Previous studies revealed that Hg I reached a frequency of ~40%–50% in two distinct regions—in Nordic populations of Scandinavia and, in southern Europe, around the Dinaric Alps—each showing different background STR modal haplotypes (Semino et al. 2000; Passarino et al. 2002; Barać et al. 2003). In addition, subclade I-M26 (Underhill et al. 2000) reaches a very high frequency (~40%) (Semino et al. 2000; Passarino et al. 2001; Francalacci et al. 2003) in Sardinia, particularly in the “archaic area” (Cappello et al. 1996; Zei et al. 2003), and is associated with the peculiar YCAIIb-11 allele (Ciminelli et al. 1995; Caglia et al. 1997; Quintana-Murci et al. 1999; Malaspina et al. 2000; Scozzari et al. 2001). Overall, these observations suggest that Hg I could have played a central role in the process of human recolonization of Europe from isolated refuge areas after the LGM and suggest the likelihood that a comprehensive phylogeographic study should be able to localize the in situ origin and spread of principal male founders.

In the present study, the M170 A→C transversion, which defines Hg I, was assessed in a total of 7,574 subjects, including 6,095 Y chromosomes from 48 European populations and 1,479 individuals from 12 populations of surrounding regions (the Near East, Macaronesia, Central Asia, and the Caucasus). The results are reported in table 1, together with 407 additional members of Hg I out of 3,859 Y chromosomes extracted from the literature. Of the 1,104 Y chromosomes from the present study (1,060 from European subjects and 44 from the adjacent regions) that showed the derived M170 C-allele, 236, representative of the entire collection, were first examined for all the Hg I mutations known to date—namely, M21, M26, P37, M72, M223, M227, M253, M258, M284, and M307, whose phylogenetic relationships are illustrated in figure 1A. Genotyping was performed in a hierarchical way, and methods are provided in the legend to figure 1. The M258 and M307 mutations were observed in all of the Hg I and I1a Y chromosomes, respectively, whereas the M21, M72, and M284 mutations were not found. Thus, all of these were subsequently omitted in the remainder of the survey. To evaluate the differentiation of the I subclades (fig. 2), 533 Hg I Y chromosomes from 34 population samples were examined for the microsatellites DYS19, DYS388, DYS390, DYS391, DYS392, and DYS393 (Roewer et al. 1992, 1996; Thomas et al. 1998, 1999); 183 of the same chromosomes (from 20 populations) were also genotyped for the DYS389 (Roewer et al. 1996), YCAIIa, and YCAIIb (Mathias et al. 1994) loci; and 128 were also genotyped for the 49a,f system (for a review, see Poloni et al. [1997]). The results obtained were used to construct the network illustrated in figure 2. The STR and 49a,f data are available upon request.

Hg I accounts for more than one-third of paternal lineages in two distinct regions of Europe: among Scandinavian populations and in the northwestern Balkans (table 1; fig. 1B). Relatively high frequencies are also characteristic of some French regions, like Low Normandy and southern France. Interestingly, a lower frequency of haplogroup I distinguished the Baltic-speaking Latvians (7.0%; 90% CI 3.8%–13.2%) from their northern neighbors, the Finnic-speaking Estonians (18.6%; 90% CI 14.6%–23.4%). Similar cases of even more significant frequency change over a short geographic distance occur between the southern Slavic-speaking populations and their adjacent neighbors: namely, the Slovenians versus the northern Italians (38.2%; 99.9% CI 19.5%–60.2% vs. 4.6%; 99.9% CI 1.4%–11.7%), and Macedonians versus Greeks (30.0%; 95% CI 19.1%–43.8% vs. 13.8%; 95% CI 10.1%–18.5%).

As reported in table 1 and illustrated in figure 1A, three subhaplogroups, defined by the M253 (I1a), P37 (I1b), and M223 (I1c) markers, account for 95% of the Hg I Y chromosomes, and the remaining subhaplogroups belong to paragroup I*.

Subhaplogroup I1a is mostly found in northern Europe, with its highest frequencies in Scandinavian populations, where it accounts for 88%–100% of Norwegian, Swedish, and Saami M170 lineages. I1a has a decreasing gradient from its peak frequency in Scandinavia toward both the Urals and the Atlantic periphery (fig. 1C). Its I1a4 subclade has been found mainly as single observations scattered in eastern and southeastern Europe. Since the Scandinavian Peninsula was completely depopulated during the LGM, two main European refugia, the Iberian Peninsula/southern France and the Ukraine/Central Russian Plain (Dolukhanov 2000), can be considered as possible source regions of Scandinavian I1a chromosomes. Although Hg I occurs in the Ukraine at a higher incidence (22%) than in western Europe (11% in France), the virtual absence in Scandinavia of the most represented eastern European I1b* subclade, together with the higher I1a microsatellite diversity background, point to western Europe as the source of the Scandinavian I1a chromosomes, since the STR diversity of I1a decreases from western to eastern Europe, showing a significant negative correlation (r=-0.63; significance level 0.99) with longitude. In France, I1a is the leading subclade, representing 45% of the Hg I lineages, which, however, occur in a focal rather than clinal pattern. Hg I is more frequent in Low Normandy (I = 23.8%; I1a = 11.9%) and southern France (I = 15.8%; I1a = 5.3%), whereas it has a much lower occurrence in the Poitier and Lyon interior regions (I = 4.0%; I1a = 2.0%). Interestingly, subclade I1a shows a distribution similar to the second PC of the synthetic maps based on classical genetic markers (Cavalli-Sforza et al. 1994) and reveals a significantly positive correlation with mtDNA haplogroups V and U5b (r_V=0.47; r_U5b=0.60; significance level 0.999), which have been suggested to mark a postglacial population expansion from Iberia (Torroni et al. 1998, 2001; Tambets et al. 2004).

A different scenario has to be envisioned for subhaplogroup I1b*, which is the most frequent clade in eastern Europe and the Balkans. It reaches its highest incidences in Croatia (31%) and Bosnia (40%), encompassing almost 80%–90% of I (table 1). In western Europe, its subclade I1b2 (M26) (fig. 1F) (Semino et al. 2000; Bosch et al. 2001; Capelli et al. 2003; Maca-Meyer et al. 2003) is found at a very low frequency (<5%), except in Sardinia (41%), Castile (19%) (Flores et al., in press), Bearnais (8%), and in the Basques (6%). Although subclade I1b2 and the paragroup I1b* (the latter present at marginal frequencies) co-occur west of the Italian Apennines, only I1b* is present east of the Adriatic. I1b* Y chromosomes rapidly dissipate west of the Balkans—they are virtually absent among Italians, Germans, French, and Swiss (table 1; fig. 1E)—but extend eastward at notable frequencies among Slavic-speaking populations. This finding suggests that, similar to I1a, I1b* also may have expanded from a glacial refuge area. However, this area was most likely located in eastern Europe or the Balkans.

In central and eastern Europe, subhaplogroups I1a and I1b show overlapping frequency gradients, although with opposite post-LGM spreading. The divergent distributions of I1b2 and I1b* suggest that their separation occurred before the LGM and that the M26 mutation arose in a I1b Y chromosome from western Europe, most likely in a population in Iberia/southern France. The exceptionally high incidence of I1b2 in the archaic zone of Sardinia (Cappello et al. 1996; Zei et al. 2003) can be explained by the presence of I1b2 chromosomes among the first humans who colonized the island, ~9,000 years ago, followed by isolation and genetic drift. The extremely low frequency of I1b2 in the Scandinavian Peninsula, where the “western European” I1a Y chromosomes account for the large majority of Hg I, suggests, in addition, that the ancestral western European population(s), characterized by the M26 mutation, probably played a minor role in the colonization of that region. A geographic and genetic subdivision within the broad western refuge area, together with differences in initial sample size, genetic drift, and expansions, could also explain the quite different distribution of Hg I subhaplogroups with respect to the west-east decreasing gradient displayed by R1b, the most frequent subhaplogroup in western Europe.

The high STR diversity of the I1b* lineages in Bosnia supports the view that the P37 SNP might have been present in the Balkan area before the LGM, as previously proposed by Semino et al. (2000). Diversity h values based on STR haplotypes for I1a are highest near Iberia but vary substantially in different populations (table 2). For I1b*, conversely, the highest h values are in the Balkan populations—among Bosnians (0.93) and Croats (0.85)—coinciding with the area of its frequency peak, but equally high values were also observed for Czechs and Slovaks (0.90). The lowest h values of I1b* were detected among Turks (0.76) and in our Moldavian sample (0.41).

Subhaplogroup I1c (fig. 1D) covers a wide range in Europe, with the highest frequencies (~5%–12%) in northwestern Europe and lower frequencies elsewhere. Its geographic and linguistic correlations across the continent were insignificant. However, the 49a,f system and the microsatellites YCAIIa-YCAIIb reveal that I1a and I1c harbor an identical compound haplotype (49a,f Ht12/YCAIIa-21/YCAIIb-19), which is different from those of I1b* (49a,f Ht10/YCAIIa-21/YCAIIb-21) and I1b2 (49a,f Ht12/YCAIIa-21/YCAIIb-11). These results indicate that subhaplogroups I1a and I1c may be part of a single monophyletic clade whose deep biallelic mutations are still undefined and that they probably share a common history of expansion. This scenario is also supported by the high positive correlation between the geographic distributions of I1a and I1c (r=0.75 when Fennoscandia is excluded; significance level 0.999).

Anatolia is at the easternmost fringe of the spread of haplogroup I, where it is found at higher frequencies in the regions that are geographically closer to Europe (Cinnioğlu et al. 2004). This observation, combined with a low haplotype diversity in Turkey plus exact haplotype matches with Europe, suggests that haplogroup I Y chromosomes in Turkey are due to migrations from Europe, as has been argued for a fraction of the Turkish mtDNAs (Richards et al. 2000).

A temporal interpretation of the phylogeography based on the results of the STR length variation in the individual subhaplogroups of I (Zhivotovsky et al. 2004) is reported in table 3. The age of STR variation for I* was estimated as 24,000±7,100 years, a value that is very close to the population divergence time (23,000±7,700 years). This finding supports the earlier suggestion that haplogroup I originated from a pool of European pre-LGM, middle Upper Paleolithic Y chromosomes (Semino et al. 2000). Our time estimates hint that its initial spread in Europe may be linked to the diffusion of the largely pan-European Gravettian technology ~28,000–23,000 years ago (Djindjian 2000; Perles 2000). On the other hand, these values represent the lower limit of the age of M170 mutation. The precedent mutation (M89) (fig. 1A) defines the overarching superhaplogroup F, whose representatives span the entire non-African gene pool, likely predating the peopling of Europe (some 40,000–50,000 years ago). Potentially more informative are the estimates of subclade divergence times. Thus, it appears that I1a, I1b, and I1c all diverged from I* in the Late Upper Paleolithic/Mesolithic period (table 3), possibly during the recolonization of Europe after the LGM. However, the expansion phase of I1a and I1b, displaying contrasting phylogeographies, seems to have occurred later, around the early Holocene. Only the less frequent subclade I1c, spread thinly over much of Europe, from Mordvin in the Volga region to southern France (table 1; fig. 1D), shows a somewhat earlier age for its STR variation (table 3), suggesting that the corresponding mutation arose earlier.