Ancient Genomics Reveals Four Prehistoric Migration Waves 2 into Southeast Asia

https://www.biorxiv.org/content/biorxiv/early/2018/03/08/278374.full.pdf

Abstract: Two distinct population models have been put forward to explain present-day
58 human diversity in Southeast Asia. The first model proposes long-term continuity
59 (Regional Continuity model) while the other suggests two waves of dispersal (Two Layer
60 model). Here, we use whole-genome capture in combination with shotgun sequencing to
61 generate 25 ancient human genome sequences from mainland and island Southeast Asia,
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
62 and directly test the two competing hypotheses. We find that early genomes from
63 Hoabinhian hunter-gatherer contexts in Laos and Malaysia have genetic affinities with
64 the Onge hunter-gatherers from the Andaman Islands, while Southeast Asian Neolithic
65 farmers have a distinct East Asian genomic ancestry related to present-day Austroasiatic66 speaking populations. We also identify two further migratory events, consistent with the
67 expansion of speakers of Austronesian languages into Island Southeast Asia ca. 4 kya,
68 and the expansion by East Asians into northern Vietnam ca. 2 kya. These findings
69 support the Two Layer model for the early peopling of Southeast Asia and highlight the
70 complexities of dispersal patterns from East Asia.

***

Main Text:
73 The population history of Southeast Asia (SEA) has been shaped by interchanging
74 periods of isolation and connectivity. Anatomically modern humans first colonized SEA at
75 least 70,000 years ago (kya) (1–3). Within SEA, the complex topography and changes in sea
76 level promoted regional expansions and contractions of populations. By the late
77 Pleistocene/early Holocene, a pan-regional lithic technological culture was established across
78 mainland SEA, named Hoabinhian (4–7). Hoabinhian foragers are thought to be the ancestors
79 of present-day SEA hunter-gatherers, sometimes referred to as ‘Negritos’ because of their
80 comparatively darker skin colour and short stature. Today, however, the majority of people in
81 SEA are believed to be descendants of rice and millet farmers with varying degrees of East
82 Asian phenotypic affinity, suggesting that human diversity in SEA was strongly influenced by
83 population expansions from the north (4). Yet, the extent to which the movements from East
84 Asia (EA) impacted on the genetic and cultural makeup of the people of SEA remains
85 controversial.
86 Two distinct population models have been proposed to account for the biological and
87 cultural diversity of human populations in present-day SEA. The Regional Continuity model,
88 based primarily on morphological evidence, argues for a long-standing evolutionary continuity
89 without significant external gene flow and for the Neolithic transition in SEA occurring as
90 hunter-gatherer groups adopted agriculture, either independently or through cultural contact
91 (8–21). While this model does not necessarily argue for the independent domestication of crops
92 across SEA, it posits that gene flow from EA farmers was not the main mechanism behind the
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
93 Neolithic transition. In contrast, the Two Layer model advocates for two major dispersal waves
94 into SEA, where EA farmers replaced the original Hoabinhian inhabitants across SEA through
95 a major demographic southward expansion ca. 4 kya. The exception to this would be the
96 isolated populations of the Andaman Islands, peninsular Thailand/Malaysia and the Philippines
97 which are considered the primary descendants of Hoabinhian hunter-gatherers (22, 23). Under
98 this model, the migratory wave of farmers originated in present-day China, where rice and
99 millet were fully domesticated in the Yangtze and Yellow River valleys between 9-5.5 kya,
100 and paddy fields developed by 4.5 kya (4, 24–26). Farming practices are thought to have
101 accompanied these populations as they spread southward through two main routes – an inland
102 wave associated with the expansion of Austroasiatic languages, and an island-hopping route
103 associated with Austronesian languages which eventually reached the Pacific (27, 28). Within
104 mainland SEA (MSEA), exchanges with EA appear to have continued in the recent past,
105 however, the extent to which these expansions had a genetic impact on the indigenous
106 populations is unknown.
107 Genetic studies of contemporary SEA populations have not resolved these
108 controversies (29–32). Ancient genomics can provide direct evidence of past population
109 events. However, SEA is characterised by tropical and monsoonal climates which cause heavy
110 weathering and acidification of soils (33), so ancient genomic studies have, so far, been
111 unsuccessful there. Though shotgun sequencing has revolutionized ancient genomic studies by
112 allowing the retrieval of all mappable DNA fragments from an ancient sample (34, 35), the
113 inverse relationship between the proportion of endogenous DNA and the cost of shotgun
114 sequencing makes this approach impractical to apply widely to regions with poor DNA
115 preservation such as SEA. Genome wide SNP capture is one way to circumvent the issue (36,
116 37), but it retrieves only a pre-selected subset of all variants of the genome and thus sacrifices
117 the full potential of rare and irreplaceable fossil samples. An alternative approach is whole
118 genome capture in which human ancient human DNA fragments are enriched through
119 hybridisation to baits that cover the entire mappable human genome (15).
120 We performed comparative testing of three different capture approaches for human
121 DNA - the SeqCap EZ Human Exome Kit v3.0 cat no. 6740294001 (Roche Nimblegen, CA,
122 USA), the SureSelect Human All Exon V5+UTRs cat. no. 5190-6213 (Agilent Technologies)
123 and the Custom MYbaits Whole Genome Enrichment (WGE) Kit version 2.0 (Arbor
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
124 Biosciences) - with the aim of applying the most effective method to ancient human remains
125 from tropical SEA (SOM1). We found a modified version of MYbaits Whole Genome
126 Enrichment to be the best-performing method. We applied this method, in combination with
127 shotgun sequencing approaches where sufficient endogenous DNA allowed it, to samples from
128 Malaysia, Thailand, Philippines, Vietnam, Indonesia and Laos, dating between 0.2 and 8 kya
129 (SOM2). We obtained 25 low-coverage ancient genomes (Table 1), along with mtDNA and
130 nuclear DNA from an additional set of 16 individuals (Table S3), belonging to hunter-gatherers
131 from the Hoabinhian culture, as well as Neolithic, Bronze Age and Iron Age farmers (SOM3).
132 All samples showed damage patterns typical of ancient DNA (38) (Table S3).
133 To address the genetic relationships among the ancient individuals, we performed a
134 principal component analysis (PCA) with our Pan-Asia Panel (see Methods) using smartpca
135 (39). We projected the ancient samples onto the first two principal components of a PCA
136 constructed solely with present-day samples (40) (SOM4). We then used ADMIXTURE (41)
137 to find reference latent ancestry components that could best fit our present-day data, and then
138 used fastNGSadmix (42, 43) to model the low-coverage ancient samples as mixtures of these
139 reference components (SOM5). Unlike all other ancient samples, the two Hoabinhian samples
(which also happen to be the oldest samples in our study) - Pha Faen, Laos (La368 - 14 140 C 7,888
± 40) and Gua Cha, Malaysia (Ma911 - 14 141 C 4,319 ± 64) - designated as Group 1, cluster
142 distantly from most East and Southeast Asians in the PCA and position closely to present-day
143 Onge (Figure 1A). Group 1 individuals also contain a mixture of several different ancestral
144 components in the fastNGSadmix plot, including components shared with Onge, the Pahari and
145 Spiti from India, Papuans and Jehai (a Malaysian ‘Negrito’ group), which are markedly
146 different from the other SEA ancient samples. This possibly results from our modeling of
147 ancient populations as a mixture of components inferred in present-day populations, via
148 fastNGSadmix (44), and from the fact the ancient samples are likely poorly represented by a
149 single present-day group. The rest of the ancient samples are defined primarily by East and
150 Southeast Asian components that are maximised in present-day Austroasiatic (Mlabri and
151 Htin), Austronesian (Ami) and Hmong (indigenous to the mountainous regions of China,
152 Vietnam, Laos and Thailand) populations, along with a broad East Asian component.
153 We used outgroup f3 statistics (f3(Mbuti;X,Ancient samples)) to determine which
154 populations have the highest levels of shared drift with each of the ancient individuals (SOM6).
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
155 Group 1 shares the most drift with certain ancient mainland samples (Figure S12, Table S4).
156 Again, we see that the closest present-day populations to Group 1 are from the Andaman
157 Islands (Onge) and then Kensiu (a Malaysian ‘Negrito’ group), Ami and Jehai, followed by a
158 mix of East and Southeast Asian populations.
159 We used D-statistics of the form D(Papuan,Tianyuan,X,Mbuti), where X is a test
160 population, to explore the relatedness of ancient and present-day Southeast Asians to two
161 highly differentiated groups: Papuans and an ancient northern East Asian individual (Tianyuan
162 - a 40 kya-old sample from Northeastern China (45)). The values of this D-statistic are
163 consistent with present-day and ancient SEA mainland samples being more closely related to
164 Tianyuan than to Papuans (SOM7). This applies to present-day northern EA populations, and -
165 more weakly - to most populations of ancient and present-day SEA. However, this D-statistic
166 is not significantly different from 0 in present-day Jehai, Onge, Jarawa and Group 1 - the
167 ancient Hoabinhians (Figure 2B, Tables S12, SOM7). While the Onge’s relationship with
168 Papuans and Tianyuan is unclear, D-statistics of the form D(Onge,Tianyuan,X,Mbuti), where
169 X is a test population, show that Jarawa, Jehai and the ancient Group 1 share more ancestry
170 with Onge than with Tianyuan (Figure 2C, SOM7). Like the Onge, both Group 1 samples carry
171 mtDNA haplogroups from the M lineage (Table S3), thought to represent the coastal migration
172 to Australasia (12, 13, 28, 46).
173 To assess the diversity among the remaining ancient individuals, we computed a new
174 PCA including only EA and SEA populations that did not have considerable Papuan or Onge175 like ancestry in the fastNGSadmix analysis (Figure S11), as it was done in the Pan-Asian SNP176 capture study (30). We observe that the remaining ancient samples form five slightly
177 differentiated clusters within the EA and SEA populations (Groups 2-6, Figure 1B), in broad
178 concordance with the fastNGSadmix (at K=13, Figure 1) and f3 results (Figure S12-S19;
179 SOM4). We thus decided to organize these samples into five more groups to facilitate further
180 analyses (Groups 2-6, Table 1), although we note that genetic differentiation among them
181 seems to be highly clinal. Samples Vt719, Th531 and Vt778 were either geographic or
182 temporal outliers to their groups and were therefore analyzed separately in groups denoted by a
183 “.1”: Group 3.1 (Th531, Vt719) and Group 4.1 (Vt778).
184 Group 2 samples from Vietnam, Laos, and the Malay Peninsula are the oldest samples
185 after Group 1, and range in age from 4.2 to 2.2 kya. At K = 6 (SOM5), Group 2 individuals, the
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
186 present-day Mlabri and a single Htin individual are the only MSEA samples in the
187 fastNGSadmix analysis to lack the broad EA component (dark green) maximised in northern
188 EA , with the exception of the Malaysian ‘Negritos’ and ‘Proto-Malays’ (Temuan). At K = 7, a
189 bright green component is maximised in these populations, and this component is also found in
190 present-day Indonesian samples west of Wallace’s Line. The two ancient Indonesian samples
191 (Group 5; 2.2 to 1.9 kya) represent a mix of Austronesian- and Austroasiatic-like ancestry,
192 similar to present-day western Indonesians. Indeed, after Mlabri and Htin, the closest
193 populations to Group 2 based on outgroup-f3 statistics are the western Indonesian samples
194 (from Bali and Java) reported to have the highest amounts of ancestry from mainland SEA (47)
195 (Figure S13).
196 These lines of evidence suggest Group 2 are possible descendants of an “Austroasiatic”
197 migration that expanded southward across MSEA and into island SEA (ISEA) by 4 kya (27,
198 47–49). We also observe a gradient in “Austronesian-like” vs. “Austroasiatic-like” ancestry in
199 the PCA (Figure 1B): while PC1 separates populations found in SEA and those found in
200 northern EA, PC2 distinguishes population based on their amounts of Austronesian-like
201 ancestry (pink component in Figure 1 - lower panel) versus Austroasiatic-like ancestry (bright
202 green component in Figure 1 - lower panel).
203 Group 6 samples are recent (between 1.8 and 0.2 kya) and come from Malaysia and the
204 Philippines. They fall within the variation of present-day populations with high Austronesian
205 ancestry in these areas. Group 6 also contains the individual (Ma554) with the highest amounts
206 of Denisovan ancestry relative to the other ancient samples, although variation in archaic
207 ancestry is not very strong across MSEA (SOM10).
208 The remaining mainland samples (Groups 3 and 4) are dated to be from 2.6 to 0.2 kya.
209 They appear similar to present-day MSEA populations and fall into two groups. Group 3 is
210 largely composed of ancient samples from Vietnam but also includes one sample from
211 Thailand (Th531); these samples cluster in the PCA with the Dai from China, Tai-Kadai
212 speakers from Thailand and Austroasiatic speakers from Vietnam, including the Kinh (Figures
213 S9). In contrast, Group 4 largely contains ancient samples from Long Long Rak, Thailand, but
214 also includes the inland-most sample from Vietnam (Vt778). These samples fall within the
215 variation of present-day Austroasiatic and Sino-Tibetan speakers from Thailand and China,
216 supporting the hypothesis that the Long Long Rak population originated in South China, and
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
217 subsequently expanded southward during the Dongson period (50). At Long Long Rak, three
218 individuals (Th387,Th530 and Th531) dated to approximately 1.6 kya were found in the same
219 chamber. Interestingly, all three individuals share the same mtDNA haplogroup (G2b1a), but
220 the nuclear ancestries for the two samples which yielded genome-wide data are quite different:
221 Th531 clusters best with Group 3, while Th530 with Group 4. These results suggests that
222 individuals with ancestry from distant regions likely cohabited at this locality.
223 To determine if any of the ancient samples had affinities to particular populations
224 outside SEA, we computed D-statistics of the form D(Group A, Group B, Not-SEA,
225 Yoruba/Mbuti) to compare each of the ancient groups. Group 2 has a significant affinity to the
226 Indian populations of Khonda Dora (Z = 6.3), relative to Group 3 (D(Group2,Group3,Khonda
227 Dora,Yoruba)), in agreement with previous reports of East Asian ancestry in tribal Indian
228 Groups (51, 52). We also investigated the affinity between certain Australasian populations
229 and particular Native American groups, like the Surui (45, 53, 54). When computing D(Mixe,
230 Surui, X,Yoruba), we find that the Group 1 samples had some suggestive but non-significant
231 affinity to Surui relative to Mixe (Z = -2.18 when X = Ma911, Z = -2.48 when X = La368;
232 Table S19), although the signal is not as robust as observed for Tianyuan (Z = -3.53), Khonda
233 Dora (Z = -3.04) and Papuans (Z = -3.02), among others (53, 54). We note, however, that there
234 are much fewer SNPs to compute this statistic on Group 1 samples than on the other
235 populations (La368: 191,797; Ma911: 47,816; Tianyuan: 295,628, Papuan: 471,703, Khonda
236 Dora: 496,097), thus we may be underpowered to detect this signal.
237 We used TreeMix (55) to explore admixture graphs that could potentially fit our data.
238 The ancient Group 1 (Onge-like) individuals are best modelled as a sister group to present-day
239 Onge (Figures 3A, S21-S23). For the highest-coverage Group 1 sample, allowing for one
240 migration, TreeMix fits Papuans as receiving admixture from Denisovans, while the second
241 migration shows East Asian populations as resulting from admixture between Tianyuan and
242 Onge.
243 We also performed a more supervised form of admixture graph modeling using
244 qpGraph (36) (SOM9). We began with a skeletal framework containing chimpanzee (PanTro2,
245 EPO alignment from Ensemb 71 (56, 57)), Denisova (58), Altai Neanderthal (59), Kostenki-14
246 (60), Mbuti, Onge, Ami and Papuan, fitting a graph based on results from Lipson and Reich
247 (61) and well-supported D-statistics (SOM7). When not including Tianyuan, we find that the
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
248 Onge-Papuan-Ami split is hard to resolve (Figure S33), in agreement with Lipson and Reich
249 (61). However, when including Tianyuan (Figure S34), we find that a best fit occurs when Ami
250 (or Han) are modelled as an admixed group, with ancestry from a population related to
251 Tianyuan and a population related to Onge (worst-fitting Z = -3.564). In support of this graph
252 assignment, we find that D(Ami, Onge, Tianyuan, Mbuti) = 0.0239 (Z = 5.148), while Papuan
253 and Onge are a clade with respect to Tianyuan: D(Papuan, Onge, Tianyuan, Mbuti) = -0.0047
254 (Z = -0.886). We then added either La368 or Ma911 to the graph. In agreement with the
255 TreeMix results, we find that La368 and Ma911 are each best modeled as a sister group to
256 Onge (Figures S35 and S36, worst-fitting Z = 3.372 and 3.803, respectively).
257 We then used qpWave/qpAdm (62, 63) to determine if La368 and Ma911 can be
258 modelled as a linear combination of ancestries from Papuans, Onge and/or Tianyuan without
259 the need to invoke partial ancestry from a population that may have split from them before
260 these populations split from each other. As outgroup populations, we used Yoruba (64), Ust261 Ishim (65), Kostenki-14 (60), Mal’ta (66), Afontova Gora 3, Vestonice 16, El Mirón and
262 Villabruna (67). All best 3-way and 2-way combinations for La368 are not feasible (have
263 negative admixture weights). There are two 1-way possibilities (La368 as a sister group to
264 either Onge or Papuans) that are feasible and are good fits (P = 0.37 and P = 0.27,
265 respectively), and this is somewhat expected as Onge and Papuans are sister clades to each
266 other - barring Denisovan introgression into Papuans. When performing the same analysis on
267 Ma911 as the target population, we find that all the best 3-way and 2-way combinations are
268 also infeasible and the only good 1-way fit is with Onge (P = 0.49). Modelling Ami as a linear
269 combination of the same three source populations results in any of the 3-, 2- or 1-way fits
270 being feasible and good fits, but the best fit is found in the 2-way combination of Tianyuan and
271 Onge (P = 0.98).
272 When modeling the Mlabri-like Group 2 in TreeMix, we see that the two samples with
273 the highest coverage in this group (La364 and Ma912) form a clade, resulting from an
274 admixture event between the ancestral populations of present-day East Asians (Han/Ami) and
275 the ancestors of La368 (Figures 3B, S24-27). Despite the low SNP overlap (~20,000 SNPs)
276 when including the Group 1 and 2 samples from Laos and Malaysia, (La368, Ma911, La364,
277 Ma912), at 3 migrations, TreeMix residuals suggest that the Onge-like ancestry in Malaysia
278 and Laos is a result of local admixture (Figure S27, SOM8). Additional data and higher
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
279 coverage samples from these regions are needed to better support a ‘local admixture’ model:
280 including all four low-depth genomes in the same admixture graph results in only 17,286
281 overlapping SNPs (including transitions), which makes inference difficult. The Jehai are best
282 fitted as an admixed population between Group 2 (Ma912) and the branch leading to present283 day Onge and La368 (Figure 3C, S28). ISEA ancient samples from Indonesia (Group 5) and
284 Borneo (Group 6) are best modelled as an admixed population carrying the signature of Group
285 2 (Figure 3D, Figures S29-33), supporting a previously reported mainland component in ISEA
286 complementary to the well-documented Austronesian expansion (47). For the ISEA samples
287 (Group 5 - In662 and Group 6 - Ma554), when a more basal migration event occurs, it
288 originates from the Papuan branch, rather than the Onge branch as seen in MSEA.
289 Consistent with the TreeMix results, La364 in qpGraph is best modeled as a mixture of
290 a population ancestral to Ami and the Group 1 / Onge-like population (Figure 3E, worst-fitting
291 Z = 3.667). Additionally, we find the best model for present-day Dai populations is a mixture
292 of Group 2 individuals and an additional pulse of admixture from East Asians (Figure S37,
293 worst-fitting Z = 3.66).
294 This is the first study to reconstruct the population history of SEA using ancient DNA.
295 We find that the genetic diversity found in present day SEA populations derives from at least
296 four prehistoric population movements by the Hoabinhians, an “Austroasiatic-like” population,
297 the Austronesians and, finally, additional EA populations into MSEA. We further show that the
298 ancient mainland Hoabinhians (Group 1) shared ancestry with present-day Onge of the
299 Andaman Islands and the Jehai of peninsular Malaysia. These results, together with the
300 absence of significant Denisovan ancestry in these populations, suggest that the Denisovan
301 admixture observed in Papuans occurred after their ancestors split from the ancestors of the
302 Onge, Jehai and the ancient Hoabinhians. This is also consistent with the presence of
303 substantial Denisovan admixture in the Mamanwa from the Philippines, which are best
304 modeled as resulting from an admixture between Austronesians and Papuans, not Onge (61).
305 Consistent with the Two Layer model, we observe a dramatic change in ancestry by 4
306 kya (Group 2) which coincides with the introduction of farming, and thus supports models that
307 posit a significant demographic expansion from EA into SEA during the Neolithic transition.
308 Group 2 are the oldest samples with distinctive EA ancestry that we find. The most closely
309 related present-day populations to Group 2 are the Mlabri and Htin - the Austroasiatic hill
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
310 tribes of Thailand - which is in agreement with hypotheses of an early Austroasiatic farmer
311 expansion into the region. They also share ancestry with the Temuan and Jehai of Peninsular
312 Malaysia and populations of Western Indonesia, supporting an Austroasiatic (“Western
313 Route”) expansion into ISEA (post-Hoabinhian, pre-Austronesian), as previously proposed
314 based on linguistic and archaeological grounds (27, 49, 68). Furthermore, a recent study also
315 identified populations of Bali and Java as the groups in ISEA with the highest frequency of
316 mainland SEA ancestry (47), also reflected in the large amounts of shared drift between Group
317 2 and the Javanese that we observe (Figure S13). The extent and nature of this Austroasiatic
318 expansion into western Indonesia prior to the Austronesian expansion could be resolved by
319 sequencing ancient genomes from ISEA prior to the Austronesian expansion.
320 By around 2 kya, all ancient mainland samples carry additional EA ancestry
321 components that are absent in Group 2. Within the variation of these recent samples, we find
322 two clusters of ancestry, possibly representing independent EA migrations into mainland SEA.
323 Group 3 has affinities to the Hmong, the Dai from China, the Thai from Thailand and the Kinh
324 from Vietnam, while Group 4 individuals - found only in inland regions - have affinities to
325 Austroasiatic Thai and Chinese speakers. Finally, we also find evidence for the arrival of
326 Austronesian ancestry into the Philippines by 1.8 kya (Group 6) and into Indonesia by 2.1 kya
327 (Group 5). By 2 kya, the population structure in MSEA was very similar to that among present328 day individuals. Despite observing a clear change in genetic structure coinciding with the
329 transition from the Hoabinhian hunter-gatherers to Neolithic farmers, we also see a degree of
330 local continuity at all sites at different points in time, suggesting that incoming waves of
331 migration did not completely replace the previous occupants in each area (Figure 4).
332 This study demonstrates that whole-genome capture is an efficient supplementary
333 approach for retrieving whole genomes from the fossil skeletal and dental remains found in the
334 tropics. As target enrichment inevitably results in subsampling library fragments, it is most
335 useful for (combined) libraries with high underlying complexity. We found a median 7.5 fold
336 enrichment, reducing the sequencing costs proportionally. By enriching the human DNA
337 content, we were able to acquire whole-genome data from selected samples in which the low
338 proportion of endogenous DNA would have been previously prohibitive. The whole genome
339 approach which we have employed here combines shotgun sequencing and capture in order to
340 maximise the potential of ancient samples.
It is made available under a CC-BY-NC-ND 4.0 International license.
was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
bioRxiv preprint first posted online Mar. 8, 2018; doi: http://dx.doi.org/10.1101/278374. The copyright holder for this preprint (which
341 The clear genetic distinction between the Onge-like Hoabinhian and EA Neolithic
342 demonstrated by this study provides an overwhelming support for the Two Layer model and
343 indicates that in SEA, like in Europe, the onset of agriculture was accompanied by a
344 demographic transition. However, on a more local level, our results point toward admixture
345 events in northern Laos and Peninsula Malaysia between the two dispersal layers. We also
346 show that the Hoabinhians of the first dispersal contributed a degree of ancestry to the
347 incoming EA populations, which may have also resulted in the passing on of some phenotypic
348 characteristics detected by proponents of the Continuity model. Finally, our results reveal that
349 the appearance of these Austroasiatic farmers at around 4 kya was followed by multiple
350 migrations of distinct EA ancestry. These subsequent migrations made significant contributions
351 to the diversity of human populations in present-day SEA.