Tài liệu Báo cáo khoa học: Co-operative effect of the isoforms of type III antifreeze protein expressed in Notched-fin eelpout, Zoarces elongatus Kner ppt

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

m
)(|T
f
) T
m
|). This is generally termed thermal hys-
teresis (TH).
AFPs categorized as type III ( 7 kDa) have been
identified in blood serum of fish living in polar sea-
water with a year-round temperature of  )1.8 °C:
Arctic and Antarctic eelpouts (Austrolycicthys brachy-
cephalus, Lycodes polaris, and Lycodichthys dearborni),
Atlantic ocean pout (Macrozoarces americanus), and
Atlantic wolffish (Anarhichas lupus) [3–6]. Type III
AFP forms a globular shape characterized by internal
Keywords
co-operative effect; Notched-fin eelpout;
type III antifreeze protein
Correspondence
S. Tsuda, Functional Protein Research
Group, Research Institute of Genome-based
Biofactory (RIGB), National Institute of
Advanced Industrial Science and Technology
(AIST), 2-17-2-1 Tsukisamu-Higashi,
Toyohira, Sapporo 062-8517, Japan
Fax: +81 11 857 8983
Tel: +81 11 857 8912
E-mail:
Note
The nucleotide and protein sequences

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.

482 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS
twofold symmetry of the ‘pretzel’ fold [7–14], which
provides a markedly flat amphipathic ice-binding
surface contributed by residues 9–10, 13–16, 18–21, 41,
42, 44 and 61 [15–18]. A large database of amino acid
sequences is available for the various isoforms of type
III AFP. The sequences of at least 12 isoforms
(denoted HPLC1–12) have been determined for
Macrozoarces americanus, which were originally categ-
orized into QAE-Sephadex-binding and SP-Sephadex-
binding groups [19]. The best-characterized AFP,
HPLC12, is the only a member of the QAE group, all
others (HPLC1–11) belonging to the SP group. The
QAE and SP isoforms share only 55% identity among
their primary sequences, although there is 90% identity
among the SP isoforms [20]. Immunological cross-reac-
tivity is also different between the QAE and SP iso-
forms [5]. Post-translational modification has been
reported only for the SP isoforms [19]. Of the residues
that form the ice-binding surface, the hydrophilic ones
are almost uniformly conserved in the QAE and SP
isoforms, whereas the hydrophobic residues tend to be
changeable. The same has been found for the AB1 and
AB2 isoforms from Austrolycicthys brachycephalus [3].
They share 83% sequential identity with the same
hydrophilic residues at the ice-binding surface, whereas
the hydrophobic residues Pro19 and Ala20 located on
the surface of AB1 are replaced with Ile19 and Val20
in AB2, respectively. At a concentration of
20 mgÆmL

)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 ( Zoarces elongatus Kner), which
lives off the north east coast of Japan ( 40° of lati-
tude) where the temperature of the seawater is )1.7 °C
in mid-winter. This fish naturally expresses at least 13
AFP isoforms (nfeAFPs). The primary sequences of
two major species were examined using a protein se-
quencer, and 13 sequences were independently deter-
mined by analysis of cDNA sequences using the
mRNA library constructed for the fresh liver of this
fish. By comparing all of the identified sequences of
nfeAFPs with those of ordinary type III AFP hetero-
geneity [19], 13 isoforms, denoted nfeAFP1–13, were
categorized into SP and QAE groups. The latter were
further divided into QAE1 and QAE2 groups which
are mainly distinguished by 10 characteristic residues.
The QAE1 isoforms exhibit high sequence similarity to
a well-known QAE isoform, HPLC12, from ocean
pout. Here we expressed five recombinant isoforms,
nfeAFP2 and nfeAFP6 (SP group), nfeAFP8 (QAE1
group), nfeAFP11 and nfeAFP13 (QAE2 group) so as
to compare the TH activity between the groups. TH
activity was also measured for a mixture of the SP and
QAE isoforms to examine their co-operative effect on
antifreeze activity.
Results
Preparation and sequence analysis of nfeAFPs
Figure 1 shows the elution profile of the soluble pro-
teins in the muscle homogenate of Notched-fin eelpout
on the cation-exchange column. The mixture of
Fig. 1. Elution profile obtained for the muscle homogenate 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-

nitrile. We analysed the amino acid sequence for peak
2 in Fig. 2A (later assigned to nfeAFP6) and its cleaved
fragments using a conventional sequence analyzer. As
shown in Fig. 3, the amino acid sequence of residues
1–56 was determined (denoted A1) for a nondigested
component of the peak. Figure 3 also shows the
sequence determinations of five fragments of peak 2
(denoted N1–N5) cleaved by asparaginyl endopeptidase
and five fragments of the peak cleaved by TPCK tryp-
sin (T1–T5). By comparing the cleaved sequences with
each other, the N1 and T1 fragments were assigned to
the N-terminal sequence, and N5 and T5 to the C-ter-
minal sequence. The C-terminus was assigned to Ala
as it would not be cleaved by the proteases used. The
other sequential overlaps between A1, N2, N3, N4,
T2, T3, and T4 allowed us to identify that A1 consists
of 56 of the 64 residues of this isoform, in which the
unknown 35th and 48th residues were assigned to Glu
and Val, respectively. The information on the com-
plete sequence analyzed for peak 2 allowed us to iden-
tify that the isoform belongs to the type III AFP the
N-terminus of which is unblocked, probably the SP
group isoform identified from ocean pout [19]. An
incomplete 62-residue sequence determined for a non-
digested component of peak 8 in Fig. 2A is also
shown in Fig. 3 (labeled A2). It should be noted that
both A1 and A2 have Gly at the N-terminus, which is
similarly identified in the SP isoforms from ocean
pout.
cDNA cloning and sequence alignment

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-

minus, which is presumably removed by the post-
translational processing [19]. Although such processing
can be assumed for nfeAFP1–5 and nfeAFP10 which
have Lys at their C-termini, we assumed a Lys-con-
taining sequence as the complete amino acid sequence
of the isoforms, as no direct evidence has been
obtained so far to indicate their deletion. For
nfeAFP6, the effect of a C-terminal lysine on activity
was examined, as described later. Interestingly, the
sequence A2 is not identical with any of the nfeAFP
sequences listed in Fig. 4, although it shows maximum
similarity to nfeAFP2: Ala23, Ala35, and Met37 of A2
are substituted by Glu, Val, and Ala of nfeAFP2,
respectively. Most of the isoforms are about 65 resi-
dues in length; only nfeAFP13 consists of 70 residues.
nfeAFP13 exceptionally contains two cysteines (Cys64
and Cys66), the sequence of which shows high similar-
ity to an isoform denoted ‘genomic k3’ identified from
ocean pout [20]. The presently identified 13 isoforms
of nfeAFP do not exhibit 100% identity with any
other reported sequences of type III AFP.
Based on the sequence similarity to the known type
III AFP isoforms from Atlantic ocean pout [20], the
13 isoforms of nfeAFP were categorized into SP and
QAE groups as listed in Fig. 4, for which there is a
typical difference in the 34th to 37th residues; therefore
sequence gaps (–) were introduced into positions 37
and 34 for the SP and QAE groups, respectively. A1,
A2, and nfeAFP1–6 were categorized into the SP
group and nfeAFP7–13 into the QAE group. For clar-

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

group); nfeAFP11, a Val9-containing isoform (QAE2
group); nfeAFP13, the largest isoform containing
Peak 2
1 10 20 30 40 50
A1 GESVVATQLIPINTALTPAMMEGKVTNPSGIPFAxMSQIVGKQVNTPxAKGQTLMP
N1 GESVVATQLIPIN
N2 TALTPAMMEGKVTNPSGIPFAxMSQIVGxQVN
N3 TALTPAMMEGKVTN
N4 PSGIPFAEMSQIVGxQVNTPVAxGQTL
N5 MVKTYVPA
T1 GESVVATQLIPINTALTPAMMEGK
T2 VTNPSGIPFAEMSQ
T3 QVNTPVAK
T4 GQTLMPGMVK
T5 TYVPA
Peak 8
1 10 20 30 40 50 60
A2 GQSVVATQLIPMNTALTPAMMAGKVTNPSGIPFAEMSQIVGKQVNVIVAKGQTLMPDMVKTY
Fig. 3. Amino acid sequences of the cleaved
peptides obtained from the HPLC peaks.
A1, 56 amino acid residues of peak 2 deter-
mined by the sequencer; N1–N5, fragments
of peak 2 cleaved by asparaginyl endopepti-
dase; T1–T5, fragments of peak 2 cleaved
by TPCK trypsin; A2, 62 amino acid residues
of peak 8 determined by sequencer.
Y. Nishimiya et al. Co-operative effect of type III AFP isoforms
FEBS Journal 272 (2005) 482–492 ª 2004 FEBS 485
Val9 and two cysteines (Cys64 and Cys66) (QAE2
group). NfeAFP6 and nfeAFP8 exhibit the highest

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

does not affect the ice-binding function, which is con-
sistent with previous indications for ocean pout AFPs
[19]. It is worth noting that nfeAFP2 and nfeAFP6
both form a folded structure, which was evidenced by
a number of secondary shifts observed throughout the
range of their
1
H-NMR spectra. TH activity at a con-
centration of 1.0 mm was 0.60 °C for nfeAFP8 (QAE1
isoform), 0.44 °C (multimers) and 0.11 °C (monomer)
for nfeAFP13 (QAE2 isoform), 0.02 °C for nfeAFP11
(QAE2 isoform), and almost zero for nfeAFPs 2 and 6
(SP isoforms). Formation of an elongated bipyramidal
ice crystal was observed for the solutions of nfeAFP2
(0.3 mm) and nfeAFP6 (3.75 mm) at a cooling rate of
Fig. 4. Alignment of amino acid of sequence
of type III AFP identified from Notched-fin
eelpout (nfeAFP). Spaces are introduced to
optimize the alignment. The yellow color
indicates the conserved residues of the
nfeAFP isoforms. Ten characteristic residues
colored blue and green distinguish the QAE1
and QAE2 groups. The putative ice-binding
residues are indicated with asterisks. The
sequences of HPLC6 and HPLC12 isoforms
from M. americanus are also shown for
comparison.
Fig. 5. SDS ⁄ PAGE (16% gel) of a purified cysteine-containing
QAE2 isoform of type III AFP from Notched-fin eelpout (nfeAFP13).
Lane A, 0.3 m

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.

Discussion
We have identified new isoforms of type III AFP from
Notched-fin eelpout living off the Japanese coast at
 40° of latitude. The isoforms were categorized into
SP and QAE groups according to the known difference
in residues 34–37. The QAE group was further divided
into two subgroups distinguished by 10 characteristic
residues (Fig. 4). A QAE1 isoform, nfeAFP8, exhibits
the highest sequence identity with HPLC12, which was
previously identified from Atlantic ocean pout (Ala24,
Met30, Val35, Glu58, and Thr64 in nfeAFP8 are
replaced by Ser24, Val30, Glu35, Asp58, and Pro64
in HPLC12). Therefore we categorized HPLC12 into
the QAE1 group. It has been shown that HPLC12 is
eluted last during HPLC [19], implying that the QAE
isoform is more hydrophobic than the SP isoform.
Although we did not complete the assignment of the
HPLC peaks (Fig. 2) to the present isoforms, we detec-
ted a difference in molecular mass between HPLC
peaks 1–10 ( 6600 Da) and peaks 11–13 ( 7000 Da)
(Fig. 2). Hence, it is highly likely that QAE isoforms
of nfeAFP are eluted later in HPLC (Fig. 2), which
agrees with the earlier elution of the SP isoforms (for
example, peak 2, nfeAFP6). In addition, the bipyrami-
dal ice crystals observed for the HPLC-pure samples
labeled 11–14 have a low c ⁄ a axial ratio (i.e. they are
thick in shape) compared with samples 2–10 of Fig. 2.
A low c ⁄ a axial ratio has been suggested to be a sign
of higher TH activity [8,18,23]. Hence, one can specu-
late that samples 11–14 correspond to QAE isoforms

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

it was assumed imperfect inhibition of the crystal
growth of f20

21g plane, a target ice surface of an
intact type I AFP. Such an unsatisfactory level of ice
growth inhibition leading to nondetection of TH
activity is also assumed for the present SP isoform.
Obviously, there are many candidates among the resi-
dues that could offer such character to the SP iso-
form. However, the residues located in the functional
site (i.e. ice-binding region) are thought to be the
prime candidates, as variants of type III AFP form a
mostly identical 3D structure irrespective of the
amino acid replacements [7–10,13,14]. As shown in
Fig. 4 (*), replacements between the three isoforms
are identified for the hydrophobic residues located at
positions 9, 13, 19, 20, 41, and 42 of the putative ice-
binding sequence. A clear difference between the SP
and QAE isoforms is identified for the 42nd residue;
Ser42 in QAE isoforms is replaced by Gly42 in SP
isoforms. However, a mutant of HPLC12 (QAE1),
S42G, has been reported to exhibit full TH activity
[11], therefore the character of the SP isoform cannot
be ascribed to Gly42 alone.
It has also been reported for HPLC12 that replace-
ment of a hydrophobic residue with a smaller aliphatic
residue results in loss of TH activity; 23–28% loss was
reported for the mutants L19A, V20A, and V41A.
Furthermore, 55% loss of activity has been reported
for the double mutant L10A⁄ I13A, and 75% loss for

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,

has the ability to form trimers and tetramers in the
absence of reductant (dithiothreitol), although mono-
mers are predominantly formed irrespective of the
presence of reductant (Fig. 5). The difference in TH
activity of nfeAFP13 in the presence and absence of
dithiothreitol (Fig. 6) implies that the TH activity of
the monomer (+ dithiothreitol) is enhanced approxi-
mately threefold by the presence of a small amount
of trimers and tetramers (– dithiothreitol). We previ-
ously reported that an artificial multimer of type III
AFP has considerably increased TH activity com-
pared with the monomer [27]. In addition, for type I
AFP isoforms from winter flounder consisting of 11
amino acid repeats, a longer isoform consisting of
four repeats showed almost twice the activity of the
ordinary three-repeat isoforms [28]. For b-helical AFP
isoforms from spruce budworm, the activity of
CfAFP-501 ( 12 kDa) with two additional loops is
twofold higher than that of CfAFP-337 ( 9 kDa)
[29]. Baardsnes et al. [30] recently reported that the
increase in activity of the type III AFP multimer
results from an increase in the size of the ice-binding
surface. These results suggest that, although the
monomer of nfeAFP13 does not exert substantial TH
activity by itself, it is enhanced with the help of a
small amount of multimers. A similar observation
was reported for antifreeze glycoprotein (AFGP) from
Antarctic cod, Pagothenia borchgrevinki, the molecular
mass of which is in the range 2.6–33 kDa: the low
TH activity of smaller sized AFGP ( 3 kDa) is

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

substantial TH activity for the less active isoform.
It is interesting to note that the less active SP iso-
form (nfeAFP6) is the major AFP species produced in
Notched-fin eelpout (Fig. 2). A similar observation
was reported for AFGP; the concentration of the smal-
ler sized AFGP in the serum of Antarctic cod is more
than eightfold that of the larger AFGP components
[32]. Ocean pout also produces many SP isoforms (11
species) compared with one QAE isoform [20]. Why is
the less active SP isoform of type III AFP dominant in
the fish serum? In winter, the level of expression of
AFPs is maximized to  20–30 mgÆmL
)1
in the serum
[33]. The SP isoform is less hydrophobic, as evidenced
by the present HPLC experiment (Fig. 2), and indeed
is more soluble than the QAE isoform (data not
shown). Hence, a plausible explanation for the sub-
stantial content of the SP isoform is as follows: (a) the
SP isoform is generated to reduce the hydrophobicity
and improve the solubility of type III AFP isoforms as
a whole at the expense of surface complementarity; (b)
accordingly, the SP isoform is the species that cannot
exert substantial TH activity by itself; (c) although the
antifreeze activity of the SP isoform is low, it can exert
TH activity with the help of the QAE isoform. The
production of a number of AFP isoforms may be a
strategy to retain AFPs in the serum at a sufficiently
high concentration to prevent the serum from freezing.
To summarize, we have succeeded in identifying 13

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;

TSKgel ODS-80Ts) with a linear gradient of 0–100% aceto-
nitrile in 0.1% trifluoroacetic acid. For observation of ice-
crystal morphology, the lyophilized powder of the AFP
collected was dissolved in 0.1 m NH
4
HCO
3
. For amino acid
sequence analysis, the AFP was digested with asparaginyl
endopeptidase (TaKaRa, Shiga, Japan) at 37 °C for 24 h in
50 mm sodium acetate (pH 5.0) containing 10 mm dithio-
threitol and 1 mm EDTA. Another digested fragment was
obtained by incubation with TPCK trypsin (Pierce, Rock-
ford, IL, USA) for 24 h at 37 °C in 0.1 m NH
4
HCO
3
(pH 7.9). These digested fragments were separated by
RP-HPLC. Sequence analysis of the fragments and native
AFP was carried out with an Applied Biosystem (Foster
City, CA, USA) 491 protein sequencer.
PCR amplification and cDNA sequencing
of nfeAFP
Fresh liver from a Notched-fin eelpout caught in the middle
of the winter of 2001 was cut into 0.5-cm-thick slices, and
soaked in RNA stabilization reagent (Qiagen, Hilden,
Germany) overnight at )4 °C. After being frozen at
)80 °C, 25 mg frozen liver was completely ground with
liquid nitrogen using a mortar and pestle, and homogenized
using a shredder spin column (Qiagen). Total RNA was

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

transformed into Esherichia coli strain BL21 (DE3), and
the transformants were grown at 28 °C in Luria–Bertani
medium supplemented with 100 lgÆmL
)1
ampicillin, until
cell growth reached the early stationary phase. To induce
expression of the recombinant nfeAFPs, 0.5 mm isopropyl
thio-b-d-galactoside was added to the medium, and the cul-
tures were grown at 28 °C overnight. Most of the nfeAFPs
were expressed as the soluble form, but nfeAFP13, which
contains two cysteines (Cys64 and Cys66), was only
expressed as the inclusion body. Hence, the nfeAFP13 sam-
ple was prepared by the methods described in [15] with the
following modifications: (a) the inclusion body was dis-
solved in 100 mm Tris ⁄ HCl (pH 8.5) containing 6 m guani-
dine hydrochloride and 10 mm 2-mercaptoethanol; (b)
nfeAFP13 was then refolded with 50 mm K
2
HPO
4
⁄ 100 mm
NaCl ⁄ 10 mm 2-mercaptoethanol (pH 10.7). The nfeAFP
isoforms obtained were dialyzed against 50 mm sodium
acetate buffer (pH 3.7) and then purified by cation-
exchange chromatography with a linear NaCl gradient
(0–0.5 m)in50mm sodium acetate buffer (pH 3.7) (50 mm
sodium citrate buffer, pH 2.9, was used for the purification
of nfeAFP11). The fractions containing the purified
nfeAFPs were stored and dialyzed against 0.1 m NH
4

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-

ics, Hartford, NY, USA) is usually used to determine the
ice-growth-initiation temperature, which is defined as the
freezing point (T
f
) of the AFP solution. However, this instru-
ment is no longer available commercially, so we used the fol-
lowing method with an alternative osmometer (model OM
802; Vogel): (a) 50 lL sample solution was placed in the
cooling pot ()7 °C) of the instrument; (b) the frosty probe
was put manually into the supercooled sample to initiate
water freezing; (c) the increase in solution temperature
caused by latent heat emission was monitored; (d) the plat-
eau temperature was determined to be the T
f
point of the
sample. All the nfeAFP samples were dissolved in 0.1 m
NH
4
HCO
3
(pH 7.9) for the T
f
measurement. The melting
temperature (T
m
) of the AFP solutions was carefully deter-
mined by monitoring the melting of an ice crystal on the
photomicroscope stage, which was manipulated by the
LK600 temperature controller. The measurement of T
f

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

fishes. Biochim Biophys Acta 915, 357–370.
5 Hew CL, Slaughter D, Joshi SB, Fletcher GL &
Ananthanarayanan VS (1984) Antifreeze polypeptides
from the Newfoundland ocean pout, Macrozoarces
americanus: presence of multiple and compositionally
diverse components. J Comp Physiol B Biochem Syst
Environ Physiol 155, 81–88.
6 Scott GK, Hayes PH, Fletcher GL & Davies PL (1988)
Wolffish antifreeze protein genes are primarily orga-
nized as tandem repeats that each contain two genes in
inverted orientation. Mol Cell Biol 8, 3670–3675.
7So
¨
nnichsen FD, Sykes BD, Chao H & Davies PL
(1993) The nonhelical structure of antifreeze protein
type III. Science 259, 1154–1157.
8 Jia Z, DeLuca CI, Chao H & Davies PL (1996) Struc-
tural basis for the binding of a globular antifreeze pro-
tein to ice. Nature 384, 285–288.
9So
¨
nnichsen FD, DeLuca CI, Davies PL & Sykes BD
(1996) Refined solution structure of type III antifreeze
protein: hydrophobic groups may be involved in the
energetics of the protein–ice interaction. Structure 4,
1325–1337.
10 Yang DS, Hon WC, Bubanko S, Xue Y, Seetharaman J,
Hew CL & Sicheri F (1998) Identification of the ice-
binding surface on a type III antifreeze protein with a
‘flatness function’ algorithm. Biophys J 74, 2142–2151.

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

275, 515–525.
17 Chen G & Jia Z (1999) Ice-binding surface of fish type
III antifreeze. Biophys J 77, 1602–1608.
18 Baardsnes J & Davies PL (2002) Contribution of hydro-
phobic residues to ice binding by fish type III antifreeze
protein. Biochim Biophys Acta 1601, 49–54.
19 Li XM Trinh KY Hew CL Buettner B Baenziger J &
Davies PL (1985) Structure of an antifreeze poly
peptide and its precursor from the ocean pout,
Macrozoarces americanus. J Biol Chem 260, 12904–
12909.
20 Hew CL Wang NC Joshi S Fletcher GL Scott GK
Hayes PH Buettner B & Davies PL (1988) Multiple
genes provide the basis for antifreeze protein diversity
and dosage in the ocean pout, Macrozoarces americanus.
J Biol Chem 263, 12049–12055.
21 Wang X DeVries AL & Cheng CH (1995) Antifreeze
peptide heterogeneity in an antarctic eel pout includes
an unusually large major variant comprised of two 7
kDa type III AFPs linked in tandem. Biochim Biophys
Acta 1247, 163–172.
22 Chao H Davies PL Sykes BD & So
¨
nnichsen FD (1993)
Use of proline mutants to help solve the NMR solution
structure of type III antifreeze protein. Protein Sci 2,
1411–1428.
23 DeLuca CI Chao H So
¨
nnichsen FD Sykes BD & Davies

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.

31 Burcham TS Knauf MJ Osuga DT Feeney RE & Yeh
Y (1984) Antifreeze glycoproteins: influence of polymer
length and ice crystal habit on activity. Biopolymers 23,
1379–1395.
32 Burcham TS Osuga DT Chino H & Feeney RE (1984)
Analysis of antifreeze glycoproteins in fish serum. Anal
Biochem 139, 197–204.
33 Fletcher GL Hew CL Li XM Haya K & Kao MH
(1985) Year-round presence of high levels of plasma
antifreeze peptides in a temperate fish, ocean pout
(Macrozoarces americanus). Can J Zool 63, 488–493.
34 Laemmli U.K. (1970) Cleavage of structural proteins
during the assembly of the head of bacteriophage T4.
Nature 227, 680–684.
Supplementary material
The following material is available from http://www.
blackwellpublishing.com/products/journals/suppmat/EJB/
EJB4490/EJB4490sm.htm
Figs S1–8. 1H-NMR spectra of the recombinant iso-
forms of nfeAFP2, 6, and 13 (+ dithiothreitol) (S1–
S3) and MALDI-TOF MS data for HPLC peaks 2, 6,
7, 8, and 13 of Fig. 2 (S4–S8).
Co-operative effect of type III AFP isoforms Y. Nishimiya et al.
492 FEBS Journal 272 (2005) 482–492 ª 2004 FEBS