Ancient proteins tell tales of our ancestors

Archaeological sites offer clues to our past. Often the clues come in the form of skeletal remains, fossils, or pottery sherds. But they also come in the form of molecules. In particular, biological molecules like DNA recovered from ancient sites can reveal much about the people who lived there and the plants and animals they raised and ate.

Over the past couple of decades, sequencing ancient DNA has become a popular approach for recovering information about ancestral humans and other animals. After all, DNA serves as a blueprint for what makes us, well, us. So it is especially good for species identification.

But there are limits to the information that DNA can provide. For one thing, in archaeological terms, it doesn’t survive all that long. The most ancient DNA that has been analyzed so far was from a horse and was about 560,000–780,000 years old (Nature 2013, DOI: 10.1038/nature12323).

In contrast, a different set of biological molecules—ancient proteins—is built of tougher stuff. For instance, scientists have analyzed proteins recovered from the dental enamel of an approximately 1.77-million-year-old rhinoceros tooth unearthed at an archaeological site in Georgia (bioRxiv 2018, DOI: 10.1101/407692). So archaeologists and anthropologists are increasingly turning to ancient proteins to ask questions about history.

Reaching back

Earlier this month, a collaborative German and Chinese team announced that it had used proteins to identify an approximately 160,000-year-old jawbone found on the Tibetan Plateau as belonging to the Denisovan group of hominins (Nature 2019, DOI: 10.1038/s41586-019-1139-x). Denisovans are an extinct sister group of Neanderthals that had previously been identified only in the Denisova Cave in Siberia. Researchers weren’t able to recover any DNA from the jawbone, so they examined its proteins instead.

“Proteins take center stage because there is no DNA,” says Frido Welker, one of the lead authors of the Nature report the team published. Welker, who did the work while he was at the Max Planck Institute for Evolutionary Anthropology, is now at the University of Copenhagen.

The researchers had planned to sample the jawbone’s proteins from the beginning, Welker says, because from the mandible’s appearance and the age of the surrounding sediments, they “already suspected that it might be a little too old for DNA preservation.”

Thanks to the protein analysis, “for the first time, we know what a Denisovan mandible looked like and therefore also for the first time we are able to make comparisons with other Chinese and Asian fossils for which we don’t have molecular evidence,” Welker says. “But none of that could be done if we didn’t have the proteomic data in this paper. Otherwise it would simply be an archaic human mandible, quite old, with unclear taxonomic identity—like so many hominin fossils from that time period.”

To analyze ancient proteins, teams like Welker’s rely on mass spectrometry. For instance, Welker and his colleagues used liquid chromatography coupled to tandem mass spectrometry to extract and identify eight proteins from dentin, the dense tissue preserved beneath the enamel of the teeth on the jawbone. All the proteins they identified were from the collagen family.

Early mass spec work with ancient proteins focused on collagen, an abundant protein found in skin, connective tissue, and, most importantly for archaeological studies, bone. Bone helps protect collagen and other proteins, making it a rich source of proteins.

In the early 2000s, while at the University of York, Matthew Collins, a pioneer in mass spectrometry of archaeological and other ancient proteins, and student Michael Buckley developed a technique called “zooarchaeology by mass spectrometry,” or ZooMS, that used the peptides in collagen for species identification. In ZooMS, collagen is extracted from a sample, broken down into peptides with enzymes, and analyzed by matrix-assisted laser desorption ionization mass spectrometry. Because different species have different sequences of peptides making up their collagen, the peptides’ masses can serve as fingerprints for species.

But “the amount of information you can get from collagen is small,” Collins says. Typically, it yields only species information. Archaeologists want to know more. So they’re turning to more comprehensive proteomic methods based on LC-MS/MS to find as many proteins in archaeological samples as they can.

When mass spec–wielding archaeologists and anthropologists go hunting for proteins, they use many of the same methods that biomedical scientists use. First, they extract proteins from a sample. Most typically they then use protease enzymes, such as trypsin, to cut the proteins into peptides, which is called bottom-up proteomics. Or, less commonly, they can skip the enzyme digestion and analyze the proteins as is, which is called top-down proteomics. Then they separate the proteins or peptides using liquid chromatography and analyze them with mass spectrometry. The final step is to identify the proteins by using the mass spectral data to search protein databases such as UniProt.

Because the best-preserved proteins tend to be in mineralized samples, protected in substances such as bone or enamel, archaeologists have an extra step of demineralizing the sample before they can do the conventional proteomic workflow. “Being able to get proteins away from the mineral while retaining the proteins in a state that they’re not being destroyed by the methods to collect them is a hard challenge,” says Timothy P. Cleland, a paleontologist and mass spectrometrist at the Museum Conservation Institute, part of the Smithsonian Institution.

And then once the proteins are extracted, “protein degradation further exacerbates the problems” of working with these samples, says Christina Warinner, an anthropologist at the Max Planck Institute for the Science of Human History who uses mass spectrometry to study ancient proteins. Although proteins survive longer than DNA, they still decay naturally over time, which scientists need to take into account when studying them in archaeological samples.

Despite these challenges, “proteins are very exciting because they give you complementary information to genomics,” says Camilla Speller, a molecular archaeologist at the University of British Columbia. “On the one hand you get information that helps to authenticate the DNA. Sometimes you arrive at the same answer with proteins and DNA. And then you get a little extra information that you can’t get with DNA.”

Chowing down

Understanding ancient diets is a research area where that little extra information from proteins can be useful.

“Let’s say I want to know what someone ate in the past. Genetics is really good for that because it will give me a species inventory,” Warinner says.

But some questions are more nuanced.

For example, Warinner has identified the remains of cattle, sheep, and goats at archaeological sites. She wants to know more than just what types of animals these were. She wants to know whether the people who raised the animals were using them for milk or just for meat. “With proteomics,” Warinner says, “I can actually detect milk proteins, so I can say a lot more about the details of the diet.”

She finds the proteins that she analyzes in an unusual place: dental calculus, a hardened form of plaque on teeth. Dietary proteins can get entrapped in calculus.

In a study with modern, living individuals, Warinner’s team analyzed calculus from 10 people (Proc. R. Soc. B 2018, DOI: 10.1098/rspb.2018.0977). Nine had no dietary proteins in their calculus, but the 10th had lots of dietary proteins.

“We don’t know why some individuals incorporate more in their calculus than others,” Warinner says. “It seems to be a random process, so we can’t apply quantitative statistics to it.”

When dietary proteins are present, they give archaeologists a window to the past that they wouldn’t otherwise have. And dairy proteins seem to preserve quite well in calculus. “We see a bias toward milk proteins over other types of proteins,” Warinner says. She thinks that may happen because many dairy proteins bind to calcium.

Warinner is using such information to study how the practice of raising animals for their milk spread across the globe. Humans, like all mammals, are genetically programmed to drink milk as infants. But we’re also programmed as adults to stop producing the lactase enzymes that allow us to digest milk sugars. This leads to lactose intolerance—and tummy troubles when we decide to have that ice cream cone anyway. Some lucky human populations, though, have acquired genetic mutations that allow them to continue producing lactase as adults and thus continue to digest milk.

Warinner and her colleagues analyzed proteins in the dental calculus of nine individuals unearthed at burial sites in northern Mongolia estimated to be more than 3,000 years old (Proc. Natl. Acad. Sci. U.S.A. 2018, DOI: 10.1073/pnas.1813608115). They found protein evidence that those people consumed cow, sheep, and goat milk. But genomic analysis revealed no evidence of mutations that would have enabled them to digest milk during adulthood.

“We’re finding that there is a 4,000-year gap between when societies in the past first started dairying intensively and when we see the first evidence of this mutation in Europe and the Near East,” Warinner says. “So we have at least 4,000 years of people dairying with no native capacity to digest these foods.”

In places such as Mongolia, many people today still don’t have the mutation that allows them to digest milk. Yet the modern Mongolian diet is dairy based. Today, people in Mongolia raise seven species for milk—cows, sheep, goats, horses, camels, yaks, and reindeer. They are able to eat such a diet because they ferment the milk, which reduces the amount of lactose, making it more digestible.

Jessica Hendy, an archaeologist at the Max Planck Institute for the Science of Human History, is interested in what proteins can tell her about ancient food preparation techniques.

“There’s a huge suite of food science literature which is about exploring how proteins get modified through food preparation,” Hendy says. For example, the side chains of amino acids can be modified during cooking. Hendy is trying to apply findings from modern food science research to archaeological contexts.

Her team has been studying what molecular modifications occur in dairy products as they’re transformed to other products. And now, “we’re looking at archaeological data to see if we also see these same transformations,” she says.

Last year, Hendy reported the extraction of proteins from ceramic vessels from Çatalhöyük, a Neolithic early farming site in south-central Turkey (Nat. Commun. 2018, DOI: 10.1038/s41467-018-06335-6). She and her colleagues were able to find proteins from grains, legumes, dairy, and meat on pottery sherds from the site. Charred cereal grains and bones found near the ceramics were radiocarbon dated to around 8,000 years ago.

They found the proteins protected within mineral deposits on the inside walls of the ceramic vessels. The mineral deposits may be analogous to the limescale calcium carbonate deposits that form inside tea kettles, Hendy says.

Previously, archaeologists would look at fats on these kinds of ceramics and get information about only very broad classes of food that would’ve been prepared in the vessels, Hendy says. By studying proteins, “we can see that this vessel was used for the preparation of milk, peas, and grains, and we’re able to see different mixes of food,” she says. “It’s like uncovering ancient recipes.”

Since the initial study, Hendy and her team have found that such residues are fairly common on pottery in the eastern Mediterranean. “One of the things I would love to do in a much bigger study is try and understand the cuisine of these early farmers.”

But she still can’t be sure if those foods were actually cooked together or if the same pot was used at different times to cook them separately. Before she can figure that out, she needs to see if the mineral deposits have a layered structure that could help answer such a question.

“It could be that there’s cooking event after cooking event after cooking event,” Hendy says. “But I think this is going to require a lot of experimental work to try and figure out if this is the signature or not. I wouldn’t say it’s impossible, but think it will be challenging.”

Caroline Tokarski, a chemist at the University of Bordeaux, has also been studying protein residues on the inside of ancient vessels. Collaborating with archaeologists, she analyzed ceramic baby bottles from the first century CE found in the Swiss Alps. The archaeologists that Tokarski was working with hoped that the milk remnants inside the bottles would prove to be human, but they turned out to be cow’s milk.

But they’re still learning more about how the milk was prepared. Using top-down proteomics, Tokarski and her colleagues were able to identify molecules resulting from a Maillard reaction. The Maillard reaction occurs between sugars and proteins, typically during cooking. “This milk was possibly heated at the beginning to be able to add this modification,” Tokarski says. She was surprised to be able to obtain such information because she thought that the proteins and their modifications would have been too degraded to be observed. “Here is an example of how the chemistry of proteins provides us information on societal and historical uses and habits.”

Facing challenges

But getting all this useful information from ancient proteins is no walk in the park. One of the biggest challenges of analyzing the biomolecules is matching mass spec data with information in modern protein databases.

“The databases really are skewed toward proteomes of species that are of interest to modern humans,” Speller says. “When we’re trying to match our ancient proteins, we are actually trying to identify every possible protein that could be there.” And some of those proteins don’t exist today because the species they came from are extinct.

But even if the proteins are in the databases, there’s another challenge. Ancient proteins are by definition degraded or otherwise modified proteins. That means the peptides created through an enzyme’s digestion of an ancient protein might not be the same ones that would show up after the digestion of modern proteins. The peptides could be shorter and thus not match what’s in a database.

“When you search against the database, the search space becomes so big because you have to account for all these possible modifications,” Speller says. “It either takes you weeks to search your data or you end up in a circumstance in which you get a lot of false matches because you’re allowing the search to be not so stringent.”

Another challenge for archaeologists working with ancient proteins is that they have to make sure their proteins are authentically old rather than modern-day contaminants. At a minimum, the proteins have to be degraded in some way to meet this criterion.

“If it’s real, it ought to be degraded,” says Beatrice Demarchi, who studies the mechanisms of ancient protein degradation at the University of Turin. Even relatively well-preserved proteins undergo changes.

“The truth, unfortunately, is we know comparatively little about what exactly makes proteins degrade and ultimately what makes proteins survive,” Demarchi says. “Survival is the exception to the rule.”

And most of what people know about protein degradation comes from methods other than mass spectrometry, although there are observable effects in mass spectra.

“Paleoproteomics is growing so fast as a field, and it’s so linked to mass spec that people sometimes forget that there has been 60 years of research on paleobiogeochemistry of amino acids. Everything we’ve learned of degradation was based on these studies,” Demarchi says. “Mass spec is great, but it’s not the only thing out there.”

Many of those studies are high-temperature experiments meant to model what happens in the real world. But they can’t actually mimic the natural breakdown process, Demarchi says. Each reaction that occurs to break down an ancient protein has a different activation energy. “If you give a higher or lower temperature, the pathways of degradation will change,” Demarchi explains.

The major forms of protein degradation are peptide bond hydrolysis, amino acid decomposition, and racemization. Hydrolysis requires water and breaks the proteins into smaller pieces, so ancient proteins are never full length. Deamidation is one type of amino acid decomposition that involves the loss of NH₂ groups from glutamine or asparagine. Those changes result in deviations from expected masses of protein fragments that can be seen in the mass spectra.

Racemization, in contrast, doesn’t affect the masses of the amino acid residues and therefore the mass spectra. Instead, it involves the postmortem conversion of the stereochemistry of amino acids from the l form usually found in biological systems to the nonnatural d form.

Demarchi uses ostrich eggshell to study protein degradation because of its robustness and ability to protect its protein contents. “It’s such a simple, tight, closed system,” she says. “You can isolate the fraction of proteins that are not affected by contamination” and determine which pieces were untouched by temperature and time.

She and her colleagues were able to identify protein sequences in 3.8-million-year-old ostrich eggshells that had been discovered close to the equator (eLife 2016, DOI: 10.7554/eLife.17092). The proteins survived despite the hot environment.

They found that binding to mineral surfaces plays a big role in protein survival over time. The binding helps explain why most ancient proteins are associated with bones, teeth, shells, and other calcified substrates.

But the survival of proteins in bones is less straightforward than it may appear, Demarchi says. “Bone is a leaky, porous system with complex interactions going on with the surrounding environment. If collagen wasn’t such an abundant molecule—and bone such a uniquely informative biological remain—it wouldn’t be the substrate of choice for paleoproteomics.”

As much as they value the new kinds of information ancient proteins give them, molecular archaeologists also worry about misinterpretation.

“At the moment, probably one in four ancient protein papers has data that is probably—to put it kindly—overinterpreted,” says Collins, the developer of ZooMS who is now at the University of Copenhagen and the University of Cambridge.

UBC’s Speller suspects that part of the problem may be naive adoption of techniques from the physical and life sciences. “As we dabble in these techniques that were not designed for us,” she says, “we have the potential to make dramatic misinterpretations about the archaeological record.”

“Everyone is doing their best,” Demarchi says. “But probably in 10 years—maybe even earlier than that—we’ll find out that some of the data sets we’re producing at the best of our ability can be reinterpreted in the light of the new genomes, new sequences, and new data that we’ll have.”

Despite the need for caution, molecular archaeologists are optimistic. Proteomics is “such a versatile tool that we can gain a lot of information from that we never really thought possible before,” the Smithsonian’s Cleland says. “We can better understand proteins and how they’re preserved” and “get molecular information from extinct species that we may never get DNA from.”