Enhancing activity and controlling stereoselectivity in a designed PLP-dependent aldolase

Article abstract of DOI:10.1002/anie.200700710

Chop and change: Site-directed mutagenesis has been employed to optimize the activity of an engineered pyridoxal phosphate-dependent aldolase and to invert its inherent threo selectivity in the cleavage of D-β-phenylserines. The modification of the active-site residues generates significant retroaldol activity that compares favorably with that of natural enzymes in terms of efficiency and selectivity.

Full text of DOI:10.1002/anie.200700710

Communications
DOI: 10.1002/anie.200700710
Aldolase Optimization
Enhancing Activity and Controlling Stereoselectivity in a Designed
PLP-Dependent Aldolase**
Miguel D. Toscano, Manuel M. Müller, and Donald Hilvert*
Biocatalysis is increasingly seen as a viable option for
performing chemical transformations in the laboratory and
on an industrial scale.[1] Although nature provides a wealth of
catalysts for such applications, natural enzymes may be
unavailable or otherwise unsuitable for specific reactions of
interest. For this reason, tailoring the properties of existing
enzyme scaffolds to access altered or completely new
activities has attracted considerable attention.[2]
Previously, we showed that a single active-site mutation is
sufficient to convert a pyridoxal phosphate (PLP)-dependent
alanine racemase from Geobacillus stearothermophilus into
an aldolase.[3] The substitution of tyrosine at position 265 in
this protein (Figure 1a) with alanine removes a catalytic
residue that is essential for racemase activity and, at the same
time, creates a cavity that accommodates d-configured b-
phenylserine isomers (Figure 1b). Native racemase activity is
decreased by greater than 103-fold at the modified active site,
whereas retroaldol cleavage of the new substrate to give
benzaldehyde and glycine (Scheme 1) is accelerated by five
Scheme 1. Mechanism of the enzyme-catalyzed retroaldol reaction of
d-b-phenylserine. B=base.
orders of magnitude. The Tyr265Ala variant also cleaves a-
disubstituted b-hydroxy amino acids, an activity that is not
reported for natural PLP-dependent aldolases.[4]
Molecular modeling shows that the side chain of d-b-
phenylserine in the aldimine complex can fit comfortably in
the space left vacant by the Tyr265Ala mutation, oriented so
À
that its Ca Cb bond is orthogonal to the PLP plane and hence
activated for scission (Figure 1b).[4] Although the alanine
substitution engenders the desired activity, this amino acid
may not be the optimal replacement for tyrosine. Conse-
quently, we systematically varied the residue at position 265.
As summarized in Table 1, introduction of serine, valine, or
glutamate at this site is detrimental for retroaldol activity,
whereas arginine has opposing effects on kcat and kcat/Km. In
contrast, the Tyr265Lys substitution results in 9- and 2-fold
increases in kcat and kcat/Km, respectively, for the cleavage of
d-b-phenylserines.[5] Although this result cannot be fully
rationalized in the absence of a structure or pre-steady-state
kinetic data, the flexible lysine side chain presumably
[*] Dr. M. D. Toscano, M. M. Müller, Prof. Dr. D. Hilvert
Figure 1. Active-site models of G. stearothermophilus alanine racemase
Laboratorium für Organische Chemie
ETH Zürich
Hönggerberg HCI F339, 8093 Zürich (Switzerland)
Fax: (+41)44-632-1486
variants.[14] Structures of the external aldimines between l-alanine
(cyan) and PLP (pink) complexed to the wild-type active site (a), the
(2R,3S)-b-phenylserine-PLP aldimine (green) complexed to the Tyr265-
Ala (b) and Tyr265Lys (c) variants, and the (2R,3R)-b-phenylserine-PLP
aldimine bound at the Met134Phe/Tyr265Lys/Ile352Trp active site (d)
E-mail: hilvert@org.chem.ethz.ch
[**] The financial support of the Schweizerischer Nationalfonds and the
Defense Advanced Research Projects Agency (DARPA) is gratefully
acknowledged.
are shown. Mutations at positions 134, 265, and 352 (red) reduce the
free space surrounding the aldimine intermediate, improving the
surface complementarity of the active site. The hydrogen bond
between the Cb-hydroxy group of the substrate and the phosphate
group of PLP (the proposed catalytic base) is shown as a white dashed
line.
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author. PLP=pyr-
idoxal phosphate.
4468
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 4468 –4470

Angewandte
Chemie
phenylserine substrates.[6,7] Inter-
Table 1: Steady-state parameters for the enzymatic conversion of d-b-phenylserine isomers to
benzaldehyde and glycine.[a]
estingly, the presence of an addi-
tional methyl group at Ca dimin-
ishes the ability of the reengi-
neered racemase to discriminate
between the different b-isomers,[4]
probably because interactions with
this extra substituent stabilize pro-
ductive orientations of the nor-
mally less-favored aldimine com-
plex.
(2R,3S)-b-phenylserine
(2R,3R)-b-phenylserine
Enzyme
kcat
Km
[mm]
k
cat/Km
kcat
Km
k
cat/Km
Selectivity[b]
9.2
[minÀ1
]
[mÀ1 minÀ1
]
[minÀ1
]
[mm] [mÀ1 minÀ1
]
Alr-WT[c]
Y265A5.7
Y265S
Y265V
Y265E
Y265R
Y265K
M134F/Y26K
Y265K/I352W
M134F/Y26K/I352W
0.0029[c]
670[c]
28
–
–
–
–
73
[c]
8.5[c]
43
–
0.044
n.d.
n.d.
n.d.
n.d.
0.43
3.4
0.60
n.d.
n.d.
n.d.
n.d.
2.3
1.2
–
n.d.
n.d.
n.d.
n.d.
190
620
–
–
–
13
52
31
48
8.4
73
35
17
15
5.9
180
1500
1800
3200
1400
7.9
2.9
2.3
0.6
The preferred conformations of
the two b-phenylserine aldimine
diastereomers differ mainly in a
5.5
0.83
2.7
0.60 1400
1.2 2300
À
308 rotation around the Ca Cb
[a] Assays were performed at 308C in 100 mm 2-(4-(2-hydroxyethyl)-1-piperazinyl)ethanesulfonic acid
(HEPES) buffer solution (pH 8.0). The standard error for all kinetic parameters is less than 20%. (–
indicates that no activity was detected above background, n.d.=not determined). For comparison,
bond.[4] Formation of a productive
hydrogen bond between the b-
hydroxy group and the catalytically
important phosphate group of the
cofactor requires swinging the aryl
ring of the less-favored (2R,3R)-
aldimine isomer closer to His166
than in the case of the
(2R,3S) isomer. Given this, we
wondered whether the inherent
threo selectivity of the engineered
aldolase could be inverted by alter-
E. coli l-threonine aldolase[5] catalyzes the conversion of (2S,3R)-b-phenylserine with kcat =225 minÀ1
,
/
kcat/Km =1.9”106 mÀ1 minÀ1, and the conversion of (2S,3S)-b-phenylserine with kcat =337 minÀ1, kcat
Km =1.4”106 mÀ1 minÀ1; A. xylosoxidans d-threonine aldolase[6] catalyzes the conversion of (2R,3S)-b-
phenylserine with kcat =1900 minÀ1, kcat/Km =1.9”106 mÀ1 minÀ1, and the conversion of (2R,3R)-b-
phenylserine with kcat =814 minÀ1, kcat/Km =1.4”106 mÀ1 minÀ1. The kcat values were estimated from the
Vmax (UmgÀ1) values reported in references [5] and [6] assuming 1 U catalyzes the formation of
1 mmolminÀ1 of product, and Mr =36495 Da for the E. coli enzyme and 42195 Da for the A. xylosoxidans
enzyme; [b] The selectivity was calculated as [kcat/Km (2R,3S)]/[kcat/Km (2R,3R)]; [c] Reference [3].
improves the packing and surface complementarity of the
substrate binding pocket; its cationic terminus may also
engage in favorable cation–p interactions with the aryl ring of
the substrate or help stabilize developing negative charge at
the more distant b-alcohol in the transition state (Figure 1c).
Whatever the ultimate origin of the improvement, it is
notable that the kcat value for the optimized enzyme is only
five times smaller than that of l-threonine aldolase from
Escherichia coli[6] and 40-fold smaller than that of a promis-
cuous d-threonine aldolase from Alcaligenes xylosoxidans[7]
for conversion of b-phenylserines. These findings highlight
the potential of single active-site mutations for remodeling
the chemical properties of existing enzymes while under-
scoring the fact that the “obvious” substitution is not
necessarily the best.
ing residues 134 and 352, which flank the pocket into which
the aryl group docks. For example, binding of the
(2R,3R) isomer might be enhanced by altering the packing
interactions on one side of the pocket (Met134Phe) while
increasing steric bulk on the other (Ile352Trp; Figure 1d). In
fact, when combined with the Tyr265Lys mutation, both
changes improve retroaldol cleavage of the (2R,3R) diaster-
eomer relative to that of the (2R,3S) diastereomer (Table 1).
The triple mutant that contains all three changes exhibits the
highest overall catalytic efficiency for (2R,3R)-b-phenylserine
(a 12-fold increase over Tyr265Lys and a 31-fold increase over
Tyr265Ala, which are achieved mostly through an improved
kcat parameter). Because these substitutions increase kcat/Km
for the d-threo substrate only slightly when compared with
the Tyr265Ala variant (approximately twofold), they effec-
tively reverse selectivity and lead to preferential retroaldol
cleavage of the erythro isomer by a factor of approximately
2:1. The diastereoselectivity of this catalyst is thus directly
responsive to the residues that line the active site and
therefore subject to rational manipulation. Nevertheless,
further enhancement of the erythro selectivity of the triple
mutant, or the inherent threo selectivity of the starting
catalyst, will likely require more extensive mutagenesis at
sites distant from the binding pocket.[10]
Because the catalytic base that initiates the retroaldol
reaction is most likely the phosphate group of the cofactor,[4]
À
C C bond cleavage should be sensitive to the configuration of
the alcohol at the Cb atom as well as to rotation around the
À
Ca Cb bond. We therefore tested (2R,3R)- and (2R,3S)-b-
phenylserine as substrates for our engineered aldolases.[8] The
relative kcat/Km values for the starting Tyr265Ala variant
indicate a 9:1 preference for the d-threo isomer, and this
preference is not significantly eroded in the kinetically
superior Tyr265Lys variant (8:1; Table 1). With b-phenyl-
serine, the engineered aldolases are thus substantially more
diastereoselective than many natural PLP-dependent aldo-
lases, which exhibit stringent stereoselectivity at the Ca atom
in the cleavage of b-hydroxy amino acids but poor stereo-
chemical control at the Cb atom.[6,7,9] The E. coli and
A. xylosoxidans aldolases, for instance, achieve threo:erythro
selectivities of only about 1.4 with their respective l- and d-b-
These experiments illustrate the adaptive potential of the
alanine racemase scaffold. Modification of the first shell of
active-site residues generates significant retroaldol activity
that compares favorably in terms of efficiency and selectivity
with natural enzymes that have evolved specifically to
promote this transformation. Given the importance of b-
hydroxy-a-amino acids as bioactive agents[11] and building
blocks for pharmaceutically important natural products, such
Angew. Chem. Int. Ed. 2007, 46, 4468 –4470
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 4469

Communications
as vancomycin[12] and thiamphenicol,[13] and the relatively
poor diastereoselectivity of many natural PLP-dependent
aldolases,[6,7,9] further optimization of the properties of these
engineered aldolases may afford practical catalysts for kinetic
[7] J. Q. Liu, M. Odani, T. Yasuoka, T. Dairi, N. Itoh, M. Kataoka, S.
Shimizu, H. Yamada, Appl. Microbiol. Biotechnol. 2000, 54, 44.
[8] (2R,3S)-b-Phenylserine was purchased as a racemic mixture of
d- and l-threo-b-phenylserine isomers. In the presence of the
aldolases, only the d-isomer is converted to product (refer-
ence [3]). The erythro isomer, (2R,3R)-b-phenylserine, was
synthesized from cinnamic acid in five steps following an
asymmetric dihydroxylation strategy: H. C. Kolb, M. S. Van-
Nieuwenhze, K. B. Sharpless, Chem. Rev. 1994, 94, 2483; D. L.
Boger, M. A. Patane, J. Zhou, J. Am. Chem. Soc. 1994, 116, 8544.
[9] T. Kimura, V. P. Vassilev, G.-J. Schen, C.-H. Wong, J. Am. Chem.
Soc. 1997, 119, 11734; J. Q. Liu, T. Dairi, N. Itoh, M. Kataoka, S.
Shimizu, H. Yamada, J. Biol. Chem. 1998, 273, 16678; H.
Misono, H. Maeda, K. Tuda, S. Ueshima, N. Miyazaki, S. Nagata,
Appl. Environ. Microbiol. 2005, 71, 4602.
[10] M. T. Reetz, Proc. Natl. Acad. Sci. USA 2004, 101, 5716; P. E.
Tomatis, R. M. Rasia, L. Segovia, A. J. Vila, Proc. Natl. Acad.
Sci. USA 2005, 102, 13761.
[11] W. Maruyama, M. Naoi, H. Narabayashi, J. Neurol. Sci. 1996,
139, 141; M. Martineau, G. Baux, J.-P. Mothet, Trends Neurosci.
2006, 29, 481.
[12] B. K. Hubbard, C. T. Walsh, Angew. Chem. 2003, 115, 752;
Angew. Chem. Int. Ed. 2003, 42, 730.
[13] J. Q. Liu, M. Odani, T. Dairi, N. Itoh, S. Shimizu, H. Yamada,
Appl. Microbiol. Biotechnol. 1999, 51, 586.
À
resolutions or, in the synthetic direction, stereocontrolled C
C bond formation.
Received: February 15, 2007
Published online: May 7, 2007
Keywords: aldolases · amino acids · biocatalysis ·
.
rational design · stereoselectivity
[1] A. Schmid, J. S. Dordick, B. Hauer, A. Kiener, M. Wubbolts, B.
Witholt, Nature 2001, 409, 258; H. E. Schoemaker, D. Mink,
M. G. Wubbolts, Science 2003, 299, 1694.
[2] For recent reviews, see: T. M. Penning, J. M. Jez, Chem. Rev.
2001, 101, 3027; M. D. Toscano, K. J. Woycechowsky, D. Hilvert,
Angew. Chem. 2007, 119, 3274; Angew. Chem. Int. Ed. 2007, 46,
3212.
[3] F. P. Seebeck, D. Hilvert, J. Am. Chem. Soc. 2003, 125, 10158.
[4] F. P. Seebeck, A. Guainazzi, C. Amoreira, K.K. Baldridge, D.
Hilvert, Angew. Chem. 2006, 118, 6978; Angew. Chem. Int. Ed.
2006, 45, 6824.
[5] As seen for the Tyr265Ala variant, introduction of lysine at
position 265 decreases the racemase activity of the enzyme by
more than three orders of magnitude. Neither d- nor l-alanine
was found to be a substrate for the Tyr265Lys mutant.
[6] J. Q. Liu, T. Dairi, N. Itoh, M. Kataoka, S. Shimizu, H. Yamada,
Eur. J. Biochem. 1998, 255, 220. For diastereoselective d-
phenylserine synthases, see: J. Steinreiber, K. Fesko, C. Rei-
singer, M. Schürmann, F. van Assema, M. Wolberg, D. Mink, H.
Griengl, Tetrahedron 2007, 63, 918.
[14] The models of the enzyme complexes are based on earlier hybrid
docking and ab initio calculations performed with aldimines of
all four a-methyl-b-phenylserine diastereomers at the Tyr265Ala
active site (reference [4]). The (2R,3S)- and (2R,3R)-b-phenyl-
serine derivatives were superimposed on their a-methyl counter-
parts, and the additional point mutations (Tyr265Lys,
Met134Phe, and Ile352Trp) were introduced with the MacPymol
software (Delano Scientific LLC, 2006), manually choosing the
rotamers that exhibit no steric clashes.
4470 www.angewandte.org
ꢀ 2007 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2007, 46, 4468 –4470