A 1H NMR study of the specificity of α-l- arabinofuranosidases on natural and unnatural substrates

Article abstract of DOI:10.1016/j.bbagen.2014.07.001

Background The detailed characterization of arabinoxylan-active enzymes, such as double-substituted xylan arabinofuranosidase activity, is still a challenging topic. Ad hoc chromogenic substrates are useful tools and can reveal subtle differences in enzymatic behavior. In this study, enzyme selectivity on natural substrates has been compared with enzyme selectivity towards aryl-glycosides. This has proven to be a suitable approach to understand how artificial substrates can be used to characterize arabinoxylan-active α-l-arabinofuranosidases (Abfs). Methods Real-time NMR using a range of artificial chromogenic, synthetic pseudo-natural and natural substrates was employed to determine the hydrolytic abilities and specificity of different Abfs. Results The way in which synthetic di-arabinofuranosylated substrates are hydrolyzed by Abfs mirrors the behavior of enzymes on natural arabinoxylo-oligosaccharide (AXOS). Family GH43 Abfs that are strictly specific for mono-substituted d-xylosyl moieties (AXH-m) do not hydrolyze synthetic di-arabinofuranosylated substrates, while those specific for di-substituted moieties (AXH-d) remove a single l-arabinofuranosyl (l-Araf) group. GH51 Abfs, which are supposedly AXH-m enzymes, can release l-Araf from disubstituted d-xylosyl moieties, when these are non-reducing terminal groups. Conclusions and general significance The present study reveals that although the activity of Abfs on artificial substrates can be quite different from that displayed on natural substrates, enzyme specificity is well conserved. This implies that carefully chosen artificial substrates bearing di-arabinofuranosyl d-xylosyl moieties are convenient tools to probe selectivity in new Abfs. Moreover, this study has further clarified the relative promiscuity of GH51 Abfs, which can apparently hydrolyze terminal disubstitutions in AXOS, albeit less efficiently than mono-substituted motifs.

Full text of DOI:10.1016/j.bbagen.2014.07.001

Biochimica et Biophysica Acta 1840 (2014) 3106–3114
Contents lists available at ScienceDirect
Biochimica et Biophysica Acta
journal homepage: www.elsevier.com/locate/bbagen
A 1H NMR study of the specificity of α-L-arabinofuranosidases on natural
and unnatural substrates
Vinciane Borsenberger a,b,c, Emmie Dornez d, Marie-Laure Desrousseaux a,b,c, Stéphane Massou a,b,c
,
Maija Tenkanen e, Christophe M. Courtin d, Claire Dumon a,b,c, Michael J. O'Donohue a,b,c, Régis Fauré a,b,c,
⁎
a
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 Food Chemistry and Biochemistry, KU Leuven, Kasteelpark Arenberg 20 bus 2463, B-3001 Leuven, Belgium
b
c
d
e
Department of Food and Environmental Sciences, Faculty of Agriculture and Forestry, University of Helsinki, P.O. Box 27, FI-00014 Helsinki, Finland
a r t i c l e i n f o
a b s t r a c t
Article history:
Background: The detailed characterization of arabinoxylan-active enzymes, such as double-substituted xylan
arabinofuranosidase activity, is still a challenging topic. Ad hoc chromogenic substrates are useful tools and
can reveal subtle differences in enzymatic behavior. In this study, enzyme selectivity on natural substrates has
been compared with enzyme selectivity towards aryl-glycosides. This has proven to be a suitable approach to un-
derstand how artificial substrates can be used to characterize arabinoxylan-active α-L-arabinofuranosidases
(Abfs).
Received 20 December 2013
Received in revised form 17 June 2014
Accepted 1 July 2014
Available online 10 July 2014
Keywords:
Methods: Real-time NMR using a range of artificial chromogenic, synthetic pseudo-natural and natural substrates
was employed to determine the hydrolytic abilities and specificity of different Abfs.
Arabinoxylan arabinofuranohydrolases
Di-arabinofuranosylated substrates
Enzymatic hydrolysis
Specificity
Screening
NMR
Results: The way in which synthetic di-arabinofuranosylated substrates are hydrolyzed by Abfs mirrors the
behavior of enzymes on natural arabinoxylo-oligosaccharide (AXOS). Family GH43 Abfs that are strictly specific
for mono-substituted D-xylosyl moieties (AXH-m) do not hydrolyze synthetic di-arabinofuranosylated sub-
strates, while those specific for di-substituted moieties (AXH-d) remove a single L-arabinofuranosyl (L-Araf)
group. GH51 Abfs, which are supposedly AXH-m enzymes, can release L-Araf from disubstituted D-xylosyl moie-
ties, when these are non-reducing terminal groups.
Conclusions and general significance: The present study reveals that although the activity of Abfs on artificial sub-
strates can be quite different from that displayed on natural substrates, enzyme specificity is well conserved. This
implies that carefully chosen artificial substrates bearing di-arabinofuranosyl D-xylosyl moieties are convenient
tools to probe selectivity in new Abfs. Moreover, this study has further clarified the relative promiscuity of
GH51 Abfs, which can apparently hydrolyze terminal disubstitutions in AXOS, albeit less efficiently than
mono-substituted motifs.
© 2014 Elsevier B.V. All rights reserved.
1. Background
(AXOS) are generally considered to be exo-enzymes that remove
non-reducing L-arabinofuranosyl (L-Araf) residues from xylan and
α-L-Arabinofuranosidases (Abfs — EC 3.2.1.55) involved in the de-
construction of arabinoxylans (AX) and arabinoxylo-oligosaccharides
xylo-oligosaccharidic backbone, being able to cleave α-1,2 and/or
α-1,3 linkages [1,2]. Currently AX-specific Abfs (i.e. arabinoxylan
arabinofuranohydrolases, designated AXH) are found in families GH2,
3, 43, 51, 54 and 62 within the CAZy database [3] and are named accord-
ing to their hydrolytic specificities (Fig. 1). Abfs that release L-Araf units
from mono-substituted main-chain D-xylopyranosyl (D-Xylp) motifs are
termed AXH-m, while those that release single L-Araf residues from
double-substituted main-chain D-Xylp motifs are termed AXH-d.
Moreover, when the exact bond specificity is known, this nomenclature
can be completed by a numeral. Hence, AXH-d3 is an Abf that cleaves
α-1,3 bonds that link L-Araf unit to a doubly-substituted D-Xylp residue.
AXH-m are the most common type of Abf and are widespread in
family GH51. Mostly, these enzymes have been reported to possess
Abbreviations: Abfs, α-L-arabinofuranosidases; AX, arabinoxylans; AXH, arabinoxylan
arabinofuranohydrolase; AXH-d3, double-substituted xylan α-1,3-L-arabinofuranosidase;
AXH-m, mono-substituted xylan α-L-arabinofuranosidase; AXOS, arabinoxylo-oligosac-
charides; Bz, benzoyl; Bn, benzyl; D-Xylp, D-xylopyranosyl; GH, glycoside hydrolase;
L-Ara, L-arabinose; L-Araf, L-arabinofuranosyl; 4NTC, 4-nitrochatecol; pNP, para-nitrophenol;
pNPA, para-nitrophenyl α-L-arabinofuranoside; SA, specific activity; TxAbf, α-L-
arabinofuranosidase from Thermobacillus xylanilyticus
Corresponding author at: Laboratoire d'Ingénierie des Systèmes Biologiques et des
Procédés, 135 Avenue de Rangueil, 31077 Toulouse cedex 4, France. Tel.: +33 5 6155
9488; fax: +33 5 6155 9400.
⁎
E-mail address: regis.faure@insa-toulouse.fr (R. Fauré).
http://dx.doi.org/10.1016/j.bbagen.2014.07.001
0304-4165/© 2014 Elsevier B.V. All rights reserved.

V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
3107
acetic acid[D4] used for the preparation of deuterated buffer were
purchased from Euriso-top (Saint-Aubin, France). DCM was distilled
from P2O5, prior to use. The evolution of synthetic reactions was moni-
tored using analytical thin-layer chromatography, employing silica gel
60 F254 precoated plates (E. Merck). Spots were visualized using ultra-
violet light (254 nm) and then by soaking in a 10% w/v orcinol solution
containing a mixture of sulfuric acid/ethanol/water (3:72.5:22.5 v/v/v)
followed by charring. Purification of reactions products was achieved
using silica gel Si 60 (40–63 μm) using a bench column, operating
with the indicated eluent, or using a Reveleris® flash chromatography
automated system (Grace, Epernon, France) equipped with prepacked
cartridges (Grace). Yields refer to chromatographically pure com-
pounds. NMR spectra were recorded at 298 K on a Bruker Avance II
500 spectrometer or on a Bruker Avance 600 spectrometer equipped
with a TCI cryoprobe. Coupling constants (J) are reported in Hz, chemi-
cal shifts (δ) are given in ppm. Multiplicities are reported as follows:
s = singlet, d = doublet, t = triplet, m = multiplet, br s = broad sin-
glet, dd = doublet of doublets, td = triplet of doublets. Analysis and as-
signments were made using 1D (1H, 13C and J-modulated spin-echo
(Jmod)) and 2D (COrrelated SpectroscopY (COSY) and Heteronuclear
Single Quantum Coherence (HSQC)) spectra. High-resolution mass
spectra (HRMS) analyses were performed by the CRMPO (Centre
régional de mesures physiques de l'Ouest, University of Rennes, France)
in positive ionization mode (ES+) on a Waters Q-Tof 2.
Fig. 1. Subclasses of arabinoxylan-active Abfs (AXH) according to substrate specificities
(hexagons: D-Xylp units; pentagons: L-Araf units).
the ability to hydrolyze simple chromogenic substrates, such as para-
nitrophenyl α-L-arabinofuranoside (pNPA) and to show fairly strict
specificity for mono-substituted D-Xylp in AX and AXOS, observations
that are consistent with structural data, which in the case of GH51
Abfs reveal that the active site is characterized by a deep −1 subsite
that tightly accommodates a single L-Araf unit [4,5]. In contrast, the
GH43 AXH-m specific AbfA from Bifidobacterium adolescentis (GenBank
accession no BAF39204.1) [6] displays almost no activity on pNPA, while
being able to hydrolyze mono-substituted 4-nitrocatechol (4NTC)
homologues [7], suggesting that the enzyme needs to recognize both
the L-Araf unit and its adjacent hydroxyl group on the D-Xylp residue.
So far, three enzymes have been reliably described as AXH-d [8–10],
with all these belonging to family GH43. The rather singular specificity of
these enzymes appears to be determined by their ability to recognize the
L-Araf unit vicinal to the departing sugar moiety. In a recent study of
HiAXH-d3 from Humicolens insolens (CAL81199.1), it was reported that
this enzyme possesses a deep active site pocket (subsite −1) that
accommodates a 1,3-linked L-Araf and an adjacent shallow binding sub-
site that accommodates the vicinal 1,2-linked L-Araf [11]. Unlike the
aforementioned GH43 AXH-d enzymes, which display tight substrate
specificity, some GH51 enzymes appear to be more promiscuous
[6,12–14], being principally AXH-m, but displaying some activity on di-
arabinofuranosylated D-Xylp moieties (denoted AXH-m,d). For instance,
it has been reported that Abf51A from Cellvibrio japonicus (ACE86344.1)
displays a weak ability to release L-Ara from double-substituted D-Xylp
moieties, although its activity on mono-substituted D-Xylp is 1000
times higher [12], and AXAH-1 from Hordeum vulgare (AAK21879.1)
[13,14] was shown to be able to hydrolyse disubstituted D-Xylp moieties
when these form the non-reducing terminal sugar in AXOS.
Overall, the comparison and classification of Abfs is not straightfor-
ward, particularly when using oversimplified synthetic substrates
(i.e. mono-glycosides linked directly to aglycon groups, such as pNP),
or complex substrates (e.g. AX) combined with informationally-poor
methods, such as the measurement of the release of reducing sugars,
neither of which allow the study of selectivity. Therefore, to address
this challenge, we have devised a strategy that combines the use of a
set of substrates displaying different structural features and 1H NMR
analysis [6,10,13]. Together, these provide the means to reveal the fine
details of Abf activity, in particular regioselectivity. To demonstrate
the power of this approach, we have addressed the question of how
the presence of a doubly-linked L-Araf motif in substrates affects the
activity of different Abfs from GH43 and GH51 families.
2.2. Artificial, pseudo-natural and natural substrates
Substrates 1 and 2 were synthesized as described in previous work
[7,15].
Substrate 3 was synthesized in-house (Supplementary Fig. 2). First,
to prepare benzyl 4-O-benzoyl-2,3-di-O-(2,3,5-tri-O-benzoyl-α-L-
arabinofuranosyl)-β-D-xylopyranoside 7, working under nitrogen, ben-
zyl 4-O-benzoyl-β-D-xylopyranoside 5 (328 mg, 0.95 mmol, 1 eq.) [16]
was partially dissolved in anhydrous DCM (25 mL) in a two-neck flask
equipped with a pressure-equalizing dropping funnel. Using a syringe,
BF3·Et2O (60 μL, 0.49 mmol, 0.5 eq.) was added to the stirred yellow
suspension. Next, trichloroacetimidate 6 (1.33 g, 2.19 mmol, 2.3 eq.)
[17] was dissolved in anhydrous DCM (10 mL) in the presence of acti-
vated molecular sieves (4 Å), and was slowly transferred to the reaction
mixture over 20 min via the dropping funnel. The funnel was rinsed
with 10 mL of anhydrous DCM, which were further added to the reac-
tion. The mixture was stirred at room temperature for 60 min, and
then neutralized with triethylamine. After, the residue was transferred
to a decanting funnel and water (50 mL) was added. After separation,
the aqueous phase was extracted twice with DCM. The combined organ-
ic extracts were first washed with a saturated solution of NaHCO3, then
with a small amount of water, and finally with brine, before being dried
(MgSO4), filtered, and evaporated. The residue was purified by auto-
mated flash column chromatography, using a 20 to 40% gradient of
EtOAc in petroleum ether, which provided compound 7 as a white
solid (676 mg, 0.55 mmol, 56%). 1H NMR (500 MHz, CDCl3, 298 K) δ
8.07–8.05 (m, 2H), 7.95–7.91 (m, 6H), 7.90–7.88 (m, 2H), 7.75–7.74
(m, 2H), 7.69–7.67 (m, 2H), 7.61–7.56 (m, 2H), 7.49–7.39 (m, 9H),
7.32–7.29 (m, 2H), 7.23–7.14 (m, 13H), 5.80 (s, 1H), 5.79 (s, 1H), 5.53
(d, 1H, J = 0.6), 5.45 (d, 1H, J = 4.3), 5.44 (d, 1H, J = 0.5), 5.41
(d, 1H, J = 4.0), 5.32–5.27 (m, 1H), 4.85 (d, 1H, J = 11.0), 4.67 (d, 1H,
J = 7.0), 4.56–4.54 (m, 1H), 4.52 (dd, 1H, J = 11.9, 3.2), 4.43–4.32
(m, 5H), 4.27 (dd, 1H, J = 12.2, 5.5), 4.24 (dd, 1H, J = 11.7, 4.4), 4.10
(dd, 1H, J = 9.7, 7.0), 3.53 (dd, 1H, J = 12.2, 7.7); 13C NMR (126 MHz,
CDCl3, 298 K) δ 166.2, 166.0, 165.5, 165.5, 165.5, 165.4, 165.3, 136.7,
133.6, 133.5, 133.5, 133.1, 133.0, 133.0, 132.9, 132.8, 129.9, 129.9,
129.8, 129.8, 129.7, 129.7, 129.6, 129.5, 129.4, 129.3, 129.1, 128.9,
128.9, 128.5, 128.4, 128.4, 128.4, 128.2, 128.2, 128.2, 128.2, 128.1,
106.5, 106.0, 101.6, 81.9, 81.8, 81.8, 81.7, 77.9, 77.7, 77.3, 77.1, 71.7,
71.2, 63.4, 63.4, 63.2. Next, to synthesize benzyl 2,3-di-O-α-L-
arabinofuranosyl-β-D-xylopyranoside 3, working under nitrogen, 7
2. Material and methods
2.1. General reagents and methods
Chemicals were purchased from either Sigma-Aldrich or ACROS and
used as received, without further purification. Sodium acetate[D3] and

3108
V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
(254 mg, 0.21 mmol, 1 eq.) was dissolved in a mixture of anhydrous
methanol and DCM (2:1 v/v, 21 mL) at 0 °C. Then, a solution of
MeONa in methanol (1 M, 0.2 mL, 0.2 mmol, 1 eq.) was added dropwise.
The reaction mixture was stirred at room temperature for 5 h, neutral-
ized with Amberlite IR-120 (H+), and filtered. The filtrates were con-
centrated under reduced pressure. Next, the residue was further
purified using an automated flash column chromatography system
equipped with a C18 silica column (12 g) and operating in a 15 to 25%
gradient of acetonitrile in water. Product-containing fractions were con-
centrated, filtered (45 μm filter), and lyophilized, affording 3 (80 mg,
0.16 mmol, 77%) as a white powder. 1H NMR (500 MHz, D2O, 298 K) δ
7.49–7.39 (m, 5H), 5.24 (d, 1H, J = 1.5), 5.18 (d, 1H, J = 1.2), 4.88 (d,
1H, J = 11.1), 4.66 (d, 1H, J = 11.1), 4.64 (d, 1H, J = 7.8), 4.19 (td,
1H, J = 5.9, 4.2), 4.16 (dd, 1H, J = 3.4, 1.7), 4.10 (dd, 1H, J = 2.6, 1.3),
4.02 (dd, 1H, J = 11.8, 4.9), 3.96 (dd, 1H, J = 6.0, 3.2), 3.94–3.90
(m, 2H), 3.81 (dd, 1H, J = 12.3, 3.4), 3.74–3.67 (m, 3H), 3.53 (dd, 1H,
J = 8.8, 7.6), 3.48 (dd, 1H, J = 12.6, 3.8), 3.44 (dd, 1H, J = 12.6, 3.2),
3.37 (dd, 1H, J = 11.8, 9.7); 13C NMR (126 MHz, D2O, 298 K) δ 136.5,
129.0, 128.7, 128.5, 108.5, 108.5, 101.0, 83.9, 83.8, 82.1, 81.4, 81.1,
77.9, 76.5, 76.4, 71.9, 68.0, 64.7, 61.2, 60.2; HRMS (ESI): m/z: calculated
for C22H32O13Na ([M + Na]+): 527.17406, found 527.1741 (0 ppm).
The AXOS, A2+ 3XX 4 [18], was isolated as previously described [19].
experiments, upon enzyme addition and mixing, the NMR tube was im-
mediately transferred to the spectrometer, and after temperature stabili-
zation, spectra consisting of an 8-scan accumulation were recorded
continuously, thus providing the first spectrum 6 to 10 min after enzyme
addition.1H NMR scans were accumulated continuously over 1.45 min
(8 scans with a repetition delay of 6 s) during 4 to 15 h, depending on
the reaction. Each NMR spectrum was acquired using an excitation flip
angle of 30° at a radiofrequency field of 29.7 kHz, and the residual
water signal was pre-saturated during the repetition delay (with a radio-
frequency field of 21 Hz). The following acquisition parameters were
used: relaxation delay (5 s) and dummy scans (2).
2.5. Choice of significant protons
All spectral chemical shifts were calibrated by setting the acetate
buffer peak at 1.92 ppm, because a separate control experiment demon-
strated that this value is unaffected within the experimental tempera-
ture range (i.e. 303 to 333 K). Similarly, signals from substrates and
products provided stable chemical shifts, within 0.01 ppm, from one
series of experiments to another. The residual acetate peak was also uti-
lized as a reference for integration scaling for the hydrolysis of 2, 3, and
4, because its concentration remained constant throughout. However,
this strategy could not be applied in the case of substrate 1, because
the methylene proton signal of the linker arm of 1c overlaps with the
acetate reference. Consequently, in this particular case, the H-6 of
4NTC in the aromatic region, which remains identical in all four 4NTC
derivatives, was used for the integration calibration.
Signals usable for hydrolysis monitoring were selected in order to
avoid overlapping of products and starting material, as well as being dis-
tant enough from the HOD peak as to remain untouched by the solvent
signal pre-saturation (Figs. 2–5). As a general rule, the zone between
4.88 and 4.14 ppm could not be exploited since H-2, H-3, H-4, and H-5
protons of the four major different forms of free L-Ara come out in this
area. As a result, most protons selected for integration were located in
the anomeric region where the H-1 of α-L-Araf is often distinctive
enough (Table 1). Hydrolysis of compound 2 was conveniently moni-
tored by following the evolution of the H-6 of 4NTC in the aromatic re-
gion, because α-L-Araf substitutions greatly affect its chemical shift. As a
result, the starting material, the intermediates, and the final product
were all characterized by fairly distinctive signals. The monitoring of
the hydrolysis of 3 was more problematic. Firstly, H-1 β-D-Xylp of 3b
overlaps with H-1 of β-L-Arap. Secondly, depending on the reaction tem-
perature, not all signals were always available for integration, as the
HOD peak interfered with the Bn methylene signals of 3b at 303 K.
Therefore, the hydrolysis of 3 by AbfB, could only be monitored by mea-
suring the disappearance of the α-L-AraB H-1 of 3 (Fig. 4c). Also, the ap-
pearance of 3a in BaAXH-d3 (303 K) experiments was easily tracked
through the H-1 β-D-Xylp proton, while its disappearance in AbfA ex-
periments (323 K) was followed with the Bn methylene signal, because
of the HOD attenuation shift. Finally, in the case of the natural substrate
4, all possible intermediates, as well as the final product have already
been characterized [19,22,23]. We observed that the signals with the
least interference were those of H-2 and H-4 of α-L-AraA of 4 when no
intermediate was present, which is acceptable in the case reactions in-
volving AbfB and TxAbf, because the formation of intermediates never
exceeded 10%. To monitor reactions catalyzed by AbfA and BaAXH-d3,
the H-1 of α-L-AraB of 4a was followed, even though a small overlap
with the β-L-Araf anomeric signal was observed.
2.3. Enzymes
AbfA, AbfB and BaAXH-d3 (BAF39204.1, BAF40305.1 and
AAO67499.1), all from B. adolescentis, and TxAbf from Thermobacillus
xylanilyticus (CAA76421.2), are recombinant enzymes whose produc-
tion has been previously described [6,10,20]. Briefly, these are produced
in Escherichia coli as recombinant proteins (cloning coding sequences
into pEXP5-CT/TOPO or pET expression vectors), each C-terminally
fused to a His6 tag, which facilitates purification from filtered lysate
using immobilized metal affinity chromatographic technology (HiTrap
HP 1 mL columns for the bifidobacterial enzymes and Clontech CellThru
10 mL disposable column containing TALON® Metal Affinity Resin for
TxAbf). After purification, the concentration of protein solutions was de-
termined spectrophotometrically at 280 nm, using relevant extinction co-
efficients (96720, 115850, 158140 and 115320 M−1·cm−1 for BaAXH-d3,
AbfA, AbfB and TxAbf, respectively) and enzyme purity (N95%) was veri-
fied using SDS-PAGE, before storing enzymes in appropriate buffers
(25 mM citrate pH 6.0 for bifidobacterial enzymes and 20 mM Tris–HCl
pH 8.0 for TxAbf) at 4 °C until use.
2.4. Real time 1H NMR monitoring of enzyme reactions
Enzyme-mediated hydrolysis of 1, 2, and 3 was monitored by 1H
NMR, performing reactions in standard 5 mm NMR tubes, containing
500 μL of deuterated sodium acetate buffer (20 mM), pD 5.87, contain-
ing 5 mM substrate. Measurement of pD was performed using a glass pH
electrode, applying the following relationship pD = pHelectrode + 0.41
[21]. Prior to the reactions, the enzyme was diluted by 10-fold in D2O
(99.90%), followed by concentration using an Amicon® Ultra filter
(regenerated cellulose 10 kDa, Millipore) system, this operation being
performed twice. Next, hydrolyses were initiated by the addition of an
aliquot of the deuterated enzyme solution and reactions were per-
formed at the optimum temperature of the studied enzyme (30 °C or
303 K for AbfB and BaAXH-d3, 50 °C or 323 K for AbfA and 60 °C or 333
K for TxAbf). Control experiments that did not contain enzyme were con-
ducted in parallel, in order to monitor the spontaneous hydrolysis of
substrates. After 24 h, a small amount of substrate 2 spontaneously hydro-
lyzed into 2a and 2b — 5% at 50 °C and 15% at 60 °C. Substrates 1 and 3
remained stable over 24 h within the considered temperature range.
Overall, the stability of the substrates was judged sufficient, since the
only sensitive experiment (TxAbf with substrate 2) did not exceed
200 min. The enzyme-mediated hydrolysis of 4 (2 mM) was performed
in 3 mm NMR tubes at 600 MHz for improved sensitivity. In typical
2.6. Evaluation of catalytic activity
Initial rates were derived by analyzing the linear part of the graphs
showing the amount of L-Araf units liberated by the enzymes (Fig. 6,
and Supplementary Figs. 3–6 and Table 1). Precise intervals between
two spectra (data points) were provided by their recording times.
For enzymes that are unable to remove both L-Araf moieties from

V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
3109
Fig. 2. 1H NMR spectra of the hydrolysis of artificial substrate 1 by different Abfs. a) 5 mM 1 in 20 mM deuterated sodium acetate, pD 5.87; b) after 136 min incubation with TxAbf at 60 °C;
c) 296 min incubation with TxAbf at 60 °C; and d) 1074 min incubation with BaAXH-d3 at 30 °C.
Fig. 3. 1H NMR spectra of the hydrolysis of artificial substrate 2 by different Abfs. a) 5 mM 1 in 20 mM deuterated sodium acetate, pD 5.87; b) after 80 min incubation with TxAbf at 60 °C;
c) 105 min incubation with AbfB at 30 °C; d) 437 min incubation with AbfB at 30 °C; and e) 221 min incubation with BaAXH-d3 at 30 °C.

3110
V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
Fig. 4. 1H NMR spectra of the hydrolysis of pseudo-natural substrate 3 by different Abfs. a) 5 mM 3 in 20 mM deuterated sodium acetate, pD 5.87; b) after 78 min incubation with TxAbf at
60 °C; c) 493 min incubation with AbfB at 30 °C; d) 140 min incubation with BaAXH-d3 at 30 °C.
disubstituted D-Xylp moieties, the total amount of sugar released was
inferred from the time-dependent consumption of the starting material,
except in the case of BaAXH-d3-catalyzed hydrolysis of 4, where the
appearance of 4a was monitored instead. In the case of enzymes able
to cleave both L-Araf moieties, the total amount of free sugar was calcu-
lated based on the consumption of the starting material, as well as the
Fig. 5. 1H NMR spectra of the hydrolysis of natural AXOS 4 by different Abfs. a) 2 mM 4 in 20 mM deuterated sodium acetate, pD 5.87; b) after 82 min incubation with TxAbf at 40 °C;
c) 291 min incubation with TxAbf at 40 °C; and d) 39 min incubation with BaAXH-d3 at 30 °C.

V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
3111
Table 1
List of the proton signals used for the tracking of the different species involved during the Abf-catalyzed enzymatic reactions.
Compound
Integrated proton
Signal multiplicity, δ in ppm (J in Hz)
Enzymatic reaction monitoring
β-L-Arap
α-L-Arap
α-L-Araf
β-L-Araf
H1
H1
H1
H1
d, 4.50 (7.8)
d, 5.22 (3.6)
d, 5.23 (2.6)
d, 5.29 (4.7)
d, 5.04 (1.6)
d, 5.16 (1.5)
d, 5.02 (1.4)
d, 7.37 (9.0)
d, 7.01 (9.0)
d, 7.29 (9.0)
d, 6.96 (9.0)
d, 5.24 (1.5)
d, 4.59 (7.6)
d, 4.68 (11.4)
d, 4.73 (11.6)
td, 4.21 (3.4, 1.7)
dd, 4.19 (5.7, 3.4)
s, 5.29
AXH-d3(a), AbfB(a), TxAbf(a) with 1 and 3
BaAXH-d3(a), AbfB(a), TxAbf(a) AbfA(a) with 2
1
H1-(L-AraA)
H1-(L-AraB)
H1-(L-AraA)
H6-(4NTC)
H6-(4NTC)
H6-(4NTC)
H6-(4NTC)
H1-(L-AraA)
H1-(D-Xylp)
CH2 (Bn)
CH2 (Bn)
H4 and H2-(L-AraA) (A3)
BaAXH-d3(d), AbfA(d), TxAbf(d), AbfB(d)
BaAXH-d3(a), AbfA(d), TxAbf(ad), AbfB(ad)
BaAXH-d3(a), AbfA(d), TxAbf(ad), AbfB(ad)
BaAXH-d3(d), AbfA(d), TxAbf(d), AbfB(d)
BaAXH-d3(a), AbfA(d), TxAbf(ad), AbfB(ad)
BaAXH-d3(a), AbfA(d), TxAbf(ad)
AbfA(a), TxAbf(a), AbfB(a)
BaAXH-d3(d), AbfB(d), TxAbf(d)
BaAXH-d3(a)
AbfA(d)
1a
1b
2
2a
2b
2c
3
3a
3a
3b
4
AbfA(a), TxAbf(a)
AbfB(d), TxAbf(d)a
4a
H1-(L-AraB) (A2)
BaAXH-d3(a), AbfA(d)a, TxAbf(ad)a, AbfB(ad)
(a): appearance, (d): disappearance.
a
Hydrolysis of 4 and 4a mediated by TxAbf and AbfA were done at 313 K due to the cryoprobe temperature limitation.
amount of final product liberated. An accurate measurement of the zero
time point was impossible due to the intrinsic limits of the experiment
(i.e. the time required to stabilize temperature).
aliquot of substrate (data not shown). The performance of BaAXH-d3
was comparable on compounds 2, 3 and 4, but substrate 1 proved to
be more resilient (Fig. 6 and Supplementary Table 1). Concerning
reaction regioselectivity, only a slight cleavage preference was ob-
served when using substrate 2, with a final product ratio (2a:2b) of
63:37. However, much higher selectivity was observed when using
compound 1, since a final product ratio (1b:1a) of 94:6 was mea-
sured. Finally, reactions involving 3 and 4 (i.e. pseudo-natural and
natural di-arabinofuranosylated D-Xylp moieties) were regiospecific,
yielding only O-2 substituted D-Xylp motifs.
3. Results
To provide a detailed investigation of Abf activity, we focused our
attention on the question of how the presence of a doubly-linked
L-Araf motif in substrates affects the activity of Abfs from different
GH families. To achieve this, we used real-time 1H NMR spectroscopy
to monitor enzyme reactions and artificial substrates 1 and 2 [7,15],
the pseudo-natural A2+ 3Bn 3 and the natural substrate A2+ 3XX
4 (Figs. 2–5 and Supplementary Fig. 1). Accordingly, we were able to
closely probe the action of four Abfs from families GH43 and 51 that
display different hydrolytic behavior on AX [2,6,10,20].
3.2. GH43 AbfA from B. adolescentis
AbfA (at a concentration of 3.8 μg·mL−1) completely hydrolyzed
the mono-substituted intermediate compounds 2a, 2b, 3a and 4a in
less than 7 h (Supplementary Fig. 6) and with hydrolytic activities for
these substrates being of the same order of magnitude (Fig. 6 and Sup-
plementary Table 1). Nevertheless, AbfA failed to hydrolyze 1b, even
after 24 h incubation. Regarding the doubly-substituted moiety and in
accordance with literature data [6,7,15], compounds 1 and 3 were unaf-
fected, even after incubation at 50 °C for 24 h using a N10-fold higher
concentration (47.2 μg·mL−1) of enzyme, although compound 2
3.1. GH43 AXH-d3 from B. adolescentis
BaAXH-d3 hydrolyzed all four substrates, in each case removing only
one L-Araf, while leaving the other one in place (Figs. 2d, 3e, 4d, 5d). This
was not due to enzyme inactivation, since the enzyme remained suffi-
ciently active at the end of the reaction to be able to hydrolyze a fresh
Fig. 6. A logarithmic-scale comparison of the hydrolytic activities of the different Abfs on various di- and mono-arabinofuranosylated substrates. Values refer to the quantity (μmol) of L-Ara
unit liberated per min and per mg of enzyme, inferred from the initial rate of the reaction measured by NMR. Since AbfA is inactive towards 1, 1b, 3 and 4 and only marginally active on 2,
the activities were measured using a mixture of 2a and 2b, 3a and 4a generated by the action of BaAXH-d3 on disubstituted moieties.

3112
V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
provided a slightly ambiguous result. Indeed, in identical conditions,
30% of compound 2 were hydrolyzed by AbfA after 24 h, which is 25%
higher than the spontaneous hydrolysis that characterized the control
reaction. Though measurable, it is noteworthy that the level of hydroly-
sis observed with 2 was still more than a 100-fold lower than that
observed in the AbfA hydrolysis of 2a and 2b.
4. Discussion
In the present study, we set out to investigate the specificities of differ-
ent Abfs using a set of substrates. Substrate 1 [15], displays a cleavable
linker arm between an unhindered double L-Araf moiety and an aryl leav-
ing group (4NTC), which offers the possibility of free rotation around the
bond between the two sugars. This implies that substrate 1 could adopt a
syn-conformation, although this would not be the favored one. Similarly,
substrate 2 bears the 4NTC moiety and the L-Araf units locked in cis posi-
tions with respect to each other. This substrate was chosen because it has
already been demonstrated that it is a good artificial substrate for AXH-d
type enzymes [7]. The non-chromogenic substrate 3, a benzyl di-
arabinofuranosylated β-D-xylopyranoside, was included in this study,
because it imitates the basic motif of natural disubstituted AXOS struc-
tures. The D-Xylp moiety of this compound is locked in β anomery, and
is linked to a benzyl (Bn) group that can occupy the subsite +2 in the ac-
tive site of enzymes. Finally, the terminally L-Araf-disubstituted AXOS,
3.3. GH51 Abfs from T. xylanilyticus and B. adolescentis
All of the substrates under study released two units of free L-Ara per
unit of initial substrate when digested by TxAbf and AbfB, including
natural AXOS 4 (Supplementary Figs. 3 and 4). The efficiency of these
enzymes was particularly high with 2 and its mono-glycosylated deriv-
atives (Figs. 6 and 7, and Supplementary Table 1). Mono-glycosylated
intermediates were observed during the hydrolysis of 1, 2 and 4, but
not in 3 (Figs. 2–5). When hydrolyzing 1, both enzymes preferentially
generated 1a, which attained 31 and 35% mol of the reaction compo-
nents when using TxAbf and AbfB, respectively, while 1b never
exceeded 5% mol (Figs. 2b–c and 7). Regarding the hydrolysis of 4 by
TxAbf or AbfB, the intermediate 4a (O-2 substituted D-Xylp moiety), de-
tectable at 5.29 ppm, reached 10% mol during the reaction, while the
characteristic signal of the O-3 substituted D-Xylp moiety (5.33 ppm)
remained undetectable (Fig. 5b–c). The only major difference in behav-
ior between the two GH51 Abfs concerned their regioselectivity on
compound 2 (Fig. 7). In this regard, AbfB only produced 2a, while
TxAbf generated both regioisomers, with a slight preference for 2b.
A
2+ 3XX 4 was included, since this is fully representative of a natural L-
Araf disubstituted substrate. In 3 and 4, the L-Arafs are locked in equatorial
positions resulting in a small dihedral angle between the two sugars (i.e.
corresponding to a syn conformation).
4.1. BaAXH-d3, a specialist GH43 enzyme that acts on double-substituted
xylan
BaAXH-d3 only released one L-Araf unit from doubly-substituted
substrates. This observation is consistent with the known properties of
Fig. 7. The time-dependent evolution of relevant (integrated) proton signals during the hydrolysis of artificial substrates, 1 (a and b) and 2 (c and d), by AbfB (a and c) and TxAbf (b and d).
The concentrations of fully hydrolyzed product were determined either by measuring the 4NTC moiety signals for the hydrolysis of 2, or calculated from NMR data for the hydrolysis of 1.
The total free L-Ara was inferred by comparing the substrate and product curves.

V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
3113
this enzyme [6,10] and can be easily explained by considering the prob-
able molecular conformations of the different compounds. In 1 the free
rotation across the C\C bond of the chain is likely to lead to preferential
positioning of the two sugars in mutual anti positions, even though
syn-positioning (required for enzyme-mediated hydrolysis) remains
possible. In practice, in 1 the sugar that is preferentially released is the
most hindered one (i.e. the one borne by the secondary hydroxyl
group). This suggests that the other sugar, which benefits from a greater
degree of freedom, is first to bind to the enzyme, an interaction that is
pursued by chain rotation and binding of the hindered sugar into the ac-
tive site pocket. This explanation is consistent with data from the study
of HiAXH-d3 from H. insolens that indicates that HiAXH-d3 specifically
recognizes the L-Araf unit in the vicinal position of the cleaved sugar
[11]. Even if the latter only displays 24% identity with BaAXH-d3, both
enzymes nonetheless hold in common eight key active site residues
(Supplementary Fig. 7) so the comparison is likely to be valid. In this
respect it is also noteworthy that when acting on 3 or 4, BaAXH-d3
only yielded O-2 substituted D-Xylp motifs. Interestingly, this finding
contrasts with the data presented for HiAXH-d3 [11], which revealed
that this enzyme's strict regioselectivity stems from the interaction of
Trp526 with the endocyclic oxygen of the D-Xylp positioned in the +2
subsite, which constitutes the only asymmetric feature of the xylan
main-chain. In BaAXH-d3 such an explanation cannot be fully valid,
first because there is no equivalent of Trp526 (Supplementary Fig. 7)
and second because substrate 3 lacks an endocyclic oxygen in the +2
position. Taken together these observations suggest that the D-Xylp res-
idue in the +1 subsite defines the orientation of the rest of the molecule
in the active site. Therefore, in spite of their very similar regiospecific
activity on natural disubstituted substrates, our results suggest that
the substrate recognition determinants in BaAXH-d3 and HiAXH-d3
are not strictly identical.
noteworthy that the positioning of O-3-linked L-Araf unit within −1
subsite is no doubt facilitated in the case of 4, where the disubstituted
D-Xylp moiety is located at the non-reducing terminal. Internal disubsti-
tuted D-Xylp moieties are likely to be much more constrained, which ex-
plains previous observations, indicating that GH51 Abfs do not generally
release L-arabinose from di-arabinofuranosylated D-Xylp moieties in
heteroxylans [24]. Moreover, this appears rather logical, since if the
O-3 L-Araf unit of 4 does indeed bind in the −1 subsite, then it is prob-
able that the O-2 L-Araf moiety is bound in subsite +2′, a subsite that ac-
commodates D-Xylp moieties when the GH51 Abfs act on internal L-Araf
units. The only major difference in behavior between the two GH51 Abfs
involves the regioselectivity associated with the hydrolysis of com-
pound 2, which suggests that the nitro group can generate a strong in-
teraction that specifically influences the hydrolysis pattern in AbfB
(Figs. 3b–d and 7). Nevertheless, the formation of up to 53 and 41%
mol of intermediates by TxAbf and AbfB respectively indicates that the
recognition of double and mono-substituted species are of the same
order of magnitude in both cases. Moreover, it is noteworthy that both
enzymes hydrolyzed 2, 2a and 2b more efficiently than the other
three compounds (Fig. 6), suggesting that these enzymes show prefer-
ence for a direct linkage to the aryl leaving group [25]. Indeed, the mea-
sured specific activities were consistent with those measured using
pNPA [7], inferring that the +1 subsites of these enzymes display strong
recognition of the aromatic ring. According to a previous structural anal-
ysis [5], the +1 subsite of TxAbf is characterized by the presence of
Trp302 and, to a lesser extent, Trp248, which are ideally placed to inter-
act with a sugar ring or an aromatic group, such as 4NTC or pNP. In
AbfB it is more difficult to pinpoint the residues that might favor the rec-
ognition of aromatic groups in its +1 subsite, but assuming that the cat-
alytic residues are localized at positions 358 and 445 (acid/base and
nucleophile) respectively, then one might assume that Trp449 in AbfB
is the functional homologue of Trp302 in TxAbf (Supplementary Fig. 8).
Finally, it is noteworthy that the +2′ subsite is partly defined by the
amino acid pair His98–Trp99, which is borne on the mobile β2α2 loop
[26]. Previously it was suggested that substrate binding in the active
site of TxAbf involves an induced fit mechanism and that the movement
of the β2α2 loop is only possible because of the presence of a shortened
β7α7 loop, which is not a universal feature in family GH51. In AbfB, it
is likely that a similar configuration prevails, with Leu263–Trp264
forming part of the β2α2 loop, and associated with a short β7α7 loop.
Overall, these observations reinforce the idea that AbfB might also fix
substrates using an induced fit mechanism and that both enzymes
possess a similar ability to bind terminally-double substituted D-Xylp
using the presence of the +1 (D-Xylp binding) and +2′ subsites
(L-Araf binding), respectively.
4.2. Fine substrate recognition in GH43 AbfA from B. adolescentis
It is known that AbfA specifically releases L-Araf from mono-
substituted D-Xylp residues, and is inactive on double-substituted
D-Xylp residues [6]. Therefore, the fact that AbfA completely hydrolyzed
the mono-substituted species 2a, 2b, 3a and 4a was expected. However,
it was more surprising to find that AbfA cannot hydrolyze 1b. This rather
high selectivity indicates that the +1 subsite of AbfA must possess rec-
ognition determinants both for the D-Xylp ring itself and the associated
vicinal hydroxyl, because the dissociation of these elements in com-
pound 1b destroys substrate recognition.
4.3. Family GH51 can completely de-ramify double-substituted D-xylosyl
moieties
Apparently, both TxAbf and AbfB display a mixed AXH-m,d activity,
since they were able to completely digest the four assayed substrates
(Supplementary Figs. 3 and 4). Moreover, the maximum level of pro-
duction of intermediate reaction products was more dependent on the
nature of the substrates than on the enzyme used, suggesting that the
activity of both GH51 enzymes is sensitive to structural changes in the
vicinity of the scissile bond. For example, when acting on 3 (Fig. 4b–c),
neither enzyme produced detectable amount of intermediates, which
suggests that the rate-limiting reaction was the first one, required to
generate the intermediate compounds. When TxAbf and AbfB hydro-
lyzed 1, the preferential production of 1a was almost certainly a witness
to the previously mentioned fact that the L-Araf unit linked to the sec-
ondary alcohol group is more hindered than the second L-Araf unit
(Figs. 2b–c and 7). Similarly, the fact that both enzymes failed to pro-
duce an O-3 substituted D-Xylp-containing intermediate compound
when hydrolyzing 4 can also be explained by the relative accessibility
of the two L-Araf moieties and thus the preference of the enzymes for
the 1,3-linked L-Araf moiety, a preference that has already been
described for a barley arabinofuranosidase [13]. Moreover, it is
5. Conclusions
Using well-designed artificial substrates we have demonstrated how
these can be used to investigate selectivity in Abfs, a property that often
remains unstudied due to the lack of appropriate substrates and readily
accessible methods. On the basis of our work, we believe that the two
chromogenic substrates (1 and 2) described here are a useful addition
for the discovery and preliminary characterization of Abfs. Beyond this
methodological contribution, our work has also revealed several inter-
esting results, but the most important of these is that GH51 Abfs,
enzymes that are generally considered to be active on the L-Araf group
of mono-substituted D-Xylp in heteroxylans, are also perfectly able to
completely debranch di-arabinofuranosylated D-Xylp moieties, when
these are located at the non-reducing end of xylo-oligosaccharides and
probably heteroxylans. In the context of AX degradation by complex
enzyme consortia this observation is significant, because it is likely that
xylo-oligosaccharides bearing a non-reducing di-arabinofuranosylated
D-Xylp moiety are a quite frequent occurrence, being a possible product
of GH10 xylanase-mediated hydrolysis [27].

3114
V. Borsenberger et al. / Biochimica et Biophysica Acta 1840 (2014) 3106–3114
either single or double substituted xylose residues in arabinoxylans, Enzym. Microb.
Acknowledgements
Technol. 48 (2011) 397–403.
[10] K.M.J. Van Laere, G. Beldman, A.G.J. Voragen, A new arabinofuranohydrolase from
Bifidobacterium adolescentis able to remove arabinosyl residues from double-
substituted xylose units in arabinoxylan, Appl. Microbiol. Biotechnol. 47 (1997)
231–235.
[11] L.S. McKee, M.J. Peña, A. Rogowski, A. Jackson, R.J. Lewis, W.S. York, K.B.R.M. Krogh,
A. Viksø-Nielsen, M. Skjøt, H.J. Gilbert, J. Marles-Wright, Introducing endo-xylanase
activity into an exo-acting arabinofuranosidase that targets side chains, Proc. Natl.
Acad. Sci. U. S. A. 109 (2012) 6537–6542.
This work was supported by Région Midi-Pyrénées grant DAER
Recherche 10008500 (to V.B.). MetaToul (Metabolomics & Fluxomics
Facitilies, Toulouse, France, www.metatoul.fr) and its staff members
(Lindsay Peyriga, Edern Cahoreau and Jean-Charles Portais) are grate-
fully acknowledged for the technical support and access to NMR facility.
MetaToul is part of the national infrastructure MetaboHUB (The
French National infrastructure for metabolomics and fluxomics, www.
metabohub.fr). MetaToul is supported by grants from the Région
Midi-Pyrénées, the European Regional Development Fund, the SICOVAL,
the Infrastructures en Biologie Sante et Agronomie (IBiSa, France), the
Centre National de la Recherche Scientifique (CNRS) and the Institut
National de la Recherche Agronomique (INRA). NMR experiments
were also performed on the PICT — Genotoul platform of Toulouse
and funded by CNRS, Université de Toulouse—UPS, IBiSa, European
structural funds and the Midi-Pyrénées region.
[12] M.-H. Beylot, V.A. McKie, A.G.J. Voragen, C.H.L. Doeswijk-Voragen, H.J. Gilbert, The
Pseudomonas cellulosa glycoside hydrolase familly 51 arabinofuranosidase exhibits
wide substrate specificity, Biochem. J. 358 (2001) 607–614.
[13] H. Ferré, A. Broberg, J.O. Duus, K.K. Thomsen, A novel type of arabinoxylan
arabinofuranohydrolase isolated from germinated barley. Analysis of substrate pref-
erence and specificity by nano-probe NMR, Eur. J. Biochem. 267 (2000) 6633–6641.
[14] R.C. Lee, R.A. Burton, M. Hrmova, G.B. Fincher, Barley arabinoxylan
arabinofuranohydrolases: purification, characterization and determination of
primary structures from cDNA clones, Biochem. J. 356 (2001) 181–189.
[15] V. Borsenberger, E. Dornez, M.-L. Desrousseaux, C.M. Courtin, M.J. O'Donohue, R. Fauré,
A substrate for the detection of broad specificity α-L-arabinofuranosidases with
indirect release of a chromogenic group, Tetrahedron Lett. 54 (2013) 3063–3066.
[16] E.V. Evtushenko, Regioselective benzoylation of glycopyranosides by benzoic
anhydride in the presence of Cu(CF3COO)2, Carbohydr. Res. 359 (2012) 111–119.
[17] Y. Du, Q. Pan, F. Kong, An efficient and concise regioselective synthesis of α-(1→5)-
linked L-arabinofuranosyl oligosaccharides, Carbohydr. Res. 329 (2000) 17–24.
[18] R. Fauré, C.M. Courtin, J.A. Delcour, C. Dumon, C.B. Faulds, G.B. Fincher, S. Fort, S.C.
Fry, S. Halila, M.A. Kabel, L. Pouvreau, B. Quemener, A. Rivet, L. Saulnier, H.A.
Schols, H. Driguez, M.J. O'Donohue, A brief and informationally rich naming system
for oligosaccharide motifs of heteroxylans found in plant cell walls, Aust. J. Chem. 62
(2009) 533–537.
Appendix A. Supplementary Data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.bbagen.2014.07.001.
References
[19] H. Pastell, P. Tuomainen, L. Virkki, M. Tenkanen, Step-wise enzymatic preparation
and structural characterization of singly and doubly substituted arabinoxylo-
oligosaccharides with non-reducing end terminal branches, Carbohydr. Res. 343
(2008) 3049–3057.
[20] T. Debeche, N. Cummings, I. Connerton, P. Debeire, M.J. O'Donohue, Genetic and
biochemical characterization of a highly thermostable α-L-arabinofuranosidase
from Thermobacillus xylanilyticus, Appl. Environ. Microbiol. 66 (2000) 1734–1736.
[21] P.K. Glasoe, F.A. Long, Use of glass electrode to mesure acidities in deuterium oxide,
J. Phys. Chem. 64 (1960) 188–190.
[22] R.A. Hoffmann, B.R. Leeflang, M.M.J. de Barse, J.P. Kamerling, J.F.G. Vliegenthart,
Characterisation by 1H-n.m.r. spectroscopy of oligosaccharides, derived from
arabinoxylans of white endosperm of wheat, that contain the elements →4)
[α-L-Araf-(1→3)]-β-D-Xylp-(1→ or →4)[α-L-Araf-(1→2)][α-L-Araf-(1→3)]-β-D-
Xylp-(1→, Carbohydr. Res. 221 (1991) 63–81.
[1] C. Dumon, L. Song, S. Bozonnet, R. Fauré, M.J. O'Donohue, Progress and future
prospects for pentose-specific biocatalysts in biorefining, Process Biochem. 47
(2012) 346–357.
[2] S. Lagaert, A. Pollet, C.M. Courtin, G. Volckaert, β-Xylosidases and α-L-
arabinofuranosidases: accessory enzymes for arabinoxylan degradation, Biotechnol.
Adv. 32 (2014) 316–332.
[3] V. Lombard, H. Golaconda Ramulu, E. Drula, P.M. Coutinho, B. Henrissat, The
carbohydrate-active enzymes database (CAZy) in 2013, Nucleic Acids Res. 42
(2014) D490–D495.
[4] K. Hövel, D. Shallom, K. Niefind, V. Belakhov, G. Shoham, T. Baasov, Y. Shoham, D.
Schomburg, Crystal structure and snapshots along the reaction pathway of a family
51 α-L-arabinofuranosidase, EMBO J. 22 (2003) 4922–4932.
[5] G. Paës, L.K. Skov, M.J. O'Donohue, C. Rémond, J.S. Kastrup, M. Gajhede, O. Mirza, The
structure of the complex between a branched pentasaccharide and Thermobacillus
xylanilyticus GH-51 arabinofuranosidase reveals xylan-binding determinants and
induced fit, Biochemistry 47 (2008) 7441–7451.
[23] H. Gruppen, R.A. Hoffmann, F.J.M. Kormelink, A.G.J. Voragen, J.P. Kamerling, J.F.G.
Vliegenthart, Characterisation by 1H NMR spectroscopy of enzymically derived oli-
gosaccharides from alkali-extractable wheat-flour arabinoxylan, Carbohydr. Res.
233 (1992) 45–64.
[6] S. Lagaert, A. Pollet, J.A. Delcour, R. Lavigne, C.M. Courtin, G. Volckaert, Substrate
specificity of three recombinant α-L-arabinofuranosidases from Bifidobacterium
adolescentis and their divergent action on arabinoxylan and arabinoxylan oligosac-
charides, Biochem. Biophys. Res. Commun. 402 (2010) 644–650.
[7] V. Borsenberger, F. Ferreira, A. Pollet, E. Dornez, M.-L. Desrousseaux, S. Massou, C.M.
Courtin, M.J. O'Donohue, R. Fauré, A versatile and colorful screening tool for the iden-
tification of arabinofuranose-acting enzymes, ChemBioChem 13 (2012) 1885–1888.
[8] H.R. Sørensen, C.T. Jørgensen, C.H. Hansen, C.I. Jørgensen, S. Pedersen, A.S. Meyer, A
novel GH43 α-L-arabinofuranosidase from Humicola insolens: mode of action and
synergy with GH51 α-L-arabinofuranosidases on wheat arabinoxylan, Appl.
Microbiol. Biotechnol. 73 (2006) 850–861.
[24] C. Rémond, I. Boukari, G. Chambat, M. O'Donohue, Action of a GH 51
α-L-arabinofuranosidase on wheat-derived arabinoxylans and arabino-
xylooligosaccharides, Carbohydr. Polym. 72 (2008) 424–430.
[25] L. Novaroli, G. Bouchard Doulakas, M. Reist, B. Rolando, R. Fruttero, A. Gasco, P.-A.
Carrupt, The lipophilicity behavior of three catechol-O-methyltransferase (COMT)
inhibitors and simple analogues, Helv. Chim. Acta 89 (2006) 144–152.
[26] F. Arab-Jaziri, B. Bissaro, S. Barbe, O. Saurel, H. Débat, C. Dumon, V. Gervais, A. Milon,
I. André, R. Fauré, M.J. O'Donohue, Functional roles of H98 and W99 and β2α2 loop
dynamics in the α-L-arabinofuranosidase from Thermobacillus xylanilyticus, FEBS J.
279 (2012) 3598–3611.
[27] G. Pell, E.J. Taylor, T.M. Gloster, J.P. Turkenburg, C.M.G.A. Fontes, L.M.A. Ferreira, T.
Nagy, S.J. Clark, G.J. Davies, H.J. Gilbert, The mechanisms by which family 10 glyco-
side hydrolases bind decorated substrates, J. Biol. Chem. 279 (2004) 9597–9605.
[9] L. Pouvreau, R. Joosten, S.W.A. Hinz, H. Gruppen, H.A. Schols, Chrysosporium
lucknowense C1 arabinofuranosidases are selective in releasing arabinose from