On the basis of the previously published phylogeny (Y Chromosome Consortium ²⁰⁰²; Jobling and Tyler-Smith ²⁰⁰³), the mutations M2/P1/M180, on the one hand, and M35/M215, on the other, further subdivide E3 in two monophyletic haplogroups: E3a and E3b. Both haplogroups are frequent in Africa (Underhill et al. ²⁰⁰⁰; Cruciani et al. ²⁰⁰²), although, to date, only E3b has also been observed in Europe (Semino et al. ²⁰⁰⁰) and western Asia (Underhill et al. ²⁰⁰⁰; Cinnioğlu et al. ²⁰⁰⁴). Recently, it has been proposed that E3b originated in sub-Saharan Africa and expanded into the Near East and northern Africa at the end of the Pleistocene (Underhill et al. ²⁰⁰¹). E3b lineages would have then been introduced from the Near East into southern Europe by immigrant farmers, during the Neolithic expansion (Hammer et al. ¹⁹⁹⁸; Semino et al. ²⁰⁰⁰; Underhill et al. ²⁰⁰¹).

The three main subclades of haplogroup E3b (E-M78, E-M81, and E-M34) and the paragroup E-M35* are not homogeneously distributed on the African continent: E-M78 has been observed in both northern and eastern Africa, E-M81 is restricted to northern Africa, E-M34 is common only in eastern Africa, and E-M35* is shared by eastern and southern Africans (Cruciani et al. ²⁰⁰²). Given the strong geographic structuring observed for the four subsets of E3b within Africa, it is possible that different E3b lineages also have different frequency profiles in western Eurasia and that the evolutionary events underlying the introduction of E3b chromosomes in this area from Africa were not as simple (Rosser et al. ²⁰⁰⁰; Richards et al. ²⁰⁰²; Jobling and Tyler-Smith ²⁰⁰³) as previously proposed (Hammer et al. ¹⁹⁹⁸; Semino et al. ²⁰⁰⁰; Underhill et al. ²⁰⁰¹).

In the present study, we address the question of the origin and dispersal of haplogroup E3b subclades within and outside of Africa by analyzing 3,401 individuals from five continents. These include 1,510 individuals analyzed here for the first time for Y-chromosome markers (see also footnotes “b,” “c,” and “d” of table 1).

All of the subjects were typed for the YAP polymorphism (Hammer and Horai ¹⁹⁹⁵), and those who were YAP⁺ (haplogroup DE) were analyzed for the SRY₄₀₆₄ (Whitfield et al. ¹⁹⁹⁵), M35, and M215 mutations (Underhill et al. ^{2000, 2001}). Two subjects were found to carry the derived state at M215 and the ancestral state at M35. This modifies the topology of the E3 branch of the tree and the nomenclature of the corresponding haplogroups, as shown in figure 1 (note that “E3b” now refers to all haplogroups with the M215 derived state). Five hundred fifteen haplogroup E3b subjects were identified and further analyzed for the biallelic markers M34, M78, M81, M123, M281 (Underhill et al. ²⁰⁰⁰; Semino et al. ²⁰⁰²), and V6. The new V6 biallelic marker was discovered in the present survey in the TBL1Y gene by denaturing high-performance liquid chromatography analysis (primer sequences available on request). This marker identifies a subset of chromosomes previously assigned to E-M35* and now classified as “E3b1e” (fig. 1). No individual was found to carry the M281 mutation. We further typed 509 of the 515 E3b subjects for seven GATA STR (A7.1, A7.2, and A10 [White et al. ¹⁹⁹⁹]; DYS19, DYS391, and DYS393 [Roewer et al. ^{1992, 1996}]; and DYS439 [Ayub et al. ²⁰⁰⁰]) and four CA dinucleotide repeat (YCAIIa, YCAIIb, DYS413a, and DYS413b [Mathias et al. ¹⁹⁹⁴; Malaspina et al. ¹⁹⁹⁷]) polymorphisms. Both tetra- and dinucleotide microsatellites were used to reconstruct haplogroup-specific networks, through use of reduced-median and median-joining procedures (Bandelt et al. ^{1995, 1999}). The seven tetranucleotide repeat polymorphisms were also used for the estimation of the time to the most recent common ancestor (TMRCA) (Goldstein et al. ¹⁹⁹⁵; Slatkin ¹⁹⁹⁵; Thomas et al. ¹⁹⁹⁸) and the time since two populations split from a common ancestor (T_D estimator [Zhivotovsky et al. ²⁰⁰⁴]). For four of the tetranucleotide loci here used, locus-specific mutation rates based on father-son transmissions (μ_i) are not available (Kayser et al. ²⁰⁰⁰). Since both TMRCA and T_D estimations critically depend on the unknown parameter μ_i, we used the averaged effective mutation rate described by Zhivotovsky et al. (²⁰⁰⁴), which is based on a list of markers close to the one used here. CIs for the TMRCA were obtained as described by Scozzari et al. (²⁰⁰¹). It should be noted that uncertainties in the mutation rate, in the shape of the genealogy, and in the mutation process would increase the CIs. Since any two chromosomes sampled from two populations have a TMRCA older than the split between populations, and since we considered as null the variance of the ancestral population at the time of its splitting, the figures reported here for the T_D estimator represent upper bounds. In all of the analyses, except the networks, the YCAIIa, YCAIIb, DYS413a, and DYS413b dinucleotide repeats were not considered, since univocal assignment of phenotypic patterns to allelic series could not be obtained.

We obtained an estimate of 25.6 thousand years (ky) (95% CI 24.3–27.4 ky) for the TMRCA of the 509 haplogroup E3b chromosomes, which is close to the 30±6 ky estimate for the age of the M35 mutation reported by Bosch et al. (²⁰⁰¹) using a different method. Several observations point to eastern Africa as the homeland for haplogroup E3b—that is, it had (1) the highest number of different E3b clades (table 1), (2) a high frequency of this haplogroup and a high microsatellite diversity, and, finally, (3) the exclusive presence of the undifferentiated E3b* paragroup.

Our data show that haplogroup E3b appears as a collection of subclades with very different evolutionary histories. Haplogroup E-M78 was observed over a wide area, including eastern (21.5%) and northern (18.5%) Africa, the Near East (5.8%), and Europe (7.2%), where it represents by far the most common E3b subhaplogroup. The high frequency of this clade (table 1) and its high microsatellite diversity suggest that it originated in eastern Africa, 23.2 ky ago (95% CI 21.1–25.4 ky). The network of the E-M78 chromosomes reveals a strong geographic structuring, since each of the clusters α, β, and γ (fig. 2B) reaches high frequencies in only one of the regions analyzed. Cluster α is largely characterized by the otherwise rare nine-repeat allele at A7.1 (we found only 3 such alleles out of 800 E[xE3b1] chromosomes analyzed [present study; R.S., unpublished data]), often associated with the uncommon DYS413 24/23 pattern and its one-step neighbors. When compared with the other clusters in the network, it displays marked starlike features, with three central haplotypes accounting for 26% of the entire cluster. This cluster is very common in the Balkans (with frequencies of 20%–32%), and its frequencies decline toward western (7.0% in continental Italy, 7.4% in Sicily, 1.1% in Sardinia, 4.3% in Corsica, 3.0% in France, and 2.2% in Iberia) and northeastern (2.6%) Europe. In the Near East, this cluster is essentially limited to Turkey (3.4%). The relatively high frequency of DYS413 24/23 haplogroup E chromosomes in Greece (A.N., unpublished data) suggests that cluster α of the E-M78 haplogroup is common in the Aegean area, too.

Cluster β, characterized by the DYS413 23/21 pattern and the rare 10-repeat allele at DYS439, is common in northwestern Africa (14.0%), representing 80% of E-M78 chromosomes in that area. Outside this region, E-M78β was observed only in five European subjects.

All of the chromosomes in cluster γ (fig. 2B) are identified by the rare short 11-repeat allele at the DYS19 STR locus. We did not find this allele in >2,000 Y(xE-M78) chromosomes analyzed (present study; R.S., unpublished data), and it is reported in only 9 of 13,447 subjects analyzed for this marker in the European Y-STR reference database (Y-STR Haplotype Reference Database Web site). The cluster E-M78γ was found in eastern Africa at an average frequency of 17.7%, with the highest frequencies in the three Cushitic-speaking groups: the Borana from Kenya (71.4%), the Oromo from Ethiopia (32.0%), and the Somali (52.2%). Outside of eastern Africa, it was found only in two subjects from Egypt (3.6%) and in one Arab from Morocco.

The fourth cluster (cluster δ in fig. 2B) is present, albeit at low frequencies, in all of the regions analyzed (4.0% in eastern and northern Africa, 3.3% in the Near East, and 1.5% in Europe) and shows a notable microsatellite differentiation (fig. 2B). The two E-M78 chromosomes found in Pakistan, at the eastern borders of the area of dispersal of haplogroup E3b, also belong to cluster δ. On the basis of these data, we suggest that cluster δ was involved in a first dispersal or dispersals of E-M78 chromosomes from eastern Africa into northern Africa and the Near East. Time-of-divergence estimates for E-M78δ chromosomes suggest a relatively great antiquity (14.7±2.7 ky) for the separation of eastern Africans from the other populations. A later range expansion from the Near East or, possibly, from northern Africa would have introduced E-M78 cluster δ into Europe. However, given the low frequencies of E-M78δ, it seems to have contributed only marginally to the shaping of the present E-M78 frequency distribution in Africa and western Eurasia. Indeed, later (and previously undetected) demographic population expansions involving clusters α in Europe (TMRCA 7.8 ky; 95% CI 6.3–9.2 ky), β in northwestern Africa (5.2 ky; 95% CI 3.2–7.5 ky), and γ in eastern Africa (9.6 ky; 95% CI 7.2–12.9 ky) should be considered the main contributors to the relatively high frequency of haplogroup E-M78 in the surveyed area.

The present distributions of these clusters also suggest episodes of range expansions. Although E-M78β and E-M78γ show only modest levels of gene flow (from northern Africa to Europe and from eastern to northern Africa, respectively), the clinal frequency distribution of E-M78α within Europe testifies to important dispersal(s), most likely Neolithic or post-Neolithic. These took place from the Balkans, where the highest frequencies are observed, in all directions, as far as Iberia to the west and, most likely, also to Turkey to the southeast. Thus, it appears that, in Europe, the overall frequency pattern of the haplogroup E-M78, the most frequent E3b haplogroup in this region, is mostly contributed by a new molecular type that distinguishes it from the aboriginal E3b chromosomes from the Near East. These data are hard to reconcile with the hypothesis of a uniform spread of a single Near Eastern gene pool into southeastern Europe. On the other hand, they might be consistent with either a small-scale leapfrog migration from Anatolia into southeastern Europe at the beginning of the Neolithic or with an expansion of indigenous people in southeastern Europe in response to the arrival of the Neolithic cultural package. At the present level of phylogenetic resolution, it is difficult to distinguish between these possibilities.

E-M81 is very common in northwestern Africa, with frequencies as high as 80% (Bosch et al. ²⁰⁰¹; Cruciani et al. ²⁰⁰²; present study), but its frequency sharply declines on the continent toward the east, and the haplogroup is not found in sub-Saharan Africa. The distribution of E-M81 chromosomes in Africa closely matches the present area of distribution of Berber-speaking populations on the continent, suggesting a close haplogroup–ethnic group parallelism: in northwestern Africa, the lowest frequencies for this haplogroup have been reported in two Arab-speaking Moroccan populations (31% and 52% vs. 65%–80% in six Berber speaking groups from Morocco and Algeria [Bosch et al. ²⁰⁰¹; Cruciani et al. ²⁰⁰²; present study]); in Egypt, where Berbers are restricted to a few villages, E-M81 is rare (1.9%), and the southernmost finding of E-M81 chromosomes on the continent is that here reported in the Tuareg from Niger (9.1%), who also speak a Berber language. Outside of Africa, E-M81 has been observed in all the six Iberian populations surveyed, with frequencies in the range of 1.6%–4.0% in northern Portuguese, southern Spaniards, Asturians, and Basques; 12.2% in southern Portuguese; and 41.1% in the Pasiegos from Cantabria. It has been suggested (Bosch et al. ²⁰⁰¹) that recent gene flow may have brought E3b chromosomes from northwestern Africa into Iberia, as a consequence of the Islamic occupation of the peninsula, and that such gene flow left only a minor contribution to the current Iberian Y-chromosome pool. The relatively young TMRCA of 5.6 ky (95% CI 4.6–6.3 ky) that we estimated for haplogroup E-M81 and the lack of differentiation between European and African haplotypes in the network of E-M81 (fig. 2C) support the hypothesis of recent gene flow between northwestern Africa and Iberia. In this context, our data refine the conclusions of Bosch et al. (²⁰⁰¹) in two ways. First, not all of the E3b chromosomes in Iberia can be regarded as a signature of African gene flow into the peninsula: in our data set, 8 of 15 E-M78 chromosomes belong to cluster α, denoting gene flow from mainland Europe (see above). Second, and more importantly, the degree of the African contribution is highly variable across different Iberian populations: the proportion of haplogroup E chromosomes of African origin (E[xE3b], E-M35*, and E-M81) was <5% in three Spanish locations; 10.0% and 14.2% in northern and southern Portugal, respectively; and >40% in the Pasiegos (table 1). A relatively high frequency of E-M81 in a different sample of Pasiegos (18%) and non-Pasiegos Cantabrians (17%) has also recently been reported (Maca-Meyer et al. ²⁰⁰³). Such differences in the relative African contribution to the male gene pool of different Iberian populations may reflect, at least in part, the different durations of Islamic influence and introgression in different parts of the peninsula, as well as drift/founder effects for the small Pasiegos group.

The E-M123 clade was found in Ethiopia (11.2%), the Near East (3.7%), Europe (1.7%), and northern Africa (0.9%). In our data set, all the E-M123 chromosomes also carry the M34 mutation (E-M34), with the exception of one E-M123* subject from Bulgaria. This paragroup has been previously reported only in one individual from Central Asia (Underhill et al. ²⁰⁰⁰). Although the frequency distribution of E-M34 could suggest that eastern Africa was the place in which the haplogroup arose, two observations point to a Near Eastern origin: (1) Within eastern Africa, the haplogroup appears to be restricted to Ethiopia, since it has not been observed in either neighboring Somalia or Kenya (present study) or Sudan (Underhill et al. ²⁰⁰⁰). By contrast, E-M34 chromosomes have been found in a large majority of the populations from the Near East so far analyzed (Underhill et al. ²⁰⁰⁰; Cinnioğlu et al. ²⁰⁰⁴; Semino et al. ²⁰⁰⁴ [in this issue]; present study). (2) E-M34 chromosomes from Ethiopia show lower variances than those from the Near East and appear closely related in the E-M34 network (fig. 2D). If our interpretation is correct, E-M34 chromosomes could have been introduced into Ethiopia from the Near East. The high frequency of E-M34 observed for some of the Ethiopian populations could be the consequence of subsequent genetic drift, which can also explain the lower frequencies (2.3% [Underhill et al. ²⁰⁰⁰] and 4.0% [Semino et al. ²⁰⁰²]) reported for two large independent samples of Ethiopians. From the Near East, E-M34 chromosomes could also have been introduced into Europe, possibly by Neolithic farmers, but the paucity of E-M34 chromosomes in southeastern Europe (Semino et al. ²⁰⁰⁴ [in this issue]; present study) weakens this hypothesis. Indeed, as for E-M78δ chromosomes, introduction of E-M34 from Africa directly to southern-central Europe cannot be excluded at the present.

Haplogroup E-V6 was observed only in eastern Africa (8.9% in Ethiopia, with a single occurrence in both Somalia and Kenya), further testifying to the richness of E3b lineages in this region. Although no clear inferences can be drawn on the basis of the current E-V6 frequency distribution data, the V6 polymorphism may prove to be a useful marker for future microevolutionary studies in eastern Africa.

The paragroup E-M35* has been observed at high frequencies in both eastern (10.5%) and southern (15.2%) Africa, with rare occurrences in northern Africa and Europe (0.4% and 0.5%, respectively). The paragroup has a high microsatellite allele variance (0.63), comparable to that of the whole set of E3b(xE3b1*) chromosomes (0.53), suggesting that E-M35* is a collection of several lineages whose relationships to other E3b haplogroups remain to be established. Nevertheless, the observed distribution of E-M35* can shed light on the history of peopling of Africa. For example, we found E-M35* and E-M78 chromosomes in Bantu-speaking populations from Kenya (14.3%) but not in those living in central Africa (Cruciani et al. ²⁰⁰²), the area in which the Bantu expansion originated (Vansina ¹⁹⁸⁴). In agreement with mtDNA data (Salas et al. ²⁰⁰²), this finding suggests a relevant contribution of eastern African peoples to the gene pool of the eastern Bantu. Also, the extensive interpopulation E-M35* microsatellite diversity (fig. 2A) between Ethiopians and Khoisan indicates that eastern Africans and Khoisan have been separated for a considerable period of time, as has been suggested elsewhere (Scozzari et al. ¹⁹⁹⁹; Cruciani et al. ²⁰⁰²; Semino et al. ²⁰⁰²).