Fine-scale Genetic Structure of Geographically Distinct Patrilineal Lineages Delineates Southward Migration Routes for Han Chinese

2.1. An Overview of the Y Chromosome Haplogroup Distributions of the Han Chinese Our dataset included a total of 372 haplogroups from 33 provincial-level administrative regions. The O haplogroup accounted for the highest proportion, with a prevalence of 78.06%, making it the most common Y-haplogroup among the patrilineal HAN (Figure 1). Notably, the O2a2 haplogroup accounted for 36.2%, indicating that approximately 240 million males in China belong to this haplogroup (2021 Population Census). This was followed by the O2a1 haplogroup, accounting for 17.9%. Collectively, these two haplogroups accounted for 54.1% of the patrilineal HAN. In addition, several other haplogroups with a proportion greater than 1% were identified, including O1a1 (12.03%), O1b1 (9.39%), C (9.15%), N (6.17%), Q (2.78%), D (2.01%), O2b1 (1.20%), and R (1.02%) (Figure 1). Additionally, other O haplogroups not included in the aforementioned haplogroups accounted for 1.34%. These haplogroups accounted for a cumulative prevalence of 99.19%, while the remaining nine haplogroups, A, E, F, G, H, I, J, L, and T, collectively accounted for only 0.81% and can be considered negligible. Therefore, our findings suggest that the O, C, N, Q, D, and R haplogroups are the predominant paternal haplogroups among the patrilineal HAN, which is consistent with the previous studies [4,20,21,22,23,24,25,26,27,28]. We constructed a phylogenetic tree of 372 haplogroups, which intuitively and clearly characterize the composition of the patrilineal HAN (Figure 1). According to the frequency and phylogeny relationship, we divided the haplogroups into 11 major haplogroups, some of which showed “star-like” expansion, suggesting possible population expansion events, some of which have been previously reported, including the three super branches: O_α, O_β, and O_γ [20]. 2.2. The Location of the Patrilineal HAN in Worldwide Populations To determine the location of the patrilineal HAN in worldwide populations, we collected data on the frequency of 28 Y-haplogroups from 221 populations worldwide, including our dataset (Table S1). We performed principal component analysis (PCA), and the first two principal components (PCs) explained 16.1% and 7.9% of the total variance, respectively. PC1 was roughly correlated with the East-West direction, while PC2 was correlated with the North-South direction (Figure 2A). Our analysis revealed a clear separation between different Eurasian populations, with the greatest divergence observed between Indo-European, Uralic, and Tai-Kadai groups. The genetic structure reflects their geographical position in Eurasia, with the patrilineal HAN is located in the core region of East Asia, surrounded by other East Asian populations such as the Japanese, Korean, and Tai-Kadai populations. This clustering result also suggests the genetic continuity of the patrilineal HAN. Within HAN, the provinces in northern China are separated from those in southern China (Figure 2B). Our PCA results, as well as additional analyses using NJ-tree, 3D-PCA, and tSNE, all support the genetic continuity and North-South divergence of the Han Chinese (Figure 3B,C; Figure 4; Figure 5).2.3. 3D-PCA and tSNE Demonstrate the Fine-scale Genetic Structure of HAN To explore the fine-scale genetic structure of the patrilineal HAN, PCA and tSNE were used. We excluded Macau, Hong Kong, and Qinghai provinces due to small sample sizes (<30). Our results revealed that the patrilineal HAN can be divided into five distinct subgroups: Northern China, Eastern Coast, Upper and Middle Yangtze River, Lingnan, and Min-Tai (Fujian and Taiwan) (Figure 3A–C; Supplementary files S1-S2). The Northern China subgroup included all 14 northern provinces, while the Eastern Coast subgroup was comprised of four provinces downstream of the Yangtze River and on the Eastern Coast area (Jiangsu, Zhejiang, Shanghai, and Anhui). The Upper and Middle Yangtze River subgroup was comprised of seven provinces along the Yangtze River, excluding the aforementioned downstream provinces. The Min-Tai subgroup included Fujian and Taiwan provinces, while the Lingnan subgroup included Guangdong, Guangxi, and Hainan provinces (Figure 3A). PCA analysis revealed that PC1, PC2, and PC3 accounted for 22.2%, 10.2%, and 9.2% of the total variance, respectively. The five subgroups in the PCA corresponded well with their actual geographic locations, with PC1 reflecting the North-South cline and PC2 roughly separating East and West China. The tSNE clustering was consistent with the PCA, as well as the consistency of each subgroup’s composition (Figure 3C; Supplementary files S1 and S2). The lack of a linear relationship between dimensions is an essential feature of the tSNE algorithm. However, a strong correlation existed between the groups in the tSNE dimensions and real geographic relationship, particularly in the first two dimensions, closely matched the geography of Chinese provinces. These indicating a complex, nonlinear relationship between geographical factors and Y-chromosome haplogroup frequencies, which could not be explained by latitude and longitude alone. To avoid bias in the comparison of Y-chromosome and mtDNA results, we downloaded an mtDNA dataset with almost the same sample size as ours (21668) from a published literature [11]. Identical PCA and tSNE were used to identify the genetic structure. In PCA, PC1, PC2, and PC3 explained 20.2%, 9.5%, and 7.9% of the variance, respectively. PC1 showed a clear North-South geographic orientation, consistent with the Y chromosome results (Figure 3D; Supplementary files S3 and S4), while PC2 and PC3 did not (Figure S2). Our PCA results indicated that the matrilineal HAN could be divided into three to five subgroups, but the subgroup boundaries were not as clear-cut as those in the patrilineal HAN. Provinces within the subgroups were geographically far apart, with Ningxia and Taiwan provinces distant from other subgroups and almost appearing as separate subgroups. Different from PCA, tSNE provided more distinct clustering findings than PCA, with distinct boundaries among the three subgroups and no outliers (Figure 3D; Supplementary files S3 and S4). tSNE and PCA results suggested that the matrilineal HAN may have had more population mixing events than patrilineal HAN, resulting in a blur of subgroup boundaries. The haplogroup diversity analysis yielded similar results (Figure 4, Table S11).2.4. Sex Bias Demonstrated by AMOVA To confirm the fine-scale genetic structure, we calculated the inter-group variation using AMOVA with various classification schemes (Table S2). We classified the HAN population into 1–8 subgroups (Table S2). The inter-group variation reflects the distance between different subgroups. These AMOVA results confirmed that the most significant inter-group variation was 1.60%, when using a five subgroup scheme for patrilineal HAN (Table 1). According to AMOVA, inter-group variation was also significant (1.41%) when the population was classified into three subgroups: Northern China, Yangtze River, and Zhujiang River. In comparison, the inter-group variation for HAN in Southern and Northern China was 1.17%, and there was considerable variation within group (Table 1). Notably, dividing the Northern China group into two subgroups resulted in an inter-group variation of 1.57%, approaching the largest inter-group variation (1.60%). This grouping pattern was consistent with the clustering results of PCA and tSNE, adding further support to the five subgroups within the patrilineal HAN. Similar to the patrilineal population, AMOVA was utilized to examine the inter-group variations in matrilineal HAN. When we divided matrilineal HAN into four subgroups based on PCA, we observed the greatest inter-group difference, at 0.995%. Inter-group variation stood at 0.981% using three subgroups according to tSNE (Table S3). These two values were similar and both markedly greater than other classification schemes, such as the five subgroup scheme based on the Y chromosome, which only demonstrated 0.613% inter-group variance (Table S3). This suggested that patrilineal and matrilineal HAN have different genetic substructures. Moreover, different subgroup patterns demonstrated by tSNE and PCA, as supported by AMOVA, reveal the complex genetic structure of matrilineal HAN. It appears that the genetic boundaries of matrilineal HAN are not as well-defined as those in patrilineal HAN. Comparison with Y haplogroup, AMOVA based on mtDNA demonstrated larger intra-population and smaller inter-group variation on average. The intra-population variation reflects the similarity between individuals within the population. When the HAN was considered as one group, intra-population variation of mtDNA stood at 99.06% and at 98.33% of Y chromosome (Table 1; Table S3), indicating that matrilineal HAN was more closely connected at individual level by comparison with patrilineal HAN. This also explained the blurred boundary among matrilineal subgroups discussed above. These findings were further supported by the subsequent HD and NJ tree results. The observation of mixed matrilineal HAN and more distinct patrilineal HAN suggests a sex bias in Han Chinese populations.2.5. Mantel Test and PC-based Correlation Demonstrate Better Geographical Aggregation of Patrilineal HAN PCA, tSNE, and AMOVA revealed a relationship between the genetic structure and geography. For quantitative characterization, we calculated the correlation between PCs and geography (Figure S4; Figure S1). In addition, we utilized a Mantel test to quantify the correlation between geographical distance and genetic distance. PC1 and latitude had the greatest Pearson correlation coefficient (PCCs) of 0.86 (p-value = 0.001) in the Y chromosome PCA (Figure S2C), while PC2 and longitude had a relatively lower correlation of 0.64 (p-value = 0.001) (Figure S2B). In the mtDNA PCA, PC1 had the highest PCCs of −0.85 (p-value 0.001), while PC2 and longitude had only 0.03 PCCs and was not statistically significant (p-value = 0.857) (Figure S3). Regardless of patrilineal or matrilineal lineage, the above results indicate that the latitude, or North-South direction, was the primary factor determining the genetic structure of the Chinese population. Longitudinal East-West direction significantly influenced the genetic structure of the patrilineal HAN but had little effect on the matrilineal HAN. This trend has been overlooked in prior investigations with limited sample sizes. Similarly, we observed that the PCs correlation on average was slightly higher in the patrilineal HAN as opposed to the matrilineal HAN, suggesting the possibility of stronger geographic correlation among patrilineal HAN. We also conducted a Mantel test to examine the relationship between the geographic and genetic distance (Table S4). Here, the correlation between geographic distance (Vincenty great circle distance) and genetic distance (pairwise Fst, Euclidean distance in PC space) was greater for patrilineal HAN, indicating more intense geographical concentration. For example, the correlation of Fst and Vincent distance for the Y chromosome was 0.4597 and 0.3610 for mtDNA. Moreover, we observed different geographical patterns between the patrilineal and matrilineal HAN in southern and northern provinces, respectively. Specifically, for the Y chromosome, the correlation was stronger in the northern provinces (0.4351) than in the southern provinces (0.3322) (Table S4), indicating the geographical environment in the northern region has shaped a more homogeneous paternal lineage. In contrast, for mtDNA results, the correlation was a little weaker in the northern provinces (0.2302) than in the southern provinces (0.2649). These results suggested that the geography of northern China has a greater impact on patrilineal HAN, while the geography of southern China exerted a greater impact on matrilineal HAN. These results reflect a geographical sex bias in the Han Chinese. 2.6. Higher Haplogroup Diversity of Northern HAN Indicates North to South Migration We calculated haplogroup diversity (HD) values for the patrilineal and matrilineal HAN for each province. We found that, on average, the matrilineal HAN exhibit higher HD values than patrilineal HAN. Specifically, the average HD value for patrilineal HAN was 0.985, while the value for matrilineal HAN was 0.994 (Figure 4A,B). In terms of geographic distribution, provinces with higher HD value of patrilineal HAN were mainly located in frontier regions such as northern China, southwestern China, and Xinjiang (Figure 4A). In contrast, the matrilineal HAN showed different pattern, showing higher values in central China (Henan, Hubei, Anhui, and Jiangxi) (Figure 4B). Interestingly, Henan, representing central China, exhibited higher haplogroup diversity for both Y chromosome and mtDNA. In central China, the higher haplogroup diversity of mtDNA probably due to the convergence of populations from surrounding regions. In border regions of China, the high diversity of Y chromosome may be the result of admixture with minority and Central Plains HAN, although this process may not have involved matrilineal populations. The lower haplogroup diversity of southern HAN may be due to genetic bottlenecks experienced during historical migrations from north to south. 2.7. NJ Tree Indicates a Possible Southward Migration Route of HAN We calculated the Euclidean distance matrix and the pairwise F_st distance matrix (Table S6; Table S11) between each province. Based on these two matrices, we constructed NJ trees for the patrilineal and matrilineal HAN, respectively (Figure 5; Figures S5 and S6). Our results revealed a clear separation between the southern and northern patrilineal HAN, with both distance-based NJ trees exhibitisng two distinct clades (Figure 5; Figures S5 and S6). However, within the same clade, the matrilineal HAN showed some intermingling between southern and northern provinces. Meanwhile, our analysis demonstrated that the northern patrilineal HAN were more closely clustered than their southern counterparts, consistent with the PCA findings (Figure 3). Interestingly, previous studies have revealed a male-dominated north-to-south migration of Chinese populations throughout history [18]. Our haplogroup diversity results supported this conclusion (Figure 4). The NJ tree also exhibited a stepped structure with four endpoints for north-to-south migration. Starting with the northern population on the NJ tree, the Eastern Coast subgroup was the first to diverge, followed by the seven provinces of the Upper and Middle Yangtze River subgroup, and finally Fujian-Taiwan and the three Lingnan provinces. This indicated that northern inhabitants may have migrate in Lingnan and Fujian-Taiwan via the middle Yangtze River, as opposed to the Eastern Coast. The PCA confirmed this result. According to PCA, the Lingnan and the Upper and Middle Yangtze River subgroups exhibited closest proximity as well (Figure 3B).