Complex N-Glycan Number and Degree

of Branching Cooperate to Regulate

Cell Proliferation and Differentiation
Ken S. Lau,1,2 Emily A. Partridge,1,3 Ani Grigorian,4 Cristina I. Silvescu,6 Vernon N. Reinhold,6
Michael Demetriou,4,5 and James W. Dennis1,3,*
1

Samuel Lunenfeld Research Institute, Mount Sinai Hospital, 600 University Avenue R988, Toronto, ON M5G 1X5, Canada
Department of Biochemistry, University of Toronto, ON M5S 1A8, Canada
3
Departments of Molecular & Medical Genetics, Laboratory Medicine and Pathology, University of Toronto, ON M5G 1L5, Canada
4
Department of Microbiology and Molecular Genetics, University of California, Irvine, CA 92697-4025, USA
5
Department of Neurology, University of California, Irvine, CA 92862-4280, USA
6
Chemistry, University of New Hampshire, Durham, NH 03824, USA
*Correspondence: dennis@mshri.on.ca
DOI 10.1016/j.cell.2007.01.049
2

SUMMARY

The number of N-glycans (n) is a distinct feature

of each glycoprotein sequence and cooperates
with the physical properties of the Golgi N-glycan-branching pathway to regulate surface
glycoprotein levels. The Golgi pathway is ultrasensitive to hexosamine flux for the production
of tri- and tetra-antennary N-glycans, which
bind to galectins and form a molecular lattice
that opposes glycoprotein endocytosis. Glycoproteins with few N-glycans (e.g., TbR, CTLA-4,
and GLUT4) exhibit enhanced cell-surface expression with switch-like responses to increasing hexosamine concentration, whereas glycoproteins with high numbers of N-glycans (e.g.,
EGFR, IGFR, FGFR, and PDGFR) exhibit hyperbolic responses. Computational and experimental data reveal that these features allow
nutrient flux stimulated by growth-promoting
high-n receptors to drive arrest/differentiation
programs by increasing surface levels of low-n
glycoproteins. We have identified a mechanism
for metabolic regulation of cellular transition
between growth and arrest in mammals arising
from apparent coevolution of N-glycan number
and branching.
INTRODUCTION
Developmental sequences in metazoan embryogenesis,
tissue repair, and adaptive immunity frequently involve
cell proliferation, followed by differentiation and cell-cycle
arrest. Although intracellular nutrient-sensing pathways,
such as SCH9 and RAS-cAMP, are largely conserved from

S. cerevisiae to mammals (Sobko, 2006), additional pathways that are sensitive to extracellular factors have
evolved in metazoans to establish more complex body
plans. In metazoans, extracellular growth factors bind receptor tyrosine kinases (RTKs) and activate PI3K/ERK signaling, stimulating glucose metabolism and proliferation
as well as membrane remodeling (Di Paolo and De Camilli,
2006). Equally important in metazoans are the opposing
pathways that antagonize growth and are sensitive to
extracellular morphogens, cell adhesion, and, indirectly,
nutrient levels. In this regard, TGF-b receptor II (TbRII), cytotoxic T-lymphocyte-associated antigen-4 (CTLA-4), and
glucose transporter-4 (GLUT4) are examples of glycoproteins with tyrosine and dileucine adaptor-binding motifs
that promote rapid constitutive endocytosis (Al-Hasani
et al., 2002; Bradshaw et al., 1997; Ehrlich et al., 2001). Remarkably, surface levels of these glycoproteins increase
to restrict proliferation despite increased endocytosis
downstream of growth signaling. Here, we explore the
possibility that cell-surface levels of glycoproteins are
regulated in a conditional manner by metabolic flux to
N-glycan biosynthesis.
For most cell-surface receptors and transporters in
eukaryotes, 70% of the Asn-X-Ser/Thr (X s Pro) motifs
exposed to the lumen of the endoplasmic reticulum are
cotranslationally N-glycosylated (Glc3Man9GlcNAc2 /
Asn; Apweiler et al., 1999). The N-glycans are trimmed
and remodeled upon transit through the Golgi, but, due
to inherent inefficiencies of the pathway enzymes, the
resulting structures on mature glycoproteins are heterogeneous. The activities of medial Golgi-branching Nacetylglucosaminyltransferases I, II, IV, and V (encoded
by Mgat1, Mgat2, Mgat4a/b, and Mgat5 in mammals)
are dependent on the availability of the common donor
substrate UDP-N-acetylglucosamine (UDP-GlcNAc; Figure S1; Schachter, 1986). UDP-GlcNAc biosynthesis
through the hexosamine pathway utilizes substrates (fructose-6P, glutamine, acetyl-CoA, and UTP) that are key

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 123

metabolites in carbon, nitrogen, and energy homeostasis

(Broschat et al., 2002). Hence, the extent of GlcNAc
branching is dependent on branching enzyme kinetics,
and metabolic flux through the hexosamine pathway to
generate UDP-GlcNAc (A.G., unpublished data; Sasai
et al., 2002). Branched N-glycans are further modified in
the trans Golgi by b1,4galactosyltransferases, variably
elongated with poly N-acetyllactosamine (linear repeats
of Galb1,4GlcNAcb1,3) by b1,3N-acetylglucosaminyltransferases (iGNTs), and capped with sialic acid and
fucose. The Mgat5 intermediate (tetra-antennary b1,
6GlcNAc-branched) is preferentially elongated with poly
N-acetyllactosamine, producing N-glycans with higher affinities for galectins than less-branched structures (Hirabayashi et al., 2002). Thus, Mgat5/ cells display reduced
galectin-3 binding to complex N-glycans on the T cell receptor (TCR; Demetriou et al., 2001) as well as EGF receptor (EGFR) and TbRII in epithelial cells (Partridge et al.,
2004). Disruption of galectin-3 binding in Mgat5/ cells
or by lactose competition in Mgat5+/+ cells increases
TCR and EGFR mobility in the plane of the membrane
and, thus, movement into both caveoli and coated-pits
(Demetriou et al., 2001; Partridge et al., 2004; unpublished
data). As such, the lattice of interactions resulting from
galectin/glycoprotein crosslinking limits receptor internalization and maintains downstream signaling sensitivity.
Although mapping binary interactions between galectins and N-glycans is one approach to study the structural
complexity of the lattice, the dependency of multiple
glycoprotein receptors and transporters on the lattice
suggests that a systems approach may lead to a more
thorough understanding of lattice-dependent regulation.
Moreover, the number of N-glycans (n, also defined as
multiplicity in this report) is highly variable for different glycoproteins and may interact with N-glycan branching to
regulate surface residency in a glycoprotein-specific manner. To explore this possibility and its impact on responsiveness to extracellular cues, we considered a relatively
basic set of N-glycan-mediated interactions between
galectins and cell-surface glycoproteins downstream of
metabolic flux to the Golgi. We show that the N-glycanbranching pathway cooperates with the number of N-glycans on glycoproteins to differentially regulate their surface levels in response to hexosamine flux. Our results
suggest that coevolution of N-glycan multiplicity of glycoproteins and hexosamine/N-glycan-branching kinetics
integrates nutrient metabolism with cellular transition between cell growth and arrest/differentiation in mammals.
RESULTS
N-Glycan Branching Is Ultrasensitive to Hexosamine
Galectins bind to complex N-glycans with affinities proportional to N-acetyllactosamine content (Hirabayashi
et al., 2002). We reasoned that increasing the molar fractions of triantennary N-glycans on glycoproteins produced
in Mgat5/ tumor cells might be sufficient to rescue galectin binding and surface levels of EGFR and TbR. There-

124 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

fore, cells were cultured with GlcNAc, which is taken up by

bulk endocytosis, phosphorylated at the six position, and
converted to UDP-GlcNAc. UDP-GlcNAc (i.e., UDP-HexNAc to UDP-Hex ratio) increased in a Michaelis-Menten
manner with GlcNAc supplementation to both Mgat5/
and Mgat5+/+ tumor cells (Figure 1A). The increase was
greater in Mgat5/ cells, possibly due to reduced feedback inhibition by downstream high-affinity products.
GlcNAc enhanced surface TbR complexes in Mgat5/
and Mgat5+/+ cells and restored surface levels of EGFR
bound to galectin-3 in Mgat5/ cells (Figure 1B). Cellsurface galectin-3 was reduced in Mgat5/ cells but
rescued by GlcNAc (Figure 1C).
To characterize the role of hexosamine and Golgidependent regulation in growth and arrest signaling, we
considered the intermediary steps downstream of UDPGlcNAc, notably N-glycan branching, receptor retention
at the cell surface, and intracellular signaling (Figure 1D).
The hexosamine-dependent changes in branching were
determined by mass spectrometry in Mgat5/ and
Mgat5+/+ cells (Figures 2A, 2B, and S2). Triantennary Nglycan levels doubled in Mgat5/ cells when supplemented
with GlcNAc, suggesting that less-branched N-glycans in
larger quantities are sufficient to restore galectin binding
and surface receptor levels (Figure 2A). In contrast,
GlcNAc supplementation of Mgat5+/+ tumor cells only produced small increases in tri- plus tetra-antennary N-glycans (Figure 2B). Oncogenic transformation increases
gene expression of Mgat4, Mgat5, and iGNT by 5- to 10fold (Buckhaults et al., 1997; Ishida et al., 2005; Takamatsu
et al., 1999), which appears to reduce the requirement for
hexosamine supplementation by supporting higher levels
of tri- and tetra-antennary N-glycan biosynthesis in
Mgat5+/+ tumor cells. Interestingly, we observed that the
triantennary N-glycan fraction, the Golgi-pathway end
product in Mgat5/ cells, increased with an ultrasensitive
response to GlcNAc (Hill coefficient nH 5.5; Figure 2A).
Ultrasensitivity describes a stimulus/response relationship
where the response rises sharply over a narrow range
of stimulus, producing a sigmoid-shaped curve with a
delay preceding the rise in response and a Hill coefficient
(nH) [ 1 (Koshland et al., 1982). The more common
Michaelis-Menten relationship describes a response curve
resembling a hyperbolic function (nH 1), which is typical
of a simple one-step substrate-product relationship.
Ultrasensitive responses are switch-like (all-or-none) responses that have been observed in many biological
systems (Ferrell and Machleder, 1998). For example, nonordered phosphorylation of yeast SIC1 increases progressively with advancing cell cycle, and, at more than five of
nine sites occupied, SIC1 affinity for CDC4 increases in
a switch-like manner (Klein et al., 2003).
Computational Model of Golgi Ultrasensitivity
To better understand branching ultrasensitivity to UDPGlcNAc, we constructed a computational model of elementary processes using ordinary differential equations
(ODEs) with enzyme concentrations and kinetic

Figure 1. Hexosamine Rescues Surface

Galectin-3 and EGF/TGF-b SurfaceReceptor Levels in Mgat5/ Cells
(A) shows UDP-HexNAc to UDP-Hex ratio
[UDP-GlcNAc + UDP-GalNAc] / [UDP-Gal +
UDP-Glc] in Mgat5+/+ and Mgat5/ tumor cells
after 48 hr of culture with GlcNAc supplementation. Normalized mean SE (standard error of
the mean) of two independent experiments is
shown.
(B) Galectin-3 association is shown with EGFR
determined by 125I-EGF crosslinking to cell surface EGFR and immunoprecipitated with antibodies to galectin-3 with and without 50 mM
GlcNAc. 125I-TGF-b1 bound to surface TbR in
cells.
(C) FACS analysis of surface galectin-3 in
Mgat5+/+ and Mgat5/ cells supplemented
with indicated amounts of GlcNAc is shown.
Mean SE.
(D) is a schematic of subsystems considered
in the rest of this report.

parameters from literature values (Table S1; Krambeck and

Betenbaugh, 2005; Umana and Bailey, 1997). N-glycan
structures are generated up to a maximum size of tetra-antennary with N-acetyllactosamine and (Galb1-4GlcNAcb13)1-2 extensions for a total of 14 structures (Table S2; structures). Substitutions of the antennae with sialic acid,
fucose, and sulfate can alter the affinity for galectins, but
we did not include these modifications in order to emphasize the effects of N-glycan branching. The GlcNAcbranching enzymes have well-characterized specificities
for acceptors (Schachter, 1986), and, therefore, the medial
Golgi pathway is modeled as a distributive series of reactions and not physically ordered in the Golgi compartments
(Figure S3A). In Mgat5/ cells, simulations of N-glycan
profiles as functions of increasing UDP-GlcNAc concentrations revealed that triantennary N-glycans indeed
increased in a switch-like fashion (nH 5.6; Figure 2C).
We next investigated the source of branching ultrasensitivity by computationally perturbing the Mgat5/ model.
Ultrasensitivity appears to be distributed over multiple reactions and therefore is a robust property of the pathway
(Table S3). However, in simulations where GlcNAcbranching enzymes are removed sequentially, ultrasensitivity is stepwise diminished, demonstrating the necessity
of multistep usage of UDP-GlcNAc (Table S2). Multistep
ultrasensitivity in a biochemical pathway, unlike allosteric
sensitivity, occurs with repetitive use of either enzymes or
substrates in sequential reactions that decline in efficiency
(Ferrell and Machleder, 1998). In this regard, affinities for
UDP-GlcNAc and enzyme concentrations decrease step-

wise from Mgat1 to Mgat5 (Table S1). Furthermore, simulation of branching as a linear pathway with no removal
of intermediate products (i.e., transit out of the Golgi) results in a hyperbolic increase of end product in response
to UDP-GlcNAc (Figure S3B). These results identified the
two conditions required for branching ultrasensitivity: (1)
sequential decreasing efficiency of Golgi enzyme reactions and (2) removal of intermediate products. In keeping
with (1), simulation of a 15-fold increase of late Golgi enzymes Mgat4, Mgat5, and iGNT) characteristic of transformed cells (Buckhaults et al., 1997; Ishida et al., 2005;
Takamatsu et al., 1999), displayed response profiles with
unchanging tri- and tetra-antennary N-glycans due to
a pathway D50 below the physiological concentration of
1.5 mM UDP-GlcNAc, similar to that observed experimentally for Mgat5+/+ tumor cells (Figure 2D).
N-Glycan Multiplicity Cooperates with Branching
to Regulate Glycoproteins
Next, we confirmed that hexosamine enhancement of
surface receptors and signaling is dependent on complex
N-glycans. Wild-type CHO cells and the Mgat1 mutant
Lec1 cells were compared for changes in b1,6GlcNAcbranched N-glycans (which bind L-PHA, P. vulgaris
leukoagglutinating lectin) and responsiveness to TGF-b
following GlcNAc supplementation. Lec1 CHO cells are
deficient in Mgat1 activity, blocking all GlcNAc-branching
reactions (Kumar and Stanley, 1989), and GlcNAc supplementation to these cells did not increase L-PHA reactivity
nor enhance responsiveness to TGF-b (Figures 3A and

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 125

Figure 2. Hexosamine Flux Stimulates

N-Glycan Branching
Distribution of branched N-glycans in (A)
Mgat5/ and (B) Mgat5+/+ cells is shown.
Branched N-glycan structures from mass
spectrometry analysis are pooled as in
Figure S2, and branching is indicated as 0 to
4+ with variable extensions as in the cartoon.
(C) Simulation of steady-state N-glycans in
Mgat5/ and (D) Mgat5+/+ cells is shown as
a function of Golgi UDP-GlcNAc concentration.
The gray box represents UDP-GlcNAc below
the estimated physiological D0 of 1.5 mM,
from where the Hill coefficient is calculated.
The larger 4+ structures are underrepresented
relative to smaller structures in the MALDI profiles, but the slopes of the response curves to
UDP-GlcNAc are consistent with computational simulations.

S4A). In contrast, GlcNAc enhanced both end points with

switch-like responses in wild-type CHO cells (nH = 4.6
6.0). Without supplementation, L-PHA binding was 2fold higher in Mgat5+/+ tumor cells than CHO cells, consistent with PyMT oncogene-dependent increases in the
late-branching Mgat4/5 enzymes.
With the multistep structure of the branching pathway,
a gain in Mgat1 activity is predicted to limit the availability
of UDP-GlcNAc to late enzymes, thus replacing high-affinity tri- and tetra-antennary N-glycans with low-affinity
monoantennary structures (Figure S3B; Table S2). Overexpression (253) of Mgat1 in Mgat5/ cells suppressed
GlcNAc-induced EGFR (Figure 3B). Simulations supported the above result (Figure S5A) and also confirmed
that EGFR suppression by Mgat1 overexpression is dependent on both declining efficiency of the pathway and
on intermediate product removal with Golgi trafficking;
these are the key properties required for ultrasensitivity
(Figures S5C and S3B). Indeed, overexpression of
Mgat1 in Mgat5+/+ tumor cells did not inhibit GlcNAc-dependent increases in EGFR signaling, as condition (1) for
ultrasensitivity was not met. (Figure S5D).
Next, we explored the effect of hexosamine-dependent
N-glycan changes on the surface levels of EGFR and TbR.
Surface TbR increased with the expected switch-like response (nH 4.4) to GlcNAc, reflecting the Golgi-pathway

126 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

response, but, surprisingly, EGFR increased in a linear

manner (Figures 3C, 3D, and S4B). GlcNAc also restored
responsiveness to EGF and TGF-b in signaling assays
with hyperbolic (nH 1.2) and switch-like (nH 5.4) response curves, respectively (Figures 3E, 3F, S4C, and
S4D). Downstream intracellular signaling does not appear
to be a confounding factor as both ERK and SMAD activation increase hyperbolically with increasing surface receptor when stimulated by excess ligand, based on computational models (Figures S4E and S4F; Schoeberl et al.,
2002; Vilar et al., 2006) and experimental data (Wiley
et al., 2003). Therefore, we conclude that branching ultrasensitivity to UDP-GlcNAc is a regulatory feature of the
pathway that interacts with glycoprotein-specific information to differentially regulate its surface levels. In this
regard, the number of N-glycosylation sites (n) is a distinct
feature defined by the protein sequence that determines
avidity for the galectin lattice. Murine TbRI and TbRII
have one and two N-glycans, respectively (Koli and Arteaga, 1997), while EGFR has eight occupied N-X-S/T
sites (Zhen et al., 2003).
Computational Model of Glycoprotein Regulation
by N-Glycan Multiplicity and Branching
To explore the potential for N-glycan multiplicity and hexosamine/N-glycan branching to differentially regulate

Figure 3. Downstream Signaling Reflects GlcNAc Dose Responses of Surface EGFR and TbR Receptors
(A) L-PHA staining (binds b1,6GlcNAcbranched N-glycans) of Mgat5+/+, Mgat5/,
CHO, and Lec1 (Mgat1/CHO) cells supplemented 48 hr with GlcNAc is shown. Mean
SE.
(B) shows overexpression of Mgat1 in
Mgat5/ cells supplemented with GlcNAc
and acutely stimulated with EGF (5 min). Infection with the Mgat1 vector and transient expression for 48 hr increased Mgat1 activity
by 25-fold, while pools of stably transfected
cells displayed an 2.5-fold increase in
Mgat1 activity. Normalized mean SE is of
minimum of six replicates from two to seven
independent experiments.
Surface (C) EGFR and (D) TbR in Mgat5+/+ and
Mgat5/ tumor cells quantified by FITC-EGF
binding and 125I-TGF-b1 binding, respectively.
(E) ERK-p nuclear translocation measured 5
min after stimulation with EGF (acute) or vehicle
(autocrine).
(F) SMAD2/3 nuclear translocation measured
45 min after stimulation with TGF-b1 or vehicle.
Normalized mean SE of three independent
experiments with at least 200 cells each is
shown.

surface glycoprotein levels, we extended our mathematical model to calculate the fractional changes in surface
glycoproteins under varying degrees of N-glycan multiplicity and Golgi enzyme activities. A glycoform is defined
here as a protein isoform with a distinct combination of Nglycan structures distributed randomly over the possible
N-X-S/T sites of the glycoproteins extracellular domain.
The protein environment of N-X-S/T sites can limit the
accessibility to Golgi enzymes (Do et al., 1994), but we
assume that site-specific bias is not a significant determinant of glycoform generation. The number of glycoforms
follows a combinatorial relationship with respect to the
number of N-glycans on a glycoprotein and the number of
N-glycan types. Although this number is actually much
larger, the 14 N-glycan types used in the model are sufficient to illustrate the concepts.
The steady-state distribution of glycoforms at the cell
surface is calculated based on the binding avidities of glycoforms for galectin-3 (Ahmad et al., 2004), a prototypic
galectin that is not limiting in this model. Avidity for the lattice is modeled to be proportional to the sum of binding affinities of the N-glycans on each glycoform (SKa). The af-

finity threshold for stable association of a glycoform with

the galectin lattice corresponds to a single tri-, two bi- or
6 monoantennary N-glycans (SKa > 1.28 mM1) and corresponds to in vitro galectin-binding data (Hirabayashi
et al., 2002). Glycoforms are retained on the cell surface
with half-lives proportional to their avidities for the lattice
and are countered by constitutive endocytosis, which is
generally held constant in our simulations to illustrate the
effects of N-glycan-dependent surface interactions.
Surface retention of glycoproteins by the lattice with a
single N-glycan follows the switch-like response of triand tetra-antennary N-glycan production. However, with
larger values of n, the number of glycoforms and their
mean avidity for the lattice increase, which essentially
suppresses the lag phase of the sigmoidal response
(Figure 4A). Moreover, the larger glycoform distributions
traverse the affinity threshold in response to increasing
UDP-GlcNAc in a graded manner, and, importantly, at
lower UDP-GlcNAc concentrations (Figures S6A and S6B).
Thus, the slopes for high-n glycoprotein responses are
steeper at low hexosamine flux (1.5 mM UDP-GlcNAc)
than at high flux (>3 mM), while the trend is reversed for

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 127

Figure 4. N-Glycan Branching Cooperates with N-Glycan Multiplicity

(A) Simulation of surface glycoprotein fractions
in Mgat5/ cells as a function of Golgi UDPGlcNAc concentration and N-glycan multiplicity (n) is shown.
(B) Change in surface expression of mycGLUT4-GFP (wild-type, n = 1; mutant, n = 0)
in HEK293T cells supplemented with GlcNAc
for 48 hr and imaged with confocal microscopy
is shown. Normalized mean SE surface/interior GLUT4 from three independent experiments of transfected cells from 10 random
fields is shown for each GlcNAc concentration.
(C) shows dose responses of ERK activation in
Mgat5/ tumor cells to IGF, PDGF, FGF, or
EGF as a function of GlcNAc supplementation.
Mean SE.
(D) Scatter plot of N-X-S/T multiplicity as a function of extracellular domain length for mammalian receptor kinases is shown. RK1 are receptor kinases with growth promoting activities
(closed, 11.3 5.1 sites/receptor), and RK2
are those with arrest/morphogenic activities
(open, 2.5 1.3 sites/receptor).

low-n glycoproteins (Figure 4A). Mgat1 overexpression at

high levels (253) suppresses hexosamine-dependent upregulation of EGFR signaling, suggesting that replacement of all high-affinity N-glycans with monoantennary
N-glycans puts glycoforms near the threshold for lattice
association, which is functionally similar to removing Nglycans. Thus, moderate Mgat1 overexpression (2.53)
changes the GlcNAc response curve from hyperbolic to
sigmoidal, characteristic of a low-n glycoprotein (Figure 3B).
Examples of Other Glycoproteins with Low and High
N-Glycan Multiplicities
The glucose transporter GLUT4 has one N-glycan, and
a dileucine motif mediates its constitutive endocytosis
(Al-Hasani et al., 2002). Intracellular stores of GLUT4 in
muscle and adipocytes are recruited into the recycling endosomes by insulin receptor activation and to the cell surface in a poorly defined manner. If GLUT4 retention at the
surface is dependent on hexosamine/N-glycan branching
and galectin binding, its upregulation by GlcNAc is expected to display a high degree of ultrasensitivity. Accordingly, GlcNAc supplementation to HEK293T cells expressing myc-GLUT4-GFP enhanced cell-surface GLUT4 levels
with an expected switch-like response (nH 10.0; Figures
4B and S7). Moreover, mutation of the single N-X-S/T site
(N 57 Q) blocked GlcNAc-dependent enhancement of surface GLUT4 (Figure 4B). In contrast, receptors of FGF,
IGF, and PDGF possess high numbers of N-glycans (8
16), and GlcNAc increased responsiveness to their cognate ligands hyperbolically (Figure 4C).

128 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

N-Glycan Multiplicity and Receptor Kinase Function

Our results suggest that hexosamine/N-glycan branching
may differentially regulate EGFR/PI3K/ERK and TGF-b/
SMAD signaling, which are antagonistic for G1/S transition (Matsuura et al., 2004; Siegel and Massague, 2003).
If more broadly applicable, this observation might be reflected in the N-glycan multiplicity of different functional
classes of receptors. To explore this possibility in a relatively well-studied set of human receptors, we ranked
the receptor kinases by the incidence of canonical N-glycosylation motifs (N-X-S/T not followed by P and X not P)
in their extracellular domains. The number of N-X-S/T in
mammalian receptor kinases varies considerably, but receptors that stimulate cell proliferation, growth, and oncogenesis have more N-X-S/T sites, longer extracellular domains, and an increased number of sites per 100 amino
acids (30 receptors designated RK1; 11.33 5.05 sites/receptor; 1.89 0.58 sites /100 amino acids; Figures 4D and
S10A; Table S4). In contrast, receptors known for their
roles in vasculature formation and organogenesis, such
as TIE1, MUSK, LTK, ROR1/2, DDR1, the EPH receptors,
and TGF-b/BMP receptors, have lower N-glycan multiplicities (27 receptors designated RK2; 2.48 1.28 sites/
receptor; 0.86 0.54 sites/100 amino acids; Figure 4D).
The occurrence of N-X-S/T sites per 100 amino acids in
a random selection of human cell-surface glycoproteins
is 1.21 0.89, which is greater than 0.60, a value calculated based solely on the amino acid composition of the
extracellular domains in this set (Figure S10A). The actual
frequency may in part reflect the conservation of some
N-X-S/T sites, where their substitution in the ER is critical

Figure 5. N-Glycan
Multiplicity Impose
Sequence

Branching and
Growth-to-Arrest

(A) L-PHA staining, UDP-GlcNAc, and UDP-Gal

in NMuMG cell lysates following GlcNAc supplementation are shown.
(B) SMAD2/3 and ERK-p nuclear translocation
in NMuMG cells preconditioned in GlcNAc for
120 hr and stimulated with either vehicle (autocrine) or 3 ng/ml of TGF-b or 5% FBS, respectively, are shown. Mean SE.
(C) ERK-p-to-SMAD2/3 activation and cell
density of NMuMG cells in GlcNAc-supplemented medium (cell viability >95% throughout). Normalized mean SE of three independent experiments.
(D) E-cadherin and F-actin staining of NMuMG
cells with or without GlcNAc supplementation
is shown.
(E) shows simulation of the relative surface
levels of glycoprotein receptor pairs (X and Y)
as a function of N-glycan multiplicity and Golgi
UDP-GlcNAc concentration.
(F) Simulation of the dynamic range of change
in surface glycoproteins as a function of endocytosis rate with an increase of Golgi UDPGlcNAc from 1.5 to 6 mM is shown.

for protein folding. However, where it has been examined

(such as the insulin receptor), most sites are not required
for protein folding or activity (Elleman et al., 2000). These
observations suggest that differences in N-glycan multiplicity of receptor kinases are conserved along with growth
and differentiation functions in mammals and are determinants of regulation via hexosamine/N-glycan branching.
Hexosamine/N-Glycan Branching Balances
Responsiveness to Growth and Arrest Cues
Epithelial-to-mesenchymal transition (EMT) requires ERK
and PI3K activation downstream of RTKs as well as autocrine TGF-b signaling at intermediate levels that are conducive to the induction of a mesenchymal program of gene
expression but below the threshold for SMAD2/3-dependent cell-cycle arrest (Oft et al., 2002). Nontransformed
NMuMG cells respond to TGF-b and undergo EMT, which
is also observed during embryogenesis and carcinoma
progression. Therefore, provided that the N-glycanbranching pathway is ultrasensitive to UDP-GlcNAc in
NMuMG cells, we anticipated that hexosamine flux

through the Golgi might enhance responsiveness to growth

cues first, then TGF-b leading to EMT and growth arrest.
GlcNAc supplementation to NMuMG cells increased UDPGlcNAc concentration (nH 1.3), b1,6GlcNAc-branched
N-glycans (nH 6.0), and responsiveness to acute TGFb/SMAD (nH 6.0) and to FCS/ERK (nH 1.3), all with
the expected response behaviors (Figures 5A and 5B).
Mgat1 overexpression in NMuMG cells also suppressed
GlcNAc-dependent signaling, confirming the branching
pathway to be multistep ultrasensitive (Figures S8AS8D).
In addition to acute signaling, hexosamine also
alters cellular responsiveness to autocrine stimulation in
NMuMG cells. In normal growth conditions, the ratio of
activated ERK/SMAD increased and then decreased in
response to hexosamine flux, mirroring changes in cellproliferation rates (Figure 5C). Accompanying growth
arrest, GlcNAc induced loss of E-cadherin and the mesenchymal morphology (Figure 5D). TGF-b/SMAD signaling in
Mgat5/ tumor cells was ultrasensitive to GlcNAc, which
also induced EMT with the expected response as autocrine TGF-b stimulation (Figures S8E and S8F). EMT was

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 129

reversible upon GlcNAc withdrawal in NMuMG cells and

Mgat5/ tumor cells. In contrast, pathologic overexpression of late-branching enzymes in PyMT Mgat5+/+ tumor
cells supplies high-affinity galectin ligands at lower UDPGlcNAc concentrations, rendering proliferation and EMT
independent of hexosamine (Figure S8E).
Computational simulations show that ratios of surface
receptors taken in pairs change in proportion to the
difference in n. Glycoproteins with high multiplicities (numerator) increase at the surface early in the UDP-GlcNAc
titration, while branching ultrasensitivity creates a delay
before low-n glycoproteins (denominator) accumulate
(Figure 5E). Simulations where endocytosis rates are varied indicate that low-n receptors are markedly more sensitive to removal from the surface due to inherently lower
avidities for the lattice (Figure 5F). Therefore, conditional
regulation of branching by nutrient flux through the hexosamine pathway stimulates growth first and at higher flux
arrest and differentiation.
Regulation of T Cell Growth Arrest
by Hexosamine/N-Glycan Branching and CTLA-4
Next, we investigated whether hexosamine/Golgi ultrasensitivity regulates growth arrest mechanisms in other
cellular systems. Naive T cells from Mgat5/ mice have
reduced thresholds for activation due to the lower affinity
of TCR for the galectin lattice, which enhances antigendriven TCR clustering and signaling at the immune synapse (Demetriou et al., 2001). Once activated, naive T cells
undergo multiple rounds of division followed by growth arrest and differentiation into effector T cells. Growth arrest
of blasting T cells is dependent on induction of cell-surface CTLA-4, a glycoprotein with low N-glycan multiplicity
and high constitutive endocytosis (Alegre et al., 2001).
CD28 and CTLA-4 are sequence-related receptors that
compete for CD80/86 ligand on antigen-presenting cells
to positively and negatively regulate T cell proliferation, respectively (Alegre et al., 2001). CD28 is constitutively expressed at the cell surface and has five N-glycan sites
(3.8/100 amino acids), while surface CTLA-4 is induced
35 days after TCR signaling and has only two sites (1.6/
100 amino acids).
Surface CTLA-4 was lower on Mgat5/ and Mgat5+/
than Mgat5+/+ T cells following anti-CD3 stimulation but
reached comparable levels with strong costimulation by
anti-CD3 and CD28 antibodies (Figure S9A). Tracking T
cells by the number of divisions revealed that the proportion of cells expressing surface CTLA-4, as well as their
associated surface levels (i.e., MFI), is always lower in
Mgat5/ T cells following low, but not high, levels of TCR
stimulation (Figure 6A). In contrast, surface CD28 levels
are similar in the two genotypes (Figure S9B). CTLA-4
gene expression increases in proportion to TCR signaling
strength (Chikuma and Bluestone, 2002), and as expected, TCR hypersensitivity of Mgat5/ T cells is
associated with increased intracellular CTLA-4 following
activation (Figure 6B), consistent with a specific defect
in surface retention. CTLA-4 surface expression was

130 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

reduced by lactose treatment, a competitive inhibitor of

galectin binding, but not by control disaccharide sucrose
(Figure S9C). Increased membrane remodeling with T
cell activation presumably increases the requirement for
lattice-dependent retention of CTLA-4. Consistent with
this notion, constitutive CTLA-4 expression in resting
CD4+FOXP3+ regulatory T cells is not different between
Mgat5+/+ and Mgat5/ cells (Figure S9E).
GlcNAc enhanced surface CTLA-4 in Mgat5+/+ more
than in Mgat5/ T cells (Figure 6C), and, as predicted
by our model, induced CTLA-4 surface expression with a
switch-like response (nH 6.4), while the UDP-HexNAc/
UDP-Hex ratio increased in a linear manner to 3-fold
(Figure 6D). Swainsonine, an inhibitor of Golgi a-mannosidase II, suppressed high-affinity galectin ligands and
blocked GlcNAc-induced surface CTLA-4 (Figure S9D).
TCR signaling induces large increases in glucose uptake
(Frauwirth et al., 2002) and Mgat5 gene expression
(Morgan et al., 2004) as well as UDP-GlcNAc, surface
b1,6GlcNAc-branched N-glycans, and, finally, CTLA-4
(Figure 6E). We conclude that TCR signaling increases
metabolite flux through the hexosamine and N-glycanbranching pathways, upregulating surface CTLA-4 with a
delay that is dependent on Golgi multistep ultrasensitivity.
DISCUSSION
Branching-Pathway Ultrasensitivity and N-Glycan
Multiplicity
Our results demonstrate multistep ultrasensitivity of
N-glycan branching downstream of hexosamine flux to
be a structural and regulatory feature of the pathway. Remarkably, ultrasensitivity of N-glycan branching appears
to be transformed inversely by N-glycan multiplicity (n),
an encoded feature of glycoprotein sequences. Receptors
engineered to have different numbers of N-glycan sites
should conform to this model provided that the alterations
do not disrupt other aspects of receptor function. Without
exception, the hexosamine response profiles for the
glycoproteins tested herein with high (EGFR, PDGF,
IGFR, and FGF) and low (TbR, CTLA-4, and GLUT4) multiplicities are consistent with a model where high-n glycoproteins are associated with growth and low-n with arrest/
differentiation. Although not a signaling receptor, GLUT4
accelerates the systemic disposal of glucose (Tsao
et al., 1996), thus reducing hexosamine flux and maintaining signaling homeostasis. Indeed, failure to reduce
glucose associated with type II diabetes induces matrix
deposition in glomerular mesangial cells in a TGF-bdependent manner (Kolm-Litty et al., 1998). Taken together,
our results indicate that differences in N-glycan multiplicities cooperate with Golgi multistep ultrasensitivity
to enhance differential regulation of surface glycoproteins downstream of metabolic flux into the hexosamine
pathway.
In addition to N-glycans, UDP-GlcNAc is utilized in
mucin, glycolipid, and hyaluronate biosynthesis and in
O-GlcNAc modifications to cytosolic proteins on Ser/Thr

Figure 6. Hexosamine and N-Glycan

Branching Promote CTLA-4 Cell-Surface
Expression
(A) Surface CTLA-4+ (%) and MFI in Mgat5+/+
and Mgat5/ T cells stimulated for 5 days
are shown, with cell division measured by
CFSE staining.
(B) Total CTLA-4 in Mgat5+/+ and Mgat5/ T
cells is shown after 4 days of stimulation with
250 ng/ml anti-CD3 antibody.
(C) Surface CTLA-4 MFI in T cells from Mgat5+/+
and Mgat5/ mice stimulated for 5 days in
the presence or absence of GlcNAc (40 mM).
Mean SE.
(D) shows UDP-HexNAc/UDP-Hex ratio at 24
hr and CTLA-4+ (%) at 4 days following GlcNAc
supplementation and stimulation of T cells with
250 ng/ml anti-CD3 antibody. Mean SE.
(E) Intracellular UDP-HexNAc/UDP-Hex ratio,
L-PHA staining, and surface CTLA-4 are shown
following stimulation with 500 ng/ml anti-CD3
antibody. Mean SE.

(Slawson et al., 2006). However, suppression of GlcNAcdependent surface-glycoprotein regulation by mutation

of the N-X-S/T site in GLUT4 and by knockout (Lec1) or
overexpression of Mgat1 confirms the involvement of Nglycan branching. Many details of N-glycan-lectin interactions remain to be defined, such as metabolic and genetic
factors regulating Golgi sugar-nucleotide pools, further
substitutions to the antennae of N-glycans, and binding
at the cell surface by other lectins. For example, cell-surface retention of B cell receptors is positively regulated by
sialoadhesin CD22, which binds a2,6sialic acid found on
the termini of lactosamine antennae (Collins et al., 2006).
Regulation of Cell Growth and Differentiation
by Fine-Tuning Ultrasensitivity
Signaling downstream of RTKs has outputs that differ in ultrasensitivity as required to control cell polarity versus the
more switch-like cell-cycle transitions. The RAF-MEKERK cascade is ultrasensitive as previously demonstrated
by computational modeling and in experiments using

Xenopus oocyte, where direct activation with progesterone triggers maturation (Ferrell and Machleder, 1998).
EGF-EGFR binding stimulates tyrosine-phosphorylation
of AP-2, EPS15, CBL-2, clathrin, and other components
of coated-pit endocytosis and promotes rapid receptor inactivation upstream of the RAF-MEK-ERK cascade (van
Delft et al., 1997). Consequently, the membrane-proximal
portion of the EGF-signaling pathway from EGFR to RAS/
PI3K is subsensitive (slope 0.1; see Supplemental Data),
presumably meeting temporal-spatial requirements for the
exploratory nature of membrane remodeling (Wang et al.,
2002). This effectively suppresses RAF-MEK-ERK ultrasensitivity (nH 10), resulting in a near hyperbolic response for EGFR to ERK. With similar considerations, high
N-glycan multiplicities of receptors produce much larger
glycoform distributions and, thus, a gradation of avidities for
the lattice that effectively suppresses Golgi ultrasensitivity
at the level of glycoprotein recruitment to the cell surface.
Importantly, higher N-glycan multiplicities increase glycoprotein avidities for the lattice at lower hexosamine levels,

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 131

which eliminates the delay in the hexosamine-dependent

titration observed for low-n receptors. This delay, as well
as the switch-like response, is a defining feature of
ultrasensitivity. Therefore, N-glycan multiplicity and the
Golgi-branching pathway impose an order of glycoprotein
titration to the cell surface with respect to increasing hexosamine flux: first growth factor receptors, then TbR and
other low-n glycoproteins. Simulations suggest that activation of cells from a quiescent state may promote endocytosis of low-n receptors at an early stage before lattice
avidity is strengthened by positive feedback to the hexosamine pathway (Figures 5F and 7).
Regulation of CTLA-4 and Growth Arrest in T Cells
After undergoing serial rounds of cell division, activated T
cells induce CTLA-4 to the cell surface and arrest in G1 of
the cell cycle (Alegre et al., 2001). b1,6GlcNAc-branching
deficiency in naive T cells reduces activation thresholds by
weakening the galectin lattice, enhancing TCR clustering,
and signaling at the immune synapse (Demetriou et al.,
2001). Metabolic supplementation of the hexosamine
pathway with glutamine, acetoacetate, uridine, or GlcNAc
enhances surface b1,6GlcNAc-branched N-glycans in naive T cells and reduces TCR sensitivity (A.G., unpublished
data). However, early changes in gene expression following antigen-driven immune synapse formation increase IL2 and Mgat5 activities (Morgan et al., 2004), which drives
positive feedback to glucose metabolism (Frauwirth et al.,
2002), hexosamine flux, N-glycan branching, lattice
strength, and, finally, CTLA-4 surface retention. Thus,
hexosamine/Golgi/lattice negatively regulates T cell responses early by inhibiting TCR/CD28 signaling in naive
T cells and later by promoting surface retention of CTLA4 in T cell blasts.
The elimination of pathogens requires rapid expansion
of reactive T cells followed by abrupt suppression to limit
inflammation and avoid autoimmune responses. Mgat5deficient mice, as well as inbred mouse strains with
deficiencies in GlcNAc-branching, are hypersensitive to
autoimmune disease (Demetriou et al., 2001; A.G., unpublished data). Hexosamine flux enhances N-glycan branching in T cells, and GlcNAc supplementation of cultured T
cells during restimulation suppresses Experimental Autoimmune Encephalomyelitis (EAE) upon transplant of the
cells into naive mice. Moreover, recent analysis of N-glycan branching in Multiple Sclerosis patients has revealed
novel alleles of Mgat1 that enhance activity and reduce
GlcNAc branching in a manner sensitive to hexosamine
flux, consistent with the data and model presented herein
(M.D., unpublished data).

glycosylation (Wang et al., 2001). Our model suggests

that Mgat2 expression provides minimal branching required for inclusion of most receptors in the galectin/glycoprotein lattice (Figure S6E). Mgat5/ mice develop normally but display postnatal deficiencies including reduced
numbers of muscle satellite and bone marrow osteoprogenitor cells with reduced nuclear ERK/SMAD ratios
(J.D., unpublished data). Mgat5/ mice also display reduced blood glucose and resistance to weight gain on
a high-fat diet and are similar in this regard to GLUT4-deficient mice (Katz et al., 1995). Mgat4a/ mice display
a predisposition to type II diabetes, and their pancreatic
b cells are deficient in glucose-stimulated insulin secretion
due to a reduction in galectin binding to surface GLUT2
(Ohtsubo et al., 2005). Interestingly, our results demonstrate that the hexosamine pathway enhances surface expression of GLUT4 and predicts a similar function for
GLUT2 in b cells during peak glucose flux. More generally,
fluctuations in metabolic pathways on the time scale of allosteric regulation (sec) appear to interact with N-glycan
branching to continually adapt cellular responsiveness to
growth and arrest cues and, consequently, homeostasis
in multiple systems.

Metabolic and Developmental Phenotypes

The hexosamine pathway and minimal branching by
Mgat1 and Mgat2 are required for mammalian embryogenesis (Boehmelt et al., 2000; Ioffe and Stanley, 1994).
Mgat2 deficiency is postpartum lethal with morphogenic
defects in multiple tissues, similar to the homologous
deficiency in human type II congenital disorder of

Coevolution of Branching and Multiplicity

Finally, complex N-glycans are expressed in very low
amounts in C. elegans and D. melanogaster (Cipollo et al.,
2005), and most mature N-glycans are mannose-terminated (i.e., paucimannose) rather than complex N-glycan
structures (Zhu et al., 2004). C. elegans expresses functional Mgat1, Mgat2, and Mgat5 enzymes (but not

132 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

Figure 7. Model of Hexosamine Regulation of Surface Glycoproteins and Responsiveness to Growth and Arrest Cues
Growth factor receptors have high numbers of N-glycans and, therefore, high avidities for the galectin lattice, while TGF-b receptors I
and II have low multiplicities (n = 1 and n = 2). Glycoforms generated
in the Golgi are above (R*) or below (R) the affinity threshold for stable
association with the galectin lattice (Figure S6A). In Mgat5/ cells, insufficient positive feedback to growth signaling (black) results in a predominance of arrest signaling (red). In wild-type cells, (1) stimulation of
RTKs in quiescent cells (2) increases PI3K signaling and promotes Factin remodeling and preferential internalization of low-n receptors
(e.g., TbR, Figure 5F). (3) This enhances positive feedback to hexosamine/N-glycan processing, which leads ultimately to (4) increasing
TbR association with galectins and autocrine arrest signaling.

Mgat4; Warren et al., 2002), while D. melanogaster has

Mgat1, Mgat2, and a sequence ortholog of the Mgat4 gene
(but not Mgat5; Figures S10BS10F). Our computational
model suggests that a minimum of 13 monoantennary
N-glycans is required to significantly increase cell-surface
glycoproteins downstream of hexosamine flux (Figures
S6CS6E). Although a survey of receptor kinases in mammals revealed that N-glycan sites are reduced in number
and density on receptor kinases that mediate arrest/differentiation, receptors in C. elegans and D. melanogaster have
relatively high N-X-S/T multiplicity and do not segregate in
this manner (Figure S10A). This suggests that selection
pressures driving the coevolution of N-glycan branching
and N-glycan multiplicity in a subset of glycoproteins may
be associated with greater complexity of the immune system and metabolic homeostasis in mammals.
EXPERIMENTAL PROCEDURES

ACKNOWLEDGMENTS
This research was supported by grants from CIHR and Genome Canada through the OGI to J.W.D. and funding from NIH and NMSS USA
to M.D. K.S.L. was funded by a NSERC PGSD scholarship. We thank
P. Cheung and G. Ruan for technical assistance and Drs. H. Schachter,
R. Linding, J. Parkinson, Y. Haffani, A. Datti, G. Bader, I.R. Nabi, and
M. Rozakis-Adcock for advice.
Received: September 15, 2006
Revised: November 15, 2006
Accepted: January 24, 2007
Published: April 5, 2007
REFERENCES
Ahmad, N., Gabius, H.J., Andre, S., Kaltner, H., Sabesan, S., Roy, R.,
Liu, B., Macaluso, F., and Brewer, C.F. (2004). Galectin-3 precipitates
as a pentamer with synthetic multivalent carbohydrates and forms heterogeneous cross-linked complexes. J. Biol. Chem. 279, 10841
10847.

Computational Model
Each process of the ordered sequential bi-bi reaction mechanism of
Golgi-branching enzymes was modeled as an ODE:

Al-Hasani, H., Kunamneni, R.K., Dawson, K., Hinck, C.S., MullerWieland, D., and Cushman, S.W. (2002). Roles of the N- and C-termini
of GLUT4 in endocytosis. J. Cell Sci. 115, 131140.

dCi 
= SVproduction  SVconsumption
dt

Alegre, M.L., Frauwirth, K.A., and Thompson, C.B. (2001). T-cell regulation by CD28 and CTLA-4. Nat. Rev. Immunol. 1, 220228.

where [Ci] is the concentration of a component. Since the Golgi concentration of UDP-GlcNAc is 15 times that of cytosol (100 mM),
the Golgi UDP-GlcNAc is estimated to be in the millimolar range.
The probability of obtaining a specific glycoform is the product of the
probabilities of specific N-glycan occurrences at each N-X-S/T site:
Pglycoform =

n
Y

PN  glycanxi

i1

was calculated using the Golgi-pathway output of P(N-glycan(x)), then

partitioned into glycoform fractions according to avidity.
Glycoform fractions are filtered on the cell surface by their avidities for
galectin-3 and modeled using another set of ODEs where membrane
recycling and protein degradation are uncoupled. The kinetic parameters of membrane recycling are fitted to the avidity-enhancement scaling factor for lattice multivalency (see Supplemental Data for details).

Cell Lines and Growth Conditions

Mgat5+/+(2.6) and Mgat5/(22.9) mammary tumor cell lines were established from spontaneous mammary carcinomas in polyoma middle
T (PyMT) transgenic mice as previously described (Granovsky et al.,
2000). Tumor cells, NMuMG (NMARU mammary gland), and
HEK293T cells were propagated in DMEM (Gibco) supplemented
with 4.5 g/L glucose and 10% fetal bovine serum plus 10 mg/ml insulin
(Sigma) for NMuMG. Cell cultures were supplemented with various
concentrations of GlcNAc (Sigma) for 48120 hr. For Mgat1 overexpression, cells were infected with pMX-PIE-carrying human Mgat1
and GFP separated by IRES. myc-GLUT4-GFP/mutant constructs
were transfected into HEK293T cells using Effectene (Qiagen), fixed,
and stained with anti-myc antibody (Santa Cruz) overnight. To reveal
cell-adhesion junctions, cells were fixed and stained with anti-E-cadherin antibodies as previously described (Partridge et al., 2004).

Supplemental Data
Supplemental Data include ten figures, four tables, Experimental Procedures, and References and can be found with this article online at
http://www.cell.com/cgi/content/full/129/1/123/DC1/.

Apweiler, R., Hermjakob, H., and Sharon, N. (1999). On the frequency

of protein glycosylation, as deduced from analysis of the SWISS-PROT
database. Biochim. Biophys. Acta 1473, 48.
Boehmelt, G., Wakeham, A., Elia, A., Sasaki, T., Plyte, S., Potter, J.,
Yang, Y., Tsang, E., Ruland, J., Iscove, N.N., et al. (2000). Decreased
UDP-GlcNAc levels abrogate proliferation control in EMeg32-deficient
cells. EMBO J. 19, 50925104.
Bradshaw, J.D., Lu, P., Leytze, G., Rodgers, J., Schieven, G.L.,
Bennett, K.L., Linsley, P.S., and Kurtz, S.E. (1997). Interaction of the
cytoplasmic tail of CTLA-4 (CD152) with a clathrin-associated protein
is negatively regulated by tyrosine phosphorylation. Biochemistry 36,
1597515982.
Broschat, K.O., Gorka, C., Page, J.D., Martin-Berger, C.L., Davies,
M.S., Huang Hc, H.C., Gulve, E.A., Salsgiver, W.J., and Kasten, T.P.
(2002). Kinetic characterization of human glutamine-fructose-6-phosphate amidotransferase I: potent feedback inhibition by glucosamine
6-phosphate. J. Biol. Chem. 277, 1476414770.
Buckhaults, P., Chen, L., Fregien, N., and Pierce, M. (1997). Transcriptional regulation of N-acetylglucosaminyltransferase V by the src
oncogene. J. Biol. Chem. 272, 1957519581.
Chikuma, S., and Bluestone, J.A. (2002). CTLA-4: Acting at the synapse. Mol. Interv. 2, 205208.
Cipollo, J.F., Awad, A.M., Costello, C.E., and Hirschberg, C.B. (2005).
N-Glycans of Caenorhabditis elegans are specific to developmental
stages. J. Biol. Chem. 280, 2606326072.
Collins, B.E., Smith, B.A., Bengtson, P., and Paulson, J.C. (2006).
Ablation of CD22 in ligand-deficient mice restores B cell receptor signaling. Nat. Immunol. 7, 199206.
Demetriou, M., Granovsky, M., Quaggin, S., and Dennis, J.W. (2001).
Negative regulation of T-cell activation and autoimmunity by mgat5
N-glycosylation. Nature 409, 733739.
Di Paolo, G., and De Camilli, P. (2006). Phosphoinositides in cell
regulation and membrane dynamics. Nature 443, 651657.
Do, K.-Y., Fregien, N., Pierce, M., and Cummings, R.D. (1994). Modification of glycoproteins by N-acetylglucosaminyltransferase V is
greatly influenced by accessibility of the enzyme to oligosacharide
acceptors. J. Biol. Chem. 269, 2345623464.

Cell 129, 123134, April 6, 2007 2007 Elsevier Inc. 133

Ehrlich, M., Shmuely, A., and Henis, Y.I. (2001). A single internalization
signal from the di-leucine family is critical for constitutive endocytosis
of the type II TGF-beta receptor. J. Cell Sci. 114, 17771786.
Elleman, T.C., Frenkel, M.J., Hoyne, P.A., McKern, N.M., Cosgrove, L.,
Hewish, D.R., Jachno, K.M., Bentley, J.D., Sankovich, S.E., and Ward,
C.W. (2000). Mutational analysis of the N-linked glycosylation sites of
the human insulin receptor. Biochem. J. 347, 771779.
Ferrell, J.E.J., and Machleder, E.M. (1998). The biochemical basis of an
all-or-none cell fate switch in Xenopus oocytes. Science 280, 895898.
Frauwirth, K.A., Riley, J.L., Harris, M.H., Parry, R.V., Rathmell, J.C.,
Plas, D.R., Elstrom, R.L., June, C.H., and Thompson, C.B. (2002).
The CD28 signaling pathway regulates glucose metabolism. Immunity
16, 769777.
Granovsky, M., Fata, J., Pawling, J., Muller, W.J., Khokha, R., and
Dennis, J.W. (2000). Suppression of tumor growth and metastasis in
mgat5-deficient mice. Nat. Med. 6, 306312.
Hirabayashi, J., Hashidate, T., Arata, Y., Nishi, N., Nakamura, T., Hirashima, M., Urashima, T., Oka, T., Futai, M., Muller, W.E., et al. (2002).
Oligosaccharide specificity of galectins: a search by frontal affinity
chromatography. Biochim. Biophys. Acta 1572, 232254.
Ioffe, E., and Stanley, P. (1994). Mice lacking N-acetylglucosaminyltransferase I activity die at mid-gestation, revealing an essential role
for complex or hybrid N-linked carbohydrates. Proc. Natl. Acad. Sci.
USA 91, 728732.

transporter 2 glycosylation promotes insulin secretion in suppressing

diabetes. Cell 123, 13071321.
Partridge, E.A., Le Roy, C., Di Guglielmo, G.M., Pawling, J., Cheung,
P., Granovsky, M., Nabi, I.R., Wrana, J.L., and Dennis, J.W. (2004).
Regulation of cytokine receptors by Golgi N-glycan processing and
endocytosis. Science 306, 120124.
Sasai, K., Ikeda, Y., Fujii, T., Tsuda, T., and Taniguchi, N. (2002). UDPGlcNAc concentration is an important factor in the biosynthesis of
beta1,6-branched oligosaccharides: regulation based on the kinetic properties of N-acetylglucosaminyltransferase V. Glycobiology 12, 119127.
Schachter, H. (1986). Biosynthetic controls that determine the branching and microheterogeneity of protein-bound oligosaccharides.
Biochem. Cell Biol. 64, 163181.
Schoeberl, B., Eichler-Jonsson, C., Gilles, E.D., and Muller, G. (2002).
Computational modeling of the dynamics of the MAP kinase cascade
activated by surface and internalized EGF receptors. Nat. Biotechnol.
20, 370375.
Siegel, P.M., and Massague, J. (2003). Cytostatic and apoptotic actions
of TGF-beta in homeostasis and cancer. Nat. Rev. Cancer 3, 807821.
Slawson, C., Housley, M.P., and Hart, G.W. (2006). O-GlcNAc cycling:
how a single sugar post-translational modification is changing the way
we think about signaling networks. J. Cell. Biochem. 97, 7183.
Sobko, A. (2006). Systems biology of AGC kinases in fungi. Sci. STKE
2006, re9.

Ishida, H., Togayachi, A., Sakai, T., Iwai, T., Hiruma, T., Sato, T.,
Okubo, R., Inaba, N., Kudo, T., Gotoh, M., et al. (2005). A novel
beta1,3-N-acetylglucosaminyltransferase (beta3Gn-T8), which synthesizes poly-N-acetyllactosamine, is dramatically upregulated in
colon cancer. FEBS Lett. 579, 7178.

Takamatsu, S., Oguri, S., Minowa, M.T., Yoshida, A., Nakamura, K.,
Takeuchi, M., and Kobata, A. (1999). Unusually high expression of
N-acetylglucosaminyltransferase-IV a in human choriocarcinoma cell
lines: a possible enzymatic basis of the formation of abnormal biantennary sugar chain. Cancer Res. 59, 39493953.

Katz, E.B., Stenbit, A.E., Hatton, K., DePinho, R., and Charron, M.J.
(1995). Cardiac and adipose tissue abnormalities but not diabetes in
mice deficient in GLUT4. Nature 377, 151155.

Tsao, T.S., Burcelin, R., Katz, E.B., Huang, L., and Charron, M.J.
(1996). Enhanced insulin action due to targeted GLUT4 overexpression
exclusively in muscle. Diabetes 45, 2836.

Klein, P., Pawson, T., and Tyers, M. (2003). Mathematical modeling

suggests cooperative interactions between a disordered polyvalent
ligand and a single receptor site. Curr. Biol. 13, 16691678.

Umana, P., and Bailey, J.E. (1997). A mathematical model of N-linked

glycoform biosynthesis. Biotechnol. Bioeng. 55, 890908.

Koli, K.M., and Arteaga, C.L. (1997). Processing of the transforming

growth factor beta type I and II receptors. Biosynthesis and ligandinduced regulation. J. Biol. Chem. 272, 64236427.
Kolm-Litty, V., Sauer, U., Nerlich, A., Lehmann, R., and Schleicher,
E.D. (1998). High glucose-induced transforming growth factor beta1
production is mediated by the hexosamine pathway in porcine glomerular mesangial cells. J. Clin. Invest. 101, 160169.
Koshland, D.E., Jr., Goldbeter, A., and Stock, J.B. (1982). Amplification
and adaptation in regulatory and sensory systems. Science 217, 220
225.
Krambeck, F.J., and Betenbaugh, M.J. (2005). A mathematical model
of N-linked glycosylation. Biotechnol. Bioeng. 92, 711728.
Kumar, R., and Stanley, P. (1989). Transfection of a human gene that
corrects the Lec1 glycosylation defect: evidence for transfer of the
structural gene for N-acetylglucosaminyltransferase I. Mol. Cell. Biol.
9, 57135717.
Matsuura, I., Denissova, N.G., Wang, G., He, D., Long, J., and Liu, F.
(2004). Cyclin-dependent kinases regulate the antiproliferative function of SMADs. Nature 430, 226231.
Morgan, R., Gao, G., Pawling, J., Dennis, J.W., Demetriou, M., and Li,
B. (2004). N-acetylglucosaminyltransferase V (mgat5)-mediated N-glycosylation negatively regulates Th1 cytokine production by T cells.
J. Immunol. 173, 72007208.
Oft, M., Akhurst, R.J., and Balmain, A. (2002). Metastasis is driven by
sequential elevation of H-ras and SMAD2 levels. Nat. Cell Biol. 4, 487
494.
Ohtsubo, K., Takamatsu, S., Minowa, M.T., Yoshida, A., Takeuchi, M.,
and Marth, J.D. (2005). Dietary and genetic control of glucose

134 Cell 129, 123134, April 6, 2007 2007 Elsevier Inc.

van Delft, S., Schumacher, C., Hage, W., Verkleij, A.J., and van Bergen
en Henegouwen, P.M. (1997). Association and colocalization of EPS15
with adaptor protein-2 and clathrin. J. Cell Biol. 136, 811821.
Vilar, J.M., Jansen, R., and Sander, C. (2006). Signal processing in the
TGF-beta superfamily ligand-receptor network. PLoS Comput. Biol. 2,
e3. 10.1371/journal.pcbi.0020003.
Wang, F., Herzmark, P., Weiner, O.D., Srinivasan, S., Servant, G., and
Bourne, H.R. (2002). Lipid products of PI(3)Ks maintain persistent cell
polarity and directed motility in neutrophils. Nat. Cell Biol. 4, 513518.
Wang, Y., Tan, J., Sutton-Smith, M., Ditto, D., Panico, M., Campbell,
R.M., Varki, N.M., Long, J.M., Jaeken, J., Levinson, S.R., et al.
(2001). Modeling human congenital disorder of glycosylation type IIa
in the mouse: conservation of asparagine-linked glycan-dependent
functions in mammalian physiology and insights into disease pathogenesis. Glycobiology 11, 10511070.
Warren, C.E., Krizus, A., Roy, P.J., Culotti, J.G., and Dennis, J.W.
(2002). The Caenorhabditis elegans gene, gly-2, can rescue the N-acetylglucosaminyltransferaseV mutation of Lec4 cells. J. Biol. Chem.
277, 2282922838.
Wiley, H.S., Shvartsman, S.Y., and Lauffenburger, D.A. (2003).
Computational modeling of the EGF-receptor system: a paradigm for
systems biology. Trends Cell Biol. 13, 4350.
Zhen, Y., Caprioli, R.M., and Staros, J.V. (2003). Characterization of
glycosylation sites of the epidermal growth factor receptor. Biochemistry 42, 54785492.
Zhu, S., Hanneman, A., Reinhold, V.N., Spence, A.M., and Schachter,
H. (2004). Caenorhabditis elegans triple null mutant lacking UDP-Nacetyl-D-glucosamine:alpha-3-D-mannoside beta1,2-N-acetylglucosaminyltransferase I. Biochem. J. 382, 9951001.