Directed divergent evolution of a thermostable D-tagatose epimerase towards improved activity for two hexose substrates

Article abstract of DOI:10.1002/cbic.201402620

Functional promiscuity of enzymes can often be harnessed as the starting point for the directed evolution of novel biocatalysts. Here we describe the divergent morphing of an engineered thermostable variant (Var8) of a promiscuous D-tagatose epimerase (DTE) into two efficient catalysts for the C3 epimerization of d-fructose to d-psicose and of L-sorbose to L-tagatose. Iterative single-site randomization and screening of 48 residues in the first and second shells around the substratebinding site of Var8 yielded the eight-site mutant IDF8 (nine-fold improved k_cat for the epimerization of D-fructose) and the six-site mutant ILS6 (14-fold improved epimerization of L-sorbose), compared to Var8. Structure analysis of IDF8 revealed a charged patch at the entrance of its active site; this presumably facilitates entry of the polar substrate. The improvement in catalytic activity of variant ILS6 is thought to relate to subtle changes in the hydration of the bound substrate. The structures can now be used to select additional sites for further directed evolution of the ketohexose epimerase.

Full text of DOI:10.1002/cbic.201402620

DOI: 10.1002/cbic.201402620
Full Papers
Directed Divergent Evolution of a Thermostable d-
Tagatose Epimerase towards Improved Activity for Two
Hexose Substrates
[a]
[b]
[a]
[b]
Andreas Bosshart, Chee Seng Hee, Matthias Bechtold, Tilman Schirmer, and
Sven Panke*
[a]
Functional promiscuity of enzymes can often be harnessed as
the starting point for the directed evolution of novel biocata-
lysts. Here we describe the divergent morphing of an engi-
neered thermostable variant (Var8) of a promiscuous d-taga-
tose epimerase (DTE) into two efficient catalysts for the C3 epi-
merization of d-fructose to d-psicose and of l-sorbose to l-ta-
gatose. Iterative single-site randomization and screening of 48
residues in the first and second shells around the substrate-
binding site of Var8 yielded the eight-site mutant IDF8 (nine-
fold improved kcat for the epimerization of d-fructose) and the
six-site mutant ILS6 (14-fold improved epimerization of l-sor-
bose), compared to Var8. Structure analysis of IDF8 revealed
a charged patch at the entrance of its active site; this presuma-
bly facilitates entry of the polar substrate. The improvement in
catalytic activity of variant ILS6 is thought to relate to subtle
changes in the hydration of the bound substrate. The struc-
tures can now be used to select additional sites for further di-
rected evolution of the ketohexose epimerase.
Introduction
In recent decades, the directed evolution of proteins has
become a powerful tool to tailor enzymes for applications in
[
1]
industrial biocatalysis. Directed evolution—iterative cycles of
mutagenesis and screening of the resulting libraries for var-
[
2]
iants that exhibit desired traits —has been shown to afford
improving almost any property of importance for an industrial-
[
3]
ly useful biocatalyst, including thermostability, enantioselec-
[
4]
[5]
tivity, and catalytic rate.
The enzyme d-tagatose epimerase (DTE) catalyzes the inter-
[
6]
conversion of C3 ketohexose epimers (Figure 1). It constitutes
the central enzyme for biocatalytic access to rare monosac-
charides, which have recently attracted great interest as low
calorie sweeteners, chiral building blocks, and active pharma-
Figure 1. Schematic representation of all eight ketohexoses in a sugar cube.
Each of The x-axis relates C3-epimers (catalyzed by the enzyme d-tagatose
epimerase from P. cichorii, PcDTE, depicted as red arrows), the y-axis relates
C4 epimers, and the z-axis relates C5-epimers.
[
7]
ceutical ingredients. By combining DTE with just two addi-
tional isomerases, the whole set of 24 hexoses can be generat-
ed in a short cascade reaction from only four cheaply available
starting materials (d-glucose, d-fructose, d-galactose, and l-
[
8]
[9]
sorbose). DTEs from various organisms are known, however
none exhibits a catalytic rate on the respective substrate such
as to make it attractive for application in an industrial context.
No active DTE homologue that could serve as an optimal
starting point for improving catalytic efficiency from thermophil-
ic origin is known. There is a growing body of evidence that en-
zymes from thermophiles are more suitable as starting points
[10]
for directed evolution than enzymes from mesophiles. The
major reason for this is that most mutations (in particular, those
altering protein function) are destabilizing, thus increasing the
chance that the mutation will results in a protein that does not
[
a] A. Bosshart, Dr. M. Bechtold, Prof. S. Panke
Department of Biosystems Science and Engineering, ETH Zꢀrich
Mattenstrasse 26, 4058 Basel (Switzerland)
E-mail: sven.panke@bsse.ethz.ch
[10c]
fold correctly and therefore will not be functional.
Proteins
from thermophilic organisms on the other hand have to func-
tion at high temperatures, and thus exhibit a larger free energy
[
b] Dr. C. S. Hee, Prof. T. Schirmer
difference (DG ) between folded (native) and unfolded states
u
Focal Area of Structural Biology and Biophysics, Biozentrum
University of Basel
[10d,11]
than mesophilic enzymes.
This free folding energy can be
considered as a “stability reservoir” that can be consumed by
accumulating mutations that encode new protein functions but
are destabilizing in terms of free energy of folding. The larger
Klingelbergstrasse 70, 4056 Basel (Switzerland)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201402620.
ChemBioChem 0000, 00, 0 – 0
1
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
stability reservoir of thermostable enzymes therefore allows ex-
ploring a larger proportion of protein sequence space (“evolva-
screening of several thousand clones. Therefore the establish-
ment of a microtiter-plate-based screening procedure was
[
12]
[10a]
[3b]
bility”) before the limit of folding stability is reached.
necessary, as the HPLC-based assay described previously
These facts clearly suggest a two-steps process for DTE.
First, a mesostable enzyme that catalyzes the desired reaction
should be (thermo-)stabilized, and in a second step the ther-
mostable enzyme variant can be evolved towards the desired
seemed infeasible. Rhodobacter sphaeroides galactitol dehydro-
genase (RsGD) was previously described for the NADH-depen-
[16]
dent oxidation of galactitol to l-tagatose. l-tagatose, the C3
epimerization product of l-sorbose, can hence be detected by
using RsGD in the reverse direction (and following the oxida-
tion of NADH at 340 nm). Indeed, RsGD reliably detected l-ta-
gatose in the presence of l-sorbose (Figure S3B in the Sup-
porting Information), a prerequisite for the screening assay.
A dehydrogenase that specifically reduces d-psicose in the
presence of d-fructose was less readily available. Klebsiella pneu-
monia strain 3321 was reported to contain a ribitol dehydro-
genase that can specifically reduce d-psicose without reducing
(novel) function.
Following this strategy, we recently described the thermo-
stabilization of a dimeric DTE from the mesophile Pseudomo-
nas cichorii (PcDTE) by systematically optimizing the dimeric
interface interactions, thus generating a variant termed PcDTE
[
3b]
Var8.
This enzyme showed promiscuous enzymatic activity
for C3 epimerization with all four ketohexose pairs, and was
therefore an appropriate starting point for further catalytic rate
optimization. We decided to divergently evolve this thermosta-
ble Var8 for improved production of two rare ketohexoses: d-
psicose from d-fructose, and l-tagatose from l-sorbose. Both
products are rare hexoses that are of considerable interest as
[17]
d-fructose. The gene coding for this protein was not speci-
fied, but the genomes of several K. pneumoniae strains have
been sequenced. Based on sequence homology, a ribitol 2-de-
hydrogenase (KpRD) gene was identified in the genome of
K. pneumoniae MGH 48. It was codon optimized for expression
in Escherichia coli, chemically synthesized, and expressed in
E. coli (see Figure S1 for amino acid sequence). Purified KpRD
showed high specificity for the reduction of d-psicose with con-
comitant oxidation of NADH, and very low activity in the reduc-
tion of d-fructose, thus making it a suitable enzyme for the de-
termination of d-psicose in our screening assay (Figure S3A).
In summary, we established a microtiter-plate-based screen-
ing system for the reliable detection of d-psicose in the pres-
ence of d-fructose by KpRD, and for l-tagatose in presence of
l-sorbose by RsGD. These UV–visible assays allowed the rapid
screening of over 6000 clones from IDF libraries (improved for
d-fructose) and over 5000 clones from ILS libraries (improved
for l-sorbose)—a number that would be highly laborious to
achieve by HPLC-based screening.
[13]
low-calorie sweetener and as chiral building blocks.
Var8
showed modest catalytic activity towards the epimerization of
ꢀ
1
d-fructose to d-psicose (kcat 12.1 s at 308C) but poor activity
ꢀ
1
in the epimerization of l-sorbose to l-tagatose (kcat 0.24 s at
08C). Regarding a suitable engineering strategy for these
3
objectives, it has been shown that a limited number of amino
acid substitutions in the vicinity of the active site are sufficient
to change the substrate specificity and catalytic activity of
[
14]
a promiscuous enzyme, thus potentially limiting the required
effort to obtain improved biocatalysts substantially. According-
ly, we reasoned that a directed-evolution strategy based on
[
15]
iterative saturation mutagenesis (ISM) of residues that sur-
round the active site of Var8 would be a promising method to
improve the catalytic efficiency towards d-fructose or l-sor-
bose. Besides the obvious benefit of superior biocatalysts for
the production of rare sugars, the central position of DTE in
sugar interconversion inspired us investigate also the develop-
ment of catalytic parameters for other ketohexose substrates.
We describe the successful divergent evolution of thermo-
stable Var8 towards the efficient epimerization of d-fructose/d-
psicose (IDF variants) and l-sorbose/l-tagatose (ILS variants) by
targeting residues in the first and second spheres around the
active site. We characterized all variants of the two divergent
evolutionary trajectories in terms of substrate specificity and
catalytic activity towards six distinct ketohexoses. The crystal
structures of Var8 in its substrate-free form as well as the crys-
tal structure of the final variants, IDF8 in complex with d-fruc-
tose and ILS6 in complex with l-sorbose, gave insight into the
molecular basis of the altered enzymatic specificities.
Selecting sites for saturation mutagenesis of Var8
Based on the available crystal structure of PcDTE WT co-crystal-
[9d]
lized with d-fructose (PDB ID: 2QUN),
we selected the 22
amino acid residues that are within 5 ꢁ of the C3-atom of d-
fructose (first-sphere residues), and 28 residues between 5 and
10 ꢁ (second-sphere residues; Figure 2). From the residues of
the first-sphere we excluded five that either make up the cata-
II
lytic duo (E152 and E246), bind the metal ion Mn (D185), or
[9d]
stabilize the cis-enediol transition state (H188 and R217;
Figures 2 and S5). To reduce the screening effort further, we
[15]
applied a “short-cut” version of the established ISM method.
Instead of screening each of the remaining 17 sites separately
and then re-diversifying each beneficial site once the first mu-
tation had been fixed, we fixed the first beneficial site encoun-
tered in the first round and then used this variant to diversify
the next site (Figure S5)
Results and Discussion
Establishment of a high-throughput screening protocol
The screening effort that comes with large libraries is arguably
still the main limiting factor in directed evolution. A strategy
that considers only the 45 residues close to the active site
reduces the workload significantly, but this still requires the
Evolving PcDTE towards improved d-fructose epimerization
Upon screening the first shell of residues, only one mutation
(S37N) improved the conversion of d-fructose to d-psicose
&
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
2
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Figure 2. Residues targeted during divergent directed evolution are depict-
ed in PcDTE WT (PDB ID: 2QUN) with bound d-fructose (green sticks): resi-
dues within 5 ꢁ of the C3 of d-fructose (first-sphere residues) are colored
cyan, and residues that are between 5 and 10 ꢁ (second sphere) are colored
orange. Catalytic residues (E152 and E248) and highly conserved residues
(D185, H188 and R217) were excluded from screening (shown as sticks).
(1.4-fold compared to Var8). This mutation resulted in the first
variant, which was termed IDF1 (Figure S2).
When screening the second-shell residues of IDF1, mutation
H209V (IDF2) yielded 1.6-fold higher activity compared to IDF1;
mutation G39E (IDF3) increased the activity 2.7-fold. In round
four the silent mutation T9T (IDF4, codon ACC to ACT) in-
creased protein expression 1.3-fold. In the next rounds, muta-
tions A258D (IDF5, 1.6-fold improvement), T109N (IDF6, 1.4-
fold improvement), L212I (IDF7, 1.3-fold improvement), and
S256G (IDF8, 1.4-fold improvement, final variant) were fixed.
The total improvement in catalysis from Var8 to IDF8 as deter-
mined from crude cell lysate during the screening was 14.4-
fold, an 8.6-fold catalytic improvement (Figure 3A) and a (cal-
culated) 1.7-fold improvement in protein expression yield. In
order to assess whether the improvement in catalytic rate
came at the price of loss in thermostability, we determined
this parameter for each IDF variant. A gradual reduction of
thermostability in the course of directed evolution was indeed
observed, especially for mutations H209V (IDF2) and A258D
m
Figure 3. Development of enzyme kinetic parameters kcat and K for each
improved variant IDF and ILS, determined at 258C. A) kcat (squares) and K
values (circles) for Var8 and IDF variants with d-fructose as substrate. B) kcat
squares) and K (circles) for Var8 and ILS variants with l-sorbose as sub-
strate.
m
(
m
For substrates d-tagatose and d-sorbose, which are believed
[18]
to be the natural substrate pair of PcDTE no net decrease in
kcat was observed over the IDF evolution trajectory (Figure 4A);
however, the catalytic efficiency (k /K ) for these two epimers
cat
m
was significantly reduced when going from Var8 to IDF8 (Fig-
ure 4C). For substrates l-sorbose and l-fructose net catalytic
rate increased moderately, whereas catalytic efficiency re-
mained largely unchanged (Figure 4A and C).
20
(
IDF5; Figure S5). The overall loss (determined as DT , the
50
Evolving PcDTE towards improved l-sorbose epimerization
temperature at which 50% of activity could be retained after
2
0 min of incubation) was ꢀ11.88C for IDF8 compared to the
Screening the first shell for improved l-sorbose to l-tagatose
conversion yielded a single mutation (Q183H, variant ILS1; Fig-
ure 3B). This mutation increased specific l-sorbose to l-taga-
tose catalytic activity 8.2-fold, as determined with crude cell
lysate (not corrected for protein expression). We then screened
the second-shell residues for improved catalytic activity based
on variant ILS1, and found mutation V153A (ILS2), which in-
creased the catalytic activity 1.4-fold. Further rounds of direct-
ed evolution identified mutations T9S (ILS3, 1.2-fold improve-
ment), G39S (ILS4, 1.3-fold improvement), and T109N (ILS5, 1.3-
fold higher activity compared to ILS4). The final mutation
M245I (ILS6) improved activity another 1.3-fold. In total, a 30-
fold activity improvement, determined with crude cell lysate,
was obtained compared to Var8. Figure 5C shows the accumu-
starting variant, Var8.
Next, all IDF variants were overexpressed and purified, and
the kcat and Km values were determined for each variant with
the substrates d-fructose and d-psicose with an HPLC-based
detection assay. For the starting template and the final variant
(IDF8), these values were also determined with the substrates
d-tagatose, d-sorbose, l-sorbose, and l-fructose (Figures 3A
and 4A, Table S4). We observed a steady increase in k from
cat
ꢀ
1
ꢀ1
Var8 (4.9 s ) to IDF8 (42.3 s ) for d-fructose, and an increase
in kcat for d-psicose from Var8 (2.8 s ) to IDF8 (57.1 s ). How-
ever, the net catalytic efficiency (k /K ) remained nearly un-
ꢀ
1
ꢀ1
cat
m
changed, as the increase in kcat was accompanied by an in-
crease in K for both d-fructose and d-psicose.
m
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
3
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
unchanged, although the catalytic efficiency for d-psicose and
l-fructose was moderately increased (Figure 4B and D).
Structural basis of the change in substrate specificity and
catalytic activity
To investigate the structural changes in PcDTE after mutagene-
sis, we determined the high resolution crystal structure of the
starting variant Var8 in the substrate-free form, and of IDF8
and ILS6 in complex with their respective substrates. The struc-
tures were solved at resolutions of 1.8 ꢁ (Var8), 2.1 ꢁ (ILS6),
and 1.9 ꢁ (IDF8; Table 1), thus giving precise insight into the
detailed geometry of the active site and the interactions be-
tween PcDTE variants and their substrates.
Structural comparison between the wild-type (WT, PDB ID:
2QUN) and the three variants (Var8, IDF8, and ILS6) showed
very little difference in tertiary structure (Figure S6). Pairwise
alignment of the dimeric structures yielded root-mean square
deviation (rmsd) values from 0.43 (IDF8/ILS6) to 0.69 ꢁ (Var8/
IDF8).
Structural features of Var8
m
Figure 4. kcat and catalytic efficiencies (kcat/K ) for the first and the final var-
iants discovered during the screening, relative to the values of parent Var8
for the same substrate. The substrate for which the variants were evolved is
The structure of Var8 at 1.8 ꢁ resolution shows that the active-
site geometry of the WT enzyme is virtually identical to that for
marked with a hash (#). A) Relative kcat for IDF1 (
ent substrates. B) Relative kcat for ILS1 ( ) and ILS6 (
efficiencies for IDF1 ( ) and IDF8 ( ). D) Relative catalytic efficiencies for ILS1
) and ILS6 ( ). Detailed kinetic parameters are listed in Table S4.
&
) and IDF8 (
&
) for six differ-
&
&
). C) Relative catalytic
&
&
(
&
&
Table 1. Data collection and structure refinement statistics.
Crystal structure
PcDTE Var8
PDB ID: 4Q7I)
PcDTE IDF8
(PDB ID: 4PFH) (PDB ID: 4PGL)
PcDTE ILS6
lated mutations around the active site and their localization in
the molecule.
(
Data collection
space group
a, b, c [ꢁ]
Enzyme kinetic parameters were determined from purified
protein of each variant in the evolutionary trajectory with sub-
strate l-sorbose, and for ILS1 and ILS6 with l-fructose, d-fruc-
tose, d-psicose, d-tagatose, and d-sorbose. The most dramatic
improvement in kinetic parameters was found for mutation
C121
1
P12 1
1
P12 1
110.5, 47.5, 124.7 57.5, 86.7, 61.8 102.8, 47.4, 126.4
90, 103.7, 90
53–1.80
a, b, g [8]
resolution [ꢁ]
90, 90.1, 90
56–1.90
(1.94–1.90)
90, 102.5, 90
36–2.10
(2.15–2.10)
[
a]
(
1.85–1.80)
unique reflections
multiplicity
Rmerge [%]
I/s(I)
completeness [%]
Refinement
58426 (5789)
3.1 (2.9)
5 (33.6)
11.3 (3.4)
99.5 (99.6)
46594 (2988) 69460 (4447)
Q183N (ILS1): a 6.9-fold higher k value (from 0.24 (Var8) to
5.4 (5.5)
11 (53.5)
9.6 (4.6)
99.2 (71.0)
6.2 (5.8)
cat
ꢀ
1
9.9 (42.5)
12.1 (3.7)
99.1 (98.4)
1
.61 s (ILS1); Figure 3B). Mutation G39S (ILS4) resulted in no
improvement in k , thus indicating that the improvement of
cat
this mutation was attributable to increased expression of solu-
ble protein. Finally, ILS6 had a 13.5-fold improvement in kcat
compared to the starting variant Var8, and hence a roughly
twofold higher level of soluble protein expression (see above).
Interestingly, and in contrast to the evolution of IDF8, catalytic
improvement for ILS6 was not accompanied by a loss in ther-
Rwork/Rfree [%]
17.6/20.8
14.4/17.7
15.2/18.1
RMSD
bond lengths [ꢁ]
bond angles [8]
No. of atoms
protein
ligand
metals
0.013
1.5
0.013
1.5
0.011
1.4
4697
29
4
4814
48
3
9384
156
5
20
mostability compared to Var8 (DT =+0.28C).
50
The mutation leading to ILS1 effectively reverses the prefer-
ence of the enzyme from the S to the R configuration for the
C4-hydroxy group of the hexose substrate, and that of the C5-
hydroxy group from R to S. This single mutation changed the
catalytic efficiency between substrates d- (3R,4S,5R) and l-sor-
bose (3S,4R,5S) by 56-fold (Figure 4D). As a result, in the final
variant (ILS6), net catalytic rates as well as catalytic efficiency
for both the native substrates d-sorbose (3R,4S,5R) and d-taga-
tose (3S,4S,5R) were reduced, whereas kcat values for d-fructose
water
Average B-factor [ꢀ ]
protein (main chain) 6.2
protein (side chain) 8.2
metals
water
other ligands
Ramachandran statistics [%]
favored regions
allowed regions
460
420
381
2
6.6
9.7
15.7
26.9
26.6
13.3
17.5
22.3
25.4
38.9
23.1
35.9
36.0
98.3
1.6
97.8
2.2
97.8
2.2
[
a] Highest-resolution shell in parenthesis.
(3S,4R,5R), d-psicose (3R,4R,5R), and l-fructose (3R,4S,5S) were
&
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
4
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
oprotection. Two of its hydroxy
groups coordinate to the bound
manganese ion (Figure 5B), thus
mimicking the interactions previ-
ously observed for d-fructose or
[9d]
d-tagatose bound to PcDTE.
The Var8 structure confirmed
our speculations on the stabiliz-
ing effect of F157Y, T194N, and
A215Q, in that these mutants
provide additional H-bonds at
the
dimer
interface
(Fig-
a
[
3b]
ure 5A).
Unexpectedly,
metal ion is bound to H116
which is a serine in WT) and the
(
equivalent residue in its dimer
partner (Figure S8). Next, muta-
tion G260C appears to increase
the complementarity of the
dimer interface, thus strengthen-
ing the subunit–subunit interac-
tion by van der Waals con-
[19]
tacts. The contribution of mu-
tations K122V and K251T to Var8
stability are likely a result of the
reduction in dimer interface en-
tropy (by substituting larger side
chains with smaller ones). The
stabilizing effect of the M265L
mutation is difficult to rationalize
based on the structure.
Structural features of IDF8
In the IDF8 structure, one of the
His6 tags (residues H293–H298)
of a neighboring dimer pro-
trudes into the cleft next to the
active site of chain A (Fig-
ure S10), thereby forming hydro-
gen bonds to H116 and R257. As
these interactions are crystallo-
graphic artifacts, only chain B is
considered in the following anal-
ysis.
Figure 5. Structural details of PcDTE variants. A) Mutations to thermostable Var8 (PDB ID: 4Q7I) at the dimeric in-
terface are depicted in chain A, with inter-chain hydrogen-bond interactions established by mutations F157Y and
T194N shown in detail (right). The active-site tunnel of each subunit is indicated by an arrow, and the catalytic res-
idues E152 and E246 and the metal-coordinating residue D185 are depicted as red sticks; glycerol in the active
site is shown as green sticks; dark blue sticks indicate residues from chain B, light cyan sticks indicate residues
from chain A, and orange sticks indicate WT residues (right). B) Activity-enhancing mutations around the active
site of IDF8 (PDB ID: 4PFH) are shown as blue sticks on chain B; chain A is shown in surface representation; muta-
tions from the precursor variant Var8 are depicted as light cyan sticks; d-fructose (substrate) is shown as green
sticks. The active-site cavity of IDF8 with the three mutations S37N, G39E, and A258D at the entrance reveal a
tightly constricted active site entrance compared to Var8 (far right). C) Mutations that accumulated during screen-
ing for activity on l-sorbose are depicted as green sticks in ILS6 chain C (PDB ID: 4PGL); chain D is shown in sur-
face representation, and mutations from Var8 are depicted as light-cyan sticks; l-sorbose (substrate) is shown as
yellow sticks, and arrows indicate the entrance of the active site.
Crystals of variant IDF8 had
been soaked in d-fructose for
10 min prior to freezing in order
to determine the structure in the
substrate-bound
state.
The
active site of IDF8 exhibits addi-
tional electron density consistent
with the presence of both d-
the substrate-bound (PDB ID: 2QUN) and substrate-free forms
PDB ID: 2QUL; Figure S7). Although Var8 was crystallized with-
fructose and d-psicose epimers, which were modeled in the
electron densities with occupancies of 0.7 and 0.3, respectively,
according to the reported thermodynamic equilibrium of the
(
out a substrate, the active site shows the presence of a ligand;
this could be accounted for by glycerol, which we used for cry-
[13]
epimerization reaction (Figure 6B).
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
5
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Hydroxy groups OH1–OH3 of both epimers are engaged in
the same interactions. However, because of the distinct C3
chirality, C1–C3 do not superimpose. This probably explains
the discontinuity in the electron density at that part of the
sugar. The electron density is also rather weak at the substrate
C4–C6 moiety, which appears to be attributable to the lack of
interactions with surrounding residues, which are usually ob-
In IDF1, asparagine (S37N) not only constricts the substrate
channel but also forms a hydrogen bond at its amide group to
O6 of the bound d-fructose. Thus, this mutation might have
improved the positioning of the substrate in the active site
(Figure 6A), as in WT the O6 group interacts only indirectly via
an ordered water molecule with S37 (Figure 6B). Mutation
G39E in IDF3 further extends this hydrogen bond network by
interacting with N37 via an ordered water molecule and at its
backbone amide (Figure 5B).
[
9d,20]
served in other PcDTE/substrate structures.
The geometry of the IDF8 active site is highly similar to
those of WT and Var8, as illustrated by the nearly perfect super-
position of catalytic residues E152 and E246 and the metal-co-
ordinating residue D185 (Figure S7). The most prominent
changes in IDF8 arose from mutation S37N, G39E, and A258D.
The substitution of small residues with larger ones constricts
the active-site entrance tunnel (Figures 5B and S9). The con-
striction of the tunnel might shield the active site from bulk
solvent and in this way lead to an increase in catalytic efficien-
Mutation H209V (which yielded IDF2) breaks a hydrogen
bond network from the manganese ion to the surface of the
enzyme (Figure 6A, B). This leads to a slight repositioning of
H211, the manganese ion, and the ordered water that contacts
O3 and O5 of the substrate. The effects of the remaining muta-
tions (T109N, L212I, and S256G) are difficult to predict from
the crystal structure.
[
21]
cy, as proposed for Kemp eliminase and haloalkane dehalo-
Structural features of ILS6
[
22]
genase. Mutations G39E and A258D increase the charge at
the tunnel opening.
Pairwise alignments with WT and Var8 reveals
a
highly similar tertiary structure for ILS6 (Figure S6 and
Table S5). The active-site geometry, as judged from
superposition of catalytic residues E152 and E246 to-
gether with metal-binding residue D185, is also very
similar (Figure S7).
ILS6 crystals were soaked in l-sorbose for 3 min
before flash freezing, in order to obtain the structure
in its substrate-bound state. For IDF8, the electron
density indicates the presence of both C3 epimers.
The l-sorbose and the l-tagatose epimers were mod-
eled with occupancies of 0.7 and 0.3, respectively
(
Figure 6D).
Mutation Q183H (first introduced in ILS1) resulted
in the highest increase in kcat for the l-sorbose/l-ta-
gatose epimerization reaction. When comparing the
substrate-binding modes of l-sorbose to ILS1 with
that of d-fructose to WT, distinct hydration patterns
at the distal part of the saccharides are evident (Fig-
ure 6C). Whereas in WT a water molecule is linking
the sugar 5-OH hydroxy with the Q183 side chain,
this interaction is lost in the Q183H mutation in ILS1.
Whether this changed hydration pattern results in
improved positioning of l-sorbose (which is the C5
epimer of d-fructose) with respect to the catalytic
center in ILS6 versus WT is hard to determine. The
structural consequences of all other mutations
appear to be very minor, as are their effects on kcat
and Km (Figure 3B), with the exception of T109N,
which improved kcat by a factor of about 1.5 (ILS5).
This mutation affects a second-shell residue that is
completely buried. Replacement of the wild-type
threonine side chain with the longer asparagine side
Figure 6. Comparisons of the interactions between WT and improved variants. Interac-
tions are shown as black dashed lines, distances are in ꢁ, manganese ions are displayed
as violet spheres, and water molecules are red spheres. 2F
o
ꢀF
C
and F
o
ꢀF
c
omit maps are
contoured at 1s (blue mesh) and at 3s (pink mesh), respectively. A) Interaction network
in vicinity of WT active site (PDB ID: 2QUN) with d-fructose (substrate) represented as
green sticks. B) The altered interaction network in IDF8 (PDB ID: 4PFH) arises from muta-
tion H209V, which interrupts the interaction network between T242 and the substrate. d- chain is accommodated by a slight displacement of
fructose is shown as green sticks, and C3-epimer d-psicose is shown as blue sticks. C) In-
the main chain segment 106–108. This structural
teraction network in vicinity of WT active site (PDB ID: 2QUN) with d-fructose shown as
change is propagated towards segment 67–69, which
green sticks. D) Changes in interactions in the active site of ILS6 (PDB ID: 4PGL) com-
forms part of the substrate channel. This small but
significant change (0.3 ꢁ) is probably not responsible
pared to WT as a result of mutations G39S and Q183H. ILS6 substrates l-sorbose and l-
tagatose are depicted as yellow and pink sticks, respectively.
&
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
6
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
alone for the altered enzyme characteristics; it is likely that the
dynamic behavior of these segments is important for substrate
entry and product release, and might well be affected by the
mutation at this buried site.
residues close to the active site showed a significant difference
in pK between WT and IDF8: E35, E39, and H209 (Table S6).
a
Hence, two different explanations are conceivable for the co-
operative pH–rate profile of IDF8. First, mutation H209V dis-
rupts a hydrogen-bond network between T242, H209, H211,
and E35 (Figure 6A, B), and thus changes the predicted pK of
a
E35 significantly (pK <0 for WT, 5.6 for IDF8). Protonation of
a
pH–rate profiles for WT, Var8, IDF8, and ILS6
E35 in IDF8 would presumably severely disturb the noncova-
II
In order to determine whether a change in pK of catalytically
a
lent interactions with H211 and thus with the catalytic Mn ion,
relevant residues might have had an impact on the increased
catalytic activity during directed evolution of IDF8 or ILS6, we
recorded pH–rate profiles for the four PcDTE variants. The epi-
merization reaction between two ketohexoses at the C3-posi-
tion by DTE has been suggested to be catalyzed by two gluta-
mate residues (E154 and E246) that are in their deprotonated
hence leading to the break-up of these noncovalent interac-
tions in a domino-like effect and abolishing the catalytic activi-
ty. Second, protonation of E39 (pK 5.9 for IDF8) could lead to
a
a loss of the extensive hydrogen-bond network at the entrance
to the substrate channel in a cooperative manner, thus leading
to an impaired entrance of the substrate.
[
9d]
states and can act as catalytic bases.
Mechanistically, the
two glutamates have been suggested to abstract a proton
from C3, thereby forming a cis-enediolate intermediate, fol-
lowed by re-protonation at C3 by the other glutamate residue.
Other ionizable residues also contribute to the binding and
positioning of the substrate, thus suggesting that pH might
significantly affect the catalytic rate.
Conclusion
Screening of the thermostable Var8 for catalytic efficiency with
two different substrates revealed two interesting evolutionary
trajectories. In both cases, we were able to find only one bene-
ficial mutation adjacent to the active site (first-sphere residue):
S37N in IDF1, and Q183H in ILS6. IDF1 showed only a moderate
increase in kcat for d-fructose, whereas ILS1 showed a much
more pronounced increase in both catalytic activity and cata-
lytic efficiency for l-sorbose. We suppose that mutation Q183H
switches the substrate preference from hexoses with a 4S,5R
configuration to those with a 4R,5S configuration, based on
the catalytic efficiencies of the ILS1 mutant for six different
substrates (Figure 4). Building on this initial improvement, cata-
lytic activity was then further improved by five additional mu-
tations, and led finally to ILS6 (>13-fold improvement in kcat
for l-sorbose, compared to Var8). These results underline the
importance of second-shell residues for catalytic activity. Toku-
riki et al. showed similarly that mutations randomly introduced
by error-prone PCR (epPCR) into a phosphotriesterase to
change it to an arylesterase accumulated stepwise in a radial
fashion (from the center out), with only four of the 18 muta-
By fitting the standard Henderson–Hasselbach equation
[Eq. (1), see the Experimental Section for details] to the normal-
ized activities, an apparent pK (pKa,app) of 6.1ꢁ0.05 was deter-
a
mined for WT, whereas the apparent pKa,app of Var8 was signifi-
cantly lower (4.5ꢁ0.02), and ILS6 showed a pK
of 5.25ꢁ
a,app
0
.06 (Figure 7). Surprisingly, the pH–rate curve of IDF8 could
not be described adequately by a simple Henderson–Hassel-
bach equation, thus suggesting cooperativity between several
different ionizable residues within the active site of IDF8. There-
fore a modified Hill equation that accommodates cooperativity
[
23]
was employed for fitting [Eq. (2)]. To assess which residues
could elicit this effect we determined the pK values of all ioniz-
a
able groups of WT and IDF8 by means of the web-based predic-
[24]
tion software H+ +. Predictions were made with explicit in-
corporation of the manganese ion in the active site, as it signifi-
cantly influences the pK values of the adjacent residues. Three
a
[25]
tions directly adjacent to the substrate-binding site. We ob-
served a similar behavior in the evolution of ILS variants, inso-
far as an initial big improvement in catalytic rate was followed
by less pronounced improvements from more distant muta-
tions in the second sphere.
In contrast to the evolution of the ILS variants of PcDTE,
where a strong initial improvement was followed by minor in-
cremental improvements, the IDF variants showed a smooth
stepwise increase in kcat over the whole trajectory (Figure 3).
Furthermore, the increase in kcat for the IDF variants was ac-
companied by a reduction in affinity (K ) for d-fructose
m
through the evolutionary trajectory, whereas the ILS branch
showed no significant change in affinity to l-sorbose despite
the increase in catalytic activity. It is difficult to explain this
finding, as the substrate concentration used for screening was
the same for both branches (200 mm). Because of the increase
in Km for IDF variants, no net increase in catalytic efficiency
Figure 7. pH–rate profiles of WT (
), thermostabilized Var8 (^), and catalyti-
&
cally improved variants IDF8 (~) and ILS6 (*). Activities were determined
with substrates d-fructose (WT, Var8, IDF8) or l-sorbose (ILS6) in acetate
buffer (pH 5–6), phosphate buffer (pH 6–7), or Tris buffer (pH 7– 9). Data
were normalized to the maximum activity of the respective variant (meanꢁ
SD; n=4 (Var8, ILS6, IDF8), or n=2 (WT)). Solid lines are best fit of the ex-
perimental values to a pH–rate equation (see the Experimental Section).
(
k /K ) was observed during the directed evolution that re-
cat m
sulted in IDF8. This does not, however, prevent the use of IDF8
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
7
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
as a promising catalyst for the production of d-psicose, as the
substrate concentration in industrial settings (in particular, in
sugar-processing plants) is usually high (>1m). Therefore the
Molecular biology: General molecular biology procedures were
[28]
performed according to standard protocols.
All PCR products
were generated using Phusion High-Fidelity Polymerase (NEB). Pri-
mers are listed in Table S1; all plasmids are listed in Table S2. All
general cloning work was done in E. coli Top10 cells (Invitrogen).
impact of K is often considered to be of secondary interest in
m
[
26]
industrial biocatalysis. We hypothesize that the stepwise im-
provement in catalytic activity and concomitant decrease in
affinity (as observed for IDF mutants) could be explained by
destabilization of the enzyme–substrate complex while stabili-
zation of the transition state is unchanged.
Cloning of KpRD from K. pneumoniae and RsGD from R. sphaer-
oides: A KpRD had been identified in Klepsiella pneumoniae previ-
[17]
ously, and the annotated genome of that strain in the NCBI data-
base indeed suggested a suitable ribitol dehydrogenase (GenBank
Nr. ESM54230). The gene was codon-optimized for expression in
E. coli and synthesized by Geneart (Regensburg, Germany). Plasmid
pAB139 was constructed as follows (see Table S2 for details): the
pBR322 origin of replication was amplified from pRK793 by using
primer pBR322_for-AscI and pBR322_rev-FseI, and used to ex-
change the origin of replication of pSEVA131 at the unique restric-
tion sites AscI and FseI to result in plasmid pAB73. The tetR-Ptet-
PT7 cassette was amplified from plasmid pKTS by using primers
Ptet-SEVA_for and Ptet-SEVA_rev, and inserted into pAB73 at re-
striction sites SpeI and HindIII to result in pAB92. The bla resistance
The loss in thermostability for IDF variants is in line with pre-
vious directed-evolution efforts that similarly improved catalyt-
[
10]
ic activity at the expense of thermostability. The unchanged
thermostability of the ILS mutants however points out that
destabilization is not necessarily a prerequisite for improved
catalytic activity.
As expected, the overall structural changes in IDF8 and ILS6
relative to WT are very small. Upon detailed inspection of the
structures, the most prominent change for IDF8 was the accu-
mulation of negatively charged residues at the entrance of the
substrate channel (Figure 5B). It has been shown previously
that modifying the residues that line the substrate channel or
its mouth has a high probability of yielding catalytic improve-
R
gene of pAB92 was exchanged to the kam resistance gene from
plasmid pSEVA231 at the unique restriction sites SwaI and FseI to
give vector pAB228. The gene for KpRD was amplified from the
vector pMA-T-KpRD by using primers KpRD_N6H_f (introducing
a His tag) and KpRD_s_rev and inserted into pAB92 at restriction
6
[
21,22]
ment.
The two mutations that encode negatively charged
sites HindIII and EcoRI to result in construct pAB139.
residues (G39E and A258D) led to a conformational change of
the two positively charged residues K70 and R257 at the en-
trance of the substrate tunnel. Together, this highly charged
patch at the mouth of the entrance might facilitate the entry
of the polar sugar substrate, although the entry tunnel of IDF8
is also believed to be significantly constricted by these muta-
tions compared to the WT enzyme.
[29]
The galactitol dehydrogenase (RsGD) gene was isolated from the
genomic DNA of R. sphaeroides (DSM No. 8371). The gene was am-
plified directly from cells by using primers RsGDH_NheI_for and
RsGDH_s_rev and inserted at restriction sites NheI and EcoRI of
pAB139 to give construct pAB140.
Expression and purification of KpRD and RsGD: E. coli BL21(DE3)
cells were transformed with plasmid pAB139 or pAB140 and pre-
cultured in lysogeny broth (5 mL) supplemented with kanamycin
(50 mgmL ). This preculture (200 mL) was used to inoculate of M9
medium (50 mL) containing d-glucose (0.4%) and kanamycin
(50 mgmL ). After incubation this was used to inoculate a 1 L fed-
batch reactor. Protein production was induced with isopropyl-b-d-
thiogalactopyranoside (0.2 mm) at OD600 =0.5, and protein was
synthesized for 6 h at 378C. Cells were harvested by centrifugation
We conclude that our strategy of improving catalytic activity
of enzymes by screening residues around the active site (pro-
gressively from the center to the periphery) is an efficient
method for directed evolution of enzymes where screening is
the limiting step, as is the case for most industrially relevant
ꢀ1
ꢀ1
[
1a]
biocatalysts. As perhaps expected, the phenotypes of the
variants cannot easily be correlated with the respective struc-
tures of the substrate complexes. A full understanding of the
effects would require extensive molecular dynamics/quantum
mechanics calculations, and would be very sensitive to correct
estimation of the on- and off-rates through the tight access
channel, as well as the structure of the transition state. The
structures of the optimized variants, however, can now be
utilized for the identification of new randomization sites for
further enzyme improvement.
(
6000g, 20 min) and stored at ꢀ808C. To obtain purified RsGD, the
wet cell pellet (10 g) from the RsGD cultivation was resuspended
in RsGD lysis buffer (15 mL, Tris (50 mm, pH 6.8), NaCl (100 mm),
ꢀ
1
imidazole (20 mm), lysozyme (0.2 mgmL )) and incubated for
20 min at room temperature before the cell suspension was frozen
at ꢀ808C for 20 min. The suspension was then thawed at room
temperature, MnCl (1 mm) and a spatula-tip of DNase (ca. 1000
2
Kunitz units) were added, and the suspension was sonicated for
0 min in an ultrasonication water bath. Cell debris was removed
by centrifugation (48384g, 20 min, 48C), and the cleared lysate
was loaded onto Ni-Sepharose 6 Fast Flow (2 mL; GE Healthcare) in
a gravity-flow column. The column was extensively washed with
lysis buffer before protein was eluted with elution buffer (Tris
1
Experimental Section
(
50 mm, pH 6.8), imidazole (200 mm), NaCl (100 mm)). The main
Except where stated otherwise, all chemicals were purchased from
Sigma–Aldrich. NADH was purchased from GERBU Biotechnik (Hei-
delberg, Germany); d-fructose, d-sorbose, and l-fructose were pur-
chased from Carbosynth (Compton, UK). d-psicose was produced
in-house by epimerization of d-fructose to d-psicose with DTE and
separation of d-psicose from d-fructose by continuous chromatog-
raphy. Restriction enzymes and polymerases were obtained from
New England Biolabs (NEB), and oligonucleotides were from Micro-
synth (Balgach, Switzerland).
fractions containing RsGD were analyzed by SDS-PAGE, and pure
fractions (>95%) were pooled and dialyzed twice against buffer A
(1 L; Tris (20 mm, pH 6.8), MnCl (1 mm), sucrose (10%)) before it
2
was dialyzed once against buffer B (200 mL; Tris (20 mm, pH 6.8),
MnCl (1 mm), sucrose (30%)). The protein solution was sterile fil-
2
tered through a 0.2 mm filter and stored at 48C. KpRD was purified
in the same way, except for the following modifications: the wet
pellet was resuspended in KpRD lysis buffer (15 mL, Tris (50 mm,
[27]
pH 8.0),
NaCl
(100 mm),
imidazole
(20 mm),
lysozyme
&
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
8
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
ꢀ
1
(
0.2 mgmL )), and IMAC-purified KpRD was dialyzed twice against
pression system with the Ptet-PT7 fusion promoter (Table S2). Plas-
mids encoding the variants were isolated from E. coli Top10 cells
and transformed into chemically competent E. coli BL21 (DE3) cells.
All variants were expressed in ZYM-5052 autoinduction medium.
Overnight culture (1 mL) of the respective variant was used to
buffer C (2 L, Tris (20 mm, pH 8.0)) before being aliquoted and
stored at ꢀ808C. Protein concentrations were determined spectro-
[31]
photometrically at 280 nm for both RsGD (M =27.5 kDa, e=21.1ꢂ
W
3
ꢀ1
ꢀ1
3
ꢀ1
ꢀ1
1
0 m cm ) and KpRD (M =27.4 kDa, e=35.1ꢂ10 m cm ).
W
inoculate ZYM-5052 (250 mL) supplemented with ampicillin
d-Psicose quantification assay using KpRD: For qualitative deter-
mination of conversion of d-fructose into d-psicose, a screening
assay was developed based on the reduction of d-psicose by KpRD
with coenzyme NADH. The screening assay was performed in
a two-step fashion. First, epimerization of d-fructose to d-psicose
by a PcDTE enzyme variant was done for a fixed time (see below).
Second, NADH and KpRD were added to convert d-psicose to alli-
tol. The concomitant oxidation of NADH was followed spectropho-
tometrically (340 nm). PcDTE was not inactivated before the KpRD
was added (for reasons of simplicity), therefore this assay allowed
only a qualitative comparison of PcDTE activity, but this was suffi-
cient for screening purposes. As d-fructose is also to a small extent
a substrate for KpRD, a calibration curve for the oxidation of differ-
ent starting concentrations of d-psicose in presence of d-fructose
was recorded by following the rate of NADH consumption at
ꢀ1
(100 mgmL ) in a 1 L Erlenmeyer flask and incubated for 16 h at
3
08C with shaking (220 rpm). Cells were harvested by centrifuga-
tion (6238g, 20 min, 48C) and stored as a pellet at ꢀ208C. Cells
were resuspended in lysis buffer (10 mL; Tris (50 mm, pH 8.0), lyso-
ꢀ1
zyme (0.2 mgmL )) and incubated for 30 min at room tempera-
ture. The cells were then lysed by one freeze/thaw cycle (20 min at
ꢀ
808C, thaw at room temperature) before MnCl2 was added
1 mm), and then DNase (5 mL of a 5 mgmL solution) was added
ꢀ1
(
to reduce viscosity. The cell lysate was heat-treated for 10 min at
08C in a water bath before cell debris and denatured host pro-
7
teins were removed by centrifugation (48384g, 20 min). Cleared
lysate was applied to Ni-Sepharose 6 Fast Flow (2 mL; GE Health-
care) in a gravity-flow column. The column was extensively washed
with buffer (Tris (50 mm, pH 8.0), NaCl (100 mm), imidazole
(
(
30 mm)), then the protein was eluted with elution buffer (Tris
50 mm, pH 8.0), NaCl (100 mm), imidazole (200 mm)). The main
3
40 nm in a Wallac 1420 Victor plate reader (PerkinElmer). Six d-
fructose calibration samples were prepared in Tris buffer (50 mm,
pH 8.0): 100, 99, 98, 95, 90, and 80 mm (the difference to a total
hexose concentration of 100 mm was made up with d-psicose).
Next, aliquots (200 mL) from each calibration sample was supple-
mented with NADH (1 mm) and KpRD (25 mg) and placed in a 96-
well microplate (Greiner Bio-One), and the rates of NADH con-
sumption at 308C were recorded. This rate was a linear function of
the d-psicose concentration (at least in the range 0–10%; Fig-
ure S4A), and this part was then used as the calibration curve to
determine the amount of d-psicose in the screening assay.
fractions containing the PcDTE variants were pooled and dialyzed
against buffer D (Tris (10 mm, pH 8.0), MnCl2 (1 mm)) and then
twice against Tris (10 mm, pH 8.0). Dialyzed proteins were aliquot-
ed and stored at ꢀ808C. All variants showed >95% protein purity
as judged by SDS-PAGE. Enzyme concentration was determined
spectrophotometrically (see Table S8 for extinction coefficients and
molecular weights of all variants).
Enzyme kinetic measurements: Enzyme kinetic constants kcat and
K were determined from progress curves with six ketohexose sub-
m
strates (d-fructose, d-psicose, d-tagatose, d-sorbose, l-sorbose, and
l-fructose) epimerizing to the respective stereoisomer. In detail,
sodium phosphate buffer (200 mL; 50 mm, pH 7.0) and substrate
l-Tagatose quantification assay using RsGD: Detection of epime-
rization of l-tagatose from l-sorbose was done as for d-psicose
(
above). Epimerization of l-sorbose to l-tagatose was detected by
(
4.4 mm, 22 mm, 110 mm, 550 mm, 1.1m, or 2.2m) were mixed with
purified enzyme (20 mL; 76–184 mg) in a U-shaped 96-well plate
NUNC) on a BioShake iQ shaker (Q. Instruments, Jena, Germany) at
58C and 1200 rpm. Reaction was stopped by adding the reaction
using RsGD, which preferentially reduces l-tagatose to galactitol
with concomitant oxidation of NADH. A calibration curve for l-ta-
gatose in the presence of l-sorbose was generated in a similar
fashion. In short, an l-tagatose calibration sample (200 mL) was pre-
pared in Tris buffer (50 mm, pH 8.0) supplemented with NADH
(
2
mixture (20 mL) to HCl (145 mL; 0.1m), followed by the addition of
NaOH (135 mL; 0.1m) after 5 min. Conversion of the substrate to
the respective epimer was determined by HPLC (see above). Kinet-
ic parameters Km and kcat were obtained by fitting initial velocities
to the Michaelis–Menten kinetic model in SigmaPlot 12.2 (Systat
Software, San Jose, CA).
(
1 mm), MgCl (1 mm), and of RsGD (4.8 mg) in a 96 well micro-
2
plate, and NADH oxidation rates were recorded at 308C. The linear
part of the slope was used as calibration curve (see Figure S4B).
Cloning of Var8 and library generation: The thermostabilized d-
[
3b]
tagatose epimerase gene from P. cichorii (Var8)
was amplified
pH-dependent activity profile: Initial catalytic rates over pH 5.0–
[3b]
from plasmid pKTS-PcDTE-Var8-C6H by using primers DTEci_Hin-
dIII_f and DTEci-ss_EcoRI_r and inserted into pAB92 to give plas-
mid pAB174, which served as the template for mutant library gen-
eration. Saturation mutagenesis libraries on pAB174 were generat-
9
.0 were determined with d-fructose (90 mm) as substrate for WT,
Var8, and IDF8, and l-sorbose (90 mm) for ILS6, in acetate buffer
80 mm, pH 5.0–6.0), phosphate buffer (80 mm, pH 6.0–7.0), or Tris
buffer (80 mm, pH 7.0–9.0). Enzyme stocks in phosphate buffer
10 mm, pH 7.0) were supplemented with MnCl (1 mm). The reac-
(
[3b]
ed as described previously. Primers (Table S7) with NNK-degener-
(
2
[30]
ated codons were used to randomize selected sites by the Quik-
Change protocol (Agilent Technologies) with Phusion High-Fidelity
Polymerase (NEB). The product was digested directly in the poly-
merase buffer with DpnI (1 mL, 10 U) for at least 2 h at 378C in
order to remove the template, then mutagens (5 mL) were used to
transform of chemically competent E. coli Top10 cells (70 mL).
tions were performed at 258C in a 96-well plate as described for
the enzyme kinetic measurements (above). Reactions were stopped
at four time-points by adding reaction mixture (20 mL) to HCl
(145 mL; 0.1m). NaOH (135 mL; 0.1m) was added after 5 min, and
conversion was determined by HPLC as described above. pH–rate
data were fitted for variants WT, Var8, and ILS6 by using Equa-
[32]
tion (1).
Expression and screening of saturation-mutagenesis libraries of
Var8: Clones from saturation mutagenesis libraries were expressed
[3b]
as described previously. For details see the Supporting Informa-
tion.
Vmax ꢂ 10ꢀ
pKa,app
VH ¼ 1
ð1Þ
ꢀpH
ꢀpKa,app
0
þ 10
Expression and purification of improved IDF and ILS variants:
Variants IDF1–IDF8 and ILS1–ILS6 were expressed by using a T7 ex-
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
9
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
where V is the pH-dependent reaction rate, Vmax is the maximum
[5] C. K. Savile, J. M. Janey, E. C. Mundorff, J. C. Moore, S. Tam, W. R. Jarvis,
J. C. Colbeck, A. Krebber, F. J. Fleitz, J. Brands, P. N. Devine, G. W. Huis-
man, G. J. Hughes, Science 2010, 329, 305–309.
H
reaction rate, and pKa,app is the apparent pK value of the acidic
a
23]
[
groups. For IDF8 a modified Hill Equation (2) was used:
[
6] S. Ito, Appl. Microbiol. Biotechnol. 2009, 84, 1053–1060.
Vmax ꢂ 10ꢀpKa,app
[7] K. Beerens, T. Desmet, W. Soetaert, J. Ind. Microbiol. Biotechnol. 2012, 39,
VH ¼ 1
ð2Þ
823–834.
ꢀpHꢂn
ꢀpKa,appꢂn
0
þ 10
[
8] T. B. Granstrçm, G. Takata, M. Tokuda, K. Izumori, J. Biosci. Bioeng. 2004,
7, 89–94.
9
where n is the Hill coefficient.
[9] a) M. Jia, W. Mu, F. Chu, X. Zhang, B. Jiang, L. L. Zhou, T. Zhang, Appl. Mi-
crobiol. Biotechnol. 2014, 98, 717–725; b) W. L. Zhang, D. Fang, Q. C.
Xing, L. Zhou, B. Jiang, W. M. Mu, PLoS One 2013, 8, e62987; c) L.
Zhang, W. Mu, B. Jiang, T. Zhang, Biotechnol. Lett. 2009, 31, 857–862;
d) H. Yoshida, M. Yamada, T. Nishitani, G. Takada, K. Izumori, S. Kamitori,
J. Mol. Biol. 2007, 374, 443–453.
Crystallization of Var8, IDF8, and ILS6: Var8, IDF8, and ILS6 were
expressed and purified as described above. Crystals were obtained
by the sitting-drop vapor diffusion method (IDF8 and ILS6) in Intel-
li-Plates 96 (Hampton Research, Aliso Viejo, CA) or MRC2 crystalliza-
tion plates (Molecular Dimensions, Suffolk, UK), or by the hanging-
drop vapor diffusion method (Var8) on siliconized cover slides on
ComboPlates (Jena Bioscience, Jena, Germany). All crystallization
trials were performed at 208C. Initial crystallization hits were ob-
tained using the commercial screens JBScreen Classic 5 (Jena Bio-
science) for Var8, NeXtal PEGs-Suite (Qiagen) for IDF8, and PEGRx
HT (Hampton Research) for ILS6. Crystallization conditions were
optimized by varying the precipitant concentration and the pH
around the initial hit conditions. The optimized crystallization con-
ditions were as follows: Tris·HCl (0.1m, pH 7.7), PEG 8K (18%, w/v),
[
10] a) W. Besenmatter, P. Kast, D. Hilvert, Proteins Struct. Funct. Bioinf. 2007,
6
6, 500–506; b) J. D. Bloom, S. T. Labthavikul, C. R. Otey, F. H. Arnold,
Proc. Natl. Acad. Sci. USA 2006, 103, 5869–5874; c) J. D. Bloom, F. H.
Arnold, Proc. Natl. Acad. Sci. USA 2009, 106, 9995–10000; d) N. Tokuriki,
D. S. Tawfik, Curr. Opin. Struct. Biol. 2009, 19, 596–604.
[
[
11] K. M. Polizzi, A. S. Bommarius, J. M. Broering, J. F. Chaparro-Riggers, Curr.
Opin. Chem. Biol. 2007, 11, 220–225.
12] N. Tokuriki, D. S. Tawfik, Science 2009, 324, 203–207.
[13] W. Mu, W. Zhang, Y. Feng, B. Jiang, L. Zhou, Appl. Microbiol. Biotechnol.
2012, 94, 1461–1467.
14] a) Y. Yoshikuni, T. E. Ferrin, J. D. Keasling, Nature 2006, 440, 1078–1082;
b) A. Aharoni, L. Gaidukov, O. Khersonsky, S. McQ Gould, C. Roodveldt,
D. S. Tawfik, Nat. Genet. 2005, 37, 73–76.
[
LiSO (0.2m) for Var8; Na-HEPES (0.1m, pH 7.5), PEG 6K (25%, w/v)
4
for IDF8; MES (0.1m, pH 5.7), PEG 4K (10%, w/v) for ILS6. Crystals
for Var8 were cryoprotected by dipping into mother liquor supple-
mented with glycerol (20%, v/v) and flash-frozen in liquid nitrogen.
Crystals for IDF8 were soaked for 10 min in mother liquor supple-
mented with d-fructose (0.8m), and crystals for ILS6 were soaked
for 3 min in mother liquor supplemented with l-sorbose (0.6m).
Soaked crystals were then flash-frozen without addition of further
cryoprotectant.
[
[
15] M. T. Reetz, J. D. Carballeira, Nat. Protoc. 2007, 2, 891–903.
16] A. Huwig, S. Emmel, G. Jꢃkel, F. Giffhorn, Carbohydrate Res. 1997, 305,
3
37–339.
17] K. Takeshita, Y. Ishida, G. Takada, K. Izumori, J. Biosci. Bioeng. 2000, 90,
45–548.
[
5
[18] Y. Ishida, T. Kamiya, H. Itoh, Y. Kimura, K. Izumori, J. Ferment. Bioeng.
1997, 83, 529–534.
19] S. J. Hubbard, P. Argos, Protein Sci. 1994, 3, 2194–2206.
20] K. Kim, H.-J. Kim, D.-K. Oh, S.-S. Cha, S. Rhee, J. Mol. Biol. 2006, 361,
920–931.
[
[
Co-ordinates and structure factors were deposited in the Protein
Data Bank under the accession codes 4Q7I (Var8), 4PFH (IDF8), and
[21] R. Blomberg, H. Kries, D. M. Pinkas, P. R. E. Mittl, M. G. Grꢄtter, H. K. Pri-
4
PGL (ILS6).
vett, S. L. Mayo, D. Hilvert, Nature 2013, 503, 418–.
[
22] a) M. Pavlova, M. Klvana, Z. Prokop, R. Chaloupkova, P. Banas, M. Otyep-
ka, R. C. Wade, M. Tsuda, Y. Nagata, J. Damborsky, Nat. Chem. Biol. 2009,
5, 727–733; b) Z. Prokop, A. Gora, J. Brezovsk y´ , R. Chaloupkovꢅ, V.
Structural analysis: The substrate entry tunnel was analyzed by
the PyMOL plugin Caver 3.0; potential hydrogen bonds and salt
bridges were predicted by PyMOL or the PDBePISA webserver
[33]
ˇ ˇ
´
Stepꢅnkovꢅ, J. Damborsky in Protein Engineering Handbook, Vol. 3 (Eds.:
S. Lutz, U. T. Bornscheuer), Wiley-VCH, Weinheim, 2012, pp. 421–464.
23] J. L. Markley, Acc. Chem. Res. 1975, 8, 70–80.
(
http://www.ebi.ac.uk/pdbe/pisa). All figures were prepared with
[
[
PyMOL.
24] R. Anandakrishnan, B. Aguilar, A. V. Onufriev, Nucleic Acids Res. 2012, 40,
W537–W541.
[
25] N. Tokuriki, C. J. Jackson, L. Afriat-Jurnou, K. T. Wyganowski, R. Tang,
D. S. Tawfik, Nat. Commun. 2012, 3, 1257.
Acknowledgements
[
[
26] R. J. Fox, M. D. Clay, Trends Biotechnol. 2009, 27, 137–140.
27] N. Wagner, M. Fuereder, A. Bosshart, S. Panke, M. Bechtold, Org. Process
Res. Dev. 2012, 16, 323–330.
The authors want to thank the beamline staff from the Swiss
Light Source (SLS) for their support during data collection. A.B. is
indebted to the Swiss National Science Foundation for funding
[
[
[
28] J. Sambrook, D. W. Russell, Molecular Cloning: A Laboratory Manual, 3rd
ed., Cold Spring Harbor Laboratory Press, 2001.
(grant 200021–121918).
29] P. Kornberger, J. Gajdzik, H. Natter, G. Wenz, F. Giffhorn, G. W. Kohring,
R. Hempelmann, Langmuir 2009, 25, 12380–12386.
30] M. T. Reetz, D. Kahakeaw, R. Lohmer, ChemBioChem 2008, 9, 1797–
Keywords: enzymes
·
epimerases
·
monosaccharides
·
1804.
mutagenesis · X-ray crystallography
[
[
31] F. W. Studier, Protein Expression Purif. 2005, 41, 207–234.
32] O. Khersonsky, D. Rçthlisberger, O. Dym, S. Albeck, C. J. Jackson, D.
Baker, D. S. Tawfik, J. Mol. Biol. 2010, 396, 1025–1042.
[
1] a) U. T. Bornscheuer, G. W. Huisman, R. J. Kazlauskas, S. Lutz, J. C. Moore,
K. Robins, Nature 2012, 485, 185–194; b) S. Luetz, L. Giver, J. Lalonde,
Biotechnol. Bioeng. 2008, 101, 647–653.
[33] E. Chovancova, A. Pavelka, P. Benes, O. Strnad, J. Brezovsky, B. Kozlikova,
A. Gora, V. Sustr, M. Klvana, P. Medek, L. Biedermannova, J. Sochor, J.
Damborsky, PLoS Comput. Biol. 2012, 8, e1002708.
[
[
2] C. A. Tracewell, F. H. Arnold, Curr. Opin. Chem. Biol. 2009, 13, 3–9.
3] a) Y. Gumulya, M. T. Reetz, ChemBioChem 2011, 12, 2502–2510; b) A.
Bosshart, S. Panke, M. Bechtold, Angew. Chem. Int. Ed. 2013, 52, 9673–
9676; Angew. Chem. 2013, 125, 9855–9858.
[4] M. T. Reetz, S. Prasad, J. D. Carballeira, Y. Gumulya, M. Bocola, J. Am.
Received: October 25, 2014
Chem. Soc. 2010, 132, 9144–9152.
Published online on && &&, 0000
&
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
10
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

FULL PAPERS
Divergent evolution: Saturation muta-
genesis of residues around the active
site of a thermostable d-tagatose epi-
merase, Var8, and screening for epimeri-
zation of two different hexose sub-
strates resulted in two variants with
much improved specific activities. Crys-
tal structure analysis of Var8 and the
two final variants revealed different mu-
tation trajectories for the increase in ac-
tivity.
A. Bosshart, C. S. Hee, M. Bechtold,
T. Schirmer, S. Panke*
&& – &&
Directed Divergent Evolution of
a Thermostable d-Tagatose Epimerase
towards Improved Activity for Two
Hexose Substrates
ChemBioChem 0000, 00, 0 – 0
www.chembiochem.org
11
ꢀ 0000 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ