Identification of two cysteine residues involved in the binding
of UDP-GalNAc to UDP-GalNAc:polypeptide
N
-acetylgalactosaminyltransferase 1 (GalNAc-T1)
Mari Tenno
1
, Shinya Toba
1
, Fere
´
nc J Ke
´
zdy
3
,A
˚
ke P. Elhammer
3
and Akira Kurosaka
1,2
1
Department of Biotechnology Faculty of Engineering, and
2
Institute for Comprehensive Research, Kyoto Sangyo University,
Kamigamo-motoyama, Kyoto, Japan;
3
Pharmacia Corporation, Kalamazoo, Michigan, USA
Biosynthesis of mucin-type O-glycans is initiated by a family
of UDP-GalNAc:polypeptide N-acetylgalactosaminyl-
transferases, which contain several conserved cysteine resi-
dues among the isozymes. We found that a cysteine-specific

sylation; UDP-GalNAc.
Mucin-type O-glycosylation is an important post-transla-
tional modification that is widely distributed on many
secretory and membrane glycoproteins [1,2]. The initial step
of this glycosylation is catalyzed by the UDP-GalNAc:poly-
peptide N-acetylgalactosaminyltransferases (GalNAc-trans-
ferases; EC 2.4.1.41). These enzymes transfer GalNAc from
UDP-GalNAc to serine or threonine residues of proteins [3].
Recent progress in molecular cloning has revealed that the
GalNAc-transferases constitute a large gene family, with 10
distinct isozymes identified to date [4–14], and that they are
type II membrane proteins with a short N-terminal
cytoplasmic tail, a hydrophobic transmembrane anchor, a
luminal stem region, and a large luminal putative catalytic
domain (Fig. 1). The luminal putative catalytic domain
contains two distinct subdomains; a central catalytic
domain and a C-terminal lectin-like domain. The central
catalytic domain can be further subdivided into two regions.
The N-terminal half is represented by a glycosyltransferase 1
(GT1) motif that is conserved among a wide range of
glycosyltransferases [15]. The extreme C-terminal end of the
GT1 motif contains a so-called DXH motif, which corres-
ponds to the DXD sequence common to many glyco-
syltransferases [16]. The C-terminal half of the catalytic
domain contains a so-called Gal/GalNAc-T motif, a
sequence segment where significant homology can be seen
between b1,4-galactosyltransferases and GalNAc-transfer-
ases [15,17]. A C-terminal lectin-like domain, called the
(QXW)
3

a disulfide bond), by cysteine-specific labeling, to study the
mechanistic involvement of the conserved cysteine residues
in the function of GalNAc-T1. Our results demonstrate that
Cys212 and Cys214, which are located at the C terminus of
the DXH motif, are free cysteine residues that interact with
the nucleotide moiety of UDP-GalNAc, possibly through
hydrogen bonding.
EXPERIMENTAL PROCEDURES
Preparation of soluble bovine GalNAc-T1
A soluble form of bovine GalNAc-T1 was expressed in
High Five cells using the baculovirus expression system. The
molecule was purified to homogeneity by apomucin-
Sepharose chromatography as described previously [20].
Construction of soluble rat recombinant GalNAc-T1
and expression in COS7 cells
Rat GalNAc-T1 cDNA was obtained as outlined by Hagen
et al. [21]. For the construction of soluble GalNAc-T1, rat
GalNAc-T1 full-length cDNA was subcloned into pcDNA4
to create the vector, prT1. prT1 was linearized with BamHI,
andthendigestedwithBal31 nuclease. The Bal31 digest was
blunt-ended and digested with NotI. The resulting digest
was ligated into the EcoRV and NotIsitesofpcDNA4,
obtaining pDN42 that encodes GalNAc-T1 with 42
N-terminal amino acid residues, including a cytoplasmic
tail and a transmembrane domain, deleted. A NheI-SmaI
fragment of the plasmid pGIR201protA (a gift from H.
Kitagawa, Kobe Pharmaceutical University) [22,23],
containing a cDNA encoding the insulin signal sequence
and the Protein A-IgG binding domain, was inserted into
NheI-HindIII digested pcDNA3.1, producing the vector,

human GalNAc-T3; hT4, human GalNAc-T4; rT5, rat GalNAc-T5; hT6, human GalNAc-T6; hT7, human GalNAc-T7; hT8, human GalNAc-T8;
hT9, human GalNAc-T9; rppGaNTase-T9, rat ppGaNTase-T9.
Ó FEBS 2002 Cys residues in GalNAc-T1 interact with UDP-GalNAc (Eur. J. Biochem. 269) 4309
secreted enzyme was purified on IgG-Sepharose. For
analysis by SDS/PAGE, the resins adsorbed with the
secreted enzyme were boiled in SDS/PAGE loading buffer.
The resulting supernatant was loaded directly on the gel.
For Western blotting, the proteins on the membrane were
visualized by incubating the blot with an affinity purified,
alkaline phosphatase-conjugated, rabbit antibody to mouse
IgG, followed by staining with nitrobluetetrazolium and
5-bromo-4-chloro-3-indolylphosphate. The protein bands
on the immunoblots was quantified by densitometry
scanning and the intensity of each band was determined
using the NIH Image software. The enzymatic activity of
the P-DN42 and mutant P-DN42 gene products was
determined as described below. The activity levels were
corrected for enzyme protein concentration in the medium.
Assay for GalNAc-transferase activity (PD-10 assay)
The enzyme activity was determined in a reaction mixture
composed of 50 m
M
imidazole buffer (pH 7.2), 10 m
M
MnCl
2
, 0.1% Triton X-100, 6 nmol UDP-
3
H-GalNAc
(approximately 10 000 d.p.m.), 150 lg apomucin [24],

M
PCMPS in the presence of UDP-GalNAc or UDP.
Identification of free cysteine residues
Labeling of bovine GalNAc-T1 with ABD-F and fraction-
ation of the labeled peptides were carried out as described
[25,26]. Briefly, GalNAc-T1 was first labeled with ABD-F,
followed by reduction with tributylphosphine and S-carbo-
xymethylation with iodoacetic acid. The alkylated protein
was digested with endoproteinase Lys-C. The digest was
then fractionated by HPLC on a C
18
HPLC column. The
fluorescent peptides were purified by rechromatography on
aC
8
column and sequenced with an automated Edman
sequencer.
Kinetic analysis
K
m
for UDP-GalNAc was obtained by varying the
concentration of UDP-GalNAc from 1.5 to 43.5 l
M
in
the presence of 1.88 mgÆmL
)1
apomucin. To determine the
K
m
for apomucin, GalNAc-transferase activity was assayed

sensitivity in this experiment, the cysteine modification
was performed with the minimal PCMPS concentration
(0.1 m
M
) required for complete inhibition of GalNAc-T1
(Fig. 2). Fig. 3 shows that GalNAc-T1 retained enzymatic
activity in the presence of either UDP or UDP-GalNAc.
This suggests that the sulfhydryl groups of free cysteine
residues modified by PCMPS may interact with UDP-
GalNAc, or at least be located in the UDP-GalNAc binding
cleft. Furthermore, the data suggest that the cysteine
residues predominantly interact with a UDP moiety of
UDP-GalNAc, as UDP and UDP-GalNAc were equally
effective at protecting the enzyme from inactivation.
Fig. 2. Inhibition of GalNAc-T1 with PCMPS. Purified bovine
GalNAc-T1 was incubated with increasing concentrations of PCMPS
for 90 min at room temperature. Following incubation, the treated
enzyme was dialyzed to remove excess PCMPS, and assayed for
activity as described in Experimental procedures.
4310 M. Tenno et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Mutagenesis of the cysteine residues in and around
the GT1 and Gal/GalNAc-T motifs
To investigate which cysteine residues are involved in the
catalytic function of GalNAc-T1, site-directed mutagenesis
was carried out on the conserved cysteine residues in the
catalytic domain. A rat GalNAc-T1 cDNA, cloned by PCR
as outlined by Hagen et al. [21], was used for this
experiment. Rat GalNAc-T1 is 98% identical to the bovine
ortholog and all of the cysteine residues are conserved
between the two enzymes (Fig. 1). For ease of purification

activities of C212A and C214A were drastically decreased,
to 6% and 17% of that of P-DN42, respectively. These
results indicate that the conserved Cys106, Cys212, and
Cys214 residues are essential for efficient enzyme function.
In contrast, the C235A mutant retained almost full activity,
as well as a high level of secretion into the culture medium.
Hence, Cys235 appears not to be required for GalNAc-T1
activity or secretion.
The Gal/GalNAc-T motif contains one conserved
cysteine residue, Cys330. In addition, there are two
conserved cysteine residues (Cys339 and Cys408) at the
C-terminal side of this motif. Each of these cysteine residues
was also mutated to alanine. Secretion of the mutants,
especially C339A, decreased significantly. Moreover, there
was a complete loss of activity in all three mutant enzymes
(Fig. 4). These results demonstrate that the cysteine residues
in positions 330, 339 and 408 are important for both
secretion and function of GalNAc-T1.
Identification of free cysteine residues in GalNAc-T1
The inactivation observed for several of the GalNAc-T1
mutants may result either from conformational changes
caused by the disruption of disulfide bridges or from
mutational effects of cysteine residues involved in enzyme
function. However, the results from modification of
GalNAc-T1 with PCMPS (Fig. 2) strongly suggest the
presence of essential, free cysteine residues. To identify
the free cysteine residues in the native enzyme, we
labeled soluble bovine GalNAc-T1 with a cysteine specific
Fig. 3. Protection of GalNAc-T1 from PCMPS inactivation. GalNAc-
T1wastreatedwith0.1m

8
column. This revealed
that peaks 1, 2, and 3 were artifacts as all of them separated
into several small peaks on the C
8
column and none of these
(secondary peaks) contained any polypeptide sequence
detectable by sequence analysis (data not shown). By
contrast, peak 4 produced a single peak on the C
8
column.
Edman analysis showed that this peak contained an ABD-F
labeled peptide. The peptide contained the N-terminal
sequence G202QVITFL
DAHC212EC214TV. The sequence
includes the GalNAc-T1 DXH motif (underlined above), a
region believed to be involved in coordination of a divalent
cation or the binding of UDP-GalNAc [16,28,29]. The two
cysteine residues in the sequence, Cys212 and Cys214, were
both labeled by ABD-F (Fig. 6), as demonstrated by the
presence of a peak corresponding to fluorescent cysteine and
the complete absence of a peak corresponding to carbo-
xymethylated cysteine, in the sequence analysis of the
peptide. Consequently, both Cys212 and Cys214 can be
considered free cysteine residues that probably are exposed
on the surface of the UDP-binding pocket. The other
essential cysteine residues, which were identified by muta-
tional analysis but not labeled by ABD-F, may form
intramolecular disulfide bonds required for proper folding
of the enzyme. Cys235, on the other hand, does not seem to

determined with an automated Edman sequencer. The solid and the
open arrows indicate the elution position of ABD-F labeled cysteine,
and S-carboxymethylated cysteine, respectively. HPLC profiles of (A)
cycle11(Cys212)and(B)13(Cys214)areshown.
4312 M. Tenno et al. (Eur. J. Biochem. 269) Ó FEBS 2002
As the fluorescent labeling was performed without dena-
turing reagents, it is possible that only free cysteine residues
exposed to the solvent were labeled by the ABD-F.
Expression and kinetic studies of cysteine-to-serine
mutant GalNAc-T1 enzymes
The findings that the activities of the C212A and C214A
mutants were drastically decreased and that UDP or UDP-
GalNAc prevented PCMPS inactivation of GalNAc-T1
suggest that electrostatic interactions through the polar
sulfhydryl groups of these cysteine residues may be
involved in the interaction(s) between the nucleotide moiety
of UDP-GalNAc and GalNAc-T1. To examine this
hypothesis, we generated two single point mutants,
C212S and C214S, and one double point mutant, C212S/
C214S. In these mutants, the cysteine residues (212 and
214) were replaced by serine residues, thereby generating
proteins with hydroxyl, instead of sulfhydryl, side chains at
positions 212 and 214. This should allow retained hydro-
gen bonding capacity at these positions and at least
theoretically, if this capacity is an essential function of these
residues, result in functional enzyme. The results shown in
Fig. 7 suggest that this is indeed the case. C212S and
C214S retained approximately 60% and 80% of parent
enzyme activity, respectively. This is significantly higher
than the activity of the corresponding alanine mutants (6%

We also examined the sensitivity of the three serine mutants
to PCMPS inactivation (shown in Fig. 8). C214S, which
contains free Cys212, was inactivated by PCMPS with a K
i
of 0.03 m
M
almost identical to that of native GalNAc-T1.
On the other hand, C212S was more resistant to the
treatment. The K
i
of  0.65 m
M
, may be due to a lower
reactivity of Cys214 as compared to Cys212. Moreover, no
inhibition was observed for C212S/C214S, even in the
presence of a large excess of PCMPS (1 m
M
), that resulted
in the complete inactivation of P-DN42. These results show
that both Cys212 and Cys214 are modified by PCMPS, and
that, consistent with the kinetic data, Cys212 is the most
Fig. 7. Enzymatic activity of cysteine-to-serine mutant GalNAc-T1
enzymes. Enzyme activity measurements and Western blotting of
mutant proteins were carried out as described in Fig. 4.
Table 1. Comparison of donor and acceptor K
m
values (apparent) for the
parent and mutant enzymes. Values shown are means of three separate
determinations.
UDP-GalNAc Apomucin

serine residues. Native bovine GalNAc-T1 is totally insen-
sitive to DFP treatment, and thus appears not to contain
any serine residues important for enzyme function. By
contrast, all three serine mutants were inactivated by DFP
to some extent (Fig. 9). The inhibition was more efficient for
C212S than C214S, again demonstrating that position 212 is
the more important site for substrate interaction. The
double mutant, C212S/C214S, was most susceptible to
DFP, confirming the cooperative involvement of the two
sites observed in the kinetic analysis (Table 1).
Taken together, the results from the kinetic, mutational
and chemical modification studies presented in this report
strongly suggests that the sulfhydryl groups at Cys212 and
Cys214, but primarily at Cys212, are involved in substrate
binding, possibly as hydrogen bond partners with UDP.
DISCUSSION
The primary aim of this study is to evaluate the functional
role(s) of the conserved cysteine residues found in the
GalNAc-transferase family. Using site-directed mutagene-
sis, in combination with identification of free cysteine
residues by cysteine-specific labeling, we demonstrated that
Cys212 and Cys214, but predominantly Cys212, is involved
in the binding of the nucleotide portion of UDP-GalNAc,
most probably through hydrogen bonding. This is consis-
tent with our previous inhibition study on GalNAc-T1
using various nucleotides and nucleotide sugars [30],
showing that the enzyme primarily recognizes the UDP
portion of the sugar donor. Recent crystallographic studies
on glycosyltransferases indicate that several interactions are
involved in binding of the UDP portion of the sugar donors

IEV (GalNAc-T8), DAHVEF
(GalNAc-T9), and DSHCE
A (ppGaNTase-T9). Of these
isozymes with variation at the two cysteine sites, GalNAc-
T5, -T7 and ppGaNTase-T9 are catalytically active. This
indicates that the two cysteine residues C-terminal to the
DXH motif may not be crucial for the basic catalytic
function of the GalNAc-transferases, but rather are
important in defining the catalytic properties of specific
isozymes. In fact, the interaction of UDP-GalNAc with
GalNAc-T5, which has a valine residue at the Cys212 site, is
less efficient than with GalNAc-T1. GalNAc-T5 has a
significantly lower affinity for UDP-GalNAc (K
m
¼ 55 l
M
)
[9], than P-DN42 and its serine-mutants (Table 1). The low
affinity of UDP-GalNAc for GalNAc-T5 may, at least in
part, be ascribed to the substitution with valine at the
Cys212 site. The two isozymes GalNAc-T8 and GalNAc-T9
lack cysteine residues at both position 212 and 214.
Consequently they may also have a low affinity for
UDP-GalNAc and consistent with this, no enzymatic
activity has so far been reported for these molecules. It is
possible that the activity of these isozymes cannot be
measured under the standard assay conditions used for
other GalNAc-transferases. Similarly, the importance of
four histidine residues for GalNAc-T1 activity has been
demonstrated by Wragg et al. [17], using site-directed

ferase VII) residue at the corresponding site. Importantly,
the NEM-sensitive cysteine residue is reported to be located
in or near the binding site for GDP-Fuc [38–41], in analogy
to what was found for GalNAc-T1.
The known glycosyltransferases have been classified into
52 different families, based both on sequence similarity and
substrate/product stereochemistry (inverting or retaining),
at the carbohydrate-active enzymes server on world
wide web, URL: />db.html [42,43]. The crystal structures of several glyco-
syltransferases belonging to different groups have been
determined recently [31–37]. Although these proteins have
no sequence identity or related functional features, the
crystallographic studies show that some of them share a
domain structure, called a UBD (UDP-binding domain)
[44], also known as a SGC (SpsA N-acetylglucosaminyl-
transferase I core) domain [28,32]. The UBD is predicted to
consist of alternating a-helices and b-sheets, constituting an
a-b-a sandwich [31,44]. The UBD of glycosyltransferases
also contains a DXD motif. In all crystallized enzymes, the
DXD motifs are located at positions closely related to one
anotherandareexpectedtobeindirectcontactwiththe
sugar donor or interact with UDP-sugars through binding
with a divalent ion [28,29]. The two cysteine residues at
positions 212 and 214 in GalNAc-T1, identified in this study
as being involved in sugar donor binding, are in the GT1
motif. A DXH motif that precedes Cys212 and Cys214 is
located at the C-terminal end of the GT1 motif. The
hydrophobic cluster analysis of several glycosyltransferases
demonstrates that the DXH motif corresponds to the DXD
motif found in most other glycosyltransferases [29]. More-

Education, Culture, Sports, Science, and Technology, Japan.
REFERENCES
1. Paulson, J.C. & Colley, K.J. (1989) Glycosyltransferases. Struc-
ture, localization, and control of cell type-specific glycosylation.
J.Biol.Chem.264, 17615–17618.
2. Jentoft, N. (1990) Why are proteins O-glycosylated? Trends
Biochem. Sci. 15, 291–294.
3. Clausen, H. & Bennett, E.P. (1996) A family of UDP-
GalNAc:polypeptide N-acetylgalactosaminyltransferases control
the initiation of mucin-type O-linked glycosylation. Glycobioogy.
6, 635–646.
4. Homa, F.L., Hollander, T., Lehman, D.J., Thomsen, D.R. &
Elhammer, A
˚
.P. (1993) Isolation and expression of a cDNA clone
encoding a bovine UDP-GalNAc:polypeptide N-acetylgalactos-
aminyltransferase. J. Biol. Chem. 268, 12609–12616.
5. White, T., Bennett, E.P., Takio, K., Sorensen, T., Bonding, N. &
Clausen, H. (1995) Purification and cDNA cloning of a human
UDP-N-acetyl-a-D-galactosamine: polypeptide N-acetylgalactos-
aminyltransferase. J. Biol. Chem. 270, 24156–24165.
6. Bennett, E.P., Hassan, H. & Clausen, H. (1996) cDNA cloning
and expression of a novel human UDP-N-acetyl-a-
D
-galactos-
amine: polypeptide N-acetylgalactosaminyltransferase, GalNAc-
T3. J. Biol. Chem. 271, 17006–17012.
7. Hagen, F.K., TenHagen, K.G., Beres, T.M., Balys, M.M.,
VanWuyckhuyse, B.C. & Tabak, L.A. (1997) cDNA cloning and
expression of a novel UDP-N-acetyl-

D
-galactosamine: polypep-
Ó FEBS 2002 Cys residues in GalNAc-T1 interact with UDP-GalNAc (Eur. J. Biochem. 269) 4315
tide N-acetylgalactosaminyltransferase, GalNAc-T7, with specifi-
city for partial GalNAc-glycosylated acceptor substrates. FEBS
Lett. 460, 226–230.
12. White, K.E., Lorenz, B., Evans, W.E., Meitinger, T., Strom, T.M.
& Econs, M.J. (2000) Molecular cloning of a novel human
UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase,
GalNAc-T8, and analysis as a candidate autosomal dominant
hypophosphatemic rickets (ADHR) gene. Gene 246, 347–356.
13.Toba,S.,Tenno,M.,Konishi,M.,Mikami,T.,Itoh,N.&
Kurosaka, A. (2000) Brain-specific expression of a novel human
UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferase
(GalNAc-T9). Biochim. Biophys. Acta 1493, 264–268.
14. Hagen, K.G., Bedi, G.S., Tetaert, D., Kingsley, P.D., Hagen,
F.K., Balys, M.M., Beres, T.M., Degand, P. & Tabak, L.A. (2001)
Cloning and characterization of a ninth member of the UDP-
GalNAc:polypeptide N-acetylgalactosaminyltransferase family,
ppGaNTase-T9. J. Biol. Chem. 276, 17395–17404.
15. Hagen, F.K., Hazes, B., Raffo, R., de Sa, D. & Tabak, L.A. (1999)
Structure-function analysis of the UDP-N-acetyl-
D
-galactosa-
mine: polypeptide N-acetylgalactosaminyltransferase. Essential
residues lie in a predicted active site cleft resembling a lactose
repressor fold. J. Biol. Chem. 274, 6797–6803.
16. Wiggins, C.A. & Munro, S. (1998) Activity of the yeast MNN1
a1,3-mannosyltransferase requires a motif conserved in many
other families of glycosyltransferases. Proc.Natl.Acad.Sci.USA

sialyltransferase gene family. J. Biol. Chem. 268, 11504–11507.
23. Kitagawa, H. & Paulson, J.C. (1993) Cloning and expression
of human Galb1,3 (4) GlcNAc a2,3-sialyltransferase. Biochem.
Biophys. Res. Commun. 194, 375–382.
24. Hagopian, A. & Eylar, E.H. (1968) Glycoprotein biosynthesis:
studies on the receptor specificity of the polypeptidyl: N-acetyl-
galactosaminyltransferase from bovine submaxillary glands. Arch.
Biochem. Biophys. 128, 422–433.
25. Toyo’oka, T. & Imai, K. (1985) Isolation and characterization of
cysteine-containing regions of proteins using 4-(aminosulfonyl)-7-
fluoro-2,1,3-benzoxadiazole and high-performance liquid chro-
matography. Anal. Chem. 57, 1931–1937.
26. Kirley, T.L. (1989) Determination of three disulfide bonds and one
free sulfhydryl in the b subunitof(Na,K)-ATPase.J.Biol.Chem.
264, 7185–7192.
27. Elhammer, A. & Kornfeld, S. (1986) Purification and character-
ization of UDP-N-acetylgalactosamine: polypeptide N-acetyl-
galactosaminyltransferase from bovine colostrum and murine
lymphoma BW5147 cells. J. Biol. Chem. 261, 5249–5255.
28. Unligil, U.M. & Rini, J.M. (2000) Glycosyltransferase structure
and mechanism. Curr. Opin. Struct. Biol. 10, 510–517.
29. Breton, C. & Imberty, A. (1999) Structure/function studies of
glycosyltransferases. Curr. Opin. Struct. Biol. 9, 563–571.
30. Elhammer, A
˚
.P., Ke
´
zdy, F.J. & Kurosaka, A. (1999) The acceptor
specificity of UDP-GalNAc:polypeptide N-acetylgalactosaminyl-
transferases. Glycoconj. J. 16, 171–180.

38. Holmes, E.H., Xu, Z.L., Sherwood, A. & Macher, B.A. (1995)
Structure-function analysis of human a1,3fucosyltrnsferases.
J. Biol. Chem. 270, 8145–8151.
39. Lowe, J.B., Kukowska-Latallo, J.F., Nair, R.P., Larsen, R.D.,
Marks, R.M., Macher, B.A., Kelly, R.J. & Ernst, L.K. (1991)
Molecular cloning of a human fucosyltransferase gene that
determines expression of the Lewis X and VIM-2 epitopes but not
ELAM-1-dependent cell adhesion. J. Biol. Chem. 266, 17467–
17477.
40. Natsuka, S., Gersten, K.M., Zenita, K., Kannagi, R. & Lowe, J.B.
(1994) Molecular cloning of a cDNA encoding a novel human
leukocyte a1,3-fucosyltransferase capable of synthesizing the sialyl
Lewis X determinant. J. Biol. Chem. 269, 16789–16794.
41. Sasaki, K., Kurata, K., Funayama, K., Nagata, M., Watanabe,
E., Ohta, S., Hanai, N. & Nishi, T. (1994) Expression cloning of a
novel a1,3-fucosyltransferase that is involved in biosynthesis of the
sialyl Lewis X carbohydrate determinants in leukocytes. J.Biol.
Chem. 269, 14730–14737.
42. Campbell, J.A., Davies, G.J., Bulone, V. & Henrissat, B. (1997) A
classification of nucleotide-diphospho-sugar glycosyltransferases
based on amino acid sequence similarities. Biochem. J. 326,929–
939.
43. Campbell, J.A., Davies, G.J., Bulone, V. & Henrissat, B. (1998) A
classification of nucleotide-diphospho-sugar glycosyltransferases
based on amino acid sequence similarities. Biochem. J. 329,719–
719.
44. Gastinel, L.N., Bignon, C., Misra, A.K., Hindsgaul, O., Shaper,
J.H. & Joziasse, D.H. (2001) Bovine a1,3-galactosyltransferase
catalytic domain structure and its relationship with ABO histo-
blood group and glycosphingolipid glycosyltransferases. EMBO J.