Molecular Design of Non-Leloir Furanose-Transferring Enzymes from an α-L-Arabinofuranosidase: A Rationale for the Engineering of Evolved Transglycosylases

Article abstract of DOI:10.1021/acscatal.5b00949

The vast biodiversity of glycoside hydrolases (GHs) constitutes a reservoir of readily available carbohydrate-acting enzymes that employ simple substrates and hold the potential to perform highly stereopecific and regioselective glycosynthetic reactions. However, most GHs preferentially hydrolyze glycosidic bonds and are thus characterized by a hydrolysis/transglycosylation partition in favor of hydrolysis. Unfortunately, current knowledge is insufficient to rationally modify this partition, specifically mutating key molecular determinants to tip the balance toward transglycosylation. In this study, in the absence of precise knowledge concerning the hydrolysis/transglycosylation partition in a hydrolytic GH51 α-L-arabinofuranosidase, we describe how an iterative protein engineering approach has been used to create the first non-Leloir transarabinofuranosylases. In the first step, random mutagenesis yielded a point mutation (R69H) at a position that is highly conserved in clan GH-A. Characterization of R69H revealed that this enzyme displays high transglycosylation activity but severely reduced (61-fold) activity on pNP-α-L-arabinofuranoside. Upon recombination of R69H with other point mutations selected using semirational or in silico approaches, transfer rates close to 100% and transarabinofuranosylation yields of the main (1→2)-linked oligosaccharide product of 80% (vs 11% for the wild-type) were obtained. Combining data presented here with knowledge drawn from the literature, we suggest that the creation of non-Leloir transglycosylases necessarily involves the destabilization of the highly developed transition states that characterize the predominantly hydrolytic exo-acting GHs; this is an efficient way to prevent ubiquitous water molecules from performing the deglycosylation step. (Figure Presented).

Full text of DOI:10.1021/acscatal.5b00949

Subscriber access provided by UB + Fachbibliothek Chemie | (FU-Bibliothekssystem)
Article
Molecular design of non-Leloir furanose-transferring
enzymes from an #-L-arabinofuranosidase: a rationale
for the engineering of evolved transglycosylases
Bastien BISSARO, Julien Durand, Xevi Biarnés, Antoni Planas,
Pierre Frédéric Monsan, Michael J. O'Donohue, and Regis Faure
ACS Catal., Just Accepted Manuscript • DOI: 10.1021/acscatal.5b00949 • Publication Date (Web): 15 Jun 2015
Downloaded from http://pubs.acs.org on July 1, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
ACS Catalysis is published by the American Chemical Society. 1155 Sixteenth Street
N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 17
ACS Catalysis
1
2
3
4
5
6
7
8
9
Molecular Design of NonꢀLeloir FuranoseꢀTransferring Enꢀ
zymes from an αꢀ ꢀArabinofuranosidase: a Rationale for the
L
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Engineering of Evolved Transglycosylases
Bastien Bissaro,†,‡,§ Julien Durand,†,‡,§ Xevi Biarnés,∥ Antoni Planas,∥ Pierre Monsan,†,‡,§,⊥, Mi-
chael J. O'Donohue*,†,‡,§ and Régis Fauré*,†,‡,§
†
Université de Toulouse; INSA,UPS,INP; LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France
‡
INRA, UMR792, Ingénierie des Systèmes Biologiques et des Procédés, F-31400 Toulouse, France
§
CNRS, UMR5504, F-31400 Toulouse, France
∥
Laboratory of Biochemistry, Institut Químic de Sarrià, Universitat Ramon Llull, Via Augusta, 08017 Barcelona, Spain
⊥
Toulouse White Biotechnology, UMS INRA/INSA 1337, UMS CNRS/INSA 3582, 3 Rue des Satellites 31400 Toulouse,
France
ABSTRACT: The vast biodiversity of glycoside hydrolases (GHs) constitutes a reservoir of readily-available carbohydrate-
acting enzymes that employ simple substrates and hold the potential to perform highly stereopecific and regioselective
glycosynthetic reactions. However, most GHs preferentially hydrolyze glycosidic bonds and are thus characterized by a
hydrolysis/transglycosylation partition in favor of hydrolysis. Unfortunately, current knowledge is insufficient to rational-
ly modify this partition, specifically mutating key molecular determinants to tip the balance towards transglycosylation.
In this study, in the absence of precise knowledge concerning the hydrolysis/transglycosylation partition in a hydrolytic
GH51 α-L-arabinofuranosidase, we describe how an iterative protein engineering approach has been used to create the
first ‘non-Leloir’ transarabinofuranosylases. In the first step, random mutagenesis yielded a point mutation (R69H) at a
position that is highly conserved in clan GH-A. Characterization of R69H revealed that this enzyme displays high
transglycosylation activity, but severely reduced (61-fold) activity on pNP-α-L-arabinofuranoside. Upon recombination of
R69H with other point mutations selected using semi-rational or in silico approaches, transfer rates close to 100% and
transarabinofuranosylation yields of the main (1ꢀ2)-linked oligosaccharide product of 80% (vs 11% for the wild-type) were
obtained. Combining data presented here with knowledge drawn from the literature, we suggest that the creation of non-
Leloir transglycosylases necessarily involves the destabilization of the highly developed transition states that characterize
the predominantly hydrolytic exo-acting GHs; this being an efficient way to prevent ubiquitous water molecules from
performing the deglycosylation step.
bohydrates.3–6 In the case of GTs, the vast majority is des-
INTRODUCTION
ignated as sugar nucleotide-dependent (Leloir) enzymes,
while r-GHs that predominantly catalyze glycosynthetic
reactions have often been referred to as non-Leloir
transglycosylases (or TGs).
Considering the importance of carbohydrates in biologi-
cal processes (i.e., in molecular recognition, the storage of
energy and also for structural purpose), the improvement
and widening of our ability to synthesize glycosidic motifs
and glycoconjugates is of utmost importance. In this re-
spect, chemo-enzymatic approaches are proving to be
highly useful alternatives to purely chemical methods,
although these have so far focused mainly on hexopyra-
noses, with furanoses being left aside, despite the fact that
the synthesis of furanosides using conventional methods
has often provided to be quite tricky.1,2
Regarding GHs, these are potentially very attractive for
chemo-enzymatic strategies, because they are highly
abundant, display a wide range of substrate specificities
and often employ readily accessible, simple activated
glycosides as donor compounds. The synthetic capability
of r-GHs resides in their double-displacement mecha-
nism, the outcome of which can be either hydrolysis or
transglycosylation, the latter being dependent on the
relative availability and the recognition of acceptors other
than water in the reaction medium (Figure 1).7 Neverthe-
less, most r-GHs principally display hydrolytic activity in
Carbohydrate active enzymes (CAZymes), such as glyco-
syltransferases (GTs) and so-called retaining glycoside
hydrolases (r-GHs), are the workhorses of chemo-
enzymatic approaches for the synthesis of complex car-
1
ACS Paragon Plus Environment

ACS Catalysis
Page 2 of 17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Two-step displacement mechanism of retaining GHs. Along the reaction mechanism, a Michaelis complex (E.S) is formed between the
enzyme and the donor substrate allowing a nucleophilic attack of the catalytic nucleophile residue onto the anomeric carbon, combined with a
protonation of the interglycosidic oxygen by the catalytic acid/base residue, leading to a release of the leaving group (LG) and formation of the
covalent glycosyl-enzyme intermediate (E-S). This step is so-called the glycosylation step. The donor LG can be either an activated group (e.g.,
pNP) or a sugar (e.g., fructose for glucansucrases). Regarding the second step, denoted as deglycosylation step, the covalent glycosyl-enzyme
intermediate can be either attacked by a water molecule (hydrolysis, R = H) or an external hydroxylated acceptor (transglycosylation, R = sugar,
alkyl chain, etc.), that are activated by the deprotonated form of the acid/base catalytic residue. In the case of secondary hydrolysis, the transgly-
cosylation product becomes a donor substrate with subsequent deglycosylation step involving water as acceptor. In this study the donor sub-
strate is pNP-α-L-Araf and the acceptor is xylotriose (Figure S1).
aqueous medium, and only a minority of r-GHs can be
described as TGs. During reactions catalyzed by TGs, the
covalent glycosyl-enzyme intermediate is most readily
deglycosylated by incoming sugar acceptors rather than
water molecules, even in reaction media displaying high
water activity, to the extent that the formation of glyco-
sidic bonds is dominant and hydrolysis is low or even
undetectable. Intriguingly, TGs are scattered among sev-
eral CAZy GH families8 (e.g., XET from GH16, amylosu-
are mostly only applicable to single GH families and fail
to answer the fundamental question of how TGs perform
glycosynthesis in environments dominated by water.10–14
Indeed, only very recently has a more integrated approach
been used to review current literature and provide more
generic understanding of the H/T
partition in r-GHs.9
Apart from the glycosynthase strategy15,16 there are no
clear-cut protein engineering guidelines to convert a
hydrolytic r-GH into a TG (using non-fluorinated activat-
ed simple substrates), which led us in the present study to
adopt a hybrid approach to engineer the hydrolytic GH51
α-L-arabinofuranosidase from Thermobacillus xylanyliti-
cus (TxAbf). For over a decade, this enzyme has been the
subject of many studies that have not only characterized
crases from GH13, trans-sialidases from GH33
-
www.cazy.org and www.cazypedia.org), although no non-
Leloir transfuranosylases have been described to date.
Significantly, TGs are very frequently highly homologous
to their hydrolytic family counterparts (e.g., amylosu-
crases/amylases from GH13). The implication of this ob-
servation is that the outcome of catalysis (i.e., hydrolysis
or transglycosylation) is not determined by any single
active site topology or protein 3-D structure, but is prob-
ably associated with a range of more subtle molecular
features.9
the ability of TxAbf to hydrolyze bonds linking α-
L-
arabinofuranosyl (α- -Araf) moieties to -xylopyranosyl
L
D
(D-Xylp) units in arabinoxylans and arabinoxylo-
oligosaccharides (AXOS) of plant origin,17 but also its
ability to perform transglycosylation. In several studies,
we have shown how, in the presence of appropriate sugar
The quest to understand the underlying molecular de-
terminants of the hydrolysis/transglycosylation (H/T)
partition in r-GHs has challenged researchers for at least
20 years and has given rise to numerous hypotheses that
or alcohol acceptors, the
wards transglycosylation, allowing the synthesis of a vari-
ety of furanose-based compounds, including -arabino-
and
-galactofuranosides.18,19 Moreover, in recent work
H/T partition can be driven to-
L
D
2
ACS Paragon Plus Environment

Page 3 of 17
ACS Catalysis
1
2
3
4
5
6
7
8
using in vitro random mutagenesis we managed to tip the
balance further towards transglycosylation, creating
novel glycosynthetic tools for the synthesis of complex
furanoside-based structures that would be tricky to access
using conventional synthetic chemistry.20,21
In the present study, we have pursued our efforts to engi-
neer TxAbf, this time combining our previously-acquired
knowledge to develop a hybrid engineering approach that
has relied on random mutagenesis, site-saturation muta-
genesis at previously identified hot-spots and site-
directed mutagenesis driven by in silico predictions. In
this step-wise manner, we have succeeded in creating
what can only be described as the very first non-Leloir
transarabinofuranosylases, the biochemical characteris-
tics of which are thoroughly described herein.
employed following the suppliers recommendations. To
H
/
T
generate optimal genetic diversity, three separate reac-
tions were performed (50 µL each) in which the quantities
of pET24-TxAbf-L352M template DNA differed (50, 100
and 150 ng), but the amounts of dNTP mix (800 µM),
primers mix (250 ng.µL-1 each) and Mutazyme II DNA
polymerase (2.5 IU) were identical. The reactions were
performed using a PCR thermocycler and the following
protocol: one initial denaturing step at 95 °C for 2 min,
followed by 25 cycles at 95 °C for 1 min, 55 °C for 1 min, 72
°C for 2 min and a final elongation step at 72 °C for 10
min. For the PCR reactions, the sequences of the forward
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
and
reverse
primers
were
5’-
GGAATTCCATATGAACGTGGCAAGCCGGGTAGTCG-3’
and 5’-CGATGGCCCACTACGTGAACCATC-3’ respective-
ly. The forward primer contains a NdeI restriction site and
the reverse primer anneals to a sequence within the
pET24 vector. Accordingly, the size of the amplicon was
1933 bp, representing the target TxAbf-encoding sequence
(1490 bp), associated with a part of the pET24 vector (433
bp). After purification using a DNA clean and concentra-
tor kit (Zymo Research, USA), the amplicon was digested
using NdeI (20 IU) and HindIII (20 IU) at 37 °C for 4 h in a
buffered (1 X Neb2 buffer) reaction (50 µL). Afterwards,
the reaction was stopped by heating at 65 °C for 20 min
and an aliquot (20 IU) of DpnI was added. Following fur-
ther incubation at 37 °C for 1 h, the DNA fragments were
purified and eluted in 10 µL of ddH2O (Zymo Research
kit). The resulting purified, mutated TxAbf-encoding
sequence was then ligated with HindIII/NdeI-digested
pET-24 vector. To achieve this, five separate ligation reac-
tions (15 µL each) were prepared in PCR tubes, each con-
taining 55 ng of mutated TxAbf-encoding DNA and 66 ng
of linearized pET24 DNA (i.e., a 3:1 molar ratio), 1.5 µL of
T4 DNA ligase buffer (10X) and 1 µL of T4 DNA ligase
(400 IU). The reactions were incubated for 12 h at 16 °C
and then 2 µL aliquots were used to transform aliquots
(50 µL) of chemically competent E. coli Top10 cells. Over-
all, 24 transformation reactions were performed, after
which the bacteria were spread onto LB-agar Petri dishes
containing kanamycin (40 µg.mL-1). Colonies were al-
lowed to develop for 14 h at 30 °C before being suspended
in 500 µL of liquid LB medium containing kanamycin,
which were added to the surface of each plate surface.
After recovery, the various cell suspensions were pooled
and the total volume was centrifuged (5000 x g, 10 min)
providing a single cell pellet, which was used to prepare
purified plasmid DNA. This DNA sample constituted the
L352M-based random library, which was stored at -20 °C
until further use.
EXPERIMENTAL SECTION
Substrates and chemicals
Unless otherwise stated, routine experimental work was
performed using chemicals purchased from Sigma-
Aldrich (Illkirch, France) and molecular biology reagents
purchased from New England Biolabs (Evry, France). The
substrate, 4-nitrophenyl α-
L
-arabinofuranoside (pNP-α-L-
Araf), was purchased from CarboSynth (Compton, UK)
and xylotriose and XOS were purchased from Wako
Chemicals GmbH (Neuss, Germany). Regarding other aryl
α-L-arabinofuranosides used in this study, these were
synthesized in-house using a modified version (Support-
ing Information) of a previously described synthetic
route.22
Mutagenesis, protein expression, and purification
Site-directed mutagenesis
The plasmid pET24-TxAbf (the original pET vector was
from Novagen, Fontenay-sous-Bois, France) was used as
template DNA for in vitro mutagenesis using the Quik-
Change Site-Directed Mutagenesis kit (Agilent, Courta-
boeuf, France), following the manufacturer’s instructions.
For recombination of mutations R69H, L352M and G179F,
the following primers (5’-3’) were employed (underlined
codon and mutated base in bold):
R69H
G179F
L352M
TCCGGTCCTCCACTGGCCGGGCG
GGCAACGAGAACTGGTTCTGCGGCGGCAACAT
CAGCTCGTCAACGTGATGCAATCCGTCATCC
The successful introduction of mutations was confirmed
by DNA sequencing (GATC Biotech, Mulhouse, France).
Large scale enzyme expression and purification
To produce larger batches of wild-type and mutated re-
combinant TxAbf, protein expression and purification
were performed as previously described.23
Screening
For library screening, 10 ng of the plasmid library DNA
stock were used to transform chemically competent E.
coli BL21 DE3 star cells, which were then grown at 37 °C
for 12 h on a nitrocellulose membrane (0.45 µm, 20 x 20
cm) placed on selective LB agar medium contained within
a Q-tray, thus allowing the development of small colonies
(Ø < 0.5 mm). Likewise, 8 Q-Trays were prepared, repre-
Creation and screening of random mutant libraries
Design of random libraries
A randomly mutated gene library was constructed using
pET24-TxAbf-L352M as DNA template. For mutagenesis,
the GeneMorph II Random Mutagenesis kit (Agilent) was
3
ACS Paragon Plus Environment

ACS Catalysis
Page 4 of 17
1
2
3
4
5
6
7
8
senting approximately 40,000 clones. Once colonies were
15 min in a Peltier heating device, before initiating the
formed, the nitrocellulose membranes were transferred
onto selective, buffered (sodium phosphate buffer 100
mM, pH 7.0, agar 15 g.L-1) minimal medium, containing
reaction through the addition of lysates. Once initiated,
the progression of reactions was monitored by removing
20 µL aliquots every 6 min (4 points in total, each reac-
tion lasting 24 min). These were immediately combined
with 180 µL of 1M Na2CO3 and absorbance (405 nm) was
determined. Using a simple, purpose-designed Excel mac-
ro, the activity (IU/mL) of the different clarified lysates
was calculated in the presence and absence of acceptor
and activity ratios were derived for each one. The relative
mean error of this method was 4% for any single lysate.
For each microtiter plate, standard deviations (σ) were
calculated and 1.96 σ was represented, which is the ap-
proximate value for the 97.5 percentile point of a normal
distribution.
IPTG (0.4 mM) and 5-bromo-4-chloro-3-indolyl α-
L-
arabinofuranoside, X-α- -Araf (0.1 mM final concentra-
L
tion, initially solubilized in DMSO), and incubated for 2 h
at 30 °C. In parallel, a control membrane was prepared on
which only colonies expressing the parental L352M vari-
ant were present. After 2 h, all membranes were examined
and colonies (~ 400) displaying pale blue color were se-
lected and transferred into four 96-well microtiter plates
containing 200 µL of LB-kanamycin (40 µg.mL-1). Plates
were incubated overnight at 30 °C with shaking (700
rpm), after which an aliquot (100 µL) was removed from
each well and combined with a glycerol solution (25%, v/v
final concentration) to create library back-ups. The back-
ups were archived at -80 °C and the remaining cell cul-
tures were used to prepare cell-free lysates.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N216X site saturation library design and screening
A site saturation mutagenesis experiment of the type NNS
was performed at position N216. To achieve this, the
TxAbf-encoding sequence was first mutated at position
216, introducing a stop codon (TAG) using QuickChange
II Site-Directed Mutagenesis kit (Stratagene, Amsterdam,
The Netherlands) and an appropriate primer pair (for-
ward primer is shown below). This mutated DNA was
then used as a template for NNS mutagenesis, which was
performed using the same mutagenesis kit and appropri-
ate primers (forward primer is shown below).
Preparation of cell-free lysates of library clones
Using the previous microtiter cultures of the selected
library clones to inoculate fresh LB-kanamycin medium
(500 µL) in 96-deepwell (2 mL/well) microtiter plates,
precultures were prepared and grown overnight (30 °C,
700 rpm). Afterwards, 100 µL of each pre-culture was used
to inoculate 900 µL aliquots of fresh LB-kanamycin con-
tained in wells of 96-well deepwell microtiter plates. The-
se cultures were incubated at 37 °C and 700 rpm during 2
h and then IPTG (0.5 mM final) was added and incuba-
tion was pursued (18 h incubation at 18 °C and 700 rpm).
After, cells from each culture were recovered using cen-
trifugation (2250 x g, 20 min) and suspended in 300 µL of
Tris-HCl buffer (20 mM, pH 8.0) containing lysozyme (0.5
mg.mL-1), and incubated 1 h at 30 °C, followed by freezing
overnight at -80 °C. After thawing (30 °C, 30 min), the cell
suspensions were heat-treated (75 °C, 30 min) and clari-
fied lysates were prepared by centrifugation (2250 x g, 50
min, 4 °C). Finally, lysates were transferred into fresh 96-
well microtiter plates and stored at 4 °C.
N216Stop TGCGTGCGGCGCGTAGACGGCCGACTACCA
N216X
TGCGTGCGGCGCGNNSACGGCCGACTACCA
Taking into account the theoretical size of the NNS li-
brary, it was considered that the selection and screening
of 96 clones would cover 95% of the diversity.24 Therefore,
96 randomly picked E. coli BL21(DE3) colonies were used
for enzyme production in microtiter plate format as pre-
viously described. After, the 96 clarified lysates were
submitted to automated kinetic analyses in the absence
and presence of a mixture of xylo-oligosaccharides as
acceptor as described above. One mutant revealed an
increased ratio (R = 1) relative to TxAbf (R=0.4), and was
identified as N216W after sequencing.
Secondary screening of selected clones
In silico screening of TxAbf against xylotriose
The binding affinity of a library of TxAbf mutants for
xylotriose was screened in silico using a model of the 3D
The 400 cleared lysates representing individual library
clones were used to perform kinetic analyses using an
automated protocol that provided the means to simulta-
neously perform 96 analyses. Likewise, it was possible to
compare the activities of individual lysates in the pres-
ence and absence of acceptor, and thus derive an activity
ratio (R), which was equal to Vi(donor + acceptor)/Vi(donor). Us-
ing a liquid handling system (Genesis RSP 200, Tecan),
reactions were prepared in wells of 96-well polypropylene
PCR microtiter plates (Corning®, Sigma Aldrich). The
basic reaction mixture used to determine Vi(donor) values
was composed of 135 µL substrate solution (15 mM pNP-α-
structure of the covalent β-L-Araf-TxAbf intermediate
involving a xylotriosyl moiety (Supporting Information).
Sequence variations were allowed only at those sequence
positions located within 10 Å of TxAbf subsites +2’, +1, +2
and +3.
Determination of kinetic parameters
Hydrolytic behavior
The kinetic parameters describing hydrolysis catalyzed by
wild-type or mutant TxAbf were determined using a con-
tinuous assay in which the release of different phenolates
was monitored using either an Eon spectrophotometer
(BioTek), or in the case of PhOH, a Cary 100 Bio UV-Vis
spectrophotometer (Agilent). Reactions were performed
at 45 °C, (a choice of temperature that translates a com-
L
-Araf in 50 mM sodium phosphate, pH 7.0) and 15 µL
(10% of final volume) of lysate. To determine the value of
Vi(donor + acceptor), the reaction was performed as before, but
10 mM xylotriose (i.e., acceptor) was added to the reac-
tion mixture. To perform reactions, the substrate solution
(with or without acceptor) was first incubated at 45 °C for
4
ACS Paragon Plus Environment

Page 5 of 17
ACS Catalysis
1
2
promise between enzyme activity and stability) in wells of
KM. In these reactions, the concentration of pNP-α-L-Araf
microtiter plates containing 50 mM sodium phosphate
buffer, pH 7.0, substrate, 0.1 mg.ml-1 BSA and enzyme, in a
final reaction volume of 200 µL. Taking into account the
fact that the substrates were initially dissolved in DMSO,
the final DMSO concentration in reactions was always
less than 2% (v/v). To perform reactions, the buffered
substrate mixture (180 µL) contained in microtiter plate
wells was equilibrated at 45 °C for 5 min, and then 20 µL
of an appropriately diluted enzyme solution was added
and the release of LG was monitored at 400 (for pNP, 3-
NP and 3,4-dNP) or 425 nm (for NCP). For reactions in-
was varied between 0.15 and 18 mM and pH was varied in
the range 4.5 to 8.7 using 100 mM citrate-sodium phos-
phate buffer (pH 4.5-7.0) and 100 mM sodium phosphate
buffer (7.0-8.7), with a common point being measured in
both buffers at pH 7.0 for each enzyme and each substrate
concentration. Control experiments in the absence of
enzyme were performed at each pH value in order to
monitor any spontaneous hydrolysis of the substrate dur-
ing the duration of the reaction (30 min). Finally the ratio
kcat/KM was plotted as a function of pH and fitted to a
bell-shaped activity profile according to equation (3).25
This analysis provided the apparent pKa values corre-
sponding to the nucleophile (pKa1) and the acid/base
(pKa2) catalytic amino acids, which were determined by
nonlinear least-squares fitting using Microsoft Excel 2010.
When a ‘tail’ was observed on the basic side of the pH-
dependent profile, equation (4) was employed.26
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
volving Ph-α-L-Araf, these were performed in a similar
way, but in a final volume of 500 µL in quartz cuvettes,
and the release of phenol was monitored at 280 nm. Con-
trol reactions containing all the reactants except the en-
zyme were used to correct for slight spontaneous hydroly-
sis of the substrates, which was observed in the case of
3,4-dNP-α-L-Araf and NCP-α-L-Araf. Furthermore, stand-
ard curves of substrates were prepared to correct for ab-
sorbance of the phenyl group in its bound form (i.e., Ph-
ꢀ
ꢊ
= (ꢀꢄ
)
ꢇꢈꢉ. (ꢊ ꢋ ꢊꢌ
)
(Eq. 3)
(Eq. 4)
ꢁꢂꢃ
ꢁꢂꢃ
ꢍꢎꢏꢐꢍꢑ
ꢍꢑꢐꢍꢎꢒ
ꢋ ꢊꢌ
ꢄ
ꢅ
ꢆ
ꢓ
ꢎ
ꢓ
ꢎ
α-L-Araf). Moreover, a standard solution of pNP was used
ꢁꢂꢃ
ꢆ
ꢁꢂꢃ
ꢅ
(ꢍꢑꢐꢍꢎ
)
ꢒ
(
)
ꢋ (
) . ꢊꢌ
ꢒ
ꢏ
ꢀ
ꢁꢂꢃ
to compare the difference in the response coefficients of
the Eon and Cary spectrophotometers. This proved to be
less than 6%. Initial reaction rates were determined from
the linear regions of time-dependent plots, which corre-
sponded to 5-15% consumption of the substrate. Substrate
concentrations were typically varied from 0.1 to 18 mM.
One unit (IU) of enzyme specific activity corresponds to
the amount of enzyme releasing one µmol of LG per mi-
nute. Steady-state kinetic parameters were determined
using non-linear regression functions embedded in the
SigmaPlot 11.0 software (Systat software Inc, Ritme, Paris,
France). Three models were employed, including the
classical Michaelis-Menten model, inhibition by excess
substrate model, and a modified Michaelis-Menten that
accounts for donor-induced activation through self-
condensation. Steady-state kinetic parameters were then
analyzed using the Brønsted correlation according to
equations (1) and (2).
=
ꢍꢎꢏꢐꢍꢑ
ꢍꢑꢐꢍꢎꢒ
ꢍꢑꢐꢍꢎꢔ
.(ꢊ ꢋ ꢊꢌ )
ꢄ
ꢊ ꢋ ꢊꢌ
ꢋ ꢊꢌ
ꢅ
Acceptor impact on global activity
To measure the influence of acceptor substrate concen-
tration on pNP release, reactions were performed in
which the donor (pNP-α-L-Araf) concentration was held
constant (5 mM), while the acceptor (xylotriose) concen-
tration was varied. When acceptor-mediated inhibition
was observed, a mixed model described by equation (5)
was employed to fit experimental data, whereas equation
(6) was used when acceptor-mediated activation oc-
curred.
ꢊ
ꢙ
ꢝ.ꢞ
ꢜ
ꢕ
ꢛ
[ꢠ]
ꢛ
ꢄ
ꢖ
[ꢠ]
ꢖ
ꢚ
=
with αꢟ = 1 +
and α = 1 +
ꢚ
ꢕ
ꢄ
ꢖ
ꢙ
ꢝ.ꢄ ꢋꢞ
ꢅ
ꢗꢂꢘ
ꢜ
ꢛ
ꢚ
(Eq. 5)
[ ]
. ꢡꢢꢣꢢꢤ .[ꢥꢦꢦꢧꢨꢩꢢꢤ]
ꢕ
ꢗꢂꢘ
=
(Eq. 6)
ꢂꢁꢁꢮꢍꢃꢫꢭ
ꢅ
ꢪꢫꢬꢫꢭ
[
]
[
] [
]
[
]
ꢄ
. ꢥꢦꢦꢧꢨꢩꢢꢤ
ꢋ
ꢡꢢꢣꢢꢤ . ꢥꢦꢦꢧꢨꢩꢢꢤ ꢋ ꢄ
. ꢡꢢꢣꢢꢤ
ꢅ
log(kcat/KM) = βLG1.pKaLG + c1
log(kcat) = βLG2.pKaLG + c2
(Eq. 1)
(Eq. 2)
Monitoring transglycosylation
Time course NMR monitoring
1
To measure solvent kinetic isotope effects (sKIE), reac-
tions were performed at 45 °C, pH 7.0 as described earlier,
using NCP-α-L-Araf in D2O. In reactions involving either
To monitor reactions by collecting H NMR spectra on a
Bruker Avance 500 spectrometer, enzyme solutions were
first prepared by 10-fold dilution in D2O (99.90%) and
then concentration on an Amicon® Ultra filter (regenerat-
ed cellulose 10 kDa, Millipore), performing the operation
twice. Afterwards, reactions were performed at 45 °C in
R69H or R69H-L352M, substrate concentration was 2
mM, while those involving TxAbf and L352M, had a sub-
strate concentration of 5 mM. The D2O content of the
reaction was varied (35, 50, 65, 80 and 100%) adding ap-
propriate amounts of H2O to reach 180 µL. The enzymes
were pre-diluted in D2O and 20 µL of enzyme solution
was used in each reaction, providing a final volume of 200
µL.
NMR tubes containing 600 µL of buffered D2O, pNP-α-L-
Araf as donor and xylotriose as acceptor, at a ratio of 1:2
(15 and 30 mM, respectively). To obtain a pD value of 5.8,
deuterated acetate (Euriso-Top, France) buffer was em-
ployed, while pD 7.1 was achieved using a deuterated
sodium phosphate buffer, which was prepared by dissolv-
ing sodium phosphate in D2O, followed by lyophilization.
This two-step protocol was repeated three times to obtain
To determine the effect of pH on enzymes, an analysis of
kcat/KM was performed, using the previously described
protocols to determine in duplicate the values of kcat and
5
ACS Paragon Plus Environment

ACS Catalysis
Page 6 of 17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2. (A) Location of R69H and L352M within TxAbf subsite -1 and (B) putative impact of the mutation R69H on the electronic displacement
system during the establishment of the transition state at the deglycosylation step. R69 is at H-bond distance (2.86 Å) from the nucleophile
catalyst E298, E176 playing the role of acid/base. Figure 2A was prepared using PyMOL Molecular Graphics System, v0.99 (Schrödinger).
sufficient deuteration. Values of pD were measured by
determining pH using a glass pH electrode and then ap-
plying the equation pD = pHelectrode + 0.41.27 To initiate
reactions, an aliquot of enzyme solution (never exceeding
5% of total reaction volume) was added to the reaction,
the quantity of enzyme (18 nM for TxAbf and N216W, 50
nM for G179F and L352M, and 2000 nM for all R69H-
containing mutants) being adjusted to suit the 8-12 h
anomeric proton (L-Araf), with X accounting for both
reaction outcomes (i.e., hydrolysis and transglycosyla-
tion), and identifying self-condensation. For all time-
course NMR kinetics, the absolute error mean value on X
ranged from 1 to 3%. The transglycosylation yield, Y(product)
,
was determined by relative integration of product ano-
meric proton signals, these being at 5.39, 5.32 and 5.27
ppm for XA3X, A3XX and A2XX respectively.29–34 Finally,
Y(product) (%) was plotted against X (%) to provide the
transfer rate (RT) of the donor substrate to a given prod-
uct (µmol of product/µmol of consumed substrate). Final-
ly, initial rates were derived from the linear regions in
which the variation of X was from 0-25% up to 0-40%,
with root mean square deviations between 0.96 and 0.99.
1
reaction timeframe. H NMR spectra were accumulated
semi-continuously, collecting a series of 5.52-min scans
1
(i.e., 32 scans) with 6 s delays between scans. Each H
NMR spectrum was acquired using an excitation flip angle
of 30° at a radiofrequency field of 29.7 kHz, and the refer-
ence residual water signal, calibrated at δ = 4.55 ppm at 45
°C,28 was pre-saturated during the repetition delay (with a
radiofrequency field of 21 Hz). The following acquisition
parameters were used: relaxation delay (6 s), dummy
scans (4). The actual duration of the reaction was de-
pendent on the enzyme used. Prior to enzyme addition,
an initial 1H NMR spectrum was acquired, which provided
starting point integrals that were subsequently corrected
using a dilution factor that accounted for the minor dilu-
tion provoked by the addition of the enzyme solution.
NMR kinetics analysis
Parameters to evaluate the effective transition from
hydrolase to transglycosidase
The activity ratio R (i.e., Vi(donor + acceptor)/Vi(donor)) putatively
reflects the T/H partition (with Vi defined as the donor
substrate consumption rate, which correlates with pNP
release rate). However, to compare different enzymes that
potentially display different activities, the transfer rate
(RT) of the donor substrate to a given product (µmol of
product/µmol of consumed substrate) was obtained by
plotting the yield of transglycosylation product(s) as a
function of donor conversion. Conveniently, being inde-
pendent of the duration of the reaction, RT provides a way
to compare different enzymes. Rho, which represents the
apparent synthesis/hydrolysis rate, is equal to |νT/νHII|;
with the apparent transglycosylation (νT, initial phase of
the plot) and the secondary hydrolysis (νHII, after maxi-
mum yield) rates being derived from the time-course
monitoring of transglycosylation product(s) yield.
Data were processed as previously described.23 Briefly, the
time-dependent evolution of donor (pNP-α-L-Araf) and
acceptor (xylotriose) concentrations was quantified by
integrating the relevant anomeric proton signals. Molar
balances, based on acceptor and donor consumption,
were used to convert the transglycosylation product sig-
nal integral into concentration, and to evaluate concen-
trations of hydrolysis products, respectively. From this,
the donor substrate conversion rate (X, in %) was calcu-
lated as the mean of three distinct signals; those of the
ortho and meta aromatic protons (pNP) and that of the
6
ACS Paragon Plus Environment

Page 7 of 17
ACS Catalysis
1
2
3
4
5
6
7
8
9
RESULTS
tion, V225M, present in clone 197. This was justified by
the fact that V225 is located far from the active site (ap-
proximately 18 Å), is not very conserved (~ 31%) in family
GH51 and is substituted at equivalent positions in other
GH51 members by a whole variety of amino acids (exclud-
ing cysteine, phenylalanine and tryptophan). Gratifyingly,
subsequent analysis of the purified R69H-L352M enzyme
confirmed the previous results obtained using the lysate
of clone 197 and revealed that activity was increased up to
260% in the presence of xylotriose (Figure 3), unlike wild-
type TxAbf, which is inhibited (e.g., 30% of residual activ-
ity in the presence of 30 mM xylotriose). Factually, rela-
tive activity ratios were multiplied up to 7.2-fold (for
R69H-L352M) compared to that of TxAbf, suggesting that
transglycosylation might be significantly reinforced. Fur-
thermore, the analysis of purified R69H demonstrated
that this mutation alone is adequate to radically alter this
ratio.
Detection of increased transfuranosylation amongst
randomly generated mutants
An earlier random mutagenesis experiment performed on
TxAbf identified L352M among several mutants that ap-
peared to improve transglycosylation.19 However, since
the transglycosylation yield of L352M was still rather
modest (approximately 18%), this mutant was used as a
starting point for a new round of engineering. In the ab-
sence of clear-cut knowledge that would allow rational
engineering, we chose to pursue the random approach
and thus created a new randomly-mutated library using
TxAbf-L352M as the parental DNA. Sequence analysis of
24 randomly selected clones from this library revealed a
mean mutation rate of 3.2 amino acid mutations/kb (Fig-
ure S2) and screening of 40,000 library clones on solid
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
medium containing the chromogenic substrate X-α-L-Araf
(Figure S3) allowed the selection of 368 pale blue colo-
nies. Referring to the results of previous work,20,35 it was
considered that the latter probably represented a subset
of hydrolytically-impaired mutants, and thus a group of
candidates for the identification of increased transglyco-
sylation activity. Therefore, further in vitro screening was
performed on these library clones in order to determine
Creation of acceptor subsite mutations G179F (+1) and
N216W (+2)
To pursue the goal of creating a finely-tuned trans-
furanosylase, two other positions were targeted for muta-
tion. The first of these, G179 located in subsite +1, was
pinpointed using an in silico screening approach (un-
published), which was used to identify point mutations
that could potentially increase hydrophobicity at positive
subsites and raise the affinity for the acceptor. When
compared to other amino acid positions, a large number
of substitutions at position G179 were predicted to modify
the binding affinity of TxAbf towards xylotriose (Figure
S7). Therefore, position G179 was selected for mutation.
Accordingly, the mutant G179F was found to display
stronger xylotriose-mediated inhibition, implying that
this mutant binds the acceptor molecule more tightly
than the wild-type enzyme (Figure S8), and significantly,
improved transglycosylation yield and transfer rate (Table
S2).
The second position targeted for mutation was N216, an
amino acid located in subsite +2 that displays only partial
conservation in family GH51 (25%, Table S3). Neverthe-
less, this residue stands out in TxAbf as a good target for
site-saturation mutagenesis, because inspection of the in
silico model suggested that N216 is well-placed to influ-
ence the positioning of acceptor molecules.36 Screening of
the subsequent mutant library revealed one clone con-
taining the mutation N216W, a substitution that is not
found in any family GH51 members. Interestingly, com-
pared to TxAbf this mutant was characterized by an al-
most total absence of inhibition in the presence of xy-
lotriose (Figure S8) and further analysis revealed that it
catalyzes the highly regioselective synthesis of (1ꢀ2)-
linked products (Figure S9 and Table S2). Finally, using
the mutants G179F and N216W, a combinatorial approach
was adopted, constructing triple mutants R69H-G179F-
their relative activities on pNP-α-L-Araf, in the presence
or absence of the xylotriosyl acceptor (Figure S4). Overall,
secondary screening revealed a ratio mean value (µ) for
the activity ratio R of 0.86, implying that most clones
displayed slight acceptor-mediated inhibition. Neverthe-
less, 4 clones lay outside of the 95% confidence interval (R
= µ ± 1.96 σ). One was the negative clone 123 (Figure S5),
while three were positive clones 165, 197 and 241, which
formed a separate population that displayed more than
200% activity in the presence of xylotriose. Further inves-
tigation of these clones, using HPAEC-PAD to monitor
the reactions confirmed that the activity ratios of clones
165, 197 and 241 were significantly above the mean value
(Figure S6). DNA sequencing of clones 165, 197 and 241
revealed that each of these contained the mutation R69H,
although the overall genotype was different in each case
(Table 1), thus providing a strong indication that R69H
substitution was responsible for the measured phenotype.
This conclusion was further supported by the examina-
tion of the molecular model of TxAbf (PDB ID: 2VRQ)
which revealed that R69 is located close (2.86 Å) to the
catalytic nucleophile (Figure 2A).
Deconvolution and recombination: synergistic ef-
fects between donor and acceptor mutations
Impact of R69H on the H/T ratio
To further investigate the consequences of the R69H
mutation, this point mutation was introduced into the
genetic backgrounds of both TxAbf and the mutant
L352M, thus creating mutant enzymes R69H and R69H-
L352M respectively, and eliminating the extra substitu-
L352M
and
R69H-N216W-L352M
(Figure
4).
7
ACS Paragon Plus Environment

ACS Catalysis
Page 8 of 17
1
2
3
4
5
6
7
8
Table 1. Steady-state kinetic parameters on pNP-α-L
-Araf for L352M, R69H and mutants thereofa
SAmax
kcat
(s-1)
KM
(mM)
0.25
8.98
0.09
1.23
kcat/ KM
(s-1.mM-1)
556.00
25.95
25.41
0.79
Nsb
(s-1.mM-1)
-
Enzyme
pHopt
(IU.mg-1)
145
TxAbf
L352M
R69H
R69H-L352M
R69H-G179F-L352M
R69H-N216W-L352M
5.8
6.5
6.5
6.5
6.5
6.5
139
233
2.287
0.972
0.535
0.580
243
2.383
1.012
0.556
0.603
-
0.049
0.031
0.021
0.034
0.69
0.48
0.78
1.21
9
a All assays were carried out at 45 °C in sodium phosphate buffer (pH 7, 50 mM) in triplicates. Relative errors were inferior to 10%.
b The Ns value corresponds to the non-specific constant attributed to autocondensation reaction and was obtained according the follow-
ing Michaelis-Menten modified model: SA = SAmax.[S]/(KM + [S]) + Ns.[S].
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Kinetic analyses of mutants combining R69H and
L352M
Steady-state kinetic parameters were determined for the
tants in the initial phase of the reaction, while plotting
synthesis as a function of time (Figure 5B) provides a
convenient method to evaluate the extent of secondary
hydrolysis. Accordingly, results confirmed that the paren-
tal mutant L352M procured a 2-fold increase in transgly-
cosylation yield compared to TxAbf (Table S2) and dis-
played a significantly reduced level of secondary hydroly-
sis (Figure 5B). In the case of the other single mutant,
R69H, the yield of A2XX reached 31% (Table 2), but strong
secondary hydrolysis at the end of the reaction led to the
degradation of the product. Nevertheless, overall R69H
displayed a 3.6-fold increase in the apparent synthe-
sis/hydrolysis rate (Rho) of A2XX (Table S5). Moreover,
this single mutation led to a 4.3-fold increase in transfer
rate (RT) for A2XX and a global RT of 99% (XA3X + A3XX +
A2XX + XA2X) in the initial phase of the reaction (Table
S6), demonstrating that in this phase R69H confers
strong transglycosylase character to TxAbf. In compari-
son, the transglycosylation yields procured by the double
mutant R69H-L352M, were slightly lower (27, 22 and 20%,
for A2XX, A3XX and XA3X respectively). However, the
contribution of L352M was reflected in the higher RT val-
ue (for A2XX) and slower secondary hydrolysis. Overall,
the combination of R69H and L352M produced a Rho
value of 55, which constitutes a 26.9-fold increase com-
pared to TxAbf.
hydrolysis of pNP-α-L-Araf catalyzed by the different
mutants bearing the substitutions R69H and L352M (Ta-
ble 1). Regarding the single mutants (i.e., those containing
only R69H or L352M), both displayed significantly modi-
fied behavior. In the case of L352M, a 37-fold increase of
the KM value led to a 22-fold reduction in the kcat/KM val-
ue, indicative of a modification in the glycosylation step.
The behavior of the R69H mutant was radically different
to that of TxAbf, since the reaction was described by a
biphasic profile (Figure S10). At pNP-α-L-Araf concentra-
tions below 5 mM, the reaction was described by the clas-
sical Michaelis-Menten model. However, at higher sub-
strate concentrations a linear relationship was observed
and overall the rate of the reaction was severely reduced.
Regarding the double mutant R69H-L352M and the triple
mutants R69H-G179F-L352M and R69H-N216W-L352M,
their behavior was highly similar to that of the single
mutant R69H. All displayed severely reduced specific
activities and biphasic reaction profiles.
Measurement of the transglycosylation activity of the
combinatorial mutants
Contrary to TxAbf, mutants R69H-L352M, R69H-G179F-
Analysis of the triple mutant, R69H-G179F-L352M, re-
vealed that the presence of the mutation G179F led to
postponed secondary hydrolysis, which had a beneficial
effect on the maximum yield of the principal product
A2XX (49%) and on the yields of A3XX and XA3X (17 and
35% respectively). In the case of R69H-N216W-L352M,
high regioselectivity (1→2 linkage towards the non-
reducing unit of xylotriose) and the alleviation of inhibi-
tion conferred by the N216W led to a dramatic increase in
maximum yield (up to 79% compared to the approximate
equimolar mixture of the different regioisomers obtained
with the wild-type, Table 2) and RT, which reached 90%
for A2XX. This is equivalent to a 6.9-fold increase of RT
compared to TxAbf (Table S6). However, inevitably the
generation of a high concentration of the transglycosyla-
tion product in the reaction mixture also led to strong
secondary hydrolysis once the donor substrate had been
consumed. Compared to R69H-L352M, this is consistent
with a lower Rho value, which was nevertheless 7-fold
higher than that measured for TxAbf (Table S5). In a pre-
L352M and R69H-N216W-L352M revealed activation in
presence of xylotriose (Figure 3), fitting to equation (6)
xylotriose
and providing apparent KM
values of 3.5, 5.9 and
45.5 mM, respectively. This resulted for example in a 17.1-
fold higher activity for R69H-N216W-L352M compared to
TxAbf in presence of 30 mM xylotriose (Table S4). Moreo-
ver, when xylotriose concentration was 50 mM, the rela-
tive activity of R69H-N216W-L352M attained 700% with-
out reaching a plateau-phase. To pursue the characteriza-
tion of the transfer ability of the different mutants, the
synthesis of arabinoxylo-oligosaccharides (AXOS, see
Fauré et al.37 for a comprehensive description of AXOS
nomenclature. Briefly, an unsubtituted
D-Xylp is desig-
nated X, while α-
L
-Araf-(1ꢀn)-β-D-Xylp segment is named
An with n corresponding to the O-position of the
D-Xylp
1
unit involved in the linkage) was monitored by H NMR
using pNP-α-L-Araf (15 mM) as donor and xylotriose (30
mM) as acceptor and the results were displayed in two
formats. The conversion-dependent plot (Figure 5A) con-
stitutes the best way to compare the performance of mu-
8
ACS Paragon Plus Environment

Page 9 of 17
ACS Catalysis
1
2
vious study23 it was revealed how a pH-dependent inhibi-
ty of the different datasets was poor, mainly because of
the two lowest LG pKa points. In this respect, it is perti-
nent to note that the electron-withdrawing character of
the LG might not be the only parameter influencing the
glycosylation step. While diffusion-related phenomena
are unlikely to exert any influence (~ 105 s-1.mM-1 >>
kcat/KM),40 steric hindrances may well affect the efficiency
of this catalytic step. It is noteworthy that similar
Brønsted profiles have been observed for other glyco-
sidases40–42 and that there is no rational or objective rea-
son to consider spurious data points such as those ob-
tained for NCP or 3,4-dNP.
3
4
5
6
7
8
9
tion of TxAbf could be used to block the degradation of
transglycosylation products. Therefore, in the present
work this strategy was successfully employed in combina-
tion with the mutant R69H-N216W-L352M. Accordingly,
by shifting the pH of the reaction from 7.0 to 5.8, it was
possible to obtain A2XX in yields > 70%, without the in-
convenience caused by secondary hydrolysis (Figure 5 and
Table 2). It is noteworthy, that the change in pH did not
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
affect the transfer rate, indicating that the H/T partition is
not pH-dependent for the mutant R69H-N216W-L352M.
Moreover, a control experiment performed in H2O/D2O
(9:1, v/v) mixture revealed that the substitution of H2O by
D2O did not affect the H/T partition, since at pH 7.0 and
Probing solvent kinetic isotope effects and the role of
water
5.8 equivalent transfer rates and yields were obtained. It is
also worth mentioning that the addition of 10% acetoni-
trile (v/v) in the reaction medium also allowed the sup-
pression of secondary hydrolysis (Figure 5), while procur-
ing a 10% increase in the final yield (Table 2) without
At the deglycosylation step, hydrolysis occurs when water
acts as an acceptor, launching a nucleophilic attack on
the anomeric carbon of the glycosyl-enzyme (Figures 1
and 2B). This catalytic sub-step is concomitant with a
second sub-step that involves covalent bond breaking and
thus the decomposition of the β–L-Araf-TxAbf intermedi-
ate. Accounting for the fact that TxAbf and mutants
thereof are deficient in the deglycosylation step and that
the reactivity of NCP-α-L-Araf is sufficient to correlate
apparent reaction rate to this step, it was assumed that
the solvent kinetic isotope effects (sKIE) approach could
be used to dissect deglycosylation, providing details on
the water-mediated sub-step. To achieve this, reactions
were performed in which H2O was replaced to different
extents by D2O (Figure S11). Overall, no major changes in
relative sKIE were observed, except for L352M, which was
less sensitive to D2O/H2O exchange than TxAbf. This
indicates that in all cases, water nucleophilic attack per se
is not a critical sub-step in this enzyme system.
changes in the initial H/T partition.
Probing the mechanisms of mutant enzymes
Hammett-Brønsted analysis: Transition state struc-
ture and rate-limiting step determination
To investigate how mutations in subsite -1 (i.e., L352M
and R69H), affect the catalytic mechanism, notably in the
switch from ground to the first transition state, a Ham-
mett-Brønsted analysis was performed using a series of
aryl α-L-arabinofuranosides to probe leaving group (LG)
effects (Table S7). Plotting log(kcat/KM) as a function of
the pKa of the LG (Figure 6A) revealed that while the
glycosylation step in reactions catalyzed by TxAbf is rela-
tively insensitive to LG pKa changes (i.e., βLG1 = -0.35 or -
0.28), the mutants R69H, L352M and R69H-L352M were
1.3 to 2.1-fold more sensitive (Table S8), indicating that
these mutants are characterized by modifications in the
transition state at the glycosylation step. Moreover, since
the catalytic constant (kcat = k2.k3/(k2+k3)) reflects the rate-
limiting step, a plot of log(kcat) as a function of LG pKa
(Figure 6B) was used to reveal whether the LG (released
during glycosylation) is involved in this step or not, with
the latter outcome implying that deglycosylation is the
rate-limiting step. This investigation demonstrated that
for most of the mutants and parental TxAbf log(kcat) was
rather insensitive to LG pKa changes, with values of βLG2
being close to zero for a wide range of pKa values up to
pKa = 9.99. Therefore, it was concluded that reactions
catalyzed by TxAbf, R69H and R69H-L352M are charac-
terized by a rate-limiting deglycosylation (k3) step, which
is consistent with previous results obtained for other
Abfs.38,39 The only exception to this was L352M, which
displayed a slight dependence of kcat on LG pKa (βLG2 = -
0.22). Taken together with the previous observation, this
result most likely indicates that this mutation strongly
affects the glycosylation step, as suggested by the increase
in the KM value. It is noteworthy that when log(kcat/KM)
was plotted as a function of LG pKa evolution, the lineari-
Figure 3. Analysis of the impact of xylotriose on global activity (pNP
release) for TxAbf, clone 197, R69H, R69H-L352M, R69H-G179F-
L352M and R69H-N216W-L352M. For the latter, an additional xy-
lotriose concentration was tested (50 mM) because the plateau-
phase was not reached. Assays were carried out in duplicate at 45 °C
in sodium phosphate buffer (50 mM, pH 7) with pNP-α-L-Araf as
donor substrate (5 mM). Apparent Ki and Ki’ values for xylotriose and
TxAbf are 15.0 and 15.6 mM. For R69H, R69H-L352M, R69H-G179F-
xylotriose
L352M and R69H-N216W-L352M mutants, apparent KM
values
for the acceptor are 8.6, 3.5, 5.9 and 45.5 mM.
9
ACS Paragon Plus Environment

ACS Catalysis
Page 10 of 17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Location of the mutations in donor subsite -1 (L352M and R69H) and acceptor subsites +1 (G179F) or +2 (N216W). (A) 3D-view of co-
crystallized TxAbf-XA3XX (PDB ID: 2VRQ)36 X-ray structure in which mutations were introduced via PyMOL Molecular Graphics System, v0.99
(Schrödinger). One can observe a steric clash between N216W and +2
D-Xylp unit, explaining (1ꢀ3) linkage deletion by N216W mutant and
putative stacking interactions between G179F and +1 -Xylp unit. (B) 2D-schematic representation of TxAbf active site.
D
Nevertheless, an absolute sKIE effect was observed (80%
residual activity), implying that H2O to D2O exchange
does affect global catalytic efficiency, probably through
modified protein dynamics or solvation,43 as suggested by
the exponential-like curve.44
using in vitro molecular evolution techniques,35,47,48 which
have provided some interesting results. Nevertheless,
these studies have fallen short of a comprehensive eluci-
dation of the mechanisms that have driven the observed
GH to TG transition.
In the area of furanose chemistry, little attention has been
focused on the development of furanose-acting transgly-
cosylases18,49,50 despite the fact that the synthesis of
furanose-based compounds figures among the more trick-
ier challenges in synthetic chemistry. Moreover, to our
knowledge the glycosynthase approach has not yet been
successfully applied to furanosidases, probably because of
the difficulty to obtain stable, but sufficiently reactive
furanoside donor substrates. For this reason and because
of our extensive knowledge of TxAbf, a rather robust,
Investigating the ionization state of catalytic resi-
dues
The pH-dependency of kcat/KM was investigated in order
to access the apparent pKa values of the two catalytic
residues of the apo-enzyme (Figure S12). The presence of
R69H and L352M in mutants was correlated with signifi-
cant shifts (0.7 to 1.2 units) in the apparent pKa value of
the nucleophile E298, while the ionization state of the
acid/base E176 remained globally unaffected (Table S9).
Nevertheless, when R69H was present as a single muta-
tion, the basic limb of the pH profile displayed an unusual
tail, which is characteristic of major ionization state mod-
ifications.26
well-characterized α-L-arabinofuranosidase from Thermo-
bacillus xylanyliticus that already displays some ability to
catalyze glycosynthetic reactions, we set out to create a
bona fide transarabinofuranosylase using a combination
of protein engineering strategies: random, in silico-driven
and site-saturation mutagenesis.
DISCUSSION
R69, a conserved residue in clan GH-A and a possible
H/T modulator
Thus far, probably the most reproducible strategy to con-
vert hydrolytic GHs into artificial TGs is the glycosynthase
approach, which calls for the use of highly reactive glyco-
side donors and hydrolytically-impotent (i.e., mutation of
the catalytic nucleophile) mutant enzymes.15,16 Beyond
this, no rational strategies for the conversion of GHs into
‘true’ TGs are currently known. This is primarily due to
the fact that it has proved very difficult to understand
how natural evolutionary processes have furnished for
example both transglycosylating and hydrolytic XTHs, or
trans-sialidase/sialidase pairs, no doubt because this in-
volves very subtle modifications of the molecular deter-
minants, which thus far have not been unequivocally
identified.13,45,46 In the absence of clear protein engineer-
ing guidelines, researchers have tackled the challenge
Screening of the randomly-mutated library created in this
study using L352M as the parental template was per-
formed using a previously described strategy, which iden-
tifies hydrolytically-impaired mutants. Although rather
simplistic, this approach is straightforward and has al-
ready proved to be efficient for the engineering of
transglycosylase activity into a GH1 β-glycosidase.35 In a
second step, the activity of the selected mutants was
measured in the presence of xylotriose, which can act as
an acceptor, but also as an inhibitor for TxAbf (Figure 3).
This provided the means to select mutants that were
either no longer inhibited by xylotriose and/or that dis-
10
ACS Paragon Plus Environment

Page 11 of 17
ACS Catalysis
1
2
3
4
5
6
7
8
9
played improved transglycosylation, taking into account
alteration of the efficiency of the catalytic acid/base. In-
terestingly, according to our data (not shown) when Y242,
a residue that also figures among the eight key residues of
GH-A (Figure 2), is substituted by phenylalanine a pheno-
the fact that in the latter case, the rate enhancements
observed in the presence of xylotriose were underestima-
tions of the actual transglycosylation improvements, since
the number of available active sites (i.e., uninhibited frac-
tion) is inferior to that in the acceptor-free reaction.
Overall, the two-stage screening strategy pinpointed the
mutation R69H. Interestingly, R69 is totally conserved
among GH51 Abfs (Table S3) and is amongst the eight key
residues that define clan GH-A,51 which is composed of 19
GH families. Beyond sequence conservation, the spatial
location of R69 is also significant, since its side chain
points towards the nucleophile residue, and lies within H-
bonding distance of the latter (Figure S13). Consequently,
it is plausible that the substitution of the equivalent argi-
nine in any member of GH-A could lead to a similar mod-
type (i.e., decreased H/T ratio) similar to that of R69H is
observed. This is unsurprising, since Y242 also interacts
with E298 and probably contributes to the definition of
its operational charge.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
It is now well-recognized that a species having oxocarbe-
nium ion-like character is formed during TS of deglyco-
sylation and that, at least in the case of retaining β-
glycosidases, a partial positive charge is developed on the
anomeric carbon,54–56 the latter feature being important
for catalysis since the incoming acceptor reacts with this
species.45,57 Therefore, the development of a charge of
appropriate magnitude is critical and no doubt facilitates
water-mediated deglycosylation.
ification of the
H/T partition and thus a similar pheno-
type. Indeed, this appears to be the case, since mutation
of R75 in the β-glucosidase from Thermus thermophilus
(GH-1) has been shown to procure a transglycosylase
phenotype.46
To better understand how the mutation of R69 favors
transglycosylation, it is important to recall that the R69H
mutation led to a drastic decrease in kcat and (consequent-
ly) kcat/KM, reflecting both slower deglycosylation and
glycosylation steps respectively. Moreover, the KM value
was lower, suggesting that the covalent glycosyl-enzyme
intermediate was subject to accumulation. Taken togeth-
er, these results are consistent with increased energy
barriers for the transition states (TS) that are involved in
each of the steps. One explanation for these effects is the
loss of a hydrogen bond linking the guanidinium group
(pKa ~ 12.5) of R69 to the carboxyl oxygen of the catalytic
nucleophile E298. One would expect this modification to
alter the electron displacement system (Figure 2B) and, in
particular, affect the ionization state of E298, which is
indeed the case, since the pKa of E298 in the mutant
R69H was increased.
It is known that along the double-displacement mecha-
nism, pKa cycling occurs,52 with the nucleophile (E298)
being negatively charged at the beginning of the reaction
and the acid/base (E176) protonated. In TxAbf, this is
possible because the acid/base catalyst (E176) displays an
apparent pKa ≈7.5, which is far higher than the pKa value
of a free glutamic acid (~ 4.5). This high pKa is probably
generated by charge-like repulsion between E176 and its
close nucleophilic neighbor, E298, such interactions being
necessary to avoid unfavorable Coulombic energetics.52
Regarding the nucleophile catalyst, the study of the GH1
β-glycosidase from Spodoptera frugiperda has revealed
that the negative ionization state of its nucleophile cata-
lyst, E399, is stabilized through hydrogen bonding to Y331
and R97, which significantly are equivalent to Y242 and
R69 in TxAbf.53 Overall, these observations support the
Figure 5. A2XX major transglycosylation product evolution as a
function of (A) donor conversion or (B) time for TxAbf (pD 5.8),
clone 197, L352M, R69H, R69H-L352M, R69H-G179F-L352M and
R69H-N216W-L352M. All Assays were carried out at 45 °C in sodium
phosphate buffer (25 mM, pD 7.0) or, when stated, in sodium acetate
buffer (25 mM, pD 5.8), with pNP-α-L-Araf as donor substrate (15
hypothesis that the altered H/T partition that characteriz-
es R69H is the result of modifications to the nucleophile’s
ability to attack the anomeric carbon, rather than a direct
mM) and xylotriose as acceptor (30 mM), using 2 µM of enzyme,
except for TxAbf and L352M or which 18 and 50 nM were employed,
respectively.
11
ACS Paragon Plus Environment

ACS Catalysis
Page 12 of 17
1
2
3
4
5
6
7
8
9
Table 2. Transglycosylation yields (%) for reactions catalyzed by TxAbf and R69H derivativesa
(L-Araf) +
Enzyme
XA3Xb
A3XXb
A2XXb
Auto 1c Auto 2c
XA2Xb
3.3
TxAbf (pD 5.8)
R69H (pD 7.0)
R69H-L352M (pD 7.0)
R69H-G179F-L352M (pD 7.0)
R69H-N216W-L352M (pD 7.0)
R69H-N216W-L352M (pD 5.8)
R69H-N216W-L352M (H2O-pH 7.0)
R69H-N216W-L352M (H2O-pH 5.8)
R69H-N216W-L352M (10% CD3CN-pD 7.0)
7.5
23.2
20.4
35.3
4.6
0.7
0.9
1.4
1.7
11.8
25.2
22.2
17.3
7.7
1.2
8.6
1.4
2
10.5
30.6
27.3
48.8
78.6
71.0
74.6
71.9
80.1
6.7
3.9
4.7
2.1
3.0
2.3
2.4
2.4
2.6
3.2
4.9
4.6
8.4
8.0
5.5
6.3
5.4
10.5
2.2
2.2
1.6
7.7
8.1
7.0
8.1
7.8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a Maximum yields can be reached at different times explaining the fact that the sum of maximum yields can be superior to 100%.
b The structures of the transglycosylation products were determined by NMR and were consistent with those previously reported (Figure S1).29–34
c Auto 1 and auto 2 correspond to two regioisomers of self-condensation products.18
In this case, although water activation is a required fea-
ture, it is likely that there is no need for any energetic
contribution from the water molecule itself. Indeed, this
is consistent with the conclusions of recent in silico anal-
yses performed on other GHs that indicate that the degly-
cosylation step is highly dissociative, meaning that the
covalent bond that links the enzyme nucleophile to the
anomeric carbon of the sugar moiety is almost broken
when it is replaced by a bond with water.58 In the light of
these mechanistic details, it appears probable that the
mutation R69H provokes a charge modification that im-
pacts on TS, reducing the efficiency of water-mediated
deglycosylation.
phile, which provides a strong ‘nucleophilic push’ towards
the anomeric carbon. In this respect, the more negative
βLG1 values measured for the mutants containing R69H
and/or L352M provide further evidence for modifications
of charge distribution within the TS, along the axis [nu-
cleophile E298 ↔ anomeric carbon ↔ interglycosidic
oxygen]. Overall, the kinetics and sKIE data presented in
this study point to the fact that the mutations R69H and
L352M, disrupt both covalent glycosyl-enzyme bond for-
mation and decomposition, and that this is caused by
changes in the distribution of electronic charge.
Key factors that transform GHs into TGs
The Hammett-Brønsted analysis revealed that TxAbf and
most of the mutants are characterized by a rate-limiting
deglycosylation step for a large range of substrates, an
observation that is consistent with data collected for oth-
er Abfs,38,39 but different to those collected for glycopyra-
nosidases.40–42,59,60 For the latter, the Brønsted plot of
log(kcat) is usually characterized by a concave breakdown
for LGs whose pKa values are above 8, this transition be-
ing indicative of a change in the rate-limiting step (from
deglycosylation to glycosylation). Accordingly, for TxAbf,
and more generally furanoside hydrolases, it is possible
that this difference is linked to the ring strain of
furanoside substrates,61 which makes the decomposition
of the glycosyl-enzyme complex (i.e., deglycosylation)
more energy demanding than its creation (i.e., glycosyla-
tion).
Taken together, the fact that deglycosylation is rate-
limiting in TxAbf and mutants thereof, irrespective of the
LG, the observation that kcat is strongly decreased in
R69H-containing mutant, and the absence of relative
changes in sKIE all argue in favor of a highly impaired
covalent bond-breaking (deglycosylation) sub-step in the
R69H-containing mutants. However, if one also considers
the rather low dependency of kcat/KM on LG pKa values, it
is also possible to deduce that only a small negative
charge is developed on the interglycosidic oxygen at TS of
the glycosylation step, which implies that either acid/base
protonation is strong enough to compensate for the elec-
tronic character of the LG,62,63 or that during glycosylation
the major contribution comes from the catalytic nucleo-
Donor subsite mutations undermine enzyme activity,
but are necessary to counter water activity
From a conceptual standpoint it is easy to imagine how
mutations in the acceptor subsite(s) of GHs could pro-
mote the acquisition of transglycosylation ability, by in-
creasing the recognition of the incoming sugar moiety.
However, it is much less obvious to casually rationalize
the postulate that donor subsite mutations are a prerequi-
site for the conversion of a GH into a TG, especially in the
light of our data that clearly demonstrate that such muta-
tions severely reduce enzyme activity. Nevertheless, the
mutation of donor subsite determinants, especially highly
conserved residues, has recently been proposed as a semi-
rational strategy to improve the transglycosylation ability
of several GHs.46,64 To better understand why this might
be, it is pertinent to recall that GHs generally operate in
aqueous environments, dominated by the presence of
water (at a concentration of 55 M). Therefore, it is clear
that it is imperative for TGs to somehow avoid water-
mediated deglycosylation, with the increased lifetime of
glycosyl-enzyme intermediates being an outward sign of
this.65 In this study, we have shown that although R69H
drastically reduced activity, it had a very positive impact
on the
H/T ratio. Transglycosylation was clearly favored
and water-mediated deglycosylation was reduced, thus
diminishing hydrolysis (Figure S14). However, R69H did
not eliminate secondary hydrolysis, a reaction that pro-
cures products that can no longer act as substrates for
further transglycosylation (Figure 1). To achieve this, it
12
ACS Paragon Plus Environment

Page 13 of 17
ACS Catalysis
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Figure 6. Brønsted plots of the Hammett relationship for the hydrolysis of aryl α-
R69H (dot) and R69H-L352M (dash-dot-dot). (A) log(kcat/KM) and (B) log(kcat) are plotted as a function of the pKa of the LG for a series of LG-α-
-Araf. All the kinetic parameters (Table S7) were derived from triplicate experiments. In Figure 6A the lowest LG pKa data point was omitted
L-arabinofuranosides catalyzed by TxAbf (solid), L352M (dash),
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
L
when establishing the linear data fit (Table S8).
was necessary to combine R69H with a second subsite -1
mutation, L352M. Clearly, the introduction of L352M
affected the glycosylation step, a fact revealed by the
marked sensitivity of L352M-containing mutants to LG
pKa variations (βLG1) and the very high KM values meas-
display without acceptor inhibition, which restrains the
apparent acceptor-mediated activity enhancement.
It is noteworthy that the overall transglycosylation yields
of the single mutants G179F and N216W remained low,
which clearly demonstrates that the modification of ac-
ceptor subsites alone is insufficient to confer good
transglycosylase activity to TxAbf. However, when the
beneficial effects of these mutations were combined with
the crippling effects of R69H-L352M, it is significant that
almost perfect transfuranosylases were obtained (Figure
7).
ured on LG-α-L-Araf. In this respect, it is noteworthy that,
compared to pNP (pKa = 7.18), the xylotriosyl moiety of
the transglycosylation products displays much lower leav-
ing group ability (pKa of glycosidic hydroxyl ~ 12), which
might explain the reduced ability of L352M-containing
mutants to hydrolyze the transglycosylation products.
Acceptor subsite modifications can produce different
effects
Towards an understanding of the evolutionary emer-
gence of TGs
In this work, two acceptor subsite mutations contributed
to the successful creation of transarabinofuranosylases.
The first one G179F, located in subsite +1, was selected
using an in silico approach that employed an algorithm
that identifies single mutations that can procure in-
creased binding for a given acceptor. Whereas the intro-
duction of G179F into the double mutant R69H-L352M
did not affect the transfer rate, it clearly altered secondary
hydrolysis, although the mechanism has yet to be clari-
fied. Nevertheless, it is highly probable that the G179F
procures increased hydrophobicity in subsite +1 and a
stacking interaction with pNP, thus specifically increasing
Numerous structural studies, including the one per-
formed on TxAbf,36 have revealed that exo-acting GHs are
characterized by extensive enzyme-substrate interaction
networks in the donor subsite. Consequently, reactions
catalyzed by exo-acting GHs are characterized by strong
stabilization of the TS (i.e., lowering of energy barriers),
which usually results in quite high catalytic efficiencies
(kcat/KM values), linked to a fast glycosylation step, and
reasonably efficient (albeit slower) water-mediated degly-
cosylation. The efficiency of the latter step is underpinned
by the fact that the vast majority of the energy required to
overcome the TS is procured by donor subsite interac-
tions (Figure S14). This last point is vital to understand
the preponderant water-mediated deglycosylation in
hydrolytic GHs operating in aqueous media. Nevertheless,
despite the overwhelming presence of water, exo-acting
GHs often display some ability to perform transglycosyla-
tion (up to 12% yields for TxAbf regarding the main prod-
uct and sometimes more for other exo-GHs),66 which is
consistent with the fact that acceptor sugars can lower
the TS2 energy level at deglycosylation step (Figures S14
and S15). Therefore, acceptor subsite mutations that im-
prove enzyme-(acceptor) substrate interactions can fur-
the affinity of the mutant enzyme for pNP-α-L-Araf. The
implication of this modification is that the donor is prob-
ably better recognized compared to the transglycosylation
products and thus secondary hydrolysis is postponed.
Regarding N216W, when present as a single mutation this
subsite +2 mutation only modestly increases the transfer
rate and diverts activity away from the formation of (1ꢀ3)
linkages, while alleviating xylotriose inhibition. Therefore,
it is probable that N216W affects both acceptor position-
ing and affinity. Additionally, as suggested by its non-
saturating acceptor dose-response relationship (Figure 3),
the triple mutant R69H-N216W-L352M may reveal the
acceptor-activation behavior that R69H-L352M should
ther exacerbate the
T/H partition and thus increase
transglycosylation yields. When coupled to other muta-
13
ACS Paragon Plus Environment

ACS Catalysis
Page 14 of 17
1
2
3
4
5
6
7
8
9
tions that deprive water-mediated deglycosylation of
much of the energy that is required to overcome the TS
(i.e., ones that perturb donor subsite interactions), potent
TGs are produced.
In a recent review, we detailed the ways in which TGs and
GHs differ and provided some insight into how we believe
Nature has provided these two different activities using a
common catalytic mechanism and a whole range of pro-
tein scaffolds.9 Overall, it appears likely that natural evo-
lutionary processes have resulted in a trade-off that inevi-
tably links good transglycosylation to poor transition
states and low overall activity, this being the only efficient
way to counter the overwhelming competition from water
molecules. Consistent with this, we observe that the over-
all TGs that have so far been thoroughly characterized can
indeed be considered catalytically-impaired enzymes. 67–69
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 7. The iterative pathway that allowed the conversion of
TxAbf into non-Leloir transarabinofuranosylases. A subtle combina-
tion of random, semi-rational and in silico approaches was used to
reach this goal.
CONCLUSION AND PERSPECTIVES
We believe that the enzymes described in this work can
be legitimately designated as non-Leloir transfuranosylas-
es (up to 80% yield of the main product, slight and con-
trollable residual hydrolysis). Therefore, to our
knowledge, this is the first example of the in vitro creation
of such enzymes. In this respect, the work described here-
in should have far-reaching consequences, notably be-
cause it provides precious tools for furanose chemistry
and a number of critical elements for the elaboration of a
more generic protein engineering strategy that can be
applied to GHs. Concerning the perspectives for furanose
ASSOCIATED CONTENT
Experimental information pertaining to the synthesis of aryl
α-L-arabinofuranoside donors and in silico calculations, as
well as further details on random mutagenesis analysis,
screening methods, structure/function relationships, conser-
vation of targeted residues, regioselectivity and biochemical
data is supplied free of charge at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
michael.odonohue@insa-toulouse.fr
regis.faure@insa-toulouse.fr
chemistry, it is noteworthy that TxAbf and other
arabinofuranosidases display the ability to transfer
L-
D-
D-
galactofuranosyl
moieties
onto
acceptors.
Author Contributions
Galactofuranose is a rare sugar of considerable im-
portance, because it is widespread in human microbial
pathogens.70 Therefore, the enzymes developed in this
work could be useful for the synthesis of medically-
relevant compounds.
The manuscript was written through contributions of all
authors. All authors have given approval to the final version
of the manuscript.
Notes
Regarding the protein engineering strategy, it is notewor-
thy that random mutagenesis coupled to a quite simple
screening method appears to be suitable to isolate mu-
tants that are hydrolytically-crippled, although in the
future it may be even more efficient to directly mutate
residues that are likely to form hydrogen bonds with the
catalytic nucleophile in the substrate donor site (begin-
ning with homologs of TxAbf R69) and/or highly con-
served around the donor subsite (-1). These could be
combined with direct mutation(s) that increase the affini-
ty or the positioning of the acceptor, predicting hot-spots
using in silico approaches such as the one described here-
in. Overall, we believe that the lessons learned in this
work provide the basis for a general GH engineering route
that should allow the successful improvement of transgly-
cosylase activity in other exo-acting GHs.
The authors declare no competing financial interest.
ACKNOWLEDGMENT
PhD fellowship of BB was supported by the Institut National
de la Recherche Agronomique (INRA) [CJS]. The NMR work
carried out in this work at the Laboratory for BioSystems &
Process Engineering (Toulouse, France) was performed with
the equipments of MetaToul (Metabolomics & Fluxomics
Facitilies, Toulouse, France, www.metatoul.fr). MetaToul is
part of the national infrastructure MetaboHUB (The French
National infrastructure for metabolomics and fluxomics,
www.metabohub.fr) and is supported by grants from the
Région Midi-Pyrénées, the European Regional Development
Fund, SICOVAL, IBiSa-France, the Centre National de la
Recherche Scientifique (CNRS) and INRA. We thank the
ICEO facility dedicated to enzyme screening and discovery,
and part of the Integrated Screening Platform of Toulouse
(PICT, IBiSA) for providing access to the TECAN liquid han-
dling equipment. Contributions from XB and AP were par-
ACS Paragon Plus Environment

Page 15 of 17
ACS Catalysis
1
2
3
4
5
tially supported by grant BIO2013-49022-C2-1-R from
MINECO, Spain.
(16)
(17)
Malet, C.; Planas, A. FEBS Lett. 1998, 440, 208–212.
Debeche, T.; Cummings, N.; Connerton, I.; Debeire,
P.; O’Donohue, M. J. Appl. Environ. Microbiol. 2000,
66, 1734–1736.
ABBREVIATIONS
(18)
(19)
(20)
Rémond, C.; Plantier-Royon, R.; Aubry, N.;
O’Donohue, M. J. Carbohydr. Res. 2005, 340, 637–
644.
Rémond, C.; Plantier-Royon, R.; Aubry, N.; Maes, E.;
Bliard, C.; O’Donohue, M. J. Carbohydr. Res. 2004,
339, 2019–2025.
Arab-Jaziri, F.; Bissaro, B.; Dion, M.; Saurel, O.;
Harrison, D.; Ferreira, F.; Milon, A.; Tellier, C., Fauré,
R.; O’Donohue, M. J. New Biotechnol. 2013, 30, 536–
544.
Arab-Jaziri, F.; Bissaro, B.; Tellier, C.; Dion, M.;
Fauré, R.; O’Donohue, M. J. Carbohydr. Res. 2015,
401, 64–72.
Kelly, M. A.; Sinnott, M. A.; Widdows, D. Carbohydr.
Res. 1988, 181, 262–266.
Bissaro, B.; Saurel, O.; Arab-Jaziri, F.; Saulnier, L.;
Milon, A.; Tenkanen, M.; Monsan, P.; O’Donohue,
M. J.; Fauré, R. Biochim. Biophys. Acta, Gen. Subj.
2014, 1840, 626–636.
Patrick, W. M.; Firth, A. E.; Blackburn, J. M. Protein
Eng. Des. Sel. 2003, 16, 451–457.
Joshi, M. D.; Sidhu, G.; Nielsen, J. E.; Brayer, G. D.;
Withers, S. G.; McIntosh, L. P. Biochemistry 2001, 40,
10115–10139.
Bicknell, R.; Knott-Hunziker, V.; Waley, S. G.
Biochem. J. 1983, 213, 61–66.
Glasoe, P. K.; Long, F. A. J. Phys. Chem. 1960, 64,
188–190.
Gottlieb, H. E.; Kotlyar, V.; Nudelman, A. J. Org.
Chem. 1997, 3263, 7512–7515.
Gruppen, H.; Hoffmann, R. A.; Kormelink, F. J.;
Voragen, A. G.; Kamerling, J. P.; Vliegenthart, J. F.
Carbohydr. Res. 1992, 233, 45–64.
Vietor, R. J.; Hoffmann, R. A.; Angelino, S. A. G. .;
Voragen, A. G. J.; Kamerling, J. P.; Vliegenthart, J. F.
G. Carbohydr. Res. 1994, 6215, 245–255.
Ferré, H.; Broberg, A.; Duus, J. O.; Thomsen, K. K.
Eur. J. Biochem. 2000, 267, 6633–6641.
Pastell, H.; Tuomainen, P.; Virkki, L.; Tenkanen, M.
Carbohydr. Res. 2008, 343, 3049–3057.
Pitkänen, L.; Tuomainen, P.; Virkki, L.; Tenkanen, M.
Int. J. Biol. Macromol. 2011, 49, 963–969.
Correia, M. A. S.; Mazumder, K.; Brás, J. L. A.;
Firbank, S. J.; Zhu, Y.; Lewis, R. J.; York, W. S.;
Fontes, C. M. G. A.; Gilbert, H. J. J. Biol. Chem. 2011,
286, 22510–22520.
Feng, H.-Y.; Drone, J.; Hoffmann, L.; Tran, V.; Tellier,
C.; Rabiller, C.; Dion, M. J. Biol. Chem. 2005, 280,
37088–37097.
Paës, G.; Skov, L. K.; O’Donohue, M. J.; Rémond, C.;
Kastrup, J. S.; Gajhede, M.; Mirza, O. Biochemistry
2008, 47, 7441–7451.
6
7
8
9
Abfs, α-
AXOS, arabinoxylo-oligosaccharides; A2XX, α-
β- -Xylp-(1ꢀ4)-β- -Xylp-(1ꢀ4)- -Xylp;
A3XX,
(1ꢀ3)-β- -Xylp-(1ꢀ4)-β- -Xylp-(1ꢀ4)- -Xylp; 3,4-dNP, 3,4-
dinitrophenol; GH, glycoside hydrolase; GT, glycosyltransfer-
ase; LG, leaving group; NCP, 2-nitro-4-chlorophenol; 3-NP, 3-
L
-arabinofuranosidases;
L
-Araf,
L
-arabinofuranosyl;
-Araf-(1ꢀ2)-
α- -Araf-
L
D
D
D
L
D
D
D
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
nitrophenol; Ph, phenol; pNP, 4-nitrophenol; pNP-α-
4-nitrophenyl α- -arabinofuranoside; r-GH, retaining GH;
sKIE, solvent kinetic isotope effects; TxAbf, α-
arabinofuranosidase from Thermobacillus xylanilyticus;
L-Araf,
L
L-
H/T,
(21)
hydrolysis/transglycosylation ratio; TG, non-Leloir transgly-
cosylase; TS, transition state; X, donor substrate conversion
rate;
arabinofuranoside; XA2X, β-
-Xylp-(1ꢀ4)-
-Xylp; XA3X, β-
β- -Xylp-(1ꢀ4)- -Xylp; XOS, xylo-oligosaccharides;
-xylopyranosyl; Y, yield.
X-α-
L
-
Araf,
5-bromo-4-chloro-3-indolyl
-Xylp-(1ꢀ4)-[α- -Araf-(1ꢀ2)]-β-
-Xylp-(1ꢀ4)-[α- -Araf-(1ꢀ3)]-
-Xylp,
α-L-
(22)
(23)
D
L
D
D
D
L
D
D
D
D
(24)
(25)
REFERENCES
(1)
(2)
Lowary, T. Curr. Opin. Chem. Biol. 2003, 7, 749–756.
Peltier, P.; Euzen, R.; Daniellou, R.; Nugier-Chauvin,
C.; Ferrières, V. Carbohydr. Res. 2008, 343, 1897–1923.
Fauré, R.; Saura-Valls, M.; Brumer, H.; Planas, A.;
Cottaz, S.; Driguez, H. J. Org. Chem. 2006, 71, 5151–
5161.
Champion, E.; André, I.; Moulis, C.; Boutet, J.;
Descroix, K.; Morel, S.; Monsan, P.; Mulard, L. A.;
Remaud-Siméon, M. J. Am. Chem. Soc. 2009, 131,
7379–7389.
(26)
(27)
(28)
(29)
(3)
(4)
(5)
(6)
Wang, L.-X.; Huang, W. Curr. Opin. Chem. Biol.
2009, 13, 592–600.
(30)
Wang, Z.; Chinoy, Z. S.; Ambre, S. G.; Peng, W.;
McBride, R.; de Vries, R. P.; Glushka, J.; Paulson, J.
C.; Boons, G.-J. Science 2013, 341, 379–383.
Koshland Jr., D. E. Biol. Rev. 1953, 28, 416–436.
Lombard, V.; Golaconda Ramulu, H.; Drula, E.;
Coutinho, P. M.; Henrissat, B. Nucleic Acids Res.
2014, 42, D490–D495.
(31)
(32)
(33)
(34)
(7)
(8)
(9)
(10)
(11)
Bissaro, B.; Monsan, P.; Fauré, R.; O’Donohue, M.
Biochem. J. 2015, 467, 17–35.
Street, I. P.; Kempton, J. B.; Withers, S. G.
Biochemistry 1992, 31, 9970–9978.
Ducros, M.-A.; Tarling, C. A.; Zechel, D. L.;
Brzozowski, A. M.; Frandsen, T. P.; Ossowski, I. Von;
Withers, S. G.; Davies, G. J. Chem. Biol. 2003, 10, 619–
628.
(35)
(36)
(37)
(12)
(13)
(14)
(15)
Zakariassen, H.; Hansen, M. C.; Jøranli, M.; Eijsink,
V. G. H.; Sørlie, M. Biochemistry 2011, 50, 5693–5703.
Eklöf, J. M.; Shojania, S.; Okon, M.; McIntosh, L. P.;
Brumer, H. J. Biol. Chem. 2013, 288, 15786–15799.
Saburi, W.; Kobayashi, M.; Mori, H.; Okuyama, M.;
Kimura, A. J. Biol. Chem. 2013, 288, 31670–31677.
Mackenzie, L. F.; Wang, Q.; Warren, R. A. J.;
Withers, S. G. J. Am. Chem. Soc. 1998, 120, 5583–5584.
Fauré, R.; Courtin, C. M.; Delcour, J. A.; Dumon, C.;
Faulds, C. B.; Fincher, G. B.; Fort, S.; Fry, S. C.; Halila,
S.; Kabel, M. A.; Pouvreau, L.; Quemener, B.; Rivet,
A.; Saulnier, L.; Schols, H. A.; Driguez, H.;
O’Donohue, M. J. Aust. J. Chem. 2009, 62, 533–537.
ACS Paragon Plus Environment

ACS Catalysis
Page 16 of 17
1
2
3
4
5
6
7
8
9
(38)
Shallom, D.; Belakhov, V.; Solomon, D.; Shoham, G.;
Baasov, T.; Shoham, Y. J. Biol. Chem. 2002, 277,
43667–43673.
Wan, C.; Chen, W.; Chen, C.; Chang, M.; Lo, L.; Li, Y.
Biochem. J. 2007, 401, 551–558.
Zechel, D. L.; Reid, S. P.; Stoll, D.; Nashiru, O.;
Warren, R. A. J.; Withers, S. G. Biochemistry 2003,
42, 7195–7204.
Bravman, T.; Zolotnitsky, G.; Belakhov, V.; Shoham,
G.; Henrissat, B.; Baasov, T.; Shoham, Y.
Biochemistry 2003, 42, 10528–10536.
Kempton, J. B.; Withers, S. G. Biochemistry 1992, 31,
9961–9969.
Combes, D.; Yoovidhya, T.; Girbal, E.; Willemot, R.-
M.; Monsan, P. Ann. N.Y. Acad. Sci. 1987, 501, 59–62.
Schowen, K. B.; Limbach, H. H.; Denisov, G. S.;
Schowen, R. L. Biochim. Biophys. Acta, Bioenergetics
2000, 1458, 43–62.
Pierdominici-Sottile, G.; Palma, J.; Roitberg, A. E.
Proteins 2013, 82, 424–435.
Teze, D.; Hendrickx, J.; Czjzek, M.; Ropartz, D.;
Sanejouand, Y.-H.; Tran, V.; Tellier, C.; Dion, M.
Protein Eng. Des. Sel. 2014, 27, 13–19.
Osanjo, G.; Dion, M.; Drone, J.; Solleux, C.; Tran, V.;
Rabiller, C.; Tellier, C. Biochemistry 2007, 46, 1022–
1033.
Placier, G.; Watzlawick, H.; Rabiller, C.; Mattes, R.
Appl. Environ. Microbiol. 2009, 75, 6312–6321.
Almendros, M.; Danalev, D.; François-Heude, M.;
Loyer, P.; Legentil, L.; Nugier-Chauvin, C.; Daniellou,
R.; Ferrières, V. Org. Biomol. Chem. 2011, 9, 8371–
8378.
Chlubnova, I.; Filipp, D.; Spiwok, V.; Dvorakova, H.;
Daniellou, R.; Nugier-Chauvin, C.; Kralova, B.;
Ferrières, V. Org. Biomol. Chem. 2010, 8, 2092–2102.
Durand, P.; Lehn, P.; Callebaut, I.; Fabrega, S.;
Henrissat, B.; Mornon, J. P. Glycobiology 1997, 7,
277–284.
McIntosh, L. P.; Hand, G.; Johnson, P. E.; Joshi, M.
D.; Ko, M.; Plesniak, L. A.; Ziser, L.; Wakarchuk, W.
W.; Withers, S. G. Biochemistry 1996, 35, 9958–9966.
Marana, S. R.; Mendonça, L. M. F.; Andrade, E. H. P.;
Terra, W. R.; Ferreira, C. Eur. J. Biochem. 2003, 270,
4866–4875.
(55)
(56)
(57)
(58)
(59)
(60)
Vocadlo, D. J.; Davies, G. J. Curr. Opin. Chem. Biol.
2008, 12, 539–555.
Davies, G. J.; Planas, A.; Rovira, C. Acc. Chem. Res.
2012, 45, 308–316.
(39)
(40)
Pierdominici-Sottile, G.; Horenstein, N. A.; Roitberg,
A. E. Biochemistry 2011, 50, 10150–10158.
Brás, N. F.; Ramos, M. J.; Fernandes, P. A. J. Mol.
Struct. THEOCHEM 2010, 946, 125–133.
Tull, D.; Withers, S. G. Biochemistry 1994, 33, 6363–
6370.
Vallmitjana, M.; Ferrer-Navarro, M.; Planell, R.; Abel,
M.; Ausín, C.; Querol, E.; Planas, A.; Pérez-Pons, J.-A.
Biochemistry 2001, 40, 5975–5982.
Taha, H. A.; Richards, M. R.; Lowary, T. L. Chem.
Rev. 2013, 113, 1851–1876.
Ly, H. D.; Withers, S. G. Annu. Rev. Biochem. 1999,
68, 487–522.
(41)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(42)
(43)
(44)
(61)
(62)
(63)
(64)
(65)
Kelly, M. A.; Sinnott, M. L.; Herrchen, M. Biochem. J.
1987, 245, 843–849.
(45)
(46)
Teze, D.; Daligault, F.; Ferrières, V.; Sanejouand, Y.-
H.; Tellier, C. Glycobiology 2015, 25, 420–427.
Piens, K.; Fauré, R.; Sundqvist, G.; Baumann, M. J.;
Saura-Valls, M.; Teeri, T. T.; Cottaz, S.; Planas, A.;
Driguez, H.; Brumer, H. J. Biol. Chem. 2008, 283,
21864–21872.
Dion, M.; Nisole, A.; Spangenberg, P.; André, C.;
Glottin-Fleury, A.; Mattes, R.; Tellier, C.; Rabiller, C.
Glycoconjugate J. 2001, 18, 215–223.
Potocki de Montalk, G.; Remaud-Siméon, M.;
Willemot, R. M.; Sarçabal, P.; Planchot, V.; Monsan,
P. FEBS Lett. 2000, 471, 219–223.
Baumann, M. J.; Eklöf, J. M.; Michel, G.; Kallas, A. M.;
Teeri, T. T.; Czjzek, M.; Brumer, H. Plant Cell 2007,
19, 1947–1963.
(47)
(66)
(67)
(68)
(69)
(70)
(48)
(49)
(50)
(51)
Ribeirāo, M.; Pereira-Chioccola, V. L.; Eichinger, D.;
Rodrigues, M. M.; Schenkman, S. Glycobiology 1997,
7, 1237–1246.
Pedersen, L. L.; Turco, S. J. Cell. Mol. Life Sci. 2003,
60, 259–266.
(52)
(53)
(54)
Notenboom, V.; Birsan, C.; Nitz, M.; Rose, D. R.;
Warren, R. A. J.; Withers, S. G. Nat. Struct. Biol.
1998, 5, 812–818.
ACS Paragon Plus Environment

Page 17 of 17
ACS Catalysis
1
2
3
4
Table of Contents artwork
5
6
7
8
9
10
11
12
13
14
15
16
17
Keywords: biocatalysis, transglycosylation, molecular evolution, glycoside hydrolase, enzyme mechanism, pentose-
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
based oligosaccharides
ACS Paragon Plus Environment