Many fructosamine 3-kinase homologues in bacteria are
ribulosamine
⁄
erythrulosamine 3-kinases potentially
involved in protein deglycation
Rita Gemayel, Juliette Fortpied, Rim Rzem, Didier Vertommen, Maria Veiga-da-Cunha and
Emile Van Schaftingen
Universite
´
Catholique de Louvain, de Duve Institute, Brussels, Belgium
Fructosamine 3-kinase (FN3K) is a recently identified
enzyme that phosphorylates the Amadori products
fructosamines, leading to their destabilization and
removal from proteins [1–3]. FN3K is therefore respon-
sible for a new protein-repair mechanism. A related
mammalian enzyme (FN3K-related protein; FN3K-RP)
sharing 65% sequence identity with FN3K does not
phosphorylate fructosamines, but does phosphorylate
other ketoamines, mainly ribulosamines and erythrulos-
amines [4–6], as does the plant homologue of FN3K [6].
Fructosamines arise through a spontaneous reaction
of glucose with amines and their formation in vivo is
Keywords
deglycation; erythrose 4-phosphate;
fructosamine; glycation; ribose 5-phosphate
Correspondence
E. Van Schaftingen, UCL 7539, Avenue
Hippocrate 75, B-1200 Brussels, Belgium
Fax: +32 27 647598
Tel: +32 27 647564
E-mail:

Abbreviations
DEAE, diethylaminoethyl; FN3K, fructosamine 3-kinase; FN3K-RP, FN3K-related protein; LMW-PTP, low-molecular-weight protein-tyrosine-
phosphatase; SP, sulfopropyl.
4360 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
well documented. By contrast, the presence of
ribulosamines in cells has not been demonstrated. We
have previously speculated that they may form through
a reaction of amines with ribose 5-phosphate, a potent
glycating agent. The resulting ribulosamine 5-phos-
phates, however, are not substrates for FN3K-RP and
they therefore need to be dephosphorylated by a
phosphatase to become a substrate of FN3K-RP
(Scheme 1). We recently purified a ribulosamine
5-phosphatase from human erythrocytes, a cell type in
which FN3K-RP is very active, and we identified this
enzyme as low-molecular-weight protein-tyrosine-phos-
phatase A (LMW-PTP-A) [7].
As homologues of FN3K are also found in bacteria
[1], where genes encoding functionally related proteins
are often arranged in operons, we proceeded to ana-
lyze bacterial genomes. In several instances, we found
that an FN3K homologue was associated in an operon
with a putative LMW-PTP. These findings led us to
express and characterize five bacterial FN3K homo-
logues and three LMW-PTP homologues, and to study
their substrate specificity.
Results
Search of FN3K homologues in databases
To identify the bacterial genomes comprising an FN3K
homologue, we performed tBLASTn searches in the

two other trypanosomatids, Trypanosoma brucei and
Leishmania major.
The sequences were aligned by ClustalX and a
neighbour-joining tree was constructed (Fig. 1). Bacte-
rial sequences formed several clusters corresponding
mostly to known groups of bacteria [e.g. Actinobacte-
ria, Cyanobacteria (two clusters) and bacteria of the
gamma subdivision (Enterobacteriales, Pasteurellaceae,
Vibrionaceae)]. Eukaryotic sequences formed one sin-
gle cluster, with the exception of the FN3K homo-
logue of T. cruzi, which clustered with bacterial
sequences.
Genome context
We also examined the genome context of the bacterial
FN3K homologues, as this could point to functionally
CH
2
O
C
HCOH
CH
2
OH
HCOH
NH
Ribulosamine
HC
HCOH
HCOH
CH

O
Ribulosamine-5-P
CH
OC
HCOH
CH
2
OH
HC
NH
2
O
O
-
P
O
-
O
Ribulosamine-3-P
HC
O
C
HCOH
CH
2
OH
HCH
O
NH
2

the origin of the substrate(s) or on the fate of the
product(s) of the FN3K homologues. Except for evolu-
tionarily related bacteria, this genome context is extre-
mely variable. However, the gene encoding the FN3K
homologue is immediately preceded by a putative
LMW-PTP in 11 genomes from phylogenetically dis-
tant bacteria: Cytophaga hutchinsonii, Thermus thermo-
philus (Fig. 2), Acidothermus cellulolyticus, Fulvimarina
pelagi, Gloeobacter violaceus, Microscilla marina,
0.1
Yersinia pestis
Photorhabdus luminescens
Erwinia carotovora
Escherichia coli
Salmonella enterica
Pasteurella multocida
Haemophilus somnus
Mannheimia succiniciproducens
Vibrio parahaemolyticus
Vibrio vulnificus
Vibrio cholerae
Vibrio fischeri
Photobacterium profundum
Pseudoalteromonas haloplanktis
Colwellia psychrerythraea
Anabaena variabilis
Nostoc punctiforme
Thermosynechococcus elongatus
Crocosphaera watsonii
Synechocystis sp.

Neurospora crassa
Arabidopsis thaliana
Giardia lamblia
Thermobifida fusca
Nocardia farcinica
Corynebacterium efficiens
Corynebacterium glutamicum
Mycobacterium avium
Propionibacterium acnes
Nocardioides sp.
Bifidobacterium breve
Bifidobacterium longum
Thermus thermophilus
Chromohalobacter salexigens
Zymomonas mobilis
Rhodobacterales bacterium
Rubrobacter xylanophilus
Rhodospirillum rubrum
Haloarcula marismortui
+
+
*
*
*
*
*
+
*
*
+

Cyanobacteria
Enterobacteriales
Pasteurellaceae
Vibrionaceae
Cyanobacteria
Lactobacillales
Eukaryote
Eukaryotes
Actinobacteria
Archaea
yes
Ribulosamine
3-kinase
activity
no
yes
yes
yes
yes
yes
yes
yes
yes
yes
Fig. 1. Fructosamine 3-kinase (FN3K) homologues: neighbour-joining tree, activity and association with putative phosphatases in various bac-
terial genomes. The Haloarcula marismortui sequence was used as an outgroup. Symbols at the nodes represent the support for each node
as obtained by 1000 bootstrap samplings: (*), > 95%; (+), 80–95%; (·), 50–80%. Nodes with no symbol were found in < 50% of the boot-
strap samplings. The branch lengths are proportional to the number of substitutions per site. The horizontal bar represents 0.1 substitutions
per site. The first column indicates the proteins that have been shown to phosphorylate ribulosamines in this work (framed) or in previous
work. The last two columns indicate the presence of homologues of low-molecular-weight protein-tyrosine-phosphatase (LMW-PTP) or the

the other orientation (YniB, called YfeE in Yersinia
pestis, or homologues) in E. coli (Fig. 2), Erwinia
carotovora, Salmonella enterica and various Shigella
and Yersinia species (data not shown). The phospha-
tase YniC is, however, absent from the genomes of
most Vibrionaceae (which comprise an FN3K homo-
logue) (Fig. 1), but present in other bacteria of the
gamma subdivision (various Shewanella species,
Marinomonas sp.) that do not comprise an FN3K
homologue. It is therefore likely that the phosphatase
YniC, contrary to LMW-PTP, is not functionally
related to FN3K homologues.
Sequence alignments
Figure 3 shows an alignment of the five bacterial pro-
teins that have been biochemically characterized in the
present work with those of eukaryotic FN3K or
FN3K-RP that have been previously studied (human
FN3K and FN3K-RP; the FN3K homologue of Ara-
bidopsis thaliana) [1,4,6,10]. All sequences share several
conserved motifs. The most striking one is the nucleo-
tide-binding motif (LHGDLWxGN; residues 214–222
in the human FN3K sequence), which is similar to that
found in aminoglycoside kinases (LHxDLHxxN). Ver-
tebrate FN3Ks and FN3K-RPs contain a stretch of
about 20 residues (residues 118–140 in human FN3K)
that is absent from the prokaryotic sequences and
from the eukaryotic sequences of plants, fungi and
protists. In relation with the lack of activity of the
E. coli FN3K homologue (see below), it is interesting
to point out that its sequence differs from the others

thermophilus
FN3KLMW-PTP
Histidine
kinase
IndA
protein
Hypothetical
protein
GTP
binding
protein
Cytophaga
hutchinsonii
FN3KLMW-PTP
Fe uptake
regulator
Alkyl
hydroperoxide
reductase
Hypothetical
proteins
Fig. 2. Genomic environment of some bac-
terial fructosamine 3-kinase (FN3K) homo-
logues.The genomic arrangements are
shown for the FN3K homologues of Ther-
mus thermophilus, Cytophaga hutchinsonii
and Escherichia coli. The most significant
finding was the association of the FN3K
homologue with a low-molecular-weight
protein-tyrosine-phosphatase (LMW-PTP)

0.1–5 mm).
Fig. 3. Alignment of human fructosamine
3-kinase (FN3K) and fructosamine 3-kinase-
related protein (FN3K-RP) with the bacterial
homologues investigated in the present
study. The sequences were aligned using
C
LUSTALX. Conserved residues are high-
lighted and the residues that differ in the
Escherichia coli FN3K homologue sequence
are underlined. The abbreviations used are:
FN3K (human FN3K), FN3KRP (human
FN3K-RP), ARATH (FN3K homologue from
Arabidopsis thaliana), ECOLI (Escherichia
coli), ENTFAE (Enterococcus faecium),
LACTPL (Lactobacillus plantarum), STAPH
(Staphylococcus aureus) and THERM (Ther-
mus thermophilus).
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4364 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
Action of bacterial FN3K homologues on
protein-bound ketoamines
We also tested the ability of the bacterial FN3K
homologues to phosphorylate protein-bound ribulosam-
ines. Two proteins, hen egg lysozyme and E. coli thio-
redoxin A, were glycated with ribose and used as
substrates (Fig. 4). All four active bacterial FN3K
homologues and mouse FN3K catalysed the phosphor-
ylation of protein-bound ribulosamines, although their
relative activity was dependent on the substrate used.

Table 1. Kinetic properties of the bacterial fructosamine 3-kinase (FN3K) homologues. The results are the means of two or three determina-
tions. In the latter case, the SEM value is given. V
max
values are expressed as nmol phosphorylated product formed per min and per mg of
protein. E. faecium, Enterococcus faecium; L. plantarum, Lactobacillus plantarum; ND, not determined; S. aureus, Staphylococcus aureus;
T. thermophilus, Thermus thermophilus.
Substrate
L. plantarum E. faecium S. aureus T. thermophilus
K
m
(lM)
V
max
(nmolÆmin
)1
Æmg
)1
)
K
m
(lM)
V
max
(nmolÆmin
)1
Æmg
)1
)
K
m

44 ± 6 220 ± 25 > 500 7 ± 1
a
Ribulosamine-
Thioredoxin A
> 500 59 ± 6
a
580 ± 40 52 ± 5 460 250 270 130
Erythrulosamine-
lysozyme
> 500 3.8 ± 0.1
a
> 500 80 ± 6
a
42 ± 4 78 ± 2 > 500 4.2 ± 0.2
a
a
Activity at 100 lM protein-bound ribulosamine or erythrulosamine.
0 5 10 15 20 25
0.00
0.05
0.10
0.15
0.20
A
B
FN3K
S. aureus
E. faecium
M. musculus
T. thermophilu

50 lgÆmL
)1
of each FN3K homologue. Incorporated phosphate was
measured at different time-points. The results are the means of
three independent measurements ± SEM.
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4365
phosphorylated by the enzymes from E. faecium and
L. plantarum, but they were slowly phosphorylated by
the enzymes from S. aureus and T. thermophilus,at
rates corresponding to 0.4 and 2%, respectively, of the
activity observed with lysozyme-bound ribulosamines.
None of these enzymes catalysed the phosphorylation
of protein-bound ribulosamine 5-phosphates (data not
shown). The E. coli FN3K homologue was also inac-
tive on all macromolecular substrates tested, which
included lysozyme-bound d- and l-ribulosamines,
d-ribulosamine 5-phosphates, fructosamines and
erythrulosamines (data not shown).
The product of the phosphorylation of lysozyme-
bound ribulosamines by the FN3K homologues from
S. aureus and T. thermophilus broke down with a half-
life of 26–28 min at 37 °C and neutral pH (data not
shown), as previously observed with the product of
human FN3K-RP [5] and the plant FN3K homologue
[6]. These results further indicated that bacterial FN3K
homologues also phosphorylated carbon 3 of the sugar
moiety of ribulosamines.
Substrate specificity of the LMW-PTP
homologues

The activity of LMW-PTPs on protein substrates was
tested through the release of
32
P from radiolabelled sub-
strates. As shown in Fig. 5, T. thermophilus LMW-PTP
acted about 10-fold faster on protein tyrosine-phos-
phates than on protein ribulosamine 5-phosphates,
whereas S. aureus PtpA acted preferentially on the latter
substrate. S. aureus PtpB was also poorly active on pro-
tein substrates. As illustrated for T. thermophilus LMW-
PTP, dephosphorylation of lysozyme glycated with
ribose 5-phosphate by this phosphatase led to the for-
mation of a substrate for the S. aureus FN3K homo-
logue (Fig. 6). The resulting phosphorylation product
was unstable and broke down, at 37 °C, with a half-life
similar to that of ribulosamine 3-phosphates (data not
shown). Similarly, incubation of lysozyme-bound
erythrulosamine 4-phosphates with T. thermophilus
LMW-PTP or S. aureus PtpA led to the formation of a
substrate for the S. aureus FN3K homologue (Fig. 7).
Discussion
Most bacterial FN3K homologues are
ribulosamine
⁄
erythrulosamine 3-kinases
Four of the five bacterial FN3K homologues that we
studied are ribulosamine ⁄ erythrulosamine 3-kinases.
This property is shared by mammalian and avian
FN3Ks and FN3K-RPs, as well as by the single
FN3K homologue present in fish and plants. This

4366 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
fructosamines, which is restricted to mammalian and
avian FN3Ks, was acquired late in evolution following
a gene duplication event that took place in the lineage
leading to mammals and birds [10]. It is not known at
present if the E. coli homologue is an inactive protein
or if it has acquired a distinct substrate specificity.
The physiological substrate of the active bacterial
FN3K homologues is presently not known, but its
structure is presumably close to that of a ribulosamine
or an erythrulosamine. The observations that no phos-
phorylation is observed with ribulose, with the reduced
forms of ribuloselysine and with xyluloselysine (C3 epi-
mer of ribuloselysine), stress the importance of the
presence of an amino group on C1, a keto function on
C2 and a hydroxyl group with a D configuration on
C3. In addition, as initially observed with FN3K [12],
ketoamine derivatives bound to the alpha amino group
of amino acids are poor substrates, whereas ketoam-
ines bound to the epsilon amino group of lysine or
cadaverine are excellent substrates.
Bacterial FN3K homologues are more than 10-fold
more active on LMW ketoamines than on protein-
bound ketoamines, which suggests that their physiolog-
ical substrates are LMW compounds. However, their
absolute activity on protein substrates is higher than
that of mammalian FN3K or FN3K-RP on similar
substrates. For instance, the V
max
of fructosamine

20
30
40
50
60
70
80
90
100
A
B
Enzyme Substrate
Enzyme Substrate
Lysozyme-RN5P
MBP-TyrP
50 µg.mL
-1
5 µg.mL
-1
Lysozyme-RN5P
MBP-TyrP
-
-
Time (min)
Release of
32
P (%)Release of
32
P (%)
0 5 10 15 20 25 30 35

[
32
P]tyrosine phosphates (MBP-TyrP) and
lysozyme-bound [
32
P]ribulosamine 5-phos-
phates (Lysozyme-RN5P), both tested at
2 l
M protein-bound [
32
P]phosphate. The
concentration of each homologue used is
shown on the graph, and conditions where
no LMW-PTP was added are shown in open
symbols. The radioactivity, corresponding to
32
P inorganic phosphate, released after
trichloroacetic acid precipitation of proteins
was measured at different time-points. The
results are the means of three independent
measurements ± SEM.
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4367
0 20 40 60 80 100
0.0
0.1
0.2
0.3
0.4
0.5

20 m
M
Time(min)
Incorporated Phosphate
(mol P/mol lysozyme)
Fig. 6. Dephosphorylation of protein
ribulosamine 5-phosphates by the Thermus
thermophilus low-molecular-weight protein-
tyrosine-phosphatase (LMW-PTP) homo-
logue and rephosphorylation by a bacterial
fructosamine 3-kinase (FN3K) homologue.
(A) Lysozyme (5 mgÆmL
)1
) glycated with
20 m
M ribose 5-phosphate was incubated
with 60 lgÆmL
)1
of T. thermophilus LMW-
PTP (closed circles), or without LMW-PTP
(open circles). Unglycated lysozyme was
also incubated with T. thermophilus LMW-
PTP (closed diamonds). The liberated inor-
ganic phosphate was measured at different
time-points. (B) The resulting product of
dephosphorylation was then incubated with
[
32
P]ATP[cP] and 50 lgÆmL
)1

S. aureus PtpA
T. thermophilus
T. thermophilus
10 m
M-
Time(min)
0 5 10 15 20 25
Time(min)
Liberated Inorganic Phosphate
(mol P/mol lysozyme)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
10 m
M
5 mM
10 mM
5 mM
-
5 m
M
10 mM
Incorporated Phosphate
(mol P/mol lysozyme)
Fig. 7. Dephosphorylation of protein
erythrulosamine 4-phosphates by bacterial

Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4368 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
homologues is an exogenous, toxic compound, which,
like aminoglycosides and macrolides, would have to be
inactivated by phosphorylation [15]. To the best of our
knowledge, no known antibiotic is a ketoamine deriva-
tive, but hypothetical ketoamine antibiotics may have
been missed in the screenings for antibacterial com-
pounds because of their instability.
An endogenous origin for the substrate(s) of FN3K
homologues has therefore to be considered. However,
except for the association with an LMW-PTP (see
below), the context in which FN3K homologues are
found in bacterial genomes is not suggestive of any
pathway leading to the formation of a ketoamine.
The association with a phosphatase suggests
that ribulosamines are formed from ribulosamine
5-phosphates
We have previously speculated that ketoamines may
arise through glycation of amino compounds by ribose
5-phosphate or erythrose 4-phosphate, which are
potent glycating agents that occur physiologically in
all cell types. The ribulosamine 5-phosphates and ery-
thrulosamine 4-phosphates that are so formed are not
substrates for mammalian or bacterial FN3K homo-
logues. A phosphatase is therefore needed to remove
the terminal phosphate before these kinases can act
(Scheme 1). We recently purified a ribulosamine
5-phosphatase from human erythrocytes and identified
it as LMW-PTP-A [7].

activity, which indicates that its function is not to
antagonize the activity of the FN3K homologues, but,
on the contrary, to complement it. The most probable
hypothesis is therefore that the phosphatase functions
physiologically as a ribulosamine 5-phosphate ⁄ erythru-
losamine 4-phosphate phosphatase, allowing the for-
mation of substrates for the FN3K homologues.
Potential sources of ribulosamine 5-phosphates
and erythrulosamine 4-phosphates
Ribose 5-phosphate and erythrose 4-phosphate are
potent glycating agents, reacting with proteins about
80- and 500-fold more rapidly than glucose, respec-
tively [7,16; R Gemayel, unpublished results]. The
information on the concentration of these phosphate
esters in bacteria is scant. The xylulose 5-phosphate
content of Oenococcus oeni (previously known as Leu-
conostoc oenos) amounts to 0.1–0.33 lmolÆg
)1
dry
weight, corresponding to concentrations of about
0.033–0.1 mm [17]. The concentration of ribose 5-phos-
phate is probably of the same order of magnitude,
indicating that in this bacterium, the glycating power
of ribose 5-phosphate is comparable to that of 2 mm
glucose. Erythrose 4-phosphate is likely to accumulate
in bacteria under some conditions, for example in the
absence of O
2
in O. oeni. This bacterium forms sub-
stantial amounts of erythritol under this condition,

different roles in different organisms.
An objection that can be raised against the role of
FN3K homologues in deglycation is that an LMW-
PTP is not found in several of the genomes that con-
tain them. However, this may be because the enzymatic
activity is carried out by another type of phosphatase.
Another objection is that protein repair does not make
sense for such rapidly dividing organisms as bacteria.
However, two other protein-repair mechanisms – those
catalysed by protein-l-isoaspartate-O-methyltransferase
[18] and methionyl sulphoxide reductases [19] – operate
in prokaryotes. Finally, as ribose 5-phosphate and
erythrose 4-phosphate are ubiquitous, why are FN3K
homologues only found in some organisms? This ques-
tion may receive multiple answers, such as (a) other
proteins – not necessarily kinases – could carry out
deglycation in organisms devoid of FN3K homologues,
(b) the levels of ribose 5-phosphate and erythrose
4-phosphate could be lower in these bacteria than in
those containing an FN3K homologue and (c) there
could be bacteria in which no process is sensitive to
glycation.
Whatever the real physiological function of bacterial
FN3K homologues, the demonstration that they act as
ribulosamine ⁄ erythrulosamine 3-kinases will be helpful
in further studies on protein deglycation by offering
tools that can be easily produced (unlike mammalian
FN3K-RPs) [4] in substantial amounts to study this
process. Thus, these enzymes could be useful to detect
protein-bound ribulosamines and erythrulosamines by

and the degree of glycation was 2.8 mol of ribulosamine
per mol of lysozyme. Lysozyme glycated with l-arabinose
(which gives rise to l-ribulosamines) was prepared as
described for ribose. Lysozyme glycated with glucose or
erythrose was prepared as previously described [6]. E. coli
thioredoxin A glycated with ribose was prepared as follows:
20 mgÆmL
)1
of purified recombinant thioredoxin A, pre-
pared as described previously for thioredoxin 2 [20], was
incubated at 60 °C for 16 h in a solution containing 20 mm
Mes, pH 6, 1 mm EGTA and 100 mm ribose. The sample
was then purified by gel filtration on a NAP-5 column
equilibrated with water. The degree of glycation was
1.1 mol of ribulosamine per mol of thioredoxin A.
Lysozyme glycated with ribose 5-phosphate was prepared
as described previously [7]. The same procedure was used
to prepare lysozyme glycated with erythrose 4-phosphate,
except that erythrose 4-phosphate was used at 5 and 10 mm
and the incubation temperature was 37 °C. Lysozyme gly-
cated with
32
P-ribose 5-phosphate and
32
P-protein tyrosine
phosphate were prepared as described previously [7,21],
except that the medium contained Tris instead of Hepes.
Synthesis of LMW substrates
Ribuloselysine, erythruloselysine, [
14

Ribuloselysine 5-phosphate was prepared as previously
described [7]. Erythruloselysine 4-phosphate was synthe-
sized as for ribuloselysine 5-phosphate, but with 50 mm
erythrose 4-phosphate and at 37 °C.
Cloning and expression of the FN3K and
LMW-PTP bacterial homologues
Expression vectors for the bacterial FN3K and LMW-PTP
homologues were constructed by PCR amplification of the
corresponding coding sequences using the appropriate
genomic DNA and primers (Table 3). The PCR products
were digested with the appropriate restriction enzymes
and inserted into pET-15b [which yields an N-terminal
poly(His)-tagged protein] or pET-3a vectors and the
sequences were verified. These expression vectors were used
to transform E. coli cells (Table 3) [23].
Expression of the recombinant proteins was carried out
as described previously [24] under the general conditions
(medium, time and temperature) indicated in Table 3. The
bacterial extracts were prepared as previously described [25]
with the exception that phenylmethanesulfonyl fluoride was
omitted from the extraction buffer.
Purification of the bacterial FN3K and LMW-PTP
homologues
All recombinant enzymes were purified from 50-mL extracts
prepared from 1-L cultures, unless stated otherwise. Protein
concentration was determined by the absorbance at
280 nm, taking into consideration the theoretical extinction
coefficient of each homologue, or by the Bradford method
[26] in the case of the T. thermophilus FN3K homologue.
All purified proteins were analysed by SDS ⁄ PAGE, supple-

umn was washed with 100 mL of buffer B and proteins were
eluted with a 150-mL NaCl gradient (0–0.75 m in buffer B).
Fractions containing the recombinant protein were pooled,
concentrated to 2 mL (in a 10-mL Amicon cell equipped with
a YM-10 membrane) and applied to a 100-cm
3
Sephacryl
S-200 column equilibrated with buffer B containing 100 mm
NaCl. The amount of purified protein was  18 mg.
T. thermophilus ‘FN3K’
The extract was heated at 80 °C for 5 min, left on ice for
10 min and then centrifuged for 30 min at 10 000 g. The
resulting supernatant was diluted twice in buffer B and
purified on DEAE-Sepharose as for the E. faecium FN3K
homologue. The amount of purified protein was  4 mg.
E. coli ‘FN3K’
The bacterial extract (25 mL) from a 500-mL culture was
brought to 20% (w ⁄ v) poly(ethylene glycol) 6000, and cen-
trifuged for 30 min at 10 000 g. The resulting supernatant
was diluted three-fold with buffer B and purified on
DEAE-Sepharose, as described for the E. faecium FN3K
homologue. The amount of purified protein obtained from
the 500-mL culture was 54 mg.
T. thermophilus LMW-PTP
The extract was heated at 80 °C for 5 min, left on ice for
10 min and then centrifuged for 30 min at 10 000 g. The
resulting supernatant was purified on DEAE-Sepharose,
as described for the E. faecium FN3K homologue, but
using buffer C (25 mm Tris pH 7.1, 1 mm dithiothreitol,
1 lgÆmL

Restriction
enzymes Vector
Escherichia coli
strain Medium Antibiotic Temp. Time
Forward (5¢)3¢)
Reverse (5¢)3¢)
FN3K homologues
Lactobacillus plantarum GTT
CATATGCACTTAACAAAAACTTGG NdeI pET15b BL21(DE3) LB
a
Ap
c
18 °C16h
GAG
GGATCCATTAATATTGCATGAGAATTC BamHI pLysS Cm
d
Enterococcus faecium AACATATGGATATCCAAACTGTTTTATC NdeI pET3a BL21(DE3) LB Ap 18 °C16h
GC
GGATCCCTTAAAAATTTTCTAGTAATTG BamHI
Staphylococcus aureus TT
CATATGAATGAACAATGGTTAGAG NdeI pET15b BL21(DE3) LB Ap 18 °C16h
C
GGATCCACTAACTTGTTGTACCTTGT BamHI
Thermus thermophilus GCAGCG
CATATGGATCCCCTAGCCCTGCTG NdeI pET3a BL21-CodonPlus LB Ap 37 °C6h
GGCAGC
AGATCTAAGAGGCGGAAATCGCCCTC BglII (DE3) Cm
Escherichia coli G
CATATGTGGCAGGCAATCAGTCGTC NdeI pET3a BL21(DE3)plysS M9 (AA)
b

cpm [
32
P]ATP[cP] and the indicated concentrations of
protein-bound ketoamines and FN3K homologue. The
incorporation of
32
P into proteins was measured as
described previously [1]. Phosphorylation of ribuloselysine
(using [
14
C]ribuloselysine) and of unlabelled compounds
(using [
32
P]ATP[cP]) was assayed as described previously
[6], in the same medium as described above. Phosphoryla-
tion of alpha-glycated amino acids was assayed spectropho-
tometrically by measuring the release of ADP through the
pyruvate kinase ⁄ lactate dehydrogenase coupled assay [13].
Phosphatase activities were assayed at 37 °C in a medium
containing 25 mm Tris, pH 7.1, 1 mm dithiothreitol, 1 mm
EGTA, 10 mm KCl (for T. thermophilus LMW-PTP) or
100 mm K acetate (for S. aureus LMW-PTPs). Dephos-
phorylation of [
32
P]protein-bound tyrosine phosphates and
ribulosamine 5-phosphates was assayed as described previ-
ously [21]. Activities on LMW substrates were assayed
through the formation of inorganic phosphate [28] in a
medium supplemented with 50 lgÆmL
)1

gaps and with correction for multiple substitutions. For the
bootstrap analysis, 1000 samplings were carried out.
Acknowledgements
This work was supported by grants from the Juvenile
Diabetes Foundation International, the Interuniversity
Attraction Poles Program-Belgian Science Policy (Net-
works P6⁄ 05 and P6 ⁄ 28), and the Concerted Research
Action Program of the French Community of Belgium.
RG is fellow of the Fonds pour l’Encouragement a
`
la
Recherche dans l’Industrie et dans l’Agriculture.
MVDC is chercheur qualifie
´
of the Fonds National de
la Recherche Scientifique. L. plantarum genomic DNA
was kindly provided by P. Hols (Universite
´
Catholique
de Louvain), E. faecium genomic DNA by P. Charlier
(Universite
´
de Lie
`
ge), S. aureus genomic DNA
by J. Van Eldere (Katholieke Universiteit Leuven),
and T. thermophilus genomic DNA by J F. Collet
(Universite
´
Catholique de Louvain).

gen E (2007) Identification of protein-ribulosamine-5-
phosphatase as human low-molecular-weight
protein-tyrosine-phosphatase-A. Biochem J 406, 139–145.
8 Vincent C, Doublet P, Grangeasse C, Vaganay E, Cozzone
AJ & Duclos B (1999) Cells of Escherichia coli contain a
protein-tyrosine kinase, Wzc, and a phosphotyrosine-
protein phosphatase, Wzb. J Bacteriol 181, 3472–3477.
9 Kuznetsova E, Proudfoot M, Gonzalez CF, Brown G,
Omelchenko MV, Borozan I, Carmel L, Wolf YI, Mori
H, Savchenko AV et al. (2006) Genome-wide analysis of
substrate specificities of the Escherichia coli haloacid
dehalogenase-like phosphatase family. J Biol Chem 281,
36149–36161.
10 Delplanque J, Delpierre G, Opperdoes FR & Van
Schaftingen E (2004) Tissue distribution and evolution
of fructosamine 3-kinase and fructosamine 3-kinase-
related protein. J Biol Chem 279, 46606–46613.
11 Soulat D, Vaganay E, Duclos B, Genestier A, Etienne J
& Cozzone AJ (2002) Staphylococcus aureus contains
two low-molecular-mass phosphotyrosine protein phos-
phatases. J Bacteriol 184, 5194–5199.
12 Delpierre G, Vertommen D, Communi D, Rider MH
& Van Schaftingen E (2004) Identification of fructos-
amine residues deglycated by fructosamine-3-kinase in
human hemoglobin. J Biol Chem 279, 27613–27620.
13 Wiame E, Delpierre G, Collard F & Van Schaftingen E
(2002) Identification of a pathway for the utilization of
the Amadori product fructoselysine in Escherichia coli.
J Biol Chem 277, 42523–42529.
14 Wiame E, Lamosa P, Santos H & Van Schaftingen E

281, 18378–18385.
22 Park JT & Johnson MJ (1949) A submicrodetermina-
tion of glucose. J Biol Chem 181, 149–151.
23 Studier FW & Moffatt BA (1986) Use of bacteriophage
T7 RNA polymerase to direct selective high-level
expression of cloned genes. J Mol Biol 189, 113–130.
24 Veiga-da-Cunha M, Courtois S, Michel A, Gosselain E
& Van Schaftingen E (1996) Amino acid conservation
in animal glucokinases. Identification of residues
implicated in the interaction with the regulatory protein.
J Biol Chem 271, 6292–6297.
25 Wiame E & Van Schaftingen E (2004) Fructoselysine
3-epimerase, an enzyme involved in the metabolism of
the unusual Amadori compound psicoselysine in
Escherichia coli. Biochem J 378, 1047–1052.
26 Bradford MM (1976) A rapid and sensitive method for
the quantification of microgram quantities of protein
utilizing the principle of protein dye-binding. Anal
Biochem 72, 248–254.
27 Maliekal P, Vertommen D, Delpierre G & Van Schaf-
tingen E (2006) Identification of the sequence encoding
N-acetylneuraminate-9-phosphate phosphatase.
Glycobiology 16, 165–172.
28 Itaya K & Ui M (1966) A new micromethod for the
colorimetric determination of inorganic phosphate.
Clin Chim Acta 14, 361–366.
29 Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang
Z, Miller W & Lipman DJ (1997) Gapped BLAST and
PSI-BLAST: a new generation of protein database
search programs. Nucleic Acids Res 25, 3389–3402.