|
Post by Admin on Jul 1, 2024 21:34:59 GMT
Results Compiling and authenticating the ancient DNA dataset We converted drilled bone and tooth powder from 359 samples of human remains into single-stranded genomic libraries, estimated endogenous DNA preservation by shotgun-sequencing a portion of the libraries (n=96), and proceeded with enrichment for 1,233,0123 variable sites (SNPs) in the human genome (‘1240k’ panel) for 343 samples (Tables S1-S3). After quality control, we removed libraries with low sequencing coverage on the targeted positions (<20,000 SNPs covered; see Methods) and substantial contamination rates (>10%) and merged libraries from genetically identical individuals. After this QC, we retained a dataset of 219 unique individuals with newly generated genome-wide data (Methods, Table S1). Where factors such as female sex, coverage, mt/nu ratio, and terminal deamination (aDNA damage) prevented a reliable autosomal contamination estimate, we kept these cases in our dataset but excluded them from our group-based analyses and interpreted the other results cautiously. Our final dataset originates from 47 sites across present-day Georgia (SI Material 1), 22 located close to Mtskheta and Tbilisi, the former and last capital of the Kingdom of Iberia, respectively (Figure 1A). By unifying the sampling range from present-day Armenia to the North Caucasus and the adjacent steppe (Figure 1B), our results synthesize into a larger picture of the population dynamics in the area from prehistory to historical periods. Several of the earliest individuals are associated with the Kura-Araxes culture (the site of Kiketi), a Bronze Age cultural complex that extended throughout the Southern Caucasus. Despite its importance and many unresolved consequential questions, such as its linguistic makeup, published ancient DNA only covers the southern and eastern end of this range 6,8. Finally, the most recent transect in our dataset (Early Middle Ages) is a unique addition to the Caucasus region that allows us to extend our understanding of the impact of Christianization and the subsequent ‘Migration Period’ on human mobility and population genetics beyond Europe 24–27. For some analysis, we partitioned the dataset based on time periods (see Methods; Data compilation). We note that a discretization of the dynamic past of the South Caucasus is challenging due to diverse cultural and archaeological contexts. Therefore, our grouping is based solely on calendar dates, with the main goal of increasing the statistical power of time series analysis. ![](https://www.biorxiv.org/content/biorxiv/early/2024/06/13/2024.06.11.597880/F1.medium.gif) Figure 1. Spatiotemporal information of new and published data analyzed in this study. A. We annotate the geographic location of new genomic data from Georgia in squares, with solid colors corresponding to their chronological period. Published data from Georgia and surrounding territories (Armenia, Azerbaijan, Russia, and Turkey) are also indicated in transparent colors. We also highlight locations with artificially elongated skulls. B. Date and archaeological period information as an average for published data and as a range for the new samples. 14C date ranges are annotated in black lines. Abbreviations of the periods discussed are provided.
|
|
|
Post by Admin on Jul 3, 2024 20:02:13 GMT
Persisting local gene pool and limited gene flow in the South Caucasus To explore genetic ancestry, we first assessed the genome-wide data by projecting the time-transect and other relevant ancient DNA datasets onto the first two axes of a principal component analysis that we constructed from 77 present-day populations from Western Eurasia (Figure 2A), including several modern groups from the Caucasus (see Methods). We find that more than 90% of the individuals join the same cluster outlined by previous studies presenting Eneolithic to Early Iron Age individuals from the southern and northern Caucasus (excluding the steppe environment to the north) 6,9,28,29. Notably, this ancient pc1-pc2 cluster also overlaps with modern populations from the southern Caucasus. None of the individuals from the Bronze to Iron Age phases fall into the distinct ‘Caucasus-Steppe BA’ cluster or a cline of mixed ancestry from the steppe to the southern Caucasus that was described previously 6. This signal indicates that this pc1-pc2 cloud comprehensively represents the genetic variability of the local South Caucasus ancestry. Figure 2. ![](https://www.biorxiv.org/content/biorxiv/early/2024/06/13/2024.06.11.597880/F2.medium.gif) Figure 2. Principal component analysis in two different datasets. A. PCA on 1300 present-day individuals from W. Eurasia. The dataset of 221 from Georgia is projected with the ‘lsqproject’ parameter of smartpca and annotated by period (color and shape). Fill color transparency is applied for individuals with low coverage and/or questionable quality (e.g., contamination rates). Other relevant ancient datasets are projected as well. B. PCA on 1486 present-day individuals from W. Eurasia and C. Asia. Published ancient groups are plotted when their members have similar PC1 and PC2 coordinates and are dated close to the ancestry outliers from Georgia. Individuals with artificial cranial deformations are annotated with respective icons. PC1 and PC2 variability in post-Hellenistic Georgia is also detected with cluster analysis with pairwise qpWave tests that indicate three different clusters at the plot’s upper-right corner. Within this cluster of local ancestry, however, we observe a temporal substructure throughout the BA, with EBA individuals occupying higher pc1 and pc2 values compared to M/LBA individuals (Figure 2A). Moreover, several individuals are placed outside this main cluster. They all date from the Late Antique and Early Middle Ages (n=17) and can be further grouped into two components: a. those on a continuum of increasing genetic affinity to Anatolia/Levant (n=10), and b. those with diverse ancestry backgrounds (n=7, referred to throughout the text as ‘ancestry outliers’). Remarkably, four out of seven of the ancestry outliers are individuals whose preserved skulls attest to the practice of ACD (Figure 2B). To better understand the geographic origin of the ancestry outliers (or of their recent ancestors), we computed another PCA that also included present-day individuals from Central Asian populations (see Methods) and projected the ancient ones as before. This PCA shows that the two individuals SMT013 and SVL015 (c. 3rd and 4th centuries AD, respectively) descended from Central Asian populations (Figure 2B), a conclusion further supported by another PCA on all available Eurasian populations (Supplementary Figure 1A). Projecting only ancient groups with similar pc1 and pc2 coordinates and dates to our ancestry outliers reveals a gradient of ancestries from European to Central/Northeast Asian, which is also visible in individuals from the Pannonian Basin starting with the invasion of the Huns and later the Avars 30,31. This observation suggests that the ancestry outliers, who were all buried in two urban contexts within a short distance of each other (Samtavro and Samshvilde), may also be linked to related contexts. To further evaluate the ancestry patterns in the PCA, we performed explicit admixture modeling using qpAdm/qpWave to address four main objectives: 1. deconstruct ancestry as broad contributions from earlier Southwest Asian and East European populations and compare the composition across time periods (distal modeling) 2. explicitly model differences across periods using as sources populations that could have been donors (proximal modeling), 3. quantify genetic variability among individuals, and 4. investigate alternative models (admixture) for the ancestry outliers. For the distal modeling, we found that all groups from Georgia and Armenia can be modeled with sources from the earliest sources from Neolithic Anatolia, the hunter-gatherers from Georgia (CHG), Chalcolithic Iran (west; Seh Gabi), Pre-Pottery Neolithic (PPN) Levant, and hunter-gatherers from Eastern Europe (EEHG) (Supplementary Figure 2A). After corroborating previous findings that CHG and Chalcolithic Iran ancestries are competing sources that are challenging to distinguish 8, we combined them into one source (Figure 3A). We then modeled the ancient populations of Georgia and Armenia as temporal groups based on chronology. This analysis shows low levels of ancestry from EEHG already during the EBA (7.7±2.4% and 6.3±1.9%, respectively, where ± denotes one standard deviation), while per-individual models had less power to detect it 8. However, it is only from the Middle Bronze and the Early Iron Ages that EEHG-related ancestry substantially increases to 13±1%, as does the coefficient from Neolithic Anatolia in Georgia (from 16±2.5% to c. 30±1.2% during the Late Antique and Early Middle Ages). Notably, in parallel to the increasing EEHG-related ancestry in the MBA, we also document the first appearance of males with the Y-haplogroup R1b-Z2103 (Supplementary Figure 4, Table S1), a lineage strongly associated with the Yamnaya, Catacomb and North Caucasus culture groups and their expansions such as the Altai Afanasievo 28,32,33.
|
|
|
Post by Admin on Jul 10, 2024 7:12:43 GMT
![](https://www.biorxiv.org/content/biorxiv/early/2024/06/13/2024.06.11.597880/F3.medium.gif) Figure 3. Admixture modeling with qpWave/qpAdm. A. qpAdm coefficients with -1SE are estimated for every period in Georgia and respectively in Armenia, using distal sources from Anatolia/Levant, Iran, the Caucasus (hunter-gatherers) and the W. Eurasian Steppe (Eastern hunter-gatherers). For some groups, adequate models are reached with CHG instead of Chl. Iran and vice-versa therefore, the two groups were merged into one source. Red outlines in the bar plots indicate models of marginal fit (0.01≤p-value<0.05). Periods are in two-to-three-letter abbreviations. B. Linear regression for EEHG-related ancestry in both Armenia and Georgia indicates a decrease in this ancestry since the Middle Bronze Age (MBA). C. Proximal qpAdm models in post-Early Bronze Age (EBA) S. Caucasus. EBA Georgia is used as the main source. With few exceptions, groups from all the periods require additional ancestry both from BA Steppe (here modeled with ‘Steppe_Catacomb’) and Anatolia (modeled here with ‘Turkey_Ikiztepe_LateC’). In the case of Hellenistic Armenia, Anatolia and Iran (‘Iran_Hasanlu_IA’) can adequately model the target with only minor components from EBA Armenia and Steppe. D. Adequate one-to three-source mixture models with qpAdm for the genetic outliers (p-value ≥ 0.05). Different possible combinations were tested informed by PC coordinates, IBD analysis, historical hypotheses, and the assumption that some of these outliers might still derive part of their ancestry locally (see also Table S8). Starting in the Iron Age, EEHG ancestry continuously declined -more substantially in Armenia than in Georgia-but reached equally low levels in modern-day Armenians and Georgians (Figure 3B, Supplementary Figure 2A). Using EBA individuals from the South Caucasus (Armenia and Georgia) to represent the local ancestry, we were able to model all post-EBA groups with additional ancestry from a steppe population north of the Caucasus (‘Russia_Steppe_Catacomb’) and a northeast Anatolian population (‘Turkey_Ikiztepe_LateC’). In the M/LBA, the proportion of the steppe-related ancestry are slightly higher than those from Anatolia, but overall, these models indicate that the observed shift in pc1 and pc2 values from EBA to M/LBA results from two different admixture events from opposite geographical directions and of comparable magnitude. From the Hellenistic period onwards, our models show more extensive gene flow from the south, particularly into Armenia (Figure 3C), as also seen in the increasing Levantine component in the distal modeling. In this line, we also find that Levantine sources (‘Lebanon_Hellenistic.SG’ and ‘Lebanon_ERoman.SG’) rather than Anatolian/Iranian adequately model Hellenistic and Late Antique Armenia (Table S4). We next explored whether genetic diversity spiked in specific periods, which would evidence recent admixture events or high individual mobility. Using qpWave, we find that in every period, some individuals cannot be modeled as being cladal with each other (i.e., genetically similar with respect to a set of reference populations), with an overall proportion of non-cladal models with p-value<0.01 of 10% or more (Supplementary Figure 2B). This proportion is highest for the periods from the Late Antique to the Early Middle Ages, reaching statistical significance in the transitional group when compared to the earliest phases (e.g., Bronze Age groups). Together with the ancestry outliers - not included in these tests - and individuals with increasing affinity to Anatolia, these results highlight that human mobility intensified starting in the Late Antique period. The increased diversity in Y further corroborates this idea (Supplementary Figure 4). We illustrate the results of the qpWave tests as a clustered heatmap of the p-values for pairs from the Hellenistic period to the Early Middle Ages (Supplementary Figure 3A). Two clusters, each consisting of 14 individuals, separate from the remaining 108 and correspond to a substructure also partially captured in the second PCA (Figure 2B). The first cluster (pink square symbols) consists of individuals with a higher affinity to north Caucasus/steppe groups, as indicated by F4-statistics (Table S5). Conversely, the second cluster (orange symbols) suggests that the high affinity to Anatolia/Levant in some individuals already visualized in the first PCA could extend to a cline with mixing southern ancestry into the local gene pool. We note that in both cases, the use of the term ‘cluster’ refers to its members being similarly distinct from the other individuals, but they cannot be treated as genetically homogeneous groups. Overall, the results of PCA and qpAdm/qpWave delineate admixture events that were continuously assimilated into a persisting local gene pool, preserving the overall population structure in the South Caucasus since the Bronze Age with only subtle shifts on the whole-population level.
|
|
|
Post by Admin on Jul 11, 2024 20:31:10 GMT
Connectivity within the South Caucasus and beyond Following the evidence from qpAdm for ‘northern’ and ‘southern’ gene flow into both Georgia and Armenia, we screened for Identical by Descent (IBD) sharing among ancient genomes using the software ancIBD (see Methods). First, we show that the average IBD sharing rate within Georgia and Armenia and between them overlap and continuously decay with geographical distance at the same rate (Figure 4A). This continuous “isolation-by-distance” pattern 34 suggests that - despite the different political and territorial influences between present-day Armenia and Georgia in the past - the populations mixed at a rate proportional to their geographical distance. By screening IBD sharing of all pairs containing one ancient Georgian individual from our dataset and another individual buried up to 1500 km away, we quantify the connections with the rest of Southwest Asia (‘South’) and Eastern Europe to Central Asia (‘North’) (Figure 4B). At shorter geographical distances (i.e., 100-240 km), we observe an increased rate of short IBD sharing with the south (i.e., Armenia), corroborating increased population connectivity within the South Caucasus compared to across the Caucasus, which therefore likely served as a barrier to mobility. However, at larger distances, rates of IBD sharing equalize between south and north, reflecting the global admixture proportions estimated with qpAdm and the rates of outliers whose ancestry is traced on either side. These signals indicate that the Caucasus mountain range was not a lasting barrier to long-distance mobility. ![](https://www.biorxiv.org/content/biorxiv/early/2024/06/13/2024.06.11.597880/F4/graphic-4.medium.gif) Figure 4. Spatial distribution of IBD connections for Georgia and Armenia. A. Sum of IBD counts for all pairs including Georgia and Armenia (c. 20,000) with IBD sum ≤ 200cM and date difference of up to 1000 years plotted in bins of increasing geographic distance. B. Distant connection between ancient individuals from Georgia and populations North and South of the Greater Caucasus presented as the mean count of IBD ≥ 8cM and geographical distance from c. 100 to 1500 km. Pairs with a date difference of more than 1000 years were removed, retaining a total of c. 35,000 pairs. The 95% confidence interval for every bin was calculated with the embedded bootstrap function ‘mean_cl_boot’ of ‘stat_summary’ in R (ggplot). C. Map of the inter-site IBD connections (≥12cM) for Georgia in early periods (Early Bronze Age to Iron Age II). The size of the symbols indicates the sample size of each site, while the line thickness represents the number of IBD pairs between two sites. D. Map of the inter-site IBD connections (≥12cM) for Georgia in late periods (Early Antique to Early Middle Ages). Dashed lines indicate the IBD connections between early and late periods. The IBD connections that were identified between the three closely located sites, Aragvispiri, Dzinvali, and Nedzikhi, have been filtered out.
|
|
|
Post by Admin on Jul 15, 2024 20:30:20 GMT
Varying population dynamics between rural settlements and urban centers We then analyzed IBD sharing ≥12cM long, which evidences recent genealogical links because IBD segments of this length originate from relatives connected within several hundred years 35. We observed that such sharing was common across all periods and distances (Figure 4C and 4D). However, we detected remarkable local patterns. The dense sampling in eastern Georgia - the area corresponding to the eastern territory of the Kingdom of Iberia - and the distribution of IBD connections in the Late Antique and Early Middle Ages allows us to investigate two groups of sites with contrasting niches of different social organization. We present a network analysis of the IBD segments connecting Samtavro, Samshvilde, and Fiqris Gora, named in order of importance as urban centers and pilgrimage sites, while the sites of Nedzikhi, Dzinvali, and Aragvispiri are all associated with rural settlements located within a very short distance from each other (Figure 5A). ![](https://www.biorxiv.org/content/biorxiv/early/2024/06/13/2024.06.11.597880/F5.medium.gif) Figure 5. Download figureOpen in new tabFigure 5. ROH and IBD network analyses. A. Network of IBD connections in Eastern Georgia where sampling is particularly dense. In the earlier period (Iron Age to early Antiquity), only individuals from the site of Kamarakhevi exhibit a high rate of intra-IBD connections within and also with sites across Georgia as shown in the panel. B. During the Late Antique and Early Middle Ages, there was an asymmetry in IBD distributions caused by some sites where all individuals from within and between sites associated with settlements of rural activity are distantly related. In contrast, nearby sites with urbanism like Samtavro exhibit minimal IBD within but higher with more distant sites as shown in panel B. C. Runs of homozygosity from the sites in the IBD network suggest a correlation of IBD connections with consanguinity revealing a contrast of more endogamic rural communities (with both ample long ROH and intra-site IBD sharing) living side-by-side with large cosmopolitan populations from urban centers in the Late Antique/Early Middle Ages (with little long ROH and intra-site IBD sharing). However, the overall levels of consanguinity have been stable over time (leftmost panel). Samtavro, located in the old capital and the urban site Mtskheta, shows the lowest rate of internal IBD sharing, with only 4 of 14 individuals connected via IBD. Overall, intrasite genetic relatedness estimated with ancIBD and BreadR is exceptionally low in Samtavro, with only two pairs of parent-offspring and no pairs of distant relatives up to the sixth degree (Figure 5A, Tables S1 and S6). Remarkably, SMT005, the individual with the most intersite connections, is also the only one from Samatvro with a signal of consanguinity (sum ROH>200cM; Figure 5C; Table S7). In addition to the high genetic diversity within the site shown with PCA, qpWave clustering, and the lack of internal biological relatedness, Samtavro displays the highest rate of IBD sharing with four sites at larger geographic distances (Figure 5B). Conversely, Dzinvali and especially Nedzikhi’s IBD sharing already declined at a short geographic radius. Taken together, these signals suggest that the urban scale of Samtavro was also reflected in the organization of its necropolis, whereas in nearby settlements, people maintained a social structure of smaller and, to some extent, endogamic communities. We observe signals of markedly different social organizations outside the urban sites as early as the Iron Age. In the site of Kamarakhevi, located next to cosmopolitan Samtavro, as many as nine of 11 individuals share IBD between them (Figure 5A). One of them presents one of the highest amounts of long ROH in the ancient DNA record, equivalent to parents related as either first or second-degree relatives. Continuing into the Late Antique and Early Middle Ages, all eight individuals from the rural site Nedzikhi show pairwise IBD connections ranging from the second to the tenth degree. We also show that four individuals from Nedzikhi exhibit short to intermediate runs of homozygosity, consistent with a population with substantial background relatedness. The remaining three have long ROH equivalent to close-kin unions (e.g., first and second cousins), indicating that all individuals sampled from this site represent an endogamous community. At rural Dzinvali - coeval and at a close distance to Nedzikhi -, the absence of ROH in half of the individuals suggests that the group was part of a larger mating pool or an endogenous community with new members joining. The qpWave and PCA results (Supplementary Figure 3) further support this idea: Individuals DZN002, DZN009, and DZN011, who are not related to anyone via IBD sharing, cluster separately from the Late Antique group due to genetic affinities from either Anatolia or possibly north of the Caucasus. The same applies to FQR002 from urban Fiqris Gora, one of the individuals with the highest affinity to Anatolia, which contrasts with the IBD-connected and consanguineous individual FQR006. Overall, the levels of consanguinity, defined as the proportion of close (up to first cousins) and distant (second cousins or equivalent) paternal relationships, remain constant through time at a rate of ∼25-30% (Figure 5C), which is elevated compared to the ancient DNA record 36. Precisely, we observe consanguinity in 16 out of 125 individuals analyzed for long ROH (Methods). The frequency in Samtavro is lower (1/25); however, the frequency of individuals consistent with distant parental relatedness (5/20) is comparable to the levels in the overall ancient post-Iron Age population (18/81). The decline in short ROH from a maximum in the Bronze Age pastoralists to a minimum in the Late Antique and Early Middle Ages shows a general trend towards locally larger population sizes, particularly in urban areas, while smaller communities continued to exist.
|
|