Báo cáo khoa học: Many fructosamine 3-kinase homologues in bacteria are ribulosamine⁄erythrulosamine 3-kinases potentially involved in protein deglycation docx

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:

(Received 11 April 2007, revised 15 June
2007, accepted 18 June 2007)
doi:10.1111/j.1742-4658.2007.05948.x
The purpose of this work was to identify the function of bacterial homo-
logues of fructosamine 3-kinase (FN3K), a mammalian enzyme responsible
for the removal of fructosamines from proteins. FN3K homologues were
identified in  200 (i.e.  27%) of the sequenced bacterial genomes. In 11
of these genomes, from phylogenetically distant bacteria, the FN3K homo-
logue was immediately preceded by a low-molecular-weight protein-tyro-
sine-phosphatase (LMW-PTP) homologue, which is therefore probably
functionally related to the FN3K homologue. Five bacterial FN3K homo-
logues (from Escherichia coli, Enterococcus faecium, Lactobacillus planta-
rum, Staphylococcus aureus and Thermus thermophilus) were overexpressed
in E. coli, purified and their kinetic properties investigated. Four were ribu-
losamine ⁄ erythrulosamine 3-kinases acting best on free lysine and cadaver-
ine derivatives, but not on ribulosamines bound to the alpha amino group
of amino acids. They also phosphorylated protein-bound ribulosamines or
erythrulosamines, but not protein-bound fructosamines, therefore having
properties similar to those of mammalian FN3K-related protein. The
E. coli FN3K homologue (YniA) was inactive on all tested substrates. The
LMW-PTP of T. thermophilus, which forms an operon with an FN3K
homologue, and an LMW-PTP of S. aureus (PtpA) were overexpressed in
E. coli, purified and shown to dephosphorylate not only protein tyrosine
phosphates, but protein ribulosamine 5-phosphates as well as free ribulose-
lysine 5-phosphate and erythruloselysine 4-phosphate. These LMW-PTPs
were devoid of ribulosamine 3-phosphatase activity. It is concluded that
most bacterial FN3K homologues are ribulosamine ⁄ erythrulosamine 3-kin-
ases. They may serve, in conjunction with a phosphatase, to deglycate
products of glycation formed from ribose 5-phosphate or erythrose 4-phos-
phate.

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

microbial genome database available at http://
www.ncbi.nlm.nih.gov. As of February 2007, 27%
(210 ⁄ 760) of all available genomes, and the same pro-
portion (124 ⁄ 453) of completely sequenced genomes,
contained an FN3K homologue. No more than one
homologue was identified per bacterial genome.
Remarkably, an FN3K homologue is present in all
Cyanobacteria, but only in some members of other
bacterial families (supplementary Table S1). For
instance, among Pasteurellaceae, Haemophilus somnus
and Actinobacillus succinogenes comprise an FN3K
homologue, but this is not the case for Haemophilus
influenzae and Act inoba cillus pleuropn eu moniae .AnFN3K
homologue was found in only 1 of the 38 sequenced
archaeal genomes, that of Haloarcula marismortui.
FN3K homologues were also identified in eukary-
otes. As previously described, two different homo-
logues, one closer to human FN3K and the other
closer to FN3K-RP, are present in mammals and
birds, whereas only one homologue is observed in fish
(and is closer to FN3K-RP). One single FN3K homo-
logue is present in Caenorhabditis elegans, Caenorhabd-
itis briggsae and Ciona intestinalis, and at least three
different homologues are present in Strongylocentrotus
purpuratus, but there are none in insects. Homologues
are also found in several fungi (e.g. Aspergilli, Neuro-
spora crassa, Magnaporthe grisea) although not in the
yeasts Saccharomyces cerevisiae and Schizosaccharomy-
ces pombe. Among protozoa, a homologue is found in
Giardia lamblia and Trypanosoma cruzi, but none in

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

2
HCOH
O
O
-
P
O
-
O
O
NH
2
Ribose-5-P
Protein
CH
2
O
C
HCOH
CH
2
HCOH
NH
2
O
O
-
P
O
-

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

4,5-dihydroxy-
1,2-pentanedione
+
FN3K
homologue
ATP
ADP
LMW-PTP
Pi
H
2
O
H
2
O
Pi
H
2
O
Scheme 1. Formation and repair of ribulosamines. Ribulosamines
presumably result from the reaction of amines with ribose 5-phos-
phate, followed by enzymatic dephosphorylation of ribulosamine
5-phosphates by a phosphatase. Ribulosamines are phosphorylated
by fructosamine 3-kinase (FN3K) homologues, which leads to their
destabilization and recovery of the unmodified amine. Erythrulosam-
ines presumably form in a similar manner from erythrose 4-phos-
phate (data not shown).
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4361
related proteins and therefore provide information on

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.

Trichodesmium erythraeum
Gloeobacter violaceus
Synechococcus elongatus
Nitrosomonas europaea
Azoarcus sp.
Thiobacillus denitrificans
Thiomicrospira crunogena
Synechococcus sp.
Prochlorococcus marinus str.
Prochlorococcus marinus
Prochlorococcus marinus subs.
Microbulbifer degradans
Staphylococcus aureus
Staphylococcus epidermidis
Lactobacillus casei
Oenococcus oeni
Lactobacillus plantarum
Leuconostoc mesenteroides
Trypanosoma cruzi
Enterococcus faecium
Cytophaga hutchinsonii
Salinibacter ruber
Gallus gallus FN3K-RP
Homo sapiens FN3K-RP
Danio rerio
Homo sapiens FN3K
Gallus gallus FN3K
Strongylocentrotus purpuratus
Caenorhabditis briggsae
Aspergillus fumigatus

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
+
+
*
*
*
*
*
+
*
*
+

+
*
*
*
*
*
*
+
+
Associated
LMW-PTP
(distance in bp)
yes (-19)
yes (0)
yes (17)
yes (12)
yes (-3)
yes (-37)
yes (66)
yes (-10)
Associated
YniC
(distance in bp)
yes (728)
yes (724)
yes (918)
yes (785)
yes (1094)
yes (335)
yes (0)

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

phosphatase YniC close to the FN3K homologue in bacterial genomes. The figure between parentheses indicates the distance (in base pairs)
separating the two ORFs. Negative values mean that the two sequences partially overlap.
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4362 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
Nocardioides sp., Rubrobacter xylanophilus, Salinibacter
ruber, Thermobifida fusca and Zymomonas mobilis (the
second column of Fig. 1, and data not shown). The
short distance between the two ORFs (average distance
15 nucleotides) and their identical orientation suggest
that they belong to the same operon. In another gen-
ome (from Rhodospirillum rubrum), the sequences
encoding the LMW-PTP and FN3K homologues are
separated by an ORF of  550 bp on the other strand
(data not shown).
blast searches with the Escherichia coli protein-
tyrosine kinase wzc [8] did not indicate the presence
of a homologue of this enzyme in several bacteria
containing the putative LMW-PTP ⁄ FN3K operon
(A. cellulolyticus, Nocardioides sp., S. ruber, T. fusca,
T. thermophilus and Z. mobilis). This makes the
presence of an LMW-PTP homologue all the more
intriguing.
Another potentially interesting association observed
in other genomes is that of the FN3K homologue
with a phosphatase (YniC) belonging to the HAD
family and shown to act, in E. coli, on a variety of
phosphate esters [9]. The FN3K homologue is imme-
diately followed by this phosphatase in the genomes
of Photobacterium profundum and Mannheimia succi-
niciproducens and is separated from it by an ORF in

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

at several positions that are conserved in all other
sequences: Ser131 (replacing Gly); Arg142 (replacing
Asp or Glu); Gln231 (replacing Phe); Arg264 (replac-
ing His); and His272 (replacing Tyr).
Action of bacterial FN3K homologues on LMW
substrates
Five bacterial FN3K homologues, from Enterococcus
faecium, E. coli, Lactobacillus plantarum, Staphylococ-
cus aureus, and T. thermophilus, which share about
30% sequence identity with the human enzyme and
30–40% sequence identity among them, were expressed
in E. coli. They were purified to homogeneity and their
kinetic properties were investigated. All bacterial
FN3K homologues, except for that from
E. coli, phos-
phorylated LMW ribulosamines and erythrulosamines
(Table 1), but not fructosamines (data not shown).
Ribulosamines and erythrulosamines bound to the
Escherichia
coli
FN3K
PFKb
Outer
Membrane
Protein
Hypothetical
protein
Hydrolase
YniCYniB
Thermus

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)

homologue.
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4363
epsilon-amino group of lysine or to cadaverine
(decarboxylated lysine) were substrates for these
enzymes, whereas the ribulosamines bound to the
alpha-amino groups of glycine, leucine and valine were
not (data not shown). Erythrulosamines were better sub-
strates than ribulosamines as indicated by the 6–20-fold
higher catalytic efficiencies observed with erythrulose-
lysine than with ribuloselysine. d-ribulose, d-erythru-
lose and reduced ribuloselysine (pentitollysine), all
tested at 1 mm, were not phosphorylated by the
L. plantarum FN3K homologue.
To check the position of the phosphorylated carbon,
ribuloselysine was phosphorylated by the S. aureus
FN3K homologue, and the phosphorylation product
was purified and analysed by tandem mass spectrome-
try, as previously described [6]. The same fragmenta-
tion spectrum was observed [6]. In particular,
fragments of m ⁄ z 349 and 319 were found, which indi-
cated that the third carbon of the sugar moiety was
phosphorylated.
The E. coli FN3K homologue was inactive on all
the above-mentioned compounds, including ribulose-
lysine and erythruloselysine. It was also inactive on
more than 50 other potential phosphate acceptors,
including d-ribulose, d-xylulose, choline, ethanol-
amine, l-serine, hydroxypyruvate, d-glycerate, thia-
mine and dl-homoserine (tested at concentrations of

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.

With glycated lysozyme, the most active enzyme was
the S. aureus FN3K homologue, and the least active
enzyme was the one from T. thermophilus. Glycated
thioredoxin A was best phosphorylated by mouse
FN3K, which apparently had access to more glycation
sites than its bacterial homologues.
The initial rate of the reaction with lysozyme-bound
ribulosamines was a hyperbolic function of the sub-
strate concentration in the case of the S. aureus
enzyme, for which K
m
and V
max
values could therefore
be determined (Table 1). The other enzymes were not
saturated at the highest concentration of lysozyme-
bound ribulosamines that we tested (500 lm) and their
activities at a substrate concentration of 100 lm are
presented in Table 1. Protein-bound ribulosamines
were poorer substrates than free ribuloselysine: from
the kinetic data, it can be calculated that the activity
on lysozyme-bound ribulosamines amounted to  10%
of the activity observed with free ribuloselysine at the
same concentration.
Lysozyme-bound erythrulosamines were also phos-
phorylated by the four active bacterial FN3K homo-
logues at rates that were about twofold higher
(E. faecium), or two- to fourfold lower (all others), than
those observed with lysozyme-bound ribulosamines.
Lysozyme-bound fructosamines were not detectably

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

(lM)
V
max
(nmolÆmin
)1
Æmg
)1
)
K
m
(lM)
V
max
(nmolÆmin
)1
Æmg
)1
)
Ribuloselysine 300 510 185 730 58 1300 58 450
Ribulosecadaverine 340 2860 44 1340 ND
a
ND ND ND
Erythruloselysine 60 610 15 480 13 1930 6 770
Erythrulosecadaverine 63 800 34 460 ND ND ND ND
Ribulosamine-
lysozyme
> 500 16 ± 0.3
a
> 500 40 ± 4
a

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

s
-
L. plantarum
Time (min)
0 5 10 15 20 25
Time (min)
Incorporated Phosphate
(mol P/mol ribulosamines)
Incorporated Phosphate
(mol P/mol ribulosamines)
0.00
0.05
0.10
0.15
FN3K
S. aureus
E. faecium
M. musculus
T. thermophilu
s
-
L. plantarum
Fig. 4. Phosphorylation of protein-bound ribulosamines by mouse
fructosamine 3-kinase (FN3K) and four bacterial FN3K homologues.
Lysozyme (A) and Escherichia coli thioredoxin A (B) glycated with
ribose were used as substrates at 50 l
M protein-bound ribulosa-
mines. The samples were incubated with [
32
P]ATP[cP] and

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

We expressed the LMW-PTP homologue belonging to
the same operon as the FN3K homologue in the T. ther-
mophilus genome, as well as the two LMW-PTP homo-
logues, PtpA and PtpB [11], present in the S. aureus
genome. PtpA and PtpB, which share, respectively, 38
and 28% sequence identity with T. thermophilus LMW-
PTP, are encoded by genes that are distant from the
gene encoding the FN3K homologue and do not appar-
ently belong to operons. All three recombinant proteins
were purified to homogeneity and their activities tested
both on LMW and macromolecular substrates.
S. aureus PtpB was poorly active or inactive on all
substrates tested, in agreement with previous results
[11]. The other two enzymes dephosphorylated previ-
ously described substrates for LMW-PTP (p-nitro-
phenyl phosphate, FMN), but also ribuloselysine
5-phosphate and erythruloselysine 4-phosphate, the
T. thermophilus enzyme being particularly active on the
latter substrate (Table 2). They did not act on ribose
5-phosphate, fructose 6-phosphate or glucose 6-phos-
phate (data not shown). We checked that dephosphory-
lation of ribuloselysine 5-phosphate by S. aureus
LMW-PTP (PtpA) led to the formation of a substrate
for a bacterial FN3K homologue (the one from E. fae-
cium was used in this experiment). That the resulting
phosphorylation product was ribuloselysine 3-phos-
phate was indicated by its instability and by the
fact that its decomposition led to the appearance of
4,5-dihydroxy-1,2-pentanedione, as determined by mass
spectrometry analysis of the quinoxaline derivative [6].

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

observation leads us to the conclusion that the ances-
tral ‘FN3K’ protein was probably a ribulosamine ⁄ ery-
thrulosamine 3-kinase. The ability to phosphorylate
Table 2. Activities of low-molecular-weight protein-tyrosine-phos-
phatase (LMW-PTP) homologues on LMW substrates. Substrates
were assayed at 0.5 m
M final concentration. Activities are
expressed as nmol inorganic phosphate formed per min and per
mg of protein. The data represent the means of three values ±
SEM. ND, not detectable; Ptp, LMW-PTP of S. aureus; S. aureus,
Staphylococcus aureus; T. thermophilus, Thermus thermophilus.
Substrate
Enzyme activity (nmolÆmin
)1
Æmg of
protein)
T. thermophilus
S. aureus
(PtpA)
S. aureus
(PtpB)
p-Nitrophenyl phosphate 1820 ± 140 7750 ± 230 42 ± 1
Flavin mononucleotide 1730 ± 130 16900 ± 520 24 ± 3
Ribuloselysine-
5-phosphate
127 ± 6 810 ± 80 ND
Erythruloselysine-
4-phosphate
5300 ± 630 1470 ± 100 ND
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.

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

3-kinase when it acts on lysozyme-bound fructosam-
ines amounts to  10 nmolÆmin
)1
Æmg
)1
of protein
(G Delpierre, E Van Schaftingen, unpublished results),
which is about 20-fold lower than the V
max
of the
S. aureus enzyme for protein-bound ribulosamines. As
FN3K and FN3K-RP have been shown to be involved
in protein repair in vivo or in intact cells [2,3,5], this
comparison suggests that this may also be true for
their bacterial homologues.
Endogenous or exogenous source for the
substrates of bacterial FN3K homologues?
The specificity of the FN3K homologues indicates that
they act on sugar derivatives. The latter could either
be of internal or external origin. The absence of associ-
ation of FN3K homologues with a transporter does
not support the idea that they play a role in the
metabolism of an exogenous substrate. This is unlike
fructosamine-6-phosphate deglycases, which are almost
always encoded by operons also containing genes for
putative fructosamine transporters [13,14]. It is
also conceivable that the substrate for the FN3K
0 20 40 60 80 100 120 140
0
10

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

0
10
20
30
40
50
60
70
80
PtpA 10 µg.mL
-1
PtpA 10 µg.mL
-1
PtpB 50 µg.mL
-1
PtpB 50 µg.mL
-1
Lysozyme-RN5P
MBP-TyrP
Lysozyme-RN5P
MBP-TyrP
Time (min)
Fig. 5. Dephosphorylation of protein tyro-
sine-phosphates and protein ribulosamine
5-phosphates by bacterial LMW-PTP homo-
logues. Thermus thermophilus (A) and
Staphylococcus aureus (B) low-molecular-
weight protein-tyrosine-phosphatase
(LMW-PTP) homologues were used to
dephosphorylate myelin basic protein-bound

[
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

A
T. thermophilus
LMW-PTP Ribose-5-P
LMW-PTP Ribose-5-P
20 m
M
-
0 m
M
20 m
M
Time (min)
Liberated Inorganic Phosphate
(mol P/mol lysozyme)
0 20 40 60 80 100
0.00
0.01
0.02
0.03
0.04
0.05
B
T. thermophilus
T. thermophilus
20 m
M
-
T. thermophilus
0 m
M

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

of Staphylo-
coccus aureus FN3K homologue for 15 min
and the incorporated phosphate was mea-
sured. The results are the means of three
independent measurements ± SEM.
0 5 10 15 20 25
0.0
0.1
0.2
0.3
0.4
0.5
A
B
LMW-PTP Erythrose-4-P
10 m
M
S. aureus PtpA
S. aureus PtpA
T. thermophilus
5 m
M
10 mM
5 mM
-
5 m
M
T. thermophilus
LMW-PTP Erythrose-4-P
S. aureus PtpA

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

low-molecular-weight protein-tyrosine-
phosphatase (LMW-PTP) homologues and
rephosphorylation by a bacterial fructos-
amine 3-kinase (FN3K) homologue.
(A) Lysozyme (5 mgÆmL
)1
) glycated with 5
or 10 mm erythrose 4-phosphate was
incubated with 20 lgÆmL
)1
of Thermus
thermophilus LMW-PTP (open squares and
triangles) or Staphylococcus aureus protein-
tyrosine-phosphatase A (closed squares and
triangles) or without LMW-PTP (open circles
and diamonds). Unglycated lysozyme was
also incubated with T. thermophilus LMW-
PTP (closed circles). The liberated inorganic
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
of S. aureus
FN3K for 5 min and the incorporated
phosphate was measured. The results
are the means of three independent
measurements ± SEM.

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].

Interestingly, an LMW-PTP homologue forms an
operon with the FN3K homologue in 11 genomes
from phylogenetically distant bacteria. This association
was not the result of recent lateral transfer events, as
indicated by the fact that the homologues of FN3K
and LMW-PTP present in these operons are very dis-
tant proteins. This operon is therefore either an extre-
mely ancient operon that has been conserved or the
result of distinct recombination events that took place
independently in several bacterial lineages. Both expla-
nations argue strongly for the physiological relevance
of this association.
The LMW-PTP homologue of T. thermophilus, and
one of the two homologues (PtpA) of S. aureus,
dephosphorylated not only phosphotyrosine, but also
ribulosamine 5-phosphates and erythrulosamine
4-phosphates, converting them to substrates for bacterial
FN3K homologues. The apparent absence of a
bacterial-type tyrosine kinase in several of the bacterial
genomes containing the putative LMW-PTP ⁄ FN3K
operon suggests that the LMW-PTP homologues may
serve physiologically to dephosphorylate substrates dif-
ferent from protein-tyrosine phosphates.
We considered the possibility that the FN3K homo-
logues which we studied could be tyrosine kinases.
However, no phosphorylation was observed when
these enzymes were allowed to act upon proteins that
had not been glycated with ribose or erythrose (data
not shown). Interestingly, LMW-PTP did not display
detectable ribulosamine 3-phosphate phosphatase

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,

because phosphoketolase acts then on fructose 6-phos-
phate rather than on xylulose 5-phosphate and there-
fore forms erythrose 4-phosphate (and acetyl-
phosphate) [17].
Although we have no proof at this stage that FN3K
and LMW-PTP homologues participate in the repair
of glycation adducts made from ribose 5-phosphate
and erythrose 4-phosphate, indirect arguments support
this hypothesis. One is the consistent presence of
FN3K and LMW-PTP homologues (Table S1) in Cy-
anobacteria, which are dependent on the Calvin cycle.
This is reminiscent of the high ribulosamine 3-kinase
activity found in spinach leaves and the fact that plant
FN3K homologues are targeted to chloroplasts [6].
Furthermore, the glycation repair hypothesis may offer
a unitary explanation for the presence of enzymes with
a similar function in organisms or cells that are so dis-
similar as bacteria, plants and erythrocytes. Unlike
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4369
bacteria and plant cells, which have tremendous meta-
bolic capacities, mammalian erythrocytes are particu-
larly poor in enzymes of intermediary metabolism,
their function being focused on O
2
transport and hae-
moglobin maintenance. Yet, in all cases, glycation by
ribose 5-phosphate and erythrose 4-phosphate is likely
to be operative. The limitation of this last argument is,
however, that FN3K homologues may actually play

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

tagging them with radiolabelled ATP. This could pro-
vide a demonstration for the presence of these protein
modifications under physiological and pathological
conditions and an avenue for the identification of pro-
teins that have this type of modification, be it linked
or not to glycation.
Experimental procedures
Materials
Reagents, of analytical grade whenever possible, were from
Sigma (St Louis, MO, USA), Acros (Geel, Belgium), Merck
(Darmstadt, Germany), MP-Biomedicals (Irvine, CA, USA)
or Roche (Mannheim, Germany). HisTrap, PD-10, NAP-5
columns, sulfopropyl (SP)-Sepharose, diethylaminoethyl
(DEAE)-Sepharose, Sephacryl S-200, and [
32
P]ATP[cP]
were from GE Healthcare (Diegem, Belgium). Vivaspin-15
centrifugal concentrators were from Vivascience (Go
¨
ttingen,
Germany). Biogel P2 was purchased from Bio-Rad (Her-
cules, CA, USA) and Dowex 1-X8 (100–200 mesh) was
purchased from Acros. N-a-t-Boc-lysine was from Nova-
biochem (Merck). Restriction enzymes were purchased from
Roche or Fermentas (St Leon-Rot, Germany). Mouse recom-
binant FN3K was prepared as previously described [1].
Preparation of ketoamines
Synthesis of glycated proteins
Lysozyme glycated with ribose was prepared as previously
described [4], except that the buffer used was Mes, pH 6,

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

C]ribuloselysine and
a-glycated amino acids were synthesized as described previ-
ously [6]. Xyluloselysine was synthesized as described for
ribuloselysine and fructoselysine was synthesized as
previously described [2]. Pentitollysine was prepared by
incubating (in 0.1 mL) 50 mm ribuloselysine with 0.5 m
NaBH
4
at 4 °C for 16 h after which 10 lL of 60% perchlo-
ric acid was added to destroy the remaining borohydride.
The sample was then purified by gel filtration on Biogel P2.
Ribulosecadaverine was synthesized as described for ri-
buloselysine [6] with the following modifications: cadaverine
(120 lmol) and ribose (240 lmol) were incubated at 65 °C
for 16 h in 3 mL of methanol. The product was then puri-
fied by cation exchange chromatography (SP-Sepharose)
and gel-filtration (Biogel P2) [2]. The yield with respect to
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4370 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
ribose was  5%. Using a similar protocol, erythruloseca-
daverine was synthesized at 50 °C from cadaverine
(240 lmol) and erythrose (60 lmol). The yield with respect
to erythrose was  6%. The titers were determined by
end-point assays using L. plantarum and E. faecium FN3K
homologues and [
32
P]ATP[cP], as previously described for
ribuloselysine [6]. They corresponded to 75% (ribulosecada-
verine) and 25% (erythrulosecadaverine) of the values
obtained by assaying the reducing power [22].

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-

mented with 10% (w ⁄ v) glycerol, and stored at )70 °C.
L. plantarum and S. aureus ‘FN3K’
Both His-tagged proteins were purified on HisTrap columns
(Ni
2+
form), as previously described [27], except that buffer
A (25 mm Hepes, pH 7.4, 1 lgÆmL
)1
of leupeptin,
1 lgÆmL
)1
of antipain and 300 mm NaCl) was used. The
proteins, eluted with 300 mm (L. plantarum) or 150 mm
(S. aureus) imidazole, were concentrated (using Vivaspin-
15) and desalted on PD-10 columns equilibrated with
25 mm Hepes, pH 7.4, 1 mm dithiothreitol and 100 mm
NaCl (L. plantarum)or25mm Hepes pH 7.1 (S. aureus).
The amount of purified protein obtained was  6mg
(L. plantarum) and  60 mg (S. aureus).
E. faecium ‘FN3K’
The extract was diluted twice in buffer B (25 mm Hepes
pH 7.4, 1 lgÆmL
)1
of leupeptin, 1 lgÆmL
)1
of antipain and
1mm dithiothreitol) and then applied to a 25-cm
3
DEAE-
Sepharose column equilibrated with the same buffer. The col-

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

)1
of leupeptin and 1 lgÆmL
)1
of antipain). Frac-
tions containing T. thermophilus LMW-PTP 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 C containing 100 mm
NaCl. The amount of purified protein was  2.4 mg.
S. aureus PtpA and PtpB
The bacterial extracts (25 mL) from 500-mL cultures
were diluted three-fold in buffer C and purified on
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4371
Table 3. Primers used for PCR amplification and expression conditions for the recombinant fructosamine 3-kinase (FN3K) and low-molecular-weight protein-tyrosine-phosphatase (LMW-
PTP) homologues. Restriction sites are written in italics. ATG and stop codons are underlined. Ap, ampicillin; Cm, chloramphenicol; LB, Luria–Bertani; M9(AA), M9 medium supplemented
with 1 m
M MgSO
4
,10mM glucose, 0.5 mgÆL
)1
of biotin, 0.5 mgÆL
)1
of thiamin, and 1 gÆL
)1
of Casamino acids; Ptp, low-molecular-weight protein-tyrosine-phosphatase of Staphylococcus
aureus; Temp., temperature.
Primers

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

Ap 18 °C16h
C
GGATCCTCATGCTGCTAATAATCTATCCAATG BamHI Cm
LMW-PTP homologues
Staphylococcus aureus PtpA GCAGCG
CATATGGTAGATGTAGCATTTGTCTG NdeI pET3a BL21(DE3)plysS LB Ap 18 °C16h
GGCAGC
GGATCCTACCCCTCTTTCAAATTTGCATC BamHI Cm
Staphylococcus aureus PtpB GCAGCG
CATATGAAGATTCTATTCGTTTGTACAG NdeI pET3a BL21(DE3)plysS LB Ap 18 °C16h
GGCAGC
GGATCCTAGCAAATAATATCTTTTAATTTTAAAA BamHI Cm
Thermus thermophilus GCAGCG
CATATGGACCGGCCCGTGCGCGT NdeI pET3a BL21-CodonPlus LB Ap 37 °C4h
GGCAGC
AGATCTCAAGCCCGGCCTTCCGCAG BglII (DE3) Cm
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4372 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS
DEAE-Sepharose and Sephacryl S-200 as described for
the T. thermophilus LMW-PTP homologue. The amount
of purified protein obtained from the 500-mL cultures was
 13 mg for PtpA and PtpB.
Measurement of enzymatic activities
Phosphorylation of protein-bound ketoamines was per-
formed at 30 °C in a medium (0.1 mL) containing 25 mm
Hepes pH 7.1, 1 mm MgCl
2
,1mm EGTA, 50 lm ATP,
10
6

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

of lysozyme.
Phosphorylation by S. aureus FN3K homologue
of the dephosphorylation products of LMW-PTP
homologues
Lysozyme glycated with ribose 5-phosphate or erythrose
4-phosphate (5 mgÆmL
)1
or 0.36 mm lysozyme) was incu-
bated at 37 °C with the indicated concentrations of LMW-
PTP homologues. At the indicated time-points, aliquots
(15 lL) were either rapidly frozen in liquid nitrogen or added
to 100 lLof10mm HCl to assay inorganic phosphate [28].
The frozen aliquots were then diluted twice in the FN3K
incubation medium and phosphorylated with the S. aureus
FN3K homologue for the indicated times (see above).
Database searches, sequence alignment and tree
construction
The NCBI nucleotide sequence database for sequenced
microbial genomes was used to search for homologues of
human FN3K using tBLASTn and the protein database was
also searched using PSI-BLAST [29]. Only the sequences with
a score of > 50 were considered as FN3K homologues. This
criterion permitted the inclusion only of sequences belonging
to COG3001 (i.e. of the FN3K family) and the exclusion
of sequences that are closer to other kinases (e.g. choline ⁄
ethanolamine kinase, methylthioribose kinase, aminoglyco-
side kinases) than to FN3K. The sequences were aligned
using ClustalX with the default parameters [30]. A neigh-
bour-joining tree was constructed using the ‘tree’ option of
the ClustalX program after exclusion of positions with

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).

References
1 Delpierre G, Rider MH, Collard F, Stroobant V, Vansta-
pel F, Santos H & Van Schaftingen E (2000) Identifica-
tion, cloning, and heterologous expression of a mammalian
fructosamine 3-kinase. Diabetes 49, 1627–1634.
2 Delpierre G, Collard F, Fortpied J & Van Schaftingen
E (2002) Fructosamine-3-kinase is involved in an intra-
cellular deglycation pathway. Biochem J 365, 801–808.
3 Veiga-da-Cunha M, Jacquemin P, Delpierre G, Godfra-
ind C, The
´
ate I, Vertommen D, Clotman F, Lemaigre
F, Devuyst O & Van Schaftingen E (2006) Increased
protein glycation in fructosamine 3-kinase-deficient
mice. Biochem J 399, 257–264.
4 Collard F, Delpierre G, Stroobant V, Matthijs G &
Van Schaftingen E (2003) A mammalian protein homo-
logous to fructosamine-3-kinase is a ketosamine-3-
kinase acting on psicosamines and ribulosamines but
not on fructosamines. Diabetes 12, 2888–2895.
5 Collard F, Wiame E, Bergans N, Fortpied J, Vertommen
D, Vanstapel F, Delpierre G & Van Schaftingen E (2004)
Fructosamine 3-kinase-related protein and deglycation
in intact erythrocytes. Biochem J 382, 1–7.
6 Fortpied J, Gemayel R, Stroobant V & Van Schaftingen
E (2005) Plant ribulosamine ⁄ erythrulosamine 3-kinase, a
putative protein-repair enzyme. Biochem J 388, 795–802.
R. Gemayel et al. Bacterial fructosamine 3-kinase homologues
FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS 4373
7 Fortpied J, Gemayel R, Vertommen D & Van Schaftin-

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

(2005) Identification of glucoselysine-6-phosphate degly-
case, an enzyme involved in the metabolism of the fructa-
tion product glucoselysine. Biochem J 392, 263–269.
15 McKay GA, Thompson PR & Wright GD (1994) Broad
spectrum aminoglycoside phosphotransferase type III
from Enterococcus: overexpression, purification, and
substrate specificity. Biochemistry 33, 6936–6944.
16 Sandwick R, Johanson M & Breuer E (2005) Maillard
reactions of ribose 5-phosphate and amino acids. Ann
N Y Acad Sci 1043, 85–96.
17 Veiga-da-Cunha M, Santos H & Van Schaftingen E
(1993) Pathway and regulation of erythritol formation in
Leuconostoc oenos. J Bacteriol 175, 3941–3948.
18 Li C & Clarke S (1992) A protein methyltransferase
specific for altered aspartyl residues is important in
Escherichia coli stationary-phase survival and heat-shock
resistance. Proc Natl Acad Sci USA 89, 9885–9889.
19 Moskovitz J, Rahman MA, Strassman J, Yancey SO,
Kushner SR, Brot N & Weissbach H (1995) Escheri-
chia coli peptide methionine sulfoxide reductase gene:
regulation of expression and role in protecting against
oxidative damage. J Bacteriol 177, 502–507.
20 Collet JF, D’Souza JC, Jakob U & Bardwell JC (2003)
Thioredoxin 2, an oxidative stress-induced protein,
contains a high affinity zinc binding site. J Biol Chem
278, 45325–45332.
21 Fortpied J, Maliekal P, Vertommen D & Van Schaftingen
E (2006) Magnesium-dependent phosphatase-1 is a
protein-fructosamine-6-phosphatase potentially involved
in glycation repair. J Biol Chem

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.

30 Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F
& Higgins DG (1997) The CLUSTAL_X windows
interface: flexible strategies for multiple sequence
alignment aided by quality analysis tools. Nucleic Acids
Res 25, 4876–4882.
Supplementary material
The following supplementary material is available online:
Table S1. Distribution of fructosamine 3-kinase
(FN3K) and low-molecular-weight protein-tyrosine-
phosphatase (LMW-PTP) homologues in prokaryotes.
This material is available as part of the online article
from
Please note: Blackwell Publishing is not responsible
for the content or functionality of any supplementary
materials supplied by the authors. Any queries (other
than missing material) should be directed to the corre-
sponding author for the article.
Bacterial fructosamine 3-kinase homologues R. Gemayel et al.
4374 FEBS Journal 274 (2007) 4360–4374 ª 2007 The Authors Journal compilation ª 2007 FEBS