Báo cáo Y học: Galactosyl-mimodye ligands for Pseudomonas fluorescens b-galactose dehydrogenase pot

Galactosyl-mimodye ligands for
Pseudomonas fluorescens
b-galactose dehydrogenase
Design, synthesis and evaluation
C. F. Mazitsos
1
, D. J. Rigden
2
, P. G. Tsoungas
3
and Y. D. Clonis
1
1
Laboratory of Enzyme Technology, Department of Agricultural Biotechnology, Agricultural University of Athens, Greece;
2
Embrapa Recursos Gene
´
ticos e Biotecnologia, Brası
´
lia, Brazil;
3
Department of Pharmaceutical and Biological Chemistry,
School of Pharmacy, University of London, UK
Protein molecular modelling and ligand docking were
employed for the design of anthraquinone galactosyl-bio-
mimetic dye ligands (galactosyl-mimodyes) for the target
enzyme galactose dehydrogenase (GaDH). Using appro-
priate modelling methodology, a GaDH model was build
based on a glucose-fructose oxidoreductase (GFO) protein
template. Subsequent computational analysis predicted
chimaeric mimodye-ligands comprising a NAD-pseudomi-

metic moiety (anthraquinone diaminobenzosulfonic acid)
and a galactosyl-mimetic moiety (2-amino-2-deoxygalactose
or shikimic acid) bearing an aliphatic ÔlinkerÕ molecule. In
addition, the designed mimodye ligands had an appropriate
in length and chemical nature ÔspacerÕ molecule via which
they can be attached onto a chromatographic support
without steric clashes upon interaction with GaDH. Fol-
lowing their synthesis, purification and analysis, the ligands
were immobilized to agarose. The respective affinity adsor-
bents, compared to other conventional adsorbents, were
shown to be superior affinity chromatography materials for
the target enzyme, Pseudomonas fluorescens b-galactose
dehydrogenase. In addition, these mimodye affinity adsor-
bents displayed good selectivity, binding low amounts of
enzymes other than GaDH. Further immobilized dye-lig-
ands, comprising different linker and/or spacer molecules, or
not having a biomimetic moiety, had inferior chromato-
graphic behavior. Therefore, these new mimodyes suggested
by computational analysis, are candidates for application in
affinity labeling and structural studies as well as for purifi-
cation of galactose dehydrogenase.
Keywords: affinity chromatography; biomimetic ligands;
galactose dehydrogenase; molecular modelling; triazine dyes.
Galactose dehydrogenase (GaDH;
D
-galactose: NAD
+
1-oxidoreductase; EC 1.1.1.48) catalyses the dehydrogena-
tion of b-
D

-galactopyranose in the presence of NAD
+
to
D
-galacto-1,5-lactone and NADH, acting on the C1 posi-
tion of the sugar substrate. The enzyme generally shows no
absolute specificity either for NAD
+
,asNADP
+
is also
used, albeit to a lesser degree. Nor is the enzyme specific for
D
-galactose, as
D
-fucose is a better substrate, although other
sugars (e.g.
L
-arabinose, 2-deoxy-
D
-galactose) are less
reactive. The kinetic mechanism is ordered Bi-Bi, with the
NAD
+
binding first to the enzyme [1]. GaDH from
Pseudomonas fluorescens is the best studied example, as it
has been cloned and expressed in Escherichia coli [2] and its
full nucleotide sequence determined [3]. The active macro-
molecule possesses two binding sites [4] and consists of two
identical subunits each of 33 kDa (304 amino-acid residues)

[3]. GaDH from Pseudomonas saccharophila has been
studied to a lesser extent [5], whereas the enzyme has been
identified in plants (e.g. green peas, oranges and Arabidopsis
thaliana), algae (e.g. Iridophycus flaccidum) and several
mammals including humans. No information is available
regarding the catalytic mechanism of GaDH, and its
structure has not been determined experimentally or
modelled.
GaDH is an important analytical tool as at alkaline pH
the product galactonolactone is hydrolysed, so that the
reaction becomes irreversible. The enzyme is therefore
useful for the determination of b-
D
-galactose and
a-
D
-galactose, after the latter is converted to the former
by the application of exogenous mutarotase. GaDH is
also exploited for the determination of lactose; the milk
sugar is hydrolysed by lactase, coupled to GaDH which
acts on the resulting b-
D
-galactose. Despite the utility of
GaDH, a simple and rapid purification method is not
available.
The ability to combine knowledge of X-ray crystallo-
graphic studies, NMR and homology structures with
defined or combinatorial chemical synthesis and advanced
computational tools has made rational design of affinity
ligands more feasible, powerful, logical and faster [6]. In the

present work, rigorous protein molecular modelling was
Correspondence to Y. D. Clonis, Laboratory of Enzyme Technology,
Department of Agricultural Biotechnology, Agricultural
University of Athens, 75 Iera Odos Street, GR-11855 Athens,
Greece. Fax: + 30 210 5294307, Tel.: + 30 210 5294311,
E-mail:
Abbreviations: ADH, alcohol dehydrogenase; BM, biomimetic ligand
or mimodye ligand; CB3GA, Cibacron blue 3GA; GaDH, galactose
dehydrogenase; GaO, galactose oxidase; GFO, glucose-fructose
oxidoreductase; GlDH, glucose dehydrogenase; GlO, glucose oxidase;
VBAR, Vilmafix Blue A-R; CDI, 1,1¢-carbonyldiimidazole.
Enzymes: galactose dehydrogenase (GaDH;
D
-galactose: NAD
+
1-oxidoreductase; EC 1.1.1.48).
(Received 31 May 2002, revised 16 August 2002,
accepted 28 August 2002)
Eur. J. Biochem. 269, 5391–5405 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03211.x
used to create an objectively sound model of GaDH using as
the best available template glucose-fructose oxidoreductase
(GFO). This model was then exploited in the design of novel
galactosyl-biomimetic chlorotriazine dye-ligands (mimodye
ligands) with bifunctional or chimaeric characteristics. In
particular, these galactosyl-mimodye ligands are designed to
bear a structural portion that interacts with the NAD
+
-
binding site and a biomimetic moiety that interacts with the
sugar-binding site of GaDH. The effectiveness of the

bifunctional (chimaeric) ligand concept has been previously
demonstrated with ketocarboxyl- [7,9] and glutathionyl-
biomimetic [10] ligands but never with sugar ones. These
mimodye ligands are expected to become useful tools for the
identification of amino-acid residues of the binding sites of
GaDH after affinity labelling. For this purpose, the
galactosyl-mimodyes were designed to bear a reactive
chloro-triazine structural scaffold, present in all reactive
triazinyl-dye ligands including the archetypal CB3GA and
VBAR. Other mimodyes and certain conventional triazine
dyes are known to act as affinity labels due to their
chlorine(s) atom(s) which react with appropriate residues
of the targeted enzyme active site [11–13]. Furthermore,
when the chlorine was substituted with a carefully chosen
spacer molecule, a nonreactive biomimetic ligand was
obtained which could be immobilized on a chromatography
support. We envisage that these immobilized ligands will be
of great use in the purification of GaDH from different
sources.
EXPERIMENTAL PROCEDURES
Materials
b-Galactose dehydrogenase (EC 1.1.1.48, P. fluorescens
gene expressed in E. coli), galactose oxidase crude lyophi-
lized powder (EC 1.1.3.9, from Dactylium dendroides),
glucose oxidase crude lyophilized powder (EC 1.1.3.4,
from Aspergillus niger, crude),
D
(+)-galactosamine
(2-amino-2-deoxy-
D

-galactopyranose; chondrosamine),
D
(+)-galactose (minimum 99%),
D
(+)-glucose, 1,3-diamino-
2-hydroxypropane, bromoacetic acid N-hydroxysuc-
cinimide ester, e-amino-n-caproic acid, ethylene-diamine,
1,5-diaminopentane, 1,6-hexane-diamine, 1,12-diaminodo-
decane, 1-ethyl-3-(3-dimethylamino-propyl) carbodiimide
(EDAC), 1,1¢-carbonyldiimidazole, o-tolidine, o-dianisi-
dine, lipophilic Sephadex LH-20, CM–Sepharose CL-6B
and DEAE–Sepharose CL-6B were obtained from Sigma
(St Louis, MO, USA). All other diaminoalkanes were
obtained from Aldrich (USA), whereas, shikimic acid was
obtained from Fluka (USA). Peroxidase (from horseradish,
grade I), NAD
+
(crystallized lithium salt c. 100%) and
crystalline bovine serum albumin (fraction V) were obtained
from Boehringer Mannheim (Germany). Hexylamine and
nutrient broth (for microbiology) were obtained from
Merck (Germany). The agarose chromatography gel
Sepharose CL-6B was obtained from Pharmacia. F324
P. fluorescens biovar V1 was kindly donated by G. J.
Nychas (Laboratory of Microbiology and Biotechnology
of Foods, Agricultural University of Athens). Baker’s
yeast, green peas and rabbit liver were purchased at the
local market. Glucose dehydrogenase was extracted from
P. fluorescens and baker’s yeast, while alcohol dehydro-
genase was extracted from baker’s yeast and green peas.

Protein modelling
Fold recognition methods [14–17] were employed to deter-
mine the best template to use for construction of a model of
GaDH. Given the low sequence identity between GaDH
and the GFO template used (17%) a rigorous modelling
strategy was used, as previously (e.g [18,19]). In this way the
challenge of modelling based on low sequence identity was
met with a strategy designed to maximize model accuracy.
Although errors will undoubtedly remain, the probability of
producing a useful model is thereby enhanced. The essential
elements of this strategy are the construction and analysis of
multiple models (20 in this case), derived from limited
randomization of initial coordinates and made with the
program
MODELLER
[20], followed by analysis of packing
and solvent exposure characteristics with
PROSA
II [21]. The
resulting profiles showed regions of unusual protein struc-
ture characteristics as peaks attaining positive values. These
regions may result from locally inaccurate target-template
alignment so that variant alignments, altered in these
doubtful regions, were tested through further cycles of
model construction and analysis. When better
PROSA
results
were obtained for the variant alignment it was assumed to
be more correct than the original. Stereochemical analysis
using

PROCHECK
was also employed, particularly when the
optimal target-template alignment had been reached. Pro-
tein models were visualized using O [22]. Structurally similar
proteins to the template were sought in the FSSP database
( [23].
STRIDE
[24] was used
for the definition of secondary structure.
Ligand design and docking
The ideal biomimetic would combine moieties that bind
both to the cofactor NAD and the substrate binding sites.
The initially considered Ôbuilding elementsÕ were two com-
mercially available compounds: (a) anthrquinone-diamino-
benzosulfonyl-dichlorotriazine (Vilmafix blue A-R or
VBAR) containing three of the four ring systems of the
well known dye Cibacron Blue 3GA (CB3GA), both known
binding mimics of NAD(P) [8,9,25] and (b) 2-amino-
2-deoxygalactose, a substrate of GaDH [1]. Both these
molecules have readily modifiable chemical groups to which
could be attached an appropriate ÔlinkerÕ molecule in order
to effect their fusion. 1-Amino-1-deoxygalactose, although
commercially available, was not considered as GaDH
attacks at the C1 position of the substrate, so that this
position was thought better preserved in the ligand.
However, in place of galactose shikimic acid was considered
which, although only moderately structurally similar to
GaDH substrates, has a clear advantage over them in terms
of chemical stability. Finally, a ÔspacerÕ molecule of appro-
priate length and chemical nature was designed to chemi-

cally attach the complete ligand, via its triazine group (ring
3), to the chromatographic matrix.
The HIC-UP database of heterocompounds [26] was
used as a source of the Cibacron Blue-derived, b-
D
-galactose
and shikimic acid components. These were rotated and
translated with respect to the protein model using
O
[22]
until optimal steric and chemical complementarity was
reached. The tendency of Cibacron Blue-like ring systems
to bind in NAD(P) binding sites with anthraquinone
mimicking adenine, along with biochemical data regarding
5392 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
sugar binding to related enzymes provided useful informa-
tion to guide the docking, as described later. Side chain
reorientations to rotameric conformations were allowed
where they significantly enhanced interactions with ligands.
The mimodye ligands (e.g. BM1 and BM2) were mod-
elled through the fusion of their respective enzyme-bound
components and the resulting complexes refined using
CNS
[27]. Topology and parameter files for energy minimization
of the ligand were generated using
XPLO
2
D
[28] and hand-
edited to reflect ideal stereochemical values.

Synthesis and purification of the dye-ligands
Amino-alkyl-VBAR dyes. (Table 1, structures aVBAR-
fVBAR). Solid commercial VBAR (50 mg, 0.045 mmol
dichloroform, purity 61.3%, w/w) was added to cold water
(2 mL) and the solution was slowly introduced under
stirring to a solution (3 mL) of the alkyl-diamines
(0.73 mmol). The pH was adjusted to 8.9–9.0 and kept at
this value with NaOH (0.1
M
) until the end of the reaction
(2.5–3 h, 25 °C). The progress of each reaction was
monitored by TLC (1-butanol-2-propanol-ethylacetate-
wate, 2 : 4 : 1 : 3 v/v/v/v) upon completion of the reaction,
solid NaCl was added (final content 3%, w/v) and the
mixture was left at 4 °C. The pH of the mixture was
adjusted with HCl (1
M
) to 1.0 and the precipitate was
filtered (Whatman paper filter 50, hardened), washed with
5mLeachofHCl(1
M
) and cold acetone, then with 7 mL
of diethyl ether and dried under reduced pressure. The solid
dye (approximately 30 mg) was dissolved in 50 : 50 water/
methanol (50%) and dimethylsulfoxide (50%) mixture, and
purified on a lipophilic Sephadex LH-20 column
(30 · 2.5 cm) [29]. The purified product was stored in a
desicator at 4 °C.
Hydrophilic spacer-VBAR dye. (Table 1, structure
gVBAR; Fig. 1). Stage 1: solid commercial VBAR

(20 mg, 0.018 mmol dichloroform, purity 61.3%, w/w)
was added to cold water (1 mL) and the solution was slowly
introduced under stirring to 1,3-diamino-2-hydroxypropane
(3 mL, 0.29 mmol). The pH was adjusted to 8.9–9.0 and
kept at this value with NaOH (0.1
M
) until the end of the
reaction (2.5–3 h, 25 °C). The progress of the reaction was
monitored by TLC (1-butanol-2-propanol-ethylacetate-
water, 2 : 4 : 1 : 3 v/v/v/v). Upon completion of the
reaction, the dye was purified according to the method
already described (see above). Stage 2: the purified product,
1,3-diamino-2-hydroxypropano-VBAR, was dissolved in
dimethylsulfoxide/water (3 mL, 50 : 50, v/v) and the pH
was adjusted to 7.5 with NaOH (0.1
M
). 0.2 mmol of
bromoacetic acid N-hydroxysuccinimide ester [30,31]
were dissolved in dioxane (1 mL) and this solution was
Table 1. The structures of amino-alkyl-VBAR dyes (a-fVBAR), hydrophilic spacer-VBAR dye (gVBAR), galactosamine-VBAR dye and archetypal
VBAR dye.
Ligand R
1
R
2
a
aVBAR –NH-(CH
2
)
2

-NH
2
–NH
2
bVBAR –NH-(CH
2
)
4
-NH
2
–NH
2
cVBAR –NH-(CH
2
)
6
-NH
2
–NH
2
dVBAR –NH-(CH
2
)
8
-NH
2
–NH
2
eVBAR –NH-(CH
2

)
10
-NH
2
–NH
2
fVBAR –NH-(CH
2
)
12
-NH
2
–NH
2
gVBAR –NH
2
Galactosamine-VBAR
b
–Cl
VBAR –Cl (– NH
2
)
a
–Cl
a
Following ligand immobilization, the -NH
2
group has replaced the -Cl atom.
b
The galactosamine-VBAR dye was synthesized employing

the procedure for amino-alkyl-VBAR dyes but using the amino-sugar instead the diamino-alkane.
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5393
introduced to the dye solution. The pH was maintained to
7.5 until the end of the reaction (1.5 h, 4 °C, as judged by
TLC). The progress of the reaction was monitored by TLC
(1-butanol-2-propanol-ethylacetate-water, 2 : 4 : 1 : 3 v/v/
v/v). Upon completion of the reaction, the mixture was
lyophilized and the dye was purified on the lipophilic
Sephadex LH-20 column [29]. Stage 3: the purified product,
bromoacetylated 1,3-diamino-2-hydroxypropano-VBAR,
was dissolved in 0.1
M
NaHCO
3
, pH 9.0 (2 mL) and the
solution was slowly introduced under stirring to a solution
of 0.4
M
1,3-diamino-2-hydroxypropane in 0.1
M
NaHCO
3
,
pH 9.0 (2 mL), while maintaining the pH to 9.0 with HCl
(1
M
). The solution was then left under stirring for another
48–72 h (25 °C), without further adjustment of the pH. The
progress of the reaction was monitored by TLC (1-butanol-
2-propanol-ethylacetate-water, 2 : 4 : 1 : 3 v/v/v/v). Upon

completion of the reaction, the dye was purified according
to the method already described (see above).
Biomimetic dye BM1. (Table 2, structure BM1; Fig. 1)
Stage 1: purified hydrophilic spacer-VBAR, structure g
(approx. 15 mg, 0.017 mmol) was dissolved in dimethyl-
sulfoxide/water (3 mL, 50 : 50, v/v) and the solution was
introduced under stirring to e-amino-n-caproic acid (2 mL,
0.17 mmol). The pH was adjusted to 9.0 and the mixture
was left shaking at 60 °C for 3 h. The progress of the reaction
was monitored by TLC (1-butanol-2-propanol-ethylacetate-
water, 2 : 4 : 1 : 3 v/v/v/v). Upon completion of the
reaction, the dye was purified according to the method
already described (see above). Control dye C
6
gVBAR
(Table 2) was synthesized in the same way. Stage 2: the
purified product obtained from stage 1, was dissolved in
dimethylsulfoxide/water (3 mL, 50 : 50, v/v), introduced to
a solution of
D
(+)-galactosamine (3 mL, 0.62 mmol), and
the pH was adjusted to 4.6, before freshly prepared solution
of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (0.3 mL,
250 mg) was introduced dropwise under stirring over a
period of 5 min, while maintaining the pH at 4.6–5.0. The
reaction was stirred for 20 h at 25 °C without pH
adjustment and monitored by TLC (1-butanol-2-propanol-
ethylacetate-water, 2 : 4 : 1 : 3 v/v/v/v). A silver nitrate
ammonia solution was used as a spray reagent for detecting
the galactose-analogue in the newly synthesized dye [32].

The product, structure BM1, was precipitated by addition
of solid NaCl (final content 15%, w/v), filtered and washed
with 7 mL of NaCl solution (15%, w/v) and 5 mL of cold
acetone, and dried under reduced pressure. The product was
re-suspended in 2 mL of water and precipitated by addition
of solid NaCl (final content 10%, w/v). The precipitate was
filtered and washed with 7 mL each of NaCl solution (10%,
w/v) and cold acetone, desiccated with 7 mL of diethyl ether
and dried under reduced pressure.
Biomimetic dye BM2. (Table 2, structure BM2; Fig. 2).
Stage 1: solid commercial VBAR (20 mg, 0.018 mmol
dichloroform, purity 61.3%, w/w) was added to cold water
(1 mL) and the solution was slowly introduced under
stirring to a solution (3 mL) of 1,3-diaminopropane
(0.29 mmol). The pH was adjusted to 8.9–9.0 and kept at
this value with NaOH (0.1
M
) until the end of the reaction
(2.5–3 h, 25 °C). The progress of each reaction was
monitored by TLC (1-butanol-2-propanol-ethylacetate-
water, 2 : 4 : 1 : 3 v/v/v/v). Upon completion of the
reaction, the dye was purified according to the method
already described (see above). Stage 2: the purified product,
VBAR-1,3-diaminopropane, was dissolved in dimethylsulf-
oxide/water (3 mL, 50 : 50, v/v), introduced to a solution of
shikimic acid (3 mL, 0.62 mmol), and the pH was adjusted
to 4.6, before freshly prepared solution of 1-ethyl-3-(3-
dimethylaminopropyl)carbodiimide (0.3 mL, 250 mg) was
introduced dropwise under stirring over a period of 5 min,
while maintaining the pH at 4.6–5.0. The reaction was

stirred for a further 20 h at 25 °C without pH adjustment
and monitored by TLC (1-butanol-2-propanol-ethylacetate-
Fig. 1. Steps for the synthesis of gVBAR dye and of mimodye BM1.
5394 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
water, 2 : 4 : 1 : 3 v/v/v/v). The product, VBAR-1,3-diami-
nopropano-shikimic acid, was precipitated by addition of
solid NaCl (final content 15%, w/v), filtered and washed
with 7 mL of NaCl solution (15%, w/v) and 5 mL of cold
acetone, and dried under reduced pressure. The product was
dissolved in a 50 : 50 water:methanol (50%) and dimeth-
ylsulfoxide (50%) mixture, and purified to homogeneity on
a lipophilic Sephadex LH-20 column (30 · 2.5 cm) [29].
Control dyes C
6
NgVBAR and C
3
NgVBAR (Table 2) were
synthesizedinthesamewasasinstages1and2.Stages3–5:
the purified product, VBAR-1,3-diaminopropano-shikimic
acid, was dissolved in dimethylsulfoxide/water (3 mL,
50 : 50, v/v/v) and the solution was introduced under
stirring to 1,3-diamino-2-hydroxypropane (2 mL,
0.17 mmol). The pH was adjusted to 9.0 and the mixture
was left shaking at 60 °C for 3 h. The progress of the reaction
was monitored by TLC (1-butanol-2-propanol-ethylacetate-
water, 2 : 4 : 1 : 3 v/v/v/v). Upon completion of the
reaction, the dye was purified according to the method
already described (as above). The purified product, 1,3-
diamino-2-hydroxypropano-VBAR-1,3-diaminopropano-
shikimic acid, was dissolved in dimethylsulfoxide/water

(3 mL, 50 : 50, v/v) and the pH was adjusted to 7.5 with
NaOH (0.1
M
). 0.2 mmol of bromoacetic acid N-hydroxy-
succinimide ester were dissolved in dioxane (1 mL) and this
solution was introduced to the dye solution. The pH was
maintained at 7.5 until the end of the reaction (1.5 h, 4 °C,
as judged by TLC). The progress of the reaction was
monitored by TLC (1-butanol-2-propanol-ethylacetate-
water, 2 : 4 : 1 : 3 v/v/v/v). Upon completion of the
reaction, the mixture was lyophilized and the dye was
purified by applying preparative TLC as follows: lyophilized
reaction mixture was dissolved in dimethylsulfoxide/water
(0.4 mL, 50 : 50, v/v) and the solution applied on a
Kieselgel 60 plate (silica gel 60, 0.2 mm, 20 · 20 cm,
Merck). The plate was developed using a 1-butanol-2-
propanol-ethylacetate-water (2 : 4 : 1 : 3 v/v/v/v) mixture.
Following completion of the chromatography, the plate was
dried and the band of interest was scraped off. The desired
dye was extracted from the silica gel with water, filtered
through a Millipore cellulose membrane filter (0.45 lm pore
size) and lyophilized. The purified product, bromoacetylated
1,3-diamino-2-hydroxypropano-VBAR-1,3-diaminopropano-
shikimic acid, was dissolved in 2 mL of 0.1
M
NaHCO
3
,
pH 9.0, and the solution was slowly introduced under
stirring to a 2-mL solution of 0.4

M
1,3-diamino-2-hydroxy-
propane in 0.1
M
NaHCO
3
, pH 9.0 (the pH maintained at 9
using 1
M
HCl). The solution was then left under stirring for
another 48–72 h (25 °C), without further adjustment of the
pH. The progress of the reaction was monitored by TLC
(1-butanol-2-propanol-ethylacetate-water, 2 : 4 : 1 : 3 v/v/
v/v). Upon completion of the reaction, the product,
hydrophilic spacer-VBAR-1,3-diaminopropano-shikimic
acid, was precipitated by addition of solid NaCl (final
content 15%, w/v), filtered and washed with 7 mL of NaCl
solution (15%, w/v) and 5 mL of cold acetone, and dried
under reduced pressure. The product was re-suspended in
2 mL of water and precipitated by addition of solid NaCl
(final content 10%, w/v). The precipitate was filtered
and washed with 7 mL each of NaCl solution (10%, w/v)
and cold acetone, desiccated with 7 mL of diethyl ether and
dried under reduced pressure.
Table 2. The structures of the mimodyes BM1 and BM2 and the control dyes.
Dye-ligand
–R
BM1
BM2
C

6
gVBAR
C
6
NgVBAR
C
3
NgVBAR
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5395
Spectroscopic characterization and analysis
of dye-ligands
Prior to their characterization, all dyes synthesized in this
work were purified by following appropriate purification
procedures, depending on the requirements of each syn-
thetic step. During the preliminary purification stage,
inorganic and certain organic contaminants were removed
by extraction with ethyl ether and precipitation with
acetone. In the next stage, complete dye purification was
achieved on a Sephadex LH-20 lipophilic column, where
salts and other organic impurities were removed. Prepara-
tive TLC has also been used as a purification technique at
certain stages. Successful dye purification was shown by
TLC analysis (single blue bands). Tables 1–3 summarize the
structures, molecular masses, molar absorption coefficients
(e), and absorption maxima (k
max
) of the purified free dyes.
The absorption maxima (k
max
) of the purified dyes were

determined by aqueous dye aliquots (50 l
M
)takeninthe
range 850–450 nm. The molar absorption coefficients
(e-values) were calculated from the linear section of
reference curves derived by plotting dye concentration vs.
absorption (620 nm, 20–100 l
M
)[29].
NMR spectra. These were recorded on a BRUKER AM
250 or 500 MHz spectrometer using standard pulse
sequences. Samples were analysed as solutions in dimethyl-
sulfoxide-d
6
or D
2
O. The ABCD and ABX patterns of the
aromatics of anthraquinone and 1,4-diamino-substituted
phenyl rings, respectively, are expectedly present in the
1
H
NMR spectra of all the compounds. The pattern is
securely based by comparison with the
1
Hand
13
CNMR
spectra of the commercially purchased reference com-
pound VBAR.
The

1
H NMR spectra of BM2 and BM1 show very
complex, yet discernible high-field multiplets, ranging from
d 1.05–4.85 p.p.m. and d 1.0–3.50 p.p.m., attributed to
-CH-CH, -CH-NH and -CH-OH couplings, respectively.
Multiplets at d 5.35–6.86 p.p.m. and d 5.20–6.80 p.p.m. are
attributed to the amide –NH resonance of both BM2 and
BM1, respectively.
Mass spectra. Electron impact (EI) and fast atom bom-
bardment (FAB) mass spectra were recorded on a VG
ZAB/SE double focusing low/high resolution spectrometer.
Electrospray ionization (ESI) spectra were run on a
Finnigan LCQ DUO spectrometer. It is known that the
reference compound VBAR does not exhibit a molecular
ion (M
+
) peak under EI ionization [49]. Indeed, no such ion
has been observed in either EI, FAB or ESI spectra.
Compounds BM2 and BM1 behaved similarly. However,
comparing the highly complexed fragmentation patterns of
BM2 and BM1, under the above ionization modes, allowed
for the detection of fragments, resulting, most probably,
from primary C–N and C–O fission.
Immobilization of amino-dyes to carbonyldiimidazole-
activated agarose and determination
of immobilized dye concentration
Agarose beads (Sepharose CL-6B) were activated with 1,1¢-
carbonyldiimidazole (CDI) by a modification of the pub-
lished method [33]. An immobilized dye concentration of
approximately 2.0 lmolÆg

)1
Sepharose CL-6B (moist gel)
was achieved by using appropriate amount of CDI in the
activation step. Exhaustively washed (300 mL of water)
Sepharose CL-6B (600 mg, moist weight) was washed
sequentially with dioxane-water (10 mL, 3 : 7, v/v), diox-
ane-water (10 mL, 7 : 3, v/v), dioxane (10 mL) and dried
dioxane (25 mL). The gel was re-suspended in dried dioxane
(1.1 mL) to which CDI (200 mg) had already been added
Fig. 2. Steps for the synthesis of mimodye BM2.
5396 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
and the mixture was tumbled for 1–2 h at 20–25 °C.
Activated gel was washed with dried dioxane (20 mL) and
used immediately. A solution of amino-dye (0.02 mmol) in
dimethylsulfoxide/water (2 mL, 50 : 50, v/v) was adjusted
to pH 10.0 with 2
M
sodium carbonate solution, whereupon
CDI activated Sepharose CL-6B (600 mg) was added. The
mixture was tumbled overnight (25 °C) and washed
sequentially with water (50 mL), NaCl solution (25 mL,
1
M
), water (25 mL), dimethylsulfoxide solution (6 mL,
50 : 50, v/v) and, finally, water (50 mL). In the case of
ligands with remaining active Cl, the adsorbent, after the
immobilization procedure, was suspended in NH
3
solution
(1

M
, pH 8.5) and tumbled for another 3 h. The dyed gels
were stored as moist gels in 20% methanol at 4 °C. Table 3
summarizes the conditions and performance of immobil-
ization reactions of the dye-ligands.
Determination of immobilized dye concentration was
achieved according to [34]. The concentration of the
immobilized dyes was calculated as micromoles of dye
per gram moist mass gel, using the molar absorption
coefficients shown in Table 3. All adsorbents were
substituted with dye-ligand at approximately the same
level (1.8–2.3 lmol dyeÆgmoistgel
)1
). When comparing
affinity adsorbents, equal ligand substitution effected by
synthesis rather than dilution with unsubstituted gel is an
important but often overlooked prerequisite. Wide var-
iations in immobilized ligand concentration are undesir-
able because the results obtained from the employment
of such affinity adsorbents may lead to misleading
conclusions. Extreme levels of ligand substitution may
lead to no binding, due to the steric effect caused by the
large number of dye molecules, or even to nonspecific
protein binding [47,48].
Assay of enzyme activities and protein, and inactivation
of galactose dehydrogenase by VBAR
Galactose dehydrogenase (GaDH), galactose oxidase
(GaO), glucose oxidase (GlO), glucose dehydrogenase
(GlDH) and alcohol dehydrogenase (ADH) assays were
performed at 25 °C with the exception of GlO, which was

performed at 35 °C. The assays were performed according
to [35] [36], [37], [38], and [39], respectively. All assays were
performed in a double beam UV-visible spectrophotometer
equipped with a thermostated cell holder (10-mm path-
length). For GaO, one unit of enzyme activity is defined as
the amount that produces a DA
425nm
of 1.0 per min at the
conditions of the assay. For the rest of the enzymes, one unit
of enzyme activity is defined as the amount that catalyses
the conversion of 1 lmol of substrate to product per min.
Protein concentration was determined by the method of
Bradford [40] or by a modified Bradford’s method [41],
using bovine serum albumin (fraction V) as standard.
Inactivation of GaDH by VBAR was performed in
incubation mixture containing in 1 mL total volume
(35 °C): 100 lmol Hepes/NaOH buffer pH 8.5, 30 nmol
VBAR, 0.13 U GaDH (enzyme assay at 25 °C). The rate of
GaDH inactivation was followed by periodically removing
samples (100 lL) from the incubation mixture for assay of
enzymatic activity. Competitive inactivation studies of
GaDH by VBAR were performed in the above reaction
mixture of 1 mL total volume (35 °C) containing also
1 lmol NAD
+
.
Preparation of cell extracts with enzyme activities
P. fluorescens dry cells (1.5 mg) were suspended in 1 mL of
10 m
M

potassium phosphate buffer containing 1 m
M
EDTA, pH 6.5, 7.0 or 7.5, and ultrasonically disintegrated
(Vibra Cell, 400 Watt, Sonics & Materials) (amplitude:
40%, 2 s sonication, 5 s pause, 8 cycles, 4 °C).Celldebris
was removed by centrifugation (5000 g,20min,4°C) and
the supernatant was dialyzed overnight at 4 °Cagainst2L
of 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA, pH 6.5, 7.0 or 7.5. The dialysate was clarified
through a Milipore cellulose membrane filter (0.45 lm pore
size), affording, typically, 0.08 U GlDHÆmL
)1
extract
(0.05 U GlDHÆmg dry cells
)1
).InthecaseofGaDH,
before dialysis, the supernatant was enriched as necessary
with commercial enzyme (P. fluorescens gene expressed in
Table 3. Characteristics of free biomimetic and nonbiomimetic dyes, and conditions and performance of their immobilization reactions onto agarose.
Dye-ligand
M
r
(sodium salt)
me
(m
M
)1

Æcm
)1
)
in water
k
max
(nm)
in water
mg dye per g
moist gel
(in reaction)
lmol dye per g
moist gel
(in adsorbent)
me
a
(m
M
)1
Æcm
)1
)
BM1 1119 7.1 628 37.3 1.9 5.3
BM2 1057 6.8 616 35.2 1.8 4.9
C
6
gVBAR 928 5.9 622 30.9 2.1 4.2
C
6
NgVBAR 985 5.8 623 31.4 2.0 4.8

C
3
NgVBAR 943 6.0 618 28.1 2.2 6.3
aVBAR 685 5.5 619 22.8 2.2 4.4
bVBAR 713 5.7 621 23.8 2.3 4.3
cVBAR 741 5.4 622 24.7 2.1 4.5
dVBAR 769 5.8 622 25.6 1.8 4.1
eVBAR 797 5.6 623 26.6 2.0 3.9
fVBAR 825 5.3 620 27.5 1.9 4.2
gVBAR 844 8.0 621 28.1 2.2 6.3
a
Determined in medium identical to the one that resulted after acid hydrolysis of the adsorbent. Values were determined from 20 l
M
dye
solutions made in the above medium. The duration of all reactions was 18 h.
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5397
E. coli) in order to achieve an initial specific activity of
about 1.1 U GaDHÆmg
)1
.
Commercial lyophilized crude powder (10 mg) of Dacty-
lium dendroides was suspended in 2 mL of 100 m
M
potas-
sium phosphate buffer, pH 7.0 or 7.5 and the suspension
was centrifuged (5000 g,20min,4°C). The supernatant
was dialyzed overnight at 4 °C against 2 L of 100 m
M
potassium phosphate buffer, pH 7.0 or 7.5. The dialysate
was clarified through a Millipore cellulose membrane filter

(0.45 lm pore size), affording specific activity, typically,
of 51.3 U GaOÆmg
)1
(18.3 U GaOÆmL
)1
extract, 3.7 U
GaOÆmg cell lyophilized powder
)1
).
Commercial lyophilized crude powder (11 mg) of
Aspergillus niger was suspended in 1.5 mL of 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA,
pH 7.0 or 7.5. The suspension was centrifuged (5000 g,
20 min, 4C) and the supernatant was dialyzed over-
night at 4 °C against 2 L of 10 m
M
potassium
phosphate buffer containing 1 m
M
EDTA, pH 7.0 or
7.5. The dialysate was clarified through a Millipore
cellulose membrane filter (0.45 lm pore size), affording,
typically, 11.2 U GlOÆmL extract
)1
(1.5 U GlOÆmg
solid
)1

).
Commercial baker’s yeast cells (9 g paste) were suspen-
ded in 12mL of 10m
M
potassium phosphate buffer
containing 1 m
M
EDTA, pH 7.0 or 7.5, or 10 m
M
potas-
sium phosphate buffer, pH 6.5, 7.0 or 7.5, before ultra-
sonically disintegrated (amplitude 40%, 5 s sonication, 5 s
pause, 12 cycles, 4 °C).Celldebriswasremovedby
centrifugation (14 000 g,50min,4°C) and the supernatant
was dialyzed overnight at 4 °Cagainst2Lof10m
M
potassium phosphate buffer containing 1 m
M
EDTA,
pH 7.0 or 7.5, or 10 m
M
potassium phosphate buffer,
pH 6.5, 7.0 or 7.5. The dialysate was clarified through a
Milipore cellulose membrane filter (0.45 lm pore size),
affording, typically, an activity of 0.06 U GaDHÆmL
extract
)1
(0.08 U GaDHÆgcellpaste
)1
), 0.39 U GlDHÆmL

extract
)1
(0.52 U GlDHÆg cell paste
)1
)and5.5U
ADHÆmL extract
)1
(7.3 U ADHÆg cell paste
)1
).
Green peas (13 g) were suspended in 20 mL of 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA, 7.0
or 7.5, or 10 m
M
potassium phosphate buffer, pH 6.5, 7.0
or 7.5, before pulped using pestle and mortar, and
homogenized (Virtishear mechanical homogenizer,
10 000 r.p.m., 1 min, 4 °C). The homogenized suspension
was filtered using cheese cloth and the filtrate was
centrifuged (18 000 g,40min,4°C). The supernatant
was dialyzed overnight at 4 °C against 5 L of 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA,
pH 7.0 or 7.5, or 10 m
M

potassium phosphate buffer,
pH 6.5, 7.0 or 7.5. The dialysate was clarified through a
Milipore cellulose membrane filter (0.45 lm pore size),
affording, typically, an activity of 0.02 U GaDHÆmL
extract
)1
(0.03 U GaDHÆg
)1
)and0.3UADHÆmL ex-
tract
)1
(0.46 U ADHÆg
)1
).
Rabbit liver (5 g) was suspended in 20 mL of 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA,
pH 6.5 or 7.0, and homogenized (Virtishear mechanical
homogenizer, 10 000 r.p.m., 3 min, 4 °C). The homogen-
ized suspension was centrifuged (750 g for 15 min, 4 °C)
and the supernatant was re-centrifuged (14 000 g,50min,
4 °C). The supernatant was dialyzed overnight at 4 °C
against 5 L of 10 m
M
potassium phosphate buffer, pH 6.5
or 7.0. The dialysate was clarified through a Milipore
cellulose membrane filter (0.45 lm pore size), affording,
typically, an activity of 0.03 U GaDHÆmL extract

)1
(0.12 U GaDHÆg
)1
).
Affinity chromatography evaluation of the amino-alkyl-
dyes, hydrophilic spacer-dye and control-dyes using
GaDH from
P. fluorescens
extract
All procedures were performed at 4 °C. Galactose dehy-
drogenase binding was assessed using analytical columns,
each packed with 0.5 mL of adsorbent bearing immobilized
ligand (amino-alkyl-VBAR dyes, structures aVBAR-
fVBAR and hydrophilic spacer-VBAR dye, structure
gVBAR of Table 1, as well as control-dyes, structures
C
6
gVBAR, C
6
NgVBAR and C
3
NgVBAR of Table 2) (1.8–
2.3 lmol dyeÆgmoistgel
)1
). Columns were equilibrated
with 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA, pH 7.0. Dialyzed P. fluorescens extract (0.5–

1.0 mL, 0.1–0.2 U GaDH, 0.09–0.17 mg protein) was
applied to each analytical column. Non-adsorbed protein
was washed off with equilibration buffer (2–3 mL). Bound
GaDH activity was eluted with 2 mL equilibration buffer
containing a mixture of 1 m
M
NAD
+
and 10 m
M
Na
2
SO
3
.
Collected fractions (1 mL) were assayed for GaDH activity.
Affinity chromatography evaluation of mimodye
adsorbents using GaDH from
P. fluorescens
extract
All procedures were performed at 4 °C. Galactose dehy-
drogenase binding was assessed using analytical columns,
each packed with 0.5 mL of mimodye adsorbent
(1.8–2.2 lmol dyeÆgmoistgel
)1
). Columns were equili-
brated with 10 m
M
potassium phosphate buffer at the
pHs shown in Table 4, containing 1 m

M
EDTA. Dialyzed
P. fluorescens enriched extract (0.5–1.0 mL, 0.10–0.38 U
GaDH, 0.09–0.33 mg protein) was applied to each analyt-
ical column. Non-adsorbed protein was washed off with
equilibration buffer (2–4 mL). Bound GaDH was eluted,
from immobilized BM1, by a mixture of 0.5 m
M
NAD
+
and 5 m
M
Na
2
SO
3
in the equilibration buffer (2–4 mL) or,
from immobilized BM2, by 0.8 m
M
NAD
+
and 8 m
M
Na
2
SO
3
in the equilibration buffer (3–4 mL). Collected
fractions (1 mL) were assayed for GaDH activity and
protein [41]. The fractions with GaDH activity were pooled

and the specific activity was determined.
Table 4. Affinity chromatography evaluation of immobilized mimodyes
and hydrophilic spacer-VBAR dye for binding GaDH activity from
P. fluorescens crude extract.
Dye-ligand pH
SA
(unitsÆmg
)1
)
Purification
(-fold)
Recovery
(%)
BM1 6.5 29.9 27.2 66
7.0 48.2 41.9 100
7.5 30.3 27.5 35
BM2 6.5 15.1 13.7 68
7.0 37.7 32.8 76
7.5 41.2 37.5 98
8.0 28.8 25.3 28
gVBAR 7.0 15.4 13.4 33
5398 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Affinity chromatography control experiments for the
evaluation of the binding selectivity of mimodye
adsorbents using enzymes other than GaDH
On each of the mimodye adsorbents (0.5 mL), previously
equilibrated with 10 m
M
potassium phosphate buffer
(pH 7.0 for BM1 and 7.5 for BM2) containing 1 m

M
EDTA, were applied the enzyme units shown in Table 5,
previously dialyzed in the same equilibration buffer (4 °C).
After the column was washed with equilibration buffer,
elution of bound proteins was effected with 1
M
KCl in the
same buffer. In the case of GaO, the equilibration buffer
used was 100 m
M
potassium phosphate (pH 7.0 for BM1,
pH 7.5 for BM2).
RESULTS AND DISCUSSION
Protein molecular modelling
Fold recognition results were near-unanimous in highlight-
ing the structure of Zymomonas mobilis glucose-fructose
oxidoreductase (GFO; PDB code 1ofg;) as the best available
template for GaDH model construction. For example, the
3D-PSSM method [16] gave GFO a score of 6 · 10
)6
with
the next best hit scoring 95 · 10
)3
. Similarly, the FFAS
method [17] gave GFO a score of 67 and the next best
template just 14. In each case, these results are strongly
significant for GFO and show it to be much more suitable as
GaDH template than the next best structures. The only
exception to the trend was
GENTHREADER

[15] which gave
GFO and rat biliverdin reductase similar high probabilites
of 0.94 and 0.95, respectively. In fact GFO, rat biliverdin
reductase and many others of the better scoring hits of the
fold recognition studies, all catalyse redox reactions and are
structurally related, sharing a dinucleotide binding fold, in
conjunction with a more variable domain responsible
principally for substrate binding [42].
Based on the FFAS alignment an initial GaDH-GFO
target-template alignment was constructed by examination
of the GFO structure to determine the most likely positions
at which the 10 insertions or deletions could be accommo-
dated. In most cases these positions were between secondary
structure elements but in others, the size of the insertion or
deletion naturally led to alteration of neighbouring helices
or strands (see Fig. 3). Although GaDH is a dimer, the
regions of the alternate subunit corresponding to those that
Table 5. Control experiments for the evaluation of the binding selectivity of immobilized BM1 and BM2 with enzymes other than GaDH. On each
affinity adsorbent (0.5 mL), previously equilibrated with 10 m
M
potassium phosphate buffer containing 1 m
M
EDTA (pH 7.0 for BM1, pH 7.5 for
BM2), were applied the enzyme units shown, previously dialyzed in the same equilibration buffer as above (4 °C). After the adsorbent was washed
with equilibration buffer, elution of bound proteins was effected with 1
M
KCl. For GaO, 100 m
M
potassium phosphate buffer (pH 7.0 for BM1,
pH 7.5 for BM2) was used as the equilibration buffer.

BM1 BM2
Enzyme Source Units applied Bound enzyme (%) Units applied Bound enzyme (%)
GaO Dactylium dendroides 15.7 5.8 13.8 5.2
GlO Aspergillus niger 8.3 6.2 4.6 12.6
GlDH P. fluorescens 0.1 4.9 0.1 4.8
GlDH Baker’s yeast 0.7 0.7 0.8 0.5
ADH Baker’s yeast 3.8 3.3 3.9 2.6
ADH green peas 0.3 19.4 0.3 15.9
Fig. 3.
ALSCRIPT
[52] alignment of GaDH with template GFO. The secondary structure of GFO is shown above the alignment and residues shared
between the two proteins are emboldened.
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5399
in GFO contribute to cofactor and substrate binding (the
N-terminal stretch and the loop around GFO residue 317),
are not present in GaDH (Fig. 3). Therefore, modelling of
an individual monomer was undertaken. The first set of 20
models was, thus, constructed and analysed as outlined in
Materials and methods. At regions of improbable protein
packing and solvent exposure, as indicated by
PROSA
II, a
series of alignment variants was constructed. These variants,
typically, involved 1–3 residues shifts of single secondary
structure elements, with the flanking loops accommodating
correspondingly altering in length. These were analysed and
the process repeated until no further alignment improve-
ments could be found. In all 17 different alignments were
tested. Special attention was then paid to stereochemical
aspects of the model. Residues in disallowed or generously

allowed areas of the Ramachandran plot were treated as
possible errors and dealt with either by flipping of peptide
bonds or ab initio regeneration with
MODELLER
.Attheend
of this process the structure best combining low
PROSA
II
score and good stereochemistry was taken as the final
model.
As previously observed, significant improvements in
model quality resulted from this careful construction
procedure. The first set of models had
PROSA
II scores in
the range )7.8 to )8.7. For the final model this improved to
)10.1. Comparison with the template also suggests a model
of high objective quality. The somewhat longer GFO (351
residues vs. 303 in the final model) scores )11.5 by
PROSA
II
analysis. The overall stereochemical quality of the model
and GFO, as measured by the G-factor calculated by
PROCHECK
, is near-identical; )0.15 for the model, )0.16 for
the crystal structure. The GaDH model places 90.5% of
residues in most-favoured regions of the Ramachandran
plot, suggestive of good structural quality and similar to the
91.5% value of the GFO template. As well as these overall
indicators, it is worth remembering that the isolated regions

of high sequence identity between target and template,
around GaDH positions 10 and 85 (Fig. 3), are situated
near the cofactor binding site. Hence, this part of the final
structure, important for docking studies, should be parti-
cularly well-modelled.
Ligand design and docking
With the good objective quality of the GaDH model
established, docking experiments were initiated to indicate
possible galactosyl-biomimetic ligands for GaDH. The
three ring systems of the CB3GA-derived portion (num-
bered 1–3: anthraquinone, diaminobenzosulfonic acid and
triazine, respectively; see Fig. 4A and Table 1, VBAR) were
first docked into the GaDH model, followed sequentially by
the galactose portion, the ÔlinkerÕ molecule between ring 3
and the galactose, and finally the chain (ÔspacerÕ molecule)
by which the ligand attaches to the chromatograpic matrix
(e.g. agarose beads). Experimental evidence regarding
residues involved in substrate and cofactor binding to
GaDH is entirely lacking, and inference of possible
important regions through their sequence conservation is
rendered impossible by the lack of any known close GaDH
homologues. Nevertheless, a variety of other indirect data
could be used to guide the docking.
The knowledge that ring systems 1–3 bind in NAD(P)
binding sites, with the anthraquinone ring system 1
Fig. 4. Stereo
MOLSCRIPT
[53] diagrams showing interactions of unre-
fined, docked components with the final GaDH model. (A) Ring systems
1–3 (B) galactose (presumed to bind similarly to 2-amino-2-deoxyga-

lactose) and (C) shikimic acid. Hydrogen bonds are shown by dotted
lines.
5400 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
generally occupying the adenine-binding part of the site [25],
was used to help their positioning. This approach is
strengthened from the experimental finding of GaDH rapid
inactivation by VBAR (approx. 95% inactivation after
6 min), a phenomenon that is prevented in the presence of
NAD (approx. 50% inactivation after 6 min). Similar
phenomena have been reported before for other dehydro-
genases and anthraquinone chlorotriazine dyes [11,13]. In
this case, an adenine binding mode similar to that seen in
GFO is unlikely to exist for GaDH, as in the template an N-
terminal extension from another subunit makes key inter-
actions. This N-terminal extension is not present in GaDH
(Fig. 3) but a suitable hydrophobic surface patch (residues
Phe64, Phe46, Leu75, Val7, Gly8) is found in the vicinity
and is presumably involved in binding the adenine portion
of the NAD cofactor of GaDH. The possibility of an
entirely different NAD binding site in GaDH may be
effectively ruled out as it shares with a series of other
enzymes a structurally conserved cofactor binding domain
to which NAD(P) binds similarly in each case [42]. The
anthraquinone ring system was positioned flat against this
hydrophobic surface patch, with its sulfonate group largely
solvent exposed but also forming hydrogen bonds with the
side chain of Thr35 and with the backbone nitrogen of
Gly37 (Fig. 4A). This sulfonate group occupies the site of
the phosphate group of NADP in the template GFO
structure. The absence of more extensive interactions

between the sulfonate group and GaDH is in accord with
the enzyme’s 10-fold greater K
m
value for NADP than for
NAD [1]. The ring system 2 (diaminobenzosulfonate) could
then be favourably positioned to interact hydrophobically
with Trp153, Pro67 and Gly11, with the sulfonate ionically
interacting with Lys12 (Fig. 4A). The sulfonate–lysine
interaction strongly resembles the interaction between the
phosphate of NADP and the corresponding lysine residue
in the template structure. Ring system 3 (triazine) was then
positioned, favourably sandwiched between the side chains
of Trp153 and Ile13 (Fig. 4A). The plane of triazine in this
position was such that the two chains, ÔlinkerÕ and ÔspacerÕ,
joining these ring systems 1–3 to the galactose moiety and to
the chromatographic matrix, respectively, could be added
without leading to steric clashes with the rest of the protein.
Attention was next turned to the 2-amino-2-deoxygalac-
tose portion of the biomimetic, which, as a weak substrate,
cansafelybeassumedtobindinthesamewayassubstrate
galactose. Galactose itself was used for docking experi-
ments. Data obtained from enzymatic characterization of
the enzyme [1] offer some clues as to the binding mode. The
comparable activity of GaDH with fucose as substrate
suggests that the OH-group attached on the C6 carbon
(6-hydroxyl group) of galactose, lacking in fucose, does not
interact strongly with GaDH. The lack of activity against
D
-galacturonic acid and
D

-galactose-6-phosphate suggests
that the protein pocket in the region of the 6-hydroxyl group
is either limited in size, lacks suitable positive residues for
ionic interactions, or both. The four-fold reduced activity of
GaDH with 2-deoxy-
D
-galactose suggests a role for the
2-hydroxyl group in binding of galactose substrate. In order
to determine the general location of the galactose binding
site we used data from the structurally homologous glucose-
6-phosphate dehydrogenase or G6PDH [43] which, while
clearly not the best template for model construction (see
above), catalyses a very similar reaction to GaDH. Infor-
mation regarding substrate binding to G6PDH is indirect,
coming from the effects of site-directed mutagenesis [43,44]
rather than from crystal structures. Nevertheless, the
corresponding region of GaDH contains a deep pocket
(often the site of binding and catalysis [45]); near to
tryptophan residues, which are commonly encountered at
carbohydrate binding sites [46]. As with G6PDH, current
ideas regarding catalytic mechanism imply that a base is
required to withdraw a proton from the C1 position [43]. A
search for possible candidates yielded only Glu259. The
location of this residue in a relatively hydrophobic
environment, surrounded by Cys114, Ile115, Ala118 and
Tyr260, would serve to raise its pKa value and thereby
facilitate its protonation. With these data in mind, favour-
able positions for b-
D
-galactose with its C1 carbon near to

Glu259 were sought, allowing the formation of a hydrogen
bond, for example, between the 1-hydroxyl group and the
side group of Glu259, as shown in Fig. 4B. Furthermore,
the chosen orientation placed the 6-hydroxyl group in a
hydrophobic pocket with no possible hydrogen bond
partners and the 2-hydroxyl group suitably positioned to
hydrogen bond to Tyr260 in accord with the experimental
data [1]. The 3- and 4-hydroxyl groups of the galactose thus
positioned were suitably placed to hydrogen bond to the
side-chain oxygen (OD1) and nitrogen atoms of Asn167 and
Ile115, respectively. One hydrophobic face of the sugar
interacted favourably with the side chain of Trp225 [46].
Little information was available for the prediction of
possible modes of binding of shikimic acid to GaDH.
However, it seemed reasonable to suppose that its six-
membered ring might adopt a similar position to that
predicted for galactose. Superimposing these rings leads to
12 possible positions for shikimic acid. However, the
presence of a carboxylate group attached to the ring proved
to be a powerful filter of possible conformations. In two of
the six possible positions for this acidic group, unfavourable
burial in a hydrophobic pocket (predicted to accommodate
the 6-OH group of galactose) was observed. In two others,
severe steric clashes with Gly241 were made, complicated by
the presence nearby of Glu259. Of the two remaining
carboxylate positions, one, interacting ionically with Lys89,
would be favoured for shikimic acid itself. However, as the
carboxyl group was the projected site of attachment to a
linker region (which would fuse the shikimate and ring 3),
only the remaining carboxylate position would be available

for the design of a biomimetic ligand. With the carboxylate
thus positioned, two conformations of shikimic acid,
superimposed on the predicted galactose conformation,
were available, related by a 180° rotation about the
carboxylate-ring bond. Both of these conformations place
a hydroxyl-bearing ring carbon in the location of the ring
oxygen of galactose leading to steric clashes. Predicted
hydrogen bonds were conspicuously absent for both these
conformations suggesting that shikimic acid might lie less
deeply in the pocket, compared to galactose. When small
translations of around 2 A
˚
out of the pocket were applied,
the steric clashes disappeared and new favourable interac-
tions were predicted, a hydrogen bond in one case and
hydrophobic interactions in the other. These favourable
hydrophobic interactions with Trp140 and Trp225 were
only attainable for one conformation and were of poten-
tially much greater significance than a single H-bond. For
this reason the position shown in Fig. 4C, with these
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5401
tryptophan interactions present, and the H-bond corres-
pondingly absent, was taken as the preferred docked model.
Overall, the binding of shikimic acid seemed less favourable
than that of galactose, as the latter, better accommodated in
the pocket, is sandwiched by predicted interacting residues
(compare Figs. 4B,C).
The respective docked positions of the 3-ring system
(Fig. 4A), galactose (Fig. 4B) and shikimic acid (Fig. 4C)
were used to predict suitable lengths for linkers with which

to join the portions occupying cofactor- and substrate-
binding sites. The measured distance to be spanned from the
3-ring system to 2-amino-2-deoxygalactose was around
9.4 A
˚
, and to shikimic acid around 6.4 A
˚
. Extended linker
distances attaching to the triazine (ring system 3) via a
nitrogen atom and to galactose or shikimate via an amide
bond were 8.7, 7.3 and 6.2 A
˚
,forC
5
-, C
4
-andC
3
-length
linkers (–(CH
2
)
n
-, n ¼ 5,4,3, respectively). Owing to the
broadness of the pocket between cofactor- and substrate-
binding sites, few interactions were predicted between linker
and enzyme so that different chemical characteristics of
different possible linkers were not predicted to strongly
influence affinity. Therefore, on the basis of the above, a C
5

linker was predicted to be suitable for 2-amino-2-deoxy-
galactose, whereas a C
3
linker for shikimate.
Finally, the connection between triazine ring and
chromatographic matrix was considered, in order to
predict an appropriate spacer molecule. As the mouth
of the catalytic cleft broadens near the enzyme surface,
this portion seemed unlikely to adopt a single conforma-
tion. For this reason, and because the large number of
degrees of freedom of an extended chain complicates
modelling, no detailed study of connection interaction
with protein was attempted. However, in contrast to the
mixed environment of the linker region, the path from the
catalytic site to the protein surface is highly hydrophilic,
being lined almost exclusively by charged side chains such
as Glu142, Glu223, Asp222, Lys12, Arg145 and His148.
Hence, a hydrophilic connection was predicted to be
favoured over a hydrophobic one.
Modelling of entire galactosyl-mimodye ligands
Based on the above analyses, entire models of GaDH
bound to BM1 (Figs 5A and 6A) and BM2 (Figs 5B and
6B) were constructed and energy-minimized. Most of the
relevant favourable interactions identified during piecemeal
docking (see above) remain in the energy-minimized final
biomimetic model (Fig. 6). Some additional favourable
interactions appeared after refinement. For example, Lys89
makes hydrophobic contacts with the linker and H-bonds to
O3 of the galactose moiety in the final complex model with
BM1 (Fig. 6A). In the case of BM2, a H-bond with Glu223

appeared during refinement. In both cases Glu88, after
refinement, was no longer suitably positioned to hydrogen
bond the linker nitrogen, but would still produce favourable
electrostatic interactions with the linker (Fig. 6A,B). It is
notable that, while NADP in the template is bound almost
entirely to the cofactor binding domain (not shown), the
mimodye ligands occupy a large cleft between the two
domains (Fig. 5A,B), with multiple contacts to both. The
positioning of the piecemeal-docked components, and hence
the final complete mimodye models, was later found to be in
accord with experimentally obtained data.
Experimental evaluation of the dye adsorbents
In view of the practical applications envisaged for the new
mimodye ligands, we made further comparisons for the
final models with experimental data obtained using
agarose-immobilized ligands. The affinity of an immobi-
lized ligand for the complementary protein is determined
Fig. 5.
MOLSCRIPT
diagrams of the final GaDH model showing (A) the
refined conformation of the biomimetic ligand BM1 and (B) the refined
conformation of the biomimetic ligand BM2. In(A)and(B)1–3areused
for ring systems 1–3 (see text), G for the galactose moiety (seen side
on), S for the shikimate moiety (also seen side on) and * for the end of
the connecting spacer (coming out of the page) which would be cov-
alently bonded to the chromatographic matrix.
5402 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
partly by the characteristics of the ligand per se and partly
by the solid support and coupling chemistry. Studies with
ligands free in solution do not fairly reflect the chemical,

geometrical and steric constrains imposed by the complex
3D solid support environment [6]. Not surprisingly
therefore earlier studies have appeared contradictory
[29,34,50]. Consequently, the experimental approach
adapted here for a practically meaningful evaluation of the
computationally predicted mimodye ligands for GaDH
was that of affinity chromatography, using adsorbents
bearing immobilsed biomimetic and nonbiomimetic/con-
trol ligands and using GaDH from P. fluorescens crude
extracts.
The two new galactosyl-mimodye adsorbents, BM1 and
BM2, at pH values 7.0 and 7.5, respectively, exhibited
complete binding of the applied GaDH, followed by
substantial enzyme purification (41.9- and 37.5-fold,
respectively) and high recovery (100% and 98%, respect-
ively) (Table 4). In contrast, the nonbiomimetic control
adsorbents, bearing immobilized ligands missing the
terminal galactosyl-biomimetic moiety (i.e. the three
control ligands C
6
gVBAR, C
6
NgVBAR C
3
NgVBAR)
have bound only 20% of the applied enzyme. Also,
immobilized ligand gVBAR, missing both the biomimetic
moiety and the ÔlinkerÕ molecule, have bound 33% of the
GaDH applied, and purified the bound enzyme only by
approx. 13-fold. The increased binding ability of gVBAR

for GaDH, compared to the three control ligands, may be
attributed to the decreased steric hindrance caused by the
less complex structure of gVBAR. The above findings
strengthen the view for a true affinity recognition mech-
anism operating between GaDH and immobilized mimo-
dyes. While BM2 is possibly the favoured mimodye for
practical applications, due to its chemical stability (gal-
actosamine vs. shikimic acid), BM1 exhibited the higher
GaDH purifying ability of the two (Table 4), a finding
that can be interpreted in terms of higher selectivity
during the binding process and, hence, higher affinity.
This result is consistent with the greater number of
favourable interactions of 2-amino-2-deoxygalactose
(BM1) with GaDH, compared to shikimic acid (BM2)
(Fig. 4B,C).
The length and nature of the spacer molecule, connecting
the ligand to the chromatographic support, were examined.
The insertion of a spacer molecule is necessary for the
immobilized ligand to overcome problems of steric con-
strains encountered during its interaction with the enzyme.
The enzyme practically does not interact either with plain
VBAR or VBAR bearing a short spacer (n ¼ 2,4). How-
ever, ligand dVBAR bearing as spacer 1,8-diaminooctane
(n ¼ 8) has shown higher binding for GaDH (16%), than
VBAR ligands bearing shorter (n ¼ 0–6; 0.2–8% binding of
GaDH) or longer (n ¼ 10,12; 6% and 11%, respectively,
binding of GaDH) hydrophobic spacers. Shorter spacers
Fig. 6.
MOLSCRIPT
[53] diagrams of the final

GaDH models of the complexes with (A) BM1
and (B) BM2. Hydrogen bonds are shown by
dotted lines.
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5403
are probably inadequate to allow VBAR reach the active
site, while longer hydrophobic spacers (eVBAR, n ¼ 10;
fVBAR, n ¼ 12), in a hydrophilic environment, may shown
unfavourable conformation changes (e.g. folding) [51], thus
failing to establish effective interaction with the enzyme. As
far as the spacer moiety is concerned, undoubtedly the
better binding was obtained with ligand gVBAR (33%
binding of GaDH) which bears a hydrophilic long spacer.
This is in full agreement with the computer visualized
strongly hydrophilic nature (Glu142, Glu223, Asp222,
Lys12, Arg145 and His148) of the passage from the catalytic
site to enzyme surface, through which the spacer molecule
must pass. The optimal length of a hydrophilic spacer
appears to be C
8
N
2
.
The designed immobilized mimodyes were evaluated
for their ability to bind enzymes other than the targeted
one, galactose dehydrogenase. For this purpose we
selected enzyme that recognize sugars (e.g. galactose
oxidase and glucose oxidase) and also enzymes from the
dehydrogenase family (glucose dehydrogenase and alco-
hol dehydrogenase). The binding observed between the
enzymes tested and mimodyes BM1 and BM2 was

generally satisfactory low (Table 5). From the six
enzymes tested, only ADH from green peas showed
some binding (19.4% for BM1 and 15.9% for BM2)
followed by glucose oxidase from A. niger (6.2% for
BM1 and 12.6% for BM2), with all other enzymes
showing negligible binding. Interestingly, while the bind-
ing of each enzyme was approximately the same for both
mimodyes, for glucose oxidase from A. niger, mimodye
BM2 was double effective compared to BM1. The
aforementioned figures are near and below those ob-
tained for the nonbiomimetic control adsorbents with
GaDH. It is not unreasonable to expect that the large
nonmimetic portion of the mimodye ligands may allow
for some interactions to occur with certain enzymes.
Nevertheless, the overall picture of the binding selectivity
for the bulky mimodyes BM1 and BM2 was satisfactory,
after testing against nonbiomimetic control adsorbents
and enzymes other than GaDH.
CONCLUSIONS
P. fluorescens galactose dehydrogenase represented a dif-
ficult molecular modelling and docking case as it lacks
close homologues and the best available template shares
with it only 17% sequence identity. Nevertheless, an
objectively sound protein model for GaDH could be built
using appropriate rigorous modelling methodology. This,
in conjunction with GaDH experimental data and infor-
mation from structural comparisons, enabled the identifi-
cation of both substrate and cofactor binding sites and the
docking of putative mimodye ligand components to the
protein. The final structure of the highest-affinity mimodye

(BM1) shows excellent steric complementarity between
enzyme and ligand and multiple favourable interactions,
thus explaining its fine GaDH-binding performance. The
computationally designed models also explain the subse-
quently observed dependence of GaDH-purifying ability
on the size and chemical nature of the individual
components making up the biomimetic ligands. These
new bifunctional galactosyl-mimodye ligands are expected
to find use for the rapid purification of GaDH from
several sources. Such use has found recently a non
bifunctional galactosyl-biomimetic ligand for galactose
oxidase [54].
REFERENCES
1.Blachnitzky,E.O.,Wengenmayer,F.&Kurz,G.(1974)
D-Galactose dehydrogenase from P. fluorescens: purification,
properties and structure. Eur. J. Biochem. 47, 235–250.
2. Buckel, P. & Zehelein, E. (1981) Expression of P. fluorescens
D
-galactose dehydrogenase in E. coli. Gene 16, 149–159.
3. Sperka, S., Zehelein, E., Fiedler, S., Fischer, S., Sommer, R. &
Buckel, P. (1989) Complete nucleotide sequence of P. fluorescens
D
-galactose dehydrogenase gene. Nucleic Acids Res. 17, 5402–
5415.
4. Maier, E. & Kurz, G. (1982) D-Galactose dehydrogenase from
P. fluorescens. Methods Enzymol. 89, 176–181.
5.Wengenmayer,F.,Ueberschar,K H.,Kurz,G.&Sund,H.
(1973) D-Galactose dehydrogenase from Pseudomonas saccharo-
phila: purification, properties and structure. Eur. J. Biochem. 40,
49–61.

6. Lowe, C.R. (2001) Combinatorial approaches to affinity chro-
matography. Current Opinion Chem Biol. 5, 248–256.
7. Labrou, N.E. & Clonis, Y.D. (1996) Biomimetic-dye affinity
chromatography for the purification of mitochondrial
L
-malate
dehydrogenase from bovine heart. J. Biotechnol. 45, 185–194.
8. Labrou, N.E., Eliopoulos, E. & Clonis, Y.D. (1996) Molecular
modelling for the design of chimaeric biomimetic dye-ligands and
their interaction with bovine heart mitochondrial malate dehy-
drogenase. Biochem. J. 315, 695–703.
9. Labrou, N.E., Eliopoulos, E. & Clonis, Y.D. (1999) Molecular
modelling for the design of a biomimetic chimeric ligand. Appli-
cation to the purification of bovine heart
L
-lactate dehydrogenase.
Biotechn Bioeng. 63, 322–332.
10. Melissis, S.C., Rigden, D.J. & Clonis, Y.D. (2001) New family of
glutathionyl-biomimetic ligands for affinity chromatography of
glutathione-recognising enzymes. J. Chromatogr. A 917, 29–42.
11. Clonis, Y.D. & Lowe, C.R. (1980) Triazine dyes, a new class of
affinity labels for nucleotide-dependent enzymes. Biochem. J. 191,
247–251.
12. Clonis, Y.D., Goldfinch, M.J. & Lowe, C.R. (1981) The interac-
tion of yeast hexokinase with Procion Green H-4G. Biochem. J.
197, 203–211.
13. Labrou, N.E., Eliopoulos, E. & Clonis, Y.D. (1996) Dye-affinity
labelling of bovine heart mitochondrial malate dehydrogenase and
study of the NADH-binding site. Biochem. J. 315, 687–693.
14. Karplus, K., Barrett, C. & Hughey, R. (1998) Hidden Markov

models for detecting remote protein homologies. Bioinformatics
14, 846–856.
15. Jones, D.T. (1999) GenTHREADER: an efficient and reliable
protein fold recognition method for genomic sequences. J. Mol
Biol. 287, 797–815.
16. Kelley, L.A., MacCallum, R.M. & Sternberg, M.J.E. (2000)
Enhanced genome annotation using structural profiles in the
program 3D-PSSM. J. Mol. Biol. 299, 501–522.
17. Rychlewski, L., Jaroszewski, L., Li, W. & Godzik, A. (2000)
Comparison of sequence profiles. Strategies for structural predic-
tions using sequence information. Protein Sci. 9, 232–241.
18. Rigden, D.J., Bagyan, I., Lamani, E., Setlow, P. & Jedrzejas, M.J.
(2001) A cofactor-dependent phosphoglycerate mutase homo-
logue from Bacillus species is actually a broad specificity acid
phosphatase. Protein Sci. 10, 1835–1846.
19. Rigden, D.J., Monteiro, A.C. & Grossi de Sa
´
, M.F. (2001) The
protease inhibitor chagasin of Trypanosoma cruzi adopts an
immunoglobulin-type fold and probably represents a consequence
of host to parasite horizontal gene transfer. FEBS Lett. 504,
41–44.
5404 C. F. Mazitsos et al. (Eur. J. Biochem. 269) Ó FEBS 2002
20. Sali, A. & Blundell, T.L. (1993) Comparative protein modelling by
satisfaction of spatial restraints. J. Mol. Biol. 234, 779–815.
21. Sippl, M.J. (1993) Recognition of errors in three-dimensional
structures of proteins. Proteins 17, 355–362.
22. Jones, T.A., Zou, J Y., Cowan, S.W. & Kjeldgaard, M. (1991)
Improved methods for building protein models in electron density
maps and the location of errors in these models. Acta. Crystallogr.

A47, 110–119.
23. Holm, L. & Sander, C. (1996) Mapping the protein universe.
Science 273, 595–602.
24. Frishman, D. & Argos, P. (1995) Knowledge-based protein sec-
ondary structure assignment. Proteins 23, 566–579.
25. Biellmann, J F., Samama, J P., Bra
¨
nden,C.I.&Eklund,H.
(1979) X-ray studies of the binding of Cibacron blue F3GA to liver
alcohol dehydrogenase. Eur. J. Biochem. 102, 107–110.
26. Kleywegt, G.J. & Jones, T.A. (1998) Databases in protein crys-
tallography. Acta Crystallogr. D54, 1119–1131.
27. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros,
P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M.,
Pannu, N.S., Read, R.J., Rice, L.M., Simonson, T. & Warren,
G.L. (1998) Crystallography & NMR system: a new software suite
for macromolecular structure determination. Acta Crystallogr.
D54, 905–921.
28. Kleywegt, G.J. & Jones, T.A. (1997) Model-building and refine-
ment practice. Methods Enzymol 277, 208–230.
29. Labrou, N.E. & Clonis, Y.D. (1995) The interaction of Candida
boidinii formate dehydrogenase with a new family of chimeric
biomimetic dye-ligands. Arch. Biochem. Biophys. 316, 169–178.
30. Cuatrecasas, P. & Anfinsen, C.B. (1971) Affinity chromatography.
Methods Enzymol 22, 345–378.
31. Cuatrecasas, P. (1970) Protein purification by affinity chroma-
tography. Derivatizations of agarose and polyacrylamide beads.
J. Biol. Chem. 245, 3059–3065.
32. Partridge, S.M. (1946) Application of the paper chromatogram to
the quantitative analysis of reducing sugars. Nature 158, 270–271.

33. Bethell, G.S., Ayers, J.S., Hancock, W.S. & Hearn, M.T.W. (1979)
A novel method of activation of cross-linked agaroses with 1,1¢-
carbonyldiimidazole which gives a matrix for affinity chroma-
tography devoid of additional charged groups. J. Biol. Chem. 254,
2572–2574.
34. Labrou,N.E.,Karagouni,A.&Clonis,Y.D.(1995)Biomimetic-
dye affinity adsorbents for enzyme purification: application to the
one-step purification of Candida boidinii formate dehydrogenase.
Biotechnol. Bioeng. 48, 278–288.
35. Wallenfels, K. & Kurz, G. (1962) Uber die spezifitat der
galaktosedehydrogenase aus Pseudomonas saccharophila und
deren anwendung als analytisches hilfsmittel. Biochem. Z. 335,
559–572.
36. Bergmeyer, H.U. (1988) Galactose oxidase from Polyporus cir-
cinatus.InMethods of Enzymatic Analysis (Bergmeyer, H.U.,
Bergmeyer, J. & Grabl, M., eds), Vol. II, 3rd edn, pp. 194–195.
VCH, Germany.
37. Bergmeyer, H.U., Gawehn, K. & Grassl, M. (1974) Methods of
Enzymatic Analysis (Bergmeyer,H.U.,eds),Vol.I,2ndedn,
pp. 457. Academic Press Inc, New York, NY.
38. Strecker, H.J. (1955) Glucose dehydrogenase from liver. Methods
Enzymol. 1, 335–339.
39. Vallee, B.L. & Hoch, F.L. (1955) Zinc, a component of
yeast alcohol dehydrogenase. Proc. Natl Acad. Sci. USA 41,
327–330.
40. Bradford, M.M. (1976) A rapid and sensitive method for the
quantitation of microgram quantities of protein utilizing the
principle of protein-dye binding. Anal. Biochem. 72, 248–254.
41. Bearden, J.C. Jr (1978) Quantitation of submicrogram quantities
of protein by an improved protein-dye binding assay. Biochim.

Biophys. Acta 533, 525–529.
42. Bellamacina, C.R. (1996) The nicotinamide dinucleotide binding
motif: a comparison of nucleotide binding proteins. FASEB J. 10,
1257–1269.
43. Rowland,P.,Basak,A.K.,Gover,S.,Levy,H.R.&Adams,M.J.
(1994) The three-dimensional structure of glucose 6-phosphate
dehydrogenase from Leuconostoc mesenteroides refined at 2.0
Angstrom resolution. Structure 2, 1073–1087.
44. Vought, V., Ciccone, T., Davino, M.H., Fairbairn, L., Lin, Y.,
Cosgrove, M.S., Adams, M.J. & Levy, H.R. (2000) Delineation of
the roles of amino acids involved in the catalytic functions of
Leuconostoc mesenteroides glucose 6-phosphate dehydrogenase.
Biochemistry 39, 15012–15021.
45. Laskowski, R.A., Luscombe, N.M., Swindells, M.B. & Thornton,
J.M. (1996) Protein clefts in molecular recognition and function.
Protein Sci. 5, 2438–2452.
46. Samanta, U. & Chakrabarti, P. (2001) Assessing the role of
tryptophan residues in the binding site. Protein Eng. 14, 7–15.
47. Liu, Y.C. & Stellwagen, E. (1986) Zonal chromatographic analysis
of the interaction of alcohol dehydrogenase with blue-sepharose.
J. Chromatogr. 376, 149–155.
48. Liu, Y.C. & Stellwagen, E. (1987) Accessibility and multivalency
of immobilized Cibacron blue F3GA. J. Biol. Chem. 262, 583–588.
49. Prestera, T., Prochaska, H.J. & Talalay, P. (1992) Inhibition of
NAD (P): (quinone-acceptor) oxidoreductase by Cibacron Blue
and related anthraquinone dyes: a structure-activity study. Bio-
chemistry 31, 824–833.
50. Burton, S.J., Stead, C.V. & Lowe, C.R. (1988) Design and
applications of biomimetic anthraquinone dyes; II. The interac-
tion of C.I. reactive blue 2 analogues bearing terminal ring mod-

ifications with horse liver alcohol dehydrogenase. J. Chromatogr.
A 455, 201–216.
51. Lowe, C.R. (1977) The synthesis of several 8-substituted deriva-
tives of adenosine 5¢-monophosphate to study the effect of the
nature of the spacer arm in affinity chromatography. Eur. J.
Biochem. 73, 265–274.
52. Barton, G.J. (1993) ALSCRIPT: a tool to format multiple
sequence alignments. Protein Eng. 6, 37–40.
53. Kraulis, P.J. (1991) MOLSCRIPT: a program to produce both
detailed and schematic plots of protein structures. J. Appl. Crys-
tallogr. 24, 946–950.
54. Mazitsos, C.F., Rigden, D.J., Tsoungas, P.G. & Clonis, Y.D.
(2002) Galactosyl-biomimetic dye-ligands for the purification of
Dactylium dendroides galactose oxidase. J. Chromatogr. A. 954,
137–150.
Ó FEBS 2002 Galactosyl-mimodyes for galactose dehydrogenase (Eur. J. Biochem. 269) 5405