White Jews are not of lineage

ARTICLE

Received 11 Jul 2013 | Accepted 4 Sep 2013 | Published 8 Oct 2013

A substantial prehistoric European ancestryamongst Ashkenazi maternal lineagesMarta D. Costa1,2,*, Joana B. Pereira1,2,*, Maria Pala3, Veronica Fernandes1,2, Anna Olivieri4, Alessandro Achilli5,

Ugo A. Perego4,6, Sergei Rychkov7, Oksana Naumova7, Jiri Hatina8, Scott R. Woodward6,9, Ken Khong Eng1,10,

Vincent Macaulay11, Martin Carr3, Pedro Soares2, Luısa Pereira2,12 & Martin B. Richards1,3

The origins of Ashkenazi Jews remain highly controversial. Like Judaism, mitochondrial DNA

is passed along the maternal line. Its variation in the Ashkenazim is highly

distinctive, with four major and numerous minor founders. However, due to their rarity in the

general population, these founders have been difficult to trace to a source. Here we show that

all four major founders, B40% of Ashkenazi mtDNA variation, have ancestry in prehistoric

Europe, rather than the Near East or Caucasus. Furthermore, most of the remaining minor

founders share a similar deep European ancestry. Thus the great majority of Ashkenazi

maternal lineages were not brought from the Levant, as commonly supposed, nor recruited in

the Caucasus, as sometimes suggested, but assimilated within Europe. These results point to

a significant role for the conversion of women in the formation of Ashkenazi communities,

and provide the foundation for a detailed reconstruction of Ashkenazi genealogical history.

DOI: 10.1038/ncomms3543 OPEN

1 Institute of Integrative and Comparative Biology, Faculty of Biological Sciences, University of Leeds, Leeds LS2 9JT, UK. 2 IPATIMUP (Instituto de Patologia eImunologia Molecular da Universidade do Porto), Porto 4200-465, Portugal. 3 School of Applied Sciences, University of Huddersfield, Queensgate,Huddersfield HD1 3DH, UK. 4 Dipartimento di Biologia e Biotecnologie, Universita di Pavia, Pavia 27100, Italy. 5 Dipartimento di Chimica, Biologia eBiotecnologie, Universita di Perugia, Perugia 06123, Italy. 6 Sorenson Molecular Genealogy Foundation, Salt Lake City, Utah 84115, USA. 7 Vavilov Institute ofGeneral Genetics, Moscow 119991, Russia. 8 Charles University, Medical Faculty in Pilsen, Institute of Biology, CZ-301 66 Pilsen, Czech Republic. 9 Ancestry,Provo, Utah 84604, USA. 10 Centre for Global Archaeological Research, Universiti Sains Malaysia, 11800 USM Penang, Malaysia. 11 School of Mathematicsand Statistics, University of Glasgow, Glasgow G12 8QQ, UK. 12 Faculdade de Medicina da Universidade do Porto, Porto 4200-319, Portugal. * These authorscontributed equally to this work. Correspondence and requests for materials should be addressed to M.B.R. (email: [email protected]).

NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications 1

& 2013 Macmillan Publishers Limited. All rights reserved.

The origins of Ashkenazi Jews—the great majority ofliving Jews—remain highly contested and enigmatic tothis day1–11. The Ashkenazim are Jews with a recent

ancestry in central and Eastern Europe, in contrast to Sephardim(with an ancestry in Iberia, followed by exile after 1492),Mizrahim (who have always resided in the Near East) andNorth African Jews (comprising both Sephardim and Mizrahim).There is consensus that all Jewish Diaspora groups, includingthe Ashkenazim, trace their ancestry, at least in part, to theLevant, B2,000–3,000 years ago5,12–14. There were Diasporacommunities throughout Mediterranean Europe and the NearEast for several centuries prior to the destruction of the SecondTemple in Jerusalem in 70 CE (Common Era), and some scholarssuggest that their scale implies proselytism and wide-scaleconversion, although this view is very controversial9,15.

The Ashkenazim are thought to have emerged from dispersalsnorth into the Rhineland of Mediterranean Jews in the earlyMiddle Ages, although there is little evidence before the twelfthcentury5,15. After expulsions from Western Europe between thethirteenth and fifteenth centuries, the communities are thought tohave expanded eastwards, especially in Poland, Lithuania andthen Russia. The implied scale of this expansion has led some toargue, again very controversially, for mass conversions in theKhazar kingdom, in the North Caucasus region to the north andeast of the Black Sea, following the Khazar leadership’s adoptionof Judaism between the ninth and tenth centuries CE8,9.

We are then faced with several competing models forAshkenazi origins: a Levantine ancestry; a Mediterranean/westEuropean ancestry; a North Caucasian ancestry; or, of course, ablend of these. This seems an ideal problem to tackle with geneticanalysis, but after decades of intensive study a definitive answerremains elusive. Although we might imagine that such anapparently straightforward admixture question might be readilyaddressed using genome-wide autosomal markers, recent studieshave proposed contradictory conclusions. Several suggest aprimarily Levantine ancestry with south/west European admix-ture3,4, but another concludes that the ancestry is largelyCaucasian16, implying a major source from converts in theKhazar kingdom17. An important reason for disagreement is thatthe Ashkenazim have undergone severe founder effects duringtheir history, drastically altering the frequencies of geneticmarkers and distorting the relationship with their ancestralpopulations.

This problem can be resolved by reconstructing the relation-ships genealogically, rather than relying on allele frequencies,using the non-recombining marker systems: the paternallyinherited male-specific part of the Y chromosome (MSY) andthe maternally inherited mitochondrial DNA (mtDNA). Thiskind of analysis can be very powerful, because nesting ofparticular lineages within clusters from a particular geographicalregion allows us to pinpoint the source for those lineages, byapplying the parsimony principle. This has indeed beenattempted, with the MSY results interpreted plausibly to suggestan overwhelming majority of Near Eastern ancestry on theAshkenazi male line of descent11,18–21, albeit with much higherlevels (450%) of European (potentially east European) lineagesin Ashkenazi Levites22, suggesting a possible Khazar source inthat particular case.

The maternal line has also been studied, and indeed AshkenazimtDNAs are highly distinctive, but they have proved difficult toassign to a source population1,2,11. Some progress has been madeby targeting whole-mtDNA genomes or mitogenomes, whichprovide much higher genealogical (and therefore geographical)and chronological resolution than the control-region sequencesused previously—although the far larger control-region databaseremains an invaluable guide to their geographic distribution.

Using this approach, Behar et al.2 identified four major founderclusters, three within haplogroup K—amounting to 32% ofsampled Ashkenazi lineages—and one within haplogroup N1b,amounting to another 9%. These lineages are extremelyinfrequent across the Near East and Europe, making theidentification of potential source populations very challenging.Nevertheless, they concluded that all four most likely arose in theNear East and were markers of a migration to Europe of peopleancestral to the Ashkenazim only B2,000 years ago1,2. Theremaining B60% of mtDNA lineages in the Ashkenazimremained unassigned to any source, with the exception of theminor haplogroup U5 and V lineages (B6% in total), whichimplied European ancestry1,23.

Here we focus on both major and minor founders, with a muchlarger database from potential source populations. We firstanalyse 956 (72 newly generated) mitogenomes from haplogroupU8 (including 909 from haplogroup K, U8’s major subclade): 477of these are from Europe and 106 from the Near East/Caucasus.We show that European and Near Eastern lineages largely fallinto discrete, ancient clusters, with minor episodes of gene flow,suggesting that haplogroup K diversified separately in Europeand the Near East during the last glacial period. Of the threeAshkenazi founders, K1a1b1a and K1a9 were most likelyassimilated in west (perhaps Mediterranean) Europe andK2a2a1 in west/central Europe. Most surprisingly, by analysingtwo new N1b2 sequences selected from a database of 278 N1bHVS-I sequences, in the context of 44 published N1b sequences24,we show that the highly distinctive N1b2 subclade, making upanother 9% of Ashkenazi lineages, was likely assimilated inMediterranean Europe, rather than in the Near East as previouslyproposed2. Moreover, from a survey of another 42,500 completemtDNA genomes and 428,000 control-region sequences fromEurope, the Near East and the Caucasus, in comparison with theavailable database of 836 Ashkenazi control-region sequences anda handful of published mitogenomes, we also evaluate the minorfounders. Overall, we estimate that most (480%) AshkenazimtDNAs were assimilated within Europe. Few derive from a NearEastern source, and despite the recent revival of the ‘Khazarhypothesis’16, virtually none are likely to have ancestry in theNorth Caucasus. Therefore, whereas on the male side there mayhave been a significant Near Eastern (and possibly east European/Caucasian) component in Ashkenazi ancestry, the maternallineages mainly trace back to prehistoric Western Europe. Theseresults emphasize the importance of recruitment of local womenand conversion in the formation of Ashkenazi communities, andrepresent a significant step in the detailed reconstruction ofAshkenazi genealogical history.

ResultsFour major founder lineages within haplogroup K and N1b.Haplogroup K arose within haplogroup U8B36 ka, in Europe orthe Near East, with the minor subclades K1b, K1c and K2 all mostlikely arising in Europe, between the last glacial period and theNeolithic (Fig. 1; Supplementary Note 1; Supplementary Data 1–3;Supplementary Figs S1–S3; Supplementary Tables S1–S3). K1aexpanded from B20 ka onwards, both in the Near East andEurope, with its major subclade, K1a1b1 (Fig. 2), mainly restrictedto Europe (with a few instances in North Africa), arriving fromthe Near East by B11.5 ka, the beginning of the Holocene(Supplementary Note 1).

Almost half of mtDNAs in west/central European AshkenaziJews belong to haplogroup K, declining to B15% in eastEuropean Jews1,11, with almost all falling into three subclades:K1a1b1a, K1a9 and K2a2a12,25 (Figs 1–4; Supplementary Fig. S4).These three founder clusters show a strong expansion signal

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543

2 NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications

& 2013 Macmillan Publishers Limited. All rights reserved.

beginning B2.3 ka, with the overall effective population size forthese lineages increasing 13-fold by 275 years ago (Fig.1).

K1a1b1a (slightly re-defined, due to the improved resolution ofthe new tree) (Fig. 2) accounts for 63% of Ashkenazi K lineages(or B20% of total Ashkenazi lineages) and dates to B4.4 ka withmaximum likelihood (ML); however, all of the samples within it,except for one, nest within a further subclade, K1a1b1a1, datingto B2.3 ka (Supplementary Data 2). K1a1b1a1 is also present innon-Ashkenazi samples, mostly from central/east Europe. As theyare nested by Ashkenazi lineages, these are likely due to gene flow

from Ashkenazi communities into the wider population. Thepattern of gene flow out into the neighbouring communities isseen in the other two major K founders, and also in haplogroupsH and J; it is especially clear when the nesting and nestedpopulations are more distinct, for example in the case ofhaplogroup HV1b, which has a deep ancestry in the Near East(Fig. 5; Supplementary Table S4).

The K1a1b1 lineages within which the K1a1b1a sequences nest(including 19 lineages of known ancestry) are solely European,pointing to an ancient European ancestry. The closest nesting

Likely near eastern origin

Undetermined origin

Likely European origin

Europeans and Ashkenazim

55

U8

U8b1

U8b

U8a

K1a1b

K1a1K1a1a K1a2

K1a9′10′15′26′30

K1a1b1

K1a1b1a

K1a1b1a1 K1a9

K1a28 K1a3a

K1a4a1

K1a

K1a3 K1a4K1a8 K1a12

K1

K1c

K1d′e′f K1b

K1b1

K1b1a K1b2

K1b1c

K2a2a1

K1c2

K1c1

K1a13′16′31

K2a

K2b

K2

K

K2c

50

45

40

35

30

25

20

15

10

5

0

1,000,000

Effe

ctiv

e po

pula

tion

size

(N

ef)

100,000

10,000

1,000

100

100 5 10

10

15 20

Time (ka)

Figure 1 | Inferred ancestry of the main subclades within haplogroup U8. The timescale (ka) is based on ML estimations for mitogenomes. Inset:

Bayesian skyline plot of 34 Ashkenazi haplogroup K lineages, showing growth in effective population size (Nef) over time.

114

2001,717G16,09216,184A

14,388 4,8236,528

8,842C

14,569

3,79613,05016,09311,765A

8276,28410,609

16,09216,223

K1a1b1d K1a1b1fK1a1b1a

K1a1b1a1

K1a1b1

K1a1b1e

K1a1b1c

K1a1b1b

10,97816,234

12,954

11,005 152

8,787 9,92114,249

8,02314,118

195 3,316

114! 114!

8,46211,020

1,393

189

15,355

16,355152

5,46016,093

1,709

16,223

513

16,278 7238,047

14,203C14,517

16,234!

6,366

477

9,21416,288

8,521

7,927

9,86112,189

114

9,932

1463,705

16,0938,291

15,04711,204

195

5,58312,007189

114

5,87616,362

39014,279

0

5,58516,222

789

5932,483

5,74611,620

11413,85116,093

2,62813,44316,093

5,74215,07416,224!16,278

9,644

Ashkenazi JewUSA

EuropeUnknown

3

6

9

12

Figure 2 | Phylogenetic tree of haplogroup K1a1b1. Time scale (ka) based on ML estimations for mitogenome sequences.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543 ARTICLE

NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications 3

& 2013 Macmillan Publishers Limited. All rights reserved.

lineages are from Italy, Germany and the British Isles, with othersubclades of K1a1b1 including lineages from west and Mediter-ranean Europe and one Hutterite (Hutterites trace their ancestryto sixteenth-century Tyrol)26. Typing/HVS-I results have alsoindicated several from Northwest Africa, matching EuropeanHVS-I types2, likely the result of gene flow from Mediter-ranean Europe. K1a1b1a is also present at low frequencies in

Spanish-exile Sephardic Jews, but absent from non-EuropeanJews, including a database of 289 North African Jews2,25. Notably,it is not seen in Libyan Jews25, who are known to have a distinctNear Eastern ancestry, with no known influx from Spanish-exileimmigrants (although Djerban Jews, with a similar history, havenot been tested to date for mtDNA, they closely resemble LibyanJews in autosomal analyses27). Thus the Ashkenazi subclade ofK1a1b1 most likely had a west European source.

K1a9 (Fig. 3; Supplementary Fig. S4), accounting for another20% of Ashkenazi K lineages (or 6% of total Ashkenazi lineages)and also dating to B2.3 ka with ML (Supplementary Data 2)again includes both Ashkenazi and non-Ashkenazi lineages solelyfrom east Europeans (again suggesting gene flow out into thewider communities). Like K1a1b1a, it is also found, at muchlower frequencies, in Sephardim. Here the ancestral branchingrelationships are less clear (Supplementary Note 1 andSupplementary Fig. S4), but K1a9 is most plausibly nested withinthe putative clade K1a9010015026030, dating to B9.8 ka, whichotherwise includes solely west European (and one Tunisian)lineages, again pointing to a west European source.

K2a2 (Fig. 4) accounts for another 16% of AshkenaziK lineages (or B5% of total Ashkenazi lineages) and dates toB8.4 ka (Supplementary Data 2). Ashkenazi lineages are oncemore found in a shallow subclade, K2a2a1, dating to B1.5 ka,that otherwise again includes only east Europeans, suggestinggene flow from the Ashkenazim. Conversely, the nesting clades,K2a2 and K2a2a, although poorly sampled, include only Frenchand German lineages. K2a2a is not found in non-EuropeanJews25.

Haplogroup K is rarer in the North Caucasus than in Europe orthe Near East (o4% (ref. 23)) and the three Ashkenazi founderclades have not been found there (Supplementary Note 2). Wetested all eight K lineages out of 208 samples from the NorthCaucasus, and all belonged to the Near Eastern subclades K1a3,K1a4 and K1a12. Haplogroup K is more common in Chuvashia,but those sampled belong to K1a4, K1a5 and pre-K2a8.

The fourth major Ashkenazi founder mtDNA falls withinhaplogroup N1b (ref. 2). The distribution of N1b is much morefocused on the Near East than that of haplogroup K (ref. 24),and the distinctive Ashkenazi N1b2 subclade has accordinglybeing assigned to a Levantine source2. N1b2 has until now been

14,947

K1a30

K1a26

13,651

16,192

67,25A

4,113

8,4008,521

16,214A16,354

152

146

4,7395,5635,964

11,98912,71115,75816,124

31012,063

14,440

16,527 11,453

11,287 2,258

3,3386,515

16,093!

14,83115,758

152

16,524

K1a9

K1a9′10′15′26′30

K1a10

K1a10a

14,160

195

5,24012,69615,226

16,048

16,291

3168,76416,04716,093!

15,43116,201

5,3009,4779,698!9,951

16,093!230T16,051

338

1,9585,6556,22715,204

5,054

Ashkenazi Jew

USANorth Africa

Europe

Unknown

16,223

2,4834,452

16,249

8,155

8

6

4

2

0

9,62912,40314,564

93 16,093!

K1a15

Figure 3 | Phylogenetic tree of haplogroup K1a9 in the context of the putative clade K1a9010015026030. Time scale (ka) based on ML estimations for

mitogenome sequences.

Europe

8,697

K2a2

K2a2a

K2a2a1

11,348

1959,263T16,390

6415,520

8,697! 63 512C9,25411,914

11,719!9,461

4,325

9,214153

14,599

Unknown

Ashkenazi Jew

USA

10

7.5

5

2.5

0

Figure 4 | Phylogenetic tree of haplogroup K2a2. Time scale (ka) based

on ML estimations for mitogenome sequences.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543

4 NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications

& 2013 Macmillan Publishers Limited. All rights reserved.

found exclusively in Ashkenazim, and although it dates to onlyB2.3 ka, it diverged from other N1b lineages B20 ka (ref. 24)(Supplementary Table S5). N1b2 can be recognized in the HVS-Idatabase by the variant 16176A, but Behar et al.2 tested 14 NearEastern samples (and some east Europeans) with this motif andidentified it as a parallel mutation. Therefore, despite the longbranch leading to N1b2, no Near Eastern samples are known tobelong to it.

In our unpublished database of 6991 HVS-I sequences,however, we identified two Italian samples with the 16176Amarker, which we completely sequenced. We confirmed that theybelong to N1b2 but diverge before the Ashkenazi lineages B5 ka,nesting the Ashkenazi cluster (Fig. 6; Supplementary Table S5).This striking result suggests that the Italian lineages may be relictsof a dispersal from the Near East into Europe before 5 ka, and thatN1b2 was assimilated into the ancestral Ashkenazi population onthe north Mediterranean B2 ka. Although we found only twosamples suggesting an Italian ancestry for N1b2, the control-region database available for inspection is very large (28,418HVS-I sequences from Europe, the Near East and the Caucasus,of which 278, or B1%, were N1b). Moreover, the conclusion issupported by our previous founder analysis of N1b HVS-Isequences, which dated the dispersal into Europe to the latePleistocene/early Holocene24.

Minor Ashkenazi mtDNA lineages. There is now a largenumber of mitogenomes from Europe, the Caucasus and theNear East (B3,500, with 470 Ashkenazim), and a substantialAshkenazi mtDNA control-region database of 836 samples1,2,11

(Supplementary Table S6). We therefore endeavoured to cross-reference the two in order to pinpoint most of the control-regiondata within the mitogenome phylogeny.

Besides the four haplogroup K and N1b founders, the majorhaplogroup in Ashkenazi Jews is haplogroup H, at 23% ofAshkenazi lineages, which is also the major haplogroup in

Europeans (40–50% in Europe, B25% in the North Caucasus andB19% in the Near East)28. There are 29 Ashkenazi Hmitogenomes available (Supplementary Table S7), 26 (90%) ofwhich nest comfortably within European subclades dating to theearly Holocene (Supplementary Note 3, Figs 7 and 8;Supplementary Figs S5–S10; Supplementary Table S8). Most, infact, nest more specifically within west/central Europeansubclades, with closely matching sequences in east Europe, aswith the pattern for the K founder clades. The Ashkenazimitogenomes from haplogroup H include 39% belonging to H1or H3, which are most frequent in west Europe and rare outsideEurope. The nesting relationships in some cases point (albeittentatively) to a central European source, but in many casescomparison with the HVS-I database indicates matches in westEurope. The phylogeographic conclusions based on the nestingrelationships are strongly supported for haplogroup H byevidence from the study of prehistoric remains, showing inalmost all cases that the lineages concerned were present inEurope since at least the early Bronze Age, B3.5 ka(Supplementary Table S7)29. There is no suggestion ofassimilation from the North Caucasus, where most H lineagesdiffer from those of Europe23 (Supplementary Note 2).

Haplogroup J comprises 7% of the Ashkenazi control-regiondatabase. Around 72% of these can be assigned to J1c, nowthought to have arisen within Late Glacial Europe30, and 19%belong to J1b1a1, also restricted to Europe. Thus 490% of theAshkenazi J lineages have a European origin, with B7% (J1b andJ2b) less clearly associated. Many have a probable west/centralEuropean source, despite (like H) being most frequent in easternAshkenazim. The four Ashkenazi J mitogenomes, in J1c5, J1c7a1aand J1c7d, once again show a striking pattern of Mediterranean,west and central European lineages enclosing Ashkenazi/eastEuropean ones (Fig. 9).

Haplogroups U5, U4 and HV0 (6.3% between them overall)arose within Europe. Some of these lineages, which are againmore frequent in the eastern than western Ashkenazi, may have

Europe12,696

HV1b

HV1b3

HV1b2

HV1b1

HV1b1b HV1b1a

152

4,047 1955,250

14,30516,15816,234

3,687

5,656

11,314

2,7559,117

13,7083,5917,9128,027

152

14,861

961

16,158 5,46014,46416,129

7094,856

1503,2905,1346,2639,585

2,626

20

15

10

5

0

4,7397,59816,274

8,0208,71510,29510,75012,87914,16116,311

1511837,664

15,17215,23615,519

16,1789,43816,129

1,6941315,033

11,081T 15,463

3,5476,02316,189

5,32713,434 16,399

10,09516,526

Anatolia, South Caucasus and the Near East

Unknown

Ashkenazi Jew

Eastern Africa

North Africa

Figure 5 | Phylogenetic tree of haplogroup HV1b. Time scale (ka) based on ML estimations for mitogenome sequences.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543 ARTICLE

NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications 5

& 2013 Macmillan Publishers Limited. All rights reserved.

been assimilated in central Europe. The haplogroup T lineages(5% overall) are more difficult to assign, but at least 60% (inT2a1b, T2b, T2e1 and T2e4) are likely of European and B10%(T1b3 and T2a2) Near Eastern origin30. The haplogroup Ilineages have evidently been present in Europe at least since theNeolithic, as indicated by both phylogeographic and ancientDNA analyses31. Haplogroup W3 may have originated in theNear East but spread to Europe as early as the Late Glacial31. TheM1a1b lineage is characteristic of the north Mediterranean andwas most likely assimilated there32, but the U6a and L2a1llineages are more difficult to pin down.

The main lineages with a potentially Near Eastern sourceinclude HV1, R0a1a and U7a5 (B8.3% in all). HV1b2 mito-genomes, in particular, date to B2 ka and nest within a clusterof Near Eastern HV1b lineages dating to B18 ka (Fig. 5;Supplementary Table S4). Others such as U1a and U1b have anultimately Near Eastern origin but, like N1b, have beensubsequently distributed around the north Mediterranean. Ingeneral, it is more difficult to assign lineages to a Near Easternsource with confidence, as the much larger control-regiondatabase indicates that (as with N1b2) many lineages with deepNear Eastern ancestry became widely dispersed along the northMediterranean during the Holocene, and may alternatively havebeen assimilated there.

If we allow for the possibility that K1a9 and N1b2 might have aNear Eastern source, then we can estimate the overall fraction ofEuropean maternal ancestry at B65%. Given the strength of thecase for even these founders having a European source, however,our best estimate is to assign B81% of Ashkenazi lineages to aEuropean source, B8% to the Near East and B1% further tothe east in Asia, with B10% remaining ambiguous (Fig. 10;

Supplementary Table S9). Thus at least two-thirds and most likelymore than four-fifths of Ashkenazi maternal lineages have aEuropean ancestry.

DiscussionThe extent to which Ashkenazi Jewry trace their ancestry to theLevant or to Europe is a long-standing question5, which remainshighly controversial3,4,6,12–14,16,17. Our results, primarily from thedetailed analysis of the four major haplogroup K and N1bfounders, but corroborated with the remaining AshkenazimtDNAs, suggest that most Ashkenazi maternal lineages tracetheir ancestry to prehistoric Europe.

Previous researchers proposed a Levantine origin for the threeAshkenazi K founders from several indirect lines of evidence:shared ancestry with non-Ashkenazi Jews, shared recent ancestrywith Mediterranean samples, and their absence from amongstnon-Jews2, and this suggestion has been widely accepted4.However, our much more detailed analyses show that two ofthe major Ashkenazi haplogroup K lineages, K1a1b1a and K2a2a1have a deep European ancestry, tracing back at least as far as theearly and mid-Holocene respectively. They both belong to ancientEuropean clades (K1a1b1 and K2) that include primarilyEuropean mtDNAs, to the virtual exclusion of any from theNear East. Despite some uncertainty in its ancestral branchingrelationships, a European ancestry seems likely for the thirdfounder clade, K1a9. The heavy concentration of Near Easternhaplogroup K lineages within particular, distinct subclades of thetree, and indeed the lack of haplogroup K lineages in Samaritans,who might be expected to have shared an ancestral gene pool withancient Israelites, both strongly imply that we are unlikely to have

Europe

Anatolia, South Caucasus and the Near East

Ashkenazi Jew

North Caucasus

North Africa

1,7033,921A4,9608,472

12,82216,145

N1b

N1b1

9,33511,362

2,26310,497

4,55.1T1,703!8,0844,136

4,22716,311 N1b1a

15,944d16,27116,343

4,967

150235

16,3118,472!12,37216,180

5,9879,921

16,093

93.1A769

4,4619,116

12,771

10,37313,41915,31716,209

7,0107,3378,4699,1339,335!14,690

16,03716,07516,311

16,569iGATC

1858,443

1,593!3,2214,8205,2915,5538,309

11,050

1996,7528,1558,251!8,2648,888

16,362

8,29013,96716,223!

27110,909

9,2309,335!9,882

13,60815,04315,88316,390!

146150320961

9,65.2C3,0838,962

9,093C

N1b1c

9,957

16,256

8,0209,43812,89113,76815,07115,07916,093

N1b1d

8673,3088,477

13,85114,56016,249

16,129

185188

8,763

8,2618,41016,291

9,86110,68816,257

3,571.1C14,053

15,813G16,297

5,5286,0458,020

14,581

551d8,676

11,719!15113,635

379563

16,390!

13,708

14,118

681789

7,526T13,114A

4,735A4,91711,92812,09213,12913,710

16,176A

N1b216,093

5,2376,272

9,92115,790G

152!12,79714,470

146195

1,4068,95015,92416,29716,311

12,29716,176A16,223!

16,126

195

N1b1b

4,904

5,4807,142

9,185G13,35016,12416,25616,400

20

15

10

5

0

Figure 6 | Phylogenetic tree of haplogroup N1b. Time scale (ka) based on ML estimations for mitogenome sequences.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543

6 NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications

& 2013 Macmillan Publishers Limited. All rights reserved.

H1a

H1b2

H1b H1c

H1e2 H1e1

H1e1a

H1e

H1e1b

H1e3

H1e4

H1e4a

H1h

H1j H1i

H1f1

H1g

H1f

1

76 1

JQ704894

JQ703268 46 1 1

EU

148452

EU

262984

JQ705236

JQ703137 14 1 44 3 3 5 1 7 2 2 1 10 6 2 73 2 1

H1b2a

H1b2a1

3 1

H1n

H1r

H1s

H1t1 H1t2

H1t H1u

H1m H1o

H1q

H1t1a

6 3 33 6 1 14 1 4 12 7 1 3

H1x

H1ae

H1af H1ah

H1aj H1ak

H1am

H1ao

H1v

H1w H1y

H1ab

H1aa

H1z

H1ac H1an

H1ap1

H1ap

H1ai H1ag

1 6 3 3

H1

6 1 3 1 5

1 9 1 4 6 1 4

JQ703788 1 1 11 1 6 2 2 3

H1aw

H1az

H1ba H1bc

H1bd

H1be H1ax

H1ar H1as H1at

H1au

H1av

13,386

H1aq

H1bb

H1as1

4 3 2 3 3 21 4 2 2 6 11 5 3 1 1

H1as2

JQ704370 1 1

H1bk

H1bp

H1bq

H1bs

H1bt

H1bv

H1bw

H1bx

H1by H1br

H1bi

H1bg

H1bf

2 2 1 1 3 1 1 2 1 2 2 3 1 4 1

H1bh

1 73 3 11

5

1

Brotherton et al. 29

JQ703655

84 1 1

73

16,162

8,251 16,189 16,080

183 8,286

16,183C 16,360

7,691

8,285.2C

16,356 477 5,460

8,512

14,902 453

15,817 16,129

93 960.1C 1,462 4,883

3,316

16,114

14,259

16,311Y

8,951

4,452 9,066s 16,093 16,189

7,013

4,733 152 6,237A

7,309

14,212 16,189

146 2,098

4,859

9,356

8,572 9,986

5,978 14,129

14,224

9,923

15,323

16,311 16,355

150 14,053

7,765 10,410A 16,037 16,256

10,314

8,966

15,047 16,189

15,299 16,189

4,131

327 11,428

11,893

15,553

16,357

6,272 14,869

8,950 12,507

6,722 15,088

14,133

207 8,618 9,621

16,172 16,192 16,456

3,666

4,763A

93

146 11,809 16,278

152 15,394

8,410

152 5,780

16,189

183 16,468

4,688

980

150 6,216

15,758

11,515A 16,148

7762s

460 16,129

5,054C 7,471 8,429

8,308

16,270 10,325

9,966

10,750 13,035

152 11,864

522dCA 11,084

10003 16,126

14,467

10,454

16,220C

16,527

12,681

8,478

789 8,740

9,921

3,745 16,239

5,054 7,849

13,768 11,377 8,701

Europe North Caucasus

North Africa America/Canada/Australia Ashkenazi Jew

Anatolia/South Caucasus/Near East

3,010

10,006R

Figure 7 | Schematic phylogenetic tree of haplogroup H1. Only the Ashkenazi lineages are shown in full detail; the distribution of other lineages is

indicated using small squares by the number present in the full tree for each subclade. Prehistoric European (all Neolithic, except for the H1aw lineage,

which dates to the Iron Age) lineages are shown using red circles29.

Europe

Unknown

Ashkenazi Jew

USA

Iran Jew

North Caucasus

Near East, South Caucasus and Anatolia

8

65,460

249d

6,54814,029

10,237

16,14810,936

5,302

1,5987,0017,364

297

150

9,77311,662

14616,482!

7482,0104,947

3,705

2,361

9,02515,287

14,970

537G980

16,092

19511,204

16,482!

936,46813,105

1507,805

14,182A16,319

7,3259,36211,61116,311

4,991

3,944

H6a1a1

239!16,218

7,202H6a1a2

H6a1a

939,05510,187

15,226C2,581

11,914

12,50116,527

8,97811,978A8,047

2,3525,23716,218

56813,953

15,884

152

13,02016,295

12,36914,560

6073,16C9,948

16,35614,094

16,311

14,944

10,3707,080

16,482!

2046,182

4,580

16,298

6,18516,145

9,0683,397

9,055

7,81316,192

14616189

41152

14,52716,526

16,16816,172

11,253

3,5485,04810,16614,211

10,58616,278

5785

6,2609,54513,191

288

70916,140

7,2697,094

16,311

827

195

H6a1a3H6a1a6H6a1a8

H6a1a8a

H6a1a3a

4116,482!

H6a1a2a

H6a1a2b

H6a1a5

H6a1a4

H6a1a7

H6a1a1a

16,297

11,923

4

2

0

H6a1a2b1

Figure 8 | Phylogenetic tree of Ashkenazi founders within haplogroup H6a1a. Time scale (ka) based on ML estimations for mitogenome sequences.

A Late Neolithic Corded Ware lineage from central Europe29 is shown in red emerging directly from the root.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543 ARTICLE

NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications 7

& 2013 Macmillan Publishers Limited. All rights reserved.

missed a hitherto undetected Levantine ‘reservoir’ of haplogroupK variation (Supplementary Note 1).

Furthermore, our results suggest that N1b2, for which a NearEastern ancestry was proposed (with much greater confidencethan for K) by Behar et al.2, is more likely to have beenassimilated into the ancestors of the Ashkenazi in the northMediterranean. Finally, our cross-comparison of control-regionand mitogenome databases shows that the great majorityof the remaining B60% of Ashkenazi lineages, belonging tohaplogroups H, J, T, HV0, U4/U5, I, W and M1 also have apredominantly European ancestry.

Overall, it seems that at least 80% of Ashkenazi maternalancestry is due to the assimilation of mtDNAs indigenous to

Europe, most likely through conversion. The phylogeneticnesting patterns suggest that the most frequent of the AshkenazimtDNA lineages were assimilated in Western Europe, B2 ka orslightly earlier. Some in particular, including N1b2, M1a1b, K1a9and perhaps even the major K1a1b1, point to a northMediterranean source. It seems likely that the major founderswere the result of the earliest and presumably most profoundwave of founder effects, from the Mediterranean northwardsinto central Europe, and that most of the minor founderswere assimilated in west/central Europe within the last 1,500years. The sharing of rarer lineages with Eastern Europeanpopulations may indicate further assimilation in some cases, butcan often be explained by exchange via intermarriage in thereverse direction.

The Ashkenazim therefore resemble Jewish communities inEastern Africa and India, and possibly also others across the NearEast, Caucasus and Central Asia, which also carry a substantialfraction of maternal lineages from their ‘host’ communities11,25.Despite widely differing interpretations of autosomal data, theseresults in fact fit well with genome-wide studies, which imply asignificant European component, with particularly closerelationships to Italians3,4,6,7. As might be expected from theautosomal picture, Y-chromosome studies generally show theopposite trend to mtDNA (with a predominantly Near Easternsource) with the exception of the large fraction of Europeanancestry seen in Ashkenazi Levites22.

Evidence for haplotype sharing with non-Ashkenazi Jews foreach of the three main haplogroup K founders may imply apartial common ancestry in Mediterranean Europe for Ashkenaziand Spanish-exile Sephardic Jews, but may also, at least in part, bedue to subsequent gene flow, especially into Bulgaria and Turkey,both of which witnessed substantial immigration from Ashkenazicommunities in the fourteenth and fifteenth centuries. Gene flowcould have been substantial in some cases—ongoing intermar-riage is likely when these communities began living in closerproximity after the Spanish exile6. A partial common ancestry forall European Jews—both Ashkenazi and Sephardic—is againstrongly supported by the autosomal results3,4.

J1c1

J1c2a

J1c2

21 6 1 3318 21 3 3

J1c3

J1c4

HM

159445

J1c6

J1c5b

20 2 1 14 1 9 4 1 2

J1c8

J1c7

J1c7a

J1c10

HM

627319

J1c11

JF812166

J1c12

J1c12a

J1c12b

J1c9

5 2 11115 2 2

J1c7a1a

J1c7a1

2 1 1 1 2 1 3 1 32 2 3 1 1 1

5

10

15

20

0

J1c

J1c5

J1c5a

EuropeNorth Caucasus

North AfricaAmerica/Canada/AustraliaAshkenazi Jew

Anatolia/South Caucasus/Near East

Asia

4823,394

16,519

188

13,934

9,63212,083G

5,198

1524,4845,460

12,3612,387

10,1924,688s

13,434s

4,4525,339

15,91616,309

228!368

11,08716,28716,527

7,3647,372

14,769

522-23d

185!

4,8389,100

5,9787,3407,888

14,798!

185!16,519

10,598

3,592

522-23d

1888,255

146

228!

185228

14,798

4,025

10,08410,398!

16,319 12,453

16,261

6,55412,127

16,092

6,464A13,681

6,887

185!5,024s

1955,291

15,103

15,26216,224

178R12,63016,255

1468,1529,117

13,91714,18215,51416,527

15,113

Figure 9 | Schematic phylogenetic tree of haplogroup J1c. Only the Ashkenazi lineages are shown in full detail; the distribution of other lineages is

indicated using small squares for each subclade with the number present in the full tree given in each case. For the full tree see Pala et al.30 Time scale (ka)

based on ML estimations for mitogenome sequences.

H: 20.5%

HV0: 4.1%I: 1.3%

J: 6.3%

K: 31.8%

M1: 0.7%

N1b: 9.2%

T: 3.0%U: 0.2%

U5: 2.0%

W: 1.6%

Near Eastern:8.3%

Asian: 1.1%

Unassigned:9.9%

European:81%

Figure 10 | Estimated contributions of European mtDNA lineages

to the Ashkenazi mtDNA pool shown by major haplogroup. The possible

overall Near Eastern contribution and fraction of unassigned lineages

are also indicated.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543

8 NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications

& 2013 Macmillan Publishers Limited. All rights reserved.

Jewish communities were already spread across the Graeco-Roman and Persian world 42,000 years ago. It is thought that asubstantial Jewish community was present in Rome from at leastthe mid-second century BCE, maintaining links to Jerusalemand numbering 30,000–50,000 by the first half of the first centuryCE15. By the end of the first millennium CE, Ashkenazicommunities were historically visible along the Rhine valley inGermany33. After the wave of expulsions in Western Europeduring the fifteenth century, they began to disperse once more,into Eastern Europe33.

These analyses suggest that the first major wave of assimilationprobably took place in Mediterranean Europe, most likely in theItalian peninsula B2 ka, with substantial further assimilation ofminor founders in west/central Europe. There is less evidence forassimilation in Eastern Europe, and almost none for a source inthe North Caucasus/Chuvashia, as would be predicted by theKhazar hypothesis8,9—rather, the results show strong geneticcontinuities between west and east European Ashkenazicommunities10, albeit with gradual clines of frequency offounders between east and west1,2 (Supplementary Note 2).

There is surprisingly little evidence for any significant founderevent from the Near East. Fewer than 10% of the AshkenazimtDNAs can be assigned to a Near Eastern source with anyconfidence, and these are found at very low frequencies (Fig. 2).The most frequent, belonging to HV1b2, R0a1a and U7, arefound at only B3, 2 and 1% respectively. All are widespreadacross Ashkenazi communities, and might conceivably be relictsof early Levantine founders, but it seems likely that other moreminor Near Eastern lineages are the result of more recent geneflow into the Ashkenazim.

The age estimates for the European founders might suggest(very tentatively, given the imprecision with present data) thatthese ancestral Jewish populations harboring haplogroup K andespecially N1b2 may have had an origin in the first millenniumBCE, rather than in the wake of the destruction of the JerusalemTemple in 70 CE. In fact, some scholars have argued fromhistorical evidence that the large-scale expansion of Judaismthroughout the Mediterranean in the Hellenistic period wasprimarily the result of proselytism and mass-conversion,especially amongst women9. We anticipate that a combinationof large-scale mitogenome and whole Y-chromosome analysis,complementing full human genome sequencing, will be able toaddress this question in much finer detail in the near future.

Despite the potential of genomic studies, the particular value offull-mitogenome sequencing should be stressed, as some studiesdismissed the value of uniparental markers because of the impactof drift in the Ashkenazim6. In fact, the reverse may be the case:autosomal studies may be confounded by drift whereas the finegenealogical resolution of full mitogenomes, given sufficientsampling, can provide a detailed reconstruction of the history ofAshkenazi women. The mtDNA genealogy may even beconsidered to have particular relevance due to the matrilinealinheritance found in Judaism since at least B200 CE and possiblyseveral centuries earlier, helping to ‘fix’ incoming lineages fromconverts within the Ashkenazi community after this time. Withsufficient resolution, a detailed genealogical history for everymaternal lineage in the Ashkenazim is now within reach; in fact,it should soon be possible to reconstruct the outlines of the entiredispersal history of each community.

MethodsSamples and analysis of mtDNA sequence variation. Although there is agrowing database of whole mitogenomes, almost all those from haplogroup U8 arefrom Europeans or individuals of European (predominantly west European)ancestry. Yet evidence from the Near East is critical in drawing up a meaningfulpicture of European (and wider west Eurasian) demographic prehistory. We

therefore selected 67 predominantly Near Eastern haplogroup K samples (identi-fied by full control-region sequencing of 111 haplogroup K samples) for mito-genome sequencing, plus five samples belonging to non-K U8 and two from Italypotentially belonging to N1b2 (Supplementary Data 1). We collected samples withthe appropriate informed consent of the subjects and the work was approved by theUniversity of Leeds, Faculty of Biological Sciences Ethics Committee, the EthicsCommittee for Clinical Experimentation at the University of Pavia, and theWestern Institution Review Board (WIRB), Olympia, WA, USA. We sequencedthem using Sanger sequencing30,34 and, to maximize the number of samples, weperformed a phylogenetic analysis alongside 884 published U8 sequences (a total of909 belonging to haplogroup K) (Supplementary Data 1) and four haplogroup Noutgroup sequences, using Network 4.6 software and the reduced-median algo-rithm35. We then constructed a putative most-parsimonious tree of the 956 U8sequences by hand from the network, following PhyloTree36 for known subclades.We used mtDNA-GeneSyn37 to convert files. As there are a number of extremelyvariable sites in K1 (positions 195 and 16,093 in particular), we confirmed theoverall topology by running networks of coding-region data only. We performedsimilar analyses for haplogroups H, J and T, and for N1b we augmented ourpreviously published tree24.

Age estimates and phylogeographic distribution. We estimated coalescencetimes of clades, using the r statistic and ML38,39, with Bayesian estimations formitogenomes using BEAST40. For the r statistic and ML, we corrected forpurifying selection using the calculator we developed previously (SupplementaryData 4)38. We defined some sub-haplogroups to be a priori monophyletic in theanalysis (U8, U8a, U8b, K, K1, K1a, K1b, K1c, K2, K2a and K2b) and assumed ageneration time of 25 years41. We also obtained Bayesian skyline plots42–44 toestimate ‘haplogroup-effective’ population sizes associated with U8 over time, andestimated the period of maximum growth39.

For a broader overview of the diversity and geographic distribution of lineages,we also compiled 1,917 haplogroup K HVS-I (hypervariable sequence I) sequences(in the range 16,051–16,400), 87 from U8a and 52 from U8b1 (from Europe, theNear East and North Africa, from a total database of 33,127 HVS-I sequences)(Supplementary Tables S1 and S2). We displayed frequency and diversitydistributions of haplogroups K, U8a1 and U8b1 sequences, identified from theirmotifs in the HVS-I database, on interpolation maps using Surfer. For thefrequency analyses, we analysed the data at the level of published regionalpopulations; for the diversity analyses we aggregated them into broader areas, asdescribed in Supplementary Table S2. For the analyses of other Ashkenazi lineageswe compared 836 published control-region sequences1,2,11 with availableAshkenazi whole mitogenomes and the global mitogenome database available onGenBank, in order to assign the Ashkenazi control-region lineages to subclades.For geographic distributions, we supplemented and checked this informationagainst a database of control-region data (38,244 records from west Eurasia,Central Asia and North Africa).

References1. Behar, D. M. et al. Differential bottleneck effects in the mtDNA gene pool of

Ashkenazi Jewish populations. Eur. J. Hum. Genet. 12, 355–364 (2004).2. Behar, D. M. et al. The matrilineal ancestry of Ashkenazi Jewry: portrait of a

recent founder event. Am. J. Hum. Genet. 78, 487–497 (2006).3. Behar, D. M. et al. The genome-wide structure of the Jewish people. Nature

466, 238–242 (2010).4. Atzmon, G. et al. Abraham’s children in the genome era: major Jewish diaspora

populations comprise distinct genetic clusters with shared Middle Easternancestry. Am. J. Hum. Genet. 86, 850–859 (2010).

5. Ostrer, H. A genetic profile of contemporary Jewish populations. Nat. Rev.Genet. 2, 891–898 (2001).

6. Zoossmann-Diskin, A. The origin of Eastern European Jews revealed by autosomal,sex chromosomal and mtDNA polymorphisms. Biol. Direct 5, 57 (2010).

7. Bray, S. M. et al. Signatures of founder effects, admixture, and selection in theAshkenazi Jewish population. Proc. Natl Acad. Sci. USA 107, 16222–16227(2010).

8. Koestler, A. The Thirteenth Tribe: The Khazar Empire and its Heritage(Random House, 1976).

9. Sand, S. The Invention of the Jewish People (Verso, 2009).10. Guha, S. et al. Implications for health and disease in the genetic signature of the

Ashkenazi Jewish population. Genome Biol. 13, R2 (2012).11. Thomas, M. et al. Founding mothers of Jewish communities: geographically

separated Jewish groups were independently founded by very few femaleancestors. Am. J. Hum. Genet. 70, 1411–1420 (2002).

12. Entine, J. Abraham’s Children: Race, Identity, and the DNA of the Chosen People(Grand Central Publishing, 2007).

13. Goldstein, D. B. Jacob’s Legacy (Yake University Press, 2008).14. Ostrer, H. Legacy: A Genetic History of the Jewish People (Oxford University

Press, 2012).15. Irshai, O. in The Illustrated History of the Jewish People. (ed De Lange, N.)

(Aurum Press, 1997).16. Elhaik, E. The missing link of Jewish European ancestry: contrasting the

Rhineland and the Khazarian hypotheses. Genome Biol. Evol. 5, 61–74 (2012).

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543 ARTICLE

NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications 9

& 2013 Macmillan Publishers Limited. All rights reserved.

17. Venton, D. Out of Khazaria—evidence for ‘Jewish genome’ lacking. GenomeBiol. Evol. 5, 75–76 (2013).

18. Hammer, M. F. et al. Jewish and Middle Eastern non-Jewish populations sharea common pool of Y-chromosome biallelic haplotypes. Proc. Natl Acad. Sci.USA 97, 6769–6774 (2000).

19. Nebel, A. et al. High-resolution Y chromosome haplotypes of Israeli andPalestinian Arabs reveal geographic substructure and substantial overlap withhaplotypes of Jews. Hum. Genet. 107, 630–641 (2000).

20. Behar, D. M. et al. Contrasting patterns of Y chromosome variation inAshkenazi Jewish and host non-Jewish European populations. Hum. Genet.114, 354–365 (2004).

21. Hammer, M. F. et al. Extended Y chromosome haplotypes resolve multiple andunique lineages of the Jewish priesthood. Hum. Genet. 126, 707–717 (2009).

22. Behar, D. M. et al. Multiple origins of Ashkenazi Levites: Y chromosomeevidence for both Near Eastern and European ancestries. Am. J. Hum. Genet.73, 768–779 (2003).

23. Richards, M. et al. Tracing European founder lineages in the Near EasternmtDNA pool. Am. J. Hum. Genet. 67, 1251–1276 (2000).

24. Fernandes, V. et al. The Arabian cradle: Mitochondrial relicts of the first stepsalong the southern route out of Africa. Am. J. Hum. Genet. 90, 347–355 (2012).

25. Behar, D. M. et al. Counting the founders: The matrilineal genetic ancestry ofthe Jewish Diaspora. PLoS One 3, e2062 (2008).

26. Pichler, I. et al. Drawing the history of the Hutterite population on a geneticlandscape, inference from Y-chromosome and mtDNA genotypes. Eur. J. Hum.Genet. 18, 463–470 (2010).

27. Campbell, C. L. et al. North African Jewish and non-Jewish populations formdistinctive, orthogonal clusters. Proc. Natl Acad. Sci. USA 109, 13865–13870(2012).

28. Soares, P. et al. The archaeogenetics of Europe. Curr. Biol. 20, R174–R183(2010).

29. Brotherton, P. et al. Neolithic mitochondrial haplogroup H genomes and thegenetic origins of Europeans. Nat. Commun. 4, 1764 (2013).

30. Pala, M. et al. Mitochondrial DNA signals of Late Glacial re-colonization ofEurope from Near Eastern refugia. Am. J. Hum. Genet. 90, 915–924 (2012).

31. Olivieri, A. et al. Mitogenomes from two uncommon haplogroups mark LateGlacial/postglacial expansions from the Near East and Neolithic dispersalswithin Europe. PLoS One 8, e70492 (2013).

32. Olivieri, A. et al. The mtDNA legacy of the Levantine Early Upper Palaeolithicin Africa. Science 314, 1767–1770 (2006).

33. Limor, O. in The Illustrated History of the Jewish People. (ed De Lange, N.)(Aurum Press, 1997).

34. Torroni, A. et al. Do the four clades of the mtDNA haplogroup L2 evolve atdifferent rates? Am. J. Hum. Genet. 69, 1348–1356 (2001).

35. Bandelt, H.-J., Forster, P., Sykes, B. C. & Richards, M. B. Mitochondrial portraitsof human populations using median networks. Genetics 141, 743–753 (1995).

36. van Oven, M. & Kayser, M. Updated comprehensive phylogenetic tree of globalhuman mitochondrial DNA variation. Hum. Mutat. 30, E386–E394 (2009).

37. Pereira, L. et al. The diversity present in 5,140 human mitochondrial genomes.Am. J. Hum. Genet. 84, 628–640 (2009).

38. Soares, P. et al. Correcting for purifying selection: an improved humanmitochondrial molecular clock. Am. J. Hum. Genet. 84, 740–759 (2009).

39. Soares, P. et al. The expansion of mtDNA haplogroup L3 within and out ofAfrica. Mol. Biol. Evol. 29, 915–927 (2012).

40. Pereira, L. et al. Population expansion in the North African Late Pleistocenesignalled by mitochondrial DNA haplogroup U6. BMC Evol. Biol. 10, 390(2010).

41. Fenner, J. N. Cross-cultural estimation of the human generation interval for usein genetics-based population divergence studies. Am. J. Phys. Anthropol. 128,415–423 (2005).

42. Drummond, A. J., Rambaut, A., Shapiro, B. & Pybus, O. G. Bayesian coalescentinference of past population dynamics from molecular sequences. Mol. Biol.Evol. 22, 1185–1192 (2005).

43. Atkinson, Q. D., Gray, R. D. & Drummond, A. J. Bayesian coalescent inferenceof major human mtDNA haplogroup expansions in Africa. Proc. R. Soc. Lond.Ser. B Biol. Sci 276, 367–373 (2009).

44. Atkinson, Q. D., Gray, R. D. & Drummond, A. J. mtDNA variation predictspopulation size in humans and Southern Asian chapter in human prehistory.Mol. Biol. Evol. 25, 468–474 (2008).

AcknowledgementsWe thank Doron Behar for discussions and suggestions, and Pierre-Marie Danze,Mukaddes Golge, Anne Cambon-Thomsen, CEPH, Steve Jones, Ariella Oppenheim,Gheorghe Stefanescu, Mark Thomas and the donors themselves for generously providingDNA samples. FCT, the Portuguese Foundation for Science and Technology, supportedthis work through the research project PTDC/CS–ANT/113832/2009 and the personalgrants to M.D.C. (SFRH/BD/48372/2008), J.B.P. (SFRH/BD/45657/2008), V.F. (SFRH/BD/61342/2009) and P.S. (SFRH/BPD/64233/2009). We also received support from theItalian Ministry of Education, University and Research: Progetti Futuro in Ricerca 2008(RBFR08U07M) and 2012 (RBFR126B8I) (to A.O. and A.A.) and Progetti RicercaInteresse Nazionale 2009 and 2012 (to A.A.), the Sorenson Molecular GenealogyFoundation (to U.A.P. and S.R.W.), the Leverhulme Trust (research project grant 10 105/D) (to MBR) and the DeLaszlo Foundation (to M.B.R./P.S.). IPATIMUP is an AssociateLaboratory of the Portuguese Ministry of Science, Technology and Higher Education andis partially supported by FCT.

Author contributionsM.B.R., L.P. and P.S. devised and supervised the project, M.D.C., J.B.P., M.C. and A.O.carried out the laboratory work, M.D.C., J.B.P., M.P., V.F., P.S., L.P. and M.B.R. carriedout the data analyses, M.D.C., J.B.P., P.S., L.P. and M.B.R. wrote the text, M.P., A.O.,A.A., U.A.P., S.R., ON., J.H., S.R.W., K.K.E., M.C. and V.M. discussed the results andhelped to revise the text.

Additional informationData access: Sequence data have been deposited in GenBank nucleotide core databaseunder accession numbers JX273243 to JX273297, KC878709 to KC878725 and KF297808to KF297809.

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Competing financial interests: The authors declare no competing financial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

How to cite this article: Costa, M.D. et al. A substantial prehistoric European ancestryamongst Ashkenazi maternal lineages. Nat. Commun. 4:2543 doi: 10.1038/ncomms3543(2013).

This article is licensed under a Creative Commons Attribution 3.0Unported Licence. To view a copy of this licence visit http://

creativecommons.org/licenses/by/3.0/.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3543

10 NATURE COMMUNICATIONS | 4:2543 | DOI: 10.1038/ncomms3543 | www.nature.com/naturecommunications

& 2013 Macmillan Publishers Limited. All rights reserved.