Content uploaded by Oldřich Říčan
Author content
All content in this area was uploaded by Oldřich Říčan on Jul 24, 2018
Content may be subject to copyright.
Phylogenetic relationships of Middle American cichlids (Cichlidae, Heroini)
based on combined evidence from nuclear genes, mtDNA, and morphology
Oldr
ˇich R
ˇíc
ˇan
a,b,*
, Rafael Zardoya
c
, Ignacio Doadrio
c
a
Department of Zoology, Faculty of Science, University of South Bohemia, Branišovská 31, 37005, C
ˇeské Bude
ˇjovice, Czech Republic
b
Institute of Animal Physiology and Genetics of the Academy of Sciences of the Czech Republic, Rumburská 89, 277 21 Libe
ˇchov, Czech Republic
c
Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales, CSIC, Madrid, Spain
article info
Article history:
Received 2 June 2008
Revised 26 July 2008
Accepted 31 July 2008
Available online 7 August 2008
Keywords:
Cichlidae
Middle America
Central America
Biogeography
Molecular systematics
Morphological systematics
Taxonomy
Revision
abstract
Heroine cichlids are the second largest and very diverse tribe of Neotropical cichlids, and the only cichlid
group that inhabits Mesoamerica. The taxonomy of heroines is complex because monophyly of most gen-
era has never been demonstrated, and many species groups are without applicable generic names after
their removal from the catch-all genus Cichlasoma (sensu Regan, 1905). Hence, a robust phylogeny for the
group is largely wanting. A rather complete heroine phylogeny based on cytb sequence data is available
[Concheiro Pérez, G.A., R
ˇíc
ˇan O., Ortí G., Bermingham, E., Doadrio, I., Zardoya, R. 2007. Phylogeny and bio-
geography of 91 species of heroine cichlids (Teleostei: Cichlidae) based on sequences of the cytochrome b
gene. Mol. Phylogenet. Evol. 43, 91–110], and in the present study, we have added and analyzed indepen-
dent data sets (nuclear and morphological) to further confirm and strengthen the cytb-phylogenetic
hypothesis. We have analyzed a combined cytb-nuclear (RAG1 and two S7 introns) data set of 48 species
representing main heroine lineages to achieve further resolution of heroine higher taxonomic levels and a
combined cytb-morphological data set of 92 species to stabilize generic taxonomy. The recovered phylog-
enies supported the circumamazonian—CAM—Heroini (sensu Concheiro Peréz et al., 2007) as a monophy-
letic group, that could be divided into six main clades: (1) australoheroines (the southernmost heroine
genus Australoheros), (2) nandopsines (the Antillean genus Nandopsis), (3) caquetaines (including the
north western Amazonian genera Caquetaia and Heroina), (4) astatheroines (including Astatheros,Heroti-
lapia and Rocio), (5) amphilophines (including Amphilophus and related genera), and (6) herichthyines
(including Herichthyis and related genera). Nuclear and mitochondrial data partitions arrived at highly
congruent topologies. Suprageneric relationships were influenced mainly by the nuclear signal, as well
as the most basal phylogenetic position of Australoheros within CAM heroines. The new phylogeny of
the tribe Heroini provides robust framework to stabilize the taxonomy of the group and for future com-
parative studies on these morphologically and ecologically diverse freshwater fishes. Morphology was
mostly informative at the genus level and aid in determining the monophyly and composition of heroine
genera. Upon acceptance of all putative genera, as recovered in this study, the Heroini would be with 35
genera the most genus-rich clade of Neotropical cichlids.
Ó2008 Elsevier Inc. All rights reserved.
1. Introduction
Heroine cichlids are secondary freshwater fishes that constitute
an important component of the Neotropical fish fauna, especially
in Mesoamerica, where they make up some 25% of the freshwater
fish diversity (Bussing, 1985). They are one of the few groups of
freshwater fishes that are distributed from southern South America
to North America (where they cross the trans-Mexican volcanic
Belt), and also are the only cichlids in the Greater Antilles. Heroine
cichlids show a wide diversity of morphologies, as well as ecolog-
ical and behavioral adaptations (e.g. Bussing, 1985; Martin and
Bermingham, 1998; Miller et al., 2005). Moreover, they constitute
a model system to study biogeography of the Neotropical region
(Concheiro Pérez et al., 2007), and for instance, the Midas cichlid
complex (Amphilophus sp.) living in crater lakes in Nicaragua has
been proposed to be a model system to study sympatric speciation
(Barluenga et al., 2006).
The taxonomical and nomenclatural history of the tribe Heroini
is inextricably connected with that of the genus Cichlasoma Swain-
son, 1839. After the revision of Regan (1905) most heroine species
were assigned to the genus Cichlasoma. However, Kullander (1983)
recognized that Cichlasoma was an unnatural catch-all group, and
restricted it to 12 morphologically very similar species closely
1055-7903/$ - see front matter Ó2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2008.07.022
* Corresponding author. Address: Department of Zoology, Faculty of Science,
University of South Bohemia, Branišovská 31, 37005, C
ˇeské Bude
ˇjovice, Czech
Republic.
E-mail address: oldrichrican@yahoo.com (O. R
ˇíc
ˇan).
Molecular Phylogenetics and Evolution 49 (2008) 941–957
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
related to the genus Aequidens Eigenmann & Bray, 1894 (but a size-
able portion of Mesoamerican heroines was still left in Cichlasoma
in Kullander, 2003). At present, the genus Cichlasoma—somewhat
ironically—is the type of Cichlasomatini. Further work (summa-
rized in Kullander, 1998) elucidated the generic taxonomy of the
Cichlasomatini, but Heroini were almost completely left out, and
this chaotic situation has changed little since then. At present,
the difficulty with assigning generic names to many heroines hin-
ders evolutionary studies with these taxa, and necessarily requires
resolving with confidence phylogenetic relationships among spe-
cies and genera.
The monophyly of the tribe Heroini as well as its sister group
relationship with the tribe Cichlasomatini, forming together the
subfamily Cichlasomatinae are well supported based on both mor-
phological and molecular grounds (Kullander, 1998; Farias et al.,
1999, 2000, 2001). However, phylogenetic relationships and gener-
ic allocation of most species within the tribe Heroini are highly
contentious, and far from being understood (Miller, 1966, 1996;
Miller et al., 2005; Kullander, 1998, 2003; Concheiro Pérez et al.,
2007).
In spite of many morphological studies published on heroine
cichlids (Appendix 1), none to date has analyzed morphological
characters in combination with molecular data. On the other hand,
early studies utilizing molecular data were based mostly on mito-
chondrial (mt) cytochrome b(cytb) gene sequences, and were char-
acterized by relatively sparse taxon samplings (Roe et al., 1997;
Martin and Bermingham, 1998; Hulsey et al., 2004; R
ˇíc
ˇan and Kul-
lander, 2006; Chakrabarty, 2006). The most inclusive molecular
study (Concheiro Pérez et al., 2007) to date on heroine phyloge-
netic relationships included virtually all heroine lineages, but
was based solely on cytb.Concheiro Pérez et al. (2007) showed that
heroines could be divided into a paraphyletic stem lineage of Ama-
zonian genera (referred to as the Amazonian Heroini; AM) and a
monophyletic lineage (termed Circumamazonian Heroini; CAM)
including all Middle American, Antillean and trans-Andean heroine
cichlids as well as three cis-Andean South American genera,
namely Caquetaia,Heroina and Australoheros. Moreover, the major-
ity of Mesoamerican heroines could be placed into either one of
two large suprageneric clades, the amphilophines and the herich-
thyines (Concheiro Pérez et al., 2007). This study set a sound start-
ing point towards resolution of heroine phylogeny and
stabilization of generic taxonomy. However, the cytb-based phy-
logeny of heroines (Concheiro Pérez et al., 2007) is characterized
by rather short internodes, which require further confirmation
both by analyzing more mt sequence data to find additional syna-
pomorphies, as well as by examining independent data sets. In this
regard, it is well known that phylogenetic analyses of combined mt
and nuclear sequence data provide resolution not achieved by each
type of data separately (e.g. Brower et al., 1996; Wiley et al., 1998;
Rüber et al., 2004; Farias et al., 2000; López-Fernández et al.,
2005a). Thus far, however, only two studies (Chakrabarty, 2006;
Higham et al., 2007) have used nuclear sequence data to address
phylogenetic relationships of Heroini, although with reduced taxon
samplings. Phylogenetic studies based on morphology could be
very informative together with molecular sequence data in further
supporting monophyly of the different genera.
Given the urgent need of having a robust phylogeny of Meso-
american cichlids as framework for the wealth of comparative
studies ongoing on these taxa, the main goal of the present study
was to strengthen and test the previous hypothesis on the phyloge-
netic relationships of CAM heroine cichlids based on cytb data
(Concheiro Pérez et al., 2007) by analyzing two additional indepen-
dent (nuclear and morphological) data sets using a thorough gen-
eric sampling. Two nuclear markers were studied to increase
phylogenetic resolution both at deeper levels (RAG1 gene), as well
as at lower taxonomic levels (two introns of the S7 gene). In addi-
tion, phylogenetic analyses of a rather complete morphological
data set were performed with the particular goal of stabilizing her-
oine genera.
2. Materials and methods
2.1. Taxon sampling
In order to further resolve the phylogeny of heroines, a nuclear
(RAG1 and S7 introns) sequence data set, which included 48 spe-
cies representing all major lineages of CAM heroines as well as
most genera was compiled, and analyzed in combination with a
mt cytb gene sequences of the same 48 species, most taken from
R
ˇíc
ˇan and Kullander (2006) and Concheiro Pérez et al. (2007),
and four newly sequenced species (‘Heros’beani,Theraps bocourti,
Theraps irregularis and Theraps nourissati).
In order to test monophyly of the different genera and stabilize
heroine taxonomy at this level, a morphological data set was gath-
ered based on an extensive review of literature coupled with a thor-
ough study of museum specimens ( Appendix 2). A total of 97 CAM
heroine species representing all putative and established CAM hero-
ine genera, as well as all type species of established CAM heroine
genera were included in the phylogenetic analyses based on mor-
phology combined with cytb gene sequence data (R
ˇíc
ˇan and Kulland-
er, 2006; Concheiro Pérez et al., 2007; this paper).
In all phylogenetic analyses, geophagines, cichlasomatines and
Amazonian heroines were used as outgroup taxa.
2.2. Generic placement of heroine species in this study
The present nomenclatural treatment of the more than 100
Mesoamerican heroine cichlid species most of which were for-
merly referred to as Cichlasoma (Regan, 1905) is both chaotic and
frustrating (Kullander, 1983, 2003; Concheiro Pérez et al., 2007).
To deal with the complex nomenclature of CAM heroines, we used
the following approach. Monophyletic lineages including type spe-
cies of established genera were considered valid, and their species
composition was adjusted to keep the genera monophyletic. Estab-
lished genera, which we found as non-monophyletic were re-
stricted to include only the type species and the monophyletic
lineage to which it belongs. The remaining lineages excluded from
these previously non-monophyletic genera, and those monophy-
letic lineages without applicable generic names were propose to
be named as new genera (we refer to these putative new genera
as ‘Heros’ species groups) if their monophyly was found to be a sig-
nificantly better hypothesis than competing hypotheses found in
the literature.
2.3. Molecular methods
DNA was extracted from small pieces of muscle or gill (10–
25 mg) using the DNeasy
TM
Tissue Kit (QIAGEN). The complete cytb
gene was PCR amplified in four species as previously described
(Concheiro Pérez et al., 2007). The 3
0
half of the RAG1 gene
(1.5 kb) was PCR amplified with primers RAG1F1 5
0
-CTG AGC
TGC AGT CAG TAC CAT AAG ATG T-3
0
and RAG1R1 5
0
-CTG AGT
CCT TGT GAG CTT CCA TRA AYT T-3
0
(López et al., 2004). Two in-
trons of the S7 ribosomal protein-coding gene were PCR amplified
using primers S7RPEX1F 5
0
-TGGCCTCTTCCTTGGCCGTC-3
0
and
S7RPEX2R 5
0
-AACTCGTCTGGCTTTTCGCC-3
0
for S7 gene intron 1
and primers S7RPEX2F 5
0
-AGCGCCAAAATAGTGAAGCC-3
0
and
S7RPEX3R 5
0
-GCCTTCAGGTCAGAGTTCAT-3
0
for S7 gene intron 2
(Chow and Hazama, 1998).
PCR amplification of RAG1 gene was carried out with an initial
denaturing step at 95 °C for 1 min, followed by 35–40 cycles of
amplification (denaturing at 95 °C for 45 s, annealing at 54 °C for
942 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
60 s and extending at 72 °C for 90 s), with a final extending step at
72 °C for 10 min. PCR amplification of the S7 gene introns was car-
ried out with an initial denaturing step at 95 °C for 1 min, followed
by 30–35 cycles of amplification (denaturing at 95 °C for 30 s,
annealing at 60 °C for 60 s, and extending at 72 °C for 120 s), with
a final extending cycle at 72 °C for 10 min. PCR products were puri-
fied either by ethanol precipitation or using Microcon PCR Filter
Units (Millipore), and directly sequenced with the corresponding
PCR primers using the BigDye
TM
Terminator Cycle Sequencing Kit
v.3.1 (PE Applied Biosystems), and following manufacturer’s
instructions. Sequencing reaction products were cleaned either
by ethanol precipitation or with DyeEx 2.0 Spin Kit (QIAGEN),
and run on ABI Prism 310 Genetic Analyzer (PE Applied Biosys-
tems). Chromatograms were assembled and checked by eye for po-
tential mistakes using SeqMan II of the DNAStar software package
(http://www.dnastar.com). Edited sequences were aligned using
ClustalX software (Thompson et al., 1997), and either the default
settings (cytb and RAG1) or three different parameter settings
(for the S7 introns; gap opening/extension: 10/5, 7/5, 10/10).
Aligned sequences were manually adjusted in BioEdit (Biological
sequence alignment editor v 5.0.9, http://www.mbio.ncsu.edu/Bio-
Edit/bioedit.html), and sites that shifted relative position in the
three alignments (S7), and could not be reconciled by eye were ex-
cluded from phylogenetic analyses (Gatesy et al., 1993).
All sequence data newly determined in this study were depos-
ited in GenBank under the Accession Nos. EF436463–EF436466
(cytb), EF433005–EF433048 (S7i1), EF433049–EF433086 (S7i2),
and EF362572–EF362613 (RAG1).
2.4. Morphological methods
Most of the characters included in Kullander’s (1998) morpho-
logical matrix were found to be uninformative to resolve phyloge-
netic relationships among closely related CAM Heroini genera.
Therefore, a thorough morphological study of a large number of
museum specimens covering all recognized or putative heroine
lineages was carried out to seek for additional characters. Refer-
ences to major literature sources from which morphological char-
acters were extracted are listed in Appendix 1. A total of 97
ingroup species (1218 museum specimens) and 31 outgroup (ama-
zonian heroines, Cichlasomatines, and Geophagines) species (135
specimens) were examined ( Appendix 2). Character descriptions
are provided in Appendix 3, and the morphological character
Table 1
Comparison of phylogenetic performance of the genes used in this study (as assessed in separate analyses with the reduced taxon-sampling)
Sites PI sites In % CI RI PI sites In % CI RI
Plus indels
RAG1 1486 66 4.440 0.5984 0.8061 66 — — —
S7i1 632 84 13.29 0.5988 0.7473 135 21.36 0.6221 0.6831
S7i2 871 191 21.92 0.5667 0.7567 296 33.98 0.5898 0.6604
Cytb 1143 443 38.75 0.2317 0.4386 443 — — —
Both introns in the S7 gene include significant amount of information in form of indels (in bold). PI, parsimony informative sites; CI, consistency index; RI, retention index.
Fig. 1. Phylogenetic analyses of nuclear genes. (A) Strict consensus MP tree based on the nuclear data set excluding indels (PI = 341; N= 15,000+; L= 1461; CI = 0.5652;
RI = 0.6907). (B) ML phylogram (lnL = 13,026.51) inferred under the GTR+I+G model. (C) BI phylogram (ln L = 12,769.69) inferred under the GTR+I+G model. Numbers show
bootstrap support (N= 1000). Asterisks denote chimerical sequences resulting from combining closely related species to increase taxon overlap between datasets. Major
clades are indicated to facilitate tree interpretation and comparisons.
Table 2
Parameters of ML analyses estimated with Modeltest
Cytb RAG1 S7
Model selected by Modeltest (AIC criteria) TrN+I+G GTR+I+G HKY+G
Nucleotide proportions A = 0.2822 A = 0.2506 A = 0.2640
C = 0.3896 C = 0.2391 C = 0.1829
G = 0.0783 G = 0.2761 G = 0.2397
T = 0.2498 T = 0.2343 T = 0.3133
Assumed proportion of invariable sites (I) 0.4908 0.4923 0
Alpha (G) 0.8995 1.022 0.8441
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 943
matrix is given in Appendix 4. A total of 81 informative characters
were analyzed. Of these, 23 (characters 1–23) were meristic char-
acters, four (characters 24–27) described external, non-meristic
characters, 19 (characters 28–46) referred to descriptive internal
and external characters (mostly from teeth and jaw morphology)
that could not be coded in a quantitative manner, and 35 (charac-
ters 47–81) were color pattern characters, which were studied in
an ontogenetic perspective (R
ˇíc
ˇan et al., 2005).
Coding of morphological characters follows recommendations
of Campbell and Frost (1993) and Wiens (1995, 1999). Qualita-
tive characters (characters 24–81) were coded using the majority
approach. Some characters, which showed more discernible
states, were coded using the scaled coding (Campbell and Frost,
1993) under the assumption that traits pass through a polymor-
phic stage between absence and fixed presence. The scaled
method is advantageous in that it allows polymorphisms to act
as synapomorphies.
Quantitative characters (characters 1–23) were coded using a
modified gap weighting method (GW) of Thiele (1993). Thiele’s
implementation of gap weighting involves finding, for a given
character, the mean value of the trait in each species in the anal-
ysis, the range of mean species values among taxa (i.e., the spe-
cies with the greatest mean value and the species with the
lowest), and then dividing this range into smaller ranges or seg-
ments equal to the maximum number of character states al-
lowed by the phylogenetic software program (i.e., 32 for
PAUP
*
). States are then assigned to species based on these
ranges. An important advantage of the gap-weighting method
is that it incorporates information on the distance between
states, weighting the changes according to the difference be-
tween mean species values (hence the name).
We used the between-state scaling (Wiens, 2001) to weight
quantitative multistate characters (i.e., those coded with the mod-
ified gap-weighting method (GW); see above) against each other.
This weighting scheme assigns transformations between species
with fixed, adjacent values of meristic variables (e.g., 13–14 verte-
brae or 5–6 anal spines) the same weight in all GW coded charac-
ters. The more fixed steps a multistate quantitative character
expresses the more information in contains, but all multistate
characters are a-priory weighted 1:1 in this method. To weight
quantitative multistate characters against qualitative characters
we used the between-character scaling (Wiens, 2001). All charac-
ters are thus in effect weighted 1:1 to each other irrespective of
their method of coding. Changes in binary variables (0–1) thus
have the same weight as the whole transformation series of a mul-
tistate character.
2.5. Phylogenetic inference
Phylogenetic analyses of the nuclear gene sequence data sepa-
rately, and combined with cytb gene sequence data were per-
formed using maximum parsimony (MP), maximum likelihood
(ML), and Bayesian Inference (BI) as implemented in PAUP
*
(Swof-
ford, 2000), PHYML (Guindon and Gascuel, 2003), and MrBayes
Fig. 2. MP Analyses of all genes combined. (A) Strict consensus MP tree based on the combined molecular data set excluding indels (PI = 784; N= 60; L= 4635; CI = 0.2956;
RI = 0.4770). (B) Strict consensus MP tree based on the combined molecular data set including indels (PI = 915; N=3;L= 5341; CI = 0.3288; RI = 0.4866). Numbers show
bootstrap support (N= 1000). Asterisks denote chimerical sequences resulting from combining closely related species to increase taxon overlap between datasets. Major
clades are indicated to facilitate tree interpretation and comparisons.
944 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
3.0 (Huelsenbeck and Ronquist, 2001), respectively. MP analyses
were conducted with heuristic searches (TBR branch swapping,
MULTREES option in effect, and 10 random stepwise additions
of taxa). We first extensively sampled the tree space using 1000
random sequence additions and keeping 10 trees per search
(commands in PAUP
*
: hsearch addseq = random nchuck = 10
chuckscore = 1 nreps = 1000). In the next step, we run a search
on the saved trees to find all the shortest trees (commands in
PAUP
*
: hsearch start = current nchuck = 0 chuckscore = 0). For
the ML and BI analyses, the best-fit model for the different genes
was selected with Modeltest 3.06 (Posada and Crandall, 1998)
using the Akaike information criterion (AIC). Bayesian analyses
were performed using two independent runs of four Metropolis-
coupled chains (MCMC) of 5 million generations each, to estimate
the posterior probability distribution. The combined sequence
matrices were partitioned per gene fragment, and independent
model parameters were estimated for each partition. Topologies
were sampled every 100 generations and majority-rule consensus
trees were estimated after discarding the first 10% generations for
the nuclear, and combined gene data sets, respectively. Robust-
ness of clades was assessed using bootstrapping (1000 pseudore-
plicates) for the MP and ML analyses, and Bayesian posterior
probabilities for the BI analyses.
Phylogenetic analyses of the morphological data set separately
and combined with the cytb gene nucleotide sequences were per-
formed using MP as implemented in PAUP
*
(Swofford, 2000; see
search strategy above). Robustness of clades was assessed using
bootstrapping (1000 pseudo-replications in PAUP
*
) and the same
approach as in the MP searches, with five random sequence addi-
tions per bootstrap replication, and saving 10 trees from each ran-
dom sequence addition. Data exploration further included Bremer
support and PBS (partitioned Bremer support) to assess congru-
ence or conflict between the data partitions at each node of inter-
est. Bremer analyses were run with the same parameters as MP
searches in PAUP
*
.
2.6. Tests of alternative topologies
In order to statistically test the significance of our results against
competing taxonomies (as found in the literature) and alternative
phylogenies, we performed statistical tests of alternative topologies.
The tests used maximum parsimony (MP) and included the com-
pare-2 T-PTP test, the Templeton test and the Shimodaira-Hasegawa
(SH) as implemented in PAUP
*
and using default settings. The com-
pare-2 T-PTP test has been run with 500 random additions.
3. Results
3.1. Performance of individual gene datasets
Comparison of phylogenetic performance of individual genes is
found in Table 1. Best-fit models and corresponding model param-
eters as estimated using Modeltest (Posada and Crandall, 1998) are
provided in Table 2. The nuclear genes showed much higher values
Fig. 3. ML and BI analyses of all genes excluding indels. (A) ML phylogram (ln L = 28,170.07) using the GTR+I+G model. (B) BI phylogram (ln L =27,594.12) using the
GTR+I+G model. Numbers show bootstrap support (N= 1000; ML analysis) and posterior probabilities for BI analysis. Asterisks denote chimerical sequences resulting from
combining closely related species to increase taxon overlap between datasets. Major clades are indicated to facilitate tree interpretation and comparisons.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 945
of consistency (CI) and retention (RI) indices, whereas cytb gene
exhibited more parsimony informative (PI) sites. The proportion
of PI sites of the two S7 introns increased when including indels,
and approached that of the cytb gene. Importantly, including PI in-
dels of the S7 introns into the phylogenetic analyses augmented
the CI index (Table 1) and phylogenetic resolution (Fig. 2). More-
over, several indels were found diagnostic even for suprageneric
clades (Fig. 6). The RAG 1 gene was very conservative with by far
the lowest amount of PI sites.
3.2. Phylogenetic relationships based on nuclear and combined
nuclear—mt gene sequence data
In order to confirm and strengthen high-level phylogenetic rela-
tionships among heroine main lineages as recovered based on cytb
gene sequence data (Concheiro Pérez et al., 2007) two additional nu-
clear markers (RAG1 + S7 introns) have been added and analyzed
using MP, ML and BI methods (Fig. 1). The results of these analyses
are in very good agreement with thecytb phylogeny (Concheiro Pérez
Fig. 4. Maximum parsimony analysis of the 81 informative morphological characters. The analysis resulted in two MP trees (L= 5248; CI = 0.21; RI = 0.67), which differed
only regarding the sister taxon of Herichthys tamasopoensis (either H. carpintis or H. cyanoguttatus). The tree shown here is one of the two MP trees (branch lengths shown),
with H. carpintis as the sister group of H. tamasopoensis in agreement with cytb-based phylogeny (Concheiro Pérez et al., 2007). Numbers above nodes show bootstrap support
(N= 1000). Numbers below nodes show Bremer support. Topology shown with interrupted lines is in conflict with molecular trees (see Figs. 1–3, 5 and 6 and Concheiro Pérez
et al., 2007). The phylogeny does not recover the six major clades found in molecular trees (see above). The major clades are however indicated to facilitate tree interpretation
and comparisons, since also combined molecular—morphological trees recover the six major clades.
946 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
et al., 2007). The only two important differences are in the positions
of Herotilapia (placed with Rocio and Astatheros), and of Caquetaia
umbrifera (placed with rest of Caquetaia and Heroina) see below).
The combined nuclear–cytb data set was analyzed under MP
either excluding indels (Fig. 2A) or considering them as a fifth char-
acter (Fig. 2B). In addition, the molecular data set was analyzed un-
der ML (Fig. 3A) and BI (Fig. 3B) without considering indels (not
implemented in these methods).
The reconstructed phylogenies based on the combined nucle-
ar-cytb gene sequence data using the different methods of phy-
logenetic inference show influence by the nuclear signal (see
differences to cytb phylogeny above), and divide CAM heroine
cichlids into four suprageneric and two generic main clades that
received relatively strong support (Figs. 2 and 3). Phylogenetic
relationships among the six main clades lacked support, with
the exception of the sister group relationship of australoheroines
(Australoheros) to the remaining CAM heroines, which received
moderate support (Figs. 2 and 3). Interestingly, MP and ML/BI
phylogenetic analyses differed in the relative phylogenetic
position of nandopsines (Nandopsis), which in the former was
Fig. 5. Maximum parsimony analysis of the mor-cytb data set. One of the three MP trees (L= 34415.75) is shown. Numbers show node numbers. See Table 3 for bootstrap
support and partitioned Bremer support. Major clades are indicated to facilitate tree interpretation and comparisons. Taxa highlighted in grey represent putative new genera.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 947
recovered as the next lineage branching out after Australoheros,
whereas in the latter appeared in a more derived position closer
to amphilophines and caquetaines (Caquetaia +Heroina)(Figs. 2
and 3), as in the phylogenies based exclusively on nuclear genes
(Fig. 1). Similarly, caquetaines were recovered as sister group of
astatheroines and herichtyines (Fig. 2B) and of amphilophines
plus nandopsines (Fig. 3) in the ML/BI phylogenetic analyses,
respectively.
3.3. Phylogenetic relationships based on morphological characters
The maximum parsimony (MP) analysis of the 81 informative
morphological characters recovered two MP trees (L= 5248;
CI = 0.21; RI = 0.67; Fig. 4), which differed only regarding the sis-
ter taxon of Herichthys tamasopoensis (either H. carpintis or H.
cyanoguttatus). The tree in Fig. 4 shows H. carpintis as sister group
of H. tamasopoensis in agreement with cytb-based phylogeny
(Concheiro Pérez et al., 2007). The recovered tree lacked support
and congruence with molecular phylogenies at deeper nodes (cf.
Figs. 1–3 and Concheiro Pérez et al., 2007), and the main split
of CAM heroines into amphilophines and herichthyines was not
supported (Fig. 4). Instead, morphology showed relative resolving
power at the genus level, and the main recovered clades sup-
ported the following non-monotypic genera: Cryptoheros (but
including also C. nigrofasciatus, and not including C. panamensis),
Australoheros,Amphilophus s. str., Parachromis,Herichthys, and
Fig. 6. Maximum parsimony analysis based on all genes (with indels) and morphology (PI = 977; N=1; L= 30,696.25; CI = 0.3146; RI = 0.4793). Branch lengths are
proportional to the number of steps. Values above branches are node numbers corresponding to Table 5 (see for bootstrap support and partitioned bremer support). Values in
boxes bellow branches show number and length of unique apomorphic indels in S7 introns (except most single indels). Thick branches identify nodes where all or most data
partitions are in agreement (see Table 5). Asterisks denote chimerical sequences resulting from combining closely related species to increase taxon overlap between datasets.
Major clades are indicated to facilitate tree interpretation and comparisons. Taxa highlighted in grey represent putative new genera.
948 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
Thorichthys. In addition, all monotypic genera with the exception
of Petenia were found as isolated lineages, not disrupting mono-
phyly of other established genera. The following species were
recovered as relatively divergent lineages without immediate
close relatives among named genera: ‘Heros’calobrensis,‘Heros’
urophthalmus,‘Heros’salvini, and to a lesser extent ‘Heros’siebol-
dii,‘Heros’tuyrensis,‘Heros’wesseli,‘Heros’istlanus,‘Heros’
grammodes,‘Heros’beani (Fig. 4), which is supported by molecu-
lar phylogenies (Figs. 1–3 and Concheiro Pérez et al., 2007). The
only intergeneric relationship that showed statistical support
was the sister group relationship of Archocentrus and Herotilapia
(which is however not recovered in molecular phylogenies; Figs.
1–3 and Concheiro Pérez et al., 2007). In agreement with molec-
ular data the morphology-based tree (Fig. 4) supported the mono-
phyly and separate status of ‘Heros’festae group, as well as the
non-synonymy of ‘Heros’sieboldii with Tomocichla. In conflict
with the molecular phylogenies (Figs. 1–3 and Concheiro Pérez
et al., 2007), Petenia was nested inside Caquetaia,Astatheros macr-
acanthus was recovered within Nandopsis,Thorichthys rendered
the rest of Astatheros paraphyletic, Vieja was recovered within
Paratheraps, and Theraps was recovered as non-monophyletic
(Fig. 4).
3.4. Generic composition of heroine cichlids based on a combined
morphological—cytb phylogeny
Only two of our data sets have a full taxon sampling at the
species level (morphology and cytb). A full taxon sampling is
however important in order to determine monophyly of all gen-
era. A combined analysis of the MOR-cytb data set was per-
formed under MP (Fig. 5). PTP congruence test did not reject
the null hypothesis of homogeneity of both data sets (P= 0.9)
and they were thus analyzed simultaneously. Morphological
characters were weighted 1:1 with respect to cytb gene sequence
Table 3
Results of bootstrap analysis (N= 1000) and partitioned bremer support (PBS) for the combined morphology plus cytb (mor-cytb) data set (see tree nodes in Fig. 5)
Node 1 2 3 4 5 6 7 8 9 10
MP bootstrap 58 94 100 100 53 65 100 98 96 52
BS 10 40 71.25 74.75 15.5 16 117 58.5 44.5 12.5
PBS Mor 0 10 21.25 0.25 20.5 26 27 31.5 9.5 32.25
Cytb 10 50 50 75 510 90 90 35 20
Node 11 12 13 14 15 16 17 18 19 20
MP bootstrap 100 100 — — 68 68 100 92 92 99
BS 54.5 66.25 12.25 9.5 16.25 15.75 39.5 22 25 58.5
PBS Mor 14.5 11.25 22.25 9.5 8.75 4.25 4.5 7 5 1.75
Cytb 40 55 10 0 25 20 35 15 20 60
Node 21 22 23 24 25 26 27 28 29 30
MP bootstrap 100 100 100 99 82 95 100 100 100 74
BS 55 50 78.5 41 30.5 25 88.5 100 151.75 20.75
PBS Mor 006.5 21 24.5 0 1.5 5 61.75 39.25
Cytb 55 50 85 20 55 25 90 95 90 60
Node 31 32 33 34 35 36 37 38 39 40
MP bootstrap 78 ——100 —95 ——100 89
BS 47.25 11.5 8.25 114 15.5 37.25 10.25 5.25 92.75 31.5
PBS Mor 37.5 68.5 17.5 94 9.5 2.25 54.75 5.25 42.75 11.5
Cytb 10 80 10 20 25 35 65 0 50 20
Node 41 42 43 44 45 46 47 48 49 50
MP bootstrap 98 79 95 86 —99 55 — 100 61
BS 35.25 17 34.75 35.5 7.25 82.5 5.5 0.75 226.55 16
PBS Mor 5.25 325.25 5.5 2.25 17.5 -29.5 -34.25 -33.5 1
Cytb 30 20 60 30 5 100 35 35 260 15
Node 51 52 53 54 55 56 57 58 59 60
MP bootstrap 94 —100 ——100 ——97 71
BS 31.5 10.5 131 0.5 5.5 67.75 5.5 0.75 41.75 10.75
PBS Mor 23.5 0.5 76 14.5 29.5 17.75 29.5 34.25 43.25 24.25
Cytb 55 10 55 15 35 50 35 35 85 35
Node 61 62 63 64 65 66 67 68 69 70
MP bootstrap — 100 57 94 75 ——100 95 —
BS 0.75 77 19.75 38.75 14.5 3.25 0.75 144.25 57.25 0.75
PBS Mor 34.25 82 85.25 11.25 4.5 16.75 34.25 0.75 17.75 34.25
Cytb 35 5105 50 10 20 35 145 75 35
Node 71 72 73 74 75 76 77 78 79 80
MP bootstrap — 63 — 76 100 —100 84 ——
BS 0.75 12 2.25 17.5 168.75 5.25 71.25 40.25 0.75 5.25
PBS Mor 34.25 7 47.75 47.5 18.75 40.25 11.25 54.75 34.25 5.25
Cytb 35 5 50 65 150 -35 60 95 35 0
Node 81 82 83 84 85 86
MP bootstrap 99 —99 ——89
BS 37.25 5.25 73.5 9 5.25 24.5
PBS Mor 52.75 54.75 11.5 9 40.25 0.5
Cytb 90 60 85 0 35 25
BS, total bremer support. Values in bold show strong support in the bootstrap analysis and agreement among the two data partitions as evaluated by PBS.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 949
Table 4
Tests of alternative hypotheses on the morphology (mor), cytb, and mor-cytb data sets
Test no. Genus/species group/species Explanation/alternative hypotheses Tested data set (length difference) Compare-2 Test Templeton Kishino-hasegawa
Tomocichla
1Tomocichla (sensu Bussing, 1975) Includes´Heros
´sieboldii Comb. data (262.75) 0.01*0.0016*0.0029*
2‘Heros’sieboldii group Monophyletic Mor data (2.25) 0.41 0.8653 0.8079
3‘Heros’sieboldii Part of herichthyines Comb. data (177.75) 0.01*0.0120*0.0132*
4‘Heros’tuyrensis Part of herichthyines Comb. data (158.5) 0.02*0.0421*0.0387*
Caquetaia-Petenia
5Caquetaia (sensu Miller) Includes Petenia splendida Comb. data (135.5) 0.04*0.0730 0.1022
6Cytb data (30) 0.02*0.0451*0.0431*
7Caquetaia (sensu Kullander, 2003) Includes C. umbrifera Cytb data (9) 0.11 0.6474 0.5702
8 Mor data (28.75) 0.22 0.3033 0.2647
9‘Heros’urophthalmus Part of Amphilophus s. str. Comb. data (278.75) 0.01*<0.0001*<0.0001*
10 ‘Heros’salvini Part of Amphilophina Comb. data (166.25) 0.02*0.0055*0.0309*
11 Sister group of´Heros
´grammodes Comb. data (107.50) 0.01*0.0067*0.0071*
12 Sister group of Herotilapia Mor data (50) 0.13 0.2786 0.1978
13 Sister group of Thorichthys Mor data (28.75) 0.06 0.3360 0.4625
14 ‘Heros’grammodes Part of amphilophines Comb. data (108.75) 0.02*0.0978 0.1156
15 Sister to´Heros
´istlanus Comb. data (136.25) 0.01*0.0319*0.0130*
Amphilophus–Archocentrus
16 Amphilophus (sensu Kullander, 2003) Monophyletic Comb. data (572.5) 0.01*<0.0001*<0.0001*
17 Amphilophus Trimaculatus group part of Parachromis Comb. data (17) 0.25 0.6504 0.5247
18 Archocentrus A. centrarchus sister to Herotilapia Comb. data (245.50) 0.01*0.0021*0.013*
19 Archocentrus (sensu Miller, 1993) Monophyletic Comb. data (76) 0.15 0.3288 0.3299
20 Cytb data (16) 0.12 0.1508 0.1573
21 Archocentrus (sensu Kullander, 2003) Monophyletic, includes´Heros
´panamenesis Comb. data (76) 0.09 0.3288 0.3299
22 Cytb data (18) 0.12 0.0297*0.0441*
23 Cryptoheros (sensu Algayer, 2001) Monophyletic Comb. data (0.5) 0.84 0.8813 0.9849
24 Monophyletic Cytb data (5) 0.44 0.5721 0.5690
25 Cryptoheros panamensis Part of Neetroplus (sister to N. nematopus) Comb. data (114.75) 0.02*0.0658 0.1121
26 Cytb data (16) 0.1 0.1255 0.1202
27 Part of Cryptoheros Cytb data (5) 0.42 0.5973 0.6043
28 Astatheros Part of Amphilophus s. str. Comb. data (256) 0.01*0.0002*0.0001*
29 monophyl. s. lat. Mor data (16.25) 0.39 0.9413 0.6368
30 Astatheros Sister group of Thorichthys Comb. data (76.25) 0.23 0.2256 0.2986
31 Cytb data (17) 0.03*0.2759 0.2330
32 Paraphyletic to Thorichthys Comb. data (164.5) 0.01*0.0137*0.0308*
33 Cytb data (37) 0.01*0.0003*0.0003*
34 ‘Heros’calobrensis Part of Astatheros Comb. data (81.5) 0.01*0.0768 0.2083
35 Cytb data (26) 0.01*0.1868 0.1462
36 Mor data (20.5) 0.36 0.5377 0.6367
37 Part of Amphilophus s. str. Comb. data (116) 0.04*0.0766 0.0836
38 ‘Heros’istlanus group Part of Amphilophus s. str. Comb. data (123) 0.02*0.0138*0.016*
39 Sister group of´Heros
´urophthalmus Comb. data (137) 0.02*0.0248*0.0131*
40 Monophyletic Mor data (13.25) 0.35 0.6774 0.7002
41 Rocio octofasciata Part of´Heros
´festae group Comb. data (10.25) 0.37 0.5321 0.8645
42 Cytb data (7) 0.51 0.4555 0.4948
43 ‘Heros’festae group Part of Amphilophus s. str. Comb. data (340.5) 0.01*<0.0001*<0.0001*
44 Part of Nandopsis Comb. data (191.25) 0.01*0.0108*0.0099*
45 Rocio +‘H.’festae gr. +Tomocichla Monophyletic Comb. data (31.25) 0.19 0.5291 0.6503
46 Cytb data (10) 0.30 0.6424 0.5080
47 Mor data (30.25) 0.01*0.3506 0.3156
950 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
data, while keeping the between-state scaling of the morpholog-
ical characters as detailed in the Section 2. The MP analysis of
the MOR-cytb data set (590 PI characters) resulted in three MP
trees (L= 34415.75; CI = 0.16; RI = 0.51). The difference between
the three recovered MP trees was only in the relative phyloge-
netic position of the genera Mesonauta,Uaru and Heros, all out-
group taxa. The topology of Fig. 5 shows the three genera as a
monophyletic group, in agreement with previously published
studies (see e.g. Farias et al., 2000). Heroines formed a monophy-
letic group with respect to outgroup taxa, and the monophyly of
CAM heroines (node 86) received high support (Table 3). As ex-
pected, bootstrap and Bremer supports for deeper nodes within
CAM heroines were generally weak (Table 3). The recovered phy-
logenetic relationships among CAM heroines based on the mor-
cytb data set were largely dominated by the cytb gene phyloge-
netic signal (as also evidenced by partitioned Bremer support,
PBS, Table 3). Six main clades as in combined molecular analyses
(see above) were recovered within CAM heroines (Fig. 5): (1)
amphilophines (node 63) with moderate support (Table 3); (2)
herichthyines (node 38), albeit without strong support (Table
3), and including Rocio octofasciata and Herotilapia multispinosa,
which were however recovered with relatively strong statistical
support as sister group of Astatheros, and forming the astathero-
ines based on the nuclear and nuclear-cytb sequence data (Figs.
1–3); (3) Australoheros (node 75) with high support (Table 3);
(4) Astatheros (node 84) with high support except A. macracan-
thus (Table 3); (5) Nandopsis (node 69) with high support (Table
3), and (6) Caquetaia +Heroina (node 67), albeit without strong
support (Table 3). The latter four clades were more closely
related to amphilophines than to herichthyines (Fig. 3).
The recovered tree showed relatively high resolving power at
the genus level (Fig. 3), and the main recovered clades supported
the following non-monotypic genera: ‘Heros’festae group, Tomo-
cichla (but not including T. sieboldii), Thorichthys,Herichthys,
Theraps,Paraneetroplus,Vieja,Paratheraps,Astatheros,Australoher-
os,Nandopsis,Caquetaia,Amphilophus s. str., Cryptoheros (includ-
ing also C. nigrofasciatus; see Concheiro Pérez et al., 2007),
Parachromis,Amphilophus, and Archocentrus. In addition, all
monotypic genera were found as isolated lineages, not disrupting
monophyly of other established genera (Fig. 5). Alternative phy-
logenetic hypotheses pertaining to the genus level taxonomy as
found in literature (or as recovered in the phylogenetic analysis
of morphological characters; Fig. 4) were tested (Table 4). The ‘
Heros’ species groups in Fig. 5 represent putative new genera
based on these results.
3.5. Phylogenetic analyses of a combined morphological and molecular
data set
The addition of morphological data to the combined molecu-
lar data set into a total evidence analysis under MP rendered a
phylogeny (Fig. 6) that is identical in topology to the MP phylog-
eny based only on the combined molecular data set (Fig. 2B).
Tests of alternative phylogenetic hypotheses pertaining to the
genus level taxonomy based on this data set are shown in Table
6. The most basal position of Nandopsis within the CAM heroine
group (instead of Australoheros;asinConcheiro Pérez et al.,
2007) could not be rejected with our data (Table 6, tests 9–
12). A consensus tree of all tree-building algorithms and ana-
lyzed data sets is given in Fig. 7 to summarize the common sig-
nal of all data sets. Important life-history traits have been
mapped onto the consensus tree in Fig. 7 in order to demon-
strate cases of probable parallel evolution (Fig. 8A). Traditional
taxonomy (the most complex and conflicting cases) has also
been mapped onto the consensus tree (Fig. 8B).
48 Vieja (sensu Kullander, 2003) Monophyletic (excluding´Heros
´tuyrensis) Comb. data (224) 0.01*0.0005*0.0026*
49 Vieja (including Paratheraps) Monophyletic Cytb data (6) 0.46 0.2888 0.2890
50 Paraneetroplus P. regani group part of Vieja s. lat. Comb. data (117.75) 0.16 0.0689 0.087
51 Cytb data (17) 0.09 0.0407*0.0406*
52 Monophyletic (including regani group) Mor data (22.75) 0.23 0.4187 0.3356
53 Theraps s. str. Includes T. tentiginosus/coeruleus Cytb data (2) 0.76 0.7849 0.7964
54 Theraps s. lat. Bocourti gr. part of Herichthys Comb. data (133.5) 0.01*0.0002*0.0004*
55 Bocourti group part of Vieja s. lat. Comb. data (83.25) 0.01*0.0026*0.0031*
56 Heterospilus part of Herichthys Comb. data (122.5) 0.01*0.0018*0.0011*
57 Heterospilus part of Vieja s. lat. Comb. data (74.5) 0.22 0.0748 0.0754
58 Cytb data (22) 0.06 0.0217*0.0217*
59 Microphthalmus group part of Vieja s. lat. Comb. data (134.5) 0.02*0.0416*0.0525
60 Incl.microp. gr., bocourti gr. and heterospilus Mor data (65.5) 0.14 0.1359 0.1334
61 T. nourissati part of Astatheros Comb. data (234.75) 0.01*0.0007*0.0005*
62 ‘Heros’wesseli Part of Theraps s. lat. Comb. data (179.50) 0.01*0.0002*0.0003*
63 Nandopsis (sensu Kullander, 2003) Non-monophyletic, excludes N. ramsdeni Comb. data (387.25) 0.01*<0.0001*<0.0001*
64 Nandopsis (sensu Miller, 1966) Paraphyletic, Astatheros sister to N. ramsdeni Comb. data (241.25) 0.01*0.0004*0.0002*
65 Monophyl., ramsdeni not sister to tetracanthus Comb. data (144.25) 0.01*0.0064*0.0050*
66 Ramsdeni sister group of bocourti group Comb. data (477.25) 0.01*<0.0001*<0.0001*
67 Monophyletic Mor data (6.75) 0.57 0.8311 0.7834
68 Amphilophines Monophyletic Mor data (235) 0.01*0.0020*0.0009*
69 Herichthyines Monophyletic Mor data (179.75) 0.01*0.0083*0.0083*
‘‘Part of” is tested as ‘‘sister group to”.
*
Statistically significant test results (at 0.05). References to alternative topologies are in Appendix 1.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 951
4. Discussion
4.1. Data sets and taxon sampling
The results of our study show that the inclusion of two nuclear
markers with markedly different substitution rates under an exten-
sive taxon-sampling scheme was crucial to increase resolution and
support of heroine cichlid phylogeny. This result was not only due
to inclusion of additional sequence data, but also to the fact that
the nuclear gene phylogenetic signal was distributed at different
taxonomic levels, and was highly congruent with cytb gene signal.
Moreover, a strong point of the study aimed to maximize resolution
was the extensive sampling strategy, which included many more
heroine lineages that any previous study thus far. We tested the rel-
ative importance of taxon-sampling versus increased number of
characters by analyzing a data set which included three additional
markers under a limited taxon-sampling of 18 lineages (species),
and resulted from combining the present study with that of Chakra-
barty(2006). Clade support did not improve with the addition of new
sequence data (not shown), in agreement with many previous
empirical and simulation studies, which concluded that addition of
key taxa is in most cases more important than inclusion of extra data
(Hillis, 1996; Graybeal, 1998; Yoder and Irwin, 1999; Zwickl and
Hillis, 2002; Hillis et al., 2003).
Fig. 7. Consensus tree of all tree-building algorithms and analyzed data sets. Support for nodes in the morphological and molecular partitions is given. A black box
demonstrates strong support in all analyses, grey box in some and white box in none (a white box does not necessarily mean conflict, especially in the morphological
partition, where resolution and support at deeper nodes is weak). The letters ‘‘M” (mitochondrial) and ‘‘N” (nuclear) in the molecular partition box show which partition
supports the grouping. Boxes at terminal branches test monophyly of genera (monotypic taxa are without boxes at terminal branches). Molecular boxes with a hash mark
show genera where too few species were included to enable testing of generic monophyly in nuclear genes. Double lines show genera with questionable monophyly
(Cryptoheros may be best synonymized with Hypsophrys; the monophyly of the ‘Heros’istlanus group and of the ‘Amphilophus’trimaculatus group is disrupted by a conflict
between nuclear and mitochondrial genes). Taxa highlighted in grey represent putative new genera. The ‘Heros’istlanus group is the least supported putative genus, held
together strongly only by the cytb data (clearly requiring further study). The ‘Amphilophus’trimaculatus group may be part of Amphilophus according to some of our analyses.
The placement of the ‘Heros’wesseli group as the sister group of the ‘Heros’sieboldii group, and the monophyly of the latter is tentative, based solely on the cytb data (nuclear
data not available for ‘Heros’wesseli and ‘Heros’tuyrensis). CGH, crown-group herichthyines.
952 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
Morphology is by far the most conflicting data partition in
our phylogenetic analyses (Fig. 4,Table 5). This result could be
attributed to a general lack of phylogenetic information of mor-
phological data at the suprageneric level, likely due to homo-
plasy and convergent evolution, which are common in cichlids
(Kocher et al., 1993; Meyer, 1993; Martin and Bermingham,
1998; Rüber et al., 1999; Rüber and Adams, 2001; López-Ferná-
ndez et al., 2005a) and found in many instances in our study
(see Fig. 8A for a few examples). For instance, in Geophaginae
cichlids, phylogentic analyses based on morphological data
grouped small-bodied taxa into a monophyletic group, which
was rejected based on molecular data (López-Fernández et al.,
2005a). Similarly, in our phylogenetic analysis based on mor-
phology (Fig. 4), small-bodied species (with high anal fin meris-
tics; Archocentrus,Herotilapia,Cryptoheros), extreme piscivores
(Petenia,Caquetaia), extreme rheophilic species (Theraps irregu-
laris,Paraneetroplus bulleri), and genera with a similar benthic
life style and morphology (Astatheros,Thorichthys) were recov-
ered together, respectively, against molecular evidence. In other
words, the reconstructed phylogenies support convergent evolu-
tion of morpho/ecotypes in different lineages of Middle Ameri-
can heroines (Fig. 8A). Nevertheless, and as expected, there
were some common agreements between morphological and
molecular evidence, but only at the generic level (e.g. monophyly
and relationships within Herichthys and Thorichthys). Overall,
conflicts in the mor-cytb analysis (Fig. 5) compared to the total
evidence analysis (Fig. 6) were less extensive as judged by PBS
(Tables 3 and 5), suggesting some degree of character
reinforcement.
4.2. Phylogeny of the tribe Heroini
Nuclear and mitochondrial data partitions arrived at highly
congruent topologies, and when combined strongly supported
the existence of six main clades of CAM heroines. The mono-
phyly and generic composition of the two largest clades i.e.,
amphilophines and herichthyines as originally described by
Concheiro Pérez et al. (2007) was further supported in the
Fig. 8. Important life-history traits and traditional taxonomy mapped onto the consensus tree of all tree-building algorithms and analyzed data sets (see Fig. 7). (A) important
life-history traits mapped in Mesquite (Maddison and Maddison, 2004) to demonstrate cases of probable parallel evolution. (B) Traditional taxonomy (the most complex and
conflicting cases) mapped in Mesquite (Maddison and Maddison, 2004) onto the consensus tree. Taxa highlighted in grey represent putative new genera.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 953
present study by nuclear genes as well as in the total evidence
analyses including also morphology. While there were indica-
tions for additional clades in the cytb gene phylogenetic analy-
ses, Concheiro Pérez et al. (2007) hesitated about their reality
based only on the cytb gene. The present study further supports
the separate status of Australoheros (australoheroines) and Nand-
opsis (nandopsines), and the likely sister group relationship of
the former to the remaining CAM heroines. Moreover, a close
relationship between Caquetaia and Heroina (caquetaines), and
of Astatheros with Herotilapia and Rocio (astatheroines), and the
separate from both main clades was also relatively well sup-
ported. Despite our phylogenetic study failed to fully resolve
phylogenetic relationships among the six main clades of CAM
heroines (Fig. 7), finding first strong evidence that these cichlids
can be divided into suprageneric groups is already a big leap for-
ward compared to the stagnation of heroine systematics
throughout the last century. In order to achieve higher resolu-
tion of deeper nodes in the Heroini phylogeny, future studies
will need to incorporate additional sequence data into the phy-
logenetic analyses while maintaining a thorough taxon-sampling
as in the present study.
4.3. Generic diversity among Heroini and Neotropical Cichlidae
Our analyses provide a major insight into the phylogenetic
relationships of heroine cichlids, which constitutes a necessary
prerequisite for any attempts aimed at solving the highly com-
plex generic taxonomy of the group. The recovered Heroini phy-
logeny calls for the stabilization of generic taxonomy, and for
diagnosability of putative undescribed genera, which will be pro-
vided elsewhere (in prep). Several lineages need description as
new putative genera, indicating that Heroini would be the most
genus-rich clade of Neotropical cichlids. The generic and species
diversity of the, predominantly Middle American, heroini is even
more striking when considering the area relationship between
Middle and South America, the former being only 9.5% of the lat-
ter. Middle American Heroini evolved several unique morpholo-
gies not encountered or rare in their South American sister
group the Cichlasomatini (several cases of rheophily, repeated
teeth adaptations for eating plant matter, repeated adaptations
for extreme piscivory, the great spectrum in body sizes and
shapes; see Fig. 8A), and the Amazonian heroini add to this al-
ready wide morphological spectrum the deep bodied and unique
Table 5
Results of bootstrap analysis (N= 1000) and partitioned bremer support (PBS) for the total evidence analysis including indels (Fig. 6)
Data/node 1 2 3 4 5 6 7 8 9 10
MP bootstrap 100 86 81 83 96 100 93 54 100 —
BS 110.25 40.5 36.5 40.5 49.25 81.25 39.25 8.25 83 8.25
PBS—mor 15.25 5.5 13.5 5.5 14.25 31.25 9.25 46.75 246.75
RAG1 5 0 0 0 10 10 5555
S7i1 10 0 5 0 5 0 0 0 10 0
S7i2 20 0 50 0 20 5 15 15 65 15
Cytb 60 35 53510352035 535
Data/node 11 12 13 14 15 16 17 18 19 20
MP bootstrap 54 100 71 50 100 91 100 91 58 100
BS 18.5 94.75 23.75 17 116.25 60.75 124.75 35 18 120
PBS—mor 16.5 15.25 21.25 33 16.25 49.25 45.25 0 23 25
RAG1 0 510 0 0 540 0 0 0
S7i1 10 0 0 5 0 0 25 0 0 0
S7i2 0 10 0 15 0 35 10 0 0 0
Cytb 45 105 55 30 100 80 95 35 595
Data/node 21 22 23 24 25 26 27 28 29 30
MP bootstrap 56 81 100 62 98 95 97 66 — 53
BS 15 33.25 100.25 15 62.25 46.75 56.5 29 2.5 4.5
PBS—mor 5 11.75 15.25 5 7.75 26.75 18.5 19 2.5 20.5
RAG1 0 0 0 0 0 515500
S7i1 0 0 0 0 5 0 10 0 0 0
S7i2 0 515 0 5 0 25 5510
Cytb 10 50 100 10 60 25 25 20 10 15
Data/node 31 32 33 34 35 36 37 38 39 40
MP bootstrap 63 52 100 65 — — — 73 57 64
BS 26 5.75 140.5 37 37 10 10 30.25 25.25 39.75
PBS—mor 914.25 4.5 8830 30 15.25 14.75 55.25
RAG1 0 0 20 5 5 55510 15
S7i1 10 0 30 5 5 55055
S7i2 25 10 10 20 20 10 10 5 20 0
Cytb 0 30 105 55 55 40 40 15 35 75
Data/node 41 42 43 44 45 46 47
MP bootstrap — — 100 — 72 100 100
BS 10 10 216.75 39.75 7.75 204 131
PBS—mor 30 35 6.75 55.25 7.25 49 19
RAG1 553015102060
S7i1 5 0 20 5 0 15 20
S7i2 10 10 80 0 0 0 10
Cytb 40 40 80 75 5 120 60
BS PBS mor PBS RAG1 PBS S7i1 PBS S7i2 PBS cytb
2516.5 383.5 205 150 450 2095
100 15 8 6 18 83
BS, total bremer support. The PBS of individual data sets is expressed also in percent contribution to total BS (below).
954 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
morphologies of Pterophyllum,Mesonauta,Uaru or Heros.
Mouthbrooding is the only trait of Cichlasomatini not encoun-
tered among Middle American Heroini, but it is present at least
in Heros severus among South American Heroini. Thus, Heroini
appear to be the Middle American counterpart of the South
American evolutionary diversification of the Geophagini (López-
Fernández et al., 2005a,b).
4.4. Biogeographic implications of Middle American Heroini
Our results show that Middle American Heroini are not a
monophyletic group, as already stressed by Concheiro Pérez
et al., 2007. Of the major clades, two are cis-Andean South
American (australoheroines and caquetaines), one is Antillean
(nandopsines), and one is entirely Middle American (astathero-
ines). The two large suprageneric clades (herichthyines and
amphilophines) have some basal members in South America or
in terrenes classified as South America, but geologically part of
Central America (trans-Andean South America), together with
more derived Middle American members. Importantly, the areas
of distribution of Middle American herichthyines and amphilo-
phines are geographically mutually exclusive around the Mota-
gua Fault in Guatemala, the former being distributed north of
it and the latter to the south (Concheiro Pérez et al., 2007), as
in other groups such as Synbranchus–Ophisternon (Perdices
et al., 2005) and some Rhamdia (Perdices et al., 2002). Hence,
Middle American Heroini diversity results from at least three
independent colonizations from South America (amphilophines,
herichthyines, astatheroines). The three major heroine clades in
Middle America may fulfill the criteria of a radiation, i.e.,
common ancestry, rapid speciation, phenotype-environment
correlation, and trait utility ( Schluter, 2000). However, the
hypothesis of a radiation of cichlids in Middle America still
needs to be tested appropriately (Pybus et al., 2002), as well
as whether geographical movement into Middle America
triggered shift in diversification rates (Moore and Donoghue,
2007).
According to phylogenetic analyses based on the nuclear
gene, combined molecular, and total evidence data sets, Austra-
loheros is the most basal CAM heroine clade, and interestingly,
it also shows the most peripheral distribution. In some of our
phylogenetic analyses, the Antillean Nandopsis was placed puta-
tively as the second most basal CAM heroine group. Such rela-
tively basal position of Nandopsis would be in agreement with
the hypothesis that these cichlids migrated via a land bridge
from South America into the Antilles (Iturralde-Vinennt and
MacPhee, 1999; Iturralde-Vinennt, 2006), and would reject the
phylogenetic hypothesis that Antillean cichlids are deeply nested
within the cichlids of Middle America, and originated through a
drift vicariance event that took place when the islands became
separated from Yucatan (Chakrabarty, 2006). This latter hypoth-
esis was based on a molecular phylogeny ( Chakrabarty, 2006)
with a rather limited taxon-sampling (e.g. neither the South
American Australoheros nor Heroina or Caquetaia were included
in the phylogenetic analyses), insufficient to address the biogeo-
graphical question at hand.
Following the isolation of the paleo-Amazonas-Orinoco from
trans-Andean NW South America due to the rise of the Andes dated
at 10.1–11.8 MYA (Lundberg, 1998) no group could probably back
colonize cis-Andean South America. Based on the results of
Concheiro Pérez et al. (2007) and confirmed in this study, invasions
of the ancestors of herichthyines, astatheroines, and amphilo-
phines into Middle America were likely concurrent, and took place
before the final separation of cis- and trans-Andean South America,
dated at 16–26 Ma (depending on calibration method used;
Concheiro Pérez et al., 2007).
Table 6
Tests of alternative hypotheses based on combined molecular and total evidence data sets
Test no Genus/species group/species Explanation/alternative hypotheses Tested data set (length difference) Compare-2 Test Templeton Kishino-hasegawa
1‘Heros’ calobrensis group n.gen. Part of Astatheros Comb. data (233.5) 0.01*0.0004*0.0002*
2 Part of Amphilophus Comb. data (124) 0.01*0.0039*0.0046*
3Rocio Part of Andinoheros Comb. data (59.75) 0.12 0.2447 0.2178
4 Molecular data + gaps (16) 0.03*0.0733 0.0736
5 Nuclear genes (13) 0.01*0.0067*0.0067*
6Archocentrus (sensu Miller, 1993) Monophyletic Comb. data (163.5) 0.01*0.0005*0.0007*
7Cryptoheros (sensu Algayer, 2001) Monophyletic Comb. data (36.5) 0.01*0.0780 0.1366
8 Molecular data + gaps (10) 0.03*0.0253*0.0253*
9Australoheros sister to all CAM heroines Nandopsis sister to all CAM heroines Comb. data (51.5) 0.01*0.2407 0.2814
10 Molecular data + gaps (8) 0.24 0.4675 0.4653
11 Nuclear genes + gaps (11) 0.05*0.1235 0.1235
12 Nuclear genes (11) 0.03*0.0870 0.0858
13 ‘Heros’ istlanus group n.gen. Trimaculatus group + istlanus Mor data (63.5) 0.1 0.1022 0.0999
14 Cytb data (34) 0.01*0.0003*0.0003*
15 Monophyletic Nuclear data (6) 0.11 0.0578 0.0578
16 Nuclear genes + gaps (7) 0.12 0.1937 0.1936
17 Amphilophus Trimaculatus group part of Parachromis All genes + MOR (12.25) 0.29 0.5692 0.6122
‘‘Part of” is tested as ‘‘sister group of”.
*
Statistically significant test results (at 0.05). References to alternative topologies are in Supplementary Appendix 1.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 955
5. Conclusions
This study presents the first comprehensive phylogeny of the
Middle American Heroini combining both morphological and
molecular (mt and nuclear) data. We have for the first time in-
cluded in the phylogenetic analyses all putative generic lineages
as discussed in the literature. Nuclear genes provide more resolu-
tion at deeper levels, and arrive at a Heroini phylogeny highly con-
gruent with that based on cytb gene data. Morphological characters
in combination with cytb sequence data provide resolution at the
generic level. The total evidence phylogenetic analysis renders a
rather resolved phylogeny, which is our best hypothesis for phylo-
genetic relationships among CAM Heroini. The new phylogenetic
framework provided in this study constitutes a major contribution
towards stabilization of the taxonomy of Middle American cichlids.
Future comparative eco-morphological studies involving heroines
should bear on the new phylogeny, whereas previous ones need
to be interpreted with caution until revised, since many of the tra-
ditionally accepted genera and sister group relationships are found
to have no support.
Acknowledgments
Laboratory procedures of this study have been done in three dif-
ferent labs. We are thankful to their staff and colleagues, namely to
Vendula Šlechtová, Vlastimil Šlechta and Petr Ráb (Libe
ˇchov, Czech
Rep.); to Mari Källersjö and Pia Eldenäs (Stockholm, Sweden) and
to Elena Guacimara (Madrid, Spain). The morphological study has
been done in NRM (Stockholm, Sweden) and MNCN (Madrid,
Spain), many thanks to Sven Kullander (NRM) for providing labora-
tory space, access to facilities and literature and to Erik Åhlander
(NRM) for help in the collection, with X-rays and many other
things on the way. Financial support was provided by a scholarship
from Stockholm University (to O.R
ˇ.), grant from the SYNTHESYS
program for Access to the Swedish Museum of Natural History
(Stockholm, Sweden; to R.Z. and O.R
ˇ.) and to the Museo Nacional
de Ciencias Naturales (Madrid, Spain; to O.R
ˇ.) within the European
Community program ‘‘Improving the Human Research Potential
and the Socio-Economic Knowledge Base’’, and by the Grant
Agency of the University of South Bohemia 04-GAJU-58, the Helge
Ax:son Johnsons Stiftelse (Stockholm), the MSM6007665801 and
the GAC
ˇR 206/08/P003 grants to O.R
ˇ. Two anonymous referees
helped to significantly improve an earlier version of the
manuscript.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ympev.2008.07.022.
References
Barluenga, M., Stolting, K.N., Salzburger, W., Muschick, M., Meyer, A., 2006.
Sympatric speciation in Nicaraguan crater lake cichlid fish. Nature 439, 719–
723.
Brower, A.V.Z., DeSalle, R., Vogler, A.P., 1996. Gene trees, species trees and
systematics: a cladistic perspective. Ann. Rev. Ecol. Syst. 27, 423–450.
Bussing, W.A., 1985. Patterns of distribution of the Central American ichthyofauna.
In: Stehli, F.G., Webb, S.D. (Eds.), The Great American Biotic Interchange.
Plenum Publishing, pp, pp. 453–473.
Campbell, J.A., Frost, D.R., 1993. Anguid lizards of the genus Abronia: revisionary
notes, description of four new species, a phylogenetic analysis, and key. Bull.
Am. Mus. Nat. Hist. 216, 1–121.
Chakrabarty, P., 2006. Systematics and historical biogeography of Greater Antillean
Cichlidae. Mol. Phylogenet. Evol. 39, 619–627.
Chow, S., Hazama, K., 1998. Universal PCR primers for S7 ribosomal protein gene
introns in fish. Mol. Ecol. 7, 1247–1263.
Concheiro Pérez, G.A., R
ˇíc
ˇan, O., Ortí, G., Bermingham, E., Doadrio, I., Zardoya, R.,
2007. Phylogeny and biogeography of 91 species of heroine cichlids (Teleostei:
Cichlidae) based on sequences of the cytochrome bgene. Mol. Phylogenet. Evol.
43, 91–110.
Farias, I.P., Ortí, G., Meyer, A., 2000. Total evidence. molecules, morphology, and the
phylogenetics of cichlid fishes. J. Exp. Zool. 288, 76–92.
Farias, I.P., Ortí, G., Sampaio, I., Schneider, H., Meyer, A., 1999. Mitochondrial DNA
phylogeny of the family Cichlidae: monophyly and fast molecular evolution of
the Neotropical assemblage. J. Mol. Evol. 48, 703–711.
Farias, I.P., Ortí, G., Sampaio, I., Schneider, H., Meyer, A., 2001. The cytochrome b
gene as a phylogenetic marker: the limits of resolution for analyzing
relationships among cichlid fishes. J. Mol. Evol. 53, 89–103.
Gatesy, J., DeSalle, R., Wheeler, W., 1993. Alignment ambiguous nucleotide sites and
the exclusion of the systematic data. Mol. Phylogenet. Evol. 2, 152–157.
Graybeal, A., 1998. Is it better to add taxa or characters to a difficult phylogenetic
problem? Syst. Biol. 47, 9–17.
Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm to estimate
large phyogenies by maximum likelihood. Syst. Biol. 52, 696–704.
Higham, T.E., Hulsey, C.D., R
ˇíc
ˇan, O., Carroll, A.M., 2007. Feeding with speed: prey
capture evolution in cichlids. J. Evol. Biol. 20, 70–78.
Hillis, D.M., 1996. Inferring complex phylogenies. Nature 383, 130–131.
Hillis, D.M., Pollock, D.D., McGuire, J.A., Zwickl, D.J., 2003. Is sparse taxon sampling a
problem for phylogenetic inference? Syst. Biol. 52, 124–126.
Huelsenbeck, J.P., Ronquist, F., 2001. MrBayes: Bayesian inference of phylogeny.
Bioinformatics 17, 754–755. Available from: <http://brahms.ucsd.edu/
papers_jph/jph.43.pdf/>.
Hulsey, D.H., García de León, F.J., Sánchez Johnson, Y., Hendrickson, D.A., Near, T.J.,
2004. Temporal diversification of Mesoamerican cichlid fishes across a major
biogeographic boundary. Mol. Phylogenet. Evol. 31, 754–764.
Iturralde-Vinennt, M.A., 2006. Meso-Cenozoic Caribbean paleogeography:
implications for the historical biogeography of the region. Int. Geol. Rev. 48,
791–827.
Iturralde-Vinennt, M.A., MacPhee, R.D.E., 1999. Paleogeography of the Caribbean
region: implications for Cenozoic biogeography. Bull. Am. Mus. Nat. Hist., 238.
Kocher, T.D., Conroy, J.A., McKaye, K.R., Stauffer, J.R., 1993. Similar morphologies of
cichlid fish in Lakes Tanganyika and Malawi are due to convergence. Mol.
Phylogenet. Evol. 2, 158–165.
Kullander, S.O., 1998. A phylogeny and classification of the South American
Cichlidae (Teleostei: Perciformes). In: Malabarba, L.R., Reis, R.E., Vari, R.P.,
Lucena, Z.M.S., Lucena, C.A.S. (Eds.), Phylogeny and Classification of Neotropical
Fishes. Edipucrs, Porto Alegre, pp. 461–498.
Kullander, S.O., 2003. Cichlidae. In: Check List of the Freshwater Fishes of South and
Central America. Edipucrs, Porto Alegre.
Kullander, S.O., 1983. Revision of the South American cichlid genus Cichlasoma.
Swedish Museum of Natural History, Stockholm.
López, J.A., Chen, W., Ortí, G., 2004. Esociform phylogeny. Copeia 2004 (3), 449–464.
López-Fernández, H., Honeycutt, R.L., Stiassny, M.L.J., Winemiller, K.O., 2005a.
Morphology, molecules, and character congruence in the phylogeny of
South American geophagine cichlids (Perciformes, Labroidei). Zool. Scr. 34,
627–651.
López-Fernández, H., Honeycutt, R.L., Winemiller, K.O., 2005b. Molecular phylogeny
and evidence for an adaptive radiation of geophagine cichlids from South
America (Perciformes: Labroidei). Mol. Phylogenet. Evol. 34, 227–244.
Lundberg, J.G., 1998. The temporal context of the diversification of Neotropical
fishes. In: Malabarba, L.R., Vari, R.E., Lucena, R.P., Lucena, Z.M.S., Lucena, C.A.S.
(Eds.), Phylogeny and Classification of Neotropical Fishes. Edipucrs, Porto
Alegre, pp. 49–68.
Maddison, W.P., Maddison, D.R. 2004. Mesquite: a modular system for evolutionary
analysis. Version 1.05. Available from: <http://mesquiteproject.org>.
Martin, A.P., Bermingham, E., 1998. Systematics and evolution of lower Central
American cichlids inferred from analysis of cytochrome b gene sequences. Mol.
Phylogenet. Evol. 9, 192–203.
Meyer, A., 1993. Phylogenetic relationships and evolutionary processes in African
cichlids. Trends Ecol. Evol. 8, 279–284.
Miller, R.R., 1966. Geographical distribution of Central American freshwater fishes.
Copeia 1966 (4), 773–801.
Miller, R.R., 1996. Theraps wesseli, a new species of cichlid fish from the Caribbean
slope of northern Honduras. Trop. Fish Hobby 44 (10), 179–183.
Miller, R.R., Minckley, W.L., Norris, S.M., 2005. Freshwater fishes of Mexico.
University of Chicago Press. 652 pp..
Moore, B.R., Donoghue, M.J., 2007. Correlates of diversification in the plant clade
dipsacales: geographic movement and evolutionary innovations. Am. Nat. 170,
S28–S55.
Perdices, A., Bermingham, E., Montilla, A., Doadrio, I., 2002. Evolutionary history of
the genus Rhamdia (Teleostei: Pimelodidae) in Central America. Mol.
Phylogenet. Evol. 25, 172–189.
Perdices, A., Doadrio, I., Bermingham, E., 2005. Evolutionary history of synbranchid
eels (Teleostei: Synbranchiformes) in Central America and the Caribbean
islands inferred from their molecular phylogeny. Mol. Phylogenet. Evol. 37,
460–473.
Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA substitution.
Bioinformatics 14, 817–818.
Pybus, O.G., Rambaut, A., Holmes, E.C., Harvey, P.H., 2002. New inferences from tree
shape: numbers of missing taxa and population growth rates. Syst. Biol. 51,
881–888.
Regan, C.T., 1905. A revision of the fishes of the American cichlid genus Cichlosoma
and of the allied genera. Ann. Mag. Nat. Hist. 7(15):60–67 7 (15), 60–67. 225–
243, 316–340, 433–445.
956 O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957
R
ˇíc
ˇan, O., Kullander, S.O., 2006. Character- and tree-based delimitation of species in
the ‘Cichlasoma’facetum group (Teleostei, Cichlidae) with the description of a
new genus. J. Zool. Syst. Evol. Res. 44, 136–152.
R
ˇíc
ˇan, O., Musilová, Z., Muška, M., Novák, J., 2005. Development of coloration
patterns in Neotropical cichlids (Perciformes: Cichlidae: Cichlasomatinae). Folia
Zool. 54 (Monogr.1), 46 pp..
Roe, J.J., Conkel, D., Lydeard, C., 1997. Molecular systematics of Middle American
cichlids and the evolution of trophic-types in ‘Cichlasoma (Amphilophus)’ and ‘C.
(Thorichthys)’. Mol. Phylogenet. Evol. 7, 366–376.
Rüber, L., Adams, D.C., 2001. Evolutionary convergence of body shape and trophic
morphology in cichlids from Lake Tanganyika. J. Evol. Biol. 14, 325–332.
Rüber, L., Britz, R., Tan, H.H., Ng, P.K.L., Zardoya, R., 2004. Evolution of
mouthbrooding and life-history correlates in the fighting fish genus Betta.
Evolution 58, 799–813.
Rüber, L., Verheyen, E., Meyer, A., 1999. Replicated evolution of trophic
specializations in an endemic cichlid fish lineage from Lake Tanganyika. Proc.
Natl. Acad. Sci. USA 96, 10230–10235.
Schluter, D., 2000. The Ecology of Adaptive Radiation. Oxford University Press, New
York.
Swofford, D.L., 2000. PAUP*: Phylogenetic Analysis Using Parsimony (and other
Methods). Version 4.0b10. Sinauer Associates, Inc., Sunderland, Massachuseetts.
Thiele, K., 1993. The holy grail of the perfect character: the cladistic treatment of
morphometric data. Cladistics 9, 275–304.
Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997.
The ClustalX windows interface: flexible strategies for multiple sequence
alignment aided by quality analysis tools. Nucleic Acids Res. 24, 4876–
4882.
Wiens, J.J., 1995. Polymorphic characters in phylogenetic systematics. Syst. Biol. 44,
482–500.
Wiens, J.J., 1999. Polymorphism in systematics and comparative biology. Annu. Rev.
Ecol. Syst. 30, 327–362.
Wiens, J.J., 2001. Character analysis in morphological phylogenetics: problems and
solutions. Syst. Biol. 50, 688–699.
Wiley, E.O., Johnson, G.D., Dimmick, W.W., 1998. The phylogenetic relationships of
lampridiform fishes (Teleostei: Acanthomorpha), based on total-evidence
analysis of morphological and molecular data. Mol. Phylogenet. Evol. 10, 417–
425.
Yoder, A.D., Irwin, J.A., 1999. Phylogeny of the Lemuridae: effects of character and
taxon sampling on resolution of species relationships within Eulemur.
Cladistics 15, 351–361.
Zwickl, D.J., Hillis, D.M., 2002. Increased taxon sampling greatly reduces
phylogenetic error. Syst. Biol. 51, 588–598.
O. R
ˇíc
ˇan et al. / Molecular Phylogenetics and Evolution 49 (2008) 941–957 957
