Genes Genet. Syst. (2003)
, p. 419–425
Molecular phylogenetic analysis of ant subfamily
relationship inferred from rDNA sequences
, Hirotami T. Imai
and Masa-Toshi Yamamoto
Drosophila Genetic Resource Center, Kyoto Institute of Technology, Saga-Ippongi-cho,
Ukyo-ku, Kyoto 616-8354, Japan
National Institute of Genetics, Mishima, Shizuoka-ken 411-8540, Japan
(Received 6 October 2003, accepted 12 December 2003)
The relationships among ant subfamilies were studied by phylogenetic analysis
of rDNA sequences of 15 species from seven subfamilies.
designed on the basis of the rDNA sequence of the Australian bulldog ant,
, previously determined.Phylogenetic trees were constructed using
sequences of a fragment of 18S rDNA (1.8 kb), a fragment of 28S rDNA (0.7 kb
excluding variable regions) and a combination of the 18S and 28S rDNAs, by
neighbor-joining (NJ), maximum parsimony (MP) and maximum likelihood (ML).
rDNA sequences corresponding to the same fragments from three non-ant hyme-
nopteran species (a sawfly, a bee and a wasp) were employed as outgroups.
trees indicated that the ant subfamilies were clustered singly, and, among the
seven subfamilies examined, Ponerinae and six other subfamilies are in a sister-
groups relationship. The relationship among the six subfamilies, however, was
not clarified. The phylogenetic trees constructed in the present study are not in
contradiction to the tree from cladistic analysis of morphological data by Baroni
Urbani et al. (1992) and the tree from morphological and molecular data (Ward
and Brady, 2003), but are inconsistent with the traditional phylogeny.
present results thus raise a question as to the status of some traditionally
employed “key” morphological characters.
traditionally treated as a member of Ponerinae as
belonging to a new subfamily.
PCR primers were
The present results also call for a reex-
ribosomal RNA gene
The ants are known as one of the most flourishing euso-
cial insects. They are taxonomically classified in the
family Formicidae in the order Hymenoptera, and about
10,000 ant species have been described from the whole
world excepting the north and south poles.
includes 16 extant subfamilies, but their phylogenetic
relationships are yet to be clarified (Bolton, 1994).
the work by Brown (1954), various alternative phyloge-
netic trees have been proposed (e.g., Wilson, 1971; Taylor,
1978; Hölldobler and Wilson, 1990; Baroni Urbani et al.,
The subfamily relationship of ants has traditionally
been constructed by using a small number of morpholog-
ical key characters (e.g., post-petiole and abdominal seg-
ment IV).Myrmicinae and Ponerinae, were traditionally
treated as members of the same group with key charac-
ters of laterally fused abdominal segment IV and a
sting.Baroni Urbani et al. (1992) used a large number
of morphological characters and reported a cladogram
that is different from the traditional subfamily relation-
ship, placing Myrmicinae and Ponerinae in separate
groups.The cladistic analysis, as in the traditional phy-
logenetic tree, resulted in a basal division of extant For-
micidae into two groups, although different in their
constitutions. However, it is not always better to
increase the number of morphological characters for phy-
logenetic analysis.Since workers in social insects have
degenerated morphologies, one finds much convergence
that is difficult to distinguish from synapomorphy (Höll-
dobler and Wilson, 1990).
Molecular phylogenetic studies may provide better res-
olution of ant phylogeny than that based solely on mor-
phology, and could reveal a precise relationship among
subfamilies.rDNAs have been used extensively for phy-
logenetic analysis (Long and Dawid, 1980).
of rDNA contains both conserved and variable regions,
the comparison of sequences of rDNAs provides informa-
Since a unit
Edited by Etsuko Matsuura
* Corresponding author. E-mail: firstname.lastname@example.org
420H. OHNISHI et al.
tion on the relationships of both distantly and closely
related groups.Variable regions of 28S rDNA have often
been used in molecular phylogenetic studies on closely
related species (Pélandakis and Solignac, 1993; Gimeno
et al., 1997). They are not suitable, however, for studies
on the subfamily relationship of ants, because highly vari-
able sequences generally can not be aligned between
distantly related species.We previously obtained a com-
plete sequence of an entire single unit of rDNA of a bull-
(Ohnishi and Yamamoto, in
press), that hopefully allowed us to design PCR primers
to amplify conserved region of rDNA of any ant species
and other hymenopteran insects.
In the present paper, we analyzed sequences of 18S and
28S rDNA fragments amplified from 15 species from
seven subfamilies of ants, employing those of three spe-
cies (a wasp, a bee and a sawfly) as outgroups.
primers were designed so as to amplify the majority of
18S rDNA and a portion of 28S rDNA including variable
regions. Based on the sequence data, we constructed
phylogenetic trees for 15 ant species and three non-ant
hymenopteran species, which do not contradict the trees
obtained by Baroni Urbani et al. (1992) and by Ward and
Brady (2003) but are inconsistent with the traditional
one. The results thus raise a question as to the rele-
vance of the traditionally employed “key” morphological
MATERIALS AND METHODS
Ants, a wasp, a bee and a sawfly specimens and
genomic DNA extraction.
cies from seven subfamilies were examined: Myrmeciinae
), Nothomyrmeciinae (
), Pseudomyrmecinae (
), Formicinae (
.), Dolichoderinae (
), Myrmicinae (
). Three non-ant hymenopteran
species used as outgroups were:
Genomic DNAs were extracted from frozen (–80
100% ethanol-stored specimens as described (Sambrook
et al., 1989).
The following 15 ant spe-
) and Ponerinae
DNA amplification and sequencing.
employed for amplification of a 18S rDNA fragment was
5'-AGTAGTCATATGCTTGTCTC-3' (18S-U1) and 5'-AAT-
primer set was 5'- ACTAAGCGGAGGAAAAGAAACTA -3'
(28S-U1) and 5'- ACTCCTTGGTCCGTGTTTCA -3' (28S-
The primer set
The 28S rDNA
database, accession numbers AB052895 and AB121787)
(Fig. 1).PCR amplification was conducted with
polymerase (Takara), according to the manufacturer’s
protocol. Amplification conditions were 30 cycles of 1
min at 96
C, 1 min at 58
C, and 2 min at 74
the initial denaturation at 96
After electrophoresis, an amplified fragment was puri-
fied by SUPREC-01 (Takara).
amplified fragments were determined by the DyeDeoxy
Terminator Cycle Sequencing kit (Perkin Elmer) on an
ABI PRISM 310 Genetic Analyzer (PE Applied Biosys-
The sequences were deposited in DDBJ/EMBL/
GenBank (accession Nos. AB126778 ~ 126810, AB121786,
AB064266 and AB064267).
These primers were designed based on the sequence
(Ohnishi and Yamamoto, in press; see also
C for 2 min.
DNA sequences of PCR-
and outgroups were aligned with the Clustal W program
version 1.74 (Thompson et al., 1994) with the default set-
tings of a gap-opening penalty 10, and a gap-extension
penalty 0.1 in pairwise and 0.05 in multiple alignments.
The resulting alignment was modified by hand to correct
a few obvious alignment errors.
regions in D2 and D3 domains in 28S rDNA fragment
(Fig. 2) were eliminated from analyses.
phylogenetic analyses were performed by neighbor-join-
ing (NJ) (Saitou and Nei, 1987), maximum parsimony
(MP) (Fitch, 1971) and maximum likelihood (ML) (Felsen-
stein, 1981) methods. The distance was computed with
Kimura two-parameter distances method (Kimura, 1980).
The NJ method with a bootstrap test (Felsenstein, 1985)
was performed using the Clustal W program.
strap values were based on 1,000 replicates.
version4.0b10 (Swofford, 2002) was used for maximum
parsimony (MP) and maximum likelihood (ML) methods
of phylogenetic analyses.Strength of nodal support in
the MP and ML analyses were estimated using the con-
ventional nonparametric bootstrap (Felsenstein, 1985;
1000 replicates, heuristic search with random input
orders, and tree-bisection-reconnection (TBR) branch-
swapping), then generate bootstrap majority-rule consen-
sus trees. We used MODELTEST version 3.06 (Posada
and Crandell, 1998) to initially estimate maximum-likeli-
hood values under 56 different substitution models, which
were then subjected to hierarchical likelihood ratio tests
to determine the most appropriate model to be used in
ML analysis (Posada and Crandell, 2001).
of different data sets may require a measure, the incon-
gruence length differences (Farris et al., 1995), for the
levels of heterogeneity among the data sets.
arguments against this procedure ( Baker et al., 2001;
Yoder et al., 2001; Dowton and Austin, 2002), however,
we did not employ the test to evaluate combinability of
Gene sequences of the ants
421Molecular Phylogenetic Analysis of Ants
Structure of rDNA.
from specimens of 15 ant species belonging to seven sub-
families (Family Formicidae) and three species belonging
to different families of the Hymenoptera.
regions of the rDNA, a large portion of 18S and a small
portion of 28S, were PCR-amplified from each species
using the same primer pairs.
obtained in each and every species examined.
were determined for the 18S rDNA fragments (1.8 kb)
and the 28S rDNA fragments (0.8 kb).
The sequences of 1.8 kb fragments of 18S rDNA were
very similar to each other among all the species studied.
There were only limited numbers of base substitutions,
some of which were subfamily-specific.
The PCR-amplified 0.8 kb fragments of 28S rDNA con-
tained three (D1, D2 and D3) of the 12 variable regions
previously identified (Hassouna et al., 1984) (Fig. 1).
There was a stretch of conserved sequences in the middle
of the D2 region when the sequences of the 18 species
were aligned. Two highly variable subregions at D2
region and a subregion at D3 region (Fig. 2) were excluded
from further phylogenetic analyses.
Genomic DNAs were extracted
Single fragments were
set yielded a NJ tree with Kimura two-parameter dis-
tances, and a single MP tree (length 220, consistency
index 0.736, retention index 0.681) and a ML tree (The
selected model (TIMef + I + G) was a transitional model
(TIM, Rodríguez et al., 1990) with
erogeneity (G) and a proportion of invariant sites (I), and
which assumes equal base frequencies (ef)).
the 28S data set yielded a NJ tree with Kimura two-
Analysis of the 18S data
-distributed rate het-
parameter distances, and a single MP tree (length 456,
consistency index 0.688, retention index 0.596) and a ML
tree (The selected model (TrN + G) was Tamura-Nei
model (TrN, Tamura and Nei, 1993) with
rate heterogeneity).Kimura two-parameter distances of
18S and 28S were significantly correlated (
0.001). The 18S rDNA evolved about one fifth as fast as
the 28S rDNA.These two data sets were combined to
yield an overall data set of about 2.5 kb.
18S+28S data set yielded a NJ tree with Kimura two-
parameter distances, a single MP tree (length 702, con-
sistency index 0.698, retention index 0.604) and a ML tree
(The selected model (TrNef + I + G) was Tamura-Nei
-distributed rate heterogeneity and a propor-
tion of invariant sites, and which assumes equal base fre-
The strict consensus topology from NJ, MP and ML
trees based on the 18S, 28S and 18S+28S sequences data
sets, respectively, are presented in Fig. 3.
employed trees of MP and ML analyses were the boot-
strap majority-rule consensus tree.
the six subfamilies of ants (excluding Ponerinae) were in
single cluster with bootstrap values of 91%, 93% and
100% for NJ, 89%, 96% and 100% for MP, 94%, 96%and
92% for ML, respectively.All five species belonging to
Ponerinae were not included within the branches of the
cluster of the 6 subfamilies of ants.
gest that Ponerinae is a sister group of the other 6
subfamilies.Figure 3 also shows that
not clustered with other Ponerinae species.
tionship among the 6 subfamilies could not be clarified,
but Myrmeciinae and Nothomyrmeciinae appeared to be
closely related on the basis of 28S and 18S + 28S data
sets, although 18S data set did not support this relation-
Analysis of the
Trees showed that
These results sug-
Fig. 1. A schematic presentation of a single unit of rDNA in
boxes.Spacer regions without repeats are indicated by thin lines and that with repeats by striped area (top).
fragments are indicated by open boxes.The PCR fragment of 28S rDNA (0.8 kb) is shown in detail, including the variable D1, D2
and D3 regions. The highly-variable subregions eliminated from analyses are shown by a, b and c.
(see Materials and Methods).
. The rRNA coding regions are indicated by filled
The PCR amplified
Arrows indicate the PCR primers
422H. OHNISHI et al.
Fig. 2. Alignment of three variable regions in the 28S rDNA sequences.
shown in Fig. 1. Each species is indicated only by genus name.
adjusted. Dashes indicate gaps created to make better alignment with conserved sequences.
rounded by the frame. Numbers are nucleotide positions as appeared in the sequences reported (accession numbers are shown in
Materials and Methods).
correspond to the highly variable subregions
Alignment was done using Clustal W and then manually
Sequences of variable regions are sur-
423Molecular Phylogenetic Analysis of Ants
consensus trees obtained by different data sets (18S, 28S, 18S+28S) are shown.
are given for each tree and indicate the results obtained from neighbor-joining (top), maximum parsimony
(second) and maximum-likelihood (bottom) analyses.
outgroups. The species studied are indicated only by genus names.
Phylogenetic trees for 15 ant species (7 subfamilies) derived from rDNA fragment sequences.
The three bootstrap values
Three non-ant hymenopteran species are used as