Ancient DNA indicates human population shifts and admixture in northern and southern China

Genome-wide data from Neolithic East Asians shows their close relationship to East Asians today

To gain insights into the genetic history of East Asians during the Neolithic, we sampled genetic material from ancient individuals across East Asia dating to 9,500-300 calibrated years before present (9.5-0.3 kBP, Fig. 1, A and B) (14). In northern East Asia (defined as north of the Qinling-Huaihe line, Fig. 1A), we sampled from the northern Chinese provinces of Inner Mongolia and Shandong. In southern East Asia, we sampled from the southern Chinese province of Fujian in mainland East Asia, as well as two Taiwan Strait islands (Fig. 1A, Table 1, and table S1). We used large scale ancient nuclear DNA capture techniques (15) to enrich for endogenous DNA at 1.2 million single nucleotide polymorphisms (SNPs) (16) in 26 individuals. 24 individuals passed our analysis filters. These filters include (1) identification of characteristic ancient DNA damage signatures suggesting the presence of endogenous DNA (17), (2) assessing modern human contamination rates to determine for each sample whether to include all fragments (≤3% contamination) or only those with the characteristic damage signature (>3% contamination) (18, 19), and (3) exclusion of individuals where a contamination rate could not be estimated (14). In total, we obtained genetic information from 24 individuals from 11 sites, sequenced to between 0.01 - 7.60x fold coverage at the targeted SNPs (Table 1), with 16 individuals sampled from southern East Asia and 8 individuals sampled from northern East Asia.

•  Download high-res image

•  Open in new tab

•  Download Powerpoint

Table 1 Ancient individuals sampled in this study.

View this table:

• View popup

• View inline

To determine whether any individuals possessed ancestry deeply diverged from present-day East Asians, we first asked the extent to which they shared ancestry with ancient Asians previously sampled who separated early from the common ancestor of East Asians (‘early Asians’, e.g., 8,000-4,000-year-old Hòabìnhians from Laos and Malaysia in Southeast Asia (7), the 3,000-year-old Ikawazu individual from Japan (7), and the 40,000-year-old Tianyuan individual from Beijing, China (20) (table S1). In a principal component analysis (PCA) including ancient and present-day Asians (14), all Neolithic East Asians cluster with populations of East Asian ancestry (Fig. 1C and fig. S1), who include present-day East Asians and Neolithic Asians from Siberia, Tibet, Southeast Asia, and the Southwest Pacific who possess primarily East Asian-related ancestry (8–10, 21). Notably, this includes Early Neolithic southern East Asians (Qihe, Liangdao), who possess cranial morphology that clustered with ‘early Asians’ (13). Thus, our results fail to support the version of the ‘two layer’ model in which these individuals are included in the first layer. Consistent results were obtained in an outgroup f3-analysis, where Neolithic East Asians share more genetic similarity with Neolithic Siberians, Tibetans, and Southwest Pacific Islanders (f3 = 0.28-0.32) than with ‘early Asians’ (f3 = 0.25-0.26, fig. S2).

A direct comparison of Neolithic East Asians to present-day East Asians and ‘early Asians’ in a symmetry test shows that Neolithic East Asians tend to be more closely related to present-day East Asians than to any ‘early Asian’, i.e., most f4(Mbuti, X; present-day East Asian, ‘early Asian’) values are significantly less than zero (Han: −22.1<Z<−4.4, table S2) and most f4(Mbuti, 'early Asian'; present-day East Asian, X) values are ~0 (Han: −2.4<Z<2.6, table S2). Thus, the newly sampled individuals, including those from Early Neolithic southern East Asia, are genetically closest to modern populations of East Asian ancestry.

We additionally investigated patterns of archaic admixture in the Neolithic East Asians and found them to be similar to present-day East Asians (table S3). Neolithic East Asians also do not show consistent connections with a 35,000-year-old individual from Belgium, nor do they show connections to any South Americans – patterns that differ from Tianyuan (14, 20) but are similar to present-day East Asians (fig. S3). Previous ancient DNA studies found deeply divergent ancestries in pre-Neolithic Asia (7, 20) associated with a potential ‘first layer’ of populations in Asia (13). However, populations from the Early Neolithic in both northern and southern China show different patterns that suggest they do not belong to the ‘first layer’ and already carry ancestry primarily associated with present-day East Asians.

Population division between northern and southern East Asians in the Early Neolithic

In the PCA, our newly sampled individuals separate along geographic lines. Coastal Neolithic southern East Asians cluster together and close to present-day southern East Asians, while coastal Neolithic northern East Asians cluster together and close to present-day northern East Asians (Fig. 1C). These results suggest population structure between northern and southern East Asia since the Early Neolithic.

To determine genetic relationships in Neolithic East Asia, we first used a pairwise comparison of outgroup f3-values (22) to assess their relationship to other Neolithic populations with East Asian-related ancestry (14). Neolithic southern East Asians share high genetic similarity to each other, Neolithic Southeast Asians, and Austronesian-related islanders from the Southwest Pacific (Fig. 1D), a pattern also observed in the PCA (Fig. 1C). In an f4-analysis (fig. S4A), Neolithic southern East Asians and the 3,000-year-old Austronesian-related islanders from the Southwest Pacific (10) consistently share a closer genetic relationship with each other than with coastal and inland Neolithic northern East Asians, Siberians (8, 9), and Tibetans (21), though some Late Neolithic southern East Asians share a connection to northern East Asians (fig. S4B), which we below suggest is related to admixture. In a maximum likelihood phylogeny allowing migration events (23), Neolithic southern East Asians group together relative to Neolithic northern East Asians, Siberians, and Tibetans (Fig. 2A and figs. S5 and S6) in 99.3-100% of the bootstrap replicates (bs). Thus, in mainland southern East Asia and Taiwan Strait islands, we find shared southern East Asian ancestry that differs from that observed in Neolithic northern East Asians – a pattern that persists even with the evidence of admixture that we describe below.

•  Download high-res image

•  Open in new tab

•  Download Powerpoint

In northern East Asians dating to the Early Neolithic, we find shared ancestry not found in Early Neolithic southern East Asians, but present in Neolithic Siberians and Tibetans. For instance, in a maximum likelihood phylogeny, Neolithic northern East Asians, Siberians, and Tibetans group together relative to Neolithic southern East Asians and Austronesian-related islanders (bs = 95.3-99%, Fig. 2A and figs. S5 and S6), and they also share high genetic similarity with each other in an outgroup f3-analysis (Fig. 1D). In particular, coastal Neolithic northern East Asians clustered together, while Yumin, an inland Neolithic northern East Asian, clustered with Neolithic Siberians from the Eastern Steppe and the Primorye region of Far East Russia (8, 9). Most Neolithic northern East Asians and Siberians share a closer genetic relationship to each other than to Neolithic southern East Asians (figs. S7 and S8). In a model developed using all f2, f3, and f4-statistics for populations included in the model (Admixture Graph, see (14)), we similarly observe that northern and southern East Asian-related ancestry separate into two lineages (Fig. 2B and figs. S16 to S18).

Though the Neolithic is highlighted by distinct northern and southern East Asian-related ancestries, gene flow also impacted populations during this time. In the PCA, the inland Early Neolithic northern East Asian Yumin does not cluster with other northern East Asians as Neolithic Siberians and coastal northern East Asians do, but rather, Yumin falls in between northern and Tibetan present-day East Asians (Fig. 1C). Additionally, two patterns improve the fit of the maximum likelihood phylogeny (Fig. 2A and figs. S5 and S6). First, Upper Paleolithic northern Siberians influenced the ancestry of Neolithic Siberians (i.e., Yana, Fig. 2A), likely mediated by populations in Siberia closely related to Native Americans (Paleosiberians, table S1) as previously observed (9) and confirmed in f4-analyses showing Paleosiberian connections primarily to Neolithic Siberians (table S4). Second, some coastal Early Neolithic northern East Asians show affinities to coastal Late Neolithic southern East Asians when compared to more inland Neolithic populations in Asia (northern East Asian, Siberian, and Tibetan, figs. S7 and S8). Coastal Early Neolithic northern East Asians show these connections to Early Neolithic southern East Asians as well (fig. S8B and fig. S9). These connections show that the population relationships among coastal northern and southern Neolithic East Asians cannot be easily clarified without the inclusion of admixture (Fig. 2, B to D, and fig. S10).

Growing influence of northern East Asian ancestry in southern East Asia

In a spatiotemporal analysis, we next assessed differences in East Asian ancestry between the Neolithic and today. A major feature of Neolithic East Asian ancestry described above is a northern ancestry, represented by Neolithic northern East Asians, Siberians, and Tibetans, and a southern ancestry, represented by Neolithic southern East Asians and Austronesian-related islanders. A test of genetic differentiation (Fst) shows that coastal Neolithic northern East Asians and coastal Neolithic southern East Asians are more highly differentiated from each other (Fst=0.042, sd=0.004) than are present-day northern and southern East Asians (Fst=0.023, sd=0.01, fig. S11). This difference indicates that East Asians today are more genetically homogeneous than Neolithic East Asians.

To determine what factors led to decreased genetic differentiation between present-day northern and southern East Asians relative to the Neolithic, we used an f4-analysis to set up a symmetry test (Fig. 3, A to C), where we compared ancient and present-day individuals against the oldest northern (Bianbian) and southern (Qihe) East Asians sampled. That is, we tested f4(Mbuti, X; Bianbian, Qihe) to determine northern and southern East Asian affinities across time and space. In the Early and Late Neolithic, there is clear spatial separation, with northern populations sharing a closer relationship to Bianbian and southern populations sharing a closer relationship to Qihe (Fig. 3, A and B, and table S5A). This pattern persists when we substitute other Early Neolithic southern East Asians for Qihe and other Early Neolithic northern East Asians for Bianbian (fig. S12 and table S5, B and C). We then used qpAdm (15) to develop mixture models (Fig. 3, D to F) to estimate ancestry proportions in Neolithic populations (14), where Neolithic northern and southern East Asians each belong to a distinct ancestry associated with their location in northern or southern East Asia (Fig. 3, D and E, and table S6A).

•  Download high-res image

•  Open in new tab

•  Download Powerpoint

In contrast, patterns in present-day East Asians show that the main factor reducing genetic differentiation in present-day East Asians is increased northern East Asian-related ancestry in southern East Asia. In present-day East Asians, we observe a dramatic change where instead of affinities based on location, all East Asians (including mainland southern East Asians) share an affinity with Neolithic northern East Asians, as shown by f4(Mbuti, present-day East Asian; Qihe, Bianbian)>0 (3<Z<8.8, Fig. 3C and table S5A). Estimates of ancestry proportions in our mixture models show that this is due to increased northern East Asian-related ancestry in southern mainland East Asia, with estimates of northern East Asian-related ancestry ranging from 21-55% (Fig. 3F, fig. S13, and table S6B). Southern East Asian ancestry also extends north as well, as it is found in Han populations from northern China (36-41%, fig. S13 and table S6B), as well as some northern East Asians (e.g., Xibo, Korean, 35-36%, Fig. 3F and table S6B). Paleosiberian-related ancestry from Siberia also greatly impacts recent East Asians, with the exception of a 300-year-old individual from coastal mainland southern East Asia, island populations, and Tibetans (Fig. 3F and table S6). This lack of Paleosiberian-related ancestry in the peripheral edges of East Asia suggests that different types of gene flow from north to south occurred in East Asia.

While we are unable to estimate the timing of gene flow events, patterns in the f4-comparisons suggest that population movement may have already started impacting East Asians by the Late Neolithic. In the test f4(Mbuti, Bianbian; Kolyma, Neolithic southern East Asians), we would expect similar results for all Neolithic southern East Asians if none of them had northern East Asian ancestry. Instead, the Late Neolithic southern East Asians share a connection to the coastal northern East Asian Bianbian that Early Neolithic southern East Asians do not share (fig. S14). Other tests of admixture support this finding: Admixture Graph (Fig. 3, C and D) and ADMIXTURE (fig. S10) results both estimate northern East Asian-related ancestry in Late Neolithic southern East Asians that is reduced in Early Neolithic southern East Asians (14).

We also investigated whether the northern East Asian affinity is closer to inland (i.e., Yumin) or coastal (i.e., Bianbian) northern East Asians. In an f4-test of symmetry, all Neolithic East Asians and coastal Neolithic Siberians share a closer relationship to Bianbian rather than Yumin, whereas all inland Neolithic Siberians and Tibetans do not (fig. S15 and table S5D). This suggests that the northern East Asian ancestry found in all present-day mainland East Asians is primarily related to populations along the lower reaches of the Yellow River. These observations are consistent with archaeological and historical studies that argue for an origin of the Han ethnic group in northern China, along the Yellow River (24).

Proto-Austronesian origins in southern China and coastal connections

Austronesian-speaking populations extend from Taiwan to the Southwest Pacific and to Madagascar. Mainland southern China has been proposed to be the likely origin of the proto-Austronesian population that entered Taiwan, based on geographic proximity across the Taiwan Strait and the occurrence of similar artifacts at archaeological sites from both regions (11, 22, 25, 26). Furthermore, genetic modeling with mitochondrial DNA (mtDNA) genomes from present-day Southeast Asians (27) also suggest a southern Chinese origin. The mtDNA of the 8,320-8,060-year-old Liangdao1 individual belongs to haplogroup E1, which is common in Austronesian-speaking populations in Taiwan, the Philippines, and Indonesia today, and is most similar to mtDNA found in the aboriginal Formosans from Taiwan (28).

Our data also support a proto-Austronesian origin in mainland southern China. Neolithic southern East Asians consistently fall on a cline leading to present-day Austronesians rather than other present-day southern East Asian populations in a PCA (Fig. 1C), and they share a connection with present-day Austronesians (Ami) relative to other present-day southern East Asians, e.g., f4(Mbuti, coastal sEastAsia_EN/LN; Dai, Ami)>0 (5.2<Z<12.7, table S7A). This connection to present-day Austronesians extending across all Neolithic samples tested from mainland southern China and Taiwan Strait islands supports a coastal southern China origin for proto-Austronesians (29).

We further investigated connections between the ancient samples and 3,000-year-old Southwest Pacific islanders from Vanuatu (10), who share a close relationship to present-day Austronesians. The Neolithic samples from mainland southern China and the Taiwan Strait islands group with the ancient Vanuatu samples in a maximum-likelihood tree (b.s. = 82.3-100%, Fig. 2A and fig. S5) and share high genetic similarity with the Vanuatu in an outgroup f3-analysis (Fig. 1D). Moreover, a direct f4-comparison demonstrates that Late Neolithic southern East Asians share a close genetic relationship with these Austronesian-related Southwest Pacific islanders (fig. S4A), with no excess connections from either group to Neolithic northern East Asians (fig. S4B). These results further support the proposed connection between Neolithic southern East Asians and proto-Austronesians (29).

A lack of genetic isolation between coastal populations can be observed all along the coast of East and Southeast Asia. While most Neolithic Southeast Asians are a mixture of Hòabìnhian-related and southern East Asian-related ancestry, a 4,000-year-old population from Vietnam shares an especially close relationship to coastal Late Neolithic southern East Asians (table S7B). Strikingly, these coastal connections extend further north. A 2,700-year-old Jōmon individual from the Japanese archipelago that separated early from present-day East Asians was suggested to share genetic affinity to Hòabìnhians (7), but comparisons including Neolithic East Asians do not show this pattern (table S8). Instead we find that the Jōmon individual shows affinities to several coastal Neolithic populations in Siberia, as well as southern East Asia (table S9). The patterns demonstrated here show that coastal regions were areas of interconnectivity and gene flow rather than isolation. These affinities between populations along the eastern coast of Asia (and island populations off the coast) that are not shared by more inland Asian populations suggest that interactions along marine-related environments played an important role in the prehistory of coastal Asia.