T
he freshwater eels of the genus Anguilla Shaw
Evolution
of the Freshwater Eels
Jun Aoyama, Katsumi Tsukamoto
Ocean Research Institute, University of Tokyo,
Minamidai, Nakano-ku, Tokyo, 164 Japan
The freshwater anguillid eels have an unusual
life history and world-wide distribution. Ques-
tions about the phylogenetic relationships of this
group and how their long spawning migrations
and larval phase may contribute to their global
distribution have not been addressed. This paper
is first presentation of molecular phylogeny of
Anguilla species, and based on this phyloge-
netic tree we suggest new aspect of the evolu-
tion of this group. Namely, ancestral eels
originated during the Eocene or earlier, in the
western Pacific Ocean near present-day Indone-
sia. A group derived from this ancestor dis-
persed westward, probably by larval transport in
the global circum-equatorial current through the
northern edge of the Tethys Sea. This group
split into the ancestor of the European and
American eels, which entered into the Atlantic
Ocean, and a second group, which dispersed
southward and split into the east African species
and Australian species. Thus the world-wide
distribution of the eel family can be understood
from knowledge of continental drift, ocean cur-
rents, a specialized larva and evolutionary
forces favoring dispersal and speciation of seg-
regated gene pool.
Naturwissenschaften 84, 17–21 (1997) © Springer-Verlag 1997
(family Anguillidae) have an unusual life his-
tory and are widely distributed throughout the
world (Fig. 1). Freshwater eels are catadromous and
have a relatively long larval development as leptoce-
phalus, transparent leaflike larvae, which is the pecu-
liar larval form for Elopmorph fishes and highly
adapted to a planktonic life in the open ocean. The
continental distribution of freshwater eels appears to
be related to the subtropical circulation of the
oceans, with most species being located in the west
side of the Atlantic, Pacific, and Indian oceans [1–3].
Warm westward-flowing equatorial currents reach
and flow along the east coast of each continent,
whereas the west coasts are swept by cold currents
which originate in high latitudes. This association
with specific hydrographic conditions may explain
why eels inhabit east coasts of Eurasia, Australia,
Africa, and North America but not the west coasts of
North and South America, Australia, and Africa [3,
4]. Although the European eel, A. anguilla, is able to
inhabit Europe and the Mediterranean because of the
warm eastward flow of the Gulf Stream and North
Atlantic drift, it is not known why there are no fresh-
water eels along the east coast of South America de-
spite the existence of the warm Brazil Current. The
geographic distribution and unique life history traits
of anguillid eels gives rise to many questions, such
as when these eels evolved to form the 18 present-
day species, including three subspecies [5], and how
they achieved their worldwide distribution. In parti-
cular, how did the two Atlantic species become iso-
lated from the majority of other eel species in the
Indo-Pacific, and why are there no eels in the South
Atlantic? To answer these questions the phylogenetic
relationships of these eels needed to be clarified and
the evolutionary history of Anguilla considered in
light of global paleogeographical events.
17
Fig. 1. Distribution of freshwater eels throughout the world and the
sampling locations of the eight species studied (asterisks). The geo-
graphic ranges of the eight species are as follows. A. marmorata: tro-
pical regions of the Indian and Pacific Oceans; A. mossambica: the
eastern coast of Africa; A. japonica: the eastern coast of the Eurasian
Continent; A. anguilla: the western coast of the Eurasian continent; A.
rostrata: the eastern coast of North America; A. reinhardti: the eastern
coast of Australia; A. australis australis: south-eastern Australia; A.
celebesensis: the Sulawesi Island region. Specimens collected during
march 1979 and September 1994 were identified according to the
morphological key of Ege [2]. Liver tissues were taken from live eels,
and minced and preserved in the following buffer: 8 M Urea, 10 mM
Tris-HCl pH 8.5, 125 mM NaCl, 50 mM EDTA, 1% v/w SDS; or in
ethanol. DNA purification details are given elsewhere [14]. The mito-
chondrial cytochrome b gene was amplified by PCR using two oligo-
nucleotide primers, GLU-L [20] and H15149 [14], and sequenced by
Uniplex Cycle Sequence kit (Millipore MBBLUNP02)
Specimens of eight species were collected from three
oceans (Fig. 1), and their mitochondrial DNA se-
quences analyzed. Phylogenetic analysis with the
neighbor-joining and maximum parsimony method
suggests that A. mossambica, A. australis, A. rostra-
ta, and A. anguilla are members of the same clade.
Interestingly, A. reinhardti and A. australis are mem-
bers of different clades in spite of their sympatry in
eastern Australia. According to neighbor-joining tree,
most of the speciation of Anguilla apparently oc-
curred early, in comparison with that of the two
Atlantic species (Fig. 2). Furthermore, the bootstrap
consensus tree created by neighbor-joining method
was in complete agreement with that of the maxi-
mum parsimony method. Thus the consensus tree
(Fig. 3) was accepted as a phylogeny of eight Angu-
18
illa species because it represented both the parsimo-
nious and neighbor-joining trees.
The one previous study of the phylogeny of fresh-
water eels based on their morphological characteris-
tics recognized 19 species, including three subspe-
cies, and divided them into four groups [2]. If the
scheme of species relationships proposed by Ege [2]
is to be interpreted as a phylogeny, and the node spe-
cies are considered to be the most likely ancestral
species, the scheme can be explained as follows: The
group including the ancestral species A. celebesensis
originated first, followed by the second group con-
taining A. marmorata and A. reinhardti, then the
third group consisting of A. japonica, A. anguilla,
and A. rostrata, and finally the fourth group includ-
ing A. australis. Except for the position of A. japoni-
ca the phylogenetic tree obtained from genetic analy-
sis in this study agrees with that resulting from the
morphological analysis of Ege [2]. He concluded that
the two Atlantic species A. anguilla and A. rostrata
were closely related to A. japonica of the North Pa-
cific [2]. However, despite these morphological simi-
larities the present geographical distributions of these
eels (separated in different oceans) and the apparent
lack of close genetic similarity suggest that the two
Atlantic species do not have a particularly close phy-
logenetic relationship with A. japonica. The morpho-
logical similarity between Atlantic species and A. ja-
ponica may be parallel or convergent developments
as a response to their temperate habitats and long
oceanic spawning migrations that may require com-
mon morphological characteristics and functions.
Fig. 2. Phylogenetic tree for the eight Anguilla species constructed
from mitochondrial DNA data by the neighbor-joining method [17].
We selected A. celebesensis as an out-group based on the morphologi-
cal study by Ege [2], which found this species to have ancestral char-
acteristics among the genus. The pairwise distance was calculated ac-
cording to Kimura’s two-parameter method [18] with weighted modi-
fication of 5:1 for unequal rates of transitions and transversion by the
DNAdist program in the software package Phylip Ver. 3.5 [19]. The
sequence divergence between the two Atlantic species, A. anguilla
and A. rostrata, in 410 base pair of the cytochrome b gene deter-
mined, was a lower (0.024) than any other combination among the
species examined (range; 0.051–0.0854)
Freshwater eels might have originated in the tropics,
because two-thirds of the recognized 18 Anguilla
species inhabit the tropics while only five species
have their growth habitat in temperate regions. Seven
of the 18 species occur in the western Pacific around
Indonesia, and A. celebesensis,which inhabits mainly
Sulawesi Island (Celebes), Indonesia, has been
thought to be one of the possible ancestral species of
the Anguillidae [2, 5, 6]. All these considerations
strongly suggest the western Pacific in and around
Indonesia as the place of origin for freshwater eels.
Present knowledge of zoogeography, paleogeography,
and ocean circulation also suggests four potential
routes of dispersal into the Atlantic ocean. One pos-
sibility is that ancestral eels moved across the wide
paleo-Pacific Ocean from the western end to the east-
ern side. However, freshwater eels do not occur on
Easter Island or the Galapagos Islands or on the west
coast of the American continents [2]. Another possi-
bility is via Cape of Good Hope (southern tip of
Africa), but there are no freshwater eels along either
the east coast of South America or the west coast of
Africa (except in the far north) [2]. The low tempera-
ture and great distance would act as a barrier to pre-
vent dispersal of ancestral eels from tropical regions
through the Arctic Ocean. Moreover, the migratory
distance for one life cycle would be too long for
these eels which are very likely to have dispersed as
larvae from their breeding place in low latitudes,
considering present knowleges of larval duration and
migratory distance of the freshwater eels. The most
likely route is through the Tethys Sea that separated
Laurasia (North American and Eurasian Continents)
from Gondwana (Africa, South America, and India)
Fig. 3. Bootstrap consensus tree for eight Anguilla species in the pre-
sent study. We accepted the consensus tree of the corresponding max-
imum parsimony and neighbor-joining analysis by means of 1000
replications of bootstrap [21], as a phylogeny for eight Anguilla spe-
cies. The alternate analysis excluding the third coddons, which might
have more frequent multiple-substitution than the other two coddons,
showed about the same result as the consensus tree. The only differ-
ence was that the positions of A. australis and A. mossambica were
switched. This suggested the possibility that A. australis branched
earlier than A. mossambica, and that the two Atlantic species derived
from A. mossambica in Africa and not from A. australis
thereby providing a circumglobal oceanic connection
at low paleolatitudes. The Tethys Sea existed during
the Mesozoic until the beginning of the Tertiary peri-
od [7]. After the break up of Pangaea to the Oligo-
cene (24–36 Ma) continents were separated from
each other and there was a global circumequatorial
current which flowed from east to west encircling
the earth along the equator [7, 8]. Therefore it is
likely that the ancestral eels dispersed westwards in
this current, from the eastern end of the Tethys Sea
to the Atlantic. The ancestors of Atlantic eels may
thus have dispersed through the Tethys Sea to enter
the paleo-Atlantic Ocean from the western Pacific.
The Tethys Sea closed in the middle to late Oligo-
cene (25–30 Ma) [7, 9]. This suggests that the node
between the ancestor of the two Atlantic species and
A. australis in the neighbor-joining tree (ND1 in
Fig. 2) occurred over 25–30 Ma, namely before the
closure of the Tethys Sea. Thus, we conventionally
dated this node (ND1 in Fig. 2) as 30 Ma. On the
basis of this dating the split between A. celebesensis
and the ancestor of the other seven species (ND2 in
Fig. 2) was approximately 43.5 Ma i.e., in the Eo-
cene (57–36 Ma). Similarly, the split between A. an-
guilla and A. rostrata (ND3 in Fig. 2) is estimated at
approximately 10.2 Ma. This aging may involve an
underestimation because the speciation between the
Atlantic species and A. australis (ND1 in Fig. 2)
may have already proceeded by reproductive isola-
tion of adults in different oceans even if the Tethys
Sea were open, and there were some supply of lepto-
19

cephali from the Indian Ocean. This means that the
timing of the Tethys Sea may have closed much later
in the phylogenetic tree, suggesting much earlier spe-
ciation of anguillid eels, for instance, in the Cretac-
eous (145–65 Ma).
Fossils in the order Anguilliformes have been de-
scribed from the Upper Cretaceous [10]. Moreover,
Regan concluded that the Anguillidae is primitive
among the order [11]. This suggests that the Anguil-
liformes appeared at the latest in the Cretaceous, ap-
proximately 100 Ma, and that the Anguillidae
evolved at an early time in the history of the order.
The sequence divergence between A. anguilla and A.
rostrata was 0.024, which is approximately congru-
ent with the 0.03 value calculated by mitochondrial
DNA restriction-site polymorphism [12, 13]. Avise et
al. [13] estimated the separation time between A. an-
guilla and A. rostrata as 1.5 million years, using a
conventional mitochondrial DNA clock calibration of
2% sequence divergence per million years. Applying
this value of separation time to the phylogenetic tree
obtained in this study, the oldest branching (ND2 in
Fig. 2) was calculated to be only 6.4 million years
ago. This would indicate that the speciation events of
the Anguilla occurred mainly during late Miocene to
early Pliocene (approximately 3–7 Ma). However,
this period was characterized by the cooling asso-
ciated with the development of the Antarctic ice
sheet [7]. If this time estimate is adopted, the west-
ward invasion of eels into the Atlantic would be im-
possible since the “Tethys Corridor” would have al-
ready closed much earlier, and as an alternative the
Arctic region would have been too cold. Although
the Isthmus of Panama was still open in that period,
the current direction was wrong to transport larvae
into the Atlantic, even if some larvae were drifted
20
Fig. 4. Diagrammatic representa-
tion of the proposed scenario for
the dispersal and speciation of an-
guillid eels. The map shows a Eo-
cene global paleogeographic recon-
struction, approximately 50–60 Ma
[7]. The ancestral Anguilla species
originated in the western Pacific,
in present-day Indonesia. The
range of the ancestral eels was ex-
panded westward by transport of
their larvae in the Tethys Sea, by
paleocircumequatorial current,
eventually reaching the north
Atlantic
across the Pacific Ocean. Therefore it seems reason-
able to conclude that molecular evolution for eels
may be much slower than for other animals, since it
is now generally known that different animals have
different molecular clocks [14, 15].
Based on our estimate of the timing of divergence
for the Anguilla species and molecular phylogenetic
trees (Fig. 3) we propose the following evolutionary
scenario for the speciation and dispersal of the fresh-
water eels Anguilla.
In the Eocene (57–36 Ma) the ancestral eel origi-
nated in the western Pacific, around present-day In-
donesia and then split into two groups. Group A dis-
persed westward by the global circumequatorial cur-
rent along the northern margin of the Tethys Sea,
while group B dispersed less widely. Subsequently,
group A which may have spread through the Tethys
Sea and split into two groups: one being the ancestor
of the two Atlantic species, which was carried
further westward by the current into the Atlantic
Ocean, and the other being the ancestor of A. mos-
sambica and A. australis which dispersed southward
in the current that flows along the east coast of the
African Continent. The ancestor of A. anguilla and
A. rostrata that entered the Atlantic ocean may have
populated the coastal regions of North Atlantic as a
result of larval dispersal in the paleo-North Atlantic
gyre since the exit of the Tethys Sea opened directly
to the North Atlantic. However, this species may not
have been able to spread into the South Atlantic be-
cause the paleo-South Atlantic had not opened suffi-
ciently to have a warm subtropical gyre suitable for
larval development area. Even after a gyre was
formed, the circulation of the northern and southern
gyres was probably removed. That is may be why no
eels now occur in the South Atlantic.
In the western Pacific group B may have remained at
its place of origin for a while before part of the
group dispersed northward by the North Pacific gyre
resulting in the lineage of A. japonica. The others
that remained near Indonesia evolved into the
lineages of A. reinhardti and A. marmorata. The for-
mer dispersed south and the latter extended east and
west.
In the latest part of the Eocene and early Oligocene
(40–35 Ma) the Australian continent had moved
north after its separation from Antarctica, and was
populated by the lineage of A. reinhardti from near-
by Indonesia. Eels that inhabit the Australian conti-
nent thus now consist of two groups that arrived
there by way of different routes at different times.
This evolutionary scenario is consistent with the two
Australian species A. reinhardti and A. australis
being resolved as members of two separate clades in
the phylogenetic tree (Fig. 3).
The Japanese eel A. japonica is apparently not closely
related to the Atlantic eels, as has previously been sup-
posed, even though recent findings have shown them
to have very similar migratory and recruitment strate-
gies [16]. Instead, it is the southern species A. mossam-
bica and A. australis that are phylogenetically closer to
the North Atlantic eels, all of which appear to have di-
verged from a common Indonesian ancestor. Keys to
this divergence were the availability of the Tethys Cor-
ridor and larval dispersal westwards and southwards
on the paleocircumglobal current. If we accept the
Tethys Corridor hypothesis presented here, the rate
of sequence divergence for Anguilla is slower than
the previously accepted standard rate. Our findings
provide a new basis for further evaluation of the evo-
lutionary implications of the remarkable life history of
freshwater eels Anguilla.
The nucleotide sequence data reported in this paper have been sub-
mitted to the DDBJ, EMBL, and GenBank nucleotide sequence data-
base, with following accession numbers, D84299 (A. marmorata),
D84300 (A. mossambica), D84301 (A. reinhardti), D84302 (A. angu-
illa), D84303 (A. australis australis), D84304 (A. celebesensis),
D84305 (A. japonica), D84306 (A. rostrata). We thank Drs. P.H.J.
Castle and M. Miller for constructive comments and for improving
the manuscript at an early stage. We also thank Drs. M. Gross and
J. C. Avise for critical reading of the manuscript and comments.
1. Schmidt, J.: Rep. Smithsonian Inst. 1924, 279 (1925)
2. Ege, V.: Dana Rep. 16, 1 (1939)
3. Tesch, F. W.: The Eel. London: Chapman and Hall 1977
4. Tsukamoto, K., in: Migratory fishes in Japan, p.42 (A. Goto, K.
Tsukamoto, K. Maekawa, eds.). Tokyo: Tokai University Press
1994
5. Castle, P. H. J., Williamson G.R.: Copeia 2, 569 (1974)
6. Matsui, I.: Eel Biology; Biological Study. Tokyo: Koseisha-Ko-
seikaku 1972
7. Haq, B. U., in: Marine Geography and Oceanography of Arabian
Sea and Costal Pakistan, p.201 (B. U. Haq, J. D. Milliman,
eds.). New York: Van Nostrand Reinhold 1984
8. Obata, I.: Natural History of the Cretaceous. Tokyo: University
of Tokyo Press 1993
9. Andel, T. H. V.: New Views on an Old Planet – Continental
Drift and the History of Earth. Tokyo: Tsukiji shokan 1985
10. Robins, C. R., in: Fishes of the Western North Atlantic, p. 9
(E. B. Bohlke, ed.). New Haven: Sears Foundation for Marine
Research 1989
11. Regan, C. T.: Ann. Mag. Nat. Hist. 8 (10), 377 (1912)
12. Avise, J. C., Helfman, G. S., Saunders N. C., Hales L. S.: Proc.
Natl. Acad. Sci. USA 83, 4350 (1986)
13. Avise, J. C., Nelson, W. S., Arnold, J., Koehn, R. K., Williams,
G. C., Thorsteinsson, V.: Evolution 44 (5), 1254 (1990)
14. Kocher, T. D., Thomas, W. K. Meyer, A., Edwards, S.V., Paa-
bo,V., Villablanca, F. X., Wilson, A. C.: Proc. Natl. Acad. Sci.
USA 86, 6196 (1989)
15. Meyer, A., in: Biochemistry and Molecular Biology of Fishes,
Vol. 2, p. 1 (Hochachka, Mommsen eds.). New York: Elsevier
1993
16. Tsukamoto, K.: Nature 356, 789 (1992)
17. Saitou, N., Nei, M.: Mol. Biol. Evol. 4, 406 (1987)
18. Kimura, M.: J. Mol. Evol. 16, 111 (1980)
19. Felsenstein, J.: Phylogenetic Inference Package, Seattle: Univer-
sity of Washington 1993
20. Palumbi, S., Martin, A., Romano, S., McMillan, W. O., Stice, L.,
Grabowski, G.: The Simple Fool’s Guide to PCR Ver. 2.0. Hono-
lulu: University of Hawaii 1991
21. Felsenstein, J.: Evolution 39, 783 (1985)
21