Reconciling migration models to the Americas with the variation of North American native mitogenomes

Dating and Tracing the Origin and Spread of B2a.

A recent study, performed at the control-region level, proposed that B2a mtDNAs mark a population expansion within the American Southwest (42), and tested the hypothesis that this expansion was because of the spread of groups of Uto-Aztecan farmers who introduced maize agriculture into this area from Mesoamerica ∼4 ka (46). Our analyses at the level of entire mitogenomes provide additional information that suggests a different scenario. The overall B2a coalescence time is at ∼12–13 ka, and we obtained a similar estimate when combining the previously published B2a control-region data with the control-region sequences of our mitogenomes. The network in Fig. S4 shows that B2a mtDNAs diverge from the root (central node) by 1.2 ± 0.3 control-region substitutions on average, which correspond to ∼11 ka [95% confidence interval (CI) 5.5–16.2 ka] when using the molecular clock of Soares et al. (47). Therefore, the onset of B2a should be predated to 11–13 ka in the late Pleistocene. This finding is further supported by our BSP analyses. Indeed, the main episode of population growth associated with B2a dates to 8–10 ka (Fig. 2); thus, it is likely too ancient to be associated with the farming/language dispersal hypothesis.

In search of alternative scenarios, one question immediately arises. Could B2a have been one of the many founder mtDNA haplogroups from Beringia? In the context of the Americas, this is a challenging issue, not only because the ancestral source populations in Siberia and Beringia no longer exists, but also because modern Asian and Siberian populations completely lack certain haplogroups. Therefore, the identification of founder mtDNA sequences from the analysis of modern individuals is based solely on the evaluation of two parameters: coalescence time and geographical distribution of the derived lineages from the postulated American founder.

Considering that B2 control-region haplotypes with the diagnostic B2a transition G16483A are not found either in present-day regions of former Beringia or further south than Central Mexico, and that our B2a mtDNAs coalesce at about 12–13 ka (more recently than the 15- to 18-ka time range reported for the 16 mitogenome founders identified to date), it is unlikely that B2a may have been a founder haplogroup from Beringia. Instead it is more likely that the diagnostic mutational motif of B2a evolved in situ (in North America) a few millennia after that B2 had already entered and spread along the double continent. If so, where did the first B2a mtDNA arise? One possibility is that it originated in the Pacific coastal regions from one of the local descendants of the B2 mitogenomes that had rapidly spread along the coast (48) and gave rise (at different stages of their spread) to other specific B2 subhaplogroups, such as B2i (49) and B2b (50) on their way from Beringia to South America. An alternative explanation is that this B2 mtDNA entered the American continent through the ice-free corridor with the same population groups that introduced the C4c and X2a mitogenomes (51, 52). Although at the moment there is no evidence that, in addition to the main Pacific coastal route, ancestral B2 mtDNAs also entered North America through the ice-free corridor, such a scenario cannot be completely ruled out. Further evidence for a B2a origin along the Pacific coast (possibly as far north as the Northwest Coast), sometime after that the Paleo-Indian migration wave had gone through, is provided by the Tsimshian (British Columbia) mitogenome that radiates directly from the B2a root (Fig. 1 and Table S1). Soon afterward, ∼10 ka, the deepest and most represented sublineages of B2a—which we termed B2a1, B2a2, and B2a4—arose possibly in the (greater) American Southwest and Mexico. It is worth noting that the main expansion signal for B2a occurs at approximately 9 ka (Fig. 2), predating both the farming diffusion linked to maize domestication (53) and the Uto-Aztecan culture (54, 55) and, according to Brown (56), even the formation of the oldest Native American language families. In this view, the spread of agriculture into the Southwest is probably marked by some of the younger B2a branches, such as B2a5, whose age and geographic distribution are both consistent with such event (Fig. S3, and Tables S1 and S2).

Finally, the high frequency of B2a in both Southern Athapaskan groups (Apache: 14.8%; Navajo: 15.1%) and its absence in all other Na-Dene groups from North America remain to be explained. The fact that none of the detected B2a sublineages are Southern Athapaskan-specific confirms that their B2a mtDNAs are most likely the result of female gene flow from surrounding non–Na-Dene populations over the last 500–1,000 y that elapsed since the first Athapaskan groups entered into the Southwest moving from their northern ancestral homeland (5, 40). When the Europeans arrived in the area, trade between the long-established Pueblo people and the Southern Athapaskans was well established. Spanish records speak of the Athapaskans traveling to the Pueblos, or living in the vicinity of them, to exchange bison meat, hides, and stone goods for maize and woven cotton (57, 58). The Navajo willingness to adopt the Pueblo lifestyle, as well as the Apache raiding and capturing of neighboring women, probably brought the B2a haplogroup, as well as others, into the Southern Athapaskan mtDNA gene pool, a scenario envisioned also by very early studies with classic genetic markers (28, 59) and mtDNA restriction fragment length polymorphism haplotypes (18).

Dating and Tracing the Origin and Spread of A2a.

Using a dataset combining both general-mixed and Native American populations, we have shown that the common ancestor of contemporary A2a mitogenomes originated much later than B2a, some 4–7 ka, in agreement with previous estimates (21). Such a temporal dissimilarity is also reflected in terms of their spatial distributions within the North American continent. Among the 130 populations considered in our control-region survey (Table S3), only seven showed both A2a and B2a mtDNAs. These populations include the two southern Athapaskan groups (Apache and Navajo), four general-mixed population samples from the American Southwest (Arizona and California) and northern Mexico (Chihuahua), and (perhaps) the Oneota specimens buried at the Norris Farms cemetery in Illinois. Most of the other A2a mitogenomes are found in the arctic regions of North America and Northeast Asia, with the three highest frequency peaks observed in northeastern Siberia, Alaska, and Greenland (Fig. 2). Most of these mtDNAs belong to the A2a3 (Fig. S3) and A2a2 subbranches (Fig. S1), even if for the latter clade, which is defined only by a coding-region marker, this is based only on a rather limited number of available mitogenomes (Fig. 1). A secondary peak was observed in the American Southwest, where most A2a mtDNAs belong to A2a4 and A2a5. However, A2a5 mtDNAs have also been observed in Alaska, the Pacific Northwest, and northwestern Canada, with a frequency peak in the Athapaskan Dogrib (26.7%) (Table S3). Even if a Siberian origin for A2a cannot be completely ruled out, these phylogeographic profiles support the scenario that one or more enclaves in Alaska (21) (or in the westernmost part of the Northwest Territories of Canada) might have functioned as incubators and ancestral homelands initially for A2a as a whole and later for its major sublineages. This finding is in agreement with the oldest archaeological evidence of the Arctic Small Tool tradition in the Americas, which is dated back 5.5 ky in the Kuzitrin Lake of Alaska (60). In the rather short period of 1–2 ky, the A2a2 and A2a3 variants, typical of contemporary circumpolar populations, evolved from the bearers of haplogroup A2a root, thus probably matching the beginnings of Paleo-Eskimo cultures. Our analyses reveal that the female effective size of these ancestral Paleo-Eskimo populations underwent a detectable bidirectional expansion temporally starting from approximately 4 ka (Fig. S2), covering all of the artic regions both in northern Canada and Greenland (on the American side) and in northeastern Siberia (on the Asian side), as a result of the back-migration process proposed by Tamm et al. (8). This expansion event is the same that led to the spread of: (i) mtDNA haplogroup D2a1, found in the northernmost Eskimos (and the single individual available from the extinct Paleo-Eskimo Saqqaq culture, dated to 3.4–4.5 ka; but hopefully other Paleo-Eskimo remains will be found and analyzed soon), Chukchi, Aleuts, Athapaskans, and possibly the Tlingit (21, 61); and (ii) Y-chromosome haplogroup Q1a*-MEH2, found in the Koryaks and Chukchi of Northeast Siberia, as well as in the ancient Paleo-Eskimo already mentioned above (12, 62, 63). Such circumpolar population dynamics neither involved the pan-American lineages that were spread all over the Americas by the Paleo-Indians, nor the subsequent Neo-Eskimo (Thule) groups that probably brought the D3 (and perhaps A2b1) mtDNAs from Alaska to South Greenland only one millennium ago (21, 61, 64).

The expansion of the native populations of Alaska and northwest Canada is also confirmed by the current geographic distributions of subhaplogroups A2a4 and, especially, A2a5, which in turn reflect the southward migration of the Athapaskans, dated less than 1 ka, and confirmed also by close Y-chromosome relationships between Athapaskans from the Subarctic and the Southwest (65). Subhaplogroup A2a5 is common in the Athapaskans (9.5%) and Tlingit (15.6%) from Alaska, but harbors a frequency peak among the Dogrib of Northwest Canada (26.7%). Among the Southern Athapaskans, A2a5’s average frequency is even higher, with a peak of 47.3% in the Apache (Fig. S3 and Table S3). Based on the available mtDNA data, the southward expansion of the ancestors of the Southern Athapaskans most likely occurred along the eastern side of the Rocky Mountains, but a parallel Pacific coastal route is also a possibility. In both cases, it is evident that the Paleo-Eskimos, who moved eastward into the American Artic (or returned to Northeast Siberia), the Tlingit, and the Athapaskans share an ancestral A2a mitogenome that, at different times and locations, differentially evolved within distinct North American population groups.

Correlating the Overall Variation of MtDNA Lineages in North America with Migration Models.

More than one migration wave was originally hypothesized to explain the introduction of distinct linguistic families, such as Eskimo–Aleut and Na-Dene, into the Americas. Recent Y-chromosome data supported the scenario that the Eskimoan- and Athapaskan-speaking groups were the result of two population expansions that occurred after the initial wave of American settlement (20). The old tripartite linguistic subdivision into Eskimo–Aleuts, Na-Dene, and Amerinds of Greenberg et al. (19) has been somehow recently revived to explain the pattern of nuclear diversity (over 364,000 SNPs) identified in 52 Native populations: 3 Eskimo–Aleutians, 1 Na-Dene (the North Athapaskan-speaking Chipewyan of the Northwest Territories of Canada), and 48 other Native American groups from North, Central, and South America (26). The whole-genome profiles were indeed interpreted as indicating that Native Americans descend from at least three streams of Asian/Beringian gene flow. However, we believe that such a stringent parallelism with the linguistic tripartite hypothesis should be regarded as challenging for at least two reasons. First, the reported data concern only a single Athapaskan population. Second, in the neighbor-joining tree [based on Fst distances and reported in figure 1C of Reich et al. (26)] the Chipewyan form a cluster with three non–Na-Dene groups (Cree, Ojibwa, and Algonquin) of the northern part of North America, the only other populations from that area included in the survey.

Taking into account these observations, could we reconcile in a different way the three-wave migration model suggested by nuclear data with the overall picture emerging from this and other recent studies which have assessed mtDNA variation at the level of entire mitogenome? As shown in Fig. 4, most of the contemporary Native American mtDNA variation (along the entire double-continent) stems from the same, presumably first wave of American settlement that was dated to 15–18 ka by applying the mitogenome molecular clock to different mtDNA founder lineages including the pan-American haplogroups A2 and B2 (52, 66, 67). In the present study we have investigated two of their subbranches (A2a and B2a) that exhibit different distributions in North America, the only geographic area in which all three language clusters proposed by Greenberg (29) are present. The estimated age of B2a and its localized geographic distribution indicate that the mutational motif of B2a most likely evolved in one of the (probably numerous) derived population groups that arose locally in North America along the trail of the coastal migration route. The early B2a carriers might have initially diffused along the northwestern Pacific coastal region, eventually resulting in a major in situ expansion in the (greater) American Southwest and northern Mexico (Fig. 4). The mutational motifs of the B2a subhaplogroups were instead fully completed more recently. Therefore, these subclusters, which arose at different times and locations, might provide new clues on the process of local linguistic and ethnic differentiation. For example, the B2a5 mtDNAs were probably carried in the Southwest ∼4–5 ka by the Uto-Aztecan farmers that introduced and diffused the maize agriculture. Such an early process of regional genetic differentiation was also evidenced by the pioneering analyses of classical nuclear markers (68⇓⇓–71).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

In this schematic overview, the 16 mtDNA founder lineages are associated with three major migratory events. Note that the location of the three spheres is approximate. In parentheses (and in italics) are indicated those founder lineages that are not yet sufficiently analyzed. Two additional events are indicated by stealth arrows. The first arrow corresponds to the recent southward spread of the Athapaskans (marked by A2a4 and A2a5). The second arrow marks the major in situ expansion of B2a.

On the other hand, A2a mtDNAs were certainly involved in the northernmost population movements that affected the Paleo-Eskimos and took place in the Arctic regions of North America (and Northeast Asia). Our phylogeographic data indicate that the present-day A2a mitogenomes coalesce to ∼4–7 ka and the variants detected among the Eskimo–Aleut speakers of both Northeast Asia and northern North America experienced the steepest population expansion ∼4 ka, thus confirming that this lineage is a good candidate marker for a separate and more recent migratory wave associated with the first Paleo-Eskimo settlement of the northernmost part of Canada and Greenland (Fig. 4), as also supported by the Y-chromosome prevalence (and diversity) of the Q1a6-NWT01 sublineage in the Western Canadian Inuit (Inuvialuit) (20). The ancestral A2a Paleo-Eskimo carriers undertook both a back-migration to Asia and an eastward path along the circumpolar region of Canada and Greenland. About 3,000 y later, from approximately the same northern regions, the Athapaskan populations moved to the Southwest carrying different A2a variants (A2a4 and A2a5). This information provides a scenario of how different languages might be associated with distinctive gene pools, not only at the level of the nuclear genome, but also at the level of mtDNA haplogroups or subhaplogroups.

Finally, there is no reason to assume that the additional (intermediate) migration wave, identified by Reich et al. (26) for the moment only in the Chipewyan, should be restricted to the Na-Dene speakers. In the American portion of their neighbor-joining tree, after the first split of the Eskimo–Aleut speakers and the far-northeastern Siberians, the Athapaskan-speaking Chipewyan and three Algonquian-speaking populations create a distinctive northern North American branch, which is well separated from the major branch encompassing all other more southern Native Americans. Thus, it is indeed conceivable that this intermediate branch indicates a migration wave that was restricted not linguistically, but rather geographically. This theory is also consistent with the observation that Chipewyans derive 90% of their nuclear genome from Paleo-Indian ancestors (26), which would imply the closest genetic relationship (and admixture) with surrounding non–Na-Dene native groups. Evidence on this issue has been recently provided by studies of mitogenomes from North America. At the moment we know that 7 of the 16 recognized mtDNA founding lineages are restricted to North America: A2b, D2a, and D3 are mostly limited to the northernmost regions (similarly to A2a), thus probably being involved only in the population dynamics of the artic areas (21, 38, 61); X2g has been identified until now only in a single Ojibwa subject; and C4c and X2a show peculiar geographic and ethnic distributions that require special attention (51, 52, 72). Phylogeographic analyses of these two latter haplogroups have led to the conclusion that they might have been carried to North America by Beringian populations, which arrived through the ice-free corridor between the Laurentide and Cordilleran ice sheets, either concomitantly or some time after the southward spread of the Beringian groups who were instead following the Pacific coastal route. Either the location of the ice-free corridor on the east side of the Rocky Mountains along the Mackenzie river basin, or a delayed arrival compared with the groups that were moving along the Pacific, would have limited the penetration into North America of the population(s) carrying haplogroups X2a and C4c. Indeed, the spatial distributions of X2a and C4c are not only limited to North America, but with an interesting exception (see below), they are restricted to the northern part of North America, encompassing numerous populations ranging from the Nuu-Chah-Nulth of Vancouver Island and the Yakima of Washington State on the Pacific coast, to the Algonquian-speakers, including the Mi’kmaq of the Atlantic regions of Canada (43, 45, 51, 52, 73). Most likely, the expansion of C4c and X2a in America occurred in the Great Plains region and our skyplot analysis suggests that the effective number of X2/C4c ancestral females began to increase from ∼8–10 ka (Fig. S2) right after the start of the stable, warm Holocene, which signed major postglacial expansions worldwide (74, 75).

At the moment there are no data concerning the distribution of X2 and C4c, as well as other haplogroups, in most of the northern Na-Dene populations, but the exception mentioned above concerns the southern Athapaskans, in which X2a has been indeed found (41, 43, 52). Overall, these observations suggest that haplogroups X2a and C4c might be present also in the northern Athapaskan groups to whom Apache and Navajo are closely related. If so, this finding would imply that X2a and C4c mtDNAs are shared between Na-Dene and non–Na-Dene of the northernmost regions of North America. Thus, the “intermediate” migration highlighted by nuclear data in the Chipewyan by Reich et al. (26) would be part of a larger-scale migratory event corresponding to the spread of the haplogroups X2a and C4c into North America, an event that did not affect only the ancestor of modern Na-Dene. An analogous differentiated male ancestry for northern Amerindians relative to more southern Native American groups is also supported by the diversity of Y-chromosome lineages (76⇓–78). This is also the highest likelihood scenario obtained by applying a copying model simulation on 2,540 linked SNPs across 32 autosomal regions (79, 80), by analyzing the diversity of X-chromosome dys44 haplotypes (10) and even by some early classic genetic data on albumin showing that the Naskapi (AL*Naskapi) variant was shared and limited to Athapaskan and Algonquian-speaking populations (59).

As shown (and dated) in the present study, the distinct, possibly intermediate, genetic input marked on the matrilineal side by mtDNA haplogroups X2a and C4c and the subsequent population expansion extensively predated the spread of A2a in the circumpolar regions of North America. Moreover, the observation that A2a, among all Amerindian populations of northern North America, is very common and almost exclusively present in the Na-Dene, with limited occurrences in other local Amerindian groups, suggests a distinctive, at least dual, genetic contribution to the formation of the Na-Dene: (i) an “older” Beringian genetic component arrived through the interior corridor and shared only with surrounding non–Na-Dene populations of northern North America, and (ii) a more recent genetic input of Alaskan origin maternally marked by mtDNA haplogroup A2a.

As a final remark, it is evident that the arrival of the first American founders, when the territory was empty, left the greatest genetic mark (81), but the original maternal and paternal makeups of North American Natives were subsequently reshaped by additional streams of gene flow and local population dynamics, making even a three-wave view by far too simplistic.