New computational technique traces ghost ancestry in modern human genomes

UC Berkeley researchers have identified regions of DNA that humans inherited from two unknown ancestors as a result of long-ago interbreeding among modern humans and at least four other archaic human relatives.

While previous research found hints that modern humans had interbred with ancient hominins in addition to now-extinct Neanderthals and Denisovans, the current study pinpoints areas of the genome inherited from these unknown ancestors and establishes a timeline. This was possible using a new technique the researchers developed that leverages hundreds of genomes from modern-day humans to find ancient genealogical relationships.

One of the unknown ancestors, which the researchers refer to as a ghost ancestor, interbred with modern humans in Africa prior to 50,000 years ago, before the most recent migration of Homo sapiens out of Africa into Europe and Asia. These genes comprise around 1% of the genomes of modern humans, comparable to the amount of Neanderthal DNA in the human genome. That unknown hominin lineage actually split off from the modern human lineage around the same time that Neanderthals and Denisovans diverged around 800,000 years ago, though they interbred with modern humans earlier.

Previous publications suggested that there might be ghost ancestry - ancestry from unknown archaic lineages in modern humans - but they hadn't concluded whether this unknown ancestry is present only in Africans or not, and when this introgression event happened. We were actually able to find and map genomic locations in modern humans that are from this ghost lineage and show that this ghost ancestry is in all modern humans, not only in Africans."

Yulin Zhang, one of two first authors of the study, Berkeley graduate student

The other mysterious ancestor - what the researchers call a super-archaic ancestor - was a 1.8 million-year-old hominin lineage that interbred with Denisovans in Eurasia; Denisovans later shared some of the super-archaic DNA with modern humans through interbreeding with Homo sapiens.

"The super-archaic finding is particularly exciting because it reveals genetic contributions from a human lineage that lived over a million years ago, despite the absence of any sequenced DNA from that population," said Arjun Biddanda, a postdoctoral researcher at Johns Hopkins University and a co-first author of the study

The findings highlight the fact that early modern humans not only lived side-by-side with numerous related hominin groups in Africa and Eurasia over the past million years, but that they were close enough genetically to interbreed with many of them.

"With ancient DNA from Neanderthals and Denisovans and with these new genealogical methods, we are learning that mixture among human populations has been very pervasive across time, and that this is also likely to be true at ancient time scales," said Priya Moorjani, a Berkeley associate professor of molecular and cell biology. "We often think of human evolution as a branching tree, but new genomic data and analytical methods reveal a much more interconnected history - more like a complex web of populations connected by repeated episodes of migration and mixing."

Though it's unclear who these ghost and super-archaic ancestors were, the inferred divergence time overlaps with the existence of Middle Pleistocene Homo groups in Africa 800,000 years ago and Homo erectus in Eurasia 1.8 million years ago, respectively.

Zhang, Biddanda, Moorjani and their colleagues will publish their findings July 30 in the journal Science.

Tracing archaic DNA

When modern humans first moved out of Africa some 50,000 years ago, they interbred with two much older hominin lineages in Eurasia, the Neanderthals and Denisovans. While those older lineages went extinct, they left some of their DNA in the modern human genome - a discovery made possible by the extraction and sequencing of ancient DNA from Neanderthal and Denisovan fossils.

But researchers also saw hints in the human genome of much earlier interbreeding. Because we have no DNA from other extinct hominins, however, it was challenging to identify their contributions.

Moorjani's team developed a technique called TRACE (TRacking Archaic Contributions via ARG Estimation) that locates these regions by analyzing complete genomes from present-day humans only. Using genome data from modern individuals from around the world, the team reconstructed genealogical relationships - what's called an ancestral recombination graph (ARG) –– across the genome, creating a detailed map of how DNA segments are related through shared ancestry over time.

"Genealogies preserve a record of our evolutionary past," said Moorjani. "TRACE reconstructs those histories across the genome. By identifying regions whose ancestry extends unusually far back in time, we can uncover genetic contributions from extinct human populations, even in the absence of ancient DNA."

Some of the regions with the oldest ancestry matched DNA from sequenced Neanderthal genome, which makes up about 1% of the human genome. Looking exclusively at genomes from people in Asia and Oceania, the only populations having substantial Denisovan DNA, they also correctly identified known regions of Denisovan introgression.

But many of the ancient regions contained neither Neanderthal nor Denisovan DNA. The researchers determined that these stretches of DNA came from two distinct lineages, with different introgression times. The ghost lineage was found in all humans, non-Africans and Africans, indicating that the gene flow occurred before modern humans made their final excursion out of Africa and spread rapidly throughout the world.

"We discovered that about 2% of the modern human genome is from archaic hominins," Zhang said. "In the case of the ghost lineage, modern-day Africans and non-African populations both inherited similar amounts of ghost ancestry. Each individual has about 0.5 to 1% of their genome inherited from this ghost lineage."

The researchers recovered the other lineage by analyzing only genomes from populations in Oceania - peoples living in the island nations of the Pacific Ocean - which have been shown to have higher proportions of Denisovan DNA, sometimes up to 4%. The super-archaic ancestry popped up in regions of Denisovan DNA, suggesting that this DNA came via gene flow through Denisovans into modern humans. Since Denisovans harbor between 3% and 5% super-archaic DNA, only a small fraction of that is inherited by modern humans, Moorjani said.

"TRACE allowed us to contextualize how the ancestry segments from these previously uncharacterized hominins are distributed throughout the human genome," Biddanda said. "We found that these contributions are widespread throughout the genome, and ghost ancestry is detected even in regions previously thought to be intolerant of Neanderthal and Denisovan ancestry."

Many archaic segments are enriched in regions associated with immunity and metabolic function, according to Moorjani.

"This pattern is not entirely surprising," she said. "Adaptation to new pathogens and food sources has been one of the strongest selective pressures in human evolution. Interbreeding with other human groups introduced new genetic variation, providing additional raw material for natural selection. Beneficial variants could then be retained and spread over many generations."

Moorjani hopes to detect faint signals of additional lineages in human DNA as the world's genome databases become more diverse, sampling a broader variety of humanity. The discovery of more Denisovan genomes - currently only one has been published - would also help. Protein sequences recently sequenced from Homo erectus fossils could even help identify who the super-archaic ancestor was.

"I think these new computational methods that allow us to reconstruct genealogical relationships are really the next frontier in this field because they are allowing us to uncover hidden episodes from our past without requiring ancient DNA," she added.

Moorjani noted that TRACE should also work with other species, "allowing us to also uncover really different patterns across the tree of life."

Other co-authors of the paper are Sarah Johnson of Berkeley's Center for Computational Biology and Colm O'Dushlaine of 54Gene, Inc., in Washington, D.C. The work was funded by the National Science Foundation.