Origin and Speciation of Haplochromine Fishes in East African Crater
Lakes Investigated by the Analysis of Their mtDNA, Mhc Genes, and SINEs
Akie Sato, Naoko Takezaki, Herbert Tichy, Felipe Figueroa, Werner E. Mayer, and Jan Klein
Max-Planck-Institut fu ¨r Biologie, Abteilung Immungenetik, Tu ¨bingen, Germany
The Western Branch of the East African Great Rift Valley is pocketed with craters of extinct or dormant volcanoes.
Many of the craters are filled with water, and the lakes are inhabited by fishes. The objective of the present study was to
determine the amount and nature of genetic variation in haplochromine fishes inhabiting two of these crater lakes, Lake
Lutoto and Lake Nshere, and to use this information to infer the origin and history of the two populations. To this end,
sequences of mitochondrial (mt) DNA control region, exon 2 of major histocompatibility complex (Mhc) class II B
genes, and short interspersed elements (SINEs) were analyzed. The results indicate that the Lake Nshere and Lake Lutoto
fishes originated from different but related large founding populations derived from the Kazinga Channel, which
connects Lake Edward and Lake George. Some of the genetic polymorphism that existed in the ancestral populations was
lost in the populations of the two lakes. The polymorphism that has been retained has persisted for some 50,000
generations (years). During this time, new mutations arose and became fixed in each of the two populations in the
mtDNA, giving rise to sets of diagnostic substitutions. Each population evolved in isolation after the colonization of the
lakes less than 50,000 years ago. There appears to be no population structure within the crater lake fishes, and their
present effective population sizes are in the order of 104to 105individuals. Comparisons with the endemic
haplochromine species of Lake Victoria reveal interesting parallels, as well as differences, which may help to understand
the nature of the speciation process.
The formation of the East African Great Rift Valley,
a complex system of faults and escarpments, was
accompanied by extensive volcanic activity that left
behind many craters along both the eastern and western
branches of the Rift (Schlu ¨ter 1997). As the volcanoes
became extinct or dormant, many of the craters filled with
water, and some of the crater lakes came to be inhabited by
fish. The circumstances under which the fish reached the
crater lakes have been a subject of speculation for many
years. Where the lakes established a temporary or
permanent connection with nearby river systems, the
origin seems obvious. For some of the crater lakes,
however, past or present contacts with rivers could not be
documented, and more fanciful explanations, such as
transfer of fish fry or fertilized eggs by birds or water
mammals, have been invoked.
Genetic characterization of a fish population can
provide information about the probable size and compo-
sition of the founding population, the time of the founding
event, and other related questions. This information can
then be used to make inferences not only about the
circumstances of the colonization but also about the
process of speciation and species divergence. Here we use
three genetic systems to probe the origin, population
dynamics, and speciation of lacustrine fishes—mitochon-
drial (mt) DNA, genes of the major histocompatibility
complex (Mhc), and short interspersed repetitive elements
(SINEs). The mtDNA control region with its high
substitution rate and maternal transmission is the tool of
choice in efforts to identify the most recent common
ancestor (MRCA) of a population (Avise et al. 1987). Our
survey of mtDNA control region sequences in lacustrine
and fluviatile haplochromine cichlids of the Lake Victoria
basin revealed the existence of seven major haplogroups,
designated I through VII (Nagl et al. 2000). Each
haplogroup is characterized by a set of diagnostic sub-
stitutions and by the clustering of its members in a single
clade on phylogenetic trees. The haplogroups are distrib-
uted differentially over the basin, the most widespread
being haplogroup V centered on Lake Victoria, the Lake
Edward region (including Lake George and Lake Albert),
and the rivers around these lakes. A number of additional
haplogroups apparently exist among the fishes of Lake
Tanganyika and Lake Malawi, which were involved in the
survey only tangentially. Haplogroup V is subdivided
further into subgroups VA through VD, which are
distinguished by a more restrictive set of diagnostic
substitutions. Each of the subgroups, however, may be
subdivisible even further on typing additional specimens.
In particular, subgroups VB and VC show a heterogeneity
indicative of future splitting.
Mhc genes with their high polymorphism and long
persistence of allelic lineages are well suited for exploring
the population dynamics in the history of a species (Klein
et al. 1998). Fish Mhc genes, like those of other jawed
vertebrates, fall into two classes, I and II, and each class
into two subclasses, A and B, encoding the a and b chains
of the ab heterodimeric protein (Klein et al. 1997; Shand
and Dixon 2001). Since most of the work on cichlid Mhc
genes was done on the class II B genes (Klein et al. 1993;
Ono et al. 1993a, 1993b; Sato et al. 1997; Ma ´laga-Trillo
et al. 1998; Figueroa et al. 2000), we focused on the most
variable part of these genes, exon 2 encoding the b1
domain of the class II b chain.
Fish SINEs are retroposons that are several hundred
base pairs long and contain a region homologous to
Key words: cichlids, haplochromine fishes, Mhc, mtDNA control
region, SINE, crater lakes.
The nucleotide sequences reported in this publication have been
submitted to the GenBank database and assigned the following accession
numbers: AY164681 to AY164696 (mitochondrial control region),
AY211028 to AY211042 (Mhc class II B), AY211024 to AY211027
Mol. Biol. Evol. 20(9):1448–1462. 2003
Molecular Biology and Evolution, Vol. 20, No. 9,
? Society for Molecular Biology and Evolution 2003; all rights reserved.
by guest on June 1, 2013
a transfer RNA gene in their 59 part and frequently also
contain a region corresponding to a segment of a long
interspersed repetitive element (LINE) in their 39 part
(Okada 1991; Schmid and Maraia 1992; Okada et al.
1997). In a recent study, Terai and colleagues (2003) have
shown that haplochromines of the Lake Victoria basin
contain two types of SINEs, fixed and polymorphic. The
former are found in all individuals of a given species or
species assemblage, whereas the latter are present in only
some individuals of a population. As there is no indication
for selection influencing the evolution of SINEs, these loci
provide a nuclear counterpart to the neutrally evolving
control region segment of the mtDNA on the one hand and
the selection driven evolution of the Mhc genes on the
other hand. Furthermore, the presumed random distribu-
tion of SINEs among the chromosomes enables sampling
of different segments of the nuclear genome.
The two main aims of the present study were (1) to
determine the origin, structure, and effective size of the
haplochromine populations in Lake Lutoto and Lake
Nshere and to ascertain the genetic relationship between
them and (2) to provide a basis for comparing the genetic
variability existing in these two isolated populations with
that present in the endemic Lake Victoria species and thus
pave a way toward a better understanding of the speciation
process in these fishes.
Materials and Methods
Lake Nshere and Lake Lutoto are located in the
Bunyaruguru county in southwest Uganda. Judging from
their location and shape, and from the geology of the
region (Boven et al. 1998), both lakes originated by
flooding of craters of extinct volcanoes. They are at
a distance of approximately 20 km from each other (fig. 1).
They are small (surface areas of , 1 km2) and are
surrounded by even smaller lakes (fig. 1), the waters of at
least some of which have a high salt concentration. Lake
Nshere (or Chibwera) is located at an altitude of 970 m in
the middle of a steppe-like flatland, the Chambura Game
Reserve administered by the Queen Elizabeth National
Parks. It is connected via a small river, which dries out
periodically, with the Kazinga Channel and so with Lake
George and Lake Edward (fig. 2). The distance from Lake
Nshere to the Kazinga Channel is approximately 15 km.
Lake Lutoto (or Rutoto, also known as Lake Nkugute) is
located at an altitude of approximately 1,250 m in
a mountainous area (the height of nearby Mt. Kasunju is
1,950 m [fig. 1]). It is connected by a small, rapidly
flowing brook to Lake Mugogo at a distance of
approximately 3.5 km, the difference in altitude between
the two lakes being approximately 180 m. Lake Mugogo
drains into the Chambura River system and so into the
FIG. 1.—Satellite photograph of the Lake Edward region, East
Africa, in which Lake Lutoto and Lake Nshere are located. The Kazinga
Channel connects Lake Edward and Lake George. (Source: Section
of the JPL / NASA SIR C 11 / X-SAR picture Nyiragongo, Zaire,
DT058_61Segment7 taken on Nov. 18, 1998.)
FIG. 2.—The localities in the Lake Victoria and the neighboring
region at which samples of haplochromines were collected.
Origin and Speciation of East African Crater Lake Fishes1449
by guest on June 1, 2013
Kazinga Channel at a distance of approximately 40 km.
Although tilapias (Oreochromis spp.) were apparently
introduced into Lake Lutoto some time ago, the studied
species seem to be the only haplochromines in the lake.
As far as we could determine, the species of
haplochromines living in Lake Nshere and Lake Lutoto
have not been described. For reasons that will become
apparent later, we regard them as two different species,
distinct from those inhabiting other East African lakes and
rivers. We refer to them tentatively as Haplochromis
‘‘Nshere’’ and Haplochromis ‘‘Lutoto.’’ The adults of both
species have an average standard body length of 55 mm
and hence belong to the category of small haplochromines.
Males of H. ‘‘Nshere’’ are dark black with prominent
orange-yellow egg dummies. The profile of their heads is
straight. Females are a pale greenish color. H. ‘‘Lutoto’’
males have dark vertical bars on a dark greenish
background. Their head profile forms a smooth concave
line (fig. 3). The fishes were caught by angling,
approximately 50 individuals in each lake, in less than
30 min using three fishing rods. Apparently, the fishes are
abundant in both lakes.
Polymerase Chain Reaction (PCR), Cloning,
Amplifications were performed in the PTC-200
Programmable Thermal Controller (MJ Research, Biozym,
Hessisch Oldendorf, Germany) or in the GeneAmp PCR
System 9700 (Applied Biosystems, Weiterstadt, Ger-
many). For the amplification of Mhc class II B exon 2,
two Hot Start PCR systems were applied. In one system,
100 ng of genomic DNA were added to a reaction mix-
ture of Mg2þ-free 1 3 PCR buffer, pH 8.5 (Invitrogen,
Karlsruhe, Germany), 0.2 mM of each of the four
deoxynucleoside triphosphates (Amersham Biosciences,
Freiburg, Germany), 1 lM of each of the sense and
antisense primers, 2.5 units of Taq DNA polymerase
(Amersham Biosciences), and 0.4 units of Pfu DNA
polymerase (Stratagene, Amsterdam, The Netherlands).
The reaction mixture was overlaid with 2.5 mM Mg2þ
HotWax Beads (Invitrogen) to initiate the reaction at high
temperature. The amplification consisted of DNA de-
naturation for 1 min at 948C, annealing for 15 sec at the
appropriate annealing temperature, and 7 min extension at
728C, followed by 34 cycles of 15 sec denaturation at
948C, 15 sec annealing at the required temperature, and 2
min extension at 728C. The reaction was completed by
a final primer extension for 7 min at 728C. Alternatively,
100 ng of genomic DNA were added to a reaction mixture
of 1 3 PCR buffer (Qiagen, Hilden, Germany), 2.5 mM
MgCl2, 0.2 mM of each of the four deoxynucleoside
triphosphates (Amersham Biosciences), 1 lM of each of
the sense and antisense primers, 2.5 units of HotStar Taq
DNA polymerase (Qiagen), and 0.4 units of Pfu DNA
polymerase (Stratagene). To activate the enzyme, the
amplification was initiated by a DNA denaturation step of
15 min at 958C. The PCR was carried out with HotStar
Taq DNA polymerase under the conditions given above.
PCR products were purified by agarose gel electrophoresis
and extraction using the QIA EX II Gel Extraction Kit
(Qiagen) and cloned into the pUC18 vector with the help
of the Sure Clone Ligation Kit (Amersham Biosciences).
The primers used in the various PCRs are listed in table 1.
The DNA was sequenced with the help of the Dyenamic
Direct Cycle Sequencing Kit with 7-deaza dGTP (Amer-
sham Biosciences). Sequencing reactions were processed
by the LI-COR Long ReadIR 4200 DNA sequencer
(MWG Biotech, Ebersberg, Germany).
FIG. 3.—The phenotypes of Haplochromis ‘‘Lutoto’’ and H.
PCR Primers Used
Mhc class II B
NOTE.—CR indicates control region; F indicates forward; R indicates reverse.
1450Sato et al.
by guest on June 1, 2013
Southern DNA Blotting and Hybridization
Five micrograms of genomic DNA were digested
with Eco RI restriction endonuclease for 18 h under the
conditions recommended by the supplier (Roche Diag-
nostics, Mannheim, Germany), and fragments were
separated by agarose gel electrophoresis and blotted onto
Hybond-Nþ nylon filters (Amersham Biosciences). Pre-
hybridization, hybridization, and probe labeling were
carried out using the AlkPhos Direct Hybridization and
Detection Kit (Amersham Biosciences). One hundred
nanograms of DNA were used to label the probe. After
the overnight hybridization, the filters were washed
according to the AlkPhos Direct protocol. After the
application of the chemiluminiscent detection reagent
CDP-Star of the kit, Hyperfilm ECL (Amersham Bio-
sciences) was exposed to the blot for 6 h and developed.
The multiple alignments of sequence data were made
by using ClustalW 1.82 (Thompson, Higgins, and Gibson
1994) and checked by eye. The phylogenetic trees were
drawn by the neighbor-joining (NJ) method (Saitou and
Nei 1987) and evaluated by 1,000 bootstrap replications.
In the analyses of mtDNA control region sequences,
the sites 518 to 535 of the alignment in Nagl et al. (2000)
were eliminated because they contain many single
nucleotide repeats and the sequences are not reliable. Sites
with indels were also deleted. NJ trees were drawn using p-
distances. Unless the substitution rate varies extensively in
different lineages, p-distance is better in obtaining a correct
tree topology than distances based on a complex sub-
stitution model, even when the substitution pattern of the
nucleotide changes follows a complex model, particularly
for closely related sequences (Nei and Kumar 2000).
The alignment of group V sequences (fig. 4 [only variable
sites are shown]) includes sequences from Nagl et al.
(2000). The NJ tree for the group V sequences (fig. 5)
was drawn based on the alignment in figure 4 with 810
In the mtDNA analyses, estimations of the mutation
rate and population size, as well as Tajima’s test for
neutrality, were based on an alignment containing
sequences from the neighboring riverine and Lake Malawi
fishes and a few Lake Tanganyika species as an outgroup
with 804 shared sites (data not shown). The major
branching patterns among the group V sequences of the
NJ tree (not shown) based on the latter alignment were
identical to those shown in figure 5. The GenBank
accession numbers of the sequences that are not included
in the group V NJ tree are as follows. Lake Tanganyika:
Petrochromis orthognathus (U12549), Lobochilotes labi-
atus (U12550), Tropheus duboisi (AF400737); Lake
Malawi: Protomelas fenestratus (AF213625), Champso-
chromis spilorhynchus (U12553), Mylochromis lateris-
Lethrinops auritus (U12551), Stigmatochromis woodi
(AF213626), Labeotropheus trewavasae (AF213623),
Pseudotropheus sp. ‘‘msobo’’ (AF213622), Melanochro-
mis melanopterus (AF213619), Melanochromis parallelus
(AF213621), Petrotilapia sp. (U12547), Metriaclima
callainos (AF213620), Rhamphochromis sp. (U12548);
Haplochromines of the rivers and lakes in the Lake
Victoria region: AF213559 to AF213609.
Since in the mtDNA control region the transition-
transversion ratio is relatively high and the substitution
rate varies extensively across sites (e.g., Tamura and Nei
1993), Kimura’s two-parameter distance with gamma
correction was used for time estimation. The gamma
parameter (a ¼ 0.12) was estimated by the maximum-
likelihood method using the tree topology of the NJ tree
and the baseml in PAML3.12 program (Yang 2002). In the
maximum-likelihood analysis, the transition-transversion
rate ratio was estimated as 7.1. Observed base frequencies
were 0.31, 0.31, 0.16, and 0.23 for A, T, G, and C,
The average distance (h) between the two descendant
clusters of interior nodes of the mtDNA control region NJ
tree was computed as
where dijis the distance between sequences i and j, A and
B are the two descendant clusters of the node, and n and m
are the numbers of the sequences in the clusters A and B,
The average distance between the mtDNA control
region sequences of Lake Malawi and Lake Victoria fishes
was 0.0445 6 0.0073. Assuming that the separation of
these two groups occurred 1 to 2 MYA (see also
Sturmbauer et al. 2001), the mutation rate for the mtDNA
control region was estimated as 2.2 ? 4.5 3 10?8per site
per year. This rate is similar to the mutation rate (1.8 3
10?8) estimated for other fish (snook [Donaldson and
Wilson 1999]), but much lower than the rate for human
mtDNA control region (0.75 3 10?7[Tamura and Nei
An NJ tree of the cichlid Mhc class II B exon 2
sequences was drawn using all the available cichlid and
closely related sequences in the GenBank database. In the
NJ tree shown in figure 5, only representative sequences
were included because of space limitations, and medaka
fish (Oryzias latipes) sequences were used as an outgroup.
In the alignment, there were 160 shared nucleotide sites
after sites with missing nucleotides or indels were
excluded. Jukes-Cantor distances were used.
Detailed information regarding the fish samples and
the sequence data obtained for mtDNA control region,
Mhc class II B exon 2, and SINE 1357 are available as
Supplementary Material online.
Computer Simulation of Coalescence Process
The coalescence theory (Hudson 1983; Tajima 1983)
and the polymorphism found at the SINE loci were used to
obtain the lower limit of the effective population size.
According to the theory, allelic lineages that exist in
a current population have a genealogy that can be traced
back to an MRCA. For a neutral nuclear locus of
a population with an effective size N, the time (tj) during
which exactly j allelic lineages exist in the population
Origin and Speciation of East African Crater Lake Fishes1451
by guest on June 1, 2013
FIG. 4.—Nucleotide alignment of mtDNA control region, haplogroup V sequences borne by East African haplochromines. Two sequences are
from haplogroups II and IV and are shown for comparison. The majority consensus sequence is given at the top and identity with the consensus is
indicated by dashes (–). Indels introduced to optimize the alignment are indicated by asterisks (*). Only variable sites are shown, and the numbering (to
be read vertically downward) at the top is that used by Nagl et al. (2000 [in the electronic attachment for that article]). The Lake Lutoto and Lake Nshere
sequences are the result of the present study; all others are from Nagl et al. (2000). A short segment (site from 518 to 535) has been omitted from the
alignment and the identification of different haplotypes for the crater lake fishes because repeated stretches of a single nucleotide are difficult to resolve
on the sequencing gels. Individual sequences are identified by a species abbreviation or sample number, followed by the GenBank accession code and
locality number (in parentheses [see fig. 2]). For the Lake Lutoto and Lake Nshere sequences, the sequence types follow the species abbreviation and
the numbers of samples that had the sequence types are shown instead of the GenBank accession code. Species abbreviations are as follows: Asnu,
1452Sato et al.
by guest on June 1, 2013
FIG. 5.—Neighbor-joining tree of mtDNA control region haplogroup V sequences. The tree is based on the alignment in figure 4. The sequence
designations are as in figure 4. Subgroups of group V are enclosed in brackets. Encircled numbers indicate important nodes; other numbers at nodes are
bootstrap values (only those .50% are shown).
Astatotilapia nubila; Asve, A. velifer; Enci, Enterochromis cinctus; Gasi, Gaurochromis simpsoni; Hali, Haplochromis lividus; Halu, H. ‘‘Lutoto’’;
Hans, H. ‘‘Nshere’’; Haro, H. rockkribensis; Havb, H. sp. ‘‘velvet black’’; Lime, Lipochromis melanopterus; Neni, Neochromis nigricans; Pabe,
Paralabidochromis beadlei; Pach, P. chilotes; Papl, P. plagiodon; Prve, Prognathochromis venator; Psri, Psammochromis riponianus; Ptsa,
Ptyochromis sauvagei; Ptxe, P. xenognathus; Ysla, Yssichromis laparagramma. Substitutions diagnostic for the bracketed subgroups are highlighted.
The accession numbers of H. ‘‘Nshere’’ (Hans) and H. ‘‘Lutoto’’ (Halu) sequences are AY164690/AY164695/AY164696 (Hans1), AF213578
(Hans2), AF213580 (Hans3), AF213579/AY164695 (Hans4), AY164687 (Hans5), AY164688 (Hans6), AY164689 (Hans7), AY164691 (Hans8),
AY164692 (Hans9), AY164693 (Hans10), AF213577/AY164682/AY164683 (Halu1), AY164681 (Halu2), AF213575 (Halu3), AF213576 (Halu4),
AY164684 (Halu5), AY164685 (Halu6), and AY164686 (Halu7). Some of the accession numbers are from Nagl et al. (2000). The samples sequenced
in Nagl et al. (2000) were resequenced and confirmed in this study.
Origin and Speciation of East African Crater Lake Fishes1453
by guest on June 1, 2013
corresponds to a random number generated according to
the probability density of the exponential distribution with
a mean of 2/[j(j?1)] in units of 2N generations (Simonsen,
Churchill, and Aquadro 1995). In the computer simulation
starting with n lineages, the coalescence process was
repeated until na? 1 lineages remained in the population,
where n is the number of genes sampled in either the H.
‘‘Nshere’’ or H. ‘‘Lutoto’’ populations, and na is the
number of alleles shared between H. ‘‘Nshere’’ or H.
‘‘Lutoto’’ and other lake populations. The time
required to reach this stage was recorded, the expectation
in units of 2N. By applying various values of N, we
searched for the case in which the probability that ta.
50,000 years (the geological estimate of the age of Lakes
Nshere and Lutoto being , 50,000 years) was smaller than
0.05. For this value of N, the hypothesis that naallelic
lineages persisted for 50,000 years could be rejected at the
5% significance level. This value of N represents the lower
limit for the effective population size of a crater lake
EðtaÞ ¼ 2
Computer Simulation for Founding Population Size and
the Number of Alleles Retained
To examine the relationship between the minimum
number of Mhc alleles retained in the population, the
founding population size, and the population growth rate
after the founding event, another computer simulation was
conducted. The simulation was a slight modification of
that used by Vincek et al. (1997) who assumed over-
dominant selection for Mhc loci with the selection co-
efficient s ¼ 0.01. The alleles for the founding population
of Nbindividuals were drawn at random from the parental
population. After the founding event, population growth
started immediately with rate r and continued until the
number of individuals in the population reached 10,000.
Then, the number of alleles retained in the population was
recorded. This process was repeated 1,000 times for each
combination of the values of Nband r. The ranges of the
Nband r values varied from 5 to 500 and from 0.01 to 0.5,
respectively. The allele frequencies in the ancestral
population were assumed to be those reported for the
HLA-DRB1 locus in Caucasians (Imanishi, Wakisaka, and
Gojobori 1991). In a preliminary study, we also assumed
that the frequencies of alleles in the ancestral population
are equal. The results for these two situations were
essentially the same.
We divide the description of the results into two parts.
In the first part we characterize the mtDNA control region,
SINE, and Mhc class II B loci of H. ‘‘Nshere’’ and H.
‘‘Lutoto.’’ In the second part we use this information to
make inferences regarding these two populations. We
begin with mtDNA.
mtDNA Control Region
Control regions of mtDNA were sequenced from 28
fish caught in Lake Nshere and 20 specimens from Lake
Lutoto. Ten distinct H. ‘‘Nshere’’ and seven H. ‘‘Lutoto’’
haplotypes were found (fig. 4) among the 48 sequences
obtained. The H. ‘‘Nshere’’ sequences showed more
variability (one to seven differences) than those of H.
‘‘Lutoto’’ (one to three differences). Phylogenetic analysis
revealed all 48 sequences to belong to haplogroup V
characterized by the presence of T, C, C, A, and A at sites
27, 87, 96, 348, and 825, respectively (see Nagl et al. 2000
and fig. 4). Each of the two sets of sequences forms
a separate cluster (a clade on a phylogenetic tree [fig. 5])
and each is identified by distinct diagnostic substitutions.
All Lake Nshere sequences have A and T at sites 107 and
830, respectively, whereas all Lake Lutoto sequences share
T, T, and G at sites 100, 347, and 830, respectively (fig. 4).
We denote the control region subgroup of the VB
haplogroup present in H. ‘‘Lutoto’’ as VE and that of H.
‘‘Nshere’’ as VF. The H. ‘‘Lutoto’’ sequences appear to be
most closely related to a particular subset of VB sequences
with which they share an A at site 635 (fig. 4). Similarly,
the H. ‘‘Nshere’’ sequences are most closely related to
another subset of VB sequences that have a G at site 635.
Some of the VF sequences also share an A at site 167 with
some members of this VB subset, but this nucleotide is
also present in some members of the VC subgroup (figs. 4
and 5). These observations lead us to three conclusions.
First, the haplochromine fishes inhabiting Lake Lutoto and
Lake Nshere originated in an area in which the haplogroup
VB is common. Second, the fishes of these two lakes
originated from two distinct but related populations,
probably occupying different parts of the area of wide
VB-haplogroup distribution (see Discussion). Third, the
fish populations of the two lakes have been isolated from
other populations long enough to attain fixation of
diagnostic substitutions. A network presentation of
mtDNA haplotypes of the VB group and the crater lake
fishes is available as Supplementary Material online.
Using the mutation rate of 2.2 ? 4.5 3 10?8per site
per year (see Material and Methods), we estimate the age
of the nodes that separate the crater lake fishes from fishes
of other localities (nodes 3 and 4 in fig. 5) as 0.18 6 0.06
to 0.09 6 0.03 MYA for H. ‘‘Nshere’’ and 0.12 6 0.05 to
0.06 6 0.03 MYA for H. ‘‘Lutoto’’ (table 2). These
estimates give the upper limit for the time when the
isolation of the crater lake fish occurred. The common
ancestral nodes of the Lake Nshere and the Lake Lutoto
fishes (nodes 5 and 6 in fig. 5) are estimated to date to 0.13
6 0.04 to 0.06 6 0.02 and 0.080 6 0.04 to 0.04 6 0.02
MYA, respectively. In contrast, the common ancestral
node (node 7 in fig. 5) of the Lake Victoria fishes is dated
to 0.23 6 0.08 to 0.12 6 0.04 MYA (table 2). These
estimates suggest that the isolation of Lake Victoria fishes
from other fishes occurred somewhat earlier than that of
the crater lake fishes.
1454Sato et al.
by guest on June 1, 2013
The specimens used were the same as those tested in
the mtDNA part of the study: 28 from Lake Nshere and 20
from Lake Lutoto. Altogether, 128 and 36 sequences were
obtained from fishes of these two lakes, respectively. After
the elimination of identical sequences and exclusion of
single nucleotide differences, these numbers reduced to
14 and five different H. ‘‘Nshere’’ and H. ‘‘Lutoto’’
sequences, respectively (fig. 6). The number of different
sequences per individual ranged from one to five (see
Supplementary Material online for detailed information),
indicating that at least some of the sequences were derived
from different loci. To identify these loci, we subjected the
sequences to phylogenetic analysis (fig. 7) and regarded
sequences containing distinct, well-separated clades on the
phylogenetic tree as representing separate loci (LC1
through LC7). Although these locus assignments must be
regarded as tentative, our confidence in them is based on
finding orthologous loci in related cichlid species from
other localities (fig. 7). Southern blot analysis of genomic
DNA isolated from these fishes and hybridized with exon
3 class II B–specific probes revealed the presence of
multiple bands corresponding to different loci (fig. 8).
(DNA of H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ was not available
in the quantity and quality necessary for Southern blot
analysis of these fishes.)
The set of cichlid genes obtained so far is split into
three major, well-separated lineages, which we refer to as
I, II, and III (fig. 7). Lineage I, as presently defined,
encompasses most known haplochromine class II B
sequences, including those of haplochromines from Lake
Victoria and Lake Malawi, as well as class II B sequences
of Alcolapia (Oreochromis alcalicus), the tilapiine cichlids
from Lake Natron and Lake Magadi, and Oreochromis
niloticus. The lineage is divided further into the sub-
lineages A and B described previously by Figueroa et al.
(2000). Sublineage A comprises the loci LC4, LC5, and
LC6 and sublineage B the loci LC1, LC2, and LC3. Both
sublineages contain genes of both haplochromine and
tilapiine cichlids. Lineage II, by contrast, contains no
known genes of haplochromines from the Lake Victoria or
Lake Edward regions (Figueroa et al. 2000), but does
contain sequences derived from Lake Malawi and
Tanganyika haplochromines and from the tilapiine O.
niloticus (but not Alcolapia). Lineage III contains Mhc
class II B sequences from Lake Malawi haplochromines
and from the tilapiine cichlid O. niloticus, but no
sequences from Lake Victoria or Lake Edward region
haplochromines are present in this lineage. To what extent
Estimated Divergence Times of Groups of East African Haplochromine Fishes Indicated by
Name of Locality or mtDNA Control Region Haplogroup
h (3 10?3) Time Estimate (MYA)
Lake Malawi versus Lake Victoria þ rivers
Lake Victoria (group V) versus rivers
1 VD versus (VA þ VB þ VC)
2 VB versus VC
3 Lake Nshere versus VB
4 Lake Lutoto versus VB
5 MRCA (Lake Nshere)
6 MRCA (Lake Lutoto)
7 MRCA (VC)
44.5 6 7.3
30.5 6 5.3
7.2 6 2.0
4.4 6 1.2
4.1 6 1.2
2.6 6 1.2
2.9 6 1.0
1.8 6 1.0
2.6 6 1.0
1.37 6 0.24
0.32 6 0.09
0.20 6 0.05
0.18 6 0.06
0.12 6 0.05
0.13 6 0.04
0.08 6 0.04
0.23 6 0.08
0.68 6 0.12
0.16 6 0.05
0.10 6 0.03
0.09 6 0.3
0.06 6 0.3
0.06 6 0.02
0.04 6 0.02
0.12 6 0.04
NOTE.—MRCA indicates most recent common ancestor. The separation of Lake Malawi sequences from Lake Victoria
first column correspond to the numbers in figure 5. h is half the average distance between the two descendant clusters of the node.
FIG. 6.—Nucleotide alignment of Mhc class II B exon 2 sequences borne by H. ‘‘Nshere’’ (Hans) and H. ‘‘Lutoto’’ (Halu). Symbols are as in figure
4. Only variable sites are shown of 167 bp obtained and the numbering (to be read vertically downward) begins with codon 1 of exon 2. Sequences are
designated by the locus (LC1 through LC7) followed by the GenBank accession number. The alleles are identified by ignoring single nucleotide
differences (a nucleotide difference that appears in a single sequence data). The frequencies of the sequences in samples from Lake Lutoto, Lake
Nshere, and the Lake Victoria region (LV) are shown on the right. The accession numbers of sequences of the LV region samples are AF232204 and
AF232210 for LC3-AY211032, AF232207 for LC3-AY211033, AF232098 for LC4-AY211037, and AF232072 for LC6-AY211041.
Origin and Speciation of East African Crater Lake Fishes1455
by guest on June 1, 2013
FIG. 7.—Neighbor-joining tree of Mhc class II B exon 2 sequences of East African cichlid fishes. H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ sequences
produced in the present study are identified as Hans or Halu, respectively, followed by the number of the individuals and the GenBank accession
number, and are shaded. Other sequences are from Figueroa et al. (2000), Ma ´laga-Trillo et al. (1998), and Ono et al. (1993a, 1993b). I, II, and III
indicate lineages, A and B indicate sublineages, and LC1 through LC7 indicate different loci. Individual sequences are identified by species
1456 Sato et al.
by guest on June 1, 2013
the separation of sequences into these clades is influenced
by the choice of primers used for PCR amplification is
unclear at present.
One unique sequence of H. ‘‘Nshere’’ is included in
lineage II; all other sequences of fishes from the two crater
lakes are included in lineage I and are approximately
equally distributed between the sublineages IA and IB
(fig. 7). Thus, a minimum of three class II B loci must exist
in fishes from crater lakes (IA, IB, and II). Additional loci
are defined by distinct clades within lineage I. Altogether,
seven possible class II B loci are defined in the
haplochromines of the two crater lakes (fig. 7). The
predicted number of loci is also supported by the results of
the Southern blot analysis (fig. 8).
Distribution of SINEs
Terai and his coworkers (2003) identified a set of
SINEs that have been inserted into the genomes of
endemic Lake Victoria and Lake Edward region haplo-
chromines after the divergence of the Lake Victoria from
the Lake Malawi flock and persist in most of the species
and populations as presence or absence polymorphisms.
To test for the presence or absence of these SINEs in the
individual H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ specimens, we
used primers specific for the flanking regions of SINEs
1350, 1840, 1424, 1801, 1807del, 1909, and 1919 in PCR.
Of the seven SINEs tested, two were found to be poly-
morphic in H. ‘‘Nshere’’ (1918 and 1801), whereas only
one (1918) was polymorphic in H. ‘‘Lutoto’’; all other
SINEs were fixed in the two populations (table 3 [see also
Terai et al. 2003]). Comparison of observed and expected
frequencies of the genotypes at the polymorphic SINEs
revealed both populations to be in Hardy-Weinberg
equilibrium (table 3). This result suggests that neither of
the two populations is structured.
One of the fixed SINEs (1357) described by Terai et
al. (2003) was shown by these authors to display sequence
polymorphism, revealing the existence of at least 14 alleles
in the haplochromines of the Lake Victoria basin. To
obtain additional information about the H. ‘‘Nshere’’ and
H. ‘‘Lutoto’’ populations, we sequenced the 1357 locus of
individual fishes from these populations (fig. 9). The
sequences obtained in the present study were shorter than
those from Terai et al. (2003) in that they lacked 36 bp of
the 59 region and 38 bp of the 39 region. For this reason,
alleles D1 and D3, which differ at site 31 in the missing
region, could not be distinguished in the sequences
obtained in this study. Of the 14 alleles, at least four (D1
or D3, D4, D5, and E1) were found in H. ‘‘Nshere,’’ but
only one (D1 or D3) was detected in H. ‘‘Lutoto.’’ No new
SINE 1357 alleles were found in either of the two
Present Population Sizes of H. ‘‘Nshere’’ and H. ‘‘Lutoto’’
Information about the effective population sizes of H.
‘‘Nshere’’ and H. ‘‘Lutoto’’ is provided by the mtDNA
control region and SINE data. In the case of the mtDNA
control region sequences, we computed the nucleotide
diversity (average number of nucleotide differences per
site, p [see Nei 1987]) as being 3.41 3 10?3and 7.46 3
10?4for the H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ populations,
respectively. In a mutation-drift equilibrium, E(p)¼ 2Nfv,
where Nfis the effective population size of females and v is
the mutation rate per generation. By taking v¼2.2310?8
to 4.5310?8per site per year (see Materials and Methods)
and assuming that 1 generation ¼ 1 year in cichlid fishes,
we obtain Nf¼ 3.4 3 104to 7.7 3 104for H. ‘‘Nshere’’
and Nf¼ 0.8 3 104to 1.7 3 104for H. ‘‘Lutoto.’’
The Lower Limit for the Effective Population Size of
Crater Lake Fishes
Since polymorphism at some of the SINE loci is
shared by the crater lake fishes and fishes of the
neighboring region, we used the number of shared alleles
and the coalescence theory to obtain the lower limit of the
effective population size for H. ‘‘Nshere’’ and H.
‘‘Lutoto.’’ As described in Materials and Methods, we
computed the probability of naallelic lineages persisting
for 50,000 years by the coalescence algorithm (Hudson
abbreviations, GenBank accession codes (see Ono et al. 1993a, 1993b; Ma ´laga-Trillo et al. 1998; Figueroa et al. 2000), and locality (LM: Lake Malawi,
LT: Lake Tanganyika), or mtDNA haplotype. Species abbreviations are as follows: Aska, Astatotilapia katavi; Asnu, A. nubila; Asve, A. velifer; Auha,
Aulonocara hansbaenschi; Cyfr, Cyphotilapia frontosa; Enci, Enterochromis cinctus; Gasi, Gaurochromis simpsoni; Habb, Haplochromis sp. ‘‘big
blue’’; Habl, H. sp. ‘‘black’’; Hahn, H. sp. ‘‘high neck’’; Hali, H. lividus; Haro, H. rockkribensis; Hast, H. sp. ‘‘stone’’; Havb, H. sp. ‘‘velvet black’’;
Meau, Melanochromis auratus; Meze, Metriaclima zebra; Neni, Neochromis nigricans; Oral, Oreochromis alcalicus; Orla, Oryzias latipes; Orni,
Oreochromis niloticus; Pabe, Paralabidochromis beadlei; Ptsa, Ptyochromis sauvagei; Ptxe, P. xenognathus; Puny, Pundamilia nyererei; Yspy,
Yssichromis pyrrhocephalus. Numbers at nodes are bootstrap values (only those .50% are shown).
FIG. 8.—Southern blot analysis of genomic DNA isolated from three
Ptyochromis xenognathus (Ptxe) and Parabidochromis chilotes (Pach)
fish. All four samples were digested with Eco RI and the blots hybridized
with an Mhc class II B exon 3a probe (Figueroa et al. 1995) (the 59 part of
exon 3, which is interrupted by an intron in these fishes). Marker sizes in
kilobases (kb) are indicated.
Origin and Speciation of East African Crater Lake Fishes1457
by guest on June 1, 2013
1983; Tajima 1983) for various population sizes (N) and
searched for the value of N for which nalineages persist
for more than 50,000 years in 5% of the iterations of the
We can assume na¼ 2 in the case of the presence/
absence polymorphism of SINEs 1918 and 1801 in H.
‘‘Nshere’’ and 1918 in H. ‘‘Lutoto’’; and na¼ 4 for the
alleles determined by sequencing SINE 1357 and found to
be shared between H. ‘‘Nshere’’ and other lake popula-
tions. Using the na¼2 value, we can reject the hypothesis
that N was smaller than 6,000 individuals at the 5%
significance level for the past 50,000 years of existence of
the H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ populations. The 6,000
individuals thus represent the lower limit for the effective
sizes of these two populations. Using the na¼4 value, the
results of the computer simulation indicate that N¼24,000
is the lower limit of the effective H. ‘‘Nshere’’ population
size that assures the persistence of at least four allelic
lineages for the last 50,000 years.
Tajima (1989) designed a test that examines
d ¼ k ? S
by using the statistic
D ¼ d
where S is the number of segregating sites in nucleotide
sequences, n is the number of sequences obtained, and k is
the averagenumber of
sequence. This test is based on relationships existing
under the mutation-drift equilibrium
nucleotide differences per
and E(k) ¼ M, where M ¼ 2Nfv for the mtDNA sequences
and v is the mutation rate per sequence in this case. It was
developed for testing the neutrality of a locus under the
assumption that a population is in the mutation-drift
equilibrium. However, if we assume that neutrality holds
at the locus, the test can also be used for examining
a population size change in the past. A negative value of D
indicates an excess of S relative to k, which suggests
a recent rapid accumulation of mutations by population
expansion. The results of the Tajima test applied to the H.
‘‘Nshere’’ and H. ‘‘Lutoto’’ populations are shown in
table 4. Tajima’s D statistic is negative for both H.
Observed and Expected Genotype Frequencies (%) at SINE Loci in Crater Lake Cichlids
H. ‘‘Nshere’’ (n ¼ 28)
H. ‘‘Lutoto’’ (n ¼ 20)
NOTE.—n is the number of individuals tested; na is the number of individuals for which data were not obtained; obs indicates observed genotype frequencies; exp is the
expected genotype frequencies;þ/þ,þ/?, and?/?indicate homozygote for the presence of SINE insertion, heterozygote, and homozygote for the absence of SINE insertion,
respectively; þd indicates SINE insertion variant with 216-bp deletion. None of the v2values indicated significant deviation from Hardy-Weinberg equilibrium.
FIG. 9.—Nucleotide sequences of SINE 1357 alleles identified in
Lake Victoria basin haplochromines (only variable sites are shown).
Alleles found in H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ are identified in
parentheses by the respective numbers of the individuals from which the
allele sequences were obtained. Sequences A1 through A4 are from the
present study and start at site 37 and end at site 455; all other sequences
are from Terai et al. (2003) and begin at site 31. Other symbols are as in
figure 4. Dots (.) indicate that no data are available. The GenBank
accession numbers for A1 to A4 are AY211024 to AY211027.
1458Sato et al.
by guest on June 1, 2013
‘‘Nshere’’ and H. ‘‘Lutoto,’’ but the result is significant
only for the latter, suggesting that the population of H.
‘‘Lutoto’’ expanded recently. The population size esti-
mates shown in the earlier section were obtained by p,
which is equivalent to k, because p ¼ k/m where m is the
number of nucleotide sites examined. If we use the
Nfis estimated as 2.43104to 4.73104for H. ‘‘Lutoto.’’
Although a large error is associated with the estimate using
S, it is likely that the female population size of H.
‘‘Lutoto’’ is currently larger than 0.8 3 104to 1.7 3 104.
Founding Population Size
Among the exon 2 Mhc class II B gene sequences we
obtained, we found 14 different ones in H. ‘‘Nshere’’ and
five different ones in H. ‘‘Lutoto.’’ All these sequences,
with one exception, belong to lineage I; the exceptional
sequence found in H. ‘‘Nshere’’ belongs to the lineage II
sequences of Metriaclima zebra (fig. 5). The lineage I
sequences can be divided into six loci (see fig. 5), or they
can be treated as one locus. By considering these two
extreme possibilities, the numbers of alleles that were
passed onto the founding populations of the crater lakes
from the ancestral population are one to four for
H. ‘‘Nshere’’ and one for H. ‘‘Lutoto’’ in the former case
and 12 for H. ‘‘Nshere’’ and five for H. ‘‘Lutoto’’ in the
In a computer simulation similar to that described by
Vincek et al. (1997), we examined the number of alleles
retained for various founding population sizes (Nb¼ 5 to
500) and growth rates (r¼0.01 to 0.5) (see Materials and
Methods). Only representative values (r ¼ 0.01, 0.05, and
0.5) are shown in figure 10. In figure 10A, the average
numbers of alleles (nave) retained when the number of
individuals reached 10,000 in 1,000 replications of
computer simulation are shown in relation to the founding
population size (Nb). In figure 10B, instead of the average
number of alleles (nave), the 95% upper limit of the number
of alleles (nm) retained in the population is shown. That is,
in at least 95% of the replications, the number of alleles
retained in the population was smaller than nm. For each
Nbvalue, we can reject that the number of alleles retained
in the population is smaller than nmat the 5% significance
level. Thus, Nbvalues in figure 10B can give a lower limit
of the founding population size for the observed number of
alleles nm. The result of the simulation suggests that four or
five alleles can be retained even in the case of a small
founding population (5 to 10 individuals) if the growth rate
of the population after the founding event is not small
(, 0.1 to 0.2). The founding population must contain at
least 20 breeding individuals to retain 12 alleles, even for
a growth rate as high as 0.5. For a relatively low growth
rate (0.01), the founding population must be at least 50 to
70 to retain four to five alleles and at least 300 individuals
to retain 12 alleles.
Mitochondrial DNA haplotyping assigns the H.
‘‘Nshere’’ and H. ‘‘Lutoto’’ populations to the VB group,
which is widely distributed throughout the Lake Edward
region. On the phylogenetic tree (fig. 5), the set of VB
sequences closest to those of H. ‘‘Nshere’’ stemmed from
localities 39 (Kazinga Channel), 40 (Lake Edward), 44
(Lake George), and 45 and 46 (Lake Albert [see fig. 2]).
Similarly, the set of VB sequences closest to those of H.
‘‘Lutoto’’ is derived from the same area (localities 39 and
40), although here related sequences were also found in
Lake Albert. We suggest, therefore, that the founders of
the two crater lake populations originated in the Kazinga
Channel but that haplotypes closely related to those they
carried were distributed across much of the entire region.
Since the nearest neighbors of H. ‘‘Nshere’’ and H.
‘‘Lutoto’’ on the tree are different subsets of the VB
subgroup, we suggest further that the founders came from
different parts of the Kazinga Channel or from different
populations and that the colonizations of the two lakes
occurred independently. There is certainly no evidence that
H. ‘‘Nshere’’ originated from H. ‘‘Lutoto,’’ or vice versa.
Geologists date the earliest volcanic activity in the
Lake Lutoto and Lake Nshere region to approximately
50,000 years ago (Boven et al. 1998). This date places an
upper limit on the origin of the two lakes and hence also on
the time of their colonization by fishes. The date is
younger than the estimated time required for the mtDNA
control region haplotypes now present in H. ‘‘Nshere’’ and
H. ‘‘Lutoto’’ to coalesce to the MRCA. The difficulty with
coalescence time estimates, however, is that the calibration
of the mtDNA molecular clock is unreliable. For the East
African cichlid fishes, the clock is commonly calibrated by
the geological age of Lake Malawi (Meyer et al. 1990;
Nagl et al. 2000; Sturmbauer et al. 2001), which, however,
is uncertain (Schlu ¨ter 1997). The estimates range from 0.5
to 4 Myr, depending on the part of the lake examined and
the dating method. In an earlier publication (Nagl et al.
2000), we calibrated the clock on both the 2 Myr and 4
Myr age of Lake Malawi. Here, we also take the younger
age estimate (1 Myr) into account. Thus calibrated, the
molecular clock dates the MRCAs of the mtDNA control
region of H. ‘‘Nshere’’ to 60,000 or 130,000 years ago and
that of H. ‘‘Lutoto’’ to 40,000 or 80,000 years ago (the two
alternatives represent datings based on the 1 or 2 Myr age
of Lake Malawi). Both estimates for H. ‘‘Nshere’’ exceed
The Nucleotide Differences in mtDNA Control Region and
Tajima's Test of Neutrality
3.41 3 10?3
7.46 3 10?4
NOTE.—n is the number of samples from which sequences are obtained; p is
the average nucleotide difference per site; k is the average nucleotide difference per
sequence; S is the number of segregating sites; D is Tajima’s D statistic.
aD is significant at 5% level.
Origin and Speciation of East African Crater Lake Fishes1459
by guest on June 1, 2013
the upper limit set by the geological dates, but the H.
‘‘Lutoto’’ estimates come reasonably close to it. Of course,
large errors are associated with these time estimates (table
2), and the geological dating of the crater lakes is subject
to uncertainty (Boven et al. 1998). It is therefore difficult
to tell whether or not the age of the MRCAs of the crater
lake fishes really exceeds the geological date of the lakes.
In principle, there can be two reasons for the discrepancy
between molecularly and geologically estimated dates.
First, the coalescence time to the MRCA need not coincide
with the time of lake colonization and emergence of
a species because polymorphism could have passed
through the speciation phase. In the case of H. ‘‘Nshere’’
and H. ‘‘Lutoto,’’ however, there is no indication that any
of the mtDNA control region haplotypes now present in
the population existed before the emergence of these
species. Second, Lake Malawi could in fact be younger
than 1 Myr or the divergence of the Lake Victoria basin
cichlids from the Lake Malawi fishes might have
significantly postdated the emergence of the lake (Sturm-
bauer et al. 2001). Be this as it may, a large standard error
of the molecular clock estimate leaves the door open for
a younger age of the MRCA. In relative terms, the ages of
H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ may be comparable to
those of the endemic Lake Victoria species.
Both the SINE and the mtDNA data indicate that the
populations of H. ‘‘Nshere’’ and H. ‘‘Lutoto’’ are quite
large. The lower limit of N (effective size of a whole
population) from the SINE data is 6,000 for H. ‘‘Lutoto’’
and 24,000 for H. ‘‘Nshere,’’ and the Nf(female effective
population size) estimate from the mtDNA control region
data is 38,000 to 77,000 for H. ‘‘Nshere’’ and 8,000 to
17,000 for H. ‘‘Lutoto.’’ For lakes that have each a surface
area of less than 1 km2, these are indeed large effective
population sizes. In a large population, selection—where
applicable—gains the upper hand over random genetic
drift. It can lead to the rapid fixation of mutations
introducing adaptive modifications. This potential does
not seem to have been exploited by fishes of Lake Nshere
and Lake Lutoto, but has apparently been capitalized on by
fishes in some other crater lakes and by haplochromines of
Lake Victoria. In West Africa, the volcanic crater lakes
Barombi Mbo and Bermin in Cameroon harbor 11 and
nine tilapiine cichlid species, respectively (Trewavas,
Green, and Corbet 1972; Stiassny, Schliewen, and
Dominey 1992), which apparently arose in situ from
a single ancestor species by sympatric speciation (Schlie-
wen, Tautz, and Pa ¨a ¨bo 1994). Similarly, Lake Tana in
Ethiopia provides a home for 14 morphotypes of large
barbels (Barbus intermedius complex), all adapted to
different ecological niches, exploiting different food
resources, and derived from a common ancestral species
within the lake (Nagelkerke, Sibbing, and Osse 1995;
Dixon et al. 1996). The tremendous morphological
diversity of the endemic Lake Victoria haplochromines is
well documented (Greenwood 1981; Seehausen 1996).
The difference between Lake Victoria on the one hand and
Lake Nshere and Lake Lutoto on the other is, of course,
that Lake Victoria, being the third largest lake in the world,
offers many different ecological niches for adaptive
radiation that the crater lakes do not. By contrast, Lake
Barombi Mbo and Lake Bermin on the one hand and Lake
Nshere and Lake Lutoto on the other are comparable in
size. Whether they are also comparable in the number of
ecological niches they potentially provide for fishes is not
known to us. The reasons why cichlids adaptively radiated
in the West African but not in the two East African crater
lakes are unclear.
The lower limits of N (6,000 for H. ‘‘Lutoto’’ and
24,000 for H. ‘‘Nshere’’) obtained in this study were based
on the assumption that the size remained constant over
50,000 years. If this was the case, the founding populations
of these two species must have been quite large. Although
a bottleneck phase is not excluded by these considerations,
if it did occur it must have been of short duration and
followed by very rapid expansion of the population to its
present size. The lower limit of the bottleneck is indicated
by the Mhc data. It is 10 to approximately 300 breeding
individuals, depending on how the data are interpreted.
These numbers make the hypothesis involving the
‘‘seeding’’ of the lakes by fish eggs or fry transmitted by
animals an unlikely proposition. Even if birds were able to
deliver up to 300 founders into each of the two lakes, one
would not expect the eggs or fry to be of the same origin in
terms of species and locality, whereas all the available
evidence suggests a monophyletic origin of both H.
‘‘Nshere’’ and H. ‘‘Lutoto’’ in a single founding event.
Most likely, the founders of H. ‘‘Nshere’’ and H. ‘‘Lutoto’’
reached the lakes by way of rivers, and perhaps the way to
Lake Lutoto was longer or more difficult than the way to
FIG. 10.—The results of computer simulations showing the relationship between the founding population size and the number of alleles retained in
the population. Nbis the number of individuals in the founding size; r is the growth rate. (A) naveis the average number of alleles retained in the
population. (B) nmis the 95% upper limit of the number of alleles. Nbgives the lower limit of the founding population size for retaining nmalleles. The
number of alleles in the population exceeds nmin less than 5% of iterations.
1460 Sato et al.
by guest on June 1, 2013
Lake Nshere, so fewer fish reached the former than the
latter body of water. In both cases, however, the genetic
diversity of the founding populations might have been
greater than was ultimately retained. This conclusion is
indicated by a comparison of the genetic variability
currently found in the crater lake populations and that
present among the Kazinga Channel fishes. Some loss of
diversity apparently occurred by random genetic drift
during the expansion phase after the founding event. The
isolation of the crater lake population might have been
another factor contributing to the loss of variability.
All the genetic systems we tested consistently
indicate that compared with H. ‘‘Nshere,’’ H. ‘‘Lutoto’’
is genetically a more homogeneous population. Whereas in
H. ‘‘Nshere’’ we detected 10 different mtDNA control
region haplotypes, found four of the six Mhc class II B loci
to be polymorphic with two to four alleles per locus, could
establish the presence or absence of polymorphism at two
of the five SINE loci, and found four alleles at the SINE
1357 locus, in H. ‘‘Lutoto’’ we found seven mtDNA
alleles, all the Mhc class II B loci to be monomorphic, one
of the SINE loci to be polymorphic, and only one SINE
1357 allele. The reason for this difference between H.
‘‘Nshere’’ and H. ‘‘Lutoto’’ populations is not immedi-
ately apparent. The founding population of H. ‘‘Lutoto’’
might have been smaller than that of H. ‘‘Nshere,’’ as
mentioned earlier, or, as the Tajima test indicates, the H.
‘‘Lutoto’’ population might have experienced a recent
bottleneck and currently be in a recovering phase as a
consequence of some event such as a change in water level.
In contrast to molecular diversification, morpholog-
ical diversification reveals just the opposite trend when the
populations of the crater lakes and Lake Victoria are
compared. The two crater lake species are rather similar in
appearance, the differences between them in body shape
and coloration being inextensive. Similarly, the morpho-
logical variation within each crater lake population is
relatively minor (H. Tichy and E. Schraml, unpublished
data). By contrast, the endemic Lake Victoria species have
diverged much more extensively from each other mor-
phologically and each shows considerable intraspecies
variation (Greenwood 1981; Seehausen 1996). These
disparate trends can be attributed to the differences in
opportunities for adaptation to distinct environmental
niches on the one hand and to opportunities to mix
diverging populations in the large and small lakes on the
At the molecular level, the most striking difference
between speciation of haplochromines in the crater lakes
and in Lake Victoria is in the fixation of substitutions.
Although, as stated earlier, the endemic haplochromines in
the crater lakes and in Lake Victoria had approximately the
same length of time to evolve, the crater lakes species each
accumulated two or three diagnostic substitutions in the
control region of their mtDNAs since the time of their
origin from the founding populations, but the endemic
Lake Victoria species tested thus far have not accrued any
such fixed changes (Nagl et al. 1998, 2000). This dif-
ference is most likely attributable to the different degrees
of isolation of the populations. Whereas the crater lake
populations apparently evolve in complete physical iso-
lation with only one species per lake, in Lake Victoria the
reproductive barriers between the emerging species may
be leaky. The barriers may be sufficiently strong to
segregate characters under selection, differentiating the
populations morphologically and behaviorally for longer
periods of time, but not strong enough to assure fixation of
neutral mutations by random genetic drift. Occasional
breaches of the barrier may keep these mutations in
a polymorphic state. An important implication of these
deductions is the tentativeness of the early phases of
speciation in the endemic Lake Victoria haplochromines.
We suggest that as long as the environmental conditions in
the lake remain stable, the different species may evolve in
isolation from one another. Dramatic changes in the
conditions may, however, usher in a phase of a limited
gene flow between some of the species and so prevent
fixation of neutral mutations. The ease with which the
Lake Victoria haplochromines produce interspecies hy-
brids in aquaria attests to the weakness of the reproductive
barriers between the species.
We thank Erwin Schraml for help in sample
collecting during the expeditions, as well as Sabine Rosner
for technical and Jane Kraushaar for editorial assistance.
Avise, J. C., J. Arnold, R. M. Ball, E. Bermingham, T. Lamb, J.
E. Neigel, C. A. Reeb, and N. C. Saunders. 1987. Intraspecific
phylogeography: the mitochondrial DNA bridge between
population genetics and systematics. Annu. Rev. Ecol. Syst.
Boven, A., P. Pastels, L. E. Punzalan, T. K. Yamba, and J. H.
Musisi. 1998. Quaternary perpotassic magmatism in Uganda
(Toro-Ankole Volcanic Province): age assessment and
significance for magmatic evolution along the East African
Rift. J. Afr. Earth Sci. 26:463–476.
Dixon, B., L. A. J. Nagelkerke, F. A. Sibbing, E. Egberts, and R.
J. M. Stet. 1996. Evolution of MHC class II b chain-encoding
genes in the Lake Tana barbel species flock (Barbus
intermedius complex). Immunogenetics 44:419–431.
Donaldson, K. A., and R. R. Wilson. 1999. Amphi-panamic
geminates of snook (Percoidei: Centropomidae) provide a
calibration of the divergence rate in the mitochondrial DNA
control region of fishes. Mol. Phylogenet. Evol. 13:208–213.
Figueroa, F., W. E. Mayer, H. Su ¨ltmann, C. O’hUigin, H. Tichy,
Y. Satta, N. Takezaki, N. Takahata, and J. Klein. 2000. Mhc
class II B gene evolution in East African cichlid fishes.
Figueroa, F., H. Ono, H. Tichy, C. O’hUigin, and J. Klein. 1995.
Evidence for insertion of a new intron into an Mhc gene of
perch-like fish. Proc. R. Soc. Lond. B Biol. Sci. 257:325–330.
Greenwood, P. H. 1981. The Haplochromine fishes of the East
African lakes. Cornell University Press, Ithaca, New York.
Hudson, R. R. 1983. Testing the constant-rate neutral allele
model with protein sequence data. Evolution 37:203–217.
Imanishi, T., A. Wakisaka, and T. Gojobori. 1991. Genetic
relationships among various human populations indicated by
MHC polymorphisms. Pp. 627–632 in K. Tsuji, M. Aizawa,
and T. Sasazuki, eds. HLA 1991. Proceedings of the eleventh
international histocompatibility workshop and conference.
Oxford University Press, Oxford.
Origin and Speciation of East African Crater Lake Fishes1461
by guest on June 1, 2013
Klein, D., H. Ono, C. O’hUigin, V. Vincek, T. Goldschmidt, and
J. Klein. 1993. Extensive MHC variability in cichlid fishes of
Lake Malawi. Nature 364:330–334.
Klein, J., D. Klein, F. Figueroa, A. Sato, and C. O’hUigin. 1997.
Major histocompatibility complex genes in the study of fish
Molecular systematics of fishes. Academic Press, New York.
Klein, J., A. Sato, S. Nagl, and C. O’hUigin. 1998. Molecular
trans-species polymorphism. Annu. Rev. Ecol. Syst. 29:1–21.
Ma ´laga-Trillo, E., Z. Zaleska-Rutczynska, B. McAndrew, V.
Vincek, F. Figueroa, H. Su ¨ltmann, and J. Klein. 1998.
Linkage relationships and haplotype polymorphism among
cichlid Mhc class II B loci. Genetics 149:1527–1537.
Meyer, A., T. D. Kocher, P. Basasibwaki, and A. C. Wilson.
1990. Monophyletic origin of Lake Victoria cichlid fishes
suggested by mitochondrial DNA sequences. Nature
Nagelkerke, L. A. J., F. A. Sibbing, and J. W. M. Osse. 1995.
Morphological divergence in the large barbs (Barbus spp.) of
Lake Tana, Ethiopia. Neth. J. Zool. 45:431–454.
Nagl, S., H. Tichy, W. E. Mayer, N. Takahata, and J. Klein.
1998. Persistence of neutral polymorphisms in Lake Victoria
cichlid fish. Proc. Natl. Acad. Sci. USA 95:14238–14243.
Nagl, S., H. Tichy, W. E. Mayer, N. Takezaki, N. Takahata, and
J. Klein. 2000. The origin and age of the haplochromine
species flock in Lake Victoria, East Africa. Proc. R. Soc.
Lond. B Biol. Sci. 267:1049–1061.
Nei, M. 1987. Molecular evolutionary genetics. Columbia
University Press, New York.
Nei, M., and S. Kumar. 2000. Molecular evolution and
phylogenetics. Oxford University Press, Oxford.
Okada, N. 1991. SINEs. Curr. Opin. Genet. Dev. 1:498–504.
Okada, N., M. Hamada, I. Ogiwara, and K. Ohshima. 1997.
SINES and LINES share common 39 sequences: a review.
Ono, H., C. O’hUigin, H. Tichy, and J. Klein. 1993b. Major-
histocompatibility-complex variation in two species of cichlid
fishes from Lake Malawi. Mol. Biol. Evol. 10:1060–1072.
Ono, H., C. O’hUigin, V. Vincek, and J. Klein. 1993a. Exon-
intron organization of fish major histocompatibility complex
class II B genes. Immunogenetics 38:223–234.
Saitou, N., and M. Nei. 1987. The neighbor-joining method:
a new method for reconstructing phylogenetic trees. Mol.
Biol. Evol. 4:406–425.
Sato, A., D. Klein, H. Su ¨ltmann, F. Figueroa, C. O’hUigin, and J.
Klein. 1997. Class I Mhc genes of cichlid fishes: identification,
expression, and polymorphism. Immunogenetics 46:63–72.
Schliewen, U. K., D. Tautz, and S. Pa ¨a ¨bo. 1994. Sympatric
speciation suggested by monophyly of crater lake cichlids.
Schlu ¨ter, T. 1997. Geology of East Africa. Gebru ¨der Borntraeger,
Schmid, C., and R. Maraia. 1992. Transcriptional regulation and
transpositional selection of active SINE sequences. Curr.
Opin. Genet. Dev. 2:874–882.
Seehausen, O. 1996. Lake Victoria rock cichlids: Taxonomy,
ecology, and distribution. Verduijn Cichlids, Zevenhuizen,
Shand, R., and B. Dixon. 2001. Teleost major histocompatibility
genes: diverse but not complex. Mod. Asp. Immunobiol.
Simonsen, K. L., G. A. Churchill, and C. F. Aquadro. 1995.
Properties of statistical tests of neutrality for DNA poly-
morphism data. Genetics 141:413–429.
Stiassny, M. L. J., U. K. Schliewen, and W. J. Dominey. 1992. A
new species flock of cichlid fishes from Lake Bermin,
Cameroon with a description of eight new species of Tilapia
(Labroidei; Cichlidae). Ichthyol. Explor. Freshwaters 3:311–
Sturmbauer, C., S. Baric, W. Salzburger, L. Ru ¨ber, and E.
Verheyen. 2001. Lake level fluctuations synchronize genetic
divergences of cichlid fishes in African lakes. Mol. Biol. Evol.
Tajima, F. 1983. Evolutionary relationship of DNA sequences in
finite populations. Genetics 105:437–460.
———. 1989. Statistical method for testing the neutral mutation
hypothesis by DNA polymorphism. Genetics 123:585–595.
Tamura, K., and M. Nei. 1993. Estimation of the number of
nucleotide substitutions in the control region of mitochondrial
DNA in humans and chimpanzees. Mol. Biol. Evol. 10:512–
Terai, Y., N. Takezaki, W. E. Mayer, H. Tichy, N. Takahata, J.
Klein, and N. Okada. 2003. Phylogenetic relationships among
East African haplochromine fishes as revealed by short
interspersed elements (SINEs). J. Mol. Evol. (in press).
Thompson, J. D., D. G. Higgins, and T. J. Gibson. 1994. Clustal
W: improving the sensitivity of progressive multiple sequence
alignment through sequence weighting, positions-specific gap
penalties and weight matrix choice. Nucleic Acids Res.
Trewavas, E., J. Green, and S. A. Corbet. 1972. Ecological
studies on crater lakes in West Cameroon fishes of Barombi
Mbo. J. Zool. 167:41–95.
Vincek, V., C. O’hUigin, Y. Satta, N. Takahata, P. T. Boag, P. R.
Grant, B. R. Grant, and J. Klein. 1997. How large was the
founding population of Darwin’s finches? Proc. R. Soc. Lond.
B Biol. Sci. 264:111–118.
Yang, Z. 2002. PAML A program package for phylogenetic
analysis by maximum likelihood. Version 3.12. University
Naruya Saitou, Associate Editor
Accepted April 20, 2003
1462 Sato et al.
by guest on June 1, 2013