Redesign of the Active Site of Sucrose Phosphorylase through a Clash-Induced Cascade of Loop Shifts

Article abstract of DOI:10.1002/cbic.201500514

Sucrose phosphorylases have been applied in the enzymatic production of glycosylated compounds for decades. However, several desirable acceptors, such as flavonoids or stilbenoids, that exhibit diverse antimicrobial, anticarcinogenic or antioxidant properties, remain poor substrates. The Q345F exchange in sucrose phosphorylase from Bifidobacterium adolescentis allows efficient glucosylation of resveratrol, (+)-catechin and (-)-epicatechin in yields of up to 97 % whereas the wild-type enzyme favours sucrose hydrolysis. Three previously undescribed products are made available. The crystal structure of the variant reveals a widened access channel with a hydrophobic aromatic surface that is likely to contribute to the improved activity towards aromatic acceptors. The generation of this channel can be explained in terms of a cascade of structural changes arising from the Q345F exchange. The observed mechanisms are likely to be relevant for the design of other tailor-made enzymes.

Full text of DOI:10.1002/cbic.201500514

DOI: 10.1002/cbic.201500514
Communications
Redesign of the Active Site of Sucrose Phosphorylase
through a Clash-Induced Cascade of Loop Shifts
Michael Kraus,[a] Clemens Grimm,*[b] and Jürgen Seibel*[a]
Sucrose phosphorylases have been applied in the enzymatic
production of glycosylated compounds for decades. However,
several desirable acceptors, such as flavonoids or stilbenoids,
that exhibit diverse antimicrobial, anticarcinogenic or antioxi-
dant properties, remain poor substrates. The Q345F exchange
in sucrose phosphorylase from Bifidobacterium adolescentis
allows efficient glucosylation of resveratrol, (+)-catechin and
(À)-epicatechin in yields of up to 97% whereas the wild-type
enzyme favours sucrose hydrolysis. Three previously unde-
scribed products are made available. The crystal structure of
the variant reveals a widened access channel with a hydropho-
bic aromatic surface that is likely to contribute to the im-
proved activity towards aromatic acceptors. The generation of
this channel can be explained in terms of a cascade of structur-
al changes arising from the Q345F exchange. The observed
mechanisms are likely to be relevant for the design of other
Scheme 1. Product spectrum and yields obtained with BaSP Q345F and
tailor-made enzymes.
600 mm sucrose as donor: 1, 2: 100 mm (+)-catechin, 3–5: 150 mm (À)-epi-
catechin, 6: 75 mm resveratrol.
Plant polyphenols such as stilbenoids or flavonoids are a focus
of interest due to their antimicrobial[1] and antitumour[2] activi-
ties and their role in lifespan/healthspan extension.[3] However,
they often suffer from low bioavailability due to their poor
water solubility.[4] Glycosylation of natural products is a general
strategy to improve their water solubility and biochemical or
pharmaceutical properties.[5] In this context, resveratrol—one
of the most popular stilbenoids—was selected as a target of
glycosylation. So far, with a few exceptions, enzymatic glycosy-
lation usually relies on glycosyltransferases (GTs), which require
expensive nucleotide diphosphate activated sugars.[5] In con-
trast, glycosidases and transglycosidases (GHs) are a class of
alternative enzymes that use cheap substrates, thus making
them suitable for industrial use, but suffer from a limited ac-
ceptor range.[6] In recent studies, redesign of enzymes through
either random or rational approaches has been used to over-
come these limitations.[7] Here we describe the structure-based
redesign of sucrose phosphorylase (EC 2.4.1.7 GH13)[8] from
Bifidobacterium adolescentis (BaSP) for efficient glycosylation of
polyphenolic substrates (Scheme 1).
BaSP was chosen for its compatibility with organic solvents
and relative thermostability.[9] In addition, crystal structures for
BaSP are available, thus allowing a rational approach to muta-
genesis.[10] Sucrose phosphorylases transfer glucose moieties to
either fructose or phosphate.[10b] Transfer to unnatural accept-
ors including stilbenoids and flavonoids has been observed,
but is generally highly inefficient, most likely because these ac-
ceptors are larger and less polar than the natural substrates.[11]
In an unconventional strategy, we aimed to introduce new
nonpolar interactions between the enzyme and the desired
substrates. We envisaged that the introduction of an aromatic
side chain into the active site of the enzyme might enable p–
p-stacking-mediated coordination of the acceptor substrate. In-
troducing a larger side chain should also cause rearrangement
of adjacent residues, such as the flexible Tyr344,[10b] resulting in
a larger binding pocket.
Gln345, located a short distance away from the acceptor
binding site (+1 site), was thus chosen for exchange with phe-
nylalanine. Interactions between Gln345 and the donor sub-
strate sucrose are limited to OH-3 and OH-6 of its fructose
moiety.[10b] Therefore the distance to the glucose-binding
pocket is sufficient, and interference with sucrose-binding ca-
pability should be tolerable. Minor interference with carbohy-
drate binding in the +1 site was intended in order to inhibit
the unwanted transfer to glucose observed with wild-type
BaSP. The enzyme is known to undergo structural changes
during its catalytic cycle.[10b] Crystal structures of the sucrose-
[a] M. Kraus, Prof. Dr. J. Seibel
Department of Organic Chemistry, University of Würzburg
Am Hubland, 97074 Würzburg (Germany)
E-mail: seibel@chemie.uni-wuerzburg.de
[b] Dr. C. Grimm
Department of Biochemistry, Theodor Boveri-Institute
University of Würzburg
Am Hubland, 97074 Würzburg (Germany)
E-mail: clemens.grimm@uni-wuerzburg.de
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201500514.
ChemBioChem 2016, 17, 33 – 36
33
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
binding conformation (PDB ID: 2GDU) as well as the proposed
phosphate-binding conformation (PDB ID: 2GDV) have been
published.[10b] The conformation of Gln345 is largely unaffected
by the structural rearrangements observed between both con-
formations. We therefore reasoned that the introduced phenyl-
alanine residue might likewise occupy similar positions in both
conformations.[10b] This is of interest because it is not known
which conformation binds aromatic acceptors. Previously, a
Gln345Ala variant was shown to possess a moderately de-
creased affinity for fructose and a strongly decreased affinity
for phosphate whereas no increase in the glucosylation of the
aromatic compound pyridoxine was observed.[12]
A reaction between (À)-epicatechin and sucrose catalysed
by the Q345F variant yielded three detectable products: (À)-
epicatechin-3’-a-O-d-glucoside (3, 23%), (À)-epicatechin-5-O-a-
d-glucoside (4, 29%) and the diglucosylated product epicate-
chin-3’,5-O-a-d-diglucoside (5). Although 3 had previously
been reported, no entries in the SciFinder database exist for
either 4 or 5.[15] Again, the NMR spectra of 5 reveal roughly 8%
of so far uncharacterised side products, presumably additional
regioisomers. To a minor extent, currently uncharacterised side
products also appear accompanying the monoglucosylation
product 4.
The variant Q345F converts resveratrol into resveratrol-3-O-
a-d-glucoside (6), a natural product found in Eleutherococcus
brachypus.[16] Yields of 6 of up to 97% were obtained with
BaSP Q345F, in comparison with 4% for the wild-type enzyme.
Glucosylation of resveratrol merely constitutes a side reaction
for wild-type BaSP. High-performance anion-exchange chroma-
tography (HPAEC) analysis revealed a preference of the wild
type for sucrose hydrolysis and consequently the production
of the glucose disaccharides kojibiose and maltose together
with an uncharacterised saccharide. Under initial reaction con-
ditions the Q345F variant uses 90% of the sucrose consumed
for transfer to resveratrol, and only about 10% hydrolysis
occurs. Increased hydrolysis and disaccharide formation is ob-
served only after most resveratrol is glucosylated. Thus, the
Q345F variant uses sucrose far more efficiently for transfer
than the wild type. Product 6 is, in contrast to resveratrol, an
inefficient substrate for BaSP Q345F. However, if the reaction
was allowed to continue after all resveratrol had been mono-
glucosylated, two further products were detectable in low
amounts of 1.5 and 2.8%. Those were not isolated; however,
the retention time suggests that at least one is a diglucosyla-
tion product.
BaSP wild type and the Q345F variant were expressed as N-
terminally hexahistidine-tagged proteins, allowing affinity pu-
rification by standard protocols. To avoid undesired phosphor-
olysis, reactions were carried out in 3-morpholinopropane-1-
sulfonic acid (MOPS) buffer. Sucrose hydrolysis was used to de-
termine activities, because no transfer reaction suitable both
for wild-type enzyme and for Q345F variant exists. Specific ac-
tivities of 0.716 UmgÀ1 for the wild type and 0.062 UmgÀ1 for
the variant were observed. Despite the fact that the targeted
Gln345 is involved in substrate binding, the Q345F variant re-
tains 8.6% of the specific activity of the wild type towards su-
crose. For the variant under the same reaction conditions, a KM
value for sucrose of 17.5Æ1.04 mm was determined. The wild-
type enzyme reaches Vmax at sucrose concentrations around
1 mm or lower; reactions with lower sucrose concentrations
were not investigated due to assay limitations. BaSP can be ef-
ficiently produced by bacterial overexpression with a yield of
50 mg purified protein per litre culture medium. In addition,
immobilisation techniques for this enzyme, enabling biocata-
lyst reuse, have been reported.[11b,13] Therefore the low specific
activity of BaSP Q345F towards the donor substrate is not con-
sidered a severe drawback for synthetic application.
To determine the mechanism responsible for the altered cat-
alytic properties of the Q345F variant, crystals of BaSP Q345F
were grown in the presence of sucrose, and a crystal structure
in complex with glucose (a hydrolysis product) with a resolu-
tion of 2.7 Š was solved (PDB ID: 5C8B). Only conformations
corresponding to the presumed phosphate-binding conforma-
tion (PDB ID: 2GDV chain B, in complex with b-d-glucose) were
observed with b-d-glucose bound to the glucose-binding site.
The most pronounced differences between the crystal struc-
ture of the Q345F variant and the wild-type enzyme are found
in the region spanning residues 86–166 (domain B), which is
shifted relative to the rest of the protein.[10b] The rearrange-
ment of loops A and B, as well as of b-sheet A (Table 1, Fig-
ure 1A, Figure S2 in the Supporting Information), are responsi-
ble for the altered catalytic properties of BaSP Q345F. Although
many of the hydrogen bonds responsible for glucose coordina-
tion in the active site are disrupted, the substrate orientation
To test for improved selectivity of polyphenols, resveratrol,
(+)-catechin and (À)-epicatechin were chosen as acceptor sub-
strates. Reactions were carried out at 378C with 30% DMSO as
a co-solvent to improve acceptor solubility. Although optimal
temperatures of 48[14] and 608C[13] are reported for BaSP, pro-
longed incubation at these temperatures with high concentra-
tions of organic solvents inactivates the protein.
BaSP Q345F produced (+)-catechin-3’-O-a-d-glucoside (1)
from (+)-catechin and sucrose in 80% yield. An additional, pre-
viously unreported product, (+)-catechin-3’,5-O-a-d-digluco-
side (2), was produced with up to 24% yield. Reactions in the
presence of the wild-type enzyme under the same conditions
afforded no significant amounts of either product. Com-
pound 1 is an inefficient acceptor for further glycosylation by
the Q345F variant. Although sufficient amounts of diglucoside
2 can be produced from catechin, a reaction with purified 1 as
sole acceptor afforded only minor amounts of 2, and signifi-
cant amounts of sucrose hydrolysis and glucose disaccharide
formation were observed. The NMR spectrum of 2 revealed
a side product that could not be fully characterised but might
be the regioisomer with glucose attached to position 7 of the
flavonoid.
Table 1. Key regions in BaSP Q345F.
Structural motif
Loop A
Loop B
Loop C
b-Sheet A
residues
shift of Ca [Š]
133–137
2.5–3.7
154–159
2.4–3.3
336–344
–
88–91, 160–162
1.8–2.1
ChemBioChem 2016, 17, 33 – 36
www.chembiochem.org
34
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
Figure 1. Structural changes in the active site of BaSP Q345F. A) Steric-hindrance-induced shifts in BaSP Q345F. Alignment of BaSP Q345F (5C8B, green) and
the wild type (PDB ID: 2GDV, chain B, white). Steric hindrance induced by Phe345 (red) causes the rotation of Tyr344. Loss of the interaction between Pro134
and Tyr344 causes loop A to shift away from loop C, resulting in a wider access channel. The shift of loop B (residues 154–159, Table 1) provides additional
space in the active site (PDB ID: 2GDV). Additionally Tyr344 in the sucrose-binding conformation of BaSP wild type is shown (PDB ID: 2GDU chain A, yellow).
B) Differences in glucose coordination: alignment of BaSP Q345F (PDB ID: 5C8B, green) and the wild type (PDB ID: 2GDV, chain B, white). In the Q345F variant
the hydrogen bonds between glucose and His88, Gln160, Asp192 and E232 are disrupted. Dashed lines: disrupted H-bonds from the wild type.
remains conserved (Figure 1B). However the bound glucose is
shifted towards b-sheet A by 0.9 Š. Whereas the hydrogen
bonds to Asp50, Asp290 and Glu232 still remain, all other hy-
drogen bonds observed in the wild-type structure (2GDV) are
disrupted (Figure 1B).[10b] In comparison to the wild-type struc-
ture, in BaSP Q345F the carboxylic acid group of Asp192, con-
stituting the catalytic nucleophile, is rotated away from the C1
atom of glucose by 1158 as the c1 angle of the residue changes
from a trans to a gauche+ conformation. In addition, His88 and
Gln160, both involved in 6-OH coordination, are likewise shift-
ed away from the bound glucose. Finally the catalytically es-
sential Glu232 is shifted by 0.9 Š (Figure 1B). These changes
explain the decreased activity towards and affinity for sucrose.
Because no crystal structure of the Q345F variant in complex
with sucrose has so far been obtained, it remains elusive how
the loss of hydrogen bonds between the fructose moiety of
sucrose and Gln345 affects the binding of the substrate. How-
ever it is likely that a weakened interaction with this substrate
due to the loss of O···H bonds between Gln345 and the fruc-
tose moiety contributes to the poor sucrose binding.
sheet A are also repositioned, as well as the rest of the do-
main B. A wider access channel, capable of accommodating
the large polyphenolic acceptors, is the result of the combined
steric-hindrance-induced shifts (Figures 1A and 2). Our crystal
structure of BaSP Q345F shows that the introduction of a steri-
cally demanding residue can result in an enlarged active site
through indirect effects. Furthermore, additional functional
changes may be introduced by this exchange, because Tyr344
and Phe345 create an aromatic surface at the opening of the
active site, and this could coordinate the desired aromatic ac-
ceptors through p–p stacking (Figure 2). Our study adds to the
recent, interesting findings of Desmet et al.[7e]
In order to create a new tool for the glucosylation of poly-
phenolic phytochemicals, we introduced an aromatic amino
acid, as a potential partner for p–p stacking, into the active
site of sucrose phosphorylase. In the Q345F variant the access
channel is not enlarged by the conventional strategy of replac-
ing a larger side chain with a smaller one. In contrast, a cascade
Tyr344 plays a key role in the wild-type catalysis and under-
goes major structural rearrangements during the catalytic
cycle along with the entire loop C.[10b] In the Q345F variant,
Tyr344 is not rotated as far out of the active site as in the
2GDU crystal structure, which corresponds to the sucrose bind-
ing conformation of the wild type, Figure 1A).[10b] In addition,
loop C (Figure 1A, Figure S2) of BaSP Q345F is found in an ori-
entation that is characteristic for the phosphate-binding con-
formation. Steric hindrance due to the exchange of Gln345 for
Phe forces the neighbouring Tyr344 to rotate by 858 out of the
active site (Figure 1A). In the phosphate-binding conformation
of the wild type (2GDV chain B), van der Waals interactions
between Tyr344 and Pro134 are present.[10b] Rotation of Tyr344
abolishes the interaction with Pro134 and favours the shifting
of loop A. As a consequence, the more distant loop B and b-
Figure 2. Access channel of BaSP Q345F displaying the nonpolar aromatic
surface created by Phe345 and Tyr344.
ChemBioChem 2016, 17, 33 – 36
www.chembiochem.org
35
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Communications
[5] R. W. Gantt, P. Peltier-Pain, J. S. Thorson, Nat. Prod. Rep. 2011, 28, 1811–
1853.
[6] T. Desmet, W. Soetaert, P. Bojarova, V. Kren, L. Dijkhuizen, V. Eastwick-
Field, A. Schiller, Chem. Eur. J. 2012, 18, 10786–10801.
[7] a) J. Gçrl, M. Timm, J. Seibel, ChemBioChem 2012, 13, 149–156; b) M.
Timm, J. Gorl, M. Kraus, S. Kralj, H. Hellmuth, R. Beine, K. Buchholz, L.
Dijkhuizen, J. Seibel, ChemBioChem 2013, 14, 2423–2426; c) F. A.
Shaikh, S. G. Withers, Biochem. Cell Biol. 2008, 86, 169–177; d) Z. Sun, A.
Ilie, M. T. Reetz, Angew. Chem. Int. Ed. 2015, 54, 9158–9160; Angew.
Chem. 2015, 127, 9288–9290; e) M. E. Dirks-Hofmeister, T. Verhaeghe, K.
De Winter, T. Desmet, Angew. Chem. Int. Ed. 2015, 54, 9289–9292;
Angew. Chem. 2015, 127, 9421–9424.
of conformational changes is induced through the replace-
ment of Gln345 with a spatially more demanding Phe residue.
By this means, glucosylation of aromatic acceptors with yields
of up to 97% is enabled. In addition, the three previously un-
described compounds 2, 5 and 3 are now available through
enzymatic glucosylation. The structural data elucidate further
how a single amino acid exchange can affect enzymatic func-
tion. We are confident that these insights will aid in further de-
velopment of tailor-made biocatalysts.
[8] V. Lombard, H. Golaconda Ramulu, E. Drula, P. M. Coutinho, B. Henrissat,
Nucleic Acids Res. 2014, 42, D490–D495.
Acknowledgements
[9] A. Cerdobbel, K. De Winter, D. Aerts, R. Kuipers, H. J. Joosten, W. Soe-
taert, T. Desmet, Protein Eng. Des. Sel. 2011, 24, 829–834.
[10] a) D. Sprogøe, L. A. van den Broek, O. Mirza, J. S. Kastrup, A. G. Voragen,
M. Gajhede, L. K. Skov, Biochemistry 2004, 43, 1156–1162; b) O. Mirza,
L. K. Skov, D. Sprogoe, L. A. van den Broek, G. Beldman, J. S. Kastrup, M.
Gajhede, J. Biol. Chem. 2006, 281, 35576–35584.
[11] a) D. Aerts, T. F. Verhaeghe, B. I. Roman, C. V. Stevens, T. Desmet, W. Soe-
taert, Carbohydr. Res. 2011, 346, 1860–1867; b) K. De Winter, K. Verlin-
den, V. Kren, L. Weignerova, W. Soetaert, T. Desmet, Green Chem. 2013,
15, 1949–1955; c) S. Kitao, T. Ariga, T. Matsudo, H. Sekine, Biosci. Bio-
technol. Biochem. 1993, 57, 2010–2015.
We thank the team of beamline ID29 of ESRF Grenoble, France
for their excellent support during data collection. We thank Jan
Wendrich for supplying the HPLC analysis and Julian Gçrl for
generating the images.
Keywords: biocatalysis · polyphenol glycosylation · protein
engineering · substrate promiscuity · sucrose phosphorylases
[12] T. Verhaeghe, M. Diricks, D. Aerts, W. Soetaert, T. Desmet, J. Mol. Catal. B
2013, 96, 81–88.
[13] A. Cerdobbel, K. De Winter, T. Desmet, W. Soetaert, Biotechnol. J. 2010,
5, 1192–1197.
[14] L. A. M. van den Broek, E. L. van Boxtel, R. P. Kievit, R. Verhoef, G. Beld-
man, A. G. J. Voragen, Appl. Microbiol. Biotechnol. 2004, 65, 219–227.
[15] P. Aramsangtienchai, W. Chavasiri, K. Ito, P. Pongsawasdi, J. Mol. Catal. B
2011, 73, 27–34.
[1] a) M. Friedman, Mol. Nutr. Food Res. 2007, 51, 116–134; b) T. P. Cushnie,
A. J. Lamb, Int. J. Antimicrob. Agents 2011, 38, 99–107; c) C. Ferreira,
D. C. Soares, M. T. Nascimento, L. H. Pinto-da-Silva, C. G. Sarzedas, L. W.
Tinoco, E. M. Saraiva, Antimicrob. Agents Chemother. 2014, 58, 6197–
6208.
[2] a) E. R. Kasala, L. N. Bodduluru, R. M. Madana, A. K. V, R. Gogoi, C. C.
Barua, Toxicol. Lett. 2015, 233, 214–225; b) M. Jang, L. Cai, G. O. Udeani,
K. V. Slowing, C. F. Thomas, C. W. Beecher, H. H. Fong, N. R. Farnsworth,
A. D. Kinghorn, R. G. Mehta, R. C. Moon, J. M. Pezzuto, Science 1997,
275, 218–220.
[16] H.-b. Hu, J. Fan, Bull. Korean Chem. Soc. 2009, 30, 703–706.
[3] K. S. Bhullar, B. P. Hubbard, Biochim. Biophys. Acta 2015, 1852, 1209–
1218.
[4] a) L. Biasutto, M. Zoratti, Curr. Drug Metab. 2014, 15, 77–95; b) C. K.
Singh, M. A. Ndiaye, N. Ahmad, Biochim. Biophys. Acta Mol. Basis Dis.
2015, 1852, 1178–1185.
Manuscript received: October 6, 2015
Accepted article published: November 3, 2015
Final article published: December 2, 2015
ChemBioChem 2016, 17, 33 – 36
www.chembiochem.org
36
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim