Báo cáo Y học: Structural and functional characterization of a C-type lectin-like antifreeze protein from rainbow smelt (Osmerus mordax) potx

Structural and functional characterization of a C-type lectin-like
antifreeze protein from rainbow smelt (
Osmerus mordax
)
John C. Achenbach
1,
* and K. Vanya Ewart
2
1
Department of Biochemistry and Molecular Biology, Dalhousie University, Halifax, Nova Scotia, Canada;
2
NRC Institute for Marine
Biosciences, Halifax, Nova Scotia, Canada
Antifreeze proteins (AFPs) are produced by several
cold-water fish species. They de press physiological freezing
temperatures by inhibiting growth of ice crystals and, in so
doing, permit the survival of these fish in seawater cooler
than their normal freezing temperatures. The type II AFP
from rainbow smelt (Osmerus mordax), which is a member
of the C-type lectin supe rfamily, was characterized in terms
of its Ca
2+
-binding quaternary structure and the role of its
single N-linked oligosaccharide. The protein core of the
smelt AFP, shown through sequence homology to be a
C-type lectin carbohydrate-recognition domain, was found
to be protease resist ant. Smelt AFP was also shown to bind
Ca
2+
, as determined by ruthenium red staining a nd a
conformational change on Ca

2+
binding detected by
intrinsic fluorescence. The N-linked oligosaccharide w as
found to have no effect on protease resistance, dimerization,
or antifreeze activity. Thus its role, if any, in the antifreeze
function of this protein remains unknown. Smelt AFP was
also shown to be a true intermolecular dimer composed of
two separate subunits. This dimerization did not require the
presence of N-linked oligosaccharide or bound Ca
2+
. Smelt
AFP dimerization has implications for the effective solution
concentration and measurement of its activity. This finding
may also lead to new interpretation of the mechanism of
ice-growth inhibition by this A FP.
Keywords: antifreeze protein; C-type lectin; dimerization;
glycosylation; rainbow smelt.
Antifreeze proteins (AFPs) are produced by many sp ecies as
an efficient means of protection from freezing. They bind
directly to growing ice crystals and thus inhibit c rystal
growth [1]. These proteins generate a far greater freezing
point depression than would be predicted from their
colligative properties. This is presumed to be due to their
direct interaction with ice crystals. The AFPs only have a
colligative effect on the melting point of a solution and
therefore cause a thermal hysteresis. This phenomenon is
the basis for quantitative measurement of AFP activity.
The AFPs are structurally and evolutionarily diverse but
are all functionally similar in that they bind ice and depress
the freezing point. They are found in a variety of insect and

plant species as well as bacteria and fungi, but fish AFPs
were the first and most extensively characterized [2]. Among
fish species, five structurally defined types of AFPs have
been identified. The antifreeze glycoproteins found in cods
and Antarctic nototheniids are composed of multiple Ala-
Ala-Thr repeats, with a disaccharide linked to each Thr
residue [1]. The type I AFPs found in certain flounders and
sculpins are alanine-rich, amphiphilic a helices with a
repeating pattern of Thr residues [3,4]. The type III AFPs
have a unique globular fold with one flattened surface
thought to take part in ice binding, a nd are f ound in
northern and Antarctic eel pouts [5]. The recently charac-
terized type IV AFP is composed of an antiparallel helix
bundle homologous to an apolipoprotein, which has been
only found in longhorn sculpin, Myoxocephalus o ctodec-
imspinosis [6]. Type II AFPs have been shown to be
homologous to C-type lectins. They are found in the serum
of three teleost fishes: sea raven (Hermitripterus americanus),
Atlantic herring (Clupea harengus harengus), and rainbow
smelt (Osmerus mordax) [2]. They consist of a long form of
the C -type c arbohydrate-recognition domain and are
C-type lectin-like domains ( CTLDs) [7–9]. This domain
consists of a t ightly folded hydrophobic core stabilized by
three or more disulfide bonds [10].
The type II AFPs of herring and smelt appear to be very
closely related but the sea raven AFP is more distinct. The
smelt and herring type II AFPs exhibit near-equivalent
molar antifreeze activity (thermal hysteresis), whereas the
sea raven AFP appears f ar more active in generating
thermal hysteresis [11]. Moreover, the smelt and herring

AFPs are structurally very similar. These AFPs share 86%
sequence identity in a 126-residue sequence overlap and they
have very similar putative signal seq uences. They share
approximately twice the percentage sequence identity with
one another than either share with the sea raven protein [7].
The smelt and herring AFPs also share the Gln-Pro-Asp
(galactose-binding) motif of galactose-binding C-type lec-
tins. This is the centre of the carbohydrate-binding site in
C-type lectins and was shown to be the ice-binding site in
herring AFP [12]. Like th ese lectins, which require Ca
2+
to
be bound before they can bind carbohydrate, the smelt and
herring AFPs require Ca
2+
in order to bind to ice [7,13].
Smelt AFP does differ from herring AFP and other fish
Correspondence to K. V. Ewart, NRC Institute for Marine Biosciences,
1411 Oxford St, Halifax, NS, B3H 3Z1, Canada.
Fax: + 1 902 426 9413, Tel.: + 1 902 426 7 620,
E-mail:
Abbreviations: AFP, antifreeze protein; BS
3
, bis(sulfosuccinimdyl)
suberate; CTLD, C-type lectin-like domain; Glu-C, endoproteinase
Glu-C.
*Present address: Department of Biochemistry, McMaster University,
Hamilton, Ontario, Canada.
(Received 26 October 2001, accepted 2 January 2002)
Eur. J. Biochem. 269, 1219–1226 (2002) Ó FEBS 2002

AFPs, however, in several important respects. It has an
18-residue N-terminal extension (GDTGKEAVMTGSSG
KNLT) and an 11-residue C-terminal extension (VNP
EVTPPSIM) not present in h erring AFP. It also has a n
N-linked glycosylation site in the N-terminal extension
sequence [ 13]. Therefore, the goal of this study was to
investigate the structural and functional characteristics of
smelt type II AFP. The effects and inter-relationship of
metal ion binding, N-linked glycosylation, and quaternary
structure in this protein were studied in order to b etter
define its structure and function as an AFP as well as its
relationship to other CTLDs.
MATERIALS AND METHODS
Materials
N-Glycosidase F and endoproteinase Glu-C were obtained
from Roche Molecular Biochemicals (Laval, Canada).
Bis(sulfosuccinimdyl) suberate (BS
3
) and Gelcode Blue
(Coomassie) stain were obtained from P ierce Chemicals
(Rockford, IL, USA). Sequencing-grade endoproteinase
Glu-C a nd trypsin were obtained from Promega (Madison,
WI, USA). Ruthen ium red was purchased from Fluka
Chemicals (Ronkonkoma, NY, USA). Nitrocellulose mem-
brane (0.45 lm) was purchased from Bio-Rad (Hercules,
CA, USA). All other chemicals we re reagent grade.
Purification of smelt AFP
Blood plasma was obtained f rom a population of rainbow
smelt (O. mordax) caught i n seawater a long the north-
eastern coast of Newfoundland on 20 February 1997 and

stored frozen until use. Approximately 2.5 mL plasma was
fractionated on a 1 · 90 cm S-200 Sephacryl (Pharmacia,
Uppsala Sweden) gel-filtration c olumn. Fractions contain-
ing AFP, as determined by SDS/PAGE, were then applied
to a 1 · 30 cm phenyl–Sepharose hydrophobic interaction
column (Amersham–Pharmacia), and eluted in 20 m
M
Tris/
HCl, pH 8.0, buffer with a continuous descending (1–0
M
)
NaCl gradient. Fractions containing AFP, as identified by
SDS/PAGE, were pooled and lyophilized. The lyophilized
sample was reconstituted in a minimal v olume of 0.1
M
NH
4
HCO
3
buffer, pH 8.0, desalted using a PD-10 column
(Amersham P harmacia), lyophilized, and stored at )20 °C.
Preparation of deglycosylated smelt AFP
Lyophilized N-glycosidase F was reconstituted to 1 UÆlL
)1
in water. Lyophilized smelt AFP was dissolved in 25 m
M
Hepes, pH 7.8, to a final concentration of 4 mgÆmL
)1
. Then
10 U N-glycosidase F was added to 100 lL of smelt AFP

solution and incubated at 37 °C; a f urther 5 U N-glyco-
sidase F added 4 h l ater, and aga in after an overnight
incubation. A further 8 U was added over a period of 4 h
followed by a second overnight incubation at 37 °C. The
reaction was considered complete after SDS/PAGE analysis
of the reaction mixture re vealed no detectable glycosylated
smelt AFP using Gelcode Blue staining. A parallel control
reaction was carried out by adding equivalent volumes of
water t o 100 lL of the smelt AFP solution, and incubating
at 37 °C for the same amount of time.
Ruthenium red staining
Ruthenium red dye binding was evaluated following the
method of Ch aruk et al. [14] with minor modifications.
Deglycosylated smelt AFP and untreated smelt AFP
samples (both 4.5 lg per lane) were run on an SDS/15%
polyacrylamide gel under nonreducing conditions and then
electrophoretically transferred to a 0.45-lm nitrocellulose
membrane. A blot section was incubated at 4 °Cfor
15 min in staining buffer (20 m
M
Hepes, pH 7.8,
10 mgÆmL
)1
rutheniumred),followedbya15-minwash
at 4 °C in wash buffer (20 m
M
Hepes, pH 7.8). An
identical section was treated the same way except
100 m
M

CaCl
2
was present in both the staining and w ash
buffer. Amido black staining was used to confirm protein
presence and to detect bands not stained by the ruthenium
red dye.
Analysis of antifreeze activity
Antifreeze activity was quantitated a s thermal hysteresis,
which is the difference b etween the melting and freezing
point of a solution. Thermal hysteresis was measured by
monitoring ice crystal behavior using a nanoliter osmometer
(Clifton Technical P hysics, Hartford, NY, USA). Five
measurements were taken o n 130-l
M
solutions of deglycos-
ylated and untreated smelt AFP in 25 m
M
Hepes, pH 7.8,
using water as a blank . As a control, the thermal hysteresis
of a solution of N-glycosidase F was measured in 25 m
M
Hepes, pH 7.8. Results were expressed a s mean ± SE.
Photographs of ice crystals viewed during these measure-
ments were taken at a magnification of 200·.
Protease protection assays on smelt AFP
Protease protection assays to compare deglyco sylated and
untreated smelt AFP were performed as previously
described [15] with minor modifications. Reaction volumes
of 16 lL containing 4.8 lg of either untreated or deglycos-
ylated smelt AFP in 10 m

M
Hepes, pH 7.8, and either 2 m
M
EDTA or 20 m
M
CaCl
2
were incubated for 30 min. This
was followed by the addition of protease (0.1 mgÆmL
)1
final
concentration), or an equal volume of water, after which all
samples were incubated for 3 h at room temperature. An
aliquot of each reaction mixture w as resolved by SDS/
PAGE (15% gel) under reducing conditions, and stained
with Gelcode Blue.
In vitro
chemical cross-linking
Untreated and deglycosylated smelt AFP samples in HCS
buffer (25 m
M
Hepes, pH 7.8, 150 m
M
NaCl, 10 m
M
CaCl
2
) or HES buffer (10 m
M
EDTA in place of CaCl

2
)
at a final AFP concentration of 33 l
M
, were aliquoted into
the w ells of a microtiter plate and left undisturbed for
15 min at room temperature. A serial dilution of BS
3
dissolved in either HCS or HES buffer was performed, and
aliquots of each dilution were added t o corresponding AFP-
containing wells. The reaction mixtures were incubated at
room temperature for 1 h, then resolved by SDS/PAGE
(15% gels) under reducing conditions and stained with
Gelcode Blue.
1220 J. C. Achenbach and K. V. Ewart (Eur. J. Biochem. 269) Ó FEBS 2002
Determination of molecular mass by gel-filtration
chromatography
A TosoHaas 4-lm particle size TSK SuperSW 2000 column
(4.6 · 300 mm) was equilibrated in HCS buffer and run at a
flow rate of 0.3 mLÆmin
)1
using an HPLC (Waters). The
column was calibrated using protein molecular-mass stand-
ards. Approximately 15 lg untreated smelt AFP was
applied to the column in duplicate runs. Similar samples
of deglycosylated AFP were applied as well. All proteins
were detected by monitoring A
230
.
Fluorescence measurements

Intrinsic fluorescence was measured using an Amin co
Bowman series 2 spectrofluorimeter at room temperature.
Untreated and deglycosylated s melt AFP samples were
diluted to equimolar concentrations in 25 m
M
Hepes,
pH 7.8, containing 1 m
M
EDTA. Emission spectra were
recorded with a 2-nm/s scan rate in duplic ate for each trial
with an excitation wavelength of 280 nm (4 nm bandpass).
CaCl
2
was added by pipetting a 0.5-
M
solution directly into
the sample cuvette with thorough mixing followed by a
10 min incubation at room temperature. A 0.5-
M
solution of
EDTA (pH 8.0) was added in a similar fashion to test the
reversibility of any Ca
2+
effect. Spectra were corrected for
the resultant dilutions.
RESULTS
Deglycosylation of smelt AFP
To study the effect of the N-linked o ligosaccharide of smelt
AFP, it was removed from the protein enzymatically using
N-glycosidase F. The reaction evaluated using SDS/PAGE

was f ound to result in a band of reduced size (17 kDa)
compared with the untreated band (22 kDa) (results not
shown). The 17-kDa molecular mass corresponds to the
17.4-kDa calculated mass o f the herring AFP sequence
(from cDNA) with the predicted signal sequence removed.
This result implies that the smelt AFP has no prosequence
and co nsists of the complete sequence p redicted from
cDNA minus the signal.
Ruthenium red staining
To detect Ca
2+
binding by smelt AFP, both untreated and
deglycosylated smelt AFP were blotted as discussed in
Materials and methods and stained with a 10 mgÆmL
)1
solution of ruthenium red. This dye was shown to bind both
the untreated and deglycosylated smelt AFP (Fig. 1) as well
as the known Ca
2+
-binding molecular mass standard
b-lactoglobulin (not shown). The specificity of binding is
demonstrated by the absence of detectable ruthenium red
staining when CaCl
2
is added to the staining and washing
buffer (Fig. 1).
Measurement of antifreeze activity
Antifreeze ac tivity was measured on equimolar amounts of
deglycosylated and untreated smelt AFP to determine
whether the N-linked oligosaccharide had any effect on this

activity. A faceted ice-crystal morphology signifies the
presence of antifreeze activity. The rounded ice crystal
formed in the control sample c ontaining only N-glycosi-
dase F shows that the enzyme does not display antifreeze
activity (Fig. 2). Equimolar solutions of deglycosylated and
untreated smelt AFP displayed similar faceted ice crystal
morphology (Fig. 2). The antifreeze activity of the untreat-
ed and d eglycosylated samples, quantitate d as thermal
hysteresis, were 0.027 ± 0.003 and 0.026 ± 0.003, respect-
ively. These are not significantly different (P < 0.001). As
expected, the enzyme control sample showed no activity,
with a hysteresis value of )0.001 ± 0.002.
Protease protection assays
To determine whether Ca
2+
or the N-linked oligosaccha-
ride has any effect on the protease susceptibility of smelt
AFP, both untreated and deglycosylated smelt AFP were
digested with Glu-C and trypsin in the presence and a bsence
of Ca
2+
. Both enzymes generated some proteolysis near the
N-terminus and/or C-terminus of the protein leaving a large
central AFP fragment intact. The sizes o f the digestion
fragments were all greater than 14 kDa, w hich is the
approximate size of the core CTLD. The addition of Ca
2+
at a concentration of 20 m
M
had no effect on the p roteolysis

seen with either deglycosylated o r untreated smelt AFP
(Fig. 3 A,B). This suggests that the CTLD of smelt AFP is
protease resistant both in the presence and absence of Ca
2+
and with a nd without the N-linked oligosaccharide. Similar
digestion patterns were also seen in both the untreated and
deglycosylated smelt AFP trials using either protease.
Digestion of either deglycosylated or un treated smelt AFP
formed products that were 1 kDa and 2 kDa smaller than
undigested smelt AFP. The size differences between the
digestion fragments of untreated and deglycosylated AFP
Fig. 1. Ruthenium red staining of smelt AFP. Ruthenium red (RR)
staining of deglycosylated (D) smelt AFP and untreated (U) smelt
AFP was carried out as described in Materials and methods. Blots
shown are stained with a mido black (Amido), ruthenium red (RR)
or ruthenium red in the p resence of 100 m
M
CaCl
2
(RR + Ca
2+
).
Molecular-mass marker sizes (kDa) are indicated.
Ó FEBS 2002 Characterization of smelt antifreeze protein (Eur. J. Biochem. 269) 1221
are all 5 kDa, which corresponds to the apparent size of the
carbohydrate determined in a separate experiment (Fig. 1).
Taken together, these results indicate that the proteases
digest both untreated and deglycosylated forms of smelt
AFP in close, if not ide ntical, locations. These results also
indicate that protease digestion under these conditions does

not cleave between the Asn residue carrying the N-linked
oligosaccharide and the core CTLD. Bands corresponding
to Glu-C and trypsin enzymes were identified on the basis of
molecular mass. The bands smaller than 14 kDa in the
trypsin digests were shown to be trypsin autolysis products
in a separate control experiment containing trypsin alone
(not shown).
Intrinsic fluorescence
Modulation o f protein conformation on addition of Ca
2+
was monitored by intrinsic fluorescence. Excitation at
280 nm produced an emi ssion spectrum with a k
max
of
343 n m from % 30 l
M
samples of both untreated (Fig. 4A)
Fig. 2. Analysis of antifreeze a ctivity. Antifreeze activity was evaluated
qualitatively by monitoring ice crystal morphology in solutions con-
taining deglycosylated and untreated smelt AFP (both 130 l
M
)as
described i n Materials and metho ds. The top row of p hotos is of ice
crystals with their c axis in the plane of the page. The bottom row
shows ice crystals with their c axis normal to t he page.
Fig. 3. Proteolytic digestions of deglycosylated and untreated smelt AFP in presence and absence of Ca
2+
. Both deglycosylated and untreated smelt
AFP were incubated with protease in the presence or absence of 20 m
M

Ca
2+
as described in Materials and methods. All reaction products were
analyzed by SDS/PAGE and stained with Gelcode Blue. (A) AF P digested with Glu -C. Bands corresp onding to Glu-C , untreated ( UT) AFP
(UT AFP), and deglycosylated AFP (DG AFP) are identified. Proteolytic fragments are seen as lower-molecu lar-mass bands in Glu-C-contain ing
lanes. ( B) AFP digested with trypsin. Bands correspond ing to trypsin, untreated AFP (UT AFP), and deglycosylated AFP (DG) are id entified.
Bands smaller than 1 4 kDa in lan es containing trypsin were shown to b e trypsin autolysis products (trypsin-only digest not shown). Molecular-mass
marker sizes (kDa) are indicated.
Fig. 4. Effect of Ca
2+
on the intrinsic fluorescence of deglycosylated and
untreated smelt AFP. (A) Emission spectra of intact (u ntreated) AFP
(UT AFP). (B) Emission spectra of deglycosylated AFP (DG AF P).
Spectra were recorded for solutions of 30 l
M
smelt AFP in 25 m
M
Hepes, pH 7.8, containing 1 m
M
EDTA, before and after the ad dition
of CaCl
2
(10 m
M
). Each spectrum shown is the average of two suc-
cessive recordings. All measurements were taken with excitation at
280 n m and a 2-nmÆs
)1
scan rate at ambient temperature.
1222 J. C. Achenbach and K. V. Ewart (Eur. J. Biochem. 269) Ó FEBS 2002

and deglycosylated (Fig. 4B) smelt AFP. The emission
intensity was shown to increase significantly i n both samples
on addition of Ca
2+
to a final concentration of 10 m
M
.The
effect was fully reversible in both samples, with the addition
of EDTA to a final c oncentration of 18 m
M
(not shown).
No blue or red shift was observed in the emission spectra of
either sample on addition of Ca
2+
or EDTA.
Detection of dimerization
To determine the quaternary structure of smelt AFP in
solution, AFP (33 l
M
) was incubated in the presence of a
range of concentrations of BS
3
, a water-soluble homo-
bifunctional cross-linking agent, in the presence of 10 m
M
CaCl
2
.IntheabsenceofBS
3
, a 22-kDa band corresponding

to the smelt AFP monomer was observed (Fig. 5A). A BS
3
concentration of 0.16 m
M
was sufficient to gen erate the
accumulation o f a larger protein form. W ith increasing
concentrations of BS
3
, the disappearance of the monomer
band corresponded to the gradual increase in intensity of a
higher-molecular-mass band. The apparent molecular mass
of this band was 38 kDa, which is substantially higher than
the molecular mass of the monomer band, indicating the
presence of a dimer. The expected mass of a smelt AFP
dimer is 44 kDa but, on SDS/PAGE, the covalently cross-
linked dimer would be expected to migrate faster than its
actual mass would predict because o f limited linearization in
SDS. The value of 38 kDa is consistent with such a d imer
band. A small amount of a high-molecular-mass aggre gate
was evident at the top of the SDS/PAGE gel lane co ntaining
the sample with t he highest concentration (20 m
M
)ofBS
3
.
This appears to be an artefact of high cross-linker concen-
tration a nd was not taken as evidence of any larger protein
aggregate. To determine whether the oligosaccharide plays a
role in dimerization, the experiment was repeated on
deglycosylated smelt AFP (Fig. 5B). Dimerization was

evident with monomer bands of 18 kDa and dimer bands
of 33 kDa. To examine whether Ca
2+
binding was required
for dimerization, smelt AFP was incubated with BS
3
in
buffer containing 10 m
M
EDTA instead of CaCl
2
.Inthe
absence of added Ca
2+
, dimerization was again observed
(Fig. 5C).
The molecular mass of smelt AFP determined by HPLC
gel-filtration analysis was 50 kDa (Fig. 6 ). As the molecular
mass of the protein on SDS/PAGE is 22 kDa, the value
obtained b y g el filtration is consistent with a dimer. There
was no e vid ence of l arge r aggregates, a s determined by
absorbance at 230 nm, nor were there any detectable peaks
corresponding to the monomer size (22 kDa). These results
are in agreement with the cross-linking experiments indica-
ting that the native smelt AFP is fully dimerized. Deglycos-
ylated AFP was also analyzed on the same column and
found to have a molecular mass of 41 kDa, which also
indicates dimer formation (not shown).
Fig. 5. Chemical cross-linking of smelt AFP. Smelt AFP (33 l
M

) was cross-linked using BS
3
as described in Materials and methods. After the 1 h
incubation, reaction c ontents were analyzed using SDS/PAGE under reducing conditions and stained using Gelcode Blue. Lanes 1–9 in all gels
correspond to BS
3
concentrations of 0, 0.16, 0.3, 0.6, 1.25, 2.5, 5, 10, and 20 m
M
, respectively. (A) I ntact AFP in the presence of CaCl
2
.(B)
Deglycosylated AFP in the presence of CaCl
2
. (C) Intact AFP in t he presence of EDTA. P osit ions of dimers (D) and monomers (M) and the
molecular-mass marker sizes (kDa) are indicated.
Fig. 6. HPLC gel-filtration analysis of smelt AFP. Smelt AFP was
applied to t he colum n un der t he co nditio ns described in Materials and
methods. The elution times of molecular-mass marker proteins (kDa)
are indicated by arrows.
Ó FEBS 2002 Characterization of smelt antifreeze protein (Eur. J. Biochem. 269) 1223
DISCUSSION
Intermolecular dimer formation by the smelt AFP makes it
unique among the fish AFPs. Other AFP types, s uch as type
I, IV, and the antifreeze glycoproteins, appear to exist as
monomers. Similarly, type III AFP is monomeric, and
shown not to self-associate when binding ice [16]. A type III
AFP f rom an Antarctic eel p out ( Rhigophila dear borni)
contains a tandem repeat of the domain that comprises
other type III AFPs on a single chain [17]. However, the
type III AFP, referred to as a n intramolecular dimer, is i n

effect a monomer (single polypeptide chain) and not a true
intermolecular dimer. The AFPs of winter rye (Secale
cereale) w ere found to form larger intermolecular com-
plexes, a lthough the precise subunit composition in the
hetero-oligomers containing different A FPs from t his plant
is unknown [18]. The dimerization of smelt AFP is
intriguing because the homologous sea raven type II AFP
was found to exist as a monomer in solution [19].
Dimerization of smelt AFP is unlikely to increase activity
by increased ice binding because only a dimer with perfectly
aligned ice-binding sites would have enhanced binding to ice
and this is not likely to be the case. This may be relevant to
the calculation of smelt AFP activity. The activity of smelt
AFP is about one-third of that of sea raven AFP on a
monomer concentration basis [11]. As sea raven and herring
AFPs bind to ice at a distinct site on the CTLD [12,21], it is
possible th at t he activity difference simply r eflects the
difference in ice binding. However, if the dimerization of
smelt AFP e ffectively prevents one subunit from binding ice,
this could a ccount f or some or all of t he difference in activity
calculated on a monomer basis. The gel-filtration and cross-
linking results shown here suggest that smelt AFP is fully
dimerized under physiological conditions, with very low or
undetectable levels of monomeric protein. Because smelt
AFP is fully dimerized, it would be more appropriate to
calculate its activity on the basis of dimer molarity t han on a
monome r basis .
Although the dimerization site of smelt AFP cannot be
determined from th e results of this study, it is unlikely that
the dimerization is due to intermolecular disulfide bonds.

Smelt AFP migrates on SDS/PAGE under nonreducing
conditions with a monomeric size, as shown in the
ruthenium red-binding assay (Fig. 3). In addition, all of
the cysteine r esidues are conserved among the three type II
AFPs and w ere shown to f orm intramolecular disulfide
bonds in the monomeric sea raven AFP structure [8].
Because it was possible to chemically cross-link the smelt
AFP dimers using the homobifunctional cross-linking agent
BS
3
, dimerization must result in the p lacement of two
primary amines, from either lysine side chains, or the
N-terminus, a maximum distance of 11.4 A
˚
apart. This
value corresponds to the spacer arm length of the BS
3
.
Several of the soluble C-type lectins associate to form dimers
[22–25]. Therefore, t he dimerization of smelt AFP is
consistent with other members of the C-type lectin super-
family and is well in keeping with the soluble lectins among
them. By analogy with other C-type lectins known to
dimerize, t here could be several types of dimerization
surfaces [22,23]. Dimerization or multimerization are com-
mon characteristics of C-type lectins, especially those
involved in the a cute-phase response o f host i mmune
defense such as the collectins. In such cases, multimerization
serves to enhance the avidity and pattern recognition of
such lectins towards pathogen carbohydrate structures [26].

The inorganic dye ruthenium red has been shown t o
selectively bind to Ca
2+
-binding sites of several proteins,
including the herring AFP [12–15]. Like the herring A FP,
smeltAFPwasshowntobindtheCa
2+
analogue,
ruthenium red, suggesting the presence of a Ca
2+
-binding
site. Further evidence for direct Ca
2+
binding by smelt AFP
was obtained using intrinsic fluorescence. An increase in
intrinsic fluorescence emission intensity was shown to result
from Ca
2+
addition, consistent with the result of similar
experiments on herring AFP [15]. This increase indicates a
change in the environment of tryptophan residues in the
protein on Ca
2+
addition, which is normally indicative of a
conformational change. The number of bound Ca
2+
ions
per smelt AFP monomer remains unknown. However,
because smelt AFP is identical with that of herring in the
positions corresponding to lectin Ca

2+
-binding sites [7], it is
reasonable to suggest that the smelt AFP binds a single
Ca
2+
in the same way as the herring AFP [15].
In comparison w ith o ther C-type lectins and he rring
AFP, smelt AFP appears to be a remarkably p rotease-
resistant protein in both the presence and absence of Ca
2+
ions. Digestion with trypsin or endoproteinase Glu-C
resulted in the formation of two slightly smaller digestion
fragments regardless of whether or no t C a
2+
was p re sent.
The large size of the digestion fragments indicates that most
of the smelt AFP is unavailable for protease digestion under
the conditions tested. The sizes of the digestion fragments
were on average only % 1–2 kDa smaller than the undi-
gested smelt AFP, suggesting that cleavage of only t he
extended N-terminus and C-terminus occurred, leaving
the remaining CTLD. These results are consistent with the
hypothesis that smelt AFP monomer is composed primarily
of a single CTLD. However, in similar studies, the CTLD-
containing herring AFP and chicken hepatic lectin did n ot
show the same level of protease resistance in the absence of
bound Ca
2+
[15,27]. The dimerization shown to o ccur
between smelt AFP monomers may enhance protease

resistance by shielding residues in the dimerization interface.
A conformational change s hown to occur in herring AFP
on Ca
2+
binding generated a shift from p rotease sensitivity
to protease resistance on Ca
2+
binding [15]. However, in t he
present investigation of smelt AFP, there was evidence for a
protease-resistant CTLD core in the absence Ca
2+
. The
addition of Ca
2+
afforded no further protection from
proteolysis. It is clear from the fl uorescence expe riments t hat
apo-(smelt AFP) undergoes a conformational change i n
response to Ca
2+
binding, which in other lectins and herring
AFP is necessary for protease resistance. It is highly unlikely
that sites susceptible to proteolysis in C a
2+
-free herring
AFP and related C-type lectins would be less so in smelt
AFP as a result of any residue differences because equiv-
alent protease resistance and fragment patterns were e vident
when trypsin was used instead of Glu-C. The cross-linking
experiments revealed that smelt AFP d imerizes in the
presence or absence of EDTA, indicating that the protein

remains dimerized in the absence of bound Ca
2+
ions. This
may protect susceptible surfaces from protease digestion in
smelt AFP, with or without Ca
2+
.
An interesting difference between smelt AFP a nd the
other known AFPs of fish, including the other type II AFPs,
is the presence of an N-linked oligosaccharide located on
1224 J. C. Achenbach and K. V. Ewart (Eur. J. Biochem. 269) Ó FEBS 2002
Asn18 [19]. The biological roles of the N-linked oligosac-
charides from many proteins have been studied (reviewed in
[28,29]). N-linked oligosaccharides have been shown to
enhance the thermal stability o f proteins, modulate and
stabilize protein secondary structures such as b turns,
mediate interc ellular t ransport of polypeptides, modulate
protein half-life, and facilitate protein–protein interactions
[28–31]. To examine possible roles of the N-linked oligo-
saccharide o f smelt AFP in native protein structure and
function, deglycosylated and untreated smelt AFP were
compared in t erms of dimerization, Ca
2+
binding, protease
resistance, and antifreeze activity. Enzymatic cleavage with
N-glycosidase F resulted in an apparent reduction in
molecular mass of % 5 kDa. The removal of the carbohy-
drate moiety had no effect on any of the characteristics of
smelt AFP that were investigated. However, the possibility
of alternative and untested roles of the N -linked oligosac-

charide such a s enhanced protein half-life or recognition by
endogenous lectins in vivo cannot be ruled out.
Rainbow smelt are unique among bony fishes in that they
have been shown to p roduce large amounts o f glycerol
(upto0.4
M
) in response to subzero temperatures. At these
concentrations, glycerol generates a substantial colligative
freezing point depression and depresses the body freezing
temperature of the smelt along with the noncolligative
activity of the smelt AFP [32]. Although smelt AFP clearly
contributes to the depression of the serum freezing tem-
perature, it does not appear to be the major contributor to
this effect. It would therefore be interesting to determine
whether the AFP has a separate undiscovered activity in
smelt plasma, in which the N-linked oligosaccharide and
dimeric character play more central roles.
Smelt AFP has been characterized in terms of Ca
2+
binding, function of the N-linked o ligosaccharide, and
quaternary structure. This study demonstrates that Ca
2+
imparts a con formational change w hen bound to smelt AFP
in the same w ay as for h erring AFP. However, unlike
herring AFP, the core CTLD of smelt AFP is protease
resistant even without Ca
2+
bound. Smelt AFP is unusual
among general AFPs in that it is N-glycosylated and
dimeric. The smelt type II AFP does not bind common

carbohydrates as many lectins do [7], but its similarity to the
soluble C-type lec tins is evident in its structural character-
istics and function. The mechanistic implications o f the
observed dimerization on smelt AFP ice binding requires
further study.
ACKNOWLEDGEMENTS
We thank Devanand Pinto (NRC IMB) for helpful review of the
manuscript. We also thank a number of IMB colleagues for their kind
assistance, including Robert Richards for helpful d iscussion, Shawna
MacKinnon for use of her HPLC, Denise LeBlanc for loan of a size
exclusion HPLC column, Steve Locke for trypsin, and Neil Ross for
time on his fluorimeter and patient instruction in software use for the
instrument. We are grateful to A/F Protein Canada Inc. for their
generous donation of smelt blood plasma. This research was supported
by NRC IMB. This is NRC p ublication number 42345.
REFERENCES
1. Yeh, Y. & Feeney, R.E. (1996) Antifreeze proteins: structures and
mechanisms of function. Chem. Rev. 96, 601–617.
2. Ewart, K.V., Lin, Q. & Hew, C.L. (1999) Structure, function and
evolution of antifreeze proteins. Cell. Mol. Life Sci. 55, 271–283.
3. Hew, C.L., Joshi, S., Wang, N.C., Kao, M.H. & Anantha-
narayanan, V.S. (1985) Structures of shorthorn sculpin antifreeze
polypeptides. Eur. J. Biochem. 151, 167–172.
4. Yang, D.S., Sax, M., C hakrabartty, A. & Hew, C.L. (1988)
Crystal structure of an antifreeze polypeptide and its m echanistic
implications. Nature (London) 333, 232–237.
5. Fletcher, G., Hew, C. & Davies, P. (2001) Antifreeze p roteins of
teleost fishes. Annu. Rev. Physiol. 63, 359–390.
6. Deng, G., Andrews, D.W. & Laursen, R.A. (1997) Amino acid
sequence of a new type of antifreeze protein, from the

longhorn sculpin Myoxocephalus octodecimspinosis. FEBS Lett.
402, 17–20.
7. Ewart, K.V. & Fletcher, G.L. (1993) Herring antifreeze protein:
primary structure and evidence f or a C-type lectin evolutionary
origin. Mol. Mar. Biol Biotechnol. 2, 20–27.
8. Gronwald, W., Loewen, M.C., Lix, B., Daugulis, A.J., Sonnich-
sen, F.D., Davies, P.L. & Sykes, B.D . (1998) The solution struc-
ture of type II antifreeze protein reveals a new member of the lectin
family. Biochemistry 37, 4712–4721.
9. Drickamer, K. (1999) C-type lectin-like domains. Curr. Opin.
Struct. Biol. 9, 5 85–590.
10. Weis, W.I., K ahn, R., Fourme, R., Drickamer, K . & Hen-
drickson, W.A. (1991) Structu re of the calcium -dependent lectin
domain from a rat mannose-binding protein determined by MAD
phasing. Science 254, 1608–1615.
11. Ewart, K.V. & Fletcher, G.L. (1990) Isolation a nd characterization
of antifreeze proteins from smelt ( Osmerus mordax) and Atlantic
herring (Clupea harengus harengus). Can. J. Zool. 68, 1652–1658.
12. Ewart, K.V., Li, Z., Yang, D.S., Fletcher, G.L. & Hew, C.L.
(1998) The ice-binding site of Atlantic herring antifreeze protein
corresponds to the carbohydrate-binding site of C-type lectins.
Biochemistry 37, 4080–4085.
13. Ewart, K.V., Rubinsky, B. & Fletcher, G.L. (1992) Structural
and functional similarity between fish antifreeze proteins and
calcium-dependent lectins. Biochem. Biophys. Res. Commun. 185 ,
335–340.
14. Charuk, J.H., Pirraglia, C.A. & Reithmeier, R.A. (1990) Interac-
tion of ruthenium red with Ca
2+
-binding proteins. Anal. Biochem.

188, 123–131.
15. Ewart, K.V., Yang, D.S., Ananthanarayanan, V.S., Fletcher, G.L.
& Hew, C.L. (1996) Ca
2+
-dependent antifreeze proteins. Mod u-
lation of conformation and a ctivity b y divalent metal ions. J. Biol.
Chem. 271, 16627–16632.
16. DeLuca, C.I., Comley, R. & Davies, P.L. (1998) Antifreeze pro-
teins bind independently to ice. Biophys. J . 74, 1502–1508.
17. Miura, K., Ohgiya, S., Hoshino, T., Nemoto, N., Suetake, T.,
Miura, A., Spyracopoulos, L., Kondo, H. & Tsuda, S. (2001)
NMR analysis of type III antifreeze protein intramolecular dimer.
Structural basis for enhanced activity. J. Biol. Chem. 276, 1304–
1310.
18. Yu, Y.M. & Griffith, M. (1999) Antifreeze proteins in winter rye
leaves form oligomeric complexes. Plant Physiol. 119, 1361–1370.
19. Ewart, K.V. (1993) Type II antifreeze proteins from smelt
(Osmerus Mordax) and Atlantic herring (Clupea Harengus
Harengus): similarity to the C-type le ctin family. Ph D thesis,
Memorial University of Newfoundland, St J ohn’s.
20. Reference withdrawn.
21. Loewen, M.C., Gronwald, W., Sonnichsen, F.D., Sykes, B.D. &
Davies, P.L. (1998) The ice-binding site of sea raven antifreeze
protein i s distinct from the ca rbohydrate-binding site of th e
homologous C-type lectin. Biochemistry 37, 17745–17753.
22. Poget, S.F., Legge, G.B., Proctor, M.R., Butler, P.J., Bycroft, M.
& Williams, R.L. ( 1999) The structure o f a tunicate C-type lectin
from Polyandrocarpa misa kiensis complexe d wit h
D
-galactose.

J. Mol. Biol. 290, 867–879.
Ó FEBS 2002 Characterization of smelt antifreeze protein (Eur. J. Biochem. 269) 1225
23. Mizuno, H., Fujimoto, Z., Koizumi, M., Kano, H., Atoda, H. &
Morita, T. ( 1997) Structure of coagulation factors IX/X-binding
protein, a heterodimer of C-type lectin domains. Nat. Struct. Biol.
4, 438–441.
24. Boyington, J.C., Riaz, A.N., Patamawenu, A., Coligan, J .E.,
Brooks, A.G. & Sun, P.D. (1999) Structure of CD94 reveals a
novel C-type lectin fold: i mplications for the NK cell-associated
CD94/NKG2 receptors. Immunity 10 , 75–82.
25. Llera, A.S., Viedma, F., Sanchez-Madrid, F. & Tormo, J. (2000)
Crystal structure of the C-type lectin-like domain from the human
hematopoietic cell receptor C D69. J. Bio l. Chem . 276, 7312–7319.
26. Weis, W.I. & Drickamer, K. (1996) Structural basis of lectin-
carbohydrate recognition. Annu. Rev. Biochem. 65, 441–473.
27. Loeb, J.A. & Drickamer, K. ( 1988) Conformational changes in the
chicken receptor for endocytosis of glycoproteins. Modulation
of ligand-binding activity by Ca
2+
and pH. J. Biol. Chem. 263,
9752–9760.
28. Varki, A. (1993) Biological roles of oligosaccharides: all of the
theories are correct. Glycobiology 3, 97–130.
29. O’Connor, S.E. & Imperiali, B. (1996) Modulation of protein
structure and function by asparagine-linked glycosylation. Chem.
Biol. 3, 803–812.
30. Wang, C., Eufemi, M., Turano, C. & Giartosio, A. (1996) Influ-
ence of the carbohydrate moiety on the stability of glycoproteins.
Biochemistry 35, 7299–7307.
31. Dwek, R.A., Lellouch, A.C. & Wormald, M.R. (1995) Glyco-

biology: Ôthe function of sugar in the IgG moleculeÕ. J. Anat. 187,
279–292.
32. Raymond, J.A. (1992) Glycerol is a colligative antifreeze in some
northern fishes. J. Exp. Zool. 262 , 347–352.
1226 J. C. Achenbach and K. V. Ewart (Eur. J. Biochem. 269) Ó FEBS 2002