Ann. Hum. Genet. (1997), 61, 507–518
Printed in Great Britain
507
A tale of two islands: population history and mitochondrial DNA sequence
variation of Bioko and Sa
4
o Tome
!
, Gulf of Guinea
E. MATEU
"
, D. COMAS
"
, F. CALAFELL*, A. PE
;
REZ-LEZAUN
"
, A. ABADE
#
J. BERTRANPETIT
"
,
†
"
Laboratori d’Antropologia, Facultat de Biologia, Universitat de Barcelona, Barcelona,
Catalonia, Spain
#
Departamento de Antropologia, Universidade de Coimbra, Coimbra, Portugal
(Received 25.7.97. Accepted 17.10.97)
The hypervariable segment I of the control region of the mtDNA was sequenced in 45 unrelated
individuals from Bioko and 50 from Sa
4
o Tome
!
, two islands in the Gulf of Guinea that have had very
different settlement patterns: Bioko was colonized around 10000 BP, while Sa
4
o Tome
!
was first
settled by the Portuguese, who brought African slaves to the island. Two different patterns of
sequence variation are evident and are also clearly a consequence of their very different demographic
histories. The Bubi present a low genetic diversity and it is likely that the island was colonized by
a small number of individuals with small later migration. Sa
4
o Tomeans might be considered a subset
of a mainland African population relocated to the island. They present high genetic diversity with
a high number of sequences being shared with many continental populations. This study, with
knowledge of the population history in island populations, strengthens the genetic approach to
unravel past demographic events.
Mitochondrial DNA sequence analysis has
proved to be a powerful tool in the study of
human population history. It has been applied to
very different space and time frames, from the
origins of anatomically modern humans (Cann et
al. 1987; Vigilant et al. 1991) to the colonization
of continents (Ward et al. 1991 ; Kolman et al.
1996; Richards et al. 1996 ; Comas et al. 1997),
and to specific populations (Santos et al. 1994:
Bertranpetit et al. 1995 ; Mountain et al. 1995;
Calafell et al. 1996 ; Comas et al. 1996). Several
properties of mitochondrial DNA (mtDNA)
* Current address: Department of Genetics, Yale
University School of Medicine, New Haven, CT, USA.
† Correspondence: Jaume Bertranpetit, Laboratori
d’Antropologia, Facultat de Biologia, Universitat de
Barcelona, Diagonal 645, 08028 Barcelona, Catalonia,
Spain. Tel.: 34–3-402 14 61. Fax: 34–3-411 08 87.
E-mail: jaumeb!porthos.bio.ub.es
make it particularly suitable for evolution
studies: absence of recombination, maternal
inheritance, and a high mutation rate (Stonek-
ing, 1993).
The impact of demographic processes in
mtDNA sequence variation has been thoroughly
modelled. In particular, the distribution of
nucleotide pairwise differences (also known as
mismatch distribution; Rogers & Harpending,
1992; Harpending et al. 1993) is particularly
sensitive to the past demographic history of a
population. In the absence of selection and with
mutation rates uniformly distributed across
nucleotides, it has been shown (Rogers & Har-
pending, 1992; Harpending et al. 1993) that
stationary populations would have irregular
mismatch distributions, while a population ex-
pansion would generate a bell-shaped distri-
bution, with a mode travelling to the right with
time. A few African populations (Pygmies and
508 E. M
!Kung in particular ; Harpending et al. 1993)
present irregular mismatch distributions, while
other African populations (Graven et al. 1995;
Watson et al. 1996) and virtually all other
populations present bell-shaped mismatch histo-
grams (Harpending et al. 1993). It has been
shown that, in Europe, mismatch modes decrease
from SE to NW, which has been interpreted as
the footprint of the colonization of Europe by
anatomically modern humans (Calafell et al.
1996; Comas et al. 1996, 1997; Francalacci et al.
1996) recently confirmed by Neandertal mtDNA
analysis (Krings et al. 1997). However, some of
the implicit assumptions in the Rogers & Har-
pending (1992) model have been reassessed by
Aris-Brosou & Excoffier (1996), who found that
mutation rate variation across nucleotides (Hase-
gawa et al. 1993; Wakeley, 1993) could also
produce bell-shaped mismatch distributions.
Therefore, alternative hypotheses for the in-
terpretation of mismatch distributions should be
considered carefully before inferring a particular
population history.
As mentioned above, some African populations
present ragged pairwise difference distributions,
while others have bell-shaped distributions. Wat-
son et al. (1996) noted that the first are hunter-
gatherers while the latter are farmers or pastor-
alists. Thus, they interpreted that the demo-
graphic expansion that produced bell-shaped
mismatch distributions in Africa was the spread
of the Neolithic. The latest direct counts of
mutation events in the mtDNA (Howell et al.
1996; Parsons et al. 1997) resulted in estimates of
the mutation rate that were higher than pre-
viously thought; this adds support to a role for a
recent demographic event, such as the Neolithic
expansion, in generating bell-shaped mismatch
distributions in Africa (Watson et al. 1996) and
maybe elsewhere (Sajantila et al. 1995; Pa
$
a
$
bo,
1996; von Haeseler et al. 1996).
Sub-Saharan African populations share a
pattern of mitochondrial DNA variation, with
high levels of genetic diversity both within and
between populations (Vigilant et al. 1991). This
has also been found in nuclear genes (Armour et
al. 1996 ; Tishkoff et al. 1966), and is compatible
with the ‘Out of Africa ’ hypothesis, according to
which non-African humans have a common,
recent origin in Africa. This genetic pattern
could increase the power of genetic population
history analysis in African populations: a popu-
lation bottleneck in an African population might
be detected by comparison with the high levels of
genetic diversity in other populations, or possible
migrations can be traced back due to interpopu-
lation variability.
We have analysed two samples from two
African islands: the Bubi from Bioko and
individuals from Sa
4
o Tome
!
. Bioko is a 2000 km
#
island in the Gulf of Guinea, 30 km off the
Cameroon coast, which, together with four
smaller islands and the mainland territory of Rio
Muni constitute the Republic of Equatorial
Guinea. The island was first colonized 10000
years BP (Vara & Bolekia, 1993), at the end of
the last glacial period. Around 2000 years BP,
farming, and possibly a Bantu language, were
introduced to the island (Martı
!
n del Molino,
1993). The Bantu-speaking Bubi are the only
population native to Bioko, and thought to be
the descendants of the original colonizers of the
island; the contact with Europeans decimated
them to a few thousand at the turn of the 20th
century. Nowadays, they number 35 000, and
share the island with mainland Fang and Fernan-
dinos, the later being descendants of former
slaves liberated by the English in the 19th
century. Sa
4
o Tome
!
e Prı
!
ncipe, a former Portu-
guese colony, is located on the Equator in the
Gulf of Guinea. It consists of two main islands
(Sa
4
o Tome
!
and Prı
!
ncipe) and a number of islets.
Their total area is 964 km
#
, of which Sa
4
o Tome
!
comprises 865 km
#
.Sa
4
o Tome
!
island was prob-
ably uninhabited when first visited by European
navigators in the 1470s. Thereafter, the Portu-
guese began to settle convicts and exiled Jews
from Portugal on the island and established
sugar plantations, using slave labour from the
African mainland; for some years Sa
4
o Tome
!
was
important in the trade and transshipment of
slaves. A recent (1995) population size estimate
for Sa
4
o Tome
!
is around 100 000 inhabitants and
is mostly of African descent.
mtDNA variation in Bioko and Sa
h
o Tome
U
509
Bioko and Sa
4
o Tome
!
are, thus, two African
islands in close proximity, with similar areas and
population sizes, but with very different settle-
ment patterns. The Bubi of Bioko are the
descendants of one or a few ancient waves of
migration from the continent, whereas the Sa
4
o
Tomeans represent an admixed population with
a recent origin. It is likely that the original
settlers of Bioko were a small number of
individuals whose descendants in around 500
generations increased to the actual number of
Bubis. However, Sa
4
o Tome
!
may have been
peopled by a larger number of imported slaves.
These different patterns of settlement may have
had genetic consequences. If the number of
settlers of Bioko was small enough, we may be
able to observe a reduction of genetic diversity in
the Bubi when compared to the mainland
populations and to Sa
4
o Tome
!
. Under isolation,
gene diversity should be reduced, whereas the
complexity of the genetic variation, as measured
by coalescence patterns, mean pairwise differ-
ences, and nucleotide diversity, should remain
comparable to that of mainland populations. In
order to discover how two very different popu-
lation histories have influenced the genetic
structure of the population, we have sequenced a
fragment of 360 base pairs in the hypervariable
segment I of the mtDNA in 45 Bubi and 50
individuals from Sa
4
o Tome
!
, and we have com-
pared the sequences to a set of 15 African and
two European populations.
Population sampling
A 360-nucleotide sequence in hypervariable
segment I (HVSI) of the mtDNA control region
was analysed in 45 individuals from Bioko island,
Equatorial Guinea. The sample comprised self-
described unrelated Bubi individuals from the
villages of Moka-Bioko, Moka-Malabo, in
southern Bioko, and Rebola, in northern
Bioko. The sample from Sa
4
o Tome
!
island com-
prises 50 individuals from different places cover-
ing the whole island.
Sample collection and DNA extraction
DNA was extracted from hair roots for the
Bubi individuals. Hairs with their roots were
plucked and stored in a vial with 95 % ethanol.
One root of each sample was introduced in a
1.5 ml sterile microfuge tube containing 0.5 ml of
extraction buffer (10 m Tris pH 8.0, 10 m
EDTA pH 8.0, 100 m NaCl, 2% SDS, 39 m
DTT, and 20 mg ml
−
"
proteinase K), then incu-
bated at 37 °C and shaken at 180–200 rpm for at
least 3 hours. After a phenol-chloroform ex-
traction (Sambrook,1989), DNA was concen-
trated in Centricon-30 tubes and stored at
®20 °C. For the Sa
4
o Tome
!
sample DNA (sup-
plied by A. Abade, Coimbra) was extracted from
fresh blood using standard protocols.
mtDNA amplification
Amplification was performed using approxi-
mately 150–250 ng of the DNA sample in a 25 µl
reaction volume; the temperature profile for 30
cycles of amplification was 94 °C for 1 min, 58 °C
for 1 min and 72 °C for 1 min. The primers used
in this reaction, L15996 (5«-CTCCACCATTAGC-
ACCCAAAGC-3«), and H16401 (5«- TGATTTCA-
CGGAGGATGGTG-3« ; Vigilant et al. 1989) amp-
lified a 446-base pair (bp) segment containing the
360-bp region that was subsequently sequenced.
mtDNA sequencing
Out of the 45 Bubi samples, 13 were sequenced
with an automatic DNA sequencer, while the
remaining 32 were sequenced manually; the
choice of method depended only on sequencer
time availability. All the Sa
4
o Tome
!
samples were
sequenced with an automatic DNA sequencer.
Automated sequencing was performed according
to manufacturer’s specifications. The sequencing
reaction was performed separately on each
strand with the DNA Sequencing Kit
TM
(Perkin
Elmer), Dye Terminator Cycle Sequencing with
AmpliTaqR DNA Polymerase. The product of
510 E. M
Fig. 1. African populations included in our study.
the sequence reaction was run in an ABI PRISM
377 (Perkin Elmer).
When sequencing manually, the amplification
product was purified with GeneClean (BIO 101)
and 7 µl of the purified amplified product were
sequenced with Sequenase Version 2.0 (USB)
following supplier’s recommendations, except for
the annealing step, which was performed by
boiling the annealing reaction mixture for 3 min
in presence of nonidet P-40, followed by a short
time in a dry-ice ethanol bath. Both strands were
sequenced using the amplification primers. Re-
action products were separated by electrophore-
sis, dried, fixed, and subjected to autoradio-
graphy.
Sequences were aligned with the ESEE pro-
gram (Cabot, 1988), and the segment from
positions 16024 to 16383 (Anderson et al. 1981)
was used for analysis.
Statistical analysis
A set of 15 African population samples,
comprising a total of 645 individuals and in-
cluding the Fang in mainland Equatorial Guinea
(Figure 1; Table 1) and two European (a British
sample, n ¯ 100, Piercy et al. 1993 and a
Portuguese sample, n ¯ 54, Co
#
rte Real et al.
1996) were used as reference.
Nucleotide diversity (Nei & Tajima, 1981) was
estimated as (n}n®1) Σ
l
i =
"
(1-x
i
#
), where n is
sample size, l is sequence length, and x
i
is the
frequency of a nucleotide (A, C, G or T) at
position i. Similarly, sequence diversity was
estimated as (n}n®1) Σ
k
i =
"
(1®p
i
#
), where p
i
is
the frequency of each of the k different sequences
in the sample. The significance of the difference
in sequence diversity between two populations
was tested through a permutation procedure: the
individuals in both populations were randomly
assigned to one of two samples of the same size as
the original ones; sequence diversities were
computed for the new random samples, and the
difference was recorded. This procedure was
repeated 10 000 times, and the probability of the
difference not being significantly smaller than
zero was estimated as the fraction of permuted
differences that were less extreme than the
observed value. Tajima’s (1989) D statistic,
which is the standardized difference between two
different estimates of Θ ¯ 2N
e
µ, was computed.
Under a number of assumptions, Tajima’s D
measures the deviation from mutation-drift
equilibrium. Bertorelle & Slatkin (1995), and
Aris-Brosou & Excoffier (1996) have examined
the effects of population expansion and mutation
rate variability on D. We also estimated Θ from
the number of segregating sites (Watterson,
1975) through equation Θ ¯ P
n
}(11}2 …
(n®1)
−
"
) in Nei (1987), where P
n
is the number
of segregating sites divided by 360 and n is the
number of the observed sequences.
The phylogeny of mtDNA sequences in the
Bubi and in the Sa
4
o Tomeans was approached
through a neighbour-joining tree (Saitou & Nei,
1987) based on Kimura’s two-parameter model
with a transition to transversion ratio set to
15:1. We used PHYLIP (Felsenstein, 1989) to
produce these sequence phylogenies. The dis-
tribution of branch lengths in a phylogeny can be
used to quantify its degree of starness, that is,
whether most of the sequences attach with long
branches to a central point, as opposed to a
mtDNA variation in Bioko and Sa
h
o Tome
U
511
Table 1. African population set used for comparison
a
Population Reference n Country Language Family (Branch)
Bubi This study 45 Equat. Guinea NK (Bantu)
Sa
4
o Tome
!
This study 50 Sa
4
o Tome
!
e Portuguese and Creole
Prı
!
ncipe
Fang Pinto et al. 1996 11 Equat. Guinea NK (Bantu)
Yoruba Vigilant et al. 1991; 35 Nigeria NK (Central Niger Congo,
Watson et al. 1996 non Bantu)
Fulbe Watson et al. 1996 61 Nigeria, Niger, NK (West Atlantic)
Benin, Cameroon,
Burkina Faso
Mandenka Graven et al. 1995 119 Senegal NK (Mande)
Kanuri Watson et al. 1996 14 Nigeria, Niger NS (Saharan)
Songhai Watson et al. 1996 10 Niger, Mali NS (Songhai)
Hausa Watson et al. 1996 20 Nigeria, Niger AA (Chadic)
Somali Watson et al. 1996 27 Kenya, Somalia, AA (Cushitic)
Ethiopia
Turkana Watson et al. 1996 37 Kenya NS (East Sudanic)
Kikuyu Watson et al. 1996 25 Kenya NK (Bantu)
Pygmy Vigilant et al. 1991 37 CAR, Zaire NK (Bantu)
!Kung Vigilant et al. 1991 25 Namibia, KH (Northern)
Botswana
Tuareg Watson et al. 1996 26 Niger, Mali AA (Berber)
Mozabite Co
#
rte-Real et al. 1996 85 Algeria AA (Berber)
Berber Pinto et al. 1996 18 Morocco AA (Berber)
a
N: Sample size. Abbreviations: NK, Niger-Kordofanian; NS, Nilo-Saharan; AA, Afro-Asiatic; KH, Khoisanid.
pattern in which branching events are more
regular. The two patterns may reflect different
population histories : a star-like tree may cor-
respond to an expanding population, whereas a
regular tree could reflect a stationary population
(von Haeseler et al. 1996). In order to characterize
the branch length distribution, we computed its
third- and fourth-degree moment (i.e. its skew-
ness and kurtosis).
Genetic distances among populations were
estimated as d ¯ d
ij
®(d
ii
d
jj
)}2 (Nei, 1987),
where d
ij
is the mean nucleotide pairwise differ-
ence between populations i and j, and d
ii
and d
jj
are the mean internal pairwise differences of
populations i and j, respectively. The standard
error of the distances was estimated by resamp-
ling nucleotide positions in 1000 bootstrap
iterations. A neighbour-joining tree (Saitou &
Nei, 1987) was built from the genetic distance
matrix, and its robustness was assessed from
1000 boostrap iterations (Felsenstein, 1985).
Nucleotide diversity
The complete sequence of a 360-bp segment of
the mtDNA control region (HVS1, positions
16024 to 16383 according to the numeration by
Anderson et al. 1981) was determined in a sample
of 45 Bubi from Bioko and 50 individuals from
Sa
4
o Tome
!
(Table 2). For the Bubi sample, 18
different sequences were found, with 32 variable
nucleotide positions. The Sa
4
o Tomeans presented
32 different sequences, with 53 variable nucleo-
tides. Overall, 48 different sequences and 61
segregating sites were found. Four of the segre-
gating sites presented transversional changes, 36
were transitions between pyrimidines (C to T or
vice versa), 19 were purine transitions, and in two
positions (16114 and 16265), we observed both
transitions and transversions. Whereas two
thirds of the transitions observed involve pyrimi-
dines, these represent 55±8% of nucleotides in
this segment. This might represent part of the
uneven distribution of mutation rates across
nucleotides (Bertorelle & Slatkin, 1995; Aris-
Brosou & Excoffier, 1996).
