A catalytically inactive b1,4-
N
-acetylglucosaminyltransferase III
(GnT-III) behaves as a dominant negative GnT-III inhibitor
Hideyuki Ihara, Yoshitaka Ikeda, Souichi Koyota, Takeshi Endo, Koichi Honke and Naoyuki Taniguchi
Department of Biochemistry, Osaka University Medical School, Suita, Osaka, Japan
b1,4 -N-Acetylglucosaminyltransferase III (GnT-III) plays a
regulatory role in the biosynthesis of N-glycans, and it has
been suggested that its product, a bisecting GlcNAc, is
involved in a variety of biological events as well as in regu-
lating the b iosynthesis of t he oligosaccharides. I n t his s tudy,
it was f ound, on the basis of sequence homology, that GnT-
III contains a small region that is signi®cantly homologous
to both snail b1,4GlcNAc transferase and b1,4Gal trans-
ferase-1. Subsequent mutational analysis demonstrated an
absolute requirement for two conserved Asp residues
(Asp321 and Asp323), which are located in the most
homologous region of rat GnT-III, for enzymatic activity.
The overexpression of Asp323-substituted, catalytically
inactive GnT-III in Huh6 cells led to the suppression of the
activity of endogenous GnT-III, but no signi®cant decrease
in its expression, and led to a speci®c inhibition of the f or-
mation of bisected sugar chains, as shown by s tructural
analysis of the total N-glycans f rom t he cells. These ®ndings
indicate that the mutant serves a dominant negative eect on
a speci®c step in N-glycan biosynthesis. This type of Ôdomi-
nant negative glycosyltransferaseÕ, identi®ed has potential
value as a powerful tool for d e®ning the precise biological
roles of the bisecting GlcNAc structure.
Keywords: GnT-III; glycosyltransferase; bisecting G lcNAc;
N-glycan synthesis; dominant negative eect.

To address these issues, a catalytically inactive GnT-III
and/or dominant negative mutant GnT-III would be
valuable because i t may potentially de®ne whether
bisected N-glycan structures are actually important.
Alternatively, the i ssue of whether G nT-III protein i s
directly involved regardless of formation of bisected sugar
chains could also be examined.
The i denti®cation of catalytic residues, the absence of
which leads to the complete loss of activity, is required to
produce a catalytically inactive GnT-III. The enzymatic
properties of t his enzyme h ave been extensively studied in
terms of s ubstrate speci®city t owards acceptors a nd
donors, and these collective data p rovide an enzymatic
basis for the role of the bisecting GlcNAc in the regulation
of N-glycan biosynthesis [1,10,11]. Nevertheless, because
the catalytic mechanism of GnT-III has not yet been
analyzed in suf®cient detail, the chemical basis of the
enzymatic reaction is not known. The r eaction catalyzed
by GnT-III is accompanied by inversion at the anomeric
center of the transferred monosaccharide and the forma-
tion of a b1 ® 4 linkage, while chemically similar
reactions are a lso catalyzed by other g lycosyltransferases
such as b1,4GlcNAc transferase (b1,4GlcNAcT) from
snail, Lymnaea stagnalis [12], and mammalian b1,4Gal
transferases, (b1,4GalTs), both of w hich ar e mutually
homologous [13]. It s eems more likely that the common
properties of these three enzymes are associated with the
catalytic mechanism, rather t han substrate binding, a s the
Correspondence to N. Taniguchi, Department of Biochemistry, Osaka
University Medical School, 2-2 Yamadaoka, Suita, O saka 565-0871,

were purchased from Takara (Kyoto, Japan), Toyobo
(Shiga, Japan) a nd New England Biolabs ( UK). UDP-
GlcNAc and GlcNAc w ere obtained f rom Sigma (MO,
USA). O ligonucleotide p rimers wer e synthes ized b y Greiner
Japan (Tokyo, Japan). Antibodies were obtained from
following sources: monoclonal anti-(GnT-III) Ig from
Fujirebio Inc. (Tokyo, Japan); horseradish peroxidase
(HRP)-conjugated anti-(mouse IgG) Ig from Promega
(WI, USA). Peptide-N-glycosidase F (PNGase F) was
obtained from Roche Diagnostics (IN, USA). Standards of
pyridylaminated sugar chains were purchased from Takara
and Seikagaku Corp. (Tokyo, Japan). Other common
chemicals were obtained from Wako pure chemicals
(Osaka, Japan), Nacalai Tesque (Kyoto, Japan) and Sigma.
Construction of expression plasmids
For transient expression in COS-1 cells, a cDNA encoding
rat GnT-III [14] was subcloned into t he Ec o RI sites of an
SV40-based expression vector, pSVK3 (Amersham Phar-
macia Biotech, Buckinghamshire, UK). Wh en the enzyme
was stably expressed in Huh6 cells, the cDNA was
subcloned into the EcoRI sites of another expression vector,
pCXNII, which contains a neo
r
gene [15]. In this vector, the
GnT-III was expressed under the contro l of t he b-a ctin
promoter and the CMV enhancer.
Site-directed mutagenesis
Site-directed mutagenesis experiments were carried out
according to Kunkel [16], as described previously [17]. A
0.6-kb fragment obtained by digestion of rat GnT-III cDNA

100 U ámL
)1
penicillin, 1 00 lgámL
)1
streptomycin and
5gáL
)1
glucose under a humidi®ed atmosphere of 95% air
and 5% CO
2
.
Protein determination
Protein concentration was determined with BCA Kit
(Pierce, IL, USA) using BSA as a standard.
Electrophoresis and immunoblot analysis
SDS/PAGE was carried out on 8% gels, according to
Laemmli [18]. The separated proteins were transferred onto
a n itrocellulose membrane (PROTORAN, Schleicher &
Schuell Inc., NH, USA). The resulting membrane was
blocked with 5 % skimmed m ilk and 0.5% BSA in NaCl/P
i
containing 0.05% Tween-20, and was then incubated with
an anti-(GnT-III) Ig. After washing with NaCl/P
i
that
contained 0.05% Tween-20, the membrane was reacted with
a HRP-conjugated goat anti-(mouse I gG) Ig. The i mmuno-
reactive protein bands w ere visualized by chemiluminescence
using an ECL system (Amersham Pharmacia). Ponceau
staining of the t ransferred membrane was performed b efore

M
Mes/NaOH buffer (pH 6.25) containing 0.5% Triton
X-100, 200 m
M
GlcNAc. The assay mixture also contained
10 m
M
MnCl
2
for GnT-III assay or 10 m
M
EDTA for
GnT-V. The concentration of pyridylaminated biantennary
sugar chain was 10 l
M
for both enzymes. The concentra-
tions of the donor, UDP-GlcNAc, for GnT-III and GnT-V
were 20 and 40 m
M
, respectively. After incubation for 2±4 h,
the reactions were terminated by rapidly heating to 100 °C.
The reaction mixtures were t hen centrifuged at 10 000 g,for
10 min and the resulting supernatants were applied to an
HPLC equipped with a TSK-gel, ODS-80TM column
(4.6 ´ 150 mm) (Tosoh, Tokyo, Japan) in order to separate
and quantitate the products. Elution was performed
isocratically at 55 °Cusinga20-m
M
acetate buffer
(pH 4 .0) containing 0.3% and 0.15% butanol for GnT-III

were then re-N-acetylated in 10 mL saturated ammonium
bicarbonate with 554 ll acetic anhydride, followed by
desalting on a column of AG 50 W-X12 resin (Bio-Rad).
Fluorescence labeling of the sugar chains was c arried out by
reductive amination involving the use of a ¯uorescent
reagent, 2-aminopyridine and sodium cyanoborohydride
[24]. After pyridylamination, the excess reagents were
removed b y gel ®ltration with H W-40F (Toy opearl, Tosoh).
The pyridylaminated sugar chains were then desialylated,
degalactosylated and defucosylated by sialidase (Arthro-
bacter ureafaciens, N acarai tesque) , b-g a lactosidase (ja ck
bean, Seikagaku Corp.) and a-fucosidase (bovine kidney,
Sigma), respectively. The digested samples were then
analyzed by HPLC using a TSK-gel ODS-80TM column
(4.6 ´ 150 mm), and elution was performed at 55 °Cbya
linear gradient of butanol from 0.1% to 0.25% in 20 m
M
ammonium/acetate buffer (pH 4.0). The eluted sugar chain
peaks w ere i denti®ed by comparison with standard pyr-
idylaminated sugar chains (Takara and Seikagaku Corp.).
RT-PCR
Total RNA from parental Huh6 cells and various transfec-
tants was prepared using TRIZOL (Gibco-BRL, MD,
USA), and the cDNAs were synthesized by reverse
transcriptase with an oligo d T-adaptor p rimer from RNA
LA PCR Kit (Takara). In order to speci®cally detect the
expression of endogenous human GnT-III, PCR was
performed with the selective primers for human GnT-III
in a PCR Thermal Cycler 480 (Takara). The primers used in
this study were designed to detect mRNA for human GnT-

region was detected in all three enzymes. These Asp residues
correspond to Asp321, Asp323 and A sp329 in the rat GnT-
III sequence, as shown by sequence alignment (Fig. 1C).
The sequence of Asp321-Val322-Asp323 appears to corre-
spond to the D -X-D motif, as have been suggested for the
catalytic importance in many other glycosyltransferases
[27±41], and, thus, the sequence comparison suggests that
these Asp residues represent likely candidates for investiga-
tion as having a role in GnT-III activity.
Site-directed mutagenesis of the conserved
aspartic acid residues in GnT-III
To explore the requirement o f the cand idate amino acids,
Asp321, Asp323 and Asp329, for enzyme activity, mutant
Ó FEBS 2002 A dominant negative glycosyltransferase (Eur. J. Biochem. 269) 195
GnT-IIIs in which these Asp residues were replaced by Ala
were prepared using site-directed mutagenesis, and the
respective mutants were designated as D321A, D323A and
D329A. When these mutants were transiently expressed in
COS-1 cells, t heir protein expression levels were f ound to b e
similar to that of wild-type, as indicated by immunoblot
analysis (Fig. 2A,B). However, in the standard activity
assay, the D321A and D323A mutants had no detectable
catalytic activity, whereas the D329A mutant was fully
active, a s shown by an activity similar to that of the wild-
type enzyme (Table 1). No activity was detected in the
D321A and D323A mutants even under the conditions of
longer incubation time and UDP-GlcNAc concentration as
high as 100 m
M
, ind icating that the replacement of Asp321

Fig. 1. Dotplot analyses for rat GnT-III vs. snail b1,4GlcNAcT or bovine b1,4G alT-1. Dotplots for (A) rat G nT-III vs. L. stagnalis b1,4GlcNAcT
and ( B) rat GnT-III vs. bovine b1,4GalT-1 are shown. The am ino-aci d sequences we re compared under the conditions of a window size of 10
residues and 50% of identity using the computer software program,
ALIGN
. The numbers beside the axis indicate the residue number of each
enzyme. Diagonal plots show the homologous regions, which satisfy the ab ove conditions. T he sequences of the most ho mologous region are given
outside the matrices. (C) Multiple alignment of the homologous regions of GnT-III, b1,4GalT-1 and L. stagnalis b1,4GlcNAcT were carried out by
CLUSTALV
. The amino-acid residues which are conserved in all enzymes are highlighted by shaded boxes, and two homologous residues are also
indicated by grey-shaded boxes. The conserved asp artic acid residues that were examined by mutational a nalysis are indicated by arrowheads.
GenBank a ccession numbers for the glycosyltransferases are: G nT-III (human), D13789; GnT-III (rat), NM_01 9239; GnT-III (mouse),
NM_010795; b1,4GlcNA cT (L. stagnalis), X80228; b1,4 GalT-1 (human), X14085; b1,4 GalT-1 (bovine), X14558; b1,4 GalT-1 (mouse), J03880;
b1,4 GalT-2 (mouse), AB019541.
196 H. Ihara et al. (Eur. J. Biochem. 269) Ó FEBS 2002
spite of i ts requirement for their enzyme activities. On the
other h and, crystallographic analyses of b1,4GalT-1 [36] as
well as other glycosyltransferases [37±41] have sugg ested
that this motif serves the coordination of a divalent cation
such as Mn
2+
along with a phosphoryl group of the donor.
The motif would thereby allow t he enzyme to interact with a
nucleotide portion of the d onor nu cleotide sugar and also
may facilitate the reaction via electrostatic catalysis involv-
ing t he divalent cation. Although the function of the D -X-D
motif is not known for GnT-III, it is possible t hat the
D-X-D m otif in GnT-III plays a similar role to that in
b1,4GalT-1 because of the signi®cant sequence homology in
the region containing this motif. In a large clostridial
glucosyltransferase mutant similar to the GnT-III D321A

geneticin resistance, clones were grown separately, and the
expression of the rat mutant enz yme was veri®ed by
immunoblot analysis. As a result, we obtained three clones,
which overexpress the D323A GnT-III mutant (Fig. 2D). In
Huh6 cells, a s was in COS-1 cells, the D323A mutant was
expressed a t a similar level to the wild-type, and appeared to
be fully glycosylated (Fig. 2D,F). In the case of the other
mutant, D321A, however, we were not successful in
establishing such clones.
The speci®c suppression of endogenous GnT-III
activity by expression of the catalytically inactive
D323A mutant in Huh6 cells
When assays for GnT-III activity were performed in the
transfected cells, it was found that the activity in the
transfected cells were as low as less than 5 % of the activity in
the parental or mock-transfected Huh6 cells, which were
used as controls. On the other hand, the activities of GnT-V,
another GlcNAc transferase which is involved in the
formation of b1,6-branches, and b1,4GalT were not essen-
tially affected in the transfe cted cells, indicating that the
overexpression of the D323A mutant has no effect on these
glycosyltransferase activities (Fig. 3). Therefore, it appears
that the overexpression of the m utant does not impair
Fig. 2. Expression of the wild-type and mutant GnT-III proteins. The
wild-type and mutant enzymes were transiently expressed in COS-1
cells (A±C), and stably expressed in Huh6 c ells (D ±F). (A,D) The cell
homogenates were separated on 8% SDS-gels and were analyz ed by
immunoblot using a nti-(GnT -III) Ig. (B ,E) The amounts of protein
loaded were veri®ed by Ponceau staining, prior to t he immunoblot
analysis. (C,F) The s amples were treated w ith PNGase F, followed b y

To examine whether the mutant GnT-III inhibits the
intrinsic G nT-III in vitro, the extracts from t he parental
cells were mixed with the extract from the D323A-
transfected ce lls, and were then analyze d by the activity
assay. As shown in Fig. 4, no inhibitory effect of the mutant
was observed, and, thus, it seemed unlikely that the in vivo
inhibition by the mutant is due to direct competition for
substrate. Furthermore, RT-PCR using a primer set that is
speci®c to the human GnT-III sequence showed that the
mRNAs for endogenous human GnT-III were not signif-
icantly d ecreased (Fig. 5), suggesting t hat the decreased
GnT-III activity is not due to the downregulation of
expression of the intrinsic human GnT-III gene. Therefore,
it is more likely that t he activity of the i ntrinsic human
enzyme is decreased as the result of control at the
translational or post-translational level including protein±
protein interactions.
Blockage of a speci®c step, namely the formation
of a bisecting GlcNAc residue, in N-glycan biosynthesis
via the expression of the D323A mutant
In order to further examine the issue of whether the
biosynthesis of bisected sugar chains are inhibited by
overexpression of the D323A mutant, total N-linked sugar
chains were prepared from the cells and labeled with
2-aminopyridine, a ¯uorescent reagent. The resulting free
oligosaccharides were digested wi th sialidase, b- galactosi-
dase and a-fucosidase, and then s ubjected to r eversed phase
HPLC, in order to analyze the core structures, i.e . the
addition of a b isecting GlcNAc and t he extent of branching.
As shown in Fig. 6, when parental cells, which express

Lanes: h uman GnT-III and rat GnT-III indic a te PCR-amp li®cation of plasmids c ontaining cDNAs for hu man and rat e nzym e, resp ec tively. D e tails
regarding the PCR procedures are described in Experimental p rocedures.
Fig. 6. Structural analysis of Asn-linked sugar
chains from parental and transfected Huh6
cells. Elution p ro®les of pyridylaminated sugar
chains in reversed phase HPLC are shown for
parental Huh6 cells ( a), wild-type GnT-III-
transfectant (b) and D 323A-transfectant (c).
Numbers at the top indicate peaks for bisected
sugar chain s: 1, as ialo -agala cto-bisec ted
biantennary sugar chain; 2, asialo-agalacto-
bisected tetraantennary sugar c hain; 3, a sialo-
agalacto-bisected triantennary sugar chain
containing a b1,4-GlcNAc residue on the
Mana1,3 arm. Arro wheads in dicate nonbi-
sected sugar chains: left arrowhead, overlap-
ping peaks of asialo-agalacto biantenna ry and
asialo-agalacto tetraantennary s ugar chains;
right, asialo, agalacto triantennary sugar chain
containing a b1,4-GlcNAc residue on the
Mana1,3 arm. These structures are shown
above the pro®le.
Ó FEBS 2002 A dominant negative glycosyltransferase (Eur. J. Biochem. 269) 199
CONCLUSIONS
Our present study identi®ed Asp321 and Asp323 as being
essential residues in the enzyme activity of GnT-I II, and
provides further support for the v iew that the D-X-D motif
in glycosyltransferases is important and signi®cant, even
though the exact roles of Asp321 and Asp323 in GnT-III
remains to be elucidated. On the other hand , the utility of

1. Narasimhan, S. (1982) Control of glycoprotein synthesis. UDP-
GlcNAc: glycopeptide beta 4-N-acetylglucosaminyltransferase III,
an enzyme in hen oviduct which adds GlcNAc in beta 1 ® 4
linkage to the beta-linked mannose of the trimannosyl core of
N-glycosyl oligosaccharides. J. Biol. Chem. 257, 10235±10242.
2. Schachter, H. (1986) Biosynthe tic controls that determine the
branching and microhete rogeneity of prote in-bou nd oligosac char-
ides. Biochem. Cell Biol. 64, 163±181.
3. Sultan, A .S., Miyoshi, E., Ihara, Y., Nishikawa, A., T sukada, Y.
& Taniguchi, N. (1997) Bisecting G lcNAc structures act as nega-
tive sorting signals for cell surface glycoproteins in forskolin-
treated rat hepatoma cells. J. Biol. Chem. 272, 2866±2872.
4. Ihara,Y.,Yoshimura,M.,Miyoshi,E.,Nishikawa,A.,Sultan,
A.S., Toyosawa, S., Ohnishi, A., Suzuki, M., Yamamura, K.,
Ijuhin, N . & Taniguch i, N. (1998) E ctopic expression of N-acety-
lglucosaminyltransferase I II in t ransgenic hepatocytes disrupts
apolipoprotein B secretion and induces aberrant cellular mor-
phology with lipid storage. Proc. Natl Acad. Sci. USA 95, 2526±
2530.
5. S heng, Y., Yoshimura, M., Inoue, S., Oritani, K., Nishiura, T.,
Yoshida,H.,Ogawa,M.,Okajima,Y.,Matsuzawa,Y.&
Taniguchi, N. (1997) Remodeling of glycoconjugates on CD44
enhances cell adhesion to hyaluronate, tumor growth and meta-
stasis in B16 melanoma cells expressing b1,4-N-acetylglucosami-
nyltransferase III. Int. J. Cancer 73, 850±858.
6. Yoshimura, M., Ihara, Y., Matsuzawa, Y. & Taniguchi, N. (1996)
Aberrant glycosylation of E-cadherin enhances cell±cell binding to
suppress metastasis. J. Biol. Chem. 271, 13811±13815.
7.Yoshimura,M.,Nishikawa,A.,Ihara,Y.,Taniguchi,S.&
Taniguchi, N. (1995) Suppression of lung metastasis of B16 mouse

Taniguchi, N. (1992) Puri®cation, c DNA c loning, and expression
of UDP-N-acetylglucosaminyltransferase III from rat kidney.
J. Biol. Chem. 267, 18199±18204.
15. Niwa, H., Yamamura, K. & Miyazaki, J. (1991) Ecient selection
for high-expression transfectants with a n ovel e ukaryotic vector.
Gene 108, 193±199.
16. Kunkel, T.A. (1985) Rapid and ecient site-speci®c mutagenesis
without phenotypic selectio n. Proc.NatlAcad.Sci.USA82, 488±
492.
17. Ikeda, Y., Fujii, J., Taniguchi, N. & Meister, A. (1995) Human
c-glutamyl transpeptidase mutants involving conserved aspartate
residues and the unique cystein e residue of the ligh t subunit. J. Biol.
Chem. 270, 12471±12475.
18. Laemmli, U.K. (1970) Cleavage of structural proteins during the
assembly of the h ead of bacteriophage T4. Nature 227, 680±685.
19. Chu, G., H ayakawa, H. & Berg, P. (1987) Elec troporation for the
ecient transfection of mammalian cells with DNA, Nucleic Acids
Res. 15, 1311±1326.
20. T aniguchi, N., Nishikawa, A., F ujii, S. & Gu, J.G. (1989) Gly-
cosyltransferase assays using p yridylaminated acceptor: N-acetyl-
glucosaminyltransferase III, IV, and V. M ethod s Enzymol. 179,
397±408.
21. Nishikawa, A., Fujii, S., Sugiyama, T. & Taniguchi, N. (1988) A
method for the determin atio n o f N-acetylglucosaminyltransferase
III activity in rat tissues involving HPLC. Anal. Biochem. 170,
349±354.
22. S eko, A., Koketsu, M., N ish izono, M., Enoki, Y ., Ibrahim, H.R.,
Juneja, L.R., Kim, M. & Yamamoto, T. (1997) Occurrence of a
sialylglycopeptide and free sialylglycans in hen's egg yolk.
Biochem. Biophys. Acta 1335, 23±32.

30. Hagen, F.K., H azes, B., de Rao, R.Sa, D. & Tabak, L.A. (1999)
Structure-function an alysis of the UDP -N-acetyl-
D
-galactos-
amine: polypeptide N-acetylgalactosaminyltransferase. Essential
residues lie in a predicted active site cleft resembling a lactose
repressor fold. J. Biol. Chem. 274, 6797±6803.
31. Munro, S. & Freeman, M. (2000) The notch signalling regulator
fringe acts in the Golgi apparatus and requires the glycosyltrans-
ferase sig natu re mo tif D XD. Curr. Biol. 10, 813±820.
32. Keusch, J.J., Manzella, S.M., Nyame, K.A., Cummings, R.D. &
Baenziger, J.U. (2000) E xpression cloning of a new member of the
ABO blood g roup g lycosyltransferases, iGb3 s ynthase, t hat directs
the synthesis of isoglobo-glycosphingolipids. J. Biol. Chem. 275,
25308±25314.
33. Keusch, J.J., Manzella, S.M., Nyame, K.A., Cummings, R.D. &
Baenziger, J.U. (2000) Cloning of G b3 syn thase, t he key e nzyme i n
globo-series glycosphingolipid synthesis, predicts a family of
a1,4-glycosyltransferases conserved inplants, insects, and mam-
mals. J. Biol. Chem. 275, 25315±25321.
34. Li, J ., Rancour, D.M., Allende, M.L., Worth, C.A., D arling, D.S.,
Gilbert,J.B.,Menon,A.K.&Young,W.W.Jr(2001)TheDXD
motif is required for GM2 synthase activity but is not critical for
nucleotide binding. Glycobiology 11, 217±229.
35. Maeda, Y., Watanabe, R., Harris, C.L., Hong, Y., Ohishi, K.,
Kinoshita, K. & K inoshita, T. (2001) PIG-M transfers the ®rst
mannose t o glycosylphosphatidylinositol on the lumenal side of
the ER. EMBO J. 20, 250±261.
36. Gastinel, L.N., Cam billau, C. & Bourne, Y. ( 1999) Crystal
structures of t he b ovine b4-galactosyltransferase catalytic domain

blastoma cell l ines: correlation with activities of a1,6,fucosyl-
transferase and N-acetylglucosaminyltransferases III and V. Int. J.
Cancer 8, 315±317.
44. Ogawa, H ., Kobayashi, I., Nagasaka, T., N amii, Y ., Hayashi, S.,
Kadomatsu, K., Muramatsu, T. & Takagi, H. (1999) Suppression
of porcine xenoantigen expression by dominant-negative eect of
a-1,3-galactosyltransferase (a-1,3-GT) splicing variants. Tr ans-
plant. Proc. 32, 58.
45. Evans, S.C., Lopez, L.C. & Shur, B.D. ( 1993) Dom inant ne gative
mutation in cell su rfac e b1,4-galactosyltransferase inhibits cell±cell
and cell±matrix interactions. J. Cell Biol. 120, 1045±1057.
Ó FEBS 2002 A dominant negative glycosyltransferase (Eur. J. Biochem. 269) 201