Báo cáo khoa học: Characterization of the flavin association in hexose oxidase from Chondrus crispus pot

Characterization of the flavin association in hexose
oxidase from Chondrus crispus
Thomas Rand
1
, Karsten B. Qvist
1
, Clive P. Walter
1
and Charlotte H. Poulsen
1,2
1 Danisco A ⁄ S, Brabrand, Denmark
2 Interdisciplinary Nanoscience Center (iNANO), University of Aarhus, Denmark
Hexose oxidase (HOX) catalyses the oxidation of a
variety of hexose sugars with concomitant reduction of
molecular oxygen to hydrogen peroxide. The oxidation
product is the corresponding lactone, which subse-
quently hydrolyses to the respective aldobionic acid.
The reaction is exemplified using glucose below:
d-glucose + O
2
fi d-d-gluconolactone + H
2
O
2
d-d-gluconolactone + H
2
O fi d-gluconic acid
The structural gene of HOX from the marine algae
Chondrus crispus has previously been expressed suc-
cessfully in both Pichia pastoris [1] and Hansenula
polymorpha [2]. Purified enzyme from both of these

expression systems has been shown to be active and to
contain FAD [1,2]. The FAD in HOX from C. crispus
has been shown to be intricately involved in the cata-
lytic mechanism of the enzyme [3], but the mechanism
Keywords
FAD; flavin; hexose oxidase; mass
spectroscopy; modeling
Correspondence
T. Rand, Danisco A ⁄ S, Edwin Rahrs Vej 38,
8220 Brabrand, Denmark
Fax: +45 8625 1077
Tel: +45 8943 5000
E-mail:
(Received 20 January 2006, revised 17 April
2006, accepted 20 April 2006)
doi:10.1111/j.1742-4658.2006.05285.x
Hexose oxidase (EC 1.1.3.5) from Hansenula polymorpha was found to
exhibit a dual covalent association of FAD with His79 via an 8a-histidyl
linkage as well as a covalent association between Cys138 and C-6 of the
isoalloxazine moiety of FAD. Spectral properties of the wild-type enzyme
exhibited maxima at 364 nm and 437 nm as well as a distinct shoulder at
445 nm. An H79K mutant enzyme exhibited only one maximum at
437 nm. The difference absorption spectrum between an oxidized and a
substrate-reduced enzyme preparation showed maxima at 360 nm and
445 nm corresponding to an apparent novel type of association. Hexose
oxidase showed a low, pH-independent fluorescence at 525 nm when exci-
ted at 450 nm. Flavin was released from the holoenzyme by treatment with
trypsin. Sequencing of the flavopeptide revealed two peptides comprising
positions 74–91 and 132–157 associated with FAD in equimolar amounts.
A homology model of hexose oxidase was constructed using the crystal

structure of glucooligosaccharide oxidase from Acremonium strictum as
template. The model placed both of the sequences found above in the close
vicinity of the FAD cofactor, and suggests covalent bonds between both
His79 and Cys138 and FAD, in accordance with the chemical evidence.
Based on the results, hexose oxidase is identified as incorporating FAD
with a double covalent association with His79 and Cys138 in the holo-
enzyme. A reaction mechanism involving the concerted action of Tyr488
and Asp409 in hexose oxidase is suggested as the initiator of the proton
abstraction from the substrate molecule in the active site.
Abbreviations
ABTS, 2,2¢-azinobis(3-ethylbenzo-6-thiazolinesulfonic acid; COX, choline oxidase; Fl(), position on the isoalloxazine ring; GOOX,
glucooligosaccharide oxidase; HOX, hexose oxidase; MeCN, acetonitrile; MSOX, monomeric sarcosine oxidase; MTOX, N-methyltryptophan
oxidase; PCMH, p-cresol methylhydroxylase; TADH, thiamine dehydrogenase; TMDH, trimethylamine dehydrogenase; VAO, vanillyl alcohol
oxidase.
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2693
of this action is not known. The main reason for this
is the lack of a crystal structure and the lack of infor-
mation regarding the association and placement of
FAD in the holoenzyme. FAD in HOX has been
shown to be covalently associated by several methods,
including acid precipitation (unpublished results) and
illuminating an acid-fixated denatured protein in
SDS ⁄ PAGE by UV light. In this case, a fluorescent
band was observed [4]. The oligomer substrates of
HOX include the b-d-forms of lactose, maltose, cello-
biose and maltotriose [5], as well as glucooligosaccha-
rides with up to seven glucose residues when
immobilized on a solid support [6]. In addition, HOX
has been shown to have a high affinity for the mono-
mer sugars glucose and galactose [5].

According to the Pfam database (ger.
ac.uk/Software/Pfam/), HOX contains an FAD-bind-
ing 4 domain, which places it in the p-cresol methyl-
hydroxylase (PCMH) FAD-binding group. This group
of enzymes is named after PCMH, the first enzyme in
the family to be crystallized [7]. These enzymes are typ-
ically composed of two major domains, one of which
binds the FAD (the F-domain) and one of which binds
the substrate (the S-domain). A recently published
novel type of flavin association found in glucooligosac-
charide oxidase (GOOX) from Acremonium strictum
shows FAD to have a double covalent association with
both a Cys and a His residue [8]. With a sequence
identity of about 24%, this enzyme has the highest
identity with HOX from C. crispus, among enzymes
with an experimentally determined three-dimensional
structure. HOX shows a somewhat peculiar spectro-
scopic profile and, based on the sequence homology to
GOOX, we set out to investigate whether wild-type
(WT) HOX, expressed recombinantly in H. polymor-
pha, also contains a His–FAD–Cys association, and if
that was the case, to develop a method of characteriz-
ing this type of association without having a crystal
structure. A homology model of the active site of the
enzyme is reported, based on the GOOX structure,
and an H. polymorpha HOX variant with a mutation
of His79 to Lys (H79K) is analysed and compared to
the WT enzyme. The biochemical evidence for the
FAD association in combination with the model will
be used to make deductions on the specific reaction

mechanism of HOX.
Results
Absorbance properties of the flavin component
of HOX
On initial viewing, the absorbance spectrum of oxidized
HOX exhibits a profile similar to that of other flavinyl-
ated proteins (Fig. 1) [9,10]. Using an enzyme concen-
tration of 0.005 mm, a first absorbance maximum is
observed at 364 nm (0.119) and a second at 437 nm
(0.123). Furthermore, the spectrum shows a distinct
shoulder at 450 nm (0.122). The spectrum did not
change significantly after incubation in 0.3% sodium
dodecyl sulfate, with 1 mm dithiothreitol (data not
shown). The molecular mass of HOX has been deter-
mined by gel filtration to be between 110 000 [1] and
130 000 Da [11]. cDNA was previously isolated from
HOX, corresponding to a polypeptide of  62 000 Da,
which was confirmed by SDS ⁄ PAGE [1]. Assuming a
homodimeric structure for HOX, these observations
suggest a 1 : 2 enzyme ⁄ FAD ratio in the WT enzyme.
0
5
10
15
20
A
1
1
2
3

2
3
310
360 410 460
Wavelength (nm)
ε (m
M
-1
cm
-1
)
B
310 360 410 460
ε (m
M
-1
cm
-1
)
0
5
10
15
20
Wavelength (nm)
C
310 360 410 460
-10
-5
0

5
10
Wavelength (nm)
HOX
ox
-HOX
red
Fig. 1. Absorbance properties of the covalently bound flavin in hexose oxidase (HOX). Absorption spectra were recorded in 0.1 M sodium
phosphate buffer, pH 6.3. Enzyme concentrations were standardized to 0.005 m
M according to a 1 : 2 enzyme ⁄ FAD stoichiometry. For other
details, see Experimental procedures. (A) (1) The spectrum of the wild-type (WT) oxidized HOX. (2) The spectrum of the H79K mutant HOX.
(3) The absorbance spectrum of an equimolar amount of free FAD (0.01 m
M) is shown in comparison. (B) (1) The spectrum of the WT HOX
anaerobically substrate-reduced enzyme using 2.5 m
M glucose. (2) The spectrum of the WT HOX aerobically substrate-reduced enzyme
using 2.5 m
M glucose. (3) The spectrum of the WT oxidized HOX. (C) The difference spectrum between WT HOX before and after sub-
strate-mediated reduction (curve 3–2 in (B)). The second absorption maximum of the difference spectrum is hypsochromically shifted in rela-
tion to free FAD. The peak maximum is observed at 360 nm, indicating covalent FAD in the holoenzyme.
Characterization of flavin in hexose oxidase T. Rand et al.
2694 FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO
This stoichiometry in HOX is further substantiated by
reports of extinction coefficients (e
445
–e
448
) for 8a- and
6-substituted flavins in the range 11.3–12.3 mm
)1
Æcm

)1
[12]. Assuming a 1 : 2 enzyme ⁄ FAD ratio, the extinc-
tion coefficients, with respect to the covalently bound
flavins in HOX, are e
364
¼ 11.9 mm
)1
Æcm
)1
, e
437
¼
12.3 mm
)1
Æcm
)1
and e
450
¼ 12.2 mm
)1
Æcm
)1
.
When HOX is reduced with substrate, its spectrum
changes significantly. The resulting reduced spectrum
exhibits a maximum at 420 nm (pH 6.3), possibly due
to the bond between the Fl(6)C atom and a Cys resi-
due, as observed for trimethylamine dehydrogenase
(TMDH) [13,14]. The apparent lack of a semiquinone
species upon substrate reduction is similar to what has

been observed for VAO [15]. Such an intermediate
probably occurs unobserved in HOX, as is common
with oxidase- and monoxygenase-type flavoenzymes
[16]. The possible contribution of the Fl(8a)C–His
bond to the spectrum of the flavin component is visu-
alized by the difference spectrum between the oxidized
and the reduced enzyme (HOX
ox
–HOX
red
), in the same
way as done previously [3]. This spectrum shows two
distinct maxima at 360 nm and at 450 nm. The second
absorption maximum is clearly more intense than
the first. The difference spectrum displays a hypso-
chromic shift of the first absorption peak in compar-
ison to free FAD of about 12 nm to 360 nm. This shift
is indicative of a modification at the Fl(8a)C atom,
although quantitatively different from the shifts previ-
ously observed in other 8a-bound flavin species [12].
The absorbance spectrum of a mutant enzyme (H79K),
in which the putative FAD-binding His had been chan-
ged to a Lys residue, showed only a single absorbance
peak at 437 nm. The fact that the near-UV flavin band
was missing from this protein can be explained by the
single bond to a cysteine residue at the Fl(6) position.
This behaviour was previously reported for TMDH
[14] and other 6-S-substituted flavins [17].
Fluorescence properties of the flavin component
in HOX

Surprisingly HOX fluorescence quantum yield was sig-
nificantly less than the fluorescence of choline oxidase
(COX) from Arthrobacter globiformis, which contains
an 8a-N1-histidyl FAD (see Fig. 2) [18]. When excited
at 450 nm at pH 8, both HOX and COX showed a
distinct fluorescence emission peak at 525 nm. The
fluorescence was considerably lower than that of free
FAD (about a factor of 3) for all the species measured.
The WT forms of both HOX and COX were buffered
in solutions of decreasing pH. Whereas the fluores-
cence emission of COX was clearly higher at lower pH
values, the fluorescence emission of HOX was largely
pH independent. An equimolar amount of purified
flavopeptide released from the holoenzyme using tryp-
sin showed the same distinct, but low, fluorescence
emission as HOX and COX, but was pH independent,
like HOX. HOX was treated with 6 m HCl (see
Experimental procedures) under conditions that were
previously shown to preserve both the thioether
Fl(6)C–S bond of TMDH [14] and the Fl(8a)C–N1
bond of thiamine dehydrogenase (TADH) [19] and
COX [18]. The fluorescence was measured on the
resulting hydrolysate. The HOX hydrolysate fluores-
cence was shown to increase about three-fold, similar
to that of COX when the pH was reduced to 3. The
H79K mutant showed no measurable fluorescence.
Kinetic parameters
The kinetic parameters determined with d-glucose as
substrate for the WT HOX were as follows: K
m

¼
4.2 mm, V
max
¼ 140 mU. These are similar to those
reported for the native enzyme [5,11]. The H79K
mutant was not active. Furthermore according to the
0
20
40
60
80
100
120
2468
pH
Relative Fluorescence (%)
Fig. 2. pH-dependent fluorescence emission at 525 nm recorded
using an excitation wavelength of 445 nm. Concentration with
respect to flavin was 0.06 l
M in all samples. Wild type (WT)
hexose oxidase (HOX) from H. polymorpha (o) shows low,
pH-independent fluorescence emission at 525 nm, as does a
trypsin-released flavopeptide from HOX (), contrary to COX from
A. globiformis (D), which shows a clear pH dependence with an
apparent pK
a
of 5.8. Flavin released from WT HOX by acid hydro-
lysis (o) shows the same pH dependence (pK
a
5.8) as COX. At

pH 8, each sample showed a similar, low, but distinct fluorescence
emission maximum at 525 nm, when excited at 445 nm.
T. Rand et al. Characterization of flavin in hexose oxidase
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2695
apparent extinction coefficient of the flavin in the
mutant enzyme (e
437
¼ 5.5 mm
)1
Æcm
)1
), only half of
the flavin is incorporated in the mutant holoenzyme as
compared with the WT enzyme. A unit (U) is defined
as the production of 1 lmol H
2
O
2
Æmin
)1
at 25 °C,
pH 6.3 [3], according to the reaction catalysed by
HOX as mentioned above.
Identification of the tryptic flavin–peptide
complex
A flavopeptide was isolated, based on the absorbance
at 450 nm, by RP-HPLC in three sequential steps (see
Experimental procedures). A fraction of the purified
flavopeptide complex was sequenced using standard
Edman sequencing(Figs 3 and 4). The sequencing of

the peptide fraction of the complex yielded two separ-
ate sequences corresponding to positions 74–91 and
132–157 in the primary sequence of HOX. There
were two deviations, however: Cycle 6 of the Edman
degradation yielded no His, indicative of a modifi-
cation of His79. Cycle 7 yielded only one equivalent of
carbamidomethylated Cys, implying that either Cys80
or Cys138 was chemically different from the carbami-
domethylated phenylhydantoin standard. These find-
ings, together with the fact that His79 and Cys80 are
in the same peptide, suggest that Cys80 was alkylated
and Cys138 is covalently associated with FAD.
Mass determination of the peptide–flavin complex
was carried out simultaneously with the chromatogra-
phy steps using an LC-MS ⁄ MS setup (see Experimen-
tal procedures). Flavopeptide-related ions in the MS
spectrum were observed at m ⁄ z 893.57, 1071.81 and
1339.49 (Fig. 4A). Indicative water loss (data not
shown) predicted that these signals would correspond
to the (M + 6H)
6+
(M + 5H)
5+
and (M + 4H)
4+
ions of the flavin–peptide complex, where M is the
mass of the neutral complex with two carbamidome-
thylations (see Fig. 3B). The singly charged ion spe-
cies (M +H)
+

was confirmed using a MALDI ion
source with a TOF analyser. Two signals that were
clearly related, based on their resolution pattern, were
Putative covalent
attachment site
Pos H79
N
A
C
M 1 K 5 46
I74 VSG G
H
CYED FVFD ECV K91
V132 LPG GS
C
YSVG LGG HIVGGG DGILA R15 7
B
N
10
N
5
9
8
8α
7α
6
4
N H
3
2

N
1
O
C H
3
O
H
2
C
N
N
O
C
H
2
C
H
C C
H
C
H
2
O P
O
O
OH
P
O
OH
O C

H
2
N
N
N H
2
OH
OH
OH
OH
OH
I74 VSGG H C*YE DFVF DEC*V K91
V132 LPG GS
C YSV GLG GHI VGG GDGI LA R157
Putative covalent
attachment site
Pos C138
Fig. 3. (A) Overview of the peptides sequenced from the isolated flavin complex. Edman degradation of a purified trypsin-released flavin
assigned two peptide sequences to the complex. The peptides were copurified with the flavin moiety through three consecutive gradients.
The first peptide is assigned to I74–K91 of the primary sequence. No His was detected in cycle 6. The second peptide stems from V132–
R157 in the primary sequence. Only one carbamidomethylated Cys could be detected in cycle 7. (B) The deduced identity of the peptide
complex linked by FAD. Cysteines marked by an asterix are modified by a carbamidomethyl group. The theoretical mass of the neutral com-
plex is 5354.129.
Characterization of flavin in hexose oxidase T. Rand et al.
2696 FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO
observed at m ⁄ z 5008.52 and 5354.05 (Fig. 4B). The
deduced mass of the singly charged ion from
the LC-MS ⁄ MS was calculated to be 5355.48 Da.
The mass discrepancy of 1.43 Da is attributed to the
relatively poor resolution of the signals. The mass

observed at 5008.52 Da corresponds to the flavopep-
tide complex with the loss of AMP () 347 Da).
MS ⁄ MS of the 893.57 and the 1071.81 signals yielded
major fragment ions of m ⁄ z 1001.78 and 1252.53,
respectively (data not shown) corresponding to
(M-AMP + 5H)
5+
and (M-AMP + 4H)
4+
. The
deduced m ⁄ z of the singly charged ion species
(M-AMP + H)
+
is 5006.01; a mass discrepancy of
2.51 Da is observed in this case, again attributable to
the poor signal resolution. MS ⁄ MS of the signals
observed at m ⁄ z 893.57 and 1071.81 both yielded an
10
20 30 40
50
60
70
Time [min]
0
10
20
30
A
Intens.
mAU

00010000.D: UV Chromatogram, 215 nm
00010000.D: UV Chromatogram, 450 nm
1071.72
1339.26
2073.52
893.57
+MS, 40.7min #616
0.0
0.5
1.0
1.5
2.0
2.5
x10
6
Intens.
200 400 600 800 1000 1200 1400 1600 1800 2000 m/z
879.98
1070.45
B
1200
1000
800
600
400
200
m/z
815.8005-
3005.1735
840.4535-

N
N
O
OP
O
OH
O C
H
2
N
N
NH
2
OH
OH
Fig. 4. Determining the mass of the pep-
tide–flavin complex isolated from a hexose
oxidase (HOX) trypsin digest by RP-HPLC.
(A) The LC-MS chromatogram of a trypsin
digest of HOX. The absorbance was con-
tinually monitored at 215 nm (red) and
450 nm (green). The trypsin-released FAD
elutes from 40 to 46 min. MS was per-
formed on the most intense peak at
40.7 min. The signals observed at m ⁄ z
893.57, 1071.81 and 1339.49 correspond to
the (M + 6H)
6+
(M + 5H)
5+

and the
(M + 4H)
4+
ions of the peptide–FAD com-
plex. (B) Part of the MALDI mass spectrum
of a purified fraction of the complex show-
ing signals at m ⁄ z 5354.05 and 5008.52.
The signals correspond to the singly
charged species of the complex (M + H)
1+
.
The difference between the two peaks
corresponds roughly to the loss of AMP
(347 Da), as shown in the insert.
T. Rand et al. Characterization of flavin in hexose oxidase
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2697
intense signal at m ⁄ z 347.96 corresponding to singly
charged AMP (348 Da), confirming that both of these
ions are FAD related.
Taken together with the Edman degradation results,
the MS data suggest that the isolated flavo component
is the peptide–FAD–peptide complex shown in
Fig. 3B. The neutral theoretical mass of this complex
is calculated to be 5354.13 Da.
Homology modelling
The model of HOX accounts for the first 451 of the
456 residues, without any gaps, and has a root
mean square (RMS) value, relative to the template,
GOOX, of 0.92 A
˚

, based on 388 Ca-atoms. Accord-
ing to the model, the N-1 atom of His79 is located
2.4 A
˚
from the 8a C atom of the isoalloxazine ring
in FAD, and the S atom of Cys138 is located 1.8 A
˚
from the C-6 atom. The distance between possible
covalent attachment points in FAD [10] and other
His, Cys or Tyr residues in HOX were in all other
cases larger than 4.5 A
˚
, suggesting covalent bonds
between both His79 and Cys138 and the FAD, in
accordance with the chemical evidence shown above.
To explore the possibility that dual covalent binding
of FAD may be common, although unnoticed until
very recently [8], we made a multiple sequence align-
ment of HOX, GOOX and the 50 sequences most
closely related to GOOX (see Experimental proce-
dures). In this alignment, 42 of the 50 GOOX
homologues had a conserved His and Cys, at posi-
tions 70 and 130, respectively (GOOX numbering),
which suggests that dual covalent binding through a
His and a Cys is not uncommon. To illustrate the
point, Fig. 5 shows the alignment of a subset of the
42 sequences, thinned such that no two sequences
are more than 70% identical, and with the sequences
of GOOX and HOX added.
Since interaction between amino acids and the pyro-

phosphate of FAD usually plays an important role in
anchoring FAD, amino acids with atoms near the O3P
atom of pyrophosphate were determined, and Val75,
Ser76, Gly77, Gly78, His79, Cys80, Val85, Val141,
Gly142, Gly144, Gly145 and His146 were found within
a distance of 6 A
˚
. In this list, the residues appearing in
italics are structurally aligned with the seven residues
found to interact with pyrophosphate in GOOX [8],
which strongly suggests that FAD is anchored in HOX
in much the same way as it is in GOOX (Fig. 6). For
Fig. 5. Truncated multiple sequence alignment of hexose oxidase (HOX), glucooligosaccharide oxidase (GOOX), and nine GOOX homologues
found and selected as described in the text. Covalent binding of His and Cys is indicated with red stars. P93762, HXO, Chondrus crispus;
Q5Z7I6, putative CPRD2, Oryza sativa; Q9AYM8, CPRD2 protein, Vigna unguiculata; Q9SVG7, hypothetical protein F21C20.150, Arabidopsis
thaliana; Q9ZPP5, berberine bridge enzyme, Berberis stolonifera; Q9SVG3, reticuline oxidase-like protein, Arabidopsis thaliana; O64743, puta-
tive berberine bridge enzyme, Arabidopsis thaliana; Q4HVL0, hypothetical protein, Gibberella zeae PH-1; Q9SA86, T5I8.16 protein, Arabidop-
sis thaliana; Q6PW77, GOOX, Acremonium strictum; Q8GTB6, tetrahydrocannabinolic acid synthase, Cannabis sativa.
Fig. 6. Residues 75–80 and 85 (blue) and 141–142 and 144–146
(red) in hexose oxidase (HOX), all having atoms within 6 A
˚
of the
O3P atom of FAD (indicated). Also shown are covalent links
between the isoalloxazine ring of FAD (yellow) and His79 and
Cys138 (green). The figure is based on a homology model of HOX,
using glucooligosaccharide oxidase (GOOX) as a template, and gen-
erated as described in the text.
Characterization of flavin in hexose oxidase T. Rand et al.
2698 FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO
GOOX, a catalytic mechanism, which involves Tyr429

acting as a general base, and Asp355 facilitating the
necessary abstraction of a proton from Tyr429, was
suggested [8], and used to explain the optimum pH of
about 10. Interestingly, in the HOX model, Tyr488
and Asp409 are structurally aligned with Tyr429 and
Asp355 in GOOX, thus suggesting a similar reaction
mechanism in the two enzymes.
Discussion
WT HOX expressed in H. polymorpha appears to con-
tain the same rare dual covalent flavin association in
the holoenzyme as the recently published GOOX from
A. strictum [8]. Based on the GOOX crystal structure
and our understanding of the reaction mechanism of
this enzyme, it has been possible to make analogous
observations on HOX as detailed in Results. In this
section a series of procedures will be highlighted that
may be generally useful in identifying this type of fla-
vin association in enzymes.
Spectral properties
As the Cys–FAD–His binding mode is novel, few spec-
tral properties have been reported to date. Summing
up some of these characteristics for HOX yields the
following. The absorbance spectrum of WT HOX
exhibits maxima at 364 and 437 nm. The extinction
coefficients with respect to flavin are 11.9 mm
)1
Æcm
)1
and 12.3 mm
)1

Æcm
)1
, respectively (Fig. 1). The differ-
ence spectrum between an oxidized and a substrate-
reduced enzyme shows maxima at 360 nm and
450 nm. The hypsochromic shift of the first absorption
maximum is less than that of other 8a-histidine-bound
flavoproteins, i.e. COX [18], VAO [15], and the shift is
larger than that observed for the 8a-cysteine-bound
flavoproteins, i.e. monoamine oxidases A and B
(MAO A and B) [20,21], monomeric sarcosine oxidase
(MSOX) and N-methyltryptophan oxidase (MTOX)
[22]. Flavoproteins with FAD bound to cysteine at the
Fl(6) carbon have a very different absorbance profile,
with a single maximum at 437 nm [14]. We consider
the second maximum of WT HOX observed at 437 nm
to be a specific feature of the C–S bond at the Fl(6)
position, an observation that is backed up by the fact
that the His79K mutant shows only one maximum at
437 nm, as expected for a 6-S-cysteinyl flavin,
although no apparent shoulder at 380 nm, as observed
for TMDH [14]. Furthermore, we consider the hypso-
chromic shift of the first maximum of the difference
spectrum to 360 nm a unique feature of the 6-S-cystei-
nyl, 8a-N1-histidyl FAD. Together, these observations
support the fact that FAD in WT HOX is indeed a
6-S-cysteinyl, 8a-N1-histidyl FAD.
The observation made in this article of inducing
pH-dependent fluorescence of the flavin in HOX by
acid hydrolysis might provide a novel and quick

method to distinguish between this type of flavin-bind-
ing motif and other covalent-binding motifs. The fol-
lowing line of reasoning may be employed in this case.
HOX exhibits a distinctly lower fluorescence at pH 8
than an equimolar amount of COX, which contains
the same linkage to a His but lacks a linkage to a Cys.
The same was evident for the denatured forms of the
enzymes, as observed when equimolar amounts of the
enzymes were separated by SDS ⁄ PAGE and illumin-
ated under a UV lamp (data not shown). In compar-
ison, 6-S-cysteinylriboflavin is virtually nonfluorescent,
both in the isolated form and when bound in enzyme
(as discussed above) unless treated with performic acid
[14]. Flavins in enzymes exhibiting the 8a-Cys flavin-
binding motif are only weakly fluorescent and show no
fluorescence pH dependency when enzyme bound or
when released as aminoacylriboflavin by acid hydrol-
ysis [12].
The reason for the sudden boost in fluorescence of
the HOX hydrolysate is not fully understood, but the
most likely reason is that some of the C–S bonds were
broken, yielding a population of N1-histidylriboflavin.
It was previously shown that the C–S bond is stable
under these conditions [14], but this was done using
isolated 6-S-cysteinylriboflavin as the starting material
and not enzyme-bound 6-S-cysteinyl, 8a-N1-histidyl-
riboflavin. The pK
a
value of the hydrolysate of HOX
containing the histidyl–cysteinylriboflavin moiety was

determined to be 5.8, which is slightly higher than the
5.2 found previously for a synthetic isomer [23]. This
observation, combined with the modelling data of the
FAD surroundings (Fig. 6), leads to the conclusion
that His79 is bound at the N-1 position of histidine.
The higher pK
a
value is attributed to the influence of
the remaining free amino acids in the hydrolysate.
HOX, GOOX comparison
According to the model of FAD in HOX presented in
this article, the surroundings of the coenzyme in the
two enzymes are highly similar in spite of only about
25% overall sequence identity. In a structural align-
ment of the model of HOX and the structure of
GOOX, the FAD-binding amino acids His79 and
Cys138 are superimposed on the two amino acids cov-
alently linked to FAD in GOOX, His70 and Cys130,
respectively. The differences in substrate specificity and
kinetic parameters, along with the fact that the
T. Rand et al. Characterization of flavin in hexose oxidase
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2699
sequence identity of the F-domain is considerably
higher (33%), suggest that the two enzymes probably
evolved a somewhat different mode of binding to the
substrate but kept the overall structure of the
F-domain [1,5,11]. It is quite possible that this is nat-
ure’s way of reusing a good concept (the F-domain) in
combination with a variable S-domain to specialize
oxidases of this type. Although the two enzymes are

very similar, several pieces of biological and biochemi-
cal evidence separate the two. The substrate specifici-
ties of the two enzymes are similar, but whereas
GOOX displays a distinct preference for glucosaccha-
rides, HOX is more promiscuous in its substrate bind-
ing, resulting in a K
m
of 3.6 mm for galactose in spite
of the axial OH4 group of the sugar [5] It has indeed
been proposed that HOX is generally nonspecific for
the C-4 position in its substrates [24]. In comparison,
GOOX did not show any measurable activity on galac-
tose when tested against a range of sugars and sugar
derivatives [8]. Both enzymes oxidize the substrates
glucose, lactose, maltose, cellobiose and glucooligosac-
charides of up to five glucose residues in solution.
GOOX is reported to be able to utilize a heptamer of
glucose [8], whereas this has only been reported for
HOX when the enzyme is immobilized on a solid sup-
port [6]. The K
m
and k
cat
values of the two enzymes
differ markedly, in that GOOX appears to have higher
affinity for longer-chain gluco- and cello-oligomers,
while HOX exhibits higher affinities for the monomer
and dimer sugars. One could hypothesize that the pro-
cessing of longer-chain sugars in the F-domain of
GOOX is enhanced by docking of a multimer of dis-

crete length in the S-domain substrate-binding groove.
This does not appear to be the case with HOX, where
K
m
increases significantly with longer-chain oligomers,
indicating that the longer-chain substrates are less
firmly anchored [5].
With the similarities of the two enzymes thus con-
firmed on the basis of similar substrate specificities and
a novel FAD-binding mode, it may be speculated that
a large family of oligosaccharide oxidases exists that
may be screened for, based solely on the FAD-binding
motif. Indeed, our results from the multiple sequence
alignment against the GOOX sequence pinpointed 42
putative oligosaccharide oxidases, all of which have
conserved Cys and His at the FAD-binding positions.
The dual flavin linkage found in HOX and GOOX
may serve to provide stability, especially in the active
site of the enzyme. Protection against hydroxylation at
the C-6 position of flavin has been suggested to be a
result of the 6-S-linkage [8] and the 8a-N1-linkage was
found to affect the activity of HOX as well as flavin
incorporation in the holoenzyme.
Experimental procedures
Purification
WT and mutant HOX were expressed in H. polymorpha and
extracted using a recently patented process ([2], example 9).
The enzyme was isolated using ion exchange chromato-
graphy at ambient temperature. The ionic strength of 2 L
of the crude sample was reduced to 11 mS with 20 mm

Bis ⁄ Tris, and the pH was adjusted to 6.3. A Q-Sepharose
FF (XK50) column (200 mL) was prepared as described
by the manufacturer (Amersham Biosciences, Hillerød,
Denmark). The column was equilibrated with 20 mm
Bis ⁄ Tris buffer (pH 6.3), and elution was performed using a
linear gradient of 0–0.5 m NaCl in 20 mm Bis ⁄ Tris buffer
(pH 6.3). Active fractions were pooled and concentrated.
The purification yielded more than 3 g of pure enzyme, cor-
responding to total activity of 450 000 U as judged by
hydrogen peroxide-dependent horseradish peroxidase reduc-
tion of 2,2¢-azinobis(3-ethylbenzo-6-thiazolinesulfonic acid)
(ABTS) [25]. The specific activity of the WT recombinant
enzyme was approximately 150 UÆmg
)1
. The final purity of
the enzyme was more than 98% as judged by SDS ⁄ PAGE
analysis and western blotting (data not shown). Assuming
that the extinction coefficient of the 8a-, 6-dual covalent
FAD is in the same range as other 8a- or 6-modified flavins
(11.3–12.3 mm
-1
Æcm
)1
) [12], absorbance at 445 nm of a
0.005 mm solution of the recombinant WT enzyme (0.115)
supports a 1 : 2 enzyme ⁄ FAD stoichiometry.
The H79K mutant enzyme was purified using a slightly
different procedure. The ionic strength of 1.7 L of the crude
sample was reduced to 11 mS with 20 mm Bis ⁄ Tris, and the
pH was adjusted to 6.3. A Q-Sepharose FF (XK50) column

(200 mL) was prepared as described by the manufacturer
(Amersham Biosciences). The column was equilibrated with
20 mm Bis ⁄ Tris buffer (pH 6.3) and elution was performed
at ambient temperature using a linear gradient of 0–1 m
NaCl in 20 mm Bis ⁄ Tris buffer (pH 6.3). SDS ⁄ PAGE with
subsequent western blotting using antibodies raised against
HOX was used to isolate relevant fractions. These were
subsequently pooled. The purification yielded 50 mg of
inactive mutant enzyme. The final purity of the mutant
enzyme was more than 98% as judged by SDS ⁄ PAGE ana-
lysis and western blotting (data not shown). A 0.005 mm
solution of the H79K mutant enzyme was found to exhibit
an A
437 nm
of 0.055. Judging by the extinction coefficient of
a 6-substituted flavin (12.3 mm
)1
Æcm
)1
) [14], the H79K
mutant incorporates flavin in a 1 : 0.9 relationship.
Substrate-mediated FAD reduction
Aerobic absorption spectra (310–500 nm) were recorded at
ambient temperature and pressure in 0.1 m sodium phos-
phate buffer, pH 6.3, using a SPECTRAMAX
TM
190
spectrophotometer with a 96-well ELISA reader (Molecular
Characterization of flavin in hexose oxidase T. Rand et al.
2700 FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO

Devices, Sunny Vale, CA). Plates were transparent. All con-
centrations of HOX used were 0.005 mm, resulting in a
0.01 mm concentration with respect to flavin content
according to the 1 : 2 enzyme ⁄ FAD stoichiometry in the
holoenzyme. Substrate sugar (glucose) was added in excess
at 2.5 mm to a total volume of 300 lL(d ¼ 1 cm). The glu-
cose solution was used following incubation at ambient
temperature in the dark for 24 h to allow equilibrium muta-
rotation to reach completion.
Anaerobic absorption spectra (310–500 nm) were recorded
in 1 mL quartz cuvettes at ambient temperature and pressure
in 0.1 m phosphate buffer containing glucose (2.5 mm),
pH 6.3, which had been bubbled through and flushed with
N
2
gas. Enzyme (0.005 mm) was added while flushing and
bubbling through with N
2
gas for 30 min. The cuvettes were
sealed before measuring. Spectra of 2.5 mm solutions of
reaction products, hydrogen peroxide, gluconic acid–lactone
and gluconic acid were recorded (Sigma-Aldrich, St Louis,
MO). These were found to exhibit negligible absorbance at
these concentrations in the 300–500 nm UV range.
Fluorescence
Fluorescence experiments were carried out using a Gemi-
niEM spectrofluorometer from SPECTRAMAX with a
96-well ELISA reader (Molecular Devices). Plates were
opaque. All solutions were standardized in 0.05 m cit-
rate ⁄ phosphate buffers adjusted to pH values between 3

and 8, so that each well (300 lL) had a flavin concentration
of 0.06 lm. The fluorescence emission intensity at 525 nm
was recorded using an excitation wavelength of 445 nm.
Native protein samples, peptide and hydrolysate showed
the same low but distinct fluorescence emission maximum
at 525 nm at pH 8.
Acid hydrolysis
Samples (0.5 mg ) of pure HOX were lyophilized and
0.5 mL of 6 m HCl was added to the lyophilized protein in
sealed glass vials. The vials were incubated in the dark for
16 h at 95 °C and lyophilized. Samples were resuspended in
1 mL of deionized water and diluted to a final concentra-
tion of 1.8 lm in phosphate ⁄ citrate buffers, pH 3–8,
respectively; 10 lL of each sample was used in the fluores-
cence experiment.
Isolation and mass determination of flavopeptide
component from a trypsin digest
A sample of purified HOX was digested using modified tryp-
sin V5111 of sequencing grade (> 99% purity) (Promega
Gmbh, Mannheim, Germany). The manufacturer’s protocol
was followed, using iodoacetamide as the alkylating agent.
The digest was separated in three consecutive RP-HPLC
runs using a different gradient of acetonitrile (MeCN) in
each run. The flow (0.1 mLÆmin
)1
) from each RP-HPLC run
was split and MS was performed simultaneously using an
ESQUIRE ion trap mass spectrometer (Bruker, Rheinstet-
ten, Germany). The initial sample injection was loaded onto
a guard column (Zorbax StableBond Guard 1.0 · 17 mm;

Agilent Technologies, Naerum, Denmark) in 0.1% trifluoro-
acetic acid for salt removal, and the flow was reversed in
0.1% formic acid when loading from the guard column onto
the analytical column (Zorbax 300 SB-C18 RP-HPLC,
2.1 · 150 mm). This was done primarily because trifluoro-
acetic acid causes severe disturbances in the MS. (Buffer A:
0.1% formic acid in HPLC grade water. Buffer B: 0.1%
formic acid, in MeCN.) The fraction absorbing most inten-
sely at 450 nm was collected from each run and reapplied to
the analytical sample column. The conditions used for the
three consecutive runs were as follows:
Run 1: Start 2% (isocratic) B; 5 min, 2% B; 7 min,
12.5% B; 150 min, 45% B; 170 min (isocratic), 80% B;
180 min, 80% B; 181 min, 2% B.
Run 2: Start 2% (isocratic) B; 5 min, 2% B; 7 min,
12.5% B; 60 min, 35% B; 70 min (isocratic), 80% B;
73 min, 80% B; 73.1 min, 2% B.
Run 3: Start 2% B (isocratic); 5 min, 2% B; 7 min,
12.5% B; 60 min, 28% B; 70 min, 80% B (isocratic);
73 min, 80% B; 73.1 min, 2% B.
MALDI-TOF using an AUTOFLEX mass spectrometer
(Bruker) was used to confirm the singly charged ion species
of the isolated flavopeptide from HOX. The sample
(0.5 lL) from the purified flavopeptide (collected from run
3) was mixed with 0.5 lL of 2,5-dihydroxybenzoic acid in
20% MeCN. MALDI target plates (as detailed above) were
used for spotting. A mass gate was inserted for spectrum
simplification (m ⁄ z > 3000).
Edman sequencing
Edman sequencing was performed on a sample of purified

flavopeptide eluted in run 3. The sequencer used was a
PROCISE, sequencer (Applied Biosystems, Naerum, Den-
mark). A total of 20 cycles were performed.
Homology modelling
WU-BLAST at EBI (European Bioinformatics Institute)
established GOOX (1zr6.pdb without and 2axr.pdb with a
product analogue) as the most suitable templates for mod-
elling of HOX. An alignment of HOX and GOOX was
determined using Tcoffee [26,27], and some manual editing.
modeller was then used through the web interface at
Berkley University to generate a three-dimensional model
for HOX, based on the GOOX structure, and its alignment
with the HOX sequence. ( />cgi-bin/homology_model/build_model_data.py.) Forty-five
T. Rand et al. Characterization of flavin in hexose oxidase
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2701
models were generated, and the one with the best value of
the modeller objective function was chosen. Molecular
graphics were obtained using the pymol Molecular Graph-
ics System (DeLano Scientific, San Carlos, CA).
Multiple alignment of selected GOOX
homologues
The 50 closest homologues to GOOX were found using the
CONSEQ-server ( [28], with
three rounds of psi-blast. The sequence of HOX was
added, and muscle [29] was used to generate a multiple
sequence alignment. ( />cgi-bin/muscle/input_muscle.py).
Acknowledgements
We thank M. R. Zargahi for kindly providing assist-
ance in the fermentation, extraction and purification of
wild-type hexose oxidase and mutant H79K hexose

oxidase, both from H. polymorpha.
References
1 Hansen OC & Stougaard P (1997) Hexose oxidase from
the red alga Chondrus crispus. Purification, molecular
cloning, and expression in Pichia pastoris. J Biol Chem
272, 11581–11587.
2 Johansen CL, Kjaerulff S, Madrid SM, Pedersen H,
Poulsen CH & Zargahi MR (2005) Extraction of pro-
teins from cells using quaternary ammonium com-
pounds. WO 2001038544. Denmark.
3 Groen BW, De Vries S & Duine JA (1997) Characteri-
zation of hexose oxidase from the red seaweed Chondrus
crispus. Eur J Biochem 244, 858–861.
4 Wolff AM, Hansen OC, Poulsen U, Madrid S & Stou-
gaard P (2001) Optimization of the production of Chon-
drus crispus hexose oxidase in Pichia pastoris. Protein
Exp Purif 22, 189–199.
5 Poulsen C & Høstrup P (1998) Purification and charac-
terization of a hexose oxidase with excellent strengthen-
ing effects in bread. Cereal Chem 1, 51–57.
6 Van der Lugt JP (1998) Evaluation of Methods for
Chemical and Biological Carbohydrate Oxidation, pp.
1–148. Technische Universiteit, Delft.
7 Mathews FS, Chen ZW, Bellamy HD & McIntire WS
(1991) Three-dimensional structure of p-cresol methyl-
hydroxylase (flavocytochrome c) from Pseudomonas
putida at 3.0-A resolution. Biochemistry 30, 238–247.
8 Huang CH, Lai WL, Lee MH, Chen CJ, Vasella A,
Tsai YC & Liaw SH (2005) Crystal structure of gluco-
oligosaccharide oxidase from Acremonium strictum:a

novel flavinylation of 6-S-cysteinyl, 8alpha-N1-histidyl
FAD. J Biol Chem 280, 38831–38838.
9 Decker K & Brandsch R (1997) Determining covalent
flavinylation. Methods Enzymol 280, 413–423.
10 Mewies M, McIntire WS & Scrutton NS (1998) Cova-
lent attachment of flavin adenine dinucleotide (FAD)
and flavin mononucleotide (FMN) to enzymes: the cur-
rent state of affairs. Protein Sci 7, 7–20.
11 Sullivan JD Jr & Ikawa M (1973) Purification and char-
acterization of hexose oxidase from the red alga Chon-
drus crispus. Biochim Biophys Acta 309, 11–22.
12 Edmondson DE & De Franscesco R (1991) Structure,
synthesis, and physical properties of covalently bound
flavins and 6- and 8-hydroxy flavins. In Chemistry and
Biochemistry of Flavoenzymes (Franz Mu
¨
ller, ed),
pp. 73–104. CRC Press, Boca Raton, Florida.
13 Steenkamp DJ, McIntire W & Kenney WC (1978)
Structure of the covalently bound coenzyme of tri-
methylamine dehydrogenase. Evidence for a 6-substi-
tuted flavin. J Biol Chem 253, 2818–2824.
14 Steenkamp DJ, Kenney WC & Singer TP (1978) A
novel type of covalently bound coenzyme in trimethyl-
amine dehydrogenase. J Biol Chem 253, 2812–2817.
15 de Jong E, van Berkel WJ, van der Zwan RP & de Bont
JA (1992) Purification and characterization of vanillyl-
alcohol oxidase from Penicillium simplicissimum.A
novel aromatic alcohol oxidase containing covalently
bound FAD. Eur J Biochem 208, 651–657.

16 Muller F (1987) Flavin radicals: chemistry and biochem-
istry. Free Radic Biol Med 3, 215–230.
17 Ghisla S, Kenney WC, Knappe WR, McIntire W &
Singer TP (1980) Chemical synthesis and some proper-
ties of 6-substituted flavins. Biochemistry 19, 2537–2544.
18 Rand T, Halkier T & Hansen OC (2003) Structural
characterization and mapping of the covalently linked
FAD cofactor in choline oxidase from Arthrobacter
globiformis. Biochemistry 42, 7188–7194.
19 Kenney WC, Edmondson DE & Seng RL (1976) Identi-
fication of the covalently bound flavin of thiamin dehy-
drogenase. J Biol Chem 251, 5386–5390.
20 Ghisla S & Hemmerich P (1971) Synthesis of the flavo-
coenzyme of monoamine oxidase. FEBS Lett 16, 229–
232.
21 Nagy J & Salach JI (1981) Identity of the active site fla-
vin-peptide fragments from the human ‘A’-form and the
bovine ‘B’-form of monoamine oxidase. Arch Biochem
Biophys 208, 388–394.
22 Wagner MA, Khanna P & Jorns MS (1999) Structure
of the flavocoenzyme of two homologous amine oxi-
dases: monomeric sarcosine oxidase and N-methyltryp-
tophan oxidase. Biochemistry 38, 5588–5595.
23 Edmondson DE, Kenney WC & Singer TP (1976) Struc-
tural elucidation and properties of 8alpha-(N1-histidyl)
riboflavin: the flavin component of thiamine dehydro-
genase and beta-cyclopiazonate oxidocyclase. Biochemis-
try 15, 2937–2945.
Characterization of flavin in hexose oxidase T. Rand et al.
2702 FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO

24 Bean RC & Hassid WZ (1956) Carbohydrate oxidase
from a red alga, Iridophycus flaccidum. J Biol Chem 218,
425–436.
25 Savary BJ, Hicks KB & O’Connor JV (2001) Hexose
oxidase from Chondrus crispus: improved purification
using perfusion chromatography. Enzyme Microb
Technol 29, 42–51.
26 Notredame C, Higgins DG & Heringa J (2000) T-Cof-
fee: a novel method for fast and accurate multiple
sequence alignment. J Mol Biol 302, 205–217.
27 Poirot O, O’Toole E & Notredame C (2003) Tcof-
fee@igs: a web server for computing, evaluating and
combining multiple sequence alignments. Nucleic Acids
Res 31, 3503–3506.
28 Berezin C, Glaser F, Rosenberg J, Paz I, Pupko T, Fari-
selli P, Casadio R & Ben Tal N (2004) ConSeq: the
identification of functionally and structurally important
residues in protein sequences. Bioinformatics 20, 1322–
1324.
29 Edgar RC (2004) MUSCLE: multiple sequence align-
ment with high accuracy and high throughput. Nucleic
Acids Res 32, 1792–1797.
T. Rand et al. Characterization of flavin in hexose oxidase
FEBS Journal 273 (2006) 2693–2703 ª 2006 DANISCO 2703