Binding of a designed substrate analogue to diisopropyl fluorophosphatase: Implications for the phosphotriesterase mechanism

Article abstract of DOI:10.1021/ja061887n

A wide range of organophosphorus nerve agents, including Soman, Sarin, and Tabun is efficiently hydrolyzed by the phosphotriesterase enzyme diisopropyl fluorophosphatase (DFPase) from Loligo vulgaris. To date, the lack of available inhibitors of DFPase has limited studies on its mechanism. The de novo design, synthesis, and characterization of substrate analogues acting as competitive inhibitors of DFPase are reported. The 1.73 A crystal structure of O,O-dicyclopentylphosphoroamidate (DcPPA) bound to DFPase shows a direct coordination of the phosphoryl oxygen by the catalytic calcium ion. The binding mode of this substrate analogue suggests a crucial role for electrostatics in the orientation of the ligand in the active site. This interpretation is further supported by the crystal structures of double mutants D229N/N120D and D229N/N175D, designed to reorient the electrostatic environment around the catalytic calcium. The structures show no differences in their calcium coordinating environment, although they are enzymatically inactive. Additional double mutants E21Q/N120D and E21Q/N175D are also inactive. On the basis of these crystal structures and kinetic and mutagenesis data as well as isotope labeling we propose a new mechanism for DFPase activity. Calcium coordinating residue D229, in concert with direct substrate activation by the metal ion, renders the phosphorus atom of the substrate susceptible for attack of water, through generation of a phosphoenzyme intermediate. Our proposed mechanism may be applicable to the structurally related enzyme paraoxonase (PON), a component of high-density lipoprotein (HDL).

Full text of DOI:10.1021/ja061887n

Published on Web 09/07/2006
Binding of a Designed Substrate Analogue to Diisopropyl
Fluorophosphatase: Implications for the Phosphotriesterase
Mechanism
Marc-Michael Blum,† Frank Lo¨hr,† Andre Richardt,‡ Heinz Ru¨terjans,*,† and
Julian C.-H. Chen*,†
Contribution from the Institute of Biophysical Chemistry, J.W. Goethe UniVersity Frankfurt,
Max-Von-Laue-Strasse 9, D-60438 Frankfurt, Germany, and Armed Forces Scientific Institute for
Protection TechnologiessNBC Protection, D-29633 Munster, Germany
Received March 20, 2006; E-mail: chen@chemie.uni-frankfurt.de; hruet@bpc.uni-frankfurt.de
Abstract: A wide range of organophosphorus nerve agents, including Soman, Sarin, and Tabun is efficiently
hydrolyzed by the phosphotriesterase enzyme diisopropyl fluorophosphatase (DFPase) from Loligo vulgaris.
To date, the lack of available inhibitors of DFPase has limited studies on its mechanism. The de novo
design, synthesis, and characterization of substrate analogues acting as competitive inhibitors of DFPase
are reported. The 1.73 Å crystal structure of O,O-dicyclopentylphosphoroamidate (DcPPA) bound to DFPase
shows a direct coordination of the phosphoryl oxygen by the catalytic calcium ion. The binding mode of
this substrate analogue suggests a crucial role for electrostatics in the orientation of the ligand in the active
site. This interpretation is further supported by the crystal structures of double mutants D229N/N120D and
D229N/N175D, designed to reorient the electrostatic environment around the catalytic calcium. The
structures show no differences in their calcium coordinating environment, although they are enzymatically
inactive. Additional double mutants E21Q/N120D and E21Q/N175D are also inactive. On the basis of these
crystal structures and kinetic and mutagenesis data as well as isotope labeling we propose a new mechanism
for DFPase activity. Calcium coordinating residue D229, in concert with direct substrate activation by the
metal ion, renders the phosphorus atom of the substrate susceptible for attack of water, through generation
of a phosphoenzyme intermediate. Our proposed mechanism may be applicable to the structurally related
enzyme paraoxonase (PON), a component of high-density lipoprotein (HDL).
Chart 1
Introduction
Diisopropyl fluorophosphatase from the squid Loligo Vulgaris
(DFPase, EC 3.1.8.2, 35 kDa) efficiently detoxifies highly toxic
organophosphorus compounds that act as suicide inhibitors of
acetylcholinesterase. Besides the compound diisopropylfluoro-
phosphate (DFP), the enzyme also detoxifies the range of G-type
nerve agents including Tabun (GA), Sarin (GB), Soman (GD),
and Cyclohexylsarin (GF)1-3 (Chart 1a). Detoxification is
achieved by hydrolysis of the bond between phosphorus and
the fluoride or cyanide leaving group. DFPase is remarkably
stable and can be highly expressed, making it a top candidate
for enzymatic decontamination of chemical warfare agents.4
According to the enzyme classification of the International
Union of Biochemistry and Molecular Biology (IUBMB),
organophosphorus-hydrolyzing enzymes are found within class
EC 3.1.8 (phosphotriesterases), although these enzymes show
wide biochemical and sequence diversity. As nerve agents are
(a) Substrates of DFPase. Nerve agents are indicated by their common
names and U.S. two-letter abbreviations; (b) compounds screened for
interaction with DFPase.
† Institute of Biophysical Chemistry.
‡ Armed Forces Scientific Institute for Protection TechnologiessNBC
Protection.
synthetic compounds, they are not the native substrates of these
enzymes. Several enzymes with activities against toxic orga-
nophosphorus compounds have been reported but only DFPase
from L. Vulgaris, organophosphorus hydrolase (OPH) from
(1) Hartleib, J.; Ruterjans, H. Protein Expression Purif. 2001, 21, 210-219.
(2) Scharff, E. I.; Koepke, J.; Fritzsch, G.; Lucke, C.; Ruterjans, H. Structure
(Cambridge, MA, U.S.) 2001, 9, 493-502.
(3) Hoskin, F. C. G.; Roush, A. H. Science 1982, 215, 1255-1257.
(4) Yang, Y. C.; Baker, J. A.; Ward, J. R. Chem. ReV. 1992, 92, 1729-1743.
9
12750
J. AM. CHEM. SOC. 2006, 128, 12750-12757
10.1021/ja061887n CCC: $33.50 © 2006 American Chemical Society

Binding of a Designed Substrate Analogue to DFPase
A R T I C L E S
Figure 1. (a) Stereoview of DcPPA in the DFPase active site with Fo-Fc simulated annealing omit map, contoured at 2.5 σ, calculated in CNS by omitting
the inhibitor from the structure factor calculation. (b) Interaction of the substrate analogue DcPPA with DFPase. Potential hydrogen bond distances are
indicated in Å, as well as the catalytic calcium-phosphoryl oxygen distance. The figure was generated using PYMOL (http://pymol.sourceforge.net).
Scheme 1. Synthetic Route to O,O-Dialkylphosphoroamidates
Pseudomonas diminuta,5,6 and organophosphorus acid anhydrase
Employing the Todd-Atherton Reaction
(OPAA) from Alteromonas7,8 can be expressed in kilogram
quantities, are stable, and show the catalytic rate enhancements
required for rapid decontamination. Interestingly, the mammalian
enzyme paraoxonase (PON, EC 3.1.8.1), a high-density lipo-
protein (HDL) component9 that may function to inactivate toxic
byproducts of lipid oxidation by the low-density lipoprotein
(LDL) complex, is structurally related to DFPase and shows
very recent investigations have shown that this proposed
activity against organophosphate triesters.10 Although the natural
mechanism is probably incorrect, as a number of H287 mutants
function of squid type DFPase and Pseudomonas OPH is
unknown, the Alteromonas OPAA has the function of a prolidase
still displayed substantial activity.14 Examples include H287F
and H287L that show specific activities of 154 U/mg and 124
(E.C. 3.4.13.9).11
U/mg compared to the wild type activity of 194 U/mg. Water
Squid-type DFPase from L. Vulgaris, which is the subject of
this work, was originally isolated from squid head ganglion.
The enzyme consists of 314 amino acids and contains two
calcium ions. DFPase has been overexpressed in E. coli and
activation by these newly introduced amino acid residues is
clearly not possible.
Since the substrates for DFPase are rapidly hydrolyzed,
attempts to characterize substrate binding to DFPase have not
purified,1 and bulk quantities are readily available from expres-
been possible to date. As substrate analogues acting as enzyme
sion in the yeast P. pastoris. Crystal structures of the DFPase
inhibitors and their structural characterization are indispensable
apoenzyme have been solved at 1.8 Å and 0.85 Å resolution.2,12
tools for studying enzyme reaction mechanisms, we sought to
The overall structure resembles a six-bladed pseudosymmetrical
â-propeller with a central water filled tunnel. A high affinity
design inhibitors of DFPase and investigate their interactions
with the protein. We report the de novo design and synthesis
calcium ion is located in the center of the molecule and has a
of a DFPase substrate analogue that functions as a competitive
structural role, while the second, low-affinity calcium ion is
essential for catalytic activity and is located at the base of the
active site, sealing the water filled tunnel (Figure 1a).
inhibitor and the 1.73 Å crystal structure of the inhibitor O,O-
dicyclopentylphosphoroamidate bound to DFPase. We further
characterize its binding mode in the active site through a
The low affinity calcium ion is crucial for catalysis, as
combination of NMR, kinetic, and computational techniques.
removal of this metal ion results in a folded, yet inactive
On the basis of these data, we propose an alternative reaction
enzyme.13 From the X-ray structure and mutagenesis experi-
mechanism, involving calcium coordinating residue D229 as a
ments, H287 in the active site was postulated to be essential
nucleophile and a phosphoenzyme intermediate which existence
for catalysis, acting as a general base. On the basis of these
findings, a reaction mechanism for DFPase was proposed. The
is proven by 18O labeling experiments.
substrate coordinates the calcium ion via the phosphoryl oxygen,
such that the fluoride leaving group points away from H287.
The histidine then activates water by proton abstraction, allowing
for a backside attack of the nucleophile on the phosphorus. From
Materials and Methods
Docking. Docking studies were carried out employing the software
AUTODOCK315 with low energy conformers of the ligands obtained
with the CORINA program.16
the pentacoordinate transition state, a fluoride ion leaves as the
Ligand Synthesis. The synthetic route to dialkylphosphoroamidates
first product followed by the phosphoric acid anion. However,
is straightforward using the dialkylphosphite17 as an intermediate with
a subsequent Todd-Atherton reaction18 to the dialkylphosphoroamidate
(5) Munnecke, D. M. Appl. EnViron. Microb. 1976, 32, 7-13.
(6) Dumas, D. P.; Caldwell, S. R.; Wild, J. R.; Raushel, F. M. J. Biol. Chem.
1989, 264, 19659-19665.
(Scheme 1). (Supporting Information)
Expression and Purification of DFPase. DFPase was expressed
and purified according to previously published procedures1 and is
described in detail in the Supporting Information.
(7) Defrank, J. J.; Cheng, T. C. J. Bacteriol. 1991, 173, 1938-1943.
(8) Cheng, T. C.; DeFrank, J. J.; Rastogi, V. K. Chem.-Biol. Interact. 1999,
119-120, 455-462.
(9) Gaidukov, L.; Tawfik, D. S. Biochemistry 2005, 44, 11843-11854.
(10) Harel, M.; Aharoni, A.; Gaidukov, L.; Brumshtein, B.; Khersonsky, O.;
Meged, R.; Dvir, H.; Ravelli, R. B.; McCarthy, A.; Toker, L.; Silman, I.;
Sussman, J. L.; Tawfik, D. S. Nat. Struct. Mol. Biol. 2004, 11, 412-419.
(11) Cheng, T. C.; Harvey, S. P.; Chen, G. L. Appl. EnViron. Microb. 1996, 62,
1636-1641.
(14) Katsemi, V.; Lucke, C.; Koepke, J.; Lohr, F.; Maurer, S.; Fritzsch, G.;
Ruterjans, H. Biochemistry 2005, 44, 9022-9033.
(15) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.;
Belew, R. K.; Olson, A. J. J. Comput. Chem. 1998, 19, 1639-1662.
(16) Sadowski, J.; Gasteiger, J. Chem. ReV. 1993, 93, 2567-2581.
(17) Teichmann, B. J. Prakt. Chem. 1965, 4. Reihe, 94-98.
(18) Atherton, F. R.; Openshaw, H. T.; Todd, A. R. J. Chem. Soc. 1945, 660-
663.
(12) Koepke, J.; Scharff, E. I.; Lucke, C.; Ruterjans, H.; Fritzsch, G. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 2003, 59, 1744-1754.
(13) Hartleib, J.; Geschwindner, S.; Scharff, E. I.; Ruterjans, H. Biochem. J.
2001, 353, 579-589.
9
J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006 12751

A R T I C L E S
Blum et al.
Table 1. Enzyme Kinetics and Inhibition
Enzyme Digestion and LC/MS Analysis. A total of 10 nmol of
DFPase in 50 µL MOPS buffer (50 mM, pH 8.0) was lyophilized. The
reaction was started by dissolving the dried enzyme in 50 µL of H218O
and adding 1 µmol of DFP (one control sample was prepared without
the addition of DFP). The samples were incubated at room temperature
for 12 h. Afterward, the samples were lyophilized again and dissolved
in ammonium carbonate buffer (25 mM, pH 7.8). Endoproteinase Glu-C
(Roche) was added (weight ratio 1:50), and the samples were incubated
at 25 °C for 12 h.
The proteolytic digest was diluted 10-fold with ammonium formate
buffer (final concentration 5 mM), loaded onto a Merck LiChroCART
Supersphere 60 column connected to an Applied Biosystems 4000 Q
TRAP mass spectrometer and then eluted with a linear gradient of 10-
90% of 80% acetonitrile in 5 mM ammonium formate over 60 min at
a flow rate of 200 µL/min. The ionspray ion source was operated in
negative ion mode.
WT
+
WT
+
WT
DcPPA (1 mM)
DcPPA (3 mM)
DcPPA
K
m (mM)
2.72 ( 0.05 23.36 ( 0.07 65.45 ( 0.13
k
k
cat (s-1
)
2107 ( 9
2111 ( 7
2091 ( 14
cat / Km (M-1 s-1
)
0.77 × 106
0.09 × 106
3.19 × 105
Ki (µM)
125 ( 12
D229N/N120D
D229N/N175D
E21Q/N120D
E21Q/N175D
<1 U/mg
<1 U/mg
<1 U/mg
<1 U/mg
NMR Ligand Screening. For saturation transfer difference (STD)
and WATER-LOGSY measurements, samples typically contained 100
µM DFPase, 1-2 mM inhibitor, in 10 mM Tris pH 7.5, 2 mM CaCl2,
20% D2O. Spectra were recorded at 277 K on a Bruker Avance 700
1
MHz equipped with a cryogenic H{13C,15N} triple resonance probe.
Crystallization, Structure Determination, and Refinement. Crys-
tals of the free enzyme, D229N/N120D, and D229N/N175D mutants
were grown at room temperature from ∼38 mg/mL (∼1 mM) protein
in G150 buffer (10 mM Tris pH 7.5, 150 mM NaCl, 2 mM CaCl2) by
the hanging drop method. For crystals of the complex, 2 mM DFPase
was mixed with a 60 mM stock solution of inhibitor dissolved in G150
buffer to a final concentration of 12 mM inhibitor. Crystals were
obtained by mixing 2 µL of the protein solution with 2-3 µL of well
solution, containing 10-15% PEG 6000, 0.1 M MES pH 6.5, and large
crystals appeared after 2-3 days.
Data were collected at room temperature from capillary mounted
crystals, on a rotating anode source, using Cu KR radiation (1.5418 Å)
and were recorded on a Mar Research 345 image plate. For the D229N/
N175D mutant, data were collected at 100 K on DESY beamline BW-6
(1.05 Å) and recorded on a Mar CCD detector. Data were integrated,
reduced, and scaled with Denzo/Scalepack (free enzyme and complex),19
and MOSFLM/SCALA (D229N/N175D).20 For all datasets, integrated
intensities were converted into structure factors using TRUNCATE
(CCP4)20 and converted into CNS format.21
For the STD, saturation of the protein was achieved by a sequence of
50 ms WURST-20 pulses (1000 Hz sweep) applied for a total duration
of 2 s at an offset of 9.7 ppm. For the WATER-LOGSY, selective
inversion of the water was achieved by a 12 ms Gaussian shaped 180
degree pulse, with a mixing time of 1.5 s. In both cases, protein signals
were suppressed by a 60 ms spin lock pulse.
Proton decoupled 31P NMR experiments were performed on a Bruker
Avance 600 spectrometer, operating at 242 MHz 31P Larmor frequency.
Measurements were done at 278 and 300 K using a conventional
1H{13C, 31P} triple resonance probe. The protein/ligand ratio was varied
from 0 (no protein) to 2; the maximum ligand concentration used was
3 mM; maximum protein concentration was 2.1 mM, in the same buffer
as above.
For the detection of intermolecular NOEs, a 0.5 mM [U-2H,15N]
labeled DFPase, in 10 mM Bis-Tris-propane pH 6.5, 2 mM CaCl2,
containing 1 mM DCPPA was employed. A 3D NOESY-[15N,1H]-
TROSY spectrum was recorded at 275 K on a 900 MHz Bruker Avance
1
spectrometer equipped with a cryogenic H{13C,15N} triple resonance
probe, using a mixing time of 300 ms.
Molecular replacement was done using the 100 K WT structure (PDB
code 1E1A),2 excluding the waters and the two calcium ions. Search
models for the mutant structures used alanines in the positions of the
mutated residues. Rotation and translation searches were performed in
CNS or MOLREP. Manual building in CHAIN22 and Coot23 was
followed by rounds of positional, individual B-factor, and simulated
annealing refinement in CNS to yield the final refined structures (Table
2).
For the cocrystal in the presence of 12 mM inhibitor, a difference
Fourier Fo-Fc map was calculated based on phases from the room-
temperature WT structure, yielding a greater than 12 σ peak for the
position of the phosphorus atom of the inhibitor. A 7 σ contoured map
showed a tetrahedral shaped peak, allowing for unambiguous placement
of the oxygen and nitrogen atoms bonded to the phosphorus.
Coordinates of the solved structures have been deposited in the PDB
under accession numbers 2GVU, 2GVV, 2GVW, and 2GVX.
Kinetic Measurements. The kinetic parameters Km and kcat of DFP
hydrolysis were determined by pH stat assays. All measurements were
carried out at 298 K in a nitrogen atmosphere. The total volume was
3.0 mL, at pH 7.5, containing 10 mM NaCl and 10% acetonitrile, and
the reaction was initiated by addition of 2 µL of 0.5 mg/mL DFPase
(28.57 pmol). Initial velocities V0 were determined at eight different
substrate concentrations (0.5-10 mM) in the absence and presence of
1 mM and 3 mM dicyclopentylphosphoroamidate and corrected by the
uncatalyzed rate of DFP hydrolysis. The kinetic parameters Km and
kcat were determined by nonlinear least-squares fitting of the data to
the Michaelis-Menten equation using the program MATHEMATICA
5. The inhibitory constant Ki was obtained from fitting of the data to
the general inhibition equation. The parameters described are the
average of at least three independent measurements (Table 1).
Single and Multiple Turnover Reactions of DFPase in H218O.
For a typical multiple turnover experiment 10 nmol of DFPase in 50
µL MOPS buffer (50 mM, pH 8.0) were lyophilized. The reaction was
started by dissolving the dried enzyme in 50 µL of H218O containing
1 µmol of DFP and incubated at room temperature for 6 h. For a single
turnover experiment, 200 nmol of DFPase in 50 µL MOPS buffer (50
mM, pH 8.0) were lyophilized. The reaction was initiated by dissolving
the dried enzyme in 50 µL of H218O containing 20 nmol of DFP and
incubated at room temperature for 6 h.
Results
Inhibitor Design, Synthesis, and Characterization. To
design an effective inhibitor, replacement of the labile fluoride
leaving group by a moiety resistant to hydrolysis was necessary.
Dialkylphosphoroamidates were chosen, as earlier work with
(19) Otwinowski, Z.; Minor, W. Macromol. Crystallogr., Part A. 1997, 276,
307-326.
After incubation, the reaction mixtures were ultrafiltrated to remove
the enzyme, diluted 10-fold with 50% acetonitrile/H2O (1:1) and intro-
duced into the mass spectrometer using a Harvard Apparatus syringe
infusion pump operating at a flow rate of 5 µL/min. The molecular mass
of the produced diisopropylphosphate was measured with an Applied
Biosystems 4000 Q TRAP LC/MS/MS mass spectrometer equipped
with an ionspray ion source in the negative ion mode (ion spray voltage
) -4 kV; decluster potential ) -15 V; entrance potential ) -10 V).
(20) Bailey, S. Acta Crystallogr., Sect. D: Biol. Crystallorgr. 1994, 50, 760-
763.
(21) Brunger, A. T.; Adams, P. D.; Clore, G. M.; DeLano, W. L.; Gros, P.;
Grosse-Kunstleve, R. W.; Jiang, J. S.; Kuszewski, J.; Nilges, M.; Pannu,
N. S.; Read, R. J.; Rice, L. M.; Simonson, T.; Warren, G. L. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 1998, 54 (5), 905-921.
(22) Sack, J. S. J. Mol. Graphics 1988, 6, 224-225.
(23) Emsley, P.; Cowtan, K. Acta Crystallogr., Sect. D: Biol. Crystallogr. 2004,
60, 2126-2132.
9
12752 J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006

Binding of a Designed Substrate Analogue to DFPase
A R T I C L E S
Table 2. Data Collection and Refinement Statistics
enzyme are nearly identical, although a minor rearrangement
of the side chains of M148 and R146 in the active site leads to
a slight expansion of the pocket. As such, there is remarkably
little change in the enzyme conformation upon ligand binding,
a textbook example of the lock and key model.
WT
WT
+
inhibitor D229N/N120D
D229N/N175D
source
wavelength
temp
rotating anode
1.5418 Å
RT
DESY BW-6
1.05 Å
100 K
2.39
0.134
RT
1.73
0.061
RT
1.9
0.119
resolution (Å) 1.86
Rsym 0.056
completeness 92.1%
space group P212121
unit cell (Å) a 43.4
The ligand adopts an orientation in which its phosphoryl oxy-
gen replaces a water molecule in the first coordination shell of
the catalytic calcium; the amino group of the ligand is oriented
toward H287, while the phosphoryl oxygen coordinates the cal-
cium ion (Figure 1b). The aliphatic rings of the ligand are
involved in van der Waals contacts with the side chains of the
active site residues. There is one weak hydrogen bond (3.33 Å
δO-N) between the δ oxygen of the ligand and one of the side
chain amide protons of N175. Although the position of the phos-
phoroamidate hydrogens are equivocal, there are possible hydro-
gen bond interactions with the carboxylate oxygens of E21 (2.78
Å) and D229 (2.73 Å). None of the interactions between protein
and ligand are bridged by water molecules, and the ligand fully
occupies the active site. The phosphoryl oxygen-calcium dis-
tance in the complex is remarkably short, 2.24 Å (Figure 1c).
Interestingly, this is the same as the phosphate oxygen-calcium
distance reported in PON,10 involving the negatively charged
phosphate (cf. the uncharged DcPPA). Thus, the short distance
observed in our structure is strongly suggestive of substrate
activation, by electron withdrawl from the phosphorus atom.
94.2%
85.0%
100%
43.6
83.0
87.5
31969
21.6
18.7
43.4
83.2
87.3
21672
24.3
18.1
42.6
82.7
86.7
16354
27.1
22.0
b
83.3
87.5
25610
20.8
17.8
c
reflections
Rfree
Rcryst
the commonly used nerve agent mimic diisopropyl-methylphos-
phonate (DIMP) was unsuccessful, although it had been
successfully cocrystallized with OPH.24 Computational docking
experiments using AUTODOCK315 suggested that diisopropyl-
phosphoroamidate (DiPPA) should have a stronger binding
affinity than DFP and DIMP. As the isopropyl groups of all
these compounds do not fully occupy the hydrophobic side
pockets of the active site, other groups were computationally
screened employing bulkier side chains. The optimum binding
affinity was predicted with cyclopentyl and cyclohexyl side
chains, and steric factors led to poorer binding with bulkier side
groups. Three different substrate mimics were therefore syn-
thesized (Scheme 1) as potential inhibitors: DiPPA, dicyclo-
propylmethyl-phosphoroamidate (DcPMPA), and dicyclopen-
tylphosphoroamidate (DcPPA) (Chart 1b).
Ligand binding was assessed by STD25 and WATER-
LOGSY26 NMR experiments. DIMP showed no observable
binding, while all three synthesized compounds show a clear
STD effect. By using mixtures of two of the three compounds
the relative binding affinities were deduced as the stronger
binder always displaces the weaker one from the binding site.
The WATER-LOGSY further characterizes the two binding
compounds; the weaker binder gives rise to a negative signal,
characteristic of displacement by the stronger binder, which has
a positive signal. On the basis of these two methods, the three
compounds were ranked in terms of their relative affinities, from
weakest to strongest: DiPPA, DcPMPA, DcPPA.
The strongest inhibitor, DcPPA, was further characterized by
measurement of the Kd by NMR, and kinetic measurements to
measure its Ki and the effect of the inhibitor on the kcat and Km
of DFPase (Table 1). The Kd was estimated to be 400 µM (water
as solvent), and the Ki was measured to be 131 µM (water plus
10% acetonitrile as solvent). The measured kcat/Km of the free
enzyme is 3-fold higher than those measured previously,27 and
is most likely due to a different solvent (ethanol vs acetonitrile).
Crystal Structure. The crystal structure of DFPase cocrys-
tallized with the substrate analogue DcPPA was solved at 1.73
Å resolution, and refined to an Rfree of 21.6%. The room-
temperature structure of the DFPase free enzyme was solved
to 1.86 Å resolution, and refined to a Rfree of 20.9% (Table 2).
The structures of the complex and of the room-temperature free
A priori, because of the symmetry of the ligand, there are
two possible interactions between the calcium and the available
coordinating atoms of the ligand, with either a calcium-amide
nitrogen interaction, or a calcium-oxygen interaction. In X-ray
structures it is difficult to distinguish nitrogen from oxygen (7
vs 8 electrons). It has been modeled as a calcium-oxygen
interaction for several reasons. First, the coordination environ-
ment of calcium is strongly biased toward calcium-oxygen
interactions, according to the theory of hard and soft acids and
bases (HSAB).28 Second, the phosphorus-oxygen bond length
is expected to be shorter than the phosphorus-nitrogen bond
length. In the crystal structure, the P-O distance is 1.48 Å,
and the P-N distance is 1.62 Å. This is in good agreement
with quantum mechanical studies of the free ligand in vacuo at
the MP2/6-31G* level of theory using SPARTAN,29,30 where
the P-O distance is 1.52 Å and the P-N distance is 1.64 Å.
Third, intermolecular NOEs are observed between the NH2
group of the inhibitor and the H287 side chain, indicative of
their proximity. For these reasons, we have modeled the ligand
with a calcium-phosphoryl oxygen coordination.
We only observe the orientation of the ligand where the
nitrogen points toward H287 (Figure 1b). A second possible
orientation with the nitrogen pointing toward F173 is not
observed. In the previously proposed reaction mechanism the
fluoride leaving group would be oriented toward F173 in order
to allow the activated water to attack from H287.2 In the crystal,
the ligand displaces the C-terminal residue F314, contributed
from a symmetry-related molecule. In spite of this, the specific
binding mode of the ligand in the active site pocket, and the
bond distances between the calcium and the phosphoryl oxygen,
suggest that the interaction is neither an artifact of the crystal-
lization nor a consequence of crystal packing.
(24) Benning, M. M.; Hong, S. B.; Raushel, F. M.; Holden, H. M. J. Biol. Chem.
2000, 275, 30556-30560.
(25) Mayer, M.; Meyer, B. Angew. Chem., Int. Ed. 1999, 38, 1784-1788.
(26) Dalvit, C.; Pevarello, P.; Tato, M.; Veronesi, M.; Vulpetti, A.; Sundstrom,
M. J. Biomol. NMR 2000, 18, 65-68.
(28) Pearson, R. G. J. Am. Chem. Soc. 1963, 85, 3533-3539.
(29) Spartan ’04; Wavefunction Inc.: Irvine, CA, 2004.
(30) Pople, J. A. et al. J. Comput. Chem. 2000, 21, 1532-1548.
(27) Hartleib, J.; Ruterjans, H. Biochim. Biophys. Acta 2001, 1546, 312-324.
9
J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006 12753

A R T I C L E S
Blum et al.
18O Incorporation Experiments. To prove the existence of
a phosphoenzyme intermediate in the catalytic cycle of DFPase
we have used 18O labeling techniques. Following a similar ap-
proach to investigate the mechanism of L-2-haloacid dehalogen-
ase,31 we investigated single and multiple turnover events of
DFPase hydrolyzing DFP in H218O. The single turnover enzyme
reactions were carried out with an excess of enzyme, whereas
the multiple turnover reactions were done using an excess
amount of substrate. If the single turnover experiment reveals the
incorporation of 16O in the product, the reaction must proceed
via a phosphoenzyme intermediate as the only source of 16O in
the solution is the enzyme itself. As the transferred 16O is
replaced by 18O after the first turnover, further turnovers will
result in the incorporation of 18O. In case the reaction proceeds
via the attack of activated water (either by an activating amino
acid or the metal), 18O will be incorporated in both the single
and multiple turnover experiment. The results clearly show the
incorporation of 16O in the single turnover experiment compared
with 18O incorporation in the multiple turnover experiments
(Figure 3). Therefore an oxygen atom must be transferred from
the enzyme to the product during the course of the catalytic
cycle, indicative of a phosphoenzyme intermediate.
To identify the amino acid residue that is acting as a nucleo-
phile the protein was digested using endoproteinase Glu-C. For
this experiment DFPase was lyophilized and H218O was added.
DFP was added to start the reaction (multiple turnover) while
there was no addition of DFP for the control reaction. After the
reaction was completed, the protein was lyophilized again to
remove H218O and endoproteinase Glu-C was used for digestion.
The fragment containing the residue D229 was the peptide with
the sequence GGADGMDFDE, residues 221-230. Visual inspec-
tion of the position of the peptide fragment within the DFPase
structure reveals that only D229 is located in the active site and
is accessible by solvent. LC/MS analysis was used to identify
the proteolytic fragments. Ion spray mass spectra shows the frag-
ment of interest in negative ion mode at m/z ) 1013.6 in the
control sample. In the sample reacted with DFP, a signal with
m/z ) 1018.1 is visible at the same elution time while the origi-
nal signal of m/z ) 1013.6 is still detectable but with signifi-
cantly reduced intensity. The signal with m/z ) 1018.1 is not
detectable in the control sample. The mass difference of 4 indi-
cates that both carboxyl oxygens of D229 incorporated the iso-
tope. This can be explained by the fact that the calcium ion has
Figure 2. 31P NMR titration of DcPPA with DFPase at 278 and 300 K.
The protein/ligand ratio ranges from 0 to 2 as indicated on the left. For a
1:1 mixture, spectra were recorded at five different temperatures. Experi-
ments were carried out at 242 MHz with the number of scans ranging from
2048 to 60000. Spectra are vertically scaled to obtain approximately equal
noise levels. Line width (full width at half-height) of the free state signal
are indicated where measurements were possible.
On a dynamic level, using 31P NMR as a chemical probe,
the inhibitor was titrated with unlabeled protein and measured
at 278 and 300 K (Figure 2).
an off rate of ∼1 s-1 13
. During the time the active site is without
a bound metal, the carboxyl group of D229 is able to flip around.
Upon addition of the protein, a second signal becomes
detectable 5.8 ppm downfield from the free inhibitor. This
downfield shift upon addition of protein can be explained by
the polarization of the phosphorus-oxygen bond due to
coordination to the calcium. This indicates that the calcium ion
participates in substrate activation. Furthermore, the appearance
of two separated signals is indicative of a slow exchange of the
free and bound states on the NMR time scale, with an estimated
upper limit of 1400 s-1. In contrast, for fast chemical exchange,
one would expect one signal at an averaged chemical shift which
changes with the protein to ligand ratio. From the line
broadening of the free state signal upon addition of DFPase,
the exchange rate can be estimated as 250-300 s-1 at 278 K.
Exchange broadening becomes more severe with increasing
temperature (Figure 2).
Discussion
The orientation of the ligand suggests that the central “core”
of the active site cleft is critical for guiding the ligand into the
proper orientation. This core contains the catalytic calcium ion,
coordinated by acidic residues E21 and D229 on one face, and
N120 and N175 on the other, such that the ligand may be
electrostatically oriented in the active site (Figure 1b). To test
this possibility, four sets of double mutants were constructed,
trying all chemically reasonable permutations, while preserving
the charge balance of the calcium coordination environment.
The double mutant D229N/N120D was found to be inactive,14
albeit with weak binding of DcPPA based on STD data. The
(31) Liu, J. Q.; Kurihara, T.; Miyagi, M.; Esaki, N.; Soda, K. J. Biol. Chem.
1995, 270, 18309-18312.
9
12754 J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006

Binding of a Designed Substrate Analogue to DFPase
A R T I C L E S
Figure 3. Ion spray mass spectra of diisopropylphosphate produced with DFPase in H218O. The spectra were obtained between 175 and 195 atomic mass
units. A single turnover cycle of DFPase yields 16O-diisopropylphosphate Mr ) 181 amu (left) while multiple turnovers result in 18O incorporation Mr ) 183
amu (right).
crystal structure of D229N/N120D was solved and refined at
1.9 Å resolution (Table 2). In the crystal structure, the active
site environment is unchanged, notably in the calcium coordina-
tion environment. A second double mutant was constructed,
D229N/N175D. This mutant also shows no activity, and
surprisingly, no inhibitor binding was detected by NMR STD
measurements, indicative of a significant effect of these muta-
tions on the Km. The crystal structure was solved at 2.4 Å
resolution (Table 2). Like D229N/N120D, the D229N/N175D
structure shows an unchanged calcium coordination environment
and an undisturbed substrate binding pocket. Double mutants
E21Q/N120D and E21Q/N175D also showed no activity (Table
1). These results strongly suggest that the correct electrostatic
Figure 4. Aligned structures of DcPPA docked into the DFPase active
site (green) and the structure of DcPPA in complex with DFPase as revealed
by X-ray crystallography (gray). Only minor differences in the protein
structure are visible upon binding.
topography of the active site is absolutely necessary for binding
and catalysis. Furthermore, these results significantly narrow
down the basis for substrate recognition, binding, and enzymatic
activity. Nevertheless, the hydrophobic ‘wings’ of the active
site pocket are also important for binding, as DiPPA was a weak
inhibitor while DcPPA was a better inhibitor (Figure 1a).
A number of studies have suggested the role of H287 as a
general base.2,27 For this to occur, the hydroxide generated by
proton abstraction by H287 must attack the phosphorus from a
position apical to the fluoride leaving group. The amino group
of the inhibitor DcPPA points toward H287 and this result is in
very good agreement with the docking studies carried out using
AUTODOCK (Figure 4). The docked conformation of DFP also
shows the fluoride leaving group pointing toward H287 (Figure
5). In this orientation the backside attack of hydroxide is not
possible. Together with the active H287L and H287F mutants
reported previously, and the lack of pH titration effects for this
residue,14 this rules out H287 as a general base.
We rule out the direct attack of hydroxide from bulk solvent.
At the lower pH limit of stability (pH 5; [OH-] ) 10-9 M) the
hydroxide concentration is 10-fold lower than the enzyme
concentration used in the kinetic experiments and is too small
to account for the activity measured.27 Furthermore, we rule
out the possibility of direct activation of water by the catalytic
calcium ion. Exchanges of the catalytic calcium (Kd ≈ 1 µM)
with Mg2+ and Ba2+ show a slight increase of activity by
approximately 5-20% relative to the wild-type. Much stronger
effects would be expected for the case where the metal ion acts
as a water activator. Furthermore, mutants of all non-calcium-
coordinating amino acids with functional groups in the active
site vicinity did not implicate any water activating residues.2,14
We propose the following mechanism (Figure 6). The
prerequisites are that the residue must be in an orientation such
that an inline attack of a nucleophile upon the phosphorus can
occur, and the only reasonable residue available for this is D229.
D229 has one available oxygen exposed in the active site; the
other oxygen is involved in coordinating the calcium.
As such, we propose that the D229 carboxylate plays a dual
role, in coordinating the calcium ion, and acting as the
nucleophile. This is similar to that proposed for the phosphatase
activity in epoxide hydrolase32a where the nucleophilic aspartate
is coordinated to magnesium and to that proposed for phospho-
noacetaldehyde hydrolase from Bacillus cereus.32b In general,
metal dependent phosphatase activities of enzymes belonging
to the haloacid dehalogenase (HAD) superfamily are thought
to work via an enzyme aspartate coordinated to the magnesium
nucleophile.33 Although the coordination of D229 to the calcium
ion should lower the nucleophilicity of the carboxylate, this
effect may be balanced by the coordination of the phosphoryl
(32) (a) Gomez, G. A.; Morisseau, C.; Hammock, B. D.; Christianson, D. W.
Biochemistry 2004, 43, 4716-4723. (b) Morais, M. C.; Zhang, W. H.;
Baker, A. S.; Zhang, G. F.; Dunaway-Mariano, D.; Allen, K. N. Biochem-
istry 2000, 39, 10385-10396.
(33) Allen, K. N.; Dunaway-Mariano, D. Trends Biochem. Sci. 2004, 29,
495-503.
9
J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006 12755

A R T I C L E S
Blum et al.
Proper arrangement of a water molecule for attack on the phos-
phoenzyme intermediate can contribute to a rate enhancement
even without water activation.35 Also it is important to note that
the intermediate is not a charged acetyl phosphate but a neutral
alkylated species. Involvement of the calcium ion in water acti-
vation at this step is highly unlikely because of steric reasons.
Even if a water molecule could coordinate to calcium during the
existence of the intermediate it would be too far away and poorly
positioned for an attack on the carboxylic carbon of D229.
Electrostatics also plays a crucial role in substrate orientation,
as all mutants switching the polarity of the calcium coordinating
residues result in inactive enzyme. Two of these mutants are
identical in structure to the wild-type enzyme, and show no
change in the calcium coordination environment.
Docking studies of the binding of the substrate ethyl N,N-
dimethyl phosphor-amidocyanidate (Tabun) give further support
to the proposed mechanism. Because of the size and steric con-
straints of the linear cyano group, both enantiomers of Tabun36
are able to bind in the DFPase active site, though neither with the
cyano group pointing toward H287 (similar to the inhibitor Dc-
PPA) nor directly away from it (similar to DFP in a previously
proposed reaction mechanism2). Results of the docking show
conformations where the cyano group is positioned for an inline
attack of D229 on the phosphorus atom (Figure 7).
Figure 5. Docking studies reveal productive (a) and nonproductive (b)
binding of DFP in the DFPase active site. The difference of the two states
is estimated to be about 0.5 kcal/mol. Nucleophilic inline attack of D229
on the phosphorus is possible in structure (a) but not in structure (b) as the
leaving group points in the wrong direction.
Our proposed mechanism makes several predictions, which
can be tested experimentally. One key feature of the mechanism
is the enzyme as the reactive nucleophile itself and not activated
water leading to the formation of a phosphoenzyme intermediate.
We were able to prove the existence of this intermediate by
isotopic labeling using H218O and analysis by mass spectroscopy
(Figure 3). Also the pKa of D229 is expected to be significantly
altered because of the pH dependence of the reaction. Although
an 0.85 Å X-ray structure of DFPase is available, it was not
possible to assign the hydrogen atoms in the active site.12
As the medically important enzyme paraoxonase (PON) is
structurally related to DFPase with both proteins showing
propeller structures, two calcium ions, a water filled tunnel, and
activity against phosphorus triesters, we investigated the crystal
structure of an engineered PON1 available from the PDB (PDB
code: 1V04) for possible mechanistic similarities.10 The cur-
rently proposed mechanism for PON1 suggests activation of
water by a His-His dyad (H115, H134). Although recent work
has shown that the His-His dyad is mediating the lactonase
and esterase functions of PON1,37 there are experimental results
that suggest that the triesterase activity of PON1 might follow
a different mechanism. Yeung et al. reported that the H115W
mutant of PON1 retains activity with paraoxon.38 They con-
cluded that H115 is not directly involved in catalysis.39 Close
inspection of the coordination environment around the catalytic
calcium reveals that the residue D269 could function as a
nucleophile similar to D229 in DFPase. Substrates with large
leaving groups like the nerve agent VX, for which a small
activity was discovered,40 are predicted to position their leaving
oxygen to the calcium. Car-Parrinello molecular dynamics on
the crucial reaction step in the phosphatase activity of expoxide
hydrolase leading toward the formation of the phosphoenzyme
intermediate showed clearly that such a step is possible.34 The
comparatively slow kcat for DFPase may be reflective of an
equilibrium between the nonproductive and productive orienta-
tions of the ligand in the binding pocket. We suggest that the
DFP substrate is bound in two orientations, a nonproductive
one similar to that seen in the inhibitor/enzyme cocrystal
structure, with the fluorine atom pointing toward H287, in
equilibrium with a productive orientation, where the fluorine is
pointing away from D229 (Figure 5). On the basis of docking
experiments, the productive orientation is estimated to be only
0.5 kcal/mol higher in energy than the nonproductive orientation.
The calcium ion is involved in activation of the substrate,
rendering the phosphorus atom electron poor and therefore primed
to attack by D229. Titration experiments with DcPPA show a
shifted 31P NMR spectrum, indicative of polarization of the phos-
phorus-oxygen bond by the calcium (Figure 2). D229 attacks
the phosphorus atom, leading to the release of fluoride and gener-
ating a phosphoenzyme intermediate. This process is possibly
assisted by hydrogen bonding of the fluorine to the amide group
of N120. In the second step, the carboxylate carbon of D229 is
attacked by a water molecule, leading to the hydrolysis of the
D229-phosphoenzyme bond under transfer of an oxygen atom
from D229 to the product diisopropylphosphate, regeneration of
the D229, and product generation (Figure 6). The rate of hydrol-
ysis of acetylphosphate is ∼10-4 s-1 under physiological
conditions. The turnover rate of DFPase is ∼200 s-1. With no
obvious general base in the vicinity, this is difficult to explain.
(35) Clausen, J. D.; McIntosh, D. B.; Woolley, D. G.; Andersen, J. P. J. Biol.
Chem. 2001, 276, 35741-35750.
(36) Benschop, H. P.; Dejong, L. P. A. Acc. Chem. Res. 1988, 21, 368-374.
(37) Khersonsky, O.; Tawfik, D. S. J. Biol. Chem. 2006, 281, 7649-7656.
(38) Yeung, D. T.; Josse, D.; Nicholson, J. D.; Khanal, A.; McAndrew, C. W.;
Bahnson, B. J.; Lenz, D. E.; Cerasoli, D. M. BBA-Proteins Proteom. 2004,
1702, 67-77.
(39) Yeung, D. T.; Lenz, D. E.; Cerasoli, D. M. FEBS J. 2005, 272, 2225-2230.
(40) Broomfield, C. A.; Morris, B. C.; Anderson, R.; Josse, D.; Masson, P.
Conference Proceedings CBMTS III, Spiez, Switzerland, May 2000.
(34) De Vivo, M.; Ensing, B.; Klein, M. L. J. Am. Chem. Soc. 2005, 127, 11226-
11227.
9
12756 J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006

Binding of a Designed Substrate Analogue to DFPase
A R T I C L E S
Figure 6. Proposed mechanism of DFP hydrolysis. After coordination of the substrate to the catalytic calcium ion, the phosphorus atom is subject to
nucleophilic attack by D229, generating a phosphoenzyme intermediate, accompanied by release of fluoride, assisted by hydrogen bonding to N120. The
intermediate is then attacked by water at the carboxylate carbon of D229 resulting in a transfer of oxygen from D229 to the phosphoric acid product leaving
the active site as diisopropylphosphate.
Figure 8. Docked conformation of VX in the active site of paraoxonase
(PON). Residue D269 is well positioned for a nucleophilic attack on the
phosphorus atom of VX.
engineering DFPase to expand the repertoire of compounds that
can be detoxified by the enzyme, notably the nerve agent VX.41
Acknowledgment. We thank Gleb Bourenkov (DESY) and
Roy Lancaster (MPI Biophysik) for access to X-ray facilities
and assistance with data collection. We thank Hans-Ju¨rgen
Figure 7. Results of computational docking of both enantiomers of Tabun
Altmann, Werner Ho¨rnemann (Armed Forces Research Institute
for Protection Technologies), and Juliana Winkler (Frankfurt)
for technical assistance, Harald John, Christian Becker, and
Erwin Wagner (Armed Forces Institute of Pharmacology and
Toxicology) for help with mass spectrometry, and Alex Koglin
and Volker Do¨tsch for critical comments on the manuscript.
This research was funded by Fraunhofer Grant E590/3X023/
M5137 and by the Hessisches Ministerium fu¨r Wissenschaft
und Kultur.
in the DFPase active site. Docked conformations are those of lowest energy
displaying a phosphoryl oxygen to calcium coordination. In both cases the
cyanide leaving group of Tabun is oriented in a way that is well suited for
an inline attack on the phosphorus by D229. Because of the size and steric
constraints of the linear cyano group, Tabun is unable to bind to DFPase
with the cyano group pointing toward H287 (like the inhibitor DcPPA) or
pointing away from H287 (as proposed for DFP in the previously proposed
reaction mechanism).
groups in an orientation such that water activated by the His-
His dyad is unable to perform the necessary inline attack. In
contrast, D269 is well positioned for this to occur (Figure 8).
Through the design and synthesis of a DFPase inhibitor, and
its kinetic characterization, combined with computational, muta-
tional, and structural studies, we have proposed a possible mech-
anism for the action of DFPase. A thorough understanding of the
mechanismwillbecrucialforeffortsaimedatlarge-scaleenzymat-
ic decontamination of existing nerve agent stocks, as well as re-
Supporting Information Available: General synthesis of
substrate analogues, protein purification protocol, and complete
ref 30. This material is available free of charge via the Internet
at http://pubs.acs.org.
JA061887N
(41) Yang, Y. C. Acc. Chem. Res. 1999, 32, 109-115.
9
J. AM. CHEM. SOC. VOL. 128, NO. 39, 2006 12757