Denisovan Ancestry in Papua New Guinea (PNG)

Around 55,000-50,000 years ago, a population of modern humans left Africa and started on the long trek that would lead them around the world. After rapidly crossing Eurasia and Southeast Asia, they traveled through the islands of Indonesia, and eventually as far as the continent of Sahul—modern-day Australia and New Guinea.

In new research published in Proceedings of the National Academy of Sciences, we detail how during this remarkable journey the ancestors of modern humans met and genetically mixed with a number of archaic human groups, including Neandertals and Denisovans, and several others for which we currently have no name. The traces of these interactions are still preserved in our genomes.



Leaving Africa

One of the first mixing events after the Neandertals appears to have taken place during the movement across southern Asia. The archaic human group involved was neither the Neandertals or Denisovans, but something similar—which currently has no name.

The genetic traces of this archaic group can be found from modern Punjabi and Bengal populations all the way through to New Guinea and Australia. As a result, we think this mixing event (marked 1 on the map) likely took place somewhere around northern India, which is the most "upstream" or westerly position it is first observed.

The ancestral population of modern humans then appears to have split as it moved across Asia with one pulse dispersing north into mainland Asia, where it met and mixed with a Denisovan group (marked 2 on the map). These Denisovans were genetically close to those we already know about from the Altai mountains. The traces of this event can be seen in East Asia today, and also in North and South America populations, who stem from northeastern Asia.



Island Southeast Asia was already crowded

The other pulse of modern humans headed south down the Malaysian Peninsula and into Island Southeast Asia (ISEA) where a big surprise awaited. They found the area was already crowded with different archaic human groups, including completely different species.

Recent fossil finds of small skeletons have shown that apparent relatives of Homo erectus (whose early fossils are common on Java) had survived on the Philippines and Flores (where they are known as "hobbits") until around 52,000 years ago. Effectively right up until the modern humans arrived.

The incoming modern human population apparently first met and mixed with a distant relative of the Denisovans in the area, leaving a signal in the genomes of Australo-Papuans and several ISEA populations. These signals are very different from the above East Asian mixing event, and instead come from a Denisovan relative that had separated genetically from the Altai/East Asian Denisovans around 280,000 years ago. This mixing event appears to have been somewhere around southern Malaysia/Borneo (marked 3 on the map).

Landfall in Australia

The wave of modern humans does not appear to have waited long to cross Wallace's Line—the famous biogeographic barrier that effectively marks the edge of the ISEA landmasses joined together during past glacial periods, when sea levels were up to 120 metres lower.

We know this because a sudden appearance of archaeological sites right across Australia around 50,000 years ago indicates that modern humans had quickly crossed the marine gaps through ISEA.

While there is one much earlier Australian site, the 65-80,000 year old Madjedbebe rock shelter in Arnhem Land, it is a complete outlier to the rest of the Australian record and the age of the site has been queried.

While moving through ISEA, the modern human population appears to have met—and mixed with—two more archaic human groups. Hunter-gatherer populations in the Philippines preserve signals of yet another Denisovan-mixing event (marked 4 on the map), after they had diverged from the main wave of modern humans moving through ISEA.

Abstract
The dispersal of anatomically modern human populations out of Africa and across much of the rest of the world around 55 to 50 thousand years before present (ka) is recorded genetically by the multiple hominin groups they met and interbred with along the way, including the Neandertals and Denisovans. The signatures of these introgression events remain preserved in the genomes of modern-day populations, and provide a powerful record of the sequence and timing of these early migrations, with Asia proving a particularly complex area. At least 3 different hominin groups appear to have been involved in Asia, of which only the Denisovans are currently known. Several interbreeding events are inferred to have taken place east of Wallace’s Line, consistent with archaeological evidence of widespread and early hominin presence in the area. However, archaeological and fossil evidence indicates archaic hominins had not spread as far as the Sahul continent (New Guinea, Australia, and Tasmania), where recent genetic evidence remains enigmatic.

PNAS first published July 12, 2019

Excavating Neandertal and Denisovan DNA from the genomes of Melanesian individuals
Benjamin Vernot1, Serena Tucci1,2, Janet Kelso3, Joshua G. Schraiber1, Aaron B. Wolf1, Rachel M. Gittelman1, Michael Dannemann3, Steffi Grote3, Rajiv C. McCoy1, Heather Norton4, Laura B. Scheinfeldt5, David A. Merriwether6, George Koki7, Jonathan S. Friedlaender8, Jon Wakefield9, Svante Pääbo3,*, Joshua M. Akey1,*

Science 08 Apr 2016:
Vol. 352, Issue 6282, pp. 235-239
DOI: 10.1126/science.aad9416

Denisovan DNA retained in Melanesians
Modern humans carry remnants of DNA from interbreeding events with archaic lineages, such as Neandertals. However, people from Oceania also retain genes from a second ancient lineage, the Denisovans. Vernot et al. surveyed archaic genomic sequences in a worldwide sample of modern humans, including 35 individuals from the Melanesian Islands. All non-African genomes surveyed contained Neandertal DNA, but a significant Denisovan component was found only in the Melanesians. Reconstruction of this genetic history suggests that Neandertals bred with modern humans multiple times, but Denosivans only once, in ancestors of modern-day Melanesians.

Science, this issue p. 235

Abstract
Although Neandertal sequences that persist in the genomes of modern humans have been identified in Eurasians, comparable studies in people whose ancestors hybridized with both Neandertals and Denisovans are lacking. We developed an approach to identify DNA inherited from multiple archaic hominin ancestors and applied it to whole-genome sequences from 1523 geographically diverse individuals, including 35 previously unknown Island Melanesian genomes. In aggregate, we recovered 1.34 gigabases and 303 megabases of the Neandertal and Denisovan genome, respectively. We use these maps of archaic sequences to show that Neandertal admixture occurred multiple times in different non-African populations, characterize genomic regions that are significantly depleted of archaic sequences, and identify signatures of adaptive introgression.

For much of human history, modern humans overlapped in time and space with other hominins (1). Analyses of the Neandertal (2, 3) and Denisovan (4, 5) genomes revealed that gene flow occurred between these archaic hominins and the ancestors of modern humans (3–8). Consequently, all non-African populations derive ~2% of their ancestry from Neandertals (3), whereas substantial levels of Denisovan ancestry (~2 to 4%) are only found in Oceanic populations (5), although low levels of Denisovan ancestry may be more widespread (9, 10). Recently, catalogs of introgressed Neandertal sequences have been created in Eurasians (11, 12). However, considerably less is known about the genomic organization and characteristics of Denisovan sequences that persist in modern individuals.

We sequenced 35 individuals from 11 locations in the Bismarck Archipelago of Northern Island Melanesia, Papua New Guinea (Fig. 1A) (13, 14), to a median depth of 40x (tables S1 and S2), including a trio to facilitate haplotype reconstruction. Unless otherwise noted, analyses were performed on a subset of 27 unrelated individuals (14). We integrated our sequencing data with single-nucleotide polymorphism genotypes from 1937 individuals spanning 159 worldwide populations (Fig. 1A and table S3) (14–16). A global principal components analysis (PCA) and ADMIXTURE analysis shows that our samples cluster most closely with other Oceanic individuals (Fig. 1B and fig. S1), and population sizes inferred by pairwise sequentially Markovian coalescent analysis are consistent with other non-African populations (fig. S2) (14). Furthermore, our Melanesian samples show genetic similarities to both Neandertals and Denisovans, whereas all other non-African populations only exhibit affinity toward Neandertals (Fig. 1C and fig. S3) (14). Using an f4 statistic (14, 15, 17), we find significant evidence of Denisovan ancestry (Z > 4) in our Melanesian samples, with admixture proportions varying between 1.9 and 3.4% (Fig. 1D and table S4).



Fig. 1 Melanesian genomic variation in a global context.
(A) Locations of the 159 geographically diverse populations studied. Information about the Melanesian individuals sequenced (blue triangles) is shown in the inset. (B) PCA of Melanesian genomes in the context of present-day worldwide genetic diversity. (C) Modern human variation projected onto the top two eigenvectors defined by PCA of the Altai Neandertal, Denisovan, and chimpanzee genome (14). Population means were plotted for each of the 11 Melanesian populations and each population of the global data set. (D) Estimates of Denisovan ancestry in Oceanic populations estimated from an f4 statistic (14). The 11 Melanesian populations are highlighted by the light blue box.

Having demonstrated that our Melanesian individuals have both Neandertal and Denisovan ancestry, we developed an approach to recover and classify archaic sequences. Briefly, we first identify putative introgressed sequences using the statistic S* (which does not use information from an archaic reference genome) (11, 18) and then refine this set by comparing significant S* haplotypes to the Neandertal and Denisovan genomes and testing to determine whether they match more than expected by chance. Variation in neutral divergence between archaic groups across loci and incomplete lineage sorting complicate classification of archaic haplotypes as Neandertal or Denisovan (fig. S4) (14). To address this issue, we developed a likelihood method that operates on the bivariate distribution of Neandertal and Denisovan match P values (14). This framework estimates the proportion of Neandertal, Denisovan, and null sequences in the set of S* significant haplotypes, identifies archaic haplotypes at a desired false discovery rate (FDR), and probabilistically categorizes them as Neandertal, Denisovan, or ambiguous (i.e., Neandertal or Denisovan status cannot be confidently distinguished) (14). In addition to our Melanesian samples, we also applied our method to whole-genome sequences from 1496 geographically diverse individuals studied as part of the 1000 Genomes Project (table S5) (14, 19).

We evaluated our approach through coalescent simulations (figs. S5 and S6) and by analyzing African populations (fig. S7). Archaic match P values calculated from null coalescent simulations without archaic admixture and significant S* sequences in African individuals show similar distributions (Fig. 2A), consistent with little to no Neandertal or Denisovan ancestry in most African populations. Notably, Luhya and Gambians do show evidence of having some Neanderthal ancestry (fig. S8 and table S6), most likely inherited indirectly through recent admixture with non-Africans (20). In Europeans, we see a strong skew of Neandertal, but not Denisovan, match P values toward zero (Fig. 2A). In contrast, Melanesians exhibit a marked skew of both Neandertal and Denisovan match P values toward zero (Fig. 2A).


Fig. 2 Identifying Neandertal and Denisovan sequences in modern human genomes.
(A) Bivariate archaic match P value distributions for simulations of nonintrogressed sequences, Esan in Nigeria, Europeans, and Melanesians. Null simulations and Esan show no skew in Neandertal or Denisovan match P values toward zero, Europeans show only a skew of Neandertal match P values toward zero, and Melanesians exhibit both Neandertal and Denisovan match P values skewed toward zero. (B) Amount of archaic introgressed sequences identified in each population. (Inset) Amount of Neandertal, Denisovan, and ambiguous (Neandertal or Denisovan) introgressed sequences for each Melanesian individual. (C) Schematic representation of introgressed haplotypes in an intronic portion of the GRM7 locus in Melanesian individuals illustrating mosaic patterns of archaic ancestry.

In aggregate, across all 1523 non-African individuals analyzed, we recovered 1340 Mb and 304 Mb of the Neandertal and Denisovan genomes, respectively, at a FDR = 5%. Melanesian individuals have on average 104 Mb of archaic sequences per individual (48.9, 42.9, and 12.2 Mb of Neandertal, Denisovan, and ambiguous sequence, respectively) (Fig. 2B). In contrast, we only call between 0.026 Mb (in Esan) to 0.5 Mb (in Luhya) of sequences per individual as archaic in Africans, highlighting that our method and error rates are well calibrated. The higher levels of archaic ancestry in Melanesians result in an average of 20 compound homozygous archaic loci per individual, with one Neandertal and one Denisovan haplotype (Fig. 2C and fig. S9). We estimate that 61% of the variability in the amount of Denisovan sequences between individuals is explained by variation in Papuan ancestry (P = 7.8 × 10−4) (table S7) (14). In other non-Africans, we identify on average 65.0, 55.2, and 51.2 Mb of archaic sequences in East Asians, South Asians, and Europeans, respectively (Fig. 2B). Virtually all of the archaic sequences in these populations are Neandertal in origin, although a small fraction (<1%) of introgressed sequences in East and South Asians are predicted to be Denisovan (fig. S8 and table S6) (14).

We developed a method to determine whether two populations have shared or unique admixture histories based on patterns of reciprocal sharing of Neandertal sequences among individuals (fig. S10) (14). We validated the expected behavior of our method in simulated data (Fig. 3A) and confirm previous observations of an additional admixture event unique to East Asians (11, 21, 22) (Fig. 3B). We find evidence for an additional pulse of Neandertal admixture in Europeans, East Asians, and South Asians compared with Melanesians (Fig. 3B), which is robust to different statistical thresholds used to call Neandertal and Denisovan sequences and determining significance (figs. S11 to S13) (14). Conversely, we find no evidence of differences in admixture histories between Europeans and South Asians (fig. S14). Collectively, these data suggest that Neandertal admixture occurred at least three distinct times in modern human history (Fig. 3C). Although most South Asian populations show shared histories of archaic admixture, we find significant evidence of differential Neandertal admixture between some European and East Asian populations (figs. S15 to S17).