We collected a total of 93 peripheral blood samples from unrelated Manchu individuals aged over 58 in different villages in the Xinbin Manchu Autonomous County, Liaoning Province, northeast China (Figure 1). These samples were collected randomly from unrelated participants whose parents and grandparents are indigenous people and have a non-consanguineous marriage of the same ethnical group for at least three generations. The ethnicities of all participates were used as their self-declaration based on their family migration history and corresponding family records. Our study and sample collection were reviewed and approved by the Ethics Committee of Jinzhou Medical University (JZMULL2021010) and followed the recommendations provided by the revised Declaration of Helsinki of 2000. Verbal and written informed consent was obtained from all participants. We used the Infinium^® Global Screening Array (GSA) covering 659,509 SNPs to genotype targeted SNPs in Manchus. Genotype calling was carried out following the default parameters. Raw data were initially filtered using PLINK 1.9 (Purcell et al., 2007) based on our predefined threshold of the genotyping success rate, missing site rates, minor allele frequency, and Hardy–Weinberg equilibrium (−maf:0.01, −hwe:0.0001, −mind:0.01, and −geno:0.01). A final dataset with 293,307 SNPs was used to perform the following population genetic analysis.

We merged our newly genotyped data with previously published modern and ancient population data from the Affymetrix Human Origins (HO) Array dataset (Wall and Yoshihara Caldeira Brandt, 2016) and the 1240K dataset from the Allen Ancient DNA Resource (AADR).³ The included modern genome-wide SNP reference data were collected from nine language families or groups (Tungusic, Mongolic, Turkic, Sinitic, Tibeto-Burman, Austronesian, Austroasiatic, Tai-Kadai, and Hmong-Mien) in China, South Siberia, and Southeast Asia (Patterson et al., 2012; Lipson et al., 2018; Liu et al., 2020; Kutanan et al., 2021; Wang et al., 2021a). Ancient reference populations were also collected from China and surrounding countries (Ning et al., 2020; Yang et al., 2020; Wang et al., 2021a). We also included recently published population data from the Guizhou Manchu people (Chen et al., 2021). Finally, two combined datasets covering 44,476 SNPs in the merged HO set and 127,435 SNPs in the merged 1240K set were used in the subsequent analysis.

Principal component analysis (PCA) was carried out using smartpca, part of the EIGENSOFT package (Patterson et al., 2006). The additional parameter of the lsqproject was set as YES (lsqproject: YES), and other parameters were used as the default setting. A total of 70 present-day and ancient worldwide populations were selected for PCA. Ancient people were projected onto the background of modern genetic variations. The PCA result was plotted by R software.⁴ We applied ADMIXTURE (Alexander et al., 2009) to conduct the model-based clustering analysis based on the merged HO dataset, which consists of 1,385 individuals from 89 populations. Prior to the analysis, we pruned SNPs in strong linkage disequilibrium with each other using PLINK 1.9 (Purcell et al., 2007) with the parameters “-indep-pairwise 200 25 0.4.” We run ADMIXTURE with the K values (number of assumed ancestral sources) ranging from 2 to 20. An optimal value of K was selected using 10-fold cross-validation errors, which is shown in Supplementary Table 1. The results of admixture analysis were visualized using AncestryPainter (Feng et al., 2018).

We calculated pairwise F_ST genetic distance between Liaoning Manchu and other included modern and ancient reference populations using smartpca of EIGENSOFT package (Patterson et al., 2006) (fstonly: YES, fsthiprecision: YES).

All f-statistics were calculated using ADMIXTOOLS (Patterson et al., 2012). We computed outgroup f₃-statistics in the form of f₃ (Manchu_Liaoning, X; Mbuti) to examine the shared genetic drift between Liaoning Manchus and non-African reference populations. Central African of Mbuti was used as the outgroup. We also used the f₃-test in the form of f₃ (source1, source2; Manchu_Liaoning) to formally test whether there was an admixture signature in Liaoning Manchus with different source pairs from modern and ancient eastern Eurasians. Negative values with the absolute Z-score of less than −3 indicated the included two source-related populations may be the plausible ancestral sources. We applied the f₄-test in the forms of f₄ (X, Y; Manchu_Liaoning, Mbuti) and f₄ (X, Manchu_Liaoning; Y, Mbuti) to estimate the differentiated allele sharing between Liaoning Manchus and other representative sources compared with focused comparative subjects, where X and Y are the included ancient and present-day populations. The tested form f₄ (X, Y; Manchu_Liaoning, Mbuti) was designed to examine the differentiated genetic ancestry between Manchus and X or Y, in which significantly negative values indicate more shared alleles between Y and Manchus compared with X, significantly positive values indicated more shared alleles between X and Manchus relative to Y, and nonsignificant values (Z-scores ranged from −3 to 3) indicated that X and Y formed one clade compared with the Manchus (Patterson et al., 2012).

We used formal tests of the qpWave/qpAdm programs in ADMIXTOOLS (Patterson et al., 2012) to determine the minimum number of streams of potential source populations contributing to the tested populations and also estimate the admixture proportions. We used the following eight outgroups including Atayal, Mbuti, Papuan, French, DevilsCave, Jomon, Malaysia_LN, and Tianyuan, which are unlikely to have been affected by recent gene flow with proposed ancestral sources and might be differentially related to the tested populations. We chose the included outgroups as the distant outgroups to dissect the northern and southern East Asian ancestries that participated in the formation of modern Manchus based on recent modern and ancient genetic studies (Chen et al., 2019; He et al., 2020; Wang et al., 2021a).

To explore the genetic relationship between the Manchu population and other references East Asians, we used the TreeMix v1.13 (Pickrell and Pritchard, 2012) program to construct the maximum likelihood trees with variable predefined mixture events. The level and time of admixture events were estimated by using ALDER v.1.0.3 (Wu, 2020).

We used shapeit v2 to phase successive SNPs into haplotype data and following used refined identity by decedent (IBD) to calculate the pairwise shared IBD (Browning and Browning, 2011). ChromoPainter (Lawson et al., 2012) and fineSTRUCTURE (Lawson et al., 2012) were further used to explore the fine-scale population structure based on the sharing haplotypes.

The mtDNA haplogroups were classified using HaploGrep2 (Weissensteiner et al., 2016) based on PhyloTree17⁵, and we used in-house scripts to assign the Y-chromosomal paternal lineage following the basic regulations reaccommodated via the International Society of Genetic Genealogy (2018)⁶.

To understand the general pattern of the genetic structure of Liaoning Manchus, we first performed PCA to explore the two-dimensional genetic relationship between Manchus and other reference eastern Eurasians. We found that the observed genetic clusters were consistent with the geographic and linguistic categories within East Asia (Figure 2). We observed in the PCA that Liaoning Manchus are genetically different with Turkic-, Tungusic-, and Mongolic-speaking groups in northwest China, Mongolia, and Siberia. Liaoning Manchus clustered closely to northern Han Chinese from Shanxi, Shandong, and Henan Provinces and also to Neolithic northern East Asians from the Yellow River and Western Liao River Basins.

In the model-based ADMIXTURE clustering analysis, we used cross-validation to identify an “optimal” number of clusters. We found the lowest CV error at K = 5 (Supplementary Table 1). At K = 5 (Figure 3), we observed there were three components of light green, dark green, and pink color reaching high proportions in Liaoning Manchus. The light green ancestry was enriched in the Tungusic people and ancient populations from the Baikal Lake region. Dark green ancestry with maximum proportion was observed in the southern Chinese populations and Southeast Asians, especially in Taiwan Hanben people. Pink ancestry was maximized in the Yellow River millet farmers. Therefore, Manchus had ancestry related to northeast Asians, Yellow River farmers, and southern East Asians. Similar genetic profiles were observed in the northern Han Chinese, suggesting a close genetic relationship between Manchu and northern Han. In pairwise F_ST analysis, we observed a similar pattern that Liaoning Manchus had the smallest genetic distance with Han Chinese in northern China such as in Shandong, Henan, and Shanxi (Figure 4 and Supplementary Table 2).

To further investigate the genetic origin and admixture of Liaoning Manchus, we performed outgroup-f₃ and admixture-f₃ statistics to measure allele sharing and detect admixture signals. In outgroup-f₃ (Manchu_Liaoning, Y; Mbuti) (Figure 5 and Supplementary Table 3A), we found that Liaoning Manchus shared the most derived alleles with northern and southern Han Chinese, She, Tujia, Ami, Miao, Japanese, and Korean. When Y represented ancient individuals, Liaoning Manchus share more alleles with ancient East Asians from the Yellow River and Western Liao River Basins, consistent with the observed patterns in PCA and ADMIXTURE. We next used admixture-f₃ statistics in the form of f₃ (X, Y; Manchu_Liaoning) to detect possible admixture signatures, in which X and Y were modern or ancient populations that might be the candidate sources for modeling the admixture of Liaoning Manchus. We observed significant signals of admixture (Z < −5) in the Liaoning Manchus when using CHS or Dai as the southern source and present-day Oroqen, Yakut, Buryat, and Hezhen or ancient Mongolia populations as the northern ancestral source. We listed the Z < −5 in the Supplementary Material (Supplementary Table 3B).

We performed f₄-statistics to explore genetic substructure between studied groups and other modern/ancient populations in the forms of f₄ (X, Y; Manchu_Liaoning, Mbuti) and f₄ (X, Manchu_Liaoning; Y, Mbuti) (Supplementary Table 4). When compared with modern Yakut, Tu, Buryat, and Thai populations, Liaoning Manchus shared more derived alleles with Han Chinese and Japanese. Compared with northern Tungusic, Mongolic, and Turkic people in Siberia, Manchu shared more alleles with the Tungusic- and Mongolic-speaking populations in China. When compared with the ancient populations from the Eurasian steppe, such as Yamnaya, EIA_Sagly, and EBA_Chemurchek, Liaoning Manchus shared more derived alleles with ancient populations from Yellow River Basin and West Liao River Basin, Boisman, and some ancient Mongolian such as EIA_SlabGrave and LBA_CenterWest. To further explore the differentiated allele sharing status between Manchu and Han Chinese, we conducted f₄ (Han, Manchu_Liaoning; X, Mbuti); and we found significant negative values when X was the Tungusic and Mongolic people, suggesting that Liaoning Manchus harbored more Tungusic-related ancestry than did Han Chinese (Supplementary Table 5A). We further observed significant negative f₄ values in the form f₄ (Mongolic/Tungusic populations, Manchu_Liaoning; southern modern East Asians, Mbuti), suggesting that Manchu people shared more alleles with southern East Asians compared with Tungusic and Mongolic people in the Amur River Region. We also observed significant negative values in f₄ (Xianbei, Manchu_Liaoning; X, Mbuti) when X represented Han, Dai, Tujia, or other southern populations, suggesting that there was gene flow from the southern part of China into Manchu after the Xianbei period. When X was ancient Eurasians, Liaoning Manchus shared more derived alleles with populations from Yellow River Basin and West Liao River Basin (Supplementary Tables 5B–H).

In the TreeMix analysis (Figure 6), we found that Tungusic-, Turkic-, and Mongolic-speaking groups in northern China tended to cluster together, but Liaoning Manchus clustered with northern Han Chinese and Guizhou Manchus clustered with southern Han Chinese and southern Tai-Kadai-speaking groups. The result was consistent with the patterns observed in the aforementioned PCA, ADMIXTURE, F_ST, and f-statistics; Liaoning Manchus and Guizhou Manchus had experienced genetic influence from surrounding Han Chinese after they were separated from northern Tungusic- and Mongolic-speaking ancestors.

We applied qpWave and qpAdm methods to further infer the possible ancestral populations and estimate the admixture proportions. We used the related available ancient northern populations (Heishui_Mohe, Boisman, Xiongnu, Xianbei, Yamnaya, Afanasievo, SlabGrave, Mongolia_North_N, CenterWest, MongunTaiga, Munkhkhairkhan, and Sagly) as the northern sources, used all available ancient populations (Miaozigou_MN, Shimao_LN, Upper_YR_IA, YR_MN, YR_LBIA, YR_LN, WLR_LN, and Upper_YR_LN) as the source of Yellow River sources, and used Iron Age Hanben (Hanben_IA) and Gongguan samples from Taiwan and Neolithic southern populations (SEastAsia_Coastal_EN, SEastAsia_Coastal_LN, SEastAsia_Island_LN, SEastAsia_Island_EN) as the southern sources. We observed that Manchus can be modeled as deriving 32.4% ancestry from Mohe people and the remaining ancestry from the farming-related ancient populations in the Yellow River Basin (Figure 7). When using Late Neolithic to Iron Age ancient southern populations (Hanben_IA, Gongguan and SEastAsia_Coastal_LN) as the southern sources, we found the proportion of northern ancestry (Heishui_Mohe and Xianbei) spanned from 61.5 to 81.2% (Figure 7). Compared with other Tungusic or Mongolic populations Hezhen, Mongolia, and Xibo, we observed Liaoning Manchus had derived more ancestry from farming-related populations in the Yellow River Basin and southern China and less ancestry from northern ancient groups, such as Heishui_Mohe and Xianbei.

We next used ALDER software to estimate when the admixture occurred. We tried different modern populations from the north and south of East Asia and Siberia as possible ancestral groups. We observed in most cases that the average time that the admixture occurred was around 500 AD (for example, 46.36 ± 20.93 generations) in the Southern and Northern Dynasties period.⁷ That period witnessed large-scale population migrations and admixtures due to the turbulence and war between Xianbei and Han Chinese.

We successfully identified 34 uniparental Y-chromosome lineages and 93 mtDNA lineages in Liaoning Manchus as shown in the Supplementary Material (Supplementary Table 6). We found that D4, A, and M8 were the dominant maternal lineages, and O2a2b1a2 was the dominant paternal lineage. Those paternal and maternal haplogroups are also dominant in Han Chinese, suggesting the possible gene flow from Han Chinese into the gene pool of Manchus.

Finally, to explore the fine-scale population structure of Manchu and other East Asians based on a higher-density dataset, we merged our data with 14 East Asian populations that were whole-genome sequenced in the Human Genome Diversity Project (HGDP) project. We performed the IBD-based clustering and calculated the pairwise F_ST genetic distances. We found that Manchus shared the most IBD with Han Chinese but had the smallest F_ST genetic distance with northern East Asians (Mongolia and Tu, Figures 8A,B). We have not detected the gene flow into Manchu people from our used plausible sources in TreeMix-based phylogeny with three gene flow events, but we found that Manchu clustered in between northern Mongolic and Tungusic people and southern Hans and other Hmong-Mien populations. Manchu clustered the closest with Han Chinese and Tujia (Figure 8C), which was further confirmed via the clustering patterns based on the fineSTRUCTURE-based chromosome painting patterns (Figures 8D,E).