−1 reported in an earlier
Figure 2. Dependence of the apparent second-order rate constant (kcat/
Km)obs for OMPDC-catalyzed turnover of EO on the concentration of
2−
FPO3
.
nm.7,10,15 Figure 2 shows the increase in the apparent second-
order rate constants (kcat/Km)obs (M−1 s−1) for decarboxylation
of EO at increasing [FPO32−]. The solid line shows the nonlinear
least-squares fit of these data to eq 3 [X2− = FPO32−] derived for
Scheme 1, using (kcat/Km)E = 0.026 M−1 s−1,46 KX = 36 2 mM
and (kcat/Km)E·X = 77 3 M−1 s−1.
⎛
kcat/Km = ⎜
⎝
⎞
KX
⎟(kcat/Km)E +
K + [X2−]
⎠
X
[X2−]
⎛
⎜
⎝
⎞
⎟(kcat/Km)E·X
K + [X2−]
⎠
(3)
X
TIM-Catalyzed Reactions of [1-13C]-GA in D2O. The
disappearance of [1-13C]-GA (10 mM ≪ Km)12 and the
formation of products [2-13C]-GA, [2-13C, 2-2H]-GA and
[1-13C, 2-2H]-GA (Scheme 2) from reactions at the active site
of wildtype ScTIM in D2O buffered by 20 mM imidazole at pD
7.0, °C, I = 0.1 (NaCl) and in the presence of dianion activators
(Scheme 2) was monitored by 1H NMR spectroscopy as
1375
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
Table 1. Kinetic Parameters for Activation of ScTIM by Oxydianions (Scheme 2) and Derived Parameters for the Binding of
a
Dianions to [E·S]‡ (Scheme 5)
b
cd
,
ce
,
f
g
dianion
none
(kcat/Km)E (M−1 s−1
)
KX (mM)
18
(kcat/Km)E·X (M−1 s−1
)
(kcat/Km)E·X/KX (M−2 s−1
)
(K‡)X (M)
RT ln(K‡)X (kcal/mol)
0.062
2.3 × 10−5
3.1 × 10−5
8.8 × 10−4
4.7 × 10−4
5.2 × 10−5
−6.3
−6.1
−4.2
−4.5
−5.8
2−
HPO3
3
48
4
2700
2000
2−
FPO3
5.1 1.3
10 0.7
h
h
i
2−
HOPO3
n.d
n.d
70 10
2−
SO4
21
4
2.8 0.3
130
2−
S2O3
7
1.0
7.3 0.3
1200
a
Reactions of 20 mM [1-13C]-GA in D2O at pD 7.0 (20 mM imidazole), 25 °C and I = 0.1 (NaCl). The quoted uncertainty in the kinetic parameters
b
is the standard error determined for the nonlinear least-squares fits of these data. Second-order rate constant for the ScTIM-catalyzed reaction of
c
[1-13C]-GA determined for a reaction in the absence of dianion activator. Kinetic parameter determined from the fit of data shown in Figure 3 to eq
d
e
3. Dissociation constant for release of the oxydianion from ScTIM. Second-order rate constant for the reactions of [1-13C]-GA catalyzed by the
f
enzyme−oxydianion complex (Scheme 2). Dissociation constant for release of the oxydianion from the transition state complex, calculated using eq
g
h
2−
6, derived for Scheme 5. Intrinsic Gibbs dianion binding free energy. Not determined: the plot for activation by HOPO3 (Figure 3) is linear
i
through [HOPO32−] = 14 mM. The slope of the linear correlation from Figure 3.
study of GPDH from rabbit muscle.16 Phosphite dianion
inhibition of hlGPDH-catalyzed reduction of DHAP by
NADH was examined by determining initial velocities, vi, at
seven different [DHAP] between 0.01 and 0.50 mM for reactions
in the presence of constant 15 mM and 30 mM [HPO32−]. A
value of Ki = 42 6 mM was determined from the nonlinear
dianion activation of human liver hlGPDH catalyzed reduction of
GA is 4-fold larger than (kcat/Km)E·X/KX = 4300 M−2 s−1 reported
in an earlier study on for rabbit muscle GPDH.16 Human liver
hlGPDH shows a similar 4-fold larger value of kcat/Km = 4.6 × 106
M−1 s−1 for catalysis of reduction of DHAP at saturating
[NADH], compared with kcat/Km = 1.0 × 106 M−1 s−1 for rabbit
muscle GPDH.16 Figure 5B−D show, respectively, kinetic data
determined under the same reaction conditions for activation of
hlGPDH-catalyzed reduction of GA by fluorophosphate, sulfate,
phosphate and thiosulfate dianion. The solid lines show the
nonlinear least-squares fit of the data to eq 5, using the kinetic
parameters reported in Table 2.
least-squares fits of the kinetic data to eq 4, for competitive
2−
inhibition by HPO3
.
vi
=
kcat[DHAP]
[DHAP] + Km(1 + [HPO32−]/Ki)
[E]
(4)
Initial velocities, vi, for the reduction of GA by NADH (0.2
mM) catalyzed by hlGPDH at pH 7.5 (10 mM TEA), 25 °C and I
= 0.12 (NaCl) in the absence and the presence of dianion
activators were determined by monitoring the decrease in
absorbance at 340 nm. Figure 4 shows the dependence of vi/[E]
vi
[E]
{(kcat)EKX + (kcat)E·X[X2−]}[GA]
=
[GA][X2−] + KGA[X2−] + KX[GA] + KGAKX
(5)
DISCUSSION
■
We have reported kinetic parameters for activation of yeast
2−
OMPDC-catalyzed decarboxylation of EO (Figure 1) by SO3
,
7
HPO32−, S2O32−, SO4 and HOPO32−. We did not examine
activation of ScTIM and hlGPDH by SO32−, which is a strong
sulfur nucleophile that will form an adduct to the carbonyl carbon
of truncated substrate GA,51−53 and added instead fluorophos-
phate dianion (FPO32−) to our series of oxydianions activators.
2−
2−
We examined HPO3 activation of rabbit muscle GPDH-
catalyzed reduction of GA by NADH,16 but use human liver
enzyme (hlGPDH) in this study. hlGPDH shows 4-fold larger
values for kcat/Km and (kcat/Km)E·X/KX for the catalyzed reactions
of DHAP and the pieces GA + HPO32−, respectively, compared
with rabbit muscle GPDH.16 Linear plots of vi/[E] (s−1) against
[HPO32−] were determined for rabbit muscle GPDH that are
consistent with KX > 100 mM for HPO32− (Scheme 3).16 This is
greater than KX = 70 mM determined for hlGPDH (Table 2).
The activation of hlGPDH-catalyzed reduction of GA by
Figure 4. Dependence of vi/[E] (s−1) on the concentration of
glycolaldehyde (GA), for the reduction by NADH (0.2 mM) catalyzed
by hlGPDH at pH 7.5 (10 mM TEA), 25 °C and I = 0.12 (NaCl).
(s−1) on the concentration of the reactive carbonyl form of GA
(fcar = 0.06)12 for the unactivated reduction of GA by NADH
catalyzed by 0.05 mM hlGPDH. The fit of these data to the
−
oxydianions (X2 ) was studied at saturating [NADH] = 0.2 mM,
−
Michaelis−Menten equation gave values of KGA = 6.2 mM, kcat
=
and proceeds through the quaternary E·NADH·GA·X2
3.1 × 10−4 s−1, and kcat/KGA = 0.05 M−1 s−1.
complex. The kinetic data were fit to Scheme 3, which neglects
steps for release of NAD+ and binding of NADH. These steps are
fast relative to kcat ≤ 10 s−1 (Table 2) for turnover of E·NADH·
Figure 5A shows the dependence of vi/[E] (s−1) on [HPO3
]
2−
for the reduction of GA by NADH (saturating) catalyzed by
hlGPDH, for reactions carried out at different fixed concen-
trations of GA. The solid lines show the nonlinear least-squares
fit of these data to eq 5, derived for Scheme 3, using the
appropriate kinetic parameters reported in Table 2. The value of
(kcat/Km)E·X/KX = 16 000 M−2 s−1 determined for phosphite
−
GA·X2 , because they support the even faster turnover of the
ternary E·NADH·DHAP complex, for which kcat = 240 s−1.
Scheme 3 shows a random order for binding of GA and X2− to E·
NADH, and a single dissociation constant KGA or KX for release
of the ligands from E·NADH·GA·X2− and from the correspond-
1376
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
Figure 5. Dependence of vi/[E] (s−1) for the reduction of GA by NADH (0.2 mM) catalyzed by hlGPDH at pH 7.5 (10 mM TEA), 25 °C and I = 0.12
2−
●
▲
○
(NaCl) at increasing concentrations of dianion for reactions at different fixed concentrations of GA. Key: (A) HPO3 : ( ), 3.6 mM GA; ( ); 3.0 mM;
2−
■
◆
■
● ○
), 2.4 mM; (Δ), 1.8 mM; ( ), 1.2 mM; ( ), 0.6 mM; ( ); 0.3 mM. (B) FPO3 : ( ), 3.6 mM GA; ( ); 3.0 mM; ( ), 2.4 mM; (Δ), 1.8 mM; ( ),
2− 2−
▲
▼
(
◆
■
◆
▼
●
▲
▼
●
1.2 mM; ( ), 0.6 mM; ( ); 0.3 mM. (C) SO4 : ( ), 3.6 mM GA; ( ), 2.4 mM ; ( ), 1.2 mM; ( ), 0.6 mM; ( ); 0.3 mM. (D) HPO4 : ( ), 3.6
2−
■
■
▲
▼
●
▲
▼
mM GA; ( ), 2.4 mM ; ( ), 1.2 mM; ( ), 0.6 mM. (E) S2O3 : ( ), 3.6 mM GA; ( ), 2.4 mM ; ( ), 1.2 mM; ( ), 0.6 mM.
interactions between GA and X2− at the quaternary complex E·
Scheme 3
NADH·GA·X2− do not result in a large change in ligand affinity
−
compared to the ternary E·NADH·X2 and E·NADH·GA
complexes.
Reactivity of Enzyme-Bound Whole Substrates and the
Substrate Pieces. Enzyme activation by phosphite dianion for
catalysis of the reactions of truncated substrate suggests similar
turnover numbers for the reactions of enzyme-bound whole
substrate and the corresponding substrate pieces. This is
supported, at least in part, by data from Table 3 which compares
kinetic parameters for the reactions of whole substrates catalyzed
by ScOMPDC (OMP), ScTIM (GAP) and hlGPDH (DHAP),
with kinetic parameters for the catalyzed reactions of substrate
pieces. The turnover numbers k′cat for complexes of the substrate
pieces (Scheme 4) were calculated from values (kcat/Km)E·HPi and
Kd, using the relationship k′cat = [(kcat/Km)E·HPi]Kd. We do not
ing E·NADH·GA or E·NADH·X2− complexes. The fits of kinetic
2−
data for activation by HPO3 to eq 5, derived for Scheme 3
(Figure 5A), give values of KGA = 4.9 mM and KX = 70 mM that
are similar, respectively, to KGA = 6.2 mM for the hlGPDH-
catalyzed reaction in the absence of activator (Figure 4) and KX =
2−
42 mM for competitive inhibition by HPO3 of hlGPDH-
catalyzed reduction of DHAP. These results show that the
Table 2. Kinetic Parameters for Activation of hlGPDH by Oxydianions (Scheme 3) and Derived Parameters for the Binding of
a
Dianions to [E·S]‡ (Scheme 5)
b
c
d
e
f
dianion
none
(kcat
)
E·X (s−1
)
KGA (mM)
6.2
KX (mM)
(kcat/Km)E·X/KX (M−2 s−1
)
(K‡)X (M)
RT ln(K‡)X (kcal/mol)
(3.1 0.3) × 10−4
5.5 0.3
1
g
g
3.3 × 10−6
6.5 × 10−7
2.5 × 10−4
4.6 × 10−5
1.5 × 10−3
−7.5
−8.4
−4.9
−5.9
−3.8
2−
HPO3
4.9 0.2 [8]
5.0 0.3
4.1 0.2
4.5 0.4
5.0 0.2
70
17
40
70
4
1
3
9
16000 1300 [4300 700]
75000 6000
2−
FPO3
6.4 0.25
2
HOPO3
0.032 0.002
0.360 0.040
0.0190 0.002
200 20
2−
SO4
1100 200
2−
S2O3
110 10
35
5
a
Reactions catalyzed by hlGPDH at pH 7.5 (10 mM TEA), 25 °C, 0.2 mM NADH, and I = 0.12 (NaCl). The quoted uncertainty in these kinetic
b
parameters is the standard error determined for the nonlinear least-squares fits of these data. First-order rate constant for turnover of the Michaelis
c
complex to form product (Scheme 3). Dissociation constant for release of GA from the binary or ternary enzyme complex (Scheme 3).
d
e
Dissociation constant for release of the oxydianion from the binary or ternary enzyme complex (Scheme 3). Dissociation constant for release of
f
g
the dianion from the transition state complex, calculated using eq 6, derived for Scheme 5. Intrinsic Gibbs dianion binding free energy. Kinetic
parameter for GPDH from rabbit muscle.16
1377
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
a
Table 3. Rate Constants for Turnover of Whole Substrates and the Substrate Pieces Catalyzed by ScOMPDC, ScTIM and hlGPDH
b
b
c
c
c
d
enzyme
kcat (s−1
)
(kcat/Km) (M−1 s−1
1.1 × 107
)
(kcat/Km)E (M−1 s−1
)
(kcat/Km)E·HPi (M−1 s−1
)
Kd (M)
≈ 0.1
k′cat (s−1
)
e
f
ScOMPDC
ScTIM
16
0.026
1600
≈ 160
g
g
h
h
8900
2.2 × 108
0.062
48
i
i
j
hlGPDH
240
4.6 × 106
0.050
1100
0.005
5.5
a
Enzyme-catalyzed reactions of the following whole substrates and substrate pieces: OMPDC; OMP and EO + HPO32−; ScTIM; GAP and GA +
b
c
HPO32−; hlGPDH; DHAP and GA + HPO3
. Kinetic parameters for turnover of the whole substrate. Kinetic parameter for turnover of the
2−
d
substrate pieces (Schemes 1−3). Rate constant for turnover of the complex between the enzyme and substrate pieces (Scheme 4), estimated as k′cat
= [(kcat/Km)E·HPi]Kd. Published kinetic parameters.10 Calculated using (kcat/Km)E·HPi = 1600 M−1 s−1 and Kd ≈ 0.1 M.10 Kinetic parameters for
e
f
g
ScTIM-catalyzed isomerization of GAP.54 Table 1. Table 2. Calculated from (kcat/Km)E·X/KX = 16000 M−2 and KX = 0.070 M (Table 2).
h
i
j
small fraction of GPDH exists as the productive complex to the
carbonyl form of GA. Nonproductive binding would result in
balancing decreases in Kd and k′cat (Scheme 4), compared to the
values for the reactive carbonyl substrate. By contrast, the
accumulation of a nonproductive hydrate complex does not
affect (kcat/Km)E·HPi determined at low [GA], where the hlGPDH
exists mainly in the unliganded form.55
Scheme 4
Dianion Activation Binding Site. The ratios of the third-
order rate constant (kcat/Km)E·X/KX for catalysis of the reaction of
the substrate pieces and the second-order rate constant (kcat/
Km)E for the truncated substrate (Table 3) range from 4.4 × 105
M−1 for yeast OMPDC-catalyzed decarboxylation to 4.4 × 104
M−1 for ScTIM-catalyzed isomerization and correspond to 7.7
and 6.3 kcal/mol intrinsic phosphite dianion binding energies (see
below).
This shows that efficient catalysis by OMPDC, ScTIM and
hlGPDH is achieved by combining a catalytic site that provides
sufficient stabilization of the transition state for reaction of the
truncated substrate to give (kcat/Km)E ≈ 0.05 M−1 s−1 and a
dianion activation site that boosts enzyme activity toward
perfection, by providing 11−13 kcal/mol transition state
stabilization from the phosphodianion binding Gibb free
energy.56
These large intrinsic dianion binding energies may be
rationalized using representations of X-ray crystal structures
shown in Figure 6.57−59 Two types of interactions are shown,
which stabilize these enzyme−phosphodianion complexes. (i)
An ionic interaction between the phosphodianion and a cationic
side chain (Arg235 for OMPDC, Arg269 for GPDH and Lys12
for ScTIM). (ii) Networks of hydrogen bonds to either backbone
report a value of (kcat)′ for the ScTIM-catalyzed reaction, because
we are unable to approach saturation of this enzyme by GA.
Table 3 shows that k′cat ≈ 160 s−1 for turnover of EO by
OMPDC is significantly larger than kcat = 16 s−1 for turnover of
OMP, as noted in earlier work.10 However, k′cat ≈ 5.5 s−1 for
turnover of GA by hlGPDH is smaller than kcat = 240 s−1 for
turnover of DHAP. It is interesting that the second order rate
constants (kcat/Km)E·HPi for OMPDC (1600 M−1 s−1) and
hlGPDH (1100 M−1 s−1) catalyzed reactions of their respective
substrate pieces are similar, but that k′cat ≈ 160 s−1 for OMPDC is
much larger than k′cat = 5.5 s−1 for GPDH-catalyzed turnover of
the respective truncated substrates, because this requires a
surprising 20-fold larger affinity of the minimal two-carbon
substrate (Kd = 0.005 M) for the hlGPDH-catalyzed reaction
compared with Kd = 0.1 M for the more functionalized substrate
EO.
GA exists mainly as the hydrate (94%),50 and the
interpretation of these data depends upon whether the
nonproductive GPDH·hydrate complex accumulates. We
suggest that Kd = 0.005 M for GA reflects mainly the formation
of the nonproductive hydrate complex and that the observed
kinetic parameter k′cat = 5.5 s−1 underestimates the turnover
number for the [E·NADH·GA·HPi] complex because only a
Figure 6. Representations of X-ray crystal structures of ScOMPDC, hlGPDH and ScTIM. (A) ScOMPDC in a complex with the intermediate analogue
6-hydroxyuridine 5′-monophosphate (PDB entry 1DQX).59 The interactions of Gln215, Tyr217 and Arg235 side chains and the Gly234 and Arg235
backbone amides with the ligand phosphodianion are shown. Reprinted with permission from ref 46. Copyright 2012 American Chemical Society. (B) A
nonproductive ternary complex of human liver hlGPDH with DHAP and NAD (PDB entry 1WPQ).57 The interactions of Arg269 and Asn270 side
chains and the Arg269 and Asn270 backbone amides with the ligand phosphodianion are shown. (C) ScTIM in a complex with the intermediate
analogue 2-phosphoglycolate (PDB entry 2YPI).58 The interactions of the K12 side chain and the Gly173, Ser213, Gly234 and Gly 235 backbone amides
are shown.
1378
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
amides, amide side chains, and in the case of ScOMPDC the
phenol side chain of Tyr217.
connection energy (ΔGS)27 arising from the covalent attachment
of the pieces bound to ScTIM (ΔGS = 13.0−6.3 = 6.7 kcal/mol)
compared with ScOMPDC (ΔGS = 11.7−7.7 = 4.0 kcal/mol) or
hlGPDH (ΔGS = 10.7−7.5 = 3.2 kcal/mol). This observation
that decreases in the total dianion binding energy are
accompanied by increases in the total phosphite dianion binding
energy suggests that the total dianion binding energy might be
sacrificed to optimize the activating dianion interactions. For
example, destabilizing steric interactions between the enzyme
and the phosphodianion of the whole substrate at a restricted
binding pocket, which are expressed in substrate binding (Km)
but not dianion activation (kcat), might be smaller or absent at the
enzyme complex to phosphite dianion and the truncated
substrate pieces.
There is no obvious correlation between oxydianion structure
and dissociation constants KX for release of oxydianions from
OMPDC, TIM and GPDH. By contrast, the strong enzyme
activation by oxydianions requires a high specificity for their
binding to the activated complexes [E·S]‡ [(K‡)X ≪ KX, Scheme
5]. We note the following trends in the values of ΔG‡ for binding
of oxydianions to the respective [E·S]‡ complexes for reactions
catalyzed by ScOMPDC, hlGPDH and ScTIM.
The effects of the K12G mutation of ScTIM (550 000 fold)
and the R235A mutation (18 000-fold) of OMPDC on kcat/Km
correspond to 7.8 and 5.8 kcal/mol stabilizing interactions
between the protein side chain and the transition states for the
respective enzyme-catalyzed reactions of GAP and OMP.54,60
The stronger transition state stabilization by Lys12 compared to
Arg235 reflects the interaction of the former side chain with both
the phosphodianion and the negatively charged enolate
oxygen,54 compared with the essentially exclusive interaction of
the side chain of Arg235 with the substrate phosphodianion.4,61
Each of the cationic side chains, which interact with substrate
phosphodianion (Figure 6), sit on the protein surface and shield
the phosphodianion from interaction with bulk solvent. This
positioning enables effective rescue of the R235A and K12G
mutants by transfer of exogenous guanidinium60 and alkylam-
monium cations,62 respectively, from water to the cleft at the
protein created by the mutation. We suggest that the R269A
mutation of hlGPDH will, likewise, result in a large falloff in
catalytic activity, which will be rescued by exogenous
guanidinium cation.
The effects of all combinations of single, double and triple
Q215A, Y217F and R235A mutations on the kinetic parameters
for ScOMPDC-catalyzed reactions of whole substrate and
substrate pieces have been determined.4 The results show that
the total 12-kcal/mol intrinsic Gibbs phosphodianion binding
free energy for OMPDC (Table 3) is equal to the sum of the
phosphodianion binding energies of the side chains of R235 (6
kcal/mol), Q215 (2 kcal/mol) Y217 (2 kcal/mol) and hydrogen
bonds to the G234 and R235 backbone amides (2 kcal/mol). We
have not quantified the contribution of interactions between the
substrate phosphodianion and the backbone amides of TIM to
the intrinsic phosphodianion binding energy. However,
stabilizing interactions of ca. 1−2 kcal/mol/amide, along with
the 7.8 kcal/mol interaction of K12, would be sufficient to
account for the 13 kcal/mol dianion binding energy.
Specificity in the Binding of Oxydianions. Values of
(K‡)X for breakdown of oxydianion complexes to the transition
states for ScOMPDC-, ScTIM- and hlGPDH-catalyzed reactions
([E·S·X2−]‡) were determined using eq 6 derived for Scheme 5.
The transition state binding energies ΔG‡ were calculated as
ΔG‡ = RT ln(K‡)X and are summarized in Chart 1. The intrinsic
Scheme 5
(1) The largest oxydianion binding energies are observed
when two negative charges are delocalized over the three oxygen
2−
of HPO3 or FPO32−. The delocalization of negative charge
2−
2−
onto the fourth heteroatom of SO4 and S2O3 results in a
reduction in the Gibbs oxydianion binding free energy for
ScOMPDC-, hlGPDH- and ScTIM-catalyzed reactions (Chart
1). The difference between the −5.8 and −4.5 kcal dianion
binding energy of S2O3 and SO4 observed for TIM may
reflect the small delocalization of charge onto the peripheral
sulfur of S2O32−, which results in a larger negative charge density
at the three oxygen.63 The generally smaller dianion binding
2−
2−
2−
2−
energy for HOPO3 compared with FPO3 may reflect
differences in the steric and electronic properties of −OH and
−F.
Chart 1
Kx(kcat/Km)E
(K‡)X =
(kcat/Km)E·X
(6)
(2) There are no interactions between ScOMPDC (Figure
6A) or ScTIM (Figure 6C) and the bridging phosphate oxygen of
the enzyme-bound ligand, but the bridging phosphate oxygen of
DHAP bound to hlGPDH forms a hydrogen bond to the cationic
side chain of Arg-269 (Figure 6B). We therefore propose that the
unusually large −8.4 kcal/mol and small −3.8 kcal/mol Gibbs
intrinsic dianion binding free energy, respectively, for activation
Gibbs phosphodianion binding free energy for ScTIM (13.0
kcal/mol, Table 3) is larger than for ScOMPDC (11.7 kcal/mol)
or hlGPDH (10.7 kcal/mol). By contrast, the intrinsic phosphite
dianion binding energy (Chart 1) for ScTIM (−6.3 kcal/mol) is
smaller than for OMPDC (−7.7 kcal/mol) or hlGPDH (−7.5
kcal), so that the latter two enzymes show the broader range of
dianion binding energies: 3.0−7.7 kcal/mol for ScOMPDC and
3.8−8.4 kcal/mol for hlGPDH, compared with 4.3−6.3 kcal/mol
for ScTIM (Chart 1). These differences reflect the larger
2−
2−
of hlGPDH by FPO3 and S2O3 (Chart 1) compared with
activation of ScTIM and hlGPDH is due to stabilizing
electrostatic interactions between the side chain of Arg-269
2−
and -F of FPO3 and destabilizing steric/electrostatic
2−
interactions between this side chain and an −S of S2O3
,
where -F and -S are bound at a position equivalent to that for the
bridging phosphate oxygen of DHAP.
1379
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
Mechanism for Enzyme Activation. Scheme 6 provides a
rationalization for phosphite dianion activation of TIM,
binding of either an oxydianion, or the transition state for the
enzymatic reaction.4−7,11,24,25,46,47,61,64−66 This Scheme and
Figure are representations of Koshland’s proposal that enzymes
exist in an inactive resting form, where the catalytic side chains
are poorly positioned to carry out their assigned functions. The
ligand binding energy is then utilized to induce a large change in
enzyme structure, which shifts these active site residues to their
catalytically active positions.28,29
Scheme 6
The many cases for which substrate Gibbs binding free energy
is utilized to induce an optimal substrate fit at enzyme active sites
confirm Koshland’s induced fit model. However, the significance
of the term “induced fit” has been reduced and its use in the
biochemical literature limited, by the lack of a generally accepted
description of the imperatives for catalysis by enzymes that exist
mainly in an inactive open form and which rely upon the
substrate binding energy to drive extensive changes in enzyme
structure.56,67 Why not the alternative, where the enzyme folds
into its active conformation and the intrinsic substrate binding
energy is used for other purposes?56,67 We propose the following
imperatives for the utilization of substrate binding energy to drive
thermodynamically unfavorable changes in enzyme conforma-
tion.
OMPDC and GPDH for catalysis of reactions of the respective
truncated substrates. These enzymes exist mainly in an open
resting form (EO), and undergo a dianion-driven conformational
change to a closed form (EC), which includes closure of a loop
over the dianion. We have proposed that this conformational
change is thermodynamically unfavorable (KC
≪
1).4−7,11,24,25,46,47,61,64,65 Part or all of energetic cost for the
enzyme conformational change is paid in the Gibbs binding free
energy of the first piece, either S‡ or X2−, as shown in Figure 7. A
(1) Wolfenden noted that substrate binding will necessarily
only occur to a solvent exposed active site, while a caged
Michaelis complex may provide for a larger number of enzyme−
ligand interactions.68 This provides a rational for conformational
changes that create a caged complex, after initial substrate
binding to the open forms of TIM, OMPDC and GPDH.6
(2) Removal of solvent from an exposed cavity during loop
closure promotes effective catalysis of polar reactions.6,69 The
energetic cost for desolvation of enzyme active sites is included in
the total cost of the enzyme conformational change, and may be
paid for by the utilization of Gibbs dianion binding free energy.24
Closure of the flexible phosphodianion gripper loop 6 of TIM
from Trypanosoma brucei brucei (TbbTIM) over the bound
substrate extrudes several water molecules to the bulk
solvent,70,71 sandwiches the carboxylate anion side chain of
Glu-167 (the active site base for TIM) between the hydrophobic
side-chains of Ile-172 and Leu-232,72 and shields the carboxylate
anion from interactions with the aqueous solvent. Loosening this
hydrophobic clamp by the I172A mutation of TbbTIM results in
a 100-fold falloff in kcat/Km for isomerization of GAP47 and a ≈2
unit decrease in the pKa for the catalytic side chain at the
enzyme−phosphoglycolate complex.73 These results are con-
sistent with the proposal that the hydrophobic side chain of Ile-
172 plays a role in effecting a strong basicity for the side chain
that acts as the Brønsted base to deprotonate the bound carbon
acid.73,74
Figure 7. Gibbs free energy diagram that shows enzyme-catalyzed
turnover of truncated substrate S by EO, and by an enzyme−oxydianion
complex (EC·X2−). The tighter binding of the oxydianion to the
transition state complex [EC·S]‡ to give [EC·X2−·S‡] [ΔGint = −RT ln(1/
Kx‡)] compared to the free enzyme EO to give EC·X2− [ΔGobs = −RT
ln(1/Kx)] is attributed to ΔGC = −RT ln(KC) for the unfavorable
conformational change that converts inactive open EO to active closed
EC. The binding energy of oxydianions is utilized to reduce, or eliminate
entirely, the barrier to ΔGC for the enzyme conformational change.
(3) There is an unclear entropic cost, paid for by the utilization
of Gibbs dianion binding free energy, to freezing motions of
flexible enzyme loops and of active site catalytic side chains as the
protein conformation changes from inactive EO to active EC.
(4) Relatively large barriers to enzyme conformational
changes may exist simply to avoid the expression of Gibbs
dianion binding free energy at the Michaelis complex and the
resulting effectively irreversible ligand binding.56 The barrier to
this unfavorable conformational change would not be strongly
expressed in the total activation barrier for the reaction under
kcat/Km conditions, so long as kcat/Km is close to the diffusion-
controlled limit for a “perfect” enzyme and conversion of the
first-formed EO·S complex to the active EC·S complex and then to
product is faster than dissociation of S from EO·S.65 In such cases,
changes in KC for the enzyme conformational change will result
larger ligand binding energy and specificity is then expressed as
tight binding of the second piece to the respective binary
complexes; either the binding of S‡ to EC·X2− or of X2− to [EC·S]‡
to form the ternary [EC·S·X2−]‡ complex (Figure 7). The small
dependence of KX on dianion structure (Tables 1, 2 and ref 7)
suggests that the specific enzyme−dianion interactions only
develop at the EC·X2− complex, and that the complexes of most
or all oxydianions to the free enzymes are nonproductive and
exist mainly in the open (EO·X2−) form.
A Role for Ligand Induced Fits in Enzyme Catalysis.
Scheme 6 and Figure 7 for catalysis by ScTIM, ScOMPDC,
hlGPDH, and other enzymes that are activated by dianions, show
two conformations: dominant inactive open EO and a proposed
high Gibbs free energy EC, whose formation is induced by
1380
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
in compensating changes in kcat and Km, but little or no change in
kcat/Km.
REFERENCES
■
(1) Herschlag, D.; Natarajan, A. Biochemistry 2013, 52, 2050−2067.
(2) Fersht, A. R. Biochemistry 1988, 27, 1577−1580.
(3) Fersht, A. R. Biochemistry 1987, 26, 8031−8037.
(4) Goldman, L. M.; Amyes, T. L.; Goryanova, B.; Gerlt, J. A.; Richard,
J. P. J. Am. Chem. Soc. 2014, 136, 10156−10165.
(5) Zhai, X.; Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc. 2014, 136,
4145−4148.
We previously reported that the L232A mutation of TbbTIM
results in a 17-fold increase in the second-order rate constant
(kcat/Km)E for TIM-catalyzed proton transfer reactions of the
truncated substrate piece [1-13C]-GA in D2O (Scheme 2).65 We
have proposed that the hydrophobic side chain of L232 functions
to cause an increase the barrier to the protein conformational
change ΔGC (Figure 7), that reduces the substrate binding
energy expressed at the Michaelis complex.47,65 The L232A
mutation then results in an increase in the fraction of enzyme
present in the active closed form (increase in ΔGC), and in
similar increases in the values of kinetic parameters that depend
upon the magnitude of ΔGC: (kcat/Km)E, KX and (kcat/Km)E·X/KX
(Figure 7)47,65
(6) Richard, J. P.; Amyes, T. L.; Goryanova, B.; Zhai, X. Curr. Opin.
Chem. Biol. 2014, 21, 1−10.
(7) Spong, K.; Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc. 2013, 135,
18343−18346.
(8) Zhai, X.; Amyes, T. L.; Wierenga, R. K.; Loria, J. P.; Richard, J. P.
Biochemistry 2013, 52, 5928−5940.
(9) Zhai, X.; Go, M. K.; Donoghue, A. C.; Amyes, T. L.; Pegan, S. D.;
Wang, Y.; Loria, J. P.; Mesecar, A. D.; Richard, J. P. Biochemistry 2014,
53, 3486−3501.
The Role of Conformational Changes in Enzyme
Catalysis. There are good reasons to refer to enzyme structures
as plastic, in recognition of the large changes in structure
observed during the catalytic cycle.9,75−78 There is an ongoing
debate about the role of these conformational changes in
catalysis.79 On the one hand, protein motions may be coupled to
motion along the reaction coordinate, and might even contribute
to the catalytic rate acceleration by promoting formation of the
transition state.80−83 However, a strong case can be made that
effective enzymatic catalysis is due to the preorganization of
catalytic side chains into their catalytically active conformations,
which optimizes their stabilizing interactions with the transition
state.69,84,85 The far from definitive experimental evidence
offered in support of the coupling of protein motions to catalysis
has been criticized.86 Our model to rationalize oxydianion
activation of the reactions catalyzed by TIM, OMPDC and
GPDH emphasizes the large barrier to the conformational
changes, which convert the inactive open enzyme to the active
closed form: a part of this barrier represents the requirement for
organization of the catalytic side chains at a caged active site
complex. We do not exclude the possibility that enzyme−dianion
interactions are first utilized to lock the Michaelis complex in the
active conformation EC, and that this is followed by coupled
motions of the protein and substrate on proceeding to the
transition state. However, this added layer of complexity is not
required to rationalize our experimental data. An examination of
Figure 6 and related structures for enzyme−ligand complexes
emphasize the enormity of the possible stabilizing interactions
between enzymes and transition states. We suggest that the
formal contribution to catalysis (if any) of coupling protein and
reaction coordinate motions is of incidental importance
compared to the large transition state stabilization obtained
from strong protein−ligand interactions.
(10) Amyes, T. L.; Richard, J. P.; Tait, J. J. J. Am. Chem. Soc. 2005, 127,
15708−15709.
(11) Go, M. K.; Amyes, T. L.; Richard, J. P. Biochemistry 2009, 48,
5769−5778.
(12) Amyes, T. L.; Richard, J. P. Biochemistry 2007, 46, 5841−5854.
(13) Richard, J. P.; Zhai, X.; Malabanan, M. M. Bioorg. Chem. 2014, 57,
206−212.
(14) Goryanova, B.; Amyes, T. L.; Gerlt, J. A.; Richard, J. P. J. Am.
Chem. Soc. 2011, 133, 6545−6548.
(15) Toth, K.; Amyes, T. L.; Wood, B. M.; Chan, K. K.; Gerlt, J. A.;
Richard, J. P. Biochemistry 2009, 48, 8006−8013.
(16) Tsang, W.-Y.; Amyes, T. L.; Richard, J. P. Biochemistry 2008, 47,
4575−4582.
(17) Ray, W. J., Jr.; Long, J. W.; Owens, J. D. Biochemistry 1976, 15,
4006−4017.
(18) Auerbach, G.; Huber, R.; Grattinger, M.; Zaiss, K.; Schurig, H.;
̈
Jaenicke, R.; Jacob, U. Structure 1997, 5, 1475−1483.
(19) Tadwal, V. S.; Sundararaman, L.; Manimekalai, M. S. S.; Hunke,
C.; Grueber, G. J. Struct. Biol. 2012, 180, 509−518.
(20) Yamaguchi, H.; Miwa, Y.; Kasa, M.; Kitano, K.; Amano, M.;
Kaibuchi, K.; Hakoshima, T. J. Biochem. 2006, 140, 305−311.
(21) Li, G.; Liang, Z. Biochem. J. 2001, 355, 681−689.
(22) Chan, K.; Fedorov, A. A.; Fedorov, E. V.; Almo, S. C.; Gerlt, J. A.
Biochemistry 2008, 47, 9608−9617.
(23) Denesyuk, A. I.; Denessiouk, K. A.; Korpela, T.; Johnson, M. S.
Biochim. Biophys. Acta 2003, 1647, 234−238.
(24) Amyes, T. L.; Richard, J. P. Biochemistry 2013, 52, 2021−2035.
(25) Malabanan, M. M.; Amyes, T. L.; Richard, J. P. Curr. Opin. Struct.
Biol. 2010, 20, 702−710.
(26) Morrow, J. R.; Amyes, T. L.; Richard, J. P. Acc. Chem. Res. 2008,
41, 539−548.
(27) Jencks, W. P. Proc. Nat. Acad. Sci. U. S. A. 1981, 78, 4046−4050.
(28) Thomas, J. A.; Koshland, D. E., Jr. J. Am. Chem. Soc. 1960, 82,
3329−3333.
(29) Koshland, D. E., Jr. Proc. Natl. Acad. Sci. U. S. A. 1958, 44, 98−104.
(30) McMillan, A. W.; Lopez, M. S.; Zhu, M.; Morse, B. C.; Yeo, I.-C.;
Amos, J.; Hull, K.; Romo, D.; Glasner, M. E. Biochemistry 2014, 53,
4434−4444.
(31) Nagar, M.; Narmandakh, A.; Khalak, Y.; Bearne, S. L. Biochemistry
2011, 50, 8846−8852.
(32) Torres, R.; Lan, B.; Latif, Y.; Chim, N.; Goulding, C. W. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 2014, 70, 1074−1085.
(33) Axe, J. M.; Boehr, D. D. J. Mol. Biol. 2013, 425, 1527−1545.
(34) Porter, C. M.; Miller, B. G. Bioorg. Chem. 2012, 43, 44−50.
(35) Johnson, T. A.; Holyoak, T. Biochemistry 2012, 51, 9547−9559.
(36) Carpenter, R. A.; Xiong, J.; Robbins, J. M.; Ellis, H. R. Biochemistry
2011, 50, 6469−6477.
(37) Wood, B. M.; Chan, K. K.; Amyes, T. L.; Richard, J. P.; Gerlt, J. A.
Biochemistry 2009, 48, 5510−5517.
(38) Bergemeyer, H. U.; Haid, E.; Nelboeck-Hochstetter, M. US
Patent US3662037 A,1972.
AUTHOR INFORMATION
■
Corresponding Author
jrichard@buffalo.edu
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We acknowledge the National Institutes of Health Grant
GM39754 for generous support of this work. We thank Prof.
Steve Withers for the suggestion to examine a fluorophosphate
dianion activator.
1381
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382

Journal of the American Chemical Society
Article
(39) Amyes, T. L.; Wood, B. M.; Chan, K.; Gerlt, J. A.; Richard, J. P. J.
Am. Chem. Soc. 2008, 130, 1574−1575.
(77) Selvaraj, M.; Roy, S.; Singh, N. S.; Sangeetha, R.; Varshney, U.;
Vijayan, M. J. Mol. Biol. 2007, 372, 186−193.
(78) Ingvarsson, H.; Mate, M. J.; Hoegbom, M.; Portnoi, D.;
Benaroudj, N.; Alzari, P. M.; Ortiz-Lombardia, M.; Unge, T. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 2007, 63, 249−259.
(79) Richard, J. P. Biochemistry 2013, 52, 2009−2011.
(80) Francis, K.; Kohen, A. Curr. Opin. Chem. Biol. 2014, 21, 19−24.
(81) Nashine, V. C.; Hammes-Schiffer, S.; Benkovic, S. J. Curr. Opin.
Chem. Biol. 2010, 14, 644−651.
(40) Barnett, S. A.; Amyes, T. L.; Wood, B. M.; Gerlt, J. A.; Richard, J.
P. Biochemistry 2008, 47, 7785−7787.
(41) Protein Identification and Analysis Tools on the ExPASy Server;
Gasteiger, E., Hoogland, C., Gattiker, A., Duvaud, S., Wilkins, M. R.,
Appel, R. D., Bairoch, A., Eds.; Humana Press, Inc.: Totowa, NJ, 2005.
(42) Gasteiger, E.; Gattiker, A.; Hoogland, C.; Ivanyi, I.; Appel, R. D.;
Bairoch, A. Nucleic Acids Res. 2003, 31, 3784−3788.
(43) Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc. 1996, 118, 3129−
3141.
(44) Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc. 1992, 114, 10297−
10302.
(45) O’Donoghue, A. C.; Amyes, T. L.; Richard, J. P. Biochemistry
2005, 44, 2610−2621.
(46) Amyes, T. L.; Ming, S. A.; Goldman, L. M.; Wood, B. M.; Desai, B.
J.; Gerlt, J. A.; Richard, J. P. Biochemistry 2012, 51, 4630−4632.
(47) Malabanan, M. M.; Koudelka, A. P.; Amyes, T. L.; Richard, J. P. J.
Am. Chem. Soc. 2012, 134, 10286−10298.
(48) Malabanan, M. M.; Go, M. K.; Amyes, T. L.; Richard, J. P.
Biochemistry 2011, 50, 5767−5779.
(49) Black, W. J. Can. J. Biochem. Phys. 1966, 44, 1301−1317.
(50) Reynolds, S. J.; Yates, D. W.; Pogson, C. I. Biochem. J. 1971, 122,
285−297.
(82) Roca, M.; Oliva, M.; Castillo, R.; Moliner, V.; Tunon, I. Chem.
Eur. J. 2010, 16, 11399−11411.
(83) Saen-Oon, S.; Ghanem, M.; Schramm, V. L.; Schwartz, S. D.
Biophys. J. 2008, 94, 4078−4088.
(84) Cannon, W. R.; Benkovic, S. J. J. Biol. Chem. 1998, 273, 26257−
26260.
(85) Warshel, A.; Florian, J.; Strajbl, M.; Villa, J. ChemBioChem 2001, 2,
109−111.
(86) Adamczyk, A. J.; Cao, J.; Kamerlin, S. C. L.; Warshel, A. Proc. Natl.
Acad. Sci. U. S. A. 2011, 108, 14115−14120.
(51) Maier, K.; Hinze, H.; Leuschel, L. Biochim. Biophys. Acta 1986,
848, 120−130.
(52) Young, P. R.; Jencks, W. P. J. Am. Chem. Soc. 1978, 100, 1228−
1235.
(53) Young, P. R.; Jencks, W. P. J. Am. Chem. Soc. 1977, 99, 1206−
1214.
(54) Go, M. K.; Koudelka, A.; Amyes, T. L.; Richard, J. P. Biochemistry
2010, 49, 5377−5389.
(55) Jencks, W. P. Catalysis in Chemistry and Enzymology; McGraw
Hill: New York, 1969.
(56) Jencks, W. P. Adv. Enzymol. Relat. Areas Mol. Biol. 1975, 43, 219−
410.
(57) Ou, X.; Ji, C.; Han, X.; Zhao, X.; Li, X.; Mao, Y.; Wong, L.-L.;
Bartlam, M.; Rao, Z. J. Mol. Biol. 2006, 357, 858−869.
(58) Lolis, E.; Petsko, G. A. Biochemistry 1990, 29, 6619−6625.
(59) Miller, B. G.; Hassell, A. M.; Wolfenden, R.; Milburn, M. V.; Short,
S. A. Proc. Nat. Acad. Sci. U. S. A. 2000, 97, 2011−2016.
(60) Barnett, S. A.; Amyes, T. L.; McKay Wood, B.; Gerlt, J. A.;
Richard, J. P. Biochemistry 2010, 49, 824−826.
(61) Goryanova, B.; Goldman, L. M.; Amyes, T. L.; Gerlt, J. A.;
Richard, J. P. Biochemistry 2013, 52, 7500−7511.
(62) Go, M. K.; Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc. 2010, 132,
13525−13532.
(63) Frey, P. A.; Sammons, R. D. Science 1985, 228, 541−545.
(64) Goryanova, B.; Spong, K.; Amyes, T. L.; Richard, J. P. Biochemistry
2013, 52, 537−546.
(65) Malabanan, M. M.; Amyes, T. L.; Richard, J. P. J. Am. Chem. Soc.
2011, 133, 16428−16431.
(66) Kholodar, S. A.; Murkin, A. S. Biochemistry 2013, 52, 2302−2308.
(67) Herschlag, D. Bioorg. Chem. 1988, 16, 62−96.
(68) Wolfenden, R. Mol. Cell. Biochem. 1974, 3, 207−211.
(69) Warshel, A. J. Biol. Chem. 1998, 273, 27035−27038.
(70) Wierenga, R. K.; Noble, M. E. M.; Vriend, G.; Nauche, S.; Hol, W.
G. J. J. Mol. Biol. 1991, 220, 995−1015.
(71) Alahuhta, M.; Wierenga, R. K. Proteins: Struct., Funct., Bioinf. 2010,
78, 1878−1888.
(72) Kursula, I.; Wierenga, R. K. J. Biol. Chem. 2003, 278, 9544−9551.
(73) Malabanan, M. M.; Nitsch-Velasquez, L.; Amyes, T. L.; Richard, J.
P. J. Am. Chem. Soc. 2013, 135, 5978−5981.
(74) Richard, J. P. Biochemistry 1998, 37, 4305−4309.
(75) He, H.; Chen, Q.; Georgiadis, M. M. Biochemistry 2014, 53,
6520−6529.
(76) Chaganti, L. K.; Kuppili, R. R.; Bose, K. FASEB J. 2013, 27, 3054−
3066.
1382
DOI: 10.1021/ja5123842
J. Am. Chem. Soc. 2015, 137, 1372−1382