When he was a PhD student at the University of Copenhagen, he was a nobody. At least, that's how the budding evolutionary geneticist seemed, unable to get his hands on one of the few coveted fossils that might still contain traces of ancient DNA.

But frustration turned to inspiration one autumn day in 2000, when he saw a dog leaving its morning droppings on the ground. Excrement contains DNA, he thought, and perhaps, even after the rain washes it away, some DNA might remain. And if it does, the genetic material of long-dead animals could persist in the environment. That would mean he could learn something about those creatures, even without access to priceless museum specimens.

Willerslev's idea was ridiculed by his professors at the time. "I've never heard anything as stupid as this," he recalls one of them remarking. But his hunch paid off handsomely. In a 2003 article in Science, he showed that plant and animal DNA could be recovered from a core of Siberian permafrost dating back 400,000 years.1
Even in the warmer temperatures of a New Zealand cave, Willerslev identified DNA from the extinct emu-like moa (Euryapteryx curtus) in 600-year-old sediments. It was the first time researchers had used sediment alone to identify long-dead, complex organisms.

Two decades later, the study of ancient DNA from sediments has matured into one of the most exciting tools for studying the past, according to researchers. Interest in soil DNA surged nearly 10 years ago when scientists discovered that human DNA could also be isolated from ancient sediments. Laboratories that once focused on extracting genetic material from valuable fossils are now turning their attention to the earth. Archaeologists are also re-examining soil collected decades ago, eager to learn more about the past using this modern technology.

The complex history archived in sediments is ripe for exploration, says Willerslev. And it's vast. In 2022, his team extracted fragments of DNA from two-million-year-old permafrost sediments in the far north of Greenland—the oldest genetic material of its kind to date.2 “It’s a huge new blue ocean” of possibilities, says Willerslev. “You have humans, you have animals, you have plants, you have the whole damn ecosystem.”

“It’s truly incredible how much molecular information is in the sediments,” says Matthias Meyer, a molecular biologist at the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany. “I think we’re really just scratching the surface of what’s possible,” he says.

According to Willerslev, if a site contains fossil remains, it could become irrelevant. “My expectation would be that we could almost just drop the bones in,” he says, “and they’d just fall into the earth.”

Sedimentary DNA has been especially influential in the study of ancient humans, revealing important clues about the earliest members of our own species, as well as about Neanderthals (Homo neanderthalensis) and the mysterious Denisovans, of whom very few bones have been found.

"Without sedimentary DNA," says Pere Gelabert, a population geneticist at the University of Vienna, uncovering these clues "would be impossible."

But even as groundbreaking findings on this new type of evidence grab headlines, some researchers are urging caution about whether enough care is being taken to ensure the results are reliable.
Hitting the Spot on
For 14 years, ancient sedimentary DNA—also known as silkDNA—remained within the realm of paleoecologists who reconstructed what life on Earth was like using lake cores and permafrost samples. But a pivotal moment for the field came in 2017, when scientists successfully identified DNA belonging to ancient humans in Ice Age soils.3

"When you have a story about humans, this is where you get people interested," says Diyendo Massilani, a paleogeneticist at Yale School of Medicine in New Haven, Connecticut, who was not involved in the 2017 study. As soon as someone found human DNA, he says, "then everyone was like, 'Let's do sediments for everything.'"

The problem is that ancient human DNA in soil is extremely rare compared to DNA from soil microorganisms and other fauna. To increase their chances of finding human DNA in samples, researchers at the Max Planck Institute for Evolutionary Anthropology—including paleogenomics' Nobel Prize-winning founder, Svante Pääbo—developed probes that selectively capture human sequences.3

The study found human DNA in locations where no human fossils have been discovered. This has highlighted the potential of sedaDNA to expand the fossil record. At Trou Al'Wesse cave in Belgium, for example, sedaDNA results confirmed a long-held suspicion—based on characteristic stone tools—that Neanderthals had occupied the site. At Denisova Cave in Siberia, researchers found DNA from both Neanderthals and a sister lineage, the Denisovans, who were named after the cave.

Some of the DNA appeared in layers without fossils. In a larger study of approximately 700 permafrost sediment samples from Denisova Cave, a sample from a deep layer indicated that Neanderthals had arrived at the site 170,000 years ago, 30,000 years earlier than suggested by fossil evidence. And although no bones have been found so far, sedaDNA places early modern humans in the cave some 45,000 years ago.

The DNA also places Neanderthals in a layer containing one type of stone tool, and Denisovans in a separate layer with another type, linking each to its likely maker. Such a connection is often difficult to establish otherwise. Meyer and others are optimistic that ancient DNA could one day identify toolmakers at other sites, and even the artists responsible for cave paintings.

Sedimentary DNA has also helped solve a mystery about a cave on the Tibetan Plateau, nearly 3,000 kilometers southeast of Denisova Cave. In 1980, a monk discovered an ancient jawbone at this site—called Baishiya Karst Cave. In 2019, Qiaomei Fu, a paleogeneticist at the Institute of Vertebrate Paleontology and Paleoanthropology in Beijing, and her colleagues reported the pioneering use of ancient proteins, revealing that the 160,000-year-old jawbone belonged to a Denisovan.5

However, the approach was novel, and there were doubts about the provenance of the jawbone, as it had been removed from the cave long ago. In 2020, Fu found Denisovan DNA in the cave sediments, confirming that these hominins had once occupied the site. It was the first conclusive evidence that archaic peoples lived outside of Siberia.6

A Step Toward Nuclear Energy

Both Meyer and Fu used 'molecular hooks' designed to retrieve human DNA from mitochondria, the tiny energy generators in cells. With thousands of copies per cell, mitochondrial DNA (mtDNA) is more abundant—and easier to find—than nuclear DNA. However, the size of the nuclear genome—three billion letters compared to only 16,000 for mtDNA—and its inheritance from both parents make it better at discerning how past human populations diverged and interbred throughout history.

The information that nuclear DNA could provide intrigued population geneticist Benjamin Vernot, who also works at the Max Planck Institute in Leipzig. He wondered if he could extract nuclear DNA from some of the samples that have yielded mtDNA.

To do this, Vernot designed a set of 1.6 million probes for sequences scattered throughout the human genome. These would bind to sequences from Neanderthals, Denisovans, and early modern humans. They would also detect DNA from ancient humans whose genetics are unknown. “There’s always the possibility that there’s Homo erectus DNA in your sample,” he says. “We wanted to be prepared, just in case.”

The probes successfully extracted nuclear sequences from the soil, but the biggest challenge was extracting meaning from the sparse data. “Our best sediment samples were still really, really bad,” says Vernot. Of the 1.6 million genomic probes, good samples contained sequences for only 10,000.

It took Vernot about eight months to develop computational methods that could handle such limited data. Using these methods, Vernot turned to sediments from the Gallery of Statues, a cave in northern Spain where excavations had uncovered a single Neanderthal foot bone and characteristic stone tools, but no genetic information.

From the mtDNA data, Vernot was able to distinguish two distinct Neanderthal populations, one of which completely replaced the other around 100,000 years ago. But from nuclear DNA, he was able to distinguish which samples were from a single man or woman, and which contained a mixture of DNA from several individuals.

Gelabert has also extracted nuclear sequences from cave sediments.8 But instead of capturing human sequences with probes, he took a brute-force approach, using shotgun sequencing to read all the DNA extracted from the soil.

NATURE