A substantial prehistoric European ancestry amongst Ashkenazi maternal lineages

1 Supplementary Figures Supplementary Figure S1. Frequency distribution maps for mtDNA haplogroups K, U8a1a and U8b1. Maps created using Surfer and based on the data presented in Supplementary Table S2. Dots represent sampling locations used for the spatial analysis. 2 Supplementary Figure S2. Distribution map of the diversity measure ρ for haplogroup K. Map created using Surfer. Dots represent sampling locations used for the spatial analysis. 3 Supplementary Figure S3. Bayesian skyline plots based on haplogroup U8 mitogenome data in the Near East/Caucasus and Europe. The hypothetical effective population size is indicated (on a logarithmic scale) through time (ka). The sample sizes were 106 and 474 respectively. The posterior effective population size through time is represented by the black line. The blue region represents the 95% confidence interval. 4 Supplementary Figure S4. Reduced-median network of haplogroup K1a9, in the context of putative clade K1a9’10’15’26’30. The network is rooted with several additional lineages from K1a. The tree in Supplementary Data 1 is resolved by assuming transitions first at position 16093 and then at position 195, with K1a20, K1a28 and K1a30 branching off earlier. The tree could also be resolved by assuming mutation at position 195 first, then at 16093, in which case K1a13, K1a16 and K1a31 would branch off earlier, rather than forming a distinct terminal subclade, K1a13’16’31. The latter resolution would considerably extend the European ancestry of the K1a9 lineage. 5 Supplementary Figure S5. Phylogenetic tree of Ashkenazi founders within haplogroup H5. Time scale (ka) based on ML estimations for mitogenome sequences. 6 Supplementary Figure S6. Phylogenetic tree of Ashkenazi founders within haplogroup H7. Time scale (ka) based on ML estimations for mitogenome sequences. 7 Supplementary Figure S7. Phylogenetic tree of Ashkenazi founders within haplogroup H11. A Bronze Age European lineage is shown with a red circle29. Time scale (ka) based on ML estimations for mitogenome sequences. 8 Supplementary Figure S8. Phylogenetic tree of Ashkenazi founder within haplogroup H13a1a1. Time scale (ka) based on ML estimations for mitogenome sequences. 9 Supplementary Figure S9. Phylogenetic tree of Ashkenazi founder within haplogroup H15. Time scale (ka) based on ML estimations for mitogenome sequences. 16 Supplementary Table S4. Age estimates using rho (ρ) and ML for haplogroup HV1b and its subclades. Ages and 95% confidence intervals (CI) in thousands of years. Clade ρ (complete sequences) ρ (synonymous) ML Age CI Age CI Age CI HV1b 17,200 [10,200–24,400] 14,200 [5000–23,300] 18,200 [10,200–26,600] HV1b+152 12,500 [5600–19,700] 12,300 [1000–23,600] 14,900 [8600–21,400] HV1b2 2100 [700–3500] 2900 [100–5700] 2200 [800–3600] HV1b3 19,000 [8600–30,000] 18,400 [3000–33,800] 12,300 [6500–18,200] HV1b1 10,300 [4800–15,900] 9900 [1700–18,000] 8700 [4500–13,100] HV1b1b 6600 [1100–12,200] 5200 [0–12,500] 5900 [1800–10,100] HV1b1a 8800 [1900–16,000] 10,500 [0–23,100] 5900 [2100 –9900] 17 Supplementary Table S5. Age estimates using rho (ρ) and ML for haplogroup N1b and its subclades. Ages and 95% confidence intervals (CI) in thousands of years. Clade ρ (complete sequences) ρ (synonymous) ML Age CI Age CI Age CI N1b 21,200 [13,600–29,100] 25,000 [9000–40,900] 19,100 [13,200–25,100] N1b1 13,900 [10,800–17,000] 12,800 [7300–18,200] 13,500 [6600–20,600] N1b1b 11,600 [7600–15,800] 16,700 [8800–24,700] 11,600 [8100–15,200] N1b1c 8700 [3900–13,400] 4000 [0–9400] 9500 [4700–14,400] N1b1+16129 15,000 [8700–21,600] 15,700 [5100–26,400] 12,600 [9200–16,000] N1b1d 13,400 [5200–21,900] 13,100 [0–26,700] 8700 [4700–12,700] N1b2 4900 [100–9800] 9300 [0–23,200] 5000 [1800–8300] N1b2+14581 2400 [700–4100] 2800 [0–6500] 2100 [700–3500] 18 Supplementary Table S6. Summary of distributions of the main haplogroups in the Ashkenazi mtDNA control-region data1,2,11,45. “HV” includes HV1, a single HV2 lineage, two HV13 lineages and five putative HV5 lineages; “U” includes U1, U2, U3, U4, U6 and a single U8b1 lineage; “others” include single N1c, N9a and two putative K1a4 lineages. Note that the British data are excluded from the Western European totals because British Ashkenazi Jews have a predominantly Eastern European ancestry, as is evident in the haplogroup frequencies. Region H HV HV0 I J K1a1b1a K1a9 K2a L2a1l M1a1b N1b2 R0a1a T U U5 U7 W X M/A Others Eastern Europe 24.3 5.8 4.3 0.9 7.5 18.7 6.0 6.0 1.5 0.3 8.8 1.8 4.3 2.3 2.3 0.9 2.0 0.6 0.9 0.6 Czech/Austria/ Hungary 30.4 1.8 6.9 1.4 3.2 21.7 5.1 7.4 0.9 — 9.7 — 4.6 1.4 1.8 — 2.3 0.5 0.5 0.5 Baltic 30.8 7.7 5.1 — — 15.4 7.7 5.1 — 2.6 15.4 7.7 — 2.6 — — — — — — Poland 13.5 7.6 2.9 1.8 10.6 25.9 4.7 7.6 2.9 0.6 5.9 1.2 3.5 4.1 1.2 1.8 2.4 0.6 1.2 — Russia/Belarus 19.2 12.3 — — 16.4 5.5 4.1 5.5 1.4 — 8.2 5.5 8.2 1.4 5.5 — 4.1 — 1.4 1.4 Romania/Moldavia 24.5 6.1 1.0 — 10.2 16.3 9.2 4.1 2.0 — 10.2 1.0 3.1 1.0 3.1 1.0 1.0 2.0 2.0 2.0 Ukraine 35.2 5.6 9.3 — 3.7 9.3 9.3 — — — 7.4 3.7 5.6 3.7 3.7 3.7 — — — — Western Europe 15.2 — 1.9 — 4.8 34.3 8.6 1.9 1.9 3.8 14.3 2.9 4.8 1.0 — 1.9 — 1.9 1.0 — France 14.3 — — — 4.1 30.6 6.1 4.1 2.0 6.1 16.3 2.0 6.1 2.0 — 2.0 — 2.0 2.0 — Netherlands/ Germany/ Switzerland 16.1 — 3.6 — 5.4 37.5 10.7 — 1.8 1.8 12.5 3.6 3.6 — — 1.8 — 1.8 — — Great Britain 21.3 5.0 5.0 6.3 3.8 13.8 6.3 3.8 — — 6.3 2.5 8.8 11.3 2.5 1.3 — 1.3 1.3 — Total 22.8 5.0 4.1 1.3 6.8 20.2 6.3 5.3 1.4 0.7 9.2 2.0 4.8 3.0 2.0 1.1 1.6 0.8 1.0 0.5 19 Supplementary Table S7. List of 29 Ashkenazi mitogenome haplogroup H sequences used in the analysis. Three Sephardic and one B'nei Israel haplogroup H sequences are also shown. Accession numbers Control-region variants motif Haplogroup Notes (e.g. presence in prehistoric central European remains29) JQ703788, JQ703655 16172C-16192T-16456ª-207A H1aj1 H1 found in Early Neolithic remains JQ704370 CRS H1as H1 found in Early Neolithic remains JQ703268, JQ704894 16080G-16189C-16356C-16360T183G H1b2a1 Subclade found mainly in central/Eastern Europe EU148452, EU262984, JQ703137, JQ705236 16114T; one has 16126C H1e4a H1e found in Early Neolithic remains JQ701904 152C-16319A H2a2a2 H2 found in Bronze Age remains JQ704133, JQ703179 16222T-16299G-146C; one has 200G H3p H3 found in mid-Neolithic remains JQ705703 195C-513A H4a1a3a H4 found in Late Neolithic Corded Ware remains JQ704150 16304C-456T H5a7 H5a found in Late Neolithic Bell Beaker remains HQ663853 16167T-16274A-16304C-73G456T H5-73G Sample designated only as “Jewish” H5 found in Early Neolithic remains JF790201 16207G-16213A-16216G-456T H5c2 H5 found in Early Neolithic remains JQ702509 16311C-16362C-16482G-239C H6a1a1a H6a1a found in Late Neolithic Corded Ware remains; Near Eastern subclade H6a1a1 HQ730608 16362C-16482G-239C H6a1a3 H6a1a in Corded Ware remains GU390313, HM015668 16362C-16482G-239C H6a1a5 H6a1a in Corded Ware remains JQ703313 152C-573.1CC H7-152C H7 found in mid-Neolithic and Bronze Age remains JQ703601 CRS H7c2 H7 found in mid-Neolithic and Bronze Age remains HQ267514 CRS H7e H7 found in mid-Neolithic and Early Bronze Age remains JQ704427 16092C-16140C-16293G-16311C152C-195C H11a2a2 H11a found in Early Bronze Age remains JQ704081 16261T-16311C-152C-195C H11b1 H11 found in Early Bronze Age remains JQ701818 150T-310C H13a1a1 H13a1a2 found in Late Neolithic Bell Beaker remains JQ704870 CRS-55C-57C-16527T H15b H15b may be Near Eastern JQ704110 146C-16129A H26c H26 found in Early Neolithic remains JQ702933 CRS H40b Source unresolved EF556184 CRS H1p Turkish (Sephardic ) Jew EF556189 CRS H25 Bulgarian Jew – Sephardic (unresolved – direct from root) JQ703478 16093C H56 Sephardic Jew from Spain (European clade) EF556150 183G-709A H13a2a B'nei Israel-Jew (Indian/Near Eastern clade) 20 Supplementary Table S8. Age estimates using rho (ρ) and ML for haplogroup H and its subclades. Ages and 95% confidence intervals (CI) in thousands of years. Subclades carried by Ashkenazim are shown in pale blue. Clade ρ ML Whole mtDNA Synonymous Whole mtDNA Age CI Age CI Age CI H1 11,000 9100–12,800 10,100 8000–12,300 11,800 10,300–13,300 H1as 7100 2400–12,000 4500 0–11,500 7600 2600–12,800 H1as1 7000 1000–13,200 0 0 4800 700–9100 H1as2 1700 0–4100 2600 0–7800 2300 0–6600 H1aj 16,200 5300–27,600 1600 0–4700 10,200 5600–14,800 H1aj1 2600 2500–24,900 0 0 2200 0–5200 H1b2 7100 1500–12,900 0 0 5300 2200–8500 H1b2a 1300 0–3800 0 0 1100 0–2300 H1e 11,400 7000–16,000 18,200 6000–30,300 10,900 8400–13,300 H1e4 3200 0–7000 0 0 3400 0–7600 H1e4a 1500 0–3300 0 0 1600 0–3400 H2a2a 7900 3300–12,600 6500 3600–9400 10,800 6900–14,800 H2a2a1 4500 3200–5900 4600 2400–6800 5000 3300–6700 H2a2a1d 2600 0–6400 0 0 2300 0–5200 H2a2a1b 2000 0–4200 2000 0–5900 2200 0–4800 H2a2a1a 400 500–8200 2600 0–7800 3300 800–5800 H2a2a2 5200 400–10,200 4000 0–8400 4900 0–17,300 H3p 9700 1700–18,000 5200 0–15,500 10,600 1800–19,800 H3p+146 1300 0–3800 0 0 1100 0–500 H4a1a-195C 10,000 3400–16,700 14,900 0–30,600 8000 4400–11,600 H4a1a3 5700 400–11,300 3100 0–9300 5400 0–11,100 H4a1a3a 3200 0–7100 3900 0–11,700 2800 0–10,500 H4a1a3a+513 0 0 0 0 0 0 H4a1a4 7900 1400–14,700 10,700 0–24,400 6300 3000–9700 H4a1a4a 2600 0–5600 0 0 3000 0–6700 H4a1a4b 3000 1600–4400 3300 700–5900 3100 1500–4700 H5 12,000 8000–16,100 12,800 6300–19,300 12,500 5500–19,800 H5-73G 7000 1000–13,200 5200 0–12,500 8900 1400–16,800 H5-73G-16167-16274 1300 0–3800 0 0 1600 0–4700 H5a7 5200 1200–9300 5200 0–11,500 4600 0–9700 H5c 7600 3300–12,000 96000 1100–18,200 8200 3100–13,200 H5c1 5200 400–10,200 7900 0–19,400 5600 600–10,700 H5c2 500 0–11,800 0 0 2000 0–5600 H5c2-16304C! 1300 0–3800 0 0 800 0–3300 H6a1a 6900 5300–8600 7100 4100–10,100 700 5100–8800 H6a1a1 8600 2600–14,800 15,700 1300–30,200 5700 2300–9200 21 H6a1a1a 1300 0–3800 3900 0–11,700 800 0–2500 H6a1a3 2600 300–4900 1600 0–4700 3100 0–6300 H6a1a5 2600 0–5300 1600 0–4700 3100 0–6300 H7 9700 7300–12,100 11,700 6800–16,600 12,900 7200–18,800 H7+152 4700 1600–7800 6300 100–12,500 6000 2000–10,200 H7c 8100 4600–11,600 7300 2500–12,200 10,400 6000–15,000 H7c1 4800 700–9000 4000 0–8400 5300 0–11,000 H7c2 900 0–2600 0 0 1200 0–10,300 H7e 3100 600–5600 4700 0–10,100 3800 700–7000 H11 13,700 6300–21,300 9400 3900–14,800 11,600 5900–17,400 H11a 8500 5400–11,700 8400 2800–14,100 6900 4900–8900 H11a1 3700 300–7200 2200 0–5400 3500 300–6800 H11a2 11,400 4500–18,600 6600 2400–10,700 6300 4200–8500 H11a2a 6700 3100–10,300 6700 1500–11,800 5000 2700–700 H11a2a1 3100 0–6700 3100 0–7500 2700 0–5800 H11a2a2 4300 500–8200 300 0–7800 2600 100–5000 H11b1 6100 2300–10,000 3900 0–8400 6500 1100–12,000 H11b1-16261T 1300 0–3800 0 0 1400 0–4100 H13a1a1 7200 4700–9700 4200 1700–6800 7200 4800–9600 H13a1a1+150 3500 100–6900 5200 0–12,500 4700 40–9100 H13a1a1d 4200 0–8800 6300 0–18,600 2800 100–5600 H13a1a1a 6800 2400–11,300 3900 500–7400 6200 3400–9000 H13a1a1e 3500 100–6900 2600 0–7800 4000 300–7900 H13a1a1b 3900 0–8400 0 0 4300 0–8800 H15 13,400 7200–19,800 22,200 5600–38,800 14,400 9400–19,500 H15a 10,400 4500–16,600 15,300 0–30,800 11,400 7100–15,900 H15a1 7900 4500–11,400 7900 2000–13,800 9000 6700–11,300 H15a1a 5200 400–10,200 0 0 5600 900–10,500 H15b 11,100 6300–16,000 11,800 3300–20,00 12,000 7200–16,800 H26 10,600 5300–16,000 19,700 5800–33,500 10,500 5500–15,600 H26a1 6400 2900–10,100 7900 2400–13,300 6000 2000–10,000 H26b 12,000 4100–20,300 23,600 4700–42,500 7100 2800–11,100 H26c 3200 0–7100 2000 0–5800 3900 0–16,800 H26c+16129 0 0 0 0 0 0 H40 5600 2400–8900 5600 0–11,500 7500 400–14,900 H40b 1300 0–3800 0 0 1400 0–12,800 22 Supplementary Table S9. The source of Ashkenazi mtDNA lineages based on 836 control-region sequences1,2,11,45. Each sequence was cross-referenced with the mitogenome phylogeny36, and either assigned to a probable European, Near Eastern or Asian source or left unassigned due to ambiguity in the phylogeographic distribution. Haplogroup H assignments based on whole mitogenomes (see Supplementary text). “HV” includes HV1, a single HV2 lineage, two HV13 lineages and five putative HV5 lineages; “U” includes U1, U2, U3, U4, U6 and a single U8b1 lineage; “others” include single N1c, N9a and two putative K1a4 lineages. Source H HV HV0 I J K1a1b1a K1a9 K2a L2a1l M1a1b N1b2 R0a1a T U U5 U7 W X M/A Others Input Europe 20.5 — 4.1 1.3 6.3 20.2 6.3 5.3 — 0.7 9.2 — 3.0 0.2 2.0 — 1.6 — — — 80.7 Near East 0.8 3.6 — — 0.1 — — — — — — 2.0 0.2 — — 1.1 — — — 0.4 8.3 Asia — — — — — — — — — — — — — — — — — — 1.0 0.1 1.1 Unassigned 1.6 1.4 — — 0.4 — — — 1.4 — — — 1.6 2.8 — — — 0.8 — — 9.9 Total 22.9 5.0 4.1 1.3 6.8 20.2 6.3 5.3 1.4 0.7 9.2 2.0 4.8 3.0 2.0 1.1 1.6 0.8 1.0 0.5 100.0 23 Supplementary Note 1 Phylogeography of haplogroup U8: context of the Ashkenazi haplogroup K founders Here we present the context for the Ashkenazi haplogroup K lineages, providing the detailed phylogeographic evidence supporting a deep ancestry for the Ashkenazi founder clades within Europe. Haplogroup K is by far the largest subclade within haplogroup U8. The phylogenetic tree (Supplementary Data 1, with its main features summarised schematically in Fig. 1) shows the complete U8 phylogeny, comprising 956 whole mitogenomes, including 909 from haplogroup K (Supplementary Data 3). The tips are colour-coded for geographical information and age estimations are provided at the main nodes (details in Supplementary Data 2). There are four estimates for each node: based on ρ with mtDNA synonymous sites only, ρ and ML (using PAML) with whole mitogenomes, with a correction for purifying selection38, and Bayesian estimates using BEAST46, based on several internal calibrations within haplogroup U40. The estimates are broadly similar (allowing for the greater imprecision of the synonymous estimates), although those from BEAST tend to be rather lower. It is worth noting that the Bayesian estimates are completely independent from those using ρ and ML, using a wholly different approach to both calibration and estimation. However, they were necessarily carried out with only a subset of the data (398 sequences) due to the limitations of the software, and given the necessary assumptions of the Bayesian approach (and, conversely the lack of any mutation model for the ρ estimates) we regard the ML values as the most reliable. The coalescence time of U8 is ~52 ka. Its age and distribution (Supplementary Fig. S1; Supplementary Table S1) suggest that U8 is one of the few extant mtDNAs dating to the time of arrival of anatomically modern humans in the Near East, ~50 24 ka47,48, and may have been one of the earliest mtDNAs to arrive in Europe – a prediction strikingly supported by the isolation of a U8 lineage, most likely pre-U8b (as it includes one of the three variants diagnostic of U8b), alongside probable pre-U5 lineages (with two HVS-I variants characteristic of U5), from human remains at Dolní Věstonice in the Czech Republic dating to ~31.5 ka49. The deepest split separates U8a, which dates to ~22 ka, from U8b, dating to ~47 ka. U8a is very rare, and appears to be exclusively European. U8b1, dating to ~44 ka and identifiable in the HVS-I database, is scattered at very low frequencies across west Eurasia, with a focus around the eastern Mediterranean and Southwest Asia. Its sister clade within U8b, the much more frequent haplogroup K, is ubiquitous throughout western Eurasia. The HVS-I data show that haplogroup K as a whole has frequency peaks in north-west Europe and between south-east Europe, Anatolia and the Levant, dipping in central and north-central Europe (Supplementary Fig. S1; Supplementary Table S1). The diversity distribution is fairly similar; there is a minor peak in south-west Europe with the major peak in south-east Europe/south-west Asia (Supplementary Fig. S2; Supplementary Table S2). However, the main subclades of haplogroup K are not visible in HVS–I data, so further investigation of their distribution using the HVS–I database alone is not possible. Using the whole-mitogenome evidence, we can see that haplogroup K dates to ~36 ka and splits into two primary subclades, K1 and K2 – although a single sequence from the South Caucasus appears to fall into a third basal branch50. This might hint at a Near Eastern origin for haplogroup K as a whole51. A Near Eastern origin for haplogroup K might also be suggested by both the Southwest Asian focus of its sister clade U8b1 and the HVS–I diversity pattern (Supplementary Fig. S2). Given the timing of the appearance of haplogroup K, just prior to the global climatic downturn, an origin 25 in the Near East – which acted as a major reservoir for mtDNA variation during the glacial period30,52 – might also more plausibly account for its survival than an origin in Europe. None of these arguments is compelling, however, and on the other hand, the majority of the major subclades within haplogroup K appear to have arisen within Europe (see below), with K1a the only likely exception, which points to a European origin for haplogroup K (Fig. 1). Furthermore, although the present distribution of U8b1 is widespread and uninformative, the probable pre-U8b lineage identified at Dolní Věstonice shows that U8b (of which haplogroup K is a subclade) has a presence in Europe dating back more than 30 ka49. Overall, we regard the place of birth of haplogroup K as unresolved, although Europe seems more likely than the Near East, given the rather weak evidence presently available. On the same grounds, the same is true for K1. Haplogroups K1 and K2 both date to ~29 ka, the cooling period from ~33ka prior to the LGM (from ~26.5 ka to ~19.5 ka53). Both divide into three further basal subclades, in each case a major one (K1a and K2a) and two minor ones (K1b, K1c; K2b, K2c). By far the most frequent of these subclades is K1a, dating to the end of the Last Glacial Maximum (LGM) at ~20 ka, and comprising numerous subclades of various ages, some of which are found predominantly in the Near East and some in Europe. K1b and K2c both date to the pre-LGM downturn, ~27 ka; K1c and K2b both date to the Late Glacial, ~17–18 ka; and K2a dates to the Neolithic, ~8.5 ka. All of the subclades, apart from K1a, are found mainly in Europe, with occasional lineages appearing in the Near East (K2c is extremely rare and remains enigmatic). All four of the minor subclades bifurcate again, with most multifurcating nodes appearing only within the last 10 ka, but within K1b some arise earlier, ~10–25 32 ambiguity, we cannot entirely rule this out, but Syrian Jews have Sephardic ancestry, and it would be unwise to draw conclusions from a single Iraqi sample, which may be the result of an “accidental”, quite possibly recent, instance of gene flow. In conclusion, the phylogeographic patterns strongly imply that of the three major Ashkenazi founders, K1a1b1a most likely traces its ancestry within Europe to the early Holocene, and K2a2a1 to at least the mid-Holocene and most likely much earlier, back to the glacial period ~30 ka. For the reasons mentioned above, K1a9 is less definite, but it most likely derives from K1a9’10’15’26’30, which would trace to Europe in the early Holocene, at a similar time and place to K1a1b1. A case might be made for a Levantine origin for the haplogroup K lineages on grounds of an insufficient Near Eastern sample database, so that the Ashkenazi lineages might simply have not yet been sampled in the Near East: a reservoir of haplogroup K variation, such as that identified by Shlush et al.52 in the Druze, might still be awaiting discovery. However, the distribution of Near Eastern lineages across the haplogroup K tree suggests that this is very unlikely. Near Eastern lineages, although diverse, are nevertheless not scattered throughout the tree, but fall mainly within particular subclades of haplogroup K1a, and are strongly concentrated in these particular clusters. The main exception, K1b1c, may be the exception that proves the rule, since it looks like a clear case of a Holocene migration from Europe. Given this pattern, it is difficult to imagine that there could be hidden Near Eastern reservoirs carrying three major clusters identified to date only in Europe, given also that 95 mitogenomes from the Near East/South Caucasus are included and that there is no suggestion of significant further variation in the much more extensive control-region dataset. Potential reservoirs with high haplogroup K frequencies (such as Georgia) follow the Near Eastern pattern and show no sign in the mitogenome data of 33 harbouring potential Ashkenazi founders. Chuvashia is another case in point (see Supplementary Note 2). Out of the 107 Near Eastern/South Caucasian lineages, there are only a small number of cases, detailed above, that “invade” the predominantly European subclades, and six of these are from Armenia or Anatolia, areas well known from previous mtDNA studies, as well as the historical record, to have experienced heavy influxes from Europe in the past few thousand years23. By comparison, for the Ashkenazi founder clades, K1a1b1 includes 37 European lineages, K1a9, 34 Europeans and one North African, and K2a2, eight Europeans; none includes a single Near Eastern lineage. Moreover, mtDNAs from Samaritans, who can reasonably be expected to have shared an ancestral Hebrew gene pool with Palestinian Jews and who, consistent with this, cluster with modern Druze and Lebanese in autosomal analyses3, lack any haplogroup K lineages, or indeed the other more common Ashkenazi founder mtDNAs (in N1b, H or J), whilst carrying Near Eastern lineages (predominantly T2h, U7 and H20)58,59 that include several that are in fact closely related to minor Ashkenazi lineages of putative Near Eastern origin (U6a, R0a)30,31. The question arises as to why an assimilation founder event might draw in several lineages from a single haplogroup (K) from a presumably diverse source population in Europe. However, we would caution against assuming that the source must have been enriched for haplogroup K. We have now shown that N1b2 underwent a similar process (main text; Fig. 6; Table S5), and numerous minor lineages may have been assimilated at the same time (with some lost later by drift). Moreover, whereas K1a1b1a and K1a9 were most likely assimilated in Mediterranean/Western Europe, the third major K founder, K2a2a1, with its slightly younger age estimate and its nesting within a German lineage, may have been assimilated in central Europe slightly later, again along with 34 numerous minor lineages (such as many of those from haplogroup H; see below). The survival of three distinct haplogroup K founder lineages through the early bottleneck may simply have been due to chance. Supplementary Note 2 Southern Russian mtDNA lineages and the “Khazar hypothesis” There is no evidence in the mtDNA pool to support the contention that lineages might have been recruited on a large scale from the North Caucasus and/or Chuvashia regions of southern Russia, as would be predicted by the “Khazar hypothesis”16. Not only are the subclades of haplogroup K that gave rise to the three major Ashkenazi founders absent, and N1b virtually so, but there is little evidence for any assimilation of the more minor Ashkenazi lineages either, most of which comprise mtDNAs found primarily in Europe. Extant southern Russians from these regions carry either distinct subclades of haplogroups from those found in the Ashkenazim (for example, amongst the more common west Eurasian clades such as haplogroups H, J and T lineages) or they lack the other haplogroups found in the Ashkenazim altogether. There are 208 sequences from the North Caucasus in the HVS-I database (excluding Kalmyks who arrived from China in the seventeenth century AD), of which eight belong to haplogroup K23. We fully sequenced seven of these and partially sequenced the eighth, in order to resolve their haplogroup status. All belonged to Near Eastern subclades, and none belonged to one of the three Ashkenazi founder clades. Two belonged to K1a3, two to K1a12a, and the remaining four belonged to K1a4 (including two belonging to K1a4a3 and one to K1a4g). There is a single N1b1 sequence, but no case of N1b2. 35 ~25% belong to haplogroup H but, unlike those in the Ashkenazim, most do not fall within typically European H subclades, with only a few H5 and H6 and a single H1b. HV1a is present, but not HV1b, which is the main HV1 subclade in the Ashkenazim. There are several examples of haplogroups C1 and D, as well as G2a and some Asian haplogroup N lineages, none of which is seen in the Ashkenazim. Similarly, most of the haplogroup J lineages in the North Caucasus are distinct from those in Europe, with far fewer J1c, which is the main European J subclade. There is, however, one exact match to the J1c7a1 sequence, also found in Eastern Europe, which may therefore be a potential candidate for a “Khazar” lineage. However, this lineage is found more commonly in west and central Europe, and the nesting pattern suggests that those in Eastern Europe may have spread from the Ashkenazim community into their neighbours, so the possibility remains remote. Again, haplogroup T in the North Caucasus lineages are markedly different, with the exception of the ubiquitous T1a, with the Ashkenazi/European T2b and T2a1b forming a very small minority of North Caucasus T2 lineages. The haplogroup U3 mtDNAs in the Caucasus are mostly U3b3, whereas those in the Ashkenazim are U3a. There is no U5b in the North Caucasus and the U5a lineages are not similar to those in the Ashkenazim. There are no haplogroup U7 or U6 lineages and those belonging to haplogroups I, W and X are also distinct. There is no haplogroup V7a, only V10a. M1a1 is present, but no M1a1b, and the situation is similar for U1b1. Overall, there is very little case to be made for any assimilation into Ashkenazi communities from this region. The language of the Chuvash, who now reside slightly to the northeast in the Middle Volga region, is thought to be close to that spoken by the Khazars, and the 36 region is considered to have been a sink for dispersals after the fall of the Khazar empire60. In our small sample of 36 Chuvashian mtDNAs, six (16.7%) belonged to haplogroup K. However, the four that we were able to test belonged to K1a4a1h, K1a5a and pre-K2a8, and a fifth can be identified as K1c1e. There are no N1b lineages, and few other lineages that might suggest any potential contribution to the Ashklenazi mtDNA pool: the only HVS-I matches are with the geographically extremely widespread root types of H5, H6a1a and V (3% of Ashkenazi lineages in total). Thus, despite the high frequency of haplogroup K, this region too cannot be considered a potential important source for the Ashkenazi maternal lineages. Supplementary Note 3 Minor Ashkenazi lineages: haplogroup H There are 29 Ashkenazi haplogroup H whole mitogenomes currently available (Supplementary Table S7). This compares favourably with the 40 from haplogroup K (overall 32% in the Ashkenazim, compared with haplogroup H at 23%), although the haplogroup H lineages are considerably more diverse. The present whole mitogenome Ashkenazi database might potentially be biased, but cross-checking of these lineages with the much larger Ashkenazi control-region database (where, for haplogroup H, n = 191) confirms their European distribution and suggests that the available mitogenome data from Ashkenazim give a fair picture of the variation. The control-region distributions for all Ashkenazi lineages (n = 836) are summarised in Supplementary Table S6. The overall frequency of haplogroup H in the Ashkenazi database is 23%, but this varies from only ~15% in Western Europe to >30% in parts of Eastern Europe. For comparison, haplogroup H is found in present-day 37 Europe at a rate of ~40%, the North Caucasus at 27%, and the Near East at 19%. Its major subclade, H1, varies from ~20–25% in Iberia to 10–15% in central and Eastern Europe to <10% in the North Caucasus and <5% in the South Caucasus/Near East, and H3 is almost entirely restricted to Western Europe23,61-64. The overall distribution of Ashkenazi haplogroup H control-region lineages is thus broadly similar to those of Europe and the Near East. Nevertheless, Ashkenazi haplogroup H lineages are highly distinctive in important respects. Virtually all of the Ashkenazi haplogroup H control-region sequences are found in the extant European mtDNA pool, but most are elevated to higher frequencies in the Ashkenazim. The fraction of singletons is correspondingly lower (17% versus 49% in a sample from Germany), as is the fraction of unique haplotypes overall (17% versus 37%). Haplogroup H, like haplogroup K, carries only a weak phylogenetic signal in the control region (especially HVS-I), which means that we cannot pinpoint the position of all of the Ashkenazi haplogroup H lineages in the control region database in the way that we can for many of the lineages in other haplogroups (see below). We have therefore estimated the fraction of Ashkenazi lineages within haplogroup H that can be allocated to a European source using the 29 mitogenomes. Even so, at least a third of the control-region lineages can be assigned to H1 and H3, and another 12% to the predominantly European H5. More specifically, almost a quarter can be allocated to specific subclades within haplogroup H1 in the global mitogenome tree (H1b, H1e4a, H1aj1) (Fig. 7) with another quarter allocated to H2e2a1, H3*, H5*, H5c2, H6a1a*, H6a1a1a, H10a1b, H11 and H15b (see Fig. 8; Supplementary Figs. S5-S10). Of these, only H5–73G (Supplementary Fig. S5), H6a1a1a (Fig. 8) and H15b (Supplementary Fig. S9) can plausibly be ascribed a potential Near Eastern origin. 38 There are several subclades within H5 (Supplementary Fig. S5), which again nest with west European lineages; and in H7 (Supplementary Fig. S6), which also appears to have arisen in Europe (although several distinct subclades have spread subsequently to the Near East). The latter (as for H1as and H40b: Supplementary Table S7) lack a clear control-region motif and are candidates for Ashkenazi control-region lineages lacking a distinctive motif. H6 is predominantly European but is also seen in the Near East. A Slovakian Ashkenazi mitogenome falls alongside an Iranian Jew into one subclade dating to the early Holocene, H6a1a1 (Fig. 8), which includes individuals from Arabia and the North Caucasus (and matching a Czech individual in the HVS-I database). This is seen in the HVS-I Ashkenazi database at 0.6%, widely dispersed from west to east European groups, and this intriguing pattern clearly calls for further study, but H6a1a has been found in a late Neolithic (Corded Ware) sample from central Europe (Fig. 8; see below)29. Further minor lineages belong to H11, H13a1a1, H15b, H26c and H40b (Supplementary Figs. S7–S10). Of these, only H15b (Supplementary Fig. S9) can be assigned a potential Near Eastern origin; H40b remains ambiguous, whilst all of the remainder nest deep within European lineages in the mitogenome tree. There is also an mtDNA from a single Spanish Sephardic Jew available, belonging to H56 which, like the Ashkenazi lineages, also nests within west European lineages, and a Bulgarian Jew belonging to H25, for which it is difficult to locate a source as the subclade emerges directly from the root of haplogroup H, with no nesting discernible (Supplementary Table S7). There is, furthermore, an mtDNA from a Turkish Jew belonging to H1p, emerging directly from the root of H1, which again most likely derives from a European source (Supplementary Table S7). There is therefore a 39 tentative suggestion of a European source for many Sephardic haplogroup H lineages as well. Although there are few whole-mitogenome data for Sephardim, the suggestion of a common European source with Ashkenazim is consistent with the autosomal evidence3,4,27. Finally, we note the presence in the whole-mitogenome database of a single haplogroup H13a2 lineage from a B’nei Israel individual from Mumbai2 (Supplementary Table S7). Despite the few Indian lineages on the haplogroup H tree, this mtDNA clusters into H13a2a with one from Andhra Pradesh36,65. H13a2 is otherwise a mixed European/Near Eastern clade that is difficult to recognise in the control region database, but the pinpointing of the Indian ancestry in this instance provides further corroboration for the phylogeographic approach adopted here. There are 22 Ashkenazi founders in total within the current haplogroup H mitogenome tree (Supplementary Tables S7, S8). Since the founder subclades at present comprise very few sequences, they can necessarily only be dated very imprecisely, but even so half of them date to <2ka (Supplementary Table S8). The remainder all include non-Ashkenazim individuals that inflate the age in at least some cases (i.e. in cases where the founder event was within the subclade, rather than cases where the whole subclade was involved in the founder event, followed by gene flow into neighbouring communities). In particular, the Near Eastern subclade H15b dates to the early Holocene, but includes a basal singleton Ashkenazi lineage, which may perhaps represent an individual recent migrant from the Near East (see below). Powerful independent evidence corroborating a deep European ancestry for these lineages comes from the study of prehistoric remains, recently achieved for a sample of 39 haplogroup H lineages from prehistoric central Europe29. Haplogroup H1, and in particular H1e, along with haplogroups H5 and H26, have been found in early 40 Neolithic remains, H3 and H7 in mid-Neolithic remains, H6 and H13 in the late Neolithic and H2, H4 and H11 in the early Bronze Age. Thus almost all of the clusters to which the Ashkenazi haplogroup H lineages belong are known to have been present in Europe by 3500 years ago – including some (such as H6) now present at appreciable levels in the Near East – confirming the phylogeographic predictions made on the basis of their extant distribution and molecular-clock estimates. The only lineages not present in the prehistoric sample (from the mitogenome tree evidence) are H15b, an early Holocene clade that may indeed have Near Eastern ancestry (but which can be identified from its HVS-II motif and is vanishingly rare in the control-region database) and H40b, which nests within mainly European lineages in the mitogenome tree. If we conservatively regard H6a1a1 as potentially having a Near Eastern source, we can therefore assign 26 out of the 29 (89.7%) Ashkenazi haplogroup H mitogenomes to a European source. Given the phylogenetic ambiguity of control-region data for haplogroup H, as mentioned above, this is the best estimate we can provide at present. 89.7% of the 22.9% of the control-region Ashkenazi database comprising haplogroup H suggests an overall European source component of 20.5%. Other minor Ashkenazi mtDNA lineages There are numerous other minor founders in the Ashkenazi mtDNA pool, as can be seen in the large control-region database (summarised in Supplementary Table S6), relatively few of which have yet been subjected to whole-mitogenome sequencing on the scale of haplogroup H, let alone haplogroups K and N1b. On the other hand, there is far less ambiguity in the phylogenetic position of most control-region lineages than for haplogroups H and K, so most of the remaining lineages can now be pinpointed fairly closely on the mitogenome tree of extant lineages from Europe and the Near East. 41 Moreover, by combining the control-region and whole-mitogenome database for west Eurasians, most of those can be further assigned to a geographical source with some confidence (Supplementary Table S9). After the four major founders and haplogroup H, haplogroup J is the next most common haplogroup in the Ashkenazim, at ~7% overall. As with the haplogroup H lineages, these provide further substantial corroboration for the importance of Europe as the primary source for Ashkenazi maternal lineages. Ashkenazi J lineages are unevenly distributed but are, like those of H, rarer in the west Ashkenazim (~5%) and more frequent in eastern groups, peaking in Russian/Belarusian Jews at ~16% (Supplementary Table S6). Again, a striking feature is elevated frequencies of rare haplotypes. (This reflects in reverse the distribution in the Ashkenazim of the main putative west Mediterranean lineages, K1a1b1a, K1a9, N1b2 and M1a1b.) The mitogenome tree for haplogroup J is well-known, with several Ashkenazi lineages included, and it is straightforward to locate most of the control-region sequences within it, as described in the main text30. As also discussed in the main text, at least 93% of Ashkenazi haplogroup J lineages, or 6.3% of total Ashkenazi lineages, have a likely European source (based on the frequencies in the control-region database, with lineages classified by cross-referencing to the whole-mitogenome tree30). Those within haplogroup J1c are shown in Fig. 10. Haplogroup T lineages comprise ~4.8% of the Ashkenazim control-region database and of these at least 60% have a likely European source, thus the European component overall is at least 3%. The haplogroup M1a1 lineages all belong to a single subclade within M1a1b (based on the HVS-I variant at position 16289). M1a1b is restricted to the north Mediterranean32, except for two individuals, from Saudi Arabia and Georgia