Highly site-selective stability increases by glycosylation of
dihydrofolate reductase
Lai-Hock Tey
1
,E. Joel Loveridge
1
, Richard S. Swanwick
1,
*, Sabine L. Flitsch
2
and Rudolf K. Allemann
1
1 School of Chemistry, Cardiff University, UK
2 School of Chemistry and Manchester Interdisciplinary Biocentre, University of Manchester, UK
Introduction
Post-translational glycosylation is one of the most
abundant forms of covalent protein modification in
eukaryotic cells and plays an important role in deter-
mining the properties of proteins, affecting many
molecular processes in vivo [1–5]. There are two main
types of protein glycosylation: N-glycosylation, in
which the oligosaccharide is attached to an asparagine
side chain, and O-glycosylation, in which it is attached
to the side chain of serine or threonine residues [4].
Surface glycoproteins act as markers for inter- and
intracellular communication, and glycosylation has
been shown to affect a number of protein properties
such as structure, dynamics, stability and catalytic
activity [6–14].
Glycosylation stabilizes many proteins against ther-
mal denaturation [14–17], whereas the removal of car-
bohydrates from naturally glycosylated proteins can
lead to decreased thermal stability and an increased
tendency towards protein aggregation [18–20]. Some
studies have shown that glycans reduce the rate of
unfolding but do not affect refolding of denatured pro-
teins, leading to the conclusion that glycans preferen-
tially bind to the folded protein and therefore stabilize
it [20–24]. Others have also shown that folding is pro-
moted in the presence of glycans [18,25,26], suggesting
that the effects are protein specific. Notably, many
proteins show considerable increases in thermostability
when in solution with high concentrations of sugars or
Keywords
enzyme; glycosylation; kinetics;
mutagenesis; stability
Correspondence
R. K. Allemann, School of Chemistry, Cardiff
University, Main Building, Park Place, Cardiff
CF10 3AT, UK
Fax: +44 29 2087 4030
Tel: +44 29 2087 9014
E-mail: allemannrk@cf.ac.uk
*Present address
Department of Life Sciences, Imperial
College, London, UK
(Received 10 January 2010, revised 26
February 2010, accepted 2 March 2010)
doi:10.1111/j.1742-4658.2010.07634.x
Post-translational glycosylation is one of the most abundant forms of cova-
lent protein modification in eukaryotic cells. It plays an important role in
determining the properties of proteins, and stabilizes many proteins against
thermal denaturation. Protein glycosylation may establish a surface micro-
environment that resembles that of unglycosylated proteins in concentrated
solutions of sugars and other polyols. We have used site-directed mutagen-
esis to introduce a series of unique cysteine residues into a cysteine-free
double mutant (DM, C85A ⁄C152S) of dihydrofolate reductase from
Escherichia coli (EcDHFR). The resulting triple mutants, DM-N18C,
DM-R52C, DM-D87C and DM-D132C EcDHFR, were alkylated with
glucose, N-acetylglucosamine, lactose and maltotriose iodoacetamides. We
found little effect on catalysis or stability in three cases. However, when
DM-D87C EcDHFR is glycosylated, stability is increased by between 1.5
and 2.6 kcalÆmol
)1
in a sugar-dependent manner. D87 is found in a hinge
region of EcDHFR that loses structure early in the thermal denaturation
process, whereas the other glycosylation sites are found in regions involved
in the later stages of temperature-induced unfolding. Glycosylation at this
site may improve the stability of EcDHFR by protecting a region of the
enzyme that is particularly prone to denaturation.
Abbreviations
DM, double mutant; EcDHFR, Escherichia coli dihydrofolate reductase.
FEBS Journal 277 (2010) 2171–2179 ª2010 The Authors Journal compilation ª2010 FEBS 2171
other polyols [27]. Protein glycosylation may therefore
establish a surface microenvironment that resembles
that of unglycosylated proteins in such solutions.
Several methods have been described for the genera-
tion of neoglycoproteins via site-selective glycosylation
of proteins using chemical modification of biotechno-
logically produced proteins [2,3,28–31]. One such
approach combines site-directed mutagenesis, to intro-
duce unique cysteine residues at the required sites, and
a highly flexible but selective chemical derivatization
strategy (Scheme 1) in which reaction of the free thiol
group of a cysteine residue with a synthetic glycosyl
iodoacetamide produces a stable linkage between the
protein and the carbohydrate [30] which resembles that
found in native glycosylation of asparagines [32–34].
We have previously used this approach to study of
the effect of site-specific glycosylation on the physical
and chemical properties of the naturally nonglycosylated
Scheme 1. Strategy used for the synthesis
of highly purified glycosylated Escherichia
coli dihydrofolate reductase triple mutants
[30]. A unique cysteine residue on the pro-
tein is first reacted with a glycosyl iodoace-
tamide (glucose is used as an example
here); unalkylated proteins are biotinylated
by reaction with 2-((biotinoyl)amino)ethyl
methanethiosulfonate. Treatment with
resin-bound avidin removes the biotinylated
protein from solution, leaving highly purified
neoglycoprotein.
Glycosylation of E. coli DHFR L.-H. Tey et al.
2172 FEBS Journal 277 (2010) 2171–2179 ª2010 The Authors Journal compilation ª2010 FEBS
enzyme dihydrofolate reductase (5,6,7,8-tetrahydro-
folate : NADP
+
oxidoreductase, EC 1.5.1.3) from
Escherichia coli (EcDHFR) [14]. EcDHFR catalyses
the stereospecific reduction of 7,8-dihydrofolate to
(6S)-5,6,7,8-tetrahydrofolate using NADPH as a cofac-
tor [35], and is therefore responsible for maintaining
the tetrahydrofolate pool within the cell. EcDHFR is a
monomeric enzyme made up of eight b-strands, four
ahelices and a number of important loop regions; it is
typically divided into three subdomains, the adenosine-
binding domain, the substrate-binding domain and the
loop domain (Fig. 1) [36]. Our previous study was
based on a cysteine-free C85A ⁄C152S double mutant
of EcDHFR (DM EcDHFR), which has similar fold-
ing, stability and kinetic properties to the wild-type
enzyme (WT EcDHFR) [37]. Cysteine residues were
introduced at two sites and the effect of glycosylation
at these sites was studied [14]. Substitution of a
cysteine residue at position 87 (to form DM-D87C
EcDHFR) caused a loss in thermostability of the
protein that was reversed on glycosylation, whereas
DM-E120C EcDHFR had similar thermostability to
the native enzyme and subsequent glycosylation led to
a smaller increase in melting temperature than that
observed at position 87 [14]. The kinetic parameters of
the steady-state reaction catalysed by EcDHFR were
not significantly affected by mutation and subsequent
glycosylation at either position [14]. This difference in
response to glycosylation at the two sites was intrigu-
ing and prompted further study. Here, we describe
the effect of glycosylation at three further sites on
EcDHFR and report the kinetic properties, thermal
stability and chemical stability at room temperature of
the resulting glycoproteins. The sites chosen were N18,
on the catalytically important M20 loop, R52, respon-
sible for binding the glutamate tail of the substrate,
and D132, ‘behind’ the active site at the end of the FG
loop (Fig. 1). Our results suggest that the local envi-
ronment of the protein is critically important in deter-
mining the effect of the glycosyl chain on protein
unfolding.
Results
Preparation of glycosylated EcDHFR mutants
Double and triple mutants of EcDHFR were prepared
using standard molecular biology techniques and the
proteins expressed, purified, glycosylated and further
purified as described previously [30]. Prior to glycosyl-
ation, all proteins were > 95% pure as judged by
SDS–PAGE. Glycosylation was confirmed by tryptic
digestion followed by MALDI-TOF MS (Fig. S1).
Ligand binding and kinetics of glycosylated EcDHFR
Quenching of the enzyme fluorescence at 340 nm was
used to determine the equilibrium dissociation con-
stants of enzyme–NADPH and enzyme–folate com-
plexes. All five mutants have K
D
values similar to WT
EcDHFR for both NADPH and folate (supporting
information). The largest change was seen for
DM-R52C with folate, where a threefold loss of affin-
ity was seen. In addition, no significant differences
between the kinetic parameters of the five mutants and
those of the wild-type protein were observed in either
the steady state or pre-steady state, nor were there any
reliable trends in the values on glycosylation (support-
ing information).
Stability of glycosylated EcDHFR
The far-UV CD spectra of the EcDHFR double and
triple mutants and of the glycosylated triple mutants
Fig. 1. Structure of Escherichia coli dihydrofolate reductase
(PBD 1RA2) [36] showing the position of the five residues mutated
to cysteine for this study. The two views are rotated 180about
the z-axis relative to each other. The adenosine-binding domain
(ABD), substrate-binding domain (SBD), loop domain (LD) and spe-
cific loops mentioned in the discussion are indicated. The enzyme
is shown as a cartoon representation; residues of interest and
ligands are shown as sticks. H
2
F, 7,8-dihydrofolate.
L.-H. Tey et al. Glycosylation of E. coli DHFR
FEBS Journal 277 (2010) 2171–2179 ª2010 The Authors Journal compilation ª2010 FEBS 2173
were all similar to those of the wild-type enzyme, indi-
cating that neither the mutations per se nor glycosyla-
tion had an effect on the secondary structure of the
proteins large enough to be detectable by CD spectros-
copy (supporting information). Thermal denaturation
of WT EcDHFR and the five mutants was reversible
from 80 to 20 C, and the melting temperatures of all
proteins except DM-D87C EcDHFR were similar
(Table 1). It has previously been shown that DM
EcDHFR has similar stability to the WT protein [37].
As previously reported, the thermal denaturation tem-
perature of DM-D87C EcDHFR is almost 10 C lower
than that of WT EcDHFR, even though there is no
significant difference in the secondary structure of the
two proteins, and stability is restored by glycosylation
[14]. The change in stability is because of the glycan
rather than the acetamide linkage [14]. Stability of the
glycosylated mutant proteins was also determined
using equilibrium urea titrations monitored by trypto-
phan fluorescence emission (Table 2 and supporting
information). Mirroring the thermal stability results,
DM EcDHFR and three of the four triple mutants
showed little change in resistance to urea denaturation,
although DM-D87C EcDHFR showed a considerably
lower free energy of unfolding, indicating a signifi-
cantly lower stability. The free energy of unfolding of
DM-D87C EcDHFR was increased by glycosylation,
although the other mutants were unaffected. The sta-
bility of glycosylated DM-D87C EcDHFRs increased
with the length of the glycosyl chain; monosaccharides
caused a similar increase in free energy of unfolding as
incubating the nonglycosylated enzyme in a 0.5 msolu-
tion of maltose, whereas larger sugars gave a more
pronounced effect.
Discussion
We have previously reported a large reduction in ther-
mal stability for DM-D87C EcDHFR and its subse-
quent ‘rescue’ by glycosylation [14]. The same study
showed a slight increase in thermal stability on glyco-
sylation of DM-E120C EcDHFR. Here we demon-
strate that three further EcDHFR triple mutants show
similar stability (against both temperature- and urea-
induced denaturation) to the wild-type protein and
that, in these cases, glycosylation does not improve
Table 1. Melting temperatures of EcDHFR, its mutants and their glycosylated forms. Values were determined by CD spectroscopy using
10 lMenzyme in 5 mMpotassium phosphate buffer (pH 7.0). DM, double mutant; EcDHFR, Escherichia coli dihydrofolate reductase; WT,
wild-type, ND, not determined.
Glycan
T
m
(C)
WT
EcDHFR-C85A ⁄
C152S (DM) DM-N18C DM-R52C DM-D87C [14] DM-E120C [14] DM-D132C
None 50.7 ± 0.2 50.9 ± 0.9 50.5 ± 0.4 49.7 ± 0.5 40.9 ± 0.3 50.8 ± 0.4 50.7 ± 0.3
Glucose 50.5 ± 0.3 49.9 ± 0.2 47.1 ± 0.3 51.8 ± 0.2 52.8 ± 0.9
N-acetylglucosamine 51.0 ± 0.1 50.2 ± 0.9 49.6 ± 1.1 52.5 ± 0.1 52.1 ± 1.6
Lactose 50.6 ± 0.9 51.6 ± 0.1 46.8 ± 2.1 54.1 ± 0.1 51.2 ± 0.7
Maltotriose 51.0 ± 0.7 49.8 ± 1.0 47.5
a
ND 52.4 ± 0.6
0.5 MMaltose 52.9 ± 0.4 53.5
a
53.5
a
43.4 ± 0.4 ND 54.6
a
a
Single measurement.
Table 2. Free energy of unfolding of Escherichia coli dihydrofolate (EcDHFR), its mutants and their glycosylated forms. Values were deter-
mined by fluorescence intensity measurement of urea-induced unfolding of 2 lMenzyme in 10 mMpotassium phosphate buffer (pH 7.0).
DM, double mutant; WT, wild-type.
Glycan
DG(kcalÆmol
)1
)
WT
EcDHFR-C85A ⁄
C152S (DM) DM-N18C DM-R52C DM-D87C DM-D132C
None 5.9 ± 0.3 5.2 ± 0.3 5.3 ± 0.3 5.7 ± 0.2 2.7 ± 0.2 5.5 ± 0.3
Glucose 5.7 ± 0.1 5.5 ± 0.2 4.2 ± 0.1 5.3 ± 0.2
N-acetylglucosamine 5.5 ± 0.2 5.4 ± 0.2 4.3 ± 0.3 5.2 ± 0.3
Lactose 5.6 ± 0.3 5.5 ± 0.2 5.1 ± 0.1 5.4 ± 0.3
Maltotriose 5.3 ± 0.3 5.4 ± 0.1 5.3 ± 0.3 5.1 ± 0.2
0.5 MMaltose 6.1 ± 0.2 5.5 ± 0.2 5.4 ± 0.2 4.1 ± 0.1 5.5 ± 0.1
Glycosylation of E. coli DHFR L.-H. Tey et al.
2174 FEBS Journal 277 (2010) 2171–2179 ª2010 The Authors Journal compilation ª2010 FEBS
stability (Table 1). In all cases except that of
DM-D87C EcDHFR, site-selective glycosylation had a
smaller effect on the thermal stability of the proteins
than the presence of 0.5 mmaltose. A similar trend
was observed for the stabilities of the proteins at
ambient temperature with respect to denaturation
induced by urea (Table 2). Notably, although the free
energy of unfolding of DM-D87C increased with
increasing glycan length, the thermal stability did not
show such a trend and was instead highest with the
monosaccharide N-acetylglucosamine. Glycosylation of
DM-D87C EcDHFR with lactose or maltotriose
acetamides had a larger effect on the free energy of
unfolding than a 0.5 msolution of maltose. Interest-
ingly, the changes in free energy of unfolding were
because of changes in the gradient of the urea depen-
dence of the free energy (supporting information),
rather than changes to the mid-point of the urea-
induced unfolding. This suggests that glycosylation of
DM-D87C EcDHFR affects the cooperativity of the
unfolding transition rather than simply the resistance
to denaturants [38]. The results presented here provide
further support [14] that, at least in the case of dihy-
drofolate reductase, increased stability through the
addition of glycans is because of highly site-specific
effects rather than nonspecific changes to the solvation
properties of the enzyme, suggesting that stabilization
of EcDHFR relies on specific interactions between the
protein and the glycan. In the case of DM-D87C
EcDHFR, it appears that glycosylation increases the
effective concentration of sugar at a critical site to
more than that provided by a 0.5 msolution of malt-
ose. In fact, the increase in melting temperature is sim-
ilar to that seen in a 1.5 m(50% w ⁄v) solution of
sucrose [39]. By contrast, site-selective glycosylation of
the protein in regions unimportant for glycan-induced
stability would produce no benefit, as observed here.
Inspection of the EcDHFR structure reveals no fea-
ture around position 87 that would be expected to
interact particularly favourably with glycans. Compu-
tational [40,41] and experimental [42] work has indi-
cated that the ‘hinge’ region of the adenosine-binding
domain in which D87 is found unfolds very early in
the denaturation process, although others [43–45] have
suggested an alternative folding pathway in which this
region would be expected to unfold slightly later. If
this region does lose structure early in the unfolding
process, this may explain why the stability of
EcDHFR is sensitive to glycosylation at this site –
regions more vulnerable to denaturation are likely to
benefit more from additional stabilizing interactions.
N18, E120 and D132 are all found in the loop
domain, which retains structure until relatively late in
the thermal unfolding process [40,41], whereas R52 is
formally located in the adenosine-binding domain but
forms part of the substrate-binding pocket. Sugars
bound at position 52 are therefore more likely to
interact with the relatively stable [40–42] substrate-
binding domain (Fig. 1). Hence glycosylation at these
positions may not exert a similarly stabilizing effect as
glycosylation at position 87.
Both ligand binding and the kinetics of EcDHFR
were remarkably robust to the mutations made and
subsequent glycosylation. The most notable difference
in K
D
values was observed for DM-R52C EcDHFR
with folate, although this is still only an approximately
threefold increase. R52 forms part of the binding site
for the glutamate tail of the folate ligand, whereas
N18 forms part of the M20 loop, which closes over
NADPH after it enters the active site (Fig. 1). It has
previously been shown that reacting DM-N18C and
DM-E17C mutants with bulky groups has little effect
on their kinetics or ligand binding relative to EcDHFR
[46,47]. E120 and D132 are both located on the FG
loop, important because of its interactions with the
M20 loop that controls progression through the cata-
lytic cycle [36,48]. Mutation of glycine 121 to bulkier
residues causes a sharp decrease in catalytic activity,
and a reduction in the affinity for NADPH [49,50].
However, this is likely to be because of global struc-
tural changes observed for the G121V mutant [40],
which disrupt the ability of the EcDHFR : NADPH
complex (and the reactive Michaelis complex) to form
its native ‘closed’ conformation [48]. Changes at posi-
tion 120, where the side chain is exposed to solvent,
would not be expected to produce so pronounced an
effect. The absence of large effects on catalysis pro-
vides further evidence that mutation and subsequent
glycosylation do not produce significant changes in the
global structure of the enzyme, but that stability of
EcDHFR may be affected by binding of sugars to
specific sites on the enzyme.
In conclusion, our previous study suggested that the
thermal stability of proteins can be increased signifi-
cantly by the attachment of even relatively small car-
bohydrates, rather than the larger oligosaccharides
typically found in nature [14]. We now report a similar
effect on chemical stability at room temperature, and
add that the local environment of the protein appears
to be critically important in determining the effect of
bound oligosaccharides. The large oligosaccharides
observed in nature may allow greater coverage of a
number of discrete, critical points of stabilization from
a single glycosylation site, rather than being simply
caused by blanket coverage of large regions of the
protein surface. Alternatively, increases in stability
L.-H. Tey et al. Glycosylation of E. coli DHFR
FEBS Journal 277 (2010) 2171–2179 ª2010 The Authors Journal compilation ª2010 FEBS 2175