Charting the Ancestry of African Americans

Abstract

The Atlantic slave trade promoted by West European empires (15th–19th centuries) forcibly moved at least 11 million people from Africa, including about one-third from west-central Africa, to European and American destinations. The mitochondrial DNA (mtDNA) genome has retained an imprint of this process, but previous analyses lacked west-central African data. Here, we make use of an African database of 4,860 mtDNAs, which include 948 mtDNA sequences from west-central Africa and a further 154 from the southwest, and compare these for the first time with a publicly available database of 1,148 African Americans from the United States that contains 1,053 mtDNAs of sub-Saharan ancestry. We show that >55% of the U.S. lineages have a West African ancestry, with <41% coming from west-central or southwestern Africa. These results are remarkably similar to the most up-to-date analyses of the historical record.

The number of people of recent African ancestry living in the Americas is comparable to that in Africa itself, largely because of the forced mass migrations of the Atlantic slave trade during the 15th–19th centuries. Historical documentation indicates that many Africans were taken from the West African coast but that a large proportion were also taken from the former Portuguese colonies of Angola and Mozambique. The perpetrators included the Portuguese, Spanish, and British, and it is estimated that they moved ∼12 million people (with ∼10.5 million surviving the voyages) in <4 centuries (Eltis et al. 1998).

There has been a growing interest in the historical roots of African Americans and other communities of recent African ancestry, and genetics can play an important role in the study of that ancestry. However, much of the work to date has focused either on the extent of the African contribution to the American gene pool (Chakraborty et al. 1992; Parra et al. 1998, 2001; Mesa et al. 2000; Sans 2000) or on particular mtDNA lineages carried to the Americas through the Atlantic slave trade (Alves-Silva et al. 2000; Bandelt et al. 2001; Pereira et al. 2001; Salas et al. 2002). Despite widespread public interest and the rapidly growing involvement of private companies in genetic ancestry testing, the only published attempt to date to assign African geographical ancestry to African American lineages by use of genetic data has been by Salas et al. (2004). This latter work suggested a majority contribution of West Africa to the mtDNA pool of North and Central Americans of recent African descent, with a slightly smaller component from west-central Africa, whereas African Americans from South America (Brazil) reversed the picture, with a larger west-central African component. No significant component was detected from other regions, including Mozambique, despite the historical evidence of the movement of ∼1 million slaves from that region (Thomas 1998). Eurasian mtDNAs of recent African ancestry were found to be made up mainly of west-central and North African components, with a larger East African component in the Near East.

However, the analysis of Salas et al. (2004) had small sample sizes, particularly the small sample size from west-central Africa and the complete lack of samples from southwestern Africa, a known source of an estimated 3 million slaves (Thomas 1998). Here, we update the analysis, using 948 sequences from west-central Africa and a further 154 from Angola and Cabinda, as well as substantially enhanced data sets from the other regions of Africa (table 1) and a database of African American lineages that is more than an order of magnitude larger than that used earlier (Monson et al. 2002; see also the work of Bandelt et al. [2004a, 2004b] and Budowle et al. [2004]). This allows us to perform a much more decisive test of African ancestry among African American maternal lineages than has previously been possible.

Table 1.

African and African American Samples Used in the Present Study

Ethnic Group	Geographic Origin	Sample Size	Reference
America:
African American	United States	1,148	Monson et al. 2002
North Africa:
Algerian	Algeria	47	Plaza et al. 2003
Arab	Morocco	18	Plaza et al. 2003
Saharawi	Northwestern Africa	56	Plaza et al. 2003
Tunisian	Tunisia	47	Plaza et al. 2003
Berber	Tunisia (Chenini, Matmata, and Sened)	155	Fadhlaoui-Zid et al. 2004
West Saharan	Western Sahara	25	Rando et al. 1998
Mauritanian	Mauritania	30	Rando et al. 1998
Berber	North-central Morocco	4	Plaza et al. 2003
Moroccan	Morocco; Souss Valley	50	Brakez et al. 2001
Berber	North Morocco	60	Rando et al. 1998
Moroccan, not Berber	North Morocco	32	Rando et al. 1998
Egyptian	Egypt	68	Krings et al. 1999
Mozabite	Algeria	85	Côrte-Real et al. 1996
Canarian	Canary Islands	300	Pinto et al. 1996; Rando et al. 1998
Azorean	Portugal (Azores)	179	Brehm et al. 2003
Madeira	Portugal (Madeira)	155	Brehm et al. 2003
West Africa:
Kanuri	Nigeria, Niger	14	Watson et al. 1997
Tuareg	Nigeria, Niger, Mali	23	Watson et al. 1997
Fulbe	Nigeria, Niger, Benin, Cameroon, and Burkina Faso	60	Watson et al. 1997
Limba	Sierra Leone	68	Jackson et al. 2005
Loko	Sierra Leone	29	Jackson et al. 2005
Temne	Sierra Leone	121	Jackson et al. 2005
Mende	Sierra Leone	59	Jackson et al. 2005
Sierra Leonan	Sierra Leone	109	Monson et al. 2002
Hausa	Nigeria and Niger	20	Watson et al. 1997
Songhai	Nigeria, Niger, and Mali	10	Watson et al. 1997
Yoruba	Nigeria	34	Vigilant et al. 1991; Watson et al. 1997
Senegalese	Senegal	50	Rando et al. 1998
Serer	Senegal	23	Rando et al. 1998
Wolof	Senegal	48	Rando et al. 1998
Mandenka	Senegal	119	Graven et al. 1995
Cavo Verde	Cavo Verde	292	Brehm et al. 2002
West-central Africa:
Mbuti	Democratic Republic of Congo	20	Vigilant et al. 1991
Biaka	Central African Republic	17	Vigilant et al. 1991
Kotoko	Northern Cameroon	18	Černý et al. 2004
Hide	Northern Cameroon	23	Černý et al. 2004
Masa	Northern Cameroon	31	Černý et al. 2004
Mafa	Northern Cameroon	32	Černý et al. 2004
Bubi	Equatorial Guinea	45	Mateu et al. 1997
São Tomé	São Tomé and Príncipe	50	Mateu et al. 1997
Angolares	São Tomé and Príncipe	30	Trovoada et al. 2004
Forros	São Tomé and Príncipe	34	Trovoada et al. 2004
Tongas	São Tomé and Príncipe	38	Trovoada et al. 2004
Fang	Equatorial Guinea	11	Pinto et al. 1996
Mbenzele-Pygmy	Central African Republic; MBelemboké	57	Destro-Bisol et al. 2004
Ewondo-Pygmy	Cameroon, Febe	53	Destro-Bisol et al. 2004
Bamileke-Pygmy	Cameroon, Bandjoun	48	Destro-Bisol et al. 2004
Bakaka	Southern Cameroon	50	Coia et al. 2005
Bamileke	Southern Cameroon	48	Coia et al. 2005
Basa	South Cameroon	46	Coia et al. 2005
Daba	Northern Cameroon	20	Coia et al. 2005
Ewondo	Southern Cameroon	53	Coia et al. 2005
Fali	Northern Cameroon	41	Coia et al. 2005
Fulbe	Northern Cameroon	34	Coia et al. 2005
Mandara	North Cameroon	37	Coia et al. 2005
Ouldeme	Northern Cameroon	28	Coia et al. 2005
Tali	Northern Cameroon	20	Coia et al. 2005
Tupuri	North Cameroon	25	Coia et al. 2005
Podokwo	North Cameroon	39	Coia et al. 2005
Southwestern Africa:
Bantu-speaking	Cabinda	110	Beleza et al. 2005
Mbundu	Angola	44	Plaza et al. 2004
East Africa:
Kenyan	Kenya (Nairobi)	100	Brandstätter et al. 2004
Ethiopian	Ethiopia	74	Thomas et al. 2002
Turkana	Kenya	37	Watson et al. 1997
Somalian	Kenya, Somalia, Ethiopia	27	Watson et al. 1997
Nubian	Sudan, Egypt	80	Krings et al. 1999
Hadza	Tanzania	12	Vigilant et al. 1991
Hadza	Tanzania	79	Knight et al. 2003
Dakota	Tanzania	18	Knight et al. 2003
Iraqw	Tanzania	12	Knight et al. 2003
Kikuyu	Kenya	24	Watson et al. 1997
Nairobi	Kenya	100	Brandstätter et al. 2004
Nuba, Shilio, Duba, and Nuer	South Sudan	76	Krings et al. 1999
Sukuma	Tanzania	21	Knight et al. 2003
Uganda	Uganda	44	Richards et al. (2004)
Southeastern Africa:
Bantu-speaking	Mozambique	109	Pereira et al. 2001
Bantu-speaking	Mozambique	307	Salas et al. 2002
Southern Africa:
!Kung	Botswana	24	Vigilant et al. 1991
!Kung	South Africa	43	Chen et al. 2000
Khwe	South Africa	31	Chen et al. 2000
Herero	Bostwana	21	Vigilant et al. 1991
Herero	Namibia	6	Vigilant et al. 1991
Herero	Namibia	12	Soodyall 1993
Sekele-!Kung	…	36	Soodyall 1993
Sotho	South Africa	20	Soodyall 1993
Zulu	South Africa	8	Soodyall 1993
Xhosa	South Africa	8	Soodyall 1993
Dama	Namibia	21	Soodyall 1993
Nama	Namibia	18	Soodyall 1993
Total		6,008

Africa is the most genetically diverse continent. A fine subdivision of African mtDNA lineages provides a powerful source of phylogeographic information: major regions of the continent display markedly different frequencies of the continent-specific mtDNA clades, or haplogroups (fig. 1a). However, the first point to make from this enhanced data set is the obvious similarity of the haplogroup frequency profiles of West Africa, west-central Africa, and southwestern Africa in comparison with the other major regions of the continent. Although there are certain differences (which become more obvious in a finer classification of the lineages than at the gross haplogroup level), these regions display a broad similarity to each other, with most of the major haplogroups showing a gradient through the three regions, with (unsurprisingly) a much closer similarity of west-central and southwestern Africa compared with West Africa.

Figure 1.

a, The mtDNA haplogroup distribution in Africa regions (NA=North Africa; WA=West Africa; WCA=west-central Africa; SWA=southwestern Africa; EA=East Africa; SEA=southeastern Africa; SA=southern Africa). Each sequence was assigned to 1 of 15 haplogroups (Salas et al. 2002). b, The mtDNA haplogroup distribution in African Americans. c, Scatter plot of the first two principal components of the African American and the African regional haplogroup frequency profiles.

The African American mtDNA sample from the United States is striking for its strong resemblance to the mtDNA composition of all three of these regions (fig. 1b). Principal-components analysis (PCA) graphically portrays the same pattern: in a scatter plot of the first two components (which account for 77% of the variance of haplogroup frequencies), African American mtDNA is almost indistinguishable from that of these three African regions (fig. 1c).

This picture can be quantified by fitting a linear model to the frequency profiles, as done by Salas et al. (2004). Our analysis (table 2) indicated, as before, that North Africa, East Africa, southern Africa, and even southeastern Africa (Mozambique) show no significant contribution to the African American mtDNA pool in the United States. Thus, the larger data set here offers no support for the weak signal from southeastern Africa detected elsewhere (Salas et al. 2004).

Table 2.

Estimated Admixture Coefficients (of Source African Regions) for African Americans

		Posterior Mean andRoot-Mean-Square Deviationa
African Region	Sample Size	Low Resolutionb	High Resolutionc
North Africa	978	.001 ± .001	.001 ± .001
West Africa	1,549	.560 ± .042	.599 ± .028
West-central Africa	855	.135 ± .083	.090 ± .027
Southwestern Africa	157	.278 ± .066	.296 ± .024
East Africa	717	.026 ± .003	.002 ± .002
Southeastern Africa	416	.021 ± .018	.009 ± .008
Southern Africa	266	.002 ± .002	.002 ± .002

^aThe posterior mean and root-mean-square deviation of the admixture coefficient for each African region.

^bValues are based on a division of haplotypes into the 15 haplogroups shown in figure 1.

^cValues are based on a division of haplotypes into 56 haplogroups.

More significantly, we are now able to substantially revise and update the estimates for the main contributing regions. We performed the analysis in several ways. In the first, low-resolution analysis, for which we used the 15 haplogroups that are illustrated in figure 1a and were used for the PCA, West Africa contributed 56.0%±4.2%, southwestern Africa 27.8%±6.6%, and west-central Africa 13.5%±8.3% (table 2). Although the last is not significantly different from zero, the admixture values for the west-central and southwestern samples are highly negatively correlated (Pearson correlation coefficient of −0.9), indicating that the large SD of the posterior distribution is most plausibly explained by the similarity of haplogroup frequency profiles of our west-central and southwestern subdivisions. A reanalysis that combined these two similar regions (and also excluded the other uninformative regions of Africa) yielded values of 52.5%±4.1% for West Africa and 47.5%±4.1% for west-central Africa.

For a high-resolution analysis, to decrease the SD of the estimates, we divided the lineages into 56 haplogroups on phylogenetic grounds, such that no haplogroup included <20 samples. In this case, West Africa contributed 60%±3%, southwestern Africa 30%±2%, and west-central Africa 9%±3% (table 2). In this case, the correlation between west-central and southwestern African values was reduced to −0.6.

These values can be compared with classic estimates provided >30 years ago by Curtin (1969, p. 157) and with more up-to-date estimates. Curtin (1969) estimated that 50.4% of North American slaves came from West Africa and 47.8% from west-central Africa, with just 1.6% from southeastern Africa. A much more recent estimate for the United States (McMillin 2004, pp. 48, 54, and 70) suggests that ∼64% were from West Africa, ∼35% from west-central Africa, and ∼1% from southeastern Africa. Interestingly, our values are closer to the more recent estimate. Given the uncertainties involved and that the admixture estimates would most likely be affected by demographic factors subsequent to the arrival in the Americas, these values derived from entirely different methodologies are strikingly similar, corroborating both lines of evidence.

Figure 1a illustrates that characteristic West African (L1b; L2b,c,d; and L3b,d) and west-central African (L1c and L3e) haplogroups have been carried into the Americas at high frequencies. Typically, eastern African lineages (e.g., within haplogroups L3*, L1f*, and L1g) are poorly represented in America, again mirroring the historically documented low contribution of East Africa to the slave trade. The small impact of North Africa is evident by the low frequency (<1%) of its autochthonous haplogroup U6. Haplogroups that characterize southern African Khoisan-speaking groups (L0d and L0k) are completely absent in America.

Of course, these results indicate only the maternal component of the American gene pool. Further research on Y-chromosome and autosomal markers is clearly needed to test whether the contribution of females and males to the African American gene pool has been symmetrical or asymmetrical—research for which, again, large African databases of such marker systems will be needed, such as are now becoming available for mtDNA. With the acceleration of data acquisition now under way, these analyses will no doubt be possible in the near future.

We conclude that mtDNA variation allows us to trace the maternal ancestry of African Americans to broad geographic regions of Africa, with results that are closely concordant with historical studies that now encompass documentation for between two-thirds and three-quarters of the estimated total voyages made during the course of the Atlantic slave trade (Eltis et al. 1998). We have previously raised the possibility of whether, with larger data sets and extensive phylogeographic analyses, more-specific reconstructions will be possible (Salas et al. 2004). However, even with this substantially augmented data set, we note that it is still not possible to go further at this stage. Even with greatly improved geographic coverage, it remains the case that many mtDNAs are very widely distributed throughout the African continent, most likely as a result largely of the Bantu dispersals (Salas et al. 2002), but no doubt also as a result of both earlier and more recent movements, including those that are due to the Atlantic slave trade itself (Salas et al. 2004). This problem will continue to hamper the allocation of African American mtDNAs to narrower geographic locations in Africa, even if the resolution of the molecular analyses is increased from the first hypervariable segment (HVS-I) to complete mtDNA genomes.

Considerable caution is therefore warranted when dealing with claims in the popular media (such as the acclaimed and prestigious BBC television documentary Motherland: A Genetic Journey, first shown in the United Kingdom in 2003) and those made by genetic ancestry–testing companies about their ability to trace the ancestry of certain American (or, for that matter, European) mtDNAs to a particular locale or population within modern-day Africa. Our analyses stand as a warning to unsuspecting members of the public who may be seduced by such promises.