Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Jul;643(8070):139-147.
doi: 10.1038/s41586-025-08913-3. Epub 2025 Apr 23.

Punic people were genetically diverse with almost no Levantine ancestors

Harald Ringbauer  1   2   3 , Ayelet Salman-Minkov  4   5 , Dalit Regev  6 , Iñigo Olalde  7   8   9 , Tomer Peled  4 , Luca Sineo  10 , Gioacchino Falsone  11 , Peter van Dommelen  12 , Alissa Mittnik  7   13   14 , Iosif Lazaridis  7   15 , Davide Pettener  16 , Maria Bofill  17 , Ana Mezquida  17 , Benjamí Costa  17 , Helena Jiménez  17 , Patricia Smith  18 , Stefania Vai  19 , Alessandra Modi  19 , Arie Shaus  7   20   21 , Kim Callan  15   22 , Elizabeth Curtis  15   22 , Aisling Kearns  15 , Ann Marie Lawson  15   22 , Matthew Mah  15   22   23 , Adam Micco  15 , Jonas Oppenheimer  15   22 , Lijun Qiu  15   22 , Kristin Stewardson  15   22 , J Noah Workman  15 , Nicholas Márquez-Grant  24 , Antonio M Sáez Romero  25 , María Luisa Lavado Florido  26 , Juan Manuel Jiménez-Arenas  27 , Isidro Jorge Toro Moyano  28 , Enrique Viguera  29 , José Suárez Padilla  29 , Sonia López Chamizo  29 , Tomas Marques-Bonet  30   31   32   33 , Esther Lizano  34   35   36 , Alicia Rodero Riaza  37 , Francesca Olivieri  38 , Pamela Toti  39 , Valentina Giuliana  40 , Alon Barash  41 , Liran Carmel  42 , Elisabetta Boaretto  43 , Marina Faerman  44 , Michaela Lucci  45 , Francesco La Pastina  10   11   45 , Alessia Nava  46 , Francesco Genchi  47 , Carla Del Vais  48 , Gabriele Lauria  10 , Francesca Meli  11 , Paola Sconzo  11 , Giulio Catalano  10 , Elisabetta Cilli  49 , Anna Chiara Fariselli  49 , Francesco Fontani  13   14   49 , Donata Luiselli  49 , Brendan J Culleton  50 , Swapan Mallick  15   22   23 , Nadin Rohland  7   15   23 , Lorenzo Nigro  51 , Alfredo Coppa  52   53   54 , David Caramelli  19 , Ron Pinhasi  52   55 , Carles Lalueza-Fox  30   56 , Ilan Gronau  57 , David Reich  58   59   60   61   62
Affiliations

Punic people were genetically diverse with almost no Levantine ancestors

Harald Ringbauer et al. Nature. 2025 Jul.

Abstract

The maritime Phoenician civilization from the Levant transformed the entire Mediterranean during the first millennium BCE1-3. However, the extent of human movement between the Levantine Phoenician homeland and Phoenician-Punic settlements in the central and western Mediterranean has been unclear in the absence of comprehensive ancient DNA studies. Here, we generated genome-wide data for 210 individuals, including 196 from 14 sites traditionally identified as Phoenician and Punic in the Levant, North Africa, Iberia, Sicily, Sardinia and Ibiza, and an early Iron Age individual from Algeria. Levantine Phoenicians made little genetic contribution to Punic settlements in the central and western Mediterranean between the sixth and second centuries BCE, despite abundant archaeological evidence of cultural, historical, linguistic and religious links4. Instead, these inheritors of Levantine Phoenician culture derived most of their ancestry from a genetic profile similar to that of Sicily and the Aegean. Much of the remaining ancestry originated from North Africa, reflecting the growing influence of Carthage5. However, this was a minority contributor of ancestry in all of the sampled sites, including in Carthage itself. Different Punic sites across the central and western Mediterranean show similar patterns of high genetic diversity. We also detect genetic relationships across the Mediterranean, reflecting shared demographic processes that shaped the Punic world.

PubMed Disclaimer

Conflict of interest statement

Competing interests: The authors declare no competing interests.

Figures

Extended Data Figure 1:
Extended Data Figure 1:. PCA of Punic individuals grouped according to site and chronology.
We project individuals sequenced on over 20,000 SNPs onto the same two PCs as in Figure 1 (calculated from modern individuals, gray dots). Here, we plot all individuals from Punic archeological sites, excluding the Phoenician site of Akhziv and including the 20 individuals dated to the Roman period (Supplementary Table 5). We split the sample into panels representing our four major geographic regions: Iberia, Sardinia, North Africa, and Sicily. The shape of the symbols indicates the site (lower left legend), and the color indicates the date range of each individual (as described in the upper right legend).
Extended Data Figure 2:
Extended Data Figure 2:
Ancestry models inferred for 122 Phoenician-Punic individuals using unsupervised ADMIXTURE with K=2, 3, 4 and 5 latent ancestry components. The 122 Phoenician-Punic individuals sequenced for more than 100,000 SNPs were jointly analyzed with 24 individuals from related ancient populations across the Mediterranean (Supplementary Table 12). (a) The model with the highest likelihood was obtained for each value of K among 50 replicate runs. Individuals are grouped based on region. Values of the ΔK score of are specified for K=3,4, with a higher score obtained for K=3, suggesting optimal fit. (b) A more detailed depiction of the best model obtained with K=3 latent ancestry components corresponding to North African ancestry (red), eastern ancestry (green), and central/western Mediterranean ancestry (blue). Individuals are partitioned within each region according to site and time range (see legend). The unsupervised ADMIXTURE model does not adequately differentiate between Levantine ancestry and ancestry found in other Mediterranean locations (e.g., Anatolia and Sicily), unlike the qpAdm models of Extended Data Figure 3.
Extended Data Figure 3:
Extended Data Figure 3:
Ancestry models inferred for the 140 Phoenician-Punic individuals in our data set by qpAdm. We partitioned individuals by region: (a) North Africa, (b) Sicily, (c) Sardinia, (d) Iberia, and (e) the Levant. Within each region, we grouped individuals by site, and for sites in Sicily, Sardinia, and Iberia, also by broad date ranges (see legend for color code). We ordered the models of each individual according to their P-values (gray bar above each model). We report P-values assuming that the LRT statistic is chi-squared distributed with degrees of freedom determined by the number of populations and of contributing source populations. We did not correct these P-values for multiple testing, but this approach is conservative since we report models with comparatively high P-values (those that are not rejected by the test). Individuals with low coverage (fewer than 100,000 SNPs) are indicated by an asterisk (*) next to the sample ID. Eastern ancestry models are indicated by a contribution of the proxy sources Levant MLBA. In contrast, western ancestry models are indicated by contributions from either Greece BA (Myc), Sicily EBA, Sardinia LBA, Iberia LBA, or Steppe MLBA. There are five individuals for whom no valid eastern or western model was inferred. For four of them, we inferred valid models under the broad ancestry scheme (marked by an asterisk above the vertical bar), and for one (I22122 from Tharros, Sardinia), we could not infer any valid model.
Extended Data Figure 4:
Extended Data Figure 4:. Proportions of North African ancestry inferred using the 2D PCA and qpAdm for 123 Punic individuals.
We exclude from this analysis the Akhziv sample, the three individuals from Sicily and Sardinia that cluster near Levantine individuals in the 2D PCA, and one individual for which we could not fit a qpAdm model. The qpAdm estimates are based on the smallest proportions estimated for the individual in a valid qpAdm model (see Methods). The PCA-based estimates of North African ancestry were computed by projecting the location of each sample in the PCA onto a cline from the cluster defined by Bronze Age individuals from Sicily to the cluster defined by North African individuals (see Supplementary Information S3 for more details). The two approaches yield similar estimates, with qpAdm being more sensitive to low ancestry proportions. Individuals from Kerkouane (depicted as squares) appear to have a broad range of North African ancestry (0 – 94%). Individuals from Sicily typically have lower proportions of North African ancestry (<20%), and we observe no significant shift in time. On the other hand, in Sardinia, none of the 12 individuals for which we inferred more than 10% North African ancestry (according to at least one of the two approaches) dated before 400 BCE, suggesting that North African ancestry was likely introduced around that time (Supplementary Information S3). We see a similar pattern in Iberia, but since we only have one individual from Iberia dating before 400 BCE, we cannot confidently infer the absence of North African ancestry during this time.
Extended Data Figure 5:
Extended Data Figure 5:. Y Haplogroup Diversity in male individuals from Phoenician and Punic contexts.
We inferred the first four characters of the ISOGG 2019 Y haplogroup classification for all Phoenician and Punic males with more than 100,000 autosomal SNPs covered (as those in almost all cases have sufficient coverage on the Y chromosome; see Methods). (a) Pie chart of Y haplogroup frequencies. (b) We visualize the Y Haplogroup diversity partitioned per Phoenician or Punic site and denote each individual’s haplotype by one circle. We set the height of the bar to the overall frequency (as depicted in panel a). The numbers in brackets indicate the total Y haplotype sample size.
Extended Data Figure 6:
Extended Data Figure 6:. Autosomal and Y Diversity without the contribution of filtered North African ancestry, per site in Phoenician-Punic contexts and the published aDNA record.
(a) Y haplogroup diversity measured using the Inverse Simpson index. This value is computed as in Figure 3a, excluding the three Punic individuals (from Kerkouane, Villaricos and Selinunte) with distinct North African Y haplogroups E1a and L (see Extended Data Figure 5). (b) autosomal diversity measured using the first two PCs from Figure 1 and the mean pairwise distance of those coordinates. This value is computed as in Figure 3b, excluding individuals with more than 10% North African ancestry based on qpAdm in Phoenician-Punic sites (see Extended Data Figure 4). Here, we combined individuals from the nearby Sicilian sites of Birgi, Motya, and Lilybaeum into one group (labeled Lilybaeum here). In both panels, the diversity measures for the context populations are as in Figure 3 (without any additional filtering), and the dashed horizontal bar in both panels indicates the maximum diversity observed in sites dating before 500 BCE. See Supplementary Information S5 for a more detailed description of this analysis.
Extended Data Figure 7:
Extended Data Figure 7:. Two reconstructed pedigrees of Punic individuals.
We reconstructed two pedigrees based on inferring biological relatives with pairwise kinship (using IBD segment sharing) and uniparental haplogroups: (a) A pedigree linking five individuals from Kerkouane, North Africa; (b) A pedigree linking three individuals from Tharros, Sardinia. In the Kerkouane pedigree in (a), individuals I24215 and I24194 are inferred to be 3rd-4th degree relatives of the two siblings I24494 and I24193, but the exact pedigree relationship cannot be resolved. Each panel depicts the projection of the related individuals onto the two major PCs used in Figure 1. Each pedigree specifies the sample IDs for all individuals, the mitochondrial (maternal) haplogroup and the Y (paternal) haplogroup for males. Both pedigrees contain individuals dating to 800–400 calBCE and link several individuals via the maternal lineage: We infer four identical maternal haplogroups in Kerkouane and a maternal grandfather in Tharrosーtwo observations that are inconsistent with strict patrilocality.
Extended Data Figure 8:
Extended Data Figure 8:. Runs of homozygosity inferred in Phoenician and Punic individuals and ancient individuals of relevant Bronze and Iron Age contexts.
We computed runs of homozygosity (ROH) in all individuals with more than 400,000 SNPs covered and recorded the total length (in cM) of ROHs binned by length into four categories (see legend). We label individuals with at least 50 and 100 cM of their genome in long ROH (> 20cM) with triangle and square marks as in - to indicate offspring of close biological parental relatives. (a) ROH in Phoenician and Punic individuals, grouped by site. (b) ROH in individuals from relevant Bronze and Iron Age contexts (as depicted in Figure 1). (c) Expected ROH for offspring of various cousin matings (according to the degree of relation between parents) and for individuals sampled in populations with small effective size (calculated as described in).
Extended Data Figure 9:
Extended Data Figure 9:
Ancestry models inferred using qpAdm for individuals from Sicily from (a) the indigenous Iron Age sites of Polizzello and Monte Falcone, (b) from Phoenician sites before Roman expansion (as shown in Extended Data Figure 3b), and (c) from Punic sites after Roman expansion. Color horizontal bars indicate radiocarbon dates. The models of each individual are sorted according to their P-values (gray bar above each model). We report P-values assuming that the LRT statistic is chi-squared distributed with degrees of freedom determined by the number of populations and of contributing source populations. We did not correct these P-values for multiple testing, but this approach is conservative since we report models with comparatively high P-values (those that are not rejected by the test). Eastern ancestry models are indicated by a contribution of the proxy source Levant MLBA. In contrast, western ancestry models are indicated by contributions from either Greece BA (Myc), Sicily EBA, Sardinia LBA, Iberia LBA, or Steppe MLBA. There are seven individuals for which no valid eastern or western model was inferred. We inferred valid models under the broad ancestry scheme (marked by an asterisk above the vertical bar) for five of them. Two individuals were inferred to be related through IBD-sharing and are indicated in the figure. The analysis suggests that indigenous populations in Sicily have similar ancestry patterns as observed in the Phoenician sites but without North African ancestry. In later periods, we see the introduction of diverse ancestry sources (Levantine and western Mediterranean), likely associated with the Roman expansion into Sicily. See Supplementary Information S3 for more details.
Extended Data Figure 10:
Extended Data Figure 10:. Additional PCA projections.
(a) Bronze and Iron Age reference and Levantine populations. We show the same PCA as in Figure 1 but focus on the ancient reference populations. (b) Zoom in PCA projections of Levantine populations. We show the same PCA depicted in (a), but zooming into the region where Levantine individuals project. We also include additional Bronze and Iron Age Levant individuals not included in Figure 1. Those previously published individuals originate from Sidon in present-day Lebanon and various sites in present-day Israel (Megiddo, Yehud, Hazor, Baq’ah, Tel Shadud, Ashkelon,). Abbreviations: MLBA: Middle-Late Bronze Age, MBA: Middle Bronze Age, IA: Iron Age. All 13 individuals from Akhziv cluster next to other Levantine individuals, together with a single outlier individual from Tharros (I22119) inferred to have Levantine ancestry (Extended Data Figure 3). Abbreviations: M/N: Mesolithic/Neolithic, MLBA: Middle-Late Bronze Age, MBA: Middle Bronze Age, LBA: Late Bronze Age, IA: Iron Age
Figure 1:
Figure 1:. Sample Overview and Principal Component Analysis.
(a) Locations of archaeological sites for which we analyzed aDNA. We indicate the number of individuals with newly generated aDNA data (>20,000 SNPs) and high-confidence archaeological association (see Methods, Supplementary Information S2) for each site. We mark the numbers of previously published ancient individuals with a “*” and depict locations of aDNA context using circles. (b) Principal component analysis (PCA). We first computed PCs using 1196 present-day individuals from Western Eurasia and North Africa (Methods). We then projected 128 individuals with >20,000 SNPs covered and confidently assigned to a Phoenician-Punic context (see Methods, Supplementary Table 5) onto the first two PCs. For context, we also projected various Bronze and Iron Age individuals across the Mediterranean (colored circles matching panel a, Supplementary Table 6). Extended Data Figure 1 depicts the PCA projections broken up according to archeological sites and dates. Extended Data Figure 10 depicts a projection of the reference populations only and a zoomed-in version of the Levantine individuals. Abbreviations: BA - Bronze Age, IA - Iron Age, M/N - Meso/Neolithic, L - Late, M - Middle, E - Early.
Figure 2:
Figure 2:. Ancestry models inferred using qpAdm.
The figure depicts the representative admixture models for 122 Phoenician-Punic individuals in our data set sequenced at more than 100,000 SNPs. We sorted individuals by region (Levant, North Africa, Sicily, Sardinia, and Iberia), then by site, estimated date range (gray horizontal bars above the site names), and, finally, by inferred North African ancestry. We combined the ancestry proportions inferred for Greece BA (Myc) and Sicily EMBA (Sicilian-Aegean ancestry), and the ancestry proportions inferred for Sardinia LBA and Iberia EBA (western Mediterranean ancestry). We combined these ancestries due to the limited ability of our qpAdm models to distinguish between them (Supplementary Information S3). Since individuals typically had several valid admixture models, we selected the model that maximized Sicilian-Aegean or Levantine ancestry (Methods). For three individuals (marked by an asterisk), we could only fit a broad ancestry model (Methods), and one individual from Tharros could not be fit by any of the ancestry models we considered. The figure highlights a pair of related individuals from Birgi and Kerkouane (see Figure 4) and related individuals buried in the same tomb in Villaricos (see Figure 5). The complete set of valid models for all 140 individuals in the data set (including those with fewer than 100,000 SNPs) is specified in Extended Data Figure 3.
Figure 3:
Figure 3:. Autosomal and Y Diversity per site in Phoenician-Punic contexts and the published aDNA record.
(a) We calculated the Y haplogroup diversity for contexts with at least five males with sufficient data using Y haplogroups at the level of the first four characters of the ISOGG classification and the Inverse Simpson index, also known as the effective number of types (see Methods). (b) We calculated the 2D PCA diversity for contexts with at least ten individuals with sufficient data using the first two PCs from Figure 1 and the mean pairwise distance of those coordinates (see Methods, values listed in Supplementary Information S5). The dashed horizontal bar in both panels indicates the maximum diversity (autosomal or Y-based) observed in sites before 500 BCE. The Mediterranean Bronze Age (BA) and Iron Age (IA) groups (diamonds) correspond to the ones depicted in Fig. 1. In each of these five groups, we included individuals from different archaeological sites.
Figure 4:
Figure 4:. IBD segments shared between individuals from Sicily and North Africa indicate they were 5–7th degree biological relatives.
(a) The sampling locations and radiocarbon date estimates of the two individuals. (b) Genomic location of the position of three long IBD segments on the 22 autosomes. (c) Posterior of not being in IBD along chromosomes 1,9, and 17 as calculated with ancIBD (red). We also visualize opposing homozygotes (upper versus lower points: yes or no) for all SNPs where the genomes of the two individuals have an imputed genotype posterior probability greater than 0.99. The dark blue segments indicate the inferred IBD segments. Our analysis suggests that both individuals had ca. 85% Sicilian-Aegean ancestry and 15% North African ancestry (highlighted in Figure 2). We list all pairs of related Punic individuals in Supplementary Table 8, including five additional pairs linking sites separated by the Mediterranean Sea.
Figure 5:
Figure 5:. A Punic tomb from Villaricos containing remains of an endogamous community.
(a) We show the mean sequencing depths for five individuals from tomb 774 in Villaricos (Iberia) for which we obtained aDNA (we indicate coverage > 0.9x by large squares). (b) This PCA plot (as in Figure 1) shows that the five individuals cluster close to Bronze Age Mycenaeans. (c) We depict the estimated kinship coefficients and their 95% confidence intervals (calculated using average pairwise mismatch rates) for all ten pairs of individuals (Methods), revealing a pattern of second-to-third-degree relatives. For the three individuals sequenced at sufficient coverage for IBD analysis, we specify the total length of long IBD segments (>12cM) shared by the three pairs, confirming that those pairs were closely related. (d) The three individuals with high coverage exhibit exceptionally high levels of ROH, close to the level expected for the offspring of first cousins.
See this image and copyright information in PMC

References

    1. Aubet ME The Phoenicians and the West: Politics, Colonies and Trade. (Cambridge University Press, 2001).
    1. Bondì SF, Garbati G, Botto M & Oggiano I Fenici e cartaginesi: una civiltà mediterranea. (Istituto poligrafico e Zecca dello Stato, Libreria dello Stato, 2009).
    1. López-Ruiz C Phoenicians and the Making of the Mediterranean. (Harvard University Press, 2022).
    1. Regev D Painting the Mediterranean Phoenician: On Canaanite-Phoenician Trade-Nets. (Equinox Publishing; (UK: ), 2021).
    1. Roppa A, Botto M & Van Dommelen P Il Mediterraneo Occidentale dalla fase fenicia all’egemonia cartaginese. Dinamiche insediative, forme rituali e cultura materiale nel V secolo a. C. Ediz. italiana, spagnola e inglese. (2021).

Publication types

Supplementary concepts

LinkOut - more resources