Initial Upper Palaeolithic humans in Europe had recent Neanderthal ancestry

Abstract

Modern humans appeared in Europe by at least 45,000 years ago1–5, but the extent of their interactions with Neanderthals, who disappeared by about 40,000 years ago6, and their relationship to the broader expansion of modern humans outside Africa are poorly understood. Here we present genome-wide data from three individuals dated to between 45,930 and 42,580 years ago from Bacho Kiro Cave, Bulgaria1,2. They are the earliest Late Pleistocene modern humans known to have been recovered in Europe so far, and were found in association with an Initial Upper Palaeolithic artefact assemblage. Unlike two previously studied individuals of similar ages from Romania7 and Siberia8 who did not contribute detectably to later populations, these individuals are more closely related to present-day and ancient populations in East Asia and the Americas than to later west Eurasian populations. This indicates that they belonged to a modern human migration into Europe that was not previously known from the genetic record, and provides evidence that there was at least some continuity between the earliest modern humans in Europe and later people in Eurasia. Moreover, we find that all three individuals had Neanderthal ancestors a few generations back in their family history, confirming that the first European modern humans mixed with Neanderthals and suggesting that such mixing could have been common. Genome-wide data for the three oldest known modern human remains in Europe, dated to around 45,000 years ago, shed light on early human migrations in Europe and suggest that mixing with Neanderthals was more common than is often assumed.

Full text

Nature | Vol 592 | 8 April 2021 | 253
Article
Initial Upper Palaeolithic humans in Europe
had recent Neanderthal ancestry
Mateja Hajdinjak1,2 ✉, Fabrizio Mafessoni1, Laurits Skov1, Benjamin Vernot1,
Alexander Hübner1,3, Qiaomei Fu4, Elena Essel1, Sarah Nagel1, Birgit Nickel1, Julia Richter1,
Oana Teodora Moldovan5,6, Silviu Constantin7, 8, Elena Endarova9, Nikolay Zahariev10,
Rosen Spasov10, Frido Welker11,12, Geoff M. Smith11, Virginie Sinet-Mathiot11, Lindsey Paskulin13,
Helen Fewlass11, Sahra Talamo11,14, Zeljko Rezek11,15 , Svoboda Sirakova16, Nikolay Sirakov16,
Shannon P. McPherron11, Tsenka Tsanova11, Jean-Jacques Hublin11,17, Benjamin M. Peter1,
Matthias Meyer1, Pontus Skoglund2, Janet Kelso1 & Svante Pääbo1 ✉
Modern humans appeared in Europe by at least 45,000 years ago1–5, but the extent of
their interactions with Neanderthals, who disappeared by about 40,000 years ago6,
and their relationship to the broader expansion of modern humans outside Africa are
poorly understood. Here we present genome-wide data from three individuals dated
to between 45,930 and 42,580 years ago from Bacho Kiro Cave, Bulgaria1,2. They are
the earliest Late Pleistocene modern humans known to have been recovered in Europe
so far, and were found in association with an Initial Upper Palaeolithic artefact
assemblage. Unlike two previously studied individuals of similar ages from Romania7
and Siberia8 who did not contribute detectably to later populations, these individuals
are more closely related to present-day and ancient populations in East Asia and the
Americas than to later west Eurasian populations. This indicates that they belonged to
a modern human migration into Europe that was not previously known from the
genetic record, and provides evidence that there was at least some continuity
between the earliest modern humans in Europe and later people in Eurasia. Moreover,
we nd that all three individuals had Neanderthal ancestors a few generations back in
their family history, conrming that the rst European modern humans mixed with
Neanderthals and suggesting that such mixing could have been common.
The transition between the Middle and Upper Palaeolithic periods in
Europe, which started about 47,000 years before present (47kyr )
1,2
,
overlapped with the spread of modern humans and the disappear-
ance of Neanderthals, which occurred by about 40kyr 
6
. Analyses
of the genomes of Neanderthals and modern humans have shown that
gene flow occurred between the two hominin groups approximately
60–50kyr 
8–11
, probably in southwestern Asia. However, owing to the
scarcity of modern human remains from Eurasia that are older than
40kyr
1–5,12
, genome-wide data are available for only three individuals
of this age
7,8,13
(Fig.1). Little is therefore known about the genetics of
the earliest modern humans in Eurasia, the extent to which they inter-
acted with archaic humans and their contribution to later populations.
For example, whereas the roughly 42,000 to 37,000-year-old ‘Oase1’
individual from Romania
7,14
and the roughly 45,000-year-old ‘Ust’Ishim’
individual from Siberia8 do not show specific genetic relationships to
subsequent Eurasian populations, the approximately 40,000-year-old
‘Tianyuan’ individual from China contributed to the genetic ancestry
of ancient and present-day East Asian populations
13
. Another open
question is the extent to which modern humans mixed with Neander-
thals when they spread across Europe and Asia. Direct evidence of local
interbreeding exists only for the Oase1 individual, who had a recent
Neanderthal ancestor7 in his family history.
Here, we analyse genome-wide data from human specimens found in
direct association with an Initial Upper Palaeolithic (IUP) assemblage of
artefacts in Bacho Kiro Cave, Bulgaria
1
(Fig.1), as well as from two more
recent specimens from the same site (Supplementary Information1).
The IUP groups together assemblages that fall chronologically between
the last Middle Palaeolithic assemblages and the first bladelet industries
of the Upper Palaeolithic. The IUP spans a broad geographical area
15
,
from southwest Asia, central and eastern Europe to Mongolia16 (Fig.1,
https://doi.org/10.1038/s41586-021-03335-3
Received: 7 July 2020
Accepted: 5 February 2021
Published online: 7 April 2021
Open access
Check for updates
1Department of Evolutionary Genetics, Max Planck Institute for Evolutionary Anthropology, Leipzig, Germany. 2Francis Crick Institute, London, UK. 3Department of Archaeogenetics, Max
Planck Institute for the Science of Human History, Jena, Germany. 4Key Laboratory of Vertebrate Evolution and Human Origins of Chinese Academy of Sciences, IVPP, Center for Excellence
in Life and Paleoenvironment, Beijing, China. 5Emil Racovita Institute of Speleology, Cluj Department, Cluj-Napoca, Romania. 6Romanian Institute of Science and Technology, Cluj-Napoca,
Romania. 7Department of Geospeleology and Paleontology, Emil Racovita Institute of Speleology, Bucharest, Romania. 8Centro Nacional de Investigación sobre la Evolución Humana,
CENIEH, Burgos, Spain. 9National History Museum, Soia, Bulgaria. 10Archaeology Department, New Bulgarian University, Soia, Bulgaria. 11Department of Human Evolution, Max Planck
Institute for Evolutionary Anthropology, Leipzig, Germany. 12Section for Evolutionary Genomics, GLOBE Institute, University of Copenhagen, Copenhagen, Denmark. 13Department of
Archaeology, University of Aberdeen, Aberdeen, UK. 14Department of Chemistry ‘G. Ciamician’, University of Bologna, Bologna, Italy. 15University of Pennsylvania Museum of Archaeology
and Anthropology, University of Pennsylvania, Philadelphia, PA, USA. 16National Institute of Archaeology with Museum, Bulgarian Academy of Sciences, Soia, Bulgaria. 17Chaire de
Paléoanthropologie, Collège de France, Paris, France. ✉e-mail: mateja_hajdinjak@eva.mpg.de; paabo@eva.mpg.de
Content courtesy of Springer Nature, terms of use apply. Rights reserved

254 | Nature | Vol 592 | 8 April 2021
Article
Supplementary Information1). Although there are reasons to group
these assemblages on the basis of their lithic technology, the IUP also
shows great regional variability. Therefore, it is debated whether the
IUP represents a dispersal of modern humans across middle-latitude
Eurasia, the diffusion of certain technological ideas, instances of inde-
pendent invention, or a combination of some or all of these15. The IUP
is contemporaneous with late Neanderthal sites in central and western
Europe
6
and precedes later Upper Palaeolithic techno-complexes in
Europe, such as the Protoaurignacian and the Aurignacian, by several
thousand years5.
Five human specimens were recovered from Bacho Kiro Cave in recent
excavations. They consist of a lower molar (F6-620) found in the upper
part of Layer J in the Main Sector, and four bone fragments (AA7-738,
BB7-240, CC7-2289 and CC7-335) from Layer I in Niche 1. They have been
directly radiocarbon-dated to between 45,930 and 42,580 calibrated
years before present (cal. )
1,2
, and their mitochondrial genomes are
of the modern human type, suggesting that they are the oldest Upper
Palaeolithic modern humans that have been recovered in Europe
1
. One
bone fragment was found in Layer B in the Main Sector (F6-597) and
another one was among the finds from excavations in the 1970s, when
it was retrieved in a position corresponding to the interface of Layers
B and C (BK1653). The two latter bone fragments were directly dated
to 36,320–35,600 cal.  and 35,290–34,610 cal. 1,2, respectively.
Although the lithic assemblages from the later layers are sparse, they
are likely to be Aurignacian
1,2
. We also produced additional data from a
mandible7,14 that was found outside any archaeological context in Petera
cu Oase, Romania (referred to as ‘Oase1’)
14
. The mandible was directly
dated to about 42–37kyr 
14
, although this may be an underestimate
as the dating was performed before recent technical improvements.
We extracted DNA from between 29.3mg and 64.7mg of tooth or
bone powder from the specimens as described1. We also treated 15mg of
bone powder from the Oase1 mandible with 0.5% hypochlorite solution
to reduce bacterial and human contamination before DNA extraction
17
.
Among DNA fragments sequenced from the DNA libraries constructed
from the Bacho Kiro Cave and Oase1 extracts, between 0.003% and 1.8%
could be mapped to the human genome (Supplementary Informa-
tion2). Owing to the low fraction of hominin DNA, we used in-solution
hybridization capture
18
to enrich the libraries for about 3.8 million
single-nucleotide polymorphisms (SNPs) that are informative about
modern human variation and archaic admixture7,19 (excluding F6-597,
which contained very little if any endogenous DNA; Supplementary
Information2).
For the six specimens, between 57,293 and 3,272,827 of the tar-
geted SNPs were covered by at least one DNA fragment (Extended
Data Table1). Of these, between 11,655 and 2,290,237 SNPs were cov-
ered by at least one fragment showing C-to-T substitutions in the first
three and/or the last three positions from the ends, suggesting the
presence of deaminated cytosine bases, which are typical of ancient
DNA
20
(Extended Data Table1, Extended Data Fig.1). On the basis of
the numbers of putatively deaminated fragments aligning to the X
chromosome and the autosomes
21
(Supplementary Information4),
we conclude that specimens F6-620, AA7-738, BB7-240 and CC7-335
belonged to males, whereas BK1653 and CC7-2289 belonged to females,
although the low amount of data makes this conclusion tentative for
CC7-2289 (Extended Data Fig.2a).
Using an approach that makes use of DNA deamination patterns
22
,
we estimated that the overall nuclear DNA contamination was between
2.2%±0.5% (F6-620) and 42.4%±0.6% (CC7-2289). In the male speci-
mens, we estimated contamination from polymorphisms on the X
chromosome
23
to between 1.6%±0.1% and 3.4%±0.5% (Supplementary
Information2). Owing to the presence of present-day human contami
-
nation, we restricted all downstream analyses to putatively deaminated
fragments for all specimens except F6-620 (for which contamination
was so low that we used all fragments). This left between 11,655 and
3,272,827 SNPs per specimen to be used for the subsequent analyses
(Supplementary Information2).
The molar F6-620 and the bone fragment AA7-738 have identical
mitochondrial genome sequences
1
and both come from males. The
pairwise mismatch rate between the two specimens at the SNPs
24
is 0.13,
similar to the mismatch rate between libraries from the same specimen
(Extended Data Fig.2b). By contrast, this number is 0.23 (interquartile
range: 0.22–0.25) for the other Bacho Kiro Cave specimens, similar to
unrelated ancient individuals from other studies (Extended Data Fig.2b).
Thus, we conclude that specimens F6-620 and AA7-738 belonged to the
same individual or to identical twins, which is much less likely.
We enriched the libraries from the male individuals using probes
that targeted about 6.9Mb of the Y chromosome
25
(Supplementary
Information4) and arrived at 15.2-fold coverage for F6-620, 2.5-fold for
BB7-240 and 1.5-fold for CC7-335. F6-620 carries a basal lineage of the
Y chromosome haplogroup F (F-M89), whereas BB7-240 and CC7-335
carry haplogroup C1 (C-F3393). Although haplogroup C is common
among males from East Asia and Oceania, both haplogroups F and C1
are rare in present-day humans and are found only at low frequencies
in mainland Southeast Asia and Japan26,27.
19 22
25,26
30–32 27
24
20
33–35 36
23
21
28
29
38
39
40
41
42
43
44
45,46
47
48–50
1
2
3
Bacho Kiro Cave
Pes¸ tera cu Oase 5,6
8
7
911
10
12 14,15
13
37
4
Ust’Ishim
Tianyuan
1, Bacho Kiro Cave; 2, Ust’Ishim; 3, Pes¸ tera cu Oase; 4, Tianyuan Cave; 5, 6, Kostenki14 (Markina Gora) and Kostenki12 (Vokovskaya);
7, Troisème Caverne of Goyet; 8, Sunghir; 9, Pes¸ tera Muierilor; 10, Grotta Paglicci; 11, Pes¸ tera Cioclovina Uscat ˘a; 12, Krems Wachtberg;
13, Yana RHS; 14, 15, Dolní V ˇestonice and Pavlov; 16, Grotta del Cavallo; 17, Kents Cavern; 18, Grotta di Fumane; 19, Brno-Bohunice;
20, Stánska Skála III; 21, Temnata; 22, Kulychivka; 23, Korolevo 1 and 2; 24, Shlyakh; 25, 26, Üçagizli and Kanal Cave; 27, Um el’Tlel; 28, Jerf Ajlah;
29, Yabrud II; 30–32, Antelias; Abou Halka and Ksar Akil; 33–35, Emireh, El Wad and Raqefet; 36, Boker Tachtit; 37, Denisova Cave;
38, Kara-Bom; 39, Ust-Karakol 1; 40, Kara-Tenesh; 41, Makarvo IV; 42, Kamenka A–C; 43, Khotyk; 44, Podzvonkaya; 45, 46, Tolbor4 and To lbor16;
47, Tsangan-Agui; 48–50, Suindonggou1, 2 and 9
17
16
18
Modern humans ~30–39 kyr BP
Modern humans >40 kyr BP
Initial Upper Palaeolithic
Without archaeological context
With EUP bladelet technology (Protoaurignacian)
Fig. 1 | Arch aeologic al sites that h ave yielded
genetic data and/or IUP assemblages. Sites
with mode rn human genome -wide data olde r
than 40kyr (red circle s) or older than 30kyr
(yellow circle s), sites in Europe with mo dern
human remain s older than 40ky r (red square s)
and sites w ith IUP assem blages (black s quares).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Nature | Vol 592 | 8 April 2021 | 255
We estimated the extent of genetic similarity among the Bacho Kiro
Cave individuals and other early modern humans using outgroup
f
3
-statistics
28
. The three roughly 45,000-year-old IUP individuals are
more similar to one another than to any other ancient individual
(Extended Data Fig.3a). By contrast, BK1653, which is about 35,000
yearsold, is more similar to later Upper Palaeolithic individuals from
Europe who are around 38,000years old or younger
29,30
(3.0 ≤|Z|≤17.4;
Extended Data Fig.4, Supplementary Information5); for example, to
the roughly 35,000-year-old ‘GoyetQ116-1’ individual from Belgium
and members of the ‘Věstonice’ genetic cluster, who are associated
with later Gravettian assemblages29 (Extended Data Figs.3a, 4b, c).
When comparing the Bacho Kiro Cave individuals to present-day
populations
31
, we found that the IUP individuals share more alleles
(that is, more genetic variants) with present-day populations from
East Asia, Central Asia and the Americas than with populations from
western Eurasia (Fig.2a, Supplementary Information5), whereas the
later BK1653 individual shares more alleles with present-day western
Eurasian populations (Extended Data Figs.3b, 4a).
We next investigated whether these observations could be due to the
fact that present-day populations in western Eurasia derive part of their
ancestry from ‘Basal Eurasians’
32,33
, an inferred population that diverged
early from other non-African populations and may have ‘diluted’ allele
sharing between western Eurasian populations and IUP individuals.
To do this, we compared the Ust’Ishim, Oase1 and IUP Bacho Kiro Cave
individuals to western Eurasian individuals such as the approximately
38,000-year-old ‘Kostenki14’ individual from Russia
29,30
, which pre-dates
the introduction of ‘Basal Eurasian’ ancestry to Europe around 8,000cal.

32
. We found that the Ust’Ishim and Oase1 individuals showed no more
affinity to western than to eastern Eurasian populations, suggesting
that they did not contribute ancestry to later Eurasian populations,
as previously shown
7,8
(Supplementary Information5, Extended Data
Fig.5). By contrast, the IUP Bacho Kiro Cave individuals shared more
alleles with the roughly 40,000-year-old Tianyuan individual13 from
China (Fig.2b) and other ancient Siberians34,35 and Native Americans36–39
(Fig.2c) than with the Kostenki14 individual (3.6≤|Z|≤5.3). Among other
western Eurasian Upper Palaeolithic humans, the IUP Bacho Kiro Cave
individuals shared more alleles with the Oase1 (3.6≤|Z|≤4.3) and roughly
35,000-year-old GoyetQ116-1
29
individuals than with the Kostenki14 indi-
vidual (3.2≤|Z|≤4.3; Fig.2c, Supplementary Information5). Notably, the
GoyetQ116-1 individual has previously been shown to share more alleles
with early East Asians than other individuals of a similar age in Europe
13
.
When we explored models of population history that are compatible
with the observations above using admixture graphs28, we found that
the IUP Bacho Kiro Cave individuals were related to populations that
contributed ancestry to the Tianyuan individual in China as well as, to
a lesser extent, to the GoyetQ116-1 and Ust’Ishim individuals (all |Z|<3;
Fig.2d, Supplementary Information6). This resolves the previously
unclear relationship between the GoyetQ116-1 and Tianyuan individu-
als13 without the need for gene flow between these two geographi-
cally distant individuals. The models also suggest that the later BK1653
b
c
a
Lovelock1
Saqqaq
LBK
Motala12
Kennewick
Loschbour
Karelia
Kotias
Kolyma
SpiritCave
USR
Natuan
Bichon
Satsurblia
Anzick
Villabruna
ElMiron
Malta1
Vestonice16
Yana old
Sunghir3
BK1653
GoyetQ116-1
Oase1
Ust’Ishim
BK CC7-335
BK BB7-240
BK F6-620
D(Tianyuan, Kostenki14; X, Mbuti) D(X, Kostenki14; BachoKiro F6-620, Mbuti)
f3(Mbuti; IUP Bacho Kiro, X)
−0.12
−0.09
−0.06
−0.03
0
0.03
0.06
D
Individual
Lovelock1
Saqqaq
LBK
Motala12
Kennewick
Loschbour
Karelia
Kotias
Kolyma
SpiritCave
USR
Natuan
Bichon
Satsurblia
Anzick
Villabruna
ElMiron
Malta1
Vestonice16
Yana old
Sunghir3
BK1653
GoyetQ116-1
Tianyuan
Oase1
Ust’Ishim
BK CC7-335
BK BB7-240
−0.03
0
0.03
0.06
0.09
0.12
D
d
0.05 0.15 0.25
f3
America Central Asia/Siberia
East Asia
Near East West Eurasia
|Z| < 3 |Z| ≥ 3
10 10
51%
81
43
7
Mbuti
7
926
96%
49%0
Vindija33.19
22
37
Denisova
37
4%
2%
8
042
3%
64
6
17
IUP Bacho Kiro
158
98%22
17% 39%
19%
83%
53 11
5
25
BK1653
148
Ust’Ishim
3
61% 21
Tianyuan
143 3
5
GoyetQ116-1
153
44%
81% 9
Kostenki14
150
97%
41
Sunghir3
10
Vestonice16
131
56%
Fig. 2 | Popul ation aff initie s of the IUP Bach o Kiro Cave indivi duals. a, Allele
sharing ( f3) bet ween the IUP Ba cho Kiro Cave indivi duals and prese nt-day
populations (X) from the S imons Genome D iversity Proj ect (SGDP)31 afte r their
separati on from an outgro up (Mbuti) (calculate d as f3(Mbuti; IUP Bac ho Kiro,
X). Warmer colour s on the map48 correspond to higher f3 values (hig her shared
genetic dr ift). b, IUP Bacho Kiro C ave individuals sha re signific antly more
alleles (proportions of allele sharing or D values plo tted on xaxis) with th e
roughly 40,000-year-old Tianyuan individual13 than with the approximately
38,000-year-old Kostenki14 individual29,30. Calculated as D(Tianyuan,
Kostenki14; X, Mbuti). c, F6-620 sha res signif icantly more alle les with the
Oase17 and Goye tQ116-129 indiv iduals, ancien t Siberians an d Native Ameri can
individua ls than with the Kos tenki14 individ ual. Calculate d as D(X, Kostenki14;
F6-62 0, Mbuti). b, c, Filled c ircles indicat e a signific ant value (|Z|≥3); open
circles, |Z|<3. Whiskers cor respond to 1 s.e . calculated a cross all autoso mes
(1,813, 821 SNPs) using a weig hted block jac kknife28 and a block s ize of 5Mb. BK,
Bacho Kiro. d, Ad mixture graph re lating Bacho K iro Cave individuals a nd
ancient hu mans older than 30k yr. This model u ses 281,732 overl apping SNPs
in all individ uals and fit s the data with a si ngle outlier (Z= 3.2 2).Anc ient
non-Afric ans (yellow circle s), Vindija 33.19 Neande rthal (orange), Denisova n
(grey) and pre sent-day African i ndividuals (li ght yellow circle) are show n.
Admixture e dges (dotted line s) show the genetic com ponent relate d to
Neander thals (red), to the IUP Ba cho Kiro Cave indiv iduals (orange) and to
BK1653 ( green). Number s on solid branche s correspon d to the estimate d drift
in f2 units of squar ed frequency d ifference; lab els on dotted e dges give
admixture proportions .
Content courtesy of Springer Nature, terms of use apply. Rights reserved

256 | Nature | Vol 592 | 8 April 2021
Article
individual belonged to a population that was related, but not identical,
to that of the GoyetQ116-1 individual (Fig.2d, Extended Data Fig.4,
Supplementary Information6) and that the Věstonice cluster, whose
members were found in association with Gravettian assemblages
29
,
derived part of their ancestry from such a population and the rest from
populations related to the roughly 34,000-year-old ‘Sunghir’ individu-
als40 from Russia (Fig.2d, Supplementary Information6).
As the IUP Bacho Kiro Cave individuals lived at the same time as some
of the last Neanderthals in Europe
6
, we estimated the proportion of Nean
-
derthal DNA in their genomes by taking advantage of two high-quality
Neanderthal genomes9,10,41. We found that the IUP individuals F6-620, BB7-
240 and CC7-335 carried 3.8% (95% confidence interval (CI): 3.3–4.4%),
3.0% (95% CI: 2.4–3.6%) and 3.4% (95% CI: 2.8–4.0%) Neanderthal DNA,
respectively. This is more than the average of 1.9% (95% CI: 1.5–2.4%) found
in other ancient or present-day humans, except for the Oase1 individual,
who had a close Neanderthal relative (6.4% (95% CI: 5.7–7.1%); Extended
Data Fig.6, Supplementary Information7). By contrast, the more recent
BK1653 individual carried only 1.9% (95% CI: 1.4–2.4%) Neanderthal DNA,
similar to other ancient and present-day humans
10,41
(Extended Data
Fig.6). As has been the case for all humans studied so far, the Neanderthal
DNA in BK1653 and the IUP Bacho Kiro Cave individuals was more similar
to the Vindija33.19
10
and Chagyrskaya8
42
Neanderthals than to the Altai
Neanderthal9 (2.8≤|Z|≤5.1; Supplementary Information7).
To study the spatial distribution of Neanderthal ancestry in the
genomes of the Bacho Kiro Cave individuals, we used around 1.7 mil-
lion SNPs at which Neanderthal
9
and/or Denisovan
43
genomes differ
from African genomes
7
and an approach
44
that detects tracts of archaic
DNA in ancient genomes. We found a total of 279.6centiMorgans (cM)
of Neanderthal DNA in F6-620, 251.6cM in CC7-335 and 220.9cM in BB7-
240, and these individuals carried seven, six and nine Neanderthal DNA
segments longer than 5cM, respectively (Fig.3, Extended Data Fig.7a,
Supplementary Information8). The longest introgressed Neanderthal
segment in F6-620 encompassed 54.3cM, and the longest segments
in CC7-335 and BB7-240 were 25.6cM and 17.4cM, respectively (Fig.3,
Extended Data Fig.7a). By contrast, the total amount of Neanderthal
DNA in the BK1653 genome was 121.7cM and the longest Neanderthal
segment was 2.5cM (Fig.3, Extended Data Fig.7a).
On the basis of the distribution of the long Neanderthal segments,
we estimate that individual F6-620 had a Neanderthal ancestor less
than six generations back in his family tree (Extended Data Table2,
Supplementary Information8). Both the CC7-335 and BB7-240 individu-
als had Neanderthal ancestors about seven generations back in their
families, with upper confidence intervals of ten and seventeen genera-
tions, respectively (Extended Data Table2, Extended Data Fig.7b, Sup-
plementary Information8). Thus, all IUP Bacho Kiro Cave individuals
had recent Neanderthal ancestors in their immediate family histories.
To further explore the extent to which the Bacho Kiro Cave individuals
contributed ancestry to later populations in Eurasia, we investigated
whether the Neanderthal DNA segments in Bacho Kiro Cave genomes
overlapped with Neanderthal segments in present-day populations
more than expected by chance. We found more overlapping of segments
between present-day East Asian populations and the IUP Bacho Kiro Cave
individuals (lowest correlation coefficient of 0.09, 95% CI: 0.08–0.1) than
with the BK1653 individual (P=0.02, Wilcoxon test). By contrast, the
BK1653 individual shows more overlapping of Neanderthal segments
with present-day western Eurasian populations (a correlation coefficient
of 0.11, 95% CI: 0.1–0.12) than do the IUP Bacho Kiro Cave individuals
(P<1×10−18, Wilcoxon test). This is compatible with the observation that
the IUP Bacho Kiro Cave population contributed more ancestry to later
populations in Asia and the Americas, whereas the BK1653 individual
contributed more ancestry to populations in western Eurasia.
We next looked for overlap between parts of the human genome
that carry little or no Neanderthal ancestry (Neanderthal ‘deserts’),
which are thought to have been caused by purifying selection against
050100 150200 250
Bacho Kiro F6-620 (43,930–42,580 cal. BP)*
050100 150200 250
BachoKiro BB7-240 (45,550–43,940 cal. BP)
050100 150 200250
Bacho Kiro CC7-335 (45,930–44,420 cal. BP)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Chromosome
050 100150 200250
279.59 cM
220.86 cM
251.63 cM
496.56 cM
Oase1 (41,860–36,950 cal. BP)
Position (Mb)
Chromosome
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Chromosome
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Chromosome
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Position (Mb)
Position (Mb)
Position (Mb)
Alleles matching modern human Alleles matching Neanderthal
Ust’Ishim
Bacho Kiro Cave
ab
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Chromosome
050100 150200 250
122.47 cM
Bacho Kiro BK 1653 (35,270–34,610 cal. BP)
Position (Mb)
Pes¸tera cu Oase
Fig. 3 | Geo graphical di stributi on of Neander thal archae ological s ites and
genome -wide distr ibution of Ne anderthal a lleles in the g enomes of an cient
modern humans. a, Neandert hal geographic al range (blue) and th e locations
of Petera c u Oase, Bacho K iro Cave and where the fem ur of the Ust’I shim
individua l was found. b, Distribution of Neanderthal DNA in ancient modern
human genom es. Neander thal DNA seg ments longe r than 0.2cM are indica ted
in blue. Pie c harts indic ate the total prop ortion of Nea nderthal DN A identifie d
in each genom e. Centrome res are shown in blac k.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Nature | Vol 592 | 8 April 2021 | 257
Neanderthal DNA shortly after introgression
45,46
. We find almost no
introgressed Neanderthal DNA in the previously described deserts in
the IUP Bacho Kiro Cave and Oase1 individuals (249kb out of 898Mb
of introgressed sequence; P=0.0079, permutation P value). When we
restricted these comparisons to the more recent Neanderthal contri-
butions (that is, segments longer than 5cM), we similarly found no
overlap (0Mb out of 415Mb, P=0.15, permutation P value), suggesting
that selection against Neanderthal DNA variants occurred within a few
generations, although additional individuals with recent Neanderthal
ancestry will be needed to fully resolve this question.
In conclusion, the Bacho Kiro Cave genomes show that several distinct
modern human populations existed during the early Upper Palaeolithic
in Eurasia. Some of these populations, represented by the Oase1 and
Ust’Ishim individuals, show no detectable affinities to later popula-
tions, whereas groups related to the IUP Bacho Kiro Cave individuals
contributed to later populations with Asian ancestry as well as some
western Eurasian humans such as the GoyetQ116-1 individual in Belgium.
This is consistent with the fact that IUP archaeological assemblages are
found from central and eastern Europe to present-day Mongolia
5,15,16
(Fig.1), and a putative IUP dispersal that reached from eastern Europe
to East Asia. Eventually populations related to the IUP Bacho Kiro Cave
individuals disappeared in western Eurasia without leaving a detectable
genetic contribution to later populations, as indicated by the fact that
later individuals, including BK1653 at Bacho Kiro Cave, were closer to
present-day Europeanpopulations than to present-day Asianpopula-
tions
29,30
. In Europe, the notion of successive population replacements is
also consistent with the archaeological record, where the IUP is clearly
intrusive against the Middle Palaeolithic background and where, apart
from the common focus on blades, there are no clear technological con-
nections between the IUP and the subsequent Aurignacian technologies.
Finally, it is striking that all four of the European individuals who over-
lapped in time with late Neanderthals
7
and from whom genome-wide
data have been retrieved had close Neanderthal relatives in their family
histories (Fig.3, Extended Data Figs.7, 8). This suggests that mixing
between Neanderthals and the first modern humans that arrived into
Europe was perhaps more common than is often assumed.
Note added in proof: A companion paper47 describes an individual
from the Czechia who—based on genetic analyses—may be of similar
age to the IUP Bacho Kiro Cave individuals and who carries a proportion
of Neanderthal ancestry similar to later Upper Palaeolithic humans.
Online content
Any methods, additional references, Nature Research reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author con-
tributions and competing interests; and statements of data and code
availability are available at https://doi.org/10.1038/s41586-021-03335-3.
1. Hublin, J.-J. etal. Initial Upper Palaeolithic Homo sapiens from Bacho Kiro Cave, Bulgaria.
Nature 581, 299–302 (2020).
2. Fewlass, H. etal. A 14C chronology for the Middle to Upper Palaeolithic transition at Bacho
Kiro Cave, Bulgaria. Nat. Ecol. Evol. 4, 794–801 (2020).
3. Higham, T. etal. The earliest evidence for anatomically modern humans in northwestern
Europe. Nature 479, 521–524 (2011).
4. Benazzi, S. etal. Early dispersal of modern humans in Europe and implications for
Neanderthal behaviour. Nature 479, 525–528 (2011).
5. Hublin, J.-J. The modern human colonization of western Eurasia: when and where? Quat.
Sci. Rev. 118, 194–210 (2015).
6. Higham, T. etal. The timing and spatiotemporal patterning of Neanderthal
disappearance. Nature 512, 306–309 (2014).
7. Fu, Q. etal. An early modern human from Romania with a recent Neanderthal ancestor.
Nature 524, 216–219 (2015).
8. Fu, Q. etal. Genome sequence of a 45,000-year-old modern human from western
Siberia. Nature 514, 445–449 (2014).
9. Prüfer, K. etal. The complete genome sequence of a Neanderthal from the Altai
Mountains. Nature 505, 43–49 (2014).
10. Prüfer, K. etal. A high-coverage Neandertal genome from Vindija Cave in Croatia. Science
358, 655–658 (2017).
11. Green, R. E. etal. A draft sequence of the Neandertal genome. Science 328, 710–722 ( 2010).
12. Benazzi, S. etal. The makers of the Protoaurignacian and implications for Neandertal
extinction. Science 348, 793–796 (2015).
13. Yang, M. A. etal. 40,000-year-old individual from Asia provides insight into early
population structure in Eurasia. Curr. Biol. 27, 3202–3208.e9 (2017).
14. Trinkaus, E. etal. An early modern human from the Petera cu Oase, Romania. Proc. Natl
Acad. Sci. USA 100, 11231–11236 (2003).
15. Kuhn, S. L. & Zwyns, N. Rethinking the initial Upper Paleolithic. Quat. Int. 347, 29–38 (2014).
16. Zwyns, N. etal. The northern route for human dispersal in central and northeast Asia: new
evidence from the site of Tolbor-16, Mongolia. Sci. Rep. 9, 11759 (2019).
17. Korlević, P. etal. Reducing microbial and human contamination in DNA extractions from
ancient bones and teeth. Biotechniques 59, 87–93 (2015).
18. Fu, Q. etal. DNA analysis of an early modern human from Tianyuan Cave, China. Proc.
Natl Acad. Sci. USA 110, 2223–2227 (2013).
19. Haak, W. etal. Massive migration from the steppe was a source for Indo-European
languages in Europe. Nature 522, 207–211 (2015).
20. Briggs, A. W. etal. Patterns of damage in genomic DNA sequences from a Neandertal.
Proc. Natl Acad. Sci. USA 104, 14616–14621 (2007).
21. Meyer, M. etal. Nuclear DNA sequences from the Middle Pleistocene Sima de los Huesos
hominins. Nature 531, 504–507 (2016).
22. Peyrégne, S. & Peter, B. M. AuthentiCT: a model of ancient DNA damage to estimate the
proportion of present-day DNA contamination. Genome Biol. 21, 246 (2020).
23. Korneliussen, T. S., Albrechtsen, A. & Nielsen, R. ANGSD: analysis of next generation
sequencing data. BMC Bioinformatics 15, 356 (2014).
24. Mittnik, A. etal. Kinship-based social inequality in Bronze Age Europe. Science 366,
731–734 (2019).
25. Petr, M. etal. The evolutionary history of Neanderthal and Denisovan Y chromosomes.
Science 369, 1653–1656 (2020).
26. Poznik, G. D. etal. Punctuated bursts in human male demography inferred from 1,244
worldwide Y-chromosome sequences. Nat. Genet. 48, 593–599 (2016).
27. Kutanan, W. etal. Contrasting paternal and maternal genetic histories of Thai and Lao
populations. Mol. Biol. Evol. 36, 1490–1506 (2019).
28. Patterson, N. etal. Ancient admixture in human history. Genetics 192, 1065–1093 (2012).
29. Fu, Q. etal. The genetic history of Ice Age Europe. Nature 534, 200–205 (2016).
30. Seguin-Orlando, A. etal. Genomic structure in Europeans dating back at least 36,200
years. Science 346, 1113–1118 (2014).
31. Mallick, S. etal. The Simons Genome Diversity Project: 300 genomes from 142 diverse
populations. Nature 538, 201–206 (2016).
32. Lazaridis, I. etal. Ancient human genomes suggest three ancestral populations for
present-day Europeans. Nature 513, 409–413 (2014).
33. Lazaridis, I. etal. Genomic insights into the origin of farming in the ancient Near East.
Nature 536, 419–424 (2016).
34. Sikora, M. etal. The population history of northeastern Siberia since the Pleistocene.
Nature 570, 182–188 (2019).
35. Rasmussen, M. etal. Ancient human genome sequence of an extinct Palaeo-Eskimo.
Nature 463, 757–762 (2010).
36. Moreno-Mayar, J. V. etal. Terminal Pleistocene Alaskan genome reveals irst founding
population of Native Americans. Nature 553, 203–207 (2018).
37. Moreno-Mayar, J. V. etal. Early human dispersals within the Americas. Science 362,
eaav2621 (2018).
38. Rasmussen, M. etal. The genome of a Late Pleistocene human from a Clovis burial site in
western Montana. Nature 506, 225–229 (2014).
39. Rasmussen, M. etal. The ancestry and afiliations of Kennewick Man. Nature 523,
455–458 (2015).
40. Sikora, M. etal. Ancient genomes show social and reproductive behavior of early Upper
Paleolithic foragers. Science 358, 659–662 (2017).
41. Petr, M., Pääbo, S., Kelso, J. & Vernot, B. Limits of long-term selection against Neandertal
introgression. Proc. Natl Acad. Sci. USA 116, 1639–1644 (2019).
42. Mafessoni, F. etal. A high-coverage Neandertal genome from Chagyrskaya Cave. Proc.
Natl Acad. Sci. USA 117, 15132–15136 (2020).
43. Meyer, M. etal. A high-coverage genome sequence from an archaic Denisovan individual.
Science 338, 222–226 (2012).
44. Peter, B. M. 100,000 years of gene low between Neandertals and Denisovans in the
Altai mountains. Preprint at https://doi.org/10.1101/2020.03.13.990523 (2020).
45. Vernot, B. etal. Excavating Neandertal and Denisovan DNA from the genomes of
Melanesian individuals. Science 352, 235–239 (2016).
46. Sankararaman, S., Mallick, S., Patterson, N. & Reich, D. The combined landscape of
Denisovan and Neanderthal ancestry in present-day humans. Curr. Biol. 26, 1241–1247 (2016).
47. Prüfer, K. etal. A genome sequence from a modern human skull over 45,000 years old from
Zlatý kůň in Czechia. Nat. Ecol. Evol.https://doi.org/10.1038/s41559-021-01443-x (2021).
48. R Core Team. R: A Language and Environment for Statistical Computing
http://www.R-project.org/ (R Foundation for Statistical Computing, 2013).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional afiliations.
Open Access This article is licensed under a Creative Commons Attribution
4.0 International License, which permits use, sharing, adaptation, distribution
and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons license, and indicate
if changes were made. The images or other third party material in this article are included in
the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your
intended use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a copy of this license,
visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2021
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Methods
Ethics declaration
All approvals for specimen handling have been obtained from the
relevant institutions. For the Oase1 specimen, the permission was
granted to S.P. by the Emil Racovita Institute of Speleology, as the
national authority in caves study. For the Bacho Kiro Cave specimens,
the permission was granted by the Bulgarian Ministry of Culture and
the National Museum of Natural History (Sofia, Bulgaria).
DNA extraction and library preparation
Data generation for the seven Bacho Kiro Cave specimens (specimen
IDs: F6-620, AA7-738, BB7-240, CC7-2289, CC7-335, F6-597 and BK1653)
was based on DNA libraries prepared and described previously
1
. To
obtain additional data from the Oase1 individual, we extracted DNA
from 15 mg of bone powder from the specimen
7,14
. As it was previously
found to be highly contaminated with microbial and present-day
human DNA, the bone powder was treated with 0.5% hypochlorite
solution before DNA extraction17. Four single-stranded DNA libraries
were prepared from the resulting extract and two additional librar-
ies were prepared, each using 5 µl of the two DNA extracts generated
previously
7
as described
49
. The pools of libraries were then sequenced
directly on Illumina MiSeq and HiSeq 2500 platforms in a double index
configuration (2×76 cycles)
50
and base calling was done using Bustard
(Illumina).
DNA captures
We enriched the selected amplified libraries for about 3.7 million SNPs
across the genome described in supplementary data 2of ref.
19
(SNP
Panel 1 or 390k array), and supplementary data 1–3of ref. 7 (SNP Pan-
els 2, 3 and 4, or 840k, 1000k and Archaic admixture arrays, respec-
tively). For the male individuals (Bacho Kiro Cave F6-620, BB7-240
and CC7-335), an aliquot of each library was additionally enriched for
about 6.9 Mb of the Y chromosome
25
. All of the enriched libraries were
sequenced on the Illumina HiSeq 2500 platforms in a double index
configuration (2×76 cycles)
50
and base calling was done using Bustard
(Illumina).
Sequencing of capture products and data processing
For all sequencing runs we trimmed the adapters and merged overlap
-
ping forward and reverse reads into single sequences using leeHom51
(version: https://bioinf.eva.mpg.de/leehom/). The Burrows-Wheeler
Aligner
52
(BWA, version: 0.5.10-evan.9-1-g44db244; https://github.
com/mpieva/network-aware-bwa) with the parameters adjusted for
ancient DNA (“-n 0.01 –o 2 –l 16500”)43 was used to align the data from
all sequencing runs to the human reference genome (GRCh37/1000
Genomes release; ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/techni-
cal/reference/phase2_reference_assembly_sequence/). Only reads
that showed perfect matches to the expected index combinations
were used for all downstream analyses. PCR duplicates were removed
using bam-rmdup (version: 0.6.3; https://github.com/mpieva/
biohazard-tools) and SAMtools (version: 1.3.1)53 was used to filter for
fragments that were at least 35 bp long and that had a mapping quality
equal to or greater than 25. BAM files of the libraries enriched for the
specific subset of the nuclear genome were further intersected with the
BED files containing target SNP positions (390k, 840k, 1000k, Archaic
admixture, a merged set of SNP Panels 1 and 2 or 1240k, and a merged
set of SNP Panels 1, 2 and 3 or 2200k) and regions (Y chromosome)
using BEDtools
54
(version: 2.24.0). In order to filter for endogenous
ancient DNA or putatively deaminated fragments, we used elevated
C-to-T substitutions relative to the reference genome at the first three
and/or last three positions of the alignment ends20. We merged librar-
ies originating from the same specimen using samtools merge
53
to
produce the final datasets for downstream analyses (Extended Data
Table1, Supplementary Information2).
Merging of the Bacho Kiro Cave and Oase1 data with other
genomes
We performed random read sampling using bam-caller (https://github.
com/bodkan/bam-caller, version: 0.1) by picking a base with a base
quality of at least 30 at each position in the 1240k and 2200k SNP Pan-
els that was covered by at least one fragment longer than 35 bp with a
mapping quality equal to or higher than 25 (L≥35bp, MQ≥25, BQ≥30).
To mitigate the effect of deamination-derived substitutions on down-
stream analyses, we did not sample any Ts on the forward strands (in
the orientation as sequenced) or any As on the reverse strands in the
first five and/or last five positions from the alignment ends. Owing to
the haploid nature of the Y chromosome, we called genotypes across
the approximately 6.9 Mb of the Y chromosome for the enriched librar-
ies of male individuals by calling a consensus allele at each position by
majority call requiring a minimum coverage of 3 for specimens F6-620
and BB7-240 and of 2 for specimen CC7-335 using using bam-caller
(https://github.com/bodkan/bam-caller, version: 0.1) (Supplementary
Information2).
We merged the data from the newly sequenced specimens with
datasets of previously published ancient and present-day humans,
as well as archaic humans, for three SNP panels (1240k, 2240k and
Archaic admixture; Supplementary Information3). Data from
the 1240k panel include genotypes of 2,109 ancient and 2,974
present-day individuals compiled from published studies and avail-
able in the EIGENSTRAT format28 at https://reich.hms.harvard.
edu/allen-ancient-dna- resource-aadr-downloadable-genotypes-
present-day-and-ancient-dna-data (version 37.2, released 22 February
2019). Data from the 2240k panel include published genetic data of
ancient modern humans obtained through hybridization captures
7,13,29
and a range of present-day9,31 and ancient modern humans8,30,32–37,55–60,
as well as the archaics
9,10,42,43,61
, for which whole-genome shotgun data
of varying coverage are available (Supplementary Information3). The
Archaic admixture panel data include 21 ancient modern humans directly
enriched for these sites7,13,29, as well as the genotypes of present-day9,31
and ancient modern humans
8,30,32–37,55–60
, as well as the archaics
9,10,42,43,61
,
for which whole-genome shotgun data are available (Supplementary
Information3) and that were intersected with about 1.7 million SNPs of
the Archaic admixture panel using BEDTools54 (version: 2.24.0).
Population genetic analyses
To determine the relationship of the Bacho Kiro Cave and Oase1 individ-
uals to other modern and archaic humans we used a range of f-statistics
from ADMIXTOOLS
28
(version: v5.1) and as implemented in the R pack-
age admixr
62
(version: 0.7.1; Supplementar y Information4). We used
qpGraph program (Admixture Graph) from ADMIXTOOLS28 (version:
v5.1) to test models of the relationship among Initial Upper Palaeolithic
Bacho Kiro Cave individuals, the roughly 35,000-year-old Bacho Kiro
Cave individual BK1653 and other ancient modern humans from Eurasia
older than 30,000 years  (Fig.2d, Supplementary Information6).
Neanderthal ancestry
We estimated the proportion of Neanderthal DNA in the genomes
of present-day and ancient modern humans by computing a direct
f4 ratio28 that takes advantage of the two high-quality Neanderthal
genomes
9,10,41
(Extended Data Fig.6, Supplementary Information7). We
used admixfrog
44
(version: 0.5.6, https://github.com/BenjaminPeter/
admixfrog/) to detect archaic introgressed segments in the genomes
of the Bacho Kiro Cave and Oase1 individuals, as well as in other ancient
modern humans and 254 present-day non-African individuals from the
SGDP
31
as a direct comparison (Supplementary Information8). We used
these introgressed segments to estimate the number of generations
since the most recent Neanderthal ancestor of the IUP Bacho Kiro Cave
and Oase1 individuals (Supplementary Information8), to investigate
the overlap of Neanderthal segments in Bacho Kiro Cave individuals
Content courtesy of Springer Nature, terms of use apply. Rights reserved

with those detected in present-day and ancient modern humans (Sup-
plementary Information9), and to investigate the overlap of Nean-
derthal segments in the IUP Bacho Kiro Cave and Oase1 individuals
with parts of the human genome devoid of Neanderthal ancestry
45,46
(Neanderthal deserts; Supplementary Information10).
Reporting summary
Further information on research design is available in theNature
Research Reporting Summary linked to this paper.
Data availability
The aligned sequences of the Bacho Kiro Cave and Oase1 individuals have
been deposited in the European Nucleotide Archive under accession
number PRJEB39134. Comparative data of present-day human genomes
from the SGDP that were used in this study are available at https://www.
simonsfoundation.org/simons-genome-diversity-project/. Compara-
tive data used in this study, which include genotypes of 2,109 ancient
and 2,974 present-day individuals compiled from published studies,
are available in the EIGENSTRAT file format at https://reich.hms.har-
vard.edu/allen-ancient-dna-resource- aadr-downloadable-genotypes-
present-day-and-ancient-dna-data (version 37.2, released 22 February
2019). To determine the Y chromosome haplogroups of male individu-
als in this study, we used the Y-haplogroup tree from the International
Society of Genetic Genealogy (ISOGG, available at http://www.isogg.
org, version: 13.38).
49. Gansauge, M.-T., Aximu-Petri, A., Nagel, S. & Meyer, M. Manual and automated
preparation of single-stranded DNA libraries for the sequencing of DNA from ancient
biological remains and other sources of highly degraded DNA. Nat. Protocols 15,
2279–2300 (2020).
50. Kircher, M., Sawyer, S. & Meyer, M. Double indexing overcomes inaccuracies in multiplex
sequencing on the Illumina platform. Nucleic Acids Res. 40, e3 (2012).
51. Renaud, G., Stenzel, U. & Kelso, J. leeHom: adaptor trimming and merging for Illumina
sequencing reads. Nucleic Acids Res. 42, e141 (2014).
52. Li, H. & Durbin, R. Fast and accurate long-read alignment with Burrows-Wheeler
transform. Bioinformatics 26, 589–595 (2010).
53. Li, H. etal. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25,
2078–2079 (2009).
54. Quinlan, A. R. & Hall, I. M. BEDTools: a lexible suite of utilities for comparing genomic
features. Bioinformatics 26, 841–842 (2010).
55. Raghavan, M. etal. Upper Palaeolithic Siberian genome reveals dual ancestry of Native
Americans. Nature 505, 87–91 (2014).
56. Olalde, I. etal. Derived immune and ancestral pigmentation alleles in a 7,000-year-old
Mesolithic European. Nature 507, 225–228 (2014).
57. Gallego Llorente, M. etal. Ancient Ethiopian genome reveals extensive Eurasian
admixture throughout the African continent. Science 350, 820–822 (2015).
58. Keller, A. etal. New insights into the Tyrolean Iceman’s origin and phenotype as inferred
by whole-genome sequencing. Nat. Commun. 3, 698 (2012).
59. Gamba, C. etal. Genome lux and stasis in a ive millennium transect of European
prehistory. Nat. Commun. 5, 5257 (2014).
60. Jones, E. R. etal. Upper Palaeolithic genomes reveal deep roots of modern Eurasians.
Nat. Commun. 6, 8912 (2015).
61. Hajdinjak, M. etal. Reconstructing the genetic history of late Neanderthals. Nature 555,
652–656 (2018).
62. Petr, M., Vernot, B. & Kelso, J. admixr-R package for reproducible analyses using
ADMIXTOOLS. Bioinformatics 35, 3194–3195 (2019).
Acknowledgements We thank A. Weihmann and B. Schellbach for their help with DNA
sequencing; R. Barr, P. Korlević and S. Tüpke for help with graphics; D. Reich and M. Slatkin for
discussions and input; the Tourist Association STD “Bacho Kiro” in Dryanovo; the History
Museum in Dryanovo; the Regional History Museum in Gabrovo; the National Museum of
Natural History (NMNH) in Soia; and N. Spassov. M.H. is supported by a Marie Skłodowska
Curie Individual Fellowship (no. 844014). Q.F. was supported by the Strategic Priority Research
Program (B) (XDB26000000) of CAS, NSFC (41925009, 41630102, 41672021). O.T.M. and S.C.
were supported by a grant from the Ministry of Research and Innovation, CNCS - UEFISCDI,
project number PN-III-P4-ID-PCCF-2016-0016, within PNCDI III and the EEA Grants 2014-2021,
under Project contract no. 3/2019. F.W. received funding from the European Research Council
(ERC) under the European Union’s Horizon 2020 research and innovation programme (grant
agreement no. 948365). P.S. was supported by the Vallee Foundation, the European Research
Council (grant no. 852558), the Wellcome Trust (217223/Z/19/Z) and Francis Crick Institute core
funding (FC001595) from Cancer Research UK, the UK Medical Research Council and the
Wellcome Trust. This study was funded by the Max Planck Society and the European Research
Council (grant agreement no. 694707 to S.P.).
Author contributions M.H., E. Essel, S.N., B.N., J.R. and Q.F. performed ancient DNA lab work.
O.T.M., S.C., E. Endarova, N.Z., R.S., F.W., G.M.S., V.S.-M., H.F., S.T., Z.R., S.S., N.S., S.P.M., T.T. and
J.-J.H. provided and analysed archaeological material. M.H., F.M., L.S., B.V., A.H. and B.M.P.
analysed DNA data. B.M.P., M.M., P.S., J.K. and S.P. supervised the study. M.H., J.K. and S.P.
wrote the manuscript with input from all co-authors.
Funding Open access funding provided by Max Planck Society.
Competing interests The authors declare no competing interests.
Additional information
Supplementary information The online version contains supplementary material available at
https://doi.org/10.1038/s41586-021-03335-3.
Correspondence and requests for materials should be addressed to M.H. or S.P.
Peer review information Nature thanks Carles Lalueza-Fox, Marie Soressi and the other,
anonymous, reviewer(s) for their contribution to the peer review of this work. Reviewer reports
are available.
Reprints and permissions information is available at http://www.nature.com/reprints.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Extende d Data Fig. 1 | C-to -T substitution f requenci es at the beg inning and
end of nucle ar alignme nts for the merg ed librarie s of the Bacho K iro Cave
and Oase1 specimens. Only fragm ents of at leas t 35bp that mapped t o the
human referen ce genome with a m apping quality o f at least 25 (MQ≥25) we re
used for this a nalysis. Solid li nes depict all f ragments a nd dashed line s the
fragments that have a C-to-T substitution at the opposing end (conditional
substitutions).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Extende d Data Fig. 2 | Sex d etermina tion, pairw ise misma tch rate bet ween
specim ens and prin cipal compo nent analyse s (PCAs). a, Sex deter mination
for Bacho Kiro C ave specimens . Only fragme nts that showed C-to-T
substitu tions in the f irst three and/or la st three positi ons and overlapping
2200k Pa nel SNPs were use d for this analysis (for the nu mber of deamin ated
fragme nts per speci men, seeSupplem entary Table 2. 8). The expected r atios of
X to (X + autoso mal) fragmen ts for a female and a male in dividual are depi cted
as dashed li nes, and circle s correspond to t he calculated va lues for each of the
Bacho Kiro C ave specimens . Whiskers corre spond to 95% binom ial confide nce
intervals. b, Pairwise mi smatch rate bet ween differe nt libraries f rom the same
specime n (intra-specim en), between dif ferent Bacho Kiro C ave specimens
(inter-specimen) and between other ancient modern humansolder than
30,000 cal. . The boxplots were drawn u sing the summar y statisti cs geom_
stat from th e R-package ggpl ot; lower and upper hin ges, first a nd third
quartile s; whiskers, maxi mum value of 1.5 × the interquar tile range; centre li ne,
median. SN Ps across all auto somes of the 2 200k Panel were u sed for the
calculati ons (number of SNPs (ns nps)=2,056,573). c, A PCA of 2 ,970
present-day hum ans genotyp ed on 597,573 SNPs wit h 22 ancient ind ividuals
older than 30,0 00cal. projected on to the plane. d, A PC A of 1,444
present-day Eura sian and Native Am erican indiv iduals genoty ped on 597,573
SNPs with 2 2 ancient indi viduals older th an 30,000cal. projecte d onto the
plane. c, d, Grey dots de note present-day hum an genomes.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Extende d Data Fig. 3 | He atmaps of out group f3-statistics corresponding to
the amount o f shared gen etic drift b etween in dividuals a nd/or
populations. a, Genetic clu stering of anc ient individua ls, including the I UP
Bacho Kiro C ave, BK1653 and Oa se1 individua ls based on the am ount of shared
genetic dr ift and calcula ted as f3(ancient1, ancient2; Mbuti). Lighter colours in
this panel indicate higher f3 values and corresp ond to higher sha red genetic
drift (nsnp s=2,056,573). b, c, Shared gene tic drift bet ween the approxi mately
35,000 -year-old BK1653 (b; n snps=825,379) or approxim ately 38,000 -year-old
Kostenki14 individuals29,30 (c; nsnps=1,676,430) and present-day human
populatio ns from the SGDP 31 calculated a s f3(Bacho Kiro BK 1653/Kostenk i14,
present-day hum ans; Mbuti). Three M buti individu als from the same p anel31
were used as a n outgroup. Hig her f3 values47 are indi cated with wa rmer colours
and corres pond to higher sh ared genetic dr ift. Plotte d f3 values were calcu lated
using ADMIXTOOLS28 as implemented in admixr61. Coordinates for pr esent-day
humans were previously published31. The heatma p scale is consi stent with
those in Fig .2a, Supplementar y Figs. 5.1, 5. 2.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Extende d Data Fig. 4 | Pop ulation af finiti es of the approxi mately
35,000-year-old BK1653 individual. a, In contras t to the IUP Bacho K iro Cave
individua ls, individual BK 1653 is signi ficantly c loser to the approxi mately
38,000-year-old Kostenki14 individual29,30 than to present-day non-Af rican
populatio ns from Central A sia and Siber ia, East Asia, S outh Asia, Oce ania or the
Americ as, as calculat ed by D(Kostenki14, present-day humans; BachoKiro
BK1653, Mbuti). D values for each compari son, plotted a s barplots, wer e
calculate d using ADMI XTOOLS28 as implemented in admixr61. Prese nt-day
human genom es from the SGD P31 were used in the se statist ics, and three Mb uti
individua ls from the same pa nel were used as an o utgroup. **| Z|≥3, *|Z|≥2. b, c,
BK1653 sh ares signif icantly more all eles with the ap proximately 35,0 00-year-
old GoyetQ1 16-129 (b) and approxi mately 31,000 -year-old Vestonice1629
individuals (c) than with mos t other ancient m odern humans . D values
calculate d as in a. Filled circle s correspond t o |Z|≥3, and open circles indi cate a
|Z|<3 (not signif icant). Error bar s in all panels show s. e. calculated u sing a
weighted block jackknife28 across a ll autosomes on t he 2200k Pan el (nsnps
(Bacho Kiro B K1653)=825,379) and a blo ck size of 5Mb.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Extende d Data Fig. 5 | D(Ko sten ki14, X; IUP Ba cho Kiro Cave indi viduals/
Ust’Ishim/Oase1, Mbuti) where X is a present-day non-African population
from Cent ral Asia and Si beria, Ea st Asia, So uth Asia, Oc eania or the
Americas. D values for each comp arison, plot ted as barplot s, were calculat ed
using ADMIXTOOLS28 as implemented in admixr61. Present-day human
genomes f rom the SGDP31 wer e used in these s tatistic s, and three Mbut i
individua ls from the same pa nel were used as a n outgroup. ** |Z|≥3, *|Z|≥2. Error
bars denot e s.e. calculat ed using a weight ed block jack knife28 across all
autosome s on the 2200 k Panel and a block s ize of 5Mb. a, A pool of thre e IUP
Bacho Kiro C ave individuals (nsnp s=1,813,821). b, Ust ’Ishim (nsnps=1,951,462).
c, Oase1 (nsnps=402 ,526).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Extende d Data Fig. 6 | Ne anderthal a ncestry i n IUP Bacho Ki ro Cave
individuals. a, The propor tion of Neand erthal ance stry in Bach o Kiro Cave
individua ls and other ancie nt and present-day mo dern humans c alculated wit h
a direct f4 rati o that takes advan tage of the two hig h-coverage Nean derthal
genomes9,10,41. f4 ratio (alpha) values calculat ed using ADM IXTOOLS28 as
implemented in admixr61. b, c, Neanderthals9,10,42 share significantly more
derived alle les with the IU P Bacho Kiro Cave indi viduals than wi th most
present-day31 ( b) or ancient mode rn humans (c). D values calc ulated using
ADMIXTOOLS28 as implemented in admixr61. Filled c ircles corres pond to |Z|≥3;
open circle s indicate |Z|<3 (not sign ificant). Error ba rs in all panels show s .e.
calculated using a weighted block jackknife28 acros s all autosomes o n the
2200k Pa nel (nsnps=2,056, 573) and a block size of 5Mb.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Extended Data Fig. 7 | Segments of Neanderthal ancestry and estimates of
the number of generations since the most recent Neanderthal ancestor. a,
A combined p lot of the inferred N eandertha l segments i n the genomes of th e
IUP Bacho K iro Cave, BK1653 an d Oase1 indivi duals, including c hromosome X,
using a hidde n Markov model approac h (admixfrog)44. b, Maxim um likelihood
estimate s (dashed red line s) of the number of gener ations since a re cent
additiona l Neanderth al introgres sion into the IUP B acho Kiro Cave and Oas e1
individuals. Dashed black lines show 95% confid ence intervals.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Extende d Data Fig. 8 | Spa tial distr ibution of Ne anderthal D NA in the
Ust’Ishim, Tianyuan and Kostenki14 genomes. Segments corresponding to
Neander thal ancest ry inferred usi ng a hidden Markov mode l approach
(admixfrog)44 longer than 0.2cM are in dicated in blue . Centromere s are
indicate d in black. Pie ch arts indica te the total amou nt of Neander thal DNA
identif ied in each gen ome.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Article
Extended Data Table 1 | Amount of data generated for Bacho Kiro Cave and Oase1 libraries for each SNP panel
Table shows number of fragments after merging all of the sequencing libraries together for each specimen. Fragments longer than 35bp with a mapping quality of at least 25 that overlapped
different target sites were used for reporting the number of SNPs on target.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Extended Data Table 2 | Estimates of the number of generations before the most recent Neanderthal introgression in Bacho
Kiro Cave and Oase1 individuals (tracts >5cM) obtained by calculating the complementary cumulative distribution (CCD) of
the lengths of Neanderthal tracts
Estimates for tracts≤5cM and therefore representative of older introgression events are also shown.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Content courtesy of Springer Nature, terms of use apply. Rights reserved

Content courtesy of Springer Nature, terms of use apply. Rights reserved

Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com