Rapid radiation, ancient incomplete lineage sorting and ancient hybridization in the endemic Lake Tanganyika cichlid tribe Tropheini

Abstract

a b s t r a c t The evolutionary history of the endemic Lake Tanganyika cichlid tribe Tropheini, the sister group of the species flocks of Lake Malawi and the Lake Victoria region, was reconstructed from 2009 bp DNA sequence of two mitochondrial genes (ND2 and control region) and from 1293 AFLP markers. A period of rapid cladogenesis at the onset of the diversification of the Tropheini produced a multitude of special-ized, predominantly rock-dwelling aufwuchs-feeders that now dominate in Lake Tanganyika's shallow habitat. Nested within the stenotopic rock-dwellers is a monophyletic group of species, which also utilize more sediment-rich habitat. Most of the extant species date back to at least 0.7 million years ago. Several instances of disagreement between AFLP and mtDNA tree topology are attributed to ancient incomplete lineage sorting, introgression and hybridization. A large degree of correspondence between AFLP cluster-ing and trophic types indicated fewer cases of parallel evolution of trophic ecomorphology than previ-ously inferred from mitochondrial data.

Full text

Page 1

Rapid radiation, ancient incomplete lineage sorting and ancient hybridization
in the endemic Lake Tanganyika cichlid tribe Tropheini
Stephan Koblmüllera,*, Bernd Eggera,b, Christian Sturmbauera, Kristina M. Sefca
aDepartment of Zoology, Karl-Franzens-University Graz, Universitätsplatz 2, A-8010 Graz, Austria
bZoological Institute, University of Basel, 4051 Basel, Switzerland
a r t i c l ei n f o
Article history:
Received 5 August 2009
Revised 28 September 2009
Accepted 29 September 2009
Available online 21 October 2009
Keywords:
AFLP
Homoplasy excess test
Introgression
mtDNA
Phylogeny
a b s t r a c t
The evolutionary history of the endemic Lake Tanganyika cichlid tribe Tropheini, the sister group of the
species flocks of Lake Malawi and the Lake Victoria region, was reconstructed from 2009 bp DNA
sequence of two mitochondrial genes (ND2 and control region) and from 1293 AFLP markers. A period
of rapid cladogenesis at the onset of the diversification of the Tropheini produced a multitude of special-
ized, predominantly rock-dwelling aufwuchs-feeders that now dominate in Lake Tanganyika’s shallow
habitat. Nested within the stenotopic rock-dwellers is a monophyletic group of species, which also utilize
more sediment-rich habitat. Most of the extant species date back to at least 0.7 million years ago. Several
instances of disagreement between AFLP and mtDNA tree topology are attributed to ancient incomplete
lineage sorting, introgression and hybridization. A large degree of correspondence between AFLP cluster-
ing and trophic types indicated fewer cases of parallel evolution of trophic ecomorphology than previ-
ously inferred from mitochondrial data.
? 2009 Elsevier Inc. All rights reserved.
1. Introduction
Together with the Darwin’s finches on the Galápagos Islands,
the Hawaiian honeycreepers or the Caribbean anoline lizards, the
cichlid species flocks of the East African Great Lakes Tanganyika
(LT), Malawi (LM) and Victoria (LV) with their hundreds of endemic
species belong to the most spectacular vertebrate examples for
rapid speciation in confined environments (Fryer and Iles, 1972;
Seehausen, 2006). Such species flocks provide ample opportunity
for evolutionary biologists to gain insights into the processes creat-
ing organismic diversity (e.g. Grant, 1981; Losos, 1994; Kornfield
and Smith, 2000; Schluter, 2000; Kocher, 2004), but also pose ma-
jor challenges to the phylogenetic analyses underlying the study of
diversification. Oftentimes, attempts to resolve the chronology of
rapid cladogenesis remain unsuccessful until exceedingly large
datasets or more suitable marker types yield sufficient phyloge-
netic resolution (e.g. Yoder and Irwin, 1999; Zwickl and Hillis,
2002; Rokas and Carroll, 2005; Jian et al., 2008). Moreover, under
particular circumstances, multiple cladogenetic events may indeed
occur simultaneously and result in hard polytomies in the phyloge-
netic reconstruction (Sturmbauer et al., 2003).
With an age of 9–12 million years, LT is the oldest of the three
East African Great Lakes and harbors the morphologically, ecolog-
ically and behaviorally most diverse cichlid species assemblage,
currently consisting of 200 valid species (Koblmüller et al.,
2008a), with several more awaiting scientific description. The age
and distinctness of the LT cichlids facilitated their systematic treat-
ment to some degree. Based on morphology, the species were orga-
nized into 12 (Poll, 1986) or 16 (Takahashi, 2003) mostly endemic
tribes, which are well supported by molecular data (Koblmüller
et al., 2008a). Intra-tribal diversification, however, proved more
difficult to resolve due to the condensed sequence of cladogenetic
events, incomplete lineage sorting and introgression between spe-
cies and genera. Interestingly, molecular phylogenies revealed a
remarkable concurrence of periods of rapid cladogenesis, perhaps
associated with environmental changes, in several different tribes:
Lamprologini (Koblmüller et al., 2007a), Limnochromini (Duftner
et al., 2005), Bathybatini (Koblmüller et al., 2005), and Ectodini
(Koblmüller et al., 2004).
With 24 described species (Koblmüller et al., 2008a), the Tro-
pheini are one of the more species-rich cichlid tribes in LT.
Although endemic to LT, the Tropheini were shown to constitute
a lineage of the Haplochromini, the most species-rich and wide-
spread cichlid tribe, and actually represent the sister group to the
two species flocks of Lake Malawi and the Lake Victoria region
and of some riverine species (Salzburger et al., 2005; Koblmüller
et al., 2008b). All Tropheini species inhabit shallow, at least par-
tially rocky habitat and feed on aufwuchs or invertebrates. In the
strictly rock-dwelling genera Tropheus and Petrochromis, most spe-
cies are highly specialized aufwuchs-feeders with restricted dis-
persal over unsuitable habitat (Sturmbauer and Dallinger, 1995),
1055-7903/$ - see front matter ? 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.ympev.2009.09.032
* Corresponding author. Fax: +43 316 380 9875.
E-mail address: stephan.koblmueller@uni-graz.at (S. Koblmüller).
Molecular Phylogenetics and Evolution 55 (2010) 318–334
Contents lists available at ScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev

Page 2

which is reflected by numerous geographical color variants and
pronounced phylogeographic and population genetic structure
(Sturmbauer et al., 2005; Egger et al., 2007; Sefc et al., 2007; Wag-
ner and McCune, 2009). In contrast, species utilizing more sedi-
ment-richareas (e.g.
Simochromis
curvifrons, Ctenochromis horei, Gnathochromis pfefferi and Limnotila-
pia dardennii) generally display less genetic and phenotypic differ-
entiation (Meyer et al., 1996; Konings, 1998; Wagner and McCune,
2009). All Tropheini are maternal mouthbrooders, and their diver-
gent mating behaviors range from temporal pair bonding and
monogamous spawning in Tropheus (Egger et al., 2006) to various
degrees of promiscuity and polyandrous spawning in the other
species (Kuwamura, 1997; Sefc et al., 2009).
Most Tropheini species are found abundantly in their respective
suitable habitat (e.g. Sturmbauer et al., 2008), and moreover, dif-
ferent species often occur in sympatry. Coexistence of the closely
related algae-feeders may be facilitated by differential aggression
towards congeners, feeding site segregation in overlapping territo-
ries and trophic specialization associated with diversification of
the trophic morphology, in particular with respect to dentition,
the pharyngeal apparatus and gut length (Yamaoka, 1982; Takam-
ura, 1983; Kuwamura, 1992; Sturmbauer et al., 1992; Kohda, 1995,
1998).
While the trophic diversity provided the morphological basis
for the generic classification of the Tropheini (Yamaoka, 1983;
Poll, 1986), it may also originate from a parallel evolution of par-
ticular feeding habits and respective morphological adaptations. A
molecular phylogeny of the Tropheini based on partial sequences
of the mitochondrial cytochrome b gene and the most variable
part of the mitochondrial control region (Sturmbauer et al.,
2003) contradicted the morphological classification. All polytypic
genera resulted as polyphyletic and eco-morphologically similar
species occupied very different branches in the phylogenetic tree.
As the ancestral branches could not be resolved due to lack of
phylogenetic signal, it was concluded that the radiation of the
Tropheini represents an example for synchronized explosive spe-
ciation, perhaps driven by a vicariance event associated with a se-
vere fluctuation in water level. When a rise of the lake level
fragmented the stenotopic species into allopatric populations,
multiple lineages could diverge simultaneously and adaptations
to equivalent habitat types may have precipitated the parallel
evolution of equivalenttrophic
et al., 2003).
Recent phylogenetic data suggest that the two gene fragments
used in the previous study are poor markers for phylogenetic anal-
ysis in the timeframe of the Tropheini radiation. With its low sub-
stitution rate in cichlid fishes, the cytochrome b gene provides little
information to reconstruct the branching order of the relevant lin-
eages, while the fast evolving control region fragment might be
equally unsuitable for resolving basal nodes due to excess homo-
plasy (see e.g. Koblmüller et al., 2005). In several recent studies,
genes with intermediate substitution rates, most prominently the
NADH dehydrogenase subunit 2 gene (ND2), improved the resolu-
tion of the LT cichlid radiation, and in combination with the rapidly
evolving control region and the slowly evolving cytochrome b gene
could address both the basal and more recent splits (Klett and
Meyer, 2002; Salzburger et al., 2002a, 2005; Koblmüller et al.,
2004, 2005, 2007a,b, 2008b; Brandstätter et al., 2005; Duftner
et al., 2005; Schelly et al.; 2006; Day et al., 2007). In any case, how-
ever, mitochondrial genes represent only a single genealogy, possi-
bly leaving ancient introgression, in particular the complete
replacement of mtDNA in a species (mitochondrial capture; see
e.g. Nyingi and Agnèse, 2007; Nevado et al., 2009), or ancient
incomplete lineage sorting (Takahashi et al., 2001) undetected.
The addition of nuclear data, ideally from multiple loci, mitigates
these problems.
spp.,
Pseudosimochromis
specializations (Sturmbauer
The present study uses both nuclear markers (AFLP) and mito-
chondrial sequences (the complete ND2 gene and the complete
control region) and increased taxon sampling in order to improve
the resolution of the Tropheini phylogeny, and explores the possi-
bility of parallel evolution of trophic specializations versus mito-
chondrial introgression and ancient incomplete lineage sorting.
Furthermore, a Bayesian relaxed molecular clock model is applied
to estimate divergence times and to establish a temporal frame-
work for the diversification of the tribe.
2. Materials and methods
2.1. Taxonomic sampling and DNA extraction
Our study comprises 104 specimens of the tribe Tropheini, and
includes all described species except Simochromis margaretae and
Petrochromis macrognathus, several conspecific populations and
color morphs, and a number of phenotypic variants lacking nomi-
nal species status (Appendix A). Within the genus Petrochromis,
several morphs have recently been described in the aquarists liter-
ature, some of which might even deserve species status (Konings,
1998). The present study includes the phenotypic variants known
as P. sp. ‘‘Texas Longola” and P. sp. ‘‘macrognathus rainbow” (both
morphologically similar to P. polyodon), and P. sp. ‘‘Moshi yellow”
(probably a yellow variant of P. ephippium) as well as additional
color morphs, which we address as P. cf. polyodon and P. cf. macro-
gnathus, the nominal species they most closely resemble. Finally, a
new morph found at Katete, Zambia, was clearly different from all
other Petrochromis species and is addressed here as Petrochromis
sp. ‘‘Katete.
The genetic diversity of the genus Tropheus, which is divided
into >100 mostly allopatrically distributed color morphs (Schupke,
2003), is represented by a small number of samples selected
according to results by Egger et al. (2007). The choice of five out-
group taxa of the tribe Haplochromini—Copadichromis borleyi, Aul-
onocara sp., and three individuals of Astatotilapia burtoni—was
based on previous phylogenetic studies (Salzburger et al., 2005;
Koblmüller et al., 2008b). Most of the specimens were sampled
during several field expeditions from 1995 to 2007, and some addi-
tional samples were obtained from the aquarium trade (Appendix
A). Sampling sites are shown in Fig. 1. Voucher specimens are
stored at the Department of Zoology, University of Graz, Austria
and the Royal Africa Museum in Tervuren, Belgium. DNA was ex-
tracted from ethanol-preserved fin clips by proteinase K digestion
followed by protein precipitation with ammonium acetate.
2.2. AFLP data collection and analysis
Restriction digestion, pre-selective and selective PCR followed
the protocol described in Koblmüller et al. (2007a). For selective
amplification we used following ten primer combinations: EcoRI-
ACA/MseI-CAA, EcoRI-ACA/MseI-CAG, EcoRI-ACA/MseI-CAC, EcoRI-
ACA/MseI-CAT, EcoRI-ACT/MseI-CAA, EcoRI-ACT/MseI-CAG, EcoRI-
ACT/MseI-CAC, EcoRI-ACT/MseI-CAT, EcoRI-ACT/MseI-CTG, EcoRI-
ACG/MseI-CTG. Selective PCR products were visualized on an ABI
3130xl automated sequencer (Applied Biosystems) along with an
internal size standard (GeneScan-500 ROX, Applied Biosystems).
Raw fragment data were analyzed using Genemapper (ver. 3.7, Ap-
plied Biosystems). Presence and absence of peaks were scored by
eye in the Genemapper software within a range of 100–500 bp,
and only distinct, major fragments were considered and assembled
as a binary (1/0) matrix. In a few cases, fragments were scored as
missing data when character states could not be determined
unambiguously.
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
319

Page 3

The program TREECON 1.3b (Van de Peer and de Wachter, 1994)
was used to construct a NJ tree based on Nei and Li’s (1979) dis-
tances. Bootstrap values from 1000 pseudo-replicates were used
as a standard measure of confidence in the reconstructed tree
topology.
To test for homoplasy excess introduced by hybridization we
performed a tree-based method outlined in Seehausen (2004). In
theory, the inclusion of a hybrid taxon into a multi-locus phylog-
eny introduces homoplasy with clades that contain the hybrid par-
ents because, theoretically, hybrid taxa are overall intermediate to
the parental taxa since they carry a mosaic of parental characters.
Thus, the removal of the hybrid taxon should increase the boot-
strap support for the clades that include the parental taxa or their
descendents by decreasing the amount of homoplasy in the data-
set. The removal of non-hybrid taxa, on the other hand, should
not affect the bootstrap support of other nodes. We removed one
species at a time (or major clades within the putative P. poly-
odon-, and P. ephippium species complexes) and calculated the
bootstrap support for nodes in the AFLP tree to obtain a distribu-
tion of bootstrap values and identify potential hybrid taxa (see also
Egger et al., 2007).
2.3. mtDNA data collection and analysis
The complete NADH dehydrogenase subunit 2 gene (ND2,
1047 bp) and the complete mitochondrial control region (CR,
962 bp; excluding the poly-T region) were amplified and se-
quenced. Polymerase chain reaction (PCR), purification of PCR
products, and sequencing followed the protocol described in Duft-
ner et al. (2005). The primers used for PCR and sequencing were L-
Fig. 1. Map of Lake Tanganyika showing the sampling sites. The three deepwater basins are indicated by grey shading. S, South; N, North.
320
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 4

Pro-F (Meyer et al., 1994), TDK-D, SC-DL and TDK-DHG (Lee et al.,
1995) for the control region, and Met, ND2.2A, Trp (Kocher et al.,
1989) and ND2.T-R (Duftner et al., 2005) for ND2. DNA fragments
were purified with SephadexTMG-50 (Amersham Biosciences) fol-
lowing the manufacturer’s instruction and subsequently visualized
on an ABI 3130xl automated sequencer (Applied Biosystems). All
sequences are available from GenBank under the accession num-
bers listed in Appendix A.
DNA sequences were aligned by eye using the Sequence Naviga-
tor software (Applied Biosystems). To assess the overall phyloge-
neticsignal weperformeda
(Strimmer and von Haeseler, 1997), using TREE-PUZZLE 5.1
(Schmidt et al., 2002). Phylogenetic inference was based on neigh-
bor joining (NJ), maximum parsimony (MP), maximum likelihood
(ML) and Bayesian inference (BI). NJ and MP were performed in
PAUP*version 4.0b10 (Swofford, 2000). For NJ analysis, we applied
the best-fit substitution model selected by Modeltest 3.06 (Posada
and Crandall, 1998): TIM + I + G (Posada and Crandall, 1998) with
nucleotide frequencies A = 0.2891, C = 0.2887, G = 0.1205; propor-
tion of invariable sites (I) = 0.5403; gamma shape parameter
(a) = 0.7444; and
R-matrix
A M G = 15.968, A M T, C M G = 1.3523, C M T = 11.3378. Heuristic
tree searches under the MP criterion applied random addition of
taxa and TBR branch swapping (1000 replicates). Statistical sup-
port for the resulting topologies was assessed by bootstrapping
(1000 pseudo-replicates for NJ and MP). ML analysis was con-
ductedusingGARLI0.951
www.bio.utexas.edu/faculty/antisense/garli/Garli.html),
default settings, with all model parameters estimated. We per-
formed 1000 runs with different random starting trees. Of the
resulting trees, the one with the best likelihood score was chosen.
To assess the statistical support of the nodes, we performed 100
non-parametric bootstrap resamplings, and the resulting trees
were used to build a 50% majority-rule consensus tree in PAUP*.
For BI, performed in MrBayes 3.0b4 (Huelsenbeck and Ronquist,
2001), data were partitioned by gene, and additionally by codon
position within ND2. Rate heterogeneity was set according to a
gamma distribution with six rate categories (GTR model; Rodrí-
guez et al., 1990) for each data partition. Posterior probabilities
were obtained from Metropolis-coupled Markov chain Monte Carlo
simulations (2 independent runs; 10 chains with 3 million genera-
tions each; chain temperature: 0.2; trees sampled every 100 gener-
ations). A burn-in period of 1 million generations was discarded to
allow likelihood values to reach stability. Chain stationarity and
run parameter convergence were checked using Tracer v1.4 (Ram-
baut and Drummond, 2008), and a 50% majority-rule consensus
tree was constructed. To test for significant differences between
the topologies obtained by the different tree-building algorithms,
we performed Shimodaira–Hasegawa (SH) tests (Shimodaira and
Hasegawa, 1999; full optimization; 1000 bootstrap replicates) in
PAUP*. Differences between the mtDNA and AFLP tree topologies
were also evaluated by means of the SH test.
MtDNA lineage divergence was dated within a Bayesian MCMC
framework in the program BEAST v1.4.8 (Drummond and Ram-
baut, 2007). In order to provide calibration points, representatives
of the following groups were included into the analysis: the ‘‘mod-
ern haplochromines” (sensu Salzburger et al., 2005), the C-lineage
(includes the majority of the mouthbrooding LT cichlid tribes; Cla-
baut et al., 2005), the Lamprologini (LT’s substrate breeding cichlid
tribe), and the Eretmodini (LT’s tribe of ‘‘goby cichlids”). As calibra-
tion points we applied (i) an age of 5–6 MY for the most recent
common ancestor (MRCA) of the cichlids assigned to the C-lineage,
assuming that this radiation (‘‘primary radiation” in Salzburger
et al., 2002a) coincided with the formation of a truly lacustrine
habitat with deep water conditions in LT (Tiercelin and Monde-
guer, 1991; Lezzar et al., 1996; Cohen et al., 1997); (ii) an age of
likelihoodmappinganalysis
A M G,G M T = 1.0000,
(Zwickl,2006;availableat
applying
5–6 MYA for the MRCA of the tribe Lamprologini, assuming that
the onset of their diversification also happened in the course of
the ‘‘primary radiation” (Salzburger et al., 2002a); (iii) a maximum
age of 0.57–1 MY for the split between the utaka and mbuna cich-
lids, based on the age of the refilling of Lake Malawi (Delvaux,
1995; Sturmbauer et al., 2001) and (iv) a maximum age of
15,000 years for the Lake Victoria cichlid species flock, based on a
recent paper by Stager and Johnson (2008), who report that geo-
physical and paleoecological data indicate that Lake Victoria dried
out at least once between 18,000 and 14,000 years ago and that it
is highly unlikely that the LV species flock could have survived this
period in remnant ponds or marshes within the desiccated basin or
anywhere else. Alternative, more ancient age estimates for the East
African cichlid radiations (Genner et al., 2007) would increase the
inferred divergence times two- to fivefold, but are difficult to rec-
oncile with the evolution of the Lake Tanganyika cichlid assem-
blage (Koblmüller et al., 2008a, b).
According to the phylogenetic relationship among LT’s cichlid
tribes (Koblmüller et al., 2008a), the bathypelagic tribes Bathyba-
tini and Hemibatini were used as outgroup. As it was impossible
to unambiguously align the control region sequences of the whole
LT cichlid species flock, only the ND2 sequences were used in this
analysis. We employed the SRD06 two-partition codon-specific
rates model of nucleotide evolution, which has fewer parameters
than the GTR + I + G model, but has been shown to provide a better
fit for protein coding sequence data (Shapiro et al., 2006) and ap-
plied the uncorrelated relaxed clock model (Drummond et al.,
2006). We used default settings for all priors except for the tree
prior which was set to Yule Process (speciation) as suggested by
Drummond et al. (2007). Monophyly of clades was not enforced,
except for those used as calibration points, and operators were
auto-optimized. A first preliminary run was used for parameter
optimization. Two final analyses were run for 15 ? 106generations
each, with a 1000-step thinning. The log files were analyzed using
Tracer v1.4 (Rambaut and Drummond, 2008) to determine the
appropriate burn-in period and to ensure that runs were sampled
from the same distribution. The burn-in phase (500,000 genera-
tions) was discarded from each log and tree file before the two runs
were combined using LogCombiner v1.4.8 (a module in the BEAST
program package). Pooled post-burn-in Effective Sample Sizes
(ESSs) for all parameters were >300, indicating that the pooled
log file accurately represented the posterior distribution (Kuhner,
2009). Divergence times were derived from the pooled post-
burn-in results and TreeAnnotator v1.4.8 (a module in the BEAST
program package) was used to compute a maximum-clade-credi-
bility tree, which was visualized in FigTree v1.2.2 (Rambaut,
2009). Divergence times were calculated as mean node heights of
the 95% highest posterior density (HPD; i.e. the shortest interval
containing 95% of all values sampled from the posterior distribu-
tion) intervals on a time-measured phylogeny.
3. Results
3.1. AFLP phylogeny
AFLP data were obtained with ten primer combinations and
consisted of 1293 fragments (35 constant, 1171 parsimony infor-
mative and 87 parsimony uninformative sites). The AFLP tree
(Fig. 2) confirmed the monophyly of all species. Of the three poly-
typic genera Tropheus, Petrochromis and Simochromis, only the
genus Tropheus was resolved as monophyletic.
Short internal branch lengths and low bootstrap support (<50%)
for several nodes at the base of the tree point to a period of very
rapid cladogenesis at the onset of diversification of the tribe Tro-
pheini. The basal position of Tropheus received low bootstrap sup-
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
321

Page 5

Fig. 2. NJ tree [Nei and Li distances (Nei and Li, 1979)] based on 1293 AFLP loci. Only bootstrap values >50 are shown. Astatotilapia burtoni, Aulonocara sp. and Copadichromis
borleyi were used as outgroup taxa. Nodes with significant results in the homoplasy excess test (Fig. 7) are labelled by grey letters (A–E). The coloured bars depict the
assignment to different tropheine genera. Question marks indicate unknown sampling localities.
322
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 6

port (<50%). The remaining taxa radiated into (a) three distinct lin-
eages represented by Lobochilotes labiatus, Simochromis diagramma
and Petrochromis famula, respectively, (b) a clade of species prefer-
ring the more sediment-rich areas of the rocky habitat (for simplic-
ity addressed as the ‘‘sediment-dwellers” in the following), which
includes Simochromis marginatus and S. babaulti plus S. pleurospilus
as sister taxa, Gnathochromis pfefferi, Pseudosimochromis curvifrons,
Ctenochromis horei, and, with lower bootstrap support, Limnotilapia
dardennii, (c) a clade comprising Petrochromis orthognathus, P. fas-
ciolatus and Interochromis loockii, and (d) the remaining Petrochr-
omis as a weakly supported monophylum comprising the large-
bodied representatives of the genus. Within the latter, Petrochromis
cf. macrognathus and P. trevawasae were resolved as sister taxa at
the most ancestral branch. Petrochromis sp. ‘‘Katete” resulted as a
distinct taxon. All P. ephippium and P. sp. ‘‘Moshi yellow”, a color
morph morphologically identical to P. ephippium, grouped in a well
supported clade, as did the P. polyodon-like taxa P. polyodon, P. cf.
polyodon, P. sp. ‘‘Texas Longola” and P. sp. ‘‘macrognathus rain-
bow”. Petrochromis sp. ‘‘macrognathus rainbow” occurs sympatri-
cally with a P. cf. polyodon at Mtosi, and constitutes a separate,
well supported clade in the P. polyodon-clade.
3.2. Mitochondrial phylogeny
Information on the sequence data is given in Table 1. Likelihood
mapping yielded 88.1% fully resolved quartets for the concatenated
dataset (Fig. 3), indicating a strong phylogenetic signal.
Our analyses (MP, NJ, ML and Bayesian inference), based on the
concatenated dataset, yielded highly congruent results with only
minor differences dependent on the tree-building algorithm used
(Fig. 4). MP yielded 418 most parsimonious trees with a length of
1614 steps (CI excluding uninformative characters = 0.4148;
RI = 0.8252; RC = 0.3809). SH tests revealed no significant differ-
ences between the likelihood values of the alternative topologies
(Table 2). In contrast, a consensus of the mitochondrial reconstruc-
tions (Fig. 5) differed significantly from the AFLP-based topology
(SH test; ?ln L of 11574.8057 versus 12524.8891 for mtDNA con-
sensus tree versus AFLP topology; P < 0.001).
The four phylogenetic methods resulted in topologies with
high bootstrap support for the monophyly of each species with
the exception of the P. polyodon-like taxon (i.e. P. (cf.) polyodon,
P. sp. ‘‘macrognathus rainbow” and P. sp. ‘‘Texas Longola”) and
the taxon consisting of Petrochromis ephippium and the ephippi-
um-like morph ‘‘Moshi yellow”, and polyphyly of all polytypic
genera. As in the AFLP tree, Tropheus occupied a basal position
within the Tropheini, but in contrast to the nuclear data, Tro-
pheus duboisi was not included in the clade containing the other
Tropheus
specimens. Amongthe
branching order of the major lineages differed dependent on
the tree-building algorithm used. As with the AFLP data, short
branches and weak statistical support indicate a rapid sequence
of cladogenesis at the onset of diversification. Six well supported
mitochondrial lineages emerged in the course of this radiation: I,
Petrochromis sp. ‘‘Katete” as a distinct lineage, as in the AFLP
tree; II, P. famula sister to P. fasciolatus and Petrochromis loockii;
III, the clade of sediment-dwellers, with lower resolution but
composed of the same species as in the AFLP tree (except for
the inclusion of the rock-dwelling P. orthognathus); IV, S. dia-
gramma plus L. labiatus; V, Petrochromis sp. ‘‘Moshi yellow” plus
P. cf. macrognathus and some of the P. cf. polyodon specimens; VI,
P. trewavasae as sister group to a clade comprising P. polyodon,
the remaining P. cf. polyodon and the polyodon-like P. sp. ‘‘mac-
rognathus rainbow” plus P. ephippium specimens from the very
remainingTropheini, the
Fig. 3. Results from the likelihood mapping analysis of the concatenated mitochondrial dataset [complete ND2 and control region (excluding the poly-T region)], indicating
the presence of a strong phylogenetic signal.
Table 1
mtDNA sequence diversity within the Tropheini.
ND2 CR
Length of alignment
Pairwise sequence divergence
Number of variable sites
Number of parsimony informative sites
1047 bp
0–6.97%
300
239
962 bp
0–7.10%
252
198
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
323

Page 7

Fig. 4. Phylogeny of the Tropheini based on the complete ND2 gene and the mitochondrial control region (excluding the poly-T region) reconstructed with different tree-
building algorithms: (a) NJ tree using the TIM + I + G (Posada and Crandall, 1998) model; (b) strict consensus tree of 418 most parsimonious trees (1614 steps; CI excluding
uninformative characters, 0.4148; RI, 0.8252; RC, 0.3809); (c) the best ML tree obtained from 1000 individual runs in Garli; (d) Bayesian 50% majority-rule consensus tree.
Astatotilapia burtoni, Aulonocara sp. and Copadichromis borleyi were used as outgroup taxa. Bootstrap values (for NJ, MP and ML) > 50 and posterior probabilities (for BI) > 0.50
are shown. The coloured bars depict the assignment to different tropheine genera.
324
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 8

south of the lake. The two P. polyodon-like taxa from Mtosi—a
P. cf. polyodon and P. sp. ‘‘macrognathus rainbow” —resulted as
closely related but were not resolved as sister taxa.
We found that the combination of ND2 and control region se-
quence data improved the resolution at the base of the phylogeny
and increased the bootstrap support for many internal nodes com-
pared to Sturmbauer et al. (2003). Differences in the tree topology
exclusively concerned nodes with low bootstrap support in
Sturmbauer et al. (2003).
Table 2
Comparison of alternative phylogenetic hypotheses.
Tree
?ln L
11534.1740
11544.2547
11506.1403
11508.0332
D ?ln L
28.0338
38.1144
Best
1.8930
P
NJ
MP
ML
BI
0.137
0.087
0.765
Shimodaira–Hasegawa tests (Shimodaira and Hasegawa, 1999) were used to
determine whether NJ, MP and BI topologies differed significantly from the ML tree
under a likelihood criterion.
Fig. 5. Strict consensus of 418 most parsimonious trees, the NJ, ML, and BI tree, representing the phylogenetic relationships of the Tropheini based on the complete ND2 gene
and the mitochondrial control region (excluding the poly-T region). Bootstrap values for NJ and MP are shown above the branches, ML bootstrap values and posterior
probabilities (for BI) below. Only bootstrap values >50 and posterior probabilities >0.50 are shown. Astatotilapia burtoni, Aulonocara sp. and Copadichromis borleyi were used as
outgroup taxa. The coloured bars depict the assignment to different tropheine genera. Question marks indicate unknown sampling localities.
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
325

Page 9

Fig. 6. A chronogram of the diversification of the Tropheini lineages based on complete ND2 sequences. Divergence time estimates are represented as the mean node height
of the 95% highest posterior density (HPD) interval from a BEAST maximum-clade-credibility tree. Blue bars span the 95% HPD interval for each well supported node. No bars
are assigned to nodes with low posterior probability. Calibration points are marked by arrows. The coloured vertical bars show the assignment of species to different
tropheine genera.
326
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 10

3.3. Dating of mitochondrial cladogenesis
The chronogram obtained from a relaxed molecular clock analy-
sisoftheND2sequencesisshowninFig.6.Themeancovariancewas
slightly positive (8.56 ? 10?5), but its 95% HPD interval spanned
zero (?0.105 to 0.113), indicating that branches with fast and slow
substitutionrates arenexttoeachother,suchthatthereisnostrong
evidence for autocorrelation of rates in the phylogeny (Drummond
et al., 2006). The parameter estimates for ‘‘ucld.stdev” and the coef-
ficient of variation were 0.605 (95% HPD: 0.451–0.771) and 0.660
(95% HPD: 0.465–0.872), respectively, and indicated a significant
but not exceptionally high rate-heterogeneity among lineages, jus-
tifying the use of a relaxed clock model (Drummond et al., 2007).
We emphasize that our divergence time estimates should be re-
garded as approximate only, since the 95% HPD intervals are large
and often overlap for successive splits; moreover, some branching
sequences were resolved differently in the nuclear tree.
The Tropheini were estimated to have diverged from the
remaining ‘‘modern haplochromines” (sensu Salzburger et al.,
2005) at ?2.8 MYA. The genus Tropheus (excluding T. duboisi) split
from the remaining Tropheini ?2.4 MYA, and the common ances-
tor of the Tropheus clade dates back to ?1.4 MYA. Between 2.1
and 1.5 MYA, a series of rapid splits gave rise to most of the
remaining major lineages, i.e. T. duboisi, Petrochromis sp. ‘‘Katete”,
several groups of the sediment-dwelling species, S. diagramma plus
L. labiatus, and the clade comprising Petrochromis sp. ‘‘Moshi yel-
low”, P. cf. macrognathus and some of the P. cf. polyodon specimens.
More recent events are concentrated around ?1.2 and ?0.7 MYA
and include the divergence of P. famula from P. fasciolatus and I.
loocki at ?1.2 MYA and the split between the latter two species
at ?0.7 MYA; the split between S. diagramma and L. labiatus at
?1.2 MYA; the split between C. horei and L. dardennii at ?0.7
MYA; the divergence of P. trevawasae from a clade containing P.
polyodon-like taxa and P. ephippium at ?1.1 MYA; the divergence
of the latter clade into three lineages, represented by individuals
from the south, the central west and the central east (Tanzanian)
coast, at ?0.8 MYA; and the divergence between S. babaulti and
S. pleurospilus at ?0.8 MYA.
3.4. Ancient and recent hybridization
Tests for excess homoplasy in the AFLP data indicated several
cases of putative hybrid origin or substantial ancient interspecific
gene flow (Fig. 7): C. horei as potential hybrid between G. pfefferi
and the ancestor of its AFLP sister clade containing P. curvifrons,
S. babaulti, S. pleurospilus and S. marginatus (node C); a potential
hybrid origin of P. sp. ‘‘Katete” involving the ancestor of the sedi-
ment-dwelling lineage (node B) and the ancestor of the large Pet-
rochromis species (node D); massive ancient gene flow from the
P. ephippium/polyodon clade (node E) into P. cf. macrognathus and
S. diagramma; and finally, ancient gene flow from the Petrochr-
omis/Interochromis ancestor (node A) into L. labiatus. The place-
ment of P. cf. macrognathus within the P. ephippium/polyodon
clade in the mitochondrial trees is consistent with the AFLP-based
inference of ancient gene flow or ancestral hybridization; in con-
trast, low statistical support makes the mitochondrial placement
of S. diagramma and L. labiatus inconclusive with regard to the
source of ancient introgression. Recent mtDNA introgression is
indicated by the occurrence of P. ephippium in the P. polyodon
clade, and the placement of two P. cf. polyodon specimens amongst
the P. ephippium-like P. sp. ‘‘Moshi yellow” in the mitochondrial
tree (Figs. 4 and 5), while both, the P. ephippium- and the P. poly-
odon-like taxa, constitute two well supported clades in the AFLP
tree (Fig. 2).
4. Discussion
4.1. Topological disagreement between nuclear and mitochondrial
trees
Mitochondrial DNA sequences and AFLP data both identified a
period of rapid cladogenesis at the onset of the diversification of
Fig. 7. Putative hybrid and introgressed taxa identified by the AFLP homoplasy excess test. The boxplots represent the distributions of bootstrap support values obtained in
21 taxon-reduced bootstrap analyses for the nodes A–E (as identified in Fig. 2). Significant outliers are represented by asterisks labelled with the taxon, whose removal caused
the increase in bootstrap support for the respective node (i.e. the putative hybrid/introgressed taxon).
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
327

Page 11

the Tropheini, but disagreed in the placement of several taxa. The
SH test indicated significant conflict between the tree topologies
derived from the two datasets. The perhaps most striking disagree-
ments concerned the phylogenetic relationship between T. duboisi
and its congeners, which was resolved as monophyletic in the AFLP
tree but not by mitochondrial data, and the placement of P. ortho-
gnathus. In the mitochondrial tree, the strictly rock-dwelling P.
orthognathus was nested among species from different genera that
prefer the shallow sediment-rich habitat, which is much at odds
with morphological and ecological aspects. In contrast, nuclear
data provide strong evidence for a close relationship of P. orthogna-
thus with the equally rock-dwelling P. fasciolatus and I. loockii,
which also fits the expectations based on the species’ morphology.
P. orthognathus and I. loockii share a terminal mouth and are similar
to P. fasciolatus with its upward-pointing mouth, whereas all other
Petrochrochromis species have a slightly downward-pointing
mouth. It is noteworthy that I. loockii is known among aquarists
as P. sp. ‘‘orthognathus tricolor” (e.g. Konings, 1998), reflecting
the morphological similarity between the two species.
Conflicts among datasets at the interspecific level are usually
attributed to past introgressive hybridization or ancient incom-
plete lineage sorting (e.g. Shaw, 2002; Machado and Hey,
2003), two processes that are difficult to distinguish using strict
hypothesis testing (Knowles, 2004). However, since both the
mitochondrial and AFLP data indicate a rather short period of ra-
pid cladogenesis with short internal branches, ancient incom-
plete lineage sorting may be the primary cause for most of the
observed inconsistencies between the phylogenetic reconstruc-
tions (Takahashi et al., 2001; McCracken and Sorenson, 2005).
In addition to lineage sorting issues, tests on the AFLP data sug-
gested several instances of ancient hybridization and perhaps the
hybrid origin of two taxa, which explains the discrepant place-
ment of the involved samples in the two trees. Overall, we sug-
gest that the tree based on the multilocus AFLP dataset, where a
large number of independently transmitted loci neutralizes dif-
ferential lineage sorting (Avise, 2004), reflects the genome-wide
genetic similarities between taxa better and thus represents a
closer approximation of the species tree than the mtDNA phylog-
eny, which essentially represents a single gene tree. Nonetheless,
the mitochondrial data supplied valuable information for the
detection of recent hybridization, and for the reconstruction of
the chronology of the evolution of the Tropheini, which is char-
acterized by a radiation into several major lineages near the base
of their phylogeny at 1.5–2.1 MYA. Most of the extant species
emerged soon after the initial diversification and date back to
some 0.7–1.2 MYA according to our calibration. The split of
the Tropheini from the remaining ‘‘modern haplochromines”
(sensu Salzburger et al., 2005) was estimated to an age of
approximately 2.8 MYA. Alternative dating following Genner
et al. (2007) increases the age of radiation and the origin of
the current species by two- to fivefold.
Some previous molecular phylogenetic studies on cichlid fishes
based on various nuclear markers included part of the taxa in-
cluded in our study. Nishida (1997) employed allozymes to derive
a comprehensive phylogenetic framework for the entire Lake Tang-
anyika cichlid species flock, including 16 tropheine species.
Although the Tropheini were resolved as monophylum, both their
monophyly and the internal branching order received only low sta-
tistical support. Similar results were obtained by Takahashi et al.
(1998), who, based on SINE insertions, found a monophylum Tro-
pheini but no further intra-tribal resolution. Kassam et al. (2006)
used AFLPs to investigate the phylogenetic relationships between
five tropheine species and seven Lake Malawi haplochromine spe-
cies. The phylogenetic relationships among the tropheine species
in their study are largely congruent to our phylogenetic hypothe-
sis. The inconsistent placement of P. famula is likely due to the ef-
fect of different taxon sampling on the phylogenetic accuracy (e.g.
Rannala et al., 1998; Hedtke et al., 2006).
Incongruence between nuclear and mitochondrial tree topolo-
gies was also observed in phylogenetic reconstructions of other
LT cichlids. Given the rapid diversification of most tribes, incom-
plete lineage sorting is always an issue, but in many cases ancient
introgression was also inferred (see Rüber et al., 2001; Salzburger
et al., 2002b; Schelly et al., 2006; Koblmüller et al., 2007b; Takah-
ashi et al., 2007), culminating in evidence for complete mtDNA
replacement in some cichlid species (Nyingi and Agnèse, 2007;
Nevado et al., 2009). The cases of ancient and recent hybridization
suspected from the present study call for additional investigations
with increased sampling at population level.
In the genus Tropheus, gene flow between color morphs can
be high periodically due to the recurrent breakdown of geo-
graphic isolation, and discrepant AFLP and mitochondrial data
were interpreted to expose evolutionary processes at different
timescales (Egger et al., 2007): The mitochondrial data retain an-
cient signals of large-scale migration events, presumably trig-
gered by major lake level fluctuations (see also Baric et al.,
2003; Sturmbauer et al., 2005), whereas the nuclear data reflect
recent gene flow and current genetic continuity among adjacent
populations and are largely congruent with species classifica-
tions, color pattern similarities and the current geographic distri-
bution of populations (Egger et al., 2007). A similar pattern may
be revealed in future studies of the genus Petrochromis, which
are also specialized rock-dwellers with pronounced geographic
color pattern variation.
4.2. Eco-morphological groups and parallel evolution of trophic types
In the AFLP-based estimation of the species tree, groups of
ecologically and morphologically similar taxa became evident:
(i) species of the strictly rock-dwelling genus Tropheus, which
have bicuspid teeth in the outer row and curved conical teeth
on the sides of the premaxillary bone (Poll, 1986) that are used
to bite and tear off pieces of filamentous epilithic algae
[=”browser”; sensu Yamaoka (1997)]; (ii) species of various gen-
era (Ctenochromis, Gnathochromis, Limnotilapia, Pseudosimochr-
omis, Simochromis), which inhabit the shallow sediment-rich,
sometimes vegetated, habitat and feed on algae and/or inverte-
brates; (iii) the genus Petrochromis, which has elongated tricus-
pid teeth as adaptation to comb unicellular algae, mainly
diatoms, and detritus from filamentous algae attached to the
rocky substrate (Yamaoka, 1997), together with I. loockii, whose
dentition, although bicuspid, resembles that of Petrochromis spp.
(Yamaoka et al., 1998); and (iv) within the Petrochromis/Inte-
rochromis clade, the distinction between species with a clearly
downwards pointing mouth (P. ephippium, P. polyodon, P. treva-
wasae, and several undescribed taxa) and those with an upwards
pointing, terminal, or only slightly downwards pointing mouth
(P. fasciolatus, P. orthognathus, P. famula, and I. loockii). Moreover,
the two Petrochromis clades largely correspond to Yamaoka’s
(1982, 1997) classification of slow (P. polyodon and P. trewava-
sae), intermediate (P. famula and P. orthognathus; I. loocki) and
fast feeders (P. fasciolatus), although Yamaoka (1982) groups P.
famula with P. polyodon and P. trewavasae based on number of
teeth and jaw musculature. The different Petrochromis species of-
ten occur sympatrically, and the differentiation in jaw morphol-
ogy and combing speed may contribute to niche-partitioning
with regard to the characteristics of the grazed surface and the
precision of food selection. Nonetheless, interspecific competition
and aggression among these species is high (Takamura, 1984;
Kohda, 1995, 1998), and while adults of the first Petrochromis-
group are quite large, aggressive and territorial, those of the sec-
ond group are usually somewhat smaller and regularly form
328
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 12

large schools to intrude the territories of the larger Petrochromis,
and those of other cichlids, for foraging (Konings, 1998; Kohda
and Takemon, 1996). In the clade comprising the species of the
sedimented habitat, G. pfefferi picks shrimp from the substrate
(Yamaoka, 1997), C. horei takes invertebrates and fry from the
substrate and the water column, and P. curvifrons, Simochromis
spp. and L. dardennii browse filamentous algae from rock sur-
faces, similar to the feeding behavior of Tropheus (see above;
Yamaoka, 1997). S. diagramma and L. dardennii were also ob-
served in muddy areas and are considered to possess great
eco-morphological versatility with a low degree of specialization
for browsing (Yamaoka, 1997). As these two species represent
ancestral splits in relation to the remaining browsers of the sed-
imented habitat in the AFLP tree, evolution in this clade appar-
ently proceeded towards higher specialization for browsing. A
distinct lineage is represented by the peculiar, strictly rock-
dwelling L. labiatus, characterized by huge lips with fleshy lobes
that are pressed around interstices between rocks or gravel,
forming a sucking disc when feeding on Trichoptera and Diptera
larvae (Hori, 1991; Yamaoka, 1997).
The correspondence between AFLP clustering and types of feed-
ing behavior suggests fewer cases of evolutionary parallelism in
the evolution of trophic specialization than inferred by Sturmbauer
et al. (2003) based on their mitochondrial tree. Nonetheless, given
the short internal branches and the weak support for some of the
clusters in the AFLP tree, we cannot rule out that some species,
e.g. within Petrochromis, acquired similar feeding habits indepen-
dently even though they seem to share a common ancestor. Also,
the observed paraphyly of the genus Simochromis may indeed re-
sult from a parallel evolution of corresponding eco-morphological
traits in S. diagramma and the S. babaulti/S. marginatus/S. pleurospi-
lus lineage. In contrast, the well-supported monophyly of the genus
Tropheus supports a single origin of their trophic morphology, and
rejects the paraphyletic relationship inferred by Sturmbauer et al.
(2003).
4.3. Taxonomy and intra-generic diversity
Some disagreement between the phylogenetic reconstructions
and the systematic classifications (Poll, 1986; Yamaoka, 1987;
Yamaoka et al., 1998) becomes evident from the poly- and para-
phyly of the two polytypic genera Simochromis and Petrochromis
in the mitochondrial and nuclear trees. Generic classifications
were primarily based on trophic morphology such as dentition,
structure of the pharyngeal apparatus and gut length, but several
more recent studies suggested that both homoplasy and ontoge-
netic plasticity compromise the phylogenetic information of
these eco-morphological traits in cichlid fishes (Meyer, 1987;
Rüber et al., 1999; Koblmüller et al., 2007b; Takahashi et al.,
2007). In this light, the AFLP phylogeny in fact corresponds
astoundingly well to the traditional classification, although it
questions the classification of S. diagramma, which does not
group with the other Simochromis species. Furthermore, although
the mitochondrial phylogeny shows a deep split (?0.8 MY) be-
tween S. pleurospilus and S. babaulti, reciprocal monophyly of
the two species is not corroborated by our AFLP data. The two
species are morphologically indistinguishable, the only difference
being rows of red dots along the sides of S. pleurospilus males,
which are not present in S. babaulti. Simochromis pleurospilus oc-
curs allopatrically from S. babaulti in a small distribution range
in the southwestern part of the lake, and may represent a geo-
graphic color morph of S. babaulti (Konings, 1998) rather than
a separate species. More data are needed to investigate the spe-
cies status of S. pleurospilus.
An alpha-taxonomic revision of the genus Petrochromis, includ-
ing extensive phylogenetic and phylogeographic work, is urgently
needed not so much because of the paraphyly with respect to Inte-
rochromis, but more importantly because the present study sug-
gests that the species diversity of Petrochromis is currently
drastically underestimated (see also Nishida, 1997). Our data indi-
cate the existence of several undescribed P. polyodon-like species
and a highly distinct Petrochromis lineage from the northwestern
part of Zambia (P. sp. ‘‘Katete”). Moreover, a plethora of color mor-
phs with and without suggested species assignment have been de-
scribed in the aquarist literature (e.g. Herrmann, 1985; Konings,
1998, 2005), only some of which are included in our study. For
example, Petrochromis sp. ‘‘Moshi yellow” has been proposed to
represent a distinct species (Konings, 1998) closely related to P.
ephippium. While our data confirm a close relationship of P. sp.
‘‘Moshi yellow” to P. ephippium, more work is required to assess
its proposed species status.
It is conspicuous that the two most strictly rock-dwelling
genera, Tropheus and Petrochromis, are the most species-rich
and also comprise numerous geographic color morphs, while
much less diversity evolved in the less specialized genera Cte-
nochromis, Gnathochromis, Limnotilapia, Pseudosimochromis and
Simochromis. One possible explanation for this pattern lies in
the correlation between ecological specialization and population
structure in cichlid fishes (Markert et al., 1999; Danley et al.,
2000; Taylor et al., 2001; Peryera et al., 2004; Duftner et al.,
2006; Koblmüller et al., 2007c, 2009; Sefc et al., 2007; Wagner
and McCune, 2009). Gene flow in Tropheus and Petrochromis
has been curbed by numerous habitat barriers along the lake-
shore (Sefc et al., 2007; Egger et al., 2007; Wagner and McCune,
2009), and the resulting high level of population structure al-
lowed for the evolution of phenotypic differentiation and even-
tuallytherise of newspecies
correlation exists in some other Lake Tanganyika cichlids; Duft-
ner et al., 2006; Sefc et al., 2007). In contrast, less stenotopic
species disperse more easily and differentiation is precluded by
continuous gene flow (Meyer et al., 1996; Wagner and McCune,
2009).
(butnotethatnosuch
5. Conclusions
Our study confirmed a period of rapid cladogenesis at the onset
of the diversification of the Tropheini, which produced the numer-
ous specialized rock-dwelling aufwuchs-feeders dominating Lake
Tanganyika’s shallow rocky habitat. A high degree of correspon-
dence between AFLP clustering and trophic types indicated fewer
cases of parallel evolution of trophic ecomorphology than inferred
by Sturmbauer et al. (2003) based on mitochondrial data alone. The
study underlines the usefulness of combining different types of
molecular markers for phylogenetic reconstructions of rapidly
radiating lineages. Our choice of mitochondrial marker genes in-
creased the resolution of the phylogenetic relationship among Tro-
pheini lineages, and nuclear multi-locus data helped to overcome
problems with incomplete lineage sorting and identified instances
of introgression and hybridization.
Acknowledgments
We thank E. Verheyen and W. Salzburger for providing DNA
samples, and the team at the Mpulungu Station of the Ministry
of Agriculture, Food and Fisheries, Republic of Zambia for their hos-
pitality and cooperation during fieldwork. This study was sup-
ported by the Austrian Science Foundation (Grants P17380 and
P20883 to K.M.S and grant I33 to C.S.).
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
329

Page 13

Appendix A
List of samples, with sampling locality and coordinates (if known) of tropheine taxa, and GenBank accession numbers for the ND2 gene
and the mitochondrial control region (CR).
SpeciesSample IDSampling locality CoordinatesGenBank Accession No.
ND2CR
Tropheini
Ctenochromis horei
6621
6636
6637
6614
6615
6618
6650
6653
6654
6658
8037
8258
6645
7190
7191
6619
6625
6628
6655
6656
7063
6613
6620
6627
7376
8241
8252
6626
6638
6657
7058
7061
7062
8031
8032
8033
8035
8248
61
149
8239
8242
8247
6612
6622
6629
6630
6631
6632
6633
7374
7375
8236
8238
8243
8249
8261
6623
8257
8259
8260
7377
Katete
Mbita Island W
Mbita Island W
Kalambo Lodge
Kalambo Lodge
Chimba
Kalambo Lodge
Wonzye
Wonzye
Wonzye
Wonzye
Mtosi
Kalambo Lodge
Mpulungu
Mpulungu
Chimba
Chimba
Mbita Island E
Wonzye
Isanga
Kalambo Lodge
Kalambo Lodge
Chimba
Chimba
Mpimbwe
S of Isonga
Mahale N
Chimba
Mbita Island W
Wonzye
Kalambo
Wonzye
Kalambo Lodge
Kalambo Lodge
Kalambo Lodge
Kalambo Lodge
Kalambo Lodge
Mahale N
Kavalla
Ikola
Mpimbwe
S of Isonga
N of Mabilibili
Kalambo Lodge
Katete
Nakaku
Nakaku
Nakaku
Nakaku
Nakaku
Nkondwe Island
N of Mpimbwe
Mpimbwe
Mpimbwe
S of Isonga
Mahale N
Mtosi
Katete
Mtosi
Mtosi
Mtosi
S of Karema
S 8?20’, E 30?30’
S 8?45’, E 31?05’
S 8?45’, E 31?05’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?19’, E 30?32’
S 8?37’, E 31?12’
S 8?43’, E 31?08’
S 8?43’, E 31?08’
S 8?43’, E 31?08’
S 8?43’, E 31?08’
S 7?35’, E 30?38’
S 8?37’, E 31?12’
S 8?46’, E 31?06’
S 8?46’, E 31?06’
S 8?19’, E 30?32’
S 8?19’, E 30?32’
S 8?45’, E 31?06’
S 8?43’, E 31?08’
S 8?39’, E 31?11’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?19’, E 30?32’
S 8?19’, E 30?32’
S 7?08’, E 30?31’
S 6?28’, E 30?09’
S 6?02’, E 29?44’
S 8?19’, E 30?32’
S 8?45’, E 31?05’
S 8?43’, E 31?08’
S 8?36’, E 31?11’
S 8?43’, E 31?08’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 6?02’, E 29?44’
S 5?38’, E 29?24’
S 6?41’, E 30?22’
S 7?08’, E 30?31’
S 6?28’, E 30?09’
S 6?26’, E 30?53’
S 8?37’, E 31?12’
S 8?20’, E 30?30’
S 8?40’, E 30?54’
S 8?40’, E 30?54’
S 8?40’, E 30?54’
S 8?40’, E 30?54’
S 8?40’, E 30?54’
S 7?23’, E 30?33’
S 7?07’, E 30?30’
S 7?08’, E 30?31’
S 7?08’, E 30?31’
S 6?28’, E 30?09’
S 6?02’, E 29?44’
S 7?35’, E 30?38’
S 08?20’, E 30?30’
S 7?35’, E 30?38’
S 7?35’, E 30?38’
S 7?35’, E 30?38’
S 6?52’, E 30?32’
GQ995716
GQ995717
GQ995718
GQ995719
GQ995720
GQ995721
GQ995742
GQ995743
GQ995744
GQ995745
GQ995746
GQ995747
GQ995722
GQ995723
GQ995724
GQ995725
GQ995726
GQ995727
GQ995728
GQ995735
GQ995736
GQ995729
GQ995730
GQ995731
GQ995732
GQ995733
GQ995734
GQ995799
GQ995793
GQ995800
GQ995801
GQ995794
GQ995795
GQ995796
GQ995797
GQ995798
GQ995802
GQ995770
GQ995737
GQ995738
GQ995739
GQ995740
GQ995741
GQ995749
GQ995750
GQ995751
GQ995752
GQ995753
GQ995754
GQ995755
GQ995756
GQ995757
GQ995774
GQ995775
GQ995767
GQ995768
GQ995769
GQ995748
GQ995771
GQ995772
GQ995773
GQ995758
GQ995825
GQ995826
GQ995827
GQ995828
GQ995829
GQ995830
GQ995851
GQ995852
GQ995853
GQ995854
GQ995855
GQ995856
GQ995831
GQ995832
GQ995833
GQ995834
GQ995835
GQ995836
GQ995837
GQ995844
GQ995845
GQ995838
GQ995839
GQ995840
GQ995841
GQ995842
GQ995843
GQ995908
GQ995902
GQ995909
GQ995910
GQ995903
GQ995904
GQ995905
GQ995906
GQ995907
GQ995911
GQ995879
GQ995846
GQ995847
GQ995848
GQ995849
GQ995850
GQ995858
GQ995859
GQ995860
GQ995861
GQ995862
GQ995863
GQ995864
GQ995865
GQ995866
GQ995883
GQ995884
GQ995876
GQ995877
GQ995878
GQ995857
GQ995880
GQ995881
GQ995882
GQ995867
Gnathochromis pfefferi
Interochromis loockii
Limnotilapia dardennii
Lobochilotes labiatus
Petrochromis ephippium
Petrochromis famula
Petrochromis fasciolatus
Petrochromis cf. macrognathus
Petrochromis orthognathus
Petrochromis polyodon
Petrochromis cf. polyodon
Petrochromis sp. ‘‘Katete”
Petrochromis sp. ‘‘macrognathus rainbow”
Petrochromis sp. ‘‘Moshi yellow”
330
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334

Page 14

Appendix A (continued)
SpeciesSample IDSampling locality CoordinatesGenBank Accession No.
ND2 CR
8235
8245
8246
8250
8234
69a
143a
6617
6624
6646
7059
7060
8240
8255
8256
6649
8038
8244
64
8253
8254
6611
6616
6639
8039
Tr10
76
85a
91a
101
3441
4518
4519
4300
5447
5448
4520a
Tr8
4370a
4374a
4376a
4385
Mpimbwe
N of Mabilibili
N of Mabilibili
Mahale N
Longola
Location unknown
Zaire
Chimba
Katete
Kalambo Lodge
Kasenga Rocks
Location unknown
Kekese
Mahale N
Mahale N
Kalambo Lodge
Wonzye
S of Isonga
Ubwari
Mahale N
Mahale N
Kalambo Lodge
Kalambo Lodge
Mbita Island W
Mbita Island W
Bulombora
Bemba
Maswa
Maswa
Bemba
Location unknown
Halembe
Halembe
Mbita Island W
Kekese
Kekese
Mabilibili
Kapampa
Kachese
Mpimbwe
Bemba
Nakaku
S 7?08’, E 30?31’
S 6?26’, E 30?53’
S 6?26’, E 30?53’
S 6?02’, E 29?44’
unknown
unknown
unknown
S 8?19’, E 30?32’
S 8?20’, E 30?30’
S 8?37’, E 31?12’
S 8?43’, E 31?09’
unknown
S 6?37’, E 30?17’
S 6?02’, E 29?44’
S 6?02’, E 29?44’
S 8?37’, E 31?12’
S 8?43’, E 31?08’
S 6?28’, E 30?09’
S 4?11’, E 29?14’
S 6?02’, E 29?44’
S 6?02’, E 29?44’
S 8?37’, E 31?12’
S 8?37’, E 31?12’
S 8?45’, E 31?05’
S 8?45’, E 31?05’
S 5?40’, E 29?46’
S 3?37’, E 20?09’
unknown
unknown
S 3?37’, E 20?09’
unknown
S 5?44’, E 29?55’
S 5?44’, E 29?55’
S 8?45’, E 31?05’
S 6?37’, E 30?17’
S 6?37’, E 30?17’
S 6?28’, E 29?56’
S 7?30’, E 30?12’
S 8?29’, E 30?28’
S 7?08’, E 30?31’
S 3?37’, E 20?09’
S 8?40’, E 30?54’
GQ995764
GQ995762
GQ995763
GQ995765
GQ995766
GQ995759
GQ995760
GQ995761
GQ995776
GQ995777
GQ995778
GQ995779
GQ995784
GQ995786
GQ995785
GQ995790
GQ995791
GQ995792
GQ995789
GQ995787
GQ995788
GQ995780
GQ995781
GQ995782
GQ995783
GQ995817
GQ995804
GQ995806
GQ995807
GQ995805
GQ995803
GQ995808
GQ995809
GQ995810
GQ995818
GQ995819
GQ995815
GQ995816
GQ995811
GQ995812
GQ995813
GQ995814
GQ995873
GQ995871
GQ995872
GQ995874
GQ995875
GQ995868
GQ995869
GQ995870
GQ995885
GQ995886
GQ995887
GQ995888
GQ995893
GQ995895
GQ995894
GQ995899
GQ995900
GQ995901
GQ995898
GQ995896
GQ995897
GQ995889
GQ995890
GQ995891
GQ995892
GQ995926
GQ995913
GQ995915
GQ995916
GQ995914
GQ995912
GQ995917
GQ995918
GQ995919
GQ995927
GQ995928
GQ995924
GQ995925
GQ995920
GQ995921
GQ995922
GQ995923
Petrochromis sp. ‘‘Texas Longola”
Petrochromis trewavasae
Pseudosimochromis curvifrons
Simochromis babaulti
Simochromis diagramma
Simochromis marginatus
Simochromis pleurospilus
Tropheus brichardi
Tropheus duboisi
Tropheus moorii
Tropheus polli
Tropheus sp. ‘‘Kirschfleck”
Tropheus sp.
Outgroup taxa
Astatotilapia burtoni
7054
7055
7056
6662a
6661a
GQ995713
GQ995714
GQ995715
GQ995712
GQ995711
AY930058
AY930105
AY930102
AY300074
AY930064
AY930077
AY930097
AY930083
AY930046
AY930063
AY930076
AF305277
AY930090
AY930092
AF305282
AY930089
AF305270
GQ995822
GQ995823
GQ995824
GQ995821
GQ995820
Aulonocara sp.
Copadichromis borleyi
Haplochromis bloyeti
Haplochromis sp.1
Haplochromis sp.2
Haplochromis sp. ‘‘dwarf big eye”
Haplochromis degeni
Haplochromis insidiae
Haplochromis obliquidens
Haplochromis squamipinnis
Haplochromis stappersii
Ptyochromis sauvagei
Xystichromis phytophagus
Astatotilapia calliptera
Cheilochromis euchilus
Copadichromis virginalis
Cyrtocara moorii
Diplotaxodon greenwoodi
(continued on next page)
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334
331

Page 15

References
Avise, J.C., 2004. Molecular Markers, Natural History, and Evolution, second ed.
Sinauer Associates, Sunderland, Massachusetts.
Baric, S., Salzburger, W., Sturmbauer, C., 2003. Phylogeography and evolution of the
Tanganyikan cichlid genus Tropheus based upon mitochondrial DNA sequences.
J. Mol. Evol. 56, 54–68.
Brandstätter, A., Salzburger, W., Sturmbauer, C., 2005. Mitochondrial phylogeny of
the Cyprichromini, a lineage of open-water cichlid fishes endemic to Lake
Tanganyika, East Africa. Mol. Phylogenet. Evol. 34, 382–391.
Clabaut, C., Salzburger, W., Meyer, A., 2005. Comparative phylogenetic analyses of
the adaptive radiation in Lake Tanganyika cichlid fishes: nuclear sequences are
less homoplasious but also less informative than mitochondrial DNA. J. Mol.
Evol. 31, 666–681.
Cohen, A.S., Lezzar, K.E., Tiercelin, J.J., Soreghan, M., 1997. New palaeographic and
lake-level reconstructions of Lake Tanganyika: implications for tectonic,
climatic and biological evolution in a rift lake. Basin Res. 9, 107–132.
Danley, P.D., Markert, J.A., Arnegard, M.E., Kocher, T.D., 2000. Divergence with
gene flow in the rock-dwelling cichlids of Lake Malawi. Evolution 54, 1725–
1737.
Day, J.J., Santini, S., Garcia-Moreno, J., 2007. Phylogenetic relationships of the Lake
Tanganyika cichlid tribe Lamprologini: the story from mitochondrial DNA. Mol.
Phylogenet. Evol. 45, 629–642.
Delvaux, D., 1995. Age of Lake Malawi (Nyasa) and water level fluctuations. Mus. R.
Afr. Centr. Tervuren (Belg.) Dept. Geol. Min. Rapp. Ann. 1995–1996, 99–108.
Drummond, A.J., Ho, S.Y.W., Phillips, M.J., Rambaut, A., 2006. Relaxed phylogenetics
and dating with confidence. PLoS Biol. 4, e88.
Drummond, A.J., Ho, S.Y.W., Rawlence, N., Rambaut, A., 2007. A rough guide to
BEAST 1.4. Available from: http://beast.bio.ed.ac.uk/.
Drummond, A.J., Rambaut, A., 2007. BEAST: Bayesian evolutionary analysis by
sampling trees. BMC Evol. Biol. 7, 214.
Duftner, N., Koblmüller, S., Sturmbauer, C., 2005. Evolutionary relationships of the
Limnochromini, a tribe of benthic deep water cichlid fishes endemic to Lake
Tanganyika, East Africa. J. Mol. Evol. 60, 277–289.
Duftner, N., Sefc, K.M., Koblmüller, S., Nevado, B., Verheyen, E., Phiri, H., Sturmbauer,
C., 2006. Distinct population structure in a phenotypically homogeneous rock-
dwelling cichlid fish from Lake Tanganyika. Mol. Ecol. 15, 2381–2396.
Egger, B., Obermüller, B., Sturmbauer, C., Phiri, H., Sefc, K.M., 2006. Monogamy in
the maternally mouthbrooding Lake Tanganyika cichlid fish Tropheus moorii.
Proc. R. Soc. Lond. B 273, 1797–1802.
Egger, B., Koblmüller, S., Sturmbauer, C., Sefc, K.M., 2007. Nuclear and mitochondrial
data reveal different evolutionary processes in the Lake Tanganyika cichlid
genus Tropheus. BMC Evol. Biol. 7, 137.
Fryer, G., Iles, T.D., 1972. The Cichlid Fishes of the Great Lakes of Africa. Oliver &
Boyd, Edinburgh, 642 pp.
Genner, M.J., Seehausen, O., Lunt, D.H., Joyce, D.A., Shaw, P.W., Carvalho, G.R.,
Turner, G.F., 2007. Age of cichlids: new dates for ancient lake fish radiations.
Mol. Biol. Evol. 24, 1269–1282.
Grant, P.R., 1981. Speciation and adaptive radiation of Darwin’s finches. Am. Sci. 69,
653–663.
Greenwood, P.H., Kullander, S.O., 1994. A taxonomic revision and redescription of
Tilapia polyacanthus and T. stormsi (Teleostei: Cichlidae), with descriptions of
two new Schwetzochromis species from the Upper Zaire River drainage. Ichthyol.
Explor. Freshwater 5, 161–180.
Hedtke, S.M., Townsend, T.M., Hillis, D.M., 2006. Resolution of phylogenetic conflict
in large data sets by increased taxon sampling. Syst. Biol. 55, 522–529.
Herrmann, H.-J., 1985. Alte und neue Petrochromis—und kein Ende in Sicht. DATZ
38, 441–444.
Hori, M., 1991. Feeding relationships among cichlid fishes in Lake Tanganyika:
Effects of intra- and interspecific variations of feeding behavior on their
coexistence. Ecol. Int. Bull. 29, 89–101.
Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference of phylogeny.
Bioinformatics 17, 754–755.
Jian, S.G., Soltis, P.S., Gitzendanner, M.A., Moore, M.J., Li, R., Hendry, T.A., Qiu, Y.L.,
Dhingra, A., Beli, C.D., Soltis, D.E., 2008. Resolving an ancient, rapid radiation in
Saxifragales. Syst. Biol. 57, 38–57.
Kassam, D., Seki, S., Hori, M., Yamaoka, K., 2006. Nuclear markers reveal that inter-
lake cichlids’ similar morphologies do not reflect similar genealogy. Mol.
Phylogenet. Evol. 40, 383–388.
Appendix A (continued)
SpeciesSample ID Sampling localityCoordinatesGenBank Accession No.
ND2 CR
Diplotaxodon limnothrissa
Lethrinops auritus
Lethrinops furcifer
Melanochromis auratus
Pseudotropheus livingstoni
Rhamphochromis esox
Rhamphochromis macrophthalmus
Astatoreochromis alluaudi
Orthochromis stormsib
Pseudocrenilabrus multicolor
Pseudocrenilabrus philander
Serranochromis stappersii
Perissodus microlepis
Plecodus straeleni
Ectodus descampsii
Xenotilapia sima
Orthochromis mazimeroensis
Orthochromis uvinzae
Cyprichromis leptosoma
Paracyprichromis brieni
Limnochromis auritus
Triglachromis otostigma
Cyphotilapia frontosa
Eretmodus cyanostictus
Spathodus erythrodon
Tanganicodus irsacae
Julidochromis marlieri
Lamprologus callipterus
Lamprologus mocquardi
Bathybates leo
Hemibates stenosoma
AF305261
U07252
AF305316
AY930069
AY930061
AF305252
AF305250
AY930071
AY930057
AY930103
AY930047
EF393698
AF398222
AF398221
AY337790
AY337785
AY930053
AY930048
AY740337
AF398223
AY337766
AY337769
U07247
AF398220
AF398218
AF398219
AF398230
AF398226
AF398225
AY663731
AY663719
Sequences not generated in the framework of this study were obtained from GenBank from following publications: Kocher et al. (1995), Shaw et al. (2000), Salzburger et al.
(2002a,b, 2005), Koblmüller et al. (2004, 2005), Brandstätter et al. (2005), Duftner et al. (2005), Katongo et al. (2007).
aSamples obtained through ornamental fish trade.
bThere is some taxonomic uncertainty about this sample. For detailed information see Greenwood and Kullander (1994) and Koblmüller et al. (2008b).
332
S. Koblmüller et al./Molecular Phylogenetics and Evolution 55 (2010) 318–334