|
|
Post by Admin on Feb 4, 2022 19:12:14 GMT
Reconstructing the spatiotemporal patterns of admixture during the European Holocene using a novel genomic dating method
Manjusha Chintalapati1#, Nick Patterson2#, Priya Moorjani1,3
12 Abstract
13
14 Recent studies have shown that gene flow or admixture has been pervasive throughout
15 human history. While several methods exist for dating admixture in contemporary
16 populations, they are not suitable for sparse, low coverage data available from ancient
17 specimens. To overcome this limitation, we developed DATES that leverages ancestry
18 covariance patterns across the genome of a single individual to infer the timing of admixture.
19 By performing simulations, we show that DATES provides reliable results under a range of
20 demographic scenarios and outperforms available methods for ancient DNA applications.
21 We apply DATES to ~1,100 ancient genomes to reconstruct gene flow events during the
22 European Holocene. Present-day Europeans derive ancestry from three distinct groups, local
23 Mesolithic hunter-gatherers, Anatolian farmers, and Yamnaya Steppe pastoralists. These
24 ancestral groups were themselves admixed. By studying the formation of Anatolian farmers,
25 we infer that the gene flow related to Iranian Neolithic farmers occurred no later than 9,600
26 BCE, predating agriculture in Anatolia. We estimate the early Steppe pastoralist groups
27 genetically formed more than a millennium before the start of steppe pastoralism, providing
28 new insights about the history of proto-Yamnaya cultures and the origin of Indo-European
29 languages. Using ancient genomes across sixteen regions in Europe, we provide a detailed
30 chronology of the Neolithization across Europe that occurred from ~6,400–4,300 BCE. This
31 movement was followed by a rapid spread of steppe ancestry from ~3,200–2,500 BCE. Our
32 analyses highlight the power of genomic dating methods to elucidate the legacy of human
33 migrations, providing insights complementary to archaeological and linguistic evidence.
34
35 Keywords
36 genomic clocks, admixture, ancient DNA, European Holocene, molecular clock, migration,
37 Neolithic, Bronze Age, Yamnaya
38 Significance
39
40 The European continent was subject to two major migrations during the Holocene: the
41 movement of Near Eastern farmers during the Neolithic and the migration of Steppe
42 pastoralists during the Bronze Age. To understand the timing and dynamics of these
43 movements, we developed DATES that leverages ancestry covariance patterns across the
44 genome of a single individual to infer the timing of admixture. Using ~1,100 ancient genomes
45 spanning ~8,000–350 BCE, we reconstruct the chronology of the formation of the ancestral
46 populations and the fine-scale details of the spread of Neolithic farming and Steppe
47 pastoralist-related ancestry to Europe. Our analysis demonstrates the power of genomic
48 dating methods to provide an independent and complementary timeline of population origins
49 and movements using genetic data
|
|
|
|
Post by Admin on Feb 4, 2022 20:28:56 GMT
50 Introduction
51
52 Recent studies have shown that population mixture (or “admixture”) is pervasive
53 throughout human history, including mixture between the ancestors of modern humans and archaic
54 hominins (i.e., Neanderthals and Denisovans), as well as in the history of many contemporary
55 human groups such as African Americans, South Asians and Europeans (1, 2). Many admixed
56 groups are formed due to population movements involving ancient migrations that pre-date
57 historical records. The recent availability of genomic data for a large number of present-day and
58 ancient genomes provides an unprecedented opportunity to reconstruct population events using
59 genetic data, providing evidence complementary to linguistics and archaeology. Understanding
60 the timing and signatures of admixture offers insights into the historical context in which the
61 mixture occurred and enables the characterization of the evolutionary and functional impact of the
62 gene flow.
63 To characterize patterns of admixture, genetic methods use the insight that the genome of
64 an admixed individual is a mosaic of chromosomal segments inherited from distinct ancestral
65 populations (3). Due to recombination, these ancestral segments get shuffled in each generation
66 and become smaller and smaller over time. The length of the segments is inversely proportional to
67 the time elapsed since the mixture (3, 4). Several genetic approaches––ROLLOFF (4), ALDER
68 (5), Globetrotter (2), and Tracts (6)–– have been developed that use this insight by characterizing
69 patterns of admixture linkage disequilibrium (LD) or haplotype lengths across the genome to infer
70 the timing of mixture. Haplotype-based methods perform chromosome painting or local ancestry
71 inference at each locus in the genome and characterize the distribution of ancestry tract lengths to
72 estimate the time of mixture (2, 6). This requires accurate phasing and inference of local ancestry,
73 which is often difficult when the admixture events are old (as ancestry blocks become smaller over
74 time) or when reference data from ancestral populations is unavailable. Admixture LD-based
75 methods, on the other hand, measure the extent of the allelic correlation across markers to infer
76 the time of admixture (4, 5). They do not require phased data from the target or reference
77 populations and work reliably for dating older admixture events (>100 generations). However,
78 they tend to be less efficient in characterizing admixture events between closely related ancestral
79 groups.
80 While highly accurate for dating admixture events using data from present-day samples,
81 current methods do not work reliably for dating admixture events using ancient genomes. Ancient
82 DNA samples often have high rates of DNA degradation, contamination (from human and other
83 sources) and low sequencing depth, leading to a large proportion of missing variants and uneven
84 coverage across the genome. Additionally, most studies generate pseudo-homozygous genotype
85 calls––consisting of a single allele call at each diploid site––that can lead to some issues in the
86 inference. In such sparse datasets, estimating admixture LD can be noisy and biased (see
87 Simulations below). Moreover, haplotype-based methods require phased data from both admixed
88 and reference populations which remains challenging for ancient DNA specimens.
89 An extension of admixture LD-based methods, recently introduced by Moorjani et al.
90 (2016), leverages ancestry covariance patterns that can be measured in a single sample using low
91 coverage data. This approach measures the allelic correlation across neighboring sites, but instead
92 of measuring admixture LD across multiple samples, it integrates data across markers within a
93 single diploid genome. Using a set of ascertained markers that are informative for Neanderthal
94 ancestry (where sub-Saharan Africans are fixed for the ancestral alleles and Neanderthals have a
95 derived allele), Moorjani et al. (2016) inferred the timing of Neanderthal gene flow in Upper
96 Paleolithic Eurasian samples and showed the approach works accurately in ancient DNA samples
97 (1). However, this approach is inapplicable for dating admixture events within modern human
98 populations, as there are very few fixed differences across populations (7).
99 Motivated by the single sample statistic in Moorjani et al. (2016), we developed DATES
100 (Distribution of Ancestry Tracts of Evolutionary Signals) that measures the ancestry covariance
101 across the genome in a single admixed individual, weighted by the allele frequency difference
102 between two ancestral populations. This method was first introduced in Narasimhan et al. (2019),
103 where it was used to infer the date of gene flow between groups related to Ancient Ancestral South
104 Indians, Iranian farmers, and Steppe pastoralists in ancient South and Central Asian populations
105 (8). In this study, we evaluate the performance of DATES by performing extensive simulations for
106 a range of demographic scenarios and compare the approach to other published genomic dating
107 methods. We then apply DATES to infer the chronology of the genetic formation of the ancestral
108 populations of Europeans and the spatiotemporal patterns of admixture during the European
109 Holocene using data from ~1,100 ancient DNA specimens spanning ~8,000–350 BCE.
|
|
|
|
Post by Admin on Feb 4, 2022 22:16:18 GMT
110 Results
111
112 Overview of DATES: Model and simulations
113
114 DATES estimates the time of admixture by measuring the weighted ancestry covariance
115 across the genome using data from a single diploid genome and two reference populations
116 (representing the ancestral source populations). DATES works like haplotype-based methods as it
117 is applicable to dating admixture in a single genome and not like admixture LD-based methods,
118 which by definition require multiple genomes to be co-analyzed; but unlike haplotype-based
119 methods, it is more flexible as it does not require local ancestry inference. There are three main
120 steps in DATES: we start by first learning the genome-wide ancestry proportions by performing a
121 simple regression analysis to model the observed genotypes in an admixed individual as a linear
122 mix of allele frequencies from the two reference populations. For each marker, we then compute
123 the likelihood of the observed genotype in the admixed individual using the estimated ancestry
124 proportions and allele frequencies in each reference population (this is similar in spirit to local
125 ancestry inference). This information is, in turn, used to compute the joint likelihood for two
126 neighboring markers to test if they derive ancestry from the same ancestral group, accounting for
127 the probability of recombination between the two markers. Finally, we compute the covariance
128 across pairs of markers located at a particular genetic distance, weighted by the allele frequency
129 differences in the reference populations (Note S1).
130 Following (1), we bin the markers that occur at a similar genetic distance across the
131 genome, rather than estimating admixture LD for each pair of markers, and compute the covariance
132 across increasing genetic distance between markers. The estimated covariance is expected to decay
133 exponentially with genetic distance, and the rate of decay is informative of the time of the mixture
134 (4). Assuming the gene flow occurred instantaneously, we infer the average date of gene flow by
135 fitting an exponential distribution to the decay pattern (Methods). In cases where data for multiple
136 individuals is available, we compute the likelihood by summing over all individuals. To make
137 DATES computationally tractable, we implemented the fast Fourier transform (FFT) for
calculating ancestry covariance as described in ALDER (5). This provides a speedup from 𝛰(𝑛
2 138 )
139 to 𝛰(𝑛 log 𝑛 ), which reduces the typical runtimes from hours to seconds (Note S1).
140 To assess the reliability of DATES, we performed simulations where we constructed ten
141 admixed diploid genomes by randomly sampling haplotypes from two source populations (Note
142 S2). Briefly, we simulated individual genomes with 20% European and 80% African ancestry by
143 using phased haplotypes of Northern Europeans (Utah European Americans, CEU) and west
144 Africans (Yoruba from Nigeria, YRI) from the 1000 Genomes Project respectively (7). As
145 reference populations in DATES, we used closely related surrogate populations of French and
146 Yoruba respectively, from the Human Genome Diversity Panel (9). We first investigated the
147 accuracy of DATES by varying the time of admixture between 10–300 generations. For
148 comparison, we also applied ALDER (5) to these simulations. Both methods reliably recovered
149 the time of admixture up to 200 generations or ~5,600 years ago, assuming a generation time of
150 28 years (1), though DATES was more precise than ALDER for older admixture events (>100
151 generations) (Table S2.4). Further, DATES shows accurate results even for single samples (Figure
152 1).
153 Next, we tested DATES for features such as varying admixture proportions and use of surrogate
154 populations as reference groups. By varying of European ancestry proportion between ~1–50%
155 (the rest derived from west Africans), we observed DATES accurately estimated the timing in all
156 cases (Figure S2.2A). However, the inferred admixture proportion was overestimated was lower
157 admixture proportions (<10%) (Figure S2.2B). Thus, we caution against using DATES for
158 estimating ancestry proportions and recommend other methods based on f-statistics (10). Using
159 reference populations which are divergent from true admixing source, we found that the inferred
160 dates were accurate even when we used Khomani San instead of Yoruba as the reference
161 population (FST ~ 0.1) (Figure S2.5). We also found that using the admixed samples themselves as
162 one of the reference populations also works reliably as ALDER (i.e., single reference setup) (5).
163 An important feature of DATES is that it does not require phased data and is applicable to
164 datasets with small sample sizes, making it in principle useful for ancient DNA applications. To
165 test the reliability of DATES for ancient genomes, we simulated data mimicking the relevant
166 features of ancient genomes, namely with large proportions of missing genotypes (between 10–
167 60%), and pseudo-homozygous genotype calls (instead of diploid genotype calls). DATES showed
168 reliable results in both cases, even only a single admixed individual was available (Figure S2.7).
169 In contrast, admixture LD-based methods require more than one sample and do not work reliably
170 with missing data. For example, ALDER estimates were very unstable for simulations with >40%
171 missing data. For older dates (>100 generations), we observed slight bias even with >10% missing
172 genotypes (Figure S2.17). As LD calculations leverage shared patterns across samples, variable
173 missingness of genotypes across individuals leads to substantial loss of data leading to unstable
174 and noisy inference. This highlights a major advantage of DATES for ancient DNA studies as it
175 provides reliable results even in sparse datasets (Note S2.5).
176 DATES assumes a model of instantaneous gene flow with a single pulse of mixture between
177 two source populations. However, many human populations have a history of multiple pulses of
178 gene flow. To test the performance of DATES for multi-way admixture events, we generated
179 admixed individuals with ancestry from three sources (East Asians, Africans, and Europeans)
180 where the gene flow occurred at two distinct time points (Note S2, Figure S2.10). By applying
181 DATES with pairs of reference populations at a time and fitting a single exponential to the ancestry
182 covariance patterns, we observed that DATES recovered both admixture times in case of equal
183 ancestry proportion from the three ancestral groups when the associated reference groups were
184 used for dating (Figure S2.11). In the case of unequal admixture proportions from three ancestral
185 groups, DATES inferred the timing of the recent admixture event in most cases, though some
186 confounding was observed, especially when the ancestry proportion of the recent event was low
187 (Figure S2.12). However, if the reference populations were set up to match the model of gene flow,
188 we observed that we could reliably recover the time of the recent gene flow event. For example,
189 there is limited confounding if the two references used in DATES include (i) the source population
190 for the recent event and (ii) either the pooled ancestral populations contributing to the first (or
191 earlier) event or the intermediate admixed group formed after the first event (Table S2.1). This
192 highlights how the choice of reference populations can help to tune the method to infer the timing
193 of specific admixture events reliably.
194 Finally, we explored the impact of more complex demographic events, including
195 continuous admixture and founder events using coalescent simulations (Note S2). In the case of
196 continuous admixture, DATES inferred an intermediate timing between the start and the end of the
197 gene flow period, similar to other methods like ALDER and Globetrotter (2, 5) (Table S2.2). In
198 the case of populations with founder events, we inferred unbiased dates of admixture in most cases
199 except when the founder event was extreme (Ne ~ 10) or the population had maintained a low
200 population size (Ne < 100) until present (i.e., no recovery bottleneck) (Figure S2.13, Table S2.3).
201 In humans, few populations have such extreme founder events, and thus, in most other cases, our
202 inferred admixture dates should be robust to founder events (11). We note that while DATES is not
203 a formal test of admixture, in simulations, we find that in the absence of gene flow, the method
204 does not infer significant dates of admixture even when the target has a complex demographic
205 history (Figure S2.15, S2.16).
207 Comparison to other methods
208
209 We assessed the reliability of DATES in real data by comparing our results with published
210 methods: Globetrotter, ALDER, and ROLLOFF. These methods are designed for the analysis of
211 present-day samples that typically have high-quality data with limited missing variants. In
212 addition, Globetrotter uses phased data which is challenging for ancient DNA samples. Thus,
213 instead of rerunning other methods, we took advantage of the published results for contemporary
214 samples presented in Hellenthal et al., (2014) (2). Following (2), we created a merged dataset
215 including individuals from Human Genome Diversity Panel (9), Behar et al. (2010) (12), and Henn
216 et al. (2012) (13) (Methods). We applied DATES and ALDER to 29 target groups using the
217 reference populations reported in Hellenthal et al. 2014 (Table S12), excluding one group where
218 the population label was unclear. Interestingly, the majority of these groups (25/29) failed
219 ALDER’s formal test of admixture; either because the results of the single reference and two
220 reference analyses yielded inconsistent estimates or because the target had long-range shared LD
221 with one of the reference populations (Table S4.1). Using DATES, we inferred significant dates of
222 admixture in 20 groups, and 14 of those were consistent with estimates based on Globetrotter. In
223 most remaining cases, recent studies suggest the target populations may have ancestry from
224 multiple gene flow events, either involving the same source populations or additional ancestral
225 groups . The estimated admixture timing based on DATES, ROLLOFF, and ALDER (assuming
226 two-way admixture regardless of the formal test results) were found to be highly concordant (Table
227 S4.1).
228
229 Fine-scale patterns of population mixtures in ancient Europe
230
231 Recent ancient DNA studies have shown that present-day Europeans derive ancestry from three
232 distinct sources: (a) hunter-gatherer-related ancestry that is closely related to Mesolithic hunter
233 gatherers (HG) from Europe; (b) Anatolian farmer-related ancestry related to Neolithic farmers
234 from the Near East and associated to the spread of farming to Europe; and (c) Steppe pastoralist
235 related ancestry that is related to the Yamnaya pastoralists from Russia and Ukraine (16–19). Many
236 open questions remain about the timing and dynamics of these population interactions, in particular
237 related to the formation of the ancestral groups (which were themselves admixed) and their
238 expansion across Europe. To characterize the spatial and temporal patterns of mixtures in Europe
239 in the past 10,000 years, we used 1,096 ancient European samples from 152 groups from the
240 publicly available Allen Ancient DNA Resource (AADR) spanning a time range of ~8,000–350
241 BCE (Methods, Table SA). Using DATES, we characterized the timing of the various gene flow
242 events, and below, we describe the key events in chronological order focusing on three main
243 periods.
|
|
|
|
Post by Admin on Feb 5, 2022 19:26:35 GMT
229 Fine-scale patterns of population mixtures in ancient Europe
230
231 Recent ancient DNA studies have shown that present-day Europeans derive ancestry from three
232 distinct sources: (a) hunter-gatherer-related ancestry that is closely related to Mesolithic hunter233 gatherers (HG) from Europe; (b) Anatolian farmer-related ancestry related to Neolithic farmers
234 from the Near East and associated to the spread of farming to Europe; and (c) Steppe pastoralist235 related ancestry that is related to the Yamnaya pastoralists from Russia and Ukraine (16–19). Many
236 open questions remain about the timing and dynamics of these population interactions, in particular
237 related to the formation of the ancestral groups (which were themselves admixed) and their
238 expansion across Europe. To characterize the spatial and temporal patterns of mixtures in Europe
239 in the past 10,000 years, we used 1,096 ancient European samples from 152 groups from the
240 publicly available Allen Ancient DNA Resource (AADR) spanning a time range of ~8,000–350
241 BCE (Methods, Table SA). Using DATES, we characterized the timing of the various gene flow
242 events, and below, we describe the key events in chronological order focusing on three main
243 periods.
244
245 Holocene to Mesolithic: Pre-Neolithic Europe was inhabited by hunter-gatherers until the arrival
246 of the first farmers from the Near East (20, 21). There was large diversity among hunter-gatherers
247 with four main groups–– western hunter-gatherers (WHG) that were related to the Villabruna
248 cluster in central Europe, eastern hunter-gatherers (EHG) from Russia and Ukraine related to the
249 Upper Paleolithic group of Ancestral North Eurasians (ANE) ancestry, Caucasus hunter-gatherers
250 (CHG) from Georgia associated to the first farmers from Iran, and the GoyetQ2-cluster associated
251 to the Magdalenian culture in Spain and Portugal (18, 22–25). Most Mesolithic HGs fall on two
252 main clines of relatedness: one cline that extends from Scandinavia to central Europe showing
253 variable WHG–EHG ancestry, and the other in southern Europe with WHG–GoyetQ2 ancestry
254 (23). This ancestry is already present in the 17,000 BCE El Mirón individual from Spain,
255 suggesting that the GoyetQ2-related gene flow occurred well before the Holocene. However, the
256 WHG–EHG cline was formed more recently during the Mesolithic period, though the precise
257 timing of the spread of EHG ancestry remains less well understood.
258 To characterize the formation of the WHG–EHG cline, we used genomic data from 16
259 ancient HG groups (n=101) with estimated ages of ~7,500–3,600 BCE. We first verified the
260 ancestry of each HG group using qpAdm that compares the allele frequency correlations between
261 the target and a set of source populations to formally test the model of admixture and then infer
262 the ancestry proportions for the best fitted model (16). For each target population, we chose the
263 most parsimonious model, i.e., fitting the data with the minimum number of source populations.
264 Consistent with previous studies, our qpAdm analysis showed that most HGs from Scandinavia,
265 the Baltic Sea region, and central Europe could be modeled as a two-way mixture of WHG and
266 EHG-related ancestry (Table S5.1, Note S5). To confirm that the target populations do not harbor
267 Anatolian farmer-related ancestry (that could lead to some confounding in estimated admixture
268 dates), we applied D-statistics of the form D(Mbuti, target, WHG, Anatolian farmers) where target
269 = Mesolithic HGs. We observed that none of the target groups had a stronger affinity to Anatolian
270 farmers than WHG, suggesting that the mixtures we date below reflect pre-Neolithic contacts
271 between the HGs (Table S5.2).
272 To infer the timing of the mixtures in the history of Mesolithic European HGs, we applied
273 DATES to hunter-gatherers from Scandinavia, the Baltic regions, and central Europe. We inferred
274 that the earliest admixture occurred in Scandinavian HGs from Norway and Sweden around ~80–
275 113 generations before the samples lived (Figure SB). Accounting for the average sampling age
276 of the specimens and the mean human generation time of 28 years (1), this translates to a timing
277 of admixture of ~10,200 BCE for Norway and Sweden Mesolithic individuals, though dates are
278 more recent (~8,000 BCE) in the Motala HG’s. In the Baltic region, we inferred admixture dates
279 of ~8,700–6,000 BCE in Latvia and Lithuania HGs, postdating the mixture in Scandinavia (Figure
280 3). In southeast Europe, the Iron Gates region of the Danube Basin shows widespread evidence of
281 mixtures between hunter-gatherer groups and, in the case of some outliers, mixture of hunter
282 gatherers and Anatolian farmer-related ancestry as early as the Mesolithic period (26). Further,
283 these groups showed strong affinity to the WHG-related ancestry in Anatolian populations,
284 suggesting ancient interactions with Near Eastern populations (26). We applied qpAdm to test the
285 model of admixture in Iron Gates HG and found that the parsimonious model with WHG and EHG
286 provides a good fit to the data. Further, when we tested the model with Anatolian-related ancestry
287 using Anatolian HG (AHG) as an additional source population, AHG was not required as the AHG
288 ancestry proportion was not significant (Table S5.1.1 and S5.1.2). Applying DATES to Iron Gates
289 HG with WHG and EHG as reference populations, we inferred this group was genetically formed
290 in ~10,000–8400 BCE. Our samples of the Iron Gates HGs include a wide range of C14 dates
291 between 8,800–5,700 BCE. We confirmed our dates were robust to the sampling age of the
292 individuals as we obtained statistically consistent dates when all samples were combined as one
293 group or when subsets of samples were grouped in bins of 500 years (Figure SA). The most recent
294 dates of ~7,500 BCE were inferred in eastern Europe in Ukraine HGs, highlighting how the WHG
295 EHG cline was formed over a period ~2000–3000 years (Figure 3, Table SC).
296
297 Early to middle Neolithic: Neolithic farming began in the Near East––the Levant, Anatolia, and
298 Iran––and spread to Europe and other parts of the world (18, 20, 27). The first farmers of Europe
299 were related to Anatolian farmers, whose origin remains unclear. The early Neolithic Anatolian
300 farmers (Aceramic Anatolian farmers) had majority ancestry from AHG with some gene flow from
301 the first farmers from Iran (26). AHG, in turn, had ancestry from Levant HG (Natufians) and some
302 mysterious hunter-gatherer group related to the ancestors of WHG individuals from central
303 Europe–– a gene flow event that likely occurred in the late Pleistocene (26). Using qpAdm, we
304 confirmed that early Anatolian farmers could be modeled as a mixture of AHG and Iran Neolithic
305 farmer-related groups (Note S5). To learn about the timing of the genetic formation of early
306 Anatolian farmers, we applied DATES using one reference group as a set of pooled individuals of
307 WHG-related and Levant Neolithic farmers-related individuals as a proxy of AHG ancestry and
308 the second reference group containing pooled Iran Neolithic farmer-related individuals. We note
309 that the application of DATES to three-way admixed groups can lead to intermediate dates between
310 the first and second pulse of gene flow unless the reference populations are chosen carefully (Table
311 S2.1). Our setup for early Anatolian farmers should have limited confounding and should recover
312 the timing of the most recent event (in this case, the gene flow from CHG or Iran Neolithic-related
313 groups) reliably (Table S2.1). We infer the Iran Neolithic farmer-related gene flow occurred
314 ~10,900 BCE (12,200–9,600 BCE), predating the origin of farming in Anatolia (28). During the
315 subsequent millennia, these early farmers further admixed with Levant Neolithic groups to form
316 Anatolian Neolithic farmers who spread towards the west to Europe and in the east to mix with
317 Iran Neolithic farmers, forming the Chalcolithic groups of Seh Gabi and Hajji Firuz. Using
318 DATES, we inferred the Chalcolithic groups were genetically formed in ~7,600–5,700 BCE (Table
319 SC).
320 In Europe, the Anatolian Neolithic farmers mixed with the local indigenous hunter
321 gatherers replacing between ~3-50% ancestry of Neolithic Europeans. To elucidate the fine-scale
322 patterns and regional dynamics of these mixtures, we applied DATES to time transect samples
323 from 94 groups (n=657) sampled from sixteen regions in Europe, ranging from ~6,000-1,900 BCE
324 and encompassing individuals from the early Neolithic to Chalcolithic periods (Table SB). Using
325 qpAdm, we first confirmed that the Neolithic Europeans could be modeled as a mixture of
326 European hunter-gatherer-related ancestry and Anatolian farmer-related ancestry and inferred their
327 ancestry proportions (Table SD). For most target populations (~80%), we found the model of gene
|
|
|
|
Post by Admin on Feb 5, 2022 21:31:06 GMT
 Figure 2: Genetic formation of early Anatolian farmers and early Bronze Age Steppe pastoralists.
328 flow between Anatolian farmer-related and WHG-related ancestry provided a good fit to the data
329 (p-value > 0.05). In some populations, we found variation in the source of the HG-related ancestry
330 and including either EHG or GoyetQ2 improved the fit of the model. In five groups, none of the
331 models fit, despite excluding outlier individuals whose ancestry profile differed from the majority
332 of the individuals in the group (Table SD, Table SE). To confirm that the target populations do not
333 harbor Steppe pastoralist-related ancestry, we applied D-statistics of the form D(Mbuti, target,
334 Anatolian farmers, Steppe pastoralists) where target = Neolithic European groups. We observed
335 that four groups had a stronger affinity to Steppe pastoralists compared to Anatolian farmers, and
336 hence we excluded these from further analysis (Table SF). After filtering, we applied DATES to
337 86 European Neolithic groups using WHG-related individuals and Anatolian farmers as reference
338 populations.
339 Earlier analysis has suggested that farming spread along two main routes in Europe, from
340 southeast to central Europe (‘continental route’) and along the Mediterranean coastline to Iberia
341 (‘coastal route') (23, 29, 30). Consistent with this, we inferred one of the earliest timings of gene
342 flow in the Balkans around 6,400 BCE. Using the most comprehensive time-transect in Hungary
343 with 19 groups (n=63) spanning from middle Neolithic to late Chalcolithic, we inferred that the
344 admixture occurred between ~6,100–4,500 BCE. Under a model of a single shared gene flow event
345 in the common ancestors of all individuals, we would expect to obtain similar dates of admixture
346 (before present) after accounting for the age of the ancient specimens. Similar to Lipson et al.
347 (2017), we observed that the estimated dates in middle Neolithic individuals were substantially
348 older than those inferred in late Neolithic or Chalcolithic individuals (Figure 3). This would be
349 expected if the underlying model of gene flow involved multiple pulses of gene flow, such that the
350 timing in the middle Neolithic samples reflects the initial two-way mixture and the timing in the
351 Chalcolithic samples captures both recent and older events. Interestingly, Lipson et al. (2017) and
352 other recent studies have documented increasing HG ancestry from ~3-15% from the Neolithic to
353 Chalcolithic period (16, 23, 31), suggesting that there was additional HG gene flow after the initial
354 mixture. This highlights that the interactions between local hunter-gatherers and incoming
355 Anatolian farmers were complex with multiple gene flow events between these two groups, which
356 explains the increasing HG ancestry and more recent dates in Chalcolithic individuals (Table SD).
357 Mirroring the pattern in Hungary, we documented the resurgence of HG ancestry in the
358 Czech Republic, France, Germany, and southern Europe. In central Europe, we inferred that the
359 Anatolian farmer-related gene flow occurred ~5,600-5,000 BCE, with some exceptions. In the
360 Blätterhöhle site from Germany, we inferred the gene flow occurred more recently (~4,000 BCE),
361 consistent with the occupation of both hunter-gatherers and farmers in this region until the late
362 Neolithic (31). In eastern Europe, using samples related to the Funnel Beaker culture (TRB; from
363 German Trichterbecher) from Poland, we dated the Anatolian farmer-related gene flow occurred
364 ~5,300–4,200 BCE. Following the TRB decline, the Baden culture and the Globular Amphora
365 culture appeared in many areas of Poland and Ukraine (25). These cultures had close contacts with
366 Corded Ware complex and steppe societies, though we did not find any evidence of Steppe
367 pastoralist-related ancestry in the GAC individuals (Table SD). Applying DATES, we inferred the
368 Anatolian farmer-related and HG mixture occurred ~5,200-3,100 BCE, predating the spread of
369 Steppe pastoralists to eastern Europe (16, 19).
370 Along the Mediterranean route, we characterized Anatolian farmer-related gene flow in
371 Italy, Iberia, France, and the British Isles. Using samples from five groups in Italy, we inferred the
372 earliest dates of Anatolian farmer-related gene flow of ~6,100 BCE, and within the millennium,
373 the ancestry spread from Sardinia to Sicily (Figure 3). In Iberia, the Anatolian farmer-related
374 mixture occurred ~6,000–3,400 BCE and showed evidence for an increase in HG ancestry from
375 ~9–20% after the initial gene flow. In France, previous studies have shown that Anatolian farmer
376 related ancestry came from both routes, along the Danubian in the north and along the
377 Mediterranean in the south (23). This is reflected in the source of the HG ancestry, which is
378 predominantly EHG and WHG-related in the north and includes WHG and Goyet-Q2 ancestry in
379 the south (23). Consistently, we also observed that the admixture dates in France were structured
380 along these routes, with the median estimate of ~5,100 BCE in the east and much older ~5,500
381 BCE in the south (Table SC). In Scandinavia, we inferred markedly more recent dates of admixture
382 of ~4,300 BCE using samples from Sweden associated with the TRB culture and Ansarve
383 Megalithic tombs, consistent with a late introduction of farming to Scandinavia (33).
384 Finally, we inferred recent dates of admixture in Neolithic samples from the British Isles
385 (England, Scotland, and Ireland) with the median timing of ~5,000 BCE across the three regions.
386 Interestingly, unlike in western and southern Europe, there was no resurgence in HG ancestry
387 during the Neolithic in Britain (34). This suggests our dates can be interpreted as the time of the
388 main mixture of HGs and Anatolian farmers in this region, implying that the farmer-related
389 ancestry reached Britain a millennium after its arrival in continental Europe. By 4,300 BCE, we
390 find that Anatolian farmer-related ancestry is present in nearly all regions in Europe.
391
|
|
