Co-operative effect of the isoforms of type III antifreeze
protein expressed in Notched-fin eelpout, Zoarces
elongatus Kner
Yoshiyuki Nishimiya
1
, Ryoko Sato
1
, Manabu Takamichi
2
, Ai Miura
1
and Sakae Tsuda
1,2
1 Functional Protein Research Group, Research Institute of Genome-based Biofactory (RIGB), National Institute of Advanced Industrial
Science and Technology (AIST), Sapporo, Japan
2 Division of Biological Sciences, Graduate School of Science, Hokkaido University, Sapporo, Japan
Antifreeze protein (AFP) possesses the unique ability
to bind to the surface of ice crystals, which permits
growth of ice at limited open spaces on the surface,
leading to the formation of numerous convex ice surfa-
ces between the bound AFPs [1]. The growing ice sur-
face becomes energetically unfavorable for further
absorption of water molecules proportionately with the
surface curvature, leading to termination of ice growth
(Kelvin effect) [2]. This AFP-induced inhibition of ice
crystal growth can be detected macroscopically as a
depression in the freezing temperature (T
f
) of the
solution without alteration of the melting temperature
(T

reported here have been deposited in the
DDBJ database under the accession
numbers AB188389–AB188401.
(Received 27 August 2004, revised 9
November 2004, accepted 17 November
2004)
doi:10.1111/j.1742-4658.2004.04490.x
We found that Notched-fin eelpout, which lives off the north east coast
of Japan, expresses an antifreeze protein (AFP). The liver of this fish
contains DNAs that encode at least 13 type III AFP isoforms (denoted
nfeAFPs). The primary sequences of the nfeAFP isoforms were categor-
ized into SP- and QAE-sephadex binding groups, and the latter were fur-
ther divided into two subgroups, QAE1 and QAE2 groups. Ice crystals
observed in HPLC-pure nfeAFP fractions are bipyramidal in shape with
different ratios of c and a axes, suggesting that all the isoforms are able
to bind ice. We expressed five recombinant isoforms of nfeAFP and ana-
lyzed the thermal hysteresis (TH) activity of each as a function of pro-
tein concentration. We also examined the change in activity on mixing
the isoforms. TH was estimated to be 0.60 °C for the QAE1 isoform,
0.11 °C for QAE2, and almost zero for the SP isoforms when the con-
centrations of these isoforms was standardized to 1.0 mm. Significantly,
the TH activity of the SP isoforms showed concentration dependence in
the presence of 0.2 mm QAE1, indicating that the less active SP isoform
becomes ‘active’ when a small amount of QAE1 is added. In contrast, it
does not become active on the addition of another SP isoform. These
results suggest that the SP and QAE isoforms of type III AFP have dif-
ferent levels of TH activity, and they accomplish the antifreeze function
in a co-operative manner.
Abbreviations
AFGP, antifreeze glycoprotein; AFP, antifreeze protein; nfeAFP, type III AFP from Notched-fin eelpout; TH, thermal hysteresis.

)1
, TH activity of AB1 is 1.27 °C, which is
slightly higher than 1.17 °C of AB2. Lycodichthys
dearborni also has three major AFPs (RD1, RD2, and
RD3) [21]. RD3 is an exceptional isoform which com-
prises two type III AFP domains connected in tandem
through a nine-residue linker. The sequential identity
between RD1 and RD2 is 94%, and the former shares
98% and 85% identities with the N-terminal and
C-terminal domains of RD3, respectively (for RD2,
the corresponding identities are 94% and 77%,
respectively). For RD1 and RD2, the amino acid
replacements of the hydrophobic residues (20th and
41st) have been identified for the ice-binding surface.
At a concentration of 10 mgÆmL
)1
, the TH activities of
RD1 and RD2 are almost identical (0.95 °C and
0.90 °C, respectively), whereas RD3 has a slightly
lower activity of 0.81 °C. Overall, the type III AFP
isoforms differ in their hydrophobic residues but not
significantly in their hydrophilic residues with regard
to the ice-binding surface. One may speculate that
these differences may differentiate the antifreeze func-
tions of the isoforms. However, not much is known
about the relationship between TH activity and the
isoforms, especially the existence of a co-operative
effect of the isoforms in the QAE and SP groups.
We recently found that a significant amount of type
III AFP can be purified from the minced muscles of

Notched-fin eelpout by cation-exchange chromatography (High-S
column). A linear gradient (dotted line) from 0 to 500 m
M NaCl at a
flow rate of 1 m LÆmin
)1
was used to elute the fractions containing
the isoforms of nfeAFP.
Y. Nishimiya et al. Co-operative effect of type III AFP isoforms
FEBS Journal 272 (2005) 482–492 ª 2004 FEBS 483
nfeAFP isoforms was eluted by application of a con-
centration gradient of NaCl ( 50–250 mm). Their
activity was confirmed by photomicroscopic observa-
tion of the bipyramidal ice crystal. The mixture of
nfeAFPs migrated as a  4.5-kDa band on SDS ⁄
PAGE, which is smaller than the actual molecular
mass ( 7 kDa), as previously observed [5,22]. The
mixture of the two fractionated nfeAFPs was resolved
into six major and eight minor peaks labeled 1–14 in
Fig. 2A by RP-HPLC using a C
18
reverse-phase col-
umn. The molecular mass of  6600 was estimated to
the peaks 1–10, and  7000 to the peaks 11–14 by
MALDI-TOF MS. An elongated bipyramidal ice crys-
tal in the c-axis direction was observed (Fig. 2B) for
peaks 1–10 eluted in the 41–54% concentration range
of acetonitrile (the hexagonal shape observed for peak
1 is attributable to the low protein concentration). In
contrast, a thick bipyramidal ice crystal was detected
for peaks 11–14 eluted in the range above 54% aceto-

of nfeAFPs
The complete amino acid sequences of the 13 isoforms
of nfeAFP were determined based on the cDNA
A
B
Fig. 2. (A) HPLC purification of the isoforms
of type III AFP from Notched-fin eelpout. The
peaks containing the HPLC-pure isoforms are
labeled 1–14. The amino acid sequence was
analysed for peaks 2 and 8 (Figs 3 and 4). (B)
Photomicrographs of the single ice crystal
observed for each HPLC-pure isoform
dissolved in 0.1
M NH
4
HCO
3
(pH 7.9).
Co-operative effect of type III AFP isoforms Y. Nishimiya et al.
484 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS
library as described in Experimental procedures. The
N-terminal residue of the 13 intact AFPs was
determined by reference to the ordinary type III AFP
sequences and the identified signal sequences of
A1 and A2: i.e. MKSVILTGLFFVLLCVDHMSSA
for nfeAFP11 and 12 and MKSVILTGLLFVLLCVD
HMSSA for nfeAFP1–10, 13. The signal sequence of
nfeAFP2 was only partially identified from cDNA.
The sequence A1 is identical with the 1st 65 residues
of nfeAFP6 with the 66th lysine residue at the C-ter-

ity, Gly1 of the SP group is defined as the 2nd residue
in the present study. We further divided the QAE iso-
forms into two subgroups, QAE1 (nfeAFP7–10) and
QAE2 (nfeAFP11–13), which are distinguished by 10
characteristic residues colored blue and green in Fig. 4.
The sequence identity within the 13 isoforms is  48%.
The identity is 77% when compared within the SP iso-
forms, 76% within the QAE1 isoforms, and 91%
within the QAE2 isoforms, respectively. With regard
to the putative ice-binding residues indicated with
asterisks (*) in Fig. 4, the 42nd residue is different
between the SP and QAE groups. In addition, Gln9,
Leu19, Val20 and Val41 in the QAE1 group are
replaced by Val9, Val19, Gly20 and Ile41, respectively,
in the QAE2 group. It is worth noting that Gln9 is
conserved in all known isoforms of type III AFP,
except HPLC7 which contains Arg9 [20]. Overall, the
hydrophilic residues are mostly conserved among the
nfeAFP isoforms for the ice-binding residues (for
example, Gln9, Asn14, Thr15, Thr18, and Gln44).
However, significant amino acid replacements are iden-
tified for the hydrophobic residues located at the 13th,
19th, 20th, and 41st positions.
Antifreeze activity of recombinant nfeAFPs
To examine the relationship between TH activity and
sequence diversity of type III AFPs, the following five
isoforms, listed in Fig. 4, were expressed and purified:
nfeAFP2 (SP group); nfeAFP6, a major isoform of
the muscle homogenate (SP group); nfeAFP8, the
sequence of which is similar to HPLC12 (QAE1

sequence identity with the isoforms HPLC6 and
HPLC12 from M. americanus, respectively (the amino
acid sequences of HPLC6 and HPLC12 are listed in
Fig. 4). Again both nfeAFP2 and nfeAFP6 were
expressed with C-terminal lysines. We also expressed
nfeAFP6minusLys66 (nfeAFP6DLys; SP group) to
examine the effect of the C-terminal lysine on activ-
ity. For nfeAFP13, TH activity was determined in the
presence and absence of reductant (dithiothreitol), as
it can form multimers via intermolecular disulfide
bridges (Fig. 5).
Figure 6 shows the molar concentration dependence
of TH activity for the six genetically produced iso-
forms of nfeAFP examined using the Vogel osmo-
meter. A nonlinear profile of TH activity typical of
ordinary AFPs was identified for nfeAFP8 (QAE1
group), although its maximum activity ( 0.7 °C) was
slightly lower than that reported for HPLC12, which
was determined using the Clifton nanoliter osomome-
ter [7]. A similar profile was detected for nfeAFP13
(QAE2 group) in the absence of reductant. The addi-
tion of reductant significantly lowered the activity of
nfeAFP13, indicating that the monomer is less active
than when a small amount of multimers is present
(Fig. 5). An extremely low level of TH activity was
detected for another QAE2 isoform, nfeAFP11. We
detected no appreciable TH activity for nfeAFP2,
nfeAFP6, and nfeAFP6DLys (SP group). The lack of a
significant difference between nfeAFP6 and
nfeAFP6DLys suggests that a lysine at the C-terminus

M nfeAFP13 in the presence of 10 mM of dithiothrei-
tol (DTT); lane B, 0.3 m
M nfeAFP13 in the absence of dithiothreitol.
The protein standards (MW) are indicated on the left. The mono-
mer of nfeAFP13 is the dominant species irrespective of the addi-
tion of reductant.
Co-operative effect of type III AFP isoforms Y. Nishimiya et al.
486 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS
0.01–0.05 °CÆmin
)1
, suggesting that these species have
the ability to inhibit ice growth.
We further examined whether the addition of a
small amount (0.2 mm) of ‘active’ nfeAFP8 influences
the TH activity of ‘less active’ nfeAFP6. Approxi-
mately 0.10 °C of TH activity was detected (Fig. 6).
As shown in Fig. 7, the TH activity of the less active
nfeAFP6 shows clear concentration dependence in the
presence of 0.2 mm nfeAFP8 (maximum TH ¼
0.60 °C). A similar TH profile was obtained for
nfeAFP6DLys in the presence of 0.2 mm of nfeAFP8.
These data indicate that ‘less active’ AFP isoforms can
exert a substantial level of antifreeze activity after the
addition of a small amount of ‘active’ isoform. The
less active nfeAFP6 does not, however, become active
after the addition of another less active isoform,
nfeAFP2 (Fig. 7). No significant difference was detec-
ted between nfeAFP6 and nfeAFP6DLys, which con-
firms that the C-terminal lysine does not directly
participate in the antifreeze function.

of nfeAFP, whereas the others correspond to SP iso-
forms, and the TH activity of the QAE isoform is
higher than that of the SP isoform.
Fig. 6. TH activity measured using an osmometer (model OM 802;
Vogel) as a function of concentration (m
M) of type III AFP iso-
forms, nfeAFP2 (h), nfeAFP6 (r), nfeAFP6DLys (e), nfeAFP8
(·), nfeAFP11 (s), nfeAFP13 in the absence of dithiothreitol (n),
and nfeAFP13 in the presence of dithiothreitol (m). The measure-
ment was repeated three times using fresh samples, and mean
values were plotted with error bars.
Fig. 7. TH activity measured as a function of concentration (mM) of
type III AFP isoforms: nfeAFP6 (m) and nfeAFP6DLys (s) in the
presence of 0.2 m
M nfeAFP8; nfeAFP6 (r) and nfeAFP6DLys (h)in
the presence of 0.2 m
M nfeAFP2. The measurements were repea-
ted three times using fresh samples and mean values were plotted
with error bars.
Y. Nishimiya et al. Co-operative effect of type III AFP isoforms
FEBS Journal 272 (2005) 482–492 ª 2004 FEBS 487
Although we identified ice-shaping activity for the
SP isoforms of nfeAFP (Fig. 2), their TH activities
were below the level of detection of the instrument
used (Vogel osmometer) (Fig. 6). TH activity could
also not be detected for 15Eklac, a 15-residue syn-
thetic peptide corresponding to the 11-residue repeat-
ing unit of a 36-residue type I AFP, using the Clifton
nanoliter osmometer [24]. This minimized peptide
forms a vertex-flat bipyramid of ice crystal, for which

L19A ⁄ V41A [18]. Leu19 and Val20 of the QAE1 iso-
form are replaced by Pro19 and Ala20 in the presently
examined SP isoforms, nfeAFP2 and nfeAFP6. The
replacement of these residues presumably alters
the ice-binding character of the AFP isoform. A ‘semi-
reversible’ ice-binding model for the kinetics of AFP-
induced ice growth inhibition has been proposed
[25,26], which includes the following adsorption steps
of AFP: (a) attachment to the ice–water interface;
(b) rearrangement of adsorbed molecules by diffu-
sion, reorientation, and ⁄ or conformational change; (c)
detachment from the interface.
Again ice-shaping ability is suggested to be a charac-
teristic of all isoforms of nfeAFP (Fig. 2). The hydro-
phobic residues located at positions 9, 13, 19, 20, 41,
and 42 are structured so as to surround the ice-binding
surface in the 3D structure of type III AFP (PDB Code
¼ 1MSI). Hence, it can be speculated that a set of
hydrophobic residues in an isoform differentiates the
surface complementarity with the target plane of the ice
crystal, which affects adsorption steps (b) and (c) espe-
cially, resulting in a different level of TH activity.
A clear concentration dependence of TH activity
was observed for a QAE1 isoform (nfeAFP8) simi-
larly to HPLC12, whereas it was below detectable
level for the QAE2 isoforms (nfeAFP11 and
nfeAFP13) (Fig. 6). It should be mentioned that
Gln9, a highly conserved residue in the known type
III AFPs, is replaced with Val9 in the QAE2 iso-
forms. The Cys-containing QAE2 isoform, nfeAFP13,

markedly enhanced by the addition of the larger spe-
cies ( 10.5 kDa) [31].
We found that TH activity of a SP isoform,
nfeAFP6, is greatly enhanced by, and showed clear
concentration dependence on, the addition of a small
amount of a QAE1 isoform, nfeAFP8 (Fig. 7). This is
similar to the case of the QAE2 isoform, nfeAFP13;
the activity of its monomer was enhanced by the pres-
ence of a small amount of the multimer. Although
Co-operative effect of type III AFP isoforms Y. Nishimiya et al.
488 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS
there is no direct experimental evidence to explain the
mechanism, one can assume the following co-operative
ice-binding mechanism: (a) the ‘active’ AFP isoform
(QAE1) firstly adsorbs to the ice crystal, which decrea-
ses its growing speed and lowers the energy barrier to
allow adsorption of the ‘less active’ isoform; (b) the
less active isoform (SP or QAE2) can then adsorb to
the ‘open space’ between the prebound AFPs of the ice
crystal surface; (c) most of the nfeAFP isoforms
adsorb to the growth-terminated ice crystal in the final
state. This hypothesis is comparable to that of Bur-
cham et al. [31]. They assumed that stabilization of the
antifreeze action of small (weak) species of AFGP
(AFGP6-8) by large (strong) species (AFGP1-5) pro-
duces co-operative coverage of a seed ice crystal,
thereby preventing further crystal growth. When we
added a less active SP isoform (nfeAFP2) to another
less active SP isoform (nfeAFP6) (Fig. 7), the
co-operative ice binding did not occur, resulting in no

new isoforms of type III AFP from Notched-fin eel-
pout, which were categorized into three groups: SP,
QAE1, and QAE2. We detected a clear difference in
TH activity between the isoforms, although ice-binding
ability was detected for all of them. This was ascribed
to differences in hydrophobic residues located in the
ice-binding region. The less active SP isoform becomes
active on addition of a small amount of the active
QAE1 isoform, whereas it does not become active on
addition of another less active SP isoform. These
results suggest that isoforms of type III AFP co-opera-
tively exert the antifreeze function.
Experimental procedures
Purification and sequence analysis of nfeAFPs
Type III AFP was purified from the muscle of Notched-fin
eelpout. After removal of the head and gut, the meat of the
fish was homogenized with water using an electric mixer
[tissue ⁄ water ratio (g) ¼ 1 : 1]. The homogenate was centri-
fuged at 3000 g for 30 min, and the supernatant obtained
dialyzed against 50 mm sodium acetate (pH 3.7) overnight
at 4 °C. After removal of the precipitate formed during
dialysis, the AFP-containing solution was loaded on to a
high-S column (1.0 · 5.0 cm; Bio-Rad, Hercules, CA,
USA), and the column-bound AFPs were eluted with a lin-
ear NaCl gradient (0–0.5 m )in50mm sodium acetate buf-
fer (pH 3.7). The fractions containing the isolated AFP
were collected and further chromatographed by RP-HPLC
using a C
18
reverse-phase column (TOSOH, Tokyo, Japan;

then isolated from the liver using an RNeasy Protect kit
(Qiagen). mRNA was purified from total RNA using the
Oligotex-dT30 mRNA Purification kit (TaKaRa). A cDNA
library was generated from 1.6 lg mRNA with the
Y. Nishimiya et al. Co-operative effect of type III AFP isoforms
FEBS Journal 272 (2005) 482–492 ª 2004 FEBS 489
ZAP-cDNA Synthesis kit (Stratagene, La Jolla, CA, USA).
PCR was performed for a major cDNA consisting of
500 bp purified from the cDNA library using the templates
of Ex-Taq DNA polymerase (TaKaRa), oligo-dT linker
primer (5¢-GAGAGAACTAGTCTCGAGTTT-3¢), and the
synthetic primer of the adapter sequence (5¢-TCGGG
AATTCGGCACGAGG-3¢). The annealing sites of these
primers were connected to 3 ¢-terminus and 5¢-terminus of
cDNA. The PCR conditions are as follows: denaturing at
94 °C for 1 min, 2 cycles pre-extending at 94 °C for 1 min,
extending at 56 °C and 72 °C for 1 min each, and 28 cycles
extending at 94 °C, 50 °C, and 72 °C for 1 min each. The
PCR products obtained were purified and ligated into
pGEM-T Easy (Promega, Madison, WI, USA). The cloned
DNAs encoding nfeAFP isoforms were sequenced using the
ABI Prism Big dye terminator cycle sequencing kit and
ABI 3100 genetic analyzer (Applied Biosystems).
Expressions and purification of the five
recombinant nfeAFPs
The five DNA fragments encoding nfeAFP2, 6, 8, 11, and
13, from which the signal sequence was removed, were
amplified by PCR using cloning plasmid vectors, and
ligated into pET20b (Novagen) with the restriction
enzymes, NdeI and XhoI. The plasmid-DNAs obtained were

HCO
3
(pH 7.9). The purity was checked by SDS ⁄ PAGE (16% gel)
[34]. It should be noted that the purified nfeAFP13
appeared to form multimers via intermolecular disulfide
bridges; nfeAFP forms a monomer in the presence of
reductant (+ dithiothreitol), while its trimer and tetramer
were also generated in the absence of reductant as shown in
Fig. 5. Therefore, for the measurement of TH activity,
nfeAFP13 was reduced for 12 h at 4 °C with 0.1 m
NH
4
HCO
3
(pH 7.9) containing 10 mm dithiothreitol, and
activity was measured on fresh samples.
Measurement of ice crystal morphology
and TH activity
Ice-crystal morphology was observed using an in-house
photomicroscope system consisting of a Leica DMLB 100
photomicroscope equipped with a Linkam LK600 (liquid
nitrogen-type) temperature controller and a CCD camera.
A droplet ( 0.5 lL) of the sample solution was frozen and
then heated until a single ice crystal was observed sepa-
rately in the solution by manipulation of the temperature
controller. The change in morphology of a single ice crystal
into a hexagonal bipyramid caused by the accumulation of
AFP on the ice surfaces was then observed at a cooling rate
of 0.05 °C per minute.
The Clifton nanoliter osmometer (Clifton Technical Phys-

and
T
m
was repeated three times using fresh samples, and mean
values were used for determination of TH activity (TH ¼|T
m
– T
f
|). It has been documented that the TH value determined
using this osmometer is slightly lower than that determined
using the Clifton nanoliter osmometer [13,27,30].
Acknowledgements
We thank Dr Tamotsu Hoshino, Michiko Kiriaki, and
Mineko Fjiwara for analysis of amino acid sequences,
and Yumika Miura for analysis of DNA sequences.
References
1 Knight CA, Cheng CC & DeVries AL (1991) Adsorp-
tion of alpha-helical antifreeze peptides on specific ice
crystal surface planes. Biophys J 59, 409–418.
Co-operative effect of type III AFP isoforms Y. Nishimiya et al.
490 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS
2 Raymond JA & DeVries AL (1977) Adsorption inhibi-
tion as a mechanism of freezing resistance in polar
fishes. Proc Natl Acad Sci USA 74, 2589–2593.
3 Cheng CH & DeVries AL (1989) Structures of antifreeze
peptides from the antarctic eel pout, Austrolycicthys
brachycephalus. Biochim Biophys Acta 997, 55–64.
4 Schrag JD, Cheng CH, Panico M, Morris HR &
DeVries AL (1987) Primary and secondary structure of
antifreeze peptides from arctic and antarctic zoarcid

11 Graether SP, DeLuca CI, Baardsnes J, Hill GA, Davies
PL & Jia Z (1999) Quantitative and qualitative analysis
of type III antifreeze protein structure and function.
J Biol Chem 274, 11842–11847.
12 Antson AA, Smith DJ, Roper DI, Lewis S, Caves LS,
Verma CS, Buckley SL, Lillford PJ & Hubbard RE
(2001) Understanding the mechanism of ice binding by
type III antifreeze proteins. J Mol Biol 305, 875–889.
13 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 intramole-
cular dimer. Structural basis for enhanced activity.
J Biol Chem 276, 1304–1310.
14 Ko TP, Robinson H, Gao YG, Cheng CH, DeVries AL
& Wang AH (2003) The refined crystal structure of an
eel pout type III antifreeze protein RD1 at 0.62-A
˚
reso-
lution reveals structural microheterogeneity of protein
and solvation. Biophys J 84, 1228–1237.
15 Chao H, So
¨
nnichsen FD, DeLuca CI, Sykes BD &
Davies PL (1994) Structure–function relationship in the
globular type III antifreeze protein: identification of a
cluster of surface residues required for binding to ice.
Protein Sci 3, 1760–1769.
16 DeLuca CI, Davies PLYe, Q & Jia Z (1998) The effects
of steric mutations on the structure of type III anti-
freeze protein and its interaction with ice. J Mol Biol

PL (1996) Effect of type III antifreeze protein dilution
and mutation on the growth inhibition of ice. Biophys J
71, 2346–2355.
24 Houston ME Chao H Hodges RS Sykes BD Kay CM
So
¨
nnichsen FD Loewen MC & Davies PL (1998) Bind-
ing of an oligopeptide to a specific plane of ice. J Biol
Chem 273, 11714–11718.
25 Hew CL & Yang DS (1992) Protein interaction with ice.
Eur J Biochem 203, 33–42.
26 Wen D & Laursen RA (1992) A model for binding of an
antifreeze polypeptide to ice. Biophys J 63, 1659–1662.
27 Nishimiya Y Ohgiya S & Tsuda S (2003) Artificial mul-
timers of the type III antifreeze protein. Effects on ther-
mal hysteresis and ice crystal morphology. J Biol Chem
278, 32307–32312.
28 Chao H Hodges RS Kay CM Gauthier SY & Davies
PL (1996) A natural variant of type I antifreeze protein
with four ice-binding repeats is a particularly potent
antifreeze. Protein Sci 5, 1150–1156.
29 Leinala EK Davies PL Doucet D Tyshenko MG
Walker VK & Jia Z (2002) A beta-helical antifreeze pro-
tein isoform with increased activity. Structural and func-
tional insights. J Biol Chem 277, 33349–33352.
Y. Nishimiya et al. Co-operative effect of type III AFP isoforms
FEBS Journal 272 (2005) 482–492 ª 2004 FEBS 491
30 Baardsnes J Kuiper MJ & Davies PL (2003) Antifreeze
protein dimer: when two ice-binding faces are better
than one. J Biol Chem 278, 38942–38947.