Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Insights into Broad-Specificity Starch Modification from the Crystal
Structure of Limosilactobacillus Reuteri NCC 2613 4,6-α-
Glucanotransferase GtfB
Tjaard Pijning,*Joana Gangoiti, Evelien M. te Poele, Tim Börner, and Lubbert Dijkhuizen
Cite This: https://doi.org/10.1021/acs.jafc.1c05657
Read Online
ACCESS Metrics & More Article Recommendations *
sıSupporting Information
ABSTRACT: GtfB-type α-glucanotransferase enzymes from glycoside hydrolase family 70 (GH70) convert starch substrates into α-
glucans that are of interest as food ingredients with a low glycemic index. Characterization of several GtfBs showed that they differ in
product- and substrate specificity, especially with regard to branching, but structural information is limited to a single GtfB,
preferring mostly linear starches and featuring a tunneled binding groove. Here, we present the second crystal structure of a 4,6-α-
glucanotransferase (Limosilactobacillus reuteri NCC 2613) and an improved homology model of a 4,3-α-glucanotransferase GtfB (L.
fermentum NCC 2970) and show that they are able to convert both linear and branched starch substrates. Compared to the
previously described GtfB structure, these two enzymes feature a much more open binding groove, reminiscent of and evolutionary
closer to starch-converting GH13 α-amylases. Sequence analysis of 287 putative GtfBs suggests that only 20% of them are similarly
“open”and thus suitable as broad-specificity starch-converting enzymes.
KEYWORDS: lactic acid bacteria, GtfB, α-glucanotransferase, glycoside hydrolase family 70, slowly digestible α-glucans
■INTRODUCTION
Lactic acid bacteria (LAB) are known for their extracellular
“coats”, largely containing α-glucan-type polysaccharides (e.g.,
dextran, mutan, and reuteran), which are synthesized from
sucrose by glucansucrase (GS) enzymes belonging to glycoside
hydrolase family 70 (GH70) (http://www.cazy.org
1
). Recently
however, within GH70, α-glucanotransferase (GT) enzyme
subfamilies were discovered in LAB and non-LAB that are
inactive on sucrose, synthesizing α-glucans from starch-like
substrates. To date, only a few GT enzymes have been
biochemically characterized;
2−12
most of them display α-4,6
transglycosylation specificity, i.e., cleavage of the substrate α-
1,4 glycosidic bond (donor half-reaction) followed by
formation of an α-1,6 glycosidic bond (acceptor half-reaction).
Three subfamilies have been described, GtfB, GtfC, and GtfD,
differing in bacterial origin, reaction and product specificity,
and (predicted) domain organization and displaying variations
of key amino acid residues in GH70 homology motifs I−IV.
13
These subfamilies also provided important insights into the
evolutionary relationships between GH13 α-amylases, which
also act on starch substrates,
14
and the GH70 glucansucrases
(acting on sucrose only), placing the GtfB, -C, and -D
subfamilies as structural and functional intermediates.
2,4,7,15
Notably, starch-degrading GH13 enzymes feature an open-
substrate binding groove, and we previously hypothesized that
the evolution of reaction specificity from α-amylases toward
GT and GS enzymes was accompanied by structural changes
involving loop architecture around the active site.
16
The GtfB-type α-glucanotransferases, exclusively found in
LAB, share 45−50% sequence identity with GSs. Several GtfB
enzymes have been characterized biochemically,
3,6,7,9−12,17
revealing different substrate and product specificities. For
example, the well-characterized GtfB from Limosilactobacillus
reuteri 121 (Lr121 GtfB) preferably converts linear starch-type
substrates, e.g., from amylose, it synthesizes linear isomalto-/
malto-polysaccharides (IMMP),
16−19
but is hardly active on
amylopectin. On the other hand, for two other GtfB enzymes,
identified on the basis of amino acid variations in homology
motifs II and/or IV, initial characterization suggested that they
act on linear (amylose) as well as branched (amylopectin)
starch substrates. From amylose, the GtfB from L. reuteri NCC
2613 (Lr2613 GtfB) synthesizes a branched reuteran-type α-
glucan,
6
while the GtfB from L. fermentum NCC 2970 (Lf2970
GtfB) remarkably displays a unique α-4,3 transglycosylation
specificity.
7
The activity of these enzymes on branched
substrates has not been studied in detail, while such knowledge
is highly relevant for the application of these enzymes for
whole starch modification. For example, the GtfB from
Streptococcus thermophilus was found to be suitable for
converting wheat starch and potato starch.
11,12
Together, the
different reaction specificities of GTs offer great opportunities
to convert starches to different α-glucans, in particular, because
introduction of α-1,3 and α-1,6 linkages decreases their
digestibility in the upper human gastrointestinal tract. This
property makes these α-glucans attractive as low-glycemic
ingredients in food applications.
20−23
Received: September 13, 2021
Revised: October 15, 2021
Accepted: October 20, 2021
Articlepubs.acs.org/JAFC
© XXXX The Authors. Published by
American Chemical Society A
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Downloaded via 213.184.95.99 on October 28, 2021 at 19:32:44 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
In order to exploit the varying specificities of GTs for α-
glucan synthesis using different starch substrates, under-
standing the GT structure−function relationship is of great
value. The accessibility of the substrate binding groove and the
architecture of acceptor substrate binding subsites are key
factors in determining substrate preference as well as the
efficiency and specificity (e.g., α-4,6 vs α-4,3) of the
transglycosylation reaction. In this regard, the so far only
available experimental 3D structure of a GH70 GT enzyme, a
truncated (ΔNΔV) GtfB from L. reuteri 121 (Lr121 GtfB),
provided important initial insights. In its groove, two long
loops (loops A1 and B) cover donor subsites -2 and -3, leading
to the hypothesis that the resulting tunnel-like feature is the
main determinant of the enzyme’s limited ability to process
branched starch substrates.
16
Interestingly, sequence align-
ments showed that the lengths of the A1 and B loops differ in
characterized GT enzymes.
6,16,21
For example, both in Lr2613
GtfB and Lf2970 GtfB these loops are much shorter; homology
models of these enzymes thus predicted a much more open
binding groove,
6,7
a feature that may determine higher ability
to process branched substrates. However, how representative
these differences are within the GtfB subfamily so far has not
been investigated. In addition to the loops, variations also
occur in homology motifs II and IV; in GH70 glucansucrases,
the structures of complexes with sucrose, maltose, or acarbose
revealed residues in these motifs that interact with the donor
substrate, acceptor substrate, or inhibitor, respectively.
24,25
Mutational studies confirmed the contribution of these
residues to reaction specificity.
13,26
However, for GH70 GTs,
no acceptor substrate complexes are available yet, and residues
that may affect reaction specificity in these enzymes could only
be identified by homology with GSs.
Deepening our understanding of GT specificity, we describe
here the second crystal structure of a GH70 α-4,6-GT from L.
reuteri NCC 2613 (Lr2613 GtfB), along with an improved
homology model of the α-4,3-GT from L. fermentum NCC
2970 (Lf2970 GtfB). Both GtfB enzymes feature shorter loops
A1 and B and consequently, instead of having a tunnel, feature
a much more open binding groove compared to Lr121 GtfB;
this allowed us to model the binding of branched starch-like
substrates. Analysis of the products synthesized from starch
substrates with various amounts of branching (amylose and
amylopectin) revealed the importance of difference in loop
architecture, in particular of loops A1 and B, of GtfB-type α-
glucanotransferases for their substrate and product specificities.
Sequence analysis of 287 putative GtfB enzymes revealed an
almost bimodal distribution of loop lengths, with the majority
of GtfBs (e.g., Lr121 GtfB) having long A1 and B loops, while
Lr2613 GtfB represents a minority with much shorter loops,
and Lf2970 GtfB has intermediate loop lengths. The structure
of Lr2613 GtfB in complex with acarbose, occupying acceptor
substrate subsites (+1 to +3), allowed the identification of
residues that likely contribute to acceptor binding. Interest-
ingly, the type of residue at these specific positions correlated
with the length of loops A1 and B. Together, our results
suggest that GH70 GtfB enzymes with a more open
architecture are evolutionary closer to starch-degrading
GH13 α-amylases, while those with a tunneled binding groove
have evolved away. This extends our structural knowledge of
GH70 GT enzymes regarding their reaction specificity and
may accelerate the discovery and application of these enzymes
for starch modification aimed at the synthesis of α-glucan food
ingredients (with a low glycemic index).
■MATERIALS AND METHODS
Sequence Analysis. A BLAST search was performed in the non-
redundant protein sequences database, using the sequence of Lr2613
GtfB (GenBank ASA47879.1). The resulting hits were aligned using
MUSCLE (default parameters) within Jalview 2.
27
The length of loop
A2 (corresponding to residues 1086−1096 in Lr2613 GtfB) was used
as a criterion to remove glucansucrases from the results, which have a
longer loop A2 blocking donor subsites beyond −1.
16,28
Sequences
not containing a full GH70 core (domains A + B + C) were also
omitted. The final set of putative GtfB sequences was numbered (1−
287), and the variation in length of loops A1 and B was determined.
Loop A1 was defined as the segment between the two α-helices that
form the subdomain between β-strand 4 and α-helix 5 of the (βα)8-
barrel in Lr2613 GtfB (residues 802−811); loop B comprises the loop
residues between the short α-helix and first β-strand in domain B
(residues 590−593). In addition to these two loops, the sequence
conservation of six residues (Y632, L635, H683, Y719, R720, and
N792 in Lr2613 GtfB) near acceptor subsites was analyzed, using the
WebLogo server
29
for visualization. Furthermore, a phylogenetic tree
was constructed for the 287 sequences using MEGA X,
30
the
maximum likelihood method based on the JTT matrix model, with
partial deletion of positions containing gaps and missing data;
31
the
bootstrap consensus tree was inferred from 1000 bootstrap
replicates.
32
Each entry was annotated with the combined number
of residues in loops A1 and B.
Expression and Purification. For the N-terminally truncated L.
reuteri NCC 2613 GtfB 4,6-α-GTase protein, plasmid pET15b/gtf B-
ΔN, constructed in a previous study and encoding amino acids 417−
1281 of the protein, fused with an N-terminal His6-tag cleavable by a
3C protease, was used for recombinant gene expression
6
using
Escherichia coli BL21 star (DE3) as a host. An overnight culture of E.
coli BL21 Star (DE3) containing the relevant plasmid was diluted to
1/100 in Luria broth with 100 μgmL
−1ampicillin and propagated to
A600 nm 0.4−0.6. Gene expression was induced by adding isopropyl-
β-D-thiogalactopyranoside (IPTG) at a final concentration of 0.1
mM, and cultivation was continued at 16 °C for 20 h. Cells were
harvested by centrifugation (10,000g×20 min) and then disrupted
with a B-PER lysis reagent (Thermo Scientific, Pierce). Following
centrifugation (15,000g×20 min), the clear supernatants were
subjected to Ni-IMAC chromatography (Sigma-Aldrich). After
washing the column with 25 mM Tris−HCl (pH 8.0) and 1 mM
CaCl2, bound proteins were eluted with 200 mM imidazole in the
same buffer and the imidazole was removed by use of a stirred
ultrafiltration unit (Amicon, Beverly, MA) with a 30,000 molecular
mass cutoff. For further purification, the protein was subjected to
anion exchange chromatography on a Resource Q column (GE
Healthcare) using 25 mM HAc/NaAc pH 5.0, 150 mM NaCl, and 1
mM CaCl2with a 0−1 M NaCl gradient. Purity and homogeneity
were analyzed by SDS-PAGE and dynamic light scattering (DynaPro
Nanostar, Wyatt Technology, Santa Barbara, USA); the protein
concentration was determined using a Nanodrop 2000 spectropho-
tometer (Isogen Life Science, De Meern, The Netherlands). Lr121
GtfB and Lf2970 GtfB were produced as previously described.
7,16
Substrate Specificity. Amylose V from potato starch (Mw≈2×
105kDa) (AVEBE, Foxhol, The Netherlands) and amylopectin from
potato starch (Mw≈5×106Da) (Sigma-Aldrich) were evaluated as
substrates for the different GtfB enzymes. Amylose V (2%, w/v) was
prepared as a stock solution in sodium hydroxide (1 M), and
amylopectin and maltodextrins (1%, w/v) were solubilized by heating
to 55 °C in Milli-Q H2O. Before incubation, the amylose V stock
solution was neutralized with 7 M HCl and diluted to a concentration
of 1% (w/v). Unless otherwise stated, the different starch-like
substrates were individually incubated with 5 μgmL
−1of Lr121 GtfB,
Lr2613 GtfB, and Lf2970 GtfB at a concentration of 0.6% (w/v). All
reactions were performed in 25 mM sodium acetate buffer, pH 5.0
with 1 mM CaCl2at 37 °C for 72 h. The reactions were stopped by
incubation at 90 °C for 10 min. The products were analyzed by high-
performance size exclusion chromatography (HPSEC), proton
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
B
nuclear magnetic resonance (1H NMR) spectroscopy, and methyl-
ation analysis as described in the Supporting Information.
Inhibitory Effect of Acarbose on the Enzyme Activity of
Lr2613 GtfB. The initial total activity of Lr2613 GtfB was assayed by
the amylose-iodine staining method as described before.
4,17
The
reaction mixture contained 0.125% (w/v) amylose V (AVEBE,
Foxhol, The Netherlands), 2−100 μgml
−1(0.02−1.0 nM) of enzyme
in 25 mM sodium acetate (pH 5.0) and 1 mM CaCl2. The decrease in
absorbance of the α-glucan-iodine complex resulting from trans-
glycosylation and/or hydrolytic activity was monitored at 660 nm for
8 min at 40 °C. One unit of activity was defined as the amount of
enzyme converting 1 mg of substrate per minute. The inhibitory effect
of acarbose (Serva Electrophoresis GmbH, Heidelberg, Germany) on
the initial total enzyme activity was measured by adding 1.0 mM
acarbose to the reaction mixture (the acarbose concentration used did
not significantly affect the amylose-iodine staining assay). The initial
activity of Lr2613 GtfB on amylose was decreased by 62% in the
presence of acarbose. Incubation of the Lr2613 GtfB with acarbose
revealed that this enzyme is not able to use acarbose as a substrate
(data not shown).
HPSEC Analysis of Products. The molecular mass distribution of
the product mixtures was determined by high-performance size-
exclusion chromatography (HPSEC) with multi-detection as
described previously.
17
The HPSEC system (Agilent Technologies
1260 Infinity) was equipped with a multi-angle laser light scattering
detector (SLD 7000 PSS, Mainz), a viscometer (ETA-2010 PSS,
Mainz), and a differential refractive index detector (G1362A 1260
RID Agilent Technologies). Separation was performed by using three
PFG-SEC columns with porosities of 100, 300, and 4000 Å, coupled
with a PFG guard column. DMSO-LiBr (0.05 M) was used as eluent
at a flow rate of 0.5 mL min−1. The system was calibrated and
validated using a standard pullulan kit (PSS, Mainz, Germany) with
Mwranging from 342 to 708,000 Da. The specific RI increment value
dn/dcwas also measured by PSS and was 0.072 mL g−1(private
communication with PSS). The multi-angle laser light scattering
signal was used to determine the molecular mass of the high
molecular mass products (>1 ×105Da). The specific RI increment
value, dn/dc, for these polysaccharides in this system was taken to be
the same as for pullulan. The molecular mass of the low molecular
mass products (<1 ×105Da) was determined by a universal
calibration method. WinGPC Unity software (PSS, Mainz, Germany)
was used for data processing.
NMR Spectroscopy. One-dimensional 1H nuclear magnetic
resonance (NMR) spectra were recorded on a Varian Inova 500
spectrometer (NMR Center, University of Groningen), using D2Oas
solvent and at a probe temperature of 298 K. Before the analysis,
samples were exchanged twice in D2O (99.9 atom % D, Cambridge
Isotope Laboratories, Inc., Andover, MA) with intermediate
lyophilization and then dissolved in 0.6 mL of D2O. One-dimensional
500-MHz 1H NMR spectra were recorded at a 4000 Hz spectral width
and 16 k complex points, using a WET1D pulse to suppress the HOD
signal. All NMR spectra were processed with MestReNova 10.0.2
(Mestrelabs Research SL, Santiago de Compostella, Spain). Manual
phase correction was performed and a Whittacker smoother baseline
correction was applied. Chemical shifts (δ) were expressed in ppm
and calibrated with the internal standard acetone (δ, 2.225 ppm). The
percentage of different linkages was estimated by integration of the
respective signal peak areas.
6,7,33
Methylation Analysis. Methylation analysis was performed as
described before.
34
Briefly, the carbohydrate samples (∼5 mg) were
per-methylated using CH3IandsolidNaOHinDMSOand
subsequently hydrolyzed with trifluoroacetic acid (2 h, 120 °C).
The partially methylated monosaccharides generated were reduced
with NaBD4(2 h, room temperature, aqueous solution). The solution
was neutralized with acetic acid and boric acid was removed by co-
evaporation with methanol. The resulting partially methylated alditols
were per-acetylated using pyridine:acetic anhydride (1:1 v/v) at 120
°C yielding mixtures of partially methylated alditol acetates (PMAAs).
PMAAs were analyzed by gas−liquid chromatography (GLC) coupled
to electron impact mass spectrometry (EI-MS) and GLC coupled to
flame ioniziation detection (FID) as previously described.
34
Crystallization and Data Collection. Crystals of Lr2613 GtfB
were obtained in hanging drop vapor diffusion experiments at 293 K,
using either streakseeding or macroseeding methods. Drops for
streakseeding experiments consisted of equal volumes of protein
solution (7 mg ml−1in 25 mM HAc/NaAc, pH 5.0, 150 mM NaCl,
and 1 mM CaCl2) and reservoir solution (1.4−1.7 M (NH4)2SO4and
0.1 M Bis-Tris−HCl, pH 5.5) and were equilibrated against reservoir
solution. The drops were streakseeded with a cat whisker 22 h after
set up, from previously grown crystals. Drops for macroseeding
experiments consisted of equal volumes of the same protein solution
and 2.7−3.1 M (NH4)2SO4, 75 mM Bis-Tris−HCl, pH 5.5, 25 mM
HAc/NaAc, and 150 mM NaCl; they were equilibrated against
reservoir solution and macroseeded immediately after set up. Crystals
using either method appeared as thin plates and were often
intergrown; for data collection, they were stabilized in 1.8 M
(NH4)2SO4, 20 mM HAc/NaAc, pH 5.0, and 75 mM NaCl and
cryoprotected by including 25% (v/v) glycerol. Acarbose-complexed
crystals were obtained by including 20 or 15 mM acarbose (Serva
Electrophoresis GmbH, Heidelberg, Germany) in the stabilization
and cryoprotectant solutions, respectively. Diffraction data were
collected at beamlines P14 (native) and P11 (acarbose complex) of
DESY (Hamburg, Germany) and indexed, integrated, and scaled
using XDS;
35
statistics are given in Table 1.
Table 1. Crystallographic Data Collection and Refinement
Statistics
parameters native acarbose complex
PDB entry 7P38 7P39
data collection
space group P212121P212121
cell dimensions a,
b,c(Å) 107.4, 134.5, 147.9 107.2, 133.8, 147.9
resolution (Å) 47.24−2.70 (2.77−
2.70) 99.21−2.90 (2.98−
2.90)
Rpim 0.130 (0.606) 0.141 (0.551)
⟨I/σ⟩6.1 (1.7) 3.6 (1.1)
completeness
(%)
a
99.9 (99.0) 98.3 (98.9)
redundancy
a
8.2 (8.3) 9.1 (9.0)
refinement
resolution (Å) 47.24−2.70 (2.77−
2.70) 99.21−2.90 (2.98−
2.90)
unique
observations
a
56,749 (4107) 44,476 (3254)
R/Rfree 0.260 (0.299) 0.293 (0.315)
number of atoms
protein 13,001 13,022
Ca2+/waters 2/96 2/19
carbohydrate ligands acarbose
other ligand molecules glycerol (1), sulfate
ion (2) sulfate ion (1)
B-factors
protein (Å2) 39.4 (molecule A),
36.9 (molecule B) 44.5 (molecule A),
42.3 (molecule B)
carbohydrate (Å2) 49.1, 46.1 (acarbose)
root-mean-square
deviations
bond lengths (Å) 0.007 0.007
bond angles (°) 1.16 1.14
Ramachandran
favored (%) 93.5 95.3
allowed (%) 5.6 3.9
outliers (%) 0.9 0.7
a
In the highest resolution shell.
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
C
Structure Determination and Refinement of Lr2613 GtfB.
The crystal structure of the Lr2613 GtfB construct was determined by
molecular replacement using PHASER;
36
a previously obtained
homology model
6
based on the crystal structure of L. reuteri 121
GtfB(-ΔNΔV)
16
was used as the search model. The asymmetric unit
of the P212121cell contains two molecules. Models of the native and
acarbose-soaked structures were refined with Refmac
37
using non-
crystallographic symmetry restraints, alternated with inspection of the
electron density and manual rebuilding with Coot.
38
The PDB-REDO
server
39
was used in the final refinement stages. The final refinement
statistics and model quality are listed in Table 1. Structural figures
were prepared with PyMOL (The PyMOL Molecular Graphics
System, Version 2.0 Schrödinger, LLC) or UCSF ChimeraX.
40
Atomic coordinates and structure factors have been deposited at the
Protein Data Bank with entries 7P38 (native) and 7P39 (acarbose-
bound).
Homology Modeling of Lf2970 GtfB. The one-to-one Phyre2
protocol
41
was used to construct a homology model of Lf2970 GtfB,
with the crystal structure of Lr2613 GtfB as the template. Residues
898−901 (in loop B) and 1114−1116 (in loop A1) were added based
on the result from a normal Phyre2 modeling protocol.
Modeling Substrate Binding. To map the substrate binding
groove of all three enzymes, first the crystal structure of Lr2613 GtfB
in complex with acarbose (subsites +3 to −1) was superimposed with
that of the Lr121 GtfB−maltopentaose complex (subsites −1to−5)
(PDB: 5JBF).
16
Guided by this superposition, a maltooctaose (G8)
was then placed in subsites +3 to −5 of Lr2613 GtfB, optimizing the
substrate by adjusting glycosidic torsion angles while avoiding clashes
with the protein. By superposition, this procedure was repeated for
the crystal structure of Lr121 GtfB and the homology model of
Lf2970 GtfB. To model a branched substrate, an α-1,6-linked glucosyl
moiety was added at the sugar unit in subsite +1.
■RESULTS AND DISCUSSION
Lr2613 GtfB, Lf2970 GtfB, and Lr121 GtfB Have
Different Substrate Specificities. To study the substrate
specificity of Lr2613 GtfB and Lf2970 GtfB, the product
mixtures synthesized from amylose or amylopectin were
analyzed regarding linkage and molecular mass distribution
and compared with that of the Lr121 GtfB (Table S1). As
reported before,
18
Lr121 GtfB synthesizes a linear IMMP
(isomalto/maltopolysaccharides). While the IMMP derived
from amylose contained 80% of α-1,6 linkages, only 14% were
introduced in the amylopectin-derived products, indicating
that the presence of α-1,6 branching points in this substrate
limits α-1,6 transglycosylating activity. In contrast, Lr2613
GtfB synthesized branched reuteran-like products with similar
linkage distributions for α-1,4 and α-1,6 (79 and 21%,
respectively) regardless of the substrate used. Finally, for
Lf2970 GtfB, we analyzed the products by methylation
analysis, as the anomeric signals of the α-1,3 and α-1,4 linked
α-D-Glcpresidues partially overlap. In the product mixture
synthesized from amylose, terminal, 3-substituted, 4-substi-
tuted, 3,4-disubstituted, and 4,6-disubstituted glucopyranose
residues are present in molar percentages of 18, 12, 59, 9, and
2%, respectively. In contrast, the relative amounts for the
amylopectin-derived products were 22, 2, 69, 1, and 6%,
respectively.
HPSEC analysis (Figure S1) of the products synthesized
from amylose or amylopectin by the three enzymes supports
the linkage distribution analyses. In the case of Lr121 GtfB, the
amylose-derived products have a different molecular mass
distribution than the amylopectin-derived ones, eluting as a
narrow peak at ∼26 mL (corresponding to a low molecular
polymer with an average Mwof ∼15 ×103Da), or with a
broader molecular mass distribution around the same Mw,
respectively. This HPSEC profile, together with the low
amount of α-1,6 linkages detected by 1H NMR in the
amylopectin-derived products (Table S1), fits with previous
results reported by Bai et al.
18
showing that this enzyme was
only capable of modifying the linear side chains of amylopectin
in wheat starches. In contrast, Lr2613 GtfB products from
amylose or amylopectin showed a rather similar molecular
mass distribution. The main peak at an elution volume of ∼29
mL corresponds to a low molecular mass polymer with an
average Mwof ∼5.6 ×103Da, together with a small shoulder
peak corresponding to maltose. Finally, for Lf2970 GtfB, the
molecular mass distribution of synthesized products clearly
depends on the substrate given. From amylose, Lf2970 GtfB
gave a bimodal molecular mass distribution containing two
main peaks corresponding to a high molecular mass polymer
with an average Mwof 26 ×106Da, and oligosaccharides.
From amylopectin, however, the synthesized α-glucans showed
more complex elution patterns because of their high
polydispersity indices. This suggests that Lf2970 GtfB
efficiently cleaves the α-1,4 linkages present in amylopectin,
while the presence of branching points limits the synthesis of
new α-1,3 linkages leading to polymer formation. As a result, a
non-uniform poly-/oligosaccharide mixture is produced from
amylopectin.
Crystal Structure of Lr2613 GtfB. The native crystal
structure of Lr2613 GtfB was determined at a resolution of 2.9
Å. The asymmetric unit of the Lr2613 GtfB crystals contains
two molecules (A and B), comprising residues 446−1277 and
445−1277, which can be superimposed with a root-mean-
square difference (rmsd) of 0.11 Å on Cαatoms. Dynamic
light scattering analysis clearly indicated a monodisperse
solution containing a monomeric protein (apparent molecular
mass 101 kDa). Additional weak electron density, likely
representing extra N-terminal residues of Lr2613 GtfB, was not
sufficient to include these in the model; also, the N-terminal
His-tag was not visible in electron density. In the native
structure, residues 805−807 of molecule B were omitted
because of weak electron density. Superposition with the
structure of Lr121 GtfB
16
gave an rmsd of 0.42 Å (for 574 Cα
atoms), reflecting the high sequence identity (79.4% for the
segments visible in the structures). In fact, these two GtfB
structures are very similar, except for some loop regions (as
will be discussed below). Lr2613 GtfB has the GH70-like
domain arrangement;
28
the crystal structure comprises the
three core domains A, B, and C, and the auxiliary domain IV
can thus be considered a ΔNΔV-structure (Figure 1).
Residues preceding the first visible residue (V446), constitut-
ing domain V and N-terminal variable domain (N), likely
extend away from domain IV. Like in other GH70 enzymes,
the catalytic domain A is circularly permuted, such that the
order of the four homology motifs is II-III-IV-I, differing from
the GH13 I-II-III-IV homology motif order.
24,28
The catalytic site of Lr2613 GtfB is located at the interface
of domains A and B with the three catalytic residues D679
(nucleophile), E717 (acid/base), and D788 (transition-state
stabilizing residue) lining a pocket that lies halfway a groove
running along the domain interface; four other residues (R677,
H787, D1135, and Q1140) that are conserved in GH70
enzymes also surround the conserved subsite −1(Figure S2).
Of these seven residues, six are also conserved in GH13 α-
amylases, while the glutamine Q1140 replaces a histidine that
is present at the corresponding position in α-amylases. Below
subsite −1 is the side chain of Y1095, known to provide an
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
D
aromatic stacking interaction with sugar moieties in this
subsite.
24
Close to the active site is a conserved Ca2+binding
site also found in other GH70 enzyme structures.
Most notably, the Lr2613 GtfB structure shows an open and
fully accessible binding groove. In agreement with earlier
modeling,
6
the short loops A1 (residues 802−811) and B
(residues 590−593) do not form a tunnel; details and
comparison with other GtfB structures are discussed below
(section “GtfB Enzymes Display Varying A1 and B Loop
Lengths and Binding Groove Architectures”).
The Lr2613 GtfB−Acarbose Complex Reveals Accept-
or Binding Subsites. Acarbose is a known inhibitor of GH13
and GH70 enzymes and the inhibitory effect on Lr2613 GtfB
activity was confirmed (Supporting Information). Soaking the
crystals with acarbose resulted in binding of the pseudote-
trasaccharide in subsites −1to+3(Figure 2). The valeinamine
moiety at the non-reducing end of the inhibitor is bound in
subsite −1 where, in case of a natural substrate (e.g., a
maltooligosaccharide fragment), the covalent enzyme−glucosyl
intermediate is formed in the first half-reaction. The valein-
amine moiety has a distorted 2H3conformation, mimicking the
transition state,
42
with the non-cleavable N-glycosidic linkage
oriented toward the catalytic acid/base residue E717. A total of
seven hydrogen bonds to the valeinamine moiety and its N-
glycosidic linkage are provided by two of the three catalytic
residues (acid/base E717 and D788) as well as the strictly
conserved H787. The orientation of this sugar moiety is
further stabilized by an aromatic stacking interaction with a
Figure 1. Crystal structure of Lr2613 GtfB-ΔNΔV with the four
domains A, B, C, and IV indicated. The active site is at the interface of
domains A and B, with the three catalytic residues (D679, E717, and
D788) shown in stick representation. The figure is prepared with
PyMOL.
Figure 2. (a) Stereo view of the Lr2613 GtfB−acarbose complex, with the pseudotetrasaccharide inhibitor occupying donor subsite −1 and
acceptor subsites +1 to +3. Residues close to acarbose are shown with side chains (catalytic residues D679, E717, and D788 are indicated with bold
labels); hydrogen bond interactions are shown as dotted lines. The figure is prepared with PyMOL. (b) Chemical structure of acarbose.
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
E
conserved tyrosine residue (Y1095). The remaining three
sugar moieties occupy subsites +1 to +3 and, for the first time,
map these three acceptor substrate subsites in a GH70 GtfB-
type enzyme. Surrounding these subsites, there are six residues
(apart from the catalytic residues) closer than 4 Å from the
bound inhibitor, shaping the binding site and providing a
hydrogen bond or aromatic stacking interactions at subsites +1
and + 2; they will be discussed below. No interactions were
observed at subsite +3, although we cannot exclude that water-
mediated hydrogen bonds are present, which are unresolved at
2.9 Å resolution.
Improved Homology Model of Lf2970 GtfB. The
previously reported homology model of Lf2970 GtfB
7
was
based on the crystal structure of glucansucrase Gtf180-ΔN
from L. reuteri 180 (sequence identity: 38.6% for segments
visible in the structure). Because of the much higher sequence
identity with Lr2613 GtfB (93.7%), the homology model of
Lf2970 GtfB presented here is more accurate. The new model
comprises residues 752−1589 of Lf2970 GtfB; with a virtually
identical chain trace, superposition with molecule A of the
Lr2613 GtfB crystal structure resulted in an rmsd of only 0.13
Å (for 750 Cαatoms). The most notable differences occur in
loops A1 and B; how the Lf2970 GtfB binding groove
compares to that of the two other GtfB enzymes will be
reviewed below.
GtfB Enzymes Display Varying A1 and B Loop
Lengths and Binding Groove Architectures. Comparing
the two crystal structures (Lr2613 GtfB and Lr121 GtfB) and
an improved homology model (Lf2970 GtfB) highlights the
differences in the architecture of the binding groove and the
importance of loops A1 and B for accessibility, as is shown in
Figure 3. Together with two other surface loops (around D601
and N625 (Lr2613 GtfB numbering), too far from the active
site to be involved in substrate binding), loops A1 and B map
to gaps in the sequence alignment. In Lr2613 GtfB, these loops
are much shorter than in Lr121 GtfB, namely, by 7 and 16
Figure 3. (a−c) Comparison of loops A1 (purple), A2 (red), and B (brown) in the crystal structure of Lr2613 GtfB (this study), the homology
model of Lf2970 GtfB (this study), and the crystal structure of Lr121 GtfB.
16
The three loops are labeled with their lengths; in the first two GtfBs,
amino acid residues at the tip of loops A1 and B are indicated. The catalytic residues (D679, E717, and D788) are shown in stick representation
(lower left). In Lr2613 GtfB and Lf2970 GtfB, loops A1 and B are relatively short, leading to a (half-)open binding groove architecture; in contrast,
in Lr121 GtfB the corresponding segments are long and cover donor subsites. The figure is prepared with PyMOL.
Figure 4. Sequence alignment of regions in Lr2613 GtfB, Lf2970 GtfB, and Lr121 GtfB in domains A (blue header) and B (green header) in which
residues closer than 4 Å to a modeled substrate are shown below the alignment with the corresponding subsite. The six acceptor subsite residues
discussed in the text are highlighted in yellow. The three catalytic residues (D679, E717, and D788) are indicated with a black triangle (NU =
nucleophile, A/B = general acid/base, TS = transition-state stabilizing residue). Color coding of the loops B, A1, and A2 corresponds to the scheme
used in Figure 3. The figure is prepared using the ESPript 3.0 server (https://espript.ibcp.fr).
45
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
F
residues, respectively (Figures 3a and 4). As a result, and in
stark contrast to Lr121 GtfB (Figure 3c), loop B hardly
protrudes from the surface; in fact, only a non-conserved
residue (Y592) does so. Also loop A1 (residues H802−Y810)
protrudes considerably less from the surface than in Lr121
GtfB.
As a result, and as was previously predicted from sequence
alignment and homology modeling,
6
the crystal structure of
Lr2613 GtfB now clearly shows that the shorter loops A1 and
B do not form a tunnel (the hydroxyl groups of residues Y806
and Y592 at their tips are about 9 Å apart); this creates a much
more open architecture. It has to be noted that the electron
density for residues K800−A811 is weak and/or ambiguous in
both molecules, suggesting possible flexibility of loop A1.
Nevertheless, the active site groove of Lr2613 GtfB is fully
accessible, reminiscent of the situation in GH13 starch-
degrading α-amylases.
43
In Lf2970 GtfB, loops A1 and B are
somewhat longer than in Lr2613 GtfB (by three and four
residues, respectively) and loop B protrudes more from the
surface (Figure 3b). Still, the residues at the tips (S900 and
G1115) are at such a distance that Lf2970 GtfB displays a
(half-)open architecture.
In order to extend our analysis of loop length beyond these
three enzymes to the entire GtfB subfamily, we performed a
BLAST search with the core domains (A, B, and C) of Lr121
GtfB. This yielded 287 putative GtfB enzyme sequences
(Table S2), including the three enzymes described in our
study; the vast majority of these (280) are from Lactobacilla-
ceae, while the rest are from Weissella (5), Leuconostoc (2), or
Streptococcus (1) species. The sequence identity of the hits
ranged from 99.9 to 44.2%. Given the importance of loops A1
and B in determining the accessibility of the substrate binding
groove (especially near donor subsites) and knowing that in
characterized GtfB enzymes their lengths differ, we focused on
the regions corresponding to these loops. Indeed, sequence
analysis and alignment of loops A1 and B in 287 putative GtfB
enzymes revealed a large variation in the length of these
elements with a rather distinct distribution (Figure 5); within
groups of enzymes having similar loop lengths, the loop
sequences are highly conserved (not shown).
The most common length for loop A1 is 17 residues (78.4%
of cases), while most of the other sequences feature a much
shorter loop A1. The length of loop B varies between 4 and 20
residues, again with the majority (80.1% of cases) having the
long version. The distribution of total loop length (A1 + B) is
similar, with two distinct groups: 78.4% (225 sequences) of the
enzymes feature long loops (37 or 40 residues combined),
while most of the remaining ones have shorter loops (17.8%
have a combined length of 12−14 residues). The three
enzymes that we compared structurally represent GtfBs with
(almost) the shortest (Lr2613 GtfB: 14 residues), intermediate
(Lf2970 GtfB: 21 residues), and (almost) the longest
combined loop lengths (Lr121 GtfB: 37 residues). The small
subset with intermediate loop lengths includes S. thermophilus
GtfB (combined loop length of 21 residues). A phylogenetic
tree (Figure S3) constructed from the 287 putative GtfB
sequences, annotated with the total loop length (A1 + B),
revealed that those with short loops cluster together in two
opposite segments, one mostly containing Lactiplantibacillus
plantarum or Fructilactobacillus sanfranciscensis species, and the
other mainly containing Limosilactobacillus reuteri and other
species (e.g., Limosilactobacillus fermentum).
Modeling Substrate Binding. The first step of the
reaction catalyzed by GTs is donor substrate binding; we
combined structural data from two GtfB enzymes to gain more
insights into the molecular details. Superposition of the Lr2613
GtfB−acarbose complex with the Lr121 GtfB−maltopentaose
complex (PDB: 5JBF
16
) showed that the (pseudo)sugar units
in subsite −1 have a similar orientation and almost all
hydrogen bond interactions are conserved. This allowed us to
construct a model of maltooctaose (G8) occupying subsites −5
to +3 in Lr2613 GtfB (Figure 6b); only minor adjustments of
glycosidic torsion angles were required to fit the oligosacchar-
ide substrate in the binding groove. Similarly, we modeled
maltooctaose in the homology model of Lf2970 GtfB (Figure
6c) and in the crystal structure of Lr121 GtfB (Figure 6d),
visualizing how linear α-1,4-linked maltooligosaccharide seg-
ments of starch substrates can be bound. Comparing our
models with the crystal structure of the α-amylase fromBacillus
halmapalus complexed with a pseudo-nonasaccharide (PDB:
1W9X)
43
(Figure 6a) showed that the global binding mode of
a maltooligosaccharide substrate in GH70 GtfBs is similar to
that observed in this α-amylase.
After modeling a bound linear donor substrate in GtfB-type
proteins, we investigated the possibility to model α-1,4/6
branched ones. As described above, the open architecture of
the binding groove of Lr2613 GtfB differs considerably from
that of the tunneled one observed in Lr121 GtfB. The (linear)
G8 models show that the positions of the O6 atoms of the
glucosyl moieties in several subsites are such that a branched
structure would fit without sterical hindrance in Lr2613 GtfB
but not in Lr121 GtfB. Therefore, analogous to the observed
binding mode of a branched oligosaccharide in the GH13
pullulanase SusG (PDB: 6BS6
44
), we modeled a branch
glucosyl unit linked to the O6 at subsite +1 (Figure S4),
showing that the branch unit may interact with residue N792
in homology motif IV, the fourth residue after the transition-
state stabilizing aspartate (D788), and almost fully conserved
in the GtfB subfamily (Figure 7). In SusG, such an interaction
is indeed experimentally observed with the corresponding
residue K541. The orientation of the branch unit is such that
the branch can be extended via α-1,4 linkages without sterical
hindrance. In Lf2970 GtfB, the situation is very similar since
Figure 5. Variation in the length of loops A1 (purple) and B (brown)
in 287 putative GH70 GtfB enzymes. The groups in which Lr2613-,
Lf2970-, and Lr121-GtfB occur are indicated (e.g., Lr2613 GtfB has a
four-residue-long loop B and a 10-residue-long loop A1).
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
G
the only main differences (somewhat longer loops A1 and B)
are far from subsite +1. In contrast, as is clear from Figure S4,
in Lr121 GtfB, residues in the region S918−D922 of loop B
would clash with branch sugar units in subsite +1′, hindering
the binding of branched starch substrates.
Acceptor Subsite Residues Correlate with Loop
Lengths. In addition to the architecture of the binding
groove, residues surrounding acceptor subsites are of
importance for the specificity of the transglycosylation
reaction, since they determine the orientation of bound
acceptor molecules after formation of the covalent glycosyl−
enyme intermediate. The Lr2613 GtfB−acarbose complex as
well as the modeled G8 substrate complex in all three enzymes
show that at subsites +1 to +3, there are six residues (apart
from the catalytic residues) that are at less than 4 Å distance
from the sugar moieties (Figures 2 and 4). Four of these
residues are from homology motifs II, III, and IV: residues
H683, R720, and N792 provide hydrogen bond interactions
with hydroxyl groups of (pseudo)sugar moieties in subsites +1,
+1′(in the case of a branched substrate), and +2, while Y719
in motif III provides an aromatic stacking interaction at subsite
+2. Finally, residues from the 630’s loop in domain B,
especially Y632 and L635, do not directly interact with the
inhibitor but they help shape the binding site. Analysis of the
conservation of these six residues in 287 putative GtfB
sequences revealed that those corresponding to positions 635,
719, and 792 (Lr2613 GtfB numbering) are almost fully
conserved, while the others show strong preference for a
certain residue type, remarkably correlating with the loop
length (Figure 7). Sequences with long loops A1 and B almost
invariantly have P, L, N, Y, H, and N (e.g., Lr121 GtfB is
representative of this group), while in sequences with short or
intermediate loop lengths, the first, third, and fifth residues are
often Y, H, and R/N, respectively. The residues at these latter
positions (632, 683, and 720 in Lr2613 GtfB) cluster together
near subsites +1 and +2, and one may speculate that a different
set of residues (with different hydrogen bond capabilities) is
needed to process starch(-like) substrates in a more open
Figure 6. (b−d) Modeled maltooctaose shown as spheres in subsites +3 to −5 of Lr2613 GtfB, Lf2970 GtfB, and Lr121 GtfB, compared to (a) the
structure of B. halmapalus α-amylase with a bound nonasaccharide.
43
With shorter loops A1 and B, Lr2613 GtfB and Lf2970 GtfB feature a more
open architecture like in the α-amylase, compared to the tunneled one in Lr121 GtfB. The figure is prepared with UCSF ChimeraX.
40
Figure 7. Sequence conservation of six residues near acceptor subsites +1 and + 2 in 287 putative GH70 GtfB enzymes and GH13 α-amylases. The
GtfB enzymes are divided in 57 sequences with short or intermediate combined loop lengths (12−21 residues) and 230 sequences with long
lengths (31−40 residues). In α-amylases, the six corresponding residues are fully conserved. Below the logos, the residues present in B. licheniformis
α-amylase, Lr2613 GtfB, Lf2970 GtfB, and Lr121 GtfB are shown; the three residues in Lf2970 GtfB that are unique are indicated with an asterisk.
The figure is based on use of the WebLogo server.
29
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
H
binding groove. In fact, a superposition with bacterial α-
amylase structures (not shown) revealed that the correspond-
ing residues are invariably Y, H, and K. Together, these
observations suggest that the GH70 GtfB subgroup with short
loops are evolutionary closer to GH13 α-amylases, while the
large subgroup with long loops have evolved further away. The
GtfB subgroup with open architecture (short loops A1 and B)
therefore provides an opportunity to study how they gained
transglycosylation capabilities (most α-amylases are hydro-
lytic).
The six aforementioned residues likely also contribute to
transglycosylation linkage specificity. Most of the GtfB
enzymes characterized so far display α-4,6 specificity,
3,6,9
but
Lf2970 GtfB is the only known GT with α-4,3 trans-
glycosylation specificity.
7
Notably, within our set of 287
putative GtfB sequences, Lf2970 GtfB has unique substitutions
at the second, third, and fifth positions (V943, D991, and
G1028) (Figure 7); one may speculate that they play a key role
in determining the orientation of acceptor substrates and thus
the type of glycosidic linkage in the product. Preliminary
modeling of acceptor substrates complexes in Lr2613 GtfB and
Lf2970 GtfB so far did not give conclusive insight into this
regard nor did single mutations targeting all six positions in
Lr2613 GtfB affect the transglycosylation linkage specificity
(personal communication). Possibly, simultaneous mutation of
several residues is needed to achieve such changes.
In addition to transglycosylation linkage specificity, the
acceptor subsite residues may also play a role in determining
the branching characteristics of the products. An interesting
comparison in this regard is the pair of GtfB-type enzymes
GtfY and GtfX from Ligilactobacillus aviarius subsp. aviarius
DSM 20655 of which ΔNΔC-constructs were characterized by
Meng et al.
9
While GtfY was shown to synthesize linear
(IMMP) products from amylose V, the GtfX product is a
branched reuteran. Notably, within our set of 287 putative
GtfBs, GtfY belongs to the majority with P, L, N, Y, H, and N
in the six positions near acceptor subsites, while GtfX belongs
to a small set of seven sequences (all from L. aviarius species)
having M, Y, N, W, D, and D at these positions (Figure 7).
This would suggest a role of these residues in facilitating
branching in GtfX, even more so because both enzymes feature
long loops A1 and B (17 and 20 residues, respectively (Table
S2)). It has to be noted that one would expect the long loops
to sterically hinder branch formation at the subsite +1 sugar
moiety just like in Lr121 GtfB. In any case, at least for this
small set of L. aviarius enzymes, loop architecture does not
entirely define product specificity.
Implications for GtfB Reaction Specificity. Together,
the observed differences between the Lr2613-, Lf2970-, and
Lr121 GtfB structures further support the hypothesis stated by
Bai et al.
16
that the evolution of GH13 α-amylases toward
GH70 α-glucanotransferases involved changes in loop
architecture that conferred changes in starch substrate
preference (and product specificity). For these three enzymes,
their binding groove architectures are in agreement with
experimental biochemical observations; while Lr121 GtfB has a
profound preference for linear starch substrates, Lr2613 GtfB
and Lf2970 process both linear and branched ones, albeit that
with the latter enzyme, the α-glucan product size and linkage
distribution varies with the amount of branching in the
substrate (Table S1). Given the high sequence similarity of
loops A1 and B (within groups of enzymes having similar loop
lengths) and having analyzed the 3D structure of three
representative GtfB α-glucanotransferases, we hypothesize that
the majority (four out of five) of GtfB enzymes feature a
tunnel-like structure like Lr121 GtfB and thus are limited to
processing mostly linear starch-like structures. In contrast,
most other GtfB enzymes feature short loops A1 and B and can
be expected to have broader substrate specificity, allowing also
more branched starches to be processed. This needs to be
validated in the future by biochemical characterization of more
GtfB enzymes. It is remarkable that intermediate loop lengths
are hardly observed, and that GtfB enzymes with long loops A1
and B moreover feature a different set of acceptor subsite
residues. This suggests that most GH70 GtfBs have evolved
away from starch-degrading GH13 α-amylases, which have an
open binding groove, and perhaps have reached optimal loop
lengths (∼37 residues). It is yet unclear what was the
evolutionary drive for such changes, but it has been suggested
that a tunneled binding groove enhances the processivity of α-
glucan synthesis
16
by keeping intermediate reaction products
bound to the enzyme. Indeed, when comparing the product
sizes of Lr2613 GtfB (short loops) and Lr121 GtfB (long
loops), the latter synthesizes higher molecular mass products
from the same substrate (Figure S1). In any case, the quest for
GH70 α-glucanotransferases with broad starch substrate
specificity should focus on enzymes with shorter A1 and B
loops. The recent findings that Streptococcus thermophilus GtfB,
with loop lengths similar to Lf2970 GtfB, can be used in potato
starch and wheat starch conversion
11,12
supports this
hypothesis. We are currently investigating the suitability of
Lf2970- and Lr2613 GtfB for the modification of different food
starches, aiming to extend the GtfB toolkit for starch
modification.
■ASSOCIATED CONTENT
*
sıSupporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.jafc.1c05657.
(Table S1) Evaluation of different starch-like substrates
(amylose V and amylopectin) as donor substrates for
different GtfB enzymes, (Figure S1) distribution of the
molecular mass of products synthesized by Lr121 GtfB,
Lr2613 GtfB, and Lf2970 GtfB from amylose and
amylopectin. (Figure S2) stereo view of the Lr2613 GtfB
active site at the interface of domains A and B, (Figure
S3) phylogenetic tree of 287 putative GtfB sequences,
and (Figure S4) modeling of a branched maltooligo-
saccharide in the substrate binding sites of Lr2613 GtfB,
Lf2970 GtfB, and Lr121 GtfB. (Table S2) List of 287
putative GtfB enzymes (PDF)
■AUTHOR INFORMATION
Corresponding Author
Tjaard Pijning −Biomolecular X-ray Crystallography,
Groningen Biomolecular Sciences and Biotechnology Institute
(GBB), University of Groningen, Groningen 9747 AG, The
Netherlands; orcid.org/0000-0003-4107-3663;
Email: t.pijning@rug.nl
Authors
Joana Gangoiti −Microbial Physiology, Groningen
Biomolecular Sciences and Biotechnology Institute (GBB),
University of Groningen, Groningen 9747 AG, The
Netherlands
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
I
Evelien M. te Poele −Microbial Physiology, Groningen
Biomolecular Sciences and Biotechnology Institute (GBB),
University of Groningen, Groningen 9747 AG, The
Netherlands; CarbExplore Research B.V., Groningen 9747
AA, The Netherlands
Tim Börner −NestléResearch, Sociétédes Produits NestléSA,
1000 Lausanne, Switzerland; orcid.org/0000-0003-
1120-225X
Lubbert Dijkhuizen −Microbial Physiology, Groningen
Biomolecular Sciences and Biotechnology Institute (GBB),
University of Groningen, Groningen 9747 AG, The
Netherlands; CarbExplore Research B.V., Groningen 9747
AA, The Netherlands
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jafc.1c05657
Funding
This work was financially supported by Nestec Ltd. and by the
University of Groningen.
Notes
The authors declare no competing financial interest.
■ACKNOWLEDGMENTS
Parts of this research were carried out at PETRA III at DESY, a
member of the Helmholtz Association (HGF). We would like
to thank E. Reddem, A. Burkhardt, and G. Bourenkov for
assistance in using the P11 and P14 beamlines.
■REFERENCES
(1) Lombard, V.; Golaconda Ramulu, H.; Drula, E.; Coutinho, P.
M.; Henrissat, B. The carbohydrate-active enzymes database (CAZy)
in 2013. Nucleic Acids Res. 2014,42, D490−D495.
(2) Kralj, S.; Grijpstra, P.; van Leeuwen, S. S.; Leemhuis, H.;
Dobruchowska, J. M.; van der Kaaij, R. M.; Malik, A.; Oetari, A.;
Kamerling, J. P.; Dijkhuizen, L. 4,6-α-glucanotransferase, a novel
enzyme that structurally and functionally provides an evolutionary
link between Glycoside Hydrolase enzyme Families 13 and 70. Appl.
Environ. Microbiol. 2011,77, 8154−8163.
(3) Leemhuis, H.; Dijkman, W. P.; Dobruchowska, J. M.; Pijning, T.;
Grijpstra, P.; Kralj, S.; Kamerling, J. P.; Dijkhuizen, L. 4,6-α-
Glucanotransferase activity occurs more widespread in Lactobacillus
strains and constitutes a separate GH70 subfamily. Appl. Microbiol.
Biotechnol. 2013,97, 181−193.
(4) Gangoiti, J.; Pijning, T.; Dijkhuizen, L. The Exiguobacterium
sibiricum 255-15 GtfC enzyme represents a novel Glycoside
Hydrolase 70 subfamily of 4,6-α-glucanotransferase enzymes. Appl.
Environ. Microbiol. 2016,82, 756−766.
(5) Gangoiti, J.; van Leeuwen, S. S.; Vafiadi, C.; Dijkhuizen, L. The
Gram-negative bacterium Azotobacter chroococcum NCIMB 8003
employs a new glycoside hydrolase family 70 4,6-α-glucanotransferase
enzyme (GtfD) to synthesize a reuteran like polymer from
maltodextrins and starch. Biochim. Biophys. Acta 2016,1860, 1224−
1236.
(6) Gangoiti, J.; van Leeuwen, S. S.; Meng, X.; Duboux, S.; Vafiadi,
C.; Pijning, T.; Dijkhuizen, L. Mining novel starch-converting
Glycoside Hydrolase 70 enzymes from the NestléCulture Collection
genome database: The Lactobacillus reuteri NCC 2613 GtfB. Sci. Rep.
2017,7, 9947.
(7) Gangoiti, J.; van Leeuwen, S. S.; Gerwig, G. J.; Duboux, S.;
Vafiadi, C.; Pijning, T.; Dijkhuizen, L. 4,3-α-Glucanotransferase, a
novel reaction specificity in glycoside hydrolase family 70 and clan
GH-H. Sci. Rep. 2017,7, 39761.
(8) Gangoiti, J.; Lamothe, L.; van Leeuwen, S. S.; Vafiadi, C.;
Dijkhuizen, L. Characterization of the Paenibacillus beijingensis DSM
24997 GtfD and its glucan polymer products representing a new
glycoside hydrolase 70 subfamily of 4,6-α-glucanotransferase enzymes.
PLoS One 2017,12, No. e0172622.
(9) Meng, X.; Gangoiti, J.; de Kok, N.; van Leeuwen, S. S.; Pijning,
T.; Dijkhuizen, L. Biochemical characterization of two GH70 family
4,6-α-glucanotransferases with distinct product specificity from
Lactobacillus aviarius subsp. aviarius DSM 20655. Food Chem. 2018,
253, 236−246.
(10) İspirli, H.; Şimşek, Ö.; Skory, C.; Sağdıç, O.; Dertli, E.
Characterization of a 4,6-α-glucanotransferase from Lactobacillus
reuteri E81 and production of malto-oligosaccharides with immune-
modulatory roles. Int. J. Biol. Macromol. 2019,124, 1213−1219.
(11) Li, X.; Fei, T.; Wang, Y.; Zhao, Y.; Pan, Y.; Li, D. Wheat starch
with low retrogradation properties produced by modification of the
GtfB enzyme 4,6-α-glucanotransferase from Streptococcus thermophi-
lus.J. Agric. Food Chem. 2018,66, 3891−3898.
(12) Li, D.; Fu, X.; Mu, S.; Fei, T.; Zhao, Y.; Fu, J.; Lee, B. H.; Ma,
Y.; Zhao, J.; Hou, J.; Li, X.; Li, Z. Potato starch modified by
Streptococcus thermophilus GtfB enzyme has low viscoelastic and
slowly digestible properties. Int. J. Biol. Macromol. 2021,183, 1248−
1256.
(13) Meng, X.; Gangoiti, J.; Bai, Y.; Pijning, T.; Van Leeuwen, S. S.;
Dijkhuizen, L. Structure-function relationships of family GH70
glucansucrase and 4,6-α-glucanotransferase enzymes, and their
evolutionary relationships with family GH13 enzymes. Cell. Mol.
Life Sci. 2016,73, 2681−2706.
(14) Janeček, S̆.; Svensson, B.; MacGregor, E. A. α-Amylase: an
enzyme specificity found in various families of glycoside hydrolases.
Cell. Mol. Life Sci. 2014,71, 1149−1170.
(15) Miao, M.; Jiang, B.; Jin, Z.; BeMiller, J. N. Microbial starch-
converting enzymes: recent insights and perspectives. Compr. Rev.
Food Sci. Food Saf. 2015,17, 1238−1260.
(16) Bai, Y.; Gangoiti, J.; Dijkstra, B. W.; Dijkhuizen, L.; Pijning, T.
Crystal structure of 4,6-α-glucanotransferase supports diet-driven
evolution of GH70 enzymes from α-amylases in oral bacteria.
Structure 2017,25, 231−242.
(17) Bai, Y.; van der Kaaij, R. M.; Leemhuis, H.; Pijning, T.; van
Leeuwen, S. S.; Jin, Z.; Dijkhuizen, L. Biochemical Characterization of
the Lactobacillus reuteri Glycoside Hydrolase Family 70 GTFB type of
4,6-α-glucanotransferase enzymes that synthesize soluble dietary
starch fibers. Appl. Environ. Microbiol. 2015,81, 7223−7232.
(18) Bai, Y.; Boger, M.; van der Kaaij, R. M.; Woortman, A. J. J.;
Pijning, T.; van Leeuwen, S. S.; Lammerts van Bueren, A.; Dijkhuizen,
L. Lactobacillus reuteri strains convert starch and maltodextrins into
homoexopolysaccharides using an extracellular and cell-associated 4,6-
α-glucanotransferase. J. Agric. Food Chem. 2016,64, 2941−2952.
(19) van der Zaal, P. H.; Schols, H. A.; Bitter, J. H.; Buwalda, P. L.
Isomalto/malto-polysaccharide structure in relation to the structural
properties of starch substrates. Carbohydr. Polym. 2018,185, 179−
186.
(20) Leemhuis, H.; Dobruchowska, J. M.; Ebbelaar, M.; Faber, F.;
Buwalda, P. L.; van der Maarel, M. J. E. C.; Kamerling, J. P.;
Dijkhuizen, L. Isomalto/malto-polysaccharide, a novel soluble dietary
fiber made via enzymatic conversion of starch. J. Agric. Food Chem.
2014,62, 12034−12044.
(21) Gangoiti, J.; Pijning, T.; Dijkhuizen, L. Biotechnological
potential of novel glycoside hydrolase family 70 enzymes synthesizing
α-glucans from starch and sucrose. Biotechnol. Adv. 2018,36, 196−
207.
(22) Gangoiti, J.; Corwin, S. F.; Lamothe, L. M.; Vafiadi, C.;
Hamaker, B. R.; Dijkhuizen, L. Synthesis of novel α-glucans with
potential health benefits through controlled glucose release in the
human gastrointestinal tract. Crit. Rev. Food Sci. Nutr. 2020,60, 123−
146.
(23) Te Poele, E. M.; Corwin, S. G.; Hamaker, B. R.; Lamothe, L.
M.; Vafiadi, C.; Dijkhuizen, L. Development of slowly digestible
starch derived α-glucans with 4,6-α-glucanotransferase and branching
sucrase enzymes. J. Agric. Food Chem. 2020,68, 6664−6671.
(24) Vujičić-Žagar, A.; Pijning, T.; Kralj, S.; López, C. A.; Eeuwema,
W.; Dijkhuizen, L.; Dijkstra, B. W. Crystal structure of a 117 kDa
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
J
glucansucrase fragment provides insight into evolution and product
specificity of GH70 enzymes. Proc. Natl. Acad. Sci. U. S. A. 2010,107,
21406−21411.
(25) Ito, K.; Ito, S.; Shimamura, T.; Weyand, S.; Kawarasaki, Y.;
Misaka, T.; Abe, K.; Kobayashi, T.; Cameron, A. D.; Iwata, S. Crystal
structure of glucansucrase from the dental caries pathogen
Streptococcus mutans.J. Mol. Biol. 2011,408, 177−186.
(26) Meng, X.; Pijning, T.; Dobruchowska, J. M.; Gerwig, G. J.;
Dijkhuizen, L. Characterization of the functional roles of amino acid
residues in acceptor-binding subsite +1 in the active site of the
glucansucrase GTF180 from Lactobacillus reuteri 180. J. Biol. Chem.
2015,290, 30131−30141.
(27) Waterhouse, A. M.; Procter, J. B.; Martin, D. M. A.; Clamp, M.;
Barton, G. J. Jalview Version 2 - a multiple sequence alignment editor
and analysis workbench. Bioinformatics 2009,25, 1189−1191.
(28) Leemhuis, H.; Pijning, T.; Dobruchowska, J. M.; Van Leeuwen,
S. S.; Kralj, S.; Dijkstra, B. W.; Dijkhuizen, L. Glucansucrases: three-
dimensional structures, reactions, mechanism, α-glucan analysis and
their implications in biotechnology and food applications. J.
Biotechnol. 2013,163, 250−272.
(29) Crooks, G. E.; Hon, G.; Chandonia, J. M.; Brenner, S. E.
WebLogo: a sequence logo generator. Genome Res. 2004,14, 1188−
1190.
(30) Kumar, S.; Stecher, G.; Li, M.; Knyaz, C.; Tamura, K. MEGA
X: Molecular evolutionary genetics analysis across computing
platforms. Mol. Biol. Evol. 2018,35, 1547−1549.
(31) Felsenstein, J. Confidence limits on phylogenies: an approach
using the bootstrap. Evolution 1985,39, 783−791.
(32) Jones, D. T.; Taylor, W. R.; Thornton, J. M. The rapid
generation of mutation data matrices from protein sequences. Comput.
Appl. Biosci. 1992,8, 275−282.
(33) Dobruchowska, J. M.; Gerwig, G. J.; Kralj, S.; Grijpstra, P.;
Leemhuis, H.; Dijkhuizen, L.; Kamerling, J. P. Structural character-
ization of linear isomalto−/malto-oligomer products synthesized by
the novel GTFB 4,6-α-glucanotransferase enzyme from Lactobacillus
reuteri 121. Glycobiology 2012,22, 517−528.
(34) van Leeuwen, S. S.; Kralj, S.; van Geel-Schutten, I. H.; Gerwig,
G. J.; Dijkhuizen, L.; Kamerling, J. P. Structural analysis of the α-D-
glucan (EPS35-5) produced by the Lactobacillus reuteri strain 35-5
glucansucrase GTFA enzyme. Carbohydr. Res. 2008,343, 1251−1265.
(35) Kabsch, W. XDS. Acta Crystallogr. D Biol. Crystallogr. 2010,66,
125−132.
(36) McCoy, A. J.; Grosse-Kunstleve, R. W.; Adams, P. D.; Winn, M.
D.; Storoni, L. C.; Read, R. J. Phaser crystallographic software. J. Appl.
Crystallogr. 2007,40, 658−674.
(37) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
macromolecular structures by the maximum-likelihood method. Acta
Crystallogr. D Biol. Crystallogr. 1997,53, 240−255.
(38) Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Features
and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 2010,
66, 486−501.
(39) Joosten, R. P.; Long, F.; Murshudov, G. N.; Perrakis, A. The
PDB_REDO server for macromolecular structure model optimiza-
tion. IUCrJ 2014,1, 213−220.
(40) Pettersen, E. F.; Goddard, T. D.; Huang, C. C.; Meng, E. C.;
Couch, G. S.; Croll, T. I.; Morris, J. H.; Ferrin, T. E. UCSF
ChimeraX: Structure visualization for researchers, educators, and
developers. Protein Sci. 2021,30,70−82.
(41) Kelley, L. A.; Mezulis, S.; Yates, C. M.; Wass, M. N.; Sternberg,
M. J. E. The Phyre2 web portal for protein modeling, prediction and
analysis. Nat. Protoc. 2015,10, 845−858.
(42) Brzozowski, A. M.; Davies, G. J. Structure of the Aspergillus
oryzae α-amylase complexed with the inhibitor acarbose at 2.0 Å
resolution. Biochemistry 1997,36, 10837−10845.
(43) Davies, G. J.; Brzozowski, A. M.; Dauter, Z.; Rasmussen, M. D.;
Borchert, T. V.; Wilson, K. S. Structure of a Bacillus halmapalus family
13 α-amylase, BHA, in complex with an acarbose-derived non-
asaccharide at 2.1 A resolution. Acta Crystallogr. D Biol. Crystallogr.
2005,61, 190−193.
(44) Arnal, G.; Cockburn, D. W.; Brumer, H.; Koropatkin, N. M.
Structural basis for the flexible recognition of α-glucan substrates by
Bacteroides thetaiotaomicron SusG. Protein Sci. 2018,27, 1093−1101.
(45) Robert, X.; Gouet, P. Deciphering key features in protein
structures with the new ENDscript server. Nucleic Acids Res. 2014,42,
W320−W324.
Journal of Agricultural and Food Chemistry pubs.acs.org/JAFC Article
https://doi.org/10.1021/acs.jafc.1c05657
J. Agric. Food Chem. XXXX, XXX, XXX−XXX
K
