Mitochondrial DNA polymorphisms in Italy. III. Population data from Sicily: a possible quantitation of maternal African ancestry
Semino et al. (1989)
Ann Hum Genet
Link to Abstract
Of particular interest is that the HpaI-3/AvaII-3 complex, which is unique to groups of African ancestry, was found in Sicily at a frequency of 4.4%. For the first time an estimate of the amount of gene flow from Blacks to the Sicilian gene pool could be obtained. […] Using the weighted mean of the frequencies of the HpaI-3/AvaII-3 marker in Senegalese and in Bantu as representative of the parental African population, the total amount of gene migration (M) from Blacks into the Sicilians was estimated according to the method of Bernstein (1931) and a value of 0.108 ± 0.053 was obtained.
The Negroid component could have been transmitted directly through the introduction of groups of Negro slaves into the island by Phoenicians and Romans and/or indirectly through Arabic migrations. […] The only Arabs for which data on mtDNA polymorphisms are available, are from Israel (Bonnk-Tamir et al. 1986). These show an incidence of the combination of interest of 12.8%, a frequency which is compatible with a Negro contribution in the Arabic gene pool of about 30%. […] If one assumes that Negro genes arrived in Sicily only through Arabs, a 0.344 ± 0.049 value would be obtained for the amount of Arab gene migration (M).
Although the actual genetic contribution from African populations could only be estimated as lying in between the two M values we calculated, this work shows that, whichever way these genes arrived, a substantial Negro component is present in the Sicilian gene pool.
This is an old study and the marker being used is not an actual haplogroup but a restriction enzyme. And the way that admixture is being "estimated" isn't very common practice in population genetics. Not only is it imprecise and unreliable, as the authors freely admit, but in this case it's based on an unusually high frequency (4.4%) of African mtDNA in Sicily, as we'll see below.
But even disregarding all that, the lower limit estimated has a margin of error of ± 5.3, so it could be as little as 5.5% (i.e. 2.75% total admixture), and the upper limit assumes huge levels of Arab admixture that are not born out by any research. Moreover, the Arab reference sample is composed of Palestinians, whom the authors "estimate" have received 30% African gene flow. Yet when Palestinians were tested with the STRUCTURE admixture program using 377 autosomal microsatellite loci, they only had 2% total African admixture (Rosenberg et al. 2002; Table 2 in the Supplementary Information), and when the same sample was typed at 642,690 SNPs, they showed no African admixture at all (Li et al. 2008; Table S1). That's a powerful example of the limitations of this study's outdated estimates.
Another study from more than a decade later, Vona et al. (2001), addressed some of these issues and the old Semino paper directly, showing that with a different sample and method, different results are obtained:
In work carried out with restriction enzymes on mtDNA in a sample of Sicilians, Semino et al. (1989) indicated the presence (4.4%) of the African complex HpaI-3/AvaII-3 (40% in Senegal and in the Bantu of South Africa). The authors hypothesized a migration of genes from Africa to Sicily, estimated at about 10%, which was introduced into the Sicilian gene pool by Black slaves brought by the Phoenicians and the Romans and/or by Arab migrations. Results at the mtDNA sequencing level, however, show no Black African influence in the Sicilian population.
Two years after that, a definitive analysis of Sicilian mtDNA was conducted by Romano et al. (2003), using a much larger sample and seven different locations. Unlike Vona et al., this study did find some Negroid maternal DNA in Sicily, but at a rate of only 0.65% (3 sequences in the sample of 465).
If we pool all of the above data — which is always a good idea — we get 7 sequences in a sample of 604, which is 1.16% maternal admixture (or 0.58% total admixture), a figure that's extremely low and comparable to admixture levels elsewhere in Europe.
This result, together with the absence of Negroid Y-chromosomes in Sicily, discredits claims that Sicilians have a Black African racial component.