Results and Discussion
Haplogroup Distribution
We have identified 20 different haplogroups (Table 2), of the 26 possible, in the sample from the Canary Islands being indicative of the informativeness of the markers selected for the study of this population. Relatedness and nomenclature of haplogroups are shown in Fig. 2
based on those from the YCC (Y Chromosome Consortium,
2002). Haplogroup R1∗(xR1a, R1b6, R1b8)
is the most frequent in all seven islands, comprising 47%
of the total sample (range 39.9–53.9%). This is, by far,
also the most abundant haplogroup in the Iberian Peninsula
(Bosch et al. 2001), reaching higher frequencies than
in the Canarian population (P < 0.05, Mann-Whitney
U-test). Excluding this haplogroup, several others are
frequent (>10%), but only in some islands: haplogroup
E3b2 in Tenerife, Gran Canaria and Fuerteventura; haplogroup
I∗(xI1b2) in La Gomera; and haplogroup R1b8
in El Hierro. Haplogroup E3b2, which has been assigned
a Northwest African origin (Bosch et al. 2001),
constitutes a clear African influence, reaching in some
islands frequencies twice as high as those reported for the
Iberian Peninsula (Bosch et al. 2001). On the other hand,
haplogroups I∗(xI1b2) and R1b8 are European related
lineages (Semino et al. 2000), the latter with an Iberian
assignation (Hurles et al. 1999). Albeit at low frequencies,
we have also detected the sub-Saharan haplogroup
E3a in four of the seven islands. Considering the geographical
location of the populations included in this
analysis, haplogroup DE∗(xE3) may also be the result of
C14 University College London 2003 Annals of Human Genetics (2003) 67,138–152 143
African influence. It must be noted that haplogroups E3aand DE∗(xE3) have not been detected in Iberian samples to date (Semino et al. 2000; Bosch et al. 2001). Also, it is striking that both an Iberian haplogroup (R1b8) and a sub-Saharan African haplogroup (E3a) were at
their highest frequencies in the archipelago in El Hierro.
The presence of some haplogroups of widespread distribution
in Europe, in combination with others of clear
Northwest and sub-Saharan African assignation, reflect
the complex demographic patterns that have shaped the
current Canarian population.
Population Structure and Affinities
Haplogroup distribution in the archipelago is heterogeneous
(P = 0.001), although few (29%) pairwise comparisons
gave results at a significant level (Table 3). Furthermore,
when we carried out an AMOVA analysis,
a low amount (0.3%) of variance was attributed to differences
between populations within the archipelago.
This value is 4.5 and 8 times lower than that obtained
for the European and Northwest African populations
considered here, respectively, although Canarian populations
are distributed with a mean distance of 228 km,
those of Europe with a mean of 587 km, and Northwest
Africans with a mean of 668 km. This result could
be perfectly explained by the way that European occupation
occurred, with groups of people of mixed origin
(Su´arez et al. 1988). This event, and subsequent
gene flow, would have reduced the between island variance.
Even though some neighbouring islands were
also genetically related (Table 3), a Mantel test revealed
that genetic and geographic distances are not correlated
(r = −0.081, P = 0.612).
Table 3
FST distances (below diagonal) and exact probability values of differentiation (above diagonal) between Canarian populations
El Hierro La Palma La Gomera Tenerife G. Canaria Lanzarote Fuerteventura
El Hierro — 0.637 0.008 0.008 0.089 0.042 0.353
La Palma 0.012 — 0.273 0.295 0.886 0.525 0.307
La Gomera 0.026 0.012 — 0.003 0.064 0.522 0.004
Tenerife 0.023 0.007 0.020 — 0.997 0.075 0.019
G. Canaria 0.027 0.009 0.026 0.002 — 0.249 0.164
Lanzarote 0.020 0.005 0.010 0.008 0.010 — 0.120
Fuerteventura 0.016 0.014 0.020 0.017 0.023 0.018 —
P < 0.05 in bold.
Table 4
Percentages of variation among different geographical
areas resulting from the AMOVA analysis
Northwest sub-Saharan
Europe Africa Africa
El Hierro 0.005 0.344 0.371
La Palma 0.000 0.319 0.371
La Gomera 0.014 0.308 0.362
Tenerife 0.001 0.299 0.380
Gran Canaria 0.000 0.327 0.412
Lanzarote 0.000 0.328 0.386
Fuerteventura 0.015 0.258 0.361
Europe — — —
Northwest Africa 0.323
Sub-Saharan Africa 0.373 0.503
Canary Islands 0.003 0.279 0.342
P > 0.05 in bold.
When AMOVA was calculated between populations
pooled by geographical areas (Table 4), the lowest, although
significant, level of variation was observed in
the European-Canarian comparison, being two orders
of magnitude lower than the comparison with both
African groups. This clearly indicates that Canarian
populations have greater differences from Africans than
from Europeans. Even so, the percentage of variance
between Canarians and Africans is lower than percentages
between Europeans and Africans, showing again
an African paternal influence on the archipelago populations.
Looking at each island, all are significantly
different from both African groups, but only La Gomera
and Fuerteventura are significantly different from the
European pool (Table 4). More precisely, these two islands
have the lowest differentiation values from both,
Northwest and sub-Saharan Africans.
Genetic affinities were also inspected from a MDS
plot of FST genetic distances between populations
144 Annals of Human Genetics (2003) 67,138–152 C University College London 2003
Y-Chromosome Variation in Canary Islands
Figure 3
Two-dimensional MDS plot relating Canarian and
continental populations.
(Fig. 3). The correlation between F
ST and the new
distances derived from a two-dimensional plot was
very high (r = 0.971). In the same way, the obtained
stress value (0.127) indicates that the chosen dimensionality
is appropriate, and that a plot in a higher
dimension is not necessary (http://www.analytictech.
com/borgatti/mds.htm). Europeans, Northwest and
sub-Saharan African populations constituted three different
clusters. The Canarian populations formed a
relatively tight cluster near the European group, but
with some islands dispersed towards Northwest Africans.
From the European group, clearly the general Iberians
have the highest affinities with Canarian populations.
Table 5 Relative admixture contributions to the Canarian populations inferred from haplogroup frequencies
Population Estimator Europe Northwest Africa sub-Saharan Africa FSTa
El Hierro mL 0.957 ± 0.036 — 0.043 ± 0.036 0.003
mY 0.959 ± 0.063 — 0.041 ± 0.063 —
La Palma mL 0.961 ± 0.053 0.019 ± 0.049 0.021 ± 0.024 0.007
mY 0.929 ± 0.054 0.051 ± 0.049 0.021 ± 0.027 —
La Gomera mL 0.997 ± 0.057 0.003 ± 0.057 — 0.019
mY 0.899 ± 0.050 0.092 ± 0.051 0.009 ± 0.024 —
Tenerife mL 0.926 ± 0.041 0.074 ± 0.041 — 0.003
mY 0.894 ± 0.044 0.097 ± 0.040 0.009 ± 0.013 —
G. Canaria mL 0.897 ± 0.043 0.093± 0.041 0.010 ± 0.012 -0.005
mY 0.885 ± 0.063 0.096± 0.059 0.019 ± 0.023 —
Lanzarote mL 0.956 ± 0.046 0.035 ± 0.044 0.009 ± 0.015 0.003
mY 0.940 ± 0.049 0.051 ± 0.046 0.009 ± 0.022 —
Fuerteventura mL 0.832 ± 0.107 0.171 ± 0.107 — 0.006
mY 0.877 ± 0.069 0.123 ± 0.069 — —
Canary Is. Total mL 0.934 ± 0.030 0.059 ± 0.029 0.006 ± 0.008 0.003
mY 0.910 ± 0.026 0.082 ± 0.026 0.008 ± 0.011 —
a Variation not explained by the model (Long, 1991).
Admixture Estimates
In previous genetic approaches, to explain the colonisation
of the Canary Islands admixture proportions have
been inferred considering the extant people on the islands
as a hybrid population with a Northwest African
ancestry, contributed, most probably, by Berber aboriginal
founders and, to a lesser degree, by sub-Saharan
Africans introduced after the conquest as a result of slave
trade. The other main contribution is from the European
population that, since the beginning of the occupation,
was steadily colonising the islands. However,
the high paternal differentiation between these clusters
of populations makes it possible to estimate their contributions
to the current Canarian populations more accurately.
Our admixture results (Table 5)were similar when
mL or mY estimators were used, although African mY
estimates always gave slightly higher values. For all comparisons
the highest paternal contribution resulted from
European descent. Northwest Africans would have also
contributed significantly, but the sub-Saharan African
contribution was almost negligible, excepting the case
of El Hierro island. Differences between islands are evident,
as the Northwest African contribution diminishes
towards the westernmost islands, being insignificant
in El Hierro. The opposite occurs for European
contributions. As suggested from results on AMOVA
C14 University College London 2003 Annals of Human Genetics (2003) 67,138–152 145 analyses and genetic distances, the Europeans were the
major contributors to the current Canarian paternal
gene pool, explaining about 90% of its whole variability,
while Northwest and sub-Saharan Africans contributed
no more than 6–8%, and less than 1%, respectively
(Table 5). Clearly, there is a sharp contrast between paternal
admixture results and those from mtDNA and
autosomal loci. As mentioned above, the higher African
contribution to the Canarian mtDNA pool compared
to the contribution estimated from the autosomal loci
has been explained by an asymmetry of female and male
aboriginal contribution to the current Canarian population
(Pinto et al. 1996a; Flores et al. 2001a). During the
Spanish conquest the European migration was constituted
mainly of soldiers, who displaced autochthonous
males by death or deportation, but mixed with aborigine
females. A corollary of these results is that the
aboriginal Y-chromosome contribution to the presentday
Canarians has to be lower than those estimated
from autosomal and mtDNA polymorphisms. This is
exactly what we found with our experimental data. Although
with stronger directional mating, similar results
have been obtained for Amerind populations, which
were also profoundly influenced by Iberian conquerors
(Carvajal-Carmona et al. 2000; Sans, 2000; Carvalho-
Silva et al. 2001).
Regarding the portion of variance not explained by
the admixture model (Table 5), and so due to genetic
drift, the highest value, by one order of magnitude, was
obtained for La Gomera, the second smallest island. Precisely,
for mtDNA data La Gomera is the only island
significantly different to the others and characterized by
the highest frequency of the North African U6 cluster
(Rando et al. 1998, 1999). In this regard, genetic drift
could be responsible for the contrasting difference in
African ancestry detected with maternal (51% of African
lineages) and paternal markers (0.3–10% of African lineages).
Alternatively, it could reflect the dramatic way
the island was conquered, producing the strongest sexual
asymmetry in the archipelago (Su´arez et al. 1988).
Aborigine Settlement Process
Three lineages deserve special attention due to their distribution
in the archipelago and in populations from
South western Europe and Africa: haplogroups J∗ (xJ2),
E3b1 and E3b2. Haplogroup J∗ (xJ2) has been suggested
to have originated in Middle East populations and spread
to Europe with the dispersion of farming during the Neolithic
period (Semino et al. 2000; Rosser et al. 2000;
Bosch et al. 2001; Nebel et al. 2001). Its arrival to Northwest
Africa could have occurred at the same time as in
Mediterranean Europe (Bosch et al. 2001) or more recently,
during the Islamization of the area (Nebel et al.
2002). Nevertheless, it reaches higher frequencies in
North Africa than in the Iberian Peninsula (Bosch et al.
2001), being especially high in Ethiopians (Semino 2002). Although a clear North African origin has been
assigned to haplogroup E3b1 (Bosch et al. 2001), it also
characterizes most of the European YAP+haplogroups,
as pointed out by Semino et al. (2002). Again, it reaches
the highest frequencies in Ethiopians (Underhill et al.
2000; Semino
et al. 2002). Strikingly, both haplogroups
are coincidentally distributed in the Canary Islands,
reaching highest frequencies in western islands and decreasing
eastwards (Fig. 1). Furthermore, they appeared
in higher frequencies in the archipelago than in the European
populations considered. On the other hand, for
haplogroup E3b2 an unambiguous Northwest African
origin has also been suggested (Bosch et al. 2001). It
has been detected in Iberian populations (Bosch et al.
2001; Flores, 2002), but never in frequencies so high as
those reported for Northwest Africa. Moreover, its frequency
in Canarians as a whole is nearly double those
for the European populations considered, being particularly
abundant in eastern islands (Fig. 1). At this point,
there are at least three different explanations for the appearance
of these three haplogroups in Canary Islands
in such relatively high frequencies with respect to continental
populations. First, they arrived in the archipelago
with Iberian colonizers, as they are also influenced by
Africans (Flores et al. 2000, 2001c; Bosch et al. 2001),
and then augmented in frequency by genetic drift. Second,
some could have come with Iberian populations
and some directly from Northwest African influence.
And third, they are an exclusive contribution from
aboriginal inhabitants. Genetic drift could hardly
explain a simultaneous increase of frequency for the
three haplogroups, considering that they are precisely
the three major haplogroups within North Africa (Underhill
2000; Bosch et al. 2001). Furthermore,
STR diversity within haplogroups by area (Table 6)
146
Annals of Human Genetics (2003) 67,138–152 C University College London 2003
Y-Chromosome Variation in Canary Islands
Table 6
STR diversity within haplogroups
by area and mY haplotype contributions
to the Canary Islands
Haplogroup
Population E3b1 E3b2 J∗ (xJ2)
Northwest Africa 1.83 ± 1.31 0.66 ± 0.51 1.12 ± 0.78
Europe 0.93 ± 0.68 0.54 ± 0.46 0.58 ± 0.50
Canary Islands 1.33 ± 0.86 0.75 ± 0.56 1.37 ± 0.87
C/NWAa 0.168 ± 0.131 0.551 ± 0.677 0.334 ± 0.127
C/Eb 0.832 ± 0.131 0.449 ± 0.677 0.666 ± 0.127
a Contributions from Northwest Africa.
b Contributions from Europe.
argues against drift being the only factor responsible for
that increase. Note that Canarian diversities, calculated
from STR haplotypes (Appendix I), are always higher
than those of Iberians and, in two instances, even than
those of Northwest Africans. However, we must note
that the particular comparison with Northwest African
samples could be biased by the comparatively low number
of samples used. We then applied mY estimator, using
STR haplotypes and their molecular differences, to
clarify the percentage of influence by which an area has
contributed to Canarians for these haplogroups. The
results (Table 6) supported a compound origin of these
three haplogroups in the Canary Islands, making the
second scenario more likely. Although J∗ (xJ2) is not
considered a genuine African marker, it seems that its
presence in the Canary Islands has an important African
cause. Even though it simplified the picture, we have
considered haplogroups J∗ (xJ2) and E3b1 as indicators
of a movement of people coming from Southern parts
of North Africa (“Saharan marker”), and haplogroup
E3b2 as their Northwest African counterpart (“Northwest
African marker”). In fact, the Saharan marker
showed a trend of increasing frequencies westwards
(rs = 0.714, P = 0.072, two-tailed test), and the contrary
is observed for the Northwest African marker
(rs = −0.857, P = 0.014, two-tailed test). Similar clinal
patterns have been found for unlinked loci, such
as mtDNA (Rando et al. 1999) and CD4/Alu haplotypes
(Flores et al. 2001b), pointing to demographic
events as a major cause, as selection would affect single
genes. Even though the aboriginal population that
occupied the archipelago was probably not a homogeneous
group, the discrepant distribution of these haplogroups
in relation to their distribution in Africa could
be compatible with the hypothesis of more than one
prehistoric settlement, particularly in the eastern islands
(Onrubia-Pintado, 1987). Congruent with this hypothesis,
cultural (Onrubia-Pintado, 1987; Navarro, 1992;
Mart´ınez, 1996) and dialectal (Reyes-Garc´ıa, 2000) aborigine
heterogeneities have been revealed within islands,
supporting more than one arrival from Africa. Furthermore,
linguistic investigation supports the idea that major
dialectal influence in aborigines, common to all the
islands, could have come from southern Algeria, but
was also notably affected by Central Moroccan dialects
(Reyes-Garc´ıa, pers. comm.). However, this possibility
could not be confirmed in a previous mtDNA analysis
(Rando et al. 1999), as the significant westward decrease
of diversity and number of mtDNA African lineages
was considered concordant to one dominant initial settlement
from the African shore following a westward
stepping-stone process.
Dating the Aboriginal Colonisations
Assuming that more than one aborigine settlement
dispersed part of these three haplogroups in the
archipelago, we tentatively tried to date those processes.
Considering all the chromosomes from the Canary Islands,
Europe and Northwest Africa, haplogroup J∗ (xJ2)
would have originated ≈3760 (CI 1610–13147) years
ago, haplogroup E3b1 ≈6060 (CI 2598–22221) years
ago, and haplogroup E3b2 ≈1840 (CI 788–6433) years
ago. We then calculated divergence percentages of haplogroups
between the Canarian population and their
parental populations. The results showed again congruent
values for haplogroup J∗ (xJ2) and E3b1, 52.7
and 52.6% respectively, bolstering the hypothesis of
their coincident dispersal in the Canarian archipelago.
Translating into years, haplogroup J∗ (xJ2) gave an age of
≈2000 years and haplogroup E3b1 ≈3200 years. These
estimates match perfectly with the most ancient age of
C human occupation in the Canary Islands as deduced
from C14 dating (Onrubia-Pintado, 1987), and that deduced
from mtDNA (Rando et al. 1999). The divergence
for haplogroup E3b2 gave only 31.6%, which
means ≈600 years, placing this colonization just before
the European conquest. However, these coalescence
time estimates must be interpreted with caution.
They are highly influenced by demographic processes
and are based on many assumptions of microsatellite
mutation rates that could be erroneous (Stumpf & Goldstein,
2001).
In summary, our Y-chromosome analysis supports
the hypothesis that the current paternal pool from the
Canarian population is, to a great extent, of Iberian descent.
The male aboriginal influence has been estimated
to be less than 10%, whereas aboriginal mtDNA lineages
represent ≈45%, reflecting the aggressiveway the islands
were conquered. Nevertheless, with such a low number
of aboriginal lineages, we detected and dated two prehistoric
settlement processes, bolstered by osteological,
cultural and linguistic data: the first one, with a Saharan
substrate, arriving during the 1st millennium B.C., and
the second, with a Northwest African ancestry, spreading
just prior to the European conquest.
http://www.geocities.com/vetinarilord/canary.pdf
