Palaeogenomic insights into the origins of early settlers on the island of Cyprus

Abstract

Archaeological evidence supports sporadic seafaring visits to the Eastern Mediterranean island of Cyprus by Epipaleolithic hunter-gatherers over 12,000 years ago, followed by permanent settlements during the early Neolithic. The geographical origins of these early seafarers have so far remained elusive. By systematically analysing all available genomes from the late Pleistocene to early Holocene Near East (c. 14,000-7000 cal BCE), we provide a comprehensive overview of the genetic landscape of the early Neolithic Fertile Crescent and Anatolia and infer the likely origins of three recently published genomes from Kissonerga-Mylouthkia (Cypriot Late Pre-Pottery Neolithic B, c. 7600-6800 cal BCE). These appear to derive roughly 80% of their ancestry from Aceramic Neolithic Central Anatolians residing in or near the Konya plain, and the remainder from a genetically basal Levantine population. Based on genome-wide weighted ancestry covariance analysis, we infer that this admixture event took place roughly between 14,000 and 10,000 BCE, coinciding with the transition from the Cypriot late Epipaleolithic to the Pre-Pottery Neolithic A (PPNA). Additionally, we identify strong genetic affinities between the examined Cypro-LPPNB individuals and later northwestern Anatolians and the earliest European Neolithic farmers. Our results inform archaeological evidence on prehistoric demographic processes in the Eastern Mediterranean, providing important insights into early seafaring, maritime connections, and insular settlement.

Figure 1

Geographical location and chronology of ancient samples used in the present study. (a) Geographical location of Epipaleolithic/Mesolithic and initial Neolithic archaeological sites in West Eurasia from which ancient samples were included in the present study. Sites and the corresponding samples are colour-coded and annotated by region and chronological period. (b) Chronology range (years BCE) and number of included individuals for all archaeological sites analysed, using the same colour-coding for regions as in panel a. Chronology abbreviations next to regions are as presented in Table 1. Country abbreviations can be found in Supplementary Information (Detailed Methods). The above information can be found in tabular form in Supplementary Tables 1 and 2.

Figure 2

Principal Components Analysis (PCA) plot displaying all ancient samples analysed in the present study, projected on modern West Eurasian genetic variation. Ancient samples are presented as coloured data points, projected on modern West Eurasian populations (light grey points). The main panel displays the first two principal components with all Epipaleolithic/Mesolithic hunter-gatherer and initial Neolithic farming populations analysed in the present study. EpiP/Meso populations can be seen at the edges of the plot (Central Zagros / Caucasus on the bottom left, North African Late Palaeolithic on the top left, Balkan Mesolithic on the bottom right). Anatolian (yellow triangle) and Levantine (brown triangles) Epipaleolithic groups, appear more centrally in the plot and have high genetic proximity to Pre-Pottery Neolithic (PPN) populations from the same regions. Anatolian Neolithic populations form two distinct clusters, the first one appearing closer to the Epipaleolithic Pınarbaşı HG and comprises Central Anatolian PPN Boncuklu and Aşıklı Höyük, Marmara PN groups and Very Early European Farmers (VEEF), as well as Cypro-LPPNB. The second Anatolian cluster shows a shift towards Levantine EpiP/PPN populations and comprises Central Anatolian PPN Musular and PN population groups. Upper Mesopotamian PPN population groups show a clear shift towards Zagros Mesolithic/Neolithic populations. A zoomed-in version of the first Anatolian cluster seen as a separate panel within the main panel, displays Cypro-LPPNB (larger orchid diamonds) clustering closely with Anatolian Marmara PN groups (yellow squares), as well as the earliest Neolithic farmers of Europe (red squares). One Cypro-LPPNB sample (I4210) clusters entirely with PN Anatolian and EN European farmers, while the other two (I4207, I4209) very slightly shift towards Levantine groups, still however clearly belonging to the Anatolian cluster. All abbreviations in the plot as in Table 1.

Figure 3

Distal genetic composition of ancient samples analysed in the present study in a spatiotemporal context. (a) Admixture weights derived using qpAdm, from a rotating model including five major, genetically distinct, ancestral Epipaleolithic/Mesolithic populations (Late Palaeolithic North African Iberomaurusian HGs, Epipaleolithic Central Anatolian HGs, Epipaleolithic Levantine Natufian HGs, Mesolithic Zagros/Caucasus HGs, Mesolithic Balkan HGs), separated by archaeologically defined periods and broad regions (Pre-Pottery Neolithic Near East, Pottery Neolithic Near East, Early Neolithic Europe), with chronological range of thus grouped samples in brackets (b) Bar plot displaying the same admixture weights grouped by region. Pre-Pottery Neolithic Near Eastern populations appear to derive all their ancestry from three of the five ancestral populations presented in panel a (Epipaleolithic Central Anatolian HGs, Epipaleolithic Levantine Natufian HGs, Mesolithic Zagros/Caucasus HGs). The Cypro-LPPNB group appears to derive 68% of its ancestry from Epipaleolithic Central Anatolians, 20% from Epipaleolithic Levantine Natufians, and 12% from Mesolithic/Neolithic Zagros. Anatolian PPN and PN population groups derive the majority of their ancestry from Epipaleolithic Central Anatolia (Pınarbaşı HG), with Central Zagros basal admixture apparent in PPN groups and additional Levantine basal admixture particularly apparent in PN groups. Levantine PPN population groups derive the majority of their ancestry from Epipaleolithic Natufians. Upper Mesopotamian and Northwest Zagros PPN population groups derive roughly equal ancestry from the three major components mentioned above, with excess basal Central Zagros admixture observed in the Northwest Zagros groups. Central Zagros PPN population groups share all of their ancestry with an earlier forager (Hotu IIIb) from the same region. The earliest European farmers have a similar profile to contemporaneous Northwest Anatolians. (c) Biplot of basal Levant to basal Zagros admixture, which helps differentiate PPN population groups, based on whether they have an excess Levantine relative to Zagros admixture or vice versa. Cypro-LPPNB belong to the former category. All abbreviations in the plot as in Table 1. All qpAdm models showing acceptable fit for all tested ancient populations, including exact admixture weights and detailed fit statistics, can be found in tabular form in Supplementary Table S3.

Figure 4

Outgroup f3-statistics displaying shared genetic drift between all ancient samples analysed in the present study and Cypro-LPPNB samples in a spatiotemporal context. (a) The four-panel map displays outgroup f3-statistics of the form f3(Mbuti; Cypro-LPPNB, comparison ancient population) estimating shared genetic drift, based on allele sharing. Each analysed population is displayed in its corresponding location on the map, with the colouring of the symbols representing the genetic proximity between the corresponding ancient population and Cypro-LPPNB farmers (orchid diamond). The darker the colour, the higher the pairwise f3 statistic between Cypro-LPPNB and the corresponding ancient population group, signifying higher allele sharing. (b) Actual value of each pairwise f3 statistic (± 3 standard errors) between Cypro-LPPNB and all tested ancient populations. The further to the right the points are in each plot, the higher the allele sharing with Cypro-LPPNB. Wider error bars indicate lower precision, as a result of smaller number of available SNPs in the given pairwise combination. Colour-coding represents geographical regions as denoted in Fig. 1. Cypro-LPPNB show high allele sharing with Central Anatolian Epipaleolithic Pınarbaşı HG and Central Anatolian PPN Boncuklu and Aşıklı Höyük, Northwest Anatolian PN Marmara groups, and the earliest Neolithic European farmers. f3 results with low precision (e.g. HotuIIIb) should be interpreted with caution, due to low number of available SNPs. All abbreviations in the plot as in Table 1. The displayed information can be found in tabular form in Supplementary Table S4. (c) MDS plot based on pairwise ‘1 minus outgroup f3-statistics’ of the form f3 (Mbuti; test, comparison). Distances between data points represent genetic differentiation based on allele sharing among Mesolithic/Epipaleolithic and Pre-Pottery Neolithic Near Eastern population groups analysed in the present study. Cypro-LPPNB Mylouthkia, cluster in the middle of a genetic cline between Epipaleolithic and early Neolithic Anatolians (Pınarbaşı HG, Boncuklu and Aşıklı Höyük) and contemporaneous Levantine groups. An apparent close genetic proximity between Mylouthkia and Anatolian Musular, is not supported by qpAdm analyses, due to the much higher admixture from Zagros-related sources among the latter. All abbreviations in the plot as in Table 1. A matrix of raw genetic distances of the form 1 minus f3 can be found in Supplementary Table S5.

Figure 5

Proximal genetic composition of Cypro-LPPNB farmers in a spatiotemporal context, estimated from qpAdm analysis, with all major Near Eastern Epipaleolithic and PPN populations (potentially ancestral to Cypro-LPPNB) as sources. (a) The presented proximal admixture composition comprising 83% ancestry from Central Anatolian Boncuklu and 17% from Epipaleolithic Levantine Natufians, provides the best fit from all tested models. The denoted regions (Konya plain and Levant) are approximations and not meant to precisely represent the actual archaeological regions. Similarly, arrows aim to provide an overall understanding of the possible migration flow from the mainland to the island of Cyprus and are not meant to represent exact departure and arrival locations. All qpAdm models involving potentially ancestral sources for Cypro-LPPNB as target, including exact admixture weights and detailed fit statistics, can be found in tabular form in Supplementary Table S7. (b) The four best fitting two-way proximal admixture models for Cypro-LPPNB as target. The first and second models present an admixture profile primarily deriving from Aceramic Neolithic Central Anatolian Boncuklu and only differ in the Levantine source detected (Epipaleolithic Natufian vs PPNB), while the third and fourth pick up Aceramic Neolithic Aşıklı Höyük and Epipaleolithic Pınarbaşı, respectively, as possible Anatolian sources, with additional input from Levantine groups. Further admixture timing analysis based on genome-wide weighted ancestry covariance favours the first model over the other three (Supplementary Table S11). (c) Inferred admixture timing using DATES, revealing a wide time range for the event giving rise to Cypro-LPPNB Mylouthkia (14,000 ± 4,000 years BCE), due to the low coverage of the relevant genomes. The same analysis reveals an admixture timing of 12,500 ± 1,700 years BCE for Aceramic Neolithic Anatolian Boncuklu. Combining these two findings and provided that all lines of evidence generated in the present study point to a Boncuklu-related group as the major ancestral source for Cypro-LPPNB Mylouthkia, it could be inferred that the admixture event for the latter took place roughly between 14,000 and 10,000 BCE (grey shaded time range), during the late Epipaleolithic or the transition to the Pre-Pottery Neolithic A (PPNA).