Ancient DNA of Phoenician remains indicates discontinuity in the settlement history of Ibiza

Ibiza was permanently settled around the 7th century BCE by founders arriving from west Phoenicia. The founding population grew significantly and reached its height during the 4th century BCE. We obtained nine complete mitochondrial genomes from skeletal remains from two Punic necropoli in Ibiza and a Bronze Age site from Formentara. We also obtained low coverage (0.47X average depth) of the genome of one individual, directly dated to 361–178 cal BCE, from the Cas Molí site on Ibiza. We analysed and compared ancient DNA results with 18 new mitochondrial genomes from modern Ibizans to determine the ancestry of the founders of Ibiza. The mitochondrial results indicate a predominantly recent European maternal ancestry for the current Ibizan population while the whole genome data suggest a significant Eastern Mediterranean component. Our mitochondrial results suggest a genetic discontinuity between the early Phoenician settlers and the island’s modern inhabitants. Our data, while limited, suggest that the Eastern or North African influence in the Punic population of Ibiza was primarily male dominated.

Processing and sequencing of samples. All ancient DNA processing was conducted in a dedicated aDNA facility at the University of Otago. DNA was extracted from archaeological teeth using standard protocols 22 . Double stranded Illumina libraries were constructed with dual barcodes for the ancient samples as described previously 23 , and the mitochondrial DNA was captured using hybridization with modern human mitochondrial bait, generated in our lab 24 .
Modern DNA samples were sent by CSIC-UPF to the University of Otago where they were processed in a separate lab from the aDNA facility, and mitogenomes were amplified, prepared and sequenced as described previously 23 .
All ancient sequences obtained were assessed for potential contamination and DNA damage patterns. Deamination and fragment lengths were all as expected for ancient samples (Supplementary Data 2). Consensus sequences were created (including indels) and deposited in GenBank. Sequences were assigned to haplogroups Scientific 25 with Phylotree build 17 26 . All aDNA reads generated have been submitted to the NCBI sequence read archive identified by lab sample number. All mtDNA haplogroup determinations for samples processed in this study and associated archaeological site names, dates, GenBank accession numbers, coverage and contamination estimates are shown in Table 1. Additional sample information is provided in Supplementary  Table 1. All ancient samples processed showed minimal estimates of contamination except for sample MS10589, from Formentara, which returned an estimate of 14%. We therefore filtered the data to only include damaged reads which resulted in reduced coverage but the same haplogroup call, so we included this pre-Phoenician sample in our analyses.
Generation of nuclear data. One ancient sample (MS10614) was selected for shotgun sequencing, based on the high genome coverage observed in the mitochondrial sequencing of this sample. Adequate preservation was confirmed with an estimated 9% endogenous DNA content. The sex of this sample was determined based on computing the number of alignments to the Y Chromosome (nY) as a fraction of the total alignments to both sex chromosomes (nX + nY) 27 . The ratio of Y:X + Y coverage was 0.0042 (SE 0.0001, 95% CI 0.0041-0.0044) and thus the individual was determined to be female, which is inconsistent with the sex determination based on skeletal characteristics. After mapping the shotgun reads to the UCSC reference genome hg19, the resulting bam file was aligned with the Human Origins SNP dataset to call a pseudo-haploid genotype 28 . This genotype was merged with previously published ancient 29,30 and modern SNP data sets [31][32][33] , resulting in a total of 119,316 SNPs used in subsequent analyses (See Supplementary To test for continuity between ancient and modern Ibizan haplotypes, we performed a temporal network analysis 40 , shown in Fig. 3, which indicates that there were no shared haplotypes between the ancient and modern Ibizans. Following the lack of DNA evidence for continuity between the Phoenician and modern Ibizan populations, we compared the modern and Phoenician mitochondrial data to other likely candidate source populations. Genetic distance (F ST ) was calculated between modern Ibizans and multiple candidate source populations 34,[41][42][43][44][45][46][47] . Pairwise F ST between all populations were visualized in a MDS plot (Supplementary Data 3), to assess the genetic distance between Ibiza and candidate source populations in comparison to the genetic distance among source populations. The genetic distance between modern Ibizans and candidate source populations was then visualized on a map, to refine our estimates of populations maternally closely related to the modern Ibizans (Fig. 4).
To determine if there were any patterns observed between Phoenician populations in the Mediterranean, we combined our ancient Phoenician data with previously published mitogenome data for a range of Mediterranean populations including pre-Phoenician, Phoenician and modern Sardinians 3,34 and pre-Phoenician 29 , Phoenician and Modern Lebanese 3 as well as modern mainland Spanish 34,41,45 and performed multidimensional scaling of pairwise genetic distance (Fig. 5).  Fig. 4a). To provide more context within the European cluster we first added to our initial PCA analyses data from modern Iberians 33 (IBS) and Levantine populations from Jordan, Syria, Lebanon, and Israel 32 . The Ibizan sample plotted within the Levantine cluster, outside of the modern Iberians (Supplemental Fig. 4b). When considering more fine-grained analysis including only the Levantine and Iberian modern data with MS10614, the Ibizan positioned between the Iberian and Levantine samples, with more affinity to the Levantine populations (Fig. 6a). We conducted a further test, adding North African populations 48 to the analyses (Fig. 6b) and see that MS10614 is situated more closely to the Levantine and Iberian samples and not the North Africans. Finally, since modern populations are likely to have changed significantly in the 2000 + years since the Phoenician expansion, we undertook similar analyses to assess   the relationship between MS10614 and other ancient populations (Fig. 6c). The ancient sample MS10614 was also plotted with the modern Human Origins (HO) dataset for reference (Supplemental Fig. 4c). The ancestry of MS10614 was also estimated using model-based clustering in ADMIXTURE 49 (Supplementary Data 5). The most appropriate model was k = 2, determined by selecting the value of k with the lowest cross-validation error, though k = 3 had only a slightly higher cross-validation score, and both are shown in Fig. 7, which shows a subset of the data, consisting of the Iberian and Levantine populations. The model k = 3 shows a new component (dark blue) that impacts all of the Eastern Mediterranean populations and the modern Iberians. Sample MS10614 has an intermediate amount of this component, with more than modern Spanish, but less than modern Lebanese and ancient Levantines.
Outgroup f 3 -statistics were used to measure shared genetic drift between MS10614 and other ancient and modern populations. These results indicated that this Phoenician individual is most closely related to Neolithic aged samples from Iberia and Anatolia, plotting away from North African samples, providing further evidence for this individual having ancestry from both mainland Europe and the Near East (Supplementary Data 6). D-statistics were similarly supportive of gene flow between the Phoenician MS10614 and both Neolithic aged Iberian and Anatolian samples, however there was no statistical significance between the inferred gene flow from each of these regions with the Phoenician Ibiza individual.

Discussion
Most Phoenician trading ports were initially founded in locations where there was already an established indigenous community, resulting in integration of locals into the growing Phoenician settlement. The settlement history of Ibiza offers a valuable opportunity to investigate the genetic makeup of a Phoenician population where there was a relatively very small indigenous population at the time of Phoenician arrival. Being a likely secondary Phoenician settlement from the city of Gadir, the early population of Ibiza also potentially provides information about Phoenician settlements on the Iberian mainland. The data presented here adds to our previous research investigating admixture in Phoenician settlements across the Western Mediterranean 3,23 . It also helps dissect and possibly identify the origins of the various populations that ultimately contributed to the genetic makeup of the modern Ibizans. Our analyses of the mitogenomes of the ancient samples from Ibiza indicate a predominantly European maternal ancestry for the population (Fig. 1). Haplogroups H1 and H3, which either emerged from the Franco-Iberian refugia of the LGM 50,51 , or arrived with early Neolithic expansions, account for 50% of the Ibizan ancient samples (see Table 1). Szecseny-Nagy et al. 38 also found similarly high rates of haplogroup H in their analyses of the Hyper-Variable Region (HVR) of Neolithic and Early Bronze Age Iberian samples. Investigating coding region SNPs, they showed that among haplogroup H, 65% belonged to hg H1 and 14% to hg H3. Particularly high frequencies of H1 and H3 are also seen in ancient and modern Sardinians 3,34 . The majority of the remaining haplogroups present in the Phoenician population of Ibiza, U4a, U5b, and T2b have all been identified in Mesolithic or Neolithic populations in Western Europe including Iberia 38,52,53 . Haplogroup U5b3 frequencies,   34 found a J1c3 haplotype in one of their ancient samples from Sardinia (MA74), dated to 6190-6000 calBP and thus its presence in our Puig des Molins samples is not unexpected, as this lineage was clearly present in the Western Mediterranean region at the time that Ibiza was settled. The primary Phoenician settlers of Ibiza are likely to have be males who sailed from Gadir and the presence of these haplogroups provides additional support to our previous work suggesting that local female integration was a hallmark of Phoenician settlement history across the Mediterranean basin 3 . Our previous analyses of Phoenician mtDNA from Monte Sirai, Sardinia, another Western Mediterranean Phoenician and Punic settlement, indicated significant integration of indigenous women among the predominantly male Phoenician settlers 3 . We suggest a similar pattern is indicated for the Phoenician settlement of Ibiza. When we investigated the relationships of mitogenomes of Phoenician populations from Lebanon, Sardinia and Ibiza (Fig. 5) we found that there was no particular affinity amongst Phoenician populations (circles) when compared to pre-Phoenician or modern populations from those locations. This may indicate that local female integration was indeed a common strategy of Phoenician settlements. We expect that the integration of indigenous women from south central coastal Iberia into the Phoenician settlement of Gadir must have also occurred and that the ancient mitochondrial lineages observed Ibiza were brought along from Gadir or, alternatively, from some other Phoenician settlement in the Central or Western Mediterranean. Analyses of ancient Phoenician samples from Gadir and complete mitogenomic sequencing of ancient samples from across the Iberian Peninsula are needed to fully test this explanation. Analyses of HVR sequences from Neolithic and Early Bronze Age samples from the Iberian peninsula 38 and modern southern Iberian populations 56 indicate that there may be differences between the southwest and southeast Iberian populations, and it may be possible to locate particular regions of the Iberian coast from which these mtDNA lineages originated.
There was a sizable increase in the number of the Punic burials at Puig des Molins during the 5 th and 4 th centuries BCE, indicating a surge in the population size of Ibiza 10 . This increase is largely attributed by historians to Despite anthropological evidence supporting the presence of North and sub-Saharan African ancestry in Punic individuals from the Island of Ibiza 57 , mitochondrial lineages exclusively associated with North Africa or the Near East were not observed in our ancient Ibizan samples, although several of these lineages were previously identified in Phoenician samples from Sardinia, i.e. haplogroup N1b1a5 3 . Our results cannot, however, rule out the arrival of admixed individuals with European maternal ancestry from Phoenician sites in Sardinia, North Africa or the Iberian mainland. In fact, we previously identified a likely Iberian mtDNA haplotype, U5b2c1, in a young Carthaginian 23 , suggesting the presence of an admixed population there by late 6 th century BCE. We also note that haplogroups H1 and H3, while European in origin, are also found at high frequency in North African populations, particularly H1, which is today found at levels of 40-45% in northwest Africa 56,58 . Haplogroup H1bc was identified in a Bronze Age sample from Sidon 29 and H1 is present in modern Lebanese 3 . It is therefore possible, that these H lineages could have been brought to Ibiza, with Phoenicians from either the Levantine homeland or North Africa. It is a more parsimonious suggestion, however, that their presence in Ibiza is due to a more direct European and probably and Iberian source, particularly given the lack of any mitogenomes of clear North African or Near Eastern origin identified in the ancient Ibizan population sampled thus far. The morphological characteristics indicating African ancestry in crania from Punic contexts in Ibiza identified by Márquez-Grant 57 could be consistent with our aDNA results if the African admixture in Ibiza was male dominated. Alternatively, aDNA analyses of additional Phoenician samples from Ibiza may indeed find non-European mtDNA lineages, but to date, we have only observed what appears to be a typical European, and primarily Western European signature. Unfortunately, to really dissect population mtDNA histories generally, and those of southwest Europe or any region with high rates of haplogroup H in particular, complete mitogenome sequencing is necessary. Despite the large sample sizes for studies of HVR sequences for the Iberian Peninsula 38,59 , there are few complete ancient mitogenome sequences publicly available particularly beyond the Basque region. As Hernandez et al. 56 point out, secure assignment to haplogroup H is not possible based on control region sequencing alone, as it requires identification of T7028C. Further sub-typing of haplogroup H also requires the identification of key coding region SNPs (for example, G3010A for H1, T6776C for H3).
In addition to the ancient Ibiza samples, we sequenced a complete mitogenome of a Late Chalcolithic/ Early-Bronze Age sample (MS10589) from the Ca na Costa site on the island of Formentera which was determined to belong to the K1a1b haplogroup. Mathieson et al. 55 identified K1a1b1 in a Middle Neolithic Iberian sample (Mina3) with an archaeological context dated to 3900-3600 BCE. In their analyses of mtDNA hypervariable region sequences obtained from ancient samples from Neolithic and Iron-Age necropoli from Mallorca and Menorca, Simón et al. 39 found that 4.34% belonged to Haplogroup K, with the highest frequency (10%) in the oldest (Neolithic) necropolis, Cova des Pas, on Minorca. Our K1a1b result may indicate that the initial settlement of Formentera, prior to the arrival of the Phoenicians, also likely came from the mainland or via Mallorca, an observation presented by Bellard 4 . Olivieri et al. 34 found K1a and K1b1b1 in pre-Phoenician Sardinians, but we have yet to find haplogroup K in any of our Phoenician samples.
Ibiza has had a complex history of resettlement periods since the time of the Phoenicians. In addition to a Phoenician founding population, modern Ibizans would have had potential genetic influences from later Punic settlements, like Carthage and our ancient DNA data are consistent with this scenario. Historic records would suggest that the Ibizan genetic signature was also likely influenced by the arrival of Arab and Berber populations during the first millennium CE and finally, from 1200 CE, it was said to be overwhelmed by mainland colonists. This might explain the discontinuity between the ancient and modern DNA results we observed, shown in Fig. 3 and is consistent with the results of Ramon and colleagues 13 , based on HLA markers and blood groups in Ibizans, and earlier mtDNA 14 and Y-chromosome 15 studies which show affinities to North African and Near Eastern populations.
Our TempNet results showed little continuity between the Phoenician inhabitants of the island and the modern Ibizan population. No shared haplotypes were found and only one haplogroup, T2b, was found in both groups. Perhaps most surprising was that neither haplogroups H1 or H3 were found in the modern Ibizans. These two haplogroups were not only the most common in our ancient population but are found at high frequencies in both modern North African and Western European populations generally 56 . The modern Ibizan population is singularly unusual in the entire Western Mediterranean region for their apparent lack of these two common haplogroups. It is possible that the mtDNA frequencies are so different from the ancient samples due to random genetic drift, which is common in small island populations. The small sample size (n = 18) could also be a factor, though our result indicating that the modern Ibizans are genetically quite different to mainland populations is consistent with previous studies 13,14,16,39 .
Haplogroups T1 and V, found in the modern Ibizans, are lineages found at reasonably high frequencies in the Near East. Haplogroup V is also found in both ancient and modern populations from the Iberian Peninsula and in ancient samples from Mallorca and Minorca, and T1 is also found at low frequency in mainland Spain today 14,32,55 . These lineages are also found in Berbers which could explain their presence in the modern Ibizans. Haplogroup L2c however, which was identified in two unrelated, modern Ibizans, has not been found in Berbers 60 . It is typically a West African, sub-Saharan lineage and it may have been introduced during the Islamic expansion which had significant exchange with the entire African continent. Alternatively, it could have been a more recent arrival, perhaps the result of the transatlantic slave trade 59 . Botigué et al. 61 , showed in their whole genome analyses that sub-Saharan ancestry was less than 1% in Europe, except for the Canary Islands, where Maca-Meyer et al. 62 found L2c in both 17 th and 18 th century historic and modern individuals. Another sub-Saharan lineage, L1b has also been identified in a Late Chalcolithic population in central Iberia 38 and in one 7 th century CE sample from Mallorca 39 .
The remaining lineages identified in modern Ibizans are all either exclusively European markers (e.g. J2b1a 63 ) or are widely distributed across Europe and the Near East and identifying the specific origin of the population influx that causes the discontinuity between ancient and modern Ibizans is difficult. The F ST results for the Scientific Fig. 4 do not indicate a single likely source, though the closest population to modern Ibizans appears to be France. France may be acting as a proxy for a more likely source of settlers to Ibiza, namely Catalonia, since it has been shown to be genetically closely related 64 . Further whole genome analyses of the modern indigenous Ibizans would provide the necessary data to fully assess the origins of this population, but from a mitochondrial perspective, it appears that they are not directly related to the Phoenician and Punic founding populations and thus this is not the explanation for the unusual genetic signature. While the mitochondrial evidence suggests that the founding female population of Ibiza was primarily derived from a mainland Iberian source, the results of the whole genome analyses of sample MS10614 indicate a significant Eastern Mediterranean/Levantine component. Sample MS10614 comes from an individual that was part of a collective burial inside a Punic hypogeum at Can Portes des Jurat, Cas Molí 20 . The sample was directly radiocarbon dated to 361-178 cal BCE. The archaeological context of the burial also suggests a 3 rd century BCE date and thus this individual lived during the period of greater Punic influence of Ibiza or just after this period of population growth.
As can be seen in Fig. 6a,b, the Ibizan Phoenician sample, MS10614, plots in between modern Levantine and Iberian populations but closer to both of these than to modern North African populations. In Fig. 6c, it is positioned most closely to a Levantine Neolithic sample, but in between a Sidon Bronze Age and European/Iberian Bronze Age samples. The ADMIXTURE result (k = 2) in Fig. 7  The mtDNA haplogroup of this individual, T2b, indicates that the direct maternal ancestry is likely local and the Levantine contribution may therefore represent paternal ancestry. This finding is in line with previous work by our group using Y-chromosome data showing East Mediterranean Y signatures across the Phoenician settlements in the West, including Ibiza. These signatures, attributed to the Phoenicians, were found to be present at the frequency of 6% among the modern male populations studied, indicating a substantial Phoenician male presence in settlements across the Western Mediterranean 65 . The Near Eastern component to the genome of this individual may represent that of the founding Phoenician population, though the possibility that this individual or a recent ancestor may have arrived directly from the Levantine homeland or from Punic settlements in North Africa or elsewhere in the Phoenician interaction network cannot be ruled out, though the sample does not appear to be closely related to modern North African populations as shown in Fig. 6b.
In conclusion, our results show a complex pattern of settlement of the island of Ibiza. We demonstrate a clear genetic discontinuity between the early Phoenician settlers and the modern inhabitants of the island based on the mtDNA results. Thus, the unusual genetic signature of modern Ibizans is not likely to be the result of their Phoenician ancestry, at least from a mitochondrial perspective. It appears that multiple population arrivals through invasions or other movements combined with periods of population instability since the early Phoenician settlement may have led to a total reshuffling of the genetic makeup of this island. Over the last several centuries Ibiza witnessed population growth supported by the arrival of mainland populations to the island, followed by significant population reduction resulting from the bubonic plague and malaria 13 . These events, combined with a founder effect, and inbreeding common in islands with a relatively small population such as Ibiza, could have resulted in the loss of the indigenous mtDNA signatures observed in the ancient samples we analysed. It appears that the lineages present in the Phoenicians of Ibiza were replaced by different European mtDNA haplogroups (which appear to be most closely related to those of modern French but likely also similar to Catalonians). Despite this lack of continuity observed in the mitochondrial genomes, previous Y chromosome analyses suggest that there is still some Phoenician signature in the modern Ibizan population. This is consistent with historical evidence suggesting that Phoenician influence in the West was male dominated and indicates that there was not a total replacement of the Ibizan founding population. Comparisons of the ancient Phoenician whole genome data with whole genome data from modern Ibizans will help clarify this point further. Finally, the whole genome data obtained from the ancient Ibizan sample belonged to an individual with a significant Eastern Mediterranean component, suggesting an admixed Phoenician community in Ibiza during the 3rd century BCE. While this result is consistent with the archaeological evidence and further indicates that diversity was a hallmark of Phoenician societies, it also highlights the complexity of island population settlements and underscores the importance of the inclusion of ancient DNA analysis in population genetics.

Methods
Archaeological sites and samples. Permits and approval for this study were obtained from the office of the Director General of Antiquities in Lebanon (permit number 4290, 6 November 2015) for all Lebanese samples and from the Museu Arqueològic d'Eivissa i Formentera for the Spanish samples and all methods were performed in accordance with the relevant guidelines and regulations.
Puig des Molins (windmill Mountain) is one of the largest necropoli discovered to date in west Phoenicia. Used since the 7 th century BCE, it spans an area around 50,000 m 2 , with close to 3000 tombs. At least three burial rituals can be distinguished in Puig Des Molins. The first between the 7 th and 6 th centuries BCE used cremation. The second coincided with significant population growth on the island during the 5 th and 4 th centuries BCE and corresponded to inhumation using the spectacular hypogea cut in the rock. The third, between the 3 rd and 2 nd Ancient DNA extraction and library preparation. All aDNA extraction and library preparation (until PCR) was carried out in a purpose built aDNA facility at the University of Otago 69 . DNA was extracted from 13 teeth recovered from two archaeological sites on Ibiza and one site on Formentera. All teeth were rinsed in 5% bleach, followed by washing with distilled water for multiple washes to remove any bleach residues and drying overnight. DNA was extracted from ~150 mg of ground tooth, following a silica-based extraction protocol 70 . One extraction negative was prepared alongside every five extracts and processed alongside extracts for the remainder of the lab work.
Mitochondrial DNA enrichment and sequencing. For the mitochondrial genome sequencing, libraries were prepared and hybridisation capture used to target the human mitochondrial genome according to the protocol described by Matisoo-Smith et al. 23 . Mitochondrial genome libraries were pooled and sequenced using a 2 × 75 bp paired-end sequencing kit on an Illumina MiSeq platform at the Otago Genomics and Bioinformatics Facility (OGBF).
Mitochondrial DNA sequence processing and sequence alignment. Resulting sequencing data from the ancient mitochondrial libraries were processed according to methods described previously 3 . Complete mitogenome sequences were subsequently aligned with one another and previously published mitochondrial sequences using mafft 71 .
Authentication and contamination assessment. Throughout the library preparation process, 'negative' libraries were handled alongside sequencing libraries and visually checked by gel electrophoresis and on the BioAnalyser to ensure that these did not contain DNA. MapDamage 72 was used during processing of sequencing data to check nucleotide mis-incorporation patterns and fragment length distribution of sequencing libraries, as an indication that the resulting data looked degraded (Supplementary Data 2). During this step, the -rescale option in map-Damage was applied to rescale the quality scores of likely damaged bases towards the ends of reads. Mitochondrial DNA contamination was estimated using contamMix on the basis of a majority rule mtDNA consensus sequence and an alignment of 311 worldwide mtDNA sequences 73 (contamMix data shown in Table 1).
Due to the relatively high contamination rate (14%) of sample MS10589, we used PMDtools with a damage threshold of 3 to select only those reads that were damaged 74 , resulting in 65.6% coverage (average coverage of 1.3X). When the VCF was loaded into HaploGrep we obtained the haplogroup call of K1a1b (which is consistent with the original call of K1a1b1) and confirmed 7 informative SNPs, ranging from between 2 and 4X coverage, thus we included this sample in further mitogenome analyses.
MtDNA sequence analysis. Network Analyses -mitogenome and HVR: In order to identify relationships between the modern and Phoenician Ibizans and potential source populations, aligned mitochondrial sequences (complete mitogenome and partial HVR) were used to construct median joining networks in PopArt 75 . A full list of populations included in each network is available in the Supplementary Tables 3 and 4.
Temp net: The modern and ancient mitogenome sequences generated in this study were aligned and used to create a TempNet temporal haplotype network 40 to assess whether the modern Ibizan population were directly descended from the Phoenician Ibizan population.
MDS: The genetic distance between modern Ibizans and candidate source populations, based on complete mitogenomes, were compared using a multi-dimensional scaling analysis. A MDS plot was also generated to compare mitogenome sequences from modern and Phoenician Ibizan individuals to modern, Phoenician and pre-Phoenician (depending on data availability) populations from the Spain, Sardinia, Lebanon and Egypt using a multi-dimensional scaling analysis. The genetic distances between populations were calculated using DnaSP v6 76 and F ST values were visualised in a two-dimensional MDS plot, using the pcoa function in the 'ape' package for R 77 . Shotgun sequencing of whole genome. One individual, MS10614, was selected for shotgun sequencing based on the high coverage of the mitochondrial genome sequencing results. The amplified sequencing library prepared for hybridisation capture was submitted for shotgun sequencing using a 2 × 75 bp paired-end sequencing kit on one lane of a HiSeq at the Otago Genomics and Bioinformatics Facility. Sequencing reads were mapped to a human reference sequence (UCSC hg19), and genotypes were called by sampling a random read per SNP in the Human Origins SNP panel, using the 'pileupCaller' tool (https://github.com/stschiff/sequenceTools/tree/ master/src-pileupCaller). The resulting genotypes were merged with ancient DNA datasets from Lazaridis et al. 30 and Haber et al. 29 and modern DNA datasets of Iberians (IBS) from the 1000 Genomes Project Consortium 33 , Tunisians from Lazaridis et al. 32 and Lebanese from Haber et al. 31 .
Analysis of whole genome data. Sex determination: To determine the genetic sex of the sample MS10614, sequenced using shotgun sequencing, we compared the number of Y chromosome alignments (3516) to the total number of X and Y chromosome alignments (827,587) as described by Skoglund and colleagues 27 . The sample was determined to be 'male' if the ratio of Y and autosomal coverage was greater than 0.25. Our result of 0.0042 would indicate the sex was female, though we suggest the low coverage overall could contribute to this result which contradicts the morphological characteristics that were recorded by the archaeologist at the time of excavation (Supplemental Data 1). PCA: When we projected MS10614 into the context of the three worldwide modern populations (Supplemental Data 4 and Fig. 6a,b), we ran the analysis twice: one taking into account the transversions only, and one including even the transitions. This approach allowed us to see if the same pattern was reproduced or not and, since in both cases the behaviour observed resulted to be the same, we decided to stand with the more inclusive file (57,221 variants).
In all the considered scenarios involving modern populations and MS10614, the Principal Component Analyses were produced calculating the eigenvectors with smartpca; the ancient sample was projected setting the lsqproject parameter to yes in the smartpca parameter file. In every case we only took into account those SNPs that were common to the merged datasets and the ancient sample. No minor allele frequency filter, or LD pruning were performed. In modern samples, only variants and samples with many missing calls were filtered out (>5% and >10%, respectively). Also, variants failing a Hardy-Weinberg equilibrium exact test were discarded (p < 10 −5 ).
For the ancient samples, PCA was performed using the merged dataset containing the Phoenician Ibizan (MS10614), ancient individuals from Lazaridis et al. 30 , and Bronze Age samples from Sidon 29 , using the 'smartpca' software from the Eigensoft package 78 .
Admixture: The ADMIXTURE v 1.3 software 49 was used on a subset of the above merged dataset, comprised of the Phoenician Ibizan, Neolithic Anatolians, Late Neolithic-Bronze Age Europeans, Bronze Age Iberians, Bronze Age Levantines, Neolithic Levantines, modern Lebanese, modern Spanish, Bronze Age Sidon samples, and modern Tunisian populations. The ancestry proportions of these populations were assessed by testing clusters of K = 1-10 (shown in Supplemental Data 5). F 3 and D statistics: Outgroup f 3 statistics were computed using the 'qp3pop' tool from the Admixtools package 79 in the form of f 3 (Mbuti; X; MS10614) where 'X' tests all populations in the merged dataset. Based on the results of the outgroup f 3 statistics, the 'qpDstat' tool from the Admixtools package 50 in the form of D(MS10614; X; Y; Mbuti), where 'X' and 'Y' tests populations identified as being ancestral to MS10614 in the f 3 and PCA analyses (Supplemental Data 6 and Table 5).

Data Availability
Modern and ancient mitogenomes presented in this paper are available under the accession numbers MH43585-43559. We are currently in the process of submitting the mapped BAM files for the mitochondrial samples and the shotgun sequence data to the NCBI short read archive.