Glycobiology vol. 19 no. 6 pp. 655–664, 2009
Advance Access publication on March 3, 2009
Association of β-1,3-N-acetylglucosaminyltransferase 1 and β-1,4-galactosyltransferase
1, trans-Golgi enzymes involved in coupled poly-N-acetyllactosamine synthesis
Peter L Lee2, Jennifer J Kohler3, and Suzanne R Pfeffer1,2
2Department of Biochemistry, Stanford University School of Medicine,
Stanford, CA 94305; and3Division of Translational Research, Department of
Internal Medicine, University of Texas Southwestern Medical Center, Dallas,
TX 75390, USA
Received on November 7, 2008; revised on February 12, 2009; accepted on
February 27, 2009
and galactose residues involved in cellular functions rang-
ing from differentiation to metastasis. PolyLacNAc also
serves as a scaffold on which other oligosaccharides such
as sialyl Lewis X are displayed. The polymerization of the
alternating N-acetylglucosamine and galactose residues is
catalyzed by the successive action of UDP-GlcNAc:βGal
β-1,3-N-acetylglucosaminyltransferase 1 (B3GNT1) and
1 (B4GALT1), respectively. The functional association be-
tween these two glycosyltransferases led us to investigate
whether the enzymes also associate physically. We show that
B3GNT1 and B4GALT1 colocalize by immunofluorescence
microscopy, interact by coimmunoprecipitation, and affect
each other’s subcellular localization when one of the two
proteins is artificially retained in the endoplasmic reticu-
physically associate in vitro and in cultured cells, providing
insight into possible mechanisms for regulation of polyLac-
Keywords: endoplasmic reticulum/enzyme complexes/
A fundamental question in oligosaccharide assembly is how
cells template the synthesis of specific carbohydrate structures.
As proteins and lipids pass through each cisterna of the Golgi
complex, they encounter distinct glycosyltransferases that gen-
erate substrates for modification by subsequently encountered
enzymes (Munro 1998; de Graffenried and Bertozzi 2004;
Maccioni 2007). Despite an enormous amount of work by many
laboratories, how glycosyltransferases achieve their unique
localizations is still poorly understood (reviewed by Colley
(1997); Opat et al. (2001); and de Graffenried and Bertozzi
1To whom correspondence should be addressed: e-mail: Pfeffer@stanford.edu
Two nonexclusive mechanisms have been proposed to ex-
plain how the type II single transmembrane-spanning glyco-
syltransferases are properly localized. According to the “bilayer
thickness” model, enzymes sort according to the size match
between the lengths of their transmembrane domains and the
local membrane thickness (Bretscher and Munro 1993; Mitra
that are localized in a common compartment interact to form
complexes that are excluded from inclusion in forward moving
cargo (Machamer 1991; Nilsson et al. 1993; Opat et al. 2000).
These models are not mutually exclusive: membrane thickness
could contribute to the localization of enzyme complexes while
enzyme complexes could establish domains of particular mem-
Importantly, the known hetero-oligomeric kin recogni-
tion complexes are composed of glycosyltransferases that
catalyze successive processing reactions (de Graffenried
and Bertozzi 2004) as proposed by Roseman (1970). The
first reported case of association was between α-1,3-1,6-
mannosidase II and β-1,2-N-acetylglucosaminyltransferase I; in
this case, the action of β-1,2-N-acetylglucosaminyltransferase
I is a necessary precondition for α-1,3-1,6-mannosidase-
Moremen 2002). Associations have also been reported in
glycolipid synthesis: β-1,4-N-acetylgalactosaminyltransferase
1 (B4GALNT1) and the UDP-Gal:GA2/GM2/GD2 β-1,3-
galactosyltransferase physically associate (Giraudo et al.
2001), and UDP-Gal:glucosylceramide galactosyltransferase,
CMP-NeuAc:lactosylceramidesialyltransferase, and ST8 α-N-
acetylneuraminide α-2,8-sialyltransferase 1 (ST8SIA1) have
ST8SIA1 can also associate with B4GALNT1 in a different cell
line (Bieberich et al. 2002). Glycosaminoglycan biosynthetic
enzymes exostosin 1 and 2 associate with one another as do
uronosyl 5-epimerase and iduronic acid 2-O-sulfotransferase
(McCormick et al. 2000; Pinhal et al. 2001). Hetero-oligomeric
glycosyltransferase complexes have been identified in organ-
isms from Saccharomyces cerevisiae to humans (Jungmann and
Munro 1998; Jungmann et al. 1999). Thus, kin recognition may
represent an important strategy to direct substrate traffic and
prevent off-target glycosylation events.
Poly-N-acetyllactosamine (polyLacNAc) is a linear carbohy-
drate polymer composed of alternating galactose (Gal) and N-
acetylglucosamine (GlcNAc) residues (Figure 1). This polysac-
charide can be incorporated into either N-linked or mucin-type
tosis, and metastasis (Kasai and Hirabayashi 1996; Elola et al.
2005). PolyLacNAc polymers can be further modified by vari-
ous glycosyltransferases to create branched structures and dis-
play terminal epitopes such as the sialyl Lewis X modification,
c ?The Author 2009. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: email@example.com
P L Lee et al.
Fig. 1. Structure of poly-N-acetyllactosamine. Poly-N-acetyllactosamine is composed of alternating residues of galactose-linked β1→4 and
N-acetylglucosamine-linked β1→3. Sugars are color-coded in gray and black, respectively.
PolyLacNAc is biosynthesized by the alternating ad-
Two of the primary glycosyltransferases within the B3GNT
and B4GALT families are UDP-GlcNAc:βGal β-1,3-N-
(B4GALT1), respectively (Trayer and Hill 1971; Narimatsu
et al. 1986; Shaper et al. 1986; Sasaki et al. 1997). While little is
known about the localization mechanism for B3GNT1, several
studies have suggested that proper B4GALT1 localization
requires all or part of its cytoplasmic and transmembrane
domains (Nilsson et al. 1991; Aoki et al. 1992; Russo et al.
1992; Teasdale et al. 1992; Evans et al. 1993; Masibay et al.
1993; Yamaguchi and Fukuda 1995). The linear and repetitive
characteristics of the polyLacNAc structure have led to the
hypothesis that the glycosyltransferases involved may form
complexes to aid in polyLacNAc assembly (de Graffenried and
Bertozzi 2004). Indeed, Seko and Yamashita (2008) recently
demonstrated the interaction of two glycosyltransferases,
B3GNT2 and B3GNT8, from the B3GNT family and showed
that the presence of B3GNT8 can stimulate the activity of
preference for the GlcNAcβ1→2Man branch while B4GALT1
shows a complementary preference for the GlcNAcβ1→6Man
branch. The equal prevalence of polyLacNAc on both branches
therefore suggests tight functional association between the two
enzymes (Ujita et al. 1999). Thus, we sought to determine
whether these enzymes, from the complementary families of
polyLacNAc catalyzing glycosyltransferases, also physically
associate. Using coimmunoprecipitation and an endoplasmic
reticulum (ER) retention assay adapted from an approach
developed by Nilsson et al. (1994), we show that B3GNT1 and
B4GALT1 interact with each other in the trans-Golgi.
1 (B3GNT1)and UDP-
B3GNT1 and B4GALT1 colocalize
To investigate the mechanism of polyLacNAc synthesis, we ex-
plored the potential association of two of the primary glycosyl-
transferases, B3GNT1 and B4GALT1, involved in polyLacNAc
production. To track subcellular localization of these enzymes,
we employed plasmids encoding B3GNT1 with a myc epitope
tag at its C-terminus and B4GALT1 with an HA epitope tag
at its C-terminus. The C-termini of these type II membrane
proteins are expected to reside in the lumen of the secretory
Fig. 2. Immunofluorescence microscopy of B4GALT1 and B3GNT1 in HeLa
cells. (A) B4GALT1 transfected cell (left panel, green in merge) with
endogenous Golgin245 (middle panel, red in merge). (B) B3GNT1 (left panel,
green in merge) and B4GALT1 (middle panel, red in merge) cotransfected cell.
(C–E) B3GNT1 transfected cells (left panels, green in merge) with
endogenous GCC185 (C), Golgin245 (D), or GM130 (E) (middle panels, red
in merge). White outlines show cell boundaries. Scale bar is 10 μm.
pathway. HeLa cells were utilized for microscopy experiments
because of their generally flat morphology; COS-1 cells were
used in coimmunoprecipitation experiments because of their
high transfection efficiency and good protein expression.
In order to associate in a physiologically relevant manner,
ment of the Golgi complex. Control experiments (Figure 2A)
Interaction of B3GNT1 and B4GALT1
Fig. 3. Reciprocal physical association of B4GALT1 and B3GNT1.
(A) Indicated COS-1 cell extracts were subjected to immunoprecipitation with
an anti-HA antibody. Input (2%) and immunoprecipitated fractions were
immunoblotted with an anti-myc antibody. (B) COS-1 cells transfected as
indicated were immunoprecipitated using an anti-myc antibody. Input (1%) and
immunoprecipitated fractions were immunoblotted with an anti-HA antibody.
demonstrated that exogenously expressed B4GALT1-HA lo-
calized near the trans-Golgi network (TGN) marker Gol-
gin245 supporting its well-established trans-Golgi localization
(Roth and Berger 1982; Gleeson et al. 1996; Llopis et al.
1998). To determine the localization of B3GNT1, we compared
the colocalization of B3GNT1 with several different markers:
B4GALT1, GCC185, Golgin245, and GM130. As shown in
Figure 2B, B3GNT1-myc and B4GALT1-HA colocalized
strongly with Pearson’s coefficient of 0.87 ± 0.02. Moreover,
B3GNT1-myc colocalized to some extent with the TGN marker
D) (Luke et al 2003). B3GNT1-myc colocalized least well with
the cis-Golgi marker GM130 (Pearson’s coefficient of 0.82 ±
0.03) (Figure 2E) (Nakamura et al. 1995). The trend in colocal-
the TGN and further from the cis-Golgi complex.
B3GNT1 and B4GALT1 coimmunoprecipitate
Next, we conducted coimmunoprecipitation experiments to test
whether B3GNT1 and B4GALT1 associate with each other. To
test whether B4GALT1 could precipitate B3GNT1, the epitope-
and lysates were incubated with an anti-HA antibody to precip-
itate B4GALT1-HA and any associated proteins. A subsequent
munoprecipitated B3GNT1-myc. As shown in Figure 3A, lane
6, B3GNT1-myc was coimmunoprecipitated with B4GALT1-
HA only when both proteins were present in the extract.
We also tested whether B3GNT1 could conversely precipi-
tate B4GALT1. Transfected COS-1 cell lysates were incubated
with an anti-myc antibody to precipitate B3GNT1-myc and its
associated proteins. Immunoblot analysis of the precipitated
proteins with an anti-HA antibody revealed that B4GALT1-HA
was coimmunoprecipitated by the anti-myc antibody only in the
presence of B3GNT1-myc (Figure 3B, lane 13). It is interesting
to note that two B4GALT1-HA bands were coimmunoprecip-
itated. Furthermore, there was a larger amount of the higher
molecular weight species (lanes 2 and 5), and the B3GNT1
precipitation seems enriched for the higher molecular weight
B4GALT1 band (lane 13). As discussed below, the larger forms
may represent more highly glycosylated forms of B4GALT1.
two enzymes that function together in polyLacNAc synthesis.
KDEL sequences cause glycosyltransferases to localize
to the ER
To determine if the physical interaction between B3GNT1 and
B4GALT1 could be demonstrated in cultured cells, we utilized
a cell-based relocalization assay. Previously, Warren and col-
glycosyltransferase interactions using an ER retention assay
(Nilsson et al. 1994). Specifically, one enzyme is retained in the
ER through an N-terminal fusion of a cytoplasmically oriented
retention signal provided by the ER-localized human invariant
chain p33. Changes in localization of putative binding partners
were then examined for complementary relocalization to the
ER. A change in localization, from the Golgi to the ER, was
interpreted as being indicative of an interaction between the two
Rather than modifying the native cytoplasmic domain se-
quences that may include localization information (Russo
et al. 1992; Evans et al. 1993), we used a lumenally oriented
C-terminal KDEL retention signal to relocalize Golgi-resident
glycosyltransferases to the ER. A similar approach was used
by Munro (1995) to investigate the retention determinants of
medial and trans-Golgi glycosyltransferases. KDEL sequences
are typically found at the C-termini of soluble, ER-resident pro-
teins (Pelham 1990). However, the unusual type II transmem-
brane topology of glycosyltransferases permits their lumenal
C-termini to interact with the KDEL receptor, thereby making
it possible to use a C-terminal KDEL sequence to recruit gly-
cosyltransferases to the ER. Indeed, KDEL-related sequences
resident glycosyltransferases (Heinonen et al. 2003; Okajima
et al. 2005; Heinonen et al. 2006 but see also Moloney and
Expression plasmids were generated that encode the amino
acid sequence SEKDEL at the C-termini of epitope-tagged
B3GNT1, B4GALT1, and ST8SIA1 constructs. ST8SIA1,
another trans-Golgi resident glycosyltransferase, participates
in the orthologous glycolipid biosynthesis pathway (Daniotti
et al. 2000). Immunofluorescence microscopy was used to de-
termine the capacity of the SEKDEL sequence to relocalize the
normally Golgi-associated glycosyltransferases to the ER. As
shown in Figure 4 (top-right column), B4GALT1-HA-KDEL,
B3GNT1-myc-KDEL, and ST8SIA1-myc-KDEL were all ef-
ficiently relocalized to the ER compared with the correspond-
ing non-KDEL-tagged constructs (top-left column). Relocal-
ization was still observed after 2 h of cycloheximide treatment
(100 μg/mL) prior to fixation, suggesting that the localization
was stable and did not reflect proteins in the process of fold-
ing prior to ER export. The cycloheximide incubation did not
perturb the normal Golgi localization of the non-KDEL-tagged
P L Lee et al.
Fig. 4. KDEL relocalizes glycosyltransferases. Representative images show
the localization of non-KDEL-tagged and KDEL-tagged B4GALT1, B3GNT1,
and ST8SIA1 (top). White outlines show cell boundaries as determined by
phase contrast microscopy. Scale bar, 20 μm. The percent ER localization was
quantified (bottom). Error bars represent 95% confidence intervals from a
sampling of ≥47 cells per condition (number of cells analyzed is shown below
Quantitative analysis of protein localization
A metric was established to permit quantitative analysis of ER
versus Golgi localization for each glycosyltransferase. First, we
approximated the cell as a flat two-dimensional structure and
defined three compartments: nucleus, Golgi, and ER (bulk cyto-
plasm). Total cell intensity was determined by summing the sig-
edge; background staining and cell autofluorescence, calculated
from cells that had been immunostained but not transfected,
was subtracted. The intensity within the Golgi and nuclear ar-
eas was determined using masks derived from a Golgi marker
protein, GM130, and a nuclear marker, DAPI, respectively. Any
signal not localized to the nuclear or Golgi area was defined
as ER. During instances where the Golgi and nucleus slightly
overlapped, the signal was attributed to the Golgi. Golgi and
nuclear areas were defined liberally to decrease the possibility
of attributing signal derived from these areas to the ER. It is im-
portant to note that ER intensity will be underestimated by this
method because some ER intensity overlaps with the nuclear
envelope and Golgi areas.
As quantified at the bottom of Figure 4, the addition of a
shifted their distributions to the ER compared with their un-
tagged counterparts even after 2 h of cycloheximide treatment.
Golgi-localized B4GALT1-HA showed a mean ER fraction of
ER fraction to 0.59, a 4.8-fold increase. Similarly, the addi-
tion of KDEL to B3GNT1-myc shifted the mean ER fraction
from 0.10 to 0.60, a 6-fold enhancement. The mean fraction of
ST8SIA1-myc intensity localized to the ER was 0.05. KDEL-
tagging ST8SIA1-myc raised this value to 0.53, a 10.6-fold
difference. These data demonstrate that the addition of KDEL
can efficiently relocalize these type II Golgi resident glycosyl-
transferases to the ER.
KDEL-tagged B3GNT1 and B4GALT1 can also
Coimmunoprecipitation experiments were carried out as an in-
and B4GALT1 can also associate in vitro. B3GNT1-myc or
B3GNT1-myc-KDEL and their associated proteins were pre-
cipitated from transfected COS-1 cell lysates with an anti-
myc antibody. Subsequent probing of an immunoblot with an
anti-HA antibody revealed that B4GALT1-HA was efficiently
precipitated by B3GNT1-myc-KDEL, and B4GALT1-HA-
KDEL was precipitated by both KDEL-tagged and non-KDEL-
tagged B3GNT1-myc (Figure 3B, lanes 14–16). Interestingly,
theKDEL-tagged B4GALT1-HA was predominantly present as
myc- and B3GNT1-myc-KDEL-precipitated B4GALT1-HA-
KDEL was enriched in the higher molecular weight form of
B4GALT1 (lane 14 and 16). While one molecular weight form
of B4GALT1-HA-KDEL is enriched, the KDEL-tagged con-
structs can efficiently associate with one another as well as their
B3GNT1 and B4GALT1 can recruit one another to the ER
Possible interactions between glycosyltransferases were tested
in our cell-based assay by cotransfecting a non-KDEL-tagged
glycosyltransferase with a KDEL-tagged glycosyltransferase
and monitoring the localization of the non-KDEL-tagged gly-
cosyltransferase after 2 h of cycloheximide treatment. When
non-KDEL-tagged B4GALT1 and B3GNT1 were coexpressed,
B4GALT1 was localized to the Golgi complex (Figure 5, upper
left). In contrast, cotransfection of B4GALT1 with B3GNT1-
KDEL readily shifted B4GALT1’s localization to the ER (Fig-
ure 5, upper left). This effect was specific, as cotransfection of
the orthologous ST8SIA1 with B3GNT1-KDEL failed to relo-
calize ST8SIA1 (Figure 5, upper left). Similarly, coexpression
of B3GNT1 with B4GALT1 did not perturb B3GNT1’s normal
Interaction of B3GNT1 and B4GALT1
Fig. 5. B4GALT1 and B3GNT1 affect each other’s subcellular localization. The localization of the one glycosyltransferase (within image) was examined to
determine the ability of a partner glycosyltransferase (indicated above image) to recruit the shown glycosyltransferase to the ER. White outlines show the cell
boundaries as determined by phase contrast microscopy. Scale bar is 20 μm.
to the ER (Figure 5, upper right).As expected, ST8SIA1-KDEL
was not capable of relocalizing B4GALT1 to the ER (Figure 5,
The ability of two proteins to alter each others’ localiza-
tions is likely to be dependent upon their relative levels of
expression. Because this aspect of our experiment is difficult
to control, we carried out careful quantitation of at least 49
cells for each cotransfected pair to ensure reliability of our
observations. Figure 6 (top-left column) shows the fraction of
the ER-localized B4GALT1 signal for individual HeLa cells
cotransfected with B4GALT1 and B3GNT1. The peak of the
distribution was centered at 0.06, consistent with the Golgi lo-
calization of B4GALT1. When B4GALT1 was cotransfected
with B3GNT1-KDEL, the fraction of B4GALT1 found in the
ER tended to increase: the mean of the distribution of ER
fraction shifted to 0.27 and the distribution overall broadened
(Figure 6, middle-left column). The distribution of ER frac-
tion of B4GALT1-KDEL alone is provided as a positive con-
trol and was centered about 0.59 (bottom panel). This repre-
sents the maximum shift that could have been observed in these
Similarly, when B3GNT1 was cotransfected with B4GALT1
(Figure 6, top-right column), the fraction of B3GNT1 local-
ized to the ER was distributed around a peak centered at 0.15.
Cotransfection with B4GALT1-KDEL resulted in a broadened
ER-fraction distribution with a mean of 0.30 (Figure 6, middle-
right column). Again, the maximum shift that could have been
observed was provided by the B3GNT1-KDEL-alone control,
data show that the presence of the KDEL-tagged partner causes
a normally Golgi-localized protein to be retained in the ER in a
large representative sampling of transfected cells.
One limitation of our approach is the fact that KDEL-
terminating proteins often leave the ER and are normally re-
trieved by the KDEL receptor. In the course of retrieval, an arti-
ficially KDEL-tagged glycosyltransferase would be expected to
traverse the Golgi compartments where it might interact prefer-
entially with untagged counterparts. Thus, we also analyzed our
data by normalizing for this possibility. Specifically, for each
non-KDEL-tagged glycosyltransferase, we calculated the frac-
partner protein’s ER fraction. This would enable a direct com-
any variation in ER retention between different KDEL-tagged
When B3GNT1 was transfected with B4GALT1-KDEL, we
detected an average normalized, ER localization value of 0.85
for B3GNT1. Similarly, when B4GALT1 was transfected with
B3GNT1-KDEL, we observed a mean normalized ER localiza-
0.36 when ST8SIA1-KDEL was cotransfected with B4GALT1
or B4GALT1-KDEL was cotransfected with ST8SIA1.
P L Lee et al.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Number of Cells
B4GALT1 (+ B3GNT1−KDEL)
B3GNT1 (+ B4GALT1−KDEL)
B4GALT1 (+ B3GNT1)
B3GNT1 (+ B4GALT1)
Fig. 6. Fraction of ER localization of glycosyltransferases in individual cells. HeLa cells were transiently transfected and imaged as in Figure 4 and the fraction of
the glycosyltransferase residing in the ER was quantified and plotted on a histogram. In each plot, the signal from the indicated glycosyltransferase was quantified
under the cotransfection conditions indicated in parenthesis. More than 48 cells were quantified per condition.
Similarly, ST8SIA1 was not relocalized to the ER by B3GNT1-
this association is stronger than that of ST8SIA1 with either
enzyme (Figure 7).
We have reported here the use of two independent methods to
demonstrate the association of B3GNT1 and B4GALT1 within
the secretory pathway. Both coimmunoprecipitation and cel-
lular relocalization assays provide evidence for an interaction
between these two enzymes. The physical association between
B3GNT1 and B4GALT1 likely contributes to the functional
coupling of these two enzymes in producing the polyLacNAc
The distinctive structures of glycosyltransferases, namely
their common type II transmembrane orientation, made it pos-
sible to adapt the ER retention assay (Nilsson et al. 1994) by
employing a C-terminal KDEL tag for recruitment. C-terminal
KDEL tags are well-established cellular signals for soluble pro-
tein retention to the ER (Pelham 1990) and have been exploited
previously to relocalize type II transmembrane proteins to the
ER (Munro 1995). Our data show that the KDEL tag can be
applied to other Golgi glycosyltransferases and used to retain
trans-Golgi glycosyltransferases and their partners in the ER.
Recent reports have suggested that Golgi localization of gly-
cosyltransferases can depend on native cytoplasmic signals
(Grabenhorst and Conradt 1999; Milland et al. 2001, 2002;
Giraudo and Maccioni 2003b; Schmitz et al. 2008; Tu et al.
2008). While the original p33-mediated ER retention assay
was able to detect interactions between medial glycosyltrans-
ferases (Nilsson et al. 1994), the use of a KDEL motif has
advantages over the original system. First, the KDEL fusion re-
quires a smaller overall change to the chimeric protein: fusion
of 6 amino acids (SEKDEL), compared to the addition of the
protein. Second, the addition of a KDEL motif does not involve
replacement of any amino acids. Lastly, the KDEL motif does
not necessitate the removal of glycosyltransferase cytoplasmic
domains that may contribute to protein localization. The effi-
ciency of partner protein ER-relocalization is likely to depend
on the individual glycosyltransferases studied, and the p33 sys-
tem may be more effective than the KDEL motif for certain
It is worth noting that the relative efficiency of relocaliza-
tion for trans-Golgi partners versus medial Golgi partners can-
not be compared directly since the different native pH envi-
ronments may influence the ability of such partners to achieve
Interaction of B3GNT1 and B4GALT1
Fig. 7. B4GALT1–B3GNT1 interaction results in a higher percentage of
relocalization to the ER when compared to interactions with non-related
glycosyltransferases. Shown is the mean ER relocalization of the
non-KDEL-tagged partner (indicated below each bar) normalized by the
efficiency with which the KDEL-tagged protein (indicated in parenthesis) is
itself ER retained. Error bars represent 95% confidence intervals from a
sampling of n = 51, 50, 52, 50, and 50, respectively.
that β-galactoside α-2,6-sialyltransferase 1, a trans-Golgi
resident enzyme, formed insoluble oligomers only when har-
vested under pH conditions similar to those found at the trans-
Golgi (Chen et al. 2000). The same pH gradient may explain
why the KDEL-mediated ER retention assay reported here did
not result in the relocalization of 100% of the examined trans-
an anti-HA antibody revealed two isoforms: one at the expected
molecular weight for unmodified B4GALT1-HA (45 kDa) and
another at a higher molecular weight (approximately 53 kDa)
ular weight band represents the presence of an extended glycan
modification added during passage through the Golgi. This hy-
transfected with B4GALT1-HA-KDEL contain predominantly
the lower molecular weight form (Figure 3B, lanes 3, 6, 8),
since the KDEL-tagged isoform is less likely to traffic through
the Golgi and be modified by glycan extending enzymes. We
were intrigued to observe that B3GNT1-myc and B3GNT1-
myc-KDEL both preferentially precipitated the higher molecu-
lar weight isoform of B4GALT1-HA-KDEL (Figure 3B, lanes
14 and 16). This finding suggests that an extended glycan on
B4GALT1 may enhance binding to B3GNT1.
B4GALT1 transmembrane residues Cys29 and His32 are re-
quired for Golgi retention and appear to also contribute to
B4GALT1 homodimerization (Aoki et al. 1992; Yamaguchi
and Fukuda 1995). An intriguing possibility is that these same
B3GNT1. Also noteworthy is B3GNT1’s unusually long, 28-
amino-acid transmembrane domain, which is particularly unex-
domain to stabilize B3GNT1 localization to the trans-Golgi.
Future experiments will be needed to test these possibilities
directly. Furthermore, experiments using purified components
will help to address whether the interaction observed is direct.
The present study has focused on two of the main enzymes
that cooperate in polyLacNAc synthesis. However, other galac-
tosyltransferases also participate in this biosynthetic process
(Lo et al. 1998; Hennet 2002). Preliminary results using the
ER recruitment assay described here provide evidence for an
association between B3GNT1 and UDP-Gal:βGlcNAc β-1,4-
galactosyltransferase, polypeptide 4, B4GALT4, in cotrans-
fected cells (data not shown). In vitro, B4GALT1 has been
shown to display substrate preference for N-linked and core
4 O-linked glycans while B4GALT4 has substrate preference
for core 2 O-linked glycans (Ujita et al. 2000). B4GALT1
and B4GALT4 might therefore compete for association with
B3GNT1 and bias B3GNT1 substrate preference in vivo. More
B4GALT complexes will aid in investigating this hypothesis.
Alternatively, B3GNT1, B4GALT1, and B4GALT4 may asso-
ciate with each other within a single, large heterocomplex that
By disrupting or strengthening associations between B3GNT1
and B4GALT1 or between B3GNT1 and B4GALT4, it may be
possible to alter the glycan synthetic capacity in cells. Further-
more, engineering an enzyme containing both B3GNT1 and
B4GALT1 catalytic domains may generate increased polyLac-
NAc length by increasing the processivity of the reaction and
allow further studies on how polyLacNAc length affects its bi-
In conclusion, we have shown the first example of enzyme
association among trans-Golgi-localized glycosyltransferases
using an ER retention assay that does not alter the N-terminal
cytoplasmic, transmembrane, and stem domains of these type
II transmembrane proteins. The physical association between
these two glycosyltransferases likely contributes to the local-
ization of both B3GNT1 and B4GALT1 to the trans-Golgi and,
importantly, to regulating the production of polyLacNAc from
Material and methods
Cell culture and transfections
HeLa and African green monkey kidney fibroblast (COS-1)
cells were maintained in Dulbecco’s modified Eagle’s medium
mycin. HeLa and COS-1 cells were transfected using FuGENE
6 (Roche Applied Science, Indianapolis, IN).
by DNA sequencing. Plasmid DNA was prepared from large-
Valencia, CA). Restriction enzymes were purchased from
New England Biolabs (Ipswich, MA).
B4GALT1-HA was amplified from pBS-β1,4 GT-1 kindly
provided by Dr. Michiko Fukuda (The Burnham Institute, La
Jolla, CA) by PCR with a 3?primer containing the HA epitope
tag, digested, and ligated into pcDNA 3.1 Zeo (+) (Invitro-
gen, Carlsbad, CA) for mammalian expression. B4GALT1-HA-
KDEL was amplified from B4GALT1-HA using a 3?primer
P L Lee et al.
containing codons for the amino acid sequence SEKDEL and
ligated into pcDNA 3.1. B3GNT1-myc was amplified from
pcDNA3.1-iGnT (B3GNT1) kindly provided by Dr. Minoru
Fukuda (The Burnham Institute, La Jolla, CA) by PCR us-
ing a 3?primer containing the myc epitope tag, digested, and
ligated into pcDNA 3.1 (+). B3GNT1-myc-KDEL was ampli-
fied from B3GNT1-myc in an analogous manner as described
for B4GALT1-HA-KDEL. Similar to the method used to pro-
duce B4GALT1 and B3GNT1 constructs above, ST8SIA1-HA,
ST8SIA1-myc, and ST8SIA1-myc-KDEL were made by PCR
amplifying the respective genes with primers encoding the ap-
propriate tags. ST8SIA1 constructs were kindly provided by
Chad Whitman (UT Southwestern, Dallas, TX).
For colocalization studies, HeLa cells were grown on coverslips
and transfected with B3GNT1-myc and/or B4GALT1-HA for
24 h. Cells were fixed using 3.7% formaldehyde and permeabi-
lized with 0.1% Triton X-100. Cells were stained with combi-
nations of chicken anti-Myc antibodies (1:750, Bethyl Labora-
tories, Montgomery, TX), rabbit anti-HA antibodies (1:1000,
Abcam, Cambridge, MA), mouse anti-GM130 antibodies
(1:500, BD Biosciences, San Jose, CA), mouse anti-Golgin245
antibodies (p230 trans-Golgi, BD Biosciences, 1:400), and rab-
bit anti-GCC185 antibodies (Reddy et al. 2006). Secondary im-
munostaining was conducted with goat anti-chicken antibodies
croscope (Nikon, Tokyo, Japan) equipped with epifluorescence
optics using a 100× numerical aperture 1.40 plan apochromat
oil immersion objective and a charge-coupled device camera
(CoolSnapHQ; Photometrics, Tucson, AZ) and MetaMorph im-
age acquisition software (Molecular Devices, Sunnyvale, CA).
Pearson’s coefficients were calculated using the Manders’ Co-
efficients plugin (Tony Collins, Wayne Rasband) for ImageJ
(NIH, Bethesda, MD).
To quantify KDEL-induced relocalization, HeLa cells were
with 100 μg/mL cycloheximide for 2 h to clear the ER of newly
synthesized proteins. Cells were subsequently fixed 24 h post-
transfection, stained and mounted as above with the exception
of different secondary antibodies: goat anti-chicken antibodies
conjugated to Alexa 633, goat anti-mouse antibodies conju-
gated to Alexa 488 or Alexa 610, and goat anti-rabbit antibod-
ies conjugated to Alexa 488 or Alexa 610 (Invitrogen). Fluores-
(Carl Zeiss, Thornwood, NY) equipped with phase contrast and
1.3 oil immersion objective and a back-thinned cooled charge-
coupled device camera (Princeton Instruments, Trenton, NJ)
(Adobe Systems, Mountain View, CA) and ImageJ (NIH). Cell
masks were defined by manually outlining the phase contrast
image of the cell. Nucleus masks were defined by thresholding
on the DAPI image of the cell. Golgi masks were defined by
thresholding on the GM130 image or the Golgi-localized sig-
nal from the glycosyltransferase. MATLAB (The MathWorks,
Natick, MA) was used for image quantitation and automated
mask joining algorithms.
COS-1 cells grown on 10 cm plates were transiently trans-
fected with combinations of B3GNT1-myc, B4GALT1-HA,
and their KDEL-tagged plasmids for 24 h. After two washes
with phosphate-buffered saline, cells were harvested in 1%
CHAPS, 30 mM Tris-HCl, pH 7.5, 150 mM NaCl sup-
plemented with protease inhibitors (complete, EDTA free,
Roche Applied Science). Protein extracts were centrifuged at
55,000 × g for 15 min, and supernatants collected. Bovine
serum albumin (0.1%) was added to the supernatants, which
were then precleared with 30 μL Protein A-agarose (Roche
Applied Science) for 20 min at 4◦C. A rabbit anti-HA antibody
(Abcam) or rabbit anti-myc antibody (Bethyl Laboratories) was
added to a final concentration of 10 μg/mL and incubated for
for 20 min at 4◦C. The beads were washed once in 30 mM Tris-
HCl, pH 7.5, 150 mM NaCl and boiled in a 20 μL SDS–PAGE
sample buffer. Samples were then resolved by SDS–PAGE and
a mouse anti-myc antibody conjugated to horseradish peroxi-
dase (Invitrogen) or a rabbit anti-HA antibody conjugated to
horseradish peroxidase (Bethyl Laboratories) and detected us-
ing the ECL Plus Western blotting detection kit (GE Healthcare
Bio-Sciences, Piscataway, NJ) and HyBlot CL autoradiography
film (Denville Scientific Inc., Metuchen, NJ).
The National Institutes of Health (R37 DK37332 to S.R.P.);
Stanford University; the Camille and Henry Dreyfus Founda-
tion (New Faculty Award to J.J.K.); and the National Science
Foundation (to P.L.L.).
We thank Dr. Minoru Fukuda for plasmids and reagents,
Dr. Michiko Fukuda for plasmids, and Chad Whitman for
the ST8SIA1 constructs. PLL is grateful to Dr. J. Theriot for
continued support and advice.
Conflict of interest statement
transferase; B3GNT1, UDP-GlcNAc:βGal β-1,3-N-acetylgluco
saminyltransferase 1; B4GALT, UDP-Gal:βGlcNAc β-1,4-
galactosyltransferase; B4GALT1, UDP-Gal:βGlcNAc β-1,4-
1; ER, endoplasmic reticulum; Gal, galactose; GlcNAc,
Interaction of B3GNT1 and B4GALT1
N-acetylglucosamine; polyLacNAc, poly-N-acetyllactosamine;
ST8SIA1, ST8 α-N-acetylneuraminide α-2,8-sialyltransferase
1; TGN, trans-Golgi network.
Aoki D, Lee N, Yamaguchi N, Dubois C, Fukuda MN. 1992. Golgi retention
of a trans-Golgi membrane protein, galactosyltransferase, requires cysteine
and histidine residues within the membrane-anchoring domain. Proc Natl
Acad Sci USA. 89(10):4319–4323.
Bieberich E, MacKinnon S, Silva J, Li DD, Tencomnao T, Irwin L, Kapitonov
D, Yu R. 2002. Regulation of ganglioside biosynthesis by enzyme
complex formation of glycosyltransferases. Biochemistry. 41(38):11479–
Bretscher MS, Munro S. 1993. Cholesterol and the Golgi apparatus. Science.
Chen C, Ma J, Lazic A, Backovic M, Colley KJ. 2000. Formation of insoluble
oligomers correlates with ST6Gal I stable localization in the Golgi. J Biol
Colley KJ. 1997. Golgi localization of glycosyltransferases: More questions
than answers. Glycobiology. 7(1):1–13.
functional role. Glycoconj J. 17(10):669–676.
Daniotti JL, Martina JA, Giraudo CG, Zurita AR, Maccioni HJF. 2000. GM3
golgi location in CHO-K1 cells. J Neurochem. 74(4):1711–1720.
de Graffenried CL, Bertozzi CR. 2004. The roles of enzyme localisation and
complex formation in glycan assembly within the Golgi apparatus. Curr
Opin Cell Biol. 16(4):356–363.
Elola MT, Chiesa ME, Alberti AF, Mordoh J, Fink NE. 2005. Galectin-1 recep-
tors in different cell types. J Biomed Sci. 12(1):13–29.
Evans S, Lopez L, Shur B. 1993. Dominant negative mutation in cell surface
J Cell Biol. 120(4):1045–1057.
Giraudo CG, Daniotti JL, Maccioni HJF. 2001. Physical and functional asso-
ciation of glycolipid N-acetyl-galactosaminyl and galactosyl transferases in
the Golgi apparatus. Proc Natl Acad Sci USA. 98(4):1625–1630.
Giraudo CG, Maccioni HJF. 2003a. Ganglioside glycosyltransferases orga-
nize in distinct multienzyme complexes in CHO-K1 cells. J Biol Chem.
Giraudo CG, Maccioni HJF. 2003b. Endoplasmic reticulum export of glyco-
syltransferases depends on interaction of a cytoplasmic dibasic motif with
Sar1. Mol Biol Cell. 14(9):3753–3766.
Gleeson PA, Anderson TJ, Stow JL, Griffiths G, Toh BH, Matheson F. 1996.
p230 is associated with vesicles budding from the trans-Golgi network.
J Cell Sci. 109(12):2811–2821.
Grabenhorst E, Conradt HS. 1999. The cytoplasmic, transmembrane, and stem
regions of glycosyltransferases specify their in vivo functional sublocaliza-
tion and stability in the Golgi. J Biol Chem. 274(51):36107–36116.
Hakomori S. 1999. Antigen structure and genetic basis of histo-blood groups
A, B and O: Their changes associated with human cancer. Biochim Biophys
Heinonen TYK, Pasternack L, Lindfors K, Breton C, Gastinel LN, M¨ aki M,
Kainulainen H. 2003. A novel human glycosyltransferase: Primary struc-
ture and characterization of the gene and transcripts. Biochem Biophys Res
Heinonen TYK, Pelto-Huikko M, Pasternack L, M¨ aki M, Kainulainen H. 2006.
Murine ortholog of the novel glycosyltransferase, B3GTL: Primary struc-
ture, characterization of the gene and transcripts, and expression in tissues.
DNA Cell Biol. 25(8):465–474.
Hennet T. 2002. The galactosyltransferase family. Cell Mol Life Sci.
Jungmann J, Munro S. 1998. Multi-protein complexes in the cis Golgi of Sac-
charomyces cerevisiae with alpha-1,6-mannosyltransferase activity. EMBO
Jungmann J, Rayner JC, Munro S. 1999. The Saccharomyces cerevisiae protein
Mnn10p/Bed1p is a subunit of a Golgi mannosyltransferase complex. J Biol
glycocodes. J Biochem. 119(1):1–8.
Kozma K, Keusch JJ, Hegemann B, Luther KB, Klein D, Hess D, Haltiwanger
RS, Hofsteenge J. 2006. Identification and characterization of a beta1,3-
glucosyltransferase that synthesizes the Glc-beta1,3-Fuc disaccharide on
thrombospondin type 1 repeats. J Biol Chem. 281(48):36742–36751.
Llopis J, McCaffery JM, Miyawaki A, Farquhar MG, Tsien RY. 1998. Mea-
surement of cytosolic, mitochondrial, and Golgi pH in single living cells
with green fluorescent proteins. Proc Natl Acad Sci USA. 95(12):6803–
Lo NW, Shaper JH, Pevsner J, Shaper NL. 1998. The expanding beta 4-
galactosyltransferase gene family: Messages from the databanks. Glyco-
Luke MR, Kjer-Nielsen L, Brown DL, Stow SL, Gleeson PA. 2003. GRIP
domain-mediated targeting of two new coiled-coil proteins, GCC88 and
GCC185, to subcompartments of the trans-Golgi network. J Biol Chem.
Maccioni HJF. 2007. Glycosylation of glycolipids in the Golgi complex.
J Neurochem. 103(Suppl 1):81–90.
Machamer CE. 1991. Golgi retention signals: Do membranes hold the key?
Trends Cell Biol. 1(6):141–144.
Masibay A, Balaji P, Boeggeman E, Qasba P. 1993. Mutational analysis of
the Golgi retention signal of bovine beta-1,4- galactosyltransferase. J Biol
McCormick C, Duncan G, Goutsos KT, Tufaro F. 2000. The putative tumor
suppressors EXT1 and EXT2 form a stable complex that accumulates in the
Golgi apparatus and catalyzes the synthesis of heparan sulfate. Proc Natl
Acad Sci USA. 97(2):668–673.
Milland J, Russell SM, Dodson HC, McKenzie IFC, Sandrin MS. 2002. The cy-
toplasmic tail of alpha 1,3-galactosyltransferase inhibits Golgi localization
of the full-length enzyme. J Biol Chem. 277(12):10374–10378.
Milland J, Taylor SG, Dodson HC, McKenzie IF, Sandrin MS. 2001. The
cytoplasmic tail of alpha 1,2-fucosyltransferase contains a sequence for
golgi localization. J Biol Chem. 276(15):12012–12018.
Mitra K, Ubarretxena-Belandia I, Taguchi T, Warren G, Engelman DM. 2004.
Modulation of the bilayer thickness of exocytic pathway membranes by
membrane proteins rather than cholesterol. Proc Natl Acad Sci USA.
Moloney D, Haltiwanger R. 1999. The O-linked fucose glycosylation pathway:
Identification and characterization of a uridine diphosphoglucose: Fucose-
beta1,3- glucosyltransferase activity from Chinese hamster ovary cells. Gly-
Moremen KW. 2002. Golgi alpha-mannosidase II deficiency in vertebrate
systems: Implications for asparagine-linked oligosaccharide processing in
mammals. Biochim Biophys Acta. 1573(3):225–235.
protein retention. EMBO J. 14(19):4695–4704.
Munro S. 1998. Localization of proteins to the Golgi apparatus. Trends Cell
Nakamura N, Rabouille C, Watson R, Nilsson T, Hui N, Slusarewicz P, Kreis
J Cell Biol. 131(6):1715–1726.
Cloning and sequencing of cDNA of bovine N-acetylglucosamine (beta
1–4)galactosyltransferase. Proc Natl Acad Sci USA. 83(13):4720–4724.
Nilsson T, Hoe MH, Slusarewicz P, Rabouille C, Watson R, Hunte F, Watzele
G, Berger EG, Warren G. 1994. Kin recognition between medial Golgi
enzymes in HeLa cells. EMBO J. 13(3):562–574.
Nilsson T, Lucocq JM, Mackay D, Warren G. 1991. The membrane spanning
domain of beta-1,4-galactosyltransferase specifies trans Golgi localization.
EMBO J. 10(12):3567–3575.
Nilsson T, Slusarewicz P, Hoe MH, Warren G. 1993. Kin recognition. A model
for the retention of Golgi enzymes. FEBS Lett. 330(1):1–4.
Okajima T, Xu A, Lei L, Irvine KD. 2005. Chaperone activity of pro-
tein O-fucosyltransferase 1 promotes notch receptor folding. Science.
Opat AS, Houghton F, Gleeson PA. 2000. Medial Golgi but not late Golgi
glycosyltransferases exist as high molecular weight complexes. Role of
luminal domain in complex formation and localization. J Biol Chem.
Golgi glycosylation enzymes. Biochimie. 83(8):763–773.
Pelham HR. 1990. The retention signal for soluble proteins of the endoplasmic
reticulum. Trends Biochem Sci. 15(12):483–486.
Pinhal MA, Smith B, Olson S, Aikawa J, Kimata K, Esko JD. 2001. En-
zyme interactions in heparan sulfate biosynthesis: Uronosyl 5-epimerase
and 2-O-sulfotransferase interact in vivo. Proc Natl Acad Sci USA.
OkayamaH, QasbaPK. 1986.
P L Lee et al.
Reddy JV, Burguete AS, Sridevi K, Ganley IG, Nottingham RM, Pfeffer SR.
2006. A functional role for the GCC185 Golgin in mannose 6-phosphate
receptor trafficking. Mol Biol Cell. 17(10):4353–4363.
Roseman S. 1970. The synthesis of complex carbohydrates by multiglycosyl-
transferase systems and their potential function in intercellular adhesion.
Chem Phys Lipids. 5(1):270–297.
Roth J, Berger EG. 1982. Immunocytochemical localization of galactosyltrans-
Golgi cisternae. J Cell Biol. 93(1):223–229.
Russo R, Shaper N, Taatjes D, Shaper J. 1992. Beta 1,4-galactosyltransferase:
A short NH2-terminal fragment that includes the cytoplasmic and trans-
membrane domain is sufficient for Golgi retention. J Biol Chem.
Sasaki K, Kurata-Miura K, Ujita M, Angata K, Nakagawa S, Sekine S,
Nishi T, Fukuda M. 1997. Expression cloning of cDNA encoding a hu-
man beta-1,3-N-acetylglucosaminyltransferase that is essential for poly-N-
Sato T, Sato M, Kiyohara K, Sogabe M, Shikanai T, Kikuchi N, Togayachi A,
Ishida H, Ito H, Kameyama A, et al. 2006. Molecular cloning and charac-
terization of a novel human beta1,3-glucosyltransferase, which is localized
at the endoplasmic reticulum and glucosylates O-linked fucosylglycan on
thrombospondin type 1 repeat domain. Glycobiology. 16(12):1194–1206.
Schmitz KR, Liu J, Li S, Setty TG, Wood CS, Burd CG, Ferguson KM. 2008.
Golgi localization of glycosyltransferases requires a Vps74p oligomer. Dev
Seko A, YamashitaK. 2008.
beta3Gn-T8. by Possible
involvement of beta3Gn-T8 in increasing poly-N-acetyllactosamine
chains in differentiated HL-60 cells. J Biol Chem. 283(48):33094–
Shaper NL, Shaper JH, Meuth JL, Fox JL, Chang H, Kirsch IR, Hollis GF.
1986. Bovine galactosyltransferase: Identification of a clone by direct im-
munological screening of a cDNA expression library. Proc Natl Acad Sci
Teasdale R, D’Agostaro G, Gleeson P. 1992. The signal for Golgi retention
of bovine beta 1,4-galactosyltransferase is in the transmembrane domain.
J Biol Chem. 267(6):4084–4096.
Trayer IP, Hill RL. 1971. The purification and properties of the A protein of
lactose synthetase. J Biol Chem. 246(21):6666–6675.
of glycosyltransferases in the Golgi. Science. 321(5887):404–407.
Ujita M, McAuliffe J, Hindsgaul O, Sasaki K, Fukuda MN, Fukuda M. 1999.
Poly-N-acetyllactosamine synthesis in branched N-glycans is controlled
by complemental branch specificity of I-extension enzyme and beta1,4-
galactosyltransferase I. J Biol Chem. 274(24):16717–16726.
Ujita M, Misra AK, McAuliffe J, Hindsgaul O, Fukuda M. 2000. Poly-N-
acetyllactosamine extension in N-glycans and core 2- and core 4-branched
O-glycans is differentially controlled by i-extension enzyme and different
members of the beta 1,4-galactosyltransferase gene family. J Biol Chem.
Yamaguchi N, Fukuda MN. 1995. Golgi retention mechanism of beta-1,4-
galactosyltransferase. J Biol Chem. 270(20):12170–12176.
Zhou D. 2003. Why are glycoproteins modified by poly-N-acetyllactosamine
glyco-conjugates? Curr Protein Pept Sci. 4(1):1–9.