Yeast glycogenin (Glg2p) produced in
Escherichia coli
is
simultaneously glucosylated at two vicinal tyrosine residues but
results in a reduced bacterial glycogen accumulation
Tanja Albrecht
1
, Sophie Haebel
2
, Anke Koch
1
, Ulrike Krause
1
, Nora Eckermann
1
and Martin Steup
1
1
Institute of Biochemistry and Biology and
2
Interdisciplinary Center for Mass Spectrometry of Biopolymers, University of Potsdam,
Potsdam-Golm, Germany
Saccharomyces cerevisiae possesses two glycogenin isoforms
(designated as Glg1p an d Glg2p) that both contain a con-
served tyrosine residue, Tyr232. Howe ver, Glg2 p posse sse s
an additional tyrosine residue, Tyr230 and therefore two
potential autoglucosylation sites. Glucosylation of Glg2p
was studied using both matrix-assisted laser desorption
ionization and electrospray quadrupole time o f flight mass
spectrometry. Glg2p, carrying a C-terminal (His
6

carbon compounds are i nsufficiently available. Both auto-
trophic and heterotrophic prokaryotes synthesize glycogen
as fungi and anim als d o [1,2], whereas almost all plastid-
containing organisms synthesize starch particles [3]. The
prokaryotic glycogen synthases ( EC 2.4.1.11) use ADPglu-
cose as the glucosyl donor (which is als o the substrate of the
various eukaryotic starch synthases), whereas the polyglucan
synthases from both fungi a nd animals rely on UDPglucose
[3]. The transfer o f glucosyl residues from either UDPglucose
or ADPglucose to the nonreducing e nd of an a-glucan-like
primer, as catalyzed by glycogen synthases, results in an
elongation of a linear oligoglu can or a polysacc h aride chain
and, in conjunction with the branching enzyme (EC
2.4.1.18), in the formation of a glycogen-like molecule
[4,5]. However, at least in eukaryotes the cooperation of
these two enzymes does not permit the de novo synthesis of a
glucan. Both in fungi and in animals glycogen biosynthesis
appears to b e initiated by the action of another U DPglucose-
dependent glucosyltransferase, designated as glycogenin (EC
2.4.1.186) [6,7]. This homodimeric protein is thought to
comprise several distinct enzymatic activities: F irst, in an
autocatalytic intersubunit reaction it transfers a glucosyl
moiety to a t yrosin residue forming a glucose 1-O-tyrosyl
linkage [8]. Second, several glucosyl residues are sequentially
transferred to the glucosylglycogenin resulting in an oligo-
glucan chain that is covalently bound to the glycogenin. It is
possible that these glucosylation reactions (or at least some
of them) are due to an intramonomer glucosyl transfer and
therefore d iffer mechanistically from the initial glucosylation
step(s) [9]. Third, glycogenin is capable of transferring

[7]. Recen tly, it has been proposed that in Agrobacterium
tumefaciens glycogen synthase catalyzes both an A DPglu-
cose-dependent autoglycosylation and an ADPglucose-
dependent glucan elongation, suggesting that it f unctionally
replaces glycogenin [18]. Similarly, the i nitial reactions of the
eukaryotic amylopectin and/or starch granule formation are
not known yet. In t he genome of Arabidopsis th aliana L., at
least seven glycogenin orthologues h ave been identified but
the biochemical functions (and the intracellular locations) of
the products of all these genes remain to be defined.
In Saccharomyces cerevisiae, two glycogenin isoforms
(designated as Glg1p and Glg2p) that a ppear to be
functionally equivalent are known. This assumption is
based o n e xperiments in which a yeast mutant deficient in
both f unctional glycogenin genes was transformed with
either the GLG1 or the GLG2 gene and each transformation
restored glycogen biosynthesis [19]. In the N-terminal
domains (which contain the autoglucosylation region)
Glg1p and Glg2p possess a 55% sequence i dentity. Both
Glg1p and Glg2p c ontain one conserved t yrosine residue,
Tyr232, which presumably corresponds to the single auto-
glucosylation site of the rabbit skeletal g lycogenin, Ty r194
[20]. Unlike Glg1p, Glg2p possesses another tyrosine
residue, Tyr230 located i n close vicinity to the conserved
Tyr232. In in vitro assays performed with Glg2p, mutation
of either Tyr230 or Tyr232 resulted in a partial loss of the
autoglucosylation activity which was completely abolished
when both tyrosine residues w ere replaced by phenylalanine
[10]. These data suggest that Glg2p possesses two self-
glucosylation sites. H owever, when t he yeast mutant d efi-

Cloning of Glg2p
S. cerevisiae strain EG328–1A (MATa trp1 leu2 ura3–52)
was used. Prior to total RNA preparation cells were grown
for 24 h at 30 °Cin
1
yeast nitrogen b ase medium c omple-
mented with amino a cids. Total RNA was isolated by using
the RNeasy midi kit (Qiagen, Hilden, Germany).
For R T-PCR t he following two primers were designed
according to the cDNA sequence of Glg2p (accession
number U25436) 5¢-ATGGCCAAGAAAGTTGCCATC
TGT; 3¢-TCAGGTATCAGGCTTTGGGAATGC. RT-
PCR was performed using SSII R NaseH
–
RT (Invitrogen,
Karlsruhe, Germany) and High Fidelity Expand Poly-
merase (Roche, Mannheim, Germany). For expression
experiments, the cDNA was subcloned i nto pET101/
D-TOPO (Invitrogen) providing a C-terminal (6xHis) tag.
The cDNA was confirmed by complete sequencing
(AGOWA,Berlin,Germany).
Production of Glg2p in
E. coli
and purification
For h eterologous expression, the E. coli s train BL21 Star
DE3 (Invitrogen) was used. Cells con taining a Glg2p
expression construct were grown on
2
tryptone–yeast extract
medium containing 100 lg ampicillin per mL a t 30 °C until

increasing concentrations o f imidazole ( pH 8.0), dissolved
in lysis buffer: 3 · 50 m
M
imidazole, 1 · 75 m
M
,and
1 · 250 m
M
.MostoftheGlg2pproteinwasreleasedby
the last e lution step.
Western blotting
Buffer-soluble proteins were separated by SDS/PAGE and
were then transferred to n itrocellulose (Protean, 0 .2 lm pore
size; Schleicher & Schuell) for 16 h at 20 V. The transfer
buffer c ontained 5 0 m
M
Tris, 150 m
M
glycine, 0.02% (w/v)
Ó FEBS 2004 Expression of yeast glycogenin (Glg2p) in E. coli (Eur. J. Biochem. 271) 3979
SDS, 20% (v/v) methanol [21]. The His-tagged Glg2p was
detected using a primary anti-(His)
5
IgG (Qiagen) and a
secondary anti-mouse immunoglobulin coupled to alkaline
phosphatase (Promega, Madison, USA).
Glycogen-related enzyme activities
E. coli cells were pelleted, washed in deionize d water
3
and

)1
) was i ncubated
at 30 °C in a mixture t hat contained, in a final volume o f
180 lL, 5 m
M
UDPglucose, 5 m
M
MnCl
2
and 50 m
M
HEPES/NaOH pH 7.5. At intervals, aliquots (30 lL) of
the reaction m ixture were withdrawn. Following the addi-
tion of 15 lL SDS containing sample buffer the protein was
denatured (5 min at 95 °C) and used for SDS/PAGE.
Protein
in-gel
digestion and extraction of peptides
Following SDS/PAGE and Coomassie blue staining, pro-
tein bands (approximately 7 .5 lg protein each) were t reated
as describe d in [25].
Protein cleavage by cyanogen bromide
Recombinant Glg2p (20 lg) was d issolved in 40 lLofa
cyanogen bromide solution (20 mgÆmL
)1
in 70% [v/v]
trifluoroacetic acid)
5
and incubated f or 4 h at room
temperature in darkness. Subsequ ently, t he reaction mixture

Pharmacia, Uppsala, Sweden) on a Pharmacia C2/C18 SC
2.1/10 column using a linear 0–50% (v/v) acetonitrile
gradient containing 0.1% (v/v) trifluoroacetic acid. A
constant flow rate of 100 lLÆmin
)1
was applied. In the
eluate, absorbance was monitored at 214 nm.
Matrix-assisted laser desorption/ionization time of flight
(MALDI-TOF) mass spectrometry
MALDI-TOF analyses were p erformed using a Reflex II
MALDI-TOF instrument (Bruker-Daltonik, B remen, Ger-
many). All spectra were recorded in the reflector mode. As
matrix 2,5-dihydroxybenzoic acid (20 mg DHB in 1 mL
20% (v/v) aqueous methanol) was used. Aliquots of the
eluate fractions of interest (2–3 lL each) were ap plied to the
target followed by t he addition of 1 lL of matrix solution
and drying under a gentle stream of air. To d etermine t he
glucosylation sites, mono-glucosylated peptides purified by
RP-HPLC were subjected to post source decay (PSD)
analysis.
Nanoelectrospray quadrupole time of flight (NanoESI
Q-TOF) mass spectrometry
MS/MS spectra were recorded using a API QSTAR pulsar I
(Applied Biosystems/MDS S ciex, Toronto, Canad a) hybrid
mass spectrometer equipped with a nanoelectrospray ion
source. The ion of interest was selected in the Q1 quadrupole.
Fragments were generated in the collision cell by c ollision
with Argon a nd analyzed in the TOF mass analyzer.
Glycogen extraction and quantification (procedure A)
Bacterial c ells (E. coli strain BL21 s tar DE3) were grown in

MgSO
4
, 1% (w/v) glucose, and 0.1 m
M
IPTG]. Under
these conditions the E. coli cells accumulate glycogen a nd
the expression of the transgene continues. At intervals
aliquots of the cell suspension (25 mL each) were with-
drawn a nd g lycogen was extracted according to [ 26].
Subsequently, the glucose content of the glycogen fraction
was determined enzymatically using t he starch kit
(r-biopharm, Darmstadt, Germany). Alternatively, glyco-
gen was extracted from the bacterial cells as described
below (procedure B ). For nitrogen starvation, NH
4
Cl was
omitted from the medium.
3980 T. Albrecht et al. (Eur. J. Biochem. 271) Ó FEBS 2004
Glycogen extraction and size distribution (procedure B)
For the determination o f the size distribution of glycogen
molecules an alternative extraction procedure was devel-
oped.
Bacterial cells (100 mL suspension) were pelleted by
centrifugation (12 min at 3000 g,4°C), resuspended in
8 mL deionized water and sonicated f or 90 s at 4 °C. The
homogenate was centrifuged for 1 2 m in at 20 000 g.The
supernatant containing low molecular mass glycans, glyco-
gen, nucleic acids, and s oluble proteins was heated (5 min at
100 °C). Denatured proteins were removedby centrifugation
(10 min at 10 000 g). High molecular mass nuc leic acids were

). Light scattering and concentra-
tion were detected with a multiangle DAWN DSP laser
photometer (He-Ne-laser; WTC, Santa Barbar a, USA) and
Optilab DSP Interferometric Refractometer (WTC), respect-
ively. The molecular mass distribution w as calculated fr om
light scattering and R I data by using the
ASTRA
software
(version 4.75 , WTC; e xtrapolation by D ebye, first order) .
Maltooligosaccharide patterns
Bacterial cells (100 mL suspension) were p ellete d by
centrifugation (12 min at 3000 g)andwashedwith
deionized water. Maltooligosaccharides were extrac ted with
10 mL 80% (v/v) aqueous ethanol for 15 min at 95 °C.
Following extraction insoluble c ompounds were removed
by centrifug ation ( 10 min at 2 0 000 g) a nd the s upernatant
containing the soluble carbohydrates was lyophilized. The
residue was resuspended in 4 mL deionized water and
proteins were removed from the aqu eous phase by treat-
ment with an equal volume o f c hloroform. Deproteinization
was repeated three times. Subsequently, the aqueous phase
was passed through a 10-kDa membrane (Millipore,
Germany) and the filtrate was lyophilized. Finally, the
residue was dissolved in 200 lL deionized water a nd used
for high performance anion exchange chromatography w ith
pulsed a mperometric d etection (HPAEC-PAD, D ionex
BioLC) using a CarboPac PA-100 column. Following
sample injection (90 lL e ach) the column was equilibrated
for10minwith5m
M

) were excised and digested with trypsin in
the gel. T he resulting peptide mixtures were eluted from the
gel pieces and analyzed by MALDI-TOF mass spectro-
metry. All major peptides of both samples could be a ssigned
to tryptic peptides d erived from the Glg2p sequence.
However, peptides containing Tyr230 and Tyr232 were
detected in neither the nonglucosylated nor the g lucosylated
form. In co ntrast, several nonglucosylated tryptic peptides
representing residues 3 60–370, 345–370, and 340–370 that
all contain the C-terminal Tyr362 were observed as major
peaks (data not shown). No traces of peaks with a mas s
increment of 162 Da, o r a multiple of it, were d etectable.
Thus, it appears that Tyr362, although essential for the
functionality of glycogenin, is not glucosylated.
In order to detect Glg2p-derived glucopeptides, the two
peptide mixtures generated by trypsination of Glg2p
0
and
Glg2p
20
were separated by RP-HPLC. For both mixtures,
essentially the same H PLC chromatograms were obtained
(data not shown). All collected fractions were analyzed by
MALDI-TOF MS. For both th e trypsinated Glg2p
0
and
Glg2p
20
glucosylated peptides were observed in fractions 12
and 13 (Fig. 2). Glucopeptides were detected as a series of

and Glg2p
0
mass spectra o f the HPLC f ractions 12 a nd 13 are s hown.
3982 T. Albrecht et al. (Eur. J. Biochem. 271) Ó FEBS 2004
amino acid sequence (residues 219–246) designated as P
1
.In
the nonglucosylated state the molecular mass of P
1
is
calculated to be 3352.8 Da. Taking into account that the
two methionine residues are likely to be oxidized during
analyte p rocessing the actual mass of the (nonglucosylated)
peptide P
1
is assumed to be 3384 .8 Da. B y using this m/z
value a nd the data shown in Fig. 2 it is estimated t hat 4–25
glucosyl residues are covalently bound to P
1
. This implies
that Glg2p i s significantly g lucosylated during p roduction in
E. coli. Glucosylation in E. coli has also been observed with
the rabbit muscle glycogenin [20]. I n the latter study 1–8
glucosyl residues were found to be linked to Tyr194.
Following se lf-glycosylation for 20 min, the glucosylation
of Glg2p is even more complex. At least 30 m/z signals
originating from the differently g lucosylated P
1
peptide were
detected (Fig. 2). In similar experiments, up to 40 glucosyl

frequently by cleavage of glycosidic linkages w hereas peptide
backbone fragments are suppressed. The latter are, however,
relevant for t he identification of glucosylation sites.
The HPLC chromatogram of a Glg2p-derived peptide
mixture, as obtained by cyanogen bromide and a-amylase
treatment, is shown in Fig. 3A. As revealed by MALDI-
TOF analysis, several eluate fractions contained an 11-mer
peptide having the sequence PNYGYQSSPAM (residues
228–238; designated as P
2
) but differing in the degree of
glucosylation. In fraction 19, P
2
was observed in the
monoglucosylated form d esignated as G -P
2a
. However, the
majority of the peptide P
2
occurs in higher glucosylated
forms as revealed by MALDI-TOF analysis of fractions
11–15. These g lucopeptides were resist ant to a prolonged or
repeated a-amylase treatment i ndicating an exhaustive
a-amylase action.
For a more effective deglucosylation of the higher
glucosylated forms of P
2
, a second enzymatic treatment
was included: RP-HPLC fractions 11–15 (Fig. 3A) were
pooled, l yophilized and incubated w ith a myloglucosidase.

2b
.
The two mono-glucosylated P
2
samples, G-P
2a
and G-P
2b
and the diglucosylated peptide, G
2
-P
2b
were analyzed by
fragmentation using both Q -TOF MS/MS a nd MALDI-
TOF P SD. F ragmentation often results from cleavage of the
peptide backbone. Fragments obtained are classified as
a-, b- or y-type fragments [28]. B oth a- a nd b-type fragments
contain the N-terminus of the peptide. Unlike a -type
fragments, b-fragments are generated by breakage of the
peptide bond and t herefore both types differ by one CO
group, i.e. a mass of 28 Da. All y-type fragments contain the
C-terminus of the p eptide. Many fragments show a satellite
peak at )17 Da. This peak is due to a loss of NH
3
,
presumably from an asparagine residue. The fragments
observed b y Q -TOF MS/MS for the two mono-glucosyl-
ated P
2
conjugates, G-P

)-162 peak. However, the occurrence of both b3
and b4 exclusively in the nonglucosylated state indicates
that in G-P
2a
, T yr230 does not carry a glucosyl residue and
therefore Tyr232 must be the glucosylated residue.
The fragment p attern obtained with G-P
2b
differs signi-
ficantly ( Table 1 and Fig. 4B). All fragments from b3 to b8
wererecoveredbothintheglucosylatedandinthe
nonglucosylated form. As the two fragments b3 and b4
Ó FEBS 2004 Expression of yeast glycogenin (Glg2p) in E. coli (Eur. J. Biochem. 271) 3983
contain Tyr230 but not Tyr232 it is obvious that in G-P
2b
the Tyr230 is a glucosyl acceptor.
In summary, the MS/MS analysis of G -P
2a
and G -P
2b
clearly shows that following production in E. coli,Glg2p
possesses two functional glucosylation sites, Tyr230 and
Tyr232. Simultaneous glucosylation of the two vicinal
tyrosine residues occurs. This conclusion was reached by
MS/MS analysis o f the sin gly charged g lucopeptide G
2
-P
2b
(Fig. 3B). The Q-TOF MS/MS spectrum is shown in Fig. 5
and the fragments observed are listed in Table 2. The

Therefore, we chose the following growth and induction
protocol: after growth in tryptone–yeast extract m edium, the
production of G lg2p was induced by IPTG under other-
wise unchanged conditions. Ninety minutes later cells were
transferred to modified M9 minimal medium. At intervals,
aliquots of the suspension were withdrawn and the cellular
glycogen content was monitored. As a control, E. coli cells
containing the plasmid without the insert were kept under
exactly the same conditions. A s revealed by Western blotting
using an a nti-(His)
5
IgG
9
, Glg2p was detectable during the
entire period of glycogen accumulation (Fig. 6A).
Bacterial glycogen was determined by either of two
methods (see Materials and methods). Procedure A does
not require homogenization of the cells but presumably
results in a partial hydrolysis of the polyglucan. P rocedure B
yields an ess entially unmodified polysaccharide fraction as
revealed by control experiments using a commercial glyco-
gen p reparation. By applying both m ethods we co nsistently
observed t hat throughout the culture in the modified M 9
minimal medium the glycogenin producing E. coli cells did
accumulate approximately 30% less glycogen than the
Table 1. List of the b-type fragments of the Glg2p-derived mono-glu-
cosylated peptides G-P
2a
and G-P
2b

2b
(Fig. 3) are shown in F ig. 4 A and B, respectively. Both a - and b-ty pe fragments c ontain the N -termin us of th e peptide, th ey differ
by one CO group, i.e. a mass of 28 Da. y-Type fragments are C-terminal [28]. Probably due to a loss of NH
3
from asparagine, many fragments show
a satellite peak at )17 Da. F ragments bearing a glucosyl moie ty are marked w ith an a sterisk (*). A su mmary of the g lucosylation state of the
observed b-type fragments i s given in Table 1.
Ó FEBS 2004 Expression of yeast glycogenin (Glg2p) in E. coli (Eur. J. Biochem. 271) 3985
control cells. In F ig. 6B, the cellular glycogen content, as
determined by procedure A, was followed o ver 20 h. At t he
end of this period of time, the protein-based glycogen level
of the G lg2p expressing bacterial cells was 70.6 ± 7.2% of
that of the c ontrol cells (average of five i ndependent
experiments). The glycogenin producing E. coli cells did
not differ from the control with r espect to growth rate and
the content of buffer soluble proteins (data not shown).
The size distribution of the bacterial glycogen formed
either in the presence or t he absence o f the eukaryotic
glycogenin was determined. As revealed by Western blotting
experiments performed with buffer soluble proteins, the
recombinant GlG2 gene was expressed throughout the e ntire
period of glycogen accumulation (Fig. 6A). E. coli c ells that
had b een transformed with t he plasmid lacking the GlG2
gene were cultivated a nd harvested using precisely the same
conditions. From all six cell s amples glycogen was prepared
and analyzed by FFF-MALLS-RI. The molecular mass
distribution of the glycogen averages from 4 · 10
7
to
1.5 · 10

-P
2b
obtained by Q -TOF M S/MS ana lysis. Frag-
ments o bse rved in the spectrum (Fig. 5) are printed in bold lette rs. For
nomenclature of the fragm ent ions s e e [28].
G
2
-P
2b
b ions b ions + 162 b ions + 324 Sequence
98.1 260.1 422.1 P
212.1 374.1 536.1 PN
375.2 537.2 699.1 PNY
432.2 594.2 756.1 PNYG
595.3 757.3 919.2 PNYGY
723.3 885.3 1047.3 PNYGYQ
810.3 972.3 1134.3 PNYGYQS
897.4 1059.4 1221.3 PNYGYQSS
994.4 1156.4 1318.4 PNYGYQSSP
1065.5 1227.5 1389.4 PNYGYQSSPA
1148.5 1310.5 1472.5 PNYGYQSSPAX
3986 T. Albrecht et al. (Eur. J. Biochem. 271) Ó FEBS 2004
It is therefore reasonable to assume that the vast majority
of the compounds resolved by HPAEC (Fig. 7) are
homoglucans.
Discussion
In this communication, we have studied glucosylation of
one o f the two yeast glycogenins, Glg2p, under both in vivo
and in vitro conditions. Following production in E. coli,
purification and trypsin treatment, a Glg2p-derived pep tide

cleavage and a llowed the identificat ion of the glucosylation
sites (Fig. 3). Following heterologous expression in E. coli ,
we observed t his peptide in a nonglucosylated form only
following tre atment with both a-amylase and amyloglucos-
idase. Thus, it seems t hat i n E. coli the eukaryotic glyco-
genin is almost quantitatively glucosylated. As deduced
from Fig. 2, the minimum number of glucosyl residues
attached to Glg2p is four.
By combining protein backbone cleavage, enzymatic
hydrolysis of glycosidic bonds and M S/MS analysis we
provide direct evidence that both Tyr230 and Tyr232 a ct as
glucosylation sites of Glg2p. Discrimination between the
two glucosylated tyrosine residues was achieved by taking
advantage of a se lectivity of the a-amylase. When P
2
was
reacted with a-amylase, glucosyl residues linked to Tyr232
were removed with the exception of the glucose that is
covalently bound to the amino acid residue. I n contrast, the
glucan chain linked to Tyr230 was incompletely hydrolyzed
even after prolonged incubation. It is likely that a-amylase
acts effectively on the glucans bound to Tyr232 only if
Tyr230 is not glucosylated. Whilst the reason for this
selective a-amylase action is unknown, it is useful for the
generation of glucosylated peptides that are accessible to
Fig. 7. Maltodextrin pattern of Glg2p-producing E. coli cells. Bacte rial
cells were grown for 2 0 h in modified M9 medium. As a c ontrol, E. coli
cells transformed with a plasmid lacking the GlG2 gene were grown
simultaneously. Following the extraction in 80% (v/v) ethanol, the
deproteinized extracts were analy zed by H PAEC-PAD. As a standard,

2
glucopeptides. T herefore,
approximately 5–10% of the protein molecules bear glucose
on Tyr232 only. The p eak containing the doubly glucosyl-
ated peptide P
2
(Fig. 3B) represents approximately 3 0% of
all P
2
-related peaks. Henc e, in at least 30% of the Glg2p
molecules t he two sites are occupied. H owever, this value is
probably largely underestimated. In 40% of the molecules
the gluco se was totally removed b y t he amyloglucosidase
treatment ( fraction 18) a nd therefore their initial g lucosy-
lation state is unknown. The monoglucosylated peptide
eluting i n fraction 15 accoun ts for a pproximately 20%. T he
corresponding MS/MS spectrum revealed that the glucose is
mainly attached to T yr230, however, a deglucosylation of
Tyr232 by the action of the amyloglucosidase cannot be
excluded. In summary: a pproximately 10% of the Glg2p
protein is glucosylated o n T yr232 only, 30% of the protein
is glucosylated on both t yrosine r esidues a nd the remaining
60% are glucosylated either on both residues or o n Tyr230
only.
As revealed by Q-TOF MS/MS a major proportion of
the Glg2p molecules is simultaneously g lucosylated at both
Tyr230 and Tyr232 (Fig. 5 and Table 2). This feature,
which has not been described for other glycogenins, is
remarkable as it implies that two glucan chains can be
synthesized simultaneously by a single g lycogenin molecule.

pyrophosphorylase activity form a carbohydrate-free gly-
cogenin [30]. Because in our study essentially all the Glg2p
extracted from t he bacterial cells was f ound to be glucos-
ylated, the autoglucosylation of the glycogenin is unlikely to
be limited by the cellular levels of UDPglucose. In E. coli,
UDPglucose plays a central role in various pathways, s uch
as the galactose or trehalose metabolism, and also in t he
biosynthesis of membrane-derived oligosaccharides [31,32].
Thus, it seems that the eukaryotic initiator of glycogen
biosynthesis, although functional, is not compatible with the
prokaryotic path of glycogen formation. This conclusion is
supported by the fact that the size distribution of the
glycogen molecules is essentially unaffected by the produc-
tion of the eukaryotic glycogenin (Fig. 6). The dual function
of the g lycogen synthase recently proposed for Agrobacte-
rium tumefaciens [18] is consistent with this assumption.
Furthermore, it concurs with some enzymatic measure-
ments performed with glycogen accumulating E. coli cells.
The G lg2p-producing bacterial cells d id not differ notice-
ably from the control in v arious glycogen-related enzymes,
such as phosphorylase, soluble glycogen synthase or e ndo-
amylase (data not shown). Because of t he lower glycogen
content and the unchanged size distribution, the f requency
of an effective initiation of glycogen b iosynthesis appears to
be even lower in the presence of the eukaryotic glycogenin.
It should, however, be noted that the turnover of the
bacterial glycogen has not yet been analyzed.
In Glg2p-producing E. coli cells, the pattern of extract-
able maltodextrins is significantly altered (Fig. 7). As the
most prominent change, the level of a relatively small

fractal structure of glycogen : a clever solution to optimize cell
metabolism. Biophys. J. 77, 132 7–1332.
3988 T. Albrecht et al. (Eur. J. Biochem. 271) Ó FEBS 2004
3. Ball, S.G. & Morell, M.K. (2003) From bacterial g lycogen to
starch: understanding the biogenesis o f the plant starch granule.
Annu. R ev. Plant Biol. 54, 207– 233.
4. Manchester, J., Sk urat, A .V., Roach , P.J., Hauschka, S.D. &
Lawrence, J.C. Jr (1996) Increased glycogen accumulation in
transgenic mice overexpressing glycogen synthase in skeletal
muscle. Proc.NatlAcad.Sci.USA93, 10707–10711.
5. Pederson, B.A., Csitkovits, A.G., Simon, R ., Schroeder, J.M.,
Wang, W ., Skurat, A .V. & Roach, P.J. (2003) Ove rexpression of
glycogen synthase in mouse muscle results in less branched gly-
cogen. Biochem. Biophys. Res. C omm. 305, 826–830.
6. Smythe, C. & Cohen, P. (199 1) The discovery of g lycogenin and
the priming mechanism f o r glycogen biogenesis. Eur. J. Biochem.
200, 625–631.
7. Lomako, J., Lom a ko, W.M. & W helan, W.J. (2004) G lyc ogenin:
the primer for mammalian and yeast glycogen synthesis. Biochim.
Biophys. A cta 1673, 45–55.
8. Lin, A ., Mu, J., Yang, J. & Roach, P.J. ( 1999) Se lf-glucosylation
of glycogenin, t he initiator of glycogen b iosynthesis, involved a n
inter-subunit reaction. Arch. B iochem. Biophys. 36 3, 163–170.
9. Gibbons,B.J.,Roach,P.J.&Hurley,T.D.(2002)Crystalstructure
of the autocatalytic initiator of glycogen biosynthesis, glycogenin.
Mol. Biol. 319 , 293–298.
10. Mu, J., Cheng, C. & Roach, P.J. ( 1996) Initiation of glycogen
synthesis in yeast: r equirement of multiple tyrosine residues for
function of the self-glucosylating Glg p roteins in vivo. J. Biol.
Chem. 271 , 26554–26554.

cerevisiae. Mol. Cell Biol. 15, 6632–6640.
20. Cao, Y., Mahrenholz, A.M., DePaoli-Roach, A.A. & Roach, P.J.
(1993) Characterization of rabbit skeletal muscle glycogenin.
Tyrosine 19 4 is essential for function. J. Biol. C hem. 268, 1 4687–
14693.
21. Towbin, H., Staehlin, T. & Gordon, J . ( 1979) Electrophoretic
transfer of proteins from polyacrylamide gel to nitrocellulose
sheets: procedure and some app lications. P roc. N atl Acad . Sci.
USA 76, 4350–4354.
22. Steup, M. & Latzko, E. (1979) I ntracellular localization of phos-
phorylases in spinach and pea leaves. Planta 145, 69–75.
23. Fox, J., Kawaguchi, K., Greenberg, E. & Preiss, J. (1976) Bio-
synthesis of b acterial glycogen: p urification and properties of the
Escherichia coli B ADPglucose: 1,4-a-
D
-glucan 4-a-glucosyl-
transferase. Biochem. 15 , 849–857.
24. Lacks, S.A. & Springhorn, S.S. (1980) Renaturation of enzymes
after polya crylamide gel electrophoresis in the presence o f sodium
dodecyl sulfate. J. Biol. Chem. 255, 7 467–7473.
25. Haebel, S. & Kehr, J. (2001) MALDI-TOF MS peptide mass
fingerprints and PSD: a t ool for the identification and a nalysis of
phloem proteins from Cucurbita m axima separated by 2D P AGE.
Planta 213, 586–593.
26.Preiss,J.,Greenberg,E.&Sabraw,A.(1975)Biosynthesisof
bacterial glycogen: kkinetic studies of a glucose-1-phosphate
adenylyltransferase (EC 2.7.7.27) from a glycogen-deficient
mutant of E scherichia coli B. J. Biol. Chem. 250, 7631–7638.
27. Wyatt, P.J. (1993) Light scattering and the absolute characteri-
zation of macromolecules. Anal. Chim. Acta 272, 1–40.