Proc. Natl. Acad. Sci. USA
Vol. 94, pp. 3811­3816, April 1997
Evolution
Evolution of antifreeze glycoprotein gene from a trypsinogen
gene in Antarctic notothenioid fish
(repetitive sequences gene duplication environmental selection de novo amplification)
LIANGBIAO CHEN, ARTHUR L. DEVRIES, AND CHI-HING C. CHENG*
Department of Molecular and Integrative Physiology, University of Illinois, Urbana, IL 61801
Communicated by C. Ladd Prosser, University of Illinois, Urbana, IL, January 10, 1997 (received for review November 14, 1996)
ABSTRACT Freezing avoidance conferred by different
types of antifreeze proteins in various polar and subpolar
fishes represents a remarkable example of cold adaptation,
but how these unique proteins arose is unknown. We have
found that the antifreeze glycoproteins (AFGPs) of the predominant
Antarctic fish taxon, the notothenioids, evolved
from a pancreatic trypsinogen. We have determined the likely
evolutionary process by which this occurred through characterization
and analyses of notothenioid AFGP and trypsinogen
genes. The primordial AFGP gene apparently arose
through recruitment of the 5 and 3 ends of an ancestral
trypsinogen gene, which provided the secretory signal and the
3 untranslated region, respectively, plus de novo amplification
of a 9-nt Thr-Ala-Ala coding element from the trypsinogen
progenitor to create a new protein coding region for the
repetitive tripeptide backbone of the antifreeze protein. The
small sequence divergence (4­7%) between notothenioid
AFGP and trypsinogen genes indicates that the transformation
of the proteinase gene into the novel ice-binding protein
gene occurred quite recently, about 5­14 million years ago
(mya), which is highly consistent with the estimated times of
the freezing of the Antarctic Ocean at 10­14 mya, and of the
main phyletic divergence of the AFGP-bearing notothenioid
families at 7­15 mya. The notothenioid trypsinogen to AFGP
conversion is the first clear example of how an old protein gene
spawned a new gene for an entirely new protein with a new
function. It also represents a rare instance in which protein
evolution, organismal adaptation, and environmental conditions
can be linked directly.
Members of a single teleost suborder, Notothenioidei, overwhelmingly
dominate the fish fauna of the freezing ( 1.9 C)
coastal regions of the Antarctic Ocean in terms of number of
species ( 50%) (1­3) and biomass (90­95%) (4, 5). Their vast
ecological success is linked to ensured survival by the presence
of special blood-borne antifreeze glycoproteins (AFGPs; ref.
6). Antifreeze proteins prevent freezing of the body fluids of
teleosts, whose equilibrium freezing point ( 0.7 to 1 C) is
significantly higher than that of seawater ( 1.9 C), by adsorbing
to small ice crystals in the body and inhibiting their growth
(6­9). Besides AFGPs, there are three other structurally
different types of antifreeze proteins from various polar and
subpolar fishes (10, 11), suggesting that these unique proteins
evolved independently at least four times. How these unique
proteins evolved in these fishes has remained one of the most
important but unanswered questions in the area of antifreeze
research. The notothenioid AFGPs exist as a family of at least
eight isoforms of different sizes all composed of a simple
glycotripeptide repeat, (Thr-Ala Pro-Ala)n, with the disaccharide
galactose-N-acetylgalactosamine attached to each Thr (6,
10), and the dipeptide Ala-Ala at the N terminus. The smallest
(n 4, Mr 2,600) and the largest (n 55, Mr 34,000) are named
AFGP8 and AFGP1, respectively; many other intermediate
sizes besides the eight originally described have been subsequently
identified (12, 13). Collectively they are maintained at
very high circulatory levels of 30­35 mg ml (6, 10). Notothenioid
AFGPs are encoded by large gene families in which each
member gene encodes a large polyprotein precursor containing
multiple AFGP molecules (12, 14). The first AFGP gene
characterized is from the Antarctic notothenioid Notothenia
coriiceps, and the AFGP polyprotein precursor it encodes
contains 46 AFGP molecules (44 AFGP8 and 2 AFGP7
isoforms) linked in direct tandem by highly conserved threeresidue
spacers, Leu-Ile Asn-Phe, which serve as posttranslational
cleavage sites to yield the individual AFGP peptides
(14). This multigene, multiple AFGP copies per gene organization
provides an extremely large gene dosage that undoubtedly
contributes to the high circulatory abundance of the
protein. The tripeptide repeating structure of the AFGP
isoforms and the highly repetitive nature of the AFGP polyprotein
gene structure suggest that extensive duplications of an
ancestral 9-nt Thr-Ala-Ala coding element had occurred to
give rise to AFGPs. However, the origin of such a coding
element and the means by which the first AFGP gene was
formed are unknown. In searching the GenBank database, we
found that the 3 flanking sequence of the N. coriiceps AFGP
gene starting from the termination codon to about 100 nt
downstream is about 80% identical to the coding sequence of
the C terminus (50 residues) of the trypsinogen cDNA of the
Atlantic plaice, suggesting that notothenioid AFGP gene and
trypsinogen gene are somehow related. To understand the
precise nature of this relationship, AFGP and trypsinogen
genes from the giant Antarctic notothenioid, Dissostichus
mawsoni were isolated and analyzed. We report here our
studies that led to the elucidation of the trypsinogen origin of
notothenioid AFGPs and the evolutionary process by which an
ancestral trypsinogen gene was transformed to an AFGP gene.
MATERIALS AND METHODS
Characterization of AFGP Genes. Specimens of the giant
Antarctic notothenioid D. mawsoni were caught by winched
cable and hook at about 300­500 m in McMurdo Sound,
Antarctica. A D. mawsoni genomic library was constructed
with liver DNA using the phage vector FIXII (Stratagene).
About a half-million plaque-forming units of the primary
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked ``advertisemenť' in
accordance with 18 U.S.C. §1734 solely to indicate this fact.
Copyright 1997 by THE NATIONAL ACADEMY OF SCIENCES OF THE USA
0027-8424 97 943811-6$2.00 0
PNAS is available online at http: www.pnas.org.
Abbreviations: AFGP, antifreeze glycoprotein; myr, million years;
mya, million years ago; RT-PCR, reverse transcription­PCR; UTR,
untranslated region.
Data deposition: The sequences reported in this paper have been
deposited in the GenBank database (accession nos. U58835, U58867,
U58868, U58944, and U58945).
*To whom reprint requests should be addressed. e-mail:
cdevries@life.uiuc.edu.
3811
library were screened with a 32
P-labeled 3.3-kb PstI fragment
of the N. coriiceps-AFGP gene (14) by the in situ filter
hybridization method (15). Of 125 putative positive clones
from the screening, several were randomly selected for analyses.
The single AFGP-positive TaqI fragment from each
phage clone was subcloned into the plasmid pBluescriptII
KS( ) (Stratagene). Nested unidirectional deletions of the
subclone insert were generated by exonuclease III digestion
(15) and sequenced. Sequence upstream of the 5 TaqI site was
obtained by PCR-amplification of the recombinant phage
DNA with the T3 promoter primer in the phage vector and an
AFGP-specific primer, cloned into plasmid pCRII (TA Cloning
Kit, Invitrogen) and sequenced.
Amplification of AFGP Genes from Genomic DNA. One
microgram of genomic DNA from each of three Antarctic
notothenioids, D. mawsoni, N. coriiceps, and Pagothenia borchgrevinki
was amplified for AFGP gene sequences by PCR using
as primers the 5 untranslated region (UTR) sequence (cTrypAF5
), and a 30-nt 3 primer (cTryp-AF3 ) that anneals to a
site immediately ahead of the putative polyadenylylation signal
site in the 3 UTR of the AFGP cDNA (see below). A Southern
blot (15) of the PCR products was hybridized to an AFGP gene
probe that contains only the repetitive Thr-Ala-Ala coding
sequence.
Characterization of Trypsinogen and AFGP cDNAs. Both
trypsinogen and AFGP cDNAs were obtained by reverse
transcription­PCR (RT-PCR) of pancreatic RNA. Total RNA
from the pancreas-associated pyloric caecal mesentery of D.
mawsoni was isolated (Ultraspec RNA Isolation System, Biotecx
Laboratories, Houston), and poly(A) RNA was then
isolated from the total RNA (PolyATtract mRNA Isolation
System, Promega). About 100 ng of poly(A) RNA was
reverse-transcribed into first-strand cDNA (Superscript Preamplification
System, Life Technologies, Gaithersburg, MD)
to which a 5 anchor oligonucleotide [5 RACE (Rapid Amplification
of cDNA Ends) System, CLONTECH] was ligated.
The 5 portion of trypsinogen cDNA was obtained by 5 RACE
(i.e., PCR-amplification of the first-strand cDNA using the
anchor primer and a primer designed from the 3 flanking
sequence of AFGP gene that shares sequence identity with
plaice trypsinogen cDNA). The 5 RACE-PCR products were
cloned as a pool into pCRII, and one of the clones was isolated
and sequenced. The first 27 nt of the trypsinogen cDNA (5
UTR) was used to fashion a primer, cTryp-AF5 , and used with
a 32-mer oligo(dT) primer to generate the full-length trypsinogen
cDNA by RT-PCR amplification of pancreas poly(A)
RNA, cloned into pCRII and sequenced.
Two partial AFGP cDNAs covering the 5 and 3 end were
generated. The 5 cDNA was obtained by RT-PCR primed by
the 27-nt cTryp-AF5 (site found in AFGP gene; see Results)
and a 27-nt primer that anneals to AFGP spacer sequence. The
3 cDNA was obtained by RT-PCR using a 29-nt AFGP-spacer
5 primer and the 30-nt 3 primer, cTryp-AF3 . The products
were cloned into pCRII and sequenced.
Characterization of Trypsinogen Gene. The genomic DNA
of D. mawsoni was amplified by PCR with the cTryp-AF5 and
the cTry-AF3 primers, which will amplify both AFGP genes
and trypsinogen genes. Trypsinogen PCR products were identified
by probing a Southern blot of the PCR products with the
trypsinogen cDNA. A trypsinogen-positive product of 2.8 kbp
was found, recovered from the gel, cloned into pCRII, and
sequenced.
RESULTS
Fig. 1 shows the structures of the characterized AFGP and
trypsinogen genes and cDNAs from the giant Antarctic notothenioid
D. mawsoni, and their regions of sequence identity.
Notothenioid AFGP Gene Structure. The complete structure
of the AFGP gene from one of the characterized clones,
Dm1A (accession no. U58944), and the structure of the 5 and
3 AFGP cDNAs (accession nos. U58867, U58868) are depicted
in Figs. 1 A and B, respectively. The AFGP polyprotein
coding region (2129 nt) in the Dm1A AFGP gene contains 41
copies of AFGP coding sequences that encode four different
sized isoforms, linked in tandem by the conserved threeresidue
spacer, LIF or LNF, similar to that reported for the
Nc-AFGP gene from the related N. coriiceps (13). Immediately
5 to the AFGP coding region is a long stretch of repetitive gt
sequences (gt20), which was assigned as part of the presumptive
signal peptide in the Nc-AFGP gene (14). However, comparing
the AFGP gene and cDNA sequences (Fig. 1 A and B)
established that the (gt)n-region in fact comprises the 3 end of
the single intron (I1), which is 1879 nt in length in Dm1A and
which intervenes at the C terminus of the true signal peptide
encoded in exon 1 (Fig. 1A). Exon 1 encodes the 5 UTR (27
nt) and most of the signal peptide (40 nt, encodes 13 amino
acids). Exon 2 encodes the rest of the signal peptide (six
residues by assigning the putative cleavage site right before the
AFGP tripeptide repeats begin), the AFGP polyprotein, and
the 3 UTR. The 3 UTR sequence is 97% identical to that of
the Nc-AFGP gene.
Notothenioid Trypsinogen Gene Structure. The structures
of notothenioid trypsinogen cDNA (accession no. U58945)
and gene (accession no. U58835) are shown in Figs. 1 C and
D, respectively. The intron exon boundaries were established
by comparing the two sequences. The trypsinogen gene contains
six exons (E1-E6) and five introns (I1-I5). Exon 1 encodes
a 27-nt 5 UTR and most of the signal peptide (40 nt, 13 amino
acids). Exon 2 encodes the remaining two amino acids of the
trypsinogen signal peptide (cleavage site inferred from the
other trypsinogen sequences in the database). The rest of exon
2 through exon 6 encode a 234-residue trypsinogen and the 3
UTR.
Alignment of Notothenioid AFGP and Trypsinogen Genes.
Alignment of the four notothenioid AFGP and trypsinogen
gene and cDNA structures showed three regions of sequence
identity between the two genes (Figs. 1 A­D). First, exon 1 (5
UTR and signal peptide) of the AFGP gene and trypsinogen
gene are 94% identical in sequence. Thus, the AFGP gene
utilizes the 5 UTR and signal peptide coding sequence of the
trypsinogen gene. Second, the 3 end region of AFGP gene
starting from the penultimate codon through 3 UTR (255 nt)
is 96% identical to the last exon (E6) of trypsinogen gene.
Thus, the AFGP gene utilizes trypsinogen E6 as 3 UTR.
Third, the entire trypsinogen intron 1 (238 nt) is present as two
segments within the single AFGP intron with an overall
sequence identity of 93%. The first 12 nt of trypsinogen I1
inclusive of 5 splice site correspond to the same in AFGP I1,
and the remaining 226 nt inclusive of 3 splice site and a (gt)36
sequence right before the 3 splice site correspond to the
(gt)20-bearing 3 end of AFGP I1. Thus, AFGP gene intron 1
is the same as trypsinogen gene I1 plus a large (1.7-kbp)
insertion. Immediately 3 to the (gt)36 sequence in the trypsinogen
gene and straddling the splice junction of I1 and E2, a 9-nt
element, acagcggca (splice sequence in italics) is found that
translates into Thr-Ala-Ala (Fig. 1D and Fig. 2), the building
block of the repetitive tripeptide sequence of AFGP. The
nucleotide sequence of the three regions of sequence identity
between the two genes are shown in Fig. 2. The trypsinogen
AFGP hybrid gene structure is found to be present in the
extant AFGP genes of the notothenioids as shown by the
multiple AFGP-positive bands on a Southern blot of PCRamplification
products from the genomic DNA of three different
notothenioids using the common trypsinogen AFGP 5
UTR (cTryp-AF5 ) and 3 UTR (cTryp-AF3 ) primers (Fig. 3)
and hybridized to an AFGP-specific probe that contains only
(Thr-Ala-Ala)n coding sequence.
3812 Evolution: Chen et al. Proc. Natl. Acad. Sci. USA 94 (1997)
DISCUSSION
Given the common sequence elements between notothenioid
trypsinogen and AFGP genes, and the occurrence of a ThrAla-Ala
coding element in the trypsinogen gene, the primordial
AFGP gene could have risen from an ancestral trypsinogen
gene by a combined process of partial gene recruitment
and de novo amplification depicted in Fig. 4. E1 (5 UTR and
signal peptide), I1, and several nucleotides of trypsinogen E2
inclusive of the 9-nt Thr-Ala-Ala coding element were recruited
to form the 5 portion of the AFGP gene. A deletion
removed the rest of trypsinogen E2 through I5, linking E6
inclusive of the 3 UTR to the 9-nt Thr-Ala-Ala coding
element, and de novo amplification of the coding element gave
rise to an entirely new coding region that encodes the repetitive
tripeptide backbone of AFGP. The deletion, recruitment,
FIG. 2. Alignment of D. mawsoni AFGP and trypsinogen (Tryp) genomic sequences showing the three regions of high sequence identity between
the two genes. x, 5 UTR (lowercase) and signal peptide coding sequences (uppercase; translated amino acids also shown) are 94% identical; y,
intron I sequences (lowercase) are 93% identical (position of the extra 1684-nt AFGP intron sequence is indicated); and z, AFGP penultimate
codon plus 3 UTR sequence (lowercase) is 96% identical to trypsinogen exon 6 (uppercase) and 3 UTR (lowercase). Positions of gene-specific
sequences are boxed. The small number of nucleotide differences in the three regions are highlighted. The dots underscore the 9-nt Thr-Ala-Ala
coding element (acagcggca) in trypsinogen that might have been amplified to give rise to the repetitive tripeptide coding sequence of AFGP (in
parentheses).
FIG. 1. Structures of AFGP and trypsinogen genes and cDNAs from Antarctic notothenioid fish D. mawsoni. (A) AFGP gene in genomic clone
Dm1A. The AFGP polyprotein coding region (2129 nt) contains 41 copies of AFGP coding sequences (boxed numbers) tandemly linked by highly
conserved 9-nt spacers (bars filled with zigzagged lines). Nucleotide and translated amino acid sequence covering copies 2, 3, and (partly) 4 (extended
above gene structure) is given. Posttranslational cleavage (1 1) of spacers produces the mature AFGPs. Exon­intron boundaries were determined
after two partial AFGP cDNAs were characterized (B). 5 UTR and signal peptide (SP) sequence are encoded by E1, and AFGP polyprotein by
E2. The single, intervening intron I1 is 1879 nt. (C) Trypsinogen cDNA contains a 27-nt 5 UTR, 747 nt of pretrypsinogen coding sequence, and
a 105-nt 3 UTR. (D) Trypsinogen gene contains six exons (E1-E6) and five introns (I1-I5) with lengths in nt as indicated. Three regions of sequence
identities are found on alignment of all four structures. The two pairs of dashed lines delimit the 5 (67-nt) and 3 (225-nt) end regions of sequence
identities (gray regions) with the segments of the gene or encoded protein each represents, as labeled. The third region of identity is indicated by
striped bars; I1 of trypsinogen gene (D) is found in two segments in I1 of AFGP gene (A) and both introns contain repetitive (gt)n at its 3 end.
The 9-nt element in trypsinogen gene that translates into the repeat unit of AFGP peptide backbone, Thr-Ala-Ala, straddles I1 and E2 (splice
sequence in italics), as shown in D. Asterisks in C (trypsinogen cDNA) indicate primer sequences cTryp-AF5 and cTryp-AF3 used for RT-PCR
of AFGP cDNA and trypsinogen gene. Structures are not necessarily drawn to scale.
Evolution: Chen et al. Proc. Natl. Acad. Sci. USA 94 (1997) 3813
and amplification events did not need to occur in the order
given. Indeed, an AFGP trypsinogen hybrid protein coding
region formed by some amount of duplication of the 9-nt
Thr-Ala-Ala coding element before bulk deletion of trypsinogen
sequence might in fact be a more stable structure for the
evolving gene than large deletion first and amplification later.
In any case, these DNA rearrangement and amplification
events together led to a frameshift resulting in a termination
codon (tga) at the start of the recruited trypsinogen E6, and
converting it into the penultimate codon that encodes the last
amino acid, Gly (ggg; 1 g from splice sequence, 2 from E6) and
the 3 flanking region (or 3 UTR) of the new gene (Fig. 4).
The same penultimate codon (ggg) is found in all notothenioid
AFGP genes sequenced (refs. 13 and 14; Fig. 2).
The initial duplication of the 9-nt Thr-Ala-Ala coding
element could be a result of slippage replication (16) at the
repetitive (gt)n sequence immediately upstream during DNA
replications, causing the Thr-Ala-Ala coding element to be
copied more than once. Subsequent amplifications could occur
through slippage replication or unequal crossing over (17, 18)
of the new duplicants. Evidence for amplification of the
trypsinogen Thr-Ala-Ala coding element to give rise to the
repetitive AFGP polyprotein coding sequence is provided by
a striking correspondence between its nucleotide sequence,
aca(Thr)-gcg(Ala)-gca(Ala), and those of the tripeptide repeats
in extant AFGP genes. In the Dm1A AFGP gene, the
preponderance of aca in coding Thr (137 170, 81%), and gca
in coding the second Ala (162 171, 95%) strongly indicates
that they are duplicants or descendants of the ancestral
trypsinogen codons. The first Ala in the tripeptide is coded
mostly by gct (65 107, 61%) and the remainder by gcg (38 107,
36%). It is likely that gct, rather than gcg, was the codon for
the first Ala in the ancestral trypsinogen Thr-Ala-Ala coding
element. Nucleotide substitutions at this codon led to either no
change in the amino acid (gct to gcg, Ala), or a Pro for Ala
replacement (gct to cct) observed at this Ala position in the
antifreeze protein. The occurrence of the latter (gct to cct) is
supported by the almost exclusive use of cct in coding the Pro
(63 64, 96%) in the Thr-Pro-Ala repeats in the Dm1A gene,
as well as in other characterized notothenioid AFGP genes (13,
14).
The origin of the three-residue spacer sequence (Leu PheIle
Asn-Phe) is unclear at this point. It could be a preexisting
element that adjoined the Thr-Ala-Ala coding element in the
recruited ancestral trypsinogen gene, or was acquired through
recombinatory events early in the duplications of the Thr-AlaAla
coding element. The two parts then could have subsequently
amplified together to give rise to the extant polyprotein
gene structure (Fig. 4). If amplification of the 9-nt
Thr-Ala-Ala coding element occurred before deletion of
trypsinogen sequence, the acquisition of the tripeptide spacer
would provide the cleavage site to excise the first AFGP
molecule from the AFGP trypsinogen hybrid protein before
the stop codon (tga) was appropriately established for the
evolving AFGP gene. Like the spacer sequence, preexisting
sequence or recombinatory events could account for the
additional sequence ( 1.7 kbp) in the intron of the AFGP
gene. The trypsinogen progenitor was lost once it was converted
into the first AFGP gene, and the apparent absence of
spacer-like sequence or extra intron 1 sequence in extant
trypsinogen genes thus does not preclude the possibility that
they might be present in the trypsinogen progenitor before its
transformation into an AFGP gene. To summarize, through
recruitment and linking of the 5 and 3 portions of a trypsinogen
gene that supplied the secretory signal and the 3 flanking
region (or 3 UTR), respectively, and de novo expansion of a
9-nt Thr-Ala-Ala coding element in the middle of the new
structure to form the AFGP coding region, the first functional
AFGP gene was formed. The hybrid trypsinogen AFGP gene
structure, like that of Dm1A, is confirmed to be present in
current members of AFGP gene families across Antarctic
notothenioid species (Fig. 3).
Deciphering the evolutionary process of notothenioid
AFGP gene from a trypsinogen gene was made possible by the
high degree of nucleotide identities (93­96%) in both the
coding and noncoding sequences between the two genes, as
well as the close correspondence between the candidate 9-nt
ancestral Thr-Ala-Ala coding element in trypsinogen gene and
the repetitive AFGP tripeptide coding sequences, all of which
indicate that the trypsinogen to AFGP conversion was a recent
event. The small divergence (7%) between AFGP and
trypsinogen intron 1 sequences particularly supports the recent
evolution of the notothenioid AFGP as intron sequences are
under no constraint to remain conserved. There are no
definitive nuclear gene sequence divergence rates for teleosts
available in the literature to estimate the time of divergence
between notothenioid trypsinogen and AFGP genes. Using
teleost (salmon) mitochondrial DNA divergence rates, 0.5­
0.9% per million years (myr; ref. 19), the 4­7% sequence
divergence between the two notothenioid genes translates into
a divergence time of about 5­14 myr. Evolution of AFGPs at
5­14 million years ago (mya) is remarkably consistent with the
estimated mid-Miocene (10­14 mya) time frame during which
the Antarctic Ocean cooled to freezing based on paleotemperatures
inferred from oxygen isotopic ratios of sea bottom
planktonic deposits (20­22). It is also highly consistent with
the molecular phylogeny of notothenioids based on mitochonFIG.
3. Southern blot of products from PCR-amplification of
genomic DNA from three different notothenioids, using common 5
and 3 trypsinogen AFGP primers (sites indicated by asterisks in Fig.
1C) and hybridized to an AFGP-specific probe containing the repetitive
AFGP coding sequence only. The multiple AFGP-positive bands
indicate that members of notothenioid AFGP gene families all have
the hybrid AFGP trypsinogen gene structure. Dm, Dissostichus mawsoni;
Nc, Notothenia coriiceps; Pb, Pagothenia borchgrevinki.
3814 Evolution: Chen et al. Proc. Natl. Acad. Sci. USA 94 (1997)
drial 12S and 16S ribosomal RNA gene sequences, which
places the main phyletic divergence of the five AFGP-bearing
notothenioid families at about 7­15 mya (23). These time
frames are not entirely definitive, since paleoceanographic
temperatures derived from oxygen isotopic methods were
subject to different interpretations, and hypotheses of phyletic
relationships among notothenioids differ depending on
whether morphological or molecular data are subject to cladistic
analysis (3, 23). However, since the driving force for
AFGP evolution came from the onset of freezing conditions in
the Antarctic waters, and emergence of AFGPs undoubtedly
enabled the phyletic radiations of the benthic ancestral notothenioids
stock (3) into ice-laden pelagic and surface habitats
where they reside today, it is reasonable to expect that the
chronology of these three events would overlap. The remarkable
agreement in the estimated Miocene times of AFGP gene
evolution in this study, the freezing of the Antarctic Ocean
inferred from paleoceanographic studies (20­22), and the
main phyletic divergence of notothenioids based on molecular
phylogeny (23) argue strongly that this time frame could not
be due to mere coincidence, but is, in fact, reliable.
A pertinent question is why a pancreatic enzyme protein
gene was selected for conversion to the new ice-binding
protein. The intestinal fluids of Antarctic notothenioids are
known to contain high concentration of AFGPs (24) which
serve to inhibit the growth of ice crystals that inevitably enter
through food and seawater ingestion (9, 24). Conversion of an
existing pancreatic enzyme gene into AFGP gene and expressing
it in the pancreas is both positionally and temporally logical
as AFGPs thus could reach the digestive tract, simultaneously
with pancreatic enzymes, to prevent the intestinal fluid from
freezing while the enzymes perform digestive functions.
AFGPs continue to be expressed in the notothenioid pancreas
today as verified by the production of AFGP cDNAs by
RT-PCR amplification of pancreatic mRNA in this study, and
the expression appears to be at high levels as indicated by the
high intensity of hybridization to an AFGP-specific probe on
a Northern blot of pancreatic mRNA (data not shown). It is
possible that the notothenioid AFGP gene evolved and became
expressed in the pancreas first to protect the intestinal
fluid, as it was readily susceptible to freezing through daily
intake of ice-associated food or seawater, and later, the
expression extended to the liver, which became the synthetic
FIG. 4. Likely mechanism by which an ancestral trypsinogen gene was transformed into an AFGP gene. The 5 end (E1, I1, and small segment
of E2) and the 3 end (I5 3 splice site and E6) of trypsinogen gene were recruited and linked, and the remainder of the gene deleted (dashed lines
and boxes). The Thr-Ala-Ala coding element was duplicated, presumably via slippage at the repetitive (gt)n sequence during replication. The
recruited E1 provided the 5 UTR and signal peptide sequences for the new AFGP gene. The deletion, linking, and amplification events led to
a 1-nt frameshift resulting in a termination codon (tga) at the start of the recruited trypsinogen E6 and converting it into the 3 flanking sequence
of the AFGP gene. The spacer sequence (bars filled with zigzagged lines) and additional I1 sequence might be existing sequence in the trypsinogen
progenitor gene or acquired through recombinatory events. The Thr-Ala-Ala coding duplicants plus a spacer became amplified de novo to form
the new AFGP polyprotein coding region. The regions of identity are illustrated as in Fig. 1. Splice sites in trypsinogen gene are given in italics.
Evolution: Chen et al. Proc. Natl. Acad. Sci. USA 94 (1997) 3815
site to provide circulatory antifreeze (25) for freezing avoidance
in other extracellular fluid compartments.
The elucidation of how a notothenioid pancreatic trypsinogen
gene was transformed into an AFGP gene provides the
first clear and plausible evolutionary process by which one of
the four known types of antifreeze proteins arose. The antifreeze
peptides (type II AFPs) from the sea raven, herring, and
smelt share partial protein sequence identity with the carbohydrate
recognition domain of C-type lectins or similar domains
in lectin-like proteins from other organisms, suggesting
evolutionary relatedness (26). Lectins in type II AFP-bearing
fishes have not been characterized, and if they indeed gave rise
to the type II AFPs in these fishes, it would represent an
example of gene recruitment and expression of the same or a
very similar protein to perform a different function, much like
the recruitment of cellular enzyme genes such as lactate
dehydrogenase and others, and express them at high levels to
form the lens crystallins (27­30). Evolution of notothenioid
AFGP genes represents another evolutionary innovation--
recruitment of segments of an existing protein gene plus de
novo amplification of a short DNA sequence to spawn a novel
protein with a new function. Despite the apparently recent
notothenioid AFGP gene evolution, powerful environmental
selectional pressure--that is, the threat of freezing death once
the Antarctic water reached perennial freezing temperatures--may
have driven rapid intragene and whole gene duplications.
These processes could readily occur for simple
repetitive sequences like the AFGPs, leading to the large
AFGP polyprotein gene families we see today.
This work is supported in part by National Science Foundation
Grant OPP-93­17629 to A.L.D.
1. Gon, O. & Heemstra, P. C. (1990) Fishes of the Southern Ocean
(JLB Smith Institute of Ichthyology, Grahamstown, South Afri-
ca).
2. Hubold, G. (1991) in Biology of Antarctic Fish, eds. di Prisco, G.,
Maresca, M. & Tota, B. (Springer, Berlin), pp. 3­22.
3. Eastman, J. T. (1993) Antarctic Fish Biology: Evolution in a
Unique Environment (Academic, San Diego).
4. DeWitt, H. H. (1971) in Antarctic Map Folio Series, Folio 15, ed
Bushnell, V. C. (Am. Geogr. Soc., New York), pp. 1­10.
5. Ekau, W. (1990) Antarct. Sci. 2, 129­137.
6. DeVries, A. L. (1988) Comp. Biochem. Physiol. B 90, 611­621.
7. Raymond, J. A. & DeVries, A. L. (1977) Proc. Natl. Acad. Sci.
USA. 74, 2589­2593.
8. DeVries, A. L. (1984) Philos. Trans. R. Soc. London B 304,
575­588.
9. DeVries, A. L. & Cheng, C.-H. C. (1992) in Water and Life, eds.
Somero, G. N., Osmond, C. B. & Bolis, C. L. (Springer, Berlin),
pp. 303­315.
10. DeVries, A. L. (1982) Comp. Biochem. Physiol. A 73, 627­640.
11. Davies, P. L. & Hew, C. L. (1990) FASEB J. 4, 2460­2468.
12. Cheng, C.-H. C. (1996) in Gene Expression and Manipulation in
Aquatic Organisms, eds. Ennion, S. & Goldspink, G. (Cambridge
Univ. Press, Cambridge, UK), pp. 1­20.
13. Chen, L., DeVries, A. L. & Cheng, C.-H. C. (1997) Proc. Natl.
Acad. Sci. USA 94, 3817­3822.
14. Hsiao, K. C., Cheng, C.-H. C., Fernandes, I. E., Detrich, H. W. &
DeVries, A. L. (1990) Proc. Natl. Acad. Sci. USA 87, 9265­9269.
15. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) Molecular
Cloning: A Laboratory Manual (Cold Spring Harbor Lab. Press,
Plainview, NY).
16. Dover, G. A. & Tautz, D. (1986) Philos. Trans. R. Soc. London B
312, 275­289.
17. Maeda, N. & Smithies, O. (1986) Annu. Rev. Genet. 20, 81­108.
18. Lewin, B. (1990) Gene IV (Oxford Univ. Press, New York and
Cell Press, Cambridge, MA), pp. 497­517.
19. Martin, A. P. & Palumbi, S. R. (1993) Proc. Natl. Acad. Sci. USA
90, 4087­4091.
20. Kennett, J. P. (1977) J. Geophys. Res. 82, 3843­3860.
21. Kennett, J. P. (1982) Marine Geology (Prentice­Hall, Englewood,
NJ).
22. Clarke, A. (1990) in Antarctic Ecosystems: Ecological Change and
Conservation, eds. Kerry, K. R. & Hempel, G. (Springer, Berlin),
pp. 9­22.
23. Bargelloni, L., Ritchie, P. A., Patarnello, T., Battaglia, B., Lambert,
D. M. & Meyer, A. (1994) Mol. Biol. Evol. 11, 854­863.
24. O'Grady, S. M., Ellory, J. C. & DeVries, A. L. (1983) J. Exp. Biol.
104, 149­162.
25. Hudson, A. P., DeVries, A. L. & Haschemeyer, A. V. E. (1979)
Comp. Biochem. Physiol. B 62, 179­183.
26. Ewart, K. V. & Fletcher, G. L. (1993) Mol. Mar. Biol. Biotechnol.
2, 20­27.
27. Wistow, G. & Piatigorsky, J. (1987) Science 236, 1554­1556.
28. Wistow, G., Anderson, A. & Piatigorsky J. (1990) Proc. Natl.
Acad. Sci. USA 87, 6277­6280.
29. Piatigorsky, J. & Wistow, G. (1991) Science 252, 1078­1079.
30. Zinovieva, R. D., Tomarev, S. I. & Piatigorsky, J. (1993) J. Biol.
Chem. 268, 11449­11455.
3816 Evolution: Chen et al. Proc. Natl. Acad. Sci. USA 94 (1997)