The Shangshihe Cemetery (110°50′47″E, 34°42′N), located in Shangshihe Village, Yima City, Henan Province, underwent two phases of excavation in 2017 and 2018 (Figure 1). The site yielded abundant cultural remains, including burials, horse pits, bronze vessels, pottery, jade artifacts, and stone tools (Yan et al., 2019). The tomb structures and burial goods assemblages are similar to those of the Guo State cemetery in Sanmenxia City. Based on the inscriptions on the bronze vessel and historical records, the cemetery is likely the burial site of the Guo State elite, their families, and retainers who fled eastward after the state was conquered (Yang and Yan, 2020). Previous studies have revealed clear social stratification at the site, with burial scale and orientation showing distinct patterns (Yang and Yan, 2020). Large and medium sized tombs are predominantly aligned north-south, consistent with the Guo State cemetery, while smaller tombs are mostly oriented east-west. These findings support the hypothesis that the Shangshihe Cemetery is linked to the Guo State during the Eastern Zhou period. We collected teeth and petrous bones from the human remains. A total of 13 individuals from the Shangshihe cemetery were selected for ancient DNA analysis.

DNA extraction and library preparation were conducted following established protocols in a dedicated ancient DNA facility at Henan University, China (Korlević et al., 2015). Human remains were decontaminated using 75% ethanol, followed by a 5% NaClO wash, and then exposed to UV light for 30 min per side. Tooth samples were processed into powder using a dental drill (Strong 90). Petrous bones were ground into fine powder using an automated grinder (JXFSTPRP-24L) after the outer layers were removed with a sandblaster machine (Renfert, Germany). DNA was extracted from 50 to 120 mg of bone powder following Dabney’s protocol (Dabney et al., 2013). Double-stranded libraries were constructed using a published protocol with minor modifications, without uracil-DNA glycosylase (UDG) treatment (Meyer and Kircher, 2010). Briefly, DNA was blunt-ended using T4 PNK and T4 polymerase at 15°C for 15 min and 25°C for 15 min, followed by purification with the MinElute PCR Purification Kit (Qiagen) using TET buffer (Tween20, EDTA, Tris-HCl) for elution. Designed adapters were then ligated to the blunt ends, and a second purification was performed. The adapter fill-in step was performed using Bst DNA polymerase at 37°C for 20 min and 80°C for 20 min. Libraries were amplified with dual-indexing primers (P5 and P7) using Q5 DNA polymerase and then purified using AMPure XP Beads (Beckman Coulter). The final libraries were quantified using a Qubit 4.0 Fluorometer (Thermo Fisher Scientific) and sequenced on an Illumina NovaSeq 6000 platform at Novogene Co., Ltd., China.

The paired-end sequencing data were merged into single-end reads using AdapterRemoval v2.3.3 (Schubert et al., 2016). The merged reads were aligned to the human reference genome (hs37d5) using BWA v0.7.17 with the parameter “-n 0.01” (Li and Durbin, 2009). The aligned reads were converted to BAM format and sorted using Samtools v1.17 (Li et al., 2009). Duplicate sequences resulting from PCR amplification were removed using DeDup v0.12.8 (Peltzer et al., 2016). The BAM files were then indexed with Samtools. Coverage of the genome was assessed using Qualimap v2.2.2 (Okonechnikov et al., 2016), and ancient DNA damage patterns were evaluated with mapDamage v2.2.2 (Jónsson et al., 2013). Quality control was performed by filtering reads with a mapping quality of at least 30 using Samtools view (-q 30), followed by trimming of 10 bp from the ends with trimBam (Danecek et al., 2021). In addition, we evaluated mitochondrial contamination levels using Schmutzi (Renaud et al., 2015) and determined nuclear genome contamination rates in males with ANGSD v0.940 (Korneliussen et al., 2014). Genotype calling was performed using SAMtools mpileup with the parameters “-q30, -Q30” and pileupCaller,¹ applying the “—randomHaploid” option to generate pseudodiploid genotypes through random sampling from the 1240K SNP dataset (Mallick et al., 2024). The resulting genotype data were then combined with worldwide genotype datasets, derived from the Human Origins (597,573 SNPs) panel² and the 1240K (1,233,013 SNPs) panel.

Sex determination was performed by calculating the ratio of Y-chromosomal and X-chromosomal coverage relative to that of autosomal coverage (Fu et al., 2016). Mitochondrial haplogroups were determined by mapping trimmed and collapsed reads to the human mitochondrial reference genome (NC_012920.1). Consensus sequences were generated using the log2fasta tool from the Schmutzi package, followed by haplogroup assignment with HaploGrep v2.4.0 (Weissensteiner et al., 2016). For Y-chromosomal haplogroup identification, the Yleaf software was used (Ralf et al., 2018), specifically employing the “Yleaf.py” and “predict_haplogroup.py” scripts to generate results that include both ancestral and derived SNPs within the ISOGG Y-chromosome phylogenetic framework. To ensure accuracy, these key SNPs were examined using the Integrative Genomics Viewer software (Thorvaldsdóttir et al., 2013).

Runs of Homozygosity (ROH) analysis serves as a genomic approach to identify continuous homozygous segments, providing insights into inbreeding and population bottleneck events. In this study, we employed hapROH³ to investigate the occurrence of consanguineous practices in the Shangshihe population (Huang and Ringbauer, 2022). For kinship estimation, we applied two different methods, including the Pairwise Mismatch Rate (PMR) and the Relationship Estimation from Ancient DNA (READ), to ensure robust relatedness inference (Kennett et al., 2017; Kuhn et al., 2018). The PMR method quantifies genetic divergence by calculating the ratio of mismatched SNP sites between paired individuals and normalizing this count by the total number of loci with available data for both. The READ method evaluates genome-wide similarity by measuring the proportion of discordant alleles across non-overlapping 1 Mb genomic windows, enabling precise estimation of kinship coefficients (Kuhn et al., 2018).

To explore the genetic structure of the Shangshihe population, we conducted principal component analysis (PCA) on the Human Origins (HO) dataset using the smartpca v18140 tool from the EIGENSOFT package (Patterson et al., 2006), enabling the “lsqproject: YES” and “shrinkmode: YES” options. Ancient samples were projected onto principal components derived from these modern populations to reveal inter-population affinities. The PCA results were visualized using the ggplot2 package in R.⁴ Additionally, we performed an unsupervised ancestry analysis with ADMIXTURE v1.3.0 (Patterson et al., 2012) on the HO dataset. Prior to analysis, we excluded SNPs with a minor allele frequency below 1% (–maf 0.01) and applied linkage disequilibrium pruning in Plink v1.90 (Chang et al., 2015) with the parameters “–indep-pairwise 200 25 0.2.” The ADMIXTURE analysis was executed across K values from 2 to 20, and cross-validation (CV) errors were calculated to determine the optimal K value, with lower CV values reflecting enhanced model stability. The resulting ancestry proportions were subsequently visualized using the AncestryPainter tool (Feng et al., 2018).

To investigate the shared genetic drift between the Shangshihe population and other groups, we calculated the outgroup f3-statistics and f4-statistics using the qp3pop v651 and qpDstat v980 programs from the ADMIXTOOLS package (Patterson et al., 2012). The outgroup f3-statistics were computed in the form f3(Mbuti.SDG; X, Shangshihe) with the “inbreed: YES” parameter enabled. Here, the Mbuti population from Central Africa served as the outgroup, while X represented various Eurasian populations. The top 30 populations with more than 10,000 SNPs, ranked by f3 values, were visualized using DataGraph v4.5.1. Additionally, the f4-statistics were performed with the parameter “f4mode: YES” on the model (Mbuti.SDG, Reference; X, Shangshihe). This model was used to explore additional gene flow between the Shangshihe population and either the reference groups or X populations. For this analysis, the reference groups comprised by ancient and modern Eurasian populations, while X consisted of ancient populations from diverse East Asian regions. We further employed the f4(Mbuti.SDG, YR_related; Shangshihe, ancient Eurasians) model to test whether ancient populations from the Yellow River Basin share more ancestry with the Shangshihe population. Standard errors were calculated using 5 cM block jackknifing, as implemented in the f3-statistics and f4-statistics analyses.

To determine whether different individuals from the Shangshihe population could be modeled as deriving from the same ancestral sources, we conducted pairwise qpWave tests using the qpWave v1520 program from ADMIXTOOLS (Patterson et al., 2012). A rank of 0 with a p-value greater than 0.01 suggests that the tested individuals share a consistent ancestry profile without evidence of distinct admixture events. Next, we performed qpAdm analysis to further explore the specific ancestry sources and their proportions, using qpAdm v1520 from the same package with the parameters “allsnp: YES” and “details: YES.” We set up a base outgroup, including Central African outgroup (Mbuti), indigenous Andamanese islanders (Onge), Western Eurasian hunter-gatherers (Loschbour), Ancient Northern Eurasian population (Botai), Neolithic farmers from Iran (Iran_N), Ancient Northeast Asian population (Shamanka), Early Asian lineage associated with the Jomon culture (Japan_Jomon), Early Neolithic population from Fujian (Fujian_EN), Indigenous Taiwan group (Ami), and early Neolithic population from Shandong (Shandong_EN). To evaluate competing admixture hypotheses, we implemented a rotation strategy, adding source populations to the base outgroup and iteratively running qpAdm to refine the ancestry proportions and assess model validity (Harney et al., 2021).

We extracted DNA from 13 human skeletal remains at the Shangshihe cemetery and performed shotgun sequencing. All individuals yielded sufficient endogenous DNA content (3.99–74.09%), resulting in autosomal coverage ranging from 0.02X to 0.93X and mitochondrial DNA coverage from 5.68X to 38.79X (Table 1; Supplementary Table S1). Analysis of the data revealed typical ancient DNA damage patterns, including C-to-T misincorporations at the 5′ end and G-to-A misincorporations at the 3′ end (Supplementary Figure S1), along with short average fragment lengths (ranging from 52.74 to 83.14 bp) (Table 1). Mitochondrial DNA contamination rates were below 5% for all individuals except SSH48, and X-chromosome contamination in male samples ranged from 0.65 to 3.24% (Supplementary Table S1). These characteristics confirm the authenticity and reliability of the genomic data obtained in this study. After excluding samples with high contamination (>5%) and low SNP counts (< 30,000) that overlapped with the 1240K SNP panel, the remaining individuals were used for downstream population genomic analyses.

We successfully assigned the mitochondrial haplotypes for all individuals. A total of 12 mtDNA haplogroups were detected (Table 1), including A (A8), B (B4c1b2c, B4a1c5*), C (C7), D (D4a5, D5a3), F (F1 + 16189, F2c, F2 + 16291), G (G2a1, G2a’c), and N (N9a1). These haplogroups are common in East Asia and indicate high maternal genetic diversity within the Shangshihe population. Most of these haplogroups were found in various sites across the Yellow River Basin over different periods. For instance, haplogroup D4a5 was found in Middle Neolithic sites such as the Wanggou and Qingtai sites, as well as in Shandong_LN populations (Ning et al., 2020; Wu, 2024; Liu et al., 2025). Haplogroups N9a1 and G2a1 were identified in Late Neolithic sites including the Shimao and Taosi sites (Xue et al., 2022). Haplogroups B4c1b2c, D5a3, and F1 + 16189 were detected in the Lusixi site during the Iron Age (Ma et al., 2025). These findings demonstrate the long-term continuity of maternal genetic structure in the Yellow River Basin.

In addition, some haplogroups were also detected in ancient populations from Southern China. For instance, haplogroup B4c1b2a (a sister branch of B4c1b2c) was found in the Dayin and Xitoucun sites during the Late Neolithic period (Wang et al., 2020; Liu et al., 2021). Haplogroup C7 was identified in the Babanqincen and Gaohuahua sites from the historical period (Liu et al., 2022). These observations suggest significant interactions between the Shangshihe population and surrounding groups in the maternal lineage. Furthermore, seven male individuals were successfully identified with Y-chromosome haplogroups, all of which belonged to the East Eurasian types (C2b1b, O2a, N1a2, N1b2, and Q1a1a1a), exhibiting high paternal genetic diversity (Table 1). These haplogroups were also found in the contemporaneous Guanzhuang site (Wu et al., 2024). Collectively, these results suggest that the Shangshihe population has diverse paternal and maternal genetic origins.

We assessed the kinship relationships among Shangshihe individuals using two complementary methods (PMR and READ) to ensure the robustness of our results. Only samples with more than 30,000 SNPs were included in the kinship analysis, which revealed no close genetic relationships among the individuals (Supplementary Figure S2; Supplementary Table S2). We further evaluated parental relatedness by measuring the length of Runs of Homozygosity (ROH) for each individual. Our findings showed that five individuals (SSH19N, SSH38N, SSH43N, SSH65, and SSH74) exhibited ROH segments exceeding 4 centiMorgans (cM) (Supplementary Table S3). Among them, SSH65 and SSH43N also displayed a limited number of longer ROH segments (> 20 cM). However, with total ROH lengths remaining below 130 cM, there was no evidence of inbreeding or close intermarriage, such as between second cousins, within the selected samples (Supplementary Figure S3).

To characterize the genetic structure of the Shangshihe population, we employed principal component analysis (PCA) to visualize population structure through dimensionality reduction. Ancient samples were projected onto the genetic variation defined by present-day Eurasian individuals. We found that the Shangshihe population clustered with the modern Han Chinese population and ancient populations from the Yellow River Basin (Figure 2). Compared to Middle Neolithic populations in the Yellow River Basin (YR_MN), the Shangshihe individuals are shifted southward along PC2, clustering with Late Neolithic populations (such as YR_Wadian_Longshan, YR_LN), Bronze Age populations (such as YR_LBIA, YR_Wangchenggang_Erlitou), and Zhou dynasty populations (such as Lusixi_Zhou and Guanzhuang). To further refine our analysis, we conducted an additional PCA focusing on East Asian genomes. This revealed that Shangshihe individuals exhibit strong genetic similarities to ancient populations dating from the Bronze Age to Iron Age in the Central Plains of China (YR_LBIA) as well as modern Northern Han populations (Supplementary Figure S4). These findings were consistent with model-based unsupervised ADMIXTURE simulation analysis. Our analysis identified the model with the lowest cross-validation error at K = 5 (Supplementary Figure S5). Most populations in the Yellow River Basin were admixed with Northern East Asian-related ancestry and Southern East Asian-related ancestry (Figure 3A). The Shangshihe population and other Zhou dynasty populations (Lusixi_Zhou and Guanzhuang) displayed similar structure, with increased Southern East Asian ancestry compared to Neolithic populations. This indicates intensified genetic interaction between Central Plains populations and Southern East Asian populations after the Neolithic period.

Similar results were observed in the outgroup f3-statistics in the form of f3(Mbuti.SDG, X; Shangshihe). Larger f3 values indicate that the Shangshihe population shares more genetic drift with the test population. The strongest genetic sharing was observed with the Guanzhuang population, followed by various Yellow River Basin groups, including those from the Yangshao and Longshan cultures (Figure 3B; Supplementary Table S4). This points to a broad genetic continuity across Yellow River Basin populations over time. Additionally, the Shangshihe population shows a remarkably close relationship with historical-period and Dawenkou culture populations from Shandong, consistent with previous studies (Du et al., 2024). This suggests that Neolithic farmers from the middle Yellow River likely contributed genetically to populations downstream, highlighting regional gene flow within the basin.

The genetic interactions influencing the Central Plains population during the Eastern Zhou period are reflected in the Shangshihe population. Our analyses using PCA, f3-statistics, and admixture modeling reveal a close genetic relationship between the Shangshihe population and YR-related populations. To further investigate potential gene flow between the Shangshihe population and other groups, we applied the f4 statistics in the form of f4(Mbuti, X; YR-related, Shangshihe). When YR-related populations were represented by the Guanzhuang population, and X included various Eurasian groups, the results of f4 statistics showed non-significant Z-values (|Z| < 2.5), confirming a close genetic relationship between Guanzhuang and Shangshihe populations (Supplementary Figure S6; Supplementary Table S5). However, when comparing the Shangshihe population to the Yangshaocun_Longshan population, we observed genetic influences from Southern East Asian populations, such as Tonga_2700BP, Vietnam_BA, and Laos_LN_BA populations, with Z-scores exceeding 2.5 (Supplementary Figure S6). Similarly, when the Shangshihe population was compared to the Longshan culture-related populations from the Yellow River Basin (e.g., YR_Wadian_Longshan), we observed distinct genetic differences (Figure 4; Supplementary Figure S6). Significant gene flow (Z-scores > 2.5) was detected with some Eurasian steppe populations (e.g., Kazakhstan_Nomad_IA.SG, Hungary_IA_Scythian, and Kazakhstan_TianShan_Saka.SG). Similar steppe-related genetic signatures were also observed when comparing Shangshihe with the Lusixi_Zhou population (Figure 4; Supplementary Table S5). These findings suggest that the Shangshihe population likely experienced gene flow with these diverse groups, demonstrating the genetic interactions during the Eastern Zhou period.

The population interactions were further reflected through qpAdm ancestry modeling. Shangshihe could be effectively modeled as a single-source population derived from YR-related groups in one-way models (Supplementary Table S6A). However, incorporating Southern East Asian (SEA)-related populations as an additional source, Shangshihe was modeled as approximately 89.3–91.8% YR-related ancestry (represented by YR_Yuzhuang_Longshan or Lusixi_Zhou) and 8.2–10.7% SEA-related ancestry (represented by Atayal or Thai) (Supplementary Table S6B). More specifically, we detected a subtle steppe influence, with Shangshihe comprising 96.4–98.8% YR-related ancestry (e.g., YR_Wadian_Longshan or Lusixi_Zhou) and 1.2–3.6% Eurasian steppe ancestry (e.g., Kazakhstan_Nomad_IA) (Figure 5). Notably, compared to the contemporary Lusixi_Zhou population, qpAdm results did not detect any Eurasian steppe ancestry (Supplementary Table S6C), highlighting the high genetic diversity of the Central Plains population during the Zhou Dynasty. These best-fitting admixture models align with the f4 findings, emphasizing diverse ancestral origins in the Shangshihe population.

In summary, the Shangshihe population exhibits a complex genetic profile characterized by dominant YR-related ancestry alongside influences from Southern East Asian-related and Eurasian steppe-related sources. This admixture pattern likely reflects intensified population interactions during the Eastern Zhou period, possibly driven by cultural exchange, population migration, or trade networks in the ancient Central Plains.