Unexpected structure of a C. difficile toxin A ligand necessitates an annotation correction in a popular screening library

Article abstract of DOI:10.1039/c1cc15344g

The structure of the pentasaccharide S259-1 in the Consortium for Functional Glycomics was investigated using a variety of techniques. Surprisingly, the structure differs from the structure assumed from the previously established specificity of the human

Full text of DOI:10.1039/c1cc15344g

View Online / Journal Homepage / Table of Contents for this issue
Dynamic Article Links
ChemComm
Cite this: Chem. Commun., 2011, 47, 12397–12399
www.rsc.org/chemcomm
COMMUNICATION
Unexpected structure of a C. difficile toxin A ligand necessitates an
annotation correction in a popular screening libraryw
Ping Zhang,ab Nahid Razi,c Luiz Eugenio,b Messele Fentabil,d Elena N. Kitova,d
John S. Klassen,d David R. Bundle,d Kenneth K.-S. Ng*b and Chang-Chun Ling*a
Received 27th August 2011, Accepted 27th September 2011
DOI: 10.1039/c1cc15344g
The structure of the pentasaccharide S259-1 in the Consortium
for Functional Glycomics was investigated using a variety of
techniques. Surprisingly, the structure differs from the structure
assumed from the previously established specificity of the human
fucosyltransferase FUT-III used in the last step of chemo-
enzymatic synthesis. When presented with a tetrasaccharide
substrate containing both type I and type II disaccharide
moieties, the enzyme generates a pentasaccharide in which the
type II moiety is preferentially fucosylated. The unexpected
product generated by FUT-III in this case highlights the
importance of performing detailed structural analysis on products
generated by enzymes.
the large intestine through the recognition of specific carbo-
hydrates whose structures remain largely unknown.4 Ng et al.
expressed several His6-tagged recombinant proteins contain-
ing fragments of the carbohydrate binding domain of TcdA1c
and submitted one of these fragments (TcdA-A2) to the glycan
array screening core facility at the Consortium for Functional
Glycomics (CFG). This screen, as well as screens using the
native TcdA holotoxin, identified a number of carbohydrate
ligands from a library containing over three hundred diverse
glycan structures.2 Confirming the results of the glycan array
screens, the free pentasaccharide S259-1 with azidoethyl agly-
cone, which was originally identified as a ‘‘LeA-LacNAc’’
structure (1, Fig. 1), showed the highest affinity for TcdA in
all screens using the direct electrospray mass spectrometry
(ESI-MS) binding assay.2 An analogous pentasaccharide
LeA-LacNAc with an aminohexyl aglycone (2) was synthesized
by an unambiguous chemical route,1d and through a series
of studies involving NMR, X-ray crystallography of
oligosaccharide-toxin complexes, MS and additional confir-
matory chemical synthesis, we establish the correct structure
of S259-1.
During research aimed at characterizing the carbohydrate
binding of the Clostridium difficile toxins,1 glycan array screening2
identified a pentasaccharide (S259-1) which was produced by
chemoenzymic synthesis with a recombinant human a(1,3)/
a(1,4)-fucosyltransferase (FUT III)3 and exhibited strong
binding to toxin A (TcdA). We show here that the structure
of S259-1 was incorrectly assigned due to the unexpected
specificity of the recombinant form of the previously charac-
terized human fucosyltransferase, and we unambiguously
prove the structure of the oligosaccharide that is active in
binding TcdA.
C. difficile causes one of the most common hospital-
acquired bacterial infections in the industrialized world, and
produces two large protein toxins, TcdA and TcdB, which are
the major virulence factors in C. difficile infections (CDI).1a,b
TcdA and TcdB appear to adhere to host epithelial cells in
Fig. 1 Structures of 1 and 2.
S259-1 was prepared using enzymatic synthesis (Scheme 1).
Glycosylation of trisaccharide 3 with b(1,3)-galactosyltrans-
ferase V (GalT V) and the UDP-Gal 4-epimerase (GalE) using
UDP-Glc (4) as the donor (the GalE converts the UDP-Glc to
an intermediate UDP-Gal, required by GalT V) yielded the desired
tetrasaccharide 5, which had both a type I (bGal(1,3)GlcNAc)
disaccharide at the non-reducing end and the type II
(bGal(1,4)GlcNAc) disaccharide at the reducing end. Subsequent
fucosylation catalyzed by FUT III generated the mono-
fucosylated S259-1 pentasaccharide.
a Department of Chemistry, Alberta Ingenuity Centre for
Carbohydrate Science, University of Calgary, Calgary AB T2N1N4,
Canada. E-mail: ccling@ucalgary.ca; Fax: 403-289-9488;
Tel: 403-220-2768
b Department of Biological Sciences, Alberta Ingenuity Centre for
Carbohydrate Science, University of Calgary, Calgary AB T2N1N4,
Canada. E-mail: ngk@ucalgary.ca; Fax: 403-289-9311;
Tel: 403-220-4320
c Department of Chemical Physiology, The Scripps Research Institute,
La Jolla, CA 92037, USA
d Department of Chemistry, Alberta Ingenuity Centre for
Carbohydrate Science, University of Alberta, Edmonton AB
T6G2G2, Canada
The transferase enzyme, FUT III used in this study was
expressed in Sf9 insect cells using a recombinant baculovirus
construct.5 FUT III is a well-characterized, bifunctional
a(1,3)/a(1,4)-transferase capable of transferring L-fucose to
w Electronic supplementary information (ESI) available: Experimental
procedures and NMR spectra of synthesized compounds 5, S259-1
and 10–14. See DOI: 10.1039/c1cc15344g
c
This journal is The Royal Society of Chemistry 2011
Chem. Commun., 2011, 47, 12397–12399 12397

View Online
Fig. 2 Comparison of partial 1H NMR spectra (D2O, 400 MHz) of
(a): compound 2, (b) S259-1, and (c) compound 14.
Scheme 1 Enzymatic synthesis of pentasaccharide S259-1.
respectively. At the resolution of S259-1, and with the presence
of disorder at the non-reducing end, it was not possible to
define the entire sequence of the S259-1, but a tetrasaccharide
was clearly observed, and the electron density unambiguously
revealed that the fucose residue was attached to O-3 of the
type II LacNAc disaccharide at the reducing end to produce
not a LeA but a LeX trisaccharide fragment. In sharp contrast,
the electron density for compound 2 clearly showed the
expected LeA-LacNAc pentasaccharide attached to the aglycone
(Fig. 3).11
either type I or type II LacNAc disaccharides. Addition of
fucose to the 3-OH of GlcNAc in type I structures yields LeA-
type structures, while the addition of fucose to the 4-OH of
GlcNAc in type II structures generates LeX-type structures.
Previous studies have established that FUT III shows higher
activity towards type I substrates.6 Watkins7 and Palcic8
observed that a tetrasaccharide containing only the type I
LacNAc disaccharide, bGal(1,3)bGlcNAc(1,3)bGal(1,4)bGlc
(type I LacNAc - lactose) gave a LeA-Lac pentasaccharide as
the main product when a FUT-III promoted fucosylation was
performed. Consistent with this data, Hakomori et al.
reported9 that FUT-III promotes the addition of a single
fucose residue to the type I site at the non-reducing end of a
hexasaccharide containing the a type I + type II tetrasacharide
identical in sequence to that of tetrasaccharide 5 and attached
to a lactose disaccharide at the reducing end. The enzymatic
fucosylation produced a heptasaccharide containing the LeA
epitope at the non-reducing end. Since FUT III generally
shows low activity towards type II substrates, and has been
shown to fucosylate type II structures only in special cases,
such as when the substrate lacks a type I component as in the
cases of polylactosamine10 and bGal(1,4)bGlcNAc(1,4)
bGlcNAc (type IIb - GlcNAc),6a the structure of the mono-
fucosylated S259-1 was assumed to be LeA-LacNAc (1).
The chemically synthesized pentasaccharide 2 and enzymatically
synthesized S259-1 were expected to have the same LeA-
LacNAc glycosidic linkages differing only in their aglycone
moieties. However, the direct ESI-MS binding assay revealed
that compound 2 binds to TcdA-A2 roughly 20-times weaker
than S259-1. From the crystal structures of TcdA-A2,1c it was
known that the protein does not make significant interactions
with the aglycone. This suggested that differences between the
two ligands were not attributable to differences in the aglycone.
When we compared the 1D 1H NMR spectra of 21d and S259-1
(Fig. 2: (a) and (b)), we found that although both compounds
contained one a- and four b-anomeric linkages, the resonances
of the anomeric protons showed distinctly different chemical
shifts, suggesting that they contained different glycosidic linkages.
Unfortunately, due to severe overlap of proton resonances, an
attempt to determine the structure of S259-1 using advanced
NMR experiments was inconclusive. Both S259-1 and
compound 2 were then co-crystallized with a fragment of
TcdA, yielding data extending to 2.2 and 2.0 A resolution
Fig. 3 Crystallographically derived model of the carbohydrate found
in complex with TcdA-A2, which is in agreement with structure of 2.
Collision-induced dissociation (CID) MS analysis of S259-1
identified a fragment with m/z = 599.3, which is also consistent
with LeX attached to the 2-azidoethyl aglycone. In addition,
the peak at m/z = 437.2 corresponds to a fragment containing
fucose, an N-acetylhexosamine residue and a 2-azidoethyl
aglycone. The presence of these two fragments confirmed that
the fucose residue was added to the type II structure attached
to the aglycone (inner disaccharide), and not to the type I
disaccharide element that constitutes the non-reducing end
(distal disaccharide). Based on the X-ray and CID-MS data,
we concluded that the tightest binding ligand S259-1 has the
structure 7 (Fig. 4).
Fig. 4 The structure of S259-1 suggested by X-ray and CID-MS analysis.
The above results are surprising, because they differ from
the predictions of studies on FUT III purified from cultured
cells6–9 or recombinantly expressed from insect cells,12 all of
which suggested a preference of type I over type II structures.
To further confirm the structural identity of S259-1, we
carried out a chemical synthesis of compound 14, which is an
c
12398 Chem. Commun., 2011, 47, 12397–12399
This journal is The Royal Society of Chemistry 2011

View Online
analogue of compound 7, differing only in the aglycone. As
shown from Scheme 2, starting from the previously available
disaccharide 8 and thioglycoside 9, we carried out a N-iodo-
succinimide mediated glycosylation to afford the LeX-trisac-
charide 10 (81% yield). Transesterification gave diol 11 (96%
yield). Glycosylation of 11 with the known thiodisaccharide 12
provided the desired pentasaccharide 13 (77% yield). After
sequential deprotections, the desired pentasaccharide 14 was
obtained (56% yield over 4 steps).
fucosylating the type II LacNAc rather than its type I counter-
part in tetrasaccharide containing both disaccharide
a
elements. Further X-ray crystallographic studies investigating
the binding of tetrasasaccharide 5 with recombinant enzyme
should help to clarify this issue.
The unexpected substrate specificity seen in this report
highlights the importance of chemical proof of structure prior
to including reference compounds in screening arrays and
underlines the risks associated with structural assignment
based on the assumed fidelity of chemoenzymatic synthesis.
Our work correcting the structure of S259-1 clarifies this
specific structure in the CFG glycan array and whilst all other
compounds in the array synthesized by the enzyme FUC-III
have been checked, our findings suggest that in future compounds
synthesized by this enzyme should be confirmed by unambiguous
spectroscopic methods. For TcdA, the work reported here
suggests that the difference in location of a fucose residue and
difference in aglycone lead to a twenty-fold difference in
affinity between compounds S259-1 and 2. This finding provides
novel insights into toxin carbohydrate-binding specificity that
may ultimately be exploited for the development of tighter
binding and more specific inhibitors of the C. difficile toxins.
Notes and references
1 (a) C. L. McDonald, G. E. Killgore, A. Thompson, R. C. Owens
Jr., S. V. Kazakova, S. P. Sambol, S. Johnson and D. N. Gerding,
N. Engl. J. Med., 2005, 353, 2433; (b) D. E. Voth and J. D. Ballard,
Clin. Microbiol. Rev., 2005, 18, 247; (c) A. Greco, J. G. S. Ho,
S.-J. Lin, M. M. Palcic, M. Rupnik and K. K.-S. Ng, Nat. Struct.
Mol. Biol., 2006, 13, 460; (d) P. Zhang, K. Ng and C.-C. Ling, Org.
Biomol. Chem., 2010, 8, 128.
2 http://www.functionalglycomics.org/CFGparadigms/index.php/
C._difficile_toxin_A_(TcdA).
Scheme 2 Chemical synthesis of pentasaccharide 14.
3 (a) Handbook of Glycosyltransferases and Related Genes, ed.
N. Taniguchi, K. Honke and M. Fukuda, Springer-Verlag, 2002;
(b) D. J. Becker and J. B. Lowe, Glycobiology, 2003, 13, 41R;
(c) T. de Vries, R. M. A. Knegtel, E. H. Holmes and B. A. Macher,
Glycobiology, 2001, 11, 119R.
4 K. D. Tucker and T. D. Wilkins, Infect. Immun., 1991, 59, 73.
5 (a) J. F. Kukowska-Latallo, R. D. Larsen, R. P. Nair and
J. B. Lowe, Genes Dev., 1990, 4, 1288; (b) Obtained from NEOSE
Technologies Inc, Horsham, PA.
6 (a) M. Nimtz, E. Grabenhorst, U. Gambert, J. Costa, V. Wray,
M. Morr, J. Thiem and H. S. Conradt, Glycoconjugate J., 1998,
15, 873; (b) E. H. Holmes and S. B. Levery, Arch. Biochem.
Biophys., 1989, 274, 633.
7 P. H. Johnson, A. S. R. Donald, J. Feeney and W. M. Watkins,
Glycoconjugate J., 1992, 9, 251.
8 T. de Vries, C. A. Srnka, M. M. Palcic, S. J. Swiedler, D. H. van
den Eijnden and B. A. Macher, J. Biol. Chem., 1995, 270, 8712.
9 M. R. Stroud, S. B. Levery, S. Martensson, M. E. K. Salyan,
H. Clausen and S.-I. Hakomori, Biochemistry, 1994, 33, 10672.
10 S. Nishihara, H. Iwasaki, M. Kaneko, A. Tawada, M. Ito and
H. Narimatsua, FEBS Lett., 1999, 462, 289.
11 K. Ng, P. Zhang and C.-C. Ling, unpublished results.
12 V. A. Morais, J. Serpa, A. S. Palma, T. Costa, L. Maranga and
J. Costa, Biochem. J., 2001, 353, 719.
A series of 1D and 2D NMR experiments were carried out
to characterize compound 14 and to compare it with S259-1.
Fig. 2 compares the partial 1D 1H NMR spectra of compound
14 (c) to S259-1 (b). The pattern of the anomeric signals of
both compounds resembles closely as four out of the five
anomeric protons (at 5.14, 4.75, 4.48, 4.47 ppm) are super-
imposable. The remaining resonances (doublet at 4.55 ppm for
compound 14 vs. 4.65 ppm for S259-1) differ since they are the
anomeric protons of the terminal reducing GlcNAc residues
each of which are attached to different aglycones. Further-
more, we also compared the 2D COSY, 2D HSQC spectra of
compound 14 and S259-1, and the results confirm again that
two compounds have the same glycosidic linkages (see supporting
informationw). ESI-MS binding measurements showed that
compounds 14 and S259-1 have similar affinities for TcdA-A2,
consistent with NMR findings.
In conclusion, the work presented here establishes that the
recombinantly produced FUT-III enzyme has a preference for
c
This journal is The Royal Society of Chemistry 2011
Chem. Commun., 2011, 47, 12397–12399 12399