Ancient oral microbiomes support gradual Neolithic dietary shifts towards agriculture

The human microbiome is shaped by lifestyle and environment and can influence host health, yet how it evolved during key cultural transitions, such as the shift to agriculture in the Neolithic, remains unclear. Prior work on modern populations shows changes in oral microbial ecology between hunter-gatherers or traditional societies and urban communities, but ancient studies on the Neolithic transition have yielded controversial results, often confounded by small sample sizes and wide geographic dispersion. Italy, particularly the Apulia region where agriculture began around 6200 BC, offers a well-documented, regionally coherent case to investigate the relationship between subsistence change and oral microbiome evolution across clearly defined cultural phases (Early, Middle, Final Middle, and Late Neolithic). The aim of this study is to test how Neolithic dietary change affected the oral microbiome by integrating metagenomic analysis of dental calculus from Palaeolithic, Neolithic, and Copper Age individuals from Central-South Italy with microscopic evidence of plant food residues, while minimizing geographic and ecological confounders.

Modern comparative studies have reported shifts in oral microbiomes between traditional/hunter-gatherer groups and urban dwellers. Ancient oral microbiome studies across the Neolithic have produced mixed results: early reports suggested species abundance shifts linked to agriculture, but a later study did not replicate those patterns. Additional research in Southern Europe detected little variation associated with the Neolithic transition, instead highlighting geographic structuring of bacterial genome variation. Many prior studies relied on small, geographically dispersed samples with distinct ecologies and subsistence strategies, potentially introducing geographic and dietary bias and obscuring cultural effects. Archaeological evidence also indicates the Neolithic transition was not monolithic, with differences in subsistence strategies across coeval communities.

Study design and sampling: The study focuses on Central-South Italy (especially Apulia), where the Neolithic began around 6200 BC, and includes culturally defined phases: Early Neolithic (EN), Middle Neolithic (MN), final Middle Neolithic (FMN), Late Neolithic (LN), plus Upper Palaeolithic (PA) hunter-gatherers from Paglicci Cave and Copper Age (CA) individuals. A total of 79 dental calculus samples were collected; 76 were processed for metagenomic profiling. Nine PA samples were analyzed to characterize pre-agricultural microbiomes and provide a regional baseline; 67 Neolithic–CA samples cover all phases of the transition. Microscopic analysis of plant microremains was performed on 27 samples (9 PA; 18 Neolithic/CA). Ancient DNA extraction and sequencing: Samples were processed in a dedicated aDNA clean lab. Calculus was UV-decontaminated, mechanically cleaned, and digested (0.45 M EDTA, Proteinase K, Tween-20). DNA was extracted via silica spin-columns optimized for short molecules. Double-indexed Illumina libraries (no UDG treatment) were prepared and sequenced on an Illumina NovaSeq 6000 (2×51+8+8 cycles). Metagenomic processing and authentication: Reads were trimmed and merged (AdapterRemoval), duplicates removed (Prinseq), and taxonomically classified with Kraken2 against a custom RefSeq database (Dec 2020) with low-complexity masking; species abundances estimated by Bracken. Profiles were compared with modern and ancient microbiome references; SourceTracker was used to assess oral source, retaining samples with high oral signal (>75%). Species with overall relative abundance >0.02% were kept, yielding 49 species that were then authenticated via deamination profiles (PMDtools) and edit distance (-4% > 0.8). Taphonomic effects were assessed across periods. Community structure and statistics: Data were clr-transformed (Aitchison framework). Unsupervised clustering used gap statistic to set k=5, followed by network analysis (NetCoMi) on Aitchison distances and hierarchical clustering (average linkage) to define clusters C1–C5. PCoA, PERMANOVA, ANOSIM, and pairwise Adonis validated cluster structure and tested effects of read counts and batch. Random forest assessed variable importance (period, site, age, sex) for cluster distinctions; associations tested with chi-squared. Differential abundance and association: DESeq2 identified species differing among clusters (adjusted p<0.05; significance in more than one comparison). MaAsLin (FDR<0.05) tested associations of taxa with clusters and periods. Significant species were contextualized using known oral ecological complexes (red, orange, yellow, green, purple). Functional profiling: HUMAnN 2.0 profiled gene content; results regrouped to UniRef90_ko, filtered to authenticated species, and compared using LEfSe (p<0.05, LDA>2). Virulence factor (VF) genes were mapped to a custom database based on VFDB. Microscopic analysis: Dental calculus and extraction pellets were processed under clean conditions; microremains (starch grains, phytoliths, diatoms, fungal spores) were identified via light microscopy (100x–630x) using a large reference collection and published keys. Controls (environmental, lab) monitored contamination. Genome reconstruction: Metagenomic assemblies (metaSPAdes) were mapped (Bowtie2), binned (MetaBAT2), and quality-checked (CheckM). High-quality MAGs were taxonomically assigned (Kraken2); authenticity assessed via damage profiling. Olsenella sp. oral taxon 807 MAGs were compared to modern references using BRIG, PATRIC/RASTtk, and a phylogeny built with RAxML; protein family content and missing regions were assessed.

• Data overview and authentication: 76 calculus samples sequenced (mean ~28 million reads/sample). On average, 12.94% reads assigned to Bacteria/Archaea; Eukaryote ~0.01%. Most samples (71) showed strong oral signal in SourceTracker (>90%). After filtering (≥0.02% abundance), 49 species were retained and authenticated by deamination profiles; no significant temporal differences in damage profiles.
• Community structure: Network analysis identified five clusters (C1–C5) driven primarily by period and secondarily by geography. Associations: C1 with Palaeolithic (PA); C2 with Early/Middle Neolithic (EN/MN); FMN split between C3 and C5 (intra-site substructure at Palagiano); C4 with Late Neolithic and Copper Age (LN/CA) across sites. EN/MN microbiomes were taxonomically/ functionally intermediate between PA and LN/CA.
• Taxonomic shifts: Species in red, orange, and green oral complexes (e.g., Porphyromonas gingivalis, Tannerella forsythia, Treponema denticola; Campylobacter spp.; Capnocytophaga spp.; Eikenella corrodens) were rare in PA but increased starting in the Neolithic and peaked in LN/CA. P. gingivalis positively associated with C4 (LN/CA). Conversely, Streptococcus gordonii/mitis/oralis/sanguinis and many Actinomyces spp. were abundant in PA and declined over time. Additional Neolithic-enriched taxa included Olsenella sp. oral taxon 807, Parvimonas micra, Desulfomicrobium orale, and Ottowia HOT 894.
• Functional shifts: PA samples were enriched in carbohydrate metabolism pathways (e.g., pyruvate, propanoate, ascorbate, amino sugar, starch metabolism), consistent with higher Streptococcus abundance and amylase-binding traits. Neolithic samples showed enrichment in galactose metabolism, glyoxylate/dicarboxylate metabolism, and glycan pathways. Amino acid biosyntheses (lysine, glycine, cysteine) increased over time, highest in LN/CA. B-vitamin-related pathways changed across periods; LN/CA enriched for one-carbon pool by folate and lipoic acid metabolism. Virulence factors (motility, immune evasion, antiphagocytosis, endotoxins) increased progressively, highest in LN/CA; contributors included P. gingivalis, T. forsythia, P. intermedia, Capnocytophaga ochracea, Fusobacterium nucleatum.
• Microremains: PA calculus contained abundant starch (215 grains), mainly Poaceae (wild Triticeae and Avena cf. barbata) and Nuphar lutea rhizome. Neolithic/CA calculus yielded fewer starches (25 grains; Triticeae morphotype A), phytoliths (16 elongate dendritic epidermal from cereal inflorescence bracts), numerous Nitzschia diatoms (freshwater/marine, pollution-tolerant), and fungal spores (Glomus; yeasts in two Neolithic individuals), suggesting cereal consumption, possible fermented foods, and polluted water exposure.
• Genome reconstruction: Six ancient Olsenella sp. oral taxon 807 MAGs formed two ancient clusters; CA genomes were phylogenetically closer to the modern reference. Ancient genomes showed 123 depleted regions, including elements related to defence, mobile elements, biofilm/interaction (e.g., CRISPR-associated proteins, antitoxin HigA, integrases, TetR), suggesting recent evolution of interaction/virulence traits.
• Overall pattern: Two major shifts were detected: (i) a PA-to-Neolithic change linked to adoption of farming; (ii) a stronger LN-phase change, consistent with intensified agriculture and dietary carbohydrate shifts, accompanied by higher oral pathogen load and virulence factors.

The study directly addresses whether and how the transition to agriculture reshaped the human oral microbiome by minimizing geographic and ecological confounders and integrating dietary microremains. The PA microbiome differed distinctly from Neolithic ones, both taxonomically and functionally, reflecting long-standing hunter-gatherer subsistence with animal proteins/fats and diverse starchy plant intake. Early Neolithic communities (EN/MN) displayed an intermediate microbiome, consistent with a gradual transition and continuity in foodways. In contrast, LN/CA communities exhibited pronounced increases in taxa from the red/orange complexes and in virulence-related functions, alongside functional signatures (galactose metabolism, amino acid biosyntheses) suggestive of higher carbohydrate/plant-based intake and possible dairy/fermented products. Botanical and isotopic records from the region support a shift toward drought-tolerant cereals and reduced animal protein consumption by LN/CA, likely driven by climatic aridity episodes. Microremains (phytoliths, diatoms) corroborate cereal consumption and indicate potential consumption of polluted water, with implications for health. The higher prevalence of periodontal-associated taxa and virulence factors in LN/CA aligns with osteological evidence of increased oral disease in later periods. Intra-site variability at Palagiano may reflect social or dietary differentiation, consistent with varying funerary practices. Finally, genomic differences in Olsenella suggest recent evolution of competitive/interaction traits that, together with dietary pressures, could have reshaped community ecology.

By integrating authenticated ancient oral metagenomes with microremain evidence within a single, well-contextualized region, the study reveals two major shifts in oral microbiome composition across the Neolithic: an initial change with the adoption of farming and a stronger transformation in the Late Neolithic, likely tied to intensified agriculture and carbohydrate-rich diets. These shifts included increases in periodontal-associated taxa and virulence factors and corresponded to archaeobotanical and isotopic evidence of dietary change and environmental pressures. The findings support a gradual, regionally specific transition to agriculture rather than an abrupt, uniform shift. Future work could expand genomic reconstructions of additional oral taxa, include presently missing temporal brackets (e.g., Mesolithic), and further integrate environmental proxies to refine links between climate, subsistence strategies, and microbiome evolution.

The dataset lacks Mesolithic individuals from the study area, leaving a chronological gap between PA and EN. Although geographic bias was minimized, some clusters showed site-specific substructure (e.g., Palagiano), and intra-site social or dietary differences could not be fully resolved. A small number of samples exhibited contamination and were excluded; one environmental taxon (Burkholderia pseudomallei) could not be confidently ruled out as ancient contamination. Dental calculus provides a partial record of oral communities and diet; microremain recovery is variable. Taphonomic processes may still influence detection of low-abundance taxa and functions despite authentication procedures.