Ancient Dental Remains Offer New Evidence of Interbreeding Between Early Human Relatives

The Teeth That Survived 400,000 Years

In the caves at Zhoukoudian — the same site near Beijing where "Peking Man" was unearthed nearly a century ago — teeth have once again rewritten the story of human evolution. A study published in Nature in May 2026, led by paleogeneticist Qiaomei Fu of the Chinese Academy of Sciences' Institute of Vertebrate Paleontology and Paleoanthropology (IVPP), extracted and analyzed ancient proteins from six Homo erectus teeth recovered at three cave sites across China [1][2]. The findings point to interbreeding between H. erectus and Denisovans roughly 400,000 years ago — and suggest that genetic material from that encounter persists in living people today.

The study represents the first molecular evidence connecting H. erectus, one of the longest-surviving hominin species, to the Denisovan lineage through gene flow rather than direct ancestry [3]. If confirmed, it extends the known web of interbreeding among human relatives far deeper into the past than previously documented.

What the Teeth Contained

The six teeth came from five males and one female — sex determined through a novel method based on the male-specific enamel protein AMELY — at three sites: Zhoukoudian (near Beijing), Hexian (Anhui Province), and Sunjiadong (Henan Province) [1][2]. All specimens date to approximately 400,000 years ago, placing them firmly in the Middle Pleistocene.

Because DNA degrades over hundreds of thousands of years, especially in warm climates, Fu's team turned to paleoproteomics: the study of ancient proteins. Using a micro-destructive acid-etching technique that leaves only slight surface discoloration, they extracted 11 enamel proteins spanning hundreds of amino acid positions from each tooth [4]. The approach is significant because it preserves the morphology of specimens that are irreplaceable.

Across all six individuals and all three sites, the team identified two amino acid variants in ameloblastin (AMBN), a protein with a key structural role in enamel formation [1][2].

The first, AMBN-A253G, had never been observed in any hominin or primate. It was absent from H. erectus specimens at Dmanisi (Georgia) and Homo antecessor at Atapuerca (Spain), as well as from Neanderthals, Denisovans, and modern humans [2]. Its presence in all six Chinese specimens, from geographically separated sites, provides molecular evidence that these individuals belonged to a single evolutionary population — a conclusion supported with 100% posterior probability in phylogenetic analysis [4].

The second variant, AMBN-M273V, is the one that changes the map of human evolution.

The Denisovan Connection

AMBN-M273V had previously been identified only in Denisovans. Finding the same variant in 400,000-year-old H. erectus teeth — older than any known Denisovan specimen — suggests the mutation originated in the East Asian H. erectus population and entered the Denisovan lineage through interbreeding [1][2].

The chain does not end there. The variant is present in living humans at measurable frequencies: 21% in the Philippines, 1.17% in India, and 0.71% in Papua New Guinea [4]. These are populations known to carry elevated Denisovan ancestry, consistent with a model in which the variant passed from H. erectus to Denisovans and then to Homo sapiens through a second wave of interbreeding.

Source: Vernot & Akey 2014; Sankararaman et al. 2014; Jacobs et al. 2019

Data as of May 14, 2026CSV

This two-step transmission model — H. erectus → Denisovans → modern humans — would represent the deepest documented chain of gene flow in the hominin family tree. Previously, the oldest confirmed interbreeding events involved Neanderthals and Homo sapiens between 45,000 and 50,000 years ago, or Neanderthals and Denisovans somewhat earlier [5][6].

The Case Against Interbreeding

Not all researchers are persuaded. Diyendo Massilani, a geneticist at Yale University, has noted that convergent mutation — the independent evolution of the same protein variant in unrelated lineages — remains a theoretically viable alternative [4]. Under this scenario, H. erectus and Denisovans would have arrived at AMBN-M273V separately, without any sexual contact between the two groups.

Ryan McRae of the Smithsonian National Museum of Natural History raised another possibility: that East Asian H. erectus could be a direct ancestor of Denisovans rather than a separate lineage that interbred with them [3]. Under this model, AMBN-M273V would simply have been inherited vertically through a parent-descendant relationship, requiring no interbreeding at all.

Kirsty Penkman, a geochemist at the University of York who specializes in ancient proteins, has pushed back on the convergence hypothesis. Enamel proteins are structurally constrained — they cannot mutate extensively without losing function — which makes the independent appearance of the same variant in two separate lineages less probable than shared ancestry or gene flow [4].

The study authors have addressed these alternatives, but the debate is unlikely to be resolved by a single protein variant. As Enrico Cappellini, a biochemist at the University of Copenhagen, noted, the work establishes a methodological framework that can now be applied to other specimens [4].

A Broader Pattern: Hualongdong and Beyond

The Nature study arrives in the context of a growing body of evidence for complex interactions among hominin species in Asia. In August 2025, a study published in the Journal of Human Evolution by Wu Xiujie and colleagues examined 21 dental elements from the Hualongdong site in Anhui Province, dating to approximately 300,000 years ago [7][8]. Those teeth displayed an unusual combination of modern features — small third molars, smooth buccal surfaces — alongside archaic traits like robust molar and premolar roots typically associated with H. erectus [7].

The Hualongdong hominins do not fit neatly into any known species category. They lack the distinctive dental features of Neanderthals and Denisovans, yet their mosaic morphology has been interpreted by some researchers as evidence of gene flow between an early Homo sapiens-like group and a more archaic population [8]. The Asian continent increasingly appears to have been a zone of overlapping hominin populations during the Middle Pleistocene, with more genetic exchange than the traditional Out-of-Africa model anticipated.

What This Means for Models of Human Origins

The strict Out-of-Africa model — in which modern humans arose in Africa around 200,000 years ago and replaced all other hominin species with little or no interbreeding — has been under revision since the sequencing of the Neanderthal genome in 2010 revealed that all non-African humans carry approximately 1–4% Neanderthal DNA [5][9].

The assimilation model, which combines African origins with limited interbreeding during expansion, has gradually become the working consensus [10]. Under this framework, modern humans evolved in Africa, migrated outward, and absorbed small but meaningful genetic contributions from archaic populations they encountered.

The new dental protein evidence fits comfortably within the assimilation framework, but it extends the scope in two ways. First, it pushes documented gene flow back to 400,000 years ago, well before the emergence of anatomically modern humans. Second, it adds H. erectus — previously considered a dead-end lineage in terms of genetic contribution — as a participant in the interbreeding network [1][2].

The multiregional hypothesis, which posits that modern humans evolved simultaneously across Africa, Europe, and Asia with gene flow maintaining species cohesion, receives at most indirect support. The dental evidence confirms gene flow between Asian populations, but the direction of that flow — from an archaic population into a lineage that later contributed to modern humans — is more consistent with assimilation than with multiregionalism proper [10].

Source: OpenAlex

Data as of Jan 1, 2026CSV

The Rise of Paleoproteomics

A significant dimension of this study is methodological. Ancient DNA typically degrades beyond recovery in specimens older than a few hundred thousand years, particularly in non-permafrost environments [4]. Proteins, by contrast, are more durable. The acid-etching technique used by Fu's team extracted usable molecular information from 400,000-year-old teeth in the temperate climate of eastern China — environments where ancient DNA approaches have largely failed.

The field of paleoproteomics has grown rapidly. According to OpenAlex data, publications on paleoproteomics and ancient proteins grew from near zero before 2015 to 78 papers in both 2023 and 2025 [11].

Source: OpenAlex

Data as of Jan 1, 2026CSV

The study also introduced a cross-validation methodology using tandem mass spectrometry and multiple data analysis pipelines, along with the AMELY-based sex determination method [2]. These tools could be applied to the vast collections of hominin teeth held in museums and research institutions worldwide — many of which are too old or too degraded for DNA extraction.

Specimens, Access, and the Question of Replication

The six teeth analyzed in this study are held at IVPP, the Chinese Academy of Sciences' paleoanthropology institute in Beijing [2]. The Zhoukoudian specimens are among the most historically significant hominin fossils in the world; the original Peking Man fossils were famously lost during World War II, making surviving material from the site all the more sensitive.

Access to hominin dental material is tightly controlled across the field. Teeth are finite and partially consumed during analysis, even with micro-destructive techniques. Independent replication would require either re-analysis of the same specimens — which depends on IVPP granting access — or application of the same methods to H. erectus teeth from other sites. The Dmanisi and Atapuerca specimens referenced in the study as negative controls suggest that the methodology has already been tested across institutions, but systematic replication remains a challenge [4].

The study's raw proteomic data would need to be deposited in public repositories for independent computational analysis. The Nature publication likely requires data availability, but the extent to which other labs can independently verify the amino acid variant calls depends on access to both the raw spectrometry data and the computational pipelines described in the paper.

Medical and Population-Genetics Implications

If the AMBN-M273V variant did indeed travel from H. erectus through Denisovans to modern humans, it joins a growing inventory of archaic alleles with measurable effects on living populations.

The medical implications of archaic introgression are already well documented for Neanderthal DNA. A Neanderthal haplotype on chromosome 3 increases the risk of severe COVID-19 and is carried by approximately 50% of South Asians and 16% of Europeans [12]. Conversely, a Neanderthal haplotype on chromosome 12, overlapping antiviral OAS genes, reduces the odds of severe COVID-19 by 22% and is present at about 30% frequency in all non-African groups [13]. Neanderthal-derived variants in the TLR6-1-10 gene cluster have been linked to increased allergy susceptibility in European and Asian populations [14]. Other archaic variants have been associated with risk for type 2 diabetes, lupus, Crohn's disease, and biliary cirrhosis [14].

The AMBN-M273V variant itself is an enamel protein, and its direct medical significance — if any — is unclear. But its geographic distribution pattern (highest in the Philippines at 21%, lower in India and Papua New Guinea) parallels known patterns of Denisovan introgression and could serve as a tracer for other, functionally significant alleles that traveled the same H. erectus → Denisovan → Homo sapiens route [4]. Population geneticists tracking archaic ancestry now have a new molecular marker that extends the search deeper into the hominin past.

Source: Vernot & Akey 2014; Sankararaman et al. 2014; Jacobs et al. 2019

Data as of May 14, 2026CSV

What Remains Unknown

Several questions are unresolved. The study identifies protein-level variants, not full genomic sequences; the two AMBN mutations provide a narrow window into what may have been a broader pattern of gene flow. The dating of the specimens to "approximately 400,000 years ago" relies on existing stratigraphic frameworks for the three cave sites, and more precise radiometric dating could refine the timeline [2].

The relationship between H. erectus and Denisovans remains ambiguous. Whether the AMBN-M273V variant entered the Denisovan gene pool through admixture or vertical inheritance from an ancestral population is, as Massilani put it, still "hypothetical" [4]. Resolving this will require either more proteomic data from additional H. erectus and Denisovan specimens or, less likely, the recovery of ancient DNA from similarly aged material.

What is clear is that the boundaries between hominin species were more permeable than textbook diagrams suggest. The evidence from these six teeth — extracted with a technique gentle enough to leave the fossils largely intact — adds another line to a story that keeps getting more complicated: the human family tree is less a tree than a braided stream.