Reversed enantioselectivity of diisopropyl fluorophosphatase against organophosphorus nerve agents by rational design

Article abstract of DOI:10.1021/ja905444g

Diisopropyl fluorophosphatase (DFPase) from Loligo vulgaris is an efficient and robust biocatalyst for the hydrolysis of a range of highly toxic organophosphorus compounds including the nerve agents sarin, soman, and cyclosarin. In contrast to the substrate diisopropyl fluorophosphate (DFP) the nerve agents possess an asymmetric phosphorus atom, which leads to pairs of enantiomers that display markedly different toxicities. Wild-type DFPase prefers the less toxic stereoisomers of the substrates which leads to slower detoxification despite rapid hydrolysis. Enzyme engineering efforts based on rational design yielded two quadruple enzyme mutants with reversed enantioselectivity and overall enhanced activity against tested nerve agents. The reversed stereochemical preference is explained through modeling studies and the crystal structures of the two mutants. Using the engineered mutants in combination with wild-type DFPase leads to significantly enhanced activity and detoxification, which is especially important for personal decontamination. Our findings may also be of relevance for the structurally related enzyme human paraoxonase (PON), which is of considerable interest as a potential catalytic in vivo scavenger in case of organophosphorus poisoning.

Full text of DOI:10.1021/ja905444g

Published on Web 11/06/2009
Reversed Enantioselectivity of Diisopropyl
Fluorophosphatase against Organophosphorus Nerve Agents
by Rational Design
Marco Melzer,†,‡ Julian C.-H. Chen,§ Anne Heidenreich,† Ju¨rgen Ga¨b,†,⊥
Marianne Koller,| Kai Kehe,| and Marc-Michael Blum*,†
Blum-Scientific SerVices, Ledererstrasse 23, 80331 Munich, Germany, Institute of Biophysical
Chemistry, Goethe UniVersity Frankfurt, Max-Von-Laue-Strasse 9, 60438 Frankfurt, Germany,
Bundeswehr Institute of Pharmacology and Toxicology, Neuherbergstrasse 11,
80937 Munich, Germany, Institute of Pathology, Johannes Gutenberg UniVersity Mainz,
55101 Mainz, Germany, and Institute of Pharmaceutical Chemistry, Philipps UniVersity
Marburg, 35032 Marburg, Germany
Received July 2, 2009; E-mail: mmblum@blum-scientific.de
Abstract: Diisopropyl fluorophosphatase (DFPase) from Loligo vulgaris is an efficient and robust biocatalyst
for the hydrolysis of a range of highly toxic organophosphorus compounds including the nerve agents
sarin, soman, and cyclosarin. In contrast to the substrate diisopropyl fluorophosphate (DFP) the nerve
agents possess an asymmetric phosphorus atom, which leads to pairs of enantiomers that display markedly
different toxicities. Wild-type DFPase prefers the less toxic stereoisomers of the substrates which leads to
slower detoxification despite rapid hydrolysis. Enzyme engineering efforts based on rational design yielded
two quadruple enzyme mutants with reversed enantioselectivity and overall enhanced activity against tested
nerve agents. The reversed stereochemical preference is explained through modeling studies and the crystal
structures of the two mutants. Using the engineered mutants in combination with wild-type DFPase leads to
significantly enhanced activity and detoxification, which is especially important for personal decontamination.
Our findings may also be of relevance for the structurally related enzyme human paraoxonase (PON), which is
of considerable interest as a potential catalytic in vivo scavenger in case of organophosphorus poisoning.
Scheme 1. Structures of DFP and Organophosphorus Nerve
Agents and Hydrolysis of DFP Catalyzed by DFPase with
Reaction Productsa
Introduction
Enzyme engineering, through methods such as site-directed,
site-saturation mutagenesis, and directed evolution are common
ways of improving upon and augmenting an enzyme’s activity.1-3
Engineering stereoselectivity into an enzyme has potential
applications in enzyme-assisted chemical synthesis and creation
of novel unnatural compounds and drug leads. The enzyme
diisopropyl fluorophosphatase (DFPase, EC 3.1.8.2) from the
squid Loligo Vulgaris is a potent and well characterized
phosphotriesterase capable of hydrolyzing a variety of organo-
phosphorus compounds that act as irreversible inhibitors of
acetylcholinesterase (AChE). Substrates include diisopropyl
fluorophosphate (DFP) and a range of highly toxic G-type
organophosphorus (OP) nerve agents such as sarin (GB), soman
(GD), and cyclosarin (GF) that still pose a credible threat to
military personnel as well as civilian populations (Scheme 1).4-7
a (a) Structures of DFP and organophosphorus nerve agents, which are
substrates of DFPase. Asterisks indicate stereocenters. (b) Hydrolysis of
DFP catalyzed by DFPase with reaction products.
† Blum-Scientific Services.
While DFP is achiral, the nerve agents possess a stereocenter
at the phosphorus atom and exist as pairs of enantiomers (four
stereoisomers for GD due to an additional stereocenter in the
‡ J. Gutenberg University Mainz.
§ Goethe University Frankfurt.
⊥ Philipps University Marburg.
| Bundeswehr Institute of Pharmacology and Toxicology.
(1) Toscano, M. D.; Woycechowsky, K. J.; Hilvert, D. Angew. Chem.,
Int. Ed. 2007, 46, 3212–3236.
(5) Blum, M. M.; Lo¨hr, F.; Richardt, A.; Ru¨terjans, H.; Chen, J. C. H.
J. Am. Chem. Soc. 2006, 128, 12750–12757.
(2) Penning, T. M.; Jez, J. M. Chem. ReV. 2001, 101, 3027–3046.
(3) Antikainen, N. M.; Martin, S. F. Bioorg. Med. Chem. 2005, 13, 2701–
2716.
(6) Blum, M. M.; Timperley, C. M.; Williams, G. R.; Thiermann, H.;
Worek, F. Biochemistry 2008, 47, 5216–5224.
(7) Blum, M. M.; Mustyakimov, M.; Ru¨terjans, H.; Kehe, K.; Schoenborn,
B. P.; Langan, P.; Chen, J. C. H. Proc. Natl. Acad. Sci. U.S.A. 2009,
106, 713–718.
(4) Scharff, E. I.; Koepke, J.; Fritzsch, G.; Lu¨cke, C.; Ru¨terjans, H.
Structure (Cambridge, MA, U.S.) 2001, 9, 493–502.
9
17226 J. AM. CHEM. SOC. 2009, 131, 17226–17232
10.1021/ja905444g CCC: $40.75  2009 American Chemical Society

Enantioselectivity of DFPase against OP Nerve Agents
A R T I C L E S
pinacolyl side chain). These optical isomers also exhibit different
inhibitory potencies toward acetylcholinesterase (AChE) and
therefore different toxicities8 and can racemize in aqueous
solution9 in a fluoride catalyzed process.
available for the screening of large mutant libraries and that
further restrictions would arise due to the high toxicity of the
substrates. With an atomic resolution X-ray19 and a neutron7
structure of DFPase available, combined with knowledge about
the phosphotriesterase mechanism,5 we aimed to create im-
proved mutants using rational design. Furthermore, the high
stability of the enzyme and the rigidity of the active site pocket
make DFPase an attractive candidate for these re-engineering
studies. In this work we report the creation of DFPase mutants
with reversed stereoselectivity and enhanced activity against a
range of nerve agent substrates by rational design and explain
the changed stereochemical preference by structural analysis
and computer modeling.
Earlier work revealed that wild-type (WT) DFPase shows a
stereochemical preference for the less toxic enantiomers of
G-type nerve agents.10 While this is not a large concern for the
decontamination of equipment as long as full detoxification is
achieved within a certain time, it is a problem for applications
requiring a rapid decrease in toxicity, such as topical applications
or in ViVo use. At the same time a very high enantioselectivity
of the enzyme, as normally required for enzymes used in organic
synthesis,11,12 is also not desirable as complete hydrolysis of
the agent is still required. Therefore, the goal is an enzyme
mutant with enhanced total activity, that displays enantioselec-
tivity for the more toxic stereoisomer leading to a rapid decrease
in toxicity while still turning over the less toxic stereoisomer
rapidly enough to ensure complete hydrolysis.
The currently proposed mechanism for DFPase assumes that
the incoming substrate replaces a water molecule in the
coordination sphere of the active site calcium ion binding via
its phosphoryl group and is activated for nucleophilic attack by
the free oxygen atom of calcium-coordinating amino acid residue
D229. This results in the formation of a phosphoenzyme
intermediate, which is subsequently hydrolyzed to form a
phosphonic or phosphoric acid as the reaction product.5 This
mechanism has been supported by the results obtained with OP-
compounds with fluorogenic leaving groups as substrates6 and
by the solvent orientation and protonation states found in a
recent neutron diffraction study.7 The phosphotriesterase activity
of paraoxonase 1 (PON1) might follow the same mechanism,5,6
a claim that has been supported by a recent in silico study.13
Like DFPase, PON1 preferably hydrolyzes the less toxic nerve
agent stereoisomers.14-16 In order to find PON1 mutants with
reversed or broader stereoselectivity, directed evolution was used
employing nerve agent derivatives with fluorogenic leaving
groups.17,18
Materials and Methods
Substrates. Chemical warfare agents were made available by
the German Ministry of Defense and handled in accordance with
the Chemical Weapons Convention. DFP was obtained from Fluka.
Organophosphorus compounds were used as approximately 2% (v/
v) stock solutions in dry acetonitrile.
Expression and Purification of DFPase. Wild-type DFPase and
mutants were expressed in E. coli BL21 cells and purified according
to the method of Hartleib and Ru¨terjans.20 Mutants were generated
by site-directed mutagenesis. Detailed procedures are described in
the Supporting Information.
Crystallization and Structure Determination. Crystals of
DFPase mutants were grown at room temperature (RT) by vapor
diffusion using the hanging drop method. A 2 µL portion of 1 mM
protein solution (ca. 35 mg/mL) was mixed with well solution
containing 0.1 M Tris buffer pH 8.5, 2% tacsimate, 16% PEG 3350
for mutant E37A/Y144A/R146A/T195M (Mut1), or 0.2 M KCl,
0.05 M HEPES buffer pH 7.5, 35% pentaerythriol propoxylate (5/4
PO/OH) for mutant E37D/Y144A/R16A/T195M (Mut2). Crystals
grew over a period of 2-4 weeks and were transferred to
cryoprotectant containing mother liquor with 20% glycerol and
flash-frozen in liquid nitrogen prior to data acquisition.
Data was collected at 100 K at ESRF beamlines ID14-1 (Mut1)
and BM16 (Mut2). Data were integrated, reduced, and scaled with
MOSFLM/SCALA (Mut1) and HKL2000 (Mut2).21,22 Molecular
replacement was done using the RT WT structure of DFPase (PDB:
2GVW) excluding the waters and the two calcium ions, with
alanines in place of the mutated residues. For Mut2, rotation and
translation searches were performed in MOLREP,23 yielding an
unambiguous set of solutions corresponding to the four molecules
in the asymmetric unit. Mut1 maps were calculated using difference
Fourier maps, clearly showing the amino acid differences at residue
37 between Mut1 and Mut2. Manual building in Coot24 was
followed by rounds of positional, individual B-factor, and simulated
annealing refinement in CNS 1.225 to yield the final refined
structures.
Realizing the complications arising from the different size
and chemical behavior of chromogenic or fluorogenic leaving
groups compared to the small fluoride leaving group found in
the original substrates, our approach focused on the original
substrates accepting that a high-throughput assay would not be
(8) Benschop, H. P.; De Jong, L. P. A. Acc. Chem. Res. 1988, 21, 368–
374.
(9) Christen, P. J.; van den Muysenberg, J. A. Biochim. Biophys. Acta
1965, 110, 217–220.
Kinetic Measurements. Specific enzymatic activities and kinetic
parameters were recorded using Fourier transform infrared (FTIR)
spectroscopy according to the method of Ga¨b et al.26 To discrimi-
(10) Schulz, W.; Scha¨fer, B.; Stroop, G.; Ru¨terjans, H. Forschungsberichte
aus der Wehrtechnik 1987, T/R770/D 00001/D 1750.
(11) Bornscheuer, U. T.; Pohl, M. Curr. Opin. Chem. Biol. 2001, 5, 137–
143.
(12) Faber, K. Biotransformations in Organic Chemistry, 4th ed.; Springer-
Verlag: Berlin, Germany, 1999.
(19) Koepke, J.; Scharff, E. I.; Lu¨cke, C.; Ru¨terjans, H.; Fritzsch, G. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 2003, 59, 1744–1754.
(20) Hartleib, J.; Ru¨terjans, H. Protein Expression Purif. 2001, 21, 210–
219.
(21) Leslie, A. G. W. Jnt CCP4/ESF-EACBM Newsl. Protein Crystallogr.
1992, 26, 27–33.
(22) Otwinowsky, Z.; Minor, W. Methods Enzymol. 1997, 276, 307–326.
(23) Vagin, A.; Teplyakov, A. J. Appl. Crystallogr. 1997, 30, 1022–1025.
(24) Emsley, P.; Cowtan, K. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2004, 60, 2126–2132.
(25) 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, 905–921.
(26) Ga¨b, J.; Melzer, M.; Kehe, K.; Richardt, A.; Blum, M. M. Anal.
Biochem. 2009, 385, 187–193.
(13) Hu, X.; Jiang, X.; Lenz, D. E.; Cerasoli, D. M.; Wallquist, A. Proteins
2009, 75, 486–498.
(14) Amitai, G.; Gaidukov, L.; Adani, R.; Yishay, S.; Yacov, G.; Kushnir,
M.; Teitlboim, S.; Lindenbaum, M.; Bel, P.; Kerhsonsky, O.; Tawfik,
D. S.; Meshulam, H. FEBS J. 2006, 273, 1906–1919.
(15) Yeung, D. T.; Smith, J. R.; Sweeney, R. E.; Lenz, D. E.; Cerasoli,
D. M. FEBS J. 2007, 274, 1183–1191.
(16) Yeung, D. T.; Smith, J. R.; Sweeney, R. E.; Lenz, D. E.; Cerasoli,
D. M. J. Anal. Toxicol. 2008, 32, 86–91.
(17) Amitai, G.; Adani, R.; Yakov, G.; Yishay, S.; Teitlboim, S.; Tveria,
L.; Limanovich, O.; Kushnir, M.; Meshulam, H. Toxicology 2007, 233,
187–198.
(18) Amitai, G.; Adani, R.; Limanovich, O.; Teitlboim, S.; Yishay, S.;
Tveria, L.; Yacov, G.; Meshulam, H.; Raveh, L. Chem. Biol. Interact.
2008, 175, 249–254.
9
J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009 17227

A R T I C L E S
Melzer et al.
nate the hydrolysis rates for the different stereoisomers of the chiral
substrates we used a combination of two methods. The overall
(racemic) hydrolysis rate was measured titrimetrically using a pH-
Stat method27 while the hydrolysis of the toxic stereoisomer was
determined by the inhibitory potency of the reaction solution against
AChE. Human hemoglobin-free erythrocyte ghosts were used as
the AChE source.28 Samples were taken at intervals, and the
reaction was quenched in formate buffer (pH 3.5). The inhibitory
potency and therefore the concentration of the toxic stereoisomer
were determined by a modified Ellman assay using acteylthiocholine
and DTNB.29-31 Concentrations were calculated using the method
of Hart and O’Brien.32 Detailed procedures are described in the
Supporting Information.
kcat/KM values for the enantiomers of GB and GF were
determined by modifying a mathematical procedure described
by Yeung et al.15 for analyzing kinetic data for the four
stereoisomers of GD. In the presence of a racemic mixture of
GB or GF, the catalyzed reaction is analogous to simultaneously
deriving the kinetic constants for the hydrolysis of two competi-
tive substrates. Instead of fitting the individual values of kcat
and KM for the two enantiomers (four variables), only kcat/KM
values (two variables) were used for the fitting procedure, which
was carried out using the software MATHEMATICA. This was
due to the fact that all reactions with GB and GF were carried
out at nonsaturating conditions.
Chiral GC/MS. The GC/MS method for the separation of
cyclosarin enantiomers was based on the original method of Reiter
et al.33 with some modifications. The GC/MS system consisted of
a 6890N gas chromatograph connected to a 5973 Network MS (both
Agilent Technologies) equipped with a cold injection system (CIS).
According to the original method the CIS was operated in the
solvent vent mode at 50 °C for 1.8 min with the split exit open.
After this initialization, the split exit was closed and the temperature
was raised to 260 °C using the maximal heating rate. After the
temperature was reached, the split exit was opened again and the
liner was purged for 3 min with carrier gas (purified helium passed
over moisture, oxygen, and two active charcoal filters) with a flow
rate of 1.3 mL/min. The oven program was initiated at 50 °C
holding the temperature for 2 min. The temperature was then raised
to 105 °C with a rate of 40 °C/min (no hold), then further to 130
°C at a rate of 3 °C/min (no hold), and finally to 170 °C at a rate
of 40 °C/min (no hold). The MS was operated in the selected ion-
monitoring mode (SIM) with a dwell-time of 100 ms for m/z 99.
The enantiomers of cyclosarin were separated using a chiral
GAMMADEX 225 column (Supelco, 30 m length, 0.25 i.d., 25
µm film-thickness). Sample preparation was carried out using solid
phase extraction (SPE). Cyclohexyl columns (Isolute) were con-
ditioned using 1 mL methanol followed by 1 mL water and loaded
with 10 µL of sample solution afterward. After being washed with
water (3 × 1 mL) cyclosarin was eluted with 1 mL of isopropanol
and diluted 10-fold using the same solvent in an autosampler vial
for GC/MS determination.
above the catalytic calcium ion. The Lamarckian genetic algorithm
with a maximum of 20 million energy evaluations in combination
with a local search algorithm was used for each of the 100 individual
docking runs. All other parameters were left to the preset values in
AutoDock-Tools. Low energy conformers of the ligands were
obtained with the CORINA program.36
Modeling. Models of the phosphoenzyme intermediate were
generated by hand using Coot24 and subjected to thorough energy
minimization using both the steepest descent and conjugate gradient
algorithm as implemented in the GROMACS37,38 simulation
software. Modifications of the GROMOS96 43a1 force field39 for
the phosphorlyated aspartate residues were carried out on the basis
of the results of the PRODRG40 server for these modified amino
acids.
Results
Stereochemical Preference of WT DFPase. In a first step the
stereochemical preference of wild-type (WT) DFPase for the
substrates sarin (GB), soman (GD), and cyclosarin (GF) was
determined using a biological assay based on the inhibition of
human acetylcholinesterase (AChE). The enzyme concentration
was maintained at a dilute 20-30 nM depending on the substrate
and enzyme mutant to ensure accurate sampling. For GF, 50%
of the agent is hydrolyzed after 300 s, but virtually no
detoxification has occurred. Only after 7500 s is complete
degradation observed (Figure 1a). It was recently shown that
the two enantiomers of GF possess markedly differing toxici-
ties.41 In the first phase of the reaction only the less toxic
enantiomer is hydrolyzed while a substantial decrease in toxicity
is only detected after prolonged reaction times. The same
behavior is seen for GD while for GB there is no clear
stereochemical preference. The results are in agreement with
qualitative observations made by Schulz and co-workers.10 It
was shown previously that the more toxic stereoisomers of GB
and GD are those with S configuration at the phosphorus atom
(SP).8,42 On the basis of the structural determinants of the AChE
active site,43,44 it can be assumed that this is also true for GF.
In a second step all stereoisomers of GB, GD, and GF were
docked into the active site of WT DFPase. The binding mode
of the substrate in the active site has to fulfill two criteria to
adopt a productive arrangement for reaction. First, the phos-
phoryl oxygen has to coordinate to the catalytic calcium ion
for substrate activation. Second, the leaving group of the
substrate must be oriented away from catalytic residue D229
(34) 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.
(35) Sanner, M. F. J. Mol. Graphics Modell. 1999, 17, 57.
(36) Sadowski, J.; Gasteiger, J. Chem. ReV. 1993, 93, 2567–2581.
(37) van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.;
Berendsen, H. J. C. J. Comput. Chem. 2005, 26, 1701–1718.
(38) Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E. J. Chem. Theory
Comput. 2008, 4, 435.
(39) van Gunsteren, W. F.; Billeter, S. R.; Eising, A. A.; Hu¨nenberger,
P. H.; Kru¨ger, P.; Mark, A. E.; Scott, W. R. P.; Tironi, I. G.
Biomolecular Simulation: The Gromos 96 Manual and User Guide;
vdf Hochschulverlag: Zu¨rich, Switzerland, 1996.
Docking. Docking experiments were carried out using Auto-
Dock434 in combination with the AutoDock-Tools35 GUI. A
gridbox (50 × 50 × 50 points with a grid spacing of 0.375 Å) was
placed over the binding site of the protein and centered slightly
(27) Hartleib, J.; Ru¨terjans, H. Biochim. Biophys. Acta 2001, 1546, 312–
324.
(28) Worek, F.; Reiter, G.; Eyer, P.; Szinicz, L. Arch. Toxicol. 2002, 76,
523–529.
(40) Schuettelkopf, A. W.; van Aalten, D. M. F. Acta Crystallogr., Sect.
D: Biol. Crystallogr. 2004, 60, 1355–1363.
(29) Ellman, G. L.; Courtney, K. D.; Andres, V.; Feather-Stone, R. M.
Biochem. Pharmacol. 1961, 7, 88–95.
(41) Harvey, S.; Kolakowski, J.; Cheng, T.; Rastogi, V.; Reiff, L. P.;
DeFrank, J. J.; Raushel, F. M.; Hill, C. Enzyme Microb. Technol. 2005,
37, 547–555.
(30) Worek, F.; Mast, U.; Kiderlen, D.; Diepold, C.; Eyer, P. Clin. Chim.
Acta 1999, 288, 73–90.
(42) Ordentlich, A.; Barak, D.; Kronman, C.; Benschop, H. P.; De Jong,
L. P.; Ariel, N.; Barak, R.; Segall, Y.; Velan, B.; Shafferman, A.
Biochemistry 1999, 38, 3055–3066.
(31) Eyer, P.; Worek, F.; Kiderlen, D.; Sinko, G.; Stuglin, A.; Simeon-
Rudolf, V.; Reiner, E. Anal. Biochem. 2003, 312, 224–227.
(32) Hart, G. J.; O’Brien, R. D. Biochemistry 1973, 12, 2940–2945.
(33) Reiter, G.; Koller, M.; Thiermann, H.; Dorandeu, F.; Mikler, J.; Worek,
F. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2007, 859,
9–15.
(43) Albaret, C.; Lacoutie`re, S.; Ashman, W. P.; Froment, D.; Fortier, P. L.
Proteins 1997, 28, 543–555.
(44) Shafferman, A.; Barak, D.; Stein, D.; Kronman, C.; Velan, B.; Greig,
N. H.; Ordentlich, A. Chem. Biol. Interact. 2008, 175, 166–172.
9
17228 J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009

Enantioselectivity of DFPase against OP Nerve Agents
A R T I C L E S
Figure 2. Docked orientation of (RP)-GF in the active site of WT DFPase.
Docking of (SP)-GF shows a similar orientation but with the phosphorus-
bound fluoride and methyl groups exchanged.
Rational Design of Enhanced DFPase Mutants. The DFPase
active site is a quasisymmetrical elongated cleft with the
catalytic calcium ion bound on the bottom of the binding pocket.
One of the determinants for the binding orientation is the
preference of the large alkoxy chain of the substrate to bind in
either the left or the right side of the binding pocket (left,
pointing toward T195; right, pointing toward I72). The docking
reveals that in WT DFPase these side chains prefer the left side
of the pocket. Therefore, our engineering efforts focused on
limiting the space of the left side of the pocket forcing the alkoxy
chain of the substrate to the right side by introducing a larger
residue at position T195. The first reaction step in the catalytic
cycle of DFPase is the attack of D229 on the substrate
phosphorus atom. This leads to the formation of the phospho-
enzyme intermediate. During the course of this first step a
Walden inversion on the phosphorus atom takes place. This
inversion is accompanied by movement and rearrangement of
the substituents on the phosphorus atom, especially the large
alkoxy group. To facilitate these movements a more open
binding pocket was generated by removing steric constraints
in the vicinity of the leaving group (E37, Y144, R146). The
final quadruple mutant generated was E37A/Y144A/R146A/
T195M (Mut1).
Structure and Activity of Designed DFPase Mutants. Mut1
and another mutant containing an aspartate at position 37 (E37D/
Y144A/R146A/T195M; Mut2) were crystallized successfully
and their crystal structures determined (PDB: 3HLH for Mut1,
3HLI for Mut2). Data collection and refinement statistics are
shown in Table 1. Crystals of the two mutants grew in a
previously uncharacterized space group (P21 compared to
P212121 for the WT), containing four molecules in the asym-
metric unit. Nevertheless, structural comparison of WT DFPase
(PDB: 1E1A) and the four mutant molecules in the asymmetric
unit reveals only minor differences (0.2-0.3 Å rmsd), localized
to loop regions and crystal contact areas. The Cꢀ atoms of the
mutated amino acids in Mut1 (A37, A144, A146) superimpose
with the Cꢀ atoms of the original side chains in the WT (E37,
Y144, R146). Cꢀ and Cγ of the new side chain of M195
(formerly T195) can be superimposed with the Cꢀ and Cγ of
the original side chain in the WT. This coincides with another
observation made with structurally characterized mutants of
DFPase: the protein backbone is highly resistant to structural
changes upon side chain mutations. This feature of DFPase is
especially useful for targeting specific mutations for rational
design efforts. Despite the overall structural rigidity of the
Figure 1. (a) Hydrolysis of enantiomers of GF by WT DFPase. (b)
Hydrolysis of enantiomers of GF by Mut1. (c) Hydrolysis of enantiomers
of GF by Mut2. (Data for substrates GB and GD can be found in the
Supporting Information.)
in order to allow an in-line attack of the free D229 oxygen on
the phosphorus.5 The lowest energy orientations of the two
enantiomers of GF bound to DFPase can be deduced from
Figure 2. The figure shows (RP)-GF in the active site of WT
DFPase. Docking of (SP)-GF shows a similar orientation but with
the phosphorus-bound fluoride and methyl groups exchanged. An
orientation that would allow a subsequent reaction can only be
found for the less toxic RP enantiomer of GF while the binding
mode of the more toxic SP enantiomer does not allow an in-line
attack of D229 on the phosphorus atom. Similar results are obtained
for the stereoisomers of GD while the small size of GB allows
productive orientations for both enantiomers. This is in agreement
with the results obtained using the biological assay.
9
J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009 17229

A R T I C L E S
Melzer et al.
Table 1. Data Collection and Refinement Statistics
Mut1
Mut2
PDB code
source
3HLH
3HLI
ESRF ID-14,
λ ) 0.933 Å
P21
a ) 67.11 Å
b ) 74.39 Å
c ) 118.64 Å
ꢀ ) 99.9°
1.8 Å
0.136 (0.458)
5.0 (1.6)
99.2% (98.7)
3.8
ESRF BM16,
λ ) 1.007 Å
P21
a ) 66.95 Å
b ) 74.49 Å
c ) 118.95 Å
ꢀ ) 100.0°
1.4 Å
0.053 (0.308)
18.8 (4.4)
94.5% (86.3)
3.1
space group
unit cell
resolution
Rsym
〈I/σ(I)〉
completeness
redundancy
indep reflections
105813
213535
(incl 5332 for Rfree
0.218/0.195
0.005 Å
)
(incl 10622 for Rfree
0.221/0.205
0.004 Å
)
Rfree/Rcryst
rmsd bond lengths
rmsd bond angles
B-factor
1.484°
1.455°
11.3 Å2
12.2 Å2
outer shell
(1.90-1.80 Å)
(1.45-1.40 Å)
Table 2. Specific Enzymatic Activities of DFPase WT, Mut1, and
Mut2 against 0.1% DFP, GB, GD, and GF
WT
Mut1
Mut2
DFP (8.2 mM)a
sarin (GB, 14.3 mM)
soman (GD, 11.0 mM)
cyclosarin (GF, 10.0 mM)
247 ( 10b
110 ( 6
390 ( 21
862 ( 32
192 ( 8
429 ( 21
418 ( 3
117 ( 7
143 ( 8
82 ( 14
198 ( 15
258 ( 13
a Substrate concentrations in parentheses. b U/mg; 1 U ) 1 µmol/min.
Figure 3. (a) Degradation of GF using WT DFPase monitored by chiral
GC/MS showing the preference for less toxic R-(+)-GF. (b) Similar results
for the degradation of GF using Mut1, showing preference for the more
toxic S-(-)-GF.
mutant structures, there is also one distinct difference relative
to the WT. B-factors for residues in the loop region 141-151
of Mut1 and Mut2, which form part of the active site rim in
DFPase, are elevated relative to the WT. This suggests a higher
mobility of this loop in the mutants, which can be attributed to
the two mutations in this region (Y144A, R146A). The
replacement of the large arginine and tyrosine side chains by
alanine reduces the contact area and removes torsional and steric
restraints, increasing the plasticity of the DFPase active site.
Mut1 shows a pronounced preference for the more toxic SP
stereoisomers of GF (Figure 1b) as well as for GB and GD
combined with a substantially enhanced enzymatic activity
(Table 2). The preference of DFPase for the more toxic
enantiomer of GF was also demonstrated by the use of a
modified GC/MS assay with a chiral column that allows for
the baseline separation of both enantiomers.33 Compared to the
indirect biological assay this method can directly visualize the
disappearance of the two individual enantiomers. GF was
hydrolyzed both with WT DFPase and Mut1, and samples were
taken at intervals and subjected to chiral GC separation and MS
detection. The more toxic S-(-)-GF eluted first and the less toxic
R-(+)-GF second. WT DFPase prefers the less toxic stereoisomer
leading to a faster degradation of this substrate (Figure 3a) while
Mut1 prefers the more toxic stereoisomer (Figure 3b). Mut2 is
slightly less active than Mut1, particularly in the case of GB where
only half the specific activity of Mut1 is observed. Attempts to
determine exact kinetic constants (KM, kcat) for the mutants with
substrates GB, GD, and GF suffered from the low affinity of the
substrates. Due to safety restrictions the maximum substrate
concentration that could be used was 0.2% in the final reaction
solution (approximately 10 mM). This is still below the KM of Mut1
and Mut2 as at the maximum substrate concentration the reaction
rate was still linearly dependent on the agent concentration and
therefore [S] , KM (data not shown). Even though kcat and KM
cannot be determined individually under these conditions, it is still
possible to determine kcat/KM if a single substrate is present. In the
present case, however, the substrate stereoisomers act as substrates
as well as competitive inhibitors of each other. Yeung et al. have
presented a mathematical model for the determination of kinetic
constants for the four stereoisomers of GD hydrolyzed by the
enzyme paraoxonase.15 The biological assay based on inhibition
of AChE used in this study is only able to discriminate the pair of
GD stereoisomers with a different configuration of the phosphorus
atom, as this is the dominant stereocenter that determines toxicity,
but is unable to individually detect all four steoroisomers. We have
only used data for GB and GF (both with Mut1 and Mut2) to
determine kcat/KM using a version of the mathematical model for
GD/paraoxonase for a system with only one pair of enantiomers.
kcat/KM for each GB and GF enantiomer was determined by
computationally fitting the variables to the model equations and
values are listed in Table 3.
Analyzing the kcat/KM values for GB, we can see that the
values for both enantiomers of GB are almost identical in case
of the wild type (ratio S/R ) 0.89). Under the experimental
conditions where [S] , KM it can be shown that the ratio of the
kcat/KM values is equal to the ratio of the reaction rates of the
two enantiomers if they are present at equal concentrations (as
in the case of a racemic mixture). Both mutants show a
preference for the more toxic (SP)-GB, and this preference is
more pronounced for Mut1. As can be seen from Table 3 this
is achieved by an enhanced kcat/KM for (SP)-GB and a decreased
value for (RP)-GB (ratio S/R Mut1 ) 5.11; Mut2 ) 2.96). In
case of GF, the DFPase WT shows a distinct preference for the
less toxic enantiomer (RP)-GF (ratio S/R ) 0.02, Figure 1a).
9
17230 J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009

Enantioselectivity of DFPase against OP Nerve Agents
A R T I C L E S
Table 3. kcat/KM for DFPase WT, Mut1, and Mut2 with Substrates DFP, GB, and GF and KM and kcat for Substrate DFP
WT
Mut1
Mut2
DFP
KM
kcat
kcat/KM
3.76 ( 0.05 mM
211 ( 9 s-1
2.12 ( 0.07 mM
291 ( 7 s-1
3.93 ( 0.14 mM
376 ( 13 s-1
a
b
b
b
b
5.6 × 104 M-1 s-1
1.4 × 105 M-1 s-1
9.6× 104 M-1 s-1
(R)-GB
(S)-GB
(R)-GF
(S)-GF
kcat/KM
kcat/KM
kcat/KM
kcat/KM
4.7 ( 0.6 × 104 M-1 s-1
4.5 ( 0.4 × 104 M-1 s-1
2.4 ( 0.1 × 104 M-1 s-1
ratio S/R
ratio S/R
0.89
5.11
2.96
4.2 ( 1.2 × 104 M-1 s-1
2.3 ( 0.3 × 105 M-1 s-1
7.1 ( 0.6 × 104 M-1 s-1
7.2 ( 1.6 × 105 M-1 s-1
1.3 ( 0.1 × 105 M-1 s-1
2.7 ( 0.4 × 104 M-1 s-1
0.02
3.77
0.89
1.7 ( 0.5 × 104 M1 s-1
4.9 ( 0.3 × 105 M-1 s-1
2.4 ( 0.2 × 104 M-1 s-1
a Determined by direct calculation from KM and kcat (KM and kcat determined from at least three independent experiments). b Determined by fitting kcat
KM for both enantiomers (determined from at least three independent experiments).
/
Figure 4. Degradation of GF using WT DFPase, Mut1, and mixtures (1:2
and 1:1 of WT and Mut1).
Figure 5. Model depicting the preferred orientation of the substrate alkoxy
groups for the phosphoenzyme intermediate in the binding site of Mut1.
The phosphorus-bound methyl group takes the position of an alkoxy oxygen
atom (omitted for clarity).
This preference is reversed in the case of Mut1 (ratio S/R )
3.77, Figure 1b), but in the case of Mut2 both enantiomers are
hydrolyzed with almost the same preference (ratio S/R ) 0.89,
Figure 1c).
stereoisomers of GB, GD, and GF, but the results were
inconclusive regarding a distinct preference for the “left” and
“right” side of the binding pocket compared to the WT. To find
out which mutational site mainly contributes to the reversed
enantioselectivity of the Mut1 the single mutants T195M and E37A
and the double mutant Y144A/R146A were generated and tested
with GF (see Supporting Information). None of these mutants
displays the behavior of Mut1, although the preference for the less
toxic enantiomers is less pronounced in T195M. Interestingly also
E37A does not lead to a different selectivity compared to the WT.
As shown above, Mut2 (containing the mutation E37D) only
leads to an equal preference for both enantiomers in the case
of GF, while Mut1 (containing mutation E37A) displays a
reversed preference. To achieve the final selectivity of Mut1
all mutations are required. For DFPase this result can be
explained structurally by modeling the different phosphoenzyme
intermediates. In addition to the effect of T195 M introducing
steric hindrance in the “left” side of the binding site, the other
mutants open the pocket to facilitate the accommodation of the
large alkoxy side chains of the substrates (Figure 5). It appears
that the phosphoenzyme intermediate formed by the toxic SP
substrates is in an energetically more favorable conformation
than those for the RP substrates. Finally, the more open binding
pocket should facilitate the structural rearrangements during the
Walden inversion on the phosphorus atom that goes along with
the nucleophilic attack by D229. Recent computational studies
have revealed that this inversion does not have to occur in a
WT DFPase and Mut1, both acting on the same chiral
substrates, not only show reversed enantioselectivity but also
have very different kcat and KM values. While the WT displays
a lower KM value compared to Mut1, which is important for
agent degradation at very low substrate concentrations, it is
clearly inferior in terms of kcat, which determines the maximum
achievable reaction rate of the enzyme. Thus, Mut1 is preferred
at higher substrate concentrations, where its preference for the
toxic enantiomers and the higher kcat will lead to rapid
detoxification. At lower substrate concentrations, the WT is
preferred, due to the better KM of the WT enzyme. To investigate
this issue, WT and Mut1 as well as 1:1, 1:2, and 2:1 mixtures
(with the same total enzyme content) were used to completely
hydrolyze a solution of 660 µM GF (Figure 4). The biphasic
hydrolysis curve for the WT is clearly visible. Although the
KM of Mut1 is much higher than the KM of WT, the use of
Mut1 reduces the required time for hydrolysis by more than
50%. Using mixtures of WT and Mut1 reduces the reaction time
even further, but apparently the effect of the ratio is minimal in
the range 1:2-2:1 (the 2:1 curve was omitted from Figure 3
for clarity but is virtually superimposable to that of the 1:1
mixture). These results clearly show that addition of DFPase
WT to Mut1 partially overcomes the effect of the unfavorable
KM of Mut1 at low substrate concentrations.
With the structures of both mutants available, docking was
used again to investigate the binding orientations of the
9
J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009 17231

A R T I C L E S
Melzer et al.
concerted manner but is also possible in a stepwise fashion.45
Studies employing QM/MM methods are currently underway
to gain a further understanding of the reaction step leading to
the phosphoenzyme intermediate.
scavenger. As the concentration of nerve agents is low in human
plasma, even in cases of severe poisoning, a much smaller KM
is required for efficient enzyme activity than that displayed by
WT DFPase and the mutants.55
The reason for choosing a rational design strategy for the
engineering of DFPase was, as pointed out above, the toxic
nature of the substrates that makes the screening of large mutant
libraries virtually impossible. At the same time, substrate
mimics, due to the presence of reporter groups and chro-
mophores, are sometimes flawed surrogates for the real sub-
strates. The creation of the DFPase mutants was aided by the
wealth of structural data for the protein, detailed knowledge
about the enzyme mechanism, and the unusual rigidity of the
protein backbone, which helps to predict mutant structures and
the effect of point mutations. It should be noted, however, that
the rational design approach employed here is unlikely to
generate an optimal enzyme mutant as the number of sampled
mutants is still small compared to the large libraries used in
evolutionary approaches.
Using rational design two quadruple mutants of DFPase have
been constructed with a reversed enantioselectivity for G-type
nerve agents. These mutants also display an enhanced enzymatic
activity against the tested substrates. Both effects lead to a rapid
detoxification of the reaction solution. Even though several
articles in the literature deal with the enantioselectivity of nerve
agent degrading enzymes like mammalian PON114-18 and PTE
from B. diminuta,56-59 they deal only with substrates mimicking
the nerve agent, and only one PTE mutant with a reversed
enantioselectivity against GF has been reported so far.41 This
was achieved by slowing down the turnover of the less toxic
enantiomer resulting in a rather low overall activity. In contrast,
the mutants reported here demonstrate an enhanced overall activity
and enantioselectivity that allow for the practical use of DFPase
for rapid detoxification. This highlights the significance of our
findings and clearly shows that enzyme engineering efforts to
enhance the activity and selectivity of DFPase are leading to
promising results and that further optimizations seem feasible.
Discussion
A survey of the literature reporting engineering enantiose-
lectivity in enzymes shows that rational design is still used but
that evolutionary approaches tend to dominate the field. A class
of enzymes where both approaches have been used are the
lipases. Lipases are valuable enzymes for the creation of
enantiomerically pure compounds, especially secondary alco-
hols, which are important building blocks in organic synthesis.
Examples of a rational design approach include early work by
Hirose et al.46 focusing on the hydrolysis of 1,4-dihydropyridines
and a more recent study by Magnusson et al.47 focusing on
secondary alcohols. Examples of evolutionary approaches can
be found in the work of Zha et al.48 using error prone PCR and
DNA shuffling and Koga et al.49 using single molecule PCR
linked in Vitro expression (SIMPLEX). These works show that
both strategies can lead to enzyme variants with reversed
enantioselectivity. Koga et al. point out, however, that improving
activity or enantioselectivity by rational design remains chal-
lenging. The reasons for this are that the structural and functional
effects of point mutations are not fully predictable and that site-
directed mutagenesis is a slow and labor-intensive process such
that only a limited number of variants can be produced and
tested. Thus, evolutionary strategies are preferred if an assay
for screening the mutant libraries exists that can be automated
and used for high-throughput screening. Several examples with
other enzymes using evolutionary methods can be found in the
literature,50-52 but also hybrid strategies have been reported,
by using error prone PCR to generate libraries and point
mutations to refine the initial hits from screening.53
Rational approaches require detailed knowledge about enzyme
structure, substrate binding, and reaction mechanism. In the
absence of this information, related enzymes with a highly
conserved active site environment and similar but stereochemi-
cally reversed substrate specificity can be used as templates.
This was demonstrated for vanillyl-alcohol oxidase (VAO) using
a similar p-cresol methylhydroxylase (PCMH).54 Analogously,
the DFPase mutants reported in this study may be used to guide
the engineering of the related enzyme paraxonase (PON).13
Human PON1 has been considered as a catalytic in ViVo
Acknowledgment. The German Ministry of Defense supported
this work under contract E/UR3G/6G115/6A801. Part of this work
was carried out by M. Melzer in partial fulfillment of the requirements
for a medical doctoral degree at the Johannes Gutenberg University,
Mainz, Germany. We thank the staff at ESRF for assistance with data
collection. J.C.-H.C. was funded by the Hessian Ministry for Science
and Culture.
(45) van Bochove, M. A.; Swart, M.; Bickelhaupt, F. M. Phys. Chem. Chem.
Phys. 2009, 11, 259–267.
Supporting Information Available: Detailed experimental
procedures and supplementary results, including reaction curves
for additional substrates and mutants as well as data to prove
the accuracy of the analytical methods. This information is
available free of charge via the Internet at http://pubs.acs.org.
(46) Hirose, Y.; Kariya, K.; Nakanishi, Y.; Kurono, Y.; Achiwa, K.
Tetrahedron Lett. 1995, 36, 1063–1066.
(47) Magnusson, A. O.; Takwa, M.; Hamberg, A.; Hult, K. Angew. Chem.,
Int. Ed. 2005, 44, 4582–4585.
(48) Zha, D.; Wilensek, S.; Hermes, M.; Jeager, K. E.; Reetz, M. T. Chem.
Commun. 2001, 2664–2665.
JA905444G
(49) Koga, Y.; Kato, K.; Nakano, H.; Yamane, T. J. Mol. Biol. 2003, 331,
585–592.
(50) May, O.; Nguyen, P. T.; Arnold, F. H. Nat. Biotechnol. 2000, 18,
317–320.
(55) Josse, D.; Lockridge, O.; Xie, W.; Bartels, C. F.; Schopfer, L. M.;
Masson, P. J. Appl. Toxicol. 2001, 21, S7–S11.
(51) Ivancic, M.; Valinger, G.; Gruber, K.; Schwab, H. J. Biotechnol. 2007,
129, 109–122.
(56) Chen-Goodspeed, M.; Sogorb, M. A.; Wu, F.; Hong, S. B.; Raushel,
F. M. Biochemistry 2001, 40, 1325–1331.
(52) Bartsch, S.; Kourist, R.; Bornscheuer, U. T. Angew. Chem., Int. Ed.
2008, 47, 1508–1511.
(57) Chen-Goodspeed, M.; Sogorb, M. A.; Wu, F.; Raushel, F. M.
Biochemistry 2001, 40, 1332–1339.
(53) Williams, G. J.; Woodhall, T.; Farnsworth, L. M.; Nelson, A.; Berry,
A. J. Am. Chem. Soc. 2006, 128, 16238–16247.
(54) van den Heuvel, R. H. H.; Fraaije, M. W.; Ferrer, M.; Mattevi, A.;
van Berkel, W. J. H. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 9455–
9460.
(58) Li, Y.; Aubert, S. D.; Raushel, F. M. J. Am. Chem. Soc. 2003, 125,
7526–7527.
(59) Hill, C. M.; Li, W. S.; Thoden, J. B.; Holden, H. M.; Raushel, F. M.
J. Am. Chem. Soc. 2004, 125, 8990–8991.
9
17232 J. AM. CHEM. SOC. VOL. 131, NO. 47, 2009